[PATCH] ARM: SAMSUNG: Bug fix spin_lock recursion in clk_enable() and clk_disable()

[PATCH] ARM: SAMSUNG: Bug fix spin_lock recursion in clk_enable() and clk_disable()

[Jaecheol Lee]

From: Minho Ban <mhban@samsung.com>

The clk_enable() and clk_disable() can be used process and ISR either.
So spin_lock_irqsave should be used instead.

Signed-off-by: Minho Ban <mhban@samsung.com>
Signed-off-by: Jaecheol Lee <jc.lee@samsung.com>
---
 arch/arm/plat-samsung/clock.c |   12 ++++++++----
 1 files changed, 8 insertions(+), 4 deletions(-)

diff --git a/arch/arm/plat-samsung/clock.c b/arch/arm/plat-samsung/clock.c
index e8d20b0..2a991a5 100644
--- a/arch/arm/plat-samsung/clock.c
+++ b/arch/arm/plat-samsung/clock.c
@@ -138,31 +138,35 @@ void clk_put(struct clk *clk)
 
 int clk_enable(struct clk *clk)
 {
+	unsigned long flags;
+
 	if (IS_ERR(clk) || clk == NULL)
 		return -EINVAL;
 
 	clk_enable(clk->parent);
 
-	spin_lock(&clocks_lock);
+	spin_lock_irqsave(&clocks_lock, flags);
 
 	if ((clk->usage++) == 0)
 		(clk->enable)(clk, 1);
 
-	spin_unlock(&clocks_lock);
+	spin_unlock_irqrestore(&clocks_lock, flags);
 	return 0;
 }
 
 void clk_disable(struct clk *clk)
 {
+	unsigned long flags;
+
 	if (IS_ERR(clk) || clk == NULL)
 		return;
 
-	spin_lock(&clocks_lock);
+	spin_lock_irqsave(&clocks_lock, flags);
 
 	if ((--clk->usage) == 0)
 		(clk->enable)(clk, 0);
 
-	spin_unlock(&clocks_lock);
+	spin_unlock_irqrestore(&clocks_lock, flags);
 	clk_disable(clk->parent);
 }
 
-- 
1.6.2.5

[PATCH] ARM: SAMSUNG: Bug fix spin_lock recursion in clk_enable() and clk_disable()

[Ben Dooks]

On 12/10/10 01:39, Jaecheol Lee wrote:
> From: Minho Ban <mhban@samsung.com>
> 
> The clk_enable() and clk_disable() can be used process and ISR either.
> So spin_lock_irqsave should be used instead.
> 
> Signed-off-by: Minho Ban <mhban@samsung.com>
> Signed-off-by: Jaecheol Lee <jc.lee@samsung.com>
> ---
>  arch/arm/plat-samsung/clock.c |   12 ++++++++----
>  1 files changed, 8 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/arm/plat-samsung/clock.c b/arch/arm/plat-samsung/clock.c
> index e8d20b0..2a991a5 100644
> --- a/arch/arm/plat-samsung/clock.c
> +++ b/arch/arm/plat-samsung/clock.c
> @@ -138,31 +138,35 @@ void clk_put(struct clk *clk)
>  
>  int clk_enable(struct clk *clk)
>  {
> +	unsigned long flags;
> +
>  	if (IS_ERR(clk) || clk == NULL)
>  		return -EINVAL;
>  
>  	clk_enable(clk->parent);
>  
> -	spin_lock(&clocks_lock);
> +	spin_lock_irqsave(&clocks_lock, flags);
>  
>  	if ((clk->usage++) == 0)
>  		(clk->enable)(clk, 1);
>  
> -	spin_unlock(&clocks_lock);
> +	spin_unlock_irqrestore(&clocks_lock, flags);
>  	return 0;
>  }
>  
>  void clk_disable(struct clk *clk)
>  {
> +	unsigned long flags;
> +
>  	if (IS_ERR(clk) || clk == NULL)
>  		return;
>  
> -	spin_lock(&clocks_lock);
> +	spin_lock_irqsave(&clocks_lock, flags);
>  
>  	if ((--clk->usage) == 0)
>  		(clk->enable)(clk, 0);
>  
> -	spin_unlock(&clocks_lock);
> +	spin_unlock_irqrestore(&clocks_lock, flags);
>  	clk_disable(clk->parent);

I'm not sure, but I don't belive that the clk_ api has
ever been callable from non-sleepable contexts such as
interrupt handlers.

