| Message ID | 2b6eb6c3f68509aa35cdf2e2a586689ae97681ab.1715255980.git.agx@sigxcpu.org (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | [v1,1/1] Input: gpio-keys - expose wakeup keys in sysfs | expand |
Hi Guido, On Thu, May 9, 2024 at 2:00 PM Guido Günther <agx@sigxcpu.org> wrote: > This helps user space to figure out which keys should be used to unidle a > device. E.g on phones the volume rocker should usually not unblank the > screen. > > Signed-off-by: Guido Günther <agx@sigxcpu.org> Thanks for your patch! This is indeed a useful feature. Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be> Tested-by: Geert Uytterhoeven <geert+renesas@glider.be> Gr{oetje,eeting}s, Geert
Hi Guido, On Thu, May 09, 2024 at 02:00:28PM +0200, Guido Günther wrote: > This helps user space to figure out which keys should be used to unidle a > device. E.g on phones the volume rocker should usually not unblank the > screen. How exactly this is supposed to be used? We have "disabled" keys and switches attribute because this function can be controlled at runtime from userspace while wakeup control is a static device setting. Kernel also does not really know if the screen should be unblanked or not, if a button or switch is configured for wake up the kernel will go through wakeup process all the same and then userspace can decide if it should stay woken up or not. Thanks.
Hi, On Mon, May 13, 2024 at 03:13:53PM -0700, Dmitry Torokhov wrote: > Hi Guido, > > On Thu, May 09, 2024 at 02:00:28PM +0200, Guido Günther wrote: > > This helps user space to figure out which keys should be used to unidle a > > device. E.g on phones the volume rocker should usually not unblank the > > screen. > > How exactly this is supposed to be used? We have "disabled" keys and > switches attribute because this function can be controlled at runtime > from userspace while wakeup control is a static device setting. Current Linux userspace usually unblanks/unidles a device on every keypress. That is usually not the expected result on phones where often only the power button and e.g. some home buttons should do this. These keys usually match the keys that are used as wakeup sources to bring a device out of suspend. So if we export the wakeup keys to userspace we can pick some sensible defaults (overridable via hwdb¹). > Kernel also does not really know if the screen should be unblanked or > not, if a button or switch is configured for wake up the kernel will go > through wakeup process all the same and then userspace can decide if it > should stay woken up or not. Yes, we merely want that as a hint to figure out sensible defaults in userspace (which might be a subset of the wakeup keys). Cherrs, -- Guido ¹) See https://gitlab.gnome.org/World/Phosh/gmobile/-/blob/main/data/61-gmobile-wakeup.hwdb?ref_type=heads#L57-L59 > > Thanks. > > -- > Dmitry >
Hi, On Tue, May 14, 2024 at 06:05:20PM +0200, Guido Günther wrote: > Hi, > On Mon, May 13, 2024 at 03:13:53PM -0700, Dmitry Torokhov wrote: > > Hi Guido, > > > > On Thu, May 09, 2024 at 02:00:28PM +0200, Guido Günther wrote: > > > This helps user space to figure out which keys should be used to unidle a > > > device. E.g on phones the volume rocker should usually not unblank the > > > screen. > > > > How exactly this is supposed to be used? We have "disabled" keys and > > switches attribute because this function can be controlled at runtime > > from userspace while wakeup control is a static device setting. > > Current Linux userspace usually unblanks/unidles a device on every > keypress. That is usually not the expected result on phones where often > only the power button and e.g. some home buttons should do this. > > These keys usually match the keys that are used as wakeup sources to > bring a device out of suspend. So if we export the wakeup keys to > userspace we can pick some sensible defaults (overridable via hwdb¹). > > > Kernel also does not really know if the screen should be unblanked or > > not, if a button or switch is configured for wake up the kernel will go > > through wakeup process all the same and then userspace can decide if it > > should stay woken up or not. > > Yes, we merely want that as a hint to figure out sensible defaults in > userspace (which might be a subset of the wakeup keys). Is there anything I can do to get this applied (currently we play catchup per device when the wakup keys change) ? Cheers, -- Guido > > Cherrs, > -- Guido > > ¹) See https://gitlab.gnome.org/World/Phosh/gmobile/-/blob/main/data/61-gmobile-wakeup.hwdb?ref_type=heads#L57-L59 > > > > > Thanks. > > > > -- > > Dmitry > >
diff --git a/drivers/input/keyboard/gpio_keys.c b/drivers/input/keyboard/gpio_keys.c index 9f3bcd41cf67..84f43d1d4375 100644 --- a/drivers/input/keyboard/gpio_keys.c +++ b/drivers/input/keyboard/gpio_keys.c @@ -198,7 +198,8 @@ static void gpio_keys_enable_button(struct gpio_button_data *bdata) */ static ssize_t gpio_keys_attr_show_helper(struct gpio_keys_drvdata *ddata, char *buf, unsigned int type, - bool only_disabled) + bool only_disabled, + bool only_wakeup) { int n_events = get_n_events_by_type(type); unsigned long *bits; @@ -218,6 +219,9 @@ static ssize_t gpio_keys_attr_show_helper(struct gpio_keys_drvdata *ddata, if (only_disabled && !bdata->disabled) continue; + if (only_wakeup && !bdata->button->wakeup) + continue; + __set_bit(*bdata->code, bits); } @@ -297,7 +301,7 @@ static ssize_t gpio_keys_attr_store_helper(struct gpio_keys_drvdata *ddata, return error; } -#define ATTR_SHOW_FN(name, type, only_disabled) \ +#define ATTR_SHOW_FN(name, type, only_disabled, only_wakeup) \ static ssize_t gpio_keys_show_##name(struct device *dev, \ struct device_attribute *attr, \ char *buf) \ @@ -306,22 +310,26 @@ static ssize_t gpio_keys_show_##name(struct device *dev, \ struct gpio_keys_drvdata *ddata = platform_get_drvdata(pdev); \ \ return gpio_keys_attr_show_helper(ddata, buf, \ - type, only_disabled); \ + type, only_disabled, \ + only_wakeup); \ } -ATTR_SHOW_FN(keys, EV_KEY, false); -ATTR_SHOW_FN(switches, EV_SW, false); -ATTR_SHOW_FN(disabled_keys, EV_KEY, true); -ATTR_SHOW_FN(disabled_switches, EV_SW, true); +ATTR_SHOW_FN(keys, EV_KEY, false, false); +ATTR_SHOW_FN(switches, EV_SW, false, false); +ATTR_SHOW_FN(disabled_keys, EV_KEY, true, false); +ATTR_SHOW_FN(disabled_switches, EV_SW, true, false); +ATTR_SHOW_FN(wakeup_keys, EV_KEY, false, true); /* * ATTRIBUTES: * * /sys/devices/platform/gpio-keys/keys [ro] * /sys/devices/platform/gpio-keys/switches [ro] + * /sys/devices/platform/gpio-keys/wakeup_keys [ro] */ static DEVICE_ATTR(keys, S_IRUGO, gpio_keys_show_keys, NULL); static DEVICE_ATTR(switches, S_IRUGO, gpio_keys_show_switches, NULL); +static DEVICE_ATTR(wakeup_keys, S_IRUGO, gpio_keys_show_wakeup_keys, NULL); #define ATTR_STORE_FN(name, type) \ static ssize_t gpio_keys_store_##name(struct device *dev, \ @@ -361,6 +369,7 @@ static struct attribute *gpio_keys_attrs[] = { &dev_attr_switches.attr, &dev_attr_disabled_keys.attr, &dev_attr_disabled_switches.attr, + &dev_attr_wakeup_keys.attr, NULL, }; ATTRIBUTE_GROUPS(gpio_keys);
This helps user space to figure out which keys should be used to unidle a device. E.g on phones the volume rocker should usually not unblank the screen. Signed-off-by: Guido Günther <agx@sigxcpu.org> --- drivers/input/keyboard/gpio_keys.c | 23 ++++++++++++++++------- 1 file changed, 16 insertions(+), 7 deletions(-)