Openmoko's S3C touchscreen driver porting question

[Maurus Cuelenaere]

Hi,

I'm working on porting the SmartQ 7 to mainline Linux, and as this
doesn't have any touchscreen drivers for the S3C6410, I tried using
Openmoko's one ([1]).

I got it compiling correctly, but when I press the screen, the driver
seems to endlessly print "stylus_action bug" (line 320 at s3c2410_ts.c).
The endless loop is probably because __kfifo_put() fails and so
S3C2410_ADCTSC doesn't get written.

What I don't understand though is why it fails (it seems to return 0),
any ideas?

[1]:
http://gitorious.org/mcuelenaere-smartq-kernel/mcuelenaere-smartq-kernel/commits/experimental-2.6.31-bd-ts

Thanks,
Maurus Cuelenaere

[Nelson Castillo-2]

On Thu, Sep 17, 2009 at 6:56 PM, Maurus Cuelenaere
<[hidden email]> wrote:

> Hi,
>
> I'm working on porting the SmartQ 7 to mainline Linux, and as this doesn't
> have any touchscreen drivers for the S3C6410, I tried using Openmoko's one
> ([1]).
>
> I got it compiling correctly, but when I press the screen, the driver seems
> to endlessly print "stylus_action bug" (line 320 at s3c2410_ts.c).
> The endless loop is probably because __kfifo_put() fails and so
> S3C2410_ADCTSC doesn't get written.

#define TS_EVENT_FIFO_SIZE (2 << 6)

Should be enough. Try a different number but I think this should be enough.
Just for a test try (2 << 10).

A) I would also check if the coordinates make sense.
B) I would also check if event_send_timer_f (case 'P') is ever reached.

> What I don't understand though is why it fails (it seems to return 0), any
> ideas?

It should return "sizeof(int) * 3".

Try checking B.

PS: There is a newer version of this driver that uses the s3c-adc API.
Since this is what upstream wants. I guess I should submit the changes
for andy-tracking (or for one of the new branches).
http://nelson-patches.googlecode.com/svn/trunk/openmoko/kernel/ts-with-s3c-adc/

[Maurus Cuelenaere]

Op 18-09-09 02:18, Nelson Castillo schreef:
> #define TS_EVENT_FIFO_SIZE (2<<  6)
>
> Should be enough. Try a different number but I think this should be enough.
> Just for a test try (2<<  10).
>    

I've tried this before, didn't work

> A) I would also check if the coordinates make sense.
>    

not sure how to do that?

> B) I would also check if event_send_timer_f (case 'P') is ever reached.
>    

doesn't get reached

> It should return "sizeof(int) * 3".
>
> Try checking B.
>
> PS: There is a newer version of this driver that uses the s3c-adc API.
> Since this is what upstream wants. I guess I should submit the changes
> for andy-tracking (or for one of the new branches).
> http://nelson-patches.googlecode.com/svn/trunk/openmoko/kernel/ts-with-s3c-adc/
>    

I haven't tried these patches yet as I'm currently chasing an SDHCI bug
which is blocking me from doing any "real" work.

Thanks for the help.

Regards,
Maurus Cuelenaere

[Nelson Castillo-2]

On Sat, Sep 19, 2009 at 7:07 AM, Maurus Cuelenaere
<[hidden email]> wrote:

> Op 18-09-09 02:18, Nelson Castillo schreef:
>>
>> #define TS_EVENT_FIFO_SIZE (2<<  6)
>>
>> Should be enough. Try a different number but I think this should be
>> enough.
>> Just for a test try (2<<  10).
>>
>
> I've tried this before, didn't work
>
>> A) I would also check if the coordinates make sense.
>>
>
> not sure how to do that?
>
>> B) I would also check if event_send_timer_f (case 'P') is ever reached.
>>
>
> doesn't get reached

Mmm.

Next thing I would check is whether __kfifo_put() succeeds at least once.

tmp = __kfifo_put(ts.event_fifo, (unsigned char *)buf,sizeof(int) * 3);
printk(..., tmp)

If it does I'd try to check if the timer is working.

[Maurus Cuelenaere]

Hi Nelson,

this is an update on your touchscreen driver on the S3C6410:

after some debugging I found out that the S3C6410's touchscreen differs
with the S3C24XX in that it has:

a) 12-bit ADC support (haven't looked at that atm)
b) additional registers used for clearing ADC and TS interrupts

Due to b), the FIFO constantly got filled with up/down messages as the
interrupt was never cleared.

I've added/hacked support for this in my tree currently, but am waiting
till your driver get in upstream
to rebase my tree and try merging my changes.

Regards,

Maurus Cuelenaere

[Nelson Castillo-2]

On Thu, Oct 1, 2009 at 12:57 PM, Maurus Cuelenaere
<[hidden email]> wrote:

> Hi Nelson,
>
> this is an update on your touchscreen driver on the S3C6410:
>
> after some debugging I found out that the S3C6410's touchscreen differs with
> the S3C24XX in that it has:
>
> a) 12-bit ADC support (haven't looked at that atm)
> b) additional registers used for clearing ADC and TS interrupts
>
> Due to b), the FIFO constantly got filled with up/down messages as the
> interrupt was never cleared.
>
> I've added/hacked support for this in my tree currently, but am waiting till
> your driver get in upstream
> to rebase my tree and try merging my changes.

Oh, I see.

Ben Dooks has been maintaining a version of the driver based in the
same sources that is not public. He said he will prepare the patches
and send them to Linux-arm-kernel.

Nelson.-