block: sanitize invalid partition table entries
authorKay Sievers <kay.sievers@vrfy.org>
Thu, 16 Oct 2008 05:04:21 +0000 (22:04 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 16 Oct 2008 18:21:47 +0000 (11:21 -0700)
We currently follow blindly what the partition table lies about the
disk, and let the kernel create block devices which can not be accessed.
Trying to identify the device leads to kernel logs full of:
  sdb: rw=0, want=73392, limit=28800
  attempt to access beyond end of device

Here is an example of a broken partition table, where sda2 starts
behind the end of the disk, and sdb3 is larger than the entire disk:
  Disk /dev/sdb: 14 MB, 14745600 bytes
  1 heads, 29 sectors/track, 993 cylinders, total 28800 sectors
     Device Boot      Start         End      Blocks   Id  System
  /dev/sdb1              29        7800        3886   83  Linux
  /dev/sdb2           37801       45601        3900+  83  Linux
  /dev/sdb3           15602       73402       28900+  83  Linux
  /dev/sdb4           23403       28796        2697   83  Linux

The kernel creates these completely invalid devices, which can not be
accessed, or may lead to other unpredictable failures:
  grep . /sys/class/block/sdb*/{start,size}
  /sys/class/block/sdb/size:28800
  /sys/class/block/sdb1/start:29
  /sys/class/block/sdb1/size:7772
  /sys/class/block/sdb2/start:37801
  /sys/class/block/sdb2/size:7801
  /sys/class/block/sdb3/start:15602
  /sys/class/block/sdb3/size:57801
  /sys/class/block/sdb4/start:23403
  /sys/class/block/sdb4/size:5394

With this patch, we ignore partitions which start behind the end of the disk,
and limit partitions to the end of the disk if they pretend to be larger:
  grep . /sys/class/block/sdb*/{start,size}
  /sys/class/block/sdb/size:28800
  /sys/class/block/sdb1/start:29
  /sys/class/block/sdb1/size:7772
  /sys/class/block/sdb3/start:15602
  /sys/class/block/sdb3/size:13198
  /sys/class/block/sdb4/start:23403
  /sys/class/block/sdb4/size:5394

These warnings are printed to the kernel log:
  sdb: p2 ignored, start 37801 is behind the end of the disk
  sdb: p3 size 57801 limited to end of disk

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Cc: Herton Ronaldo Krzesinski <herton@mandriva.com.br>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/partitions/check.c

index 7408227c49c91cefed99f861a90a8bbb85e8b346..fbeb2f372a93dba06f0d9aab1cfb438389a5668e 100644 (file)
@@ -538,10 +538,23 @@ int rescan_partitions(struct gendisk *disk, struct block_device *bdev)
                sector_t from = state->parts[p].from;
                if (!size)
                        continue;
+               if (from >= get_capacity(disk)) {
+                       printk(KERN_WARNING
+                              "%s: p%d ignored, start %llu is behind the end of the disk\n",
+                              disk->disk_name, p, (unsigned long long) from);
+                       continue;
+               }
                if (from + size > get_capacity(disk)) {
+                       /*
+                        * we can not ignore partitions of broken tables
+                        * created by for example camera firmware, but we
+                        * limit them to the end of the disk to avoid
+                        * creating invalid block devices
+                        */
                        printk(KERN_WARNING
-                               "%s: p%d exceeds device capacity\n",
-                               disk->disk_name, p);
+                              "%s: p%d size %llu limited to end of disk\n",
+                              disk->disk_name, p, (unsigned long long) size);
+                       size = get_capacity(disk) - from;
                }
                res = add_partition(disk, p, from, size, state->parts[p].flags);
                if (res) {