summaryrefslogtreecommitdiffstats
path: root/numa.c
blob: 9f56be960fb381c4d35a29c7a2e193cbf7a6404c (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
/*
 * NUMA parameter parsing routines
 *
 * Copyright (c) 2014 Fujitsu Ltd.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 * THE SOFTWARE.
 */

#include "qemu/osdep.h"
#include "sysemu/numa.h"
#include "exec/cpu-common.h"
#include "exec/ramlist.h"
#include "qemu/bitmap.h"
#include "qom/cpu.h"
#include "qemu/error-report.h"
#include "include/exec/cpu-common.h" /* for RAM_ADDR_FMT */
#include "qapi-visit.h"
#include "qapi/opts-visitor.h"
#include "hw/boards.h"
#include "sysemu/hostmem.h"
#include "qmp-commands.h"
#include "hw/mem/pc-dimm.h"
#include "qemu/option.h"
#include "qemu/config-file.h"

QemuOptsList qemu_numa_opts = {
    .name = "numa",
    .implied_opt_name = "type",
    .head = QTAILQ_HEAD_INITIALIZER(qemu_numa_opts.head),
    .desc = { { 0 } } /* validated with OptsVisitor */
};

static int have_memdevs = -1;
static int max_numa_nodeid; /* Highest specified NUMA node ID, plus one.
                             * For all nodes, nodeid < max_numa_nodeid
                             */
int nb_numa_nodes;
NodeInfo numa_info[MAX_NODES];

void numa_set_mem_node_id(ram_addr_t addr, uint64_t size, uint32_t node)
{
    struct numa_addr_range *range;

    /*
     * Memory-less nodes can come here with 0 size in which case,
     * there is nothing to do.
     */
    if (!size) {
        return;
    }

    range = g_malloc0(sizeof(*range));
    range->mem_start = addr;
    range->mem_end = addr + size - 1;
    QLIST_INSERT_HEAD(&numa_info[node].addr, range, entry);
}

void numa_unset_mem_node_id(ram_addr_t addr, uint64_t size, uint32_t node)
{
    struct numa_addr_range *range, *next;

    QLIST_FOREACH_SAFE(range, &numa_info[node].addr, entry, next) {
        if (addr == range->mem_start && (addr + size - 1) == range->mem_end) {
            QLIST_REMOVE(range, entry);
            g_free(range);
            return;
        }
    }
}

static void numa_set_mem_ranges(void)
{
    int i;
    ram_addr_t mem_start = 0;

    /*
     * Deduce start address of each node and use it to store
     * the address range info in numa_info address range list
     */
    for (i = 0; i < nb_numa_nodes; i++) {
        numa_set_mem_node_id(mem_start, numa_info[i].node_mem, i);
        mem_start += numa_info[i].node_mem;
    }
}

/*
 * Check if @addr falls under NUMA @node.
 */
static bool numa_addr_belongs_to_node(ram_addr_t addr, uint32_t node)
{
    struct numa_addr_range *range;

    QLIST_FOREACH(range, &numa_info[node].addr, entry) {
        if (addr >= range->mem_start && addr <= range->mem_end) {
            return true;
        }
    }
    return false;
}

/*
 * Given an address, return the index of the NUMA node to which the
 * address belongs to.
 */
uint32_t numa_get_node(ram_addr_t addr, Error **errp)
{
    uint32_t i;

    /* For non NUMA configurations, check if the addr falls under node 0 */
    if (!nb_numa_nodes) {
        if (numa_addr_belongs_to_node(addr, 0)) {
            return 0;
        }
    }

    for (i = 0; i < nb_numa_nodes; i++) {
        if (numa_addr_belongs_to_node(addr, i)) {
            return i;
        }
    }

    error_setg(errp, "Address 0x" RAM_ADDR_FMT " doesn't belong to any "
                "NUMA node", addr);
    return -1;
}

static void numa_node_parse(NumaNodeOptions *node, QemuOpts *opts, Error **errp)
{
    uint16_t nodenr;
    uint16List *cpus = NULL;

    if (node->has_nodeid) {
        nodenr = node->nodeid;
    } else {
        nodenr = nb_numa_nodes;
    }

    if (nodenr >= MAX_NODES) {
        error_setg(errp, "Max number of NUMA nodes reached: %"
                   PRIu16 "", nodenr);
        return;
    }

    if (numa_info[nodenr].present) {
        error_setg(errp, "Duplicate NUMA nodeid: %" PRIu16, nodenr);
        return;
    }

