distributor: fix crash in flush
authorDavid Hunt <david.hunt@intel.com>
Fri, 14 Apr 2017 08:59:53 +0000 (09:59 +0100)
committerThomas Monjalon <thomas@monjalon.net>
Thu, 20 Apr 2017 23:03:35 +0000 (01:03 +0200)
Occasionally, the distributor single-packet-at-a-time mode will
segfault because it inadvertently calls some burst mode code when
flushing packets.

This patch ensures that only the v20 (single mode) codepath is used, and
returns without falling through to the burst mode code.

Fixes: 775003ad2f96 ("distributor: add new burst-capable library")

Signed-off-by: David Hunt <david.hunt@intel.com>
Tested-by: Yong Liu <yong.liu@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
lib/librte_distributor/rte_distributor.c

index 4725904..e4dfa7f 100644 (file)
@@ -555,7 +555,7 @@ total_outstanding(const struct rte_distributor *d)
 int
 rte_distributor_flush_v1705(struct rte_distributor *d)
 {
-       const unsigned int flushed = total_outstanding(d);
+       unsigned int flushed;
        unsigned int wkr;
 
        if (d->alg_type == RTE_DIST_ALG_SINGLE) {
@@ -563,6 +563,8 @@ rte_distributor_flush_v1705(struct rte_distributor *d)
                return rte_distributor_flush_v20(d->d_v20);
        }
 
+       flushed = total_outstanding(d);
+
        while (total_outstanding(d) > 0)
                rte_distributor_process(d, NULL, 0);
 
@@ -590,6 +592,7 @@ rte_distributor_clear_returns_v1705(struct rte_distributor *d)
        if (d->alg_type == RTE_DIST_ALG_SINGLE) {
                /* Call the old API */
                rte_distributor_clear_returns_v20(d->d_v20);
+               return;
        }
 
        /* throw away returns, so workers can exit */