Re: namespaces?: bug at mm/slub.c:2750

[Vegard Nossum <vegard.nossum@gmail.com>]

On Wed, Feb 11, 2009 at 6:24 PM, Serge E. Hallyn <serue@us.ibm.com> wrote:
> Quoting David Howells (dhowells@redhat.com):
>> Serge E. Hallyn <serue@us.ibm.com> wrote:
>>
>> >  static void uid_hash_remove(struct user_struct *up)
>> >  {
>> > +   put_user_ns(up->user_ns);
>> >     hlist_del_init(&up->uidhash_node);
>> >  }
>>
>> Don't you need to do the hlist_del_init() first?  Otherwise, mightn't the
>> put_user_ns() cause the namespace to be freed before hlist_del_init() removes
>> the user_struct from it?
>
> It's called under uidhash_lock spinlock so should be ok, but in
> principle you're right so it's probably a good idea.
>
> The main point is that without this patch, put_user_ns is done before
> the hlist_del_init and *not* atomically under uidhash_lock.

Congrats, your (unmodified) patch made it through the first 20 minutes
of testing! :-D

(In comparison, the unpatched kernel would usually crash after ~3 minutes)

I wonder why you couldn't reproduce it, though.

KOSAKI Motohiro: You might want to see if this patch helps too. It is
here: http://lkml.org/lkml/2009/2/11/251


Vegard

-- 
"The animistic metaphor of the bug that maliciously sneaked in while
the programmer was not looking is intellectually dishonest as it
disguises that the error is the programmer's own creation."
	-- E. W. Dijkstra, EWD1036