From: Benny Halevy <bhalevy@panasas.com>
To: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: linux-nfs@vger.kernel.org
Subject: Re: [RFC 07/27] pnfs: encode_layoutcommit
Date: Fri, 22 Apr 2011 11:48:54 +0300 [thread overview]
Message-ID: <4DB140F6.7020604@panasas.com> (raw)
In-Reply-To: <1303330705.23206.38.camel@lade.trondhjem.org>
On 2011-04-20 23:18, Trond Myklebust wrote:
> On Wed, 2011-04-20 at 20:27 +0300, Benny Halevy wrote:
>> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
>> ---
>> fs/nfs/nfs4xdr.c | 16 +++++++++++++---
>> fs/nfs/pnfs.h | 4 ++++
>> 2 files changed, 17 insertions(+), 3 deletions(-)
>>
>> diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
>> index 6b64dd8..4f7bef9 100644
>> --- a/fs/nfs/nfs4xdr.c
>> +++ b/fs/nfs/nfs4xdr.c
>> @@ -1877,6 +1877,7 @@ encode_layoutget(struct xdr_stream *xdr,
>>
>> static int
>> encode_layoutcommit(struct xdr_stream *xdr,
>> + struct inode *inode,
>> const struct nfs4_layoutcommit_args *args,
>> struct compound_hdr *hdr)
>> {
>> @@ -1885,7 +1886,7 @@ encode_layoutcommit(struct xdr_stream *xdr,
>> dprintk("%s: lbw: %llu type: %d\n", __func__, args->lastbytewritten,
>> NFS_SERVER(args->inode)->pnfs_curr_ld->id);
>>
>> - p = reserve_space(xdr, 48 + NFS4_STATEID_SIZE);
>> + p = reserve_space(xdr, 44 + NFS4_STATEID_SIZE);
>> *p++ = cpu_to_be32(OP_LAYOUTCOMMIT);
>> /* Only whole file layouts */
>> p = xdr_encode_hyper(p, 0); /* offset */
>> @@ -1896,7 +1897,14 @@ encode_layoutcommit(struct xdr_stream *xdr,
>> p = xdr_encode_hyper(p, args->lastbytewritten);
>> *p++ = cpu_to_be32(0); /* Never send time_modify_changed */
>> *p++ = cpu_to_be32(NFS_SERVER(args->inode)->pnfs_curr_ld->id);/* type */
>> - *p++ = cpu_to_be32(0); /* no file layout payload */
>> +
>> + if (NFS_SERVER(inode)->pnfs_curr_ld->encode_layoutcommit)
>> + NFS_SERVER(inode)->pnfs_curr_ld->encode_layoutcommit(
>> + NFS_I(inode)->layout, xdr, args);
>> + else {
>> + p = reserve_space(xdr, 4);
>> + *p = cpu_to_be32(0); /* no layout-type payload */
>> + }
>>
>> hdr->nops++;
>> hdr->replen += decode_layoutcommit_maxsz;
>> @@ -2759,6 +2767,8 @@ static void nfs4_xdr_enc_layoutcommit(struct rpc_rqst *req,
>> struct xdr_stream *xdr,
>> struct nfs4_layoutcommit_args *args)
>> {
>> + struct nfs4_layoutcommit_data *data =
>> + container_of(args, struct nfs4_layoutcommit_data, args);
>> struct compound_hdr hdr = {
>> .minorversion = nfs4_xdr_minorversion(&args->seq_args),
>> };
>> @@ -2766,7 +2776,7 @@ static void nfs4_xdr_enc_layoutcommit(struct rpc_rqst *req,
>> encode_compound_hdr(xdr, req, &hdr);
>> encode_sequence(xdr, &args->seq_args, &hdr);
>> encode_putfh(xdr, NFS_FH(args->inode), &hdr);
>> - encode_layoutcommit(xdr, args, &hdr);
>> + encode_layoutcommit(xdr, data->args.inode, args, &hdr);
>> encode_getfattr(xdr, args->bitmask, &hdr);
>> encode_nops(&hdr);
>> }
>> diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
>> index 51dcbc1..011885e 100644
>> --- a/fs/nfs/pnfs.h
>> +++ b/fs/nfs/pnfs.h
>> @@ -99,6 +99,10 @@ struct pnfs_layoutdriver_type {
>> /* device notification methods */
>> void (*delete_deviceid)(struct nfs4_deviceid *);
>>
>> + void (*encode_layoutcommit) (struct pnfs_layout_hdr *layoutid,
>> + struct xdr_stream *xdr,
>> + const struct nfs4_layoutcommit_args *args);
>
> This too is way too ugly. Can't the layout payload be pre-encoded by the
> layout driver?
>
>
The contents for the objects layout driver are dynamic, representing
attribute-level metadata rather than something derived from the layout
that can be pre-encoded. In addition, I'd like to avoid having to
alloc and free a buffer for this stuff and rather have the layout driver
encode whatever it needs in-line.
