linux/drivers/pci/pci.c
<<
>>
Prefs
   1/*
   2 *      PCI Bus Services, see include/linux/pci.h for further explanation.
   3 *
   4 *      Copyright 1993 -- 1997 Drew Eckhardt, Frederic Potter,
   5 *      David Mosberger-Tang
   6 *
   7 *      Copyright 1997 -- 2000 Martin Mares <mj@ucw.cz>
   8 */
   9
  10#include <linux/kernel.h>
  11#include <linux/delay.h>
  12#include <linux/init.h>
  13#include <linux/of.h>
  14#include <linux/of_pci.h>
  15#include <linux/pci.h>
  16#include <linux/pm.h>
  17#include <linux/slab.h>
  18#include <linux/module.h>
  19#include <linux/spinlock.h>
  20#include <linux/string.h>
  21#include <linux/log2.h>
  22#include <linux/pci-aspm.h>
  23#include <linux/pm_wakeup.h>
  24#include <linux/interrupt.h>
  25#include <linux/device.h>
  26#include <linux/pm_runtime.h>
  27#include <linux/pci_hotplug.h>
  28#include <asm/setup.h>
  29#include <linux/aer.h>
  30#include "pci.h"
  31
  32const char *pci_power_names[] = {
  33        "error", "D0", "D1", "D2", "D3hot", "D3cold", "unknown",
  34};
  35EXPORT_SYMBOL_GPL(pci_power_names);
  36
  37int isa_dma_bridge_buggy;
  38EXPORT_SYMBOL(isa_dma_bridge_buggy);
  39
  40int pci_pci_problems;
  41EXPORT_SYMBOL(pci_pci_problems);
  42
  43unsigned int pci_pm_d3_delay;
  44
  45static void pci_pme_list_scan(struct work_struct *work);
  46
  47static LIST_HEAD(pci_pme_list);
  48static DEFINE_MUTEX(pci_pme_list_mutex);
  49static DECLARE_DELAYED_WORK(pci_pme_work, pci_pme_list_scan);
  50
  51struct pci_pme_device {
  52        struct list_head list;
  53        struct pci_dev *dev;
  54};
  55
  56#define PME_TIMEOUT 1000 /* How long between PME checks */
  57
  58static void pci_dev_d3_sleep(struct pci_dev *dev)
  59{
  60        unsigned int delay = dev->d3_delay;
  61
  62        if (delay < pci_pm_d3_delay)
  63                delay = pci_pm_d3_delay;
  64
  65        msleep(delay);
  66}
  67
  68#ifdef CONFIG_PCI_DOMAINS
  69int pci_domains_supported = 1;
  70#endif
  71
  72#define DEFAULT_CARDBUS_IO_SIZE         (256)
  73#define DEFAULT_CARDBUS_MEM_SIZE        (64*1024*1024)
  74/* pci=cbmemsize=nnM,cbiosize=nn can override this */
  75unsigned long pci_cardbus_io_size = DEFAULT_CARDBUS_IO_SIZE;
  76unsigned long pci_cardbus_mem_size = DEFAULT_CARDBUS_MEM_SIZE;
  77
  78#define DEFAULT_HOTPLUG_IO_SIZE         (256)
  79#define DEFAULT_HOTPLUG_MEM_SIZE        (2*1024*1024)
  80/* pci=hpmemsize=nnM,hpiosize=nn can override this */
  81unsigned long pci_hotplug_io_size  = DEFAULT_HOTPLUG_IO_SIZE;
  82unsigned long pci_hotplug_mem_size = DEFAULT_HOTPLUG_MEM_SIZE;
  83
  84enum pcie_bus_config_types pcie_bus_config = PCIE_BUS_DEFAULT;
  85
  86/*
  87 * The default CLS is used if arch didn't set CLS explicitly and not
  88 * all pci devices agree on the same value.  Arch can override either
  89 * the dfl or actual value as it sees fit.  Don't forget this is
  90 * measured in 32-bit words, not bytes.
  91 */
  92u8 pci_dfl_cache_line_size = L1_CACHE_BYTES >> 2;
  93u8 pci_cache_line_size;
  94
  95/*
  96 * If we set up a device for bus mastering, we need to check the latency
  97 * timer as certain BIOSes forget to set it properly.
  98 */
  99unsigned int pcibios_max_latency = 255;
 100
 101/* If set, the PCIe ARI capability will not be used. */
 102static bool pcie_ari_disabled;
 103
 104/**
 105 * pci_bus_max_busnr - returns maximum PCI bus number of given bus' children
 106 * @bus: pointer to PCI bus structure to search
 107 *
 108 * Given a PCI bus, returns the highest PCI bus number present in the set
 109 * including the given PCI bus and its list of child PCI buses.
 110 */
 111unsigned char pci_bus_max_busnr(struct pci_bus *bus)
 112{
 113        struct pci_bus *tmp;
 114        unsigned char max, n;
 115
 116        max = bus->busn_res.end;
 117        list_for_each_entry(tmp, &bus->children, node) {
 118                n = pci_bus_max_busnr(tmp);
 119                if (n > max)
 120                        max = n;
 121        }
 122        return max;
 123}
 124EXPORT_SYMBOL_GPL(pci_bus_max_busnr);
 125
 126#ifdef CONFIG_HAS_IOMEM
 127void __iomem *pci_ioremap_bar(struct pci_dev *pdev, int bar)
 128{
 129        struct resource *res = &pdev->resource[bar];
 130
 131        /*
 132         * Make sure the BAR is actually a memory resource, not an IO resource
 133         */
 134        if (res->flags & IORESOURCE_UNSET || !(res->flags & IORESOURCE_MEM)) {
 135                dev_warn(&pdev->dev, "can't ioremap BAR %d: %pR\n", bar, res);
 136                return NULL;
 137        }
 138        return ioremap_nocache(res->start, resource_size(res));
 139}
 140EXPORT_SYMBOL_GPL(pci_ioremap_bar);
 141
 142void __iomem *pci_ioremap_wc_bar(struct pci_dev *pdev, int bar)
 143{
 144        /*
 145         * Make sure the BAR is actually a memory resource, not an IO resource
 146         */
 147        if (!(pci_resource_flags(pdev, bar) & IORESOURCE_MEM)) {
 148                WARN_ON(1);
 149                return NULL;
 150        }
 151        return ioremap_wc(pci_resource_start(pdev, bar),
 152                          pci_resource_len(pdev, bar));
 153}
 154EXPORT_SYMBOL_GPL(pci_ioremap_wc_bar);
 155#endif
 156
 157
 158static int __pci_find_next_cap_ttl(struct pci_bus *bus, unsigned int devfn,
 159                                   u8 pos, int cap, int *ttl)
 160{
 161        u8 id;
 162        u16 ent;
 163
 164        pci_bus_read_config_byte(bus, devfn, pos, &pos);
 165
 166        while ((*ttl)--) {
 167                if (pos < 0x40)
 168                        break;
 169                pos &= ~3;
 170                pci_bus_read_config_word(bus, devfn, pos, &ent);
 171
 172                id = ent & 0xff;
 173                if (id == 0xff)
 174                        break;
 175                if (id == cap)
 176                        return pos;
 177                pos = (ent >> 8);
 178        }
 179        return 0;
 180}
 181
 182static int __pci_find_next_cap(struct pci_bus *bus, unsigned int devfn,
 183                               u8 pos, int cap)
 184{
 185        int ttl = PCI_FIND_CAP_TTL;
 186
 187        return __pci_find_next_cap_ttl(bus, devfn, pos, cap, &ttl);
 188}
 189
 190int pci_find_next_capability(struct pci_dev *dev, u8 pos, int cap)
 191{
 192        return __pci_find_next_cap(dev->bus, dev->devfn,
 193                                   pos + PCI_CAP_LIST_NEXT, cap);
 194}
 195EXPORT_SYMBOL_GPL(pci_find_next_capability);
 196
 197static int __pci_bus_find_cap_start(struct pci_bus *bus,
 198                                    unsigned int devfn, u8 hdr_type)
 199{
 200        u16 status;
 201
 202        pci_bus_read_config_word(bus, devfn, PCI_STATUS, &status);
 203        if (!(status & PCI_STATUS_CAP_LIST))
 204                return 0;
 205
 206        switch (hdr_type) {
 207        case PCI_HEADER_TYPE_NORMAL:
 208        case PCI_HEADER_TYPE_BRIDGE:
 209                return PCI_CAPABILITY_LIST;
 210        case PCI_HEADER_TYPE_CARDBUS:
 211                return PCI_CB_CAPABILITY_LIST;
 212        }
 213
 214        return 0;
 215}
 216
 217/**
 218 * pci_find_capability - query for devices' capabilities
 219 * @dev: PCI device to query
 220 * @cap: capability code
 221 *
 222 * Tell if a device supports a given PCI capability.
 223 * Returns the address of the requested capability structure within the
 224 * device's PCI configuration space or 0 in case the device does not
 225 * support it.  Possible values for @cap:
 226 *
 227 *  %PCI_CAP_ID_PM           Power Management
 228 *  %PCI_CAP_ID_AGP          Accelerated Graphics Port
 229 *  %PCI_CAP_ID_VPD          Vital Product Data
 230 *  %PCI_CAP_ID_SLOTID       Slot Identification
 231 *  %PCI_CAP_ID_MSI          Message Signalled Interrupts
 232 *  %PCI_CAP_ID_CHSWP        CompactPCI HotSwap
 233 *  %PCI_CAP_ID_PCIX         PCI-X
 234 *  %PCI_CAP_ID_EXP          PCI Express
 235 */
 236int pci_find_capability(struct pci_dev *dev, int cap)
 237{
 238        int pos;
 239
 240        pos = __pci_bus_find_cap_start(dev->bus, dev->devfn, dev->hdr_type);
 241        if (pos)
 242                pos = __pci_find_next_cap(dev->bus, dev->devfn, pos, cap);
 243
 244        return pos;
 245}
 246EXPORT_SYMBOL(pci_find_capability);
 247
 248/**
 249 * pci_bus_find_capability - query for devices' capabilities
 250 * @bus:   the PCI bus to query
 251 * @devfn: PCI device to query
 252 * @cap:   capability code
 253 *
 254 * Like pci_find_capability() but works for pci devices that do not have a
 255 * pci_dev structure set up yet.
 256 *
 257 * Returns the address of the requested capability structure within the
 258 * device's PCI configuration space or 0 in case the device does not
 259 * support it.
 260 */
 261int pci_bus_find_capability(struct pci_bus *bus, unsigned int devfn, int cap)
 262{
 263        int pos;
 264        u8 hdr_type;
 265
 266        pci_bus_read_config_byte(bus, devfn, PCI_HEADER_TYPE, &hdr_type);
 267
 268        pos = __pci_bus_find_cap_start(bus, devfn, hdr_type & 0x7f);
 269        if (pos)
 270                pos = __pci_find_next_cap(bus, devfn, pos, cap);
 271
 272        return pos;
 273}
 274EXPORT_SYMBOL(pci_bus_find_capability);
 275
 276/**
 277 * pci_find_next_ext_capability - Find an extended capability
 278 * @dev: PCI device to query
 279 * @start: address at which to start looking (0 to start at beginning of list)
 280 * @cap: capability code
 281 *
 282 * Returns the address of the next matching extended capability structure
 283 * within the device's PCI configuration space or 0 if the device does
 284 * not support it.  Some capabilities can occur several times, e.g., the
 285 * vendor-specific capability, and this provides a way to find them all.
 286 */
 287int pci_find_next_ext_capability(struct pci_dev *dev, int start, int cap)
 288{
 289        u32 header;
 290        int ttl;
 291        int pos = PCI_CFG_SPACE_SIZE;
 292
 293        /* minimum 8 bytes per capability */
 294        ttl = (PCI_CFG_SPACE_EXP_SIZE - PCI_CFG_SPACE_SIZE) / 8;
 295
 296        if (dev->cfg_size <= PCI_CFG_SPACE_SIZE)
 297                return 0;
 298
 299        if (start)
 300                pos = start;
 301
 302        if (pci_read_config_dword(dev, pos, &header) != PCIBIOS_SUCCESSFUL)
 303                return 0;
 304
 305        /*
 306         * If we have no capabilities, this is indicated by cap ID,
 307         * cap version and next pointer all being 0.
 308         */
 309        if (header == 0)
 310                return 0;
 311
 312        while (ttl-- > 0) {
 313                if (PCI_EXT_CAP_ID(header) == cap && pos != start)
 314                        return pos;
 315
 316                pos = PCI_EXT_CAP_NEXT(header);
 317                if (pos < PCI_CFG_SPACE_SIZE)
 318                        break;
 319
 320                if (pci_read_config_dword(dev, pos, &header) != PCIBIOS_SUCCESSFUL)
 321                        break;
 322        }
 323
 324        return 0;
 325}
 326EXPORT_SYMBOL_GPL(pci_find_next_ext_capability);
 327
 328/**
 329 * pci_find_ext_capability - Find an extended capability
 330 * @dev: PCI device to query
 331 * @cap: capability code
 332 *
 333 * Returns the address of the requested extended capability structure
 334 * within the device's PCI configuration space or 0 if the device does
 335 * not support it.  Possible values for @cap:
 336 *
 337 *  %PCI_EXT_CAP_ID_ERR         Advanced Error Reporting
 338 *  %PCI_EXT_CAP_ID_VC          Virtual Channel
 339 *  %PCI_EXT_CAP_ID_DSN         Device Serial Number
 340 *  %PCI_EXT_CAP_ID_PWR         Power Budgeting
 341 */
 342int pci_find_ext_capability(struct pci_dev *dev, int cap)
 343{
 344        return pci_find_next_ext_capability(dev, 0, cap);
 345}
 346EXPORT_SYMBOL_GPL(pci_find_ext_capability);
 347
 348static int __pci_find_next_ht_cap(struct pci_dev *dev, int pos, int ht_cap)
 349{
 350        int rc, ttl = PCI_FIND_CAP_TTL;
 351        u8 cap, mask;
 352
 353        if (ht_cap == HT_CAPTYPE_SLAVE || ht_cap == HT_CAPTYPE_HOST)
 354                mask = HT_3BIT_CAP_MASK;
 355        else
 356                mask = HT_5BIT_CAP_MASK;
 357
 358        pos = __pci_find_next_cap_ttl(dev->bus, dev->devfn, pos,
 359                                      PCI_CAP_ID_HT, &ttl);
 360        while (pos) {
 361                rc = pci_read_config_byte(dev, pos + 3, &cap);
 362                if (rc != PCIBIOS_SUCCESSFUL)
 363                        return 0;
 364
 365                if ((cap & mask) == ht_cap)
 366                        return pos;
 367
 368                pos = __pci_find_next_cap_ttl(dev->bus, dev->devfn,
 369                                              pos + PCI_CAP_LIST_NEXT,
 370                                              PCI_CAP_ID_HT, &ttl);
 371        }
 372
 373        return 0;
 374}
 375/**
 376 * pci_find_next_ht_capability - query a device's Hypertransport capabilities
 377 * @dev: PCI device to query
 378 * @pos: Position from which to continue searching
 379 * @ht_cap: Hypertransport capability code
 380 *
 381 * To be used in conjunction with pci_find_ht_capability() to search for
 382 * all capabilities matching @ht_cap. @pos should always be a value returned
 383 * from pci_find_ht_capability().
 384 *
 385 * NB. To be 100% safe against broken PCI devices, the caller should take
 386 * steps to avoid an infinite loop.
 387 */
 388int pci_find_next_ht_capability(struct pci_dev *dev, int pos, int ht_cap)
 389{
 390        return __pci_find_next_ht_cap(dev, pos + PCI_CAP_LIST_NEXT, ht_cap);
 391}
 392EXPORT_SYMBOL_GPL(pci_find_next_ht_capability);
 393
 394/**
 395 * pci_find_ht_capability - query a device's Hypertransport capabilities
 396 * @dev: PCI device to query
 397 * @ht_cap: Hypertransport capability code
 398 *
 399 * Tell if a device supports a given Hypertransport capability.
 400 * Returns an address within the device's PCI configuration space
 401 * or 0 in case the device does not support the request capability.
 402 * The address points to the PCI capability, of type PCI_CAP_ID_HT,
 403 * which has a Hypertransport capability matching @ht_cap.
 404 */
 405int pci_find_ht_capability(struct pci_dev *dev, int ht_cap)
 406{
 407        int pos;
 408
 409        pos = __pci_bus_find_cap_start(dev->bus, dev->devfn, dev->hdr_type);
 410        if (pos)
 411                pos = __pci_find_next_ht_cap(dev, pos, ht_cap);
 412
 413        return pos;
 414}
 415EXPORT_SYMBOL_GPL(pci_find_ht_capability);
 416
 417/**
 418 * pci_find_parent_resource - return resource region of parent bus of given region
 419 * @dev: PCI device structure contains resources to be searched
 420 * @res: child resource record for which parent is sought
 421 *
 422 *  For given resource region of given device, return the resource
 423 *  region of parent bus the given region is contained in.
 424 */
 425struct resource *pci_find_parent_resource(const struct pci_dev *dev,
 426                                          struct resource *res)
 427{
 428        const struct pci_bus *bus = dev->bus;
 429        struct resource *r;
 430        int i;
 431
 432        pci_bus_for_each_resource(bus, r, i) {
 433                if (!r)
 434                        continue;
 435                if (res->start && resource_contains(r, res)) {
 436
 437                        /*
 438                         * If the window is prefetchable but the BAR is
 439                         * not, the allocator made a mistake.
 440                         */
 441                        if (r->flags & IORESOURCE_PREFETCH &&
 442                            !(res->flags & IORESOURCE_PREFETCH))
 443                                return NULL;
 444
 445                        /*
 446                         * If we're below a transparent bridge, there may
 447                         * be both a positively-decoded aperture and a
 448                         * subtractively-decoded region that contain the BAR.
 449                         * We want the positively-decoded one, so this depends
 450                         * on pci_bus_for_each_resource() giving us those
 451                         * first.
 452                         */
 453                        return r;
 454                }
 455        }
 456        return NULL;
 457}
 458EXPORT_SYMBOL(pci_find_parent_resource);
 459
 460/**
 461 * pci_find_pcie_root_port - return PCIe Root Port
 462 * @dev: PCI device to query
 463 *
 464 * Traverse up the parent chain and return the PCIe Root Port PCI Device
 465 * for a given PCI Device.
 466 */
 467struct pci_dev *pci_find_pcie_root_port(struct pci_dev *dev)
 468{
 469        struct pci_dev *bridge, *highest_pcie_bridge = NULL;
 470
 471        bridge = pci_upstream_bridge(dev);
 472        while (bridge && pci_is_pcie(bridge)) {
 473                highest_pcie_bridge = bridge;
 474                bridge = pci_upstream_bridge(bridge);
 475        }
 476
 477        if (pci_pcie_type(highest_pcie_bridge) != PCI_EXP_TYPE_ROOT_PORT)
 478                return NULL;
 479
 480        return highest_pcie_bridge;
 481}
 482EXPORT_SYMBOL(pci_find_pcie_root_port);
 483
 484/**
 485 * pci_wait_for_pending - wait for @mask bit(s) to clear in status word @pos
 486 * @dev: the PCI device to operate on
 487 * @pos: config space offset of status word
 488 * @mask: mask of bit(s) to care about in status word
 489 *
 490 * Return 1 when mask bit(s) in status word clear, 0 otherwise.
 491 */
 492int pci_wait_for_pending(struct pci_dev *dev, int pos, u16 mask)
 493{
 494        int i;
 495
 496        /* Wait for Transaction Pending bit clean */
 497        for (i = 0; i < 4; i++) {
 498                u16 status;
 499                if (i)
 500                        msleep((1 << (i - 1)) * 100);
 501
 502                pci_read_config_word(dev, pos, &status);
 503                if (!(status & mask))
 504                        return 1;
 505        }
 506
 507        return 0;
 508}
 509
 510/**
 511 * pci_restore_bars - restore a device's BAR values (e.g. after wake-up)
 512 * @dev: PCI device to have its BARs restored
 513 *
 514 * Restore the BAR values for a given device, so as to make it
 515 * accessible by its driver.
 516 */
 517static void pci_restore_bars(struct pci_dev *dev)
 518{
 519        int i;
 520
 521        /* Per SR-IOV spec 3.4.1.11, VF BARs are RO zero */
 522        if (dev->is_virtfn)
 523                return;
 524
 525        for (i = 0; i < PCI_BRIDGE_RESOURCES; i++)
 526                pci_update_resource(dev, i);
 527}
 528
 529static const struct pci_platform_pm_ops *pci_platform_pm;
 530
 531int pci_set_platform_pm(const struct pci_platform_pm_ops *ops)
 532{
 533        if (!ops->is_manageable || !ops->set_state || !ops->choose_state
 534            || !ops->sleep_wake)
 535                return -EINVAL;
 536        pci_platform_pm = ops;
 537        return 0;
 538}
 539
 540static inline bool platform_pci_power_manageable(struct pci_dev *dev)
 541{
 542        return pci_platform_pm ? pci_platform_pm->is_manageable(dev) : false;
 543}
 544
 545static inline int platform_pci_set_power_state(struct pci_dev *dev,
 546                                               pci_power_t t)
 547{
 548        return pci_platform_pm ? pci_platform_pm->set_state(dev, t) : -ENOSYS;
 549}
 550
 551static inline pci_power_t platform_pci_choose_state(struct pci_dev *dev)
 552{
 553        return pci_platform_pm ?
 554                        pci_platform_pm->choose_state(dev) : PCI_POWER_ERROR;
 555}
 556
 557static inline int platform_pci_sleep_wake(struct pci_dev *dev, bool enable)
 558{
 559        return pci_platform_pm ?
 560                        pci_platform_pm->sleep_wake(dev, enable) : -ENODEV;
 561}
 562
 563static inline int platform_pci_run_wake(struct pci_dev *dev, bool enable)
 564{
 565        return pci_platform_pm ?
 566                        pci_platform_pm->run_wake(dev, enable) : -ENODEV;
 567}
 568
 569static inline bool platform_pci_need_resume(struct pci_dev *dev)
 570{
 571        return pci_platform_pm ? pci_platform_pm->need_resume(dev) : false;
 572}
 573
 574/**
 575 * pci_raw_set_power_state - Use PCI PM registers to set the power state of
 576 *                           given PCI device
 577 * @dev: PCI device to handle.
 578 * @state: PCI power state (D0, D1, D2, D3hot) to put the device into.
 579 *
 580 * RETURN VALUE:
 581 * -EINVAL if the requested state is invalid.
