This series adds support for running protected VMs using KVM under the
Arm Confidential Compute Architecture (CCA). The purpose of this
series is to gather feedback on the proposed changes to the architecture
code for CCA.
The main change from the previous RFC is that it updates the code to use
a guest_memfd descriptor to back the private memory of the guest. This
avoids any issues where a malicious VMM could potentially cause a fatal
Granule Protection Fault elsewhere in the kernel.
The ABI to the RMM (the RMI) is based on the final RMM v1.0 (EAC 5)
specification[1].
This series is based on v6.9-rc1. It is also available as a git
repository:
https://gitlab.arm.com/linux-arm/linux-cca cca-host/v2
Work in progress changes for kvmtool are available from the git
repository below, these changes are based on Fuad Tabba's repository for
pKVM to provide some alignment with the ongoing pKVM work:
https://gitlab.arm.com/linux-arm/kvmtool-cca cca/v2
Introduction
============
A more general introduction to Arm CCA is available on the Arm
website[2], and links to the other components involved are available in
the overall cover letter.
Arm Confidential Compute Architecture adds two new 'worlds' to the
architecture: Root and Realm. A new software component known as the RMM
(Realm Management Monitor) runs in Realm EL2 and is trusted by both the
Normal World and VMs running within Realms. This enables mutual
distrust between the Realm VMs and the Normal World.
Virtual machines running within a Realm can decide on a (4k)
page-by-page granularity whether to share a page with the (Normal World)
host or to keep it private (protected). This protection is provided by
the hardware and attempts to access a page which isn't shared by the
Normal World will trigger a Granule Protection Fault. The series starts
by adding handling for these; faults within user space can be handled by
killing the process, faults within kernel space are considered fatal.
The Normal World host can communicate with the RMM via an SMC interface
known as RMI (Realm Management Interface), and Realm VMs can communicate
with the RMM via another SMC interface known as RSI (Realm Services
Interface). This series adds wrappers for the full set of RMI commands
and uses them to manage the realm guests.
The Normal World can use RMI commands to delegate pages to the Realm
world and to create, manage and run Realm VMs. Once delegated the pages
are inaccessible to the Normal World (unless explicitly shared by the
guest). However the Normal World may destroy the Realm VM at any time to
be able to reclaim (undelegate) the pages.
Realm VMs are identified by the KVM_CREATE_VM command, where the 'type'
argument has a new field to describe whether the guest is 'normal' or a
'realm'.
Entry/exit of a Realm VM attempts to reuse the KVM infrastructure, but
ultimately the final mechanism is different. So this series has a bunch
of commits handling the differences. As much as possible is placed in a
two new files: rme.c and rme-exit.c.
KVM also handles some of the PSCI requests for a realm and helps the RMM
complete the PSCI service requests.
Interrupts are managed by KVM, and are injected into the Realm with the
help of the RMM.
The RMM specification provides a new mechanism for a guest to
communicate with host which goes by the name "Host Call". This is simply
hooked up to the existing support for HVC calls from a normal
guest.
[1] https://developer.arm.com/documentation/den0137/1-0eac5/
[2] https://www.arm.com/architecture/security-features/arm-confidential-compute-architecture
Jean-Philippe Brucker (7):
arm64: RME: Propagate number of breakpoints and watchpoints to
userspace
arm64: RME: Set breakpoint parameters through SET_ONE_REG
arm64: RME: Initialize PMCR.N with number counter supported by RMM
arm64: RME: Propagate max SVE vector length from RMM
arm64: RME: Configure max SVE vector length for a Realm
arm64: RME: Provide register list for unfinalized RME RECs
arm64: RME: Provide accurate register list
Joey Gouly (2):
arm64: rme: allow userspace to inject aborts
arm64: rme: support RSI_HOST_CALL
Sean Christopherson (1):
KVM: Prepare for handling only shared mappings in mmu_notifier events
Steven Price (29):
arm64: RME: Handle Granule Protection Faults (GPFs)
arm64: RME: Add SMC definitions for calling the RMM
arm64: RME: Add wrappers for RMI calls
arm64: RME: Check for RME support at KVM init
arm64: RME: Define the user ABI
arm64: RME: ioctls to create and configure realms
arm64: kvm: Allow passing machine type in KVM creation
arm64: RME: Keep a spare page delegated to the RMM
arm64: RME: RTT handling
arm64: RME: Allocate/free RECs to match vCPUs
arm64: RME: Support for the VGIC in realms
KVM: arm64: Support timers in realm RECs
arm64: RME: Allow VMM to set RIPAS
arm64: RME: Handle realm enter/exit
KVM: arm64: Handle realm MMIO emulation
arm64: RME: Allow populating initial contents
arm64: RME: Runtime faulting of memory
KVM: arm64: Handle realm VCPU load
KVM: arm64: Validate register access for a Realm VM
KVM: arm64: Handle Realm PSCI requests
KVM: arm64: WARN on injected undef exceptions
arm64: Don't expose stolen time for realm guests
arm64: RME: Always use 4k pages for realms
arm64: rme: Prevent Device mappings for Realms
arm_pmu: Provide a mechanism for disabling the physical IRQ
arm64: rme: Enable PMU support with a realm guest
kvm: rme: Hide KVM_CAP_READONLY_MEM for realm guests
arm64: kvm: Expose support for private memory
KVM: arm64: Allow activating realms
Suzuki K Poulose (4):
kvm: arm64: pgtable: Track the number of pages in the entry level
kvm: arm64: Include kvm_emulate.h in kvm/arm_psci.h
kvm: arm64: Expose debug HW register numbers for Realm
arm64: rme: Allow checking SVE on VM instance
Documentation/virt/kvm/api.rst | 3 +
arch/arm64/include/asm/kvm_emulate.h | 35 +
arch/arm64/include/asm/kvm_host.h | 13 +-
arch/arm64/include/asm/kvm_pgtable.h | 2 +
arch/arm64/include/asm/kvm_rme.h | 154 +++
arch/arm64/include/asm/rmi_cmds.h | 509 +++++++++
arch/arm64/include/asm/rmi_smc.h | 250 ++++
arch/arm64/include/asm/virt.h | 1 +
arch/arm64/include/uapi/asm/kvm.h | 49 +
arch/arm64/kvm/Kconfig | 1 +
arch/arm64/kvm/Makefile | 3 +-
arch/arm64/kvm/arch_timer.c | 45 +-
arch/arm64/kvm/arm.c | 178 ++-
arch/arm64/kvm/guest.c | 99 +-
arch/arm64/kvm/hyp/pgtable.c | 5 +-
arch/arm64/kvm/hypercalls.c | 4 +-
arch/arm64/kvm/inject_fault.c | 2 +
arch/arm64/kvm/mmio.c | 10 +-
arch/arm64/kvm/mmu.c | 172 ++-
arch/arm64/kvm/pmu-emul.c | 7 +-
arch/arm64/kvm/psci.c | 29 +
arch/arm64/kvm/reset.c | 23 +-
arch/arm64/kvm/rme-exit.c | 211 ++++
arch/arm64/kvm/rme.c | 1590 ++++++++++++++++++++++++++
arch/arm64/kvm/sys_regs.c | 83 +-
arch/arm64/kvm/vgic/vgic-v3.c | 9 +-
arch/arm64/kvm/vgic/vgic.c | 37 +-
arch/arm64/mm/fault.c | 29 +-
drivers/perf/arm_pmu.c | 15 +
include/kvm/arm_arch_timer.h | 2 +
include/kvm/arm_psci.h | 2 +
include/linux/kvm_host.h | 2 +
include/linux/perf/arm_pmu.h | 1 +
include/uapi/linux/kvm.h | 30 +-
virt/kvm/kvm_main.c | 7 +
35 files changed, 3514 insertions(+), 98 deletions(-)
create mode 100644 arch/arm64/include/asm/kvm_rme.h
create mode 100644 arch/arm64/include/asm/rmi_cmds.h
create mode 100644 arch/arm64/include/asm/rmi_smc.h
create mode 100644 arch/arm64/kvm/rme-exit.c
create mode 100644 arch/arm64/kvm/rme.c
--
2.34.1
From: Sean Christopherson <[email protected]>
Add flags to "struct kvm_gfn_range" to let notifier events target only
shared and only private mappings, and write up the existing mmu_notifier
events to be shared-only (private memory is never associated with a
userspace virtual address, i.e. can't be reached via mmu_notifiers).
Add two flags so that KVM can handle the three possibilities (shared,
private, and shared+private) without needing something like a tri-state
enum.
Link: https://lore.kernel.org/all/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
include/linux/kvm_host.h | 2 ++
virt/kvm/kvm_main.c | 7 +++++++
2 files changed, 9 insertions(+)
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 48f31dcd318a..c7581360fd88 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -268,6 +268,8 @@ struct kvm_gfn_range {
gfn_t start;
gfn_t end;
union kvm_mmu_notifier_arg arg;
+ bool only_private;
+ bool only_shared;
bool may_block;
};
bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index fb49c2a60200..3486ceef6f4e 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -633,6 +633,13 @@ static __always_inline kvm_mn_ret_t __kvm_handle_hva_range(struct kvm *kvm,
* the second or later invocation of the handler).
*/
gfn_range.arg = range->arg;
+
+ /*
+ * HVA-based notifications aren't relevant to private
+ * mappings as they don't have a userspace mapping.
+ */
+ gfn_range.only_private = false;
+ gfn_range.only_shared = true;
gfn_range.may_block = range->may_block;
/*
--
2.34.1
If the host attempts to access granules that have been delegated for use
in a realm these accesses will be caught and will trigger a Granule
Protection Fault (GPF).
A fault during a page walk signals a bug in the kernel and is handled by
oopsing the kernel. A non-page walk fault could be caused by user space
having access to a page which has been delegated to the kernel and will
trigger a SIGBUS to allow debugging why user space is trying to access a
delegated page.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/mm/fault.c | 29 ++++++++++++++++++++++++-----
1 file changed, 24 insertions(+), 5 deletions(-)
diff --git a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c
index 8251e2fea9c7..91da0f446dd9 100644
--- a/arch/arm64/mm/fault.c
+++ b/arch/arm64/mm/fault.c
@@ -765,6 +765,25 @@ static int do_tag_check_fault(unsigned long far, unsigned long esr,
return 0;
}
+static int do_gpf_ptw(unsigned long far, unsigned long esr, struct pt_regs *regs)
+{
+ const struct fault_info *inf = esr_to_fault_info(esr);
+
+ die_kernel_fault(inf->name, far, esr, regs);
+ return 0;
+}
+
+static int do_gpf(unsigned long far, unsigned long esr, struct pt_regs *regs)
+{
+ const struct fault_info *inf = esr_to_fault_info(esr);
+
+ if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
+ return 0;
+
+ arm64_notify_die(inf->name, regs, inf->sig, inf->code, far, esr);
+ return 0;
+}
+
static const struct fault_info fault_info[] = {
{ do_bad, SIGKILL, SI_KERNEL, "ttbr address size fault" },
{ do_bad, SIGKILL, SI_KERNEL, "level 1 address size fault" },
@@ -802,11 +821,11 @@ static const struct fault_info fault_info[] = {
{ do_alignment_fault, SIGBUS, BUS_ADRALN, "alignment fault" },
{ do_bad, SIGKILL, SI_KERNEL, "unknown 34" },
{ do_bad, SIGKILL, SI_KERNEL, "unknown 35" },
- { do_bad, SIGKILL, SI_KERNEL, "unknown 36" },
- { do_bad, SIGKILL, SI_KERNEL, "unknown 37" },
- { do_bad, SIGKILL, SI_KERNEL, "unknown 38" },
- { do_bad, SIGKILL, SI_KERNEL, "unknown 39" },
- { do_bad, SIGKILL, SI_KERNEL, "unknown 40" },
+ { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection Fault at level 0" },
+ { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection Fault at level 1" },
+ { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection Fault at level 2" },
+ { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection Fault at level 3" },
+ { do_gpf, SIGBUS, SI_KERNEL, "Granule Protection Fault not on table walk" },
{ do_bad, SIGKILL, SI_KERNEL, "level -1 address size fault" },
{ do_bad, SIGKILL, SI_KERNEL, "unknown 42" },
{ do_translation_fault, SIGSEGV, SEGV_MAPERR, "level -1 translation fault" },
--
2.34.1
The RMM (Realm Management Monitor) provides functionality that can be
accessed by SMC calls from the host.
The SMC definitions are based on DEN0137[1] version 1.0-eac5
[1] https://developer.arm.com/documentation/den0137/1-0eac5/
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/rmi_smc.h | 250 +++++++++++++++++++++++++++++++
1 file changed, 250 insertions(+)
create mode 100644 arch/arm64/include/asm/rmi_smc.h
diff --git a/arch/arm64/include/asm/rmi_smc.h b/arch/arm64/include/asm/rmi_smc.h
new file mode 100644
index 000000000000..c205efdb18d8
--- /dev/null
+++ b/arch/arm64/include/asm/rmi_smc.h
@@ -0,0 +1,250 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2023 ARM Ltd.
+ *
+ * The values and structures in this file are from the Realm Management Monitor
+ * specification (DEN0137) version A-bet0:
+ * https://developer.arm.com/documentation/den0137/1-0bet0/
+ */
+
+#ifndef __ASM_RME_SMC_H
+#define __ASM_RME_SMC_H
+
+#include <linux/arm-smccc.h>
+
+#define SMC_RxI_CALL(func) \
+ ARM_SMCCC_CALL_VAL(ARM_SMCCC_FAST_CALL, \
+ ARM_SMCCC_SMC_64, \
+ ARM_SMCCC_OWNER_STANDARD, \
+ (func))
+
+#define SMC_RMI_DATA_CREATE SMC_RxI_CALL(0x0153)
+#define SMC_RMI_DATA_CREATE_UNKNOWN SMC_RxI_CALL(0x0154)
+#define SMC_RMI_DATA_DESTROY SMC_RxI_CALL(0x0155)
+#define SMC_RMI_FEATURES SMC_RxI_CALL(0x0165)
+#define SMC_RMI_GRANULE_DELEGATE SMC_RxI_CALL(0x0151)
+#define SMC_RMI_GRANULE_UNDELEGATE SMC_RxI_CALL(0x0152)
+#define SMC_RMI_PSCI_COMPLETE SMC_RxI_CALL(0x0164)
+#define SMC_RMI_REALM_ACTIVATE SMC_RxI_CALL(0x0157)
+#define SMC_RMI_REALM_CREATE SMC_RxI_CALL(0x0158)
+#define SMC_RMI_REALM_DESTROY SMC_RxI_CALL(0x0159)
+#define SMC_RMI_REC_AUX_COUNT SMC_RxI_CALL(0x0167)
+#define SMC_RMI_REC_CREATE SMC_RxI_CALL(0x015a)
+#define SMC_RMI_REC_DESTROY SMC_RxI_CALL(0x015b)
+#define SMC_RMI_REC_ENTER SMC_RxI_CALL(0x015c)
+#define SMC_RMI_RTT_CREATE SMC_RxI_CALL(0x015d)
+#define SMC_RMI_RTT_DESTROY SMC_RxI_CALL(0x015e)
+#define SMC_RMI_RTT_FOLD SMC_RxI_CALL(0x0166)
+#define SMC_RMI_RTT_INIT_RIPAS SMC_RxI_CALL(0x0168)
+#define SMC_RMI_RTT_MAP_UNPROTECTED SMC_RxI_CALL(0x015f)
+#define SMC_RMI_RTT_READ_ENTRY SMC_RxI_CALL(0x0161)
+#define SMC_RMI_RTT_SET_RIPAS SMC_RxI_CALL(0x0169)
+#define SMC_RMI_RTT_UNMAP_UNPROTECTED SMC_RxI_CALL(0x0162)
+#define SMC_RMI_VERSION SMC_RxI_CALL(0x0150)
+
+#define RMI_ABI_MAJOR_VERSION 1
+#define RMI_ABI_MINOR_VERSION 0
+
+#define RMI_UNASSIGNED 0
+#define RMI_ASSIGNED 1
+#define RMI_TABLE 2
+
+#define RMI_ABI_VERSION_GET_MAJOR(version) ((version) >> 16)
+#define RMI_ABI_VERSION_GET_MINOR(version) ((version) & 0xFFFF)
+#define RMI_ABI_VERSION(major, minor) (((major) << 16) | (minor))
+
+#define RMI_RETURN_STATUS(ret) ((ret) & 0xFF)
+#define RMI_RETURN_INDEX(ret) (((ret) >> 8) & 0xFF)
+
+#define RMI_SUCCESS 0
+#define RMI_ERROR_INPUT 1
+#define RMI_ERROR_REALM 2
+#define RMI_ERROR_REC 3
+#define RMI_ERROR_RTT 4
+
+#define RMI_EMPTY 0
+#define RMI_RAM 1
+#define RMI_DESTROYED 2
+
+#define RMI_NO_MEASURE_CONTENT 0
+#define RMI_MEASURE_CONTENT 1
+
+#define RMI_FEATURE_REGISTER_0_S2SZ GENMASK(7, 0)
+#define RMI_FEATURE_REGISTER_0_LPA2 BIT(8)
+#define RMI_FEATURE_REGISTER_0_SVE_EN BIT(9)
+#define RMI_FEATURE_REGISTER_0_SVE_VL GENMASK(13, 10)
+#define RMI_FEATURE_REGISTER_0_NUM_BPS GENMASK(17, 14)
+#define RMI_FEATURE_REGISTER_0_NUM_WPS GENMASK(21, 18)
+#define RMI_FEATURE_REGISTER_0_PMU_EN BIT(22)
+#define RMI_FEATURE_REGISTER_0_PMU_NUM_CTRS GENMASK(27, 23)
+#define RMI_FEATURE_REGISTER_0_HASH_SHA_256 BIT(28)
+#define RMI_FEATURE_REGISTER_0_HASH_SHA_512 BIT(29)
+
+#define RMI_REALM_PARAM_FLAG_LPA2 BIT(0)
+#define RMI_REALM_PARAM_FLAG_SVE BIT(1)
+#define RMI_REALM_PARAM_FLAG_PMU BIT(2)
+
+/*
+ * Note many of these fields are smaller than u64 but all fields have u64
+ * alignment, so use u64 to ensure correct alignment.
+ */
+struct realm_params {
+ union { /* 0x0 */
+ struct {
+ u64 flags;
+ u64 s2sz;
+ u64 sve_vl;
+ u64 num_bps;
+ u64 num_wps;
+ u64 pmu_num_ctrs;
+ u64 hash_algo;
+ };
+ u8 padding_1[0x400];
+ };
+ union { /* 0x400 */
+ u8 rpv[64];
+ u8 padding_2[0x400];
+ };
+ union { /* 0x800 */
+ struct {
+ u64 vmid;
+ u64 rtt_base;
+ s64 rtt_level_start;
+ u64 rtt_num_start;
+ };
+ u8 padding_3[0x800];
+ };
+};
+
+/*
+ * The number of GPRs (starting from X0) that are
+ * configured by the host when a REC is created.
+ */
+#define REC_CREATE_NR_GPRS 8
+
+#define REC_PARAMS_FLAG_RUNNABLE BIT_ULL(0)
+
+#define REC_PARAMS_AUX_GRANULES 16
+
+struct rec_params {
+ union { /* 0x0 */
+ u64 flags;
+ u8 padding1[0x100];
+ };
+ union { /* 0x100 */
+ u64 mpidr;
+ u8 padding2[0x100];
+ };
+ union { /* 0x200 */
+ u64 pc;
+ u8 padding3[0x100];
+ };
+ union { /* 0x300 */
+ u64 gprs[REC_CREATE_NR_GPRS];
+ u8 padding4[0x500];
+ };
+ union { /* 0x800 */
+ struct {
+ u64 num_rec_aux;
+ u64 aux[REC_PARAMS_AUX_GRANULES];
+ };
+ u8 padding5[0x800];
+ };
+};
+
+#define RMI_EMULATED_MMIO BIT(0)
+#define RMI_INJECT_SEA BIT(1)
+#define RMI_TRAP_WFI BIT(2)
+#define RMI_TRAP_WFE BIT(3)
+
+#define REC_RUN_GPRS 31
+#define REC_GIC_NUM_LRS 16
+
+struct rec_entry {
+ union { /* 0x000 */
+ u64 flags;
+ u8 padding0[0x200];
+ };
+ union { /* 0x200 */
+ u64 gprs[REC_RUN_GPRS];
+ u8 padding2[0x100];
+ };
+ union { /* 0x300 */
+ struct {
+ u64 gicv3_hcr;
+ u64 gicv3_lrs[REC_GIC_NUM_LRS];
+ };
+ u8 padding3[0x100];
+ };
+ u8 padding4[0x400];
+};
+
+struct rec_exit {
+ union { /* 0x000 */
+ u8 exit_reason;
+ u8 padding0[0x100];
+ };
+ union { /* 0x100 */
+ struct {
+ u64 esr;
+ u64 far;
+ u64 hpfar;
+ };
+ u8 padding1[0x100];
+ };
+ union { /* 0x200 */
+ u64 gprs[REC_RUN_GPRS];
+ u8 padding2[0x100];
+ };
+ union { /* 0x300 */
+ struct {
+ u64 gicv3_hcr;
+ u64 gicv3_lrs[REC_GIC_NUM_LRS];
+ u64 gicv3_misr;
+ u64 gicv3_vmcr;
+ };
+ u8 padding3[0x100];
+ };
+ union { /* 0x400 */
+ struct {
+ u64 cntp_ctl;
+ u64 cntp_cval;
+ u64 cntv_ctl;
+ u64 cntv_cval;
+ };
+ u8 padding4[0x100];
+ };
+ union { /* 0x500 */
+ struct {
+ u64 ripas_base;
+ u64 ripas_top;
+ u64 ripas_value;
+ };
+ u8 padding5[0x100];
+ };
+ union { /* 0x600 */
+ u16 imm;
+ u8 padding6[0x100];
+ };
+ union { /* 0x700 */
+ struct {
+ u64 pmu_ovf_status;
+ };
+ u8 padding7[0x100];
+ };
+};
+
+struct rec_run {
+ struct rec_entry entry;
+ struct rec_exit exit;
+};
+
+#define RMI_EXIT_SYNC 0x00
+#define RMI_EXIT_IRQ 0x01
+#define RMI_EXIT_FIQ 0x02
+#define RMI_EXIT_PSCI 0x03
+#define RMI_EXIT_RIPAS_CHANGE 0x04
+#define RMI_EXIT_HOST_CALL 0x05
+#define RMI_EXIT_SERROR 0x06
+
+#endif
--
2.34.1
The wrappers make the call sites easier to read and deal with the
boiler plate of handling the error codes from the RMM.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/rmi_cmds.h | 509 ++++++++++++++++++++++++++++++
1 file changed, 509 insertions(+)
create mode 100644 arch/arm64/include/asm/rmi_cmds.h
diff --git a/arch/arm64/include/asm/rmi_cmds.h b/arch/arm64/include/asm/rmi_cmds.h
new file mode 100644
index 000000000000..c21414127e8e
--- /dev/null
+++ b/arch/arm64/include/asm/rmi_cmds.h
@@ -0,0 +1,509 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2023 ARM Ltd.
+ */
+
+#ifndef __ASM_RMI_CMDS_H
+#define __ASM_RMI_CMDS_H
+
+#include <linux/arm-smccc.h>
+
+#include <asm/rmi_smc.h>
+
+struct rtt_entry {
+ unsigned long walk_level;
+ unsigned long desc;
+ int state;
+ int ripas;
+};
+
+/**
+ * rmi_data_create() - Create a Data Granule
+ * @rd: PA of the RD
+ * @data: PA of the target granule
+ * @ipa: IPA at which the granule will be mapped in the guest
+ * @src: PA of the source granule
+ * @flags: RMI_MEASURE_CONTENT if the contents should be measured
+ *
+ * Create a new Data Granule, copying contents from a Non-secure Granule.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_data_create(unsigned long rd, unsigned long data,
+ unsigned long ipa, unsigned long src,
+ unsigned long flags)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_DATA_CREATE, rd, data, ipa, src,
+ flags, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_data_create_unknown() - Create a Data Granule with unknown contents
+ * @rd: PA of the RD
+ * @data: PA of the target granule
+ * @ipa: IPA at which the granule will be mapped in the guest
+ *
+ * Create a new Data Granule with unknown contents
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_data_create_unknown(unsigned long rd,
+ unsigned long data,
+ unsigned long ipa)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_DATA_CREATE_UNKNOWN, rd, data, ipa, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_data_destroy() - Destroy a Data Granule
+ * @rd: PA of the RD
+ * @ipa: IPA at which the granule is mapped in the guest
+ * @data_out: PA of the granule which was destroyed
+ * @top_out: Top IPA of non-live RTT entries
+ *
+ * Transitions the granule to DESTROYED state, the address cannot be used by
+ * the guest for the lifetime of the Realm.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_data_destroy(unsigned long rd, unsigned long ipa,
+ unsigned long *data_out,
+ unsigned long *top_out)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_DATA_DESTROY, rd, ipa, &res);
+
+ *data_out = res.a1;
+ *top_out = res.a2;
+
+ return res.a0;
+}
+
+/**
+ * rmi_features() - Read feature register
+ * @index: Feature register index
+ * @out: Feature register value is written to this pointer
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_features(unsigned long index, unsigned long *out)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_FEATURES, index, &res);
+
+ *out = res.a1;
+ return res.a0;
+}
+
+/**
+ * rmi_granule_delegate() - Delegate a Granule
+ * @phys: PA of the Granule
+ *
+ * Delegate a Granule for use by the Realm World.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_granule_delegate(unsigned long phys)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_GRANULE_DELEGATE, phys, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_granule_undelegate() - Undelegate a Granule
+ * @phys: PA of the Granule
+ *
+ * Undelegate a Granule to allow use by the Normal World. Will fail if the
+ * Granule is in use.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_granule_undelegate(unsigned long phys)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_GRANULE_UNDELEGATE, phys, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_psci_complete() - Complete pending PSCI command
+ * @calling_rec: PA of the calling REC
+ * @target_rec: PA of the target REC
+ * @status: Status of the PSCI request
+ *
+ * Completes a pending PSCI command which was called with an MPIDR argument, by
+ * providing the corresponding REC.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_psci_complete(unsigned long calling_rec,
+ unsigned long target_rec,
+ unsigned long status)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_PSCI_COMPLETE, calling_rec, target_rec,
+ status, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_realm_activate() - Active a Realm
+ * @rd: PA of the RD
+ *
+ * Mark a Realm as Active signalling that creation is complete and allowing
+ * execution of the Realm.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_realm_activate(unsigned long rd)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_REALM_ACTIVATE, rd, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_realm_create() - Create a Realm
+ * @rd: PA of the RD
+ * @params_ptr: PA of Realm parameters
+ *
+ * Create a new Realm using the given parameters.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_realm_create(unsigned long rd, unsigned long params_ptr)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_REALM_CREATE, rd, params_ptr, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_realm_destroy() - Destroy a Realm
+ * @rd: PA of the RD
+ *
+ * Destroys a Realm, all objects belonging to the Realm must be destroyed first.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_realm_destroy(unsigned long rd)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_REALM_DESTROY, rd, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_rec_aux_count() - Get number of auxiliary Granules required
+ * @rd: PA of the RD
+ * @aux_count: Number of pages written to this pointer
+ *
+ * A REC may require extra auxiliary pages to be delegateed for the RMM to
+ * store metadata (not visible to the normal world) in. This function provides
+ * the number of pages that are required.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_rec_aux_count(unsigned long rd, unsigned long *aux_count)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_REC_AUX_COUNT, rd, &res);
+
+ *aux_count = res.a1;
+ return res.a0;
+}
+
+/**
+ * rmi_rec_create() - Create a REC
+ * @rd: PA of the RD
+ * @rec: PA of the target REC
+ * @params_ptr: PA of REC parameters
+ *
+ * Create a REC using the parameters specified in the struct rec_params pointed
+ * to by @params_ptr.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_rec_create(unsigned long rd, unsigned long rec,
+ unsigned long params_ptr)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_REC_CREATE, rd, rec, params_ptr, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_rec_destroy() - Destroy a REC
+ * @rec: PA of the target REC
+ *
+ * Destroys a REC. The REC must not be running.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_rec_destroy(unsigned long rec)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_REC_DESTROY, rec, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_rec_enter() - Enter a REC
+ * @rec: PA of the target REC
+ * @run_ptr: PA of RecRun structure
+ *
+ * Starts (or continues) execution within a REC.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_rec_enter(unsigned long rec, unsigned long run_ptr)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_REC_ENTER, rec, run_ptr, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_rtt_create() - Creates an RTT
+ * @rd: PA of the RD
+ * @rtt: PA of the target RTT
+ * @ipa: Base of the IPA range described by the RTT
+ * @level: Depth of the RTT within the tree
+ *
+ * Creates an RTT (Realm Translation Table) at the specified address and level
+ * within the realm.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_rtt_create(unsigned long rd, unsigned long rtt,
+ unsigned long ipa, long level)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_RTT_CREATE, rd, rtt, ipa, level, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_rtt_destroy() - Destroy an RTT
+ * @rd: PA of the RD
+ * @ipa: Base of the IPA range described by the RTT
+ * @level: Depth of the RTT within the tree
+ * @out_rtt: Pointer to write the PA of the RTT which was destroyed
+ * @out_top: Pointer to write the top IPA of non-live RTT entries
+ *
+ * Destroys an RTT. The RTT must be empty.
+ *
+ * Return: RMI return code.
+ */
+static inline int rmi_rtt_destroy(unsigned long rd,
+ unsigned long ipa,
+ long level,
+ unsigned long *out_rtt,
+ unsigned long *out_top)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_RTT_DESTROY, rd, ipa, level, &res);
+
+ *out_rtt = res.a1;
+ *out_top = res.a2;
+
+ return res.a0;
+}
+
+/**
+ * rmi_rtt_fold() - Fold an RTT
+ * @rd: PA of the RD
+ * @ipa: Base of the IPA range described by the RTT
+ * @level: Depth of the RTT within the tree
+ * @out_rtt: Pointer to write the PA of the RTT which was destroyed
+ *
+ * Folds an RTT. If all entries with the RTT are 'homogeneous' the RTT can be
+ * folded into the parent and the RTT destroyed.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_rtt_fold(unsigned long rd, unsigned long ipa,
+ long level, unsigned long *out_rtt)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_RTT_FOLD, rd, ipa, level, &res);
+
+ *out_rtt = res.a1;
+
+ return res.a0;
+}
+
+/**
+ * rmi_rtt_init_ripas() - Set RIPAS for new Realm
+ * @rd: PA of the RD
+ * @base: Base of target IPA region
+ * @top: Top of target IPA region
+ * @out_top: Top IPA of range whose RIPAS was modified
+ *
+ * Sets the RIPAS of a target IPA range to RAM, for a Realm in the NEW state.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_rtt_init_ripas(unsigned long rd, unsigned long base,
+ unsigned long top, unsigned long *out_top)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_RTT_INIT_RIPAS, rd, base, top, &res);
+
+ *out_top = res.a1;
+
+ return res.a0;
+}
+
+/**
+ * rmi_rtt_map_unprotected() - Map NS pages into a Realm
+ * @rd: PA of the RD
+ * @ipa: Base IPA of the mapping
+ * @level: Depth within the RTT tree
+ * @desc: RTTE descriptor
+ *
+ * Create a mapping from an Unprotected IPA to a Non-secure PA.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_rtt_map_unprotected(unsigned long rd,
+ unsigned long ipa,
+ long level,
+ unsigned long desc)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_RTT_MAP_UNPROTECTED, rd, ipa, level,
+ desc, &res);
+
+ return res.a0;
+}
+
+/**
+ * rmi_rtt_read_entry() - Read an RTTE
+ * @rd: PA of the RD
+ * @ipa: IPA for which to read the RTTE
+ * @level: RTT level at which to read the RTTE
+ * @rtt: Output structure describing the RTTE
+ *
+ * Reads a RTTE (Realm Translation Table Entry).
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_rtt_read_entry(unsigned long rd, unsigned long ipa,
+ long level, struct rtt_entry *rtt)
+{
+ struct arm_smccc_1_2_regs regs = {
+ SMC_RMI_RTT_READ_ENTRY,
+ rd, ipa, level
+ };
+
+ arm_smccc_1_2_smc(®s, ®s);
+
+ rtt->walk_level = regs.a1;
+ rtt->state = regs.a2 & 0xFF;
+ rtt->desc = regs.a3;
+ rtt->ripas = regs.a4;
+
+ return regs.a0;
+}
+
+/**
+ * rmi_rtt_set_ripas() - Set RIPAS for an running realm
+ * @rd: PA of the RD
+ * @rec: PA of the REC making the request
+ * @base: Base of target IPA region
+ * @top: Top of target IPA region
+ * @out_top: Pointer to write top IPA of range whose RIPAS was modified
+ *
+ * Completes a request made by the Realm to change the RIPAS of a target IPA
+ * range.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_rtt_set_ripas(unsigned long rd, unsigned long rec,
+ unsigned long base, unsigned long top,
+ unsigned long *out_top)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_RTT_SET_RIPAS, rd, rec, base, top, &res);
+
+ *out_top = res.a1;
+
+ return res.a0;
+}
+
+/**
+ * rmi_rtt_unmap_unprotected() - Remove a NS mapping
+ * @rd: PA of the RD
+ * @ipa: Base IPA of the mapping
+ * @level: Depth within the RTT tree
+ * @out_top: Pointer to write top IPA of non-live RTT entries
+ *
+ * Removes a mapping at an Unprotected IPA.
+ *
+ * Return: RMI return code
+ */
+static inline int rmi_rtt_unmap_unprotected(unsigned long rd,
+ unsigned long ipa,
+ long level,
+ unsigned long *out_top)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_1_1_invoke(SMC_RMI_RTT_UNMAP_UNPROTECTED, rd, ipa,
+ level, &res);
+
+ *out_top = res.a1;
+
+ return res.a0;
+}
+
+/**
+ * rmi_rtt_get_phys() - Get the PA from a RTTE
+ * @rtt: The RTTE
+ *
+ * Return: the physical address from a RTT entry.
+ */
+static inline phys_addr_t rmi_rtt_get_phys(struct rtt_entry *rtt)
+{
+ return rtt->desc & GENMASK(47, 12);
+}
+
+#endif
--
2.34.1
Query the RMI version number and check if it is a compatible version. A
static key is also provided to signal that a supported RMM is available.
Functions are provided to query if a VM or VCPU is a realm (or rec)
which currently will always return false.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_emulate.h | 18 +++++++++
arch/arm64/include/asm/kvm_host.h | 4 ++
arch/arm64/include/asm/kvm_rme.h | 56 ++++++++++++++++++++++++++++
arch/arm64/include/asm/virt.h | 1 +
arch/arm64/kvm/Makefile | 3 +-
arch/arm64/kvm/arm.c | 9 +++++
arch/arm64/kvm/rme.c | 52 ++++++++++++++++++++++++++
7 files changed, 142 insertions(+), 1 deletion(-)
create mode 100644 arch/arm64/include/asm/kvm_rme.h
create mode 100644 arch/arm64/kvm/rme.c
diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
index 975af30af31f..6f08398537e2 100644
--- a/arch/arm64/include/asm/kvm_emulate.h
+++ b/arch/arm64/include/asm/kvm_emulate.h
@@ -611,4 +611,22 @@ static __always_inline void kvm_reset_cptr_el2(struct kvm_vcpu *vcpu)
kvm_write_cptr_el2(val);
}
+
+static inline bool kvm_is_realm(struct kvm *kvm)
+{
+ if (static_branch_unlikely(&kvm_rme_is_available))
+ return kvm->arch.is_realm;
+ return false;
+}
+
+static inline enum realm_state kvm_realm_state(struct kvm *kvm)
+{
+ return READ_ONCE(kvm->arch.realm.state);
+}
+
+static inline bool vcpu_is_rec(struct kvm_vcpu *vcpu)
+{
+ return false;
+}
+
#endif /* __ARM64_KVM_EMULATE_H__ */
diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index 9e8a496fb284..63b68b85db3f 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -27,6 +27,7 @@
#include <asm/fpsimd.h>
#include <asm/kvm.h>
#include <asm/kvm_asm.h>
+#include <asm/kvm_rme.h>
#include <asm/vncr_mapping.h>
#define __KVM_HAVE_ARCH_INTC_INITIALIZED
@@ -348,6 +349,9 @@ struct kvm_arch {
* the associated pKVM instance in the hypervisor.
*/
struct kvm_protected_vm pkvm;
+
+ bool is_realm;
+ struct realm realm;
};
struct kvm_vcpu_fault_info {
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
new file mode 100644
index 000000000000..922da3f47227
--- /dev/null
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -0,0 +1,56 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2023 ARM Ltd.
+ */
+
+#ifndef __ASM_KVM_RME_H
+#define __ASM_KVM_RME_H
+
+/**
+ * enum realm_state - State of a Realm
+ */
+enum realm_state {
+ /**
+ * @REALM_STATE_NONE:
+ * Realm has not yet been created. rmi_realm_create() may be
+ * called to create the realm.
+ */
+ REALM_STATE_NONE,
+ /**
+ * @REALM_STATE_NEW:
+ * Realm is under construction, not eligible for execution. Pages
+ * may be populated with rmi_data_create().
+ */
+ REALM_STATE_NEW,
+ /**
+ * @REALM_STATE_ACTIVE:
+ * Realm has been created and is eligible for execution with
+ * rmi_rec_enter(). Pages may no longer be populated with
+ * rmi_data_create().
+ */
+ REALM_STATE_ACTIVE,
+ /**
+ * @REALM_STATE_DYING:
+ * Realm is in the process of being destroyed or has already been
+ * destroyed.
+ */
+ REALM_STATE_DYING,
+ /**
+ * @REALM_STATE_DEAD:
+ * Realm has been destroyed.
+ */
+ REALM_STATE_DEAD
+};
+
+/**
+ * struct realm - Additional per VM data for a Realm
+ *
+ * @state: The lifetime state machine for the realm
+ */
+struct realm {
+ enum realm_state state;
+};
+
+int kvm_init_rme(void);
+
+#endif
diff --git a/arch/arm64/include/asm/virt.h b/arch/arm64/include/asm/virt.h
index 261d6e9df2e1..12cf36c38189 100644
--- a/arch/arm64/include/asm/virt.h
+++ b/arch/arm64/include/asm/virt.h
@@ -81,6 +81,7 @@ void __hyp_reset_vectors(void);
bool is_kvm_arm_initialised(void);
DECLARE_STATIC_KEY_FALSE(kvm_protected_mode_initialized);
+DECLARE_STATIC_KEY_FALSE(kvm_rme_is_available);
/* Reports the availability of HYP mode */
static inline bool is_hyp_mode_available(void)
diff --git a/arch/arm64/kvm/Makefile b/arch/arm64/kvm/Makefile
index c0c050e53157..1c1d8cdf381f 100644
--- a/arch/arm64/kvm/Makefile
+++ b/arch/arm64/kvm/Makefile
@@ -20,7 +20,8 @@ kvm-y += arm.o mmu.o mmio.o psci.o hypercalls.o pvtime.o \
vgic/vgic-v3.o vgic/vgic-v4.o \
vgic/vgic-mmio.o vgic/vgic-mmio-v2.o \
vgic/vgic-mmio-v3.o vgic/vgic-kvm-device.o \
- vgic/vgic-its.o vgic/vgic-debug.o
+ vgic/vgic-its.o vgic/vgic-debug.o \
+ rme.o
kvm-$(CONFIG_HW_PERF_EVENTS) += pmu-emul.o pmu.o
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index 3dee5490eea9..2056c660c5ee 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -38,6 +38,7 @@
#include <asm/kvm_mmu.h>
#include <asm/kvm_nested.h>
#include <asm/kvm_pkvm.h>
+#include <asm/kvm_rme.h>
#include <asm/kvm_emulate.h>
#include <asm/sections.h>
@@ -47,6 +48,8 @@
static enum kvm_mode kvm_mode = KVM_MODE_DEFAULT;
+DEFINE_STATIC_KEY_FALSE(kvm_rme_is_available);
+
DECLARE_KVM_HYP_PER_CPU(unsigned long, kvm_hyp_vector);
DEFINE_PER_CPU(unsigned long, kvm_arm_hyp_stack_page);
@@ -2562,6 +2565,12 @@ static __init int kvm_arm_init(void)
in_hyp_mode = is_kernel_in_hyp_mode();
+ if (in_hyp_mode) {
+ err = kvm_init_rme();
+ if (err)
+ return err;
+ }
+
if (cpus_have_final_cap(ARM64_WORKAROUND_DEVICE_LOAD_ACQUIRE) ||
cpus_have_final_cap(ARM64_WORKAROUND_1508412))
kvm_info("Guests without required CPU erratum workarounds can deadlock system!\n" \
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
new file mode 100644
index 000000000000..3dbbf9d046bf
--- /dev/null
+++ b/arch/arm64/kvm/rme.c
@@ -0,0 +1,52 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2023 ARM Ltd.
+ */
+
+#include <linux/kvm_host.h>
+
+#include <asm/rmi_cmds.h>
+#include <asm/virt.h>
+
+static int rmi_check_version(void)
+{
+ struct arm_smccc_res res;
+ int version_major, version_minor;
+ unsigned long host_version = RMI_ABI_VERSION(RMI_ABI_MAJOR_VERSION,
+ RMI_ABI_MINOR_VERSION);
+
+ arm_smccc_1_1_invoke(SMC_RMI_VERSION, host_version, &res);
+
+ if (res.a0 == SMCCC_RET_NOT_SUPPORTED)
+ return -ENXIO;
+
+ version_major = RMI_ABI_VERSION_GET_MAJOR(res.a1);
+ version_minor = RMI_ABI_VERSION_GET_MINOR(res.a1);
+
+ if (version_major != RMI_ABI_MAJOR_VERSION) {
+ kvm_err("Unsupported RMI ABI (v%d.%d) host supports v%d.%d\n",
+ version_major, version_minor,
+ RMI_ABI_MAJOR_VERSION,
+ RMI_ABI_MINOR_VERSION);
+ return -ENXIO;
+ }
+
+ kvm_info("RMI ABI version %d.%d\n", version_major, version_minor);
+
+ return 0;
+}
+
+int kvm_init_rme(void)
+{
+ if (PAGE_SIZE != SZ_4K)
+ /* Only 4k page size on the host is supported */
+ return 0;
+
+ if (rmi_check_version())
+ /* Continue without realm support */
+ return 0;
+
+ /* Future patch will enable static branch kvm_rme_is_available */
+
+ return 0;
+}
--
2.34.1
There is one (multiplexed) CAP which can be used to create, populate and
then activate the realm.
Co-developed-by: Suzuki K Poulose <[email protected]>
Signed-off-by: Suzuki K Poulose <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
Documentation/virt/kvm/api.rst | 1 +
arch/arm64/include/uapi/asm/kvm.h | 49 +++++++++++++++++++++++++++++++
include/uapi/linux/kvm.h | 11 +++++++
3 files changed, 61 insertions(+)
diff --git a/Documentation/virt/kvm/api.rst b/Documentation/virt/kvm/api.rst
index 0b5a33ee71ee..b4bd3d0928a2 100644
--- a/Documentation/virt/kvm/api.rst
+++ b/Documentation/virt/kvm/api.rst
@@ -5064,6 +5064,7 @@ Recognised values for feature:
===== ===========================================
arm64 KVM_ARM_VCPU_SVE (requires KVM_CAP_ARM_SVE)
+ arm64 KVM_ARM_VCPU_REC (requires KVM_CAP_ARM_RME)
===== ===========================================
Finalizes the configuration of the specified vcpu feature.
diff --git a/arch/arm64/include/uapi/asm/kvm.h b/arch/arm64/include/uapi/asm/kvm.h
index 964df31da975..86110a8cc6be 100644
--- a/arch/arm64/include/uapi/asm/kvm.h
+++ b/arch/arm64/include/uapi/asm/kvm.h
@@ -108,6 +108,7 @@ struct kvm_regs {
#define KVM_ARM_VCPU_PTRAUTH_ADDRESS 5 /* VCPU uses address authentication */
#define KVM_ARM_VCPU_PTRAUTH_GENERIC 6 /* VCPU uses generic authentication */
#define KVM_ARM_VCPU_HAS_EL2 7 /* Support nested virtualization */
+#define KVM_ARM_VCPU_REC 8 /* VCPU REC state as part of Realm */
struct kvm_vcpu_init {
__u32 target;
@@ -418,6 +419,54 @@ enum {
#define KVM_DEV_ARM_VGIC_SAVE_PENDING_TABLES 3
#define KVM_DEV_ARM_ITS_CTRL_RESET 4
+/* KVM_CAP_ARM_RME on VM fd */
+#define KVM_CAP_ARM_RME_CONFIG_REALM 0
+#define KVM_CAP_ARM_RME_CREATE_RD 1
+#define KVM_CAP_ARM_RME_INIT_IPA_REALM 2
+#define KVM_CAP_ARM_RME_POPULATE_REALM 3
+#define KVM_CAP_ARM_RME_ACTIVATE_REALM 4
+
+#define KVM_CAP_ARM_RME_MEASUREMENT_ALGO_SHA256 0
+#define KVM_CAP_ARM_RME_MEASUREMENT_ALGO_SHA512 1
+
+#define KVM_CAP_ARM_RME_RPV_SIZE 64
+
+/* List of configuration items accepted for KVM_CAP_ARM_RME_CONFIG_REALM */
+#define KVM_CAP_ARM_RME_CFG_RPV 0
+#define KVM_CAP_ARM_RME_CFG_HASH_ALGO 1
+
+struct kvm_cap_arm_rme_config_item {
+ __u32 cfg;
+ union {
+ /* cfg == KVM_CAP_ARM_RME_CFG_RPV */
+ struct {
+ __u8 rpv[KVM_CAP_ARM_RME_RPV_SIZE];
+ };
+
+ /* cfg == KVM_CAP_ARM_RME_CFG_HASH_ALGO */
+ struct {
+ __u32 hash_algo;
+ };
+
+ /* Fix the size of the union */
+ __u8 reserved[256];
+ };
+};
+
+#define KVM_ARM_RME_POPULATE_FLAGS_MEASURE (1U << 0)
+struct kvm_cap_arm_rme_populate_realm_args {
+ __u64 populate_ipa_base;
+ __u64 populate_ipa_size;
+ __u32 flags;
+ __u32 reserved[3];
+};
+
+struct kvm_cap_arm_rme_init_ipa_args {
+ __u64 init_ipa_base;
+ __u64 init_ipa_size;
+ __u32 reserved[4];
+};
+
/* Device Control API on vcpu fd */
#define KVM_ARM_VCPU_PMU_V3_CTRL 0
#define KVM_ARM_VCPU_PMU_V3_IRQ 0
diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
index 2190adbe3002..a1147036d1bd 100644
--- a/include/uapi/linux/kvm.h
+++ b/include/uapi/linux/kvm.h
@@ -918,6 +918,8 @@ struct kvm_enable_cap {
#define KVM_CAP_GUEST_MEMFD 234
#define KVM_CAP_VM_TYPES 235
+#define KVM_CAP_ARM_RME 300 /* FIXME: Large number to prevent conflicts */
+
struct kvm_irq_routing_irqchip {
__u32 irqchip;
__u32 pin;
@@ -1548,4 +1550,13 @@ struct kvm_create_guest_memfd {
__u64 reserved[6];
};
+/* Available with KVM_CAP_ARM_RME, only for VMs with KVM_VM_TYPE_ARM_REALM */
+struct kvm_arm_rmm_psci_complete {
+ __u64 target_mpidr;
+ __u32 psci_status;
+ __u32 padding[3];
+};
+
+#define KVM_ARM_VCPU_RMM_PSCI_COMPLETE _IOW(KVMIO, 0xd2, struct kvm_arm_rmm_psci_complete)
+
#endif /* __LINUX_KVM_H */
--
2.34.1
Add the KVM_CAP_ARM_RME_CREATE_FD ioctl to create a realm. This involves
delegating pages to the RMM to hold the Realm Descriptor (RD) and for
the base level of the Realm Translation Tables (RTT). A VMID also need
to be picked, since the RMM has a separate VMID address space a
dedicated allocator is added for this purpose.
KVM_CAP_ARM_RME_CONFIG_REALM is provided to allow configuring the realm
before it is created.
Co-developed-by: Suzuki K Poulose <[email protected]>
Signed-off-by: Suzuki K Poulose <[email protected]>
Signed-off-by: Steven Price <[email protected]>
Signed-off-by: Jean-Philippe Brucker <[email protected]>
---
arch/arm64/include/asm/kvm_emulate.h | 5 +
arch/arm64/include/asm/kvm_rme.h | 19 ++
arch/arm64/kvm/arm.c | 18 ++
arch/arm64/kvm/mmu.c | 15 +-
arch/arm64/kvm/rme.c | 282 +++++++++++++++++++++++++++
5 files changed, 337 insertions(+), 2 deletions(-)
diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
index 6f08398537e2..c606316f4729 100644
--- a/arch/arm64/include/asm/kvm_emulate.h
+++ b/arch/arm64/include/asm/kvm_emulate.h
@@ -624,6 +624,11 @@ static inline enum realm_state kvm_realm_state(struct kvm *kvm)
return READ_ONCE(kvm->arch.realm.state);
}
+static inline bool kvm_realm_is_created(struct kvm *kvm)
+{
+ return kvm_is_realm(kvm) && kvm_realm_state(kvm) != REALM_STATE_NONE;
+}
+
static inline bool vcpu_is_rec(struct kvm_vcpu *vcpu)
{
return false;
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index 922da3f47227..cf8cc4d30364 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -6,6 +6,8 @@
#ifndef __ASM_KVM_RME_H
#define __ASM_KVM_RME_H
+#include <uapi/linux/kvm.h>
+
/**
* enum realm_state - State of a Realm
*/
@@ -46,11 +48,28 @@ enum realm_state {
* struct realm - Additional per VM data for a Realm
*
* @state: The lifetime state machine for the realm
+ * @rd: Kernel mapping of the Realm Descriptor (RD)
+ * @params: Parameters for the RMI_REALM_CREATE command
+ * @num_aux: The number of auxiliary pages required by the RMM
+ * @vmid: VMID to be used by the RMM for the realm
+ * @ia_bits: Number of valid Input Address bits in the IPA
*/
struct realm {
enum realm_state state;
+
+ void *rd;
+ struct realm_params *params;
+
+ unsigned long num_aux;
+ unsigned int vmid;
+ unsigned int ia_bits;
};
int kvm_init_rme(void);
+u32 kvm_realm_ipa_limit(void);
+
+int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap);
+int kvm_init_realm_vm(struct kvm *kvm);
+void kvm_destroy_realm(struct kvm *kvm);
#endif
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index 2056c660c5ee..5729ea430d6d 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -119,6 +119,13 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
}
mutex_unlock(&kvm->slots_lock);
break;
+ case KVM_CAP_ARM_RME:
+ if (!kvm_is_realm(kvm))
+ return -EINVAL;
+ mutex_lock(&kvm->lock);
+ r = kvm_realm_enable_cap(kvm, cap);
+ mutex_unlock(&kvm->lock);
+ break;
default:
r = -EINVAL;
break;
@@ -179,6 +186,13 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
bitmap_zero(kvm->arch.vcpu_features, KVM_VCPU_MAX_FEATURES);
+ /* Initialise the realm bits after the generic bits are enabled */
+ if (kvm_is_realm(kvm)) {
+ ret = kvm_init_realm_vm(kvm);
+ if (ret)
+ goto err_free_cpumask;
+ }
+
return 0;
err_free_cpumask:
@@ -219,6 +233,7 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
kvm_unshare_hyp(kvm, kvm + 1);
kvm_arm_teardown_hypercalls(kvm);
+ kvm_destroy_realm(kvm);
}
int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
@@ -328,6 +343,9 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
case KVM_CAP_ARM_SUPPORTED_REG_MASK_RANGES:
r = BIT(0);
break;
+ case KVM_CAP_ARM_RME:
+ r = static_key_enabled(&kvm_rme_is_available);
+ break;
default:
r = 0;
}
diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index 18680771cdb0..aae365647b62 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -872,6 +872,10 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
struct kvm_pgtable *pgt;
u64 mmfr0, mmfr1;
u32 phys_shift;
+ u32 ipa_limit = kvm_ipa_limit;
+
+ if (kvm_is_realm(kvm))
+ ipa_limit = kvm_realm_ipa_limit();
if (type & ~KVM_VM_TYPE_ARM_IPA_SIZE_MASK)
return -EINVAL;
@@ -880,12 +884,12 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
if (is_protected_kvm_enabled()) {
phys_shift = kvm_ipa_limit;
} else if (phys_shift) {
- if (phys_shift > kvm_ipa_limit ||
+ if (phys_shift > ipa_limit ||
phys_shift < ARM64_MIN_PARANGE_BITS)
return -EINVAL;
} else {
phys_shift = KVM_PHYS_SHIFT;
- if (phys_shift > kvm_ipa_limit) {
+ if (phys_shift > ipa_limit) {
pr_warn_once("%s using unsupported default IPA limit, upgrade your VMM\n",
current->comm);
return -EINVAL;
@@ -1014,6 +1018,13 @@ void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu)
struct kvm_pgtable *pgt = NULL;
write_lock(&kvm->mmu_lock);
+ if (kvm_is_realm(kvm) &&
+ (kvm_realm_state(kvm) != REALM_STATE_DEAD &&
+ kvm_realm_state(kvm) != REALM_STATE_NONE)) {
+ /* TODO: teardown rtts */
+ write_unlock(&kvm->mmu_lock);
+ return;
+ }
pgt = mmu->pgt;
if (pgt) {
mmu->pgd_phys = 0;
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 3dbbf9d046bf..658d14e8d87d 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -5,9 +5,20 @@
#include <linux/kvm_host.h>
+#include <asm/kvm_emulate.h>
+#include <asm/kvm_mmu.h>
#include <asm/rmi_cmds.h>
#include <asm/virt.h>
+#include <asm/kvm_pgtable.h>
+
+static unsigned long rmm_feat_reg0;
+
+static bool rme_supports(unsigned long feature)
+{
+ return !!u64_get_bits(rmm_feat_reg0, feature);
+}
+
static int rmi_check_version(void)
{
struct arm_smccc_res res;
@@ -36,8 +47,272 @@ static int rmi_check_version(void)
return 0;
}
+u32 kvm_realm_ipa_limit(void)
+{
+ return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
+}
+
+static int get_start_level(struct realm *realm)
+{
+ return 4 - stage2_pgtable_levels(realm->ia_bits);
+}
+
+static int realm_create_rd(struct kvm *kvm)
+{
+ struct realm *realm = &kvm->arch.realm;
+ struct realm_params *params = realm->params;
+ void *rd = NULL;
+ phys_addr_t rd_phys, params_phys;
+ struct kvm_pgtable *pgt = kvm->arch.mmu.pgt;
+ int i, r;
+
+ if (WARN_ON(realm->rd) || WARN_ON(!realm->params))
+ return -EEXIST;
+
+ rd = (void *)__get_free_page(GFP_KERNEL);
+ if (!rd)
+ return -ENOMEM;
+
+ rd_phys = virt_to_phys(rd);
+ if (rmi_granule_delegate(rd_phys)) {
+ r = -ENXIO;
+ goto out;
+ }
+
+ for (i = 0; i < pgt->pgd_pages; i++) {
+ phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
+
+ if (rmi_granule_delegate(pgd_phys)) {
+ r = -ENXIO;
+ goto out_undelegate_tables;
+ }
+ }
+
+ realm->ia_bits = VTCR_EL2_IPA(kvm->arch.mmu.vtcr);
+
+ params->rtt_level_start = get_start_level(realm);
+ params->rtt_num_start = pgt->pgd_pages;
+ params->rtt_base = kvm->arch.mmu.pgd_phys;
+ params->vmid = realm->vmid;
+
+ params_phys = virt_to_phys(params);
+
+ if (rmi_realm_create(rd_phys, params_phys)) {
+ r = -ENXIO;
+ goto out_undelegate_tables;
+ }
+
+ realm->rd = rd;
+
+ if (WARN_ON(rmi_rec_aux_count(rd_phys, &realm->num_aux))) {
+ WARN_ON(rmi_realm_destroy(rd_phys));
+ goto out_undelegate_tables;
+ }
+
+ return 0;
+
+out_undelegate_tables:
+ while (--i >= 0) {
+ phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
+
+ WARN_ON(rmi_granule_undelegate(pgd_phys));
+ }
+ WARN_ON(rmi_granule_undelegate(rd_phys));
+out:
+ free_page((unsigned long)rd);
+ return r;
+}
+
+/* Protects access to rme_vmid_bitmap */
+static DEFINE_SPINLOCK(rme_vmid_lock);
+static unsigned long *rme_vmid_bitmap;
+
+static int rme_vmid_init(void)
+{
+ unsigned int vmid_count = 1 << kvm_get_vmid_bits();
+
+ rme_vmid_bitmap = bitmap_zalloc(vmid_count, GFP_KERNEL);
+ if (!rme_vmid_bitmap) {
+ kvm_err("%s: Couldn't allocate rme vmid bitmap\n", __func__);
+ return -ENOMEM;
+ }
+
+ return 0;
+}
+
+static int rme_vmid_reserve(void)
+{
+ int ret;
+ unsigned int vmid_count = 1 << kvm_get_vmid_bits();
+
+ spin_lock(&rme_vmid_lock);
+ ret = bitmap_find_free_region(rme_vmid_bitmap, vmid_count, 0);
+ spin_unlock(&rme_vmid_lock);
+
+ return ret;
+}
+
+static void rme_vmid_release(unsigned int vmid)
+{
+ spin_lock(&rme_vmid_lock);
+ bitmap_release_region(rme_vmid_bitmap, vmid, 0);
+ spin_unlock(&rme_vmid_lock);
+}
+
+static int kvm_create_realm(struct kvm *kvm)
+{
+ struct realm *realm = &kvm->arch.realm;
+ int ret;
+
+ if (!kvm_is_realm(kvm) || kvm_realm_is_created(kvm))
+ return -EEXIST;
+
+ ret = rme_vmid_reserve();
+ if (ret < 0)
+ return ret;
+ realm->vmid = ret;
+
+ ret = realm_create_rd(kvm);
+ if (ret) {
+ rme_vmid_release(realm->vmid);
+ return ret;
+ }
+
+ WRITE_ONCE(realm->state, REALM_STATE_NEW);
+
+ /* The realm is up, free the parameters. */
+ free_page((unsigned long)realm->params);
+ realm->params = NULL;
+
+ return 0;
+}
+
+static int config_realm_hash_algo(struct realm *realm,
+ struct kvm_cap_arm_rme_config_item *cfg)
+{
+ switch (cfg->hash_algo) {
+ case KVM_CAP_ARM_RME_MEASUREMENT_ALGO_SHA256:
+ if (!rme_supports(RMI_FEATURE_REGISTER_0_HASH_SHA_256))
+ return -EINVAL;
+ break;
+ case KVM_CAP_ARM_RME_MEASUREMENT_ALGO_SHA512:
+ if (!rme_supports(RMI_FEATURE_REGISTER_0_HASH_SHA_512))
+ return -EINVAL;
+ break;
+ default:
+ return -EINVAL;
+ }
+ realm->params->hash_algo = cfg->hash_algo;
+ return 0;
+}
+
+static int kvm_rme_config_realm(struct kvm *kvm, struct kvm_enable_cap *cap)
+{
+ struct kvm_cap_arm_rme_config_item cfg;
+ struct realm *realm = &kvm->arch.realm;
+ int r = 0;
+
+ if (kvm_realm_is_created(kvm))
+ return -EINVAL;
+
+ if (copy_from_user(&cfg, (void __user *)cap->args[1], sizeof(cfg)))
+ return -EFAULT;
+
+ switch (cfg.cfg) {
+ case KVM_CAP_ARM_RME_CFG_RPV:
+ memcpy(&realm->params->rpv, &cfg.rpv, sizeof(cfg.rpv));
+ break;
+ case KVM_CAP_ARM_RME_CFG_HASH_ALGO:
+ r = config_realm_hash_algo(realm, &cfg);
+ break;
+ default:
+ r = -EINVAL;
+ }
+
+ return r;
+}
+
+int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
+{
+ int r = 0;
+
+ if (!kvm_is_realm(kvm))
+ return -EINVAL;
+
+ switch (cap->args[0]) {
+ case KVM_CAP_ARM_RME_CONFIG_REALM:
+ r = kvm_rme_config_realm(kvm, cap);
+ break;
+ case KVM_CAP_ARM_RME_CREATE_RD:
+ r = kvm_create_realm(kvm);
+ break;
+ default:
+ r = -EINVAL;
+ break;
+ }
+
+ return r;
+}
+
+void kvm_destroy_realm(struct kvm *kvm)
+{
+ struct realm *realm = &kvm->arch.realm;
+ struct kvm_pgtable *pgt = kvm->arch.mmu.pgt;
+ int i;
+
+ if (realm->params) {
+ free_page((unsigned long)realm->params);
+ realm->params = NULL;
+ }
+
+ if (!kvm_realm_is_created(kvm))
+ return;
+
+ WRITE_ONCE(realm->state, REALM_STATE_DYING);
+
+ if (realm->rd) {
+ phys_addr_t rd_phys = virt_to_phys(realm->rd);
+
+ if (WARN_ON(rmi_realm_destroy(rd_phys)))
+ return;
+ if (WARN_ON(rmi_granule_undelegate(rd_phys)))
+ return;
+ free_page((unsigned long)realm->rd);
+ realm->rd = NULL;
+ }
+
+ rme_vmid_release(realm->vmid);
+
+ for (i = 0; i < pgt->pgd_pages; i++) {
+ phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
+
+ if (WARN_ON(rmi_granule_undelegate(pgd_phys)))
+ return;
+ }
+
+ WRITE_ONCE(realm->state, REALM_STATE_DEAD);
+
+ kvm_free_stage2_pgd(&kvm->arch.mmu);
+}
+
+int kvm_init_realm_vm(struct kvm *kvm)
+{
+ struct realm_params *params;
+
+ params = (struct realm_params *)get_zeroed_page(GFP_KERNEL);
+ if (!params)
+ return -ENOMEM;
+
+ /* Default parameters, not exposed to user space */
+ params->s2sz = VTCR_EL2_IPA(kvm->arch.mmu.vtcr);
+ kvm->arch.realm.params = params;
+ return 0;
+}
+
int kvm_init_rme(void)
{
+ int ret;
+
if (PAGE_SIZE != SZ_4K)
/* Only 4k page size on the host is supported */
return 0;
@@ -46,6 +321,13 @@ int kvm_init_rme(void)
/* Continue without realm support */
return 0;
+ if (WARN_ON(rmi_features(0, &rmm_feat_reg0)))
+ return 0;
+
+ ret = rme_vmid_init();
+ if (ret)
+ return ret;
+
/* Future patch will enable static branch kvm_rme_is_available */
return 0;
--
2.34.1
From: Suzuki K Poulose <[email protected]>
Expose VM specific Debug HW register numbers.
Signed-off-by: Suzuki K Poulose <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/arm.c | 24 +++++++++++++++++++++---
1 file changed, 21 insertions(+), 3 deletions(-)
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index 5729ea430d6d..22da6493912a 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -67,6 +67,22 @@ bool is_kvm_arm_initialised(void)
return kvm_arm_initialised;
}
+static u32 kvm_arm_get_num_brps(struct kvm *kvm)
+{
+ if (!kvm || !kvm_is_realm(kvm))
+ return get_num_brps();
+ /* Realm guest is not debuggable. */
+ return 0;
+}
+
+static u32 kvm_arm_get_num_wrps(struct kvm *kvm)
+{
+ if (!kvm || !kvm_is_realm(kvm))
+ return get_num_wrps();
+ /* Realm guest is not debuggable. */
+ return 0;
+}
+
int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
{
return kvm_vcpu_exiting_guest_mode(vcpu) == IN_GUEST_MODE;
@@ -257,7 +273,6 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
case KVM_CAP_ARM_IRQ_LINE_LAYOUT_2:
case KVM_CAP_ARM_NISV_TO_USER:
case KVM_CAP_ARM_INJECT_EXT_DABT:
- case KVM_CAP_SET_GUEST_DEBUG:
case KVM_CAP_VCPU_ATTRIBUTES:
case KVM_CAP_PTP_KVM:
case KVM_CAP_ARM_SYSTEM_SUSPEND:
@@ -265,6 +280,9 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
case KVM_CAP_COUNTER_OFFSET:
r = 1;
break;
+ case KVM_CAP_SET_GUEST_DEBUG:
+ r = !kvm_is_realm(kvm);
+ break;
case KVM_CAP_SET_GUEST_DEBUG2:
return KVM_GUESTDBG_VALID_MASK;
case KVM_CAP_ARM_SET_DEVICE_ADDR:
@@ -310,10 +328,10 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
r = cpus_have_final_cap(ARM64_HAS_32BIT_EL1);
break;
case KVM_CAP_GUEST_DEBUG_HW_BPS:
- r = get_num_brps();
+ r = kvm_arm_get_num_brps(kvm);
break;
case KVM_CAP_GUEST_DEBUG_HW_WPS:
- r = get_num_wrps();
+ r = kvm_arm_get_num_wrps(kvm);
break;
case KVM_CAP_ARM_PMU_V3:
r = kvm_arm_support_pmu_v3();
--
2.34.1
Previously machine type was used purely for specifying the physical
address size of the guest. Reserve the higher bits to specify an ARM
specific machine type and declare a new type 'KVM_VM_TYPE_ARM_REALM'
used to create a realm guest.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/arm.c | 17 +++++++++++++++++
arch/arm64/kvm/mmu.c | 3 ---
include/uapi/linux/kvm.h | 19 +++++++++++++++----
3 files changed, 32 insertions(+), 7 deletions(-)
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index 22da6493912a..c5a6139d5454 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -173,6 +173,23 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
mutex_unlock(&kvm->lock);
#endif
+ if (type & ~(KVM_VM_TYPE_ARM_MASK | KVM_VM_TYPE_ARM_IPA_SIZE_MASK))
+ return -EINVAL;
+
+ switch (type & KVM_VM_TYPE_ARM_MASK) {
+ case KVM_VM_TYPE_ARM_NORMAL:
+ break;
+ case KVM_VM_TYPE_ARM_REALM:
+ kvm->arch.is_realm = true;
+ if (!kvm_is_realm(kvm)) {
+ /* Realm support unavailable */
+ return -EINVAL;
+ }
+ break;
+ default:
+ return -EINVAL;
+ }
+
ret = kvm_share_hyp(kvm, kvm + 1);
if (ret)
return ret;
diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index aae365647b62..af4564f3add5 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -877,9 +877,6 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
if (kvm_is_realm(kvm))
ipa_limit = kvm_realm_ipa_limit();
- if (type & ~KVM_VM_TYPE_ARM_IPA_SIZE_MASK)
- return -EINVAL;
-
phys_shift = KVM_VM_TYPE_ARM_IPA_SIZE(type);
if (is_protected_kvm_enabled()) {
phys_shift = kvm_ipa_limit;
diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
index a1147036d1bd..5153c837c8c7 100644
--- a/include/uapi/linux/kvm.h
+++ b/include/uapi/linux/kvm.h
@@ -635,14 +635,25 @@ struct kvm_enable_cap {
#define KVM_S390_SIE_PAGE_OFFSET 1
/*
- * On arm64, machine type can be used to request the physical
- * address size for the VM. Bits[7-0] are reserved for the guest
- * PA size shift (i.e, log2(PA_Size)). For backward compatibility,
- * value 0 implies the default IPA size, 40bits.
+ * On arm64, machine type can be used to request both the machine type and
+ * the physical address size for the VM.
+ *
+ * Bits[11-8] are reserved for the ARM specific machine type.
+ *
+ * Bits[7-0] are reserved for the guest PA size shift (i.e, log2(PA_Size)).
+ * For backward compatibility, value 0 implies the default IPA size, 40bits.
*/
+#define KVM_VM_TYPE_ARM_SHIFT 8
+#define KVM_VM_TYPE_ARM_MASK (0xfULL << KVM_VM_TYPE_ARM_SHIFT)
+#define KVM_VM_TYPE_ARM(_type) \
+ (((_type) << KVM_VM_TYPE_ARM_SHIFT) & KVM_VM_TYPE_ARM_MASK)
+#define KVM_VM_TYPE_ARM_NORMAL KVM_VM_TYPE_ARM(0)
+#define KVM_VM_TYPE_ARM_REALM KVM_VM_TYPE_ARM(1)
+
#define KVM_VM_TYPE_ARM_IPA_SIZE_MASK 0xffULL
#define KVM_VM_TYPE_ARM_IPA_SIZE(x) \
((x) & KVM_VM_TYPE_ARM_IPA_SIZE_MASK)
+
/*
* ioctls for /dev/kvm fds:
*/
--
2.34.1
Pages can only be populated/destroyed on the RMM at the 4KB granule,
this requires creating the full depth of RTTs. However if the pages are
going to be combined into a 4MB huge page the last RTT is only
temporarily needed. Similarly when freeing memory the huge page must be
temporarily split requiring temporary usage of the full depth oF RTTs.
To avoid needing to perform a temporary allocation and delegation of a
page for this purpose we keep a spare delegated page around. In
particular this avoids the need for memory allocation while destroying
the realm guest.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_rme.h | 5 +++++
arch/arm64/kvm/rme.c | 8 ++++++++
2 files changed, 13 insertions(+)
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index cf8cc4d30364..fba85e9ce3ae 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -50,6 +50,9 @@ enum realm_state {
* @state: The lifetime state machine for the realm
* @rd: Kernel mapping of the Realm Descriptor (RD)
* @params: Parameters for the RMI_REALM_CREATE command
+ * @spare_page: A physical page that has been delegated to the Realm world but
+ * is otherwise free. Used to avoid temporary allocation during
+ * RTT operations.
* @num_aux: The number of auxiliary pages required by the RMM
* @vmid: VMID to be used by the RMM for the realm
* @ia_bits: Number of valid Input Address bits in the IPA
@@ -60,6 +63,8 @@ struct realm {
void *rd;
struct realm_params *params;
+ phys_addr_t spare_page;
+
unsigned long num_aux;
unsigned int vmid;
unsigned int ia_bits;
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 658d14e8d87d..9652ec6ab2fd 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -103,6 +103,7 @@ static int realm_create_rd(struct kvm *kvm)
}
realm->rd = rd;
+ realm->spare_page = PHYS_ADDR_MAX;
if (WARN_ON(rmi_rec_aux_count(rd_phys, &realm->num_aux))) {
WARN_ON(rmi_realm_destroy(rd_phys));
@@ -283,6 +284,13 @@ void kvm_destroy_realm(struct kvm *kvm)
rme_vmid_release(realm->vmid);
+ if (realm->spare_page != PHYS_ADDR_MAX) {
+ /* Leak the page if the undelegate fails */
+ if (!WARN_ON(rmi_granule_undelegate(realm->spare_page)))
+ free_page((unsigned long)phys_to_virt(realm->spare_page));
+ realm->spare_page = PHYS_ADDR_MAX;
+ }
+
for (i = 0; i < pgt->pgd_pages; i++) {
phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
--
2.34.1
The RMM owns the stage 2 page tables for a realm, and KVM must request
that the RMM creates/destroys entries as necessary. The physical pages
to store the page tables are delegated to the realm as required, and can
be undelegated when no longer used.
Creating new RTTs is the easy part, tearing down is a little more
tricky. The result of realm_rtt_destroy() can be used to effectively
walk the tree and destroy the entries (undelegating pages that were
given to the realm).
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_rme.h | 19 ++++
arch/arm64/kvm/mmu.c | 6 +-
arch/arm64/kvm/rme.c | 171 +++++++++++++++++++++++++++++++
3 files changed, 193 insertions(+), 3 deletions(-)
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index fba85e9ce3ae..4ab5cb5e91b3 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -76,5 +76,24 @@ u32 kvm_realm_ipa_limit(void);
int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap);
int kvm_init_realm_vm(struct kvm *kvm);
void kvm_destroy_realm(struct kvm *kvm);
+void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits);
+
+#define RME_RTT_BLOCK_LEVEL 2
+#define RME_RTT_MAX_LEVEL 3
+
+#define RME_PAGE_SHIFT 12
+#define RME_PAGE_SIZE BIT(RME_PAGE_SHIFT)
+/* See ARM64_HW_PGTABLE_LEVEL_SHIFT() */
+#define RME_RTT_LEVEL_SHIFT(l) \
+ ((RME_PAGE_SHIFT - 3) * (4 - (l)) + 3)
+#define RME_L2_BLOCK_SIZE BIT(RME_RTT_LEVEL_SHIFT(2))
+
+static inline unsigned long rme_rtt_level_mapsize(int level)
+{
+ if (WARN_ON(level > RME_RTT_MAX_LEVEL))
+ return RME_PAGE_SIZE;
+
+ return (1UL << RME_RTT_LEVEL_SHIFT(level));
+}
#endif
diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index af4564f3add5..46f0c4e80ace 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -1012,17 +1012,17 @@ void stage2_unmap_vm(struct kvm *kvm)
void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu)
{
struct kvm *kvm = kvm_s2_mmu_to_kvm(mmu);
- struct kvm_pgtable *pgt = NULL;
+ struct kvm_pgtable *pgt;
write_lock(&kvm->mmu_lock);
+ pgt = mmu->pgt;
if (kvm_is_realm(kvm) &&
(kvm_realm_state(kvm) != REALM_STATE_DEAD &&
kvm_realm_state(kvm) != REALM_STATE_NONE)) {
- /* TODO: teardown rtts */
write_unlock(&kvm->mmu_lock);
+ kvm_realm_destroy_rtts(kvm, pgt->ia_bits);
return;
}
- pgt = mmu->pgt;
if (pgt) {
mmu->pgd_phys = 0;
mmu->pgt = NULL;
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 9652ec6ab2fd..09b59bcad8b6 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -47,6 +47,53 @@ static int rmi_check_version(void)
return 0;
}
+static phys_addr_t __alloc_delegated_page(struct realm *realm,
+ struct kvm_mmu_memory_cache *mc,
+ gfp_t flags)
+{
+ phys_addr_t phys = PHYS_ADDR_MAX;
+ void *virt;
+
+ if (realm->spare_page != PHYS_ADDR_MAX) {
+ swap(realm->spare_page, phys);
+ goto out;
+ }
+
+ if (mc)
+ virt = kvm_mmu_memory_cache_alloc(mc);
+ else
+ virt = (void *)__get_free_page(flags);
+
+ if (!virt)
+ goto out;
+
+ phys = virt_to_phys(virt);
+
+ if (rmi_granule_delegate(phys)) {
+ free_page((unsigned long)virt);
+
+ phys = PHYS_ADDR_MAX;
+ }
+
+out:
+ return phys;
+}
+
+static void free_delegated_page(struct realm *realm, phys_addr_t phys)
+{
+ if (realm->spare_page == PHYS_ADDR_MAX) {
+ realm->spare_page = phys;
+ return;
+ }
+
+ if (WARN_ON(rmi_granule_undelegate(phys))) {
+ /* Undelegate failed: leak the page */
+ return;
+ }
+
+ free_page((unsigned long)phys_to_virt(phys));
+}
+
u32 kvm_realm_ipa_limit(void)
{
return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
@@ -124,6 +171,130 @@ static int realm_create_rd(struct kvm *kvm)
return r;
}
+static int realm_rtt_destroy(struct realm *realm, unsigned long addr,
+ int level, phys_addr_t *rtt_granule,
+ unsigned long *next_addr)
+{
+ unsigned long out_rtt;
+ unsigned long out_top;
+ int ret;
+
+ ret = rmi_rtt_destroy(virt_to_phys(realm->rd), addr, level,
+ &out_rtt, &out_top);
+
+ if (rtt_granule)
+ *rtt_granule = out_rtt;
+ if (next_addr)
+ *next_addr = out_top;
+
+ return ret;
+}
+
+static int realm_tear_down_rtt_level(struct realm *realm, int level,
+ unsigned long start, unsigned long end)
+{
+ ssize_t map_size;
+ unsigned long addr, next_addr;
+
+ if (WARN_ON(level > RME_RTT_MAX_LEVEL))
+ return -EINVAL;
+
+ map_size = rme_rtt_level_mapsize(level - 1);
+
+ for (addr = start; addr < end; addr = next_addr) {
+ phys_addr_t rtt_granule;
+ int ret;
+ unsigned long align_addr = ALIGN(addr, map_size);
+
+ next_addr = ALIGN(addr + 1, map_size);
+
+ if (next_addr <= end && align_addr == addr) {
+ ret = realm_rtt_destroy(realm, addr, level,
+ &rtt_granule, &next_addr);
+ } else {
+ /* Recurse a level deeper */
+ ret = realm_tear_down_rtt_level(realm,
+ level + 1,
+ addr,
+ min(next_addr, end));
+ if (ret)
+ return ret;
+ continue;
+ }
+
+ switch (RMI_RETURN_STATUS(ret)) {
+ case RMI_SUCCESS:
+ if (!WARN_ON(rmi_granule_undelegate(rtt_granule)))
+ free_page((unsigned long)phys_to_virt(rtt_granule));
+ break;
+ case RMI_ERROR_RTT:
+ if (next_addr > addr) {
+ /* unassigned or destroyed */
+ break;
+ }
+ if (WARN_ON(RMI_RETURN_INDEX(ret) != level))
+ return -EBUSY;
+ if (WARN_ON(level == RME_RTT_MAX_LEVEL)) {
+ // Live entry
+ return -EBUSY;
+ }
+ /* Recurse a level deeper */
+ next_addr = ALIGN(addr + 1, map_size);
+ ret = realm_tear_down_rtt_level(realm,
+ level + 1,
+ addr,
+ next_addr);
+ if (ret)
+ return ret;
+ /* Try again at this level */
+ next_addr = addr;
+ break;
+ default:
+ WARN_ON(1);
+ return -ENXIO;
+ }
+ }
+
+ return 0;
+}
+
+static int realm_tear_down_rtt_range(struct realm *realm,
+ unsigned long start, unsigned long end)
+{
+ return realm_tear_down_rtt_level(realm, get_start_level(realm) + 1,
+ start, end);
+}
+
+static void ensure_spare_page(struct realm *realm)
+{
+ phys_addr_t tmp_rtt;
+
+ /*
+ * Make sure we have a spare delegated page for tearing down the
+ * block mappings. We do this by allocating then freeing a page.
+ * We must use Atomic allocations as we are called with kvm->mmu_lock
+ * held.
+ */
+ tmp_rtt = __alloc_delegated_page(realm, NULL, GFP_ATOMIC);
+
+ /*
+ * If the allocation failed, continue as we may not have a block level
+ * mapping so it may not be fatal, otherwise free it to assign it
+ * to the spare page.
+ */
+ if (tmp_rtt != PHYS_ADDR_MAX)
+ free_delegated_page(realm, tmp_rtt);
+}
+
+void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits)
+{
+ struct realm *realm = &kvm->arch.realm;
+
+ ensure_spare_page(realm);
+
+ WARN_ON(realm_tear_down_rtt_range(realm, 0, (1UL << ia_bits)));
+}
+
/* Protects access to rme_vmid_bitmap */
static DEFINE_SPINLOCK(rme_vmid_lock);
static unsigned long *rme_vmid_bitmap;
--
2.34.1
The RMM maintains a data structure known as the Realm Execution Context
(or REC). It is similar to struct kvm_vcpu and tracks the state of the
virtual CPUs. KVM must delegate memory and request the structures are
created when vCPUs are created, and suitably tear down on destruction.
See Realm Management Monitor specification (DEN0137) for more information:
https://developer.arm.com/documentation/den0137/
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_emulate.h | 2 +
arch/arm64/include/asm/kvm_host.h | 3 +
arch/arm64/include/asm/kvm_rme.h | 18 ++++
arch/arm64/kvm/arm.c | 2 +
arch/arm64/kvm/reset.c | 11 ++
arch/arm64/kvm/rme.c | 150 +++++++++++++++++++++++++++
6 files changed, 186 insertions(+)
diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
index c606316f4729..2209a7c6267f 100644
--- a/arch/arm64/include/asm/kvm_emulate.h
+++ b/arch/arm64/include/asm/kvm_emulate.h
@@ -631,6 +631,8 @@ static inline bool kvm_realm_is_created(struct kvm *kvm)
static inline bool vcpu_is_rec(struct kvm_vcpu *vcpu)
{
+ if (static_branch_unlikely(&kvm_rme_is_available))
+ return vcpu->arch.rec.mpidr != INVALID_HWID;
return false;
}
diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index 63b68b85db3f..f7ac40ce0caf 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -694,6 +694,9 @@ struct kvm_vcpu_arch {
/* Per-vcpu CCSIDR override or NULL */
u32 *ccsidr;
+
+ /* Realm meta data */
+ struct realm_rec rec;
};
/*
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index 4ab5cb5e91b3..915e76068b00 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -6,6 +6,7 @@
#ifndef __ASM_KVM_RME_H
#define __ASM_KVM_RME_H
+#include <asm/rmi_smc.h>
#include <uapi/linux/kvm.h>
/**
@@ -70,6 +71,21 @@ struct realm {
unsigned int ia_bits;
};
+/**
+ * struct realm_rec - Additional per VCPU data for a Realm
+ *
+ * @mpidr: MPIDR (Multiprocessor Affinity Register) value to identify this VCPU
+ * @rec_page: Kernel VA of the RMM's private page for this REC
+ * @aux_pages: Additional pages private to the RMM for this REC
+ * @run: Kernel VA of the RmiRecRun structure shared with the RMM
+ */
+struct realm_rec {
+ unsigned long mpidr;
+ void *rec_page;
+ struct page *aux_pages[REC_PARAMS_AUX_GRANULES];
+ struct rec_run *run;
+};
+
int kvm_init_rme(void);
u32 kvm_realm_ipa_limit(void);
@@ -77,6 +93,8 @@ int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap);
int kvm_init_realm_vm(struct kvm *kvm);
void kvm_destroy_realm(struct kvm *kvm);
void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits);
+int kvm_create_rec(struct kvm_vcpu *vcpu);
+void kvm_destroy_rec(struct kvm_vcpu *vcpu);
#define RME_RTT_BLOCK_LEVEL 2
#define RME_RTT_MAX_LEVEL 3
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index c5a6139d5454..d70c511e16a0 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -432,6 +432,8 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
/* Force users to call KVM_ARM_VCPU_INIT */
vcpu_clear_flag(vcpu, VCPU_INITIALIZED);
+ vcpu->arch.rec.mpidr = INVALID_HWID;
+
vcpu->arch.mmu_page_cache.gfp_zero = __GFP_ZERO;
/*
diff --git a/arch/arm64/kvm/reset.c b/arch/arm64/kvm/reset.c
index 68d1d05672bd..6e6eb4a15095 100644
--- a/arch/arm64/kvm/reset.c
+++ b/arch/arm64/kvm/reset.c
@@ -134,6 +134,11 @@ int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature)
return -EPERM;
return kvm_vcpu_finalize_sve(vcpu);
+ case KVM_ARM_VCPU_REC:
+ if (!kvm_is_realm(vcpu->kvm))
+ return -EINVAL;
+
+ return kvm_create_rec(vcpu);
}
return -EINVAL;
@@ -144,6 +149,11 @@ bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu)
if (vcpu_has_sve(vcpu) && !kvm_arm_vcpu_sve_finalized(vcpu))
return false;
+ if (kvm_is_realm(vcpu->kvm) &&
+ !(vcpu_is_rec(vcpu) &&
+ READ_ONCE(vcpu->kvm->arch.realm.state) == REALM_STATE_ACTIVE))
+ return false;
+
return true;
}
@@ -157,6 +167,7 @@ void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu)
kvm_unshare_hyp(sve_state, sve_state + vcpu_sve_state_size(vcpu));
kfree(sve_state);
kfree(vcpu->arch.ccsidr);
+ kvm_destroy_rec(vcpu);
}
static void kvm_vcpu_reset_sve(struct kvm_vcpu *vcpu)
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 09b59bcad8b6..629a095bea61 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -474,6 +474,156 @@ void kvm_destroy_realm(struct kvm *kvm)
kvm_free_stage2_pgd(&kvm->arch.mmu);
}
+static void free_rec_aux(struct page **aux_pages,
+ unsigned int num_aux)
+{
+ unsigned int i;
+
+ for (i = 0; i < num_aux; i++) {
+ phys_addr_t aux_page_phys = page_to_phys(aux_pages[i]);
+
+ /* If the undelegate fails then leak the page */
+ if (WARN_ON(rmi_granule_undelegate(aux_page_phys)))
+ continue;
+
+ __free_page(aux_pages[i]);
+ }
+}
+
+static int alloc_rec_aux(struct page **aux_pages,
+ u64 *aux_phys_pages,
+ unsigned int num_aux)
+{
+ int ret;
+ unsigned int i;
+
+ for (i = 0; i < num_aux; i++) {
+ struct page *aux_page;
+ phys_addr_t aux_page_phys;
+
+ aux_page = alloc_page(GFP_KERNEL);
+ if (!aux_page) {
+ ret = -ENOMEM;
+ goto out_err;
+ }
+ aux_page_phys = page_to_phys(aux_page);
+ if (rmi_granule_delegate(aux_page_phys)) {
+ __free_page(aux_page);
+ ret = -ENXIO;
+ goto out_err;
+ }
+ aux_pages[i] = aux_page;
+ aux_phys_pages[i] = aux_page_phys;
+ }
+
+ return 0;
+out_err:
+ free_rec_aux(aux_pages, i);
+ return ret;
+}
+
+int kvm_create_rec(struct kvm_vcpu *vcpu)
+{
+ struct user_pt_regs *vcpu_regs = vcpu_gp_regs(vcpu);
+ unsigned long mpidr = kvm_vcpu_get_mpidr_aff(vcpu);
+ struct realm *realm = &vcpu->kvm->arch.realm;
+ struct realm_rec *rec = &vcpu->arch.rec;
+ unsigned long rec_page_phys;
+ struct rec_params *params;
+ int r, i;
+
+ if (kvm_realm_state(vcpu->kvm) != REALM_STATE_NEW)
+ return -ENOENT;
+
+ /*
+ * The RMM will report PSCI v1.0 to Realms and the KVM_ARM_VCPU_PSCI_0_2
+ * flag covers v0.2 and onwards.
+ */
+ if (!vcpu_has_feature(vcpu, KVM_ARM_VCPU_PSCI_0_2))
+ return -EINVAL;
+
+ BUILD_BUG_ON(sizeof(*params) > PAGE_SIZE);
+ BUILD_BUG_ON(sizeof(*rec->run) > PAGE_SIZE);
+
+ params = (struct rec_params *)get_zeroed_page(GFP_KERNEL);
+ rec->rec_page = (void *)__get_free_page(GFP_KERNEL);
+ rec->run = (void *)get_zeroed_page(GFP_KERNEL);
+ if (!params || !rec->rec_page || !rec->run) {
+ r = -ENOMEM;
+ goto out_free_pages;
+ }
+
+ for (i = 0; i < ARRAY_SIZE(params->gprs); i++)
+ params->gprs[i] = vcpu_regs->regs[i];
+
+ params->pc = vcpu_regs->pc;
+
+ if (vcpu->vcpu_id == 0)
+ params->flags |= REC_PARAMS_FLAG_RUNNABLE;
+
+ rec_page_phys = virt_to_phys(rec->rec_page);
+
+ if (rmi_granule_delegate(rec_page_phys)) {
+ r = -ENXIO;
+ goto out_free_pages;
+ }
+
+ r = alloc_rec_aux(rec->aux_pages, params->aux, realm->num_aux);
+ if (r)
+ goto out_undelegate_rmm_rec;
+
+ params->num_rec_aux = realm->num_aux;
+ params->mpidr = mpidr;
+
+ if (rmi_rec_create(virt_to_phys(realm->rd),
+ rec_page_phys,
+ virt_to_phys(params))) {
+ r = -ENXIO;
+ goto out_free_rec_aux;
+ }
+
+ rec->mpidr = mpidr;
+
+ free_page((unsigned long)params);
+ return 0;
+
+out_free_rec_aux:
+ free_rec_aux(rec->aux_pages, realm->num_aux);
+out_undelegate_rmm_rec:
+ if (WARN_ON(rmi_granule_undelegate(rec_page_phys)))
+ rec->rec_page = NULL;
+out_free_pages:
+ free_page((unsigned long)rec->run);
+ free_page((unsigned long)rec->rec_page);
+ free_page((unsigned long)params);
+ return r;
+}
+
+void kvm_destroy_rec(struct kvm_vcpu *vcpu)
+{
+ struct realm *realm = &vcpu->kvm->arch.realm;
+ struct realm_rec *rec = &vcpu->arch.rec;
+ unsigned long rec_page_phys;
+
+ if (!vcpu_is_rec(vcpu))
+ return;
+
+ rec_page_phys = virt_to_phys(rec->rec_page);
+
+ /* If the REC destroy fails, leak all pages relating to the REC */
+ if (WARN_ON(rmi_rec_destroy(rec_page_phys)))
+ return;
+
+ free_rec_aux(rec->aux_pages, realm->num_aux);
+
+ /* If the undelegate fails then leak the REC page */
+ if (WARN_ON(rmi_granule_undelegate(rec_page_phys)))
+ return;
+
+ free_page((unsigned long)rec->rec_page);
+ free_page((unsigned long)rec->run);
+}
+
int kvm_init_realm_vm(struct kvm *kvm)
{
struct realm_params *params;
--
2.34.1
The RMM keeps track of the timer while the realm REC is running, but on
exit to the normal world KVM is responsible for handling the timers.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/arch_timer.c | 45 ++++++++++++++++++++++++++++++++----
include/kvm/arm_arch_timer.h | 2 ++
2 files changed, 43 insertions(+), 4 deletions(-)
diff --git a/arch/arm64/kvm/arch_timer.c b/arch/arm64/kvm/arch_timer.c
index 879982b1cc73..0b2be34a9ba3 100644
--- a/arch/arm64/kvm/arch_timer.c
+++ b/arch/arm64/kvm/arch_timer.c
@@ -162,6 +162,13 @@ static void timer_set_cval(struct arch_timer_context *ctxt, u64 cval)
static void timer_set_offset(struct arch_timer_context *ctxt, u64 offset)
{
+ struct kvm_vcpu *vcpu = ctxt->vcpu;
+
+ if (kvm_is_realm(vcpu->kvm)) {
+ WARN_ON(offset);
+ return;
+ }
+
if (!ctxt->offset.vm_offset) {
WARN(offset, "timer %ld\n", arch_timer_ctx_index(ctxt));
return;
@@ -460,6 +467,21 @@ static void kvm_timer_update_irq(struct kvm_vcpu *vcpu, bool new_level,
}
}
+void kvm_realm_timers_update(struct kvm_vcpu *vcpu)
+{
+ struct arch_timer_cpu *arch_timer = &vcpu->arch.timer_cpu;
+ int i;
+
+ for (i = 0; i < NR_KVM_EL0_TIMERS; i++) {
+ struct arch_timer_context *timer = &arch_timer->timers[i];
+ bool status = timer_get_ctl(timer) & ARCH_TIMER_CTRL_IT_STAT;
+ bool level = kvm_timer_irq_can_fire(timer) && status;
+
+ if (level != timer->irq.level)
+ kvm_timer_update_irq(vcpu, level, timer);
+ }
+}
+
/* Only called for a fully emulated timer */
static void timer_emulate(struct arch_timer_context *ctx)
{
@@ -831,6 +853,8 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu)
if (unlikely(!timer->enabled))
return;
+ kvm_timer_unblocking(vcpu);
+
get_timer_map(vcpu, &map);
if (static_branch_likely(&has_gic_active_state)) {
@@ -844,8 +868,6 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu)
kvm_timer_vcpu_load_nogic(vcpu);
}
- kvm_timer_unblocking(vcpu);
-
timer_restore_state(map.direct_vtimer);
if (map.direct_ptimer)
timer_restore_state(map.direct_ptimer);
@@ -988,7 +1010,9 @@ static void timer_context_init(struct kvm_vcpu *vcpu, int timerid)
ctxt->vcpu = vcpu;
- if (timerid == TIMER_VTIMER)
+ if (kvm_is_realm(vcpu->kvm))
+ ctxt->offset.vm_offset = NULL;
+ else if (timerid == TIMER_VTIMER)
ctxt->offset.vm_offset = &kvm->arch.timer_data.voffset;
else
ctxt->offset.vm_offset = &kvm->arch.timer_data.poffset;
@@ -1011,13 +1035,19 @@ static void timer_context_init(struct kvm_vcpu *vcpu, int timerid)
void kvm_timer_vcpu_init(struct kvm_vcpu *vcpu)
{
struct arch_timer_cpu *timer = vcpu_timer(vcpu);
+ u64 cntvoff;
for (int i = 0; i < NR_KVM_TIMERS; i++)
timer_context_init(vcpu, i);
+ if (kvm_is_realm(vcpu->kvm))
+ cntvoff = 0;
+ else
+ cntvoff = kvm_phys_timer_read();
+
/* Synchronize offsets across timers of a VM if not already provided */
if (!test_bit(KVM_ARCH_FLAG_VM_COUNTER_OFFSET, &vcpu->kvm->arch.flags)) {
- timer_set_offset(vcpu_vtimer(vcpu), kvm_phys_timer_read());
+ timer_set_offset(vcpu_vtimer(vcpu), cntvoff);
timer_set_offset(vcpu_ptimer(vcpu), 0);
}
@@ -1525,6 +1555,13 @@ int kvm_timer_enable(struct kvm_vcpu *vcpu)
return -EINVAL;
}
+ /*
+ * We don't use mapped IRQs for Realms because the RMI doesn't allow
+ * us setting the LR.HW bit in the VGIC.
+ */
+ if (vcpu_is_rec(vcpu))
+ return 0;
+
get_timer_map(vcpu, &map);
ret = kvm_vgic_map_phys_irq(vcpu,
diff --git a/include/kvm/arm_arch_timer.h b/include/kvm/arm_arch_timer.h
index c819c5d16613..d8ab297560d0 100644
--- a/include/kvm/arm_arch_timer.h
+++ b/include/kvm/arm_arch_timer.h
@@ -112,6 +112,8 @@ int kvm_arm_timer_set_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr);
int kvm_arm_timer_get_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr);
int kvm_arm_timer_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr);
+void kvm_realm_timers_update(struct kvm_vcpu *vcpu);
+
u64 kvm_phys_timer_read(void);
void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu);
--
2.34.1
Entering a realm is done using a SMC call to the RMM. On exit the
exit-codes need to be handled slightly differently to the normal KVM
path so define our own functions for realm enter/exit and hook them
in if the guest is a realm guest.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_rme.h | 3 +
arch/arm64/kvm/Makefile | 2 +-
arch/arm64/kvm/arm.c | 19 +++-
arch/arm64/kvm/rme-exit.c | 180 +++++++++++++++++++++++++++++++
arch/arm64/kvm/rme.c | 11 ++
5 files changed, 209 insertions(+), 6 deletions(-)
create mode 100644 arch/arm64/kvm/rme-exit.c
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index cc8f81cfc3c0..749f2eb97bd4 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -96,6 +96,9 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits);
int kvm_create_rec(struct kvm_vcpu *vcpu);
void kvm_destroy_rec(struct kvm_vcpu *vcpu);
+int kvm_rec_enter(struct kvm_vcpu *vcpu);
+int handle_rme_exit(struct kvm_vcpu *vcpu, int rec_run_status);
+
void kvm_realm_unmap_range(struct kvm *kvm,
unsigned long ipa,
u64 size,
diff --git a/arch/arm64/kvm/Makefile b/arch/arm64/kvm/Makefile
index 1c1d8cdf381f..2eec980cbe5c 100644
--- a/arch/arm64/kvm/Makefile
+++ b/arch/arm64/kvm/Makefile
@@ -21,7 +21,7 @@ kvm-y += arm.o mmu.o mmio.o psci.o hypercalls.o pvtime.o \
vgic/vgic-mmio.o vgic/vgic-mmio-v2.o \
vgic/vgic-mmio-v3.o vgic/vgic-kvm-device.o \
vgic/vgic-its.o vgic/vgic-debug.o \
- rme.o
+ rme.o rme-exit.o
kvm-$(CONFIG_HW_PERF_EVENTS) += pmu-emul.o pmu.o
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index f8daf478114f..e92afb3835f6 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -1130,7 +1130,10 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
trace_kvm_entry(*vcpu_pc(vcpu));
guest_timing_enter_irqoff();
- ret = kvm_arm_vcpu_enter_exit(vcpu);
+ if (vcpu_is_rec(vcpu))
+ ret = kvm_rec_enter(vcpu);
+ else
+ ret = kvm_arm_vcpu_enter_exit(vcpu);
vcpu->mode = OUTSIDE_GUEST_MODE;
vcpu->stat.exits++;
@@ -1184,10 +1187,13 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
local_irq_enable();
- trace_kvm_exit(ret, kvm_vcpu_trap_get_class(vcpu), *vcpu_pc(vcpu));
-
/* Exit types that need handling before we can be preempted */
- handle_exit_early(vcpu, ret);
+ if (!vcpu_is_rec(vcpu)) {
+ trace_kvm_exit(ret, kvm_vcpu_trap_get_class(vcpu),
+ *vcpu_pc(vcpu));
+
+ handle_exit_early(vcpu, ret);
+ }
preempt_enable();
@@ -1210,7 +1216,10 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
ret = ARM_EXCEPTION_IL;
}
- ret = handle_exit(vcpu, ret);
+ if (vcpu_is_rec(vcpu))
+ ret = handle_rme_exit(vcpu, ret);
+ else
+ ret = handle_exit(vcpu, ret);
}
/* Tell userspace about in-kernel device output levels */
diff --git a/arch/arm64/kvm/rme-exit.c b/arch/arm64/kvm/rme-exit.c
new file mode 100644
index 000000000000..5bf58c9b42b7
--- /dev/null
+++ b/arch/arm64/kvm/rme-exit.c
@@ -0,0 +1,180 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2023 ARM Ltd.
+ */
+
+#include <linux/kvm_host.h>
+#include <kvm/arm_hypercalls.h>
+#include <kvm/arm_psci.h>
+
+#include <asm/rmi_smc.h>
+#include <asm/kvm_emulate.h>
+#include <asm/kvm_rme.h>
+#include <asm/kvm_mmu.h>
+
+typedef int (*exit_handler_fn)(struct kvm_vcpu *vcpu);
+
+static int rec_exit_reason_notimpl(struct kvm_vcpu *vcpu)
+{
+ struct realm_rec *rec = &vcpu->arch.rec;
+
+ pr_err("[vcpu %d] Unhandled exit reason from realm (ESR: %#llx)\n",
+ vcpu->vcpu_id, rec->run->exit.esr);
+ return -ENXIO;
+}
+
+static int rec_exit_sync_dabt(struct kvm_vcpu *vcpu)
+{
+ return kvm_handle_guest_abort(vcpu);
+}
+
+static int rec_exit_sync_iabt(struct kvm_vcpu *vcpu)
+{
+ struct realm_rec *rec = &vcpu->arch.rec;
+
+ pr_err("[vcpu %d] Unhandled instruction abort (ESR: %#llx).\n",
+ vcpu->vcpu_id, rec->run->exit.esr);
+ return -ENXIO;
+}
+
+static int rec_exit_sys_reg(struct kvm_vcpu *vcpu)
+{
+ struct realm_rec *rec = &vcpu->arch.rec;
+ unsigned long esr = kvm_vcpu_get_esr(vcpu);
+ int rt = kvm_vcpu_sys_get_rt(vcpu);
+ bool is_write = !(esr & 1);
+ int ret;
+
+ if (is_write)
+ vcpu_set_reg(vcpu, rt, rec->run->exit.gprs[0]);
+
+ ret = kvm_handle_sys_reg(vcpu);
+
+ if (ret >= 0 && !is_write)
+ rec->run->entry.gprs[0] = vcpu_get_reg(vcpu, rt);
+
+ return ret;
+}
+
+static exit_handler_fn rec_exit_handlers[] = {
+ [0 ... ESR_ELx_EC_MAX] = rec_exit_reason_notimpl,
+ [ESR_ELx_EC_SYS64] = rec_exit_sys_reg,
+ [ESR_ELx_EC_DABT_LOW] = rec_exit_sync_dabt,
+ [ESR_ELx_EC_IABT_LOW] = rec_exit_sync_iabt
+};
+
+static int rec_exit_psci(struct kvm_vcpu *vcpu)
+{
+ struct realm_rec *rec = &vcpu->arch.rec;
+ int i;
+ int ret;
+
+ for (i = 0; i < REC_RUN_GPRS; i++)
+ vcpu_set_reg(vcpu, i, rec->run->exit.gprs[i]);
+
+ ret = kvm_smccc_call_handler(vcpu);
+
+ for (i = 0; i < REC_RUN_GPRS; i++)
+ rec->run->entry.gprs[i] = vcpu_get_reg(vcpu, i);
+
+ return ret;
+}
+
+static int rec_exit_ripas_change(struct kvm_vcpu *vcpu)
+{
+ struct kvm *kvm = vcpu->kvm;
+ struct realm *realm = &kvm->arch.realm;
+ struct realm_rec *rec = &vcpu->arch.rec;
+ unsigned long base = rec->run->exit.ripas_base;
+ unsigned long top = rec->run->exit.ripas_top;
+ unsigned long ripas = rec->run->exit.ripas_value & 1;
+ int ret = -EINVAL;
+
+ if (realm_is_addr_protected(realm, base) &&
+ realm_is_addr_protected(realm, top - 1)) {
+ kvm_mmu_topup_memory_cache(&vcpu->arch.mmu_page_cache,
+ kvm_mmu_cache_min_pages(vcpu->arch.hw_mmu));
+ write_lock(&kvm->mmu_lock);
+ ret = realm_set_ipa_state(vcpu, base, top, ripas);
+ write_unlock(&kvm->mmu_lock);
+ }
+
+ WARN(ret && ret != -ENOMEM,
+ "Unable to satisfy SET_IPAS for %#lx - %#lx, ripas: %#lx\n",
+ base, top, ripas);
+
+ /* Exit to VMM to complete the change */
+ kvm_prepare_memory_fault_exit(vcpu, base, top - base, false, false,
+ ripas == 1);
+
+ return 0;
+}
+
+static void update_arch_timer_irq_lines(struct kvm_vcpu *vcpu)
+{
+ struct realm_rec *rec = &vcpu->arch.rec;
+
+ __vcpu_sys_reg(vcpu, CNTV_CTL_EL0) = rec->run->exit.cntv_ctl;
+ __vcpu_sys_reg(vcpu, CNTV_CVAL_EL0) = rec->run->exit.cntv_cval;
+ __vcpu_sys_reg(vcpu, CNTP_CTL_EL0) = rec->run->exit.cntp_ctl;
+ __vcpu_sys_reg(vcpu, CNTP_CVAL_EL0) = rec->run->exit.cntp_cval;
+
+ kvm_realm_timers_update(vcpu);
+}
+
+/*
+ * Return > 0 to return to guest, < 0 on error, 0 (and set exit_reason) on
+ * proper exit to userspace.
+ */
+int handle_rme_exit(struct kvm_vcpu *vcpu, int rec_run_ret)
+{
+ struct realm_rec *rec = &vcpu->arch.rec;
+ u8 esr_ec = ESR_ELx_EC(rec->run->exit.esr);
+ unsigned long status, index;
+
+ status = RMI_RETURN_STATUS(rec_run_ret);
+ index = RMI_RETURN_INDEX(rec_run_ret);
+
+ /*
+ * If a PSCI_SYSTEM_OFF request raced with a vcpu executing, we might
+ * see the following status code and index indicating an attempt to run
+ * a REC when the RD state is SYSTEM_OFF. In this case, we just need to
+ * return to user space which can deal with the system event or will try
+ * to run the KVM VCPU again, at which point we will no longer attempt
+ * to enter the Realm because we will have a sleep request pending on
+ * the VCPU as a result of KVM's PSCI handling.
+ */
+ if (status == RMI_ERROR_REALM && index == 1) {
+ vcpu->run->exit_reason = KVM_EXIT_UNKNOWN;
+ return 0;
+ }
+
+ if (rec_run_ret)
+ return -ENXIO;
+
+ vcpu->arch.fault.esr_el2 = rec->run->exit.esr;
+ vcpu->arch.fault.far_el2 = rec->run->exit.far;
+ vcpu->arch.fault.hpfar_el2 = rec->run->exit.hpfar;
+
+ update_arch_timer_irq_lines(vcpu);
+
+ /* Reset the emulation flags for the next run of the REC */
+ rec->run->entry.flags = 0;
+
+ switch (rec->run->exit.exit_reason) {
+ case RMI_EXIT_SYNC:
+ return rec_exit_handlers[esr_ec](vcpu);
+ case RMI_EXIT_IRQ:
+ case RMI_EXIT_FIQ:
+ return 1;
+ case RMI_EXIT_PSCI:
+ return rec_exit_psci(vcpu);
+ case RMI_EXIT_RIPAS_CHANGE:
+ return rec_exit_ripas_change(vcpu);
+ }
+
+ kvm_pr_unimpl("Unsupported exit reason: %u\n",
+ rec->run->exit.exit_reason);
+ vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+ return 0;
+}
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 9e5983c51393..0a3f823b2446 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -864,6 +864,17 @@ void kvm_destroy_realm(struct kvm *kvm)
kvm_free_stage2_pgd(&kvm->arch.mmu);
}
+int kvm_rec_enter(struct kvm_vcpu *vcpu)
+{
+ struct realm_rec *rec = &vcpu->arch.rec;
+
+ if (kvm_realm_state(vcpu->kvm) != REALM_STATE_ACTIVE)
+ return -EINVAL;
+
+ return rmi_rec_enter(virt_to_phys(rec->rec_page),
+ virt_to_phys(rec->run));
+}
+
static void free_rec_aux(struct page **aux_pages,
unsigned int num_aux)
{
--
2.34.1
MMIO emulation for a realm cannot be done directly with the VM's
registers as they are protected from the host. However, for emulatable
data aborts, the RMM uses GPRS[0] to provide the read/written value.
We can transfer this from/to the equivalent VCPU's register entry and
then depend on the generic MMIO handling code in KVM.
For a MMIO read, the value is placed in the shared RecExit structure
during kvm_handle_mmio_return() rather than in the VCPU's register
entry.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/mmio.c | 10 +++++++++-
arch/arm64/kvm/rme-exit.c | 6 ++++++
2 files changed, 15 insertions(+), 1 deletion(-)
diff --git a/arch/arm64/kvm/mmio.c b/arch/arm64/kvm/mmio.c
index 200c8019a82a..589a1a7601ad 100644
--- a/arch/arm64/kvm/mmio.c
+++ b/arch/arm64/kvm/mmio.c
@@ -6,6 +6,7 @@
#include <linux/kvm_host.h>
#include <asm/kvm_emulate.h>
+#include <asm/rmi_smc.h>
#include <trace/events/kvm.h>
#include "trace.h"
@@ -90,6 +91,9 @@ int kvm_handle_mmio_return(struct kvm_vcpu *vcpu)
vcpu->mmio_needed = 0;
+ if (vcpu_is_rec(vcpu))
+ vcpu->arch.rec.run->entry.flags |= RMI_EMULATED_MMIO;
+
if (!kvm_vcpu_dabt_iswrite(vcpu)) {
struct kvm_run *run = vcpu->run;
@@ -108,7 +112,11 @@ int kvm_handle_mmio_return(struct kvm_vcpu *vcpu)
trace_kvm_mmio(KVM_TRACE_MMIO_READ, len, run->mmio.phys_addr,
&data);
data = vcpu_data_host_to_guest(vcpu, data, len);
- vcpu_set_reg(vcpu, kvm_vcpu_dabt_get_rd(vcpu), data);
+
+ if (vcpu_is_rec(vcpu))
+ vcpu->arch.rec.run->entry.gprs[0] = data;
+ else
+ vcpu_set_reg(vcpu, kvm_vcpu_dabt_get_rd(vcpu), data);
}
/*
diff --git a/arch/arm64/kvm/rme-exit.c b/arch/arm64/kvm/rme-exit.c
index 5bf58c9b42b7..2ddaec09cc98 100644
--- a/arch/arm64/kvm/rme-exit.c
+++ b/arch/arm64/kvm/rme-exit.c
@@ -25,6 +25,12 @@ static int rec_exit_reason_notimpl(struct kvm_vcpu *vcpu)
static int rec_exit_sync_dabt(struct kvm_vcpu *vcpu)
{
+ struct realm_rec *rec = &vcpu->arch.rec;
+
+ if (kvm_vcpu_dabt_iswrite(vcpu) && kvm_vcpu_dabt_isvalid(vcpu))
+ vcpu_set_reg(vcpu, kvm_vcpu_dabt_get_rd(vcpu),
+ rec->run->exit.gprs[0]);
+
return kvm_handle_guest_abort(vcpu);
}
--
2.34.1
The RMM needs to be informed of the target REC when a PSCI call is made
with an MPIDR argument. Expose an ioctl to the userspace in case the PSCI
is handled by it.
Co-developed-by: Suzuki K Poulose <[email protected]>
Signed-off-by: Suzuki K Poulose <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_rme.h | 3 +++
arch/arm64/kvm/arm.c | 25 +++++++++++++++++++++++++
arch/arm64/kvm/psci.c | 29 +++++++++++++++++++++++++++++
arch/arm64/kvm/rme.c | 15 +++++++++++++++
4 files changed, 72 insertions(+)
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index 48c7766fadeb..fef04706ff9d 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -116,6 +116,9 @@ int realm_map_non_secure(struct realm *realm,
int realm_set_ipa_state(struct kvm_vcpu *vcpu,
unsigned long addr, unsigned long end,
unsigned long ripas);
+int realm_psci_complete(struct kvm_vcpu *calling,
+ struct kvm_vcpu *target,
+ unsigned long status);
#define RME_RTT_BLOCK_LEVEL 2
#define RME_RTT_MAX_LEVEL 3
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index 4fd58a1d3351..a0dcae0391d0 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -1605,6 +1605,22 @@ static int kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
return __kvm_arm_vcpu_set_events(vcpu, events);
}
+static int kvm_arm_vcpu_rmm_psci_complete(struct kvm_vcpu *vcpu,
+ struct kvm_arm_rmm_psci_complete *arg)
+{
+ struct kvm_vcpu *target = kvm_mpidr_to_vcpu(vcpu->kvm, arg->target_mpidr);
+
+ if (!target)
+ return -EINVAL;
+
+ /*
+ * RMM v1.0 only supports PSCI_RET_SUCCESS or PSCI_RET_DENIED
+ * for the status. But, let us leave it to the RMM to filter
+ * for making this future proof.
+ */
+ return realm_psci_complete(vcpu, target, arg->psci_status);
+}
+
long kvm_arch_vcpu_ioctl(struct file *filp,
unsigned int ioctl, unsigned long arg)
{
@@ -1727,6 +1743,15 @@ long kvm_arch_vcpu_ioctl(struct file *filp,
return kvm_arm_vcpu_finalize(vcpu, what);
}
+ case KVM_ARM_VCPU_RMM_PSCI_COMPLETE: {
+ struct kvm_arm_rmm_psci_complete req;
+
+ if (!kvm_is_realm(vcpu->kvm))
+ return -EINVAL;
+ if (copy_from_user(&req, argp, sizeof(req)))
+ return -EFAULT;
+ return kvm_arm_vcpu_rmm_psci_complete(vcpu, &req);
+ }
default:
r = -EINVAL;
}
diff --git a/arch/arm64/kvm/psci.c b/arch/arm64/kvm/psci.c
index 1f69b667332b..f9abab5d50d7 100644
--- a/arch/arm64/kvm/psci.c
+++ b/arch/arm64/kvm/psci.c
@@ -103,6 +103,12 @@ static unsigned long kvm_psci_vcpu_on(struct kvm_vcpu *source_vcpu)
reset_state->reset = true;
kvm_make_request(KVM_REQ_VCPU_RESET, vcpu);
+ /*
+ * Make sure we issue PSCI_COMPLETE before the VCPU can be
+ * scheduled.
+ */
+ if (vcpu_is_rec(vcpu))
+ realm_psci_complete(source_vcpu, vcpu, PSCI_RET_SUCCESS);
/*
* Make sure the reset request is observed if the RUNNABLE mp_state is
@@ -115,6 +121,10 @@ static unsigned long kvm_psci_vcpu_on(struct kvm_vcpu *source_vcpu)
out_unlock:
spin_unlock(&vcpu->arch.mp_state_lock);
+ if (vcpu_is_rec(vcpu) && ret != PSCI_RET_SUCCESS)
+ realm_psci_complete(source_vcpu, vcpu,
+ ret == PSCI_RET_ALREADY_ON ?
+ PSCI_RET_SUCCESS : PSCI_RET_DENIED);
return ret;
}
@@ -142,6 +152,25 @@ static unsigned long kvm_psci_vcpu_affinity_info(struct kvm_vcpu *vcpu)
/* Ignore other bits of target affinity */
target_affinity &= target_affinity_mask;
+ if (vcpu_is_rec(vcpu)) {
+ struct kvm_vcpu *target_vcpu;
+
+ /* RMM supports only zero affinity level */
+ if (lowest_affinity_level != 0)
+ return PSCI_RET_INVALID_PARAMS;
+
+ target_vcpu = kvm_mpidr_to_vcpu(kvm, target_affinity);
+ if (!target_vcpu)
+ return PSCI_RET_INVALID_PARAMS;
+
+ /*
+ * Provide the references of running and target RECs to the RMM
+ * so that the RMM can complete the PSCI request.
+ */
+ realm_psci_complete(vcpu, target_vcpu, PSCI_RET_SUCCESS);
+ return PSCI_RET_SUCCESS;
+ }
+
/*
* If one or more VCPU matching target affinity are running
* then ON else OFF
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 72f6f5f542c4..9544653accb6 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -101,6 +101,21 @@ static void free_delegated_page(struct realm *realm, phys_addr_t phys)
free_page((unsigned long)phys_to_virt(phys));
}
+int realm_psci_complete(struct kvm_vcpu *calling, struct kvm_vcpu *target,
+ unsigned long status)
+{
+ int ret;
+
+ ret = rmi_psci_complete(virt_to_phys(calling->arch.rec.rec_page),
+ virt_to_phys(target->arch.rec.rec_page),
+ status);
+
+ if (ret)
+ return -EINVAL;
+
+ return 0;
+}
+
static int realm_rtt_create(struct realm *realm,
unsigned long addr,
int level,
--
2.34.1
Always split up huge pages to avoid problems managing huge pages. There
are two issues currently:
1. The uABI for the VMM allows populating memory on 4k boundaries even
if the underlying allocator (e.g. hugetlbfs) is using a larger page
size. Using a memfd for private allocations will push this issue onto
the VMM as it will need to respect the granularity of the allocator.
2. The guest is able to request arbitrary ranges to be remapped as
shared. Again with a memfd approach it will be up to the VMM to deal
with the complexity and either overmap (need the huge mapping and add
an additional 'overlapping' shared mapping) or reject the request as
invalid due to the use of a huge page allocator.
For now just break everything down to 4k pages in the RMM controlled
stage 2.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/mmu.c | 4 ++++
arch/arm64/kvm/rme.c | 4 +++-
2 files changed, 7 insertions(+), 1 deletion(-)
diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index 50a49e4e2020..d891fc3c1a10 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -1532,6 +1532,10 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
if (logging_active) {
force_pte = true;
vma_shift = PAGE_SHIFT;
+ } else if (kvm_is_realm(kvm)) {
+ // Force PTE level mappings for realms
+ force_pte = true;
+ vma_shift = PAGE_SHIFT;
} else {
vma_shift = get_vma_page_shift(vma, hva);
}
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 9593e8e35913..ae9fd12c4e7d 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -847,7 +847,9 @@ static int populate_par_region(struct kvm *kvm,
break;
}
- if (is_vm_hugetlb_page(vma))
+ // FIXME: To avoid the overmapping issue (see below comment)
+ // force the use of 4k pages
+ if (is_vm_hugetlb_page(vma) && 0)
vma_shift = huge_page_shift(hstate_vma(vma));
else
vma_shift = PAGE_SHIFT;
--
2.34.1
Physical device assignment is not yet supported by the RMM, so it
doesn't make much sense to allow device mappings within the realm.
Prevent them when the guest is a realm.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/mmu.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index d891fc3c1a10..48c957e21c83 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -1101,6 +1101,10 @@ int kvm_phys_addr_ioremap(struct kvm *kvm, phys_addr_t guest_ipa,
if (is_protected_kvm_enabled())
return -EPERM;
+ /* We don't support mapping special pages into a Realm */
+ if (kvm_is_realm(kvm))
+ return -EINVAL;
+
size += offset_in_page(guest_ipa);
guest_ipa &= PAGE_MASK;
--
2.34.1
Arm CCA assigns the physical PMU device to the guest running in realm
world, however the IRQs are routed via the host. To enter a realm guest
while a PMU IRQ is pending it is necessary to block the physical IRQ to
prevent an immediate exit. Provide a mechanism in the PMU driver for KVM
to control the physical IRQ.
Signed-off-by: Steven Price <[email protected]>
---
drivers/perf/arm_pmu.c | 15 +++++++++++++++
include/linux/perf/arm_pmu.h | 1 +
2 files changed, 16 insertions(+)
diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
index 8458fe2cebb4..1e024acc98bb 100644
--- a/drivers/perf/arm_pmu.c
+++ b/drivers/perf/arm_pmu.c
@@ -735,6 +735,21 @@ static int arm_perf_teardown_cpu(unsigned int cpu, struct hlist_node *node)
return 0;
}
+void arm_pmu_set_phys_irq(bool enable)
+{
+ int cpu = get_cpu();
+ struct arm_pmu *pmu = per_cpu(cpu_armpmu, cpu);
+ int irq;
+
+ irq = armpmu_get_cpu_irq(pmu, cpu);
+ if (irq && !enable)
+ per_cpu(cpu_irq_ops, cpu)->disable_pmuirq(irq);
+ else if (irq && enable)
+ per_cpu(cpu_irq_ops, cpu)->enable_pmuirq(irq);
+
+ put_cpu();
+}
+
#ifdef CONFIG_CPU_PM
static void cpu_pm_pmu_setup(struct arm_pmu *armpmu, unsigned long cmd)
{
diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
index b3b34f6670cf..fd5e2e63b7fb 100644
--- a/include/linux/perf/arm_pmu.h
+++ b/include/linux/perf/arm_pmu.h
@@ -168,6 +168,7 @@ void kvm_host_pmu_init(struct arm_pmu *pmu);
#endif
bool arm_pmu_irq_is_nmi(void);
+void arm_pmu_set_phys_irq(bool enable);
/* Internal functions only for core arm_pmu code */
struct arm_pmu *armpmu_alloc(void);
--
2.34.1
Use the PMU registers from the RmiRecExit structure to identify when an
overflow interrupt is due and inject it into the guest. Also hook up the
configuration option for enabling the PMU within the guest.
When entering a realm guest with a PMU interrupt pending, it is
necessary to disable the physical interrupt. Otherwise when the RMM
restores the PMU state the physical interrupt will trigger causing an
immediate exit back to the host. The guest is expected to acknowledge
the interrupt causing a host exit (to update the GIC state) which gives
the opportunity to re-enable the physical interrupt before the next PMU
event.
Number of PMU counters is configured by the VMM by writing to PMCR.N.
Signed-off-by: Steven Price <[email protected]>
Signed-off-by: Jean-Philippe Brucker <[email protected]>
---
arch/arm64/kvm/arm.c | 15 +++++++++++++++
arch/arm64/kvm/guest.c | 7 +++++++
arch/arm64/kvm/pmu-emul.c | 4 +++-
arch/arm64/kvm/rme.c | 8 ++++++++
arch/arm64/kvm/sys_regs.c | 2 +-
5 files changed, 34 insertions(+), 2 deletions(-)
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index dcd9089877f3..2aad83053b62 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -15,6 +15,7 @@
#include <linux/vmalloc.h>
#include <linux/fs.h>
#include <linux/mman.h>
+#include <linux/perf/arm_pmu.h>
#include <linux/sched.h>
#include <linux/kvm.h>
#include <linux/kvm_irqfd.h>
@@ -1075,6 +1076,8 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
run->exit_reason = KVM_EXIT_UNKNOWN;
run->flags = 0;
while (ret > 0) {
+ bool pmu_stopped = false;
+
/*
* Check conditions before entering the guest
*/
@@ -1106,6 +1109,15 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
kvm_pmu_flush_hwstate(vcpu);
+ if (vcpu_is_rec(vcpu)) {
+ struct kvm_pmu *pmu = &vcpu->arch.pmu;
+
+ if (pmu->irq_level) {
+ pmu_stopped = true;
+ arm_pmu_set_phys_irq(false);
+ }
+ }
+
local_irq_disable();
kvm_vgic_flush_hwstate(vcpu);
@@ -1208,6 +1220,9 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
preempt_enable();
+ if (pmu_stopped)
+ arm_pmu_set_phys_irq(true);
+
/*
* The ARMv8 architecture doesn't give the hypervisor
* a mechanism to prevent a guest from dropping to AArch32 EL0
diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c
index 5223a828a344..d35367cf527d 100644
--- a/arch/arm64/kvm/guest.c
+++ b/arch/arm64/kvm/guest.c
@@ -782,6 +782,8 @@ int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
return kvm_arm_sys_reg_get_reg(vcpu, reg);
}
+#define KVM_REG_ARM_PMCR_EL0 ARM64_SYS_REG(3, 3, 9, 12, 0)
+
/*
* The RMI ABI only enables setting the lower GPRs (x0-x7) and PC.
* All other registers are reset to architectural or otherwise defined reset
@@ -800,6 +802,11 @@ static bool validate_realm_set_reg(struct kvm_vcpu *vcpu,
case KVM_REG_ARM_CORE_REG(regs.pc):
return true;
}
+ } else {
+ switch (reg->id) {
+ case KVM_REG_ARM_PMCR_EL0:
+ return true;
+ }
}
return false;
diff --git a/arch/arm64/kvm/pmu-emul.c b/arch/arm64/kvm/pmu-emul.c
index a35ce10e0a9f..ce7c8e55d904 100644
--- a/arch/arm64/kvm/pmu-emul.c
+++ b/arch/arm64/kvm/pmu-emul.c
@@ -341,7 +341,9 @@ static u64 kvm_pmu_overflow_status(struct kvm_vcpu *vcpu)
{
u64 reg = 0;
- if ((kvm_vcpu_read_pmcr(vcpu) & ARMV8_PMU_PMCR_E)) {
+ if (vcpu_is_rec(vcpu)) {
+ reg = vcpu->arch.rec.run->exit.pmu_ovf_status;
+ } else if ((kvm_vcpu_read_pmcr(vcpu) & ARMV8_PMU_PMCR_E)) {
reg = __vcpu_sys_reg(vcpu, PMOVSSET_EL0);
reg &= __vcpu_sys_reg(vcpu, PMCNTENSET_EL0);
reg &= __vcpu_sys_reg(vcpu, PMINTENSET_EL1);
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index ae9fd12c4e7d..e60a1196a2fe 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -314,6 +314,11 @@ static int realm_create_rd(struct kvm *kvm)
params->rtt_base = kvm->arch.mmu.pgd_phys;
params->vmid = realm->vmid;
+ if (kvm->arch.arm_pmu) {
+ params->pmu_num_ctrs = kvm->arch.pmcr_n;
+ params->flags |= RMI_REALM_PARAM_FLAG_PMU;
+ }
+
params_phys = virt_to_phys(params);
if (rmi_realm_create(rd_phys, params_phys)) {
@@ -1366,6 +1371,9 @@ int kvm_create_rec(struct kvm_vcpu *vcpu)
if (!vcpu_has_feature(vcpu, KVM_ARM_VCPU_PSCI_0_2))
return -EINVAL;
+ if (vcpu->kvm->arch.arm_pmu && !kvm_vcpu_has_pmu(vcpu))
+ return -EINVAL;
+
BUILD_BUG_ON(sizeof(*params) > PAGE_SIZE);
BUILD_BUG_ON(sizeof(*rec->run) > PAGE_SIZE);
diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c
index c9f4f387155f..60452c6519a4 100644
--- a/arch/arm64/kvm/sys_regs.c
+++ b/arch/arm64/kvm/sys_regs.c
@@ -1279,7 +1279,7 @@ static int set_pmcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r,
* implements. Ignore this error to maintain compatibility
* with the existing KVM behavior.
*/
- if (!kvm_vm_has_ran_once(kvm) &&
+ if (!kvm_vm_has_ran_once(kvm) && !kvm_realm_is_created(kvm) &&
new_n <= kvm_arm_pmu_get_max_counters(kvm))
kvm->arch.pmcr_n = new_n;
--
2.34.1
From: Jean-Philippe Brucker <[email protected]>
Allow userspace to configure the number of breakpoints and watchpoints
of a Realm VM through KVM_SET_ONE_REG ID_AA64DFR0_EL1.
The KVM sys_reg handler checks the user value against the maximum value
given by RMM (arm64_check_features() gets it from the
read_sanitised_id_aa64dfr0_el1() reset handler).
Userspace discovers that it can write these fields by issuing a
KVM_ARM_GET_REG_WRITABLE_MASKS ioctl.
Signed-off-by: Jean-Philippe Brucker <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/guest.c | 2 ++
arch/arm64/kvm/rme.c | 3 +++
arch/arm64/kvm/sys_regs.c | 21 ++++++++++++++-------
3 files changed, 19 insertions(+), 7 deletions(-)
diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c
index d35367cf527d..f9a47ce71a26 100644
--- a/arch/arm64/kvm/guest.c
+++ b/arch/arm64/kvm/guest.c
@@ -783,6 +783,7 @@ int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
}
#define KVM_REG_ARM_PMCR_EL0 ARM64_SYS_REG(3, 3, 9, 12, 0)
+#define KVM_REG_ARM_ID_AA64DFR0_EL1 ARM64_SYS_REG(3, 0, 0, 5, 0)
/*
* The RMI ABI only enables setting the lower GPRs (x0-x7) and PC.
@@ -805,6 +806,7 @@ static bool validate_realm_set_reg(struct kvm_vcpu *vcpu,
} else {
switch (reg->id) {
case KVM_REG_ARM_PMCR_EL0:
+ case KVM_REG_ARM_ID_AA64DFR0_EL1:
return true;
}
}
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 4edc8d98e1e6..51ac8c3462ea 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -305,6 +305,7 @@ static int realm_create_rd(struct kvm *kvm)
void *rd = NULL;
phys_addr_t rd_phys, params_phys;
struct kvm_pgtable *pgt = kvm->arch.mmu.pgt;
+ u64 dfr0 = IDREG(kvm, SYS_ID_AA64DFR0_EL1);
int i, r;
if (WARN_ON(realm->rd) || WARN_ON(!realm->params))
@@ -335,6 +336,8 @@ static int realm_create_rd(struct kvm *kvm)
params->rtt_num_start = pgt->pgd_pages;
params->rtt_base = kvm->arch.mmu.pgd_phys;
params->vmid = realm->vmid;
+ params->num_bps = SYS_FIELD_GET(ID_AA64DFR0_EL1, BRPs, dfr0) + 1;
+ params->num_wps = SYS_FIELD_GET(ID_AA64DFR0_EL1, WRPs, dfr0) + 1;
if (kvm->arch.arm_pmu) {
params->pmu_num_ctrs = kvm->arch.pmcr_n;
diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c
index cb3dc640a4a5..5703b63186fd 100644
--- a/arch/arm64/kvm/sys_regs.c
+++ b/arch/arm64/kvm/sys_regs.c
@@ -1717,6 +1717,9 @@ static int set_id_aa64dfr0_el1(struct kvm_vcpu *vcpu,
{
u8 debugver = SYS_FIELD_GET(ID_AA64DFR0_EL1, DebugVer, val);
u8 pmuver = SYS_FIELD_GET(ID_AA64DFR0_EL1, PMUVer, val);
+ u8 bps = SYS_FIELD_GET(ID_AA64DFR0_EL1, BRPs, val);
+ u8 wps = SYS_FIELD_GET(ID_AA64DFR0_EL1, WRPs, val);
+ u8 ctx_cmps = SYS_FIELD_GET(ID_AA64DFR0_EL1, CTX_CMPs, val);
/*
* Prior to commit 3d0dba5764b9 ("KVM: arm64: PMU: Move the
@@ -1736,10 +1739,11 @@ static int set_id_aa64dfr0_el1(struct kvm_vcpu *vcpu,
val &= ~ID_AA64DFR0_EL1_PMUVer_MASK;
/*
- * ID_AA64DFR0_EL1.DebugVer is one of those awkward fields with a
- * nonzero minimum safe value.
+ * ID_AA64DFR0_EL1.DebugVer, BRPs and WRPs all have to be greater than
+ * zero. CTX_CMPs is never greater than BRPs.
*/
- if (debugver < ID_AA64DFR0_EL1_DebugVer_IMP)
+ if (debugver < ID_AA64DFR0_EL1_DebugVer_IMP || !bps || !wps ||
+ ctx_cmps > bps)
return -EINVAL;
return set_id_reg(vcpu, rd, val);
@@ -1822,10 +1826,11 @@ static int set_id_reg(struct kvm_vcpu *vcpu, const struct sys_reg_desc *rd,
mutex_lock(&vcpu->kvm->arch.config_lock);
/*
- * Once the VM has started the ID registers are immutable. Reject any
- * write that does not match the final register value.
+ * Once the VM has started or the Realm descriptor is created, the ID
+ * registers are immutable. Reject any write that does not match the
+ * final register value.
*/
- if (kvm_vm_has_ran_once(vcpu->kvm)) {
+ if (kvm_vm_has_ran_once(vcpu->kvm) || kvm_realm_is_created(vcpu->kvm)) {
if (val != read_id_reg(vcpu, rd))
ret = -EBUSY;
else
@@ -2307,7 +2312,9 @@ static const struct sys_reg_desc sys_reg_descs[] = {
.set_user = set_id_aa64dfr0_el1,
.reset = read_sanitised_id_aa64dfr0_el1,
.val = ID_AA64DFR0_EL1_PMUVer_MASK |
- ID_AA64DFR0_EL1_DebugVer_MASK, },
+ ID_AA64DFR0_EL1_DebugVer_MASK |
+ ID_AA64DFR0_EL1_BRPs_MASK |
+ ID_AA64DFR0_EL1_WRPs_MASK, },
ID_SANITISED(ID_AA64DFR1_EL1),
ID_UNALLOCATED(5,2),
ID_UNALLOCATED(5,3),
--
2.34.1
From: Jean-Philippe Brucker <[email protected]>
Provide an accurate number of available PMU counters to userspace when
setting up a Realm.
Signed-off-by: Jean-Philippe Brucker <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_rme.h | 1 +
arch/arm64/kvm/pmu-emul.c | 3 +++
arch/arm64/kvm/rme.c | 5 +++++
3 files changed, 9 insertions(+)
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index 44aa95befc14..9c00bcb018f8 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -88,6 +88,7 @@ struct realm_rec {
int kvm_init_rme(void);
u32 kvm_realm_ipa_limit(void);
+u8 kvm_realm_max_pmu_counters(void);
u64 kvm_realm_reset_id_aa64dfr0_el1(struct kvm_vcpu *vcpu, u64 val);
bool kvm_rme_supports_sve(void);
diff --git a/arch/arm64/kvm/pmu-emul.c b/arch/arm64/kvm/pmu-emul.c
index ce7c8e55d904..39790736f1e2 100644
--- a/arch/arm64/kvm/pmu-emul.c
+++ b/arch/arm64/kvm/pmu-emul.c
@@ -912,6 +912,9 @@ u8 kvm_arm_pmu_get_max_counters(struct kvm *kvm)
{
struct arm_pmu *arm_pmu = kvm->arch.arm_pmu;
+ if (kvm_is_realm(kvm))
+ return kvm_realm_max_pmu_counters();
+
/*
* The arm_pmu->num_events considers the cycle counter as well.
* Ignore that and return only the general-purpose counters.
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 51ac8c3462ea..1bd97e206846 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -271,6 +271,11 @@ u32 kvm_realm_ipa_limit(void)
return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
}
+u8 kvm_realm_max_pmu_counters(void)
+{
+ return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_PMU_NUM_CTRS);
+}
+
u64 kvm_realm_reset_id_aa64dfr0_el1(struct kvm_vcpu *vcpu, u64 val)
{
u32 bps = u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_NUM_BPS);
--
2.34.1
From: Jean-Philippe Brucker <[email protected]>
RMM provides the maximum vector length it supports for a guest in its
feature register. Make it visible to the rest of KVM and to userspace
via KVM_REG_ARM64_SVE_VLS.
Signed-off-by: Jean-Philippe Brucker <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_host.h | 2 +-
arch/arm64/include/asm/kvm_rme.h | 1 +
arch/arm64/kvm/guest.c | 2 +-
arch/arm64/kvm/reset.c | 12 ++++++++++--
arch/arm64/kvm/rme.c | 6 ++++++
5 files changed, 19 insertions(+), 4 deletions(-)
diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index f7ac40ce0caf..902923402f6e 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -76,8 +76,8 @@ static inline enum kvm_mode kvm_get_mode(void) { return KVM_MODE_NONE; };
DECLARE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
-extern unsigned int __ro_after_init kvm_sve_max_vl;
int __init kvm_arm_init_sve(void);
+unsigned int kvm_sve_get_max_vl(struct kvm *kvm);
u32 __attribute_const__ kvm_target_cpu(void);
void kvm_reset_vcpu(struct kvm_vcpu *vcpu);
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index 9c00bcb018f8..be24be001aaa 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -89,6 +89,7 @@ struct realm_rec {
int kvm_init_rme(void);
u32 kvm_realm_ipa_limit(void);
u8 kvm_realm_max_pmu_counters(void);
+unsigned int kvm_realm_sve_max_vl(void);
u64 kvm_realm_reset_id_aa64dfr0_el1(struct kvm_vcpu *vcpu, u64 val);
bool kvm_rme_supports_sve(void);
diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c
index f9a47ce71a26..c62fda66cdc5 100644
--- a/arch/arm64/kvm/guest.c
+++ b/arch/arm64/kvm/guest.c
@@ -355,7 +355,7 @@ static int set_sve_vls(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
if (vq_present(vqs, vq))
max_vq = vq;
- if (max_vq > sve_vq_from_vl(kvm_sve_max_vl))
+ if (max_vq > sve_vq_from_vl(kvm_sve_get_max_vl(vcpu->kvm)))
return -EINVAL;
/*
diff --git a/arch/arm64/kvm/reset.c b/arch/arm64/kvm/reset.c
index 6e6eb4a15095..a90b7c2d35bb 100644
--- a/arch/arm64/kvm/reset.c
+++ b/arch/arm64/kvm/reset.c
@@ -45,7 +45,7 @@ static u32 __ro_after_init kvm_ipa_limit;
#define VCPU_RESET_PSTATE_SVC (PSR_AA32_MODE_SVC | PSR_AA32_A_BIT | \
PSR_AA32_I_BIT | PSR_AA32_F_BIT)
-unsigned int __ro_after_init kvm_sve_max_vl;
+static unsigned int __ro_after_init kvm_sve_max_vl;
int __init kvm_arm_init_sve(void)
{
@@ -73,9 +73,17 @@ int __init kvm_arm_init_sve(void)
return 0;
}
+unsigned int kvm_sve_get_max_vl(struct kvm *kvm)
+{
+ if (kvm_is_realm(kvm))
+ return kvm_realm_sve_max_vl();
+ else
+ return kvm_sve_max_vl;
+}
+
static void kvm_vcpu_enable_sve(struct kvm_vcpu *vcpu)
{
- vcpu->arch.sve_max_vl = kvm_sve_max_vl;
+ vcpu->arch.sve_max_vl = kvm_sve_get_max_vl(vcpu->kvm);
/*
* Userspace can still customize the vector lengths by writing
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 1bd97e206846..cd5b74aac092 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -276,6 +276,12 @@ u8 kvm_realm_max_pmu_counters(void)
return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_PMU_NUM_CTRS);
}
+unsigned int kvm_realm_sve_max_vl(void)
+{
+ return sve_vl_from_vq(u64_get_bits(rmm_feat_reg0,
+ RMI_FEATURE_REGISTER_0_SVE_VL) + 1);
+}
+
u64 kvm_realm_reset_id_aa64dfr0_el1(struct kvm_vcpu *vcpu, u64 val)
{
u32 bps = u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_NUM_BPS);
--
2.34.1
From: Jean-Philippe Brucker <[email protected]>
Obtain the max vector length configured by userspace on the vCPUs, and
write it into the Realm parameters. By default the vCPU is configured
with the max vector length reported by RMM, and userspace can reduce it
with a write to KVM_REG_ARM64_SVE_VLS.
