linux-kernel - [PATCH regression fix] misc: lis3lv02d_i2c: Fix regulators getting en-/dis-abled twice on suspend/resume

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <20240220190035.53402-1-hdegoede@redhat.com>
Date: Tue, 20 Feb 2024 20:00:35 +0100
From: Hans de Goede <hdegoede@...hat.com>
To: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Cc: Hans de Goede <hdegoede@...hat.com>,
	Arnd Bergmann <arnd@...db.de>,
	Eric Piel <eric.piel@...mplin-utc.net>,
	linux-kernel@...r.kernel.org,
	Paul Menzel <pmenzel@...gen.mpg.de>,
	stable@...r.kernel.org,
	regressions@...ts.linux.dev
Subject: [PATCH regression fix] misc: lis3lv02d_i2c: Fix regulators getting en-/dis-abled twice on suspend/resume

When not configured for wakeup lis3lv02d_i2c_suspend() will call
lis3lv02d_poweroff() even if the device has already been turned off
by the runtime-suspend handler and if configured for wakeup and
the device is runtime-suspended at this point then it is not turned
back on to serve as a wakeup source.

Before commit b1b9f7a49440 ("misc: lis3lv02d_i2c: Add missing setting
of the reg_ctrl callback"), lis3lv02d_poweroff() failed to disable
the regulators which as a side effect made calling poweroff() twice ok.

Now that poweroff() correctly disables the regulators, doing this twice
triggers a WARN() in the regulator core:

unbalanced disables for regulator-dummy
WARNING: CPU: 1 PID: 92 at drivers/regulator/core.c:2999 _regulator_disable
..

Fix lis3lv02d_i2c_suspend() to not call poweroff() a second time if
already runtime-suspended and add a poweron() call when necessary to
make wakeup work.

lis3lv02d_i2c_resume() has similar issues, with an added weirness that
it always powers on the device if it is runtime suspended, after which
the first runtime-resume will call poweron() again, causing the enabled
count for the regulator to increase by 1 every suspend/resume. These
unbalanced regulator_enable() calls cause the regulator to never
be turned off and trigger the following WARN() on driver unbind:

WARNING: CPU: 1 PID: 1724 at drivers/regulator/core.c:2396 _regulator_put

Fix this by making lis3lv02d_i2c_resume() mirror the new suspend().

Fixes: b1b9f7a49440 ("misc: lis3lv02d_i2c: Add missing setting of the reg_ctrl callback")
Reported-by: Paul Menzel <pmenzel@...gen.mpg.de>
Closes: https://lore.kernel.org/regressions/5fc6da74-af0a-4aac-b4d5-a000b39a63a5@molgen.mpg.de/
Cc: stable@...r.kernel.org
Cc: regressions@...ts.linux.dev
Signed-off-by: Hans de Goede <hdegoede@...hat.com>
---
 drivers/misc/lis3lv02d/lis3lv02d_i2c.c | 21 +++++++++++++--------
 1 file changed, 13 insertions(+), 8 deletions(-)

diff --git a/drivers/misc/lis3lv02d/lis3lv02d_i2c.c b/drivers/misc/lis3lv02d/lis3lv02d_i2c.c
index c6eb27d46cb0..15119584473c 100644
--- a/drivers/misc/lis3lv02d/lis3lv02d_i2c.c
+++ b/drivers/misc/lis3lv02d/lis3lv02d_i2c.c
@@ -198,8 +198,14 @@ static int lis3lv02d_i2c_suspend(struct device *dev)
 	struct i2c_client *client = to_i2c_client(dev);
 	struct lis3lv02d *lis3 = i2c_get_clientdata(client);

-	if (!lis3->pdata || !lis3->pdata->wakeup_flags)
+	/* Turn on for wakeup if turned off by runtime suspend */
+	if (lis3->pdata && lis3->pdata->wakeup_flags) {
+		if (pm_runtime_suspended(dev))
+			lis3lv02d_poweron(lis3);
+	/* For non wakeup turn off if not already turned off by runtime suspend */
+	} else if (!pm_runtime_suspended(dev))
 		lis3lv02d_poweroff(lis3);
+
 	return 0;
 }

@@ -208,13 +214,12 @@ static int lis3lv02d_i2c_resume(struct device *dev)
 	struct i2c_client *client = to_i2c_client(dev);
 	struct lis3lv02d *lis3 = i2c_get_clientdata(client);

-	/*
-	 * pm_runtime documentation says that devices should always
-	 * be powered on at resume. Pm_runtime turns them off after system
-	 * wide resume is complete.
-	 */
-	if (!lis3->pdata || !lis3->pdata->wakeup_flags ||
-		pm_runtime_suspended(dev))
+	/* Turn back off if turned on for wakeup and runtime suspended*/
+	if (lis3->pdata && lis3->pdata->wakeup_flags) {
+		if (pm_runtime_suspended(dev))
+			lis3lv02d_poweroff(lis3);
+	/* For non wakeup turn back on if not runtime suspended */
+	} else if (!pm_runtime_suspended(dev))
 		lis3lv02d_poweron(lis3);

 	return 0;
-- 
2.43.0