ext4: reject too-large filesystems on 32-bit kernels
authorEric Sandeen <sandeen@redhat.com>
Tue, 18 Aug 2009 03:48:51 +0000 (23:48 -0400)
committerTheodore Ts'o <tytso@mit.edu>
Tue, 18 Aug 2009 03:48:51 +0000 (23:48 -0400)
ext4 will happily mount a > 16T filesystem on a 32-bit box, but
this is not safe; writes to the block device will wrap past 16T
and the page cache can't index past 16T (232 index * 4k pages).

Adding another test to the existing "too many sectors" test
should do the trick.

Add a comment, a relevant return value, and fix the reference
to the CONFIG_LBD(AF) option as well.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
fs/ext4/super.c

index fe3f376b7df24482dec5e38b3337ce8ec761070f..de67c3657b83fa0f013d5f4d6401906c9f2c5ec6 100644 (file)
@@ -2550,12 +2550,19 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
                goto failed_mount;
        }
 
-       if (ext4_blocks_count(es) >
-                   (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
+       /*
+        * Test whether we have more sectors than will fit in sector_t,
+        * and whether the max offset is addressable by the page cache.
+        */
+       if ((ext4_blocks_count(es) >
+            (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
+           (ext4_blocks_count(es) >
+            (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
                ext4_msg(sb, KERN_ERR, "filesystem"
-                       " too large to mount safely");
+                        " too large to mount safely on this system");
                if (sizeof(sector_t) < 8)
                        ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
+               ret = -EFBIG;
                goto failed_mount;
        }