qemu_capabilities.c 113.2 KB
Newer Older
1 2 3
/*
 * qemu_capabilities.c: QEMU capabilities generation
 *
4
 * Copyright (C) 2006-2014 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16 17
 * Copyright (C) 2006 Daniel P. Berrange
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with this library.  If not, see
O
Osier Yang 已提交
19
 * <http://www.gnu.org/licenses/>.
20 21 22 23 24 25 26
 *
 * Author: Daniel P. Berrange <berrange@redhat.com>
 */

#include <config.h>

#include "qemu_capabilities.h"
27
#include "viralloc.h"
28
#include "vircrypto.h"
29
#include "virlog.h"
30
#include "virerror.h"
E
Eric Blake 已提交
31
#include "virfile.h"
32 33
#include "virpidfile.h"
#include "virprocess.h"
34 35 36
#include "nodeinfo.h"
#include "cpu/cpu.h"
#include "domain_conf.h"
37
#include "vircommand.h"
38
#include "virbitmap.h"
39
#include "virnodesuspend.h"
40
#include "virnuma.h"
41
#include "qemu_monitor.h"
42
#include "virstring.h"
43
#include "qemu_hostdev.h"
44

45
#include <fcntl.h>
46 47 48
#include <sys/stat.h>
#include <unistd.h>
#include <sys/wait.h>
49
#include <stdarg.h>
50 51 52

#define VIR_FROM_THIS VIR_FROM_QEMU

53 54
VIR_LOG_INIT("qemu.qemu_capabilities");

55 56 57 58
/* While not public, these strings must not change. They
 * are used in domain status files which are read on
 * daemon restarts
 */
59
VIR_ENUM_IMPL(virQEMUCaps, QEMU_CAPS_LAST,
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
              "kqemu",  /* 0 */
              "vnc-colon",
              "no-reboot",
              "drive",
              "drive-boot",

              "name", /* 5 */
              "uuid",
              "domid",
              "vnet-hdr",
              "migrate-kvm-stdio",

              "migrate-qemu-tcp", /* 10 */
              "migrate-qemu-exec",
              "drive-cache-v2",
              "kvm",
              "drive-format",

              "vga", /* 15 */
              "0.10",
              "pci-device",
              "mem-path",
              "drive-serial",

              "xen-domid", /* 20 */
              "migrate-qemu-unix",
              "chardev",
              "enable-kvm",
              "monitor-json",

              "balloon", /* 25 */
              "device",
              "sdl",
              "smp-topology",
              "netdev",

              "rtc", /* 30 */
97
              "vhost-net",
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
              "rtc-td-hack",
              "no-hpet",
              "no-kvm-pit",

              "tdf", /* 35 */
              "pci-configfd",
              "nodefconfig",
              "boot-menu",
              "enable-kqemu",

              "fsdev", /* 40 */
              "nesting",
              "name-process",
              "drive-readonly",
              "smbios-type",

              "vga-qxl", /* 45 */
              "spice",
              "vga-none",
              "migrate-qemu-fd",
              "boot-index",

              "hda-duplex", /* 50 */
              "drive-aio",
              "pci-multibus",
              "pci-bootindex",
              "ccid-emulated",

              "ccid-passthru", /* 55 */
              "chardev-spicevmc",
              "device-spicevmc",
              "virtio-tx-alg",
              "device-qxl-vga",
131 132

              "pci-multifunction", /* 60 */
133
              "virtio-blk-pci.ioeventfd",
M
Michal Privoznik 已提交
134
              "sga",
135 136
              "virtio-blk-pci.event_idx",
              "virtio-net-pci.event_idx",
137 138 139 140 141 142 143 144 145 146

              "cache-directsync", /* 65 */
              "piix3-usb-uhci",
              "piix4-usb-uhci",
              "usb-ehci",
              "ich9-usb-ehci1",

              "vt82c686b-usb-uhci", /* 70 */
              "pci-ohci",
              "usb-redir",
M
Marc-André Lureau 已提交
147
              "usb-hub",
148
              "no-shutdown",
149 150

              "cache-unsafe", /* 75 */
151
              "rombar",
J
Jim Fehlig 已提交
152
              "ich9-ahci",
153
              "no-acpi",
154
              "fsdev-readonly",
155

156
              "virtio-blk-pci.scsi", /* 80 */
157
              "blk-sg-io",
O
Osier Yang 已提交
158
              "drive-copy-on-read",
159
              "cpu-host",
160
              "fsdev-writeout",
161 162

              "drive-iotune", /* 85 */
163
              "system_wakeup",
164
              "scsi-disk.channel",
165
              "scsi-block",
166
              "transaction",
167 168 169

              "block-job-sync", /* 90 */
              "block-job-async",
170
              "scsi-cd",
171
              "ide-cd",
172
              "no-user-config",
M
Marc-André Lureau 已提交
173 174

              "hda-micro", /* 95 */
175
              "dump-guest-memory",
G
Gerd Hoffmann 已提交
176
              "nec-usb-xhci",
177
              "virtio-s390",
178
              "balloon-event",
M
Marc-André Lureau 已提交
179

R
Richa Marwaha 已提交
180
              "bridge", /* 100 */
181 182
              "lsi",
              "virtio-scsi-pci",
V
Viktor Mihajlovski 已提交
183
              "blockio",
184
              "disable-s3",
R
Richa Marwaha 已提交
185

186
              "disable-s4", /* 105 */
187
              "usb-redir.filter",
188 189
              "ide-drive.wwn",
              "scsi-disk.wwn",
190
              "seccomp-sandbox",
191 192

              "reboot-timeout", /* 110 */
193
              "dump-guest-core",
194
              "seamless-migration",
195
              "block-commit",
196
              "vnc",
197 198

              "drive-mirror", /* 115 */
199 200
              "usb-redir.bootindex",
              "usb-host.bootindex",
201
              "blockdev-snapshot-sync",
202 203 204 205 206 207
              "qxl",

              "VGA", /* 120 */
              "cirrus-vga",
              "vmware-svga",
              "device-video-primary",
208
              "s390-sclp",
209 210

              "usb-serial", /* 125 */
G
Guannan Ren 已提交
211
              "usb-net",
212
              "add-fd",
213
              "nbd-server",
214 215 216
              "virtio-rng",

              "rng-random", /* 130 */
217
              "rng-egd",
O
Olivia Yin 已提交
218 219
              "virtio-ccw",
              "dtb",
220
              "megasas",
J
Ján Tomko 已提交
221 222

              "ipv6-migration", /* 135 */
223
              "machine-opt",
L
Li Zhang 已提交
224
              "machine-usb-opt",
S
Stefan Berger 已提交
225 226
              "tpm-passthrough",
              "tpm-tis",
227 228

              "nvram",  /* 140 */
H
Han Cheng 已提交
229 230 231 232 233 234
              "pci-bridge",
              "vfio-pci",
              "vfio-pci.bootindex",
              "scsi-generic",

              "scsi-generic.bootindex", /* 145 */
235
              "mem-merge",
236
              "vnc-websocket",
O
Osier Yang 已提交
237
              "drive-discard",
238
              "mlock",
239 240

              "vnc-share-policy", /* 150 */
241
              "device-del-event",
242
              "dmi-to-pci-bridge",
243 244
              "i440fx-pci-hole64-size",
              "q35-pci-hole64-size",
245 246

              "usb-storage", /* 155 */
247
              "usb-storage.removable",
248
              "virtio-mmio",
249
              "ich9-intel-hda",
250
              "kvm-pit-lost-tick-policy",
251 252

              "boot-strict", /* 160 */
253 254
              "pvpanic",
              "enable-fips",
255 256
              "spice-file-xfer-disable",
              "spiceport",
L
Li Zhang 已提交
257 258

              "usb-kbd", /* 165 */
259
              "host-pci-multidomain",
260
              "msg-timestamp",
261
              "active-commit",
262
              "change-backing-file",
263 264

              "memory-backend-ram", /* 170 */
265
              "numa",
266
              "memory-backend-file",
267
              "usb-audio",
268
              "rtc-reset-reinjection",
269 270

              "splash-timeout", /* 175 */
J
John Ferlan 已提交
271
              "iothread",
272
              "migrate-rdma",
273
              "ivshmem",
274
              "drive-iotune-max",
275 276
    );

277 278 279 280 281 282 283 284

/*
 * Update the XML parser/formatter when adding more
 * information to this struct so that it gets cached
 * correctly. It does not have to be ABI-stable, as
 * the cache will be discarded & repopulated if the
 * timestamp on the libvirtd binary changes.
 */
285
struct _virQEMUCaps {
286 287
    virObject object;

288 289
    bool usedQMP;

290
    char *binary;
291
    time_t ctime;
292

293
    virBitmapPtr flags;
294 295 296 297

    unsigned int version;
    unsigned int kvmVersion;

298
    virArch arch;
299 300 301 302 303 304 305

    size_t ncpuDefinitions;
    char **cpuDefinitions;

    size_t nmachineTypes;
    char **machineTypes;
    char **machineAliases;
306
    unsigned int *machineMaxCpus;
307 308
};

309
struct _virQEMUCapsCache {
310 311
    virMutex lock;
    virHashTablePtr binaries;
312
    char *libDir;
313
    char *cacheDir;
314 315
    uid_t runUid;
    gid_t runGid;
316 317
};

318 319 320 321
struct virQEMUCapsSearchData {
    virArch arch;
};

322

323 324
static virClassPtr virQEMUCapsClass;
static void virQEMUCapsDispose(void *obj);
325

326
static int virQEMUCapsOnceInit(void)
327
{
328 329 330 331
    if (!(virQEMUCapsClass = virClassNew(virClassForObject(),
                                         "virQEMUCaps",
                                         sizeof(virQEMUCaps),
                                         virQEMUCapsDispose)))
332 333 334 335 336
        return -1;

    return 0;
}

337
VIR_ONCE_GLOBAL_INIT(virQEMUCaps)
338

339
static virArch virQEMUCapsArchFromString(const char *arch)
340 341 342 343 344 345 346 347 348 349
{
    if (STREQ(arch, "i386"))
        return VIR_ARCH_I686;
    if (STREQ(arch, "arm"))
        return VIR_ARCH_ARMV7L;

    return virArchFromString(arch);
}


350
static const char *virQEMUCapsArchToString(virArch arch)
351 352 353 354 355 356 357 358 359 360
{
    if (arch == VIR_ARCH_I686)
        return "i386";
    else if (arch == VIR_ARCH_ARMV7L)
        return "arm";

    return virArchToString(arch);
}


361
static virCommandPtr
362 363
virQEMUCapsProbeCommand(const char *qemu,
                        virQEMUCapsPtr qemuCaps,
364
                        uid_t runUid, gid_t runGid)
365 366 367
{
    virCommandPtr cmd = virCommandNew(qemu);

368 369
    if (qemuCaps) {
        if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_NO_USER_CONFIG))
370
            virCommandAddArg(cmd, "-no-user-config");
371
        else if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_NODEFCONFIG))
372 373 374 375 376
            virCommandAddArg(cmd, "-nodefconfig");
    }

    virCommandAddEnvPassCommon(cmd);
    virCommandClearCaps(cmd);
377 378
    virCommandSetGID(cmd, runGid);
    virCommandSetUID(cmd, runUid);
379 380 381 382 383

    return cmd;
}


384
static void
385 386
virQEMUCapsSetDefaultMachine(virQEMUCapsPtr qemuCaps,
                             size_t defIdx)
387
{
388 389
    char *name = qemuCaps->machineTypes[defIdx];
    char *alias = qemuCaps->machineAliases[defIdx];
390
    unsigned int maxCpus = qemuCaps->machineMaxCpus[defIdx];
391 392 393 394 395 396 397

    memmove(qemuCaps->machineTypes + 1,
            qemuCaps->machineTypes,
            sizeof(qemuCaps->machineTypes[0]) * defIdx);
    memmove(qemuCaps->machineAliases + 1,
            qemuCaps->machineAliases,
            sizeof(qemuCaps->machineAliases[0]) * defIdx);
398 399 400
    memmove(qemuCaps->machineMaxCpus + 1,
            qemuCaps->machineMaxCpus,
            sizeof(qemuCaps->machineMaxCpus[0]) * defIdx);
401 402
    qemuCaps->machineTypes[0] = name;
    qemuCaps->machineAliases[0] = alias;
403
    qemuCaps->machineMaxCpus[0] = maxCpus;
404 405
}

406 407 408 409
/* Format is:
 * <machine> <desc> [(default)|(alias of <canonical>)]
 */
static int
410 411
virQEMUCapsParseMachineTypesStr(const char *output,
                                virQEMUCapsPtr qemuCaps)
412 413 414
{
    const char *p = output;
    const char *next;
415
    size_t defIdx = 0;
416 417 418

    do {
        const char *t;
419 420
        char *name;
        char *canonical = NULL;
421 422 423 424 425 426 427 428 429 430

        if ((next = strchr(p, '\n')))
            ++next;

        if (STRPREFIX(p, "Supported machines are:"))
            continue;

        if (!(t = strchr(p, ' ')) || (next && t >= next))
            continue;

431 432
        if (VIR_STRNDUP(name, p, t - p) < 0)
            return -1;
433 434

        p = t;
435
        if ((t = strstr(p, "(default)")) && (!next || t < next))
436
            defIdx = qemuCaps->nmachineTypes;
437 438 439

        if ((t = strstr(p, "(alias of ")) && (!next || t < next)) {
            p = t + strlen("(alias of ");
440 441
            if (!(t = strchr(p, ')')) || (next && t >= next)) {
                VIR_FREE(name);
442
                continue;
443
            }
444

445
            if (VIR_STRNDUP(canonical, p, t - p) < 0) {
446
                VIR_FREE(name);
447
                return -1;
448 449 450
            }
        }

451
        if (VIR_REALLOC_N(qemuCaps->machineTypes, qemuCaps->nmachineTypes + 1) < 0 ||
452 453
            VIR_REALLOC_N(qemuCaps->machineAliases, qemuCaps->nmachineTypes + 1) < 0 ||
            VIR_REALLOC_N(qemuCaps->machineMaxCpus, qemuCaps->nmachineTypes + 1) < 0) {
454 455
            VIR_FREE(name);
            VIR_FREE(canonical);
456
            return -1;
457
        }
458
        qemuCaps->nmachineTypes++;
459
        if (canonical) {
460 461
            qemuCaps->machineTypes[qemuCaps->nmachineTypes-1] = canonical;
            qemuCaps->machineAliases[qemuCaps->nmachineTypes-1] = name;
462
        } else {
463 464
            qemuCaps->machineTypes[qemuCaps->nmachineTypes-1] = name;
            qemuCaps->machineAliases[qemuCaps->nmachineTypes-1] = NULL;
465
        }
466 467
        /* When parsing from command line we don't have information about maxCpus */
        qemuCaps->machineMaxCpus[qemuCaps->nmachineTypes-1] = 0;
468 469
    } while ((p = next));

470

471
    if (defIdx)
472
        virQEMUCapsSetDefaultMachine(qemuCaps, defIdx);
473 474 475 476

    return 0;
}

477
static int
478 479
virQEMUCapsProbeMachineTypes(virQEMUCapsPtr qemuCaps,
                             uid_t runUid, gid_t runGid)
480 481
{
    char *output;
482 483
    int ret = -1;
    virCommandPtr cmd;
484
    int status;
485

486 487 488 489
    /* Make sure the binary we are about to try exec'ing exists.
     * Technically we could catch the exec() failure, but that's
     * in a sub-process so it's hard to feed back a useful error.
     */
490
    if (!virFileIsExecutable(qemuCaps->binary)) {
491
        virReportSystemError(errno, _("Cannot find QEMU binary %s"),
492
                             qemuCaps->binary);
493 494 495
        return -1;
    }

496
    cmd = virQEMUCapsProbeCommand(qemuCaps->binary, qemuCaps, runUid, runGid);
497
    virCommandAddArgList(cmd, "-M", "?", NULL);
498
    virCommandSetOutputBuffer(cmd, &output);
499

500 501
    /* Ignore failure from older qemu that did not understand '-M ?'.  */
    if (virCommandRun(cmd, &status) < 0)
502 503
        goto cleanup;

504
    if (virQEMUCapsParseMachineTypesStr(output, qemuCaps) < 0)
505
        goto cleanup;
506 507 508

    ret = 0;

509
 cleanup:
510 511
    VIR_FREE(output);
    virCommandFree(cmd);
512 513 514 515 516 517

    return ret;
}


typedef int
518 519
(*virQEMUCapsParseCPUModels)(const char *output,
                             virQEMUCapsPtr qemuCaps);
520 521 522 523 524 525 526

/* Format:
 *      <arch> <model>
 * qemu-0.13 encloses some model names in []:
 *      <arch> [<model>]
 */
static int
527 528
virQEMUCapsParseX86Models(const char *output,
                          virQEMUCapsPtr qemuCaps)
529 530 531
{
    const char *p = output;
    const char *next;
532
    int ret = -1;
533 534 535

    do {
        const char *t;
536
        size_t len;
537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553

        if ((next = strchr(p, '\n')))
            next++;

        if (!(t = strchr(p, ' ')) || (next && t >= next))
            continue;

        if (!STRPREFIX(p, "x86"))
            continue;

        p = t;
        while (*p == ' ')
            p++;

        if (*p == '\0' || *p == '\n')
            continue;

554
        if (VIR_EXPAND_N(qemuCaps->cpuDefinitions, qemuCaps->ncpuDefinitions, 1) < 0)
555
            goto cleanup;
556

557 558 559 560
        if (next)
            len = next - p - 1;
        else
            len = strlen(p);
561

562 563 564 565
        if (len > 2 && *p == '[' && p[len - 1] == ']') {
            p++;
            len -= 2;
        }
566

567
        if (VIR_STRNDUP(qemuCaps->cpuDefinitions[qemuCaps->ncpuDefinitions - 1], p, len) < 0)
568
            goto cleanup;
569 570
    } while ((p = next));

