Hi Al, list,
I think the latest blockdev (maybe the do_open) changes broke the raid
subsystem. In order to 'activate' a raid device, the userland tools open
the device node (e.g. /dev/md0) to perform ioctls against it, even though
that device isn't up and running yet. In 2.5.43 it returns ENXIO.
In 2.5.42 we used to get through to bdev->bd_op->open() without a
'gendisk' structure, but now we bail with ENXIO. Is this the pertinent
difference?
David
--
/==============================\
| David Mansfield |
| [email protected] |
\==============================/
On Wed, 16 Oct 2002, David Mansfield wrote:
>
> Hi Al, list,
>
> I think the latest blockdev (maybe the do_open) changes broke the raid
> subsystem. In order to 'activate' a raid device, the userland tools open
> the device node (e.g. /dev/md0) to perform ioctls against it, even though
> that device isn't up and running yet. In 2.5.43 it returns ENXIO.
I know, patches that handle that are coming to Linus in an hour.
... going net.dead for two days didn't help any ;-/