Re: [RFC] New locking primitive for 2.5

[Martin Wirth <Martin.Wirth@dlr.de>]

Robert Love wrote:

> Some of the talk I've heard has been toward an adaptive lock.  These
> are locks like Solaris's that can spin or sleep, usually depending on
> the state of the lock's holder.  Another alternative, which I prefer
> since it is much less overhead, is a lock that spins-then-sleeps
> unconditionally.

Dave Hanson wrote:

> he spin-then-sleep lock could be interesting as a replacement for the 
> BKL in places where a semaphore causes performance degredation.  In 
> quite a few places where we replaced the BKL with a more finely grained 
> semapore (not a spinlock because we needed to sleep during the hold), 
> instead of spinning for a bit, it would schedule instead.  This was bad 
> :).  Spin-then-sleep would be great behaviour in this situation.


Wouldn't it be sufficient to include the following patch of code
at the beginning of __combi_wait(...):

        if (smp_processor_id() != owner->cpu) {
                int cnt=MAX_LOOP_CNT;
retry:
                spin_unlock(&x->wait.lock) 
                do {
                        barrier();
                while (--cnt && x->owner);
                spin_lock(&x->wait.lock);
                if (!x->owner)
                        return;
                if (cnt)
                        goto retry;
        }
       then the sleep code of __combi_wait(...)

If one fears that the owner (or current if the kernel is made
preemptible) migrated to the same cpu while we are spinning 
for x->owner and hence may
make no progress, one could let the waiting loop last about a typical
process switch time and add an outer loop that checks if the cpu
of the owner is still different.


  Martin Wirth