Re: [RESEND PATCH v6 2/3] leds: Add driver for the TLC5925 LED controller

[Jean-Jacques Hiblot]

On 24/08/2022 10:55, Andy Shevchenko wrote:
> On Wed, Aug 24, 2022 at 11:39 AM Jean-Jacques Hiblot
> <jjhiblot@xxxxxxxxxxxxxxx> wrote:
> > On 04/08/2022 23:04, Pavel Machek wrote:
> > > On Thu 2022-08-04 22:23:00, Jean-Jacques Hiblot wrote:
> > > > On 31/07/2022 21:28, Andy Shevchenko wrote:
> > > > > On Fri, Jul 22, 2022 at 10:14 AM Jean-Jacques Hiblot
> > > > > <jjhiblot@xxxxxxxxxxxxxxx> wrote:
> ...
>
> > > > > Sorry for my slowpokeness, but I just realized that this driver may
> > > > > not be needed. What is the difference to existing gpio-74x164?
> > > > It might work. However it might not be as practical and efficient as the
> > > > dedicated LED driver.
> > > >
> > > > I'll give a try.
> > > It is certainly preffered solution. If you decide to re-submit the
> > > driver anyway, please mention that we already have GPIO driver for
> > > compatible chip, and explain why this is superior.
> > sorry for the delay. I tried with the  74x164 gpio driver and it works
> > as expected.
> >
> > The only drawbacks are:
> >
> > - as-is the 74x164 gpio driver supports only one output-enable gpio.
> > However in practice I don't think multiple OE GPIOs will ever be used.
> Let's leave it to the case when it will be needed. So, we can skip this point.
>
> > - with this approach, every time a LED status is changed the whole
> > register has to be sent on the SPI bus. In other words, changes cannot
> > be coalesced.
> But isn't it the same as what you do in your driver? To me it looks
> like you send the entire range of the values each time you change one
> LED's brightness. I don't see any differences with the GPIO driver.
No. The TLC5925 driver updates the register asynchronously: the cached 
value of the register is updated synchronously and then it is 
transferred over SPI using a workqueue. This way if multiple LED are set 
in a short time, the changes are coalesced into a single SPI transfer. 
This is however probably not a must-have feature.
> > I don't know if this is enough to make a dedicated TLC5925 driver
> > desirable in the kernel.
> I don't think you have enough justification for a new driver.