idr: fix a critical misallocation bug, take#2

This is retry of reverted 859ddf0
("idr: fix a critical misallocation bug") which contained two bugs.

* pa[idp->layers] should be cleared even if it's not used by
  sub_alloc() because it's used by mark idr_mark_full().

* The original condition check also assigned pa[l] to p which the new
  code didn't do thus leaving p pointing at the wrong layer.

Both problems have been fixed and the idr code has received good amount
testing using userland testing setup where simple bitmap allocator is
run parallel to verify the result of idr allocation.

The bug this patch fixes is caused by sub_alloc() optimization path
bypassing out-of-room condition check and restarting allocation loop
with starting value higher than maximum allowed value.  For detailed
description, please read commit message of 859ddf0.

Signed-off-by: Tejun Heo <[email protected]>
Based-on-patch-from: Eric Paris <[email protected]>
Reported-by: Eric Paris <[email protected]>
Tested-by: Stefan Lippers-Hollmann <[email protected]>
Tested-by: Serge Hallyn <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib/idr.c

@@ -156,10 +156,12 @@ static int sub_alloc(struct idr *idp, int *starting_id, struct idr_layer **pa)
 			id = (id | ((1 << (IDR_BITS * l)) - 1)) + 1;
 
 			/* if already at the top layer, we need to grow */
-			if (!(p = pa[l])) {
+			if (id >= 1 << (idp->layers * IDR_BITS)) {
 				*starting_id = id;
 				return IDR_NEED_TO_GROW;
 			}
+			p = pa[l];
+			BUG_ON(!p);
 
 			/* If we need to go up one layer, continue the
 			 * loop; otherwise, restart from the top.