[PATCH 0/4] KVM: selftests: Fastops DIV testcases

[PATCH 0/4] KVM: selftests: Fastops DIV testcases

[Sean Christopherson]

Add DIV and IDIV support to the so called fastops test to add coverage for #DE
(the only exception that can occur during fastops execution), which was broken
due to a typo in PeterZ's overhaul[*].

I meant to post this back in July and forgot...

[*] https://lore.kernel.org/all/aIF7ZhWZxlkcpm4y@google.com

Sean Christopherson (4):
  KVM: selftests: Add support for #DE exception fixup
  KVM: selftests: Add coverage for 'b' (byte) sized fastops emulation
  KVM: selftests: Dedup the gnarly constraints of the fastops tests
    (more macros!)
  KVM: selftests: Add support for DIV and IDIV in the fastops test

 .../selftests/kvm/include/x86/processor.h     |  6 ++
 .../testing/selftests/kvm/lib/x86/processor.c |  2 +-
 .../testing/selftests/kvm/x86/fastops_test.c  | 82 ++++++++++++++-----
 3 files changed, 70 insertions(+), 20 deletions(-)


base-commit: 33f843444e28920d6e624c6c24637b4bb5d3c8de
-- 
2.51.0.384.g4c02a37b29-goog

[PATCH 1/4] KVM: selftests: Add support for #DE exception fixup

[Sean Christopherson]

Add support for handling #DE (divide error) exceptions in KVM selftests
so that the fastops test can verify KVM correctly handles #DE when
emulating DIV or IDIV on behalf of the guest.  Morph #DE to 0xff (i.e.
to -1) as a mostly-arbitrary vector to indicate #DE, so that '0' (the
real #DE vector) can still be used to indicate "no exception".

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 tools/testing/selftests/kvm/include/x86/processor.h | 6 ++++++
 tools/testing/selftests/kvm/lib/x86/processor.c     | 2 +-
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/kvm/include/x86/processor.h b/tools/testing/selftests/kvm/include/x86/processor.h
index 2efb05c2f2fb..5142d1bf15a2 100644
--- a/tools/testing/selftests/kvm/include/x86/processor.h
+++ b/tools/testing/selftests/kvm/include/x86/processor.h
@@ -1179,6 +1179,12 @@ struct idt_entry {
 void vm_install_exception_handler(struct kvm_vm *vm, int vector,
 			void (*handler)(struct ex_regs *));
 
+/*
+ * Exception fixup morphs #DE to an arbitrary magic vector so that '0' can be
+ * used to signal "no expcetion".
+ */
+#define KVM_MAGIC_DE_VECTOR 0xff
+
 /* If a toddler were to say "abracadabra". */
 #define KVM_EXCEPTION_MAGIC 0xabacadabaULL
 
diff --git a/tools/testing/selftests/kvm/lib/x86/processor.c b/tools/testing/selftests/kvm/lib/x86/processor.c
index d4c19ac885a9..3b63c99f7b96 100644
--- a/tools/testing/selftests/kvm/lib/x86/processor.c
+++ b/tools/testing/selftests/kvm/lib/x86/processor.c
@@ -557,7 +557,7 @@ static bool kvm_fixup_exception(struct ex_regs *regs)
 		return false;
 
 	if (regs->vector == DE_VECTOR)
-		return false;
+		regs->vector = KVM_MAGIC_DE_VECTOR;
 
 	regs->rip = regs->r11;
 	regs->r9 = regs->vector;
-- 
2.51.0.384.g4c02a37b29-goog

[PATCH 2/4] KVM: selftests: Add coverage for 'b' (byte) sized fastops emulation

[Sean Christopherson]

Extend the fastops test to cover instructions that operate on 8-bit data.
Support for 8-bit instructions was omitted from the original commit purely
due to complications with BT not having a r/m8 variant.  To keep the
RFLAGS.CF behavior deterministic and not heavily biased to '0' or '1',
continue using BT, but cast and load the to-be-tested value into a
dedicated 32-bit constraint.

Supporting 8-bit operations will allow using guest_test_fastops() as-is to
provide full coverage for DIV and IDIV.  For divide operations, covering
all operand sizes _is_ interesting, because KVM needs provide exception
fixup for each size (failure to handle a #DE could panic the host).

