Clean up mentions of mux tile

... which is now fd_stem

Author: riptl

contrib/test/run_script_tests.sh

@@ -51,17 +51,6 @@ fi
 # FIXME: Needs a /tmp/test.pcap file
 #$UNIT_TEST/test_replay         --tile-cpus 38-42/2 --tx-pcap /tmp/test.pcap 2> $LOG_PATH/replay
 
-
-# todo(mmcgee-jump): reenable those tests https://github.com/firedancer-io/firedancer/issues/761
-# if $UNIT_TEST/test_mux_ipc_init $OBJDIR && \
-#     $UNIT_TEST/test_mux_ipc_meta 16 16  && \
-#     $UNIT_TEST/test_mux_ipc_full 16 16  && \
-#     $UNIT_TEST/test_mux_ipc_fini; then
-#   echo pass > $LOG_PATH/mux_ipc
-# else
-#   echo FAIL > $LOG_PATH/mux_ipc
-# fi
-
 wait
 
 for f in `ls $LOG_PATH`; do

src/app/fdctl/topology.c

@@ -191,7 +191,7 @@ fd_topo_initialize( config_t * config ) {
 
   /* Sign links don't need to be reliable because they are synchronous,
      so there's at most one fragment in flight at a time anyway.  The
-     sign links are also not polled by the mux, instead the tiles will
+     sign links are also not polled by fd_stem, instead the tiles will
      read the sign responses out of band in a dedicated spin loop. */
 
   for( ulong i=0UL; i<shred_tile_cnt; i++ ) {

src/app/firedancer/topology.c

@@ -667,7 +667,7 @@ fd_topo_initialize( config_t * config ) {
 
   /* Sign links don't need to be reliable because they are synchronous,
     so there's at most one fragment in flight at a time anyway.  The
-    sign links are also not polled by the mux, instead the tiles will
+    sign links are also not polled by fd_stem, instead the tiles will
     read the sign responses out of band in a dedicated spin loop. */
   for( ulong i=0UL; i<shred_tile_cnt; i++ ) {
     /**/               fd_topob_tile_in(  topo, "sign",   0UL,           "metric_in", "shred_sign",    i,            FD_TOPOB_UNRELIABLE, FD_TOPOB_POLLED   );

src/app/shared/commands/monitor/monitor.c

@@ -123,10 +123,10 @@ find_producer_out_idx( fd_topo_t const *      topo,
      and then returns the position of the consumer (specified by tile
      and index of the in of that tile) in that list. The list ordering
      is not important, except that it matches the ordering of fseqs
-     provided to the mux tile, so that metrics written for each link
-     index are retrieved at the same index here.
+     provided to fd_stem, so that metrics written for each link index
+     are retrieved at the same index here.
 
-     This is why we only count reliable links, because the mux tile only
+     This is why we only count reliable links, because fd_stem only
      looks at and writes producer side diagnostics (is the link slow)
      for reliable links. */
 

src/disco/bundle/fd_bundle_tile_private.h

@@ -41,8 +41,8 @@ struct fd_bundle_metrics {
 
 typedef struct fd_bundle_metrics fd_bundle_metrics_t;
 
-/* fd_bundle_tile_t is the context object provided to callbacks from the
-   mux tile, and contains all state needed to progress the tile. */
+/* fd_bundle_tile_t is the context object provided to callbacks from
+   stem, and contains all state needed to progress the tile. */
 
 struct fd_bundle_tile {
   /* Key switch */

src/disco/cswtch/fd_cswtch_tile.c

@@ -36,9 +36,9 @@ scratch_footprint( fd_topo_tile_t const * tile ) {
 
 static void
 before_credit( fd_cswtch_ctx_t *   ctx,
-               fd_stem_context_t * mux,
+               fd_stem_context_t * stem,
                int *               charge_busy ) {
-  (void)mux;
+  (void)stem;
 
   long now = fd_log_wallclock();
   if( now<ctx->next_report_nanos ) {

src/disco/dedup/fd_dedup_tile.c

@@ -10,11 +10,7 @@
 /* fd_dedup provides services to deduplicate multiple streams of input
    fragments and present them to a mix of reliable and unreliable
    consumers as though they were generated by a single multi-stream
-   producer.
-
-   The dedup tile is simply a wrapper around the mux tile, that also
-   checks the transaction signature field for duplicates and filters
-   them out. */
+   producer. */
 
 #define IN_KIND_GOSSIP       (0UL)
 #define IN_KIND_VERIFY       (1UL)
@@ -30,8 +26,8 @@ typedef struct {
   ulong       mtu;
 } fd_dedup_in_ctx_t;
 
-/* fd_dedup_ctx_t is the context object provided to callbacks from the
-   mux tile, and contains all state needed to progress the tile. */
+/* fd_dedup_ctx_t is the context object provided to callbacks from
+   fd_stem, and contains all state needed to progress the tile. */
 
 typedef struct {
   ulong   tcache_depth;   /* == fd_tcache_depth( tcache ), depth of this dedups's tcache (const) */

src/disco/metrics/fd_metrics.h

@@ -45,8 +45,8 @@
 
    where every value is a ulong.  Tile metrics come after link metrics,
    so this base pointer points at the very start of the layout.  You
-   shouldn't need to use this directly, instead it's used by the mux
-   tile when it's computing the metrics for specific links. */
+   shouldn't need to use this directly, instead it's used by fd_stem
+   when it's computing the metrics for specific links. */
 extern FD_TL ulong * fd_metrics_base_tl;
 
 /* All metrics in the application are ulongs, and are laid out

src/disco/quic/fd_quic_tile.c

@@ -11,22 +11,20 @@
 #include <linux/unistd.h>
 #include <sys/random.h>
 
-/* fd_quic provides a TPU server tile.
+/* fd_quic_tile provides a TPU server tile.
 
    This tile handles incoming transactions that clients request to be
    included in blocks.  Supported protocols currently include TPU/UDP
    and TPU/QUIC.
 
-   The fd_quic tile acts as a plain old Tango producer writing to a cnc
-   and an mcache.  The tile will defragment multi-packet TPU streams
-   coming in from QUIC, such that each mcache/dcache pair forms a
-   complete txn.  This requires the dcache mtu to be at least that of
-   the largest allowed serialized txn size.
+   The fd_quic tile acts as a plain old Tango producer.  The tile will
+   defragment multi-packet TPU streams coming in from QUIC, such tha
+   each frag_meta refers to a complete txn.  This requires the dcache
+   mtu to be at least that of the largest allowed serialized txn size.
 
-   QUIC tiles don't service network devices directly, but rely on
-   packets being received by net tiles and forwarded on via. a mux
-   (multiplexer).  An arbitrary number of QUIC tiles can be run.  Each
-   UDP flow must stick to one QUIC tile. */
+   QUIC tiles don't service network devices directly, but rely on net
+   tiles to send and receive packets.  An arbitrary number of QUIC tiles
+   can be run.  Each UDP flow must stick to one QUIC tile. */
 
 static inline fd_quic_limits_t
 quic_limits( fd_topo_tile_t const * tile ) {
@@ -92,7 +90,7 @@ legacy_stream_notify( fd_quic_ctx_t * ctx,
 }
 
 /* Because of the separate mcache for publishing network fragments
-   back to networking tiles, which is not managed by the mux, we
+   back to networking tiles, which is not managed by fd_stem, we
    need to periodically update the sync. */
 static void
 during_housekeeping( fd_quic_ctx_t * ctx ) {

src/disco/topo/fd_topob.h

@@ -95,10 +95,9 @@ fd_topob_tile( fd_topo_t *    topo,
                int            is_agave,
                int            uses_keyswitch );
 
-/* Add an input link to the tile.  If the tile is created with the
-   standard mux runner, it will automatically poll the in link and
-   forward fragments to the user code (unless the link is specified
-   as unpolled).
+/* Add an input link to the tile.  If the tile is created with fd_stem,
+   it will automatically poll the in link and forward fragments to the
+   user code (unless the link is specified as unpolled).
 
    An input link has an fseq which is a ulong used for returning the
    current reader position in sequence space, used for wiring flow