Hello Conor,

On Wed, Nov 30, 2022 at 09:53:51AM +0000, Conor Dooley wrote:
> On Mon, Nov 21, 2022 at 03:29:39PM +0000, Conor Dooley wrote:
> > On Thu, Nov 17, 2022 at 10:03:13PM +0000, Conor Dooley wrote:
> > > On Thu, Nov 17, 2022 at 10:04:33PM +0100, Uwe Kleine-König wrote:
> > > > On Thu, Nov 17, 2022 at 05:38:26PM +0000, Conor Dooley wrote:
> > > > > On Thu, Nov 17, 2022 at 05:49:50PM +0100, Uwe Kleine-König wrote:
> > > > > > Hello Conor,
> > > > > 
> > > > > Hello Uwe,
> > > > > 
> > > > > > On Thu, Nov 10, 2022 at 09:35:12AM +0000, Conor Dooley wrote:
> > > > > > > [...]
> > > > > > > +
> > > > > > > +static void mchp_core_pwm_enable(struct pwm_chip *chip, struct pwm_device *pwm,
> > > > > > > +				 bool enable, u64 period)
> > > > > > > +{
> > > > > > > +	struct mchp_core_pwm_chip *mchp_core_pwm = to_mchp_core_pwm(chip);
> > > > > > > +	u8 channel_enable, reg_offset, shift;
> > > > > > > +
> > > > > > > +	/*
> > > > > > > +	 * There are two adjacent 8 bit control regs, the lower reg controls
> > > > > > > +	 * 0-7 and the upper reg 8-15. Check if the pwm is in the upper reg
> > > > > > > +	 * and if so, offset by the bus width.
> > > > > > > +	 */
> > > > > > > +	reg_offset = MCHPCOREPWM_EN(pwm->hwpwm >> 3);
> > > > > > > +	shift = pwm->hwpwm & 7;
> > > > > > > +
> > > > > > > +	channel_enable = readb_relaxed(mchp_core_pwm->base + reg_offset);
> > > > > > > +	channel_enable &= ~(1 << shift);
> > > > > > > +	channel_enable |= (enable << shift);
> > > > > > > +
> > > > > > > +	writel_relaxed(channel_enable, mchp_core_pwm->base + reg_offset);
> > > > > > > +	mchp_core_pwm->channel_enabled &= ~BIT(pwm->hwpwm);
> > > > > > > +	mchp_core_pwm->channel_enabled |= enable << pwm->hwpwm;
> > > > > > > +
> > > > > > > +	/*
> > > > > > > +	 * Notify the block to update the waveform from the shadow registers.
> > > > > > > +	 * The updated values will not appear on the bus until they have been
> > > > > > > +	 * applied to the waveform at the beginning of the next period. We must
> > > > > > > +	 * write these registers and wait for them to be applied before
> > > > > > > +	 * considering the channel enabled.
> > > > > > > +	 * If the delay is under 1 us, sleep for at least 1 us anyway.
> > > > > > > +	 */
> > > > > > > +	if (mchp_core_pwm->sync_update_mask & (1 << pwm->hwpwm)) {
> > > > > > > +		u64 delay;
> > > > > > > +
> > > > > > > +		delay = div_u64(period, 1000u) ? : 1u;
> > > > > > > +		writel_relaxed(1U, mchp_core_pwm->base + MCHPCOREPWM_SYNC_UPD);
> > > > > > > +		usleep_range(delay, delay * 2);
> > > > > > > +	}
> > > > > > 
> > > > > > In some cases the delay could be prevented. e.g. when going from one
> > > > > > disabled state to another. If you don't want to complicate the driver
> > > > > > here, maybe point it out in a comment at least?
> > > > > 
> > > > > Maybe this is my naivity talking, but I'd rather wait. Is there not the
> > > > > chance that we re-enter pwm_apply() before the update has actually gone
> > > > > through?
> > > > 
> > > > My idea was to do something like that:
> > > > 
> > > > 	int mchp_core_pwm_apply(....)
> > > > 	{
> > > > 		if (mchp_core_pwm->sync_update_mask & (1 << pwm->hwpwm)) {
> > > > 			/*
> > > > 			 * We're still waiting for an update, don't
> > > > 			 * interfer until it's completed.
> > > > 			 */
> > > > 			while (readl_relaxed(mchp_core_pwm->base + MCHPCOREPWM_SYNC_UPD)) {
> > > > 				cpu_relax();
> > > > 				if (waited_unreasonably_long())
> > > > 					return -ETIMEOUT;
> > > > 			}
> > > > 		}
> > > > 
> > > > 		update_period_and_duty(...);
> > > > 		return 0;
> > > > 	}
> > 
> > So I was doing some fiddling, and the following works reasonably well:
> > 	if (mchp_core_pwm->sync_update_mask & (1 << pwm->hwpwm)) {
> > 		u32 delay = MCHPCOREPWM_TIMEOUT_US;
> > 		u32 sync_upd;
> > 		int ret;
> > 
> > 		writel_relaxed(1u, mchp_core_pwm->base + MCHPCOREPWM_SYNC_UPD);
> > 
> > 		ret = read_poll_timeout(readl, sync_upd, !sync_upd, delay/100, delay,
> > 					false, mchp_core_pwm->base + MCHPCOREPWM_SYNC_UPD);
> > 		if (ret)
> > 			dev_dbg(mchp_core_pwm->chip.dev,
> > 				"timed out waiting for shadow register sync\n");
> > 	}
> > 
> > but...
> > 
> > > > This way you don't have to wait at all if the calls to pwm_apply() are
> > > > infrequent. Of course this only works this way, if you can determine if
> > > > there is a pending update.
> > > 
> > > Ah I think I get what you mean now about waiting for completion &
> > > reading the bit. I don't know off the top of my head if that bit is
> > > readable. Docs say that they're R/W but I don't know if that means that
> > > an AXI read works or if the value is actually readable. I'll try
> > > something like this if I can.
> > 
> > ...it does not implement what I think you suggested & comes with the
> > drawback of inconsistent behaviour depending on whether the timeout is
> > hit or not.
> > 
> > Instead, waiting in apply(), as you suggested, & get_state() looks to be the
> > better option, using the same sort of logic as above, say:
> > static int mchp_core_pwm_wait_for_sync_update(struct mchp_core_pwm_chip *mchp_core_pwm,
> > 					      unsigned int channel)
> > {
> > 	int ret;
> > 
> > 	/*
> > 	 * If a shadow register is used for this PWM channel, and iff there is
> > 	 * a pending update to the waveform, we must wait for it to be applied
> > 	 * before attempting to read its state, as reading the registers yields
> > 	 * the currently implemented settings, the new ones are only readable
> > 	 * once the current period has ended.
> > 	 *
> > 	 * Rather large delays are possible, in the seconds, so to avoid waiting
> > 	 * around for **too** long - cap the wait at 100 ms.
> > 	 */
> > 	if (mchp_core_pwm->sync_update_mask & (1 << channel)) {
> > 		u32 delay = MCHPCOREPWM_TIMEOUT_US;
> > 		u32 sync_upd;
> > 
> > 		writel_relaxed(1u, mchp_core_pwm->base + MCHPCOREPWM_SYNC_UPD);
> > 
> > 		ret = read_poll_timeout(readl, sync_upd, !sync_upd, delay/100, delay,
> > 					false, mchp_core_pwm->base + MCHPCOREPWM_SYNC_UPD);
> > 		if (ret)
> > 			return -ETIMEDOUT;
> > 	}
> > 
> > 	return 0;
> > }
> > 
> > I think that strikes a good balance? We return quickly & don't blocker
> > the caller, but simultaneously try to prevent them from either trying to
> > apply new settings or get the current settings until the last request
> > has gone though?
> > 
> > get_state() returns void though, is it valid behaviour to wait for the
> > timeout there?

