Message-ID: <8f404284c29a6e7736de49ede9a44a2c.squirrel@intranet.cs.nmsu.edu>
Date:	Thu, 14 Jan 2010 16:03:23 -0700
From:	"Rick L. Vinyard, Jr." <rvinyard@...nmsu.edu>
To:	"Miguel Ojeda" <miguel.ojeda.sandonis@...il.com>
Cc:	"Jaya Kumar" <jayakumar.lkml@...il.com>,
	"Pavel Machek" <pavel@....cz>, "Jiri Kosina" <jkosina@...e.cz>,
	linux-kernel@...r.kernel.org, krzysztof.h1@...pl,
	"Andrew Morton" <akpm@...ux-foundation.org>,
	linux-usb@...r.kernel.org, "Oliver Neukum" <oliver@...kum.org>,
	linux-input@...r.kernel.org
Subject: Re: [PATCH] Logitech G13 driver (fixed cc list --- ignore others)

Miguel Ojeda wrote:
> On Thu, Jan 14, 2010 at 10:48 PM, Rick L. Vinyard, Jr.
> <rvinyard@...nmsu.edu> wrote:
>> Miguel Ojeda wrote:
>>> On Tue, Jan 5, 2010 at 1:14 AM, Jaya Kumar <jayakumar.lkml@...il.com>
>>> wrote:
>>>> On Tue, Jan 5, 2010 at 6:48 AM, Pavel Machek <pavel@....cz> wrote:
>>>>>
>>>>> Ok, but I guess you should cc auxdisplay people in future.
>>>>
>>>> Hi Pavel,
>>>>
>>>> I just looked at the drivers/auxdisplay directory and got a bit
>>>> confused. The reason I got confused is because auxdisplay is actually
>>>> an fbdev driver but it is outside of the drivers/video directory. It
>>>> looks like there has only been 1 commit and that was for the Samsung
>>>> KS0108 controller. It also sort of uses platform support but the way
>>>> it is abstracted is odd to my thinking. The controller is ks0108, so
>>>> in my mind that would be the code that would be platform independent,
>>>> and then that would use a board specific IO driver to push data (eg:
>>>> parport or gpio or usb). I think in the long term it would probably
>>>> make sense to write a cleaner approach with a drivers/video/ks0108.c
>>>> which is cleanly platform independent (and back within fbdev proper)
>>>> and then a board specific driver in the appropriate location that
>>>> handles the IO.
>>>
>>> I wrote long ago the driver(s) and people that reviewed it thought it
>>> was better to keep it outside. I think that if someone else is going
>>> to need ks0108, then I agree: we should write a independent driver.
>>>
>>> It should not be hard, it is an easy controller to play with and the
>>> code is already there. I would try to do it; however, I am not sure if
>>> I would be the most appropriate person to code such generic driver, as
>>> I know almost nothing about all drivers/video/* stuff and the ways of
>>> making it truly generic for future video/ users. Still, I will help
>>> gladly.
>>>
>>
>> When I started to look at writing the G13 framebuffer the first code I
>> looked at was the cfag12864b, and started off trying to adapt it.
>>
>
> I hope it was useful, at least at first. : )
>
>> However, as I was digging through the video/* directory looking for
>> something (I forget now what) I came across the hecubafb and patterned
>> the
>> G13 after it instead.
>>
>> In moving between the two, the biggest difference was that I was able to
>> strip out alot of the workqueue code you had since all that was provided
>> by defio. Otherwise, the general structure was almost identical.
>>
>> In particular, what would change is the lower half of cfag12864b.c and
>> you
>> would be able to eliminate almost everything from the /* Update work */
>> and below comment with the exception of cfag12864b_update().
>>
>> cfag12864b_update() would become almost analogous to the g13_fb_update()
>> I
>> have in the G13 driver which is triggered by the deferred_io member of
>> the
>> fb_deferred_io structure.
>>
>> You would have something like:
>>
>> /* Callback from deferred IO workqueue */
>> static void cfag12864b_deferred_io(struct fb_info *info, struct
>> list_head
>> *pagelist)
>> {
>>        cfag12864b_update(info->par);
>> }
>>
>> static struct fb_deferred_io cfag12864b_defio = {
>>        .delay = HZ / CFAG12864B_UPDATE_RATE_DEFAULT,
>>        .deferred_io = cfag12864b_deferred_io,
>> };
>>
>
> Thank you for the analysis of cfag12864b. See below.
>
>>
>> The other major change is that you could eliminate the periodic memcmp()
>> to see if the buffer has change since the deferred_io is only going to
>> trigger on a page write fault.
>
> Yeah, I admit the memcmp() is pretty ugly knowing about deferred_io,
> which I did not. It is strange that anyone pointed it out long before,
> is it new? Are there any known drawbacks?
>

Not sure how old it is... I don't know of any drawbacks.

>>
>> But, that isn't a major change in the code... only in performance.
>>
>
> So less code and greater performance. That sounds like a winning deal!
>
> About ks0108, have you got any thoughts on how to write a generic
> driver? Do you need something special about ks0108? I only needed raw
> output operations so I just implemented that. Also, cfag12864b uses
> two ks0108 controllers and I suppose other LCD's use many more, so
> there are many points that may need a "research".
>

Actually, I don't need the ks0108 code. Way back when Alan Cox suggested
taking a framebuffer approach for the G13, Pavel suggested looking at the
auxdisplay code.

But, the LCD in the G13 is really a USB device that ships the image out as
an interrupt message with the framebuffer image as the payload. So, in
essence, the callback in the G13 is really a usbhid_submit_message() after
some other work to massage the bits from an xbm format to a format
specific to the Logitech game panel.

---

Rick


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists