internal backtrace support - libunwind or backtrace_symbols?

Michael Tokarev mjt at
Mon May 30 07:49:53 UTC 2022

30.05.2022 01:48, Andrew Bartlett wrote:
> Michael,
> You mentioned that you didn't get a backtrace until buidling debian
> packages with libunwind.  Can you double-check?  The reason I ask is

Well. I only wrote that I'm getting backtraces after enabling libunwind,
I didn't say I weren't getting them before enabling libunwind.

I just checked - and you're right, the backtraces are there without libunwind
too, here's an example (just sending SIGSEGV to smbd):

[2022/05/30 10:38:06.000715,  0] ../../lib/util/fault.c:173(smb_panic_log)
   INTERNAL ERROR: Signal 11: Segmentation fault in pid 340388 (4.16.1-Debian)
[2022/05/30 10:38:06.000754,  0] ../../lib/util/fault.c:177(smb_panic_log)
   If you are running a recent Samba version, and if you think this problem is not yet fixed in the latest versions, please consider reporting this 
bug, see
[2022/05/30 10:38:06.000787,  0] ../../lib/util/fault.c:182(smb_panic_log)
[2022/05/30 10:38:06.000817,  0] ../../lib/util/fault.c:183(smb_panic_log)
   PANIC (pid 340388): Signal 11: Segmentation fault in 4.16.1-Debian
[2022/05/30 10:38:06.007755,  0] ../../lib/util/fault.c:287(log_stack_trace)
   BACKTRACE: 21 stack frames:
    #0 /usr/lib/x86_64-linux-gnu/ [0x7fddf44640f0]
    #1 /usr/lib/x86_64-linux-gnu/ [0x7fddf4464339]
    #2 /usr/lib/x86_64-linux-gnu/ [0x7fddf44643d1]
    #3 /lib/x86_64-linux-gnu/ [0x7fddf4292140]
    #4 /lib/x86_64-linux-gnu/ [0x7fddf41810d6]
    #5 /usr/lib/x86_64-linux-gnu/ [0x7fddf4264e07]
    #6 /usr/lib/x86_64-linux-gnu/ [0x7fddf4263087]
    #7 /usr/lib/x86_64-linux-gnu/ [0x7fddf425de24]
    #8 /usr/lib/x86_64-linux-gnu/ [0x7fddf425e10b]
    #9 /usr/lib/x86_64-linux-gnu/ [0x7fddf4263027]
    #10 /usr/lib/x86_64-linux-gnu/samba/ [0x7fddf4746518]
    #11 /usr/sbin/smbd(+0xa161) [0x55be897c6161]
    #12 /usr/lib/x86_64-linux-gnu/ [0x7fddf425e9e1]
    #13 /usr/lib/x86_64-linux-gnu/ [0x7fddf4264f47]
    #14 /usr/lib/x86_64-linux-gnu/ [0x7fddf4263087]
    #15 /usr/lib/x86_64-linux-gnu/ [0x7fddf425de24]
    #16 /usr/lib/x86_64-linux-gnu/ [0x7fddf425e10b]
    #17 /usr/lib/x86_64-linux-gnu/ [0x7fddf4263027]
    #18 /usr/sbin/smbd(main+0x1c22) [0x55be897c39f2]
    #19 /lib/x86_64-linux-gnu/ [0x7fddf40a9d0a]
    #20 /usr/sbin/smbd(_start+0x2a) [0x55be897c3baa]
[2022/05/30 10:38:06.008228,  0] ../../source3/lib/dumpcore.c:315(dump_core)
   dumping core in /var/log/samba/cores/smbd

> that when I disabled libunwind with this MR, I get the
> backtrace_symbols one, and I need to know if this isn't working in the
> 'real world'.

It looks like it's the same here.

What happened here is that after issues which Dirk faced in another tread,
and your reply there, I *thought* backtraces does not work at all without
libunwind, but I didn't verify this, having no time to spare. I just enabled
libunwind for debian build and verified the backtrace is there *after* doing
so, - in a hope to become closer to understanding the problem which Dirk is
facing (which is quite bad experience).

Now I looked at the traces he's getting, and it looks like the prob there is
not the lack of a backtrace but a corrupt stack which can't be backtraced.

I'll disable libunwind on debian samba build now once things are clear.
And this mess is what I'm getting when trying to be just a bit helpful
but in a rush.  Please excuse me for the noise.  Unfortunately I still
don't know what to do with Dirk's case.


More information about the samba-technical mailing list