public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Willy Tarreau <w@1wt.eu>
To: linux-kernel@vger.kernel.org, stable@vger.kernel.org
Cc: Eric Dumazet <edumazet@google.com>,
	Vlad Yasevich <vyasevic@redhat.com>,
	"David S. Miller" <davem@davemloft.net>, Willy Tarreau <w@1wt.eu>
Subject: [ 26/48] net: Correctly set segment mac_len in skb_segment().
Date: Sun, 16 Nov 2014 22:53:54 +0100	[thread overview]
Message-ID: <20141116215329.732657503@1wt.eu> (raw)
In-Reply-To: <28c765bc23bd4bae1611534e510f49f8@local>

2.6.32-longterm review patch.  If anyone has any objections, please let me know.

------------------

From: Vlad Yasevich <vyasevic@redhat.com>

[ Upstream commit fcdfe3a7fa4cb74391d42b6a26dc07c20dab1d82 ]

When performing segmentation, the mac_len value is copied right
out of the original skb.  However, this value is not always set correctly
(like when the packet is VLAN-tagged) and we'll end up copying a bad
value.

One way to demonstrate this is to configure a VM which tags
packets internally and turn off VLAN acceleration on the forwarding
bridge port.  The packets show up corrupt like this:
16:18:24.985548 52:54:00:ab:be:25 > 52:54:00:26:ce:a3, ethertype 802.1Q
(0x8100), length 1518: vlan 100, p 0, ethertype 0x05e0,
        0x0000:  8cdb 1c7c 8cdb 0064 4006 b59d 0a00 6402 ...|...d@.....d.
        0x0010:  0a00 6401 9e0d b441 0a5e 64ec 0330 14fa ..d....A.^d..0..
        0x0020:  29e3 01c9 f871 0000 0101 080a 000a e833)....q.........3
        0x0030:  000f 8c75 6e65 7470 6572 6600 6e65 7470 ...unetperf.netp
        0x0040:  6572 6600 6e65 7470 6572 6600 6e65 7470 erf.netperf.netp
        0x0050:  6572 6600 6e65 7470 6572 6600 6e65 7470 erf.netperf.netp
        0x0060:  6572 6600 6e65 7470 6572 6600 6e65 7470 erf.netperf.netp
        ...

This also leads to awful throughput as GSO packets are dropped and
cause retransmissions.

The solution is to set the mac_len using the values already available
in then new skb.  We've already adjusted all of the header offset, so we
might as well correctly figure out the mac_len using skb_reset_mac_len().
After this change, packets are segmented correctly and performance
is restored.

CC: Eric Dumazet <edumazet@google.com>
Signed-off-by: Vlad Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
[wt: open-code skb_mac_len() as 2.6.32 doesn't have it]
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
 net/core/skbuff.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 72ff527..b6707b8 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -2573,7 +2573,6 @@ struct sk_buff *skb_segment(struct sk_buff *skb, int features)
 		tail = nskb;
 
 		__copy_skb_header(nskb, skb);
-		nskb->mac_len = skb->mac_len;
 
 		/* nskb and skb might have different headroom */
 		if (nskb->ip_summed == CHECKSUM_PARTIAL)
@@ -2583,6 +2582,7 @@ struct sk_buff *skb_segment(struct sk_buff *skb, int features)
 		skb_set_network_header(nskb, skb->mac_len);
 		nskb->transport_header = (nskb->network_header +
 					  skb_network_header_len(skb));
+		nskb->mac_len = nskb->network_header - nskb->mac_header;
 		skb_copy_from_linear_data(skb, nskb->data, doffset);
 
 		if (fskb != skb_shinfo(skb)->frag_list)
-- 
1.7.12.2.21.g234cd45.dirty




  parent reply	other threads:[~2014-11-16 22:01 UTC|newest]

