2024-02-13 07:07:17

by Jerry Hoemann

Subject: [PATCH] watchdog/hpwdt: Support Suspend and Resume

Add call backs to support suspend and resume.

Signed-off-by: Jerry Hoemann <[email protected]>
---
drivers/watchdog/hpwdt.c | 27 +++++++++++++++++++++++++++
1 file changed, 27 insertions(+)

diff --git a/drivers/watchdog/hpwdt.c b/drivers/watchdog/hpwdt.c
index 138dc8d8ca3d..6565cfaa8e57 100644
--- a/drivers/watchdog/hpwdt.c
+++ b/drivers/watchdog/hpwdt.c
@@ -378,11 +378,38 @@ static void hpwdt_exit(struct pci_dev *dev)
pci_disable_device(dev);
}

+static int hpwdt_suspend(struct device *dev)
+{
+ dev_dbg(dev, "Suspend watchdog\n");
+
+ if (watchdog_active(&hpwdt_dev))
+ hpwdt_stop();
+
+ return 0;
+}
+
+static int hpwdt_resume(struct device *dev)
+{
+ dev_dbg(dev, "Resume watchdog\n");
+
+ if (watchdog_active(&hpwdt_dev))
+ hpwdt_start(&hpwdt_dev);
+
+ return 0;
+}
+
+static DEFINE_SIMPLE_DEV_PM_OPS(hpwdt_pm_ops, hpwdt_suspend, hpwdt_resume);
+
static struct pci_driver hpwdt_driver = {
.name = "hpwdt",
.id_table = hpwdt_devices,
.probe = hpwdt_init_one,
.remove = hpwdt_exit,
+
+ .driver = {
+ .name = "hpwdt",
+ .pm = &hpwdt_pm_ops,
+ }
};

MODULE_AUTHOR("Tom Mingarelli");
--
2.43.0

2024-02-13 16:34:50

by Guenter Roeck

Subject: Re: [PATCH] watchdog/hpwdt: Support Suspend and Resume

On Tue, Feb 13, 2024 at 12:02:03AM -0700, Jerry Hoemann wrote:
> Add call backs to support suspend and resume.
>

That makes me wonder if we should add something like
watchdog_stop_on_suspend()
to the watchdog core. Almost every watchdog driver supporting
suspend/resume repeats the same sequence (except for the debug
message). That is a separate question, though.

> Signed-off-by: Jerry Hoemann <[email protected]>
> ---
> drivers/watchdog/hpwdt.c | 27 +++++++++++++++++++++++++++
> 1 file changed, 27 insertions(+)
>
> diff --git a/drivers/watchdog/hpwdt.c b/drivers/watchdog/hpwdt.c
> index 138dc8d8ca3d..6565cfaa8e57 100644
> --- a/drivers/watchdog/hpwdt.c
> +++ b/drivers/watchdog/hpwdt.c
> @@ -378,11 +378,38 @@ static void hpwdt_exit(struct pci_dev *dev)
> pci_disable_device(dev);
> }
>
> +static int hpwdt_suspend(struct device *dev)
> +{
> + dev_dbg(dev, "Suspend watchdog\n");
> +

Doesn't the suspend / resume code already display such debug messages ?

> + if (watchdog_active(&hpwdt_dev))
> + hpwdt_stop();
> +
> + return 0;
> +}
> +
> +static int hpwdt_resume(struct device *dev)
> +{
> + dev_dbg(dev, "Resume watchdog\n");
> +
> + if (watchdog_active(&hpwdt_dev))
> + hpwdt_start(&hpwdt_dev);
> +
> + return 0;
> +}
> +
> +static DEFINE_SIMPLE_DEV_PM_OPS(hpwdt_pm_ops, hpwdt_suspend, hpwdt_resume);

That disables / enables the watchdog as part of regular suspend/resume
handling, meaning any hang during suspend/resume that happens later will
hang the system. Sure you don't want to use SET_LATE_SYSTEM_SLEEP_PM_OPS()
instead ?

Never mind, though. Your call, obviously.

Reviewed-by: Guenter Roeck <[email protected]>

Thanks,
Guenter

2024-02-14 03:16:21

by Jerry Hoemann

Subject: Re: [PATCH] watchdog/hpwdt: Support Suspend and Resume

On Tue, Feb 13, 2024 at 08:12:57AM -0800, Guenter Roeck wrote:
> On Tue, Feb 13, 2024 at 12:02:03AM -0700, Jerry Hoemann wrote:
> > Add call backs to support suspend and resume.
> >
>
> That makes me wonder if we should add something like
> watchdog_stop_on_suspend()
> to the watchdog core. Almost every watchdog driver supporting
> suspend/resume repeats the same sequence (except for the debug
> message). That is a separate question, though.

I think that is a good idea.

>
> > Signed-off-by: Jerry Hoemann <[email protected]>
> > ---
> > drivers/watchdog/hpwdt.c | 27 +++++++++++++++++++++++++++
> > 1 file changed, 27 insertions(+)
> >
> > diff --git a/drivers/watchdog/hpwdt.c b/drivers/watchdog/hpwdt.c
> > index 138dc8d8ca3d..6565cfaa8e57 100644
> > --- a/drivers/watchdog/hpwdt.c
> > +++ b/drivers/watchdog/hpwdt.c
> > @@ -378,11 +378,38 @@ static void hpwdt_exit(struct pci_dev *dev)
> > pci_disable_device(dev);
> > }
> >
> > +static int hpwdt_suspend(struct device *dev)
> > +{
> > + dev_dbg(dev, "Suspend watchdog\n");
> > +
>
> Doesn't the suspend / resume code already display such debug messages ?

It displays some. I mostly was using to make sure that the
callback was being called. I can drop it.


>
> > + if (watchdog_active(&hpwdt_dev))
> > + hpwdt_stop();
> > +
> > + return 0;
> > +}
> > +
> > +static int hpwdt_resume(struct device *dev)
> > +{
> > + dev_dbg(dev, "Resume watchdog\n");
> > +
> > + if (watchdog_active(&hpwdt_dev))
> > + hpwdt_start(&hpwdt_dev);
> > +
> > + return 0;
> > +}
> > +
> > +static DEFINE_SIMPLE_DEV_PM_OPS(hpwdt_pm_ops, hpwdt_suspend, hpwdt_resume);
>
> That disables / enables the watchdog as part of regular suspend/resume
> handling, meaning any hang during suspend/resume that happens later will
> hang the system. Sure you don't want to use SET_LATE_SYSTEM_SLEEP_PM_OPS()
> instead ?
>
> Never mind, though. Your call, obviously.

Your suggestion would be an improvement.
Will change in version 2 of patch.


Thanks for the help.


>
> Reviewed-by: Guenter Roeck <[email protected]>
>
> Thanks,
> Guenter

--

-----------------------------------------------------------------------------
Jerry Hoemann Software Engineer Hewlett Packard Enterprise
-----------------------------------------------------------------------------