Merge branch 'for-linus' of git://oss.sgi.com/xfs/xfs

* 'for-linus' of git://oss.sgi.com/xfs/xfs:
  xfs: stop calling filemap_fdatawait inside ->fsync
  fix readahead calculations in xfs_dir2_leaf_getdents()
  xfs: make sure xfs_sync_fsdata covers the log
  xfs: mark inodes dirty before issuing I/O
  xfs: cleanup ->sync_fs
  xfs: fix xfs_quiesce_data
  xfs: implement ->dirty_inode to fix timestamp handling

6 files changed, 112 insertions, 73 deletions

diff --git a/fs/xfs/linux-2.6/xfs_aops.c b/fs/xfs/linux-2.6/xfs_aops.c
index 381854461b28..c2e30eea74dc 100644
--- a/fs/xfs/linux-2.6/xfs_aops.c
+++ b/fs/xfs/linux-2.6/xfs_aops.c
@@ -186,19 +186,37 @@ xfs_destroy_ioend(
 }
 
 /*
+ * If the end of the current ioend is beyond the current EOF,
+ * return the new EOF value, otherwise zero.
+ */
+STATIC xfs_fsize_t
+xfs_ioend_new_eof(
+        xfs_ioend_t             *ioend)
+{
+        xfs_inode_t             *ip = XFS_I(ioend->io_inode);
+        xfs_fsize_t             isize;
+        xfs_fsize_t             bsize;
+
+        bsize = ioend->io_offset + ioend->io_size;
+        isize = MAX(ip->i_size, ip->i_new_size);
+        isize = MIN(isize, bsize);
+        return isize > ip->i_d.di_size ? isize : 0;
+}
+
+/*
  * Update on-disk file size now that data has been written to disk.
  * The current in-memory file size is i_size.  If a write is beyond
  * eof i_new_size will be the intended file size until i_size is
  * updated.  If this write does not extend all the way to the valid
  * file size then restrict this update to the end of the write.
  */
+
 STATIC void
 xfs_setfilesize(
         xfs_ioend_t             *ioend)
 {
         xfs_inode_t             *ip = XFS_I(ioend->io_inode);
         xfs_fsize_t             isize;
-        xfs_fsize_t             bsize;
 
         ASSERT((ip->i_d.di_mode & S_IFMT) == S_IFREG);
         ASSERT(ioend->io_type != IOMAP_READ);
@@ -206,16 +224,10 @@ xfs_setfilesize(
         if (unlikely(ioend->io_error))
                 return;
 
-        bsize = ioend->io_offset + ioend->io_size;
-
         xfs_ilock(ip, XFS_ILOCK_EXCL);
-
-        isize = MAX(ip->i_size, ip->i_new_size);
-        isize = MIN(isize, bsize);
-
-        if (ip->i_d.di_size < isize) {
+        isize = xfs_ioend_new_eof(ioend);
+        if (isize) {
                 ip->i_d.di_size = isize;
-                ip->i_update_core = 1;
                 xfs_mark_inode_dirty_sync(ip);
         }
 
@@ -404,10 +416,16 @@ xfs_submit_ioend_bio(
         struct bio      *bio)
 {
         atomic_inc(&ioend->io_remaining);
-
         bio->bi_private = ioend;
         bio->bi_end_io = xfs_end_bio;
 
+        /*
+         * If the I/O is beyond EOF we mark the inode dirty immediately
+         * but don't update the inode size until I/O completion.
+         */
+        if (xfs_ioend_new_eof(ioend))
+                xfs_mark_inode_dirty_sync(XFS_I(ioend->io_inode));
+
         submit_bio(WRITE, bio);
         ASSERT(!bio_flagged(bio, BIO_EOPNOTSUPP));
         bio_put(bio);
diff --git a/fs/xfs/linux-2.6/xfs_file.c b/fs/xfs/linux-2.6/xfs_file.c
index 629370974e57..eff61e2732af 100644
--- a/fs/xfs/linux-2.6/xfs_file.c
+++ b/fs/xfs/linux-2.6/xfs_file.c
@@ -176,14 +176,7 @@ xfs_file_fsync(
         struct dentry           *dentry,
         int                     datasync)
 {
-        struct inode            *inode = dentry->d_inode;
-        struct xfs_inode        *ip = XFS_I(inode);
-        int                     error;
-
-        /* capture size updates in I/O completion before writing the inode. */
-        error = filemap_fdatawait(inode->i_mapping);
-        if (error)
-                return error;
+        struct xfs_inode        *ip = XFS_I(dentry->d_inode);
 
