Imported Upstream version 16.11.2

Change-Id: I947038e46a2c747296dc7aa7522239733ca2f659
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>

5 files changed, 59 insertions, 13 deletions

diff --git a/examples/performance-thread/common/lthread_timer.h b/examples/performance-thread/common/lthread_timer.h
index b5e6fb0e..2d147577 100644
--- a/examples/performance-thread/common/lthread_timer.h
+++ b/examples/performance-thread/common/lthread_timer.h
@@ -42,11 +42,22 @@
 static inline uint64_t
 _ns_to_clks(uint64_t ns)
 {
-	unsigned __int128 clkns = rte_get_tsc_hz();
+	/*
+	 * clkns needs to be divided by 1E9 to get ns clocks. However,
+	 * dividing by this first would lose a lot of accuracy.
+	 * Dividing after a multiply by ns, could cause overflow of
+	 * uint64_t if ns is about 5 seconds [if we assume a max tsc
+	 * rate of 4GHz]. Therefore we first divide by 1E4, then
+	 * multiply and finally divide by 1E5. This allows ns to be
+	 * values many hours long, without overflow, while still keeping
+	 * reasonable accuracy.
+	 */
+	uint64_t clkns = rte_get_tsc_hz() / 1e4;
 
 	clkns *= ns;
-	clkns /= 1000000000;
-	return (uint64_t) clkns;
+	clkns /= 1e5;
+
+	return clkns;
 }
 
 
diff --git a/examples/performance-thread/l3fwd-thread/main.c b/examples/performance-thread/l3fwd-thread/main.c
index fdc90b28..dd403ca8 100644
--- a/examples/performance-thread/l3fwd-thread/main.c
+++ b/examples/performance-thread/l3fwd-thread/main.c
@@ -90,6 +90,10 @@
 #define APP_LOOKUP_METHOD             APP_LOOKUP_LPM
 #endif
 
+#ifndef __GLIBC__ /* sched_getcpu() is glibc specific */
+#define sched_getcpu() rte_lcore_id()
+#endif
+
 /*
  *  When set to zero, simple forwaring path is eanbled.
  *  When set to one, optimized forwarding path is enabled.
@@ -282,7 +286,7 @@ static struct rte_eth_conf port_conf = {
 		.hw_ip_checksum = 1, /**< IP checksum offload enabled */
 		.hw_vlan_filter = 0, /**< VLAN filtering disabled */
 		.jumbo_frame    = 0, /**< Jumbo Frame Support disabled */
-		.hw_strip_crc   = 0, /**< CRC stripped by hardware */
+		.hw_strip_crc   = 1, /**< CRC stripped by hardware */
 	},
 	.rx_adv_conf = {
 		.rss_conf = {
diff --git a/examples/performance-thread/pthread_shim/main.c b/examples/performance-thread/pthread_shim/main.c
index f0357218..850b009d 100644
--- a/examples/performance-thread/pthread_shim/main.c
+++ b/examples/performance-thread/pthread_shim/main.c
@@ -59,6 +59,10 @@
 #define DEBUG_APP 0
 #define HELLOW_WORLD_MAX_LTHREADS 10
 
+#ifndef __GLIBC__ /* sched_getcpu() is glibc-specific */
+#define sched_getcpu() rte_lcore_id()
+#endif
+
 __thread int print_count;
 __thread pthread_mutex_t print_lock;
 
@@ -175,12 +179,12 @@ static void initial_lthread(void *args __attribute__((unused)))
 		 * use an attribute to pass the desired lcore
 		 */
 		pthread_attr_t attr;
-		cpu_set_t cpuset;
+		rte_cpuset_t cpuset;
 
 		CPU_ZERO(&cpuset);
 		CPU_SET(lcore, &cpuset);
 		pthread_attr_init(&attr);
-		pthread_attr_setaffinity_np(&attr, sizeof(cpu_set_t), &cpuset);
+		pthread_attr_setaffinity_np(&attr, sizeof(rte_cpuset_t), &cpuset);
 
