aboutsummaryrefslogtreecommitdiff
path: root/numa.c
blob: 53bd65a05df94a3418d378c19d609f19b7b06f8c (plain)
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
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
/*
 * 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 "qapi/error.h"
#include "qapi/opts-visitor.h"
#include "qapi/qapi-commands-misc.h"
#include "qapi/qapi-visit-misc.h"
#include "hw/boards.h"
#include "sysemu/hostmem.h"
#include "hw/mem/pc-dimm.h"
#include "hw/mem/memory-device.h"
#include "qemu/option.h"
#include "qemu/config-file.h"
#include "qemu/cutils.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;
bool have_numa_distance;
NodeInfo numa_info[MAX_NODES];


static void parse_numa_node(MachineState *ms, NumaNodeOptions *node,
                            Error **errp)
{
    uint16_t nodenr;
    uint16List *cpus = NULL;
    MachineClass *mc = MACHINE_GET_CLASS(ms);

    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;
    }

    if (!mc->cpu_index_to_instance_props || !mc->get_default_cpu_node_id) {
        error_report("NUMA is not supported by this machine-type");
        exit(1);
    }
    for (cpus = node->cpus; cpus; cpus = cpus->next) {
        CpuInstanceProperties props;
        if (cpus->value >= max_cpus) {
            error_setg(errp,
                       "CPU index (%" PRIu16 ")"
                       " should be smaller than maxcpus (%d)",
                       cpus->value, max_cpus);
            return;
        }
        props = mc->cpu_index_to_instance_props(ms, cpus->value);
        props.node_id = nodenr;
        props.has_node_id = true;
        machine_set_cpu_numa_node(ms, &props, &error_fatal);
    }

    if (node->has_mem && node->has_memdev) {
        error_setg(errp, "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, "memdev option must be specified for either "
                   "all or no nodes");
        return;
    }

    if (node->has_mem) {
        numa_info[nodenr].node_mem = node->mem;
    }
    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_uint(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);
    nb_numa_nodes++;
}

static void parse_numa_distance(NumaDistOptions *dist, Error **errp)
{
    uint16_t src = dist->src;
    uint16_t dst = dist->dst;
    uint8_t val = dist->val;

    if (src >= MAX_NODES || dst >= MAX_NODES) {
        error_setg(errp, "Parameter '%s' expects an integer between 0 and %d",
                   src >= MAX_NODES ? "src" : "dst", MAX_NODES - 1);
        return;
    }

    if (!numa_info[src].present || !numa_info[dst].present) {
        error_setg(errp, "Source/Destination NUMA node is missing. "
                   "Please use '-numa node' option to declare it first.");
        return;
    }

    if (val < NUMA_DISTANCE_MIN) {
        error_setg(errp, "NUMA distance (%" PRIu8 ") is invalid, "
                   "it shouldn't be less than %d.",
                   val, NUMA_DISTANCE_MIN);
        return;
    }

    if (src == dst && val != NUMA_DISTANCE_MIN) {
        error_setg(errp, "Local distance of node %d should be %d.",
                   src, NUMA_DISTANCE_MIN);
        return;
    }

    numa_info[src].distance[dst] = val;
    have_numa_distance = true;
}

static
void set_numa_options(MachineState *ms, NumaOptions *object, Error **errp)
{
    Error *err = NULL;

    switch (object->type) {
    case NUMA_OPTIONS_TYPE_NODE:
        parse_numa_node(ms, &object->u.node, &err);
        if (err) {
            goto end;
        }
        break;
    case NUMA_OPTIONS_TYPE_DIST:
        parse_numa_distance(&object->u.dist, &err);
        if (err) {
            goto end;
        }
        break;
    case NUMA_OPTIONS_TYPE_CPU:
        if (!object->u.cpu.has_node_id) {
            error_setg(&err, "Missing mandatory node-id property");
            goto end;
        }
        if (!numa_info[object->u.cpu.node_id].present) {
            error_setg(&err, "Invalid node-id=%" PRId64 ", NUMA node must be "
                "defined with -numa node,nodeid=ID before it's used with "
                "-numa cpu,node-id=ID", object->u.cpu.node_id);
            goto end;
        }

        machine_set_cpu_numa_node(ms, qapi_NumaCpuOptions_base(&object->u.cpu),
                                  &err);
        break;
    default:
        abort();
    }

end:
    error_propagate(errp, err);
}

int parse_numa(void *opaque, QemuOpts *opts, Error **errp)
{
    NumaOptions *object = NULL;
    MachineState *ms = MACHINE(opaque);
    Error *err = NULL;
    Visitor *v = opts_visitor_new(opts);

    visit_type_NumaOptions(v, NULL, &object, &err);
    visit_free(v);
    if (err) {
        goto end;
    }

