On Tue, Jul 10 2018, Vijay Sankar <[email protected]> wrote:
> Quoting Jeremie Courreges-Anglas <[email protected]>:
>
>> On Tue, Jul 10 2018, Ian McWilliam <[email protected]> wrote:
>>> Ouch, that's a new lock order problem. The one I encountered with
>>> 4.8.2 was with the vfs/acl_tdb module.
>>
>> They are not the same thing: you hit a locking problem in samba code,
>> the log from Vijay shows a lock ordering problem in the OpenBSD kernel.

[...]

>>> 2) When samba 4.8.3 starts, the following is logged
>>>
>>> lock order reversal:
>>> 1st 0xffffff07811695c0 vmmaplk (&map->lock) @
>>> /usr/src/sys/uvm/uvm_map.c:4433
>>> 2nd 0xffffff087c37bf80 inode (&ip->i_lock) @
>>> /usr/src/sys/ufs/ufs/ufs_vnops.c:1555
>>> lock order "&ip->i_lock"(rrwlock) -> "&map->lock"(rwlock) first seen at:
>>> #0  witness_checkorder+0x4c0
>>> #1  _rw_enter_read+0x49
>>> #2  uvmfault_lookup+0x8d
>>> #3  uvm_fault+0x72
>>> #4  pageflttrap+0x14c
>>> #5  trap+0x319
>>> #6  alltraps_kern+0x7e
>>> #7  copyout+0x48
>>> #8  ffs_read+0x1f0
>>> #9  VOP_READ+0x49
>>> #10 vn_read+0xca
>>> #11 dofilereadv+0x216
>>> #12 sys_read+0x82
>>> #13 syscall+0x32a
>>> #14 Xsyscall_untramp+0xe4
>>> lock order "&map->lock"(rwlock) -> "&ip->i_lock"(rrwlock) first seen at:
>>> #0  witness_checkorder+0x4c0
>>> #1  _rw_enter+0x68
>>> #2  _rrw_enter+0x3e
>>> #3  VOP_LOCK+0x3d
>>> #4  vn_lock+0x34
>>> #5  uvn_io+0x1b8
>>> #6  uvm_pager_put+0x109
>>> #7  uvn_flush+0x424
>>> #8  uvm_map_clean+0x3e7
>>> #9  syscall+0x32a
>>> #10 Xsyscall_untramp+0xe4
>>>
>>> Is this to be expected or is there a problem here?
>>>
>>> If there are any other tests that will be helpful please let me know.
>>>
>>> Thanks very much,
>>>
>>> Vijay
>>>
>>>
>>> Vijay Sankar, M.Eng., P.Eng.
>>> ForeTell Technologies Limited
>>> [email protected]
>>>
>>
>>
>> --
>> jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF  DDCC 0DFA 74AE 1524 E7EE
>
> I will do the tests again on a set of clean machines with
> a newer  -current. I am also wondering if testing with bsd.sp instead of
> bsd.mp  is a valid approach.

I'm not sure that using bsd.sp would avoid those log entries, but that's
not the point: those witness(4) traces are here to help kernel
developers track down lock ordering issues.  Please try a fresh -current
and if that message appears again please report it on bugs@.

Those witness(4) traces may point out a problem that could affect samba,
but they don't *look* samba-specific to me.  There's a good chance that
samba will keep working correctly, whether those warnings appear or not.

btw,

[...]

