On Mon, 2012-09-24 at 10:47 -0400, Andy Adamson wrote:
> On Fri, Sep 21, 2012 at 4:50 PM, Trond Myklebust
> <Trond.Myklebust@xxxxxxxxxx> wrote:
> > If pnfs_layout_io_test_failed() authorises a retry of the failed layoutgets,
> > it should first clear the existing layout segments so that we start afresh.
> >
> > Signed-off-by: Trond Myklebust <Trond.Myklebust@xxxxxxxxxx>
> > ---
> > fs/nfs/pnfs.c | 8 ++++++--
> > 1 file changed, 6 insertions(+), 2 deletions(-)
> >
> > diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
> > index cbbb0fc..80aaaba 100644
> > --- a/fs/nfs/pnfs.c
> > +++ b/fs/nfs/pnfs.c
> > @@ -256,7 +256,8 @@ pnfs_layout_io_set_failed(struct pnfs_layout_hdr *lo, u32 iomode)
> > }
> >
> > static bool
> > -pnfs_layout_io_test_failed(struct pnfs_layout_hdr *lo, u32 iomode)
> > +pnfs_layout_io_test_failed(struct pnfs_layout_hdr *lo, u32 iomode,
> > + struct list_head *head)
> > {
> > unsigned long start, end;
> > if (test_bit(pnfs_iomode_to_fail_bit(iomode), &lo->plh_flags) == 0)
> > @@ -267,6 +268,7 @@ pnfs_layout_io_test_failed(struct pnfs_layout_hdr *lo, u32 iomode)
> > /* It is time to retry the failed layoutgets */
> > clear_bit(NFS_LAYOUT_RW_FAILED, &lo->plh_flags);
> > clear_bit(NFS_LAYOUT_RO_FAILED, &lo->plh_flags);
> > + pnfs_mark_matching_lsegs_invalid(lo, head, NULL);
>
> By clearing all iomode failed bits and passing a NULL as the
> recall_range to pnfs_mark_matching_lsegs, you are discarding layouts
> not authorized by the iomode passed into pnfs_layout_io_test_failed.
> Why not clear just the failed bit and setup a pnfs_layout_range
> structure (for the whole file) and pass the authorized failed iomode
> to pnfs_mark_matching_lsegs so as to remove only the failed iomode
> layout segments?
There were other reasons for changing this part of the code too. The
main problem is that if the first layoutget fails, so that we never
associate an lseg to the pnfs_layout_hdr, then the plh_refcount will go
to zero, and we end up not caching the fail bit.
Version 3 of these patches therefore change the code in this region so
that we keep a reference to the pnfs_layout_hdr with the fail bit. Then
it clears out the existing lsegs (matching iomodes) in
pnfs_layout_io_set_failed.
> -->Andy
>
>
> > return false;
> > }
> > return true;
> > @@ -1058,6 +1060,7 @@ pnfs_update_layout(struct inode *ino,
> > struct nfs_client *clp = server->nfs_client;
> > struct pnfs_layout_hdr *lo;
> > struct pnfs_layout_segment *lseg = NULL;
> > + LIST_HEAD(tmp_list);
> > bool first = false;
> >
> > if (!pnfs_enabled_sb(NFS_SERVER(ino)))
> > @@ -1081,7 +1084,7 @@ pnfs_update_layout(struct inode *ino,
> > }
> >
> > /* if LAYOUTGET already failed once we don't try again */
> > - if (pnfs_layout_io_test_failed(nfsi->layout, iomode))
> > + if (pnfs_layout_io_test_failed(nfsi->layout, iomode, &tmp_list))
> > goto out_unlock;
> >
> > /* Check to see if the layout for the given range already exists */
> > @@ -1127,6 +1130,7 @@ pnfs_update_layout(struct inode *ino,
> > }
> > atomic_dec(&lo->plh_outstanding);
> > out:
> > + pnfs_free_lseg_list(&tmp_list);
> > pnfs_put_layout_hdr(lo);
> > dprintk("%s end, state 0x%lx lseg %p\n", __func__,
> > nfsi->layout ? nfsi->layout->plh_flags : -1, lseg);
> > --
> > 1.7.11.4
> >
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> > the body of a message to majordomo@xxxxxxxxxxxxxxx
> > More majordomo info at http://vger.kernel.org/majordomo-info.html
--
Trond Myklebust
Linux NFS client maintainer
NetApp
Trond.Myklebust@xxxxxxxxxx
www.netapp.com
��.n��������+%������w��{.n�����{��w��� jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
