W83627HF SuperIO support in lm_sensors

W83627HF SuperIO support in lm_sensors

[Axel Thimm]

Hi Alfredo,

copying sensors@Stimpy.netroedge.com on this one.

On Mon, Sep 29, 2003 at 05:49:47PM +0200, Alfredo Milani-Comparetti wrote:
> The main problem is that the W83627HF module is not included amongst the 
> loadable modules that can be MODPROBEd with your latest kernel for RH9.
> This leads to a SENSORS-DETECT routine that suggests to load W83627HF, which, 
> in turn, is actually unavailable.
> At present I've reverted to the original kernel from RH9 and written by hand a 
> PERL script which enables some bits in the SuperIO chip in order to let it be 
> found on the SMBus, but I think it would be nice to be able to find all 
> available modules for all known chips :-)

I used the kernel-patching method described in the lm_sensors 2.8.0
documentation. Could you try to do the same, to reproduce whether the
problem is in the packaging or in the upstream sources?

Thanks!
-- 
Axel.Thimm@physik.fu-berlin.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20030929/508c97d7/attachment.bin

W83627HF SuperIO support in lm_sensors

[Jean Delvare]

> > If you want more drivers, either tweak the FILES and Config.in files
> > in mkpatch, or build the modules directly in the lm_sensors
> > directory instead of patching the kernel.
> 
> What do you think is better for an rpm (and what is better supported
> by you in general)? I checked the first way and it would require
> someone providing me with the changes, and these would have to be
> synced with the next lm_sensors release.
>
> The second proposal is something I could do propably without
> assistance (building kernel modules), but this would have to wait for
> some time (a couple of weeks), as it means ripping lm_sensors out of
> the atrpms kernels and providing it seperately, and also rebuilding
> and redistributing all further atrpms kernel modules.

I would use solution 2. This is what I do as a user. It doesn't change a
thing as far as packaging is concerned. You may build the modules inside
the lm_sensors tree rather than inside the kernel tree, they still end
up in /lib/modules, where they have to be. The only difference is that
you can't choose which module you want to install with solution 2, but
being packaging a kernel, you obviously don't want to exclude any driver
anyway.

That said, I think that we could add many drivers (if not all) to the
list of "considered stable" drivers. I can't think of any of our drivers
that may cause serious problems [1] so there is no real reason not to
include them all.
> 
> On Tue, Sep 30, 2003 at 10:40:20AM +0200, Alfredo Milani-Comparetti
> wrote:

Aha, you're the almico.com guy. Nice to see you here actually, I did not
know you were using Linux too :)

> > I see :-)
> > Anyway: since yours is a kernel recompiled for those who "know what
> > they're doing" and since setting up lm_sensors needs some knowledge,
> > I think it would be nice to include all those modules in your
> > distribution. Consider that sensors-detect script already knows
> > which modules are dangerous and which ones are not and acts
> > accordingly. Just a suggestion. Thank you for your time :-)
> 
> OK, sounds reasonable, what do you think, Jean?

Yes, let's go. Either you switch so solution 2, or you wait for us to
update our list, if nobody on the list objects it, of course.



[1] Actually, the adm1021 driver seems to cause much trouble these days,
for an unknown reason, and the i2c-piix4 used to cause trouble to
Thinkpad users - and they *are* both in the list of stable drivers.

-- 
Jean Delvare
http://www.ensicaen.ismra.fr/~delvare/

W83627HF SuperIO support in lm_sensors

[Axel Thimm]

Hi,

On Mon, Sep 29, 2003 at 09:42:48PM +0200, Jean Delvare wrote:
> > On Mon, Sep 29, 2003 at 05:49:47PM +0200, Alfredo Milani-Comparetti
> > wrote:
> > > The main problem is that the W83627HF module is not included
> > > amongst the loadable modules that can be MODPROBEd with your
> > > latest kernel for RH9. [...]
> > I used the kernel-patching method described in the lm_sensors 2.8.0
> > documentation. [...]

> Known "problem" (much more of a feature actually). Our patching script
> does only include the more stable drivers. Other drivers are only
> available when you compile modules outside of the kernel tree. Our
> sensors-detect script doesn't know about that of course, and detects
> whatever it can.

OK, I was not aware of that.

> If you want more drivers, either tweak the FILES and Config.in files in
> mkpatch, or build the modules directly in the lm_sensors directory
> instead of patching the kernel.

What do you think is better for an rpm (and what is better supported
by you in general)? I checked the first way and it would require
someone providing me with the changes, and these would have to be
synced with the next lm_sensors release.

The second proposal is something I could do propably without
assistance (building kernel modules), but this would have to wait for
some time (a couple of weeks), as it means ripping lm_sensors out of
the atrpms kernels and providing it seperately, and also rebuilding
and redistributing all further atrpms kernel modules.

