Re: [PATCH net v4] smsc911x: only update stats when interface is up

[Jakub Kicinski]

On Wed, 29 Mar 2023 21:48:55 +0200 Wolfram Sang wrote:
> > Hm, so you opted to not add the flag in the core?
> > To keep the backport small? I think we should just add it..
> > Clearly multiple drivers would benefit and it's not a huge change.  
> 
> I did it this way for two reasons. First, yes, this is a minimal patch
> for backporting. No dependency on core changes, very easy. Second, this
> is a solution I could develop quickly. I am interested in finding
> another solution, but I guess it needs more time, especially as it
> probably touches 15 drivers. I created an action item for it. I hope
> I'll be able to work on it somewhen. But for now, I just need the SMSC
> bug fixed and need to move to the next issue. If we later have the
> generic solution, converting this driver also won't make a lot of a
> difference.

Okay, core changes aside - does pm_runtime_put() imply an RCU sync?
Otherwise your check in get_stats is racy...