Re: [PATCH] More i2c driver changes for 2.5.66

[Albert Cranford <ac9410@attbi.com>]

Your right, nobody outside of sensors uses i2c-proc, but
why not create a new i2c-sysfs.h in drivers/i2c locally
and and adjust the couple of existing drivers.
We all know the application library is not going to have
access to include/linux/include/i2c-xxxx.h anyhow.

Later we can completely remove linux/include/linux/i2c-proc.h
which the existing application library relies upon.
ALbert

Greg KH wrote:
> On Thu, Apr 03, 2003 at 01:21:13AM -0500, Albert Cranford wrote:
> 
>>I read the thread concerning the removal of proc.c & proc.h
>>but hope that this does not go to Linus until the interface
>>between i2c & sensors to application is somewhat defined.
>>
>>At the moment we have a sysctl API used by sensors, video and
>>other i2c kernel applications that is working.
> 
> 
> The only in-kernel drivers that were using the sysctl/proc interface was
> the lm75 and adm1021 drivers.  The video and other i2c kernel drivers do
> not use this interface at all.
> 
> Those two drivers, and the two other chip drivers that I added to the
> kernel in this set of patches were converted over to the sysfs interface
> (well one of the new ones were, the other one will build and run, but
> doesn't export any sysfs files yet, that will change soon.)
> 
> 
>>Our desire to switch the sensors to sysfs interface should not
>>break other applications.  At least until we have a model/api
>>to propose to these other drivers.
> 
> 
> Yes, any applications that used the sysctl interface to get data from
> those two driver will break.  However we have to switch at some point in
> time, and the userspace library can't be worked on very well if the
> kernel can't support it yet :)
> 
> So I'm choosing to update the kernel first, and will be working on the
> library in the coming weeks.  As there is no real sensors support
> besides those two drivers in the main kernel, I didn't break much :)
> 
> 
>>In my personal home systems I use it87 driver and have been
>>somewhat successful in switching to sysfs.  A big blocking
>>point is there is no application to read/set/monitor the
>>driver, so it is basically unverified.
> 
> 
> I tested the changes I did by using echo and cat, no library or other
> applications are needed just yet.
> 
> thanks,
> 
> greg k-h
> 
>