Re: [PATCH 00/11] nfsd: add nfsdcltrack support and limit the use of md5 hashes in nfsdv4 code

From: "J. Bruce Fields" <bfields@fieldses.org>
To: Jeff Layton <jlayton@redhat.com>
Cc: linux-nfs@vger.kernel.org
Subject: Re: [PATCH 00/11] nfsd: add nfsdcltrack support and limit the use of md5 hashes in nfsdv4 code
Date: Mon, 12 Nov 2012 18:57:33 -0500	[thread overview]
Message-ID: <20121112235733.GB2219@fieldses.org> (raw)
In-Reply-To: <1352750458-16857-1-git-send-email-jlayton@redhat.com>

On Mon, Nov 12, 2012 at 03:00:47PM -0500, Jeff Layton wrote:
> This is the first "official" posting of the patchset that I sent as an
> RFC last week. I've spent some time testing the set now and I'm fairly
> convinced that it works properly. At this point, it's probably ready to
> soak in -next for a bit and if all goes well, merge in 3.8.

Thanks, applying, should be pushed out tomorrow.

Please send any further bugfixes as incremental patches.

--b.

> 
> The main changes from the last set are:
> 
> 1/ I've prefixed this set with the patches to add nfsdcltrack
> 
> 2/ fixed an off-by-one bug in nfsd4_cltrack_legacy_recdir()
> 
> 3/ error handling for nfs4_make_rec_clidname() has been cleaned up so
>    that errors from the functions called are returned to the caller
> 
> 4/ the callers of nfs4_make_rec_clidname in the legacy tracker now call
>    it earlier, which means that it's computed under fewer locks
> 
> 5/ if nfs4_make_rec_clidname returns -ENOENT, then the legacy tracker
>    will now disable the client ID tracking altogether and emit a
>    printk to warn that recovery won't work properly
> 
> 6/ the legacy tracker also frees the contents of the reclaim list in
>    its gracedone operation. There's no point in keeping that around
>    afterward.
> 
> Bruce, let me know if you see anything that needs addressing before
> you can put this into your for-next branch.
> 
> Thanks,
> 
> Jeff Layton (11):
>   nfsd: add a usermodehelper upcall for NFSv4 client ID tracking
>   nfsd: change heuristic for selecting the client_tracking_ops
>   nfsd: pass info about the legacy recoverydir in environment variables
>   nfsd: warn about impending removal of nfsdcld upcall
>   nfsd: have nfsd4_find_reclaim_client take a char * argument
>   nfsd: break out reclaim record removal into separate function
>   nfsd: make nfs4_client_to_reclaim return a pointer to the reclaim
>     record
>   nfsd: don't search for client by hash on legacy reboot recovery
>     gracedone
>   nfsd: move the confirmed and unconfirmed hlists to a rbtree
>   nfsd: get rid of cl_recdir field
>   nfsd: release the legacy reclaimable clients list in grace_done
> 
>  fs/nfsd/nfs4recover.c | 354 ++++++++++++++++++++++++++++++++++++++++++++++----
>  fs/nfsd/nfs4state.c   | 217 ++++++++++++++++++-------------
>  fs/nfsd/state.h       |  14 +-
>  3 files changed, 461 insertions(+), 124 deletions(-)
> 
> -- 
> 1.7.11.7
> 