Skip to contentSkip to conversation

x86/kvm/fpu: Fix guest migration bugs that can crash guest

PATCH Applied Leonardo Bras 3 participants 8 messages Feb 17, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v4 cover x86/kvm/fpu: Fix guest migration bugs that can crash guest

This patchset comes from a bug I found during qemu guest migration from a host with newer CPU to a host with an older version of this CPU, and thus having less FPU features.

When the guests were created, the one with less features is used as config, so migration is possible.

Patch 1 fix a bug that always happens during this migration, and is related to the fact that xsave saves all feature flags, but xrstor does not touch the PKRU flag. It also changes how fpstate->user_xfeatures is set, going from kvm_check_cpuid() to the later called kvm_vcpu_after_set_cpuid().

Patch 2 removes kvm_vcpu_arch.guest_supported_xcr0 since it now duplicates guest_fpu.fpstate->user_xfeatures. Some wrappers were introduced in order to make it easier to read the replaced version.

Patches were compile-tested, and could fix the bug found.

Please let me know of anything to improve!

Best regards,
Leo

--
Changes since v3:
- Add new patch to remove the use of kvm_vcpu_arch.guest_supported_xcr0, since it is now duplicating guest_fpu.fpstate->user_xfeatures.
- On patch 1, also avoid setting user_xfeatures on kvm_check_cpuid(), since it is already set in kvm_vcpu_after_set_cpuid() now.
Changes since v2:
- Fix building error because I forgot to EXPORT_SYMBOL(fpu_user_cfg) Changes since v1:
- Instead of masking xfeatures, mask user_xfeatures instead. This will only change the value sent to user, instead of the one saved in buf.
- Above change removed the need of the patch 2/2
- Instead of masking the current value of user_xfeatures, save on it fpu_user_cfg.default_features & vcpu->arch.guest_supported_xcr0

Leonardo Bras (2):
x86/kvm/fpu: Mask guest fpstate->xfeatures with guest_supported_xcr0 x86/kvm/fpu: Remove kvm_vcpu_arch.guest_supported_xcr0

 arch/x86/include/asm/kvm_host.h |  1 -
 arch/x86/kernel/fpu/xstate.c    |  5 ++++-
 arch/x86/kvm/cpuid.c            |  5 ++++-
 arch/x86/kvm/x86.c              | 20 +++++++++++++++-----
 4 files changed, 23 insertions(+), 8 deletions(-)
Signature · 1 line
2.35.1
#1 lore raw

PATCH v4 1/2 x86/kvm/fpu: Mask guest fpstate->xfeatures with guest_supported_xcr0

During host/guest switch (like in kvm_arch_vcpu_ioctl_run()), the kernel swaps the fpu between host/guest contexts, by using fpu_swap_kvm_fpstate().

When xsave feature is available, the fpu swap is done by:
- xsave(s) instruction, with guest's fpstate->xfeatures as mask, is used to store the current state of the fpu registers to a buffer.
- xrstor(s) instruction, with (fpu_kernel_cfg.max_features & XFEATURE_MASK_FPSTATE) as mask, is used to put the buffer into fpu regs.

For xsave(s) the mask is used to limit what parts of the fpu regs will be copied to the buffer. Likewise on xrstor(s), the mask is used to limit what parts of the fpu regs will be changed.

The mask for xsave(s), the guest's fpstate->xfeatures, is defined on kvm_arch_vcpu_create(), which (in summary) sets it to all features supported by the cpu which are enabled on kernel config.

This means that xsave(s) will save to guest buffer all the fpu regs contents the cpu has enabled when the guest is paused, even if they are not used.

This would not be an issue, if xrstor(s) would also do that.

xrstor(s)'s mask for host/guest swap is basically every valid feature contained in kernel config, except XFEATURE_MASK_PKRU. Accordingto kernel src, it is instead switched in switch_to() and flush_thread().

