All of lore.kernel.org
 help / color / mirror / Atom feed
From: Dmitry Torokhov <dmitry.torokhov@gmail.com>
To: Danilo Krummrich <danilokrummrich@dk-develop.de>
Cc: linux-input@vger.kernel.org, linux-kernel@vger.kernel.org,
	linus.walleij@linaro.org
Subject: Re: [PATCH 1/3] input: ps2-gpio: use ktime for IRQ timekeeping
Date: Mon, 14 Feb 2022 22:14:24 -0800	[thread overview]
Message-ID: <YgtEwMQMrp3uQinK@google.com> (raw)
In-Reply-To: <20220211212258.80345-2-danilokrummrich@dk-develop.de>

Hi Danilo,

On Fri, Feb 11, 2022 at 10:22:56PM +0100, Danilo Krummrich wrote:
> Using jiffies for the IRQ timekeeping is not sufficient for two reasons:
> 
> (1) Usually jiffies have a resolution of 1ms to 10ms. The IRQ intervals
>     based on the clock frequency of PS2 protocol specification (10kHz -
>     16.7kHz) are between ~60us and 100us only. Therefore only those IRQ
>     intervals can be detected which are either at the end of a transfer
>     or are overly delayed. While this is sufficient in most cases, since
>     we have quite a lot of ways to detect faulty transfers, it can
>     produce false positives in rare cases: When the jiffies value
>     changes right between two interrupt that are in time, we wrongly
>     assume that we missed one or more clock cycles.
> 
> (2) Some gpio controllers (e.g. the one in the bcm283x chips) may generate
>     spurious IRQs when processing interrupts in the frequency given by PS2
>     devices.
> 
> Both issues can be fixed by using ktime resolution for IRQ timekeeping.
> 
> However, it is still possible to miss clock cycles without detecting
> them. When the PS2 device generates the falling edge of the clock signal
> we have between ~30us and 50us to sample the data line, because after
> this time we reach the next rising edge at which the device changes the
> data signal already. But, the only thing we can detect is whether the
> IRQ interval is within the given period. Therefore it is possible to
> have an IRQ latency greater than ~30us to 50us, sample the wrong bit on
> the data line and still be on time with the next IRQ. However, this can
> only happen when within a given transfer the IRQ latency increases
> slowly.
> 
> ___            ______            ______            ______            ___
>    \          /      \          /      \          /      \          /
>     \        /        \        /        \        /        \        /
>      \______/          \______/          \______/          \______/
> 
>     |-----------------|                 |--------|
>          60us/100us                      30us/50us
> 
> Signed-off-by: Danilo Krummrich <danilokrummrich@dk-develop.de>
> ---
>  drivers/input/serio/ps2-gpio.c | 81 ++++++++++++++++++++++++++++------
>  1 file changed, 67 insertions(+), 14 deletions(-)
> 
> diff --git a/drivers/input/serio/ps2-gpio.c b/drivers/input/serio/ps2-gpio.c
> index 8970b49ea09a..7fef4176bdd1 100644
> --- a/drivers/input/serio/ps2-gpio.c
> +++ b/drivers/input/serio/ps2-gpio.c
> @@ -19,6 +19,7 @@
>  #include <linux/of.h>
>  #include <linux/jiffies.h>
>  #include <linux/delay.h>
> +#include <linux/timekeeping.h>
>  
>  #define DRIVER_NAME		"ps2-gpio"
>  
> @@ -44,6 +45,29 @@
>  
>  #define PS2_CMD_RESEND		0xfe
>  
> +/* The PS2 protocol specifies a clock frequency between 10kHz and 16.7kHz,
> + * therefore the maximal interrupt interval should be 100us and the minimum
> + * interrupt interval should be ~60us. Let's allow +/- 20us for frequency
> + * deviations and interrupt latency.
> + *
> + * The data line must be samples after ~30us to 50us after the falling edge,
> + * since the device updates the data line at the rising edge.
> + *
> + * ___            ______            ______            ______            ___
> + *    \          /      \          /      \          /      \          /
> + *     \        /        \        /        \        /        \        /
> + *      \______/          \______/          \______/          \______/
> + *
> + *     |-----------------|                 |--------|
> + *          60us/100us                      30us/50us
> + */
> +#define PS2_CLK_FREQ_MIN_HZ		10000
> +#define PS2_CLK_FREQ_MAX_HZ		16700
> +#define PS2_CLK_MIN_INTERVAL_US		((1000 * 1000) / PS2_CLK_FREQ_MAX_HZ)
> +#define PS2_CLK_MAX_INTERVAL_US		((1000 * 1000) / PS2_CLK_FREQ_MIN_HZ)
> +#define PS2_IRQ_MIN_INTERVAL_US		(PS2_CLK_MIN_INTERVAL_US - 20)
> +#define PS2_IRQ_MAX_INTERVAL_US		(PS2_CLK_MAX_INTERVAL_US + 20)
> +
>  struct ps2_gpio_data {
>  	struct device *dev;
>  	struct serio *serio;
> @@ -59,6 +83,8 @@ struct ps2_gpio_data {
>  	struct completion tx_done;
>  	struct mutex tx_mutex;
>  	struct delayed_work tx_work;
> +	ktime_t tx_start;
> +	ktime_t tx_end;
>  };
>  
>  static int ps2_gpio_open(struct serio *serio)
> @@ -118,6 +144,7 @@ static void ps2_gpio_tx_work_fn(struct work_struct *work)
>  						    struct ps2_gpio_data,
>  						    tx_work);
>  
> +	drvdata->tx_start = ktime_get();
>  	enable_irq(drvdata->irq);
>  	gpiod_direction_output(drvdata->gpio_data, 0);
>  	gpiod_direction_input(drvdata->gpio_clk);
> @@ -128,20 +155,33 @@ static irqreturn_t ps2_gpio_irq_rx(struct ps2_gpio_data *drvdata)
>  	unsigned char byte, cnt;
>  	int data;
>  	int rxflags = 0;
> -	static unsigned long old_jiffies;
> +	static ktime_t t_last, t_now;
> +	s64 us_delta;
>  
>  	byte = drvdata->rx_byte;
>  	cnt = drvdata->rx_cnt;
>  
> -	if (old_jiffies == 0)
> -		old_jiffies = jiffies;
> +	t_now = ktime_get();
> +	if (t_last == 0)

Instead of checking this every time, do you think we could seed the
value in ps2_gpio_open() (and also make it per-port, not static)?

> +		t_last = t_now;
>  
> -	if ((jiffies - old_jiffies) > usecs_to_jiffies(100)) {
> +	/* We need to consider spurious interrupts happening right after a TX xfer
> +	 * finished.
> +	 */
> +	if (unlikely(ktime_us_delta(t_now, drvdata->tx_end) <
> +		     PS2_IRQ_MIN_INTERVAL_US))
> +		goto end;
> +
> +	us_delta = ktime_us_delta(t_now, t_last);
> +	if (us_delta > PS2_IRQ_MAX_INTERVAL_US && cnt) {
>  		dev_err(drvdata->dev,
>  			"RX: timeout, probably we missed an interrupt\n");
>  		goto err;
> +	} else if (us_delta < PS2_IRQ_MIN_INTERVAL_US && t_now != t_last) {
> +		/* Ignore spurious IRQs. */
> +		goto end;
>  	}
> -	old_jiffies = jiffies;
> +	t_last = t_now;
>  
>  	data = gpiod_get_value(drvdata->gpio_data);
>  	if (unlikely(data < 0)) {
> @@ -205,7 +245,7 @@ static irqreturn_t ps2_gpio_irq_rx(struct ps2_gpio_data *drvdata)
>  			goto err;
>  		}
>  		cnt = byte = 0;
> -		old_jiffies = 0;
> +
>  		goto end; /* success */
>  	default:
>  		dev_err(drvdata->dev, "RX: got out of sync with the device\n");
> @@ -217,7 +257,6 @@ static irqreturn_t ps2_gpio_irq_rx(struct ps2_gpio_data *drvdata)
>  
>  err:
>  	cnt = byte = 0;
> -	old_jiffies = 0;
>  	__ps2_gpio_write(drvdata->serio, PS2_CMD_RESEND);
>  end:
>  	drvdata->rx_cnt = cnt;
> @@ -229,20 +268,34 @@ static irqreturn_t ps2_gpio_irq_tx(struct ps2_gpio_data *drvdata)
>  {
>  	unsigned char byte, cnt;
>  	int data;
> -	static unsigned long old_jiffies;
> +	static ktime_t t_last, t_now;
> +	s64 us_delta;
>  
>  	cnt = drvdata->tx_cnt;
>  	byte = drvdata->tx_byte;
>  
> -	if (old_jiffies == 0)
> -		old_jiffies = jiffies;
> +	t_now = ktime_get();
> +	if (t_last == 0)
> +		t_last = t_now;
> +
> +	/* There might be pending IRQs since we disabled IRQs in __ps2_gpio_write().
> +	 * We can expect at least one clock period until the device generates the
> +	 * first falling edge after releasing the clock line.
> +	 */
> +	if (unlikely(ktime_us_delta(t_now, drvdata->tx_start) <
> +		     PS2_CLK_MIN_INTERVAL_US))
> +		goto end;
>  
> -	if ((jiffies - old_jiffies) > usecs_to_jiffies(100)) {
> +	us_delta = ktime_us_delta(t_now, t_last);
> +	if (us_delta > PS2_IRQ_MAX_INTERVAL_US && cnt > 1) {
>  		dev_err(drvdata->dev,
>  			"TX: timeout, probably we missed an interrupt\n");
>  		goto err;
> +	} else if (us_delta < PS2_IRQ_MIN_INTERVAL_US && t_now != t_last) {
> +		/* Ignore spurious IRQs. */
> +		goto end;
>  	}
> -	old_jiffies = jiffies;
> +	t_last = t_now;
>  
>  	switch (cnt) {
>  	case PS2_START_BIT:
> @@ -283,11 +336,11 @@ static irqreturn_t ps2_gpio_irq_tx(struct ps2_gpio_data *drvdata)
>  			goto err;
>  		}
>  
> +		drvdata->tx_end = ktime_get();
>  		drvdata->mode = PS2_MODE_RX;
>  		complete(&drvdata->tx_done);
>  
>  		cnt = 1;
> -		old_jiffies = 0;
>  		goto end; /* success */
>  	default:
>  		/* Probably we missed the stop bit. Therefore we release data
> @@ -303,7 +356,6 @@ static irqreturn_t ps2_gpio_irq_tx(struct ps2_gpio_data *drvdata)
>  
>  err:
>  	cnt = 1;
> -	old_jiffies = 0;
>  	gpiod_direction_input(drvdata->gpio_data);
>  	__ps2_gpio_write(drvdata->serio, drvdata->tx_byte);
>  end:
> @@ -399,6 +451,7 @@ static int ps2_gpio_probe(struct platform_device *pdev)
>  	drvdata->serio = serio;
>  	drvdata->dev = dev;
>  	drvdata->mode = PS2_MODE_RX;
> +	drvdata->tx_end = 0;
>  
>  	/* Tx count always starts at 1, as the start bit is sent implicitly by
>  	 * host-to-device communication initialization.
> -- 
> 2.34.1
> 

Thanks.

-- 
Dmitry

  reply	other threads:[~2022-02-15  6:14 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-02-11 21:22 input: ps2-gpio: use ktime for IRQ timekeeping Danilo Krummrich
2022-02-11 21:22 ` [PATCH 1/3] " Danilo Krummrich
2022-02-15  6:14   ` Dmitry Torokhov [this message]
2022-02-15 12:09     ` Danilo Krummrich
2022-02-11 21:22 ` [PATCH 2/3] input: ps2-gpio: remove tx timeout from ps2_gpio_irq_tx() Danilo Krummrich
2022-02-11 21:22 ` [PATCH 3/3] input: ps2-gpio: don't send rx data before the stop bit Danilo Krummrich

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=YgtEwMQMrp3uQinK@google.com \
    --to=dmitry.torokhov@gmail.com \
    --cc=danilokrummrich@dk-develop.de \
    --cc=linus.walleij@linaro.org \
    --cc=linux-input@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.