Re: kmemleak: Protect the seq start/next/stop sequence by rcu_read_lock()

[Catalin Marinas <catalin.marinas@arm.com>]

On Wed, 2009-08-12 at 08:32 -0700, Linus Torvalds wrote:
> 
> On Wed, 12 Aug 2009, Catalin Marinas wrote:
> > 
> > If the object size is bigger than a predefined value (4K in this case),
> > release the object lock during scanning and call cond_resched().
> 
> What guarantees that the object isn't released during this?

There is no guarantee but the while() loop rechecks that object->flags
still has the OBJECT_ALLOCATED bit set after every cond_resched():

+               while (start < end && (object->flags & OBJECT_ALLOCATED)) {
+                       scan_block(start, min(start + MAX_SCAN_SIZE, end),
+                                  object, 0);
+                       start += MAX_SCAN_SIZE;
+
+                       spin_unlock_irqrestore(&object->lock, flags);
+                       cond_resched();
+                       spin_lock_irqsave(&object->lock, flags);
+               }

It seems to be working fine with LTP running in parallel with kmemleak
scanning for nearly a day.

-- 
Catalin