qemu_capabilities.c 48.0 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 18 19 20 21 22 23 24 25 26 27 28 29 30
 * 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
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307  USA
 *
 * 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",
M
Marc-André Lureau 已提交
168

169 170
    );

171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206
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 },
207 208
    {  "microblaze", 32, NULL, "qemu-system-microblaze",   NULL, NULL, 0 },
    {  "microblazeel", 32, NULL, "qemu-system-microblazeel",   NULL, NULL, 0 },
209 210 211 212
    {  "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
213
    {  "ppc64",    64, NULL, "qemu-system-ppc64",    NULL, NULL, 0 },
214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296
    {  "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 },
};


/* Format is:
 * <machine> <desc> [(default)|(alias of <canonical>)]
 */
static int
qemuCapsParseMachineTypesStr(const char *output,
                             virCapsGuestMachinePtr **machines,
                             int *nmachines)
{
    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,
297
                          virBitmapPtr qemuCaps,
298 299 300 301
                          virCapsGuestMachinePtr **machines,
                          int *nmachines)
{
    char *output;
302 303
    int ret = -1;
    virCommandPtr cmd;
304
    int status;
305

306 307 308 309
    /* 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
310
    if (!virFileIsExecutable(binary)) {
311 312 313 314
        virReportSystemError(errno, _("Cannot find QEMU binary %s"), binary);
        return -1;
    }

315 316
    cmd = qemuCapsProbeCommand(binary, qemuCaps);
    virCommandAddArgList(cmd, "-M", "?", NULL);
317
    virCommandSetOutputBuffer(cmd, &output);
318

319 320
    /* Ignore failure from older qemu that did not understand '-M ?'.  */
    if (virCommandRun(cmd, &status) < 0)
321 322 323
        goto cleanup;

    if (qemuCapsParseMachineTypesStr(output, machines, nmachines) < 0)
324
        goto cleanup;
325 326 327 328

    ret = 0;

cleanup:
329 330
    VIR_FREE(output);
    virCommandFree(cmd);
331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469

    return ret;
}

static int
qemuCapsGetOldMachinesFromInfo(virCapsGuestDomainInfoPtr info,
                               const char *emulator,
                               time_t emulator_mtime,
                               virCapsGuestMachinePtr **machines,
                               int *nmachines)
{
    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,
                       int *nmachines)
{
    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,
                       unsigned int *retcount,
                       const char ***retcpus);

/* Format:
 *      <arch> <model>
 * qemu-0.13 encloses some model names in []:
 *      <arch> [<model>]
 */
static int
qemuCapsParseX86Models(const char *output,
                       unsigned int *retcount,
                       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;

470 471
            if (VIR_REALLOC_N(cpus, count + 1) < 0) {
                virReportOOMError();
472
                goto error;
473
            }
474 475 476 477 478 479 480 481 482 483 484

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

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

485 486
            if (!(cpus[count] = strndup(p, len))) {
                virReportOOMError();
487
                goto error;
488
            }
489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509
        }
        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
510 511 512 513 514 515 516 517 518 519 520 521
/* ppc64 parser.
 * Format : PowerPC <machine> <description>
 */
static int
qemuCapsParsePPCModels(const char *output,
                       unsigned int *retcount,
                       const char ***retcpus)
{
    const char *p = output;
    const char *next;
    unsigned int count = 0;
    const char **cpus = NULL;
522
    int i, ret = -1;
P
Prerna Saxena 已提交
523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548

    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;

549
            if (VIR_REALLOC_N(cpus, count + 1) < 0) {
P
Prerna Saxena 已提交
550
                virReportOOMError();
551 552
                goto cleanup;
            }
P
Prerna Saxena 已提交
553 554 555

            len = t - p - 1;

556
            if (!(cpus[count] = strndup(p, len))) {
P
Prerna Saxena 已提交
557
                virReportOOMError();
558 559
                goto cleanup;
            }
P
Prerna Saxena 已提交
560 561 562 563 564 565
        }
        count++;
    } while ((p = next));

    if (retcount)
        *retcount = count;
566
    if (retcpus) {
P
Prerna Saxena 已提交
567
        *retcpus = cpus;
568 569 570
        cpus = NULL;
    }
    ret = 0;
P
Prerna Saxena 已提交
571

572
cleanup:
P
Prerna Saxena 已提交
573 574 575
    if (cpus) {
        for (i = 0; i < count; i++)
            VIR_FREE(cpus[i]);
576
        VIR_FREE(cpus);
P
Prerna Saxena 已提交
577
    }
578
    return ret;
P
Prerna Saxena 已提交
579
}
580 581 582

int
qemuCapsProbeCPUModels(const char *qemu,
583
                       virBitmapPtr qemuCaps,
584 585 586 587 588 589 590
                       const char *arch,
                       unsigned int *count,
                       const char ***cpus)
{
    char *output = NULL;
    int ret = -1;
    qemuCapsParseCPUModels parse;
591
    virCommandPtr cmd;
592 593 594 595 596 597 598 599

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

    if (STREQ(arch, "i686") || STREQ(arch, "x86_64"))
        parse = qemuCapsParseX86Models;
P
Prerna Saxena 已提交
600 601
    else if (STREQ(arch, "ppc64"))
        parse = qemuCapsParsePPCModels;
602 603 604 605 606
    else {
        VIR_DEBUG("don't know how to parse %s CPU models", arch);
        return 0;
    }

607 608
    cmd = qemuCapsProbeCommand(qemu, qemuCaps);
    virCommandAddArgList(cmd, "-cpu", "?", NULL);
609
    virCommandSetOutputBuffer(cmd, &output);
610

611
    if (virCommandRun(cmd, NULL) < 0)
612 613
        goto cleanup;

614
    if (parse(output, count, cpus) < 0)
615 616 617 618 619 620
        goto cleanup;

    ret = 0;

cleanup:
    VIR_FREE(output);
621
    virCommandFree(cmd);
622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644

    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;
    int nmachines = 0;
    struct stat st;
    unsigned int ncpus;
645
    virBitmapPtr qemuCaps = NULL;
646 647 648 649 650 651 652
    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
653
    if (binary == NULL || !virFileIsExecutable(binary)) {
654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673
        VIR_FREE(binary);
        binary = virFindFileInPath(info->altbinary);
    }

    /* Can use acceleration for KVM/KQEMU if
     *  - 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"))) {
        if (access("/dev/kvm", F_OK) == 0) {
            const char *const kvmbins[] = { "/usr/libexec/qemu-kvm", /* RHEL */
                                            "qemu-kvm", /* Fedora */
                                            "kvm" }; /* Upstream .spec */

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

E
Eric Blake 已提交
674
                if (!kvmbin)
675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691
                    continue;

                haskvm = 1;
                if (!binary)
                    binary = kvmbin;

                break;
            }
        }

        if (access("/dev/kqemu", F_OK) == 0)
            haskqemu = 1;
    }

    if (!binary)
        return 0;

J
Jiri Denemark 已提交
692 693 694 695 696 697
    /* Ignore binary if extracting version info fails */
    if (qemuCapsExtractVersionInfo(binary, info->arch, NULL, &qemuCaps) < 0) {
        ret = 0;
        goto cleanup;
    }

698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
    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 &&
735 736
            qemuCapsProbeMachineTypes(binary, qemuCaps,
                                      &machines, &nmachines) < 0)
737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757
            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 &&
758
        qemuCapsProbeCPUModels(binary, NULL, info->arch, &ncpus, NULL) == 0 &&
759 760 761 762
        ncpus > 0 &&
        !virCapabilitiesAddGuestFeature(guest, "cpuselection", 1, 0))
        goto error;

J
Jiri Denemark 已提交
763 764
    if (qemuCapsGet(qemuCaps, QEMU_CAPS_BOOTINDEX) &&
        !virCapabilitiesAddGuestFeature(guest, "deviceboot", 1, 0))
765 766
        goto error;

767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803
    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;

            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;
            }

            if (!STREQ(binary, kvmbin)) {
                int probe = 1;
                if (old_caps && binary_mtime)
                    probe = !qemuCapsGetOldMachines("hvm", info->arch, info->wordsize,
                                                    kvmbin, binary_mtime,
                                                    old_caps, &machines, &nmachines);
                if (probe &&
804 805
                    qemuCapsProbeMachineTypes(kvmbin, qemuCaps,
                                              &machines, &nmachines) < 0)
806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852
                    goto error;
            }

            if ((dom = virCapabilitiesAddGuestDomain(guest,
                                                     "kvm",
                                                     kvmbin,
                                                     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:
    if (binary == kvmbin) {
        /* don't double free */
        VIR_FREE(binary);
    } else {
        VIR_FREE(binary);
        VIR_FREE(kvmbin);
    }
853
    qemuCapsFree(qemuCaps);
854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908

    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;
}


909 910 911 912 913 914
static int qemuDefaultConsoleType(const char *ostype ATTRIBUTE_UNUSED)
{
    return VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_SERIAL;
}


915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937
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);
938
        VIR_WARN("Failed to query host NUMA topology, disabling NUMA capabilities");
939 940 941 942
    }

    if (old_caps == NULL || old_caps->host.cpu == NULL) {
        if (qemuCapsInitCPU(caps, utsname.machine) < 0)
943
            VIR_WARN("Failed to get host CPU");
944 945 946 947 948 949
    }
    else {
        caps->host.cpu = old_caps->host.cpu;
        old_caps->host.cpu = NULL;
    }

950 951
    /* Add the power management features of the host */

952
    if (virNodeSuspendGetTargetMask(&caps->host.powerMgmt) < 0)
953 954
        VIR_WARN("Failed to get host power management capabilities");

955 956 957 958 959 960 961 962 963 964 965 966 967
    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
968
    if (xenner != NULL && virFileIsExecutable(xenner) == 0 &&
969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986
        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);

987
    caps->defaultConsoleTargetType = qemuDefaultConsoleType;
988 989 990 991 992 993 994 995 996 997

    return caps;

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


998
static int
999 1000 1001
qemuCapsComputeCmdFlags(const char *help,
                        unsigned int version,
                        unsigned int is_kvm,
1002
                        unsigned int kvm_version,
1003 1004
                        virBitmapPtr flags,
                        bool check_yajl ATTRIBUTE_UNUSED)
1005 1006
{
    const char *p;
1007
    const char *fsdev;
1008 1009

    if (strstr(help, "-no-kqemu"))
1010
        qemuCapsSet(flags, QEMU_CAPS_KQEMU);
1011
    if (strstr(help, "-enable-kqemu"))
1012
        qemuCapsSet(flags, QEMU_CAPS_ENABLE_KQEMU);
1013
    if (strstr(help, "-no-kvm"))
1014
        qemuCapsSet(flags, QEMU_CAPS_KVM);
1015
    if (strstr(help, "-enable-kvm"))
1016
        qemuCapsSet(flags, QEMU_CAPS_ENABLE_KVM);
1017
    if (strstr(help, "-no-reboot"))
1018
        qemuCapsSet(flags, QEMU_CAPS_NO_REBOOT);
1019
    if (strstr(help, "-name")) {
1020
        qemuCapsSet(flags, QEMU_CAPS_NAME);
1021
        if (strstr(help, ",process="))
1022
            qemuCapsSet(flags, QEMU_CAPS_NAME_PROCESS);
1023 1024
    }
    if (strstr(help, "-uuid"))
1025
        qemuCapsSet(flags, QEMU_CAPS_UUID);
1026
    if (strstr(help, "-xen-domid"))
1027
        qemuCapsSet(flags, QEMU_CAPS_XEN_DOMID);
1028
    else if (strstr(help, "-domid"))
1029
        qemuCapsSet(flags, QEMU_CAPS_DOMID);
1030
    if (strstr(help, "-drive")) {
1031 1032
        const char *cache = strstr(help, "cache=");

1033
        qemuCapsSet(flags, QEMU_CAPS_DRIVE);
1034 1035 1036 1037
        if (cache && (p = strchr(cache, ']'))) {
            if (memmem(cache, p - cache, "on|off", sizeof("on|off") - 1) == NULL)
                qemuCapsSet(flags, QEMU_CAPS_DRIVE_CACHE_V2);
            if (memmem(cache, p - cache, "directsync", sizeof("directsync") - 1))
1038
                qemuCapsSet(flags, QEMU_CAPS_DRIVE_CACHE_DIRECTSYNC);
1039 1040
            if (memmem(cache, p - cache, "unsafe", sizeof("unsafe") - 1))
                qemuCapsSet(flags, QEMU_CAPS_DRIVE_CACHE_UNSAFE);
1041
        }
1042
        if (strstr(help, "format="))
1043
            qemuCapsSet(flags, QEMU_CAPS_DRIVE_FORMAT);
1044
        if (strstr(help, "readonly="))
1045
            qemuCapsSet(flags, QEMU_CAPS_DRIVE_READONLY);
1046
        if (strstr(help, "aio=threads|native"))
1047
            qemuCapsSet(flags, QEMU_CAPS_DRIVE_AIO);
O
Osier Yang 已提交
1048 1049
        if (strstr(help, "copy-on-read=on|off"))
            qemuCapsSet(flags, QEMU_CAPS_DRIVE_COPY_ON_READ);
1050 1051
        if (strstr(help, "bps="))
            qemuCapsSet(flags, QEMU_CAPS_DRIVE_IOTUNE);
1052 1053 1054 1055
    }
    if ((p = strstr(help, "-vga")) && !strstr(help, "-std-vga")) {
        const char *nl = strstr(p, "\n");

1056
        qemuCapsSet(flags, QEMU_CAPS_VGA);
1057 1058

        if (strstr(p, "|qxl"))
1059
            qemuCapsSet(flags, QEMU_CAPS_VGA_QXL);
1060
        if ((p = strstr(p, "|none")) && p < nl)
1061
            qemuCapsSet(flags, QEMU_CAPS_VGA_NONE);
1062 1063
    }
    if (strstr(help, "-spice"))
1064
        qemuCapsSet(flags, QEMU_CAPS_SPICE);
1065
    if (strstr(help, "boot=on"))
1066
        qemuCapsSet(flags, QEMU_CAPS_DRIVE_BOOT);
1067
    if (strstr(help, "serial=s"))
1068
        qemuCapsSet(flags, QEMU_CAPS_DRIVE_SERIAL);
1069
    if (strstr(help, "-pcidevice"))
1070
        qemuCapsSet(flags, QEMU_CAPS_PCIDEVICE);
1071
    if (strstr(help, "-mem-path"))
1072
        qemuCapsSet(flags, QEMU_CAPS_MEM_PATH);
1073
    if (strstr(help, "-chardev")) {
1074
        qemuCapsSet(flags, QEMU_CAPS_CHARDEV);
1075
        if (strstr(help, "-chardev spicevmc"))
1076
            qemuCapsSet(flags, QEMU_CAPS_CHARDEV_SPICEVMC);
1077
    }
1078
    if (strstr(help, "-balloon"))
1079
        qemuCapsSet(flags, QEMU_CAPS_BALLOON);
1080
    if (strstr(help, "-device")) {
1081
        qemuCapsSet(flags, QEMU_CAPS_DEVICE);
1082 1083 1084 1085
        /*
         * When -device was introduced, qemu already supported drive's
         * readonly option but didn't advertise that.
         */
1086
        qemuCapsSet(flags, QEMU_CAPS_DRIVE_READONLY);
1087 1088
    }
    if (strstr(help, "-nodefconfig"))
1089
        qemuCapsSet(flags, QEMU_CAPS_NODEFCONFIG);
1090 1091
    if (strstr(help, "-no-user-config"))
        qemuCapsSet(flags, QEMU_CAPS_NO_USER_CONFIG);
1092 1093
    /* The trailing ' ' is important to avoid a bogus match */
    if (strstr(help, "-rtc "))
1094
        qemuCapsSet(flags, QEMU_CAPS_RTC);
1095 1096
    /* to wit */
    if (strstr(help, "-rtc-td-hack"))
1097
        qemuCapsSet(flags, QEMU_CAPS_RTC_TD_HACK);
1098
    if (strstr(help, "-no-hpet"))
1099
        qemuCapsSet(flags, QEMU_CAPS_NO_HPET);
1100 1101
    if (strstr(help, "-no-acpi"))
        qemuCapsSet(flags, QEMU_CAPS_NO_ACPI);
1102
    if (strstr(help, "-no-kvm-pit-reinjection"))
1103
        qemuCapsSet(flags, QEMU_CAPS_NO_KVM_PIT);
1104
    if (strstr(help, "-tdf"))
1105
        qemuCapsSet(flags, QEMU_CAPS_TDF);
1106
    if (strstr(help, "-enable-nesting"))
1107
        qemuCapsSet(flags, QEMU_CAPS_NESTING);
1108
    if (strstr(help, ",menu=on"))
1109
        qemuCapsSet(flags, QEMU_CAPS_BOOT_MENU);
1110
    if ((fsdev = strstr(help, "-fsdev"))) {
1111
        qemuCapsSet(flags, QEMU_CAPS_FSDEV);
1112 1113
        if (strstr(fsdev, "readonly"))
            qemuCapsSet(flags, QEMU_CAPS_FSDEV_READONLY);
1114 1115
        if (strstr(fsdev, "writeout"))
            qemuCapsSet(flags, QEMU_CAPS_FSDEV_WRITEOUT);
1116
    }
1117
    if (strstr(help, "-smbios type"))
1118
        qemuCapsSet(flags, QEMU_CAPS_SMBIOS_TYPE);
1119 1120 1121 1122

    if (strstr(help, "-netdev")) {
        /* Disable -netdev on 0.12 since although it exists,
         * the corresponding netdev_add/remove monitor commands
1123 1124
         * do not, and we need them to be able to do hotplug.
         * But see below about RHEL build. */
1125
        if (version >= 13000)
1126
            qemuCapsSet(flags, QEMU_CAPS_NETDEV);
1127 1128 1129
    }

    if (strstr(help, "-sdl"))
1130
        qemuCapsSet(flags, QEMU_CAPS_SDL);
1131 1132 1133
    if (strstr(help, "cores=") &&
        strstr(help, "threads=") &&
        strstr(help, "sockets="))
1134
        qemuCapsSet(flags, QEMU_CAPS_SMP_TOPOLOGY);
1135 1136

    if (version >= 9000)
1137
        qemuCapsSet(flags, QEMU_CAPS_VNC_COLON);
1138 1139

    if (is_kvm && (version >= 10000 || kvm_version >= 74))
1140
        qemuCapsSet(flags, QEMU_CAPS_VNET_HDR);
1141

1142
    if (strstr(help, ",vhost=")) {
1143
        qemuCapsSet(flags, QEMU_CAPS_VHOST_NET);
1144 1145
    }

1146 1147
    /* 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
1148
     * 0.14.* and 0.15.0)
1149
     */
1150
    if (strstr(help, "-no-shutdown") && (version < 14000 || version > 15000))
1151 1152
        qemuCapsSet(flags, QEMU_CAPS_NO_SHUTDOWN);

1153 1154 1155 1156
    /*
     * Handling of -incoming arg with varying features
     *  -incoming tcp    (kvm >= 79, qemu >= 0.10.0)
     *  -incoming exec   (kvm >= 80, qemu >= 0.10.0)
1157 1158
     *  -incoming unix   (qemu >= 0.12.0)
     *  -incoming fd     (qemu >= 0.12.0)
1159 1160 1161 1162 1163 1164 1165
     *  -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) {
1166 1167
        qemuCapsSet(flags, QEMU_CAPS_MIGRATE_QEMU_TCP);
        qemuCapsSet(flags, QEMU_CAPS_MIGRATE_QEMU_EXEC);
1168
        if (version >= 12000) {
1169 1170
            qemuCapsSet(flags, QEMU_CAPS_MIGRATE_QEMU_UNIX);
            qemuCapsSet(flags, QEMU_CAPS_MIGRATE_QEMU_FD);
1171
        }
1172
    } else if (kvm_version >= 79) {
1173
        qemuCapsSet(flags, QEMU_CAPS_MIGRATE_QEMU_TCP);
1174
        if (kvm_version >= 80)
1175
            qemuCapsSet(flags, QEMU_CAPS_MIGRATE_QEMU_EXEC);
1176
    } else if (kvm_version > 0) {
1177
        qemuCapsSet(flags, QEMU_CAPS_MIGRATE_KVM_STDIO);
1178 1179 1180
    }

    if (version >= 10000)
1181
        qemuCapsSet(flags, QEMU_CAPS_0_10);
1182

1183 1184 1185
    if (version >= 11000)
        qemuCapsSet(flags, QEMU_CAPS_VIRTIO_BLK_SG_IO);

1186 1187 1188
    /* 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
1189 1190 1191 1192 1193
     * 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.
1194
     */
E
Eric Blake 已提交
1195
#if HAVE_YAJL
1196
    if (version >= 13000) {
1197
        qemuCapsSet(flags, QEMU_CAPS_MONITOR_JSON);
1198 1199 1200 1201 1202
    } else if (version >= 12000 &&
               strstr(help, "libvirt")) {
        qemuCapsSet(flags, QEMU_CAPS_MONITOR_JSON);
        qemuCapsSet(flags, QEMU_CAPS_NETDEV);
    }
1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
#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
     * dependency, so distros won't hit this).  */
    if (version >= 15000 ||
        (version >= 12000 && strstr(help, "libvirt"))) {
        if (check_yajl) {
            qemuReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                            _("this qemu binary requires libvirt to be "
                              "compiled with yajl"));
            return -1;
        }
        qemuCapsSet(flags, QEMU_CAPS_NETDEV);
    }
E
Eric Blake 已提交
1221
#endif
1222 1223 1224

    if (version >= 13000)
        qemuCapsSet(flags, QEMU_CAPS_PCI_MULTIFUNCTION);
1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237

    /* 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)
        qemuCapsSet(flags, QEMU_CAPS_PCI_ROMBAR);
1238 1239 1240

    if (version >= 11000)
        qemuCapsSet(flags, QEMU_CAPS_CPU_HOST);
1241
    return 0;
1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
}

/* 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,
1270
                         virBitmapPtr flags,
1271 1272
                         unsigned int *version,
                         unsigned int *is_kvm,
1273 1274
                         unsigned int *kvm_version,
                         bool check_yajl)
1275 1276 1277
{
    unsigned major, minor, micro;
    const char *p = help;
1278
    char *strflags;
1279

1280
    *version = *is_kvm = *kvm_version = 0;
1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297

    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
1298
    if (minor == -1)
1299 1300
        goto fail;

J
Jiri Denemark 已提交
1301 1302 1303 1304 1305 1306 1307 1308
    if (*p != '.') {
        micro = 0;
    } else {
        ++p;
        micro = virParseNumber(&p);
        if (micro == -1)
            goto fail;
    }
1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329

    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;

1330 1331 1332
    if (qemuCapsComputeCmdFlags(help, *version, *is_kvm, *kvm_version,
                                flags, check_yajl) < 0)
        goto cleanup;
1333 1334 1335 1336 1337

    strflags = virBitmapString(flags);
    VIR_DEBUG("Version %u.%u.%u, cooked version %u, flags %s",
              major, minor, micro, *version, NULLSTR(strflags));
    VIR_FREE(strflags);
1338 1339 1340 1341 1342 1343 1344 1345 1346 1347

    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');
1348 1349
    if (!p)
        p = strchr(help, '\0');
1350 1351

    qemuReportError(VIR_ERR_INTERNAL_ERROR,
1352 1353
                    _("cannot parse %s version number in '%.*s'"),
                    qemu, (int) (p - help), help);
1354

1355
cleanup:
1356 1357 1358
    return -1;
}

E
Eric Blake 已提交
1359 1360
static int
qemuCapsExtractDeviceStr(const char *qemu,
1361
                         virBitmapPtr flags)
1362
{
E
Eric Blake 已提交
1363
    char *output = NULL;
1364
    virCommandPtr cmd;
E
Eric Blake 已提交
1365
    int ret = -1;
1366

E
Eric Blake 已提交
1367 1368
    /* Cram together all device-related queries into one invocation;
     * the output format makes it possible to distinguish what we
1369 1370
     * need.  With qemu 0.13.0 and later, unrecognized '-device
     * bogus,?' cause an error in isolation, but are silently ignored
1371
     * in combination with '-device ?'.  Upstream qemu 0.12.x doesn't
1372 1373
     * understand '-device name,?', and always exits with status 1 for
     * the simpler '-device ?', so this function is really only useful
1374
     * if -help includes "device driver,?".  */
1375 1376 1377 1378 1379 1380 1381 1382
    cmd = qemuCapsProbeCommand(qemu, flags);
    virCommandAddArgList(cmd,
                         "-device", "?",
                         "-device", "pci-assign,?",
                         "-device", "virtio-blk-pci,?",
                         "-device", "virtio-net-pci,?",
                         "-device", "scsi-disk,?",
                         NULL);
1383
    /* qemu -help goes to stdout, but qemu -device ? goes to stderr.  */
E
Eric Blake 已提交
1384
    virCommandSetErrorBuffer(cmd, &output);
1385

1386
    if (virCommandRun(cmd, NULL) < 0)
1387 1388
        goto cleanup;

E
Eric Blake 已提交
1389
    ret = qemuCapsParseDeviceStr(output, flags);
1390 1391

cleanup:
E
Eric Blake 已提交
1392
    VIR_FREE(output);
1393
    virCommandFree(cmd);
E
Eric Blake 已提交
1394 1395 1396 1397 1398
    return ret;
}


int
1399
qemuCapsParseDeviceStr(const char *str, virBitmapPtr flags)
E
Eric Blake 已提交
1400
{
1401 1402
    /* Which devices exist. */
    if (strstr(str, "name \"hda-duplex\""))
1403
        qemuCapsSet(flags, QEMU_CAPS_HDA_DUPLEX);
M
Marc-André Lureau 已提交
1404 1405
    if (strstr(str, "name \"hda-micro\""))
        qemuCapsSet(flags, QEMU_CAPS_HDA_MICRO);
1406
    if (strstr(str, "name \"ccid-card-emulated\""))
1407
        qemuCapsSet(flags, QEMU_CAPS_CCID_EMULATED);
1408
    if (strstr(str, "name \"ccid-card-passthru\""))
1409
        qemuCapsSet(flags, QEMU_CAPS_CCID_PASSTHRU);
1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424

    if (strstr(str, "name \"piix3-usb-uhci\""))
        qemuCapsSet(flags, QEMU_CAPS_PIIX3_USB_UHCI);
    if (strstr(str, "name \"piix4-usb-uhci\""))
        qemuCapsSet(flags, QEMU_CAPS_PIIX4_USB_UHCI);
    if (strstr(str, "name \"usb-ehci\""))
        qemuCapsSet(flags, QEMU_CAPS_USB_EHCI);
    if (strstr(str, "name \"ich9-usb-ehci1\""))
        qemuCapsSet(flags, QEMU_CAPS_ICH9_USB_EHCI1);
    if (strstr(str, "name \"vt82c686b-usb-uhci\""))
        qemuCapsSet(flags, QEMU_CAPS_VT82C686B_USB_UHCI);
    if (strstr(str, "name \"pci-ohci\""))
        qemuCapsSet(flags, QEMU_CAPS_PCI_OHCI);
    if (strstr(str, "name \"usb-redir\""))
        qemuCapsSet(flags, QEMU_CAPS_USB_REDIR);
M
Marc-André Lureau 已提交
1425 1426
    if (strstr(str, "name \"usb-hub\""))
        qemuCapsSet(flags, QEMU_CAPS_USB_HUB);
J
Jim Fehlig 已提交
1427 1428
    if (strstr(str, "name \"ich9-ahci\""))
        qemuCapsSet(flags, QEMU_CAPS_ICH9_AHCI);
1429

1430
    /* Prefer -chardev spicevmc (detected earlier) over -device spicevmc */
1431
    if (!qemuCapsGet(flags, QEMU_CAPS_CHARDEV_SPICEVMC) &&
1432
        strstr(str, "name \"spicevmc\""))
1433
        qemuCapsSet(flags, QEMU_CAPS_DEVICE_SPICEVMC);
1434 1435

    /* Features of given devices. */
E
Eric Blake 已提交
1436
    if (strstr(str, "pci-assign.configfd"))
1437
        qemuCapsSet(flags, QEMU_CAPS_PCI_CONFIGFD);
1438 1439
    if (strstr(str, "virtio-blk-pci.multifunction"))
        qemuCapsSet(flags, QEMU_CAPS_PCI_MULTIFUNCTION);
1440
    if (strstr(str, "virtio-blk-pci.bootindex")) {
1441
        qemuCapsSet(flags, QEMU_CAPS_BOOTINDEX);
1442
        if (strstr(str, "pci-assign.bootindex"))
1443
            qemuCapsSet(flags, QEMU_CAPS_PCI_BOOTINDEX);
1444
    }
1445
    if (strstr(str, "virtio-net-pci.tx="))
1446
        qemuCapsSet(flags, QEMU_CAPS_VIRTIO_TX_ALG);
1447 1448
    if (strstr(str, "name \"qxl-vga\""))
        qemuCapsSet(flags, QEMU_CAPS_DEVICE_QXL_VGA);
1449 1450
    if (strstr(str, "virtio-blk-pci.ioeventfd"))
        qemuCapsSet(flags, QEMU_CAPS_VIRTIO_IOEVENTFD);
M
Michal Privoznik 已提交
1451 1452
    if (strstr(str, "name \"sga\""))
        qemuCapsSet(flags, QEMU_CAPS_SGA);
1453 1454 1455 1456
    if (strstr(str, "virtio-blk-pci.event_idx"))
        qemuCapsSet(flags, QEMU_CAPS_VIRTIO_BLK_EVENT_IDX);
    if (strstr(str, "virtio-net-pci.event_idx"))
        qemuCapsSet(flags, QEMU_CAPS_VIRTIO_NET_EVENT_IDX);
1457 1458
    if (strstr(str, "virtio-blk-pci.scsi"))
        qemuCapsSet(flags, QEMU_CAPS_VIRTIO_BLK_SCSI);
1459 1460
    if (strstr(str, "scsi-disk.channel"))
        qemuCapsSet(flags, QEMU_CAPS_SCSI_DISK_CHANNEL);
1461 1462
    if (strstr(str, "scsi-block"))
        qemuCapsSet(flags, QEMU_CAPS_SCSI_BLOCK);
1463 1464
    if (strstr(str, "scsi-cd"))
        qemuCapsSet(flags, QEMU_CAPS_SCSI_CD);
1465 1466
    if (strstr(str, "ide-cd"))
        qemuCapsSet(flags, QEMU_CAPS_IDE_CD);
E
Eric Blake 已提交
1467 1468

    return 0;
1469 1470
}

1471
int qemuCapsExtractVersionInfo(const char *qemu, const char *arch,
1472
                               unsigned int *retversion,
1473
                               virBitmapPtr *retflags)
1474
{
1475
    int ret = -1;
1476
    unsigned int version, is_kvm, kvm_version;
1477
    virBitmapPtr flags = NULL;
1478 1479
    char *help = NULL;
    virCommandPtr cmd;
1480 1481

    if (retflags)
1482
        *retflags = NULL;
1483 1484 1485 1486 1487 1488 1489
    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
1490
    if (!virFileIsExecutable(qemu)) {
1491 1492 1493 1494
        virReportSystemError(errno, _("Cannot find QEMU binary %s"), qemu);
        return -1;
    }

1495 1496
    cmd = qemuCapsProbeCommand(qemu, NULL);
    virCommandAddArgList(cmd, "-help", NULL);
1497
    virCommandSetOutputBuffer(cmd, &help);
1498

1499 1500
    if (virCommandRun(cmd, NULL) < 0)
        goto cleanup;
1501

1502 1503
    if (!(flags = qemuCapsNew()) ||
        qemuCapsParseHelpStr(qemu, help, flags,
1504
                             &version, &is_kvm, &kvm_version, true) == -1)
1505
        goto cleanup;
1506

1507 1508 1509
    /* Currently only x86_64 and i686 support PCI-multibus. */
    if (STREQLEN(arch, "x86_64", 6) ||
        STREQLEN(arch, "i686", 4)) {
1510
        qemuCapsSet(flags, QEMU_CAPS_PCI_MULTIBUS);
1511 1512
    }

1513 1514
    /* qemuCapsExtractDeviceStr will only set additional flags if qemu
     * understands the 0.13.0+ notion of "-device driver,".  */
1515
    if (qemuCapsGet(flags, QEMU_CAPS_DEVICE) &&
1516
        strstr(help, "-device driver,?") &&
1517
        qemuCapsExtractDeviceStr(qemu, flags) < 0)
E
Eric Blake 已提交
1518
        goto cleanup;
1519 1520 1521

    if (retversion)
        *retversion = version;
1522
    if (retflags) {
1523
        *retflags = flags;
1524 1525
        flags = NULL;
    }
1526 1527 1528

    ret = 0;

1529
cleanup:
1530
    VIR_FREE(help);
1531
    virCommandFree(cmd);
1532
    qemuCapsFree(flags);
1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576

    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) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        _("Cannot find suitable emulator for %s"), ut.machine);
        return -1;
    }

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

1577
    if (qemuCapsExtractVersionInfo(binary, ut.machine, version, NULL) < 0) {
1578 1579 1580 1581 1582
        return -1;
    }

    return 0;
}
1583 1584


1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596
virBitmapPtr
qemuCapsNew(void)
{
    virBitmapPtr caps;

    if (!(caps = virBitmapAlloc(QEMU_CAPS_LAST)))
        virReportOOMError();

    return caps;
}


1597
void
1598
qemuCapsSet(virBitmapPtr caps,
1599 1600
            enum qemuCapsFlags flag)
{
1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614
    ignore_value(virBitmapSetBit(caps, flag));
}


void
qemuCapsSetList(virBitmapPtr caps, ...)
{
    va_list list;
    int flag;

    va_start(list, caps);
    while ((flag = va_arg(list, int)) < QEMU_CAPS_LAST)
        ignore_value(virBitmapSetBit(caps, flag));
    va_end(list);
1615 1616 1617 1618
}


void
1619
qemuCapsClear(virBitmapPtr caps,
1620 1621
              enum qemuCapsFlags flag)
{
1622
    ignore_value(virBitmapClearBit(caps, flag));
1623 1624 1625 1626
}


bool
1627
qemuCapsGet(virBitmapPtr caps,
1628 1629
            enum qemuCapsFlags flag)
{
1630 1631 1632 1633 1634 1635
    bool b;

    if (!caps || virBitmapGetBit(caps, flag, &b) < 0)
        return false;
    else
        return b;
1636
}
1637 1638 1639 1640 1641 1642 1643 1644 1645


virCommandPtr
qemuCapsProbeCommand(const char *qemu,
                     virBitmapPtr qemuCaps)
{
    virCommandPtr cmd = virCommandNew(qemu);

    if (qemuCaps) {
1646 1647 1648
        if (qemuCapsGet(qemuCaps, QEMU_CAPS_NO_USER_CONFIG))
            virCommandAddArg(cmd, "-no-user-config");
        else if (qemuCapsGet(qemuCaps, QEMU_CAPS_NODEFCONFIG))
1649 1650 1651 1652 1653 1654 1655 1656
            virCommandAddArg(cmd, "-nodefconfig");
    }

    virCommandAddEnvPassCommon(cmd);
    virCommandClearCaps(cmd);

    return cmd;
}