Re: [syzbot] [netfs?] INFO: task hung in netfs_unbuffered_write_iter

[Oleg Nesterov <oleg@redhat.com>]

On 03/25, K Prateek Nayak wrote:
>
> I chased this down to p9_client_rpc() net/9p/client.c specifically:
>
>         err = c->trans_mod->request(c, req);
>         if (err < 0) {
>                 /* write won't happen */
>                 p9_req_put(c, req);
>                 if (err != -ERESTARTSYS && err != -EFAULT)
>                         c->status = Disconnected;
>                 goto recalc_sigpending;
>         }
>
> c->trans_mod->request() calls p9_fd_request() in net/9p/trans_fd.c
> which basically does a p9_fd_poll().

Again, I know nothing about 9p... but if p9_fd_request() returns
an err < 0, then it comes from p9_conn->err and p9_fd_request()
does nothing else.

> Previously, the above would fail with err as -EIO which would
> cause the client to "Disconnect" and the retry logic would make
> progress. Now however, the err returned is -ERESTARTSYS

OK... So p9_conn->err = -ERESTARTSYS was set by p9_conn_cancel()
called by p9_write_work() because pipe_write() returns ERESTARTSYS?

But I don't understand -EIO with the reverted commit aaec5a95d59615

Oleg.