pci.c 59.3 KB
Newer Older
1
/*
2
 * Copyright (C) 2009-2012 Red Hat, Inc.
3 4 5 6 7 8 9 10 11 12 13 14
 *
 * 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
15
 * License along with this library.  If not, see
O
Osier Yang 已提交
16
 * <http://www.gnu.org/licenses/>.
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
 *
 * Authors:
 *     Mark McLoughlin <markmc@redhat.com>
 */

#include <config.h>

#include "pci.h"

#include <dirent.h>
#include <fcntl.h>
#include <inttypes.h>
#include <limits.h>
#include <stdio.h>
#include <string.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
35
#include <stdlib.h>
36 37 38

#include "logging.h"
#include "memory.h"
39
#include "command.h"
40
#include "virterror_internal.h"
E
Eric Blake 已提交
41
#include "virfile.h"
42 43 44 45 46

#define PCI_SYSFS "/sys/bus/pci/"
#define PCI_ID_LEN 10   /* "XXXX XXXX" */
#define PCI_ADDR_LEN 13 /* "XXXX:XX:XX.X" */

47 48 49 50
#define SRIOV_FOUND 0
#define SRIOV_NOT_FOUND 1
#define SRIOV_ERROR -1

51 52 53 54 55 56 57 58
struct _pciDevice {
    unsigned      domain;
    unsigned      bus;
    unsigned      slot;
    unsigned      function;

    char          name[PCI_ADDR_LEN]; /* domain:bus:slot.function */
    char          id[PCI_ID_LEN];     /* product vendor */
E
Eric Blake 已提交
59
    char          *path;
60
    const char    *used_by;           /* The domain which uses the device */
61 62 63 64 65 66 67
    int           fd;

    unsigned      initted;
    unsigned      pcie_cap_pos;
    unsigned      pci_pm_cap_pos;
    unsigned      has_flr : 1;
    unsigned      has_pm_reset : 1;
68
    unsigned      managed : 1;
69 70 71 72 73

    /* used by reattach function */
    unsigned      unbind_from_stub : 1;
    unsigned      remove_slot : 1;
    unsigned      reprobe : 1;
74 75
};

76 77 78 79 80 81
struct _pciDeviceList {
    unsigned count;
    pciDevice **devs;
};


82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98
/* For virReportOOMError()  and virReportSystemError() */
#define VIR_FROM_THIS VIR_FROM_NONE

/* Specifications referenced in comments:
 *  PCI30  - PCI Local Bus Specification 3.0
 *  PCIe20 - PCI Express Base Specification 2.0
 *  BR12   - PCI-to-PCI Bridge Architecture Specification 1.2
 *  PM12   - PCI Bus Power Management Interface Specification 1.2
 *  ECN_AF - Advanced Capabilities for Conventional PCI ECN
 */

/* Type 0 config space header length; PCI30 Section 6.1 Configuration Space Organization */
#define PCI_CONF_LEN            0x100
#define PCI_CONF_HEADER_LEN     0x40

/* PCI30 6.2.1 */
#define PCI_HEADER_TYPE         0x0e    /* Header type */
99 100 101
#define PCI_HEADER_TYPE_BRIDGE 0x1
#define PCI_HEADER_TYPE_MASK   0x7f
#define PCI_HEADER_TYPE_MULTI  0x80
102 103 104 105 106 107 108 109 110

/* PCI30 6.2.1  Device Identification */
#define PCI_CLASS_DEVICE        0x0a    /* Device class */

/* Class Code for bridge; PCI30 D.7  Base Class 06h */
#define PCI_CLASS_BRIDGE_PCI    0x0604

/* PCI30 6.2.3  Device Status */
#define PCI_STATUS              0x06    /* 16 bits */
111
#define PCI_STATUS_CAP_LIST    0x10    /* Support Capability List */
112 113 114 115 116 117 118 119 120 121 122 123 124

/* PCI30 6.7  Capabilities List */
#define PCI_CAPABILITY_LIST     0x34    /* Offset of first capability list entry */

/* PM12 3.2.1  Capability Identifier */
#define PCI_CAP_ID_PM           0x01    /* Power Management */
/* PCI30 H Capability IDs */
#define PCI_CAP_ID_EXP          0x10    /* PCI Express */
/* ECN_AF 6.x.1.1  Capability ID for AF */
#define PCI_CAP_ID_AF           0x13    /* Advanced Features */

/* PCIe20 7.8.3  Device Capabilities Register (Offset 04h) */
#define PCI_EXP_DEVCAP          0x4     /* Device capabilities */
125
#define PCI_EXP_DEVCAP_FLR     (1<<28) /* Function Level Reset */
126 127 128 129 130 131 132

/* Header type 1 BR12 3.2 PCI-to-PCI Bridge Configuration Space Header Format */
#define PCI_PRIMARY_BUS         0x18    /* BR12 3.2.5.2 Primary bus number */
#define PCI_SECONDARY_BUS       0x19    /* BR12 3.2.5.3 Secondary bus number */
#define PCI_SUBORDINATE_BUS     0x1a    /* BR12 3.2.5.4 Highest bus number behind the bridge */
#define PCI_BRIDGE_CONTROL      0x3e
/* BR12 3.2.5.18  Bridge Control Register */
133
#define PCI_BRIDGE_CTL_RESET   0x40    /* Secondary bus reset */
134 135 136

/* PM12 3.2.4  Power Management Control/Status (Offset = 4) */
#define PCI_PM_CTRL                4    /* PM control and status register */
137 138 139 140
#define PCI_PM_CTRL_STATE_MASK    0x3  /* Current power state (D0 to D3) */
#define PCI_PM_CTRL_STATE_D0      0x0  /* D0 state */
#define PCI_PM_CTRL_STATE_D3hot   0x3  /* D3 state */
#define PCI_PM_CTRL_NO_SOFT_RESET 0x8  /* No reset for D3hot->D0 */
141 142 143

/* ECN_AF 6.x.1  Advanced Features Capability Structure */
#define PCI_AF_CAP              0x3     /* Advanced features capabilities */
144
#define PCI_AF_CAP_FLR         0x2     /* Function Level Reset */
145

J
Jiri Denemark 已提交
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167
#define PCI_EXP_FLAGS           0x2
#define PCI_EXP_FLAGS_TYPE      0x00f0
#define PCI_EXP_TYPE_DOWNSTREAM 0x6

#define PCI_EXT_CAP_BASE          0x100
#define PCI_EXT_CAP_LIMIT         0x1000
#define PCI_EXT_CAP_ID_MASK       0x0000ffff
#define PCI_EXT_CAP_OFFSET_SHIFT  20
#define PCI_EXT_CAP_OFFSET_MASK   0x00000ffc

#define PCI_EXT_CAP_ID_ACS      0x000d
#define PCI_EXT_ACS_CTRL        0x06

#define PCI_EXT_CAP_ACS_SV      0x01
#define PCI_EXT_CAP_ACS_RR      0x04
#define PCI_EXT_CAP_ACS_CR      0x08
#define PCI_EXT_CAP_ACS_UF      0x10
#define PCI_EXT_CAP_ACS_ENABLED (PCI_EXT_CAP_ACS_SV | \
                                 PCI_EXT_CAP_ACS_RR | \
                                 PCI_EXT_CAP_ACS_CR | \
                                 PCI_EXT_CAP_ACS_UF)

168 169 170 171 172 173 174 175 176 177 178
static int
pciOpenConfig(pciDevice *dev)
{
    int fd;

    if (dev->fd > 0)
        return 0;

    fd = open(dev->path, O_RDWR);
    if (fd < 0) {
        char ebuf[1024];
179
        VIR_WARN("Failed to open config space file '%s': %s",
180 181 182 183 184 185 186 187
                 dev->path, virStrerror(errno, ebuf, sizeof(ebuf)));
        return -1;
    }
    VIR_DEBUG("%s %s: opened %s", dev->id, dev->name, dev->path);
    dev->fd = fd;
    return 0;
}

188 189 190 191 192 193
static void
pciCloseConfig(pciDevice *dev)
{
    if (!dev)
        return;

194
    VIR_FORCE_CLOSE(dev->fd);
195 196
}

197 198 199 200 201 202 203 204
static int
pciRead(pciDevice *dev, unsigned pos, uint8_t *buf, unsigned buflen)
{
    memset(buf, 0, buflen);

    if (pciOpenConfig(dev) < 0)
        return -1;

205 206
    if (lseek(dev->fd, pos, SEEK_SET) != pos ||
        saferead(dev->fd, buf, buflen) != buflen) {
207
        char ebuf[1024];
208
        VIR_WARN("Failed to read from '%s' : %s", dev->path,
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
                 virStrerror(errno, ebuf, sizeof(ebuf)));
        return -1;
    }
    return 0;
}

static uint8_t
pciRead8(pciDevice *dev, unsigned pos)
{
    uint8_t buf;
    pciRead(dev, pos, &buf, sizeof(buf));
    return buf;
}

static uint16_t
pciRead16(pciDevice *dev, unsigned pos)
{
    uint8_t buf[2];
    pciRead(dev, pos, &buf[0], sizeof(buf));
    return (buf[0] << 0) | (buf[1] << 8);
}

static uint32_t
pciRead32(pciDevice *dev, unsigned pos)
{
    uint8_t buf[4];
    pciRead(dev, pos, &buf[0], sizeof(buf));
    return (buf[0] << 0) | (buf[1] << 8) | (buf[2] << 16) | (buf[3] << 24);
}

static int
pciWrite(pciDevice *dev, unsigned pos, uint8_t *buf, unsigned buflen)
{
    if (pciOpenConfig(dev) < 0)
        return -1;

245 246
    if (lseek(dev->fd, pos, SEEK_SET) != pos ||
        safewrite(dev->fd, buf, buflen) != buflen) {
247
        char ebuf[1024];
248
        VIR_WARN("Failed to write to '%s' : %s", dev->path,
249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268
                 virStrerror(errno, ebuf, sizeof(ebuf)));
        return -1;
    }
    return 0;
}

static void
pciWrite16(pciDevice *dev, unsigned pos, uint16_t val)
{
    uint8_t buf[2] = { (val >> 0), (val >> 8) };
    pciWrite(dev, pos, &buf[0], sizeof(buf));
}

static void
pciWrite32(pciDevice *dev, unsigned pos, uint32_t val)
{
    uint8_t buf[4] = { (val >> 0), (val >> 8), (val >> 16), (val >> 14) };
    pciWrite(dev, pos, &buf[0], sizeof(buf));
}

269
typedef int (*pciIterPredicate)(pciDevice *, pciDevice *, void *);
270 271 272 273 274 275 276

/* Iterate over available PCI devices calling @predicate
 * to compare each one to @dev.
 * Return -1 on error since we don't want to assume it is
 * safe to reset if there is an error.
 */
static int
277
pciIterDevices(pciIterPredicate predicate,
278
               pciDevice *dev,
279 280
               pciDevice **matched,
               void *data)
281 282 283
{
    DIR *dir;
    struct dirent *entry;
284
    int ret = 0;
285
    int rc;
286 287 288 289 290 291 292

    *matched = NULL;

    VIR_DEBUG("%s %s: iterating over " PCI_SYSFS "devices", dev->id, dev->name);

    dir = opendir(PCI_SYSFS "devices");
    if (!dir) {
293
        VIR_WARN("Failed to open " PCI_SYSFS "devices");
294 295 296 297
        return -1;
    }

    while ((entry = readdir(dir))) {
298
        unsigned int domain, bus, slot, function;
299
        pciDevice *check;
300
        char *tmp;
301 302 303 304 305

        /* Ignore '.' and '..' */
        if (entry->d_name[0] == '.')
            continue;

306 307 308 309 310 311 312 313 314
        /* expected format: <domain>:<bus>:<slot>.<function> */
        if (/* domain */
            virStrToLong_ui(entry->d_name, &tmp, 16, &domain) < 0 || *tmp != ':' ||
            /* bus */
            virStrToLong_ui(tmp + 1, &tmp, 16, &bus) < 0 || *tmp != ':' ||
            /* slot */
            virStrToLong_ui(tmp + 1, &tmp, 16, &slot) < 0 || *tmp != '.' ||
            /* function */
            virStrToLong_ui(tmp + 1, NULL, 16, &function) < 0) {
315 316 317 318
            VIR_WARN("Unusual entry in " PCI_SYSFS "devices: %s", entry->d_name);
            continue;
        }

319
        check = pciGetDevice(domain, bus, slot, function);
320
        if (!check) {
321 322 323
            ret = -1;
            break;
        }
324

325 326 327 328 329 330 331 332
        rc = predicate(dev, check, data);
        if (rc < 0) {
            /* the predicate returned an error, bail */
            pciFreeDevice(check);
            ret = -1;
            break;
        }
        else if (rc == 1) {
333 334
            VIR_DEBUG("%s %s: iter matched on %s", dev->id, dev->name, check->name);
            *matched = check;
335
            ret = 1;
336 337
            break;
        }
338

339
        pciFreeDevice(check);
340 341
    }
    closedir(dir);
342
    return ret;
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
}

static uint8_t
pciFindCapabilityOffset(pciDevice *dev, unsigned capability)
{
    uint16_t status;
    uint8_t pos;

    status = pciRead16(dev, PCI_STATUS);
    if (!(status & PCI_STATUS_CAP_LIST))
        return 0;

    pos = pciRead8(dev, PCI_CAPABILITY_LIST);

    /* Zero indicates last capability, capabilities can't
     * be in the config space header and 0xff is returned
     * by the kernel if we don't have access to this region
     *
     * Note: we're not handling loops or extended
     * capabilities here.
     */
    while (pos >= PCI_CONF_HEADER_LEN && pos != 0xff) {
        uint8_t capid = pciRead8(dev, pos);
        if (capid == capability) {
            VIR_DEBUG("%s %s: found cap 0x%.2x at 0x%.2x",
                      dev->id, dev->name, capability, pos);
            return pos;
        }

        pos = pciRead8(dev, pos + 1);
    }

    VIR_DEBUG("%s %s: failed to find cap 0x%.2x", dev->id, dev->name, capability);

    return 0;
}

J
Jiri Denemark 已提交
380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403
static unsigned int
pciFindExtendedCapabilityOffset(pciDevice *dev, unsigned capability)
{
    int ttl;
    unsigned int pos;
    uint32_t header;

    /* minimum 8 bytes per capability */
    ttl = (PCI_EXT_CAP_LIMIT - PCI_EXT_CAP_BASE) / 8;
    pos = PCI_EXT_CAP_BASE;

    while (ttl > 0 && pos >= PCI_EXT_CAP_BASE) {
        header = pciRead32(dev, pos);

        if ((header & PCI_EXT_CAP_ID_MASK) == capability)
            return pos;

        pos = (header >> PCI_EXT_CAP_OFFSET_SHIFT) & PCI_EXT_CAP_OFFSET_MASK;
        ttl--;
    }

    return 0;
}

404 405 406 407
/* detects whether this device has FLR.  Returns 0 if the device does
 * not have FLR, 1 if it does, and -1 on error
 */
static int
408 409
pciDetectFunctionLevelReset(pciDevice *dev)
{
M
Mark McLoughlin 已提交
410
    uint32_t caps;
411
    uint8_t pos;
412 413
    char *path;
    int found;
414 415 416 417 418 419 420 421

    /* The PCIe Function Level Reset capability allows
     * individual device functions to be reset without
     * affecting any other functions on the device or
     * any other devices on the bus. This is only common
     * on SR-IOV NICs at the moment.
     */
    if (dev->pcie_cap_pos) {
M
Mark McLoughlin 已提交
422
        caps = pciRead32(dev, dev->pcie_cap_pos + PCI_EXP_DEVCAP);
423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441
        if (caps & PCI_EXP_DEVCAP_FLR) {
            VIR_DEBUG("%s %s: detected PCIe FLR capability", dev->id, dev->name);
            return 1;
        }
    }

    /* The PCI AF Function Level Reset capability is
     * the same thing, except for conventional PCI
     * devices. This is not common yet.
     */
    pos = pciFindCapabilityOffset(dev, PCI_CAP_ID_AF);
    if (pos) {
        caps = pciRead16(dev, pos + PCI_AF_CAP);
        if (caps & PCI_AF_CAP_FLR) {
            VIR_DEBUG("%s %s: detected PCI FLR capability", dev->id, dev->name);
            return 1;
        }
    }

442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460
    /* there are some buggy devices that do support FLR, but forget to
     * advertise that fact in their capabilities.  However, FLR is *required*
     * to be present for virtual functions (VFs), so if we see that this
     * device is a VF, we just assume FLR works
     */

    if (virAsprintf(&path, PCI_SYSFS "devices/%s/physfn", dev->name) < 0) {
        virReportOOMError();
        return -1;
    }

    found = virFileExists(path);
    VIR_FREE(path);
    if (found) {
        VIR_DEBUG("%s %s: buggy device didn't advertise FLR, but is a VF; forcing flr on",
                  dev->id, dev->name);
        return 1;
    }

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
    VIR_DEBUG("%s %s: no FLR capability found", dev->id, dev->name);

    return 0;
}

/* Require the device has the PCI Power Management capability
 * and that a D3hot->D0 transition will results in a full
 * internal reset, not just a soft reset.
 */
static unsigned
pciDetectPowerManagementReset(pciDevice *dev)
{
    if (dev->pci_pm_cap_pos) {
        uint32_t ctl;

        /* require the NO_SOFT_RESET bit is clear */
        ctl = pciRead32(dev, dev->pci_pm_cap_pos + PCI_PM_CTRL);
        if (!(ctl & PCI_PM_CTRL_NO_SOFT_RESET)) {
            VIR_DEBUG("%s %s: detected PM reset capability", dev->id, dev->name);
            return 1;
        }
    }

    VIR_DEBUG("%s %s: no PM reset capability found", dev->id, dev->name);

    return 0;
}

489
/* Any active devices on the same domain/bus ? */
490
static int
491
pciSharesBusWithActive(pciDevice *dev, pciDevice *check, void *data)
492
{
493
    pciDeviceList *inactiveDevs = data;
494

495
    /* Different domain, different bus, or simply identical device */
496 497
    if (dev->domain != check->domain ||
        dev->bus != check->bus ||
498 499
        (dev->slot == check->slot &&
         dev->function == check->function))
500 501
        return 0;

502 503
    /* same bus, but inactive, i.e. about to be assigned to guest */
    if (inactiveDevs && pciDeviceListFind(inactiveDevs, check))
504
        return 0;
505

506
    return 1;
507 508
}

509
static pciDevice *
510
pciBusContainsActiveDevices(pciDevice *dev,
511
                            pciDeviceList *inactiveDevs)
512 513
{
    pciDevice *active = NULL;
514
    if (pciIterDevices(pciSharesBusWithActive,
515
                       dev, &active, inactiveDevs) < 0)
516 517 518 519 520
        return NULL;
    return active;
}

