numa.c

   1 /*
   2  * NUMA parameter parsing routines
   3  *
   4  * Copyright (c) 2014 Fujitsu Ltd.
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a copy
   7  * of this software and associated documentation files (the "Software"), to deal
   8  * in the Software without restriction, including without limitation the rights
   9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  10  * copies of the Software, and to permit persons to whom the Software is
  11  * furnished to do so, subject to the following conditions:
  12  *
  13  * The above copyright notice and this permission notice shall be included in
  14  * all copies or substantial portions of the Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  22  * THE SOFTWARE.
  23  */
  24
  25 #include "sysemu/sysemu.h"
  26 #include "exec/cpu-common.h"
  27 #include "qemu/bitmap.h"
  28 #include "qom/cpu.h"
  29
  30 static void numa_node_parse_cpus(int nodenr, const char *cpus)
  31 {
  32     char *endptr;
  33     unsigned long long value, endvalue;
  34
  35     /* Empty CPU range strings will be considered valid, they will simply
  36      * not set any bit in the CPU bitmap.
  37      */
  38     if (!*cpus) {
  39         return;
  40     }
  41
  42     if (parse_uint(cpus, &value, &endptr, 10) < 0) {
  43         goto error;
  44     }
  45     if (*endptr == '-') {
  46         if (parse_uint_full(endptr + 1, &endvalue, 10) < 0) {
  47             goto error;
  48         }
  49     } else if (*endptr == '\0') {
  50         endvalue = value;
  51     } else {
  52         goto error;
  53     }
  54
  55     if (endvalue >= MAX_CPUMASK_BITS) {
  56         endvalue = MAX_CPUMASK_BITS - 1;
  57         fprintf(stderr,
  58             "qemu: NUMA: A max of %d VCPUs are supported\n",
  59              MAX_CPUMASK_BITS);
  60     }
  61
  62     if (endvalue < value) {
  63         goto error;
  64     }
  65
  66     bitmap_set(node_cpumask[nodenr], value, endvalue-value+1);
  67     return;
  68
  69 error:
  70     fprintf(stderr, "qemu: Invalid NUMA CPU range: %s\n", cpus);
  71     exit(1);
  72 }
  73
  74 void numa_add(const char *optarg)
  75 {
  76     char option[128];
  77     char *endptr;
  78     unsigned long long nodenr;
  79
  80     optarg = get_opt_name(option, 128, optarg, ',');
  81     if (*optarg == ',') {
  82         optarg++;
  83     }
  84     if (!strcmp(option, "node")) {
  85
  86         if (nb_numa_nodes >= MAX_NODES) {
  87             fprintf(stderr, "qemu: too many NUMA nodes\n");
  88             exit(1);
  89         }
  90
  91         if (get_param_value(option, 128, "nodeid", optarg) == 0) {
  92             nodenr = nb_numa_nodes;
  93         } else {
  94             if (parse_uint_full(option, &nodenr, 10) < 0) {
  95                 fprintf(stderr, "qemu: Invalid NUMA nodeid: %s\n", option);
  96                 exit(1);
  97             }
  98         }
  99
 100         if (nodenr >= MAX_NODES) {
 101             fprintf(stderr, "qemu: invalid NUMA nodeid: %llu\n", nodenr);
 102             exit(1);
 103         }
 104
 105         if (get_param_value(option, 128, "mem", optarg) == 0) {
 106             node_mem[nodenr] = 0;
 107         } else {
 108             int64_t sval;
 109             sval = strtosz(option, &endptr);
 110             if (sval < 0 || *endptr) {
 111                 fprintf(stderr, "qemu: invalid numa mem size: %s\n", optarg);
 112                 exit(1);
 113             }
 114             node_mem[nodenr] = sval;
 115         }
 116         if (get_param_value(option, 128, "cpus", optarg) != 0) {
 117             numa_node_parse_cpus(nodenr, option);
 118         }
 119         nb_numa_nodes++;
 120     } else {
 121         fprintf(stderr, "Invalid -numa option: %s\n", option);
 122         exit(1);
 123     }
 124 }
 125
 126 void set_numa_nodes(void)
 127 {
 128     if (nb_numa_nodes > 0) {
 129         int i;
 130
 131         if (nb_numa_nodes > MAX_NODES) {
 132             nb_numa_nodes = MAX_NODES;
 133         }
 134
 135         /* If no memory size if given for any node, assume the default case
 136          * and distribute the available memory equally across all nodes
 137          */
 138         for (i = 0; i < nb_numa_nodes; i++) {
 139             if (node_mem[i] != 0) {
 140                 break;
 141             }
 142         }
 143         if (i == nb_numa_nodes) {
 144             uint64_t usedmem = 0;
 145
 146             /* On Linux, the each node's border has to be 8MB aligned,
 147              * the final node gets the rest.
 148              */
 149             for (i = 0; i < nb_numa_nodes - 1; i++) {
 150                 node_mem[i] = (ram_size / nb_numa_nodes) & ~((1 << 23UL) - 1);
 151                 usedmem += node_mem[i];
 152             }
 153             node_mem[i] = ram_size - usedmem;
 154         }
 155
 156         for (i = 0; i < nb_numa_nodes; i++) {
 157             if (!bitmap_empty(node_cpumask[i], MAX_CPUMASK_BITS)) {
 158                 break;
 159             }
 160         }
 161         /* assigning the VCPUs round-robin is easier to implement, guest OSes
 162          * must cope with this anyway, because there are BIOSes out there in
 163          * real machines which also use this scheme.
 164          */
 165         if (i == nb_numa_nodes) {
 166             for (i = 0; i < max_cpus; i++) {
 167                 set_bit(i, node_cpumask[i % nb_numa_nodes]);
 168             }
 169         }
 170     }
 171 }
 172
 173 void set_numa_modes(void)
 174 {
 175     CPUState *cpu;
 176     int i;
 177
 178     CPU_FOREACH(cpu) {
 179         for (i = 0; i < nb_numa_nodes; i++) {
 180             if (test_bit(cpu->cpu_index, node_cpumask[i])) {
 181                 cpu->numa_node = i;
 182             }
 183         }
 184     }
 185 }