linux-watchdog.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [RFC][PATCH] watchdog: rti-wdt: Provide set_timeout handler to make existing userspace happy
@ 2021-09-13 11:41 Jan Kiszka
  2021-09-13 14:37 ` Guenter Roeck
  0 siblings, 1 reply; 2+ messages in thread
From: Jan Kiszka @ 2021-09-13 11:41 UTC (permalink / raw)
  To: Wim Van Sebroeck, Guenter Roeck
  Cc: linux-watchdog, Linux Kernel Mailing List, Tero Kristo, Su,
	Bao Cheng (RC-CN DF FA R&D)

From: Jan Kiszka <jan.kiszka@siemens.com>

Prominent userspace - systemd - cannot handle watchdogs without
WDIOF_SETTIMEOUT, even if it was configured to the same time as the
driver selected or was used by firmware to start the watchdog. To avoid
failing in this case, implement a handler that only fails if a deviating
set_timeout is requested.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---

See also https://github.com/systemd/systemd/issues/20683

 drivers/watchdog/rti_wdt.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)

diff --git a/drivers/watchdog/rti_wdt.c b/drivers/watchdog/rti_wdt.c
index 359302f71f7e..365255b15a0d 100644
--- a/drivers/watchdog/rti_wdt.c
+++ b/drivers/watchdog/rti_wdt.c
@@ -173,13 +173,27 @@ static unsigned int rti_wdt_get_timeleft_ms(struct watchdog_device *wdd)
 	return timer_counter;
 }
 
+static int rti_wdt_set_timeout(struct watchdog_device *wdd,
+			       unsigned int timeout)
+{
+	/*
+	 * Updating the timeout after start is actually not supported, but
+	 * let's ignore requests for the already configured value. Helps
+	 * existing userspace such as systemd.
+	 */
+	if (timeout != heartbeat)
+		return -EOPNOTSUPP;
+
+	return 0;
+}
+
 static unsigned int rti_wdt_get_timeleft(struct watchdog_device *wdd)
 {
 	return rti_wdt_get_timeleft_ms(wdd) / 1000;
 }
 
 static const struct watchdog_info rti_wdt_info = {
-	.options = WDIOF_KEEPALIVEPING,
+	.options = WDIOF_KEEPALIVEPING | WDIOF_SETTIMEOUT,
 	.identity = "K3 RTI Watchdog",
 };
 
@@ -187,6 +201,7 @@ static const struct watchdog_ops rti_wdt_ops = {
 	.owner		= THIS_MODULE,
 	.start		= rti_wdt_start,
 	.ping		= rti_wdt_ping,
+	.set_timeout	= rti_wdt_set_timeout,
 	.get_timeleft	= rti_wdt_get_timeleft,
 };
 
-- 
2.31.1

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [RFC][PATCH] watchdog: rti-wdt: Provide set_timeout handler to make existing userspace happy
  2021-09-13 11:41 [RFC][PATCH] watchdog: rti-wdt: Provide set_timeout handler to make existing userspace happy Jan Kiszka
@ 2021-09-13 14:37 ` Guenter Roeck
  0 siblings, 0 replies; 2+ messages in thread
From: Guenter Roeck @ 2021-09-13 14:37 UTC (permalink / raw)
  To: Jan Kiszka, Wim Van Sebroeck
  Cc: linux-watchdog, Linux Kernel Mailing List, Tero Kristo, Su,
	Bao Cheng (RC-CN DF FA R&D)

On 9/13/21 4:41 AM, Jan Kiszka wrote:
> From: Jan Kiszka <jan.kiszka@siemens.com>
> 
> Prominent userspace - systemd - cannot handle watchdogs without
> WDIOF_SETTIMEOUT, even if it was configured to the same time as the
> driver selected or was used by firmware to start the watchdog. To avoid
> failing in this case, implement a handler that only fails if a deviating
> set_timeout is requested.
> 
> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>

NACK. This is a userspace problem. The ABI clearly states that
WDIOF_SETTIMEOUT is optional, and userspace must not depend on it.
We can not start making kernel changes just to make broken userspace
code happy. systemd should be fixed instead.

Guenter

> ---
> 
> See also https://github.com/systemd/systemd/issues/20683
> 
>   drivers/watchdog/rti_wdt.c | 17 ++++++++++++++++-
>   1 file changed, 16 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/watchdog/rti_wdt.c b/drivers/watchdog/rti_wdt.c
> index 359302f71f7e..365255b15a0d 100644
> --- a/drivers/watchdog/rti_wdt.c
> +++ b/drivers/watchdog/rti_wdt.c
> @@ -173,13 +173,27 @@ static unsigned int rti_wdt_get_timeleft_ms(struct watchdog_device *wdd)
>   	return timer_counter;
>   }
>   
> +static int rti_wdt_set_timeout(struct watchdog_device *wdd,
> +			       unsigned int timeout)
> +{
> +	/*
> +	 * Updating the timeout after start is actually not supported, but
> +	 * let's ignore requests for the already configured value. Helps
> +	 * existing userspace such as systemd.
> +	 */
> +	if (timeout != heartbeat)
> +		return -EOPNOTSUPP;
> +
> +	return 0;
> +}
> +
>   static unsigned int rti_wdt_get_timeleft(struct watchdog_device *wdd)
>   {
>   	return rti_wdt_get_timeleft_ms(wdd) / 1000;
>   }
>   
>   static const struct watchdog_info rti_wdt_info = {
> -	.options = WDIOF_KEEPALIVEPING,
> +	.options = WDIOF_KEEPALIVEPING | WDIOF_SETTIMEOUT,
>   	.identity = "K3 RTI Watchdog",
>   };
>   
> @@ -187,6 +201,7 @@ static const struct watchdog_ops rti_wdt_ops = {
>   	.owner		= THIS_MODULE,
>   	.start		= rti_wdt_start,
>   	.ping		= rti_wdt_ping,
> +	.set_timeout	= rti_wdt_set_timeout,
>   	.get_timeleft	= rti_wdt_get_timeleft,
>   };
>   
> 


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2021-09-13 15:31 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-09-13 11:41 [RFC][PATCH] watchdog: rti-wdt: Provide set_timeout handler to make existing userspace happy Jan Kiszka
2021-09-13 14:37 ` Guenter Roeck

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).