 582 * -EIO if device does not support PCI PM or its PM capabilities register has a
 583 * wrong version, or device doesn't support the requested state.
 584 * 0 if device already is in the requested state.
 585 * 0 if device's power state has been successfully changed.
 586 */
 587static int pci_raw_set_power_state(struct pci_dev *dev, pci_power_t state)
 588{
 589        u16 pmcsr;
 590        bool need_restore = false;
 591
 592        /* Check if we're already there */
 593        if (dev->current_state == state)
 594                return 0;
 595
 596        if (!dev->pm_cap)
 597                return -EIO;
 598
 599        if (state < PCI_D0 || state > PCI_D3hot)
 600                return -EINVAL;
 601
 602        /* Validate current state:
 603         * Can enter D0 from any state, but if we can only go deeper
 604         * to sleep if we're already in a low power state
 605         */
 606        if (state != PCI_D0 && dev->current_state <= PCI_D3cold
 607            && dev->current_state > state) {
 608                dev_err(&dev->dev, "invalid power transition (from state %d to %d)\n",
 609                        dev->current_state, state);
 610                return -EINVAL;
 611        }
 612
 613        /* check if this device supports the desired state */
 614        if ((state == PCI_D1 && !dev->d1_support)
 615           || (state == PCI_D2 && !dev->d2_support))
 616                return -EIO;
 617
 618        pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
 619
 620        /* If we're (effectively) in D3, force entire word to 0.
 621         * This doesn't affect PME_Status, disables PME_En, and
 622         * sets PowerState to 0.
 623         */
 624        switch (dev->current_state) {
 625        case PCI_D0:
 626        case PCI_D1:
 627        case PCI_D2:
 628                pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
 629                pmcsr |= state;
 630                break;
 631        case PCI_D3hot:
 632        case PCI_D3cold:
 633        case PCI_UNKNOWN: /* Boot-up */
 634                if ((pmcsr & PCI_PM_CTRL_STATE_MASK) == PCI_D3hot
 635                 && !(pmcsr & PCI_PM_CTRL_NO_SOFT_RESET))
 636                        need_restore = true;
 637                /* Fall-through: force to D0 */
 638        default:
 639                pmcsr = 0;
 640                break;
 641        }
 642
 643        /* enter specified state */
 644        pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, pmcsr);
 645
 646        /* Mandatory power management transition delays */
 647        /* see PCI PM 1.1 5.6.1 table 18 */
 648        if (state == PCI_D3hot || dev->current_state == PCI_D3hot)
 649                pci_dev_d3_sleep(dev);
 650        else if (state == PCI_D2 || dev->current_state == PCI_D2)
 651                udelay(PCI_PM_D2_DELAY);
 652
 653        pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
 654        dev->current_state = (pmcsr & PCI_PM_CTRL_STATE_MASK);
 655        if (dev->current_state != state && printk_ratelimit())
 656                dev_info(&dev->dev, "Refused to change power state, currently in D%d\n",
 657                         dev->current_state);
 658
 659        /*
 660         * According to section 5.4.1 of the "PCI BUS POWER MANAGEMENT
 661         * INTERFACE SPECIFICATION, REV. 1.2", a device transitioning
 662         * from D3hot to D0 _may_ perform an internal reset, thereby
 663         * going to "D0 Uninitialized" rather than "D0 Initialized".
 664         * For example, at least some versions of the 3c905B and the
 665         * 3c556B exhibit this behaviour.
 666         *
 667         * At least some laptop BIOSen (e.g. the Thinkpad T21) leave
 668         * devices in a D3hot state at boot.  Consequently, we need to
 669         * restore at least the BARs so that the device will be
 670         * accessible to its driver.
 671         */
 672        if (need_restore)
 673                pci_restore_bars(dev);
 674
 675        if (dev->bus->self)
 676                pcie_aspm_pm_state_change(dev->bus->self);
 677
 678        return 0;
 679}
 680
 681/**
 682 * pci_update_current_state - Read PCI power state of given device from its
 683 *                            PCI PM registers and cache it
 684 * @dev: PCI device to handle.
 685 * @state: State to cache in case the device doesn't have the PM capability
 686 */
 687void pci_update_current_state(struct pci_dev *dev, pci_power_t state)
 688{
 689        if (dev->pm_cap) {
 690                u16 pmcsr;
 691
 692                /*
 693                 * Configuration space is not accessible for device in
 694                 * D3cold, so just keep or set D3cold for safety
 695                 */
 696                if (dev->current_state == PCI_D3cold)
 697                        return;
 698                if (state == PCI_D3cold) {
 699                        dev->current_state = PCI_D3cold;
 700                        return;
 701                }
 702                pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
 703                dev->current_state = (pmcsr & PCI_PM_CTRL_STATE_MASK);
 704        } else {
 705                dev->current_state = state;
 706        }
 707}
 708
 709/**
 710 * pci_power_up - Put the given device into D0 forcibly
 711 * @dev: PCI device to power up
 712 */
 713void pci_power_up(struct pci_dev *dev)
 714{
 715        if (platform_pci_power_manageable(dev))
 716                platform_pci_set_power_state(dev, PCI_D0);
 717
 718        pci_raw_set_power_state(dev, PCI_D0);
 719        pci_update_current_state(dev, PCI_D0);
 720}
 721
 722/**
 723 * pci_platform_power_transition - Use platform to change device power state
 724 * @dev: PCI device to handle.
 725 * @state: State to put the device into.
 726 */
 727static int pci_platform_power_transition(struct pci_dev *dev, pci_power_t state)
 728{
 729        int error;
 730
 731        if (platform_pci_power_manageable(dev)) {
 732                error = platform_pci_set_power_state(dev, state);
 733                if (!error)
 734                        pci_update_current_state(dev, state);
 735        } else
 736                error = -ENODEV;
 737
 738        if (error && !dev->pm_cap) /* Fall back to PCI_D0 */
 739                dev->current_state = PCI_D0;
 740
 741        return error;
 742}
 743
 744/**
 745 * pci_wakeup - Wake up a PCI device
 746 * @pci_dev: Device to handle.
 747 * @ign: ignored parameter
 748 */
 749static int pci_wakeup(struct pci_dev *pci_dev, void *ign)
 750{
 751        pci_wakeup_event(pci_dev);
 752        pm_request_resume(&pci_dev->dev);
 753        return 0;
 754}
 755
 756/**
 757 * pci_wakeup_bus - Walk given bus and wake up devices on it
 758 * @bus: Top bus of the subtree to walk.
 759 */
 760static void pci_wakeup_bus(struct pci_bus *bus)
 761{
 762        if (bus)
 763                pci_walk_bus(bus, pci_wakeup, NULL);
 764}
 765
 766/**
 767 * __pci_start_power_transition - Start power transition of a PCI device
 768 * @dev: PCI device to handle.
 769 * @state: State to put the device into.
 770 */
 771static void __pci_start_power_transition(struct pci_dev *dev, pci_power_t state)
 772{
 773        if (state == PCI_D0) {
 774                pci_platform_power_transition(dev, PCI_D0);
 775                /*
 776                 * Mandatory power management transition delays, see
 777                 * PCI Express Base Specification Revision 2.0 Section
 778                 * 6.6.1: Conventional Reset.  Do not delay for
 779                 * devices powered on/off by corresponding bridge,
 780                 * because have already delayed for the bridge.
 781                 */
 782                if (dev->runtime_d3cold) {
 783                        msleep(dev->d3cold_delay);
 784                        /*
 785                         * When powering on a bridge from D3cold, the
 786                         * whole hierarchy may be powered on into
 787                         * D0uninitialized state, resume them to give
 788                         * them a chance to suspend again
 789                         */
 790                        pci_wakeup_bus(dev->subordinate);
 791                }
 792        }
 793}
 794
 795/**
 796 * __pci_dev_set_current_state - Set current state of a PCI device
 797 * @dev: Device to handle
 798 * @data: pointer to state to be set
 799 */
 800static int __pci_dev_set_current_state(struct pci_dev *dev, void *data)
 801{
 802        pci_power_t state = *(pci_power_t *)data;
 803
 804        dev->current_state = state;
 805        return 0;
 806}
 807
 808/**
 809 * __pci_bus_set_current_state - Walk given bus and set current state of devices
 810 * @bus: Top bus of the subtree to walk.
 811 * @state: state to be set
 812 */
 813static void __pci_bus_set_current_state(struct pci_bus *bus, pci_power_t state)
 814{
 815        if (bus)
 816                pci_walk_bus(bus, __pci_dev_set_current_state, &state);
 817}
 818
 819/**
 820 * __pci_complete_power_transition - Complete power transition of a PCI device
 821 * @dev: PCI device to handle.
 822 * @state: State to put the device into.
 823 *
 824 * This function should not be called directly by device drivers.
 825 */
 826int __pci_complete_power_transition(struct pci_dev *dev, pci_power_t state)
 827{
 828        int ret;
 829
 830        if (state <= PCI_D0)
 831                return -EINVAL;
 832        ret = pci_platform_power_transition(dev, state);
 833        /* Power off the bridge may power off the whole hierarchy */
 834        if (!ret && state == PCI_D3cold)
 835                __pci_bus_set_current_state(dev->subordinate, PCI_D3cold);
 836        return ret;
 837}
 838EXPORT_SYMBOL_GPL(__pci_complete_power_transition);
 839
 840/**
 841 * pci_set_power_state - Set the power state of a PCI device
 842 * @dev: PCI device to handle.
 843 * @state: PCI power state (D0, D1, D2, D3hot) to put the device into.
 844 *
 845 * Transition a device to a new power state, using the platform firmware and/or
 846 * the device's PCI PM registers.
 847 *
 848 * RETURN VALUE:
 849 * -EINVAL if the requested state is invalid.
 850 * -EIO if device does not support PCI PM or its PM capabilities register has a
 851 * wrong version, or device doesn't support the requested state.
 852 * 0 if device already is in the requested state.
 853 * 0 if device's power state has been successfully changed.
 854 */
 855int pci_set_power_state(struct pci_dev *dev, pci_power_t state)
 856{
 857        int error;
 858
 859        /* bound the state we're entering */
 860        if (state > PCI_D3cold)
 861                state = PCI_D3cold;
 862        else if (state < PCI_D0)
 863                state = PCI_D0;
 864        else if ((state == PCI_D1 || state == PCI_D2) && pci_no_d1d2(dev))
 865                /*
 866                 * If the device or the parent bridge do not support PCI PM,
 867                 * ignore the request if we're doing anything other than putting
 868                 * it into D0 (which would only happen on boot).
 869                 */
 870                return 0;
 871
 872        /* Check if we're already there */
 873        if (dev->current_state == state)
 874                return 0;
 875
 876        __pci_start_power_transition(dev, state);
 877
 878        /* This device is quirked not to be put into D3, so
 879           don't put it in D3 */
 880        if (state >= PCI_D3hot && (dev->dev_flags & PCI_DEV_FLAGS_NO_D3))
 881                return 0;
 882
 883        /*
 884         * To put device in D3cold, we put device into D3hot in native
 885         * way, then put device into D3cold with platform ops
 886         */
 887        error = pci_raw_set_power_state(dev, state > PCI_D3hot ?
 888                                        PCI_D3hot : state);
 889
 890        if (!__pci_complete_power_transition(dev, state))
 891                error = 0;
 892
 893        return error;
 894}
 895EXPORT_SYMBOL(pci_set_power_state);
 896
 897/**
 898 * pci_choose_state - Choose the power state of a PCI device
 899 * @dev: PCI device to be suspended
 900 * @state: target sleep state for the whole system. This is the value
 901 *      that is passed to suspend() function.
 902 *
 903 * Returns PCI power state suitable for given device and given system
 904 * message.
 905 */
 906
 907pci_power_t pci_choose_state(struct pci_dev *dev, pm_message_t state)
 908{
 909        pci_power_t ret;
 910
 911        if (!dev->pm_cap)
 912                return PCI_D0;
 913
 914        ret = platform_pci_choose_state(dev);
 915        if (ret != PCI_POWER_ERROR)
 916                return ret;
 917
 918        switch (state.event) {
 919        case PM_EVENT_ON:
 920                return PCI_D0;
 921        case PM_EVENT_FREEZE:
 922        case PM_EVENT_PRETHAW:
 923                /* REVISIT both freeze and pre-thaw "should" use D0 */
 924        case PM_EVENT_SUSPEND:
 925        case PM_EVENT_HIBERNATE:
 926                return PCI_D3hot;
 927        default:
 928                dev_info(&dev->dev, "unrecognized suspend event %d\n",
 929                         state.event);
 930                BUG();
 931        }
 932        return PCI_D0;
 933}
 934EXPORT_SYMBOL(pci_choose_state);
 935
 936#define PCI_EXP_SAVE_REGS       7
 937
 938static struct pci_cap_saved_state *_pci_find_saved_cap(struct pci_dev *pci_dev,
 939                                                       u16 cap, bool extended)
 940{
 941        struct pci_cap_saved_state *tmp;
 942
 943        hlist_for_each_entry(tmp, &pci_dev->saved_cap_space, next) {
 944                if (tmp->cap.cap_extended == extended && tmp->cap.cap_nr == cap)
 945                        return tmp;
 946        }
 947        return NULL;
 948}
 949
 950struct pci_cap_saved_state *pci_find_saved_cap(struct pci_dev *dev, char cap)
 951{
 952        return _pci_find_saved_cap(dev, cap, false);
 953}
 954
 955struct pci_cap_saved_state *pci_find_saved_ext_cap(struct pci_dev *dev, u16 cap)
 956{
 957        return _pci_find_saved_cap(dev, cap, true);
 958}
 959
 960static int pci_save_pcie_state(struct pci_dev *dev)
 961{
 962        int i = 0;
 963        struct pci_cap_saved_state *save_state;
 964        u16 *cap;
 965
 966        if (!pci_is_pcie(dev))
 967                return 0;
 968
 969        save_state = pci_find_saved_cap(dev, PCI_CAP_ID_EXP);
 970        if (!save_state) {
 971                dev_err(&dev->dev, "buffer not found in %s\n", __func__);
 972                return -ENOMEM;
 973        }
 974
 975        cap = (u16 *)&save_state->cap.data[0];
 976        pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &cap[i++]);
 977        pcie_capability_read_word(dev, PCI_EXP_LNKCTL, &cap[i++]);
 978        pcie_capability_read_word(dev, PCI_EXP_SLTCTL, &cap[i++]);
 979        pcie_capability_read_word(dev, PCI_EXP_RTCTL,  &cap[i++]);
 980        pcie_capability_read_word(dev, PCI_EXP_DEVCTL2, &cap[i++]);
 981        pcie_capability_read_word(dev, PCI_EXP_LNKCTL2, &cap[i++]);
 982        pcie_capability_read_word(dev, PCI_EXP_SLTCTL2, &cap[i++]);
 983
 984        return 0;
 985}
 986
 987static void pci_restore_pcie_state(struct pci_dev *dev)
 988{
 989        int i = 0;
 990        struct pci_cap_saved_state *save_state;
 991        u16 *cap;
 992
 993        save_state = pci_find_saved_cap(dev, PCI_CAP_ID_EXP);
 994        if (!save_state)
 995                return;
 996
 997        cap = (u16 *)&save_state->cap.data[0];
 998        pcie_capability_write_word(dev, PCI_EXP_DEVCTL, cap[i++]);
 999        pcie_capability_write_word(dev, PCI_EXP_LNKCTL, cap[i++]);
1000        pcie_capability_write_word(dev, PCI_EXP_SLTCTL, cap[i++]);
1001        pcie_capability_write_word(dev, PCI_EXP_RTCTL, cap[i++]);
1002        pcie_capability_write_word(dev, PCI_EXP_DEVCTL2, cap[i++]);
1003        pcie_capability_write_word(dev, PCI_EXP_LNKCTL2, cap[i++]);
1004        pcie_capability_write_word(dev, PCI_EXP_SLTCTL2, cap[i++]);
1005}
1006
1007
1008static int pci_save_pcix_state(struct pci_dev *dev)
1009{
1010        int pos;
1011        struct pci_cap_saved_state *save_state;
1012
1013        pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1014        if (!pos)
1015                return 0;
1016
1017        save_state = pci_find_saved_cap(dev, PCI_CAP_ID_PCIX);
1018        if (!save_state) {
1019                dev_err(&dev->dev, "buffer not found in %s\n", __func__);
1020                return -ENOMEM;
1021        }
1022
1023        pci_read_config_word(dev, pos + PCI_X_CMD,
1024                             (u16 *)save_state->cap.data);
1025
1026        return 0;
1027}
1028
1029static void pci_restore_pcix_state(struct pci_dev *dev)
1030{
1031        int i = 0, pos;
1032        struct pci_cap_saved_state *save_state;
1033        u16 *cap;
1034
1035        save_state = pci_find_saved_cap(dev, PCI_CAP_ID_PCIX);
1036        pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1037        if (!save_state || !pos)
1038                return;
1039        cap = (u16 *)&save_state->cap.data[0];
1040
1041        pci_write_config_word(dev, pos + PCI_X_CMD, cap[i++]);
1042}
1043
1044
1045/**
1046 * pci_save_state - save the PCI configuration space of a device before suspending
1047 * @dev: - PCI device that we're dealing with
1048 */
1049int pci_save_state(struct pci_dev *dev)
1050{
1051        int i;
1052        /* XXX: 100% dword access ok here? */
1053        for (i = 0; i < 16; i++)
1054                pci_read_config_dword(dev, i * 4, &dev->saved_config_space[i]);
1055        dev->state_saved = true;
1056
1057        i = pci_save_pcie_state(dev);
1058        if (i != 0)
1059                return i;
1060
1061        i = pci_save_pcix_state(dev);
1062        if (i != 0)
1063                return i;
1064
1065        return pci_save_vc_state(dev);
1066}
1067EXPORT_SYMBOL(pci_save_state);
1068
1069static void pci_restore_config_dword(struct pci_dev *pdev, int offset,
1070                                     u32 saved_val, int retry)
1071{
1072        u32 val;
1073
1074        pci_read_config_dword(pdev, offset, &val);
1075        if (val == saved_val)
1076                return;
1077
1078        for (;;) {
1079                dev_dbg(&pdev->dev, "restoring config space at offset %#x (was %#x, writing %#x)\n",
1080                        offset, val, saved_val);
1081                pci_write_config_dword(pdev, offset, saved_val);
1082                if (retry-- <= 0)
1083                        return;
1084
1085                pci_read_config_dword(pdev, offset, &val);
1086                if (val == saved_val)
1087                        return;
1088
1089                mdelay(1);
1090        }
1091}
1092
1093static void pci_restore_config_space_range(struct pci_dev *pdev,
1094                                           int start, int end, int retry)
1095{
1096        int index;
1097
1098        for (index = end; index >= start; index--)
1099                pci_restore_config_dword(pdev, 4 * index,
1100                                         pdev->saved_config_space[index],
1101                                         retry);
1102}
1103
1104static void pci_restore_config_space(struct pci_dev *pdev)
1105{
1106        if (pdev->hdr_type == PCI_HEADER_TYPE_NORMAL) {
1107                pci_restore_config_space_range(pdev, 10, 15, 0);
1108                /* Restore BARs before the command register. */
1109                pci_restore_config_space_range(pdev, 4, 9, 10);
1110                pci_restore_config_space_range(pdev, 0, 3, 0);
1111        } else {
1112                pci_restore_config_space_range(pdev, 0, 15, 0);
1113        }
1114}
1115
1116/**
1117 * pci_restore_state - Restore the saved state of a PCI device
1118 * @dev: - PCI device that we're dealing with
1119 */
1120void pci_restore_state(struct pci_dev *dev)
1121{
1122        if (!dev->state_saved)
1123                return;
1124
1125        /* PCI Express register must be restored first */
1126        pci_restore_pcie_state(dev);
1127        pci_restore_ats_state(dev);
1128        pci_restore_vc_state(dev);
1129
1130        pci_cleanup_aer_error_status_regs(dev);
1131
1132        pci_restore_config_space(dev);
1133
1134        pci_restore_pcix_state(dev);
1135        pci_restore_msi_state(dev);
1136
1137        /* Restore ACS and IOV configuration state */
1138        pci_enable_acs(dev);
1139        pci_restore_iov_state(dev);
1140
1141        dev->state_saved = false;
1142}
1143EXPORT_SYMBOL(pci_restore_state);
1144
1145struct pci_saved_state {
1146        u32 config_space[16];
1147        struct pci_cap_saved_data cap[0];
1148};
1149
1150/**
1151 * pci_store_saved_state - Allocate and return an opaque struct containing
1152 *                         the device saved state.
1153 * @dev: PCI device that we're dealing with
1154 *
1155 * Return NULL if no state or error.
1156 */
1157struct pci_saved_state *pci_store_saved_state(struct pci_dev *dev)
1158{
1159        struct pci_saved_state *state;
1160        struct pci_cap_saved_state *tmp;
1161        struct pci_cap_saved_data *cap;
1162        size_t size;
1163
1164        if (!dev->state_saved)
1165                return NULL;
1166
1167        size = sizeof(*state) + sizeof(struct pci_cap_saved_data);
1168
1169        hlist_for_each_entry(tmp, &dev->saved_cap_space, next)
1170                size += sizeof(struct pci_cap_saved_data) + tmp->cap.size;
1171
1172        state = kzalloc(size, GFP_KERNEL);
1173        if (!state)
1174                return NULL;
1175
1176        memcpy(state->config_space, dev->saved_config_space,
1177               sizeof(state->config_space));
1178
1179        cap = state->cap;
1180        hlist_for_each_entry(tmp, &dev->saved_cap_space, next) {
1181                size_t len = sizeof(struct pci_cap_saved_data) + tmp->cap.size;
1182                memcpy(cap, &tmp->cap, len);
1183                cap = (struct pci_cap_saved_data *)((u8 *)cap + len);
1184        }
1185        /* Empty cap_save terminates list */
1186
1187        return state;
1188}
1189EXPORT_SYMBOL_GPL(pci_store_saved_state);
1190
1191/**
1192 * pci_load_saved_state - Reload the provided save state into struct pci_dev.
