[dpdk-dev] [PATCH v7] eal: provide API for querying valid socket id's

[Anatoly Burakov <anatoly.burakov@intel.com>]

During lcore scan, find all socket ID's and store them, and
provide public API to query valid socket id's. This will break
the ABI, so bump ABI version.

Also, remove deprecation notice corresponding to this change.

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Acked-by: Gowrishankar Muthukrishnan <gowrishankar.m@linux.vnet.ibm.com>
---

Notes:
    v7:
    - Renamed rte_num_socket_ids() to rte_socket_count()
    - Removed deprecation notice associated with this change
    - Addressed review comments
    
    v6:
    - Fixed meson ABI version header
    
    v5:
    - Move API to experimental
    - Store list of valid socket id's instead of simply
      recording the biggest one
    
    v4:
    - Remove backwards ABI compatibility, bump ABI instead
    
    v3:
    - Added ABI compatibility
    
    v2:
    - checkpatch changes
    - check socket before deciding if the core is not to be used

 doc/guides/rel_notes/deprecation.rst      |  3 --
 lib/librte_eal/bsdapp/eal/Makefile        |  2 +-
 lib/librte_eal/common/eal_common_lcore.c  | 75 ++++++++++++++++++++++++++-----
 lib/librte_eal/common/include/rte_eal.h   |  2 +
 lib/librte_eal/common/include/rte_lcore.h | 30 +++++++++++++
 lib/librte_eal/linuxapp/eal/Makefile      |  2 +-
 lib/librte_eal/meson.build                |  2 +-
 lib/librte_eal/rte_eal_version.map        |  2 +
 8 files changed, 100 insertions(+), 18 deletions(-)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 74c18ed..80472f5 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -38,9 +38,6 @@ Deprecation Notices
   success and failure, respectively.  This will change to 1 and 0 for true and
   false, respectively, to make use of the function more intuitive.
 
-* eal: new ``numa_node_count`` member will be added to ``rte_config`` structure
-  in v18.05.
-
 * eal: due to internal data layout reorganization, there will be changes to
   several structures and functions as a result of coming changes to support
   memory hotplug in v18.05.
diff --git a/lib/librte_eal/bsdapp/eal/Makefile b/lib/librte_eal/bsdapp/eal/Makefile
index dd455e6..ed1d17b 100644
--- a/lib/librte_eal/bsdapp/eal/Makefile
+++ b/lib/librte_eal/bsdapp/eal/Makefile
@@ -21,7 +21,7 @@ LDLIBS += -lgcc_s
 
 EXPORT_MAP := ../../rte_eal_version.map
 
-LIBABIVER := 6
+LIBABIVER := 7
 
 # specific to bsdapp exec-env
 SRCS-$(CONFIG_RTE_EXEC_ENV_BSDAPP) := eal.c
diff --git a/lib/librte_eal/common/eal_common_lcore.c b/lib/librte_eal/common/eal_common_lcore.c
index 7724fa4..3167e9d 100644
--- a/lib/librte_eal/common/eal_common_lcore.c
+++ b/lib/librte_eal/common/eal_common_lcore.c
@@ -7,6 +7,7 @@
 #include <string.h>
 #include <dirent.h>
 
+#include <rte_errno.h>
 #include <rte_log.h>
 #include <rte_eal.h>
 #include <rte_lcore.h>
@@ -16,6 +17,19 @@
 #include "eal_private.h"
 #include "eal_thread.h"
 
+static int
+socket_id_cmp(const void *a, const void *b)
+{
+	const int *lcore_id_a = a;
+	const int *lcore_id_b = b;
+
+	if (*lcore_id_a < *lcore_id_b)
+		return -1;
+	if (*lcore_id_a > *lcore_id_b)
+		return 1;
+	return 0;
+}
+
 /*
  * Parse /sys/devices/system/cpu to get the number of physical and logical
  * processors on the machine. The function will fill the cpu_info
@@ -28,6 +42,8 @@ rte_eal_cpu_init(void)
 	struct rte_config *config = rte_eal_get_configuration();
 	unsigned lcore_id;
 	unsigned count = 0;
+	unsigned int socket_id, prev_socket_id;
+	int lcore_to_socket_id[RTE_MAX_LCORE];
 