571
    ret = 0;
572

573
 cleanup:
574
    return ret;
575 576
}

P
Prerna Saxena 已提交
577 578 579 580
/* ppc64 parser.
 * Format : PowerPC <machine> <description>
 */
static int
581 582
virQEMUCapsParsePPCModels(const char *output,
                          virQEMUCapsPtr qemuCaps)
P
Prerna Saxena 已提交
583 584 585
{
    const char *p = output;
    const char *next;
586
    int ret = -1;
P
Prerna Saxena 已提交
587 588 589

    do {
        const char *t;
590
        size_t len;
P
Prerna Saxena 已提交
591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610

        if ((next = strchr(p, '\n')))
            next++;

        if (!STRPREFIX(p, "PowerPC "))
            continue;

        /* Skip the preceding sub-string "PowerPC " */
        p += 8;

        /*Malformed string, does not obey the format 'PowerPC <model> <desc>'*/
        if (!(t = strchr(p, ' ')) || (next && t >= next))
            continue;

        if (*p == '\0')
            break;

        if (*p == '\n')
            continue;

611
        if (VIR_EXPAND_N(qemuCaps->cpuDefinitions, qemuCaps->ncpuDefinitions, 1) < 0)
612
            goto cleanup;
P
Prerna Saxena 已提交
613

614
        len = t - p - 1;
P
Prerna Saxena 已提交
615

616
        if (VIR_STRNDUP(qemuCaps->cpuDefinitions[qemuCaps->ncpuDefinitions - 1], p, len) < 0)
617
            goto cleanup;
P
Prerna Saxena 已提交
618 619
    } while ((p = next));

620
    ret = 0;
P
Prerna Saxena 已提交
621

622
 cleanup:
623
    return ret;
P
Prerna Saxena 已提交
624
}
625

626
static int
627
virQEMUCapsProbeCPUModels(virQEMUCapsPtr qemuCaps, uid_t runUid, gid_t runGid)
628 629 630
{
    char *output = NULL;
    int ret = -1;
631
    virQEMUCapsParseCPUModels parse;
632
    virCommandPtr cmd;
633

634
    if (qemuCaps->arch == VIR_ARCH_I686 ||
635
        qemuCaps->arch == VIR_ARCH_X86_64) {
636
        parse = virQEMUCapsParseX86Models;
637
    } else if ARCH_IS_PPC64(qemuCaps->arch) {
638
        parse = virQEMUCapsParsePPCModels;
639
    } else {
640
        VIR_DEBUG("don't know how to parse %s CPU models",
641
                  virArchToString(qemuCaps->arch));
642 643 644
        return 0;
    }

645
    cmd = virQEMUCapsProbeCommand(qemuCaps->binary, qemuCaps, runUid, runGid);
646
    virCommandAddArgList(cmd, "-cpu", "?", NULL);
647
    virCommandSetOutputBuffer(cmd, &output);
648

649
    if (virCommandRun(cmd, NULL) < 0)
650 651
        goto cleanup;

652
    if (parse(output, qemuCaps) < 0)
653 654 655 656
        goto cleanup;

    ret = 0;

657
 cleanup:
658
    VIR_FREE(output);
659
    virCommandFree(cmd);
660 661 662 663 664

    return ret;
}


665
static char *
666 667
virQEMUCapsFindBinaryForArch(virArch hostarch,
                             virArch guestarch)
668 669
{
    char *ret;
670
    const char *archstr = virQEMUCapsArchToString(guestarch);
671
    char *binary;
672

673
    if (virAsprintf(&binary, "qemu-system-%s", archstr) < 0)
674 675 676 677 678 679 680 681 682 683 684
        return NULL;

    ret = virFindFileInPath(binary);
    VIR_FREE(binary);
    if (ret && !virFileIsExecutable(ret))
        VIR_FREE(ret);

    if (guestarch == VIR_ARCH_I686 &&
        !ret &&
        hostarch == VIR_ARCH_X86_64) {
        ret = virFindFileInPath("qemu-system-x86_64");
685 686
        if (ret && !virFileIsExecutable(ret))
            VIR_FREE(ret);
687
    }
688

689 690 691 692 693
    if (guestarch == VIR_ARCH_I686 &&
        !ret) {
        ret = virFindFileInPath("qemu");
        if (ret && !virFileIsExecutable(ret))
            VIR_FREE(ret);
694
    }
695

696 697 698 699 700
    return ret;
}


static bool
701 702
virQEMUCapsIsValidForKVM(virArch hostarch,
                         virArch guestarch)
703
{
704
    if (hostarch == guestarch)
705
        return true;
706 707
    if (hostarch == VIR_ARCH_X86_64 &&
        guestarch == VIR_ARCH_I686)
708 709 710 711
        return true;
    return false;
}

712
static int
713 714 715 716
virQEMUCapsInitGuest(virCapsPtr caps,
                     virQEMUCapsCachePtr cache,
                     virArch hostarch,
                     virArch guestarch)
717
{
718
    size_t i;
719 720
    char *kvmbin = NULL;
    char *binary = NULL;
721 722
    virQEMUCapsPtr qemubinCaps = NULL;
    virQEMUCapsPtr kvmbinCaps = NULL;
723 724
    int ret = -1;

J
Ján Tomko 已提交
725
    /* Check for existence of base emulator, or alternate base
726 727
     * which can be used with magic cpu choice
     */
728
    binary = virQEMUCapsFindBinaryForArch(hostarch, guestarch);
729

730
    /* Ignore binary if extracting version info fails */
731
    if (binary) {
732
        if (!(qemubinCaps = virQEMUCapsCacheLookup(cache, binary))) {
733 734 735 736
            virResetLastError();
            VIR_FREE(binary);
        }
    }
737 738

    /* qemu-kvm/kvm binaries can only be used if
739 740 741 742 743
     *  - host & guest arches match
     * Or
     *  - hostarch is x86_64 and guest arch is i686
     * The latter simply needs "-cpu qemu32"
     */
744
    if (virQEMUCapsIsValidForKVM(hostarch, guestarch)) {
745 746 747
        const char *const kvmbins[] = { "/usr/libexec/qemu-kvm", /* RHEL */
                                        "qemu-kvm", /* Fedora */
                                        "kvm" }; /* Upstream .spec */
748

749 750
        for (i = 0; i < ARRAY_CARDINALITY(kvmbins); ++i) {
            kvmbin = virFindFileInPath(kvmbins[i]);
751

752 753
            if (!kvmbin)
                continue;
754

755
            if (!(kvmbinCaps = virQEMUCapsCacheLookup(cache, kvmbin))) {
756
                virResetLastError();
757 758 759
                VIR_FREE(kvmbin);
                continue;
            }
760

761 762
            if (!binary) {
                binary = kvmbin;
763
                qemubinCaps = kvmbinCaps;
764
                kvmbin = NULL;
765
                kvmbinCaps = NULL;
766
            }
767
            break;
768 769 770
        }
    }

771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799
    ret = virQEMUCapsInitGuestFromBinary(caps,
                                         binary, qemubinCaps,
                                         kvmbin, kvmbinCaps,
                                         guestarch);

    VIR_FREE(binary);
    VIR_FREE(kvmbin);
    virObjectUnref(qemubinCaps);
    virObjectUnref(kvmbinCaps);

    return ret;
}

int
virQEMUCapsInitGuestFromBinary(virCapsPtr caps,
                               const char *binary,
                               virQEMUCapsPtr qemubinCaps,
                               const char *kvmbin,
                               virQEMUCapsPtr kvmbinCaps,
                               virArch guestarch)
{
    virCapsGuestPtr guest;
    bool haskvm = false;
    bool haskqemu = false;
    virCapsGuestMachinePtr *machines = NULL;
    size_t nmachines = 0;
    int ret = -1;
    bool hasdisksnapshot = false;

800 801 802
    if (!binary)
        return 0;

803
    if (virFileExists("/dev/kvm") &&
804 805
        (virQEMUCapsGet(qemubinCaps, QEMU_CAPS_KVM) ||
         virQEMUCapsGet(qemubinCaps, QEMU_CAPS_ENABLE_KVM) ||
806
         kvmbin))
807
        haskvm = true;
808

809
    if (virFileExists("/dev/kqemu") &&
810
        virQEMUCapsGet(qemubinCaps, QEMU_CAPS_KQEMU))
811
        haskqemu = true;
J
Jiri Denemark 已提交
812

813
    if (virQEMUCapsGetMachineTypesCaps(qemubinCaps, &nmachines, &machines) < 0)
814
        goto cleanup;
815 816 817 818

    /* We register kvm as the base emulator too, since we can
     * just give -no-kvm to disable acceleration if required */
    if ((guest = virCapabilitiesAddGuest(caps,
D
Daniel P. Berrange 已提交
819
                                         "hvm",
820
                                         guestarch,
821 822 823 824
                                         binary,
                                         NULL,
                                         nmachines,
                                         machines)) == NULL)
825
        goto cleanup;
826 827 828 829 830

    machines = NULL;
    nmachines = 0;

    if (caps->host.cpu &&
J
Jiri Denemark 已提交
831
        caps->host.cpu->model &&
832
        virQEMUCapsGetCPUDefinitions(qemubinCaps, NULL) > 0 &&
833
        !virCapabilitiesAddGuestFeature(guest, "cpuselection", true, false))
834
        goto cleanup;
835

836
    if (virQEMUCapsGet(qemubinCaps, QEMU_CAPS_BOOTINDEX) &&
837
        !virCapabilitiesAddGuestFeature(guest, "deviceboot", true, false))
838
        goto cleanup;
839

840 841 842
    if (virQEMUCapsGet(qemubinCaps, QEMU_CAPS_DISK_SNAPSHOT))
        hasdisksnapshot = true;

843 844
    if (!virCapabilitiesAddGuestFeature(guest, "disksnapshot", hasdisksnapshot,
                                        false))
845
        goto cleanup;
846

D
Daniel P. Berrange 已提交
847 848 849 850 851 852
    if (virCapabilitiesAddGuestDomain(guest,
                                      "qemu",
                                      NULL,
                                      NULL,
                                      0,
                                      NULL) == NULL)
853
        goto cleanup;
854

D
Daniel P. Berrange 已提交
855 856 857 858 859 860 861
    if (haskqemu &&
        virCapabilitiesAddGuestDomain(guest,
                                      "kqemu",
                                      NULL,
                                      NULL,
                                      0,
                                      NULL) == NULL)
862
        goto cleanup;
863

D
Daniel P. Berrange 已提交
864 865
    if (haskvm) {
        virCapsGuestDomainPtr dom;
866

D
Daniel P. Berrange 已提交
867
        if (kvmbin &&
868
            virQEMUCapsGetMachineTypesCaps(kvmbinCaps, &nmachines, &machines) < 0)
869
            goto cleanup;
870

D
Daniel P. Berrange 已提交
871 872 873 874 875 876
        if ((dom = virCapabilitiesAddGuestDomain(guest,
                                                 "kvm",
                                                 kvmbin ? kvmbin : binary,
                                                 NULL,
                                                 nmachines,
                                                 machines)) == NULL) {
877
            goto cleanup;
D
Daniel P. Berrange 已提交
878
        }
879

D
Daniel P. Berrange 已提交
880 881
        machines = NULL;
        nmachines = 0;
882 883 884

    }

885 886
    if (((guestarch == VIR_ARCH_I686) ||
         (guestarch == VIR_ARCH_X86_64)) &&
887 888
        (virCapabilitiesAddGuestFeature(guest, "acpi", true, true) == NULL ||
         virCapabilitiesAddGuestFeature(guest, "apic", true, false) == NULL))
889
        goto cleanup;
890

891
    if ((guestarch == VIR_ARCH_I686) &&
892 893
        (virCapabilitiesAddGuestFeature(guest, "pae", true, false) == NULL ||
         virCapabilitiesAddGuestFeature(guest, "nonpae", true, false) == NULL))
894
        goto cleanup;
895 896 897

    ret = 0;

898
 cleanup:
899 900 901

    virCapabilitiesFreeMachines(machines, nmachines);

902
    return ret;
903 904 905 906
}


static int
907 908
virQEMUCapsInitCPU(virCapsPtr caps,
                   virArch arch)
909 910
{
    virCPUDefPtr cpu = NULL;
911
    virCPUDataPtr data = NULL;
912 913 914
    virNodeInfo nodeinfo;
    int ret = -1;

915
    if (VIR_ALLOC(cpu) < 0)
916 917
        goto error;

918 919
    cpu->arch = arch;

920
    if (nodeGetInfo(&nodeinfo))
921 922 923 924 925 926
        goto error;

    cpu->type = VIR_CPU_TYPE_HOST;
    cpu->sockets = nodeinfo.sockets;
    cpu->cores = nodeinfo.cores;
    cpu->threads = nodeinfo.threads;
J
Jiri Denemark 已提交
927
    caps->host.cpu = cpu;
928

929
    if (!(data = cpuNodeData(arch))
930
        || cpuDecode(cpu, data, NULL, 0, NULL) < 0)
J
Jiri Denemark 已提交
931
        goto cleanup;
932 933 934

    ret = 0;

935
 cleanup:
J
Jiri Denemark 已提交
936
    cpuDataFree(data);
937 938 939

    return ret;

940
 error:
941 942 943 944 945
    virCPUDefFree(cpu);
    goto cleanup;
}


M
Michal Privoznik 已提交
946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968
static int
virQEMUCapsInitPages(virCapsPtr caps)
{
    int ret = -1;
    unsigned int *pages_size = NULL;
    size_t npages;

    if (virNumaGetPages(-1 /* Magic constant for overall info */,
                        &pages_size, NULL, NULL, &npages) < 0)
        goto cleanup;

    caps->host.pagesSize = pages_size;
    pages_size = NULL;
    caps->host.nPagesSize = npages;
    npages = 0;

    ret = 0;
 cleanup:
    VIR_FREE(pages_size);
    return ret;
}


969
virCapsPtr virQEMUCapsInit(virQEMUCapsCachePtr cache)
970 971
{
    virCapsPtr caps;
972
    size_t i;
T
Tal Kain 已提交
973
    virArch hostarch = virArchFromHost();
974

T
Tal Kain 已提交
975
    if ((caps = virCapabilitiesNew(hostarch,
976
                                   true, true)) == NULL)
977
        goto error;
978 979 980 981 982 983 984

    /* Some machines have problematic NUMA toplogy causing
     * unexpected failures. We don't want to break the QEMU
     * driver in this scenario, so log errors & carry on
     */
    if (nodeCapsInitNUMA(caps) < 0) {
        virCapabilitiesFreeNUMAInfo(caps);
985
        VIR_WARN("Failed to query host NUMA topology, disabling NUMA capabilities");
986 987
    }

T
Tal Kain 已提交
988
    if (virQEMUCapsInitCPU(caps, hostarch) < 0)
989
        VIR_WARN("Failed to get host CPU");
990

991
    /* Add the power management features of the host */
992
    if (virNodeSuspendGetTargetMask(&caps->host.powerMgmt) < 0)
993 994
        VIR_WARN("Failed to get host power management capabilities");

M
Michal Privoznik 已提交
995 996 997 998
    /* Add huge pages info */
    if (virQEMUCapsInitPages(caps) < 0)
        VIR_WARN("Failed to get pages info");

999 1000 1001
    /* Add domain migration transport URIs */
    virCapabilitiesAddHostMigrateTransport(caps, "tcp");
    virCapabilitiesAddHostMigrateTransport(caps, "rdma");
1002

1003 1004 1005 1006
    /* QEMU can support pretty much every arch that exists,
     * so just probe for them all - we gracefully fail
     * if a qemu-system-$ARCH binary can't be found
     */
1007
    for (i = 0; i < VIR_ARCH_LAST; i++)
1008
        if (virQEMUCapsInitGuest(caps, cache,
T
Tal Kain 已提交
1009
                                 hostarch,
1010
                                 i) < 0)
1011
            goto error;
1012 1013 1014

    return caps;

1015
 error:
1016
    virObjectUnref(caps);
1017 1018 1019 1020
    return NULL;
}


1021
static int
1022 1023
virQEMUCapsComputeCmdFlags(const char *help,
                           unsigned int version,
1024
                           bool is_kvm,
1025 1026 1027
                           unsigned int kvm_version,
                           virQEMUCapsPtr qemuCaps,
                           bool check_yajl ATTRIBUTE_UNUSED)
1028 1029
{
    const char *p;
R
Richa Marwaha 已提交
1030
    const char *fsdev, *netdev;
1031 1032

    if (strstr(help, "-no-kqemu"))
1033
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_KQEMU);
1034
    if (strstr(help, "-enable-kqemu"))
1035
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_ENABLE_KQEMU);
1036
    if (strstr(help, "-no-kvm"))
1037
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_KVM);
1038
    if (strstr(help, "-enable-kvm"))
1039
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_ENABLE_KVM);
1040
    if (strstr(help, "-no-reboot"))
1041
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_REBOOT);
1042
    if (strstr(help, "-name")) {
1043
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NAME);
1044
        if (strstr(help, ",process="))
1045
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_NAME_PROCESS);
1046 1047
    }
    if (strstr(help, "-uuid"))
1048
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_UUID);
1049
    if (strstr(help, "-xen-domid"))
1050
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_XEN_DOMID);
1051
    else if (strstr(help, "-domid"))
1052
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DOMID);
1053
    if (strstr(help, "-drive")) {
1054 1055
        const char *cache = strstr(help, "cache=");

1056
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE);
1057 1058
        if (cache && (p = strchr(cache, ']'))) {
            if (memmem(cache, p - cache, "on|off", sizeof("on|off") - 1) == NULL)
1059
                virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_V2);
1060
            if (memmem(cache, p - cache, "directsync", sizeof("directsync") - 1))
1061
                virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_DIRECTSYNC);
1062
            if (memmem(cache, p - cache, "unsafe", sizeof("unsafe") - 1))
1063
                virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_UNSAFE);
1064
        }
1065
        if (strstr(help, "format="))
1066
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_FORMAT);
1067
        if (strstr(help, "readonly="))
1068
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_READONLY);
1069
        if (strstr(help, "aio=threads|native"))
1070
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_AIO);
O
Osier Yang 已提交
1071
        if (strstr(help, "copy-on-read=on|off"))
1072
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_COPY_ON_READ);
1073
        if (strstr(help, "bps="))
1074
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_IOTUNE);
1075 1076 1077 1078
    }
    if ((p = strstr(help, "-vga")) && !strstr(help, "-std-vga")) {
        const char *nl = strstr(p, "\n");

1079
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA);
1080 1081

        if (strstr(p, "|qxl"))
1082
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA_QXL);
1083
        if ((p = strstr(p, "|none")) && p < nl)
1084
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA_NONE);
1085 1086
    }
    if (strstr(help, "-spice"))
1087
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SPICE);
1088
    if (strstr(help, "-vnc"))
1089
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC);
1090
    if (strstr(help, "seamless-migration="))
1091
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION);
1092
    if (strstr(help, "boot=on"))
1093
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_BOOT);
1094
    if (strstr(help, "serial=s"))
1095
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_SERIAL);
1096
    if (strstr(help, "-pcidevice"))
1097
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_PCIDEVICE);
1098 1099
    if (strstr(help, "host=[seg:]bus"))
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_HOST_PCI_MULTIDOMAIN);
1100
    if (strstr(help, "-mem-path"))
1101
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MEM_PATH);
1102
    if (strstr(help, "-chardev")) {
1103
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_CHARDEV);
1104
        if (strstr(help, "-chardev spicevmc"))
1105
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_CHARDEV_SPICEVMC);
1106 1107
        if (strstr(help, "-chardev spiceport"))
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_CHARDEV_SPICEPORT);
1108
    }
1109
    if (strstr(help, "-balloon"))
1110
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_BALLOON);
1111
    if (strstr(help, "-device")) {
1112
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE);
1113 1114 1115 1116
        /*
         * When -device was introduced, qemu already supported drive's
         * readonly option but didn't advertise that.
         */
1117
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_READONLY);
1118 1119
    }
    if (strstr(help, "-nodefconfig"))
1120
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NODEFCONFIG);
1121
    if (strstr(help, "-no-user-config"))
1122
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_USER_CONFIG);
1123 1124
    /* The trailing ' ' is important to avoid a bogus match */
    if (strstr(help, "-rtc "))
1125
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_RTC);
1126 1127
    /* to wit */
    if (strstr(help, "-rtc-td-hack"))
1128
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_RTC_TD_HACK);
1129
    if (strstr(help, "-no-hpet"))
1130
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_HPET);
1131
    if (strstr(help, "-no-acpi"))
1132
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_ACPI);
1133
    if (strstr(help, "-no-kvm-pit-reinjection"))
1134
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_KVM_PIT);
1135
    if (strstr(help, "-tdf"))
1136
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_TDF);
1137
    if (strstr(help, "-enable-nesting"))
1138
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NESTING);
1139
    if (strstr(help, ",menu=on"))
1140
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_BOOT_MENU);
1141
    if (strstr(help, ",reboot-timeout=rb_time"))
1142
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_REBOOT_TIMEOUT);
1143 1144
    if (strstr(help, ",splash-time=sp_time"))
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SPLASH_TIMEOUT);
1145
    if ((fsdev = strstr(help, "-fsdev"))) {
1146
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV);
1147
        if (strstr(fsdev, "readonly"))
1148
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV_READONLY);
1149
        if (strstr(fsdev, "writeout"))
1150
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV_WRITEOUT);
1151
    }
1152
    if (strstr(help, "-smbios type"))
1153
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SMBIOS_TYPE);
1154
    if (strstr(help, "-sandbox"))
1155
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SECCOMP_SANDBOX);
1156

R
Richa Marwaha 已提交
1157
    if ((netdev = strstr(help, "-netdev"))) {
1158 1159
        /* Disable -netdev on 0.12 since although it exists,
         * the corresponding netdev_add/remove monitor commands
1160 1161
         * do not, and we need them to be able to do hotplug.
         * But see below about RHEL build. */
R
Richa Marwaha 已提交
1162 1163
        if (version >= 13000) {
            if (strstr(netdev, "bridge"))
1164 1165
                virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV_BRIDGE);
           virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV);
R
Richa Marwaha 已提交
1166
        }
1167 1168 1169
    }

    if (strstr(help, "-sdl"))
1170
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SDL);
1171 1172 1173
    if (strstr(help, "cores=") &&
        strstr(help, "threads=") &&
        strstr(help, "sockets="))
1174
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SMP_TOPOLOGY);
1175 1176

    if (version >= 9000)
1177
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_COLON);
1178 1179

    if (is_kvm && (version >= 10000 || kvm_version >= 74))
1180
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNET_HDR);
1181

1182
    if (strstr(help, ",vhost="))
1183
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VHOST_NET);
1184

1185 1186
    /* Do not use -no-shutdown if qemu doesn't support it or SIGTERM handling
     * is most likely buggy when used with -no-shutdown (which applies for qemu
1187
     * 0.14.* and 0.15.0)
1188
     */
1189
    if (strstr(help, "-no-shutdown") && (version < 14000 || version > 15000))
1190
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_SHUTDOWN);
1191

1192
    if (strstr(help, "dump-guest-core=on|off"))
1193
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DUMP_GUEST_CORE);
1194

O
Olivia Yin 已提交
1195 1196 1197
    if (strstr(help, "-dtb"))
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DTB);

1198 1199 1200
    if (strstr(help, "-machine"))
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MACHINE_OPT);

1201 1202 1203 1204
    /*
     * Handling of -incoming arg with varying features
     *  -incoming tcp    (kvm >= 79, qemu >= 0.10.0)
     *  -incoming exec   (kvm >= 80, qemu >= 0.10.0)
1205 1206
     *  -incoming unix   (qemu >= 0.12.0)
     *  -incoming fd     (qemu >= 0.12.0)
1207 1208 1209 1210 1211 1212 1213
     *  -incoming stdio  (all earlier kvm)
     *
     * NB, there was a pre-kvm-79 'tcp' support, but it
     * was broken, because it blocked the monitor console
     * while waiting for data, so pretend it doesn't exist
     */
    if (version >= 10000) {
1214 1215
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_TCP);
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_EXEC);
1216
        if (version >= 12000) {
1217 1218
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX);
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD);
1219
        }
1220
    } else if (kvm_version >= 79) {
1221
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_TCP);
1222
        if (kvm_version >= 80)
1223
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_EXEC);
1224
    } else if (kvm_version > 0) {
1225
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_KVM_STDIO);
1226 1227 1228
    }

    if (version >= 10000)
1229
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_0_10);
1230

1231
    if (version >= 11000)
1232
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VIRTIO_BLK_SG_IO);
1233

1234 1235 1236
    /* While JSON mode was available in 0.12.0, it was too
     * incomplete to contemplate using. The 0.13.0 release
     * is good enough to use, even though it lacks one or
1237 1238 1239 1240 1241
     * two features. This is also true of versions of qemu
     * built for RHEL, labeled 0.12.1, but with extra text
     * in the help output that mentions that features were
     * backported for libvirt. The benefits of JSON mode now
     * outweigh the downside.
1242
     */
1243
#if WITH_YAJL
1244
    if (version >= 13000) {
1245
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MONITOR_JSON);
1246 1247
    } else if (version >= 12000 &&
               strstr(help, "libvirt")) {
1248 1249
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MONITOR_JSON);
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV);
1250
    }
1251 1252 1253 1254 1255 1256 1257
#else
    /* Starting with qemu 0.15 and newer, upstream qemu no longer
     * promises to keep the human interface stable, but requests that
     * we use QMP (the JSON interface) for everything.  If the user
     * forgot to include YAJL libraries when building their own
     * libvirt but is targetting a newer qemu, we are better off
     * telling them to recompile (the spec file includes the
1258
     * dependency, so distros won't hit this).  This check is
1259
     * also in m4/virt-yajl.m4 (see $with_yajl).  */
1260 1261 1262
    if (version >= 15000 ||
        (version >= 12000 && strstr(help, "libvirt"))) {
        if (check_yajl) {
1263 1264 1265
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("this qemu binary requires libvirt to be "
                             "compiled with yajl"));
1266 1267
            return -1;
        }
1268
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV);
1269
    }
E
Eric Blake 已提交
1270
#endif
1271 1272

    if (version >= 13000)
1273
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_PCI_MULTIFUNCTION);
1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285

    /* Although very new versions of qemu advertise the presence of
     * the rombar option in the output of "qemu -device pci-assign,?",
     * this advertisement was added to the code long after the option
     * itself. According to qemu developers, though, rombar is
     * available in all qemu binaries from release 0.12 onward.
     * Setting the capability this way makes it available in more
     * cases where it might be needed, and shouldn't cause any false
     * positives (in the case that it did, qemu would produce an error
     * log and refuse to start, so it would be immediately obvious).
     */
    if (version >= 12000)
1286
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_PCI_ROMBAR);
1287 1288

    if (version >= 11000)
1289
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_CPU_HOST);
1290

1291
    if (version >= 1001000) {
J
Ján Tomko 已提交
1292
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_IPV6_MIGRATION);
1293 1294
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_SHARE_POLICY);
    }
J
Ján Tomko 已提交
1295

1296
    return 0;
1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322
}

/* We parse the output of 'qemu -help' to get the QEMU
 * version number. The first bit is easy, just parse
 * 'QEMU PC emulator version x.y.z'
 * or
 * 'QEMU emulator version x.y.z'.
 *
 * With qemu-kvm, however, that is followed by a string
 * in parenthesis as follows:
 *  - qemu-kvm-x.y.z in stable releases
 *  - kvm-XX for kvm versions up to kvm-85
 *  - qemu-kvm-devel-XX for kvm version kvm-86 and later
 *
 * For qemu-kvm versions before 0.10.z, we need to detect
 * the KVM version number for some features. With 0.10.z
 * and later, we just need the QEMU version number and
 * whether it is KVM QEMU or mainline QEMU.
 */
#define QEMU_VERSION_STR_1  "QEMU emulator version"
#define QEMU_VERSION_STR_2  "QEMU PC emulator version"
#define QEMU_KVM_VER_PREFIX "(qemu-kvm-"
#define KVM_VER_PREFIX      "(kvm-"

#define SKIP_BLANKS(p) do { while ((*(p) == ' ') || (*(p) == '\t')) (p)++; } while (0)

1323 1324 1325 1326
int virQEMUCapsParseHelpStr(const char *qemu,
                            const char *help,
                            virQEMUCapsPtr qemuCaps,
                            unsigned int *version,
1327
                            bool *is_kvm,
1328 1329
                            unsigned int *kvm_version,
                            bool check_yajl)
1330 1331 1332
{
    unsigned major, minor, micro;
    const char *p = help;
1333
    char *strflags;
1334

1335 1336
    *version = *kvm_version = 0;
    *is_kvm = false;
1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353

    if (STRPREFIX(p, QEMU_VERSION_STR_1))
        p += strlen(QEMU_VERSION_STR_1);
    else if (STRPREFIX(p, QEMU_VERSION_STR_2))
        p += strlen(QEMU_VERSION_STR_2);
    else
        goto fail;

    SKIP_BLANKS(p);

    major = virParseNumber(&p);
    if (major == -1 || *p != '.')
        goto fail;

    ++p;

    minor = virParseNumber(&p);
J
Jiri Denemark 已提交
1354
    if (minor == -1)
1355 1356
        goto fail;

J
Jiri Denemark 已提交
1357 1358 1359 1360 1361 1362 1363 1364
    if (*p != '.') {
        micro = 0;
    } else {
        ++p;
        micro = virParseNumber(&p);
        if (micro == -1)
            goto fail;
    }
1365 1366 1367 1368

    SKIP_BLANKS(p);

    if (STRPREFIX(p, QEMU_KVM_VER_PREFIX)) {
1369
        *is_kvm = true;
1370 1371 1372 1373
        p += strlen(QEMU_KVM_VER_PREFIX);
    } else if (STRPREFIX(p, KVM_VER_PREFIX)) {
        int ret;

1374
        *is_kvm = true;
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385
        p += strlen(KVM_VER_PREFIX);

        ret = virParseNumber(&p);
        if (ret == -1)
            goto fail;

        *kvm_version = ret;
    }

    *version = (major * 1000 * 1000) + (minor * 1000) + micro;

1386 1387 1388 1389 1390 1391 1392 1393 1394 1395
    /* Refuse to parse -help output for QEMU releases >= 1.2.0 that should be
     * using QMP probing.
     */
    if (*version >= 1002000) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       _("QEMU %u.%u.%u is too new for help parsing"),
                       major, minor, micro);
        goto cleanup;
    }

1396 1397
    if (virQEMUCapsComputeCmdFlags(help, *version, *is_kvm, *kvm_version,
                                   qemuCaps, check_yajl) < 0)
1398
        goto cleanup;
1399

1400
    strflags = virBitmapString(qemuCaps->flags);
1401 1402 1403
    VIR_DEBUG("Version %u.%u.%u, cooked version %u, flags %s",
              major, minor, micro, *version, NULLSTR(strflags));
    VIR_FREE(strflags);
1404 1405 1406 1407 1408 1409 1410 1411

    if (*kvm_version)
        VIR_DEBUG("KVM version %d detected", *kvm_version);
    else if (*is_kvm)
        VIR_DEBUG("qemu-kvm version %u.%u.%u detected", major, minor, micro);

    return 0;

1412
 fail:
1413
    p = strchr(help, '\n');
1414 1415
    if (!p)
        p = strchr(help, '\0');
1416

1417 1418 1419
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("cannot parse %s version number in '%.*s'"),
                   qemu, (int) (p - help), help);
1420

1421
 cleanup:
1422 1423 1424
    return -1;
}

1425

1426
struct virQEMUCapsStringFlags {
1427 1428 1429 1430 1431
    const char *value;
    int flag;
};


1432 1433 1434
struct virQEMUCapsStringFlags virQEMUCapsCommands[] = {
    { "system_wakeup", QEMU_CAPS_WAKEUP },
    { "transaction", QEMU_CAPS_TRANSACTION },
1435 1436
    { "block_stream", QEMU_CAPS_BLOCKJOB_SYNC },
    { "block-stream", QEMU_CAPS_BLOCKJOB_ASYNC },
1437 1438 1439 1440 1441 1442 1443 1444 1445
    { "dump-guest-memory", QEMU_CAPS_DUMP_GUEST_MEMORY },
    { "query-spice", QEMU_CAPS_SPICE },
    { "query-kvm", QEMU_CAPS_KVM },
    { "block-commit", QEMU_CAPS_BLOCK_COMMIT },
    { "query-vnc", QEMU_CAPS_VNC },
    { "drive-mirror", QEMU_CAPS_DRIVE_MIRROR },
    { "blockdev-snapshot-sync", QEMU_CAPS_DISK_SNAPSHOT },
    { "add-fd", QEMU_CAPS_ADD_FD },
    { "nbd-server-start", QEMU_CAPS_NBD_SERVER },
1446
    { "change-backing-file", QEMU_CAPS_CHANGE_BACKING_FILE },
1447
    { "rtc-reset-reinjection", QEMU_CAPS_RTC_RESET_REINJECTION },
1448 1449
};

1450 1451 1452 1453
struct virQEMUCapsStringFlags virQEMUCapsMigration[] = {
    { "rdma-pin-all", QEMU_CAPS_MIGRATE_RDMA },
};

1454 1455 1456
struct virQEMUCapsStringFlags virQEMUCapsEvents[] = {
    { "BALLOON_CHANGE", QEMU_CAPS_BALLOON_EVENT },
    { "SPICE_MIGRATE_COMPLETED", QEMU_CAPS_SEAMLESS_MIGRATION },
1457
    { "DEVICE_DELETED", QEMU_CAPS_DEVICE_DEL_EVENT },
1458 1459
};

