Problem with s3c24xx pwm driver

Problem with s3c24xx pwm driver

[Vasily Khoruzhick]

Hi,

I'm encountering very strange problem with s3c24xx pwm driver. I'm using it 
with pwm-bl driver. When backlight level is 0, pwm-bl calls pwm_config(0, 
period); pwm_disable(), otherwise it calls pwm_config(duty, period); 
pwm_enable().

Problem is that after pwm was disabled, it can't be enabled with 
pwm_config(duty, period); pwm_enable() for some (but not all) _valid_ 
duty/period combinations.

pwm-bl works fine if pwm_disable() call is commented out.

Does anybody know how to solve this problem? Any suggestions?

Regards
Vasily
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20090923/436b5290/attachment.sig>

Problem with s3c24xx pwm driver

[Eric Miao]

On Wed, Sep 23, 2009 at 11:24 PM, Vasily Khoruzhick <anarsoul@gmail.com> wrote:
> Hi,
>
> I'm encountering very strange problem with s3c24xx pwm driver. I'm using it
> with pwm-bl driver. When backlight level is 0, pwm-bl calls pwm_config(0,
> period); pwm_disable(), otherwise it calls pwm_config(duty, period);
> pwm_enable().
>
> Problem is that after pwm was disabled, it can't be enabled with
> pwm_config(duty, period); pwm_enable() for some (but not all) _valid_
> duty/period combinations.
>
> pwm-bl works fine if pwm_disable() call is commented out.
>
> Does anybody know how to solve this problem? Any suggestions?
>

Sounds like there are cases where the pwm_enable() and pwm_disable()
are not paired? You may want to print out the reference count each time
to see if this is true.

Problem with s3c24xx pwm driver

[Vasily Khoruzhick]

? ????????? ?? 23 ???????? 2009 22:03:17 ????? Eric Miao ???????:
> Sounds like there are cases where the pwm_enable() and pwm_disable()
> are not paired? You may want to print out the reference count each time
> to see if this is true.


I'm pretty sure they are paired, it's unlikely that it's pwm-bl bug.

Regards
Vasily
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20090923/7afea097/attachment.sig>

Problem with s3c24xx pwm driver

[Vasily Khoruzhick]

? ????????? ?? 23 ???????? 2009 22:03:17 ????? Eric Miao ???????:
> Sounds like there are cases where the pwm_enable() and pwm_disable()
> are not paired? You may want to print out the reference count each time
> to see if this is true.

And, btw, pwm_enable()/pwm_disable for s3c24xx just sets/clears one bit. It 
doesn't harm to set same bit twice, does it?

Regards
Vasily
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20090923/1c740fa4/attachment.sig>

Problem with s3c24xx pwm driver

[Eric Miao]

On Thu, Sep 24, 2009 at 3:23 AM, Vasily Khoruzhick <anarsoul@gmail.com> wrote:
> ? ????????? ?? 23 ???????? 2009 22:03:17 ????? Eric Miao ???????:
>> Sounds like there are cases where the pwm_enable() and pwm_disable()
>> are not paired? You may want to print out the reference count each time
>> to see if this is true.
>
> And, btw, pwm_enable()/pwm_disable for s3c24xx just sets/clears one bit. It
> doesn't harm to set same bit twice, does it?
>

No, it doesn't.

So my suggestion at the moment is that you may:

1. try dump the registers after pwm_enable() and see if all are setup
correctly,

2. and make sure those registers are write-able at that time (some
SoC doesn't allow register writing when the controller is disabled).

Problem with s3c24xx pwm driver

[Vasily Khoruzhick]

? ????????? ?? 24 ???????? 2009 01:25:46 ????? Eric Miao ???????:

> No, it doesn't.
> 
> So my suggestion at the moment is that you may:
> 
> 1. try dump the registers after pwm_enable() and see if all are setup
> correctly,
> 
> 2. and make sure those registers are write-able at that time (some
> SoC doesn't allow register writing when the controller is disabled).
> 

Here's a trace with my comments, but I can't see anything wrong in this trace:

