diff mbox series

[09/11] KVM: SVM: restore host save area from assembly

Message ID 20221109145156.84714-10-pbonzini@redhat.com
State New
Headers
Series KVM: SVM: fixes for vmentry code |

Commit Message

Paolo Bonzini Nov. 9, 2022, 2:51 p.m. UTC 
  Allow access to the percpu area via the GS segment base, which is
needed in order to access the saved host spec_ctrl value.  In linux-next
FILL_RETURN_BUFFER also needs to access percpu data.

For simplicity, the physical address of the save area is added to struct
svm_cpu_data.

Cc: stable@vger.kernel.org
Fixes: a149180fbcf3 ("x86: Add magic AMD return-thunk")
Reported-by: Nathan Chancellor <nathan@kernel.org>
Analyzed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 arch/x86/kvm/kvm-asm-offsets.c |  1 +
 arch/x86/kvm/svm/svm.c         | 14 ++++++--------
 arch/x86/kvm/svm/svm.h         |  2 ++
 arch/x86/kvm/svm/svm_ops.h     |  5 -----
 arch/x86/kvm/svm/vmenter.S     | 17 +++++++++++++++++
 5 files changed, 26 insertions(+), 13 deletions(-)

Comments

Sean Christopherson Nov. 9, 2022, 3:54 p.m. UTC | #1 
On Wed, Nov 09, 2022, Paolo Bonzini wrote:
> Allow access to the percpu area via the GS segment base, which is
> needed in order to access the saved host spec_ctrl value.  In linux-next
> FILL_RETURN_BUFFER also needs to access percpu data.
> 
> For simplicity, the physical address of the save area is added to struct
> svm_cpu_data.
> 
> Cc: stable@vger.kernel.org
> Fixes: a149180fbcf3 ("x86: Add magic AMD return-thunk")
> Reported-by: Nathan Chancellor <nathan@kernel.org>
> Analyzed-by: Andrew Cooper <andrew.cooper3@citrix.com>
> Tested-by: Nathan Chancellor <nathan@kernel.org>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---

Reviewed-by: Sean Christopherson <seanjc@google.com>

> diff --git a/arch/x86/kvm/svm/svm.h b/arch/x86/kvm/svm/svm.h
> index 2af6a71126c1..83955a4e520e 100644
> --- a/arch/x86/kvm/svm/svm.h
> +++ b/arch/x86/kvm/svm/svm.h
> @@ -287,6 +287,8 @@ struct svm_cpu_data {
>  	struct kvm_ldttss_desc *tss_desc;
>  
>  	struct page *save_area;
> +	unsigned long save_area_pa;

I really dislike storing both the page and the address, but that's more about
storing the "struct page" instead of the virtual address, and that can be cleaned
up in a follow-up series.

Specifically, the ugly pointer arithmetic in svm_prepare_switch_to_guest() can
be avoided by updating "struct vmcb" to capture SEV-ES+, and by tracking the save
area as a VMCB (which it is).

E.g. as a very partial patch

diff --git a/arch/x86/include/asm/svm.h b/arch/x86/include/asm/svm.h
index 0361626841bc..64ba98d32689 100644
--- a/arch/x86/include/asm/svm.h
+++ b/arch/x86/include/asm/svm.h
@@ -513,7 +513,10 @@ static inline void __unused_size_checks(void)
 
 struct vmcb {
        struct vmcb_control_area control;
-       struct vmcb_save_area save;
+       union {
+               struct sev_es_save_area sev_es_save;
+               struct vmcb_save_area save;
+       }
 } __packed;
 
 #define SVM_CPUID_FUNC 0x8000000a
diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index 9f88c8e6766e..b23b7633033b 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -1462,12 +1462,8 @@ static void svm_prepare_switch_to_guest(struct kvm_vcpu *vcpu)
         * or subsequent vmload of host save area.
         */
        vmsave(sd->save_area_pa);
-       if (sev_es_guest(vcpu->kvm)) {
-               struct sev_es_save_area *hostsa;
-               hostsa = (struct sev_es_save_area *)(page_address(sd->save_area) + 0x400);
-
-               sev_es_prepare_switch_to_guest(hostsa);
-       }
+       if (sev_es_guest(vcpu->kvm))
+               sev_es_prepare_switch_to_guest(sd->save_area->sev_es_save);
 
        if (tsc_scaling)
                __svm_write_tsc_multiplier(vcpu->arch.tsc_scaling_ratio);
diff --git a/arch/x86/kvm/svm/svm.h b/arch/x86/kvm/svm/svm.h
index 199a2ecef1ce..802ed393d860 100644
--- a/arch/x86/kvm/svm/svm.h
+++ b/arch/x86/kvm/svm/svm.h
@@ -286,7 +286,7 @@ struct svm_cpu_data {
        u32 min_asid;
        struct kvm_ldttss_desc *tss_desc;
 
-       struct page *save_area;
+       struct vmcb *save_area;
        unsigned long save_area_pa;
 
        struct vmcb *current_vmcb;
diff mbox series

Patch

diff --git a/arch/x86/kvm/kvm-asm-offsets.c b/arch/x86/kvm/kvm-asm-offsets.c
index f83e88b85bf2..1b805cd24d66 100644
--- a/arch/x86/kvm/kvm-asm-offsets.c
+++ b/arch/x86/kvm/kvm-asm-offsets.c
@@ -18,6 +18,7 @@  static void __used common(void)
 		OFFSET(SVM_current_vmcb, vcpu_svm, current_vmcb);
 		OFFSET(SVM_vmcb01, vcpu_svm, vmcb01);
 		OFFSET(KVM_VMCB_pa, kvm_vmcb_info, pa);
+		OFFSET(SD_save_area_pa, svm_cpu_data, save_area_pa);
 	}
 
 	if (IS_ENABLED(CONFIG_KVM_INTEL)) {
diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index 4e3a47eb5002..469c1b5617af 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -592,7 +592,7 @@  static int svm_hardware_enable(void)
 
 	wrmsrl(MSR_EFER, efer | EFER_SVME);
 
