lkml.org 
[lkml]   [2014]   [Jan]   [5]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] kobject: provide kobject_put_wait to fix module unload race
On Sun, Jan 05, 2014 at 05:11:23PM -0500, Mikulas Patocka wrote:
>
>
> On Sat, 4 Jan 2014, Dmitry Torokhov wrote:
>
> > On Sat, Jan 04, 2014 at 06:34:03PM +0000, Al Viro wrote:
> > > On Sat, Jan 04, 2014 at 10:16:20AM -0800, Greg Kroah-Hartman wrote:
> > >
> > > > > I came up with a simpler patch to achieve the same purpose - this patch
> > > > > makes fixing the drivers easy - the driver is fixed just by replacing
> > > > > "kobject_put" with "kobject_put_wait" in the unload routine.
> > > >
> > > > No, that's not ok at all.
> > >
> > > Agreed - all it takes is one cargo-culter who religoiusly does such
> > > conversion and drops a ref to parent before that to child.
> > >
> > > > > However, this pattern is buggy with respect to modules. The release method
> > > > > is placed in the driver's module. When the driver exits, the module
> > > > > reference count is zero, thus the module may be freed. However, there may
> > > > > still be references to the kobject. If the module is unloaded and then the
> > > > > release method is called, a crash happens.
> > > >
> > > > Yes, module unloading while a kobject is still "active" is not a good
> > > > thing, what modules do you have that cause this problem? Why not just
> > > > grab the module reference in your kobject if you need this type of
> > > > protection? It's not the kobject's code fault that this issue is there,
> > > > or that we now have a "delayed release" function to expose this type of
> > > > thing, it's the user of the kobject.
> > > >
> > > > Please fix the broken users of the kobject first.
> > >
> > > <snide> Are you saying that there is another kind? </snide>
> > >
> > > When would you grab that reference to module? More to the point, when
> > > would you *drop* it? Doing so from module_exit is not going to work,
> > > obviously...
> >
> > You normally have subsystem core module that does handle release of its
> > objects and users of said objects so it is usually OK for objects to
> > outlive the users, you just need to make sure the core stays around.
> >
> > In input we grab module reference to input core when we allocate input
> > device and drop it when input device is freed. This way we can be sure
> > that input core stays around until all input devices are gone. The same
> > for serio.
> >
> > Thanks.
> >
> > --
> > Dmitry
>
> But sometimes, the driver itself needs to create nodes in the sysfs
> filesystem (for example drivers/md/dm-sysfs.c). I don't quite see how
> would you push all driver-specific sysfs nodes into the generic non-module
> code.

Then you need to make sure your driver does not allow unloading while
its objects are active. I.e. require that all your devices are gone
(by increasing module count when you create a DM object and decreasing
it when you release DM object) before you allow unloading the driver.

Basically we should avoid kobject_put() in exit paths of the module.
Then we are left with that tiny race with release being preempted and
module getting unloaded.

Thanks.

--
Dmitry


\
 
 \ /
  Last update: 2014-01-06 00:01    [W:0.061 / U:1.176 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site