Skip to contentSkip to conversation

KVM: selftests: Stress save+restore and #PF (ft. nested)

PATCH Superseded Yosry Ahmed 2 participants 22 messages Jun 4 – Jun 11

Subsystems (3)

v2 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v2 cover KVM: selftests: Stress save+restore and #PF (ft. nested)

Add a stress test for save+restore while the guest is triggering and handling #PFs, in both L1 and L2. The goal was to create a generic selftest that would catch bugs like the one fixed by commit 5c247d08bc81 ("KVM: nSVM: Use vcpu->arch.cr2 when updating vmcb12 on nested #VMEXIT"), instead of relying on high-level testing (e.g. building GCC in L2) to catch it.

The test tries to be as generic as possible by triggering #PFs in a guest and installing a proper #PF handler, while the host is continuously doing save+restore cycles. Exiting to userspace is randomly triggered by a second thread that constantly signals the vCPU thread.

Patches (1-6) are prep patches, fixing GPR switching for nSVM and generalizing it to cover nVMX, which is needed for the test to run properly with nVMX. Patch 4 removes HORRIFIC_L2_UCALL_CLOBBER_HACK, as it is no longer needed. While this series does not have the "complete" fix added by commit 6783ca4105a7 ("KVM: selftests: Add a shameful hack to preserve/clobber GPRs across ucall"), it's a good step in the right direction.

Patches (7-10) add the actual test. The test is first introduced as a simple (read: dummy) stress test that just explicitly syncs to userspace after each #PF handling to do save+restore, then gradually evolves to add the random signaling and nested support. After the last patch, the test reliably reproduces the CR2 bug.

This series conflicts with reworking L2 stack allocation in [1], but the conflict should be trivial to fix regardless of which series lands first.

[1]https://lore.kernel.org/kvm/[email protected]/

v1 -> v2:
- Switch guest_regs to an array, which simplifies the offsets calculation and forgoes the dependency on using OFFSET() or defining the struct offsets for assembly otherwise.
- Move page table mapping to the test (instead of a generic helper), as the helper mistakenly tried to map the entire memslot, not just page tables.
- Do not use identity mappings for page tables as it collisions with GVAs used for ELF in some cases.
- Simplify page table walking by using loops.
- Make sure the signals are ignored before creating the signaling thread [Sashiko]
- Assert that the guest actually ran and had page faults [Sashiko]
- Add a patch to fix RAX and RFLAGS offsets in run_guest() [Sashiko]
- Initialize exception_has_payload when injecting a #UD [Sashiko]
- Only check KVM_STATE_NESTED_GUEST_MODE when running in nested mode [Sashiko]

v1: https://lore.kernel.org/all/[email protected]/

Yosry Ahmed (10):
KVM: selftests: Move STR() and XSTR() definitions to test_util.h
KVM: selftests: Fix RAX and RFLAGS VMCB offsets when running L2
KVM: selftests: Use an array for guest_regs (and fix offsets)
KVM: selftests: Move GPR load/save definitions outside of nSVM code
KVM: selftests: Reuse GPR switching logic for nVMX
KVM: selftests: Drop HORRIFIC_L2_UCALL_CLOBBER_HACK
KVM: selftests: Add basic stress test for save+restore and #PF
handling
KVM: selftests: Trigger save+restore randomly in the #PF stress test
KVM: selftests: Support running stress save+restore and #PF test in L2
KVM: selftests: Trigger L2->L1 exits stress save+restore and #PF test

 tools/testing/selftests/kvm/Makefile.kvm      |   1 +
 .../testing/selftests/kvm/include/test_util.h |   3 +
 .../selftests/kvm/include/x86/processor.h     |  55 ++-
 tools/testing/selftests/kvm/include/x86/vmx.h |  63 ++--
 .../testing/selftests/kvm/lib/x86/processor.c |   2 +
 tools/testing/selftests/kvm/lib/x86/svm.c     |  58 +--
 tools/testing/selftests/kvm/lib/x86/ucall.c   |  32 +-
 .../kvm/x86/evmcs_smm_controls_test.c         |   3 -
 tools/testing/selftests/kvm/x86/smm_test.c    |   3 -
 .../kvm/x86/stress_save_restore_pf_test.c     | 333 ++++++++++++++++++
 10 files changed, 435 insertions(+), 118 deletions(-)
 create mode 100644 tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c

base-commit: d4bfaa66fa171089b9b9fb2dc17af9245f2b9b34

Signature · 1 line
2.54.0.1032.g2f8565e1d1-goog
#1 lore raw

PATCH v2 1/10 KVM: selftests: Move STR() and XSTR() definitions to test_util.h

The macros are defined in two tests, and future changes will use them elsewhere. Move their definition into test_util.h to deduplicate them.

No functional change intended.

---
 tools/testing/selftests/kvm/include/test_util.h           | 3 +++
 tools/testing/selftests/kvm/x86/evmcs_smm_controls_test.c | 3 ---
 tools/testing/selftests/kvm/x86/smm_test.c                | 3 ---
 3 files changed, 3 insertions(+), 6 deletions(-)
tools/testing/selftests/kvm/include/test_util.h +3 -0
index a56271c237ae9..c55148ebfe934 100644
@@ -240,4 +240,7 @@ char *strdup_printf(const char *fmt, ...) __attribute__((format(printf, 1, 2), n
 
 char *sys_get_cur_clocksource(void);
 
+#define STR(x) #x
+#define XSTR(x) STR(x)
+
 #endif /* SELFTEST_KVM_TEST_UTIL_H */
tools/testing/selftests/kvm/x86/evmcs_smm_controls_test.c +0 -3
index 5b3aef109cfc5..0c655c48a75d6 100644
@@ -22,9 +22,6 @@
 
 #define SYNC_PORT	0xe
 
-#define STR(x) #x
-#define XSTR(s) STR(s)
-
 /*
  * SMI handler: runs in real-address mode.
  * Reports SMRAM_STAGE via port IO, then does RSM.
tools/testing/selftests/kvm/x86/smm_test.c +0 -3
index 740051167dbd4..c260701d4691b 100644
@@ -22,9 +22,6 @@
 #define SMRAM_GPA 0x1000000
 #define SMRAM_STAGE 0xfe
 
-#define STR(x) #x
-#define XSTR(s) STR(s)
-
 #define SYNC_PORT 0xe
 #define DONE 0xff
 
Signature · 1 line
2.54.0.1032.g2f8565e1d1-goog
#2 lore raw

PATCH v2 2/10 KVM: selftests: Fix RAX and RFLAGS VMCB offsets when running L2

In reply to Yosry Ahmed

The offsets used (0x170 and 0x1f8) are offsets within vmcb_save_area, not vmcb. The correct offsets should include the base of vmcb_save_area within vmcb (which is 0x400 -- so 0x570 and 0x5f8).

Instead of just correcting the offsets, use vmcb->save.rax and vmcb->save.rflags as parameters to the asm block and avoid hardcoding offsets completely. While at it, also use guest_regs.rax directly instead of assuming it's at offset 0 of guest_regs.

Note: "+m" must be used for vmcb_rax and vmcb_rflags, as caching those fields in registers would be wrong as the underlying KVM will update them in memory.

The same problem was recently fixed (differently) for kvm-unit-tests [1].

[1]https://lore.kernel.org/all/[email protected]/

---
 tools/testing/selftests/kvm/lib/x86/svm.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)
tools/testing/selftests/kvm/lib/x86/svm.c +11 -8
index 3b01605ab016c..f9ea1a2aeb8bc 100644
@@ -162,19 +162,22 @@ void run_guest(struct vmcb *vmcb, u64 vmcb_gpa)
 {
 	asm volatile (
 		"vmload %[vmcb_gpa]\n\t"
-		"mov rflags, %%r15\n\t"	// rflags
-		"mov %%r15, 0x170(%[vmcb])\n\t"
-		"mov guest_regs, %%r15\n\t"	// rax
-		"mov %%r15, 0x1f8(%[vmcb])\n\t"
+		"mov rflags, %%r15\n\t"
+		"mov %%r15, %[vmcb_rflags]\n\t"
+		"mov %[guest_regs_rax], %%r15\n\t"
+		"mov %%r15, %[vmcb_rax]\n\t"
 		LOAD_GPR_C
 		"vmrun %[vmcb_gpa]\n\t"
 		SAVE_GPR_C
-		"mov 0x170(%[vmcb]), %%r15\n\t"	// rflags
+		"mov %[vmcb_rflags], %%r15\n\t"
 		"mov %%r15, rflags\n\t"
-		"mov 0x1f8(%[vmcb]), %%r15\n\t"	// rax
-		"mov %%r15, guest_regs\n\t"
+		"mov %[vmcb_rax], %%r15\n\t"	// rax
+		"mov %%r15, %[guest_regs_rax]\n\t"
 		"vmsave %[vmcb_gpa]\n\t"
-		: : [vmcb] "r" (vmcb), [vmcb_gpa] "a" (vmcb_gpa)
+		: [vmcb_rflags] "+m" (vmcb->save.rflags),
+		  [vmcb_rax] "+m" (vmcb->save.rax),
+		  [guest_regs_rax] "+rm" (guest_regs.rax)
+		: [vmcb_gpa] "a" (vmcb_gpa)
 		: "r15", "memory");
 }
 
Signature · 1 line
2.54.0.1032.g2f8565e1d1-goog
#3 lore raw

PATCH v2 3/10 KVM: selftests: Use an array for guest_regs (and fix offsets)

In reply to Yosry Ahmed

The assembly code defined by SAVE_GPR_C uses the wrong offsets for some registers in guest_regs. For example, the offset of RCX should 0x08 not 0x10. Also, the last offset in the struct (R15) is 0x78, not 0x80, so the code actually saves and restore beyond the end of gpr64_regs.