Benny
next prev parent reply other threads:[~2011-04-22 8:49 UTC|newest]
Thread overview: 54+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-04-20 16:46 [RFC 0/27] pnfs-submit for 2.6.40 Benny Halevy
2011-04-20 17:26 ` [RFC 01/27] pnfs: CB_NOTIFY_DEVICEID Benny Halevy
2011-04-20 19:41 ` Trond Myklebust
2011-04-22 6:22 ` Benny Halevy
2011-04-20 17:26 ` [RFC 02/27] pnfs: direct i/o Benny Halevy
2011-04-20 17:26 ` [RFC 03/27] pnfs: layoutreturn Benny Halevy
2011-04-20 19:53 ` Trond Myklebust
2011-04-22 6:52 ` Benny Halevy
2011-04-22 8:04 ` [PATCH 1/6] SQUASHME: call pnfs_return_layout right before pnfs_destroy_layout Benny Halevy
2011-04-22 8:04 ` [PATCH 2/6] SQUASHME: remove assert_spin_locked from pnfs_clear_lseg_list Benny Halevy
2011-04-22 8:04 ` [PATCH 3/6] SQUASHME: remove wait parameter from the layoutreturn path Benny Halevy
2011-04-22 8:31 ` Benny Halevy
2011-04-22 8:05 ` [PATCH 4/6] SQUASHME: remove return_type field from nfs4_layoutreturn_args Benny Halevy
2011-04-22 8:05 ` [PATCH 5/6] SQUASHME: remove range " Benny Halevy
2011-04-22 8:05 ` [PATCH 6/6] SQUASHME: no need to send layoutcommit from _pnfs_return_layout Benny Halevy
2011-04-20 17:26 ` [RFC 04/27] pnfs: layoutret_on_setattr Benny Halevy
2011-04-20 20:03 ` Trond Myklebust
2011-04-22 8:23 ` Benny Halevy
2011-04-20 17:26 ` [RFC 05/27] pnfs: Use byte-range layout segments Benny Halevy
2011-04-20 17:26 ` [RFC 06/27] pnfs: encode_layoutreturn Benny Halevy
2011-04-20 20:16 ` Trond Myklebust
2011-04-22 8:26 ` Benny Halevy
2011-04-20 17:27 ` [RFC 07/27] pnfs: encode_layoutcommit Benny Halevy
2011-04-20 20:18 ` Trond Myklebust
2011-04-22 8:48 ` Benny Halevy [this message]
2011-04-20 17:27 ` [RFC 08/27] pnfs: {setup,cleanup}_layoutcommit Benny Halevy
2011-04-20 20:22 ` Trond Myklebust
2011-04-20 17:27 ` [RFC 09/27] pnfs: support for non-rpc layout drivers Benny Halevy
2011-04-20 20:34 ` Trond Myklebust
2011-04-22 9:03 ` Benny Halevy
2011-04-20 17:27 ` [RFC 10/27] pnfs: {,un}set_layoutdriver methods Benny Halevy
2011-04-20 17:27 ` [RFC 11/27] pnfs: per mount layout driver private data Benny Halevy
2011-04-20 20:36 ` Trond Myklebust
2011-04-22 9:05 ` Benny Halevy
2011-04-20 17:27 ` [RFC 12/27] pnfs: alloc and free layout_hdr layoutdriver methods Benny Halevy
2011-04-20 20:43 ` Trond Myklebust
2011-04-22 9:09 ` Benny Halevy
2011-04-20 17:27 ` [RFC 13/27] pnfs: client stats Benny Halevy
2011-04-20 17:28 ` [RFC 14/27] pnfsd: introduce exp_xdr.h Benny Halevy
2011-04-20 17:28 ` [RFC 15/27] pnfs-obj: pnfs_osd XDR definitions Benny Halevy
2011-04-20 20:49 ` Trond Myklebust
2011-04-22 9:11 ` Benny Halevy
2011-04-20 17:28 ` [RFC 16/27] pnfs-obj: pnfs_osd XDR client implementations Benny Halevy
2011-04-20 17:28 ` [RFC 17/27] exofs: pnfs-tree: Remove pnfs-osd private definitions Benny Halevy
2011-04-20 17:28 ` [RFC 18/27] pnfs-obj: Define PNFS_OBJLAYOUT Kconfig option Benny Halevy
2011-04-20 17:28 ` [RFC 19/27] pnfs-obj: objlayout driver skeleton Benny Halevy
2011-04-20 17:28 ` [RFC 20/27] pnfs-obj: objio_osd device information retrieval and caching Benny Halevy
2011-04-20 17:28 ` [RFC 21/27] pnfs-obj: objio_osd real IO implementation Benny Halevy
2011-04-20 17:29 ` [RFC 22/27] sunrpc: New xdr_rewind_stream() Benny Halevy
2011-04-20 17:29 ` [RFC 23/27] pnfs-obj: objlayout_encode_layoutreturn Implementation Benny Halevy
2011-04-20 17:29 ` [RFC 24/27] pnfs-obj: objio_osd report osd_errors for layoutreturn Benny Halevy
2011-04-20 17:29 ` [RFC 25/27] pnfs-obj: objlayout_encode_layoutcommit implementation Benny Halevy
2011-04-20 17:29 ` [RFC 26/27] pnfs-obj: objio_osd: RAID0 support Benny Halevy
2011-04-20 17:29 ` [RFC 27/27] pnfs-obj: objio_osd: groups support Benny Halevy
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=4DB140F6.7020604@panasas.com \
--to=bhalevy@panasas.com \
--cc=Trond.Myklebust@netapp.com \
--cc=linux-nfs@vger.kernel.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