Then, the following happens with a host supporting PKRU starts a guest that does not support it:
1 - Host has XFEATURE_MASK_PKRU set. 1st switch to guest,
2 - xsave(s) fpu regs to host fpustate (buffer has XFEATURE_MASK_PKRU) 3 - xrstor(s) guest fpustate to fpu regs (fpu regs have XFEATURE_MASK_PKRU) 4 - guest runs, then switch back to host,
5 - xsave(s) fpu regs to guest fpstate (buffer now have XFEATURE_MASK_PKRU) 6 - xrstor(s) host fpstate to fpu regs.
7 - kvm_vcpu_ioctl_x86_get_xsave() copy guest fpstate to userspace (with
XFEATURE_MASK_PKRU, which should not be supported by guest vcpu)

On 5, even though the guest does not support PKRU, it does have the flag set on guest fpstate, which is transferred to userspace via vcpu ioctl KVM_GET_XSAVE.

This becomes a problem when the user decides on migrating the above guest to another machine that does not support PKRU:
The new host restores guest's fpu regs to as they were before (xrstor(s)), but since the new host don't support PKRU, a general-protection exception ocurs in xrstor(s) and that crashes the guest.

This can be solved by making the guest's fpstate->user_xfeatures hold a copy of guest_supported_xcr0. This way, on 7 the only flags copied to userspace will be the ones compatible to guest requirements, and thus there will be no issue during migration.

As a bonus, it will also fail if userspace tries to set fpu features that are not compatible to the guest configuration. (KVM_SET_XSAVE ioctl)

Also, since kvm_vcpu_after_set_cpuid() now sets fpstate->user_xfeatures, there is not need to set it in kvm_check_cpuid(). So, change fpstate_realloc() so it does not touch fpstate->user_xfeatures if a non-NULL guest_fpu is passed, which is the case when kvm_check_cpuid() calls it.

---
 arch/x86/kernel/fpu/xstate.c | 5 ++++-
 arch/x86/kvm/cpuid.c         | 2 ++
 2 files changed, 6 insertions(+), 1 deletion(-)
arch/x86/kernel/fpu/xstate.c +4 -1
index 02b3ddaf4f75..7c7824ae7862 100644
@@ -1558,7 +1558,10 @@ static int fpstate_realloc(u64 xfeatures, unsigned int ksize,
 		fpregs_restore_userregs();
 
 	newfps->xfeatures = curfps->xfeatures | xfeatures;
-	newfps->user_xfeatures = curfps->user_xfeatures | xfeatures;
+
+	if (!guest_fpu)
+		newfps->user_xfeatures = curfps->user_xfeatures | xfeatures;
+
 	newfps->xfd = curfps->xfd & ~xfeatures;
 
 	/* Do the final updates within the locked region */
arch/x86/kvm/cpuid.c +2 -0
index 494d4d351859..71125291c578 100644
@@ -296,6 +296,8 @@ static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
 	vcpu->arch.guest_supported_xcr0 =
 		cpuid_get_supported_xcr0(vcpu->arch.cpuid_entries, vcpu->arch.cpuid_nent);
 
+	vcpu->arch.guest_fpu.fpstate->user_xfeatures = vcpu->arch.guest_supported_xcr0;
+
 	kvm_update_pv_runtime(vcpu);
 
 	vcpu->arch.maxphyaddr = cpuid_query_maxphyaddr(vcpu);
Signature · 1 line
2.35.1
#2 lore raw

Re: [PATCH v4 1/2] x86/kvm/fpu: Mask guest fpstate->xfeatures with guest_supported_xcr0

The single line summary is now out of date - there's no new masking.

Quoted 90 lines · Leonardo Bras

During host/guest switch (like in kvm_arch_vcpu_ioctl_run()), the kernel swaps the fpu between host/guest contexts, by using fpu_swap_kvm_fpstate().

When xsave feature is available, the fpu swap is done by:
- xsave(s) instruction, with guest's fpstate->xfeatures as mask, is used to store the current state of the fpu registers to a buffer.
- xrstor(s) instruction, with (fpu_kernel_cfg.max_features & XFEATURE_MASK_FPSTATE) as mask, is used to put the buffer into fpu regs.

