Since the PWM framework is switching struct pwm_args.period's datatype
to u64, prepare for this transition by typecasting it to u32.
Also, since the dividend is still a 32-bit number, any divisor greater
than the numerator will cause the quotient to be zero, so return 0 in
that case to efficiently skip the division.
Signed-off-by: Guru Das Srinagesh <[email protected]>
---
drivers/pwm/pwm-clps711x.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/drivers/pwm/pwm-clps711x.c b/drivers/pwm/pwm-clps711x.c
index 924d39a..da771b1 100644
--- a/drivers/pwm/pwm-clps711x.c
+++ b/drivers/pwm/pwm-clps711x.c
@@ -43,7 +43,10 @@ static void clps711x_pwm_update_val(struct clps711x_chip *priv, u32 n, u32 v)
static unsigned int clps711x_get_duty(struct pwm_device *pwm, unsigned int v)
{
/* Duty cycle 0..15 max */
- return DIV_ROUND_CLOSEST(v * 0xf, pwm->args.period);
+ if (pwm->args.period > (v * 0xf))
+ return 0;
+
+ return DIV_ROUND_CLOSEST(v * 0xf, (u32)pwm->args.period);
}
static int clps711x_pwm_request(struct pwm_chip *chip, struct pwm_device *pwm)
--
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
On Wed, May 20, 2020 at 03:55:57PM -0700, Guru Das Srinagesh wrote:
> Since the PWM framework is switching struct pwm_args.period's datatype
> to u64, prepare for this transition by typecasting it to u32.
>
> Also, since the dividend is still a 32-bit number, any divisor greater
> than the numerator will cause the quotient to be zero, so return 0 in
> that case to efficiently skip the division.
>
> Signed-off-by: Guru Das Srinagesh <[email protected]>
> ---
> drivers/pwm/pwm-clps711x.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/pwm/pwm-clps711x.c b/drivers/pwm/pwm-clps711x.c
> index 924d39a..da771b1 100644
> --- a/drivers/pwm/pwm-clps711x.c
> +++ b/drivers/pwm/pwm-clps711x.c
> @@ -43,7 +43,10 @@ static void clps711x_pwm_update_val(struct clps711x_chip *priv, u32 n, u32 v)
> static unsigned int clps711x_get_duty(struct pwm_device *pwm, unsigned int v)
> {
> /* Duty cycle 0..15 max */
> - return DIV_ROUND_CLOSEST(v * 0xf, pwm->args.period);
> + if (pwm->args.period > (v * 0xf))
> + return 0;
This doesn't look right to me.
DIV_ROUND_CLOSEST() does rounded division and the short circuit doesn't
implement that.
Daniel.
> +
> + return DIV_ROUND_CLOSEST(v * 0xf, (u32)pwm->args.period);
> }
>
> static int clps711x_pwm_request(struct pwm_chip *chip, struct pwm_device *pwm)
> --
> The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
> a Linux Foundation Collaborative Project
>
On Thu, May 21, 2020 at 11:19:34AM +0100, Daniel Thompson wrote:
> On Wed, May 20, 2020 at 03:55:57PM -0700, Guru Das Srinagesh wrote:
> > Since the PWM framework is switching struct pwm_args.period's datatype
> > to u64, prepare for this transition by typecasting it to u32.
> >
> > Also, since the dividend is still a 32-bit number, any divisor greater
> > than the numerator will cause the quotient to be zero, so return 0 in
> > that case to efficiently skip the division.
> >
> > Signed-off-by: Guru Das Srinagesh <[email protected]>
> > ---
> > drivers/pwm/pwm-clps711x.c | 5 ++++-
> > 1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/pwm/pwm-clps711x.c b/drivers/pwm/pwm-clps711x.c
> > index 924d39a..da771b1 100644
> > --- a/drivers/pwm/pwm-clps711x.c
> > +++ b/drivers/pwm/pwm-clps711x.c
> > @@ -43,7 +43,10 @@ static void clps711x_pwm_update_val(struct clps711x_chip *priv, u32 n, u32 v)
> > static unsigned int clps711x_get_duty(struct pwm_device *pwm, unsigned int v)
> > {
> > /* Duty cycle 0..15 max */
> > - return DIV_ROUND_CLOSEST(v * 0xf, pwm->args.period);
> > + if (pwm->args.period > (v * 0xf))
> > + return 0;
>
> This doesn't look right to me.
>
> DIV_ROUND_CLOSEST() does rounded division and the short circuit doesn't
> implement that.
My initial patch [1] was to simply use DIV64_U64_ROUND_CLOSEST(), but I
got review feedback to add a short-circuit (same thread, [2]). I feel
like I should skip the short-circuiting and type casting and simply just
use DIV64_U64_ROUND_CLOSEST() - what do you think?
[1] https://lore.kernel.org/lkml/587f9ccae68ad7e1ce97fa8da6037292af1a5095.1584473399.git.gurus@codeaurora.org/
[2] https://lore.kernel.org/lkml/CAK8P3a2Hi_AoRC3g7qKth4e_Y1jZrbBDhWUb3YPZm10FWMu-ig@mail.gmail.com/
On Thu, May 21, 2020 at 01:25:25PM -0700, Guru Das Srinagesh wrote:
> On Thu, May 21, 2020 at 11:19:34AM +0100, Daniel Thompson wrote:
> > On Wed, May 20, 2020 at 03:55:57PM -0700, Guru Das Srinagesh wrote:
> > > Since the PWM framework is switching struct pwm_args.period's datatype
> > > to u64, prepare for this transition by typecasting it to u32.
> > >
> > > Also, since the dividend is still a 32-bit number, any divisor greater
> > > than the numerator will cause the quotient to be zero, so return 0 in
> > > that case to efficiently skip the division.
> > >
> > > Signed-off-by: Guru Das Srinagesh <[email protected]>
> > > ---
> > > drivers/pwm/pwm-clps711x.c | 5 ++++-
> > > 1 file changed, 4 insertions(+), 1 deletion(-)> > >
> > > diff --git a/drivers/pwm/pwm-clps711x.c b/drivers/pwm/pwm-clps711x.c
> > > index 924d39a..da771b1 100644
> > > --- a/drivers/pwm/pwm-clps711x.c
> > > +++ b/drivers/pwm/pwm-clps711x.c
> > > @@ -43,7 +43,10 @@ static void clps711x_pwm_update_val(struct clps711x_chip *priv, u32 n, u32 v)
> > > static unsigned int clps711x_get_duty(struct pwm_device *pwm, unsigned int v)
> > > {
> > > /* Duty cycle 0..15 max */
> > > - return DIV_ROUND_CLOSEST(v * 0xf, pwm->args.period);
> > > + if (pwm->args.period > (v * 0xf))
> > > + return 0;
> >
> > This doesn't look right to me.
> >
> > DIV_ROUND_CLOSEST() does rounded division and the short circuit doesn't
> > implement that.
>
> My initial patch [1] was to simply use DIV64_U64_ROUND_CLOSEST(), but I
> got review feedback to add a short-circuit (same thread, [2]). I feel
> like I should skip the short-circuiting and type casting and simply just
> use DIV64_U64_ROUND_CLOSEST() - what do you think?
A trivial review of pwm-clps711x.c suggests that the period is always
32-bit anyway so why not just throw away the short circuit entirely and
replace with a comment saying that CLPS711X has a hard coded period
that is always >1000000000 ?
Daniel.
On Fri, May 22, 2020 at 10:37:38AM +0100, Daniel Thompson wrote:
> On Thu, May 21, 2020 at 01:25:25PM -0700, Guru Das Srinagesh wrote:
> > On Thu, May 21, 2020 at 11:19:34AM +0100, Daniel Thompson wrote:
> > > On Wed, May 20, 2020 at 03:55:57PM -0700, Guru Das Srinagesh wrote:
> > > > Since the PWM framework is switching struct pwm_args.period's datatype
> > > > to u64, prepare for this transition by typecasting it to u32.
> > > >
> > > > Also, since the dividend is still a 32-bit number, any divisor greater
> > > > than the numerator will cause the quotient to be zero, so return 0 in
> > > > that case to efficiently skip the division.
> > > >
> > > > Signed-off-by: Guru Das Srinagesh <[email protected]>
> > > > ---
> > > > drivers/pwm/pwm-clps711x.c | 5 ++++-
> > > > 1 file changed, 4 insertions(+), 1 deletion(-)> > >
> > > > diff --git a/drivers/pwm/pwm-clps711x.c b/drivers/pwm/pwm-clps711x.c
> > > > index 924d39a..da771b1 100644
> > > > --- a/drivers/pwm/pwm-clps711x.c
> > > > +++ b/drivers/pwm/pwm-clps711x.c
> > > > @@ -43,7 +43,10 @@ static void clps711x_pwm_update_val(struct clps711x_chip *priv, u32 n, u32 v)
> > > > static unsigned int clps711x_get_duty(struct pwm_device *pwm, unsigned int v)
> > > > {
> > > > /* Duty cycle 0..15 max */
> > > > - return DIV_ROUND_CLOSEST(v * 0xf, pwm->args.period);
> > > > + if (pwm->args.period > (v * 0xf))
> > > > + return 0;
> > >
> > > This doesn't look right to me.
> > >
> > > DIV_ROUND_CLOSEST() does rounded division and the short circuit doesn't
> > > implement that.
> >
> > My initial patch [1] was to simply use DIV64_U64_ROUND_CLOSEST(), but I
> > got review feedback to add a short-circuit (same thread, [2]). I feel
> > like I should skip the short-circuiting and type casting and simply just
> > use DIV64_U64_ROUND_CLOSEST() - what do you think?
>
> A trivial review of pwm-clps711x.c suggests that the period is always
> 32-bit anyway so why not just throw away the short circuit entirely and
> replace with a comment saying that CLPS711X has a hard coded period
> that is always >1000000000 ?
Sorry, I don't follow the significance of 1000000000 - could you please
explain?
Just to clarify, what I was saying in my previous email was the
following: I think it might be simpler to just throw away the short
circuit and just do:
s/DIV_ROUND_CLOSEST/DIV64_U64_ROUND_CLOSEST
like in another patch in this series [1]. That should handle the
rounding properly as per design. Is that okay?
[1] https://lore.kernel.org/lkml/ca783e0f5ff7b517ce0854908f0e89b07551bfe5.1588616856.git.gurus@codeaurora.org/
Thank you.
Guru Das.
On Fri, May 22, 2020 at 04:19:04PM -0700, Guru Das Srinagesh wrote:
> On Fri, May 22, 2020 at 10:37:38AM +0100, Daniel Thompson wrote:
> > On Thu, May 21, 2020 at 01:25:25PM -0700, Guru Das Srinagesh wrote:
> > > On Thu, May 21, 2020 at 11:19:34AM +0100, Daniel Thompson wrote:
> > > > On Wed, May 20, 2020 at 03:55:57PM -0700, Guru Das Srinagesh wrote:
> > > > > Since the PWM framework is switching struct pwm_args.period's datatype
> > > > > to u64, prepare for this transition by typecasting it to u32.
> > > > >
> > > > > Also, since the dividend is still a 32-bit number, any divisor greater
> > > > > than the numerator will cause the quotient to be zero, so return 0 in
> > > > > that case to efficiently skip the division.
> > > > >
> > > > > Signed-off-by: Guru Das Srinagesh <[email protected]>
> > > > > ---
> > > > > drivers/pwm/pwm-clps711x.c | 5 ++++-
> > > > > 1 file changed, 4 insertions(+), 1 deletion(-)> > >
> > > > > diff --git a/drivers/pwm/pwm-clps711x.c b/drivers/pwm/pwm-clps711x.c
> > > > > index 924d39a..da771b1 100644
> > > > > --- a/drivers/pwm/pwm-clps711x.c
> > > > > +++ b/drivers/pwm/pwm-clps711x.c
> > > > > @@ -43,7 +43,10 @@ static void clps711x_pwm_update_val(struct clps711x_chip *priv, u32 n, u32 v)
> > > > > static unsigned int clps711x_get_duty(struct pwm_device *pwm, unsigned int v)
> > > > > {
> > > > > /* Duty cycle 0..15 max */
> > > > > - return DIV_ROUND_CLOSEST(v * 0xf, pwm->args.period);
> > > > > + if (pwm->args.period > (v * 0xf))
> > > > > + return 0;
> > > >
> > > > This doesn't look right to me.
> > > >
> > > > DIV_ROUND_CLOSEST() does rounded division and the short circuit doesn't
> > > > implement that.
> > >
> > > My initial patch [1] was to simply use DIV64_U64_ROUND_CLOSEST(), but I
> > > got review feedback to add a short-circuit (same thread, [2]). I feel
> > > like I should skip the short-circuiting and type casting and simply just
> > > use DIV64_U64_ROUND_CLOSEST() - what do you think?
> >
> > A trivial review of pwm-clps711x.c suggests that the period is always
> > 32-bit anyway so why not just throw away the short circuit entirely and
> > replace with a comment saying that CLPS711X has a hard coded period
> > that is always >1000000000 ?
>
> Sorry, I don't follow the significance of 1000000000 - could you please
> explain?
One of the questions you are asked (by Arnd) was whether the period
could ever be larger than UINT_MAX. I think you gave the wrong answer
to that question when you said the divisor could be 64-bit. For this
driver I don't see how the period could ever be larger than 1000000000
(a number that is approximately 4x smaller than UINT_MAX).
> Just to clarify, what I was saying in my previous email was the
> following: I think it might be simpler to just throw away the short
> circuit and just do:
>
> s/DIV_ROUND_CLOSEST/DIV64_U64_ROUND_CLOSEST
>
> like in another patch in this series [1]. That should handle the
> rounding properly as per design. Is that okay?
The short circuit must go because it is broken and we wouldn't want
that code copied somewhere where the code would actually be reachable.
Personally I don't much care which macro you use although given the
divisor cannot be greater then UINT_MAX I guess DIV_ROUND_CLOSEST
is marginally better.
Daniel.