Eliminate hardcoded offsets by using an array instead of a struct (similar to KVM's per-vCPU regs), and use the array index to generate the offset. While at it, rename SAVE_GPR_C and LOAD_GPR_C to a single macro, SVM_SWITCH_GPRS_ASM.

---
 .../selftests/kvm/include/x86/processor.h     | 36 +++++++--------
 tools/testing/selftests/kvm/lib/x86/svm.c     | 45 ++++++++++---------
 2 files changed, 40 insertions(+), 41 deletions(-)
tools/testing/selftests/kvm/include/x86/processor.h +17 -19
index 513e4a1075fac..ff0b270bfe27f 100644
@@ -396,25 +396,23 @@ static inline unsigned int x86_model(unsigned int eax)
 #define PTE_GET_PA(pte)		((pte) & PHYSICAL_PAGE_MASK)
 #define PTE_GET_PFN(pte)        (PTE_GET_PA(pte) >> PAGE_SHIFT)
 
-/* General Registers in 64-Bit Mode */
-struct gpr64_regs {
-	u64 rax;
-	u64 rcx;
-	u64 rdx;
-	u64 rbx;
-	u64 rsp;
-	u64 rbp;
-	u64 rsi;
-	u64 rdi;
-	u64 r8;
-	u64 r9;
-	u64 r10;
-	u64 r11;
-	u64 r12;
-	u64 r13;
-	u64 r14;
-	u64 r15;
-};
+#define GUEST_REGS_RAX	0
+#define GUEST_REGS_RCX	1
+#define GUEST_REGS_RDX	2
+#define GUEST_REGS_RBX	3
+#define GUEST_REGS_RSP	4
+#define GUEST_REGS_RBP	5
+#define GUEST_REGS_RSI	6
+#define GUEST_REGS_RDI	7
+#define GUEST_REGS_R8	8
+#define GUEST_REGS_R9	9
+#define GUEST_REGS_R10	10
+#define GUEST_REGS_R11	11
+#define GUEST_REGS_R12	12
+#define GUEST_REGS_R13	13
+#define GUEST_REGS_R14	14
+#define GUEST_REGS_R15	15
+#define NR_GUEST_REGS	GUEST_REGS_R15
 
 struct desc64 {
 	u16 limit0;
tools/testing/selftests/kvm/lib/x86/svm.c +23 -22

80 lines not on this page · raw message

Signature · 1 line
2.54.0.1032.g2f8565e1d1-goog
#4 lore raw

Re: [PATCH v2 03/10] KVM: selftests: Use an array for guest_regs (and fix offsets)

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] The `guest_regs` array is under-allocated by one element, leading to an out-of-bounds write that…

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] The guest_regs array is under-allocated by one element, leading to an out-of-bounds write that clobbers adjacent memory when swapping the r15 register. --

commit 2b93b9d75a4b05f37bd5b56a9795d649f317160b
Author: Yosry Ahmed <[email protected]>

KVM: selftests: Use an array for guest_regs (and fix offsets)

This commit replaces the gpr64_regs struct with an array to eliminate hardcoded offsets. It also renames the SAVE_GPR_C and LOAD_GPR_C macros to a single SVM_SWITCH_GPRS_ASM macro.

Quoted 4 lines
tools/testing/selftests/kvm/include/x86/processor.h +0 -0
index 513e4a1075fac..ff0b270bfe27f 100644

[ ... ]

tools/testing/selftests/kvm/include/x86/processor.h static inline unsigned int x86_model(unsigned int eax) · line 415

Quoted 4 lines
tools/testing/selftests/kvm/include/x86/processor.h +4 -0
+#define GUEST_REGS_R13	13
+#define GUEST_REGS_R14	14
+#define GUEST_REGS_R15	15
+#define NR_GUEST_REGS	GUEST_REGS_R15

[Severity: Medium]
Does this under-allocate the guest_regs array?

GUEST_REGS_R15 evaluates to 15, making NR_GUEST_REGS 15. However, there are 16 registers in total when counting from index 0 (RAX) through 15 (R15).

tools/testing/selftests/kvm/lib/x86/svm.c line 17

Quoted 13 lines

struct desc64 {

tools/testing/selftests/kvm/lib/x86/svm.c +1 -1
index f9ea1a2aeb8bc..1bd111e8bed26 100644
@@ -13,7 +13,7 @@
 
 #define SEV_DEV_PATH "/dev/sev"
 
-struct gpr64_regs guest_regs;
+u64 guest_regs[NR_GUEST_REGS];
 u64 rflags;

[Severity: Medium]
Since NR_GUEST_REGS is 15, does this array allocation leave out space for the final register?

[ ... ]

Quoted 2 lines
tools/testing/selftests/kvm/lib/x86/svm.c +2 -0
+#define SVM_SWITCH_GPRS_ASM \
+	GUEST_SWITCH_GPR_ASM(rbx, GUEST_REGS_RBX) \

[ ... ]

tools/testing/selftests/kvm/lib/x86/svm.c void generic_svm_setup(struct svm_test_data *svm, void *guest_rip, void *guest_r · line 155

Quoted 2 lines
tools/testing/selftests/kvm/lib/x86/svm.c +2 -0
+	GUEST_SWITCH_GPR_ASM(r14, GUEST_REGS_R14) \
+	GUEST_SWITCH_GPR_ASM(r15, GUEST_REGS_R15)

[Severity: Medium]
Can this write to index 15 clobber adjacent memory?

Because the guest_regs array only has 15 elements, writing the r15 register to GUEST_REGS_R15 (index 15) writes past the end of the array. This looks like it will corrupt adjacent variables in memory, such as the rflags variable defined right below it.

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=3
#5 lore raw

Re: [PATCH v2 03/10] KVM: selftests: Use an array for guest_regs (and fix offsets)

Quoted 55 lines · On Thu, Jun 4, 2026 at 1:44 PM <[email protected]> wrote:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] The guest_regs array is under-allocated by one element, leading to an out-of-bounds write that clobbers adjacent memory when swapping the r15 register. --

commit 2b93b9d75a4b05f37bd5b56a9795d649f317160b
Author: Yosry Ahmed <[email protected]>

KVM: selftests: Use an array for guest_regs (and fix offsets)

This commit replaces the gpr64_regs struct with an array to eliminate hardcoded offsets. It also renames the SAVE_GPR_C and LOAD_GPR_C macros to a single SVM_SWITCH_GPRS_ASM macro.

Quoted 4 lines
tools/testing/selftests/kvm/include/x86/processor.h +0 -0
index 513e4a1075fac..ff0b270bfe27f 100644

[ ... ]

tools/testing/selftests/kvm/include/x86/processor.h static inline unsigned int x86_model(unsigned int eax) · line 415

Quoted 4 lines
tools/testing/selftests/kvm/include/x86/processor.h +4 -0
+#define GUEST_REGS_R13       13
+#define GUEST_REGS_R14       14
+#define GUEST_REGS_R15       15
+#define NR_GUEST_REGS        GUEST_REGS_R15

[Severity: Medium]
Does this under-allocate the guest_regs array?

GUEST_REGS_R15 evaluates to 15, making NR_GUEST_REGS 15. However, there are 16 registers in total when counting from index 0 (RAX) through 15 (R15).

tools/testing/selftests/kvm/lib/x86/svm.c line 17

Quoted 13 lines

struct desc64 {

tools/testing/selftests/kvm/lib/x86/svm.c +1 -1
index f9ea1a2aeb8bc..1bd111e8bed26 100644
@@ -13,7 +13,7 @@
 
 #define SEV_DEV_PATH "/dev/sev"
 
-struct gpr64_regs guest_regs;
+u64 guest_regs[NR_GUEST_REGS];
 u64 rflags;

[Severity: Medium]
Since NR_GUEST_REGS is 15, does this array allocation leave out space for the final register?

