[PATCH] softmac: Fix deadlock of wx_set_essid with assoc work

[PATCH] softmac: Fix deadlock of wx_set_essid with assoc work

[Michael Buesch]

The essid wireless extension does deadlock against the assoc mutex,
as we don't unlock the assoc mutex.

Signed-off-by: Michael Buesch <mb@bu3sch.de>

--

David, please test if this fixes your deadlock.

Index: bu3sch-wireless-dev/net/ieee80211/softmac/ieee80211softmac_wx.c
===================================================================
--- bu3sch-wireless-dev.orig/net/ieee80211/softmac/ieee80211softmac_wx.c	2007-03-05 18:42:18.000000000 +0100
+++ bu3sch-wireless-dev/net/ieee80211/softmac/ieee80211softmac_wx.c	2007-07-31 14:51:46.000000000 +0200
@@ -98,9 +98,12 @@ ieee80211softmac_wx_set_essid(struct net
 				cancel_delayed_work(&authptr->work);
 			sm->associnfo.bssvalid = 0;
 			sm->associnfo.bssfixed = 0;
-			flush_scheduled_work();
 			sm->associnfo.associating = 0;
 			sm->associnfo.associated = 0;
+			/* Unlock mutex, otherwise we deadlock with the assoc work handler. */
+			mutex_unlock(&sm->associnfo.mutex);
+			flush_scheduled_work();
+			mutex_lock(&sm->associnfo.mutex);
 		}
 	}
 

-- 
Greetings Michael.

Re: [PATCH] softmac: Fix deadlock of wx_set_essid with assoc work

[David Woodhouse]

On Tue, 2007-07-31 at 14:57 +0200, Michael Buesch wrote:
> The essid wireless extension does deadlock against the assoc mutex,
> as we don't unlock the assoc mutex.
> 
> Signed-off-by: Michael Buesch <mb@bu3sch.de>
> 
> --
> 
> David, please test if this fixes your deadlock. 

Seems to, although I'll have to leave it a few more hours before I'm
sure. Thanks.

-- 
dwmw2

Re: [PATCH] softmac: Fix deadlock of wx_set_essid with assoc work

[Michael Buesch]

On Tuesday 31 July 2007 17:01:00 David Woodhouse wrote:
> On Tue, 2007-07-31 at 14:57 +0200, Michael Buesch wrote:
> > The essid wireless extension does deadlock against the assoc mutex,
> > as we don't unlock the assoc mutex.
> > 
> > Signed-off-by: Michael Buesch <mb@bu3sch.de>
> > 
> > --
> > 
> > David, please test if this fixes your deadlock. 
> 
> Seems to, although I'll have to leave it a few more hours before I'm
> sure. Thanks.

Ok, this is a deadlock, anyway. Even if it's not "your" deadlock. :)
So I'll build a correct patch for this (this was an incomplete
hack) and submit it.

-- 
Greetings Michael.

[PATCH] softmac: Fix deadlock of wx_set_essid with assoc work

[Michael Buesch]

The essid wireless extension does deadlock against the assoc mutex,
as we don't unlock the assoc mutex.

Signed-off-by: Michael Buesch <mb@bu3sch.de>

--

This is included in John's wireless-2.6 tree.

This doesn't fix the whole issue completely, as we shouldn't
flush the workqueue while we are holding the rtnl lock.
But, it turns this bug from a "will 100% always deadlock" into
a "it might deadlock" bug.
A followup patch will come (if someone has some time to do so)
to fix this crap code completely.

Index: bu3sch-wireless-dev/net/ieee80211/softmac/ieee80211softmac_wx.c
===================================================================
--- bu3sch-wireless-dev.orig/net/ieee80211/softmac/ieee80211softmac_wx.c	2007-03-05 18:42:18.000000000 +0100
+++ bu3sch-wireless-dev/net/ieee80211/softmac/ieee80211softmac_wx.c	2007-07-31 14:51:46.000000000 +0200
@@ -98,9 +98,12 @@ ieee80211softmac_wx_set_essid(struct net
 				cancel_delayed_work(&authptr->work);
 			sm->associnfo.bssvalid = 0;
 			sm->associnfo.bssfixed = 0;
-			flush_scheduled_work();
 			sm->associnfo.associating = 0;
 			sm->associnfo.associated = 0;
+			/* Unlock mutex, otherwise we deadlock with the assoc work handler. */
+			mutex_unlock(&sm->associnfo.mutex);
+			flush_scheduled_work();
+			mutex_lock(&sm->associnfo.mutex);
 		}
 	}
 


-- 
Greetings Michael.

Re: [PATCH] softmac: Fix deadlock of wx_set_essid with assoc work

[Michael Buesch]

On Tuesday 07 August 2007 12:02:58 Michael Buesch wrote:
> The essid wireless extension does deadlock against the assoc mutex,
> as we don't unlock the assoc mutex.
> 
> Signed-off-by: Michael Buesch <mb@bu3sch.de>
> 
> --
> 
> This is included in John's wireless-2.6 tree.
> 
> This doesn't fix the whole issue completely, as we shouldn't
> flush the workqueue while we are holding the rtnl lock.
> But, it turns this bug from a "will 100% always deadlock" into
> a "it might deadlock" bug.
> A followup patch will come (if someone has some time to do so)
> to fix this crap code completely.
> 
> Index: bu3sch-wireless-dev/net/ieee80211/softmac/ieee80211softmac_wx.c
> ===================================================================
> --- bu3sch-wireless-dev.orig/net/ieee80211/softmac/ieee80211softmac_wx.c	2007-03-05 18:42:18.000000000 +0100
> +++ bu3sch-wireless-dev/net/ieee80211/softmac/ieee80211softmac_wx.c	2007-07-31 14:51:46.000000000 +0200
> @@ -98,9 +98,12 @@ ieee80211softmac_wx_set_essid(struct net
>  				cancel_delayed_work(&authptr->work);
>  			sm->associnfo.bssvalid = 0;
>  			sm->associnfo.bssfixed = 0;
> -			flush_scheduled_work();
>  			sm->associnfo.associating = 0;
>  			sm->associnfo.associated = 0;
> +			/* Unlock mutex, otherwise we deadlock with the assoc work handler. */
> +			mutex_unlock(&sm->associnfo.mutex);
> +			flush_scheduled_work();
> +			mutex_lock(&sm->associnfo.mutex);
>  		}
>  	}
>  
> 
> 

Whoops, no. I attached the wrong patch. Sorry!

-- 
Greetings Michael.