Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Mon, 13 May 2019 12:17:08 -0700
Message-Id: <20190513191708.87956-1-rajatja@google.com>
Mime-Version: 1.0
Subject: [RFC PATCH] ACPI: PM: Enable wake-up device GPEs for suspend-to-idle
From:   Rajat Jain <rajatja@google.com>
To:     rjw@rjwysocki.net, lenb@kernel.org, linux-acpi@vger.kernel.org,
        linux-kernel@vger.kernel.org
Cc:     Rajat Jain <rajatja@google.com>, rajatxjain@gmail.com,
        furquan@google.com
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

I noticed that recently multiple systems (chromebooks) couldn't wake
from S0ix using LID or Keyboard after updating to a newer kernel, I
bisected and the issue is seen starting the following commit:

commit f941d3e41da7 ("ACPI: EC / PM: Disable non-wakeup GPEs for
suspend-to-idle")

and found that the issue gets fixed if I revert it. I debugged and
found that although PNP0C0D:00 (representing the LID) is wake capable
and should wakeup the system per the code in acpi_wakeup_gpe_init()
and in drivers/acpi/button.c:

localhost /sys # cat /proc/acpi/wakeup
Device  S-state   Status   Sysfs node
LID0      S4    *enabled   platform:PNP0C0D:00
CREC      S5    *disabled  platform:GOOG0004:00
                *disabled  platform:cros-ec-dev.1.auto
                *disabled  platform:cros-ec-accel.0
                *disabled  platform:cros-ec-accel.1
                *disabled  platform:cros-ec-gyro.0
                *disabled  platform:cros-ec-ring.0
                *disabled  platform:cros-usbpd-charger.2.auto
                *disabled  platform:cros-usbpd-logger.3.auto
D015      S3    *enabled   i2c:i2c-ELAN0000:00
PENH      S3    *enabled   platform:PRP0001:00
XHCI      S3    *enabled   pci:0000:00:14.0
GLAN      S4    *disabled
WIFI      S3    *disabled  pci:0000:00:14.3
localhost /sys #

On debugging, I found that its corresponding GPE is not being enabled.
The particular GPE's "gpe_register_info->enable_for_wake" does not have any
bits set when acpi_enable_all_wakeup_gpes() comes around to use it. I
looked at code and could not find any other code path that should set the
bits in "enable_for_wake" bitmask for the wake enabled devices for s2idle
(I do see that it happens for S3 in acpi_sleep_prepare()).

Thus I used the same call to enable the GPEs for wake enabled devices,
and verified that this fixes the regression I was seeing on multiple of
my devices.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=203579
Signed-off-by: Rajat Jain <rajatja@google.com>
---
 drivers/acpi/sleep.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/acpi/sleep.c b/drivers/acpi/sleep.c
index 403c4ff15349..e52f1238d2d6 100644
--- a/drivers/acpi/sleep.c
+++ b/drivers/acpi/sleep.c
@@ -977,6 +977,8 @@ static int acpi_s2idle_prepare(void)
 	if (acpi_sci_irq_valid())
 		enable_irq_wake(acpi_sci_irq);
 
+	acpi_enable_wakeup_devices(ACPI_STATE_S0);
+
 	/* Change the configuration of GPEs to avoid spurious wakeup. */
 	acpi_enable_all_wakeup_gpes();
 	acpi_os_wait_events_complete();
@@ -1027,6 +1029,8 @@ static void acpi_s2idle_restore(void)
 {
 	acpi_enable_all_runtime_gpes();
 
+	acpi_disable_wakeup_devices(ACPI_STATE_S0);
+
 	if (acpi_sci_irq_valid())
 		disable_irq_wake(acpi_sci_irq);
 
-- 
2.21.0.1020.gf2820cf01a-goog