The UFS2 filesystem allocates new blocks of inodes as they are needed.

When a cylinder group runs short of inodes, a new block for inodes is
allocated, zero'ed, and written to the disk. The zero'ed inodes must
be on the disk before the cylinder group can be updated to claim them.
If the cylinder group claiming the new inodes were written before the
zero'ed block of inodes, the system could crash with the filesystem in
an unrecoverable state.

Rather than adding a soft updates dependency to ensure that the new
inode block is written before it is claimed by the cylinder group
map, we just do a barrier write of the zero'ed inode block to ensure
that it will get written before the updated cylinder group map can
be written. This change should only slow down bulk loading of newly
created filesystems since that is the primary time that new inode
blocks need to be created.

Reported by: Robert Watson
Reviewed by: kib
Tested by:   Peter Holm

1 files changed, 10 insertions, 3 deletions

diff --git a/sys/ufs/ffs/ffs_alloc.c b/sys/ufs/ffs/ffs_alloc.c
index d462cbb..abe4073 100644
--- a/sys/ufs/ffs/ffs_alloc.c
+++ b/sys/ufs/ffs/ffs_alloc.c
@@ -1861,7 +1861,6 @@ gotit:
 	/*
 	 * Check to see if we need to initialize more inodes.
 	 */
-	ibp = NULL;
 	if (fs->fs_magic == FS_UFS2_MAGIC &&
 	    ipref + INOPB(fs) > cgp->cg_initediblk &&
 	    cgp->cg_initediblk < cgp->cg_niblk) {
@@ -1874,6 +1873,16 @@ gotit:
 			dp2->di_gen = arc4random() / 2 + 1;
 			dp2++;
 		}
+		/*
+		 * Rather than adding a soft updates dependency to ensure
+		 * that the new inode block is written before it is claimed
+		 * by the cylinder group map, we just do a barrier write
+		 * here. The barrier write will ensure that the inode block
+		 * gets written before the updated cylinder group map can be
+		 * written. The barrier write should only slow down bulk
+		 * loading of newly created filesystems.
+		 */
+		babarrierwrite(ibp);
 		cgp->cg_initediblk += INOPB(fs);
 	}
 	UFS_LOCK(ump);
@@ -1892,8 +1901,6 @@ gotit:
 	if (DOINGSOFTDEP(ITOV(ip)))
 		softdep_setup_inomapdep(bp, ip, cg * fs->fs_ipg + ipref, mode);
 	bdwrite(bp);
-	if (ibp != NULL)
-		bawrite(ibp);
 	return ((ino_t)(cg * fs->fs_ipg + ipref));
 }