On Sat, Jun 10, 2017 at 10:33:28AM +0200, Stephan Müller wrote: > > Right. Shouldn't we drop the ctx->inflight completely? > > The code in the current patch set contains: > > when an async operation is queued: > > sock_hold(sk); > ctx->inflight++; > > upon completion of the callback: > > __sock_put(sk); > ctx->inflight--; > > Thus, the socket is grabbed already. Hence, when dropping the inflight code > including the wait queue entirely, I would think we are still save as we hold > the socket. Indeed, we should just kill the wait or perhaps convert it to a WARN_ON. Thanks, -- Email: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