Signed-off-by: Jean-Philippe Brucker <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/guest.c | 3 ++-
arch/arm64/kvm/rme.c | 42 ++++++++++++++++++++++++++++++++++++++++++
2 files changed, 44 insertions(+), 1 deletion(-)
diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c
index c62fda66cdc5..d72e59e79185 100644
--- a/arch/arm64/kvm/guest.c
+++ b/arch/arm64/kvm/guest.c
@@ -341,7 +341,7 @@ static int set_sve_vls(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
if (!vcpu_has_sve(vcpu))
return -ENOENT;
- if (kvm_arm_vcpu_sve_finalized(vcpu))
+ if (kvm_arm_vcpu_sve_finalized(vcpu) || kvm_realm_is_created(vcpu->kvm))
return -EPERM; /* too late! */
if (WARN_ON(vcpu->arch.sve_state))
@@ -807,6 +807,7 @@ static bool validate_realm_set_reg(struct kvm_vcpu *vcpu,
switch (reg->id) {
case KVM_REG_ARM_PMCR_EL0:
case KVM_REG_ARM_ID_AA64DFR0_EL1:
+ case KVM_REG_ARM64_SVE_VLS:
return true;
}
}
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index cd5b74aac092..93aab6caddf5 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -309,6 +309,44 @@ static int get_start_level(struct realm *realm)
return 4 - stage2_pgtable_levels(realm->ia_bits);
}
+static int realm_init_sve_param(struct kvm *kvm, struct realm_params *params)
+{
+ int ret = 0;
+ unsigned long i;
+ struct kvm_vcpu *vcpu;
+ int max_vl, realm_max_vl = -1;
+
+ /*
+ * Get the preferred SVE configuration, set by userspace with the
+ * KVM_ARM_VCPU_SVE feature and KVM_REG_ARM64_SVE_VLS pseudo-register.
+ */
+ kvm_for_each_vcpu(i, vcpu, kvm) {
+ mutex_lock(&vcpu->mutex);
+ if (vcpu_has_sve(vcpu)) {
+ if (!kvm_arm_vcpu_sve_finalized(vcpu))
+ ret = -EINVAL;
+ max_vl = vcpu->arch.sve_max_vl;
+ } else {
+ max_vl = 0;
+ }
+ mutex_unlock(&vcpu->mutex);
+ if (ret)
+ return ret;
+
+ /* We need all vCPUs to have the same SVE config */
+ if (realm_max_vl >= 0 && realm_max_vl != max_vl)
+ return -EINVAL;
+
+ realm_max_vl = max_vl;
+ }
+
+ if (realm_max_vl > 0) {
+ params->sve_vl = sve_vq_from_vl(realm_max_vl) - 1;
+ params->flags |= RMI_REALM_PARAM_FLAG_SVE;
+ }
+ return 0;
+}
+
static int realm_create_rd(struct kvm *kvm)
{
struct realm *realm = &kvm->arch.realm;
@@ -355,6 +393,10 @@ static int realm_create_rd(struct kvm *kvm)
params->flags |= RMI_REALM_PARAM_FLAG_PMU;
}
+ r = realm_init_sve_param(kvm, params);
+ if (r)
+ goto out_undelegate_tables;
+
params_phys = virt_to_phys(params);
if (rmi_realm_create(rd_phys, params_phys)) {
--
2.34.1
From: Jean-Philippe Brucker <[email protected]>
KVM_GET_REG_LIST should not be called before SVE is finalized. The ioctl
handler currently returns -EPERM in this case. But because it uses
kvm_arm_vcpu_is_finalized(), it now also rejects the call for
unfinalized REC even though finalizing the REC can only be done late,
after Realm descriptor creation.
Move the check to copy_sve_reg_indices(). One adverse side effect of
this change is that a KVM_GET_REG_LIST call that only probes for the
array size will now succeed even if SVE is not finalized, but that seems
harmless since the following KVM_GET_REG_LIST with the full array will
fail.
Signed-off-by: Jean-Philippe Brucker <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/arm.c | 4 ----
arch/arm64/kvm/guest.c | 9 +++------
2 files changed, 3 insertions(+), 10 deletions(-)
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index 69d29797c2ed..2dd014d3c366 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -1696,10 +1696,6 @@ long kvm_arch_vcpu_ioctl(struct file *filp,
if (unlikely(!kvm_vcpu_initialized(vcpu)))
break;
- r = -EPERM;
- if (!kvm_arm_vcpu_is_finalized(vcpu))
- break;
-
r = -EFAULT;
if (copy_from_user(®_list, user_list, sizeof(reg_list)))
break;
diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c
index d72e59e79185..f1fe51775649 100644
--- a/arch/arm64/kvm/guest.c
+++ b/arch/arm64/kvm/guest.c
@@ -652,12 +652,9 @@ static unsigned long num_sve_regs(const struct kvm_vcpu *vcpu)
{
const unsigned int slices = vcpu_sve_slices(vcpu);
- if (!vcpu_has_sve(vcpu))
+ if (!vcpu_has_sve(vcpu) || !kvm_arm_vcpu_sve_finalized(vcpu))
return 0;
- /* Policed by KVM_GET_REG_LIST: */
- WARN_ON(!kvm_arm_vcpu_sve_finalized(vcpu));
-
return slices * (SVE_NUM_PREGS + SVE_NUM_ZREGS + 1 /* FFR */)
+ 1; /* KVM_REG_ARM64_SVE_VLS */
}
@@ -673,8 +670,8 @@ static int copy_sve_reg_indices(const struct kvm_vcpu *vcpu,
if (!vcpu_has_sve(vcpu))
return 0;
- /* Policed by KVM_GET_REG_LIST: */
- WARN_ON(!kvm_arm_vcpu_sve_finalized(vcpu));
+ if (!kvm_arm_vcpu_sve_finalized(vcpu))
+ return -EPERM;
/*
* Enumerate this first, so that userspace can save/restore in
--
2.34.1
Select KVM_GENERIC_PRIVATE_MEM and provide the necessary support
functions.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_host.h | 4 ++++
arch/arm64/kvm/Kconfig | 1 +
arch/arm64/kvm/arm.c | 5 +++++
arch/arm64/kvm/mmu.c | 19 +++++++++++++++++++
4 files changed, 29 insertions(+)
diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index 902923402f6e..93de7f5009fe 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -1259,6 +1259,10 @@ static inline bool kvm_vm_is_protected(struct kvm *kvm)
return false;
}
+#ifdef CONFIG_KVM_PRIVATE_MEM
+bool kvm_arch_has_private_mem(struct kvm *kvm);
+#endif
+
int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
diff --git a/arch/arm64/kvm/Kconfig b/arch/arm64/kvm/Kconfig
index 58f09370d17e..8da57e74c86a 100644
--- a/arch/arm64/kvm/Kconfig
+++ b/arch/arm64/kvm/Kconfig
@@ -37,6 +37,7 @@ menuconfig KVM
select HAVE_KVM_VCPU_RUN_PID_CHANGE
select SCHED_INFO
select GUEST_PERF_EVENTS if PERF_EVENTS
+ select KVM_GENERIC_PRIVATE_MEM
help
Support hosting virtualized guest machines.
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index 2dd014d3c366..a66d0a6eb4fa 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -89,6 +89,11 @@ int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
return kvm_vcpu_exiting_guest_mode(vcpu) == IN_GUEST_MODE;
}
+bool kvm_arch_has_private_mem(struct kvm *kvm)
+{
+ return kvm_is_realm(kvm);
+}
+
int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
struct kvm_enable_cap *cap)
{
diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index 48c957e21c83..808bceebad4d 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -2171,6 +2171,25 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
return ret;
}
+bool kvm_arch_pre_set_memory_attributes(struct kvm *kvm,
+ struct kvm_gfn_range *range)
+{
+ WARN_ON_ONCE(!kvm_arch_has_private_mem(kvm));
+ return false;
+}
+
+bool kvm_arch_post_set_memory_attributes(struct kvm *kvm,
+ struct kvm_gfn_range *range)
+{
+ WARN_ON_ONCE(!kvm_arch_has_private_mem(kvm));
+
+ if (range->arg.attributes & KVM_MEMORY_ATTRIBUTE_PRIVATE)
+ range->only_shared = true;
+ kvm_unmap_gfn_range(kvm, range);
+
+ return false;
+}
+
void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
{
}
--
2.34.1
The RMM only allows setting the lower GPRS (x0-x7) and PC for a realm
guest. Check this in kvm_arm_set_reg() so that the VMM can receive a
suitable error return if other registers are accessed.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/guest.c | 26 ++++++++++++++++++++++++++
1 file changed, 26 insertions(+)
diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c
index e2f762d959bb..a035689dc39b 100644
--- a/arch/arm64/kvm/guest.c
+++ b/arch/arm64/kvm/guest.c
@@ -782,12 +782,38 @@ int kvm_arm_get_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
return kvm_arm_sys_reg_get_reg(vcpu, reg);
}
+/*
+ * The RMI ABI only enables setting the lower GPRs (x0-x7) and PC.
+ * All other registers are reset to architectural or otherwise defined reset
+ * values by the RMM, except for a few configuration fields that correspond to
+ * Realm parameters.
+ */
+static bool validate_realm_set_reg(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg)
+{
+ if ((reg->id & KVM_REG_ARM_COPROC_MASK) == KVM_REG_ARM_CORE) {
+ u64 off = core_reg_offset_from_id(reg->id);
+
+ switch (off) {
+ case KVM_REG_ARM_CORE_REG(regs.regs[0]) ...
+ KVM_REG_ARM_CORE_REG(regs.regs[7]):
+ case KVM_REG_ARM_CORE_REG(regs.pc):
+ return true;
+ }
+ }
+
+ return false;
+}
+
int kvm_arm_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
{
/* We currently use nothing arch-specific in upper 32 bits */
if ((reg->id & ~KVM_REG_SIZE_MASK) >> 32 != KVM_REG_ARM64 >> 32)
return -EINVAL;
+ if (kvm_is_realm(vcpu->kvm) && !validate_realm_set_reg(vcpu, reg))
+ return -EINVAL;
+
switch (reg->id & KVM_REG_ARM_COPROC_MASK) {
case KVM_REG_ARM_CORE: return set_core_reg(vcpu, reg);
case KVM_REG_ARM_FW:
--
2.34.1
Add the ioctl to activate a realm and set the static branch to enable
access to the realm functionality if the RMM is detected.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/rme.c | 19 ++++++++++++++++++-
1 file changed, 18 insertions(+), 1 deletion(-)
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 93aab6caddf5..5901d57ca9d0 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -1167,6 +1167,20 @@ static int kvm_init_ipa_range_realm(struct kvm *kvm,
return ret;
}
+static int kvm_activate_realm(struct kvm *kvm)
+{
+ struct realm *realm = &kvm->arch.realm;
+
+ if (kvm_realm_state(kvm) != REALM_STATE_NEW)
+ return -EINVAL;
+
+ if (rmi_realm_activate(virt_to_phys(realm->rd)))
+ return -ENXIO;
+
+ WRITE_ONCE(realm->state, REALM_STATE_ACTIVE);
+ return 0;
+}
+
/* Protects access to rme_vmid_bitmap */
static DEFINE_SPINLOCK(rme_vmid_lock);
static unsigned long *rme_vmid_bitmap;
@@ -1314,6 +1328,9 @@ int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
r = kvm_populate_realm(kvm, &args);
break;
}
+ case KVM_CAP_ARM_RME_ACTIVATE_REALM:
+ r = kvm_activate_realm(kvm);
+ break;
default:
r = -EINVAL;
break;
@@ -1567,7 +1584,7 @@ int kvm_init_rme(void)
if (ret)
return ret;
- /* Future patch will enable static branch kvm_rme_is_available */
+ static_branch_enable(&kvm_rme_is_available);
return 0;
}
--
2.34.1
When loading a realm VCPU much of the work is handled by the RMM so only
some of the actions are required. Rearrange kvm_arch_vcpu_load()
slightly so we can bail out early for a realm guest.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/arm.c | 13 +++++++++----
1 file changed, 9 insertions(+), 4 deletions(-)
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index e92afb3835f6..4fd58a1d3351 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -512,10 +512,6 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
kvm_vgic_load(vcpu);
kvm_timer_vcpu_load(vcpu);
- if (has_vhe())
- kvm_vcpu_load_vhe(vcpu);
- kvm_arch_vcpu_load_fp(vcpu);
- kvm_vcpu_pmu_restore_guest(vcpu);
if (kvm_arm_is_pvtime_enabled(&vcpu->arch))
kvm_make_request(KVM_REQ_RECORD_STEAL, vcpu);
@@ -524,6 +520,15 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
else
vcpu_set_wfx_traps(vcpu);
+ /* No additional state needs to be loaded on Realmed VMs */
+ if (vcpu_is_rec(vcpu))
+ return;
+
+ if (has_vhe())
+ kvm_vcpu_load_vhe(vcpu);
+ kvm_arch_vcpu_load_fp(vcpu);
+ kvm_vcpu_pmu_restore_guest(vcpu);
+
if (vcpu_has_ptrauth(vcpu))
vcpu_ptrauth_disable(vcpu);
kvm_arch_vcpu_load_debug_state_flags(vcpu);
--
2.34.1
The RMM doesn't allow injection of a undefined exception into a realm
guest. Add a WARN to catch if this ever happens.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/inject_fault.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/kvm/inject_fault.c b/arch/arm64/kvm/inject_fault.c
index a640e839848e..44ce1c9bdc2e 100644
--- a/arch/arm64/kvm/inject_fault.c
+++ b/arch/arm64/kvm/inject_fault.c
@@ -224,6 +224,8 @@ void kvm_inject_size_fault(struct kvm_vcpu *vcpu)
*/
void kvm_inject_undefined(struct kvm_vcpu *vcpu)
{
+ if (vcpu_is_rec(vcpu))
+ WARN(1, "Cannot inject undefined exception into REC. Continuing with unknown behaviour");
if (vcpu_el1_is_32bit(vcpu))
inject_undef32(vcpu);
else
--
2.34.1
From: Suzuki K Poulose <[email protected]>
Fix a potential build error (like below, when asm/kvm_emulate.h gets included after
the kvm/arm_psci.h) by including the missing header file in kvm/arm_psci.h:
/include/kvm/arm_psci.h: In function ‘kvm_psci_version’:
/include/kvm/arm_psci.h:29:13: error: implicit declaration of function
‘vcpu_has_feature’; did you mean ‘cpu_have_feature’? [-Werror=implicit-function-declaration]
29 | if (vcpu_has_feature(vcpu, KVM_ARM_VCPU_PSCI_0_2)) {
| ^~~~~~~~~~~~~~~~
| cpu_have_feature
Signed-off-by: Suzuki K Poulose <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
include/kvm/arm_psci.h | 2 ++
1 file changed, 2 insertions(+)
diff --git a/include/kvm/arm_psci.h b/include/kvm/arm_psci.h
index e8fb624013d1..1801c6fd3f10 100644
--- a/include/kvm/arm_psci.h
+++ b/include/kvm/arm_psci.h
@@ -10,6 +10,8 @@
#include <linux/kvm_host.h>
#include <uapi/linux/psci.h>
+#include <asm/kvm_emulate.h>
+
#define KVM_ARM_PSCI_0_1 PSCI_VERSION(0, 1)
#define KVM_ARM_PSCI_0_2 PSCI_VERSION(0, 2)
#define KVM_ARM_PSCI_1_0 PSCI_VERSION(1, 0)
--
2.34.1
From: Joey Gouly <[email protected]>
Forward RSI_HOST_CALLS to KVM's HVC handler.
Signed-off-by: Joey Gouly <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/rme-exit.c | 25 +++++++++++++++++++++++++
1 file changed, 25 insertions(+)
diff --git a/arch/arm64/kvm/rme-exit.c b/arch/arm64/kvm/rme-exit.c
index 2ddaec09cc98..b442e9a00d07 100644
--- a/arch/arm64/kvm/rme-exit.c
+++ b/arch/arm64/kvm/rme-exit.c
@@ -116,6 +116,29 @@ static int rec_exit_ripas_change(struct kvm_vcpu *vcpu)
return 0;
}
+static int rec_exit_host_call(struct kvm_vcpu *vcpu)
+{
+ int ret, i;
+ struct realm_rec *rec = &vcpu->arch.rec;
+
+ vcpu->stat.hvc_exit_stat++;
+
+ for (i = 0; i < REC_RUN_GPRS; i++)
+ vcpu_set_reg(vcpu, i, rec->run->exit.gprs[i]);
+
+ ret = kvm_smccc_call_handler(vcpu);
+
+ if (ret < 0) {
+ vcpu_set_reg(vcpu, 0, ~0UL);
+ ret = 1;
+ }
+
+ for (i = 0; i < REC_RUN_GPRS; i++)
+ rec->run->entry.gprs[i] = vcpu_get_reg(vcpu, i);
+
+ return ret;
+}
+
static void update_arch_timer_irq_lines(struct kvm_vcpu *vcpu)
{
struct realm_rec *rec = &vcpu->arch.rec;
@@ -177,6 +200,8 @@ int handle_rme_exit(struct kvm_vcpu *vcpu, int rec_run_ret)
return rec_exit_psci(vcpu);
case RMI_EXIT_RIPAS_CHANGE:
return rec_exit_ripas_change(vcpu);
+ case RMI_EXIT_HOST_CALL:
+ return rec_exit_host_call(vcpu);
}
kvm_pr_unimpl("Unsupported exit reason: %u\n",
--
2.34.1
From: Suzuki K Poulose <[email protected]>
Given we have different types of VMs supported, check the
support for SVE for the given instance of the VM to accurately
report the status.
Signed-off-by: Suzuki K Poulose <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_rme.h | 2 ++
arch/arm64/kvm/arm.c | 5 ++++-
arch/arm64/kvm/rme.c | 5 +++++
3 files changed, 11 insertions(+), 1 deletion(-)
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index fef04706ff9d..62d1fa82ac92 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -89,6 +89,8 @@ struct realm_rec {
int kvm_init_rme(void);
u32 kvm_realm_ipa_limit(void);
+bool kvm_rme_supports_sve(void);
+
int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap);
int kvm_init_realm_vm(struct kvm *kvm);
void kvm_destroy_realm(struct kvm *kvm);
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index f2279ab45add..dcd9089877f3 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -363,7 +363,10 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
r = get_kvm_ipa_limit();
break;
case KVM_CAP_ARM_SVE:
- r = system_supports_sve();
+ if (kvm && kvm_is_realm(kvm))
+ r = kvm_rme_supports_sve();
+ else
+ r = system_supports_sve();
break;
case KVM_CAP_ARM_PTRAUTH_ADDRESS:
case KVM_CAP_ARM_PTRAUTH_GENERIC:
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 9544653accb6..9593e8e35913 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -20,6 +20,11 @@ static bool rme_supports(unsigned long feature)
return !!u64_get_bits(rmm_feat_reg0, feature);
}
+bool kvm_rme_supports_sve(void)
+{
+ return rme_supports(RMI_FEATURE_REGISTER_0_SVE_EN);
+}
+
static int rmi_check_version(void)
{
struct arm_smccc_res res;
--
2.34.1
For protected memory read only isn't supported. While it may be possible
to support read only for unprotected memory, this isn't supported at the
present time.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/arm.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index 2aad83053b62..69d29797c2ed 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -284,7 +284,6 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
case KVM_CAP_ONE_REG:
case KVM_CAP_ARM_PSCI:
case KVM_CAP_ARM_PSCI_0_2:
- case KVM_CAP_READONLY_MEM:
case KVM_CAP_MP_STATE:
case KVM_CAP_IMMEDIATE_EXIT:
case KVM_CAP_VCPU_EVENTS:
@@ -298,6 +297,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
case KVM_CAP_COUNTER_OFFSET:
r = 1;
break;
+ case KVM_CAP_READONLY_MEM:
case KVM_CAP_SET_GUEST_DEBUG:
r = !kvm_is_realm(kvm);
break;
--
2.34.1
The RMM provides emulation of a VGIC to the realm guest but delegates
much of the handling to the host. Implement support in KVM for
saving/restoring state to/from the REC structure.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/arm.c | 15 +++++++++++---
arch/arm64/kvm/vgic/vgic-v3.c | 9 +++++++--
arch/arm64/kvm/vgic/vgic.c | 37 +++++++++++++++++++++++++++++++++--
3 files changed, 54 insertions(+), 7 deletions(-)
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index d70c511e16a0..f8daf478114f 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -534,17 +534,22 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
{
+ kvm_timer_vcpu_put(vcpu);
+ kvm_vgic_put(vcpu);
+
+ vcpu->cpu = -1;
+
+ if (vcpu_is_rec(vcpu))
+ return;
+
kvm_arch_vcpu_put_debug_state_flags(vcpu);
kvm_arch_vcpu_put_fp(vcpu);
if (has_vhe())
kvm_vcpu_put_vhe(vcpu);
- kvm_timer_vcpu_put(vcpu);
- kvm_vgic_put(vcpu);
kvm_vcpu_pmu_restore_host(vcpu);
kvm_arm_vmid_clear_active();
vcpu_clear_on_unsupported_cpu(vcpu);
- vcpu->cpu = -1;
}
static void __kvm_arm_vcpu_power_off(struct kvm_vcpu *vcpu)
@@ -758,6 +763,10 @@ int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
}
if (!irqchip_in_kernel(kvm)) {
+ /* Userspace irqchip not yet supported with Realms */
+ if (kvm_is_realm(vcpu->kvm))
+ return -EOPNOTSUPP;
+
/*
* Tell the rest of the code that there are userspace irqchip
* VMs in the wild.
diff --git a/arch/arm64/kvm/vgic/vgic-v3.c b/arch/arm64/kvm/vgic/vgic-v3.c
index 4ea3340786b9..75ae98b0f700 100644
--- a/arch/arm64/kvm/vgic/vgic-v3.c
+++ b/arch/arm64/kvm/vgic/vgic-v3.c
@@ -7,9 +7,11 @@
#include <linux/kvm.h>
#include <linux/kvm_host.h>
#include <kvm/arm_vgic.h>
+#include <asm/kvm_emulate.h>
#include <asm/kvm_hyp.h>
#include <asm/kvm_mmu.h>
#include <asm/kvm_asm.h>
+#include <asm/rmi_smc.h>
#include "vgic.h"
@@ -667,7 +669,8 @@ int vgic_v3_probe(const struct gic_kvm_info *info)
(unsigned long long)info->vcpu.start);
} else if (kvm_get_mode() != KVM_MODE_PROTECTED) {
kvm_vgic_global_state.vcpu_base = info->vcpu.start;
- kvm_vgic_global_state.can_emulate_gicv2 = true;
+ if (!static_branch_unlikely(&kvm_rme_is_available))
+ kvm_vgic_global_state.can_emulate_gicv2 = true;
ret = kvm_register_vgic_device(KVM_DEV_TYPE_ARM_VGIC_V2);
if (ret) {
kvm_err("Cannot register GICv2 KVM device.\n");
@@ -742,7 +745,9 @@ void vgic_v3_vmcr_sync(struct kvm_vcpu *vcpu)
{
struct vgic_v3_cpu_if *cpu_if = &vcpu->arch.vgic_cpu.vgic_v3;
- if (likely(cpu_if->vgic_sre))
+ if (vcpu_is_rec(vcpu))
+ cpu_if->vgic_vmcr = vcpu->arch.rec.run->exit.gicv3_vmcr;
+ else if (likely(cpu_if->vgic_sre))
cpu_if->vgic_vmcr = kvm_call_hyp_ret(__vgic_v3_read_vmcr);
}
diff --git a/arch/arm64/kvm/vgic/vgic.c b/arch/arm64/kvm/vgic/vgic.c
index 4ec93587c8cd..02561d4e6447 100644
--- a/arch/arm64/kvm/vgic/vgic.c
+++ b/arch/arm64/kvm/vgic/vgic.c
@@ -10,7 +10,9 @@
#include <linux/list_sort.h>
#include <linux/nospec.h>
+#include <asm/kvm_emulate.h>
#include <asm/kvm_hyp.h>
+#include <asm/rmi_smc.h>
#include "vgic.h"
@@ -845,10 +847,23 @@ static inline bool can_access_vgic_from_kernel(void)
return !static_branch_unlikely(&kvm_vgic_global_state.gicv3_cpuif) || has_vhe();
}
+static inline void vgic_rmm_save_state(struct kvm_vcpu *vcpu)
+{
+ struct vgic_v3_cpu_if *cpu_if = &vcpu->arch.vgic_cpu.vgic_v3;
+ int i;
+
+ for (i = 0; i < kvm_vgic_global_state.nr_lr; i++) {
+ cpu_if->vgic_lr[i] = vcpu->arch.rec.run->exit.gicv3_lrs[i];
+ vcpu->arch.rec.run->entry.gicv3_lrs[i] = 0;
+ }
+}
+
static inline void vgic_save_state(struct kvm_vcpu *vcpu)
{
if (!static_branch_unlikely(&kvm_vgic_global_state.gicv3_cpuif))
vgic_v2_save_state(vcpu);
+ else if (vcpu_is_rec(vcpu))
+ vgic_rmm_save_state(vcpu);
else
__vgic_v3_save_state(&vcpu->arch.vgic_cpu.vgic_v3);
}
@@ -875,10 +890,28 @@ void kvm_vgic_sync_hwstate(struct kvm_vcpu *vcpu)
vgic_prune_ap_list(vcpu);
}
+static inline void vgic_rmm_restore_state(struct kvm_vcpu *vcpu)
+{
+ struct vgic_v3_cpu_if *cpu_if = &vcpu->arch.vgic_cpu.vgic_v3;
+ int i;
+
+ for (i = 0; i < kvm_vgic_global_state.nr_lr; i++) {
+ vcpu->arch.rec.run->entry.gicv3_lrs[i] = cpu_if->vgic_lr[i];
+ /*
+ * Also populate the rec.run->exit copies so that a late
+ * decision to back out from entering the realm doesn't cause
+ * the state to be lost
+ */
+ vcpu->arch.rec.run->exit.gicv3_lrs[i] = cpu_if->vgic_lr[i];
+ }
+}
+
static inline void vgic_restore_state(struct kvm_vcpu *vcpu)
{
if (!static_branch_unlikely(&kvm_vgic_global_state.gicv3_cpuif))
vgic_v2_restore_state(vcpu);
+ else if (vcpu_is_rec(vcpu))
+ vgic_rmm_restore_state(vcpu);
else
__vgic_v3_restore_state(&vcpu->arch.vgic_cpu.vgic_v3);
}
@@ -919,7 +952,7 @@ void kvm_vgic_flush_hwstate(struct kvm_vcpu *vcpu)
void kvm_vgic_load(struct kvm_vcpu *vcpu)
{
- if (unlikely(!vgic_initialized(vcpu->kvm)))
+ if (unlikely(!vgic_initialized(vcpu->kvm)) || vcpu_is_rec(vcpu))
return;
if (kvm_vgic_global_state.type == VGIC_V2)
@@ -930,7 +963,7 @@ void kvm_vgic_load(struct kvm_vcpu *vcpu)
void kvm_vgic_put(struct kvm_vcpu *vcpu)
{
- if (unlikely(!vgic_initialized(vcpu->kvm)))
+ if (unlikely(!vgic_initialized(vcpu->kvm)) || vcpu_is_rec(vcpu))
return;
if (kvm_vgic_global_state.type == VGIC_V2)
--
2.34.1
From: Jean-Philippe Brucker <[email protected]>
The RMM describes the maximum number of BPs/WPs available to the guest
in the Feature Register 0. Propagate those numbers into ID_AA64DFR0_EL1,
which is visible to userspace. A VMM needs this information in order to
set up realm parameters.
Signed-off-by: Jean-Philippe Brucker <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_rme.h | 1 +
arch/arm64/kvm/rme.c | 22 ++++++++++++++++++++++
arch/arm64/kvm/sys_regs.c | 2 +-
3 files changed, 24 insertions(+), 1 deletion(-)
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index 62d1fa82ac92..44aa95befc14 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -88,6 +88,7 @@ struct realm_rec {
int kvm_init_rme(void);
u32 kvm_realm_ipa_limit(void);
+u64 kvm_realm_reset_id_aa64dfr0_el1(struct kvm_vcpu *vcpu, u64 val);
bool kvm_rme_supports_sve(void);
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index e60a1196a2fe..4edc8d98e1e6 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -271,6 +271,28 @@ u32 kvm_realm_ipa_limit(void)
return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
}
+u64 kvm_realm_reset_id_aa64dfr0_el1(struct kvm_vcpu *vcpu, u64 val)
+{
+ u32 bps = u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_NUM_BPS);
+ u32 wps = u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_NUM_WPS);
+ u32 ctx_cmps;
+
+ if (!kvm_is_realm(vcpu->kvm))
+ return val;
+
+ /* Ensure CTX_CMPs is still valid */
+ ctx_cmps = FIELD_GET(ID_AA64DFR0_EL1_CTX_CMPs, val) + 1;
+ ctx_cmps = min(bps, ctx_cmps);
+
+ val &= ~(ID_AA64DFR0_EL1_BRPs_MASK | ID_AA64DFR0_EL1_WRPs_MASK |
+ ID_AA64DFR0_EL1_CTX_CMPs);
+ val |= FIELD_PREP(ID_AA64DFR0_EL1_BRPs_MASK, bps - 1) |
+ FIELD_PREP(ID_AA64DFR0_EL1_WRPs_MASK, wps - 1) |
+ FIELD_PREP(ID_AA64DFR0_EL1_CTX_CMPs, ctx_cmps - 1);
+
+ return val;
+}
+
static int get_start_level(struct realm *realm)
{
return 4 - stage2_pgtable_levels(realm->ia_bits);
diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c
index 60452c6519a4..cb3dc640a4a5 100644
--- a/arch/arm64/kvm/sys_regs.c
+++ b/arch/arm64/kvm/sys_regs.c
@@ -1708,7 +1708,7 @@ static u64 read_sanitised_id_aa64dfr0_el1(struct kvm_vcpu *vcpu,
/* Hide SPE from guests */
val &= ~ID_AA64DFR0_EL1_PMSVer_MASK;
- return val;
+ return kvm_realm_reset_id_aa64dfr0_el1(vcpu, val);
}
static int set_id_aa64dfr0_el1(struct kvm_vcpu *vcpu,
--
2.34.1
Each page within the protected region of the realm guest can be marked
as either RAM or EMPTY. Allow the VMM to control this before the guest
has started and provide the equivalent functions to change this (with
the guest's approval) at runtime.
When transitioning from RIPAS RAM (1) to RIPAS EMPTY (0) the memory is
unmapped from the guest and undelegated allowing the memory to be reused
by the host. When transitioning to RIPAS RAM the actual population of
the leaf RTTs is done later on stage 2 fault, however it may be
necessary to allocate additional RTTs to represent the range requested.
When freeing a block mapping it is necessary to temporarily unfold the
RTT which requires delegating an extra page to the RMM, this page can
then be recovered once the contents of the block mapping have been
freed. A spare, delegated page (spare_page) is used for this purpose.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_rme.h | 16 ++
arch/arm64/kvm/mmu.c | 8 +-
arch/arm64/kvm/rme.c | 390 +++++++++++++++++++++++++++++++
3 files changed, 411 insertions(+), 3 deletions(-)
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index 915e76068b00..cc8f81cfc3c0 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -96,6 +96,14 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits);
int kvm_create_rec(struct kvm_vcpu *vcpu);
void kvm_destroy_rec(struct kvm_vcpu *vcpu);
+void kvm_realm_unmap_range(struct kvm *kvm,
+ unsigned long ipa,
+ u64 size,
+ bool unmap_private);
+int realm_set_ipa_state(struct kvm_vcpu *vcpu,
+ unsigned long addr, unsigned long end,
+ unsigned long ripas);
+
#define RME_RTT_BLOCK_LEVEL 2
#define RME_RTT_MAX_LEVEL 3
@@ -114,4 +122,12 @@ static inline unsigned long rme_rtt_level_mapsize(int level)
return (1UL << RME_RTT_LEVEL_SHIFT(level));
}
+static inline bool realm_is_addr_protected(struct realm *realm,
+ unsigned long addr)
+{
+ unsigned int ia_bits = realm->ia_bits;
+
+ return !(addr & ~(BIT(ia_bits - 1) - 1));
+}
+
#endif
diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index 46f0c4e80ace..8a7b5449697f 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -310,6 +310,7 @@ static void invalidate_icache_guest_page(void *va, size_t size)
* @start: The intermediate physical base address of the range to unmap
* @size: The size of the area to unmap
* @may_block: Whether or not we are permitted to block
+ * @only_shared: If true then protected mappings should not be unmapped
*
* Clear a range of stage-2 mappings, lowering the various ref-counts. Must
* be called while holding mmu_lock (unless for freeing the stage2 pgd before
@@ -317,7 +318,7 @@ static void invalidate_icache_guest_page(void *va, size_t size)
* with things behind our backs.
*/
static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64 size,
- bool may_block)
+ bool may_block, bool only_shared)
{
struct kvm *kvm = kvm_s2_mmu_to_kvm(mmu);
phys_addr_t end = start + size;
@@ -330,7 +331,7 @@ static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64
static void unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64 size)
{
- __unmap_stage2_range(mmu, start, size, true);
+ __unmap_stage2_range(mmu, start, size, true, false);
}
static void stage2_flush_memslot(struct kvm *kvm,
@@ -1771,7 +1772,8 @@ bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range)
__unmap_stage2_range(&kvm->arch.mmu, range->start << PAGE_SHIFT,
(range->end - range->start) << PAGE_SHIFT,
- range->may_block);
+ range->may_block,
+ range->only_shared);
return false;
}
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 629a095bea61..9e5983c51393 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -79,6 +79,12 @@ static phys_addr_t __alloc_delegated_page(struct realm *realm,
return phys;
}
+static phys_addr_t alloc_delegated_page(struct realm *realm,
+ struct kvm_mmu_memory_cache *mc)
+{
+ return __alloc_delegated_page(realm, mc, GFP_KERNEL);
+}
+
static void free_delegated_page(struct realm *realm, phys_addr_t phys)
{
if (realm->spare_page == PHYS_ADDR_MAX) {
@@ -94,6 +100,151 @@ static void free_delegated_page(struct realm *realm, phys_addr_t phys)
free_page((unsigned long)phys_to_virt(phys));
}
+static int realm_rtt_create(struct realm *realm,
+ unsigned long addr,
+ int level,
+ phys_addr_t phys)
+{
+ addr = ALIGN_DOWN(addr, rme_rtt_level_mapsize(level - 1));
+ return rmi_rtt_create(virt_to_phys(realm->rd), phys, addr, level);
+}
+
+static int realm_rtt_fold(struct realm *realm,
+ unsigned long addr,
+ int level,
+ phys_addr_t *rtt_granule)
+{
+ unsigned long out_rtt;
+ int ret;
+
+ ret = rmi_rtt_fold(virt_to_phys(realm->rd), addr, level, &out_rtt);
+
+ if (RMI_RETURN_STATUS(ret) == RMI_SUCCESS && rtt_granule)
+ *rtt_granule = out_rtt;
+
+ return ret;
+}
+
+static int realm_destroy_protected(struct realm *realm,
+ unsigned long ipa,
+ unsigned long *next_addr)
+{
+ unsigned long rd = virt_to_phys(realm->rd);
+ unsigned long addr;
+ phys_addr_t rtt;
+ int ret;
+
+loop:
+ ret = rmi_data_destroy(rd, ipa, &addr, next_addr);
+ if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
+ if (*next_addr > ipa)
+ return 0; /* UNASSIGNED */
+ rtt = alloc_delegated_page(realm, NULL);
+ if (WARN_ON(rtt == PHYS_ADDR_MAX))
+ return -1;
+ /* ASSIGNED - ipa is mapped as a block, so split */
+ ret = realm_rtt_create(realm, ipa,
+ RMI_RETURN_INDEX(ret) + 1, rtt);
+ if (WARN_ON(ret)) {
+ free_delegated_page(realm, rtt);
+ return -1;
+ }
+ /* retry */
+ goto loop;
+ } else if (WARN_ON(ret)) {
+ return -1;
+ }
+ ret = rmi_granule_undelegate(addr);
+
+ /*
+ * If the undelegate fails then something has gone seriously
+ * wrong: take an extra reference to just leak the page
+ */
+ if (WARN_ON(ret))
+ get_page(phys_to_page(addr));
+
+ return 0;
+}
+
+static void realm_unmap_range_shared(struct kvm *kvm,
+ int level,
+ unsigned long start,
+ unsigned long end)
+{
+ struct realm *realm = &kvm->arch.realm;
+ unsigned long rd = virt_to_phys(realm->rd);
+ ssize_t map_size = rme_rtt_level_mapsize(level);
+ unsigned long next_addr, addr;
+ unsigned long shared_bit = BIT(realm->ia_bits - 1);
+
+ if (WARN_ON(level > RME_RTT_MAX_LEVEL))
+ return;
+
+ start |= shared_bit;
+ end |= shared_bit;
+
+ for (addr = start; addr < end; addr = next_addr) {
+ unsigned long align_addr = ALIGN(addr, map_size);
+ int ret;
+
+ next_addr = ALIGN(addr + 1, map_size);
+
+ if (align_addr != addr || next_addr > end) {
+ /* Need to recurse deeper */
+ if (addr < align_addr)
+ next_addr = align_addr;
+ realm_unmap_range_shared(kvm, level + 1, addr,
+ min(next_addr, end));
+ continue;
+ }
+
+ ret = rmi_rtt_unmap_unprotected(rd, addr, level, &next_addr);
+ switch (RMI_RETURN_STATUS(ret)) {
+ case RMI_SUCCESS:
+ break;
+ case RMI_ERROR_RTT:
+ if (next_addr == addr) {
+ next_addr = ALIGN(addr + 1, map_size);
+ realm_unmap_range_shared(kvm, level + 1, addr,
+ next_addr);
+ }
+ break;
+ default:
+ WARN_ON(1);
+ }
+ }
+}
+
+static void realm_unmap_range_private(struct kvm *kvm,
+ unsigned long start,
+ unsigned long end)
+{
+ struct realm *realm = &kvm->arch.realm;
+ ssize_t map_size = RME_PAGE_SIZE;
+ unsigned long next_addr, addr;
+
+ for (addr = start; addr < end; addr = next_addr) {
+ int ret;
+
+ next_addr = ALIGN(addr + 1, map_size);
+
+ ret = realm_destroy_protected(realm, addr, &next_addr);
+
+ if (WARN_ON(ret))
+ break;
+ }
+}
+
+static void realm_unmap_range(struct kvm *kvm,
+ unsigned long start,
+ unsigned long end,
+ bool unmap_private)
+{
+ realm_unmap_range_shared(kvm, RME_RTT_MAX_LEVEL - 1, start, end);
+ if (unmap_private)
+ realm_unmap_range_private(kvm, start, end);
+}
+
u32 kvm_realm_ipa_limit(void)
{
return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
@@ -190,6 +341,30 @@ static int realm_rtt_destroy(struct realm *realm, unsigned long addr,
return ret;
}
+static int realm_create_rtt_levels(struct realm *realm,
+ unsigned long ipa,
+ int level,
+ int max_level,
+ struct kvm_mmu_memory_cache *mc)
+{
+ if (WARN_ON(level == max_level))
+ return 0;
+
+ while (level++ < max_level) {
+ phys_addr_t rtt = alloc_delegated_page(realm, mc);
+
+ if (rtt == PHYS_ADDR_MAX)
+ return -ENOMEM;
+
+ if (realm_rtt_create(realm, ipa, level, rtt)) {
+ free_delegated_page(realm, rtt);
+ return -ENXIO;
+ }
+ }
+
+ return 0;
+}
+
static int realm_tear_down_rtt_level(struct realm *realm, int level,
unsigned long start, unsigned long end)
{
@@ -265,6 +440,68 @@ static int realm_tear_down_rtt_range(struct realm *realm,
start, end);
}
+/*
+ * Returns 0 on successful fold, a negative value on error, a positive value if
+ * we were not able to fold all tables at this level.
+ */
+static int realm_fold_rtt_level(struct realm *realm, int level,
+ unsigned long start, unsigned long end)
+{
+ int not_folded = 0;
+ ssize_t map_size;
+ unsigned long addr, next_addr;
+
+ if (WARN_ON(level > RME_RTT_MAX_LEVEL))
+ return -EINVAL;
+
+ map_size = rme_rtt_level_mapsize(level - 1);
+
+ for (addr = start; addr < end; addr = next_addr) {
+ phys_addr_t rtt_granule;
+ int ret;
+ unsigned long align_addr = ALIGN(addr, map_size);
+
+ next_addr = ALIGN(addr + 1, map_size);
+
+ ret = realm_rtt_fold(realm, align_addr, level, &rtt_granule);
+
+ switch (RMI_RETURN_STATUS(ret)) {
+ case RMI_SUCCESS:
+ if (!WARN_ON(rmi_granule_undelegate(rtt_granule)))
+ free_page((unsigned long)phys_to_virt(rtt_granule));
+ break;
+ case RMI_ERROR_RTT:
+ if (level == RME_RTT_MAX_LEVEL ||
+ RMI_RETURN_INDEX(ret) < level) {
+ not_folded++;
+ break;
+ }
+ /* Recurse a level deeper */
+ ret = realm_fold_rtt_level(realm,
+ level + 1,
+ addr,
+ next_addr);
+ if (ret < 0)
+ return ret;
+ else if (ret == 0)
+ /* Try again at this level */
+ next_addr = addr;
+ break;
+ default:
+ return -ENXIO;
+ }
+ }
+
+ return not_folded;
+}
+
+static int realm_fold_rtt_range(struct realm *realm,
+ unsigned long start, unsigned long end)
+{
+ return realm_fold_rtt_level(realm, get_start_level(realm) + 1,
+ start, end);
+}
+
static void ensure_spare_page(struct realm *realm)
{
phys_addr_t tmp_rtt;
@@ -295,6 +532,147 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits)
WARN_ON(realm_tear_down_rtt_range(realm, 0, (1UL << ia_bits)));
}
+void kvm_realm_unmap_range(struct kvm *kvm, unsigned long ipa, u64 size,
+ bool unmap_private)
+{
+ unsigned long end = ipa + size;
+ struct realm *realm = &kvm->arch.realm;
+
+ end = min(BIT(realm->ia_bits - 1), end);
+
+ ensure_spare_page(realm);
+
+ realm_unmap_range(kvm, ipa, end, unmap_private);
+
+ realm_fold_rtt_range(realm, ipa, end);
+}
+
+static int find_map_level(struct realm *realm,
+ unsigned long start,
+ unsigned long end)
+{
+ int level = RME_RTT_MAX_LEVEL;
+
+ while (level > get_start_level(realm)) {
+ unsigned long map_size = rme_rtt_level_mapsize(level - 1);
+
+ if (!IS_ALIGNED(start, map_size) ||
+ (start + map_size) > end)
+ break;
+
+ level--;
+ }
+
+ return level;
+}
+
+int realm_set_ipa_state(struct kvm_vcpu *vcpu,
+ unsigned long start,
+ unsigned long end,
+ unsigned long ripas)
+{
+ struct kvm *kvm = vcpu->kvm;
+ struct realm *realm = &kvm->arch.realm;
+ struct realm_rec *rec = &vcpu->arch.rec;
+ phys_addr_t rd_phys = virt_to_phys(realm->rd);
+ phys_addr_t rec_phys = virt_to_phys(rec->rec_page);
+ struct kvm_mmu_memory_cache *memcache = &vcpu->arch.mmu_page_cache;
+ unsigned long ipa = start;
+ int ret = 0;
+
+ while (ipa < end) {
+ unsigned long next;
+
+ ret = rmi_rtt_set_ripas(rd_phys, rec_phys, ipa, end, &next);
+
+ if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
+ int walk_level = RMI_RETURN_INDEX(ret);
+ int level = find_map_level(realm, ipa, end);
+
+ if (walk_level < level) {
+ ret = realm_create_rtt_levels(realm, ipa,
+ walk_level,
+ level,
+ memcache);
+ if (!ret)
+ continue;
+ } else {
+ ret = -EINVAL;
+ }
+
+ break;
+ } else if (RMI_RETURN_STATUS(ret) != RMI_SUCCESS) {
+ WARN(1, "Unexpected error in %s: %#x\n", __func__,
+ ret);
+ ret = -EINVAL;
+ break;
+ }
+ ipa = next;
+ }
+
+ if (ripas == RMI_EMPTY && ipa != start)
+ kvm_realm_unmap_range(kvm, start, ipa - start, true);
+
+ return ret;
+}
+
+static int realm_init_ipa_state(struct realm *realm,
+ unsigned long ipa,
+ unsigned long end)
+{
+ phys_addr_t rd_phys = virt_to_phys(realm->rd);
+ int ret;
+
+ while (ipa < end) {
+ unsigned long next;
+
+ ret = rmi_rtt_init_ripas(rd_phys, ipa, end, &next);
+
+ if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
+ int err_level = RMI_RETURN_INDEX(ret);
+ int level = find_map_level(realm, ipa, end);
+
+ if (WARN_ON(err_level >= level))
+ return -ENXIO;
+
+ ret = realm_create_rtt_levels(realm, ipa,
+ err_level,
+ level, NULL);
+ if (ret)
+ return ret;
+ /* Retry with the RTT levels in place */
+ continue;
+ } else if (WARN_ON(ret)) {
+ return -ENXIO;
+ }
+
+ ipa = next;
+ }
+
+ return 0;
+}
+
+static int kvm_init_ipa_range_realm(struct kvm *kvm,
+ struct kvm_cap_arm_rme_init_ipa_args *args)
+{
+ int ret = 0;
+ gpa_t addr, end;
+ struct realm *realm = &kvm->arch.realm;
+
+ addr = args->init_ipa_base;
+ end = addr + args->init_ipa_size;
+
+ if (end < addr)
+ return -EINVAL;
+
+ if (kvm_realm_state(kvm) != REALM_STATE_NEW)
+ return -EINVAL;
+
+ ret = realm_init_ipa_state(realm, addr, end);
+
+ return ret;
+}
+
/* Protects access to rme_vmid_bitmap */
static DEFINE_SPINLOCK(rme_vmid_lock);
static unsigned long *rme_vmid_bitmap;
@@ -418,6 +796,18 @@ int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
case KVM_CAP_ARM_RME_CREATE_RD:
r = kvm_create_realm(kvm);
break;
+ case KVM_CAP_ARM_RME_INIT_IPA_REALM: {
+ struct kvm_cap_arm_rme_init_ipa_args args;
+ void __user *argp = u64_to_user_ptr(cap->args[1]);
+
+ if (copy_from_user(&args, argp, sizeof(args))) {
+ r = -EFAULT;
+ break;
+ }
+
+ r = kvm_init_ipa_range_realm(kvm, &args);
+ break;
+ }
default:
r = -EINVAL;
break;
--
2.34.1
The VMM needs to populate the realm with some data before starting (e.g.
a kernel and initrd). This is measured by the RMM and used as part of
the attestation later on.
Co-developed-by: Suzuki K Poulose <[email protected]>
Signed-off-by: Suzuki K Poulose <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/rme.c | 234 +++++++++++++++++++++++++++++++++++++++++++
1 file changed, 234 insertions(+)
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 0a3f823b2446..4aab507f896e 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -4,6 +4,7 @@
*/
#include <linux/kvm_host.h>
+#include <linux/hugetlb.h>
#include <asm/kvm_emulate.h>
#include <asm/kvm_mmu.h>
@@ -547,6 +548,227 @@ void kvm_realm_unmap_range(struct kvm *kvm, unsigned long ipa, u64 size,
realm_fold_rtt_range(realm, ipa, end);
}
+static int realm_create_protected_data_page(struct realm *realm,
+ unsigned long ipa,
+ struct page *dst_page,
+ struct page *src_page,
+ unsigned long flags)
+{
+ phys_addr_t dst_phys, src_phys;
+ int ret;
+
+ dst_phys = page_to_phys(dst_page);
+ src_phys = page_to_phys(src_page);
+
+ if (rmi_granule_delegate(dst_phys))
+ return -ENXIO;
+
+ ret = rmi_data_create(virt_to_phys(realm->rd), dst_phys, ipa, src_phys,
+ flags);
+
+ if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
+ /* Create missing RTTs and retry */
+ int level = RMI_RETURN_INDEX(ret);
+
+ ret = realm_create_rtt_levels(realm, ipa, level,
+ RME_RTT_MAX_LEVEL, NULL);
+ if (ret)
+ goto err;
+
+ ret = rmi_data_create(virt_to_phys(realm->rd), dst_phys, ipa,
+ src_phys, flags);
+ }
+
+ if (ret)
+ goto err;
+
+ return 0;
+
+err:
+ if (WARN_ON(rmi_granule_undelegate(dst_phys))) {
+ /* Page can't be returned to NS world so is lost */
+ get_page(dst_page);
+ }
+ return -ENXIO;
+}
+
+static int fold_rtt(struct realm *realm, unsigned long addr, int level)
+{
+ phys_addr_t rtt_addr;
+ int ret;
+
+ ret = realm_rtt_fold(realm, addr, level + 1, &rtt_addr);
+ if (ret)
+ return ret;
+
+ free_delegated_page(realm, rtt_addr);
+
+ return 0;
+}
+
+static int populate_par_region(struct kvm *kvm,
+ phys_addr_t ipa_base,
+ phys_addr_t ipa_end,
+ u32 flags)
+{
+ struct realm *realm = &kvm->arch.realm;
+ struct kvm_memory_slot *memslot;
+ gfn_t base_gfn, end_gfn;
+ int idx;
+ phys_addr_t ipa;
+ int ret = 0;
+ struct page *tmp_page;
+ unsigned long data_flags = 0;
+
+ base_gfn = gpa_to_gfn(ipa_base);
+ end_gfn = gpa_to_gfn(ipa_end);
+
+ if (flags & KVM_ARM_RME_POPULATE_FLAGS_MEASURE)
+ data_flags = RMI_MEASURE_CONTENT;
+
+ idx = srcu_read_lock(&kvm->srcu);
+ memslot = gfn_to_memslot(kvm, base_gfn);
+ if (!memslot) {
+ ret = -EFAULT;
+ goto out;
+ }
+
+ /* We require the region to be contained within a single memslot */
+ if (memslot->base_gfn + memslot->npages < end_gfn) {
+ ret = -EINVAL;
+ goto out;
+ }
+
+ tmp_page = alloc_page(GFP_KERNEL);
+ if (!tmp_page) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ mmap_read_lock(current->mm);
+
+ ipa = ipa_base;
+ while (ipa < ipa_end) {
+ struct vm_area_struct *vma;
+ unsigned long map_size;
+ unsigned int vma_shift;
+ unsigned long offset;
+ unsigned long hva;
+ struct page *page;
+ kvm_pfn_t pfn;
+ int level;
+
+ hva = gfn_to_hva_memslot(memslot, gpa_to_gfn(ipa));
+ vma = vma_lookup(current->mm, hva);
+ if (!vma) {
+ ret = -EFAULT;
+ break;
+ }
+
+ if (is_vm_hugetlb_page(vma))
+ vma_shift = huge_page_shift(hstate_vma(vma));
+ else
+ vma_shift = PAGE_SHIFT;
+
+ map_size = 1 << vma_shift;
+
+ /*
+ * FIXME: This causes over mapping, but there's no good
+ * solution here with the ABI as it stands
+ */
+ ipa = ALIGN_DOWN(ipa, map_size);
+
+ switch (map_size) {
+ case RME_L2_BLOCK_SIZE:
+ level = 2;
+ break;
+ case PAGE_SIZE:
+ level = 3;
+ break;
+ default:
+ WARN_ONCE(1, "Unsupport vma_shift %d", vma_shift);
+ ret = -EFAULT;
+ break;
+ }
+
+ pfn = gfn_to_pfn_memslot(memslot, gpa_to_gfn(ipa));
+
+ if (is_error_pfn(pfn)) {
+ ret = -EFAULT;
+ break;
+ }
+
+ if (level < RME_RTT_MAX_LEVEL) {
+ /*
+ * A temporary RTT is needed during the map, precreate
+ * it, however if there is an error (e.g. missing
+ * parent tables) this will be handled in the
+ * realm_create_protected_data_page() call.
+ */
+ realm_create_rtt_levels(realm, ipa, level,
+ RME_RTT_MAX_LEVEL, NULL);
+ }
+
+ page = pfn_to_page(pfn);
+
+ for (offset = 0; offset < map_size && !ret;
+ offset += PAGE_SIZE, page++) {
+ phys_addr_t page_ipa = ipa + offset;
+
+ ret = realm_create_protected_data_page(realm, page_ipa,
+ page, tmp_page,
+ data_flags);
+ }
+ if (ret)
+ goto err_release_pfn;
+
+ if (level == 2) {
+ ret = fold_rtt(realm, ipa, level);
+ if (ret)
+ goto err_release_pfn;
+ }
+
+ ipa += map_size;
+ kvm_release_pfn_dirty(pfn);
+err_release_pfn:
+ if (ret) {
+ kvm_release_pfn_clean(pfn);
+ break;
+ }
+ }
+
+ mmap_read_unlock(current->mm);
+ __free_page(tmp_page);
+
+out:
+ srcu_read_unlock(&kvm->srcu, idx);
+ return ret;
+}
+
+static int kvm_populate_realm(struct kvm *kvm,
+ struct kvm_cap_arm_rme_populate_realm_args *args)
+{
+ phys_addr_t ipa_base, ipa_end;
+
+ if (kvm_realm_state(kvm) != REALM_STATE_NEW)
+ return -EINVAL;
+
+ if (!IS_ALIGNED(args->populate_ipa_base, PAGE_SIZE) ||
+ !IS_ALIGNED(args->populate_ipa_size, PAGE_SIZE))
+ return -EINVAL;
+
+ if (args->flags & ~RMI_MEASURE_CONTENT)
+ return -EINVAL;
+
+ ipa_base = args->populate_ipa_base;
+ ipa_end = ipa_base + args->populate_ipa_size;
+
+ if (ipa_end < ipa_base)
+ return -EINVAL;
+
+ return populate_par_region(kvm, ipa_base, ipa_end, args->flags);
+}
+
static int find_map_level(struct realm *realm,
unsigned long start,
unsigned long end)
@@ -808,6 +1030,18 @@ int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
r = kvm_init_ipa_range_realm(kvm, &args);
break;
}
+ case KVM_CAP_ARM_RME_POPULATE_REALM: {
+ struct kvm_cap_arm_rme_populate_realm_args args;
+ void __user *argp = u64_to_user_ptr(cap->args[1]);
+
+ if (copy_from_user(&args, argp, sizeof(args))) {
+ r = -EFAULT;
+ break;
+ }
+
+ r = kvm_populate_realm(kvm, &args);
+ break;
+ }
default:
r = -EINVAL;
break;
--
2.34.1
At runtime if the realm guest accesses memory which hasn't yet been
mapped then KVM needs to either populate the region or fault the guest.
For memory in the lower (protected) region of IPA a fresh page is
provided to the RMM which will zero the contents. For memory in the
upper (shared) region of IPA, the memory from the memslot is mapped
into the realm VM non secure.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/include/asm/kvm_emulate.h | 10 ++
arch/arm64/include/asm/kvm_rme.h | 10 ++
arch/arm64/kvm/mmu.c | 119 +++++++++++++++-
arch/arm64/kvm/rme.c | 199 ++++++++++++++++++++++++---
4 files changed, 316 insertions(+), 22 deletions(-)
diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
index 2209a7c6267f..d40d998d9be2 100644
--- a/arch/arm64/include/asm/kvm_emulate.h
+++ b/arch/arm64/include/asm/kvm_emulate.h
@@ -629,6 +629,16 @@ static inline bool kvm_realm_is_created(struct kvm *kvm)
return kvm_is_realm(kvm) && kvm_realm_state(kvm) != REALM_STATE_NONE;
}
+static inline gpa_t kvm_gpa_stolen_bits(struct kvm *kvm)
+{
+ if (kvm_is_realm(kvm)) {
+ struct realm *realm = &kvm->arch.realm;
+
+ return BIT(realm->ia_bits - 1);
+ }
+ return 0;
+}
+
static inline bool vcpu_is_rec(struct kvm_vcpu *vcpu)
{
if (static_branch_unlikely(&kvm_rme_is_available))
diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
index 749f2eb97bd4..48c7766fadeb 100644
--- a/arch/arm64/include/asm/kvm_rme.h
+++ b/arch/arm64/include/asm/kvm_rme.h
@@ -103,6 +103,16 @@ void kvm_realm_unmap_range(struct kvm *kvm,
unsigned long ipa,
u64 size,
bool unmap_private);
+int realm_map_protected(struct realm *realm,
+ unsigned long base_ipa,
+ struct page *dst_page,
+ unsigned long map_size,
+ struct kvm_mmu_memory_cache *memcache);
+int realm_map_non_secure(struct realm *realm,
+ unsigned long ipa,
+ struct page *page,
+ unsigned long map_size,
+ struct kvm_mmu_memory_cache *memcache);
int realm_set_ipa_state(struct kvm_vcpu *vcpu,
unsigned long addr, unsigned long end,
unsigned long ripas);
diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index 8a7b5449697f..50a49e4e2020 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -325,8 +325,13 @@ static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64
lockdep_assert_held_write(&kvm->mmu_lock);
WARN_ON(size & ~PAGE_MASK);
- WARN_ON(stage2_apply_range(mmu, start, end, kvm_pgtable_stage2_unmap,
- may_block));
+
+ if (kvm_is_realm(kvm))
+ kvm_realm_unmap_range(kvm, start, size, !only_shared);
+ else
+ WARN_ON(stage2_apply_range(mmu, start, end,
+ kvm_pgtable_stage2_unmap,
+ may_block));
}
static void unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64 size)
@@ -340,7 +345,11 @@ static void stage2_flush_memslot(struct kvm *kvm,
phys_addr_t addr = memslot->base_gfn << PAGE_SHIFT;
phys_addr_t end = addr + PAGE_SIZE * memslot->npages;
- stage2_apply_range_resched(&kvm->arch.mmu, addr, end, kvm_pgtable_stage2_flush);
+ if (kvm_is_realm(kvm))
+ kvm_realm_unmap_range(kvm, addr, end - addr, false);
+ else
+ stage2_apply_range_resched(&kvm->arch.mmu, addr, end,
+ kvm_pgtable_stage2_flush);
}
/**
@@ -997,6 +1006,10 @@ void stage2_unmap_vm(struct kvm *kvm)
struct kvm_memory_slot *memslot;
int idx, bkt;
+ /* For realms this is handled by the RMM so nothing to do here */
+ if (kvm_is_realm(kvm))
+ return;
+
idx = srcu_read_lock(&kvm->srcu);
mmap_read_lock(current->mm);
write_lock(&kvm->mmu_lock);
@@ -1020,6 +1033,7 @@ void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu)
if (kvm_is_realm(kvm) &&
(kvm_realm_state(kvm) != REALM_STATE_DEAD &&
kvm_realm_state(kvm) != REALM_STATE_NONE)) {
+ unmap_stage2_range(mmu, 0, (~0ULL) & PAGE_MASK);
write_unlock(&kvm->mmu_lock);
kvm_realm_destroy_rtts(kvm, pgt->ia_bits);
return;
@@ -1383,6 +1397,69 @@ static bool kvm_vma_mte_allowed(struct vm_area_struct *vma)
return vma->vm_flags & VM_MTE_ALLOWED;
}
+static int realm_map_ipa(struct kvm *kvm, phys_addr_t ipa,
+ kvm_pfn_t pfn, unsigned long map_size,
+ enum kvm_pgtable_prot prot,
+ struct kvm_mmu_memory_cache *memcache)
+{
+ struct realm *realm = &kvm->arch.realm;
+ struct page *page = pfn_to_page(pfn);
+
+ if (WARN_ON(!(prot & KVM_PGTABLE_PROT_W)))
+ return -EFAULT;
+
+ if (!realm_is_addr_protected(realm, ipa))
+ return realm_map_non_secure(realm, ipa, page, map_size,
+ memcache);
+
+ return realm_map_protected(realm, ipa, page, map_size, memcache);
+}
+
+static int private_memslot_fault(struct kvm_vcpu *vcpu,
+ phys_addr_t fault_ipa,
+ struct kvm_memory_slot *memslot)
+{
+ struct kvm *kvm = vcpu->kvm;
+ gpa_t gpa_stolen_mask = kvm_gpa_stolen_bits(kvm);
+ gfn_t gfn = (fault_ipa & ~gpa_stolen_mask) >> PAGE_SHIFT;
+ bool is_priv_gfn = !((fault_ipa & gpa_stolen_mask) == gpa_stolen_mask);
+ bool priv_exists = kvm_mem_is_private(kvm, gfn);
+ struct kvm_mmu_memory_cache *memcache = &vcpu->arch.mmu_page_cache;
+ int order;
+ kvm_pfn_t pfn;
+ int ret;
+
+ if (priv_exists != is_priv_gfn) {
+ kvm_prepare_memory_fault_exit(vcpu,
+ fault_ipa & ~gpa_stolen_mask,
+ PAGE_SIZE,
+ kvm_is_write_fault(vcpu),
+ false, is_priv_gfn);
+
+ return 0;
+ }
+
+ if (!is_priv_gfn) {
+ /* Not a private mapping, handling normally */
+ return -EAGAIN;
+ }
+
+ if (kvm_gmem_get_pfn(kvm, memslot, gfn, &pfn, &order))
+ return 1; /* Retry */
+
+ ret = kvm_mmu_topup_memory_cache(memcache,
+ kvm_mmu_cache_min_pages(vcpu->arch.hw_mmu));
+ if (ret)
+ return ret;
+
+ /* FIXME: Should be able to use bigger than PAGE_SIZE mappings */
+ ret = realm_map_ipa(kvm, fault_ipa, pfn, PAGE_SIZE, KVM_PGTABLE_PROT_W,
+ memcache);
+ if (!ret)
+ return 1; /* Handled */
+ return ret;
+}
+
static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
struct kvm_memory_slot *memslot, unsigned long hva,
bool fault_is_perm)
@@ -1402,10 +1479,19 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
long vma_pagesize, fault_granule;
enum kvm_pgtable_prot prot = KVM_PGTABLE_PROT_R;
struct kvm_pgtable *pgt;
+ gpa_t gpa_stolen_mask = kvm_gpa_stolen_bits(vcpu->kvm);
if (fault_is_perm)
fault_granule = kvm_vcpu_trap_get_perm_fault_granule(vcpu);
write_fault = kvm_is_write_fault(vcpu);
+
+ /*
+ * Realms cannot map protected pages read-only
+ * FIXME: It should be possible to map unprotected pages read-only
+ */
+ if (vcpu_is_rec(vcpu))
+ write_fault = true;
+
exec_fault = kvm_vcpu_trap_is_exec_fault(vcpu);
VM_BUG_ON(write_fault && exec_fault);
@@ -1478,7 +1564,7 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
if (vma_pagesize == PMD_SIZE || vma_pagesize == PUD_SIZE)
fault_ipa &= ~(vma_pagesize - 1);
- gfn = fault_ipa >> PAGE_SHIFT;
+ gfn = (fault_ipa & ~gpa_stolen_mask) >> PAGE_SHIFT;
mte_allowed = kvm_vma_mte_allowed(vma);
vfio_allow_any_uc = vma->vm_flags & VM_ALLOW_ANY_UNCACHED;
@@ -1538,7 +1624,8 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
* If we are not forced to use page mapping, check if we are
* backed by a THP and thus use block mapping if possible.
*/
- if (vma_pagesize == PAGE_SIZE && !(force_pte || device)) {
+ /* FIXME: We shouldn't need to disable this for realms */
+ if (vma_pagesize == PAGE_SIZE && !(force_pte || device || kvm_is_realm(kvm))) {
if (fault_is_perm && fault_granule > PAGE_SIZE)
vma_pagesize = fault_granule;
else
@@ -1584,6 +1671,9 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
*/
if (fault_is_perm && vma_pagesize == fault_granule)
ret = kvm_pgtable_stage2_relax_perms(pgt, fault_ipa, prot);
+ else if (kvm_is_realm(kvm))
+ ret = realm_map_ipa(kvm, fault_ipa, pfn, vma_pagesize,
+ prot, memcache);
else
ret = kvm_pgtable_stage2_map(pgt, fault_ipa, vma_pagesize,
__pfn_to_phys(pfn), prot,
@@ -1638,6 +1728,7 @@ int kvm_handle_guest_abort(struct kvm_vcpu *vcpu)
struct kvm_memory_slot *memslot;
unsigned long hva;
bool is_iabt, write_fault, writable;
+ gpa_t gpa_stolen_mask = kvm_gpa_stolen_bits(vcpu->kvm);
gfn_t gfn;
int ret, idx;
@@ -1693,8 +1784,15 @@ int kvm_handle_guest_abort(struct kvm_vcpu *vcpu)
idx = srcu_read_lock(&vcpu->kvm->srcu);
- gfn = fault_ipa >> PAGE_SHIFT;
+ gfn = (fault_ipa & ~gpa_stolen_mask) >> PAGE_SHIFT;
memslot = gfn_to_memslot(vcpu->kvm, gfn);
+
+ if (kvm_slot_can_be_private(memslot)) {
+ ret = private_memslot_fault(vcpu, fault_ipa, memslot);
+ if (ret != -EAGAIN)
+ goto out;
+ }
+
hva = gfn_to_hva_memslot_prot(memslot, gfn, &writable);
write_fault = kvm_is_write_fault(vcpu);
if (kvm_is_error_hva(hva) || (write_fault && !writable)) {
@@ -1738,6 +1836,7 @@ int kvm_handle_guest_abort(struct kvm_vcpu *vcpu)
* of the page size.
