KVM: VMX: INTR, NMI and #MC cleanup
cover KVM: VMX: INTR, NMI and #MC cleanup
This series' primarily focus is to refine VMX's handling of INTRs, NMIs and #MCs after a VM-Exit, making a few small optimizations and hopefully resulting in more readable code.
There's also a bug fix related to handling #MCs that occur during VM-Entry that was found by inspection when doing the aforementioned cleanup.
Sean Christopherson (5):
KVM: VMX: Fix handling of #MC that occurs during VM-Entry
KVM: VMX: Read cached VM-Exit reason to detect external interrupt
KVM: VMX: Store the host kernel's IDT base in a global variable
KVM: x86: Move kvm_{before,after}_interrupt() calls to vendor code
KVM: VMX: Handle NMIs, #MCs and async #PFs in common irqs-disabled fn
arch/x86/include/asm/kvm_host.h | 2 +- arch/x86/kvm/svm.c | 6 +- arch/x86/kvm/vmx/vmcs.h | 6 ++ arch/x86/kvm/vmx/vmx.c | 111 +++++++++++++++++--------------- arch/x86/kvm/vmx/vmx.h | 1 - arch/x86/kvm/x86.c | 4 +- 6 files changed, 71 insertions(+), 59 deletions(-)
Signature · 1 line
2.21.0
PATCH 1/5 KVM: VMX: Fix handling of #MC that occurs during VM-Entry
A previous fix to prevent KVM from consuming stale VMCS state after a failed VM-Entry inadvertantly blocked KVM's handling of machine checks that occur during VM-Entry.
Per Intel's SDM, a #MC during VM-Entry is handled in one of three ways, depending on when the #MC is recognoized. As it pertains to this bug fix, the third case explicitly states EXIT_REASON_MCE_DURING_VMENTRY is handled like any other VM-Exit during VM-Entry, i.e. sets bit 31 to indicate the VM-Entry failed.
If a machine-check event occurs during a VM entry, one of the following occurs:
- The machine-check event is handled as if it occurred before the VM entry:
...
- The machine-check event is handled after VM entry completes:
...- A VM-entry failure occurs as described in Section 26.7. The basic exit reason is 41, for "VM-entry failure due to machine-check event".
Explicitly handle EXIT_REASON_MCE_DURING_VMENTRY as a one-off case in vmx_vcpu_run() instead of binning it into vmx_complete_atomic_exit(). Doing so allows vmx_vcpu_run() to handle VMX_EXIT_REASONS_FAILED_VMENTRY in a sane fashion and also simplifies vmx_complete_atomic_exit() since VMCS.VM_EXIT_INTR_INFO is guaranteed to be fresh.
--- arch/x86/kvm/vmx/vmx.c | 20 ++++++++------------ 1 file changed, 8 insertions(+), 12 deletions(-)
arch/x86/kvm/vmx/vmx.c +8 -12
@@ -6103,28 +6103,21 @@ static void vmx_apicv_post_state_restore(struct kvm_vcpu *vcpu)static void vmx_complete_atomic_exit (struct vcpu_vmx *vmx) { -u32 exit_intr_info =0 ; -u16 basic_exit_reason = (u16 )vmx->exit_reason; - -if (!(basic_exit_reason ==EXIT_REASON_MCE_DURING_VMENTRY - || basic_exit_reason ==EXIT_REASON_EXCEPTION_NMI )) +if (vmx->exit_reason !=EXIT_REASON_EXCEPTION_NMI )return ; -if (!(vmx->exit_reason &VMX_EXIT_REASONS_FAILED_VMENTRY )) - exit_intr_info =vmcs_read32 (VM_EXIT_INTR_INFO ); - vmx->exit_intr_info = exit_intr_info; + vmx->exit_intr_info =vmcs_read32 (VM_EXIT_INTR_INFO );/* if exit due to PF check for async PF */ -if (is_page_fault (exit_intr_info)) +if (is_page_fault (vmx->exit_intr_info)) vmx->vcpu.arch.apf.host_apf_reason =kvm_read_and_reset_pf_reason ();/* Handle machine checks before interrupts are enabled */ -if (basic_exit_reason ==EXIT_REASON_MCE_DURING_VMENTRY || -is_machine_check (exit_intr_info)) +if (is_machine_check (vmx->exit_intr_info))kvm_machine_check ();/* We need to handle NMIs before interrupts are enabled */ -if (is_nmi (exit_intr_info)) { +if (is_nmi (vmx->exit_intr_info)) {kvm_before_interrupt (&vmx->vcpu); asm("int $2" );kvm_after_interrupt (&vmx->vcpu);
@@ -6527,6 +6520,9 @@ static void vmx_vcpu_run(struct kvm_vcpu *vcpu) vmx->idt_vectoring_info =0 ; vmx->exit_reason = vmx->fail ?0xdead :vmcs_read32 (VM_EXIT_REASON ); +if ((u16 )vmx->exit_reason ==EXIT_REASON_MCE_DURING_VMENTRY ) +kvm_machine_check (); +if (vmx->fail || (vmx->exit_reason &VMX_EXIT_REASONS_FAILED_VMENTRY ))return ;
Signature · 1 line
2.21.0
Paolo Bonzini
maintainerRe: [PATCH 1/5] KVM: VMX: Fix handling of #MC that occurs during VM-Entry
Quoted 60 lines · Sean Christopherson
A previous fix to prevent KVM from consuming stale VMCS state after a failed VM-Entry inadvertantly blocked KVM's handling of machine checks that occur during VM-Entry.
Per Intel's SDM, a #MC during VM-Entry is handled in one of three ways, depending on when the #MC is recognoized. As it pertains to this bug fix, the third case explicitly states
EXIT_REASON_MCE_DURING_VMENTRYis handled like any other VM-Exit during VM-Entry, i.e. sets bit 31 to indicate the VM-Entry failed.If a machine-check event occurs during a VM entry, one of the following occurs:
- The machine-check event is handled as if it occurred before the VM entry:... - The machine-check event is handled after VM entry completes: ...- A VM-entry failure occurs as described in Section 26.7. The basic exit reason is 41, for "VM-entry failure due to machine-check event".
Explicitly handle
EXIT_REASON_MCE_DURING_VMENTRYas a one-off case invmx_vcpu_run()instead of binning it intovmx_complete_atomic_exit(). Doing so allowsvmx_vcpu_run()to handleVMX_EXIT_REASONS_FAILED_VMENTRYin a sane fashion and also simplifiesvmx_complete_atomic_exit()since VMCS.VM_EXIT_INTR_INFO is guaranteed to be fresh.Fixes b060ca3b2e9e7 ("kvm: vmx: Handle VMLAUNCH/VMRESUME failure properly")Cc Jim Mattson <[email protected]>Signed-off-by Sean Christopherson <[email protected]>--- arch/x86/kvm/vmx/vmx.c | 20 ++++++++------------ 1 file changed, 8 insertions(+), 12 deletions(-)arch/x86/kvm/vmx/vmx.c +3
-10@@ -6103,28 +6103,21 @@ static void vmx_apicv_post_state_restore(struct kvm_vcpu *vcpu)static void vmx_complete_atomic_exit (struct vcpu_vmx *vmx) { -u32 exit_intr_info =0 ; -u16 basic_exit_reason = (u16 )vmx->exit_reason; - -if (!(basic_exit_reason ==EXIT_REASON_MCE_DURING_VMENTRY - || basic_exit_reason ==EXIT_REASON_EXCEPTION_NMI )) +if (vmx->exit_reason !=EXIT_REASON_EXCEPTION_NMI )return ; -if (!(vmx->exit_reason &VMX_EXIT_REASONS_FAILED_VMENTRY )) - exit_intr_info =vmcs_read32 (VM_EXIT_INTR_INFO ); - vmx->exit_intr_info = exit_intr_info; + vmx->exit_intr_info =vmcs_read32 (VM_EXIT_INTR_INFO );/* if exit due to PF check for async PF */ -if (is_page_fault (exit_intr_info)) +if (is_page_fault (vmx->exit_intr_info)) vmx->vcpu.arch.apf.host_apf_reason =kvm_read_and_reset_pf_reason ();/* Handle machine checks before interrupts are enabled */ - if (basic_exit_reason == EXIT_REASON_MCE_DURING_VMENTRY ||
arch/x86/kvm/vmx/vmx.c static void vmx_apicv_post_state_restore(struct kvm_vcpu *vcpu) · line 6123
Quoted 10 lines
arch/x86/kvm/vmx/vmx.c +2
-2- is_machine_check(exit_intr_info)) +if (is_machine_check (vmx->exit_intr_info))kvm_machine_check ();/* We need to handle NMIs before interrupts are enabled */ -if (is_nmi (exit_intr_info)) { +if (is_nmi (vmx->exit_intr_info)) {kvm_before_interrupt (&vmx->vcpu); asm("int $2" );kvm_after_interrupt (&vmx->vcpu);
This is indeed cleaner in addition to fixing the bug. I'm also applying this
-------------- 8< --------------
Subject: [PATCH] kvm: nVMX: small cleanup in handle_exception
From: Paolo Bonzini <[email protected]>
The reason for skipping handling of NMI and #MC in handle_exception is the same, namely they are handled earlier by vmx_complete_atomic_exit. Calling the machine check handler (which just returns 1) is misleading, don't do it.
arch/x86/kvm/vmx/vmx.c +2 -5
@@ -4455,11 +4455,8 @@ static int handle_exception(struct kvm_vcpu *vcpu) vect_info = vmx->idt_vectoring_info; intr_info = vmx->exit_intr_info; -if (is_machine_check (intr_info)) -return handle_machine_check (vcpu); - -if (is_nmi (intr_info)) -return 1 ;/* already handled by vmx_vcpu_run() */ +if (is_machine_check (intr_info) ||is_nmi (intr_info)) +return 1 ;/* already handled by vmx_complete_atomic_exit */ if (is_invalid_opcode (intr_info))return handle_ud (vcpu);
PATCH 2/5 KVM: VMX: Read cached VM-Exit reason to detect external interrupt
In reply to Sean ChristophersonGeneric x86 code blindly invokes the dedicated external interrupt handler blindly, i.e. vmx_handle_external_intr() is called on all VM-Exits regardless of the actual exit type. Use the already-cached EXIT_REASON to determine if the VM-Exit was due to an interrupt, thus avoiding an extra VMREAD (to query VM_EXIT_INTR_INFO) for all other types of VM-Exit.
