Re: [PATCH] [V9fs-developer] [PATCH] /net/9p/trans_fd.c: fix race-condition by flushing workqueue before the kfree()

[Dominique Martinet <asmadeus@codewreck.org>]

Tomas Bortoli wrote on Fri, Jul 20, 2018:
> The patch adds the flush in p9_mux_poll_stop() as it the function used by
> p9_conn_destroy(), in turn called by p9_fd_close() to stop the async
> polling associated with the data regarding the connection.
> 
> Signed-off-by: Tomas Bortoli <tomasbortoli@gmail.com>
> Reported-by: syzbot+39749ed7d9ef6dfb23f6@syzkaller.appspotmail.com

Looks good to me, I'm taking this patch.

If I had to say something, try to aim for slightly shorter subject lines
if possible :)

> ---
> As shown by Syzbot, it is possible to provoke a race between p9_fd_close() 
> and p9_poll_workfn() that is called to take care of the async read/write work
> to do. To make sure p9_fd_close() frees "trans" when it is not used anymore, 
> it has to explicitly flush p9_poll_work before the kfree().
> 
>  net/9p/trans_fd.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/net/9p/trans_fd.c b/net/9p/trans_fd.c
> index bf459ee0feab..a64b01c56e30 100644
> --- a/net/9p/trans_fd.c
> +++ b/net/9p/trans_fd.c
> @@ -185,6 +185,8 @@ static void p9_mux_poll_stop(struct p9_conn *m)
>  	spin_lock_irqsave(&p9_poll_lock, flags);
>  	list_del_init(&m->poll_pending_link);
>  	spin_unlock_irqrestore(&p9_poll_lock, flags);
> +
> +	flush_work(&p9_poll_work);
>  }
>  
>  /**
-- 
Dominique Martinet