Hi all,
This v2 of the nVHE hypervisor stack unwinder. The previous version was
posted at:
v1: https://lore.kernel.org/lkml/[email protected]/
The series is based on arm64 for-next/core and is also dependent on
kvm-arm64/hyp-stack-guard [1].
The main changes in this version is splitting some patches into more
incremental changes, per Mark Brown
The relevant parts of the previous cover letter are copied below for
convenience.
[1] https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/log/?h=kvm-arm64/hyp-stack-guard
Thanks,
Kalesh
-------
This new version of the unwinder splits the unwinding and dumping
of the stack between the hypervisor and host:
- The hypervisor unwinds its stack and dumps the address entries
into a page shared with the host.
- The host then symnolizes and prints the hyp stacktrace from
the shared page.
The new approach doesn't depend on CONFIG_NVHE_EL2_DEBUG,
and allows dumping hyp stacktraces in prodcution environments
(!CONFIG_NVHE_EL2_DEBUG).
arm64/kernel/stacktrace.c is compiled twice: stacktrace.o for the
host kernel and stacktrace.nvhe.o for the hypervisor. This allows
reusing most of the host unwinding logic in the nVHE hypervisor.
Kalesh Singh (5):
KVM: arm64: Factor out common stack unwinding logic
KVM: arm64: Compile stacktrace.nvhe.o
KVM: arm64: Add hypervisor overflow stack
KVM: arm64: Allocate shared stacktrace pages
KVM: arm64: Unwind and dump nVHE hypervisor stacktrace
arch/arm64/include/asm/kvm_asm.h | 1 +
arch/arm64/include/asm/stacktrace.h | 58 +++++++++--
arch/arm64/kernel/stacktrace.c | 151 +++++++++++++++++++++++-----
arch/arm64/kvm/arm.c | 34 +++++++
arch/arm64/kvm/handle_exit.c | 4 +
arch/arm64/kvm/hyp/nvhe/Makefile | 3 +-
arch/arm64/kvm/hyp/nvhe/host.S | 9 +-
arch/arm64/kvm/hyp/nvhe/setup.c | 11 ++
arch/arm64/kvm/hyp/nvhe/switch.c | 4 +
9 files changed, 231 insertions(+), 44 deletions(-)
--
2.36.0.464.gb9c8b46e94-goog
On hyp_panic(), the hypervisor dumps the addresses for its stacktrace
entries to a page shared with the host. The host then symbolizes and
prints the hyp stacktrace before panicking itself.
Example stacktrace:
[ 122.051187] kvm [380]: Invalid host exception to nVHE hyp!
[ 122.052467] kvm [380]: nVHE HYP call trace:
[ 122.052814] kvm [380]: [<ffff800008f5b550>] __kvm_nvhe___pkvm_vcpu_init_traps+0x1f0/0x1f0
[ 122.053865] kvm [380]: [<ffff800008f560f0>] __kvm_nvhe_hyp_panic+0x130/0x1c0
[ 122.054367] kvm [380]: [<ffff800008f56190>] __kvm_nvhe___kvm_vcpu_run+0x10/0x10
[ 122.054878] kvm [380]: [<ffff800008f57a40>] __kvm_nvhe_handle___kvm_vcpu_run+0x30/0x50
[ 122.055412] kvm [380]: [<ffff800008f57d2c>] __kvm_nvhe_handle_trap+0xbc/0x160
[ 122.055911] kvm [380]: [<ffff800008f56864>] __kvm_nvhe___host_exit+0x64/0x64
[ 122.056417] kvm [380]: ---- end of nVHE HYP call trace ----
Signed-off-by: Kalesh Singh <[email protected]>
Reviewed-by: Mark Brown <[email protected]>
---
Changes in v2:
- Add Mark's Reviewed-by tag
arch/arm64/include/asm/stacktrace.h | 42 ++++++++++++++--
arch/arm64/kernel/stacktrace.c | 75 +++++++++++++++++++++++++++++
arch/arm64/kvm/handle_exit.c | 4 ++
arch/arm64/kvm/hyp/nvhe/switch.c | 4 ++
4 files changed, 121 insertions(+), 4 deletions(-)
diff --git a/arch/arm64/include/asm/stacktrace.h b/arch/arm64/include/asm/stacktrace.h
index f5af9a94c5a6..3063912107b0 100644
--- a/arch/arm64/include/asm/stacktrace.h
+++ b/arch/arm64/include/asm/stacktrace.h
@@ -5,6 +5,7 @@
#ifndef __ASM_STACKTRACE_H
#define __ASM_STACKTRACE_H
+#include <linux/kvm_host.h>
