Re: a problem tcp_v4_err()

[Eric Dumazet <eric.dumazet@gmail.com>]

Le vendredi 12 novembre 2010 à 21:29 +0300, Alexey Kuznetsov a écrit :
> Hello!
> 
> On Fri, Nov 12, 2010 at 07:12:58PM +0100, Eric Dumazet wrote:
> > I see socket is locked around line 368,
> > 
> >         bh_lock_sock(sk);
> >         /* If too many ICMPs get dropped on busy
> >          * servers this needs to be solved differently.
> >          */
> >         if (sock_owned_by_user(sk))
> >                 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
> > 
> > 
> > Hmm, maybe some goto is missing ;)
> 
> It is not missing, sock_owned_by_user() is checked later when some operation which
> cannot be done without lock is required. It was done to save error in sk_err_soft even
> when socket is locked.
> 
> This code also _understands_ this: look at
> 
>                 } else if (sock_owned_by_user(sk)) {
>                         /* RTO revert clocked out retransmission,
>                          * but socket is locked. Will defer. */
>                         inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
>                                                   HZ/20, TCP_RTO_MAX);
> 
> but somehow it considers the manipulations with rto/backoff/write_queue as safe.
> Seems, they are not.

Indeed, right you are, I came to same conclusion.

I CC Damian Lukowski in my previous answer (and this one too)

Thanks !