RE: [PATCH 6/6] ptp_ocelot: support 4 programmable pins

[Y.b. Lu]

> -----Original Message-----
> From: Richard Cochran <richardcochran@xxxxxxxxx>
> Sent: Tuesday, March 24, 2020 9:08 PM
> To: Y.b. Lu <yangbo.lu@xxxxxxx>
> Cc: linux-kernel@xxxxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx; David S . Miller
> <davem@xxxxxxxxxxxxx>; Vladimir Oltean <vladimir.oltean@xxxxxxx>;
> Claudiu Manoil <claudiu.manoil@xxxxxxx>; Andrew Lunn <andrew@xxxxxxx>;
> Vivien Didelot <vivien.didelot@xxxxxxxxx>; Florian Fainelli
> <f.fainelli@xxxxxxxxx>; Alexandre Belloni <alexandre.belloni@xxxxxxxxxxx>;
> Microchip Linux Driver Support <UNGLinuxDriver@xxxxxxxxxxxxx>
> Subject: Re: [PATCH 6/6] ptp_ocelot: support 4 programmable pins
> 
> On Fri, Mar 20, 2020 at 06:37:26PM +0800, Yangbo Lu wrote:
> > +static int ocelot_ptp_enable(struct ptp_clock_info *ptp,
> > +			     struct ptp_clock_request *rq, int on)
> > +{
> > +	struct ocelot *ocelot = container_of(ptp, struct ocelot, ptp_info);
> > +	enum ocelot_ptp_pins ptp_pin;
> > +	struct timespec64 ts;
> > +	unsigned long flags;
> > +	int pin = -1;
> > +	u32 val;
> > +	s64 ns;
> > +
> > +	switch (rq->type) {
> > +	case PTP_CLK_REQ_PEROUT:
> > +		/* Reject requests with unsupported flags */
> > +		if (rq->perout.flags)
> > +			return -EOPNOTSUPP;
> > +
> > +		/*
> > +		 * TODO: support disabling function
> > +		 * When ptp_disable_pinfunc() is to disable function,
> > +		 * it has already held pincfg_mux.
> > +		 * However ptp_find_pin() in .enable() called also needs
> > +		 * to hold pincfg_mux.
> > +		 * This causes dead lock. So, just return for function
> > +		 * disabling, and this needs fix-up.
> 
> What dead lock?
> 
> When enable(PTP_CLK_REQ_PEROUT, on=0) is called, you don't need to
> call ptp_disable_pinfunc().  Just stop the periodic waveform
> generator.  The assignment of function to pin remains unchanged.

This happens when we try to change pin function through ptp_ioctl PTP_PIN_SETFUNC.
When software holds pincfg_mux and calls ptp_set_pinfunc, it will disable the function previous assigned and the current function of current pin calling ptp_disable_pinfunc.
The problem is the enable callback in ptp_disable_pinfunc may have to hold pincfg_mux for ptp_find_pin.

The calling should be like this,
ptp_set_pinfunc (hold pincfg_mux)
---> ptp_disable_pinfunc
   ---> .enable
      ---> ptp_find_pin (hold pincfg_mux)

Thanks.

> 
> > +		 */
> > +		if (!on)
> > +			break;
> > +
> > +		pin = ptp_find_pin(ocelot->ptp_clock, PTP_PF_PEROUT,
> > +				   rq->perout.index);
> > +		if (pin == 0)
> > +			ptp_pin = PTP_PIN_0;
> > +		else if (pin == 1)
> > +			ptp_pin = PTP_PIN_1;
> > +		else if (pin == 2)
> > +			ptp_pin = PTP_PIN_2;
> > +		else if (pin == 3)
> > +			ptp_pin = PTP_PIN_3;
> > +		else
> > +			return -EINVAL;
> 
> Return -EBUSY here instead.

Thanks. Will modify it in next version.

> 
> Thanks,
> Richard