kvm: WARNING in vmx_handle_exit

[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

[Radim Krčmář]

2017-01-02 11:23+0100, Dmitry Vyukov:

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
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 <rkr...@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

[Dmitry Vyukov]

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.

[Dmitry Vyukov]

Was this ever merged? Still happens for me.

[Radim Krčmář]

2017-03-10 19:56+0100, Dmitry Vyukov:

No, it was forgotten ... queued now,

Thanks