In addition to avoiding the extra VMREAD, checking the EXIT_REASON instead of VM_EXIT_INTR_INFO makes it more obvious that vmx_handle_external_intr() is called for all VM-Exits, e.g. someone unfamiliar with the flow might wonder under what condition(s) VM_EXIT_INTR_INFO does not contain a valid interrupt, which is simply not possible since KVM always runs with "ack interrupt on exit".
WARN once if VM_EXIT_INTR_INFO doesn't contain a valid interrupt on an EXTERNAL_INTERRUPT VM-Exit, as such a condition would indicate a hardware bug.
--- arch/x86/kvm/vmx/vmcs.h | 6 +++++ arch/x86/kvm/vmx/vmx.c | 60 +++++++++++++++++++++-------------------- 2 files changed, 37 insertions(+), 29 deletions(-)
arch/x86/kvm/vmx/vmcs.h +6 -0
@@ -115,6 +115,12 @@ static inline bool is_nmi(u32 intr_info) == (INTR_TYPE_NMI_INTR | INTR_INFO_VALID_MASK); } +static inline bool is_external_intr (u32 intr_info) +{ +return (intr_info & (INTR_INFO_VALID_MASK |INTR_INFO_INTR_TYPE_MASK )) + == (INTR_INFO_VALID_MASK |INTR_TYPE_EXT_INTR ); +} +enum vmcs_field_width {VMCS_FIELD_WIDTH_U16 =0 ,VMCS_FIELD_WIDTH_U64 =1 ,
arch/x86/kvm/vmx/vmx.c +31 -29
73 lines not on this page · raw message
Signature · 1 line
2.21.0
Paolo Bonzini
maintainerRe: [PATCH 2/5] KVM: VMX: Read cached VM-Exit reason to detect external interrupt
Quoted 3 lines · Sean Christopherson
Generic x86 code blindly invokes the dedicated external interrupt handler blindly, i.e.
vmx_handle_external_intr()is called on all VM-Exits regardless of the actual exit type.
That's *really* blindly. :) Rephrased to
Generic x86 code invokes the kvm_x86_ops external interrupt handler on all VM-Exits regardless of the actual exit type.
- unsigned long entry;
- gate_desc *desc;
+ unsigned long entry;
I'd rather keep the desc variable to simplify review (with "diff -b") and because the code is more readable that way. Unless you have a strong reason not to do so, I can do the change when applying.
Paolo
Re: [PATCH 2/5] KVM: VMX: Read cached VM-Exit reason to detect external interrupt
Quoted 6 lines · Paolo Bonzini
Quoted 3 lines · Sean Christopherson
Generic x86 code blindly invokes the dedicated external interrupt handler blindly, i.e.
vmx_handle_external_intr()is called on all VM-Exits regardless of the actual exit type.That's *really* blindly. :) Rephrased to
Hmm, I must not have seen the first one.
Quoted 10 lines
Generic x86 code invokes the
kvm_x86_opsexternal interrupt handler on all VM-Exits regardless of the actual exit type.arch/x86/kvm/vmx/vmx.c +1
-2-unsigned long entry; -gate_desc *desc; +unsigned long entry;I'd rather keep the desc variable to simplify review (with "diff -b") and because the code is more readable that way. Unless you have a strong reason not to do so, I can do the change when applying.
No strong reason, I found the code to be more readable without it :-)
PATCH 3/5 KVM: VMX: Store the host kernel's IDT base in a global variable
In reply to Sean ChristophersonAlthough the kernel may use multiple IDTs, KVM should only ever see the "real" IDT, e.g. the early init IDT is long gone by the time KVM runs and the debug stack IDT is only used for small windows of time in very specific flows.
Before commit a547c6db4d2f1 ("KVM: VMX: Enable acknowledge interupt on vmexit"), the kernel's IDT base was consumed by KVM only when setting constant VMCS state, i.e. to set VMCS.HOST_IDTR_BASE. Because constant host state is done once per vCPU, there was ostensibly no need to cache the kernel's IDT base.
