[dpdk-dev] [PATCH v3 2/2] eal: restrict ctrl threads to startup cpu affinity

[David Marchand <david.marchand@redhat.com>]

Spawning the ctrl threads on anything that is not part of the eal
coremask is not that polite to the rest of the system, especially
when you took good care to pin your processes on cpu resources with
tools like taskset (linux) / cpuset (freebsd).

Rather than introduce yet another eal options to control on which cpu
those ctrl threads are created, let's take the startup cpu affinity
as a reference and remove the eal coremask from it.
If no cpu is left, then we default to the master core.

The cpuset is computed once at init before the original cpu affinity
is lost.

Introduced a RTE_CPU_AND macro to abstract the differences between linux
and freebsd respective macros.

Examples in a 4 cores FreeBSD vm:

$ ./build/app/testpmd -l 2,3 --no-huge --no-pci -m 512 \
 -- -i --total-num-mbufs=2048

$ procstat -S 1057
  PID    TID COMM                TDNAME              CPU CSID CPU MASK
 1057 100131 testpmd             -                     2    1 2
 1057 100140 testpmd             eal-intr-thread       1    1 0-1
 1057 100141 testpmd             rte_mp_handle         1    1 0-1
 1057 100142 testpmd             lcore-slave-3         3    1 3

$ cpuset -l 1,2,3 ./build/app/testpmd -l 2,3 --no-huge --no-pci -m 512 \
 -- -i --total-num-mbufs=2048

$ procstat -S 1061
  PID    TID COMM                TDNAME              CPU CSID CPU MASK
 1061 100131 testpmd             -                     2    2 2
 1061 100144 testpmd             eal-intr-thread       1    2 1
 1061 100145 testpmd             rte_mp_handle         1    2 1
 1061 100147 testpmd             lcore-slave-3         3    2 3

$ cpuset -l 2,3 ./build/app/testpmd -l 2,3 --no-huge --no-pci -m 512 \
 -- -i --total-num-mbufs=2048

$ procstat -S 1065
  PID    TID COMM                TDNAME              CPU CSID CPU MASK
 1065 100131 testpmd             -                     2    2 2
 1065 100148 testpmd             eal-intr-thread       2    2 2
 1065 100149 testpmd             rte_mp_handle         2    2 2
 1065 100150 testpmd             lcore-slave-3         3    2 3

Fixes: d651ee4919cd ("eal: set affinity for control threads")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
---
Changes since v2:
- added missing Cc: stable@dpdk.org
- re-indent doc to comply with 80 columns
- fix incorrect check on pthread_getaffinity_np

Changes since v1:
- added some description in the prog guide
- fixed FreeBSD build

---
 doc/guides/prog_guide/env_abstraction_layer.rst | 22 +++++++++++++++++++
 lib/librte_eal/common/eal_common_options.c      | 28 +++++++++++++++++++++++++
 lib/librte_eal/common/eal_common_thread.c       | 21 ++++---------------
 lib/librte_eal/common/eal_internal_cfg.h        |  3 +++
 lib/librte_eal/common/include/rte_lcore.h       | 17 +++++++++++----
 5 files changed, 70 insertions(+), 21 deletions(-)

diff --git a/doc/guides/prog_guide/env_abstraction_layer.rst b/doc/guides/prog_guide/env_abstraction_layer.rst
index 929d76d..1fd6ece 100644
--- a/doc/guides/prog_guide/env_abstraction_layer.rst
+++ b/doc/guides/prog_guide/env_abstraction_layer.rst
@@ -498,6 +498,28 @@ Those TLS include *_cpuset* and *_socket_id*:
 *	*_socket_id* stores the NUMA node of the CPU set. If the CPUs in CPU set belong to different NUMA node, the *_socket_id* will be set to SOCKET_ID_ANY.
 
