linux-kernel - [PATCH RFC 05/10] clk: Move runtime PM calls out of the prepare_lock in clk

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

Message-Id: <20250326-cross-lock-dep-v1-5-3199e49e8652@bootlin.com>
Date: Wed, 26 Mar 2025 19:26:20 +0100
From: Miquel Raynal <miquel.raynal@...tlin.com>
To: "Rafael J. Wysocki" <rafael@...nel.org>, Pavel Machek <pavel@....cz>, 
 Len Brown <len.brown@...el.com>, 
 Greg Kroah-Hartman <gregkh@...uxfoundation.org>, 
 Danilo Krummrich <dakr@...nel.org>, 
 Michael Turquette <mturquette@...libre.com>, 
 Stephen Boyd <sboyd@...nel.org>
Cc: Thomas Petazzoni <thomas.petazzoni@...tlin.com>, 
 linux-pm@...r.kernel.org, linux-kernel@...r.kernel.org, 
 linux-clk@...r.kernel.org, Chen-Yu Tsai <wenst@...omium.org>, 
 Lucas Stach <l.stach@...gutronix.de>, 
 Laurent Pinchart <laurent.pinchart@...asonboard.com>, 
 Marek Vasut <marex@...x.de>, Ulf Hansson <ulf.hansson@...aro.org>, 
 Kevin Hilman <khilman@...nel.org>, Fabio Estevam <festevam@...x.de>, 
 Jacky Bai <ping.bai@....com>, Peng Fan <peng.fan@....com>, 
 Shawn Guo <shawnguo@...nel.org>, Shengjiu Wang <shengjiu.wang@....com>, 
 linux-imx@....com, Ian Ray <ian.ray@...ealthcare.com>, 
 Hervé Codina <herve.codina@...tlin.com>, 
 Luca Ceresoli <luca.ceresoli@...tlin.com>, 
 Saravana Kannan <saravanak@...gle.com>, 
 Miquel Raynal <miquel.raynal@...tlin.com>
Subject: [PATCH RFC 05/10] clk: Move runtime PM calls out of the
 prepare_lock in clk_init()

In order to fix the ABBA locking situation between clock and power
domains, let's disimburse these two locks by preventing any runtime PM
call to happen with the clk prepare_lock mutex acquired.

The __clock_core_init() routine shall preferably call the runtime PM
functions before acquiring the prepare lock, which can be achieved by
the existing clk_pm_runtime_get_all() call. We have no other choice than
waking up all clocks at this stage because we are going the possibly
reparent any random orphan clock in the system at this stage, and thus
will require all its parents and children to be enabled.

Signed-off-by: Miquel Raynal <miquel.raynal@...tlin.com>
---
 drivers/clk/clk.c | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)

diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c
index a6148eae7b6615d23d6ac665e8f94e2ae69f93b6..b6decfaf8aa6bd50c2173d19f1ece8f08a95afd8 100644
--- a/drivers/clk/clk.c
+++ b/drivers/clk/clk.c
@@ -3906,6 +3906,14 @@ static int __clk_core_init(struct clk_core *core)
 	unsigned long rate;
 	int phase;

+	/*
+	 * While reparenting the orphans, it is too late to resume the children
+	 * that will get their rates recalculated.
+	 */
+	ret = clk_pm_runtime_get_all();
+	if (ret)
+		return ret;
+
 	clk_prepare_lock();

 	/*
@@ -3916,10 +3924,6 @@ static int __clk_core_init(struct clk_core *core)
 	 */
 	core->hw->core = core;

-	ret = clk_pm_runtime_get(core);
-	if (ret)
-		goto unlock;
-
 	/* check to see if a clock with this name is already registered */
 	if (clk_core_lookup(core->name)) {
 		pr_debug("%s: clk %s already initialized\n",
@@ -4082,8 +4086,6 @@ static int __clk_core_init(struct clk_core *core)

 	clk_core_reparent_orphans_nolock();
 out:
-	clk_pm_runtime_put(core);
-unlock:
 	if (ret) {
 		hlist_del_init(&core->child_node);
 		core->hw->core = NULL;
@@ -4091,6 +4093,8 @@ static int __clk_core_init(struct clk_core *core)

 	clk_prepare_unlock();

+	clk_pm_runtime_put_all();
+
 	if (!ret)
 		clk_debug_register(core);

-- 
2.48.1