When support for "ack interrupt on exit" was introduced, KVM added a second consumer of the IDT base as handling already-acked interrupts requires directly calling the interrupt handler, i.e. KVM uses the IDT base to find the address of the handler. Because interrupts are a fast path, KVM cached the IDT base to avoid having to VMREAD HOST_IDTR_BASE. Presumably, the IDT base was cached on a per-vCPU basis simply because the existing code grabbed the IDT base on a per-vCPU (VMCS) basis.
Note, all post-boot IDTs use the same handlers for external interrupts, i.e. the "ack interrupt on exit" use of the IDT base would be unaffected even if the cached IDT somehow did not match the current IDT. And as for the original use case of setting VMCS.HOST_IDTR_BASE, if any of the above analysis is wrong then KVM has had a bug since the beginning of time since KVM has effectively been caching the IDT at vCPU creation since commit a8b732ca01c ("[PATCH] kvm: userspace interface").
--- arch/x86/kvm/vmx/vmx.c | 12 +++++++----- arch/x86/kvm/vmx/vmx.h | 1 - 2 files changed, 7 insertions(+), 6 deletions(-)
arch/x86/kvm/vmx/vmx.c +7 -5
@@ -389,6 +389,7 @@ static const struct kvm_vmx_segment_field { };u64 host_efer; +unsigned long host_idt_base;/* * Though SYSCALL is only supported in 64-bit mode on Intel CPUs, kvm
@@ -3732,7 +3733,6 @@ void vmx_set_constant_host_state(struct vcpu_vmx *vmx) {u32 low32, high32;unsigned long tmpl; -struct desc_ptr dt;unsigned long cr0, cr3, cr4; cr0 =read_cr0 ();
@@ -3768,9 +3768,7 @@ void vmx_set_constant_host_state(struct vcpu_vmx *vmx)vmcs_write16 (HOST_SS_SELECTOR , __KERNEL_DS);/* 22.2.4 */ vmcs_write16 (HOST_TR_SELECTOR ,GDT_ENTRY_TSS *8 );/* 22.2.4 */ -store_idt (&dt); -vmcs_writel (HOST_IDTR_BASE , dt.address);/* 22.2.4 */ - vmx->host_idt_base = dt.address; +vmcs_writel (HOST_IDTR_BASE , host_idt_base);/* 22.2.4 */ vmcs_writel (HOST_RIP , (unsigned long )vmx_vmexit);/* 22.2.5 */
@@ -6142,7 +6140,7 @@ static void vmx_handle_external_intr(struct kvm_vcpu *vcpu)return ; vector = intr_info &INTR_INFO_VECTOR_MASK ; - entry =gate_offset ((gate_desc *)to_vmx (vcpu)->host_idt_base + vector); + entry =gate_offset ((gate_desc *)host_idt_base + vector); asm volatile(#ifdef CONFIG_X86_64
@@ -7443,10 +7441,14 @@ static bool vmx_need_emulation_on_page_fault(struct kvm_vcpu *vcpu)static __init int hardware_setup (void ) {unsigned long host_bndcfgs; +struct desc_ptr dt;int r, i;rdmsrl_safe (MSR_EFER , &host_efer); +store_idt (&dt); + host_idt_base = dt.address; +for (i =0 ; i <ARRAY_SIZE (vmx_msr_index); ++i)kvm_define_shared_msr (i, vmx_msr_index[i]);
arch/x86/kvm/vmx/vmx.h +0 -1
@@ -184,7 +184,6 @@ struct vcpu_vmx {int nmsrs;int save_nmsrs;bool guest_msrs_dirty; -unsigned long host_idt_base;#ifdef CONFIG_X86_64 u64 msr_host_kernel_gs_base;u64 msr_guest_kernel_gs_base;
Signature · 1 line
2.21.0
kbuild test robot
build bot[RFC PATCH] KVM: VMX: host_idt_base can be static
Fixes: 94952ddb99f7 ("KVM: VMX: Store the host kernel's IDT base in a global variable")
--- vmx.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/kvm/vmx/vmx.c +1 -1
@@ -389,7 +389,7 @@ static const struct kvm_vmx_segment_field { };u64 host_efer; -unsigned long host_idt_base; +static unsigned long host_idt_base;/* * Though SYSCALL is only supported in 64-bit mode on Intel CPUs, kvm
PATCH 4/5 KVM: x86: Move kvm_{before,after}_interrupt() calls to vendor code
In reply to Sean ChristophersonVMX can conditionally call kvm_{before,after}_interrupt() since KVM always uses "ack interrupt on exit" and therefore explicitly handles interrupts as opposed to blindly enabling irqs.