I would welcome RMK's response (or any other response)
about this issue?

Personally, given that some clk_ calls may sleep esp.
for pll enables, I would prefer to see this patch
dropped in favour of fixing the users.

-- 
Ben

[PATCH] ARM: SAMSUNG: Bug fix spin_lock recursion in clk_enable() and clk_disable()

[Jassi Brar]

On Tue, Oct 19, 2010 at 8:09 AM, Ben Dooks <ben-linux@fluff.org> wrote:
> On 12/10/10 01:39, Jaecheol Lee wrote:
>> From: Minho Ban <mhban@samsung.com>
>>
>> The clk_enable() and clk_disable() can be used process and ISR either.
>> So spin_lock_irqsave should be used instead.
>>
>> Signed-off-by: Minho Ban <mhban@samsung.com>
>> Signed-off-by: Jaecheol Lee <jc.lee@samsung.com>
>> ---
>> ?arch/arm/plat-samsung/clock.c | ? 12 ++++++++----
>> ?1 files changed, 8 insertions(+), 4 deletions(-)
>>
>> diff --git a/arch/arm/plat-samsung/clock.c b/arch/arm/plat-samsung/clock.c
>> index e8d20b0..2a991a5 100644
>> --- a/arch/arm/plat-samsung/clock.c
>> +++ b/arch/arm/plat-samsung/clock.c
>> @@ -138,31 +138,35 @@ void clk_put(struct clk *clk)
>>
>> ?int clk_enable(struct clk *clk)
>> ?{
>> + ? ? unsigned long flags;
>> +
>> ? ? ? if (IS_ERR(clk) || clk == NULL)
>> ? ? ? ? ? ? ? return -EINVAL;
>>
>> ? ? ? clk_enable(clk->parent);
>>
>> - ? ? spin_lock(&clocks_lock);
>> + ? ? spin_lock_irqsave(&clocks_lock, flags);
>>
>> ? ? ? if ((clk->usage++) == 0)
>> ? ? ? ? ? ? ? (clk->enable)(clk, 1);
>>
>> - ? ? spin_unlock(&clocks_lock);
>> + ? ? spin_unlock_irqrestore(&clocks_lock, flags);
>> ? ? ? return 0;
>> ?}
>>
>> ?void clk_disable(struct clk *clk)
>> ?{
>> + ? ? unsigned long flags;
>> +
>> ? ? ? if (IS_ERR(clk) || clk == NULL)
>> ? ? ? ? ? ? ? return;
>>
>> - ? ? spin_lock(&clocks_lock);
>> + ? ? spin_lock_irqsave(&clocks_lock, flags);
>>
>> ? ? ? if ((--clk->usage) == 0)
>> ? ? ? ? ? ? ? (clk->enable)(clk, 0);
>>
>> - ? ? spin_unlock(&clocks_lock);
>> + ? ? spin_unlock_irqrestore(&clocks_lock, flags);
>> ? ? ? clk_disable(clk->parent);
>
> I'm not sure, but I don't belive that the clk_ api has
> ever been callable from non-sleepable contexts such as
> interrupt handlers.
>
> I would welcome RMK's response (or any other response)
> about this issue?
>
> Personally, given that some clk_ calls may sleep esp.
> for pll enables, I would prefer to see this patch
> dropped in favour of fixing the users.

I tend to agree.
I would like to know any such case where enabling/disabling of some
clock is so urgent as to require doing from ISR and not from a tasklet
scheduled from ISR.
Also let us not make it impossible in future to switch to common clock api
by Jeremy Kerr, that doesn't allow calls from IRQ context.

[PATCH] ARM: SAMSUNG: Bug fix spin_lock recursion in clk_enable() and clk_disable()

[mhban]