Link: https://lore.kernel.org/all/aIF7ZhWZxlkcpm4y@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 .../testing/selftests/kvm/x86/fastops_test.c  | 20 ++++++++++++-------
 1 file changed, 13 insertions(+), 7 deletions(-)

diff --git a/tools/testing/selftests/kvm/x86/fastops_test.c b/tools/testing/selftests/kvm/x86/fastops_test.c
index 2ac89d6c1e46..6c9a2dbf6365 100644
--- a/tools/testing/selftests/kvm/x86/fastops_test.c
+++ b/tools/testing/selftests/kvm/x86/fastops_test.c
@@ -10,12 +10,13 @@
  */
 #define guest_execute_fastop_1(FEP, insn, __val, __flags)				\
 ({											\
-	__asm__ __volatile__("bt $0, %[val]\n\t"					\
+	__asm__ __volatile__("bt $0, %[ro_val]\n\t"					\
 			     FEP insn " %[val]\n\t"					\
 			     "pushfq\n\t"						\
 			     "pop %[flags]\n\t"						\
 			     : [val]"+r"(__val), [flags]"=r"(__flags)			\
-			     : : "cc", "memory");					\
+			     : [ro_val]"rm"((uint32_t)__val)				\
+			     : "cc", "memory");						\
 })
 
 #define guest_test_fastop_1(insn, type_t, __val)					\
@@ -36,12 +37,13 @@
 
 #define guest_execute_fastop_2(FEP, insn, __input, __output, __flags)			\
 ({											\
-	__asm__ __volatile__("bt $0, %[output]\n\t"					\
+	__asm__ __volatile__("bt $0, %[ro_val]\n\t"					\
 			     FEP insn " %[input], %[output]\n\t"			\
 			     "pushfq\n\t"						\
 			     "pop %[flags]\n\t"						\
 			     : [output]"+r"(__output), [flags]"=r"(__flags)		\
-			     : [input]"r"(__input) : "cc", "memory");			\
+			     : [input]"r"(__input), [ro_val]"rm"((uint32_t)__output)	\
+			     : "cc", "memory");						\
 })
 
 #define guest_test_fastop_2(insn, type_t, __val1, __val2)				\
@@ -63,12 +65,13 @@
 
 #define guest_execute_fastop_cl(FEP, insn, __shift, __output, __flags)			\
 ({											\
-	__asm__ __volatile__("bt $0, %[output]\n\t"					\
+	__asm__ __volatile__("bt $0, %[ro_val]\n\t"					\
 			     FEP insn " %%cl, %[output]\n\t"				\
 			     "pushfq\n\t"						\
 			     "pop %[flags]\n\t"						\
 			     : [output]"+r"(__output), [flags]"=r"(__flags)		\
-			     : "c"(__shift) : "cc", "memory");				\
+			     : "c"(__shift), [ro_val]"rm"((uint32_t)__output)		\
+			     : "cc", "memory");						\
 })
 
 #define guest_test_fastop_cl(insn, type_t, __val1, __val2)				\
@@ -115,14 +118,16 @@ do {											\
 			guest_test_fastop_2("add" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_2("adc" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_2("and" suffix, type_t, vals[i], vals[j]);	\
+if (sizeof(type_t) != 1) {							\
 			guest_test_fastop_2("bsf" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_2("bsr" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_2("bt" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_2("btc" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_2("btr" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_2("bts" suffix, type_t, vals[i], vals[j]);	\
-			guest_test_fastop_2("cmp" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_2("imul" suffix, type_t, vals[i], vals[j]);	\
+}											\
+			guest_test_fastop_2("cmp" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_2("or" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_2("sbb" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_2("sub" suffix, type_t, vals[i], vals[j]);	\
@@ -142,6 +147,7 @@ do {											\
 
 static void guest_code(void)
 {
+	guest_test_fastops(uint8_t, "b");
 	guest_test_fastops(uint16_t, "w");
 	guest_test_fastops(uint32_t, "l");
 	guest_test_fastops(uint64_t, "q");
-- 
2.51.0.384.g4c02a37b29-goog

[PATCH 3/4] KVM: selftests: Dedup the gnarly constraints of the fastops tests (more macros!)

[Sean Christopherson]

Add a fastop() macro along with macros to define its required constraints,
and use the macros to dedup the innermost guts of the fastop testcases.

