Clear the IN_SIZEMOD and IN_IBLKDATA flags only when doing a

synchronous inode update.

The IN_SIZEMOD and IN_IBLKDATA flags indicate changes to the
file size and block pointer fields in the inode. When these
fields have been changed, the fsync() and fsyncdata() system
calls must write the inode to ensure their semantics that the
file is on stable store.

The IN_SIZEMOD and IN_IBLKDATA flags cannot be cleared until
a synchronous write of the inode is done. If they are cleared
on an asynchronous write, then the inode may not yet have been
written to the disk when an fsync() or fsyncdata() call is done.
Absent these flags, these calls would not know that they needed
to write the inode. Thus, these flags only can be cleared on
synchronous writes of the inode. Since the inode will be locked
for the duration of the I/O that writes it to disk, no fsync()
or fsyncdata() will be able to run before the on-disk inode
is complete.

Reviewed by: kib
MFC with: -r361785
Differential revision:  https://reviews.freebsd.org/D25072

1 files changed, 21 insertions, 1 deletions

diff --git a/sys/ufs/ffs/ffs_inode.c b/sys/ufs/ffs/ffs_inode.c
index dbe3c9b8aa16..93836a222e5b 100644
--- a/sys/ufs/ffs/ffs_inode.c
+++ b/sys/ufs/ffs/ffs_inode.c
@@ -94,7 +94,27 @@ ffs_update(vp, waitfor)
 	ip = VTOI(vp);
 	if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
 		return (0);
-	ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED | IN_IBLKDATA);
+	ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED);
+	/*
+	 * The IN_SIZEMOD and IN_IBLKDATA flags indicate changes to the
+	 * file size and block pointer fields in the inode. When these
+	 * fields have been changed, the fsync() and fsyncdata() system 
+	 * calls must write the inode to ensure their semantics that the 
+	 * file is on stable store.
+	 *
+	 * The IN_SIZEMOD and IN_IBLKDATA flags cannot be cleared until
+	 * a synchronous write of the inode is done. If they are cleared
+	 * on an asynchronous write, then the inode may not yet have been
+	 * written to the disk when an fsync() or fsyncdata() call is done.
+	 * Absent these flags, these calls would not know that they needed
+	 * to write the inode. Thus, these flags only can be cleared on
+	 * synchronous writes of the inode. Since the inode will be locked
+	 * for the duration of the I/O that writes it to disk, no fsync()
+	 * or fsyncdata() will be able to run before the on-disk inode
+	 * is complete.
+	 */
+	if (waitfor)
+		ip->i_flag &= ~(IN_SIZEMOD | IN_IBLKDATA);
 	fs = ITOFS(ip);
 	if (fs->fs_ronly && ITOUMP(ip)->um_fsckpid == 0)
 		return (0);