qemu_hostdev.c 28.6 KB
Newer Older
1 2 3
/*
 * qemu_hostdev.c: QEMU hostdev management
 *
4
 * Copyright (C) 2006-2007, 2009-2012 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 31
 * 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_hostdev.h"
#include "logging.h"
#include "virterror_internal.h"
#include "memory.h"
#include "pci.h"
#include "hostusb.h"
32
#include "virnetdev.h"
33

34 35
#define VIR_FROM_THIS VIR_FROM_QEMU

36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74
static pciDeviceList *
qemuGetPciHostDeviceList(virDomainHostdevDefPtr *hostdevs, int nhostdevs)
{
    pciDeviceList *list;
    int i;

    if (!(list = pciDeviceListNew()))
        return NULL;

    for (i = 0 ; i < nhostdevs ; i++) {
        virDomainHostdevDefPtr hostdev = hostdevs[i];
        pciDevice *dev;

        if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
            continue;
        if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI)
            continue;

        dev = pciGetDevice(hostdev->source.subsys.u.pci.domain,
                           hostdev->source.subsys.u.pci.bus,
                           hostdev->source.subsys.u.pci.slot,
                           hostdev->source.subsys.u.pci.function);
        if (!dev) {
            pciDeviceListFree(list);
            return NULL;
        }

        if (pciDeviceListAdd(list, dev) < 0) {
            pciFreeDevice(dev);
            pciDeviceListFree(list);
            return NULL;
        }

        pciDeviceSetManaged(dev, hostdev->managed);
    }

    return list;
}

75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117
static pciDeviceList *
qemuGetActivePciHostDeviceList(struct qemud_driver *driver,
                               virDomainHostdevDefPtr *hostdevs,
                               int nhostdevs)
{
    pciDeviceList *list;
    int i;

    if (!(list = pciDeviceListNew()))
        return NULL;

    for (i = 0 ; i < nhostdevs ; i++) {
        virDomainHostdevDefPtr hostdev = hostdevs[i];
        pciDevice *dev;
        pciDevice *activeDev;

        if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
            continue;
        if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI)
            continue;

        dev = pciGetDevice(hostdev->source.subsys.u.pci.domain,
                           hostdev->source.subsys.u.pci.bus,
                           hostdev->source.subsys.u.pci.slot,
                           hostdev->source.subsys.u.pci.function);
        if (!dev) {
            pciDeviceListFree(list);
            return NULL;
        }

        if ((activeDev = pciDeviceListFind(driver->activePciHostdevs, dev))) {
            if (pciDeviceListAdd(list, activeDev) < 0) {
                pciFreeDevice(dev);
                pciDeviceListFree(list);
                return NULL;
            }
        }

        pciFreeDevice(dev);
    }

    return list;
}
118 119 120 121

int qemuUpdateActivePciHostdevs(struct qemud_driver *driver,
                                virDomainDefPtr def)
{
122 123
    virDomainHostdevDefPtr hostdev = NULL;
    int i;
124 125 126 127

    if (!def->nhostdevs)
        return 0;

128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151
    for (i = 0; i < def->nhostdevs; i++) {
        pciDevice *dev = NULL;
        hostdev = def->hostdevs[i];

        if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
            continue;
        if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI)
            continue;

        dev = pciGetDevice(hostdev->source.subsys.u.pci.domain,
                           hostdev->source.subsys.u.pci.bus,
                           hostdev->source.subsys.u.pci.slot,
                           hostdev->source.subsys.u.pci.function);

        if (!dev)
            return -1;

        pciDeviceSetManaged(dev, hostdev->managed);
        pciDeviceSetUsedBy(dev, def->name);