/* Is @check the parent of @dev ? */
521
static int
522
pciIsParent(pciDevice *dev, pciDevice *check, void *data)
523 524 525
{
    uint16_t device_class;
    uint8_t header_type, secondary, subordinate;
526
    pciDevice **best = data;
527

528
    if (dev->domain != check->domain)
529 530 531
        return 0;

    /* Is it a bridge? */
532
    device_class = pciRead16(check, PCI_CLASS_DEVICE);
533 534 535 536
    if (device_class != PCI_CLASS_BRIDGE_PCI)
        return 0;

    /* Is it a plane? */
537
    header_type = pciRead8(check, PCI_HEADER_TYPE);
538 539 540
    if ((header_type & PCI_HEADER_TYPE_MASK) != PCI_HEADER_TYPE_BRIDGE)
        return 0;

541 542
    secondary   = pciRead8(check, PCI_SECONDARY_BUS);
    subordinate = pciRead8(check, PCI_SUBORDINATE_BUS);
543

544
    VIR_DEBUG("%s %s: found parent device %s", dev->id, dev->name, check->name);
545

546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578
    /* if the secondary bus exactly equals the device's bus, then we found
     * the direct parent.  No further work is necessary
     */
    if (dev->bus == secondary)
        return 1;

    /* otherwise, SRIOV allows VFs to be on different busses then their PFs.
     * In this case, what we need to do is look for the "best" match; i.e.
     * the most restrictive match that still satisfies all of the conditions.
     */
    if (dev->bus > secondary && dev->bus <= subordinate) {
        if (*best == NULL) {
            *best = pciGetDevice(check->domain, check->bus, check->slot,
                                 check->function);
            if (*best == NULL)
                return -1;
        }
        else {
            /* OK, we had already recorded a previous "best" match for the
             * parent.  See if the current device is more restrictive than the
             * best, and if so, make it the new best
             */
            if (secondary > pciRead8(*best, PCI_SECONDARY_BUS)) {
                pciFreeDevice(*best);
                *best = pciGetDevice(check->domain, check->bus, check->slot,
                                     check->function);
                if (*best == NULL)
                    return -1;
            }
        }
    }

    return 0;
579 580
}

581 582
static int
pciGetParentDevice(pciDevice *dev, pciDevice **parent)
583
{
584 585 586 587 588 589 590 591 592 593
    pciDevice *best = NULL;
    int ret;

    *parent = NULL;
    ret = pciIterDevices(pciIsParent, dev, parent, &best);
    if (ret == 1)
        pciFreeDevice(best);
    else if (ret == 0)
        *parent = best;
    return ret;
594 595 596 597 598 599
}

/* Secondary Bus Reset is our sledgehammer - it resets all
 * devices behind a bus.
 */
static int
600
pciTrySecondaryBusReset(pciDevice *dev,
601
                        pciDeviceList *inactiveDevs)
602
{
603
    pciDevice *parent, *conflict;
604 605 606 607
    uint8_t config_space[PCI_CONF_LEN];
    uint16_t ctl;
    int ret = -1;

608 609 610
    /* Refuse to do a secondary bus reset if there are other
     * devices/functions behind the bus are used by the host
     * or other guests.
611
     */
612
    if ((conflict = pciBusContainsActiveDevices(dev, inactiveDevs))) {
613
        virReportError(VIR_ERR_INTERNAL_ERROR,
614 615
                       _("Active %s devices on bus with %s, not doing bus reset"),
                       conflict->name, dev->name);
616 617 618 619
        return -1;
    }

    /* Find the parent bus */
620 621
    if (pciGetParentDevice(dev, &parent) < 0)
        return -1;
622
    if (!parent) {
623
        virReportError(VIR_ERR_INTERNAL_ERROR,
624 625
                       _("Failed to find parent device for %s"),
                       dev->name);
626 627 628 629 630 631 632 633 634 635
        return -1;
    }

    VIR_DEBUG("%s %s: doing a secondary bus reset", dev->id, dev->name);

    /* Save and restore the device's config space; we only do this
     * for the supplied device since we refuse to do a reset if there
     * are multiple devices/functions
     */
    if (pciRead(dev, 0, config_space, PCI_CONF_LEN) < 0) {
636
        virReportError(VIR_ERR_INTERNAL_ERROR,
637
                       _("Failed to read PCI config space for %s"),
638
                       dev->name);
639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654
        goto out;
    }

    /* Read the control register, set the reset flag, wait 200ms,
     * unset the reset flag and wait 200ms.
     */
    ctl = pciRead16(dev, PCI_BRIDGE_CONTROL);

    pciWrite16(parent, PCI_BRIDGE_CONTROL, ctl | PCI_BRIDGE_CTL_RESET);

    usleep(200 * 1000); /* sleep 200ms */

    pciWrite16(parent, PCI_BRIDGE_CONTROL, ctl);

    usleep(200 * 1000); /* sleep 200ms */

655
    if (pciWrite(dev, 0, config_space, PCI_CONF_LEN) < 0) {
656
        virReportError(VIR_ERR_INTERNAL_ERROR,
657 658 659 660
                       _("Failed to restore PCI config space for %s"),
                       dev->name);
        goto out;
    }
661 662
    ret = 0;
out:
663
    pciFreeDevice(parent);
664 665 666 667 668 669 670 671
    return ret;
}

/* Power management reset attempts to reset a device using a
 * D-state transition from D3hot to D0. Note, in detect_pm_reset()
 * above we require the device supports a full internal reset.
 */
static int
672
pciTryPowerManagementReset(pciDevice *dev)
673 674 675 676 677 678 679 680 681
{
    uint8_t config_space[PCI_CONF_LEN];
    uint32_t ctl;

    if (!dev->pci_pm_cap_pos)
        return -1;

    /* Save and restore the device's config space. */
    if (pciRead(dev, 0, &config_space[0], PCI_CONF_LEN) < 0) {
682
        virReportError(VIR_ERR_INTERNAL_ERROR,
683
                       _("Failed to read PCI config space for %s"),
684
                       dev->name);
685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700
        return -1;
    }

    VIR_DEBUG("%s %s: doing a power management reset", dev->id, dev->name);

    ctl = pciRead32(dev, dev->pci_pm_cap_pos + PCI_PM_CTRL);
    ctl &= ~PCI_PM_CTRL_STATE_MASK;

    pciWrite32(dev, dev->pci_pm_cap_pos + PCI_PM_CTRL, ctl|PCI_PM_CTRL_STATE_D3hot);

    usleep(10 * 1000); /* sleep 10ms */

    pciWrite32(dev, dev->pci_pm_cap_pos + PCI_PM_CTRL, ctl|PCI_PM_CTRL_STATE_D0);

    usleep(10 * 1000); /* sleep 10ms */

701
    if (pciWrite(dev, 0, &config_space[0], PCI_CONF_LEN) < 0) {
702
        virReportError(VIR_ERR_INTERNAL_ERROR,
703 704 705 706
                       _("Failed to restore PCI config space for %s"),
                       dev->name);
        return -1;
    }
707 708 709 710 711

    return 0;
}

static int
712
pciInitDevice(pciDevice *dev)
713
{
714 715
    int flr;

716
    if (pciOpenConfig(dev) < 0) {
717
        virReportSystemError(errno,
718 719 720 721 722 723 724
                             _("Failed to open config space file '%s'"),
                             dev->path);
        return -1;
    }

    dev->pcie_cap_pos   = pciFindCapabilityOffset(dev, PCI_CAP_ID_EXP);
    dev->pci_pm_cap_pos = pciFindCapabilityOffset(dev, PCI_CAP_ID_PM);
725
    flr = pciDetectFunctionLevelReset(dev);
726
    if (flr < 0)
727 728
        return flr;
    dev->has_flr        = flr;
729 730 731 732 733 734
    dev->has_pm_reset   = pciDetectPowerManagementReset(dev);
    dev->initted        = 1;
    return 0;
}

int
735
pciResetDevice(pciDevice *dev,
736 737
               pciDeviceList *activeDevs,
               pciDeviceList *inactiveDevs)
738 739 740
{
    int ret = -1;

741
    if (activeDevs && pciDeviceListFind(activeDevs, dev)) {
742
        virReportError(VIR_ERR_INTERNAL_ERROR,
743 744 745 746
                       _("Not resetting active device %s"), dev->name);
        return -1;
    }

747
    if (!dev->initted && pciInitDevice(dev) < 0)
748 749 750 751 752 753 754 755
        return -1;

    /* KVM will perform FLR when starting and stopping
     * a guest, so there is no need for us to do it here.
     */
    if (dev->has_flr)
        return 0;

756 757 758 759 760
    /* If the device supports PCI power management reset,
     * that's the next best thing because it only resets
     * the function, not the whole device.
     */
    if (dev->has_pm_reset)
761
        ret = pciTryPowerManagementReset(dev);
762

763
    /* Bus reset is not an option with the root bus */
764
    if (ret < 0 && dev->bus != 0)
765
        ret = pciTrySecondaryBusReset(dev, inactiveDevs);
766

767 768
    if (ret < 0) {
        virErrorPtr err = virGetLastError();
769
        virReportError(VIR_ERR_INTERNAL_ERROR,
770 771 772 773 774
                       _("Unable to reset PCI device %s: %s"),
                       dev->name,
                       err ? err->message : _("no FLR, PM reset or bus reset available"));
    }

