Skip to contentSkip to conversation

don't divide by 0 when trying to mount ext3

PATCH In review Andries Brouwer 3 participants 4 messages Nov 8, 2004

lore
Share & tools
lore

Andries Brouwer

author
#0 lore raw

PATCH 1/1 don't divide by 0 when trying to mount ext3

Not surprisingly, the ext3 code crashes in the same way the ext2 code does when dividing by zero.

fs/ext3/super.c +18 -12
diff -uprN -X /linux/dontdiff a/fs/ext3/super.c b/fs/ext3/super.c
@@ -1259,13 +1259,8 @@ static int ext3_fill_super (struct super
 	es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
 	sbi->s_es = es;
 	sb->s_magic = le16_to_cpu(es->s_magic);
-	if (sb->s_magic != EXT3_SUPER_MAGIC) {
-		if (!silent)
-			printk(KERN_ERR 
-			       "VFS: Can't find ext3 filesystem on dev %s.\n",
-			       sb->s_id);
-		goto failed_mount;
-	}
+	if (sb->s_magic != EXT3_SUPER_MAGIC)
+		goto cantfind_ext3;
 
 	/* Set defaults before we parse the mount options */
 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);

@@ -1397,8 +1392,13 @@ static int ext3_fill_super (struct super
 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 	sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
+	if (EXT3_INODE_SIZE(sb) == 0)
+		goto cantfind_ext3;
 	sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
-	sbi->s_itb_per_group = sbi->s_inodes_per_group /sbi->s_inodes_per_block;
+	if (sbi->s_inodes_per_block == 0)
+		goto cantfind_ext3;
+	sbi->s_itb_per_group = sbi->s_inodes_per_group /
+					sbi->s_inodes_per_block;
 	sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
 	sbi->s_sbh = bh;
 	sbi->s_mount_state = le16_to_cpu(es->s_state);

@@ -1427,6 +1427,8 @@ static int ext3_fill_super (struct super
 		goto failed_mount;
 	}
 
+	if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
+		goto cantfind_ext3;
 	sbi->s_groups_count = (le32_to_cpu(es->s_blocks_count) -
 			       le32_to_cpu(es->s_first_data_block) +
 			       EXT3_BLOCKS_PER_GROUP(sb) - 1) /

@@ -1579,6 +1581,12 @@ static int ext3_fill_super (struct super
 
 	return 0;
 
+cantfind_ext3:
+	if (!silent)
+		printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
+		       sb->s_id);
+	goto failed_mount;
+
 failed_mount3:
 	journal_destroy(sbi->s_journal);
 failed_mount2:

@@ -1588,10 +1596,8 @@ failed_mount2:
 	kfree(sbi->s_group_desc);
 failed_mount:
 #ifdef CONFIG_QUOTA
-	for (i = 0; i < MAXQUOTAS; i++) {
-		if (sbi->s_qf_names[i])
-			kfree(sbi->s_qf_names[i]);
-	}
+	for (i = 0; i < MAXQUOTAS; i++)
+		kfree(sbi->s_qf_names[i]);
 #endif
 	ext3_blkdev_remove(sbi);
 	brelse(bh);
#1 lore raw
Quoted 6 lines · Andries Brouwer

Not surprisingly, the ext3 code crashes in the same way the ext2 code does when dividing by zero.

fs/ext3/super.c +2 -0
+	if (sb->s_magic != EXT3_SUPER_MAGIC)
+		goto cantfind_ext3;

[...]

Quoted 2 lines
(quoted diff) +2 -0
+	if (EXT3_INODE_SIZE(sb) == 0)
+		goto cantfind_ext3;

[...]

Quoted 2 lines
fs/ext3/super.c +2 -0
+	if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
+		goto cantfind_ext3;

[...]

fs/ext3/super.c static int ext3_fill_super (struct super · line 1588

Quoted 5 lines
fs/ext3/super.c +5 -0
+cantfind_ext3:
+	if (!silent)
+		printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
+		       sb->s_id);
+	goto failed_mount;

There are now three cases that end up with the same message and same error from userspace viewpoint. There are many cases where debugging a problem is helped when it's possible to find out exactly which test determined that the filesystem could not be mounted.

How about:

	[ ... ]	{
	errstr = "no magic";
	goto cantfind_ext3;
	}
Signature · 5 lines
** [email protected] ** http://www.BitWizard.nl/ ** +31-15-2600998 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
Q: It doesn't work. A: Look buddy, doesn't work is an ambiguous statement. 
Does it sit on the couch all day? Is it unemployed? Please be specific! 
Define 'it' and what it isn't doing. --------- Adapted from lxrbot FAQ
#2 lore raw
Quoted 4 lines · Rogier Wolff

There are now three cases that end up with the same message and same error from userspace viewpoint. There are many cases where debugging a problem is helped when it's possible to find out exactly which test determined that the filesystem could not be mounted.

Strings are expensive. Don't like to add worthless code. We lived without this for years, so it is not a frequent occurrence. If you have a bad ext2/ext3 system, e2fsck will find what is wrong.

#3 lore raw
Quoted 10 lines · Andries Brouwer
Quoted 4 lines · Rogier Wolff

There are now three cases that end up with the same message and same error from userspace viewpoint. There are many cases where debugging a problem is helped when it's possible to find out exactly which test determined that the filesystem could not be mounted.

Strings are expensive. Don't like to add worthless code. We lived without this for years, so it is not a frequent occurrence. If you have a bad ext2/ext3 system, e2fsck will find what is wrong.

   int cpos=0;

if (++cpos && (errorposibility1 ) ) goto error_handling;
[...]
if (++cpos && (errorposibility2 ) ) goto error_handling;
[...]
if (++cpos && (errorposibility3 ) ) goto error_handling;
[...]
error_handling:

     printk (".... %d ...", cpos);

Rogier.

Signature · 5 lines
** [email protected] ** http://www.BitWizard.nl/ ** +31-15-2600998 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
Q: It doesn't work. A: Look buddy, doesn't work is an ambiguous statement. 
Does it sit on the couch all day? Is it unemployed? Please be specific! 
Define 'it' and what it isn't doing. --------- Adapted from lxrbot FAQ