This series extends KVM RISC-V to detect and use Svpbmt for both
G-stage (hypervisor) and VS-stage (guest) page table.
The corresponding KVMTOOL patches used for testing this series
can be found in riscv_svpbmt_sstc_v1 branch at:
https://github.com/avpatel/kvmtool.git
These patches can also be found in riscv_kvm_svpbmt_v1 branch at:
https://github.com/avpatel/linux.git
Alexandre Ghiti (1):
riscv: Fix missing PAGE_PFN_MASK
Anup Patel (4):
KVM: Add gfp_custom flag in struct kvm_mmu_memory_cache
RISC-V: KVM: Add G-stage ioremap() and iounmap() functions
RISC-V: KVM: Use PAGE_KERNEL_IO in kvm_riscv_gstage_ioremap()
RISC-V: KVM: Add support for Svpbmt inside Guest/VM
arch/riscv/include/asm/csr.h | 16 ++++++++++++++++
arch/riscv/include/asm/kvm_host.h | 5 +++++
arch/riscv/include/asm/pgtable-64.h | 12 ++++++------
arch/riscv/include/asm/pgtable.h | 6 +++---
arch/riscv/include/uapi/asm/kvm.h | 1 +
arch/riscv/kvm/mmu.c | 22 ++++++++++++++++------
arch/riscv/kvm/vcpu.c | 16 ++++++++++++++++
include/linux/kvm_types.h | 1 +
virt/kvm/kvm_main.c | 4 +++-
9 files changed, 67 insertions(+), 16 deletions(-)
--
2.34.1
The kvm_mmu_topup_memory_cache() always uses GFP_KERNEL_ACCOUNT for
memory allocation which prevents it's use in atomic context. To address
this limitation of kvm_mmu_topup_memory_cache(), we add gfp_custom flag
in struct kvm_mmu_memory_cache. When the gfp_custom flag is set to some
GFP_xyz flags, the kvm_mmu_topup_memory_cache() will use that instead of
GFP_KERNEL_ACCOUNT.
Signed-off-by: Anup Patel <[email protected]>
---
include/linux/kvm_types.h | 1 +
virt/kvm/kvm_main.c | 4 +++-
2 files changed, 4 insertions(+), 1 deletion(-)
diff --git a/include/linux/kvm_types.h b/include/linux/kvm_types.h
index ac1ebb37a0ff..1dcfba68076a 100644
--- a/include/linux/kvm_types.h
+++ b/include/linux/kvm_types.h
@@ -87,6 +87,7 @@ struct gfn_to_pfn_cache {
struct kvm_mmu_memory_cache {
int nobjs;
gfp_t gfp_zero;
+ gfp_t gfp_custom;
struct kmem_cache *kmem_cache;
void *objects[KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE];
};
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index a49df8988cd6..e3a6f7647474 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -386,7 +386,9 @@ int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min)
if (mc->nobjs >= min)
return 0;
while (mc->nobjs < ARRAY_SIZE(mc->objects)) {
- obj = mmu_memory_cache_alloc_obj(mc, GFP_KERNEL_ACCOUNT);
+ obj = mmu_memory_cache_alloc_obj(mc, (mc->gfp_custom) ?
+ mc->gfp_custom :
+ GFP_KERNEL_ACCOUNT);
if (!obj)
return mc->nobjs >= min ? 0 : -ENOMEM;
mc->objects[mc->nobjs++] = obj;
--
2.34.1
From: Alexandre Ghiti <[email protected]>
There are a bunch of functions that use the PFN from a page table entry
that end up with the svpbmt upper-bits because they are missing the newly
introduced PAGE_PFN_MASK which leads to wrong addresses conversions and
then crash: fix this by adding this mask.
Fixes: 100631b48ded ("riscv: Fix accessing pfn bits in PTEs for non-32bit variants")
Signed-off-by: Alexandre Ghiti <[email protected]>
Reviewed-by: Anup Patel <[email protected]>
---
arch/riscv/include/asm/pgtable-64.h | 12 ++++++------
arch/riscv/include/asm/pgtable.h | 6 +++---
arch/riscv/kvm/mmu.c | 2 +-
3 files changed, 10 insertions(+), 10 deletions(-)
diff --git a/arch/riscv/include/asm/pgtable-64.h b/arch/riscv/include/asm/pgtable-64.h
index 5c2aba5efbd0..dc42375c2357 100644
--- a/arch/riscv/include/asm/pgtable-64.h
+++ b/arch/riscv/include/asm/pgtable-64.h
@@ -175,7 +175,7 @@ static inline pud_t pfn_pud(unsigned long pfn, pgprot_t prot)
static inline unsigned long _pud_pfn(pud_t pud)
{
- return pud_val(pud) >> _PAGE_PFN_SHIFT;
+ return __page_val_to_pfn(pud_val(pud));
}
static inline pmd_t *pud_pgtable(pud_t pud)
@@ -278,13 +278,13 @@ static inline p4d_t pfn_p4d(unsigned long pfn, pgprot_t prot)
static inline unsigned long _p4d_pfn(p4d_t p4d)
{
- return p4d_val(p4d) >> _PAGE_PFN_SHIFT;
+ return __page_val_to_pfn(p4d_val(p4d));
}
static inline pud_t *p4d_pgtable(p4d_t p4d)
{
if (pgtable_l4_enabled)
- return (pud_t *)pfn_to_virt(p4d_val(p4d) >> _PAGE_PFN_SHIFT);
+ return (pud_t *)pfn_to_virt(__page_val_to_pfn(p4d_val(p4d)));
return (pud_t *)pud_pgtable((pud_t) { p4d_val(p4d) });
}
@@ -292,7 +292,7 @@ static inline pud_t *p4d_pgtable(p4d_t p4d)
static inline struct page *p4d_page(p4d_t p4d)
{
- return pfn_to_page(p4d_val(p4d) >> _PAGE_PFN_SHIFT);
+ return pfn_to_page(__page_val_to_pfn(p4d_val(p4d)));
}
#define pud_index(addr) (((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1))
@@ -347,7 +347,7 @@ static inline void pgd_clear(pgd_t *pgd)
static inline p4d_t *pgd_pgtable(pgd_t pgd)
{
if (pgtable_l5_enabled)
- return (p4d_t *)pfn_to_virt(pgd_val(pgd) >> _PAGE_PFN_SHIFT);
+ return (p4d_t *)pfn_to_virt(__page_val_to_pfn(pgd_val(pgd)));
return (p4d_t *)p4d_pgtable((p4d_t) { pgd_val(pgd) });
}
@@ -355,7 +355,7 @@ static inline p4d_t *pgd_pgtable(pgd_t pgd)
static inline struct page *pgd_page(pgd_t pgd)
{
- return pfn_to_page(pgd_val(pgd) >> _PAGE_PFN_SHIFT);
+ return pfn_to_page(__page_val_to_pfn(pgd_val(pgd)));
}
#define pgd_page(pgd) pgd_page(pgd)
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 1d1be9d9419c..5dbd6610729b 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -261,7 +261,7 @@ static inline pgd_t pfn_pgd(unsigned long pfn, pgprot_t prot)
static inline unsigned long _pgd_pfn(pgd_t pgd)
{
- return pgd_val(pgd) >> _PAGE_PFN_SHIFT;
+ return __page_val_to_pfn(pgd_val(pgd));
}
static inline struct page *pmd_page(pmd_t pmd)
@@ -590,14 +590,14 @@ static inline pmd_t pmd_mkinvalid(pmd_t pmd)
return __pmd(pmd_val(pmd) & ~(_PAGE_PRESENT|_PAGE_PROT_NONE));
}
-#define __pmd_to_phys(pmd) (pmd_val(pmd) >> _PAGE_PFN_SHIFT << PAGE_SHIFT)
+#define __pmd_to_phys(pmd) (__page_val_to_pfn(pmd_val(pmd)) << PAGE_SHIFT)
static inline unsigned long pmd_pfn(pmd_t pmd)
{
return ((__pmd_to_phys(pmd) & PMD_MASK) >> PAGE_SHIFT);
}
-#define __pud_to_phys(pud) (pud_val(pud) >> _PAGE_PFN_SHIFT << PAGE_SHIFT)
+#define __pud_to_phys(pud) (__page_val_to_pfn(pud_val(pud)) << PAGE_SHIFT)
static inline unsigned long pud_pfn(pud_t pud)
{
diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
index 2965284a490d..b75d4e200064 100644
--- a/arch/riscv/kvm/mmu.c
+++ b/arch/riscv/kvm/mmu.c
@@ -54,7 +54,7 @@ static inline unsigned long gstage_pte_index(gpa_t addr, u32 level)
static inline unsigned long gstage_pte_page_vaddr(pte_t pte)
{
- return (unsigned long)pfn_to_virt(pte_val(pte) >> _PAGE_PFN_SHIFT);
+ return (unsigned long)pfn_to_virt(__page_val_to_pfn(pte_val(pte)));
}
static int gstage_page_size_to_level(unsigned long page_size, u32 *out_level)
--
2.34.1
The in-kernel AIA IMSIC support requires on-demand mapping / unmapping
of Guest IMSIC address to Host IMSIC guest files. To help achieve this,
we add kvm_riscv_stage2_ioremap() and kvm_riscv_stage2_iounmap()
functions. These new functions for updating G-stage page table mappings
will be called in atomic context so we have special "in_atomic" parameter
for this purpose.
Signed-off-by: Anup Patel <[email protected]>
---
arch/riscv/include/asm/kvm_host.h | 5 +++++
arch/riscv/kvm/mmu.c | 18 ++++++++++++++----
2 files changed, 19 insertions(+), 4 deletions(-)
diff --git a/arch/riscv/include/asm/kvm_host.h b/arch/riscv/include/asm/kvm_host.h
index 59a0cf2ca7b9..60c517e4d576 100644
--- a/arch/riscv/include/asm/kvm_host.h
+++ b/arch/riscv/include/asm/kvm_host.h
@@ -284,6 +284,11 @@ void kvm_riscv_hfence_vvma_gva(struct kvm *kvm,
void kvm_riscv_hfence_vvma_all(struct kvm *kvm,
unsigned long hbase, unsigned long hmask);
+int kvm_riscv_gstage_ioremap(struct kvm *kvm, gpa_t gpa,
+ phys_addr_t hpa, unsigned long size,
+ bool writable, bool in_atomic);
+void kvm_riscv_gstage_iounmap(struct kvm *kvm, gpa_t gpa,
+ unsigned long size);
int kvm_riscv_gstage_map(struct kvm_vcpu *vcpu,
struct kvm_memory_slot *memslot,
gpa_t gpa, unsigned long hva, bool is_write);
diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
index b75d4e200064..f7862ca4c4c6 100644
--- a/arch/riscv/kvm/mmu.c
+++ b/arch/riscv/kvm/mmu.c
@@ -343,8 +343,9 @@ static void gstage_wp_memory_region(struct kvm *kvm, int slot)
kvm_flush_remote_tlbs(kvm);
}
-static int gstage_ioremap(struct kvm *kvm, gpa_t gpa, phys_addr_t hpa,
- unsigned long size, bool writable)
+int kvm_riscv_gstage_ioremap(struct kvm *kvm, gpa_t gpa,
+ phys_addr_t hpa, unsigned long size,
+ bool writable, bool in_atomic)
{
pte_t pte;
int ret = 0;
@@ -353,6 +354,7 @@ static int gstage_ioremap(struct kvm *kvm, gpa_t gpa, phys_addr_t hpa,
struct kvm_mmu_memory_cache pcache;
memset(&pcache, 0, sizeof(pcache));
+ pcache.gfp_custom = (in_atomic) ? GFP_ATOMIC | __GFP_ACCOUNT : 0;
pcache.gfp_zero = __GFP_ZERO;
end = (gpa + size + PAGE_SIZE - 1) & PAGE_MASK;
@@ -382,6 +384,13 @@ static int gstage_ioremap(struct kvm *kvm, gpa_t gpa, phys_addr_t hpa,
return ret;
}
+void kvm_riscv_gstage_iounmap(struct kvm *kvm, gpa_t gpa, unsigned long size)
+{
+ spin_lock(&kvm->mmu_lock);
+ gstage_unmap_range(kvm, gpa, size, false);
+ spin_unlock(&kvm->mmu_lock);
+}
+
void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
struct kvm_memory_slot *slot,
gfn_t gfn_offset,
@@ -517,8 +526,9 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
goto out;
}
- ret = gstage_ioremap(kvm, gpa, pa,
- vm_end - vm_start, writable);
+ ret = kvm_riscv_gstage_ioremap(kvm, gpa, pa,
+ vm_end - vm_start,
+ writable, false);
if (ret)
break;
}
--
2.34.1
The Guest/VM can use Svpbmt in VS-stage page tables when allowed by the
Hypervisor using the henvcfg.PBMTE bit.
We add Svpbmt support for the KVM Guest/VM which can be enabled/disabled
by the KVM user-space (QEMU/KVMTOOL) using the ISA extension ONE_REG
interface.
Signed-off-by: Anup Patel <[email protected]>
---
arch/riscv/include/asm/csr.h | 16 ++++++++++++++++
arch/riscv/include/uapi/asm/kvm.h | 1 +
arch/riscv/kvm/vcpu.c | 16 ++++++++++++++++
3 files changed, 33 insertions(+)
diff --git a/arch/riscv/include/asm/csr.h b/arch/riscv/include/asm/csr.h
index 6d85655e7edf..17516afc389a 100644
--- a/arch/riscv/include/asm/csr.h
+++ b/arch/riscv/include/asm/csr.h
@@ -156,6 +156,18 @@
(_AC(1, UL) << IRQ_S_TIMER) | \
(_AC(1, UL) << IRQ_S_EXT))
+/* xENVCFG flags */
+#define ENVCFG_STCE (_AC(1, ULL) << 63)
+#define ENVCFG_PBMTE (_AC(1, ULL) << 62)
+#define ENVCFG_CBZE (_AC(1, UL) << 7)
+#define ENVCFG_CBCFE (_AC(1, UL) << 6)
+#define ENVCFG_CBIE_SHIFT 4
+#define ENVCFG_CBIE (_AC(0x3, UL) << ENVCFG_CBIE_SHIFT)
+#define ENVCFG_CBIE_ILL _AC(0x0, UL)
+#define ENVCFG_CBIE_FLUSH _AC(0x1, UL)
+#define ENVCFG_CBIE_INV _AC(0x3, UL)
+#define ENVCFG_FIOM _AC(0x1, UL)
+
/* symbolic CSR names: */
#define CSR_CYCLE 0xc00
#define CSR_TIME 0xc01
@@ -252,7 +264,9 @@
#define CSR_HTIMEDELTA 0x605
#define CSR_HCOUNTEREN 0x606
#define CSR_HGEIE 0x607
+#define CSR_HENVCFG 0x60a
#define CSR_HTIMEDELTAH 0x615
+#define CSR_HENVCFGH 0x61a
#define CSR_HTVAL 0x643
#define CSR_HIP 0x644
#define CSR_HVIP 0x645
@@ -264,6 +278,8 @@
#define CSR_MISA 0x301
#define CSR_MIE 0x304
#define CSR_MTVEC 0x305
+#define CSR_MENVCFG 0x30a
+#define CSR_MENVCFGH 0x31a
#define CSR_MSCRATCH 0x340
#define CSR_MEPC 0x341
#define CSR_MCAUSE 0x342
diff --git a/arch/riscv/include/uapi/asm/kvm.h b/arch/riscv/include/uapi/asm/kvm.h
index 6119368ba6d5..24b2a6e27698 100644
--- a/arch/riscv/include/uapi/asm/kvm.h
+++ b/arch/riscv/include/uapi/asm/kvm.h
@@ -96,6 +96,7 @@ enum KVM_RISCV_ISA_EXT_ID {
KVM_RISCV_ISA_EXT_H,
KVM_RISCV_ISA_EXT_I,
KVM_RISCV_ISA_EXT_M,
+ KVM_RISCV_ISA_EXT_SVPBMT,
KVM_RISCV_ISA_EXT_MAX,
};
diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
index 6dd9cf729614..b7a433c54d0f 100644
--- a/arch/riscv/kvm/vcpu.c
+++ b/arch/riscv/kvm/vcpu.c
@@ -51,6 +51,7 @@ static const unsigned long kvm_isa_ext_arr[] = {
RISCV_ISA_EXT_h,
RISCV_ISA_EXT_i,
RISCV_ISA_EXT_m,
+ RISCV_ISA_EXT_SVPBMT,
};
static unsigned long kvm_riscv_vcpu_base2isa_ext(unsigned long base_ext)
@@ -777,6 +778,19 @@ int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
return -EINVAL;
}
+static void kvm_riscv_vcpu_update_config(const unsigned long *isa)
+{
+ u64 henvcfg = 0;
+
+ if (__riscv_isa_extension_available(isa, RISCV_ISA_EXT_SVPBMT))
+ henvcfg |= ENVCFG_PBMTE;
+
+ csr_write(CSR_HENVCFG, henvcfg);
+#ifdef CONFIG_32BIT
+ csr_write(CSR_HENVCFGH, henvcfg >> 32);
+#endif
+}
+
void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
{
struct kvm_vcpu_csr *csr = &vcpu->arch.guest_csr;
@@ -791,6 +805,8 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
csr_write(CSR_HVIP, csr->hvip);
csr_write(CSR_VSATP, csr->vsatp);
+ kvm_riscv_vcpu_update_config(vcpu->arch.isa);
+
kvm_riscv_gstage_update_hgatp(vcpu);
kvm_riscv_vcpu_timer_restore(vcpu);
--
2.34.1
When the host has Svpbmt extension, we should use page based memory
type 2 (i.e. IO) for IO mappings in the G-stage page table.
To achieve this, we replace use of PAGE_KERNEL with PAGE_KERNEL_IO
in the kvm_riscv_gstage_ioremap().
Signed-off-by: Anup Patel <[email protected]>
---
arch/riscv/kvm/mmu.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
index f7862ca4c4c6..bc545aef6034 100644
--- a/arch/riscv/kvm/mmu.c
+++ b/arch/riscv/kvm/mmu.c
@@ -361,7 +361,7 @@ int kvm_riscv_gstage_ioremap(struct kvm *kvm, gpa_t gpa,
pfn = __phys_to_pfn(hpa);
for (addr = gpa; addr < end; addr += PAGE_SIZE) {
- pte = pfn_pte(pfn, PAGE_KERNEL);
+ pte = pfn_pte(pfn, PAGE_KERNEL_IO);
if (!writable)
pte = pte_wrprotect(pte);
--
2.34.1
On Thu, Jul 7, 2022 at 8:23 PM Anup Patel <[email protected]> wrote:
>
> From: Alexandre Ghiti <[email protected]>
>
> There are a bunch of functions that use the PFN from a page table entry
> that end up with the svpbmt upper-bits because they are missing the newly
> introduced PAGE_PFN_MASK which leads to wrong addresses conversions and
> then crash: fix this by adding this mask.
>
> Fixes: 100631b48ded ("riscv: Fix accessing pfn bits in PTEs for non-32bit variants")
> Signed-off-by: Alexandre Ghiti <[email protected]>
> Reviewed-by: Anup Patel <[email protected]>
I have queued this patch for 5.19-rcX fixes which I will send-out this week.
