> > From some light research, I can't find a trace of RCU sync. Pity, I'll > > need to move furhter investigation to later. I'll report back if I have > > something, but it may take a while. Thanks for the help! > > If you don't want to spend too much time you can just call it yourself > right? :) Put a synchronize_rcu() with a comment about concurrent stat > reading after pdata->is_open = false; and that's it? Probably. Although I trust you, I would still add code which I a) don't fully understand myself yet and b) feels somehow wrong because something like synchronize_rcu() should likely not be in drivers but somewhere more generic. Given that the bug in this driver went unnoticed for years (same with the race in other drivers), I do prefer to come back later with the good feeling that I understood the problem and fixed it thoroughly.
Attachment:
signature.asc
Description: PGP signature