*/
fault_ipa |= kvm_vcpu_get_hfar(vcpu) & ((1 << 12) - 1);
+ fault_ipa &= ~gpa_stolen_mask;
ret = io_mem_abort(vcpu, fault_ipa);
goto out_unlock;
}
@@ -1819,6 +1918,10 @@ bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
if (!kvm->arch.mmu.pgt)
return false;
+ /* We don't support aging for Realms */
+ if (kvm_is_realm(kvm))
+ return true;
+
return kvm_pgtable_stage2_test_clear_young(kvm->arch.mmu.pgt,
range->start << PAGE_SHIFT,
size, true);
@@ -1831,6 +1934,10 @@ bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
if (!kvm->arch.mmu.pgt)
return false;
+ /* We don't support aging for Realms */
+ if (kvm_is_realm(kvm))
+ return true;
+
return kvm_pgtable_stage2_test_clear_young(kvm->arch.mmu.pgt,
range->start << PAGE_SHIFT,
size, false);
diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
index 4aab507f896e..72f6f5f542c4 100644
--- a/arch/arm64/kvm/rme.c
+++ b/arch/arm64/kvm/rme.c
@@ -606,6 +606,170 @@ static int fold_rtt(struct realm *realm, unsigned long addr, int level)
return 0;
}
+int realm_map_protected(struct realm *realm,
+ unsigned long base_ipa,
+ struct page *dst_page,
+ unsigned long map_size,
+ struct kvm_mmu_memory_cache *memcache)
+{
+ phys_addr_t dst_phys = page_to_phys(dst_page);
+ phys_addr_t rd = virt_to_phys(realm->rd);
+ unsigned long phys = dst_phys;
+ unsigned long ipa = base_ipa;
+ unsigned long size;
+ int map_level;
+ int ret = 0;
+
+ if (WARN_ON(!IS_ALIGNED(ipa, map_size)))
+ return -EINVAL;
+
+ switch (map_size) {
+ case PAGE_SIZE:
+ map_level = 3;
+ break;
+ case RME_L2_BLOCK_SIZE:
+ map_level = 2;
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ if (map_level < RME_RTT_MAX_LEVEL) {
+ /*
+ * A temporary RTT is needed during the map, precreate it,
+ * however if there is an error (e.g. missing parent tables)
+ * this will be handled below.
+ */
+ realm_create_rtt_levels(realm, ipa, map_level,
+ RME_RTT_MAX_LEVEL, memcache);
+ }
+
+ for (size = 0; size < map_size; size += PAGE_SIZE) {
+ if (rmi_granule_delegate(phys)) {
+ struct rtt_entry rtt;
+
+ /*
+ * It's possible we raced with another VCPU on the same
+ * fault. If the entry exists and matches then exit
+ * early and assume the other VCPU will handle the
+ * mapping.
+ */
+ if (rmi_rtt_read_entry(rd, ipa, RME_RTT_MAX_LEVEL, &rtt))
+ goto err;
+
+ // FIXME: For a block mapping this could race at level
+ // 2 or 3...
+ if (WARN_ON((rtt.walk_level != RME_RTT_MAX_LEVEL ||
+ rtt.state != RMI_ASSIGNED ||
+ rtt.desc != phys))) {
+ goto err;
+ }
+
+ return 0;
+ }
+
+ ret = rmi_data_create_unknown(rd, phys, ipa);
+
+ if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
+ /* Create missing RTTs and retry */
+ int level = RMI_RETURN_INDEX(ret);
+
+ ret = realm_create_rtt_levels(realm, ipa, level,
+ RME_RTT_MAX_LEVEL,
+ memcache);
+ WARN_ON(ret);
+ if (ret)
+ goto err_undelegate;
+
+ ret = rmi_data_create_unknown(rd, phys, ipa);
+ }
+ WARN_ON(ret);
+
+ if (ret)
+ goto err_undelegate;
+
+ phys += PAGE_SIZE;
+ ipa += PAGE_SIZE;
+ }
+
+ if (map_size == RME_L2_BLOCK_SIZE)
+ ret = fold_rtt(realm, base_ipa, map_level);
+ if (WARN_ON(ret))
+ goto err;
+
+ return 0;
+
+err_undelegate:
+ if (WARN_ON(rmi_granule_undelegate(phys))) {
+ /* Page can't be returned to NS world so is lost */
+ get_page(phys_to_page(phys));
+ }
+err:
+ while (size > 0) {
+ unsigned long data, top;
+
+ phys -= PAGE_SIZE;
+ size -= PAGE_SIZE;
+ ipa -= PAGE_SIZE;
+
+ WARN_ON(rmi_data_destroy(rd, ipa, &data, &top));
+
+ if (WARN_ON(rmi_granule_undelegate(phys))) {
+ /* Page can't be returned to NS world so is lost */
+ get_page(phys_to_page(phys));
+ }
+ }
+ return -ENXIO;
+}
+
+int realm_map_non_secure(struct realm *realm,
+ unsigned long ipa,
+ struct page *page,
+ unsigned long map_size,
+ struct kvm_mmu_memory_cache *memcache)
+{
+ phys_addr_t rd = virt_to_phys(realm->rd);
+ int map_level;
+ int ret = 0;
+ unsigned long desc = page_to_phys(page) |
+ PTE_S2_MEMATTR(MT_S2_FWB_NORMAL) |
+ /* FIXME: Read+Write permissions for now */
+ (3 << 6) |
+ PTE_SHARED;
+
+ if (WARN_ON(!IS_ALIGNED(ipa, map_size)))
+ return -EINVAL;
+
+ switch (map_size) {
+ case PAGE_SIZE:
+ map_level = 3;
+ break;
+ case RME_L2_BLOCK_SIZE:
+ map_level = 2;
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ ret = rmi_rtt_map_unprotected(rd, ipa, map_level, desc);
+
+ if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
+ /* Create missing RTTs and retry */
+ int level = RMI_RETURN_INDEX(ret);
+
+ ret = realm_create_rtt_levels(realm, ipa, level, map_level,
+ memcache);
+ if (WARN_ON(ret))
+ return -ENXIO;
+
+ ret = rmi_rtt_map_unprotected(rd, ipa, map_level, desc);
+ }
+ if (WARN_ON(ret))
+ return -ENXIO;
+
+ return 0;
+}
+
static int populate_par_region(struct kvm *kvm,
phys_addr_t ipa_base,
phys_addr_t ipa_end,
@@ -617,7 +781,6 @@ static int populate_par_region(struct kvm *kvm,
int idx;
phys_addr_t ipa;
int ret = 0;
- struct page *tmp_page;
unsigned long data_flags = 0;
base_gfn = gpa_to_gfn(ipa_base);
@@ -639,9 +802,8 @@ static int populate_par_region(struct kvm *kvm,
goto out;
}
- tmp_page = alloc_page(GFP_KERNEL);
- if (!tmp_page) {
- ret = -ENOMEM;
+ if (!kvm_slot_can_be_private(memslot)) {
+ ret = -EINVAL;
goto out;
}
@@ -714,31 +876,36 @@ static int populate_par_region(struct kvm *kvm,
for (offset = 0; offset < map_size && !ret;
offset += PAGE_SIZE, page++) {
phys_addr_t page_ipa = ipa + offset;
+ kvm_pfn_t priv_pfn;
+ int order;
- ret = realm_create_protected_data_page(realm, page_ipa,
- page, tmp_page,
- data_flags);
+ ret = kvm_gmem_get_pfn(kvm, memslot,
+ page_ipa >> PAGE_SHIFT,
+ &priv_pfn, &order);
+ if (ret)
+ break;
+
+ ret = realm_create_protected_data_page(
+ realm, page_ipa,
+ pfn_to_page(priv_pfn),
+ page, data_flags);
}
+
+ kvm_release_pfn_clean(pfn);
+
if (ret)
- goto err_release_pfn;
+ break;
if (level == 2) {
ret = fold_rtt(realm, ipa, level);
if (ret)
- goto err_release_pfn;
+ break;
}
ipa += map_size;
- kvm_release_pfn_dirty(pfn);
-err_release_pfn:
- if (ret) {
- kvm_release_pfn_clean(pfn);
- break;
- }
}
mmap_read_unlock(current->mm);
- __free_page(tmp_page);
out:
srcu_read_unlock(&kvm->srcu, idx);
--
2.34.1
From: Joey Gouly <[email protected]>
Extend KVM_SET_VCPU_EVENTS to support realms, where KVM cannot set the
system registers, and the RMM must perform it on next REC entry.
Signed-off-by: Joey Gouly <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
Documentation/virt/kvm/api.rst | 2 ++
arch/arm64/kvm/guest.c | 24 ++++++++++++++++++++++++
2 files changed, 26 insertions(+)
diff --git a/Documentation/virt/kvm/api.rst b/Documentation/virt/kvm/api.rst
index b4bd3d0928a2..0b2386ee4f15 100644
--- a/Documentation/virt/kvm/api.rst
+++ b/Documentation/virt/kvm/api.rst
@@ -1278,6 +1278,8 @@ User space may need to inject several types of events to the guest.
Set the pending SError exception state for this VCPU. It is not possible to
'cancel' an Serror that has been made pending.
+User space cannot inject SErrors into Realms.
+
If the guest performed an access to I/O memory which could not be handled by
userspace, for example because of missing instruction syndrome decode
information or because there is no device mapped at the accessed IPA, then
diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c
index a035689dc39b..5223a828a344 100644
--- a/arch/arm64/kvm/guest.c
+++ b/arch/arm64/kvm/guest.c
@@ -865,6 +865,30 @@ int __kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
bool has_esr = events->exception.serror_has_esr;
bool ext_dabt_pending = events->exception.ext_dabt_pending;
+ if (vcpu_is_rec(vcpu)) {
+ /* Cannot inject SError into a Realm. */
+ if (serror_pending)
+ return -EINVAL;
+
+ /*
+ * If a data abort is pending, set the flag and let the RMM
+ * inject an SEA when the REC is scheduled to be run.
+ */
+ if (ext_dabt_pending) {
+ /*
+ * Can only inject SEA into a Realm if the previous exit
+ * was due to a data abort of an Unprotected IPA.
+ */
+ if (!(vcpu->arch.rec.run->entry.flags & RMI_EMULATED_MMIO))
+ return -EINVAL;
+
+ vcpu->arch.rec.run->entry.flags &= ~RMI_EMULATED_MMIO;
+ vcpu->arch.rec.run->entry.flags |= RMI_INJECT_SEA;
+ }
+
+ return 0;
+ }
+
if (serror_pending && has_esr) {
if (!cpus_have_final_cap(ARM64_HAS_RAS_EXTN))
return -EINVAL;
--
2.34.1
It doesn't make much sense and with the ABI as it is it's a footgun for
the VMM which makes fatal granule protection faults easy to trigger.
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/arm.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index a0dcae0391d0..f2279ab45add 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -339,7 +339,10 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
r = system_supports_mte();
break;
case KVM_CAP_STEAL_TIME:
- r = kvm_arm_pvtime_supported();
+ if (kvm && kvm_is_realm(kvm))
+ r = 0;
+ else
+ r = kvm_arm_pvtime_supported();
break;
case KVM_CAP_ARM_EL1_32BIT:
r = cpus_have_final_cap(ARM64_HAS_32BIT_EL1);
--
2.34.1
From: Jean-Philippe Brucker <[email protected]>
Userspace can set a few registers with KVM_SET_ONE_REG (9 GP registers
at runtime, and 3 system registers during initialization). Update the
register list returned by KVM_GET_REG_LIST.
Signed-off-by: Jean-Philippe Brucker <[email protected]>
Signed-off-by: Steven Price <[email protected]>
---
arch/arm64/kvm/guest.c | 40 ++++++++++++++++++-------
arch/arm64/kvm/hypercalls.c | 4 +--
arch/arm64/kvm/sys_regs.c | 58 ++++++++++++++++++++++++++++---------
3 files changed, 75 insertions(+), 27 deletions(-)
diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c
index f1fe51775649..b6d64a03024c 100644
--- a/arch/arm64/kvm/guest.c
+++ b/arch/arm64/kvm/guest.c
@@ -73,6 +73,17 @@ static u64 core_reg_offset_from_id(u64 id)
return id & ~(KVM_REG_ARCH_MASK | KVM_REG_SIZE_MASK | KVM_REG_ARM_CORE);
}
+static bool kvm_realm_validate_core_reg(u64 off)
+{
+ switch (off) {
+ case KVM_REG_ARM_CORE_REG(regs.regs[0]) ...
+ KVM_REG_ARM_CORE_REG(regs.regs[7]):
+ case KVM_REG_ARM_CORE_REG(regs.pc):
+ return true;
+ }
+ return false;
+}
+
static int core_reg_size_from_offset(const struct kvm_vcpu *vcpu, u64 off)
{
int size;
@@ -115,6 +126,9 @@ static int core_reg_size_from_offset(const struct kvm_vcpu *vcpu, u64 off)
if (vcpu_has_sve(vcpu) && core_reg_offset_is_vreg(off))
return -EINVAL;
+ if (kvm_is_realm(vcpu->kvm) && !kvm_realm_validate_core_reg(off))
+ return -EPERM;
+
return size;
}
@@ -599,8 +613,6 @@ static const u64 timer_reg_list[] = {
KVM_REG_ARM_PTIMER_CVAL,
};
-#define NUM_TIMER_REGS ARRAY_SIZE(timer_reg_list)
-
static bool is_timer_reg(u64 index)
{
switch (index) {
@@ -615,9 +627,14 @@ static bool is_timer_reg(u64 index)
return false;
}
+static unsigned long num_timer_regs(struct kvm_vcpu *vcpu)
+{
+ return kvm_is_realm(vcpu->kvm) ? 0 : ARRAY_SIZE(timer_reg_list);
+}
+
static int copy_timer_indices(struct kvm_vcpu *vcpu, u64 __user *uindices)
{
- for (int i = 0; i < NUM_TIMER_REGS; i++) {
+ for (int i = 0; i < num_timer_regs(vcpu); i++) {
if (put_user(timer_reg_list[i], uindices))
return -EFAULT;
uindices++;
@@ -655,6 +672,9 @@ static unsigned long num_sve_regs(const struct kvm_vcpu *vcpu)
if (!vcpu_has_sve(vcpu) || !kvm_arm_vcpu_sve_finalized(vcpu))
return 0;
+ if (kvm_is_realm(vcpu->kvm))
+ return 1; /* KVM_REG_ARM64_SVE_VLS */
+
return slices * (SVE_NUM_PREGS + SVE_NUM_ZREGS + 1 /* FFR */)
+ 1; /* KVM_REG_ARM64_SVE_VLS */
}
@@ -682,6 +702,9 @@ static int copy_sve_reg_indices(const struct kvm_vcpu *vcpu,
return -EFAULT;
++num_regs;
+ if (kvm_is_realm(vcpu->kvm))
+ return num_regs;
+
for (i = 0; i < slices; i++) {
for (n = 0; n < SVE_NUM_ZREGS; n++) {
reg = KVM_REG_ARM64_SVE_ZREG(n, i);
@@ -720,7 +743,7 @@ unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu)
res += num_sve_regs(vcpu);
res += kvm_arm_num_sys_reg_descs(vcpu);
res += kvm_arm_get_fw_num_regs(vcpu);
- res += NUM_TIMER_REGS;
+ res += num_timer_regs(vcpu);
return res;
}
@@ -754,7 +777,7 @@ int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices)
ret = copy_timer_indices(vcpu, uindices);
if (ret < 0)
return ret;
- uindices += NUM_TIMER_REGS;
+ uindices += num_timer_regs(vcpu);
return kvm_arm_copy_sys_reg_indices(vcpu, uindices);
}
@@ -794,12 +817,7 @@ static bool validate_realm_set_reg(struct kvm_vcpu *vcpu,
if ((reg->id & KVM_REG_ARM_COPROC_MASK) == KVM_REG_ARM_CORE) {
u64 off = core_reg_offset_from_id(reg->id);
- switch (off) {
- case KVM_REG_ARM_CORE_REG(regs.regs[0]) ...
- KVM_REG_ARM_CORE_REG(regs.regs[7]):
- case KVM_REG_ARM_CORE_REG(regs.pc):
- return true;
- }
+ return kvm_realm_validate_core_reg(off);
} else {
switch (reg->id) {
case KVM_REG_ARM_PMCR_EL0:
diff --git a/arch/arm64/kvm/hypercalls.c b/arch/arm64/kvm/hypercalls.c
index 5763d979d8ca..28b4166cf234 100644
--- a/arch/arm64/kvm/hypercalls.c
+++ b/arch/arm64/kvm/hypercalls.c
@@ -407,14 +407,14 @@ void kvm_arm_teardown_hypercalls(struct kvm *kvm)
int kvm_arm_get_fw_num_regs(struct kvm_vcpu *vcpu)
{
- return ARRAY_SIZE(kvm_arm_fw_reg_ids);
+ return kvm_is_realm(vcpu->kvm) ? 0 : ARRAY_SIZE(kvm_arm_fw_reg_ids);
}
int kvm_arm_copy_fw_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices)
{
int i;
- for (i = 0; i < ARRAY_SIZE(kvm_arm_fw_reg_ids); i++) {
+ for (i = 0; i < kvm_arm_get_fw_num_regs(vcpu); i++) {
if (put_user(kvm_arm_fw_reg_ids[i], uindices++))
return -EFAULT;
}
diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c
index 5703b63186fd..f1239a8c041b 100644
--- a/arch/arm64/kvm/sys_regs.c
+++ b/arch/arm64/kvm/sys_regs.c
@@ -3869,18 +3869,18 @@ int kvm_arm_sys_reg_set_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg
sys_reg_descs, ARRAY_SIZE(sys_reg_descs));
}
-static unsigned int num_demux_regs(void)
+static unsigned int num_demux_regs(struct kvm_vcpu *vcpu)
{
- return CSSELR_MAX;
+ return kvm_is_realm(vcpu->kvm) ? 0 : CSSELR_MAX;
}
-static int write_demux_regids(u64 __user *uindices)
+static int write_demux_regids(struct kvm_vcpu *vcpu, u64 __user *uindices)
{
u64 val = KVM_REG_ARM64 | KVM_REG_SIZE_U32 | KVM_REG_ARM_DEMUX;
unsigned int i;
val |= KVM_REG_ARM_DEMUX_ID_CCSIDR;
- for (i = 0; i < CSSELR_MAX; i++) {
+ for (i = 0; i < num_demux_regs(vcpu); i++) {
if (put_user(val | i, uindices))
return -EFAULT;
uindices++;
@@ -3888,6 +3888,23 @@ static int write_demux_regids(u64 __user *uindices)
return 0;
}
+static unsigned int num_invariant_regs(struct kvm_vcpu *vcpu)
+{
+ return kvm_is_realm(vcpu->kvm) ? 0 : ARRAY_SIZE(invariant_sys_regs);
+}
+
+static int write_invariant_regids(struct kvm_vcpu *vcpu, u64 __user *uindices)
+{
+ unsigned int i;
+
+ for (i = 0; i < num_invariant_regs(vcpu); i++) {
+ if (put_user(sys_reg_to_index(&invariant_sys_regs[i]), uindices))
+ return -EFAULT;
+ uindices++;
+ }
+ return 0;
+}
+
static u64 sys_reg_to_index(const struct sys_reg_desc *reg)
{
return (KVM_REG_ARM64 | KVM_REG_SIZE_U64 |
@@ -3911,11 +3928,27 @@ static bool copy_reg_to_user(const struct sys_reg_desc *reg, u64 __user **uind)
return true;
}
+static bool kvm_realm_sys_reg_hidden_user(const struct kvm_vcpu *vcpu, u64 reg)
+{
+ if (!kvm_is_realm(vcpu->kvm))
+ return false;
+
+ switch (reg) {
+ case SYS_ID_AA64DFR0_EL1:
+ case SYS_PMCR_EL0:
+ return false;
+ }
+ return true;
+}
+
static int walk_one_sys_reg(const struct kvm_vcpu *vcpu,
const struct sys_reg_desc *rd,
u64 __user **uind,
unsigned int *total)
{
+ if (kvm_realm_sys_reg_hidden_user(vcpu, reg_to_encoding(rd)))
+ return 0;
+
/*
* Ignore registers we trap but don't save,
* and for which no custom user accessor is provided.
@@ -3953,29 +3986,26 @@ static int walk_sys_regs(struct kvm_vcpu *vcpu, u64 __user *uind)
unsigned long kvm_arm_num_sys_reg_descs(struct kvm_vcpu *vcpu)
{
- return ARRAY_SIZE(invariant_sys_regs)
- + num_demux_regs()
+ return num_invariant_regs(vcpu)
+ + num_demux_regs(vcpu)
+ walk_sys_regs(vcpu, (u64 __user *)NULL);
}
int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices)
{
- unsigned int i;
int err;
- /* Then give them all the invariant registers' indices. */
- for (i = 0; i < ARRAY_SIZE(invariant_sys_regs); i++) {
- if (put_user(sys_reg_to_index(&invariant_sys_regs[i]), uindices))
- return -EFAULT;
- uindices++;
- }
+ err = write_invariant_regids(vcpu, uindices);
+ if (err)
+ return err;
+ uindices += num_invariant_regs(vcpu);
err = walk_sys_regs(vcpu, uindices);
if (err < 0)
return err;
uindices += err;
- return write_demux_regids(uindices);
+ return write_demux_regids(vcpu, uindices);
}
#define KVM_ARM_FEATURE_ID_RANGE_INDEX(r) \
--
2.34.1
Hi Steven,
kernel test robot noticed the following build errors:
[auto build test ERROR on kvmarm/next]
[also build test ERROR on kvm/queue arm64/for-next/core linus/master v6.9-rc3 next-20240412]
[cannot apply to kvm/linux-next]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Steven-Price/KVM-Prepare-for-handling-only-shared-mappings-in-mmu_notifier-events/20240412-170311
base: https://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm.git next
patch link: https://lore.kernel.org/r/20240412084309.1733783-34-steven.price%40arm.com
patch subject: [PATCH v2 33/43] arm64: rme: Enable PMU support with a realm guest
config: arm64-randconfig-r064-20240414 (https://download.01.org/0day-ci/archive/20240414/[email protected]/config)
compiler: clang version 19.0.0git (https://github.com/llvm/llvm-project 8b3b4a92adee40483c27f26c478a384cd69c6f05)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240414/[email protected]/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
All errors (new ones prefixed by >>):
In file included from arch/arm64/kvm/arm.c:9:
In file included from include/linux/entry-kvm.h:6:
In file included from include/linux/resume_user_mode.h:8:
In file included from include/linux/memcontrol.h:21:
In file included from include/linux/mm.h:2208:
include/linux/vmstat.h:508:43: warning: arithmetic between different enumeration types ('enum zone_stat_item' and 'enum numa_stat_item') [-Wenum-enum-conversion]
508 | return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
| ~~~~~~~~~~~~~~~~~~~~~ ^
509 | item];
| ~~~~
include/linux/vmstat.h:515:43: warning: arithmetic between different enumeration types ('enum zone_stat_item' and 'enum numa_stat_item') [-Wenum-enum-conversion]
515 | return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
| ~~~~~~~~~~~~~~~~~~~~~ ^
516 | NR_VM_NUMA_EVENT_ITEMS +
| ~~~~~~~~~~~~~~~~~~~~~~
include/linux/vmstat.h:522:36: warning: arithmetic between different enumeration types ('enum node_stat_item' and 'enum lru_list') [-Wenum-enum-conversion]
522 | return node_stat_name(NR_LRU_BASE + lru) + 3; // skip "nr_"
| ~~~~~~~~~~~ ^ ~~~
include/linux/vmstat.h:527:43: warning: arithmetic between different enumeration types ('enum zone_stat_item' and 'enum numa_stat_item') [-Wenum-enum-conversion]
527 | return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
| ~~~~~~~~~~~~~~~~~~~~~ ^
528 | NR_VM_NUMA_EVENT_ITEMS +
| ~~~~~~~~~~~~~~~~~~~~~~
include/linux/vmstat.h:536:43: warning: arithmetic between different enumeration types ('enum zone_stat_item' and 'enum numa_stat_item') [-Wenum-enum-conversion]
536 | return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
| ~~~~~~~~~~~~~~~~~~~~~ ^
537 | NR_VM_NUMA_EVENT_ITEMS +
| ~~~~~~~~~~~~~~~~~~~~~~
>> arch/arm64/kvm/arm.c:1115:13: error: no member named 'irq_level' in 'struct kvm_pmu'
1115 | if (pmu->irq_level) {
| ~~~ ^
>> arch/arm64/kvm/arm.c:1117:5: error: call to undeclared function 'arm_pmu_set_phys_irq'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
1117 | arm_pmu_set_phys_irq(false);
| ^
arch/arm64/kvm/arm.c:1224:4: error: call to undeclared function 'arm_pmu_set_phys_irq'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
1224 | arm_pmu_set_phys_irq(true);
| ^
5 warnings and 3 errors generated.
vim +1115 arch/arm64/kvm/arm.c
1044
1045 /**
1046 * kvm_arch_vcpu_ioctl_run - the main VCPU run function to execute guest code
1047 * @vcpu: The VCPU pointer
1048 *
1049 * This function is called through the VCPU_RUN ioctl called from user space. It
1050 * will execute VM code in a loop until the time slice for the process is used
1051 * or some emulation is needed from user space in which case the function will
1052 * return with return value 0 and with the kvm_run structure filled in with the
1053 * required data for the requested emulation.
1054 */
1055 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
1056 {
1057 struct kvm_run *run = vcpu->run;
1058 int ret;
1059
1060 if (run->exit_reason == KVM_EXIT_MMIO) {
1061 ret = kvm_handle_mmio_return(vcpu);
1062 if (ret)
1063 return ret;
1064 }
1065
1066 vcpu_load(vcpu);
1067
1068 if (run->immediate_exit) {
1069 ret = -EINTR;
1070 goto out;
1071 }
1072
1073 kvm_sigset_activate(vcpu);
1074
1075 ret = 1;
1076 run->exit_reason = KVM_EXIT_UNKNOWN;
1077 run->flags = 0;
1078 while (ret > 0) {
1079 bool pmu_stopped = false;
1080
1081 /*
1082 * Check conditions before entering the guest
1083 */
1084 ret = xfer_to_guest_mode_handle_work(vcpu);
1085 if (!ret)
1086 ret = 1;
1087
1088 if (ret > 0)
1089 ret = check_vcpu_requests(vcpu);
1090
1091 /*
1092 * Preparing the interrupts to be injected also
1093 * involves poking the GIC, which must be done in a
1094 * non-preemptible context.
1095 */
1096 preempt_disable();
1097
1098 /*
1099 * The VMID allocator only tracks active VMIDs per
1100 * physical CPU, and therefore the VMID allocated may not be
1101 * preserved on VMID roll-over if the task was preempted,
1102 * making a thread's VMID inactive. So we need to call
1103 * kvm_arm_vmid_update() in non-premptible context.
1104 */
1105 if (kvm_arm_vmid_update(&vcpu->arch.hw_mmu->vmid) &&
1106 has_vhe())
1107 __load_stage2(vcpu->arch.hw_mmu,
1108 vcpu->arch.hw_mmu->arch);
1109
1110 kvm_pmu_flush_hwstate(vcpu);
1111
1112 if (vcpu_is_rec(vcpu)) {
1113 struct kvm_pmu *pmu = &vcpu->arch.pmu;
1114
> 1115 if (pmu->irq_level) {
1116 pmu_stopped = true;
> 1117 arm_pmu_set_phys_irq(false);
1118 }
1119 }
1120
1121 local_irq_disable();
1122
1123 kvm_vgic_flush_hwstate(vcpu);
1124
1125 kvm_pmu_update_vcpu_events(vcpu);
1126
1127 /*
1128 * Ensure we set mode to IN_GUEST_MODE after we disable
1129 * interrupts and before the final VCPU requests check.
1130 * See the comment in kvm_vcpu_exiting_guest_mode() and
1131 * Documentation/virt/kvm/vcpu-requests.rst
1132 */
1133 smp_store_mb(vcpu->mode, IN_GUEST_MODE);
1134
1135 if (ret <= 0 || kvm_vcpu_exit_request(vcpu, &ret)) {
1136 vcpu->mode = OUTSIDE_GUEST_MODE;
1137 isb(); /* Ensure work in x_flush_hwstate is committed */
1138 kvm_pmu_sync_hwstate(vcpu);
1139 if (static_branch_unlikely(&userspace_irqchip_in_use))
1140 kvm_timer_sync_user(vcpu);
1141 kvm_vgic_sync_hwstate(vcpu);
1142 local_irq_enable();
1143 preempt_enable();
1144 continue;
1145 }
1146
1147 kvm_arm_setup_debug(vcpu);
1148 kvm_arch_vcpu_ctxflush_fp(vcpu);
1149
1150 /**************************************************************
1151 * Enter the guest
1152 */
1153 trace_kvm_entry(*vcpu_pc(vcpu));
1154 guest_timing_enter_irqoff();
1155
1156 if (vcpu_is_rec(vcpu))
1157 ret = kvm_rec_enter(vcpu);
1158 else
1159 ret = kvm_arm_vcpu_enter_exit(vcpu);
1160
1161 vcpu->mode = OUTSIDE_GUEST_MODE;
1162 vcpu->stat.exits++;
1163 /*
1164 * Back from guest
1165 *************************************************************/
1166
1167 kvm_arm_clear_debug(vcpu);
1168
1169 /*
1170 * We must sync the PMU state before the vgic state so
1171 * that the vgic can properly sample the updated state of the
1172 * interrupt line.
1173 */
1174 kvm_pmu_sync_hwstate(vcpu);
1175
1176 /*
1177 * Sync the vgic state before syncing the timer state because
1178 * the timer code needs to know if the virtual timer
1179 * interrupts are active.
1180 */
1181 kvm_vgic_sync_hwstate(vcpu);
1182
1183 /*
1184 * Sync the timer hardware state before enabling interrupts as
1185 * we don't want vtimer interrupts to race with syncing the
1186 * timer virtual interrupt state.
1187 */
1188 if (static_branch_unlikely(&userspace_irqchip_in_use))
1189 kvm_timer_sync_user(vcpu);
1190
1191 kvm_arch_vcpu_ctxsync_fp(vcpu);
1192
1193 /*
1194 * We must ensure that any pending interrupts are taken before
1195 * we exit guest timing so that timer ticks are accounted as
1196 * guest time. Transiently unmask interrupts so that any
1197 * pending interrupts are taken.
1198 *
1199 * Per ARM DDI 0487G.b section D1.13.4, an ISB (or other
1200 * context synchronization event) is necessary to ensure that
1201 * pending interrupts are taken.
1202 */
1203 if (ARM_EXCEPTION_CODE(ret) == ARM_EXCEPTION_IRQ) {
1204 local_irq_enable();
1205 isb();
1206 local_irq_disable();
1207 }
1208
1209 guest_timing_exit_irqoff();
1210
1211 local_irq_enable();
1212
1213 /* Exit types that need handling before we can be preempted */
1214 if (!vcpu_is_rec(vcpu)) {
1215 trace_kvm_exit(ret, kvm_vcpu_trap_get_class(vcpu),
1216 *vcpu_pc(vcpu));
1217
1218 handle_exit_early(vcpu, ret);
1219 }
1220
1221 preempt_enable();
1222
1223 if (pmu_stopped)
1224 arm_pmu_set_phys_irq(true);
1225
1226 /*
1227 * The ARMv8 architecture doesn't give the hypervisor
1228 * a mechanism to prevent a guest from dropping to AArch32 EL0
1229 * if implemented by the CPU. If we spot the guest in such
1230 * state and that we decided it wasn't supposed to do so (like
1231 * with the asymmetric AArch32 case), return to userspace with
1232 * a fatal error.
1233 */
1234 if (vcpu_mode_is_bad_32bit(vcpu)) {
1235 /*
1236 * As we have caught the guest red-handed, decide that
1237 * it isn't fit for purpose anymore by making the vcpu
1238 * invalid. The VMM can try and fix it by issuing a
1239 * KVM_ARM_VCPU_INIT if it really wants to.
1240 */
1241 vcpu_clear_flag(vcpu, VCPU_INITIALIZED);
1242 ret = ARM_EXCEPTION_IL;
1243 }
1244
1245 if (vcpu_is_rec(vcpu))
1246 ret = handle_rme_exit(vcpu, ret);
1247 else
1248 ret = handle_exit(vcpu, ret);
1249 }
1250
1251 /* Tell userspace about in-kernel device output levels */
1252 if (unlikely(!irqchip_in_kernel(vcpu->kvm))) {
1253 kvm_timer_update_run(vcpu);
1254 kvm_pmu_update_run(vcpu);
1255 }
1256
1257 kvm_sigset_deactivate(vcpu);
1258
1259 out:
1260 /*
1261 * In the unlikely event that we are returning to userspace
1262 * with pending exceptions or PC adjustment, commit these
1263 * adjustments in order to give userspace a consistent view of
1264 * the vcpu state. Note that this relies on __kvm_adjust_pc()
1265 * being preempt-safe on VHE.
1266 */
1267 if (unlikely(vcpu_get_flag(vcpu, PENDING_EXCEPTION) ||
1268 vcpu_get_flag(vcpu, INCREMENT_PC)))
1269 kvm_call_hyp(__kvm_adjust_pc, vcpu);
1270
1271 vcpu_put(vcpu);
1272 return ret;
1273 }
1274
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
On 12/04/2024 09:42, Steven Price wrote:
> If the host attempts to access granules that have been delegated for use
> in a realm these accesses will be caught and will trigger a Granule
> Protection Fault (GPF).
>
> A fault during a page walk signals a bug in the kernel and is handled by
> oopsing the kernel. A non-page walk fault could be caused by user space
> having access to a page which has been delegated to the kernel and will
> trigger a SIGBUS to allow debugging why user space is trying to access a
> delegated page.
>
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/mm/fault.c | 29 ++++++++++++++++++++++++-----
> 1 file changed, 24 insertions(+), 5 deletions(-)
>
> diff --git a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c
> index 8251e2fea9c7..91da0f446dd9 100644
> --- a/arch/arm64/mm/fault.c
> +++ b/arch/arm64/mm/fault.c
> @@ -765,6 +765,25 @@ static int do_tag_check_fault(unsigned long far, unsigned long esr,
> return 0;
> }
>
> +static int do_gpf_ptw(unsigned long far, unsigned long esr, struct pt_regs *regs)
> +{
> + const struct fault_info *inf = esr_to_fault_info(esr);
> +
> + die_kernel_fault(inf->name, far, esr, regs);
> + return 0;
> +}
> +
> +static int do_gpf(unsigned long far, unsigned long esr, struct pt_regs *regs)
> +{
> + const struct fault_info *inf = esr_to_fault_info(esr);
> +
> + if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
> + return 0;
> +
> + arm64_notify_die(inf->name, regs, inf->sig, inf->code, far, esr);
> + return 0;
> +}
> +
> static const struct fault_info fault_info[] = {
> { do_bad, SIGKILL, SI_KERNEL, "ttbr address size fault" },
> { do_bad, SIGKILL, SI_KERNEL, "level 1 address size fault" },
> @@ -802,11 +821,11 @@ static const struct fault_info fault_info[] = {
> { do_alignment_fault, SIGBUS, BUS_ADRALN, "alignment fault" },
> { do_bad, SIGKILL, SI_KERNEL, "unknown 34" },
> { do_bad, SIGKILL, SI_KERNEL, "unknown 35" },
Should this also be converted to do_gpf_ptw, "GPF at level -1", given we
support LPA2 ?
> - { do_bad, SIGKILL, SI_KERNEL, "unknown 36" },
> - { do_bad, SIGKILL, SI_KERNEL, "unknown 37" },
> - { do_bad, SIGKILL, SI_KERNEL, "unknown 38" },
> - { do_bad, SIGKILL, SI_KERNEL, "unknown 39" },
> - { do_bad, SIGKILL, SI_KERNEL, "unknown 40" },
> + { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection Fault at level 0" },
> + { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection Fault at level 1" },
> + { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection Fault at level 2" },
> + { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection Fault at level 3" },
> + { do_gpf, SIGBUS, SI_KERNEL, "Granule Protection Fault not on table walk" },
> { do_bad, SIGKILL, SI_KERNEL, "level -1 address size fault" },
> { do_bad, SIGKILL, SI_KERNEL, "unknown 42" },
> { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level -1 translation fault" },
Rest looks fine to me.
Suzuki
Hi Steven
On 12/04/2024 09:42, Steven Price wrote:
> The RMM (Realm Management Monitor) provides functionality that can be
> accessed by SMC calls from the host.
>
> The SMC definitions are based on DEN0137[1] version 1.0-eac5
>
> [1] https://developer.arm.com/documentation/den0137/1-0eac5/
>
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/include/asm/rmi_smc.h | 250 +++++++++++++++++++++++++++++++
> 1 file changed, 250 insertions(+)
> create mode 100644 arch/arm64/include/asm/rmi_smc.h
>
> diff --git a/arch/arm64/include/asm/rmi_smc.h b/arch/arm64/include/asm/rmi_smc.h
> new file mode 100644
> index 000000000000..c205efdb18d8
> --- /dev/null
> +++ b/arch/arm64/include/asm/rmi_smc.h
> @@ -0,0 +1,250 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (C) 2023 ARM Ltd.
> + *
> + * The values and structures in this file are from the Realm Management Monitor
> + * specification (DEN0137) version A-bet0:
> + * https://developer.arm.com/documentation/den0137/1-0bet0/
This should now point to eac5 instead.
> + */
> +
> +#ifndef __ASM_RME_SMC_H
> +#define __ASM_RME_SMC_H
> +
> +#include <linux/arm-smccc.h>
> +
> +#define SMC_RxI_CALL(func) \
> + ARM_SMCCC_CALL_VAL(ARM_SMCCC_FAST_CALL, \
> + ARM_SMCCC_SMC_64, \
> + ARM_SMCCC_OWNER_STANDARD, \
> + (func))
> +
> +#define SMC_RMI_DATA_CREATE SMC_RxI_CALL(0x0153)
> +#define SMC_RMI_DATA_CREATE_UNKNOWN SMC_RxI_CALL(0x0154)
> +#define SMC_RMI_DATA_DESTROY SMC_RxI_CALL(0x0155)
> +#define SMC_RMI_FEATURES SMC_RxI_CALL(0x0165)
> +#define SMC_RMI_GRANULE_DELEGATE SMC_RxI_CALL(0x0151)
> +#define SMC_RMI_GRANULE_UNDELEGATE SMC_RxI_CALL(0x0152)
> +#define SMC_RMI_PSCI_COMPLETE SMC_RxI_CALL(0x0164)
> +#define SMC_RMI_REALM_ACTIVATE SMC_RxI_CALL(0x0157)
> +#define SMC_RMI_REALM_CREATE SMC_RxI_CALL(0x0158)
> +#define SMC_RMI_REALM_DESTROY SMC_RxI_CALL(0x0159)
> +#define SMC_RMI_REC_AUX_COUNT SMC_RxI_CALL(0x0167)
> +#define SMC_RMI_REC_CREATE SMC_RxI_CALL(0x015a)
> +#define SMC_RMI_REC_DESTROY SMC_RxI_CALL(0x015b)
> +#define SMC_RMI_REC_ENTER SMC_RxI_CALL(0x015c)
> +#define SMC_RMI_RTT_CREATE SMC_RxI_CALL(0x015d)
> +#define SMC_RMI_RTT_DESTROY SMC_RxI_CALL(0x015e)
> +#define SMC_RMI_RTT_FOLD SMC_RxI_CALL(0x0166)
> +#define SMC_RMI_RTT_INIT_RIPAS SMC_RxI_CALL(0x0168)
> +#define SMC_RMI_RTT_MAP_UNPROTECTED SMC_RxI_CALL(0x015f)
> +#define SMC_RMI_RTT_READ_ENTRY SMC_RxI_CALL(0x0161)
> +#define SMC_RMI_RTT_SET_RIPAS SMC_RxI_CALL(0x0169)
> +#define SMC_RMI_RTT_UNMAP_UNPROTECTED SMC_RxI_CALL(0x0162)
> +#define SMC_RMI_VERSION SMC_RxI_CALL(0x0150)
> +
> +#define RMI_ABI_MAJOR_VERSION 1
> +#define RMI_ABI_MINOR_VERSION 0
> +
> +#define RMI_UNASSIGNED 0
> +#define RMI_ASSIGNED 1
> +#define RMI_TABLE 2
> +
> +#define RMI_ABI_VERSION_GET_MAJOR(version) ((version) >> 16)
> +#define RMI_ABI_VERSION_GET_MINOR(version) ((version) & 0xFFFF)
> +#define RMI_ABI_VERSION(major, minor) (((major) << 16) | (minor))
> +
> +#define RMI_RETURN_STATUS(ret) ((ret) & 0xFF)
> +#define RMI_RETURN_INDEX(ret) (((ret) >> 8) & 0xFF)
> +
> +#define RMI_SUCCESS 0
> +#define RMI_ERROR_INPUT 1
> +#define RMI_ERROR_REALM 2
> +#define RMI_ERROR_REC 3
> +#define RMI_ERROR_RTT 4
> +
> +#define RMI_EMPTY 0
> +#define RMI_RAM 1
> +#define RMI_DESTROYED 2
> +
> +#define RMI_NO_MEASURE_CONTENT 0
> +#define RMI_MEASURE_CONTENT 1
> +
> +#define RMI_FEATURE_REGISTER_0_S2SZ GENMASK(7, 0)
> +#define RMI_FEATURE_REGISTER_0_LPA2 BIT(8)
> +#define RMI_FEATURE_REGISTER_0_SVE_EN BIT(9)
> +#define RMI_FEATURE_REGISTER_0_SVE_VL GENMASK(13, 10)
> +#define RMI_FEATURE_REGISTER_0_NUM_BPS GENMASK(17, 14)
> +#define RMI_FEATURE_REGISTER_0_NUM_WPS GENMASK(21, 18)
> +#define RMI_FEATURE_REGISTER_0_PMU_EN BIT(22)
> +#define RMI_FEATURE_REGISTER_0_PMU_NUM_CTRS GENMASK(27, 23)
> +#define RMI_FEATURE_REGISTER_0_HASH_SHA_256 BIT(28)
> +#define RMI_FEATURE_REGISTER_0_HASH_SHA_512 BIT(29)
> +
> +#define RMI_REALM_PARAM_FLAG_LPA2 BIT(0)
> +#define RMI_REALM_PARAM_FLAG_SVE BIT(1)
> +#define RMI_REALM_PARAM_FLAG_PMU BIT(2)
> +
> +/*
> + * Note many of these fields are smaller than u64 but all fields have u64
> + * alignment, so use u64 to ensure correct alignment.
> + */
> +struct realm_params {
> + union { /* 0x0 */
> + struct {
> + u64 flags;
> + u64 s2sz;
> + u64 sve_vl;
> + u64 num_bps;
> + u64 num_wps;
> + u64 pmu_num_ctrs;
> + u64 hash_algo;
> + };
> + u8 padding_1[0x400];
> + };
> + union { /* 0x400 */
> + u8 rpv[64];
> + u8 padding_2[0x400];
> + };
> + union { /* 0x800 */
> + struct {
> + u64 vmid;
> + u64 rtt_base;
> + s64 rtt_level_start;
> + u64 rtt_num_start;
> + };
> + u8 padding_3[0x800];
> + };
> +};
> +
> +/*
> + * The number of GPRs (starting from X0) that are
> + * configured by the host when a REC is created.
> + */
> +#define REC_CREATE_NR_GPRS 8
> +
> +#define REC_PARAMS_FLAG_RUNNABLE BIT_ULL(0)
> +
> +#define REC_PARAMS_AUX_GRANULES 16
> +
> +struct rec_params {
> + union { /* 0x0 */
> + u64 flags;
> + u8 padding1[0x100];
> + };
> + union { /* 0x100 */
> + u64 mpidr;
> + u8 padding2[0x100];
> + };
> + union { /* 0x200 */
> + u64 pc;
> + u8 padding3[0x100];
> + };
> + union { /* 0x300 */
> + u64 gprs[REC_CREATE_NR_GPRS];
> + u8 padding4[0x500];
> + };
> + union { /* 0x800 */
> + struct {
> + u64 num_rec_aux;
> + u64 aux[REC_PARAMS_AUX_GRANULES];
> + };
> + u8 padding5[0x800];
> + };
> +};
> +
> +#define RMI_EMULATED_MMIO BIT(0)
> +#define RMI_INJECT_SEA BIT(1)
> +#define RMI_TRAP_WFI BIT(2)
> +#define RMI_TRAP_WFE BIT(3)
For completeness, we could add :
#define RMI_RIPAS_RESPONSE BIT(4)
Not sure if we use it later in the series.
> +
> +#define REC_RUN_GPRS 31
> +#define REC_GIC_NUM_LRS 16
> +
> +struct rec_entry {
> + union { /* 0x000 */
> + u64 flags;
> + u8 padding0[0x200];
> + };
> + union { /* 0x200 */
> + u64 gprs[REC_RUN_GPRS];
> + u8 padding2[0x100];
> + };
> + union { /* 0x300 */
> + struct {
> + u64 gicv3_hcr;
> + u64 gicv3_lrs[REC_GIC_NUM_LRS];
> + };
> + u8 padding3[0x100];
> + };
> + u8 padding4[0x400];
> +};
> +
> +struct rec_exit {
> + union { /* 0x000 */
> + u8 exit_reason;
> + u8 padding0[0x100];
> + };
> + union { /* 0x100 */
> + struct {
> + u64 esr;
> + u64 far;
> + u64 hpfar;
> + };
> + u8 padding1[0x100];
> + };
> + union { /* 0x200 */
> + u64 gprs[REC_RUN_GPRS];
> + u8 padding2[0x100];
> + };
> + union { /* 0x300 */
> + struct {
> + u64 gicv3_hcr;
> + u64 gicv3_lrs[REC_GIC_NUM_LRS];
> + u64 gicv3_misr;
> + u64 gicv3_vmcr;
> + };
> + u8 padding3[0x100];
> + };
> + union { /* 0x400 */
> + struct {
> + u64 cntp_ctl;
> + u64 cntp_cval;
> + u64 cntv_ctl;
> + u64 cntv_cval;
> + };
> + u8 padding4[0x100];
> + };
> + union { /* 0x500 */
> + struct {
> + u64 ripas_base;
> + u64 ripas_top;
> + u64 ripas_value;
> + };
> + u8 padding5[0x100];
> + };
> + union { /* 0x600 */
> + u16 imm;
> + u8 padding6[0x100];
> + };
> + union { /* 0x700 */
> + struct {
> + u64 pmu_ovf_status;
This is u8 as per section B4.4.10 RmiPmuOverflowStatus type.
> + };
> + u8 padding7[0x100];
> + };
> +};
> +
> +struct rec_run {
> + struct rec_entry entry;
> + struct rec_exit exit;
> +};
> +
> +#define RMI_EXIT_SYNC 0x00
> +#define RMI_EXIT_IRQ 0x01
> +#define RMI_EXIT_FIQ 0x02
> +#define RMI_EXIT_PSCI 0x03
> +#define RMI_EXIT_RIPAS_CHANGE 0x04
> +#define RMI_EXIT_HOST_CALL 0x05
> +#define RMI_EXIT_SERROR 0x06
Minor nit: Like the other definitions, it may be good to keep the
defintions of the "exit_reason" above the field declaration.
Rest looks fine to me.
Suzuki
> +
> +#endif
Hi Steven
On 12/04/2024 09:42, Steven Price wrote:
> The wrappers make the call sites easier to read and deal with the
> boiler plate of handling the error codes from the RMM.
>
I have compared the parameters and output values to that of the RMM spec
and they match. There are some minor nits below.
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/include/asm/rmi_cmds.h | 509 ++++++++++++++++++++++++++++++
> 1 file changed, 509 insertions(+)
> create mode 100644 arch/arm64/include/asm/rmi_cmds.h
>
> diff --git a/arch/arm64/include/asm/rmi_cmds.h b/arch/arm64/include/asm/rmi_cmds.h
> new file mode 100644
> index 000000000000..c21414127e8e
> --- /dev/null
> +++ b/arch/arm64/include/asm/rmi_cmds.h
> @@ -0,0 +1,509 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (C) 2023 ARM Ltd.
> + */
> +
> +#ifndef __ASM_RMI_CMDS_H
> +#define __ASM_RMI_CMDS_H
> +
> +#include <linux/arm-smccc.h>
> +
> +#include <asm/rmi_smc.h>
> +
> +struct rtt_entry {
> + unsigned long walk_level;
> + unsigned long desc;
> + int state;
> + int ripas;
> +};
> +
..
> +/**
> + * rmi_data_destroy() - Destroy a Data Granule
> + * @rd: PA of the RD
> + * @ipa: IPA at which the granule is mapped in the guest
> + * @data_out: PA of the granule which was destroyed
> + * @top_out: Top IPA of non-live RTT entries
> + *
> + * Transitions the granule to DESTROYED state, the address cannot be used by
> + * the guest for the lifetime of the Realm.
> + *
> + * Return: RMI return code
> + */
> +static inline int rmi_data_destroy(unsigned long rd, unsigned long ipa,
> + unsigned long *data_out,
> + unsigned long *top_out)
> +{
> + struct arm_smccc_res res;
> +
> + arm_smccc_1_1_invoke(SMC_RMI_DATA_DESTROY, rd, ipa, &res);
> +
> + *data_out = res.a1;
> + *top_out = res.a2;
minor nit: Do we need to be safer by checking the parameters before
filling them in ? i.e.,
if (ptr)
*ptr = result_out;
This applies for others calls below.
> +
> + return res.a0;
> +}
> +
> +/**
> + * rmi_realm_destroy() - Destroy a Realm
> + * @rd: PA of the RD
> + *
> + * Destroys a Realm, all objects belonging to the Realm must be destroyed first.
> + *
> + * Return: RMI return code
> + */
> +static inline int rmi_realm_destroy(unsigned long rd)
> +{
> + struct arm_smccc_res res;
> +
> + arm_smccc_1_1_invoke(SMC_RMI_REALM_DESTROY, rd, &res);
> +
> + return res.a0;
> +}
> +
> +/**
> + * rmi_rec_aux_count() - Get number of auxiliary Granules required
> + * @rd: PA of the RD
> + * @aux_count: Number of pages written to this pointer
> + *
> + * A REC may require extra auxiliary pages to be delegateed for the RMM to
minor nit: "s/delegateed/delegated/"
..
> +/**
> + * rmi_rtt_read_entry() - Read an RTTE
> + * @rd: PA of the RD
> + * @ipa: IPA for which to read the RTTE
> + * @level: RTT level at which to read the RTTE
> + * @rtt: Output structure describing the RTTE
> + *
> + * Reads a RTTE (Realm Translation Table Entry).
> + *
> + * Return: RMI return code
> + */
> +static inline int rmi_rtt_read_entry(unsigned long rd, unsigned long ipa,
> + long level, struct rtt_entry *rtt)
> +{
> + struct arm_smccc_1_2_regs regs = {
> + SMC_RMI_RTT_READ_ENTRY,
> + rd, ipa, level
> + };
> +
> + arm_smccc_1_2_smc(®s, ®s);
> +
> + rtt->walk_level = regs.a1;
> + rtt->state = regs.a2 & 0xFF;
minor nit: We mask the state, but not the "ripas". Both of them are u8.
For consistency, we should mask both or neither.
> + rtt->desc = regs.a3;
> + rtt->ripas = regs.a4;
> +
> + return regs.a0;
> +}
> +
..
> +/**
> + * rmi_rtt_get_phys() - Get the PA from a RTTE
> + * @rtt: The RTTE
> + *
> + * Return: the physical address from a RTT entry.
> + */
> +static inline phys_addr_t rmi_rtt_get_phys(struct rtt_entry *rtt)
> +{
> + return rtt->desc & GENMASK(47, 12);
> +}
I guess this may need to change with the LPA2 support in RMM and must be
used in conjunction with the "realm" object to make the correct
conversion.
Suzuki
> +
> +#endif
Hi Steven
On 12/04/2024 09:42, Steven Price wrote:
> Query the RMI version number and check if it is a compatible version. A
> static key is also provided to signal that a supported RMM is available.
>
> Functions are provided to query if a VM or VCPU is a realm (or rec)
> which currently will always return false.
>
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/include/asm/kvm_emulate.h | 18 +++++++++
> arch/arm64/include/asm/kvm_host.h | 4 ++
> arch/arm64/include/asm/kvm_rme.h | 56 ++++++++++++++++++++++++++++
> arch/arm64/include/asm/virt.h | 1 +
> arch/arm64/kvm/Makefile | 3 +-
> arch/arm64/kvm/arm.c | 9 +++++
> arch/arm64/kvm/rme.c | 52 ++++++++++++++++++++++++++
> 7 files changed, 142 insertions(+), 1 deletion(-)
> create mode 100644 arch/arm64/include/asm/kvm_rme.h
> create mode 100644 arch/arm64/kvm/rme.c
>
> diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
> index 975af30af31f..6f08398537e2 100644
> --- a/arch/arm64/include/asm/kvm_emulate.h
> +++ b/arch/arm64/include/asm/kvm_emulate.h
> @@ -611,4 +611,22 @@ static __always_inline void kvm_reset_cptr_el2(struct kvm_vcpu *vcpu)
>
> kvm_write_cptr_el2(val);
> }
> +
> +static inline bool kvm_is_realm(struct kvm *kvm)
> +{
> + if (static_branch_unlikely(&kvm_rme_is_available))
> + return kvm->arch.is_realm;
> + return false;
> +}
> +
> +static inline enum realm_state kvm_realm_state(struct kvm *kvm)
> +{
> + return READ_ONCE(kvm->arch.realm.state);
> +}
> +
> +static inline bool vcpu_is_rec(struct kvm_vcpu *vcpu)
> +{
> + return false;
> +}
> +
> #endif /* __ARM64_KVM_EMULATE_H__ */
> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> index 9e8a496fb284..63b68b85db3f 100644
> --- a/arch/arm64/include/asm/kvm_host.h
> +++ b/arch/arm64/include/asm/kvm_host.h
> @@ -27,6 +27,7 @@
> #include <asm/fpsimd.h>
> #include <asm/kvm.h>
> #include <asm/kvm_asm.h>
> +#include <asm/kvm_rme.h>
> #include <asm/vncr_mapping.h>
>
> #define __KVM_HAVE_ARCH_INTC_INITIALIZED
> @@ -348,6 +349,9 @@ struct kvm_arch {
> * the associated pKVM instance in the hypervisor.
> */
> struct kvm_protected_vm pkvm;
> +
> + bool is_realm;
> + struct realm realm;
> };
>
> struct kvm_vcpu_fault_info {
> diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
> new file mode 100644
> index 000000000000..922da3f47227
> --- /dev/null
> +++ b/arch/arm64/include/asm/kvm_rme.h
> @@ -0,0 +1,56 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (C) 2023 ARM Ltd.
> + */
> +
> +#ifndef __ASM_KVM_RME_H
> +#define __ASM_KVM_RME_H
> +
> +/**
> + * enum realm_state - State of a Realm
> + */
> +enum realm_state {
> + /**
> + * @REALM_STATE_NONE:
> + * Realm has not yet been created. rmi_realm_create() may be
> + * called to create the realm.
> + */
> + REALM_STATE_NONE,
> + /**
> + * @REALM_STATE_NEW:
> + * Realm is under construction, not eligible for execution. Pages
> + * may be populated with rmi_data_create().
> + */
> + REALM_STATE_NEW,
> + /**
> + * @REALM_STATE_ACTIVE:
> + * Realm has been created and is eligible for execution with
> + * rmi_rec_enter(). Pages may no longer be populated with
> + * rmi_data_create().
> + */
> + REALM_STATE_ACTIVE,
> + /**
> + * @REALM_STATE_DYING:
> + * Realm is in the process of being destroyed or has already been
> + * destroyed.
> + */
> + REALM_STATE_DYING,
> + /**
> + * @REALM_STATE_DEAD:
> + * Realm has been destroyed.
> + */
> + REALM_STATE_DEAD
> +};
> +
> +/**
> + * struct realm - Additional per VM data for a Realm
> + *
> + * @state: The lifetime state machine for the realm
> + */
> +struct realm {
> + enum realm_state state;
> +};
> +
> +int kvm_init_rme(void);
> +
> +#endif
> diff --git a/arch/arm64/include/asm/virt.h b/arch/arm64/include/asm/virt.h
> index 261d6e9df2e1..12cf36c38189 100644
> --- a/arch/arm64/include/asm/virt.h
> +++ b/arch/arm64/include/asm/virt.h
> @@ -81,6 +81,7 @@ void __hyp_reset_vectors(void);
> bool is_kvm_arm_initialised(void);
>
> DECLARE_STATIC_KEY_FALSE(kvm_protected_mode_initialized);
> +DECLARE_STATIC_KEY_FALSE(kvm_rme_is_available);
>
> /* Reports the availability of HYP mode */
> static inline bool is_hyp_mode_available(void)
> diff --git a/arch/arm64/kvm/Makefile b/arch/arm64/kvm/Makefile
> index c0c050e53157..1c1d8cdf381f 100644
> --- a/arch/arm64/kvm/Makefile
> +++ b/arch/arm64/kvm/Makefile
> @@ -20,7 +20,8 @@ kvm-y += arm.o mmu.o mmio.o psci.o hypercalls.o pvtime.o \
> vgic/vgic-v3.o vgic/vgic-v4.o \
> vgic/vgic-mmio.o vgic/vgic-mmio-v2.o \
> vgic/vgic-mmio-v3.o vgic/vgic-kvm-device.o \
> - vgic/vgic-its.o vgic/vgic-debug.o
> + vgic/vgic-its.o vgic/vgic-debug.o \
> + rme.o
>
> kvm-$(CONFIG_HW_PERF_EVENTS) += pmu-emul.o pmu.o
>
> diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
> index 3dee5490eea9..2056c660c5ee 100644
> --- a/arch/arm64/kvm/arm.c
> +++ b/arch/arm64/kvm/arm.c
> @@ -38,6 +38,7 @@
> #include <asm/kvm_mmu.h>
> #include <asm/kvm_nested.h>
> #include <asm/kvm_pkvm.h>
> +#include <asm/kvm_rme.h>
> #include <asm/kvm_emulate.h>
> #include <asm/sections.h>
>
> @@ -47,6 +48,8 @@
>
> static enum kvm_mode kvm_mode = KVM_MODE_DEFAULT;
>
> +DEFINE_STATIC_KEY_FALSE(kvm_rme_is_available);
> +
> DECLARE_KVM_HYP_PER_CPU(unsigned long, kvm_hyp_vector);
>
> DEFINE_PER_CPU(unsigned long, kvm_arm_hyp_stack_page);
> @@ -2562,6 +2565,12 @@ static __init int kvm_arm_init(void)
>
> in_hyp_mode = is_kernel_in_hyp_mode();
>
> + if (in_hyp_mode) {
> + err = kvm_init_rme();
> + if (err)
> + return err;
> + }
> +
> if (cpus_have_final_cap(ARM64_WORKAROUND_DEVICE_LOAD_ACQUIRE) ||
> cpus_have_final_cap(ARM64_WORKAROUND_1508412))
> kvm_info("Guests without required CPU erratum workarounds can deadlock system!\n" \
> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
> new file mode 100644
> index 000000000000..3dbbf9d046bf
> --- /dev/null
> +++ b/arch/arm64/kvm/rme.c
> @@ -0,0 +1,52 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright (C) 2023 ARM Ltd.
> + */
> +
> +#include <linux/kvm_host.h>
> +
> +#include <asm/rmi_cmds.h>
> +#include <asm/virt.h>
> +
> +static int rmi_check_version(void)
> +{
> + struct arm_smccc_res res;
> + int version_major, version_minor;
> + unsigned long host_version = RMI_ABI_VERSION(RMI_ABI_MAJOR_VERSION,
> + RMI_ABI_MINOR_VERSION);
> +
> + arm_smccc_1_1_invoke(SMC_RMI_VERSION, host_version, &res);
> +
> + if (res.a0 == SMCCC_RET_NOT_SUPPORTED)
> + return -ENXIO;
> +
> + version_major = RMI_ABI_VERSION_GET_MAJOR(res.a1);
> + version_minor = RMI_ABI_VERSION_GET_MINOR(res.a1);
> +
We don't seem to be using the res.a0 to determin if the RMM supports our
requested version. As per RMM spec, section B4.3.23 :
"
The status code and lower revision output values indicate which of the
following is true, in order of precedence:
a) The RMM supports an interface revision which is compatible with the
requested revision.
• The status code is RMI_SUCCESS.
• The lower revision is equal to the requested revision.
b) The RMM does not support an interface revision which is compatible
with the requested revision The RMM supports an interface revision
which is incompatible with and less than the requested revision.
• The status code is RMI_ERROR_INPUT.
• The lower revision is the highest interface revision which is
both less than the requested revision and supported by the RMM.
c) The RMM does not support an interface revision which is compatible
with the requested revision The RMM supports an interface revision
which is incompatible with and greater than the requested revision.
• The status code is RMI_ERROR_INPUT.
• The lower revision is equal to the higher revision.
So, we could simply check the res.a0 for RMI_SUCCESS and proceed with
marking RMM available.
> + if (version_major != RMI_ABI_MAJOR_VERSION) {
> + kvm_err("Unsupported RMI ABI (v%d.%d) host supports v%d.%d\n",
> + version_major, version_minor,
> + RMI_ABI_MAJOR_VERSION,
> + RMI_ABI_MINOR_VERSION);
> + return -ENXIO;
> + }
> +
> + kvm_info("RMI ABI version %d.%d\n", version_major, version_minor);
> +
> + return 0;
> +}
> +
> +int kvm_init_rme(void)
> +{
> + if (PAGE_SIZE != SZ_4K)
> + /* Only 4k page size on the host is supported */
> + return 0;
> +
> + if (rmi_check_version())
> + /* Continue without realm support */
> + return 0;
> +
> + /* Future patch will enable static branch kvm_rme_is_available */
> +
> + return 0;
Do we ever expect this to fail the kvm initialisation ? Otherwise, we
could leave it as a void ?
Suzuki
> +}
Hi Steven
On 12/04/2024 09:42, Steven Price wrote:
> Add the KVM_CAP_ARM_RME_CREATE_FD ioctl to create a realm. This involves
minor nit: s/FD/RD
> delegating pages to the RMM to hold the Realm Descriptor (RD) and for
> the base level of the Realm Translation Tables (RTT). A VMID also need
> to be picked, since the RMM has a separate VMID address space a
> dedicated allocator is added for this purpose.
>
> KVM_CAP_ARM_RME_CONFIG_REALM is provided to allow configuring the realm
> before it is created.
It might be helpful to provide a bit more background on the Realm
parameters. Something like:
Realm parameters for Realm Descriptor creation could be classified as:
1. Parameters specific to the Realm stage2 (e.g. IPA Size, vmid, stage2
entry level, entry level RTTs, number of RTTs in start level, LPA2)
Most of these are not measured by RMM and comes from KVM book
keeping.
2. Parameters controlling "Arm Architecture features for the VM". (e.g.
SVE VL, PMU counters, number of HW BRPs/WPs), configured by the VMM
using the "user ID register write" mechanism. These will be
supported in the later patches.
3. Parameters are not part of the core Arm architecture but defined
by the RMM spec (e.g. Hash algorithm for measurement,
Personalisation value). These are programmed via
KVM_CAP_ARM_RME_CONFIG_REALM.
Also it may be a good idea to call out one of the issues that we have
with the UABI w.r.t the IPA Size. The IPA Size supported by RMM *could*
be different from the normal IPA Size as supported by KVM. We do not
expect this to be common, but is not impossible.
If the RMM_IPA_Size < Normal_IPA_Size, we have a problem with
advertising the "IPA Size" to the VMM. Right now we advertise
the "normal limit" by KVM_CAP_ARM_VM_IPA_SIZE and the IPA Size
is configured via vm_type[7:0] in KVM_CREATE_VM. Given we have
to configure the IPA size for a "Realm VM" at CREATE_VM time too,
the VMM is unable to choose a valid IPA Size for the Realm. We
have the following options:
1. Given IPA Size for a Realm is measured, the user must get
what they choose. i.e., if the platform cannot support the
requested size, don't run your Realm VM. In this case, we
don't need to do anything.
2. Add KVM_CAP_ARM_VM_RMM_IPA_SIZE to expose the RMM limit
for the VMM to choose.
3. VMM to create a Realm VM using the default IPA Size and then
check the KVM_CAP_ARM_VM_IPA_SIZE on the "kvm" instance (which
is Realm) and get the RMM IPA limit.
I prefer 2 or 1, in that order of preference. Happy to hear suggestions.