1460
struct virQEMUCapsStringFlags virQEMUCapsObjectTypes[] = {
1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475
    { "hda-duplex", QEMU_CAPS_HDA_DUPLEX },
    { "hda-micro", QEMU_CAPS_HDA_MICRO },
    { "ccid-card-emulated", QEMU_CAPS_CCID_EMULATED },
    { "ccid-card-passthru", QEMU_CAPS_CCID_PASSTHRU },
    { "piix3-usb-uhci", QEMU_CAPS_PIIX3_USB_UHCI },
    { "piix4-usb-uhci", QEMU_CAPS_PIIX4_USB_UHCI },
    { "usb-ehci", QEMU_CAPS_USB_EHCI },
    { "ich9-usb-ehci1", QEMU_CAPS_ICH9_USB_EHCI1 },
    { "vt82c686b-usb-uhci", QEMU_CAPS_VT82C686B_USB_UHCI },
    { "pci-ohci", QEMU_CAPS_PCI_OHCI },
    { "nec-usb-xhci", QEMU_CAPS_NEC_USB_XHCI },
    { "usb-redir", QEMU_CAPS_USB_REDIR },
    { "usb-hub", QEMU_CAPS_USB_HUB },
    { "ich9-ahci", QEMU_CAPS_ICH9_AHCI },
    { "virtio-blk-s390", QEMU_CAPS_VIRTIO_S390 },
1476
    { "virtio-blk-ccw", QEMU_CAPS_VIRTIO_CCW },
1477
    { "sclpconsole", QEMU_CAPS_SCLP_S390 },
1478
    { "lsi53c895a", QEMU_CAPS_SCSI_LSI },
1479
    { "virtio-scsi-pci", QEMU_CAPS_VIRTIO_SCSI },
1480 1481
    { "virtio-scsi-s390", QEMU_CAPS_VIRTIO_SCSI },
    { "virtio-scsi-ccw", QEMU_CAPS_VIRTIO_SCSI },
1482
    { "megasas", QEMU_CAPS_SCSI_MEGASAS },
1483 1484
    { "spicevmc", QEMU_CAPS_DEVICE_SPICEVMC },
    { "qxl-vga", QEMU_CAPS_DEVICE_QXL_VGA },
1485
    { "qxl", QEMU_CAPS_DEVICE_QXL },
1486 1487 1488 1489
    { "sga", QEMU_CAPS_SGA },
    { "scsi-block", QEMU_CAPS_SCSI_BLOCK },
    { "scsi-cd", QEMU_CAPS_SCSI_CD },
    { "ide-cd", QEMU_CAPS_IDE_CD },
1490 1491 1492
    { "VGA", QEMU_CAPS_DEVICE_VGA },
    { "cirrus-vga", QEMU_CAPS_DEVICE_CIRRUS_VGA },
    { "vmware-svga", QEMU_CAPS_DEVICE_VMWARE_SVGA },
H
Han Cheng 已提交
1493 1494
    { "usb-serial", QEMU_CAPS_DEVICE_USB_SERIAL },
    { "usb-net", QEMU_CAPS_DEVICE_USB_NET },
1495
    { "virtio-rng-pci", QEMU_CAPS_DEVICE_VIRTIO_RNG },
1496 1497
    { "virtio-rng-s390", QEMU_CAPS_DEVICE_VIRTIO_RNG },
    { "virtio-rng-ccw", QEMU_CAPS_DEVICE_VIRTIO_RNG },
1498
    { "rng-random", QEMU_CAPS_OBJECT_RNG_RANDOM },
1499
    { "rng-egd", QEMU_CAPS_OBJECT_RNG_EGD },
1500
    { "spapr-nvram", QEMU_CAPS_DEVICE_NVRAM },
1501
    { "pci-bridge", QEMU_CAPS_DEVICE_PCI_BRIDGE },
1502
    { "vfio-pci", QEMU_CAPS_DEVICE_VFIO_PCI },
H
Han Cheng 已提交
1503
    { "scsi-generic", QEMU_CAPS_DEVICE_SCSI_GENERIC },
1504
    { "i82801b11-bridge", QEMU_CAPS_DEVICE_DMI_TO_PCI_BRIDGE },
1505
    { "usb-storage", QEMU_CAPS_DEVICE_USB_STORAGE },
1506
    { "virtio-mmio", QEMU_CAPS_DEVICE_VIRTIO_MMIO },
1507
    { "ich9-intel-hda", QEMU_CAPS_DEVICE_ICH9_INTEL_HDA },
H
Hu Tao 已提交
1508
    { "pvpanic", QEMU_CAPS_DEVICE_PANIC },
L
Li Zhang 已提交
1509
    { "usb-kbd", QEMU_CAPS_DEVICE_USB_KBD },
1510
    { "memory-backend-ram", QEMU_CAPS_OBJECT_MEMORY_RAM },
1511
    { "memory-backend-file", QEMU_CAPS_OBJECT_MEMORY_FILE },
1512
    { "usb-audio", QEMU_CAPS_OBJECT_USB_AUDIO },
J
John Ferlan 已提交
1513
    { "iothread", QEMU_CAPS_OBJECT_IOTHREAD},
1514
    { "ivshmem", QEMU_CAPS_DEVICE_IVSHMEM },
1515 1516
};

1517
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsVirtioBlk[] = {
1518 1519 1520 1521 1522 1523 1524 1525
    { "multifunction", QEMU_CAPS_PCI_MULTIFUNCTION },
    { "bootindex", QEMU_CAPS_BOOTINDEX },
    { "ioeventfd", QEMU_CAPS_VIRTIO_IOEVENTFD },
    { "event_idx", QEMU_CAPS_VIRTIO_BLK_EVENT_IDX },
    { "scsi", QEMU_CAPS_VIRTIO_BLK_SCSI },
    { "logical_block_size", QEMU_CAPS_BLOCKIO },
};

1526
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsVirtioNet[] = {
1527 1528 1529 1530
    { "tx", QEMU_CAPS_VIRTIO_TX_ALG },
    { "event_idx", QEMU_CAPS_VIRTIO_NET_EVENT_IDX },
};

1531
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsPCIAssign[] = {
1532
    { "rombar", QEMU_CAPS_PCI_ROMBAR },
1533 1534 1535 1536
    { "configfd", QEMU_CAPS_PCI_CONFIGFD },
    { "bootindex", QEMU_CAPS_PCI_BOOTINDEX },
};

1537
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsVfioPCI[] = {
1538 1539 1540
    { "bootindex", QEMU_CAPS_VFIO_PCI_BOOTINDEX },
};

1541
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsSCSIDisk[] = {
1542 1543 1544 1545
    { "channel", QEMU_CAPS_SCSI_DISK_CHANNEL },
    { "wwn", QEMU_CAPS_SCSI_DISK_WWN },
};

1546
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsIDEDrive[] = {
1547 1548 1549
    { "wwn", QEMU_CAPS_IDE_DRIVE_WWN },
};

1550
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsPixx4PM[] = {
1551 1552 1553 1554
    { "disable_s3", QEMU_CAPS_DISABLE_S3 },
    { "disable_s4", QEMU_CAPS_DISABLE_S4 },
};

1555
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsUSBRedir[] = {
1556
    { "filter", QEMU_CAPS_USB_REDIR_FILTER },
1557 1558 1559
    { "bootindex", QEMU_CAPS_USB_REDIR_BOOTINDEX },
};

1560
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsUSBHost[] = {
1561
    { "bootindex", QEMU_CAPS_USB_HOST_BOOTINDEX },
1562 1563
};

1564
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsSCSIGeneric[] = {
H
Han Cheng 已提交
1565 1566 1567
    { "bootindex", QEMU_CAPS_DEVICE_SCSI_GENERIC_BOOTINDEX },
};

1568
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsI440FXPCIHost[] = {
1569 1570 1571
    { "pci-hole64-size", QEMU_CAPS_I440FX_PCI_HOLE64_SIZE },
};

1572
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsQ35PCIHost[] = {
1573 1574 1575
    { "pci-hole64-size", QEMU_CAPS_Q35_PCI_HOLE64_SIZE },
};

1576
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsUSBStorage[] = {
1577 1578 1579
    { "removable", QEMU_CAPS_USB_STORAGE_REMOVABLE },
};

1580 1581 1582 1583
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsKVMPit[] = {
    { "lost_tick_policy", QEMU_CAPS_KVM_PIT_TICK_POLICY },
};

1584
struct virQEMUCapsObjectTypeProps {
1585
    const char *type;
1586
    struct virQEMUCapsStringFlags *props;
1587 1588 1589
    size_t nprops;
};

1590 1591 1592 1593
static struct virQEMUCapsObjectTypeProps virQEMUCapsObjectProps[] = {
    { "virtio-blk-pci", virQEMUCapsObjectPropsVirtioBlk,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioBlk) },
    { "virtio-net-pci", virQEMUCapsObjectPropsVirtioNet,
1594 1595 1596 1597
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioNet) },
    { "virtio-blk-ccw", virQEMUCapsObjectPropsVirtioBlk,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioBlk) },
    { "virtio-net-ccw", virQEMUCapsObjectPropsVirtioNet,
1598 1599 1600 1601 1602
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioNet) },
    { "virtio-blk-s390", virQEMUCapsObjectPropsVirtioBlk,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioBlk) },
    { "virtio-net-s390", virQEMUCapsObjectPropsVirtioNet,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioNet) },
1603 1604 1605 1606 1607 1608
    { "pci-assign", virQEMUCapsObjectPropsPCIAssign,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsPCIAssign) },
    { "kvm-pci-assign", virQEMUCapsObjectPropsPCIAssign,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsPCIAssign) },
    { "vfio-pci", virQEMUCapsObjectPropsVfioPCI,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVfioPCI) },
1609 1610
    { "scsi-disk", virQEMUCapsObjectPropsSCSIDisk,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsSCSIDisk) },
1611 1612 1613 1614
    { "ide-drive", virQEMUCapsObjectPropsIDEDrive,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsIDEDrive) },
    { "PIIX4_PM", virQEMUCapsObjectPropsPixx4PM,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsPixx4PM) },
1615 1616 1617 1618
    { "usb-redir", virQEMUCapsObjectPropsUSBRedir,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsUSBRedir) },
    { "usb-host", virQEMUCapsObjectPropsUSBHost,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsUSBHost) },
1619 1620
    { "scsi-generic", virQEMUCapsObjectPropsSCSIGeneric,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsSCSIGeneric) },
1621 1622 1623 1624
    { "i440FX-pcihost", virQEMUCapsObjectPropsI440FXPCIHost,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsI440FXPCIHost) },
    { "q35-pcihost", virQEMUCapsObjectPropsQ35PCIHost,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsQ35PCIHost) },
1625 1626
    { "usb-storage", virQEMUCapsObjectPropsUSBStorage,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsUSBStorage) },
1627 1628
    { "kvm-pit", virQEMUCapsObjectPropsKVMPit,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsKVMPit) },
1629 1630 1631 1632
};


static void
1633 1634 1635 1636 1637
virQEMUCapsProcessStringFlags(virQEMUCapsPtr qemuCaps,
                              size_t nflags,
                              struct virQEMUCapsStringFlags *flags,
                              size_t nvalues,
                              char *const*values)
1638 1639
{
    size_t i, j;
1640 1641
    for (i = 0; i < nflags; i++) {
        for (j = 0; j < nvalues; j++) {
1642
            if (STREQ(values[j], flags[i].value)) {
1643
                virQEMUCapsSet(qemuCaps, flags[i].flag);
1644 1645 1646 1647 1648 1649 1650 1651
                break;
            }
        }
    }
}


static void
1652 1653
virQEMUCapsFreeStringList(size_t len,
                          char **values)
1654 1655
{
    size_t i;
1656
    for (i = 0; i < len; i++)
1657 1658 1659 1660 1661 1662 1663 1664
        VIR_FREE(values[i]);
    VIR_FREE(values);
}


#define OBJECT_TYPE_PREFIX "name \""

static int
1665 1666
virQEMUCapsParseDeviceStrObjectTypes(const char *str,
                                     char ***types)
1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684
{
    const char *tmp = str;
    int ret = -1;
    size_t ntypelist = 0;
    char **typelist = NULL;

    *types = NULL;

    while ((tmp = strstr(tmp, OBJECT_TYPE_PREFIX))) {
        char *end;
        tmp += strlen(OBJECT_TYPE_PREFIX);
        end = strstr(tmp, "\"");
        if (!end) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Malformed QEMU device list string, missing quote"));
            goto cleanup;
        }

1685
        if (VIR_EXPAND_N(typelist, ntypelist, 1) < 0)
1686
            goto cleanup;
1687
        if (VIR_STRNDUP(typelist[ntypelist - 1], tmp, end-tmp) < 0)
1688 1689 1690 1691 1692 1693
            goto cleanup;
    }

    *types = typelist;
    ret = ntypelist;

1694
 cleanup:
1695
    if (ret < 0)
1696
        virQEMUCapsFreeStringList(ntypelist, typelist);
1697 1698 1699 1700 1701
    return ret;
}


static int
1702 1703 1704
virQEMUCapsParseDeviceStrObjectProps(const char *str,
                                     const char *type,
                                     char ***props)
1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737
{
    const char *tmp = str;
    int ret = -1;
    size_t nproplist = 0;
    char **proplist = NULL;

    VIR_DEBUG("Extract type %s", type);
    *props = NULL;

    while ((tmp = strchr(tmp, '\n'))) {
        char *end;
        tmp += 1;

        if (*tmp == '\0')
            break;

        if (STRPREFIX(tmp, OBJECT_TYPE_PREFIX))
            continue;

        if (!STRPREFIX(tmp, type))
            continue;

        tmp += strlen(type);
        if (*tmp != '.')
            continue;
        tmp++;

        end = strstr(tmp, "=");
        if (!end) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Malformed QEMU device list string, missing '='"));
            goto cleanup;
        }
1738
        if (VIR_EXPAND_N(proplist, nproplist, 1) < 0)
1739
            goto cleanup;
1740
        if (VIR_STRNDUP(proplist[nproplist - 1], tmp, end-tmp) < 0)
1741 1742 1743 1744 1745 1746
            goto cleanup;
    }

    *props = proplist;
    ret = nproplist;

1747
 cleanup:
1748
    if (ret < 0 && proplist)
1749
        virQEMUCapsFreeStringList(nproplist, proplist);
1750 1751 1752 1753 1754
    return ret;
}


int
1755
virQEMUCapsParseDeviceStr(virQEMUCapsPtr qemuCaps, const char *str)
1756 1757 1758 1759 1760
{
    int nvalues;
    char **values;
    size_t i;

1761
    if ((nvalues = virQEMUCapsParseDeviceStrObjectTypes(str, &values)) < 0)
1762
        return -1;
1763 1764 1765 1766 1767 1768
    virQEMUCapsProcessStringFlags(qemuCaps,
                                  ARRAY_CARDINALITY(virQEMUCapsObjectTypes),
                                  virQEMUCapsObjectTypes,
                                  nvalues, values);
    virQEMUCapsFreeStringList(nvalues, values);

1769
    for (i = 0; i < ARRAY_CARDINALITY(virQEMUCapsObjectProps); i++) {
1770 1771 1772 1773
        const char *type = virQEMUCapsObjectProps[i].type;
        if ((nvalues = virQEMUCapsParseDeviceStrObjectProps(str,
                                                            type,
                                                            &values)) < 0)
1774
            return -1;
1775 1776 1777 1778 1779
        virQEMUCapsProcessStringFlags(qemuCaps,
                                      virQEMUCapsObjectProps[i].nprops,
                                      virQEMUCapsObjectProps[i].props,
                                      nvalues, values);
        virQEMUCapsFreeStringList(nvalues, values);
1780 1781 1782
    }

    /* Prefer -chardev spicevmc (detected earlier) over -device spicevmc */
1783 1784
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_CHARDEV_SPICEVMC))
        virQEMUCapsClear(qemuCaps, QEMU_CAPS_DEVICE_SPICEVMC);
1785 1786 1787 1788 1789

    return 0;
}


E
Eric Blake 已提交
1790
static int
1791 1792
virQEMUCapsExtractDeviceStr(const char *qemu,
                            virQEMUCapsPtr qemuCaps,
1793
                            uid_t runUid, gid_t runGid)
1794
{
E
Eric Blake 已提交
1795
    char *output = NULL;
1796
    virCommandPtr cmd;
E
Eric Blake 已提交
1797
    int ret = -1;
1798

E
Eric Blake 已提交
1799 1800
    /* Cram together all device-related queries into one invocation;
     * the output format makes it possible to distinguish what we
1801 1802
     * need.  With qemu 0.13.0 and later, unrecognized '-device
     * bogus,?' cause an error in isolation, but are silently ignored
1803
     * in combination with '-device ?'.  Upstream qemu 0.12.x doesn't
1804 1805
     * understand '-device name,?', and always exits with status 1 for
     * the simpler '-device ?', so this function is really only useful
1806
     * if -help includes "device driver,?".  */
1807
    cmd = virQEMUCapsProbeCommand(qemu, qemuCaps, runUid, runGid);
1808 1809 1810 1811 1812 1813
    virCommandAddArgList(cmd,
                         "-device", "?",
                         "-device", "pci-assign,?",
                         "-device", "virtio-blk-pci,?",
                         "-device", "virtio-net-pci,?",
                         "-device", "scsi-disk,?",
1814
                         "-device", "PIIX4_PM,?",
1815
                         "-device", "usb-redir,?",
1816
                         "-device", "ide-drive,?",
1817
                         "-device", "usb-host,?",
H
Han Cheng 已提交
1818
                         "-device", "scsi-generic,?",
1819
                         "-device", "usb-storage,?",
1820
                         NULL);
1821
    /* qemu -help goes to stdout, but qemu -device ? goes to stderr.  */
E
Eric Blake 已提交
1822
    virCommandSetErrorBuffer(cmd, &output);
1823

1824
    if (virCommandRun(cmd, NULL) < 0)
1825 1826
        goto cleanup;

1827
    ret = virQEMUCapsParseDeviceStr(qemuCaps, output);
1828

1829
 cleanup:
E
Eric Blake 已提交
1830
    VIR_FREE(output);
1831
    virCommandFree(cmd);
E
Eric Blake 已提交
1832 1833 1834
    return ret;
}

1835

1836 1837 1838
int virQEMUCapsGetDefaultVersion(virCapsPtr caps,
                                 virQEMUCapsCachePtr capsCache,
                                 unsigned int *version)
1839 1840
{
    const char *binary;
1841
    virQEMUCapsPtr qemucaps;
T
Tal Kain 已提交
1842
    virArch hostarch;
1843 1844 1845 1846

    if (*version > 0)
        return 0;

T
Tal Kain 已提交
1847
    hostarch = virArchFromHost();
1848 1849
    if ((binary = virCapabilitiesDefaultGuestEmulator(caps,
                                                      "hvm",
T
Tal Kain 已提交
1850
                                                      hostarch,
1851
                                                      "qemu")) == NULL) {
1852
        virReportError(VIR_ERR_INTERNAL_ERROR,
1853
                       _("Cannot find suitable emulator for %s"),
T
Tal Kain 已提交
1854
                       virArchToString(hostarch));
1855 1856 1857
        return -1;
    }

1858
    if (!(qemucaps = virQEMUCapsCacheLookup(capsCache, binary)))
1859 1860
        return -1;

1861
    *version = virQEMUCapsGetVersion(qemucaps);
1862
    virObjectUnref(qemucaps);
1863 1864
    return 0;
}
1865 1866


1867 1868


1869 1870
virQEMUCapsPtr
virQEMUCapsNew(void)
1871
{
1872
    virQEMUCapsPtr qemuCaps;
1873

1874
    if (virQEMUCapsInitialize() < 0)
1875 1876
        return NULL;

1877
    if (!(qemuCaps = virObjectNew(virQEMUCapsClass)))
1878 1879
        return NULL;

1880
    if (!(qemuCaps->flags = virBitmapNew(QEMU_CAPS_LAST)))
1881
        goto error;
1882

1883
    return qemuCaps;
1884

1885
 error:
1886
    virObjectUnref(qemuCaps);
1887
    return NULL;
1888 1889 1890
}


1891
virQEMUCapsPtr virQEMUCapsNewCopy(virQEMUCapsPtr qemuCaps)
1892
{
1893
    virQEMUCapsPtr ret = virQEMUCapsNew();
1894 1895 1896 1897 1898
    size_t i;

    if (!ret)
        return NULL;

1899
    virBitmapCopy(ret->flags, qemuCaps->flags);
1900

1901 1902 1903 1904
    ret->usedQMP = qemuCaps->usedQMP;
    ret->version = qemuCaps->version;
    ret->kvmVersion = qemuCaps->kvmVersion;
    ret->arch = qemuCaps->arch;
1905

1906
    if (VIR_ALLOC_N(ret->cpuDefinitions, qemuCaps->ncpuDefinitions) < 0)
1907
        goto error;
1908
    ret->ncpuDefinitions = qemuCaps->ncpuDefinitions;
1909
    for (i = 0; i < qemuCaps->ncpuDefinitions; i++) {
1910 1911
        if (VIR_STRDUP(ret->cpuDefinitions[i], qemuCaps->cpuDefinitions[i]) < 0)
            goto error;
1912 1913
    }

1914
    if (VIR_ALLOC_N(ret->machineTypes, qemuCaps->nmachineTypes) < 0)
1915
        goto error;
1916
    if (VIR_ALLOC_N(ret->machineAliases, qemuCaps->nmachineTypes) < 0)
1917
        goto error;
1918
    if (VIR_ALLOC_N(ret->machineMaxCpus, qemuCaps->nmachineTypes) < 0)
1919
        goto error;
1920
    ret->nmachineTypes = qemuCaps->nmachineTypes;
1921
    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
1922 1923 1924
        if (VIR_STRDUP(ret->machineTypes[i], qemuCaps->machineTypes[i]) < 0 ||
            VIR_STRDUP(ret->machineAliases[i], qemuCaps->machineAliases[i]) < 0)
            goto error;
1925
        ret->machineMaxCpus[i] = qemuCaps->machineMaxCpus[i];
1926 1927 1928 1929
    }

    return ret;

1930
 error:
1931 1932 1933 1934 1935
    virObjectUnref(ret);
    return NULL;
}


1936
void virQEMUCapsDispose(void *obj)
1937
{
1938
    virQEMUCapsPtr qemuCaps = obj;
1939 1940
    size_t i;

1941
    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
1942 1943
        VIR_FREE(qemuCaps->machineTypes[i]);
        VIR_FREE(qemuCaps->machineAliases[i]);
1944
    }
1945 1946
    VIR_FREE(qemuCaps->machineTypes);
    VIR_FREE(qemuCaps->machineAliases);
1947
    VIR_FREE(qemuCaps->machineMaxCpus);
1948

1949
    for (i = 0; i < qemuCaps->ncpuDefinitions; i++)
1950 1951
        VIR_FREE(qemuCaps->cpuDefinitions[i]);
    VIR_FREE(qemuCaps->cpuDefinitions);
1952

1953
    virBitmapFree(qemuCaps->flags);
1954

1955
    VIR_FREE(qemuCaps->binary);
1956 1957
}

1958
void
1959
virQEMUCapsSet(virQEMUCapsPtr qemuCaps,
1960
               virQEMUCapsFlags flag)
1961
{
1962
    ignore_value(virBitmapSetBit(qemuCaps->flags, flag));
1963 1964 1965 1966
}


void
1967
virQEMUCapsSetList(virQEMUCapsPtr qemuCaps, ...)
1968 1969 1970 1971
{
    va_list list;
    int flag;

1972
    va_start(list, qemuCaps);
1973
    while ((flag = va_arg(list, int)) < QEMU_CAPS_LAST)
1974
        ignore_value(virBitmapSetBit(qemuCaps->flags, flag));
1975
    va_end(list);
1976 1977 1978 1979
}


void
1980
virQEMUCapsClear(virQEMUCapsPtr qemuCaps,
1981
                 virQEMUCapsFlags flag)
1982
{
1983
    ignore_value(virBitmapClearBit(qemuCaps->flags, flag));
1984 1985 1986
}


1987
char *virQEMUCapsFlagsString(virQEMUCapsPtr qemuCaps)
1988
{
1989
    return virBitmapString(qemuCaps->flags);
1990 1991 1992 1993
}


bool
1994
virQEMUCapsGet(virQEMUCapsPtr qemuCaps,
1995
               virQEMUCapsFlags flag)
1996
{
1997 1998
    bool b;

1999
    if (!qemuCaps || virBitmapGetBit(qemuCaps->flags, flag, &b) < 0)
2000 2001 2002
        return false;
    else
        return b;
2003
}
2004 2005


D
Daniel P. Berrange 已提交
2006 2007 2008 2009 2010 2011 2012 2013 2014
bool virQEMUCapsHasPCIMultiBus(virQEMUCapsPtr qemuCaps,
                               virDomainDefPtr def)
{
    bool hasMultiBus = virQEMUCapsGet(qemuCaps, QEMU_CAPS_PCI_MULTIBUS);

    if (hasMultiBus)
        return true;

    if (def->os.arch == VIR_ARCH_PPC ||
2015
        ARCH_IS_PPC64(def->os.arch)) {
D
Daniel P. Berrange 已提交
2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
        /*
         * Usage of pci.0 naming:
         *
         *    ref405ep: no pci
         *       taihu: no pci
         *      bamboo: 1.1.0
         *       mac99: 2.0.0
         *     g3beige: 2.0.0
         *        prep: 1.4.0
         *     pseries: 2.0.0
         *   mpc8544ds: forever
         * virtex-m507: no pci
         *     ppce500: 1.6.0
         */

        if (qemuCaps->version >= 2000000)
            return true;

        if (qemuCaps->version >= 1006000 &&
            STREQ(def->os.machine, "ppce500"))
            return true;

        if (qemuCaps->version >= 1004000 &&
            STREQ(def->os.machine, "prep"))
            return true;

        if (qemuCaps->version >= 1001000 &&
            STREQ(def->os.machine, "bamboo"))
            return true;

        if (STREQ(def->os.machine, "mpc8544ds"))
            return true;

        return false;
    }

    return false;
}


2056
const char *virQEMUCapsGetBinary(virQEMUCapsPtr qemuCaps)
2057
{
2058
    return qemuCaps->binary;
2059 2060
}

2061
virArch virQEMUCapsGetArch(virQEMUCapsPtr qemuCaps)
2062
{
2063
    return qemuCaps->arch;
2064 2065 2066
}


2067
unsigned int virQEMUCapsGetVersion(virQEMUCapsPtr qemuCaps)
2068
{
2069
    return qemuCaps->version;
2070 2071 2072
}


2073
unsigned int virQEMUCapsGetKVMVersion(virQEMUCapsPtr qemuCaps)
2074
{
2075
    return qemuCaps->kvmVersion;
2076 2077 2078
}


2079 2080
int virQEMUCapsAddCPUDefinition(virQEMUCapsPtr qemuCaps,
                                const char *name)
2081
{
2082 2083 2084
    char *tmp;

    if (VIR_STRDUP(tmp, name) < 0)
2085
        return -1;
2086
    if (VIR_EXPAND_N(qemuCaps->cpuDefinitions, qemuCaps->ncpuDefinitions, 1) < 0) {
2087 2088 2089
        VIR_FREE(tmp);
        return -1;
    }
2090
    qemuCaps->cpuDefinitions[qemuCaps->ncpuDefinitions-1] = tmp;
2091 2092 2093 2094
    return 0;
}


2095 2096
size_t virQEMUCapsGetCPUDefinitions(virQEMUCapsPtr qemuCaps,
                                    char ***names)
2097
{
2098
    if (names)
2099 2100
        *names = qemuCaps->cpuDefinitions;
    return qemuCaps->ncpuDefinitions;
2101 2102 2103
}


2104 2105
size_t virQEMUCapsGetMachineTypes(virQEMUCapsPtr qemuCaps,
                                  char ***names)
2106
{
2107
    if (names)
2108 2109
        *names = qemuCaps->machineTypes;
    return qemuCaps->nmachineTypes;
2110 2111
}

2112 2113 2114
int virQEMUCapsGetMachineTypesCaps(virQEMUCapsPtr qemuCaps,
                                   size_t *nmachines,
                                   virCapsGuestMachinePtr **machines)
2115 2116 2117 2118
{
    size_t i;

    *machines = NULL;
2119
    *nmachines = qemuCaps->nmachineTypes;
2120

2121 2122 2123 2124
    if (*nmachines &&
        VIR_ALLOC_N(*machines, qemuCaps->nmachineTypes) < 0)
        goto error;

2125
    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
2126 2127
        virCapsGuestMachinePtr mach;
        if (VIR_ALLOC(mach) < 0)
2128
            goto error;
2129
        (*machines)[i] = mach;
2130
        if (qemuCaps->machineAliases[i]) {
2131 2132 2133
            if (VIR_STRDUP(mach->name, qemuCaps->machineAliases[i]) < 0 ||
                VIR_STRDUP(mach->canonical, qemuCaps->machineTypes[i]) < 0)
                goto error;
2134
        } else {
2135 2136
            if (VIR_STRDUP(mach->name, qemuCaps->machineTypes[i]) < 0)
                goto error;
2137
        }
2138
        mach->maxCpus = qemuCaps->machineMaxCpus[i];
2139 2140 2141 2142
    }

    return 0;

2143
 error:
2144 2145 2146 2147 2148 2149 2150 2151
    virCapabilitiesFreeMachines(*machines, *nmachines);
    *nmachines = 0;
    *machines = NULL;
    return -1;
}



2152

2153 2154
const char *virQEMUCapsGetCanonicalMachine(virQEMUCapsPtr qemuCaps,
                                           const char *name)
2155 2156 2157
{
    size_t i;

2158 2159 2160
    if (!name)
        return NULL;

2161
    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
2162
        if (!qemuCaps->machineAliases[i])
2163
            continue;
2164 2165
        if (STREQ(qemuCaps->machineAliases[i], name))
            return qemuCaps->machineTypes[i];
2166 2167 2168 2169
    }

    return name;
}
2170 2171


2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190
int virQEMUCapsGetMachineMaxCpus(virQEMUCapsPtr qemuCaps,
                                 const char *name)
{
    size_t i;

    if (!name)
        return 0;

    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
        if (!qemuCaps->machineMaxCpus[i])
            continue;
        if (STREQ(qemuCaps->machineTypes[i], name))
            return qemuCaps->machineMaxCpus[i];
    }

    return 0;
}


2191
static int
2192 2193
virQEMUCapsProbeQMPCommands(virQEMUCapsPtr qemuCaps,
                            qemuMonitorPtr mon)
2194 2195 2196 2197 2198 2199 2200
{
    char **commands = NULL;
    int ncommands;

    if ((ncommands = qemuMonitorGetCommands(mon, &commands)) < 0)
        return -1;

2201 2202 2203 2204 2205
    virQEMUCapsProcessStringFlags(qemuCaps,
                                  ARRAY_CARDINALITY(virQEMUCapsCommands),
                                  virQEMUCapsCommands,
                                  ncommands, commands);
    virQEMUCapsFreeStringList(ncommands, commands);
2206

2207 2208 2209 2210
    /* QMP add-fd was introduced in 1.2, but did not support
     * management control of set numbering, and did not have a
     * counterpart -add-fd command line option.  We require the
     * add-fd features from 1.3 or later.  */
2211
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_ADD_FD)) {
2212 2213 2214 2215 2216 2217 2218
        int fd = open("/dev/null", O_RDONLY);
        if (fd < 0) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("unable to probe for add-fd"));
            return -1;
        }
        if (qemuMonitorAddFd(mon, 0, fd, "/dev/null") < 0)
2219
            virQEMUCapsClear(qemuCaps, QEMU_CAPS_ADD_FD);
2220 2221 2222
        VIR_FORCE_CLOSE(fd);
    }

2223 2224 2225 2226 2227 2228
    /* Probe for active commit of qemu 2.1 (for now, we are choosing
     * to ignore the fact that qemu 2.0 can also do active commit) */
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_BLOCK_COMMIT) &&
        qemuMonitorSupportsActiveCommit(mon))
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_ACTIVE_COMMIT);

2229 2230 2231 2232 2233
    return 0;
}


static int
2234 2235
virQEMUCapsProbeQMPEvents(virQEMUCapsPtr qemuCaps,
                          qemuMonitorPtr mon)
2236 2237 2238 2239 2240 2241 2242
{
    char **events = NULL;
    int nevents;

    if ((nevents = qemuMonitorGetEvents(mon, &events)) < 0)
        return -1;

2243 2244 2245 2246 2247
    virQEMUCapsProcessStringFlags(qemuCaps,
                                  ARRAY_CARDINALITY(virQEMUCapsEvents),
                                  virQEMUCapsEvents,
                                  nevents, events);
    virQEMUCapsFreeStringList(nevents, events);
2248 2249 2250 2251 2252

    return 0;
}


2253
static int
2254 2255
virQEMUCapsProbeQMPObjects(virQEMUCapsPtr qemuCaps,
                           qemuMonitorPtr mon)
2256 2257 2258 2259 2260 2261 2262
{
    int nvalues;
    char **values;
    size_t i;

    if ((nvalues = qemuMonitorGetObjectTypes(mon, &values)) < 0)
        return -1;
2263 2264 2265 2266 2267 2268
    virQEMUCapsProcessStringFlags(qemuCaps,
                                  ARRAY_CARDINALITY(virQEMUCapsObjectTypes),
                                  virQEMUCapsObjectTypes,
                                  nvalues, values);
    virQEMUCapsFreeStringList(nvalues, values);

2269
    for (i = 0; i < ARRAY_CARDINALITY(virQEMUCapsObjectProps); i++) {
2270
        const char *type = virQEMUCapsObjectProps[i].type;
2271 2272 2273 2274
        if ((nvalues = qemuMonitorGetObjectProps(mon,
                                                 type,
                                                 &values)) < 0)
            return -1;
2275 2276 2277 2278 2279
        virQEMUCapsProcessStringFlags(qemuCaps,
                                      virQEMUCapsObjectProps[i].nprops,
                                      virQEMUCapsObjectProps[i].props,
                                      nvalues, values);
        virQEMUCapsFreeStringList(nvalues, values);
2280 2281 2282
    }

    /* Prefer -chardev spicevmc (detected earlier) over -device spicevmc */
2283 2284
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_CHARDEV_SPICEVMC))
        virQEMUCapsClear(qemuCaps, QEMU_CAPS_DEVICE_SPICEVMC);
2285
    /* If qemu supports newer -device qxl it supports -vga qxl as well */
2286 2287
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_QXL))
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA_QXL);
2288 2289 2290 2291 2292 2293

    return 0;
}


static int
2294 2295
virQEMUCapsProbeQMPMachineTypes(virQEMUCapsPtr qemuCaps,
                                qemuMonitorPtr mon)
2296 2297 2298 2299 2300
{
    qemuMonitorMachineInfoPtr *machines = NULL;
    int nmachines = 0;
    int ret = -1;
    size_t i;
2301
    size_t defIdx = 0;
2302 2303

    if ((nmachines = qemuMonitorGetMachines(mon, &machines)) < 0)
2304
        return -1;
2305

2306
    if (VIR_ALLOC_N(qemuCaps->machineTypes, nmachines) < 0)
2307
        goto cleanup;
2308
    if (VIR_ALLOC_N(qemuCaps->machineAliases, nmachines) < 0)
2309
        goto cleanup;
2310
    if (VIR_ALLOC_N(qemuCaps->machineMaxCpus, nmachines) < 0)
2311
        goto cleanup;
2312

2313
    for (i = 0; i < nmachines; i++) {
2314 2315
        if (STREQ(machines[i]->name, "none"))
            continue;
2316 2317 2318 2319 2320
        qemuCaps->nmachineTypes++;
        if (VIR_STRDUP(qemuCaps->machineAliases[qemuCaps->nmachineTypes -1],
                       machines[i]->alias) < 0 ||
            VIR_STRDUP(qemuCaps->machineTypes[qemuCaps->nmachineTypes - 1],
                       machines[i]->name) < 0)
2321
            goto cleanup;
2322
        if (machines[i]->isDefault)
2323
            defIdx = qemuCaps->nmachineTypes - 1;
2324 2325
        qemuCaps->machineMaxCpus[qemuCaps->nmachineTypes - 1] =
            machines[i]->maxCpus;
2326
    }
2327 2328

    if (defIdx)
2329
        virQEMUCapsSetDefaultMachine(qemuCaps, defIdx);
2330 2331 2332

    ret = 0;

2333
 cleanup:
2334
    for (i = 0; i < nmachines; i++)
2335 2336 2337 2338 2339 2340 2341
        qemuMonitorMachineInfoFree(machines[i]);
    VIR_FREE(machines);
    return ret;
}


static int
2342 2343
virQEMUCapsProbeQMPCPUDefinitions(virQEMUCapsPtr qemuCaps,
                                  qemuMonitorPtr mon)
2344 2345 2346 2347 2348 2349 2350
{
    int ncpuDefinitions;
    char **cpuDefinitions;

    if ((ncpuDefinitions = qemuMonitorGetCPUDefinitions(mon, &cpuDefinitions)) < 0)
        return -1;

