Request for proper hardware dependencies

Request for proper hardware dependencies

[Uwe Bonnes]

Hallo,

the number of kernel driver exploded in the last years and so the kernel
configuration. Distribution mantainers must at some point limit compiled
driver at least partly to the driver needed one on a given arch. However the
distribution maintainers mosty can't always know what they deal about.

For me this resulted in loosing socketcan from the Opensuse Kernels in
between 12.3 up to some weeks ago, as one maintainer (not Jean) matched
socketcan in error with "SPI" which is not related to normal desktop
machines. See
http://lists.opensuse.org/archive/opensuse-bugs/2014-01/msg00970.html.
Opensuse 13.1 recent kernels and the next release will have socketcan again,
but in the course of the discussion Jean Delvare pointed some weak points in
socketcan configuration, leading to compilation on driver not needed on some
archs:

>>>>> "Jean" == Jean Delvare <jdelvare@suse.com> writes:

    Jean> For the record, hardware dependencies on CAN device drivers aren't
    Jean> that bad in general, I've seen worse in several other subsystems.

    Jean> That being said, there is still some room for improvement. For
    Jean> example, CAN_AT91 could depend on ARCH_AT91 instead of just ARM,
    Jean> PCH_CAN could depend on X86, and a few drivers with "ISA" in their
    Jean> description should probably depend on ISA. I'm also not sure why
    Jean> CAN_MSCAN alternatively depends on M68K instead of just
    Jean> ARM. That's just from a quick look from the non-initiate I am,
    Jean> there may be more dependency fix-ups possible.  I'll send a few
    Jean> patches out if I can find some time for that today.

Please consider openly his patches or even better consider a cleanup yourself.

Bye
-- 
Uwe Bonnes                bon@elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------

Re: Request for proper hardware dependencies

[Marc Kleine-Budde]

[-- Attachment #1: Type: text/plain, Size: 2186 bytes --]

On 05/06/2014 11:13 AM, Uwe Bonnes wrote:
> Hallo,
> 
> the number of kernel driver exploded in the last years and so the kernel
> configuration. Distribution mantainers must at some point limit compiled
> driver at least partly to the driver needed one on a given arch. However the
> distribution maintainers mosty can't always know what they deal about.
> 
> For me this resulted in loosing socketcan from the Opensuse Kernels in
> between 12.3 up to some weeks ago, as one maintainer (not Jean) matched
> socketcan in error with "SPI" which is not related to normal desktop
> machines. See
> http://lists.opensuse.org/archive/opensuse-bugs/2014-01/msg00970.html.
> Opensuse 13.1 recent kernels and the next release will have socketcan again,
> but in the course of the discussion Jean Delvare pointed some weak points in
> socketcan configuration, leading to compilation on driver not needed on some
> archs:
> 
>>>>>> "Jean" == Jean Delvare <jdelvare@suse.com> writes:
> 
>     Jean> For the record, hardware dependencies on CAN device drivers aren't
>     Jean> that bad in general, I've seen worse in several other subsystems.
> 
>     Jean> That being said, there is still some room for improvement. For
>     Jean> example, CAN_AT91 could depend on ARCH_AT91 instead of just ARM,
>     Jean> PCH_CAN could depend on X86, and a few drivers with "ISA" in their
>     Jean> description should probably depend on ISA. I'm also not sure why
>     Jean> CAN_MSCAN alternatively depends on M68K instead of just
>     Jean> ARM. That's just from a quick look from the non-initiate I am,
>     Jean> there may be more dependency fix-ups possible.  I'll send a few
>     Jean> patches out if I can find some time for that today.
> 
> Please consider openly his patches or even better consider a cleanup yourself.

I've applied all 3 patches by Jean to can-next.

Tnx,
Marc

-- 
Pengutronix e.K.                  | Marc Kleine-Budde           |
Industrial Linux Solutions        | Phone: +49-231-2826-924     |
Vertretung West/Dortmund          | Fax:   +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686  | http://www.pengutronix.de   |


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 242 bytes --]

Re: Request for proper hardware dependencies

[Jean Delvare]

Le Tuesday 13 May 2014 à 16:37 +0200, Marc Kleine-Budde a écrit :
> On 05/06/2014 11:13 AM, Uwe Bonnes wrote:
> > Hallo,
> > 
> > the number of kernel driver exploded in the last years and so the kernel
> > configuration. Distribution mantainers must at some point limit compiled
> > driver at least partly to the driver needed one on a given arch. However the
> > distribution maintainers mosty can't always know what they deal about.
> > 
> > For me this resulted in loosing socketcan from the Opensuse Kernels in
> > between 12.3 up to some weeks ago, as one maintainer (not Jean) matched
> > socketcan in error with "SPI" which is not related to normal desktop
> > machines. See
> > http://lists.opensuse.org/archive/opensuse-bugs/2014-01/msg00970.html.
> > Opensuse 13.1 recent kernels and the next release will have socketcan again,
> > but in the course of the discussion Jean Delvare pointed some weak points in
> > socketcan configuration, leading to compilation on driver not needed on some
> > archs:
> > 
> >>>>>> "Jean" == Jean Delvare <jdelvare@suse.com> writes:
> > 
> >     Jean> For the record, hardware dependencies on CAN device drivers aren't
> >     Jean> that bad in general, I've seen worse in several other subsystems.
> > 
> >     Jean> That being said, there is still some room for improvement. For
> >     Jean> example, CAN_AT91 could depend on ARCH_AT91 instead of just ARM,
> >     Jean> PCH_CAN could depend on X86, and a few drivers with "ISA" in their
> >     Jean> description should probably depend on ISA. I'm also not sure why
> >     Jean> CAN_MSCAN alternatively depends on M68K instead of just
> >     Jean> ARM. That's just from a quick look from the non-initiate I am,
> >     Jean> there may be more dependency fix-ups possible.  I'll send a few
> >     Jean> patches out if I can find some time for that today.
> > 
> > Please consider openly his patches or even better consider a cleanup yourself.
> 
> I've applied all 3 patches by Jean to can-next.

Great, thank you!

-- 
Jean Delvare
SUSE L3 Support