>
> Co-developed-by: Suzuki K Poulose <[email protected]>
> Signed-off-by: Suzuki K Poulose <[email protected]>
> Signed-off-by: Steven Price <[email protected]>
> Signed-off-by: Jean-Philippe Brucker <[email protected]>
> ---
> arch/arm64/include/asm/kvm_emulate.h | 5 +
> arch/arm64/include/asm/kvm_rme.h | 19 ++
> arch/arm64/kvm/arm.c | 18 ++
> arch/arm64/kvm/mmu.c | 15 +-
> arch/arm64/kvm/rme.c | 282 +++++++++++++++++++++++++++
> 5 files changed, 337 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
> index 6f08398537e2..c606316f4729 100644
> --- a/arch/arm64/include/asm/kvm_emulate.h
> +++ b/arch/arm64/include/asm/kvm_emulate.h
> @@ -624,6 +624,11 @@ static inline enum realm_state kvm_realm_state(struct kvm *kvm)
> return READ_ONCE(kvm->arch.realm.state);
> }
>
> +static inline bool kvm_realm_is_created(struct kvm *kvm)
> +{
> + return kvm_is_realm(kvm) && kvm_realm_state(kvm) != REALM_STATE_NONE;
> +}
> +
> static inline bool vcpu_is_rec(struct kvm_vcpu *vcpu)
> {
> return false;
> diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
> index 922da3f47227..cf8cc4d30364 100644
> --- a/arch/arm64/include/asm/kvm_rme.h
> +++ b/arch/arm64/include/asm/kvm_rme.h
> @@ -6,6 +6,8 @@
> #ifndef __ASM_KVM_RME_H
> #define __ASM_KVM_RME_H
>
> +#include <uapi/linux/kvm.h>
> +
> /**
> * enum realm_state - State of a Realm
> */
> @@ -46,11 +48,28 @@ enum realm_state {
> * struct realm - Additional per VM data for a Realm
> *
> * @state: The lifetime state machine for the realm
> + * @rd: Kernel mapping of the Realm Descriptor (RD)
> + * @params: Parameters for the RMI_REALM_CREATE command
> + * @num_aux: The number of auxiliary pages required by the RMM
> + * @vmid: VMID to be used by the RMM for the realm
> + * @ia_bits: Number of valid Input Address bits in the IPA
> */
> struct realm {
> enum realm_state state;
> +
> + void *rd;
> + struct realm_params *params;
> +
> + unsigned long num_aux;
> + unsigned int vmid;
> + unsigned int ia_bits;
> };
>
> int kvm_init_rme(void);
> +u32 kvm_realm_ipa_limit(void);
> +
> +int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap);
> +int kvm_init_realm_vm(struct kvm *kvm);
> +void kvm_destroy_realm(struct kvm *kvm);
>
> #endif
> diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
> index 2056c660c5ee..5729ea430d6d 100644
> --- a/arch/arm64/kvm/arm.c
> +++ b/arch/arm64/kvm/arm.c
> @@ -119,6 +119,13 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
> }
> mutex_unlock(&kvm->slots_lock);
> break;
> + case KVM_CAP_ARM_RME:
> + if (!kvm_is_realm(kvm))
> + return -EINVAL;
> + mutex_lock(&kvm->lock);
> + r = kvm_realm_enable_cap(kvm, cap);
> + mutex_unlock(&kvm->lock);
> + break;
> default:
> r = -EINVAL;
> break;
> @@ -179,6 +186,13 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
>
> bitmap_zero(kvm->arch.vcpu_features, KVM_VCPU_MAX_FEATURES);
>
> + /* Initialise the realm bits after the generic bits are enabled */
> + if (kvm_is_realm(kvm)) {
> + ret = kvm_init_realm_vm(kvm);
> + if (ret)
> + goto err_free_cpumask;
> + }
> +
> return 0;
>
> err_free_cpumask:
> @@ -219,6 +233,7 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
> kvm_unshare_hyp(kvm, kvm + 1);
>
> kvm_arm_teardown_hypercalls(kvm);
> + kvm_destroy_realm(kvm);
> }
>
> int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
> @@ -328,6 +343,9 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
> case KVM_CAP_ARM_SUPPORTED_REG_MASK_RANGES:
> r = BIT(0);
> break;
> + case KVM_CAP_ARM_RME:
> + r = static_key_enabled(&kvm_rme_is_available);
> + break;
> default:
> r = 0;
> }
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index 18680771cdb0..aae365647b62 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -872,6 +872,10 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
> struct kvm_pgtable *pgt;
> u64 mmfr0, mmfr1;
> u32 phys_shift;
> + u32 ipa_limit = kvm_ipa_limit;
> +
> + if (kvm_is_realm(kvm))
> + ipa_limit = kvm_realm_ipa_limit();
>
> if (type & ~KVM_VM_TYPE_ARM_IPA_SIZE_MASK)
> return -EINVAL;
> @@ -880,12 +884,12 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
> if (is_protected_kvm_enabled()) {
> phys_shift = kvm_ipa_limit;
> } else if (phys_shift) {
> - if (phys_shift > kvm_ipa_limit ||
> + if (phys_shift > ipa_limit ||
> phys_shift < ARM64_MIN_PARANGE_BITS)
> return -EINVAL;
> } else {
> phys_shift = KVM_PHYS_SHIFT;
> - if (phys_shift > kvm_ipa_limit) {
> + if (phys_shift > ipa_limit) {
> pr_warn_once("%s using unsupported default IPA limit, upgrade your VMM\n",
> current->comm);
> return -EINVAL;
> @@ -1014,6 +1018,13 @@ void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu)
> struct kvm_pgtable *pgt = NULL;
>
> write_lock(&kvm->mmu_lock);
> + if (kvm_is_realm(kvm) &&
> + (kvm_realm_state(kvm) != REALM_STATE_DEAD &&
> + kvm_realm_state(kvm) != REALM_STATE_NONE)) {
> + /* TODO: teardown rtts */
> + write_unlock(&kvm->mmu_lock);
> + return;
> + }
This needs a comment to explain the rationale of deferring the
Stage2 pgt freeing. Something like :
/*
* For realms, we can free the entry level RTTs
* only after :
* 1. All of the stage2 mappings are torn down.
* 2. The Realm has been destroyed.
*
* So, come back later once the RD has been destroyed.
*/
> pgt = mmu->pgt;
> if (pgt) {
> mmu->pgd_phys = 0;
> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
> index 3dbbf9d046bf..658d14e8d87d 100644
> --- a/arch/arm64/kvm/rme.c
> +++ b/arch/arm64/kvm/rme.c
> @@ -5,9 +5,20 @@
>
> #include <linux/kvm_host.h>
>
> +#include <asm/kvm_emulate.h>
> +#include <asm/kvm_mmu.h>
> #include <asm/rmi_cmds.h>
> #include <asm/virt.h>
>
> +#include <asm/kvm_pgtable.h>
> +
> +static unsigned long rmm_feat_reg0;
> +
> +static bool rme_supports(unsigned long feature)
> +{
> + return !!u64_get_bits(rmm_feat_reg0, feature);
> +}
> +
> static int rmi_check_version(void)
> {
> struct arm_smccc_res res;
> @@ -36,8 +47,272 @@ static int rmi_check_version(void)
> return 0;
> }
>
> +u32 kvm_realm_ipa_limit(void)
> +{
> + return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
> +}
> +
> +static int get_start_level(struct realm *realm)
> +{
> + return 4 - stage2_pgtable_levels(realm->ia_bits);
> +}
> +
> +static int realm_create_rd(struct kvm *kvm)
> +{
> + struct realm *realm = &kvm->arch.realm;
> + struct realm_params *params = realm->params;
> + void *rd = NULL;
> + phys_addr_t rd_phys, params_phys;
> + struct kvm_pgtable *pgt = kvm->arch.mmu.pgt;
> + int i, r;
> +
> + if (WARN_ON(realm->rd) || WARN_ON(!realm->params))
> + return -EEXIST;
> +
> + rd = (void *)__get_free_page(GFP_KERNEL);
> + if (!rd)
> + return -ENOMEM;
> +
> + rd_phys = virt_to_phys(rd);
> + if (rmi_granule_delegate(rd_phys)) {
> + r = -ENXIO;
> + goto out;
super minor nit: s/out/free_rd/ is a bit more readable. Here "out" is
only used for error exits and could be confusing.
> + }
> +
> + for (i = 0; i < pgt->pgd_pages; i++) {
> + phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
> +
> + if (rmi_granule_delegate(pgd_phys)) {
> + r = -ENXIO;
> + goto out_undelegate_tables;
> + }
> + }
> +
> + realm->ia_bits = VTCR_EL2_IPA(kvm->arch.mmu.vtcr);
> +
> + params->rtt_level_start = get_start_level(realm);
> + params->rtt_num_start = pgt->pgd_pages;
> + params->rtt_base = kvm->arch.mmu.pgd_phys;
> + params->vmid = realm->vmid;
> +
> + params_phys = virt_to_phys(params);
> +
> + if (rmi_realm_create(rd_phys, params_phys)) {
> + r = -ENXIO;
> + goto out_undelegate_tables;
> + }
> +
> + realm->rd = rd;
> +
> + if (WARN_ON(rmi_rec_aux_count(rd_phys, &realm->num_aux))) {
> + WARN_ON(rmi_realm_destroy(rd_phys));
> + goto out_undelegate_tables;
> + }
> +
> + return 0;
> +
> +out_undelegate_tables:
> + while (--i >= 0) {
> + phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
> +
> + WARN_ON(rmi_granule_undelegate(pgd_phys));
> + }
> + WARN_ON(rmi_granule_undelegate(rd_phys));
> +out:
> + free_page((unsigned long)rd);
> + return r;
> +}
> +
> +/* Protects access to rme_vmid_bitmap */
> +static DEFINE_SPINLOCK(rme_vmid_lock);
> +static unsigned long *rme_vmid_bitmap;
> +
> +static int rme_vmid_init(void)
> +{
> + unsigned int vmid_count = 1 << kvm_get_vmid_bits();
minor nit: RMM has a fixed VMID width of 16bits. Do we need to
explicitly use that, instead of relying what kvm thinks ? (Though
in practise, this would only be a problem if the architecture
evolves to support something more).
> +
> + rme_vmid_bitmap = bitmap_zalloc(vmid_count, GFP_KERNEL);
> + if (!rme_vmid_bitmap) {
> + kvm_err("%s: Couldn't allocate rme vmid bitmap\n", __func__);
> + return -ENOMEM;
> + }
> +
> + return 0;
> +}
> +
> +static int rme_vmid_reserve(void)
> +{
> + int ret;
> + unsigned int vmid_count = 1 << kvm_get_vmid_bits();
> +
> + spin_lock(&rme_vmid_lock);
> + ret = bitmap_find_free_region(rme_vmid_bitmap, vmid_count, 0);
> + spin_unlock(&rme_vmid_lock);
> +
> + return ret;
> +}
> +
> +static void rme_vmid_release(unsigned int vmid)
> +{
> + spin_lock(&rme_vmid_lock);
> + bitmap_release_region(rme_vmid_bitmap, vmid, 0);
> + spin_unlock(&rme_vmid_lock);
> +}
> +
> +static int kvm_create_realm(struct kvm *kvm)
> +{
> + struct realm *realm = &kvm->arch.realm;
> + int ret;
> +
> + if (!kvm_is_realm(kvm) || kvm_realm_is_created(kvm))
> + return -EEXIST;
Minor nit:
if (!kvm_is_realm(kvm)
return -EIO or even -EINVAL ?
if (kvm_realm_is_created(kvm))
return -EEXIST;
> +
> + ret = rme_vmid_reserve();
> + if (ret < 0)
> + return ret;
> + realm->vmid = ret;
> +
> + ret = realm_create_rd(kvm);
> + if (ret) {
> + rme_vmid_release(realm->vmid);
> + return ret;
> + }
> +
> + WRITE_ONCE(realm->state, REALM_STATE_NEW);
> +
> + /* The realm is up, free the parameters. */
> + free_page((unsigned long)realm->params);
> + realm->params = NULL;
> +
> + return 0;
> +}
> +
> +static int config_realm_hash_algo(struct realm *realm,
> + struct kvm_cap_arm_rme_config_item *cfg)
> +{
> + switch (cfg->hash_algo) {
> + case KVM_CAP_ARM_RME_MEASUREMENT_ALGO_SHA256:
> + if (!rme_supports(RMI_FEATURE_REGISTER_0_HASH_SHA_256))
> + return -EINVAL;
> + break;
> + case KVM_CAP_ARM_RME_MEASUREMENT_ALGO_SHA512:
> + if (!rme_supports(RMI_FEATURE_REGISTER_0_HASH_SHA_512))
> + return -EINVAL;
Do we need to add a comment here on why we don't expose the supported
"hash" algo as part of the UABI ? Something like :
/*
* The hash algorithm for the measurements is a choosen by
* the Realm owner (since it affects the attestation), we
* would like the owner to get what they wants.
*/
> + break;
> + default:
> + return -EINVAL;
> + }
> + realm->params->hash_algo = cfg->hash_algo;
> + return 0;
> +}
> +
> +static int kvm_rme_config_realm(struct kvm *kvm, struct kvm_enable_cap *cap)
> +{
> + struct kvm_cap_arm_rme_config_item cfg;
> + struct realm *realm = &kvm->arch.realm;
> + int r = 0;
> +
> + if (kvm_realm_is_created(kvm))
> + return -EINVAL;
minor nit: May be return -EEXIST or -EIO rather than, "Invalid
(parameter)" ?
> +
> + if (copy_from_user(&cfg, (void __user *)cap->args[1], sizeof(cfg)))
> + return -EFAULT;
> +
> + switch (cfg.cfg) {
> + case KVM_CAP_ARM_RME_CFG_RPV:
> + memcpy(&realm->params->rpv, &cfg.rpv, sizeof(cfg.rpv));
> + break;
> + case KVM_CAP_ARM_RME_CFG_HASH_ALGO:
> + r = config_realm_hash_algo(realm, &cfg);
> + break;
> + default:
> + r = -EINVAL;
> + }
> +
> + return r;
> +}
> +
> +int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
> +{
> + int r = 0;
> +
> + if (!kvm_is_realm(kvm))
> + return -EINVAL;
> +
> + switch (cap->args[0]) {
> + case KVM_CAP_ARM_RME_CONFIG_REALM:
> + r = kvm_rme_config_realm(kvm, cap);
> + break;
> + case KVM_CAP_ARM_RME_CREATE_RD:
> + r = kvm_create_realm(kvm);
> + break;
> + default:
> + r = -EINVAL;
> + break;
> + }
> +
> + return r;
> +}
> +
> +void kvm_destroy_realm(struct kvm *kvm)
> +{
> + struct realm *realm = &kvm->arch.realm;
> + struct kvm_pgtable *pgt = kvm->arch.mmu.pgt;
> + int i;
> +
> + if (realm->params) {
> + free_page((unsigned long)realm->params);
> + realm->params = NULL;
> + }
> +
> + if (!kvm_realm_is_created(kvm))
> + return;
> +
> + WRITE_ONCE(realm->state, REALM_STATE_DYING);
> +
> + if (realm->rd) {
> + phys_addr_t rd_phys = virt_to_phys(realm->rd);
> +
> + if (WARN_ON(rmi_realm_destroy(rd_phys)))
> + return;
> + if (WARN_ON(rmi_granule_undelegate(rd_phys)))
> + return;
> + free_page((unsigned long)realm->rd);
> + realm->rd = NULL;
> + }
> +
> + rme_vmid_release(realm->vmid);
> +
> + for (i = 0; i < pgt->pgd_pages; i++) {
> + phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
> +
> + if (WARN_ON(rmi_granule_undelegate(pgd_phys)))
> + return;
> + }
> +
> + WRITE_ONCE(realm->state, REALM_STATE_DEAD);
> +
May be add in a comment here:
/* Now that the Realm is destroyed, free the entry level RTTs */
> + kvm_free_stage2_pgd(&kvm->arch.mmu);
> +}
> +
> +int kvm_init_realm_vm(struct kvm *kvm)
> +{
> + struct realm_params *params;
> +
> + params = (struct realm_params *)get_zeroed_page(GFP_KERNEL);
> + if (!params)
> + return -ENOMEM;
> +
> + /* Default parameters, not exposed to user space */
This is a bit misleading. The value comes from the userspace and...
> + params->s2sz = VTCR_EL2_IPA(kvm->arch.mmu.vtcr);
(minor nit) we initialise most of the params, including those that come
from KVM later. So, as such may be a good idea to move it together at
create_realm, unless we need it.
> + kvm->arch.realm.params = params;
> + return 0;
> +}
> +
> int kvm_init_rme(void)
> {
> + int ret;
> +
> if (PAGE_SIZE != SZ_4K)
> /* Only 4k page size on the host is supported */
> return 0;
> @@ -46,6 +321,13 @@ int kvm_init_rme(void)
> /* Continue without realm support */
> return 0;
>
> + if (WARN_ON(rmi_features(0, &rmm_feat_reg0)))
> + return 0;
> +
> + ret = rme_vmid_init();
> + if (ret)
> + return ret;
> +
> /* Future patch will enable static branch kvm_rme_is_available */
>
> return 0;
Suzuki
Hi Steven
On 12/04/2024 09:42, Steven Price wrote:
> Pages can only be populated/destroyed on the RMM at the 4KB granule,
> this requires creating the full depth of RTTs. However if the pages are
> going to be combined into a 4MB huge page the last RTT is only
minor nit: 2MB huge page.
> temporarily needed. Similarly when freeing memory the huge page must be
> temporarily split requiring temporary usage of the full depth oF RTTs.
>
> To avoid needing to perform a temporary allocation and delegation of a
> page for this purpose we keep a spare delegated page around. In
> particular this avoids the need for memory allocation while destroying
> the realm guest.
>
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/include/asm/kvm_rme.h | 5 +++++
> arch/arm64/kvm/rme.c | 8 ++++++++
> 2 files changed, 13 insertions(+)
>
> diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
> index cf8cc4d30364..fba85e9ce3ae 100644
> --- a/arch/arm64/include/asm/kvm_rme.h
> +++ b/arch/arm64/include/asm/kvm_rme.h
> @@ -50,6 +50,9 @@ enum realm_state {
> * @state: The lifetime state machine for the realm
> * @rd: Kernel mapping of the Realm Descriptor (RD)
> * @params: Parameters for the RMI_REALM_CREATE command
> + * @spare_page: A physical page that has been delegated to the Realm world but
> + * is otherwise free. Used to avoid temporary allocation during
> + * RTT operations.
> * @num_aux: The number of auxiliary pages required by the RMM
> * @vmid: VMID to be used by the RMM for the realm
> * @ia_bits: Number of valid Input Address bits in the IPA
> @@ -60,6 +63,8 @@ struct realm {
> void *rd;
> struct realm_params *params;
>
> + phys_addr_t spare_page;
> +
> unsigned long num_aux;
> unsigned int vmid;
> unsigned int ia_bits;
> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
> index 658d14e8d87d..9652ec6ab2fd 100644
> --- a/arch/arm64/kvm/rme.c
> +++ b/arch/arm64/kvm/rme.c
> @@ -103,6 +103,7 @@ static int realm_create_rd(struct kvm *kvm)
> }
>
> realm->rd = rd;
> + realm->spare_page = PHYS_ADDR_MAX;
>
> if (WARN_ON(rmi_rec_aux_count(rd_phys, &realm->num_aux))) {
> WARN_ON(rmi_realm_destroy(rd_phys));
> @@ -283,6 +284,13 @@ void kvm_destroy_realm(struct kvm *kvm)
>
> rme_vmid_release(realm->vmid);
>
> + if (realm->spare_page != PHYS_ADDR_MAX) {
> + /* Leak the page if the undelegate fails */
> + if (!WARN_ON(rmi_granule_undelegate(realm->spare_page)))
> + free_page((unsigned long)phys_to_virt(realm->spare_page));
> + realm->spare_page = PHYS_ADDR_MAX;
> + }
> +
> for (i = 0; i < pgt->pgd_pages; i++) {
> phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
>
Reviewed-by: Suzuki K Poulose <[email protected]>
On 12/04/2024 09:42, Steven Price wrote:
> Previously machine type was used purely for specifying the physical
> address size of the guest. Reserve the higher bits to specify an ARM
> specific machine type and declare a new type 'KVM_VM_TYPE_ARM_REALM'
> used to create a realm guest.
>
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/kvm/arm.c | 17 +++++++++++++++++
> arch/arm64/kvm/mmu.c | 3 ---
> include/uapi/linux/kvm.h | 19 +++++++++++++++----
> 3 files changed, 32 insertions(+), 7 deletions(-)
>
> diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
> index 22da6493912a..c5a6139d5454 100644
> --- a/arch/arm64/kvm/arm.c
> +++ b/arch/arm64/kvm/arm.c
> @@ -173,6 +173,23 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
> mutex_unlock(&kvm->lock);
> #endif
>
> + if (type & ~(KVM_VM_TYPE_ARM_MASK | KVM_VM_TYPE_ARM_IPA_SIZE_MASK))
> + return -EINVAL;
> +
> + switch (type & KVM_VM_TYPE_ARM_MASK) {
> + case KVM_VM_TYPE_ARM_NORMAL:
> + break;
> + case KVM_VM_TYPE_ARM_REALM:
> + kvm->arch.is_realm = true;
> + if (!kvm_is_realm(kvm)) {
> + /* Realm support unavailable */
> + return -EINVAL;
> + }
> + break;
> + default:
> + return -EINVAL;
> + }
> +
> ret = kvm_share_hyp(kvm, kvm + 1);
> if (ret)
> return ret;
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index aae365647b62..af4564f3add5 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -877,9 +877,6 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
> if (kvm_is_realm(kvm))
> ipa_limit = kvm_realm_ipa_limit();
>
> - if (type & ~KVM_VM_TYPE_ARM_IPA_SIZE_MASK)
> - return -EINVAL;
> -
> phys_shift = KVM_VM_TYPE_ARM_IPA_SIZE(type);
> if (is_protected_kvm_enabled()) {
> phys_shift = kvm_ipa_limit;
> diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
> index a1147036d1bd..5153c837c8c7 100644
> --- a/include/uapi/linux/kvm.h
> +++ b/include/uapi/linux/kvm.h
> @@ -635,14 +635,25 @@ struct kvm_enable_cap {
> #define KVM_S390_SIE_PAGE_OFFSET 1
>
> /*
> - * On arm64, machine type can be used to request the physical
> - * address size for the VM. Bits[7-0] are reserved for the guest
> - * PA size shift (i.e, log2(PA_Size)). For backward compatibility,
> - * value 0 implies the default IPA size, 40bits.
> + * On arm64, machine type can be used to request both the machine type and
> + * the physical address size for the VM.
> + *
> + * Bits[11-8] are reserved for the ARM specific machine type.
> + *
> + * Bits[7-0] are reserved for the guest PA size shift (i.e, log2(PA_Size)).
> + * For backward compatibility, value 0 implies the default IPA size, 40bits.
> */
> +#define KVM_VM_TYPE_ARM_SHIFT 8
> +#define KVM_VM_TYPE_ARM_MASK (0xfULL << KVM_VM_TYPE_ARM_SHIFT)
> +#define KVM_VM_TYPE_ARM(_type) \
> + (((_type) << KVM_VM_TYPE_ARM_SHIFT) & KVM_VM_TYPE_ARM_MASK)
> +#define KVM_VM_TYPE_ARM_NORMAL KVM_VM_TYPE_ARM(0)
> +#define KVM_VM_TYPE_ARM_REALM KVM_VM_TYPE_ARM(1)
> +
> #define KVM_VM_TYPE_ARM_IPA_SIZE_MASK 0xffULL
> #define KVM_VM_TYPE_ARM_IPA_SIZE(x) \
> ((x) & KVM_VM_TYPE_ARM_IPA_SIZE_MASK)
> +
> /*
> * ioctls for /dev/kvm fds:
> */
Looks good to me.
Reviewed-by: Suzuki K Poulose <[email protected]>
Hi Steven
minort nit, Subject: arm64: RME: RTT tear down
This patch is all about tearing the RTTs, so may be the subject could
be adjusted accordingly.
On 12/04/2024 09:42, Steven Price wrote:
> The RMM owns the stage 2 page tables for a realm, and KVM must request
> that the RMM creates/destroys entries as necessary. The physical pages
> to store the page tables are delegated to the realm as required, and can
> be undelegated when no longer used.
>
> Creating new RTTs is the easy part, tearing down is a little more
> tricky. The result of realm_rtt_destroy() can be used to effectively
> walk the tree and destroy the entries (undelegating pages that were
> given to the realm).
The patch looks functionally correct to me. Some minor style related
comments below.
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/include/asm/kvm_rme.h | 19 ++++
> arch/arm64/kvm/mmu.c | 6 +-
> arch/arm64/kvm/rme.c | 171 +++++++++++++++++++++++++++++++
> 3 files changed, 193 insertions(+), 3 deletions(-)
>
> diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
> index fba85e9ce3ae..4ab5cb5e91b3 100644
> --- a/arch/arm64/include/asm/kvm_rme.h
> +++ b/arch/arm64/include/asm/kvm_rme.h
> @@ -76,5 +76,24 @@ u32 kvm_realm_ipa_limit(void);
> int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap);
> int kvm_init_realm_vm(struct kvm *kvm);
> void kvm_destroy_realm(struct kvm *kvm);
> +void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits);
> +
> +#define RME_RTT_BLOCK_LEVEL 2
> +#define RME_RTT_MAX_LEVEL 3
> +
> +#define RME_PAGE_SHIFT 12
> +#define RME_PAGE_SIZE BIT(RME_PAGE_SHIFT)
> +/* See ARM64_HW_PGTABLE_LEVEL_SHIFT() */
> +#define RME_RTT_LEVEL_SHIFT(l) \
> + ((RME_PAGE_SHIFT - 3) * (4 - (l)) + 3)
> +#define RME_L2_BLOCK_SIZE BIT(RME_RTT_LEVEL_SHIFT(2))
> +
> +static inline unsigned long rme_rtt_level_mapsize(int level)
> +{
> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
> + return RME_PAGE_SIZE;
> +
> + return (1UL << RME_RTT_LEVEL_SHIFT(level));
> +}
>
super minor nit: We only support 4K for now, so may be could reuse
the ARM64 generic macro helpers. I am fine either way.
> #endif
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index af4564f3add5..46f0c4e80ace 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -1012,17 +1012,17 @@ void stage2_unmap_vm(struct kvm *kvm)
> void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu)
> {
> struct kvm *kvm = kvm_s2_mmu_to_kvm(mmu);
> - struct kvm_pgtable *pgt = NULL;
> + struct kvm_pgtable *pgt;
>
> write_lock(&kvm->mmu_lock);
> + pgt = mmu->pgt;
> if (kvm_is_realm(kvm) &&
> (kvm_realm_state(kvm) != REALM_STATE_DEAD &&
> kvm_realm_state(kvm) != REALM_STATE_NONE)) {
> - /* TODO: teardown rtts */
> write_unlock(&kvm->mmu_lock);
> + kvm_realm_destroy_rtts(kvm, pgt->ia_bits);
> return;
> }
> - pgt = mmu->pgt;
> if (pgt) {
> mmu->pgd_phys = 0;
> mmu->pgt = NULL;
> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
> index 9652ec6ab2fd..09b59bcad8b6 100644
> --- a/arch/arm64/kvm/rme.c
> +++ b/arch/arm64/kvm/rme.c
> @@ -47,6 +47,53 @@ static int rmi_check_version(void)
> return 0;
> }
>
> +static phys_addr_t __alloc_delegated_page(struct realm *realm,
> + struct kvm_mmu_memory_cache *mc,
> + gfp_t flags)
minor nit: Do we need "__" here ? The counter part is plain
free_delegated_page without the "__". We could drop the prefix
Or we could split the function as:
alloc_delegated_page()
{
if (spare_page_available)
return spare_page;
return __alloc_delegated_page(); /* Alloc and delegate a page */
}
> +{
> + phys_addr_t phys = PHYS_ADDR_MAX;
> + void *virt;
> +
> + if (realm->spare_page != PHYS_ADDR_MAX) {
> + swap(realm->spare_page, phys);
> + goto out;
> + }
> +
> + if (mc)
> + virt = kvm_mmu_memory_cache_alloc(mc);
> + else
> + virt = (void *)__get_free_page(flags);
> +
> + if (!virt)
> + goto out;
> +
> + phys = virt_to_phys(virt);
> +
> + if (rmi_granule_delegate(phys)) {
> + free_page((unsigned long)virt);
> +
> + phys = PHYS_ADDR_MAX;
> + }
> +
> +out:
> + return phys;
> +}
> +
> +static void free_delegated_page(struct realm *realm, phys_addr_t phys)
> +{
> + if (realm->spare_page == PHYS_ADDR_MAX) {
> + realm->spare_page = phys;
> + return;
> + }
> +
> + if (WARN_ON(rmi_granule_undelegate(phys))) {
> + /* Undelegate failed: leak the page */
> + return;
> + }
> +
> + free_page((unsigned long)phys_to_virt(phys));
> +}
> +
> u32 kvm_realm_ipa_limit(void)
> {
> return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
> @@ -124,6 +171,130 @@ static int realm_create_rd(struct kvm *kvm)
> return r;
> }
>
> +static int realm_rtt_destroy(struct realm *realm, unsigned long addr,
> + int level, phys_addr_t *rtt_granule,
> + unsigned long *next_addr)
> +{
> + unsigned long out_rtt;
> + unsigned long out_top;
> + int ret;
> +
> + ret = rmi_rtt_destroy(virt_to_phys(realm->rd), addr, level,
> + &out_rtt, &out_top);
> +
> + if (rtt_granule)
> + *rtt_granule = out_rtt;
> + if (next_addr)
> + *next_addr = out_top;
minor nit: As mentioned in the previous patch, we could move this check
to the rmi_rtt_destroy().
> +
> + return ret;
> +}
> +
> +static int realm_tear_down_rtt_level(struct realm *realm, int level,
> + unsigned long start, unsigned long end)
> +{
> + ssize_t map_size;
> + unsigned long addr, next_addr;
> +
> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
> + return -EINVAL;
> +
> + map_size = rme_rtt_level_mapsize(level - 1);
> +
> + for (addr = start; addr < end; addr = next_addr) {
> + phys_addr_t rtt_granule;
> + int ret;
> + unsigned long align_addr = ALIGN(addr, map_size);
> +
> + next_addr = ALIGN(addr + 1, map_size);
> +
> + if (next_addr <= end && align_addr == addr) {
> + ret = realm_rtt_destroy(realm, addr, level,
> + &rtt_granule, &next_addr);
> + } else {
> + /* Recurse a level deeper */
> + ret = realm_tear_down_rtt_level(realm,
> + level + 1,
> + addr,
> + min(next_addr, end));
> + if (ret)
> + return ret;
> + continue;
> + }
I think it would be better readable if we did something like :
/*
* The target range is smaller than what this level
* covers. Go deeper.
*/
if (next_addr > end || align_addr != addr) {
ret = realm_tear_down_rtt_level(realm,
level + 1, addr,
min(next_addr, end));
if (ret)
return ret;
continue;
}
ret = realm_rtt_destroy(realm, addr, level,
&rtt_granule, &next_addr);
> +
> + switch (RMI_RETURN_STATUS(ret)) {
> + case RMI_SUCCESS:
> + if (!WARN_ON(rmi_granule_undelegate(rtt_granule)))
> + free_page((unsigned long)phys_to_virt(rtt_granule));
> + break;
> + case RMI_ERROR_RTT:
> + if (next_addr > addr) {
> + /* unassigned or destroyed */
minor nit:
/* RTT doesn't exist, skip */
> + break;
> + }
> + if (WARN_ON(RMI_RETURN_INDEX(ret) != level))
> + return -EBUSY;
In practise, we only call this for the full IPA range and we wouldn't go
deeper, if the top level entry was missing. So, there is no reason why
the RMM didn't walk to the requested level. May be we could add a
comment here :
/*
* We tear down the RTT range for the full IPA
* space, after everything is unmapped. Also we
* descend down only if we cannot tear down a
* top level RTT. Thus RMM must be able to walk
* to the requested level. e.g., a block mapping
* exists at L1 or L2.
*/
> + if (WARN_ON(level == RME_RTT_MAX_LEVEL)) {
> + // Live entry
> + return -EBUSY;
The first part of the comment above applies to this. So may be it is
good to have it.
> + }
> + /* Recurse a level deeper */
minor nit:
/*
* The table has active entries in it, recurse
* deeper and tear down the RTTs.
*/
> + next_addr = ALIGN(addr + 1, map_size);
> + ret = realm_tear_down_rtt_level(realm,
> + level + 1,
> + addr,
> + next_addr);
> + if (ret)
> + return ret;
> + /* Try again at this level */
/*
* Now that the children RTTs are destroyed,
* retry at this level.
*/
> + next_addr = addr;
> + break;
> + default:
> + WARN_ON(1);
> + return -ENXIO;
> + }
> + }
> +
> + return 0;
> +}
> +
> +static int realm_tear_down_rtt_range(struct realm *realm,
> + unsigned long start, unsigned long end)
> +{
> + return realm_tear_down_rtt_level(realm, get_start_level(realm) + 1,
> + start, end);
> +}
> +
> +static void ensure_spare_page(struct realm *realm)
> +{
> + phys_addr_t tmp_rtt;
> +
> + /*
> + * Make sure we have a spare delegated page for tearing down the
> + * block mappings. We do this by allocating then freeing a page.
> + * We must use Atomic allocations as we are called with kvm->mmu_lock
> + * held.
> + */
> + tmp_rtt = __alloc_delegated_page(realm, NULL, GFP_ATOMIC);
> +
> + /*
> + * If the allocation failed, continue as we may not have a block level
> + * mapping so it may not be fatal, otherwise free it to assign it
> + * to the spare page.
> + */
> + if (tmp_rtt != PHYS_ADDR_MAX)
> + free_delegated_page(realm, tmp_rtt);
> +}
> +
> +void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits)
> +{
> + struct realm *realm = &kvm->arch.realm;
> +
> + ensure_spare_page(realm);
> +
> + WARN_ON(realm_tear_down_rtt_range(realm, 0, (1UL << ia_bits)));
> +}
minor nit: We don't seem to be using the "spare_page" yet in this patch.
May be a good idea to move all the related changes
(alloc_delegated_page() / free_delegated_page, ensure_spare_page() etc)
to the patch where it may be better suited ?
Suzuki
> +
> /* Protects access to rme_vmid_bitmap */
> static DEFINE_SPINLOCK(rme_vmid_lock);
> static unsigned long *rme_vmid_bitmap;
On 12/04/2024 09:42, Steven Price wrote:
> The RMM maintains a data structure known as the Realm Execution Context
> (or REC). It is similar to struct kvm_vcpu and tracks the state of the
> virtual CPUs. KVM must delegate memory and request the structures are
> created when vCPUs are created, and suitably tear down on destruction.
>
May be a good idea to add a note about the AUX granules, to help the
reader with the context.
RECs must be supplied with additional pages (AUX Granules) for storing
the larger registers state (e.g., SVE). The number of AUX granules for
a REC depends on the "parameters" with which the Realm was created.
Also the register states for the REC cannot be modified by KVM after the
REC is created.
> See Realm Management Monitor specification (DEN0137) for more information:
> https://developer.arm.com/documentation/den0137/
>
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/include/asm/kvm_emulate.h | 2 +
> arch/arm64/include/asm/kvm_host.h | 3 +
> arch/arm64/include/asm/kvm_rme.h | 18 ++++
> arch/arm64/kvm/arm.c | 2 +
> arch/arm64/kvm/reset.c | 11 ++
> arch/arm64/kvm/rme.c | 150 +++++++++++++++++++++++++++
> 6 files changed, 186 insertions(+)
>
> diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
> index c606316f4729..2209a7c6267f 100644
> --- a/arch/arm64/include/asm/kvm_emulate.h
> +++ b/arch/arm64/include/asm/kvm_emulate.h
> @@ -631,6 +631,8 @@ static inline bool kvm_realm_is_created(struct kvm *kvm)
>
> static inline bool vcpu_is_rec(struct kvm_vcpu *vcpu)
> {
> + if (static_branch_unlikely(&kvm_rme_is_available))
> + return vcpu->arch.rec.mpidr != INVALID_HWID;
> return false;
> }
>
> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> index 63b68b85db3f..f7ac40ce0caf 100644
> --- a/arch/arm64/include/asm/kvm_host.h
> +++ b/arch/arm64/include/asm/kvm_host.h
> @@ -694,6 +694,9 @@ struct kvm_vcpu_arch {
>
> /* Per-vcpu CCSIDR override or NULL */
> u32 *ccsidr;
> +
> + /* Realm meta data */
> + struct realm_rec rec;
> };
>
> /*
> diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
> index 4ab5cb5e91b3..915e76068b00 100644
> --- a/arch/arm64/include/asm/kvm_rme.h
> +++ b/arch/arm64/include/asm/kvm_rme.h
> @@ -6,6 +6,7 @@
> #ifndef __ASM_KVM_RME_H
> #define __ASM_KVM_RME_H
>
> +#include <asm/rmi_smc.h>
> #include <uapi/linux/kvm.h>
>
> /**
> @@ -70,6 +71,21 @@ struct realm {
> unsigned int ia_bits;
> };
>
> +/**
> + * struct realm_rec - Additional per VCPU data for a Realm
> + *
> + * @mpidr: MPIDR (Multiprocessor Affinity Register) value to identify this VCPU
> + * @rec_page: Kernel VA of the RMM's private page for this REC
> + * @aux_pages: Additional pages private to the RMM for this REC
> + * @run: Kernel VA of the RmiRecRun structure shared with the RMM
> + */
> +struct realm_rec {
> + unsigned long mpidr;
> + void *rec_page;
> + struct page *aux_pages[REC_PARAMS_AUX_GRANULES];
> + struct rec_run *run;
> +};
> +
> int kvm_init_rme(void);
> u32 kvm_realm_ipa_limit(void);
>
> @@ -77,6 +93,8 @@ int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap);
> int kvm_init_realm_vm(struct kvm *kvm);
> void kvm_destroy_realm(struct kvm *kvm);
> void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits);
> +int kvm_create_rec(struct kvm_vcpu *vcpu);
> +void kvm_destroy_rec(struct kvm_vcpu *vcpu);
>
> #define RME_RTT_BLOCK_LEVEL 2
> #define RME_RTT_MAX_LEVEL 3
> diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
> index c5a6139d5454..d70c511e16a0 100644
> --- a/arch/arm64/kvm/arm.c
> +++ b/arch/arm64/kvm/arm.c
> @@ -432,6 +432,8 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
> /* Force users to call KVM_ARM_VCPU_INIT */
> vcpu_clear_flag(vcpu, VCPU_INITIALIZED);
>
> + vcpu->arch.rec.mpidr = INVALID_HWID;
> +
> vcpu->arch.mmu_page_cache.gfp_zero = __GFP_ZERO;
>
> /*
> diff --git a/arch/arm64/kvm/reset.c b/arch/arm64/kvm/reset.c
> index 68d1d05672bd..6e6eb4a15095 100644
> --- a/arch/arm64/kvm/reset.c
> +++ b/arch/arm64/kvm/reset.c
> @@ -134,6 +134,11 @@ int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature)
> return -EPERM;
>
> return kvm_vcpu_finalize_sve(vcpu);
> + case KVM_ARM_VCPU_REC:
> + if (!kvm_is_realm(vcpu->kvm))
> + return -EINVAL;
> +
> + return kvm_create_rec(vcpu);
> }
>
> return -EINVAL;
> @@ -144,6 +149,11 @@ bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu)
> if (vcpu_has_sve(vcpu) && !kvm_arm_vcpu_sve_finalized(vcpu))
> return false;
>
> + if (kvm_is_realm(vcpu->kvm) &&
> + !(vcpu_is_rec(vcpu) &&
> + READ_ONCE(vcpu->kvm->arch.realm.state) == REALM_STATE_ACTIVE))
> + return false;
> +
> return true;
> }
>
> @@ -157,6 +167,7 @@ void kvm_arm_vcpu_destroy(struct kvm_vcpu *vcpu)
> kvm_unshare_hyp(sve_state, sve_state + vcpu_sve_state_size(vcpu));
> kfree(sve_state);
> kfree(vcpu->arch.ccsidr);
> + kvm_destroy_rec(vcpu);
> }
>
> static void kvm_vcpu_reset_sve(struct kvm_vcpu *vcpu)
> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
> index 09b59bcad8b6..629a095bea61 100644
> --- a/arch/arm64/kvm/rme.c
> +++ b/arch/arm64/kvm/rme.c
> @@ -474,6 +474,156 @@ void kvm_destroy_realm(struct kvm *kvm)
> kvm_free_stage2_pgd(&kvm->arch.mmu);
> }
>
> +static void free_rec_aux(struct page **aux_pages,
> + unsigned int num_aux)
> +{
> + unsigned int i;
> +
> + for (i = 0; i < num_aux; i++) {
> + phys_addr_t aux_page_phys = page_to_phys(aux_pages[i]);
> +
> + /* If the undelegate fails then leak the page */
> + if (WARN_ON(rmi_granule_undelegate(aux_page_phys)))
> + continue;
> +
> + __free_page(aux_pages[i]);
> + }
> +}
> +
> +static int alloc_rec_aux(struct page **aux_pages,
> + u64 *aux_phys_pages,
> + unsigned int num_aux)
> +{
> + int ret;
> + unsigned int i;
> +
> + for (i = 0; i < num_aux; i++) {
> + struct page *aux_page;
> + phys_addr_t aux_page_phys;
> +
> + aux_page = alloc_page(GFP_KERNEL);
> + if (!aux_page) {
> + ret = -ENOMEM;
> + goto out_err;
> + }
> + aux_page_phys = page_to_phys(aux_page);
> + if (rmi_granule_delegate(aux_page_phys)) {
> + __free_page(aux_page);
> + ret = -ENXIO;
> + goto out_err;
> + }
> + aux_pages[i] = aux_page;
> + aux_phys_pages[i] = aux_page_phys;
> + }
> +
> + return 0;
> +out_err:
> + free_rec_aux(aux_pages, i);
> + return ret;
> +}
> +
> +int kvm_create_rec(struct kvm_vcpu *vcpu)
> +{
> + struct user_pt_regs *vcpu_regs = vcpu_gp_regs(vcpu);
> + unsigned long mpidr = kvm_vcpu_get_mpidr_aff(vcpu);
> + struct realm *realm = &vcpu->kvm->arch.realm;
> + struct realm_rec *rec = &vcpu->arch.rec;
> + unsigned long rec_page_phys;
> + struct rec_params *params;
> + int r, i;
> +
> + if (kvm_realm_state(vcpu->kvm) != REALM_STATE_NEW)
> + return -ENOENT;
> +
> + /*
> + * The RMM will report PSCI v1.0 to Realms and the KVM_ARM_VCPU_PSCI_0_2
> + * flag covers v0.2 and onwards.
> + */
> + if (!vcpu_has_feature(vcpu, KVM_ARM_VCPU_PSCI_0_2))
> + return -EINVAL;
> +
> + BUILD_BUG_ON(sizeof(*params) > PAGE_SIZE);
> + BUILD_BUG_ON(sizeof(*rec->run) > PAGE_SIZE);
> +
> + params = (struct rec_params *)get_zeroed_page(GFP_KERNEL);
> + rec->rec_page = (void *)__get_free_page(GFP_KERNEL);
> + rec->run = (void *)get_zeroed_page(GFP_KERNEL);
> + if (!params || !rec->rec_page || !rec->run) {
> + r = -ENOMEM;
> + goto out_free_pages;
> + }
> +
> + for (i = 0; i < ARRAY_SIZE(params->gprs); i++)
> + params->gprs[i] = vcpu_regs->regs[i];
> +
> + params->pc = vcpu_regs->pc;
> +
> + if (vcpu->vcpu_id == 0)
> + params->flags |= REC_PARAMS_FLAG_RUNNABLE;
> +
> + rec_page_phys = virt_to_phys(rec->rec_page);
> +
> + if (rmi_granule_delegate(rec_page_phys)) {
> + r = -ENXIO;
> + goto out_free_pages;
> + }
> +
> + r = alloc_rec_aux(rec->aux_pages, params->aux, realm->num_aux);
> + if (r)
> + goto out_undelegate_rmm_rec;
> +
> + params->num_rec_aux = realm->num_aux;
> + params->mpidr = mpidr;
> +
> + if (rmi_rec_create(virt_to_phys(realm->rd),
> + rec_page_phys,
> + virt_to_phys(params))) {
> + r = -ENXIO;
> + goto out_free_rec_aux;
> + }
> +
> + rec->mpidr = mpidr;
> +
> + free_page((unsigned long)params);
> + return 0;
> +
> +out_free_rec_aux:
> + free_rec_aux(rec->aux_pages, realm->num_aux);
> +out_undelegate_rmm_rec:
> + if (WARN_ON(rmi_granule_undelegate(rec_page_phys)))
> + rec->rec_page = NULL;
> +out_free_pages:
> + free_page((unsigned long)rec->run);
> + free_page((unsigned long)rec->rec_page);
> + free_page((unsigned long)params);
> + return r;
> +}
> +
> +void kvm_destroy_rec(struct kvm_vcpu *vcpu)
> +{
> + struct realm *realm = &vcpu->kvm->arch.realm;
> + struct realm_rec *rec = &vcpu->arch.rec;
> + unsigned long rec_page_phys;
> +
> + if (!vcpu_is_rec(vcpu))
> + return;
> +
> + rec_page_phys = virt_to_phys(rec->rec_page);
> +
> + /* If the REC destroy fails, leak all pages relating to the REC */
minor nit: May be we could clarify the situation of AUX granules.
/*
* We cannot reclaim the REC page and any AUX pages
* until the REC is destroyed. So, if we fail to destroy
* the REC, leak the REC and AUX pages.
*/
> + if (WARN_ON(rmi_rec_destroy(rec_page_phys)))
> + return;
> +
> + free_rec_aux(rec->aux_pages, realm->num_aux);
> +
> + /* If the undelegate fails then leak the REC page */
> + if (WARN_ON(rmi_granule_undelegate(rec_page_phys)))
> + return;
> +
> + free_page((unsigned long)rec->rec_page);
> + free_page((unsigned long)rec->run);
I think this can be freed irrespective of the delegated pages.
So, may be we could do move it to the top.
> +}
> +
Suzuki
> int kvm_init_realm_vm(struct kvm *kvm)
> {
> struct realm_params *params;
On 12/04/2024 09:42, Steven Price wrote:
> The RMM keeps track of the timer while the realm REC is running, but on
> exit to the normal world KVM is responsible for handling the timers.
>
minor nit: It may be worth mentioning this will be hooked in, when we
add the Realm exit handling.
Otherwise looks good to me.
Suzuki
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/kvm/arch_timer.c | 45 ++++++++++++++++++++++++++++++++----
> include/kvm/arm_arch_timer.h | 2 ++
> 2 files changed, 43 insertions(+), 4 deletions(-)
>
> diff --git a/arch/arm64/kvm/arch_timer.c b/arch/arm64/kvm/arch_timer.c
> index 879982b1cc73..0b2be34a9ba3 100644
> --- a/arch/arm64/kvm/arch_timer.c
> +++ b/arch/arm64/kvm/arch_timer.c
> @@ -162,6 +162,13 @@ static void timer_set_cval(struct arch_timer_context *ctxt, u64 cval)
>
> static void timer_set_offset(struct arch_timer_context *ctxt, u64 offset)
> {
> + struct kvm_vcpu *vcpu = ctxt->vcpu;
> +
> + if (kvm_is_realm(vcpu->kvm)) {
> + WARN_ON(offset);
> + return;
> + }
> +
> if (!ctxt->offset.vm_offset) {
> WARN(offset, "timer %ld\n", arch_timer_ctx_index(ctxt));
> return;
> @@ -460,6 +467,21 @@ static void kvm_timer_update_irq(struct kvm_vcpu *vcpu, bool new_level,
> }
> }
>
> +void kvm_realm_timers_update(struct kvm_vcpu *vcpu)
> +{
> + struct arch_timer_cpu *arch_timer = &vcpu->arch.timer_cpu;
> + int i;
> +
> + for (i = 0; i < NR_KVM_EL0_TIMERS; i++) {
> + struct arch_timer_context *timer = &arch_timer->timers[i];
> + bool status = timer_get_ctl(timer) & ARCH_TIMER_CTRL_IT_STAT;
> + bool level = kvm_timer_irq_can_fire(timer) && status;
> +
> + if (level != timer->irq.level)
> + kvm_timer_update_irq(vcpu, level, timer);
> + }
> +}
> +
> /* Only called for a fully emulated timer */
> static void timer_emulate(struct arch_timer_context *ctx)
> {
> @@ -831,6 +853,8 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu)
> if (unlikely(!timer->enabled))
> return;
>
> + kvm_timer_unblocking(vcpu);
> +
> get_timer_map(vcpu, &map);
>
> if (static_branch_likely(&has_gic_active_state)) {
> @@ -844,8 +868,6 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu)
> kvm_timer_vcpu_load_nogic(vcpu);
> }
>
> - kvm_timer_unblocking(vcpu);
> -
> timer_restore_state(map.direct_vtimer);
> if (map.direct_ptimer)
> timer_restore_state(map.direct_ptimer);
> @@ -988,7 +1010,9 @@ static void timer_context_init(struct kvm_vcpu *vcpu, int timerid)
>
> ctxt->vcpu = vcpu;
>
> - if (timerid == TIMER_VTIMER)
> + if (kvm_is_realm(vcpu->kvm))
> + ctxt->offset.vm_offset = NULL;
> + else if (timerid == TIMER_VTIMER)
> ctxt->offset.vm_offset = &kvm->arch.timer_data.voffset;
> else
> ctxt->offset.vm_offset = &kvm->arch.timer_data.poffset;
> @@ -1011,13 +1035,19 @@ static void timer_context_init(struct kvm_vcpu *vcpu, int timerid)
> void kvm_timer_vcpu_init(struct kvm_vcpu *vcpu)
> {
> struct arch_timer_cpu *timer = vcpu_timer(vcpu);
> + u64 cntvoff;
>
> for (int i = 0; i < NR_KVM_TIMERS; i++)
> timer_context_init(vcpu, i);
>
> + if (kvm_is_realm(vcpu->kvm))
> + cntvoff = 0;
> + else
> + cntvoff = kvm_phys_timer_read();
> +
> /* Synchronize offsets across timers of a VM if not already provided */
> if (!test_bit(KVM_ARCH_FLAG_VM_COUNTER_OFFSET, &vcpu->kvm->arch.flags)) {
> - timer_set_offset(vcpu_vtimer(vcpu), kvm_phys_timer_read());
> + timer_set_offset(vcpu_vtimer(vcpu), cntvoff);
> timer_set_offset(vcpu_ptimer(vcpu), 0);
> }
>
> @@ -1525,6 +1555,13 @@ int kvm_timer_enable(struct kvm_vcpu *vcpu)
> return -EINVAL;
> }
>
> + /*
> + * We don't use mapped IRQs for Realms because the RMI doesn't allow
> + * us setting the LR.HW bit in the VGIC.
> + */
> + if (vcpu_is_rec(vcpu))
> + return 0;
> +
> get_timer_map(vcpu, &map);
>
> ret = kvm_vgic_map_phys_irq(vcpu,
> diff --git a/include/kvm/arm_arch_timer.h b/include/kvm/arm_arch_timer.h
> index c819c5d16613..d8ab297560d0 100644
> --- a/include/kvm/arm_arch_timer.h
> +++ b/include/kvm/arm_arch_timer.h
> @@ -112,6 +112,8 @@ int kvm_arm_timer_set_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr);
> int kvm_arm_timer_get_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr);
> int kvm_arm_timer_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr);
>
> +void kvm_realm_timers_update(struct kvm_vcpu *vcpu);
> +
> u64 kvm_phys_timer_read(void);
>
> void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu);
On 16/04/2024 13:38, Suzuki K Poulose wrote:
> Hi Steven
>
> On 12/04/2024 09:42, Steven Price wrote:
>> The RMM (Realm Management Monitor) provides functionality that can be
>> accessed by SMC calls from the host.
>>
>> The SMC definitions are based on DEN0137[1] version 1.0-eac5
>>
>> [1] https://developer.arm.com/documentation/den0137/1-0eac5/
>>
>> Signed-off-by: Steven Price <[email protected]>
>> ---
>> arch/arm64/include/asm/rmi_smc.h | 250 +++++++++++++++++++++++++++++++
>> 1 file changed, 250 insertions(+)
>> create mode 100644 arch/arm64/include/asm/rmi_smc.h
>>
>> diff --git a/arch/arm64/include/asm/rmi_smc.h
>> b/arch/arm64/include/asm/rmi_smc.h
>> new file mode 100644
>> index 000000000000..c205efdb18d8
>> --- /dev/null
>> +++ b/arch/arm64/include/asm/rmi_smc.h
>> @@ -0,0 +1,250 @@
>> +/* SPDX-License-Identifier: GPL-2.0 */
>> +/*
>> + * Copyright (C) 2023 ARM Ltd.
>> + *
>> + * The values and structures in this file are from the Realm
>> Management Monitor
>> + * specification (DEN0137) version A-bet0:
>> + * https://developer.arm.com/documentation/den0137/1-0bet0/
>
> This should now point to eac5 instead.
Typical - I searched through the commit logs, but forgot I'd put a
reference in the code too! Thanks for spotting.
>> + */
>> +
>> +#ifndef __ASM_RME_SMC_H
>> +#define __ASM_RME_SMC_H
>> +
>> +#include <linux/arm-smccc.h>
>> +
>> +#define SMC_RxI_CALL(func) \
>> + ARM_SMCCC_CALL_VAL(ARM_SMCCC_FAST_CALL, \
>> + ARM_SMCCC_SMC_64, \
>> + ARM_SMCCC_OWNER_STANDARD, \
>> + (func))
>> +
>> +#define SMC_RMI_DATA_CREATE SMC_RxI_CALL(0x0153)
>> +#define SMC_RMI_DATA_CREATE_UNKNOWN SMC_RxI_CALL(0x0154)
>> +#define SMC_RMI_DATA_DESTROY SMC_RxI_CALL(0x0155)
>> +#define SMC_RMI_FEATURES SMC_RxI_CALL(0x0165)
>> +#define SMC_RMI_GRANULE_DELEGATE SMC_RxI_CALL(0x0151)
>> +#define SMC_RMI_GRANULE_UNDELEGATE SMC_RxI_CALL(0x0152)
>> +#define SMC_RMI_PSCI_COMPLETE SMC_RxI_CALL(0x0164)
>> +#define SMC_RMI_REALM_ACTIVATE SMC_RxI_CALL(0x0157)
>> +#define SMC_RMI_REALM_CREATE SMC_RxI_CALL(0x0158)
>> +#define SMC_RMI_REALM_DESTROY SMC_RxI_CALL(0x0159)
>> +#define SMC_RMI_REC_AUX_COUNT SMC_RxI_CALL(0x0167)
>> +#define SMC_RMI_REC_CREATE SMC_RxI_CALL(0x015a)
>> +#define SMC_RMI_REC_DESTROY SMC_RxI_CALL(0x015b)
>> +#define SMC_RMI_REC_ENTER SMC_RxI_CALL(0x015c)
>> +#define SMC_RMI_RTT_CREATE SMC_RxI_CALL(0x015d)
>> +#define SMC_RMI_RTT_DESTROY SMC_RxI_CALL(0x015e)
>> +#define SMC_RMI_RTT_FOLD SMC_RxI_CALL(0x0166)
>> +#define SMC_RMI_RTT_INIT_RIPAS SMC_RxI_CALL(0x0168)
>> +#define SMC_RMI_RTT_MAP_UNPROTECTED SMC_RxI_CALL(0x015f)
>> +#define SMC_RMI_RTT_READ_ENTRY SMC_RxI_CALL(0x0161)
>> +#define SMC_RMI_RTT_SET_RIPAS SMC_RxI_CALL(0x0169)
>> +#define SMC_RMI_RTT_UNMAP_UNPROTECTED SMC_RxI_CALL(0x0162)
>> +#define SMC_RMI_VERSION SMC_RxI_CALL(0x0150)
>> +
>> +#define RMI_ABI_MAJOR_VERSION 1
>> +#define RMI_ABI_MINOR_VERSION 0
>> +
>> +#define RMI_UNASSIGNED 0
>> +#define RMI_ASSIGNED 1
>> +#define RMI_TABLE 2
>> +
>> +#define RMI_ABI_VERSION_GET_MAJOR(version) ((version) >> 16)
>> +#define RMI_ABI_VERSION_GET_MINOR(version) ((version) & 0xFFFF)
>> +#define RMI_ABI_VERSION(major, minor) (((major) << 16) | (minor))
>> +
>> +#define RMI_RETURN_STATUS(ret) ((ret) & 0xFF)
>> +#define RMI_RETURN_INDEX(ret) (((ret) >> 8) & 0xFF)
>> +
>> +#define RMI_SUCCESS 0
>> +#define RMI_ERROR_INPUT 1
>> +#define RMI_ERROR_REALM 2
>> +#define RMI_ERROR_REC 3
>> +#define RMI_ERROR_RTT 4
>> +
>> +#define RMI_EMPTY 0
>> +#define RMI_RAM 1
>> +#define RMI_DESTROYED 2
>> +
>> +#define RMI_NO_MEASURE_CONTENT 0
>> +#define RMI_MEASURE_CONTENT 1
>> +
>> +#define RMI_FEATURE_REGISTER_0_S2SZ GENMASK(7, 0)
>> +#define RMI_FEATURE_REGISTER_0_LPA2 BIT(8)
>> +#define RMI_FEATURE_REGISTER_0_SVE_EN BIT(9)
>> +#define RMI_FEATURE_REGISTER_0_SVE_VL GENMASK(13, 10)
>> +#define RMI_FEATURE_REGISTER_0_NUM_BPS GENMASK(17, 14)
>> +#define RMI_FEATURE_REGISTER_0_NUM_WPS GENMASK(21, 18)
>> +#define RMI_FEATURE_REGISTER_0_PMU_EN BIT(22)
>> +#define RMI_FEATURE_REGISTER_0_PMU_NUM_CTRS GENMASK(27, 23)
>> +#define RMI_FEATURE_REGISTER_0_HASH_SHA_256 BIT(28)
>> +#define RMI_FEATURE_REGISTER_0_HASH_SHA_512 BIT(29)
>> +
>> +#define RMI_REALM_PARAM_FLAG_LPA2 BIT(0)
>> +#define RMI_REALM_PARAM_FLAG_SVE BIT(1)
>> +#define RMI_REALM_PARAM_FLAG_PMU BIT(2)
>> +
>> +/*
>> + * Note many of these fields are smaller than u64 but all fields have
>> u64
>> + * alignment, so use u64 to ensure correct alignment.
>> + */
>> +struct realm_params {
>> + union { /* 0x0 */
>> + struct {
>> + u64 flags;
>> + u64 s2sz;
>> + u64 sve_vl;
>> + u64 num_bps;
>> + u64 num_wps;
>> + u64 pmu_num_ctrs;
>> + u64 hash_algo;
>> + };
>> + u8 padding_1[0x400];
>> + };
>> + union { /* 0x400 */
>> + u8 rpv[64];
>> + u8 padding_2[0x400];
>> + };
>> + union { /* 0x800 */
>> + struct {
>> + u64 vmid;
>> + u64 rtt_base;
>> + s64 rtt_level_start;
>> + u64 rtt_num_start;
>> + };
>> + u8 padding_3[0x800];
>> + };
>> +};
>> +
>> +/*
>> + * The number of GPRs (starting from X0) that are
>> + * configured by the host when a REC is created.
>> + */
>> +#define REC_CREATE_NR_GPRS 8
>> +
>> +#define REC_PARAMS_FLAG_RUNNABLE BIT_ULL(0)
>> +
>> +#define REC_PARAMS_AUX_GRANULES 16
>> +
>> +struct rec_params {
>> + union { /* 0x0 */
>> + u64 flags;
>> + u8 padding1[0x100];
>> + };
>> + union { /* 0x100 */
>> + u64 mpidr;
>> + u8 padding2[0x100];
>> + };
>> + union { /* 0x200 */
>> + u64 pc;
>> + u8 padding3[0x100];
>> + };
>> + union { /* 0x300 */
>> + u64 gprs[REC_CREATE_NR_GPRS];
>> + u8 padding4[0x500];
>> + };
>> + union { /* 0x800 */
>> + struct {
>> + u64 num_rec_aux;
>> + u64 aux[REC_PARAMS_AUX_GRANULES];
>> + };
>> + u8 padding5[0x800];
>> + };
>> +};
>> +
>> +#define RMI_EMULATED_MMIO BIT(0)
>> +#define RMI_INJECT_SEA BIT(1)
>> +#define RMI_TRAP_WFI BIT(2)
>> +#define RMI_TRAP_WFE BIT(3)
>
> For completeness, we could add :
>
> #define RMI_RIPAS_RESPONSE BIT(4)
>
> Not sure if we use it later in the series.
Yes, I'll add for completeness. Currently KVM will never reject a RIPAS
change request from the guest. I'm not sure in what situation it would
make sense to do such a thing. The current uABI doesn't allow the VMM to
have a say in it either as the RIPAS change is completed before the exit
to the VMM. The expectation is therefore that the VMM would simply
terminate a Realm guest that attempted a RIPAS change that it disagreed
with.
>> +
>> +#define REC_RUN_GPRS 31
>> +#define REC_GIC_NUM_LRS 16
>> +
>> +struct rec_entry {
While I'm reading this (and the spec) again - I notice that the spec
says "RecEnter" not 'entry' - I'll rename this to be consistent.
>> + union { /* 0x000 */
>> + u64 flags;
>> + u8 padding0[0x200];
>> + };
>> + union { /* 0x200 */
>> + u64 gprs[REC_RUN_GPRS];
>> + u8 padding2[0x100];
>> + };
>> + union { /* 0x300 */
>> + struct {
>> + u64 gicv3_hcr;
>> + u64 gicv3_lrs[REC_GIC_NUM_LRS];
>> + };
>> + u8 padding3[0x100];
>> + };
>> + u8 padding4[0x400];
>> +};
>> +
>> +struct rec_exit {
>> + union { /* 0x000 */
>> + u8 exit_reason;
>> + u8 padding0[0x100];
>> + };
>> + union { /* 0x100 */
>> + struct {
>> + u64 esr;
>> + u64 far;
>> + u64 hpfar;
>> + };
>> + u8 padding1[0x100];
>> + };
>> + union { /* 0x200 */
>> + u64 gprs[REC_RUN_GPRS];
>> + u8 padding2[0x100];
>> + };
>> + union { /* 0x300 */
>> + struct {
>> + u64 gicv3_hcr;
>> + u64 gicv3_lrs[REC_GIC_NUM_LRS];
>> + u64 gicv3_misr;
>> + u64 gicv3_vmcr;
>> + };
>> + u8 padding3[0x100];
>> + };
>> + union { /* 0x400 */
>> + struct {
>> + u64 cntp_ctl;
>> + u64 cntp_cval;
>> + u64 cntv_ctl;
>> + u64 cntv_cval;
>> + };
>> + u8 padding4[0x100];
>> + };
>> + union { /* 0x500 */
>> + struct {
>> + u64 ripas_base;
>> + u64 ripas_top;
>> + u64 ripas_value;
>> + };
>> + u8 padding5[0x100];
>> + };
>> + union { /* 0x600 */
>> + u16 imm;
>> + u8 padding6[0x100];
>> + };
>> + union { /* 0x700 */
>> + struct {
>> + u64 pmu_ovf_status;
>
> This is u8 as per section B4.4.10 RmiPmuOverflowStatus type.
Indeed - I'm not sure where I got u64 from - it was probably to provide
padding in an older version of the spec.
>> + };
>> + u8 padding7[0x100];
>> + };
>> +};
>> +
>> +struct rec_run {
>> + struct rec_entry entry;
>> + struct rec_exit exit;
>> +};
>> +
>> +#define RMI_EXIT_SYNC 0x00
>> +#define RMI_EXIT_IRQ 0x01
>> +#define RMI_EXIT_FIQ 0x02
>> +#define RMI_EXIT_PSCI 0x03
>> +#define RMI_EXIT_RIPAS_CHANGE 0x04
>> +#define RMI_EXIT_HOST_CALL 0x05
>> +#define RMI_EXIT_SERROR 0x06
>
> Minor nit: Like the other definitions, it may be good to keep the
> defintions of the "exit_reason" above the field declaration.
Yes, makes sense - I'll move these.
Thanks for the review!
Steve
>
> Rest looks fine to me.
>
> Suzuki
>> +
>> +#endif
>
On 16/04/2024 12:17, Suzuki K Poulose wrote:
> On 12/04/2024 09:42, Steven Price wrote:
>> If the host attempts to access granules that have been delegated for use
>> in a realm these accesses will be caught and will trigger a Granule
>> Protection Fault (GPF).
>>
>> A fault during a page walk signals a bug in the kernel and is handled by
>> oopsing the kernel. A non-page walk fault could be caused by user space
>> having access to a page which has been delegated to the kernel and will
>> trigger a SIGBUS to allow debugging why user space is trying to access a
>> delegated page.
>>
>> Signed-off-by: Steven Price <[email protected]>
>> ---
>> arch/arm64/mm/fault.c | 29 ++++++++++++++++++++++++-----
>> 1 file changed, 24 insertions(+), 5 deletions(-)
>>
>> diff --git a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c
>> index 8251e2fea9c7..91da0f446dd9 100644
>> --- a/arch/arm64/mm/fault.c
>> +++ b/arch/arm64/mm/fault.c
>> @@ -765,6 +765,25 @@ static int do_tag_check_fault(unsigned long far,
>> unsigned long esr,
>> return 0;
>> }
>> +static int do_gpf_ptw(unsigned long far, unsigned long esr, struct
>> pt_regs *regs)
>> +{
>> + const struct fault_info *inf = esr_to_fault_info(esr);
>> +
>> + die_kernel_fault(inf->name, far, esr, regs);
>> + return 0;
>> +}
>> +
>> +static int do_gpf(unsigned long far, unsigned long esr, struct
>> pt_regs *regs)
>> +{
>> + const struct fault_info *inf = esr_to_fault_info(esr);
>> +
>> + if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
>> + return 0;
>> +
>> + arm64_notify_die(inf->name, regs, inf->sig, inf->code, far, esr);
>> + return 0;
>> +}
>> +
>> static const struct fault_info fault_info[] = {
>> { do_bad, SIGKILL, SI_KERNEL, "ttbr address size
>> fault" },
>> { do_bad, SIGKILL, SI_KERNEL, "level 1 address size
>> fault" },
>> @@ -802,11 +821,11 @@ static const struct fault_info fault_info[] = {
>> { do_alignment_fault, SIGBUS, BUS_ADRALN, "alignment
>> fault" },
>> { do_bad, SIGKILL, SI_KERNEL, "unknown 34" },
>> { do_bad, SIGKILL, SI_KERNEL, "unknown 35" },
>
> Should this also be converted to do_gpf_ptw, "GPF at level -1", given we
> support LPA2 ?
Ah, yes I somehow missed that. Although something has gone majorly wrong
if this triggers! ;)
Steve
>> - { do_bad, SIGKILL, SI_KERNEL, "unknown 36" },
>> - { do_bad, SIGKILL, SI_KERNEL, "unknown 37" },
>> - { do_bad, SIGKILL, SI_KERNEL, "unknown 38" },
>> - { do_bad, SIGKILL, SI_KERNEL, "unknown 39" },
>> - { do_bad, SIGKILL, SI_KERNEL, "unknown 40" },
>> + { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection
>> Fault at level 0" },
>> + { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection
>> Fault at level 1" },
>> + { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection
>> Fault at level 2" },
>> + { do_gpf_ptw, SIGKILL, SI_KERNEL, "Granule Protection
>> Fault at level 3" },
>> + { do_gpf, SIGBUS, SI_KERNEL, "Granule Protection Fault
>> not on table walk" },
>> { do_bad, SIGKILL, SI_KERNEL, "level -1 address size
>> fault" },
>> { do_bad, SIGKILL, SI_KERNEL, "unknown 42" },
>> { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level -1
>> translation fault" },
>
>
> Rest looks fine to me.
>
> Suzuki
On 12/04/2024 09:42, Steven Price wrote:
> Add the KVM_CAP_ARM_RME_CREATE_FD ioctl to create a realm. This involves
> delegating pages to the RMM to hold the Realm Descriptor (RD) and for
> the base level of the Realm Translation Tables (RTT). A VMID also need
> to be picked, since the RMM has a separate VMID address space a
> dedicated allocator is added for this purpose.
>
> KVM_CAP_ARM_RME_CONFIG_REALM is provided to allow configuring the realm
> before it is created.
>
> Co-developed-by: Suzuki K Poulose <[email protected]>
> Signed-off-by: Suzuki K Poulose <[email protected]>
> Signed-off-by: Steven Price <[email protected]>
> Signed-off-by: Jean-Philippe Brucker <[email protected]>
> ---
> arch/arm64/include/asm/kvm_emulate.h | 5 +
> arch/arm64/include/asm/kvm_rme.h | 19 ++
> arch/arm64/kvm/arm.c | 18 ++
> arch/arm64/kvm/mmu.c | 15 +-
> arch/arm64/kvm/rme.c | 282 +++++++++++++++++++++++++++
> 5 files changed, 337 insertions(+), 2 deletions(-)
>
> @@ -1014,6 +1018,13 @@ void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu)
> struct kvm_pgtable *pgt = NULL;
>
> write_lock(&kvm->mmu_lock);
> + if (kvm_is_realm(kvm) &&
> + (kvm_realm_state(kvm) != REALM_STATE_DEAD &&
> + kvm_realm_state(kvm) != REALM_STATE_NONE)) {
> + /* TODO: teardown rtts */
> + write_unlock(&kvm->mmu_lock);
> + return;
> + }
> pgt = mmu->pgt;
> if (pgt) {
> mmu->pgd_phys = 0;
See my comment below.
..
> +
> +void kvm_destroy_realm(struct kvm *kvm)
> +{
..
