mds4 at verizon.net
Sat Mar 27 15:54:43 CET 2004
Note that there are three already-in-kernel drivers/macintosh/therm_*.c
which suffer from many of the same problems below.
Mark Studebaker wrote:
> agreed. This doesn't conform to either our 2.4 or 2.6 driver template in
> several ways:
> - no chip detection
> - no bus functionality checking
> - no conformance to sysfs guidelines
> - uses i2c_master_send rather than i2c_smbus_xxx
> - thermal control loop in kernel space
> - Apple-specific and keywest-specific code
> Jean Delvare wrote:
>>> Sure. The current version, which is quite limited, can be found at
>>> I have the docs, so I'll be able to add functionalities if needed.
>> Mmm, as I read it, it doesn't belong to our i2c subsystem (well, it
>> makes use of it but is really different from the hardware monitoring
>> drivers we have). Usually our drivers are passive and simply export
>> their registers, with some conversions, to sysfs. Any active behavior
>> either belongs to the hardware, or to the userspace, but not to the
>> driver, as you have done if I'm not mistaking.
>> So to make it short, I don't think I'm qualified to review your code.
>>>> Make sure
>>>> you follow the sysfs naming conventions as found in
>>>> Documentation/i2c/sysfs-interface of the same Linux tree.
>>> I'm sure I don't ! But since someone is interested, I'll try to adapt.
>> Oh well, this doesn't apply then. The sysfs interface is meant for our
>> driver design, and obviously doesn't apply to yours at the moment.
>> Unless you want to rewrite a driver that fits into our architecture.
More information about the lm-sensors