[pnfs] [PATCH 05/14] 2.6-latest pnfs client read path layoutget
William A. (Andy) Adamson
andros at citi.umich.edu
Tue Nov 20 11:36:35 EST 2007
hi benny
comments in-line.
On 11/19/07, Benny Halevy <bhalevy at panasas.com> wrote:
>
> andros at umich.edu wrote:
> > From: Andy Adamson <andros at umich.edu>
> >
> > Use the pNFS ds_rsize for read pageio and ask for a layout.
> >
> > Signed-off by: Andy Adamson<andros at umich.edu>
> > ---
> > fs/nfs/nfs4proc.c | 1 -
> > fs/nfs/pnfs.c | 35 +++++++++++++++++++++++++++++++++--
> > fs/nfs/pnfs.h | 20 +++++++++++++++++++-
> > fs/nfs/read.c | 7 +++++++
> > 4 files changed, 59 insertions(+), 4 deletions(-)
> >
> > diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
> > index 79f8e93..7e1eb68 100644
> > --- a/fs/nfs/nfs4proc.c
> > +++ b/fs/nfs/nfs4proc.c
> > @@ -5252,7 +5252,6 @@ const struct nfs_rpc_ops pnfs_v41_clientops = {
> > .file_open = nfs_open,
> > .file_release = nfs_release,
> > .lock = nfs4_proc_lock,
> > - .rsize = pnfs_rsize,
> > .wsize = pnfs_wsize,
> > .rpages = pnfs_rpages,
> > .wpages = pnfs_wpages,
> > diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
> > index 3f765ad..90b78db 100644
> > --- a/fs/nfs/pnfs.c
> > +++ b/fs/nfs/pnfs.c
> > @@ -439,6 +439,38 @@ pnfs_inject_layout(struct nfs_inode* nfsi,
> > return io_ops->set_layout(layid, inode, lgr);
> > }
> >
> > +void
> > +pnfs_update_layout_read(struct inode *inode,
> > + struct nfs_open_context *ctx,
> > + struct list_head *pages,
> > + loff_t offset,
> > + size_t *rsize)
> > +{
> > + struct nfs_server *nfss = NFS_SERVER(inode);
> > + struct page *page;
> > + size_t count = 0;
> > + int status = 0;
> > +
> > + dprintk("--> %s inode %p ctx %p pages %p offset %lu\n",
> > + __func__, inode, ctx, pages, (unsigned long)offset);
> > +
> > + if (!pnfs_enabled_sb(nfss))
> > + return;
> > +
> > + /* Calculate the total read-ahead count */
> > + list_for_each_entry(page, pages, lru)
> > + count += pnfs_page_length(page, inode);
> > +
> > + dprintk("%s count %ld\n", __func__,(long int)count);
> > +
> > + *rsize = pnfs_rsize(inode, count);
>
> I'd consider implementing pnfs_rsize inline here since if
> count >= 0 && below_threshold(inode, count, 0)
ok - good idea.
I don't see
> the point of getting the layout.
i had the ordering wrong, the idea is to try to get a layout, check the
threshold, then set the rsize to ds_rsize. if we can't get a layout, don't
set the rsize to ds_rsize, leave it as MDS rsize, and do I/O throught the
MDS. if we setup the nfs page dache using ds_rsize, and then can't get a
layout, we may have to come back and set-up pages as MDS rsize.
> +
> > + status = virtual_update_layout(inode, ctx, count,
> > + offset, IOMODE_RW);
> > + dprintk("%s *rsize %Zd virt update returned %d\n",
> > + __func__, *rsize, status);
> > +}
> > +
> > /* Check to see if the module is handling which layouts need to be
> > * retrieved from the server. If they are not, then use retrieve based
> > * upon the returned data ranges from get_layout.
