Hi,
On 25-11-2019 10:25, Andy Shevchenko wrote:
On Fri, Nov 22, 2019 at 08:23:34PM +0100, Hans de Goede wrote:
On some laptops enabling wakeup on the GPIO interrupts used for ACPI _AEI
event handling causes spurious wakeups.
This commit adds a new honor_wakeup option, defaulting to true (our current
behavior), which can be used to disable wakeup on troublesome hardware
to avoid these spurious wakeups.
This is a workaround for an architectural problem with s2idle under Linux
where we do not have any mechanism to immediately go back to sleep after
wakeup events, other then for embedded-controller events using the standard
ACPI EC interface, for details see:
https://lore.kernel.org/linux-acpi/61450f9b-cbc6-0c09-8b3a-aff6bf9a0b3c@xxxxxxxxxx/
One series of laptops which is not able to suspend without this workaround
is the HP x2 10 Cherry Trail models, this commit adds a DMI based quirk
which makes sets honor_wakeup to false on these models.
I'm not against this approach (yeah, it seems we will always have a stream of
quirks for BIOS enabled platforms, especially cheapest ones), though last word
is by Rafael.
Reviewed-by: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
Thanks.
One nit below, though.
Signed-off-by: Hans de Goede <hdegoede@xxxxxxxxxx>
---
drivers/gpio/gpiolib-acpi.c | 33 ++++++++++++++++++++++++++++++++-
1 file changed, 32 insertions(+), 1 deletion(-)
diff --git a/drivers/gpio/gpiolib-acpi.c b/drivers/gpio/gpiolib-acpi.c
index 2b47d906d536..9ce9b449ac4b 100644
--- a/drivers/gpio/gpiolib-acpi.c
+++ b/drivers/gpio/gpiolib-acpi.c
@@ -22,12 +22,18 @@
#include "gpiolib-acpi.h"
#define QUIRK_NO_EDGE_EVENTS_ON_BOOT 0x01l
+#define QUIRK_NO_WAKEUP 0x02l
static int run_edge_events_on_boot = -1;
module_param(run_edge_events_on_boot, int, 0444);
MODULE_PARM_DESC(run_edge_events_on_boot,
"Run edge _AEI event-handlers at boot: 0=no, 1=yes, -1=auto");
+static int honor_wakeup = -1;
+module_param(honor_wakeup, int, 0444);
+MODULE_PARM_DESC(honor_wakeup,
+ "Honor the ACPI wake-capable flag: 0=no, 1=yes, -1=auto");
+
/**
* struct acpi_gpio_event - ACPI GPIO event handler data
*
@@ -283,7 +289,8 @@ static acpi_status acpi_gpiochip_alloc_event(struct acpi_resource *ares,
event->handle = evt_handle;
event->handler = handler;
event->irq = irq;
- event->irq_is_wake = agpio->wake_capable == ACPI_WAKE_CAPABLE;
+ if (honor_wakeup)
+ event->irq_is_wake = agpio->wake_capable == ACPI_WAKE_CAPABLE;
Perhaps:
event->irq_is_wake = honor_wakeup && agpio->wake_capable == ACPI_WAKE_CAPABLE;
Yes that is better, I also noticed some typos in the comment explaining why the
quirk is necessary, I will submit a v2 fixing both.
Regards,
Hans
(I don't care about 80 limit here)
event->pin = pin;
event->desc = desc;
@@ -1337,6 +1344,23 @@ static const struct dmi_system_id gpiolib_acpi_quirks[] = {
},
.driver_data = (void *)QUIRK_NO_EDGE_EVENTS_ON_BOOT,
},
+ {
+ /*
+ * Various HP X2 10 Cherry Trail models use external
+ * embedded-controller connected via I2C + a ACPI GPIO
+ * event handler. The embedded controller generates various
+ * spurious wakeup events when suspended. So disable wakeup
+ * for its handler (it used the only ACPI GPIO event handler).
+ * This breaks wakeup when opening the lid, the user needs
+ * to press the power-button to wakeup the system. The
+ * alternative is suspend simply not working, which is worse.
+ */
+ .matches = {
+ DMI_MATCH(DMI_SYS_VENDOR, "HP"),
+ DMI_MATCH(DMI_PRODUCT_NAME, "HP x2 Detachable 10-p0XX"),
+ },
+ .driver_data = (void *)QUIRK_NO_WAKEUP,
+ },
{} /* Terminating entry */
};
@@ -1356,6 +1380,13 @@ static int acpi_gpio_setup_params(void)
run_edge_events_on_boot = 1;
}
+ if (honor_wakeup < 0) {
+ if (quirks & QUIRK_NO_WAKEUP)
+ honor_wakeup = 0;
+ else
+ honor_wakeup = 1;
+ }
+
return 0;
}
--
2.23.0
