It seems I will never get my OS to run on a smp system, because of the many race conditions

I attached my scheduler code. It seems to be a race condition when my semaphore code calls the function "scheduler_add_scheduler". I get a page fault because ebx (lastThread) is NULL. But it can´t be NULL when the code works right. If here is someone who has time and could have a look at my code, maybe he will see something which wont work on a smp system!

Please can you describe the algo your using for protecting resources from concurrent use ?

SMP issues typically stem from thinking a single lock is going to protect a resource, when in fact more than one is needed in some circumstances.

I use a spinlock to protect the priority queue for the threads. I think that it could be that when cpu1 has the semaphore and wakes up a thread that cpu0 hasn´t finished with that thread. But for this there is also a spinlock for every thread that the scheduler takes.