Re: [PATCH v12 2/2] pwm: Add support for Xilinx AXI Timer

["Uwe Kleine-König" <u.kleine-koenig@pengutronix.de>]

[-- Attachment #1: Type: text/plain, Size: 9768 bytes --]

On Fri, Feb 04, 2022 at 12:51:28PM -0500, Sean Anderson wrote:
> 
> 
> On 1/31/22 11:40 AM, Sean Anderson wrote:
> > 
> > 
> > On 1/31/22 9:10 AM, Uwe Kleine-König wrote:
> >> Hello,
> >> 
> >> first of all: Sorry for taking so long for the next review round.
> >> 
> >>> diff --git a/arch/microblaze/kernel/timer.c b/arch/microblaze/kernel/timer.c
> >>> index f8832cf49384..26c385582c3b 100644
> >>> --- a/arch/microblaze/kernel/timer.c
> >>> +++ b/arch/microblaze/kernel/timer.c
> >>> @@ -251,6 +251,10 @@ static int __init xilinx_timer_init(struct device_node *timer)
> >>>  	u32 timer_num = 1;
> >>>  	int ret;
> >>>  
> >>> +	/* If this property is present, the device is a PWM and not a timer */
> >>> +	if (of_property_read_bool(timer, "#pwm-cells"))
> >>> +		return 0;
> >>> +
> >>>  	if (initialized)
> >>>  		return -EINVAL;
> >>>  
> >>> diff --git a/drivers/pwm/Kconfig b/drivers/pwm/Kconfig
> >>> index 21e3b05a5153..cefbf00b4c7e 100644
> >>> --- a/drivers/pwm/Kconfig
> >>> +++ b/drivers/pwm/Kconfig
> >>> @@ -640,4 +640,18 @@ config PWM_VT8500
> >>>  	  To compile this driver as a module, choose M here: the module
> >>>  	  will be called pwm-vt8500.
> >>>  
> >>> +config PWM_XILINX
> >>> +	tristate "Xilinx AXI Timer PWM support"
> >>> +	depends on OF_ADDRESS
> >>> +	depends on COMMON_CLK
> >>> +	select REGMAP_MMIO
> >>> +	help
> >>> +	  PWM driver for Xilinx LogiCORE IP AXI timers. This timer is
> >>> +	  typically a soft core which may be present in Xilinx FPGAs.
> >>> +	  This device may also be present in Microblaze soft processors.
> >>> +	  If you don't have this IP in your design, choose N.
> >>> +
> >>> +	  To compile this driver as a module, choose M here: the module
> >>> +	  will be called pwm-xilinx.
> >>> +
> >>>  endif
> >>> diff --git a/drivers/pwm/Makefile b/drivers/pwm/Makefile
> >>> index 708840b7fba8..ea785480359b 100644
> >>> --- a/drivers/pwm/Makefile
> >>> +++ b/drivers/pwm/Makefile
> >>> @@ -60,3 +60,4 @@ obj-$(CONFIG_PWM_TWL)		+= pwm-twl.o
> >>>  obj-$(CONFIG_PWM_TWL_LED)	+= pwm-twl-led.o
> >>>  obj-$(CONFIG_PWM_VISCONTI)	+= pwm-visconti.o
> >>>  obj-$(CONFIG_PWM_VT8500)	+= pwm-vt8500.o
> >>> +obj-$(CONFIG_PWM_XILINX)	+= pwm-xilinx.o
> >>> diff --git a/drivers/pwm/pwm-xilinx.c b/drivers/pwm/pwm-xilinx.c
> >>> new file mode 100644
> >>> index 000000000000..b4d93e8812c6
> >>> --- /dev/null
> >>> +++ b/drivers/pwm/pwm-xilinx.c
> >>> @@ -0,0 +1,319 @@
> >>> +// SPDX-License-Identifier: GPL-2.0+
> >>> +/*
> >>> + * Copyright (C) 2021 Sean Anderson <sean.anderson@seco.com>
> >>> + *
> >>> + * Limitations:
> >>> + * - When changing both duty cycle and period, we may end up with one cycle
> >>> + *   with the old duty cycle and the new period. This is because the counters
> >>> + *   may only be reloaded by first stopping them, or by letting them be
> >>> + *   automatically reloaded at the end of a cycle. If this automatic reload
> >>> + *   happens after we set TLR0 but before we set TLR1 then we will have a
> >>> + *   bad cycle. This could probably be fixed by reading TCR0 just before
