On Sat, 11 Mar 2000, Roman V. Shaposhnick wrote:
> Hi,
>
> I would like to mention that in 2.3.51 there is a huge problem with
> modules related to recent changes in fs-register-unregister interface.
>
> First let's see the bug itself:
[snip]
> yellow:/usr/src/linux/fs/isofs # mount /dev/hdc /cdrom
> mount: block device /dev/hdc is write-protected, mounting read-only
> yellow:/usr/src/linux/fs/isofs # lsmod
> Module Size Used by
> isofs 19164 3
Oh, fsck! OK, I see what happens. get_fs_type() grabs the reference to
module (as it should). However, we call it in sys_mount() and then in
do_mount(). Reference obtained in the latter is properly handled, no
problems here (either retained or dropped, depending on results of
->read_super()). But reference from call in sys_mount() is a leak.
And since you didn't say -o ro mount(8) called mount(2) twice. Resulting
in reference counter 3: 1 from successful mount + 2 leaked.
Oh, well... Fix being:
--- linux-2.3.51/fs/super.c Sat Mar 11 01:33:11 2000
+++ linux-bird.vfs/fs/super.c Sat Mar 11 06:21:54 2000
@@ -1134,7 +1134,7 @@
dentry = namei(dev_name);
retval = PTR_ERR(dentry);
if (IS_ERR(dentry))
- goto out;
+ goto fs_out;
inode = dentry->d_inode;
retval = -ENOTBLK;
@@ -1163,6 +1163,8 @@
dput_and_out:
dput(dentry);
+fs_out:
+ put_filesystem(fstype);
out:
unlock_kernel();
return retval;
Linus, could you apply it?
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Mar 15 2000 - 21:00:19 EST