Re: [PATCH] kernel/timer.c: using spin_lock_irqsave instead of spin_lock + local_irq_save, especially when CONFIG_LOCKDEP not defined

[Chen Gang <gang.chen@asianux.com>]

On 06/19/2013 06:53 PM, Thomas Gleixner wrote:
> On Wed, 19 Jun 2013, Chen Gang wrote:
> 
>> > On 06/19/2013 05:59 PM, Thomas Gleixner wrote:
>>> > > I'm well aware how that works. And there is no difference whether you
>>> > > do:
>>> > > 
>>> > > 	local_irq_save(flags);
>>> > > 	spin_lock(&lock);
>>> > > or
>>> > > 	spin_lock_irqsave(&lock, flags);
>> > 
>> > if CONFIG_LOCKDEP is not defined, they are not semantically the same.
> Care to explain _your_ spinlock semantics to me?
> 
> The factual ones are:
> 
>     spin_lock_irqsave() returns with the lock held, interrupts and
>     preemption disabled. 
> 

Yes.

>     spin_lock() returns with the lock held, preemption disabled. It
>     does not affect interrupt disabled/enabled state
> 

Yes.

> So
> 	local_irq_save(flags);
> 	spin_lock(&lock);
> 
> is semantically the same as 
> 
> 	spin_lock_irqsave(&lock, flags);
> 

Yes (but reverse is NO).

> And this is completely independent of LOCKDEP.

NO.

 	spin_lock_irqsave(&lock, flags);

 is not semantically the same as

 	local_irq_save(flags);
 	spin_lock(&lock);

It depend on the spin_lock_irqsave() implementation, if the parameters
has no relation ship with each other, semantically the same.


Thanks.
-- 
Chen Gang

Asianux Corporation