[ ... ]

tools/testing/selftests/kvm/lib/x86/svm.c void generic_svm_setup(struct svm_test_data *svm, void *guest_rip, void *guest_r · line 142

Quoted 2 lines
tools/testing/selftests/kvm/lib/x86/svm.c +2 -0
+#define SVM_SWITCH_GPRS_ASM \
+     GUEST_SWITCH_GPR_ASM(rbx, GUEST_REGS_RBX) \
Quoted 7 lines

[ ... ]

tools/testing/selftests/kvm/lib/x86/svm.c void generic_svm_setup(struct svm_test_data *svm, void *guest_rip, void *guest_r · line 155

Quoted 2 lines
tools/testing/selftests/kvm/lib/x86/svm.c +2 -0
+     GUEST_SWITCH_GPR_ASM(r14, GUEST_REGS_R14) \
+     GUEST_SWITCH_GPR_ASM(r15, GUEST_REGS_R15)

[Severity: Medium]
Can this write to index 15 clobber adjacent memory?

It's ironic that the changelog says:
Also, the last offset in the struct (R15) is 0x78, not 0x80, so the code actually saves and restore beyond the end of gpr64_regs.

Yet I reintroduced the same exact bug lol.

Obviously NR_GUEST_REGS should be defined as GUEST_REGS_R15+1.

I also failed to document why an enum cannot be used here, which would have been ideal to avoid such problems. The preprocessor doesn't seem to be able to use enum values, passing an enum value to GUEST_SWITCH_GPR_ASM() causes compilation failures.

#6 lore raw

Re: [PATCH v2 03/10] KVM: selftests: Use an array for guest_regs (and fix offsets)

Quoted 4 lines

It's ironic that the changelog says:
Also, the last offset in the struct (R15) is 0x78, not 0x80, so the code actually saves and restore beyond the end of gpr64_regs.

Quoted 8 lines

Yet I reintroduced the same exact bug lol.

Obviously NR_GUEST_REGS should be defined as GUEST_REGS_R15+1.

I also failed to document why an enum cannot be used here, which would have been ideal to avoid such problems. The preprocessor doesn't seem to be able to use enum values, passing an enum value to GUEST_SWITCH_GPR_ASM() causes compilation failures.

One thing I can do is create macros like __GUEST_REGS_RAX, and then an enum using those values similar to enum kvm_reg, something like this:

#define __GUEST_REGS_RAX 0
...

enum guest_regs {
GUEST_REGS_RAX = __GUEST_REGS_RAX,
...
NR_GUEST_REGS,
};

Although this might be too verbose and the only real benefit is getting NR_GUEST_REGS right.

#7 lore raw

PATCH v2 4/10 KVM: selftests: Move GPR load/save definitions outside of nSVM code

In reply to Yosry Ahmed

In preparation for reusing the code for nVMX tests, move the guest_regs array declaration as well as GPR switching macro to processor.h.

No functional change intended.

---
 tools/testing/selftests/kvm/include/x86/processor.h | 5 +++++
 tools/testing/selftests/kvm/lib/x86/processor.c     | 2 ++
 tools/testing/selftests/kvm/lib/x86/svm.c           | 4 ----
 3 files changed, 7 insertions(+), 4 deletions(-)
tools/testing/selftests/kvm/include/x86/processor.h +5 -0
index ff0b270bfe27f..57f81f0bcc42d 100644
@@ -414,6 +414,11 @@ static inline unsigned int x86_model(unsigned int eax)
 #define GUEST_REGS_R15	15
 #define NR_GUEST_REGS	GUEST_REGS_R15
 
+extern u64 guest_regs[NR_GUEST_REGS];
+
+#define GUEST_SWITCH_GPR_ASM(reg, idx) \
+	"xchg %%" #reg ", guest_regs + 8 *" XSTR(idx) "\n\t"
+
 struct desc64 {
 	u16 limit0;
 	u16 base0;
tools/testing/selftests/kvm/lib/x86/processor.c +2 -0
index 4ca48de7a926d..44a7233d69900 100644
@@ -29,6 +29,8 @@ bool host_cpu_is_amd_compatible;
 bool is_forced_emulation_enabled;
 u64 guest_tsc_khz;
 
+u64 guest_regs[NR_GUEST_REGS];
+
 const char *ex_str(int vector)
 {
 	switch (vector) {
tools/testing/selftests/kvm/lib/x86/svm.c +0 -4
index 1bd111e8bed26..6528a6416c178 100644
@@ -13,7 +13,6 @@
 
 #define SEV_DEV_PATH "/dev/sev"
 
-u64 guest_regs[NR_GUEST_REGS];
 u64 rflags;
 
 /* Allocate memory regions for nested SVM tests.
@@ -131,9 +130,6 @@ void generic_svm_setup(struct svm_test_data *svm, void *guest_rip, void *guest_r
 	}
 }
 
-#define GUEST_SWITCH_GPR_ASM(reg, idx) \
-	"xchg %%" #reg ", guest_regs + 8 *" XSTR(idx) "\n\t"
-
 /*
  * save/restore 64-bit general registers except rax, rip, rsp
  * which are directly handed through the VMCB guest processor state
Signature · 1 line
2.54.0.1032.g2f8565e1d1-goog
#8 lore raw

Re: [PATCH v2 04/10] KVM: selftests: Move GPR load/save definitions outside of nSVM code

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] The `guest_regs` array is incorrectly sized (15 instead of 16), leading to an out-of-bounds…

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] The guest_regs array is incorrectly sized (15 instead of 16), leading to an out-of-bounds memory access when saving/restoring the R15 register. --

commit 776bec433fbb092e9b3d73813e7a34673cb5fbbb
Author: Yosry Ahmed <[email protected]>

KVM: selftests: Move GPR load/save definitions outside of nSVM code

This commit moves the guest_regs array declaration and the GPR switching macro to processor.h. This is done in preparation for reusing the code for nVMX tests.

tools/testing/selftests/kvm/include/x86/processor.h static inline unsigned int x86_model(unsigned int eax) · line 420

Quoted 12 lines
tools/testing/selftests/kvm/include/x86/processor.h +4 -0
index ff0b270bfe27f..57f81f0bcc42d 100644
@@ -414,6 +414,11 @@ static inline unsigned int x86_model(unsigned int eax)
 #define GUEST_REGS_R15	15
 #define NR_GUEST_REGS	GUEST_REGS_R15
 
+extern u64 guest_regs[NR_GUEST_REGS];
+
+#define GUEST_SWITCH_GPR_ASM(reg, idx) \
+	"xchg %%" #reg ", guest_regs + 8 *" XSTR(idx) "\n\t"

[Severity: Medium]
Since NR_GUEST_REGS is defined as GUEST_REGS_R15 (15), does this allocate guest_regs with only 15 elements (indices 0 to 14)?

If so, won't saving and restoring the R15 register use index 15, which evaluates to guest_regs + 120, resulting in an 8-byte out-of-bounds read/write?

This off-by-one error was introduced earlier in the patch series when replacing the 16-element gpr64_regs struct with an array. Moving guest_regs to processor.c in this patch alters the memory layout. Previously, the out-of-bounds access silently corrupted the adjacently declared rflags variable (which was safely overwritten immediately afterward), but now it will silently corrupt whichever variable immediately follows guest_regs in the .bss section.

Quoted 4 lines
tools/testing/selftests/kvm/lib/x86/processor.c +0 -0
index 4ca48de7a926d..44a7233d69900 100644

[ ... ]

Quoted 12 lines
tools/testing/selftests/kvm/lib/x86/processor.c +2 -0
@@ -29,6 +29,8 @@ bool host_cpu_is_amd_compatible;
 bool is_forced_emulation_enabled;
 u64 guest_tsc_khz;
 
+u64 guest_regs[NR_GUEST_REGS];
+
 const char *ex_str(int vector)
 
tools/testing/selftests/kvm/lib/x86/svm.c +0 -0
index 1bd111e8bed26..6528a6416c178 100644

[ ... ]

Quoted 8 lines
tools/testing/selftests/kvm/lib/x86/svm.c +0 -1
@@ -13,7 +13,6 @@
 
 #define SEV_DEV_PATH "/dev/sev"
 
-u64 guest_regs[NR_GUEST_REGS];
 u64 rflags;
 
 /* Allocate memory regions for nested SVM tests.
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=4
#9 lore raw

PATCH v2 5/10 KVM: selftests: Reuse GPR switching logic for nVMX

In reply to Yosry Ahmed

Reuse the GPR switching logic for nVMX by defining VMX_SWITCH_GPRS_ASM, which is essentially the same as SVM_SWITCH_GPRS_ASM but also switches RAX, replacing the push/pop of a subset of the registers.

The long clobber list of registers is no longer needed as registers are saved and restored appropriately (and not clobbered by L2).

