[dm-devel] confusion about multipath_prepare_ioctl

[Mike Snitzer]

On Mon, Mar 05 2018 at 10:35pm -0500,
Wang Sheng-Hui <shhuiw at foxmail.com> wrote:

> Dear,
> 
> Sorry to trouble you.
> 
> I noticed some code in dm-*.c like:
> "
> static int multipath_prepare_ioctl(struct dm_target *ti,
> 		struct block_device **bdev, fmode_t *mode)
> {
> ...
> 	/*
> 	 * Only pass ioctls through if the device sizes match exactly.
> 	 */
> 	if (!r && ti->len != i_size_read((*bdev)->bd_inode) >> SECTOR_SHIFT)
> 		return 1;
> ...
> }
> "
> Here, return value 1 means 
> "ioctl is being issued against a subset of the parent bdev; require extra privileges."
> (comment in dm_blk_ioctl)
> 
> I'm confused by the comment and '!=' test for multipath. 
> In which cases, the size of low level single device is not equal to the parent 
> size of multipath device?

Given that ti->len is sent down from userspace, the DM multipath
target's ti->len _could_ be smaller than the underlying path(s).  But in
practice that doesn't occur with multipathd.. a partitioned multipath
device is generally done, via kpartx, in terms of linear mappings ontop
of the multipath device.

The same != test is done in the dm linear target and is much more
relevant to concerns about ioctls being sent to partition.

Mike