nvmf,direct: add AER support in direct mode.

After checking the code, aerl in our session is 0,
so there will be only 1 AER. So currently,
we will only handle 1 AER case.

When the AER event is triggered by real NVMe device owned
by the subsystem, it notifies all sessions belonging to
the subsystem.

Change-Id: Ia80fb0f03e893c20d8dd14afbed8db10db38301c
Signed-off-by: Ziye Yang <ziye.yang@intel.com>

autotest.sh

@@ -125,6 +125,7 @@ fi
 
 timing_enter host
 
+run_test test/nvmf/host/aer.sh
 run_test test/nvmf/host/identify.sh
 run_test test/nvmf/host/perf.sh
 run_test test/nvmf/host/identify_kernel_nvmf.sh

include/spdk/nvmf.h

@@ -84,6 +84,11 @@ struct spdk_nvmf_host {
 };
 
 struct spdk_nvmf_ctrlr_ops {
+	/**
+	 * Set NVMe ctrlr AER.
+	 */
+	void (*set_aer_callback)(struct spdk_nvmf_subsystem *subsys);
+
 	/**
 	 * Get NVMe identify controller data.
 	 */

lib/nvmf/direct.c

@@ -177,9 +177,9 @@ nvmf_direct_ctrlr_process_admin_cmd(struct spdk_nvmf_request *req)
 		break;
 	case SPDK_NVME_OPC_ASYNC_EVENT_REQUEST:
 		SPDK_TRACELOG(SPDK_TRACE_NVMF, "Async Event Request\n");
-		/* TODO: Just release the request as consumed. AER events will never
+		session->aer_req = req;
-		 * be triggered. */
+
-		return SPDK_NVMF_REQUEST_EXEC_STATUS_RELEASE;
+		return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS;
 	case SPDK_NVME_OPC_KEEP_ALIVE:
 		SPDK_TRACELOG(SPDK_TRACE_NVMF, "Keep Alive\n");
 		/*
@@ -249,7 +249,29 @@ nvmf_direct_ctrlr_detach(struct spdk_nvmf_subsystem *subsystem)
 	}
 }
 
+static void
+nvmf_direct_ctrlr_complete_aer(void *arg, const struct spdk_nvme_cpl *cpl)
+{
+	struct spdk_nvmf_subsystem *subsystem = (struct spdk_nvmf_subsystem *) arg;
+	struct spdk_nvmf_session *session;
+
+	TAILQ_FOREACH(session, &subsystem->sessions, link) {
+		if (session->aer_req) {
+			nvmf_direct_ctrlr_complete_cmd(session->aer_req, cpl);
+			session->aer_req = NULL;
+		}
+	}
+}
+
+static void
+nvmf_direct_ctrlr_set_aer_callback(struct spdk_nvmf_subsystem *subsys)
+{
+	spdk_nvme_ctrlr_register_aer_callback(subsys->dev.direct.ctrlr,
+					      nvmf_direct_ctrlr_complete_aer, subsys);
+}
+
 const struct spdk_nvmf_ctrlr_ops spdk_nvmf_direct_ctrlr_ops = {
+	.set_aer_callback		= nvmf_direct_ctrlr_set_aer_callback,
 	.ctrlr_get_data			= nvmf_direct_ctrlr_get_data,
 	.process_admin_cmd		= nvmf_direct_ctrlr_process_admin_cmd,
 	.process_io_cmd			= nvmf_direct_ctrlr_process_io_cmd,

lib/nvmf/session.h

@@ -91,6 +91,7 @@ struct spdk_nvmf_session {
 			uint8_t fw_activation_notice : 1;
 		} bits;
 	} async_event_config;
+	struct spdk_nvmf_request *aer_req;
 	uint8_t hostid[16];
 	const struct spdk_nvmf_transport	*transport;
 

lib/nvmf/subsystem.c

@@ -311,6 +311,8 @@ nvmf_subsystem_add_ctrlr(struct spdk_nvmf_subsystem *subsystem,
 		SPDK_ERRLOG("spdk_nvme_ctrlr_alloc_io_qpair() failed\n");
 		return -1;
 	}
+
+	subsystem->ops->set_aer_callback(subsystem);
 	return 0;
 }
 

lib/nvmf/virtual.c

@@ -532,7 +532,13 @@ nvmf_virtual_ctrlr_detach(struct spdk_nvmf_subsystem *subsystem)
 	}
 }
 
+static void
+nvmf_virtual_ctrlr_set_aer_callback(struct spdk_nvmf_subsystem *subsys)
+{
+}
+
 const struct spdk_nvmf_ctrlr_ops spdk_nvmf_virtual_ctrlr_ops = {
+	.set_aer_callback		= nvmf_virtual_ctrlr_set_aer_callback,
 	.ctrlr_get_data			= nvmf_virtual_ctrlr_get_data,
 	.process_admin_cmd		= nvmf_virtual_ctrlr_process_admin_cmd,
 	.process_io_cmd			= nvmf_virtual_ctrlr_process_io_cmd,

test/lib/nvme/aer/aer.c

@@ -31,13 +31,17 @@
  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+#include <inttypes.h>
 #include <stdbool.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <unistd.h>
+#include <string.h>
 
 #include <rte_config.h>
 #include <rte_lcore.h>
 
+#include "spdk/log.h"
 #include "spdk/nvme.h"
 #include "spdk/env.h"
 
@@ -62,6 +66,7 @@ static int aer_done = 0;
 
 static int temperature_done = 0;
 static int failed = 0;
+static struct spdk_nvme_transport_id g_trid;
 
 static void set_feature_completion(void *cb_arg, const struct spdk_nvme_cpl *cpl)
 {
@@ -183,6 +188,69 @@ static void aer_cb(void *arg, const struct spdk_nvme_cpl *cpl)
 	get_health_log_page(dev);
 }
 