Assisted-by Gemini:gemini-3.1-pro
---
 tools/testing/selftests/kvm/include/x86/vmx.h | 63 +++++++++----------
 1 file changed, 29 insertions(+), 34 deletions(-)
tools/testing/selftests/kvm/include/x86/vmx.h +29 -34
index 90fffaf915958..64996c310fe4c 100644
@@ -363,9 +363,23 @@ static inline u64 vmptrstz(void)
 	return value;
 }
 
-/*
- * No guest state (e.g. GPRs) is established by this vmlaunch.
- */
+#define VMX_SWITCH_GPRS_ASM \
+	GUEST_SWITCH_GPR_ASM(rax, GUEST_REGS_RAX) \
+	GUEST_SWITCH_GPR_ASM(rbx, GUEST_REGS_RBX) \
+	GUEST_SWITCH_GPR_ASM(rcx, GUEST_REGS_RCX) \
+	GUEST_SWITCH_GPR_ASM(rdx, GUEST_REGS_RDX) \
+	GUEST_SWITCH_GPR_ASM(rbp, GUEST_REGS_RBP) \
+	GUEST_SWITCH_GPR_ASM(rsi, GUEST_REGS_RSI) \
+	GUEST_SWITCH_GPR_ASM(rdi, GUEST_REGS_RDI) \
+	GUEST_SWITCH_GPR_ASM(r8,  GUEST_REGS_R8)  \
+	GUEST_SWITCH_GPR_ASM(r9,  GUEST_REGS_R9)  \
+	GUEST_SWITCH_GPR_ASM(r10, GUEST_REGS_R10) \
+	GUEST_SWITCH_GPR_ASM(r11, GUEST_REGS_R11) \
+	GUEST_SWITCH_GPR_ASM(r12, GUEST_REGS_R12) \
+	GUEST_SWITCH_GPR_ASM(r13, GUEST_REGS_R13) \
+	GUEST_SWITCH_GPR_ASM(r14, GUEST_REGS_R14) \
+	GUEST_SWITCH_GPR_ASM(r15, GUEST_REGS_R15)
+
 static inline int vmlaunch(void)
 {
 	int ret;
@@ -373,34 +387,23 @@ static inline int vmlaunch(void)
 	if (enable_evmcs)
 		return evmcs_vmlaunch();
 
-	__asm__ __volatile__("push %%rbp;"
-			     "push %%rcx;"
-			     "push %%rdx;"
-			     "push %%rsi;"
-			     "push %%rdi;"
-			     "push $0;"
+	__asm__ __volatile__("push $0;"
 			     "vmwrite %%rsp, %[host_rsp];"
 			     "lea 1f(%%rip), %%rax;"
 			     "vmwrite %%rax, %[host_rip];"
+			     VMX_SWITCH_GPRS_ASM
 			     "vmlaunch;"
 			     "incq (%%rsp);"
-			     "1: pop %%rax;"
-			     "pop %%rdi;"
-			     "pop %%rsi;"
-			     "pop %%rdx;"
-			     "pop %%rcx;"
-			     "pop %%rbp;"
+			     "1: ;"
+			     VMX_SWITCH_GPRS_ASM
+			     "pop %%rax;"
 			     : [ret]"=&a"(ret)
 			     : [host_rsp]"r"((u64)HOST_RSP),
 			       [host_rip]"r"((u64)HOST_RIP)
-			     : "memory", "cc", "rbx", "r8", "r9", "r10",
-			       "r11", "r12", "r13", "r14", "r15");
+			     : "memory", "cc");
 	return ret;
 }
 
-/*
- * No guest state (e.g. GPRs) is established by this vmresume.
- */
 static inline int vmresume(void)
 {
 	int ret;
@@ -408,28 +411,20 @@ static inline int vmresume(void)
 	if (enable_evmcs)
 		return evmcs_vmresume();
 
-	__asm__ __volatile__("push %%rbp;"
-			     "push %%rcx;"
-			     "push %%rdx;"
-			     "push %%rsi;"
-			     "push %%rdi;"
-			     "push $0;"
+	__asm__ __volatile__("push $0;"
 			     "vmwrite %%rsp, %[host_rsp];"
 			     "lea 1f(%%rip), %%rax;"
 			     "vmwrite %%rax, %[host_rip];"
+			     VMX_SWITCH_GPRS_ASM
 			     "vmresume;"
 			     "incq (%%rsp);"
-			     "1: pop %%rax;"
-			     "pop %%rdi;"
-			     "pop %%rsi;"
-			     "pop %%rdx;"
-			     "pop %%rcx;"
-			     "pop %%rbp;"
+			     "1: ;"
+			     VMX_SWITCH_GPRS_ASM
+			     "pop %%rax;"
 			     : [ret]"=&a"(ret)
 			     : [host_rsp]"r"((u64)HOST_RSP),
 			       [host_rip]"r"((u64)HOST_RIP)
-			     : "memory", "cc", "rbx", "r8", "r9", "r10",
-			       "r11", "r12", "r13", "r14", "r15");
+			     : "memory", "cc");
 	return ret;
 }
 
Signature · 1 line
2.54.0.1032.g2f8565e1d1-goog
#10 lore raw

Re: [PATCH v2 05/10] KVM: selftests: Reuse GPR switching logic for nVMX

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider: - [High] The `guest_regs` array is sized for 15 elements, but 16 elements are accessed, causing an…

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] The guest_regs array is sized for 15 elements, but 16 elements are accessed, causing an out-of-bounds read/write to the .bss section.
- [Medium] vmlaunch() and vmresume() behavior for guest GPRs diverges based on the enable_evmcs flag, silently failing tests that rely on the new GPR switching logic under Enlightened VMCS. --

commit 496a5916183241f499c38e4f427dd5cbecd4dc37
Author: Yosry Ahmed <[email protected]>

KVM: selftests: Reuse GPR switching logic for nVMX

Reuse the GPR switching logic for nVMX by defining VMX_SWITCH_GPRS_ASM, which is essentially the same as SVM_SWITCH_GPRS_ASM but also switches RAX, replacing the push/pop of a subset of the registers.

The long clobber list of registers is no longer needed as registers are saved and restored appropriately (and not clobbered by L2).

Quoted 17 lines
tools/testing/selftests/kvm/include/x86/vmx.h +6 -3
index 90fffaf915958..64996c310fe4c 100644
@@ -363,9 +363,23 @@ static inline u64 vmptrstz(void)
 	return value;
 }
 
-/*
- * No guest state (e.g. GPRs) is established by this vmlaunch.
- */
+#define VMX_SWITCH_GPRS_ASM \
+	GUEST_SWITCH_GPR_ASM(rax, GUEST_REGS_RAX) \
+	GUEST_SWITCH_GPR_ASM(rbx, GUEST_REGS_RBX) \
+	GUEST_SWITCH_GPR_ASM(rcx, GUEST_REGS_RCX) \
+	GUEST_SWITCH_GPR_ASM(rdx, GUEST_REGS_RDX) \
+	GUEST_SWITCH_GPR_ASM(rbp, GUEST_REGS_RBP) \

tools/testing/selftests/kvm/include/x86/vmx.h static inline u64 vmptrstz(void) · line 381

Quoted 10 lines
tools/testing/selftests/kvm/include/x86/vmx.h +10 -0
+	GUEST_SWITCH_GPR_ASM(rsi, GUEST_REGS_RSI) \
+	GUEST_SWITCH_GPR_ASM(rdi, GUEST_REGS_RDI) \
+	GUEST_SWITCH_GPR_ASM(r8,  GUEST_REGS_R8)  \
+	GUEST_SWITCH_GPR_ASM(r9,  GUEST_REGS_R9)  \
+	GUEST_SWITCH_GPR_ASM(r10, GUEST_REGS_R10) \
+	GUEST_SWITCH_GPR_ASM(r11, GUEST_REGS_R11) \
+	GUEST_SWITCH_GPR_ASM(r12, GUEST_REGS_R12) \
+	GUEST_SWITCH_GPR_ASM(r13, GUEST_REGS_R13) \
+	GUEST_SWITCH_GPR_ASM(r14, GUEST_REGS_R14) \
+	GUEST_SWITCH_GPR_ASM(r15, GUEST_REGS_R15)

[Severity: High]
Does this access past the end of the guest_regs array?

GUEST_REGS_R15 evaluates to 15, and the array is defined in tools/testing/selftests/kvm/include/x86/processor.h as:

#define NR_GUEST_REGS GUEST_REGS_R15
extern u64 guest_regs[NR_GUEST_REGS];

Since NR_GUEST_REGS is 15, the array only holds 15 elements (indices 0 to 14). Calling GUEST_SWITCH_GPR_ASM(r15, GUEST_REGS_R15) will read and write the 16th element at index 15, resulting in an out-of-bounds access that corrupts memory in the .bss section.

tools/testing/selftests/kvm/include/x86/vmx.h static inline int vmlaunch(void) · line 388

Quoted 7 lines
tools/testing/selftests/kvm/include/x86/vmx.h +1 -0
+
 static inline int vmlaunch(void)
 {
 	int ret;
@@ -373,34 +387,23 @@ static inline int vmlaunch(void)
 	if (enable_evmcs)
 		return evmcs_vmlaunch();

[Severity: Medium]
Will this break EVMCS test configurations?