#include <linux/percpu.h>
#include <linux/sched.h>
#include <linux/sched/task_stack.h>
@@ -19,10 +20,12 @@ enum stack_type {
#ifndef __KVM_NVHE_HYPERVISOR__
STACK_TYPE_TASK,
STACK_TYPE_IRQ,
- STACK_TYPE_OVERFLOW,
STACK_TYPE_SDEI_NORMAL,
STACK_TYPE_SDEI_CRITICAL,
+#else /* __KVM_NVHE_HYPERVISOR__ */
+ STACK_TYPE_HYP,
#endif /* !__KVM_NVHE_HYPERVISOR__ */
+ STACK_TYPE_OVERFLOW,
STACK_TYPE_UNKNOWN,
__NR_STACK_TYPES
};
@@ -55,6 +58,9 @@ static inline bool on_stack(unsigned long sp, unsigned long size,
extern void dump_backtrace(struct pt_regs *regs, struct task_struct *tsk,
const char *loglvl);
+extern void hyp_dump_backtrace(unsigned long hyp_offset);
+
+DECLARE_PER_CPU(unsigned long, kvm_arm_hyp_stacktrace_page);
DECLARE_PER_CPU(unsigned long *, irq_stack_ptr);
static inline bool on_irq_stack(unsigned long sp, unsigned long size,
@@ -91,8 +97,32 @@ static inline bool on_overflow_stack(unsigned long sp, unsigned long size,
static inline bool on_overflow_stack(unsigned long sp, unsigned long size,
struct stack_info *info) { return false; }
#endif
-#endif /* !__KVM_NVHE_HYPERVISOR__ */
+#else /* __KVM_NVHE_HYPERVISOR__ */
+
+extern void hyp_save_backtrace(void);
+
+DECLARE_PER_CPU(unsigned long [PAGE_SIZE/sizeof(long)], overflow_stack);
+DECLARE_PER_CPU(struct kvm_nvhe_init_params, kvm_init_params);
+
+static inline bool on_overflow_stack(unsigned long sp, unsigned long size,
+ struct stack_info *info)
+{
+ unsigned long low = (unsigned long)this_cpu_ptr(overflow_stack);
+ unsigned long high = low + PAGE_SIZE;
+
+ return on_stack(sp, size, low, high, STACK_TYPE_OVERFLOW, info);
+}
+
+static inline bool on_hyp_stack(unsigned long sp, unsigned long size,
+ struct stack_info *info)
+{
+ struct kvm_nvhe_init_params *params = this_cpu_ptr(&kvm_init_params);
+ unsigned long high = params->stack_hyp_va;
+ unsigned long low = high - PAGE_SIZE;
+ return on_stack(sp, size, low, high, STACK_TYPE_HYP, info);
+}
+#endif /* !__KVM_NVHE_HYPERVISOR__ */
/*
* We can only safely access per-cpu stacks from current in a non-preemptible
@@ -105,6 +135,9 @@ static inline bool on_accessible_stack(const struct task_struct *tsk,
if (info)
info->type = STACK_TYPE_UNKNOWN;
+ if (on_overflow_stack(sp, size, info))
+ return true;
+
#ifndef __KVM_NVHE_HYPERVISOR__
if (on_task_stack(tsk, sp, size, info))
return true;
@@ -112,10 +145,11 @@ static inline bool on_accessible_stack(const struct task_struct *tsk,
return false;
if (on_irq_stack(sp, size, info))
return true;
- if (on_overflow_stack(sp, size, info))
- return true;
if (on_sdei_stack(sp, size, info))
return true;
+#else /* __KVM_NVHE_HYPERVISOR__ */
+ if (on_hyp_stack(sp, size, info))
+ return true;
#endif /* !__KVM_NVHE_HYPERVISOR__ */
return false;
diff --git a/arch/arm64/kernel/stacktrace.c b/arch/arm64/kernel/stacktrace.c
index f346b4c66f1c..c81dea9760ac 100644
--- a/arch/arm64/kernel/stacktrace.c
+++ b/arch/arm64/kernel/stacktrace.c
@@ -104,6 +104,7 @@ static int notrace __unwind_next(struct task_struct *tsk,
*
* TASK -> IRQ -> OVERFLOW -> SDEI_NORMAL
* TASK -> SDEI_NORMAL -> SDEI_CRITICAL -> OVERFLOW
+ * HYP -> OVERFLOW
*
* ... but the nesting itself is strict. Once we transition from one
* stack to another, it's never valid to unwind back to that first
@@ -242,7 +243,81 @@ noinline notrace void arch_stack_walk(stack_trace_consume_fn consume_entry,
unwind(task, &state, consume_entry, cookie);
}
+
+/**
+ * Symbolizes and dumps the hypervisor backtrace from the shared
+ * stacktrace page.
+ */
+noinline notrace void hyp_dump_backtrace(unsigned long hyp_offset)
+{
+ unsigned long *stacktrace_pos =
+ (unsigned long *)*this_cpu_ptr(&kvm_arm_hyp_stacktrace_page);
+ unsigned long va_mask = GENMASK_ULL(vabits_actual - 1, 0);
+ unsigned long pc = *stacktrace_pos++;
+
+ kvm_err("nVHE HYP call trace:\n");
+
+ while (pc) {
+ pc &= va_mask; /* Mask tags */
+ pc += hyp_offset; /* Convert to kern addr */
+ kvm_err("[<%016lx>] %pB\n", pc, (void *)pc);
+ pc = *stacktrace_pos++;
+ }
+
+ kvm_err("---- end of nVHE HYP call trace ----\n");
+}
#else /* __KVM_NVHE_HYPERVISOR__ */
DEFINE_PER_CPU(unsigned long [PAGE_SIZE/sizeof(long)], overflow_stack)
__aligned(16);
+
+static int notrace unwind_next(struct task_struct *tsk,
+ struct unwind_state *state)
+{
+ struct stack_info info;
+
+ return __unwind_next(tsk, state, &info);
+}
+
+/**
+ * Saves a hypervisor stacktrace entry (address) to the shared stacktrace page.
+ */
+static bool hyp_save_backtrace_entry(void *arg, unsigned long where)
+{
+ struct kvm_nvhe_init_params *params = this_cpu_ptr(&kvm_init_params);
+ unsigned long **stacktrace_pos = (unsigned long **)arg;
+ unsigned long stacktrace_start, stacktrace_end;
+
+ stacktrace_start = (unsigned long)params->stacktrace_hyp_va;
+ stacktrace_end = stacktrace_start + PAGE_SIZE - (2 * sizeof(long));
+
+ if ((unsigned long) *stacktrace_pos > stacktrace_end)
+ return false;
+
+ /* Save the entry to the current pos in stacktrace page */
+ **stacktrace_pos = where;
+
+ /* A zero entry delimits the end of the stacktrace. */
+ *(*stacktrace_pos + 1) = 0UL;
+
+ /* Increment the current pos */
+ ++*stacktrace_pos;
+
+ return true;
+}
+
+/**
+ * Saves hypervisor stacktrace to the shared stacktrace page.
+ */
+noinline notrace void hyp_save_backtrace(void)
+{
+ struct kvm_nvhe_init_params *params = this_cpu_ptr(&kvm_init_params);
+ void *stacktrace_start = (void *)params->stacktrace_hyp_va;
+ struct unwind_state state;
+
+ unwind_init(&state, (unsigned long)__builtin_frame_address(0),
+ _THIS_IP_);
+
+ unwind(NULL, &state, hyp_save_backtrace_entry, &stacktrace_start);
+}
+
#endif /* !__KVM_NVHE_HYPERVISOR__ */
diff --git a/arch/arm64/kvm/handle_exit.c b/arch/arm64/kvm/handle_exit.c
index 2e61a987b0d5..f1a6b556ec32 100644
--- a/arch/arm64/kvm/handle_exit.c
+++ b/arch/arm64/kvm/handle_exit.c
@@ -17,6 +17,7 @@
#include <asm/kvm_emulate.h>
#include <asm/kvm_mmu.h>
#include <asm/debug-monitors.h>
+#include <asm/stacktrace.h>
#include <asm/traps.h>
#include <kvm/arm_hypercalls.h>
@@ -325,6 +326,9 @@ void __noreturn __cold nvhe_hyp_panic_handler(u64 esr, u64 spsr,
(void *)panic_addr);
}
+ /* Dump the hypervisor stacktrace */
+ hyp_dump_backtrace(hyp_offset);
+
/*
* Hyp has panicked and we're going to handle that by panicking the
* kernel. The kernel offset will be revealed in the panic so we're
diff --git a/arch/arm64/kvm/hyp/nvhe/switch.c b/arch/arm64/kvm/hyp/nvhe/switch.c
index 978f1b94fb25..95d810e86c7d 100644
--- a/arch/arm64/kvm/hyp/nvhe/switch.c
+++ b/arch/arm64/kvm/hyp/nvhe/switch.c
@@ -25,6 +25,7 @@
#include <asm/fpsimd.h>
#include <asm/debug-monitors.h>
#include <asm/processor.h>
+#include <asm/stacktrace.h>
#include <nvhe/fixed_config.h>
#include <nvhe/mem_protect.h>
@@ -395,6 +396,9 @@ asmlinkage void __noreturn hyp_panic(void)
__sysreg_restore_state_nvhe(host_ctxt);
}
+ /* Save the hypervisor stacktrace */
+ hyp_save_backtrace();
+
__hyp_do_panic(host_ctxt, spsr, elr, par);
unreachable();
}
--
2.36.0.464.gb9c8b46e94-goog