For xsave(s) the mask is used to limit what parts of the fpu regs will be copied to the buffer. Likewise on xrstor(s), the mask is used to limit what parts of the fpu regs will be changed.

The mask for xsave(s), the guest's fpstate->xfeatures, is defined on kvm_arch_vcpu_create(), which (in summary) sets it to all features supported by the cpu which are enabled on kernel config.

This means that xsave(s) will save to guest buffer all the fpu regs contents the cpu has enabled when the guest is paused, even if they are not used.

This would not be an issue, if xrstor(s) would also do that.

xrstor(s)'s mask for host/guest swap is basically every valid feature contained in kernel config, except XFEATURE_MASK_PKRU. Accordingto kernel src, it is instead switched in switch_to() and flush_thread().

Then, the following happens with a host supporting PKRU starts a guest that does not support it:
1 - Host has XFEATURE_MASK_PKRU set. 1st switch to guest,
2 - xsave(s) fpu regs to host fpustate (buffer has XFEATURE_MASK_PKRU) 3 - xrstor(s) guest fpustate to fpu regs (fpu regs have XFEATURE_MASK_PKRU) 4 - guest runs, then switch back to host,
5 - xsave(s) fpu regs to guest fpstate (buffer now have XFEATURE_MASK_PKRU) 6 - xrstor(s) host fpstate to fpu regs.
7 - kvm_vcpu_ioctl_x86_get_xsave() copy guest fpstate to userspace (with
XFEATURE_MASK_PKRU, which should not be supported by guest vcpu)

On 5, even though the guest does not support PKRU, it does have the flag set on guest fpstate, which is transferred to userspace via vcpu ioctl KVM_GET_XSAVE.

This becomes a problem when the user decides on migrating the above guest to another machine that does not support PKRU:
The new host restores guest's fpu regs to as they were before (xrstor(s)), but since the new host don't support PKRU, a general-protection exception ocurs in xrstor(s) and that crashes the guest.

This can be solved by making the guest's fpstate->user_xfeatures hold a copy of guest_supported_xcr0. This way, on 7 the only flags copied to userspace will be the ones compatible to guest requirements, and thus there will be no issue during migration.

As a bonus, it will also fail if userspace tries to set fpu features that are not compatible to the guest configuration. (KVM_SET_XSAVE ioctl)

Also, since kvm_vcpu_after_set_cpuid() now sets fpstate->user_xfeatures, there is not need to set it in kvm_check_cpuid(). So, change fpstate_realloc() so it does not touch fpstate->user_xfeatures if a non-NULL guest_fpu is passed, which is the case when kvm_check_cpuid() calls it.

---
 arch/x86/kernel/fpu/xstate.c | 5 ++++-
 arch/x86/kvm/cpuid.c         | 2 ++
 2 files changed, 6 insertions(+), 1 deletion(-)
arch/x86/kernel/fpu/xstate.c +4 -1
index 02b3ddaf4f75..7c7824ae7862 100644
@@ -1558,7 +1558,10 @@ static int fpstate_realloc(u64 xfeatures, unsigned int ksize,
 		fpregs_restore_userregs();
 
 	newfps->xfeatures = curfps->xfeatures | xfeatures;
-	newfps->user_xfeatures = curfps->user_xfeatures | xfeatures;
+
+	if (!guest_fpu)
+		newfps->user_xfeatures = curfps->user_xfeatures | xfeatures;
+
 	newfps->xfd = curfps->xfd & ~xfeatures;
 
 	/* Do the final updates within the locked region */
arch/x86/kvm/cpuid.c +0 -0
index 494d4d351859..71125291c578 100644

arch/x86/kvm/cpuid.c static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu) · line 303

Quoted 8 lines
arch/x86/kvm/cpuid.c +2 -0
@@ -296,6 +296,8 @@ static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
 	vcpu->arch.guest_supported_xcr0 =
 		cpuid_get_supported_xcr0(vcpu->arch.cpuid_entries, vcpu->arch.cpuid_nent);
 
+	vcpu->arch.guest_fpu.fpstate->user_xfeatures = vcpu->arch.guest_supported_xcr0;
+
 	kvm_update_pv_runtime(vcpu);
 
 	vcpu->arch.maxphyaddr = cpuid_query_maxphyaddr(vcpu);

dme.

