2 * PCI Bus Services, see include/linux/pci.h for further explanation.
4 * Copyright 1993 -- 1997 Drew Eckhardt, Frederic Potter,
7 * Copyright 1997 -- 2000 Martin Mares <mj@ucw.cz>
10 #include <linux/acpi.h>
11 #include <linux/kernel.h>
12 #include <linux/delay.h>
13 #include <linux/dmi.h>
14 #include <linux/init.h>
16 #include <linux/of_pci.h>
17 #include <linux/pci.h>
19 #include <linux/slab.h>
20 #include <linux/module.h>
21 #include <linux/spinlock.h>
22 #include <linux/string.h>
23 #include <linux/log2.h>
24 #include <linux/pci-aspm.h>
25 #include <linux/pm_wakeup.h>
26 #include <linux/interrupt.h>
27 #include <linux/device.h>
28 #include <linux/pm_runtime.h>
29 #include <linux/pci_hotplug.h>
30 #include <linux/vmalloc.h>
31 #include <asm/setup.h>
33 #include <linux/aer.h>
36 const char *pci_power_names[] = {
37 "error", "D0", "D1", "D2", "D3hot", "D3cold", "unknown",
39 EXPORT_SYMBOL_GPL(pci_power_names);
41 int isa_dma_bridge_buggy;
42 EXPORT_SYMBOL(isa_dma_bridge_buggy);
45 EXPORT_SYMBOL(pci_pci_problems);
47 unsigned int pci_pm_d3_delay;
49 static void pci_pme_list_scan(struct work_struct *work);
51 static LIST_HEAD(pci_pme_list);
52 static DEFINE_MUTEX(pci_pme_list_mutex);
53 static DECLARE_DELAYED_WORK(pci_pme_work, pci_pme_list_scan);
55 struct pci_pme_device {
56 struct list_head list;
60 #define PME_TIMEOUT 1000 /* How long between PME checks */
62 static void pci_dev_d3_sleep(struct pci_dev *dev)
64 unsigned int delay = dev->d3_delay;
66 if (delay < pci_pm_d3_delay)
67 delay = pci_pm_d3_delay;
72 #ifdef CONFIG_PCI_DOMAINS
73 int pci_domains_supported = 1;
76 #define DEFAULT_CARDBUS_IO_SIZE (256)
77 #define DEFAULT_CARDBUS_MEM_SIZE (64*1024*1024)
78 /* pci=cbmemsize=nnM,cbiosize=nn can override this */
79 unsigned long pci_cardbus_io_size = DEFAULT_CARDBUS_IO_SIZE;
80 unsigned long pci_cardbus_mem_size = DEFAULT_CARDBUS_MEM_SIZE;
82 #define DEFAULT_HOTPLUG_IO_SIZE (256)
83 #define DEFAULT_HOTPLUG_MEM_SIZE (2*1024*1024)
84 /* pci=hpmemsize=nnM,hpiosize=nn can override this */
85 unsigned long pci_hotplug_io_size = DEFAULT_HOTPLUG_IO_SIZE;
86 unsigned long pci_hotplug_mem_size = DEFAULT_HOTPLUG_MEM_SIZE;
88 #define DEFAULT_HOTPLUG_BUS_SIZE 1
89 unsigned long pci_hotplug_bus_size = DEFAULT_HOTPLUG_BUS_SIZE;
91 enum pcie_bus_config_types pcie_bus_config = PCIE_BUS_DEFAULT;
94 * The default CLS is used if arch didn't set CLS explicitly and not
95 * all pci devices agree on the same value. Arch can override either
96 * the dfl or actual value as it sees fit. Don't forget this is
97 * measured in 32-bit words, not bytes.
99 u8 pci_dfl_cache_line_size = L1_CACHE_BYTES >> 2;
100 u8 pci_cache_line_size;
103 * If we set up a device for bus mastering, we need to check the latency
104 * timer as certain BIOSes forget to set it properly.
106 unsigned int pcibios_max_latency = 255;
108 /* If set, the PCIe ARI capability will not be used. */
109 static bool pcie_ari_disabled;
111 /* Disable bridge_d3 for all PCIe ports */
112 static bool pci_bridge_d3_disable;
113 /* Force bridge_d3 for all PCIe ports */
114 static bool pci_bridge_d3_force;
116 static int __init pcie_port_pm_setup(char *str)
118 if (!strcmp(str, "off"))
119 pci_bridge_d3_disable = true;
120 else if (!strcmp(str, "force"))
121 pci_bridge_d3_force = true;
124 __setup("pcie_port_pm=", pcie_port_pm_setup);
127 * pci_bus_max_busnr - returns maximum PCI bus number of given bus' children
128 * @bus: pointer to PCI bus structure to search
130 * Given a PCI bus, returns the highest PCI bus number present in the set
131 * including the given PCI bus and its list of child PCI buses.
133 unsigned char pci_bus_max_busnr(struct pci_bus *bus)
136 unsigned char max, n;
138 max = bus->busn_res.end;
139 list_for_each_entry(tmp, &bus->children, node) {
140 n = pci_bus_max_busnr(tmp);
146 EXPORT_SYMBOL_GPL(pci_bus_max_busnr);
148 #ifdef CONFIG_HAS_IOMEM
149 void __iomem *pci_ioremap_bar(struct pci_dev *pdev, int bar)
151 struct resource *res = &pdev->resource[bar];
154 * Make sure the BAR is actually a memory resource, not an IO resource
156 if (res->flags & IORESOURCE_UNSET || !(res->flags & IORESOURCE_MEM)) {
157 dev_warn(&pdev->dev, "can't ioremap BAR %d: %pR\n", bar, res);
160 return ioremap_nocache(res->start, resource_size(res));
162 EXPORT_SYMBOL_GPL(pci_ioremap_bar);
164 void __iomem *pci_ioremap_wc_bar(struct pci_dev *pdev, int bar)
167 * Make sure the BAR is actually a memory resource, not an IO resource
169 if (!(pci_resource_flags(pdev, bar) & IORESOURCE_MEM)) {
173 return ioremap_wc(pci_resource_start(pdev, bar),
174 pci_resource_len(pdev, bar));
176 EXPORT_SYMBOL_GPL(pci_ioremap_wc_bar);
180 static int __pci_find_next_cap_ttl(struct pci_bus *bus, unsigned int devfn,
181 u8 pos, int cap, int *ttl)
186 pci_bus_read_config_byte(bus, devfn, pos, &pos);
192 pci_bus_read_config_word(bus, devfn, pos, &ent);
204 static int __pci_find_next_cap(struct pci_bus *bus, unsigned int devfn,
207 int ttl = PCI_FIND_CAP_TTL;
209 return __pci_find_next_cap_ttl(bus, devfn, pos, cap, &ttl);
212 int pci_find_next_capability(struct pci_dev *dev, u8 pos, int cap)
214 return __pci_find_next_cap(dev->bus, dev->devfn,
215 pos + PCI_CAP_LIST_NEXT, cap);
217 EXPORT_SYMBOL_GPL(pci_find_next_capability);
219 static int __pci_bus_find_cap_start(struct pci_bus *bus,
220 unsigned int devfn, u8 hdr_type)
224 pci_bus_read_config_word(bus, devfn, PCI_STATUS, &status);
225 if (!(status & PCI_STATUS_CAP_LIST))
229 case PCI_HEADER_TYPE_NORMAL:
230 case PCI_HEADER_TYPE_BRIDGE:
231 return PCI_CAPABILITY_LIST;
232 case PCI_HEADER_TYPE_CARDBUS:
233 return PCI_CB_CAPABILITY_LIST;
240 * pci_find_capability - query for devices' capabilities
241 * @dev: PCI device to query
242 * @cap: capability code
244 * Tell if a device supports a given PCI capability.
245 * Returns the address of the requested capability structure within the
246 * device's PCI configuration space or 0 in case the device does not
247 * support it. Possible values for @cap:
249 * %PCI_CAP_ID_PM Power Management
250 * %PCI_CAP_ID_AGP Accelerated Graphics Port
251 * %PCI_CAP_ID_VPD Vital Product Data
252 * %PCI_CAP_ID_SLOTID Slot Identification
253 * %PCI_CAP_ID_MSI Message Signalled Interrupts
254 * %PCI_CAP_ID_CHSWP CompactPCI HotSwap
255 * %PCI_CAP_ID_PCIX PCI-X
256 * %PCI_CAP_ID_EXP PCI Express
258 int pci_find_capability(struct pci_dev *dev, int cap)
262 pos = __pci_bus_find_cap_start(dev->bus, dev->devfn, dev->hdr_type);
264 pos = __pci_find_next_cap(dev->bus, dev->devfn, pos, cap);
268 EXPORT_SYMBOL(pci_find_capability);
271 * pci_bus_find_capability - query for devices' capabilities
272 * @bus: the PCI bus to query
273 * @devfn: PCI device to query
274 * @cap: capability code
276 * Like pci_find_capability() but works for pci devices that do not have a
277 * pci_dev structure set up yet.
279 * Returns the address of the requested capability structure within the
280 * device's PCI configuration space or 0 in case the device does not
283 int pci_bus_find_capability(struct pci_bus *bus, unsigned int devfn, int cap)
288 pci_bus_read_config_byte(bus, devfn, PCI_HEADER_TYPE, &hdr_type);
290 pos = __pci_bus_find_cap_start(bus, devfn, hdr_type & 0x7f);
292 pos = __pci_find_next_cap(bus, devfn, pos, cap);
296 EXPORT_SYMBOL(pci_bus_find_capability);
299 * pci_find_next_ext_capability - Find an extended capability
300 * @dev: PCI device to query
301 * @start: address at which to start looking (0 to start at beginning of list)
302 * @cap: capability code
304 * Returns the address of the next matching extended capability structure
305 * within the device's PCI configuration space or 0 if the device does
306 * not support it. Some capabilities can occur several times, e.g., the
307 * vendor-specific capability, and this provides a way to find them all.
309 int pci_find_next_ext_capability(struct pci_dev *dev, int start, int cap)
313 int pos = PCI_CFG_SPACE_SIZE;
315 /* minimum 8 bytes per capability */
316 ttl = (PCI_CFG_SPACE_EXP_SIZE - PCI_CFG_SPACE_SIZE) / 8;
318 if (dev->cfg_size <= PCI_CFG_SPACE_SIZE)
324 if (pci_read_config_dword(dev, pos, &header) != PCIBIOS_SUCCESSFUL)
328 * If we have no capabilities, this is indicated by cap ID,
329 * cap version and next pointer all being 0.
335 if (PCI_EXT_CAP_ID(header) == cap && pos != start)
338 pos = PCI_EXT_CAP_NEXT(header);
339 if (pos < PCI_CFG_SPACE_SIZE)
342 if (pci_read_config_dword(dev, pos, &header) != PCIBIOS_SUCCESSFUL)
348 EXPORT_SYMBOL_GPL(pci_find_next_ext_capability);
351 * pci_find_ext_capability - Find an extended capability
352 * @dev: PCI device to query
353 * @cap: capability code
355 * Returns the address of the requested extended capability structure
356 * within the device's PCI configuration space or 0 if the device does
357 * not support it. Possible values for @cap:
359 * %PCI_EXT_CAP_ID_ERR Advanced Error Reporting
360 * %PCI_EXT_CAP_ID_VC Virtual Channel
361 * %PCI_EXT_CAP_ID_DSN Device Serial Number
362 * %PCI_EXT_CAP_ID_PWR Power Budgeting
364 int pci_find_ext_capability(struct pci_dev *dev, int cap)
366 return pci_find_next_ext_capability(dev, 0, cap);
368 EXPORT_SYMBOL_GPL(pci_find_ext_capability);
370 static int __pci_find_next_ht_cap(struct pci_dev *dev, int pos, int ht_cap)
372 int rc, ttl = PCI_FIND_CAP_TTL;
375 if (ht_cap == HT_CAPTYPE_SLAVE || ht_cap == HT_CAPTYPE_HOST)
376 mask = HT_3BIT_CAP_MASK;
378 mask = HT_5BIT_CAP_MASK;
380 pos = __pci_find_next_cap_ttl(dev->bus, dev->devfn, pos,
381 PCI_CAP_ID_HT, &ttl);
383 rc = pci_read_config_byte(dev, pos + 3, &cap);
384 if (rc != PCIBIOS_SUCCESSFUL)
387 if ((cap & mask) == ht_cap)
390 pos = __pci_find_next_cap_ttl(dev->bus, dev->devfn,
391 pos + PCI_CAP_LIST_NEXT,
392 PCI_CAP_ID_HT, &ttl);
398 * pci_find_next_ht_capability - query a device's Hypertransport capabilities
399 * @dev: PCI device to query
400 * @pos: Position from which to continue searching
401 * @ht_cap: Hypertransport capability code
403 * To be used in conjunction with pci_find_ht_capability() to search for
404 * all capabilities matching @ht_cap. @pos should always be a value returned
405 * from pci_find_ht_capability().
407 * NB. To be 100% safe against broken PCI devices, the caller should take
408 * steps to avoid an infinite loop.
410 int pci_find_next_ht_capability(struct pci_dev *dev, int pos, int ht_cap)
412 return __pci_find_next_ht_cap(dev, pos + PCI_CAP_LIST_NEXT, ht_cap);
414 EXPORT_SYMBOL_GPL(pci_find_next_ht_capability);
417 * pci_find_ht_capability - query a device's Hypertransport capabilities
418 * @dev: PCI device to query
419 * @ht_cap: Hypertransport capability code
421 * Tell if a device supports a given Hypertransport capability.
422 * Returns an address within the device's PCI configuration space
423 * or 0 in case the device does not support the request capability.
424 * The address points to the PCI capability, of type PCI_CAP_ID_HT,
425 * which has a Hypertransport capability matching @ht_cap.
427 int pci_find_ht_capability(struct pci_dev *dev, int ht_cap)
431 pos = __pci_bus_find_cap_start(dev->bus, dev->devfn, dev->hdr_type);
433 pos = __pci_find_next_ht_cap(dev, pos, ht_cap);
437 EXPORT_SYMBOL_GPL(pci_find_ht_capability);
440 * pci_find_parent_resource - return resource region of parent bus of given region
441 * @dev: PCI device structure contains resources to be searched
442 * @res: child resource record for which parent is sought
444 * For given resource region of given device, return the resource
445 * region of parent bus the given region is contained in.
447 struct resource *pci_find_parent_resource(const struct pci_dev *dev,
448 struct resource *res)
450 const struct pci_bus *bus = dev->bus;
454 pci_bus_for_each_resource(bus, r, i) {
457 if (res->start && resource_contains(r, res)) {
460 * If the window is prefetchable but the BAR is
461 * not, the allocator made a mistake.
463 if (r->flags & IORESOURCE_PREFETCH &&
464 !(res->flags & IORESOURCE_PREFETCH))
468 * If we're below a transparent bridge, there may
469 * be both a positively-decoded aperture and a
470 * subtractively-decoded region that contain the BAR.
471 * We want the positively-decoded one, so this depends
472 * on pci_bus_for_each_resource() giving us those
480 EXPORT_SYMBOL(pci_find_parent_resource);
483 * pci_find_pcie_root_port - return PCIe Root Port
484 * @dev: PCI device to query
486 * Traverse up the parent chain and return the PCIe Root Port PCI Device
487 * for a given PCI Device.
489 struct pci_dev *pci_find_pcie_root_port(struct pci_dev *dev)
491 struct pci_dev *bridge, *highest_pcie_bridge = NULL;
493 bridge = pci_upstream_bridge(dev);
494 while (bridge && pci_is_pcie(bridge)) {
495 highest_pcie_bridge = bridge;
496 bridge = pci_upstream_bridge(bridge);
499 if (pci_pcie_type(highest_pcie_bridge) != PCI_EXP_TYPE_ROOT_PORT)
502 return highest_pcie_bridge;
504 EXPORT_SYMBOL(pci_find_pcie_root_port);
507 * pci_wait_for_pending - wait for @mask bit(s) to clear in status word @pos
508 * @dev: the PCI device to operate on
509 * @pos: config space offset of status word
510 * @mask: mask of bit(s) to care about in status word
512 * Return 1 when mask bit(s) in status word clear, 0 otherwise.
514 int pci_wait_for_pending(struct pci_dev *dev, int pos, u16 mask)
518 /* Wait for Transaction Pending bit clean */
519 for (i = 0; i < 4; i++) {
522 msleep((1 << (i - 1)) * 100);
524 pci_read_config_word(dev, pos, &status);
525 if (!(status & mask))
533 * pci_restore_bars - restore a device's BAR values (e.g. after wake-up)
534 * @dev: PCI device to have its BARs restored
536 * Restore the BAR values for a given device, so as to make it
537 * accessible by its driver.
539 static void pci_restore_bars(struct pci_dev *dev)
543 /* Per SR-IOV spec 3.4.1.11, VF BARs are RO zero */
547 for (i = 0; i < PCI_BRIDGE_RESOURCES; i++)
548 pci_update_resource(dev, i);
551 static const struct pci_platform_pm_ops *pci_platform_pm;
553 int pci_set_platform_pm(const struct pci_platform_pm_ops *ops)
555 if (!ops->is_manageable || !ops->set_state || !ops->get_state ||
556 !ops->choose_state || !ops->sleep_wake || !ops->run_wake ||
559 pci_platform_pm = ops;
563 static inline bool platform_pci_power_manageable(struct pci_dev *dev)
565 return pci_platform_pm ? pci_platform_pm->is_manageable(dev) : false;
568 static inline int platform_pci_set_power_state(struct pci_dev *dev,
571 return pci_platform_pm ? pci_platform_pm->set_state(dev, t) : -ENOSYS;
574 static inline pci_power_t platform_pci_get_power_state(struct pci_dev *dev)
576 return pci_platform_pm ? pci_platform_pm->get_state(dev) : PCI_UNKNOWN;
579 static inline pci_power_t platform_pci_choose_state(struct pci_dev *dev)
581 return pci_platform_pm ?
582 pci_platform_pm->choose_state(dev) : PCI_POWER_ERROR;
585 static inline int platform_pci_sleep_wake(struct pci_dev *dev, bool enable)
587 return pci_platform_pm ?
588 pci_platform_pm->sleep_wake(dev, enable) : -ENODEV;
591 static inline int platform_pci_run_wake(struct pci_dev *dev, bool enable)
593 return pci_platform_pm ?
594 pci_platform_pm->run_wake(dev, enable) : -ENODEV;
597 static inline bool platform_pci_need_resume(struct pci_dev *dev)
599 return pci_platform_pm ? pci_platform_pm->need_resume(dev) : false;
603 * pci_raw_set_power_state - Use PCI PM registers to set the power state of
605 * @dev: PCI device to handle.
606 * @state: PCI power state (D0, D1, D2, D3hot) to put the device into.
609 * -EINVAL if the requested state is invalid.
610 * -EIO if device does not support PCI PM or its PM capabilities register has a
611 * wrong version, or device doesn't support the requested state.
612 * 0 if device already is in the requested state.
613 * 0 if device's power state has been successfully changed.
615 static int pci_raw_set_power_state(struct pci_dev *dev, pci_power_t state)
618 bool need_restore = false;
620 /* Check if we're already there */
621 if (dev->current_state == state)
627 if (state < PCI_D0 || state > PCI_D3hot)
630 /* Validate current state:
631 * Can enter D0 from any state, but if we can only go deeper
632 * to sleep if we're already in a low power state
634 if (state != PCI_D0 && dev->current_state <= PCI_D3cold
635 && dev->current_state > state) {
636 dev_err(&dev->dev, "invalid power transition (from state %d to %d)\n",
637 dev->current_state, state);
641 /* check if this device supports the desired state */
642 if ((state == PCI_D1 && !dev->d1_support)
643 || (state == PCI_D2 && !dev->d2_support))
646 pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
648 /* If we're (effectively) in D3, force entire word to 0.
649 * This doesn't affect PME_Status, disables PME_En, and
650 * sets PowerState to 0.
652 switch (dev->current_state) {
656 pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
661 case PCI_UNKNOWN: /* Boot-up */
662 if ((pmcsr & PCI_PM_CTRL_STATE_MASK) == PCI_D3hot
663 && !(pmcsr & PCI_PM_CTRL_NO_SOFT_RESET))
665 /* Fall-through: force to D0 */
671 /* enter specified state */
672 pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, pmcsr);
674 /* Mandatory power management transition delays */
675 /* see PCI PM 1.1 5.6.1 table 18 */
676 if (state == PCI_D3hot || dev->current_state == PCI_D3hot)
677 pci_dev_d3_sleep(dev);
678 else if (state == PCI_D2 || dev->current_state == PCI_D2)
679 udelay(PCI_PM_D2_DELAY);
681 pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
682 dev->current_state = (pmcsr & PCI_PM_CTRL_STATE_MASK);
683 if (dev->current_state != state && printk_ratelimit())
684 dev_info(&dev->dev, "Refused to change power state, currently in D%d\n",
688 * According to section 5.4.1 of the "PCI BUS POWER MANAGEMENT
689 * INTERFACE SPECIFICATION, REV. 1.2", a device transitioning
690 * from D3hot to D0 _may_ perform an internal reset, thereby
691 * going to "D0 Uninitialized" rather than "D0 Initialized".
692 * For example, at least some versions of the 3c905B and the
693 * 3c556B exhibit this behaviour.
695 * At least some laptop BIOSen (e.g. the Thinkpad T21) leave
696 * devices in a D3hot state at boot. Consequently, we need to
697 * restore at least the BARs so that the device will be
698 * accessible to its driver.
701 pci_restore_bars(dev);
704 pcie_aspm_pm_state_change(dev->bus->self);
710 * pci_update_current_state - Read power state of given device and cache it
711 * @dev: PCI device to handle.
712 * @state: State to cache in case the device doesn't have the PM capability
714 * The power state is read from the PMCSR register, which however is
715 * inaccessible in D3cold. The platform firmware is therefore queried first
716 * to detect accessibility of the register. In case the platform firmware
717 * reports an incorrect state or the device isn't power manageable by the
718 * platform at all, we try to detect D3cold by testing accessibility of the
719 * vendor ID in config space.
721 void pci_update_current_state(struct pci_dev *dev, pci_power_t state)
723 if (platform_pci_get_power_state(dev) == PCI_D3cold ||
724 !pci_device_is_present(dev)) {
725 dev->current_state = PCI_D3cold;
726 } else if (dev->pm_cap) {
729 pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
730 dev->current_state = (pmcsr & PCI_PM_CTRL_STATE_MASK);
732 dev->current_state = state;
737 * pci_power_up - Put the given device into D0 forcibly
738 * @dev: PCI device to power up
740 void pci_power_up(struct pci_dev *dev)
742 if (platform_pci_power_manageable(dev))
743 platform_pci_set_power_state(dev, PCI_D0);
745 pci_raw_set_power_state(dev, PCI_D0);
746 pci_update_current_state(dev, PCI_D0);
750 * pci_platform_power_transition - Use platform to change device power state
751 * @dev: PCI device to handle.
752 * @state: State to put the device into.
754 static int pci_platform_power_transition(struct pci_dev *dev, pci_power_t state)
758 if (platform_pci_power_manageable(dev)) {
759 error = platform_pci_set_power_state(dev, state);
761 pci_update_current_state(dev, state);
765 if (error && !dev->pm_cap) /* Fall back to PCI_D0 */
766 dev->current_state = PCI_D0;
772 * pci_wakeup - Wake up a PCI device
773 * @pci_dev: Device to handle.
774 * @ign: ignored parameter
776 static int pci_wakeup(struct pci_dev *pci_dev, void *ign)
778 pci_wakeup_event(pci_dev);
779 pm_request_resume(&pci_dev->dev);
784 * pci_wakeup_bus - Walk given bus and wake up devices on it
785 * @bus: Top bus of the subtree to walk.
787 static void pci_wakeup_bus(struct pci_bus *bus)
790 pci_walk_bus(bus, pci_wakeup, NULL);
794 * __pci_start_power_transition - Start power transition of a PCI device
795 * @dev: PCI device to handle.
796 * @state: State to put the device into.
798 static void __pci_start_power_transition(struct pci_dev *dev, pci_power_t state)
800 if (state == PCI_D0) {
801 pci_platform_power_transition(dev, PCI_D0);
803 * Mandatory power management transition delays, see
804 * PCI Express Base Specification Revision 2.0 Section
805 * 6.6.1: Conventional Reset. Do not delay for
806 * devices powered on/off by corresponding bridge,
807 * because have already delayed for the bridge.
809 if (dev->runtime_d3cold) {
810 msleep(dev->d3cold_delay);
812 * When powering on a bridge from D3cold, the
813 * whole hierarchy may be powered on into
814 * D0uninitialized state, resume them to give
815 * them a chance to suspend again
817 pci_wakeup_bus(dev->subordinate);
823 * __pci_dev_set_current_state - Set current state of a PCI device
824 * @dev: Device to handle
825 * @data: pointer to state to be set
827 static int __pci_dev_set_current_state(struct pci_dev *dev, void *data)
829 pci_power_t state = *(pci_power_t *)data;
831 dev->current_state = state;
836 * __pci_bus_set_current_state - Walk given bus and set current state of devices
837 * @bus: Top bus of the subtree to walk.
838 * @state: state to be set
840 static void __pci_bus_set_current_state(struct pci_bus *bus, pci_power_t state)
843 pci_walk_bus(bus, __pci_dev_set_current_state, &state);
847 * __pci_complete_power_transition - Complete power transition of a PCI device
848 * @dev: PCI device to handle.
849 * @state: State to put the device into.
851 * This function should not be called directly by device drivers.
853 int __pci_complete_power_transition(struct pci_dev *dev, pci_power_t state)
859 ret = pci_platform_power_transition(dev, state);
860 /* Power off the bridge may power off the whole hierarchy */
861 if (!ret && state == PCI_D3cold)
862 __pci_bus_set_current_state(dev->subordinate, PCI_D3cold);
865 EXPORT_SYMBOL_GPL(__pci_complete_power_transition);
868 * pci_set_power_state - Set the power state of a PCI device
869 * @dev: PCI device to handle.
870 * @state: PCI power state (D0, D1, D2, D3hot) to put the device into.
872 * Transition a device to a new power state, using the platform firmware and/or
873 * the device's PCI PM registers.
876 * -EINVAL if the requested state is invalid.
877 * -EIO if device does not support PCI PM or its PM capabilities register has a
878 * wrong version, or device doesn't support the requested state.
879 * 0 if device already is in the requested state.
880 * 0 if device's power state has been successfully changed.
882 int pci_set_power_state(struct pci_dev *dev, pci_power_t state)
886 /* bound the state we're entering */
887 if (state > PCI_D3cold)
889 else if (state < PCI_D0)
891 else if ((state == PCI_D1 || state == PCI_D2) && pci_no_d1d2(dev))
893 * If the device or the parent bridge do not support PCI PM,
894 * ignore the request if we're doing anything other than putting
895 * it into D0 (which would only happen on boot).
899 /* Check if we're already there */
900 if (dev->current_state == state)
903 __pci_start_power_transition(dev, state);
905 /* This device is quirked not to be put into D3, so
906 don't put it in D3 */
907 if (state >= PCI_D3hot && (dev->dev_flags & PCI_DEV_FLAGS_NO_D3))
911 * To put device in D3cold, we put device into D3hot in native
912 * way, then put device into D3cold with platform ops
914 error = pci_raw_set_power_state(dev, state > PCI_D3hot ?
917 if (!__pci_complete_power_transition(dev, state))
922 EXPORT_SYMBOL(pci_set_power_state);
925 * pci_choose_state - Choose the power state of a PCI device
926 * @dev: PCI device to be suspended
927 * @state: target sleep state for the whole system. This is the value
928 * that is passed to suspend() function.
930 * Returns PCI power state suitable for given device and given system
934 pci_power_t pci_choose_state(struct pci_dev *dev, pm_message_t state)
941 ret = platform_pci_choose_state(dev);
942 if (ret != PCI_POWER_ERROR)
945 switch (state.event) {
948 case PM_EVENT_FREEZE:
949 case PM_EVENT_PRETHAW:
950 /* REVISIT both freeze and pre-thaw "should" use D0 */
951 case PM_EVENT_SUSPEND:
952 case PM_EVENT_HIBERNATE:
955 dev_info(&dev->dev, "unrecognized suspend event %d\n",
961 EXPORT_SYMBOL(pci_choose_state);
963 #define PCI_EXP_SAVE_REGS 7
965 static struct pci_cap_saved_state *_pci_find_saved_cap(struct pci_dev *pci_dev,
966 u16 cap, bool extended)
968 struct pci_cap_saved_state *tmp;
970 hlist_for_each_entry(tmp, &pci_dev->saved_cap_space, next) {
971 if (tmp->cap.cap_extended == extended && tmp->cap.cap_nr == cap)
977 struct pci_cap_saved_state *pci_find_saved_cap(struct pci_dev *dev, char cap)
979 return _pci_find_saved_cap(dev, cap, false);
982 struct pci_cap_saved_state *pci_find_saved_ext_cap(struct pci_dev *dev, u16 cap)
984 return _pci_find_saved_cap(dev, cap, true);
987 static int pci_save_pcie_state(struct pci_dev *dev)
990 struct pci_cap_saved_state *save_state;
993 if (!pci_is_pcie(dev))
996 save_state = pci_find_saved_cap(dev, PCI_CAP_ID_EXP);
998 dev_err(&dev->dev, "buffer not found in %s\n", __func__);
1002 cap = (u16 *)&save_state->cap.data[0];
1003 pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &cap[i++]);
1004 pcie_capability_read_word(dev, PCI_EXP_LNKCTL, &cap[i++]);
1005 pcie_capability_read_word(dev, PCI_EXP_SLTCTL, &cap[i++]);
1006 pcie_capability_read_word(dev, PCI_EXP_RTCTL, &cap[i++]);
1007 pcie_capability_read_word(dev, PCI_EXP_DEVCTL2, &cap[i++]);
1008 pcie_capability_read_word(dev, PCI_EXP_LNKCTL2, &cap[i++]);
1009 pcie_capability_read_word(dev, PCI_EXP_SLTCTL2, &cap[i++]);
1014 static void pci_restore_pcie_state(struct pci_dev *dev)
1017 struct pci_cap_saved_state *save_state;
1020 save_state = pci_find_saved_cap(dev, PCI_CAP_ID_EXP);
1024 cap = (u16 *)&save_state->cap.data[0];
1025 pcie_capability_write_word(dev, PCI_EXP_DEVCTL, cap[i++]);
1026 pcie_capability_write_word(dev, PCI_EXP_LNKCTL, cap[i++]);
1027 pcie_capability_write_word(dev, PCI_EXP_SLTCTL, cap[i++]);
1028 pcie_capability_write_word(dev, PCI_EXP_RTCTL, cap[i++]);
1029 pcie_capability_write_word(dev, PCI_EXP_DEVCTL2, cap[i++]);
1030 pcie_capability_write_word(dev, PCI_EXP_LNKCTL2, cap[i++]);
1031 pcie_capability_write_word(dev, PCI_EXP_SLTCTL2, cap[i++]);
1035 static int pci_save_pcix_state(struct pci_dev *dev)
1038 struct pci_cap_saved_state *save_state;
1040 pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1044 save_state = pci_find_saved_cap(dev, PCI_CAP_ID_PCIX);
1046 dev_err(&dev->dev, "buffer not found in %s\n", __func__);
1050 pci_read_config_word(dev, pos + PCI_X_CMD,
1051 (u16 *)save_state->cap.data);
1056 static void pci_restore_pcix_state(struct pci_dev *dev)
1059 struct pci_cap_saved_state *save_state;
1062 save_state = pci_find_saved_cap(dev, PCI_CAP_ID_PCIX);
1063 pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1064 if (!save_state || !pos)
1066 cap = (u16 *)&save_state->cap.data[0];
1068 pci_write_config_word(dev, pos + PCI_X_CMD, cap[i++]);
1073 * pci_save_state - save the PCI configuration space of a device before suspending
1074 * @dev: - PCI device that we're dealing with
1076 int pci_save_state(struct pci_dev *dev)
1079 /* XXX: 100% dword access ok here? */
1080 for (i = 0; i < 16; i++)
1081 pci_read_config_dword(dev, i * 4, &dev->saved_config_space[i]);
1082 dev->state_saved = true;
1084 i = pci_save_pcie_state(dev);
1088 i = pci_save_pcix_state(dev);
1092 return pci_save_vc_state(dev);
1094 EXPORT_SYMBOL(pci_save_state);
1096 static void pci_restore_config_dword(struct pci_dev *pdev, int offset,
1097 u32 saved_val, int retry)
1101 pci_read_config_dword(pdev, offset, &val);
1102 if (val == saved_val)
1106 dev_dbg(&pdev->dev, "restoring config space at offset %#x (was %#x, writing %#x)\n",
1107 offset, val, saved_val);
1108 pci_write_config_dword(pdev, offset, saved_val);
1112 pci_read_config_dword(pdev, offset, &val);
1113 if (val == saved_val)
1120 static void pci_restore_config_space_range(struct pci_dev *pdev,
1121 int start, int end, int retry)
1125 for (index = end; index >= start; index--)
1126 pci_restore_config_dword(pdev, 4 * index,
1127 pdev->saved_config_space[index],
1131 static void pci_restore_config_space(struct pci_dev *pdev)
1133 if (pdev->hdr_type == PCI_HEADER_TYPE_NORMAL) {
1134 pci_restore_config_space_range(pdev, 10, 15, 0);
1135 /* Restore BARs before the command register. */
1136 pci_restore_config_space_range(pdev, 4, 9, 10);
1137 pci_restore_config_space_range(pdev, 0, 3, 0);
1139 pci_restore_config_space_range(pdev, 0, 15, 0);
1144 * pci_restore_state - Restore the saved state of a PCI device
1145 * @dev: - PCI device that we're dealing with
1147 void pci_restore_state(struct pci_dev *dev)
1149 if (!dev->state_saved)
1152 /* PCI Express register must be restored first */
1153 pci_restore_pcie_state(dev);
1154 pci_restore_ats_state(dev);
1155 pci_restore_vc_state(dev);
1157 pci_cleanup_aer_error_status_regs(dev);
1159 pci_restore_config_space(dev);
1161 pci_restore_pcix_state(dev);
1162 pci_restore_msi_state(dev);
1164 /* Restore ACS and IOV configuration state */
1165 pci_enable_acs(dev);
1166 pci_restore_iov_state(dev);
1168 dev->state_saved = false;
1170 EXPORT_SYMBOL(pci_restore_state);
1172 struct pci_saved_state {
1173 u32 config_space[16];
1174 struct pci_cap_saved_data cap[0];
1178 * pci_store_saved_state - Allocate and return an opaque struct containing
1179 * the device saved state.