--- arch/x86/kvm/svm.c | 2 ++ arch/x86/kvm/vmx/vmx.c | 4 ++++ arch/x86/kvm/x86.c | 2 -- 3 files changed, 6 insertions(+), 2 deletions(-)
arch/x86/kvm/svm.c +2 -0
@@ -6162,6 +6162,7 @@ static int svm_check_intercept(struct kvm_vcpu *vcpu,static void svm_handle_external_intr (struct kvm_vcpu *vcpu) { +kvm_before_interrupt (vcpu);local_irq_enable ();/* * We must have an instruction with interrupts enabled, so
@@ -6169,6 +6170,7 @@ static void svm_handle_external_intr(struct kvm_vcpu *vcpu)*/ asm("nop" );local_irq_disable (); +kvm_after_interrupt (vcpu); }static void svm_sched_in (struct kvm_vcpu *vcpu,int cpu)
arch/x86/kvm/vmx/vmx.c +4 -0
@@ -6142,6 +6142,8 @@ static void vmx_handle_external_intr(struct kvm_vcpu *vcpu) vector = intr_info &INTR_INFO_VECTOR_MASK ; entry =gate_offset ((gate_desc *)host_idt_base + vector); +kvm_before_interrupt (vcpu); + asmvolatile ( #ifdef CONFIG_X86_64 "mov %%" _ASM_SP ", %[sp]\n\t"
@@ -6162,6 +6164,8 @@ static void vmx_handle_external_intr(struct kvm_vcpu *vcpu) [ss]"i"(__KERNEL_DS), [cs]"i"(__KERNEL_CS) ); + +kvm_after_interrupt (vcpu); }STACK_FRAME_NON_STANDARD (vmx_handle_external_intr);
arch/x86/kvm/x86.c +0 -2
@@ -7945,9 +7945,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) vcpu->mode =OUTSIDE_GUEST_MODE ;smp_wmb (); -kvm_before_interrupt (vcpu); kvm_x86_ops->handle_external_intr (vcpu); -kvm_after_interrupt (vcpu); ++vcpu->stat.exits;
Signature · 1 line
2.21.0
PATCH 5/5 KVM: VMX: Handle NMIs, #MCs and async #PFs in common irqs-disabled fn
In reply to Sean ChristophersonPer commit 1b6269db3f833 ("KVM: VMX: Handle NMIs before enabling interrupts and preemption"), NMIs are handled directly in vmx_vcpu_run() to "make sure we handle NMI on the current cpu, and that we don't service maskable interrupts before non-maskable ones". The other exceptions handled by complete_atomic_exit(), e.g. async #PF and #MC, have similar requirements, and are located there to avoid extra VMREADs since VMX bins hardware exceptions and NMIs into a single exit reason.
Clean up the code and eliminate the vaguely named complete_atomic_exit() by moving the interrupts-disabled exception and NMI handling into the existing handle_external_intrs() callback, and rename the callback to a more appropriate name.
In addition to improving code readability, this also ensures the NMI handler is run with the host's debug registers loaded in the unlikely event that the user is debugging NMIs. Accuracy of the last_guest_tsc field is also improved when handling NMIs (and #MCs) as the handler will run after updating said field.
--- arch/x86/include/asm/kvm_host.h | 2 +- arch/x86/kvm/svm.c | 4 ++-- arch/x86/kvm/vmx/vmx.c | 25 ++++++++++++++----------- arch/x86/kvm/x86.c | 2 +- 4 files changed, 18 insertions(+), 15 deletions(-)
arch/x86/include/asm/kvm_host.h +1 -1
@@ -1109,7 +1109,7 @@ struct kvm_x86_ops {int (*check_intercept)(struct kvm_vcpu *vcpu,struct x86_instruction_info *info,enum x86_intercept_stage stage); -void (*handle_external_intr)(struct kvm_vcpu *vcpu); +void (*handle_events_irqs_disabled)(struct kvm_vcpu *vcpu);bool (*mpx_supported)(void );bool (*xsaves_supported)(void );bool (*umip_emulated)(void );
arch/x86/kvm/svm.c +2 -2
@@ -6160,7 +6160,7 @@ static int svm_check_intercept(struct kvm_vcpu *vcpu,return ret; } -static void svm_handle_external_intr (struct kvm_vcpu *vcpu) +static void svm_handle_events_irqs_disabled (struct kvm_vcpu *vcpu) {kvm_before_interrupt (vcpu);local_irq_enable ();
@@ -7256,7 +7256,7 @@ static struct kvm_x86_ops svm_x86_ops __ro_after_init = { .set_tdp_cr3 = set_tdp_cr3, .check_intercept = svm_check_intercept, - .handle_external_intr = svm_handle_external_intr, + .handle_events_irqs_disabled = svm_handle_events_irqs_disabled, .request_immediate_exit = __kvm_request_immediate_exit,
arch/x86/kvm/vmx/vmx.c +14 -11
@@ -4441,7 +4441,7 @@ static void kvm_machine_check(void)static int handle_machine_check (struct kvm_vcpu *vcpu) { -/* already handled by vcpu_run */ +/* handled by vmx_handle_events_irqs_disabled() */ return 1 ; }