        /* Setup the original states for the PCI device */
        pciDeviceSetUnbindFromStub(dev, hostdev->origstates.states.pci.unbind_from_stub);
        pciDeviceSetRemoveSlot(dev, hostdev->origstates.states.pci.remove_slot);
        pciDeviceSetReprobe(dev, hostdev->origstates.states.pci.reprobe);
152 153 154

        if (pciDeviceListAdd(driver->activePciHostdevs, dev) < 0) {
            pciFreeDevice(dev);
155
            return -1;
156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195
        }
    }

    return 0;
}

int
qemuUpdateActiveUsbHostdevs(struct qemud_driver *driver,
                            virDomainDefPtr def)
{
    virDomainHostdevDefPtr hostdev = NULL;
    int i;

    if (!def->nhostdevs)
        return 0;

    for (i = 0; i < def->nhostdevs; i++) {
        usbDevice *usb = NULL;
        hostdev = def->hostdevs[i];

        if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
            continue;
        if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB)
            continue;

        usb = usbGetDevice(hostdev->source.subsys.u.usb.bus,
                           hostdev->source.subsys.u.usb.device);
        if (!usb) {
            VIR_WARN("Unable to reattach USB device %03d.%03d on domain %s",
                     hostdev->source.subsys.u.usb.bus,
                     hostdev->source.subsys.u.usb.device,
                     def->name);
            continue;
        }

        usbDeviceSetUsedBy(usb, def->name);

        if (usbDeviceListAdd(driver->activeUsbHostdevs, usb) < 0) {
            usbFreeDevice(usb);
            return -1;
196 197 198
        }
    }

199
    return 0;
200 201
}

202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261
static int
qemuDomainHostdevPciSysfsPath(virDomainHostdevDefPtr hostdev, char **sysfs_path)
{
    struct pci_config_address config_address;

    config_address.domain = hostdev->source.subsys.u.pci.domain;
    config_address.bus = hostdev->source.subsys.u.pci.bus;
    config_address.slot = hostdev->source.subsys.u.pci.slot;
    config_address.function = hostdev->source.subsys.u.pci.function;

    return pciConfigAddressToSysfsFile(&config_address, sysfs_path);
}

int
qemuDomainHostdevIsVirtualFunction(virDomainHostdevDefPtr hostdev)
{
    char *sysfs_path = NULL;
    int ret = -1;

    if (qemuDomainHostdevPciSysfsPath(hostdev, &sysfs_path) < 0)
        return ret;

    ret = pciDeviceIsVirtualFunction(sysfs_path);

    VIR_FREE(sysfs_path);

    return ret;
}

static int
qemuDomainHostdevNetDevice(virDomainHostdevDefPtr hostdev, char **linkdev,
                           int *vf)
{
    int ret = -1;
    char *sysfs_path = NULL;

    if (qemuDomainHostdevPciSysfsPath(hostdev, &sysfs_path) < 0)
        return ret;

    if (pciDeviceIsVirtualFunction(sysfs_path) == 1) {
        if (pciDeviceGetVirtualFunctionInfo(sysfs_path, linkdev,
                                            vf) < 0)
            goto cleanup;
    } else {
        if (pciDeviceNetName(sysfs_path, linkdev) < 0)
            goto cleanup;
        *vf = -1;
    }

    ret = 0;

cleanup:
    VIR_FREE(sysfs_path);

    return ret;
}

static int
qemuDomainHostdevNetConfigVirtPortProfile(const char *linkdev, int vf,
                                          virNetDevVPortProfilePtr virtPort,
262
                                          const virMacAddrPtr macaddr,
263 264 265 266 267 268 269 270 271 272 273 274 275
                                          const unsigned char *uuid,
                                          int associate)
{
    int ret = -1;

    if (!virtPort)
        return ret;

    switch(virtPort->virtPortType) {
    case VIR_NETDEV_VPORT_PROFILE_NONE:
    case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
    case VIR_NETDEV_VPORT_PROFILE_8021QBG:
    case VIR_NETDEV_VPORT_PROFILE_LAST:
276 277 278 279 280
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       _("virtualport type %s is "
                         "currently not supported on interfaces of type "
                         "hostdev"),
                       virNetDevVPortTypeToString(virtPort->virtPortType));
281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312
        break;

    case VIR_NETDEV_VPORT_PROFILE_8021QBH:
        if (associate)
            ret = virNetDevVPortProfileAssociate(NULL, virtPort, macaddr,
                                                 linkdev, vf, uuid,
                                                 VIR_NETDEV_VPORT_PROFILE_OP_CREATE, false);
        else
            ret = virNetDevVPortProfileDisassociate(NULL, virtPort,
                                                    macaddr, linkdev, vf,
                                                    VIR_NETDEV_VPORT_PROFILE_OP_DESTROY);
        break;
    }

    return ret;
}

int
qemuDomainHostdevNetConfigReplace(virDomainHostdevDefPtr hostdev,
                                  const unsigned char *uuid,
                                  char *stateDir)
{
    char *linkdev = NULL;
    virNetDevVPortProfilePtr virtPort;
    int ret = -1;
    int vf = -1;
    int vlanid = -1;
    int port_profile_associate = 1;
    int isvf;

    isvf = qemuDomainHostdevIsVirtualFunction(hostdev);
    if (isvf <= 0) {
313 314 315
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Interface type hostdev is currently supported on"
                         " SR-IOV Virtual Functions only"));
316 317 318 319 320 321 322 323 324 325
        return ret;
    }

    if (qemuDomainHostdevNetDevice(hostdev, &linkdev, &vf) < 0)
        return ret;

    virtPort = virDomainNetGetActualVirtPortProfile(
                                 hostdev->parent.data.net);
    if (virtPort)
        ret = qemuDomainHostdevNetConfigVirtPortProfile(linkdev, vf,
326
                            virtPort, &hostdev->parent.data.net->mac, uuid,
327 328 329 330
                            port_profile_associate);
    else
        /* Set only mac */
        ret = virNetDevReplaceNetConfig(linkdev, vf,
331
                                        &hostdev->parent.data.net->mac, vlanid,
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350
                                        stateDir);
    VIR_FREE(linkdev);

    return ret;
}

int
qemuDomainHostdevNetConfigRestore(virDomainHostdevDefPtr hostdev,
                                  char *stateDir)
{
    char *linkdev = NULL;
    virNetDevVPortProfilePtr virtPort;
    int ret = -1;
    int vf = -1;
    int port_profile_associate = 0;
    int isvf;

    isvf = qemuDomainHostdevIsVirtualFunction(hostdev);
    if (isvf <= 0) {
351 352 353
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Interface type hostdev is currently supported on"
                         " SR-IOV Virtual Functions only"));
354 355 356 357 358 359 360 361 362 363
        return ret;
    }

    if (qemuDomainHostdevNetDevice(hostdev, &linkdev, &vf) < 0)
        return ret;

    virtPort = virDomainNetGetActualVirtPortProfile(
                                 hostdev->parent.data.net);
    if (virtPort)
        ret = qemuDomainHostdevNetConfigVirtPortProfile(linkdev, vf, virtPort,
364
                                          &hostdev->parent.data.net->mac, NULL,
365 366 367 368 369 370 371 372 373
                                          port_profile_associate);
    else
        ret = virNetDevRestoreNetConfig(linkdev, vf, stateDir);

    VIR_FREE(linkdev);

    return ret;
}

374
int qemuPrepareHostdevPCIDevices(struct qemud_driver *driver,
375
                                 const char *name,
376
                                 const unsigned char *uuid,
377 378 379 380
                                 virDomainHostdevDefPtr *hostdevs,
                                 int nhostdevs)
{
    pciDeviceList *pcidevs;
381
    int last_processed_hostdev_vf = -1;
382 383 384 385 386 387
    int i;
    int ret = -1;

    if (!(pcidevs = qemuGetPciHostDeviceList(hostdevs, nhostdevs)))
        return -1;

388
    /* We have to use 9 loops here. *All* devices must
389 390 391 392 393 394
     * be detached before we reset any of them, because
     * in some cases you have to reset the whole PCI,
     * which impacts all devices on it. Also, all devices
     * must be reset before being marked as active.
     */

395
    /* Loop 1: validate that non-managed device isn't in use, eg
396 397 398 399 400 401
     * by checking that device is either un-bound, or bound
     * to pci-stub.ko
     */

    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
402
        pciDevice *other;
403

404
        if (!pciDeviceIsAssignable(dev, !driver->relaxedACS)) {
405 406 407
            virReportError(VIR_ERR_OPERATION_INVALID,
                           _("PCI device %s is not assignable"),
                           pciDeviceGetName(dev));
408 409 410 411 412 413 414 415 416
            goto cleanup;
        }
        /* The device is in use by other active domain if
         * the dev is in list driver->activePciHostdevs.
         */
        if ((other = pciDeviceListFind(driver->activePciHostdevs, dev))) {
            const char *other_name = pciDeviceGetUsedBy(other);

            if (other_name)
417 418 419
                virReportError(VIR_ERR_OPERATION_INVALID,
                               _("PCI device %s is in use by domain %s"),
                               pciDeviceGetName(dev), other_name);
420
            else
421 422 423
                virReportError(VIR_ERR_OPERATION_INVALID,
                               _("PCI device %s is already in use"),
                               pciDeviceGetName(dev));
424 425 426 427 428 429 430
            goto cleanup;
        }
    }

    /* Loop 2: detach managed devices */
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
431
        if (pciDeviceGetManaged(dev) &&
432
            pciDettachDevice(dev, driver->activePciHostdevs, NULL) < 0)
433
            goto reattachdevs;
434 435
    }

436 437
    /* Loop 3: Now that all the PCI hostdevs have been detached, we
     * can safely reset them */
438 439
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
440 441
        if (pciResetDevice(dev, driver->activePciHostdevs,
                           driver->inactivePciHostdevs) < 0)
442
            goto reattachdevs;
443 444
    }

445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463
    /* Loop 4: For SRIOV network devices, Now that we have detached the
     * the network device, set the netdev config */
    for (i = 0; i < nhostdevs; i++) {
         virDomainHostdevDefPtr hostdev = hostdevs[i];
         if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
             continue;
         if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI)
             continue;
         if (hostdev->parent.type == VIR_DOMAIN_DEVICE_NET &&
             hostdev->parent.data.net) {
             if (qemuDomainHostdevNetConfigReplace(hostdev, uuid,
                                                   driver->stateDir) < 0) {
                 goto resetvfnetconfig;
             }
         }
         last_processed_hostdev_vf = i;
    }

    /* Loop 5: Now mark all the devices as active */
464 465 466 467
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
        if (pciDeviceListAdd(driver->activePciHostdevs, dev) < 0) {
            pciFreeDevice(dev);
468
            goto inactivedevs;
469 470 471
        }
    }

472
    /* Loop 6: Now remove the devices from inactive list. */
473 474 475 476 477
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
         pciDevice *dev = pciDeviceListGet(pcidevs, i);
         pciDeviceListDel(driver->inactivePciHostdevs, dev);
    }

478
    /* Loop 7: Now set the used_by_domain of the device in
479 480 481 482 483 484 485 486 487 488 489
     * driver->activePciHostdevs as domain name.
     */
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev, *activeDev;

        dev = pciDeviceListGet(pcidevs, i);
        activeDev = pciDeviceListFind(driver->activePciHostdevs, dev);

        pciDeviceSetUsedBy(activeDev, name);
    }

490
    /* Loop 8: Now set the original states for hostdev def */
491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521
    for (i = 0; i < nhostdevs; i++) {
        pciDevice *dev;
        pciDevice *pcidev;
        virDomainHostdevDefPtr hostdev = hostdevs[i];

        if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
            continue;
        if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI)
            continue;

        dev = pciGetDevice(hostdev->source.subsys.u.pci.domain,
                           hostdev->source.subsys.u.pci.bus,
                           hostdev->source.subsys.u.pci.slot,
                           hostdev->source.subsys.u.pci.function);

        /* original states "unbind_from_stub", "remove_slot",
         * "reprobe" were already set by pciDettachDevice in
         * loop 2.
         */
        if ((pcidev = pciDeviceListFind(pcidevs, dev))) {
            hostdev->origstates.states.pci.unbind_from_stub =
                pciDeviceGetUnbindFromStub(pcidev);
            hostdev->origstates.states.pci.remove_slot =
                pciDeviceGetRemoveSlot(pcidev);
            hostdev->origstates.states.pci.reprobe =
                pciDeviceGetReprobe(pcidev);
        }

        pciFreeDevice(dev);
    }

522
    /* Loop 9: Now steal all the devices from pcidevs */
523 524
    while (pciDeviceListCount(pcidevs) > 0) {
        pciDevice *dev = pciDeviceListGet(pcidevs, 0);
525 526 527
        pciDeviceListSteal(pcidevs, dev);
    }

528
    ret = 0;
529 530 531 532 533 534
    goto cleanup;

inactivedevs:
    /* Only steal all the devices from driver->activePciHostdevs. We will
     * free them in pciDeviceListFree().
     */
535 536
    while (pciDeviceListCount(pcidevs) > 0) {
        pciDevice *dev = pciDeviceListGet(pcidevs, 0);
537 538
        pciDeviceListSteal(driver->activePciHostdevs, dev);
    }
539

540 541 542 543 544 545 546 547 548
resetvfnetconfig:
    for (i = 0; i < last_processed_hostdev_vf; i++) {
         virDomainHostdevDefPtr hostdev = hostdevs[i];
         if (hostdev->parent.type == VIR_DOMAIN_DEVICE_NET &&
             hostdev->parent.data.net) {
             qemuDomainHostdevNetConfigRestore(hostdev, driver->stateDir);
         }
    }

549 550 551
reattachdevs:
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
552
        pciReAttachDevice(dev, driver->activePciHostdevs, NULL);
553 554
    }

555 556 557 558 559 560 561 562 563
cleanup:
    pciDeviceListFree(pcidevs);
    return ret;
}

static int
qemuPrepareHostPCIDevices(struct qemud_driver *driver,
                          virDomainDefPtr def)
{
564 565
    return qemuPrepareHostdevPCIDevices(driver, def->name, def->uuid,
                                        def->hostdevs, def->nhostdevs);
566 567
}

568 569 570
int
qemuPrepareHostdevUSBDevices(struct qemud_driver *driver,
                             const char *name,
571
                             usbDeviceList *list)
572
{
573
    int i, j;
574 575 576 577 578 579 580 581 582 583 584
    unsigned int count;
    usbDevice *tmp;

    count = usbDeviceListCount(list);

    for (i = 0; i < count; i++) {
        usbDevice *usb = usbDeviceListGet(list, i);
        if ((tmp = usbDeviceListFind(driver->activeUsbHostdevs, usb))) {
            const char *other_name = usbDeviceGetUsedBy(tmp);

            if (other_name)
585 586 587
                virReportError(VIR_ERR_OPERATION_INVALID,
                               _("USB device %s is in use by domain %s"),
                               usbDeviceGetName(tmp), other_name);
588
            else
589 590 591
                virReportError(VIR_ERR_OPERATION_INVALID,
                               _("USB device %s is already in use"),
                               usbDeviceGetName(tmp));
592
            goto error;
593 594 595 596 597 598 599 600 601 602 603
        }

        usbDeviceSetUsedBy(usb, name);
        VIR_DEBUG("Adding %03d.%03d dom=%s to activeUsbHostdevs",
                  usbDeviceGetBus(usb), usbDeviceGetDevno(usb), name);
        /*
         * The caller is responsible to steal these usb devices
         * from the usbDeviceList that passed in on success,
         * perform rollback on failure.
         */
        if (usbDeviceListAdd(driver->activeUsbHostdevs, usb) < 0)
604
            goto error;
605 606
    }
    return 0;
607 608 609 610 611 612 613

error:
    for (j = 0; j < i; j++) {
        tmp = usbDeviceListGet(list, i);
        usbDeviceListSteal(driver->activeUsbHostdevs, tmp);
    }
    return -1;
614 615 616 617 618 619 620
}

static int
qemuPrepareHostUSBDevices(struct qemud_driver *driver,
                          virDomainDefPtr def)
{
    int i, ret = -1;
621 622
    usbDeviceList *list;
    usbDevice *tmp;
623 624
    virDomainHostdevDefPtr *hostdevs = def->hostdevs;
    int nhostdevs = def->nhostdevs;
625 626 627 628 629 630 631 632 633

    /* To prevent situation where USB device is assigned to two domains
     * we need to keep a list of currently assigned USB devices.
     * This is done in several loops which cannot be joined into one big
     * loop. See qemuPrepareHostdevPCIDevices()
     */
    if (!(list = usbDeviceListNew()))
        goto cleanup;

634
    /* Loop 1: build temporary list
635 636 637
     */
    for (i = 0 ; i < nhostdevs ; i++) {
        virDomainHostdevDefPtr hostdev = hostdevs[i];
638
        usbDevice *usb = NULL;
639 640 641 642 643 644

        if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
            continue;
        if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB)
            continue;