Signature · 1 line
All those lines and circles, to me, a mystery.
maintainer
#3 lore raw

Re: [PATCH v4 1/2] x86/kvm/fpu: Mask guest fpstate->xfeatures with guest_supported_xcr0

Quoted 1 line · David Edmondson

The single line summary is now out of date - there's no new masking.

Thanks for the review, I made the adjustments and pushed to master.

Paolo

Quoted 102 lines · Leonardo Bras
Quoted 91 lines

During host/guest switch (like in kvm_arch_vcpu_ioctl_run()), the kernel swaps the fpu between host/guest contexts, by using fpu_swap_kvm_fpstate().

When xsave feature is available, the fpu swap is done by:
- xsave(s) instruction, with guest's fpstate->xfeatures as mask, is used to store the current state of the fpu registers to a buffer.
- xrstor(s) instruction, with (fpu_kernel_cfg.max_features & XFEATURE_MASK_FPSTATE) as mask, is used to put the buffer into fpu regs.

For xsave(s) the mask is used to limit what parts of the fpu regs will be copied to the buffer. Likewise on xrstor(s), the mask is used to limit what parts of the fpu regs will be changed.

The mask for xsave(s), the guest's fpstate->xfeatures, is defined on kvm_arch_vcpu_create(), which (in summary) sets it to all features supported by the cpu which are enabled on kernel config.

This means that xsave(s) will save to guest buffer all the fpu regs contents the cpu has enabled when the guest is paused, even if they are not used.

This would not be an issue, if xrstor(s) would also do that.

xrstor(s)'s mask for host/guest swap is basically every valid feature contained in kernel config, except XFEATURE_MASK_PKRU. Accordingto kernel src, it is instead switched in switch_to() and flush_thread().

Then, the following happens with a host supporting PKRU starts a guest that does not support it:
1 - Host has XFEATURE_MASK_PKRU set. 1st switch to guest,
2 - xsave(s) fpu regs to host fpustate (buffer has XFEATURE_MASK_PKRU) 3 - xrstor(s) guest fpustate to fpu regs (fpu regs have XFEATURE_MASK_PKRU) 4 - guest runs, then switch back to host,
5 - xsave(s) fpu regs to guest fpstate (buffer now have XFEATURE_MASK_PKRU) 6 - xrstor(s) host fpstate to fpu regs.
7 - kvm_vcpu_ioctl_x86_get_xsave() copy guest fpstate to userspace (with
XFEATURE_MASK_PKRU, which should not be supported by guest vcpu)

On 5, even though the guest does not support PKRU, it does have the flag set on guest fpstate, which is transferred to userspace via vcpu ioctl KVM_GET_XSAVE.

This becomes a problem when the user decides on migrating the above guest to another machine that does not support PKRU:
The new host restores guest's fpu regs to as they were before (xrstor(s)), but since the new host don't support PKRU, a general-protection exception ocurs in xrstor(s) and that crashes the guest.

This can be solved by making the guest's fpstate->user_xfeatures hold a copy of guest_supported_xcr0. This way, on 7 the only flags copied to userspace will be the ones compatible to guest requirements, and thus there will be no issue during migration.

As a bonus, it will also fail if userspace tries to set fpu features that are not compatible to the guest configuration. (KVM_SET_XSAVE ioctl)

Also, since kvm_vcpu_after_set_cpuid() now sets fpstate->user_xfeatures, there is not need to set it in kvm_check_cpuid(). So, change fpstate_realloc() so it does not touch fpstate->user_xfeatures if a non-NULL guest_fpu is passed, which is the case when kvm_check_cpuid() calls it.