1193 * @dev: PCI device that we're dealing with
1194 * @state: Saved state returned from pci_store_saved_state()
1195 */
1196int pci_load_saved_state(struct pci_dev *dev,
1197                         struct pci_saved_state *state)
1198{
1199        struct pci_cap_saved_data *cap;
1200
1201        dev->state_saved = false;
1202
1203        if (!state)
1204                return 0;
1205
1206        memcpy(dev->saved_config_space, state->config_space,
1207               sizeof(state->config_space));
1208
1209        cap = state->cap;
1210        while (cap->size) {
1211                struct pci_cap_saved_state *tmp;
1212
1213                tmp = _pci_find_saved_cap(dev, cap->cap_nr, cap->cap_extended);
1214                if (!tmp || tmp->cap.size != cap->size)
1215                        return -EINVAL;
1216
1217                memcpy(tmp->cap.data, cap->data, tmp->cap.size);
1218                cap = (struct pci_cap_saved_data *)((u8 *)cap +
1219                       sizeof(struct pci_cap_saved_data) + cap->size);
1220        }
1221
1222        dev->state_saved = true;
1223        return 0;
1224}
1225EXPORT_SYMBOL_GPL(pci_load_saved_state);
1226
1227/**
1228 * pci_load_and_free_saved_state - Reload the save state pointed to by state,
1229 *                                 and free the memory allocated for it.
1230 * @dev: PCI device that we're dealing with
1231 * @state: Pointer to saved state returned from pci_store_saved_state()
1232 */
1233int pci_load_and_free_saved_state(struct pci_dev *dev,
1234                                  struct pci_saved_state **state)
1235{
1236        int ret = pci_load_saved_state(dev, *state);
1237        kfree(*state);
1238        *state = NULL;
1239        return ret;
1240}
1241EXPORT_SYMBOL_GPL(pci_load_and_free_saved_state);
1242
1243int __weak pcibios_enable_device(struct pci_dev *dev, int bars)
1244{
1245        return pci_enable_resources(dev, bars);
1246}
1247
1248static int do_pci_enable_device(struct pci_dev *dev, int bars)
1249{
1250        int err;
1251        struct pci_dev *bridge;
1252        u16 cmd;
1253        u8 pin;
1254
1255        err = pci_set_power_state(dev, PCI_D0);
1256        if (err < 0 && err != -EIO)
1257                return err;
1258
1259        bridge = pci_upstream_bridge(dev);
1260        if (bridge)
1261                pcie_aspm_powersave_config_link(bridge);
1262
1263        err = pcibios_enable_device(dev, bars);
1264        if (err < 0)
1265                return err;
1266        pci_fixup_device(pci_fixup_enable, dev);
1267
1268        if (dev->msi_enabled || dev->msix_enabled)
1269                return 0;
1270
1271        pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &pin);
1272        if (pin) {
1273                pci_read_config_word(dev, PCI_COMMAND, &cmd);
1274                if (cmd & PCI_COMMAND_INTX_DISABLE)
1275                        pci_write_config_word(dev, PCI_COMMAND,
1276                                              cmd & ~PCI_COMMAND_INTX_DISABLE);
1277        }
1278
1279        return 0;
1280}
1281
1282/**
1283 * pci_reenable_device - Resume abandoned device
1284 * @dev: PCI device to be resumed
1285 *
1286 *  Note this function is a backend of pci_default_resume and is not supposed
1287 *  to be called by normal code, write proper resume handler and use it instead.
1288 */
1289int pci_reenable_device(struct pci_dev *dev)
1290{
1291        if (pci_is_enabled(dev))
1292                return do_pci_enable_device(dev, (1 << PCI_NUM_RESOURCES) - 1);
1293        return 0;
1294}
1295EXPORT_SYMBOL(pci_reenable_device);
1296
1297static void pci_enable_bridge(struct pci_dev *dev)
1298{
1299        struct pci_dev *bridge;
1300        int retval;
1301
1302        bridge = pci_upstream_bridge(dev);
1303        if (bridge)
1304                pci_enable_bridge(bridge);
1305
1306        if (pci_is_enabled(dev)) {
1307                if (!dev->is_busmaster)
1308                        pci_set_master(dev);
1309                return;
1310        }
1311
1312        retval = pci_enable_device(dev);
1313        if (retval)
1314                dev_err(&dev->dev, "Error enabling bridge (%d), continuing\n",
1315                        retval);
1316        pci_set_master(dev);
1317}
1318
1319static int pci_enable_device_flags(struct pci_dev *dev, unsigned long flags)
1320{
1321        struct pci_dev *bridge;
1322        int err;
1323        int i, bars = 0;
1324
1325        /*
1326         * Power state could be unknown at this point, either due to a fresh
1327         * boot or a device removal call.  So get the current power state
1328         * so that things like MSI message writing will behave as expected
1329         * (e.g. if the device really is in D0 at enable time).
1330         */
1331        if (dev->pm_cap) {
1332                u16 pmcsr;
1333                pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
1334                dev->current_state = (pmcsr & PCI_PM_CTRL_STATE_MASK);
1335        }
1336
1337        if (atomic_inc_return(&dev->enable_cnt) > 1)
1338                return 0;               /* already enabled */
1339
1340        bridge = pci_upstream_bridge(dev);
1341        if (bridge)
1342                pci_enable_bridge(bridge);
1343
1344        /* only skip sriov related */
1345        for (i = 0; i <= PCI_ROM_RESOURCE; i++)
1346                if (dev->resource[i].flags & flags)
1347                        bars |= (1 << i);
1348        for (i = PCI_BRIDGE_RESOURCES; i < DEVICE_COUNT_RESOURCE; i++)
1349                if (dev->resource[i].flags & flags)
1350                        bars |= (1 << i);
1351
1352        err = do_pci_enable_device(dev, bars);
1353        if (err < 0)
1354                atomic_dec(&dev->enable_cnt);
1355        return err;
1356}
1357
1358/**
1359 * pci_enable_device_io - Initialize a device for use with IO space
1360 * @dev: PCI device to be initialized
1361 *
1362 *  Initialize device before it's used by a driver. Ask low-level code
1363 *  to enable I/O resources. Wake up the device if it was suspended.
1364 *  Beware, this function can fail.
1365 */
1366int pci_enable_device_io(struct pci_dev *dev)
1367{
1368        return pci_enable_device_flags(dev, IORESOURCE_IO);
1369}
1370EXPORT_SYMBOL(pci_enable_device_io);
1371
1372/**
1373 * pci_enable_device_mem - Initialize a device for use with Memory space
1374 * @dev: PCI device to be initialized
1375 *
1376 *  Initialize device before it's used by a driver. Ask low-level code
1377 *  to enable Memory resources. Wake up the device if it was suspended.
1378 *  Beware, this function can fail.
1379 */
1380int pci_enable_device_mem(struct pci_dev *dev)
1381{
1382        return pci_enable_device_flags(dev, IORESOURCE_MEM);
1383}
1384EXPORT_SYMBOL(pci_enable_device_mem);
1385
1386/**
1387 * pci_enable_device - Initialize device before it's used by a driver.
1388 * @dev: PCI device to be initialized
1389 *
1390 *  Initialize device before it's used by a driver. Ask low-level code
1391 *  to enable I/O and memory. Wake up the device if it was suspended.
1392 *  Beware, this function can fail.
1393 *
1394 *  Note we don't actually enable the device many times if we call
1395 *  this function repeatedly (we just increment the count).
1396 */
1397int pci_enable_device(struct pci_dev *dev)
1398{
1399        return pci_enable_device_flags(dev, IORESOURCE_MEM | IORESOURCE_IO);
1400}
1401EXPORT_SYMBOL(pci_enable_device);
1402
1403/*
1404 * Managed PCI resources.  This manages device on/off, intx/msi/msix
1405 * on/off and BAR regions.  pci_dev itself records msi/msix status, so
1406 * there's no need to track it separately.  pci_devres is initialized
1407 * when a device is enabled using managed PCI device enable interface.
1408 */
1409struct pci_devres {
1410        unsigned int enabled:1;
1411        unsigned int pinned:1;
1412        unsigned int orig_intx:1;
1413        unsigned int restore_intx:1;
1414        u32 region_mask;
1415};
1416
1417static void pcim_release(struct device *gendev, void *res)
1418{
1419        struct pci_dev *dev = to_pci_dev(gendev);
1420        struct pci_devres *this = res;
1421        int i;
1422
1423        if (dev->msi_enabled)
1424                pci_disable_msi(dev);
1425        if (dev->msix_enabled)
1426                pci_disable_msix(dev);
1427
1428        for (i = 0; i < DEVICE_COUNT_RESOURCE; i++)
1429                if (this->region_mask & (1 << i))
1430                        pci_release_region(dev, i);
1431
1432        if (this->restore_intx)
1433                pci_intx(dev, this->orig_intx);
1434
1435        if (this->enabled && !this->pinned)
1436                pci_disable_device(dev);
1437}
1438
1439static struct pci_devres *get_pci_dr(struct pci_dev *pdev)
1440{
1441        struct pci_devres *dr, *new_dr;
1442
1443        dr = devres_find(&pdev->dev, pcim_release, NULL, NULL);
1444        if (dr)
1445                return dr;
1446
1447        new_dr = devres_alloc(pcim_release, sizeof(*new_dr), GFP_KERNEL);
1448        if (!new_dr)
1449                return NULL;
1450        return devres_get(&pdev->dev, new_dr, NULL, NULL);
1451}
1452
1453static struct pci_devres *find_pci_dr(struct pci_dev *pdev)
1454{
1455        if (pci_is_managed(pdev))
1456                return devres_find(&pdev->dev, pcim_release, NULL, NULL);
1457        return NULL;
1458}
1459
1460/**
1461 * pcim_enable_device - Managed pci_enable_device()
1462 * @pdev: PCI device to be initialized
1463 *
1464 * Managed pci_enable_device().
1465 */
1466int pcim_enable_device(struct pci_dev *pdev)
1467{
1468        struct pci_devres *dr;
1469        int rc;
1470
1471        dr = get_pci_dr(pdev);
1472        if (unlikely(!dr))
1473                return -ENOMEM;
1474        if (dr->enabled)
1475                return 0;
1476
1477        rc = pci_enable_device(pdev);
1478        if (!rc) {
1479                pdev->is_managed = 1;
1480                dr->enabled = 1;
1481        }
1482        return rc;
1483}
1484EXPORT_SYMBOL(pcim_enable_device);
1485
1486/**
1487 * pcim_pin_device - Pin managed PCI device
1488 * @pdev: PCI device to pin
1489 *
1490 * Pin managed PCI device @pdev.  Pinned device won't be disabled on
1491 * driver detach.  @pdev must have been enabled with
1492 * pcim_enable_device().
1493 */
1494void pcim_pin_device(struct pci_dev *pdev)
1495{
1496        struct pci_devres *dr;
1497
1498        dr = find_pci_dr(pdev);
1499        WARN_ON(!dr || !dr->enabled);
1500        if (dr)
1501                dr->pinned = 1;
1502}
1503EXPORT_SYMBOL(pcim_pin_device);
1504
1505/*
1506 * pcibios_add_device - provide arch specific hooks when adding device dev
1507 * @dev: the PCI device being added
1508 *
1509 * Permits the platform to provide architecture specific functionality when
1510 * devices are added. This is the default implementation. Architecture
1511 * implementations can override this.
1512 */
1513int __weak pcibios_add_device(struct pci_dev *dev)
1514{
1515        return 0;
1516}
1517
1518/**
1519 * pcibios_release_device - provide arch specific hooks when releasing device dev
1520 * @dev: the PCI device being released
1521 *
1522 * Permits the platform to provide architecture specific functionality when
1523 * devices are released. This is the default implementation. Architecture
1524 * implementations can override this.
1525 */
1526void __weak pcibios_release_device(struct pci_dev *dev) {}
1527
1528/**
1529 * pcibios_disable_device - disable arch specific PCI resources for device dev
1530 * @dev: the PCI device to disable
1531 *
1532 * Disables architecture specific PCI resources for the device. This
1533 * is the default implementation. Architecture implementations can
1534 * override this.
1535 */
1536void __weak pcibios_disable_device(struct pci_dev *dev) {}
1537
1538/**
1539 * pcibios_penalize_isa_irq - penalize an ISA IRQ
1540 * @irq: ISA IRQ to penalize
1541 * @active: IRQ active or not
1542 *
1543 * Permits the platform to provide architecture-specific functionality when
1544 * penalizing ISA IRQs. This is the default implementation. Architecture
1545 * implementations can override this.
1546 */
1547void __weak pcibios_penalize_isa_irq(int irq, int active) {}
1548
1549static void do_pci_disable_device(struct pci_dev *dev)
1550{
1551        u16 pci_command;
1552
1553        pci_read_config_word(dev, PCI_COMMAND, &pci_command);
1554        if (pci_command & PCI_COMMAND_MASTER) {
1555                pci_command &= ~PCI_COMMAND_MASTER;
1556                pci_write_config_word(dev, PCI_COMMAND, pci_command);
1557        }
1558
1559        pcibios_disable_device(dev);
1560}
1561
1562/**
1563 * pci_disable_enabled_device - Disable device without updating enable_cnt
1564 * @dev: PCI device to disable
1565 *
1566 * NOTE: This function is a backend of PCI power management routines and is
1567 * not supposed to be called drivers.
1568 */
1569void pci_disable_enabled_device(struct pci_dev *dev)
1570{
1571        if (pci_is_enabled(dev))
1572                do_pci_disable_device(dev);
1573}
1574
1575/**
1576 * pci_disable_device - Disable PCI device after use
1577 * @dev: PCI device to be disabled
1578 *
1579 * Signal to the system that the PCI device is not in use by the system
1580 * anymore.  This only involves disabling PCI bus-mastering, if active.
1581 *
1582 * Note we don't actually disable the device until all callers of
1583 * pci_enable_device() have called pci_disable_device().
1584 */
1585void pci_disable_device(struct pci_dev *dev)
1586{
1587        struct pci_devres *dr;
1588
1589        dr = find_pci_dr(dev);
1590        if (dr)
1591                dr->enabled = 0;
1592
1593        dev_WARN_ONCE(&dev->dev, atomic_read(&dev->enable_cnt) <= 0,
1594                      "disabling already-disabled device");
1595
1596        if (atomic_dec_return(&dev->enable_cnt) != 0)
1597                return;
1598
1599        do_pci_disable_device(dev);
1600
1601        dev->is_busmaster = 0;
1602}
1603EXPORT_SYMBOL(pci_disable_device);
1604
1605/**
1606 * pcibios_set_pcie_reset_state - set reset state for device dev
1607 * @dev: the PCIe device reset
1608 * @state: Reset state to enter into
1609 *
1610 *
1611 * Sets the PCIe reset state for the device. This is the default
1612 * implementation. Architecture implementations can override this.
1613 */
1614int __weak pcibios_set_pcie_reset_state(struct pci_dev *dev,
1615                                        enum pcie_reset_state state)
1616{
1617        return -EINVAL;
1618}
1619
1620/**
1621 * pci_set_pcie_reset_state - set reset state for device dev
1622 * @dev: the PCIe device reset
1623 * @state: Reset state to enter into
1624 *
1625 *
1626 * Sets the PCI reset state for the device.
1627 */
1628int pci_set_pcie_reset_state(struct pci_dev *dev, enum pcie_reset_state state)
1629{
1630        return pcibios_set_pcie_reset_state(dev, state);
1631}
1632EXPORT_SYMBOL_GPL(pci_set_pcie_reset_state);
1633
1634/**
1635 * pci_check_pme_status - Check if given device has generated PME.
1636 * @dev: Device to check.
1637 *
1638 * Check the PME status of the device and if set, clear it and clear PME enable
1639 * (if set).  Return 'true' if PME status and PME enable were both set or
1640 * 'false' otherwise.
1641 */
1642bool pci_check_pme_status(struct pci_dev *dev)
1643{
1644        int pmcsr_pos;
1645        u16 pmcsr;
1646        bool ret = false;
1647
1648        if (!dev->pm_cap)
1649                return false;
1650
1651        pmcsr_pos = dev->pm_cap + PCI_PM_CTRL;
1652        pci_read_config_word(dev, pmcsr_pos, &pmcsr);
1653        if (!(pmcsr & PCI_PM_CTRL_PME_STATUS))
1654                return false;
1655
1656        /* Clear PME status. */
1657        pmcsr |= PCI_PM_CTRL_PME_STATUS;
1658        if (pmcsr & PCI_PM_CTRL_PME_ENABLE) {
1659                /* Disable PME to avoid interrupt flood. */
1660                pmcsr &= ~PCI_PM_CTRL_PME_ENABLE;
1661                ret = true;
1662        }
1663
1664        pci_write_config_word(dev, pmcsr_pos, pmcsr);
1665
1666        return ret;
1667}
1668
1669/**
1670 * pci_pme_wakeup - Wake up a PCI device if its PME Status bit is set.
1671 * @dev: Device to handle.
1672 * @pme_poll_reset: Whether or not to reset the device's pme_poll flag.
1673 *
1674 * Check if @dev has generated PME and queue a resume request for it in that
1675 * case.
1676 */
1677static int pci_pme_wakeup(struct pci_dev *dev, void *pme_poll_reset)
1678{
1679        if (pme_poll_reset && dev->pme_poll)
1680                dev->pme_poll = false;
1681
1682        if (pci_check_pme_status(dev)) {
1683                pci_wakeup_event(dev);
1684                pm_request_resume(&dev->dev);
1685        }
1686        return 0;
1687}
1688
1689/**
1690 * pci_pme_wakeup_bus - Walk given bus and wake up devices on it, if necessary.
1691 * @bus: Top bus of the subtree to walk.
1692 */
1693void pci_pme_wakeup_bus(struct pci_bus *bus)
1694{
1695        if (bus)
1696                pci_walk_bus(bus, pci_pme_wakeup, (void *)true);
1697}
1698
1699
1700/**
1701 * pci_pme_capable - check the capability of PCI device to generate PME#
1702 * @dev: PCI device to handle.
1703 * @state: PCI state from which device will issue PME#.
1704 */
1705bool pci_pme_capable(struct pci_dev *dev, pci_power_t state)
1706{
1707        if (!dev->pm_cap)
1708                return false;
1709
1710        return !!(dev->pme_support & (1 << state));
1711}
1712EXPORT_SYMBOL(pci_pme_capable);
1713
1714static void pci_pme_list_scan(struct work_struct *work)
1715{
1716        struct pci_pme_device *pme_dev, *n;
1717
1718        mutex_lock(&pci_pme_list_mutex);
1719        list_for_each_entry_safe(pme_dev, n, &pci_pme_list, list) {
1720                if (pme_dev->dev->pme_poll) {
1721                        struct pci_dev *bridge;
1722
1723                        bridge = pme_dev->dev->bus->self;
1724                        /*
1725                         * If bridge is in low power state, the
1726                         * configuration space of subordinate devices
1727                         * may be not accessible
1728                         */
1729                        if (bridge && bridge->current_state != PCI_D0)
1730                                continue;
1731                        pci_pme_wakeup(pme_dev->dev, NULL);
1732                } else {
1733                        list_del(&pme_dev->list);
1734                        kfree(pme_dev);
1735                }
1736        }
1737        if (!list_empty(&pci_pme_list))
1738                schedule_delayed_work(&pci_pme_work,
1739                                      msecs_to_jiffies(PME_TIMEOUT));
1740        mutex_unlock(&pci_pme_list_mutex);
1741}
1742
1743static void __pci_pme_active(struct pci_dev *dev, bool enable)
1744{
1745        u16 pmcsr;
1746
1747        if (!dev->pme_support)
1748                return;
1749
1750        pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
1751        /* Clear PME_Status by writing 1 to it and enable PME# */
1752        pmcsr |= PCI_PM_CTRL_PME_STATUS | PCI_PM_CTRL_PME_ENABLE;
1753        if (!enable)
1754                pmcsr &= ~PCI_PM_CTRL_PME_ENABLE;
1755
1756        pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, pmcsr);
1757}
1758
1759/**
1760 * pci_pme_active - enable or disable PCI device's PME# function
1761 * @dev: PCI device to handle.
1762 * @enable: 'true' to enable PME# generation; 'false' to disable it.
1763 *
1764 * The caller must verify that the device is capable of generating PME# before
1765 * calling this function with @enable equal to 'true'.
1766 */
1767void pci_pme_active(struct pci_dev *dev, bool enable)
1768{
1769        __pci_pme_active(dev, enable);
1770
1771        /*
1772         * PCI (as opposed to PCIe) PME requires that the device have
1773         * its PME# line hooked up correctly. Not all hardware vendors
1774         * do this, so the PME never gets delivered and the device
1775         * remains asleep. The easiest way around this is to
1776         * periodically walk the list of suspended devices and check
1777         * whether any have their PME flag set. The assumption is that
1778         * we'll wake up often enough anyway that this won't be a huge
1779         * hit, and the power savings from the devices will still be a
1780         * win.
1781         *
1782         * Although PCIe uses in-band PME message instead of PME# line
1783         * to report PME, PME does not work for some PCIe devices in
1784         * reality.  For example, there are devices that set their PME
1785         * status bits, but don't really bother to send a PME message;
1786         * there are PCI Express Root Ports that don't bother to
1787         * trigger interrupts when they receive PME messages from the
1788         * devices below.  So PME poll is used for PCIe devices too.
1789         */
1790
1791        if (dev->pme_poll) {
1792                struct pci_pme_device *pme_dev;
1793                if (enable) {
1794                        pme_dev = kmalloc(sizeof(struct pci_pme_device),
1795                                          GFP_KERNEL);
1796                        if (!pme_dev) {
1797                                dev_warn(&dev->dev, "can't enable PME#\n");
1798                                return;
1799                        }
1800                        pme_dev->dev = dev;
1801                        mutex_lock(&pci_pme_list_mutex);
1802                        list_add(&pme_dev->list, &pci_pme_list);
1803                        if (list_is_singular(&pci_pme_list))
1804                                schedule_delayed_work(&pci_pme_work,
1805                                                      msecs_to_jiffies(PME_TIMEOUT));
1806                        mutex_unlock(&pci_pme_list_mutex);
1807                } else {
1808                        mutex_lock(&pci_pme_list_mutex);
1809                        list_for_each_entry(pme_dev, &pci_pme_list, list) {
1810                                if (pme_dev->dev == dev) {
1811                                        list_del(&pme_dev->list);
1812                                        kfree(pme_dev);
1813                                        break;
1814                                }
1815                        }
1816                        mutex_unlock(&pci_pme_list_mutex);
1817                }
1818        }
1819
1820        dev_dbg(&dev->dev, "PME# %s\n", enable ? "enabled" : "disabled");
1821}
1822EXPORT_SYMBOL(pci_pme_active);
1823
1824/**
1825 * __pci_enable_wake - enable PCI device as wakeup event source
1826 * @dev: PCI device affected
1827 * @state: PCI state from which device will issue wakeup events
1828 * @runtime: True if the events are to be generated at run time
1829 * @enable: True to enable event generation; false to disable
1830 *
1831 * This enables the device as a wakeup event source, or disables it.