1180 * @dev: PCI device that we're dealing with
1182 * Return NULL if no state or error.
1184 struct pci_saved_state *pci_store_saved_state(struct pci_dev *dev)
1186 struct pci_saved_state *state;
1187 struct pci_cap_saved_state *tmp;
1188 struct pci_cap_saved_data *cap;
1191 if (!dev->state_saved)
1194 size = sizeof(*state) + sizeof(struct pci_cap_saved_data);
1196 hlist_for_each_entry(tmp, &dev->saved_cap_space, next)
1197 size += sizeof(struct pci_cap_saved_data) + tmp->cap.size;
1199 state = kzalloc(size, GFP_KERNEL);
1203 memcpy(state->config_space, dev->saved_config_space,
1204 sizeof(state->config_space));
1207 hlist_for_each_entry(tmp, &dev->saved_cap_space, next) {
1208 size_t len = sizeof(struct pci_cap_saved_data) + tmp->cap.size;
1209 memcpy(cap, &tmp->cap, len);
1210 cap = (struct pci_cap_saved_data *)((u8 *)cap + len);
1212 /* Empty cap_save terminates list */
1216 EXPORT_SYMBOL_GPL(pci_store_saved_state);
1219 * pci_load_saved_state - Reload the provided save state into struct pci_dev.
1220 * @dev: PCI device that we're dealing with
1221 * @state: Saved state returned from pci_store_saved_state()
1223 int pci_load_saved_state(struct pci_dev *dev,
1224 struct pci_saved_state *state)
1226 struct pci_cap_saved_data *cap;
1228 dev->state_saved = false;
1233 memcpy(dev->saved_config_space, state->config_space,
1234 sizeof(state->config_space));
1238 struct pci_cap_saved_state *tmp;
1240 tmp = _pci_find_saved_cap(dev, cap->cap_nr, cap->cap_extended);
1241 if (!tmp || tmp->cap.size != cap->size)
1244 memcpy(tmp->cap.data, cap->data, tmp->cap.size);
1245 cap = (struct pci_cap_saved_data *)((u8 *)cap +
1246 sizeof(struct pci_cap_saved_data) + cap->size);
1249 dev->state_saved = true;
1252 EXPORT_SYMBOL_GPL(pci_load_saved_state);
1255 * pci_load_and_free_saved_state - Reload the save state pointed to by state,
1256 * and free the memory allocated for it.
1257 * @dev: PCI device that we're dealing with
1258 * @state: Pointer to saved state returned from pci_store_saved_state()
1260 int pci_load_and_free_saved_state(struct pci_dev *dev,
1261 struct pci_saved_state **state)
1263 int ret = pci_load_saved_state(dev, *state);
1268 EXPORT_SYMBOL_GPL(pci_load_and_free_saved_state);
1270 int __weak pcibios_enable_device(struct pci_dev *dev, int bars)
1272 return pci_enable_resources(dev, bars);
1275 static int do_pci_enable_device(struct pci_dev *dev, int bars)
1278 struct pci_dev *bridge;
1282 err = pci_set_power_state(dev, PCI_D0);
1283 if (err < 0 && err != -EIO)
1286 bridge = pci_upstream_bridge(dev);
1288 pcie_aspm_powersave_config_link(bridge);
1290 err = pcibios_enable_device(dev, bars);
1293 pci_fixup_device(pci_fixup_enable, dev);
1295 if (dev->msi_enabled || dev->msix_enabled)
1298 pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &pin);
1300 pci_read_config_word(dev, PCI_COMMAND, &cmd);
1301 if (cmd & PCI_COMMAND_INTX_DISABLE)
1302 pci_write_config_word(dev, PCI_COMMAND,
1303 cmd & ~PCI_COMMAND_INTX_DISABLE);
1310 * pci_reenable_device - Resume abandoned device
1311 * @dev: PCI device to be resumed
1313 * Note this function is a backend of pci_default_resume and is not supposed
1314 * to be called by normal code, write proper resume handler and use it instead.
1316 int pci_reenable_device(struct pci_dev *dev)
1318 if (pci_is_enabled(dev))
1319 return do_pci_enable_device(dev, (1 << PCI_NUM_RESOURCES) - 1);
1322 EXPORT_SYMBOL(pci_reenable_device);
1324 static void pci_enable_bridge(struct pci_dev *dev)
1326 struct pci_dev *bridge;
1329 bridge = pci_upstream_bridge(dev);
1331 pci_enable_bridge(bridge);
1333 if (pci_is_enabled(dev)) {
1334 if (!dev->is_busmaster)
1335 pci_set_master(dev);
1339 retval = pci_enable_device(dev);
1341 dev_err(&dev->dev, "Error enabling bridge (%d), continuing\n",
1343 pci_set_master(dev);
1346 static int pci_enable_device_flags(struct pci_dev *dev, unsigned long flags)
1348 struct pci_dev *bridge;
1353 * Power state could be unknown at this point, either due to a fresh
1354 * boot or a device removal call. So get the current power state
1355 * so that things like MSI message writing will behave as expected
1356 * (e.g. if the device really is in D0 at enable time).
1360 pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
1361 dev->current_state = (pmcsr & PCI_PM_CTRL_STATE_MASK);
1364 if (atomic_inc_return(&dev->enable_cnt) > 1)
1365 return 0; /* already enabled */
1367 bridge = pci_upstream_bridge(dev);
1369 pci_enable_bridge(bridge);
1371 /* only skip sriov related */
1372 for (i = 0; i <= PCI_ROM_RESOURCE; i++)
1373 if (dev->resource[i].flags & flags)
1375 for (i = PCI_BRIDGE_RESOURCES; i < DEVICE_COUNT_RESOURCE; i++)
1376 if (dev->resource[i].flags & flags)
1379 err = do_pci_enable_device(dev, bars);
1381 atomic_dec(&dev->enable_cnt);
1386 * pci_enable_device_io - Initialize a device for use with IO space
1387 * @dev: PCI device to be initialized
1389 * Initialize device before it's used by a driver. Ask low-level code
1390 * to enable I/O resources. Wake up the device if it was suspended.
1391 * Beware, this function can fail.
1393 int pci_enable_device_io(struct pci_dev *dev)
1395 return pci_enable_device_flags(dev, IORESOURCE_IO);
1397 EXPORT_SYMBOL(pci_enable_device_io);
1400 * pci_enable_device_mem - Initialize a device for use with Memory space
1401 * @dev: PCI device to be initialized
1403 * Initialize device before it's used by a driver. Ask low-level code
1404 * to enable Memory resources. Wake up the device if it was suspended.
1405 * Beware, this function can fail.
1407 int pci_enable_device_mem(struct pci_dev *dev)
1409 return pci_enable_device_flags(dev, IORESOURCE_MEM);
1411 EXPORT_SYMBOL(pci_enable_device_mem);
1414 * pci_enable_device - Initialize device before it's used by a driver.
1415 * @dev: PCI device to be initialized
1417 * Initialize device before it's used by a driver. Ask low-level code
1418 * to enable I/O and memory. Wake up the device if it was suspended.
1419 * Beware, this function can fail.
1421 * Note we don't actually enable the device many times if we call
1422 * this function repeatedly (we just increment the count).
1424 int pci_enable_device(struct pci_dev *dev)
1426 return pci_enable_device_flags(dev, IORESOURCE_MEM | IORESOURCE_IO);
1428 EXPORT_SYMBOL(pci_enable_device);
1431 * Managed PCI resources. This manages device on/off, intx/msi/msix
1432 * on/off and BAR regions. pci_dev itself records msi/msix status, so
1433 * there's no need to track it separately. pci_devres is initialized
1434 * when a device is enabled using managed PCI device enable interface.
1437 unsigned int enabled:1;
1438 unsigned int pinned:1;
1439 unsigned int orig_intx:1;
1440 unsigned int restore_intx:1;
1444 static void pcim_release(struct device *gendev, void *res)
1446 struct pci_dev *dev = to_pci_dev(gendev);
1447 struct pci_devres *this = res;
1450 if (dev->msi_enabled)
1451 pci_disable_msi(dev);
1452 if (dev->msix_enabled)
1453 pci_disable_msix(dev);
1455 for (i = 0; i < DEVICE_COUNT_RESOURCE; i++)
1456 if (this->region_mask & (1 << i))
1457 pci_release_region(dev, i);
1459 if (this->restore_intx)
1460 pci_intx(dev, this->orig_intx);
1462 if (this->enabled && !this->pinned)
1463 pci_disable_device(dev);
1466 static struct pci_devres *get_pci_dr(struct pci_dev *pdev)
1468 struct pci_devres *dr, *new_dr;
1470 dr = devres_find(&pdev->dev, pcim_release, NULL, NULL);
1474 new_dr = devres_alloc(pcim_release, sizeof(*new_dr), GFP_KERNEL);
1477 return devres_get(&pdev->dev, new_dr, NULL, NULL);
1480 static struct pci_devres *find_pci_dr(struct pci_dev *pdev)
1482 if (pci_is_managed(pdev))
1483 return devres_find(&pdev->dev, pcim_release, NULL, NULL);
1488 * pcim_enable_device - Managed pci_enable_device()
1489 * @pdev: PCI device to be initialized
1491 * Managed pci_enable_device().
1493 int pcim_enable_device(struct pci_dev *pdev)
1495 struct pci_devres *dr;
1498 dr = get_pci_dr(pdev);
1504 rc = pci_enable_device(pdev);
1506 pdev->is_managed = 1;
1511 EXPORT_SYMBOL(pcim_enable_device);
1514 * pcim_pin_device - Pin managed PCI device
1515 * @pdev: PCI device to pin
1517 * Pin managed PCI device @pdev. Pinned device won't be disabled on
1518 * driver detach. @pdev must have been enabled with
1519 * pcim_enable_device().
1521 void pcim_pin_device(struct pci_dev *pdev)
1523 struct pci_devres *dr;
1525 dr = find_pci_dr(pdev);
1526 WARN_ON(!dr || !dr->enabled);
1530 EXPORT_SYMBOL(pcim_pin_device);
1533 * pcibios_add_device - provide arch specific hooks when adding device dev
1534 * @dev: the PCI device being added
1536 * Permits the platform to provide architecture specific functionality when
1537 * devices are added. This is the default implementation. Architecture
1538 * implementations can override this.
1540 int __weak pcibios_add_device(struct pci_dev *dev)
1546 * pcibios_release_device - provide arch specific hooks when releasing device dev
1547 * @dev: the PCI device being released
1549 * Permits the platform to provide architecture specific functionality when
1550 * devices are released. This is the default implementation. Architecture
1551 * implementations can override this.
1553 void __weak pcibios_release_device(struct pci_dev *dev) {}
1556 * pcibios_disable_device - disable arch specific PCI resources for device dev
1557 * @dev: the PCI device to disable
1559 * Disables architecture specific PCI resources for the device. This
1560 * is the default implementation. Architecture implementations can
1563 void __weak pcibios_disable_device(struct pci_dev *dev) {}
1566 * pcibios_penalize_isa_irq - penalize an ISA IRQ
1567 * @irq: ISA IRQ to penalize
1568 * @active: IRQ active or not
1570 * Permits the platform to provide architecture-specific functionality when
1571 * penalizing ISA IRQs. This is the default implementation. Architecture
1572 * implementations can override this.
1574 void __weak pcibios_penalize_isa_irq(int irq, int active) {}
1576 static void do_pci_disable_device(struct pci_dev *dev)
1580 pci_read_config_word(dev, PCI_COMMAND, &pci_command);
1581 if (pci_command & PCI_COMMAND_MASTER) {
1582 pci_command &= ~PCI_COMMAND_MASTER;
1583 pci_write_config_word(dev, PCI_COMMAND, pci_command);
1586 pcibios_disable_device(dev);
1590 * pci_disable_enabled_device - Disable device without updating enable_cnt
1591 * @dev: PCI device to disable
1593 * NOTE: This function is a backend of PCI power management routines and is
1594 * not supposed to be called drivers.
1596 void pci_disable_enabled_device(struct pci_dev *dev)
1598 if (pci_is_enabled(dev))
1599 do_pci_disable_device(dev);
1603 * pci_disable_device - Disable PCI device after use
1604 * @dev: PCI device to be disabled
1606 * Signal to the system that the PCI device is not in use by the system
1607 * anymore. This only involves disabling PCI bus-mastering, if active.
1609 * Note we don't actually disable the device until all callers of
1610 * pci_enable_device() have called pci_disable_device().
1612 void pci_disable_device(struct pci_dev *dev)
1614 struct pci_devres *dr;
1616 dr = find_pci_dr(dev);
1620 dev_WARN_ONCE(&dev->dev, atomic_read(&dev->enable_cnt) <= 0,
1621 "disabling already-disabled device");
1623 if (atomic_dec_return(&dev->enable_cnt) != 0)
1626 do_pci_disable_device(dev);
1628 dev->is_busmaster = 0;
1630 EXPORT_SYMBOL(pci_disable_device);
1633 * pcibios_set_pcie_reset_state - set reset state for device dev
1634 * @dev: the PCIe device reset
1635 * @state: Reset state to enter into
1638 * Sets the PCIe reset state for the device. This is the default
1639 * implementation. Architecture implementations can override this.
1641 int __weak pcibios_set_pcie_reset_state(struct pci_dev *dev,
1642 enum pcie_reset_state state)
1648 * pci_set_pcie_reset_state - set reset state for device dev
1649 * @dev: the PCIe device reset
1650 * @state: Reset state to enter into
1653 * Sets the PCI reset state for the device.
1655 int pci_set_pcie_reset_state(struct pci_dev *dev, enum pcie_reset_state state)
1657 return pcibios_set_pcie_reset_state(dev, state);
1659 EXPORT_SYMBOL_GPL(pci_set_pcie_reset_state);
1662 * pci_check_pme_status - Check if given device has generated PME.
1663 * @dev: Device to check.
1665 * Check the PME status of the device and if set, clear it and clear PME enable
1666 * (if set). Return 'true' if PME status and PME enable were both set or
1667 * 'false' otherwise.
1669 bool pci_check_pme_status(struct pci_dev *dev)
1678 pmcsr_pos = dev->pm_cap + PCI_PM_CTRL;
1679 pci_read_config_word(dev, pmcsr_pos, &pmcsr);
1680 if (!(pmcsr & PCI_PM_CTRL_PME_STATUS))
1683 /* Clear PME status. */
1684 pmcsr |= PCI_PM_CTRL_PME_STATUS;
1685 if (pmcsr & PCI_PM_CTRL_PME_ENABLE) {
1686 /* Disable PME to avoid interrupt flood. */
1687 pmcsr &= ~PCI_PM_CTRL_PME_ENABLE;
1691 pci_write_config_word(dev, pmcsr_pos, pmcsr);
1697 * pci_pme_wakeup - Wake up a PCI device if its PME Status bit is set.
1698 * @dev: Device to handle.
1699 * @pme_poll_reset: Whether or not to reset the device's pme_poll flag.
1701 * Check if @dev has generated PME and queue a resume request for it in that
1704 static int pci_pme_wakeup(struct pci_dev *dev, void *pme_poll_reset)
1706 if (pme_poll_reset && dev->pme_poll)
1707 dev->pme_poll = false;
1709 if (pci_check_pme_status(dev)) {
1710 pci_wakeup_event(dev);
1711 pm_request_resume(&dev->dev);
1717 * pci_pme_wakeup_bus - Walk given bus and wake up devices on it, if necessary.
1718 * @bus: Top bus of the subtree to walk.
1720 void pci_pme_wakeup_bus(struct pci_bus *bus)
1723 pci_walk_bus(bus, pci_pme_wakeup, (void *)true);
1728 * pci_pme_capable - check the capability of PCI device to generate PME#
1729 * @dev: PCI device to handle.
1730 * @state: PCI state from which device will issue PME#.
1732 bool pci_pme_capable(struct pci_dev *dev, pci_power_t state)
1737 return !!(dev->pme_support & (1 << state));
1739 EXPORT_SYMBOL(pci_pme_capable);
1741 static void pci_pme_list_scan(struct work_struct *work)
1743 struct pci_pme_device *pme_dev, *n;
1745 mutex_lock(&pci_pme_list_mutex);
1746 list_for_each_entry_safe(pme_dev, n, &pci_pme_list, list) {
1747 if (pme_dev->dev->pme_poll) {
1748 struct pci_dev *bridge;
1750 bridge = pme_dev->dev->bus->self;
1752 * If bridge is in low power state, the
1753 * configuration space of subordinate devices
1754 * may be not accessible
1756 if (bridge && bridge->current_state != PCI_D0)
1758 pci_pme_wakeup(pme_dev->dev, NULL);
1760 list_del(&pme_dev->list);
1764 if (!list_empty(&pci_pme_list))
1765 schedule_delayed_work(&pci_pme_work,
1766 msecs_to_jiffies(PME_TIMEOUT));
1767 mutex_unlock(&pci_pme_list_mutex);
1770 static void __pci_pme_active(struct pci_dev *dev, bool enable)
1774 if (!dev->pme_support)
1777 pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
1778 /* Clear PME_Status by writing 1 to it and enable PME# */
1779 pmcsr |= PCI_PM_CTRL_PME_STATUS | PCI_PM_CTRL_PME_ENABLE;
1781 pmcsr &= ~PCI_PM_CTRL_PME_ENABLE;
1783 pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, pmcsr);
1787 * pci_pme_active - enable or disable PCI device's PME# function
1788 * @dev: PCI device to handle.
1789 * @enable: 'true' to enable PME# generation; 'false' to disable it.
1791 * The caller must verify that the device is capable of generating PME# before
1792 * calling this function with @enable equal to 'true'.
1794 void pci_pme_active(struct pci_dev *dev, bool enable)
1796 __pci_pme_active(dev, enable);
1799 * PCI (as opposed to PCIe) PME requires that the device have
1800 * its PME# line hooked up correctly. Not all hardware vendors
1801 * do this, so the PME never gets delivered and the device
1802 * remains asleep. The easiest way around this is to
1803 * periodically walk the list of suspended devices and check
1804 * whether any have their PME flag set. The assumption is that
1805 * we'll wake up often enough anyway that this won't be a huge
1806 * hit, and the power savings from the devices will still be a
1809 * Although PCIe uses in-band PME message instead of PME# line
1810 * to report PME, PME does not work for some PCIe devices in
1811 * reality. For example, there are devices that set their PME
1812 * status bits, but don't really bother to send a PME message;
1813 * there are PCI Express Root Ports that don't bother to
1814 * trigger interrupts when they receive PME messages from the
1815 * devices below. So PME poll is used for PCIe devices too.
1818 if (dev->pme_poll) {
1819 struct pci_pme_device *pme_dev;
1821 pme_dev = kmalloc(sizeof(struct pci_pme_device),
1824 dev_warn(&dev->dev, "can't enable PME#\n");
1828 mutex_lock(&pci_pme_list_mutex);
1829 list_add(&pme_dev->list, &pci_pme_list);
1830 if (list_is_singular(&pci_pme_list))
1831 schedule_delayed_work(&pci_pme_work,
1832 msecs_to_jiffies(PME_TIMEOUT));
1833 mutex_unlock(&pci_pme_list_mutex);
1835 mutex_lock(&pci_pme_list_mutex);
1836 list_for_each_entry(pme_dev, &pci_pme_list, list) {
1837 if (pme_dev->dev == dev) {
1838 list_del(&pme_dev->list);
1843 mutex_unlock(&pci_pme_list_mutex);
1847 dev_dbg(&dev->dev, "PME# %s\n", enable ? "enabled" : "disabled");
1849 EXPORT_SYMBOL(pci_pme_active);
1852 * __pci_enable_wake - enable PCI device as wakeup event source
1853 * @dev: PCI device affected
1854 * @state: PCI state from which device will issue wakeup events
1855 * @runtime: True if the events are to be generated at run time
1856 * @enable: True to enable event generation; false to disable
1858 * This enables the device as a wakeup event source, or disables it.
1859 * When such events involves platform-specific hooks, those hooks are
1860 * called automatically by this routine.
1862 * Devices with legacy power management (no standard PCI PM capabilities)
1863 * always require such platform hooks.
1866 * 0 is returned on success
1867 * -EINVAL is returned if device is not supposed to wake up the system
1868 * Error code depending on the platform is returned if both the platform and
1869 * the native mechanism fail to enable the generation of wake-up events
1871 int __pci_enable_wake(struct pci_dev *dev, pci_power_t state,
1872 bool runtime, bool enable)
1876 if (enable && !runtime && !device_may_wakeup(&dev->dev))
1879 /* Don't do the same thing twice in a row for one device. */
1880 if (!!enable == !!dev->wakeup_prepared)
1884 * According to "PCI System Architecture" 4th ed. by Tom Shanley & Don
1885 * Anderson we should be doing PME# wake enable followed by ACPI wake
1886 * enable. To disable wake-up we call the platform first, for symmetry.
1892 if (pci_pme_capable(dev, state))
1893 pci_pme_active(dev, true);
1896 error = runtime ? platform_pci_run_wake(dev, true) :
1897 platform_pci_sleep_wake(dev, true);
1901 dev->wakeup_prepared = true;
1904 platform_pci_run_wake(dev, false);
1906 platform_pci_sleep_wake(dev, false);
1907 pci_pme_active(dev, false);
1908 dev->wakeup_prepared = false;
1913 EXPORT_SYMBOL(__pci_enable_wake);
1916 * pci_wake_from_d3 - enable/disable device to wake up from D3_hot or D3_cold
1917 * @dev: PCI device to prepare
1918 * @enable: True to enable wake-up event generation; false to disable
1920 * Many drivers want the device to wake up the system from D3_hot or D3_cold
1921 * and this function allows them to set that up cleanly - pci_enable_wake()
1922 * should not be called twice in a row to enable wake-up due to PCI PM vs ACPI
1923 * ordering constraints.
1925 * This function only returns error code if the device is not capable of
1926 * generating PME# from both D3_hot and D3_cold, and the platform is unable to
1927 * enable wake-up power for it.
1929 int pci_wake_from_d3(struct pci_dev *dev, bool enable)
1931 return pci_pme_capable(dev, PCI_D3cold) ?
1932 pci_enable_wake(dev, PCI_D3cold, enable) :
1933 pci_enable_wake(dev, PCI_D3hot, enable);
1935 EXPORT_SYMBOL(pci_wake_from_d3);
1938 * pci_target_state - find an appropriate low power state for a given PCI dev
1941 * Use underlying platform code to find a supported low power state for @dev.
1942 * If the platform can't manage @dev, return the deepest state from which it
1943 * can generate wake events, based on any available PME info.
1945 static pci_power_t pci_target_state(struct pci_dev *dev)
1947 pci_power_t target_state = PCI_D3hot;
1949 if (platform_pci_power_manageable(dev)) {
1951 * Call the platform to choose the target state of the device
1952 * and enable wake-up from this state if supported.
1954 pci_power_t state = platform_pci_choose_state(dev);
1957 case PCI_POWER_ERROR:
1962 if (pci_no_d1d2(dev))
1965 target_state = state;
1968 return target_state;
1972 target_state = PCI_D0;
1975 * If the device is in D3cold even though it's not power-manageable by
1976 * the platform, it may have been powered down by non-standard means.
1977 * Best to let it slumber.
1979 if (dev->current_state == PCI_D3cold)
1980 target_state = PCI_D3cold;
1982 if (device_may_wakeup(&dev->dev)) {
1984 * Find the deepest state from which the device can generate
1985 * wake-up events, make it the target state and enable device
1988 if (dev->pme_support) {
1990 && !(dev->pme_support & (1 << target_state)))
1995 return target_state;
1999 * pci_prepare_to_sleep - prepare PCI device for system-wide transition into a sleep state
2000 * @dev: Device to handle.
2002 * Choose the power state appropriate for the device depending on whether
2003 * it can wake up the system and/or is power manageable by the platform
2004 * (PCI_D3hot is the default) and put the device into that state.
2006 int pci_prepare_to_sleep(struct pci_dev *dev)
2008 pci_power_t target_state = pci_target_state(dev);
2011 if (target_state == PCI_POWER_ERROR)
2014 pci_enable_wake(dev, target_state, device_may_wakeup(&dev->dev));
2016 error = pci_set_power_state(dev, target_state);
2019 pci_enable_wake(dev, target_state, false);
2023 EXPORT_SYMBOL(pci_prepare_to_sleep);
2026 * pci_back_from_sleep - turn PCI device on during system-wide transition into working state
2027 * @dev: Device to handle.
2029 * Disable device's system wake-up capability and put it into D0.
2031 int pci_back_from_sleep(struct pci_dev *dev)
2033 pci_enable_wake(dev, PCI_D0, false);
2034 return pci_set_power_state(dev, PCI_D0);
2036 EXPORT_SYMBOL(pci_back_from_sleep);
2039 * pci_finish_runtime_suspend - Carry out PCI-specific part of runtime suspend.
2040 * @dev: PCI device being suspended.
2042 * Prepare @dev to generate wake-up events at run time and put it into a low
2045 int pci_finish_runtime_suspend(struct pci_dev *dev)
2047 pci_power_t target_state = pci_target_state(dev);
2050 if (target_state == PCI_POWER_ERROR)
2053 dev->runtime_d3cold = target_state == PCI_D3cold;
2055 __pci_enable_wake(dev, target_state, true, pci_dev_run_wake(dev));
2057 error = pci_set_power_state(dev, target_state);
2060 __pci_enable_wake(dev, target_state, true, false);
2061 dev->runtime_d3cold = false;
2068 * pci_dev_run_wake - Check if device can generate run-time wake-up events.
2069 * @dev: Device to check.
2071 * Return true if the device itself is capable of generating wake-up events
2072 * (through the platform or using the native PCIe PME) or if the device supports
2073 * PME and one of its upstream bridges can generate wake-up events.
2075 bool pci_dev_run_wake(struct pci_dev *dev)
2077 struct pci_bus *bus = dev->bus;
2079 if (device_run_wake(&dev->dev))
2082 if (!dev->pme_support)
2085 while (bus->parent) {
2086 struct pci_dev *bridge = bus->self;
2088 if (device_run_wake(&bridge->dev))
2094 /* We have reached the root bus. */
2096 return device_run_wake(bus->bridge);
2100 EXPORT_SYMBOL_GPL(pci_dev_run_wake);
2103 * pci_dev_keep_suspended - Check if the device can stay in the suspended state.
2104 * @pci_dev: Device to check.
2106 * Return 'true' if the device is runtime-suspended, it doesn't have to be
2107 * reconfigured due to wakeup settings difference between system and runtime
2108 * suspend and the current power state of it is suitable for the upcoming
2109 * (system) transition.
2111 * If the device is not configured for system wakeup, disable PME for it before
2112 * returning 'true' to prevent it from waking up the system unnecessarily.
2114 bool pci_dev_keep_suspended(struct pci_dev *pci_dev)
2116 struct device *dev = &pci_dev->dev;
2118 if (!pm_runtime_suspended(dev)
2119 || pci_target_state(pci_dev) != pci_dev->current_state
2120 || platform_pci_need_resume(pci_dev))
2124 * At this point the device is good to go unless it's been configured
2125 * to generate PME at the runtime suspend time, but it is not supposed
2126 * to wake up the system. In that case, simply disable PME for it
2127 * (it will have to be re-enabled on exit from system resume).
2129 * If the device's power state is D3cold and the platform check above
2130 * hasn't triggered, the device's configuration is suitable and we don't
2131 * need to manipulate it at all.
2133 spin_lock_irq(&dev->power.lock);
2135 if (pm_runtime_suspended(dev) && pci_dev->current_state < PCI_D3cold &&
2136 !device_may_wakeup(dev))
2137 __pci_pme_active(pci_dev, false);
2139 spin_unlock_irq(&dev->power.lock);
2144 * pci_dev_complete_resume - Finalize resume from system sleep for a device.
2145 * @pci_dev: Device to handle.
2147 * If the device is runtime suspended and wakeup-capable, enable PME for it as
2148 * it might have been disabled during the prepare phase of system suspend if
2149 * the device was not configured for system wakeup.
2151 void pci_dev_complete_resume(struct pci_dev *pci_dev)
2153 struct device *dev = &pci_dev->dev;
2155 if (!pci_dev_run_wake(pci_dev))
2158 spin_lock_irq(&dev->power.lock);
2160 if (pm_runtime_suspended(dev) && pci_dev->current_state < PCI_D3cold)
2161 __pci_pme_active(pci_dev, true);
2163 spin_unlock_irq(&dev->power.lock);
2166 void pci_config_pm_runtime_get(struct pci_dev *pdev)
2168 struct device *dev = &pdev->dev;
2169 struct device *parent = dev->parent;
2172 pm_runtime_get_sync(parent);
2173 pm_runtime_get_noresume(dev);
2175 * pdev->current_state is set to PCI_D3cold during suspending,
2176 * so wait until suspending completes
2178 pm_runtime_barrier(dev);
2180 * Only need to resume devices in D3cold, because config
2181 * registers are still accessible for devices suspended but
2184 if (pdev->current_state == PCI_D3cold)
2185 pm_runtime_resume(dev);
2188 void pci_config_pm_runtime_put(struct pci_dev *pdev)
2190 struct device *dev = &pdev->dev;
2191 struct device *parent = dev->parent;
2193 pm_runtime_put(dev);
2195 pm_runtime_put_sync(parent);
2199 * pci_bridge_d3_possible - Is it possible to put the bridge into D3
2200 * @bridge: Bridge to check
2202 * This function checks if it is possible to move the bridge to D3.
2203 * Currently we only allow D3 for recent enough PCIe ports.
2205 static bool pci_bridge_d3_possible(struct pci_dev *bridge)
2209 if (!pci_is_pcie(bridge))
2212 switch (pci_pcie_type(bridge)) {
2213 case PCI_EXP_TYPE_ROOT_PORT:
2214 case PCI_EXP_TYPE_UPSTREAM:
2215 case PCI_EXP_TYPE_DOWNSTREAM:
2216 if (pci_bridge_d3_disable)
2218 if (pci_bridge_d3_force)
2222 * It should be safe to put PCIe ports from 2015 or newer
2225 if (dmi_get_date(DMI_BIOS_DATE, &year, NULL, NULL) &&
2235 static int pci_dev_check_d3cold(struct pci_dev *dev, void *data)
2237 bool *d3cold_ok = data;
2241 * The device needs to be allowed to go D3cold and if it is wake
2242 * capable to do so from D3cold.
