libringbuffer shm: be resilient to app crash between shm_open and shm_unlink

Leak at most a single shm if an application crashes. By unlinking the
shm before every open attempt, we are certain to never retry endlessly
in case of crash of another application between the open and unlink.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

diff --git a/libringbuffer/shm.c b/libringbuffer/shm.c
index 873e6e87597311ae9a5cc03c3d1c27a53dae025c..36f7e36b6a573dc2738b4444d92da563ed5830c9 100644 (file)
--- a/libringbuffer/shm.c
+++ b/libringbuffer/shm.c
@@ -67,8 +67,14 @@ struct shm_object *shm_object_table_append(struct shm_object_table *table,
         * We specifically do _not_ use the / at the beginning of the
         * pathname so that some OS implementations can keep it local to
         * the process (POSIX leaves this implementation-defined).
+        * Ignore the shm_unlink errors, because we handle leaks that
+        * could occur by applications crashing between shm_open and
+        * shm_unlink by unlinking the shm before every open. Therefore,
+        * we can only leak one single shm (and only if the application
+        * crashes between shm_open and the following shm_unlink).
         */
        do {
+               (void) shm_unlink("ust-shm-tmp");
                shmfd = shm_open("ust-shm-tmp",
                                 O_CREAT | O_EXCL | O_RDWR, 0700);
        } while (shmfd < 0 && errno == EEXIST);