[RFC v3] PCMCIA locking updates for 2.6.34

[Dominik Brodowski]

Hey,

On Tue, Jan 19, 2010 at 11:11:26PM +0100, Wolfram Sang wrote:
> > No lockdep-warnings anymore whatever I tried. kmemleak reports something, but
> 
> Sorry, today I got this (after applying the patch to make my D-Link DMF560TX
> work):
> 
> Jan 19 23:03:11 goodluck kernel: [   71.618969] =======================================================
> Jan 19 23:03:11 goodluck kernel: [   71.618975] [ INFO: possible circular locking dependency detected ]
> Jan 19 23:03:11 goodluck kernel: [   71.618982] 2.6.33-rc4-ninja-00148-gf842b63-dirty #69
> Jan 19 23:03:11 goodluck kernel: [   71.618988] -------------------------------------------------------
> Jan 19 23:03:11 goodluck kernel: [   71.618994] pccardd/1284 is trying to acquire lock:
> Jan 19 23:03:11 goodluck kernel: [   71.618999]  (events){+.+.+.}, at: [<801482d0>] flush_workqueue+0x0/0x70
> Jan 19 23:03:11 goodluck kernel: [   71.619019] 
> Jan 19 23:03:11 goodluck kernel: [   71.619021] but task is already holding lock:
> Jan 19 23:03:11 goodluck kernel: [   71.619027]  (&socket->skt_mutex){+.+.+.}, at: [<c1afe721>] pccardd+0x181/0x3a0 [pcmcia_core]

On Wed, Jan 20, 2010 at 10:30:54PM +0100, Wolfram Sang wrote:
> > The slot continues to work, though. It does not happen every time I release the
> > card. I could not trigger it a second time up to now.
> 
> Update: Just happens once after a cold-boot.

It seems that lockdep is disabled after one such warning anyways... Still, I
think I found the cause of the bug; it's solved in v4 of my locking update
patches: you can't use schedule_work holding a lock if the scheduled work
takes said lock, too...

Best,
	Dominik