Unix domain: Enable logic to clean up the FIFOs underlying stream sockets with those sockets are disconnected. Tehre is still no corresponding clean-up logic in place for Unix domain datagram sockets because the life of the FIFO is not as well known in that case

2 files changed, 48 insertions, 28 deletions

diff --git a/nuttx/TODO b/nuttx/TODO
index 14dcaf1f5..545420d5b 100644
--- a/nuttx/TODO
+++ b/nuttx/TODO
@@ -826,13 +826,12 @@ o Network (net/, drivers/net)
 
   Title:       SOCKETS DO NOT ALWAYS SUPPORT O_NONBLOCK
   Description: sockets do not support all modes for O_NONBLOCK. Sockets
-               support only (1) TCP/IP non-blocking read operations when read-ahead
-               buffering is enabled, (2) TCP/IP accept() operations when TCP/IP
-               connection backlog is enabled, and (3) nonblocking operations on
-               Unix domain sockets.
-               REVISIT:  UDP read-ahead buffering has been implemented.  But I
-               do not think that the necessary bits of logic are in place to
-               permit non-clocking UDP reads.
+               support nonblocking operations only (1) for TCP/IP non-
+               blocking read operations when read-ahead buffering is
+               enabled, (2) TCP/IP accept() operations when TCP/IP
+               connection backlog is enabled, (2) UDP/IP read() operations
+               when UDP read-ahead is enabled, and (3) non-blocking
+               operations on Unix domain sockets.
   Status:      Open
   Priority:    Low.
 
@@ -923,20 +922,23 @@ o Network (net/, drivers/net)
   Status:      Open
   Priority:    Low, this is just a nuisance in most cases.
 
-  Title:       FIFO CLEAN-UP AFTER CLOSING UNIX DOMAIN SOCKET
-  Description: FIFOs are used as the IPC underlying local Unix domain sockets.
-               In NuttX, FIFOs are implemented as device drivers (not as
-               special files).  The FIFO device driver is instantiated when
-               the Unix domain socket communications begin and will
-               automatically be released when (1) the driver is unlinked and
-               (2) all open references to the driver have been closed.  But
-               there is no mechanism in place now to unline the FIFO when
-               the Unix domain socket is no longer used.  The primary issue
-               is timing.. the FIFO should persist until it is no longer
-               needed.  Perhaps there should be a delayed call to unlink()
-               (using a watchdog or the work queue).  If the driver is re-
-               opened, the delayed unlink could be canceled?  Needs more
-               thought.
+  Title:       FIFO CLEAN-UP AFTER CLOSING UNIX DOMAIN DATAGRAM SOCKET
+  Description: FIFOs are used as the IPC underlying all local Unix domain
+               sockets.  In NuttX, FIFOs are implemented as device drivers
+               (not as a special FIFO files).  The FIFO device driver is
+               instantiated when the Unix domain socket communications begin
+               and will automatically be released when (1) the driver is
+               unlinked and (2) all open references to the driver have been
+               closed.  But there is no mechanism in place now to unlink the
+               FIFO when the Unix domain datagram socket is no longer used.
+               The primary issue is timing.. the FIFO should persist until
+               it is no longer needed.  Perhaps there should be a delayed
+               call to unlink() (using a watchdog or the work queue).  If
+               the driver is re-opened, the delayed unlink could be
+               cancelled?  Needs more thought.
+               NOTE: This is not an issue for Unix domain streams sockets:
+               The end-of-life of the FIFO is well determined when sockets
+               are disconnected and support for that case is fully implemented.
   Status:      Open
   Priority:    Low for now because I don't have a situation where this is a
                problem for me.  If you use the same Unix domain paths, then
diff --git a/nuttx/net/local/local_fifo.c b/nuttx/net/local/local_fifo.c
index 42a59fa4c..cb94f7519 100644
--- a/nuttx/net/local/local_fifo.c
+++ b/nuttx/net/local/local_fifo.c
@@ -193,17 +193,17 @@ static int local_create_fifo(FAR const char *path)
 
 static int local_release_fifo(FAR const char *path)
 {
+  int ret;
+
   /* Unlink the client-to-server FIFO if it exists. */
 
   if (local_fifo_exists(path))
     {
-      /* REVISIT:  This is wrong!  Un-linking the FIFO does not eliminate it;
-       * it only removes it from the namespace.  A new interface will be
-       * required to destroy the FIFO driver instance and all of its resources.
+      /* Un-linking the FIFO removes the FIFO from the namespace.  It will
+       * also mark the FIFO device "unlinked".  When all of the open
+       * references to the FIFO device are closed, the resources consumed
+       * by the device instance will also be freed.
        */
-#warning Missing logic
-#if 0
-      int ret;
 
       ret = unlink(path);
       if (ret < 0)
@@ -214,7 +214,6 @@ static int local_release_fifo(FAR const char *path)
           ndbg("ERROR: Failed to unlink FIFO %s: %d\n", path, errcode);
           return -errcode;
         }
-#endif
     }
 
   /* The FIFO does not exist or we successfully unlinked it. */
@@ -440,6 +439,24 @@ int local_release_halfduplex(FAR struct local_conn_s *conn)
 #ifdef CONFIG_NET_LOCAL_STREAM
 int local_open_client_rx(FAR struct local_conn_s *client, bool nonblock)
 {
+#if 1
+  /* REVIST: We need to think about this carefully.  Unlike the connection-
+   * oriented Unix domain socket, we don't really know the best time to
+   * release the FIFO resource.  It would be extremely inefficient to create
+   * and destroy the FIFO on each packet. But, on the other hand, failing
+   * to destory the FIFO will leave the FIFO resources in place after the
+   * communications have completed.
+   *
+   * I am thinking that ther should be something like a timer.  The timer
+   * would be started at the completion of each transfer and cancelled at
+   * the beginning of each transfer.  If the timer expires, then the FIFO
+   * would be destroyed.
+   */
+
+#  warning Missing logic
+  return OK;
+
+#else
   char path[LOCAL_FULLPATH_LEN];
   int ret;
 
@@ -458,6 +475,7 @@ int local_open_client_rx(FAR struct local_conn_s *client, bool nonblock)
     }
 
   return ret;
+#endif
 }
 #endif /* CONFIG_NET_LOCAL_STREAM */