On Mon, Jul 22, 2024 at 04:09:46PM +1200, Chris Packham wrote:
>
> On 22/07/24 15:53, Guenter Roeck wrote:
> > On 7/21/24 17:58, Chris Packham wrote:
> > > By default the PWM duty cycle in hardware is 100%. On some systems this
> > > can cause unwanted fan noise. Add the ability to specify the fan
> > > connections and initial state of the PWMs via device properties.
> > >
> > > Signed-off-by: Chris Packham <chris.packham@xxxxxxxxxxxxxxxxxxx>
> > > ---
> > >
> > > Notes:
> > > Changes in v6:
> > > - Use do_div() instead of plain /
> > > - Use a helper function to avoid repetition between the of and
> > > non-of
> > > code paths.
> > > Changes in v5:
> > > - Deal with PWM frequency and duty cycle being specified in
> > > nanoseconds
> > > Changes in v4:
> > > - Support DT and ACPI fwnodes
> > > - Put PWM into manual mode
> > > Changes in v3:
> > > - Use the pwm provider/consumer bindings
> > > Changes in v2:
> > > - Use correct device property string for frequency
> > > - Allow -EINVAL and only warn on error
> > > - Use a frequency of 0 to indicate that the hardware should be
> > > left as-is
> > >
> > > drivers/hwmon/adt7475.c | 130 ++++++++++++++++++++++++++++++++++++++++
> > > 1 file changed, 130 insertions(+)
> > >
> > > diff --git a/drivers/hwmon/adt7475.c b/drivers/hwmon/adt7475.c
> > > index 4224ffb30483..fc5605d34f36 100644
> > > --- a/drivers/hwmon/adt7475.c
> > > +++ b/drivers/hwmon/adt7475.c
> > > @@ -21,6 +21,8 @@
> > > #include <linux/of.h>
> > > #include <linux/util_macros.h>
> > > +#include <dt-bindings/pwm/pwm.h>
> > > +
> > > /* Indexes for the sysfs hooks */
> > > #define INPUT 0
> > > @@ -1662,6 +1664,130 @@ static int adt7475_set_pwm_polarity(struct
> > > i2c_client *client)
> > > return 0;
> > > }
> > > +struct adt7475_pwm_config {
> > > + int index;
> > > + int freq;
> > > + int flags;
> > > + int duty;
> > > +};
> > > +
> > > +static int _adt7475_pwm_properties_parse_args(u32 args[4], struct
> > > adt7475_pwm_config *cfg)
> > > +{
> > > + unsigned long freq_hz;
> > > + unsigned long duty;
> > > +
> > > + if (args[1] == 0)
> > > + return -EINVAL;
> > > +
> > > + freq_hz = 1000000000UL;
> > > + do_div(freq_hz, args[1]);
> > > + duty = 255 * args[3];
> > > + do_div(duty, args[1]);
> > > +
> >
> > Gues I am a bit at loss here, just as 0-day. Why use do_div ? It is only
> > needed
> > for 64-bit divide operations.
>
> Mainly because of Uwe's comment on v5. I think I've avoided the original u64
> issue now that I'm converting fwnode_reference_args::args to a u32 array.
My comment was only about the build bot finding a division where the gcc
stub was missing with is an indication that do_div should be used.
Usually for PWMs perdiod and duty_cycle are u64, but here it's only
about values from the dtb, so they are u32 and a plain / should be fine.
> can probably get away with plain division, although 255 * args[3] / args[1]
> might overflow in theory but shouldn't in practice.
I don't like possible overflows, but I don't care enough for hwmon
drivers to object. Still a check for args[3] <= 0x1010101 would be easy
enough.
Best regards
Uwe
Attachment:
signature.asc
Description: PGP signature