    /* Fix up legacy suffix-less format */
    if ((object->type == NUMA_OPTIONS_TYPE_NODE) && object->u.node.has_mem) {
        const char *mem_str = qemu_opt_get(opts, "mem");
        qemu_strtosz_MiB(mem_str, NULL, &object->u.node.mem);
    }

    set_numa_options(ms, object, &err);

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

    return 0;
}

/* If all node pair distances are symmetric, then only distances
 * in one direction are enough. If there is even one asymmetric
 * pair, though, then all distances must be provided. The
 * distance from a node to itself is always NUMA_DISTANCE_MIN,
 * so providing it is never necessary.
 */
static void validate_numa_distance(void)
{
    int src, dst;
    bool is_asymmetrical = false;

    for (src = 0; src < nb_numa_nodes; src++) {
        for (dst = src; dst < nb_numa_nodes; dst++) {
            if (numa_info[src].distance[dst] == 0 &&
                numa_info[dst].distance[src] == 0) {
                if (src != dst) {
                    error_report("The distance between node %d and %d is "
                                 "missing, at least one distance value "
                                 "between each nodes should be provided.",
                                 src, dst);
                    exit(EXIT_FAILURE);
                }
            }

            if (numa_info[src].distance[dst] != 0 &&
                numa_info[dst].distance[src] != 0 &&
                numa_info[src].distance[dst] !=
                numa_info[dst].distance[src]) {
                is_asymmetrical = true;
            }
        }
    }

    if (is_asymmetrical) {
        for (src = 0; src < nb_numa_nodes; src++) {
            for (dst = 0; dst < nb_numa_nodes; dst++) {
                if (src != dst && numa_info[src].distance[dst] == 0) {
                    error_report("At least one asymmetrical pair of "
                            "distances is given, please provide distances "
                            "for both directions of all node pairs.");
                    exit(EXIT_FAILURE);
                }
            }
        }
    }
}

static void complete_init_numa_distance(void)
{
    int src, dst;

    /* Fixup NUMA distance by symmetric policy because if it is an
     * asymmetric distance table, it should be a complete table and
     * there would not be any missing distance except local node, which
     * is verified by validate_numa_distance above.
     */
    for (src = 0; src < nb_numa_nodes; src++) {
        for (dst = 0; dst < nb_numa_nodes; dst++) {
            if (numa_info[src].distance[dst] == 0) {
                if (src == dst) {
                    numa_info[src].distance[dst] = NUMA_DISTANCE_MIN;
                } else {
                    numa_info[src].distance[dst] = numa_info[dst].distance[src];
                }
            }
        }
    }
}

void numa_legacy_auto_assign_ram(MachineClass *mc, NodeInfo *nodes,
                                 int nb_nodes, ram_addr_t size)
{
    int i;
    uint64_t usedmem = 0;

    /* Align each node according to the alignment
     * requirements of the machine class
     */

    for (i = 0; i < nb_nodes - 1; i++) {
        nodes[i].node_mem = (size / nb_nodes) &
                            ~((1 << mc->numa_mem_align_shift) - 1);
        usedmem += nodes[i].node_mem;
    }
    nodes[i].node_mem = size - usedmem;
}

void numa_default_auto_assign_ram(MachineClass *mc, NodeInfo *nodes,
                                  int nb_nodes, ram_addr_t size)
{
    int i;
    uint64_t usedmem = 0, node_mem;
    uint64_t granularity = size / nb_nodes;
    uint64_t propagate = 0;

    for (i = 0; i < nb_nodes - 1; i++) {
        node_mem = (granularity + propagate) &
                   ~((1 << mc->numa_mem_align_shift) - 1);
        propagate = granularity + propagate - node_mem;
        nodes[i].node_mem = node_mem;
        usedmem += node_mem;
    }
    nodes[i].node_mem = size - usedmem;
}

void numa_complete_configuration(MachineState *ms)
{
    int i;
    MachineClass *mc = MACHINE_GET_CLASS(ms);

