[ros-diffs] [gdalsnes] 18419: add comment about KeCancelTimer pitfall

[gdalsnes at svn.reactos.com]

add comment about KeCancelTimer pitfall
Modified: trunk/reactos/ntoskrnl/ke/timer.c
  _____  

Modified: trunk/reactos/ntoskrnl/ke/timer.c
--- trunk/reactos/ntoskrnl/ke/timer.c	2005-10-12 22:15:22 UTC (rev
18418)
+++ trunk/reactos/ntoskrnl/ke/timer.c	2005-10-12 23:10:31 UTC (rev
18419)
@@ -34,6 +34,19 @@

  *       Timer = timer to cancel
  * RETURNS: True if the timer was running
  *          False otherwise
+ *
+ * DANGER!
+ * The statement in the DDK for KeCancelTimer that "if a DPC object is
+ * associated with the timer, it too is canceled" is wrong -- nothing
is
+ * done with the DPC object when the timer is removed from the system
+ * queue. So its very likely that the DPC will run after you have
canceled
+ * the timer!
+ * For what it's worth, calling KeRemoveQueueDpc after KeCancelTimer
would
+ * be sufficient to prevent any problems associated with destroying the
DPC
+ * object, at least as the OS is currently implemented. This is because
the
+ * DPC dispatcher doesn't need access to the object once the DPC is
+ * dequeued, and the dequeuing happens before the DPC routine gets
called."
+ * -Gunnar (article by Walter Oney)
  */
 BOOLEAN
 STDCALL
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.reactos.org/pipermail/ros-diffs/attachments/20051013/16165edb/attachment.html