         xfs_iflags_clear(ip, XFS_ITRUNCATED);
         return -xfs_fsync(ip);
diff --git a/fs/xfs/linux-2.6/xfs_iops.c b/fs/xfs/linux-2.6/xfs_iops.c
index da0159d99f82..cd42ef78f6b5 100644
--- a/fs/xfs/linux-2.6/xfs_iops.c
+++ b/fs/xfs/linux-2.6/xfs_iops.c
@@ -57,19 +57,22 @@
 #include <linux/fiemap.h>
 
 /*
- * Bring the atime in the XFS inode uptodate.
- * Used before logging the inode to disk or when the Linux inode goes away.
+ * Bring the timestamps in the XFS inode uptodate.
+ *
+ * Used before writing the inode to disk.
  */
 void
-xfs_synchronize_atime(
+xfs_synchronize_times(
         xfs_inode_t     *ip)
 {
         struct inode    *inode = VFS_I(ip);
 
-        if (!(inode->i_state & I_CLEAR)) {
-                ip->i_d.di_atime.t_sec = (__int32_t)inode->i_atime.tv_sec;
-                ip->i_d.di_atime.t_nsec = (__int32_t)inode->i_atime.tv_nsec;
-        }
+        ip->i_d.di_atime.t_sec = (__int32_t)inode->i_atime.tv_sec;
+        ip->i_d.di_atime.t_nsec = (__int32_t)inode->i_atime.tv_nsec;
+        ip->i_d.di_ctime.t_sec = (__int32_t)inode->i_ctime.tv_sec;
+        ip->i_d.di_ctime.t_nsec = (__int32_t)inode->i_ctime.tv_nsec;
+        ip->i_d.di_mtime.t_sec = (__int32_t)inode->i_mtime.tv_sec;
+        ip->i_d.di_mtime.t_nsec = (__int32_t)inode->i_mtime.tv_nsec;
 }
 
 /*
@@ -106,32 +109,20 @@ xfs_ichgtime(
         if ((flags & XFS_ICHGTIME_MOD) &&
             !timespec_equal(&inode->i_mtime, &tv)) {
                 inode->i_mtime = tv;
-                ip->i_d.di_mtime.t_sec = (__int32_t)tv.tv_sec;
-                ip->i_d.di_mtime.t_nsec = (__int32_t)tv.tv_nsec;
                 sync_it = 1;
         }
         if ((flags & XFS_ICHGTIME_CHG) &&
             !timespec_equal(&inode->i_ctime, &tv)) {
                 inode->i_ctime = tv;
-                ip->i_d.di_ctime.t_sec = (__int32_t)tv.tv_sec;
-                ip->i_d.di_ctime.t_nsec = (__int32_t)tv.tv_nsec;
                 sync_it = 1;
         }
 
         /*
-         * We update the i_update_core field _after_ changing
-         * the timestamps in order to coordinate properly with
-         * xfs_iflush() so that we don't lose timestamp updates.
-         * This keeps us from having to hold the inode lock
-         * while doing this.  We use the SYNCHRONIZE macro to
-         * ensure that the compiler does not reorder the update
-         * of i_update_core above the timestamp updates above.
+         * Update complete - now make sure everyone knows that the inode
+         * is dirty.
          */
-        if (sync_it) {
-                SYNCHRONIZE();
-                ip->i_update_core = 1;
+        if (sync_it)
                 xfs_mark_inode_dirty_sync(ip);
-        }
 }
 