+static void
+usage(const char *program_name)
+{
+	printf("%s [options]", program_name);
+	printf("\n");
+	printf("options:\n");
+	printf(" -r trid    remote NVMe over Fabrics target address\n");
+	printf("    Format: 'key:value [key:value] ...'\n");
+	printf("    Keys:\n");
+	printf("     trtype      Transport type (e.g. RDMA)\n");
+	printf("     adrfam      Address family (e.g. IPv4, IPv6)\n");
+	printf("     traddr      Transport address (e.g. 192.168.100.8)\n");
+	printf("     trsvcid     Transport service identifier (e.g. 4420)\n");
+	printf("     subnqn      Subsystem NQN (default: %s)\n", SPDK_NVMF_DISCOVERY_NQN);
+	printf("    Example: -r 'trtype:RDMA adrfam:IPv4 traddr:192.168.100.8 trsvcid:4420'\n");
+
+	spdk_tracelog_usage(stdout, "-t");
+
+	printf(" -v         verbose (enable warnings)\n");
+	printf(" -H         show this usage\n");
+}
+
+static int
+parse_args(int argc, char **argv)
+{
+	int op, rc;
+
+	g_trid.trtype = SPDK_NVME_TRANSPORT_PCIE;
+	snprintf(g_trid.subnqn, sizeof(g_trid.subnqn), "%s", SPDK_NVMF_DISCOVERY_NQN);
+
+	while ((op = getopt(argc, argv, "r:t:H")) != -1) {
+		switch (op) {
+		case 't':
+			rc = spdk_log_set_trace_flag(optarg);
+			if (rc < 0) {
+				fprintf(stderr, "unknown flag\n");
+				usage(argv[0]);
+				exit(EXIT_FAILURE);
+			}
+#ifndef DEBUG
+			fprintf(stderr, "%s must be rebuilt with CONFIG_DEBUG=y for -t flag.\n",
+				argv[0]);
+			usage(argv[0]);
+			return 0;
+#endif
+			break;
+		case 'r':
+			if (spdk_nvme_transport_id_parse(&g_trid, optarg) != 0) {
+				fprintf(stderr, "Error parsing transport address\n");
+				return 1;
+			}
+			break;
+		case 'H':
+		default:
+			usage(argv[0]);
+			return 1;
+		}
+	}
+
+	optind = 1;
+
+	return 0;
+}
 
 static bool
 probe_cb(void *cb_ctx, const struct spdk_nvme_transport_id *trid,
@@ -221,6 +289,12 @@ int main(int argc, char **argv)
 	struct dev		*dev;
 	int			i;
 	struct spdk_env_opts	opts;
+	int			rc;
+
+	rc = parse_args(argc, argv);
+	if (rc != 0) {
+		return rc;
+	}
 
 	spdk_env_opts_init(&opts);
 	opts.name = "aer";
@@ -229,7 +303,7 @@ int main(int argc, char **argv)
 
 	printf("Asynchronous Event Request test\n");
 
-	if (spdk_nvme_probe(NULL, NULL, probe_cb, attach_cb, NULL) != 0) {
+	if (spdk_nvme_probe(&g_trid, NULL, probe_cb, attach_cb, NULL) != 0) {
 		fprintf(stderr, "spdk_nvme_probe() failed\n");
 		return 1;
 	}

test/lib/nvmf/direct/direct_ut.c

@@ -66,6 +66,13 @@ spdk_nvme_ctrlr_get_data(struct spdk_nvme_ctrlr *ctrlr)
 	return NULL;
 }
 
+void
+spdk_nvme_ctrlr_register_aer_callback(struct spdk_nvme_ctrlr *ctrlr,
+				      spdk_nvme_aer_cb aer_cb_fn,
+				      void *aer_cb_arg)
+{
+}
+
 int
 spdk_nvme_ctrlr_cmd_io_raw(struct spdk_nvme_ctrlr *ctrlr,
 			   struct spdk_nvme_qpair *qpair,

test/nvmf/host/aer.sh

@@ -0,0 +1,41 @@
+#!/usr/bin/env bash
+
+testdir=$(readlink -f $(dirname $0))
+rootdir=$(readlink -f $testdir/../../..)
+source $rootdir/scripts/autotest_common.sh
+source $rootdir/test/nvmf/common.sh
+
+rpc_py="python $rootdir/scripts/rpc.py"
+
+set -e
+
+if ! rdma_nic_available; then
+        echo "no NIC for nvmf test"
+        exit 0
+fi
+
+timing_enter aer
+
+# Start up the NVMf target in another process
+$rootdir/app/nvmf_tgt/nvmf_tgt -c $testdir/../nvmf.conf -m 0x2 -p 1 -s 512 -t nvmf &
+nvmfpid=$!
+
+trap "killprocess $nvmfpid; exit 1" SIGINT SIGTERM EXIT
+
+waitforlisten $nvmfpid ${RPC_PORT}
+
+$rpc_py construct_nvmf_subsystem Direct nqn.2016-06.io.spdk:cnode1 'transport:RDMA traddr:192.168.100.8 trsvcid:4420' '' -p "*"
+
+$rootdir/test/lib/nvme/aer/aer -r "\
+        trtype:RDMA \
+        adrfam:IPv4 \
+        traddr:$NVMF_FIRST_TARGET_IP \
+        trsvcid:$NVMF_PORT \
+        subnqn:nqn.2014-08.org.nvmexpress.discovery" -t all
+sync
+$rpc_py delete_nvmf_subsystem nqn.2016-06.io.spdk:cnode1
+
+trap - SIGINT SIGTERM EXIT
+
+killprocess $nvmfpid
+timing_exit aer