| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Fri, 13 Oct 2017 18:15:00 +0800
From: "Du, Changbin" <changbin.du@...el.com>
To: "Du, Changbin" <changbin.du@...el.com>
Cc: Arnaldo Carvalho de Melo <acme@...nel.org>, peterz@...radead.org,
mingo@...hat.com, alexander.shishkin@...ux.intel.com,
linux-kernel@...r.kernel.org, jolsa@...nel.org
Subject: Re: Does perf-annotate work correctly?
Hi Jiri,
Sorry, missed you (but get_maintainer.pl doesn't list you). Here is ealier email.
https://lkml.org/lkml/2017/9/12/158
Do you think if this is a real issue?
btw, is their a dedicated mailist for perf? Thanks.
On Wed, Sep 13, 2017 at 05:14:09PM +0800, Du, Changbin wrote:
>
> Here is another example which may be easy to analyse. I beleve that something wrong
> in parsing objdump output.
>
> The 'test %esi,%esi' is not of 'if (unlikely(error))', but 'if (!idtv_info_valid)'.
>
>
> perf annotate of function vmx_complete_interrupts():
> vmx_complete_interrupts /lib/modules/4.13.0+/kernel/arch/x86/kvm/kvm-intel.ko
> Percent│ mov 0x4590(%rdi),%esi ▒
> │ vmx_cancel_injection(): ▒
> │ vmcs_read32(VM_ENTRY_INTR_INFO_FIELD), ▒
> │ VM_ENTRY_INSTRUCTION_LEN, ▒
> │ VM_ENTRY_EXCEPTION_ERROR_CODE); ▒
> │ ▒
> │ vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0); ▒
> │ } ▒
> │ movb $0x0,0x32c8(%rdi) ▒
> │ __vmx_complete_interrupts(): ▒
> │ movb $0x0,0x1a7c(%rdi) ▒
> │ movb $0x0,0x1a88(%rdi) ▒
> │ __vmcs_writel(): ▒
> │ u8 error; ▒
> │ ▒
> │ asm volatile (__ex(ASM_VMX_VMWRITE_RAX_RDX) "; setna %0" ▒
> │ : "=q"(error) : "a"(value), "d"(field) : "cc"); ▒
> │ if (unlikely(error)) ▒
> │ vmwrite_error(field, value); ▒
> 0.02 │ test %esi,%esi ▒
> │ ↓ js 25 ▒
> 99.98 │ ← retq ▒
> │25: push %rbp ▒
> │ mov $0x440a,%ecx ▒
> │ mov $0x440c,%edx ▒
> │ vmx_complete_interrupts(): ◆
> │ break; ▒
> │ } ▒
> │ } ▒
> │ ▒
> │ static void vmx_complete_interrupts(struct vcpu_vmx *vmx) ▒
> │ { ▒
> │ mov %rsp,%rbp ▒
> │ → callq __vmx_complete_interrupts.part.64 ▒
> │ __vmx_complete_interrupts(&vmx->vcpu, vmx->idt_vectoring_info, ▒
> │ pop %rbp ▒
> │ ← retq ▒
>
>
>
>
> GDB disassable:
> (gdb) disassemble /s vmx_complete_interrupts
> Dump of assembler code for function vmx_complete_interrupts:
> arch/x86/kvm/vmx.c:
> 8917 {
> 0x0000000000007110 <+0>: callq 0x7115 <vmx_complete_interrupts+5>
>
> 8918 __vmx_complete_interrupts(&vmx->vcpu, vmx->idt_vectoring_info,
> 0x0000000000007115 <+5>: mov 0x4590(%rdi),%esi
>
> 8873 vcpu->arch.nmi_injected = false;
> 0x000000000000711b <+11>: movb $0x0,0x32c8(%rdi)
>
> arch/x86/kvm/x86.h:
> 14 vcpu->arch.exception.pending = false;
> 0x0000000000007122 <+18>: movb $0x0,0x1a7c(%rdi)
>
> 15 }
> 16
> 17 static inline void kvm_queue_interrupt(struct kvm_vcpu *vcpu, u8 vector,
> 18 bool soft)
> 19 {
> 20 vcpu->arch.interrupt.pending = true;
> 21 vcpu->arch.interrupt.soft = soft;
> 22 vcpu->arch.interrupt.nr = vector;
> 23 }
> 24
> 25 static inline void kvm_clear_interrupt_queue(struct kvm_vcpu *vcpu)
> 26 {
> 27 vcpu->arch.interrupt.pending = false;
> 0x0000000000007129 <+25>: movb $0x0,0x1a88(%rdi)
>
> arch/x86/kvm/vmx.c:
> 8877 if (!idtv_info_valid)
> 0x0000000000007130 <+32>: test %esi,%esi
> 0x0000000000007132 <+34>: js 0x7135 <vmx_complete_interrupts+37>
>
> 8919 VM_EXIT_INSTRUCTION_LEN,
> 8920 IDT_VECTORING_ERROR_CODE);
> 8921 }
> 0x0000000000007134 <+36>: retq
>
> 8917 {
> 0x0000000000007135 <+37>: push %rbp
> 0x0000000000007136 <+38>: mov $0x440a,%ecx
> 0x000000000000713b <+43>: mov $0x440c,%edx
> ---Type <return> to continue, or q <return> to quit---
> 0x0000000000007140 <+48>: mov %rsp,%rbp
> 0x0000000000007143 <+51>: callq 0x7000 <__vmx_complete_interrupts>
>
> 8919 VM_EXIT_INSTRUCTION_LEN,
> 8920 IDT_VECTORING_ERROR_CODE);
> 8921 }
>
--
Thanks,
Changbin Du
Download attachment "signature.asc" of type "application/pgp-signature" (474 bytes)
Powered by blists - more mailing lists