Squashfs: add sanity checks to fragment reading at mount time

Fsfuzzer generates corrupted filesystems which throw a warn_on in
kmalloc.  One of these is due to a corrupted superblock fragments field.
Fix this by checking that the number of bytes to be read (and allocated)
does not extend into the next filesystem structure.

Also add a couple of other sanity checks of the mount-time fragment table
structures.

Signed-off-by: Phillip Lougher <phillip@lougher.demon.co.uk>
This commit is contained in:
Phillip Lougher
2011-05-24 04:33:34 +01:00
parent ac51a0a713
commit 1cac63cc9b
3 changed files with 25 additions and 4 deletions

View File

@@ -261,6 +261,7 @@ allocate_id_index_table:
msblk->inode_lookup_table = NULL;
goto failed_mount;
}
next_table = msblk->inode_lookup_table[0];
sb->s_export_op = &squashfs_export_ops;
@@ -278,7 +279,7 @@ handle_fragments:
/* Allocate and read fragment index table */
msblk->fragment_index = squashfs_read_fragment_index_table(sb,
le64_to_cpu(sblk->fragment_table_start), fragments);
le64_to_cpu(sblk->fragment_table_start), next_table, fragments);
if (IS_ERR(msblk->fragment_index)) {
ERROR("unable to read fragment index table\n");
err = PTR_ERR(msblk->fragment_index);