From: Chris Darroch <chrisd@apache.org>
Date: Tue, 4 Mar 2008 20:03:51 +0000 (+0000)
Subject: add comment from event MPM's fdqueue ap_pop_pool() regarding
X-Git-Tag: 2.3.0~908
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=46ccf01ae7e1caadf26818afe2c212c9ccc8938d;p=apache

add comment from event MPM's fdqueue ap_pop_pool() regarding
ABA problem and avoidance because only caller is single listener thread


git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/trunk@633612 13f79535-47bb-0310-9956-ffa450edef68
---

diff --git a/server/mpm/worker/fdqueue.c b/server/mpm/worker/fdqueue.c
index 8a75d24d1e..d46dd5361b 100644
--- a/server/mpm/worker/fdqueue.c
+++ b/server/mpm/worker/fdqueue.c
@@ -188,6 +188,14 @@ apr_status_t ap_queue_info_wait_for_idler(fd_queue_info_t *queue_info,
     apr_atomic_dec32(&(queue_info->idlers));
 
     /* Atomically pop a pool from the recycled list */
+
+    /* This function is safe only as long as it is single threaded because
+     * it reaches into the queue and accesses "next" which can change.
+     * We are OK today because it is only called from the listener thread.
+     * cas-based pushes do not have the same limitation - any number can
+     * happen concurrently with a single cas-based pop.
+     */
+
     for (;;) {
         struct recycled_pool *first_pool = queue_info->recycled_pools;
         if (first_pool == NULL) {