2018-01-05 01:46:26

by David Lechner

Subject: [PATCH v2] clk: fix reentrancy of clk_enable() on UP systems

Reentrant calls to clk_enable() are not working on UP systems. This is
caused by the fact spin_trylock_irqsave() always returns true when
CONFIG_SMP=n (and CONFIG_DEBUG_SPINLOCK=n) which causes the reference
counting to not work correctly when clk_enable_lock() is called twice
before clk_enable_unlock() is called (this happens when clk_enable()
is called from within another clk_enable()).

This fixes the problem by skipping the call to spin_trylock_irqsave() on UP
systems and relying solely on reference counting. We also make sure to set
flags in this case so that we are not returning an uninitialized value.

Suggested-by: Stephen Boyd <[email protected]>
Signed-off-by: David Lechner <[email protected]>
---

v2 changes:
* Use fix suggested by Stephen Boyd

drivers/clk/clk.c | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c
index 83c8df7..5add75b 100644
--- a/drivers/clk/clk.c
+++ b/drivers/clk/clk.c
@@ -143,10 +143,18 @@ static unsigned long clk_enable_lock(void)
{
unsigned long flags;

- if (!spin_trylock_irqsave(&enable_lock, flags)) {
+ /*
+ * On UP systems, spin_trylock_irqsave() always returns true, even if
+ * we already hold the lock. So, in that case, we rely only on
+ * reference counting.
+ */
+ if (!IS_ENABLED(CONFIG_SMP) ||
+ !spin_trylock_irqsave(&enable_lock, flags)) {
if (enable_owner == current) {
enable_refcnt++;
__acquire(enable_lock);
+ if (!IS_ENABLED(CONFIG_SMP))
+ local_save_flags(flags);
return flags;
}
spin_lock_irqsave(&enable_lock, flags);
--
2.7.4

2018-01-10 21:40:48

by Stephen Boyd

Subject: Re: [PATCH v2] clk: fix reentrancy of clk_enable() on UP systems

On 01/04, David Lechner wrote:
> Reentrant calls to clk_enable() are not working on UP systems. This is
> caused by the fact spin_trylock_irqsave() always returns true when
> CONFIG_SMP=n (and CONFIG_DEBUG_SPINLOCK=n) which causes the reference
> counting to not work correctly when clk_enable_lock() is called twice
> before clk_enable_unlock() is called (this happens when clk_enable()
> is called from within another clk_enable()).
>
> This fixes the problem by skipping the call to spin_trylock_irqsave() on UP
> systems and relying solely on reference counting. We also make sure to set
> flags in this case so that we are not returning an uninitialized value.
>
> Suggested-by: Stephen Boyd <[email protected]>
> Signed-off-by: David Lechner <[email protected]>
> ---

Applied to clk-next

--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project