[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <56d8a231-3af7-4e12-a08b-f7faa549c96a@roeck-us.net>
Date: Wed, 29 Jan 2025 06:48:04 -0800
From: Guenter Roeck <linux@...ck-us.net>
To: Marcus Folkesson <marcus.folkesson@...il.com>,
Support Opensource <support.opensource@...semi.com>,
Wim Van Sebroeck <wim@...ux-watchdog.org>
Cc: linux-watchdog@...r.kernel.org, linux-kernel@...r.kernel.org,
Marcus Folkesson <marcus.folkesson@...bitech.com>
Subject: Re: [PATCH] watchdog: da9052_wdt: respect TWDMIN
On 1/29/25 01:24, Marcus Folkesson wrote:
> From: Marcus Folkesson <marcus.folkesson@...bitech.com>
>
> We have to wait at least the minimium time for the watchdog window
> (TWDMIN) before writings to the wdt register after the
> watchdog is activated.
> Otherwise the chip will assert TWD_ERROR and power down to reset mode.
>
This is missing a key explanation:
Why does setting @min_hw_heartbeat_ms in struct watchdog_device
not work for this driver ?
Guenter
> Signed-off-by: Marcus Folkesson <marcus.folkesson@...il.com>
> ---
> The current implementation does not respect the time window for all
> cases not it handle when the jiffie wrap around.
> E.g. setting a new timeout and immediately perform a 'ping' causes the
> chip to assert TWD_ERROR.
> ---
> drivers/watchdog/da9052_wdt.c | 33 +++++++++++++++++++++++----------
> 1 file changed, 23 insertions(+), 10 deletions(-)
>
> diff --git a/drivers/watchdog/da9052_wdt.c b/drivers/watchdog/da9052_wdt.c
> index 77039f2f0be54273df1666fe40c413b6c89285a1..5ca957a6c2219d99cb3e653205e3f556f7cfb56f 100644
> --- a/drivers/watchdog/da9052_wdt.c
> +++ b/drivers/watchdog/da9052_wdt.c
> @@ -45,6 +45,21 @@ static const struct {
> { 7, 131 },
> };
>
> +static void da9052_wdt_wait_for_twdmin(struct watchdog_device *wdt_dev)
> +{
> + struct da9052_wdt_data *driver_data = watchdog_get_drvdata(wdt_dev);
> + unsigned long msecs, twdmin, jnow = jiffies;
> +
> + /*
> + * The host must wait at least TWDMIN ms between writings to the watchdog
> + * register or the DA9052 will assert TWD_ERROR and power down to RESET mode.
> + */
> + twdmin = driver_data->jpast + msecs_to_jiffies(DA9052_TWDMIN);
> + if (time_before(jnow, twdmin)) {
> + msecs = jiffies_to_msecs(twdmin - jnow);
> + mdelay(msecs);
> + }
> +}
>
> static int da9052_wdt_set_timeout(struct watchdog_device *wdt_dev,
> unsigned int timeout)
> @@ -53,6 +68,7 @@ static int da9052_wdt_set_timeout(struct watchdog_device *wdt_dev,
> struct da9052 *da9052 = driver_data->da9052;
> int ret, i;
>
> + da9052_wdt_wait_for_twdmin(wdt_dev);
> /*
> * Disable the Watchdog timer before setting
> * new time out.
> @@ -89,8 +105,8 @@ static int da9052_wdt_set_timeout(struct watchdog_device *wdt_dev,
> }
>
> wdt_dev->timeout = timeout;
> - driver_data->jpast = jiffies;
> }
> + driver_data->jpast = jiffies;
>
> return 0;
> }
> @@ -109,16 +125,9 @@ static int da9052_wdt_ping(struct watchdog_device *wdt_dev)
> {
> struct da9052_wdt_data *driver_data = watchdog_get_drvdata(wdt_dev);
> struct da9052 *da9052 = driver_data->da9052;
> - unsigned long msec, jnow = jiffies;
> int ret;
>
> - /*
> - * We have a minimum time for watchdog window called TWDMIN. A write
> - * to the watchdog before this elapsed time should cause an error.
> - */
> - msec = (jnow - driver_data->jpast) * 1000/HZ;
> - if (msec < DA9052_TWDMIN)
> - mdelay(msec);
> + da9052_wdt_wait_for_twdmin(wdt_dev);
>
> /* Reset the watchdog timer */
> ret = da9052_reg_update(da9052, DA9052_CONTROL_D_REG,
> @@ -130,8 +139,11 @@ static int da9052_wdt_ping(struct watchdog_device *wdt_dev)
> * FIXME: Reset the watchdog core, in general PMIC
> * is supposed to do this
> */
> - return da9052_reg_update(da9052, DA9052_CONTROL_D_REG,
> + ret = da9052_reg_update(da9052, DA9052_CONTROL_D_REG,
> DA9052_CONTROLD_WATCHDOG, 0 << 7);
> +
> + driver_data->jpast = jiffies;
> + return ret;
> }
>
> static const struct watchdog_info da9052_wdt_info = {
> @@ -187,6 +199,7 @@ static int da9052_wdt_probe(struct platform_device *pdev)
> return ret;
> }
>
> + driver_data->jpast = jiffies;
> return devm_watchdog_register_device(dev, &driver_data->wdt);
> }
>
>
> ---
> base-commit: 05dbaf8dd8bf537d4b4eb3115ab42a5fb40ff1f5
> change-id: 20250129-da9052-661dfbb7de38
>
> Best regards,
Powered by blists - more mailing lists