 /*
@@ -506,10 +497,8 @@ xfs_vn_getattr(
         stat->gid = ip->i_d.di_gid;
         stat->ino = ip->i_ino;
         stat->atime = inode->i_atime;
-        stat->mtime.tv_sec = ip->i_d.di_mtime.t_sec;
-        stat->mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
-        stat->ctime.tv_sec = ip->i_d.di_ctime.t_sec;
-        stat->ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
+        stat->mtime = inode->i_mtime;
+        stat->ctime = inode->i_ctime;
         stat->blocks =
                 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
 
diff --git a/fs/xfs/linux-2.6/xfs_lrw.c b/fs/xfs/linux-2.6/xfs_lrw.c
index 49e4a6aea73c..072050f8d346 100644
--- a/fs/xfs/linux-2.6/xfs_lrw.c
+++ b/fs/xfs/linux-2.6/xfs_lrw.c
@@ -667,7 +667,7 @@ start:
                 xip->i_new_size = new_size;
 
         if (likely(!(ioflags & IO_INVIS)))
-                xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
+                file_update_time(file);
 
         /*
          * If the offset is beyond the size of the file, we have a couple
diff --git a/fs/xfs/linux-2.6/xfs_super.c b/fs/xfs/linux-2.6/xfs_super.c
index bdd41c8c342f..18a4b8e11df2 100644
--- a/fs/xfs/linux-2.6/xfs_super.c
+++ b/fs/xfs/linux-2.6/xfs_super.c
@@ -977,6 +977,28 @@ xfs_fs_inode_init_once(
 }
 
 /*
+ * Dirty the XFS inode when mark_inode_dirty_sync() is called so that
+ * we catch unlogged VFS level updates to the inode. Care must be taken
+ * here - the transaction code calls mark_inode_dirty_sync() to mark the
+ * VFS inode dirty in a transaction and clears the i_update_core field;
+ * it must clear the field after calling mark_inode_dirty_sync() to
+ * correctly indicate that the dirty state has been propagated into the
+ * inode log item.
+ *
+ * We need the barrier() to maintain correct ordering between unlogged
+ * updates and the transaction commit code that clears the i_update_core
+ * field. This requires all updates to be completed before marking the
+ * inode dirty.
+ */
+STATIC void
+xfs_fs_dirty_inode(
+        struct inode    *inode)
+{
+        barrier();
+        XFS_I(inode)->i_update_core = 1;
+}
+
+/*
  * Attempt to flush the inode, this will actually fail
  * if the inode is pinned, but we dirty the inode again
  * at the point when it is unpinned after a log write,
@@ -1126,7 +1148,7 @@ xfs_fs_put_super(
 }
 
 STATIC int
-xfs_fs_sync_super(
+xfs_fs_sync_fs(
         struct super_block      *sb,
         int                     wait)
 {
@@ -1134,23 +1156,23 @@ xfs_fs_sync_super(
         int                     error;
 
         /*
-         * Treat a sync operation like a freeze.  This is to work
-         * around a race in sync_inodes() which works in two phases
-         * - an asynchronous flush, which can write out an inode
-         * without waiting for file size updates to complete, and a
-         * synchronous flush, which wont do anything because the
-         * async flush removed the inode's dirty flag.  Also
-         * sync_inodes() will not see any files that just have
-         * outstanding transactions to be flushed because we don't
-         * dirty the Linux inode until after the transaction I/O
-         * completes.
+         * Not much we can do for the first async pass.  Writing out the
+         * superblock would be counter-productive as we are going to redirty
+         * when writing out other data and metadata (and writing out a single
+         * block is quite fast anyway).
+         *
+         * Try to asynchronously kick off quota syncing at least.
          */
-        if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE))
-                error = xfs_quiesce_data(mp);
-        else
-                error = xfs_sync_fsdata(mp, 0);
+        if (!wait) {
+                xfs_qm_sync(mp, SYNC_TRYLOCK);
+                return 0;
+        }
+
+        error = xfs_quiesce_data(mp);
+        if (error)
+                return -error;
 
-        if (unlikely(laptop_mode)) {
+        if (laptop_mode) {
                 int     prev_sync_seq = mp->m_sync_seq;
 
                 /*
@@ -1169,7 +1191,7 @@ xfs_fs_sync_super(
                                 mp->m_sync_seq != prev_sync_seq);
         }
 