---
  arch/x86/kernel/fpu/xstate.c | 5 ++++-
  arch/x86/kvm/cpuid.c         | 2 ++
  2 files changed, 6 insertions(+), 1 deletion(-)
arch/x86/kernel/fpu/xstate.c +4 -1
index 02b3ddaf4f75..7c7824ae7862 100644
@@ -1558,7 +1558,10 @@ static int fpstate_realloc(u64 xfeatures, unsigned int ksize,
  		fpregs_restore_userregs();
 
  	newfps->xfeatures = curfps->xfeatures | xfeatures;
-	newfps->user_xfeatures = curfps->user_xfeatures | xfeatures;
+
+	if (!guest_fpu)
+		newfps->user_xfeatures = curfps->user_xfeatures | xfeatures;
+
  	newfps->xfd = curfps->xfd & ~xfeatures;
 
  	/* Do the final updates within the locked region */
arch/x86/kvm/cpuid.c +0 -0
index 494d4d351859..71125291c578 100644

arch/x86/kvm/cpuid.c static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu) · line 303

Quoted 8 lines
arch/x86/kvm/cpuid.c +2 -0
@@ -296,6 +296,8 @@ static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
  	vcpu->arch.guest_supported_xcr0 =
  		cpuid_get_supported_xcr0(vcpu->arch.cpuid_entries, vcpu->arch.cpuid_nent);
 
+	vcpu->arch.guest_fpu.fpstate->user_xfeatures = vcpu->arch.guest_supported_xcr0;
+
  	kvm_update_pv_runtime(vcpu);
 
  	vcpu->arch.maxphyaddr = cpuid_query_maxphyaddr(vcpu);

dme.

#4 lore raw

PATCH v4 2/2 x86/kvm/fpu: Remove kvm_vcpu_arch.guest_supported_xcr0

In reply to Leonardo Bras

kvm_vcpu_arch currently contains the guest supported features in both guest_supported_xcr0 and guest_fpu.fpstate->user_xfeatures field.

Currently both fields are set to the same value in kvm_vcpu_after_set_cpuid() and are not changed anywhere else after that.

Since it's not good to keep duplicated data, remove guest_supported_xcr0.

To keep the code more readable, introduce kvm_guest_supported_xcr() and kvm_guest_supported_xfd() to replace the previous usages of guest_supported_xcr0.

---
 arch/x86/include/asm/kvm_host.h |  1 -
 arch/x86/kvm/cpuid.c            |  5 +++--
 arch/x86/kvm/x86.c              | 20 +++++++++++++++-----
 3 files changed, 18 insertions(+), 8 deletions(-)
arch/x86/include/asm/kvm_host.h +0 -1
index 6dcccb304775..ec9830d2aabf 100644
@@ -703,7 +703,6 @@ struct kvm_vcpu_arch {
 	struct fpu_guest guest_fpu;
 
 	u64 xcr0;
-	u64 guest_supported_xcr0;
 
 	struct kvm_pio_request pio;
 	void *pio_data;
arch/x86/kvm/cpuid.c +3 -2
index 71125291c578..b8f8d268d058 100644
@@ -282,6 +282,7 @@ static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
 {
 	struct kvm_lapic *apic = vcpu->arch.apic;
 	struct kvm_cpuid_entry2 *best;
+	u64 guest_supported_xcr0;
 
 	best = kvm_find_cpuid_entry(vcpu, 1, 0);
 	if (best && apic) {
@@ -293,10 +294,10 @@ static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
 		kvm_apic_set_version(vcpu);
 	}
 
-	vcpu->arch.guest_supported_xcr0 =
+	guest_supported_xcr0 =
 		cpuid_get_supported_xcr0(vcpu->arch.cpuid_entries, vcpu->arch.cpuid_nent);
 