> + for (i = 0; i < pgt->pgd_pages; i++) {
> + phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
> +
> + if (WARN_ON(rmi_granule_undelegate(pgd_phys)))
> + return;
I think we need to either:
a. memset() the root RTT pages to 0 here.
OR
b. for Realms, avoid walking the page table triggered via
kvm_pgtable_stage2_destroy()->kvm_pgtable_walk().
Even though the root RTTs are all empty (invalid entries, written using
RMM's memory encryption.), the Host might be seeing "garbage" which
might look like "valid" entries and thus triggering crashes.
I prefer not walking the RTTs for a Realm and thus simply skip the walk.
Suzuki
> + }
> +
> + WRITE_ONCE(realm->state, REALM_STATE_DEAD);
> +
> + kvm_free_stage2_pgd(&kvm->arch.mmu);
> +}
> +
> +int kvm_init_realm_vm(struct kvm *kvm)
> +{
> + struct realm_params *params;
> +
> + params = (struct realm_params *)get_zeroed_page(GFP_KERNEL);
> + if (!params)
> + return -ENOMEM;
> +
> + /* Default parameters, not exposed to user space */
> + params->s2sz = VTCR_EL2_IPA(kvm->arch.mmu.vtcr);
> + kvm->arch.realm.params = params;
> + return 0;
> +}
> +
> int kvm_init_rme(void)
> {
> + int ret;
> +
> if (PAGE_SIZE != SZ_4K)
> /* Only 4k page size on the host is supported */
> return 0;
> @@ -46,6 +321,13 @@ int kvm_init_rme(void)
> /* Continue without realm support */
> return 0;
>
> + if (WARN_ON(rmi_features(0, &rmm_feat_reg0)))
> + return 0;
> +
> + ret = rme_vmid_init();
> + if (ret)
> + return ret;
> +
> /* Future patch will enable static branch kvm_rme_is_available */
>
> return 0;
On 14/04/2024 00:44, kernel test robot wrote:
> Hi Steven,
>
> kernel test robot noticed the following build errors:
>
> [auto build test ERROR on kvmarm/next]
> [also build test ERROR on kvm/queue arm64/for-next/core linus/master v6.9-rc3 next-20240412]
> [cannot apply to kvm/linux-next]
> [If your patch is applied to the wrong git tree, kindly drop us a note.
> And when submitting patch, we suggest to use '--base' as documented in
> https://git-scm.com/docs/git-format-patch#_base_tree_information]
>
> url: https://github.com/intel-lab-lkp/linux/commits/Steven-Price/KVM-Prepare-for-handling-only-shared-mappings-in-mmu_notifier-events/20240412-170311
> base: https://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm.git next
> patch link: https://lore.kernel.org/r/20240412084309.1733783-34-steven.price%40arm.com
> patch subject: [PATCH v2 33/43] arm64: rme: Enable PMU support with a realm guest
> config: arm64-randconfig-r064-20240414 (https://download.01.org/0day-ci/archive/20240414/[email protected]/config)
> compiler: clang version 19.0.0git (https://github.com/llvm/llvm-project 8b3b4a92adee40483c27f26c478a384cd69c6f05)
> reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240414/[email protected]/reproduce)
>
> If you fix the issue in a separate patch/commit (i.e. not just a new version of
> the same patch/commit), kindly add following tags
> | Reported-by: kernel test robot <[email protected]>
> | Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
I guess the problem is with CONFIG_HW_PERF_EVENT not set, arm_pmu is an
empty struct, triggering all these errors.
Suzuki
On 12/04/2024 09:42, Steven Price wrote:
> Each page within the protected region of the realm guest can be marked
> as either RAM or EMPTY. Allow the VMM to control this before the guest
> has started and provide the equivalent functions to change this (with
> the guest's approval) at runtime.
>
> When transitioning from RIPAS RAM (1) to RIPAS EMPTY (0) the memory is
> unmapped from the guest and undelegated allowing the memory to be reused
> by the host. When transitioning to RIPAS RAM the actual population of
> the leaf RTTs is done later on stage 2 fault, however it may be
> necessary to allocate additional RTTs to represent the range requested.
minor nit: To give a bit more context:
"however it may be necessary to allocate additional RTTs in order for
the RMM to track the RIPAS for the requested range".
>
> When freeing a block mapping it is necessary to temporarily unfold the
> RTT which requires delegating an extra page to the RMM, this page can
> then be recovered once the contents of the block mapping have been
> freed. A spare, delegated page (spare_page) is used for this purpose.
>
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/include/asm/kvm_rme.h | 16 ++
> arch/arm64/kvm/mmu.c | 8 +-
> arch/arm64/kvm/rme.c | 390 +++++++++++++++++++++++++++++++
> 3 files changed, 411 insertions(+), 3 deletions(-)
>
> diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
> index 915e76068b00..cc8f81cfc3c0 100644
> --- a/arch/arm64/include/asm/kvm_rme.h
> +++ b/arch/arm64/include/asm/kvm_rme.h
> @@ -96,6 +96,14 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits);
> int kvm_create_rec(struct kvm_vcpu *vcpu);
> void kvm_destroy_rec(struct kvm_vcpu *vcpu);
>
> +void kvm_realm_unmap_range(struct kvm *kvm,
> + unsigned long ipa,
> + u64 size,
> + bool unmap_private);
> +int realm_set_ipa_state(struct kvm_vcpu *vcpu,
> + unsigned long addr, unsigned long end,
> + unsigned long ripas);
> +
> #define RME_RTT_BLOCK_LEVEL 2
> #define RME_RTT_MAX_LEVEL 3
>
> @@ -114,4 +122,12 @@ static inline unsigned long rme_rtt_level_mapsize(int level)
> return (1UL << RME_RTT_LEVEL_SHIFT(level));
> }
>
> +static inline bool realm_is_addr_protected(struct realm *realm,
> + unsigned long addr)
> +{
> + unsigned int ia_bits = realm->ia_bits;
> +
> + return !(addr & ~(BIT(ia_bits - 1) - 1));
> +}
> +
> #endif
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index 46f0c4e80ace..8a7b5449697f 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -310,6 +310,7 @@ static void invalidate_icache_guest_page(void *va, size_t size)
> * @start: The intermediate physical base address of the range to unmap
> * @size: The size of the area to unmap
> * @may_block: Whether or not we are permitted to block
> + * @only_shared: If true then protected mappings should not be unmapped
> *
> * Clear a range of stage-2 mappings, lowering the various ref-counts. Must
> * be called while holding mmu_lock (unless for freeing the stage2 pgd before
> @@ -317,7 +318,7 @@ static void invalidate_icache_guest_page(void *va, size_t size)
> * with things behind our backs.
> */
> static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64 size,
> - bool may_block)
> + bool may_block, bool only_shared)
> {
> struct kvm *kvm = kvm_s2_mmu_to_kvm(mmu);
> phys_addr_t end = start + size;
> @@ -330,7 +331,7 @@ static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64
>
> static void unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64 size)
> {
> - __unmap_stage2_range(mmu, start, size, true);
> + __unmap_stage2_range(mmu, start, size, true, false);
> }
>
> static void stage2_flush_memslot(struct kvm *kvm,
> @@ -1771,7 +1772,8 @@ bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range)
>
> __unmap_stage2_range(&kvm->arch.mmu, range->start << PAGE_SHIFT,
> (range->end - range->start) << PAGE_SHIFT,
> - range->may_block);
> + range->may_block,
> + range->only_shared);
>
> return false;
> }
> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
> index 629a095bea61..9e5983c51393 100644
> --- a/arch/arm64/kvm/rme.c
> +++ b/arch/arm64/kvm/rme.c
> @@ -79,6 +79,12 @@ static phys_addr_t __alloc_delegated_page(struct realm *realm,
> return phys;
> }
>
> +static phys_addr_t alloc_delegated_page(struct realm *realm,
> + struct kvm_mmu_memory_cache *mc)
> +{
> + return __alloc_delegated_page(realm, mc, GFP_KERNEL);
> +}
> +
> static void free_delegated_page(struct realm *realm, phys_addr_t phys)
> {
> if (realm->spare_page == PHYS_ADDR_MAX) {
> @@ -94,6 +100,151 @@ static void free_delegated_page(struct realm *realm, phys_addr_t phys)
> free_page((unsigned long)phys_to_virt(phys));
> }
>
> +static int realm_rtt_create(struct realm *realm,
> + unsigned long addr,
> + int level,
> + phys_addr_t phys)
> +{
> + addr = ALIGN_DOWN(addr, rme_rtt_level_mapsize(level - 1));
> + return rmi_rtt_create(virt_to_phys(realm->rd), phys, addr, level);
> +}
> +
> +static int realm_rtt_fold(struct realm *realm,
> + unsigned long addr,
> + int level,
> + phys_addr_t *rtt_granule)
> +{
> + unsigned long out_rtt;
> + int ret;
> +
> + ret = rmi_rtt_fold(virt_to_phys(realm->rd), addr, level, &out_rtt);
> +
> + if (RMI_RETURN_STATUS(ret) == RMI_SUCCESS && rtt_granule)
> + *rtt_granule = out_rtt;
> +
> + return ret;
> +}
> +
> +static int realm_destroy_protected(struct realm *realm,
> + unsigned long ipa,
> + unsigned long *next_addr)
> +{
> + unsigned long rd = virt_to_phys(realm->rd);
> + unsigned long addr;
> + phys_addr_t rtt;
> + int ret;
> +
> +loop:
> + ret = rmi_data_destroy(rd, ipa, &addr, next_addr);
> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
> + if (*next_addr > ipa)
> + return 0; /* UNASSIGNED */
> + rtt = alloc_delegated_page(realm, NULL);
> + if (WARN_ON(rtt == PHYS_ADDR_MAX))
> + return -1;
> + /* ASSIGNED - ipa is mapped as a block, so split */
> + ret = realm_rtt_create(realm, ipa,
> + RMI_RETURN_INDEX(ret) + 1, rtt);
Could we not go all the way to L3 (rather than 1 level deeper) and try
again ? That way, we are covered for block mappings at L1 (1G).
> + if (WARN_ON(ret)) {
> + free_delegated_page(realm, rtt);
> + return -1;
> + }
> + /* retry */
> + goto loop;
> + } else if (WARN_ON(ret)) {
> + return -1;
> + }
> + ret = rmi_granule_undelegate(addr);
> +
> + /*
> + * If the undelegate fails then something has gone seriously
> + * wrong: take an extra reference to just leak the page
> + */
> + if (WARN_ON(ret))
> + get_page(phys_to_page(addr));
> +
> + return 0;
> +}
> +
> +static void realm_unmap_range_shared(struct kvm *kvm,
> + int level,
> + unsigned long start,
> + unsigned long end)
> +{
> + struct realm *realm = &kvm->arch.realm;
> + unsigned long rd = virt_to_phys(realm->rd);
> + ssize_t map_size = rme_rtt_level_mapsize(level);
> + unsigned long next_addr, addr;
> + unsigned long shared_bit = BIT(realm->ia_bits - 1);
> +
> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
> + return;
> +
> + start |= shared_bit;
> + end |= shared_bit;
> +
> + for (addr = start; addr < end; addr = next_addr) {
> + unsigned long align_addr = ALIGN(addr, map_size);
> + int ret;
> +
> + next_addr = ALIGN(addr + 1, map_size);
> +
> + if (align_addr != addr || next_addr > end) {
> + /* Need to recurse deeper */
> + if (addr < align_addr)
> + next_addr = align_addr;
> + realm_unmap_range_shared(kvm, level + 1, addr,
> + min(next_addr, end));
> + continue;
> + }
> +
> + ret = rmi_rtt_unmap_unprotected(rd, addr, level, &next_addr);
minor nit: We could potentially use rmi_rtt_destroy() to tear down
shared mappings without unmapping them individually, if the range
is big enough. All such optimisations could come later though.
> + switch (RMI_RETURN_STATUS(ret)) {
> + case RMI_SUCCESS:
> + break;
> + case RMI_ERROR_RTT:
> + if (next_addr == addr) {
At this point we have a block aligned address, but the mapping is
further deep. Given, start from top to down, we implicitly handle
the case of block mappings. Not sure if that needs to be in a comment
here.
> + next_addr = ALIGN(addr + 1, map_size);
Reset to the "actual next" as it was overwritten by the RMI call.
> + realm_unmap_range_shared(kvm, level + 1, addr,
> + next_addr);
> + }
> + break;
> + default:
> + WARN_ON(1);
> + }
> + }
> +}
> +
> +static void realm_unmap_range_private(struct kvm *kvm,
> + unsigned long start,
> + unsigned long end)
> +{
> + struct realm *realm = &kvm->arch.realm;
> + ssize_t map_size = RME_PAGE_SIZE;
> + unsigned long next_addr, addr;
> +
> + for (addr = start; addr < end; addr = next_addr) {
> + int ret;
> +
> + next_addr = ALIGN(addr + 1, map_size);
> +
> + ret = realm_destroy_protected(realm, addr, &next_addr);
> +
> + if (WARN_ON(ret))
> + break;
> + }
> +}
> +
> +static void realm_unmap_range(struct kvm *kvm,
> + unsigned long start,
> + unsigned long end,
> + bool unmap_private)
> +{
> + realm_unmap_range_shared(kvm, RME_RTT_MAX_LEVEL - 1, start, end);
minor nit: We already have a helper to find a suitable start level
(defined below), may be we could use that ? And even do the rtt_destroy
optimisation for unprotected range.
> + if (unmap_private)
> + realm_unmap_range_private(kvm, start, end);
> +}
> +
> u32 kvm_realm_ipa_limit(void)
> {
> return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
> @@ -190,6 +341,30 @@ static int realm_rtt_destroy(struct realm *realm, unsigned long addr,
> return ret;
> }
>
> +static int realm_create_rtt_levels(struct realm *realm,
> + unsigned long ipa,
> + int level,
> + int max_level,
> + struct kvm_mmu_memory_cache *mc)
> +{
> + if (WARN_ON(level == max_level))
> + return 0;
> +
> + while (level++ < max_level) {
> + phys_addr_t rtt = alloc_delegated_page(realm, mc);
> +
> + if (rtt == PHYS_ADDR_MAX)
> + return -ENOMEM;
> +
> + if (realm_rtt_create(realm, ipa, level, rtt)) {
> + free_delegated_page(realm, rtt);
> + return -ENXIO;
> + }
> + }
> +
> + return 0;
> +}
> +
> static int realm_tear_down_rtt_level(struct realm *realm, int level,
> unsigned long start, unsigned long end)
> {
> @@ -265,6 +440,68 @@ static int realm_tear_down_rtt_range(struct realm *realm,
> start, end);
> }
>
> +/*
> + * Returns 0 on successful fold, a negative value on error, a positive value if
> + * we were not able to fold all tables at this level.
> + */
> +static int realm_fold_rtt_level(struct realm *realm, int level,
> + unsigned long start, unsigned long end)
> +{
> + int not_folded = 0;
> + ssize_t map_size;
> + unsigned long addr, next_addr;
> +
> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
> + return -EINVAL;
> +
> + map_size = rme_rtt_level_mapsize(level - 1);
> +
> + for (addr = start; addr < end; addr = next_addr) {
> + phys_addr_t rtt_granule;
> + int ret;
> + unsigned long align_addr = ALIGN(addr, map_size);
> +
> + next_addr = ALIGN(addr + 1, map_size);
> +
> + ret = realm_rtt_fold(realm, align_addr, level, &rtt_granule);
> +
> + switch (RMI_RETURN_STATUS(ret)) {
> + case RMI_SUCCESS:
> + if (!WARN_ON(rmi_granule_undelegate(rtt_granule)))
> + free_page((unsigned long)phys_to_virt(rtt_granule));
minor nit: Do we need a wrapper function for things like this, abd
leaking the page if undelegate fails, something like
rme_reclaim_delegated_page() ?
> + break;
> + case RMI_ERROR_RTT:
> + if (level == RME_RTT_MAX_LEVEL ||
> + RMI_RETURN_INDEX(ret) < level) {
> + not_folded++;
> + break;
> + }
> + /* Recurse a level deeper */
> + ret = realm_fold_rtt_level(realm,
> + level + 1,
> + addr,
> + next_addr);
> + if (ret < 0)
> + return ret;
> + else if (ret == 0)
> + /* Try again at this level */
> + next_addr = addr;
> + break;
> + default:
> + return -ENXIO;
> + }
> + }
> +
> + return not_folded;
> +}
> +
> +static int realm_fold_rtt_range(struct realm *realm,
> + unsigned long start, unsigned long end)
> +{
> + return realm_fold_rtt_level(realm, get_start_level(realm) + 1,
> + start, end);
> +}
> +
> static void ensure_spare_page(struct realm *realm)
> {
> phys_addr_t tmp_rtt;
> @@ -295,6 +532,147 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits)
> WARN_ON(realm_tear_down_rtt_range(realm, 0, (1UL << ia_bits)));
> }
>
> +void kvm_realm_unmap_range(struct kvm *kvm, unsigned long ipa, u64 size,
> + bool unmap_private)
> +{
> + unsigned long end = ipa + size;
> + struct realm *realm = &kvm->arch.realm;
> +
> + end = min(BIT(realm->ia_bits - 1), end);
> +
> + ensure_spare_page(realm);
> +
> + realm_unmap_range(kvm, ipa, end, unmap_private);
> +
> + realm_fold_rtt_range(realm, ipa, end);
Shouldn't this be :
if (unmap_private)
realm_fold_rtt_range(realm, ipa, end);
Also it is fine to reclaim RTTs from the protected space, not the
unprotected half, as long as we use RTT_DESTROY in unmap_shared routine.
> +}
> +
> +static int find_map_level(struct realm *realm,
> + unsigned long start,
> + unsigned long end)
> +{
> + int level = RME_RTT_MAX_LEVEL;
> +
> + while (level > get_start_level(realm)) {
> + unsigned long map_size = rme_rtt_level_mapsize(level - 1);
> +
> + if (!IS_ALIGNED(start, map_size) ||
> + (start + map_size) > end)
> + break;
> +
> + level--;
> + }
> +
> + return level;
> +}
> +
> +int realm_set_ipa_state(struct kvm_vcpu *vcpu,
> + unsigned long start,
> + unsigned long end,
> + unsigned long ripas)
> +{
> + struct kvm *kvm = vcpu->kvm;
> + struct realm *realm = &kvm->arch.realm;
> + struct realm_rec *rec = &vcpu->arch.rec;
> + phys_addr_t rd_phys = virt_to_phys(realm->rd);
> + phys_addr_t rec_phys = virt_to_phys(rec->rec_page);
> + struct kvm_mmu_memory_cache *memcache = &vcpu->arch.mmu_page_cache;
> + unsigned long ipa = start;
> + int ret = 0;
> +
> + while (ipa < end) {
> + unsigned long next;
> +
> + ret = rmi_rtt_set_ripas(rd_phys, rec_phys, ipa, end, &next);
> +
> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
> + int walk_level = RMI_RETURN_INDEX(ret);
> + int level = find_map_level(realm, ipa, end);
Might be worth adding a comment here. Check if the RMM needs tables to
create deeper level tables.
> +
> + if (walk_level < level) {
> + ret = realm_create_rtt_levels(realm, ipa,
> + walk_level,
> + level,
> + memcache);
/* Retry with RTTs created */
> + if (!ret)
> + continue;
> + } else {
> + ret = -EINVAL;
> + }
> +
> + break;
> + } else if (RMI_RETURN_STATUS(ret) != RMI_SUCCESS) {
> + WARN(1, "Unexpected error in %s: %#x\n", __func__,
> + ret);
> + ret = -EINVAL;
> + break;
> + }
> + ipa = next;
> + }
> +
> + if (ripas == RMI_EMPTY && ipa != start)
> + kvm_realm_unmap_range(kvm, start, ipa - start, true);
This triggers unmapping the "shared" aliases too, which is not necessary.
> +
> + return ret;
> +}
> +
> +static int realm_init_ipa_state(struct realm *realm,
> + unsigned long ipa,
> + unsigned long end)
> +{
> + phys_addr_t rd_phys = virt_to_phys(realm->rd);
> + int ret;
> +
> + while (ipa < end) {
> + unsigned long next;
> +
> + ret = rmi_rtt_init_ripas(rd_phys, ipa, end, &next);
> +
> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
> + int err_level = RMI_RETURN_INDEX(ret);
> + int level = find_map_level(realm, ipa, end);
> +
> + if (WARN_ON(err_level >= level))
> + return -ENXIO;
> +
> + ret = realm_create_rtt_levels(realm, ipa,
> + err_level,
> + level, NULL);
> + if (ret)
> + return ret;
> + /* Retry with the RTT levels in place */
> + continue;
> + } else if (WARN_ON(ret)) {
> + return -ENXIO;
> + }
> +
> + ipa = next;
> + }
> +
> + return 0;
> +}
> +
> +static int kvm_init_ipa_range_realm(struct kvm *kvm,
> + struct kvm_cap_arm_rme_init_ipa_args *args)
> +{
> + int ret = 0;
> + gpa_t addr, end;
> + struct realm *realm = &kvm->arch.realm;
> +
> + addr = args->init_ipa_base;
> + end = addr + args->init_ipa_size;
> +
> + if (end < addr)
> + return -EINVAL;
> +
> + if (kvm_realm_state(kvm) != REALM_STATE_NEW)
> + return -EINVAL;
> +
> + ret = realm_init_ipa_state(realm, addr, end);
> +
> + return ret;
super minor nit:
return realm_init_ipa_state(realm, addr, end);
> +}
> +
> /* Protects access to rme_vmid_bitmap */
> static DEFINE_SPINLOCK(rme_vmid_lock);
> static unsigned long *rme_vmid_bitmap;
> @@ -418,6 +796,18 @@ int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
> case KVM_CAP_ARM_RME_CREATE_RD:
> r = kvm_create_realm(kvm);
> break;
> + case KVM_CAP_ARM_RME_INIT_IPA_REALM: {
> + struct kvm_cap_arm_rme_init_ipa_args args;
> + void __user *argp = u64_to_user_ptr(cap->args[1]);
> +
> + if (copy_from_user(&args, argp, sizeof(args))) {
> + r = -EFAULT;
> + break;
> + }
> +
> + r = kvm_init_ipa_range_realm(kvm, &args);
> + break;
> + }
Suzuki
On 19/04/2024 10:34, Suzuki K Poulose wrote:
> On 12/04/2024 09:42, Steven Price wrote:
>> Each page within the protected region of the realm guest can be marked
>> as either RAM or EMPTY. Allow the VMM to control this before the guest
>> has started and provide the equivalent functions to change this (with
>> the guest's approval) at runtime.
>>
>> When transitioning from RIPAS RAM (1) to RIPAS EMPTY (0) the memory is
>> unmapped from the guest and undelegated allowing the memory to be reused
>> by the host. When transitioning to RIPAS RAM the actual population of
>> the leaf RTTs is done later on stage 2 fault, however it may be
>> necessary to allocate additional RTTs to represent the range requested.
>
> minor nit: To give a bit more context:
>
> "however it may be necessary to allocate additional RTTs in order for
> the RMM to track the RIPAS for the requested range".
>
>>
>> When freeing a block mapping it is necessary to temporarily unfold the
>> RTT which requires delegating an extra page to the RMM, this page can
>> then be recovered once the contents of the block mapping have been
>> freed. A spare, delegated page (spare_page) is used for this purpose.
>>
>> Signed-off-by: Steven Price <[email protected]>
>> ---
>> arch/arm64/include/asm/kvm_rme.h | 16 ++
>> arch/arm64/kvm/mmu.c | 8 +-
>> arch/arm64/kvm/rme.c | 390 +++++++++++++++++++++++++++++++
>> 3 files changed, 411 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/arm64/include/asm/kvm_rme.h
>> b/arch/arm64/include/asm/kvm_rme.h
>> index 915e76068b00..cc8f81cfc3c0 100644
>> --- a/arch/arm64/include/asm/kvm_rme.h
>> +++ b/arch/arm64/include/asm/kvm_rme.h
>> @@ -96,6 +96,14 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32
>> ia_bits);
>> int kvm_create_rec(struct kvm_vcpu *vcpu);
>> void kvm_destroy_rec(struct kvm_vcpu *vcpu);
>> +void kvm_realm_unmap_range(struct kvm *kvm,
>> + unsigned long ipa,
>> + u64 size,
>> + bool unmap_private);
>> +int realm_set_ipa_state(struct kvm_vcpu *vcpu,
>> + unsigned long addr, unsigned long end,
>> + unsigned long ripas);
>> +
>> #define RME_RTT_BLOCK_LEVEL 2
>> #define RME_RTT_MAX_LEVEL 3
>> @@ -114,4 +122,12 @@ static inline unsigned long
>> rme_rtt_level_mapsize(int level)
>> return (1UL << RME_RTT_LEVEL_SHIFT(level));
>> }
>> +static inline bool realm_is_addr_protected(struct realm *realm,
>> + unsigned long addr)
>> +{
>> + unsigned int ia_bits = realm->ia_bits;
>> +
>> + return !(addr & ~(BIT(ia_bits - 1) - 1));
>> +}
>> +
>> #endif
>> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
>> index 46f0c4e80ace..8a7b5449697f 100644
>> --- a/arch/arm64/kvm/mmu.c
>> +++ b/arch/arm64/kvm/mmu.c
>> @@ -310,6 +310,7 @@ static void invalidate_icache_guest_page(void *va,
>> size_t size)
>> * @start: The intermediate physical base address of the range to unmap
>> * @size: The size of the area to unmap
>> * @may_block: Whether or not we are permitted to block
>> + * @only_shared: If true then protected mappings should not be unmapped
>> *
>> * Clear a range of stage-2 mappings, lowering the various
>> ref-counts. Must
>> * be called while holding mmu_lock (unless for freeing the stage2
>> pgd before
>> @@ -317,7 +318,7 @@ static void invalidate_icache_guest_page(void *va,
>> size_t size)
>> * with things behind our backs.
>> */
>> static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t
>> start, u64 size,
>> - bool may_block)
>> + bool may_block, bool only_shared)
>> {
>> struct kvm *kvm = kvm_s2_mmu_to_kvm(mmu);
>> phys_addr_t end = start + size;
>> @@ -330,7 +331,7 @@ static void __unmap_stage2_range(struct kvm_s2_mmu
>> *mmu, phys_addr_t start, u64
>> static void unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t
>> start, u64 size)
>> {
>> - __unmap_stage2_range(mmu, start, size, true);
>> + __unmap_stage2_range(mmu, start, size, true, false);
>> }
>> static void stage2_flush_memslot(struct kvm *kvm,
>> @@ -1771,7 +1772,8 @@ bool kvm_unmap_gfn_range(struct kvm *kvm, struct
>> kvm_gfn_range *range)
>> __unmap_stage2_range(&kvm->arch.mmu, range->start << PAGE_SHIFT,
>> (range->end - range->start) << PAGE_SHIFT,
>> - range->may_block);
>> + range->may_block,
>> + range->only_shared);
>> return false;
>> }
>> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
>> index 629a095bea61..9e5983c51393 100644
>> --- a/arch/arm64/kvm/rme.c
>> +++ b/arch/arm64/kvm/rme.c
>> @@ -79,6 +79,12 @@ static phys_addr_t __alloc_delegated_page(struct
>> realm *realm,
>> return phys;
>> }
>> +static phys_addr_t alloc_delegated_page(struct realm *realm,
>> + struct kvm_mmu_memory_cache *mc)
>> +{
>> + return __alloc_delegated_page(realm, mc, GFP_KERNEL);
>> +}
>> +
>> static void free_delegated_page(struct realm *realm, phys_addr_t phys)
>> {
>> if (realm->spare_page == PHYS_ADDR_MAX) {
>> @@ -94,6 +100,151 @@ static void free_delegated_page(struct realm
>> *realm, phys_addr_t phys)
>> free_page((unsigned long)phys_to_virt(phys));
>> }
>> +static int realm_rtt_create(struct realm *realm,
>> + unsigned long addr,
>> + int level,
>> + phys_addr_t phys)
>> +{
>> + addr = ALIGN_DOWN(addr, rme_rtt_level_mapsize(level - 1));
>> + return rmi_rtt_create(virt_to_phys(realm->rd), phys, addr, level);
>> +}
>> +
>> +static int realm_rtt_fold(struct realm *realm,
>> + unsigned long addr,
>> + int level,
>> + phys_addr_t *rtt_granule)
>> +{
>> + unsigned long out_rtt;
>> + int ret;
>> +
>> + ret = rmi_rtt_fold(virt_to_phys(realm->rd), addr, level, &out_rtt);
>> +
>> + if (RMI_RETURN_STATUS(ret) == RMI_SUCCESS && rtt_granule)
>> + *rtt_granule = out_rtt;
>> +
>> + return ret;
>> +}
>> +
>> +static int realm_destroy_protected(struct realm *realm,
>> + unsigned long ipa,
>> + unsigned long *next_addr)
>> +{
>> + unsigned long rd = virt_to_phys(realm->rd);
>> + unsigned long addr;
>> + phys_addr_t rtt;
>> + int ret;
>> +
>> +loop:
>> + ret = rmi_data_destroy(rd, ipa, &addr, next_addr);
>> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
>> + if (*next_addr > ipa)
>> + return 0; /* UNASSIGNED */
>> + rtt = alloc_delegated_page(realm, NULL);
>> + if (WARN_ON(rtt == PHYS_ADDR_MAX))
>> + return -1;
>> + /* ASSIGNED - ipa is mapped as a block, so split */
>> + ret = realm_rtt_create(realm, ipa,
>> + RMI_RETURN_INDEX(ret) + 1, rtt);
>
> Could we not go all the way to L3 (rather than 1 level deeper) and try
> again ? That way, we are covered for block mappings at L1 (1G).
>
>> + if (WARN_ON(ret)) {
>> + free_delegated_page(realm, rtt);
>> + return -1;
>> + }
>> + /* retry */
>> + goto loop;
>> + } else if (WARN_ON(ret)) {
>> + return -1;
>> + }
>> + ret = rmi_granule_undelegate(addr);
>> +
>> + /*
>> + * If the undelegate fails then something has gone seriously
>> + * wrong: take an extra reference to just leak the page
>> + */
>> + if (WARN_ON(ret))
>> + get_page(phys_to_page(addr));
>> +
>> + return 0;
>> +}
>> +
>> +static void realm_unmap_range_shared(struct kvm *kvm,
>> + int level,
>> + unsigned long start,
>> + unsigned long end)
>> +{
>> + struct realm *realm = &kvm->arch.realm;
>> + unsigned long rd = virt_to_phys(realm->rd);
>> + ssize_t map_size = rme_rtt_level_mapsize(level);
>> + unsigned long next_addr, addr;
>> + unsigned long shared_bit = BIT(realm->ia_bits - 1);
>> +
>> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
>> + return;
>> +
>> + start |= shared_bit;
>> + end |= shared_bit;
>> +
>> + for (addr = start; addr < end; addr = next_addr) {
>> + unsigned long align_addr = ALIGN(addr, map_size);
>> + int ret;
>> +
>> + next_addr = ALIGN(addr + 1, map_size);
>> +
>> + if (align_addr != addr || next_addr > end) {
>> + /* Need to recurse deeper */
>> + if (addr < align_addr)
>> + next_addr = align_addr;
>> + realm_unmap_range_shared(kvm, level + 1, addr,
>> + min(next_addr, end));
>> + continue;
>> + }
>> +
>> + ret = rmi_rtt_unmap_unprotected(rd, addr, level, &next_addr);
>
> minor nit: We could potentially use rmi_rtt_destroy() to tear down
> shared mappings without unmapping them individually, if the range
> is big enough. All such optimisations could come later though.
>
>> + switch (RMI_RETURN_STATUS(ret)) {
>> + case RMI_SUCCESS:
>> + break;
>> + case RMI_ERROR_RTT:
>> + if (next_addr == addr) {
>
> At this point we have a block aligned address, but the mapping is
> further deep. Given, start from top to down, we implicitly handle
> the case of block mappings. Not sure if that needs to be in a comment
> here.
>
>> + next_addr = ALIGN(addr + 1, map_size);
>
> Reset to the "actual next" as it was overwritten by the RMI call.
>
>> + realm_unmap_range_shared(kvm, level + 1, addr,
>> + next_addr);
>> + }
>> + break;
>> + default:
>> + WARN_ON(1);
>> + }
>> + }
>> +}
>> +
>> +static void realm_unmap_range_private(struct kvm *kvm,
>> + unsigned long start,
>> + unsigned long end)
>> +{
>> + struct realm *realm = &kvm->arch.realm;
>> + ssize_t map_size = RME_PAGE_SIZE;
>> + unsigned long next_addr, addr;
>> +
>> + for (addr = start; addr < end; addr = next_addr) {
>> + int ret;
>> +
>> + next_addr = ALIGN(addr + 1, map_size);
>> +
>> + ret = realm_destroy_protected(realm, addr, &next_addr);
>> +
>> + if (WARN_ON(ret))
>> + break;
>> + }
>> +}
>> +
>> +static void realm_unmap_range(struct kvm *kvm,
>> + unsigned long start,
>> + unsigned long end,
>> + bool unmap_private)
>> +{
>> + realm_unmap_range_shared(kvm, RME_RTT_MAX_LEVEL - 1, start, end);
>
> minor nit: We already have a helper to find a suitable start level
> (defined below), may be we could use that ? And even do the rtt_destroy
> optimisation for unprotected range.
>
>> + if (unmap_private)
>> + realm_unmap_range_private(kvm, start, end);
>> +}
>> +
>> u32 kvm_realm_ipa_limit(void)
>> {
>> return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
>> @@ -190,6 +341,30 @@ static int realm_rtt_destroy(struct realm *realm,
>> unsigned long addr,
>> return ret;
>> }
>> +static int realm_create_rtt_levels(struct realm *realm,
>> + unsigned long ipa,
>> + int level,
>> + int max_level,
>> + struct kvm_mmu_memory_cache *mc)
>> +{
>> + if (WARN_ON(level == max_level))
>> + return 0;
>> +
>> + while (level++ < max_level) {
>> + phys_addr_t rtt = alloc_delegated_page(realm, mc);
>> +
>> + if (rtt == PHYS_ADDR_MAX)
>> + return -ENOMEM;
>> +
>> + if (realm_rtt_create(realm, ipa, level, rtt)) {
>> + free_delegated_page(realm, rtt);
>> + return -ENXIO;
>> + }
>> + }
>> +
>> + return 0;
>> +}
>> +
>> static int realm_tear_down_rtt_level(struct realm *realm, int level,
>> unsigned long start, unsigned long end)
>> {
>> @@ -265,6 +440,68 @@ static int realm_tear_down_rtt_range(struct realm
>> *realm,
>> start, end);
>> }
>> +/*
>> + * Returns 0 on successful fold, a negative value on error, a
>> positive value if
>> + * we were not able to fold all tables at this level.
>> + */
>> +static int realm_fold_rtt_level(struct realm *realm, int level,
>> + unsigned long start, unsigned long end)
>> +{
>> + int not_folded = 0;
>> + ssize_t map_size;
>> + unsigned long addr, next_addr;
>> +
>> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
>> + return -EINVAL;
>> +
>> + map_size = rme_rtt_level_mapsize(level - 1);
>> +
>> + for (addr = start; addr < end; addr = next_addr) {
>> + phys_addr_t rtt_granule;
>> + int ret;
>> + unsigned long align_addr = ALIGN(addr, map_size);
>> +
>> + next_addr = ALIGN(addr + 1, map_size);
>> +
>> + ret = realm_rtt_fold(realm, align_addr, level, &rtt_granule);
>> +
>> + switch (RMI_RETURN_STATUS(ret)) {
>> + case RMI_SUCCESS:
>> + if (!WARN_ON(rmi_granule_undelegate(rtt_granule)))
>> + free_page((unsigned long)phys_to_virt(rtt_granule));
>
> minor nit: Do we need a wrapper function for things like this, abd
> leaking the page if undelegate fails, something like
> rme_reclaim_delegated_page() ?
>
>
>> + break;
>> + case RMI_ERROR_RTT:
>> + if (level == RME_RTT_MAX_LEVEL ||
>> + RMI_RETURN_INDEX(ret) < level) {
>> + not_folded++;
>> + break;
>> + }
>> + /* Recurse a level deeper */
>> + ret = realm_fold_rtt_level(realm,
>> + level + 1,
>> + addr,
>> + next_addr);
>> + if (ret < 0)
>> + return ret;
>> + else if (ret == 0)
>> + /* Try again at this level */
>> + next_addr = addr;
>> + break;
>> + default:
>> + return -ENXIO;
>> + }
>> + }
>> +
>> + return not_folded;
>> +}
>> +
>> +static int realm_fold_rtt_range(struct realm *realm,
>> + unsigned long start, unsigned long end)
>> +{
>> + return realm_fold_rtt_level(realm, get_start_level(realm) + 1,
>> + start, end);
>> +}
>> +
>> static void ensure_spare_page(struct realm *realm)
>> {
>> phys_addr_t tmp_rtt;
>> @@ -295,6 +532,147 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32
>> ia_bits)
>> WARN_ON(realm_tear_down_rtt_range(realm, 0, (1UL << ia_bits)));
>> }
>> +void kvm_realm_unmap_range(struct kvm *kvm, unsigned long ipa, u64 size,
>> + bool unmap_private)
>> +{
>> + unsigned long end = ipa + size;
>> + struct realm *realm = &kvm->arch.realm;
>> +
>> + end = min(BIT(realm->ia_bits - 1), end);
>> +
>> + ensure_spare_page(realm);
>> +
>> + realm_unmap_range(kvm, ipa, end, unmap_private);
>> +
>> + realm_fold_rtt_range(realm, ipa, end);
>
> Shouldn't this be :
>
> if (unmap_private)
> realm_fold_rtt_range(realm, ipa, end);
>
> Also it is fine to reclaim RTTs from the protected space, not the
> unprotected half, as long as we use RTT_DESTROY in unmap_shared routine.
Thinking about this a bit more, we could :
1. Rename this to realm_reclaim_rtts_range()
2. Use "FOLD" vs "DESTROY" depending on the state of the Realm. If the
realm is DYING (or add a state in the kvm_pgtable_stage2_destroy() to
indicate that stage2 can now be "destroyed") and use DESTROY
wherever it is safe to do so.
Suzuki
On 16/04/2024 14:14, Suzuki K Poulose wrote:
> Hi Steven
>
> On 12/04/2024 09:42, Steven Price wrote:
>> The wrappers make the call sites easier to read and deal with the
>> boiler plate of handling the error codes from the RMM.
>>
>
> I have compared the parameters and output values to that of the RMM spec
> and they match. There are some minor nits below.
>
>> Signed-off-by: Steven Price <[email protected]>
>> ---
>> arch/arm64/include/asm/rmi_cmds.h | 509 ++++++++++++++++++++++++++++++
>> 1 file changed, 509 insertions(+)
>> create mode 100644 arch/arm64/include/asm/rmi_cmds.h
>>
>> diff --git a/arch/arm64/include/asm/rmi_cmds.h
>> b/arch/arm64/include/asm/rmi_cmds.h
>> new file mode 100644
>> index 000000000000..c21414127e8e
>> --- /dev/null
>> +++ b/arch/arm64/include/asm/rmi_cmds.h
>> @@ -0,0 +1,509 @@
>> +/* SPDX-License-Identifier: GPL-2.0 */
>> +/*
>> + * Copyright (C) 2023 ARM Ltd.
>> + */
>> +
>> +#ifndef __ASM_RMI_CMDS_H
>> +#define __ASM_RMI_CMDS_H
>> +
>> +#include <linux/arm-smccc.h>
>> +
>> +#include <asm/rmi_smc.h>
>> +
>> +struct rtt_entry {
>> + unsigned long walk_level;
>> + unsigned long desc;
>> + int state;
>> + int ripas;
>> +};
>> +
>
> ...
>
>> +/**
>> + * rmi_data_destroy() - Destroy a Data Granule
>> + * @rd: PA of the RD
>> + * @ipa: IPA at which the granule is mapped in the guest
>> + * @data_out: PA of the granule which was destroyed
>> + * @top_out: Top IPA of non-live RTT entries
>> + *
>> + * Transitions the granule to DESTROYED state, the address cannot be
>> used by
>> + * the guest for the lifetime of the Realm.
>> + *
>> + * Return: RMI return code
>> + */
>> +static inline int rmi_data_destroy(unsigned long rd, unsigned long ipa,
>> + unsigned long *data_out,
>> + unsigned long *top_out)
>> +{
>> + struct arm_smccc_res res;
>> +
>> + arm_smccc_1_1_invoke(SMC_RMI_DATA_DESTROY, rd, ipa, &res);
>> +
>> + *data_out = res.a1;
>> + *top_out = res.a2;
>
> minor nit: Do we need to be safer by checking the parameters before
> filling them in ? i.e.,
>
> if (ptr)
> *ptr = result_out;
>
> This applies for others calls below.
I had taken the approach of making all the out-parameters required (i.e.
non-NULL). But I guess I can switch over to allowing NULL - hopefully
the compiler will optimise these checks away, but there are some
situations where we are currently ignoring the extra out-parameters that
could be tidied up.
>
>> +
>> + return res.a0;
>> +}
>
>> +
>> +/**
>> + * rmi_realm_destroy() - Destroy a Realm
>> + * @rd: PA of the RD
>> + *
>> + * Destroys a Realm, all objects belonging to the Realm must be
>> destroyed first.
>> + *
>> + * Return: RMI return code
>> + */
>> +static inline int rmi_realm_destroy(unsigned long rd)
>> +{
>> + struct arm_smccc_res res;
>> +
>> + arm_smccc_1_1_invoke(SMC_RMI_REALM_DESTROY, rd, &res);
>> +
>> + return res.a0;
>> +}
>> +
>> +/**
>> + * rmi_rec_aux_count() - Get number of auxiliary Granules required
>> + * @rd: PA of the RD
>> + * @aux_count: Number of pages written to this pointer
>> + *
>> + * A REC may require extra auxiliary pages to be delegateed for the
>> RMM to
>
> minor nit: "s/delegateed/delegated/"
>
> ...
>
>> +/**
>> + * rmi_rtt_read_entry() - Read an RTTE
>> + * @rd: PA of the RD
>> + * @ipa: IPA for which to read the RTTE
>> + * @level: RTT level at which to read the RTTE
>> + * @rtt: Output structure describing the RTTE
>> + *
>> + * Reads a RTTE (Realm Translation Table Entry).
>> + *
>> + * Return: RMI return code
>> + */
>> +static inline int rmi_rtt_read_entry(unsigned long rd, unsigned long
>> ipa,
>> + long level, struct rtt_entry *rtt)
>> +{
>> + struct arm_smccc_1_2_regs regs = {
>> + SMC_RMI_RTT_READ_ENTRY,
>> + rd, ipa, level
>> + };
>> +
>> + arm_smccc_1_2_smc(®s, ®s);
>> +
>> + rtt->walk_level = regs.a1;
>> + rtt->state = regs.a2 & 0xFF;
>
> minor nit: We mask the state, but not the "ripas". Both of them are u8.
> For consistency, we should mask both or neither.
Good point - I'll mask ripas as well. I suspect this is a bug that crept
in when I was updating for the new RIPAS state.
>> + rtt->desc = regs.a3;
>> + rtt->ripas = regs.a4;
>> +
>> + return regs.a0;
>> +}
>> +
>
> ...
>
>> +/**
>> + * rmi_rtt_get_phys() - Get the PA from a RTTE
>> + * @rtt: The RTTE
>> + *
>> + * Return: the physical address from a RTT entry.
>> + */
>> +static inline phys_addr_t rmi_rtt_get_phys(struct rtt_entry *rtt)
>> +{
>> + return rtt->desc & GENMASK(47, 12);
>> +}
>
> I guess this may need to change with the LPA2 support in RMM and must be
> used in conjunction with the "realm" object to make the correct
> conversion.
Actually this is currently unused, and there's a potential bug lurking
in realm_map_protected() where rtt->desc is assumed to be a valid
physical address. I'll move the function there and fix it up by also
taking a realm argument. I've tried to keep the realm structure out of
this file.
Thanks,
Steve
Hi Steven
On 12/04/2024 09:42, Steven Price wrote:
> Entering a realm is done using a SMC call to the RMM. On exit the
> exit-codes need to be handled slightly differently to the normal KVM
> path so define our own functions for realm enter/exit and hook them
> in if the guest is a realm guest.
>
> Signed-off-by: Steven Price <[email protected]>
Please find a comment about RIPAS change exit handling below. Rest
looks good to me.
> ---
> arch/arm64/include/asm/kvm_rme.h | 3 +
> arch/arm64/kvm/Makefile | 2 +-
> arch/arm64/kvm/arm.c | 19 +++-
> arch/arm64/kvm/rme-exit.c | 180 +++++++++++++++++++++++++++++++
> arch/arm64/kvm/rme.c | 11 ++
> 5 files changed, 209 insertions(+), 6 deletions(-)
> create mode 100644 arch/arm64/kvm/rme-exit.c
>
..
>
> /* Tell userspace about in-kernel device output levels */
> diff --git a/arch/arm64/kvm/rme-exit.c b/arch/arm64/kvm/rme-exit.c
> new file mode 100644
> index 000000000000..5bf58c9b42b7
> --- /dev/null
> +++ b/arch/arm64/kvm/rme-exit.c
> @@ -0,0 +1,180 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +/*
> + * Copyright (C) 2023 ARM Ltd.
> + */
> +
> +#include <linux/kvm_host.h>
> +#include <kvm/arm_hypercalls.h>
> +#include <kvm/arm_psci.h>
> +
> +#include <asm/rmi_smc.h>
> +#include <asm/kvm_emulate.h>
> +#include <asm/kvm_rme.h>
> +#include <asm/kvm_mmu.h>
> +
> +typedef int (*exit_handler_fn)(struct kvm_vcpu *vcpu);
> +
> +static int rec_exit_reason_notimpl(struct kvm_vcpu *vcpu)
> +{
> + struct realm_rec *rec = &vcpu->arch.rec;
> +
> + pr_err("[vcpu %d] Unhandled exit reason from realm (ESR: %#llx)\n",
> + vcpu->vcpu_id, rec->run->exit.esr);
> + return -ENXIO;
> +}
> +
> +static int rec_exit_sync_dabt(struct kvm_vcpu *vcpu)
> +{
> + return kvm_handle_guest_abort(vcpu);
> +}
> +
> +static int rec_exit_sync_iabt(struct kvm_vcpu *vcpu)
> +{
> + struct realm_rec *rec = &vcpu->arch.rec;
> +
> + pr_err("[vcpu %d] Unhandled instruction abort (ESR: %#llx).\n",
> + vcpu->vcpu_id, rec->run->exit.esr);
> + return -ENXIO;
> +}
> +
> +static int rec_exit_sys_reg(struct kvm_vcpu *vcpu)
> +{
> + struct realm_rec *rec = &vcpu->arch.rec;
> + unsigned long esr = kvm_vcpu_get_esr(vcpu);
> + int rt = kvm_vcpu_sys_get_rt(vcpu);
> + bool is_write = !(esr & 1);
> + int ret;
> +
> + if (is_write)
> + vcpu_set_reg(vcpu, rt, rec->run->exit.gprs[0]);
> +
> + ret = kvm_handle_sys_reg(vcpu);
> +
> + if (ret >= 0 && !is_write)
> + rec->run->entry.gprs[0] = vcpu_get_reg(vcpu, rt);
> +
> + return ret;
> +}
> +
> +static exit_handler_fn rec_exit_handlers[] = {
> + [0 ... ESR_ELx_EC_MAX] = rec_exit_reason_notimpl,
> + [ESR_ELx_EC_SYS64] = rec_exit_sys_reg,
> + [ESR_ELx_EC_DABT_LOW] = rec_exit_sync_dabt,
> + [ESR_ELx_EC_IABT_LOW] = rec_exit_sync_iabt
> +};
> +
> +static int rec_exit_psci(struct kvm_vcpu *vcpu)
> +{
> + struct realm_rec *rec = &vcpu->arch.rec;
> + int i;
> + int ret;
> +
> + for (i = 0; i < REC_RUN_GPRS; i++)
> + vcpu_set_reg(vcpu, i, rec->run->exit.gprs[i]);
> +
> + ret = kvm_smccc_call_handler(vcpu);
> +
> + for (i = 0; i < REC_RUN_GPRS; i++)
> + rec->run->entry.gprs[i] = vcpu_get_reg(vcpu, i);
> +
> + return ret;
> +}
> +
> +static int rec_exit_ripas_change(struct kvm_vcpu *vcpu)
> +{
> + struct kvm *kvm = vcpu->kvm;
> + struct realm *realm = &kvm->arch.realm;
> + struct realm_rec *rec = &vcpu->arch.rec;
> + unsigned long base = rec->run->exit.ripas_base;
> + unsigned long top = rec->run->exit.ripas_top;
> + unsigned long ripas = rec->run->exit.ripas_value & 1;
> + int ret = -EINVAL;
> +
> + if (realm_is_addr_protected(realm, base) &&
> + realm_is_addr_protected(realm, top - 1)) {
> + kvm_mmu_topup_memory_cache(&vcpu->arch.mmu_page_cache,
> + kvm_mmu_cache_min_pages(vcpu->arch.hw_mmu));
> + write_lock(&kvm->mmu_lock);
> + ret = realm_set_ipa_state(vcpu, base, top, ripas);
> + write_unlock(&kvm->mmu_lock);
> + }
There are couple of issues here :
1. We don't verify if the range is covered by a memslot, for ripas ==
RIPAS_RAM conversions and reject the request if not backed by memory.
2. The realm_set_ipa_state() might have partially completed the request
before it ran out of pages in the cache. And as such we should only
let the VMM know about what was completed, below. So, we need some
feedback from realm_set_ipa_state() to indicate the completed range.
Rest looks fine to me.
Suzuki
> +
> + WARN(ret && ret != -ENOMEM,
> + "Unable to satisfy SET_IPAS for %#lx - %#lx, ripas: %#lx\n",
> + base, top, ripas);
> +
> + /* Exit to VMM to complete the change */
> + kvm_prepare_memory_fault_exit(vcpu, base, top - base, false, false,
> + ripas == 1);
> +
> + return 0;
> +}
> +
> +static void update_arch_timer_irq_lines(struct kvm_vcpu *vcpu)
> +{
> + struct realm_rec *rec = &vcpu->arch.rec;
> +
> + __vcpu_sys_reg(vcpu, CNTV_CTL_EL0) = rec->run->exit.cntv_ctl;
> + __vcpu_sys_reg(vcpu, CNTV_CVAL_EL0) = rec->run->exit.cntv_cval;
> + __vcpu_sys_reg(vcpu, CNTP_CTL_EL0) = rec->run->exit.cntp_ctl;
> + __vcpu_sys_reg(vcpu, CNTP_CVAL_EL0) = rec->run->exit.cntp_cval;
> +
> + kvm_realm_timers_update(vcpu);
> +}
> +
> +/*
> + * Return > 0 to return to guest, < 0 on error, 0 (and set exit_reason) on
> + * proper exit to userspace.
> + */
> +int handle_rme_exit(struct kvm_vcpu *vcpu, int rec_run_ret)
> +{
> + struct realm_rec *rec = &vcpu->arch.rec;
> + u8 esr_ec = ESR_ELx_EC(rec->run->exit.esr);
> + unsigned long status, index;
> +
> + status = RMI_RETURN_STATUS(rec_run_ret);
> + index = RMI_RETURN_INDEX(rec_run_ret);
> +
> + /*
> + * If a PSCI_SYSTEM_OFF request raced with a vcpu executing, we might
> + * see the following status code and index indicating an attempt to run
> + * a REC when the RD state is SYSTEM_OFF. In this case, we just need to
> + * return to user space which can deal with the system event or will try
> + * to run the KVM VCPU again, at which point we will no longer attempt
> + * to enter the Realm because we will have a sleep request pending on
> + * the VCPU as a result of KVM's PSCI handling.
> + */
> + if (status == RMI_ERROR_REALM && index == 1) {
> + vcpu->run->exit_reason = KVM_EXIT_UNKNOWN;
> + return 0;
> + }
> +
> + if (rec_run_ret)
> + return -ENXIO;
> +
> + vcpu->arch.fault.esr_el2 = rec->run->exit.esr;
> + vcpu->arch.fault.far_el2 = rec->run->exit.far;
> + vcpu->arch.fault.hpfar_el2 = rec->run->exit.hpfar;
> +
> + update_arch_timer_irq_lines(vcpu);
> +
> + /* Reset the emulation flags for the next run of the REC */
> + rec->run->entry.flags = 0;
> +
> + switch (rec->run->exit.exit_reason) {
> + case RMI_EXIT_SYNC:
> + return rec_exit_handlers[esr_ec](vcpu);
> + case RMI_EXIT_IRQ:
> + case RMI_EXIT_FIQ:
> + return 1;
> + case RMI_EXIT_PSCI:
> + return rec_exit_psci(vcpu);
> + case RMI_EXIT_RIPAS_CHANGE:
> + return rec_exit_ripas_change(vcpu);
> + }
I guess, the RMI_EXIT_HOST_CALL could be handled here similar to the
rec_exit_psci() ?
Suzuki
On 12/04/2024 09:42, Steven Price wrote:
> The VMM needs to populate the realm with some data before starting (e.g.
> a kernel and initrd). This is measured by the RMM and used as part of
> the attestation later on.
>
> Co-developed-by: Suzuki K Poulose <[email protected]>
> Signed-off-by: Suzuki K Poulose <[email protected]>
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/kvm/rme.c | 234 +++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 234 insertions(+)
>
> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
> index 0a3f823b2446..4aab507f896e 100644
> --- a/arch/arm64/kvm/rme.c
> +++ b/arch/arm64/kvm/rme.c
> @@ -4,6 +4,7 @@
> */
>
> #include <linux/kvm_host.h>
> +#include <linux/hugetlb.h>
>
> #include <asm/kvm_emulate.h>
> #include <asm/kvm_mmu.h>
> @@ -547,6 +548,227 @@ void kvm_realm_unmap_range(struct kvm *kvm, unsigned long ipa, u64 size,
> realm_fold_rtt_range(realm, ipa, end);
> }
>
> +static int realm_create_protected_data_page(struct realm *realm,
> + unsigned long ipa,
> + struct page *dst_page,
> + struct page *src_page,
> + unsigned long flags)
> +{
> + phys_addr_t dst_phys, src_phys;
> + int ret;
> +
> + dst_phys = page_to_phys(dst_page);
> + src_phys = page_to_phys(src_page);
> +
> + if (rmi_granule_delegate(dst_phys))
> + return -ENXIO;
> +
> + ret = rmi_data_create(virt_to_phys(realm->rd), dst_phys, ipa, src_phys,
> + flags);
> +
> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
> + /* Create missing RTTs and retry */
> + int level = RMI_RETURN_INDEX(ret);
> +
> + ret = realm_create_rtt_levels(realm, ipa, level,
> + RME_RTT_MAX_LEVEL, NULL);
> + if (ret)
> + goto err;
> +
> + ret = rmi_data_create(virt_to_phys(realm->rd), dst_phys, ipa,
> + src_phys, flags);
> + }
> +
> + if (ret)
> + goto err;
ultra minor nit:
if (!ret)
return 0;
> +
> + return 0;
> +
> +err:
> + if (WARN_ON(rmi_granule_undelegate(dst_phys))) {
> + /* Page can't be returned to NS world so is lost */
> + get_page(dst_page);
> + }
> + return -ENXIO;
> +}
> +
> +static int fold_rtt(struct realm *realm, unsigned long addr, int level)
> +{
> + phys_addr_t rtt_addr;
> + int ret;
> +
> + ret = realm_rtt_fold(realm, addr, level + 1, &rtt_addr);
> + if (ret)
> + return ret;
> +
> + free_delegated_page(realm, rtt_addr);
> +
> + return 0;
> +}
> +
> +static int populate_par_region(struct kvm *kvm,
> + phys_addr_t ipa_base,
> + phys_addr_t ipa_end,
> + u32 flags)
> +{
> + struct realm *realm = &kvm->arch.realm;
> + struct kvm_memory_slot *memslot;
> + gfn_t base_gfn, end_gfn;
> + int idx;
> + phys_addr_t ipa;
> + int ret = 0;
> + struct page *tmp_page;
> + unsigned long data_flags = 0;
> +
> + base_gfn = gpa_to_gfn(ipa_base);
> + end_gfn = gpa_to_gfn(ipa_end);
> +
> + if (flags & KVM_ARM_RME_POPULATE_FLAGS_MEASURE)
> + data_flags = RMI_MEASURE_CONTENT;
> +
> + idx = srcu_read_lock(&kvm->srcu);
> + memslot = gfn_to_memslot(kvm, base_gfn);
> + if (!memslot) {
> + ret = -EFAULT;
> + goto out;
> + }
> +
> + /* We require the region to be contained within a single memslot */
> + if (memslot->base_gfn + memslot->npages < end_gfn) {
> + ret = -EINVAL;
> + goto out;
> + }
> +
> + tmp_page = alloc_page(GFP_KERNEL);
> + if (!tmp_page) {
> + ret = -ENOMEM;
> + goto out;
> + }
> +
> + mmap_read_lock(current->mm);
> +
> + ipa = ipa_base;
> + while (ipa < ipa_end) {
> + struct vm_area_struct *vma;
> + unsigned long map_size;
> + unsigned int vma_shift;
> + unsigned long offset;
> + unsigned long hva;
> + struct page *page;
> + kvm_pfn_t pfn;
> + int level;
> +
> + hva = gfn_to_hva_memslot(memslot, gpa_to_gfn(ipa));
> + vma = vma_lookup(current->mm, hva);
> + if (!vma) {
> + ret = -EFAULT;
> + break;
> + }
> +
> + if (is_vm_hugetlb_page(vma))
> + vma_shift = huge_page_shift(hstate_vma(vma));
> + else
> + vma_shift = PAGE_SHIFT;
> +
> + map_size = 1 << vma_shift;
> +
> + /*
> + * FIXME: This causes over mapping, but there's no good
> + * solution here with the ABI as it stands
> + */
> + ipa = ALIGN_DOWN(ipa, map_size);
> +
> + switch (map_size) {
> + case RME_L2_BLOCK_SIZE:
> + level = 2;
> + break;
> + case PAGE_SIZE:
> + level = 3;
> + break;
> + default:
> + WARN_ONCE(1, "Unsupport vma_shift %d", vma_shift);
Do we really need this WARNing ? Could we not fallback to the next
possible mapping size ? e.g: if 1G, we could at least try 2M. I guess
this is more or less similar to what we would do on a fault and it may
be a good idea to see if could reuse the core bits from user_mem_abort ?
> + ret = -EFAULT;
> + break;
> + }
> +
> + pfn = gfn_to_pfn_memslot(memslot, gpa_to_gfn(ipa));
> +
> + if (is_error_pfn(pfn)) {
> + ret = -EFAULT;
> + break;
> + }
> +
> + if (level < RME_RTT_MAX_LEVEL) {
> + /*
> + * A temporary RTT is needed during the map, precreate
> + * it, however if there is an error (e.g. missing
> + * parent tables) this will be handled in the
> + * realm_create_protected_data_page() call.
> + */
> + realm_create_rtt_levels(realm, ipa, level,
> + RME_RTT_MAX_LEVEL, NULL);
> + }
> +
> + page = pfn_to_page(pfn);
> +
> + for (offset = 0; offset < map_size && !ret;
> + offset += PAGE_SIZE, page++) {
> + phys_addr_t page_ipa = ipa + offset;
> +
> + ret = realm_create_protected_data_page(realm, page_ipa,
> + page, tmp_page,
> + data_flags);
> + }
> + if (ret)
> + goto err_release_pfn;
> +
> + if (level == 2) {
> + ret = fold_rtt(realm, ipa, level);
> + if (ret)
> + goto err_release_pfn;
Do we care about the FOLD error here ? Ideally we shouldn't hit an
error, but unable to fold is not an error case ? We could live with
L3 mappings ?
Suzuki
> + }
> +
> + ipa += map_size;
> + kvm_release_pfn_dirty(pfn);
> +err_release_pfn:
> + if (ret) {
> + kvm_release_pfn_clean(pfn);
> + break;
> + }
> + }
> +
> + mmap_read_unlock(current->mm);
> + __free_page(tmp_page);
> +
> +out:
> + srcu_read_unlock(&kvm->srcu, idx);
> + return ret;
> +}
> +
> +static int kvm_populate_realm(struct kvm *kvm,
> + struct kvm_cap_arm_rme_populate_realm_args *args)
> +{
> + phys_addr_t ipa_base, ipa_end;
> +
> + if (kvm_realm_state(kvm) != REALM_STATE_NEW)
> + return -EINVAL;
> +
> + if (!IS_ALIGNED(args->populate_ipa_base, PAGE_SIZE) ||
> + !IS_ALIGNED(args->populate_ipa_size, PAGE_SIZE))
> + return -EINVAL;
> +
> + if (args->flags & ~RMI_MEASURE_CONTENT)
> + return -EINVAL;
> +
> + ipa_base = args->populate_ipa_base;
> + ipa_end = ipa_base + args->populate_ipa_size;
> +
> + if (ipa_end < ipa_base)
> + return -EINVAL;
> +
> + return populate_par_region(kvm, ipa_base, ipa_end, args->flags);
> +}
> +
> static int find_map_level(struct realm *realm,
> unsigned long start,
> unsigned long end)
> @@ -808,6 +1030,18 @@ int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
> r = kvm_init_ipa_range_realm(kvm, &args);
> break;
> }
> + case KVM_CAP_ARM_RME_POPULATE_REALM: {
> + struct kvm_cap_arm_rme_populate_realm_args args;
> + void __user *argp = u64_to_user_ptr(cap->args[1]);
> +
> + if (copy_from_user(&args, argp, sizeof(args))) {
> + r = -EFAULT;
> + break;
> + }
> +
> + r = kvm_populate_realm(kvm, &args);
> + break;
> + }
> default:
> r = -EINVAL;
> break;
On 16/04/2024 14:30, Suzuki K Poulose wrote:
> Hi Steven
>
> On 12/04/2024 09:42, Steven Price wrote:
>> Query the RMI version number and check if it is a compatible version. A
>> static key is also provided to signal that a supported RMM is available.
>>
>> Functions are provided to query if a VM or VCPU is a realm (or rec)
>> which currently will always return false.
>>
>> Signed-off-by: Steven Price <[email protected]>
>> ---
>> arch/arm64/include/asm/kvm_emulate.h | 18 +++++++++
>> arch/arm64/include/asm/kvm_host.h | 4 ++
>> arch/arm64/include/asm/kvm_rme.h | 56 ++++++++++++++++++++++++++++
>> arch/arm64/include/asm/virt.h | 1 +
>> arch/arm64/kvm/Makefile | 3 +-
>> arch/arm64/kvm/arm.c | 9 +++++
>> arch/arm64/kvm/rme.c | 52 ++++++++++++++++++++++++++
>> 7 files changed, 142 insertions(+), 1 deletion(-)
>> create mode 100644 arch/arm64/include/asm/kvm_rme.h
>> create mode 100644 arch/arm64/kvm/rme.c
>>
>> diff --git a/arch/arm64/include/asm/kvm_emulate.h
>> b/arch/arm64/include/asm/kvm_emulate.h
>> index 975af30af31f..6f08398537e2 100644
>> --- a/arch/arm64/include/asm/kvm_emulate.h
>> +++ b/arch/arm64/include/asm/kvm_emulate.h
>> @@ -611,4 +611,22 @@ static __always_inline void
>> kvm_reset_cptr_el2(struct kvm_vcpu *vcpu)
>> kvm_write_cptr_el2(val);
>> }
>> +
>> +static inline bool kvm_is_realm(struct kvm *kvm)
>> +{
>> + if (static_branch_unlikely(&kvm_rme_is_available))
>> + return kvm->arch.is_realm;
>> + return false;
>> +}
>> +
>> +static inline enum realm_state kvm_realm_state(struct kvm *kvm)
>> +{
>> + return READ_ONCE(kvm->arch.realm.state);
>> +}
>> +
>> +static inline bool vcpu_is_rec(struct kvm_vcpu *vcpu)
>> +{
>> + return false;
>> +}
>> +
>> #endif /* __ARM64_KVM_EMULATE_H__ */
>> diff --git a/arch/arm64/include/asm/kvm_host.h
>> b/arch/arm64/include/asm/kvm_host.h
>> index 9e8a496fb284..63b68b85db3f 100644
>> --- a/arch/arm64/include/asm/kvm_host.h
>> +++ b/arch/arm64/include/asm/kvm_host.h
>> @@ -27,6 +27,7 @@
>> #include <asm/fpsimd.h>
>> #include <asm/kvm.h>
>> #include <asm/kvm_asm.h>
>> +#include <asm/kvm_rme.h>
>> #include <asm/vncr_mapping.h>
>> #define __KVM_HAVE_ARCH_INTC_INITIALIZED
>> @@ -348,6 +349,9 @@ struct kvm_arch {
>> * the associated pKVM instance in the hypervisor.
>> */
>> struct kvm_protected_vm pkvm;
>> +
>> + bool is_realm;
>> + struct realm realm;
>> };
>> struct kvm_vcpu_fault_info {
>> diff --git a/arch/arm64/include/asm/kvm_rme.h
>> b/arch/arm64/include/asm/kvm_rme.h
>> new file mode 100644
>> index 000000000000..922da3f47227
>> --- /dev/null
>> +++ b/arch/arm64/include/asm/kvm_rme.h
>> @@ -0,0 +1,56 @@
>> +/* SPDX-License-Identifier: GPL-2.0 */
>> +/*
>> + * Copyright (C) 2023 ARM Ltd.
>> + */
>> +
>> +#ifndef __ASM_KVM_RME_H
>> +#define __ASM_KVM_RME_H
>> +
>> +/**
>> + * enum realm_state - State of a Realm
>> + */
>> +enum realm_state {
>> + /**
>> + * @REALM_STATE_NONE:
>> + * Realm has not yet been created. rmi_realm_create() may be
>> + * called to create the realm.
>> + */
>> + REALM_STATE_NONE,
>> + /**
>> + * @REALM_STATE_NEW:
>> + * Realm is under construction, not eligible for execution.
>> Pages
>> + * may be populated with rmi_data_create().
>> + */
>> + REALM_STATE_NEW,
>> + /**
>> + * @REALM_STATE_ACTIVE:
>> + * Realm has been created and is eligible for execution with
>> + * rmi_rec_enter(). Pages may no longer be populated with
>> + * rmi_data_create().
>> + */
>> + REALM_STATE_ACTIVE,
>> + /**
>> + * @REALM_STATE_DYING:
>> + * Realm is in the process of being destroyed or has already
>> been
>> + * destroyed.
>> + */
>> + REALM_STATE_DYING,
>> + /**
>> + * @REALM_STATE_DEAD:
>> + * Realm has been destroyed.
>> + */
>> + REALM_STATE_DEAD
>> +};
>> +
>> +/**
>> + * struct realm - Additional per VM data for a Realm
>> + *
>> + * @state: The lifetime state machine for the realm
>> + */
>> +struct realm {
>> + enum realm_state state;
>> +};
>> +
>> +int kvm_init_rme(void);
>> +
>> +#endif
>> diff --git a/arch/arm64/include/asm/virt.h
>> b/arch/arm64/include/asm/virt.h
>> index 261d6e9df2e1..12cf36c38189 100644
>> --- a/arch/arm64/include/asm/virt.h
>> +++ b/arch/arm64/include/asm/virt.h
>> @@ -81,6 +81,7 @@ void __hyp_reset_vectors(void);
>> bool is_kvm_arm_initialised(void);
>> DECLARE_STATIC_KEY_FALSE(kvm_protected_mode_initialized);
>> +DECLARE_STATIC_KEY_FALSE(kvm_rme_is_available);
>> /* Reports the availability of HYP mode */
>> static inline bool is_hyp_mode_available(void)
>> diff --git a/arch/arm64/kvm/Makefile b/arch/arm64/kvm/Makefile
>> index c0c050e53157..1c1d8cdf381f 100644
>> --- a/arch/arm64/kvm/Makefile
>> +++ b/arch/arm64/kvm/Makefile
>> @@ -20,7 +20,8 @@ kvm-y += arm.o mmu.o mmio.o psci.o hypercalls.o
>> pvtime.o \
>> vgic/vgic-v3.o vgic/vgic-v4.o \
>> vgic/vgic-mmio.o vgic/vgic-mmio-v2.o \
>> vgic/vgic-mmio-v3.o vgic/vgic-kvm-device.o \
>> - vgic/vgic-its.o vgic/vgic-debug.o
>> + vgic/vgic-its.o vgic/vgic-debug.o \
>> + rme.o
>> kvm-$(CONFIG_HW_PERF_EVENTS) += pmu-emul.o pmu.o
>> diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
>> index 3dee5490eea9..2056c660c5ee 100644
>> --- a/arch/arm64/kvm/arm.c
>> +++ b/arch/arm64/kvm/arm.c
>> @@ -38,6 +38,7 @@
>> #include <asm/kvm_mmu.h>
>> #include <asm/kvm_nested.h>
>> #include <asm/kvm_pkvm.h>
>> +#include <asm/kvm_rme.h>
>> #include <asm/kvm_emulate.h>
>> #include <asm/sections.h>
>> @@ -47,6 +48,8 @@
>> static enum kvm_mode kvm_mode = KVM_MODE_DEFAULT;
>> +DEFINE_STATIC_KEY_FALSE(kvm_rme_is_available);
>> +
>> DECLARE_KVM_HYP_PER_CPU(unsigned long, kvm_hyp_vector);
>> DEFINE_PER_CPU(unsigned long, kvm_arm_hyp_stack_page);
>> @@ -2562,6 +2565,12 @@ static __init int kvm_arm_init(void)
>> in_hyp_mode = is_kernel_in_hyp_mode();
>> + if (in_hyp_mode) {
>> + err = kvm_init_rme();
>> + if (err)
>> + return err;
>> + }
>> +
>> if (cpus_have_final_cap(ARM64_WORKAROUND_DEVICE_LOAD_ACQUIRE) ||
>> cpus_have_final_cap(ARM64_WORKAROUND_1508412))
>> kvm_info("Guests without required CPU erratum workarounds
>> can deadlock system!\n" \
>> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
>> new file mode 100644
>> index 000000000000..3dbbf9d046bf
>> --- /dev/null
>> +++ b/arch/arm64/kvm/rme.c
>> @@ -0,0 +1,52 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +/*
>> + * Copyright (C) 2023 ARM Ltd.
>> + */
>> +
>> +#include <linux/kvm_host.h>
>> +
>> +#include <asm/rmi_cmds.h>
>> +#include <asm/virt.h>
>> +
>> +static int rmi_check_version(void)
>> +{
>> + struct arm_smccc_res res;
>> + int version_major, version_minor;
>> + unsigned long host_version = RMI_ABI_VERSION(RMI_ABI_MAJOR_VERSION,
>> + RMI_ABI_MINOR_VERSION);
>> +
>> + arm_smccc_1_1_invoke(SMC_RMI_VERSION, host_version, &res);
>> +
>> + if (res.a0 == SMCCC_RET_NOT_SUPPORTED)
>> + return -ENXIO;
>> +
>> + version_major = RMI_ABI_VERSION_GET_MAJOR(res.a1);
>> + version_minor = RMI_ABI_VERSION_GET_MINOR(res.a1);
>> +
>
> We don't seem to be using the res.a0 to determin if the RMM supports our
> requested version. As per RMM spec, section B4.3.23 :
>
> "
> The status code and lower revision output values indicate which of the
> following is true, in order of precedence:
> a) The RMM supports an interface revision which is compatible with the
> requested revision.
> • The status code is RMI_SUCCESS.
> • The lower revision is equal to the requested revision.
> b) The RMM does not support an interface revision which is compatible
> with the requested revision The RMM supports an interface revision
> which is incompatible with and less than the requested revision.
> • The status code is RMI_ERROR_INPUT.
> • The lower revision is the highest interface revision which is
> both less than the requested revision and supported by the RMM.
>
> c) The RMM does not support an interface revision which is compatible
> with the requested revision The RMM supports an interface revision
> which is incompatible with and greater than the requested revision.
> • The status code is RMI_ERROR_INPUT.
> • The lower revision is equal to the higher revision.
>
> So, we could simply check the res.a0 for RMI_SUCCESS and proceed with
> marking RMM available.
Good point - this didn't work in a previous version of the spec, but we
should be able to rely on the return value now.
>> + if (version_major != RMI_ABI_MAJOR_VERSION) {
>> + kvm_err("Unsupported RMI ABI (v%d.%d) host supports v%d.%d\n",
>> + version_major, version_minor,
>> + RMI_ABI_MAJOR_VERSION,
>> + RMI_ABI_MINOR_VERSION);
>> + return -ENXIO;
>> + }
>> +
>> + kvm_info("RMI ABI version %d.%d\n", version_major, version_minor);
>> +
>> + return 0;
>> +}
>> +
>> +int kvm_init_rme(void)
>> +{
>> + if (PAGE_SIZE != SZ_4K)
>> + /* Only 4k page size on the host is supported */
>> + return 0;
>> +
>> + if (rmi_check_version())
>> + /* Continue without realm support */
>> + return 0;
>> +
>> + /* Future patch will enable static branch kvm_rme_is_available */
>> +
>> + return 0;
>
> Do we ever expect this to fail the kvm initialisation ? Otherwise, we
> could leave it as a void ?
Technically in a later patch the return from rme_vmid_init() can cause
such a failure. But it's not clear that it makes any sense to completely
kill KVM because of that. So I'll change this to a void return.
Thanks,
Steve
On 17/04/2024 10:51, Suzuki K Poulose wrote:
> Hi Steven
>
> On 12/04/2024 09:42, Steven Price wrote:
>> Add the KVM_CAP_ARM_RME_CREATE_FD ioctl to create a realm. This involves
>
> minor nit: s/FD/RD
>
>> delegating pages to the RMM to hold the Realm Descriptor (RD) and for
>> the base level of the Realm Translation Tables (RTT). A VMID also need
>> to be picked, since the RMM has a separate VMID address space a
>> dedicated allocator is added for this purpose.
>>
>> KVM_CAP_ARM_RME_CONFIG_REALM is provided to allow configuring the realm
>> before it is created.
>
> It might be helpful to provide a bit more background on the Realm
> parameters. Something like:
>
> Realm parameters for Realm Descriptor creation could be classified as:
>
> 1. Parameters specific to the Realm stage2 (e.g. IPA Size, vmid, stage2
> entry level, entry level RTTs, number of RTTs in start level, LPA2)
> Most of these are not measured by RMM and comes from KVM book
> keeping.
>
> 2. Parameters controlling "Arm Architecture features for the VM". (e.g.
> SVE VL, PMU counters, number of HW BRPs/WPs), configured by the VMM
> using the "user ID register write" mechanism. These will be
> supported in the later patches.
>
> 3. Parameters are not part of the core Arm architecture but defined
> by the RMM spec (e.g. Hash algorithm for measurement,
> Personalisation value). These are programmed via
> KVM_CAP_ARM_RME_CONFIG_REALM.
>
Thanks, I'll steal that wording ;)
>
> Also it may be a good idea to call out one of the issues that we have
> with the UABI w.r.t the IPA Size. The IPA Size supported by RMM *could*
> be different from the normal IPA Size as supported by KVM. We do not
> expect this to be common, but is not impossible.
>
> If the RMM_IPA_Size < Normal_IPA_Size, we have a problem with
> advertising the "IPA Size" to the VMM. Right now we advertise
> the "normal limit" by KVM_CAP_ARM_VM_IPA_SIZE and the IPA Size
> is configured via vm_type[7:0] in KVM_CREATE_VM. Given we have
> to configure the IPA size for a "Realm VM" at CREATE_VM time too,
> the VMM is unable to choose a valid IPA Size for the Realm. We
> have the following options:
>
> 1. Given IPA Size for a Realm is measured, the user must get
> what they choose. i.e., if the platform cannot support the
> requested size, don't run your Realm VM. In this case, we
> don't need to do anything.
>
> 2. Add KVM_CAP_ARM_VM_RMM_IPA_SIZE to expose the RMM limit
> for the VMM to choose.
>
> 3. VMM to create a Realm VM using the default IPA Size and then
> check the KVM_CAP_ARM_VM_IPA_SIZE on the "kvm" instance (which
> is Realm) and get the RMM IPA limit.
>
> I prefer 2 or 1, in that order of preference. Happy to hear suggestions.
My gut feeling is that we should go for 1, and add 2 if we actually see
a need for it.
I'm struggling to see why the VMM would want to choose the IPA size
dynamically (surely the attestation server doesn't want to deal with
different IPA sizes?). So the IPA size is going to be picked by the
container owner based on actual need and the VMM can attempt to create a
VM and fail if the platform doesn't support that size. In that case
because attestation would fail with a smaller IPA size the correct
response is failure.
I'll stick something in the commit message to record that.
>>
>> Co-developed-by: Suzuki K Poulose <[email protected]>
>> Signed-off-by: Suzuki K Poulose <[email protected]>
>> Signed-off-by: Steven Price <[email protected]>
>> Signed-off-by: Jean-Philippe Brucker <[email protected]>
>> ---
>> arch/arm64/include/asm/kvm_emulate.h | 5 +
>> arch/arm64/include/asm/kvm_rme.h | 19 ++
>> arch/arm64/kvm/arm.c | 18 ++
>> arch/arm64/kvm/mmu.c | 15 +-
>> arch/arm64/kvm/rme.c | 282 +++++++++++++++++++++++++++
>> 5 files changed, 337 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/arm64/include/asm/kvm_emulate.h
>> b/arch/arm64/include/asm/kvm_emulate.h
>> index 6f08398537e2..c606316f4729 100644
>> --- a/arch/arm64/include/asm/kvm_emulate.h
>> +++ b/arch/arm64/include/asm/kvm_emulate.h
>> @@ -624,6 +624,11 @@ static inline enum realm_state
>> kvm_realm_state(struct kvm *kvm)
>> return READ_ONCE(kvm->arch.realm.state);
>> }
>> +static inline bool kvm_realm_is_created(struct kvm *kvm)
>> +{
>> + return kvm_is_realm(kvm) && kvm_realm_state(kvm) !=
>> REALM_STATE_NONE;
>> +}
>> +
>> static inline bool vcpu_is_rec(struct kvm_vcpu *vcpu)
>> {
>> return false;
>> diff --git a/arch/arm64/include/asm/kvm_rme.h
>> b/arch/arm64/include/asm/kvm_rme.h
>> index 922da3f47227..cf8cc4d30364 100644
>> --- a/arch/arm64/include/asm/kvm_rme.h
>> +++ b/arch/arm64/include/asm/kvm_rme.h
>> @@ -6,6 +6,8 @@
>> #ifndef __ASM_KVM_RME_H
>> #define __ASM_KVM_RME_H
>> +#include <uapi/linux/kvm.h>
>> +
>> /**
>> * enum realm_state - State of a Realm
>> */
>> @@ -46,11 +48,28 @@ enum realm_state {
>> * struct realm - Additional per VM data for a Realm
>> *
>> * @state: The lifetime state machine for the realm
>> + * @rd: Kernel mapping of the Realm Descriptor (RD)
>> + * @params: Parameters for the RMI_REALM_CREATE command
>> + * @num_aux: The number of auxiliary pages required by the RMM
>> + * @vmid: VMID to be used by the RMM for the realm
>> + * @ia_bits: Number of valid Input Address bits in the IPA
>> */
>> struct realm {
>> enum realm_state state;
>> +
>> + void *rd;
>> + struct realm_params *params;
>> +
>> + unsigned long num_aux;
>> + unsigned int vmid;
>> + unsigned int ia_bits;
>> };
>> int kvm_init_rme(void);
>> +u32 kvm_realm_ipa_limit(void);
>> +
>> +int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap);
>> +int kvm_init_realm_vm(struct kvm *kvm);
>> +void kvm_destroy_realm(struct kvm *kvm);
>> #endif
>> diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
>> index 2056c660c5ee..5729ea430d6d 100644
>> --- a/arch/arm64/kvm/arm.c
>> +++ b/arch/arm64/kvm/arm.c
>> @@ -119,6 +119,13 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
>> }
>> mutex_unlock(&kvm->slots_lock);
>> break;
>> + case KVM_CAP_ARM_RME:
>> + if (!kvm_is_realm(kvm))
>> + return -EINVAL;
>> + mutex_lock(&kvm->lock);
>> + r = kvm_realm_enable_cap(kvm, cap);
>> + mutex_unlock(&kvm->lock);
>> + break;
>> default:
>> r = -EINVAL;
>> break;
>> @@ -179,6 +186,13 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned
>> long type)
>> bitmap_zero(kvm->arch.vcpu_features, KVM_VCPU_MAX_FEATURES);
>> + /* Initialise the realm bits after the generic bits are enabled */
>> + if (kvm_is_realm(kvm)) {
>> + ret = kvm_init_realm_vm(kvm);
>> + if (ret)
>> + goto err_free_cpumask;
>> + }
>> +
>> return 0;
>> err_free_cpumask:
>> @@ -219,6 +233,7 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
>> kvm_unshare_hyp(kvm, kvm + 1);
>> kvm_arm_teardown_hypercalls(kvm);
>> + kvm_destroy_realm(kvm);
>> }
>> int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
>> @@ -328,6 +343,9 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm,
>> long ext)
>> case KVM_CAP_ARM_SUPPORTED_REG_MASK_RANGES:
>> r = BIT(0);
>> break;
>> + case KVM_CAP_ARM_RME:
>> + r = static_key_enabled(&kvm_rme_is_available);
>> + break;
>> default:
>> r = 0;
>> }
>> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
>> index 18680771cdb0..aae365647b62 100644
>> --- a/arch/arm64/kvm/mmu.c
>> +++ b/arch/arm64/kvm/mmu.c
>> @@ -872,6 +872,10 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct
>> kvm_s2_mmu *mmu, unsigned long t
>> struct kvm_pgtable *pgt;
>> u64 mmfr0, mmfr1;
>> u32 phys_shift;
>> + u32 ipa_limit = kvm_ipa_limit;
>> +
>> + if (kvm_is_realm(kvm))
>> + ipa_limit = kvm_realm_ipa_limit();
>> if (type & ~KVM_VM_TYPE_ARM_IPA_SIZE_MASK)
>> return -EINVAL;
>> @@ -880,12 +884,12 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct
>> kvm_s2_mmu *mmu, unsigned long t
>> if (is_protected_kvm_enabled()) {
>> phys_shift = kvm_ipa_limit;
>> } else if (phys_shift) {
>> - if (phys_shift > kvm_ipa_limit ||
>> + if (phys_shift > ipa_limit ||
>> phys_shift < ARM64_MIN_PARANGE_BITS)
>> return -EINVAL;
>> } else {
>> phys_shift = KVM_PHYS_SHIFT;
>> - if (phys_shift > kvm_ipa_limit) {
>> + if (phys_shift > ipa_limit) {
>> pr_warn_once("%s using unsupported default IPA limit,
>> upgrade your VMM\n",
>> current->comm);
>> return -EINVAL;
>> @@ -1014,6 +1018,13 @@ void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu)
>> struct kvm_pgtable *pgt = NULL;
>> write_lock(&kvm->mmu_lock);
>> + if (kvm_is_realm(kvm) &&
>> + (kvm_realm_state(kvm) != REALM_STATE_DEAD &&
>> + kvm_realm_state(kvm) != REALM_STATE_NONE)) {
>> + /* TODO: teardown rtts */
>> + write_unlock(&kvm->mmu_lock);
>> + return;
>> + }
>
> This needs a comment to explain the rationale of deferring the
> Stage2 pgt freeing. Something like :
>
> /*
> * For realms, we can free the entry level RTTs
> * only after :
> * 1. All of the stage2 mappings are torn down.
> * 2. The Realm has been destroyed.
> *
> * So, come back later once the RD has been destroyed.
> */
Actually the 'TODO' is really "this is handled in a future patch". I'll
make that more obvious.
>> pgt = mmu->pgt;
>> if (pgt) {
>> mmu->pgd_phys = 0;
>> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
>> index 3dbbf9d046bf..658d14e8d87d 100644
>> --- a/arch/arm64/kvm/rme.c
>> +++ b/arch/arm64/kvm/rme.c
>> @@ -5,9 +5,20 @@
>> #include <linux/kvm_host.h>
>> +#include <asm/kvm_emulate.h>
>> +#include <asm/kvm_mmu.h>
>> #include <asm/rmi_cmds.h>
>> #include <asm/virt.h>
>> +#include <asm/kvm_pgtable.h>
>> +
>> +static unsigned long rmm_feat_reg0;
>> +
>> +static bool rme_supports(unsigned long feature)
>> +{
>> + return !!u64_get_bits(rmm_feat_reg0, feature);
>> +}
>> +
>> static int rmi_check_version(void)
>> {
>> struct arm_smccc_res res;
>> @@ -36,8 +47,272 @@ static int rmi_check_version(void)
>> return 0;
>> }
>> +u32 kvm_realm_ipa_limit(void)
>> +{
>> + return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
>> +}
>> +
>> +static int get_start_level(struct realm *realm)
>> +{
>> + return 4 - stage2_pgtable_levels(realm->ia_bits);
>> +}
>> +
>> +static int realm_create_rd(struct kvm *kvm)
>> +{
>> + struct realm *realm = &kvm->arch.realm;
>> + struct realm_params *params = realm->params;
>> + void *rd = NULL;
>> + phys_addr_t rd_phys, params_phys;
>> + struct kvm_pgtable *pgt = kvm->arch.mmu.pgt;
>> + int i, r;
>> +
>> + if (WARN_ON(realm->rd) || WARN_ON(!realm->params))
>> + return -EEXIST;
>> +
>> + rd = (void *)__get_free_page(GFP_KERNEL);
>> + if (!rd)
>> + return -ENOMEM;
>> +
>> + rd_phys = virt_to_phys(rd);
>> + if (rmi_granule_delegate(rd_phys)) {
>> + r = -ENXIO;
>> + goto out;
>
> super minor nit: s/out/free_rd/ is a bit more readable. Here "out" is
> only used for error exits and could be confusing.
It's a good point - I'll fix.
>> + }
>> +
>> + for (i = 0; i < pgt->pgd_pages; i++) {
>> + phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
>> +
>> + if (rmi_granule_delegate(pgd_phys)) {
>> + r = -ENXIO;
>> + goto out_undelegate_tables;
>> + }
>> + }
>> +
>> + realm->ia_bits = VTCR_EL2_IPA(kvm->arch.mmu.vtcr);
>> +
>> + params->rtt_level_start = get_start_level(realm);
>> + params->rtt_num_start = pgt->pgd_pages;
>> + params->rtt_base = kvm->arch.mmu.pgd_phys;
>> + params->vmid = realm->vmid;
>> +
>> + params_phys = virt_to_phys(params);
>> +
>> + if (rmi_realm_create(rd_phys, params_phys)) {
>> + r = -ENXIO;
>> + goto out_undelegate_tables;
>> + }
>> +
>> + realm->rd = rd;
>> +
>> + if (WARN_ON(rmi_rec_aux_count(rd_phys, &realm->num_aux))) {
>> + WARN_ON(rmi_realm_destroy(rd_phys));
>> + goto out_undelegate_tables;
>> + }
>> +
>> + return 0;
>> +
>> +out_undelegate_tables:
>> + while (--i >= 0) {
>> + phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
>> +
>> + WARN_ON(rmi_granule_undelegate(pgd_phys));
>> + }
>> + WARN_ON(rmi_granule_undelegate(rd_phys));
>> +out:
>> + free_page((unsigned long)rd);
>> + return r;
>> +}
>> +
>> +/* Protects access to rme_vmid_bitmap */
>> +static DEFINE_SPINLOCK(rme_vmid_lock);
>> +static unsigned long *rme_vmid_bitmap;
>> +
>> +static int rme_vmid_init(void)
>> +{
>> + unsigned int vmid_count = 1 << kvm_get_vmid_bits();
>
> minor nit: RMM has a fixed VMID width of 16bits. Do we need to
> explicitly use that, instead of relying what kvm thinks ? (Though
> in practise, this would only be a problem if the architecture
> evolves to support something more).
I don't think the RMM has a fixed VMID width according to the spec. It says:
The VMID of a Realm is chosen by the Host. The VMID must be within the
range supported by the hardware platform. The RMM ensures that every
Realm on the system has a unique VMID.
It also declares a pseudo-code VmidIsValid function which says:
If the underlying hardware platform does not implement FEAT_VMID16 then
a VMID value with vmid[15:8] != 0 is invalid.
So it would seem from the spec that a VMID width <16 bits is supported.
On the other hand if we ever get >16 bit VMID support in the
architecture then the 16 bit vmid field in RmiRealmParams is clearly
going to trip us up.
At the moment my feeling is that 16 bit VMIDs ought to be enough for
anybody ... ;)
>> +
>> + rme_vmid_bitmap = bitmap_zalloc(vmid_count, GFP_KERNEL);
>> + if (!rme_vmid_bitmap) {
>> + kvm_err("%s: Couldn't allocate rme vmid bitmap\n", __func__);
>> + return -ENOMEM;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static int rme_vmid_reserve(void)
>> +{
>> + int ret;
>> + unsigned int vmid_count = 1 << kvm_get_vmid_bits();
>> +
>> + spin_lock(&rme_vmid_lock);
>> + ret = bitmap_find_free_region(rme_vmid_bitmap, vmid_count, 0);
>> + spin_unlock(&rme_vmid_lock);
>> +
>> + return ret;
>> +}
>> +
>> +static void rme_vmid_release(unsigned int vmid)
>> +{
>> + spin_lock(&rme_vmid_lock);
>> + bitmap_release_region(rme_vmid_bitmap, vmid, 0);
>> + spin_unlock(&rme_vmid_lock);
>> +}
>> +
>> +static int kvm_create_realm(struct kvm *kvm)
>> +{
>> + struct realm *realm = &kvm->arch.realm;
>> + int ret;
>> +
>> + if (!kvm_is_realm(kvm) || kvm_realm_is_created(kvm))
>> + return -EEXIST;
>
> Minor nit:
>
> if (!kvm_is_realm(kvm)
> return -EIO or even -EINVAL ?
>
> if (kvm_realm_is_created(kvm))
> return -EEXIST;
Indeed EEXIST is a terrible error code for !realm. My preference is EIO.
>> +
>> + ret = rme_vmid_reserve();
>> + if (ret < 0)
>> + return ret;
>> + realm->vmid = ret;
>> +
>> + ret = realm_create_rd(kvm);
>> + if (ret) {
>> + rme_vmid_release(realm->vmid);
>> + return ret;
>> + }
>> +
>> + WRITE_ONCE(realm->state, REALM_STATE_NEW);
>> +
>> + /* The realm is up, free the parameters. */
>> + free_page((unsigned long)realm->params);
>> + realm->params = NULL;
>> +
>> + return 0;
>> +}
>> +
>> +static int config_realm_hash_algo(struct realm *realm,
>> + struct kvm_cap_arm_rme_config_item *cfg)
>> +{
>> + switch (cfg->hash_algo) {
>> + case KVM_CAP_ARM_RME_MEASUREMENT_ALGO_SHA256:
>> + if (!rme_supports(RMI_FEATURE_REGISTER_0_HASH_SHA_256))
>> + return -EINVAL;
>> + break;
>> + case KVM_CAP_ARM_RME_MEASUREMENT_ALGO_SHA512:
>> + if (!rme_supports(RMI_FEATURE_REGISTER_0_HASH_SHA_512))
>> + return -EINVAL;
>
> Do we need to add a comment here on why we don't expose the supported
> "hash" algo as part of the UABI ? Something like :
>
> /*
> * The hash algorithm for the measurements is a choosen by
> * the Realm owner (since it affects the attestation), we
> * would like the owner to get what they wants.
> */
I'm not really sure here's the right place for such a comment. Clearly
here we're just doing whatever the VMM requested and returning -EINVAL
if it's not supported (by KVM or the RMM).
There is the bigger question on whether discovery of the RMM's feature
set is necessary and easy enough with the current uAPI. As you say there
is a reasonable argument that the VMM doesn't have a choice so there's
no need for discovery. Equally adding an extra CAP ioctl for discovery
later is easy enough.
>> + break;
>> + default:
>> + return -EINVAL;
>> + }
>> + realm->params->hash_algo = cfg->hash_algo;
>> + return 0;
>> +}
>> +
>> +static int kvm_rme_config_realm(struct kvm *kvm, struct
>> kvm_enable_cap *cap)
>> +{
>> + struct kvm_cap_arm_rme_config_item cfg;
>> + struct realm *realm = &kvm->arch.realm;
>> + int r = 0;
>> +
>> + if (kvm_realm_is_created(kvm))
>> + return -EINVAL;
>
> minor nit: May be return -EEXIST or -EIO rather than, "Invalid
> (parameter)" ?
Yep, although EBUSY seems like the most descriptive to me.
>
>> +
>> + if (copy_from_user(&cfg, (void __user *)cap->args[1], sizeof(cfg)))
>> + return -EFAULT;
>> +
>> + switch (cfg.cfg) {
>> + case KVM_CAP_ARM_RME_CFG_RPV:
>> + memcpy(&realm->params->rpv, &cfg.rpv, sizeof(cfg.rpv));
>> + break;
>> + case KVM_CAP_ARM_RME_CFG_HASH_ALGO:
>> + r = config_realm_hash_algo(realm, &cfg);
>> + break;
>> + default:
>> + r = -EINVAL;
>> + }
>> +
>> + return r;
>> +}
>> +
>> +int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
>> +{
>> + int r = 0;
>> +
>> + if (!kvm_is_realm(kvm))
>> + return -EINVAL;
>> +
>> + switch (cap->args[0]) {
>> + case KVM_CAP_ARM_RME_CONFIG_REALM:
>> + r = kvm_rme_config_realm(kvm, cap);
>> + break;
>> + case KVM_CAP_ARM_RME_CREATE_RD:
>> + r = kvm_create_realm(kvm);
>> + break;
>> + default:
>> + r = -EINVAL;
>> + break;
>> + }
>> +
>> + return r;
>> +}
>> +
>> +void kvm_destroy_realm(struct kvm *kvm)
>> +{
>> + struct realm *realm = &kvm->arch.realm;
>> + struct kvm_pgtable *pgt = kvm->arch.mmu.pgt;
>> + int i;
>> +
>> + if (realm->params) {
>> + free_page((unsigned long)realm->params);
>> + realm->params = NULL;
>> + }
>> +
>> + if (!kvm_realm_is_created(kvm))
>> + return;
>> +
>> + WRITE_ONCE(realm->state, REALM_STATE_DYING);
>> +
>> + if (realm->rd) {
>> + phys_addr_t rd_phys = virt_to_phys(realm->rd);
>> +
>> + if (WARN_ON(rmi_realm_destroy(rd_phys)))
>> + return;
>> + if (WARN_ON(rmi_granule_undelegate(rd_phys)))
>> + return;
>> + free_page((unsigned long)realm->rd);
>> + realm->rd = NULL;
>> + }
>> +
>> + rme_vmid_release(realm->vmid);
>> +
>> + for (i = 0; i < pgt->pgd_pages; i++) {
>> + phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
>> +
>> + if (WARN_ON(rmi_granule_undelegate(pgd_phys)))
>> + return;
>> + }
>> +
>> + WRITE_ONCE(realm->state, REALM_STATE_DEAD);
>> +
>
> May be add in a comment here:
>
> /* Now that the Realm is destroyed, free the entry level RTTs */
>> + kvm_free_stage2_pgd(&kvm->arch.mmu);
>
>
>
>> +}
>> +
>> +int kvm_init_realm_vm(struct kvm *kvm)
>> +{
>> + struct realm_params *params;
>> +
>> + params = (struct realm_params *)get_zeroed_page(GFP_KERNEL);
>> + if (!params)
>> + return -ENOMEM;
>> +
>> + /* Default parameters, not exposed to user space */
>
> This is a bit misleading. The value comes from the userspace and...
>
>> + params->s2sz = VTCR_EL2_IPA(kvm->arch.mmu.vtcr);
>
> (minor nit) we initialise most of the params, including those that come
> from KVM later. So, as such may be a good idea to move it together at
> create_realm, unless we need it.
Indeed I'll move this to create_realm.
Thanks,
Steve
>> + kvm->arch.realm.params = params;
>> + return 0;
>> +}
>> +
>> int kvm_init_rme(void)
>> {
>> + int ret;
>> +
>> if (PAGE_SIZE != SZ_4K)
>> /* Only 4k page size on the host is supported */
>> return 0;
>> @@ -46,6 +321,13 @@ int kvm_init_rme(void)
>> /* Continue without realm support */
>> return 0;
>> + if (WARN_ON(rmi_features(0, &rmm_feat_reg0)))
>> + return 0;
>> +
>> + ret = rme_vmid_init();
>> + if (ret)
>> + return ret;
>> +
>> /* Future patch will enable static branch kvm_rme_is_available */
>> return 0;
>
>
> Suzuki
>
On 17/04/2024 14:37, Suzuki K Poulose wrote:
> Hi Steven
>
> minort nit, Subject: arm64: RME: RTT tear down
>
> This patch is all about tearing the RTTs, so may be the subject could
> be adjusted accordingly.
Good point, this patch has evolved and is now all about tearing down.
> On 12/04/2024 09:42, Steven Price wrote:
>> The RMM owns the stage 2 page tables for a realm, and KVM must request
>> that the RMM creates/destroys entries as necessary. The physical pages
>> to store the page tables are delegated to the realm as required, and can
>> be undelegated when no longer used.
>>
>> Creating new RTTs is the easy part, tearing down is a little more
>> tricky. The result of realm_rtt_destroy() can be used to effectively
>> walk the tree and destroy the entries (undelegating pages that were
>> given to the realm).
>
> The patch looks functionally correct to me. Some minor style related
> comments below.
>
>> Signed-off-by: Steven Price <[email protected]>
>
>> ---
>> arch/arm64/include/asm/kvm_rme.h | 19 ++++
>> arch/arm64/kvm/mmu.c | 6 +-
>> arch/arm64/kvm/rme.c | 171 +++++++++++++++++++++++++++++++
>> 3 files changed, 193 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/arm64/include/asm/kvm_rme.h
>> b/arch/arm64/include/asm/kvm_rme.h
>> index fba85e9ce3ae..4ab5cb5e91b3 100644
>> --- a/arch/arm64/include/asm/kvm_rme.h
>> +++ b/arch/arm64/include/asm/kvm_rme.h
>> @@ -76,5 +76,24 @@ u32 kvm_realm_ipa_limit(void);
>> int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap);
>> int kvm_init_realm_vm(struct kvm *kvm);
>> void kvm_destroy_realm(struct kvm *kvm);
>> +void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits);
>> +
>> +#define RME_RTT_BLOCK_LEVEL 2
>> +#define RME_RTT_MAX_LEVEL 3
>> +
>> +#define RME_PAGE_SHIFT 12
>> +#define RME_PAGE_SIZE BIT(RME_PAGE_SHIFT)
>> +/* See ARM64_HW_PGTABLE_LEVEL_SHIFT() */
>> +#define RME_RTT_LEVEL_SHIFT(l) \
>> + ((RME_PAGE_SHIFT - 3) * (4 - (l)) + 3)
>> +#define RME_L2_BLOCK_SIZE BIT(RME_RTT_LEVEL_SHIFT(2))
>> +
>> +static inline unsigned long rme_rtt_level_mapsize(int level)
>> +{
>> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
>> + return RME_PAGE_SIZE;
>> +
>> + return (1UL << RME_RTT_LEVEL_SHIFT(level));
>> +}
>>
>
> super minor nit: We only support 4K for now, so may be could reuse
> the ARM64 generic macro helpers. I am fine either way.
Given we'll likely want to support host granules other than 4k in the
future I'd like to avoid using the generic ones. It's also a clear
signal that the code is referring to the RTTs rather than the host's
page tables.
>
>> #endif
>> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
>> index af4564f3add5..46f0c4e80ace 100644
>> --- a/arch/arm64/kvm/mmu.c
>> +++ b/arch/arm64/kvm/mmu.c
>> @@ -1012,17 +1012,17 @@ void stage2_unmap_vm(struct kvm *kvm)
>> void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu)
>> {
>> struct kvm *kvm = kvm_s2_mmu_to_kvm(mmu);
>> - struct kvm_pgtable *pgt = NULL;
>> + struct kvm_pgtable *pgt;
>> write_lock(&kvm->mmu_lock);
>> + pgt = mmu->pgt;
>> if (kvm_is_realm(kvm) &&
>> (kvm_realm_state(kvm) != REALM_STATE_DEAD &&
>> kvm_realm_state(kvm) != REALM_STATE_NONE)) {
>> - /* TODO: teardown rtts */
>> write_unlock(&kvm->mmu_lock);
>> + kvm_realm_destroy_rtts(kvm, pgt->ia_bits);
>> return;
>> }
>> - pgt = mmu->pgt;
>> if (pgt) {
>> mmu->pgd_phys = 0;
>> mmu->pgt = NULL;
>> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
>> index 9652ec6ab2fd..09b59bcad8b6 100644
>> --- a/arch/arm64/kvm/rme.c
>> +++ b/arch/arm64/kvm/rme.c
>> @@ -47,6 +47,53 @@ static int rmi_check_version(void)
>> return 0;
>> }
>> +static phys_addr_t __alloc_delegated_page(struct realm *realm,
>> + struct kvm_mmu_memory_cache *mc,
>> + gfp_t flags)
>
> minor nit: Do we need "__" here ? The counter part is plain
> free_delegated_page without the "__". We could drop the prefix
>
> Or we could split the function as:
>
> alloc_delegated_page()
> {
> if (spare_page_available)
> return spare_page;
> return __alloc_delegated_page(); /* Alloc and delegate a page */
> }
I'm not really sure I follow. The reason for the 'wrapper' function
alloc_delegated_page() is because most call sites don't care about the
GFP flags (defaults to GFP_KERNEL), but for ensure_spare_page() we need
to pass GFP_ATOMIC.
Admittedly there are only 3 call sites in total and the wrapper isn't
added yet. I'll tidy this up by simply adding the GFP_KERNEL flag onto
the two call sites.
>
>> +{
>> + phys_addr_t phys = PHYS_ADDR_MAX;
>> + void *virt;
>> +
>> + if (realm->spare_page != PHYS_ADDR_MAX) {
>> + swap(realm->spare_page, phys);
>> + goto out;
>> + }
>> +
>> + if (mc)
>> + virt = kvm_mmu_memory_cache_alloc(mc);
>> + else
>> + virt = (void *)__get_free_page(flags);
>> +
>> + if (!virt)
>> + goto out;
>> +
>> + phys = virt_to_phys(virt);
>> +
>> + if (rmi_granule_delegate(phys)) {
>> + free_page((unsigned long)virt);
>> +
>> + phys = PHYS_ADDR_MAX;
>> + }
>> +
>> +out:
>> + return phys;
>> +}
>> +
>> +static void free_delegated_page(struct realm *realm, phys_addr_t phys)
>> +{
>> + if (realm->spare_page == PHYS_ADDR_MAX) {
>> + realm->spare_page = phys;
>> + return;
>> + }
>> +
>> + if (WARN_ON(rmi_granule_undelegate(phys))) {
>> + /* Undelegate failed: leak the page */
>> + return;
>> + }
>> +
>> + free_page((unsigned long)phys_to_virt(phys));
>> +}
>> +
>> u32 kvm_realm_ipa_limit(void)
>> {
>> return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
>> @@ -124,6 +171,130 @@ static int realm_create_rd(struct kvm *kvm)
>> return r;
>> }
>> +static int realm_rtt_destroy(struct realm *realm, unsigned long addr,
>> + int level, phys_addr_t *rtt_granule,
>> + unsigned long *next_addr)
>> +{
>> + unsigned long out_rtt;
>> + unsigned long out_top;
>> + int ret;
>> +
>> + ret = rmi_rtt_destroy(virt_to_phys(realm->rd), addr, level,
>> + &out_rtt, &out_top);
>> +
>> + if (rtt_granule)
>> + *rtt_granule = out_rtt;
>> + if (next_addr)
>> + *next_addr = out_top;
>
> minor nit: As mentioned in the previous patch, we could move this check
> to the rmi_rtt_destroy().
Done, I've also dropped the check for rtt_granule - it's a bug to be
passing that as NULL.
>> +
>> + return ret;
>> +}
>> +
>> +static int realm_tear_down_rtt_level(struct realm *realm, int level,
>> + unsigned long start, unsigned long end)
>> +{
>> + ssize_t map_size;
>> + unsigned long addr, next_addr;
>> +
>> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
>> + return -EINVAL;
>> +
>> + map_size = rme_rtt_level_mapsize(level - 1);
>> +
>> + for (addr = start; addr < end; addr = next_addr) {
>> + phys_addr_t rtt_granule;
>> + int ret;
>> + unsigned long align_addr = ALIGN(addr, map_size);
>> +
>> + next_addr = ALIGN(addr + 1, map_size);
>> +
>> + if (next_addr <= end && align_addr == addr) {
>> + ret = realm_rtt_destroy(realm, addr, level,
>> + &rtt_granule, &next_addr);
>> + } else {
>> + /* Recurse a level deeper */
>> + ret = realm_tear_down_rtt_level(realm,
>> + level + 1,
>> + addr,
>> + min(next_addr, end));
>> + if (ret)
>> + return ret;
>> + continue;
>> + }
>
> I think it would be better readable if we did something like :
>
> /*
> * The target range is smaller than what this level
> * covers. Go deeper.
> */
> if (next_addr > end || align_addr != addr) {
> ret = realm_tear_down_rtt_level(realm,
> level + 1, addr,
> min(next_addr, end));
> if (ret)
> return ret;
> continue;
> }
>
> ret = realm_rtt_destroy(realm, addr, level,
> &rtt_granule, &next_addr);
Yes that seems clearly, thanks for the suggestion.
>> +
>> + switch (RMI_RETURN_STATUS(ret)) {
>> + case RMI_SUCCESS:
>> + if (!WARN_ON(rmi_granule_undelegate(rtt_granule)))
>> + free_page((unsigned long)phys_to_virt(rtt_granule));
>> + break;
>> + case RMI_ERROR_RTT:
>> + if (next_addr > addr) {
>> + /* unassigned or destroyed */
>
> minor nit:
> /* RTT doesn't exist, skip */
Indeed, this comment is out of date - the spec now calls this condition
"Missing RTT" so I'll use that wording.
>
>> + break;
>> + }
>
>> + if (WARN_ON(RMI_RETURN_INDEX(ret) != level))
>> + return -EBUSY;
>
> In practise, we only call this for the full IPA range and we wouldn't go
> deeper, if the top level entry was missing. So, there is no reason why
> the RMM didn't walk to the requested level. May be we could add a
> comment here :
> /*
> * We tear down the RTT range for the full IPA
> * space, after everything is unmapped. Also we
> * descend down only if we cannot tear down a
> * top level RTT. Thus RMM must be able to walk
> * to the requested level. e.g., a block mapping
> * exists at L1 or L2.
> */
Sure, will add.
>> + if (WARN_ON(level == RME_RTT_MAX_LEVEL)) {
>> + // Live entry
>> + return -EBUSY;
>
>
> The first part of the comment above applies to this. So may be it is
> good to have it.
>
>
>> + }
>
>> + /* Recurse a level deeper */
>
> minor nit:
> /*
> * The table has active entries in it, recurse
> * deeper and tear down the RTTs.
> */
Sure
>> + next_addr = ALIGN(addr + 1, map_size);
>> + ret = realm_tear_down_rtt_level(realm,
>> + level + 1,
>> + addr,
>> + next_addr);
>> + if (ret)
>> + return ret;
>> + /* Try again at this level */
>
> /*
> * Now that the children RTTs are destroyed,
> * retry at this level.
> */
Sure
>> + next_addr = addr;
>> + break;
>> + default:
>> + WARN_ON(1);
>> + return -ENXIO;
>> + }
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static int realm_tear_down_rtt_range(struct realm *realm,
>> + unsigned long start, unsigned long end)
>> +{
>> + return realm_tear_down_rtt_level(realm, get_start_level(realm) + 1,
>> + start, end);
>> +}
>> +
>> +static void ensure_spare_page(struct realm *realm)
>> +{
>> + phys_addr_t tmp_rtt;
>> +
>> + /*
>> + * Make sure we have a spare delegated page for tearing down the
>> + * block mappings. We do this by allocating then freeing a page.
>> + * We must use Atomic allocations as we are called with
>> kvm->mmu_lock
>> + * held.
>> + */
>> + tmp_rtt = __alloc_delegated_page(realm, NULL, GFP_ATOMIC);
>> +
>> + /*
>> + * If the allocation failed, continue as we may not have a block
>> level
>> + * mapping so it may not be fatal, otherwise free it to assign it
>> + * to the spare page.
>> + */
>> + if (tmp_rtt != PHYS_ADDR_MAX)
>> + free_delegated_page(realm, tmp_rtt);
>> +}
>> +
>> +void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits)
>> +{
>> + struct realm *realm = &kvm->arch.realm;
>> +
>> + ensure_spare_page(realm);
>> +
>> + WARN_ON(realm_tear_down_rtt_range(realm, 0, (1UL << ia_bits)));
>> +}
>
> minor nit: We don't seem to be using the "spare_page" yet in this patch.
> May be a good idea to move all the related changes
> (alloc_delegated_page() / free_delegated_page, ensure_spare_page() etc)
> to the patch where it may be better suited ?
Good point - I think the calls get added in "arm64: RME: Allow VMM to
set RIPAS". I'll try to move them there.
Thanks,
Steve
> Suzuki
>
>> +
>> /* Protects access to rme_vmid_bitmap */
>> static DEFINE_SPINLOCK(rme_vmid_lock);
>> static unsigned long *rme_vmid_bitmap;
>
Hi,
On Fri, Apr 12, 2024 at 9:43 AM Steven Price <[email protected]> wrote:
>
> From: Sean Christopherson <[email protected]>
>
> Add flags to "struct kvm_gfn_range" to let notifier events target only
> shared and only private mappings, and write up the existing mmu_notifier
> events to be shared-only (private memory is never associated with a
> userspace virtual address, i.e. can't be reached via mmu_notifiers).
>
> Add two flags so that KVM can handle the three possibilities (shared,
> private, and shared+private) without needing something like a tri-state
> enum.
>
> Link: https://lore.kernel.org/all/[email protected]
> Signed-off-by: Sean Christopherson <[email protected]>
> Signed-off-by: Steven Price <[email protected]>
> ---
> include/linux/kvm_host.h | 2 ++
> virt/kvm/kvm_main.c | 7 +++++++
> 2 files changed, 9 insertions(+)
>
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index 48f31dcd318a..c7581360fd88 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -268,6 +268,8 @@ struct kvm_gfn_range {
> gfn_t start;
> gfn_t end;
> union kvm_mmu_notifier_arg arg;
> + bool only_private;
> + bool only_shared;
> bool may_block;
> };
> bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index fb49c2a60200..3486ceef6f4e 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -633,6 +633,13 @@ static __always_inline kvm_mn_ret_t __kvm_handle_hva_range(struct kvm *kvm,
> * the second or later invocation of the handler).
> */
> gfn_range.arg = range->arg;
> +
> + /*
> + * HVA-based notifications aren't relevant to private
> + * mappings as they don't have a userspace mapping.
> + */
> + gfn_range.only_private = false;
> + gfn_range.only_shared = true;
> gfn_range.may_block = range->may_block;
I'd discussed this with Sean when he posted this earlier. Having two
booleans to encode three valid states could be confusing. In response,
Sean suggested using an enum instead:
https://lore.kernel.org/all/[email protected]/
Cheers,
/fuad
>
> /*
> --
> 2.34.1
>
Hi,
On Fri, Apr 12, 2024 at 9:43 AM Steven Price <[email protected]> wrote:
>
> Each page within the protected region of the realm guest can be marked
> as either RAM or EMPTY. Allow the VMM to control this before the guest
> has started and provide the equivalent functions to change this (with
> the guest's approval) at runtime.
>
> When transitioning from RIPAS RAM (1) to RIPAS EMPTY (0) the memory is
> unmapped from the guest and undelegated allowing the memory to be reused
> by the host. When transitioning to RIPAS RAM the actual population of
> the leaf RTTs is done later on stage 2 fault, however it may be
> necessary to allocate additional RTTs to represent the range requested.
>
> When freeing a block mapping it is necessary to temporarily unfold the
> RTT which requires delegating an extra page to the RMM, this page can
> then be recovered once the contents of the block mapping have been
> freed. A spare, delegated page (spare_page) is used for this purpose.
>
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/include/asm/kvm_rme.h | 16 ++
> arch/arm64/kvm/mmu.c | 8 +-
> arch/arm64/kvm/rme.c | 390 +++++++++++++++++++++++++++++++
> 3 files changed, 411 insertions(+), 3 deletions(-)
>
> diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
> index 915e76068b00..cc8f81cfc3c0 100644
> --- a/arch/arm64/include/asm/kvm_rme.h
> +++ b/arch/arm64/include/asm/kvm_rme.h
> @@ -96,6 +96,14 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits);
> int kvm_create_rec(struct kvm_vcpu *vcpu);
> void kvm_destroy_rec(struct kvm_vcpu *vcpu);
>
> +void kvm_realm_unmap_range(struct kvm *kvm,
> + unsigned long ipa,
> + u64 size,
> + bool unmap_private);
> +int realm_set_ipa_state(struct kvm_vcpu *vcpu,
> + unsigned long addr, unsigned long end,
> + unsigned long ripas);
> +
> #define RME_RTT_BLOCK_LEVEL 2
> #define RME_RTT_MAX_LEVEL 3
>
> @@ -114,4 +122,12 @@ static inline unsigned long rme_rtt_level_mapsize(int level)
> return (1UL << RME_RTT_LEVEL_SHIFT(level));
> }
>
> +static inline bool realm_is_addr_protected(struct realm *realm,
> + unsigned long addr)
> +{
> + unsigned int ia_bits = realm->ia_bits;
> +
> + return !(addr & ~(BIT(ia_bits - 1) - 1));
> +}
> +
> #endif
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index 46f0c4e80ace..8a7b5449697f 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -310,6 +310,7 @@ static void invalidate_icache_guest_page(void *va, size_t size)
> * @start: The intermediate physical base address of the range to unmap
> * @size: The size of the area to unmap
> * @may_block: Whether or not we are permitted to block
> + * @only_shared: If true then protected mappings should not be unmapped
> *
> * Clear a range of stage-2 mappings, lowering the various ref-counts. Must
> * be called while holding mmu_lock (unless for freeing the stage2 pgd before
> @@ -317,7 +318,7 @@ static void invalidate_icache_guest_page(void *va, size_t size)
> * with things behind our backs.
> */
> static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64 size,
> - bool may_block)
> + bool may_block, bool only_shared)
I found the new added only_shared parameter to be a bit confusing,
since this patch also introduces kvm_realm_unmap_range(..., bool
unmap_private), where unmap_private has a different meaning, which I
think is really unmap_all.. It might be better if the parameter meant
the same thing everywhere. Having helpers to sort this out, similar
to what you have with realm_unmap_range_*() might be even clearer.
Thanks,
/fuad
> {
> struct kvm *kvm = kvm_s2_mmu_to_kvm(mmu);
> phys_addr_t end = start + size;
> @@ -330,7 +331,7 @@ static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64
>
> static void unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64 size)
> {
> - __unmap_stage2_range(mmu, start, size, true);
> + __unmap_stage2_range(mmu, start, size, true, false);
> }
>
> static void stage2_flush_memslot(struct kvm *kvm,
> @@ -1771,7 +1772,8 @@ bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range)
>
> __unmap_stage2_range(&kvm->arch.mmu, range->start << PAGE_SHIFT,
> (range->end - range->start) << PAGE_SHIFT,
> - range->may_block);
> + range->may_block,
> + range->only_shared);
>
> return false;
> }
> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
> index 629a095bea61..9e5983c51393 100644
> --- a/arch/arm64/kvm/rme.c
> +++ b/arch/arm64/kvm/rme.c
> @@ -79,6 +79,12 @@ static phys_addr_t __alloc_delegated_page(struct realm *realm,
> return phys;
> }
>
> +static phys_addr_t alloc_delegated_page(struct realm *realm,
> + struct kvm_mmu_memory_cache *mc)
> +{
> + return __alloc_delegated_page(realm, mc, GFP_KERNEL);
> +}
> +
> static void free_delegated_page(struct realm *realm, phys_addr_t phys)
> {
> if (realm->spare_page == PHYS_ADDR_MAX) {
> @@ -94,6 +100,151 @@ static void free_delegated_page(struct realm *realm, phys_addr_t phys)
> free_page((unsigned long)phys_to_virt(phys));
> }
>
> +static int realm_rtt_create(struct realm *realm,
> + unsigned long addr,
> + int level,
> + phys_addr_t phys)
> +{
> + addr = ALIGN_DOWN(addr, rme_rtt_level_mapsize(level - 1));
> + return rmi_rtt_create(virt_to_phys(realm->rd), phys, addr, level);
> +}
> +
> +static int realm_rtt_fold(struct realm *realm,
> + unsigned long addr,
> + int level,
> + phys_addr_t *rtt_granule)
> +{
> + unsigned long out_rtt;
> + int ret;
> +
> + ret = rmi_rtt_fold(virt_to_phys(realm->rd), addr, level, &out_rtt);
> +
> + if (RMI_RETURN_STATUS(ret) == RMI_SUCCESS && rtt_granule)
> + *rtt_granule = out_rtt;
> +
> + return ret;
> +}
> +
> +static int realm_destroy_protected(struct realm *realm,
> + unsigned long ipa,
> + unsigned long *next_addr)
> +{
> + unsigned long rd = virt_to_phys(realm->rd);
> + unsigned long addr;
> + phys_addr_t rtt;
> + int ret;
> +
> +loop:
> + ret = rmi_data_destroy(rd, ipa, &addr, next_addr);
> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
> + if (*next_addr > ipa)
> + return 0; /* UNASSIGNED */
> + rtt = alloc_delegated_page(realm, NULL);
> + if (WARN_ON(rtt == PHYS_ADDR_MAX))
> + return -1;
> + /* ASSIGNED - ipa is mapped as a block, so split */
> + ret = realm_rtt_create(realm, ipa,
> + RMI_RETURN_INDEX(ret) + 1, rtt);
> + if (WARN_ON(ret)) {
> + free_delegated_page(realm, rtt);
> + return -1;
> + }
> + /* retry */
> + goto loop;
> + } else if (WARN_ON(ret)) {
> + return -1;
> + }
> + ret = rmi_granule_undelegate(addr);
> +
> + /*
> + * If the undelegate fails then something has gone seriously
> + * wrong: take an extra reference to just leak the page
> + */
> + if (WARN_ON(ret))
> + get_page(phys_to_page(addr));
> +
> + return 0;
> +}
> +
> +static void realm_unmap_range_shared(struct kvm *kvm,
> + int level,
> + unsigned long start,
> + unsigned long end)
> +{
> + struct realm *realm = &kvm->arch.realm;
> + unsigned long rd = virt_to_phys(realm->rd);
> + ssize_t map_size = rme_rtt_level_mapsize(level);
> + unsigned long next_addr, addr;
> + unsigned long shared_bit = BIT(realm->ia_bits - 1);
> +
> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
> + return;
> +
> + start |= shared_bit;
> + end |= shared_bit;
> +
> + for (addr = start; addr < end; addr = next_addr) {
> + unsigned long align_addr = ALIGN(addr, map_size);
> + int ret;
> +
> + next_addr = ALIGN(addr + 1, map_size);
> +
> + if (align_addr != addr || next_addr > end) {
> + /* Need to recurse deeper */
> + if (addr < align_addr)
> + next_addr = align_addr;
> + realm_unmap_range_shared(kvm, level + 1, addr,
> + min(next_addr, end));
> + continue;
> + }
> +
> + ret = rmi_rtt_unmap_unprotected(rd, addr, level, &next_addr);
> + switch (RMI_RETURN_STATUS(ret)) {
> + case RMI_SUCCESS:
> + break;
> + case RMI_ERROR_RTT:
> + if (next_addr == addr) {
> + next_addr = ALIGN(addr + 1, map_size);
> + realm_unmap_range_shared(kvm, level + 1, addr,
> + next_addr);
> + }
> + break;
> + default:
> + WARN_ON(1);
> + }
> + }
> +}
> +
> +static void realm_unmap_range_private(struct kvm *kvm,
> + unsigned long start,
> + unsigned long end)
> +{
> + struct realm *realm = &kvm->arch.realm;
> + ssize_t map_size = RME_PAGE_SIZE;
> + unsigned long next_addr, addr;
> +
> + for (addr = start; addr < end; addr = next_addr) {
> + int ret;
> +
> + next_addr = ALIGN(addr + 1, map_size);
> +
> + ret = realm_destroy_protected(realm, addr, &next_addr);
> +
> + if (WARN_ON(ret))
> + break;
> + }
> +}
> +
> +static void realm_unmap_range(struct kvm *kvm,
> + unsigned long start,
> + unsigned long end,
> + bool unmap_private)
> +{
> + realm_unmap_range_shared(kvm, RME_RTT_MAX_LEVEL - 1, start, end);
> + if (unmap_private)
> + realm_unmap_range_private(kvm, start, end);
> +}
> +
> u32 kvm_realm_ipa_limit(void)
> {
> return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
> @@ -190,6 +341,30 @@ static int realm_rtt_destroy(struct realm *realm, unsigned long addr,
> return ret;
> }
>
> +static int realm_create_rtt_levels(struct realm *realm,
> + unsigned long ipa,
> + int level,
> + int max_level,
> + struct kvm_mmu_memory_cache *mc)
> +{
> + if (WARN_ON(level == max_level))
> + return 0;
> +
> + while (level++ < max_level) {
> + phys_addr_t rtt = alloc_delegated_page(realm, mc);
> +
> + if (rtt == PHYS_ADDR_MAX)
> + return -ENOMEM;
> +
> + if (realm_rtt_create(realm, ipa, level, rtt)) {
> + free_delegated_page(realm, rtt);
> + return -ENXIO;
> + }
> + }
> +
> + return 0;
> +}
> +
> static int realm_tear_down_rtt_level(struct realm *realm, int level,
> unsigned long start, unsigned long end)
> {
> @@ -265,6 +440,68 @@ static int realm_tear_down_rtt_range(struct realm *realm,
> start, end);
> }
>
> +/*
> + * Returns 0 on successful fold, a negative value on error, a positive value if
> + * we were not able to fold all tables at this level.
> + */
> +static int realm_fold_rtt_level(struct realm *realm, int level,
> + unsigned long start, unsigned long end)
> +{
> + int not_folded = 0;
> + ssize_t map_size;
> + unsigned long addr, next_addr;
> +
> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
> + return -EINVAL;
> +
> + map_size = rme_rtt_level_mapsize(level - 1);
> +
> + for (addr = start; addr < end; addr = next_addr) {
> + phys_addr_t rtt_granule;
> + int ret;
> + unsigned long align_addr = ALIGN(addr, map_size);
> +
> + next_addr = ALIGN(addr + 1, map_size);
> +
> + ret = realm_rtt_fold(realm, align_addr, level, &rtt_granule);
> +
> + switch (RMI_RETURN_STATUS(ret)) {
> + case RMI_SUCCESS:
> + if (!WARN_ON(rmi_granule_undelegate(rtt_granule)))
> + free_page((unsigned long)phys_to_virt(rtt_granule));
> + break;
> + case RMI_ERROR_RTT:
> + if (level == RME_RTT_MAX_LEVEL ||
> + RMI_RETURN_INDEX(ret) < level) {
> + not_folded++;
> + break;
> + }
> + /* Recurse a level deeper */
> + ret = realm_fold_rtt_level(realm,
> + level + 1,
> + addr,
> + next_addr);
> + if (ret < 0)
> + return ret;
> + else if (ret == 0)
> + /* Try again at this level */
> + next_addr = addr;
> + break;
> + default:
> + return -ENXIO;
> + }
> + }
> +
> + return not_folded;
> +}
> +
> +static int realm_fold_rtt_range(struct realm *realm,
> + unsigned long start, unsigned long end)
> +{
> + return realm_fold_rtt_level(realm, get_start_level(realm) + 1,
> + start, end);
> +}
> +
> static void ensure_spare_page(struct realm *realm)
> {
> phys_addr_t tmp_rtt;
> @@ -295,6 +532,147 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32 ia_bits)
> WARN_ON(realm_tear_down_rtt_range(realm, 0, (1UL << ia_bits)));
> }
>
> +void kvm_realm_unmap_range(struct kvm *kvm, unsigned long ipa, u64 size,
> + bool unmap_private)
> +{
> + unsigned long end = ipa + size;
> + struct realm *realm = &kvm->arch.realm;
> +
> + end = min(BIT(realm->ia_bits - 1), end);
> +
> + ensure_spare_page(realm);
> +
> + realm_unmap_range(kvm, ipa, end, unmap_private);
> +
> + realm_fold_rtt_range(realm, ipa, end);
> +}
> +
> +static int find_map_level(struct realm *realm,
> + unsigned long start,
> + unsigned long end)
> +{
> + int level = RME_RTT_MAX_LEVEL;
> +
> + while (level > get_start_level(realm)) {
> + unsigned long map_size = rme_rtt_level_mapsize(level - 1);
> +
> + if (!IS_ALIGNED(start, map_size) ||
> + (start + map_size) > end)
> + break;
> +
> + level--;
> + }
> +
> + return level;
> +}
> +
> +int realm_set_ipa_state(struct kvm_vcpu *vcpu,
> + unsigned long start,
> + unsigned long end,
> + unsigned long ripas)
> +{
> + struct kvm *kvm = vcpu->kvm;
> + struct realm *realm = &kvm->arch.realm;
> + struct realm_rec *rec = &vcpu->arch.rec;
> + phys_addr_t rd_phys = virt_to_phys(realm->rd);
> + phys_addr_t rec_phys = virt_to_phys(rec->rec_page);
> + struct kvm_mmu_memory_cache *memcache = &vcpu->arch.mmu_page_cache;
> + unsigned long ipa = start;
> + int ret = 0;
> +
> + while (ipa < end) {
> + unsigned long next;
> +
> + ret = rmi_rtt_set_ripas(rd_phys, rec_phys, ipa, end, &next);
> +
> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
> + int walk_level = RMI_RETURN_INDEX(ret);
> + int level = find_map_level(realm, ipa, end);
> +
> + if (walk_level < level) {
> + ret = realm_create_rtt_levels(realm, ipa,
> + walk_level,
> + level,
> + memcache);
> + if (!ret)
> + continue;
> + } else {
> + ret = -EINVAL;
> + }
> +
> + break;
> + } else if (RMI_RETURN_STATUS(ret) != RMI_SUCCESS) {
> + WARN(1, "Unexpected error in %s: %#x\n", __func__,
> + ret);
> + ret = -EINVAL;
> + break;
> + }
> + ipa = next;
> + }
> +
> + if (ripas == RMI_EMPTY && ipa != start)
> + kvm_realm_unmap_range(kvm, start, ipa - start, true);
> +
> + return ret;
> +}
> +
> +static int realm_init_ipa_state(struct realm *realm,
> + unsigned long ipa,
> + unsigned long end)
> +{
> + phys_addr_t rd_phys = virt_to_phys(realm->rd);
> + int ret;
> +
> + while (ipa < end) {
> + unsigned long next;
> +
> + ret = rmi_rtt_init_ripas(rd_phys, ipa, end, &next);
> +
> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
> + int err_level = RMI_RETURN_INDEX(ret);
> + int level = find_map_level(realm, ipa, end);
> +
> + if (WARN_ON(err_level >= level))
> + return -ENXIO;
> +
> + ret = realm_create_rtt_levels(realm, ipa,
> + err_level,
> + level, NULL);
> + if (ret)
> + return ret;
> + /* Retry with the RTT levels in place */
> + continue;
> + } else if (WARN_ON(ret)) {
> + return -ENXIO;
> + }
> +
> + ipa = next;
> + }
> +
> + return 0;
> +}
> +
> +static int kvm_init_ipa_range_realm(struct kvm *kvm,
> + struct kvm_cap_arm_rme_init_ipa_args *args)
> +{
> + int ret = 0;
> + gpa_t addr, end;
> + struct realm *realm = &kvm->arch.realm;
> +
> + addr = args->init_ipa_base;
> + end = addr + args->init_ipa_size;
> +
> + if (end < addr)
> + return -EINVAL;
> +
> + if (kvm_realm_state(kvm) != REALM_STATE_NEW)
> + return -EINVAL;
> +
> + ret = realm_init_ipa_state(realm, addr, end);
> +
> + return ret;
> +}
> +
> /* Protects access to rme_vmid_bitmap */
> static DEFINE_SPINLOCK(rme_vmid_lock);
> static unsigned long *rme_vmid_bitmap;
> @@ -418,6 +796,18 @@ int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
> case KVM_CAP_ARM_RME_CREATE_RD:
> r = kvm_create_realm(kvm);
> break;
> + case KVM_CAP_ARM_RME_INIT_IPA_REALM: {
> + struct kvm_cap_arm_rme_init_ipa_args args;
> + void __user *argp = u64_to_user_ptr(cap->args[1]);
> +
> + if (copy_from_user(&args, argp, sizeof(args))) {
> + r = -EFAULT;
> + break;
> + }
> +
> + r = kvm_init_ipa_range_realm(kvm, &args);
> + break;
> + }
> default:
> r = -EINVAL;
> break;
> --
> 2.34.1
>
Hi,
On Fri, Apr 12, 2024 at 9:44 AM Steven Price <[email protected]> wrote:
>
> At runtime if the realm guest accesses memory which hasn't yet been
> mapped then KVM needs to either populate the region or fault the guest.
>
> For memory in the lower (protected) region of IPA a fresh page is
> provided to the RMM which will zero the contents. For memory in the
> upper (shared) region of IPA, the memory from the memslot is mapped
> into the realm VM non secure.
>
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/include/asm/kvm_emulate.h | 10 ++
> arch/arm64/include/asm/kvm_rme.h | 10 ++
> arch/arm64/kvm/mmu.c | 119 +++++++++++++++-
> arch/arm64/kvm/rme.c | 199 ++++++++++++++++++++++++---
> 4 files changed, 316 insertions(+), 22 deletions(-)
>
> diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
> index 2209a7c6267f..d40d998d9be2 100644
> --- a/arch/arm64/include/asm/kvm_emulate.h
> +++ b/arch/arm64/include/asm/kvm_emulate.h
> @@ -629,6 +629,16 @@ static inline bool kvm_realm_is_created(struct kvm *kvm)
> return kvm_is_realm(kvm) && kvm_realm_state(kvm) != REALM_STATE_NONE;
> }
>
> +static inline gpa_t kvm_gpa_stolen_bits(struct kvm *kvm)
> +{
> + if (kvm_is_realm(kvm)) {
> + struct realm *realm = &kvm->arch.realm;
> +
> + return BIT(realm->ia_bits - 1);
> + }
> + return 0;
> +}
> +
> static inline bool vcpu_is_rec(struct kvm_vcpu *vcpu)
> {
> if (static_branch_unlikely(&kvm_rme_is_available))
> diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
> index 749f2eb97bd4..48c7766fadeb 100644
> --- a/arch/arm64/include/asm/kvm_rme.h
> +++ b/arch/arm64/include/asm/kvm_rme.h
> @@ -103,6 +103,16 @@ void kvm_realm_unmap_range(struct kvm *kvm,
> unsigned long ipa,
> u64 size,
> bool unmap_private);
> +int realm_map_protected(struct realm *realm,
> + unsigned long base_ipa,
> + struct page *dst_page,
> + unsigned long map_size,
> + struct kvm_mmu_memory_cache *memcache);
> +int realm_map_non_secure(struct realm *realm,
> + unsigned long ipa,
> + struct page *page,
> + unsigned long map_size,
> + struct kvm_mmu_memory_cache *memcache);
> int realm_set_ipa_state(struct kvm_vcpu *vcpu,
> unsigned long addr, unsigned long end,
> unsigned long ripas);
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index 8a7b5449697f..50a49e4e2020 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -325,8 +325,13 @@ static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64
>
> lockdep_assert_held_write(&kvm->mmu_lock);
> WARN_ON(size & ~PAGE_MASK);
> - WARN_ON(stage2_apply_range(mmu, start, end, kvm_pgtable_stage2_unmap,
> - may_block));
> +
> + if (kvm_is_realm(kvm))
> + kvm_realm_unmap_range(kvm, start, size, !only_shared);
> + else
> + WARN_ON(stage2_apply_range(mmu, start, end,
> + kvm_pgtable_stage2_unmap,
> + may_block));
> }
>
> static void unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64 size)
> @@ -340,7 +345,11 @@ static void stage2_flush_memslot(struct kvm *kvm,
> phys_addr_t addr = memslot->base_gfn << PAGE_SHIFT;
> phys_addr_t end = addr + PAGE_SIZE * memslot->npages;
>
> - stage2_apply_range_resched(&kvm->arch.mmu, addr, end, kvm_pgtable_stage2_flush);
> + if (kvm_is_realm(kvm))
> + kvm_realm_unmap_range(kvm, addr, end - addr, false);
> + else
> + stage2_apply_range_resched(&kvm->arch.mmu, addr, end,
> + kvm_pgtable_stage2_flush);
> }
>
> /**
> @@ -997,6 +1006,10 @@ void stage2_unmap_vm(struct kvm *kvm)
> struct kvm_memory_slot *memslot;
> int idx, bkt;
>
> + /* For realms this is handled by the RMM so nothing to do here */
> + if (kvm_is_realm(kvm))
> + return;
> +
> idx = srcu_read_lock(&kvm->srcu);
> mmap_read_lock(current->mm);
> write_lock(&kvm->mmu_lock);
> @@ -1020,6 +1033,7 @@ void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu)
> if (kvm_is_realm(kvm) &&
> (kvm_realm_state(kvm) != REALM_STATE_DEAD &&
> kvm_realm_state(kvm) != REALM_STATE_NONE)) {
> + unmap_stage2_range(mmu, 0, (~0ULL) & PAGE_MASK);
> write_unlock(&kvm->mmu_lock);
> kvm_realm_destroy_rtts(kvm, pgt->ia_bits);
> return;
> @@ -1383,6 +1397,69 @@ static bool kvm_vma_mte_allowed(struct vm_area_struct *vma)
> return vma->vm_flags & VM_MTE_ALLOWED;
> }
>
> +static int realm_map_ipa(struct kvm *kvm, phys_addr_t ipa,
> + kvm_pfn_t pfn, unsigned long map_size,
> + enum kvm_pgtable_prot prot,
> + struct kvm_mmu_memory_cache *memcache)
> +{
> + struct realm *realm = &kvm->arch.realm;
> + struct page *page = pfn_to_page(pfn);
> +
> + if (WARN_ON(!(prot & KVM_PGTABLE_PROT_W)))
> + return -EFAULT;
> +
> + if (!realm_is_addr_protected(realm, ipa))
> + return realm_map_non_secure(realm, ipa, page, map_size,
> + memcache);
> +
> + return realm_map_protected(realm, ipa, page, map_size, memcache);
> +}
> +
> +static int private_memslot_fault(struct kvm_vcpu *vcpu,
> + phys_addr_t fault_ipa,
> + struct kvm_memory_slot *memslot)
> +{
> + struct kvm *kvm = vcpu->kvm;
> + gpa_t gpa_stolen_mask = kvm_gpa_stolen_bits(kvm);
> + gfn_t gfn = (fault_ipa & ~gpa_stolen_mask) >> PAGE_SHIFT;
> + bool is_priv_gfn = !((fault_ipa & gpa_stolen_mask) == gpa_stolen_mask);
> + bool priv_exists = kvm_mem_is_private(kvm, gfn);
> + struct kvm_mmu_memory_cache *memcache = &vcpu->arch.mmu_page_cache;
> + int order;
> + kvm_pfn_t pfn;
> + int ret;
> +
> + if (priv_exists != is_priv_gfn) {
> + kvm_prepare_memory_fault_exit(vcpu,
> + fault_ipa & ~gpa_stolen_mask,
> + PAGE_SIZE,
> + kvm_is_write_fault(vcpu),
> + false, is_priv_gfn);
> +
> + return 0;
> + }
> +
> + if (!is_priv_gfn) {
> + /* Not a private mapping, handling normally */
> + return -EAGAIN;
> + }
> +
> + if (kvm_gmem_get_pfn(kvm, memslot, gfn, &pfn, &order))
> + return 1; /* Retry */
You don't need to pass a variable to hold the order if you don't need
it. You can pass NULL.
