Re: [patch] preempt-smp.patch, 2.6.8-rc3-mm2

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

Hi Ingo,

from the preempt-smp patch:

@@ -306,6 +306,21 @@ static int invalidate_list(struct list_h
 		struct list_head * tmp = next;
 		struct inode * inode;
 
+		/*
+		 * Preempt if necessary. To make this safe we use a dummy
+		 * inode as a marker - we can continue off that point.
+		 * We rely on this sb's inodes (including the marker) not
+		 * getting reordered within the list during umount. Other
+		 * inodes might get reordered.
+		 */
+		if (need_resched_lock()) {
+			list_add_tail(mark, next);
+			spin_unlock(&inode_lock);
+			cond_resched();
+			spin_lock(&inode_lock);
+			tmp = next = mark->next;
+			list_del(mark);
+		}
 		next = next->next;
 		if (tmp == head)
 			break;


why use cond_resched in the loop if you use need_resched_lock in the condition?
cond_resched does not do the cpu_relax. Nor is it quite nice to use 
cond_resched_lock there since it would increment preempt_check_count again
causing the step to be 2 which in turn will make one miss the cpu_relax condition.

Peter Zijlstra