Re: programming for preemption (was: [PATCH] 2.5.46: access

[Olaf Dietsche]

Andrew Morton <[e-mail]> writes:

> Olaf Dietsche wrote:
>> 
>> Ben Clifford <[e-mail]> writes:
>> 
>> > I still get those stack traces, though...
>> 
>> I retested with CONFIG_PREEMPT=y and now I get those stack traces,
>> too. So, it seems my code is not preempt safe.
>> 
>
> It's not that your code is unsafe with preemption.  It's just that
> CONFIG_PREEMPT=y turns on the debugging infrastructure which allows
> us to detect things like calling kmalloc(GFP_KERNEL) inside spinlock.

Thanks for this hint. So this means kmalloc(GFP_KERNEL) inside
spinlock is not necessarily dangerous, but should be avoided if
possible? Is using a semaphore better than using spinlocks? Is
there a list of dos and don'ts for preempt kernels beside
Documentation/preempt-locking.txt?

And btw, who is "us"?

Regards, Olaf.