775 776 777
    return ret;
}

778

779 780
static int
pciDriverDir(char **buffer, const char *driver)
781
{
782 783 784 785 786 787 788 789
    VIR_FREE(*buffer);

    if (virAsprintf(buffer, PCI_SYSFS "drivers/%s", driver) < 0) {
        virReportOOMError();
        return -1;
    }

    return 0;
790 791
}

792 793
static int
pciDriverFile(char **buffer, const char *driver, const char *file)
794
{
795 796 797 798 799 800 801 802
    VIR_FREE(*buffer);

    if (virAsprintf(buffer, PCI_SYSFS "drivers/%s/%s", driver, file) < 0) {
        virReportOOMError();
        return -1;
    }

    return 0;
803
}
804

805 806
static int
pciDeviceFile(char **buffer, const char *device, const char *file)
807
{
808 809 810 811 812 813 814 815
    VIR_FREE(*buffer);

    if (virAsprintf(buffer, PCI_SYSFS "devices/%s/%s", device, file) < 0) {
        virReportOOMError();
        return -1;
    }

    return 0;
816
}
817 818


819
static const char *
820
pciFindStubDriver(void)
821
{
822
    char *drvpath = NULL;
823 824 825
    int probed = 0;

recheck:
826 827 828 829 830 831
    if (pciDriverDir(&drvpath, "pci-stub") < 0) {
        return NULL;
    }

    if (virFileExists(drvpath)) {
        VIR_FREE(drvpath);
832
        return "pci-stub";
833 834 835 836 837 838 839 840
    }

    if (pciDriverDir(&drvpath, "pciback") < 0) {
        return NULL;
    }

    if (virFileExists(drvpath)) {
        VIR_FREE(drvpath);
841
        return "pciback";
842 843 844
    }

    VIR_FREE(drvpath);
845 846 847 848 849 850 851 852 853 854 855 856

    if (!probed) {
        const char *const stubprobe[] = { MODPROBE, "pci-stub", NULL };
        const char *const backprobe[] = { MODPROBE, "pciback", NULL };

        probed = 1;
        /*
         * Probing for pci-stub will succeed regardless of whether
         * on native or Xen kernels.
         * On Xen though, we want to prefer pciback, so probe
         * for that first, because that will only work on Xen
         */
857 858
        if (virRun(backprobe, NULL) < 0 &&
            virRun(stubprobe, NULL) < 0) {
859
            char ebuf[1024];
860
            VIR_WARN("failed to load pci-stub or pciback drivers: %s",
861
                     virStrerror(errno, ebuf, sizeof(ebuf)));
862
            return NULL;
863
        }
864 865

        goto recheck;
866 867
    }

868 869 870
    return NULL;
}

871 872 873 874 875 876 877
static int
pciUnbindDeviceFromStub(pciDevice *dev, const char *driver)
{
    int result = -1;
    char *drvdir = NULL;
    char *path = NULL;

E
Eric Blake 已提交
878 879 880
    if (pciDriverDir(&drvdir, driver) < 0)
        goto cleanup;

881 882 883
    if (!dev->unbind_from_stub)
        goto remove_slot;

884 885
    /* If the device is bound to stub, unbind it.
     */
E
Eric Blake 已提交
886
    if (pciDeviceFile(&path, dev->name, "driver") < 0)
887 888 889 890 891 892 893 894 895
        goto cleanup;

    if (virFileExists(drvdir) && virFileLinkPointsTo(path, drvdir)) {
        if (pciDriverFile(&path, driver, "unbind") < 0) {
            goto cleanup;
        }

        if (virFileWriteStr(path, dev->name, 0) < 0) {
            virReportSystemError(errno,
896
                                 _("Failed to unbind PCI device '%s' from %s"),
897 898 899 900
                                 dev->name, driver);
            goto cleanup;
        }
    }
901 902 903 904 905
    dev->unbind_from_stub = 0;

remove_slot:
    if (!dev->remove_slot)
        goto reprobe;
906 907 908 909 910 911 912 913

    /* Xen's pciback.ko wants you to use remove_slot on the specific device */
    if (pciDriverFile(&path, driver, "remove_slot") < 0) {
        goto cleanup;
    }

    if (virFileExists(path) && virFileWriteStr(path, dev->name, 0) < 0) {
        virReportSystemError(errno,
914
                             _("Failed to remove slot for PCI device '%s' from %s"),
915 916 917
                             dev->name, driver);
        goto cleanup;
    }
918 919 920 921 922 923 924
    dev->remove_slot = 0;

reprobe:
    if (!dev->reprobe) {
        result = 0;
        goto cleanup;
    }
925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946

    /* Trigger a re-probe of the device is not in the stub's dynamic
     * ID table. If the stub is available, but 'remove_id' isn't
     * available, then re-probing would just cause the device to be
     * re-bound to the stub.
     */
    if (pciDriverFile(&path, driver, "remove_id") < 0) {
        goto cleanup;
    }

    if (!virFileExists(drvdir) || virFileExists(path)) {
        if (virFileWriteStr(PCI_SYSFS "drivers_probe", dev->name, 0) < 0) {
            virReportSystemError(errno,
                                 _("Failed to trigger a re-probe for PCI device '%s'"),
                                 dev->name);
            goto cleanup;
        }
    }

    result = 0;

cleanup:
947 948 949 950 951
    /* do not do it again */
    dev->unbind_from_stub = 0;
    dev->remove_slot = 0;
    dev->reprobe = 0;

952 953 954 955 956 957
    VIR_FREE(drvdir);
    VIR_FREE(path);

    return result;
}

958 959

static int
960
pciBindDeviceToStub(pciDevice *dev, const char *driver)
961
{
962 963 964
    int result = -1;
    char *drvdir = NULL;
    char *path = NULL;
965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980
    int reprobe = 0;

    /* check whether the device is already bound to a driver */
    if (pciDriverDir(&drvdir, driver) < 0 ||
        pciDeviceFile(&path, dev->name, "driver") < 0) {
        goto cleanup;
    }

    if (virFileExists(path)) {
        if (virFileLinkPointsTo(path, drvdir)) {
            /* The device is already bound to pci-stub */
            result = 0;
            goto cleanup;
        }
        reprobe = 1;
    }
981 982 983 984 985 986 987 988 989

    /* Add the PCI device ID to the stub's dynamic ID table;
     * this is needed to allow us to bind the device to the stub.
     * Note: if the device is not currently bound to any driver,
     * stub will immediately be bound to the device. Also, note
     * that if a new device with this ID is hotplugged, or if a probe
     * is triggered for such a device, it will also be immediately
     * bound by the stub.
     */
990
    if (pciDriverFile(&path, driver, "new_id") < 0) {
991
        goto cleanup;
992 993
    }

994
    if (virFileWriteStr(path, dev->id, 0) < 0) {
995
        virReportSystemError(errno,
996 997
                             _("Failed to add PCI device ID '%s' to %s"),
                             dev->id, driver);
998
        goto cleanup;
999 1000
    }

1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014
    /* check whether the device is bound to pci-stub when we write dev->id to
     * new_id.
     */
    if (pciDriverDir(&drvdir, driver) < 0 ||
        pciDeviceFile(&path, dev->name, "driver") < 0) {
        goto remove_id;
    }

    if (virFileLinkPointsTo(path, drvdir)) {
        dev->unbind_from_stub = 1;
        dev->remove_slot = 1;
        goto remove_id;
    }

1015 1016 1017 1018 1019
    /* If the device is already bound to a driver, unbind it.
     * Note, this will have rather unpleasant side effects if this
     * PCI device happens to be IDE controller for the disk hosting
     * your root filesystem.
     */
1020 1021 1022 1023
    if (pciDeviceFile(&path, dev->name, "driver/unbind") < 0) {
        goto cleanup;
    }

1024 1025 1026 1027 1028 1029 1030 1031
    if (virFileExists(path)) {
        if (virFileWriteStr(path, dev->name, 0) < 0) {
            virReportSystemError(errno,
                                 _("Failed to unbind PCI device '%s'"),
                                 dev->name);
            goto cleanup;
        }
        dev->reprobe = reprobe;
1032 1033
    }

1034 1035
    /* If the device isn't already bound to pci-stub, try binding it now.
     */
1036 1037
    if (pciDriverDir(&drvdir, driver) < 0 ||
        pciDeviceFile(&path, dev->name, "driver") < 0) {
1038
        goto remove_id;
1039 1040
    }

1041 1042
    if (!virFileLinkPointsTo(path, drvdir)) {
        /* Xen's pciback.ko wants you to use new_slot first */
1043
        if (pciDriverFile(&path, driver, "new_slot") < 0) {
1044
            goto remove_id;
1045 1046
        }

1047
        if (virFileExists(path) && virFileWriteStr(path, dev->name, 0) < 0) {
1048
            virReportSystemError(errno,
1049 1050
                                 _("Failed to add slot for PCI device '%s' to %s"),
                                 dev->name, driver);
1051
            goto remove_id;
1052
        }
1053
        dev->remove_slot = 1;
1054 1055

        if (pciDriverFile(&path, driver, "bind") < 0) {
1056
            goto remove_id;
1057 1058
        }

1059
        if (virFileWriteStr(path, dev->name, 0) < 0) {
1060
            virReportSystemError(errno,
1061 1062
                                 _("Failed to bind PCI device '%s' to %s"),
                                 dev->name, driver);
1063
            goto remove_id;
1064
        }
1065
        dev->unbind_from_stub = 1;
1066 1067
    }

