[v2] pwm: imx-tpm: Use correct MODULO value for EPWM mode

Commit Message

Erik Schumacher Oct. 24, 2024, 6:41 a.m. UTC

The modulo register defines the period of the edge-aligned PWM mode
(which is the only mode implemented). The reference manual states:
"The EPWM period is determined by (MOD + 0001h) ..." So the value that
is written to the MOD register must therefore be one less than the
calculated period length.
A correct MODULO value is particularly relevant if the PWM has to output
a high frequency due to a low period value.

Fixes: 738a1cfec2ed ("pwm: Add i.MX TPM PWM driver support")
Cc: [e-mail]

Signed-off-by: Erik Schumacher <[e-mail]>
---
v2:
 - Add Fixes: and Cc: tags

 drivers/pwm/pwm-imx-tpm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Uwe Kleine-König Oct. 24, 2024, 9:01 p.m. UTC | #1

Hello,

[dropping [e-mail] from Cc]

On Thu, Oct 24, 2024 at 06:41:07AM +0000, Erik Schumacher wrote:
> The modulo register defines the period of the edge-aligned PWM mode
> (which is the only mode implemented). The reference manual states:
> "The EPWM period is determined by (MOD + 0001h) ..." So the value that
> is written to the MOD register must therefore be one less than the
> calculated period length.
> A correct MODULO value is particularly relevant if the PWM has to output
> a high frequency due to a low period value.
> 
> Fixes: 738a1cfec2ed ("pwm: Add i.MX TPM PWM driver support")
> Cc: [e-mail]
> 
> Signed-off-by: Erik Schumacher <[e-mail]>

No empty line between these trailer lines please.

> diff --git a/drivers/pwm/pwm-imx-tpm.c b/drivers/pwm/pwm-imx-tpm.c
> index 96ea343856f0..a05b66ffe208 100644
> --- a/drivers/pwm/pwm-imx-tpm.c
> +++ b/drivers/pwm/pwm-imx-tpm.c
> @@ -106,7 +106,7 @@ static int pwm_imx_tpm_round_state(struct pwm_chip *chip,
>  	p->prescale = prescale;
>  
>  	period_count = (clock_unit + ((1 << prescale) >> 1)) >> prescale;
> -	p->mod = period_count;
> +	p->mod = period_count - 1;

This looks bogus if period_count is 0.

Best regards
Uwe

Erik Schumacher Oct. 25, 2024, 7:25 a.m. UTC | #2

Hello,

thanks for the review.

Am Donnerstag, dem 24.10.2024 um 23:01 +0200 schrieb Uwe Kleine-König:
> Hello,
> 
> [dropping [e-mail] from Cc]
> 
> On Thu, Oct 24, 2024 at 06:41:07AM +0000, Erik Schumacher wrote:
> > The modulo register defines the period of the edge-aligned PWM mode
> > (which is the only mode implemented). The reference manual states:
> > "The EPWM period is determined by (MOD + 0001h) ..." So the value that
> > is written to the MOD register must therefore be one less than the
> > calculated period length.
> > A correct MODULO value is particularly relevant if the PWM has to output
> > a high frequency due to a low period value.
> > 
> > Fixes: 738a1cfec2ed ("pwm: Add i.MX TPM PWM driver support")
> > Cc: [e-mail]
> > 
> > Signed-off-by: Erik Schumacher <[e-mail]>
> 
> No empty line between these trailer lines please.

Noted for v3!

> 
> > diff --git a/drivers/pwm/pwm-imx-tpm.c b/drivers/pwm/pwm-imx-tpm.c
> > index 96ea343856f0..a05b66ffe208 100644
> > --- a/drivers/pwm/pwm-imx-tpm.c
> > +++ b/drivers/pwm/pwm-imx-tpm.c
> > @@ -106,7 +106,7 @@ static int pwm_imx_tpm_round_state(struct pwm_chip *chip,
> >  	p->prescale = prescale;
> >  
> >  	period_count = (clock_unit + ((1 << prescale) >> 1)) >> prescale;
> > -	p->mod = period_count;
> > +	p->mod = period_count - 1;
> 
> This looks bogus if period_count is 0.
> 

That's true. It can only be 0 if someone is requesting a period from
the PWM that is lower than the period of the clock feeding the PWM.
But, it would be an invalid config and returning -EINVAL there seems to
be the right thing (instead of underflowing the integer). I will add
such a check.

> Best regards
> Uwe

Kind regards
Erik

Patch

diff --git a/drivers/pwm/pwm-imx-tpm.c b/drivers/pwm/pwm-imx-tpm.c
index 96ea343856f0..a05b66ffe208 100644
--- a/drivers/pwm/pwm-imx-tpm.c
+++ b/drivers/pwm/pwm-imx-tpm.c
@@ -106,7 +106,7 @@  static int pwm_imx_tpm_round_state(struct pwm_chip *chip,
 	p->prescale = prescale;
 
 	period_count = (clock_unit + ((1 << prescale) >> 1)) >> prescale;
-	p->mod = period_count;
+	p->mod = period_count - 1;
 
 	/* calculate real period HW can support */
 	tmp = (u64)period_count << prescale;