[XFS] write barrier support Issue all log sync operations as ordered

writes.  In addition flush the disk cache on fsync if the sync cached
operation didn't sync the log to disk (this requires some additional
bookeping in the transaction and log code). If the device doesn't claim to
support barriers, the filesystem has an extern log volume or the trial
superblock write with barriers enabled failed we disable barriers and
print a warning.  We should probably fail the mount completely, but that
could lead to nasty boot failures for the root filesystem.  Not enabled by
default yet, needs more destructive testing first.

SGI-PV: 912426
SGI-Modid: xfs-linux:xfs-kern:198723a

Signed-off-by: Christoph Hellwig <hch@sgi.com>
Signed-off-by: Nathan Scott <nathans@sgi.com>

4 files changed, 76 insertions, 4 deletions

diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c
index 188cbbd5b74a..4663f7dbff1c 100644
--- a/fs/xfs/linux-2.6/xfs_buf.c
+++ b/fs/xfs/linux-2.6/xfs_buf.c
@@ -1295,6 +1295,11 @@ _pagebuf_ioapply(
                 rw = (pb->pb_flags & PBF_READ) ? READ : WRITE;
         }
 
+        if (pb->pb_flags & PBF_ORDERED) {
+                ASSERT(!(pb->pb_flags & PBF_READ));
+                rw = WRITE_BARRIER;
+        }
+
         /* Special code path for reading a sub page size pagebuf in --
          * we populate up the whole page, and hence the other metadata
          * in the same page.  This optimization is only valid when the
diff --git a/fs/xfs/linux-2.6/xfs_buf.h b/fs/xfs/linux-2.6/xfs_buf.h
index 39c8ca122534..fa21d1f9cb0b 100644
--- a/fs/xfs/linux-2.6/xfs_buf.h
+++ b/fs/xfs/linux-2.6/xfs_buf.h
@@ -74,7 +74,7 @@ typedef enum page_buf_flags_e {		/* pb_flags values */
         PBF_DELWRI = (1 << 6),  /* buffer has dirty pages                  */
         PBF_STALE = (1 << 7),   /* buffer has been staled, do not find it  */
         PBF_FS_MANAGED = (1 << 8),  /* filesystem controls freeing memory  */
-        PBF_FLUSH = (1 << 11),      /* flush disk write cache              */
+        PBF_ORDERED = (1 << 11),    /* use ordered writes                  */
         PBF_READ_AHEAD = (1 << 12), /* asynchronous read-ahead             */
 
         /* flags used only as arguments to access routines */
@@ -383,9 +383,9 @@ extern void pagebuf_trace(
 #define XFS_BUF_UNASYNC(x)       ((x)->pb_flags &= ~PBF_ASYNC)
 #define XFS_BUF_ISASYNC(x)       ((x)->pb_flags & PBF_ASYNC)
 
-#define XFS_BUF_FLUSH(x)         ((x)->pb_flags |= PBF_FLUSH)
-#define XFS_BUF_UNFLUSH(x)       ((x)->pb_flags &= ~PBF_FLUSH)
-#define XFS_BUF_ISFLUSH(x)       ((x)->pb_flags & PBF_FLUSH)
+#define XFS_BUF_ORDERED(x)       ((x)->pb_flags |= PBF_ORDERED)
+#define XFS_BUF_UNORDERED(x)     ((x)->pb_flags &= ~PBF_ORDERED)
+#define XFS_BUF_ISORDERED(x)     ((x)->pb_flags & PBF_ORDERED)
 
 #define XFS_BUF_SHUT(x)          printk("XFS_BUF_SHUT not implemented yet\n")
 #define XFS_BUF_UNSHUT(x)        printk("XFS_BUF_UNSHUT not implemented yet\n")
diff --git a/fs/xfs/linux-2.6/xfs_super.c b/fs/xfs/linux-2.6/xfs_super.c
index 2302454d8d47..d2701cc624b9 100644
--- a/fs/xfs/linux-2.6/xfs_super.c
+++ b/fs/xfs/linux-2.6/xfs_super.c
@@ -278,6 +278,72 @@ xfs_blkdev_put(
                 close_bdev_excl(bdev);
 }
 
+/*
+ * Try to write out the superblock using barriers.
+ */
+STATIC int
+xfs_barrier_test(
+        xfs_mount_t     *mp)
+{
+        xfs_buf_t       *sbp = xfs_getsb(mp, 0);
+        int             error;
+
+        XFS_BUF_UNDONE(sbp);
+        XFS_BUF_UNREAD(sbp);
+        XFS_BUF_UNDELAYWRITE(sbp);
+        XFS_BUF_WRITE(sbp);
+        XFS_BUF_UNASYNC(sbp);
+        XFS_BUF_ORDERED(sbp);
+
+        xfsbdstrat(mp, sbp);
+        error = xfs_iowait(sbp);
+
+        /*
+         * Clear all the flags we set and possible error state in the
+         * buffer.  We only did the write to try out whether barriers
+         * worked and shouldn't leave any traces in the superblock
+         * buffer.
+         */
+        XFS_BUF_DONE(sbp);
+        XFS_BUF_ERROR(sbp, 0);
+        XFS_BUF_UNORDERED(sbp);
+
+        xfs_buf_relse(sbp);
+        return error;
+}
+
+void
+xfs_mountfs_check_barriers(xfs_mount_t *mp)
+{
+        int error;
+
+        if (mp->m_logdev_targp != mp->m_ddev_targp) {
+                xfs_fs_cmn_err(CE_NOTE, mp,
+                  "Disabling barriers, not supported with external log device");
+                mp->m_flags &= ~XFS_MOUNT_BARRIER;
+        }
+
+        if (mp->m_ddev_targp->pbr_bdev->bd_disk->queue->ordered ==
+                                        QUEUE_ORDERED_NONE) {
+                xfs_fs_cmn_err(CE_NOTE, mp,
+                  "Disabling barriers, not supported by the underlying device");
+                mp->m_flags &= ~XFS_MOUNT_BARRIER;
+        }
+
+        error = xfs_barrier_test(mp);
+        if (error) {
+                xfs_fs_cmn_err(CE_NOTE, mp,
+                  "Disabling barriers, trial barrier write failed");
+                mp->m_flags &= ~XFS_MOUNT_BARRIER;
+        }
+}
+
+void
+xfs_blkdev_issue_flush(
+        xfs_buftarg_t           *buftarg)
+{
+        blkdev_issue_flush(buftarg->pbr_bdev, NULL);
+}
 
 STATIC struct inode *
 linvfs_alloc_inode(
diff --git a/fs/xfs/linux-2.6/xfs_super.h b/fs/xfs/linux-2.6/xfs_super.h
index ec7e0035c731..ad77e3743e04 100644
--- a/fs/xfs/linux-2.6/xfs_super.h
+++ b/fs/xfs/linux-2.6/xfs_super.h
@@ -132,6 +132,7 @@ extern void xfs_flush_device(struct xfs_inode *);
 extern int  xfs_blkdev_get(struct xfs_mount *, const char *,
                                 struct block_device **);
 extern void xfs_blkdev_put(struct block_device *);
+extern void xfs_blkdev_issue_flush(struct xfs_buftarg *);
 
 extern struct export_operations linvfs_export_ops;