1068
remove_id:
1069 1070 1071
    /* If 'remove_id' exists, remove the device id from pci-stub's dynamic
     * ID table so that 'drivers_probe' works below.
     */
1072
    if (pciDriverFile(&path, driver, "remove_id") < 0) {
E
Eric Blake 已提交
1073
        /* We do not remove PCI ID from pci-stub, and we cannot reprobe it */
1074 1075 1076 1077 1078
        if (dev->reprobe) {
            VIR_WARN("Could not remove PCI ID '%s' from %s, and the device "
                     "cannot be probed again.", dev->id, driver);
        }
        dev->reprobe = 0;
1079 1080 1081
        goto cleanup;
    }

1082
    if (virFileExists(path) && virFileWriteStr(path, dev->id, 0) < 0) {
1083
        virReportSystemError(errno,
1084 1085
                             _("Failed to remove PCI ID '%s' from %s"),
                             dev->id, driver);
1086

E
Eric Blake 已提交
1087
        /* remove PCI ID from pci-stub failed, and we cannot reprobe it */
1088 1089 1090 1091 1092
        if (dev->reprobe) {
            VIR_WARN("Failed to remove PCI ID '%s' from %s, and the device "
                     "cannot be probed again.", dev->id, driver);
        }
        dev->reprobe = 0;
1093
        goto cleanup;
1094 1095
    }

1096 1097 1098 1099 1100 1101
    result = 0;

cleanup:
    VIR_FREE(drvdir);
    VIR_FREE(path);

1102 1103 1104 1105
    if (result < 0) {
        pciUnbindDeviceFromStub(dev, driver);
    }

1106
    return result;
1107 1108 1109
}

int
1110 1111 1112
pciDettachDevice(pciDevice *dev,
                 pciDeviceList *activeDevs,
                 pciDeviceList *inactiveDevs)
1113
{
1114
    const char *driver = pciFindStubDriver();
1115
    if (!driver) {
1116
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
1117 1118 1119 1120
                       _("cannot find any PCI stub module"));
        return -1;
    }

1121
    if (activeDevs && pciDeviceListFind(activeDevs, dev)) {
1122
        virReportError(VIR_ERR_INTERNAL_ERROR,
1123 1124 1125 1126
                       _("Not detaching active device %s"), dev->name);
        return -1;
    }

1127 1128 1129 1130 1131 1132 1133 1134 1135 1136
    if (pciBindDeviceToStub(dev, driver) < 0)
        return -1;

    /* Add the dev into list inactiveDevs */
    if (inactiveDevs && !pciDeviceListFind(inactiveDevs, dev)) {
        if (pciDeviceListAdd(inactiveDevs, dev) < 0)
            return -1;
    }

    return 0;
1137 1138 1139
}

int
1140 1141 1142
pciReAttachDevice(pciDevice *dev,
                  pciDeviceList *activeDevs,
                  pciDeviceList *inactiveDevs)
1143
{
1144
    const char *driver = pciFindStubDriver();
1145
    if (!driver) {
1146
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
1147 1148 1149 1150
                       _("cannot find any PCI stub module"));
        return -1;
    }

1151
    if (activeDevs && pciDeviceListFind(activeDevs, dev)) {
1152
        virReportError(VIR_ERR_INTERNAL_ERROR,
1153 1154 1155 1156
                       _("Not reattaching active device %s"), dev->name);
        return -1;
    }

1157 1158 1159 1160 1161 1162 1163 1164
    if (pciUnbindDeviceFromStub(dev, driver) < 0)
        return -1;

    /* Steal the dev from list inactiveDevs */
    if (inactiveDevs)
        pciDeviceListSteal(inactiveDevs, dev);

    return 0;
1165 1166
}

1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200
/* Certain hypervisors (like qemu/kvm) map the PCI bar(s) on
 * the host when doing device passthrough.  This can lead to a race
 * condition where the hypervisor is still cleaning up the device while
 * libvirt is trying to re-attach it to the host device driver.  To avoid
 * this situation, we look through /proc/iomem, and if the hypervisor is
 * still holding onto the bar (denoted by the string in the matcher variable),
 * then we can wait around a bit for that to clear up.
 *
 * A typical /proc/iomem looks like this (snipped for brevity):
 * 00010000-0008efff : System RAM
 * 0008f000-0008ffff : reserved
 * ...
 * 00100000-cc9fcfff : System RAM
 *   00200000-00483d3b : Kernel code
 *   00483d3c-005c88df : Kernel data
 * cc9fd000-ccc71fff : ACPI Non-volatile Storage
 * ...
 * d0200000-d02fffff : PCI Bus #05
 *   d0200000-d021ffff : 0000:05:00.0
 *     d0200000-d021ffff : e1000e
 *   d0220000-d023ffff : 0000:05:00.0
 *     d0220000-d023ffff : e1000e
 * ...
 * f0000000-f0003fff : 0000:00:1b.0
 *   f0000000-f0003fff : kvm_assigned_device
 *
 * Returns 0 if we are clear to continue, and 1 if the hypervisor is still
 * holding onto the resource.
 */
int
pciWaitForDeviceCleanup(pciDevice *dev, const char *matcher)
{
    FILE *fp;
    char line[160];
1201
    char *tmp;
1202
    unsigned long long start, end;
1203
    unsigned int domain, bus, slot, function;
1204 1205 1206 1207 1208 1209 1210 1211 1212 1213
    int in_matching_device;
    int ret;
    size_t match_depth;

    fp = fopen("/proc/iomem", "r");
    if (!fp) {
        /* If we failed to open iomem, we just basically ignore the error.  The
         * unbind might succeed anyway, and besides, it's very likely we have
         * no way to report the error
         */
1214
        VIR_DEBUG("Failed to open /proc/iomem, trying to continue anyway");
1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230
        return 0;
    }

    ret = 0;
    in_matching_device = 0;
    match_depth = 0;
    while (fgets(line, sizeof(line), fp) != 0) {
        /* the logic here is a bit confusing.  For each line, we look to
         * see if it matches the domain:bus:slot.function we were given.
         * If this line matches the DBSF, then any subsequent lines indented
         * by 2 spaces are the PCI regions for this device.  It's also
         * possible that none of the PCI regions are currently mapped, in
         * which case we have no indented regions.  This code handles all
         * of these situations
         */
        if (in_matching_device && (strspn(line, " ") == (match_depth + 2))) {
1231 1232 1233 1234 1235 1236
            /* expected format: <start>-<end> : <suffix> */
            if (/* start */
                virStrToLong_ull(line, &tmp, 16, &start) < 0 || *tmp != '-' ||
                /* end */
                virStrToLong_ull(tmp + 1, &tmp, 16, &end) < 0 ||
                (tmp = STRSKIP(tmp, " : ")) == NULL)
1237 1238
                continue;

1239
            if (STRPREFIX(tmp, matcher)) {
1240 1241 1242 1243 1244 1245 1246
                ret = 1;
                break;
            }
        }
        else {
            in_matching_device = 0;

1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260
            /* expected format: <start>-<end> : <domain>:<bus>:<slot>.<function> */
            if (/* start */
                virStrToLong_ull(line, &tmp, 16, &start) < 0 || *tmp != '-' ||
                /* end */
                virStrToLong_ull(tmp + 1, &tmp, 16, &end) < 0 ||
                (tmp = STRSKIP(tmp, " : ")) == NULL ||
                /* domain */
                virStrToLong_ui(tmp, &tmp, 16, &domain) < 0 || *tmp != ':' ||
                /* bus */
                virStrToLong_ui(tmp + 1, &tmp, 16, &bus) < 0 || *tmp != ':' ||
                /* slot */
                virStrToLong_ui(tmp + 1, &tmp, 16, &slot) < 0 || *tmp != '.' ||
                /* function */
                virStrToLong_ui(tmp + 1, &tmp, 16, &function) < 0 || *tmp != '\n')
1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
                continue;

            if (domain != dev->domain || bus != dev->bus || slot != dev->slot ||
                function != dev->function)
                continue;
            in_matching_device = 1;
            match_depth = strspn(line, " ");
        }
    }

E
Eric Blake 已提交
1271
    VIR_FORCE_FCLOSE(fp);
1272 1273 1274 1275

    return ret;
}

1276 1277 1278
static char *
pciReadDeviceID(pciDevice *dev, const char *id_name)
{
1279
    char *path = NULL;
1280 1281
    char *id_str;

1282 1283 1284
    if (pciDeviceFile(&path, dev->name, id_name) < 0) {
        return NULL;
    }
1285 1286

    /* ID string is '0xNNNN\n' ... i.e. 7 bytes */
1287 1288
    if (virFileReadAll(path, 7, &id_str) < 0) {
        VIR_FREE(path);
1289
        return NULL;
1290 1291 1292
    }

    VIR_FREE(path);
1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305

    /* Check for 0x suffix */
    if (id_str[0] != '0' || id_str[1] != 'x') {
        VIR_FREE(id_str);
        return NULL;
    }

    /* Chop off the newline; we know the string is 7 bytes */
    id_str[6] = '\0';

    return id_str;
}

1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329
int
pciGetDeviceAddrString(unsigned domain,
                       unsigned bus,
                       unsigned slot,
                       unsigned function,
                       char **pciConfigAddr)
{
    pciDevice *dev = NULL;
    int ret = -1;

    dev = pciGetDevice(domain, bus, slot, function);
    if (dev != NULL) {
        if ((*pciConfigAddr = strdup(dev->name)) == NULL) {
            virReportOOMError();
            goto cleanup;
        }
        ret = 0;
    }

cleanup:
    pciFreeDevice(dev);
    return ret;
}

1330
pciDevice *
1331
pciGetDevice(unsigned domain,
1332 1333 1334 1335 1336
             unsigned bus,
             unsigned slot,
             unsigned function)
{
    pciDevice *dev;
E
Eric Blake 已提交
1337 1338
    char *vendor = NULL;
    char *product = NULL;
1339 1340

    if (VIR_ALLOC(dev) < 0) {
1341
        virReportOOMError();
1342 1343 1344
        return NULL;
    }

1345
    dev->fd       = -1;
1346 1347 1348 1349 1350
    dev->domain   = domain;
    dev->bus      = bus;
    dev->slot     = slot;
    dev->function = function;

E
Eric Blake 已提交
1351 1352 1353
    if (snprintf(dev->name, sizeof(dev->name), "%.4x:%.2x:%.2x.%.1x",
                 dev->domain, dev->bus, dev->slot,
                 dev->function) >= sizeof(dev->name)) {
1354
        virReportError(VIR_ERR_INTERNAL_ERROR,
E
Eric Blake 已提交
1355 1356
                       _("dev->name buffer overflow: %.4x:%.2x:%.2x.%.1x"),
                       dev->domain, dev->bus, dev->slot, dev->function);
E
Eric Blake 已提交
1357
        goto error;
E
Eric Blake 已提交
1358 1359 1360 1361
    }
    if (virAsprintf(&dev->path, PCI_SYSFS "devices/%s/config",
                    dev->name) < 0) {
        virReportOOMError();
E
Eric Blake 已提交
1362
        goto error;
E
Eric Blake 已提交
1363
    }
1364

1365 1366 1367 1368
    if (access(dev->path, F_OK) != 0) {
        virReportSystemError(errno,
                             _("Device %s not found: could not access %s"),
                             dev->name, dev->path);
E
Eric Blake 已提交
1369
        goto error;
1370 1371
    }

1372 1373 1374 1375
    vendor  = pciReadDeviceID(dev, "vendor");
    product = pciReadDeviceID(dev, "device");

    if (!vendor || !product) {
1376
        virReportError(VIR_ERR_INTERNAL_ERROR,
1377 1378
                       _("Failed to read product/vendor ID for %s"),
                       dev->name);
E
Eric Blake 已提交
1379
        goto error;
1380 1381 1382
    }

    /* strings contain '0x' prefix */
E
Eric Blake 已提交
1383 1384
    if (snprintf(dev->id, sizeof(dev->id), "%s %s", &vendor[2],
                 &product[2]) >= sizeof(dev->id)) {
1385
        virReportError(VIR_ERR_INTERNAL_ERROR,
E
Eric Blake 已提交
1386 1387
                       _("dev->id buffer overflow: %s %s"),
                       &vendor[2], &product[2]);
E
Eric Blake 已提交
1388
        goto error;
E
Eric Blake 已提交
1389
    }
1390 1391 1392

    VIR_DEBUG("%s %s: initialized", dev->id, dev->name);

E
Eric Blake 已提交
1393 1394 1395
cleanup:
    VIR_FREE(product);
    VIR_FREE(vendor);
1396
    return dev;
E
Eric Blake 已提交
1397 1398 1399 1400 1401

error:
    pciFreeDevice(dev);
    dev = NULL;
    goto cleanup;
1402 1403 1404
}

void
1405
pciFreeDevice(pciDevice *dev)
1406
{
1407 1408
    if (!dev)
        return;
1409
    VIR_DEBUG("%s %s: freeing", dev->id, dev->name);
1410
    pciCloseConfig(dev);
E
Eric Blake 已提交
1411
    VIR_FREE(dev->path);
1412 1413
    VIR_FREE(dev);
}
1414

1415 1416 1417 1418 1419 1420
const char *
pciDeviceGetName(pciDevice *dev)
{
    return dev->name;
}

1421 1422 1423 1424 1425 1426 1427 1428 1429 1430
void pciDeviceSetManaged(pciDevice *dev, unsigned managed)
{
    dev->managed = !!managed;
}

unsigned pciDeviceGetManaged(pciDevice *dev)
{
    return dev->managed;
}

1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466
unsigned
pciDeviceGetUnbindFromStub(pciDevice *dev)
{
    return dev->unbind_from_stub;
}

void
pciDeviceSetUnbindFromStub(pciDevice *dev, unsigned unbind)
{
    dev->unbind_from_stub = !!unbind;
}

unsigned
pciDeviceGetRemoveSlot(pciDevice *dev)
{
    return dev->remove_slot;
}

void
pciDeviceSetRemoveSlot(pciDevice *dev, unsigned remove_slot)
{
    dev->remove_slot = !!remove_slot;
}

unsigned
pciDeviceGetReprobe(pciDevice *dev)
{
    return dev->reprobe;
}

void
pciDeviceSetReprobe(pciDevice *dev, unsigned reprobe)
{
    dev->reprobe = !!reprobe;
}

1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478
void
pciDeviceSetUsedBy(pciDevice *dev, const char *name)
{
    dev->used_by = name;
}

const char *
pciDeviceGetUsedBy(pciDevice *dev)
{
    return dev->used_by;
}

1479 1480 1481 1482 1483 1484 1485 1486
void pciDeviceReAttachInit(pciDevice *pci)
{
    pci->unbind_from_stub = 1;
    pci->remove_slot = 1;
    pci->reprobe = 1;
}


1487
pciDeviceList *
1488
pciDeviceListNew(void)
1489 1490 1491 1492
{
    pciDeviceList *list;

    if (VIR_ALLOC(list) < 0) {
1493
        virReportOOMError();
1494 1495 1496 1497 1498 1499 1500
        return NULL;
    }

    return list;
}

void
1501
pciDeviceListFree(pciDeviceList *list)
1502 1503 1504 1505 1506 1507 1508
{
    int i;

    if (!list)
        return;

    for (i = 0; i < list->count; i++) {
1509
        pciFreeDevice(list->devs[i]);
1510 1511 1512 1513 1514 1515 1516 1517 1518
        list->devs[i] = NULL;
    }

    list->count = 0;
    VIR_FREE(list->devs);
    VIR_FREE(list);
}

int
1519
pciDeviceListAdd(pciDeviceList *list,
1520 1521 1522
                 pciDevice *dev)
{
    if (pciDeviceListFind(list, dev)) {
1523
        virReportError(VIR_ERR_INTERNAL_ERROR,
1524 1525 1526 1527 1528
                       _("Device %s is already in use"), dev->name);
        return -1;
    }

    if (VIR_REALLOC_N(list->devs, list->count+1) < 0) {
1529
        virReportOOMError();
1530 1531 1532 1533 1534 1535 1536 1537
        return -1;
    }

    list->devs[list->count++] = dev;

    return 0;
}

1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551
pciDevice *
pciDeviceListGet(pciDeviceList *list,
                 int idx)
{
    if (idx >= list->count)
        return NULL;
    if (idx < 0)
        return NULL;

    return list->devs[idx];
}

int
pciDeviceListCount(pciDeviceList *list)
1552
{
1553 1554 1555 1556
    return list->count;
}

pciDevice *
1557
pciDeviceListSteal(pciDeviceList *list,
1558 1559 1560
                   pciDevice *dev)
{
    pciDevice *ret = NULL;
1561 1562 1563 1564 1565 1566 1567 1568 1569
    int i;

    for (i = 0; i < list->count; i++) {
        if (list->devs[i]->domain   != dev->domain ||
            list->devs[i]->bus      != dev->bus    ||
            list->devs[i]->slot     != dev->slot   ||
            list->devs[i]->function != dev->function)
            continue;

1570
        ret = list->devs[i];
1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582

        if (i != --list->count)
            memmove(&list->devs[i],
                    &list->devs[i+1],
                    sizeof(*list->devs) * (list->count-i));

        if (VIR_REALLOC_N(list->devs, list->count) < 0) {
            ; /* not fatal */
        }

        break;
    }
1583 1584 1585 1586
    return ret;
}

void
1587
pciDeviceListDel(pciDeviceList *list,
1588 1589
                 pciDevice *dev)
{
1590
    pciDevice *ret = pciDeviceListSteal(list, dev);
1591
    if (ret)
1592
        pciFreeDevice(ret);
1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607
}

pciDevice *
pciDeviceListFind(pciDeviceList *list, pciDevice *dev)
{
    int i;

    for (i = 0; i < list->count; i++)
        if (list->devs[i]->domain   == dev->domain &&
            list->devs[i]->bus      == dev->bus    &&
            list->devs[i]->slot     == dev->slot   &&
            list->devs[i]->function == dev->function)
            return list->devs[i];
    return NULL;
}
1608 1609


1610
int pciDeviceFileIterate(pciDevice *dev,
1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621
                         pciDeviceFileActor actor,
                         void *opaque)
{
    char *pcidir = NULL;
    char *file = NULL;
    DIR *dir = NULL;
    int ret = -1;
    struct dirent *ent;

    if (virAsprintf(&pcidir, "/sys/bus/pci/devices/%04x:%02x:%02x.%x",
                    dev->domain, dev->bus, dev->slot, dev->function) < 0) {
1622
        virReportOOMError();
1623 1624 1625 1626
        goto cleanup;
    }

    if (!(dir = opendir(pcidir))) {
1627
        virReportSystemError(errno,
1628 1629 1630 1631 1632 1633
                             _("cannot open %s"), pcidir);
        goto cleanup;
    }

    while ((ent = readdir(dir)) != NULL) {
        /* Device assignment requires:
A
Alex Williamson 已提交
1634 1635
         *   $PCIDIR/config, $PCIDIR/resource, $PCIDIR/resourceNNN,
         *   $PCIDIR/rom, $PCIDIR/reset
1636 1637 1638
         */
        if (STREQ(ent->d_name, "config") ||
            STRPREFIX(ent->d_name, "resource") ||
A
Alex Williamson 已提交
1639 1640
            STREQ(ent->d_name, "rom") ||
            STREQ(ent->d_name, "reset")) {
1641
            if (virAsprintf(&file, "%s/%s", pcidir, ent->d_name) < 0) {
1642
                virReportOOMError();
1643 1644
                goto cleanup;
            }
1645
            if ((actor)(dev, file, opaque) < 0)
1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
                goto cleanup;