No functional change intended.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 .../testing/selftests/kvm/x86/fastops_test.c  | 36 +++++++++----------
 1 file changed, 18 insertions(+), 18 deletions(-)

diff --git a/tools/testing/selftests/kvm/x86/fastops_test.c b/tools/testing/selftests/kvm/x86/fastops_test.c
index 6c9a2dbf6365..26a381c8303a 100644
--- a/tools/testing/selftests/kvm/x86/fastops_test.c
+++ b/tools/testing/selftests/kvm/x86/fastops_test.c
@@ -8,14 +8,20 @@
  * to set RFLAGS.CF based on whether or not the input is even or odd, so that
  * instructions like ADC and SBB are deterministic.
  */
+#define fastop(__insn)									\
+	"bt $0, %[bt_val]\n\t"								\
+	__insn "\n\t"									\
+	"pushfq\n\t"									\
+	"pop %[flags]\n\t"
+
+#define flags_constraint(flags_val) [flags]"=r"(flags_val)
+#define bt_constraint(__bt_val) [bt_val]"rm"((uint32_t)__bt_val)
+
 #define guest_execute_fastop_1(FEP, insn, __val, __flags)				\
 ({											\
-	__asm__ __volatile__("bt $0, %[ro_val]\n\t"					\
-			     FEP insn " %[val]\n\t"					\
-			     "pushfq\n\t"						\
-			     "pop %[flags]\n\t"						\
-			     : [val]"+r"(__val), [flags]"=r"(__flags)			\
-			     : [ro_val]"rm"((uint32_t)__val)				\
+	__asm__ __volatile__(fastop(FEP insn " %[val]")					\
+			     : [val]"+r"(__val), flags_constraint(__flags)		\
+			     : bt_constraint(__val)					\
 			     : "cc", "memory");						\
 })
 
@@ -37,12 +43,9 @@
 
 #define guest_execute_fastop_2(FEP, insn, __input, __output, __flags)			\
 ({											\
-	__asm__ __volatile__("bt $0, %[ro_val]\n\t"					\
-			     FEP insn " %[input], %[output]\n\t"			\
-			     "pushfq\n\t"						\
-			     "pop %[flags]\n\t"						\
-			     : [output]"+r"(__output), [flags]"=r"(__flags)		\
-			     : [input]"r"(__input), [ro_val]"rm"((uint32_t)__output)	\
+	__asm__ __volatile__(fastop(FEP insn " %[input], %[output]")			\
+			     : [output]"+r"(__output), flags_constraint(__flags)	\
+			     : [input]"r"(__input), bt_constraint(__output)		\
 			     : "cc", "memory");						\
 })
 
@@ -65,12 +68,9 @@
 
 #define guest_execute_fastop_cl(FEP, insn, __shift, __output, __flags)			\
 ({											\
-	__asm__ __volatile__("bt $0, %[ro_val]\n\t"					\
-			     FEP insn " %%cl, %[output]\n\t"				\
-			     "pushfq\n\t"						\
-			     "pop %[flags]\n\t"						\
-			     : [output]"+r"(__output), [flags]"=r"(__flags)		\
-			     : "c"(__shift), [ro_val]"rm"((uint32_t)__output)		\
+	__asm__ __volatile__(fastop(FEP insn " %%cl, %[output]")			\
+			     : [output]"+r"(__output), flags_constraint(__flags)	\
+			     : "c"(__shift), bt_constraint(__output)			\
 			     : "cc", "memory");						\
 })
 
-- 
2.51.0.384.g4c02a37b29-goog

[PATCH 4/4] KVM: selftests: Add support for DIV and IDIV in the fastops test

[Sean Christopherson]

Extend the fastops test coverage to DIV and IDIV, specifically to provide
coverage for #DE (divide error) exceptions, as #DE is the only exception
that can occur in KVM's fastops path, i.e. that requires exception fixup.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 .../testing/selftests/kvm/x86/fastops_test.c  | 38 +++++++++++++++++++
 1 file changed, 38 insertions(+)

diff --git a/tools/testing/selftests/kvm/x86/fastops_test.c b/tools/testing/selftests/kvm/x86/fastops_test.c
index 26a381c8303a..8926cfe0e209 100644
--- a/tools/testing/selftests/kvm/x86/fastops_test.c
+++ b/tools/testing/selftests/kvm/x86/fastops_test.c
@@ -92,6 +92,42 @@
 			ex_flags, insn, shift, (uint64_t)input, flags);			\
 })
 