 
+Control Thread API
+~~~~~~~~~~~~~~~~~~
+
+It is possible to create Control Threads using the public API
+``rte_ctrl_thread_create()``.
+Those threads can be used for management/infrastructure tasks and are used
+internally by DPDK for multi process support and interrupt handling.
+
+Those threads will be scheduled on cpus part of the original process cpu
+affinity from which the dataplane and service lcores are excluded.
+
+For example, on a 8 cpus system, starting a dpdk application with -l 2,3
+(dataplane cores), then depending on the affinity configuration which can be
+controlled with tools like taskset (Linux) or cpuset (FreeBSD),
+
+- with no affinity configuration, the Control Threads will end up on
+  0-1,4-7 cpus.
+- with affinity restricted to 2-4, the Control Threads will end up on
+  cpu 4.
+- with affinity restricted to 2-3, the Control Threads will end up on
+  cpu 2 (master lcore, which is the default when no cpu is available).
+
 .. _known_issue_label:
 
 Known Issues
diff --git a/lib/librte_eal/common/eal_common_options.c b/lib/librte_eal/common/eal_common_options.c
index 1f45f82..f8b7ab9 100644
--- a/lib/librte_eal/common/eal_common_options.c
+++ b/lib/librte_eal/common/eal_common_options.c
@@ -217,6 +217,7 @@ struct device_option {
 	internal_cfg->create_uio_dev = 0;
 	internal_cfg->iova_mode = RTE_IOVA_DC;
 	internal_cfg->user_mbuf_pool_ops_name = NULL;
+	CPU_ZERO(&internal_cfg->ctrl_cpuset);
 	internal_cfg->init_complete = 0;
 }
 
@@ -1359,6 +1360,31 @@ static int xdigit2val(unsigned char c)
 	cfg->lcore_count -= removed;
 }
 
+static void
+compute_ctrl_threads_cpuset(struct internal_config *internal_cfg)
+{
+	rte_cpuset_t *cpuset = &internal_cfg->ctrl_cpuset;
+	rte_cpuset_t default_set;
+	unsigned int lcore_id;
+
+	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
+		if (eal_cpu_detected(lcore_id) &&
+				rte_lcore_has_role(lcore_id, ROLE_OFF)) {
+			CPU_SET(lcore_id, cpuset);
+		}
+	}
+
+	if (pthread_getaffinity_np(pthread_self(), sizeof(rte_cpuset_t),
+				&default_set))
+		CPU_ZERO(&default_set);
+
+	RTE_CPU_AND(cpuset, cpuset, &default_set);
+
+	/* if no detected cpu is off, use master core */
+	if (!CPU_COUNT(cpuset))
+		CPU_SET(rte_get_master_lcore(), cpuset);
+}
+
 int
 eal_cleanup_config(struct internal_config *internal_cfg)
 {
@@ -1392,6 +1418,8 @@ static int xdigit2val(unsigned char c)
 		lcore_config[cfg->master_lcore].core_role = ROLE_RTE;
 	}
 
+	compute_ctrl_threads_cpuset(internal_cfg);
+
 	/* if no memory amounts were requested, this will result in 0 and
 	 * will be overridden later, right after eal_hugepage_info_init() */
 	for (i = 0; i < RTE_MAX_NUMA_NODES; i++)
diff --git a/lib/librte_eal/common/eal_common_thread.c b/lib/librte_eal/common/eal_common_thread.c
index a3985ce..14f206c 100644
--- a/lib/librte_eal/common/eal_common_thread.c
+++ b/lib/librte_eal/common/eal_common_thread.c
@@ -16,6 +16,7 @@
 #include <rte_memory.h>
 #include <rte_log.h>
 
+#include "eal_internal_cfg.h"
 #include "eal_private.h"
 #include "eal_thread.h"
 
@@ -168,10 +169,9 @@ static void *rte_thread_init(void *arg)
 		const pthread_attr_t *attr,
 		void *(*start_routine)(void *), void *arg)
 {
+	rte_cpuset_t *cpuset = &internal_config.ctrl_cpuset;
 	struct rte_thread_ctrl_params *params;
-	unsigned int lcore_id;
-	rte_cpuset_t cpuset;
-	int cpu_found, ret;
+	int ret;
 
 	params = malloc(sizeof(*params));
 	if (!params)
@@ -195,20 +195,7 @@ static void *rte_thread_init(void *arg)
 				"Cannot set name for ctrl thread\n");
 	}
 
-	cpu_found = 0;
-	CPU_ZERO(&cpuset);
-	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
-		if (eal_cpu_detected(lcore_id) &&
-				rte_lcore_has_role(lcore_id, ROLE_OFF)) {
-			CPU_SET(lcore_id, &cpuset);
-			cpu_found = 1;
-		}
-	}
-	/* if no detected cpu is off, use master core */
-	if (!cpu_found)
-		CPU_SET(rte_get_master_lcore(), &cpuset);
-
-	ret = pthread_setaffinity_np(*thread, sizeof(cpuset), &cpuset);
+	ret = pthread_setaffinity_np(*thread, sizeof(*cpuset), cpuset);
 	if (ret)
 		goto fail;
 
diff --git a/lib/librte_eal/common/eal_internal_cfg.h b/lib/librte_eal/common/eal_internal_cfg.h
index 60eaead..edff09d 100644
--- a/lib/librte_eal/common/eal_internal_cfg.h
+++ b/lib/librte_eal/common/eal_internal_cfg.h
@@ -13,6 +13,8 @@
 #include <rte_eal.h>
 #include <rte_pci_dev_feature_defs.h>
 
+#include "eal_thread.h"
+
 #define MAX_HUGEPAGE_SIZES 3  /**< support up to 3 page sizes */
 
 /*
@@ -73,6 +75,7 @@ struct internal_config {
 	unsigned num_hugepage_sizes;      /**< how many sizes on this system */
 	struct hugepage_info hugepage_info[MAX_HUGEPAGE_SIZES];
 	enum rte_iova_mode iova_mode ;    /**< Set IOVA mode on this system  */
+	rte_cpuset_t ctrl_cpuset;         /**< cpuset for ctrl threads */
 	volatile unsigned int init_complete;
 	/**< indicates whether EAL has completed initialization */
 };
diff --git a/lib/librte_eal/common/include/rte_lcore.h b/lib/librte_eal/common/include/rte_lcore.h
index 6e09d91..eb49132 100644
--- a/lib/librte_eal/common/include/rte_lcore.h
+++ b/lib/librte_eal/common/include/rte_lcore.h
@@ -23,10 +23,18 @@
 #define LCORE_ID_ANY     UINT32_MAX       /**< Any lcore. */
 
 #if defined(__linux__)
-	typedef	cpu_set_t rte_cpuset_t;
+typedef	cpu_set_t rte_cpuset_t;
+#define RTE_CPU_AND(dst, src1, src2) CPU_AND(dst, src1, src2)
 #elif defined(__FreeBSD__)
 #include <pthread_np.h>
-	typedef cpuset_t rte_cpuset_t;
+typedef cpuset_t rte_cpuset_t;
+#define RTE_CPU_AND(dst, src1, src2) do \
+{ \
+	cpuset_t tmp; \
+	CPU_COPY(src1, &tmp); \
+	CPU_AND(&tmp, src2); \
+	CPU_COPY(&tmp, dst); \
+} while (0)
 #endif
 
 /**
@@ -280,8 +288,9 @@ struct lcore_config {
  * Create a control thread.
  *
  * Wrapper to pthread_create(), pthread_setname_np() and
- * pthread_setaffinity_np(). The dataplane and service lcores are
- * excluded from the affinity of the new thread.
+ * pthread_setaffinity_np(). The affinity of the new thread is based
+ * on the cpu affinity retrieved at the time rte_eal_init() was called,
+ * the dataplane and service lcores are then excluded.
  *
  * @param thread
  *   Filled with the thread id of the new created thread.
-- 
1.8.3.1