-	vcpu->arch.guest_fpu.fpstate->user_xfeatures = vcpu->arch.guest_supported_xcr0;
+	vcpu->arch.guest_fpu.fpstate->user_xfeatures = guest_supported_xcr0;
 
 	kvm_update_pv_runtime(vcpu);
 
arch/x86/kvm/x86.c +15 -5
index 641044db415d..92177e2ff664 100644
@@ -984,6 +984,18 @@ void kvm_load_host_xsave_state(struct kvm_vcpu *vcpu)
 }
 EXPORT_SYMBOL_GPL(kvm_load_host_xsave_state);
 
+static inline u64 kvm_guest_supported_xcr(struct kvm_vcpu *vcpu)
+{
+	u64 guest_supported_xcr0 = vcpu->arch.guest_fpu.fpstate->user_xfeatures;
+
+	return guest_supported_xcr0;
+}
+
+static inline u64 kvm_guest_supported_xfd(struct kvm_vcpu *vcpu)
+{
+	return kvm_guest_supported_xcr(vcpu) & XFEATURE_MASK_USER_DYNAMIC;
+}
+
 static int __kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr)
 {
 	u64 xcr0 = xcr;
@@ -1003,7 +1015,7 @@ static int __kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr)
 	 * saving.  However, xcr0 bit 0 is always set, even if the
 	 * emulated CPU does not support XSAVE (see kvm_vcpu_reset()).
 	 */
-	valid_bits = vcpu->arch.guest_supported_xcr0 | XFEATURE_MASK_FP;
+	valid_bits = kvm_guest_supported_xcr(vcpu) | XFEATURE_MASK_FP;
 	if (xcr0 & ~valid_bits)
 		return 1;
 
@@ -3706,8 +3718,7 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 		    !guest_cpuid_has(vcpu, X86_FEATURE_XFD))
 			return 1;
 
-		if (data & ~(XFEATURE_MASK_USER_DYNAMIC &
-			     vcpu->arch.guest_supported_xcr0))
+		if (data & ~(kvm_guest_supported_xfd(vcpu)))
 			return 1;
 
 		fpu_update_guest_xfd(&vcpu->arch.guest_fpu, data);
@@ -3717,8 +3728,7 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 		    !guest_cpuid_has(vcpu, X86_FEATURE_XFD))
 			return 1;
 
-		if (data & ~(XFEATURE_MASK_USER_DYNAMIC &
-			     vcpu->arch.guest_supported_xcr0))
+		if (data & ~(kvm_guest_supported_xfd(vcpu)))
 			return 1;
 
 		vcpu->arch.guest_fpu.xfd_err = data;
Signature · 1 line
2.35.1
#5 lore raw

Re: [PATCH v4 2/2] x86/kvm/fpu: Remove kvm_vcpu_arch.guest_supported_xcr0

Quoted 36 lines · Leonardo Bras

kvm_vcpu_arch currently contains the guest supported features in both guest_supported_xcr0 and guest_fpu.fpstate->user_xfeatures field.

Currently both fields are set to the same value in kvm_vcpu_after_set_cpuid() and are not changed anywhere else after that.

Since it's not good to keep duplicated data, remove guest_supported_xcr0.

To keep the code more readable, introduce kvm_guest_supported_xcr() and kvm_guest_supported_xfd() to replace the previous usages of guest_supported_xcr0.

---
 arch/x86/include/asm/kvm_host.h |  1 -
 arch/x86/kvm/cpuid.c            |  5 +++--
 arch/x86/kvm/x86.c              | 20 +++++++++++++++-----
 3 files changed, 18 insertions(+), 8 deletions(-)
arch/x86/include/asm/kvm_host.h +0 -1
index 6dcccb304775..ec9830d2aabf 100644
@@ -703,7 +703,6 @@ struct kvm_vcpu_arch {
 	struct fpu_guest guest_fpu;
 
 	u64 xcr0;
-	u64 guest_supported_xcr0;
 
 	struct kvm_pio_request pio;
 	void *pio_data;
arch/x86/kvm/cpuid.c +0 -0
index 71125291c578..b8f8d268d058 100644

arch/x86/kvm/cpuid.c static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu) · line 285

Quoted 4 lines
arch/x86/kvm/cpuid.c +1 -0
@@ -282,6 +282,7 @@ static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
 {
 	struct kvm_lapic *apic = vcpu->arch.apic;
 	struct kvm_cpuid_entry2 *best;
+	u64 guest_supported_xcr0;

The intermediate variable seems unnecessary.