            VIR_FREE(file);
        }
    }

    ret = 0;

cleanup:
    if (dir)
        closedir(dir);
    VIR_FREE(file);
    VIR_FREE(pcidir);
    return ret;
}
J
Jiri Denemark 已提交
1661 1662

static int
1663
pciDeviceDownstreamLacksACS(pciDevice *dev)
J
Jiri Denemark 已提交
1664 1665 1666 1667 1668
{
    uint16_t flags;
    uint16_t ctrl;
    unsigned int pos;

1669
    if (!dev->initted && pciInitDevice(dev) < 0)
J
Jiri Denemark 已提交
1670 1671 1672
        return -1;

    pos = dev->pcie_cap_pos;
1673
    if (!pos || pciRead16(dev, PCI_CLASS_DEVICE) != PCI_CLASS_BRIDGE_PCI)
J
Jiri Denemark 已提交
1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696
        return 0;

    flags = pciRead16(dev, pos + PCI_EXP_FLAGS);
    if (((flags & PCI_EXP_FLAGS_TYPE) >> 4) != PCI_EXP_TYPE_DOWNSTREAM)
        return 0;

    pos = pciFindExtendedCapabilityOffset(dev, PCI_EXT_CAP_ID_ACS);
    if (!pos) {
        VIR_DEBUG("%s %s: downstream port lacks ACS", dev->id, dev->name);
        return 1;
    }

    ctrl = pciRead16(dev, pos + PCI_EXT_ACS_CTRL);
    if ((ctrl & PCI_EXT_CAP_ACS_ENABLED) != PCI_EXT_CAP_ACS_ENABLED) {
        VIR_DEBUG("%s %s: downstream port has ACS disabled",
                  dev->id, dev->name);
        return 1;
    }

    return 0;
}

static int
1697
pciDeviceIsBehindSwitchLackingACS(pciDevice *dev)
J
Jiri Denemark 已提交
1698 1699 1700
{
    pciDevice *parent;

1701 1702
    if (pciGetParentDevice(dev, &parent) < 0)
        return -1;
1703 1704 1705 1706 1707 1708 1709 1710
    if (!parent) {
        /* if we have no parent, and this is the root bus, ACS doesn't come
         * into play since devices on the root bus can't P2P without going
         * through the root IOMMU.
         */
        if (dev->bus == 0)
            return 0;
        else {
1711
            virReportError(VIR_ERR_INTERNAL_ERROR,
1712 1713 1714 1715
                           _("Failed to find parent device for %s"),
                           dev->name);
            return -1;
        }
J
Jiri Denemark 已提交
1716 1717 1718 1719 1720 1721 1722 1723 1724
    }

    /* XXX we should rather fail when we can't find device's parent and
     * stop the loop when we get to root instead of just stopping when no
     * parent can be found
     */
    do {
        pciDevice *tmp;
        int acs;
1725
        int ret;
J
Jiri Denemark 已提交
1726

1727
        acs = pciDeviceDownstreamLacksACS(parent);
J
Jiri Denemark 已提交
1728 1729

        if (acs) {
1730
            pciFreeDevice(parent);
J
Jiri Denemark 已提交
1731 1732 1733 1734 1735 1736 1737
            if (acs < 0)
                return -1;
            else
                return 1;
        }

        tmp = parent;
1738
        ret = pciGetParentDevice(parent, &parent);
1739
        pciFreeDevice(tmp);
1740 1741
        if (ret < 0)
            return -1;
J
Jiri Denemark 已提交
1742 1743 1744 1745 1746
    } while (parent);

    return 0;
}

1747
int pciDeviceIsAssignable(pciDevice *dev,
J
Jiri Denemark 已提交
1748 1749 1750 1751 1752 1753 1754 1755 1756
                          int strict_acs_check)
{
    int ret;

    /* XXX This could be a great place to actually check that a non-managed
     * device isn't in use, e.g. by checking that device is either un-bound
     * or bound to a stub driver.
     */

1757
    ret = pciDeviceIsBehindSwitchLackingACS(dev);
J
Jiri Denemark 已提交
1758 1759 1760 1761 1762 1763 1764 1765
    if (ret < 0)
        return 0;

    if (ret) {
        if (!strict_acs_check) {
            VIR_DEBUG("%s %s: strict ACS check disabled; device assignment allowed",
                      dev->id, dev->name);
        } else {
1766
            virReportError(VIR_ERR_INTERNAL_ERROR,
J
Jiri Denemark 已提交
1767 1768 1769 1770 1771 1772 1773 1774 1775
                           _("Device %s is behind a switch lacking ACS and "
                             "cannot be assigned"),
                           dev->name);
            return 0;
        }
    }

    return 1;
}
1776

1777 1778
#ifdef __linux__

1779
/*
1780
 * returns true if equal
1781
 */
1782
static bool
1783 1784 1785 1786 1787 1788 1789 1790 1791
pciConfigAddressEqual(struct pci_config_address *bdf1,
                      struct pci_config_address *bdf2)
{
    return ((bdf1->domain == bdf2->domain) &&
            (bdf1->bus == bdf2->bus) &&
            (bdf1->slot == bdf2->slot) &&
            (bdf1->function == bdf2->function));
}

1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859
static int
logStrToLong_ui(char const *s,
                char **end_ptr,
                int base,
                unsigned int *result)
{
    int ret = 0;

    ret = virStrToLong_ui(s, end_ptr, base, result);
    if (ret != 0) {
        VIR_ERROR(_("Failed to convert '%s' to unsigned int"), s);
    } else {
        VIR_DEBUG("Converted '%s' to unsigned int %u", s, *result);
    }

    return ret;
}

static int
pciParsePciConfigAddress(char *address,
                         struct pci_config_address *bdf)
{
    char *p = NULL;
    int ret = -1;

    if ((address == NULL) || (logStrToLong_ui(address, &p, 16,
                                              &bdf->domain) == -1)) {
        goto out;
    }

    if ((p == NULL) || (logStrToLong_ui(p+1, &p, 16,
                                        &bdf->bus) == -1)) {
        goto out;
    }

    if ((p == NULL) || (logStrToLong_ui(p+1, &p, 16,
                                        &bdf->slot) == -1)) {
        goto out;
    }

    if ((p == NULL) || (logStrToLong_ui(p+1, &p, 16,
                                        &bdf->function) == -1)) {
        goto out;
    }

    ret = 0;

out:
    return ret;
}

static int
pciGetPciConfigAddressFromSysfsDeviceLink(const char *device_link,
                                          struct pci_config_address **bdf)
{
    char *config_address = NULL;
    char *device_path = NULL;
    char errbuf[64];
    int ret = -1;

    VIR_DEBUG("Attempting to resolve device path from device link '%s'",
              device_link);

    if (!virFileExists(device_link)) {
        VIR_DEBUG("sysfs_path '%s' does not exist", device_link);
        return ret;
    }

1860
    device_path = canonicalize_file_name(device_link);
1861 1862
    if (device_path == NULL) {
        memset(errbuf, '\0', sizeof(errbuf));
1863 1864 1865
        virReportSystemError(errno,
                             _("Failed to resolve device link '%s'"),
                             device_link);
1866 1867 1868 1869 1870 1871 1872 1873 1874 1875
        return ret;
    }

    config_address = basename(device_path);
    if (VIR_ALLOC(*bdf) != 0) {
        virReportOOMError();
        goto out;
    }

    if (pciParsePciConfigAddress(config_address, *bdf) != 0) {
1876
        virReportError(VIR_ERR_INTERNAL_ERROR,
1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942
                       _("Failed to parse PCI config address '%s'"),
                       config_address);
        VIR_FREE(*bdf);
        goto out;
    }

    VIR_DEBUG("pci_config_address %.4x:%.2x:%.2x.%.1x",
              (*bdf)->domain,
              (*bdf)->bus,
              (*bdf)->slot,
              (*bdf)->function);

    ret = 0;

out:
    VIR_FREE(device_path);

    return ret;
}

/*
 * Returns Physical function given a virtual function
 */
int
pciGetPhysicalFunction(const char *vf_sysfs_path,
                       struct pci_config_address **physical_function)
{
    int ret = -1;
    char *device_link = NULL;

    VIR_DEBUG("Attempting to get SR IOV physical function for device "
              "with sysfs path '%s'", vf_sysfs_path);

    if (virBuildPath(&device_link, vf_sysfs_path, "physfn") == -1) {
        virReportOOMError();
        return ret;
    } else {
        ret = pciGetPciConfigAddressFromSysfsDeviceLink(device_link,
                                                        physical_function);
    }

    VIR_FREE(device_link);

    return ret;
}

/*
 * Returns virtual functions of a physical function
 */
int
pciGetVirtualFunctions(const char *sysfs_path,
                       struct pci_config_address ***virtual_functions,
                       unsigned int *num_virtual_functions)
{
    int ret = -1;
    DIR *dir = NULL;
    struct dirent *entry = NULL;
    char *device_link = NULL;
    char errbuf[64];