Thanks,
Anup
> ---
> arch/riscv/include/asm/pgtable-64.h | 12 ++++++------
> arch/riscv/include/asm/pgtable.h | 6 +++---
> arch/riscv/kvm/mmu.c | 2 +-
> 3 files changed, 10 insertions(+), 10 deletions(-)
>
> diff --git a/arch/riscv/include/asm/pgtable-64.h b/arch/riscv/include/asm/pgtable-64.h
> index 5c2aba5efbd0..dc42375c2357 100644
> --- a/arch/riscv/include/asm/pgtable-64.h
> +++ b/arch/riscv/include/asm/pgtable-64.h
> @@ -175,7 +175,7 @@ static inline pud_t pfn_pud(unsigned long pfn, pgprot_t prot)
>
> static inline unsigned long _pud_pfn(pud_t pud)
> {
> - return pud_val(pud) >> _PAGE_PFN_SHIFT;
> + return __page_val_to_pfn(pud_val(pud));
> }
>
> static inline pmd_t *pud_pgtable(pud_t pud)
> @@ -278,13 +278,13 @@ static inline p4d_t pfn_p4d(unsigned long pfn, pgprot_t prot)
>
> static inline unsigned long _p4d_pfn(p4d_t p4d)
> {
> - return p4d_val(p4d) >> _PAGE_PFN_SHIFT;
> + return __page_val_to_pfn(p4d_val(p4d));
> }
>
> static inline pud_t *p4d_pgtable(p4d_t p4d)
> {
> if (pgtable_l4_enabled)
> - return (pud_t *)pfn_to_virt(p4d_val(p4d) >> _PAGE_PFN_SHIFT);
> + return (pud_t *)pfn_to_virt(__page_val_to_pfn(p4d_val(p4d)));
>
> return (pud_t *)pud_pgtable((pud_t) { p4d_val(p4d) });
> }
> @@ -292,7 +292,7 @@ static inline pud_t *p4d_pgtable(p4d_t p4d)
>
> static inline struct page *p4d_page(p4d_t p4d)
> {
> - return pfn_to_page(p4d_val(p4d) >> _PAGE_PFN_SHIFT);
> + return pfn_to_page(__page_val_to_pfn(p4d_val(p4d)));
> }
>
> #define pud_index(addr) (((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1))
> @@ -347,7 +347,7 @@ static inline void pgd_clear(pgd_t *pgd)
> static inline p4d_t *pgd_pgtable(pgd_t pgd)
> {
> if (pgtable_l5_enabled)
> - return (p4d_t *)pfn_to_virt(pgd_val(pgd) >> _PAGE_PFN_SHIFT);
> + return (p4d_t *)pfn_to_virt(__page_val_to_pfn(pgd_val(pgd)));
>
> return (p4d_t *)p4d_pgtable((p4d_t) { pgd_val(pgd) });
> }
> @@ -355,7 +355,7 @@ static inline p4d_t *pgd_pgtable(pgd_t pgd)
>
> static inline struct page *pgd_page(pgd_t pgd)
> {
> - return pfn_to_page(pgd_val(pgd) >> _PAGE_PFN_SHIFT);
> + return pfn_to_page(__page_val_to_pfn(pgd_val(pgd)));
> }
> #define pgd_page(pgd) pgd_page(pgd)
>
> diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
> index 1d1be9d9419c..5dbd6610729b 100644
> --- a/arch/riscv/include/asm/pgtable.h
> +++ b/arch/riscv/include/asm/pgtable.h
> @@ -261,7 +261,7 @@ static inline pgd_t pfn_pgd(unsigned long pfn, pgprot_t prot)
>
> static inline unsigned long _pgd_pfn(pgd_t pgd)
> {
> - return pgd_val(pgd) >> _PAGE_PFN_SHIFT;
> + return __page_val_to_pfn(pgd_val(pgd));
> }
>
> static inline struct page *pmd_page(pmd_t pmd)
> @@ -590,14 +590,14 @@ static inline pmd_t pmd_mkinvalid(pmd_t pmd)
> return __pmd(pmd_val(pmd) & ~(_PAGE_PRESENT|_PAGE_PROT_NONE));
> }
>
> -#define __pmd_to_phys(pmd) (pmd_val(pmd) >> _PAGE_PFN_SHIFT << PAGE_SHIFT)
> +#define __pmd_to_phys(pmd) (__page_val_to_pfn(pmd_val(pmd)) << PAGE_SHIFT)
>
> static inline unsigned long pmd_pfn(pmd_t pmd)
> {
> return ((__pmd_to_phys(pmd) & PMD_MASK) >> PAGE_SHIFT);
> }
>
> -#define __pud_to_phys(pud) (pud_val(pud) >> _PAGE_PFN_SHIFT << PAGE_SHIFT)
> +#define __pud_to_phys(pud) (__page_val_to_pfn(pud_val(pud)) << PAGE_SHIFT)
>
> static inline unsigned long pud_pfn(pud_t pud)
> {
> diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
> index 2965284a490d..b75d4e200064 100644
> --- a/arch/riscv/kvm/mmu.c
> +++ b/arch/riscv/kvm/mmu.c
> @@ -54,7 +54,7 @@ static inline unsigned long gstage_pte_index(gpa_t addr, u32 level)
>
> static inline unsigned long gstage_pte_page_vaddr(pte_t pte)
> {
> - return (unsigned long)pfn_to_virt(pte_val(pte) >> _PAGE_PFN_SHIFT);
> + return (unsigned long)pfn_to_virt(__page_val_to_pfn(pte_val(pte)));
> }
>
> static int gstage_page_size_to_level(unsigned long page_size, u32 *out_level)
> --
> 2.34.1
>
On Thu, Jul 7, 2022 at 7:53 AM Anup Patel <[email protected]> wrote:
>
> The kvm_mmu_topup_memory_cache() always uses GFP_KERNEL_ACCOUNT for
> memory allocation which prevents it's use in atomic context. To address
> this limitation of kvm_mmu_topup_memory_cache(), we add gfp_custom flag
> in struct kvm_mmu_memory_cache. When the gfp_custom flag is set to some
> GFP_xyz flags, the kvm_mmu_topup_memory_cache() will use that instead of
> GFP_KERNEL_ACCOUNT.
>
> Signed-off-by: Anup Patel <[email protected]>
> ---
> include/linux/kvm_types.h | 1 +
> virt/kvm/kvm_main.c | 4 +++-
> 2 files changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/include/linux/kvm_types.h b/include/linux/kvm_types.h
> index ac1ebb37a0ff..1dcfba68076a 100644
> --- a/include/linux/kvm_types.h
> +++ b/include/linux/kvm_types.h
> @@ -87,6 +87,7 @@ struct gfn_to_pfn_cache {
> struct kvm_mmu_memory_cache {
> int nobjs;
> gfp_t gfp_zero;
> + gfp_t gfp_custom;
> struct kmem_cache *kmem_cache;
> void *objects[KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE];
> };
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index a49df8988cd6..e3a6f7647474 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -386,7 +386,9 @@ int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min)
> if (mc->nobjs >= min)
> return 0;
> while (mc->nobjs < ARRAY_SIZE(mc->objects)) {
> - obj = mmu_memory_cache_alloc_obj(mc, GFP_KERNEL_ACCOUNT);
> + obj = mmu_memory_cache_alloc_obj(mc, (mc->gfp_custom) ?
