Skip to contentSkip to conversation

Kernel lock bug detected (kseriod)

PATCH In review Linux Portal 4 participants 7 messages Jun 2, 2006 – Jun 3, 2006 Input drivers

lore
Share & tools
lore
#0 lore raw

Yes, it was observed on 2.6.17-rc5-mm2. Of specific stuff I have synaptics driver compiled in (together with psmouse - bug was observed on a laptop - during the boot sequence!). If you need more info about the machine or its configuration, feel free to ask. For the time being I'm sending just what the kernel lock validator left in my kernel log. And keep up the good work, the lock validator is definitely some fine piece of art!

Synaptics Touchpad, model: 1, fw: 5.9, id: 0x1b6eb1, caps: 0xa84793/0x100000
serio: Synaptics pass-through port at isa0060/serio4/input0
input: SynPS/2 Synaptics TouchPad as /class/input/input2
====================================
[ BUG: possible deadlock detected! ]
------------------------------------
kseriod/133 is trying to acquire lock:
(&ps2dev->cmd_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10

but task is already holding lock:
(&ps2dev->cmd_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10

which could potentially lead to deadlocks!

other info that might help us debug this:
4 locks held by kseriod/133:

#0:  (serio_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
#1:  (&serio->drv_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
#2:  (psmouse_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
#3:  (&ps2dev->cmd_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
stack backtrace:
<78105572> show_trace+0x12/0x20  <78105599> dump_stack+0x19/0x20
<7813920e> __lockdep_acquire+0x54e/0xe00  <78139f2a> lockdep_acquire+0x7a/0xa0
<7846b3c9> __mutex_lock_slowpath+0x49/0x160  <7846b4e8> mutex_lock+0x8/0x10
<7834497b> ps2_command+0x3b/0x3c0  <7834ad22> psmouse_sliced_command+0x22/0x70
<7834f471> synaptics_pt_write+0x21/0x50  <78344736> ps2_sendbyte+0x46/0x120
<78344a29> ps2_command+0xe9/0x3c0  <7834ae8d> psmouse_probe+0x1d/0xa0
<7834c537> psmouse_connect+0x137/0x200  <78341649>
serio_connect_driver+0x29/0x50
<783419b6> serio_driver_probe+0x16/0x20  <782a8fb4>
driver_probe_device+0x44/0xd0
<782a9048> __device_attach+0x8/0x10  <782a8563> bus_for_each_drv+0x63/0x90
<782a90a6> device_attach+0x56/0x60  <782a868e> bus_attach_device+0x1e/0x40
<782a7763> device_add+0x113/0x180  <7834299d> serio_thread+0x1cd/0x2bb
<781322c6> kthread+0xc6/0xca  <78101005> kernel_thread_helper+0x5/0xb
Signature · 1 line
http://linux.inet.hr/
#1 lore raw
Quoted 31 lines · Linux Portal

Yes, it was observed on 2.6.17-rc5-mm2. Of specific stuff I have synaptics driver compiled in (together with psmouse - bug was observed on a laptop - during the boot sequence!). If you need more info about the machine or its configuration, feel free to ask. For the time being I'm sending just what the kernel lock validator left in my kernel log. And keep up the good work, the lock validator is definitely some fine piece of art!

Synaptics Touchpad, model: 1, fw: 5.9, id: 0x1b6eb1, caps: 0xa84793/0x100000
serio: Synaptics pass-through port at isa0060/serio4/input0
input: SynPS/2 Synaptics TouchPad as /class/input/input2
====================================
[ BUG: possible deadlock detected! ]
------------------------------------
kseriod/133 is trying to acquire lock:
(&ps2dev->cmd_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10

but task is already holding lock:
(&ps2dev->cmd_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10

which could potentially lead to deadlocks!

other info that might help us debug this:
4 locks held by kseriod/133:

 #0:  (serio_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
 #1:  (&serio->drv_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
 #2:  (psmouse_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
 #3:  (&ps2dev->cmd_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
Quoted 16 lines
stack backtrace:
 <78105572> show_trace+0x12/0x20  <78105599> dump_stack+0x19/0x20
 <7813920e> __lockdep_acquire+0x54e/0xe00  <78139f2a> lockdep_acquire+0x7a/0xa0
 <7846b3c9> __mutex_lock_slowpath+0x49/0x160  <7846b4e8> mutex_lock+0x8/0x10
 <7834497b> ps2_command+0x3b/0x3c0  <7834ad22> psmouse_sliced_command+0x22/0x70
 <7834f471> synaptics_pt_write+0x21/0x50  <78344736> ps2_sendbyte+0x46/0x120
 <78344a29> ps2_command+0xe9/0x3c0  <7834ae8d> psmouse_probe+0x1d/0xa0
 <7834c537> psmouse_connect+0x137/0x200  <78341649>
serio_connect_driver+0x29/0x50
 <783419b6> serio_driver_probe+0x16/0x20  <782a8fb4>
driver_probe_device+0x44/0xd0
 <782a9048> __device_attach+0x8/0x10  <782a8563> bus_for_each_drv+0x63/0x90
 <782a90a6> device_attach+0x56/0x60  <782a868e> bus_attach_device+0x1e/0x40
 <782a7763> device_add+0x113/0x180  <7834299d> serio_thread+0x1cd/0x2bb
 <781322c6> kthread+0xc6/0xca  <78101005> kernel_thread_helper+0x5/0xb

Thanks.

So we're taking ps2->cmd_mutex and then we're recurring back into ps2_command() and then taking ps2->serio->cmd_mutex.

I suspect that's all correct/natural/expected and needs another make-lockdep-shut-up patch.

#2 lore raw
Quoted 49 lines · Andrew Morton
Quoted 31 lines · Linux Portal

Yes, it was observed on 2.6.17-rc5-mm2. Of specific stuff I have synaptics driver compiled in (together with psmouse - bug was observed on a laptop - during the boot sequence!). If you need more info about the machine or its configuration, feel free to ask. For the time being I'm sending just what the kernel lock validator left in my kernel log. And keep up the good work, the lock validator is definitely some fine piece of art!

Synaptics Touchpad, model: 1, fw: 5.9, id: 0x1b6eb1, caps: 0xa84793/0x100000
serio: Synaptics pass-through port at isa0060/serio4/input0
input: SynPS/2 Synaptics TouchPad as /class/input/input2
====================================
[ BUG: possible deadlock detected! ]
------------------------------------
kseriod/133 is trying to acquire lock:
(&ps2dev->cmd_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10

but task is already holding lock:
(&ps2dev->cmd_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10

which could potentially lead to deadlocks!

other info that might help us debug this:
4 locks held by kseriod/133:

 #0:  (serio_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
 #1:  (&serio->drv_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
 #2:  (psmouse_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
 #3:  (&ps2dev->cmd_mutex){--..}, at: [<7846b4e8>] mutex_lock+0x8/0x10
Quoted 16 lines
stack backtrace:
 <78105572> show_trace+0x12/0x20  <78105599> dump_stack+0x19/0x20
 <7813920e> __lockdep_acquire+0x54e/0xe00  <78139f2a> lockdep_acquire+0x7a/0xa0
 <7846b3c9> __mutex_lock_slowpath+0x49/0x160  <7846b4e8> mutex_lock+0x8/0x10
 <7834497b> ps2_command+0x3b/0x3c0  <7834ad22> psmouse_sliced_command+0x22/0x70
 <7834f471> synaptics_pt_write+0x21/0x50  <78344736> ps2_sendbyte+0x46/0x120
 <78344a29> ps2_command+0xe9/0x3c0  <7834ae8d> psmouse_probe+0x1d/0xa0
 <7834c537> psmouse_connect+0x137/0x200  <78341649>
serio_connect_driver+0x29/0x50
 <783419b6> serio_driver_probe+0x16/0x20  <782a8fb4>
driver_probe_device+0x44/0xd0
 <782a9048> __device_attach+0x8/0x10  <782a8563> bus_for_each_drv+0x63/0x90
 <782a90a6> device_attach+0x56/0x60  <782a868e> bus_attach_device+0x1e/0x40
 <782a7763> device_add+0x113/0x180  <7834299d> serio_thread+0x1cd/0x2bb
 <781322c6> kthread+0xc6/0xca  <78101005> kernel_thread_helper+0x5/0xb
Quoted 5 lines

Thanks.

So we're taking ps2->cmd_mutex and then we're recurring back into ps2_command() and then taking ps2->serio->cmd_mutex.

Right, these are 2 different mutextes, one protects the child PS/2 device and the other protects parent PS/2 device accessed via pass-through port (synaptics_pt_write).

Quoted 4 lines

I suspect that's all correct/natural/expected and needs another make-lockdep-shut-up patch.

