patch 'eal/unix: optimize thread creation' has been queued to stable release 22.11.7

[luca.boccassi at gmail.com]

Hi,

FYI, your patch has been queued to stable release 22.11.7

Note it hasn't been pushed to http://dpdk.org/browse/dpdk-stable yet.
It will be pushed if I get no objections before 11/14/24. So please
shout if anyone has objections.

Also note that after the patch there's a diff of the upstream commit vs the
patch applied to the branch. This will indicate if there was any rebasing
needed to apply to the stable branch. If there were code changes for rebasing
(ie: not only metadata diffs), please double check that the rebase was
correctly done.

Queued patches are on a temporary branch at:
https://github.com/bluca/dpdk-stable

This queued commit can be viewed at:
https://github.com/bluca/dpdk-stable/commit/02097752bb88526197af8b0dd0a8da9e6321cf9a

Thanks.

Luca Boccassi

---
>From 02097752bb88526197af8b0dd0a8da9e6321cf9a Mon Sep 17 00:00:00 2001
From: David Marchand <david.marchand at redhat.com>
Date: Sat, 2 Nov 2024 10:38:16 +0100
Subject: [PATCH] eal/unix: optimize thread creation

[ upstream commit 64f27886b8bf127cd365a8a3ed5c05852a5ae81d ]

Setting the cpu affinity of the child thread from the parent thread is
racy when using pthread_setaffinity_np, as the child thread may start
running and initialize before affinity is set.

On the other hand, setting the cpu affinity from the child thread itself
may fail, so the parent thread waits for the child thread to report
whether this call succeeded.

This synchronisation point resulted in a significant slow down of
rte_thread_create() (as seen in the lcores_autotest unit tests, in OBS
for some ARM systems).

Another option for setting cpu affinity is to use the not portable
pthread_attr_setaffinity_np available in FreeBSD and glibc,
but not available in musl.

Fixes: b28c6196b132 ("eal/unix: fix thread creation")

Signed-off-by: David Marchand <david.marchand at redhat.com>
Acked-by: Luca Boccassi <bluca at debian.org>
Acked-by: Stephen Hemminger <stephen at networkplumber.org>
Acked-by: Chengwen Feng <fengchengwen at huawei.com>
---
 lib/eal/unix/meson.build  |  5 +++++
 lib/eal/unix/rte_thread.c | 25 +++++++++++++++++++++++++
 2 files changed, 30 insertions(+)

diff --git a/lib/eal/unix/meson.build b/lib/eal/unix/meson.build
index cc7d67dd32..f1eb82e16a 100644
--- a/lib/eal/unix/meson.build
+++ b/lib/eal/unix/meson.build
@@ -11,3 +11,8 @@ sources += files(
         'eal_unix_timer.c',
         'rte_thread.c',
 )
+
+if is_freebsd or cc.has_function('pthread_attr_setaffinity_np', args: '-D_GNU_SOURCE',
+                                 prefix : '#include <pthread.h>')
+    cflags += '-DRTE_EAL_PTHREAD_ATTR_SETAFFINITY_NP'
+endif
diff --git a/lib/eal/unix/rte_thread.c b/lib/eal/unix/rte_thread.c
index f4076122a4..e546329120 100644
--- a/lib/eal/unix/rte_thread.c
+++ b/lib/eal/unix/rte_thread.c
@@ -17,6 +17,7 @@ struct eal_tls_key {
 	pthread_key_t thread_index;
 };
 
+#ifndef RTE_EAL_PTHREAD_ATTR_SETAFFINITY_NP
 struct thread_start_context {
 	rte_thread_func thread_func;
 	void *thread_args;
@@ -26,6 +27,7 @@ struct thread_start_context {
 	int wrapper_ret;
 	bool wrapper_done;
 };
+#endif
 
 static int
 thread_map_priority_to_os_value(enum rte_thread_priority eal_pri, int *os_pri,
@@ -86,6 +88,7 @@ thread_map_os_priority_to_eal_priority(int policy, int os_pri,
 	return 0;
 }
 
+#ifndef RTE_EAL_PTHREAD_ATTR_SETAFFINITY_NP
 static void *
 thread_start_wrapper(void *arg)
 {
@@ -111,6 +114,7 @@ thread_start_wrapper(void *arg)
 
 	return (void *)(uintptr_t)thread_func(thread_args);
 }
+#endif
 
 int
 rte_thread_create(rte_thread_t *thread_id,
@@ -124,6 +128,7 @@ rte_thread_create(rte_thread_t *thread_id,
 		.sched_priority = 0,
 	};
 	int policy = SCHED_OTHER;
+#ifndef RTE_EAL_PTHREAD_ATTR_SETAFFINITY_NP
 	struct thread_start_context ctx = {
 		.thread_func = thread_func,
 		.thread_args = args,
@@ -132,6 +137,7 @@ rte_thread_create(rte_thread_t *thread_id,
 		.wrapper_mutex = PTHREAD_MUTEX_INITIALIZER,
 		.wrapper_cond = PTHREAD_COND_INITIALIZER,
 	};
+#endif
 
 	if (thread_attr != NULL) {
 		ret = pthread_attr_init(&attr);
@@ -142,6 +148,16 @@ rte_thread_create(rte_thread_t *thread_id,
 
 		attrp = &attr;
 