2351 2352
    qemuCaps->ncpuDefinitions = ncpuDefinitions;
    qemuCaps->cpuDefinitions = cpuDefinitions;
2353 2354 2355 2356

    return 0;
}

2357 2358
struct tpmTypeToCaps {
    int type;
2359
    virQEMUCapsFlags caps;
2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379
};

static const struct tpmTypeToCaps virQEMUCapsTPMTypesToCaps[] = {
    {
        .type = VIR_DOMAIN_TPM_TYPE_PASSTHROUGH,
        .caps = QEMU_CAPS_DEVICE_TPM_PASSTHROUGH,
    },
};

const struct tpmTypeToCaps virQEMUCapsTPMModelsToCaps[] = {
    {
        .type = VIR_DOMAIN_TPM_MODEL_TIS,
        .caps = QEMU_CAPS_DEVICE_TPM_TIS,
    },
};

static int
virQEMUCapsProbeQMPTPM(virQEMUCapsPtr qemuCaps,
                       qemuMonitorPtr mon)
{
2380 2381
    int nentries;
    size_t i;
2382
    char **entries = NULL;
S
Stefan Berger 已提交
2383

2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413
    if ((nentries = qemuMonitorGetTPMModels(mon, &entries)) < 0)
        return -1;

    if (nentries > 0) {
        for (i = 0; i < ARRAY_CARDINALITY(virQEMUCapsTPMModelsToCaps); i++) {
            const char *needle = virDomainTPMModelTypeToString(
                virQEMUCapsTPMModelsToCaps[i].type);
            if (virStringArrayHasString(entries, needle))
                virQEMUCapsSet(qemuCaps,
                               virQEMUCapsTPMModelsToCaps[i].caps);
        }
    }
    virStringFreeList(entries);

    if ((nentries = qemuMonitorGetTPMTypes(mon, &entries)) < 0)
        return -1;

    if (nentries > 0) {
        for (i = 0; i < ARRAY_CARDINALITY(virQEMUCapsTPMTypesToCaps); i++) {
            const char *needle = virDomainTPMBackendTypeToString(
                virQEMUCapsTPMTypesToCaps[i].type);
            if (virStringArrayHasString(entries, needle))
                virQEMUCapsSet(qemuCaps, virQEMUCapsTPMTypesToCaps[i].caps);
        }
    }
    virStringFreeList(entries);

    return 0;
}

2414

2415
static int
2416 2417
virQEMUCapsProbeQMPKVMState(virQEMUCapsPtr qemuCaps,
                            qemuMonitorPtr mon)
2418 2419 2420 2421
{
    bool enabled = false;
    bool present = false;

2422
    if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_KVM))
2423 2424 2425 2426 2427 2428
        return 0;

    if (qemuMonitorGetKVMState(mon, &enabled, &present) < 0)
        return -1;

    /* The QEMU_CAPS_KVM flag was initially set according to the QEMU
2429
     * reporting the recognition of 'query-kvm' QMP command. That merely
N
Nehal J Wani 已提交
2430
     * indicates existence of the command though, not whether KVM support
2431 2432 2433 2434 2435 2436
     * is actually available, nor whether it is enabled by default.
     *
     * If it is not present we need to clear the flag, and if it is
     * not enabled by default we need to change the flag.
     */
    if (!present) {
2437
        virQEMUCapsClear(qemuCaps, QEMU_CAPS_KVM);
2438
    } else if (!enabled) {
2439 2440
        virQEMUCapsClear(qemuCaps, QEMU_CAPS_KVM);
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_ENABLE_KVM);
2441 2442 2443 2444 2445
    }

    return 0;
}

2446 2447 2448 2449 2450 2451 2452 2453
struct virQEMUCapsCommandLineProps {
    const char *option;
    const char *param;
    int flag;
};

static struct virQEMUCapsCommandLineProps virQEMUCapsCommandLine[] = {
    { "machine", "mem-merge", QEMU_CAPS_MEM_MERGE },
O
Osier Yang 已提交
2454
    { "drive", "discard", QEMU_CAPS_DRIVE_DISCARD },
2455
    { "realtime", "mlock", QEMU_CAPS_MLOCK },
2456
    { "boot-opts", "strict", QEMU_CAPS_BOOT_STRICT },
2457
    { "boot-opts", "reboot-timeout", QEMU_CAPS_REBOOT_TIMEOUT },
2458
    { "boot-opts", "splash-time", QEMU_CAPS_SPLASH_TIMEOUT },
2459
    { "spice", "disable-agent-file-xfer", QEMU_CAPS_SPICE_FILE_XFER_DISABLE },
2460
    { "msg", "timestamp", QEMU_CAPS_MSG_TIMESTAMP },
2461
    { "numa", NULL, QEMU_CAPS_NUMA },
2462
    { "drive", "throttling.bps-total-max", QEMU_CAPS_DRIVE_IOTUNE_MAX},
2463 2464 2465 2466 2467 2468
};

static int
virQEMUCapsProbeQMPCommandLine(virQEMUCapsPtr qemuCaps,
                               qemuMonitorPtr mon)
{
2469
    bool found = false;
2470 2471 2472 2473 2474 2475 2476
    int nvalues;
    char **values;
    size_t i, j;

    for (i = 0; i < ARRAY_CARDINALITY(virQEMUCapsCommandLine); i++) {
        if ((nvalues = qemuMonitorGetCommandLineOptionParameters(mon,
                                                                 virQEMUCapsCommandLine[i].option,
2477 2478
                                                                 &values,
                                                                 &found)) < 0)
2479
            return -1;
2480 2481 2482 2483

        if (found && !virQEMUCapsCommandLine[i].param)
            virQEMUCapsSet(qemuCaps, virQEMUCapsCommandLine[i].flag);

2484
        for (j = 0; j < nvalues; j++) {
2485
            if (STREQ_NULLABLE(virQEMUCapsCommandLine[i].param, values[j])) {
2486 2487 2488 2489 2490 2491 2492 2493 2494
                virQEMUCapsSet(qemuCaps, virQEMUCapsCommandLine[i].flag);
                break;
            }
        }
        virStringFreeList(values);
    }

    return 0;
}
2495

2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514
static int
virQEMUCapsProbeQMPMigrationCapabilities(virQEMUCapsPtr qemuCaps,
                                         qemuMonitorPtr mon)
{
    char **caps = NULL;
    int ncaps;

    if ((ncaps = qemuMonitorGetMigrationCapabilities(mon, &caps)) < 0)
        return -1;

    virQEMUCapsProcessStringFlags(qemuCaps,
                                  ARRAY_CARDINALITY(virQEMUCapsMigration),
                                  virQEMUCapsMigration,
                                  ncaps, caps);
    virQEMUCapsFreeStringList(ncaps, caps);

    return 0;
}

2515 2516
int virQEMUCapsProbeQMP(virQEMUCapsPtr qemuCaps,
                        qemuMonitorPtr mon)
2517
{
2518
    VIR_DEBUG("qemuCaps=%p mon=%p", qemuCaps, mon);
2519

2520
    if (qemuCaps->usedQMP)
2521 2522
        return 0;

2523
    if (virQEMUCapsProbeQMPCommands(qemuCaps, mon) < 0)
2524 2525
        return -1;

2526
    if (virQEMUCapsProbeQMPEvents(qemuCaps, mon) < 0)
2527 2528 2529 2530 2531 2532
        return -1;

    return 0;
}


2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558
/*
 * Parsing a doc that looks like
 *
 * <qemuCaps>
 *   <qemuctime>234235253</qemuctime>
 *   <selfctime>234235253</selfctime>
 *   <usedQMP/>
 *   <flag name='foo'/>
 *   <flag name='bar'/>
 *   ...
 *   <cpu name="pentium3"/>
 *   ...
 *   <machine name="pc-1.0" alias="pc" maxCpus="4"/>
 *   ...
 * </qemuCaps>
 */
static int
virQEMUCapsLoadCache(virQEMUCapsPtr qemuCaps, const char *filename,
                     time_t *qemuctime, time_t *selfctime)
{
    xmlDocPtr doc = NULL;
    int ret = -1;
    size_t i;
    int n;
    xmlNodePtr *nodes = NULL;
    xmlXPathContextPtr ctxt = NULL;
J
Ján Tomko 已提交
2559
    char *str = NULL;
2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644
    long long int l;

    if (!(doc = virXMLParseFile(filename)))
        goto cleanup;

    if (!(ctxt = xmlXPathNewContext(doc))) {
        virReportOOMError();
        goto cleanup;
    }

    ctxt->node = xmlDocGetRootElement(doc);

    if (STRNEQ((const char *)ctxt->node->name, "qemuCaps")) {
        virReportError(VIR_ERR_XML_ERROR,
                       _("unexpected root element <%s>, "
                         "expecting <qemuCaps>"),
                       ctxt->node->name);
        goto cleanup;
    }

    if (virXPathLongLong("string(./qemuctime)", ctxt, &l) < 0) {
        virReportError(VIR_ERR_XML_ERROR, "%s",
                       _("missing qemuctime in QEMU capabilities XML"));
        goto cleanup;
    }
    *qemuctime = (time_t)l;

    if (virXPathLongLong("string(./selfctime)", ctxt, &l) < 0) {
        virReportError(VIR_ERR_XML_ERROR, "%s",
                       _("missing selfctime in QEMU capabilities XML"));
        goto cleanup;
    }
    *selfctime = (time_t)l;

    qemuCaps->usedQMP = virXPathBoolean("count(./usedQMP) > 0",
                                        ctxt) > 0;

    if ((n = virXPathNodeSet("./flag", ctxt, &nodes)) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("failed to parse qemu capabilities flags"));
        goto cleanup;
    }
    VIR_DEBUG("Got flags %d", n);
    if (n > 0) {
        for (i = 0; i < n; i++) {
            int flag;
            if (!(str = virXMLPropString(nodes[i], "name"))) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("missing flag name in QEMU capabilities cache"));
                goto cleanup;
            }
            flag = virQEMUCapsTypeFromString(str);
            if (flag < 0) {
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unknown qemu capabilities flag %s"), str);
                goto cleanup;
            }
            VIR_FREE(str);
            virQEMUCapsSet(qemuCaps, flag);
        }
    }
    VIR_FREE(nodes);

    if (virXPathUInt("string(./version)", ctxt, &qemuCaps->version) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("missing version in QEMU capabilities cache"));
        goto cleanup;
    }

    if (virXPathUInt("string(./kvmVersion)", ctxt, &qemuCaps->kvmVersion) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("missing version in QEMU capabilities cache"));
        goto cleanup;
    }

    if (!(str = virXPathString("string(./arch)", ctxt))) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("missing arch in QEMU capabilities cache"));
        goto cleanup;
    }
    if (!(qemuCaps->arch = virArchFromString(str))) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unknown arch %s in QEMU capabilities cache"), str);
        goto cleanup;
    }
J
Ján Tomko 已提交
2645
    VIR_FREE(str);
2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658

    if ((n = virXPathNodeSet("./cpu", ctxt, &nodes)) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("failed to parse qemu capabilities cpus"));
        goto cleanup;
    }
    if (n > 0) {
        qemuCaps->ncpuDefinitions = n;
        if (VIR_ALLOC_N(qemuCaps->cpuDefinitions,
                        qemuCaps->ncpuDefinitions) < 0)
            goto cleanup;

        for (i = 0; i < n; i++) {
J
Ján Tomko 已提交
2659
            if (!(qemuCaps->cpuDefinitions[i] = virXMLPropString(nodes[i], "name"))) {
2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("missing cpu name in QEMU capabilities cache"));
                goto cleanup;
            }
        }
    }
    VIR_FREE(nodes);


    if ((n = virXPathNodeSet("./machine", ctxt, &nodes)) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("failed to parse qemu capabilities machines"));
        goto cleanup;
    }
    if (n > 0) {
        qemuCaps->nmachineTypes = n;
        if (VIR_ALLOC_N(qemuCaps->machineTypes,
                        qemuCaps->nmachineTypes) < 0 ||
            VIR_ALLOC_N(qemuCaps->machineAliases,
                        qemuCaps->nmachineTypes) < 0 ||
            VIR_ALLOC_N(qemuCaps->machineMaxCpus,
                        qemuCaps->nmachineTypes) < 0)
            goto cleanup;

        for (i = 0; i < n; i++) {
J
Ján Tomko 已提交
2685
            if (!(qemuCaps->machineTypes[i] = virXMLPropString(nodes[i], "name"))) {
2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("missing machine name in QEMU capabilities cache"));
                goto cleanup;
            }
            qemuCaps->machineAliases[i] = virXMLPropString(nodes[i], "alias");

            str = virXMLPropString(nodes[i], "maxCpus");
            if (str &&
                virStrToLong_ui(str, NULL, 10, &(qemuCaps->machineMaxCpus[i])) < 0) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("malformed machine cpu count in QEMU capabilities cache"));
                goto cleanup;
            }
J
Ján Tomko 已提交
2699
            VIR_FREE(str);
2700 2701 2702 2703 2704
        }
    }
    VIR_FREE(nodes);

    ret = 0;
2705
 cleanup:
J
Ján Tomko 已提交
2706
    VIR_FREE(str);
2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717
    VIR_FREE(nodes);
    xmlXPathFreeContext(ctxt);
    xmlFreeDoc(doc);
    return ret;
}


static int
virQEMUCapsSaveCache(virQEMUCapsPtr qemuCaps, const char *filename)
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;
M
Michal Privoznik 已提交
2718
    char *xml = NULL;
2719 2720 2721 2722
    int ret = -1;
    size_t i;

    virBufferAddLit(&buf, "<qemuCaps>\n");
2723
    virBufferAdjustIndent(&buf, 2);
2724

2725
    virBufferAsprintf(&buf, "<qemuctime>%llu</qemuctime>\n",
2726
                      (long long)qemuCaps->ctime);
2727
    virBufferAsprintf(&buf, "<selfctime>%llu</selfctime>\n",
2728 2729 2730
                      (long long)virGetSelfLastChanged());

    if (qemuCaps->usedQMP)
2731
        virBufferAddLit(&buf, "<usedQMP/>\n");
2732 2733 2734

    for (i = 0; i < QEMU_CAPS_LAST; i++) {
        if (virQEMUCapsGet(qemuCaps, i)) {
2735
            virBufferAsprintf(&buf, "<flag name='%s'/>\n",
2736 2737 2738 2739
                              virQEMUCapsTypeToString(i));
        }
    }

2740
    virBufferAsprintf(&buf, "<version>%d</version>\n",
2741 2742
                      qemuCaps->version);

2743
    virBufferAsprintf(&buf, "<kvmVersion>%d</kvmVersion>\n",
2744 2745
                      qemuCaps->kvmVersion);

2746
    virBufferAsprintf(&buf, "<arch>%s</arch>\n",
2747 2748 2749
                      virArchToString(qemuCaps->arch));

    for (i = 0; i < qemuCaps->ncpuDefinitions; i++) {
2750
        virBufferEscapeString(&buf, "<cpu name='%s'/>\n",
2751 2752 2753 2754
                              qemuCaps->cpuDefinitions[i]);
    }

    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
2755
        virBufferEscapeString(&buf, "<machine name='%s'",
2756 2757 2758 2759 2760 2761 2762 2763
                              qemuCaps->machineTypes[i]);
        if (qemuCaps->machineAliases[i])
            virBufferEscapeString(&buf, " alias='%s'",
                              qemuCaps->machineAliases[i]);
        virBufferAsprintf(&buf, " maxCpus='%u'/>\n",
                          qemuCaps->machineMaxCpus[i]);
    }

2764
    virBufferAdjustIndent(&buf, -2);
2765 2766
    virBufferAddLit(&buf, "</qemuCaps>\n");

2767
    if (virBufferCheckError(&buf) < 0)
2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837
        goto cleanup;

    xml = virBufferContentAndReset(&buf);

    if (virFileWriteStr(filename, xml, 0600) < 0) {
        virReportSystemError(errno,
                             _("Failed to save '%s' for '%s'"),
                             filename, qemuCaps->binary);
        goto cleanup;
    }

    VIR_DEBUG("Saved caps '%s' for '%s' with (%lld, %lld)",
              filename, qemuCaps->binary,
              (long long)qemuCaps->ctime,
              (long long)virGetSelfLastChanged());

    ret = 0;
 cleanup:
    VIR_FREE(xml);
    return ret;
}

static int
virQEMUCapsRememberCached(virQEMUCapsPtr qemuCaps, const char *cacheDir)
{
    char *capsdir = NULL;
    char *capsfile = NULL;
    int ret = -1;
    char *binaryhash = NULL;

    if (virAsprintf(&capsdir, "%s/capabilities", cacheDir) < 0)
        goto cleanup;

    if (virCryptoHashString(VIR_CRYPTO_HASH_SHA256,
                            qemuCaps->binary,
                            &binaryhash) < 0)
        goto cleanup;

    if (virAsprintf(&capsfile, "%s/%s.xml", capsdir, binaryhash) < 0)
        goto cleanup;

    if (virFileMakePath(capsdir) < 0) {
        virReportSystemError(errno,
                             _("Unable to create directory '%s'"),
                             capsdir);
        goto cleanup;
    }

    if (virQEMUCapsSaveCache(qemuCaps, capsfile) < 0)
        goto cleanup;

    ret = 0;
 cleanup:
    VIR_FREE(binaryhash);
    VIR_FREE(capsfile);
    VIR_FREE(capsdir);
    return ret;
}


static void
virQEMUCapsReset(virQEMUCapsPtr qemuCaps)
{
    size_t i;

    virBitmapClearAll(qemuCaps->flags);
    qemuCaps->version = qemuCaps->kvmVersion = 0;
    qemuCaps->arch = VIR_ARCH_NONE;
    qemuCaps->usedQMP = false;

2838
    for (i = 0; i < qemuCaps->ncpuDefinitions; i++)
2839 2840 2841 2842 2843 2844 2845 2846 2847 2848
        VIR_FREE(qemuCaps->cpuDefinitions[i]);
    VIR_FREE(qemuCaps->cpuDefinitions);
    qemuCaps->ncpuDefinitions = 0;

    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
        VIR_FREE(qemuCaps->machineTypes[i]);
        VIR_FREE(qemuCaps->machineAliases[i]);
    }
    VIR_FREE(qemuCaps->machineTypes);
    VIR_FREE(qemuCaps->machineAliases);
J
Ján Tomko 已提交
2849
    VIR_FREE(qemuCaps->machineMaxCpus);
2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933
    qemuCaps->nmachineTypes = 0;
}


static int
virQEMUCapsInitCached(virQEMUCapsPtr qemuCaps, const char *cacheDir)
{
    char *capsdir = NULL;
    char *capsfile = NULL;
    int ret = -1;
    char *binaryhash = NULL;
    struct stat sb;
    time_t qemuctime;
    time_t selfctime;

    if (virAsprintf(&capsdir, "%s/capabilities", cacheDir) < 0)
        goto cleanup;

    if (virCryptoHashString(VIR_CRYPTO_HASH_SHA256,
                            qemuCaps->binary,
                            &binaryhash) < 0)
        goto cleanup;

    if (virAsprintf(&capsfile, "%s/%s.xml", capsdir, binaryhash) < 0)
        goto cleanup;

    if (virFileMakePath(capsdir) < 0) {
        virReportSystemError(errno,
                             _("Unable to create directory '%s'"),
                             capsdir);
        goto cleanup;
    }

    if (stat(capsfile, &sb) < 0) {
        if (errno == ENOENT) {
            VIR_DEBUG("No cached capabilities '%s' for '%s'",
                      capsfile, qemuCaps->binary);
            ret = 0;
            goto cleanup;
        }
        virReportSystemError(errno,
                             _("Unable to access cache '%s' for '%s'"),
                             capsfile, qemuCaps->binary);
        goto cleanup;
    }

    if (virQEMUCapsLoadCache(qemuCaps, capsfile, &qemuctime, &selfctime) < 0) {
        virErrorPtr err = virGetLastError();
        VIR_WARN("Failed to load cached caps from '%s' for '%s': %s",
                 capsfile, qemuCaps->binary, err ? NULLSTR(err->message) :
                 _("unknown error"));
        virResetLastError();
        ret = 0;
        virQEMUCapsReset(qemuCaps);
        goto cleanup;
    }

    /* Discard if cache is older that QEMU binary */
    if (qemuctime != qemuCaps->ctime ||
        selfctime < virGetSelfLastChanged()) {
        VIR_DEBUG("Outdated cached capabilities '%s' for '%s' "
                  "(%lld vs %lld, %lld vs %lld)",
                  capsfile, qemuCaps->binary,
                  (long long)qemuctime, (long long)qemuCaps->ctime,
                  (long long)selfctime, (long long)virGetSelfLastChanged());
        ignore_value(unlink(capsfile));
        virQEMUCapsReset(qemuCaps);
        ret = 0;
        goto cleanup;
    }

    VIR_DEBUG("Loaded '%s' for '%s' ctime %lld usedQMP=%d",
              capsfile, qemuCaps->binary,
              (long long)qemuCaps->ctime, qemuCaps->usedQMP);

    ret = 1;
 cleanup:
    VIR_FREE(binaryhash);
    VIR_FREE(capsfile);
    VIR_FREE(capsdir);
    return ret;
}


2934 2935
#define QEMU_SYSTEM_PREFIX "qemu-system-"

2936
static int
2937
virQEMUCapsInitHelp(virQEMUCapsPtr qemuCaps, uid_t runUid, gid_t runGid)
2938
{
2939
    virCommandPtr cmd = NULL;
2940
    bool is_kvm;
2941
    char *help = NULL;
2942 2943
    int ret = -1;
    const char *tmp;
2944

2945
    VIR_DEBUG("qemuCaps=%p", qemuCaps);
2946

2947
    tmp = strstr(qemuCaps->binary, QEMU_SYSTEM_PREFIX);
2948 2949
    if (tmp) {
        tmp += strlen(QEMU_SYSTEM_PREFIX);
2950

2951
        qemuCaps->arch = virQEMUCapsArchFromString(tmp);
2952
    } else {
2953
        qemuCaps->arch = virArchFromHost();
2954 2955
    }

2956
    cmd = virQEMUCapsProbeCommand(qemuCaps->binary, NULL, runUid, runGid);
2957 2958 2959 2960
    virCommandAddArgList(cmd, "-help", NULL);
    virCommandSetOutputBuffer(cmd, &help);

    if (virCommandRun(cmd, NULL) < 0)
2961
        goto cleanup;
2962

2963 2964 2965 2966 2967 2968
    if (virQEMUCapsParseHelpStr(qemuCaps->binary,
                                help, qemuCaps,
                                &qemuCaps->version,
                                &is_kvm,
                                &qemuCaps->kvmVersion,
                                false) < 0)
2969
        goto cleanup;
2970

D
Daniel P. Berrange 已提交
2971 2972 2973 2974 2975 2976 2977
    /* x86_64 and i686 support PCI-multibus on all machine types
     * since forever. For other architectures, it has been changing
     * across releases, per machine type, so we can't simply detect
     * it here. Thus the rest of the logic is provided in a separate
     * helper virQEMUCapsHasPCIMultiBus() which keys off the machine
     * stored in virDomainDef and QEMU version number
     */
2978
    if (qemuCaps->arch == VIR_ARCH_X86_64 ||
D
Daniel P. Berrange 已提交
2979
        qemuCaps->arch == VIR_ARCH_I686)
2980
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_PCI_MULTIBUS);
D
Daniel P. Berrange 已提交
2981 2982 2983 2984 2985

    /* -no-acpi is not supported on non-x86
     * even if qemu reports it in -help */
    if (qemuCaps->arch != VIR_ARCH_X86_64 &&
        qemuCaps->arch != VIR_ARCH_I686)
2986
        virQEMUCapsClear(qemuCaps, QEMU_CAPS_NO_ACPI);
2987

2988
    /* virQEMUCapsExtractDeviceStr will only set additional caps if qemu
2989
     * understands the 0.13.0+ notion of "-device driver,".  */
2990
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE) &&
2991
        strstr(help, "-device driver,?") &&
2992 2993
        virQEMUCapsExtractDeviceStr(qemuCaps->binary,
                                    qemuCaps, runUid, runGid) < 0) {
2994
        goto cleanup;
2995
    }
2996

2997
    if (virQEMUCapsProbeCPUModels(qemuCaps, runUid, runGid) < 0)
2998
        goto cleanup;
2999

3000
    if (virQEMUCapsProbeMachineTypes(qemuCaps, runUid, runGid) < 0)
3001
        goto cleanup;
3002

3003
    ret = 0;
3004
 cleanup:
3005
    virCommandFree(cmd);
3006
    VIR_FREE(help);
3007 3008 3009 3010
    return ret;
}


3011
static void virQEMUCapsMonitorNotify(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
3012 3013
                                     virDomainObjPtr vm ATTRIBUTE_UNUSED,
                                     void *opaque ATTRIBUTE_UNUSED)
3014 3015 3016 3017
{
}

static qemuMonitorCallbacks callbacks = {
3018 3019
    .eofNotify = virQEMUCapsMonitorNotify,
    .errorNotify = virQEMUCapsMonitorNotify,
3020 3021 3022 3023 3024 3025 3026
};


/* Capabilities that we assume are always enabled
 * for QEMU >= 1.2.0
 */
static void
3027
virQEMUCapsInitQMPBasic(virQEMUCapsPtr qemuCaps)
3028
{
3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_COLON);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_REBOOT);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NAME);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_UUID);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNET_HDR);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_TCP);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_EXEC);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_V2);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_FORMAT);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_0_10);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MEM_PATH);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_SERIAL);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_CHARDEV);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MONITOR_JSON);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_BALLOON);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_SDL);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_SMP_TOPOLOGY);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_RTC);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VHOST_NET);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NODEFCONFIG);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_BOOT_MENU);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NAME_PROCESS);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_READONLY);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_SMBIOS_TYPE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA_NONE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_AIO);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_CHARDEV_SPICEVMC);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE_QXL_VGA);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_DIRECTSYNC);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_SHUTDOWN);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_UNSAFE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV_READONLY);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VIRTIO_BLK_SG_IO);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_COPY_ON_READ);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_CPU_HOST);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV_WRITEOUT);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_IOTUNE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_WAKEUP);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_USER_CONFIG);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV_BRIDGE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_SECCOMP_SANDBOX);
O
Olivia Yin 已提交
3077
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DTB);
J
Ján Tomko 已提交
3078
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_IPV6_MIGRATION);
3079 3080
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MACHINE_OPT);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DUMP_GUEST_CORE);
3081
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_SHARE_POLICY);
3082
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_HOST_PCI_MULTIDOMAIN);
3083 3084
}

3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103
/* Capabilities that are architecture depending
 * initialized for QEMU.
 */
static int
virQEMUCapsInitArchQMPBasic(virQEMUCapsPtr qemuCaps,
                            qemuMonitorPtr mon)
{
    char *archstr = NULL;
    int ret = -1;

    if (!(archstr = qemuMonitorGetTargetArch(mon)))
        return -1;

    if ((qemuCaps->arch = virQEMUCapsArchFromString(archstr)) == VIR_ARCH_NONE) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Unknown QEMU arch %s"), archstr);
        goto cleanup;
    }

D
Daniel P. Berrange 已提交
3104 3105 3106 3107 3108 3109 3110 3111
    /* x86_64 and i686 support PCI-multibus on all machine types
     * since forever. For other architectures, it has been changing
     * across releases, per machine type, so we can't simply detect
     * it here. Thus the rest of the logic is provided in a separate
     * helper virQEMUCapsHasPCIMultiBus() which keys off the machine
     * stored in virDomainDef and QEMU version number
     *
     * ACPI/HPET/KVM PIT are also x86 specific
3112 3113 3114 3115 3116
     */
    if (qemuCaps->arch == VIR_ARCH_X86_64 ||
        qemuCaps->arch == VIR_ARCH_I686) {
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_PCI_MULTIBUS);
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_ACPI);
J
Ján Tomko 已提交
3117
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_HPET);
3118 3119 3120 3121 3122
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_KVM_PIT);
    }

    ret = 0;

3123
 cleanup:
3124 3125 3126
    VIR_FREE(archstr);
    return ret;
}
3127

3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180
int
virQEMUCapsInitQMPMonitor(virQEMUCapsPtr qemuCaps,
                          qemuMonitorPtr mon)
{
    int ret = -1;
    int major, minor, micro;
    char *package = NULL;

    /* @mon is supposed to be locked by callee */

    if (qemuMonitorSetCapabilities(mon) < 0) {
        virErrorPtr err = virGetLastError();
        VIR_DEBUG("Failed to set monitor capabilities %s",
                  err ? err->message : "<unknown problem>");
        ret = 0;
        goto cleanup;
    }

    if (qemuMonitorGetVersion(mon,
                              &major, &minor, &micro,
                              &package) < 0) {
        virErrorPtr err = virGetLastError();
        VIR_DEBUG("Failed to query monitor version %s",
                  err ? err->message : "<unknown problem>");
        ret = 0;
        goto cleanup;
    }

    VIR_DEBUG("Got version %d.%d.%d (%s)",
              major, minor, micro, NULLSTR(package));

    if (major < 1 || (major == 1 && minor < 2)) {
        VIR_DEBUG("Not new enough for QMP capabilities detection");
        ret = 0;
        goto cleanup;
    }

    qemuCaps->version = major * 1000000 + minor * 1000 + micro;
    qemuCaps->usedQMP = true;

    virQEMUCapsInitQMPBasic(qemuCaps);

    if (virQEMUCapsInitArchQMPBasic(qemuCaps, mon) < 0)
        goto cleanup;

    /* USB option is supported v1.3.0 onwards */
    if (qemuCaps->version >= 1003000)
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MACHINE_USB_OPT);

    /* WebSockets were introduced between 1.3.0 and 1.3.1 */
    if (qemuCaps->version >= 1003001)
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_WEBSOCKET);

3181 3182 3183 3184 3185 3186
    /* -chardev spiceport is supported from 1.4.0, but usable through
     * qapi only since 1.5.0, however, it still cannot be queried
     * for as a capability */
    if (qemuCaps->version >= 1005000)
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_CHARDEV_SPICEPORT);

3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205
    if (qemuCaps->version >= 1006000)
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE_VIDEO_PRIMARY);

    if (virQEMUCapsProbeQMPCommands(qemuCaps, mon) < 0)
        goto cleanup;
    if (virQEMUCapsProbeQMPEvents(qemuCaps, mon) < 0)
        goto cleanup;
    if (virQEMUCapsProbeQMPObjects(qemuCaps, mon) < 0)
        goto cleanup;
    if (virQEMUCapsProbeQMPMachineTypes(qemuCaps, mon) < 0)
        goto cleanup;
    if (virQEMUCapsProbeQMPCPUDefinitions(qemuCaps, mon) < 0)
        goto cleanup;
    if (virQEMUCapsProbeQMPKVMState(qemuCaps, mon) < 0)
        goto cleanup;
    if (virQEMUCapsProbeQMPTPM(qemuCaps, mon) < 0)
        goto cleanup;
    if (virQEMUCapsProbeQMPCommandLine(qemuCaps, mon) < 0)
        goto cleanup;
3206 3207
    if (virQEMUCapsProbeQMPMigrationCapabilities(qemuCaps, mon) < 0)
        goto cleanup;
3208 3209

    ret = 0;
3210
 cleanup:
3211 3212 3213 3214
    VIR_FREE(package);
    return ret;
}

3215
static int
3216 3217 3218 3219
virQEMUCapsInitQMP(virQEMUCapsPtr qemuCaps,
                   const char *libDir,
                   uid_t runUid,
                   gid_t runGid)
3220 3221 3222 3223 3224 3225 3226 3227
{
    int ret = -1;
    virCommandPtr cmd = NULL;
    qemuMonitorPtr mon = NULL;
    int status = 0;
    virDomainChrSourceDef config;
    char *monarg = NULL;
    char *monpath = NULL;
3228
    char *pidfile = NULL;
3229
    pid_t pid = 0;
3230 3231
    virDomainObjPtr vm = NULL;
    virDomainXMLOptionPtr xmlopt = NULL;
3232

3233 3234 3235
    /* the ".sock" sufix is important to avoid a possible clash with a qemu
     * domain called "capabilities"
     */
3236
    if (virAsprintf(&monpath, "%s/%s", libDir, "capabilities.monitor.sock") < 0)
3237
        goto cleanup;
3238
    if (virAsprintf(&monarg, "unix:%s,server,nowait", monpath) < 0)
3239 3240
        goto cleanup;

3241 3242
    /* ".pidfile" suffix is used rather than ".pid" to avoid a possible clash
     * with a qemu domain called "capabilities"
3243 3244 3245
     * Normally we'd use runDir for pid files, but because we're using
     * -daemonize we need QEMU to be allowed to create them, rather
     * than libvirtd. So we're using libDir which QEMU can write to
3246
     */
3247
    if (virAsprintf(&pidfile, "%s/%s", libDir, "capabilities.pidfile") < 0)
3248 3249
        goto cleanup;

3250 3251 3252 3253 3254
    memset(&config, 0, sizeof(config));
    config.type = VIR_DOMAIN_CHR_TYPE_UNIX;
    config.data.nix.path = monpath;
    config.data.nix.listen = false;

3255 3256
    virPidFileForceCleanupPath(pidfile);

3257
    VIR_DEBUG("Try to get caps via QMP qemuCaps=%p", qemuCaps);
3258

3259 3260 3261 3262 3263 3264 3265
    /*
     * We explicitly need to use -daemonize here, rather than
     * virCommandDaemonize, because we need to synchronize
     * with QEMU creating its monitor socket API. Using
     * daemonize guarantees control won't return to libvirt
     * until the socket is present.
     */
3266
    cmd = virCommandNewArgList(qemuCaps->binary,
3267 3268 3269 3270 3271 3272
                               "-S",
                               "-no-user-config",
                               "-nodefaults",
                               "-nographic",
                               "-M", "none",
                               "-qmp", monarg,
3273 3274
                               "-pidfile", pidfile,
                               "-daemonize",
3275 3276 3277
                               NULL);
    virCommandAddEnvPassCommon(cmd);
    virCommandClearCaps(cmd);
3278 3279
    virCommandSetGID(cmd, runGid);
    virCommandSetUID(cmd, runUid);
3280

3281
    /* Log, but otherwise ignore, non-zero status.  */
3282 3283 3284 3285 3286
    if (virCommandRun(cmd, &status) < 0)
        goto cleanup;

    if (status != 0) {
        ret = 0;
3287
        VIR_DEBUG("QEMU %s exited with status %d", qemuCaps->binary, status);
3288 3289 3290
        goto cleanup;
    }

3291 3292 3293 3294 3295 3296
    if (virPidFileReadPath(pidfile, &pid) < 0) {
        VIR_DEBUG("Failed to read pidfile %s", pidfile);
        ret = 0;
        goto cleanup;
    }

3297 3298 3299 3300 3301
    if (!(xmlopt = virDomainXMLOptionNew(NULL, NULL, NULL)) ||
        !(vm = virDomainObjNew(xmlopt)))
        goto cleanup;

    vm->pid = pid;
3302

3303
    if (!(mon = qemuMonitorOpen(vm, &config, true, &callbacks, NULL))) {
3304
        ret = 0;
3305
        goto cleanup;
3306
    }
3307

3308
    virObjectLock(mon);
3309

3310
    if (virQEMUCapsInitQMPMonitor(qemuCaps, mon) < 0)
3311
        goto cleanup;
3312 3313 3314

    ret = 0;

3315
 cleanup:
3316
    if (mon)
3317
        virObjectUnlock(mon);
3318 3319
    qemuMonitorClose(mon);
    virCommandAbort(cmd);
3320
    virCommandFree(cmd);
3321
    VIR_FREE(monarg);
3322 3323
    if (monpath)
        ignore_value(unlink(monpath));
3324
    VIR_FREE(monpath);
3325 3326
    virObjectUnref(vm);
    virObjectUnref(xmlopt);
3327

3328
    if (pid != 0) {
3329 3330
        char ebuf[1024];