Quoted 21 lines
arch/x86/kvm/cpuid.c +2 -2
 	best = kvm_find_cpuid_entry(vcpu, 1, 0);
 	if (best && apic) {
@@ -293,10 +294,10 @@ static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
 		kvm_apic_set_version(vcpu);
 	}
 
-	vcpu->arch.guest_supported_xcr0 =
+	guest_supported_xcr0 =
 		cpuid_get_supported_xcr0(vcpu->arch.cpuid_entries, vcpu->arch.cpuid_nent);
 
-	vcpu->arch.guest_fpu.fpstate->user_xfeatures = vcpu->arch.guest_supported_xcr0;
+	vcpu->arch.guest_fpu.fpstate->user_xfeatures = guest_supported_xcr0;
 
 	kvm_update_pv_runtime(vcpu);
 
arch/x86/kvm/x86.c +0 -0
index 641044db415d..92177e2ff664 100644

arch/x86/kvm/x86.c void kvm_load_host_xsave_state(struct kvm_vcpu *vcpu) · line 989

Quoted 6 lines
arch/x86/kvm/x86.c +3 -0
@@ -984,6 +984,18 @@ void kvm_load_host_xsave_state(struct kvm_vcpu *vcpu)
 }
 EXPORT_SYMBOL_GPL(kvm_load_host_xsave_state);
 
+static inline u64 kvm_guest_supported_xcr(struct kvm_vcpu *vcpu)
+{
+	u64 guest_supported_xcr0 = vcpu->arch.guest_fpu.fpstate->user_xfeatures;

...and here.

Quoted 22 lines
arch/x86/kvm/x86.c +10 -1
+
+	return guest_supported_xcr0;
+}
+
+static inline u64 kvm_guest_supported_xfd(struct kvm_vcpu *vcpu)
+{
+	return kvm_guest_supported_xcr(vcpu) & XFEATURE_MASK_USER_DYNAMIC;
+}
+
 static int __kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr)
 {
 	u64 xcr0 = xcr;
@@ -1003,7 +1015,7 @@ static int __kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr)
 	 * saving.  However, xcr0 bit 0 is always set, even if the
 	 * emulated CPU does not support XSAVE (see kvm_vcpu_reset()).
 	 */
-	valid_bits = vcpu->arch.guest_supported_xcr0 | XFEATURE_MASK_FP;
+	valid_bits = kvm_guest_supported_xcr(vcpu) | XFEATURE_MASK_FP;
 	if (xcr0 & ~valid_bits)
 		return 1;
 

arch/x86/kvm/x86.c int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info) · line 3721

Quoted 6 lines
arch/x86/kvm/x86.c +1 -2
@@ -3706,8 +3718,7 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 		    !guest_cpuid_has(vcpu, X86_FEATURE_XFD))
 			return 1;
 
-		if (data & ~(XFEATURE_MASK_USER_DYNAMIC &
-			     vcpu->arch.guest_supported_xcr0))
+		if (data & ~(kvm_guest_supported_xfd(vcpu)))

Brackets could be removed...

arch/x86/kvm/x86.c int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info) · line 3731

Quoted 10 lines
arch/x86/kvm/x86.c +1 -2
 			return 1;
 
 		fpu_update_guest_xfd(&vcpu->arch.guest_fpu, data);
@@ -3717,8 +3728,7 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 		    !guest_cpuid_has(vcpu, X86_FEATURE_XFD))
 			return 1;
 
-		if (data & ~(XFEATURE_MASK_USER_DYNAMIC &
-			     vcpu->arch.guest_supported_xcr0))
+		if (data & ~(kvm_guest_supported_xfd(vcpu)))

...and here.

arch/x86/kvm/x86.c int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info) · line 3734

Quoted 3 lines
arch/x86/kvm/x86.c +0 -0
 			return 1;
 
 		vcpu->arch.guest_fpu.xfd_err = data;

dme.