1832 * When such events involves platform-specific hooks, those hooks are
1833 * called automatically by this routine.
1834 *
1835 * Devices with legacy power management (no standard PCI PM capabilities)
1836 * always require such platform hooks.
1837 *
1838 * RETURN VALUE:
1839 * 0 is returned on success
1840 * -EINVAL is returned if device is not supposed to wake up the system
1841 * Error code depending on the platform is returned if both the platform and
1842 * the native mechanism fail to enable the generation of wake-up events
1843 */
1844int __pci_enable_wake(struct pci_dev *dev, pci_power_t state,
1845                      bool runtime, bool enable)
1846{
1847        int ret = 0;
1848
1849        if (enable && !runtime && !device_may_wakeup(&dev->dev))
1850                return -EINVAL;
1851
1852        /* Don't do the same thing twice in a row for one device. */
1853        if (!!enable == !!dev->wakeup_prepared)
1854                return 0;
1855
1856        /*
1857         * According to "PCI System Architecture" 4th ed. by Tom Shanley & Don
1858         * Anderson we should be doing PME# wake enable followed by ACPI wake
1859         * enable.  To disable wake-up we call the platform first, for symmetry.
1860         */
1861
1862        if (enable) {
1863                int error;
1864
1865                if (pci_pme_capable(dev, state))
1866                        pci_pme_active(dev, true);
1867                else
1868                        ret = 1;
1869                error = runtime ? platform_pci_run_wake(dev, true) :
1870                                        platform_pci_sleep_wake(dev, true);
1871                if (ret)
1872                        ret = error;
1873                if (!ret)
1874                        dev->wakeup_prepared = true;
1875        } else {
1876                if (runtime)
1877                        platform_pci_run_wake(dev, false);
1878                else
1879                        platform_pci_sleep_wake(dev, false);
1880                pci_pme_active(dev, false);
1881                dev->wakeup_prepared = false;
1882        }
1883
1884        return ret;
1885}
1886EXPORT_SYMBOL(__pci_enable_wake);
1887
1888/**
1889 * pci_wake_from_d3 - enable/disable device to wake up from D3_hot or D3_cold
1890 * @dev: PCI device to prepare
1891 * @enable: True to enable wake-up event generation; false to disable
1892 *
1893 * Many drivers want the device to wake up the system from D3_hot or D3_cold
1894 * and this function allows them to set that up cleanly - pci_enable_wake()
1895 * should not be called twice in a row to enable wake-up due to PCI PM vs ACPI
1896 * ordering constraints.
1897 *
1898 * This function only returns error code if the device is not capable of
1899 * generating PME# from both D3_hot and D3_cold, and the platform is unable to
1900 * enable wake-up power for it.
1901 */
1902int pci_wake_from_d3(struct pci_dev *dev, bool enable)
1903{
1904        return pci_pme_capable(dev, PCI_D3cold) ?
1905                        pci_enable_wake(dev, PCI_D3cold, enable) :
1906                        pci_enable_wake(dev, PCI_D3hot, enable);
1907}
1908EXPORT_SYMBOL(pci_wake_from_d3);
1909
1910/**
1911 * pci_target_state - find an appropriate low power state for a given PCI dev
1912 * @dev: PCI device
1913 *
1914 * Use underlying platform code to find a supported low power state for @dev.
1915 * If the platform can't manage @dev, return the deepest state from which it
1916 * can generate wake events, based on any available PME info.
1917 */
1918static pci_power_t pci_target_state(struct pci_dev *dev)
1919{
1920        pci_power_t target_state = PCI_D3hot;
1921
1922        if (platform_pci_power_manageable(dev)) {
1923                /*
1924                 * Call the platform to choose the target state of the device
1925                 * and enable wake-up from this state if supported.
1926                 */
1927                pci_power_t state = platform_pci_choose_state(dev);
1928
1929                switch (state) {
1930                case PCI_POWER_ERROR:
1931                case PCI_UNKNOWN:
1932                        break;
1933                case PCI_D1:
1934                case PCI_D2:
1935                        if (pci_no_d1d2(dev))
1936                                break;
1937                default:
1938                        target_state = state;
1939                }
1940        } else if (!dev->pm_cap) {
1941                target_state = PCI_D0;
1942        } else if (device_may_wakeup(&dev->dev)) {
1943                /*
1944                 * Find the deepest state from which the device can generate
1945                 * wake-up events, make it the target state and enable device
1946                 * to generate PME#.
1947                 */
1948                if (dev->pme_support) {
1949                        while (target_state
1950                              && !(dev->pme_support & (1 << target_state)))
1951                                target_state--;
1952                }
1953        }
1954
1955        return target_state;
1956}
1957
1958/**
1959 * pci_prepare_to_sleep - prepare PCI device for system-wide transition into a sleep state
1960 * @dev: Device to handle.
1961 *
1962 * Choose the power state appropriate for the device depending on whether
1963 * it can wake up the system and/or is power manageable by the platform
1964 * (PCI_D3hot is the default) and put the device into that state.
1965 */
1966int pci_prepare_to_sleep(struct pci_dev *dev)
1967{
1968        pci_power_t target_state = pci_target_state(dev);
1969        int error;
1970
1971        if (target_state == PCI_POWER_ERROR)
1972                return -EIO;
1973
1974        pci_enable_wake(dev, target_state, device_may_wakeup(&dev->dev));
1975
1976        error = pci_set_power_state(dev, target_state);
1977
1978        if (error)
1979                pci_enable_wake(dev, target_state, false);
1980
1981        return error;
1982}
1983EXPORT_SYMBOL(pci_prepare_to_sleep);
1984
1985/**
1986 * pci_back_from_sleep - turn PCI device on during system-wide transition into working state
1987 * @dev: Device to handle.
1988 *
1989 * Disable device's system wake-up capability and put it into D0.
1990 */
1991int pci_back_from_sleep(struct pci_dev *dev)
1992{
1993        pci_enable_wake(dev, PCI_D0, false);
1994        return pci_set_power_state(dev, PCI_D0);
1995}
1996EXPORT_SYMBOL(pci_back_from_sleep);
1997
1998/**
1999 * pci_finish_runtime_suspend - Carry out PCI-specific part of runtime suspend.
2000 * @dev: PCI device being suspended.
2001 *
2002 * Prepare @dev to generate wake-up events at run time and put it into a low
2003 * power state.
2004 */
2005int pci_finish_runtime_suspend(struct pci_dev *dev)
2006{
2007        pci_power_t target_state = pci_target_state(dev);
2008        int error;
2009
2010        if (target_state == PCI_POWER_ERROR)
2011                return -EIO;
2012
2013        dev->runtime_d3cold = target_state == PCI_D3cold;
2014
2015        __pci_enable_wake(dev, target_state, true, pci_dev_run_wake(dev));
2016
2017        error = pci_set_power_state(dev, target_state);
2018
2019        if (error) {
2020                __pci_enable_wake(dev, target_state, true, false);
2021                dev->runtime_d3cold = false;
2022        }
2023
2024        return error;
2025}
2026
2027/**
2028 * pci_dev_run_wake - Check if device can generate run-time wake-up events.
2029 * @dev: Device to check.
2030 *
2031 * Return true if the device itself is capable of generating wake-up events
2032 * (through the platform or using the native PCIe PME) or if the device supports
2033 * PME and one of its upstream bridges can generate wake-up events.
2034 */
2035bool pci_dev_run_wake(struct pci_dev *dev)
2036{
2037        struct pci_bus *bus = dev->bus;
2038
2039        if (device_run_wake(&dev->dev))
2040                return true;
2041
2042        if (!dev->pme_support)
2043                return false;
2044
2045        while (bus->parent) {
2046                struct pci_dev *bridge = bus->self;
2047
2048                if (device_run_wake(&bridge->dev))
2049                        return true;
2050
2051                bus = bus->parent;
2052        }
2053
2054        /* We have reached the root bus. */
2055        if (bus->bridge)
2056                return device_run_wake(bus->bridge);
2057
2058        return false;
2059}
2060EXPORT_SYMBOL_GPL(pci_dev_run_wake);
2061
2062/**
2063 * pci_dev_keep_suspended - Check if the device can stay in the suspended state.
2064 * @pci_dev: Device to check.
2065 *
2066 * Return 'true' if the device is runtime-suspended, it doesn't have to be
2067 * reconfigured due to wakeup settings difference between system and runtime
2068 * suspend and the current power state of it is suitable for the upcoming
2069 * (system) transition.
2070 *
2071 * If the device is not configured for system wakeup, disable PME for it before
2072 * returning 'true' to prevent it from waking up the system unnecessarily.
2073 */
2074bool pci_dev_keep_suspended(struct pci_dev *pci_dev)
2075{
2076        struct device *dev = &pci_dev->dev;
2077
2078        if (!pm_runtime_suspended(dev)
2079            || pci_target_state(pci_dev) != pci_dev->current_state
2080            || platform_pci_need_resume(pci_dev))
2081                return false;
2082
2083        /*
2084         * At this point the device is good to go unless it's been configured
2085         * to generate PME at the runtime suspend time, but it is not supposed
2086         * to wake up the system.  In that case, simply disable PME for it
2087         * (it will have to be re-enabled on exit from system resume).
2088         *
2089         * If the device's power state is D3cold and the platform check above
2090         * hasn't triggered, the device's configuration is suitable and we don't
2091         * need to manipulate it at all.
2092         */
2093        spin_lock_irq(&dev->power.lock);
2094
2095        if (pm_runtime_suspended(dev) && pci_dev->current_state < PCI_D3cold &&
2096            !device_may_wakeup(dev))
2097                __pci_pme_active(pci_dev, false);
2098
2099        spin_unlock_irq(&dev->power.lock);
2100        return true;
2101}
2102
2103/**
2104 * pci_dev_complete_resume - Finalize resume from system sleep for a device.
2105 * @pci_dev: Device to handle.
2106 *
2107 * If the device is runtime suspended and wakeup-capable, enable PME for it as
2108 * it might have been disabled during the prepare phase of system suspend if
2109 * the device was not configured for system wakeup.
2110 */
2111void pci_dev_complete_resume(struct pci_dev *pci_dev)
2112{
2113        struct device *dev = &pci_dev->dev;
2114
2115        if (!pci_dev_run_wake(pci_dev))
2116                return;
2117
2118        spin_lock_irq(&dev->power.lock);
2119
2120        if (pm_runtime_suspended(dev) && pci_dev->current_state < PCI_D3cold)
2121                __pci_pme_active(pci_dev, true);
2122
2123        spin_unlock_irq(&dev->power.lock);
2124}
2125
2126void pci_config_pm_runtime_get(struct pci_dev *pdev)
2127{
2128        struct device *dev = &pdev->dev;
2129        struct device *parent = dev->parent;
2130
2131        if (parent)
2132                pm_runtime_get_sync(parent);
2133        pm_runtime_get_noresume(dev);
2134        /*
2135         * pdev->current_state is set to PCI_D3cold during suspending,
2136         * so wait until suspending completes
2137         */
2138        pm_runtime_barrier(dev);
2139        /*
2140         * Only need to resume devices in D3cold, because config
2141         * registers are still accessible for devices suspended but
2142         * not in D3cold.
2143         */
2144        if (pdev->current_state == PCI_D3cold)
2145                pm_runtime_resume(dev);
2146}
2147
2148void pci_config_pm_runtime_put(struct pci_dev *pdev)
2149{
2150        struct device *dev = &pdev->dev;
2151        struct device *parent = dev->parent;
2152
2153        pm_runtime_put(dev);
2154        if (parent)
2155                pm_runtime_put_sync(parent);
2156}
2157
2158/**
2159 * pci_pm_init - Initialize PM functions of given PCI device
2160 * @dev: PCI device to handle.
2161 */
2162void pci_pm_init(struct pci_dev *dev)
2163{
2164        int pm;
2165        u16 pmc;
2166
2167        pm_runtime_forbid(&dev->dev);
2168        pm_runtime_set_active(&dev->dev);
2169        pm_runtime_enable(&dev->dev);
2170        device_enable_async_suspend(&dev->dev);
2171        dev->wakeup_prepared = false;
2172
2173        dev->pm_cap = 0;
2174        dev->pme_support = 0;
2175
2176        /* find PCI PM capability in list */
2177        pm = pci_find_capability(dev, PCI_CAP_ID_PM);
2178        if (!pm)
2179                return;
2180        /* Check device's ability to generate PME# */
2181        pci_read_config_word(dev, pm + PCI_PM_PMC, &pmc);
2182
2183        if ((pmc & PCI_PM_CAP_VER_MASK) > 3) {
2184                dev_err(&dev->dev, "unsupported PM cap regs version (%u)\n",
2185                        pmc & PCI_PM_CAP_VER_MASK);
2186                return;
2187        }
2188
2189        dev->pm_cap = pm;
2190        dev->d3_delay = PCI_PM_D3_WAIT;
2191        dev->d3cold_delay = PCI_PM_D3COLD_WAIT;
2192        dev->d3cold_allowed = true;
2193
2194        dev->d1_support = false;
2195        dev->d2_support = false;
2196        if (!pci_no_d1d2(dev)) {
2197                if (pmc & PCI_PM_CAP_D1)
2198                        dev->d1_support = true;
2199                if (pmc & PCI_PM_CAP_D2)
2200                        dev->d2_support = true;
2201
2202                if (dev->d1_support || dev->d2_support)
2203                        dev_printk(KERN_DEBUG, &dev->dev, "supports%s%s\n",
2204                                   dev->d1_support ? " D1" : "",
2205                                   dev->d2_support ? " D2" : "");
2206        }
2207
2208        pmc &= PCI_PM_CAP_PME_MASK;
2209        if (pmc) {
2210                dev_printk(KERN_DEBUG, &dev->dev,
2211                         "PME# supported from%s%s%s%s%s\n",
2212                         (pmc & PCI_PM_CAP_PME_D0) ? " D0" : "",
2213                         (pmc & PCI_PM_CAP_PME_D1) ? " D1" : "",
2214                         (pmc & PCI_PM_CAP_PME_D2) ? " D2" : "",
2215                         (pmc & PCI_PM_CAP_PME_D3) ? " D3hot" : "",
2216                         (pmc & PCI_PM_CAP_PME_D3cold) ? " D3cold" : "");
2217                dev->pme_support = pmc >> PCI_PM_CAP_PME_SHIFT;
2218                dev->pme_poll = true;
2219                /*
2220                 * Make device's PM flags reflect the wake-up capability, but
2221                 * let the user space enable it to wake up the system as needed.
2222                 */
2223                device_set_wakeup_capable(&dev->dev, true);
2224                /* Disable the PME# generation functionality */
2225                pci_pme_active(dev, false);
2226        }
2227}
2228
2229static unsigned long pci_ea_flags(struct pci_dev *dev, u8 prop)
2230{
2231        unsigned long flags = IORESOURCE_PCI_FIXED;
2232
2233        switch (prop) {
2234        case PCI_EA_P_MEM:
2235        case PCI_EA_P_VF_MEM:
2236                flags |= IORESOURCE_MEM;
2237                break;
2238        case PCI_EA_P_MEM_PREFETCH:
2239        case PCI_EA_P_VF_MEM_PREFETCH:
2240                flags |= IORESOURCE_MEM | IORESOURCE_PREFETCH;
2241                break;
2242        case PCI_EA_P_IO:
2243                flags |= IORESOURCE_IO;
2244                break;
2245        default:
2246                return 0;
2247        }
2248
2249        return flags;
2250}
2251
2252static struct resource *pci_ea_get_resource(struct pci_dev *dev, u8 bei,
2253                                            u8 prop)
2254{
2255        if (bei <= PCI_EA_BEI_BAR5 && prop <= PCI_EA_P_IO)
2256                return &dev->resource[bei];
2257#ifdef CONFIG_PCI_IOV
2258        else if (bei >= PCI_EA_BEI_VF_BAR0 && bei <= PCI_EA_BEI_VF_BAR5 &&
2259                 (prop == PCI_EA_P_VF_MEM || prop == PCI_EA_P_VF_MEM_PREFETCH))
2260                return &dev->resource[PCI_IOV_RESOURCES +
2261                                      bei - PCI_EA_BEI_VF_BAR0];
2262#endif
2263        else if (bei == PCI_EA_BEI_ROM)
2264                return &dev->resource[PCI_ROM_RESOURCE];
2265        else
2266                return NULL;
2267}
2268
2269/* Read an Enhanced Allocation (EA) entry */
2270static int pci_ea_read(struct pci_dev *dev, int offset)
2271{
2272        struct resource *res;
2273        int ent_size, ent_offset = offset;
2274        resource_size_t start, end;
2275        unsigned long flags;
2276        u32 dw0, bei, base, max_offset;
2277        u8 prop;
2278        bool support_64 = (sizeof(resource_size_t) >= 8);
2279
2280        pci_read_config_dword(dev, ent_offset, &dw0);
2281        ent_offset += 4;
2282
2283        /* Entry size field indicates DWORDs after 1st */
2284        ent_size = ((dw0 & PCI_EA_ES) + 1) << 2;
2285
2286        if (!(dw0 & PCI_EA_ENABLE)) /* Entry not enabled */
2287                goto out;
2288
2289        bei = (dw0 & PCI_EA_BEI) >> 4;
2290        prop = (dw0 & PCI_EA_PP) >> 8;
2291
2292        /*
2293         * If the Property is in the reserved range, try the Secondary
2294         * Property instead.
2295         */
2296        if (prop > PCI_EA_P_BRIDGE_IO && prop < PCI_EA_P_MEM_RESERVED)
2297                prop = (dw0 & PCI_EA_SP) >> 16;
2298        if (prop > PCI_EA_P_BRIDGE_IO)
2299                goto out;
2300
2301        res = pci_ea_get_resource(dev, bei, prop);
2302        if (!res) {
2303                dev_err(&dev->dev, "Unsupported EA entry BEI: %u\n", bei);
2304                goto out;
2305        }
2306
2307        flags = pci_ea_flags(dev, prop);
2308        if (!flags) {
2309                dev_err(&dev->dev, "Unsupported EA properties: %#x\n", prop);
2310                goto out;
2311        }
2312
2313        /* Read Base */
2314        pci_read_config_dword(dev, ent_offset, &base);
2315        start = (base & PCI_EA_FIELD_MASK);
2316        ent_offset += 4;
2317
2318        /* Read MaxOffset */
2319        pci_read_config_dword(dev, ent_offset, &max_offset);
2320        ent_offset += 4;
2321
2322        /* Read Base MSBs (if 64-bit entry) */
2323        if (base & PCI_EA_IS_64) {
2324                u32 base_upper;
2325
2326                pci_read_config_dword(dev, ent_offset, &base_upper);
2327                ent_offset += 4;
2328
2329                flags |= IORESOURCE_MEM_64;
2330
2331                /* entry starts above 32-bit boundary, can't use */
2332                if (!support_64 && base_upper)
2333                        goto out;
2334
2335                if (support_64)
2336                        start |= ((u64)base_upper << 32);
2337        }
2338
2339        end = start + (max_offset | 0x03);
2340
2341        /* Read MaxOffset MSBs (if 64-bit entry) */
2342        if (max_offset & PCI_EA_IS_64) {
2343                u32 max_offset_upper;
2344
2345                pci_read_config_dword(dev, ent_offset, &max_offset_upper);
2346                ent_offset += 4;
2347
2348                flags |= IORESOURCE_MEM_64;
2349
2350                /* entry too big, can't use */
2351                if (!support_64 && max_offset_upper)
2352                        goto out;
2353
2354                if (support_64)
2355                        end += ((u64)max_offset_upper << 32);
2356        }
2357
2358        if (end < start) {
2359                dev_err(&dev->dev, "EA Entry crosses address boundary\n");
2360                goto out;
2361        }
2362
2363        if (ent_size != ent_offset - offset) {
2364                dev_err(&dev->dev,
2365                        "EA Entry Size (%d) does not match length read (%d)\n",
2366                        ent_size, ent_offset - offset);
2367                goto out;
2368        }
2369
2370        res->name = pci_name(dev);
2371        res->start = start;
2372        res->end = end;
2373        res->flags = flags;
2374
2375        if (bei <= PCI_EA_BEI_BAR5)
2376                dev_printk(KERN_DEBUG, &dev->dev, "BAR %d: %pR (from Enhanced Allocation, properties %#02x)\n",
2377                           bei, res, prop);
2378        else if (bei == PCI_EA_BEI_ROM)
2379                dev_printk(KERN_DEBUG, &dev->dev, "ROM: %pR (from Enhanced Allocation, properties %#02x)\n",
2380                           res, prop);
2381        else if (bei >= PCI_EA_BEI_VF_BAR0 && bei <= PCI_EA_BEI_VF_BAR5)
2382                dev_printk(KERN_DEBUG, &dev->dev, "VF BAR %d: %pR (from Enhanced Allocation, properties %#02x)\n",
2383                           bei - PCI_EA_BEI_VF_BAR0, res, prop);
2384        else
2385                dev_printk(KERN_DEBUG, &dev->dev, "BEI %d res: %pR (from Enhanced Allocation, properties %#02x)\n",
2386                           bei, res, prop);
2387
2388out:
2389        return offset + ent_size;
2390}
2391
2392/* Enhanced Allocation Initalization */
2393void pci_ea_init(struct pci_dev *dev)
2394{
2395        int ea;
2396        u8 num_ent;
2397        int offset;
2398        int i;
2399
2400        /* find PCI EA capability in list */
2401        ea = pci_find_capability(dev, PCI_CAP_ID_EA);
2402        if (!ea)
2403                return;
2404
2405        /* determine the number of entries */
2406        pci_bus_read_config_byte(dev->bus, dev->devfn, ea + PCI_EA_NUM_ENT,
2407                                        &num_ent);
2408        num_ent &= PCI_EA_NUM_ENT_MASK;
2409
2410        offset = ea + PCI_EA_FIRST_ENT;
2411
2412        /* Skip DWORD 2 for type 1 functions */
2413        if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE)
2414                offset += 4;
2415
2416        /* parse each EA entry */
2417        for (i = 0; i < num_ent; ++i)
2418                offset = pci_ea_read(dev, offset);
2419}
2420
2421static void pci_add_saved_cap(struct pci_dev *pci_dev,
2422        struct pci_cap_saved_state *new_cap)
2423{
2424        hlist_add_head(&new_cap->next, &pci_dev->saved_cap_space);
2425}
2426
2427/**
2428 * _pci_add_cap_save_buffer - allocate buffer for saving given
2429 *                            capability registers
2430 * @dev: the PCI device
2431 * @cap: the capability to allocate the buffer for
2432 * @extended: Standard or Extended capability ID
2433 * @size: requested size of the buffer
2434 */
2435static int _pci_add_cap_save_buffer(struct pci_dev *dev, u16 cap,
2436                                    bool extended, unsigned int size)
2437{
2438        int pos;
2439        struct pci_cap_saved_state *save_state;
2440
2441        if (extended)
2442                pos = pci_find_ext_capability(dev, cap);
2443        else
2444                pos = pci_find_capability(dev, cap);
2445
2446        if (!pos)
2447                return 0;
2448
2449        save_state = kzalloc(sizeof(*save_state) + size, GFP_KERNEL);
2450        if (!save_state)
2451                return -ENOMEM;
2452
2453        save_state->cap.cap_nr = cap;
2454        save_state->cap.cap_extended = extended;
2455        save_state->cap.size = size;
2456        pci_add_saved_cap(dev, save_state);
2457
2458        return 0;
2459}
2460
2461int pci_add_cap_save_buffer(struct pci_dev *dev, char cap, unsigned int size)
2462{
2463        return _pci_add_cap_save_buffer(dev, cap, false, size);
2464}
2465
2466int pci_add_ext_cap_save_buffer(struct pci_dev *dev, u16 cap, unsigned int size)
2467{
2468        return _pci_add_cap_save_buffer(dev, cap, true, size);
2469}
2470
2471/**
2472 * pci_allocate_cap_save_buffers - allocate buffers for saving capabilities
2473 * @dev: the PCI device
2474 */
2475void pci_allocate_cap_save_buffers(struct pci_dev *dev)
2476{
2477        int error;
2478
2479        error = pci_add_cap_save_buffer(dev, PCI_CAP_ID_EXP,
2480                                        PCI_EXP_SAVE_REGS * sizeof(u16));
2481        if (error)
2482                dev_err(&dev->dev,
2483                        "unable to preallocate PCI Express save buffer\n");
2484
2485        error = pci_add_cap_save_buffer(dev, PCI_CAP_ID_PCIX, sizeof(u16));
2486        if (error)
2487                dev_err(&dev->dev,
2488                        "unable to preallocate PCI-X save buffer\n");
2489
2490        pci_allocate_vc_save_buffers(dev);
2491}
2492
2493void pci_free_cap_save_buffers(struct pci_dev *dev)
2494{
2495        struct pci_cap_saved_state *tmp;
2496        struct hlist_node *n;
2497
2498        hlist_for_each_entry_safe(tmp, n, &dev->saved_cap_space, next)
2499                kfree(tmp);
2500}
2501
2502/**
2503 * pci_configure_ari - enable or disable ARI forwarding
2504 * @dev: the PCI device
2505 *
2506 * If @dev and its upstream bridge both support ARI, enable ARI in the
2507 * bridge.  Otherwise, disable ARI in the bridge.
