testing the /dev/cciss/c0d0 device names

testing the /dev/cciss/c0d0 device names

[Loic Dachary]

[-- Attachment #1: Type: text/plain, Size: 786 bytes --]

Hi Ilya,

ceph-disk has special handling for device names like /dev/cciss/c0d1 [1] and it was partially broken when support for device mapper was introduced. Ideally there would be a way to test that support when running the ceph-disk suite [2]. Do you know of a way to do that without having the hardware for which this driver is designed ? 

Maybe this convention (/dev/cciss/c0d0 being mapped to /sys/block/cciss!c0d0 is not unique to this driver and I could use another to validate the name conversion from X/Y to X!Y and vice versa is handled as it should ?

Cheers

[1] https://github.com/ceph/ceph/blob/infernalis/src/ceph-disk#L438
[2] https://github.com/ceph/ceph/blob/infernalis/qa/workunits/ceph-disk/ceph-disk-test.py
-- 
Loïc Dachary, Artisan Logiciel Libre


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: testing the /dev/cciss/c0d0 device names

[Ilya Dryomov]

On Sat, Dec 5, 2015 at 7:36 PM, Loic Dachary <loic@dachary.org> wrote:
> Hi Ilya,
>
> ceph-disk has special handling for device names like /dev/cciss/c0d1 [1] and it was partially broken when support for device mapper was introduced. Ideally there would be a way to test that support when running the ceph-disk suite [2]. Do you know of a way to do that without having the hardware for which this driver is designed ?
>
> Maybe this convention (/dev/cciss/c0d0 being mapped to /sys/block/cciss!c0d0 is not unique to this driver and I could use another to validate the name conversion from X/Y to X!Y and vice versa is handled as it should ?

No, it's not unique.  driver core does strreplace(s, '/', '!') at
register time to work around such block devices.  The list includes
DAC960, aoeblk, cciss, cpqarray, sx8 and probably more, but I don't
think anything widespread uses this naming scheme.  IIRC dm actually
won't let you name a device with anything that contains a slash.

If you really wanted you could set up aoeblk I guess, but a simple unit
test should be be more than enough ;)

Thanks,

                Ilya

Re: testing the /dev/cciss/c0d0 device names

[Loic Dachary]

[-- Attachment #1: Type: text/plain, Size: 1410 bytes --]


Hi,

On 06/12/2015 20:15, Ilya Dryomov wrote:
> On Sat, Dec 5, 2015 at 7:36 PM, Loic Dachary <loic@dachary.org> wrote:
>> Hi Ilya,
>>
>> ceph-disk has special handling for device names like /dev/cciss/c0d1 [1] and it was partially broken when support for device mapper was introduced. Ideally there would be a way to test that support when running the ceph-disk suite [2]. Do you know of a way to do that without having the hardware for which this driver is designed ?
>>
>> Maybe this convention (/dev/cciss/c0d0 being mapped to /sys/block/cciss!c0d0 is not unique to this driver and I could use another to validate the name conversion from X/Y to X!Y and vice versa is handled as it should ?
> 
> No, it's not unique.  driver core does strreplace(s, '/', '!') at
> register time to work around such block devices.  The list includes
> DAC960, aoeblk, cciss, cpqarray, sx8 and probably more, but I don't
> think anything widespread uses this naming scheme.  IIRC dm actually
> won't let you name a device with anything that contains a slash.
> 
> If you really wanted you could set up aoeblk I guess, but a simple unit
> test should be be more than enough ;)

Thanks for the information. I'll refrain from attempting integration tests and try to make it solid with unit tests :-)

Cheers

> 
> Thanks,
> 
>                 Ilya
> 

-- 
Loïc Dachary, Artisan Logiciel Libre


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]