Re: [PATCH] Proposed module init race fix.

[Roman Zippel]

Hi,

Rusty Russell wrote:

> diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .17886-linux-2.5-bk/drivers/block/genhd.c .17886-linux-2.5-bk.updated/drivers/block/genhd.c
> --- .17886-linux-2.5-bk/drivers/block/genhd.c   2003-01-13 16:56:23.000000000 +1100
> +++ .17886-linux-2.5-bk.updated/drivers/block/genhd.c   2003-01-13 22:58:07.000000000 +1100
> @@ -104,10 +104,13 @@ static int exact_lock(dev_t dev, void *d
>   * @gp: per-device partitioning information
>   *
>   * This function registers the partitioning information in @gp
> - * with the kernel.
> + * with the kernel.  Your init function MUST NOT FAIL after this.
>   */
>  void add_disk(struct gendisk *disk)
>  {
> +       /* It needs to be accessible so we can read partitions. */
> +       make_module_live(disk->fops->owner);
> +

After this the module can be removed without problems.

>         disk->flags |= GENHD_FL_UP;
>         blk_register_region(MKDEV(disk->major, disk->first_minor), disk->minors,
>                         NULL, exact_match, exact_lock, disk);

blk_register_region() allocates memory, which can fail?

bye, Roman