2508 */
2509void pci_configure_ari(struct pci_dev *dev)
2510{
2511        u32 cap;
2512        struct pci_dev *bridge;
2513
2514        if (pcie_ari_disabled || !pci_is_pcie(dev) || dev->devfn)
2515                return;
2516
2517        bridge = dev->bus->self;
2518        if (!bridge)
2519                return;
2520
2521        pcie_capability_read_dword(bridge, PCI_EXP_DEVCAP2, &cap);
2522        if (!(cap & PCI_EXP_DEVCAP2_ARI))
2523                return;
2524
2525        if (pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ARI)) {
2526                pcie_capability_set_word(bridge, PCI_EXP_DEVCTL2,
2527                                         PCI_EXP_DEVCTL2_ARI);
2528                bridge->ari_enabled = 1;
2529        } else {
2530                pcie_capability_clear_word(bridge, PCI_EXP_DEVCTL2,
2531                                           PCI_EXP_DEVCTL2_ARI);
2532                bridge->ari_enabled = 0;
2533        }
2534}
2535
2536static int pci_acs_enable;
2537
2538/**
2539 * pci_request_acs - ask for ACS to be enabled if supported
2540 */
2541void pci_request_acs(void)
2542{
2543        pci_acs_enable = 1;
2544}
2545
2546/**
2547 * pci_std_enable_acs - enable ACS on devices using standard ACS capabilites
2548 * @dev: the PCI device
2549 */
2550static int pci_std_enable_acs(struct pci_dev *dev)
2551{
2552        int pos;
2553        u16 cap;
2554        u16 ctrl;
2555
2556        pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ACS);
2557        if (!pos)
2558                return -ENODEV;
2559
2560        pci_read_config_word(dev, pos + PCI_ACS_CAP, &cap);
2561        pci_read_config_word(dev, pos + PCI_ACS_CTRL, &ctrl);
2562
2563        /* Source Validation */
2564        ctrl |= (cap & PCI_ACS_SV);
2565
2566        /* P2P Request Redirect */
2567        ctrl |= (cap & PCI_ACS_RR);
2568
2569        /* P2P Completion Redirect */
2570        ctrl |= (cap & PCI_ACS_CR);
2571
2572        /* Upstream Forwarding */
2573        ctrl |= (cap & PCI_ACS_UF);
2574
2575        pci_write_config_word(dev, pos + PCI_ACS_CTRL, ctrl);
2576
2577        return 0;
2578}
2579
2580/**
2581 * pci_enable_acs - enable ACS if hardware support it
2582 * @dev: the PCI device
2583 */
2584void pci_enable_acs(struct pci_dev *dev)
2585{
2586        if (!pci_acs_enable)
2587                return;
2588
2589        if (!pci_std_enable_acs(dev))
2590                return;
2591
2592        pci_dev_specific_enable_acs(dev);
2593}
2594
2595static bool pci_acs_flags_enabled(struct pci_dev *pdev, u16 acs_flags)
2596{
2597        int pos;
2598        u16 cap, ctrl;
2599
2600        pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ACS);
2601        if (!pos)
2602                return false;
2603
2604        /*
2605         * Except for egress control, capabilities are either required
2606         * or only required if controllable.  Features missing from the
2607         * capability field can therefore be assumed as hard-wired enabled.
2608         */
2609        pci_read_config_word(pdev, pos + PCI_ACS_CAP, &cap);
2610        acs_flags &= (cap | PCI_ACS_EC);
2611
2612        pci_read_config_word(pdev, pos + PCI_ACS_CTRL, &ctrl);
2613        return (ctrl & acs_flags) == acs_flags;
2614}
2615
2616/**
2617 * pci_acs_enabled - test ACS against required flags for a given device
2618 * @pdev: device to test
2619 * @acs_flags: required PCI ACS flags
2620 *
2621 * Return true if the device supports the provided flags.  Automatically
2622 * filters out flags that are not implemented on multifunction devices.
2623 *
2624 * Note that this interface checks the effective ACS capabilities of the
2625 * device rather than the actual capabilities.  For instance, most single
2626 * function endpoints are not required to support ACS because they have no
2627 * opportunity for peer-to-peer access.  We therefore return 'true'
2628 * regardless of whether the device exposes an ACS capability.  This makes
2629 * it much easier for callers of this function to ignore the actual type
2630 * or topology of the device when testing ACS support.
2631 */
2632bool pci_acs_enabled(struct pci_dev *pdev, u16 acs_flags)
2633{
2634        int ret;
2635
2636        ret = pci_dev_specific_acs_enabled(pdev, acs_flags);
2637        if (ret >= 0)
2638                return ret > 0;
2639
2640        /*
2641         * Conventional PCI and PCI-X devices never support ACS, either
2642         * effectively or actually.  The shared bus topology implies that
2643         * any device on the bus can receive or snoop DMA.
2644         */
2645        if (!pci_is_pcie(pdev))
2646                return false;
2647
2648        switch (pci_pcie_type(pdev)) {
2649        /*
2650         * PCI/X-to-PCIe bridges are not specifically mentioned by the spec,
2651         * but since their primary interface is PCI/X, we conservatively
2652         * handle them as we would a non-PCIe device.
2653         */
2654        case PCI_EXP_TYPE_PCIE_BRIDGE:
2655        /*
2656         * PCIe 3.0, 6.12.1 excludes ACS on these devices.  "ACS is never
2657         * applicable... must never implement an ACS Extended Capability...".
2658         * This seems arbitrary, but we take a conservative interpretation
2659         * of this statement.
2660         */
2661        case PCI_EXP_TYPE_PCI_BRIDGE:
2662        case PCI_EXP_TYPE_RC_EC:
2663                return false;
2664        /*
2665         * PCIe 3.0, 6.12.1.1 specifies that downstream and root ports should
2666         * implement ACS in order to indicate their peer-to-peer capabilities,
2667         * regardless of whether they are single- or multi-function devices.
2668         */
2669        case PCI_EXP_TYPE_DOWNSTREAM:
2670        case PCI_EXP_TYPE_ROOT_PORT:
2671                return pci_acs_flags_enabled(pdev, acs_flags);
2672        /*
2673         * PCIe 3.0, 6.12.1.2 specifies ACS capabilities that should be
2674         * implemented by the remaining PCIe types to indicate peer-to-peer
2675         * capabilities, but only when they are part of a multifunction
2676         * device.  The footnote for section 6.12 indicates the specific
2677         * PCIe types included here.
2678         */
2679        case PCI_EXP_TYPE_ENDPOINT:
2680        case PCI_EXP_TYPE_UPSTREAM:
2681        case PCI_EXP_TYPE_LEG_END:
2682        case PCI_EXP_TYPE_RC_END:
2683                if (!pdev->multifunction)
2684                        break;
2685
2686                return pci_acs_flags_enabled(pdev, acs_flags);
2687        }
2688
2689        /*
2690         * PCIe 3.0, 6.12.1.3 specifies no ACS capabilities are applicable
2691         * to single function devices with the exception of downstream ports.
2692         */
2693        return true;
2694}
2695
2696/**
2697 * pci_acs_path_enable - test ACS flags from start to end in a hierarchy
2698 * @start: starting downstream device
2699 * @end: ending upstream device or NULL to search to the root bus
2700 * @acs_flags: required flags
2701 *
2702 * Walk up a device tree from start to end testing PCI ACS support.  If
2703 * any step along the way does not support the required flags, return false.
2704 */
2705bool pci_acs_path_enabled(struct pci_dev *start,
2706                          struct pci_dev *end, u16 acs_flags)
2707{
2708        struct pci_dev *pdev, *parent = start;
2709
2710        do {
2711                pdev = parent;
2712
2713                if (!pci_acs_enabled(pdev, acs_flags))
2714                        return false;
2715
2716                if (pci_is_root_bus(pdev->bus))
2717                        return (end == NULL);
2718
2719                parent = pdev->bus->self;
2720        } while (pdev != end);
2721
2722        return true;
2723}
2724
2725/**
2726 * pci_swizzle_interrupt_pin - swizzle INTx for device behind bridge
2727 * @dev: the PCI device
2728 * @pin: the INTx pin (1=INTA, 2=INTB, 3=INTC, 4=INTD)
2729 *
2730 * Perform INTx swizzling for a device behind one level of bridge.  This is
2731 * required by section 9.1 of the PCI-to-PCI bridge specification for devices
2732 * behind bridges on add-in cards.  For devices with ARI enabled, the slot
2733 * number is always 0 (see the Implementation Note in section 2.2.8.1 of
2734 * the PCI Express Base Specification, Revision 2.1)
2735 */
2736u8 pci_swizzle_interrupt_pin(const struct pci_dev *dev, u8 pin)
2737{
2738        int slot;
2739
2740        if (pci_ari_enabled(dev->bus))
2741                slot = 0;
2742        else
2743                slot = PCI_SLOT(dev->devfn);
2744
2745        return (((pin - 1) + slot) % 4) + 1;
2746}
2747
2748int pci_get_interrupt_pin(struct pci_dev *dev, struct pci_dev **bridge)
2749{
2750        u8 pin;
2751
2752        pin = dev->pin;
2753        if (!pin)
2754                return -1;
2755
2756        while (!pci_is_root_bus(dev->bus)) {
2757                pin = pci_swizzle_interrupt_pin(dev, pin);
2758                dev = dev->bus->self;
2759        }
2760        *bridge = dev;
2761        return pin;
2762}
2763
2764/**
2765 * pci_common_swizzle - swizzle INTx all the way to root bridge
2766 * @dev: the PCI device
2767 * @pinp: pointer to the INTx pin value (1=INTA, 2=INTB, 3=INTD, 4=INTD)
2768 *
2769 * Perform INTx swizzling for a device.  This traverses through all PCI-to-PCI
2770 * bridges all the way up to a PCI root bus.
2771 */
2772u8 pci_common_swizzle(struct pci_dev *dev, u8 *pinp)
2773{
2774        u8 pin = *pinp;
2775
2776        while (!pci_is_root_bus(dev->bus)) {
2777                pin = pci_swizzle_interrupt_pin(dev, pin);
2778                dev = dev->bus->self;
2779        }
2780        *pinp = pin;
2781        return PCI_SLOT(dev->devfn);
2782}
2783EXPORT_SYMBOL_GPL(pci_common_swizzle);
2784
2785/**
2786 *      pci_release_region - Release a PCI bar
2787 *      @pdev: PCI device whose resources were previously reserved by pci_request_region
2788 *      @bar: BAR to release
2789 *
2790 *      Releases the PCI I/O and memory resources previously reserved by a
2791 *      successful call to pci_request_region.  Call this function only
2792 *      after all use of the PCI regions has ceased.
2793 */
2794void pci_release_region(struct pci_dev *pdev, int bar)
2795{
2796        struct pci_devres *dr;
2797
2798        if (pci_resource_len(pdev, bar) == 0)
2799                return;
2800        if (pci_resource_flags(pdev, bar) & IORESOURCE_IO)
2801                release_region(pci_resource_start(pdev, bar),
2802                                pci_resource_len(pdev, bar));
2803        else if (pci_resource_flags(pdev, bar) & IORESOURCE_MEM)
2804                release_mem_region(pci_resource_start(pdev, bar),
2805                                pci_resource_len(pdev, bar));
2806
2807        dr = find_pci_dr(pdev);
2808        if (dr)
2809                dr->region_mask &= ~(1 << bar);
2810}
2811EXPORT_SYMBOL(pci_release_region);
2812
2813/**
2814 *      __pci_request_region - Reserved PCI I/O and memory resource
2815 *      @pdev: PCI device whose resources are to be reserved
2816 *      @bar: BAR to be reserved
2817 *      @res_name: Name to be associated with resource.
2818 *      @exclusive: whether the region access is exclusive or not
2819 *
2820 *      Mark the PCI region associated with PCI device @pdev BR @bar as
2821 *      being reserved by owner @res_name.  Do not access any
2822 *      address inside the PCI regions unless this call returns
2823 *      successfully.
2824 *
2825 *      If @exclusive is set, then the region is marked so that userspace
2826 *      is explicitly not allowed to map the resource via /dev/mem or
2827 *      sysfs MMIO access.
2828 *
2829 *      Returns 0 on success, or %EBUSY on error.  A warning
2830 *      message is also printed on failure.
2831 */
2832static int __pci_request_region(struct pci_dev *pdev, int bar,
2833                                const char *res_name, int exclusive)
2834{
2835        struct pci_devres *dr;
2836
2837        if (pci_resource_len(pdev, bar) == 0)
2838                return 0;
2839
2840        if (pci_resource_flags(pdev, bar) & IORESOURCE_IO) {
2841                if (!request_region(pci_resource_start(pdev, bar),
2842                            pci_resource_len(pdev, bar), res_name))
2843                        goto err_out;
2844        } else if (pci_resource_flags(pdev, bar) & IORESOURCE_MEM) {
2845                if (!__request_mem_region(pci_resource_start(pdev, bar),
2846                                        pci_resource_len(pdev, bar), res_name,
2847                                        exclusive))
2848                        goto err_out;
2849        }
2850
2851        dr = find_pci_dr(pdev);
2852        if (dr)
2853                dr->region_mask |= 1 << bar;
2854
2855        return 0;
2856
2857err_out:
2858        dev_warn(&pdev->dev, "BAR %d: can't reserve %pR\n", bar,
2859                 &pdev->resource[bar]);
2860        return -EBUSY;
2861}
2862
2863/**
2864 *      pci_request_region - Reserve PCI I/O and memory resource
2865 *      @pdev: PCI device whose resources are to be reserved
2866 *      @bar: BAR to be reserved
2867 *      @res_name: Name to be associated with resource
2868 *
2869 *      Mark the PCI region associated with PCI device @pdev BAR @bar as
2870 *      being reserved by owner @res_name.  Do not access any
2871 *      address inside the PCI regions unless this call returns
2872 *      successfully.
2873 *
2874 *      Returns 0 on success, or %EBUSY on error.  A warning
2875 *      message is also printed on failure.
2876 */
2877int pci_request_region(struct pci_dev *pdev, int bar, const char *res_name)
2878{
2879        return __pci_request_region(pdev, bar, res_name, 0);
2880}
2881EXPORT_SYMBOL(pci_request_region);
2882
2883/**
2884 *      pci_request_region_exclusive - Reserved PCI I/O and memory resource
2885 *      @pdev: PCI device whose resources are to be reserved
2886 *      @bar: BAR to be reserved
2887 *      @res_name: Name to be associated with resource.
2888 *
2889 *      Mark the PCI region associated with PCI device @pdev BR @bar as
2890 *      being reserved by owner @res_name.  Do not access any
2891 *      address inside the PCI regions unless this call returns
2892 *      successfully.
2893 *
2894 *      Returns 0 on success, or %EBUSY on error.  A warning
2895 *      message is also printed on failure.
2896 *
2897 *      The key difference that _exclusive makes it that userspace is
2898 *      explicitly not allowed to map the resource via /dev/mem or
2899 *      sysfs.
2900 */
2901int pci_request_region_exclusive(struct pci_dev *pdev, int bar,
2902                                 const char *res_name)
2903{
2904        return __pci_request_region(pdev, bar, res_name, IORESOURCE_EXCLUSIVE);
2905}
2906EXPORT_SYMBOL(pci_request_region_exclusive);
2907
2908/**
2909 * pci_release_selected_regions - Release selected PCI I/O and memory resources
2910 * @pdev: PCI device whose resources were previously reserved
2911 * @bars: Bitmask of BARs to be released
2912 *
2913 * Release selected PCI I/O and memory resources previously reserved.
2914 * Call this function only after all use of the PCI regions has ceased.
2915 */
2916void pci_release_selected_regions(struct pci_dev *pdev, int bars)
2917{
2918        int i;
2919
2920        for (i = 0; i < 6; i++)
2921                if (bars & (1 << i))
2922                        pci_release_region(pdev, i);
2923}
2924EXPORT_SYMBOL(pci_release_selected_regions);
2925
2926static int __pci_request_selected_regions(struct pci_dev *pdev, int bars,
2927                                          const char *res_name, int excl)
2928{
2929        int i;
2930
2931        for (i = 0; i < 6; i++)
2932                if (bars & (1 << i))
2933                        if (__pci_request_region(pdev, i, res_name, excl))
2934                                goto err_out;
2935        return 0;
2936
2937err_out:
2938        while (--i >= 0)
2939                if (bars & (1 << i))
2940                        pci_release_region(pdev, i);
2941
2942        return -EBUSY;
2943}
2944
2945
2946/**
2947 * pci_request_selected_regions - Reserve selected PCI I/O and memory resources
2948 * @pdev: PCI device whose resources are to be reserved
2949 * @bars: Bitmask of BARs to be requested
2950 * @res_name: Name to be associated with resource
2951 */
2952int pci_request_selected_regions(struct pci_dev *pdev, int bars,
2953                                 const char *res_name)
2954{
2955        return __pci_request_selected_regions(pdev, bars, res_name, 0);
2956}
2957EXPORT_SYMBOL(pci_request_selected_regions);
2958
2959int pci_request_selected_regions_exclusive(struct pci_dev *pdev, int bars,
2960                                           const char *res_name)
2961{
2962        return __pci_request_selected_regions(pdev, bars, res_name,
2963                        IORESOURCE_EXCLUSIVE);
2964}
2965EXPORT_SYMBOL(pci_request_selected_regions_exclusive);
2966
2967/**
2968 *      pci_release_regions - Release reserved PCI I/O and memory resources
2969 *      @pdev: PCI device whose resources were previously reserved by pci_request_regions
2970 *
2971 *      Releases all PCI I/O and memory resources previously reserved by a
2972 *      successful call to pci_request_regions.  Call this function only
2973 *      after all use of the PCI regions has ceased.
2974 */
2975
2976void pci_release_regions(struct pci_dev *pdev)
2977{
2978        pci_release_selected_regions(pdev, (1 << 6) - 1);
2979}
2980EXPORT_SYMBOL(pci_release_regions);
2981
2982/**
2983 *      pci_request_regions - Reserved PCI I/O and memory resources
2984 *      @pdev: PCI device whose resources are to be reserved
2985 *      @res_name: Name to be associated with resource.
2986 *
2987 *      Mark all PCI regions associated with PCI device @pdev as
2988 *      being reserved by owner @res_name.  Do not access any
2989 *      address inside the PCI regions unless this call returns
2990 *      successfully.
2991 *
2992 *      Returns 0 on success, or %EBUSY on error.  A warning
2993 *      message is also printed on failure.
2994 */
2995int pci_request_regions(struct pci_dev *pdev, const char *res_name)
2996{
2997        return pci_request_selected_regions(pdev, ((1 << 6) - 1), res_name);
2998}
2999EXPORT_SYMBOL(pci_request_regions);
3000
3001/**
3002 *      pci_request_regions_exclusive - Reserved PCI I/O and memory resources
3003 *      @pdev: PCI device whose resources are to be reserved
3004 *      @res_name: Name to be associated with resource.
3005 *
3006 *      Mark all PCI regions associated with PCI device @pdev as
3007 *      being reserved by owner @res_name.  Do not access any
3008 *      address inside the PCI regions unless this call returns
3009 *      successfully.
3010 *
3011 *      pci_request_regions_exclusive() will mark the region so that
3012 *      /dev/mem and the sysfs MMIO access will not be allowed.
3013 *
3014 *      Returns 0 on success, or %EBUSY on error.  A warning
3015 *      message is also printed on failure.
