Re: [PATCH v6.6 RESEND 2/2] x86/irq: Plug vector setup race

[Jinjie Ruan]

On 2025/8/22 16:57, Greg KH wrote:
> On Fri, Aug 22, 2025 at 03:38:25AM +0000, Jinjie Ruan wrote:
>> From: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
>>
>> commit ce0b5eedcb753697d43f61dd2e27d68eb5d3150f upstream.
>>

[...]

>>  
>>  /*
>> @@ -273,7 +308,9 @@ DEFINE_IDTENTRY_IRQ(common_interrupt)
>>  	/* entry code tells RCU that we're not quiescent.  Check it. */
>>  	RCU_LOCKDEP_WARN(!rcu_is_watching(), "IRQ failed to wake up RCU");
>>  
>> -	call_irq_handler(vector, regs);
>> +	if (unlikely(!call_irq_handler(vector, regs)))
>> +		apic_eoi();
>> +
> 
> This chunk does not look correct.  The original commit did not have
> this, so why add it here?  Where did it come from?
> 
> The original patch said:
> 	-       if (unlikely(call_irq_handler(vector, regs)))
> 	+       if (unlikely(!call_irq_handler(vector, regs)))
> 
> And was not an if statement.
> 
> So did you forget to backport something else here?  Why is this not
> identical to what the original was?

The if statement is introduced in commit 1b03d82ba15e ("x86/irq: Install
posted MSI notification handler") which is a patch in patch set
https://lore.kernel.org/all/20240423174114.526704-1-jacob.jun.pan@xxxxxxxxxxxxxxx/,
but it seems to be a performance optimization patch set, and this patch
includes additional modifications. The context conflict is merely a
simple refactoring, but the cost of the entire round of this patch set
is too high.

> 
> thanks,
> 
> greg k-h