On Tue, 2010-10-19 at 10:12 +0900, Jassi Brar wrote:
> On Tue, Oct 19, 2010 at 8:09 AM, Ben Dooks <ben-linux@fluff.org> wrote:
> > On 12/10/10 01:39, Jaecheol Lee wrote:
> >> From: Minho Ban <mhban@samsung.com>
> >>
> >> The clk_enable() and clk_disable() can be used process and ISR either.
> >> So spin_lock_irqsave should be used instead.
> >>
> >> Signed-off-by: Minho Ban <mhban@samsung.com>
> >> Signed-off-by: Jaecheol Lee <jc.lee@samsung.com>
> >> ---
> >>  arch/arm/plat-samsung/clock.c |   12 ++++++++----
> >>  1 files changed, 8 insertions(+), 4 deletions(-)
> >>
> >> diff --git a/arch/arm/plat-samsung/clock.c b/arch/arm/plat-samsung/clock.c
> >> index e8d20b0..2a991a5 100644
> >> --- a/arch/arm/plat-samsung/clock.c
> >> +++ b/arch/arm/plat-samsung/clock.c
> >> @@ -138,31 +138,35 @@ void clk_put(struct clk *clk)
> >>
> >>  int clk_enable(struct clk *clk)
> >>  {
> >> +     unsigned long flags;
> >> +
> >>       if (IS_ERR(clk) || clk == NULL)
> >>               return -EINVAL;
> >>
> >>       clk_enable(clk->parent);
> >>
> >> -     spin_lock(&clocks_lock);
> >> +     spin_lock_irqsave(&clocks_lock, flags);
> >>
> >>       if ((clk->usage++) == 0)
> >>               (clk->enable)(clk, 1);
> >>
> >> -     spin_unlock(&clocks_lock);
> >> +     spin_unlock_irqrestore(&clocks_lock, flags);
> >>       return 0;
> >>  }
> >>
> >>  void clk_disable(struct clk *clk)
> >>  {
> >> +     unsigned long flags;
> >> +
> >>       if (IS_ERR(clk) || clk == NULL)
> >>               return;
> >>
> >> -     spin_lock(&clocks_lock);
> >> +     spin_lock_irqsave(&clocks_lock, flags);
> >>
> >>       if ((--clk->usage) == 0)
> >>               (clk->enable)(clk, 0);
> >>
> >> -     spin_unlock(&clocks_lock);
> >> +     spin_unlock_irqrestore(&clocks_lock, flags);
> >>       clk_disable(clk->parent);
> >
> > I'm not sure, but I don't belive that the clk_ api has
> > ever been callable from non-sleepable contexts such as
> > interrupt handlers.
> >
> > I would welcome RMK's response (or any other response)
> > about this issue?
> >
> > Personally, given that some clk_ calls may sleep esp.
> > for pll enables, I would prefer to see this patch
> > dropped in favour of fixing the users.
> 
> I tend to agree.
> I would like to know any such case where enabling/disabling of some
> clock is so urgent as to require doing from ISR and not from a tasklet
> scheduled from ISR.
> Also let us not make it impossible in future to switch to common clock api
> by Jeremy Kerr, that doesn't allow calls from IRQ context.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-samsung-soc" in
> the body of a message to majordomo at vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Hello,

Mostly you are right but it would require some explicit comment in clk.h
(eg, clk_get/put) not to be confusing to the driver developer.

Thanks.

Regards,
Minho Ban

-- 
Minho Ban (mhban at samsung.com)
Senior Engineer
Platform R&D Team,  Mobile Comm.
SAMSUNG ELECTRONICS Co.Ltd.

[PATCH] ARM: SAMSUNG: Bug fix spin_lock recursion in clk_enable() and clk_disable()

[Russell King - ARM Linux]

On Tue, Oct 19, 2010 at 12:09:25AM +0100, Ben Dooks wrote:
> I'm not sure, but I don't belive that the clk_ api has
> ever been callable from non-sleepable contexts such as
> interrupt handlers.
> 
> I would welcome RMK's response (or any other response)
> about this issue?

We have traditionally allowed it from non-sleepable contexts, mainly
because the platforms I've been using it with have had such simple
clock implementations that it's not worth artifically restricting it,
or the implementation only requires a few us delay after enabling a
clock.

[PATCH] ARM: SAMSUNG: Bug fix spin_lock recursion in clk_enable() and clk_disable()

[Ben Dooks]

On 19/10/10 19:57, Russell King - ARM Linux wrote:
> On Tue, Oct 19, 2010 at 12:09:25AM +0100, Ben Dooks wrote:
>> I'm not sure, but I don't belive that the clk_ api has
>> ever been callable from non-sleepable contexts such as
>> interrupt handlers.
>>
>> I would welcome RMK's response (or any other response)
>> about this issue?
> 
> We have traditionally allowed it from non-sleepable contexts, mainly
> because the platforms I've been using it with have had such simple
> clock implementations that it's not worth artifically restricting it,
> or the implementation only requires a few us delay after enabling a
> clock.

Some on chip PLLs can be 150-300uS to get running.

-- 
Ben