3016 */
3017int pci_request_regions_exclusive(struct pci_dev *pdev, const char *res_name)
3018{
3019        return pci_request_selected_regions_exclusive(pdev,
3020                                        ((1 << 6) - 1), res_name);
3021}
3022EXPORT_SYMBOL(pci_request_regions_exclusive);
3023
3024/**
3025 *      pci_remap_iospace - Remap the memory mapped I/O space
3026 *      @res: Resource describing the I/O space
3027 *      @phys_addr: physical address of range to be mapped
3028 *
3029 *      Remap the memory mapped I/O space described by the @res
3030 *      and the CPU physical address @phys_addr into virtual address space.
3031 *      Only architectures that have memory mapped IO functions defined
3032 *      (and the PCI_IOBASE value defined) should call this function.
3033 */
3034int __weak pci_remap_iospace(const struct resource *res, phys_addr_t phys_addr)
3035{
3036#if defined(PCI_IOBASE) && defined(CONFIG_MMU)
3037        unsigned long vaddr = (unsigned long)PCI_IOBASE + res->start;
3038
3039        if (!(res->flags & IORESOURCE_IO))
3040                return -EINVAL;
3041
3042        if (res->end > IO_SPACE_LIMIT)
3043                return -EINVAL;
3044
3045        return ioremap_page_range(vaddr, vaddr + resource_size(res), phys_addr,
3046                                  pgprot_device(PAGE_KERNEL));
3047#else
3048        /* this architecture does not have memory mapped I/O space,
3049           so this function should never be called */
3050        WARN_ONCE(1, "This architecture does not support memory mapped I/O\n");
3051        return -ENODEV;
3052#endif
3053}
3054
3055static void __pci_set_master(struct pci_dev *dev, bool enable)
3056{
3057        u16 old_cmd, cmd;
3058
3059        pci_read_config_word(dev, PCI_COMMAND, &old_cmd);
3060        if (enable)
3061                cmd = old_cmd | PCI_COMMAND_MASTER;
3062        else
3063                cmd = old_cmd & ~PCI_COMMAND_MASTER;
3064        if (cmd != old_cmd) {
3065                dev_dbg(&dev->dev, "%s bus mastering\n",
3066                        enable ? "enabling" : "disabling");
3067                pci_write_config_word(dev, PCI_COMMAND, cmd);
3068        }
3069        dev->is_busmaster = enable;
3070}
3071
3072/**
3073 * pcibios_setup - process "pci=" kernel boot arguments
3074 * @str: string used to pass in "pci=" kernel boot arguments
3075 *
3076 * Process kernel boot arguments.  This is the default implementation.
3077 * Architecture specific implementations can override this as necessary.
3078 */
3079char * __weak __init pcibios_setup(char *str)
3080{
3081        return str;
3082}
3083
3084/**
3085 * pcibios_set_master - enable PCI bus-mastering for device dev
3086 * @dev: the PCI device to enable
3087 *
3088 * Enables PCI bus-mastering for the device.  This is the default
3089 * implementation.  Architecture specific implementations can override
3090 * this if necessary.
3091 */
3092void __weak pcibios_set_master(struct pci_dev *dev)
3093{
3094        u8 lat;
3095
3096        /* The latency timer doesn't apply to PCIe (either Type 0 or Type 1) */
3097        if (pci_is_pcie(dev))
3098                return;
3099
3100        pci_read_config_byte(dev, PCI_LATENCY_TIMER, &lat);
3101        if (lat < 16)
3102                lat = (64 <= pcibios_max_latency) ? 64 : pcibios_max_latency;
3103        else if (lat > pcibios_max_latency)
3104                lat = pcibios_max_latency;
3105        else
3106                return;
3107
3108        pci_write_config_byte(dev, PCI_LATENCY_TIMER, lat);
3109}
3110
3111/**
3112 * pci_set_master - enables bus-mastering for device dev
3113 * @dev: the PCI device to enable
3114 *
3115 * Enables bus-mastering on the device and calls pcibios_set_master()
3116 * to do the needed arch specific settings.
3117 */
3118void pci_set_master(struct pci_dev *dev)
3119{
3120        __pci_set_master(dev, true);
3121        pcibios_set_master(dev);
3122}
3123EXPORT_SYMBOL(pci_set_master);
3124
3125/**
3126 * pci_clear_master - disables bus-mastering for device dev
3127 * @dev: the PCI device to disable
3128 */
3129void pci_clear_master(struct pci_dev *dev)
3130{
3131        __pci_set_master(dev, false);
3132}
3133EXPORT_SYMBOL(pci_clear_master);
3134
3135/**
3136 * pci_set_cacheline_size - ensure the CACHE_LINE_SIZE register is programmed
3137 * @dev: the PCI device for which MWI is to be enabled
3138 *
3139 * Helper function for pci_set_mwi.
3140 * Originally copied from drivers/net/acenic.c.
3141 * Copyright 1998-2001 by Jes Sorensen, <jes@trained-monkey.org>.
3142 *
3143 * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
3144 */
3145int pci_set_cacheline_size(struct pci_dev *dev)
3146{
3147        u8 cacheline_size;
3148
3149        if (!pci_cache_line_size)
3150                return -EINVAL;
3151
3152        /* Validate current setting: the PCI_CACHE_LINE_SIZE must be
3153           equal to or multiple of the right value. */
3154        pci_read_config_byte(dev, PCI_CACHE_LINE_SIZE, &cacheline_size);
3155        if (cacheline_size >= pci_cache_line_size &&
3156            (cacheline_size % pci_cache_line_size) == 0)
3157                return 0;
3158
3159        /* Write the correct value. */
3160        pci_write_config_byte(dev, PCI_CACHE_LINE_SIZE, pci_cache_line_size);
3161        /* Read it back. */
3162        pci_read_config_byte(dev, PCI_CACHE_LINE_SIZE, &cacheline_size);
3163        if (cacheline_size == pci_cache_line_size)
3164                return 0;
3165
3166        dev_printk(KERN_DEBUG, &dev->dev, "cache line size of %d is not supported\n",
3167                   pci_cache_line_size << 2);
3168
3169        return -EINVAL;
3170}
3171EXPORT_SYMBOL_GPL(pci_set_cacheline_size);
3172
3173/**
3174 * pci_set_mwi - enables memory-write-invalidate PCI transaction
3175 * @dev: the PCI device for which MWI is enabled
3176 *
3177 * Enables the Memory-Write-Invalidate transaction in %PCI_COMMAND.
3178 *
3179 * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
3180 */
3181int pci_set_mwi(struct pci_dev *dev)
3182{
3183#ifdef PCI_DISABLE_MWI
3184        return 0;
3185#else
3186        int rc;
3187        u16 cmd;
3188
3189        rc = pci_set_cacheline_size(dev);
3190        if (rc)
3191                return rc;
3192
3193        pci_read_config_word(dev, PCI_COMMAND, &cmd);
3194        if (!(cmd & PCI_COMMAND_INVALIDATE)) {
3195                dev_dbg(&dev->dev, "enabling Mem-Wr-Inval\n");
3196                cmd |= PCI_COMMAND_INVALIDATE;
3197                pci_write_config_word(dev, PCI_COMMAND, cmd);
3198        }
3199        return 0;
3200#endif
3201}
3202EXPORT_SYMBOL(pci_set_mwi);
3203
3204/**
3205 * pci_try_set_mwi - enables memory-write-invalidate PCI transaction
3206 * @dev: the PCI device for which MWI is enabled
3207 *
3208 * Enables the Memory-Write-Invalidate transaction in %PCI_COMMAND.
3209 * Callers are not required to check the return value.
3210 *
3211 * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
3212 */
3213int pci_try_set_mwi(struct pci_dev *dev)
3214{
3215#ifdef PCI_DISABLE_MWI
3216        return 0;
3217#else
3218        return pci_set_mwi(dev);
3219#endif
3220}
3221EXPORT_SYMBOL(pci_try_set_mwi);
3222
3223/**
3224 * pci_clear_mwi - disables Memory-Write-Invalidate for device dev
3225 * @dev: the PCI device to disable
3226 *
3227 * Disables PCI Memory-Write-Invalidate transaction on the device
3228 */
3229void pci_clear_mwi(struct pci_dev *dev)
3230{
3231#ifndef PCI_DISABLE_MWI
3232        u16 cmd;
3233
3234        pci_read_config_word(dev, PCI_COMMAND, &cmd);
3235        if (cmd & PCI_COMMAND_INVALIDATE) {
3236                cmd &= ~PCI_COMMAND_INVALIDATE;
3237                pci_write_config_word(dev, PCI_COMMAND, cmd);
3238        }
3239#endif
3240}
3241EXPORT_SYMBOL(pci_clear_mwi);
3242
3243/**
3244 * pci_intx - enables/disables PCI INTx for device dev
3245 * @pdev: the PCI device to operate on
3246 * @enable: boolean: whether to enable or disable PCI INTx
3247 *
3248 * Enables/disables PCI INTx for device dev
3249 */
3250void pci_intx(struct pci_dev *pdev, int enable)
3251{
3252        u16 pci_command, new;
3253
3254        pci_read_config_word(pdev, PCI_COMMAND, &pci_command);
3255
3256        if (enable)
3257                new = pci_command & ~PCI_COMMAND_INTX_DISABLE;
3258        else
3259                new = pci_command | PCI_COMMAND_INTX_DISABLE;
3260
3261        if (new != pci_command) {
3262                struct pci_devres *dr;
3263
3264                pci_write_config_word(pdev, PCI_COMMAND, new);
3265
3266                dr = find_pci_dr(pdev);
3267                if (dr && !dr->restore_intx) {
3268                        dr->restore_intx = 1;
3269                        dr->orig_intx = !enable;
3270                }
3271        }
3272}
3273EXPORT_SYMBOL_GPL(pci_intx);
3274
3275/**
3276 * pci_intx_mask_supported - probe for INTx masking support
3277 * @dev: the PCI device to operate on
3278 *
3279 * Check if the device dev support INTx masking via the config space
3280 * command word.
3281 */
3282bool pci_intx_mask_supported(struct pci_dev *dev)
3283{
3284        bool mask_supported = false;
3285        u16 orig, new;
3286
3287        if (dev->broken_intx_masking)
3288                return false;
3289
3290        pci_cfg_access_lock(dev);
3291
3292        pci_read_config_word(dev, PCI_COMMAND, &orig);
3293        pci_write_config_word(dev, PCI_COMMAND,
3294                              orig ^ PCI_COMMAND_INTX_DISABLE);
3295        pci_read_config_word(dev, PCI_COMMAND, &new);
3296
3297        /*
3298         * There's no way to protect against hardware bugs or detect them
3299         * reliably, but as long as we know what the value should be, let's
3300         * go ahead and check it.
3301         */
3302        if ((new ^ orig) & ~PCI_COMMAND_INTX_DISABLE) {
3303                dev_err(&dev->dev, "Command register changed from 0x%x to 0x%x: driver or hardware bug?\n",
3304                        orig, new);
3305        } else if ((new ^ orig) & PCI_COMMAND_INTX_DISABLE) {
3306                mask_supported = true;
3307                pci_write_config_word(dev, PCI_COMMAND, orig);
3308        }
3309
3310        pci_cfg_access_unlock(dev);
3311        return mask_supported;
3312}
3313EXPORT_SYMBOL_GPL(pci_intx_mask_supported);
3314
3315static bool pci_check_and_set_intx_mask(struct pci_dev *dev, bool mask)
3316{
3317        struct pci_bus *bus = dev->bus;
3318        bool mask_updated = true;
3319        u32 cmd_status_dword;
3320        u16 origcmd, newcmd;
3321        unsigned long flags;
3322        bool irq_pending;
3323
3324        /*
3325         * We do a single dword read to retrieve both command and status.
3326         * Document assumptions that make this possible.
3327         */
3328        BUILD_BUG_ON(PCI_COMMAND % 4);
3329        BUILD_BUG_ON(PCI_COMMAND + 2 != PCI_STATUS);
3330
3331        raw_spin_lock_irqsave(&pci_lock, flags);
3332
3333        bus->ops->read(bus, dev->devfn, PCI_COMMAND, 4, &cmd_status_dword);
3334
3335        irq_pending = (cmd_status_dword >> 16) & PCI_STATUS_INTERRUPT;
3336
3337        /*
3338         * Check interrupt status register to see whether our device
3339         * triggered the interrupt (when masking) or the next IRQ is
3340         * already pending (when unmasking).
3341         */
3342        if (mask != irq_pending) {
3343                mask_updated = false;
3344                goto done;
3345        }
3346
3347        origcmd = cmd_status_dword;
3348        newcmd = origcmd & ~PCI_COMMAND_INTX_DISABLE;
3349        if (mask)
3350                newcmd |= PCI_COMMAND_INTX_DISABLE;
3351        if (newcmd != origcmd)
3352                bus->ops->write(bus, dev->devfn, PCI_COMMAND, 2, newcmd);
3353
3354done:
3355        raw_spin_unlock_irqrestore(&pci_lock, flags);
3356
3357        return mask_updated;
3358}
3359
3360/**
3361 * pci_check_and_mask_intx - mask INTx on pending interrupt
3362 * @dev: the PCI device to operate on
3363 *
3364 * Check if the device dev has its INTx line asserted, mask it and
3365 * return true in that case. False is returned if not interrupt was
3366 * pending.
3367 */
3368bool pci_check_and_mask_intx(struct pci_dev *dev)
3369{
3370        return pci_check_and_set_intx_mask(dev, true);
3371}
3372EXPORT_SYMBOL_GPL(pci_check_and_mask_intx);
3373
3374/**
3375 * pci_check_and_unmask_intx - unmask INTx if no interrupt is pending
3376 * @dev: the PCI device to operate on
3377 *
3378 * Check if the device dev has its INTx line asserted, unmask it if not
3379 * and return true. False is returned and the mask remains active if
3380 * there was still an interrupt pending.
3381 */
3382bool pci_check_and_unmask_intx(struct pci_dev *dev)
3383{
3384        return pci_check_and_set_intx_mask(dev, false);
3385}
3386EXPORT_SYMBOL_GPL(pci_check_and_unmask_intx);
3387
3388/**
3389 * pci_wait_for_pending_transaction - waits for pending transaction
3390 * @dev: the PCI device to operate on
3391 *
3392 * Return 0 if transaction is pending 1 otherwise.
3393 */
3394int pci_wait_for_pending_transaction(struct pci_dev *dev)
3395{
3396        if (!pci_is_pcie(dev))
3397                return 1;
3398
3399        return pci_wait_for_pending(dev, pci_pcie_cap(dev) + PCI_EXP_DEVSTA,
3400                                    PCI_EXP_DEVSTA_TRPND);
3401}
3402EXPORT_SYMBOL(pci_wait_for_pending_transaction);
3403
3404/*
3405 * We should only need to wait 100ms after FLR, but some devices take longer.
3406 * Wait for up to 1000ms for config space to return something other than -1.
3407 * Intel IGD requires this when an LCD panel is attached.  We read the 2nd
3408 * dword because VFs don't implement the 1st dword.
3409 */
3410static void pci_flr_wait(struct pci_dev *dev)
3411{
3412        int i = 0;
3413        u32 id;
3414
3415        do {
3416                msleep(100);
3417                pci_read_config_dword(dev, PCI_COMMAND, &id);
3418        } while (i++ < 10 && id == ~0);
3419
3420        if (id == ~0)
3421                dev_warn(&dev->dev, "Failed to return from FLR\n");
3422        else if (i > 1)
3423                dev_info(&dev->dev, "Required additional %dms to return from FLR\n",
3424                         (i - 1) * 100);
3425}
3426
3427static int pcie_flr(struct pci_dev *dev, int probe)
3428{
3429        u32 cap;
3430
3431        pcie_capability_read_dword(dev, PCI_EXP_DEVCAP, &cap);
3432        if (!(cap & PCI_EXP_DEVCAP_FLR))
3433                return -ENOTTY;
3434
3435        if (probe)
3436                return 0;
3437
3438        if (!pci_wait_for_pending_transaction(dev))
3439                dev_err(&dev->dev, "timed out waiting for pending transaction; performing function level reset anyway\n");
3440
3441        pcie_capability_set_word(dev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
3442        pci_flr_wait(dev);
3443        return 0;
3444}
3445
3446static int pci_af_flr(struct pci_dev *dev, int probe)
3447{
3448        int pos;
3449        u8 cap;
3450
3451        pos = pci_find_capability(dev, PCI_CAP_ID_AF);
3452        if (!pos)
3453                return -ENOTTY;
3454
3455        pci_read_config_byte(dev, pos + PCI_AF_CAP, &cap);
3456        if (!(cap & PCI_AF_CAP_TP) || !(cap & PCI_AF_CAP_FLR))
3457                return -ENOTTY;
3458
3459        if (probe)
3460                return 0;
3461
3462        /*
3463         * Wait for Transaction Pending bit to clear.  A word-aligned test
3464         * is used, so we use the conrol offset rather than status and shift
3465         * the test bit to match.
3466         */
3467        if (!pci_wait_for_pending(dev, pos + PCI_AF_CTRL,
3468                                 PCI_AF_STATUS_TP << 8))
3469                dev_err(&dev->dev, "timed out waiting for pending transaction; performing AF function level reset anyway\n");
3470
3471        pci_write_config_byte(dev, pos + PCI_AF_CTRL, PCI_AF_CTRL_FLR);
3472        pci_flr_wait(dev);
3473        return 0;
3474}
3475
3476/**
3477 * pci_pm_reset - Put device into PCI_D3 and back into PCI_D0.
3478 * @dev: Device to reset.
3479 * @probe: If set, only check if the device can be reset this way.
3480 *
3481 * If @dev supports native PCI PM and its PCI_PM_CTRL_NO_SOFT_RESET flag is
3482 * unset, it will be reinitialized internally when going from PCI_D3hot to
3483 * PCI_D0.  If that's the case and the device is not in a low-power state
3484 * already, force it into PCI_D3hot and back to PCI_D0, causing it to be reset.
3485 *
3486 * NOTE: This causes the caller to sleep for twice the device power transition
3487 * cooldown period, which for the D0->D3hot and D3hot->D0 transitions is 10 ms
3488 * by default (i.e. unless the @dev's d3_delay field has a different value).
3489 * Moreover, only devices in D0 can be reset by this function.
3490 */
3491static int pci_pm_reset(struct pci_dev *dev, int probe)
3492{
3493        u16 csr;
3494
3495        if (!dev->pm_cap || dev->dev_flags & PCI_DEV_FLAGS_NO_PM_RESET)
3496                return -ENOTTY;
3497
3498        pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &csr);
3499        if (csr & PCI_PM_CTRL_NO_SOFT_RESET)
3500                return -ENOTTY;
3501
3502        if (probe)
3503                return 0;
3504
3505        if (dev->current_state != PCI_D0)
3506                return -EINVAL;
3507
3508        csr &= ~PCI_PM_CTRL_STATE_MASK;
3509        csr |= PCI_D3hot;
3510        pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, csr);
3511        pci_dev_d3_sleep(dev);
3512
3513        csr &= ~PCI_PM_CTRL_STATE_MASK;
3514        csr |= PCI_D0;
3515        pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, csr);
3516        pci_dev_d3_sleep(dev);
3517
3518        return 0;
3519}
3520
3521void pci_reset_secondary_bus(struct pci_dev *dev)
3522{
3523        u16 ctrl;
3524
3525        pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &ctrl);
3526        ctrl |= PCI_BRIDGE_CTL_BUS_RESET;
3527        pci_write_config_word(dev, PCI_BRIDGE_CONTROL, ctrl);
3528        /*
3529         * PCI spec v3.0 7.6.4.2 requires minimum Trst of 1ms.  Double
3530         * this to 2ms to ensure that we meet the minimum requirement.
3531         */
3532        msleep(2);
3533
3534        ctrl &= ~PCI_BRIDGE_CTL_BUS_RESET;
3535        pci_write_config_word(dev, PCI_BRIDGE_CONTROL, ctrl);
3536
3537        /*
3538         * Trhfa for conventional PCI is 2^25 clock cycles.
3539         * Assuming a minimum 33MHz clock this results in a 1s
3540         * delay before we can consider subordinate devices to
3541         * be re-initialized.  PCIe has some ways to shorten this,
3542         * but we don't make use of them yet.
3543         */
3544        ssleep(1);
3545}
3546
3547void __weak pcibios_reset_secondary_bus(struct pci_dev *dev)
3548{
3549        pci_reset_secondary_bus(dev);
3550}
3551
3552/**
3553 * pci_reset_bridge_secondary_bus - Reset the secondary bus on a PCI bridge.
3554 * @dev: Bridge device
3555 *
3556 * Use the bridge control register to assert reset on the secondary bus.
3557 * Devices on the secondary bus are left in power-on state.
3558 */
3559void pci_reset_bridge_secondary_bus(struct pci_dev *dev)
3560{
3561        pcibios_reset_secondary_bus(dev);
3562}
3563EXPORT_SYMBOL_GPL(pci_reset_bridge_secondary_bus);
3564
3565static int pci_parent_bus_reset(struct pci_dev *dev, int probe)
3566{
3567        struct pci_dev *pdev;
3568
3569        if (pci_is_root_bus(dev->bus) || dev->subordinate ||
3570            !dev->bus->self || dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET)
3571                return -ENOTTY;
3572
3573        list_for_each_entry(pdev, &dev->bus->devices, bus_list)
3574                if (pdev != dev)
3575                        return -ENOTTY;
3576
3577        if (probe)
3578                return 0;
3579
3580        pci_reset_bridge_secondary_bus(dev->bus->self);
3581
3582        return 0;
3583}
3584
3585static int pci_reset_hotplug_slot(struct hotplug_slot *hotplug, int probe)
3586{
3587        int rc = -ENOTTY;
3588
3589        if (!hotplug || !try_module_get(hotplug->ops->owner))
3590                return rc;
3591
3592        if (hotplug->ops->reset_slot)
3593                rc = hotplug->ops->reset_slot(hotplug, probe);
3594
3595        module_put(hotplug->ops->owner);
3596
3597        return rc;
3598}
3599
3600static int pci_dev_reset_slot_function(struct pci_dev *dev, int probe)
3601{
3602        struct pci_dev *pdev;
3603
3604        if (dev->subordinate || !dev->slot ||
3605            dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET)
3606                return -ENOTTY;
3607
3608        list_for_each_entry(pdev, &dev->bus->devices, bus_list)
3609                if (pdev != dev && pdev->slot == dev->slot)
3610                        return -ENOTTY;
3611
3612        return pci_reset_hotplug_slot(dev->slot->hotplug, probe);
3613}
3614
3615static int __pci_dev_reset(struct pci_dev *dev, int probe)
3616{
3617        int rc;
3618
3619        might_sleep();
3620
3621        rc = pci_dev_specific_reset(dev, probe);
3622        if (rc != -ENOTTY)
3623                goto done;
3624
3625        rc = pcie_flr(dev, probe);
3626        if (rc != -ENOTTY)
3627                goto done;
3628
3629        rc = pci_af_flr(dev, probe);
3630        if (rc != -ENOTTY)
3631                goto done;
3632
3633        rc = pci_pm_reset(dev, probe);
3634        if (rc != -ENOTTY)
3635                goto done;
3636
3637        rc = pci_dev_reset_slot_function(dev, probe);
3638        if (rc != -ENOTTY)
3639                goto done;
3640
3641        rc = pci_parent_bus_reset(dev, probe);
3642done:
3643        return rc;
3644}
3645
3646static void pci_dev_lock(struct pci_dev *dev)
3647{
3648        pci_cfg_access_lock(dev);
3649        /* block PM suspend, driver probe, etc. */
3650        device_lock(&dev->dev);
3651}
3652
3653/* Return 1 on successful lock, 0 on contention */
3654static int pci_dev_trylock(struct pci_dev *dev)
3655{
3656        if (pci_cfg_access_trylock(dev)) {
3657                if (device_trylock(&dev->dev))
3658                        return 1;
3659                pci_cfg_access_unlock(dev);
3660        }
3661
3662        return 0;
3663}
3664
3665static void pci_dev_unlock(struct pci_dev *dev)
3666{
3667        device_unlock(&dev->dev);
3668        pci_cfg_access_unlock(dev);
3669}
3670
3671/**
3672 * pci_reset_notify - notify device driver of reset
3673 * @dev: device to be notified of reset
3674 * @prepare: 'true' if device is about to be reset; 'false' if reset attempt
3675 *           completed
3676 *
3677 * Must be called prior to device access being disabled and after device
3678 * access is restored.
