ext3: avoid printk floods in the face of directory corruption
authorEric Sandeen <sandeen@redhat.com>
Sun, 19 Oct 2008 03:28:00 +0000 (20:28 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Mon, 20 Oct 2008 15:52:38 +0000 (08:52 -0700)
A very large directory with many read failures (either due to storage
problems, or due to invalid size & blocks from corruption) will generate a
printk storm as the filesystem continues to try to read all the blocks.
This flood of messages can tie up the box until it is complete - which may
be a very long time, especially for very large corrupted values.

This is fixed by only reporting the corruption once each time we try to
read the directory.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Eugene Teo <eugeneteo@kernel.sg>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/ext3/dir.c

index 28b681ef47e8b9686630edc002283b2056e9fa12..4c82531ea0a8b0d13ce61b63eb7daeecbe5d1423 100644 (file)
@@ -102,6 +102,7 @@ static int ext3_readdir(struct file * filp,
        int err;
        struct inode *inode = filp->f_path.dentry->d_inode;
        int ret = 0;
+       int dir_has_error = 0;
 
        sb = inode->i_sb;
 
@@ -148,9 +149,12 @@ static int ext3_readdir(struct file * filp,
                 * of recovering data when there's a bad sector
                 */
                if (!bh) {
-                       ext3_error (sb, "ext3_readdir",
-                               "directory #%lu contains a hole at offset %lu",
-                               inode->i_ino, (unsigned long)filp->f_pos);
+                       if (!dir_has_error) {
+                               ext3_error(sb, __func__, "directory #%lu "
+                                       "contains a hole at offset %lld",
+                                       inode->i_ino, filp->f_pos);
+                               dir_has_error = 1;
+                       }
                        /* corrupt size?  Maybe no more blocks to read */
                        if (filp->f_pos > inode->i_blocks << 9)
                                break;