> + mc->gfp_custom :
> + GFP_KERNEL_ACCOUNT);
> if (!obj)
> return mc->nobjs >= min ? 0 : -ENOMEM;
> mc->objects[mc->nobjs++] = obj;
> --
> 2.34.1
>
Acked-by: Atish Patra <[email protected]>
--
Regards,
Atish
On Thu, Jul 7, 2022 at 7:53 AM Anup Patel <[email protected]> wrote:
>
> When the host has Svpbmt extension, we should use page based memory
> type 2 (i.e. IO) for IO mappings in the G-stage page table.
>
> To achieve this, we replace use of PAGE_KERNEL with PAGE_KERNEL_IO
> in the kvm_riscv_gstage_ioremap().
>
> Signed-off-by: Anup Patel <[email protected]>
> ---
> arch/riscv/kvm/mmu.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
> index f7862ca4c4c6..bc545aef6034 100644
> --- a/arch/riscv/kvm/mmu.c
> +++ b/arch/riscv/kvm/mmu.c
> @@ -361,7 +361,7 @@ int kvm_riscv_gstage_ioremap(struct kvm *kvm, gpa_t gpa,
> pfn = __phys_to_pfn(hpa);
>
> for (addr = gpa; addr < end; addr += PAGE_SIZE) {
> - pte = pfn_pte(pfn, PAGE_KERNEL);
> + pte = pfn_pte(pfn, PAGE_KERNEL_IO);
>
> if (!writable)
> pte = pte_wrprotect(pte);
> --
> 2.34.1
>
LGTM.
Reviewed-by: Atish Patra <[email protected]>
--
Regards,
Atish
On Thu, Jul 7, 2022 at 7:53 AM Anup Patel <[email protected]> wrote:
>
> The Guest/VM can use Svpbmt in VS-stage page tables when allowed by the
> Hypervisor using the henvcfg.PBMTE bit.
>
> We add Svpbmt support for the KVM Guest/VM which can be enabled/disabled
> by the KVM user-space (QEMU/KVMTOOL) using the ISA extension ONE_REG
> interface.
>
> Signed-off-by: Anup Patel <[email protected]>
> ---
> arch/riscv/include/asm/csr.h | 16 ++++++++++++++++
> arch/riscv/include/uapi/asm/kvm.h | 1 +
> arch/riscv/kvm/vcpu.c | 16 ++++++++++++++++
> 3 files changed, 33 insertions(+)
>
> diff --git a/arch/riscv/include/asm/csr.h b/arch/riscv/include/asm/csr.h
> index 6d85655e7edf..17516afc389a 100644
> --- a/arch/riscv/include/asm/csr.h
> +++ b/arch/riscv/include/asm/csr.h
> @@ -156,6 +156,18 @@
> (_AC(1, UL) << IRQ_S_TIMER) | \
> (_AC(1, UL) << IRQ_S_EXT))
>
> +/* xENVCFG flags */
> +#define ENVCFG_STCE (_AC(1, ULL) << 63)
> +#define ENVCFG_PBMTE (_AC(1, ULL) << 62)
> +#define ENVCFG_CBZE (_AC(1, UL) << 7)
> +#define ENVCFG_CBCFE (_AC(1, UL) << 6)
> +#define ENVCFG_CBIE_SHIFT 4
> +#define ENVCFG_CBIE (_AC(0x3, UL) << ENVCFG_CBIE_SHIFT)
> +#define ENVCFG_CBIE_ILL _AC(0x0, UL)
> +#define ENVCFG_CBIE_FLUSH _AC(0x1, UL)
> +#define ENVCFG_CBIE_INV _AC(0x3, UL)
> +#define ENVCFG_FIOM _AC(0x1, UL)
> +
> /* symbolic CSR names: */
> #define CSR_CYCLE 0xc00
> #define CSR_TIME 0xc01
> @@ -252,7 +264,9 @@
> #define CSR_HTIMEDELTA 0x605
> #define CSR_HCOUNTEREN 0x606
> #define CSR_HGEIE 0x607
> +#define CSR_HENVCFG 0x60a
> #define CSR_HTIMEDELTAH 0x615
> +#define CSR_HENVCFGH 0x61a
> #define CSR_HTVAL 0x643
> #define CSR_HIP 0x644
> #define CSR_HVIP 0x645
> @@ -264,6 +278,8 @@
> #define CSR_MISA 0x301
> #define CSR_MIE 0x304
> #define CSR_MTVEC 0x305
> +#define CSR_MENVCFG 0x30a
> +#define CSR_MENVCFGH 0x31a
> #define CSR_MSCRATCH 0x340
> #define CSR_MEPC 0x341
> #define CSR_MCAUSE 0x342
> diff --git a/arch/riscv/include/uapi/asm/kvm.h b/arch/riscv/include/uapi/asm/kvm.h
> index 6119368ba6d5..24b2a6e27698 100644
> --- a/arch/riscv/include/uapi/asm/kvm.h
> +++ b/arch/riscv/include/uapi/asm/kvm.h
> @@ -96,6 +96,7 @@ enum KVM_RISCV_ISA_EXT_ID {
> KVM_RISCV_ISA_EXT_H,
> KVM_RISCV_ISA_EXT_I,
> KVM_RISCV_ISA_EXT_M,
> + KVM_RISCV_ISA_EXT_SVPBMT,
> KVM_RISCV_ISA_EXT_MAX,
> };
>
> diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
> index 6dd9cf729614..b7a433c54d0f 100644
> --- a/arch/riscv/kvm/vcpu.c
> +++ b/arch/riscv/kvm/vcpu.c
> @@ -51,6 +51,7 @@ static const unsigned long kvm_isa_ext_arr[] = {
> RISCV_ISA_EXT_h,
> RISCV_ISA_EXT_i,
> RISCV_ISA_EXT_m,
> + RISCV_ISA_EXT_SVPBMT,
> };
>
> static unsigned long kvm_riscv_vcpu_base2isa_ext(unsigned long base_ext)
> @@ -777,6 +778,19 @@ int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
> return -EINVAL;
> }
>
> +static void kvm_riscv_vcpu_update_config(const unsigned long *isa)
> +{
> + u64 henvcfg = 0;
> +
> + if (__riscv_isa_extension_available(isa, RISCV_ISA_EXT_SVPBMT))
> + henvcfg |= ENVCFG_PBMTE;
> +
> + csr_write(CSR_HENVCFG, henvcfg);
> +#ifdef CONFIG_32BIT
> + csr_write(CSR_HENVCFGH, henvcfg >> 32);
> +#endif
> +}
> +
> void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
> {
> struct kvm_vcpu_csr *csr = &vcpu->arch.guest_csr;
> @@ -791,6 +805,8 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
> csr_write(CSR_HVIP, csr->hvip);
> csr_write(CSR_VSATP, csr->vsatp);
>
> + kvm_riscv_vcpu_update_config(vcpu->arch.isa);
> +
> kvm_riscv_gstage_update_hgatp(vcpu);
>
> kvm_riscv_vcpu_timer_restore(vcpu);
> --
> 2.34.1
>
LGTM.
