mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Paolo Bonzini <pbonzini@redhat.com>
To: Sean Christopherson <sean.j.christopherson@intel.com>
Cc: Vitaly Kuznetsov <vkuznets@redhat.com>,
	Wanpeng Li <wanpengli@tencent.com>,
	Jim Mattson <jmattson@google.com>, Joerg Roedel <joro@8bytes.org>,
	kvm@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] KVM: VMX: Add a trampoline to fix VMREAD error handling
Date: Tue, 31 Mar 2020 17:03:06 +0200	[thread overview]
Message-ID: <140483ed-f1f9-067e-6b60-eccd7115bb0d@redhat.com> (raw)
In-Reply-To: <20200326160712.28803-1-sean.j.christopherson@intel.com>

On 26/03/20 17:07, Sean Christopherson wrote:
> Add a hand coded assembly trampoline to preserve volatile registers
> across vmread_error(), and to handle the calling convention differences
> between 64-bit and 32-bit due to asmlinkage on vmread_error().  Pass
> @field and @fault on the stack when invoking the trampoline to avoid
> clobbering volatile registers in the context of the inline assembly.
> 
> Calling vmread_error() directly from inline assembly is partially broken
> on 64-bit, and completely broken on 32-bit.  On 64-bit, it will clobber
> %rdi and %rsi (used to pass @field and @fault) and any volatile regs
> written by vmread_error().  On 32-bit, asmlinkage means vmread_error()
> expects the parameters to be passed on the stack, not via regs.
> 
> Opportunistically zero out the result in the trampoline to save a few
> bytes of code for every VMREAD.  A happy side effect of the trampoline
> is that the inline code footprint is reduced by three bytes on 64-bit
> due to PUSH/POP being more efficent (in terms of opcode bytes) than MOV.
> 
> Fixes: 6e2020977e3e6 ("KVM: VMX: Add error handling to VMREAD helper")
> Cc: stable@vger.kernel.org
> Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
> ---
> 
> Becuase there just isn't enough custom assembly in VMX :-)
> 
> Simply reverting isn't a great option because we'd lose error reporting
> for VMREAD failure, i.e. it'd return garbage with no other indication that
> something went awry.
> 
> Tested all paths (fail, fault w/o rebooting, fault w/ rebooting) on both
> 64-bit and 32-bit.
> 
>  arch/x86/kvm/vmx/ops.h     | 28 +++++++++++++-----
>  arch/x86/kvm/vmx/vmenter.S | 58 ++++++++++++++++++++++++++++++++++++++
>  2 files changed, 79 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/x86/kvm/vmx/ops.h b/arch/x86/kvm/vmx/ops.h
> index 45eaedee2ac0..09b0937d56b1 100644
> --- a/arch/x86/kvm/vmx/ops.h
> +++ b/arch/x86/kvm/vmx/ops.h
> @@ -12,7 +12,8 @@
>  
>  #define __ex(x) __kvm_handle_fault_on_reboot(x)
>  
> -asmlinkage void vmread_error(unsigned long field, bool fault);
> +__attribute__((regparm(0))) void vmread_error_trampoline(unsigned long field,
> +							 bool fault);
>  void vmwrite_error(unsigned long field, unsigned long value);
>  void vmclear_error(struct vmcs *vmcs, u64 phys_addr);
>  void vmptrld_error(struct vmcs *vmcs, u64 phys_addr);
> @@ -70,15 +71,28 @@ static __always_inline unsigned long __vmcs_readl(unsigned long field)
>  	asm volatile("1: vmread %2, %1\n\t"
>  		     ".byte 0x3e\n\t" /* branch taken hint */
>  		     "ja 3f\n\t"
> -		     "mov %2, %%" _ASM_ARG1 "\n\t"
> -		     "xor %%" _ASM_ARG2 ", %%" _ASM_ARG2 "\n\t"
> -		     "2: call vmread_error\n\t"
> -		     "xor %k1, %k1\n\t"
> +
> +		     /*
> +		      * VMREAD failed.  Push '0' for @fault, push the failing
> +		      * @field, and bounce through the trampoline to preserve
> +		      * volatile registers.
> +		      */
> +		     "push $0\n\t"
> +		     "push %2\n\t"
> +		     "2:call vmread_error_trampoline\n\t"
> +
> +		     /*
> +		      * Unwind the stack.  Note, the trampoline zeros out the
> +		      * memory for @fault so that the result is '0' on error.
> +		      */
> +		     "pop %2\n\t"
> +		     "pop %1\n\t"
>  		     "3:\n\t"
>  
> +		     /* VMREAD faulted.  As above, except push '1' for @fault. */
>  		     ".pushsection .fixup, \"ax\"\n\t"
> -		     "4: mov %2, %%" _ASM_ARG1 "\n\t"
> -		     "mov $1, %%" _ASM_ARG2 "\n\t"
> +		     "4: push $1\n\t"
> +		     "push %2\n\t"
>  		     "jmp 2b\n\t"
>  		     ".popsection\n\t"
>  		     _ASM_EXTABLE(1b, 4b)
> diff --git a/arch/x86/kvm/vmx/vmenter.S b/arch/x86/kvm/vmx/vmenter.S
> index 81ada2ce99e7..861ae40e7144 100644
> --- a/arch/x86/kvm/vmx/vmenter.S
> +++ b/arch/x86/kvm/vmx/vmenter.S
> @@ -234,3 +234,61 @@ SYM_FUNC_START(__vmx_vcpu_run)
>  2:	mov $1, %eax
>  	jmp 1b
>  SYM_FUNC_END(__vmx_vcpu_run)
> +
> +/**
> + * vmread_error_trampoline - Trampoline from inline asm to vmread_error()
> + * @field:	VMCS field encoding that failed
> + * @fault:	%true if the VMREAD faulted, %false if it failed
> +
> + * Save and restore volatile registers across a call to vmread_error().  Note,
> + * all parameters are passed on the stack.
> + */
> +SYM_FUNC_START(vmread_error_trampoline)
> +	push %_ASM_BP
> +	mov  %_ASM_SP, %_ASM_BP
> +
> +	push %_ASM_AX
> +	push %_ASM_CX
> +	push %_ASM_DX
> +#ifdef CONFIG_X86_64
> +	push %rdi
> +	push %rsi
> +	push %r8
> +	push %r9
> +	push %r10
> +	push %r11
> +#endif
> +#ifdef CONFIG_X86_64
> +	/* Load @field and @fault to arg1 and arg2 respectively. */
> +	mov 3*WORD_SIZE(%rbp), %_ASM_ARG2
> +	mov 2*WORD_SIZE(%rbp), %_ASM_ARG1
> +#else
> +	/* Parameters are passed on the stack for 32-bit (see asmlinkage). */
> +	push 3*WORD_SIZE(%ebp)
> +	push 2*WORD_SIZE(%ebp)
> +#endif
> +
> +	call vmread_error
> +
> +#ifndef CONFIG_X86_64
> +	add $8, %esp
> +#endif
> +
> +	/* Zero out @fault, which will be popped into the result register. */
> +	_ASM_MOV $0, 3*WORD_SIZE(%_ASM_BP)
> +
> +#ifdef CONFIG_X86_64
> +	pop %r11
> +	pop %r10
> +	pop %r9
> +	pop %r8
> +	pop %rsi
> +	pop %rdi
> +#endif
> +	pop %_ASM_DX
> +	pop %_ASM_CX
> +	pop %_ASM_AX
> +	pop %_ASM_BP
> +
> +	ret
> +SYM_FUNC_END(vmread_error_trampoline)
> 

Queued, thanks.

Paolo


      reply	other threads:[~2020-03-31 15:03 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-03-26 16:07 Sean Christopherson
2020-03-31 15:03 ` Paolo Bonzini [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=140483ed-f1f9-067e-6b60-eccd7115bb0d@redhat.com \
    --to=pbonzini@redhat.com \
    --cc=jmattson@google.com \
    --cc=joro@8bytes.org \
    --cc=kvm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=sean.j.christopherson@intel.com \
    --cc=vkuznets@redhat.com \
    --cc=wanpengli@tencent.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®