3679 */
3680static void pci_reset_notify(struct pci_dev *dev, bool prepare)
3681{
3682        const struct pci_error_handlers *err_handler =
3683                        dev->driver ? dev->driver->err_handler : NULL;
3684        if (err_handler && err_handler->reset_notify)
3685                err_handler->reset_notify(dev, prepare);
3686}
3687
3688static void pci_dev_save_and_disable(struct pci_dev *dev)
3689{
3690        pci_reset_notify(dev, true);
3691
3692        /*
3693         * Wake-up device prior to save.  PM registers default to D0 after
3694         * reset and a simple register restore doesn't reliably return
3695         * to a non-D0 state anyway.
3696         */
3697        pci_set_power_state(dev, PCI_D0);
3698
3699        pci_save_state(dev);
3700        /*
3701         * Disable the device by clearing the Command register, except for
3702         * INTx-disable which is set.  This not only disables MMIO and I/O port
3703         * BARs, but also prevents the device from being Bus Master, preventing
3704         * DMA from the device including MSI/MSI-X interrupts.  For PCI 2.3
3705         * compliant devices, INTx-disable prevents legacy interrupts.
3706         */
3707        pci_write_config_word(dev, PCI_COMMAND, PCI_COMMAND_INTX_DISABLE);
3708}
3709
3710static void pci_dev_restore(struct pci_dev *dev)
3711{
3712        pci_restore_state(dev);
3713        pci_reset_notify(dev, false);
3714}
3715
3716static int pci_dev_reset(struct pci_dev *dev, int probe)
3717{
3718        int rc;
3719
3720        if (!probe)
3721                pci_dev_lock(dev);
3722
3723        rc = __pci_dev_reset(dev, probe);
3724
3725        if (!probe)
3726                pci_dev_unlock(dev);
3727
3728        return rc;
3729}
3730
3731/**
3732 * __pci_reset_function - reset a PCI device function
3733 * @dev: PCI device to reset
3734 *
3735 * Some devices allow an individual function to be reset without affecting
3736 * other functions in the same device.  The PCI device must be responsive
3737 * to PCI config space in order to use this function.
3738 *
3739 * The device function is presumed to be unused when this function is called.
3740 * Resetting the device will make the contents of PCI configuration space
3741 * random, so any caller of this must be prepared to reinitialise the
3742 * device including MSI, bus mastering, BARs, decoding IO and memory spaces,
3743 * etc.
3744 *
3745 * Returns 0 if the device function was successfully reset or negative if the
3746 * device doesn't support resetting a single function.
3747 */
3748int __pci_reset_function(struct pci_dev *dev)
3749{
3750        return pci_dev_reset(dev, 0);
3751}
3752EXPORT_SYMBOL_GPL(__pci_reset_function);
3753
3754/**
3755 * __pci_reset_function_locked - reset a PCI device function while holding
3756 * the @dev mutex lock.
3757 * @dev: PCI device to reset
3758 *
3759 * Some devices allow an individual function to be reset without affecting
3760 * other functions in the same device.  The PCI device must be responsive
3761 * to PCI config space in order to use this function.
3762 *
3763 * The device function is presumed to be unused and the caller is holding
3764 * the device mutex lock when this function is called.
3765 * Resetting the device will make the contents of PCI configuration space
3766 * random, so any caller of this must be prepared to reinitialise the
3767 * device including MSI, bus mastering, BARs, decoding IO and memory spaces,
3768 * etc.
3769 *
3770 * Returns 0 if the device function was successfully reset or negative if the
3771 * device doesn't support resetting a single function.
3772 */
3773int __pci_reset_function_locked(struct pci_dev *dev)
3774{
3775        return __pci_dev_reset(dev, 0);
3776}
3777EXPORT_SYMBOL_GPL(__pci_reset_function_locked);
3778
3779/**
3780 * pci_probe_reset_function - check whether the device can be safely reset
3781 * @dev: PCI device to reset
3782 *
3783 * Some devices allow an individual function to be reset without affecting
3784 * other functions in the same device.  The PCI device must be responsive
3785 * to PCI config space in order to use this function.
3786 *
3787 * Returns 0 if the device function can be reset or negative if the
3788 * device doesn't support resetting a single function.
3789 */
3790int pci_probe_reset_function(struct pci_dev *dev)
3791{
3792        return pci_dev_reset(dev, 1);
3793}
3794
3795/**
3796 * pci_reset_function - quiesce and reset a PCI device function
3797 * @dev: PCI device to reset
3798 *
3799 * Some devices allow an individual function to be reset without affecting
3800 * other functions in the same device.  The PCI device must be responsive
3801 * to PCI config space in order to use this function.
3802 *
3803 * This function does not just reset the PCI portion of a device, but
3804 * clears all the state associated with the device.  This function differs
3805 * from __pci_reset_function in that it saves and restores device state
3806 * over the reset.
3807 *
3808 * Returns 0 if the device function was successfully reset or negative if the
3809 * device doesn't support resetting a single function.
3810 */
3811int pci_reset_function(struct pci_dev *dev)
3812{
3813        int rc;
3814
3815        rc = pci_dev_reset(dev, 1);
3816        if (rc)
3817                return rc;
3818
3819        pci_dev_save_and_disable(dev);
3820
3821        rc = pci_dev_reset(dev, 0);
3822
3823        pci_dev_restore(dev);
3824
3825        return rc;
3826}
3827EXPORT_SYMBOL_GPL(pci_reset_function);
3828
3829/**
3830 * pci_try_reset_function - quiesce and reset a PCI device function
3831 * @dev: PCI device to reset
3832 *
3833 * Same as above, except return -EAGAIN if unable to lock device.
3834 */
3835int pci_try_reset_function(struct pci_dev *dev)
3836{
3837        int rc;
3838
3839        rc = pci_dev_reset(dev, 1);
3840        if (rc)
3841                return rc;
3842
3843        pci_dev_save_and_disable(dev);
3844
3845        if (pci_dev_trylock(dev)) {
3846                rc = __pci_dev_reset(dev, 0);
3847                pci_dev_unlock(dev);
3848        } else
3849                rc = -EAGAIN;
3850
3851        pci_dev_restore(dev);
3852
3853        return rc;
3854}
3855EXPORT_SYMBOL_GPL(pci_try_reset_function);
3856
3857/* Do any devices on or below this bus prevent a bus reset? */
3858static bool pci_bus_resetable(struct pci_bus *bus)
3859{
3860        struct pci_dev *dev;
3861
3862        list_for_each_entry(dev, &bus->devices, bus_list) {
3863                if (dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET ||
3864                    (dev->subordinate && !pci_bus_resetable(dev->subordinate)))
3865                        return false;
3866        }
3867
3868        return true;
3869}
3870
3871/* Lock devices from the top of the tree down */
3872static void pci_bus_lock(struct pci_bus *bus)
3873{
3874        struct pci_dev *dev;
3875
3876        list_for_each_entry(dev, &bus->devices, bus_list) {
3877                pci_dev_lock(dev);
3878                if (dev->subordinate)
3879                        pci_bus_lock(dev->subordinate);
3880        }
3881}
3882
3883/* Unlock devices from the bottom of the tree up */
3884static void pci_bus_unlock(struct pci_bus *bus)
3885{
3886        struct pci_dev *dev;
3887
3888        list_for_each_entry(dev, &bus->devices, bus_list) {
3889                if (dev->subordinate)
3890                        pci_bus_unlock(dev->subordinate);
3891                pci_dev_unlock(dev);
3892        }
3893}
3894
3895/* Return 1 on successful lock, 0 on contention */
3896static int pci_bus_trylock(struct pci_bus *bus)
3897{
3898        struct pci_dev *dev;
3899
3900        list_for_each_entry(dev, &bus->devices, bus_list) {
3901                if (!pci_dev_trylock(dev))
3902                        goto unlock;
3903                if (dev->subordinate) {
3904                        if (!pci_bus_trylock(dev->subordinate)) {
3905                                pci_dev_unlock(dev);
3906                                goto unlock;
3907                        }
3908                }
3909        }
3910        return 1;
3911
3912unlock:
3913        list_for_each_entry_continue_reverse(dev, &bus->devices, bus_list) {
3914                if (dev->subordinate)
3915                        pci_bus_unlock(dev->subordinate);
3916                pci_dev_unlock(dev);
3917        }
3918        return 0;
3919}
3920
3921/* Do any devices on or below this slot prevent a bus reset? */
3922static bool pci_slot_resetable(struct pci_slot *slot)
3923{
3924        struct pci_dev *dev;
3925
3926        list_for_each_entry(dev, &slot->bus->devices, bus_list) {
3927                if (!dev->slot || dev->slot != slot)
3928                        continue;
3929                if (dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET ||
3930                    (dev->subordinate && !pci_bus_resetable(dev->subordinate)))
3931                        return false;
3932        }
3933
3934        return true;
3935}
3936
3937/* Lock devices from the top of the tree down */
3938static void pci_slot_lock(struct pci_slot *slot)
3939{
3940        struct pci_dev *dev;
3941
3942        list_for_each_entry(dev, &slot->bus->devices, bus_list) {
3943                if (!dev->slot || dev->slot != slot)
3944                        continue;
3945                pci_dev_lock(dev);
3946                if (dev->subordinate)
3947                        pci_bus_lock(dev->subordinate);
3948        }
3949}
3950
3951/* Unlock devices from the bottom of the tree up */
3952static void pci_slot_unlock(struct pci_slot *slot)
3953{
3954        struct pci_dev *dev;
3955
3956        list_for_each_entry(dev, &slot->bus->devices, bus_list) {
3957                if (!dev->slot || dev->slot != slot)
3958                        continue;
3959                if (dev->subordinate)
3960                        pci_bus_unlock(dev->subordinate);
3961                pci_dev_unlock(dev);
3962        }
3963}
3964
3965/* Return 1 on successful lock, 0 on contention */
3966static int pci_slot_trylock(struct pci_slot *slot)
3967{
3968        struct pci_dev *dev;
3969
3970        list_for_each_entry(dev, &slot->bus->devices, bus_list) {
3971                if (!dev->slot || dev->slot != slot)
3972                        continue;
3973                if (!pci_dev_trylock(dev))
3974                        goto unlock;
3975                if (dev->subordinate) {
3976                        if (!pci_bus_trylock(dev->subordinate)) {
3977                                pci_dev_unlock(dev);
3978                                goto unlock;
3979                        }
3980                }
3981        }
3982        return 1;
3983
3984unlock:
3985        list_for_each_entry_continue_reverse(dev,
3986                                             &slot->bus->devices, bus_list) {
3987                if (!dev->slot || dev->slot != slot)
3988                        continue;
3989                if (dev->subordinate)
3990                        pci_bus_unlock(dev->subordinate);
3991                pci_dev_unlock(dev);
3992        }
3993        return 0;
3994}
3995
3996/* Save and disable devices from the top of the tree down */
3997static void pci_bus_save_and_disable(struct pci_bus *bus)
3998{
3999        struct pci_dev *dev;
4000
4001        list_for_each_entry(dev, &bus->devices, bus_list) {
4002                pci_dev_save_and_disable(dev);
4003                if (dev->subordinate)
4004                        pci_bus_save_and_disable(dev->subordinate);
4005        }
4006}
4007
4008/*
4009 * Restore devices from top of the tree down - parent bridges need to be
4010 * restored before we can get to subordinate devices.
4011 */
4012static void pci_bus_restore(struct pci_bus *bus)
4013{
4014        struct pci_dev *dev;
4015
4016        list_for_each_entry(dev, &bus->devices, bus_list) {
4017                pci_dev_restore(dev);
4018                if (dev->subordinate)
4019                        pci_bus_restore(dev->subordinate);
4020        }
4021}
4022
4023/* Save and disable devices from the top of the tree down */
4024static void pci_slot_save_and_disable(struct pci_slot *slot)
4025{
4026        struct pci_dev *dev;
4027
4028        list_for_each_entry(dev, &slot->bus->devices, bus_list) {
4029                if (!dev->slot || dev->slot != slot)
4030                        continue;
4031                pci_dev_save_and_disable(dev);
4032                if (dev->subordinate)
4033                        pci_bus_save_and_disable(dev->subordinate);
4034        }
4035}
4036
4037/*
4038 * Restore devices from top of the tree down - parent bridges need to be
4039 * restored before we can get to subordinate devices.
4040 */
4041static void pci_slot_restore(struct pci_slot *slot)
4042{
4043        struct pci_dev *dev;
4044
4045        list_for_each_entry(dev, &slot->bus->devices, bus_list) {
4046                if (!dev->slot || dev->slot != slot)
4047                        continue;
4048                pci_dev_restore(dev);
4049                if (dev->subordinate)
4050                        pci_bus_restore(dev->subordinate);
4051        }
4052}
4053
4054static int pci_slot_reset(struct pci_slot *slot, int probe)
4055{
4056        int rc;
4057
4058        if (!slot || !pci_slot_resetable(slot))
4059                return -ENOTTY;
4060
4061        if (!probe)
4062                pci_slot_lock(slot);
4063
4064        might_sleep();
4065
4066        rc = pci_reset_hotplug_slot(slot->hotplug, probe);
4067
4068        if (!probe)
4069                pci_slot_unlock(slot);
4070
4071        return rc;
4072}
4073
4074/**
4075 * pci_probe_reset_slot - probe whether a PCI slot can be reset
4076 * @slot: PCI slot to probe
4077 *
4078 * Return 0 if slot can be reset, negative if a slot reset is not supported.
4079 */
4080int pci_probe_reset_slot(struct pci_slot *slot)
4081{
4082        return pci_slot_reset(slot, 1);
4083}
4084EXPORT_SYMBOL_GPL(pci_probe_reset_slot);
4085
4086/**
4087 * pci_reset_slot - reset a PCI slot
4088 * @slot: PCI slot to reset
4089 *
4090 * A PCI bus may host multiple slots, each slot may support a reset mechanism
4091 * independent of other slots.  For instance, some slots may support slot power
4092 * control.  In the case of a 1:1 bus to slot architecture, this function may
4093 * wrap the bus reset to avoid spurious slot related events such as hotplug.
4094 * Generally a slot reset should be attempted before a bus reset.  All of the
4095 * function of the slot and any subordinate buses behind the slot are reset
4096 * through this function.  PCI config space of all devices in the slot and
4097 * behind the slot is saved before and restored after reset.
4098 *
4099 * Return 0 on success, non-zero on error.
4100 */
4101int pci_reset_slot(struct pci_slot *slot)
4102{
4103        int rc;
4104
4105        rc = pci_slot_reset(slot, 1);
4106        if (rc)
4107                return rc;
4108
4109        pci_slot_save_and_disable(slot);
4110
4111        rc = pci_slot_reset(slot, 0);
4112
4113        pci_slot_restore(slot);
4114
4115        return rc;
4116}
4117EXPORT_SYMBOL_GPL(pci_reset_slot);
4118
4119/**
4120 * pci_try_reset_slot - Try to reset a PCI slot
4121 * @slot: PCI slot to reset
4122 *
4123 * Same as above except return -EAGAIN if the slot cannot be locked
4124 */
4125int pci_try_reset_slot(struct pci_slot *slot)
4126{
4127        int rc;
4128
4129        rc = pci_slot_reset(slot, 1);
4130        if (rc)
4131                return rc;
4132
4133        pci_slot_save_and_disable(slot);
4134
4135        if (pci_slot_trylock(slot)) {
4136                might_sleep();
4137                rc = pci_reset_hotplug_slot(slot->hotplug, 0);
4138                pci_slot_unlock(slot);
4139        } else
4140                rc = -EAGAIN;
4141
4142        pci_slot_restore(slot);
4143
4144        return rc;
4145}
4146EXPORT_SYMBOL_GPL(pci_try_reset_slot);
4147
4148static int pci_bus_reset(struct pci_bus *bus, int probe)
4149{
4150        if (!bus->self || !pci_bus_resetable(bus))
4151                return -ENOTTY;
4152
4153        if (probe)
4154                return 0;
4155
4156        pci_bus_lock(bus);
4157
4158        might_sleep();
4159
4160        pci_reset_bridge_secondary_bus(bus->self);
4161
4162        pci_bus_unlock(bus);
4163
4164        return 0;
4165}
4166
4167/**
4168 * pci_probe_reset_bus - probe whether a PCI bus can be reset
4169 * @bus: PCI bus to probe
4170 *
4171 * Return 0 if bus can be reset, negative if a bus reset is not supported.
4172 */
4173int pci_probe_reset_bus(struct pci_bus *bus)
4174{
4175        return pci_bus_reset(bus, 1);
4176}
4177EXPORT_SYMBOL_GPL(pci_probe_reset_bus);
4178
4179/**
4180 * pci_reset_bus - reset a PCI bus
4181 * @bus: top level PCI bus to reset
4182 *
4183 * Do a bus reset on the given bus and any subordinate buses, saving
4184 * and restoring state of all devices.
4185 *
4186 * Return 0 on success, non-zero on error.
4187 */
4188int pci_reset_bus(struct pci_bus *bus)
4189{
4190        int rc;
4191
4192        rc = pci_bus_reset(bus, 1);
4193        if (rc)
4194                return rc;
4195
4196        pci_bus_save_and_disable(bus);
4197
4198        rc = pci_bus_reset(bus, 0);
4199
4200        pci_bus_restore(bus);
4201
4202        return rc;
4203}
4204EXPORT_SYMBOL_GPL(pci_reset_bus);
4205
4206/**
4207 * pci_try_reset_bus - Try to reset a PCI bus
4208 * @bus: top level PCI bus to reset
4209 *
4210 * Same as above except return -EAGAIN if the bus cannot be locked
4211 */
4212int pci_try_reset_bus(struct pci_bus *bus)
4213{
4214        int rc;
4215
4216        rc = pci_bus_reset(bus, 1);
4217        if (rc)
4218                return rc;
4219
4220        pci_bus_save_and_disable(bus);
4221
4222        if (pci_bus_trylock(bus)) {
4223                might_sleep();
4224                pci_reset_bridge_secondary_bus(bus->self);
4225                pci_bus_unlock(bus);
4226        } else
4227                rc = -EAGAIN;
4228
4229        pci_bus_restore(bus);
4230
4231        return rc;
4232}
4233EXPORT_SYMBOL_GPL(pci_try_reset_bus);
4234
4235/**
4236 * pcix_get_max_mmrbc - get PCI-X maximum designed memory read byte count
4237 * @dev: PCI device to query
4238 *
4239 * Returns mmrbc: maximum designed memory read count in bytes
4240 *    or appropriate error value.
4241 */
4242int pcix_get_max_mmrbc(struct pci_dev *dev)
4243{
4244        int cap;
4245        u32 stat;
4246
4247        cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
4248        if (!cap)
4249                return -EINVAL;
4250
4251        if (pci_read_config_dword(dev, cap + PCI_X_STATUS, &stat))
4252                return -EINVAL;
4253
4254        return 512 << ((stat & PCI_X_STATUS_MAX_READ) >> 21);
4255}
4256EXPORT_SYMBOL(pcix_get_max_mmrbc);
4257
4258/**
4259 * pcix_get_mmrbc - get PCI-X maximum memory read byte count
4260 * @dev: PCI device to query
4261 *
4262 * Returns mmrbc: maximum memory read count in bytes
4263 *    or appropriate error value.
4264 */
4265int pcix_get_mmrbc(struct pci_dev *dev)
4266{
4267        int cap;
4268        u16 cmd;
4269
4270        cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
4271        if (!cap)
4272                return -EINVAL;
4273
4274        if (pci_read_config_word(dev, cap + PCI_X_CMD, &cmd))
4275                return -EINVAL;
4276
4277        return 512 << ((cmd & PCI_X_CMD_MAX_READ) >> 2);
4278}
4279EXPORT_SYMBOL(pcix_get_mmrbc);
4280
4281/**
4282 * pcix_set_mmrbc - set PCI-X maximum memory read byte count
4283 * @dev: PCI device to query
4284 * @mmrbc: maximum memory read count in bytes
4285 *    valid values are 512, 1024, 2048, 4096
4286 *
4287 * If possible sets maximum memory read byte count, some bridges have erratas
4288 * that prevent this.
4289 */
4290int pcix_set_mmrbc(struct pci_dev *dev, int mmrbc)
4291{
4292        int cap;
4293        u32 stat, v, o;
4294        u16 cmd;
4295
4296        if (mmrbc < 512 || mmrbc > 4096 || !is_power_of_2(mmrbc))
4297                return -EINVAL;
4298
4299        v = ffs(mmrbc) - 10;
4300
4301        cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
4302        if (!cap)
4303                return -EINVAL;
4304
4305        if (pci_read_config_dword(dev, cap + PCI_X_STATUS, &stat))
4306                return -EINVAL;
4307
4308        if (v > (stat & PCI_X_STATUS_MAX_READ) >> 21)
4309                return -E2BIG;
4310
4311        if (pci_read_config_word(dev, cap + PCI_X_CMD, &cmd))
4312                return -EINVAL;
4313
4314        o = (cmd & PCI_X_CMD_MAX_READ) >> 2;
4315        if (o != v) {
4316                if (v > o && (dev->bus->bus_flags & PCI_BUS_FLAGS_NO_MMRBC))
4317                        return -EIO;
4318
4319                cmd &= ~PCI_X_CMD_MAX_READ;
4320                cmd |= v << 2;
4321                if (pci_write_config_word(dev, cap + PCI_X_CMD, cmd))
4322                        return -EIO;
4323        }
4324        return 0;
4325}
4326EXPORT_SYMBOL(pcix_set_mmrbc);
4327
4328/**
4329 * pcie_get_readrq - get PCI Express read request size
4330 * @dev: PCI device to query
4331 *
4332 * Returns maximum memory read request in bytes
4333 *    or appropriate error value.
4334 */
4335int pcie_get_readrq(struct pci_dev *dev)
4336{
4337        u16 ctl;
4338
4339        pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &ctl);
4340
4341        return 128 << ((ctl & PCI_EXP_DEVCTL_READRQ) >> 12);
4342}
4343EXPORT_SYMBOL(pcie_get_readrq);
4344
4345/**
4346 * pcie_set_readrq - set PCI Express maximum memory read request
4347 * @dev: PCI device to query
4348 * @rq: maximum memory read count in bytes
4349 *    valid values are 128, 256, 512, 1024, 2048, 4096
4350 *
4351 * If possible sets maximum memory read request in bytes
4352 */
4353int pcie_set_readrq(struct pci_dev *dev, int rq)
4354{
4355        u16 v;
4356
4357        if (rq < 128 || rq > 4096 || !is_power_of_2(rq))
4358                return -EINVAL;
4359
4360        /*
4361         * If using the "performance" PCIe config, we clamp the
4362         * read rq size to the max packet size to prevent the
4363         * host bridge generating requests larger than we can
4364         * cope with
4365         */
4366        if (pcie_bus_config == PCIE_BUS_PERFORMANCE) {
4367                int mps = pcie_get_mps(dev);
4368
4369                if (mps < rq)
4370                        rq = mps;
4371        }
4372
4373        v = (ffs(rq) - 8) << 12;
4374
4375        return pcie_capability_clear_and_set_word(dev, PCI_EXP_DEVCTL,
4376                                                  PCI_EXP_DEVCTL_READRQ, v);
4377}
4378EXPORT_SYMBOL(pcie_set_readrq);
4379
4380/**
4381 * pcie_get_mps - get PCI Express maximum payload size
4382 * @dev: PCI device to query
4383 *
4384 * Returns maximum payload size in bytes
4385 */
4386int pcie_get_mps(struct pci_dev *dev)
4387{
4388        u16 ctl;
4389
4390        pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &ctl);
4391
4392        return 128 << ((ctl & PCI_EXP_DEVCTL_PAYLOAD) >> 5);
4393}
4394EXPORT_SYMBOL(pcie_get_mps);
4395
4396/**
4397 * pcie_set_mps - set PCI Express maximum payload size
4398 * @dev: PCI device to query
4399 * @mps: maximum payload size in bytes
4400 *    valid values are 128, 256, 512, 1024, 2048, 4096
4401 *
4402 * If possible sets maximum payload size
4403 */
4404int pcie_set_mps(struct pci_dev *dev, int mps)
4405{
4406        u16 v;
4407
4408        if (mps < 128 || mps > 4096 || !is_power_of_2(mps))
4409                return -EINVAL;
4410
4411        v = ffs(mps) - 8;
4412        if (v > dev->pcie_mpss)
4413                return -EINVAL;
4414        v <<= 5;
4415
4416        return pcie_capability_clear_and_set_word(dev, PCI_EXP_DEVCTL,
4417                                                  PCI_EXP_DEVCTL_PAYLOAD, v);
4418}
4419EXPORT_SYMBOL(pcie_set_mps);
4420
4421/**
4422 * pcie_get_minimum_link - determine minimum link settings of a PCI device
4423 * @dev: PCI device to query
4424 * @speed: storage for minimum speed
4425 * @width: storage for minimum width
4426 *
4427 * This function will walk up the PCI device chain and determine the minimum
4428 * link width and speed of the device.
4429 */
4430int pcie_get_minimum_link(struct pci_dev *dev, enum pci_bus_speed *speed,
4431                          enum pcie_link_width *width)
4432{
4433        int ret;
4434
4435        *speed = PCI_SPEED_UNKNOWN;
4436        *width = PCIE_LNK_WIDTH_UNKNOWN;
4437
4438        while (dev) {
4439                u16 lnksta;
4440                enum pci_bus_speed next_speed;
4441                enum pcie_link_width next_width;
4442
4443                ret = pcie_capability_read_word(dev, PCI_EXP_LNKSTA, &lnksta);
4444                if (ret)
4445                        return ret;
4446
4447                next_speed = pcie_link_speed[lnksta & PCI_EXP_LNKSTA_CLS];
4448                next_width = (lnksta & PCI_EXP_LNKSTA_NLW) >>
4449                        PCI_EXP_LNKSTA_NLW_SHIFT;
4450
4451                if (next_speed < *speed)
4452                        *speed = next_speed;
4453
4454                if (next_width < *width)
4455                        *width = next_width;
4456
4457                dev = dev->bus->self;
4458        }
4459
4460        return 0;
4461}
4462EXPORT_SYMBOL(pcie_get_minimum_link);
4463
4464/**
4465 * pci_select_bars - Make BAR mask from the type of resource
4466 * @dev: the PCI device for which BAR mask is made
4467 * @flags: resource type mask to be selected
4468 *
4469 * This helper routine makes bar mask from the type of resource.
4470 */
4471int pci_select_bars(struct pci_dev *dev, unsigned long flags)
4472{
4473        int i, bars = 0;
4474        for (i = 0; i < PCI_NUM_RESOURCES; i++)
4475                if (pci_resource_flags(dev, i) & flags)
4476                        bars |= (1 << i);
4477        return bars;
4478}
4479EXPORT_SYMBOL(pci_select_bars);
4480
4481/**
4482 * pci_resource_bar - get position of the BAR associated with a resource
4483 * @dev: the PCI device
4484 * @resno: the resource number
4485 * @type: the BAR type to be filled in
4486 *
4487 * Returns BAR position in config space, or 0 if the BAR is invalid.
4488 */
4489int pci_resource_bar(struct pci_dev *dev, int resno, enum pci_bar_type *type)
4490{
4491        int reg;
4492
4493        if (resno < PCI_ROM_RESOURCE) {
4494                *type = pci_bar_unknown;
4495                return PCI_BASE_ADDRESS_0 + 4 * resno;
4496        } else if (resno == PCI_ROM_RESOURCE) {
4497                *type = pci_bar_mem32;
4498                return dev->rom_base_reg;
4499        } else if (resno < PCI_BRIDGE_RESOURCES) {
4500                /* device specific resource */
4501                *type = pci_bar_unknown;
4502                reg = pci_iov_resource_bar(dev, resno);
4503                if (reg)
4504                        return reg;
4505        }
4506
4507        dev_err(&dev->dev, "BAR %d: invalid resource\n", resno);
4508        return 0;
4509}
4510
4511/* Some architectures require additional programming to enable VGA */
4512static arch_set_vga_state_t arch_set_vga_state;
4513
4514void __init pci_register_set_vga_state(arch_set_vga_state_t func)
4515{
4516        arch_set_vga_state = func;      /* NULL disables */
4517}
4518
4519static int pci_set_vga_state_arch(struct pci_dev *dev, bool decode,
4520                                  unsigned int command_bits, u32 flags)
4521{
4522        if (arch_set_vga_state)
4523                return arch_set_vga_state(dev, decode, command_bits,
4524                                                flags);
4525        return 0;
4526}
4527
4528/**
4529 * pci_set_vga_state - set VGA decode state on device and parents if requested
4530 * @dev: the PCI device
4531 * @decode: true = enable decoding, false = disable decoding
4532 * @command_bits: PCI_COMMAND_IO and/or PCI_COMMAND_MEMORY
4533 * @flags: traverse ancestors and change bridges
4534 * CHANGE_BRIDGE_ONLY / CHANGE_BRIDGE
4535 */
4536int pci_set_vga_state(struct pci_dev *dev, bool decode,
4537                      unsigned int command_bits, u32 flags)
4538{
4539        struct pci_bus *bus;
4540        struct pci_dev *bridge;
4541        u16 cmd;
4542        int rc;
4543
4544        WARN_ON((flags & PCI_VGA_STATE_CHANGE_DECODES) && (command_bits & ~(PCI_COMMAND_IO|PCI_COMMAND_MEMORY)));
4545
4546        /* ARCH specific VGA enables */
4547        rc = pci_set_vga_state_arch(dev, decode, command_bits, flags);
4548        if (rc)
4549                return rc;
4550
4551        if (flags & PCI_VGA_STATE_CHANGE_DECODES) {
4552                pci_read_config_word(dev, PCI_COMMAND, &cmd);
4553                if (decode == true)
4554                        cmd |= command_bits;
4555                else
4556                        cmd &= ~command_bits;
4557                pci_write_config_word(dev, PCI_COMMAND, cmd);
4558        }
4559
4560        if (!(flags & PCI_VGA_STATE_CHANGE_BRIDGE))
4561                return 0;
4562
4563        bus = dev->bus;
4564        while (bus) {
4565                bridge = bus->self;
4566                if (bridge) {
4567                        pci_read_config_word(bridge, PCI_BRIDGE_CONTROL,
4568                                             &cmd);
4569                        if (decode == true)
4570                                cmd |= PCI_BRIDGE_CTL_VGA;
4571                        else
4572                                cmd &= ~PCI_BRIDGE_CTL_VGA;
4573                        pci_write_config_word(bridge, PCI_BRIDGE_CONTROL,
4574                                              cmd);
4575                }
4576                bus = bus->parent;
4577        }
4578        return 0;
4579}
4580
4581bool pci_device_is_present(struct pci_dev *pdev)
4582{
4583        u32 v;
4584
4585        return pci_bus_read_dev_vendor_id(pdev->bus, pdev->devfn, &v, 0);
4586}
4587EXPORT_SYMBOL_GPL(pci_device_is_present);
4588
4589void pci_ignore_hotplug(struct pci_dev *dev)
4590{
4591        struct pci_dev *bridge = dev->bus->self;
4592
4593        dev->ignore_hotplug = 1;
4594        /* Propagate the "ignore hotplug" setting to the parent bridge. */
4595        if (bridge)
4596                bridge->ignore_hotplug = 1;
4597}
4598EXPORT_SYMBOL_GPL(pci_ignore_hotplug);
4599
4600#define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
4601static char resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
4602static DEFINE_SPINLOCK(resource_alignment_lock);
4603
4604/**
4605 * pci_specified_resource_alignment - get resource alignment specified by user.
4606 * @dev: the PCI device to get
4607 *
4608 * RETURNS: Resource alignment if it is specified.
4609 *          Zero if it is not specified.
4610 */
4611static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev)
4612{
4613        int seg, bus, slot, func, align_order, count;
4614        resource_size_t align = 0;
4615        char *p;
4616
4617        spin_lock(&resource_alignment_lock);
4618        p = resource_alignment_param;
4619        while (*p) {
4620                count = 0;
4621                if (sscanf(p, "%d%n", &align_order, &count) == 1 &&
4622                                                        p[count] == '@') {
4623                        p += count + 1;
4624                } else {
4625                        align_order = -1;
4626                }
4627                if (sscanf(p, "%x:%x:%x.%x%n",
4628                        &seg, &bus, &slot, &func, &count) != 4) {
4629                        seg = 0;
4630                        if (sscanf(p, "%x:%x.%x%n",
4631                                        &bus, &slot, &func, &count) != 3) {
4632                                /* Invalid format */
4633                                printk(KERN_ERR "PCI: Can't parse resource_alignment parameter: %s\n",
4634                                        p);
4635                                break;
4636                        }
4637                }
4638                p += count;
4639                if (seg == pci_domain_nr(dev->bus) &&
4640                        bus == dev->bus->number &&
4641                        slot == PCI_SLOT(dev->devfn) &&
4642                        func == PCI_FUNC(dev->devfn)) {
4643                        if (align_order == -1)
4644                                align = PAGE_SIZE;
4645                        else
4646                                align = 1 << align_order;
4647                        /* Found */
4648                        break;
4649                }
4650                if (*p != ';' && *p != ',') {
4651                        /* End of param or invalid format */
4652                        break;
4653                }
4654                p++;
4655        }
4656        spin_unlock(&resource_alignment_lock);
4657        return align;
4658}
4659
4660/*
4661 * This function disables memory decoding and releases memory resources
4662 * of the device specified by kernel's boot parameter 'pci=resource_alignment='.
4663 * It also rounds up size to specified alignment.
4664 * Later on, the kernel will assign page-aligned memory resource back
4665 * to the device.
4666 */
4667void pci_reassigndev_resource_alignment(struct pci_dev *dev)
4668{
4669        int i;
4670        struct resource *r;
4671        resource_size_t align, size;
4672        u16 command;
4673
4674        /* check if specified PCI is target device to reassign */
4675        align = pci_specified_resource_alignment(dev);
4676        if (!align)
4677                return;
4678
4679        if (dev->hdr_type == PCI_HEADER_TYPE_NORMAL &&
4680            (dev->class >> 8) == PCI_CLASS_BRIDGE_HOST) {
4681                dev_warn(&dev->dev,
4682                        "Can't reassign resources to host bridge.\n");
4683                return;
4684        }
4685
4686        dev_info(&dev->dev,
4687                "Disabling memory decoding and releasing memory resources.\n");
4688        pci_read_config_word(dev, PCI_COMMAND, &command);
4689        command &= ~PCI_COMMAND_MEMORY;
4690        pci_write_config_word(dev, PCI_COMMAND, command);
4691
4692        for (i = 0; i < PCI_BRIDGE_RESOURCES; i++) {
4693                r = &dev->resource[i];
4694                if (!(r->flags & IORESOURCE_MEM))
4695                        continue;
4696                size = resource_size(r);
4697                if (size < align) {
4698                        size = align;
4699                        dev_info(&dev->dev,
4700                                "Rounding up size of resource #%d to %#llx.\n",
4701                                i, (unsigned long long)size);
4702                }
4703                r->flags |= IORESOURCE_UNSET;
4704                r->end = size - 1;
4705                r->start = 0;
4706        }
4707        /* Need to disable bridge's resource window,
4708         * to enable the kernel to reassign new resource
4709         * window later on.
4710         */
4711        if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE &&
4712            (dev->class >> 8) == PCI_CLASS_BRIDGE_PCI) {
4713                for (i = PCI_BRIDGE_RESOURCES; i < PCI_NUM_RESOURCES; i++) {
4714                        r = &dev->resource[i];
4715                        if (!(r->flags & IORESOURCE_MEM))
4716                                continue;
4717                        r->flags |= IORESOURCE_UNSET;
4718                        r->end = resource_size(r) - 1;
4719                        r->start = 0;
4720                }
4721                pci_disable_bridge_window(dev);
4722        }
4723}
4724
4725static ssize_t pci_set_resource_alignment_param(const char *buf, size_t count)
4726{
4727        if (count > RESOURCE_ALIGNMENT_PARAM_SIZE - 1)
4728                count = RESOURCE_ALIGNMENT_PARAM_SIZE - 1;
4729        spin_lock(&resource_alignment_lock);
4730        strncpy(resource_alignment_param, buf, count);
4731        resource_alignment_param[count] = '\0';
4732        spin_unlock(&resource_alignment_lock);
4733        return count;
4734}
4735
4736static ssize_t pci_get_resource_alignment_param(char *buf, size_t size)
4737{
4738        size_t count;
4739        spin_lock(&resource_alignment_lock);
4740        count = snprintf(buf, size, "%s", resource_alignment_param);
4741        spin_unlock(&resource_alignment_lock);
4742        return count;
4743}
4744
4745static ssize_t pci_resource_alignment_show(struct bus_type *bus, char *buf)
4746{
4747        return pci_get_resource_alignment_param(buf, PAGE_SIZE);
4748}
4749
4750static ssize_t pci_resource_alignment_store(struct bus_type *bus,
4751                                        const char *buf, size_t count)
4752{
4753        return pci_set_resource_alignment_param(buf, count);
4754}
4755
4756BUS_ATTR(resource_alignment, 0644, pci_resource_alignment_show,
4757                                        pci_resource_alignment_store);
4758
4759static int __init pci_resource_alignment_sysfs_init(void)
4760{
4761        return bus_create_file(&pci_bus_type,
4762                                        &bus_attr_resource_alignment);
4763}
4764late_initcall(pci_resource_alignment_sysfs_init);
4765
4766static void pci_no_domains(void)
4767{
4768#ifdef CONFIG_PCI_DOMAINS
4769        pci_domains_supported = 0;
4770#endif
4771}
4772
4773#ifdef CONFIG_PCI_DOMAINS
4774static atomic_t __domain_nr = ATOMIC_INIT(-1);
4775
4776int pci_get_new_domain_nr(void)
4777{
4778        return atomic_inc_return(&__domain_nr);
4779}
4780
4781#ifdef CONFIG_PCI_DOMAINS_GENERIC
4782void pci_bus_assign_domain_nr(struct pci_bus *bus, struct device *parent)
4783{
4784        static int use_dt_domains = -1;
4785        int domain = -1;
4786
4787        if (parent)
4788                domain = of_get_pci_domain_nr(parent->of_node);
4789        /*
4790         * Check DT domain and use_dt_domains values.
4791         *
4792         * If DT domain property is valid (domain >= 0) and
4793         * use_dt_domains != 0, the DT assignment is valid since this means
4794         * we have not previously allocated a domain number by using
4795         * pci_get_new_domain_nr(); we should also update use_dt_domains to
4796         * 1, to indicate that we have just assigned a domain number from
4797         * DT.
4798         *
4799         * If DT domain property value is not valid (ie domain < 0), and we
4800         * have not previously assigned a domain number from DT
4801         * (use_dt_domains != 1) we should assign a domain number by
4802         * using the:
4803         *
4804         * pci_get_new_domain_nr()
4805         *
4806         * API and update the use_dt_domains value to keep track of method we
4807         * are using to assign domain numbers (use_dt_domains = 0).
4808         *
4809         * All other combinations imply we have a platform that is trying
4810         * to mix domain numbers obtained from DT and pci_get_new_domain_nr(),
4811         * which is a recipe for domain mishandling and it is prevented by
4812         * invalidating the domain value (domain = -1) and printing a
4813         * corresponding error.
4814         */
4815        if (domain >= 0 && use_dt_domains) {
4816                use_dt_domains = 1;
4817        } else if (domain < 0 && use_dt_domains != 1) {
4818                use_dt_domains = 0;
4819                domain = pci_get_new_domain_nr();
4820        } else {
4821                dev_err(parent, "Node %s has inconsistent \"linux,pci-domain\" property in DT\n",
4822                        parent->of_node->full_name);
4823                domain = -1;
4824        }
4825
4826        bus->domain_nr = domain;
4827}
4828#endif
4829#endif
4830
4831/**
4832 * pci_ext_cfg_avail - can we access extended PCI config space?
4833 *
4834 * Returns 1 if we can access PCI extended config space (offsets
4835 * greater than 0xff). This is the default implementation. Architecture
4836 * implementations can override this.
4837 */
4838int __weak pci_ext_cfg_avail(void)
4839{
4840        return 1;
4841}
4842
4843void __weak pci_fixup_cardbus(struct pci_bus *bus)
4844{
4845}
4846EXPORT_SYMBOL(pci_fixup_cardbus);
4847
4848static int __init pci_setup(char *str)
4849{
4850        while (str) {
4851                char *k = strchr(str, ',');
4852                if (k)
4853                        *k++ = 0;
4854                if (*str && (str = pcibios_setup(str)) && *str) {
4855                        if (!strcmp(str, "nomsi")) {
4856                                pci_no_msi();
4857                        } else if (!strcmp(str, "noaer")) {
4858                                pci_no_aer();
4859                        } else if (!strncmp(str, "realloc=", 8)) {
4860                                pci_realloc_get_opt(str + 8);
4861                        } else if (!strncmp(str, "realloc", 7)) {
4862                                pci_realloc_get_opt("on");
4863                        } else if (!strcmp(str, "nodomains")) {
4864                                pci_no_domains();
4865                        } else if (!strncmp(str, "noari", 5)) {
4866                                pcie_ari_disabled = true;
4867                        } else if (!strncmp(str, "cbiosize=", 9)) {
4868                                pci_cardbus_io_size = memparse(str + 9, &str);
4869                        } else if (!strncmp(str, "cbmemsize=", 10)) {
4870                                pci_cardbus_mem_size = memparse(str + 10, &str);
4871                        } else if (!strncmp(str, "resource_alignment=", 19)) {
4872                                pci_set_resource_alignment_param(str + 19,
4873                                                        strlen(str + 19));
4874                        } else if (!strncmp(str, "ecrc=", 5)) {
4875                                pcie_ecrc_get_policy(str + 5);
4876                        } else if (!strncmp(str, "hpiosize=", 9)) {
4877                                pci_hotplug_io_size = memparse(str + 9, &str);
4878                        } else if (!strncmp(str, "hpmemsize=", 10)) {
4879                                pci_hotplug_mem_size = memparse(str + 10, &str);
4880                        } else if (!strncmp(str, "pcie_bus_tune_off", 17)) {
4881                                pcie_bus_config = PCIE_BUS_TUNE_OFF;
4882                        } else if (!strncmp(str, "pcie_bus_safe", 13)) {
4883                                pcie_bus_config = PCIE_BUS_SAFE;
4884                        } else if (!strncmp(str, "pcie_bus_perf", 13)) {
4885                                pcie_bus_config = PCIE_BUS_PERFORMANCE;
4886                        } else if (!strncmp(str, "pcie_bus_peer2peer", 18)) {
4887                                pcie_bus_config = PCIE_BUS_PEER2PEER;
4888                        } else if (!strncmp(str, "pcie_scan_all", 13)) {
4889                                pci_add_flags(PCI_SCAN_ALL_PCIE_DEVS);
4890                        } else {
4891                                printk(KERN_ERR "PCI: Unknown option `%s'\n",
4892                                                str);
4893                        }
4894                }
4895                str = k;
4896        }
4897        return 0;
4898}
4899early_param("pci", pci_setup);
4900