    for (cpus = node->cpus; cpus; cpus = cpus->next) {
        if (cpus->value >= max_cpus) {
            error_setg(errp,
                       "CPU index (%" PRIu16 ")"
                       " should be smaller than maxcpus (%d)",
                       cpus->value, max_cpus);
            return;
        }
        bitmap_set(numa_info[nodenr].node_cpu, cpus->value, 1);
    }

    if (node->has_mem && node->has_memdev) {
        error_setg(errp, "qemu: cannot specify both mem= and memdev=");
        return;
    }

    if (have_memdevs == -1) {
        have_memdevs = node->has_memdev;
    }
    if (node->has_memdev != have_memdevs) {
        error_setg(errp, "qemu: memdev option must be specified for either "
                   "all or no nodes");
        return;
    }

    if (node->has_mem) {
        uint64_t mem_size = node->mem;
        const char *mem_str = qemu_opt_get(opts, "mem");
        /* Fix up legacy suffix-less format */
        if (g_ascii_isdigit(mem_str[strlen(mem_str) - 1])) {
            mem_size <<= 20;
        }
        numa_info[nodenr].node_mem = mem_size;
    }
    if (node->has_memdev) {
        Object *o;
        o = object_resolve_path_type(node->memdev, TYPE_MEMORY_BACKEND, NULL);
        if (!o) {
            error_setg(errp, "memdev=%s is ambiguous", node->memdev);
            return;
        }

        object_ref(o);
        numa_info[nodenr].node_mem = object_property_get_int(o, "size", NULL);
        numa_info[nodenr].node_memdev = MEMORY_BACKEND(o);
    }
    numa_info[nodenr].present = true;
    max_numa_nodeid = MAX(max_numa_nodeid, nodenr + 1);
}

static int parse_numa(void *opaque, QemuOpts *opts, Error **errp)
{
    NumaOptions *object = NULL;
    Error *err = NULL;

    {
        Visitor *v = opts_visitor_new(opts);
        visit_type_NumaOptions(v, NULL, &object, &err);
        visit_free(v);
    }

    if (err) {
        goto end;
    }

    switch (object->type) {
    case NUMA_OPTIONS_KIND_NODE:
        numa_node_parse(object->u.node.data, opts, &err);
        if (err) {
            goto end;
        }
        nb_numa_nodes++;
        break;
    default:
        abort();
    }

end:
    qapi_free_NumaOptions(object);
    if (err) {
        error_report_err(err);
        return -1;
    }

    return 0;
}

static char *enumerate_cpus(unsigned long *cpus, int max_cpus)
{
    int cpu;
    bool first = true;
    GString *s = g_string_new(NULL);

    for (cpu = find_first_bit(cpus, max_cpus);
        cpu < max_cpus;
        cpu = find_next_bit(cpus, max_cpus, cpu + 1)) {
        g_string_append_printf(s, "%s%d", first ? "" : " ", cpu);
        first = false;
    }
    return g_string_free(s, FALSE);
}

static void validate_numa_cpus(void)
{
    int i;
    unsigned long *seen_cpus = bitmap_new(max_cpus);

    for (i = 0; i < nb_numa_nodes; i++) {
        if (bitmap_intersects(seen_cpus, numa_info[i].node_cpu, max_cpus)) {
            bitmap_and(seen_cpus, seen_cpus,
                       numa_info[i].node_cpu, max_cpus);
            error_report("CPU(s) present in multiple NUMA nodes: %s",
                         enumerate_cpus(seen_cpus, max_cpus));
            g_free(seen_cpus);
            exit(EXIT_FAILURE);
        }
        bitmap_or(seen_cpus, seen_cpus,
                  numa_info[i].node_cpu, max_cpus);
    }

    if (!bitmap_full(seen_cpus, max_cpus)) {
        char *msg;
        bitmap_complement(seen_cpus, seen_cpus, max_cpus);
        msg = enumerate_cpus(seen_cpus, max_cpus);
        error_report("warning: CPU(s) not present in any NUMA nodes: %s", msg);
        error_report("warning: All CPU(s) up to maxcpus should be described "
                     "in NUMA config");
        g_free(msg);
    }
    g_free(seen_cpus);
}

void parse_numa_opts(MachineClass *mc)
{
    int i;

    for (i = 0; i < MAX_NODES; i++) {
        numa_info[i].node_cpu = bitmap_new(max_cpus);
    }

    if (qemu_opts_foreach(qemu_find_opts("numa"), parse_numa, NULL, NULL)) {
        exit(1);
    }

    assert(max_numa_nodeid <= MAX_NODES);

    /* No support for sparse NUMA node IDs yet: */
    for (i = max_numa_nodeid - 1; i >= 0; i--) {
        /* Report large node IDs first, to make mistakes easier to spot */
        if (!numa_info[i].present) {
            error_report("numa: Node ID missing: %d", i);
            exit(1);
        }
    }

    /* This must be always true if all nodes are present: */
    assert(nb_numa_nodes == max_numa_nodeid);

    if (nb_numa_nodes > 0) {
        uint64_t numa_total;

        if (nb_numa_nodes > MAX_NODES) {
            nb_numa_nodes = MAX_NODES;
        }

        /* If no memory size is given for any node, assume the default case
         * and distribute the available memory equally across all nodes
         */
        for (i = 0; i < nb_numa_nodes; i++) {
            if (numa_info[i].node_mem != 0) {
                break;
            }
        }
        if (i == nb_numa_nodes) {
            uint64_t usedmem = 0;

            /* On Linux, each node's border has to be 8MB aligned,
             * the final node gets the rest.
             */
            for (i = 0; i < nb_numa_nodes - 1; i++) {
                numa_info[i].node_mem = (ram_size / nb_numa_nodes) &
                                        ~((1 << 23UL) - 1);
                usedmem += numa_info[i].node_mem;
            }
            numa_info[i].node_mem = ram_size - usedmem;
        }

        numa_total = 0;
        for (i = 0; i < nb_numa_nodes; i++) {
            numa_total += numa_info[i].node_mem;
        }
        if (numa_total != ram_size) {
            error_report("total memory for NUMA nodes (0x%" PRIx64 ")"
                         " should equal RAM size (0x" RAM_ADDR_FMT ")",
                         numa_total, ram_size);
            exit(1);
        }

        for (i = 0; i < nb_numa_nodes; i++) {
            QLIST_INIT(&numa_info[i].addr);
        }

        numa_set_mem_ranges();

        for (i = 0; i < nb_numa_nodes; i++) {
            if (!bitmap_empty(numa_info[i].node_cpu, max_cpus)) {
                break;
            }
        }
        /* Historically VCPUs were assigned in round-robin order to NUMA
         * nodes. However it causes issues with guest not handling it nice
         * in case where cores/threads from a multicore CPU appear on
         * different nodes. So allow boards to override default distribution
         * rule grouping VCPUs by socket so that VCPUs from the same socket
         * would be on the same node.
         */
        if (i == nb_numa_nodes) {
            for (i = 0; i < max_cpus; i++) {
                unsigned node_id = i % nb_numa_nodes;
                if (mc->cpu_index_to_socket_id) {
                    node_id = mc->cpu_index_to_socket_id(i) % nb_numa_nodes;
                }

                set_bit(i, numa_info[node_id].node_cpu);
            }
        }

        validate_numa_cpus();
    } else {
        numa_set_mem_node_id(0, ram_size, 0);
    }
}

void numa_post_machine_init(void)
{
    CPUState *cpu;
    int i;

    CPU_FOREACH(cpu) {
        for (i = 0; i < nb_numa_nodes; i++) {
            assert(cpu->cpu_index < max_cpus);
            if (test_bit(cpu->cpu_index, numa_info[i].node_cpu)) {
                cpu->numa_node = i;
            }
        }
    }
}

static void allocate_system_memory_nonnuma(MemoryRegion *mr, Object *owner,
                                           const char *name,
                                           uint64_t ram_size)
{
    if (mem_path) {
#ifdef __linux__
        Error *err = NULL;
        memory_region_init_ram_from_file(mr, owner, name, ram_size, false,
                                         mem_path, &err);
        if (err) {
            error_report_err(err);
            if (mem_prealloc) {
                exit(1);
            }

            /* Legacy behavior: if allocation failed, fall back to
             * regular RAM allocation.
             */
            memory_region_init_ram(mr, owner, name, ram_size, &error_fatal);
        }
#else
        fprintf(stderr, "-mem-path not supported on this host\n");
        exit(1);
#endif
    } else {
        memory_region_init_ram(mr, owner, name, ram_size, &error_fatal);
    }
    vmstate_register_ram_global(mr);
}

void memory_region_allocate_system_memory(MemoryRegion *mr, Object *owner,
                                          const char *name,
                                          uint64_t ram_size)
{
    uint64_t addr = 0;
    int i;

    if (nb_numa_nodes == 0 || !have_memdevs) {
        allocate_system_memory_nonnuma(mr, owner, name, ram_size);
        return;
    }

    memory_region_init(mr, owner, name, ram_size);
    for (i = 0; i < MAX_NODES; i++) {
        uint64_t size = numa_info[i].node_mem;
        HostMemoryBackend *backend = numa_info[i].node_memdev;
        if (!backend) {
            continue;
        }
        MemoryRegion *seg = host_memory_backend_get_memory(backend,
                                                           &error_fatal);

        if (memory_region_is_mapped(seg)) {
            char *path = object_get_canonical_path_component(OBJECT(backend));
            error_report("memory backend %s is used multiple times. Each "
                         "-numa option must use a different memdev value.",
                         path);
            exit(1);
        }

        host_memory_backend_set_mapped(backend, true);
        memory_region_add_subregion(mr, addr, seg);
        vmstate_register_ram_global(seg);
        addr += size;
    }
}

static void numa_stat_memory_devices(uint64_t node_mem[])
{
    MemoryDeviceInfoList *info_list = NULL;
    MemoryDeviceInfoList **prev = &info_list;
    MemoryDeviceInfoList *info;

    qmp_pc_dimm_device_list(qdev_get_machine(), &prev);
    for (info = info_list; info; info = info->next) {
        MemoryDeviceInfo *value = info->value;

        if (value) {
            switch (value->type) {
            case MEMORY_DEVICE_INFO_KIND_DIMM:
                node_mem[value->u.dimm.data->node] += value->u.dimm.data->size;
                break;
            default:
                break;
            }
        }
    }
    qapi_free_MemoryDeviceInfoList(info_list);
}

void query_numa_node_mem(uint64_t node_mem[])
{
    int i;

    if (nb_numa_nodes <= 0) {
        return;
    }

    numa_stat_memory_devices(node_mem);
    for (i = 0; i < nb_numa_nodes; i++) {
        node_mem[i] += numa_info[i].node_mem;
    }
}

static int query_memdev(Object *obj, void *opaque)
{
    MemdevList **list = opaque;
    MemdevList *m = NULL;

    if (object_dynamic_cast(obj, TYPE_MEMORY_BACKEND)) {
        m = g_malloc0(sizeof(*m));

        m->value = g_malloc0(sizeof(*m->value));

        m->value->id = object_property_get_str(obj, "id", NULL);
        m->value->has_id = !!m->value->id;

        m->value->size = object_property_get_int(obj, "size",
                                                 &error_abort);
        m->value->merge = object_property_get_bool(obj, "merge",
                                                   &error_abort);
        m->value->dump = object_property_get_bool(obj, "dump",
                                                  &error_abort);
        m->value->prealloc = object_property_get_bool(obj,
                                                      "prealloc",
                                                      &error_abort);
        m->value->policy = object_property_get_enum(obj,
                                                    "policy",
                                                    "HostMemPolicy",
                                                    &error_abort);
        object_property_get_uint16List(obj, "host-nodes",
                                       &m->value->host_nodes,
                                       &error_abort);

        m->next = *list;
        *list = m;
    }

    return 0;
}

MemdevList *qmp_query_memdev(Error **errp)
{
    Object *obj = object_get_objects_root();
    MemdevList *list = NULL;

    object_child_foreach(obj, query_memdev, &list);
    return list;
}

int numa_get_node_for_cpu(int idx)
{
    int i;

    assert(idx < max_cpus);

    for (i = 0; i < nb_numa_nodes; i++) {
        if (test_bit(idx, numa_info[i].node_cpu)) {
            break;
        }
    }
    return i;
}

void ram_block_notifier_add(RAMBlockNotifier *n)
{
    QLIST_INSERT_HEAD(&ram_list.ramblock_notifiers, n, next);
}

void ram_block_notifier_remove(RAMBlockNotifier *n)
{
    QLIST_REMOVE(n, next);
}

void ram_block_notify_add(void *host, size_t size)
{
    RAMBlockNotifier *notifier;

    QLIST_FOREACH(notifier, &ram_list.ramblock_notifiers, next) {
        notifier->ram_block_added(notifier, host, size);
    }
}

void ram_block_notify_remove(void *host, size_t size)
{
    RAMBlockNotifier *notifier;

    QLIST_FOREACH(notifier, &ram_list.ramblock_notifiers, next) {
        notifier->ram_block_removed(notifier, host, size);
    }
}