Re: [PATCH] tracefs: avoid setting i_mode to a temp value

[Sishuai Gong]

> On Aug 17, 2023, at 8:00 PM, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
> 
> On Thu, 17 Aug 2023 19:47:34 -0400
> Sishuai Gong <sishuai.system@xxxxxxxxx> wrote:
> 
>>> Can you produce this race?  
>> This data race was detected when I was testing the kernel (e.g., fuzzing)
>> but I did not make the attempt to reproduce it.
> 
> Now, I'm curious to what exactly is this fixing? The intermediate value is
> the S_IALLUGO bits cleared. Doesn't that mean that nothing has permission?
> 
> It's not a big deal if that's the case, as it just means things are locked
> down a bit more than normal.
You are right. Even if the intermediate value is read, it is unlikely to cause anything
serious. The reader I observed is acl_permission_check(), which will not be affected
by the race.

> 
> My question is, do we really care, and why should we?
This shouldn’t be a serious problem. Maybe we could consider this patch as an
annotation to the race.

> 
> -- Steve
>