-
-
Notifications
You must be signed in to change notification settings - Fork 342
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
unbound crashing with SIGABRT #38
Comments
svn trunk@5175 is git 2a78803 |
Hi bleve, |
Problem is I only see that on heavily loaded production email servers where I can't experiment. Upgraded to new snapshot from previous one last night and more than half of the servers had unbound crash this morning. |
The debug symbol package should only affect the printout and not the running server. On some package systems this is just a package you can install next to it, with the debug symbols, it would make all the anonymous unbound+xx functions have a name. |
I know - but I can't run crashing version on production. I downgaded already to previous working commit. |
Ok. I downgraded unbound on one server to crashing version and installed debuginfos - we can only hope it hits the issue. |
Thanks! |
Crahser happened this fast first time but no better backtrace, installing abrtd to get proper coredump. |
(gdb) bt full |
Hi bleve, |
I build rpm packages in mock build which is completely clean for all builds. |
And previous commit id works just fine with exactly same build falgs. |
That is excellent. So I would like to disable the optimizations, eg. custom CFLAGS with optimization options, and with --disable-flto passed to configure.... what really? |
You mean that last commit, I mean that one commit is the cause of the problem? Or you mean the older version that you cited above built fine. Yeah I guess the code changed in the meantime... |
I mean the previous non-problematic commit id works. |
Ok, but I still think it could be the optimizer. Could try optimization options disabled. Or could try to 'bisect' to search which commit(s) cause the problem, eg. what caused the problem to appear. |
Used build flags are: CFLAGS='-O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -m64 -mtune=generic -fPIE -pie' |
I cna bisect that but it takes some time, it takes about one hour to crash and there is about five steps to bisect. |
Okay thanks for the build flags. So the --disable-flto flag to configure disables another optimization argument that is added by configure and not in the list put there. Trying that or the bisect could maybe show up something, or show a better stack trace (i.e. |
I removed -O2 and added --disable-flto for bisect builds. d4f697f was good. |
Tested with latest git to make sure I get crash with new build options. (gdb) bt full |
So that |
There is no other app, just unbound caching dns. |
over here we decided to downgrade to 1.8.3 in the meantime |
Hi Antonio, Thanks for reporting. You are downgrading to 1.8.3 for the same reason (that is SIGABRT)? |
git bisect start |
Now testing with commit 0b77c9d reverted to verify bisect. |
yes, for the very same reason on both our 1.9.1 instances |
This issue is not in 1.9.1. You are talking about different issue which has been fixed already. |
you mean this one is a different issue? May 29 11:11:44 ns2-rec kernel: pid 78000 (unbound), uid 59: exited on signal 6 |
That is different because that issue we are talking here is introduced in last two weeks in git version. |
Can you get a backtrace from your crasher? And if it's not same as this one, you could report that as a separate issue - and of course if it's same, please give that information too. |
well, as previously mentioned, we just downgraded to 1.8.3 so currently we're no more observing that issue. |
No crashes since reverting bisected bad commit. |
Thank you very much for the bisect! The commit has to be removed. Issue is that the bug it fixes needs a different fix now... |
Fix with the original fix and a fix for the fix, this makes several error cases also use the correct buffer. In 6067ce6 |
Testing fix, I'll report back in few hours. |
Looks good in production. |
Wow Thanks! That is good to hear, since rc2 already contains it, that should go to the released version soon. I'll close the issue then. |
Jun 06 07:46:51 mx3-fi1 unbound[20686]: *** stack smashing detected ***: /usr/sbin/unbound terminated
Jun 06 07:46:51 mx3-fi1 unbound[20686]: ======= Backtrace: =========
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /lib64/libc.so.6(__fortify_fail+0x37)[0x7fcfe5cf0b67]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /lib64/libc.so.6(+0x117b22)[0x7fcfe5cf0b22]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0x2f351)[0x55cc1f6e1351]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0x36822)[0x55cc1f6e8822]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0x2e703)[0x55cc1f6e0703]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0x2a23e)[0x55cc1f6dc23e]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0x3cdd8)[0x55cc1f6eedd8]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(iter_operate+0x335)[0x55cc1f6efb05]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0x4d782)[0x55cc1f6ff782]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(worker_handle_request+0x1b45)[0x55cc1f6da055]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0xc91db)[0x55cc1f77b1db]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0xc5868)[0x55cc1f777868]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(comm_point_tcp_handle_callback+0xf4)[0x55cc1f777ba4]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /lib64/libevent-2.0.so.5(event_base_loop+0x774)[0x7fcfe6a00a14]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0xc2eac)[0x55cc1f774eac]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0x1ddf1)[0x55cc1f6cfdf1]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0x1995f)[0x55cc1f6cb95f]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /lib64/libc.so.6(__libc_start_main+0xf5)[0x7fcfe5bfb495]
Jun 06 07:46:51 mx3-fi1 unbound[20686]: /usr/sbin/unbound(+0x1a542)[0x55cc1f6cc542]
This is new issue, This didn't happen yet with svn trunk@5175 but happens with git a4f4d7b
The text was updated successfully, but these errors were encountered: