[PATCH 0/3] RFC: Add a drm_aux-dev module.

[Daniel Vetter]

On Thu, Sep 17, 2015 at 12:01 AM, Jani Nikula
<jani.nikula at linux.intel.com> wrote:
> On Wed, 16 Sep 2015, Ville Syrjälä <ville.syrjala at linux.intel.com> wrote:
>> On Wed, Sep 16, 2015 at 01:09:54PM -0700, Daniel Vetter wrote:
>>> On Tue, Sep 15, 2015 at 10:03:09AM -0700, Rafael Antognolli wrote:
>>> > On Tue, Sep 15, 2015 at 07:46:55PM +0300, Ville Syrjälä wrote:
>>> > > On Tue, Sep 15, 2015 at 07:41:06PM +0300, Ville Syrjälä wrote:
>>> > > > On Tue, Sep 15, 2015 at 09:34:15AM -0700, Rafael Antognolli wrote:
>>> > > > > On Tue, Sep 15, 2015 at 10:35:19AM +0300, Ville Syrjälä wrote:
>>> > > > > > On Mon, Sep 14, 2015 at 04:12:29PM -0700, Rafael Antognolli wrote:
>>> > > > > > > This is a tentative implementation of a module that allows reading/writing
>>> > > > > > > arbitrary dpcd registers, following the suggestion from Daniel Vetter. It
>>> > > > > > > assumes the drm aux helpers were used by the driver.
>>> > > > > > >
>>> > > > > > > I tried to follow the i2c-dev implementation as close as possible, but the only
>>> > > > > > > operations that are provided on the dev node are two different ioctl's, one for
>>> > > > > > > reading a register and another one for writing it.
>>> > > > > >
>>> > > > > > Why would you use ioctls instead of normal read/write syscalls?
>>> > > > > >
>>> > > > >
>>> > > > > For writing, that should work fine, I can easily change that if you
>>> > > > > prefer.
>>> > > > >
>>> > > > > For reading, one has to first tell which register address is going to be
>>> > > > > read.
>>> > > >
>>> > > > seek()
>>> > >
>>> > > Sorry typo. Make that lseek().
>>> > >
>>> > Oh, nice, I'll update the patch with this and remove the notifier stuff,
>>> > thanks!
>>>
>>> Well there's also other modes like i2c over aux, and that would be easier
>>> to support with an ioctl in a uniform way. So not sure how much value
>>> there is in reusing read/write for i2c ...
>>
>> We already have i2c-dev for i2c. So not sure what you're after here.
>
> Yeah, definitely don't reinvent the wheel for this.

Of course I didn't mean to reinvent the i2c-dev wheel, but just expose
the underlying low-level dp aux ops to do i2c transaction - that might
be useful for hacking around to figure out some of the recent i2c
changes we've done. Otoh we could add such a raw mode later on as an
ioctl, for plain dpcd reads/writes read/write/lseek seems indeed a
good fit.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch