qemu_capabilities.c 54.1 KB
Newer Older
1 2 3
/*
 * qemu_capabilities.c: QEMU capabilities generation
 *
4
 * Copyright (C) 2006-2012 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
O
Osier Yang 已提交
18 19
 * License along with this library;  If not, see
 * <http://www.gnu.org/licenses/>.
20 21 22 23 24 25 26 27 28 29 30
 *
 * Author: Daniel P. Berrange <berrange@redhat.com>
 */

#include <config.h>

#include "qemu_capabilities.h"
#include "memory.h"
#include "logging.h"
#include "virterror_internal.h"
#include "util.h"
E
Eric Blake 已提交
31
#include "virfile.h"
32 33 34 35
#include "nodeinfo.h"
#include "cpu/cpu.h"
#include "domain_conf.h"
#include "qemu_conf.h"
36
#include "command.h"
37
#include "virnodesuspend.h"
38 39 40 41 42

#include <sys/stat.h>
#include <unistd.h>
#include <sys/wait.h>
#include <sys/utsname.h>
43
#include <stdarg.h>
44 45 46

#define VIR_FROM_THIS VIR_FROM_QEMU

47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88
/* While not public, these strings must not change. They
 * are used in domain status files which are read on
 * daemon restarts
 */
VIR_ENUM_IMPL(qemuCaps, QEMU_CAPS_LAST,
              "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 */
89
              "vhost-net",
90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122
              "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",
123 124

              "pci-multifunction", /* 60 */
125
              "virtio-blk-pci.ioeventfd",
M
Michal Privoznik 已提交
126
              "sga",
127 128
              "virtio-blk-pci.event_idx",
              "virtio-net-pci.event_idx",
129 130 131 132 133 134 135 136 137 138

              "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
139
              "usb-hub",
140
              "no-shutdown",
141 142

              "cache-unsafe", /* 75 */
143
              "rombar",
J
Jim Fehlig 已提交
144
              "ich9-ahci",
145
              "no-acpi",
146
              "fsdev-readonly",
147

148
              "virtio-blk-pci.scsi", /* 80 */
149
              "blk-sg-io",
O
Osier Yang 已提交
150
              "drive-copy-on-read",
151
              "cpu-host",
152
              "fsdev-writeout",
153 154

              "drive-iotune", /* 85 */
155
              "system_wakeup",
156
              "scsi-disk.channel",
157
              "scsi-block",
158
              "transaction",
159 160 161

              "block-job-sync", /* 90 */
              "block-job-async",
162
              "scsi-cd",
163
              "ide-cd",
164
              "no-user-config",
M
Marc-André Lureau 已提交
165 166

              "hda-micro", /* 95 */
167
              "dump-guest-memory",
G
Gerd Hoffmann 已提交
168
              "nec-usb-xhci",
169
              "virtio-s390",
170
              "balloon-event",
M
Marc-André Lureau 已提交
171

R
Richa Marwaha 已提交
172
              "bridge", /* 100 */
173 174
              "lsi",
              "virtio-scsi-pci",
V
Viktor Mihajlovski 已提交
175
              "blockio",
176
              "disable-s3",
R
Richa Marwaha 已提交
177

178
              "disable-s4", /* 105 */
179
              "usb-redir.filter",
180 181
    );

182 183 184 185
struct _qemuCaps {
    virObject object;

    virBitmapPtr flags;
186 187 188 189 190 191 192 193 194 195 196 197

    unsigned int version;
    unsigned int kvmVersion;

    char *arch;

    size_t ncpuDefinitions;
    char **cpuDefinitions;

    size_t nmachineTypes;
    char **machineTypes;
    char **machineAliases;
198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215
};


static virClassPtr qemuCapsClass;
static void qemuCapsDispose(void *obj);

static int qemuCapsOnceInit(void)
{
    if (!(qemuCapsClass = virClassNew("qemuCaps",
                                      sizeof(qemuCaps),
                                      qemuCapsDispose)))
        return -1;

    return 0;
}

VIR_ONCE_GLOBAL_INIT(qemuCaps)

216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251
struct qemu_feature_flags {
    const char *name;
    const int default_on;
    const int toggle;
};

struct qemu_arch_info {
    const char *arch;
    int wordsize;
    const char *machine;
    const char *binary;
    const char *altbinary;
    const struct qemu_feature_flags *flags;
    int nflags;
};

/* Feature flags for the architecture info */
static const struct qemu_feature_flags const arch_info_i686_flags [] = {
    { "pae",  1, 0 },
    { "nonpae",  1, 0 },
    { "acpi", 1, 1 },
    { "apic", 1, 0 },
};

static const struct qemu_feature_flags const arch_info_x86_64_flags [] = {
    { "acpi", 1, 1 },
    { "apic", 1, 0 },
};

/* The archicture tables for supported QEMU archs */
static const struct qemu_arch_info const arch_info_hvm[] = {
    {  "i686",   32, NULL, "qemu",
       "qemu-system-x86_64", arch_info_i686_flags, 4 },
    {  "x86_64", 64, NULL, "qemu-system-x86_64",
       NULL, arch_info_x86_64_flags, 2 },
    {  "arm",    32, NULL, "qemu-system-arm",    NULL, NULL, 0 },
252 253
    {  "microblaze", 32, NULL, "qemu-system-microblaze",   NULL, NULL, 0 },
    {  "microblazeel", 32, NULL, "qemu-system-microblazeel",   NULL, NULL, 0 },
254 255 256 257
    {  "mips",   32, NULL, "qemu-system-mips",   NULL, NULL, 0 },
    {  "mipsel", 32, NULL, "qemu-system-mipsel", NULL, NULL, 0 },
    {  "sparc",  32, NULL, "qemu-system-sparc",  NULL, NULL, 0 },
    {  "ppc",    32, NULL, "qemu-system-ppc",    NULL, NULL, 0 },
P
Prerna Saxena 已提交
258
    {  "ppc64",    64, NULL, "qemu-system-ppc64",    NULL, NULL, 0 },
259 260 261 262 263 264 265 266 267 268
    {  "itanium", 64, NULL, "qemu-system-ia64",  NULL, NULL, 0 },
    {  "s390x",  64, NULL, "qemu-system-s390x",  NULL, NULL, 0 },
};

static const struct qemu_arch_info const arch_info_xen[] = {
    {  "i686",   32, "xenner", "xenner", NULL, arch_info_i686_flags, 4 },
    {  "x86_64", 64, "xenner", "xenner", NULL, arch_info_x86_64_flags, 2 },
};


269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288
static virCommandPtr
qemuCapsProbeCommand(const char *qemu,
                     qemuCapsPtr caps)
{
    virCommandPtr cmd = virCommandNew(qemu);

    if (caps) {
        if (qemuCapsGet(caps, QEMU_CAPS_NO_USER_CONFIG))
            virCommandAddArg(cmd, "-no-user-config");
        else if (qemuCapsGet(caps, QEMU_CAPS_NODEFCONFIG))
            virCommandAddArg(cmd, "-nodefconfig");
    }

    virCommandAddEnvPassCommon(cmd);
    virCommandClearCaps(cmd);

    return cmd;
}


289 290 291 292 293 294
/* Format is:
 * <machine> <desc> [(default)|(alias of <canonical>)]
 */
static int
qemuCapsParseMachineTypesStr(const char *output,
                             virCapsGuestMachinePtr **machines,
295
                             size_t *nmachines)
296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361
{
    const char *p = output;
    const char *next;
    virCapsGuestMachinePtr *list = NULL;
    int nitems = 0;

    do {
        const char *t;
        virCapsGuestMachinePtr machine;

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

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

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

        if (VIR_ALLOC(machine) < 0)
            goto no_memory;

        if (!(machine->name = strndup(p, t - p))) {
            VIR_FREE(machine);
            goto no_memory;
        }

        if (VIR_REALLOC_N(list, nitems + 1) < 0) {
            VIR_FREE(machine->name);
            VIR_FREE(machine);
            goto no_memory;
        }

        p = t;
        if (!(t = strstr(p, "(default)")) || (next && t >= next)) {
            list[nitems++] = machine;
        } else {
            /* put the default first in the list */
            memmove(list + 1, list, sizeof(*list) * nitems);
            list[0] = machine;
            nitems++;
        }

        if ((t = strstr(p, "(alias of ")) && (!next || t < next)) {
            p = t + strlen("(alias of ");
            if (!(t = strchr(p, ')')) || (next && t >= next))
                continue;

            if (!(machine->canonical = strndup(p, t - p)))
                goto no_memory;
        }
    } while ((p = next));

    *machines = list;
    *nmachines = nitems;

    return 0;

  no_memory:
    virReportOOMError();
    virCapabilitiesFreeMachines(list, nitems);
    return -1;
}

int
qemuCapsProbeMachineTypes(const char *binary,
362
                          qemuCapsPtr caps,
363
                          virCapsGuestMachinePtr **machines,
364
                          size_t *nmachines)
365 366
{
    char *output;
367 368
    int ret = -1;
    virCommandPtr cmd;
369
    int status;
370

371 372 373 374
    /* 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.
     */
E
Eric Blake 已提交
375
    if (!virFileIsExecutable(binary)) {
376 377 378 379
        virReportSystemError(errno, _("Cannot find QEMU binary %s"), binary);
        return -1;
    }

380
    cmd = qemuCapsProbeCommand(binary, caps);
381
    virCommandAddArgList(cmd, "-M", "?", NULL);
382
    virCommandSetOutputBuffer(cmd, &output);
383

384 385
    /* Ignore failure from older qemu that did not understand '-M ?'.  */
    if (virCommandRun(cmd, &status) < 0)
386 387 388
        goto cleanup;

    if (qemuCapsParseMachineTypesStr(output, machines, nmachines) < 0)
389
        goto cleanup;
390 391 392 393

    ret = 0;

cleanup:
394 395
    VIR_FREE(output);
    virCommandFree(cmd);
396 397 398 399 400 401 402 403 404

    return ret;
}

static int
qemuCapsGetOldMachinesFromInfo(virCapsGuestDomainInfoPtr info,
                               const char *emulator,
                               time_t emulator_mtime,
                               virCapsGuestMachinePtr **machines,
405
                               size_t *nmachines)
406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
{
    virCapsGuestMachinePtr *list;
    int i;

    if (!info->nmachines)
        return 0;

    if (!info->emulator || !STREQ(emulator, info->emulator))
        return 0;

    if (emulator_mtime != info->emulator_mtime) {
        VIR_DEBUG("mtime on %s has changed, refreshing machine types",
                  info->emulator);
        return 0;
    }

    if (VIR_ALLOC_N(list, info->nmachines) < 0) {
        virReportOOMError();
        return 0;
    }

    for (i = 0; i < info->nmachines; i++) {
        if (VIR_ALLOC(list[i]) < 0) {
            goto no_memory;
        }
        if (info->machines[i]->name &&
            !(list[i]->name = strdup(info->machines[i]->name))) {
            goto no_memory;
        }
        if (info->machines[i]->canonical &&
            !(list[i]->canonical = strdup(info->machines[i]->canonical))) {
            goto no_memory;
        }
    }

    *machines = list;
    *nmachines = info->nmachines;

    return 1;

  no_memory:
    virReportOOMError();
    virCapabilitiesFreeMachines(list, info->nmachines);
    return 0;
}

static int
qemuCapsGetOldMachines(const char *ostype,
                       const char *arch,
                       int wordsize,
                       const char *emulator,
                       time_t emulator_mtime,
                       virCapsPtr old_caps,
                       virCapsGuestMachinePtr **machines,
460
                       size_t *nmachines)
461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493
{
    int i;

    for (i = 0; i < old_caps->nguests; i++) {
        virCapsGuestPtr guest = old_caps->guests[i];
        int j;

        if (!STREQ(ostype, guest->ostype) ||
            !STREQ(arch, guest->arch.name) ||
            wordsize != guest->arch.wordsize)
            continue;

        for (j = 0; j < guest->arch.ndomains; j++) {
            virCapsGuestDomainPtr dom = guest->arch.domains[j];

            if (qemuCapsGetOldMachinesFromInfo(&dom->info,
                                               emulator, emulator_mtime,
                                               machines, nmachines))
                return 1;
        }

        if (qemuCapsGetOldMachinesFromInfo(&guest->arch.defaultInfo,
                                           emulator, emulator_mtime,
                                           machines, nmachines))
            return 1;
    }

    return 0;
}


typedef int
(*qemuCapsParseCPUModels)(const char *output,
494 495
                          size_t *retcount,
                          const char ***retcpus);
496 497 498 499 500 501 502 503

/* Format:
 *      <arch> <model>
 * qemu-0.13 encloses some model names in []:
 *      <arch> [<model>]
 */
static int
qemuCapsParseX86Models(const char *output,
504
                       size_t *retcount,
505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534
                       const char ***retcpus)
{
    const char *p = output;
    const char *next;
    unsigned int count = 0;
    const char **cpus = NULL;
    int i;

    do {
        const char *t;

        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;

        if (retcpus) {
            unsigned int len;

535 536
            if (VIR_REALLOC_N(cpus, count + 1) < 0) {
                virReportOOMError();
537
                goto error;
538
            }
539 540 541 542 543 544 545 546 547 548 549

            if (next)
                len = next - p - 1;
            else
                len = strlen(p);

            if (len > 2 && *p == '[' && p[len - 1] == ']') {
                p++;
                len -= 2;
            }

550 551
            if (!(cpus[count] = strndup(p, len))) {
                virReportOOMError();
552
                goto error;
553
            }
554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574
        }
        count++;
    } while ((p = next));

    if (retcount)
        *retcount = count;
    if (retcpus)
        *retcpus = cpus;

    return 0;

error:
    if (cpus) {
        for (i = 0; i < count; i++)
            VIR_FREE(cpus[i]);
    }
    VIR_FREE(cpus);

    return -1;
}

P
Prerna Saxena 已提交
575 576 577 578 579
/* ppc64 parser.
 * Format : PowerPC <machine> <description>
 */
static int
qemuCapsParsePPCModels(const char *output,
580
                       size_t *retcount,
P
Prerna Saxena 已提交
581 582 583 584 585 586
                       const char ***retcpus)
{
    const char *p = output;
    const char *next;
    unsigned int count = 0;
    const char **cpus = NULL;
587
    int i, ret = -1;
P
Prerna Saxena 已提交
588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613

    do {
        const char *t;

        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;

        if (retcpus) {
            unsigned int len;

614
            if (VIR_REALLOC_N(cpus, count + 1) < 0) {
P
Prerna Saxena 已提交
615
                virReportOOMError();
616 617
                goto cleanup;
            }
P
Prerna Saxena 已提交
618 619 620

            len = t - p - 1;

621
            if (!(cpus[count] = strndup(p, len))) {
P
Prerna Saxena 已提交
622
                virReportOOMError();
623 624
                goto cleanup;
            }
P
Prerna Saxena 已提交
625 626 627 628 629 630
        }
        count++;
    } while ((p = next));

    if (retcount)
        *retcount = count;
631
    if (retcpus) {
P
Prerna Saxena 已提交
632
        *retcpus = cpus;
633 634 635
        cpus = NULL;
    }
    ret = 0;
P
Prerna Saxena 已提交
636

637
cleanup:
P
Prerna Saxena 已提交
638 639 640
    if (cpus) {
        for (i = 0; i < count; i++)
            VIR_FREE(cpus[i]);
641
        VIR_FREE(cpus);
P
Prerna Saxena 已提交
642
    }
643
    return ret;
P
Prerna Saxena 已提交
644
}
645 646 647

int
qemuCapsProbeCPUModels(const char *qemu,
648
                       qemuCapsPtr caps,
649
                       const char *arch,
650
                       size_t *count,
651 652 653 654 655
                       const char ***cpus)
{
    char *output = NULL;
    int ret = -1;
    qemuCapsParseCPUModels parse;
656
    virCommandPtr cmd;
657 658 659 660 661 662 663 664

    if (count)
        *count = 0;
    if (cpus)
        *cpus = NULL;

    if (STREQ(arch, "i686") || STREQ(arch, "x86_64"))
        parse = qemuCapsParseX86Models;
P
Prerna Saxena 已提交
665 666
    else if (STREQ(arch, "ppc64"))
        parse = qemuCapsParsePPCModels;
667 668 669 670 671
    else {
        VIR_DEBUG("don't know how to parse %s CPU models", arch);
        return 0;
    }

672
    cmd = qemuCapsProbeCommand(qemu, caps);
673
    virCommandAddArgList(cmd, "-cpu", "?", NULL);
674
    virCommandSetOutputBuffer(cmd, &output);
675

676
    if (virCommandRun(cmd, NULL) < 0)
677 678
        goto cleanup;

679
    if (parse(output, count, cpus) < 0)
680 681 682 683 684 685
        goto cleanup;

    ret = 0;

cleanup:
    VIR_FREE(output);
686
    virCommandFree(cmd);
687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706

    return ret;
}


static int
qemuCapsInitGuest(virCapsPtr caps,
                  virCapsPtr old_caps,
                  const char *hostmachine,
                  const struct qemu_arch_info *info,
                  int hvm)
{
    virCapsGuestPtr guest;
    int i;
    int haskvm = 0;
    int haskqemu = 0;
    char *kvmbin = NULL;
    char *binary = NULL;
    time_t binary_mtime;
    virCapsGuestMachinePtr *machines = NULL;
707
    size_t nmachines = 0;
708
    struct stat st;
709
    size_t ncpus;
710 711
    qemuCapsPtr qemubinCaps = NULL;
    qemuCapsPtr kvmbinCaps = NULL;
712 713 714 715 716 717 718
    int ret = -1;

    /* Check for existance of base emulator, or alternate base
     * which can be used with magic cpu choice
     */
    binary = virFindFileInPath(info->binary);

E
Eric Blake 已提交
719
    if (binary == NULL || !virFileIsExecutable(binary)) {
720 721 722 723
        VIR_FREE(binary);
        binary = virFindFileInPath(info->altbinary);
    }

724 725
    /* Ignore binary if extracting version info fails */
    if (binary &&
726
        qemuCapsExtractVersionInfo(binary, info->arch,
727
                                   false, NULL, &qemubinCaps) < 0)
728 729 730
        VIR_FREE(binary);

    /* qemu-kvm/kvm binaries can only be used if
731 732 733 734 735 736 737
     *  - host & guest arches match
     * Or
     *  - hostarch is x86_64 and guest arch is i686
     * The latter simply needs "-cpu qemu32"
     */
    if (STREQ(info->arch, hostmachine) ||
        (STREQ(hostmachine, "x86_64") && STREQ(info->arch, "i686"))) {
738 739 740
        const char *const kvmbins[] = { "/usr/libexec/qemu-kvm", /* RHEL */
                                        "qemu-kvm", /* Fedora */
                                        "kvm" }; /* Upstream .spec */
741

742 743
        for (i = 0; i < ARRAY_CARDINALITY(kvmbins); ++i) {
            kvmbin = virFindFileInPath(kvmbins[i]);
744

745 746
            if (!kvmbin)
                continue;
747

748
            if (qemuCapsExtractVersionInfo(kvmbin, info->arch,
749
                                           false, NULL,
750
                                           &kvmbinCaps) < 0) {
751 752 753
                VIR_FREE(kvmbin);
                continue;
            }
754

755 756
            if (!binary) {
                binary = kvmbin;
757
                qemubinCaps = kvmbinCaps;
758
                kvmbin = NULL;
759
                kvmbinCaps = NULL;
760
            }
761
            break;
762 763 764 765 766 767
        }
    }

    if (!binary)
        return 0;

768
    if (access("/dev/kvm", F_OK) == 0 &&
769 770
        (qemuCapsGet(qemubinCaps, QEMU_CAPS_KVM) ||
         qemuCapsGet(qemubinCaps, QEMU_CAPS_ENABLE_KVM) ||
771 772 773 774
         kvmbin))
        haskvm = 1;

    if (access("/dev/kqemu", F_OK) == 0 &&
775
        qemuCapsGet(qemubinCaps, QEMU_CAPS_KQEMU))
776
        haskqemu = 1;
J
Jiri Denemark 已提交
777

778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814
    if (stat(binary, &st) == 0) {
        binary_mtime = st.st_mtime;
    } else {
        char ebuf[1024];
        VIR_WARN("Failed to stat %s, most peculiar : %s",
                 binary, virStrerror(errno, ebuf, sizeof(ebuf)));
        binary_mtime = 0;
    }

    if (info->machine) {
        virCapsGuestMachinePtr machine;

        if (VIR_ALLOC(machine) < 0) {
            goto no_memory;
        }

        if (!(machine->name = strdup(info->machine))) {
            VIR_FREE(machine);
            goto no_memory;
        }

        nmachines = 1;

        if (VIR_ALLOC_N(machines, nmachines) < 0) {
            VIR_FREE(machine->name);
            VIR_FREE(machine);
            goto no_memory;
        }

        machines[0] = machine;
    } else {
        int probe = 1;
        if (old_caps && binary_mtime)
            probe = !qemuCapsGetOldMachines(hvm ? "hvm" : "xen", info->arch,
                                            info->wordsize, binary, binary_mtime,
                                            old_caps, &machines, &nmachines);
        if (probe &&
815
            qemuCapsProbeMachineTypes(binary, qemubinCaps,
816
                                      &machines, &nmachines) < 0)
817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837
            goto error;
    }

    /* We register kvm as the base emulator too, since we can
     * just give -no-kvm to disable acceleration if required */
    if ((guest = virCapabilitiesAddGuest(caps,
                                         hvm ? "hvm" : "xen",
                                         info->arch,
                                         info->wordsize,
                                         binary,
                                         NULL,
                                         nmachines,
                                         machines)) == NULL)
        goto error;

    machines = NULL;
    nmachines = 0;

    guest->arch.defaultInfo.emulator_mtime = binary_mtime;

    if (caps->host.cpu &&
838
        qemuCapsProbeCPUModels(binary, NULL, info->arch, &ncpus, NULL) == 0 &&
839 840 841 842
        ncpus > 0 &&
        !virCapabilitiesAddGuestFeature(guest, "cpuselection", 1, 0))
        goto error;

843
    if (qemuCapsGet(qemubinCaps, QEMU_CAPS_BOOTINDEX) &&
J
Jiri Denemark 已提交
844
        !virCapabilitiesAddGuestFeature(guest, "deviceboot", 1, 0))
845 846
        goto error;

847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867
    if (hvm) {
        if (virCapabilitiesAddGuestDomain(guest,
                                          "qemu",
                                          NULL,
                                          NULL,
                                          0,
                                          NULL) == NULL)
            goto error;

        if (haskqemu &&
            virCapabilitiesAddGuestDomain(guest,
                                          "kqemu",
                                          NULL,
                                          NULL,
                                          0,
                                          NULL) == NULL)
            goto error;

        if (haskvm) {
            virCapsGuestDomainPtr dom;

868
            if (kvmbin) {
869
                int probe = 1;
870 871 872 873 874 875 876 877 878 879

                if (stat(kvmbin, &st) == 0) {
                    binary_mtime = st.st_mtime;
                } else {
                    char ebuf[1024];
                    VIR_WARN("Failed to stat %s, most peculiar : %s",
                             binary, virStrerror(errno, ebuf, sizeof(ebuf)));
                    binary_mtime = 0;
                }

880
                if (old_caps && binary_mtime)
881 882 883 884
                    probe = !qemuCapsGetOldMachines("hvm", info->arch,
                                                    info->wordsize, kvmbin,
                                                    binary_mtime, old_caps,
                                                    &machines, &nmachines);
885
                if (probe &&
886
                    qemuCapsProbeMachineTypes(kvmbin, kvmbinCaps,
887
                                              &machines, &nmachines) < 0)
888 889 890 891 892
                    goto error;
            }

            if ((dom = virCapabilitiesAddGuestDomain(guest,
                                                     "kvm",
893
                                                     kvmbin ? kvmbin : binary,
894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927
                                                     NULL,
                                                     nmachines,
                                                     machines)) == NULL) {
                goto error;
            }

            machines = NULL;
            nmachines = 0;

            dom->info.emulator_mtime = binary_mtime;
        }
    } else {
        if (virCapabilitiesAddGuestDomain(guest,
                                          "kvm",
                                          NULL,
                                          NULL,
                                          0,
                                          NULL) == NULL)
            goto error;
    }

    if (info->nflags) {
        for (i = 0 ; i < info->nflags ; i++) {
            if (virCapabilitiesAddGuestFeature(guest,
                                               info->flags[i].name,
                                               info->flags[i].default_on,
                                               info->flags[i].toggle) == NULL)
                goto error;
        }
    }

    ret = 0;

cleanup:
928 929
    VIR_FREE(binary);
    VIR_FREE(kvmbin);
930 931
    virObjectUnref(qemubinCaps);
    virObjectUnref(kvmbinCaps);
932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986

    return ret;

no_memory:
    virReportOOMError();

error:
    virCapabilitiesFreeMachines(machines, nmachines);

    goto cleanup;
}


static int
qemuCapsInitCPU(virCapsPtr caps,
                 const char *arch)
{
    virCPUDefPtr cpu = NULL;
    union cpuData *data = NULL;
    virNodeInfo nodeinfo;
    int ret = -1;

    if (VIR_ALLOC(cpu) < 0
        || !(cpu->arch = strdup(arch))) {
        virReportOOMError();
        goto error;
    }

    if (nodeGetInfo(NULL, &nodeinfo))
        goto error;

    cpu->type = VIR_CPU_TYPE_HOST;
    cpu->sockets = nodeinfo.sockets;
    cpu->cores = nodeinfo.cores;
    cpu->threads = nodeinfo.threads;

    if (!(data = cpuNodeData(arch))
        || cpuDecode(cpu, data, NULL, 0, NULL) < 0)
        goto error;

    caps->host.cpu = cpu;

    ret = 0;

cleanup:
    cpuDataFree(arch, data);

    return ret;

error:
    virCPUDefFree(cpu);
    goto cleanup;
}


987 988 989 990 991 992
static int qemuDefaultConsoleType(const char *ostype ATTRIBUTE_UNUSED)
{
    return VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_SERIAL;
}


993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
virCapsPtr qemuCapsInit(virCapsPtr old_caps)
{
    struct utsname utsname;
    virCapsPtr caps;
    int i;
    char *xenner = NULL;

    /* Really, this never fails - look at the man-page. */
    uname (&utsname);

    if ((caps = virCapabilitiesNew(utsname.machine,
                                   1, 1)) == NULL)
        goto no_memory;

    /* Using KVM's mac prefix for QEMU too */
    virCapabilitiesSetMacPrefix(caps, (unsigned char[]){ 0x52, 0x54, 0x00 });

    /* 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);
1016
        VIR_WARN("Failed to query host NUMA topology, disabling NUMA capabilities");
1017 1018 1019 1020
    }

    if (old_caps == NULL || old_caps->host.cpu == NULL) {
        if (qemuCapsInitCPU(caps, utsname.machine) < 0)
1021
            VIR_WARN("Failed to get host CPU");
1022 1023 1024 1025 1026 1027
    }
    else {
        caps->host.cpu = old_caps->host.cpu;
        old_caps->host.cpu = NULL;
    }

1028 1029
    /* Add the power management features of the host */

1030
    if (virNodeSuspendGetTargetMask(&caps->host.powerMgmt) < 0)
1031 1032
        VIR_WARN("Failed to get host power management capabilities");

1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045
    virCapabilitiesAddHostMigrateTransport(caps,
                                           "tcp");

    /* First the pure HVM guests */
    for (i = 0 ; i < ARRAY_CARDINALITY(arch_info_hvm) ; i++)
        if (qemuCapsInitGuest(caps, old_caps,
                              utsname.machine,
                              &arch_info_hvm[i], 1) < 0)
            goto no_memory;

    /* Then possibly the Xen paravirt guests (ie Xenner */
    xenner = virFindFileInPath("xenner");

E
Eric Blake 已提交
1046
    if (xenner != NULL && virFileIsExecutable(xenner) == 0 &&
1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064
        access("/dev/kvm", F_OK) == 0) {
        for (i = 0 ; i < ARRAY_CARDINALITY(arch_info_xen) ; i++)
            /* Allow Xen 32-on-32, 32-on-64 and 64-on-64 */
            if (STREQ(arch_info_xen[i].arch, utsname.machine) ||
                (STREQ(utsname.machine, "x86_64") &&
                 STREQ(arch_info_xen[i].arch, "i686"))) {
                if (qemuCapsInitGuest(caps, old_caps,
                                      utsname.machine,
                                      &arch_info_xen[i], 0) < 0)
                    goto no_memory;
            }
    }

    VIR_FREE(xenner);

    /* QEMU Requires an emulator in the XML */
    virCapabilitiesSetEmulatorRequired(caps);

1065
    caps->defaultConsoleTargetType = qemuDefaultConsoleType;
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075

    return caps;

 no_memory:
    VIR_FREE(xenner);
    virCapabilitiesFree(caps);
    return NULL;
}


1076
static int
1077 1078 1079
qemuCapsComputeCmdFlags(const char *help,
                        unsigned int version,
                        unsigned int is_kvm,
1080
                        unsigned int kvm_version,
1081
                        qemuCapsPtr caps,
1082
                        bool check_yajl ATTRIBUTE_UNUSED)
1083 1084
{
    const char *p;
R
Richa Marwaha 已提交
1085
    const char *fsdev, *netdev;
1086 1087

    if (strstr(help, "-no-kqemu"))
1088
        qemuCapsSet(caps, QEMU_CAPS_KQEMU);
1089
    if (strstr(help, "-enable-kqemu"))
1090
        qemuCapsSet(caps, QEMU_CAPS_ENABLE_KQEMU);
1091
    if (strstr(help, "-no-kvm"))
1092
        qemuCapsSet(caps, QEMU_CAPS_KVM);
1093
    if (strstr(help, "-enable-kvm"))
1094
        qemuCapsSet(caps, QEMU_CAPS_ENABLE_KVM);
1095
    if (strstr(help, "-no-reboot"))
1096
        qemuCapsSet(caps, QEMU_CAPS_NO_REBOOT);
1097
    if (strstr(help, "-name")) {
1098
        qemuCapsSet(caps, QEMU_CAPS_NAME);
1099
        if (strstr(help, ",process="))
1100
            qemuCapsSet(caps, QEMU_CAPS_NAME_PROCESS);
1101 1102
    }
    if (strstr(help, "-uuid"))
1103
        qemuCapsSet(caps, QEMU_CAPS_UUID);
1104
    if (strstr(help, "-xen-domid"))
1105
        qemuCapsSet(caps, QEMU_CAPS_XEN_DOMID);
1106
    else if (strstr(help, "-domid"))
1107
        qemuCapsSet(caps, QEMU_CAPS_DOMID);
1108
    if (strstr(help, "-drive")) {
1109 1110
        const char *cache = strstr(help, "cache=");

1111
        qemuCapsSet(caps, QEMU_CAPS_DRIVE);
1112 1113
        if (cache && (p = strchr(cache, ']'))) {
            if (memmem(cache, p - cache, "on|off", sizeof("on|off") - 1) == NULL)
1114
                qemuCapsSet(caps, QEMU_CAPS_DRIVE_CACHE_V2);
1115
            if (memmem(cache, p - cache, "directsync", sizeof("directsync") - 1))
1116
                qemuCapsSet(caps, QEMU_CAPS_DRIVE_CACHE_DIRECTSYNC);
1117
            if (memmem(cache, p - cache, "unsafe", sizeof("unsafe") - 1))
1118
                qemuCapsSet(caps, QEMU_CAPS_DRIVE_CACHE_UNSAFE);
1119
        }
1120
        if (strstr(help, "format="))
1121
            qemuCapsSet(caps, QEMU_CAPS_DRIVE_FORMAT);
1122
        if (strstr(help, "readonly="))
1123
            qemuCapsSet(caps, QEMU_CAPS_DRIVE_READONLY);
1124
        if (strstr(help, "aio=threads|native"))
1125
            qemuCapsSet(caps, QEMU_CAPS_DRIVE_AIO);
O
Osier Yang 已提交
1126
        if (strstr(help, "copy-on-read=on|off"))
1127
            qemuCapsSet(caps, QEMU_CAPS_DRIVE_COPY_ON_READ);
1128
        if (strstr(help, "bps="))
1129
            qemuCapsSet(caps, QEMU_CAPS_DRIVE_IOTUNE);
1130 1131 1132 1133
    }
    if ((p = strstr(help, "-vga")) && !strstr(help, "-std-vga")) {
        const char *nl = strstr(p, "\n");

1134
        qemuCapsSet(caps, QEMU_CAPS_VGA);
1135 1136

        if (strstr(p, "|qxl"))
1137
            qemuCapsSet(caps, QEMU_CAPS_VGA_QXL);
1138
        if ((p = strstr(p, "|none")) && p < nl)
1139
            qemuCapsSet(caps, QEMU_CAPS_VGA_NONE);
1140 1141
    }
    if (strstr(help, "-spice"))
1142
        qemuCapsSet(caps, QEMU_CAPS_SPICE);
1143
    if (strstr(help, "boot=on"))
1144
        qemuCapsSet(caps, QEMU_CAPS_DRIVE_BOOT);
1145
    if (strstr(help, "serial=s"))
1146
        qemuCapsSet(caps, QEMU_CAPS_DRIVE_SERIAL);
1147
    if (strstr(help, "-pcidevice"))
1148
        qemuCapsSet(caps, QEMU_CAPS_PCIDEVICE);
1149
    if (strstr(help, "-mem-path"))
1150
        qemuCapsSet(caps, QEMU_CAPS_MEM_PATH);
1151
    if (strstr(help, "-chardev")) {
1152
        qemuCapsSet(caps, QEMU_CAPS_CHARDEV);
1153
        if (strstr(help, "-chardev spicevmc"))
1154
            qemuCapsSet(caps, QEMU_CAPS_CHARDEV_SPICEVMC);
1155
    }
1156
    if (strstr(help, "-balloon"))
1157
        qemuCapsSet(caps, QEMU_CAPS_BALLOON);
1158
    if (strstr(help, "-device")) {
1159
        qemuCapsSet(caps, QEMU_CAPS_DEVICE);
1160 1161 1162 1163
        /*
         * When -device was introduced, qemu already supported drive's
         * readonly option but didn't advertise that.
         */
1164
        qemuCapsSet(caps, QEMU_CAPS_DRIVE_READONLY);
1165 1166
    }
    if (strstr(help, "-nodefconfig"))
1167
        qemuCapsSet(caps, QEMU_CAPS_NODEFCONFIG);
1168
    if (strstr(help, "-no-user-config"))
1169
        qemuCapsSet(caps, QEMU_CAPS_NO_USER_CONFIG);
1170 1171
    /* The trailing ' ' is important to avoid a bogus match */
    if (strstr(help, "-rtc "))
1172
        qemuCapsSet(caps, QEMU_CAPS_RTC);
1173 1174
    /* to wit */
    if (strstr(help, "-rtc-td-hack"))
1175
        qemuCapsSet(caps, QEMU_CAPS_RTC_TD_HACK);
1176
    if (strstr(help, "-no-hpet"))
1177
        qemuCapsSet(caps, QEMU_CAPS_NO_HPET);
1178
    if (strstr(help, "-no-acpi"))
1179
        qemuCapsSet(caps, QEMU_CAPS_NO_ACPI);
1180
    if (strstr(help, "-no-kvm-pit-reinjection"))
1181
        qemuCapsSet(caps, QEMU_CAPS_NO_KVM_PIT);
1182
    if (strstr(help, "-tdf"))
1183
        qemuCapsSet(caps, QEMU_CAPS_TDF);
1184
    if (strstr(help, "-enable-nesting"))
1185
        qemuCapsSet(caps, QEMU_CAPS_NESTING);
1186
    if (strstr(help, ",menu=on"))
1187
        qemuCapsSet(caps, QEMU_CAPS_BOOT_MENU);
1188
    if ((fsdev = strstr(help, "-fsdev"))) {
1189
        qemuCapsSet(caps, QEMU_CAPS_FSDEV);
1190
        if (strstr(fsdev, "readonly"))
1191
            qemuCapsSet(caps, QEMU_CAPS_FSDEV_READONLY);
1192
        if (strstr(fsdev, "writeout"))
1193
            qemuCapsSet(caps, QEMU_CAPS_FSDEV_WRITEOUT);
1194
    }
1195
    if (strstr(help, "-smbios type"))
1196
        qemuCapsSet(caps, QEMU_CAPS_SMBIOS_TYPE);
1197

R
Richa Marwaha 已提交
1198
    if ((netdev = strstr(help, "-netdev"))) {
1199 1200
        /* Disable -netdev on 0.12 since although it exists,
         * the corresponding netdev_add/remove monitor commands
1201 1202
         * do not, and we need them to be able to do hotplug.
         * But see below about RHEL build. */
R
Richa Marwaha 已提交
1203 1204
        if (version >= 13000) {
            if (strstr(netdev, "bridge"))
1205 1206
                qemuCapsSet(caps, QEMU_CAPS_NETDEV_BRIDGE);
           qemuCapsSet(caps, QEMU_CAPS_NETDEV);
R
Richa Marwaha 已提交
1207
        }
1208 1209 1210
    }

    if (strstr(help, "-sdl"))
1211
        qemuCapsSet(caps, QEMU_CAPS_SDL);
1212 1213 1214
    if (strstr(help, "cores=") &&
        strstr(help, "threads=") &&
        strstr(help, "sockets="))
1215
        qemuCapsSet(caps, QEMU_CAPS_SMP_TOPOLOGY);
1216 1217

    if (version >= 9000)
1218
        qemuCapsSet(caps, QEMU_CAPS_VNC_COLON);
1219 1220

    if (is_kvm && (version >= 10000 || kvm_version >= 74))
1221
        qemuCapsSet(caps, QEMU_CAPS_VNET_HDR);
1222

1223
    if (strstr(help, ",vhost=")) {
1224
        qemuCapsSet(caps, QEMU_CAPS_VHOST_NET);
1225 1226
    }

1227 1228
    /* 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
1229
     * 0.14.* and 0.15.0)
1230
     */
1231
    if (strstr(help, "-no-shutdown") && (version < 14000 || version > 15000))
1232
        qemuCapsSet(caps, QEMU_CAPS_NO_SHUTDOWN);
1233

1234 1235 1236 1237
    /*
     * Handling of -incoming arg with varying features
     *  -incoming tcp    (kvm >= 79, qemu >= 0.10.0)
     *  -incoming exec   (kvm >= 80, qemu >= 0.10.0)
1238 1239
     *  -incoming unix   (qemu >= 0.12.0)
     *  -incoming fd     (qemu >= 0.12.0)
1240 1241 1242 1243 1244 1245 1246
     *  -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) {
1247 1248
        qemuCapsSet(caps, QEMU_CAPS_MIGRATE_QEMU_TCP);
        qemuCapsSet(caps, QEMU_CAPS_MIGRATE_QEMU_EXEC);
1249
        if (version >= 12000) {
1250 1251
            qemuCapsSet(caps, QEMU_CAPS_MIGRATE_QEMU_UNIX);
            qemuCapsSet(caps, QEMU_CAPS_MIGRATE_QEMU_FD);
1252
        }
1253
    } else if (kvm_version >= 79) {
1254
        qemuCapsSet(caps, QEMU_CAPS_MIGRATE_QEMU_TCP);
1255
        if (kvm_version >= 80)
1256
            qemuCapsSet(caps, QEMU_CAPS_MIGRATE_QEMU_EXEC);
1257
    } else if (kvm_version > 0) {
1258
        qemuCapsSet(caps, QEMU_CAPS_MIGRATE_KVM_STDIO);
1259 1260 1261
    }

    if (version >= 10000)
1262
        qemuCapsSet(caps, QEMU_CAPS_0_10);
1263

1264
    if (version >= 11000)
1265
        qemuCapsSet(caps, QEMU_CAPS_VIRTIO_BLK_SG_IO);
1266

1267 1268 1269
    /* 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
1270 1271 1272 1273 1274
     * 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.
1275
     */
E
Eric Blake 已提交
1276
#if HAVE_YAJL
1277
    if (version >= 13000) {
1278
        qemuCapsSet(caps, QEMU_CAPS_MONITOR_JSON);
1279 1280
    } else if (version >= 12000 &&
               strstr(help, "libvirt")) {
1281 1282
        qemuCapsSet(caps, QEMU_CAPS_MONITOR_JSON);
        qemuCapsSet(caps, QEMU_CAPS_NETDEV);
1283
    }
1284 1285 1286 1287 1288 1289 1290
#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
1291 1292
     * dependency, so distros won't hit this).  This check is
     * also in configure.ac (see $with_yajl).  */
1293 1294 1295
    if (version >= 15000 ||
        (version >= 12000 && strstr(help, "libvirt"))) {
        if (check_yajl) {
1296 1297 1298
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("this qemu binary requires libvirt to be "
                             "compiled with yajl"));
1299 1300
            return -1;
        }
1301
        qemuCapsSet(caps, QEMU_CAPS_NETDEV);
1302
    }
E
Eric Blake 已提交
1303
#endif
1304 1305

    if (version >= 13000)
1306
        qemuCapsSet(caps, QEMU_CAPS_PCI_MULTIFUNCTION);
1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318

    /* 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)
1319
        qemuCapsSet(caps, QEMU_CAPS_PCI_ROMBAR);
1320 1321

    if (version >= 11000)
1322
        qemuCapsSet(caps, QEMU_CAPS_CPU_HOST);
1323
    return 0;
1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351
}

/* 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)

int qemuCapsParseHelpStr(const char *qemu,
                         const char *help,
1352
                         qemuCapsPtr caps,
1353 1354
                         unsigned int *version,
                         unsigned int *is_kvm,
1355 1356
                         unsigned int *kvm_version,
                         bool check_yajl)
1357 1358 1359
{
    unsigned major, minor, micro;
    const char *p = help;
1360
    char *strflags;
1361

1362
    *version = *is_kvm = *kvm_version = 0;
1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379

    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
1380
    if (minor == -1)
1381 1382
        goto fail;

J
Jiri Denemark 已提交
1383 1384 1385 1386 1387 1388 1389 1390
    if (*p != '.') {
        micro = 0;
    } else {
        ++p;
        micro = virParseNumber(&p);
        if (micro == -1)
            goto fail;
    }
1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411

    SKIP_BLANKS(p);

    if (STRPREFIX(p, QEMU_KVM_VER_PREFIX)) {
        *is_kvm = 1;
        p += strlen(QEMU_KVM_VER_PREFIX);
    } else if (STRPREFIX(p, KVM_VER_PREFIX)) {
        int ret;

        *is_kvm = 1;
        p += strlen(KVM_VER_PREFIX);

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

        *kvm_version = ret;
    }

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

1412
    if (qemuCapsComputeCmdFlags(help, *version, *is_kvm, *kvm_version,
1413
                                caps, check_yajl) < 0)
1414
        goto cleanup;
1415

1416
    strflags = virBitmapString(caps->flags);
1417 1418 1419
    VIR_DEBUG("Version %u.%u.%u, cooked version %u, flags %s",
              major, minor, micro, *version, NULLSTR(strflags));
    VIR_FREE(strflags);
1420 1421 1422 1423 1424 1425 1426 1427 1428 1429

    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;

fail:
    p = strchr(help, '\n');
1430 1431
    if (!p)
        p = strchr(help, '\0');
1432

1433 1434 1435
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("cannot parse %s version number in '%.*s'"),
                   qemu, (int) (p - help), help);
1436

1437
cleanup:
1438 1439 1440
    return -1;
}

E
Eric Blake 已提交
1441 1442
static int
qemuCapsExtractDeviceStr(const char *qemu,
1443
                         qemuCapsPtr caps)
1444
{
E
Eric Blake 已提交
1445
    char *output = NULL;
1446
    virCommandPtr cmd;
E
Eric Blake 已提交
1447
    int ret = -1;
1448

E
Eric Blake 已提交
1449 1450
    /* Cram together all device-related queries into one invocation;
     * the output format makes it possible to distinguish what we
1451 1452
     * need.  With qemu 0.13.0 and later, unrecognized '-device
     * bogus,?' cause an error in isolation, but are silently ignored
1453
     * in combination with '-device ?'.  Upstream qemu 0.12.x doesn't
1454 1455
     * understand '-device name,?', and always exits with status 1 for
     * the simpler '-device ?', so this function is really only useful
1456
     * if -help includes "device driver,?".  */
1457
    cmd = qemuCapsProbeCommand(qemu, caps);
1458 1459 1460 1461 1462 1463
    virCommandAddArgList(cmd,
                         "-device", "?",
                         "-device", "pci-assign,?",
                         "-device", "virtio-blk-pci,?",
                         "-device", "virtio-net-pci,?",
                         "-device", "scsi-disk,?",
1464
                         "-device", "PIIX4_PM,?",
1465
                         "-device", "usb-redir,?",
1466
                         NULL);
1467
    /* qemu -help goes to stdout, but qemu -device ? goes to stderr.  */
E
Eric Blake 已提交
1468
    virCommandSetErrorBuffer(cmd, &output);
1469

1470
    if (virCommandRun(cmd, NULL) < 0)
1471 1472
        goto cleanup;

1473
    ret = qemuCapsParseDeviceStr(output, caps);
1474 1475

cleanup:
E
Eric Blake 已提交
1476
    VIR_FREE(output);
1477
    virCommandFree(cmd);
E
Eric Blake 已提交
1478 1479 1480 1481 1482
    return ret;
}


int
1483
qemuCapsParseDeviceStr(const char *str, qemuCapsPtr caps)
E
Eric Blake 已提交
1484
{
1485 1486
    /* Which devices exist. */
    if (strstr(str, "name \"hda-duplex\""))
1487
        qemuCapsSet(caps, QEMU_CAPS_HDA_DUPLEX);
M
Marc-André Lureau 已提交
1488
    if (strstr(str, "name \"hda-micro\""))
1489
        qemuCapsSet(caps, QEMU_CAPS_HDA_MICRO);
1490
    if (strstr(str, "name \"ccid-card-emulated\""))
1491
        qemuCapsSet(caps, QEMU_CAPS_CCID_EMULATED);
1492
    if (strstr(str, "name \"ccid-card-passthru\""))
1493
        qemuCapsSet(caps, QEMU_CAPS_CCID_PASSTHRU);
1494 1495

    if (strstr(str, "name \"piix3-usb-uhci\""))
1496
        qemuCapsSet(caps, QEMU_CAPS_PIIX3_USB_UHCI);
1497
    if (strstr(str, "name \"piix4-usb-uhci\""))
1498
        qemuCapsSet(caps, QEMU_CAPS_PIIX4_USB_UHCI);
1499
    if (strstr(str, "name \"usb-ehci\""))
1500
        qemuCapsSet(caps, QEMU_CAPS_USB_EHCI);
1501
    if (strstr(str, "name \"ich9-usb-ehci1\""))
1502
        qemuCapsSet(caps, QEMU_CAPS_ICH9_USB_EHCI1);
1503
    if (strstr(str, "name \"vt82c686b-usb-uhci\""))
1504
        qemuCapsSet(caps, QEMU_CAPS_VT82C686B_USB_UHCI);
1505
    if (strstr(str, "name \"pci-ohci\""))
1506
        qemuCapsSet(caps, QEMU_CAPS_PCI_OHCI);
G
Gerd Hoffmann 已提交
1507
    if (strstr(str, "name \"nec-usb-xhci\""))
1508
        qemuCapsSet(caps, QEMU_CAPS_NEC_USB_XHCI);
1509
    if (strstr(str, "name \"usb-redir\""))
1510
        qemuCapsSet(caps, QEMU_CAPS_USB_REDIR);
1511
    if (strstr(str, "usb-redir.filter"))
1512
        qemuCapsSet(caps, QEMU_CAPS_USB_REDIR_FILTER);
M
Marc-André Lureau 已提交
1513
    if (strstr(str, "name \"usb-hub\""))
1514
        qemuCapsSet(caps, QEMU_CAPS_USB_HUB);
J
Jim Fehlig 已提交
1515
    if (strstr(str, "name \"ich9-ahci\""))
1516
        qemuCapsSet(caps, QEMU_CAPS_ICH9_AHCI);
1517 1518 1519
    if (strstr(str, "name \"virtio-blk-s390\"") ||
        strstr(str, "name \"virtio-net-s390\"") ||
        strstr(str, "name \"virtio-serial-s390\""))
1520
        qemuCapsSet(caps, QEMU_CAPS_VIRTIO_S390);
1521

1522
    if (strstr(str, "name \"lsi53c895a\""))
1523
        qemuCapsSet(caps, QEMU_CAPS_SCSI_LSI);
1524
    if (strstr(str, "name \"virtio-scsi-pci\""))
1525
        qemuCapsSet(caps, QEMU_CAPS_VIRTIO_SCSI_PCI);
1526

1527
    /* Prefer -chardev spicevmc (detected earlier) over -device spicevmc */
1528
    if (!qemuCapsGet(caps, QEMU_CAPS_CHARDEV_SPICEVMC) &&
1529
        strstr(str, "name \"spicevmc\""))
1530
        qemuCapsSet(caps, QEMU_CAPS_DEVICE_SPICEVMC);
1531 1532

    /* Features of given devices. */
E
Eric Blake 已提交
1533
    if (strstr(str, "pci-assign.configfd"))
1534
        qemuCapsSet(caps, QEMU_CAPS_PCI_CONFIGFD);
1535
    if (strstr(str, "virtio-blk-pci.multifunction"))
1536
        qemuCapsSet(caps, QEMU_CAPS_PCI_MULTIFUNCTION);
1537
    if (strstr(str, "virtio-blk-pci.bootindex")) {
1538
        qemuCapsSet(caps, QEMU_CAPS_BOOTINDEX);
1539
        if (strstr(str, "pci-assign.bootindex"))
1540
            qemuCapsSet(caps, QEMU_CAPS_PCI_BOOTINDEX);
1541
    }
1542
    if (strstr(str, "virtio-net-pci.tx="))
1543
        qemuCapsSet(caps, QEMU_CAPS_VIRTIO_TX_ALG);
1544
    if (strstr(str, "name \"qxl-vga\""))
1545
        qemuCapsSet(caps, QEMU_CAPS_DEVICE_QXL_VGA);
1546
    if (strstr(str, "virtio-blk-pci.ioeventfd"))
1547
        qemuCapsSet(caps, QEMU_CAPS_VIRTIO_IOEVENTFD);
M
Michal Privoznik 已提交
1548
    if (strstr(str, "name \"sga\""))
1549
        qemuCapsSet(caps, QEMU_CAPS_SGA);
1550
    if (strstr(str, "virtio-blk-pci.event_idx"))
1551
        qemuCapsSet(caps, QEMU_CAPS_VIRTIO_BLK_EVENT_IDX);
1552
    if (strstr(str, "virtio-net-pci.event_idx"))
1553
        qemuCapsSet(caps, QEMU_CAPS_VIRTIO_NET_EVENT_IDX);
1554
    if (strstr(str, "virtio-blk-pci.scsi"))
1555
        qemuCapsSet(caps, QEMU_CAPS_VIRTIO_BLK_SCSI);
1556
    if (strstr(str, "scsi-disk.channel"))
1557
        qemuCapsSet(caps, QEMU_CAPS_SCSI_DISK_CHANNEL);
1558
    if (strstr(str, "scsi-block"))
1559
        qemuCapsSet(caps, QEMU_CAPS_SCSI_BLOCK);
1560
    if (strstr(str, "scsi-cd"))
1561
        qemuCapsSet(caps, QEMU_CAPS_SCSI_CD);
1562
    if (strstr(str, "ide-cd"))
1563
        qemuCapsSet(caps, QEMU_CAPS_IDE_CD);
1564 1565 1566 1567 1568 1569 1570 1571 1572
    /*
     * the iolimit detection is not really straight forward:
     * in qemu this is a capability of the block layer, if
     * present any of -device scsi-disk, virtio-blk-*, ...
     * will offer to specify logical and physical block size
     * and other properties...
     */
    if (strstr(str, ".logical_block_size") &&
        strstr(str, ".physical_block_size"))
1573
        qemuCapsSet(caps, QEMU_CAPS_BLOCKIO);
1574
    if (strstr(str, "PIIX4_PM.disable_s3="))
1575
        qemuCapsSet(caps, QEMU_CAPS_DISABLE_S3);
1576
    if (strstr(str, "PIIX4_PM.disable_s4="))
1577
        qemuCapsSet(caps, QEMU_CAPS_DISABLE_S4);
E
Eric Blake 已提交
1578 1579

    return 0;
1580 1581
}

1582 1583 1584
int qemuCapsExtractVersionInfo(const char *qemu,
                               const char *arch,
                               bool check_yajl,
1585
                               unsigned int *retversion,
1586
                               qemuCapsPtr *retcaps)
1587
{
1588
    int ret = -1;
1589
    unsigned int version, is_kvm, kvm_version;
1590
    qemuCapsPtr caps = NULL;
1591 1592
    char *help = NULL;
    virCommandPtr cmd;
1593

1594 1595
    if (retcaps)
        *retcaps = NULL;
1596 1597 1598 1599 1600 1601 1602
    if (retversion)
        *retversion = 0;

    /* 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.
     */
E
Eric Blake 已提交
1603
    if (!virFileIsExecutable(qemu)) {
1604 1605 1606 1607
        virReportSystemError(errno, _("Cannot find QEMU binary %s"), qemu);
        return -1;
    }

1608 1609
    cmd = qemuCapsProbeCommand(qemu, NULL);
    virCommandAddArgList(cmd, "-help", NULL);
1610
    virCommandSetOutputBuffer(cmd, &help);
1611

1612 1613
    if (virCommandRun(cmd, NULL) < 0)
        goto cleanup;
1614

1615 1616
    if (!(caps = qemuCapsNew()) ||
        qemuCapsParseHelpStr(qemu, help, caps,
1617 1618
                             &version, &is_kvm, &kvm_version,
                             check_yajl) == -1)
1619
        goto cleanup;
1620

1621 1622 1623
    /* Currently only x86_64 and i686 support PCI-multibus. */
    if (STREQLEN(arch, "x86_64", 6) ||
        STREQLEN(arch, "i686", 4)) {
1624
        qemuCapsSet(caps, QEMU_CAPS_PCI_MULTIBUS);
1625 1626
    }

1627 1628 1629
    /* S390 and probably other archs do not support no-acpi -
       maybe the qemu option parsing should be re-thought. */
    if (STRPREFIX(arch, "s390"))
1630
        qemuCapsClear(caps, QEMU_CAPS_NO_ACPI);
1631

1632
    /* qemuCapsExtractDeviceStr will only set additional caps if qemu
1633
     * understands the 0.13.0+ notion of "-device driver,".  */
1634
    if (qemuCapsGet(caps, QEMU_CAPS_DEVICE) &&
1635
        strstr(help, "-device driver,?") &&
1636
        qemuCapsExtractDeviceStr(qemu, caps) < 0)
E
Eric Blake 已提交
1637
        goto cleanup;
1638 1639 1640

    if (retversion)
        *retversion = version;
1641 1642 1643
    if (retcaps) {
        *retcaps = caps;
        caps = NULL;
1644
    }
1645 1646 1647

    ret = 0;

1648
cleanup:
1649
    VIR_FREE(help);
1650
    virCommandFree(cmd);
1651
    virObjectUnref(caps);
1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684

    return ret;
}

static void
uname_normalize (struct utsname *ut)
{
    uname(ut);

    /* Map i386, i486, i586 to i686.  */
    if (ut->machine[0] == 'i' &&
        ut->machine[1] != '\0' &&
        ut->machine[2] == '8' &&
        ut->machine[3] == '6' &&
        ut->machine[4] == '\0')
        ut->machine[1] = '6';
}

int qemuCapsExtractVersion(virCapsPtr caps,
                           unsigned int *version)
{
    const char *binary;
    struct stat sb;
    struct utsname ut;

    if (*version > 0)
        return 0;

    uname_normalize(&ut);
    if ((binary = virCapabilitiesDefaultGuestEmulator(caps,
                                                      "hvm",
                                                      ut.machine,
                                                      "qemu")) == NULL) {
1685 1686
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Cannot find suitable emulator for %s"), ut.machine);
1687 1688 1689 1690 1691 1692 1693 1694 1695
        return -1;
    }

    if (stat(binary, &sb) < 0) {
        virReportSystemError(errno,
                             _("Cannot find QEMU binary %s"), binary);
        return -1;
    }

1696 1697
    if (qemuCapsExtractVersionInfo(binary, ut.machine, false,
                                   version, NULL) < 0) {
1698 1699 1700 1701 1702
        return -1;
    }

    return 0;
}
1703 1704


1705 1706 1707


qemuCapsPtr
1708 1709
qemuCapsNew(void)
{
1710
    qemuCapsPtr caps;
1711

1712 1713 1714 1715 1716 1717 1718 1719
    if (qemuCapsInitialize() < 0)
        return NULL;

    if (!(caps = virObjectNew(qemuCapsClass)))
        return NULL;

    if (!(caps->flags = virBitmapAlloc(QEMU_CAPS_LAST)))
        goto no_memory;
1720 1721

    return caps;
1722 1723 1724 1725 1726

no_memory:
    virReportOOMError();
    virObjectUnref(caps);
    return NULL;
1727 1728 1729
}


1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776
qemuCapsPtr qemuCapsNewCopy(qemuCapsPtr caps)
{
    qemuCapsPtr ret = qemuCapsNew();
    size_t i;

    if (!ret)
        return NULL;

    virBitmapCopy(ret->flags, caps->flags);

    ret->version = caps->version;
    ret->kvmVersion = caps->kvmVersion;

    if (caps->arch &&
        !(ret->arch = strdup(caps->arch)))
        goto no_memory;

    if (VIR_ALLOC_N(ret->cpuDefinitions, caps->ncpuDefinitions) < 0)
        goto no_memory;
    ret->ncpuDefinitions = caps->ncpuDefinitions;
    for (i = 0 ; i < caps->ncpuDefinitions ; i++) {
        if (!(ret->cpuDefinitions[i] = strdup(caps->cpuDefinitions[i])))
            goto no_memory;
    }

    if (VIR_ALLOC_N(ret->machineTypes, caps->nmachineTypes) < 0)
        goto no_memory;
    if (VIR_ALLOC_N(ret->machineAliases, caps->nmachineTypes) < 0)
        goto no_memory;
    ret->nmachineTypes = caps->nmachineTypes;
    for (i = 0 ; i < caps->nmachineTypes ; i++) {
        if (!(ret->machineTypes[i] = strdup(caps->machineTypes[i])))
            goto no_memory;
        if (caps->machineAliases[i] &&
            !(ret->machineAliases[i] = strdup(caps->machineAliases[i])))
            goto no_memory;
    }

    return ret;

no_memory:
    virReportOOMError();
    virObjectUnref(ret);
    return NULL;
}


1777 1778 1779
void qemuCapsDispose(void *obj)
{
    qemuCapsPtr caps = obj;
1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794
    size_t i;

    VIR_FREE(caps->arch);

    for (i = 0 ; i < caps->nmachineTypes ; i++) {
        VIR_FREE(caps->machineTypes[i]);
        VIR_FREE(caps->machineAliases[i]);
    }
    VIR_FREE(caps->machineTypes);
    VIR_FREE(caps->machineAliases);

    for (i = 0 ; i < caps->ncpuDefinitions ; i++) {
        VIR_FREE(caps->cpuDefinitions[i]);
    }
    VIR_FREE(caps->cpuDefinitions);
1795 1796 1797 1798

    virBitmapFree(caps->flags);
}

1799
void
1800
qemuCapsSet(qemuCapsPtr caps,
1801 1802
            enum qemuCapsFlags flag)
{
1803
    ignore_value(virBitmapSetBit(caps->flags, flag));
1804 1805 1806 1807
}


void
1808
qemuCapsSetList(qemuCapsPtr caps, ...)
1809 1810 1811 1812 1813 1814
{
    va_list list;
    int flag;

    va_start(list, caps);
    while ((flag = va_arg(list, int)) < QEMU_CAPS_LAST)
1815
        ignore_value(virBitmapSetBit(caps->flags, flag));
1816
    va_end(list);
1817 1818 1819 1820
}


void
1821
qemuCapsClear(qemuCapsPtr caps,
1822 1823
              enum qemuCapsFlags flag)
{
1824 1825 1826 1827 1828 1829 1830
    ignore_value(virBitmapClearBit(caps->flags, flag));
}


char *qemuCapsFlagsString(qemuCapsPtr caps)
{
    return virBitmapString(caps->flags);
1831 1832 1833 1834
}


bool
1835
qemuCapsGet(qemuCapsPtr caps,
1836 1837
            enum qemuCapsFlags flag)
{
1838 1839
    bool b;

1840
    if (!caps || virBitmapGetBit(caps->flags, flag, &b) < 0)
1841 1842 1843
        return false;
    else
        return b;
1844
}
1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895


const char *qemuCapsGetArch(qemuCapsPtr caps)
{
    return caps->arch;
}


unsigned int qemuCapsGetVersion(qemuCapsPtr caps)
{
    return caps->version;
}


unsigned int qemuCapsGetKVMVersion(qemuCapsPtr caps)
{
    return caps->kvmVersion;
}


size_t qemuCapsGetCPUDefinitions(qemuCapsPtr caps,
                                 char ***names)
{
    *names = caps->cpuDefinitions;
    return caps->ncpuDefinitions;
}


size_t qemuCapsGetMachineTypes(qemuCapsPtr caps,
                               char ***names)
{
    *names = caps->machineTypes;
    return caps->nmachineTypes;
}


const char *qemuCapsGetCanonicalMachine(qemuCapsPtr caps,
                                        const char *name)

{
    size_t i;

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

    return name;
}