Thread overview: 51+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <28c765bc23bd4bae1611534e510f49f8@local>
2014-11-16 21:53 ` [ 00/48] 2.6.32.64-longterm review Willy Tarreau
2014-11-16 21:53 ` [ 01/48] x86_32, entry: Do syscall exit work on badsys Willy Tarreau
2014-11-16 21:53 ` [ 02/48] x86_32, entry: Store badsys error code in %eax Willy Tarreau
2014-11-16 21:53 ` [ 03/48] x86_32, entry: Clean up sysenter_badsys declaration Willy Tarreau
2014-11-16 21:53 ` [ 04/48] MIPS: Cleanup flags in syscall flags handlers Willy Tarreau
2014-11-16 21:53 ` [ 05/48] MIPS: asm: thread_info: Add _TIF_SECCOMP flag Willy Tarreau
2014-11-16 21:53 ` [ 06/48] fix autofs/afs/etc. magic mountpoint breakage Willy Tarreau
2014-11-16 21:53 ` [ 07/48] ALSA: control: Make sure that id->index does not Willy Tarreau
2014-11-16 21:53 ` [ 08/48] ALSA: control: Handle numid overflow Willy Tarreau
2014-11-16 21:53 ` [ 09/48] sctp: Fix sk_ack_backlog wrap-around problem Willy Tarreau
2014-11-16 21:53 ` [ 10/48] mm: try_to_unmap_cluster() should lock_page() before Willy Tarreau
2014-11-16 21:53 ` [ 11/48] filter: prevent nla extensions to peek beyond the end Willy Tarreau
2014-11-16 21:53 ` [ 12/48] ALSA: control: Protect user controls against Willy Tarreau
2014-11-16 21:53 ` [ 13/48] ptrace,x86: force IRET path after a ptrace_stop() Willy Tarreau
2014-11-16 21:53 ` [ 14/48] sym53c8xx_2: Set DID_REQUEUE return code when aborting Willy Tarreau
2014-11-16 21:53 ` [ 15/48] tcp: fix tcp_match_skb_to_sack() for unaligned SACK at Willy Tarreau
2014-11-16 21:53 ` [ 16/48] igmp: fix the problem when mc leave group Willy Tarreau
2014-11-16 21:53 ` [ 17/48] appletalk: Fix socket referencing in skb Willy Tarreau
2014-11-16 21:53 ` [ 18/48] net: sctp: fix information leaks in ulpevent layer Willy Tarreau
2014-11-16 21:53 ` [ 19/48] sunvnet: clean up objects created in vnet_new() on Willy Tarreau
2014-11-16 21:53 ` [ 20/48] ipv4: fix buffer overflow in ip_options_compile() Willy Tarreau
2014-11-16 21:53 ` [ 21/48] net: sctp: inherit auth_capable on INIT collisions Willy Tarreau
2014-11-16 21:53 ` [ 22/48] net: sendmsg: fix NULL pointer dereference Willy Tarreau
2014-12-01 11:45   ` Luis Henriques
2014-12-01 12:30     ` Willy Tarreau
2014-11-16 21:53 ` [ 23/48] tcp: Fix integer-overflows in TCP veno Willy Tarreau
2014-11-16 21:53 ` [ 24/48] tcp: Fix integer-overflow in TCP vegas Willy Tarreau
2014-11-16 21:53 ` [ 25/48] macvlan: Initialize vlan_features to turn on offload Willy Tarreau
2014-11-16 21:53 ` Willy Tarreau [this message]
2014-11-16 21:53 ` [ 27/48] iovec: make sure the caller actually wants anything in Willy Tarreau
2014-11-16 21:53 ` [ 28/48] sctp: fix possible seqlock seadlock in Willy Tarreau
2014-11-16 21:53 ` [ 29/48] Revert "nfsd: correctly handle return value from Willy Tarreau
2014-11-16 21:53 ` [ 30/48] dm crypt: fix access beyond the end of allocated space Willy Tarreau
2014-11-16 21:53 ` [ 31/48] gianfar: disable vlan tag insertion by default Willy Tarreau
2014-11-16 21:54 ` [ 32/48] USB: kobil_sct: fix non-atomic allocation in write Willy Tarreau
2014-11-16 21:54 ` [ 33/48] fix misuses of f_count() in ppp and netlink Willy Tarreau
2014-11-16 21:54 ` [ 34/48] net: sctp: fix skb_over_panic when receiving malformed Willy Tarreau
2014-11-16 21:54 ` [ 35/48] tty: Fix high cpu load if tty is unreleaseable Willy Tarreau
2014-11-16 21:54 ` [ 36/48] netfilter: nf_log: account for size of NLMSG_DONE Willy Tarreau
2014-11-16 21:54 ` [ 37/48] netfilter: nfnetlink_log: fix maximum packet length Willy Tarreau
2014-11-16 21:54 ` [ 38/48] ring-buffer: Always reset iterator to reader page Willy Tarreau
2014-11-16 21:54 ` [ 39/48] md/raid6: avoid data corruption during recovery of Willy Tarreau
2014-11-16 21:54 ` [ 40/48] net: pppoe: use correct channel MTU when using Willy Tarreau
2014-11-16 21:54 ` [ 41/48] ARM: 7668/1: fix memset-related crashes caused by Willy Tarreau
2014-11-16 21:54 ` [ 42/48] ARM: 7670/1: fix the memset fix Willy Tarreau
2014-11-16 21:54 ` [ 43/48] lib/lzo: Update LZO compression to current upstream Willy Tarreau
2014-11-16 21:54 ` [ 44/48] Documentation: lzo: document part of the encoding Willy Tarreau
2014-11-16 21:54 ` [ 45/48] lzo: check for length overrun in variable length Willy Tarreau
2014-11-16 21:54 ` [ 46/48] USB: add new zte 3g-dongles pid to option.c Willy Tarreau
2014-11-16 21:54 ` [ 47/48] futex: Unlock hb->lock in futex_wait_requeue_pi() Willy Tarreau
2014-11-16 21:54 ` [ 48/48] isofs: Fix unbounded recursion when processing Willy Tarreau

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20141116215329.732657503@1wt.eu \
    --to=w@1wt.eu \
    --cc=davem@davemloft.net \
    --cc=edumazet@google.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=stable@vger.kernel.org \
    --cc=vyasevic@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox