kernel/linux.git/drivers/watchdog/softdog.c, branch linux-4.16.y

watchdog: softdog: fire watchdog even if softirqs do not get to run

2017-03-01T14:15:10+00:00

Checking for timer expiration is done from the softirq TIMER_SOFTIRQ. Since commit 4cd13c21b207 ("softirq: Let ksoftirqd do its job"), pending softirqs are no longer always handled immediately, instead, if there are pending softirqs, and ksoftirqd is in state TASK_RUNNING, the handling of the softirqs are deferred, and are instead supposed to be handled by ksoftirqd, when ksoftirqd gets scheduled. If a user space process with a real-time policy starts to misbehave by never relinquishing the CPU while ksoftirqd is in state TASK_RUNNING, what will happen is that all softirqs will get deferred, while ksoftirqd, which is supposed to handle the deferred softirqs, will never get to run. To make sure that the watchdog is able to fire even when we do not get to run softirqs, replace the timers with hrtimers. Signed-off-by: Niklas Cassel Signed-off-by: Guenter Roeck

watchdog: softdog: make pretimeout support a compile option

2017-02-24T22:00:23+00:00

It occurred to me that the panic pretimeout governor will stall the softdog, because it is purely software which simply breaks when the kernel panics. Testing governors with the softdog on the other hand is really useful, so make this feature a compile time option which nees to be enabled explicitly. This also removes the overhead if pretimeout support is not used because it will now be compiled away (saving ~10% on ARM32). Signed-off-by: Wolfram Sang Reviewed-by: Vladimir Zapolskiy Reviewed-by: Guenter Roeck Signed-off-by: Guenter Roeck

watchdog: softdog: implement pretimeout support

2016-10-08T08:27:27+00:00

Give devices which do not have hardware support for pretimeout at least a software version of it. Signed-off-by: Wolfram Sang Signed-off-by: Vladimir Zapolskiy Reviewed-by: Guenter Roeck Signed-off-by: Guenter Roeck Signed-off-by: Wim Van Sebroeck

watchdog: constify watchdog_ops structures

2016-09-24T06:50:10+00:00

Check for watchdog_ops structures that are only stored in the ops field of a watchdog_device structure. This field is declared const, so watchdog_ops structures that have this property can be declared as const also. The semantic patch that makes this change is as follows: (http://coccinelle.lip6.fr/) // @r disable optional_qualifier@ identifier i; position p; @@ static struct watchdog_ops i@p = { ... }; @ok@ identifier r.i; struct watchdog_device e; position p; @@ e.ops = &i@p; @bad@ position p != {r.p,ok.p}; identifier r.i; struct watchdog_ops e; @@ e@i@p @depends on !bad disable optional_qualifier@ identifier r.i; @@ static +const struct watchdog_ops i = { ... }; // Signed-off-by: Julia Lawall Reviewed-by: Guenter Roeck Signed-off-by: Guenter Roeck Signed-off-by: Wim Van Sebroeck

watchdog: softdog: improve coding style

2016-07-17T18:54:52+00:00

Signed-off-by: Wolfram Sang Reviewed-by: Guenter Roeck Signed-off-by: Guenter Roeck Signed-off-by: Wim Van Sebroeck

watchdog: softdog: drop superfluous set_timeout callback

2016-07-17T18:54:49+00:00

If we leave set_timeout empty, the core will do exactly what is implemented here anyway. Signed-off-by: Wolfram Sang Reviewed-by: Guenter Roeck Signed-off-by: Guenter Roeck Signed-off-by: Wim Van Sebroeck

watchdog: softdog: sort includes to avoid duplicates

2016-07-17T18:54:45+00:00

Signed-off-by: Wolfram Sang Reviewed-by: Guenter Roeck Signed-off-by: Guenter Roeck Signed-off-by: Wim Van Sebroeck

watchdog: softdog: remove forward declaration

2016-07-17T18:54:42+00:00

Signed-off-by: Wolfram Sang Reviewed-by: Guenter Roeck Signed-off-by: Guenter Roeck Signed-off-by: Wim Van Sebroeck

watchdog: softdog: consistently use softdog_ prefix

2016-07-17T18:54:38+00:00

And move module_init/exit to the proper place while here. Signed-off-by: Wolfram Sang Reviewed-by: Guenter Roeck Signed-off-by: Guenter Roeck Signed-off-by: Wim Van Sebroeck

watchdog: softdog: use watchdog core to init timeout value

2016-07-17T18:54:34+00:00

Error string and comment say we fall back to a default, but in reality we bailed out. Refactor the code to use the core helper which then matches the described behaviour. While updating the init message anyhow, shorten it while we are here; no need for versioning there as well and the name is already given via pr_fmt. Signed-off-by: Wolfram Sang Reviewed-by: Guenter Roeck Signed-off-by: Guenter Roeck Signed-off-by: Wim Van Sebroeck