@@ -4461,7 +4461,7 @@ static int handle_exception(struct kvm_vcpu *vcpu)return handle_machine_check (vcpu);if (is_nmi (intr_info)) -return 1 ;/* already handled by vmx_vcpu_run() */ +return 1 ;/* handled by vmx_handle_events_irqs_disabled() */ if (is_invalid_opcode (intr_info))return handle_ud (vcpu);
@@ -6099,11 +6099,8 @@ static void vmx_apicv_post_state_restore(struct kvm_vcpu *vcpu)memset (vmx->pi_desc.pir,0 ,sizeof (vmx->pi_desc.pir)); } -static void vmx_complete_atomic_exit (struct vcpu_vmx *vmx) +static void vmx_handle_exception_nmi_irqs_disabled (struct vcpu_vmx *vmx) { -if (vmx->exit_reason !=EXIT_REASON_EXCEPTION_NMI ) -return ; - vmx->exit_intr_info =vmcs_read32 (VM_EXIT_INTR_INFO );/* if exit due to PF check for async PF */
@@ -6131,9 +6128,6 @@ static void vmx_handle_external_intr(struct kvm_vcpu *vcpu)#endif u32 intr_info; -if (to_vmx (vcpu)->exit_reason !=EXIT_REASON_EXTERNAL_INTERRUPT ) -return ; - intr_info =vmcs_read32 (VM_EXIT_INTR_INFO );if (WARN_ONCE (!is_external_intr (intr_info),"KVM: unexpected VM-Exit interrupt info: 0x%x" , intr_info))
@@ -6169,6 +6163,16 @@ static void vmx_handle_external_intr(struct kvm_vcpu *vcpu) }STACK_FRAME_NON_STANDARD (vmx_handle_external_intr); +static void vmx_handle_events_irqs_disabled (struct kvm_vcpu *vcpu) +{ +struct vcpu_vmx *vmx =to_vmx (vcpu); + +if (vmx->exit_reason ==EXIT_REASON_EXCEPTION_NMI ) +vmx_handle_exception_nmi_irqs_disabled (vmx); +else if (vmx->exit_reason ==EXIT_REASON_EXTERNAL_INTERRUPT ) +vmx_handle_external_intr (vcpu); +} +static bool vmx_has_emulated_msr (int index) {switch (index) {
@@ -6533,7 +6537,6 @@ static void vmx_vcpu_run(struct kvm_vcpu *vcpu) vmx->loaded_vmcs->launched =1 ; vmx->idt_vectoring_info =vmcs_read32 (IDT_VECTORING_INFO_FIELD ); -vmx_complete_atomic_exit (vmx);vmx_recover_nmi_blocking (vmx);vmx_complete_interrupts (vmx); }
@@ -7708,7 +7711,7 @@ static struct kvm_x86_ops vmx_x86_ops __ro_after_init = { .set_tdp_cr3 = vmx_set_cr3, .check_intercept = vmx_check_intercept, - .handle_external_intr = vmx_handle_external_intr, + .handle_events_irqs_disabled = vmx_handle_events_irqs_disabled, .mpx_supported = vmx_mpx_supported, .xsaves_supported = vmx_xsaves_supported, .umip_emulated = vmx_umip_emulated,
arch/x86/kvm/x86.c +1 -1
@@ -7945,7 +7945,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) vcpu->mode =OUTSIDE_GUEST_MODE ;smp_wmb (); - kvm_x86_ops->handle_external_intr (vcpu); + kvm_x86_ops->handle_events_irqs_disabled (vcpu); ++vcpu->stat.exits;
Signature · 1 line
2.21.0
Paolo Bonzini
maintainerRe: [PATCH 5/5] KVM: VMX: Handle NMIs, #MCs and async #PFs in common irqs-disabled fn
Quoted 13 lines · Sean Christopherson
Per commit
1b6269db3f833("KVM: VMX: Handle NMIs before enabling interrupts and preemption"), NMIs are handled directly invmx_vcpu_run()to "make sure we handle NMI on the current cpu, and that we don't service maskable interrupts before non-maskable ones". The other exceptions handled bycomplete_atomic_exit(), e.g. async #PF and #MC, have similar requirements, and are located there to avoid extra VMREADs since VMX bins hardware exceptions and NMIs into a single exit reason.Clean up the code and eliminate the vaguely named
complete_atomic_exit()by moving the interrupts-disabled exception and NMI handling into the existinghandle_external_intrs()callback, and rename the callback to a more appropriate name.
Quoted 7 lines
In addition to improving code readability, this also ensures the NMI handler is run with the host's debug registers loaded in the unlikely event that the user is debugging NMIs. Accuracy of the
last_guest_tscfield is also improved when handling NMIs (and #MCs) as the handler will run after updating said field.Signed-off-by Sean Christopherson <[email protected]>
Very nice, just some changes I'd like to propose. "atomic" is Linux lingo for "irqs disabled", so I'd like to rename the handler to handle_exit_atomic so it has a correspondance with handle_exit. Likewise we could have handle_exception_nmi_atomic and handle_external_interrupt_atomic.
