vm locking question

[Borislav Deianov]

Hi,

For the atomic ps I'd like to scan the memory maps of all processes
without sleeping. This means I can't grab mmap_sem like the procfs
code does. I'm thinking of doing something like this:

read_lock(&tasklist_lock);
for_each_task(p) {
	...
	if (p->mm) {
		atomic_inc(&p->mm_count);
		vmlist_access_lock(p->mm);
		for (vma = p->mm->mmap; vma; vma = vma->vm_next) {
			...
		}
		vmlist_access_unlock(p->mm);
		mmdrop(p->mm);
	}
	...
}
read_unlock(&tasklist_lock);

Is this OK or will it get me in trouble (latency issues aside)?

Thanks,
Borislav