[parted-devel] "linux: use devicemapper task name instead of device node name" causes dmraid breakage

[Hans de Goede]

Hi,

On 05/17/2010 03:37 PM, Phillip Susi wrote:
> On 5/17/2010 7:37 AM, Colin Watson wrote:
>> IIRC the member of udev upstream who works for us disagrees with this
>> naming scheme; Scott, could you comment?
>
> I'm not sure I like the naming scheme either.  The p suffix on dev nodes
> normally indicates that the minor number differs from the containing dev
> node without the p to indicate it is a partition.  Since the kernel
> device mapper does not currently support partitions, dmraid creates new
> dm devices to map the partitions.  This means that if dm-0 is the raid
> array, then it sounds like with these changes, the partition will be
> created as /dev/dm-0p1

Erm, no the purpose of the patch you are suggesting to revert is exactly
to make sure that the partition will be created as /dev/mapper/foop1 independent
of the lvm2 version. And dependent on the lvm2 version /dev/mapper/foo1 will
either be a device node with the same major and minor as /dev/dm-1, or a symlink
to /dev/dm-1.

> though the kernel name of the device is dm-1.
> How would udev even know it should name the node dm-0p1 if the kernel
> name is dm-1?
>

Because it uses the devicemapper name, as shown for example in "dmsetup table"
output as the name for the symlink under /dev/mapper.

>>> Likewise kpartx (which one should use rather then dmraid's build
>>> in partition code as kpartx can handle gpt labels too and dmraid
>>> cannot), also creates partition mappings with a p in the name
>>> by default.
>
> I would really like to remove the partition code from the kernel and let
> udev handle it with partx.  One thing I've been wondering about is why
> it seems others like kpartx instead of the original partx?
>

Because partx only works on devices for which the kernel supports partitions,
where as kpartx also works on other devices as it uses devicemapper

Regards,

Hans