-	wrmsrl(MSR_VM_HSAVE_PA, __sme_page_pa(sd->save_area));
+	wrmsrl(MSR_VM_HSAVE_PA, sd->save_area_pa);
 
 	if (static_cpu_has(X86_FEATURE_TSCRATEMSR)) {
 		/*
@@ -648,6 +648,7 @@  static void svm_cpu_uninit(int cpu)
 
 	kfree(sd->sev_vmcbs);
 	__free_page(sd->save_area);
+	sd->save_area_pa = 0;
 	sd->save_area = NULL;
 }
 
@@ -665,6 +666,7 @@  static int svm_cpu_init(int cpu)
 	if (ret)
 		goto free_save_area;
 
+	sd->save_area_pa = __sme_page_pa(sd->save_area);
 	return 0;
 
 free_save_area:
@@ -1450,7 +1452,7 @@  static void svm_prepare_switch_to_guest(struct kvm_vcpu *vcpu)
 	 * Save additional host state that will be restored on VMEXIT (sev-es)
 	 * or subsequent vmload of host save area.
 	 */
-	vmsave(__sme_page_pa(sd->save_area));
+	vmsave(sd->save_area_pa);
 	if (sev_es_guest(vcpu->kvm)) {
 		struct sev_es_save_area *hostsa;
 		hostsa = (struct sev_es_save_area *)(page_address(sd->save_area) + 0x400);
@@ -3905,14 +3907,10 @@  static noinstr void svm_vcpu_enter_exit(struct kvm_vcpu *vcpu)
 
 	guest_state_enter_irqoff();
 
-	if (sev_es_guest(vcpu->kvm)) {
+	if (sev_es_guest(vcpu->kvm))
 		__svm_sev_es_vcpu_run(svm);
-	} else {
-		struct svm_cpu_data *sd = per_cpu_ptr(&svm_data, vcpu->cpu);
-
+	else
 		__svm_vcpu_run(svm);
-		vmload(__sme_page_pa(sd->save_area));
-	}
 
 	guest_state_exit_irqoff();
 }
diff --git a/arch/x86/kvm/svm/svm.h b/arch/x86/kvm/svm/svm.h
index 2af6a71126c1..83955a4e520e 100644
--- a/arch/x86/kvm/svm/svm.h
+++ b/arch/x86/kvm/svm/svm.h
@@ -287,6 +287,8 @@  struct svm_cpu_data {
 	struct kvm_ldttss_desc *tss_desc;
 
 	struct page *save_area;
+	unsigned long save_area_pa;
+
 	struct vmcb *current_vmcb;
 
 	/* index = sev_asid, value = vmcb pointer */
diff --git a/arch/x86/kvm/svm/svm_ops.h b/arch/x86/kvm/svm/svm_ops.h
index 9430d6437c9f..36c8af87a707 100644
--- a/arch/x86/kvm/svm/svm_ops.h
+++ b/arch/x86/kvm/svm/svm_ops.h
@@ -61,9 +61,4 @@  static __always_inline void vmsave(unsigned long pa)
 	svm_asm1(vmsave, "a" (pa), "memory");
 }
 
-static __always_inline void vmload(unsigned long pa)
-{
-	svm_asm1(vmload, "a" (pa), "memory");
-}
-
 #endif /* __KVM_X86_SVM_OPS_H */
diff --git a/arch/x86/kvm/svm/vmenter.S b/arch/x86/kvm/svm/vmenter.S
index 5bc2ed7d79c0..57440acfc73e 100644
--- a/arch/x86/kvm/svm/vmenter.S
+++ b/arch/x86/kvm/svm/vmenter.S
@@ -49,6 +49,14 @@  SYM_FUNC_START(__svm_vcpu_run)
 #endif
 	push %_ASM_BX
 
+	/*
+	 * Save variables needed after vmexit on the stack, in inverse
+	 * order compared to when they are needed.
+	 */
+
+	/* Needed to restore access to percpu variables.  */
+	__ASM_SIZE(push) PER_CPU_VAR(svm_data + SD_save_area_pa)
+
 	/* Save @svm. */
 	push %_ASM_ARG1
 
@@ -124,6 +132,11 @@  SYM_FUNC_START(__svm_vcpu_run)
 5:	vmsave %_ASM_AX
 6:
 
+	/* Restores GSBASE among other things, allowing access to percpu data.  */
+	pop %_ASM_AX
+7:	vmload %_ASM_AX
+8:
+
 #ifdef CONFIG_RETPOLINE
 	/* IMPORTANT: Stuff the RSB immediately after VM-Exit, before RET! */
 	FILL_RETURN_BUFFER %_ASM_AX, RSB_CLEAR_LOOPS, X86_FEATURE_RETPOLINE
@@ -187,10 +200,14 @@  SYM_FUNC_START(__svm_vcpu_run)
 50:	cmpb $0, kvm_rebooting
 	jne 6b
 	ud2
+70:	cmpb $0, kvm_rebooting
+	jne 8b
+	ud2
 
 	_ASM_EXTABLE(1b, 10b)
 	_ASM_EXTABLE(3b, 30b)
 	_ASM_EXTABLE(5b, 50b)
+	_ASM_EXTABLE(7b, 70b)
 
 SYM_FUNC_END(__svm_vcpu_run)