xfs: reimplement fdatasync support

Add an in-memory only flag to say we logged timestamps only, and use it to
check if fdatasync can optimize away the log force.

Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>

4 files changed, 18 insertions, 5 deletions

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 78d8b0299592..54a67dd9ac0a 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -197,8 +197,11 @@ xfs_file_fsync(
          * to flush the log up to the latest LSN that touched the inode.
          */
         xfs_ilock(ip, XFS_ILOCK_SHARED);
-        if (xfs_ipincount(ip))
-                lsn = ip->i_itemp->ili_last_lsn;
+        if (xfs_ipincount(ip)) {
+                if (!datasync ||
+                    (ip->i_itemp->ili_fields & ~XFS_ILOG_TIMESTAMP))
+                        lsn = ip->i_itemp->ili_last_lsn;
+        }
         xfs_iunlock(ip, XFS_ILOCK_SHARED);
 
         if (lsn)
diff --git a/fs/xfs/xfs_inode_item.c b/fs/xfs/xfs_inode_item.c
index 8becef4f9e6a..05d924efceaf 100644
--- a/fs/xfs/xfs_inode_item.c
+++ b/fs/xfs/xfs_inode_item.c
@@ -438,7 +438,8 @@ out:
          * games in recovery easier, which isn't a big deal as just about any
          * transaction would dirty it anyway.
          */
-        iip->ili_format.ilf_fields = XFS_ILOG_CORE | iip->ili_fields;
+        iip->ili_format.ilf_fields = XFS_ILOG_CORE |
+                (iip->ili_fields & ~XFS_ILOG_TIMESTAMP);
         iip->ili_format.ilf_size = nvecs;
 }
 
diff --git a/fs/xfs/xfs_inode_item.h b/fs/xfs/xfs_inode_item.h
index bc183d81ad32..41d61c3b7a36 100644
--- a/fs/xfs/xfs_inode_item.h
+++ b/fs/xfs/xfs_inode_item.h
@@ -86,6 +86,15 @@ typedef struct xfs_inode_log_format_64 {
 #define XFS_ILOG_AEXT   0x080   /* log i_af.if_extents */
 #define XFS_ILOG_ABROOT 0x100   /* log i_af.i_broot */
 
+
+/*
+ * The timestamps are dirty, but not necessarily anything else in the inode
+ * core.  Unlike the other fields above this one must never make it to disk
+ * in the ilf_fields of the inode_log_format, but is purely store in-memory in
+ * ili_fields in the inode_log_item.
+ */
+#define XFS_ILOG_TIMESTAMP      0x4000
+
 #define XFS_ILOG_NONCORE        (XFS_ILOG_DDATA | XFS_ILOG_DEXT | \
                                  XFS_ILOG_DBROOT | XFS_ILOG_DEV | \
                                  XFS_ILOG_UUID | XFS_ILOG_ADATA | \
@@ -101,7 +110,7 @@ typedef struct xfs_inode_log_format_64 {
                                  XFS_ILOG_DEXT | XFS_ILOG_DBROOT | \
                                  XFS_ILOG_DEV | XFS_ILOG_UUID | \
                                  XFS_ILOG_ADATA | XFS_ILOG_AEXT | \
-                                 XFS_ILOG_ABROOT)
+                                 XFS_ILOG_ABROOT | XFS_ILOG_TIMESTAMP)
 
 static inline int xfs_ilog_fbroot(int w)
 {
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index e602c8c67c5c..e9ad7894648e 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -907,7 +907,7 @@ xfs_fs_dirty_inode(
         ip->i_d.di_mtime.t_nsec = (__int32_t)inode->i_mtime.tv_nsec;
 
         xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
-        xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
+        xfs_trans_log_inode(tp, ip, XFS_ILOG_TIMESTAMP);
         error = xfs_trans_commit(tp, 0);
         if (error)
                 goto trouble;