Re: kvm: WARNING In kvm_apic_accept_events

[Paolo Bonzini <pbonzini@redhat.com>]

On 23/01/2017 17:03, David Hildenbrand wrote:
> Am 22.01.2017 um 09:52 schrieb Dmitry Vyukov:
>> On Sat, Nov 12, 2016 at 2:56 AM, Dmitry Vyukov <dvyukov@google.com> wrote:
>>> Hello,
>>>
>>> The following program triggers WARNING in kvm_apic_accept_events:
>>> https://gist.githubusercontent.com/dvyukov/95b845a2e637485568ea1ef181a72370/raw/d90717dd67128b21715c5e794568a1600f613d97/gistfile1.txt
>>>
>>> On commit 015ed9433be2b476ec7e2e6a9a411a56e3b5b035 (Nov 11).
>>>
>>> ------------[ cut here ]------------
>>> WARNING: CPU: 1 PID: 23523 at arch/x86/kvm/lapic.c:2330
>>> kvm_apic_accept_events+0x3f6/0x460
>>> Modules linked in:[   55.632648] CPU: 1 PID: 2812 Comm: a.out Not
>>> tainted 4.9.0-rc4+ #41
>>> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
>>>  ffff88006be77978 ffffffff81c2f79b 0000000000000000 0000000000000000
>>>  ffffffff8322d6a0 ffffffff81127426 ffff88006be779c0 ffffffff8123ffef
>>>  ffffffff811acc46 ffffffff0000091a ffffffff8322d6a0 000000000000091a
>>> Call Trace:
>>>  [<     inline     >] __dump_stack lib/dump_stack.c:15
>>>  [<ffffffff81c2f79b>] dump_stack+0xb3/0x118 lib/dump_stack.c:51
>>>  [<ffffffff8123ffef>] __warn+0x19f/0x1e0 kernel/panic.c:550
>>>  [<ffffffff812401fc>] warn_slowpath_null+0x2c/0x40 kernel/panic.c:585
>>>  [<ffffffff81127426>] kvm_apic_accept_events+0x3f6/0x460
>>> arch/x86/kvm/lapic.c:2330
>>>  [<     inline     >] vcpu_block arch/x86/kvm/x86.c:6793
>>>  [<     inline     >] vcpu_run arch/x86/kvm/x86.c:6828
>>>  [<ffffffff810b8f20>] kvm_arch_vcpu_ioctl_run+0x1580/0x5a90
>>> arch/x86/kvm/x86.c:6984
>>>  [<ffffffff8106091e>] kvm_vcpu_ioctl+0x61e/0xdd0
>>> arch/x86/kvm/../../../virt/kvm/kvm_main.c:2557
>>>  [<     inline     >] vfs_ioctl fs/ioctl.c:43
>>>  [<ffffffff816b16fc>] do_vfs_ioctl+0x18c/0x1040 fs/ioctl.c:679
>>>  [<     inline     >] SYSC_ioctl fs/ioctl.c:694
>>>  [<ffffffff816b263f>] SyS_ioctl+0x8f/0xc0 fs/ioctl.c:685
>>>  [<ffffffff831ebd01>] entry_SYSCALL_64_fastpath+0x1f/0xc2
>>> arch/x86/entry/entry_64.S:209
>>> ---[ end trace b2415ddd1b26dce6 ]---
>>
>>
>> ping
>>
> 
> Our VCPU is in SMM mode but is switched into KVM_MP_STATE_INIT_RECEIVED.
> 
> The relevant warning is (arch/x86/kvm/lapic.c:kvm_apic_accept_events())
> 
> if (is_smm(vcpu)) {
>         WARN_ON_ONCE(vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED);
>         ....
> 
> Something like that might help (untested):
> 
> 
> From: David Hildenbrand <david@redhat.com>
> Date: Mon, 23 Jan 2017 16:42:23 +0100
> Subject: [PATCH] KVM: x86: fix illegal MP_STATE when in/entering SMM
> 
> If we already entered/are about to enter SMM, don't allow
> switching to INIT/SIPI_RECEIVED, otherwise the next call to
> kvm_apic_accept_events() will report a warning.
> 
> Fixes: cd7764fe9f73 ("KVM: x86: latch INITs while in system management mode")
> Reported-by: Dmitry Vyukov <dvyukov@google.com>
> Signed-off-by: David Hildenbrand <david@redhat.com>
> ---
>  arch/x86/kvm/x86.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 57d8a85..06ffc91 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -7247,6 +7247,12 @@ int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
>  	    mp_state->mp_state != KVM_MP_STATE_RUNNABLE)
>  		return -EINVAL;
>  
> +	/* INITs are latched while in SMM */
> +	if ((is_smm(vcpu) || vcpu->arch.smi_pending) &&
> +	    (mp_state->mp_state == KVM_MP_STATE_SIPI_RECEIVED ||
> +	     mp_state->mp_state == KVM_MP_STATE_INIT_RECEIVED))
> +		return -EINVAL;
> +
>  	if (mp_state->mp_state == KVM_MP_STATE_SIPI_RECEIVED) {
>  		vcpu->arch.mp_state = KVM_MP_STATE_INIT_RECEIVED;
>  		set_bit(KVM_APIC_SIPI, &vcpu->arch.apic->pending_events);
> 

Looks good, thanks for analyzing it.  I'll try making a minimal testcase...

Paolo