Re: [PATCH] slub: Fix sysfs circular locking dependency

[Christoph Lameter]

On Tue, 4 Jan 2011, David Rientjes wrote:

> slub_lock protects slab_state following kmem_cache_init() if caches are
> created/destroyed prior to the sysfs slab initcall setting the global
> variable, so couldn't this leak the kobject if its initialization and add
> was deferred on kmem_cache_create() and added by slab_sysfs_init()?

Potentially.

Also note that the same lock inversion occurs in kmem_cache_create().
slub lock is taken and then sysfs operations are performed.

The problem is that slub_lock has multiple roles:

1. Protect the list of slab caches and the slab_state (thats why its
   mainly used in kmem_cache_destroy and kmem_cache_create)

2. Protect against removal and adding of new cpus and nodes
   (show_slab_objects)


If we could create another lock that protects against new cpus / nodes
being added and removed then we could take that lock in
show_slab_objects() instead.

IMHO the lock order must have the slub_lock at the top.

The other lock that would prevent adding/removal of nodes / cpu can be
taken in show_slab_objects and in the corresponding hotplug functions. We
do not need to take the lock at all in show_slab_objects cpu and memory
hotplug are not enabled. Maybe there is a hotplug specific lock that can
be used instead?


--
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/