If enable_evmcs is true, the code redirects to evmcs_vmlaunch() which wasn't updated to use VMX_SWITCH_GPRS_ASM. It still uses the old inline assembly that completely ignores guest_regs:

tools/testing/selftests/kvm/include/x86/evmcs.h:evmcs_vmlaunch() {
    ...
    __asm__ __volatile__("push %%rbp;"
                 "push %%rcx;"
                 "push %%rdx;"
                 "push %%rsi;"
                 "push %%rdi;"
                 "push $0;"
    ...
}

Any KVM selftest that relies on guest_regs to establish L2 state before VM entry will now silently execute with uninitialized L2 GPRs when run with EVMCS enabled.

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=5
#11 lore raw

PATCH v2 6/10 KVM: selftests: Drop HORRIFIC_L2_UCALL_CLOBBER_HACK

In reply to Yosry Ahmed

Now that nVMX test codes preserves GPRs across nested VM-Exits (specifically RBP, RDX, and RDI among others), drop the ucall-specific hack to avoid clobbering these registers.

Assisted-by Gemini:gemini-3.1-pro
---
 tools/testing/selftests/kvm/lib/x86/ucall.c | 32 ++-------------------
 1 file changed, 2 insertions(+), 30 deletions(-)
tools/testing/selftests/kvm/lib/x86/ucall.c +2 -30
index e7dd5791959ba..38050c60a0670 100644
@@ -10,36 +10,8 @@
 
 void ucall_arch_do_ucall(gva_t uc)
 {
-	/*
-	 * FIXME: Revert this hack (the entire commit that added it) once nVMX
-	 * preserves L2 GPRs across a nested VM-Exit.  If a ucall from L2, e.g.
-	 * to do a GUEST_SYNC(), lands the vCPU in L1, any and all GPRs can be
-	 * clobbered by L1.  Save and restore non-volatile GPRs (clobbering RBP
-	 * in particular is problematic) along with RDX and RDI (which are
-	 * inputs), and clobber volatile GPRs. *sigh*
-	 */
-#define HORRIFIC_L2_UCALL_CLOBBER_HACK	\
-	"rcx", "rsi", "r8", "r9", "r10", "r11"
-
-	asm volatile("push %%rbp\n\t"
-		     "push %%r15\n\t"
-		     "push %%r14\n\t"
-		     "push %%r13\n\t"
-		     "push %%r12\n\t"
-		     "push %%rbx\n\t"
-		     "push %%rdx\n\t"
-		     "push %%rdi\n\t"
-		     "in %[port], %%al\n\t"
-		     "pop %%rdi\n\t"
-		     "pop %%rdx\n\t"
-		     "pop %%rbx\n\t"
-		     "pop %%r12\n\t"
-		     "pop %%r13\n\t"
-		     "pop %%r14\n\t"
-		     "pop %%r15\n\t"
-		     "pop %%rbp\n\t"
-		: : [port] "d" (UCALL_PIO_PORT), "D" (uc) : "rax", "memory",
-		     HORRIFIC_L2_UCALL_CLOBBER_HACK);
+	asm volatile("in %[port], %%al"
+		: : [port] "d" (UCALL_PIO_PORT), "D" (uc) : "rax", "memory");
 }
 
 void *ucall_arch_get_ucall(struct kvm_vcpu *vcpu)
Signature · 1 line
2.54.0.1032.g2f8565e1d1-goog
#12 lore raw

Re: [PATCH v2 06/10] KVM: selftests: Drop HORRIFIC_L2_UCALL_CLOBBER_HACK

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] Removal of the ucall GPR hack causes L2 register corruption in EVMCS nested virtualization…

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Removal of the ucall GPR hack causes L2 register corruption in EVMCS nested virtualization tests. --

commit 15ef07c71219fb9256d8d4da6427386aaf993fbd
Author: Yosry Ahmed <[email protected]>
KVM: selftests: Drop HORRIFIC_L2_UCALL_CLOBBER_HACK

This commit drops the HORRIFIC_L2_UCALL_CLOBBER_HACK from ucall_arch_do_ucall(), relying on the nVMX test code to preserve L2 GPRs across nested VM-exits.

Quoted 8 lines
tools/testing/selftests/kvm/lib/x86/ucall.c +0 -0
index e7dd5791959ba..38050c60a0670 100644
@@ -10,36 +10,8 @@
 
 void ucall_arch_do_ucall(gva_t uc)
 {

[ ... ]

tools/testing/selftests/kvm/lib/x86/ucall.c line 14

Quoted 3 lines
tools/testing/selftests/kvm/lib/x86/ucall.c +2 -1
-		     HORRIFIC_L2_UCALL_CLOBBER_HACK);
+	asm volatile("in %[port], %%al"
+		: : [port] "d" (UCALL_PIO_PORT), "D" (uc) : "rax", "memory");

[Severity: Medium]
Does the removal of this hack cause L2 register corruption in Enlightened VMX (EVMCS) tests?

While standard nVMX tests preserve GPRs using VMX_SWITCH_GPRS_ASM, it appears the EVMCS helpers evmcs_vmlaunch() and evmcs_vmresume() in tools/testing/selftests/kvm/include/x86/evmcs.h were not updated.

These functions still use a hardcoded inline assembly sequence that explicitly clobbers %rbx and %r8-%r15.

If L2 triggers a ucall and an asynchronous event forces a nested VM-Exit to L1, wouldn't the L1 compiler clobber L2's GPRs, causing L2 to resume with corrupted registers and leading to a test failure?

Quoted 3 lines

}

void *ucall_arch_get_ucall(struct kvm_vcpu *vcpu)

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=6
#13 lore raw

Re: [PATCH v2 06/10] KVM: selftests: Drop HORRIFIC_L2_UCALL_CLOBBER_HACK

Quoted 21 lines · On Thu, Jun 4, 2026 at 1:50 PM <[email protected]> wrote:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Removal of the ucall GPR hack causes L2 register corruption in EVMCS nested virtualization tests. --

commit 15ef07c71219fb9256d8d4da6427386aaf993fbd
Author: Yosry Ahmed <[email protected]>
KVM: selftests: Drop HORRIFIC_L2_UCALL_CLOBBER_HACK

This commit drops the HORRIFIC_L2_UCALL_CLOBBER_HACK from ucall_arch_do_ucall(), relying on the nVMX test code to preserve L2 GPRs across nested VM-exits.

Quoted 8 lines
tools/testing/selftests/kvm/lib/x86/ucall.c +0 -0
index e7dd5791959ba..38050c60a0670 100644
@@ -10,36 +10,8 @@
 
 void ucall_arch_do_ucall(gva_t uc)
 {

[ ... ]

Quoted 7 lines

tools/testing/selftests/kvm/lib/x86/ucall.c line 14

Quoted 3 lines
tools/testing/selftests/kvm/lib/x86/ucall.c +2 -1
-                  HORRIFIC_L2_UCALL_CLOBBER_HACK);
+     asm volatile("in %[port], %%al"
+             : : [port] "d" (UCALL_PIO_PORT), "D" (uc) : "rax", "memory");

[Severity: Medium]
Does the removal of this hack cause L2 register corruption in Enlightened VMX (EVMCS) tests?

This is interesting because:
- Yes, I did forget to handle evmcs_vmlaunch() and evmcs_vmresume().
- Yes, HORRIFIC_L2_UCALL_CLOBBER_HACK was introduced specifically due to a failure in hyperv_evmcs (according to the changelog of 6783ca4105a7).
- Yes, I did run hyperv_evmcs before sending this patch

...but NO, the test did not fail.

I guess it depends on what registers the compiler ends up using and how the code is compiled? Not sure.

Looking at the code, it actually seems evmcs_vmlaunch() and evmcs_vmresume() properly save and restore RBP, RDX, and RDI (the registers mentioned with HORRIFIC_L2_UCALL_CLOBBER_HACK), so I don't understand how HORRIFIC_L2_UCALL_CLOBBER_HACK ever made a difference.

All that being said, yes, I need to fix evmcs_vmlaunch() and evmcs_vmresume().

Quoted 18 lines

While standard nVMX tests preserve GPRs using VMX_SWITCH_GPRS_ASM, it appears the EVMCS helpers evmcs_vmlaunch() and evmcs_vmresume() in tools/testing/selftests/kvm/include/x86/evmcs.h were not updated.

These functions still use a hardcoded inline assembly sequence that explicitly clobbers %rbx and %r8-%r15.

If L2 triggers a ucall and an asynchronous event forces a nested VM-Exit to L1, wouldn't the L1 compiler clobber L2's GPRs, causing L2 to resume with corrupted registers and leading to a test failure?

Quoted 3 lines

}

void *ucall_arch_get_ucall(struct kvm_vcpu *vcpu)

--
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=6

#14 lore raw

PATCH v2 7/10 KVM: selftests: Add basic stress test for save+restore and #PF handling

In reply to Yosry Ahmed

Add a basic stress test for handling #PFs in a guest while the host is doing save+restore cycles. The guest periodically accesses non-present memory causing a #PF, and the #PF handler walks the page tables and updates the PTE to be present, like a proper #PF handler.