-        return -error;
+        return 0;
 }
 
 STATIC int
@@ -1539,10 +1561,11 @@ xfs_fs_get_sb(
 static const struct super_operations xfs_super_operations = {
         .alloc_inode            = xfs_fs_alloc_inode,
         .destroy_inode          = xfs_fs_destroy_inode,
+        .dirty_inode            = xfs_fs_dirty_inode,
         .write_inode            = xfs_fs_write_inode,
         .clear_inode            = xfs_fs_clear_inode,
         .put_super              = xfs_fs_put_super,
-        .sync_fs                = xfs_fs_sync_super,
+        .sync_fs                = xfs_fs_sync_fs,
         .freeze_fs              = xfs_fs_freeze,
         .statfs                 = xfs_fs_statfs,
         .remount_fs             = xfs_fs_remount,
diff --git a/fs/xfs/linux-2.6/xfs_sync.c b/fs/xfs/linux-2.6/xfs_sync.c
index 320be6aea492..961df0a22c78 100644
--- a/fs/xfs/linux-2.6/xfs_sync.c
+++ b/fs/xfs/linux-2.6/xfs_sync.c
@@ -309,11 +309,15 @@ xfs_sync_attr(
 STATIC int
 xfs_commit_dummy_trans(
         struct xfs_mount        *mp,
-        uint                    log_flags)
+        uint                    flags)
 {
         struct xfs_inode        *ip = mp->m_rootip;
         struct xfs_trans        *tp;
         int                     error;
+        int                     log_flags = XFS_LOG_FORCE;
+
+        if (flags & SYNC_WAIT)
+                log_flags |= XFS_LOG_SYNC;
 
         /*
          * Put a dummy transaction in the log to tell recovery
@@ -331,13 +335,12 @@ xfs_commit_dummy_trans(
         xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
         xfs_trans_ihold(tp, ip);
         xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
-        /* XXX(hch): ignoring the error here.. */
         error = xfs_trans_commit(tp, 0);
-
         xfs_iunlock(ip, XFS_ILOCK_EXCL);
 
+        /* the log force ensures this transaction is pushed to disk */
         xfs_log_force(mp, 0, log_flags);
-        return 0;
+        return error;
 }
 
 int
@@ -385,7 +388,20 @@ xfs_sync_fsdata(
         else
                 XFS_BUF_ASYNC(bp);
 
-        return xfs_bwrite(mp, bp);
+        error = xfs_bwrite(mp, bp);
+        if (error)
+                return error;
+
+        /*
+         * If this is a data integrity sync make sure all pending buffers
+         * are flushed out for the log coverage check below.
+         */
+        if (flags & SYNC_WAIT)
+                xfs_flush_buftarg(mp->m_ddev_targp, 1);
+
+        if (xfs_log_need_covered(mp))
+                error = xfs_commit_dummy_trans(mp, flags);
+        return error;
 
  out_brelse:
         xfs_buf_relse(bp);
@@ -419,14 +435,16 @@ xfs_quiesce_data(
         /* push non-blocking */
         xfs_sync_data(mp, 0);
         xfs_qm_sync(mp, SYNC_TRYLOCK);
-        xfs_filestream_flush(mp);
 
-        /* push and block */
+        /* push and block till complete */
         xfs_sync_data(mp, SYNC_WAIT);
         xfs_qm_sync(mp, SYNC_WAIT);
 
+        /* drop inode references pinned by filestreams */
+        xfs_filestream_flush(mp);
+
         /* write superblock and hoover up shutdown errors */
-        error = xfs_sync_fsdata(mp, 0);
+        error = xfs_sync_fsdata(mp, SYNC_WAIT);
 
         /* flush data-only devices */
         if (mp->m_rtdev_targp)
@@ -570,8 +588,6 @@ xfs_sync_worker(
                 /* dgc: errors ignored here */
                 error = xfs_qm_sync(mp, SYNC_TRYLOCK);
                 error = xfs_sync_fsdata(mp, SYNC_TRYLOCK);
-                if (xfs_log_need_covered(mp))
-                        error = xfs_commit_dummy_trans(mp, XFS_LOG_FORCE);
         }
         mp->m_sync_seq++;
         wake_up(&mp->m_wait_single_sync_task);