dpdk: dpdk-input optimizations and fixes

- fix issue caused by assumption that ethertype starts at 0
- intoduce buffer templates to speed-up vlib_buffer_t metadata
initialization
- avoid check for multiseg buffers inside loop if multiseg is disabled
- interleave prefetches to reduce load on L1 cache

Change-Id: I3b76e6d3e1e15ed28f01625edb7fbe9f38112e03
Signed-off-by: Damjan Marion <damarion@cisco.com>

src/plugins/dpdk.am

@@ -16,6 +16,10 @@ vppplugins_LTLIBRARIES += dpdk_plugin.la
 
 dpdk_plugin_la_LDFLAGS = $(AM_LDFLAGS) -Wl,--whole-archive,-l:libdpdk.a,--no-whole-archive,-lm,-ldl
 
+# due to internal compiler error in GCC when compiling dpdk/device/node.c
+# debug sysmbols level reduced to 1. See GCC PR #79953 for details
+CFLAGS += -g1
+
 dpdk_plugin_la_SOURCES =					\
   dpdk/main.c							\
   dpdk/buffer.c							\

src/plugins/dpdk/device/dpdk.h

@@ -338,6 +338,9 @@ typedef struct
   /* per-thread recycle lists */
   u32 **recycle;
 
+  /* per-thread buffer templates */
+  vlib_buffer_t *buffer_templates;
+
   /* buffer flags template, configurable to enable/disable tcp / udp cksum */
   u32 buffer_flags_template;
 

src/plugins/dpdk/device/init.c

@@ -572,6 +572,21 @@ dpdk_lib_init (dpdk_main_t * dm)
     dm->buffer_flags_template &= ~(IP_BUFFER_L4_CHECKSUM_CORRECT
 				   | IP_BUFFER_L4_CHECKSUM_COMPUTED);
 
+  /* vlib_buffer_t template */
+  vec_validate_aligned (dm->buffer_templates, tm->n_vlib_mains - 1,
+			CLIB_CACHE_LINE_BYTES);
+  for (i = 0; i < tm->n_vlib_mains; i++)
+    {
+      vlib_buffer_free_list_t *fl;
+      vlib_buffer_t *bt = vec_elt_at_index (dm->buffer_templates, i);
+      fl = vlib_buffer_get_free_list (vm,
+				      VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX);
+      vlib_buffer_init_for_free_list (bt, fl);
+      bt->flags = dm->buffer_flags_template;
+      bt->current_data = -RTE_PKTMBUF_HEADROOM;
+      vnet_buffer (bt)->sw_if_index[VLIB_TX] = (u32) ~ 0;
+    }
+
   for (i = 0; i < nports; i++)
     {
       u8 addr[6];