Date: Thu, 16 Sep 2010 12:45:48 -0700 From: Matthew Fleming <mdf356@gmail.com> To: Matthew Jacob <mj@feral.com> Cc: freebsd-hackers@freebsd.org Subject: Re: race conditions for destroying and opening a dev Message-ID: <AANLkTimiLCDtHKa22Yjhx2vEJ5Va4pPHOTvfADQo9%2Bw4@mail.gmail.com> In-Reply-To: <4C92694D.1070705@feral.com> References: <4C92694D.1070705@feral.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Sep 16, 2010 at 12:00 PM, Matthew Jacob <mj@feral.com> wrote:
>
> Has anyone seen this scenario before? I am seeing it in RELENG_7, but the
> code in question exists through to head.
>
> Thread 1:
>
> (kgdb) where
> #0 =A0sched_switch (td=3D0xffffff003a04ea80, newtd=3D0xffffff00210b4000,
> flags=3DVariable "flags" is not available.
> ) at ../../../kern/sched_ule.c:1944
> #1 =A00xffffffff803b6091 in mi_switch (flags=3D1, newtd=3D0x0) at
> ../../../kern/kern_synch.c:450
> #2 =A00xffffffff80402399 in sleepq_switch (wchan=3D0xffffff8413b50b60) at
> ../../../kern/subr_sleepqueue.c:497
> #3 =A00xffffffff80402e8c in sleepq_timedwait (wchan=3D0xffffff8413b50b60)=
at
> ../../../kern/subr_sleepqueue.c:615
> #4 =A00xffffffff803b682d in _sleep (ident=3D0xffffff8413b50b60,
> lock=3D0xffffffff80b0ee00, priority=3D76, wmesg=3D0xffffffff806583bb "dev=
drn",
> timo=3D100) at ../../../kern/kern_synch.c:228
> #5 =A00xffffffff8037640c in destroy_devl (dev=3D0xffffff003aaf0000) at
> ../../../kern/kern_conf.c:874
> #6 =A00xffffffff80376759 in destroy_dev (dev=3D0xffffff003aaf0000) at
> ../../../kern/kern_conf.c:916
> #7 =A00xffffffff8034c939 in g_dev_orphan (cp=3D0xffffff003a544800) at
> ../../../geom/geom_dev.c:438
> #8 =A00xffffffff803506a0 in g_run_events () at ../../../geom/geom_event.c=
:164
> #9 =A00xffffffff80351f1c in g_event_procbody () at
> ../../../geom/geom_kern.c:141
> #10 0xffffffff8038a73a in fork_exit (callout=3D0xffffffff80351eb0
> <g_event_procbody at ../../../geom/geom_kern.c:132>, arg=3D0x0,
> frame=3D0xffffff8413b50c80) at ../../../kern/kern_fork.c:829
> #11 0xffffffff805a747e in fork_trampoline () at
> ../../../amd64/amd64/exception.S:564
> #12 0x0000000000000000 in ?? ()
>
> This thread is waiting on the threadcount to go away- i.e., the last clos=
e
> of the device to occur ("da16" in this case).
>
> Thread 2:
>
> (kgdb) where
> #0 =A0sched_switch (td=3D0xffffff009bb4ca80, newtd=3D0xffffff003af43380,
> flags=3DVariable "flags" is not available.
> ) at ../../../kern/sched_ule.c:1944
> #1 =A00xffffffff803b6091 in mi_switch (flags=3D1, newtd=3D0x0) at
> ../../../kern/kern_synch.c:450
> #2 =A00xffffffff80402399 in sleepq_switch (wchan=3D0xffffffff80b0e040) at
> ../../../kern/subr_sleepqueue.c:497
> #3 =A00xffffffff80402f84 in sleepq_wait (wchan=3D0xffffffff80b0e040) at
> ../../../kern/subr_sleepqueue.c:580
> #4 =A00xffffffff803b5385 in _sx_xlock_hard (sx=3D0xffffffff80b0e040,
> tid=3D18446742976810240640, opts=3DVariable "opts" is not available.
> ) at ../../../kern/kern_sx.c:562
> #5 =A00xffffffff803b5731 in _sx_xlock (sx=3D0xffffffff80b0e040, opts=3D0,
> file=3D0xffffffff80652d27 "../../../geom/geom_dev.c", line=3D196) at sx.h=
:154
> #6 =A00xffffffff8034d1bc in g_dev_open (dev=3D0xffffff003aaf0000, flags=
=3D1,
> fmt=3DVariable "fmt" is not available.
> ) at ../../../geom/geom_dev.c:196
> #7 =A00xffffffff80333741 in devfs_open (ap=3D0xffffff841dea88b0) at
> ../../../fs/devfs/devfs_vnops.c:902
> #8 =A00xffffffff80601daf in VOP_OPEN_APV (vop=3D0xffffffff8089fb80,
> a=3D0xffffff841dea88b0) at vnode_if.c:371
> #9 =A00xffffffff80467246 in vn_open_cred (ndp=3D0xffffff841dea8a00,
> flagp=3D0xffffff841dea894c, cmode=3DVariable "cmode" is not available.
> ) at vnode_if.h:199
> #10 0xffffffff80463770 in kern_open (td=3D0xffffff009bb4ca80, path=3D0x51=
14a0
> <Address 0x5114a0 out of bounds>, pathseg=3DVariable "pathseg" is not
> available.
> ) at ../../../kern/vfs_syscalls.c:1054
> #11 0xffffffff805c599e in syscall (frame=3D0xffffff841dea8c80) at
> ../../../amd64/amd64/trap.c:911
> #12 0xffffffff805a723b in Xfast_syscall () at
> ../../../amd64/amd64/exception.S:349
> #13 0x00000008009a219c in ?? ()
>
> This thread was opening the device, bumped the refcount, but then wedged =
on
> the geom topology lock .....
>
> the refcount field is protected under devmtx....
>
> Anyone seen this?
>
> I'm half inclined to either add in CDP_SCHED_DTR when one calls destroy_d=
ev,
> or make dev_refthread look at CDP_ACTIVE, leaning more toward the latter.
>
> Any thoughts on this?
We had a similar bug at Isilon, but in our case it was in
cam/scsi/scsi_pass.c::passcleanup() calling destroy_dev(). We
switched it to destroy_dev_sched() to fix the si_threadcount deadlock.
Cheers,
matthew
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?AANLkTimiLCDtHKa22Yjhx2vEJ5Va4pPHOTvfADQo9%2Bw4>