2244 no_d3cold = dev->no_d3cold || !dev->d3cold_allowed ||
2245 (device_may_wakeup(&dev->dev) && !pci_pme_capable(dev, PCI_D3cold)) ||
2246 !pci_power_manageable(dev);
2248 *d3cold_ok = !no_d3cold;
2254 * pci_bridge_d3_update - Update bridge D3 capabilities
2255 * @dev: PCI device which is changed
2256 * @remove: Is the device being removed
2258 * Update upstream bridge PM capabilities accordingly depending on if the
2259 * device PM configuration was changed or the device is being removed. The
2260 * change is also propagated upstream.
2262 static void pci_bridge_d3_update(struct pci_dev *dev, bool remove)
2264 struct pci_dev *bridge;
2265 bool d3cold_ok = true;
2267 bridge = pci_upstream_bridge(dev);
2268 if (!bridge || !pci_bridge_d3_possible(bridge))
2271 pci_dev_get(bridge);
2273 * If the device is removed we do not care about its D3cold
2277 pci_dev_check_d3cold(dev, &d3cold_ok);
2281 * We need to go through all children to find out if all of
2282 * them can still go to D3cold.
2284 pci_walk_bus(bridge->subordinate, pci_dev_check_d3cold,
2288 if (bridge->bridge_d3 != d3cold_ok) {
2289 bridge->bridge_d3 = d3cold_ok;
2290 /* Propagate change to upstream bridges */
2291 pci_bridge_d3_update(bridge, false);
2294 pci_dev_put(bridge);
2298 * pci_bridge_d3_device_changed - Update bridge D3 capabilities on change
2299 * @dev: PCI device that was changed
2301 * If a device is added or its PM configuration, such as is it allowed to
2302 * enter D3cold, is changed this function updates upstream bridge PM
2303 * capabilities accordingly.
2305 void pci_bridge_d3_device_changed(struct pci_dev *dev)
2307 pci_bridge_d3_update(dev, false);
2311 * pci_bridge_d3_device_removed - Update bridge D3 capabilities on remove
2312 * @dev: PCI device being removed
2314 * Function updates upstream bridge PM capabilities based on other devices
2315 * still left on the bus.
2317 void pci_bridge_d3_device_removed(struct pci_dev *dev)
2319 pci_bridge_d3_update(dev, true);
2323 * pci_d3cold_enable - Enable D3cold for device
2324 * @dev: PCI device to handle
2326 * This function can be used in drivers to enable D3cold from the device
2327 * they handle. It also updates upstream PCI bridge PM capabilities
2330 void pci_d3cold_enable(struct pci_dev *dev)
2332 if (dev->no_d3cold) {
2333 dev->no_d3cold = false;
2334 pci_bridge_d3_device_changed(dev);
2337 EXPORT_SYMBOL_GPL(pci_d3cold_enable);
2340 * pci_d3cold_disable - Disable D3cold for device
2341 * @dev: PCI device to handle
2343 * This function can be used in drivers to disable D3cold from the device
2344 * they handle. It also updates upstream PCI bridge PM capabilities
2347 void pci_d3cold_disable(struct pci_dev *dev)
2349 if (!dev->no_d3cold) {
2350 dev->no_d3cold = true;
2351 pci_bridge_d3_device_changed(dev);
2354 EXPORT_SYMBOL_GPL(pci_d3cold_disable);
2357 * pci_pm_init - Initialize PM functions of given PCI device
2358 * @dev: PCI device to handle.
2360 void pci_pm_init(struct pci_dev *dev)
2365 pm_runtime_forbid(&dev->dev);
2366 pm_runtime_set_active(&dev->dev);
2367 pm_runtime_enable(&dev->dev);
2368 device_enable_async_suspend(&dev->dev);
2369 dev->wakeup_prepared = false;
2372 dev->pme_support = 0;
2374 /* find PCI PM capability in list */
2375 pm = pci_find_capability(dev, PCI_CAP_ID_PM);
2378 /* Check device's ability to generate PME# */
2379 pci_read_config_word(dev, pm + PCI_PM_PMC, &pmc);
2381 if ((pmc & PCI_PM_CAP_VER_MASK) > 3) {
2382 dev_err(&dev->dev, "unsupported PM cap regs version (%u)\n",
2383 pmc & PCI_PM_CAP_VER_MASK);
2388 dev->d3_delay = PCI_PM_D3_WAIT;
2389 dev->d3cold_delay = PCI_PM_D3COLD_WAIT;
2390 dev->bridge_d3 = pci_bridge_d3_possible(dev);
2391 dev->d3cold_allowed = true;
2393 dev->d1_support = false;
2394 dev->d2_support = false;
2395 if (!pci_no_d1d2(dev)) {
2396 if (pmc & PCI_PM_CAP_D1)
2397 dev->d1_support = true;
2398 if (pmc & PCI_PM_CAP_D2)
2399 dev->d2_support = true;
2401 if (dev->d1_support || dev->d2_support)
2402 dev_printk(KERN_DEBUG, &dev->dev, "supports%s%s\n",
2403 dev->d1_support ? " D1" : "",
2404 dev->d2_support ? " D2" : "");
2407 pmc &= PCI_PM_CAP_PME_MASK;
2409 dev_printk(KERN_DEBUG, &dev->dev,
2410 "PME# supported from%s%s%s%s%s\n",
2411 (pmc & PCI_PM_CAP_PME_D0) ? " D0" : "",
2412 (pmc & PCI_PM_CAP_PME_D1) ? " D1" : "",
2413 (pmc & PCI_PM_CAP_PME_D2) ? " D2" : "",
2414 (pmc & PCI_PM_CAP_PME_D3) ? " D3hot" : "",
2415 (pmc & PCI_PM_CAP_PME_D3cold) ? " D3cold" : "");
2416 dev->pme_support = pmc >> PCI_PM_CAP_PME_SHIFT;
2417 dev->pme_poll = true;
2419 * Make device's PM flags reflect the wake-up capability, but
2420 * let the user space enable it to wake up the system as needed.
2422 device_set_wakeup_capable(&dev->dev, true);
2423 /* Disable the PME# generation functionality */
2424 pci_pme_active(dev, false);
2428 static unsigned long pci_ea_flags(struct pci_dev *dev, u8 prop)
2430 unsigned long flags = IORESOURCE_PCI_FIXED | IORESOURCE_PCI_EA_BEI;
2434 case PCI_EA_P_VF_MEM:
2435 flags |= IORESOURCE_MEM;
2437 case PCI_EA_P_MEM_PREFETCH:
2438 case PCI_EA_P_VF_MEM_PREFETCH:
2439 flags |= IORESOURCE_MEM | IORESOURCE_PREFETCH;
2442 flags |= IORESOURCE_IO;
2451 static struct resource *pci_ea_get_resource(struct pci_dev *dev, u8 bei,
2454 if (bei <= PCI_EA_BEI_BAR5 && prop <= PCI_EA_P_IO)
2455 return &dev->resource[bei];
2456 #ifdef CONFIG_PCI_IOV
2457 else if (bei >= PCI_EA_BEI_VF_BAR0 && bei <= PCI_EA_BEI_VF_BAR5 &&
2458 (prop == PCI_EA_P_VF_MEM || prop == PCI_EA_P_VF_MEM_PREFETCH))
2459 return &dev->resource[PCI_IOV_RESOURCES +
2460 bei - PCI_EA_BEI_VF_BAR0];
2462 else if (bei == PCI_EA_BEI_ROM)
2463 return &dev->resource[PCI_ROM_RESOURCE];
2468 /* Read an Enhanced Allocation (EA) entry */
2469 static int pci_ea_read(struct pci_dev *dev, int offset)
2471 struct resource *res;
2472 int ent_size, ent_offset = offset;
2473 resource_size_t start, end;
2474 unsigned long flags;
2475 u32 dw0, bei, base, max_offset;
2477 bool support_64 = (sizeof(resource_size_t) >= 8);
2479 pci_read_config_dword(dev, ent_offset, &dw0);
2482 /* Entry size field indicates DWORDs after 1st */
2483 ent_size = ((dw0 & PCI_EA_ES) + 1) << 2;
2485 if (!(dw0 & PCI_EA_ENABLE)) /* Entry not enabled */
2488 bei = (dw0 & PCI_EA_BEI) >> 4;
2489 prop = (dw0 & PCI_EA_PP) >> 8;
2492 * If the Property is in the reserved range, try the Secondary
2495 if (prop > PCI_EA_P_BRIDGE_IO && prop < PCI_EA_P_MEM_RESERVED)
2496 prop = (dw0 & PCI_EA_SP) >> 16;
2497 if (prop > PCI_EA_P_BRIDGE_IO)
2500 res = pci_ea_get_resource(dev, bei, prop);
2502 dev_err(&dev->dev, "Unsupported EA entry BEI: %u\n", bei);
2506 flags = pci_ea_flags(dev, prop);
2508 dev_err(&dev->dev, "Unsupported EA properties: %#x\n", prop);
2513 pci_read_config_dword(dev, ent_offset, &base);
2514 start = (base & PCI_EA_FIELD_MASK);
2517 /* Read MaxOffset */
2518 pci_read_config_dword(dev, ent_offset, &max_offset);
2521 /* Read Base MSBs (if 64-bit entry) */
2522 if (base & PCI_EA_IS_64) {
2525 pci_read_config_dword(dev, ent_offset, &base_upper);
2528 flags |= IORESOURCE_MEM_64;
2530 /* entry starts above 32-bit boundary, can't use */
2531 if (!support_64 && base_upper)
2535 start |= ((u64)base_upper << 32);
2538 end = start + (max_offset | 0x03);
2540 /* Read MaxOffset MSBs (if 64-bit entry) */
2541 if (max_offset & PCI_EA_IS_64) {
2542 u32 max_offset_upper;
2544 pci_read_config_dword(dev, ent_offset, &max_offset_upper);
2547 flags |= IORESOURCE_MEM_64;
2549 /* entry too big, can't use */
2550 if (!support_64 && max_offset_upper)
2554 end += ((u64)max_offset_upper << 32);
2558 dev_err(&dev->dev, "EA Entry crosses address boundary\n");
2562 if (ent_size != ent_offset - offset) {
2564 "EA Entry Size (%d) does not match length read (%d)\n",
2565 ent_size, ent_offset - offset);
2569 res->name = pci_name(dev);
2574 if (bei <= PCI_EA_BEI_BAR5)
2575 dev_printk(KERN_DEBUG, &dev->dev, "BAR %d: %pR (from Enhanced Allocation, properties %#02x)\n",
2577 else if (bei == PCI_EA_BEI_ROM)
2578 dev_printk(KERN_DEBUG, &dev->dev, "ROM: %pR (from Enhanced Allocation, properties %#02x)\n",
2580 else if (bei >= PCI_EA_BEI_VF_BAR0 && bei <= PCI_EA_BEI_VF_BAR5)
2581 dev_printk(KERN_DEBUG, &dev->dev, "VF BAR %d: %pR (from Enhanced Allocation, properties %#02x)\n",
2582 bei - PCI_EA_BEI_VF_BAR0, res, prop);
2584 dev_printk(KERN_DEBUG, &dev->dev, "BEI %d res: %pR (from Enhanced Allocation, properties %#02x)\n",
2588 return offset + ent_size;
2591 /* Enhanced Allocation Initialization */
2592 void pci_ea_init(struct pci_dev *dev)
2599 /* find PCI EA capability in list */
2600 ea = pci_find_capability(dev, PCI_CAP_ID_EA);
2604 /* determine the number of entries */
2605 pci_bus_read_config_byte(dev->bus, dev->devfn, ea + PCI_EA_NUM_ENT,
2607 num_ent &= PCI_EA_NUM_ENT_MASK;
2609 offset = ea + PCI_EA_FIRST_ENT;
2611 /* Skip DWORD 2 for type 1 functions */
2612 if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE)
2615 /* parse each EA entry */
2616 for (i = 0; i < num_ent; ++i)
2617 offset = pci_ea_read(dev, offset);
2620 static void pci_add_saved_cap(struct pci_dev *pci_dev,
2621 struct pci_cap_saved_state *new_cap)
2623 hlist_add_head(&new_cap->next, &pci_dev->saved_cap_space);
2627 * _pci_add_cap_save_buffer - allocate buffer for saving given
2628 * capability registers
2629 * @dev: the PCI device
2630 * @cap: the capability to allocate the buffer for
2631 * @extended: Standard or Extended capability ID
2632 * @size: requested size of the buffer
2634 static int _pci_add_cap_save_buffer(struct pci_dev *dev, u16 cap,
2635 bool extended, unsigned int size)
2638 struct pci_cap_saved_state *save_state;
2641 pos = pci_find_ext_capability(dev, cap);
2643 pos = pci_find_capability(dev, cap);
2648 save_state = kzalloc(sizeof(*save_state) + size, GFP_KERNEL);
2652 save_state->cap.cap_nr = cap;
2653 save_state->cap.cap_extended = extended;
2654 save_state->cap.size = size;
2655 pci_add_saved_cap(dev, save_state);
2660 int pci_add_cap_save_buffer(struct pci_dev *dev, char cap, unsigned int size)
2662 return _pci_add_cap_save_buffer(dev, cap, false, size);
2665 int pci_add_ext_cap_save_buffer(struct pci_dev *dev, u16 cap, unsigned int size)
2667 return _pci_add_cap_save_buffer(dev, cap, true, size);
2671 * pci_allocate_cap_save_buffers - allocate buffers for saving capabilities
2672 * @dev: the PCI device
2674 void pci_allocate_cap_save_buffers(struct pci_dev *dev)
2678 error = pci_add_cap_save_buffer(dev, PCI_CAP_ID_EXP,
2679 PCI_EXP_SAVE_REGS * sizeof(u16));
2682 "unable to preallocate PCI Express save buffer\n");
2684 error = pci_add_cap_save_buffer(dev, PCI_CAP_ID_PCIX, sizeof(u16));
2687 "unable to preallocate PCI-X save buffer\n");
2689 pci_allocate_vc_save_buffers(dev);
2692 void pci_free_cap_save_buffers(struct pci_dev *dev)
2694 struct pci_cap_saved_state *tmp;
2695 struct hlist_node *n;
2697 hlist_for_each_entry_safe(tmp, n, &dev->saved_cap_space, next)
2702 * pci_configure_ari - enable or disable ARI forwarding
2703 * @dev: the PCI device
2705 * If @dev and its upstream bridge both support ARI, enable ARI in the
2706 * bridge. Otherwise, disable ARI in the bridge.
2708 void pci_configure_ari(struct pci_dev *dev)
2711 struct pci_dev *bridge;
2713 if (pcie_ari_disabled || !pci_is_pcie(dev) || dev->devfn)
2716 bridge = dev->bus->self;
2720 pcie_capability_read_dword(bridge, PCI_EXP_DEVCAP2, &cap);
2721 if (!(cap & PCI_EXP_DEVCAP2_ARI))
2724 if (pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ARI)) {
2725 pcie_capability_set_word(bridge, PCI_EXP_DEVCTL2,
2726 PCI_EXP_DEVCTL2_ARI);
2727 bridge->ari_enabled = 1;
2729 pcie_capability_clear_word(bridge, PCI_EXP_DEVCTL2,
2730 PCI_EXP_DEVCTL2_ARI);
2731 bridge->ari_enabled = 0;
2735 static int pci_acs_enable;
2738 * pci_request_acs - ask for ACS to be enabled if supported
2740 void pci_request_acs(void)
2746 * pci_std_enable_acs - enable ACS on devices using standard ACS capabilites
2747 * @dev: the PCI device
2749 static void pci_std_enable_acs(struct pci_dev *dev)
2755 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ACS);
2759 pci_read_config_word(dev, pos + PCI_ACS_CAP, &cap);
2760 pci_read_config_word(dev, pos + PCI_ACS_CTRL, &ctrl);
2762 /* Source Validation */
2763 ctrl |= (cap & PCI_ACS_SV);
2765 /* P2P Request Redirect */
2766 ctrl |= (cap & PCI_ACS_RR);
2768 /* P2P Completion Redirect */
2769 ctrl |= (cap & PCI_ACS_CR);
2771 /* Upstream Forwarding */
2772 ctrl |= (cap & PCI_ACS_UF);
2774 pci_write_config_word(dev, pos + PCI_ACS_CTRL, ctrl);
2778 * pci_enable_acs - enable ACS if hardware support it
2779 * @dev: the PCI device
2781 void pci_enable_acs(struct pci_dev *dev)
2783 if (!pci_acs_enable)
2786 if (!pci_dev_specific_enable_acs(dev))
2789 pci_std_enable_acs(dev);
2792 static bool pci_acs_flags_enabled(struct pci_dev *pdev, u16 acs_flags)
2797 pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ACS);
2802 * Except for egress control, capabilities are either required
2803 * or only required if controllable. Features missing from the
2804 * capability field can therefore be assumed as hard-wired enabled.
