[dpdk-stable] [PATCH] telemetry: fix error when using ports of different types

[dpdk-stable] [PATCH] telemetry: fix error when using ports of different types

[Bruce Richardson]

Different NIC ports can have different numbers of xstats on them, which
means that we can't just use the xstats list from the first port registered
in the telemetry library. Instead, we need to check the type of each port -
by checking its ops structure pointer - and register each port type once
with the metrics lib.

CC: stable@dpdk.org
Fixes: fdbdb3f9ce46 ("telemetry: add initial connection socket")

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 lib/librte_telemetry/rte_telemetry.c          | 40 ++++++++++++++-----
 lib/librte_telemetry/rte_telemetry_internal.h |  2 +-
 2 files changed, 32 insertions(+), 10 deletions(-)

diff --git a/lib/librte_telemetry/rte_telemetry.c b/lib/librte_telemetry/rte_telemetry.c
index 016431f12..7fb247eaa 100644
--- a/lib/librte_telemetry/rte_telemetry.c
+++ b/lib/librte_telemetry/rte_telemetry.c
@@ -558,7 +558,7 @@ rte_telemetry_send_ports_stats_values(uint32_t *metric_ids, int num_metric_ids,
 		}
 
 		ret = rte_telemetry_update_metrics_ethdev(telemetry,
-				port_ids[i], telemetry->reg_index);
+				port_ids[i], telemetry->reg_index[i]);
 		if (ret < 0) {
 			TELEMETRY_LOG_ERR("Failed to update ethdev metrics");
 			return -1;
@@ -658,23 +658,45 @@ rte_telemetry_reg_ethdev_to_metrics(uint16_t port_id)
 static int32_t
 rte_telemetry_initial_accept(struct telemetry_impl *telemetry)
 {
+	struct driver_index {
+		const void *dev_ops;
+		int reg_index;
+	} drv_idx[RTE_MAX_ETHPORTS];
+	int nb_drv_idx = 0;
 	uint16_t pid;
 	int ret;
 	int selftest = 0;
 
 	RTE_ETH_FOREACH_DEV(pid) {
-		telemetry->reg_index = rte_telemetry_reg_ethdev_to_metrics(pid);
-		break;
-	}
+		int i;
+		/* Different device types have different numbers of stats, so
+		 * first check if the stats for this type of device have
+		 * already been registered
+		 */
+		for (i = 0; i < nb_drv_idx; i++) {
+			if (rte_eth_devices[pid].dev_ops == drv_idx[i].dev_ops) {
+				telemetry->reg_index[pid] = drv_idx[i].reg_index;
+				break;
+			}
+		}
+		if (i < nb_drv_idx)
+			continue; /* we found a match, go to next port */
 
-	if (telemetry->reg_index < 0) {
-		TELEMETRY_LOG_ERR("Failed to register ethdev metrics");
-		return -1;
+		/* No match, register a new set of xstats for this port */
+		ret = rte_telemetry_reg_ethdev_to_metrics(pid);
+		if (ret < 0) {
+			TELEMETRY_LOG_ERR("Failed to register ethdev metrics");
+			return -1;
+		}
+		telemetry->reg_index[pid] = ret;
+		drv_idx[nb_drv_idx].dev_ops = rte_eth_devices[pid].dev_ops;
+		drv_idx[nb_drv_idx].reg_index = ret;
+		nb_drv_idx++;
 	}
 
 	telemetry->metrics_register_done = 1;
 	if (selftest) {
-		ret = rte_telemetry_socket_messaging_testing(telemetry->reg_index,
+		ret = rte_telemetry_socket_messaging_testing(telemetry->reg_index[0],
 				telemetry->server_fd);
 		if (ret < 0)
 			return -1;
@@ -1299,7 +1321,7 @@ rte_telemetry_socket_messaging_testing(int index, int socket)
 	}
 
 	telemetry->server_fd = socket;
-	telemetry->reg_index = index;
+	telemetry->reg_index[0] = index;
 	TELEMETRY_LOG_INFO("Beginning Telemetry socket message Selftest");
 	rte_telemetry_socket_test_setup(telemetry, &send_fd, &recv_fd);
 	TELEMETRY_LOG_INFO("Register valid client test");
diff --git a/lib/librte_telemetry/rte_telemetry_internal.h b/lib/librte_telemetry/rte_telemetry_internal.h
index de7afda30..c298c3919 100644
--- a/lib/librte_telemetry/rte_telemetry_internal.h
+++ b/lib/librte_telemetry/rte_telemetry_internal.h
@@ -36,7 +36,7 @@ typedef struct telemetry_impl {
 	pthread_t thread_id;
 	int thread_status;
 	uint32_t socket_id;
-	int reg_index;
+	int reg_index[RTE_MAX_ETHPORTS];
 	int metrics_register_done;
 	TAILQ_HEAD(, telemetry_client) client_list_head;
 	struct telemetry_client *request_client;
-- 
2.19.2

Re: [dpdk-stable] [dpdk-dev] [PATCH] telemetry: fix error when using ports of different types

[Thomas Monjalon]

19/12/2018 12:59, Bruce Richardson:
> Different NIC ports can have different numbers of xstats on them, which
> means that we can't just use the xstats list from the first port registered
> in the telemetry library. Instead, we need to check the type of each port -
> by checking its ops structure pointer - and register each port type once
> with the metrics lib.
> 
> CC: stable@dpdk.org
> Fixes: fdbdb3f9ce46 ("telemetry: add initial connection socket")
> 
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>

Kevin, any comment please?

Re: [dpdk-stable] [PATCH] telemetry: fix error when using ports of different types

[Laatz, Kevin]

On 19/12/2018 11:59, Bruce Richardson wrote:
> Different NIC ports can have different numbers of xstats on them, which
> means that we can't just use the xstats list from the first port registered
> in the telemetry library. Instead, we need to check the type of each port -
> by checking its ops structure pointer - and register each port type once
> with the metrics lib.
>
> CC: stable@dpdk.org
> Fixes: fdbdb3f9ce46 ("telemetry: add initial connection socket")
>
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>

Acked-by: Kevin Laatz <kevin.laatz@intel.com>

Re: [dpdk-stable] [PATCH] telemetry: fix error when using ports of different types

[Thomas Monjalon]

21/12/2018 14:27, Laatz, Kevin:
> On 19/12/2018 11:59, Bruce Richardson wrote:
> > Different NIC ports can have different numbers of xstats on them, which
> > means that we can't just use the xstats list from the first port registered
> > in the telemetry library. Instead, we need to check the type of each port -
> > by checking its ops structure pointer - and register each port type once
> > with the metrics lib.
> >
> > CC: stable@dpdk.org
> > Fixes: fdbdb3f9ce46 ("telemetry: add initial connection socket")
> >
> > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> 
> Acked-by: Kevin Laatz <kevin.laatz@intel.com>

Applied, thanks