Signature · 1 line
Dmitry
#3 lore raw

PATCH 1/1 Declare explicit, hardware based lock ranking in serio

In reply to Andrew Morton
Quoted 8 lines

Thanks.

So we're taking ps2->cmd_mutex and then we're recurring back into ps2_command() and then taking ps2->serio->cmd_mutex.

I suspect that's all correct/natural/expected and needs another make-lockdep-shut-up patch.

The PS/2 code has a natural device order and there is a one level recursion in this device order in terms of the cmd_mutex; annotate this explicit recursion as ok

---
 drivers/input/serio/libps2.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/input/serio/libps2.c +1 -1
Index: linux-2.6.17-rc5-mm2/drivers/input/serio/libps2.c
===================================================================
@@ -177,7 +177,7 @@ int ps2_command(struct ps2dev *ps2dev, u
 		return -1;
 	}
 
-	mutex_lock(&ps2dev->cmd_mutex);
+	mutex_lock_nested(&ps2dev->cmd_mutex, SINGLE_DEPTH_NESTING);
 
 	serio_pause_rx(ps2dev->serio);
 	ps2dev->flags = command == PS2_CMD_GETID ? PS2_FLAG_WAITID : 0;
#4 lore raw

Re: [patch] Declare explicit, hardware based lock ranking in serio

Quoted 5 lines · Arjan van de Ven
Quoted 5 lines

Thanks.

So we're taking ps2->cmd_mutex and then we're recurring back into ps2_command() and then taking ps2->serio->cmd_mutex.

Quoted 8 lines
Quoted 2 lines

I suspect that's all correct/natural/expected and needs another make-lockdep-shut-up patch.

The PS/2 code has a natural device order and there is a one level recursion in this device order in terms of the cmd_mutex; annotate this explicit recursion as ok

It is not necessarily single depth - one could have 2 or more pass-through ports chained together, although currently in kernel we only have Synaptics pass-through. If we were ever to implement pass-through port for IBMs trackpoints then we'd have:

	Synaptics<->pass-through<->TP<->pass-through<->some mouse
Signature · 1 line
Dmitry
#5 lore raw

Re: [patch] Declare explicit, hardware based lock ranking in serio

Quoted 15 lines · Dmitry Torokhov
Quoted 5 lines · Arjan van de Ven
Quoted 5 lines

Thanks.

So we're taking ps2->cmd_mutex and then we're recurring back into ps2_command() and then taking ps2->serio->cmd_mutex.

Quoted 8 lines
Quoted 2 lines

I suspect that's all correct/natural/expected and needs another make-lockdep-shut-up patch.

The PS/2 code has a natural device order and there is a one level recursion in this device order in terms of the cmd_mutex; annotate this explicit recursion as ok

Quoted 6 lines

It is not necessarily single depth - one could have 2 or more pass-through ports chained together, although currently in kernel we only have Synaptics pass-through. If we were ever to implement pass-through port for IBMs trackpoints then we'd have:

	Synaptics<->pass-through<->TP<->pass-through<->some mouse

is there a bound to this? lockdep can deal with upto 8 or so recursions....

#6 lore raw

Re: [patch] Declare explicit, hardware based lock ranking in serio

Quoted 21 lines · Arjan van de Ven
Quoted 20 lines · Dmitry Torokhov
Quoted 5 lines · Arjan van de Ven
Quoted 5 lines

Thanks.

So we're taking ps2->cmd_mutex and then we're recurring back into ps2_command() and then taking ps2->serio->cmd_mutex.

Quoted 8 lines
Quoted 2 lines

I suspect that's all correct/natural/expected and needs another make-lockdep-shut-up patch.

The PS/2 code has a natural device order and there is a one level recursion in this device order in terms of the cmd_mutex; annotate this explicit recursion as ok

It is not necessarily single depth - one could have 2 or more pass-through ports chained together, although currently in kernel we only have Synaptics pass-through. If we were ever to implement pass-through port for IBMs trackpoints then we'd have:

Quoted 5 lines
Quoted 1 line
	Synaptics<->pass-through<->TP<->pass-through<->some mouse

is there a bound to this? lockdep can deal with upto 8 or so recursions....

Theoretically - no, practically - 2 as shown above. At the moment the only devices that are using/may be using pass-through ports are Synaptics touchpad and IBM trackpoint.

Signature · 1 line
Dmitry