    VIR_DEBUG("Attempting to get SR IOV virtual functions for device"
              "with sysfs path '%s'", sysfs_path);

    dir = opendir(sysfs_path);
    if (dir == NULL) {
        memset(errbuf, '\0', sizeof(errbuf));
1943 1944 1945
        virReportSystemError(errno,
                             _("Failed to open dir '%s'"),
                             sysfs_path);
1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972
        return ret;
    }

    *virtual_functions = NULL;
    *num_virtual_functions = 0;
    while ((entry = readdir(dir))) {
        if (STRPREFIX(entry->d_name, "virtfn")) {

            if (virBuildPath(&device_link, sysfs_path, entry->d_name) == -1) {
                virReportOOMError();
                goto out;
            }

            VIR_DEBUG("Number of virtual functions: %d",
                      *num_virtual_functions);
            if (VIR_REALLOC_N(*virtual_functions,
                (*num_virtual_functions) + 1) != 0) {
                virReportOOMError();
                VIR_FREE(device_link);
                goto out;
            }

            if (pciGetPciConfigAddressFromSysfsDeviceLink(device_link,
                &((*virtual_functions)[*num_virtual_functions])) !=
                SRIOV_FOUND) {
                /* We should not get back SRIOV_NOT_FOUND in this
                 * case, so if we do, it's an error. */
1973
                virReportError(VIR_ERR_INTERNAL_ERROR,
1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992
                               _("Failed to get SR IOV function from device "
                               "link '%s'"), device_link);
                VIR_FREE(device_link);
                goto out;
            } else {
                (*num_virtual_functions)++;
            }
            VIR_FREE(device_link);
        }
    }

    ret = 0;

out:
    if (dir)
        closedir(dir);

    return ret;
}
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029

/*
 * Returns 1 if vf device is a virtual function, 0 if not, -1 on error
 */
int
pciDeviceIsVirtualFunction(const char *vf_sysfs_device_link)
{
    char *vf_sysfs_physfn_link = NULL;
    int ret = -1;

    if (virAsprintf(&vf_sysfs_physfn_link, "%s/physfn",
        vf_sysfs_device_link) < 0) {
        virReportOOMError();
        return ret;
    }

    ret = virFileExists(vf_sysfs_physfn_link);

    VIR_FREE(vf_sysfs_physfn_link);

    return ret;
}

/*
 * Returns the sriov virtual function index of vf given its pf
 */
int
pciGetVirtualFunctionIndex(const char *pf_sysfs_device_link,
                           const char *vf_sysfs_device_link,
                           int *vf_index)
{
    int ret = -1, i;
    unsigned int num_virt_fns = 0;
    struct pci_config_address *vf_bdf = NULL;
    struct pci_config_address **virt_fns = NULL;

    if (pciGetPciConfigAddressFromSysfsDeviceLink(vf_sysfs_device_link,
2030
        &vf_bdf) < 0)
2031 2032 2033
        return ret;

    if (pciGetVirtualFunctions(pf_sysfs_device_link, &virt_fns,
2034
        &num_virt_fns) < 0) {
2035
        virReportError(VIR_ERR_INTERNAL_ERROR,
2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054
                       _("Error getting physical function's '%s' "
                      "virtual_functions"), pf_sysfs_device_link);
        goto out;
    }

    for (i = 0; i < num_virt_fns; i++) {
         if (pciConfigAddressEqual(vf_bdf, virt_fns[i])) {
             *vf_index = i;
             ret = 0;
             break;
         }
    }

out:

    /* free virtual functions */
    for (i = 0; i < num_virt_fns; i++)
         VIR_FREE(virt_fns[i]);

A
ajia@redhat.com 已提交
2055
    VIR_FREE(virt_fns);
2056 2057 2058 2059 2060
    VIR_FREE(vf_bdf);

    return ret;
}

2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076
/*
 * Returns a path to the PCI sysfs file given the BDF of the PCI function
 */

int
pciSysfsFile(char *pciDeviceName, char **pci_sysfs_device_link)
{
    if (virAsprintf(pci_sysfs_device_link, PCI_SYSFS "devices/%s",
                    pciDeviceName) < 0) {
        virReportOOMError();
        return -1;
    }

    return 0;
}

R
Roopa Prabhu 已提交
2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090
int
pciConfigAddressToSysfsFile(struct pci_config_address *dev,
                            char **pci_sysfs_device_link)
{
    if (virAsprintf(pci_sysfs_device_link,
                    PCI_SYSFS "devices/%04x:%02x:%02x.%x", dev->domain,
                    dev->bus, dev->slot, dev->function) < 0) {
        virReportOOMError();
        return -1;
    }

    return 0;
}

2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118
/*
 * Returns the network device name of a pci device
 */
int
pciDeviceNetName(char *device_link_sysfs_path, char **netname)
{
     char *pcidev_sysfs_net_path = NULL;
     int ret = -1;
     DIR *dir = NULL;
     struct dirent *entry = NULL;

     if (virBuildPath(&pcidev_sysfs_net_path, device_link_sysfs_path,
         "net") == -1) {
         virReportOOMError();
         return -1;
     }

     dir = opendir(pcidev_sysfs_net_path);
     if (dir == NULL)
         goto out;

     while ((entry = readdir(dir))) {
            if (STREQ(entry->d_name, ".") ||
                STREQ(entry->d_name, ".."))
                continue;

            /* Assume a single directory entry */
            *netname = strdup(entry->d_name);
2119 2120 2121 2122
            if (!*netname)
                virReportOOMError();
            else
                ret = 0;
2123 2124 2125 2126 2127 2128 2129 2130 2131 2132
            break;
     }

     closedir(dir);

out:
     VIR_FREE(pcidev_sysfs_net_path);

     return ret;
}
R
Roopa Prabhu 已提交
2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164

int
pciDeviceGetVirtualFunctionInfo(const char *vf_sysfs_device_path,
                                char **pfname, int *vf_index)
{
    struct pci_config_address *pf_config_address = NULL;
    char *pf_sysfs_device_path = NULL;
    int ret = -1;

    if (pciGetPhysicalFunction(vf_sysfs_device_path, &pf_config_address) < 0)
        return ret;

    if (pciConfigAddressToSysfsFile(pf_config_address,
                                    &pf_sysfs_device_path) < 0) {

        VIR_FREE(pf_config_address);
        return ret;
    }

    if (pciGetVirtualFunctionIndex(pf_sysfs_device_path, vf_sysfs_device_path,
        vf_index) < 0)
        goto cleanup;

    ret = pciDeviceNetName(pf_sysfs_device_path, pfname);

cleanup:
    VIR_FREE(pf_config_address);
    VIR_FREE(pf_sysfs_device_path);

    return ret;
}

2165
#else
2166 2167
static const char *unsupported = N_("not supported on non-linux platforms");

2168
int
2169 2170
pciGetPhysicalFunction(const char *vf_sysfs_path ATTRIBUTE_UNUSED,
              struct pci_config_address **physical_function ATTRIBUTE_UNUSED)
2171
{
2172
    virReportError(VIR_ERR_INTERNAL_ERROR, "%s", _(unsupported));
2173 2174 2175 2176
    return -1;
}

int
2177 2178 2179
pciGetVirtualFunctions(const char *sysfs_path ATTRIBUTE_UNUSED,
             struct pci_config_address ***virtual_functions ATTRIBUTE_UNUSED,
             unsigned int *num_virtual_functions ATTRIBUTE_UNUSED)
2180
{
2181
    virReportError(VIR_ERR_INTERNAL_ERROR, "%s", _(unsupported));
2182 2183
    return -1;
}
2184 2185

int
2186
pciDeviceIsVirtualFunction(const char *vf_sysfs_device_link ATTRIBUTE_UNUSED)
2187
{
2188
    virReportError(VIR_ERR_INTERNAL_ERROR, "%s", _(unsupported));
2189 2190 2191 2192
    return -1;
}

int
2193 2194 2195
pciGetVirtualFunctionIndex(const char *pf_sysfs_device_link ATTRIBUTE_UNUSED,
                           const char *vf_sysfs_device_link ATTRIBUTE_UNUSED,
                           int *vf_index ATTRIBUTE_UNUSED)
2196
{
2197
    virReportError(VIR_ERR_INTERNAL_ERROR, "%s", _(unsupported));
2198 2199 2200 2201
    return -1;

}

2202 2203 2204 2205
int
pciConfigAddressToSysfsFile(struct pci_config_address *dev ATTRIBUTE_UNUSED,
                            char **pci_sysfs_device_link ATTRIBUTE_UNUSED)
{
2206
    virReportError(VIR_ERR_INTERNAL_ERROR, "%s", _(unsupported));
2207 2208 2209
    return -1;
}

2210
int
2211 2212
pciDeviceNetName(char *device_link_sysfs_path ATTRIBUTE_UNUSED,
                 char **netname ATTRIBUTE_UNUSED)
2213
{
2214
    virReportError(VIR_ERR_INTERNAL_ERROR, "%s", _(unsupported));
2215 2216
    return -1;
}
R
Roopa Prabhu 已提交
2217 2218 2219

int
pciDeviceGetVirtualFunctionInfo(const char *vf_sysfs_device_path ATTRIBUTE_UNUSED,
2220 2221
                                char **pfname ATTRIBUTE_UNUSED,
                                int *vf_index ATTRIBUTE_UNUSED)
R
Roopa Prabhu 已提交
2222
{
2223
    virReportError(VIR_ERR_INTERNAL_ERROR, "%s", _(unsupported));
R
Roopa Prabhu 已提交
2224 2225
    return -1;
}
2226
#endif /* __linux__ */