/* Configuring backlight level */
pwm_config:271: TCMPB, wrote: 0000010a, read: 0000010a <- Configure duty
pwm_config:274: TCNTB, wrote: 0000012a, read: 0000012a <- Configure period
pwm_config:281: TCON, wrote: 0051990f, read: 0051990f <- Set manual update 
flag
pwm_config:286: TCON, wrote: 0051990d, read: 0051990d <- Clear manual update 
flag
pwm_enable:149: TCON, wrote: 0051990d, read: 0051990d <- Enable called each 
time if backlight level != 0
/* Configuring backlight level */
pwm_config:271: TCMPB, wrote: 000000a3, read: 000000a3
pwm_config:274: TCNTB, wrote: 0000012a, read: 0000012a
pwm_config:281: TCON, wrote: 0051990f, read: 0051990f
pwm_config:286: TCON, wrote: 0051990d, read: 0051990d
pwm_enable:149: TCON, wrote: 0051990d, read: 0051990d
/* Configuring backlight level */
pwm_config:271: TCMPB, wrote: 00000112, read: 00000112
pwm_config:274: TCNTB, wrote: 0000012a, read: 0000012a
pwm_config:281: TCON, wrote: 0051990f, read: 0051990f
pwm_config:286: TCON, wrote: 0051990d, read: 0051990d
pwm_enable:149: TCON, wrote: 0051990d, read: 0051990d
/* Works fine until next sequence */
/* Disabling backlight */
pwm_config:271: TCMPB, wrote: 00000129, read: 00000129
pwm_config:274: TCNTB, wrote: 0000012a, read: 0000012a
pwm_config:281: TCON, wrote: 0051990f, read: 0051990f
pwm_config:286: TCON, wrote: 0051990d, read: 0051990d
pwm_disable:170: TCON, wrote: 0051990c, read: 0051990c <- Clear start flag

/* Enabling backlight back, but it doesn't work :( */
pwm_config:271: TCMPB, wrote: 00000124, read: 00000124 <- Configure duty
pwm_config:274: TCNTB, wrote: 0000012a, read: 0000012a <- Configure period
pwm_config:281: TCON, wrote: 0051990e, read: 0051990e <- Set manual update 
flag
pwm_config:286: TCON, wrote: 0051990c, read: 0051990c <- Clear manual update 
flag
pwm_enable:149: TCON, wrote: 0051990d, read: 0051990d <- Set start flag
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20090924/cd49e4b9/attachment.sig>

Problem with s3c24xx pwm driver

[Vasily Khoruzhick]

? ????????? ?? 24 ???????? 2009 01:25:46 ?? ????????:

> No, it doesn't.
> 
> So my suggestion at the moment is that you may:
> 
> 1. try dump the registers after pwm_enable() and see if all are setup
> correctly,
> 
> 2. and make sure those registers are write-able at that time (some
> SoC doesn't allow register writing when the controller is disabled).

Ok, I've found source of  problem: it seems to be some issue with backlight 
LED driver (hardware, not software) with selected PWM clock. It seems that  
part of period with "1" is not long enought to charge capacitor, or something 
like this. PWM itself works fine.

Sorry for buzz.

Regards
Vasily
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20090924/3da13106/attachment.sig>

Problem with s3c24xx pwm driver

[Russell King - ARM Linux]

On Thu, Sep 24, 2009 at 05:33:24PM +0300, Vasily Khoruzhick wrote:
> Ok, I've found source of  problem: it seems to be some issue with backlight 
> LED driver (hardware, not software) with selected PWM clock. It seems that  
> part of period with "1" is not long enought to charge capacitor, or something 
> like this. PWM itself works fine.

Is it LED or is it CCFL?  If it's CCFL, it needs to be 'struck' and then
you can reduce the intensity level - by going to max brightness when you
turn it on and then back the intensity down.

That's fairly normal for CCFL stuff.

Problem with s3c24xx pwm driver

[Vasily Khoruzhick]

? ????????? ?? 24 ???????? 2009 23:34:37 ????? Russell King - ARM Linux 
???????:

> Is it LED or is it CCFL?  If it's CCFL, it needs to be 'struck' and then
> you can reduce the intensity level - by going to max brightness when you
> turn it on and then back the intensity down.
> 
> That's fairly normal for CCFL stuff.

I'm not sure, and I don't know how to check it without disassembling PDA. 
Anyway, thanks for suggestion, let's assume that backlight is CCFL.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20090925/0528c02a/attachment.sig>