linux-kernel - [PATCH v2 2/2] regulator: core: Avoid propagating to supplies when possible

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 for Android: free password hash cracker in your pocket

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

Message-Id: <20181120175255.227783-2-dianders@chromium.org>
Date:   Tue, 20 Nov 2018 09:52:54 -0800
From:   Douglas Anderson <dianders@...omium.org>
To:     Mark Brown <broonie@...nel.org>
Cc:     Bjorn Andersson <bjorn.andersson@...aro.org>,
        Marc Gonzalez <marc.w.gonzalez@...e.fr>, evgreen@...omium.org,
        swboyd@...omium.org, Dmitry Osipenko <digetx@...il.com>,
        ryandcase@...omium.org, David Collins <collinsd@...eaurora.org>,
        linux-arm-msm@...r.kernel.org,
        Douglas Anderson <dianders@...omium.org>,
        Liam Girdwood <lgirdwood@...il.com>,
        linux-kernel@...r.kernel.org
Subject: [PATCH v2 2/2] regulator: core: Avoid propagating to supplies when possible

When we called regulator_enable() on a regulator we'd end up
propagating that call all the way up the chain every time.  This is a
bit of a waste of time.  A child regulator already refcounts its own
enables so it should avoid passing on to its parent unless the
refcount transitioned between 0 and 1.

Historically this hasn't been a huge problem since we skipped dealing
with enable for always-on regulators.  In a previous patch, however,
we removed the always-on optimization.  On one system, the debugfs
regulator_summary was now showing a "use_count" of 33 for a top-level
regulator.

Let's implement this optimization.  This turns out to be fairly
trivial with the recent reorganization of the regulator core.

NOTE: as part of this patch I'll make "always-on" regulators start
with a use count of 1.  This keeps the counts clean when recursively
resolving regulators.

ALSO NOTE: this commit also contains somewhat of a bug fix to
regulator_force_disable().  It was incorrectly looping over
"rdev->open_count" when it should have been looping over use_count.
We have to touch that code anyway (since we should no longer loop at
all), so we'll fix it together in one patch.  Also: since this comes
after commit f8702f9e4aa7 ("regulator: core: Use ww_mutex for
regulators locking") we can now move to use _regulator_disable() for
our supply and keep it in the lock.

Signed-off-by: Douglas Anderson <dianders@...omium.org>
---
NOTE: this patch is based atop the patch ("regulator: core: Only count
load for enabled consumers") because that patch changed the signature
of _regulator_disable() and this patch touches calls to
_regulator_disable().  If desired the order of the two patches could
be swapped.

Changes in v2:
- Squashed in ("Remove loop disabling supplies in regulator_force_disable()")

 drivers/regulator/core.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/regulator/core.c b/drivers/regulator/core.c
index 1271b5406ad7..dbe2f2e6e625 100644
--- a/drivers/regulator/core.c
+++ b/drivers/regulator/core.c
@@ -1824,6 +1824,7 @@ static int regulator_resolve_supply(struct regulator_dev *rdev)
 			rdev->supply = NULL;
 			return ret;
 		}
+		rdev->use_count = 1;
 	}

 	return 0;
@@ -2493,7 +2494,7 @@ static int _regulator_enable(struct regulator *regulator)

 	lockdep_assert_held_once(&rdev->mutex.base);

-	if (rdev->supply) {
+	if (rdev->use_count == 0 && rdev->supply) {
 		ret = _regulator_enable(rdev->supply);
 		if (ret < 0)
 			return ret;
@@ -2541,7 +2542,7 @@ static int _regulator_enable(struct regulator *regulator)
 	_regulator_handle_consumer_disable(regulator);

 err_disable_supply:
-	if (rdev->supply)
+	if (rdev->use_count == 0 && rdev->supply)
 		_regulator_disable(rdev->supply);

 	return ret;
@@ -2650,7 +2651,7 @@ static int _regulator_disable(struct regulator *regulator)
 	if (ret == 0 && rdev->coupling_desc.n_coupled > 1)
 		ret = regulator_balance_voltage(rdev, PM_SUSPEND_ON);

-	if (ret == 0 && rdev->supply)
+	if (ret == 0 && rdev->use_count == 0 && rdev->supply)
 		ret = _regulator_disable(rdev->supply);

 	return ret;
@@ -2735,11 +2736,10 @@ int regulator_force_disable(struct regulator *regulator)
 		ret = drms_uA_update(rdev);
 	}

-	regulator_unlock_dependent(rdev, &ww_ctx);
+	if (rdev->use_count != 0 && rdev->supply)
+		_regulator_disable(rdev->supply);

-	if (rdev->supply)
-		while (rdev->open_count--)
-			regulator_disable(rdev->supply);
+	regulator_unlock_dependent(rdev, &ww_ctx);

 	return ret;
 }
-- 
2.19.1.1215.g8438c0b245-goog