Signature · 1 line
But he said, leave me alone, I'm a family man.
maintainer Applied
#6 lore raw
In reply to Leonardo Bras
Quoted 13 lines · Leonardo Bras

This patchset comes from a bug I found during qemu guest migration from a host with newer CPU to a host with an older version of this CPU, and thus having less FPU features.

When the guests were created, the one with less features is used as config, so migration is possible.

Patch 1 fix a bug that always happens during this migration, and is related to the fact that xsave saves all feature flags, but xrstor does not touch the PKRU flag. It also changes how fpstate->user_xfeatures is set, going from kvm_check_cpuid() to the later called kvm_vcpu_after_set_cpuid().

Quoted 5 lines

Patch 2 removes kvm_vcpu_arch.guest_supported_xcr0 since it now duplicates guest_fpu.fpstate->user_xfeatures. Some wrappers were introduced in order to make it easier to read the replaced version.

Patches were compile-tested, and could fix the bug found.

Queued, thanks (for 5.17 of course)! For patch 2, I renamed the function to kvm_guest_supported_xcr0.

Paolo

Quoted 30 lines

Please let me know of anything to improve!

Best regards,
Leo

--
Changes since v3:
- Add new patch to remove the use of kvm_vcpu_arch.guest_supported_xcr0, since it is now duplicating guest_fpu.fpstate->user_xfeatures.
- On patch 1, also avoid setting user_xfeatures on kvm_check_cpuid(), since it is already set in kvm_vcpu_after_set_cpuid() now.
Changes since v2:
- Fix building error because I forgot to EXPORT_SYMBOL(fpu_user_cfg) Changes since v1:
- Instead of masking xfeatures, mask user_xfeatures instead. This will only change the value sent to user, instead of the one saved in buf.
- Above change removed the need of the patch 2/2
- Instead of masking the current value of user_xfeatures, save on it fpu_user_cfg.default_features & vcpu->arch.guest_supported_xcr0

Leonardo Bras (2):
x86/kvm/fpu: Mask guest fpstate->xfeatures with guest_supported_xcr0 x86/kvm/fpu: Remove kvm_vcpu_arch.guest_supported_xcr0

  arch/x86/include/asm/kvm_host.h |  1 -
  arch/x86/kernel/fpu/xstate.c    |  5 ++++-
  arch/x86/kvm/cpuid.c            |  5 ++++-
  arch/x86/kvm/x86.c              | 20 +++++++++++++++-----
  4 files changed, 23 insertions(+), 8 deletions(-)
#7 lore raw
Quoted 19 lines · Paolo Bonzini
Quoted 17 lines · Leonardo Bras

This patchset comes from a bug I found during qemu guest migration from a host with newer CPU to a host with an older version of this CPU, and thus having less FPU features.

When the guests were created, the one with less features is used as config, so migration is possible.

Patch 1 fix a bug that always happens during this migration, and is related to the fact that xsave saves all feature flags, but xrstor does not touch the PKRU flag. It also changes how fpstate->user_xfeatures is set, going from kvm_check_cpuid() to the later called kvm_vcpu_after_set_cpuid().

Patch 2 removes kvm_vcpu_arch.guest_supported_xcr0 since it now duplicates guest_fpu.fpstate->user_xfeatures. Some wrappers were introduced in order to make it easier to read the replaced version.

Quoted 7 lines
Quoted 1 line

Patches were compile-tested, and could fix the bug found.

Queued, thanks (for 5.17 of course)! For patch 2, I renamed the function to kvm_guest_supported_xcr0.

Paolo

That's great!
Thanks Paolo!