Re: [question] why not get rid of handle_softirq from schedule()?

[Manfred Spraul]

[e-mail] wrote:
> 
> Hello!
> 
> > If this is correct, then why do we need to call do_softirq() in
> > schedule() at all - isn't it enough to just call it on every
> > interrupt/syscall return?
> 
> OK, let's go further. Why to call it on return from syscall?
> 

We would have to check for pending softirq's in local_bh_enable().

* user space
	syscall
* within kernel space:
	spin_lock_bh();
* interrupt arrives
* within interrupt: mark_bh
* interrupt returns, but can't handle softirq now
	...
	spin_unlock_bh();
	...
	iretd	

--
	Manfred