changes for optization

diff --git a/ti/runtime/netapi/netapi.h b/ti/runtime/netapi/netapi.h
index fdce74ccc3be2e3220463fbed1cc3a353deb3230..d1dd141bad5cda56fe06e6fc4ba491136285e074 100755 (executable)
--- a/ti/runtime/netapi/netapi.h
+++ b/ti/runtime/netapi/netapi.h
@@ -102,7 +102,22 @@ typedef void * NETAPI_T;
 \r
 #define NETAPI_SYS_MASTER  2  //master for system\r
 #define NETAPI_CORE_MASTER 1  //master for core\r
-#define NETAPI_NO_MASTE    0  //data only\r
+#define NETAPI_NO_MASTER   0  //data only\r
+\r
+/***********************************************\r
+*************RUN TIME CONTROLS*****************\r
+***********************************************/\r
+typedef struct NETAPI_CFG_Tag\r
+{\r
+ int def_mem_size;             //bytes of CMA memory we have allocated \r
+ int def_flow_pkt_rx_offset;   //offset in pkt buffer for hw to start RX\r
+ int def_max_descriptors;      //# of descriptors in system  (must be power of 2), 2^14 max\r
+ int def_tot_descriptors_for_us;  //#of descriptors to create in our region (must be power of 2)\r
+ int def_heap_n_descriptors;   //# descriptors+buffers in default heap\r
+ int def_heap_n_zdescriptors;  //# zero len descriptors in defaule heap\r
+ int def_heap_buf_size;        //size of buffers in default heap\r
+}NETAPI_CFG_T;\r
+\r
 \r
 \r
 #include "netapi_err.h"\r
@@ -125,7 +140,7 @@ typedef void * NETAPI_T;
  **********BUILD TIME CONTROLS *****************\r
  ***********************************************/\r
 /* see netapi_tune.h */\r
- \r
+\r
 \r
 /*************************************\r
  **************NETAPI****************\r
@@ -145,10 +160,11 @@ typedef void * NETAPI_T;
  *     - cppi\r
  *     - nwal\r
  *  @param[in]  master mode: NETAPI_SYS_MASTER or NETAPI_NO_MASTER\r
+ *  @param[in]  configuration (master mode).  pointer to NETAPI_CFG_T above or NULL\r
  *  @retval     @ref NETAPI_T: handle to the instance or NULL on error \r
  *  @pre        none \r
  */\r
-NETAPI_T  netapi_init(int master);\r
+NETAPI_T  netapi_init(int master, NETAPI_CFG_T * p_cfg);\r
 \r
 \r
 /**  @ingroup netapi_api_functions */\r

diff --git a/ti/runtime/netapi/netapi_timer.h b/ti/runtime/netapi/netapi_timer.h
index f8ed7ba32fae42d2a93086040b24605d30fe3137..033b757bf81c1496175a629117349b0760413aa6 100755 (executable)
--- a/ti/runtime/netapi/netapi_timer.h
+++ b/ti/runtime/netapi/netapi_timer.h
@@ -92,7 +92,7 @@ unsigned long t64_ticks_sec(void);
  *  @pre        @ref netapi_init \r
  */\r
 //return 64 bit timestamp from hw timer\r
-static inline unsigned long long netapi_getTimestamp(void) \r
+volatile static inline unsigned long long netapi_getTimestamp(void) \r
 {\r
  volatile unsigned long long t1;\r
  volatile unsigned long long t2;\r

diff --git a/ti/runtime/netapi/netapi_tune.h b/ti/runtime/netapi/netapi_tune.h
index 7cea8b1296470050222eb3f5a093ca98bd182c4c..54dcad9ff100c2342280a4b727e9217f7a6e055d 100755 (executable)
--- a/ti/runtime/netapi/netapi_tune.h
+++ b/ti/runtime/netapi/netapi_tune.h
@@ -87,9 +87,10 @@
 /**\r
  *  @def TUNE_NETAPI_PERM_MEM_SZ\r
  * (1) how much contiguous memory to grab. This is used for\r
- * descriptors and buffers.  Can't be bigger than MSMC (internal SOC memory area) \r
+ * descriptors and buffers.  Can't be bigger than CMA memory area (or msmc if\r
+ * we are using uncached .  This can be set at netapi_init via NETAPI_CFG_T\r
  */\r
-#define TUNE_NETAPI_PERM_MEM_SZ   (2*1024*1024)\r
+#define TUNE_NETAPI_PERM_MEM_SZ   (2*1024*1024) \r
 \r
 /**\r
  * @def TUNE_NETAPI_MAX_PKTIO\r
@@ -104,6 +105,7 @@
 /**\r
  * @def TUNE_NETAPI_DEFAULT_BUFFER_SIZE\r
  * (3) size of netapi default pktlib heap buffers\r
+*  This can be set at netapi_init() \r
  */\r
 #define TUNE_NETAPI_DEFAULT_BUFFER_SIZE 1600  \r
 \r
@@ -112,12 +114,14 @@
 /**\r
  * @def TUNE_NETAPI_DEFAULT_NUM_BUFFERS\r
  *(4) number of netapi default pktlib heap buffers (and assoc descriptors)\r
+ * this can be set at netapi_init()\r
  */\r
 #define TUNE_NETAPI_DEFAULT_NUM_BUFFERS  200 \r
 \r
 /*\r
  * @def TUNE_NETAPI_DEFAULT_NUM_SOLO_DESCRIPTORS\r
  * (5) number of netapi default pkt lib heap solo descriptors\r
+ * this can be set at netapi_init\r
  */\r
 #define TUNE_NETAPI_DEFAULT_NUM_SOLO_DESCRIPTORS  100 \r
 \r
@@ -128,6 +132,8 @@
 #define NETAPI_INCLUDE_SCHED \r
 \r
 //(7) # of QM descriptors (total)\r
+//     this can be set set in netapi_init\r
+// MUST BE POWER OF 2\r
 #define TUNE_NETAPI_QM_CONFIG_MAX_DESC_NUM  1024 /* 16384 is abs max */\r
 \r
 //(8) Region info\r
@@ -135,6 +141,7 @@
 \r
 /* at least as big as DEFAULT_NUM_SOLO+DEFAULT_NUM_BUFFERS but also a power of 2*/\r
 #define TUNE_NETAPI_NUM_GLOBAL_DESC  512 \r
+\r
 #define TUNE_NETAPI_DESC_SIZE  128 //don't change!!\r
 #ifdef NETAPI_USE_DDR\r
 #define TUNE_NETAPI_QM_START_INDEX  0x2000  //WARNING:  must reflect what kernel is using for their region, see device tree blob\r

diff --git a/ti/runtime/netapi/netcp_cfg.h b/ti/runtime/netapi/netcp_cfg.h
index 5c6e3bb307f446bcb99a7a4765c0e9e9417a9dfe..073a65fcfde23f2a4207c5db0cbf91bf82f4e689 100755 (executable)
--- a/ti/runtime/netapi/netcp_cfg.h
+++ b/ti/runtime/netapi/netcp_cfg.h
@@ -71,6 +71,7 @@ NETCP_CFG_FLOW_HANDLE_T netcp_cfgAddFlow(NETAPI_T ,
                                             int n, \r
                                             Pktlib_HeapHandle handles[],\r
                                            int sizes[],\r
+                                            int recv_offset,  //bytes to save in front of packet\r
                                             int * err );\r
 void netcp_cfgDelFlow(NETAPI_T , NETCP_CFG_FLOW_HANDLE_T , int * err);\r
 \r

diff --git a/ti/runtime/netapi/pktio.h b/ti/runtime/netapi/pktio.h
index 91eb045c06c6e6a6068007d9472f23d7f9fa0ade..ef7bc775a2a3e3534631b0f374c3d36da7cd6adf 100755 (executable)
--- a/ti/runtime/netapi/pktio.h
+++ b/ti/runtime/netapi/pktio.h
@@ -79,11 +79,23 @@ typedef struct PKTIO_METADATA_Tag
 /* the callback function */\r
 struct PKTIO_HANDLE_tag;\r
 \r
+//polling control\r
+typedef struct PKTIO_POLL_Tag\r
+{\r
+/* future */\r
+} PKTIO_POLL_T;\r
+\r
 #define PKTIO_MAX_RECV  (TUNE_NETAPI_MAX_BURST_RCV)\r
 typedef void (*PKTIO_CB)(struct PKTIO_HANDLE_tag * channel, Ti_Pkt* p_recv[],\r
                          PKTIO_METADATA_T p_meta[], int n_pkts,\r
                          uint64_t ts );\r
 \r
+typedef int (*PKTIO_SEND)(struct PKTIO_HANDLE_tag * channel, Ti_Pkt* p_send,\r
+                         PKTIO_METADATA_T *p_meta, int * p_err);\r
+\r
+typedef int (*PKTIO_POLL)(struct PKTIO_HANDLE_tag * channel,PKTIO_POLL_T * p_poll_cfg ,\r
+                          int * p_err);\r
+\r
 /** channel configuration */\r
 #define PKTIO_NA 0\r
 typedef struct PKTIO_CFG_Tag\r
@@ -129,6 +141,8 @@ typedef struct PKTIO_HANDLE_Tag
        Qmss_Queue qInfo;        /*   and its qm#/q# */\r
        int max_n;        /* max # of pkts to read in one poll */\r
        void * cookie;    /* app specific */\r
+        PKTIO_SEND _send;  /* pktio type specific send function */\r
+        PKTIO_POLL _poll;  /* pktio type specific POLL function */\r
         char name[PKTIO_MAX_NAME+1];\r
 }  PKTIO_HANDLE_T;\r
 \r
@@ -143,11 +157,6 @@ typedef struct PKTIO_CONTROL_Tag
 } PKTIO_CONTROL_T;\r
 \r
 \r
-//polling control\r
-typedef struct PKTIO_POLL_Tag\r
-{\r
-/* future */\r
-} PKTIO_POLL_T;\r
 \r
 /*---------------------------------------------------*/\r
 /*-------------------------API-----------------------*/\r
@@ -227,10 +236,13 @@ void pktio_delete(PKTIO_HANDLE_T * channel, int * err);
  *  @retval      int npkts: 1 if packet sent, 0 if error \r
  *  @pre       @ref netapi_init, @ref pktio_create, @ref pktio_open\r
  */\r
-int  pktio_send(PKTIO_HANDLE_T * channel, /* the channel */\r
+static inline int  pktio_send(PKTIO_HANDLE_T * channel, /* the channel */\r
                 Ti_Pkt *pkt,              /* pointer to packet */\r
                 PKTIO_METADATA_T *m,      /* pointer to meta data */\r
-                int * err);\r
+                int * err)\r
+{\r
+   return channel->_send((struct PKTIO_HANDLE_tag *)channel, pkt, m, err);\r
+}\r
 \r
 /*\r
 *  @brief  API sends data to a pktio channel\r
@@ -273,9 +285,12 @@ int pktio_sendMulti(PKTIO_HANDLE_T *channel, /* the channel handle */
  *  @retval      int npkts: number of packets received by poll \r
  *  @pre       @ref netapi_init, @ref pktio_create, @ref pktio_open\r
  */\r
-int pktio_poll(PKTIO_HANDLE_T * handle,   //handle to pktio\r
+static inline int pktio_poll(PKTIO_HANDLE_T * handle,   //handle to pktio\r
                PKTIO_POLL_T * p_poll_cfg,  //polling configuration\r
-               int * err);\r
+               int * err)\r
+{\r
+       handle->_poll((struct PKTIO_HANDLE_tag *) handle, p_poll_cfg, err);\r
+}\r
 \r
 /*\r
 *  @brief  API polls all pkto channels associarted with @ref NETAPI_T instance\r
@@ -303,4 +318,80 @@ int pktio_pollAll(NETAPI_T handle, PKTIO_POLL_T * p_poll_cfg, int *err);
 #define pktio_set_cookie(handle, cookie) (handle)->cookie = cookie\r
 #define pktio_get_cookie(handle) (handle)->cookie\r
 #define pktio_get_q(handle) (handle)->q\r
+\r
+/*-----------------Extra Fast Path pkt meta data macros--------------------*/\r
+#include "cppi_desc.h"\r
+#include "ti/drv/pa/pa.h"\r
+#include "ti/drv/pa/pasahost.h"\r
+\r
+\r
+//return default packet queue to poll for netcp RX\r
+//these are expensive calls, so call once and save\r
+static inline Qmss_QueueHnd PKTIO_GET_DEFAULT_NETCP_Q(PKTIO_HANDLE_T *h)\r
+{\r
+nwalGlobCxtInfo_t Info;\r
+nwal_getGlobCxtInfo(h->nwalInstanceHandle,&Info);\r
+return Info.rxDefPktQ;\r
+}\r
+\r
+//return L4Queue to poll for netcp RX (L4 classifier queue)\r
+//these are expensive calls, so call once and save\r
+static inline Qmss_QueueHnd PKTIO_GET_DEFAULT_NETCP_L4Q(PKTIO_HANDLE_T *h)\r
+{\r
+nwalLocCxtInfo_t Info;\r
+nwal_getLocCxtInfo(h->nwalInstanceHandle,&Info);\r
+return Info.rxL4PktQ;\r
+}\r
+\r
+\r
+/* find pointer to proto info fields in descriptor */\r
+static inline pasahoLongInfo_t* PKTIO_GET_PROTO_INFO( Ti_Pkt * pkt)\r
+{\r
+pasahoLongInfo_t*   pinfo;\r
+uint32_t            infoLen;\r
+Cppi_getPSData (Cppi_DescType_HOST, Cppi_PSLoc_PS_IN_DESC,(Cppi_Desc *)pkt, (uint8_t **)&pinfo, &infoLen);\r
+return pinfo;\r
+}\r
+\r
+/** "p" below is return of PKTIO_GET_PROTO_INFO() above**/\r
+\r
+/* offset to L3 header */\r
+#define PKTIO_GET_L3_OFFSET(p)  PASAHO_LINFO_READ_L3_OFFSET(p)\r
+\r
+/* offset to L4 header */\r
+#define PKTIO_GET_L4_OFFSET(p)  PASAHO_LINFO_READ_L4_OFFSET(p)\r
+\r
+/* next proto header */\r
+#define PKTIO_GET_NEXT_HEADER_TYPE(P) PASAHO_LINFO_READ_NXT_HDR_TYPE(p)\r
+\r
+/* offset to L4 payload */\r
+#define PKTIO_GET_L5_OFFSET(p)  PASAHO_LINFO_READ_L5_OFFSET(p)\r
+\r
+/* end of L4 payload */\r
+#define PKTIO_GET_PAYLOAD_END(p) PASAHO_LINFO_READ_END_OFFSET(p)\r
+\r
+/* IPSEC ESP done ? */\r
+#define PKTIO_ESP_DONE(p) PASAHO_LINFO_IS_IPSEC_ESP(p)\r
+\r
+/* IPSEC ESP done ? */\r
+#define PKTIO_AH_DONE(p) PASAHO_LINFO_IS_IPSEC_AH(p)\r
+\r
+/* MAC info */\r
+#define PKTIO_IS_MAC_BROADCAST(p)    PASAHO_READ_BITFIELD(p)\r
+#define PKTIO_IS_MAC_MULTICAST(p)    PASAHO_READ_BITFIELD(p)\r
+#define PKTIO_GET_MAC_TYPE(p)        PASAHO_LINFO_READ_MAC_PKTTYPE(p)\r
+\r
+/* read input port */\r
+#define PKTIO_GET_INPUT_PORT(p) PASAHO_LINFO_READ_INPORT(p) \r
+\r
+/* AppId */\r
+static inline unsigned int PKTIO_GET_APPID( Ti_Pkt * pkt) \r
+{\r
+unsigned char * p_swinfo0;\r
+Cppi_getSoftwareInfo (Cppi_DescType_HOST,\r
+                     (Cppi_Desc *)pkt,\r
+                      &p_swinfo0);\r
+return *((unsigned int *)p_swinfo0);\r
+} \r
+\r
 #endif\r

diff --git a/ti/runtime/netapi/src/netapi.c b/ti/runtime/netapi/src/netapi.c
index 7cf66d5c2ae16cbf8335000a9942f0dd206d44e8..fe785af69fd0ce3f094b7858ad2b09186cd67fa5 100755 (executable)
--- a/ti/runtime/netapi/src/netapi.c
+++ b/ti/runtime/netapi/src/netapi.c
@@ -64,13 +64,24 @@ static void netapi_cleanup_at_start(void);
 /*------------globals-----------------*/\r
 #define NUM_HOST_DESC  (TUNE_NETAPI_NUM_LOCAL_DESC)\r
 #define SIZE_LOCAL_DESC  (TUNE_NETAPI_DESC_SIZE) \r
-#define NUM_SHARED_DESC  (TUNE_NETAPI_NUM_GLOBAL_DESC) \r
 #define SIZE_SHARED_DESC (TUNE_NETAPI_DESC_SIZE)\r
 \r
 #define CONFIG_BUFSIZE_PA_INST      256\r
 #define CONFIG_BUFSIZE_L2_TABLE     1000\r
 #define CONFIG_BUFSIZE_L3_TABLE     4000\r
 \r
+static NETAPI_CFG_T netapi_default_cfg=\r
+{\r
+TUNE_NETAPI_PERM_MEM_SZ,\r
+0,  //start of packet offset for hw to place data on rx for default flow\r
+TUNE_NETAPI_QM_CONFIG_MAX_DESC_NUM, //max number of descriptors in system\r
+TUNE_NETAPI_NUM_GLOBAL_DESC,        //total we will use\r
+TUNE_NETAPI_DEFAULT_NUM_BUFFERS,   //#descriptors+buffers in default heap\r
+TUNE_NETAPI_DEFAULT_NUM_SOLO_DESCRIPTORS, //#descriptors w/o buffers in default heap\r
+TUNE_NETAPI_DEFAULT_BUFFER_SIZE   //size of buffers in default heap\r
+\r
+};\r
+\r
 static Pktlib_HeapIfTable  netapi_pktlib_ifTable;\r
 static NETAPI_GLOBAL_T netapi_global;\r
 NETAPI_GLOBAL_T * netapi_get_global(){ return &netapi_global;}\r
@@ -83,7 +94,7 @@ void netapi_zapQ(int queueNum);
 /*-------------------------------------\r
  *  initialize NETAPI instance \r
  *-------------------------------------*/\r
-NETAPI_T netapi_init(int master)\r
+NETAPI_T netapi_init(int master, NETAPI_CFG_T * p_cfg)\r
 {\r
  int i;\r
  int err;\r
@@ -107,6 +118,8 @@ NETAPI_T netapi_init(int master)
  /* global stuff (if master) */\r
  if (master==NETAPI_SYS_MASTER)\r
  {\r
+   if (p_cfg) memcpy(&netapi_global.cfg,p_cfg, sizeof(NETAPI_CFG_T));\r
+   else      memcpy(&netapi_global.cfg,&netapi_default_cfg, sizeof(NETAPI_CFG_T));\r
    for(i=0;i<NETAPI_MAX_PKTIO;i++) \r
    { \r
        netapi_global.pktios[i].qn.qNum=-1;\r
@@ -206,7 +219,7 @@ static int system_init(NETAPI_HANDLE_T * handle)
     netapi_init_timer();\r
 \r
     /* Initialize Queue Manager Sub System */\r
-    result = netapi_init_qm (); \r
+    result = netapi_init_qm (netapi_global.cfg.def_max_descriptors); \r
     if (result != 1)\r
     {\r
         return -1;\r
@@ -223,7 +236,7 @@ static int system_init(NETAPI_HANDLE_T * handle)
 \r
     /* Initialize the global descriptor memory region. */\r
     result= netapi_qm_setup_mem_region( \r
-                      NUM_SHARED_DESC,\r
+                      netapi_global.cfg.def_tot_descriptors_for_us,\r
                       SIZE_SHARED_DESC,\r
                       (unsigned int *) netapi_VM_QMemGlobalDescRam,\r
                       NETAPI_GLOBAL_REGION);\r
@@ -239,6 +252,7 @@ static int system_init(NETAPI_HANDLE_T * handle)
     if(result <0) {printf("can't setup local region\n"); return -1;}\r
 #endif\r
     /* Initialize CPPI CPDMA */\r
+\r
     result = netapi_init_cppi ();\r
     if (result != 1)\r
     {\r
@@ -258,12 +272,11 @@ static int system_init(NETAPI_HANDLE_T * handle)
     netapi_pktlib_ifTable.data_free               = netapiSharedMemoryFree;\r
 \r
     /* Create Shared Heap with specified configuration. */\r
-#define SHARED_MAX_DATA_SIZE      (TUNE_NETAPI_DEFAULT_BUFFER_SIZE) \r
    sharedHeapHandle = Pktlib_createHeap("netapi", NETAPI_GLOBAL_REGION,  //was 0\r
                                                 1,\r
-                                                SHARED_MAX_DATA_SIZE,\r
-                                                TUNE_NETAPI_DEFAULT_NUM_BUFFERS,\r
-                                                TUNE_NETAPI_DEFAULT_NUM_SOLO_DESCRIPTORS,\r
+                                                netapi_global.cfg.def_heap_buf_size,\r
+                                                netapi_global.cfg.def_heap_n_descriptors,\r
+                                                netapi_global.cfg.def_heap_n_zdescriptors,\r
                                                 &netapi_pktlib_ifTable);\r
    //todo -> cleanup on failure\r
    if (!sharedHeapHandle) { printf(">'netapi' heap create failed\n"); return -1;}\r
@@ -284,7 +297,8 @@ static int system_init(NETAPI_HANDLE_T * handle)
     result = netapi_init_nwal(\r
                               NETAPI_GLOBAL_REGION,\r
                               &netapi_pktlib_ifTable, \r
-                              &netapi_global.nwal_context);\r
+                              &netapi_global.nwal_context,\r
+                              &netapi_global.cfg);\r
     if (result<0) {printf(">netapi  init_nwal() failed\n"); return -1; }\r
 \r
     /* start NWAL */\r

diff --git a/ti/runtime/netapi/src/netapi_init.c b/ti/runtime/netapi/src/netapi_init.c
index 59c0b3df26974b3dcea8720a68ad49028e4e9d76..51936b396fabb9f6843196997d8206e06f51e04d 100755 (executable)
--- a/ti/runtime/netapi/src/netapi_init.c
+++ b/ti/runtime/netapi/src/netapi_init.c
@@ -140,7 +140,7 @@ int netapi_init_cpsw(void)
 //****************************************************\r
 // initialize QM (per SOC)\r
 //***************************************************\r
-int netapi_init_qm(void)\r
+int netapi_init_qm(int max_descriptors)\r
 {\r
   Qmss_InitCfg     qmssInitConfig;\r
   int32_t          result;\r
@@ -152,7 +152,7 @@ int netapi_init_qm(void)
   qmssInitConfig.linkingRAM0Base = 0;\r
   qmssInitConfig.linkingRAM0Size = 0;\r
   qmssInitConfig.linkingRAM1Base = 0;\r
-  qmssInitConfig.maxDescNum      = TUNE_NETAPI_QM_CONFIG_MAX_DESC_NUM;\r
+  qmssInitConfig.maxDescNum      = max_descriptors;\r
   qmssInitConfig.qmssHwStatus =QMSS_HW_INIT_COMPLETE; //bypass some of the hw init\r
   nwalTest_qmssGblCfgParams = qmssGblCfgParams[0];\r
 \r
@@ -393,7 +393,8 @@ uint8_t salldChanHandle[NETAPI_NWAL_CONFIG_BUFSIZE_SA_LLD_HANDLE_PER_CHAN * TUNE
 int netapi_init_nwal(\r
      int region2use,  \r
      Pktlib_HeapIfTable * p_table,\r
-     NETAPI_NWAL_GLOBAL_CONTEXT_T * p_nwal_context )\r
+     NETAPI_NWAL_GLOBAL_CONTEXT_T * p_nwal_context, \r
+     NETAPI_CFG_T*p_cfg )\r
 {\r
     nwalSizeInfo_t  nwalSizeInfo;\r
     nwalMemBuf_t    nwalMemBuf[nwal_N_BUFS];\r
@@ -411,12 +412,12 @@ int netapi_init_nwal(
     /* Initialize Buffer Pool for NetCP PA to SA packets */\r
     nwalGlobCfg.pa2SaBufPool.numBufPools = 1;\r
     nwalGlobCfg.pa2SaBufPool.bufPool[0].descSize = TUNE_NETAPI_DESC_SIZE;\r
-    nwalGlobCfg.pa2SaBufPool.bufPool[0].bufSize =  TUNE_NETAPI_DEFAULT_BUFFER_SIZE;\r
+    nwalGlobCfg.pa2SaBufPool.bufPool[0].bufSize =  p_cfg->def_heap_buf_size;\r
     nwalGlobCfg.pa2SaBufPool.bufPool[0].heapHandle = p_nwal_context->pa2sa_heap=\r
                       Pktlib_createHeap("nwal PA2SA",\r
                        region2use,\r
                        0,\r
-                       TUNE_NETAPI_DEFAULT_BUFFER_SIZE,\r
+                       p_cfg->def_heap_buf_size,\r
                        TUNE_NETAPI_CONFIG_MAX_PA_TO_SA_DESC,\r
                        0,\r
                        p_table);\r
@@ -430,13 +431,13 @@ int netapi_init_nwal(
  /* Initialize Buffer Pool for NetCP SA to PA packets */\r
     nwalGlobCfg.sa2PaBufPool.numBufPools = 1;\r
     nwalGlobCfg.sa2PaBufPool.bufPool[0].descSize = TUNE_NETAPI_DESC_SIZE;\r
-    nwalGlobCfg.sa2PaBufPool.bufPool[0].bufSize =  TUNE_NETAPI_DEFAULT_BUFFER_SIZE;\r
+    nwalGlobCfg.sa2PaBufPool.bufPool[0].bufSize =  p_cfg->def_heap_buf_size;\r
 \r
     nwalGlobCfg.sa2PaBufPool.bufPool[0].heapHandle = p_nwal_context->sa2pa_heap=\r
         Pktlib_createHeap("nwal SA2PA",\r
                           region2use,\r
                           0,\r
-                          TUNE_NETAPI_DEFAULT_BUFFER_SIZE,\r
+                          p_cfg->def_heap_buf_size,\r
                           TUNE_NETAPI_CONFIG_MAX_SA_TO_PA_DESC,\r
                           0,\r
                           p_table);\r

diff --git a/ti/runtime/netapi/src/netapi_loc.h b/ti/runtime/netapi/src/netapi_loc.h
index cc9029ac21f586c625ff43515f794fa94b288594..b7244aa7760fe9a2eed809df884297ba343b041d 100755 (executable)
--- a/ti/runtime/netapi/src/netapi_loc.h
+++ b/ti/runtime/netapi/src/netapi_loc.h
@@ -164,7 +164,8 @@ typedef struct NETAPI_GLOBAL_tag
 #define NETAPI_MAX_PKTIO (TUNE_NETAPI_MAX_PKTIO) \r
 PKTIO_ENTRY_T pktios[NETAPI_MAX_PKTIO];\r
 \r
-/* pktlib heap */\r
+/* configuration */\r
+NETAPI_CFG_T  cfg;\r
 \r
 /* global timers */\r
 \r
@@ -212,14 +213,15 @@ void * cookie;  /*set by calling thread */
 \r
 \r
 //internal initialization routines */\r
-int netapi_init_qm(void);\r
+int netapi_init_qm(int max_descriptors);\r
 int netapi_init_cppi(void);\r
 int netapi_init_cpsw(void);\r
 int netapi_start_qm(void);\r
 int netapi_init_nwal(\r
      int region2use,\r
      Pktlib_HeapIfTable * p_table,\r
-     NETAPI_NWAL_GLOBAL_CONTEXT_T * p_nwal_context );\r
+     NETAPI_NWAL_GLOBAL_CONTEXT_T * p_nwal_context,\r
+     NETAPI_CFG_T *p_cfg );\r
 int netapi_start_nwal(Pktlib_HeapHandle pkt_heap,\r
                     Pktlib_HeapHandle cmd_heap,\r
                      NETAPI_NWAL_LOCAL_CONTEXT_T *p ,\r

diff --git a/ti/runtime/netapi/src/netapi_sec.c b/ti/runtime/netapi/src/netapi_sec.c
index 3742ec997d50e6c87b28ef1e7c533fdb49e653d5..a71882112ce4dcf9af87efa6988108dabc2e379e 100644 (file)
--- a/ti/runtime/netapi/src/netapi_sec.c
+++ b/ti/runtime/netapi/src/netapi_sec.c
@@ -208,7 +208,7 @@ nwalCreateSAParams_t    createParam =
                 netcp_cfgp_delete_sa(&netapi_get_global()->nwal_context,tunnelId);
                 return -1;
         }
-        printf(">netapisec %d Creating sideband mode SA for %d ( mac %d)\n", tunnelId, iface_no); 
+        printf(">netapisec.  Creating sideband mode SA for %d ( mac %d)\n", tunnelId, iface_no); 
         *p_data_mode_handle = dm_handle;
     }
     //save stuff

diff --git a/ti/runtime/netapi/src/netapi_util.h b/ti/runtime/netapi/src/netapi_util.h
index 8830112bc029095d9d500cdf258edb0e29d16ec4..b0bae824753fc542c3904b7fcd0f5ff1356db221 100755 (executable)
--- a/ti/runtime/netapi/src/netapi_util.h
+++ b/ti/runtime/netapi/src/netapi_util.h
@@ -42,6 +42,7 @@
  *****************************************************/\r
 #ifndef __NETAPI_UTIL__H\r
 #define __NETAPI_UTIL__H\r
+\r
 /* timing */\r
 static inline unsigned long netapi_timing_start(void)\r
 {\r

diff --git a/ti/runtime/netapi/src/netapi_vm.c b/ti/runtime/netapi/src/netapi_vm.c
index 03a80494f79b0ccc998e4569535bbefe90a9bb0e..a7e1756d441f22bc8663e65ba68b9df60a3d9767 100755 (executable)
--- a/ti/runtime/netapi/src/netapi_vm.c
+++ b/ti/runtime/netapi/src/netapi_vm.c
@@ -63,20 +63,6 @@
 /* Macro to align x to y */\r
 #define align(x,y)   ((x + y) & (~y))\r
 \r
-#define NETAPI_PERM_MEM_SZ  (TUNE_NETAPI_PERM_MEM_SZ) \r
-\r
-/* Physical address map & size for various subsystems */\r
-#define QMSS_CFG_BASE_ADDR  CSL_QM_SS_CFG_QUE_PEEK_REGS\r
-#define QMSS_CFG_BLK_SZ (1*1024*1024)\r
-#define QMSS_DATA_BASE_ADDR 0x44020000 \r
-#define QMSS_DATA_BLK_SZ (0x60000)\r
-#define SRIO_CFG_BASE_ADDR  CSL_SRIO_CONFIG_REGS\r
-#define SRIO_CFG_BLK_SZ (132*1024)\r
-#define PASS_CFG_BASE_ADDR  CSL_PA_SS_CFG_REGS \r
-#define PASS_CFG_BLK_SZ (1*1024*1024)\r
-\r
-#define MSMC_SRAM_BASE_ADDR CSL_MSMC_SRAM_REGS\r
-\r
 uint8_t *netapi_VM_mem_start_phy = (uint8_t*)0;\r
 uint8_t *netapi_VM_mem_start = (uint8_t*)0;\r
 uint8_t *netapi_VM_mem_end = (uint8_t*)0;\r
@@ -85,6 +71,7 @@ static uint8_t *netapi_VM_mem_alloc_ptr = (uint8_t*)0;
 static uint32_t netapi_VM_mem_size = 0;\r
 \r
 \r
+\r
 /* File descriptor for /dev/mem */ \r
 static int dev_mem_fd;\r
 static int  our_netapi_module_fd;\r
@@ -131,7 +118,7 @@ nwal_Bool_t netapi_VM_memAllocInit
     addr=  ( uint8_t *) netapi_utilGetPhysOfBufferArea();  //get address that was allocated for us by kernela module */\r
     size = netapi_utilGetSizeOfBufferArea();   //get the size that was allocated\r
 #ifdef USE_MODULE_MMAP\r
-    map_base = (void *) netapi_utilGetVaOfBufferArea(NETAPIMOD_MMAP_DMA_MEM_OFFSET, size); //mmap into our space, return va\r
+    map_base = (void *) netapi_utilGetVaOfBufferArea(0,size); //mmap into our space, return va\r
 #else\r
     if(  (temp_fd = open("/dev/mem", O_RDWR )) == -1) {\r
       printf(">netapi_VM_memAllocInit: failed to open dev/mem again cached err=%d\n",errno);\r
@@ -246,6 +233,20 @@ void *netapi_VM_memMap
 /* Total Permanent memory required in NWAL test\r
  * for Packet buffers & descriptor buffers\r
  */\r
+#define NETAPI_PERM_MEM_SZ  (TUNE_NETAPI_PERM_MEM_SZ) \r
+\r
+/* Physical address map & size for various subsystems */\r
+#define QMSS_CFG_BASE_ADDR  CSL_QM_SS_CFG_QUE_PEEK_REGS\r
+#define QMSS_CFG_BLK_SZ (1*1024*1024)\r
+#define QMSS_DATA_BASE_ADDR 0x44020000 \r
+#define QMSS_DATA_BLK_SZ (0x60000)\r
+#define SRIO_CFG_BASE_ADDR  CSL_SRIO_CONFIG_REGS\r
+#define SRIO_CFG_BLK_SZ (132*1024)\r
+#define PASS_CFG_BASE_ADDR  CSL_PA_SS_CFG_REGS \r
+#define PASS_CFG_BLK_SZ (1*1024*1024)\r
+\r
+#define MSMC_SRAM_BASE_ADDR CSL_MSMC_SRAM_REGS\r
+\r
 /* Global variables to hold virtual address of various subsystems */\r
 void *netapi_VM_qmssCfgVaddr;\r
 void *netapi_VM_qmssDataVaddr;\r
@@ -301,13 +302,8 @@ int netapi_VM_memory_setup(void)
     printf(">netapi QMSS_CFG_BASE_ADDR:0x%x Memory mapped at address %p.\n",(void*)QMSS_CFG_BASE_ADDR, netapi_VM_qmssCfgVaddr);\r
 \r
     /* (2b) QMSS DATA Regs */\r
-#ifdef USE_MODULE_MMAP\r
-    netapi_VM_qmssDataVaddr = (void *) netapi_utilGetVaOfBufferArea(NETAPIMOD_MMAP_QM_DATA_REG_MEM_OFFSET, QMSS_DATA_BLK_SZ);\r
-#else\r
     netapi_VM_qmssDataVaddr = netapi_VM_memMap((void*)QMSS_DATA_BASE_ADDR,\r
                                             QMSS_DATA_BLK_SZ);\r
-#endif\r
-\r
     if (!netapi_VM_qmssDataVaddr)\r
     {\r
         printf(">netapi ERROR: Failed to map QMSS DATA registers\n");\r

diff --git a/ti/runtime/netapi/src/netcp_cfg.c b/ti/runtime/netapi/src/netcp_cfg.c
index e9a7791e4e0e09e4cd8fbb3f0b4f67dced86a892..2ae66eb34b75873a7bc1197b2d3486e9a5981f99 100755 (executable)
--- a/ti/runtime/netapi/src/netcp_cfg.c
+++ b/ti/runtime/netapi/src/netcp_cfg.c
@@ -1114,6 +1114,7 @@ NETCP_CFG_FLOW_HANDLE_T netcp_cfgAddFlow(NETAPI_T h,
                                             int n,\r
                                             Pktlib_HeapHandle handles[],\r
                                             int sizes[],\r
+                                            int byte_offset,\r
                                             int * err )\r
 {\r
   Cppi_RxFlowCfg  rxFlowCfg;\r
@@ -1155,7 +1156,7 @@ NETCP_CFG_FLOW_HANDLE_T netcp_cfgAddFlow(NETAPI_T h,
   rxFlowCfg.flowIdNum      = CPPI_PARAM_NOT_SPECIFIED;\r
   rxFlowCfg.rx_dest_qnum   = 100;   //DANGEROUS>  TODO PUT VALID Q HERE\r
   rxFlowCfg.rx_dest_qmgr   = 0;\r
-  rxFlowCfg.rx_sop_offset  = 0;\r
+  rxFlowCfg.rx_sop_offset  = byte_offset;\r
   rxFlowCfg.rx_ps_location = Cppi_PSLoc_PS_IN_DESC;\r
   rxFlowCfg.rx_desc_type   = Cppi_DescType_HOST;\r
   rxFlowCfg.rx_error_handling = 0;\r

diff --git a/ti/runtime/netapi/src/osal.c b/ti/runtime/netapi/src/osal.c
index 5e6a334ffef18cc50443cfed65164de5237fee61..6ee7c33497390ee1205df449f1670fd2f422b7b0 100755 (executable)
--- a/ti/runtime/netapi/src/osal.c
+++ b/ti/runtime/netapi/src/osal.c
@@ -220,11 +220,11 @@ unsigned long netapi_utilGetVaOfBufferArea(unsigned int offset, unsigned int siz
 \r
     /* Map the physical address to user space */\r
     userp = mmap(0,                       // Preferred start address\r
-                 size,                           // Length to be mapped\r
+                 size,            // Length to be mapped\r
                  PROT_WRITE | PROT_READ,  // Read and write access\r
                  MAP_SHARED,              // Shared memory\r
                  netapi_fd,               // File descriptor\r
-                 offset);                 // The byte offset from fd\r
+                 offset);                      // The byte offset from fd\r
 \r
     if (userp == MAP_FAILED) {\r
         return 0;\r
@@ -272,7 +272,7 @@ static inline void Osal_writeBackCache (void *blockPtr, uint32_t size)
     if ((blockPtr <netapi_VM_mem_start)||( blockPtr>netapi_VM_mem_end)) return;\r
     //netapi_utilCacheWbInv(blockPtr, size);\r
     //printf("osal> wbiv %x %x %d ..",  blockPtr, Osal_saGetSCPhyAddr(blockPtr), size);\r
-    _netapi_utilCacheWbInv(blockPtr, size);\r
+    _netapi_utilCacheWbInv(_Osal_qmssVirtToPhy(blockPtr), size);\r
     v2= netapi_timing_stop();\r
     cache_op_cycles += (v2-v1); \r
     n_cache_op_cycles+=1;\r
@@ -660,13 +660,13 @@ void* Osal_qmssConvertDescVirtToPhy(void *descAddr)
         Cppi_HostDesc *prevBDPtr = 0;\r
         while (nextBDPtr)\r
         {\r
-            void *buffPtr;\r
+            void *buffPtr=NULL;\r
             if (nextBDPtr->buffPtr)\r
             {\r
                 buffPtr = (void *)nextBDPtr->buffPtr;\r
                 nextBDPtr->buffPtr = (uint32_t)_Osal_qmssVirtToPhy((void *)(nextBDPtr->buffPtr));\r
                 if (!(nextBDPtr->buffPtr)) return (void *)0;\r
-                Osal_writeBackCache(buffPtr, nextBDPtr->buffLen);\r
+                Osal_writeBackCache(buffPtr, nextBDPtr->buffLen);          \r
             }\r
 \r
             if (nextBDPtr->origBuffPtr)\r
@@ -683,6 +683,7 @@ void* Osal_qmssConvertDescVirtToPhy(void *descAddr)
                 if (!(prevBDPtr->nextBDPtr)) return (void *)0;\r
             }\r
 \r
+            if (buffPtr) Osal_writeBackCache(buffPtr, prevBDPtr->buffLen);\r
             Osal_writeBackCache(prevBDPtr, TUNE_NETAPI_DESC_SIZE);\r
         }\r
         descAddr = _Osal_qmssVirtToPhy(descAddr);\r
@@ -717,12 +718,12 @@ void* Osal_qmssConvertDescPhyToVirt(void *descAddr)
         Cppi_HostDesc *nextBDPtr = (Cppi_HostDesc *)QMSS_DESC_PTR(descAddr);\r
         while (nextBDPtr)\r
         {\r
-            //Osal_invalidateCache(nextBDPtr, TUNE_NETAPI_DESC_SIZE);\r
+            //Qmss_osalBeginMemAccess(nextBDPtr, TUNE_NETAPI_DESC_SIZE);\r
             if (nextBDPtr->buffPtr)\r
             {\r
                 nextBDPtr->buffPtr = (uint32_t)_Osal_qmssPhyToVirt((void *)(nextBDPtr->buffPtr));\r
                 if (!(nextBDPtr->buffPtr)) return (void *)0;\r
-                //Osal_invalidateCache((void *)(nextBDPtr->buffPtr), nextBDPtr->buffLen);\r
+                //Qmss_osalBeginMemAccess((void *)(nextBDPtr->buffPtr), nextBDPtr->buffLen);\r
             }\r
 \r
             if (nextBDPtr->origBuffPtr)\r
@@ -745,7 +746,6 @@ void* Osal_qmssConvertDescPhyToVirt(void *descAddr)
     {\r
         descAddr = _Osal_qmssPhyToVirt(descAddr);\r
         if (!descAddr) return (void *)0;\r
-        //Osal_invalidateCache(descAddr, TUNE_NETAPI_DESC_SIZE);\r
     }\r
 #endif\r
     return descAddr;\r

diff --git a/ti/runtime/netapi/src/pktio.c b/ti/runtime/netapi/src/pktio.c
index 0f2b15d1355d18657cb073520b61814eedadc542..c805913fee5f24c56741a9cd1e2f9608a3b21f5c 100755 (executable)
--- a/ti/runtime/netapi/src/pktio.c
+++ b/ti/runtime/netapi/src/pktio.c
@@ -55,6 +55,10 @@ extern unsigned int vv9p;
 extern unsigned int vv10p;\r
 extern unsigned int vv11p;\r
 extern unsigned int vv12p;\r
+extern unsigned int vv13p;  //rcv path\r
+extern unsigned int vv14p;\r
+extern unsigned int vv15p;\r
+\r
 \r
 #ifdef DAL_BENCH\r
 unsigned int pktio_get_qop_time(){return (BENCH_get_qop_time());}\r
@@ -84,6 +88,190 @@ static PKTIO_HANDLE_T * pktiop_get_free_channel_slot(NETAPI_T n)
        return NULL;\r
 }\r
 \r
+\r
+/*-----------------------------------------------------*/\r
+/*          optimized send/rcv functions              */\r
+/*----------------------------------------------------*/\r
+\r
+//********************************************\r
+//send pkt via ipc queue \r
+//********************************************\r
+static int pktio_send_ipc(PKTIO_HANDLE_T * p, Ti_Pkt *pkt, PKTIO_METADATA_T *m, int * err)\r
+{\r
+        *err=0;\r
+#ifdef DO_BENCH\r
+        vv8p=netapi_timing_stop();\r
+#endif\r
+#ifdef DAL_BENCH\r
+        BENCH_Qmss_queuePushDesc (p->q, (void*)pkt);\r
+#else\r
+        Qmss_queuePushDesc (p->q, (void*)pkt);\r
+#endif\r
+#ifdef DO_BENCH\r
+        vv9p=netapi_timing_stop();\r
+#endif\r
+       return 1;\r
+}\r
+//********************************************\r
+//send pkt to NETCP via NWAL\r
+//********************************************\r
+static int pktio_send_nwal(PKTIO_HANDLE_T * p, Ti_Pkt *pkt, PKTIO_METADATA_T *m, int * err)\r
+{\r
+        nwalTxPktInfo_t * pPktInfo=m->u.tx_meta;\r
+        nwal_RetValue res;\r
+       *err=0;\r
+       pPktInfo->pPkt = pkt;\r
+#ifdef DO_BENCH\r
+        vv11p=netapi_timing_stop();\r
+#endif\r
+        res=nwal_send(p->nwalInstanceHandle, m->sa_handle,pPktInfo);\r
+#ifdef DO_BENCH\r
+        vv12p=netapi_timing_stop();\r
+#endif\r
+        if (res != nwal_OK) {  *err = NETAPI_ERR_NWAL_TX_ERR -res;}\r
+        return 1;\r
+}\r
+//********************************************\r
+//send to SA via SB queue\r
+//********************************************\r
+static int pktio_send_sb(PKTIO_HANDLE_T * p, Ti_Pkt *pkt, PKTIO_METADATA_T *m, int * err)\r
+{\r
+        nwalDmTxPayloadInfo_t *pPktInfoSB =  m->u.tx_sb_meta;\r
+        nwal_RetValue res;\r
+       *err=0;\r
+        pPktInfoSB->pPkt = pkt;\r
+        res=nwal_sendDM(p->nwalInstanceHandle, m->sa_handle,pPktInfoSB);\r
+        if (res != nwal_OK) *err = NETAPI_ERR_NWAL_TX_ERR -res;\r
+        return 1;\r
+}\r
+//********************************************\r
+//dummy.  return err\r
+//********************************************\r
+static int pktio_send_dummy(PKTIO_HANDLE_T * p, Ti_Pkt *pkt, PKTIO_METADATA_T *m, int * err)\r
+{\r
+       *err =  NETAPI_ERR_BAD_INPUT;\r
+       return -1;\r
+} \r
+//*******************************************\r
+//dummy poll\r
+//*******************************************\r
+static int pktio_poll_dummy(PKTIO_HANDLE_T * p, PKTIO_POLL_T * p_poll_cfg,  int * err)\r
+{\r
+       *err= NETAPI_ERR_BAD_INPUT;\r
+       return 0;\r
+}\r
+//********************************************\r
+//poll IPC queue\r
+//********************************************\r
+static int pktio_poll_ipc(PKTIO_HANDLE_T * p, PKTIO_POLL_T * p_poll_cfg,  int * err)\r
+{\r
+Ti_Pkt * pkt_list[PKTIO_MAX_RECV];\r
+PKTIO_METADATA_T meta_s[PKTIO_MAX_RECV];\r
+int r=0;\r
+int n;\r
+Ti_Pkt * temp;\r
+*err=0;\r
+n= (p->max_n< PKTIO_MAX_RECV) ? p->max_n : PKTIO_MAX_RECV;\r
+for(r=0;r<n;r++)\r
+{\r
+#ifdef DO_BENCH\r
+      if (r==0) vv7p=netapi_timing_stop();\r
+#endif\r
+#ifdef DAL_BENCH\r
+      temp=(Ti_Pkt*)(Cppi_HostDesc*)QMSS_DESC_PTR(BENCH_Qmss_queuePop(p->q));\r
+#else\r
+      temp=(Ti_Pkt*)(Cppi_HostDesc*)QMSS_DESC_PTR(Qmss_queuePop(p->q));\r
+#endif\r
+#ifdef DO_BENCH\r
+      if (r==0) vv10p=netapi_timing_stop();\r
+#endif\r
+      if(!temp) break;\r
+         /* process meta data */\r
+         pkt_list[r]= temp;\r
+         meta_s[r].flags1=0x1;\r
+      }\r
+      if (r) p->cb((struct PKTIO_HANDLE_tag *)p, pkt_list, &meta_s[0], r, 0LL);\r
+      return r;\r
+}\r
+//********************************************\r
+//poll nwal data queues for pkts from netcp\r
+//********************************************\r
+static int pktio_poll_nwal(PKTIO_HANDLE_T * p, PKTIO_POLL_T * p_poll_cfg,  int * err)\r
+{\r
+int r=0;\r
+*err=0;\r
+        /* Poll for common L2/L3 packets and L4 class pkts (todo-> only do L4 if classifiers are\r
+         set.. optimizaion maybe? */\r
+#ifdef DO_BENCH\r
+        vv13p=netapi_timing_stop();\r
+#endif\r
+        r=nwal_pollPkt(p->nwalInstanceHandle,\r
+                   nwal_POLL_DEFAULT_GLOB_PKT_Q|  nwal_POLL_DEFAULT_PER_PROC_PKT_Q,\r
+                   (uint32_t) p,\r
+                   p->max_n,\r
+                   QMSS_PARAM_NOT_SPECIFIED,\r
+                   (void*) NULL);\r
+        return r;\r
+}\r
+//********************************************\r
+//poll nwal sideband queues for pkts from SA\r
+//********************************************\r
+static int  pktio_poll_sb(PKTIO_HANDLE_T * p, PKTIO_POLL_T * p_poll_cfg,  int * err)\r
+{\r
+int r=0;\r
+*err=0;\r
+      r=nwal_pollDm(p->nwalInstanceHandle,\r
+                    nwal_POLL_DM_DEF_GLOB_ENCRYPT_Q,\r
+                     (uint32_t) p,\r
+                     p->max_n,\r
+                     QMSS_PARAM_NOT_SPECIFIED,\r
+                     (void *)  NULL);\r
+      r+=nwal_pollDm(p->nwalInstanceHandle,\r
+                     nwal_POLL_DM_DEF_GLOB_DECRYPT_Q,\r
+                     (uint32_t) p,\r
+                     p->max_n,\r
+                     QMSS_PARAM_NOT_SPECIFIED,\r
+                    (void *) NULL);\r
+      return r;\r
+}\r
+\r
+//********************************************\r
+//poll app-provided netcp rx queue\r
+//********************************************\r
+static int pktio_poll_nwal_adj(PKTIO_HANDLE_T * p, PKTIO_POLL_T * p_poll_cfg,  int * err)\r
+{\r
+int r=0;\r
+*err=0;\r
+        /* Poll for common L2/L3 packets and L4 class pkts (todo-> only do L4 if classifiers are\r
+         set.. optimizaion maybe? */\r
+#ifdef DO_BENCH\r
+        vv14p=netapi_timing_stop();\r
+#endif\r
+        r=nwal_pollPkt(p->nwalInstanceHandle,\r
+                     nwal_POLL_APP_MANAGED_PKT_Q,\r
+                     (uint32_t) p,\r
+                     p->max_n,\r
+                     p->q,\r
+                     (void *)  NULL);\r
+        return r;\r
+}\r
+\r
+//********************************************\r
+//poll app-defined sideband queues for pkts from SA\r
+//********************************************\r
+static int  pktio_poll_sb_adj(PKTIO_HANDLE_T * p, PKTIO_POLL_T * p_poll_cfg,  int * err)\r
+{\r
+int r=0;\r
+*err=0;\r
+      r=nwal_pollDm(p->nwalInstanceHandle,\r
+                     nwal_POLL_DM_APP_MANAGED_Q,\r
+                     (uint32_t) p,\r
+                     p->max_n,\r
+                     p->q,\r
+                     (void *)  NULL);\r
+      return r;\r
+}\r
+\r
 /*-----------------------MAIN API----------------------*/\r
 /* create a channel */\r
 PKTIO_HANDLE_T * pktio_create(NETAPI_T n, char * name,\r
@@ -107,6 +295,8 @@ p->inuse= PKTIO_INUSE;
 p->back = n;\r
 p->cb = cb;\r
 p->max_n = p_cfg->max_n;\r
+p->_poll=pktio_poll_dummy;\r
+p->_send=pktio_send_dummy;\r
 memcpy((char *)&p->cfg, (char*) p_cfg, sizeof(PKTIO_CFG_T));\r
 \r
     /* create a  general queue (for now). todo: allow qnum to be passed in */\r
@@ -123,14 +313,21 @@ memcpy((char *)&p->cfg, (char*) p_cfg, sizeof(PKTIO_CFG_T));
     if (p_cfg->flags2 & PKTIO_PKT)\r
     {\r
        p->use_nwal = PKTIO_4_ADJ_NWAL;\r
+       p->_poll=pktio_poll_nwal_adj;\r
        p->nwalInstanceHandle = netapi_return_nwal_instance_handle(n);\r
     }\r
     else if (p_cfg->flags2 & PKTIO_SB)\r
     {\r
        p->use_nwal = PKTIO_4_ADJ_SB;\r
+       p->_poll=pktio_poll_sb_adj;\r
        p->nwalInstanceHandle = netapi_return_nwal_instance_handle(n);\r
     }\r
-    else p->use_nwal=0;\r
+    else\r
+    {\r
+        p->use_nwal=0;\r
+        if (p_cfg->flags1& PKTIO_W) p->_send=pktio_send_ipc;\r
+        if (p_cfg->flags1& PKTIO_R) p->_poll=pktio_poll_ipc;\r
+    }\r
 \r
     /* save name */ \r
     strncpy(p->name,name,\r
@@ -174,6 +371,8 @@ p->inuse= PKTIO_INUSE;
 p->back = n;\r
 p->cb = cb;\r
 p->max_n = p_cfg->max_n;\r
+p->_poll=pktio_poll_dummy;\r
+p->_send=pktio_send_dummy;\r
 memcpy((char *)&p->cfg, (char*) p_cfg, sizeof(PKTIO_CFG_T));\r
 \r
 /* special handling of NETCP_RX, NETCP_TX */\r
@@ -183,6 +382,8 @@ if( (!strcmp(name, NETCP_RX)) || (!strcmp(name,NETCP_TX)) )
     p->use_nwal = PKTIO_DEF_NWAL;\r
     p->q = 0;  \r
     p->nwalInstanceHandle = netapi_return_nwal_instance_handle(n); \r
+    if (!strcmp(name,NETCP_RX)) p->_poll=pktio_poll_nwal;\r
+    if (!strcmp(name,NETCP_TX)) p->_send=pktio_send_nwal;\r
 }\r
 else if( (!strcmp(name, NETCP_SB_RX)) || (!strcmp(name,NETCP_SB_TX)) )\r
 {\r
@@ -190,6 +391,8 @@ else if( (!strcmp(name, NETCP_SB_RX)) || (!strcmp(name,NETCP_SB_TX)) )
     p->use_nwal =  PKTIO_DEF_SB;\r
     p->q = 0;\r
     p->nwalInstanceHandle = netapi_return_nwal_instance_handle(n);\r
+    if (!strcmp(name,NETCP_SB_RX)) p->_poll=pktio_poll_sb;\r
+    if (!strcmp(name,NETCP_SB_TX)) p->_send=pktio_send_sb;\r
 }\r
 else\r
 {\r
@@ -217,15 +420,21 @@ else
     if (p_cfg->flags2 & PKTIO_PKT)\r
     {\r
        p->use_nwal = PKTIO_4_ADJ_NWAL;  //additonal RX q for nwal\r
+       p->_poll = pktio_poll_nwal_adj;\r
        p->nwalInstanceHandle = netapi_return_nwal_instance_handle(n); \r
     } \r
     else if (p_cfg->flags2 & PKTIO_SB)\r
     {\r
        p->use_nwal = PKTIO_4_ADJ_SB;  //additional RX q for sideband with NWAL\r
+       p->_poll = pktio_poll_sb_adj;\r
        p->nwalInstanceHandle = netapi_return_nwal_instance_handle(n);\r
     } \r
-\r
-    else p->use_nwal=0; //not handled by nwal\r
+    else \r
+    {\r
+       p->use_nwal=0; //not handled by nwal\r
+        if (p_cfg->flags1& PKTIO_W) p->_send=pktio_send_ipc;\r
+        if (p_cfg->flags1& PKTIO_R) p->_poll=pktio_poll_ipc;\r
+    }\r
 }\r
 \r
     /* save name */\r
@@ -316,7 +525,7 @@ void pktio_delete(PKTIO_HANDLE_T * p, int * err)
 /***********************************************************/\r
 /*****************send *************************/\r
 /***********************************************************/\r
-int pktio_send(PKTIO_HANDLE_T * p, Ti_Pkt *pkt, PKTIO_METADATA_T *m, int * err)\r
+int pktio_send_genric(PKTIO_HANDLE_T * p, Ti_Pkt *pkt, PKTIO_METADATA_T *m, int * err)\r
 {\r
         nwalTxPktInfo_t * pPktInfo=m->u.tx_meta;\r
        nwal_RetValue res;\r
@@ -328,7 +537,7 @@ int pktio_send(PKTIO_HANDLE_T * p, Ti_Pkt *pkt, PKTIO_METADATA_T *m, int * err)
                 vv11p=netapi_timing_stop();\r
                res=nwal_send(p->nwalInstanceHandle, m->sa_handle,pPktInfo);\r
                 vv12p=netapi_timing_stop();\r
-               if (res != nwal_OK) *err = NETAPI_ERR_NWAL_TX_ERR -res;\r
+               if (res != nwal_OK) { printf("nwal send err= %d (%x)\n", res, res);  *err = NETAPI_ERR_NWAL_TX_ERR -res;}\r
        }\r
         else if ((p->use_nwal== PKTIO_DEF_SB) )\r
         {\r
@@ -397,14 +606,15 @@ int pktio_sendMulti(PKTIO_HANDLE_T * p, Ti_Pkt * pkt[], PKTIO_METADATA_T * m[],
 /***********************************************************/\r
 \r
 /* poll a particular channel */\r
-int pktio_poll(PKTIO_HANDLE_T * p, PKTIO_POLL_T * p_poll_cfg,  int * err)\r
+int pktio_poll_generic(PKTIO_HANDLE_T * p, PKTIO_POLL_T * p_poll_cfg,  int * err)\r
 {\r
 int r=0;\r
 int n;\r
 Ti_Pkt * temp;\r
 Ti_Pkt * pkt_list[PKTIO_MAX_RECV];\r
 PKTIO_METADATA_T meta_s[PKTIO_MAX_RECV];\r
-uint64_t ts= netapi_getTimestamp();  //get_ts\r
+//uint64_t ts= netapi_getTimestamp();  //get_ts\r
+uint64_t ts=0LL;\r
 \r
         if(! p->cfg.flags1&PKTIO_R) return 0;\r
 \r
@@ -413,6 +623,7 @@ uint64_t ts= netapi_getTimestamp();  //get_ts
        {\r
                /* Poll for common L2/L3 packets and L4 class pkts (todo-> only do L4 if classifiers are\r
                    set.. optimizaion maybe? */\r
+                vv13p=netapi_timing_stop();\r
                r=nwal_pollPkt(p->nwalInstanceHandle,\r
                      nwal_POLL_DEFAULT_GLOB_PKT_Q|  nwal_POLL_DEFAULT_PER_PROC_PKT_Q,\r
                      (uint32_t) p,\r
@@ -424,6 +635,7 @@ uint64_t ts= netapi_getTimestamp();  //get_ts
        else if (p->use_nwal==PKTIO_4_ADJ_NWAL)\r
        {\r
                /* Poll an additional NETCP RX queue */\r
+                vv14p=netapi_timing_stop();\r
                r=nwal_pollPkt(p->nwalInstanceHandle,\r
                      nwal_POLL_APP_MANAGED_PKT_Q,\r
                      (uint32_t) p,\r
@@ -519,6 +731,7 @@ int r=0;
 int n;\r
 Ti_Pkt * pkt_list[PKTIO_MAX_RECV];\r
 PKTIO_METADATA_T meta_s[PKTIO_MAX_RECV];\r
+    vv15p=netapi_timing_stop();\r
     for(r=0;r<numPkts;r++)\r
     {\r
          pkt_list[r] = pPktInfo[r].pPkt;\r

diff --git a/ti/runtime/netapi/src/tim64.c b/ti/runtime/netapi/src/tim64.c
index ad17c5a8f29a796c46f330fc7a04cd98532ee542..00c1a80d19c5f64bf4b1ef5b0a0cb17116e6cf13 100755 (executable)
--- a/ti/runtime/netapi/src/tim64.c
+++ b/ti/runtime/netapi/src/tim64.c
@@ -129,11 +129,19 @@ t64_cpu_cycle_per_tick = (unsigned long) ((t2-t1)/(t264-t164));
 printf(">netapi:  tim64 calibration - n=%d t2-t1=%lu t264-t164=%llu ccpt=%ld tps=%ld\n",\r
        n, t2-t1, t264-t164, t64_cpu_cycle_per_tick,  t64_ticks_sec());\r
 \r
+t1=timing_stop();\r
+t164=netapi_getTimestamp();\r
+t264=netapi_getTimestamp();\r
+t2=timing_stop();\r
+printf(">netapi:  tim64 cycle cost= %d(cpu ticks),  back2back= %lld\n", (t2-t1)/2, t264-t164);\r
+\r
 }\r
 \r
 \r
 /*********************************\r
  * memory map t64 into user space\r
+t264=read_t64();\r
+t264=read_t64();\r
  *   input: pass in fd for /dev/mem\r
  **********************************/\r
 int t64_memmap(int fd)\r

diff --git a/ti/runtime/netapi/test/net_test.c b/ti/runtime/netapi/test/net_test.c
index 3877e14ed088661a89696f691fe30b93cd659bde..311fc3a39d5b87114561093ffc93beab9471e8cf 100644 (file)
--- a/ti/runtime/netapi/test/net_test.c
+++ b/ti/runtime/netapi/test/net_test.c
@@ -59,6 +59,8 @@
 //#define IPSEC_MODE_RX_SIDEBAND\r
 //#define IPSEC_MODE_TX_SIDEBAND\r
 \r
+//#define TEST_TIMERS\r
+\r
 /*************debug********************/\r
 void dump_descr(unsigned long *p, int n)\r
 {\r
@@ -83,6 +85,7 @@ void dump_header(unsigned long *p, int n, int a, int r)
 //************for multi pkt burst  xfer test in loopback mode\r
 #define TX_BURST 800 \r
 int pktloopback=TUNE_NETAPI_NWAL_ENABLE_PASS_LOOPBACK;\r
+nwalTxFlowCookie_t  flowCookie;\r
 \r
 //this device: 10.0.0.100, mac 0x,01,02,03,04,05  and .. 0x6\r
 \r
@@ -201,6 +204,7 @@ uint16_t test_utilGetIpv4PsudoChkSum (uint8_t *data, uint16_t payloadLen)
 typedef struct stats_t\r
 {\r
         long itx;  //initially generated\r
+        long itx2;\r
        long rx;\r
        long tx;\r
        long n_bad;\r
@@ -269,6 +273,17 @@ int perslow= PERSLOW;
 /*******************************************\r
  *************NETAPI OBJECTS***************\r
  *****************************************/\r
+static NETAPI_CFG_T our_netapi_default_cfg=\r
+{\r
+TUNE_NETAPI_PERM_MEM_SZ,\r
+256,  //start of packet offset for hw to place data on rx for default flow\r
+TUNE_NETAPI_QM_CONFIG_MAX_DESC_NUM, //max number of descriptors in system\r
+TUNE_NETAPI_NUM_GLOBAL_DESC,        //total we will use\r
+TUNE_NETAPI_DEFAULT_NUM_BUFFERS,   //#descriptors+buffers in default heap\r
+0, //#descriptors w/o buffers in default heap\r
+TUNE_NETAPI_DEFAULT_BUFFER_SIZE+256   //size of buffers in default heap\r
+};\r
+\r
 Pktlib_HeapHandle OurHeap;\r
 Pktlib_HeapHandle specialSmall;\r
 Pktlib_HeapHandle specialLarge;\r
@@ -290,7 +305,7 @@ void house(NETAPI_SCHED_HANDLE_T *s);
 NETAPI_T netapi_handle;\r
 NETAPI_SCHED_HANDLE_T * our_sched;\r
 NETAPI_SCHED_CONFIG_T our_sched_cfg={\r
-  NETAPI_SCHED_DURATION|NETAPI_SCHED_CBV, 0, house, 500  //every 500 poll loops\r
+  NETAPI_SCHED_DURATION|NETAPI_SCHED_CBV, 0, house, 50000  //every 5000 poll loops\r
 };\r
 void our_stats_cb(NETAPI_T h, paSysStats_t* pPaStats);\r
 NETAPI_TIMER_GROUP_HANDLE_T ourTimerBlock; \r
@@ -565,6 +580,55 @@ unsigned int vv12p;
 \r
 unsigned int vv11;\r
 \r
+unsigned int vv13p;  //rcv path\r
+unsigned int vv14p;\r
+unsigned int vv15p;\r
+\r
+extern unsigned int nwal_prof1,nwal_prof2,nwal_prof3,nwal_prof4,nwal_prof5,nwal_prof6;\r
+\r
+//#define REASSEMBLE_BENCH\r
+#ifdef REASSEMBLE_BENCH\r
+#include <ti/drv/pa/example/reassemLib/reassemLib.h>\r
+/*--------------reassembly benchmark--------------------------------*/\r
+void our_reassemble_bench(int nfrags)\r
+{\r
+paIPReassemblyConfig_t Config={5,128,10000 };\r
+int i,j;\r
+int len;\r
+Ti_Pkt tip;\r
+char *buffer;\r
+unsigned long v1;\r
+unsigned long v2;\r
+unsigned long sum1=0;\r
+unsigned long sum2=0;\r
+paEx_reassemLibInit(&Config);\r
+for(j=0;j<200/nfrags;j++)\r
+{\r
+  for(i=0;i<nfrags;i++)\r
+  {\r
+    short temp;\r
+    tip=Pktlib_allocPacket(OurHeap,PKT_LEN);\r
+    Pktlib_getDataBuffer(tip,(uint8_t**)&buffer,&len);\r
+    memcpy(&buffer[0],&testPkt[14],20); //IP header\r
+    if (i < (nfrags-1)) buffer[6] = 0x20;\r
+    temp = i*40; \r
+    buffer[6]|= (temp&0x1f00)>>8;\r
+    buffer[7]= (temp&0xff);\r
+    temp = 20+40*8; \r
+    buffer[2]= (temp&0xff00)>>8;\r
+    buffer[3]= (temp&0xff);\r
+    Pktlib_setPacketLen(tip, temp);\r
+    v1= netapi_timing_stop();\r
+    paEx_reassemLibProc(tip, 0xffff);\r
+    v2= netapi_timing_stop();\r
+    sum1+= v2-v1;\r
+  }\r
+  sum2 += v2-v1;\r
+}\r
+printf("reasssembly test:  %d trials, %d frags/pkt  %d cycles/frag %d cycles/last frag\n",j,nfrags, sum1/(j*nfrags), sum2/(j));\r
+}\r
+#endif\r
+\r
 /*--------------basic pktio send/recv benchmark----------------------*/\r
 unsigned int timings[10];\r
 void our_pktio_bench(int ntrials)\r
@@ -653,6 +717,8 @@ Ti_Pkt * get_pkt(int n, unsigned int *p_len)
 \r
    //get pointer to buffer area of packet\r
    Pktlib_getDataBuffer(b,(uint8_t**)&buffer,&len);\r
+   if (!buffer) \r
+    {printf("net_test: get_pkt() heap returned empty buffer %d \n", n); return NULL;};\r
 \r
 #if 0 \r
 if (pktloopback==0)\r
@@ -849,7 +915,10 @@ void recv_cb_bench(struct PKTIO_HANDLE_Tag * channel, Ti_Pkt* p_recv[],
 {\r
    int i;\r
    vv6=   netapi_timing_stop();\r
-   for (i=0;i<n_pkts; i++) Pktlib_freePacket(p_recv[i]);\r
+   for (i=0;i<n_pkts; i++) \r
+   {\r
+     Pktlib_freePacket(p_recv[i]);\r
+   }\r
    vv11 = netapi_timing_stop();\r
 }\r
 \r
@@ -1213,6 +1282,18 @@ unsigned int sum_vv3=0;
 unsigned int sum_vv4=0;\r
 unsigned int sum_vv5=0;\r
 \r
+unsigned int nwal_flow_vv1,nwal_flow_vv2;\r
+unsigned int nwal_sum_vv1=0;\r
+unsigned int nwal_sum_vv2=0;\r
+unsigned int nwal_sum_vv3=0;\r
+unsigned int nwal_sum_vv4=0;\r
+unsigned int nwal_sum_vv5=0;\r
+unsigned int nwal_sum_vv6=0;\r
+\r
+unsigned int nwal_sum_flow_vv1=0;\r
+unsigned int nwal_sum_flow_vv2=0;\r
+static int first =0;\r
+\r
 Osal_cache_op_measure_reset();\r
 memset(&meta_tx,0,sizeof(meta_tx));\r
 for(p=0;p<TX_BURST;p++) {  \r
@@ -1246,7 +1327,7 @@ if ((house_pkts_gened>0) && (! (house_pkts_gened%1000)) )
 \r
    /* set up meta data */\r
     meta.sa_handle=nwal_HANDLE_INVALID;\r
-    meta_tx.txFlag1 = (NWAL_TX_FLAG1_DO_IPV4_CHKSUM | NWAL_TX_FLAG1_DO_UDP_CHKSUM| NWAL_TX_FLAG1_META_DATA_VALID);\r
+    meta_tx.txFlag1 = (NWAL_TX_FLAG1_DO_IPV4_CHKSUM | NWAL_TX_FLAG1_DO_UDP_CHKSUM| NWAL_TX_FLAG1_META_DATA_VALID|NWAL_TX_FLAG1_COPY_TX_FLOW_COOKIE);\r
     meta_tx.startOffset = 0;\r
     //GONE in V2 meta_tx.pktLen = len;\r
     meta_tx.ipOffBytes = TEST_PKT_IP_OFFSET_BYTES;\r
@@ -1254,6 +1335,7 @@ if ((house_pkts_gened>0) && (! (house_pkts_gened%1000)) )
     meta_tx.l4HdrLen = TEST_PKT_UDP_HDR_LEN;\r
     //GONE in V2 meta_tx.ploadOffBytes = TEST_PKT_PLOAD_OFFSET_BYTES;\r
     meta_tx.ploadLen = TEST_PAYLOAD_LEN;\r
+    meta_tx.pTxFlowCookie = &flowCookie;\r
 \r
     Pktlib_getDataBuffer(tip,&pData,&len);\r
     pIpHdr = pData + meta_tx.ipOffBytes;\r
@@ -1266,16 +1348,52 @@ if ((house_pkts_gened>0) && (! (house_pkts_gened%1000)) )
    if (house_pkts_gened<16) dump_descr((long *) tip, house_pkts_gened);\r
    else if (house_pkts_gened>99) dump_descr((long *) tip,house_pkts_gened);\r
 #endif\r
-   vv2= netapi_timing_stop();\r
-   pktio_send(netcp_tx_chan,tip,&meta,&err);\r
-   vv3= netapi_timing_stop();\r
-   sum_vv1 += (vv3-vv1);\r
-   sum_vv2 += (vv2-vv1);\r
-   sum_vv3 += (vv3-vv2);\r
-   sum_vv4 += (vv11p-vv2);\r
-   sum_vv5 += (vv12p-vv11p);\r
+\r
+   if(first < TX_BURST)\r
+   {\r
+       first++;\r
+       vv2= netapi_timing_stop();\r
+       meta_tx.txFlag1 |= NWAL_TX_FLAG1_COPY_TX_FLOW_COOKIE;\r
+       pktio_send(netcp_tx_chan,tip,&meta,&err);\r
+       vv3= netapi_timing_stop();\r
+       if(err != 0)\r
+       {\r
+           printf("pktIO Error %d \n",err);\r
+       }\r
+       sum_vv1 += (vv3-vv1);\r
+       sum_vv2 += (vv2-vv1);\r
+       sum_vv3 += (vv3-vv2);\r
+       sum_vv4 += (vv11p-vv2);\r
+       sum_vv5 += (vv12p-vv11p);\r
+\r
+       nwal_sum_vv1+= (nwal_prof1-vv11p);\r
+       nwal_sum_vv2+= (nwal_prof2-nwal_prof1);\r
+       nwal_sum_vv3+= (nwal_prof3-nwal_prof2);\r
+       nwal_sum_vv4+= (nwal_prof4-nwal_prof3);\r
+       nwal_sum_vv5+= (nwal_prof5-nwal_prof4);\r
+      nwal_sum_vv6+= (nwal_prof6-nwal_prof5);\r
+      if (err == 0) stats.itx +=1;\r
+   }\r
+   else\r
+   {\r
+       nwal_RetValue    retVal;\r
+       nwal_flow_vv1= netapi_timing_stop();\r
+       meta_tx.pPkt = tip;\r
+       retVal = nwal_flowSend(&meta_tx);\r
+       nwal_flow_vv2= netapi_timing_stop();\r
+       if(retVal != nwal_OK)\r
+       {           \r
+           printf("nwal_flowSend Error %d \n",retVal);\r
+       }\r
+\r
+       nwal_sum_flow_vv1 += (nwal_flow_vv1-vv1); \r
+       nwal_sum_flow_vv2 += (nwal_flow_vv2-nwal_flow_vv1); \r
+\r
+       if(retVal == nwal_OK)stats.itx2 +=1;\r
+   }\r
+   \r
    // printf("pktio send. full=%d metadata=%d pktio_send=%d\n", vv3-vv1,  vv2-vv1,  vv3-vv2);\r
-   if (err == 0) stats.itx +=1;\r
+   \r
 \r
    house_pkts_gened +=1;\r
  }\r
@@ -1284,10 +1402,24 @@ if ((house_pkts_gened>0) && (! (house_pkts_gened%1000)) )
    int n_c_ops;\r
    ccycles =Osal_cache_op_measure(&n_c_ops);\r
    if (sum_vv1) \r
+   {\r
       printf("BURST pktio send %d pkts. full=%d metadata=%d pktio_send=%d to_nwal=%d nwal_send= %d n_c_ops=%d cache_op_time=%d (pp-> %d)\n", stats.itx,\r
    sum_vv1/stats.itx,  sum_vv2/stats.itx,  sum_vv3/stats.itx, \r
    sum_vv4/stats.itx,  sum_vv5/stats.itx, n_c_ops, ccycles,  \r
    n_c_ops? (ccycles/(n_c_ops/2)) : 0);\r
+\r
+      printf("NWAL Profile Cycles: Prof1= %d,Prof2=%d,Prof3=%d,Prof4=%d,Prof5=%d ,Prof6=%d \n",\r
+              nwal_sum_vv1/stats.itx,nwal_sum_vv2/stats.itx,nwal_sum_vv3/stats.itx,\r
+              nwal_sum_vv4/stats.itx,nwal_sum_vv5/stats.itx,nwal_sum_vv6/stats.itx);\r
+    \r
+\r
+      if(stats.itx2)\r
+      {\r
+          printf("nwal_flowSend Profile Cycles: Prof1= %d,Prof2=%d \n",\r
+              nwal_sum_flow_vv1/stats.itx2,nwal_sum_flow_vv2/stats.itx2);\r
+      }\r
+\r
+   }\r
 }\r
 }\r
 \r
@@ -1328,7 +1460,7 @@ memcpy(&testPkt[26],&real_ip_addr[0],8);//overrite test pkt ip addresses
 /*******************************************/\r
 \r
 /* create netapi */\r
-netapi_handle = netapi_init(NETAPI_SYS_MASTER);\r
+netapi_handle = netapi_init(NETAPI_SYS_MASTER, &our_netapi_default_cfg);\r
 \r
 /* open the main heap */\r
 OurHeap = Pktlib_findHeapByName("netapi");\r
@@ -1351,6 +1483,10 @@ specialLarge = Pktlib_createHeap("netapi-small", NETAPI_GLOBAL_REGION,
 netapi_registerHeap(netapi_handle, specialSmall);\r
 netapi_registerHeap(netapi_handle, specialLarge);\r
 \r
+#ifdef REASSEMBLE_BENCH\r
+our_reassemble_bench(2);\r
+exit(1);\r
+#endif\r
 \r
 /* create a pktio channel */\r
 our_chan=pktio_create(netapi_handle,"our1stq",(PKTIO_CB) recv_cb_bench, &our_chan_cfg,&err);\r
@@ -1380,7 +1516,7 @@ if (!our_sched) {printf("sched create failed\n"); exit(1);}
 /********************************************\r
 * Basic pkt loopback test\r
 *********************************************/\r
-printf("...runnining pure push/pop benchmark\n");\r
+printf("...running pure push/pop benchmark\n");\r
 our_pktio_bench(1000);\r
 our_pktio_bench(1000);\r
 our_pktio_bench(1000);\r
@@ -1413,7 +1549,7 @@ ip_rule0=netcp_addIp(
                   &err\r
                   );\r
 if (err) {printf("addip0 failed %d\n",err); exit(1); } \r
-#if 1\r
+\r
 //create a 2nd mac instance\r
 netcp_cfgCreateMacInterface(\r
                   netapi_handle,\r
@@ -1470,9 +1606,11 @@ specialFlow = netcp_cfgAddFlow( netapi_handle,
                                2,\r
                                 heaps,\r
                                 sizes,\r
+                                0,  //offset to start rx is 0\r
                                 &err);\r
 if (err) {printf("add flow failed\n", err); exit(1);}\r
 }\r
+#if 0\r
 //special route for this classifier:  different flow + destination q\r
 class2_route.p_dest_q = netcp_rx_chan2;\r
 class2_route.p_flow = specialFlow;\r
@@ -1544,7 +1682,7 @@ tx_tunnel = netapi_secAddSA( netapi_handle,
 if (err) {printf("addTxSa failed %d\n",err); exit(1);}\r
 \r
 \r
-\r
+#ifdef TEST_TIMERS\r
 //timers\r
 ourTimerBlock = netapi_TimerGroupCreate(\r
         netapi_handle,\r
@@ -1577,7 +1715,7 @@ t3 = netapi_TimerGroupStartTimer(
         300LL,  //timer group ticks\r
         &err);\r
 if (err) {printf("timerstart failed %d\n");}\r
-\r
+#endif\r
 netcp_cfgReqStats(netapi_handle, our_stats_cb, 1,&err);\r
 if (err!=0) {printf("stats req failed\n");}\r
 \r
@@ -1593,9 +1731,6 @@ printf("...runnining pure push/pop benchmark\n");
 our_pktio_bench(100);\r
 \r
 \r
-\r
-\r
-\r
 /**************unused stuff******************/\r
 /* create TRIE */\r
 P_trie = trie_new();\r
@@ -1624,6 +1759,12 @@ netapi_schedWaitForEvents(our_sched, &err);
 /* done */\r
 our_stats_cb(netapi_handle, NULL);\r
 \r
+\r
+#define DO_FAST_POLL\r
+#ifdef DO_FAST_POLL\r
+example_fast_poll(netcp_rx_chan);\r
+#endif\r
+\r
 /*************************************************\r
  ************CLEAN UP****************************\r
  ************************************************/\r
@@ -1631,7 +1772,7 @@ our_stats_cb(netapi_handle, NULL);
 //delete Classifiers\r
 netcp_cfgDelClass(netapi_handle, class_0, &err); \r
 netcp_cfgDelClass(netapi_handle, class_1, &err); \r
-netcp_cfgDelClass(netapi_handle, class_2, &err); \r
+//netcp_cfgDelClass(netapi_handle, class_2, &err); \r
 \r
 //delete flow \r
 netcp_cfgDelFlow(netapi_handle, specialFlow, &err);\r
@@ -1670,3 +1811,141 @@ pktio_delete(netcp_rx_chan2,&err);
 netapi_shutdown(netapi_handle);\r
 \r
 }\r
+\r
+//EXAMPLE FAST POLL\r
+/* PLD */\r
+void netapi_pld(void * x)\r
+{   \r
+           asm volatile("pld [r0]");\r
+}\r
+\r
+#define M 1008\r
+static int l3_off[M], l4_off[M], L3_chk_ok[M], L4_chk_ok[M], appid[M], len[M] ;\r
+static unsigned char * buf[M];\r
+#define N2POLL 8\r
+void example_fast_poll( PKTIO_HANDLE_T * p_pktio)\r
+{\r
+int j=0;\r
+int jj=0;\r
+int i,k,l=0,ltot=0;\r
+int n= N2POLL; //max # of pkts to poll\r
+Ti_Pkt * pHd[N2POLL];\r
+Ti_Pkt * tempVA;\r
+pasahoLongInfo_t* pinfo;\r
+unsigned long t1;\r
+unsigned long t2;\r
+unsigned long t11;\r
+unsigned long t12;\r
+unsigned long np;\r
+unsigned long sumt=0;\r
+unsigned long sumf=0;\r
+unsigned long sump=0;\r
+unsigned long totlen=0;\r
+int max_batch=0;\r
+\r
+//this should  be done once and saved\r
+Qmss_QueueHnd rxQ= PKTIO_GET_DEFAULT_NETCP_Q(p_pktio);\r
+Qmss_QueueHnd freeQ;\r
+//loop forever\r
+for(;;)\r
+{\r
+    t1= netapi_timing_stop();\r
+    pHd[0] = (Ti_Pkt *)QMSS_DESC_PTR(Qmss_queuePopRAW (rxQ));\r
+    if (!pHd[0]) continue;\r
+    //got pkt\r
+    for(i=1;(i<n) && (pHd[i-1]);i++)\r
+    {\r
+        //convert previous descriptor PA -> VA\r
+        tempVA  =  Osal_qmssPhyToVirt(pHd[i-1]); \r
+        Cppi_getData (Cppi_DescType_HOST, (Cppi_Desc*)tempVA, &buf[jj], &len[jj]);\r
+\r
+        //try and preload desriptor\r
+        //__builtin_prefetch(tempVA);\r
+        //netapi_pld(tempVA);\r
+\r
+        //read next descriptor from queue \r
+        pHd[i] = (Ti_Pkt *)QMSS_DESC_PTR(Qmss_queuePopRAW (rxQ));\r
+       \r
+        /* extract some meta data */\r
+        pinfo =  PKTIO_GET_PROTO_INFO(tempVA);\r
+        l3_off[jj]= PKTIO_GET_L3_OFFSET(pinfo);\r
+        l4_off[jj]= PKTIO_GET_L4_OFFSET(pinfo);\r
+        appid[jj]= PKTIO_GET_APPID(tempVA);\r
+        //get ptr (Physical address) and length of associate buffer\r
+        //Cppi_getData (Cppi_DescType_HOST, (Cppi_Desc*)tempVA, &buf[jj], &len[jj]);\r
+        jj+=1;\r
+    }\r
+    //finish last pkt in burst\r
+    if(pHd[i-1])\r
+    {\r
+        //convert previous descriptor PA -> VA\r
+        tempVA  =  Osal_qmssPhyToVirt(pHd[i-1]); \r
+\r
+        /* extract some meta data */\r
+        pinfo =  PKTIO_GET_PROTO_INFO(tempVA);\r
+        l3_off[jj]= PKTIO_GET_L3_OFFSET(pinfo);\r
+        l4_off[jj]= PKTIO_GET_L4_OFFSET(pinfo);\r
+        appid[jj]= PKTIO_GET_APPID(tempVA);\r
+        //get ptr (Physical address) and length of associate buffer\r
+        Cppi_getData (Cppi_DescType_HOST, (Cppi_Desc*)tempVA, &buf[jj], &len[jj]);\r
+        jj+=1;\r
+    }\r
+    t2= netapi_timing_stop();\r
+    j+=(pHd[i-1]? i: (i-1)) ;\r
+    if (jj>(M-n)) jj=0;\r
+    l+=1;  //n batches\r
+    ltot+=1;\r
+    if(pHd[i-1])\r
+    {\r
+      if (i>max_batch) max_batch= i;\r
+    }\r
+    else\r
+    {\r
+      if( (i-1) >max_batch) max_batch = i-1;\r
+    }\r
+\r
+    //cleanup\r
+    //printf("cleanup %d\n",i);\r
+    for(k=0;k<i;k++)\r
+    {\r
+        //cleanup.  need to covert all of desriptor to VA so that i can use freePacket() \r
+        //alternative would be to just do cache ops plus descriptor raw push to pktlib\r
+        // heap free queue\r
+        if(pHd[k])\r
+        {\r
+\r
+        //tempVA=Qmss_osalConvertDescPhyToVirt(pHd[k]);\r
+        tempVA  =  Osal_qmssPhyToVirt(pHd[k]);\r
+        freeQ=Qmss_getQueueHandle(Cppi_getReturnQueue (Cppi_DescType_HOST, tempVA));\r
+        netapi_utilCacheWbInv(pHd[k],128);\r
+        //would need to wbInv buffer also in practice. Also need to walk\r
+        // descriptor chain \r
+        t11= netapi_timing_stop();\r
+       // Pktlib_freePacket(tempVA);\r
+        Qmss_queuePushDescSizeRaw (freeQ,\r
+                                (void *) pHd[k],\r
+                                128);\r
+\r
+        t12= netapi_timing_stop();\r
+        sumf += (t12-t11);       \r
+        }\r
+    }\r
+    sumt += (t2-t1);\r
+    sump +=(pHd[i-1]? i: (i-1));\r
+    //printf("end cleanup %d %d %d\n",sumt,sumf,sump );\r
+    if (sump > 10000) {\r
+       printf("pkts rx %d batches=%d  appid=%x l3_off=%d l4_off=%d len=%d buf=0x%x rxcycle= %d pkts/batchx1000=%d maxbatch=%d cycles per rawpush = %d\n", \r
+                  j,ltot, appid[j%M], \r
+                  l3_off[j%M],l4_off[j%M],  \r
+                  len[j%M],buf[j%M],\r
+                   sumt/sump,  (sump*1000)/l, max_batch,\r
+                   sumf/sump);\r
+       sumt=sump=sumf=0;\r
+       l=0;\r
+    }\r
+}\r
+\r
+\r
+}\r
+\r
+\r