BUG/MAJOR: ring: free the ring storage not the ring itself when using…

… maps

A recent issue was uncovered by the CI which started to randomly report
segfaults on a few tests, and more systematically on FreeBSD. It turn
out that it was introduced by recent commit 03816cc ("MAJOR: ring:
insert an intermediary ring_storage level"), which overlooked the munmap()
path of the sink and startup logs: once the ring and its storage were
split, it was no longer correct to munmap() the ring, only its storage
area needs to be unmapped, and the ring must always be freed separately.

Thanks to Christopher and William for their help at trying to reproduce
it and figure the circumstances that triggers it.

No backport is needed.

src/errors.c

@@ -190,10 +190,9 @@ void startup_logs_free(struct ring *r)
 {
 #ifdef USE_SHM_OPEN
 	if (r == shm_startup_logs)
-		munmap(r, STARTUP_LOG_SIZE);
-	else
+		munmap(ring_area(r), STARTUP_LOG_SIZE);
 #endif /* ! USE_SHM_OPEN */
-		ring_free(r);
+	ring_free(r);
 }
 
 /* duplicate a startup logs which was previously allocated in a shm */

src/sink.c

@@ -696,14 +696,13 @@ static void sink_free(struct sink *sink)
 	if (sink->type == SINK_TYPE_BUFFER) {
 		if (sink->store) {
 			size_t size = (ring_size(sink->ctx.ring) + 4095UL) & -4096UL;
-			void *area = (ring_area(sink->ctx.ring) - sizeof(*sink->ctx.ring));
+			void *area = ring_area(sink->ctx.ring);
 
 			msync(area, size, MS_SYNC);
 			munmap(area, size);
 			ha_free(&sink->store);
 		}
-		else
-			ring_free(sink->ctx.ring);
+		ring_free(sink->ctx.ring);
 	}
 	LIST_DEL_INIT(&sink->sink_list); // remove from parent list
 	task_destroy(sink->forward_task);