public inbox for linux-nfs@vger.kernel.org
 help / color / mirror / Atom feed
From: "J. Bruce Fields" <bfields@fieldses.org>
To: andros@netapp.com
Cc: linux-nfs@vger.kernel.org, pnfs@linux-nfs.org
Subject: Re: [PATCH 02/29] nfsd41: encode create_session result into cache
Date: Thu, 23 Apr 2009 19:21:05 -0400	[thread overview]
Message-ID: <20090423232105.GI1906@fieldses.org> (raw)
In-Reply-To: <1240504988-9572-3-git-send-email-andros@netapp.com>

On Thu, Apr 23, 2009 at 12:42:41PM -0400, andros@netapp.com wrote:
> From: Andy Adamson <andros@netapp.com>
> 
> CREATE_SESSION can be preceeded by a SEQUENCE operation and the
> create session single slot cache must be maintained. Encode the results of
> a create session call into the cache at the end of processing.
> 
> The create session result will also be encoded into the RPC response, and if
> it is preceeded by a SEQUENCE operation, will also be encoded into the
> session slot table cache.
> 
> Errors that do not change the create session cache:
> A create session NFS4ERR_STALE_CLIENTID error means that a client record
> (and associated create session slot) could not be found and therefore can't
> be changed.  NFSERR_SEQ_MISORDERED errors do not change the slot cache.
> 
> All other errors get cached.

Thanks, this all sounds sensible.  (One simple thing I think I was
confused about: the create_session sequencing is done at a layer above
sessions; so in the case of a create_session preceded by a sequence, if
we discover at the point of processing the sequence that this is a
replay, we use the sequence replay cache and never even get to any
create_session processing.  OK!)

> Signed-off-by: Andy Adamson <andros@netapp.com>
> ---
>  fs/nfsd/nfs4state.c       |    6 +++++-
>  fs/nfsd/nfs4xdr.c         |   19 +++++++++++++++++++
>  include/linux/nfsd/xdr4.h |    2 ++
>  3 files changed, 26 insertions(+), 1 deletions(-)
> 
> diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> index 27ad37f..279b47e 100644
> --- a/fs/nfsd/nfs4state.c
> +++ b/fs/nfsd/nfs4state.c
> @@ -1377,7 +1377,7 @@ nfsd4_create_session(struct svc_rqst *rqstp,
>  		if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
>  		    (ip_addr != unconf->cl_addr)) {

Could we apply the patch removing these ip_addr checks before any of
these patches?  Also I seem to recall expressing some other doubts about
the correctness of exchange_id and create_session--I'd like to see those
addressed first, as this builds on those.

>  			status = nfserr_clid_inuse;
> -			goto out;
> +			goto out_cache;
>  		}
>  
>  		slot = &unconf->cl_slot;
> @@ -1424,6 +1424,10 @@ nfsd4_create_session(struct svc_rqst *rqstp,
>  	cstate->slot = slot;
>  	/* Ensure a page is used for the cache */
>  	slot->sl_cache_entry.ce_cachethis = 1;

Is this still needed?

--b.

> +out_cache:
> +	/* cache solo and embedded create sessions under the state lock */
> +	nfsd4_cache_create_session(cr_ses, slot, status);
> +
>  out:
>  	nfs4_unlock_state();
>  	dprintk("%s returns %d\n", __func__, ntohl(status));
> diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
> index b2f8d74..6b34fac 100644
> --- a/fs/nfsd/nfs4xdr.c
> +++ b/fs/nfsd/nfs4xdr.c
> @@ -3046,6 +3046,25 @@ nfsd4_encode_create_session(struct nfsd4_compoundres *resp, int nfserr,
>  	return 0;
>  }
>  
> +/*
> + * Encode the create_session result into the create session single DRC
> + * slot cache. Do this for solo or embedded create session operations.
> + */
> +void
> +nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
> +			   struct nfsd4_slot *slot, int nfserr)
> +{
> +	struct nfsd4_cache_entry *entry = &slot->sl_cache_entry;
> +	__be32 *p = (__be32 *)entry->ce_datav.iov_base;
> +	struct nfsd4_compoundres tmp = {
> +		.p = p,
> +		.end = p + XDR_QUADLEN(CS_MAX_ENC_SZ),
> +	}, *resp = &tmp;
> +
> +	entry->ce_status = nfsd4_encode_create_session(resp, nfserr, cr_ses);
> +	entry->ce_datav.iov_len = (char *)resp->p - (char *)p;
> +}
> +
>  static __be32
>  nfsd4_encode_destroy_session(struct nfsd4_compoundres *resp, int nfserr,
>  			     struct nfsd4_destroy_session *destroy_session)
> diff --git a/include/linux/nfsd/xdr4.h b/include/linux/nfsd/xdr4.h
> index afd4464..cc50ca0 100644
> --- a/include/linux/nfsd/xdr4.h
> +++ b/include/linux/nfsd/xdr4.h
> @@ -517,6 +517,8 @@ extern __be32 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
>  extern void nfsd4_store_cache_entry(struct nfsd4_compoundres *resp);
>  extern __be32 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
>  		struct nfsd4_sequence *seq);
> +extern void nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
> +		struct nfsd4_slot *slot, int nfserr);
>  extern __be32 nfsd4_exchange_id(struct svc_rqst *rqstp,
>  		struct nfsd4_compound_state *,
>  struct nfsd4_exchange_id *);
> -- 
> 1.5.4.3
> 

  parent reply	other threads:[~2009-04-23 23:21 UTC|newest]