2806 pci_read_config_word(pdev, pos + PCI_ACS_CAP, &cap);
2807 acs_flags &= (cap | PCI_ACS_EC);
2809 pci_read_config_word(pdev, pos + PCI_ACS_CTRL, &ctrl);
2810 return (ctrl & acs_flags) == acs_flags;
2814 * pci_acs_enabled - test ACS against required flags for a given device
2815 * @pdev: device to test
2816 * @acs_flags: required PCI ACS flags
2818 * Return true if the device supports the provided flags. Automatically
2819 * filters out flags that are not implemented on multifunction devices.
2821 * Note that this interface checks the effective ACS capabilities of the
2822 * device rather than the actual capabilities. For instance, most single
2823 * function endpoints are not required to support ACS because they have no
2824 * opportunity for peer-to-peer access. We therefore return 'true'
2825 * regardless of whether the device exposes an ACS capability. This makes
2826 * it much easier for callers of this function to ignore the actual type
2827 * or topology of the device when testing ACS support.
2829 bool pci_acs_enabled(struct pci_dev *pdev, u16 acs_flags)
2833 ret = pci_dev_specific_acs_enabled(pdev, acs_flags);
2838 * Conventional PCI and PCI-X devices never support ACS, either
2839 * effectively or actually. The shared bus topology implies that
2840 * any device on the bus can receive or snoop DMA.
2842 if (!pci_is_pcie(pdev))
2845 switch (pci_pcie_type(pdev)) {
2847 * PCI/X-to-PCIe bridges are not specifically mentioned by the spec,
2848 * but since their primary interface is PCI/X, we conservatively
2849 * handle them as we would a non-PCIe device.
2851 case PCI_EXP_TYPE_PCIE_BRIDGE:
2853 * PCIe 3.0, 6.12.1 excludes ACS on these devices. "ACS is never
2854 * applicable... must never implement an ACS Extended Capability...".
2855 * This seems arbitrary, but we take a conservative interpretation
2856 * of this statement.
2858 case PCI_EXP_TYPE_PCI_BRIDGE:
2859 case PCI_EXP_TYPE_RC_EC:
2862 * PCIe 3.0, 6.12.1.1 specifies that downstream and root ports should
2863 * implement ACS in order to indicate their peer-to-peer capabilities,
2864 * regardless of whether they are single- or multi-function devices.
2866 case PCI_EXP_TYPE_DOWNSTREAM:
2867 case PCI_EXP_TYPE_ROOT_PORT:
2868 return pci_acs_flags_enabled(pdev, acs_flags);
2870 * PCIe 3.0, 6.12.1.2 specifies ACS capabilities that should be
2871 * implemented by the remaining PCIe types to indicate peer-to-peer
2872 * capabilities, but only when they are part of a multifunction
2873 * device. The footnote for section 6.12 indicates the specific
2874 * PCIe types included here.
2876 case PCI_EXP_TYPE_ENDPOINT:
2877 case PCI_EXP_TYPE_UPSTREAM:
2878 case PCI_EXP_TYPE_LEG_END:
2879 case PCI_EXP_TYPE_RC_END:
2880 if (!pdev->multifunction)
2883 return pci_acs_flags_enabled(pdev, acs_flags);
2887 * PCIe 3.0, 6.12.1.3 specifies no ACS capabilities are applicable
2888 * to single function devices with the exception of downstream ports.
2894 * pci_acs_path_enable - test ACS flags from start to end in a hierarchy
2895 * @start: starting downstream device
2896 * @end: ending upstream device or NULL to search to the root bus
2897 * @acs_flags: required flags
2899 * Walk up a device tree from start to end testing PCI ACS support. If
2900 * any step along the way does not support the required flags, return false.
2902 bool pci_acs_path_enabled(struct pci_dev *start,
2903 struct pci_dev *end, u16 acs_flags)
2905 struct pci_dev *pdev, *parent = start;
2910 if (!pci_acs_enabled(pdev, acs_flags))
2913 if (pci_is_root_bus(pdev->bus))
2914 return (end == NULL);
2916 parent = pdev->bus->self;
2917 } while (pdev != end);
2923 * pci_swizzle_interrupt_pin - swizzle INTx for device behind bridge
2924 * @dev: the PCI device
2925 * @pin: the INTx pin (1=INTA, 2=INTB, 3=INTC, 4=INTD)
2927 * Perform INTx swizzling for a device behind one level of bridge. This is
2928 * required by section 9.1 of the PCI-to-PCI bridge specification for devices
2929 * behind bridges on add-in cards. For devices with ARI enabled, the slot
2930 * number is always 0 (see the Implementation Note in section 2.2.8.1 of
2931 * the PCI Express Base Specification, Revision 2.1)
2933 u8 pci_swizzle_interrupt_pin(const struct pci_dev *dev, u8 pin)
2937 if (pci_ari_enabled(dev->bus))
2940 slot = PCI_SLOT(dev->devfn);
2942 return (((pin - 1) + slot) % 4) + 1;
2945 int pci_get_interrupt_pin(struct pci_dev *dev, struct pci_dev **bridge)
2953 while (!pci_is_root_bus(dev->bus)) {
2954 pin = pci_swizzle_interrupt_pin(dev, pin);
2955 dev = dev->bus->self;
2962 * pci_common_swizzle - swizzle INTx all the way to root bridge
2963 * @dev: the PCI device
2964 * @pinp: pointer to the INTx pin value (1=INTA, 2=INTB, 3=INTD, 4=INTD)
2966 * Perform INTx swizzling for a device. This traverses through all PCI-to-PCI
2967 * bridges all the way up to a PCI root bus.
2969 u8 pci_common_swizzle(struct pci_dev *dev, u8 *pinp)
2973 while (!pci_is_root_bus(dev->bus)) {
2974 pin = pci_swizzle_interrupt_pin(dev, pin);
2975 dev = dev->bus->self;
2978 return PCI_SLOT(dev->devfn);
2980 EXPORT_SYMBOL_GPL(pci_common_swizzle);
2983 * pci_release_region - Release a PCI bar
2984 * @pdev: PCI device whose resources were previously reserved by pci_request_region
2985 * @bar: BAR to release
2987 * Releases the PCI I/O and memory resources previously reserved by a
2988 * successful call to pci_request_region. Call this function only
2989 * after all use of the PCI regions has ceased.
2991 void pci_release_region(struct pci_dev *pdev, int bar)
2993 struct pci_devres *dr;
2995 if (pci_resource_len(pdev, bar) == 0)
2997 if (pci_resource_flags(pdev, bar) & IORESOURCE_IO)
2998 release_region(pci_resource_start(pdev, bar),
2999 pci_resource_len(pdev, bar));
3000 else if (pci_resource_flags(pdev, bar) & IORESOURCE_MEM)
3001 release_mem_region(pci_resource_start(pdev, bar),
3002 pci_resource_len(pdev, bar));
3004 dr = find_pci_dr(pdev);
3006 dr->region_mask &= ~(1 << bar);
3008 EXPORT_SYMBOL(pci_release_region);
3011 * __pci_request_region - Reserved PCI I/O and memory resource
3012 * @pdev: PCI device whose resources are to be reserved
3013 * @bar: BAR to be reserved
3014 * @res_name: Name to be associated with resource.
3015 * @exclusive: whether the region access is exclusive or not
3017 * Mark the PCI region associated with PCI device @pdev BR @bar as
3018 * being reserved by owner @res_name. Do not access any
3019 * address inside the PCI regions unless this call returns
3022 * If @exclusive is set, then the region is marked so that userspace
3023 * is explicitly not allowed to map the resource via /dev/mem or
3024 * sysfs MMIO access.
3026 * Returns 0 on success, or %EBUSY on error. A warning
3027 * message is also printed on failure.
3029 static int __pci_request_region(struct pci_dev *pdev, int bar,
3030 const char *res_name, int exclusive)
3032 struct pci_devres *dr;
3034 if (pci_resource_len(pdev, bar) == 0)
3037 if (pci_resource_flags(pdev, bar) & IORESOURCE_IO) {
3038 if (!request_region(pci_resource_start(pdev, bar),
3039 pci_resource_len(pdev, bar), res_name))
3041 } else if (pci_resource_flags(pdev, bar) & IORESOURCE_MEM) {
3042 if (!__request_mem_region(pci_resource_start(pdev, bar),
3043 pci_resource_len(pdev, bar), res_name,
3048 dr = find_pci_dr(pdev);
3050 dr->region_mask |= 1 << bar;
3055 dev_warn(&pdev->dev, "BAR %d: can't reserve %pR\n", bar,
3056 &pdev->resource[bar]);
3061 * pci_request_region - Reserve PCI I/O and memory resource
3062 * @pdev: PCI device whose resources are to be reserved
3063 * @bar: BAR to be reserved
3064 * @res_name: Name to be associated with resource
3066 * Mark the PCI region associated with PCI device @pdev BAR @bar as
3067 * being reserved by owner @res_name. Do not access any
3068 * address inside the PCI regions unless this call returns
3071 * Returns 0 on success, or %EBUSY on error. A warning
3072 * message is also printed on failure.
3074 int pci_request_region(struct pci_dev *pdev, int bar, const char *res_name)
3076 return __pci_request_region(pdev, bar, res_name, 0);
3078 EXPORT_SYMBOL(pci_request_region);
3081 * pci_request_region_exclusive - Reserved PCI I/O and memory resource
3082 * @pdev: PCI device whose resources are to be reserved
3083 * @bar: BAR to be reserved
3084 * @res_name: Name to be associated with resource.
3086 * Mark the PCI region associated with PCI device @pdev BR @bar as
3087 * being reserved by owner @res_name. Do not access any
3088 * address inside the PCI regions unless this call returns
3091 * Returns 0 on success, or %EBUSY on error. A warning
3092 * message is also printed on failure.
3094 * The key difference that _exclusive makes it that userspace is
3095 * explicitly not allowed to map the resource via /dev/mem or
3098 int pci_request_region_exclusive(struct pci_dev *pdev, int bar,
3099 const char *res_name)
3101 return __pci_request_region(pdev, bar, res_name, IORESOURCE_EXCLUSIVE);
3103 EXPORT_SYMBOL(pci_request_region_exclusive);
3106 * pci_release_selected_regions - Release selected PCI I/O and memory resources
3107 * @pdev: PCI device whose resources were previously reserved
3108 * @bars: Bitmask of BARs to be released
3110 * Release selected PCI I/O and memory resources previously reserved.
3111 * Call this function only after all use of the PCI regions has ceased.
3113 void pci_release_selected_regions(struct pci_dev *pdev, int bars)
3117 for (i = 0; i < 6; i++)
3118 if (bars & (1 << i))
3119 pci_release_region(pdev, i);
3121 EXPORT_SYMBOL(pci_release_selected_regions);
3123 static int __pci_request_selected_regions(struct pci_dev *pdev, int bars,
3124 const char *res_name, int excl)
3128 for (i = 0; i < 6; i++)
3129 if (bars & (1 << i))
3130 if (__pci_request_region(pdev, i, res_name, excl))
3136 if (bars & (1 << i))
3137 pci_release_region(pdev, i);
3144 * pci_request_selected_regions - Reserve selected PCI I/O and memory resources
3145 * @pdev: PCI device whose resources are to be reserved
3146 * @bars: Bitmask of BARs to be requested
3147 * @res_name: Name to be associated with resource
3149 int pci_request_selected_regions(struct pci_dev *pdev, int bars,
3150 const char *res_name)
3152 return __pci_request_selected_regions(pdev, bars, res_name, 0);
3154 EXPORT_SYMBOL(pci_request_selected_regions);
3156 int pci_request_selected_regions_exclusive(struct pci_dev *pdev, int bars,
3157 const char *res_name)
3159 return __pci_request_selected_regions(pdev, bars, res_name,
3160 IORESOURCE_EXCLUSIVE);
3162 EXPORT_SYMBOL(pci_request_selected_regions_exclusive);
3165 * pci_release_regions - Release reserved PCI I/O and memory resources
3166 * @pdev: PCI device whose resources were previously reserved by pci_request_regions
3168 * Releases all PCI I/O and memory resources previously reserved by a
3169 * successful call to pci_request_regions. Call this function only
3170 * after all use of the PCI regions has ceased.
3173 void pci_release_regions(struct pci_dev *pdev)
3175 pci_release_selected_regions(pdev, (1 << 6) - 1);
3177 EXPORT_SYMBOL(pci_release_regions);
3180 * pci_request_regions - Reserved PCI I/O and memory resources
3181 * @pdev: PCI device whose resources are to be reserved
3182 * @res_name: Name to be associated with resource.
3184 * Mark all PCI regions associated with PCI device @pdev as
3185 * being reserved by owner @res_name. Do not access any
3186 * address inside the PCI regions unless this call returns
3189 * Returns 0 on success, or %EBUSY on error. A warning
3190 * message is also printed on failure.
3192 int pci_request_regions(struct pci_dev *pdev, const char *res_name)
3194 return pci_request_selected_regions(pdev, ((1 << 6) - 1), res_name);
3196 EXPORT_SYMBOL(pci_request_regions);
3199 * pci_request_regions_exclusive - Reserved PCI I/O and memory resources
3200 * @pdev: PCI device whose resources are to be reserved
3201 * @res_name: Name to be associated with resource.
3203 * Mark all PCI regions associated with PCI device @pdev as
3204 * being reserved by owner @res_name. Do not access any
3205 * address inside the PCI regions unless this call returns
3208 * pci_request_regions_exclusive() will mark the region so that
3209 * /dev/mem and the sysfs MMIO access will not be allowed.
3211 * Returns 0 on success, or %EBUSY on error. A warning
3212 * message is also printed on failure.
3214 int pci_request_regions_exclusive(struct pci_dev *pdev, const char *res_name)
3216 return pci_request_selected_regions_exclusive(pdev,
3217 ((1 << 6) - 1), res_name);
3219 EXPORT_SYMBOL(pci_request_regions_exclusive);
3223 struct list_head list;
3225 resource_size_t size;
3228 static LIST_HEAD(io_range_list);
3229 static DEFINE_SPINLOCK(io_range_lock);
3233 * Record the PCI IO range (expressed as CPU physical address + size).
3234 * Return a negative value if an error has occured, zero otherwise
3236 int __weak pci_register_io_range(phys_addr_t addr, resource_size_t size)
3241 struct io_range *range;
3242 resource_size_t allocated_size = 0;
3244 /* check if the range hasn't been previously recorded */
3245 spin_lock(&io_range_lock);
3246 list_for_each_entry(range, &io_range_list, list) {
3247 if (addr >= range->start && addr + size <= range->start + size) {
3248 /* range already registered, bail out */
3251 allocated_size += range->size;
3254 /* range not registed yet, check for available space */
3255 if (allocated_size + size - 1 > IO_SPACE_LIMIT) {
3256 /* if it's too big check if 64K space can be reserved */
3257 if (allocated_size + SZ_64K - 1 > IO_SPACE_LIMIT) {
3263 pr_warn("Requested IO range too big, new size set to 64K\n");
3266 /* add the range to the list */
3267 range = kzalloc(sizeof(*range), GFP_ATOMIC);
3273 range->start = addr;
3276 list_add_tail(&range->list, &io_range_list);
3279 spin_unlock(&io_range_lock);
3285 phys_addr_t pci_pio_to_address(unsigned long pio)
3287 phys_addr_t address = (phys_addr_t)OF_BAD_ADDR;
3290 struct io_range *range;
3291 resource_size_t allocated_size = 0;
3293 if (pio > IO_SPACE_LIMIT)
3296 spin_lock(&io_range_lock);
3297 list_for_each_entry(range, &io_range_list, list) {
3298 if (pio >= allocated_size && pio < allocated_size + range->size) {
3299 address = range->start + pio - allocated_size;
3302 allocated_size += range->size;
3304 spin_unlock(&io_range_lock);
3310 unsigned long __weak pci_address_to_pio(phys_addr_t address)
3313 struct io_range *res;
3314 resource_size_t offset = 0;
3315 unsigned long addr = -1;
3317 spin_lock(&io_range_lock);
3318 list_for_each_entry(res, &io_range_list, list) {
3319 if (address >= res->start && address < res->start + res->size) {
3320 addr = address - res->start + offset;
3323 offset += res->size;
3325 spin_unlock(&io_range_lock);
3329 if (address > IO_SPACE_LIMIT)
3330 return (unsigned long)-1;
3332 return (unsigned long) address;
3337 * pci_remap_iospace - Remap the memory mapped I/O space
3338 * @res: Resource describing the I/O space
3339 * @phys_addr: physical address of range to be mapped
3341 * Remap the memory mapped I/O space described by the @res
3342 * and the CPU physical address @phys_addr into virtual address space.
