Skip to contentSkip to conversation

fix for initrd panic in 2.4.5

PATCH In review James Bottomley 2 participants 2 messages Jun 6, 2001

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fix for initrd panic in 2.4.5

Not many people have seen this, but the i_bdev field in fake_inode in ioctl_by_bdev() is uninitialised. Since this field is dereferenced by BLKFLSBUF in rd_ioctl, it can lead to a panic (depending on what happens to be on the stack). The attached fixes the problem and clears all the fields in fake_inode to make any other problems like this show up.

James Bottomley

Index: fs/block_dev.c
===================================================================
RCS file: /home/jejb/CVSROOT/linux/2.4/fs/block_dev.c,v retrieving revision 1.1.1.10

block_dev.c +2 -1
diff -u -r1.1.1.10 block_dev.c
@@ -596,13 +596,14 @@
 int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg)
 {
 	kdev_t rdev = to_kdev_t(bdev->bd_dev);
-	struct inode inode_fake;
+	struct inode inode_fake = { 0 };
 	int res;
 	mm_segment_t old_fs = get_fs();
 
 	if (!bdev->bd_op->ioctl)
 		return -EINVAL;
 	inode_fake.i_rdev=rdev;
+	inode_fake.i_bdev = bdev;
 	init_waitqueue_head(&inode_fake.i_wait);
 	set_fs(KERNEL_DS);
 	res = bdev->bd_op->ioctl(&inode_fake, NULL, cmd, arg);
#1 lore raw
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"james" == James Bottomley <[email protected]> writes:

james> Not many people have seen this, but the i_bdev field in fake_inode in james> ioctl_by_bdev() is uninitialised. Since this field is dereferenced by james> BLKFLSBUF in rd_ioctl, it can lead to a panic (depending on what happens to be james> on the stack). The attached fixes the problem and clears all the fields in james> fake_inode to make any other problems like this show up.

see the fix in ac kernels, you have a bdev there to put in the fake_inode.

Later, Juan.

james> Index: fs/block_dev.c
james> =================================================================== james> RCS file: /home/jejb/CVSROOT/linux/2.4/fs/block_dev.c,v
james> retrieving revision 1.1.1.10
james> diff -u -r1.1.1.10 block_dev.c
james> --- fs/block_dev.c 2001/05/26 15:33:37 1.1.1.10
james> +++ fs/block_dev.c 2001/06/02 13:14:35
james> @@ -596,13 +596,14 @@
james> int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg) james> {
james> kdev_t rdev = to_kdev_t(bdev->bd_dev);
james> - struct inode inode_fake;
james> + struct inode inode_fake = { 0 };
james> int res;
james> mm_segment_t old_fs = get_fs();

james> if (!bdev->bd_op->ioctl)
james> return -EINVAL;
james> inode_fake.i_rdev=rdev;
james> + inode_fake.i_bdev = bdev;
james> init_waitqueue_head(&inode_fake.i_wait); james> set_fs(KERNEL_DS);
james> res = bdev->bd_op->ioctl(&inode_fake, NULL, cmd, arg);

Signature · 2 lines
In theory, practice and theory are the same, but in practice they 
are different -- Larry McVoy