[lm-sensors] Proposal: remove watchdog support from fscher and

[lm-sensors] Proposal: remove watchdog support from fscher and

[Hans de Goede]

Hi all,

As you all know I've been working lately on moving the fscher driver to 
individual alarm files and on merging the fscpos driver into the fscher driver.

While typing a reply to Jean's review of the fscpos support for the fscher 
driver, I came to the conclusion that it might be best to remove the watchdog 
supporting sysfs attr from these 2 drivers, These sysfs attr are nothing more 
then a raw export of the watchdog registers. If people want raw access they can 
and should use i2c-dev. Thus I think it would best to just remove the watchdog 
sysfs attr, reducing the driver size and complexity. Notice that there are no 
userspace programs which are currently using these sysfs attr.

Maybe in the future a proper watchdog driver can be written for the watchdog 
functions on these chips, under the same subsystem as the other watchdog 
drivers, since its an i2c chip, sharing it with other drivers should not be a 
problem (I think).

Regards,

Hans

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Proposal: remove watchdog support from fscher and

[Jean Delvare]

Hi Hans,

On Thu, 19 Jul 2007 17:32:10 +0200, Hans de Goede wrote:
> As you all know I've been working lately on moving the fscher driver to 
> individual alarm files and on merging the fscpos driver into the fscher driver.
> 
> While typing a reply to Jean's review of the fscpos support for the fscher 
> driver, I came to the conclusion that it might be best to remove the watchdog 
> supporting sysfs attr from these 2 drivers, These sysfs attr are nothing more 
> then a raw export of the watchdog registers. If people want raw access they can 
> and should use i2c-dev. Thus I think it would best to just remove the watchdog 
> sysfs attr, reducing the driver size and complexity.

(Repeating what I wrote in another thread, as this new thread is a
better place:)

I agree that the current watchdog implementation is poor, but you can't
rip it away right now without offering a replacement. Either deprecate
it and plan it for removal at a later point in time, or replace it with
a proper implementation (or both.)

Having i2c-dev and a kernel driver access the same chip isn't a good
idea, and might no longer work in the future.

>                                                      Notice that there are no 
> userspace programs which are currently using these sysfs attr.

How can you be sure? Just because nothing in the lm-sensors package
uses them, doesn't mean that some users don't have dedicated script
using them directly.

> Maybe in the future a proper watchdog driver can be written for the watchdog 
> functions on these chips, under the same subsystem as the other watchdog 
> drivers, since its an i2c chip, sharing it with other drivers should not be a 
> problem (I think).

As the i2c subsystem is being improved to better follow the driver
model, "sharing it with other drivers" doesn't sound good. But nothing
prevents you from having a single driver handling both the hardware
monitoring function and the watchdog function. The problem with the
current driver isn't that it handles both, but only the non-standard
watchdog interface.

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Proposal: remove watchdog support from fscher and

[Hans de Goede]

Jean Delvare wrote:
> Hi Hans,
> 
> On Thu, 19 Jul 2007 17:32:10 +0200, Hans de Goede wrote:
>> As you all know I've been working lately on moving the fscher driver to 
>> individual alarm files and on merging the fscpos driver into the fscher driver.
>>
>> While typing a reply to Jean's review of the fscpos support for the fscher 
>> driver, I came to the conclusion that it might be best to remove the watchdog 
>> supporting sysfs attr from these 2 drivers, These sysfs attr are nothing more 
>> then a raw export of the watchdog registers. If people want raw access they can 
>> and should use i2c-dev. Thus I think it would best to just remove the watchdog 
>> sysfs attr, reducing the driver size and complexity.
> 
> (Repeating what I wrote in another thread, as this new thread is a
> better place:)
> 

Yes, I saw your reply there too, but this whole discussion has got kinda 
obsoleted by the fscscy driver also getting into the picture.

> I agree that the current watchdog implementation is poor, but you can't
> rip it away right now without offering a replacement. Either deprecate
> it and plan it for removal at a later point in time, or replace it with
> a proper implementation (or both.)
> 

I understand that compatibility is important. Since with some bad sysfs api 
decisions were made in the whole fscxxx series, my plan is now to do a new 
driver called fscxxx without any of the baggage, and once that is reviewed an 
in the kernel mark the entire fscher and fscpos driver as obsolete.

Regards,

Hans

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors