Re: [PATCH v2 5/8] pwm: Add support for pwmchip devices for faster and easier userspace access

[David Lechner <dlechner@baylibre.com>]

On 7/30/24 5:26 AM, Uwe Kleine-König wrote:
> On Mon, Jul 15, 2024 at 02:37:57PM -0500, David Lechner wrote:
>> On 7/15/24 6:16 AM, Uwe Kleine-König wrote:
>>
>>> +#define PWM_IOCTL_GET_NUM_PWMS	_IO(0x75, 0)
>>
>> What is the use case for PWM_IOCTL_GET_NUM_PWMS? This info is already available
>> from sysfs, and it doesn't seem like there would be any performance consideration
>> for using an ioctl to get the same info.
> 
> Thinking about that (and sending out a v3 that didn't change anything
> here), I think it would be sensible to drop this. Maybe in favour of an
> ioctl that gets the chip id. This way a consumer could find the
> respective device directory below /sys/class/pwm without parsing the
> chip id from the filename (assuming a sane udev configuration).
> 
> Would that make sense to you, too?

How do we expect users to find the "right" PWM to use in the first
place? If libpwm is going to use libudev to enumerate all PWM devices
to find a match then we will already be able to get both the /sys/ and
/dev/ paths for the device from libudev.

And wouldn't the file name in both cases be "pwmchipX" (e.g.
/dev/pwm/pwmchip0 and /sys/class/pwm/pwmchip0) so we wouldn't need
to scrape the number out of the name if we wanted to do the matching
that way?

So I'm not convinced yet that having an IOCTL to get the device ID
it is especially useful.