[RESEND,v4,23/23] KVM: x86: smm: preserve interrupt shadow in SMRAM

Message ID 20221025124741.228045-24-mlevitsk@redhat.com
State New
Headers
Series SMM emulation and interrupt shadow fixes |

Commit Message

Maxim Levitsky Oct. 25, 2022, 12:47 p.m. UTC
  When #SMI is asserted, the CPU can be in interrupt shadow due to sti or
mov ss.

It is not mandatory in  Intel/AMD prm to have the #SMI blocked during the
shadow, and on top of that, since neither SVM nor VMX has true support
for SMI window, waiting for one instruction would mean single stepping
the guest.

Instead, allow #SMI in this case, but both reset the interrupt window and
stash its value in SMRAM to restore it on exit from SMM.

This fixes rare failures seen mostly on windows guests on VMX, when #SMI
falls on the sti instruction which mainfest in VM entry failure due
to EFLAGS.IF not being set, but STI interrupt window still being set
in the VMCS.

Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
---
 arch/x86/kvm/smm.c | 24 +++++++++++++++++++++---
 arch/x86/kvm/smm.h |  5 +++--
 2 files changed, 24 insertions(+), 5 deletions(-)
  

Comments

Paolo Bonzini Oct. 28, 2022, 10:35 a.m. UTC | #1
On 10/25/22 14:47, Maxim Levitsky wrote:
> @@ -19,7 +19,8 @@ struct kvm_smram_state_32 {
>   	u32 reserved1[62];
>   	u32 smbase;
>   	u32 smm_revision;
> -	u32 reserved2[5];
> +	u32 reserved2[4];
> +	u32 int_shadow; /* KVM extension */
>   	u32 cr4; /* CR4 is not present in Intel/AMD SMRAM image */
>   	u32 reserved3[5];

Of course the placement of KVM-specific fields is somewhat arbitrary, 
but based on sandpile.org data I would place it at 0xFF1A ("reserved", 
you have to search for 7F1Ah in the web page).

> @@ -86,7 +87,7 @@ struct kvm_smram_state_64 {
>   	u64 io_restart_rsi;
>   	u64 io_restart_rdi;
>   	u32 io_restart_dword;
> -	u32 reserved1;
> +	u32 int_shadow;
>   	u8 io_inst_restart;
>   	u8 auto_hlt_restart;
>   	u8 reserved2[6];

Likewise, based on AMD BKDG I would place this at 0xFECB after the "NMI 
Mask" field (which unfortunately I learnt about only after "inventing" 
HF_SMM_INSIDE_NMI_MASK).

I can do the changes myself, but please ack.

Paolo
  
Maxim Levitsky Oct. 30, 2022, 8:23 a.m. UTC | #2
On Fri, 2022-10-28 at 12:35 +0200, Paolo Bonzini wrote:
> On 10/25/22 14:47, Maxim Levitsky wrote:
> > @@ -19,7 +19,8 @@ struct kvm_smram_state_32 {
> >         u32 reserved1[62];
> >         u32 smbase;
> >         u32 smm_revision;
> > -       u32 reserved2[5];
> > +       u32 reserved2[4];
> > +       u32 int_shadow; /* KVM extension */
> >         u32 cr4; /* CR4 is not present in Intel/AMD SMRAM image */
> >         u32 reserved3[5];
> 
> Of course the placement of KVM-specific fields is somewhat arbitrary, 
> but based on sandpile.org data I would place it at 0xFF1A ("reserved", 
> you have to search for 7F1Ah in the web page).
> 
> > @@ -86,7 +87,7 @@ struct kvm_smram_state_64 {
> >         u64 io_restart_rsi;
> >         u64 io_restart_rdi;
> >         u32 io_restart_dword;
> > -       u32 reserved1;
> > +       u32 int_shadow;
> >         u8 io_inst_restart;
> >         u8 auto_hlt_restart;
> >         u8 reserved2[6];
> 
> Likewise, based on AMD BKDG I would place this at 0xFECB after the "NMI 
> Mask" field (which unfortunately I learnt about only after "inventing" 
> HF_SMM_INSIDE_NMI_MASK).

I don't see any problem with this, makes sense.

I wish AMD would keep on releaseing the BKDG - I haven't looked there
because last public version is very old.


Thanks!
Best regards,
	Maxim Levitsky

> 
> I can do the changes myself, but please ack.
> 
> Paolo
>
  

Patch

diff --git a/arch/x86/kvm/smm.c b/arch/x86/kvm/smm.c
index 82761384a8664d..46d2656937a71c 100644
--- a/arch/x86/kvm/smm.c
+++ b/arch/x86/kvm/smm.c
@@ -21,6 +21,7 @@  static void check_smram_offsets(void)
 	CHECK_SMRAM32_OFFSET(smbase,			0xFEF8);
 	CHECK_SMRAM32_OFFSET(smm_revision,		0xFEFC);
 	CHECK_SMRAM32_OFFSET(reserved2,			0xFF00);
+	CHECK_SMRAM32_OFFSET(int_shadow,		0xFF10);
 	CHECK_SMRAM32_OFFSET(cr4,			0xFF14);
 	CHECK_SMRAM32_OFFSET(reserved3,			0xFF18);
 	CHECK_SMRAM32_OFFSET(ds,			0xFF2C);
@@ -65,7 +66,7 @@  static void check_smram_offsets(void)
 	CHECK_SMRAM64_OFFSET(io_restart_rsi,		0xFEB0);
 	CHECK_SMRAM64_OFFSET(io_restart_rdi,		0xFEB8);
 	CHECK_SMRAM64_OFFSET(io_restart_dword,		0xFEC0);
-	CHECK_SMRAM64_OFFSET(reserved1,			0xFEC4);
+	CHECK_SMRAM64_OFFSET(int_shadow,		0xFEC4);
 	CHECK_SMRAM64_OFFSET(io_inst_restart,		0xFEC8);
 	CHECK_SMRAM64_OFFSET(auto_hlt_restart,		0xFEC9);
 	CHECK_SMRAM64_OFFSET(reserved2,			0xFECA);
@@ -212,6 +213,8 @@  static void enter_smm_save_state_32(struct kvm_vcpu *vcpu,
 	smram->cr4 = kvm_read_cr4(vcpu);
 	smram->smm_revision = 0x00020000;
 	smram->smbase = vcpu->arch.smbase;
+
+	smram->int_shadow = static_call(kvm_x86_get_interrupt_shadow)(vcpu);
 }
 
 #ifdef CONFIG_X86_64
@@ -261,6 +264,8 @@  static void enter_smm_save_state_64(struct kvm_vcpu *vcpu,
 	enter_smm_save_seg_64(vcpu, &smram->ds, VCPU_SREG_DS);
 	enter_smm_save_seg_64(vcpu, &smram->fs, VCPU_SREG_FS);
 	enter_smm_save_seg_64(vcpu, &smram->gs, VCPU_SREG_GS);
+
+	smram->int_shadow = static_call(kvm_x86_get_interrupt_shadow)(vcpu);
 }
 #endif
 
@@ -306,6 +311,8 @@  void enter_smm(struct kvm_vcpu *vcpu)
 	kvm_set_rflags(vcpu, X86_EFLAGS_FIXED);
 	kvm_rip_write(vcpu, 0x8000);
 
+	static_call(kvm_x86_set_interrupt_shadow)(vcpu, 0);
+
 	cr0 = vcpu->arch.cr0 & ~(X86_CR0_PE | X86_CR0_EM | X86_CR0_TS | X86_CR0_PG);
 	static_call(kvm_x86_set_cr0)(vcpu, cr0);
 	vcpu->arch.cr0 = cr0;
@@ -453,7 +460,7 @@  static int rsm_load_state_32(struct x86_emulate_ctxt *ctxt,
 {
 	struct kvm_vcpu *vcpu = ctxt->vcpu;
 	struct desc_ptr dt;
-	int i;
+	int i, r;
 
 	ctxt->eflags =  smstate->eflags | X86_EFLAGS_FIXED;
 	ctxt->_eip =  smstate->eip;
@@ -487,8 +494,16 @@  static int rsm_load_state_32(struct x86_emulate_ctxt *ctxt,
 
 	vcpu->arch.smbase = smstate->smbase;
 
-	return rsm_enter_protected_mode(vcpu, smstate->cr0,
+	r = rsm_enter_protected_mode(vcpu, smstate->cr0,
 					smstate->cr3, smstate->cr4);
+
+	if (r != X86EMUL_CONTINUE)
+		return r;
+
+	static_call(kvm_x86_set_interrupt_shadow)(vcpu, 0);
+	ctxt->interruptibility = (u8)smstate->int_shadow;
+
+	return r;
 }
 
 #ifdef CONFIG_X86_64
@@ -539,6 +554,9 @@  static int rsm_load_state_64(struct x86_emulate_ctxt *ctxt,
 	rsm_load_seg_64(vcpu, &smstate->fs, VCPU_SREG_FS);
 	rsm_load_seg_64(vcpu, &smstate->gs, VCPU_SREG_GS);
 
+	static_call(kvm_x86_set_interrupt_shadow)(vcpu, 0);
+	ctxt->interruptibility = (u8)smstate->int_shadow;
+
 	return X86EMUL_CONTINUE;
 }
 #endif
diff --git a/arch/x86/kvm/smm.h b/arch/x86/kvm/smm.h
index 8d96bff3f4d54f..2eaec53bcc9504 100644
--- a/arch/x86/kvm/smm.h
+++ b/arch/x86/kvm/smm.h
@@ -19,7 +19,8 @@  struct kvm_smram_state_32 {
 	u32 reserved1[62];
 	u32 smbase;
 	u32 smm_revision;
-	u32 reserved2[5];
+	u32 reserved2[4];
+	u32 int_shadow; /* KVM extension */
 	u32 cr4; /* CR4 is not present in Intel/AMD SMRAM image */
 	u32 reserved3[5];
 
@@ -86,7 +87,7 @@  struct kvm_smram_state_64 {
 	u64 io_restart_rsi;
 	u64 io_restart_rdi;
 	u32 io_restart_dword;
-	u32 reserved1;
+	u32 int_shadow;
 	u8 io_inst_restart;
 	u8 auto_hlt_restart;
 	u8 reserved2[6];