+#define guest_execute_fastop_div(__KVM_ASM_SAFE, insn, __a, __d, __rm, __flags)		\
+({											\
+	uint64_t ign_error_code;							\
+	uint8_t vector;									\
+											\
+	__asm__ __volatile__(fastop(__KVM_ASM_SAFE(insn " %[denom]"))			\
+			     : "+a"(__a), "+d"(__d), flags_constraint(__flags),		\
+			       KVM_ASM_SAFE_OUTPUTS(vector, ign_error_code)		\
+			     : [denom]"rm"(__rm), bt_constraint(__rm)			\
+			     : "cc", "memory", KVM_ASM_SAFE_CLOBBERS);			\
+	vector;										\
+})
+
+#define guest_test_fastop_div(insn, type_t, __val1, __val2)				\
+({											\
+	type_t _a = __val1, _d = __val1, rm = __val2;					\
+	type_t a = _a, d = _d, ex_a = _a, ex_d = _d;					\
+	uint64_t flags, ex_flags;							\
+	uint8_t v, ex_v;								\
+											\
+	ex_v = guest_execute_fastop_div(KVM_ASM_SAFE, insn, ex_a, ex_d, rm, ex_flags);	\
+	v = guest_execute_fastop_div(KVM_ASM_SAFE_FEP, insn, a, d, rm, flags);		\
+											\
+	GUEST_ASSERT_EQ(v, ex_v);							\
+	__GUEST_ASSERT(v == ex_v,							\
+		       "Wanted vector 0x%x for '%s 0x%lx:0x%lx/0x%lx', got 0x%x",	\
+		       ex_v, insn, (uint64_t)_a, (uint64_t)_d, (uint64_t)rm, v);	\
+	__GUEST_ASSERT(a == ex_a && d == ex_d,						\
+		       "Wanted 0x%lx:0x%lx for '%s 0x%lx:0x%lx/0x%lx', got 0x%lx:0x%lx",\
+		       (uint64_t)ex_a, (uint64_t)ex_d, insn, (uint64_t)_a,		\
+		       (uint64_t)_d, (uint64_t)rm, (uint64_t)a, (uint64_t)d);		\
+	__GUEST_ASSERT(v || ex_v || (flags == ex_flags),				\
+			"Wanted flags 0x%lx for '%s  0x%lx:0x%lx/0x%lx', got 0x%lx",	\
+			ex_flags, insn, (uint64_t)_a, (uint64_t)_d, (uint64_t)rm, flags);\
+})
+
 static const uint64_t vals[] = {
 	0,
 	1,
@@ -141,6 +177,8 @@ if (sizeof(type_t) != 1) {							\
 			guest_test_fastop_cl("sar" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_cl("shl" suffix, type_t, vals[i], vals[j]);	\
 			guest_test_fastop_cl("shr" suffix, type_t, vals[i], vals[j]);	\
+											\
+			guest_test_fastop_div("div" suffix, type_t, vals[i], vals[j]);	\
 		}									\
 	}										\
 } while (0)
-- 
2.51.0.384.g4c02a37b29-goog

Re: [PATCH 0/4] KVM: selftests: Fastops DIV testcases

[Sean Christopherson]

On Tue, 09 Sep 2025 13:28:31 -0700, Sean Christopherson wrote:
> Add DIV and IDIV support to the so called fastops test to add coverage for #DE
> (the only exception that can occur during fastops execution), which was broken
> due to a typo in PeterZ's overhaul[*].
> 
> I meant to post this back in July and forgot...
> 
> [*] https://lore.kernel.org/all/aIF7ZhWZxlkcpm4y@google.com
> 
> [...]

Applied to kvm-x86 selftests, thanks!

[1/4] KVM: selftests: Add support for #DE exception fixup
      https://github.com/kvm-x86/linux/commit/7b39b6c76942
[2/4] KVM: selftests: Add coverage for 'b' (byte) sized fastops emulation
      https://github.com/kvm-x86/linux/commit/9bf5da1ca427
[3/4] KVM: selftests: Dedup the gnarly constraints of the fastops tests (more macros!)
      https://github.com/kvm-x86/linux/commit/fe08478b1d51
[4/4] KVM: selftests: Add support for DIV and IDIV in the fastops test
      https://github.com/kvm-x86/linux/commit/aebc62b3dedc

--
https://github.com/kvm-x86/linux/tree/next