 	/*
 	 * Parse the maximum set of logical cores, detect the subset of running
@@ -39,6 +55,19 @@ rte_eal_cpu_init(void)
 		/* init cpuset for per lcore config */
 		CPU_ZERO(&lcore_config[lcore_id].cpuset);
 
+		/* find socket first */
+		socket_id = eal_cpu_socket_id(lcore_id);
+		if (socket_id >= RTE_MAX_NUMA_NODES) {
+#ifdef RTE_EAL_ALLOW_INV_SOCKET_ID
+			socket_id = 0;
+#else
+			RTE_LOG(ERR, EAL, "Socket ID (%u) is greater than RTE_MAX_NUMA_NODES (%d)\n",
+					socket_id, RTE_MAX_NUMA_NODES);
+			return -1;
+#endif
+		}
+		lcore_to_socket_id[lcore_id] = socket_id;
+
 		/* in 1:1 mapping, record related cpu detected state */
 		lcore_config[lcore_id].detected = eal_cpu_detected(lcore_id);
 		if (lcore_config[lcore_id].detected == 0) {
@@ -54,18 +83,7 @@ rte_eal_cpu_init(void)
 		config->lcore_role[lcore_id] = ROLE_RTE;
 		lcore_config[lcore_id].core_role = ROLE_RTE;
 		lcore_config[lcore_id].core_id = eal_cpu_core_id(lcore_id);
-		lcore_config[lcore_id].socket_id = eal_cpu_socket_id(lcore_id);
-		if (lcore_config[lcore_id].socket_id >= RTE_MAX_NUMA_NODES) {
-#ifdef RTE_EAL_ALLOW_INV_SOCKET_ID
-			lcore_config[lcore_id].socket_id = 0;
-#else
-			RTE_LOG(ERR, EAL, "Socket ID (%u) is greater than "
-				"RTE_MAX_NUMA_NODES (%d)\n",
-				lcore_config[lcore_id].socket_id,
-				RTE_MAX_NUMA_NODES);
-			return -1;
-#endif
-		}
+		lcore_config[lcore_id].socket_id = socket_id;
 		RTE_LOG(DEBUG, EAL, "Detected lcore %u as "
 				"core %u on socket %u\n",
 				lcore_id, lcore_config[lcore_id].core_id,
@@ -79,5 +97,38 @@ rte_eal_cpu_init(void)
 		RTE_MAX_LCORE);
 	RTE_LOG(INFO, EAL, "Detected %u lcore(s)\n", config->lcore_count);
 
+	/* sort all socket id's in ascending order */
+	qsort(lcore_to_socket_id, RTE_DIM(lcore_to_socket_id),
+			sizeof(lcore_to_socket_id[0]), socket_id_cmp);
+
+	prev_socket_id = -1;
+	config->numa_node_count = 0;
+	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
+		socket_id = lcore_to_socket_id[lcore_id];
+		if (socket_id != prev_socket_id)
+			config->numa_nodes[config->numa_node_count++] =
+					socket_id;
+		prev_socket_id = socket_id;
+	}
+	RTE_LOG(INFO, EAL, "Detected %u NUMA nodes\n", config->numa_node_count);
+
 	return 0;
 }
+
+unsigned int __rte_experimental
+rte_socket_count(void)
+{
+	const struct rte_config *config = rte_eal_get_configuration();
+	return config->numa_node_count;
+}
+
+int __rte_experimental
+rte_socket_id_by_idx(unsigned int idx)
+{
+	const struct rte_config *config = rte_eal_get_configuration();
+	if (idx >= config->numa_node_count) {
+		rte_errno = EINVAL;
+		return -1;
+	}
+	return config->numa_nodes[idx];
+}
diff --git a/lib/librte_eal/common/include/rte_eal.h b/lib/librte_eal/common/include/rte_eal.h
index 93ca4cc..991cbe0 100644
--- a/lib/librte_eal/common/include/rte_eal.h
+++ b/lib/librte_eal/common/include/rte_eal.h
@@ -57,6 +57,8 @@ enum rte_proc_type_t {
 struct rte_config {
 	uint32_t master_lcore;       /**< Id of the master lcore */
 	uint32_t lcore_count;        /**< Number of available logical cores. */
+	uint32_t numa_node_count;    /**< Number of detected NUMA nodes. */
+	uint32_t numa_nodes[RTE_MAX_NUMA_NODES]; /**< List of detected numa nodes. */
 	uint32_t service_lcore_count;/**< Number of available service cores. */
 	enum rte_lcore_role_t lcore_role[RTE_MAX_LCORE]; /**< State of cores. */
 