Thread overview: 46+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-04-23 16:42 [PATCH 0/29] NFSv4.1 Server DRC rewrite andros
2009-04-23 16:42 ` [PATCH 01/29] nfsd41: add create session slot buffer to struc nfs4_client andros
2009-04-23 16:42   ` [PATCH 02/29] nfsd41: encode create_session result into cache andros
2009-04-23 16:42     ` [PATCH 03/29] nfsd41: create_session check replay first andros
2009-04-23 16:42       ` [PATCH 04/29] nfsd41: replay solo and embedded create session andros
2009-04-23 16:42         ` [PATCH 05/29] nfsd41: create_session cache hold client reference andros
2009-04-23 16:42           ` [PATCH 06/29] nfsd41: no nfsd4_release_respages for the clientid cache andros
2009-04-23 16:42             ` [PATCH 07/29] nfsd41: slots are freed with session andros
2009-04-23 16:42               ` [PATCH 08/29] nfsd41: protect sv_drc_pages_used with spinlock andros
2009-04-23 16:42                 ` [PATCH 09/29] nfsd41: sanity check client drc maxreqs andros
2009-04-23 16:42                   ` [PATCH 10/29] nfsd41: change from page to memory based drc limits andros
2009-04-23 16:42                     ` [PATCH 11/29] nfsd41: set the session maximum response size cached andros
2009-04-23 16:42                       ` [PATCH 12/29] nfsd41: allocate and use drc cache buffers andros
2009-04-23 16:42                         ` [PATCH 13/29] nfsd41: free " andros
2009-04-23 16:42                           ` [PATCH 14/29] nfsd41: obliterate nfsd4_copy_pages andros
2009-04-23 16:42                             ` [PATCH 15/29] nfsd41: obliterate nfsd41_copy_replay_data andros
2009-04-23 16:42                               ` [PATCH 16/29] nfsd41: obliterate nfsd4_release_respages andros
2009-04-23 16:42                                 ` [PATCH 17/29] nfsd41: remove unused nfsd4_cache_entry fields andros
2009-04-23 16:42                                   ` [PATCH 18/29] nfsd41: obliterate nfsd4_set_statp andros
2009-04-23 16:42                                     ` [PATCH 19/29] nfsd41: rename nfsd4_enc_uncached_replay andros
2009-04-23 16:42                                       ` [PATCH 20/29] nfsd41: encode replay sequence from the slot values andros
2009-04-23 16:43                                         ` [PATCH 21/29] nfsd41: remove iovlen field from nfsd4_compound_state andros
2009-04-23 16:43                                           ` [PATCH 22/29] nfsd41: obliterate nfsd41_copy_replay_data andros
2009-04-23 16:43                                             ` [PATCH 23/29] nfsd41: fix nfsd4_store_cache_entry comments andros
2009-04-23 16:43                                               ` [PATCH 24/29] nfsd41: support 16 slots per session andros
2009-04-23 16:43                                                 ` [PATCH 25/29] nfsd41: use the maximum operations per compound in nfsd4_compoundargs andros
2009-04-23 16:43                                                   ` [PATCH 26/29] nfsd41: fix nfsd4_store_cache_entry dprintk andros
2009-04-23 16:43                                                     ` [PATCH 27/29] nfsd41: add test for failed sequence operation andros
2009-04-23 16:43                                                       ` [PATCH 28/29] nfsd41: remove redundant failed sequence check andros
2009-04-23 16:43                                                         ` [PATCH 29/29] nfsd41: remove check for session andros
2009-04-23 23:39                           ` [PATCH 13/29] nfsd41: free drc cache buffers J. Bruce Fields
2009-04-24 14:11                             ` [pnfs] " William A. (Andy) Adamson
2009-04-23 23:36                 ` [PATCH 08/29] nfsd41: protect sv_drc_pages_used with spinlock J. Bruce Fields
2009-04-24 14:11                   ` [pnfs] " William A. (Andy) Adamson
2009-04-23 23:28           ` [PATCH 05/29] nfsd41: create_session cache hold client reference J. Bruce Fields
2009-04-24 13:52             ` Andy Adamson
2009-04-24 14:02               ` J. Bruce Fields
2009-04-24 14:06                 ` [pnfs] " William A. (Andy) Adamson
2009-04-23 23:25         ` [PATCH 04/29] nfsd41: replay solo and embedded create session J. Bruce Fields
2009-04-23 23:21     ` J. Bruce Fields [this message]
2009-04-23 23:32       ` [PATCH 02/29] nfsd41: encode create_session result into cache J. Bruce Fields
2009-04-24 13:56         ` Andy Adamson
2009-04-24 13:52       ` Andy Adamson
2009-04-23 22:55   ` [PATCH 01/29] nfsd41: add create session slot buffer to struc nfs4_client J. Bruce Fields
2009-04-23 23:41 ` [PATCH 0/29] NFSv4.1 Server DRC rewrite J. Bruce Fields
2009-04-24 14:12   ` [pnfs] " William A. (Andy) Adamson

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20090423232105.GI1906@fieldses.org \
    --to=bfields@fieldses.org \
    --cc=andros@netapp.com \
    --cc=linux-nfs@vger.kernel.org \
    --cc=pnfs@linux-nfs.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox