[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1423709570.4847.21.camel@edumazet-glaptop2.roam.corp.google.com>
Date: Wed, 11 Feb 2015 18:52:50 -0800
From: Eric Dumazet <eric.dumazet@...il.com>
To: David L Stevens <david.stevens@...cle.com>
Cc: David Miller <davem@...emloft.net>, netdev@...r.kernel.org
Subject: Re: [PATCH net-next] sunvnet: don't change gso data on clones
On Wed, 2015-02-11 at 12:35 -0500, David L Stevens wrote:
> I don't think I understand your concern. This problem can result in a
> panic using sunvnet because the sunvnet driver is changing the original
> skb, which is always, or at least almost always, a clone. TCP uses gso_segs
> to track packet counts, so changing it in the driver can result in bad math--
> TCP assumes its copy of the clone's data shouldn't change (of course).
>
> A driver that doesn't change the segmentation or original data doesn't
> need to care whether it's a clone or not-- it'll free it and drop a
> reference. Since sunvnet is changing the gso_size and gso_segs, it needs
> to unclone first.
Well, we had a very hard to find bug in TCP stack, I want to make sure
we fixed all relevant points.
commit c52e2421f7368fd36cbe330d2cf41b10452e39a9
Author: Eric Dumazet <edumazet@...gle.com>
Date: Tue Oct 15 11:54:30 2013 -0700
tcp: must unclone packets before mangling them
TCP stack should make sure it owns skbs before mangling them.
We had various crashes using bnx2x, and it turned out gso_size
was cleared right before bnx2x driver was populating TC descriptor
of the _previous_ packet send. TCP stack can sometime retransmit
packets that are still in Qdisc.
Of course we could make bnx2x driver more robust (using
ACCESS_ONCE(shinfo->gso_size) for example), but the bug is TCP stack.
We have identified two points where skb_unclone() was needed.
This patch adds a WARN_ON_ONCE() to warn us if we missed another
fix of this kind.
Kudos to Neal for finding the root cause of this bug. Its visible
using small MSS.
Signed-off-by: Eric Dumazet <edumazet@...gle.com>
Signed-off-by: Neal Cardwell <ncardwell@...gle.com>
Cc: Yuchung Cheng <ycheng@...gle.com>
Signed-off-by: David S. Miller <davem@...emloft.net>
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists