Date: Mon, 21 Jul 2014 12:01:20 -0400
From: "J. Bruce Fields" <bfields@fieldses.org>
To: Jeff Layton <jlayton@primarydata.com>
Cc: hch@infradead.org, linux-nfs@vger.kernel.org
Subject: Re: [PATCH v5 00/10] nfsd: more delegation fixes to prepare for
 client_mutex removal
Message-ID: <20140721160120.GE8438@fieldses.org>
References: <1405949706-27757-1-git-send-email-jlayton@primarydata.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <1405949706-27757-1-git-send-email-jlayton@primarydata.com>
Sender: linux-nfs-owner@vger.kernel.org

On Mon, Jul 21, 2014 at 09:34:56AM -0400, Jeff Layton wrote:
> v5:
> - fix spinlocking in block_delegations. Lock should be held around all
>   of the set_bit calls so we don't race with the swap of the two
>   fields.
> 
> - eliminate destroy_revoked_delegation (just use nfs4_put_delegation)
> 
> - eliminate unneeded NULL pointer check in nfs4_setlease
> 
> v4:
> - close more potential races in setlease code, and fix some bugs in
>   error handling in that code.
> 
> - clean up delegation setting functions, eliminating unused arguments
>   and avoiding allocations when there has already been a delegation
>   break
> 
> - add separate spinlock for block_delegations/delegation_blocked code
> 
> v3:
> - use alternate method for checking for delegation break races after
>   getting a lease (just check fi_had_conflict instead)
> 
> - drop file_has_lease patch -- no longer needed
> 
> - move cl_revoked handling patch into this set. It means altering a
>   few of the later patches, but it keeps the set more topically
>   coherent
> 
> v2:
> - move remove_stid call from nfs4_free_stid and into callers
> 
> Ok, I'm hoping we're now on the final respin here. This just cleans
> up some nits that Christoph noticed in the last set. It also fixes
> a potential bug that Neil noticed around the block_delegations
> spinlocking.
> 
> Bruce, I noticed that you had merged the last set into your nfsd-next
> branch, so let me know if you'd prefer me to do incremental changes
> on top of that instead.

As long as it's just in nfsd-next I'll take replacements, if it's in
master (== for-3.17) then I'm considering it really and truly committed
and would rather take incremental patches.  (Though I haven't been
really great about sticking to that.)

--b.

> 
> Jeff Layton (7):
>   nfsd: Protect the nfs4_file delegation fields using the fi_lock
>   nfsd: Fix delegation revocation
>   nfsd: ensure that clp->cl_revoked list is protected by clp->cl_lock
>   nfsd: drop unused stp arg to alloc_init_deleg
>   nfsd: clean up arguments to nfs4_open_delegation
>   nfsd: clean up nfs4_set_delegation
>   nfsd: give block_delegation and delegation_blocked its own spinlock
> 
> Trond Myklebust (3):
>   nfsd: Move the delegation reference counter into the struct nfs4_stid
>   nfsd: simplify stateid allocation and file handling
>   nfsd: Convert delegation counter to an atomic_long_t type
> 
>  fs/nfsd/nfs4state.c | 253 ++++++++++++++++++++++++++++++++--------------------
>  fs/nfsd/state.h     |   2 +-
>  2 files changed, 159 insertions(+), 96 deletions(-)
> 
> -- 
> 1.9.3
>