]> git.ipfire.org Git - thirdparty/kernel/stable.git/commit
tcp: Do not apply TSO segment limit to non-TSO packets
authorHerbert Xu <herbert@gondor.apana.org.au>
Wed, 31 Dec 2014 13:39:23 +0000 (00:39 +1100)
committerJiri Slaby <jslaby@suse.cz>
Mon, 26 Jan 2015 13:39:17 +0000 (14:39 +0100)
commitdc5e8861c1ab671c1fd3908d63a03fe5f2f4fb11
tree2919c3a4894dfb1a2407a7b8e9c743eb1f4e9c20
parent00c47292b1cdf7e721dccc5659879732eca6451a
tcp: Do not apply TSO segment limit to non-TSO packets

[ Upstream commit 843925f33fcc293d80acf2c5c8a78adf3344d49b ]

Thomas Jarosch reported IPsec TCP stalls when a PMTU event occurs.

In fact the problem was completely unrelated to IPsec.  The bug is
also reproducible if you just disable TSO/GSO.

The problem is that when the MSS goes down, existing queued packet
on the TX queue that have not been transmitted yet all look like
TSO packets and get treated as such.

This then triggers a bug where tcp_mss_split_point tells us to
generate a zero-sized packet on the TX queue.  Once that happens
we're screwed because the zero-sized packet can never be removed
by ACKs.

Fixes: 1485348d242 ("tcp: Apply device TSO segment limit earlier")
Reported-by: Thomas Jarosch <thomas.jarosch@intra2net.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Cheers,
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
net/ipv4/tcp_output.c