Prev: Western Union Office
Next: [PATCHv2] tcp: fix crash in tcp_xmit_retransmit_queue
From: Lubomir Rintel on 19 Jul 2010 07:20
Newly mkfs-ed filesystems from Seventh Edition have last modification
time set to zero, but are otherwise perfectly valid.

Also, tighten up other sanity checks to filter out most filesystems with
different bytesex than we're using.

Cc: Christoph Hellwig <hch(a)lst.de>
Signed-off-by: Lubomir Rintel <lkundrak(a)v3.sk>
---
fs/sysv/super.c | 8 +++++---
include/linux/sysv_fs.h | 11 +++++++++++
2 files changed, 16 insertions(+), 3 deletions(-)

diff --git a/fs/sysv/super.c b/fs/sysv/super.c
index bd8d141..17ac83d 100644
--- a/fs/sysv/super.c
+++ b/fs/sysv/super.c
@@ -470,8 +470,8 @@ static int v7_fill_super(struct super_block *sb, void *data, int silent)
v7sb = (struct v7_super_block *) bh->b_data;
if (fs16_to_cpu(sbi, v7sb->s_nfree) > V7_NICFREE ||
fs16_to_cpu(sbi, v7sb->s_ninode) > V7_NICINOD ||
- fs32_to_cpu(sbi, v7sb->s_time) == 0)
- goto failed;
+ fs32_to_cpu(sbi, v7sb->s_fsize) > V7_MAXSIZE)
+ return 0;

/* plausibility check on root inode: it is a directory,
with a nonzero size that is a multiple of 16 */
@@ -480,7 +480,9 @@ static int v7_fill_super(struct super_block *sb, void *data, int silent)
v7i = (struct sysv_inode *)(bh2->b_data + 64);
if ((fs16_to_cpu(sbi, v7i->i_mode) & ~0777) != S_IFDIR ||
(fs32_to_cpu(sbi, v7i->i_size) == 0) ||
- (fs32_to_cpu(sbi, v7i->i_size) & 017) != 0)
+ (fs32_to_cpu(sbi, v7i->i_size) & 017)
+ (fs32_to_cpu(sbi, v7i->i_size) > V7_NFILES *
+ sizeof (struct sysv_dir_entry))) {
goto failed;
brelse(bh2);
bh2 = NULL;
diff --git a/include/linux/sysv_fs.h b/include/linux/sysv_fs.h
index 9641130..0a7a232 100644
--- a/include/linux/sysv_fs.h
+++ b/include/linux/sysv_fs.h
@@ -148,6 +148,17 @@ struct v7_super_block {
char s_fname[6]; /* file system name */
char s_fpack[6]; /* file system pack name */
};
+/* Constants to aid sanity checking */
+/* This is not a hard limit, nor enforced by v7 kernel. It's actually just
+ * the limit used by Seventh Edition's ls, though is high enough to assume
+ * that no reasonable file system would have that much entries in root
+ * directory. Thus, if we see anything higher, we just probably got the
+ * endiannes wrong. */
+#define V7_NFILES 1024
+/* Indirect blocks hold just three-byte addresses, therefore if see a file
+ * system whose length has the most significant byte non-zero something is
+ * most likely wrong (not a filesystem, bad bytesex). */
+#define V7_MAXSIZE 0x00ffffff

/* Coherent super-block data on disk */
#define COH_NICINOD 100 /* number of inode cache entries */
--
1.6.5.2

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
 | 
Pages: 1
Prev: Western Union Office
Next: [PATCHv2] tcp: fix crash in tcp_xmit_retransmit_queue