3343 * Only architectures that have memory mapped IO functions defined
3344 * (and the PCI_IOBASE value defined) should call this function.
3346 int __weak pci_remap_iospace(const struct resource *res, phys_addr_t phys_addr)
3348 #if defined(PCI_IOBASE) && defined(CONFIG_MMU)
3349 unsigned long vaddr = (unsigned long)PCI_IOBASE + res->start;
3351 if (!(res->flags & IORESOURCE_IO))
3354 if (res->end > IO_SPACE_LIMIT)
3357 return ioremap_page_range(vaddr, vaddr + resource_size(res), phys_addr,
3358 pgprot_device(PAGE_KERNEL));
3360 /* this architecture does not have memory mapped I/O space,
3361 so this function should never be called */
3362 WARN_ONCE(1, "This architecture does not support memory mapped I/O\n");
3368 * pci_unmap_iospace - Unmap the memory mapped I/O space
3369 * @res: resource to be unmapped
3371 * Unmap the CPU virtual address @res from virtual address space.
3372 * Only architectures that have memory mapped IO functions defined
3373 * (and the PCI_IOBASE value defined) should call this function.
3375 void pci_unmap_iospace(struct resource *res)
3377 #if defined(PCI_IOBASE) && defined(CONFIG_MMU)
3378 unsigned long vaddr = (unsigned long)PCI_IOBASE + res->start;
3380 unmap_kernel_range(vaddr, resource_size(res));
3384 static void __pci_set_master(struct pci_dev *dev, bool enable)
3388 pci_read_config_word(dev, PCI_COMMAND, &old_cmd);
3390 cmd = old_cmd | PCI_COMMAND_MASTER;
3392 cmd = old_cmd & ~PCI_COMMAND_MASTER;
3393 if (cmd != old_cmd) {
3394 dev_dbg(&dev->dev, "%s bus mastering\n",
3395 enable ? "enabling" : "disabling");
3396 pci_write_config_word(dev, PCI_COMMAND, cmd);
3398 dev->is_busmaster = enable;
3402 * pcibios_setup - process "pci=" kernel boot arguments
3403 * @str: string used to pass in "pci=" kernel boot arguments
3405 * Process kernel boot arguments. This is the default implementation.
3406 * Architecture specific implementations can override this as necessary.
3408 char * __weak __init pcibios_setup(char *str)
3414 * pcibios_set_master - enable PCI bus-mastering for device dev
3415 * @dev: the PCI device to enable
3417 * Enables PCI bus-mastering for the device. This is the default
3418 * implementation. Architecture specific implementations can override
3419 * this if necessary.
3421 void __weak pcibios_set_master(struct pci_dev *dev)
3425 /* The latency timer doesn't apply to PCIe (either Type 0 or Type 1) */
3426 if (pci_is_pcie(dev))
3429 pci_read_config_byte(dev, PCI_LATENCY_TIMER, &lat);
3431 lat = (64 <= pcibios_max_latency) ? 64 : pcibios_max_latency;
3432 else if (lat > pcibios_max_latency)
3433 lat = pcibios_max_latency;
3437 pci_write_config_byte(dev, PCI_LATENCY_TIMER, lat);
3441 * pci_set_master - enables bus-mastering for device dev
3442 * @dev: the PCI device to enable
3444 * Enables bus-mastering on the device and calls pcibios_set_master()
3445 * to do the needed arch specific settings.
3447 void pci_set_master(struct pci_dev *dev)
3449 __pci_set_master(dev, true);
3450 pcibios_set_master(dev);
3452 EXPORT_SYMBOL(pci_set_master);
3455 * pci_clear_master - disables bus-mastering for device dev
3456 * @dev: the PCI device to disable
3458 void pci_clear_master(struct pci_dev *dev)
3460 __pci_set_master(dev, false);
3462 EXPORT_SYMBOL(pci_clear_master);
3465 * pci_set_cacheline_size - ensure the CACHE_LINE_SIZE register is programmed
3466 * @dev: the PCI device for which MWI is to be enabled
3468 * Helper function for pci_set_mwi.
3469 * Originally copied from drivers/net/acenic.c.
3470 * Copyright 1998-2001 by Jes Sorensen, <jes@trained-monkey.org>.
3472 * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
3474 int pci_set_cacheline_size(struct pci_dev *dev)
3478 if (!pci_cache_line_size)
3481 /* Validate current setting: the PCI_CACHE_LINE_SIZE must be
3482 equal to or multiple of the right value. */
3483 pci_read_config_byte(dev, PCI_CACHE_LINE_SIZE, &cacheline_size);
3484 if (cacheline_size >= pci_cache_line_size &&
3485 (cacheline_size % pci_cache_line_size) == 0)
3488 /* Write the correct value. */
3489 pci_write_config_byte(dev, PCI_CACHE_LINE_SIZE, pci_cache_line_size);
3491 pci_read_config_byte(dev, PCI_CACHE_LINE_SIZE, &cacheline_size);
3492 if (cacheline_size == pci_cache_line_size)
3495 dev_printk(KERN_DEBUG, &dev->dev, "cache line size of %d is not supported\n",
3496 pci_cache_line_size << 2);
3500 EXPORT_SYMBOL_GPL(pci_set_cacheline_size);
3503 * pci_set_mwi - enables memory-write-invalidate PCI transaction
3504 * @dev: the PCI device for which MWI is enabled
3506 * Enables the Memory-Write-Invalidate transaction in %PCI_COMMAND.
3508 * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
3510 int pci_set_mwi(struct pci_dev *dev)
3512 #ifdef PCI_DISABLE_MWI
3518 rc = pci_set_cacheline_size(dev);
3522 pci_read_config_word(dev, PCI_COMMAND, &cmd);
3523 if (!(cmd & PCI_COMMAND_INVALIDATE)) {
3524 dev_dbg(&dev->dev, "enabling Mem-Wr-Inval\n");
3525 cmd |= PCI_COMMAND_INVALIDATE;
3526 pci_write_config_word(dev, PCI_COMMAND, cmd);
3531 EXPORT_SYMBOL(pci_set_mwi);
3534 * pci_try_set_mwi - enables memory-write-invalidate PCI transaction
3535 * @dev: the PCI device for which MWI is enabled
3537 * Enables the Memory-Write-Invalidate transaction in %PCI_COMMAND.
3538 * Callers are not required to check the return value.
3540 * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
3542 int pci_try_set_mwi(struct pci_dev *dev)
3544 #ifdef PCI_DISABLE_MWI
3547 return pci_set_mwi(dev);
3550 EXPORT_SYMBOL(pci_try_set_mwi);
3553 * pci_clear_mwi - disables Memory-Write-Invalidate for device dev
3554 * @dev: the PCI device to disable
3556 * Disables PCI Memory-Write-Invalidate transaction on the device
3558 void pci_clear_mwi(struct pci_dev *dev)
3560 #ifndef PCI_DISABLE_MWI
3563 pci_read_config_word(dev, PCI_COMMAND, &cmd);
3564 if (cmd & PCI_COMMAND_INVALIDATE) {
3565 cmd &= ~PCI_COMMAND_INVALIDATE;
3566 pci_write_config_word(dev, PCI_COMMAND, cmd);
3570 EXPORT_SYMBOL(pci_clear_mwi);
3573 * pci_intx - enables/disables PCI INTx for device dev
3574 * @pdev: the PCI device to operate on
3575 * @enable: boolean: whether to enable or disable PCI INTx
3577 * Enables/disables PCI INTx for device dev
3579 void pci_intx(struct pci_dev *pdev, int enable)
3581 u16 pci_command, new;
3583 pci_read_config_word(pdev, PCI_COMMAND, &pci_command);
3586 new = pci_command & ~PCI_COMMAND_INTX_DISABLE;
3588 new = pci_command | PCI_COMMAND_INTX_DISABLE;
3590 if (new != pci_command) {
3591 struct pci_devres *dr;
3593 pci_write_config_word(pdev, PCI_COMMAND, new);
3595 dr = find_pci_dr(pdev);
3596 if (dr && !dr->restore_intx) {
3597 dr->restore_intx = 1;
3598 dr->orig_intx = !enable;
3602 EXPORT_SYMBOL_GPL(pci_intx);
3605 * pci_intx_mask_supported - probe for INTx masking support
3606 * @dev: the PCI device to operate on
3608 * Check if the device dev support INTx masking via the config space
3611 bool pci_intx_mask_supported(struct pci_dev *dev)
3613 bool mask_supported = false;
3616 if (dev->broken_intx_masking)
3619 pci_cfg_access_lock(dev);
3621 pci_read_config_word(dev, PCI_COMMAND, &orig);
3622 pci_write_config_word(dev, PCI_COMMAND,
3623 orig ^ PCI_COMMAND_INTX_DISABLE);
3624 pci_read_config_word(dev, PCI_COMMAND, &new);
3627 * There's no way to protect against hardware bugs or detect them
3628 * reliably, but as long as we know what the value should be, let's
3629 * go ahead and check it.
3631 if ((new ^ orig) & ~PCI_COMMAND_INTX_DISABLE) {
3632 dev_err(&dev->dev, "Command register changed from 0x%x to 0x%x: driver or hardware bug?\n",
3634 } else if ((new ^ orig) & PCI_COMMAND_INTX_DISABLE) {
3635 mask_supported = true;
3636 pci_write_config_word(dev, PCI_COMMAND, orig);
3639 pci_cfg_access_unlock(dev);
3640 return mask_supported;
3642 EXPORT_SYMBOL_GPL(pci_intx_mask_supported);
3644 static bool pci_check_and_set_intx_mask(struct pci_dev *dev, bool mask)
3646 struct pci_bus *bus = dev->bus;
3647 bool mask_updated = true;
3648 u32 cmd_status_dword;
3649 u16 origcmd, newcmd;
3650 unsigned long flags;
3654 * We do a single dword read to retrieve both command and status.
3655 * Document assumptions that make this possible.
3657 BUILD_BUG_ON(PCI_COMMAND % 4);
3658 BUILD_BUG_ON(PCI_COMMAND + 2 != PCI_STATUS);
3660 raw_spin_lock_irqsave(&pci_lock, flags);
3662 bus->ops->read(bus, dev->devfn, PCI_COMMAND, 4, &cmd_status_dword);
3664 irq_pending = (cmd_status_dword >> 16) & PCI_STATUS_INTERRUPT;
3667 * Check interrupt status register to see whether our device
3668 * triggered the interrupt (when masking) or the next IRQ is
3669 * already pending (when unmasking).
3671 if (mask != irq_pending) {
3672 mask_updated = false;
3676 origcmd = cmd_status_dword;
3677 newcmd = origcmd & ~PCI_COMMAND_INTX_DISABLE;
3679 newcmd |= PCI_COMMAND_INTX_DISABLE;
3680 if (newcmd != origcmd)
3681 bus->ops->write(bus, dev->devfn, PCI_COMMAND, 2, newcmd);
3684 raw_spin_unlock_irqrestore(&pci_lock, flags);
3686 return mask_updated;
3690 * pci_check_and_mask_intx - mask INTx on pending interrupt
3691 * @dev: the PCI device to operate on
3693 * Check if the device dev has its INTx line asserted, mask it and
3694 * return true in that case. False is returned if not interrupt was
3697 bool pci_check_and_mask_intx(struct pci_dev *dev)
3699 return pci_check_and_set_intx_mask(dev, true);
3701 EXPORT_SYMBOL_GPL(pci_check_and_mask_intx);
3704 * pci_check_and_unmask_intx - unmask INTx if no interrupt is pending
3705 * @dev: the PCI device to operate on
3707 * Check if the device dev has its INTx line asserted, unmask it if not
3708 * and return true. False is returned and the mask remains active if
3709 * there was still an interrupt pending.
3711 bool pci_check_and_unmask_intx(struct pci_dev *dev)
3713 return pci_check_and_set_intx_mask(dev, false);
3715 EXPORT_SYMBOL_GPL(pci_check_and_unmask_intx);
3718 * pci_wait_for_pending_transaction - waits for pending transaction
3719 * @dev: the PCI device to operate on
3721 * Return 0 if transaction is pending 1 otherwise.
3723 int pci_wait_for_pending_transaction(struct pci_dev *dev)
3725 if (!pci_is_pcie(dev))
3728 return pci_wait_for_pending(dev, pci_pcie_cap(dev) + PCI_EXP_DEVSTA,
3729 PCI_EXP_DEVSTA_TRPND);
3731 EXPORT_SYMBOL(pci_wait_for_pending_transaction);
3734 * We should only need to wait 100ms after FLR, but some devices take longer.
3735 * Wait for up to 1000ms for config space to return something other than -1.
3736 * Intel IGD requires this when an LCD panel is attached. We read the 2nd
3737 * dword because VFs don't implement the 1st dword.
3739 static void pci_flr_wait(struct pci_dev *dev)
3746 pci_read_config_dword(dev, PCI_COMMAND, &id);
3747 } while (i++ < 10 && id == ~0);
3750 dev_warn(&dev->dev, "Failed to return from FLR\n");
3752 dev_info(&dev->dev, "Required additional %dms to return from FLR\n",
3756 static int pcie_flr(struct pci_dev *dev, int probe)
3760 pcie_capability_read_dword(dev, PCI_EXP_DEVCAP, &cap);
3761 if (!(cap & PCI_EXP_DEVCAP_FLR))
3767 if (!pci_wait_for_pending_transaction(dev))
3768 dev_err(&dev->dev, "timed out waiting for pending transaction; performing function level reset anyway\n");
3770 pcie_capability_set_word(dev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
3775 static int pci_af_flr(struct pci_dev *dev, int probe)
3780 pos = pci_find_capability(dev, PCI_CAP_ID_AF);
3784 pci_read_config_byte(dev, pos + PCI_AF_CAP, &cap);
3785 if (!(cap & PCI_AF_CAP_TP) || !(cap & PCI_AF_CAP_FLR))
3792 * Wait for Transaction Pending bit to clear. A word-aligned test
3793 * is used, so we use the conrol offset rather than status and shift
3794 * the test bit to match.
3796 if (!pci_wait_for_pending(dev, pos + PCI_AF_CTRL,
3797 PCI_AF_STATUS_TP << 8))
3798 dev_err(&dev->dev, "timed out waiting for pending transaction; performing AF function level reset anyway\n");
3800 pci_write_config_byte(dev, pos + PCI_AF_CTRL, PCI_AF_CTRL_FLR);
3806 * pci_pm_reset - Put device into PCI_D3 and back into PCI_D0.
3807 * @dev: Device to reset.
3808 * @probe: If set, only check if the device can be reset this way.
3810 * If @dev supports native PCI PM and its PCI_PM_CTRL_NO_SOFT_RESET flag is
3811 * unset, it will be reinitialized internally when going from PCI_D3hot to
3812 * PCI_D0. If that's the case and the device is not in a low-power state
3813 * already, force it into PCI_D3hot and back to PCI_D0, causing it to be reset.
3815 * NOTE: This causes the caller to sleep for twice the device power transition
3816 * cooldown period, which for the D0->D3hot and D3hot->D0 transitions is 10 ms
3817 * by default (i.e. unless the @dev's d3_delay field has a different value).
3818 * Moreover, only devices in D0 can be reset by this function.