Putting everything together we get:
arch/x86/include/asm/kvm_host.h +1 -1
@@ -1117,7 +1117,7 @@ struct kvm_x86_ops {int (*check_intercept)(struct kvm_vcpu *vcpu,struct x86_instruction_info *info,enum x86_intercept_stage stage); -void (*handle_external_intr)(struct kvm_vcpu *vcpu); +void (*handle_exit_atomic)(struct kvm_vcpu *vcpu);bool (*mpx_supported)(void );bool (*xsaves_supported)(void );bool (*umip_emulated)(void );
arch/x86/kvm/svm.c +2 -2
@@ -6172,7 +6172,7 @@ static int svm_check_intercept(struct kvm_vcpu *vcpu,return ret; } -static void svm_handle_external_intr (struct kvm_vcpu *vcpu) +static void svm_handle_exit_atomic (struct kvm_vcpu *vcpu) {kvm_before_interrupt (vcpu);local_irq_enable ();
@@ -7268,7 +7268,7 @@ static bool svm_need_emulation_on_page_fault(struct kvm_vcpu *vcpu) .set_tdp_cr3 = set_tdp_cr3, .check_intercept = svm_check_intercept, - .handle_external_intr = svm_handle_external_intr, + .handle_exit_atomic = svm_handle_exit_atomic, .request_immediate_exit = __kvm_request_immediate_exit,
arch/x86/kvm/vmx/vmx.c +18 -15
@@ -4437,11 +4437,11 @@ static void kvm_machine_check(void)static int handle_machine_check (struct kvm_vcpu *vcpu) { -/* already handled by vcpu_run */ +/* handled by vmx_vcpu_run() */ return 1 ; } -static int handle_exception (struct kvm_vcpu *vcpu) +static int handle_exception_nmi (struct kvm_vcpu *vcpu) {struct vcpu_vmx *vmx =to_vmx (vcpu);struct kvm_run *kvm_run = vcpu->run;
@@ -4454,7 +4454,7 @@ static int handle_exception(struct kvm_vcpu *vcpu) intr_info = vmx->exit_intr_info;if (is_machine_check (intr_info) ||is_nmi (intr_info)) -return 1 ;/* already handled by vmx_complete_atomic_exit */ +return 1 ;/* handled by handle_exception_nmi_atomic() */ if (is_invalid_opcode (intr_info))return handle_ud (vcpu);
@@ -5462,7 +5462,7 @@ static int handle_encls(struct kvm_vcpu *vcpu)* to be done to userspace and return 0. */ static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu) = { - [EXIT_REASON_EXCEPTION_NMI ] = handle_exception, + [EXIT_REASON_EXCEPTION_NMI ] = handle_exception_nmi, [EXIT_REASON_EXTERNAL_INTERRUPT ] = handle_external_interrupt, [EXIT_REASON_TRIPLE_FAULT ] = handle_triple_fault, [EXIT_REASON_NMI_WINDOW ] = handle_nmi_window,
@@ -6100,11 +6100,8 @@ static void vmx_apicv_post_state_restore(struct kvm_vcpu *vcpu)memset (vmx->pi_desc.pir,0 ,sizeof (vmx->pi_desc.pir)); } -static void vmx_complete_atomic_exit (struct vcpu_vmx *vmx) +static void handle_exception_nmi_atomic (struct vcpu_vmx *vmx) { -if (vmx->exit_reason !=EXIT_REASON_EXCEPTION_NMI ) -return ; - vmx->exit_intr_info =vmcs_read32 (VM_EXIT_INTR_INFO );/* if exit due to PF check for async PF */
@@ -6123,7 +6120,7 @@ static void vmx_complete_atomic_exit(struct vcpu_vmx *vmx) } } -static void vmx_handle_external_intr (struct kvm_vcpu *vcpu) +static void handle_external_interrupt_atomic (struct kvm_vcpu *vcpu) {unsigned int vector;unsigned long entry;
@@ -6133,9 +6130,6 @@ static void vmx_handle_external_intr(struct kvm_vcpu *vcpu)gate_desc *desc;u32 intr_info; -if (to_vmx (vcpu)->exit_reason !=EXIT_REASON_EXTERNAL_INTERRUPT ) -return ; - intr_info =vmcs_read32 (VM_EXIT_INTR_INFO );if (WARN_ONCE (!is_external_intr (intr_info),"KVM: unexpected VM-Exit interrupt info: 0x%x" , intr_info))
@@ -6170,7 +6164,17 @@ static void vmx_handle_external_intr(struct kvm_vcpu *vcpu)kvm_after_interrupt (vcpu); } -STACK_FRAME_NON_STANDARD (vmx_handle_external_intr); +STACK_FRAME_NON_STANDARD (handle_external_interrupt_atomic); + +static void vmx_handle_exit_atomic (struct kvm_vcpu *vcpu) +{ +struct vcpu_vmx *vmx =to_vmx (vcpu); + +if (vmx->exit_reason ==EXIT_REASON_EXTERNAL_INTERRUPT ) +handle_external_interrupt_atomic (vcpu); +else if (vmx->exit_reason ==EXIT_REASON_EXCEPTION_NMI ) +handle_exception_nmi_atomic (vmx); +}static bool vmx_has_emulated_msr (int index) {
@@ -6540,7 +6544,6 @@ static void vmx_vcpu_run(struct kvm_vcpu *vcpu) vmx->loaded_vmcs->launched =1 ; vmx->idt_vectoring_info =vmcs_read32 (IDT_VECTORING_INFO_FIELD ); -vmx_complete_atomic_exit (vmx);vmx_recover_nmi_blocking (vmx);vmx_complete_interrupts (vmx); }
@@ -7694,7 +7697,7 @@ static __exit void hardware_unsetup(void) .set_tdp_cr3 = vmx_set_cr3, .check_intercept = vmx_check_intercept, - .handle_external_intr = vmx_handle_external_intr, + .handle_exit_atomic = vmx_handle_exit_atomic, .mpx_supported = vmx_mpx_supported, .xsaves_supported = vmx_xsaves_supported, .umip_emulated = vmx_umip_emulated,
arch/x86/kvm/x86.c +1 -1
@@ -7999,7 +7999,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) vcpu->mode =OUTSIDE_GUEST_MODE ;smp_wmb (); - kvm_x86_ops->handle_external_intr (vcpu); + kvm_x86_ops->handle_exit_atomic (vcpu); ++vcpu->stat.exits;
Re: [PATCH 5/5] KVM: VMX: Handle NMIs, #MCs and async #PFs in common irqs-disabled fn
Quoted 20 lines · Paolo Bonzini
Quoted 19 lines · Sean Christopherson
Per commit
1b6269db3f833("KVM: VMX: Handle NMIs before enabling interrupts and preemption"), NMIs are handled directly invmx_vcpu_run()to "make sure we handle NMI on the current cpu, and that we don't service maskable interrupts before non-maskable ones". The other exceptions handled bycomplete_atomic_exit(), e.g. async #PF and #MC, have similar requirements, and are located there to avoid extra VMREADs since VMX bins hardware exceptions and NMIs into a single exit reason.Clean up the code and eliminate the vaguely named
complete_atomic_exit()by moving the interrupts-disabled exception and NMI handling into the existinghandle_external_intrs()callback, and rename the callback to a more appropriate name.In addition to improving code readability, this also ensures the NMI handler is run with the host's debug registers loaded in the unlikely event that the user is debugging NMIs. Accuracy of the
last_guest_tscfield is also improved when handling NMIs (and #MCs) as the handler will run after updating said field.
Quoted 4 lines
Quoted 1 line
Signed-off-by Sean Christopherson <[email protected]>Very nice, just some changes I'd like to propose. "atomic" is Linux lingo for "irqs disabled", so I'd like to rename the handler to
The code disagrees, e.g.
/*
* Are we running in atomic context? WARNING: this macro cannot
* always detect atomic context; in particular, it cannot know about
* held spinlocks in non-preemptible kernels. Thus it should not be
* used in the general case to determine whether sleeping is possible.
* Do not use in_atomic() in driver code.
*/
#define in_atomic() (preempt_count() != 0)
and
void ___might_sleep(...)
{
...
printk(KERN_ERR
"in_atomic(): %d, irqs_disabled(): %d, pid: %d, name: %s\n",
in_atomic(), irqs_disabled(),
current->pid, current->comm);}
and
static inline void *kmap_atomic(struct page *page)
{
preempt_disable();
pagefault_disable();
return page_address(page);}
My interpretation of things is that the kernel's definition of an atomic context is with respect to preemption. Disabling IRQs would also provide atomicity, but the reverse is not true, i.e. entering an atomic context does not imply IRQs are disabled.
As it pertains to KVM, we specifically care about IRQs being disabled, e.g. VMX needs to ensure #MC and NMI are handled before any pending IRQs, and both VMX and SVM need to ensure a pending perf interrupt is handled in the callback.
And if "atomic" is interpreted as "IRQs disabled", one could argue that the SVM behavior is buggy since enabling IRQs would break atomicity.
Quoted 3 lines
handle_exit_atomicso it has a correspondance withhandle_exit. Likewise we could havehandle_exception_nmi_atomicandhandle_external_interrupt_atomic.
Paolo Bonzini
maintainerRe: [PATCH 5/5] KVM: VMX: Handle NMIs, #MCs and async #PFs in common irqs-disabled fn
Quoted 10 lines · Sean Christopherson
The code disagrees, e.g.
/*
* Are we running in atomic context? WARNING: this macro cannot
* always detect atomic context; in particular, it cannot know about
* held spinlocks in non-preemptible kernels. Thus it should not be
* used in the general case to determine whether sleeping is possible.
* Do not usein_atomic()in driver code.
*/
#definein_atomic()(preempt_count()!= 0)
You're totally right. "_irqoff" seems to be the common suffix for irq-disabled functions.
Paolo