> So I installed samba 4.8.3 on a different box (desktop computer with
> gnome) thinking that the problem was due to my test setup.
> Unfortunately starting samba with the default smb.conf (not a domain
> member) resulted in the following:
>
> error: [drm:pid48841:intel_pipe_update_start] *ERROR* Potential atomic
> update failure on pipe A
> lock order reversal:
>   1st 0xffffffff81df30b0 &sched_lock (&sched_lock) @
> /usr/src/sys/kern/kern_synch.c:444
>   2nd 0xffff8000000ff9f0 &dev_priv->uncore.lock
> (&dev_priv->uncore.lock) @
> /usr/src/sys/dev/pci/drm/i915/intel_uncore.c:811
> lock order "&dev_priv->uncore.lock"(mutex) ->  "&sched_lock"(sched_lock)
> first seen at:
> #0  witness_checkorder+0x4c0
> #1  ___mp_lock+0x70
> #2  schedclock+0x30
> #3  hardclock+0xe3
> #4  lapic_clockintr+0x3d
> #5  Xresume_lapic_ltimer+0x22
> #6  x86_bus_space_mem_read_4+0x14
> #7  gen6_read32+0x184
> #8  drm_update_vblank_count+0x65
> #9  drm_handle_vblank+0xee
> #10 ironlake_irq_handler+0x4e4
> #11 intr_handler+0x74
> #12 Xintr_ioapic_edge18_untramp+0x161
> #13 ___mp_lock+0xca
> #14 solock+0x50
> #15 sosend+0x117
> #16 sendit+0x3fb
> #17 sys_sendmsg+0x15a
> #18 syscall+0x32a
> lock order "&sched_lock"(sched_lock) ->  "&dev_priv->uncore.lock"(mutex)
> first seen at:
> #0  witness_checkorder+0x4c0
> #1  _mtx_enter+0x31
> #2  gen6_read32+0x8f
> #3  gen6_ring_get_seqno+0x3a
> #4  __i915_wait_request+0x232
> #5  i915_gem_object_wait_rendering__nonblocking+0x1d6
> #6  i915_gem_set_domain_ioctl+0xdb
> #7  drm_do_ioctl+0x221
> #8  drmioctl+0xf9
> #9  VOP_IOCTL+0x5a
> #10 vn_ioctl+0x6b
> #11 sys_ioctl+0x477
> #12 syscall+0x32a
> #13 Xsyscall_untramp+0xe4
> lock order reversal:
>   1st 0xffffffff81df30b0 &sched_lock (&sched_lock) @
> /usr/src/sys/kern/kern_synch.c:444
>   2nd 0xffff800000106270 &dev_priv->irq_lock (&dev_priv->irq_lock) @
> /usr/src/sys/dev/pci/drm/i915/intel_ringbuffer.c:1787
> lock order "&dev_priv->irq_lock"(mutex) -> "&sched_lock"(sched_lock)
> first seen at:
> #0  witness_checkorder+0x4c0
> #1  ___mp_lock+0x70
> #2  wakeup_n+0x39
> #3  task_add+0x93
> #4  gen6_rps_boost+0x129
> #5  __i915_wait_request+0x155
> #6  i915_gem_object_wait_rendering__nonblocking+0x1d6
> #7  i915_gem_set_domain_ioctl+0xdb
> #8  drm_do_ioctl+0x221
> #9  drmioctl+0xf9
> #10 VOP_IOCTL+0x5a
> #11 vn_ioctl+0x6b
> #12 sys_ioctl+0x477
> #13 syscall+0x32a
> #14 Xsyscall_untramp+0xe4
> lock order "&sched_lock"(sched_lock) -> "&dev_priv->irq_lock"(mutex)
> first seen at:
> #0  witness_checkorder+0x4c0
> #1  _mtx_enter+0x31
> #2  gen6_ring_put_irq+0x36
> #3  __i915_wait_request+0x367
> #4  i915_gem_object_wait_rendering__nonblocking+0x1d6
> #5  i915_gem_set_domain_ioctl+0xdb
> #6  drm_do_ioctl+0x221
> #7  drmioctl+0xf9
> #8  VOP_IOCTL+0x5a
> #9  vn_ioctl+0x6b
> #10 sys_ioctl+0x477
> #11 syscall+0x32a
> #12 Xsyscall_untramp+0xe4

Those two traces don't look samba-related at all, and should be
reported. ;)

> lock order reversal:
>   1st 0xffffff07059c4a00 vmmaplk (&map->lock) @
> /usr/src/sys/uvm/uvm_map.c:4433
>   2nd 0xffffff070d217a30 inode (&ip->i_lock) @
> /usr/src/sys/ufs/ufs/ufs_vnops.c:1555
> lock order "&ip->i_lock"(rrwlock) -> "&map->lock"(rwlock) first seen at:
> #0  witness_checkorder+0x4c0
> #1  _rw_enter+0x68
> #2  vm_map_lock_ln+0xbc
> #3  uvm_map+0x1a1
> #4  km_alloc+0x16a
> #5  pool_multi_alloc_ni+0xbb
> #6  pool_p_alloc+0x56
> #7  pool_do_get+0xe4
> #8  pool_get+0xaf
> #9  ufsdirhash_build+0x31e
> #10 ufs_lookup+0x19d
> #11 VOP_LOOKUP+0x4f
> #12 vfs_lookup+0x27e
> #13 namei+0x226
> #14 start_init+0xb2
> lock order "&map->lock"(rwlock) -> "&ip->i_lock"(rrwlock) first seen at:
> #0  witness_checkorder+0x4c0
> #1  _rw_enter+0x68
> #2  _rrw_enter+0x3e
> #3  VOP_LOCK+0x3d
> #4  vn_lock+0x34
> #5  uvn_io+0x1b8
> #6  uvm_pager_put+0x109
> #7  uvn_flush+0x424
> #8  uvm_map_clean+0x3e7
> #9  syscall+0x32a
> #10 Xsyscall_untramp+0xe4
>
> Did a pkg_add -r samba-4.8.2 and ldb-1.3.3; restarted the desktop and
> verified there are no lock reversal messages

I would of course prefer that you keep on testing with the latest 4.8.3
samba package. ;)

-- 
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF  DDCC 0DFA 74AE 1524 E7EE

Reply via email to