[Samba] Problem with heaps of sleeping smb processes due to panic action

[Dieter Modig]

Thanks for replying! 

----- Ursprungligt meddelande -----

> Från: "Andrew Bartlett" <abartlet at samba.org>
> Till: "Dieter Modig" <dieter.m at inputinterior.se>
> Kopia: samba at lists.samba.org, metze at samba.org
> Skickat: lördag, 18 aug 2012 8:38:49
> Ämne: Re: [Samba] Problem with heaps of sleeping smb processes due to
> panic action

> On Tue, 2012-08-14 at 15:21 +0200, Dieter Modig wrote:
> > Hi!
> >
> > We're running Samba4 (Version 4.0.0beta4) and are experiencing
> > problems with smb_panic actions which result in loads of sleep
> > processes in the end taking down the entire machine. This problem
> > did not exist in prior builds for us (upgraded from alpha17 to
> > beta3 and then beta4). Are there any specific log extracts I can
> > supply that can help someone pinpoint the problem? log.smbd seems
> > to indicate the following:
> >
> > [2012/07/06 13:52:36.425367, 0]
> > ../source3/lib/util.c:974(log_stack_trace)
> > BACKTRACE: 27 stack frames:
> > #0 /usr/local/samba/lib/libsmbconf.so.0(log_stack_trace+0x1f)
> > [0x7f37c35011df]
> > #1 /usr/local/samba/lib/libsmbconf.so.0(smb_panic_s3+0x6d)
> > [0x7f37c350105d]
> > #2 /usr/local/samba/lib/libsamba-util.so.0(smb_panic+0x28)
> > [0x7f37c5323aee]
> > #3 /usr/local/samba/lib/private/libsmbd_base.so(+0x129b89)
> > [0x7f37c4ac6b89]
> > #4 /usr/local/samba/lib/private/libsmbd_base.so(+0x129ea9)
> > [0x7f37c4ac6ea9]
> > #5 /usr/local/samba/lib/private/libsmbd_base.so(+0x12c769)
> > [0x7f37c4ac9769]
> > #6 /usr/local/samba/lib/private/libsmbd_base.so(+0x130045)
> > [0x7f37c4acd045]
> > #7
> > /usr/local/samba/lib/private/libsmbd_base.so(create_file_default+0x2f8)
> > [0x7f37c4acdb7f]
> > #8 /usr/local/samba/lib/private/libsmbd_base.so(+0x23f813)
> > [0x7f37c4bdc813]
> > #9
> > /usr/local/samba/lib/private/libsmbd_base.so(smb_vfs_call_create_file+0xcb)
> > [0x7f37c4ad8fc8]
> > #10 /usr/local/samba/lib/private/libsmbd_base.so(+0x1771b2)
> > [0x7f37c4b141b2]
> > #11
> > /usr/local/samba/lib/private/libsmbd_base.so(smbd_smb2_request_process_create+0x7ac)
> > [0x7f37c4b1224c]
> > #12
> > /usr/local/samba/lib/private/libsmbd_base.so(smbd_smb2_request_dispatch+0x6fe)
> > [0x7f37c4b0a42d]
> > #13 /usr/local/samba/lib/private/libsmbd_base.so(+0x1781ce)
> > [0x7f37c4b151ce]
> > #14
> > /usr/local/samba/lib/private/libtevent.so.0(tevent_common_loop_immediate+0x1f9)
> > [0x7f37c376a090]
> > #15 /usr/local/samba/lib/libsmbconf.so.0(run_events_poll+0x57)
> > [0x7f37c351d23f]
> > #16 /usr/local/samba/lib/libsmbconf.so.0(+0x44ac2) [0x7f37c351dac2]
> > #17
> > /usr/local/samba/lib/private/libtevent.so.0(_tevent_loop_once+0xe8)
> > [0x7f37c376918f]
> > #18
> > /usr/local/samba/lib/private/libsmbd_base.so(smbd_process+0x10ed)
> > [0x7f37c4af4569]
> > #19 /usr/local/samba/sbin/smbd() [0x409c48]
> > #20 /usr/local/samba/lib/libsmbconf.so.0(run_events_poll+0x71a)
> > [0x7f37c351d902]
> > #21 /usr/local/samba/lib/libsmbconf.so.0(+0x44ba2) [0x7f37c351dba2]
> > #22
> > /usr/local/samba/lib/private/libtevent.so.0(_tevent_loop_once+0xe8)
> > [0x7f37c376918f]
> > #23 /usr/local/samba/sbin/smbd() [0x40a838]
> > #24 /usr/local/samba/sbin/smbd(main+0x14b9) [0x40be42]
> > #25 /lib/libc.so.6(__libc_start_main+0xfd) [0x7f37c1e02c8d]
> > #26 /usr/local/samba/sbin/smbd() [0x405969]
> > [2012/07/06 13:52:36.430994, 0]
> > ../source3/lib/util.c:875(smb_panic_s3)
> >
> > but I'm honestly not good enough to dissect the problem.

> In developer mode, the default panic action is to run 'sleep' so you
> can
> attach with a debugger (see testparm -v output).

> To instead produce a stack trace that we can use, get 'gdb_backtrace'
> from selftest/gdb_backtrace and set in your smb.conf:

> panic action = /path/to/gdb_backtrace %d

Done! Where does this backtrace end up once it's been created? 

> I've CC'ed metze, one of the key developers working on the smbd file
> server to see if he wants to persue this with you further on beta4.

> Otherwise, please update to current master as this is a fast-moving
> area
> that may have already been fixed.
I've upgraded to beta6 (which at least last week was the latest version) and the problem still persists. 

Regards, 
/Dieter