Re: [PATCH] KVM: VMX: downgrade warning on unexpected exit code

[Dmitry Vyukov <dvyukov@google.com>]

On Fri, Jan 13, 2017 at 6:59 PM, Radim Krčmář <rkrcmar@redhat.com> wrote:
> 2017-01-02 11:23+0100, Dmitry Vyukov:
>> Hello,
>>
>> I've got the following warning while running syzkaller fuzzer:
>>
>> WARNING: CPU: 2 PID: 13257 at arch/x86/kvm/vmx.c:8633
>> vmx_handle_exit+0x262b/0x38b0 arch/x86/kvm/vmx.c:8633
>> vmx: unexpected exit reason 0xb
>> CPU: 2 PID: 13257 Comm: syz-executor7 Not tainted 4.10.0-rc1+ #118
>> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
>> Call Trace:
>>  __dump_stack lib/dump_stack.c:15 [inline]
>>  dump_stack+0x292/0x3a2 lib/dump_stack.c:51
>>  panic+0x1cb/0x3a9 kernel/panic.c:179
>>  __warn+0x1c4/0x1e0 kernel/panic.c:539
>>  warn_slowpath_fmt+0xc5/0x100 kernel/panic.c:562
>>  vmx_handle_exit+0x262b/0x38b0 arch/x86/kvm/vmx.c:8633
>>  vcpu_enter_guest arch/x86/kvm/x86.c:6884 [inline]
>>  vcpu_run arch/x86/kvm/x86.c:6943 [inline]
>>  kvm_arch_vcpu_ioctl_run+0xf3d/0x45f0 arch/x86/kvm/x86.c:7101
>>  kvm_vcpu_ioctl+0x673/0x1120 arch/x86/kvm/../../../virt/kvm/kvm_main.c:2569
>>  vfs_ioctl fs/ioctl.c:43 [inline]
>>  do_vfs_ioctl+0x1bf/0x1780 fs/ioctl.c:683
>>  SYSC_ioctl fs/ioctl.c:698 [inline]
>>  SyS_ioctl+0x8f/0xc0 fs/ioctl.c:689
>>  entry_SYSCALL_64_fastpath+0x1f/0xc2
>> RIP: 0033:0x4438a9
>> RSP: 002b:00007f935aa2fb58 EFLAGS: 00000286 ORIG_RAX: 0000000000000010
>> RAX: ffffffffffffffda RBX: 0000000000000015 RCX: 00000000004438a9
>> RDX: 0000000000000000 RSI: 000000000000ae80 RDI: 0000000000000015
>> RBP: 00000000006ddb30 R08: 0000000000000000 R09: 0000000000000000
>> R10: 0000000000000000 R11: 0000000000000286 R12: 0000000000700000
>> R13: 0000000000000006 R14: 0000000000000015 R15: 0000000020014000
>> Dumping ftrace buffer:
>>    (ftrace buffer empty)
>> Kernel Offset: disabled
>>
>> Exit reason 0xb is GETSEC instruction. This does not look harmful as
>> it is handled as #UD. But I think we should print a single line
>> message regarding non-emulated instruction as in other cases, just to
>> not scare cloud admins and to make syzkaller ignore it.
>>
>> FTR, a raw reproducer is here:
>> https://gist.githubusercontent.com/dvyukov/c762f6ea04ebbba49cdee0a6caca31b4/raw/f21deb04cdc70ae74100c12447d71bb0cd2025c7/gistfile1.txt
>
> I can't reproduce -- maybe a nested bug, which will take a while to
> figure out.  Still, host dump at that point is useless, so the change
> makes sense.
>
> A guest dump would be useful (e.g. I think that GETSEC should not exit

Thanks!

I had to disable dump_vmcs locally because it constantly produces tons
of output even when no bugs happen.
If you really-really need it, I can enable dump_vmcs and re-reproduce the bug.


> if guest CR4.SMX is disabled), but just giving the error is about as
> good and we want something short and rate-limited if the message can be
> trigerred by a guest in production ...
> ---8<---
> We never needed the call trace and we better rate-limit if it can be
> triggered by a guest.
>
> Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
> ---
>  arch/x86/kvm/vmx.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index a236decb81e4..7cd606daa01e 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -8595,7 +8595,8 @@ static int vmx_handle_exit(struct kvm_vcpu *vcpu)
>             && kvm_vmx_exit_handlers[exit_reason])
>                 return kvm_vmx_exit_handlers[exit_reason](vcpu);
>         else {
> -               WARN_ONCE(1, "vmx: unexpected exit reason 0x%x\n", exit_reason);
> +               vcpu_unimpl(vcpu, "vmx: unexpected exit reason 0x%x\n",
> +                               exit_reason);
>                 kvm_queue_exception(vcpu, UD_VECTOR);
>                 return 1;
>         }
> --
> 2.11.0
>