+#ifdef RTE_EAL_PTHREAD_ATTR_SETAFFINITY_NP
+		if (CPU_COUNT(&thread_attr->cpuset) > 0) {
+			ret = pthread_attr_setaffinity_np(attrp, sizeof(thread_attr->cpuset),
+				&thread_attr->cpuset);
+			if (ret != 0) {
+				RTE_LOG(DEBUG, EAL, "pthread_attr_setaffinity_np failed");
+				goto cleanup;
+			}
+		}
+#endif
 		/*
 		 * Set the inherit scheduler parameter to explicit,
 		 * otherwise the priority attribute is ignored.
@@ -176,6 +192,14 @@ rte_thread_create(rte_thread_t *thread_id,
 		}
 	}
 
+#ifdef RTE_EAL_PTHREAD_ATTR_SETAFFINITY_NP
+	ret = pthread_create((pthread_t *)&thread_id->opaque_id, attrp,
+		(void *)(void *)thread_func, args);
+	if (ret != 0) {
+		RTE_LOG(DEBUG, EAL, "pthread_create failed\n");
+		goto cleanup;
+	}
+#else /* !RTE_EAL_PTHREAD_ATTR_SETAFFINITY_NP */
 	ret = pthread_create((pthread_t *)&thread_id->opaque_id, attrp,
 		thread_start_wrapper, &ctx);
 	if (ret != 0) {
@@ -191,6 +215,7 @@ rte_thread_create(rte_thread_t *thread_id,
 
 	if (ret != 0)
 		pthread_join((pthread_t)thread_id->opaque_id, NULL);
+#endif /* RTE_EAL_PTHREAD_ATTR_SETAFFINITY_NP */
 
 cleanup:
 	if (attrp != NULL)
-- 
2.45.2

---
  Diff of the applied patch vs upstream commit (please double-check if non-empty:
---
--- -	2024-11-12 22:06:58.987392626 +0000
+++ 0009-eal-unix-optimize-thread-creation.patch	2024-11-12 22:06:58.639306763 +0000
@@ -1 +1 @@
-From 64f27886b8bf127cd365a8a3ed5c05852a5ae81d Mon Sep 17 00:00:00 2001
+From 02097752bb88526197af8b0dd0a8da9e6321cf9a Mon Sep 17 00:00:00 2001
@@ -5,0 +6,2 @@
+[ upstream commit 64f27886b8bf127cd365a8a3ed5c05852a5ae81d ]
+
@@ -23 +24,0 @@
-Cc: stable at dpdk.org
@@ -48 +49 @@
-index 1b4c73f58e..ea629c2065 100644
+index f4076122a4..e546329120 100644
@@ -51 +52 @@
-@@ -19,6 +19,7 @@ struct eal_tls_key {
+@@ -17,6 +17,7 @@ struct eal_tls_key {
@@ -59 +60 @@
-@@ -28,6 +29,7 @@ struct thread_start_context {
+@@ -26,6 +27,7 @@ struct thread_start_context {
@@ -67 +68 @@
-@@ -88,6 +90,7 @@ thread_map_os_priority_to_eal_priority(int policy, int os_pri,
+@@ -86,6 +88,7 @@ thread_map_os_priority_to_eal_priority(int policy, int os_pri,
@@ -75 +76 @@
-@@ -113,6 +116,7 @@ thread_start_wrapper(void *arg)
+@@ -111,6 +114,7 @@ thread_start_wrapper(void *arg)
@@ -83 +84 @@
-@@ -126,6 +130,7 @@ rte_thread_create(rte_thread_t *thread_id,
+@@ -124,6 +128,7 @@ rte_thread_create(rte_thread_t *thread_id,
@@ -91 +92 @@
-@@ -134,6 +139,7 @@ rte_thread_create(rte_thread_t *thread_id,
+@@ -132,6 +137,7 @@ rte_thread_create(rte_thread_t *thread_id,
@@ -99 +100 @@
-@@ -144,6 +150,16 @@ rte_thread_create(rte_thread_t *thread_id,
+@@ -142,6 +148,16 @@ rte_thread_create(rte_thread_t *thread_id,
@@ -108 +109 @@
-+				EAL_LOG(DEBUG, "pthread_attr_setaffinity_np failed");
++				RTE_LOG(DEBUG, EAL, "pthread_attr_setaffinity_np failed");
@@ -116 +117 @@
-@@ -178,6 +194,14 @@ rte_thread_create(rte_thread_t *thread_id,
+@@ -176,6 +192,14 @@ rte_thread_create(rte_thread_t *thread_id,
@@ -124 +125 @@
-+		EAL_LOG(DEBUG, "pthread_create failed");
++		RTE_LOG(DEBUG, EAL, "pthread_create failed\n");
@@ -131 +132 @@
-@@ -193,6 +217,7 @@ rte_thread_create(rte_thread_t *thread_id,
+@@ -191,6 +215,7 @@ rte_thread_create(rte_thread_t *thread_id,
@@ -134 +135 @@
- 		rte_thread_join(*thread_id, NULL);
+ 		pthread_join((pthread_t)thread_id->opaque_id, NULL);