From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from dpdk.org (dpdk.org [92.243.14.124]) by inbox.dpdk.org (Postfix) with ESMTP id 5C0FEA051F; Wed, 10 Jun 2020 16:29:37 +0200 (CEST) Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id 9EDBA1BFA6; Wed, 10 Jun 2020 16:28:06 +0200 (CEST) Received: from mail-lj1-f194.google.com (mail-lj1-f194.google.com [209.85.208.194]) by dpdk.org (Postfix) with ESMTP id 926D34C89 for ; Wed, 10 Jun 2020 16:28:00 +0200 (CEST) Received: by mail-lj1-f194.google.com with SMTP id z9so2708216ljh.13 for ; Wed, 10 Jun 2020 07:28:00 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=7mHawFcZfFG6dLsXvfQmzhcLT0fAkkidrZdVwU+GiSw=; b=YcWhL4RO+nmCgUghDH0+2CJaCXP3glp6eTcvuX5esCZfaCwnPDUNFwODlHPgY0Qvfp AFrQGHJHnx+n0oe3mqWLPfnrZFteRZ+EQsqJdfqhKOSDnN27pSPleQ0FF9sIgazK7PUj R0FNfJP4wc/isrbBafHHDgihNREs42nmvZLknG8gmAhqRjoXU2ZQKF04f1tiO28M6ETu Aukhj/7R3saxX1R4da/QmZtRzRnY3cwp8TWI1Nkx3yrsPWSnZc+da/qX1XqYDWvjad3o fzd+rmx3QudHxfRLd/8gSnYpa/mIl1XivFxayUU41lcB2HFTxTEHN+H+JypE1vjI7i3d xz/A== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=7mHawFcZfFG6dLsXvfQmzhcLT0fAkkidrZdVwU+GiSw=; b=XElKuKncedQ2ixHk/azdqJAxvazm5SX/hE/4SCByMocbtEPcZ4N8uu+I9ukPm/ztlC VQUxGBf/P++w9pLhgI0fUsYNW9SQWtR5gASXDadASElStnaj20zSvB4jN/MdkgDXYxxP 4tQe9XhQygoBKuLuwZHigFxd7i+t9aUZwYQXGXITiPHeglRS0BEKAjgB7SZNsOt7Hksa raezlGvqBXYqWDNwO1z+FWiR/IjHU10tRaXK4htuRESBwGqU7K/MvA1RkBRTaTU/Y3/g yinydj5RFkDrZPZHirwHGAxwUcJiTj8E0/S+BGPqaLfqNG0IY/+rLdzKJ+7nXrhd25Nr Bphw== X-Gm-Message-State: AOAM530FDSW+g0GxGAM5Pe/OigxS/M5YLL/pOnSPQV1o4Y8OBTJLTllT MhQVrdlHP1pRNdc7fPpLMqC8YdHySpUBfg== X-Google-Smtp-Source: ABdhPJzbBMak2rwyFrRWeNylpuS+jliuXsZt4Y9U/NVwoXch/VNoX9QiKjX20lqitWcRqmwCszK3YA== X-Received: by 2002:a2e:9ac2:: with SMTP id p2mr2114169ljj.223.1591799279860; Wed, 10 Jun 2020 07:27:59 -0700 (PDT) Received: from localhost.localdomain (broadband-37-110-65-23.ip.moscow.rt.ru. [37.110.65.23]) by smtp.gmail.com with ESMTPSA id j12sm5051733ljg.15.2020.06.10.07.27.58 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Wed, 10 Jun 2020 07:27:59 -0700 (PDT) From: Dmitry Kozlyuk To: dev@dpdk.org Cc: Dmitry Malloy , Narcisa Ana Maria Vasile , Fady Bader , Tal Shnaiderman , Dmitry Kozlyuk , Harini Ramakrishnan , Omar Cardona , Pallavi Kadam , Ranjit Menon Date: Wed, 10 Jun 2020 17:27:28 +0300 Message-Id: <20200610142730.31376-10-dmitry.kozliuk@gmail.com> X-Mailer: git-send-email 2.25.4 In-Reply-To: <20200610142730.31376-1-dmitry.kozliuk@gmail.com> References: <20200608074153.29611-1-dmitry.kozliuk@gmail.com> <20200610142730.31376-1-dmitry.kozliuk@gmail.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Subject: [dpdk-dev] [PATCH v8 09/11] eal/windows: improve CPU and NUMA node detection X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" 1. Map CPU cores to their respective NUMA nodes as reported by system. 2. Support systems with more than 64 cores (multiple processor groups). 3. Fix magic constants, styling issues, and compiler warnings. 