Red Hat Bugzilla – Bug 1465599
Fix multiple issues with xt action
Last modified: 2017-12-13 12:19:19 EST
Bug for remaining issues from bug bz#1326726. Still not fully fixed. Tested on
+++ This bug was initially created as a clone of Bug #1326726 +++
--- Additional comment from Phil Sutter on 2017-05-18 14:19:34 EDT ---
(In reply to Jaroslav Aster from comment #6)
> Hi Phil,
> unfortunately I see still a lot of fails and segfaults in my test case. Yes,
> there is some progress, but it is not definitely fixed. See
Yes, xt action is still buggy. The segfaults are all related to using any of the NAT targets (SNAT, DNAT, MASQUERADE and REDIRECT). I found the reason why it segfaults (which happens in libipt, BTW) and I have a workaround at hand. Problem is, these targets will still refuse to work then, just the segfault does not happen anymore.
Other non-functional targets are internal ones (ACCEPT, DROP, RETURN). This is also a known issue upstream.
You see, xt action has very limited use even upstream and so I don't see much sense in trying to get all these things functional.
Upstream commit to backport (which prevents the segfault at least):
Author: Phil Sutter <firstname.lastname@example.org>
Date: Tue May 23 15:40:57 2017 +0200
tc: m_xt: Prevent a segfault in libipt
This happens with NAT targets, such as SNAT, DNAT and MASQUERADE. These
are still not usable with this patch, but at least tc doesn't crash
anymore when one tries to use them.
Signed-off-by: Phil Sutter <email@example.com>
I can confirm that there are no segfaults anymore, but there are still a lot of issues to solve.
I'm not sure what we should do with this bug.
(In reply to Jaroslav Aster from comment #3)
> Hi Phil,
> I can confirm that there are no segfaults anymore, but there are still a lot
> of issues to solve.
> I'm not sure what we should do with this bug.
There are conceptual problems with xt action, I don't think it is even possible to address all of them. Assuming that anyone using it should be aware of its limitations, I'd say making sure tc doesn't crash is about the best we can do.
*** Bug 1513973 has been marked as a duplicate of this bug. ***