There was an approach to change that, see
https://lore.kernel.org/linux-pwm/20220916151506.298488-1-u.kleine-koenig@pengutronix.de

I need to send a v2.

> > I had a check in the core code and found some places where the call in
> > looks like:
> > 	struct pwm_state s1, s2; 
> > 	chip->ops->get_state(chip, pwm, &s1);
> > In this case, exiting early would leave us with a completely wrong
> > idead of the state, if it was to time out.
> > 
> > Either way, it seems like either way we would be misleading the caller
> > of get_state() - perhaps the way around that is to do the wait & then
> > just carry on with get_state()?
> > In that scenario, you'd get the new settings where possible and the old ones
> > otherwise.
> > Returning if the timeout is hit would give you the new settings where possible
> > & otherwise you'd get whatever was passed to get_state().
> > I'm not really sure which of those two situations would be preferred?

Hmm, .get_state should not return the old state. We really want
.get_state to return an error code. Maybe postpone that question until
we have that?

> Apologies for bumping this, I was wondering if any thoughts on the
> above? I'm not sure which is the lesser evil here (or if I have
> misunderstood something).

That's fine. I'm sorry to be not more responsive. This development cycle
is somehow crazy and there are so many open mails in my inbox ... :-\

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | https://www.pengutronix.de/ |