RE: [PATCH 03/34] pnfs: let layoutcommit code handle multiple segments

[<tao.peng@emc.com>]

Hi, Fred,

> -----Original Message-----
> From: linux-nfs-owner@vger.kernel.org [mailto:linux-nfs-owner@vger.kernel.org]
> On Behalf Of Fred Isaman
> Sent: Monday, June 13, 2011 10:37 PM
> To: Jim Rees
> Cc: linux-nfs@vger.kernel.org; peter honeyman
> Subject: Re: [PATCH 03/34] pnfs: let layoutcommit code handle multiple segments
> 
> On Sun, Jun 12, 2011 at 7:43 PM, Jim Rees <rees@umich.edu> wrote:
> > From: Peng Tao <bergwolf@gmail.com>
> >
> > Some layout driver like block will have multiple segments.
> > Generic code should be able to handle it.
> >
> > Signed-off-by: Peng Tao <peng_tao@emc.com>
> > ---
> >  fs/nfs/pnfs.c |   17 +++++++++++++----
> >  fs/nfs/pnfs.h |    1 +
> >  2 files changed, 14 insertions(+), 4 deletions(-)
> >
> > diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
> > index e3d618b..f03a5e0 100644
> > --- a/fs/nfs/pnfs.c
> > +++ b/fs/nfs/pnfs.c
> > @@ -892,7 +892,7 @@ pnfs_find_lseg(struct pnfs_layout_hdr *lo,
> >        dprintk("%s:Begin\n", __func__);
> >
> >        assert_spin_locked(&lo->plh_inode->i_lock);
> > -       list_for_each_entry(lseg, &lo->plh_segs, pls_list) {
> > +       list_for_each_entry_reverse(lseg, &lo->plh_segs, pls_list) {
> 
> This is a sortred list, and the order of search matters.  You can't
> just reverse it here.
The layout segment list is in offset increasing order. But the lookup code here assumes it's a decreasing ordered list.
To fix it, we should either reverse lookup the list, or change the break condition test. Otherwise lookup always fails if not matching the first one.

> 
> >                if (test_bit(NFS_LSEG_VALID, &lseg->pls_flags) &&
> >                    is_matching_lseg(&lseg->pls_range, range)) {
> >                        ret = get_lseg(lseg);
> > @@ -1193,10 +1193,18 @@ pnfs_try_to_read_data(struct nfs_read_data *rdata,
> >  static struct pnfs_layout_segment *pnfs_list_write_lseg(struct inode *inode)
> >  {
> >        struct pnfs_layout_segment *lseg, *rv = NULL;
> > +       loff_t max_pos = 0;
> > +
> > +       list_for_each_entry(lseg, &NFS_I(inode)->layout->plh_segs, pls_list) {
> > +               if (lseg->pls_range.iomode == IOMODE_RW) {
> > +                       if (max_pos < lseg->pls_end_pos)
> > +                               max_pos = lseg->pls_end_pos;
> > +                       if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
> &lseg->pls_flags))
> > +                               rv = lseg;
> > +               }
> > +       }
> > +       rv->pls_end_pos = max_pos;
> >
> 
> The idea here was that it could be extended to use segment by
> returning a list of affected lsegs,
> not so,e random one.  Because otherwise you have problems with the
> fact that relevant but not
> returned lsegs are going to get there refcounts messed up.
The above code relies on NFS_INO_LAYOUTCOMMIT bit to ensure that only one inode lseg has NFS_LSEG_LAYOUTCOMMIT set. But, you are right. The layoutcommit code needs a second thought.
How about making it return a list of affected lsegs and pass them around layoutcommit_procs?

Thanks,
Tao

> 
> Fred
> 
> > -       list_for_each_entry(lseg, &NFS_I(inode)->layout->plh_segs, pls_list)
> > -               if (lseg->pls_range.iomode == IOMODE_RW)
> > -                       rv = lseg;
> >        return rv;
> >  }
> >
> > @@ -1211,6 +1219,7 @@ pnfs_set_layoutcommit(struct nfs_write_data *wdata)
> >        if (!test_and_set_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) {
> >                /* references matched in nfs4_layoutcommit_release */
> >                get_lseg(wdata->lseg);
> > +               set_bit(NFS_LSEG_LAYOUTCOMMIT,
> &wdata->lseg->pls_flags);
> >                wdata->lseg->pls_lc_cred =
> >                        get_rpccred(wdata->args.context->state->owner-
> >so_cred);
> >                mark_as_dirty = true;
> > diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
> > index b071b56..a3fc0f2 100644
> > --- a/fs/nfs/pnfs.h
> > +++ b/fs/nfs/pnfs.h
> > @@ -36,6 +36,7 @@
> >  enum {
> >        NFS_LSEG_VALID = 0,     /* cleared when lseg is recalled/returned */
> >        NFS_LSEG_ROC,           /* roc bit received from server */
> > +       NFS_LSEG_LAYOUTCOMMIT,  /* layoutcommit bit set for
> layoutcommit */
> >  };
> >
> >  struct pnfs_layout_segment {
> > --
> > 1.7.4.1
> >
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> >
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html