4. Add EAL private function to map DPDK socket ID to NUMA node number. Fixes: 53ffd9f080fc ("eal/windows: add minimum viable code") Signed-off-by: Dmitry Kozlyuk --- lib/librte_eal/windows/eal_lcore.c | 185 +++++++++++++++++---------- lib/librte_eal/windows/eal_windows.h | 10 ++ 2 files changed, 124 insertions(+), 71 deletions(-) diff --git a/lib/librte_eal/windows/eal_lcore.c b/lib/librte_eal/windows/eal_lcore.c index 82ee45413..9d931d50a 100644 --- a/lib/librte_eal/windows/eal_lcore.c +++ b/lib/librte_eal/windows/eal_lcore.c @@ -3,103 +3,146 @@ */ #include +#include #include #include +#include +#include +#include #include "eal_private.h" #include "eal_thread.h" #include "eal_windows.h" -/* global data structure that contains the CPU map */ -static struct _wcpu_map { - unsigned int total_procs; - unsigned int proc_sockets; - unsigned int proc_cores; - unsigned int reserved; - struct _win_lcore_map { - uint8_t socket_id; - uint8_t core_id; - } wlcore_map[RTE_MAX_LCORE]; -} wcpu_map = { 0 }; - -/* - * Create a map of all processors and associated cores on the system - */ +/** Number of logical processors (cores) in a processor group (32 or 64). */ +#define EAL_PROCESSOR_GROUP_SIZE (sizeof(KAFFINITY) * CHAR_BIT) + +struct lcore_map { + uint8_t socket_id; + uint8_t core_id; +}; + +struct socket_map { + uint16_t node_id; +}; + +struct cpu_map { + unsigned int socket_count; + unsigned int lcore_count; + struct lcore_map lcores[RTE_MAX_LCORE]; + struct socket_map sockets[RTE_MAX_NUMA_NODES]; +}; + +static struct cpu_map cpu_map = { 0 }; + void -eal_create_cpu_map() +eal_create_cpu_map(void) { - wcpu_map.total_procs = - GetActiveProcessorCount(ALL_PROCESSOR_GROUPS); - - LOGICAL_PROCESSOR_RELATIONSHIP lprocRel; - DWORD lprocInfoSize = 0; - BOOL ht_enabled = FALSE; - - /* First get the processor package information */ - lprocRel = RelationProcessorPackage; - /* Determine the size of buffer we need (pass NULL) */ - GetLogicalProcessorInformationEx(lprocRel, NULL, &lprocInfoSize); - wcpu_map.proc_sockets = lprocInfoSize / 48; - - lprocInfoSize = 0; - /* Next get the processor core information */ - lprocRel = RelationProcessorCore; - GetLogicalProcessorInformationEx(lprocRel, NULL, &lprocInfoSize); - wcpu_map.proc_cores = lprocInfoSize / 48; - - if (wcpu_map.total_procs > wcpu_map.proc_cores) - ht_enabled = TRUE; - - /* Distribute the socket and core ids appropriately - * across the logical cores. For now, split the cores - * equally across the sockets. - */ - unsigned int lcore = 0; - for (unsigned int socket = 0; socket < - wcpu_map.proc_sockets; ++socket) { - for (unsigned int core = 0; - core < (wcpu_map.proc_cores / wcpu_map.proc_sockets); - ++core) { - wcpu_map.wlcore_map[lcore] - .socket_id = socket; - wcpu_map.wlcore_map[lcore] - .core_id = core; - lcore++; - if (ht_enabled) { - wcpu_map.wlcore_map[lcore] - .socket_id = socket; - wcpu_map.wlcore_map[lcore] - .core_id = core; - lcore++; + SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX *infos, *info; + DWORD infos_size; + bool full = false; + + infos_size = 0; + if (!GetLogicalProcessorInformationEx( + RelationNumaNode, NULL, &infos_size)) { + DWORD error = GetLastError(); + if (error != ERROR_INSUFFICIENT_BUFFER) { + rte_panic("cannot get NUMA node info size, error %lu", + GetLastError()); + } + } + + infos = malloc(infos_size); + if (infos == NULL) { + rte_panic("cannot allocate memory for NUMA node information"); + return; + } + + if (!GetLogicalProcessorInformationEx( + RelationNumaNode, infos, &infos_size)) { + rte_panic("cannot get NUMA node information, error %lu", + GetLastError()); + } + + info = infos; + while ((uint8_t *)info - (uint8_t *)infos < infos_size) { + unsigned int node_id = info->NumaNode.NodeNumber; + GROUP_AFFINITY *cores = &info->NumaNode.GroupMask; + struct lcore_map *lcore; + unsigned int i, socket_id; + + /* NUMA node may be reported multiple times if it includes + * cores from different processor groups, e. g. 80 cores + * of a physical processor comprise one NUMA node, but two + * processor groups, because group size is limited by 32/64. + */ + for (socket_id = 0; socket_id < cpu_map.socket_count; + socket_id++) { + if (cpu_map.sockets[socket_id].node_id == node_id) + break; + } + + if (socket_id == cpu_map.socket_count) { + if (socket_id == RTE_DIM(cpu_map.sockets)) { + full = true; + goto exit; } + + cpu_map.sockets[socket_id].node_id = node_id; + cpu_map.socket_count++; + } + + for (i = 0; i < EAL_PROCESSOR_GROUP_SIZE; i++) { + if ((cores->Mask & ((KAFFINITY)1 << i)) == 0) + continue; + + if (cpu_map.lcore_count == RTE_DIM(cpu_map.lcores)) { + full = true; + goto exit; + } + + lcore = &cpu_map.lcores[cpu_map.lcore_count]; + lcore->socket_id = socket_id; + lcore->core_id = + cores->Group * EAL_PROCESSOR_GROUP_SIZE + i; + cpu_map.lcore_count++; } + + info = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX *)( + (uint8_t *)info + info->Size); + } + +exit: + if (full) { + /* RTE_LOG() may not be available, but this is important. */ + fprintf(stderr, "Enumerated maximum of %u NUMA nodes and %u cores\n", + cpu_map.socket_count, cpu_map.lcore_count); } + + free(infos); } -/* - * Check if a cpu is present by the presence of the cpu information for it - */ int eal_cpu_detected(unsigned int lcore_id) { - return (lcore_id < wcpu_map.total_procs); + return lcore_id < cpu_map.lcore_count; } -/* - * Get CPU socket id for a logical core - */ unsigned eal_cpu_socket_id(unsigned int lcore_id) { - return wcpu_map.wlcore_map[lcore_id].socket_id; + return cpu_map.lcores[lcore_id].socket_id; } -/* - * Get CPU socket id (NUMA node) for a logical core - */ unsigned eal_cpu_core_id(unsigned int lcore_id) { - return wcpu_map.wlcore_map[lcore_id].core_id; + return cpu_map.lcores[lcore_id].core_id; +} + +unsigned int +eal_socket_numa_node(unsigned int socket_id) +{ + return cpu_map.sockets[socket_id].node_id; } diff --git a/lib/librte_eal/windows/eal_windows.h b/lib/librte_eal/windows/eal_windows.h index fadd676b2..390d2fd66 100644 --- a/lib/librte_eal/windows/eal_windows.h +++ b/lib/librte_eal/windows/eal_windows.h @@ -26,4 +26,14 @@ void eal_create_cpu_map(void); */ int eal_thread_create(pthread_t *thread); +/** + * Get system NUMA node number for a socket ID. + * + * @param socket_id + * Valid EAL socket ID. + * @return + * NUMA node number to use with Win32 API. + */ +unsigned int eal_socket_numa_node(unsigned int socket_id); + #endif /* _EAL_WINDOWS_H_ */ -- 2.25.4