Reviewed-by: Atish Patra <[email protected]>
--
Regards,
Atish
On Thu, Jul 7, 2022 at 7:53 AM Anup Patel <[email protected]> wrote:
>
> The in-kernel AIA IMSIC support requires on-demand mapping / unmapping
> of Guest IMSIC address to Host IMSIC guest files. To help achieve this,
> we add kvm_riscv_stage2_ioremap() and kvm_riscv_stage2_iounmap()
> functions. These new functions for updating G-stage page table mappings
> will be called in atomic context so we have special "in_atomic" parameter
> for this purpose.
>
> Signed-off-by: Anup Patel <[email protected]>
> ---
> arch/riscv/include/asm/kvm_host.h | 5 +++++
> arch/riscv/kvm/mmu.c | 18 ++++++++++++++----
> 2 files changed, 19 insertions(+), 4 deletions(-)
>
> diff --git a/arch/riscv/include/asm/kvm_host.h b/arch/riscv/include/asm/kvm_host.h
> index 59a0cf2ca7b9..60c517e4d576 100644
> --- a/arch/riscv/include/asm/kvm_host.h
> +++ b/arch/riscv/include/asm/kvm_host.h
> @@ -284,6 +284,11 @@ void kvm_riscv_hfence_vvma_gva(struct kvm *kvm,
> void kvm_riscv_hfence_vvma_all(struct kvm *kvm,
> unsigned long hbase, unsigned long hmask);
>
> +int kvm_riscv_gstage_ioremap(struct kvm *kvm, gpa_t gpa,
> + phys_addr_t hpa, unsigned long size,
> + bool writable, bool in_atomic);
> +void kvm_riscv_gstage_iounmap(struct kvm *kvm, gpa_t gpa,
> + unsigned long size);
> int kvm_riscv_gstage_map(struct kvm_vcpu *vcpu,
> struct kvm_memory_slot *memslot,
> gpa_t gpa, unsigned long hva, bool is_write);
> diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
> index b75d4e200064..f7862ca4c4c6 100644
> --- a/arch/riscv/kvm/mmu.c
> +++ b/arch/riscv/kvm/mmu.c
> @@ -343,8 +343,9 @@ static void gstage_wp_memory_region(struct kvm *kvm, int slot)
> kvm_flush_remote_tlbs(kvm);
> }
>
> -static int gstage_ioremap(struct kvm *kvm, gpa_t gpa, phys_addr_t hpa,
> - unsigned long size, bool writable)
> +int kvm_riscv_gstage_ioremap(struct kvm *kvm, gpa_t gpa,
> + phys_addr_t hpa, unsigned long size,
> + bool writable, bool in_atomic)
> {
> pte_t pte;
> int ret = 0;
> @@ -353,6 +354,7 @@ static int gstage_ioremap(struct kvm *kvm, gpa_t gpa, phys_addr_t hpa,
> struct kvm_mmu_memory_cache pcache;
>
> memset(&pcache, 0, sizeof(pcache));
> + pcache.gfp_custom = (in_atomic) ? GFP_ATOMIC | __GFP_ACCOUNT : 0;
> pcache.gfp_zero = __GFP_ZERO;
>
> end = (gpa + size + PAGE_SIZE - 1) & PAGE_MASK;
> @@ -382,6 +384,13 @@ static int gstage_ioremap(struct kvm *kvm, gpa_t gpa, phys_addr_t hpa,
> return ret;
> }
>
> +void kvm_riscv_gstage_iounmap(struct kvm *kvm, gpa_t gpa, unsigned long size)
> +{
> + spin_lock(&kvm->mmu_lock);
> + gstage_unmap_range(kvm, gpa, size, false);
> + spin_unlock(&kvm->mmu_lock);
> +}
> +
> void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
> struct kvm_memory_slot *slot,
> gfn_t gfn_offset,
> @@ -517,8 +526,9 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
> goto out;
> }
>
> - ret = gstage_ioremap(kvm, gpa, pa,
> - vm_end - vm_start, writable);
> + ret = kvm_riscv_gstage_ioremap(kvm, gpa, pa,
> + vm_end - vm_start,
> + writable, false);
> if (ret)
> break;
> }
> --
> 2.34.1
>
Reviewed-by: Atish Patra <[email protected]>
--
Regards,
Atish
On Wed, Jul 13, 2022 at 6:53 AM Atish Patra <[email protected]> wrote:
>
> On Thu, Jul 7, 2022 at 7:53 AM Anup Patel <[email protected]> wrote:
> >
> > When the host has Svpbmt extension, we should use page based memory
> > type 2 (i.e. IO) for IO mappings in the G-stage page table.
> >
> > To achieve this, we replace use of PAGE_KERNEL with PAGE_KERNEL_IO
> > in the kvm_riscv_gstage_ioremap().
> >
> > Signed-off-by: Anup Patel <[email protected]>
> > ---
> > arch/riscv/kvm/mmu.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
> > index f7862ca4c4c6..bc545aef6034 100644
> > --- a/arch/riscv/kvm/mmu.c
> > +++ b/arch/riscv/kvm/mmu.c
> > @@ -361,7 +361,7 @@ int kvm_riscv_gstage_ioremap(struct kvm *kvm, gpa_t gpa,
> > pfn = __phys_to_pfn(hpa);
> >
> > for (addr = gpa; addr < end; addr += PAGE_SIZE) {
> > - pte = pfn_pte(pfn, PAGE_KERNEL);
> > + pte = pfn_pte(pfn, PAGE_KERNEL_IO);
> >
> > if (!writable)
> > pte = pte_wrprotect(pte);
> > --
> > 2.34.1
> >
>
> LGTM.
>
> Reviewed-by: Atish Patra <[email protected]>
Queued this patch for 5.20.
Thanks,
Anup
On Wed, Jul 13, 2022 at 6:54 AM Atish Patra <[email protected]> wrote:
>
> On Thu, Jul 7, 2022 at 7:53 AM Anup Patel <[email protected]> wrote:
> >
> > The Guest/VM can use Svpbmt in VS-stage page tables when allowed by the
> > Hypervisor using the henvcfg.PBMTE bit.
> >
> > We add Svpbmt support for the KVM Guest/VM which can be enabled/disabled
> > by the KVM user-space (QEMU/KVMTOOL) using the ISA extension ONE_REG
> > interface.
> >
> > Signed-off-by: Anup Patel <[email protected]>
> > ---
> > arch/riscv/include/asm/csr.h | 16 ++++++++++++++++
> > arch/riscv/include/uapi/asm/kvm.h | 1 +
> > arch/riscv/kvm/vcpu.c | 16 ++++++++++++++++
> > 3 files changed, 33 insertions(+)
> >
> > diff --git a/arch/riscv/include/asm/csr.h b/arch/riscv/include/asm/csr.h
> > index 6d85655e7edf..17516afc389a 100644
> > --- a/arch/riscv/include/asm/csr.h
> > +++ b/arch/riscv/include/asm/csr.h
> > @@ -156,6 +156,18 @@
> > (_AC(1, UL) << IRQ_S_TIMER) | \
> > (_AC(1, UL) << IRQ_S_EXT))
> >
> > +/* xENVCFG flags */
> > +#define ENVCFG_STCE (_AC(1, ULL) << 63)
> > +#define ENVCFG_PBMTE (_AC(1, ULL) << 62)
> > +#define ENVCFG_CBZE (_AC(1, UL) << 7)
> > +#define ENVCFG_CBCFE (_AC(1, UL) << 6)
> > +#define ENVCFG_CBIE_SHIFT 4
> > +#define ENVCFG_CBIE (_AC(0x3, UL) << ENVCFG_CBIE_SHIFT)
> > +#define ENVCFG_CBIE_ILL _AC(0x0, UL)
> > +#define ENVCFG_CBIE_FLUSH _AC(0x1, UL)
> > +#define ENVCFG_CBIE_INV _AC(0x3, UL)
> > +#define ENVCFG_FIOM _AC(0x1, UL)
> > +
> > /* symbolic CSR names: */
> > #define CSR_CYCLE 0xc00
> > #define CSR_TIME 0xc01
> > @@ -252,7 +264,9 @@
> > #define CSR_HTIMEDELTA 0x605
> > #define CSR_HCOUNTEREN 0x606
> > #define CSR_HGEIE 0x607
> > +#define CSR_HENVCFG 0x60a
> > #define CSR_HTIMEDELTAH 0x615
> > +#define CSR_HENVCFGH 0x61a
> > #define CSR_HTVAL 0x643
> > #define CSR_HIP 0x644
> > #define CSR_HVIP 0x645
> > @@ -264,6 +278,8 @@
> > #define CSR_MISA 0x301
> > #define CSR_MIE 0x304
> > #define CSR_MTVEC 0x305
> > +#define CSR_MENVCFG 0x30a
> > +#define CSR_MENVCFGH 0x31a
> > #define CSR_MSCRATCH 0x340
> > #define CSR_MEPC 0x341
> > #define CSR_MCAUSE 0x342
> > diff --git a/arch/riscv/include/uapi/asm/kvm.h b/arch/riscv/include/uapi/asm/kvm.h
> > index 6119368ba6d5..24b2a6e27698 100644
> > --- a/arch/riscv/include/uapi/asm/kvm.h
> > +++ b/arch/riscv/include/uapi/asm/kvm.h
> > @@ -96,6 +96,7 @@ enum KVM_RISCV_ISA_EXT_ID {
> > KVM_RISCV_ISA_EXT_H,
> > KVM_RISCV_ISA_EXT_I,
> > KVM_RISCV_ISA_EXT_M,
> > + KVM_RISCV_ISA_EXT_SVPBMT,
> > KVM_RISCV_ISA_EXT_MAX,
> > };
> >
> > diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
> > index 6dd9cf729614..b7a433c54d0f 100644
> > --- a/arch/riscv/kvm/vcpu.c
> > +++ b/arch/riscv/kvm/vcpu.c
> > @@ -51,6 +51,7 @@ static const unsigned long kvm_isa_ext_arr[] = {
> > RISCV_ISA_EXT_h,
> > RISCV_ISA_EXT_i,
> > RISCV_ISA_EXT_m,
> > + RISCV_ISA_EXT_SVPBMT,
> > };
> >
> > static unsigned long kvm_riscv_vcpu_base2isa_ext(unsigned long base_ext)
> > @@ -777,6 +778,19 @@ int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
> > return -EINVAL;
> > }
> >
> > +static void kvm_riscv_vcpu_update_config(const unsigned long *isa)
> > +{
> > + u64 henvcfg = 0;
> > +
> > + if (__riscv_isa_extension_available(isa, RISCV_ISA_EXT_SVPBMT))
> > + henvcfg |= ENVCFG_PBMTE;
> > +
> > + csr_write(CSR_HENVCFG, henvcfg);
> > +#ifdef CONFIG_32BIT
> > + csr_write(CSR_HENVCFGH, henvcfg >> 32);
> > +#endif
> > +}
> > +
> > void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
> > {
> > struct kvm_vcpu_csr *csr = &vcpu->arch.guest_csr;
> > @@ -791,6 +805,8 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
> > csr_write(CSR_HVIP, csr->hvip);
> > csr_write(CSR_VSATP, csr->vsatp);
> >
> > + kvm_riscv_vcpu_update_config(vcpu->arch.isa);
> > +
> > kvm_riscv_gstage_update_hgatp(vcpu);
> >
> > kvm_riscv_vcpu_timer_restore(vcpu);
> > --
> > 2.34.1
> >
>
> LGTM.
>
>
> Reviewed-by: Atish Patra <[email protected]>
>
Queued this patch for 5.20.
Thanks,
Anup
Hi Paolo,
On Wed, Jul 13, 2022 at 6:57 AM Atish Patra <[email protected]> wrote:
>
> On Thu, Jul 7, 2022 at 7:53 AM Anup Patel <[email protected]> wrote:
> >
> > The kvm_mmu_topup_memory_cache() always uses GFP_KERNEL_ACCOUNT for
> > memory allocation which prevents it's use in atomic context. To address
> > this limitation of kvm_mmu_topup_memory_cache(), we add gfp_custom flag
> > in struct kvm_mmu_memory_cache. When the gfp_custom flag is set to some
> > GFP_xyz flags, the kvm_mmu_topup_memory_cache() will use that instead of
> > GFP_KERNEL_ACCOUNT.
