On Sat 26-03-22 15:50:46, Christophe JAILLET wrote:
> ida_alloc_range(..., min, max, ...) returns values from min to max,
> inclusive.
>
> So, NR_EXT_DEVT is a valid idx returned by blk_alloc_ext_minor().
>
> This is an issue because in device_add_disk(), this value is used in:
> ddev->devt = MKDEV(disk->major, disk->first_minor);
> and NR_EXT_DEVT is '(1 << MINORBITS)'.
>
> So, should 'disk->first_minor' be NR_EXT_DEVT, it would overflow.
>
> Fixes: 22ae8ce8b892 ("block: simplify bdev/disk lookup in blkdev_get")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@xxxxxxxxxx>
Indeed. The patch looks good to me so feel free to add:
Reviewed-by: Jan Kara <jack@xxxxxxx>
> ---
> #define MKDEV(ma,mi) (((ma) << MINORBITS) | (mi))
>
> This patch is completely speculative, but it seems that idr_alloc() and
> ida_alloc_range() don't have the same semantic regarding the upper bound.
> idr_alloc() looks exclusive, while ida_alloc_range() is inclusive.
>
> We changed from the first one to the other one in the commit in Fixes:.
Yes, this difference is really a landmine. Matthew, why is the semantics of
max parameter for idr_alloc() different from ida_alloc_range() or say
idr_alloc_u32()? It is really easy to introduce subtle bugs with this...
Honza
> ---
> block/genhd.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/block/genhd.c b/block/genhd.c
> index c9a4fc90d3e9..b8b6759d670f 100644
> --- a/block/genhd.c
> +++ b/block/genhd.c
> @@ -335,7 +335,7 @@ int blk_alloc_ext_minor(void)
> {
> int idx;
>
> - idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
> + idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT - 1, GFP_KERNEL);
> if (idx == -ENOSPC)
> return -EBUSY;
> return idx;
> --
> 2.32.0
>
--
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR
