Refresh of dapl-ep

diff --git a/trunk/ulp/dapl2/dapl/openib_scm/SOURCES b/trunk/ulp/dapl2/dapl/openib_scm/SOURCES
index 6e4ad30cc03c46a31bef9375fc79e5a622b64e7c..6372f2d555111abdb778efafb72e30fc6b7c77a5 100644 (file)
--- a/trunk/ulp/dapl2/dapl/openib_scm/SOURCES
+++ b/trunk/ulp/dapl2/dapl/openib_scm/SOURCES
@@ -28,7 +28,7 @@ INCLUDES = ..\include;..\openib_common\;..\common;windows;..\..\dat\include;\
                   ..\..\dat\udat\windows;..\udapl\windows;\\r
                   ..\..\..\..\inc;..\..\..\..\inc\user;..\..\..\libibverbs\include\r
 \r
-DAPL_OPTS = -DEXPORT_DAPL_SYMBOLS -DDAT_EXTENSIONS -DSOCK_CM -DOPENIB -DCQ_WAIT_OBJECT\r
+DAPL_OPTS = -DEXPORT_DAPL_SYMBOLS -DDAT_EXTENSIONS -DDAPL_COUNTERS -DOPENIB -DCQ_WAIT_OBJECT\r
 \r
 USER_C_FLAGS = $(USER_C_FLAGS) $(DAPL_OPTS)\r
 \r

diff --git a/trunk/ulp/dapl2/dapl/openib_scm/cm.c b/trunk/ulp/dapl2/dapl/openib_scm/cm.c
index 29694835c0ad39cca313a137094f8dc3b74ed263..31b5c94bbf5d2d524497d1a6f546440140bc9e90 100644 (file)
--- a/trunk/ulp/dapl2/dapl/openib_scm/cm.c
+++ b/trunk/ulp/dapl2/dapl/openib_scm/cm.c
@@ -60,6 +60,44 @@
 #include "dapl_ep_util.h"
 #include "dapl_osd.h"
 
+/* Check for EP linking to IA and proper connect state */\r
+void dapli_ep_check(DAPL_EP *ep)\r
+{\r
+       DAPL_IA *ia_ptr = ep->header.owner_ia;\r
+       DAPL_EP *ep_ptr, *next_ep_ptr;\r
+       int found = 0;\r
+\r
+       dapl_os_lock(&ia_ptr->header.lock);\r
+       ep_ptr = (dapl_llist_is_empty (&ia_ptr->ep_list_head)\r
+               ? NULL : dapl_llist_peek_head (&ia_ptr->ep_list_head));\r
+\r
+       while (ep_ptr != NULL) {\r
+               next_ep_ptr = \r
+                       dapl_llist_next_entry(&ia_ptr->ep_list_head,\r
+                                             &ep_ptr->header.ia_list_entry);\r
+               if (ep == ep_ptr) {\r
+                       found++;\r
+                       if ((ep->cr_ptr && ep->param.ep_state \r
+                               != DAT_EP_STATE_COMPLETION_PENDING) ||\r
+                           (!ep->cr_ptr && ep->param.ep_state \r
+                               != DAT_EP_STATE_ACTIVE_CONNECTION_PENDING))\r
+                               goto err;\r
+                       else \r
+                               goto match;\r
+               }\r
+               ep_ptr = next_ep_ptr;\r
+       }\r
+err:\r
+       dapl_log(DAPL_DBG_TYPE_ERR,\r
+                " dapli_ep_check ERR: %s %s ep=%p state=%d magic=0x%x\n", \r
+                ep->cr_ptr ? "PASSIVE":"ACTIVE", \r
+                found ? "WRONG_STATE":"NOT_FOUND" ,\r
+                ep, ep->param.ep_state, ep->header.magic);\r
+match:\r
+       dapl_os_unlock(&ia_ptr->header.lock);\r
+       return;\r
+}\r
+
 #if defined(_WIN32) || defined(_WIN64)
 enum DAPL_FD_EVENTS {
        DAPL_FD_READ = 0x1,
@@ -452,7 +490,7 @@ static void dapli_socket_connected(dp_ib_cm_handle_t cm_ptr, int err)
                dapl_log(DAPL_DBG_TYPE_ERR,
                         " CONN_PENDING: %s ERR %s -> %s %d\n",
                         err == -1 ? "POLL" : "SOCKOPT",
-                        err == -1 ? strerror(errno) : strerror(err), 
+                        err == -1 ? strerror(dapl_socket_errno()) : strerror(err), 
                         inet_ntoa(((struct sockaddr_in *)
                                &cm_ptr->addr)->sin_addr), 
                         ntohs(((struct sockaddr_in *)
@@ -475,9 +513,10 @@ static void dapli_socket_connected(dp_ib_cm_handle_t cm_ptr, int err)
        }
 
        if (len != (exp + ntohs(cm_ptr->msg.p_size))) {
+               int err = dapl_socket_errno();
                dapl_log(DAPL_DBG_TYPE_ERR,
-                        " CONN_PENDING len ERR %s, wcnt=%d(%d) -> %s\n",
-                        strerror(errno), len, 
+                        " CONN_PENDING len ERR 0x%x %s, wcnt=%d(%d) -> %s\n",
+                        err, strerror(err), len, 
                         exp + ntohs(cm_ptr->msg.p_size), 
                         inet_ntoa(((struct sockaddr_in *)
                                   ep_ptr->param.
@@ -530,16 +569,19 @@ dapli_socket_connect(DAPL_EP * ep_ptr,
        /* create, connect, sockopt, and exchange QP information */
        if ((cm_ptr->socket =
             socket(AF_INET, SOCK_STREAM, IPPROTO_TCP)) == DAPL_INVALID_SOCKET) {
+               int err = dapl_socket_errno();
                dapl_log(DAPL_DBG_TYPE_ERR,
-                        " connect: socket create ERR %s\n", strerror(errno));
+                        " connect: socket create ERR 0x%x %s\n", 
+                        err, strerror(err));
                goto bail;
        }
 
        ret = dapl_config_socket(cm_ptr->socket);
        if (ret < 0) {
                dapl_log(DAPL_DBG_TYPE_ERR,
-                        " connect: config socket %d ERR %d %s\n",
-                        cm_ptr->socket, ret, strerror(dapl_socket_errno()));
+                        " connect: config socket %d RET %d ERR 0x%x %s\n",
+                        cm_ptr->socket, ret, 
+                        dapl_socket_errno(), strerror(dapl_socket_errno()));
                dat_ret = DAT_INTERNAL_ERROR;
                goto bail;
        }
@@ -556,6 +598,10 @@ dapli_socket_connect(DAPL_EP * ep_ptr,
        ret = dapl_connect_socket(cm_ptr->socket, (struct sockaddr *)&cm_ptr->addr,
                                  sizeof(cm_ptr->addr));
        if (ret && ret != EAGAIN) {
+               dapl_log(DAPL_DBG_TYPE_ERR,
+                        " connect: dapl_connect_socket RET %d ERR 0x%x %s\n",
+                        ret, dapl_socket_errno(), 
+                        strerror(dapl_socket_errno()));
                dat_ret = DAT_INVALID_ADDRESS;
                goto bail;
        }
@@ -572,9 +618,10 @@ dapli_socket_connect(DAPL_EP * ep_ptr,
        /* get local address information from socket */
        sl = sizeof(cm_ptr->msg.daddr.so);
        if (getsockname(cm_ptr->socket, (struct sockaddr *)&cm_ptr->msg.daddr.so, &sl)) {
+               int err = dapl_socket_errno();
                dapl_log(DAPL_DBG_TYPE_ERR,
-                       " connect getsockname ERROR: %s -> %s r_qual %d\n",
-                       strerror(errno), 
+                       " connect getsockname ERROR: 0x%x %s -> %s r_qual %d\n",
+                       err, strerror(err), 
                        inet_ntoa(((struct sockaddr_in *)r_addr)->sin_addr),
                        (unsigned int)r_qual);;
        }
@@ -604,8 +651,7 @@ dapli_socket_connect(DAPL_EP * ep_ptr,
        return DAT_SUCCESS;
 bail:
        dapl_log(DAPL_DBG_TYPE_ERR,
-                " connect ERROR: %s -> %s r_qual %d\n",
-                strerror(errno), 
+                " connect ERROR: -> %s r_qual %d\n",
                 inet_ntoa(((struct sockaddr_in *)r_addr)->sin_addr),
                 (unsigned int)r_qual);
 
@@ -629,9 +675,10 @@ static void dapli_socket_connect_rtu(dp_ib_cm_handle_t cm_ptr)
 
        len = recv(cm_ptr->socket, (char *)&cm_ptr->msg, exp, 0);
        if (len != exp || ntohs(cm_ptr->msg.ver) != DCM_VER) {
+               int err = dapl_socket_errno();
                dapl_log(DAPL_DBG_TYPE_WARN,
-                        " CONN_RTU read: sk %d ERR %s, rcnt=%d, v=%d -> %s PORT L-%x R-%x PID L-%x R-%x\n",
-                        cm_ptr->socket, strerror(errno), len, ntohs(cm_ptr->msg.ver),
+                        " CONN_RTU read: sk %d ERR 0x%x, rcnt=%d, v=%d -> %s PORT L-%x R-%x PID L-%x R-%x\n",
+                        cm_ptr->socket, err, len, ntohs(cm_ptr->msg.ver),
                         inet_ntoa(((struct sockaddr_in *)&cm_ptr->addr)->sin_addr),
                         ntohs(((struct sockaddr_in *)&cm_ptr->msg.daddr.so)->sin_port),
                         ntohs(((struct sockaddr_in *)&cm_ptr->addr)->sin_port),
@@ -639,7 +686,7 @@ static void dapli_socket_connect_rtu(dp_ib_cm_handle_t cm_ptr)
                         ntohs(*(uint16_t*)&cm_ptr->msg.resv[2]));
 
                /* Retry; corner case where server tcp stack resets under load */
-               if (dapl_socket_errno() == ECONNRESET) {
+               if (err == ECONNRESET) {
                        closesocket(cm_ptr->socket);
                        cm_ptr->socket = DAPL_INVALID_SOCKET;
                        dapli_socket_connect(cm_ptr->ep, (DAT_IA_ADDRESS_PTR)&cm_ptr->addr, 
@@ -692,9 +739,10 @@ static void dapli_socket_connect_rtu(dp_ib_cm_handle_t cm_ptr)
        if (exp) {
                len = recv(cm_ptr->socket, cm_ptr->msg.p_data, exp, 0);
                if (len != exp) {
+                       int err = dapl_socket_errno();
                        dapl_log(DAPL_DBG_TYPE_ERR,
-                                " CONN_RTU read pdata: ERR %s, rcnt=%d -> %s\n",
-                                strerror(errno), len,
+                                " CONN_RTU read pdata: ERR 0x%x %s, rcnt=%d -> %s\n",
+                                err, strerror(err), len,
                                 inet_ntoa(((struct sockaddr_in *)
                                            ep_ptr->param.
                                            remote_ia_address_ptr)->sin_addr));
@@ -761,8 +809,10 @@ static void dapli_socket_connect_rtu(dp_ib_cm_handle_t cm_ptr)
        cm_ptr->state = DCM_CONNECTED;
        cm_ptr->msg.op = ntohs(DCM_RTU);
        if (send(cm_ptr->socket, (char *)&cm_ptr->msg, 4, 0) == -1) {
+               int err = dapl_socket_errno();
                dapl_log(DAPL_DBG_TYPE_ERR,
-                        " CONN_RTU: write error = %s\n", strerror(errno));
+                        " CONN_RTU: write ERR = 0x%x %s\n", 
+                        err, strerror(err));
                goto bail;
        }
        /* post the event with private data */
@@ -821,6 +871,7 @@ ud_bail:
        } else
 #endif
        {
+               dapli_ep_check(cm_ptr->ep);
                dapl_evd_connection_callback(cm_ptr, event, cm_ptr->msg.p_data,
                                             DCM_MAX_PDATA_SIZE, ep_ptr);
        }
@@ -864,8 +915,10 @@ dapli_socket_listen(DAPL_IA * ia_ptr, DAT_CONN_QUAL serviceID, DAPL_SP * sp_ptr)
        /* bind, listen, set sockopt, accept, exchange data */
        if ((cm_ptr->socket =
             socket(AF_INET, SOCK_STREAM, IPPROTO_TCP)) == DAPL_INVALID_SOCKET) {
-               dapl_log(DAPL_DBG_TYPE_ERR, " ERR: listen socket create: %s\n",
-                        strerror(errno));
+               int err = dapl_socket_errno();
+               dapl_log(DAPL_DBG_TYPE_ERR, 
+                        " listen: socket create: ERR 0x%x %s\n",
+                        err, strerror(err));
                dat_status = DAT_INSUFFICIENT_RESOURCES;
                goto bail;
        }
@@ -877,10 +930,11 @@ dapli_socket_listen(DAPL_IA * ia_ptr, DAT_CONN_QUAL serviceID, DAPL_SP * sp_ptr)
 
        if ((bind(cm_ptr->socket, (struct sockaddr *)&addr, sizeof(addr)) < 0)
            || (listen(cm_ptr->socket, 128) < 0)) {
+               int err = dapl_socket_errno();
                dapl_log(DAPL_DBG_TYPE_CM,
-                        " listen: ERROR %s on port %d\n",
-                        strerror(errno), serviceID + 1000);
-               if (dapl_socket_errno() == EADDRINUSE)
+                        " listen: ERROR 0x%x %s on port %d\n",
+                        err, strerror(err), serviceID + 1000);
+               if (err == EADDRINUSE)
                        dat_status = DAT_CONN_QUAL_IN_USE;
                else
                        dat_status = DAT_CONN_QUAL_UNAVAILABLE;
@@ -933,9 +987,10 @@ static void dapli_socket_accept(ib_cm_srvc_handle_t cm_ptr)
                                        &acm_ptr->msg.daddr.so,
                                        (socklen_t *) &len);
                if (acm_ptr->socket == DAPL_INVALID_SOCKET) {
+                       int err = dapl_socket_errno();
                        dapl_log(DAPL_DBG_TYPE_ERR,
-                               " ACCEPT: ERR %s on FD %d l_cr %p\n",
-                               strerror(errno), cm_ptr->socket, cm_ptr);
+                               " ACCEPT: ERR 0x%x %s on FD %d l_cr %p\n",
+                               err, strerror(err), cm_ptr->socket, cm_ptr);
                        dapls_cm_release(acm_ptr);
                        return;
                }
@@ -948,11 +1003,14 @@ static void dapli_socket_accept(ib_cm_srvc_handle_t cm_ptr)
                /* no delay for small packets */
                ret = setsockopt(acm_ptr->socket, IPPROTO_TCP, TCP_NODELAY,
                           (char *)&opt, sizeof(opt));
-               if (ret)
+               if (ret) {
+                       int err = dapl_socket_errno();
                        dapl_log(DAPL_DBG_TYPE_ERR,
-                                " ACCEPT: NODELAY setsockopt: 0x%x 0x%x %s\n",
-                                ret, dapl_socket_errno(), strerror(dapl_socket_errno()));
-               
+                                " ACCEPT: NODELAY setsockopt:"
+                                " RET %d ERR 0x%x %s\n",
+                                ret, err, strerror(err));
+               }
+
                /* get local address information from socket */
                sl = sizeof(acm_ptr->addr);
                getsockname(acm_ptr->socket, (struct sockaddr *)&acm_ptr->addr, &sl);
@@ -975,9 +1033,10 @@ static void dapli_socket_accept_data(ib_cm_srvc_handle_t acm_ptr)
        /* read in DST QP info, IA address. check for private data */
        len = recv(acm_ptr->socket, (char *)&acm_ptr->msg, exp, 0);
        if (len != exp || ntohs(acm_ptr->msg.ver) != DCM_VER) {
+               int err = dapl_socket_errno();
                dapl_log(DAPL_DBG_TYPE_ERR,
-                        " ACCEPT read: ERR %s, rcnt=%d, ver=%d\n",
-                        strerror(errno), len, ntohs(acm_ptr->msg.ver));
+                        " ACCEPT read: ERR 0x%x %s, rcnt=%d, ver=%d\n",
+                        err, strerror(err), len, ntohs(acm_ptr->msg.ver));
                goto bail;
        }
 
@@ -996,9 +1055,10 @@ static void dapli_socket_accept_data(ib_cm_srvc_handle_t acm_ptr)
        if (exp) {
                len = recv(acm_ptr->socket, acm_ptr->msg.p_data, exp, 0);
                if (len != exp) {
+                       int err = dapl_socket_errno();
                        dapl_log(DAPL_DBG_TYPE_ERR,
-                                " accept read pdata: ERR %s, rcnt=%d\n",
-                                strerror(errno), len);
+                                " accept read pdata: ERR 0x%x %s, rcnt=%d\n",
+                                err, strerror(err), len);
                        goto bail;
                }
                p_data = acm_ptr->msg.p_data;
@@ -1143,6 +1203,7 @@ dapli_socket_accept_usr(DAPL_EP * ep_ptr,
        cm_ptr->hca = ia_ptr->hca_ptr;
        cm_ptr->state = DCM_ACCEPTED;
 
+       /* Link CM to EP, already queued on work thread */
        dapl_ep_link_cm(ep_ptr, cm_ptr);
        cm_ptr->ep = ep_ptr;
 
@@ -1158,9 +1219,10 @@ dapli_socket_accept_usr(DAPL_EP * ep_ptr,
                len = writev(cm_ptr->socket, iov, 1);
        
        if (len != (p_size + exp)) {
+               int err = dapl_socket_errno();
                dapl_log(DAPL_DBG_TYPE_ERR,
-                        " ACCEPT_USR: ERR %s, wcnt=%d -> %s\n",
-                        strerror(errno), len, 
+                        " ACCEPT_USR: ERR 0x%x %s, wcnt=%d -> %s\n",
+                        err, strerror(err), len, 
                         inet_ntoa(((struct sockaddr_in *)
                                   &cm_ptr->msg.daddr.so)->sin_addr));
                dapl_ep_unlink_cm(ep_ptr, cm_ptr);
@@ -1180,6 +1242,7 @@ dapli_socket_accept_usr(DAPL_EP * ep_ptr,
                     htonll(*(uint64_t*)&local.saddr.ib.gid[8]));
 
        dapl_dbg_log(DAPL_DBG_TYPE_EP, " PASSIVE: accepted!\n");
+
        return DAT_SUCCESS;
 bail:
        /* schedule cleanup from workq */
@@ -1261,6 +1324,7 @@ ud_bail:
        } else 
 #endif
        {
+               dapli_ep_check(cm_ptr->ep);
                dapls_cr_callback(cm_ptr, event, NULL, 0, cm_ptr->sp);
        }
        return;