After every access (and #PF), the guest triggers a sync and the test performs save+restore of the VM. This is not very meaningful as save+restore are performed after the access and #PF handling complete, but following changes will change that.

Assisted-by Gemini:gemini-3.1-pro
---
 tools/testing/selftests/kvm/Makefile.kvm      |   1 +
 .../selftests/kvm/include/x86/processor.h     |  14 ++
 .../kvm/x86/stress_save_restore_pf_test.c     | 182 ++++++++++++++++++
 3 files changed, 197 insertions(+)
 create mode 100644 tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c
tools/testing/selftests/kvm/Makefile.kvm +1 -0
index e0ddd3ff9472d..731ef7de1afb4 100644
@@ -111,6 +111,7 @@ TEST_GEN_PROGS_x86 += x86/set_sregs_test
 TEST_GEN_PROGS_x86 += x86/smaller_maxphyaddr_emulation_test
 TEST_GEN_PROGS_x86 += x86/smm_test
 TEST_GEN_PROGS_x86 += x86/state_test
+TEST_GEN_PROGS_x86 += x86/stress_save_restore_pf_test
 TEST_GEN_PROGS_x86 += x86/vmx_preemption_timer_test
 TEST_GEN_PROGS_x86 += x86/svm_vmcall_test
 TEST_GEN_PROGS_x86 += x86/svm_int_ctl_test
tools/testing/selftests/kvm/include/x86/processor.h +14 -0
index 57f81f0bcc42d..10d26e079a1a8 100644
@@ -585,6 +585,15 @@ static inline void set_cr0(u64 val)
 	__asm__ __volatile__("mov %0, %%cr0" : : "r" (val) : "memory");
 }
 
+static inline u64 get_cr2(void)
+{
+	u64 cr2;
+
+	__asm__ __volatile__("mov %%cr2, %[cr2]"
+			     : /* output */ [cr2]"=r"(cr2));
+	return cr2;
+}
+
 static inline u64 get_cr3(void)
 {
 	u64 cr3;
@@ -880,6 +889,11 @@ static inline void write_sse_reg(int reg, const sse128_t *data)
 	}
 }
 
+static inline void invlpg(u64 addr)
+{
+	__asm__ __volatile__("invlpg (%0)" : : "r"(addr) : "memory");
+}
+
 static inline void cpu_relax(void)
 {
 	asm volatile("rep; nop" ::: "memory");
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c new file +182 -0

182 lines not on this page · raw message

Signature · 1 line
2.54.0.1032.g2f8565e1d1-goog
#15 lore raw

Re: [PATCH v2 07/10] KVM: selftests: Add basic stress test for save+restore and #PF handling

tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c line 165

Quoted 12 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c new file +12 -0
+       while (count++ < NR_ITERATIONS) {
+               r = __vcpu_run(vcpu);
+               TEST_ASSERT(!r, "vcpu_run failed");
+               TEST_ASSERT_KVM_EXIT_REASON(vcpu, KVM_EXIT_IO);
+
+               get_ucall(vcpu, &uc);
+               if (uc.cmd == UCALL_ABORT) {
+                       REPORT_GUEST_ASSERT(uc);
+                       break;
+               }
+               TEST_ASSERT_EQ(uc.cmd, UCALL_SYNC);
+               TEST_ASSERT_EQ(uc.args[1], count - 1);

This is off-by-one, it should be 'count', otherwise the assertion always fails. I made changes between versions and did not run the test in between changes. The assertion goes away in the next patch so I didn't notice.

Quoted 20 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c new file +17 -1
+
+               state = vcpu_save_state(vcpu);
+
+               kvm_vm_release(vm);
+               vcpu = vm_recreate_with_one_vcpu(vm);
+               vcpu_load_state(vcpu, state);
+               kvm_x86_state_cleanup(state);
+
+               PRINT_ITER("Save+restore iterations: ", count);
+       }
+
+       sync_global_from_guest(vm, guest_faults);
+       pr_info("Guest page faults: %lu\n", guest_faults);
+
+       kvm_vm_free(vm);
+       return 0;
+}
--

2.54.0.1032.g2f8565e1d1-goog

#16 lore raw

PATCH v2 8/10 KVM: selftests: Trigger save+restore randomly in the #PF stress test

In reply to Yosry Ahmed

Instead of an explicit GUEST_SYNC() after each access+#PF, run another thread that keeps sending SIGUSR to the vCPU thread, essentially triggering exits to userspace and save+restore on random points in guest execution. This makes the test a lot more meaningful as it opens the door to exercising race conditions between #PF handling in the guest and save+restore in the host.

The signals are ignored using SIG_IGN outside of __vcpu_run() to avoid interrupting other ioctls/sysctls performed by the test.

Assisted-by Gemini:gemini-3.1-pro
---
 .../kvm/x86/stress_save_restore_pf_test.c     | 59 ++++++++++++++++---
 1 file changed, 51 insertions(+), 8 deletions(-)
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +51 -8
index 622d102179e66..92efb7ac9d95f 100644
@@ -5,6 +5,8 @@
 #include <errno.h>
 #include <sys/types.h>
 #include <time.h>
+#include <pthread.h>
+#include <signal.h>
 #include <unistd.h>
 
 #include "test_util.h"
@@ -94,15 +96,41 @@ static void guest_access_memory(void *arg)
 		/* Clear the present bit again so it faults next time */
 		*guest_get_pte(vaddr) &= ~pte_present_mask;
 		invlpg(vaddr);
+	}
+}
+
+static void *sigusr_thread_fn(void *arg)
+{
+	pthread_t vcpu_thread = (pthread_t)arg;
 
-		GUEST_SYNC(guest_faults);
+	for (;;) {
+		pthread_testcancel();
+		pthread_kill(vcpu_thread, SIGUSR1);
+		usleep(100);
 	}
+	return NULL;
+}
+
+static void dummy_signal_handler(int signo) {}
+static struct sigaction sa;
+
+static void vcpu_sigusr_listen(void)
+{
+	sa.sa_handler = dummy_signal_handler;
+	sigaction(SIGUSR1, &sa, NULL);
+}
+
+static void vcpu_sigusr_ignore(void)
+{
+	sa.sa_handler = SIG_IGN;
+	sigaction(SIGUSR1, &sa, NULL);
 }
 
 int main(int argc, char *argv[])
 {
 	struct kvm_x86_state *state;
 	int r, i, level, count = 0;
+	pthread_t sigusr_thread;
 	gpa_t gpa, pgtable_gpa;
 	struct kvm_vcpu *vcpu;
 	struct kvm_vm *vm;
@@ -151,18 +179,30 @@ int main(int argc, char *argv[])
 		pgtable_gpa = PTE_GET_PA(pte);
 	}
 
+	/* Initialize the thread sending SIGUSR and install the handler */
+	vcpu_sigusr_ignore();
+	r = pthread_create(&sigusr_thread, NULL, sigusr_thread_fn,
+			   (void *)pthread_self());
+	TEST_ASSERT(!r, "pthread_create() failed: %d", r);
+
 	while (count++ < NR_ITERATIONS) {
+		/*
+		 * Only handle SIGUSR while the vCPU is running, otherwise
+		 * ignore it to avoid interrupting other ioctls/syscalls.
+		 */
+		vcpu_sigusr_listen();
 		r = __vcpu_run(vcpu);
-		TEST_ASSERT(!r, "vcpu_run failed");
-		TEST_ASSERT_KVM_EXIT_REASON(vcpu, KVM_EXIT_IO);
-
-		get_ucall(vcpu, &uc);
-		if (uc.cmd == UCALL_ABORT) {
+		if (r == -1)
+			TEST_ASSERT_EQ(errno, EINTR);
+		vcpu_sigusr_ignore();
+
+		/* The guest only exits due to a signal or failed assertion */
+		if (!r) {
+			TEST_ASSERT_KVM_EXIT_REASON(vcpu, KVM_EXIT_IO);
+			TEST_ASSERT_EQ(get_ucall(vcpu, &uc), UCALL_ABORT);
 			REPORT_GUEST_ASSERT(uc);
 			break;
 		}
-		TEST_ASSERT_EQ(uc.cmd, UCALL_SYNC);
-		TEST_ASSERT_EQ(uc.args[1], count - 1);
 
 		state = vcpu_save_state(vcpu);
 
@@ -175,8 +215,11 @@ int main(int argc, char *argv[])
 	}
 
 	sync_global_from_guest(vm, guest_faults);
+	TEST_ASSERT(guest_faults > 0, "No guest page faults triggered");
 	pr_info("Guest page faults: %lu\n", guest_faults);
 
+	pthread_cancel(sigusr_thread);
+	pthread_join(sigusr_thread, NULL);
 	kvm_vm_free(vm);
 	return 0;
 }
Signature · 1 line
2.54.0.1032.g2f8565e1d1-goog
#17 lore raw

Re: [PATCH v2 08/10] KVM: selftests: Trigger save+restore randomly in the #PF stress test

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] Aggressive 100us signal interval causes test livelock and spurious failures on slower hosts. --

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Aggressive 100us signal interval causes test livelock and spurious failures on slower hosts. --

commit bf3ad650cf1e2ff2e1e4bd996b52e985a3657486
Author: Yosry Ahmed <[email protected]>

KVM: selftests: Trigger save+restore randomly in the #PF stress test

This commit updates the #PF stress test to trigger exits and save/restore operations randomly via a background thread sending SIGUSR1. This replaces the explicit GUEST_SYNC() calls to better exercise race conditions during guest execution.