> > @@ -1483,12 +1515,11 @@ struct pnfs_client_operations pnfs_ops = {
> > };
> >
> > int
> > -pnfs_rsize(struct inode *inode, unsigned int count, struct
> nfs_read_data *rdata)
> > +pnfs_rsize(struct inode *inode, unsigned int count)
> > {
> > if (count >= 0 && below_threshold(inode, count, 0))
> > return NFS_SERVER(inode)->rsize;
> >
> > - rdata->pnfsflags |= PNFS_USE_DS;
> > return NFS_SERVER(inode)->ds_rsize;
> > }
> >
> > diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
> > index edfc732..681443c 100644
> > --- a/fs/nfs/pnfs.h
> > +++ b/fs/nfs/pnfs.h
> > @@ -46,13 +46,31 @@ int pnfs_use_nfsv4_rproto(struct inode *inode,
> ssize_t count);
> > unsigned int pnfs_getiosize(struct nfs_server *server);
> > int pnfs_commit(struct inode* inode, struct list_head *head, int sync,
> struct nfs_write_data *data);
> > int pnfs_try_to_commit(struct inode *, struct nfs_write_data *, struct
> list_head *, int);
> > -int pnfs_rsize(struct inode *, unsigned int, struct nfs_read_data *);
> > +int pnfs_rsize(struct inode *, unsigned int);
> > int pnfs_wsize(struct inode *, unsigned int, struct nfs_write_data *);
> > int pnfs_rpages(struct inode *);
> > int pnfs_wpages(struct inode *);
> > void pnfs_readpage_result_norpc(struct rpc_task *task, void *calldata);
> > void pnfs_writeback_done_norpc(struct rpc_task *, void *);
> > void pnfs_commit_done_norpc(struct rpc_task *, void *);
> > +void pnfs_update_layout_read(struct inode *, struct nfs_open_context *,
> struct list_head *, loff_t, size_t *);
> > +
> > +
> > +static inline
> > +unsigned int pnfs_page_length(struct page *page, struct inode *inode)
> > +{
> > + loff_t i_size = i_size_read(inode);
> > +
> > + if (i_size > 0) {
> > + pgoff_t end_index = (i_size - 1) >> PAGE_CACHE_SHIFT;
>
> 4 lines below should be indented one notch to the right.
ok
> + if (page->index < end_index)
> > + return PAGE_CACHE_SIZE;
> > + if (page->index == end_index)
>
> why not just "else"?
we want to return 0 if end_index > page->index....
> + return ((i_size - 1) & ~PAGE_CACHE_MASK) + 1;
> > + }
> > + return 0;
> > +}
> > +
> > #endif /* CONFIG_PNFS */
> >
> > #endif /* FS_NFS_PNFS_H */
> > diff --git a/fs/nfs/read.c b/fs/nfs/read.c
> > index dd98eca..7b07865 100644
> > --- a/fs/nfs/read.c
> > +++ b/fs/nfs/read.c
> > @@ -22,6 +22,10 @@
> >
> > #include <asm/system.h>
> > #include <linux/module.h>
> > +#ifdef CONFIG_PNFS
>
> At some point we'd want to separate the client side config option
> from the server's.
true. this should be a separate patch...
> +#include <linux/pnfs_xdr.h>
> > +#include "pnfs.h"
> > +#endif /* CONFIG_PNFS */
> >
> > #include "nfs4_fs.h"
> > #include "internal.h"
> > @@ -635,6 +639,9 @@ int nfs_readpages(struct file *filp, struct
> address_space *mapping,
> > return -EBADF;
> > } else
> > desc.ctx =
> get_nfs_open_context(nfs_file_open_context(filp));
> > +#ifdef CONFIG_PNFS
> > + pnfs_update_layout_read(inode, desc.ctx, pages, filp->f_pos,
> &rsize);
> > +#endif /* CONFIG_PNFS */
> > if (rsize < PAGE_CACHE_SIZE)
> > nfs_pageio_init(&pgio, inode, nfs_pagein_multi, rsize, 0);
> > else
>
> _______________________________________________
> pNFS mailing list
> pNFS at linux-nfs.org
> http://linux-nfs.org/cgi-bin/mailman/listinfo/pnfs
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://linux-nfs.org/pipermail/pnfs/attachments/20071120/092e1241/attachment.htm
More information about the pNFS
mailing list