Re: [PATCH] thermal: core: Send a sysfs notification on trip points

[Vincent Whitchurch]

On Mon, Apr 06, 2020 at 11:45:10AM +0200, Daniel Lezcano wrote:
> On 06/04/2020 09:45, Vincent Whitchurch wrote:
> > On Fri, Apr 03, 2020 at 05:26:39PM +0200, Daniel Lezcano wrote:
> >> On 03/04/2020 16:40, Vincent Whitchurch wrote:
> >>> Normally sysfs_notify() is used to notify userspace that the
> >>> value of the sysfs file has changed, but in this case it's
> >>> being used on a sysfs file whose value never changes.  I don't
> >>> know if there are other drivers that do something similar.
> >>
> >> I think so:
> >>
> >> eg.
> >>
> >> drivers/hwmon/adt7x10.c: sysfs_notify(&dev->kobj, NULL,
> >> "temp1_max_alarm"); drivers/hwmon/adt7x10.c:
> >> sysfs_notify(&dev->kobj, NULL, "temp1_min_alarm");
> >> drivers/hwmon/adt7x10.c: sysfs_notify(&dev->kobj, NULL,
> >> "temp1_crit_alarm");
> >>
> >> drivers/hwmon/abx500.c: sysfs_notify(&data->pdev->dev.kobj, NULL,
> >> alarm_node); drivers/hwmon/abx500.c:
> >> sysfs_notify(&data->pdev->dev.kobj, NULL, alarm_node);
> >>
> >> drivers/hwmon/stts751.c: sysfs_notify(&priv->dev->kobj, NULL,
> >> "temp1_max_alarm"); drivers/hwmon/stts751.c:
> >> sysfs_notify(&priv->dev->kobj, NULL, "temp1_min_alarm");
> >>
> >> There are also some other places I believe they are doing the
> >> same like:
> >>
> >> drivers/md/md.c: sysfs_notify(&mddev->kobj, NULL,
> >> "sync_completed"); drivers/md/md.c: sysfs_notify(&mddev->kobj,
> >> NULL, "degraded");
> >
> > AFAICS all these drivers (including the hwmon ones) use
> > sysfs_notify() to notify that the value of the sysfs file has
> > changed, unlike your proposed patch.
> 
> Sorry, I don't have the same understanding:
> 
> drivers/hwmon/adt7x10.c:
> 
>  - receives an interrupt because one of the programmed temperature is
> reached
>  - reads the status to know which one and sends a sysfs notification

In the sysfs file implementation, you can see that the value in the
sysfs file changes based on the same condition:

static ssize_t adt7x10_alarm_show(struct device *dev,
                                  struct device_attribute *da, char *buf)
{
        struct sensor_device_attribute *attr = to_sensor_dev_attr(da);
        int ret;

        ret = adt7x10_read_byte(dev, ADT7X10_STATUS);
        if (ret < 0)
                return ret;

        return sprintf(buf, "%d\n", !!(ret & attr->index));
}

static SENSOR_DEVICE_ATTR_RO(temp1_max_alarm, adt7x10_alarm,
                             ADT7X10_STAT_T_HIGH);

It's the same case with the other examples: the sysfs file's value
changes.

Anyway, as you say it probably doesn't matter as long as it is
documented.