diff --git a/lib/librte_eal/common/include/rte_lcore.h b/lib/librte_eal/common/include/rte_lcore.h
index 0472220..7312975 100644
--- a/lib/librte_eal/common/include/rte_lcore.h
+++ b/lib/librte_eal/common/include/rte_lcore.h
@@ -132,6 +132,36 @@ rte_lcore_index(int lcore_id)
 unsigned rte_socket_id(void);
 
 /**
+ * Return number of physical sockets detected on the system.
+ *
+ * Note that number of nodes may not be correspondent to their physical id's:
+ * for example, a system may report two socket id's, but the actual socket id's
+ * may be 0 and 8.
+ *
+ * @return
+ *   the number of physical sockets as recognized by EAL
+ */
+unsigned int __rte_experimental
+rte_socket_count(void);
+
+/**
+ * Return socket id with a particular index.
+ *
+ * This will return socket id at a particular position in list of all detected
+ * physical socket id's. For example, on a machine with sockets [0, 8], passing
+ * 1 as a parameter will return 8.
+ *
+ * @param idx
+ *   index of physical socket id to return
+ *
+ * @return
+ *   - physical socket id as recognized by EAL
+ *   - -1 on error, with errno set to EINVAL
+ */
+int __rte_experimental
+rte_socket_id_by_idx(unsigned int idx);
+
+/**
  * Get the ID of the physical socket of the specified lcore
  *
  * @param lcore_id
diff --git a/lib/librte_eal/linuxapp/eal/Makefile b/lib/librte_eal/linuxapp/eal/Makefile
index 7e5bbe8..b9c7727 100644
--- a/lib/librte_eal/linuxapp/eal/Makefile
+++ b/lib/librte_eal/linuxapp/eal/Makefile
@@ -10,7 +10,7 @@ ARCH_DIR ?= $(RTE_ARCH)
 EXPORT_MAP := ../../rte_eal_version.map
 VPATH += $(RTE_SDK)/lib/librte_eal/common/arch/$(ARCH_DIR)
 
-LIBABIVER := 6
+LIBABIVER := 7
 
 VPATH += $(RTE_SDK)/lib/librte_eal/common
 
diff --git a/lib/librte_eal/meson.build b/lib/librte_eal/meson.build
index 15d1c6a..4aa63e3 100644
--- a/lib/librte_eal/meson.build
+++ b/lib/librte_eal/meson.build
@@ -21,7 +21,7 @@ else
 	error('unsupported system type @0@'.format(hostmachine.system()))
 endif
 
-version = 6  # the version of the EAL API
+version = 7  # the version of the EAL API
 allow_experimental_apis = true
 deps += 'compat'
 cflags += '-D_GNU_SOURCE'
diff --git a/lib/librte_eal/rte_eal_version.map b/lib/librte_eal/rte_eal_version.map
index 1d88437..30ec1fc 100644
--- a/lib/librte_eal/rte_eal_version.map
+++ b/lib/librte_eal/rte_eal_version.map
@@ -257,5 +257,7 @@ EXPERIMENTAL {
 	rte_service_set_runstate_mapped_check;
 	rte_service_set_stats_enable;
 	rte_service_start_with_defaults;
+	rte_socket_count;
+	rte_socket_id_by_idx;
 
 } DPDK_18.02;
-- 
2.7.4