ARM: OMAP: clock: fix race in disable all clocks
authorNishanth Menon <nm@ti.com>
Thu, 15 Mar 2012 18:24:31 +0000 (13:24 -0500)
committerPaul Walmsley <paul@pwsan.com>
Wed, 4 Apr 2012 15:23:34 +0000 (09:23 -0600)
clk_disable_unused is invoked when CONFIG_OMAP_RESET_CLOCKS=y.
Since clk_disable_unused is called as lateinitcall, there can
be more than a few workqueues executing off secondary CPU(s).
The current code does the following:
a) checks if clk is unused
b) holds lock
c) disables clk
d) unlocks

Between (a) and (b) being executed on CPU0, It is possible to
have a driver executing on CPU1 which could do a get_sync->clk_get
(and increase the use_count) of the clock which was just about
to be disabled by clk_disable_unused.

We ensure instead that the entire list traversal is protected by
the lock allowing for parent child clock traversal which could be
potentially be done by runtime operations to be safe as well.

Reported-by: Todd Poynor <toddpoynor@google.com>
Signed-off-by: Nishanth Menon <nm@ti.com>
Signed-off-by: Paul Walmsley <paul@pwsan.com>
arch/arm/plat-omap/clock.c

index 56b6f8b..8506cbb 100644 (file)
@@ -441,6 +441,8 @@ static int __init clk_disable_unused(void)
                return 0;
 
        pr_info("clock: disabling unused clocks to save power\n");
+
+       spin_lock_irqsave(&clockfw_lock, flags);
        list_for_each_entry(ck, &clocks, node) {
                if (ck->ops == &clkops_null)
                        continue;
@@ -448,10 +450,9 @@ static int __init clk_disable_unused(void)
                if (ck->usecount > 0 || !ck->enable_reg)
                        continue;
 
-               spin_lock_irqsave(&clockfw_lock, flags);
                arch_clock->clk_disable_unused(ck);
-               spin_unlock_irqrestore(&clockfw_lock, flags);
        }
+       spin_unlock_irqrestore(&clockfw_lock, flags);
 
        return 0;
 }