Quoted 4 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +0 -0
index 622d102179e66..92efb7ac9d95f 100644

[ ... ]

Quoted 1 line
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +0 -0
@@ -94,15 +96,41 @@ static void guest_access_memory(void *arg)

[ ... ]

tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c static void guest_access_memory(void *arg) · line 109

Quoted 9 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +7 -1
+static void *sigusr_thread_fn(void *arg)
+{
+	pthread_t vcpu_thread = (pthread_t)arg;
 
-		GUEST_SYNC(guest_faults);
+	for (;;) {
+		pthread_testcancel();
+		pthread_kill(vcpu_thread, SIGUSR1);
+		usleep(100);

[Severity: Medium]
Could this aggressive 100-microsecond signal interval cause a livelock on slower hosts or nested environments?

If the KVM VM-entry and EPT/NPT setup takes longer than 100 microseconds, the pending signal might cause KVM to abort the entry and return -EINTR to userspace before the guest makes any forward progress.

Because the main loop fully destroys and recreates the VM on every iteration (via kvm_vm_release() and vm_recreate_with_one_vcpu()), the MMU state is dropped, meaning the next loop starts completely cold. If every entry takes longer than 100us, the test will repeatedly fail to enter the guest.

Quoted 3 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +2 -0
 	}
+	return NULL;
+}

[ ... ]

tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c int main(int argc, char *argv[]) · line 218

Quoted 5 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +1 -0
@@ -175,8 +215,11 @@ int main(int argc, char *argv[])
 	}
 
 	sync_global_from_guest(vm, guest_faults);
+	TEST_ASSERT(guest_faults > 0, "No guest page faults triggered");

[Severity: Medium]
If the VM is livelocked as described above, will this assertion spuriously fail because the guest never executed an instruction?

Quoted 7 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +2 -0
 	pr_info("Guest page faults: %lu\n", guest_faults);
 
+	pthread_cancel(sigusr_thread);
+	pthread_join(sigusr_thread, NULL);
 	kvm_vm_free(vm);
 	return 0;
 }
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=8
#18 lore raw

Re: [PATCH v2 08/10] KVM: selftests: Trigger save+restore randomly in the #PF stress test

Quoted 30 lines · On Thu, Jun 4, 2026 at 1:49 PM <[email protected]> wrote:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Aggressive 100us signal interval causes test livelock and spurious failures on slower hosts. --

commit bf3ad650cf1e2ff2e1e4bd996b52e985a3657486
Author: Yosry Ahmed <[email protected]>

KVM: selftests: Trigger save+restore randomly in the #PF stress test

This commit updates the #PF stress test to trigger exits and save/restore operations randomly via a background thread sending SIGUSR1. This replaces the explicit GUEST_SYNC() calls to better exercise race conditions during guest execution.

Quoted 4 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +0 -0
index 622d102179e66..92efb7ac9d95f 100644

[ ... ]

Quoted 1 line
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +0 -0
@@ -94,15 +96,41 @@ static void guest_access_memory(void *arg)

[ ... ]

Quoted 4 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +3 -0
+static void *sigusr_thread_fn(void *arg)
+{
+     pthread_t vcpu_thread = (pthread_t)arg;
Quoted 9 lines

tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c static void guest_access_memory(void *arg) · line 109

Quoted 5 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +4 -1
-             GUEST_SYNC(guest_faults);
+     for (;;) {
+             pthread_testcancel();
+             pthread_kill(vcpu_thread, SIGUSR1);
+             usleep(100);

[Severity: Medium]
Could this aggressive 100-microsecond signal interval cause a livelock on slower hosts or nested environments?

Not sure if that's practically a concern.

In earlier iterations of the test, I remember higher intervals not reliably reproducing the bug, but trying now, seems like this is no longer the case. With 1 msec the bug still seems to reproduce reliably, at least on Turin, and we get a lot more page faults. The overall test runtime doesn't change much either.

So I think we can do usleep(msecs_to_usecs(1)) in the next version.

Quoted 21 lines

If the KVM VM-entry and EPT/NPT setup takes longer than 100 microseconds, the pending signal might cause KVM to abort the entry and return -EINTR to userspace before the guest makes any forward progress.

Because the main loop fully destroys and recreates the VM on every iteration (via kvm_vm_release() and vm_recreate_with_one_vcpu()), the MMU state is dropped, meaning the next loop starts completely cold. If every entry takes longer than 100us, the test will repeatedly fail to enter the guest.

Quoted 3 lines
      }
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +2 -0
+     return NULL;
+}

[ ... ]

Quoted 5 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +1 -0
@@ -175,8 +215,11 @@ int main(int argc, char *argv[])
      }
 
      sync_global_from_guest(vm, guest_faults);
+     TEST_ASSERT(guest_faults > 0, "No guest page faults triggered");
#19 lore raw

Re: [PATCH v2 08/10] KVM: selftests: Trigger save+restore randomly in the #PF stress test

Quoted 14 lines
Quoted 4 lines
Quoted 4 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +3 -0
+static void *sigusr_thread_fn(void *arg)
+{
+     pthread_t vcpu_thread = (pthread_t)arg;
Quoted 9 lines

tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c static void guest_access_memory(void *arg) · line 109

Quoted 5 lines
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +4 -1
-             GUEST_SYNC(guest_faults);
+     for (;;) {
+             pthread_testcancel();
+             pthread_kill(vcpu_thread, SIGUSR1);
+             usleep(100);

[Severity: Medium]
Could this aggressive 100-microsecond signal interval cause a livelock on slower hosts or nested environments?

Quoted 1 line

Not sure if that's practically a concern.

Spoiler alert: it is.

Quoted 8 lines

In earlier iterations of the test, I remember higher intervals not reliably reproducing the bug, but trying now, seems like this is no longer the case. With 1 msec the bug still seems to reproduce reliably, at least on Turin, and we get a lot more page faults. The overall test runtime doesn't change much either.

So I think we can do usleep(msecs_to_usecs(1)) in the next version.

We observed the test failing internally on slow builds (e.g. with ASAN), due to the test not completing any accesses. Using 1 msec instead of 100us does fix it indeed.

#20 lore raw

PATCH v2 9/10 KVM: selftests: Support running stress save+restore and #PF test in L2

In reply to Yosry Ahmed

Extend the stress test to allow running the access+#PF code in L2 instead of L1 by adding proper L1 guest code to bootstrap L2. The test runs in nested mode if a '-n' flag is added.

Assisted-by Gemini:gemini-3.1-pro
---
 .../kvm/x86/stress_save_restore_pf_test.c     | 75 ++++++++++++++++++-
 1 file changed, 73 insertions(+), 2 deletions(-)
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +73 -2
index 92efb7ac9d95f..c285ef093400d 100644
@@ -8,10 +8,13 @@
 #include <pthread.h>
 #include <signal.h>
 #include <unistd.h>
+#include <getopt.h>
 
 #include "test_util.h"
 #include "kvm_util.h"
 #include "processor.h"
+#include "svm_util.h"
+#include "vmx.h"
 
 #define NR_ITERATIONS		500
 
@@ -25,6 +28,8 @@ do {								\
 	fflush(stdout);						\
 } while (0)
 
+#define L2_GUEST_STACK_SIZE	64
+
 #define PTRS_PER_PTE		512
 #define PXD_INDEX(vaddr, level)	(((vaddr) >> PG_LEVEL_SHIFT(level)) & (PTRS_PER_PTE - 1))
 
@@ -99,6 +104,41 @@ static void guest_access_memory(void *arg)
 	}
 }
 
+static void l1_svm_code(struct svm_test_data *svm)
+{
+	unsigned long l2_guest_stack[L2_GUEST_STACK_SIZE];
+
+	generic_svm_setup(svm, guest_access_memory, &l2_guest_stack[L2_GUEST_STACK_SIZE]);
+
+	run_guest(svm->vmcb, svm->vmcb_gpa);
+	GUEST_ASSERT(false);
+}
+
+static void l1_vmx_code(struct vmx_pages *vmx)
+{
+	unsigned long l2_guest_stack[L2_GUEST_STACK_SIZE];
+
+	GUEST_ASSERT(prepare_for_vmx_operation(vmx));
+	GUEST_ASSERT(load_vmcs(vmx));
+	prepare_vmcs(vmx, guest_access_memory, &l2_guest_stack[L2_GUEST_STACK_SIZE]);
+
+	/* Ignore any #PF */
+	GUEST_ASSERT(!vmwrite(EXCEPTION_BITMAP, BIT(PF_VECTOR)));
+	GUEST_ASSERT(!vmwrite(PAGE_FAULT_ERROR_CODE_MASK, 0));
+	GUEST_ASSERT(!vmwrite(PAGE_FAULT_ERROR_CODE_MATCH, -1));
+
+	GUEST_ASSERT(!vmlaunch());
+	GUEST_ASSERT(false);
+}
+
+static void l1_guest_code(void *test_data)
+{
+	if (this_cpu_has(X86_FEATURE_SVM))
+		l1_svm_code(test_data);
+	else
+		l1_vmx_code(test_data);
+}
+
 static void *sigusr_thread_fn(void *arg)
 {
 	pthread_t vcpu_thread = (pthread_t)arg;
@@ -126,6 +166,25 @@ static void vcpu_sigusr_ignore(void)
 	sigaction(SIGUSR1, &sa, NULL);
 }
 
+static bool parse_args_nested(int argc, char *argv[])
+{
+	bool nested = false;
+	int opt;
+
+	while ((opt = getopt(argc, argv, "n")) != -1) {
+		switch (opt) {
+		case 'n':
+			nested = true;
+			break;
+		default:
+			printf("Usage: %s [-n]\n", argv[0]);
+			exit(1);
+		}
+	}
+
+	return nested;
+}
+
 int main(int argc, char *argv[])
 {
 	struct kvm_x86_state *state;
@@ -136,12 +195,24 @@ int main(int argc, char *argv[])
 	struct kvm_vm *vm;
 	struct ucall uc;
 	u64 *pgtable;
+	bool nested;
 	gva_t gva;
 	u64 pte;
 
-	vm = vm_create_with_one_vcpu(&vcpu, guest_access_memory);
+	nested = parse_args_nested(argc, argv);
+
+	vm = vm_create_with_one_vcpu(&vcpu, nested ? l1_guest_code : guest_access_memory);
 	vm_install_exception_handler(vm, PF_VECTOR, guest_pf_handler);
 
+	if (nested) {
+		TEST_REQUIRE(kvm_cpu_has(X86_FEATURE_SVM) || kvm_cpu_has(X86_FEATURE_VMX));
+		if (kvm_cpu_has(X86_FEATURE_SVM))
+			vcpu_alloc_svm(vm, &gva);
+		else
+			vcpu_alloc_vmx(vm, &gva);
+		vcpu_args_set(vcpu, 1, gva);
+	}
+
 	pte_present_mask = PTE_PRESENT_MASK(&vm->mmu);
 	pte_huge_mask = PTE_HUGE_MASK(&vm->mmu);
 	sync_global_to_guest(vm, pte_present_mask);
@@ -216,7 +287,7 @@ int main(int argc, char *argv[])
 
 	sync_global_from_guest(vm, guest_faults);
 	TEST_ASSERT(guest_faults > 0, "No guest page faults triggered");
-	pr_info("Guest page faults: %lu\n", guest_faults);
+	pr_info("Guest page faults%s: %lu\n", nested ? " (in L2)" : "", guest_faults);
 
 	pthread_cancel(sigusr_thread);
 	pthread_join(sigusr_thread, NULL);
Signature · 1 line
2.54.0.1032.g2f8565e1d1-goog
#21 lore raw

PATCH v2 10/10 KVM: selftests: Trigger L2->L1 exits stress save+restore and #PF test

In reply to Yosry Ahmed

Extend the testing coverage in L2 by injecting a #UD into the vCPU every other iteration during restore, and intercepting #UD from L1, essentially forcing an L2 -> L1 VM-Exit directly after save+restore.

With this change, the test reliably reproduces the CR2 bug fixed by commit 5c247d08bc81 ("KVM: nSVM: Use vcpu->arch.cr2 when updating vmcb12 on nested #VMEXIT") -- at least on Milan, Genoa, and Turin CPUs.

Assisted-by Gemini:gemini-3.1-pro
---
 .../kvm/x86/stress_save_restore_pf_test.c     | 47 +++++++++++++++++--
 1 file changed, 42 insertions(+), 5 deletions(-)
tools/testing/selftests/kvm/x86/stress_save_restore_pf_test.c +42 -5
index c285ef093400d..6aae34cf8f13f 100644
@@ -110,8 +110,12 @@ static void l1_svm_code(struct svm_test_data *svm)
 
 	generic_svm_setup(svm, guest_access_memory, &l2_guest_stack[L2_GUEST_STACK_SIZE]);
 
-	run_guest(svm->vmcb, svm->vmcb_gpa);
-	GUEST_ASSERT(false);
+	svm->vmcb->control.intercept_exceptions |= BIT(UD_VECTOR);
+
+	while (1) {
+		run_guest(svm->vmcb, svm->vmcb_gpa);
+		GUEST_ASSERT_EQ(svm->vmcb->control.exit_code, (SVM_EXIT_EXCP_BASE + UD_VECTOR));
+	}
 }
 
 static void l1_vmx_code(struct vmx_pages *vmx)
@@ -122,13 +126,17 @@ static void l1_vmx_code(struct vmx_pages *vmx)
 	GUEST_ASSERT(load_vmcs(vmx));
 	prepare_vmcs(vmx, guest_access_memory, &l2_guest_stack[L2_GUEST_STACK_SIZE]);
 
-	/* Ignore any #PF */
-	GUEST_ASSERT(!vmwrite(EXCEPTION_BITMAP, BIT(PF_VECTOR)));
+	/* Intercept UD, ignore any #PF */
+	GUEST_ASSERT(!vmwrite(EXCEPTION_BITMAP, BIT(UD_VECTOR) | BIT(PF_VECTOR)));
 	GUEST_ASSERT(!vmwrite(PAGE_FAULT_ERROR_CODE_MASK, 0));
 	GUEST_ASSERT(!vmwrite(PAGE_FAULT_ERROR_CODE_MATCH, -1));
 
 	GUEST_ASSERT(!vmlaunch());
-	GUEST_ASSERT(false);
+	while (1) {
+		GUEST_ASSERT_EQ(vmreadz(VM_EXIT_REASON), EXIT_REASON_EXCEPTION_NMI);
+		GUEST_ASSERT_EQ(vmreadz(VM_EXIT_INTR_INFO) & 0xff, UD_VECTOR);
+		GUEST_ASSERT(!vmresume());
+	}
 }
 
 static void l1_guest_code(void *test_data)
@@ -166,6 +174,24 @@ static void vcpu_sigusr_ignore(void)
 	sigaction(SIGUSR1, &sa, NULL);
 }
 
+static bool vcpu_state_is_guest_mode(struct kvm_x86_state *state)
+{
+	return !!(state->nested.flags & KVM_STATE_NESTED_GUEST_MODE);
+}
+
+static void vcpu_state_inject_ud(struct kvm_x86_state *state)
+{
+	if (state->events.exception.pending || state->events.exception.injected)
+		return;
+
+	state->events.flags |= KVM_VCPUEVENT_VALID_PAYLOAD;
+	state->events.exception.pending = true;
+	state->events.exception.injected = false;
+	state->events.exception.nr = UD_VECTOR;
+	state->events.exception.has_error_code = false;
+	state->events.exception_has_payload = false;
+}
+
 static bool parse_args_nested(int argc, char *argv[])
 {
 	bool nested = false;
@@ -199,10 +225,13 @@ int main(int argc, char *argv[])
 	gva_t gva;
 	u64 pte;
 
+	TEST_REQUIRE(kvm_has_cap(KVM_CAP_EXCEPTION_PAYLOAD));
+
 	nested = parse_args_nested(argc, argv);
 
 	vm = vm_create_with_one_vcpu(&vcpu, nested ? l1_guest_code : guest_access_memory);
 	vm_install_exception_handler(vm, PF_VECTOR, guest_pf_handler);
+	vm_enable_cap(vm, KVM_CAP_EXCEPTION_PAYLOAD, -2ul);
 
 	if (nested) {
 		TEST_REQUIRE(kvm_cpu_has(X86_FEATURE_SVM) || kvm_cpu_has(X86_FEATURE_VMX));
@@ -277,8 +306,16 @@ int main(int argc, char *argv[])
 
 		state = vcpu_save_state(vcpu);
 
+		/*
+		 * If the vCPU is in guest mode, inject a #UD to trigger an
+		 * L2->L1 VM-Exit every other iteration.
+		 */
+		if (nested && vcpu_state_is_guest_mode(state) && count % 2 == 0)
+			vcpu_state_inject_ud(state);
+
 		kvm_vm_release(vm);
 		vcpu = vm_recreate_with_one_vcpu(vm);
+		vm_enable_cap(vm, KVM_CAP_EXCEPTION_PAYLOAD, -2ul);
 		vcpu_load_state(vcpu, state);
 		kvm_x86_state_cleanup(state);
 
Signature · 1 line
2.54.0.1032.g2f8565e1d1-goog