lkml.org 
[lkml]   [2010]   [Feb]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH] sysfs: differentiate between locking links and non-links
From
On Wed, Feb 10, 2010 at 9:09 AM, Neil Brown <neilb@suse.de> wrote:
>
> Hi,
>  I've just spent a while sorting out some lockdep complaints triggered
>  by the recent addition of the "s_active" lockdep annotation in sysfs
>  (commit 846f99749ab68bbc7f75c74fec305de675b1a1bf)
>
>  Some of them are genuine and I have submitted a fix for those.
>  Some are, I think, debatable and I get to that is a minute.  I've
>  submitted a fix for them anyway.
>  But some are to my mind clearly bogus and I'm hoping that can be
>  fixed by the change below (or similar).
>  The 'bogus' ones are triggered by writing to a sysfs attribute file
>  for which the handler tries to delete a symlink from sysfs.
>  This appears to be a recursion on s_active as s_active is held while
>  the handler runs and is again needed to effect the delete.  However
>  as the thing being deleted is a symlink, it is very clearly a
>  different object to the thing triggering the delete, so there is no
>  real loop.
>
>  The following patch splits the lockdep context in two - one for
>  symlink and one for everything else.  This removes the apparent loop.
>  (An example report can be seen in
>     http://bugzilla.kernel.org/show_bug.cgi?id=15142).
>
>  The "debatable" dependency loops happen when writing to one attribute
>  causes a different attribute to be deleted.  In my (md) case this can
>  actually cause a deadlock as both the attributes take the same lock
>  while the handler is running.  This is because deleting the attribute
>  will block until the all accesses of that attribute have completed (I
>  think).
>  However it should be possible to delete a name from sysfs while there
>  are still accesses pending (it works for normal files!!).  So if
>  sysfs could be changed to simply unlink the file and leave deletion to
>  happen when the refcount become zero it would certainly make my life
>  a lot easier, and allow the removal of some ugly code from md.c.
>  I don't know sysfs well enough to suggest a patch though.
>

Hi, Neil,

Thanks for your patch.

This bug report is new for me. Recently we received lots of sysfs lockdep
warnings, I am working on a patch to fix all the bogus ones.

However, this one is _not_ similar to the other cases, as you decribed.
This patch could fix the problem, but not a good fix, IMO. We need more
work in sysfs layer to fix this kind of things. I will take care of this.

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

\
 
 \ /
  Last update: 2010-02-10 03:11    [W:0.119 / U:2.952 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site