On Wed, 2012-02-15 at 15:07 +0000, Adamson, Andy wrote:
>
> Please don't merge nfs4_reset_slot_tables and nfs4_init_slot_tables. It assumes a static array (realloc) which goes away with the dynamic slot table code. Instead, take a look at the patch set I sent on Feb 12
Why not? There is _no_ functional difference between what
nfs4_reset_slot_tables() and nfs4_init_slot_tables() need to do.
They both need to allocate new slots (conditionally in the case of
nfs4_reset_slot_tables, but the condition is compatible with the
nfs4_init_slot_tables case), and they both need to initialise those
slots to the value '1'.
AFAICS There is no reason for keeping those as 2 separate functions, and
I don't see how the dynamic session patches change anything to that
conclusion.
> [PATCH Version 7 3/3] NFSv4.1 avoid freeing slot when tasks are waiting
> [PATCH Version 1 2/3] NFSv4.1 prepare for dyamic session slots
> [PATCH Version 7 1/3] NFS4.1 clean up nfs4_alloc_session
I need a fix for the 3.3 final...
Those patches can be cleaned up and made ready for 3.4 (needs work -
they won't apply to the 'nfs-for-next' branch), but right now they're
not sufficiently tested nor are they sufficiently reviewed. For
instance, I'm not happy with the idea of adding a 'tk_private' field in
the struct rpc_task.
--
Trond Myklebust
Linux NFS client maintainer
NetApp
Trond.Myklebust@xxxxxxxxxx
www.netapp.com
��.n��������+%������w��{.n�����{��w��� jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