 		/* create the thread */
 		pthread_create(&tid[i], &attr, helloworld_pthread, (void *) i);
diff --git a/examples/performance-thread/pthread_shim/pthread_shim.c b/examples/performance-thread/pthread_shim/pthread_shim.c
index 0d6100c9..113bafa0 100644
--- a/examples/performance-thread/pthread_shim/pthread_shim.c
+++ b/examples/performance-thread/pthread_shim/pthread_shim.c
@@ -48,6 +48,21 @@
 
 #define POSIX_ERRNO(x)  (x)
 
+/* some releases of FreeBSD 10, e.g. 10.0, don't have CPU_COUNT macro */
+#ifndef CPU_COUNT
+#define CPU_COUNT(x) __cpu_count(x)
+
+static inline unsigned int
+__cpu_count(const rte_cpuset_t *cpuset)
+{
+	unsigned int i, count = 0;
+	for (i = 0; i < RTE_MAX_LCORE; i++)
+		if (CPU_ISSET(i, cpuset))
+			count++;
+	return count;
+}
+#endif
+
 /*
  * this flag determines at run time if we override pthread
  * calls and map then to equivalent lthread calls
@@ -159,7 +174,7 @@ int (*f_pthread_setschedparam)
 int (*f_pthread_yield)
 	(void);
 int (*f_pthread_setaffinity_np)
-	(pthread_t thread, size_t cpusetsize, const cpu_set_t *cpuset);
+	(pthread_t thread, size_t cpusetsize, const rte_cpuset_t *cpuset);
 int (*f_nanosleep)
 	(const struct timespec *req, struct timespec *rem);
 } _sys_pthread_funcs = {
@@ -390,11 +405,11 @@ pthread_create(pthread_t *__restrict tid,
 
 		if (attr != NULL) {
 			/* determine CPU being requested */
-			cpu_set_t cpuset;
+			rte_cpuset_t cpuset;
 
 			CPU_ZERO(&cpuset);
 			pthread_attr_getaffinity_np(attr,
-						sizeof(cpu_set_t),
+						sizeof(rte_cpuset_t),
 						&cpuset);
 
 			if (CPU_COUNT(&cpuset) != 1)
@@ -576,15 +591,26 @@ int pthread_rwlock_wrlock(pthread_rwlock_t *a)
 	return _sys_pthread_funcs.f_pthread_rwlock_wrlock(a);
 }
 
-int pthread_yield(void)
+#ifdef RTE_EXEC_ENV_LINUXAPP
+int
+pthread_yield(void)
 {
 	if (override) {
 		lthread_yield();
 		return 0;
 	}
 	return _sys_pthread_funcs.f_pthread_yield();
-
 }
+#else
+void
+pthread_yield(void)
+{
+	if (override)
+		lthread_yield();
+	else
+		_sys_pthread_funcs.f_pthread_yield();
+}
+#endif
 
 pthread_t pthread_self(void)
 {
@@ -686,7 +712,7 @@ int nanosleep(const struct timespec *req, struct timespec *rem)
 
 int
 pthread_setaffinity_np(pthread_t thread, size_t cpusetsize,
-		       const cpu_set_t *cpuset)
+		       const rte_cpuset_t *cpuset)
 {
 	if (override) {
 		/* we only allow affinity with a single CPU */
diff --git a/examples/performance-thread/pthread_shim/pthread_shim.h b/examples/performance-thread/pthread_shim/pthread_shim.h
index 78bbb5ac..10f87894 100644
--- a/examples/performance-thread/pthread_shim/pthread_shim.h
+++ b/examples/performance-thread/pthread_shim/pthread_shim.h
@@ -33,7 +33,8 @@
 
 #ifndef _PTHREAD_SHIM_H_
 #define _PTHREAD_SHIM_H_
-#include <pthread.h>
+
+#include <rte_lcore.h>
 
 /*
  * This pthread shim is an example that demonstrates how legacy code