net/failsafe: support xstats
authorStephen Hemminger <stephen@networkplumber.org>
Fri, 1 Nov 2019 20:12:56 +0000 (13:12 -0700)
committerFerruh Yigit <ferruh.yigit@intel.com>
Mon, 11 Nov 2019 13:23:02 +0000 (14:23 +0100)
Add support for extended statistics in failsafe driver.
Reports detailed statistics for each sub device.

Example:

testpmd> show port xstats 1
rx_good_packets: 0
tx_good_packets: 0
rx_good_bytes: 0
tx_good_bytes: 0
rx_missed_errors: 0
rx_errors: 0
tx_errors: 0
rx_mbuf_allocation_errors: 0
rx_q0packets: 0
rx_q0bytes: 0
rx_q0errors: 0
tx_q0packets: 0
tx_q0bytes: 0
rx_sub0_good_packets: 0
tx_sub0_good_packets: 0
...
rx_sub1_good_packets: 0
tx_sub1_good_packets: 0
rx_sub1_good_bytes: 0

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
drivers/net/failsafe/failsafe_ops.c

index 0afae6c..a87e49b 100644 (file)
@@ -14,6 +14,7 @@
 #include <rte_flow.h>
 #include <rte_cycles.h>
 #include <rte_ethdev.h>
+#include <rte_string_fns.h>
 
 #include "failsafe_private.h"
 
@@ -881,6 +882,154 @@ fs_stats_reset(struct rte_eth_dev *dev)
        return 0;
 }
 
+static int
+__fs_xstats_count(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       int count = 0;
+       uint8_t i;
+       int ret;
+
+       FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
+               ret = rte_eth_xstats_get_names(PORT_ID(sdev), NULL, 0);
+               if (ret < 0)
+                       return ret;
+               count += ret;
+       }
+
+       return count;
+}
+
+static int
+__fs_xstats_get_names(struct rte_eth_dev *dev,
+                   struct rte_eth_xstat_name *xstats_names,
+                   unsigned int limit)
+{
+       struct sub_device *sdev;
+       unsigned int count = 0;
+       uint8_t i;
+
+       /* Caller only cares about count */
+       if (!xstats_names)
+               return  __fs_xstats_count(dev);
+
+       FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
+               struct rte_eth_xstat_name *sub_names = xstats_names + count;
+               int j, r;
+
+               if (count >= limit)
+                       break;
+
+               r = rte_eth_xstats_get_names(PORT_ID(sdev),
+                                            sub_names, limit - count);
+               if (r < 0)
+                       return r;
+
+               /* add subN_ prefix to names */
+               for (j = 0; j < r; j++) {
+                       char *xname = sub_names[j].name;
+                       char tmp[RTE_ETH_XSTATS_NAME_SIZE];
+
+                       if ((xname[0] == 't' || xname[0] == 'r') &&
+                           xname[1] == 'x' && xname[2] == '_')
+                               snprintf(tmp, sizeof(tmp), "%.3ssub%u_%s",
+                                        xname, i, xname + 3);
+                       else
+                               snprintf(tmp, sizeof(tmp), "sub%u_%s",
+                                        i, xname);
+
+                       strlcpy(xname, tmp, RTE_ETH_XSTATS_NAME_SIZE);
+               }
+               count += r;
+       }
+       return count;
+}
+
+static int
+fs_xstats_get_names(struct rte_eth_dev *dev,
+                   struct rte_eth_xstat_name *xstats_names,
+                   unsigned int limit)
+{
+       int ret;
+
+       fs_lock(dev, 0);
+       ret = __fs_xstats_get_names(dev, xstats_names, limit);
+       fs_unlock(dev, 0);
+       return ret;
+}
+
+static int
+__fs_xstats_get(struct rte_eth_dev *dev,
+             struct rte_eth_xstat *xstats,
+             unsigned int n)
+{
+       unsigned int count = 0;
+       struct sub_device *sdev;
+       uint8_t i;
+       int j, ret;
+
+       ret = __fs_xstats_count(dev);
+       /*
+        * if error
+        * or caller did not give enough space
+        * or just querying
+        */
+       if (ret < 0 || ret > (int)n || xstats == NULL)
+               return ret;
+
+       FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
+               ret = rte_eth_xstats_get(PORT_ID(sdev), xstats, n);
+               if (ret < 0)
+                       return ret;
+
+               if (ret > (int)n)
+                       return n + count;
+
+               /* add offset to id's from sub-device */
+               for (j = 0; j < ret; j++)
+                       xstats[j].id += count;
+
+               xstats += ret;
+               n -= ret;
+               count += ret;
+       }
+
+       return count;
+}
+
+static int
+fs_xstats_get(struct rte_eth_dev *dev,
+             struct rte_eth_xstat *xstats,
+             unsigned int n)
+{
+       int ret;
+
+       fs_lock(dev, 0);
+       ret = __fs_xstats_get(dev, xstats, n);
+       fs_unlock(dev, 0);
+
+       return ret;
+}
+
+
+static int
+fs_xstats_reset(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+       int r = 0;
+
+       fs_lock(dev, 0);
+       FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
+               r = rte_eth_xstats_reset(PORT_ID(sdev));
+               if (r < 0)
+                       break;
+       }
+       fs_unlock(dev, 0);
+
+       return r;
+}
+
 static void
 fs_dev_merge_desc_lim(struct rte_eth_desc_lim *to,
                      const struct rte_eth_desc_lim *from)
@@ -1331,6 +1480,9 @@ const struct eth_dev_ops failsafe_ops = {
        .link_update = fs_link_update,
        .stats_get = fs_stats_get,
        .stats_reset = fs_stats_reset,
+       .xstats_get = fs_xstats_get,
+       .xstats_get_names = fs_xstats_get_names,
+       .xstats_reset = fs_xstats_reset,
        .dev_infos_get = fs_dev_infos_get,
        .dev_supported_ptypes_get = fs_dev_supported_ptypes_get,
        .mtu_set = fs_mtu_set,