> >>> + *   reprogramming, but I think it would add complexity for little gain.
> >>> + * - Cannot produce 100% duty cycle by configuring the TLRs. This might be
> >>> + *   possible by stopping the counters at an appropriate point in the cycle,
> >>> + *   but this is not (yet) implemented.
> >>> + * - Only produces "normal" output.
> >>> + * - Always produces low output if disabled.
> >>> + */
> >>> +
> >>> +#include <clocksource/timer-xilinx.h>
> >>> +#include <linux/clk.h>
> >>> +#include <linux/clk-provider.h>
> >>> +#include <linux/device.h>
> >>> +#include <linux/module.h>
> >>> +#include <linux/of.h>
> >>> +#include <linux/platform_device.h>
> >>> +#include <linux/pwm.h>
> >>> +#include <linux/regmap.h>
> >>> +
> >>> +/*
> >>> + * The following functions are "common" to drivers for this device, and may be
> >>> + * exported at a future date.
> >>> + */
> >>> +u32 xilinx_timer_tlr_cycles(struct xilinx_timer_priv *priv, u32 tcsr,
> >>> +			    u64 cycles)
> >>> +{
> >>> +	WARN_ON(cycles < 2 || cycles - 2 > priv->max);
> >>> +
> >>> +	if (tcsr & TCSR_UDT)
> >>> +		return cycles - 2;
> >>> +	return priv->max - cycles + 2;
> >>> +}
> >>> +
> >>> +unsigned int xilinx_timer_get_period(struct xilinx_timer_priv *priv,
> >>> +				     u32 tlr, u32 tcsr)
> >>> +{
> >>> +	u64 cycles;
> >>> +
> >>> +	if (tcsr & TCSR_UDT)
> >>> +		cycles = tlr + 2;
> >>> +	else
> >>> +		cycles = (u64)priv->max - tlr + 2;
> >>> +
> >>> +	/* cycles has a max of 2^32 + 2 */
> >> 
> >> If you add "... so the multiplication doesn't overflow." it becomes more
> >> obvious why this comment is there.
> >> 
> >>> +	return DIV64_U64_ROUND_UP(cycles * NSEC_PER_SEC,
> >>> +				  clk_get_rate(priv->clk));
> >>> +}
> >>> +
> >>> +/*
> >>> + * The idea here is to capture whether the PWM is actually running (e.g.
> >>> + * because we or the bootloader set it up) and we need to be careful to ensure
> >>> + * we don't cause a glitch. According to the data sheet, to enable the PWM we
> >>> + * need to
> >>> + *
> >>> + * - Set both timers to generate mode (MDT=1)
> >>> + * - Set both timers to PWM mode (PWMA=1)
> >>> + * - Enable the generate out signals (GENT=1)
> >>> + *
> >>> + * In addition,
> >>> + *
> >>> + * - The timer must be running (ENT=1)
> >>> + * - The timer must auto-reload TLR into TCR (ARHT=1)
> >>> + * - We must not be in the process of loading TLR into TCR (LOAD=0)
> >>> + * - Cascade mode must be disabled (CASC=0)
> >>> + *
> >>> + * If any of these differ from usual, then the PWM is either disabled, or is
> >>> + * running in a mode that this driver does not support.
> >>> + */
> >>> +#define TCSR_PWM_SET (TCSR_GENT | TCSR_ARHT | TCSR_ENT | TCSR_PWMA)
> >>> +#define TCSR_PWM_CLEAR (TCSR_MDT | TCSR_LOAD)
> >>> +#define TCSR_PWM_MASK (TCSR_PWM_SET | TCSR_PWM_CLEAR)
> >>> +
> >>> +struct xilinx_pwm_device {
> >>> +	struct pwm_chip chip;
> >>> +	struct xilinx_timer_priv priv;
> >>> +};
> >>> +
> >>> +static inline struct xilinx_timer_priv
> >>> +*xilinx_pwm_chip_to_priv(struct pwm_chip *chip)
> >>> +{
> >>> +	return &container_of(chip, struct xilinx_pwm_device, chip)->priv;
> >>> +}
> >>> +
> >>> +static bool xilinx_timer_pwm_enabled(u32 tcsr0, u32 tcsr1)
> >>> +{
