Re: [PATCH] x86/kvm: Fix broken irq restoration in kvm_wait

[Paolo Bonzini <pbonzini@redhat.com>]

On 11/03/21 16:54, Sean Christopherson wrote:
> On Tue, Feb 23, 2021, Wanpeng Li wrote:
>> On Tue, 23 Feb 2021 at 13:25, Wanpeng Li <kernellwp@gmail.com> wrote:
>>>
>>> From: Wanpeng Li <wanpengli@tencent.com>
>>>
>>> After commit 997acaf6b4b59c (lockdep: report broken irq restoration), the guest
>>> splatting below during boot:
>>>
>>>   raw_local_irq_restore() called with IRQs enabled
>>>   WARNING: CPU: 1 PID: 169 at kernel/locking/irqflag-debug.c:10 warn_bogus_irq_restore+0x26/0x30
>>>   Modules linked in: hid_generic usbhid hid
>>>   CPU: 1 PID: 169 Comm: systemd-udevd Not tainted 5.11.0+ #25
>>>   RIP: 0010:warn_bogus_irq_restore+0x26/0x30
>>>   Call Trace:
>>>    kvm_wait+0x76/0x90
>>>    __pv_queued_spin_lock_slowpath+0x285/0x2e0
>>>    do_raw_spin_lock+0xc9/0xd0
>>>    _raw_spin_lock+0x59/0x70
>>>    lockref_get_not_dead+0xf/0x50
>>>    __legitimize_path+0x31/0x60
>>>    legitimize_root+0x37/0x50
>>>    try_to_unlazy_next+0x7f/0x1d0
>>>    lookup_fast+0xb0/0x170
>>>    path_openat+0x165/0x9b0
>>>    do_filp_open+0x99/0x110
>>>    do_sys_openat2+0x1f1/0x2e0
>>>    do_sys_open+0x5c/0x80
>>>    __x64_sys_open+0x21/0x30
>>>    do_syscall_64+0x32/0x50
>>>    entry_SYSCALL_64_after_hwframe+0x44/0xae
>>>
>>> The irqflags handling in kvm_wait() which ends up doing:
>>>
>>>          local_irq_save(flags);
>>>          safe_halt();
>>>          local_irq_restore(flags);
>>>
>>> which triggered a new consistency checking, we generally expect
>>> local_irq_save() and local_irq_restore() to be pared and sanely
>>> nested, and so local_irq_restore() expects to be called with
>>> irqs disabled.
>>>
>>> This patch fixes it by adding a local_irq_disable() after safe_halt()
>>> to avoid this warning.
>>>
>>> Cc: Mark Rutland <mark.rutland@arm.com>
>>> Cc: Thomas Gleixner <tglx@linutronix.de>
>>> Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
>>> ---
>>>   arch/x86/kernel/kvm.c | 4 +++-
>>>   1 file changed, 3 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
>>> index 5e78e01..688c84a 100644
>>> --- a/arch/x86/kernel/kvm.c
>>> +++ b/arch/x86/kernel/kvm.c
>>> @@ -853,8 +853,10 @@ static void kvm_wait(u8 *ptr, u8 val)
>>>           */
>>>          if (arch_irqs_disabled_flags(flags))
>>>                  halt();
>>> -       else
>>> +       else {
>>>                  safe_halt();
>>> +               local_irq_disable();
>>> +       }
>>
>> An alternative fix:
>>
>> diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
>> index 5e78e01..7127aef 100644
>> --- a/arch/x86/kernel/kvm.c
>> +++ b/arch/x86/kernel/kvm.c
>> @@ -836,12 +836,13 @@ static void kvm_kick_cpu(int cpu)
>>
>>   static void kvm_wait(u8 *ptr, u8 val)
>>   {
>> -    unsigned long flags;
>> +    bool disabled = irqs_disabled();
>>
>>       if (in_nmi())
>>           return;
>>
>> -    local_irq_save(flags);
>> +    if (!disabled)
>> +        local_irq_disable();
>>
>>       if (READ_ONCE(*ptr) != val)
>>           goto out;
>> @@ -851,13 +852,14 @@ static void kvm_wait(u8 *ptr, u8 val)
>>        * for irq enabled case to avoid hang when lock info is overwritten
>>        * in irq spinlock slowpath and no spurious interrupt occur to save us.
>>        */
>> -    if (arch_irqs_disabled_flags(flags))
>> +    if (disabled)
>>           halt();
>>       else
>>           safe_halt();
>>
>>   out:
>> -    local_irq_restore(flags);
>> +    if (!disabled)
>> +        local_irq_enable();
>>   }
>>
>>   #ifdef CONFIG_X86_32
> 
> A third option would be to split the paths.  In the end, it's only the ptr/val
> line that's shared.
> 
> ---
>   arch/x86/kernel/kvm.c | 23 ++++++++++-------------
>   1 file changed, 10 insertions(+), 13 deletions(-)
> 
> diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
> index 5e78e01ca3b4..78bb0fae3982 100644
> --- a/arch/x86/kernel/kvm.c
> +++ b/arch/x86/kernel/kvm.c
> @@ -836,28 +836,25 @@ static void kvm_kick_cpu(int cpu)
> 
>   static void kvm_wait(u8 *ptr, u8 val)
>   {
> -	unsigned long flags;
> -
>   	if (in_nmi())
>   		return;
> 
> -	local_irq_save(flags);
> -
> -	if (READ_ONCE(*ptr) != val)
> -		goto out;
> -
>   	/*
>   	 * halt until it's our turn and kicked. Note that we do safe halt
>   	 * for irq enabled case to avoid hang when lock info is overwritten
>   	 * in irq spinlock slowpath and no spurious interrupt occur to save us.
>   	 */
> -	if (arch_irqs_disabled_flags(flags))
> -		halt();
> -	else
> -		safe_halt();
> +	if (irqs_disabled()) {
> +		if (READ_ONCE(*ptr) == val)
> +			halt();
> +	} else {
> +		local_irq_disable();
> 
> -out:
> -	local_irq_restore(flags);
> +		if (READ_ONCE(*ptr) == val)
> +			safe_halt();
> +
> +		local_irq_enable();
> +	}
>   }
> 
>   #ifdef CONFIG_X86_32
> --
> 

I'll send this one tomorrow.

Paolo