> >
> > Signed-off-by: Anup Patel <[email protected]>
> > ---
> > include/linux/kvm_types.h | 1 +
> > virt/kvm/kvm_main.c | 4 +++-
> > 2 files changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/include/linux/kvm_types.h b/include/linux/kvm_types.h
> > index ac1ebb37a0ff..1dcfba68076a 100644
> > --- a/include/linux/kvm_types.h
> > +++ b/include/linux/kvm_types.h
> > @@ -87,6 +87,7 @@ struct gfn_to_pfn_cache {
> > struct kvm_mmu_memory_cache {
> > int nobjs;
> > gfp_t gfp_zero;
> > + gfp_t gfp_custom;
> > struct kmem_cache *kmem_cache;
> > void *objects[KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE];
> > };
> > diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> > index a49df8988cd6..e3a6f7647474 100644
> > --- a/virt/kvm/kvm_main.c
> > +++ b/virt/kvm/kvm_main.c
> > @@ -386,7 +386,9 @@ int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min)
> > if (mc->nobjs >= min)
> > return 0;
> > while (mc->nobjs < ARRAY_SIZE(mc->objects)) {
> > - obj = mmu_memory_cache_alloc_obj(mc, GFP_KERNEL_ACCOUNT);
> > + obj = mmu_memory_cache_alloc_obj(mc, (mc->gfp_custom) ?
> > + mc->gfp_custom :
> > + GFP_KERNEL_ACCOUNT);
> > if (!obj)
> > return mc->nobjs >= min ? 0 : -ENOMEM;
> > mc->objects[mc->nobjs++] = obj;
> > --
> > 2.34.1
> >
>
> Acked-by: Atish Patra <[email protected]>
I have queued this patch for 5.20.
Please let me know if you are not okay or need changes in this patch.
Thanks,
Anup
On Wed, Jul 13, 2022 at 6:56 AM Atish Patra <[email protected]> wrote:
>
> On Thu, Jul 7, 2022 at 7:53 AM Anup Patel <[email protected]> wrote:
> >
> > The in-kernel AIA IMSIC support requires on-demand mapping / unmapping
> > of Guest IMSIC address to Host IMSIC guest files. To help achieve this,
> > we add kvm_riscv_stage2_ioremap() and kvm_riscv_stage2_iounmap()
> > functions. These new functions for updating G-stage page table mappings
> > will be called in atomic context so we have special "in_atomic" parameter
> > for this purpose.
> >
> > Signed-off-by: Anup Patel <[email protected]>
> > ---
> > arch/riscv/include/asm/kvm_host.h | 5 +++++
> > arch/riscv/kvm/mmu.c | 18 ++++++++++++++----
> > 2 files changed, 19 insertions(+), 4 deletions(-)
> >
> > diff --git a/arch/riscv/include/asm/kvm_host.h b/arch/riscv/include/asm/kvm_host.h
> > index 59a0cf2ca7b9..60c517e4d576 100644
> > --- a/arch/riscv/include/asm/kvm_host.h
> > +++ b/arch/riscv/include/asm/kvm_host.h
> > @@ -284,6 +284,11 @@ void kvm_riscv_hfence_vvma_gva(struct kvm *kvm,
> > void kvm_riscv_hfence_vvma_all(struct kvm *kvm,
> > unsigned long hbase, unsigned long hmask);
> >
> > +int kvm_riscv_gstage_ioremap(struct kvm *kvm, gpa_t gpa,
> > + phys_addr_t hpa, unsigned long size,
> > + bool writable, bool in_atomic);
> > +void kvm_riscv_gstage_iounmap(struct kvm *kvm, gpa_t gpa,
> > + unsigned long size);
> > int kvm_riscv_gstage_map(struct kvm_vcpu *vcpu,
> > struct kvm_memory_slot *memslot,
> > gpa_t gpa, unsigned long hva, bool is_write);
> > diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
> > index b75d4e200064..f7862ca4c4c6 100644
> > --- a/arch/riscv/kvm/mmu.c
> > +++ b/arch/riscv/kvm/mmu.c
> > @@ -343,8 +343,9 @@ static void gstage_wp_memory_region(struct kvm *kvm, int slot)
> > kvm_flush_remote_tlbs(kvm);
> > }
> >
> > -static int gstage_ioremap(struct kvm *kvm, gpa_t gpa, phys_addr_t hpa,
> > - unsigned long size, bool writable)
> > +int kvm_riscv_gstage_ioremap(struct kvm *kvm, gpa_t gpa,
> > + phys_addr_t hpa, unsigned long size,
> > + bool writable, bool in_atomic)
> > {
> > pte_t pte;
> > int ret = 0;
> > @@ -353,6 +354,7 @@ static int gstage_ioremap(struct kvm *kvm, gpa_t gpa, phys_addr_t hpa,
> > struct kvm_mmu_memory_cache pcache;
> >
> > memset(&pcache, 0, sizeof(pcache));
> > + pcache.gfp_custom = (in_atomic) ? GFP_ATOMIC | __GFP_ACCOUNT : 0;
> > pcache.gfp_zero = __GFP_ZERO;
> >
> > end = (gpa + size + PAGE_SIZE - 1) & PAGE_MASK;
> > @@ -382,6 +384,13 @@ static int gstage_ioremap(struct kvm *kvm, gpa_t gpa, phys_addr_t hpa,
> > return ret;
> > }
> >
> > +void kvm_riscv_gstage_iounmap(struct kvm *kvm, gpa_t gpa, unsigned long size)
> > +{
> > + spin_lock(&kvm->mmu_lock);
> > + gstage_unmap_range(kvm, gpa, size, false);
> > + spin_unlock(&kvm->mmu_lock);
> > +}
> > +
> > void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
> > struct kvm_memory_slot *slot,
> > gfn_t gfn_offset,
> > @@ -517,8 +526,9 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
> > goto out;
> > }
> >
> > - ret = gstage_ioremap(kvm, gpa, pa,
> > - vm_end - vm_start, writable);
> > + ret = kvm_riscv_gstage_ioremap(kvm, gpa, pa,
> > + vm_end - vm_start,
> > + writable, false);
> > if (ret)
> > break;
> > }
> > --
> > 2.34.1
> >
>
> Reviewed-by: Atish Patra <[email protected]>
Queued this patch for 5.20.
Thanks,
Anup