From: Fred Isaman <iisaman@xxxxxxxxxxxxxx>
There is currently no good way for pnfs to communicate problems. For
example - the linux read code first tries to do readahead through
nfs_readpages. Failure there is ignored, and it will later call
nfs_readpage. Failure there is also ignored, except that the lack of
PG_uptodate is communicated back via -EIO.
With pnfs, it would be useful to be able to communicate to
nfs_readpage that direct disk IO failed on readahead, and that it
should failover to using the MDS.
Making the page flag PG_owner_priv_1 available as PG_pnfserr is one
way to do so. (An alternative would be to embed this in the layout,
but then pg_test can't easily access the info.)
This may be better as generic pnfs code, in which case it should be
put in pnfs.h, or even page-flags.h
Signed-off-by: Fred Isaman <iisaman@xxxxxxxxxxxxxx>
Signed-off-by: Benny Halevy <bhalevy@xxxxxxxxxxx>
---
fs/nfs/blocklayout/blocklayout.h | 5 +++++
1 files changed, 5 insertions(+), 0 deletions(-)
diff --git a/fs/nfs/blocklayout/blocklayout.h b/fs/nfs/blocklayout/blocklayout.h
index a596e75..0a54feb 100644
--- a/fs/nfs/blocklayout/blocklayout.h
+++ b/fs/nfs/blocklayout/blocklayout.h
@@ -35,6 +35,11 @@
#include <linux/nfs_fs.h>
#include "../pnfs.h"
+#define PG_pnfserr PG_owner_priv_1
+#define PagePnfsErr(page) test_bit(PG_pnfserr, &(page)->flags)
+#define SetPagePnfsErr(page) set_bit(PG_pnfserr, &(page)->flags)
+#define ClearPagePnfsErr(page) clear_bit(PG_pnfserr, &(page)->flags)
+
struct block_mount_id {
spinlock_t bm_lock; /* protects list */
struct list_head bm_devlist; /* holds pnfs_block_dev */
--
1.7.4.1
--
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
