Re: Why won't slapd shutdown (kill -0)?

On Wed, 2004-11-17 at 18:36 -0800, Doug White wrote:
> On Wed, 17 Nov 2004, Sean McNeil wrote:
> 
> > On Wed, 2004-11-17 at 10:28 -0800, Doug White wrote:
> > > On Tue, 16 Nov 2004, Sean McNeil wrote:
> > >
> > > > This has been happening for a long time with current and hasn't been
> > > > resolved.  When I start up slapd, I cannot stop it without kill -9 ing
> > > > it.  It would appear stuck in kse and probably has something to do with
> > > > kill -0:
> > >
> > > Mind expanding on this? The backtrace looks normal for a pthread process.
> > > kill -0 just tests signal delivery; the process is completely unaware that
> > > the probe occured, though.  The process may also be unkillable if its
> > > stuck in some sort of I/O wait.
> > >
> > > Is the server busy when you signal it?
> >
> > Oh, OK.  I didn't look at /usr/local/etc/rc.subr too closely.  I have
> > additional information, though....
> >
> > It appears that all the threads are destroyed yet it is still in the
> > thread processing loop.  The process is no longer active at all.  I just
> > had a similar problem happen with vlc where I closed it yet it is
> > hanging in the same place as slapd with all the threads gone.
> 
> Interesting... what scheduler are you using?

4BSD with PREEMPTION on.  -CURRENT as of yesterday.  This has been an
issue for quite some time now, however.

> >
> > Here is the one from vlc:
> >
> > (gdb) bt full
> > #0  _thr_sched_switch_unlocked (curthread=0x955000) at pthread_md.h:226
> 
> I can't find a reference to this in that file.  Can you run ldd against
> your vlc binary? I('m curious what thread library it thinks its running.

/usr/X11R6/bin/vlc:
        libiconv.so.3 => /usr/local/lib/libiconv.so.3 (0x800955000)
        libintl.so.6 => /usr/local/lib/libintl.so.6 (0x800b48000)
        libz.so.2 => /lib/libz.so.2 (0x800c52000)
        libxvidcore.so.4 => /usr/local/lib/libxvidcore.so.4
(0x800d65000)
        libfaad.so.0 => /usr/local/lib/libfaad.so.0 (0x800f3f000)
        libvorbis.so.3 => /usr/local/lib/libvorbis.so.3 (0x801087000)
        libvorbisenc.so.2 => /usr/local/lib/libvorbisenc.so.2
(0x8011b3000)
        libmp3lame.so.0 => /usr/local/lib/libmp3lame.so.0 (0x801491000)
        libm.so.3 => /lib/libm.so.3 (0x80162c000)
        liba52.so.0 => /usr/local/lib/liba52.so.0 (0x80174e000)
        libtheora.so.0 => /usr/local/lib/libtheora.so.0 (0x80185a000)
        libogg.so.5 => /usr/local/lib/libogg.so.5 (0x801979000)
        libstdc++.so.4 => /usr/lib/libstdc++.so.4 (0x801a7e000)
        libpthread.so.1 => /usr/lib/libpthread.so.1 (0x801c7c000)
        libc.so.6 => /lib/libc.so.6 (0x801da6000)

I just ran it again and had a thread sitting at
_thr_sched_switch_unlocked.  Here is the trace for that thread:

(gdb) bt
#0  _thr_sched_switch_unlocked (curthread=0x1b38c00) at pthread_md.h:226
#1  0x0000000801c8da8f in _nanosleep (time_to_sleep=0x7fffffe8df50,
    time_remaining=0x0)
at /usr/src/lib/libpthread/thread/thr_nanosleep.c:71
#2  0x0000000801c8dbd3 in __nanosleep (time_to_sleep=0x7fffffe8df50,
    time_remaining=0x0)
at /usr/src/lib/libpthread/thread/thr_nanosleep.c:125
#3  0x000000000042b1f9 in msleep (delay=34369437808) at
src/misc/mtime.c:309
#4  0x0000000806a51e02 in OSSThread (p_aout=0x1b38800) at oss.c:624
#5  0x0000000801c87139 in thread_start (curthread=0x800940070,
    start_routine=0x94d068, arg=0x800940070)
    at /usr/src/lib/libpthread/thread/thr_create.c:343
#6  0x0000000801df1ff4 in makectx_wrapper (ucp=0x800940060, func=0x6636,
    args=0x1) at /usr/src/lib/libc/amd64/gen/makecontext.c:100