On Tue, Sep 30, 2003 at 10:40:20AM +0200, Alfredo Milani-Comparetti wrote:
> I see :-)
> Anyway: since yours is a kernel recompiled for those who "know what they're 
> doing" and since setting up lm_sensors needs some knowledge, I think it would 
> be nice to include all those modules in your distribution.
> Consider that sensors-detect script already knows which modules are dangerous 
> and which ones are not and acts accordingly.
> Just a suggestion. Thank you for your time :-)

OK, sounds reasonable, what do you think, Jean?
-- 
Axel.Thimm@physik.fu-berlin.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20030930/9815bb5b/attachment.bin

W83627HF SuperIO support in lm_sensors

[Axel Thimm]

On Tue, Sep 30, 2003 at 07:56:45PM +0200, Jean Delvare wrote:
> > > If you want more drivers, either tweak the FILES and Config.in files
> > > in mkpatch, or build the modules directly in the lm_sensors
> > > directory instead of patching the kernel.
> > 
> > What do you think is better for an rpm (and what is better supported
> > by you in general)?

> I would use solution 2. This is what I do as a user. It doesn't change a
> thing as far as packaging is concerned. You may build the modules inside
> the lm_sensors tree rather than inside the kernel tree, they still end
> up in /lib/modules, where they have to be. The only difference is that
> you can't choose which module you want to install with solution 2, but
> being packaging a kernel, you obviously don't want to exclude any driver
> anyway.

Ok, the packaging scheme looks like the following:
lm_sensors-2.8.x-...:			   Contains userland stuff.
lm_sensors-kmdl-2.4.20-20_29...-2.8.x-...: Contains kernel modules
					   built for the kernel
					   version embedded in the rpm
					   name.
I.e. there will be one main non-kernel rpm and several kmdl rpms (one
for each supported kernel rpm). The userland and kernel rpms would be
built out of the same src.rpm with different passes.

> > OK, sounds reasonable, what do you think, Jean?
> 
> Yes, let's go. Either you switch so solution 2, or you wait for us to
> update our list, if nobody on the list objects it, of course.

I will attack it in about a week, as I am currently on the road again.
-- 
Axel.Thimm@physik.fu-berlin.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20031001/6775a2e3/attachment.bin

W83627HF SuperIO support in lm_sensors

[Alfredo Milani-Comparetti]

> Known "problem" (much more of a feature actually). Our patching script
> does only include the more stable drivers. Other drivers are only
> available when you compile modules outside of the kernel tree. Our
> sensors-detect script doesn't know about that of course, and detects
> whatever it can.
> 
> If you want more drivers, either tweak the FILES and Config.in files in
> mkpatch, or build the modules directly in the lm_sensors directory
> instead of patching the kernel.

I see :-)
Anyway: since yours is a kernel recompiled for those who "know what they're 
doing" and since setting up lm_sensors needs some knowledge, I think it would 
be nice to include all those modules in your distribution.
Consider that sensors-detect script already knows which modules are dangerous 
and which ones are not and acts accordingly.
Just a suggestion. Thank you for your time :-)

Alfredo Milani Comparetti
alfredo@milanicomparetti.com
SpeedFan home page --> http://www.almico.com/speedfan.php
Delphi home page --> http://www.almico.com/delphi
Elf home page    --> http://www.radsite.com/elf

W83627HF SuperIO support in lm_sensors

[Jean Delvare]

> On Mon, Sep 29, 2003 at 05:49:47PM +0200, Alfredo Milani-Comparetti
> wrote:
> > The main problem is that the W83627HF module is not included amongst
> > the loadable modules that can be MODPROBEd with your latest kernel
> > for RH9. This leads to a SENSORS-DETECT routine that suggests to
> > load W83627HF, which, in turn, is actually unavailable.
> > At present I've reverted to the original kernel from RH9 and written
> > by hand a PERL script which enables some bits in the SuperIO chip in
> > order to let it be found on the SMBus, but I think it would be nice
> > to be able to find all available modules for all known chips :-)
> 
> I used the kernel-patching method described in the lm_sensors 2.8.0
> documentation. Could you try to do the same, to reproduce whether the
> problem is in the packaging or in the upstream sources?

Known "problem" (much more of a feature actually). Our patching script
does only include the more stable drivers. Other drivers are only
available when you compile modules outside of the kernel tree. Our
sensors-detect script doesn't know about that of course, and detects
whatever it can.

If you want more drivers, either tweak the FILES and Config.in files in
mkpatch, or build the modules directly in the lm_sensors directory
instead of patching the kernel.

Hope that helps.

-- 
Jean Delvare
http://www.ensicaen.ismra.fr/~delvare/