> >>> +	return ((TCSR_PWM_MASK | TCSR_CASC) & tcsr0) == TCSR_PWM_SET &&
> >>> +		(TCSR_PWM_MASK & tcsr1) == TCSR_PWM_SET;
> >>> +}
> >>> +
> >>> +static int xilinx_pwm_apply(struct pwm_chip *chip, struct pwm_device *unused,
> >>> +			    const struct pwm_state *state)
> >>> +{
> >>> +	struct xilinx_timer_priv *priv = xilinx_pwm_chip_to_priv(chip);
> >>> +	u32 tlr0, tlr1, tcsr0, tcsr1;
> >>> +	u64 period_cycles, duty_cycles;
> >>> +	unsigned long rate;
> >>> +
> >>> +	if (state->polarity != PWM_POLARITY_NORMAL)
> >>> +		return -EINVAL;
> >>> +
> >>> +	/*
> >>> +	 * To be representable by TLR, cycles must be between 2 and
> >>> +	 * priv->max + 2. To enforce this we can reduce the cycles, but we may
> >>> +	 * not increase them. Caveat emptor: while this does result in more
> >>> +	 * predictable rounding, it may also result in a completely different
> >>> +	 * duty cycle (% high time) than what was requested.
> >>> +	 */
> >>> +	rate = clk_get_rate(priv->clk);
> >>> +	/* Avoid overflow */
> >>> +	period_cycles = min_t(u64, state->period, ULONG_MAX * NSEC_PER_SEC);
> >> 
> >> on a 64 bit platform ULONG_MAX * NSEC_PER_SEC doesn't fit into an u64
> >> ... I think if you replace ULONG_MAX by U32_MAX it works as intended.
> >> 
> >>> +	period_cycles = mul_u64_u32_div(period_cycles, rate, NSEC_PER_SEC);
> >>> +	period_cycles = min_t(u64, period_cycles, priv->max + 2);
> >>> +	if (period_cycles < 2)
> >>> +		return -ERANGE;
> >>> +
> >>> +[...]
> >>> +static void xilinx_pwm_get_state(struct pwm_chip *chip,
> >>> +				 struct pwm_device *unused,
> >>> +				 struct pwm_state *state)
> >>> +{
> >>> +	struct xilinx_timer_priv *priv = xilinx_pwm_chip_to_priv(chip);
> >>> +	u32 tlr0, tlr1, tcsr0, tcsr1;
> >>> +
> >>> +	regmap_read(priv->map, TLR0, &tlr0);
> >>> +	regmap_read(priv->map, TLR1, &tlr1);
> >>> +	regmap_read(priv->map, TCSR0, &tcsr0);
> >>> +	regmap_read(priv->map, TCSR1, &tcsr1);
> >>> +	state->period = xilinx_timer_get_period(priv, tlr0, tcsr0);
> >>> +	state->duty_cycle = xilinx_timer_get_period(priv, tlr1, tcsr1);
> >>> +	state->enabled = xilinx_timer_pwm_enabled(tcsr0, tcsr1);
> >>> +	state->polarity = PWM_POLARITY_NORMAL;
> >>> +
> >>> +	/* 100% duty cycle results in constant low output */
> >>> +	if (state->period == state->duty_cycle)
> >> 
> >> There is a corner case: It can happen that
> >> xilinx_timer_get_period(priv, tlr0, tcsr0) ==
> >> xilinx_timer_get_period(priv, tlr1, tcsr1) but not tlr0 == tlr1.
> >> 
> >> This only happens for clkrate > 1000000000, but given that the fix is
> >> cheap (i.e. check tlr0 == tlr1 instead of state->period ==
> >> state->duty_cycle) I'd suggest to do that.
> 
> This is intentional. xilinx_timer_get_period abstracts over whether UDT
> is set or not. I will fix this when you find me this hardware
> implemented with a 1GHz clock.

So it can happen, that UDT is set for tlr0 but not for tlr1?

For a future you: Can I convince you to add a check for the maximal clk
freq such that finding that issue again when 1 GHz becomes normal is
easier?

Having said that, I wonder what happens if the TLR1 > TLR0?

Best regards
Uwe

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

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]