[XFS] Start writeout earlier (on last close) in the case where we have a

truncate down followed by delayed allocation (buffered writes) - worst
case scenario for the notorious NULL files problem.  This reduces the
window where we are exposed to that problem significantly.

SGI-PV: 917976
SGI-Modid: xfs-linux-melb:xfs-kern:26100a

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

1 files changed, 52 insertions, 36 deletions

diff --git a/fs/xfs/xfs_vnodeops.c b/fs/xfs/xfs_vnodeops.c
index cb36a56392e7..35906bae92e1 100644
--- a/fs/xfs/xfs_vnodeops.c
+++ b/fs/xfs/xfs_vnodeops.c
@@ -16,8 +16,6 @@
  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  */
 
-#include <linux/capability.h>
-
 #include "xfs.h"
 #include "xfs_fs.h"
 #include "xfs_types.h"
@@ -58,32 +56,14 @@
 #include "xfs_log_priv.h"
 #include "xfs_mac.h"
 
-
-/*
- * The maximum pathlen is 1024 bytes. Since the minimum file system
- * blocksize is 512 bytes, we can get a max of 2 extents back from
- * bmapi.
- */
-#define SYMLINK_MAPS 2
-
-/*
- * For xfs, we check that the file isn't too big to be opened by this kernel.
- * No other open action is required for regular files.  Devices are handled
- * through the specfs file system, pipes through fifofs.  Device and
- * fifo vnodes are "wrapped" by specfs and fifofs vnodes, respectively,
- * when a new vnode is first looked up or created.
- */
 STATIC int
 xfs_open(
         bhv_desc_t      *bdp,
         cred_t          *credp)
 {
         int             mode;
-        vnode_t         *vp;
-        xfs_inode_t     *ip;
-
-        vp = BHV_TO_VNODE(bdp);
-        ip = XFS_BHVTOI(bdp);
+        vnode_t         *vp = BHV_TO_VNODE(bdp);
+        xfs_inode_t     *ip = XFS_BHVTOI(bdp);
 
         if (XFS_FORCED_SHUTDOWN(ip->i_mount))
                 return XFS_ERROR(EIO);
@@ -101,6 +81,36 @@ xfs_open(
         return 0;
 }
 
+STATIC int
+xfs_close(
+        bhv_desc_t      *bdp,
+        int             flags,
+        lastclose_t     lastclose,
+        cred_t          *credp)
+{
+        vnode_t         *vp = BHV_TO_VNODE(bdp);
+        xfs_inode_t     *ip = XFS_BHVTOI(bdp);
+        int             error = 0;
+
+        if (XFS_FORCED_SHUTDOWN(ip->i_mount))
+                return XFS_ERROR(EIO);
+
+        if (lastclose != L_TRUE || !VN_ISREG(vp))
+                return 0;
+
+        /*
+         * If we previously truncated this file and removed old data in
+         * the process, we want to initiate "early" writeout on the last
+         * close.  This is an attempt to combat the notorious NULL files
+         * problem which is particularly noticable from a truncate down,
+         * buffered (re-)write (delalloc), followed by a crash.  What we
+         * are effectively doing here is significantly reducing the time
+         * window where we'd otherwise be exposed to that problem.
+         */
+        if (VUNTRUNCATE(vp) && VN_DIRTY(vp) && ip->i_delayed_blks > 0)
+                VOP_FLUSH_PAGES(vp, 0, -1, XFS_B_ASYNC, FI_NONE, error);
+        return error;
+}
 
 /*
  * xfs_getattr
@@ -665,9 +675,17 @@ xfs_setattr(
                                             ((ip->i_d.di_nlink != 0 ||
                                               !(mp->m_flags & XFS_MOUNT_WSYNC))
                                              ? 1 : 0));
-                        if (code) {
+                        if (code)
                                 goto abort_return;
-                        }
+                        /*
+                         * Truncated "down", so we're removing references
+                         * to old data here - if we now delay flushing for
+                         * a long time, we expose ourselves unduly to the
+                         * notorious NULL files problem.  So, we mark this
+                         * vnode and flush it when the file is closed, and
+                         * do not wait the usual (long) time for writeout.
+                         */
+                        VTRUNCATE(vp);
                 }
                 /*
                  * Have to do this even if the file's size doesn't change.
@@ -937,6 +955,13 @@ xfs_access(
 
 
 /*
+ * The maximum pathlen is 1024 bytes. Since the minimum file system
+ * blocksize is 512 bytes, we can get a max of 2 extents back from
+ * bmapi.
+ */
+#define SYMLINK_MAPS 2
+
+/*
  * xfs_readlink
  *
  */
@@ -1470,9 +1495,6 @@ xfs_inactive_symlink_local(
         return 0;
 }
 
-/*
- *
- */
 STATIC int
 xfs_inactive_attrs(
         xfs_inode_t     *ip,
@@ -1531,10 +1553,10 @@ xfs_release(
 
         vp = BHV_TO_VNODE(bdp);
         ip = XFS_BHVTOI(bdp);
+        mp = ip->i_mount;
 
-        if (!VN_ISREG(vp) || (ip->i_d.di_mode == 0)) {
+        if (!VN_ISREG(vp) || (ip->i_d.di_mode == 0))
                 return 0;
-        }
 
         /* If this is a read-only mount, don't do this (would generate I/O) */
         if (vp->v_vfsp->vfs_flag & VFS_RDONLY)
@@ -1546,8 +1568,6 @@ xfs_release(
                 return 0;
 #endif
 
-        mp = ip->i_mount;
-
         if (ip->i_d.di_nlink != 0) {
                 if ((((ip->i_d.di_mode & S_IFMT) == S_IFREG) &&
                      ((ip->i_d.di_size > 0) || (VN_CACHED(vp) > 0 ||
@@ -3745,7 +3765,6 @@ xfs_inode_flush(
         return error;
 }
 
-
 int
 xfs_set_dmattrs (
         bhv_desc_t      *bdp,
@@ -3786,10 +3805,6 @@ xfs_set_dmattrs (
         return error;
 }
 
-
-/*
- * xfs_reclaim
- */
 STATIC int
 xfs_reclaim(
         bhv_desc_t      *bdp)
@@ -4645,6 +4660,7 @@ xfs_change_file_space(
 vnodeops_t xfs_vnodeops = {
         BHV_IDENTITY_INIT(VN_BHV_XFS,VNODE_POSITION_XFS),
         .vop_open               = xfs_open,
+        .vop_close              = xfs_close,
         .vop_read               = xfs_read,
 #ifdef HAVE_SENDFILE
         .vop_sendfile           = xfs_sendfile,