What do you actually save? The longjmp() to the coroutine code,
linking in to the mutex wait queue, and another switch back to the
main coroutine? Given that we don't expect to block often, it seems
hardly a cost worth optimizing.
It's a matter of correctness, not optimization.
Consider the following example:
coroutine {
l2 = find_l2(offset);
// allocates but does not increment max cluster offset
l2[l2_offset(offset)] = alloc_cluster();
co_mutex_lock(&lock);
write_l2(l2);
co_mutex_unlock(&lock);
l1[l1_offset(offset)] = l2;
co_mutex_lock(&lock);
write_l1(l1);
co_mutex_unlock(&lock);
commit_cluster(l2[l2_offset(offset)]);
}
This code is incorrect.