    /*
     * If memory hotplug is enabled (slots > 0) but without '-numa'
     * options explicitly on CLI, guestes will break.
     *
     *   Windows: won't enable memory hotplug without SRAT table at all
     *
     *   Linux: if QEMU is started with initial memory all below 4Gb
     *   and no SRAT table present, guest kernel will use nommu DMA ops,
     *   which breaks 32bit hw drivers when memory is hotplugged and
     *   guest tries to use it with that drivers.
     *
     * Enable NUMA implicitly by adding a new NUMA node automatically.
     */
    if (ms->ram_slots > 0 && nb_numa_nodes == 0 &&
        mc->auto_enable_numa_with_memhp) {
            NumaNodeOptions node = { };
            parse_numa_node(ms, &node, NULL);
    }

    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) {
            assert(mc->numa_auto_assign_ram);
            mc->numa_auto_assign_ram(mc, numa_info, nb_numa_nodes, ram_size);
        }

        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);
        }

        /* QEMU needs at least all unique node pair distances to build
         * the whole NUMA distance table. QEMU treats the distance table
         * as symmetric by default, i.e. distance A->B == distance B->A.
         * Thus, QEMU is able to complete the distance table
         * initialization even though only distance A->B is provided and
         * distance B->A is not. QEMU knows the distance of a node to
         * itself is always 10, so A->A distances may be omitted. When
         * the distances of two nodes of a pair differ, i.e. distance
         * A->B != distance B->A, then that means the distance table is
         * asymmetric. In this case, the distances for both directions
         * of all node pairs are required.
         */
        if (have_numa_distance) {
            /* Validate enough NUMA distance information was provided. */
            validate_numa_distance();

            /* Validation succeeded, now fill in any missing distances. */
            complete_init_numa_distance();
        }
    }
}

void parse_numa_opts(MachineState *ms)
{
    if (qemu_opts_foreach(qemu_find_opts("numa"), parse_numa, ms, NULL)) {
        exit(1);
    }
}

void numa_cpu_pre_plug(const CPUArchId *slot, DeviceState *dev, Error **errp)
{
    int node_id = object_property_get_int(OBJECT(dev), "node-id", &error_abort);

    if (node_id == CPU_UNSET_NUMA_NODE_ID) {
        /* due to bug in libvirt, it doesn't pass node-id from props on
         * device_add as expected, so we have to fix it up here */
        if (slot->props.has_node_id) {
            object_property_set_int(OBJECT(dev), slot->props.node_id,
                                    "node-id", errp);
        }
    } else if (node_id != slot->props.node_id) {
        error_setg(errp, "node-id=%d must match numa node specified "
                   "with -numa option", node_id);
    }
}

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, 0, false,
                                         mem_path, &err);
        if (err) {
            error_report_err(err);
            if (mem_prealloc) {
                exit(1);
            }
            error_report("falling back to regular RAM allocation.");

            /* Legacy behavior: if allocation failed, fall back to
             * regular RAM allocation.
             */
            mem_path = NULL;
            memory_region_init_ram_nomigrate(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_nomigrate(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 < nb_numa_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(NumaNodeMem node_mem[])
{
    MemoryDeviceInfoList *info_list = qmp_memory_device_list();
    MemoryDeviceInfoList *info;
    PCDIMMDeviceInfo     *pcdimm_info;

    for (info = info_list; info; info = info->next) {
        MemoryDeviceInfo *value = info->value;

        if (value) {
            switch (value->type) {
            case MEMORY_DEVICE_INFO_KIND_DIMM:
                pcdimm_info = value->u.dimm.data;
                break;

            case MEMORY_DEVICE_INFO_KIND_NVDIMM:
                pcdimm_info = value->u.nvdimm.data;
                break;

            default:
                pcdimm_info = NULL;
                break;
            }

            if (pcdimm_info) {
                node_mem[pcdimm_info->node].node_mem += pcdimm_info->size;
                if (pcdimm_info->hotpluggable && pcdimm_info->hotplugged) {
                    node_mem[pcdimm_info->node].node_plugged_mem +=
                        pcdimm_info->size;
                }
            }
        }
    }
    qapi_free_MemoryDeviceInfoList(info_list);
}

void query_numa_node_mem(NumaNodeMem 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].node_mem += 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_get_canonical_path_component(obj);
        m->value->has_id = !!m->value->id;

        m->value->size = object_property_get_uint(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;
}

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);
    }
}