linux-kernel - Re: drm_vm.c:drm_mmap: possible circular locking dependency detected

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Mon, 4 Jan 2010 10:57:55 -0800
From:	Dmitry Torokhov <dmitry.torokhov@...il.com>
To:	"Eric W. Biederman" <ebiederm@...ssion.com>
Cc:	Tejun Heo <tj@...nel.org>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	KOSAKI Motohiro <kosaki.motohiro@...fujitsu.com>,
	Borislav Petkov <petkovbb@...glemail.com>,
	David Airlie <airlied@...ux.ie>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Greg KH <greg@...ah.com>, Al Viro <viro@...IV.linux.org.uk>
Subject: Re: drm_vm.c:drm_mmap: possible circular locking dependency detected

On Sun, Jan 03, 2010 at 02:57:15AM -0800, Eric W. Biederman wrote:
> Dmitry Torokhov <dmitry.torokhov@...il.com> writes:
> 
> >
> > Overall I am not concerned about lockdep bitching about serio because it
> > still bitches if you simply reload psmouse on a box with Synaptics with a
> > pass-through port even though there are nested annotations and it is
> > silent first time around.
> 
> This is a new lockdep annotation, and looking into it this appears to
> be a true possible deadlock in the serio/sysfs interactions.
> 
> We have serio_pin_driver() called from all of the sysfs attributes
> which does:
>    mutex_lock(&serio->drv_mutex);
> 
> We have serio_disconnect_driver() called on an unplug which does:
>    mutex_lock(&serio->drv_mutex);
> 
> The deadlock potential is if someone reads say the psmouse rate
> sysfs file while the mouse is being unplugged.  There is a race
> such that we can have:
> 
> 						  sysfs_read_file()
>                                                     fill_read_buffer()
> 						       sysfs_get_active_two()
> 							 psmouse_attr_show_helper()
>                                         		   serio_pin_driver()
> serio_disconnect_driver()		
>   mutex_lock(&serio->drv_mutex);		
> 				<----------------->	   mutex_lock(&serio_drv_mutex);
>     psmouse_disconnect()
>       sysfs_remove_group(... psmouse_attr_group);
> 	....
> 	sysfs_deactivate(); 
> 	  wait_for_completion();
> 
> 
> So it is unlikely but possible to deadlock by accessing a serio
> attribute of a serio device that is being removed.

Hmm, I guess I was too quick dismissing lockdep complaints here. Now
that sysfs remove waits deadlock indeed is possible. Actually the locks
on serio->drv_mutex in attributes were added to make sure we don't
access device that was unbound from the driver through stale sysfs
attributes.

> 
> What to do about it is another question.

I think we should simply not take serio->drv_mutex in attributes and use
driver-private mutex to serialize "set" methods that may alter device
state.

-- 
Dmitry
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/