BUG: Problem with your patches for sysfs from 2 weeks ago

[Alan Stern]

Maneesh:

I've been tracing a problem with sysfs that starting showing up just
recently, and it seems likely to have originated with your patches from a
couple of weeks ago.  I can't tell exactly what's wrong or fix it because
I don't understand the filesystem layer well enough.

The problem I found (there may be others) shows up when trying to delete a 
nonexistent symlink -- presumably trying to delete a nonexistent file 
would have a similar result.  The code in sysfs_hash_and_remove() does a 
lookup on the nonexistent name and sysfs_get_dentry() returns a 
newly-allocated dentry.  Creating the new entry increments the parent 
directory's d_count, of course.  But at the end of the routine, when 
dput() is called for the new dentry, the parent's d_count does _not_ get 
decremented.  The new dentry is placed on the dentry_unused list and the 
parent is left with an anomalously large d_count.  This doesn't ever seem 
to get resolved, and when the directory's kobject is deleted the reference 
you added doesn't get dropped.

Alan Stern

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