ar9170 deadlock report

ar9170 deadlock report

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 2031 bytes --]

[  328.121951] =======================================================
[  328.122169] [ INFO: possible circular locking dependency detected ]
[  328.122339] 2.6.30-rc5-wl-23310-ge0995c7-dirty #54
[  328.122468] -------------------------------------------------------
[  328.122636] pm-hibernate/7289 is trying to acquire lock:
[  328.122779]  (&(&ar->tx_status_janitor)->work){+.+...}, at: [<c00000000007eb14>] .wait_on_work+0x3c/0x194
[  328.123058] 
[  328.123059] but task is already holding lock:
[  328.123216]  (&ar->mutex){+.+...}, at: [<d0000000015c29a4>] .ar9170_op_stop+0x60/0x120 [ar9170usb]
[  328.123482] 
[  328.123482] which lock already depends on the new lock.
[  328.123484] 
[  328.123702] 
[  328.123703] the existing dependency chain (in reverse order) is:
[  328.123903] 
[  328.123904] -> #1 (&ar->mutex){+.+...}:
[  328.124066]        [<c00000000009ee40>] .check_prev_add+0x3d4/0x5f8
[  328.124242]        [<c00000000009f7b0>] .validate_chain+0x74c/0x930
[  328.128871]        [<c0000000000a01a0>] .__lock_acquire+0x80c/0x8f4
[  328.133531]        [<c0000000000a03d4>] .lock_acquire+0x14c/0x19c
[  328.138139]        [<c0000000004cb0f0>] .mutex_lock_nested+0xa8/0x504
[  328.142769]        [<d0000000015c414c>] .ar9170_tx_status_janitor+0x58/0x114 [ar9170usb]


[  328.156599] -> #0 (&(&ar->tx_status_janitor)->work){+.+...}:
[  328.156604]        [<c00000000009eb2c>] .check_prev_add+0xc0/0x5f8
[  328.156609]        [<c00000000009f7b0>] .validate_chain+0x74c/0x930
[  328.156614]        [<c0000000000a01a0>] .__lock_acquire+0x80c/0x8f4
[  328.156618]        [<c0000000000a03d4>] .lock_acquire+0x14c/0x19c
[  328.156623]        [<c00000000007eb54>] .wait_on_work+0x7c/0x194
[  328.156627]        [<c00000000007ed50>] .__cancel_work_timer+0xe4/0x164
[  328.156632]        [<c00000000007ee00>] .cancel_delayed_work_sync+0x30/0x48
[  328.156636]        [<d0000000015c29b0>] .ar9170_op_stop+0x6c/0x120 [ar9170usb]

Don't cancel a work under a lock that it needs!

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: ar9170 deadlock report

[Christian Lamparter]

On Friday 15 May 2009 12:50:42 Johannes Berg wrote:
> [  328.121951] =======================================================
> [  328.122169] [ INFO: possible circular locking dependency detected ]
> [  328.122339] 2.6.30-rc5-wl-23310-ge0995c7-dirty #54
> [  328.122468] -------------------------------------------------------
> [  328.122636] pm-hibernate/7289 is trying to acquire lock:
> [  328.122779]  (&(&ar->tx_status_janitor)->work){+.+...}, at: [<c00000000007eb14>] .wait_on_work+0x3c/0x194
> [  328.123058] 
> [  328.123059] but task is already holding lock:
> [  328.123216]  (&ar->mutex){+.+...}, at: [<d0000000015c29a4>] .ar9170_op_stop+0x60/0x120 [ar9170usb]
> [  328.123482] 
> [  328.123482] which lock already depends on the new lock.
> [  328.123484] 
> [  328.123702] 
> [  328.123703] the existing dependency chain (in reverse order) is:
> [  328.123903] 
> [  328.123904] -> #1 (&ar->mutex){+.+...}:
> [  328.124066]        [<c00000000009ee40>] .check_prev_add+0x3d4/0x5f8
> [  328.124242]        [<c00000000009f7b0>] .validate_chain+0x74c/0x930
> [  328.128871]        [<c0000000000a01a0>] .__lock_acquire+0x80c/0x8f4
> [  328.133531]        [<c0000000000a03d4>] .lock_acquire+0x14c/0x19c
> [  328.138139]        [<c0000000004cb0f0>] .mutex_lock_nested+0xa8/0x504
> [  328.142769]        [<d0000000015c414c>] .ar9170_tx_status_janitor+0x58/0x114 [ar9170usb]
> 
> 
> [  328.156599] -> #0 (&(&ar->tx_status_janitor)->work){+.+...}:
> [  328.156604]        [<c00000000009eb2c>] .check_prev_add+0xc0/0x5f8
> [  328.156609]        [<c00000000009f7b0>] .validate_chain+0x74c/0x930
> [  328.156614]        [<c0000000000a01a0>] .__lock_acquire+0x80c/0x8f4
> [  328.156618]        [<c0000000000a03d4>] .lock_acquire+0x14c/0x19c
> [  328.156623]        [<c00000000007eb54>] .wait_on_work+0x7c/0x194
> [  328.156627]        [<c00000000007ed50>] .__cancel_work_timer+0xe4/0x164
> [  328.156632]        [<c00000000007ee00>] .cancel_delayed_work_sync+0x30/0x48
> [  328.156636]        [<d0000000015c29b0>] .ar9170_op_stop+0x6c/0x120 [ar9170usb]
> 
> Don't cancel a work under a lock that it needs!
> 
> johannes
> 
do you have the module this time?

Regards,
	Chr

Re: ar9170 deadlock report

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 534 bytes --]

On Fri, 2009-05-15 at 13:18 +0200, Christian Lamparter wrote:

> > Don't cancel a work under a lock that it needs!
> > 
> > johannes
> > 
> do you have the module this time?

Yes, why? The bug is trivial:

        mutex_lock(&ar->mutex);
        cancel_delayed_work_sync(&ar->tx_status_janitor);
        cancel_work_sync(&ar->filter_config_work);
        cancel_work_sync(&ar->beacon_work);

vs.

static void ar9170_tx_status_janitor(struct work_struct *work)
{
	...
        mutex_lock(&ar->mutex);


johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 801 bytes --]