3331 3332 3333 3334 3335 3336 3337
        VIR_DEBUG("Killing QMP caps process %lld", (long long) pid);
        if (virProcessKill(pid, SIGKILL) < 0 && errno != ESRCH)
            VIR_ERROR(_("Failed to kill process %lld: %s"),
                      (long long) pid,
                      virStrerror(errno, ebuf, sizeof(ebuf)));
    }
    if (pidfile) {
3338 3339 3340
        unlink(pidfile);
        VIR_FREE(pidfile);
    }
3341 3342 3343 3344
    return ret;
}


3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356
#define MESSAGE_ID_CAPS_PROBE_FAILURE "8ae2f3fb-2dbe-498e-8fbd-012d40afa361"

static void
virQEMUCapsLogProbeFailure(const char *binary)
{
    virLogMetadata meta[] = {
        { .key = "MESSAGE_ID", .s = MESSAGE_ID_CAPS_PROBE_FAILURE, .iv = 0 },
        { .key = "LIBVIRT_QEMU_BINARY", .s = binary, .iv = 0 },
        { .key = NULL },
    };
    virErrorPtr err = virGetLastError();

3357
    virLogMessage(&virLogSelf,
3358 3359 3360 3361 3362 3363 3364 3365 3366
                  VIR_LOG_WARN,
                  __FILE__, __LINE__, __func__,
                  meta,
                  _("Failed to probe capabilities for %s: %s"),
                  binary, err && err->message ? err->message :
                  _("unknown failure"));
}


3367 3368
virQEMUCapsPtr virQEMUCapsNewForBinary(const char *binary,
                                       const char *libDir,
3369
                                       const char *cacheDir,
3370 3371
                                       uid_t runUid,
                                       gid_t runGid)
3372
{
3373
    virQEMUCapsPtr qemuCaps;
3374 3375 3376
    struct stat sb;
    int rv;

3377 3378 3379
    if (!(qemuCaps = virQEMUCapsNew()))
        goto error;

3380 3381
    if (VIR_STRDUP(qemuCaps->binary, binary) < 0)
        goto error;
3382 3383 3384 3385 3386 3387 3388 3389

    /* We would also want to check faccessat if we cared about ACLs,
     * but we don't.  */
    if (stat(binary, &sb) < 0) {
        virReportSystemError(errno, _("Cannot check QEMU binary %s"),
                             binary);
        goto error;
    }
3390
    qemuCaps->ctime = sb.st_ctime;
3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401

    /* Make sure the binary we are about to try exec'ing exists.
     * Technically we could catch the exec() failure, but that's
     * in a sub-process so it's hard to feed back a useful error.
     */
    if (!virFileIsExecutable(binary)) {
        virReportSystemError(errno, _("QEMU binary %s is not executable"),
                             binary);
        goto error;
    }

3402
    if ((rv = virQEMUCapsInitCached(qemuCaps, cacheDir)) < 0)
3403 3404
        goto error;

3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418
    if (rv == 0) {
        if (virQEMUCapsInitQMP(qemuCaps, libDir, runUid, runGid) < 0) {
            virQEMUCapsLogProbeFailure(binary);
            goto error;
        }

        if (!qemuCaps->usedQMP &&
            virQEMUCapsInitHelp(qemuCaps, runUid, runGid) < 0) {
            virQEMUCapsLogProbeFailure(binary);
            goto error;
        }

        if (virQEMUCapsRememberCached(qemuCaps, cacheDir) < 0)
            goto error;
3419
    }
3420

3421
    return qemuCaps;
3422

3423
 error:
3424 3425
    virObjectUnref(qemuCaps);
    qemuCaps = NULL;
3426
    return NULL;
3427 3428 3429
}


3430
bool virQEMUCapsIsValid(virQEMUCapsPtr qemuCaps)
3431 3432 3433
{
    struct stat sb;

3434
    if (!qemuCaps->binary)
3435 3436
        return true;

3437
    if (stat(qemuCaps->binary, &sb) < 0)
3438 3439
        return false;

3440
    return sb.st_ctime == qemuCaps->ctime;
3441
}
3442 3443


3444 3445
virQEMUCapsCachePtr
virQEMUCapsCacheNew(const char *libDir,
3446
                    const char *cacheDir,
3447 3448
                    uid_t runUid,
                    gid_t runGid)
3449
{
3450
    virQEMUCapsCachePtr cache;
3451

3452
    if (VIR_ALLOC(cache) < 0)
3453 3454 3455 3456 3457 3458 3459 3460 3461
        return NULL;

    if (virMutexInit(&cache->lock) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Unable to initialize mutex"));
        VIR_FREE(cache);
        return NULL;
    }

E
Eric Blake 已提交
3462
    if (!(cache->binaries = virHashCreate(10, virObjectFreeHashData)))
3463
        goto error;
3464
    if (VIR_STRDUP(cache->libDir, libDir) < 0)
3465
        goto error;
3466 3467
    if (VIR_STRDUP(cache->cacheDir, cacheDir) < 0)
        goto error;
3468

3469 3470 3471
    cache->runUid = runUid;
    cache->runGid = runGid;

3472 3473
    return cache;

3474
 error:
3475
    virQEMUCapsCacheFree(cache);
3476 3477 3478 3479
    return NULL;
}


3480 3481
virQEMUCapsPtr
virQEMUCapsCacheLookup(virQEMUCapsCachePtr cache, const char *binary)
3482
{
3483
    virQEMUCapsPtr ret = NULL;
3484 3485 3486
    virMutexLock(&cache->lock);
    ret = virHashLookup(cache->binaries, binary);
    if (ret &&
3487
        !virQEMUCapsIsValid(ret)) {
3488 3489 3490 3491 3492 3493 3494 3495
        VIR_DEBUG("Cached capabilities %p no longer valid for %s",
                  ret, binary);
        virHashRemoveEntry(cache->binaries, binary);
        ret = NULL;
    }
    if (!ret) {
        VIR_DEBUG("Creating capabilities for %s",
                  binary);
3496
        ret = virQEMUCapsNewForBinary(binary, cache->libDir,
3497
                                      cache->cacheDir,
3498
                                      cache->runUid, cache->runGid);
3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514
        if (ret) {
            VIR_DEBUG("Caching capabilities %p for %s",
                      ret, binary);
            if (virHashAddEntry(cache->binaries, binary, ret) < 0) {
                virObjectUnref(ret);
                ret = NULL;
            }
        }
    }
    VIR_DEBUG("Returning caps %p for %s", ret, binary);
    virObjectRef(ret);
    virMutexUnlock(&cache->lock);
    return ret;
}


3515 3516
virQEMUCapsPtr
virQEMUCapsCacheLookupCopy(virQEMUCapsCachePtr cache, const char *binary)
3517
{
3518 3519
    virQEMUCapsPtr qemuCaps = virQEMUCapsCacheLookup(cache, binary);
    virQEMUCapsPtr ret;
3520

3521
    if (!qemuCaps)
3522 3523
        return NULL;

3524 3525
    ret = virQEMUCapsNewCopy(qemuCaps);
    virObjectUnref(qemuCaps);
3526 3527 3528 3529
    return ret;
}


3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558
static int
virQEMUCapsCompareArch(const void *payload,
                       const void *name ATTRIBUTE_UNUSED,
                       const void *opaque)
{
    struct virQEMUCapsSearchData *data = (struct virQEMUCapsSearchData *) opaque;
    const virQEMUCaps *qemuCaps = payload;

    return qemuCaps->arch == data->arch;
}


virQEMUCapsPtr
virQEMUCapsCacheLookupByArch(virQEMUCapsCachePtr cache,
                             virArch arch)
{
    virQEMUCapsPtr ret = NULL;
    struct virQEMUCapsSearchData data = { .arch = arch };

    virMutexLock(&cache->lock);
    ret = virHashSearch(cache->binaries, virQEMUCapsCompareArch, &data);
    VIR_DEBUG("Returning caps %p for arch %s", ret, virArchToString(arch));
    virObjectRef(ret);
    virMutexUnlock(&cache->lock);

    return ret;
}


3559
void
3560
virQEMUCapsCacheFree(virQEMUCapsCachePtr cache)
3561 3562 3563 3564
{
    if (!cache)
        return;

3565
    VIR_FREE(cache->libDir);
3566
    VIR_FREE(cache->cacheDir);
3567 3568 3569 3570
    virHashFree(cache->binaries);
    virMutexDestroy(&cache->lock);
    VIR_FREE(cache);
}
3571 3572

bool
3573
virQEMUCapsUsedQMP(virQEMUCapsPtr qemuCaps)
3574
{
3575
    return qemuCaps->usedQMP;
3576
}
3577 3578 3579 3580

bool
virQEMUCapsSupportsChardev(virDomainDefPtr def,
                           virQEMUCapsPtr qemuCaps,
3581
                           virDomainChrDefPtr chr)
3582 3583 3584 3585 3586
{
    if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_CHARDEV) ||
        !virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE))
        return false;

3587
    if ((def->os.arch == VIR_ARCH_PPC) || ARCH_IS_PPC64(def->os.arch)) {
3588 3589 3590 3591 3592
        /* only pseries need -device spapr-vty with -chardev */
        return (chr->deviceType == VIR_DOMAIN_CHR_DEVICE_TYPE_SERIAL &&
                chr->info.type == VIR_DOMAIN_DEVICE_ADDRESS_TYPE_SPAPRVIO);
    }

3593
    if ((def->os.arch != VIR_ARCH_ARMV7L) && (def->os.arch != VIR_ARCH_AARCH64))
3594
        return true;
3595

3596 3597 3598 3599 3600 3601
    /* This may not be true for all ARM machine types, but at least
     * the only supported non-virtio serial devices of vexpress and versatile
     * don't have the -chardev property wired up. */
    return (chr->info.type == VIR_DOMAIN_DEVICE_ADDRESS_TYPE_VIRTIO_MMIO ||
            (chr->deviceType == VIR_DOMAIN_CHR_DEVICE_TYPE_CONSOLE &&
             chr->targetType == VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_VIRTIO));
3602
}
3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616


bool
virQEMUCapsIsMachineSupported(virQEMUCapsPtr qemuCaps,
                              const char *canonical_machine)
{
    size_t i;

    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
        if (STREQ(canonical_machine, qemuCaps->machineTypes[i]))
            return true;
    }
    return false;
}
3617 3618 3619 3620 3621 3622 3623 3624 3625


const char *
virQEMUCapsGetDefaultMachine(virQEMUCapsPtr qemuCaps)
{
    if (!qemuCaps->nmachineTypes)
        return NULL;
    return qemuCaps->machineTypes[0];
}
3626 3627


3628
static int
3629
virQEMUCapsFillDomainLoaderCaps(virQEMUCapsPtr qemuCaps,
3630
                                virDomainCapsLoaderPtr capsLoader,
3631
                                virArch arch,
3632 3633
                                char **loader,
                                size_t nloader)
3634
{
3635 3636
    size_t i;

3637
    capsLoader->device.supported = true;
3638

3639
    if (VIR_ALLOC_N(capsLoader->values.values, nloader) < 0)
3640 3641
        return -1;

3642 3643
    for (i = 0; i < nloader; i++) {
        const char *filename = loader[i];
3644 3645 3646 3647 3648 3649

        if (!virFileExists(filename)) {
            VIR_DEBUG("loader filename=%s does not exist", filename);
            continue;
        }

3650
        if (VIR_STRDUP(capsLoader->values.values[capsLoader->values.nvalues],
3651 3652
                       filename) < 0)
            return -1;
3653
        capsLoader->values.nvalues++;
3654 3655
    }

3656
    VIR_DOMAIN_CAPS_ENUM_SET(capsLoader->type,
3657 3658 3659 3660 3661
                             VIR_DOMAIN_LOADER_TYPE_ROM);

    if (arch == VIR_ARCH_X86_64 &&
        virQEMUCapsGet(qemuCaps, QEMU_CAPS_DRIVE) &&
        virQEMUCapsGet(qemuCaps, QEMU_CAPS_DRIVE_FORMAT))
3662
        VIR_DOMAIN_CAPS_ENUM_SET(capsLoader->type,
3663 3664 3665 3666
                                 VIR_DOMAIN_LOADER_TYPE_PFLASH);


    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_DRIVE_READONLY))
3667
        VIR_DOMAIN_CAPS_ENUM_SET(capsLoader->readonly,
3668 3669
                                 VIR_TRISTATE_BOOL_YES,
                                 VIR_TRISTATE_BOOL_NO);
3670
    return 0;
3671 3672 3673
}


3674
static int
3675 3676
virQEMUCapsFillDomainOSCaps(virQEMUCapsPtr qemuCaps,
                            virDomainCapsOSPtr os,
3677
                            virArch arch,
3678 3679
                            char **loader,
                            size_t nloader)
3680
{
3681
    virDomainCapsLoaderPtr capsLoader = &os->loader;
3682 3683

    os->device.supported = true;
3684 3685
    if (virQEMUCapsFillDomainLoaderCaps(qemuCaps, capsLoader, arch,
                                        loader, nloader) < 0)
3686 3687
        return -1;
    return 0;
3688 3689 3690
}


3691
static int
3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713
virQEMUCapsFillDomainDeviceDiskCaps(virQEMUCapsPtr qemuCaps,
                                    virDomainCapsDeviceDiskPtr disk)
{
    disk->device.supported = true;
    /* QEMU supports all of these */
    VIR_DOMAIN_CAPS_ENUM_SET(disk->diskDevice,
                             VIR_DOMAIN_DISK_DEVICE_DISK,
                             VIR_DOMAIN_DISK_DEVICE_CDROM,
                             VIR_DOMAIN_DISK_DEVICE_FLOPPY);

    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_VIRTIO_BLK_SG_IO))
        VIR_DOMAIN_CAPS_ENUM_SET(disk->diskDevice, VIR_DOMAIN_DISK_DEVICE_LUN);

    VIR_DOMAIN_CAPS_ENUM_SET(disk->bus,
                             VIR_DOMAIN_DISK_BUS_IDE,
                             VIR_DOMAIN_DISK_BUS_FDC,
                             VIR_DOMAIN_DISK_BUS_SCSI,
                             VIR_DOMAIN_DISK_BUS_VIRTIO,
                             /* VIR_DOMAIN_DISK_BUS_SD */);

    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_USB_STORAGE))
        VIR_DOMAIN_CAPS_ENUM_SET(disk->bus, VIR_DOMAIN_DISK_BUS_USB);
3714
    return 0;
3715 3716 3717
}


3718
static int
3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762
virQEMUCapsFillDomainDeviceHostdevCaps(virQEMUCapsPtr qemuCaps,
                                       virDomainCapsDeviceHostdevPtr hostdev)
{
    bool supportsPassthroughKVM = qemuHostdevHostSupportsPassthroughLegacy();
    bool supportsPassthroughVFIO = qemuHostdevHostSupportsPassthroughVFIO();

    hostdev->device.supported = true;
    /* VIR_DOMAIN_HOSTDEV_MODE_CAPABILITIES is for containers only */
    VIR_DOMAIN_CAPS_ENUM_SET(hostdev->mode,
                             VIR_DOMAIN_HOSTDEV_MODE_SUBSYS);

    VIR_DOMAIN_CAPS_ENUM_SET(hostdev->startupPolicy,
                             VIR_DOMAIN_STARTUP_POLICY_DEFAULT,
                             VIR_DOMAIN_STARTUP_POLICY_MANDATORY,
                             VIR_DOMAIN_STARTUP_POLICY_REQUISITE,
                             VIR_DOMAIN_STARTUP_POLICY_OPTIONAL);

    VIR_DOMAIN_CAPS_ENUM_SET(hostdev->subsysType,
                             VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB,
                             VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI);
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_DRIVE) &&
        virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE) &&
        virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_SCSI_GENERIC))
        VIR_DOMAIN_CAPS_ENUM_SET(hostdev->subsysType,
                                 VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_SCSI);

    /* No virDomainHostdevCapsType for QEMU */
    virDomainCapsEnumClear(&hostdev->capsType);

    virDomainCapsEnumClear(&hostdev->pciBackend);
    if (supportsPassthroughVFIO &&
        virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_VFIO_PCI)) {
        VIR_DOMAIN_CAPS_ENUM_SET(hostdev->pciBackend,
                                 VIR_DOMAIN_HOSTDEV_PCI_BACKEND_DEFAULT,
                                 VIR_DOMAIN_HOSTDEV_PCI_BACKEND_VFIO);
    }

    if (supportsPassthroughKVM &&
        (virQEMUCapsGet(qemuCaps, QEMU_CAPS_PCIDEVICE) ||
         virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE))) {
        VIR_DOMAIN_CAPS_ENUM_SET(hostdev->pciBackend,
                                 VIR_DOMAIN_HOSTDEV_PCI_BACKEND_DEFAULT,
                                 VIR_DOMAIN_HOSTDEV_PCI_BACKEND_KVM);
    }
3763
    return 0;
3764 3765 3766
}


3767
int
3768
virQEMUCapsFillDomainCaps(virDomainCapsPtr domCaps,
3769
                          virQEMUCapsPtr qemuCaps,
3770 3771
                          char **loader,
                          size_t nloader)
3772
{
3773
    virDomainCapsOSPtr os = &domCaps->os;
3774 3775 3776 3777 3778 3779
    virDomainCapsDeviceDiskPtr disk = &domCaps->disk;
    virDomainCapsDeviceHostdevPtr hostdev = &domCaps->hostdev;
    int maxvcpus = virQEMUCapsGetMachineMaxCpus(qemuCaps, domCaps->machine);

    domCaps->maxvcpus = maxvcpus;

3780 3781
    if (virQEMUCapsFillDomainOSCaps(qemuCaps, os, domCaps->arch,
                                    loader, nloader) < 0 ||
3782 3783 3784 3785
        virQEMUCapsFillDomainDeviceDiskCaps(qemuCaps, disk) < 0 ||
        virQEMUCapsFillDomainDeviceHostdevCaps(qemuCaps, hostdev) < 0)
        return -1;
    return 0;
3786
}