latency: fix calculation for multi-thread
authorReshma Pattan <reshma.pattan@intel.com>
Thu, 5 Dec 2019 17:18:11 +0000 (17:18 +0000)
committerThomas Monjalon <thomas@monjalon.net>
Mon, 20 Jan 2020 00:32:50 +0000 (01:32 +0100)
Make latency calculation multithread safe by
using spinlock.

Fixes: 5cd3cac9ed ("latency: added new library for latency stats")
Cc: stable@dpdk.org
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
lib/librte_latencystats/rte_latencystats.c

index 98e0189..ba2fff3 100644 (file)
@@ -42,6 +42,7 @@ struct rte_latency_stats {
        float avg_latency; /**< Average latency in nano seconds */
        float max_latency; /**< Maximum latency in nano seconds */
        float jitter; /** Latency variation */
+       rte_spinlock_t lock; /** Latency calculation lock */
 };
 
 static struct rte_latency_stats *glob_stats;
@@ -164,6 +165,7 @@ calc_latency(uint16_t pid __rte_unused,
                        latency[cnt++] = now - pkts[i]->timestamp;
        }
 
+       rte_spinlock_lock(&glob_stats->lock);
        for (i = 0; i < cnt; i++) {
                /*
                 * The jitter is calculated as statistical mean of interpacket
@@ -193,6 +195,7 @@ calc_latency(uint16_t pid __rte_unused,
                        alpha * (latency[i] - glob_stats->avg_latency);
                prev_latency = latency[i];
        }
+       rte_spinlock_unlock(&glob_stats->lock);
 
        return nb_pkts;
 }
@@ -223,6 +226,7 @@ rte_latencystats_init(uint64_t app_samp_intvl,
        }
 
        glob_stats = mz->addr;
+       rte_spinlock_init(&glob_stats->lock);
        samp_intvl = app_samp_intvl * latencystat_cycles_per_ns();
 
        /** Register latency stats with stats library */