I am also confused about the return, why do you return 1 regardless of
the reason kvm_gmem_get_pfn() fails?
> + ret = kvm_mmu_topup_memory_cache(memcache,
> + kvm_mmu_cache_min_pages(vcpu->arch.hw_mmu));
> + if (ret)
> + return ret;
If this fails you should release the page you got earlier (e.g.,
kvm_release_pfn_clean()), or you could move it before
kvm_gmem_get_pfn().
> + /* FIXME: Should be able to use bigger than PAGE_SIZE mappings */
> + ret = realm_map_ipa(kvm, fault_ipa, pfn, PAGE_SIZE, KVM_PGTABLE_PROT_W,
> + memcache);
> + if (!ret)
> + return 1; /* Handled */
Should also release the page if it fails. Speaking of which,
where/when do you eventually release the page?
Cheers,
/fuad
> + return ret;
> +}
> +
> static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
> struct kvm_memory_slot *memslot, unsigned long hva,
> bool fault_is_perm)
> @@ -1402,10 +1479,19 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
> long vma_pagesize, fault_granule;
> enum kvm_pgtable_prot prot = KVM_PGTABLE_PROT_R;
> struct kvm_pgtable *pgt;
> + gpa_t gpa_stolen_mask = kvm_gpa_stolen_bits(vcpu->kvm);
>
> if (fault_is_perm)
> fault_granule = kvm_vcpu_trap_get_perm_fault_granule(vcpu);
> write_fault = kvm_is_write_fault(vcpu);
> +
> + /*
> + * Realms cannot map protected pages read-only
> + * FIXME: It should be possible to map unprotected pages read-only
> + */
> + if (vcpu_is_rec(vcpu))
> + write_fault = true;
> +
> exec_fault = kvm_vcpu_trap_is_exec_fault(vcpu);
> VM_BUG_ON(write_fault && exec_fault);
>
> @@ -1478,7 +1564,7 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
> if (vma_pagesize == PMD_SIZE || vma_pagesize == PUD_SIZE)
> fault_ipa &= ~(vma_pagesize - 1);
>
> - gfn = fault_ipa >> PAGE_SHIFT;
> + gfn = (fault_ipa & ~gpa_stolen_mask) >> PAGE_SHIFT;
> mte_allowed = kvm_vma_mte_allowed(vma);
>
> vfio_allow_any_uc = vma->vm_flags & VM_ALLOW_ANY_UNCACHED;
> @@ -1538,7 +1624,8 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
> * If we are not forced to use page mapping, check if we are
> * backed by a THP and thus use block mapping if possible.
> */
> - if (vma_pagesize == PAGE_SIZE && !(force_pte || device)) {
> + /* FIXME: We shouldn't need to disable this for realms */
> + if (vma_pagesize == PAGE_SIZE && !(force_pte || device || kvm_is_realm(kvm))) {
> if (fault_is_perm && fault_granule > PAGE_SIZE)
> vma_pagesize = fault_granule;
> else
> @@ -1584,6 +1671,9 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
> */
> if (fault_is_perm && vma_pagesize == fault_granule)
> ret = kvm_pgtable_stage2_relax_perms(pgt, fault_ipa, prot);
> + else if (kvm_is_realm(kvm))
> + ret = realm_map_ipa(kvm, fault_ipa, pfn, vma_pagesize,
> + prot, memcache);
> else
> ret = kvm_pgtable_stage2_map(pgt, fault_ipa, vma_pagesize,
> __pfn_to_phys(pfn), prot,
> @@ -1638,6 +1728,7 @@ int kvm_handle_guest_abort(struct kvm_vcpu *vcpu)
> struct kvm_memory_slot *memslot;
> unsigned long hva;
> bool is_iabt, write_fault, writable;
> + gpa_t gpa_stolen_mask = kvm_gpa_stolen_bits(vcpu->kvm);
> gfn_t gfn;
> int ret, idx;
>
> @@ -1693,8 +1784,15 @@ int kvm_handle_guest_abort(struct kvm_vcpu *vcpu)
>
> idx = srcu_read_lock(&vcpu->kvm->srcu);
>
> - gfn = fault_ipa >> PAGE_SHIFT;
> + gfn = (fault_ipa & ~gpa_stolen_mask) >> PAGE_SHIFT;
> memslot = gfn_to_memslot(vcpu->kvm, gfn);
> +
> + if (kvm_slot_can_be_private(memslot)) {
> + ret = private_memslot_fault(vcpu, fault_ipa, memslot);
> + if (ret != -EAGAIN)
> + goto out;
> + }
> +
> hva = gfn_to_hva_memslot_prot(memslot, gfn, &writable);
> write_fault = kvm_is_write_fault(vcpu);
> if (kvm_is_error_hva(hva) || (write_fault && !writable)) {
> @@ -1738,6 +1836,7 @@ int kvm_handle_guest_abort(struct kvm_vcpu *vcpu)
> * of the page size.
> */
> fault_ipa |= kvm_vcpu_get_hfar(vcpu) & ((1 << 12) - 1);
> + fault_ipa &= ~gpa_stolen_mask;
> ret = io_mem_abort(vcpu, fault_ipa);
> goto out_unlock;
> }
> @@ -1819,6 +1918,10 @@ bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
> if (!kvm->arch.mmu.pgt)
> return false;
>
> + /* We don't support aging for Realms */
> + if (kvm_is_realm(kvm))
> + return true;
> +
> return kvm_pgtable_stage2_test_clear_young(kvm->arch.mmu.pgt,
> range->start << PAGE_SHIFT,
> size, true);
> @@ -1831,6 +1934,10 @@ bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
> if (!kvm->arch.mmu.pgt)
> return false;
>
> + /* We don't support aging for Realms */
> + if (kvm_is_realm(kvm))
> + return true;
> +
> return kvm_pgtable_stage2_test_clear_young(kvm->arch.mmu.pgt,
> range->start << PAGE_SHIFT,
> size, false);
> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
> index 4aab507f896e..72f6f5f542c4 100644
> --- a/arch/arm64/kvm/rme.c
> +++ b/arch/arm64/kvm/rme.c
> @@ -606,6 +606,170 @@ static int fold_rtt(struct realm *realm, unsigned long addr, int level)
> return 0;
> }
>
> +int realm_map_protected(struct realm *realm,
> + unsigned long base_ipa,
> + struct page *dst_page,
> + unsigned long map_size,
> + struct kvm_mmu_memory_cache *memcache)
> +{
> + phys_addr_t dst_phys = page_to_phys(dst_page);
> + phys_addr_t rd = virt_to_phys(realm->rd);
> + unsigned long phys = dst_phys;
> + unsigned long ipa = base_ipa;
> + unsigned long size;
> + int map_level;
> + int ret = 0;
> +
> + if (WARN_ON(!IS_ALIGNED(ipa, map_size)))
> + return -EINVAL;
> +
> + switch (map_size) {
> + case PAGE_SIZE:
> + map_level = 3;
> + break;
> + case RME_L2_BLOCK_SIZE:
> + map_level = 2;
> + break;
> + default:
> + return -EINVAL;
> + }
> +
> + if (map_level < RME_RTT_MAX_LEVEL) {
> + /*
> + * A temporary RTT is needed during the map, precreate it,
> + * however if there is an error (e.g. missing parent tables)
> + * this will be handled below.
> + */
> + realm_create_rtt_levels(realm, ipa, map_level,
> + RME_RTT_MAX_LEVEL, memcache);
> + }
> +
> + for (size = 0; size < map_size; size += PAGE_SIZE) {
> + if (rmi_granule_delegate(phys)) {
> + struct rtt_entry rtt;
> +
> + /*
> + * It's possible we raced with another VCPU on the same
> + * fault. If the entry exists and matches then exit
> + * early and assume the other VCPU will handle the
> + * mapping.
> + */
> + if (rmi_rtt_read_entry(rd, ipa, RME_RTT_MAX_LEVEL, &rtt))
> + goto err;
> +
> + // FIXME: For a block mapping this could race at level
> + // 2 or 3...
> + if (WARN_ON((rtt.walk_level != RME_RTT_MAX_LEVEL ||
> + rtt.state != RMI_ASSIGNED ||
> + rtt.desc != phys))) {
> + goto err;
> + }
> +
> + return 0;
> + }
> +
> + ret = rmi_data_create_unknown(rd, phys, ipa);
> +
> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
> + /* Create missing RTTs and retry */
> + int level = RMI_RETURN_INDEX(ret);
> +
> + ret = realm_create_rtt_levels(realm, ipa, level,
> + RME_RTT_MAX_LEVEL,
> + memcache);
> + WARN_ON(ret);
> + if (ret)
> + goto err_undelegate;
> +
> + ret = rmi_data_create_unknown(rd, phys, ipa);
> + }
> + WARN_ON(ret);
> +
> + if (ret)
> + goto err_undelegate;
> +
> + phys += PAGE_SIZE;
> + ipa += PAGE_SIZE;
> + }
> +
> + if (map_size == RME_L2_BLOCK_SIZE)
> + ret = fold_rtt(realm, base_ipa, map_level);
> + if (WARN_ON(ret))
> + goto err;
> +
> + return 0;
> +
> +err_undelegate:
> + if (WARN_ON(rmi_granule_undelegate(phys))) {
> + /* Page can't be returned to NS world so is lost */
> + get_page(phys_to_page(phys));
> + }
> +err:
> + while (size > 0) {
> + unsigned long data, top;
> +
> + phys -= PAGE_SIZE;
> + size -= PAGE_SIZE;
> + ipa -= PAGE_SIZE;
> +
> + WARN_ON(rmi_data_destroy(rd, ipa, &data, &top));
> +
> + if (WARN_ON(rmi_granule_undelegate(phys))) {
> + /* Page can't be returned to NS world so is lost */
> + get_page(phys_to_page(phys));
> + }
> + }
> + return -ENXIO;
> +}
> +
> +int realm_map_non_secure(struct realm *realm,
> + unsigned long ipa,
> + struct page *page,
> + unsigned long map_size,
> + struct kvm_mmu_memory_cache *memcache)
> +{
> + phys_addr_t rd = virt_to_phys(realm->rd);
> + int map_level;
> + int ret = 0;
> + unsigned long desc = page_to_phys(page) |
> + PTE_S2_MEMATTR(MT_S2_FWB_NORMAL) |
> + /* FIXME: Read+Write permissions for now */
> + (3 << 6) |
> + PTE_SHARED;
> +
> + if (WARN_ON(!IS_ALIGNED(ipa, map_size)))
> + return -EINVAL;
> +
> + switch (map_size) {
> + case PAGE_SIZE:
> + map_level = 3;
> + break;
> + case RME_L2_BLOCK_SIZE:
> + map_level = 2;
> + break;
> + default:
> + return -EINVAL;
> + }
> +
> + ret = rmi_rtt_map_unprotected(rd, ipa, map_level, desc);
> +
> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
> + /* Create missing RTTs and retry */
> + int level = RMI_RETURN_INDEX(ret);
> +
> + ret = realm_create_rtt_levels(realm, ipa, level, map_level,
> + memcache);
> + if (WARN_ON(ret))
> + return -ENXIO;
> +
> + ret = rmi_rtt_map_unprotected(rd, ipa, map_level, desc);
> + }
> + if (WARN_ON(ret))
> + return -ENXIO;
> +
> + return 0;
> +}
> +
> static int populate_par_region(struct kvm *kvm,
> phys_addr_t ipa_base,
> phys_addr_t ipa_end,
> @@ -617,7 +781,6 @@ static int populate_par_region(struct kvm *kvm,
> int idx;
> phys_addr_t ipa;
> int ret = 0;
> - struct page *tmp_page;
> unsigned long data_flags = 0;
>
> base_gfn = gpa_to_gfn(ipa_base);
> @@ -639,9 +802,8 @@ static int populate_par_region(struct kvm *kvm,
> goto out;
> }
>
> - tmp_page = alloc_page(GFP_KERNEL);
> - if (!tmp_page) {
> - ret = -ENOMEM;
> + if (!kvm_slot_can_be_private(memslot)) {
> + ret = -EINVAL;
> goto out;
> }
>
> @@ -714,31 +876,36 @@ static int populate_par_region(struct kvm *kvm,
> for (offset = 0; offset < map_size && !ret;
> offset += PAGE_SIZE, page++) {
> phys_addr_t page_ipa = ipa + offset;
> + kvm_pfn_t priv_pfn;
> + int order;
>
> - ret = realm_create_protected_data_page(realm, page_ipa,
> - page, tmp_page,
> - data_flags);
> + ret = kvm_gmem_get_pfn(kvm, memslot,
> + page_ipa >> PAGE_SHIFT,
> + &priv_pfn, &order);
> + if (ret)
> + break;
> +
> + ret = realm_create_protected_data_page(
> + realm, page_ipa,
> + pfn_to_page(priv_pfn),
> + page, data_flags);
> }
> +
> + kvm_release_pfn_clean(pfn);
> +
> if (ret)
> - goto err_release_pfn;
> + break;
>
> if (level == 2) {
> ret = fold_rtt(realm, ipa, level);
> if (ret)
> - goto err_release_pfn;
> + break;
> }
>
> ipa += map_size;
> - kvm_release_pfn_dirty(pfn);
> -err_release_pfn:
> - if (ret) {
> - kvm_release_pfn_clean(pfn);
> - break;
> - }
> }
>
> mmap_read_unlock(current->mm);
> - __free_page(tmp_page);
>
> out:
> srcu_read_unlock(&kvm->srcu, idx);
> --
> 2.34.1
>
Hi,
On Fri, Apr 12, 2024 at 9:44 AM Steven Price <[email protected]> wrote:
>
> Select KVM_GENERIC_PRIVATE_MEM and provide the necessary support
> functions.
>
> Signed-off-by: Steven Price <[email protected]>
> ---
> arch/arm64/include/asm/kvm_host.h | 4 ++++
> arch/arm64/kvm/Kconfig | 1 +
> arch/arm64/kvm/arm.c | 5 +++++
> arch/arm64/kvm/mmu.c | 19 +++++++++++++++++++
> 4 files changed, 29 insertions(+)
>
> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> index 902923402f6e..93de7f5009fe 100644
> --- a/arch/arm64/include/asm/kvm_host.h
> +++ b/arch/arm64/include/asm/kvm_host.h
> @@ -1259,6 +1259,10 @@ static inline bool kvm_vm_is_protected(struct kvm *kvm)
> return false;
> }
>
> +#ifdef CONFIG_KVM_PRIVATE_MEM
> +bool kvm_arch_has_private_mem(struct kvm *kvm);
> +#endif
> +
I think it might be better to define kvm_arch_has_private_mem() for
both cases, whether KVM_PRIVATE_MEM is enabled or not, similar to the
way it's defined in arch/x86/include/asm/kvm_host.h
> int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
> bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
>
> diff --git a/arch/arm64/kvm/Kconfig b/arch/arm64/kvm/Kconfig
> index 58f09370d17e..8da57e74c86a 100644
> --- a/arch/arm64/kvm/Kconfig
> +++ b/arch/arm64/kvm/Kconfig
> @@ -37,6 +37,7 @@ menuconfig KVM
> select HAVE_KVM_VCPU_RUN_PID_CHANGE
> select SCHED_INFO
> select GUEST_PERF_EVENTS if PERF_EVENTS
> + select KVM_GENERIC_PRIVATE_MEM
I don't think this should be enabled by default, but should depend on
whether RME is configured. That said, I can't find the config option
for RME...
> help
> Support hosting virtualized guest machines.
>
> diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
> index 2dd014d3c366..a66d0a6eb4fa 100644
> --- a/arch/arm64/kvm/arm.c
> +++ b/arch/arm64/kvm/arm.c
> @@ -89,6 +89,11 @@ int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
> return kvm_vcpu_exiting_guest_mode(vcpu) == IN_GUEST_MODE;
> }
>
> +bool kvm_arch_has_private_mem(struct kvm *kvm)
> +{
> + return kvm_is_realm(kvm);
> +}
> +
Related to my earlier comment on kvm_arch_has_private_mem(), and
considering how often this function is called, wouldn't it be better
to define this in a way similar to arch/x86/include/asm/kvm_host.h ?
> int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
> struct kvm_enable_cap *cap)
> {
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index 48c957e21c83..808bceebad4d 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -2171,6 +2171,25 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
> return ret;
> }
The following two functions should be gated by
#ifdef CONFIG_KVM_GENERIC_MEMORY_ATTRIBUTES
> +bool kvm_arch_pre_set_memory_attributes(struct kvm *kvm,
> + struct kvm_gfn_range *range)
> +{
> + WARN_ON_ONCE(!kvm_arch_has_private_mem(kvm));
> + return false;
> +}
> +
> +bool kvm_arch_post_set_memory_attributes(struct kvm *kvm,
> + struct kvm_gfn_range *range)
> +{
> + WARN_ON_ONCE(!kvm_arch_has_private_mem(kvm));
I think this should return here, not just warn.
Cheers,
/fuad
> +
> + if (range->arg.attributes & KVM_MEMORY_ATTRIBUTE_PRIVATE)
> + range->only_shared = true;
> + kvm_unmap_gfn_range(kvm, range);
> +
> + return false;
> +}
> +
> void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
> {
> }
> --
> 2.34.1
>
On 25/04/2024 10:48, Fuad Tabba wrote:
> Hi,
>
> On Fri, Apr 12, 2024 at 9:43 AM Steven Price <[email protected]> wrote:
>>
>> From: Sean Christopherson <[email protected]>
>>
>> Add flags to "struct kvm_gfn_range" to let notifier events target only
>> shared and only private mappings, and write up the existing mmu_notifier
>> events to be shared-only (private memory is never associated with a
>> userspace virtual address, i.e. can't be reached via mmu_notifiers).
>>
>> Add two flags so that KVM can handle the three possibilities (shared,
>> private, and shared+private) without needing something like a tri-state
>> enum.
>>
>> Link: https://lore.kernel.org/all/[email protected]
>> Signed-off-by: Sean Christopherson <[email protected]>
>> Signed-off-by: Steven Price <[email protected]>
>> ---
>> include/linux/kvm_host.h | 2 ++
>> virt/kvm/kvm_main.c | 7 +++++++
>> 2 files changed, 9 insertions(+)
>>
>> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
>> index 48f31dcd318a..c7581360fd88 100644
>> --- a/include/linux/kvm_host.h
>> +++ b/include/linux/kvm_host.h
>> @@ -268,6 +268,8 @@ struct kvm_gfn_range {
>> gfn_t start;
>> gfn_t end;
>> union kvm_mmu_notifier_arg arg;
>> + bool only_private;
>> + bool only_shared;
>> bool may_block;
>> };
>> bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
>> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
>> index fb49c2a60200..3486ceef6f4e 100644
>> --- a/virt/kvm/kvm_main.c
>> +++ b/virt/kvm/kvm_main.c
>> @@ -633,6 +633,13 @@ static __always_inline kvm_mn_ret_t __kvm_handle_hva_range(struct kvm *kvm,
>> * the second or later invocation of the handler).
>> */
>> gfn_range.arg = range->arg;
>> +
>> + /*
>> + * HVA-based notifications aren't relevant to private
>> + * mappings as they don't have a userspace mapping.
>> + */
>> + gfn_range.only_private = false;
>> + gfn_range.only_shared = true;
>> gfn_range.may_block = range->may_block;
>
> I'd discussed this with Sean when he posted this earlier. Having two
> booleans to encode three valid states could be confusing. In response,
> Sean suggested using an enum instead:
> https://lore.kernel.org/all/[email protected]/
That would work fine too! Unless I've missed it Sean hasn't posted an
updated patch. My assumption is that this will get merged (in whatever
form) before the rest of the series as part of that other series. It
shouldn't be too hard to adapt.
Thanks,
Steve
> Cheers,
> /fuad
>
>>
>> /*
>
>
>> --
>> 2.34.1
>>
>
On Thu, Apr 25, 2024, Steven Price wrote:
> On 25/04/2024 10:48, Fuad Tabba wrote:
> > On Fri, Apr 12, 2024 at 9:43 AM Steven Price <[email protected]> wrote:
> >> bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
> >> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> >> index fb49c2a60200..3486ceef6f4e 100644
> >> --- a/virt/kvm/kvm_main.c
> >> +++ b/virt/kvm/kvm_main.c
> >> @@ -633,6 +633,13 @@ static __always_inline kvm_mn_ret_t __kvm_handle_hva_range(struct kvm *kvm,
> >> * the second or later invocation of the handler).
> >> */
> >> gfn_range.arg = range->arg;
> >> +
> >> + /*
> >> + * HVA-based notifications aren't relevant to private
> >> + * mappings as they don't have a userspace mapping.
> >> + */
> >> + gfn_range.only_private = false;
> >> + gfn_range.only_shared = true;
> >> gfn_range.may_block = range->may_block;
> >
> > I'd discussed this with Sean when he posted this earlier. Having two
> > booleans to encode three valid states could be confusing. In response,
> > Sean suggested using an enum instead:
> > https://lore.kernel.org/all/[email protected]/
>
> That would work fine too! Unless I've missed it Sean hasn't posted an
> updated patch. My assumption is that this will get merged (in whatever
> form) before the rest of the series as part of that other series. It
> shouldn't be too hard to adapt.
Yeah, there's no updated patch.
Fuad, if you have a strong preference, I recommend chiming in on the TDX series[*],
as that is the series that's likely going to be the first user, and I don't have
a strong preference on bools versus an enum.
[*] https://lore.kernel.org/all/e324ff5e47e07505648c0092a5370ac9ddd72f0b.1708933498.git.isaku.yamahata@intel.com
On Fri, Apr 12, 2024 at 09:42:43AM +0100, Steven Price wrote:
> +static inline bool realm_is_addr_protected(struct realm *realm,
> + unsigned long addr)
> +{
> + unsigned int ia_bits = realm->ia_bits;
> +
> + return !(addr & ~(BIT(ia_bits - 1) - 1));
Is it enough to return !(addr & BIT(realm->ia_bits - 1))?
> +static void realm_unmap_range_shared(struct kvm *kvm,
> + int level,
> + unsigned long start,
> + unsigned long end)
> +{
> + struct realm *realm = &kvm->arch.realm;
> + unsigned long rd = virt_to_phys(realm->rd);
> + ssize_t map_size = rme_rtt_level_mapsize(level);
> + unsigned long next_addr, addr;
> + unsigned long shared_bit = BIT(realm->ia_bits - 1);
> +
> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
> + return;
> +
> + start |= shared_bit;
> + end |= shared_bit;
> +
> + for (addr = start; addr < end; addr = next_addr) {
> + unsigned long align_addr = ALIGN(addr, map_size);
> + int ret;
> +
> + next_addr = ALIGN(addr + 1, map_size);
> +
> + if (align_addr != addr || next_addr > end) {
> + /* Need to recurse deeper */
> + if (addr < align_addr)
> + next_addr = align_addr;
> + realm_unmap_range_shared(kvm, level + 1, addr,
> + min(next_addr, end));
> + continue;
> + }
> +
> + ret = rmi_rtt_unmap_unprotected(rd, addr, level, &next_addr);
> + switch (RMI_RETURN_STATUS(ret)) {
> + case RMI_SUCCESS:
> + break;
> + case RMI_ERROR_RTT:
> + if (next_addr == addr) {
> + next_addr = ALIGN(addr + 1, map_size);
> + realm_unmap_range_shared(kvm, level + 1, addr,
> + next_addr);
> + }
> + break;
> + default:
> + WARN_ON(1);
In this case we also need to return, because RMM returns with next_addr ==
0, causing an infinite loop. At the moment a VMM can trigger this easily
by creating guest memfd before creating a RD, see below
> + }
> + }
> +}
> +
> +static void realm_unmap_range_private(struct kvm *kvm,
> + unsigned long start,
> + unsigned long end)
> +{
> + struct realm *realm = &kvm->arch.realm;
> + ssize_t map_size = RME_PAGE_SIZE;
> + unsigned long next_addr, addr;
> +
> + for (addr = start; addr < end; addr = next_addr) {
> + int ret;
> +
> + next_addr = ALIGN(addr + 1, map_size);
> +
> + ret = realm_destroy_protected(realm, addr, &next_addr);
> +
> + if (WARN_ON(ret))
> + break;
> + }
> +}
> +
> +static void realm_unmap_range(struct kvm *kvm,
> + unsigned long start,
> + unsigned long end,
> + bool unmap_private)
> +{
Should this check for a valid kvm->arch.realm.rd, or a valid realm state?
I'm not sure what the best place is but none of the RMM calls will succeed
if the RD is NULL, causing some WARNs.
I can trigger this with set_memory_attributes() ioctls before creating a
RD for example.
> + realm_unmap_range_shared(kvm, RME_RTT_MAX_LEVEL - 1, start, end);
> + if (unmap_private)
> + realm_unmap_range_private(kvm, start, end);
> +}
> +
> u32 kvm_realm_ipa_limit(void)
> {
> return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
> @@ -190,6 +341,30 @@ static int realm_rtt_destroy(struct realm *realm, unsigned long addr,
> return ret;
> }
>
> +static int realm_create_rtt_levels(struct realm *realm,
> + unsigned long ipa,
> + int level,
> + int max_level,
> + struct kvm_mmu_memory_cache *mc)
> +{
> + if (WARN_ON(level == max_level))
> + return 0;
> +
> + while (level++ < max_level) {
> + phys_addr_t rtt = alloc_delegated_page(realm, mc);
> +
> + if (rtt == PHYS_ADDR_MAX)
> + return -ENOMEM;
> +
> + if (realm_rtt_create(realm, ipa, level, rtt)) {
> + free_delegated_page(realm, rtt);
> + return -ENXIO;
> + }
> + }
> +
> + return 0;
> +}
> +
> static int realm_tear_down_rtt_level(struct realm *realm, int level,
> unsigned long start, unsigned long end)
> {
> @@ -265,6 +440,68 @@ static int realm_tear_down_rtt_range(struct realm *realm,
> start, end);
> }
>
> +/*
> + * Returns 0 on successful fold, a negative value on error, a positive value if
> + * we were not able to fold all tables at this level.
> + */
> +static int realm_fold_rtt_level(struct realm *realm, int level,
> + unsigned long start, unsigned long end)
> +{
> + int not_folded = 0;
> + ssize_t map_size;
> + unsigned long addr, next_addr;
> +
> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
> + return -EINVAL;
> +
> + map_size = rme_rtt_level_mapsize(level - 1);
> +
> + for (addr = start; addr < end; addr = next_addr) {
> + phys_addr_t rtt_granule;
> + int ret;
> + unsigned long align_addr = ALIGN(addr, map_size);
> +
> + next_addr = ALIGN(addr + 1, map_size);
> +
> + ret = realm_rtt_fold(realm, align_addr, level, &rtt_granule);
> +
> + switch (RMI_RETURN_STATUS(ret)) {
> + case RMI_SUCCESS:
> + if (!WARN_ON(rmi_granule_undelegate(rtt_granule)))
> + free_page((unsigned long)phys_to_virt(rtt_granule));
> + break;
> + case RMI_ERROR_RTT:
> + if (level == RME_RTT_MAX_LEVEL ||
> + RMI_RETURN_INDEX(ret) < level) {
> + not_folded++;
> + break;
> + }
> + /* Recurse a level deeper */
> + ret = realm_fold_rtt_level(realm,
> + level + 1,
> + addr,
> + next_addr);
> + if (ret < 0)
> + return ret;
> + else if (ret == 0)
> + /* Try again at this level */
> + next_addr = addr;
> + break;
> + default:
Maybe this also deserves a WARN() to be consistent with the other RMI
calls
Thanks,
Jean
> + return -ENXIO;
> + }
> + }
> +
> + return not_folded;
> +}
On 01/05/2024 15:27, Jean-Philippe Brucker wrote:
> On Fri, Apr 12, 2024 at 09:42:43AM +0100, Steven Price wrote:
>> +static inline bool realm_is_addr_protected(struct realm *realm,
>> + unsigned long addr)
>> +{
>> + unsigned int ia_bits = realm->ia_bits;
>> +
>> + return !(addr & ~(BIT(ia_bits - 1) - 1));
>
> Is it enough to return !(addr & BIT(realm->ia_bits - 1))?
I thought about that too. But if we are dealing with an IPA
that is > (BIT(realm->ia_bits)), we don't want to be treating
that as a protected address. This could only happen if the Realm
is buggy (or the VMM has tricked it). So the existing check
looks safer.
>
>> +static void realm_unmap_range_shared(struct kvm *kvm,
>> + int level,
>> + unsigned long start,
>> + unsigned long end)
>> +{
>> + struct realm *realm = &kvm->arch.realm;
>> + unsigned long rd = virt_to_phys(realm->rd);
>> + ssize_t map_size = rme_rtt_level_mapsize(level);
>> + unsigned long next_addr, addr;
>> + unsigned long shared_bit = BIT(realm->ia_bits - 1);
>> +
>> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
>> + return;
>> +
>> + start |= shared_bit;
>> + end |= shared_bit;
>> +
>> + for (addr = start; addr < end; addr = next_addr) {
>> + unsigned long align_addr = ALIGN(addr, map_size);
>> + int ret;
>> +
>> + next_addr = ALIGN(addr + 1, map_size);
>> +
>> + if (align_addr != addr || next_addr > end) {
>> + /* Need to recurse deeper */
>> + if (addr < align_addr)
>> + next_addr = align_addr;
>> + realm_unmap_range_shared(kvm, level + 1, addr,
>> + min(next_addr, end));
>> + continue;
>> + }
>> +
>> + ret = rmi_rtt_unmap_unprotected(rd, addr, level, &next_addr);
>> + switch (RMI_RETURN_STATUS(ret)) {
>> + case RMI_SUCCESS:
>> + break;
>> + case RMI_ERROR_RTT:
>> + if (next_addr == addr) {
>> + next_addr = ALIGN(addr + 1, map_size);
>> + realm_unmap_range_shared(kvm, level + 1, addr,
>> + next_addr);
>> + }
>> + break;
>> + default:
>> + WARN_ON(1);
>
> In this case we also need to return, because RMM returns with next_addr ==
> 0, causing an infinite loop. At the moment a VMM can trigger this easily
> by creating guest memfd before creating a RD, see below
Thats a good point. I agree.
>
>> + }
>> + }
>> +}
>> +
>> +static void realm_unmap_range_private(struct kvm *kvm,
>> + unsigned long start,
>> + unsigned long end)
>> +{
>> + struct realm *realm = &kvm->arch.realm;
>> + ssize_t map_size = RME_PAGE_SIZE;
>> + unsigned long next_addr, addr;
>> +
>> + for (addr = start; addr < end; addr = next_addr) {
>> + int ret;
>> +
>> + next_addr = ALIGN(addr + 1, map_size);
>> +
>> + ret = realm_destroy_protected(realm, addr, &next_addr);
>> +
>> + if (WARN_ON(ret))
>> + break;
>> + }
>> +}
>> +
>> +static void realm_unmap_range(struct kvm *kvm,
>> + unsigned long start,
>> + unsigned long end,
>> + bool unmap_private)
>> +{
>
> Should this check for a valid kvm->arch.realm.rd, or a valid realm state?
> I'm not sure what the best place is but none of the RMM calls will succeed
> if the RD is NULL, causing some WARNs.
>
> I can trigger this with set_memory_attributes() ioctls before creating a
> RD for example.
>
True, this could be triggered by a buggy VMM in other ways, and we could
easily gate it on the Realm state >= NEW.
Suzuki
On 19/04/2024 10:34, Suzuki K Poulose wrote:
> On 12/04/2024 09:42, Steven Price wrote:
>> Each page within the protected region of the realm guest can be marked
>> as either RAM or EMPTY. Allow the VMM to control this before the guest
>> has started and provide the equivalent functions to change this (with
>> the guest's approval) at runtime.
>>
>> When transitioning from RIPAS RAM (1) to RIPAS EMPTY (0) the memory is
>> unmapped from the guest and undelegated allowing the memory to be reused
>> by the host. When transitioning to RIPAS RAM the actual population of
>> the leaf RTTs is done later on stage 2 fault, however it may be
>> necessary to allocate additional RTTs to represent the range requested.
>
> minor nit: To give a bit more context:
>
> "however it may be necessary to allocate additional RTTs in order for
> the RMM to track the RIPAS for the requested range".
That is what I meant - but your wording is probably clearer ;)
Technically there's also the case where a RIPAS change will cause a
block mapping to be split which isn't just about tracking RIPAS but also
about breaking up the block mapping for the pages that remain.
>>
>> When freeing a block mapping it is necessary to temporarily unfold the
>> RTT which requires delegating an extra page to the RMM, this page can
>> then be recovered once the contents of the block mapping have been
>> freed. A spare, delegated page (spare_page) is used for this purpose.
>>
>> Signed-off-by: Steven Price <[email protected]>
>> ---
>> arch/arm64/include/asm/kvm_rme.h | 16 ++
>> arch/arm64/kvm/mmu.c | 8 +-
>> arch/arm64/kvm/rme.c | 390 +++++++++++++++++++++++++++++++
>> 3 files changed, 411 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/arm64/include/asm/kvm_rme.h
>> b/arch/arm64/include/asm/kvm_rme.h
>> index 915e76068b00..cc8f81cfc3c0 100644
>> --- a/arch/arm64/include/asm/kvm_rme.h
>> +++ b/arch/arm64/include/asm/kvm_rme.h
>> @@ -96,6 +96,14 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32
>> ia_bits);
>> int kvm_create_rec(struct kvm_vcpu *vcpu);
>> void kvm_destroy_rec(struct kvm_vcpu *vcpu);
>> +void kvm_realm_unmap_range(struct kvm *kvm,
>> + unsigned long ipa,
>> + u64 size,
>> + bool unmap_private);
>> +int realm_set_ipa_state(struct kvm_vcpu *vcpu,
>> + unsigned long addr, unsigned long end,
>> + unsigned long ripas);
>> +
>> #define RME_RTT_BLOCK_LEVEL 2
>> #define RME_RTT_MAX_LEVEL 3
>> @@ -114,4 +122,12 @@ static inline unsigned long
>> rme_rtt_level_mapsize(int level)
>> return (1UL << RME_RTT_LEVEL_SHIFT(level));
>> }
>> +static inline bool realm_is_addr_protected(struct realm *realm,
>> + unsigned long addr)
>> +{
>> + unsigned int ia_bits = realm->ia_bits;
>> +
>> + return !(addr & ~(BIT(ia_bits - 1) - 1));
>> +}
>> +
>> #endif
>> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
>> index 46f0c4e80ace..8a7b5449697f 100644
>> --- a/arch/arm64/kvm/mmu.c
>> +++ b/arch/arm64/kvm/mmu.c
>> @@ -310,6 +310,7 @@ static void invalidate_icache_guest_page(void *va,
>> size_t size)
>> * @start: The intermediate physical base address of the range to unmap
>> * @size: The size of the area to unmap
>> * @may_block: Whether or not we are permitted to block
>> + * @only_shared: If true then protected mappings should not be unmapped
>> *
>> * Clear a range of stage-2 mappings, lowering the various
>> ref-counts. Must
>> * be called while holding mmu_lock (unless for freeing the stage2
>> pgd before
>> @@ -317,7 +318,7 @@ static void invalidate_icache_guest_page(void *va,
>> size_t size)
>> * with things behind our backs.
>> */
>> static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t
>> start, u64 size,
>> - bool may_block)
>> + bool may_block, bool only_shared)
>> {
>> struct kvm *kvm = kvm_s2_mmu_to_kvm(mmu);
>> phys_addr_t end = start + size;
>> @@ -330,7 +331,7 @@ static void __unmap_stage2_range(struct kvm_s2_mmu
>> *mmu, phys_addr_t start, u64
>> static void unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t
>> start, u64 size)
>> {
>> - __unmap_stage2_range(mmu, start, size, true);
>> + __unmap_stage2_range(mmu, start, size, true, false);
>> }
>> static void stage2_flush_memslot(struct kvm *kvm,
>> @@ -1771,7 +1772,8 @@ bool kvm_unmap_gfn_range(struct kvm *kvm, struct
>> kvm_gfn_range *range)
>> __unmap_stage2_range(&kvm->arch.mmu, range->start << PAGE_SHIFT,
>> (range->end - range->start) << PAGE_SHIFT,
>> - range->may_block);
>> + range->may_block,
>> + range->only_shared);
>> return false;
>> }
>> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
>> index 629a095bea61..9e5983c51393 100644
>> --- a/arch/arm64/kvm/rme.c
>> +++ b/arch/arm64/kvm/rme.c
>> @@ -79,6 +79,12 @@ static phys_addr_t __alloc_delegated_page(struct
>> realm *realm,
>> return phys;
>> }
>> +static phys_addr_t alloc_delegated_page(struct realm *realm,
>> + struct kvm_mmu_memory_cache *mc)
>> +{
>> + return __alloc_delegated_page(realm, mc, GFP_KERNEL);
>> +}
>> +
>> static void free_delegated_page(struct realm *realm, phys_addr_t phys)
>> {
>> if (realm->spare_page == PHYS_ADDR_MAX) {
>> @@ -94,6 +100,151 @@ static void free_delegated_page(struct realm
>> *realm, phys_addr_t phys)
>> free_page((unsigned long)phys_to_virt(phys));
>> }
>> +static int realm_rtt_create(struct realm *realm,
>> + unsigned long addr,
>> + int level,
>> + phys_addr_t phys)
>> +{
>> + addr = ALIGN_DOWN(addr, rme_rtt_level_mapsize(level - 1));
>> + return rmi_rtt_create(virt_to_phys(realm->rd), phys, addr, level);
>> +}
>> +
>> +static int realm_rtt_fold(struct realm *realm,
>> + unsigned long addr,
>> + int level,
>> + phys_addr_t *rtt_granule)
>> +{
>> + unsigned long out_rtt;
>> + int ret;
>> +
>> + ret = rmi_rtt_fold(virt_to_phys(realm->rd), addr, level, &out_rtt);
>> +
>> + if (RMI_RETURN_STATUS(ret) == RMI_SUCCESS && rtt_granule)
>> + *rtt_granule = out_rtt;
>> +
>> + return ret;
>> +}
>> +
>> +static int realm_destroy_protected(struct realm *realm,
>> + unsigned long ipa,
>> + unsigned long *next_addr)
>> +{
>> + unsigned long rd = virt_to_phys(realm->rd);
>> + unsigned long addr;
>> + phys_addr_t rtt;
>> + int ret;
>> +
>> +loop:
>> + ret = rmi_data_destroy(rd, ipa, &addr, next_addr);
>> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
>> + if (*next_addr > ipa)
>> + return 0; /* UNASSIGNED */
>> + rtt = alloc_delegated_page(realm, NULL);
>> + if (WARN_ON(rtt == PHYS_ADDR_MAX))
>> + return -1;
>> + /* ASSIGNED - ipa is mapped as a block, so split */
>> + ret = realm_rtt_create(realm, ipa,
>> + RMI_RETURN_INDEX(ret) + 1, rtt);
>
> Could we not go all the way to L3 (rather than 1 level deeper) and try
> again ? That way, we are covered for block mappings at L1 (1G).
I don't think this situation can happen. The spec states that for
RMI_ERROR_RTT either top>ipa (UNASSIGNED, in which case we've bailed out
above) or top==ipa (ASSIGNED) in which case RMI_RETURN_INDEX(ret) must
equal 2.
So in this case realm_rtt_create() will always be called with level=3.
I can simplify this by explicitly passing 3 and adding a WARN_ON() for
RMI_RETURN_INDEX(ret) != 2.
>> + if (WARN_ON(ret)) {
>> + free_delegated_page(realm, rtt);
>> + return -1;
>> + }
>> + /* retry */
>> + goto loop;
>> + } else if (WARN_ON(ret)) {
>> + return -1;
>> + }
>> + ret = rmi_granule_undelegate(addr);
>> +
>> + /*
>> + * If the undelegate fails then something has gone seriously
>> + * wrong: take an extra reference to just leak the page
>> + */
>> + if (WARN_ON(ret))
>> + get_page(phys_to_page(addr));
>> +
>> + return 0;
>> +}
>> +
>> +static void realm_unmap_range_shared(struct kvm *kvm,
>> + int level,
>> + unsigned long start,
>> + unsigned long end)
>> +{
>> + struct realm *realm = &kvm->arch.realm;
>> + unsigned long rd = virt_to_phys(realm->rd);
>> + ssize_t map_size = rme_rtt_level_mapsize(level);
>> + unsigned long next_addr, addr;
>> + unsigned long shared_bit = BIT(realm->ia_bits - 1);
>> +
>> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
>> + return;
>> +
>> + start |= shared_bit;
>> + end |= shared_bit;
>> +
>> + for (addr = start; addr < end; addr = next_addr) {
>> + unsigned long align_addr = ALIGN(addr, map_size);
>> + int ret;
>> +
>> + next_addr = ALIGN(addr + 1, map_size);
>> +
>> + if (align_addr != addr || next_addr > end) {
>> + /* Need to recurse deeper */
>> + if (addr < align_addr)
>> + next_addr = align_addr;
>> + realm_unmap_range_shared(kvm, level + 1, addr,
>> + min(next_addr, end));
>> + continue;
>> + }
>> +
>> + ret = rmi_rtt_unmap_unprotected(rd, addr, level, &next_addr);
>
> minor nit: We could potentially use rmi_rtt_destroy() to tear down
> shared mappings without unmapping them individually, if the range
> is big enough. All such optimisations could come later though.
Apparently so, I do feel the "liveness" concept in the RMM spec is very
confusing. But it looks like an RTT in the unprotected IPA range isn't
considered 'live' if it only contains ASSIGNED_NS entries. The wording
here isn't great though:
An RTT is live if, for any of its entries, either of the following is
true:
• The RTTE state is ASSIGNED
• The RTTE state is TABLE.
Note that an RTT can be non-live, even if one of its entries is live.
This would be the case for example if the RTT corresponds to an
Unprotected IPA range and the state of one of its entries is
ASSIGNED_NS.
Like you say this is an optimisation, so I think we should leave it
until later for now.
>> + switch (RMI_RETURN_STATUS(ret)) {
>> + case RMI_SUCCESS:
>> + break;
>> + case RMI_ERROR_RTT:
>> + if (next_addr == addr) {
>
> At this point we have a block aligned address, but the mapping is
> further deep. Given, start from top to down, we implicitly handle
> the case of block mappings. Not sure if that needs to be in a comment
> here.
>
>> + next_addr = ALIGN(addr + 1, map_size);
>
> Reset to the "actual next" as it was overwritten by the RMI call.
I'll add a comment explaining what's happening here.
>> + realm_unmap_range_shared(kvm, level + 1, addr,
>> + next_addr);
>> + }
>> + break;
>> + default:
>> + WARN_ON(1);
>> + }
>> + }
>> +}
>> +
>> +static void realm_unmap_range_private(struct kvm *kvm,
>> + unsigned long start,
>> + unsigned long end)
>> +{
>> + struct realm *realm = &kvm->arch.realm;
>> + ssize_t map_size = RME_PAGE_SIZE;
>> + unsigned long next_addr, addr;
>> +
>> + for (addr = start; addr < end; addr = next_addr) {
>> + int ret;
>> +
>> + next_addr = ALIGN(addr + 1, map_size);
>> +
>> + ret = realm_destroy_protected(realm, addr, &next_addr);
>> +
>> + if (WARN_ON(ret))
>> + break;
>> + }
>> +}
>> +
>> +static void realm_unmap_range(struct kvm *kvm,
>> + unsigned long start,
>> + unsigned long end,
>> + bool unmap_private)
>> +{
>> + realm_unmap_range_shared(kvm, RME_RTT_MAX_LEVEL - 1, start, end);
>
> minor nit: We already have a helper to find a suitable start level
> (defined below), may be we could use that ? And even do the rtt_destroy
> optimisation for unprotected range.
Good point, that should work.
>> + if (unmap_private)
>> + realm_unmap_range_private(kvm, start, end);
>> +}
>> +
>> u32 kvm_realm_ipa_limit(void)
>> {
>> return u64_get_bits(rmm_feat_reg0, RMI_FEATURE_REGISTER_0_S2SZ);
>> @@ -190,6 +341,30 @@ static int realm_rtt_destroy(struct realm *realm,
>> unsigned long addr,
>> return ret;
>> }
>> +static int realm_create_rtt_levels(struct realm *realm,
>> + unsigned long ipa,
>> + int level,
>> + int max_level,
>> + struct kvm_mmu_memory_cache *mc)
>> +{
>> + if (WARN_ON(level == max_level))
>> + return 0;
>> +
>> + while (level++ < max_level) {
>> + phys_addr_t rtt = alloc_delegated_page(realm, mc);
>> +
>> + if (rtt == PHYS_ADDR_MAX)
>> + return -ENOMEM;
>> +
>> + if (realm_rtt_create(realm, ipa, level, rtt)) {
>> + free_delegated_page(realm, rtt);
>> + return -ENXIO;
>> + }
>> + }
>> +
>> + return 0;
>> +}
>> +
>> static int realm_tear_down_rtt_level(struct realm *realm, int level,
>> unsigned long start, unsigned long end)
>> {
>> @@ -265,6 +440,68 @@ static int realm_tear_down_rtt_range(struct realm
>> *realm,
>> start, end);
>> }
>> +/*
>> + * Returns 0 on successful fold, a negative value on error, a
>> positive value if
>> + * we were not able to fold all tables at this level.
>> + */
>> +static int realm_fold_rtt_level(struct realm *realm, int level,
>> + unsigned long start, unsigned long end)
>> +{
>> + int not_folded = 0;
>> + ssize_t map_size;
>> + unsigned long addr, next_addr;
>> +
>> + if (WARN_ON(level > RME_RTT_MAX_LEVEL))
>> + return -EINVAL;
>> +
>> + map_size = rme_rtt_level_mapsize(level - 1);
>> +
>> + for (addr = start; addr < end; addr = next_addr) {
>> + phys_addr_t rtt_granule;
>> + int ret;
>> + unsigned long align_addr = ALIGN(addr, map_size);
>> +
>> + next_addr = ALIGN(addr + 1, map_size);
>> +
>> + ret = realm_rtt_fold(realm, align_addr, level, &rtt_granule);
>> +
>> + switch (RMI_RETURN_STATUS(ret)) {
>> + case RMI_SUCCESS:
>> + if (!WARN_ON(rmi_granule_undelegate(rtt_granule)))
>> + free_page((unsigned long)phys_to_virt(rtt_granule));
>
> minor nit: Do we need a wrapper function for things like this, abd
> leaking the page if undelegate fails, something like
> rme_reclaim_delegated_page() ?
I'll take a look at that - mostly sites don't have the same argument
repeated (unlike this one), so I'll have to have a think about the best
form.
>> + break;
>> + case RMI_ERROR_RTT:
>> + if (level == RME_RTT_MAX_LEVEL ||
>> + RMI_RETURN_INDEX(ret) < level) {
>> + not_folded++;
>> + break;
>> + }
>> + /* Recurse a level deeper */
>> + ret = realm_fold_rtt_level(realm,
>> + level + 1,
>> + addr,
>> + next_addr);
>> + if (ret < 0)
>> + return ret;
>> + else if (ret == 0)
>> + /* Try again at this level */
>> + next_addr = addr;
>> + break;
>> + default:
>> + return -ENXIO;
>> + }
>> + }
>> +
>> + return not_folded;
>> +}
>> +
>> +static int realm_fold_rtt_range(struct realm *realm,
>> + unsigned long start, unsigned long end)
>> +{
>> + return realm_fold_rtt_level(realm, get_start_level(realm) + 1,
>> + start, end);
>> +}
>> +
>> static void ensure_spare_page(struct realm *realm)
>> {
>> phys_addr_t tmp_rtt;
>> @@ -295,6 +532,147 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32
>> ia_bits)
>> WARN_ON(realm_tear_down_rtt_range(realm, 0, (1UL << ia_bits)));
>> }
>> +void kvm_realm_unmap_range(struct kvm *kvm, unsigned long ipa, u64
>> size,
>> + bool unmap_private)
>> +{
>> + unsigned long end = ipa + size;
>> + struct realm *realm = &kvm->arch.realm;
>> +
>> + end = min(BIT(realm->ia_bits - 1), end);
>> +
>> + ensure_spare_page(realm);
>> +
>> + realm_unmap_range(kvm, ipa, end, unmap_private);
>> +
>> + realm_fold_rtt_range(realm, ipa, end);
>
> Shouldn't this be :
>
> if (unmap_private)
> realm_fold_rtt_range(realm, ipa, end);
Indeed it's a little pointless folding if we haven't touched the private
mappings.
> Also it is fine to reclaim RTTs from the protected space, not the
> unprotected half, as long as we use RTT_DESTROY in unmap_shared routine.
>
>> +}
>> +
>> +static int find_map_level(struct realm *realm,
>> + unsigned long start,
>> + unsigned long end)
>> +{
>> + int level = RME_RTT_MAX_LEVEL;
>> +
>> + while (level > get_start_level(realm)) {
>> + unsigned long map_size = rme_rtt_level_mapsize(level - 1);
>> +
>> + if (!IS_ALIGNED(start, map_size) ||
>> + (start + map_size) > end)
>> + break;
>> +
>> + level--;
>> + }
>> +
>> + return level;
>> +}
>> +
>> +int realm_set_ipa_state(struct kvm_vcpu *vcpu,
>> + unsigned long start,
>> + unsigned long end,
>> + unsigned long ripas)
>> +{
>> + struct kvm *kvm = vcpu->kvm;
>> + struct realm *realm = &kvm->arch.realm;
>> + struct realm_rec *rec = &vcpu->arch.rec;
>> + phys_addr_t rd_phys = virt_to_phys(realm->rd);
>> + phys_addr_t rec_phys = virt_to_phys(rec->rec_page);
>> + struct kvm_mmu_memory_cache *memcache = &vcpu->arch.mmu_page_cache;
>> + unsigned long ipa = start;
>> + int ret = 0;
>> +
>> + while (ipa < end) {
>> + unsigned long next;
>> +
>> + ret = rmi_rtt_set_ripas(rd_phys, rec_phys, ipa, end, &next);
>> +
>> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
>> + int walk_level = RMI_RETURN_INDEX(ret);
>> + int level = find_map_level(realm, ipa, end);
>
> Might be worth adding a comment here. Check if the RMM needs tables to
> create deeper level tables.
Ok
>> +
>> + if (walk_level < level) {
>> + ret = realm_create_rtt_levels(realm, ipa,
>> + walk_level,
>> + level,
>> + memcache);
> /* Retry with RTTs created */
Ok
>> + if (!ret)
>> + continue;
>> + } else {
>> + ret = -EINVAL;
>> + }
>> +
>> + break;
>> + } else if (RMI_RETURN_STATUS(ret) != RMI_SUCCESS) {
>> + WARN(1, "Unexpected error in %s: %#x\n", __func__,
>> + ret);
>> + ret = -EINVAL;
>> + break;
>> + }
>> + ipa = next;
>> + }
>> +
>> + if (ripas == RMI_EMPTY && ipa != start)
>> + kvm_realm_unmap_range(kvm, start, ipa - start, true);
>
> This triggers unmapping the "shared" aliases too, which is not necessary.
Yeah, the kvm_realm_unmap_range() function is a bit problematic - the
various "unmap_private" and "only_shared" flags need a bit of a rethink.
>> +
>> + return ret;
>> +}
>> +
>> +static int realm_init_ipa_state(struct realm *realm,
>> + unsigned long ipa,
>> + unsigned long end)
>> +{
>> + phys_addr_t rd_phys = virt_to_phys(realm->rd);
>> + int ret;
>> +
>> + while (ipa < end) {
>> + unsigned long next;
>> +
>> + ret = rmi_rtt_init_ripas(rd_phys, ipa, end, &next);
>> +
>> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
>> + int err_level = RMI_RETURN_INDEX(ret);
>> + int level = find_map_level(realm, ipa, end);
>> +
>> + if (WARN_ON(err_level >= level))
>> + return -ENXIO;
>> +
>> + ret = realm_create_rtt_levels(realm, ipa,
>> + err_level,
>> + level, NULL);
>> + if (ret)
>> + return ret;
>> + /* Retry with the RTT levels in place */
>> + continue;
>> + } else if (WARN_ON(ret)) {
>> + return -ENXIO;
>> + }
>> +
>> + ipa = next;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static int kvm_init_ipa_range_realm(struct kvm *kvm,
>> + struct kvm_cap_arm_rme_init_ipa_args *args)
>> +{
>> + int ret = 0;
>> + gpa_t addr, end;
>> + struct realm *realm = &kvm->arch.realm;
>> +
>> + addr = args->init_ipa_base;
>> + end = addr + args->init_ipa_size;
>> +
>> + if (end < addr)
>> + return -EINVAL;
>> +
>> + if (kvm_realm_state(kvm) != REALM_STATE_NEW)
>> + return -EINVAL;
>> +
>> + ret = realm_init_ipa_state(realm, addr, end);
>> +
>> + return ret;
>
> super minor nit:
>
> return realm_init_ipa_state(realm, addr, end);
That was probably left over from some printk() debugging ;)
Thanks,
Steve
>> +}
>> +
>> /* Protects access to rme_vmid_bitmap */
>> static DEFINE_SPINLOCK(rme_vmid_lock);
>> static unsigned long *rme_vmid_bitmap;
>> @@ -418,6 +796,18 @@ int kvm_realm_enable_cap(struct kvm *kvm, struct
>> kvm_enable_cap *cap)
>> case KVM_CAP_ARM_RME_CREATE_RD:
>> r = kvm_create_realm(kvm);
>> break;
>> + case KVM_CAP_ARM_RME_INIT_IPA_REALM: {
>> + struct kvm_cap_arm_rme_init_ipa_args args;
>> + void __user *argp = u64_to_user_ptr(cap->args[1]);
>> +
>> + if (copy_from_user(&args, argp, sizeof(args))) {
>> + r = -EFAULT;
>> + break;
>> + }
>> +
>> + r = kvm_init_ipa_range_realm(kvm, &args);
>> + break;
>> + }
>
>
> Suzuki
On 01/05/2024 16:47, Steven Price wrote:
> On 19/04/2024 10:34, Suzuki K Poulose wrote:
>> On 12/04/2024 09:42, Steven Price wrote:
>>> Each page within the protected region of the realm guest can be marked
>>> as either RAM or EMPTY. Allow the VMM to control this before the guest
>>> has started and provide the equivalent functions to change this (with
>>> the guest's approval) at runtime.
>>>
>>> When transitioning from RIPAS RAM (1) to RIPAS EMPTY (0) the memory is
>>> unmapped from the guest and undelegated allowing the memory to be reused
>>> by the host. When transitioning to RIPAS RAM the actual population of
>>> the leaf RTTs is done later on stage 2 fault, however it may be
>>> necessary to allocate additional RTTs to represent the range requested.
>>
>> minor nit: To give a bit more context:
>>
>> "however it may be necessary to allocate additional RTTs in order for
>> the RMM to track the RIPAS for the requested range".
>
> That is what I meant - but your wording is probably clearer ;)
> Technically there's also the case where a RIPAS change will cause a
> block mapping to be split which isn't just about tracking RIPAS but also
> about breaking up the block mapping for the pages that remain.
>
>>>
>>> When freeing a block mapping it is necessary to temporarily unfold the
>>> RTT which requires delegating an extra page to the RMM, this page can
>>> then be recovered once the contents of the block mapping have been
>>> freed. A spare, delegated page (spare_page) is used for this purpose.
>>>
>>> Signed-off-by: Steven Price <[email protected]>
>>> ---
>>> arch/arm64/include/asm/kvm_rme.h | 16 ++
>>> arch/arm64/kvm/mmu.c | 8 +-
>>> arch/arm64/kvm/rme.c | 390 +++++++++++++++++++++++++++++++
>>> 3 files changed, 411 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/arch/arm64/include/asm/kvm_rme.h
>>> b/arch/arm64/include/asm/kvm_rme.h
>>> index 915e76068b00..cc8f81cfc3c0 100644
>>> --- a/arch/arm64/include/asm/kvm_rme.h
>>> +++ b/arch/arm64/include/asm/kvm_rme.h
>>> @@ -96,6 +96,14 @@ void kvm_realm_destroy_rtts(struct kvm *kvm, u32
>>> ia_bits);
>>> int kvm_create_rec(struct kvm_vcpu *vcpu);
>>> void kvm_destroy_rec(struct kvm_vcpu *vcpu);
>>> +void kvm_realm_unmap_range(struct kvm *kvm,
>>> + unsigned long ipa,
>>> + u64 size,
>>> + bool unmap_private);
>>> +int realm_set_ipa_state(struct kvm_vcpu *vcpu,
>>> + unsigned long addr, unsigned long end,
>>> + unsigned long ripas);
>>> +
>>> #define RME_RTT_BLOCK_LEVEL 2
>>> #define RME_RTT_MAX_LEVEL 3
>>> @@ -114,4 +122,12 @@ static inline unsigned long
>>> rme_rtt_level_mapsize(int level)
>>> return (1UL << RME_RTT_LEVEL_SHIFT(level));
>>> }
>>> +static inline bool realm_is_addr_protected(struct realm *realm,
>>> + unsigned long addr)
>>> +{
>>> + unsigned int ia_bits = realm->ia_bits;
>>> +
>>> + return !(addr & ~(BIT(ia_bits - 1) - 1));
>>> +}
>>> +
>>> #endif
>>> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
>>> index 46f0c4e80ace..8a7b5449697f 100644
>>> --- a/arch/arm64/kvm/mmu.c
>>> +++ b/arch/arm64/kvm/mmu.c
>>> @@ -310,6 +310,7 @@ static void invalidate_icache_guest_page(void *va,
>>> size_t size)
>>> * @start: The intermediate physical base address of the range to unmap
>>> * @size: The size of the area to unmap
>>> * @may_block: Whether or not we are permitted to block
>>> + * @only_shared: If true then protected mappings should not be unmapped
>>> *
>>> * Clear a range of stage-2 mappings, lowering the various
>>> ref-counts. Must
>>> * be called while holding mmu_lock (unless for freeing the stage2
>>> pgd before
>>> @@ -317,7 +318,7 @@ static void invalidate_icache_guest_page(void *va,
>>> size_t size)
>>> * with things behind our backs.
>>> */
>>> static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t
>>> start, u64 size,
>>> - bool may_block)
>>> + bool may_block, bool only_shared)
>>> {
>>> struct kvm *kvm = kvm_s2_mmu_to_kvm(mmu);
>>> phys_addr_t end = start + size;
>>> @@ -330,7 +331,7 @@ static void __unmap_stage2_range(struct kvm_s2_mmu
>>> *mmu, phys_addr_t start, u64
>>> static void unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t
>>> start, u64 size)
>>> {
>>> - __unmap_stage2_range(mmu, start, size, true);
>>> + __unmap_stage2_range(mmu, start, size, true, false);
>>> }
>>> static void stage2_flush_memslot(struct kvm *kvm,
>>> @@ -1771,7 +1772,8 @@ bool kvm_unmap_gfn_range(struct kvm *kvm, struct
>>> kvm_gfn_range *range)
>>> __unmap_stage2_range(&kvm->arch.mmu, range->start << PAGE_SHIFT,
>>> (range->end - range->start) << PAGE_SHIFT,
>>> - range->may_block);
>>> + range->may_block,
>>> + range->only_shared);
>>> return false;
>>> }
>>> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
>>> index 629a095bea61..9e5983c51393 100644
>>> --- a/arch/arm64/kvm/rme.c
>>> +++ b/arch/arm64/kvm/rme.c
>>> @@ -79,6 +79,12 @@ static phys_addr_t __alloc_delegated_page(struct
>>> realm *realm,
>>> return phys;
>>> }
>>> +static phys_addr_t alloc_delegated_page(struct realm *realm,
>>> + struct kvm_mmu_memory_cache *mc)
>>> +{
>>> + return __alloc_delegated_page(realm, mc, GFP_KERNEL);
>>> +}
>>> +
>>> static void free_delegated_page(struct realm *realm, phys_addr_t phys)
>>> {
>>> if (realm->spare_page == PHYS_ADDR_MAX) {
>>> @@ -94,6 +100,151 @@ static void free_delegated_page(struct realm
>>> *realm, phys_addr_t phys)
>>> free_page((unsigned long)phys_to_virt(phys));
>>> }
>>> +static int realm_rtt_create(struct realm *realm,
>>> + unsigned long addr,
>>> + int level,
>>> + phys_addr_t phys)
>>> +{
>>> + addr = ALIGN_DOWN(addr, rme_rtt_level_mapsize(level - 1));
>>> + return rmi_rtt_create(virt_to_phys(realm->rd), phys, addr, level);
>>> +}
>>> +
>>> +static int realm_rtt_fold(struct realm *realm,
>>> + unsigned long addr,
>>> + int level,
>>> + phys_addr_t *rtt_granule)
>>> +{
>>> + unsigned long out_rtt;
>>> + int ret;
>>> +
>>> + ret = rmi_rtt_fold(virt_to_phys(realm->rd), addr, level, &out_rtt);
>>> +
>>> + if (RMI_RETURN_STATUS(ret) == RMI_SUCCESS && rtt_granule)
>>> + *rtt_granule = out_rtt;
>>> +
>>> + return ret;
>>> +}
>>> +
>>> +static int realm_destroy_protected(struct realm *realm,
>>> + unsigned long ipa,
>>> + unsigned long *next_addr)
>>> +{
>>> + unsigned long rd = virt_to_phys(realm->rd);
>>> + unsigned long addr;
>>> + phys_addr_t rtt;
>>> + int ret;
>>> +
>>> +loop:
>>> + ret = rmi_data_destroy(rd, ipa, &addr, next_addr);
>>> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
>>> + if (*next_addr > ipa)
>>> + return 0; /* UNASSIGNED */
>>> + rtt = alloc_delegated_page(realm, NULL);
>>> + if (WARN_ON(rtt == PHYS_ADDR_MAX))
>>> + return -1;
>>> + /* ASSIGNED - ipa is mapped as a block, so split */
>>> + ret = realm_rtt_create(realm, ipa,
>>> + RMI_RETURN_INDEX(ret) + 1, rtt);
>>
>> Could we not go all the way to L3 (rather than 1 level deeper) and try
>> again ? That way, we are covered for block mappings at L1 (1G).
>
> I don't think this situation can happen. The spec states that for
> RMI_ERROR_RTT either top>ipa (UNASSIGNED, in which case we've bailed out
> above) or top==ipa (ASSIGNED) in which case RMI_RETURN_INDEX(ret) must
> equal 2.
It could be 1 as RMM spec allows Block mapping at L1. But if we don't
fold to that level in KVM, the current code may be fine.
>
> So in this case realm_rtt_create() will always be called with level=3.
>
> I can simplify this by explicitly passing 3 and adding a WARN_ON() for
> RMI_RETURN_INDEX(ret) != 2.
Yes please, if we don't fold all the way up to L1.
Thanks
Suzuki
On 25/04/2024 11:43, Fuad Tabba wrote:
> Hi,
Hi,
Thanks for the review. Sorry I didn't respond earlier.
> On Fri, Apr 12, 2024 at 9:44 AM Steven Price <[email protected]> wrote:
>>
<snip>
>> +static int private_memslot_fault(struct kvm_vcpu *vcpu,
>> + phys_addr_t fault_ipa,
>> + struct kvm_memory_slot *memslot)
>> +{
>> + struct kvm *kvm = vcpu->kvm;
>> + gpa_t gpa_stolen_mask = kvm_gpa_stolen_bits(kvm);
>> + gfn_t gfn = (fault_ipa & ~gpa_stolen_mask) >> PAGE_SHIFT;
>> + bool is_priv_gfn = !((fault_ipa & gpa_stolen_mask) == gpa_stolen_mask);
>> + bool priv_exists = kvm_mem_is_private(kvm, gfn);
>> + struct kvm_mmu_memory_cache *memcache = &vcpu->arch.mmu_page_cache;
>> + int order;
>> + kvm_pfn_t pfn;
>> + int ret;
>> +
>> + if (priv_exists != is_priv_gfn) {
>> + kvm_prepare_memory_fault_exit(vcpu,
>> + fault_ipa & ~gpa_stolen_mask,
>> + PAGE_SIZE,
>> + kvm_is_write_fault(vcpu),
>> + false, is_priv_gfn);
>> +
>> + return 0;
>> + }
>> +
>> + if (!is_priv_gfn) {
>> + /* Not a private mapping, handling normally */
>> + return -EAGAIN;
>> + }
>> +
>> + if (kvm_gmem_get_pfn(kvm, memslot, gfn, &pfn, &order))
>> + return 1; /* Retry */
>
> You don't need to pass a variable to hold the order if you don't need
> it. You can pass NULL.
Ah, good point - that simplifies things.
> I am also confused about the return, why do you return 1 regardless of
> the reason kvm_gmem_get_pfn() fails?
Thinking about this, I don't think we actually expect kvm_gmem_get_pfn()
to fail, so it's actually more appropriate to just pass return any error
value.
>> + ret = kvm_mmu_topup_memory_cache(memcache,
>> + kvm_mmu_cache_min_pages(vcpu->arch.hw_mmu));
>> + if (ret)
>> + return ret;
>
> If this fails you should release the page you got earlier (e.g.,
> kvm_release_pfn_clean()), or you could move it before
> kvm_gmem_get_pfn().
Good point, however...
>> + /* FIXME: Should be able to use bigger than PAGE_SIZE mappings */
>> + ret = realm_map_ipa(kvm, fault_ipa, pfn, PAGE_SIZE, KVM_PGTABLE_PROT_W,
>> + memcache);
>> + if (!ret)
>> + return 1; /* Handled */
>
> Should also release the page if it fails. Speaking of which,
> where/when do you eventually release the page?
.. I messed this up ;) It seems I'm managing to leak all guestmem
pages. I'm not sure what I was thinking but I think I'd got it into my
head guestmem wasn't reference counting the pages. I'll fix this up in
the next version.
Thanks,
Steve