Hardy: [NETFILTER]: {ip,ip6,nfnetlink}_queue: fix SKB_LINEAR_ASSERT when mangling packet data

Package set to linux.

Hi yanaventer,

Looks like this patch has made it's way into the Intrepid kernel. I'll open a Hardy nomination for them to consider backporting. Thanks.

ogasawara@yoji:~/ubuntu-intrepid$ git log -p e2b58a67b91dec07dfb40ca2056c64011ce8489d
commit e2b58a67b91dec07dfb40ca2056c64011ce8489d
Author: Patrick McHardy <email address hidden>
Date: Tue Feb 19 17:17:52 2008 -0800

    [NETFILTER]: {ip,ip6,nfnetlink}_queue: fix SKB_LINEAR_ASSERT when mangling packet data

http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-hardy.git;a=commit;h=19cf29258654c2d55422540993482a8ca88bbfce

SRU Justification

Impact: Fix kernel assert crash

Patch Description: Use skb_copy_expand instead of pskb_expand_head.

Patch: http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-hardy.git;a=commit;h=19cf29258654c2d55422540993482a8ca88bbfce

Test Case: See bug description

Accepted into -proposed, please test and give feedback here. Please see https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

If we're going to apply this fix, do we not also need to commit http://lkml.org/lkml/2008/5/13/392 as well?

Marking this as verification-failed, unless it can be shown this fix doesn't introduce the regression discussed in the previously mentioned http://lkml.org/lkml/2008/5/13/392 , which states:

  The patch below asks for allocation of a copy with enough space for
  mangled packet and the same amount of headroom as old sk_buff. While
  looking at how the regression appeared (e2b58a67), I noticed the same
  pattern in ipq_mangle_ipv6() and ipq_mangle_ipv4(). The patch corrects
  those locations too.

git commit e2b58a67 is the fix referred to in comment 2 by Leann Ogasawara.

@Steve - http://lkml.org/lkml/2008/5/13/392 indicates that the above commit works fine with 2.6.24. Arnaud said he had trouble with 2.6.25.

On Fri, Oct 03, 2008 at 03:21:25AM -0000, Tim Gardner wrote:
> @Steve - http://lkml.org/lkml/2008/5/13/392 indicates that the above
> commit works fine with 2.6.24. Arnaud said he had trouble with 2.6.25.

@Tim - while that's true (he also claims that his mangling
works fine with 2.6.22), note that the original patch reported in
http://lkml.org/lkml/2008/3/21/449 for this bug was from the -stable
2.6.24.4 thread starting at http://lkml.org/lkml/2008/3/21/447; it's not
exactly clear *which* 2.6.24 kernel Arnaud tried his mangling against
when he went back to look to see where the regression was introduced.

What I'd really like is a testcase that triggers the original bug, and
then we could work from there, but I expect we're not going to get that.

--
Steve Beattie
<email address hidden>
http://NxNW.org/~steve/

linux 2.6.24-21 copied to hardy-updates.