645 646 647 648
        unsigned vendor = hostdev->source.subsys.u.usb.vendor;
        unsigned product = hostdev->source.subsys.u.usb.product;
        unsigned bus = hostdev->source.subsys.u.usb.bus;
        unsigned device = hostdev->source.subsys.u.usb.device;
649

650 651
        if (vendor && bus) {
            usb = usbFindDevice(vendor, product, bus, device);
652

653 654
        } else if (vendor && !bus) {
            usbDeviceList *devs = usbFindDeviceByVendor(vendor, product);
655
            if (!devs)
656
                goto cleanup;
657

658
            if (usbDeviceListCount(devs) > 1) {
659 660 661
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("multiple USB devices for %x:%x, "
                                 "use <address> to specify one"), vendor, product);
662 663 664
                usbDeviceListFree(devs);
                goto cleanup;
            }
665 666 667 668
            usb = usbDeviceListGet(devs, 0);
            usbDeviceListSteal(devs, usb);
            usbDeviceListFree(devs);

669 670 671
            hostdev->source.subsys.u.usb.bus = usbDeviceGetBus(usb);
            hostdev->source.subsys.u.usb.device = usbDeviceGetDevno(usb);

672 673 674 675 676 677
        } else if (!vendor && bus) {
            usb = usbFindDeviceByBus(bus, device);
        }

        if (!usb)
            goto cleanup;
678

679 680 681
        if (usbDeviceListAdd(list, usb) < 0) {
            usbFreeDevice(usb);
            goto cleanup;
682 683 684
        }
    }

685
    /* Mark devices in temporary list as used by @name
686 687 688
     * and add them do driver list. However, if something goes
     * wrong, perform rollback.
     */
689
    if (qemuPrepareHostdevUSBDevices(driver, def->name, list) < 0)
690
        goto cleanup;
691

692
    /* Loop 2: Temporary list was successfully merged with
693 694 695 696 697 698 699 700 701 702 703 704 705
     * driver list, so steal all items to avoid freeing them
     * in cleanup label.
     */
    while (usbDeviceListCount(list) > 0) {
        tmp = usbDeviceListGet(list, 0);
        usbDeviceListSteal(list, tmp);
    }

    ret = 0;

cleanup:
    usbDeviceListFree(list);
    return ret;
706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727
}

int qemuPrepareHostDevices(struct qemud_driver *driver,
                           virDomainDefPtr def)
{
    if (!def->nhostdevs)
        return 0;

    if (qemuPrepareHostPCIDevices(driver, def) < 0)
        return -1;

    if (qemuPrepareHostUSBDevices(driver, def) < 0)
        return -1;

    return 0;
}


void qemuReattachPciDevice(pciDevice *dev, struct qemud_driver *driver)
{
    int retries = 100;

728 729 730 731 732
    /* If the device is not managed and was attached to guest
     * successfully, it must have been inactive.
     */
    if (!pciDeviceGetManaged(dev)) {
        pciDeviceListAdd(driver->inactivePciHostdevs, dev);
733
        return;
734
    }
735

736 737 738 739 740 741
    while (pciWaitForDeviceCleanup(dev, "kvm_assigned_device")
           && retries) {
        usleep(100*1000);
        retries--;
    }

742 743
    if (pciReAttachDevice(dev, driver->activePciHostdevs,
                          driver->inactivePciHostdevs) < 0) {
744 745 746 747
        virErrorPtr err = virGetLastError();
        VIR_ERROR(_("Failed to re-attach PCI device: %s"),
                  err ? err->message : _("unknown error"));
        virResetError(err);
748 749 750 751 752
    }
}


void qemuDomainReAttachHostdevDevices(struct qemud_driver *driver,
753
                                      const char *name,
754 755 756 757 758 759
                                      virDomainHostdevDefPtr *hostdevs,
                                      int nhostdevs)
{
    pciDeviceList *pcidevs;
    int i;

760 761 762
    if (!(pcidevs = qemuGetActivePciHostDeviceList(driver,
                                                   hostdevs,
                                                   nhostdevs))) {
763 764 765 766 767 768 769
        virErrorPtr err = virGetLastError();
        VIR_ERROR(_("Failed to allocate pciDeviceList: %s"),
                  err ? err->message : _("unknown error"));
        virResetError(err);
        return;
    }

