MPTSAS problems in 2.6.26-rc2-mm1

I have 2.6.26-rc2-mm1 with Rik's splitvm patches. While booting the kernel, I ran into the following. I searched my email quickly to see that no one else has reported this problem. Lockdep seems to be complaining about the same lock being held at the same location (the message is a bit confusing to me).

=============================================
[ INFO: possible recursive locking detected ] 2.6.26-rc2-mm1 #2
---------------------------------------------
insmod/1072 is trying to acquire lock:
(&cls->mutex){--..}, at: [<ffffffff803eb2cc>] device_add+0x46e/0x5d7

but task is already holding lock:
(&cls->mutex){--..}, at: [<ffffffff803eb2cc>] device_add+0x46e/0x5d7

other info that might help us debug this:
3 locks held by insmod/1072:
#0: (&ioc->sas_discovery_mutex){--..}, at: [<ffffffffa003e7e9>]

mptsas_probe+0x3a1/0x442 [mptsas]
 #1:  (&shost->scan_mutex){--..}, at: [<ffffffff80466cbd>]
scsi_scan_target+0x71/0xb0
 #2:  (&cls->mutex){--..}, at: [<ffffffff803eb2cc>] device_add+0x46e/0x5d7

stack backtrace:

Pid: 1072, comm: insmod Not tainted 2.6.26-rc2-mm1 #2

Call Trace:
 [<ffffffff80254545>] __lock_acquire+0x911/0xc4c
 [<ffffffff80254835>] ? __lock_acquire+0xc01/0xc4c
 [<ffffffff803eb2cc>] ? device_add+0x46e/0x5d7
 [<ffffffff8025490e>] lock_acquire+0x8e/0xb2
 [<ffffffff803eb2cc>] ? device_add+0x46e/0x5d7
 [<ffffffff8059e067>] mutex_lock_nested+0xf2/0x27f
 [<ffffffff803eb2cc>] ? device_add+0x46e/0x5d7
 [<ffffffff8059f80b>] ? _spin_unlock+0x26/0x2a
 [<ffffffff803eb2cc>] device_add+0x46e/0x5d7
 [<ffffffff803eb44e>] device_register+0x19/0x1d
 [<ffffffff803eb531>] device_create+0xdf/0x110
 [<ffffffff80253794>] ? trace_hardirqs_on_caller+0xf9/0x124
 [<ffffffff802537cc>] ? trace_hardirqs_on+0xd/0xf
 [<ffffffff8059df73>] ? mutex_unlock+0x9/0xb
 [<ffffffff803eed13>] ? kobj_map+0x119/0x12e
 [<ffffffff802a4488>] ? exact_lock+0x0/0x14
 [<ffffffff802a4364>] ? exact_match+0x0/0x9
 [<ffffffff80488bb4>] sg_add+0x262/0x394
 [<ffffffff8059f80b>] ? _spin_unlock+0x26/0x2a
 [<ffffffff803eb31b>] device_add+0x4bd/0x5d7
 [<ffffffff80467d9b>] scsi_sysfs_add_sdev+0xde/0x200
 [<ffffffff80465feb>] scsi_probe_and_add_lun+0x847/0x977
 [<ffffffff8046639d>] __scsi_scan_target+0xbf/0x522
 [<ffffffff80253794>] ? trace_hardirqs_on_caller+0xf9/0x124
 [<ffffffff80250b9d>] ? debug_mutex_free_waiter+0x46/0x4a
 [<ffffffff8059e1e5>] ? mutex_lock_nested+0x270/0x27f
 [<ffffffff80466cbd>] ? scsi_scan_target+0x71/0xb0
 [<ffffffff80466ce8>] scsi_scan_target+0x9c/0xb0
 [<ffffffffa001ad60>] :scsi_transport_sas:sas_rphy_add+0x14c/0x15b
 [<ffffffffa003cc17>] :mptsas:mptsas_probe_one_phy+0x364/0x3a9
 [<ffffffffa003d651>] :mptsas:mptsas_probe_hba_phys+0x5bb/0x614
 [<ffffffffa0028aeb>] ? :mptbase:mpt_timer_expired+0x0/0x2e
 [<ffffffffa0028aeb>] ? :mptbase:mpt_timer_expired+0x0/0x2e
 [<ffffffffa0028aeb>] ? :mptbase:mpt_timer_expired+0x0/0x2e
 [<ffffffffa003e7f1>] :mptsas:mptsas_probe+0x3a9/0x442
 [<ffffffff8039126a>] pci_device_probe+0xbd/0x116
 [<ffffffff803ed1f0>] driver_probe_device+0xdb/0x162
 [<ffffffff803ed2c6>] __driver_attach+0x4f/0x79
 [<ffffffff803ed277>] ? __driver_attach+0x0/0x79
 [<ffffffff803ec832>] bus_for_each_dev+0x4d/0x7e
 [<ffffffff803ed01c>] driver_attach+0x1c/0x1e
 [<ffffffff803eccc0>] bus_add_driver+0xba/0x212
 [<ffffffff803ed4a6>] driver_register+0x93/0x10a
 [<ffffffff80391526>] __pci_register_driver+0x73/0xad
 [<ffffffffa00450e2>] :mptsas:mptsas_init+0xe2/0xfa
 [<ffffffff8025c25e>] sys_init_module+0x9e/0x1b2
 [<ffffffff8020b20b>] system_call_after_swapgs+0x7b/0x80

Quoted 4 lines · Balbir Singh

I have 2.6.26-rc2-mm1 with Rik's splitvm patches. While booting the kernel, I ran into the following. I searched my email quickly to see that no one else has reported this problem. Lockdep seems to be complaining about the same lock being held at the same location (the message is a bit confusing to me).

looks like device_add() recursing - most likely with a different device though.

Quoted 7 lines · Peter Zijlstra

Quoted 4 lines · Balbir Singh

I have 2.6.26-rc2-mm1 with Rik's splitvm patches. While booting the kernel, I ran into the following. I searched my email quickly to see that no one else has reported this problem. Lockdep seems to be complaining about the same lock being held at the same location (the message is a bit confusing to me).

looks like device_add() recursing - most likely with a different device though.

This is another instance of a problem being caused by semaphore to mutex conversion in struct class;

There's another thread on this here:

http://marc.info/?t=121074904600001