770 771 772 773
    /* Again 4 loops; mark all devices as inactive before reset
     * them and reset all the devices before re-attach.
     * Attach mac and port profile parameters to devices
     */
774 775
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
776 777 778 779 780 781 782
        pciDevice *activeDev = NULL;

        /* Never delete the dev from list driver->activePciHostdevs
         * if it's used by other domain.
         */
        activeDev = pciDeviceListFind(driver->activePciHostdevs, dev);
        if (activeDev &&
783 784
            STRNEQ_NULLABLE(name, pciDeviceGetUsedBy(activeDev))) {
            pciDeviceListSteal(pcidevs, dev);
785
            continue;
786
        }
787

788 789
        /* pciDeviceListFree() will take care of freeing the dev. */
        pciDeviceListSteal(driver->activePciHostdevs, dev);
790 791
    }

792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807
    /*
     * For SRIOV net host devices, unset mac and port profile before
     * reset and reattach device
     */
    for (i = 0; i < nhostdevs; i++) {
         virDomainHostdevDefPtr hostdev = hostdevs[i];
         if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
             continue;
         if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI)
             continue;
         if (hostdev->parent.type == VIR_DOMAIN_DEVICE_NET &&
             hostdev->parent.data.net) {
             qemuDomainHostdevNetConfigRestore(hostdev, driver->stateDir);
         }
    }

808 809
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
810 811
        if (pciResetDevice(dev, driver->activePciHostdevs,
                           driver->inactivePciHostdevs) < 0) {
812 813 814 815 816 817 818 819 820 821 822 823 824 825 826
            virErrorPtr err = virGetLastError();
            VIR_ERROR(_("Failed to reset PCI device: %s"),
                      err ? err->message : _("unknown error"));
            virResetError(err);
        }
    }

    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
        qemuReattachPciDevice(dev, driver);
    }

    pciDeviceListFree(pcidevs);
}

827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883
static void
qemuDomainReAttachHostUsbDevices(struct qemud_driver *driver,
                                 const char *name,
                                 virDomainHostdevDefPtr *hostdevs,
                                 int nhostdevs)
{
    int i;

    for (i = 0; i < nhostdevs; i++) {
        virDomainHostdevDefPtr hostdev = hostdevs[i];
        usbDevice *usb, *tmp;
        const char *used_by = NULL;

        if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
            continue;
        if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB)
            continue;

        usb = usbGetDevice(hostdev->source.subsys.u.usb.bus,
                           hostdev->source.subsys.u.usb.device);

        if (!usb) {
            VIR_WARN("Unable to reattach USB device %03d.%03d on domain %s",
                     hostdev->source.subsys.u.usb.bus,
                     hostdev->source.subsys.u.usb.device,
                     name);
            continue;
        }

        /* Delete only those USB devices which belongs
         * to domain @name because qemuProcessStart() might
         * have failed because USB device is already taken.
         * Therefore we want to steal only those devices from
         * the list which were taken by @name */

        tmp = usbDeviceListFind(driver->activeUsbHostdevs, usb);
        usbFreeDevice(usb);

        if (!tmp) {
            VIR_WARN("Unable to find device %03d.%03d "
                     "in list of active USB devices",
                     hostdev->source.subsys.u.usb.bus,
                     hostdev->source.subsys.u.usb.device);
            continue;
        }

        used_by = usbDeviceGetUsedBy(tmp);
        if (STREQ_NULLABLE(used_by, name)) {
            VIR_DEBUG("Removing %03d.%03d dom=%s from activeUsbHostdevs",
                      hostdev->source.subsys.u.usb.bus,
                      hostdev->source.subsys.u.usb.device,
                      name);

            usbDeviceListDel(driver->activeUsbHostdevs, tmp);
        }
    }
}
884 885 886 887 888 889 890

void qemuDomainReAttachHostDevices(struct qemud_driver *driver,
                                   virDomainDefPtr def)
{
    if (!def->nhostdevs)
        return;

891 892
    qemuDomainReAttachHostdevDevices(driver, def->name, def->hostdevs,
                                     def->nhostdevs);
893 894 895

    qemuDomainReAttachHostUsbDevices(driver, def->name, def->hostdevs,
                                     def->nhostdevs);
896
}