3820 static int pci_pm_reset(struct pci_dev *dev, int probe)
3824 if (!dev->pm_cap || dev->dev_flags & PCI_DEV_FLAGS_NO_PM_RESET)
3827 pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &csr);
3828 if (csr & PCI_PM_CTRL_NO_SOFT_RESET)
3834 if (dev->current_state != PCI_D0)
3837 csr &= ~PCI_PM_CTRL_STATE_MASK;
3839 pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, csr);
3840 pci_dev_d3_sleep(dev);
3842 csr &= ~PCI_PM_CTRL_STATE_MASK;
3844 pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, csr);
3845 pci_dev_d3_sleep(dev);
3850 void pci_reset_secondary_bus(struct pci_dev *dev)
3854 pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &ctrl);
3855 ctrl |= PCI_BRIDGE_CTL_BUS_RESET;
3856 pci_write_config_word(dev, PCI_BRIDGE_CONTROL, ctrl);
3858 * PCI spec v3.0 7.6.4.2 requires minimum Trst of 1ms. Double
3859 * this to 2ms to ensure that we meet the minimum requirement.
3863 ctrl &= ~PCI_BRIDGE_CTL_BUS_RESET;
3864 pci_write_config_word(dev, PCI_BRIDGE_CONTROL, ctrl);
3867 * Trhfa for conventional PCI is 2^25 clock cycles.
3868 * Assuming a minimum 33MHz clock this results in a 1s
3869 * delay before we can consider subordinate devices to
3870 * be re-initialized. PCIe has some ways to shorten this,
3871 * but we don't make use of them yet.
3876 void __weak pcibios_reset_secondary_bus(struct pci_dev *dev)
3878 pci_reset_secondary_bus(dev);
3882 * pci_reset_bridge_secondary_bus - Reset the secondary bus on a PCI bridge.
3883 * @dev: Bridge device
3885 * Use the bridge control register to assert reset on the secondary bus.
3886 * Devices on the secondary bus are left in power-on state.
3888 void pci_reset_bridge_secondary_bus(struct pci_dev *dev)
3890 pcibios_reset_secondary_bus(dev);
3892 EXPORT_SYMBOL_GPL(pci_reset_bridge_secondary_bus);
3894 static int pci_parent_bus_reset(struct pci_dev *dev, int probe)
3896 struct pci_dev *pdev;
3898 if (pci_is_root_bus(dev->bus) || dev->subordinate ||
3899 !dev->bus->self || dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET)
3902 list_for_each_entry(pdev, &dev->bus->devices, bus_list)
3909 pci_reset_bridge_secondary_bus(dev->bus->self);
3914 static int pci_reset_hotplug_slot(struct hotplug_slot *hotplug, int probe)
3918 if (!hotplug || !try_module_get(hotplug->ops->owner))
3921 if (hotplug->ops->reset_slot)
3922 rc = hotplug->ops->reset_slot(hotplug, probe);
3924 module_put(hotplug->ops->owner);
3929 static int pci_dev_reset_slot_function(struct pci_dev *dev, int probe)
3931 struct pci_dev *pdev;
3933 if (dev->subordinate || !dev->slot ||
3934 dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET)
3937 list_for_each_entry(pdev, &dev->bus->devices, bus_list)
3938 if (pdev != dev && pdev->slot == dev->slot)
3941 return pci_reset_hotplug_slot(dev->slot->hotplug, probe);
3944 static int __pci_dev_reset(struct pci_dev *dev, int probe)
3950 rc = pci_dev_specific_reset(dev, probe);
3954 rc = pcie_flr(dev, probe);
3958 rc = pci_af_flr(dev, probe);
3962 rc = pci_pm_reset(dev, probe);
3966 rc = pci_dev_reset_slot_function(dev, probe);
3970 rc = pci_parent_bus_reset(dev, probe);
3975 static void pci_dev_lock(struct pci_dev *dev)
3977 pci_cfg_access_lock(dev);
3978 /* block PM suspend, driver probe, etc. */
3979 device_lock(&dev->dev);
3982 /* Return 1 on successful lock, 0 on contention */
3983 static int pci_dev_trylock(struct pci_dev *dev)
3985 if (pci_cfg_access_trylock(dev)) {
3986 if (device_trylock(&dev->dev))
3988 pci_cfg_access_unlock(dev);
3994 static void pci_dev_unlock(struct pci_dev *dev)
3996 device_unlock(&dev->dev);
3997 pci_cfg_access_unlock(dev);
4001 * pci_reset_notify - notify device driver of reset
4002 * @dev: device to be notified of reset
4003 * @prepare: 'true' if device is about to be reset; 'false' if reset attempt
4006 * Must be called prior to device access being disabled and after device
4007 * access is restored.
4009 static void pci_reset_notify(struct pci_dev *dev, bool prepare)
4011 const struct pci_error_handlers *err_handler =
4012 dev->driver ? dev->driver->err_handler : NULL;
4013 if (err_handler && err_handler->reset_notify)
4014 err_handler->reset_notify(dev, prepare);
4017 static void pci_dev_save_and_disable(struct pci_dev *dev)
4019 pci_reset_notify(dev, true);
4022 * Wake-up device prior to save. PM registers default to D0 after
4023 * reset and a simple register restore doesn't reliably return
4024 * to a non-D0 state anyway.
4026 pci_set_power_state(dev, PCI_D0);
4028 pci_save_state(dev);
4030 * Disable the device by clearing the Command register, except for
4031 * INTx-disable which is set. This not only disables MMIO and I/O port
4032 * BARs, but also prevents the device from being Bus Master, preventing
4033 * DMA from the device including MSI/MSI-X interrupts. For PCI 2.3
4034 * compliant devices, INTx-disable prevents legacy interrupts.
4036 pci_write_config_word(dev, PCI_COMMAND, PCI_COMMAND_INTX_DISABLE);
4039 static void pci_dev_restore(struct pci_dev *dev)
4041 pci_restore_state(dev);
4042 pci_reset_notify(dev, false);
4045 static int pci_dev_reset(struct pci_dev *dev, int probe)
4052 rc = __pci_dev_reset(dev, probe);
4055 pci_dev_unlock(dev);
4061 * __pci_reset_function - reset a PCI device function
4062 * @dev: PCI device to reset
4064 * Some devices allow an individual function to be reset without affecting
4065 * other functions in the same device. The PCI device must be responsive
4066 * to PCI config space in order to use this function.
4068 * The device function is presumed to be unused when this function is called.
4069 * Resetting the device will make the contents of PCI configuration space
4070 * random, so any caller of this must be prepared to reinitialise the
4071 * device including MSI, bus mastering, BARs, decoding IO and memory spaces,
4074 * Returns 0 if the device function was successfully reset or negative if the
4075 * device doesn't support resetting a single function.
4077 int __pci_reset_function(struct pci_dev *dev)
4079 return pci_dev_reset(dev, 0);
4081 EXPORT_SYMBOL_GPL(__pci_reset_function);
4084 * __pci_reset_function_locked - reset a PCI device function while holding
4085 * the @dev mutex lock.
4086 * @dev: PCI device to reset
4088 * Some devices allow an individual function to be reset without affecting
4089 * other functions in the same device. The PCI device must be responsive
4090 * to PCI config space in order to use this function.
4092 * The device function is presumed to be unused and the caller is holding
4093 * the device mutex lock when this function is called.
4094 * Resetting the device will make the contents of PCI configuration space
4095 * random, so any caller of this must be prepared to reinitialise the
4096 * device including MSI, bus mastering, BARs, decoding IO and memory spaces,
4099 * Returns 0 if the device function was successfully reset or negative if the
4100 * device doesn't support resetting a single function.
4102 int __pci_reset_function_locked(struct pci_dev *dev)
4104 return __pci_dev_reset(dev, 0);
4106 EXPORT_SYMBOL_GPL(__pci_reset_function_locked);
4109 * pci_probe_reset_function - check whether the device can be safely reset
4110 * @dev: PCI device to reset
4112 * Some devices allow an individual function to be reset without affecting
4113 * other functions in the same device. The PCI device must be responsive
4114 * to PCI config space in order to use this function.
4116 * Returns 0 if the device function can be reset or negative if the
4117 * device doesn't support resetting a single function.
4119 int pci_probe_reset_function(struct pci_dev *dev)
4121 return pci_dev_reset(dev, 1);
4125 * pci_reset_function - quiesce and reset a PCI device function
4126 * @dev: PCI device to reset
4128 * Some devices allow an individual function to be reset without affecting
4129 * other functions in the same device. The PCI device must be responsive
4130 * to PCI config space in order to use this function.
4132 * This function does not just reset the PCI portion of a device, but
4133 * clears all the state associated with the device. This function differs
4134 * from __pci_reset_function in that it saves and restores device state
4137 * Returns 0 if the device function was successfully reset or negative if the
4138 * device doesn't support resetting a single function.
4140 int pci_reset_function(struct pci_dev *dev)
4144 rc = pci_dev_reset(dev, 1);
4148 pci_dev_save_and_disable(dev);
4150 rc = pci_dev_reset(dev, 0);
4152 pci_dev_restore(dev);
4156 EXPORT_SYMBOL_GPL(pci_reset_function);
4159 * pci_try_reset_function - quiesce and reset a PCI device function
4160 * @dev: PCI device to reset
4162 * Same as above, except return -EAGAIN if unable to lock device.
4164 int pci_try_reset_function(struct pci_dev *dev)
4168 rc = pci_dev_reset(dev, 1);
4172 pci_dev_save_and_disable(dev);
4174 if (pci_dev_trylock(dev)) {
4175 rc = __pci_dev_reset(dev, 0);
4176 pci_dev_unlock(dev);
4180 pci_dev_restore(dev);
4184 EXPORT_SYMBOL_GPL(pci_try_reset_function);
4186 /* Do any devices on or below this bus prevent a bus reset? */
4187 static bool pci_bus_resetable(struct pci_bus *bus)
4189 struct pci_dev *dev;
4191 list_for_each_entry(dev, &bus->devices, bus_list) {
4192 if (dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET ||
4193 (dev->subordinate && !pci_bus_resetable(dev->subordinate)))
4200 /* Lock devices from the top of the tree down */
4201 static void pci_bus_lock(struct pci_bus *bus)
4203 struct pci_dev *dev;
4205 list_for_each_entry(dev, &bus->devices, bus_list) {
4207 if (dev->subordinate)
4208 pci_bus_lock(dev->subordinate);
4212 /* Unlock devices from the bottom of the tree up */
4213 static void pci_bus_unlock(struct pci_bus *bus)
4215 struct pci_dev *dev;
4217 list_for_each_entry(dev, &bus->devices, bus_list) {
4218 if (dev->subordinate)
4219 pci_bus_unlock(dev->subordinate);
4220 pci_dev_unlock(dev);
4224 /* Return 1 on successful lock, 0 on contention */
4225 static int pci_bus_trylock(struct pci_bus *bus)
4227 struct pci_dev *dev;
4229 list_for_each_entry(dev, &bus->devices, bus_list) {
4230 if (!pci_dev_trylock(dev))
4232 if (dev->subordinate) {
4233 if (!pci_bus_trylock(dev->subordinate)) {
4234 pci_dev_unlock(dev);
4242 list_for_each_entry_continue_reverse(dev, &bus->devices, bus_list) {
4243 if (dev->subordinate)
4244 pci_bus_unlock(dev->subordinate);
4245 pci_dev_unlock(dev);
4250 /* Do any devices on or below this slot prevent a bus reset? */
4251 static bool pci_slot_resetable(struct pci_slot *slot)
4253 struct pci_dev *dev;
4255 list_for_each_entry(dev, &slot->bus->devices, bus_list) {
4256 if (!dev->slot || dev->slot != slot)
4258 if (dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET ||
4259 (dev->subordinate && !pci_bus_resetable(dev->subordinate)))
4266 /* Lock devices from the top of the tree down */
4267 static void pci_slot_lock(struct pci_slot *slot)
4269 struct pci_dev *dev;
4271 list_for_each_entry(dev, &slot->bus->devices, bus_list) {
4272 if (!dev->slot || dev->slot != slot)
4275 if (dev->subordinate)
4276 pci_bus_lock(dev->subordinate);
4280 /* Unlock devices from the bottom of the tree up */
4281 static void pci_slot_unlock(struct pci_slot *slot)
4283 struct pci_dev *dev;
4285 list_for_each_entry(dev, &slot->bus->devices, bus_list) {
4286 if (!dev->slot || dev->slot != slot)
4288 if (dev->subordinate)
4289 pci_bus_unlock(dev->subordinate);
4290 pci_dev_unlock(dev);
4294 /* Return 1 on successful lock, 0 on contention */
4295 static int pci_slot_trylock(struct pci_slot *slot)
4297 struct pci_dev *dev;
4299 list_for_each_entry(dev, &slot->bus->devices, bus_list) {
4300 if (!dev->slot || dev->slot != slot)
4302 if (!pci_dev_trylock(dev))
4304 if (dev->subordinate) {
4305 if (!pci_bus_trylock(dev->subordinate)) {
4306 pci_dev_unlock(dev);
4314 list_for_each_entry_continue_reverse(dev,
4315 &slot->bus->devices, bus_list) {
4316 if (!dev->slot || dev->slot != slot)
4318 if (dev->subordinate)
4319 pci_bus_unlock(dev->subordinate);
4320 pci_dev_unlock(dev);
4325 /* Save and disable devices from the top of the tree down */
4326 static void pci_bus_save_and_disable(struct pci_bus *bus)
4328 struct pci_dev *dev;
4330 list_for_each_entry(dev, &bus->devices, bus_list) {
4331 pci_dev_save_and_disable(dev);
4332 if (dev->subordinate)
4333 pci_bus_save_and_disable(dev->subordinate);
4338 * Restore devices from top of the tree down - parent bridges need to be
4339 * restored before we can get to subordinate devices.
4341 static void pci_bus_restore(struct pci_bus *bus)
4343 struct pci_dev *dev;
4345 list_for_each_entry(dev, &bus->devices, bus_list) {
4346 pci_dev_restore(dev);
4347 if (dev->subordinate)
4348 pci_bus_restore(dev->subordinate);
4352 /* Save and disable devices from the top of the tree down */
4353 static void pci_slot_save_and_disable(struct pci_slot *slot)
4355 struct pci_dev *dev;
4357 list_for_each_entry(dev, &slot->bus->devices, bus_list) {
4358 if (!dev->slot || dev->slot != slot)
4360 pci_dev_save_and_disable(dev);
4361 if (dev->subordinate)
4362 pci_bus_save_and_disable(dev->subordinate);
4367 * Restore devices from top of the tree down - parent bridges need to be
4368 * restored before we can get to subordinate devices.
4370 static void pci_slot_restore(struct pci_slot *slot)
4372 struct pci_dev *dev;
4374 list_for_each_entry(dev, &slot->bus->devices, bus_list) {
4375 if (!dev->slot || dev->slot != slot)
4377 pci_dev_restore(dev);
4378 if (dev->subordinate)
4379 pci_bus_restore(dev->subordinate);
4383 static int pci_slot_reset(struct pci_slot *slot, int probe)
4387 if (!slot || !pci_slot_resetable(slot))
4391 pci_slot_lock(slot);
4395 rc = pci_reset_hotplug_slot(slot->hotplug, probe);
4398 pci_slot_unlock(slot);
4404 * pci_probe_reset_slot - probe whether a PCI slot can be reset
4405 * @slot: PCI slot to probe
4407 * Return 0 if slot can be reset, negative if a slot reset is not supported.
4409 int pci_probe_reset_slot(struct pci_slot *slot)
4411 return pci_slot_reset(slot, 1);
4413 EXPORT_SYMBOL_GPL(pci_probe_reset_slot);
4416 * pci_reset_slot - reset a PCI slot
4417 * @slot: PCI slot to reset
4419 * A PCI bus may host multiple slots, each slot may support a reset mechanism
4420 * independent of other slots. For instance, some slots may support slot power
4421 * control. In the case of a 1:1 bus to slot architecture, this function may
4422 * wrap the bus reset to avoid spurious slot related events such as hotplug.
4423 * Generally a slot reset should be attempted before a bus reset. All of the
4424 * function of the slot and any subordinate buses behind the slot are reset
4425 * through this function. PCI config space of all devices in the slot and
4426 * behind the slot is saved before and restored after reset.
4428 * Return 0 on success, non-zero on error.
4430 int pci_reset_slot(struct pci_slot *slot)
4434 rc = pci_slot_reset(slot, 1);
4438 pci_slot_save_and_disable(slot);
4440 rc = pci_slot_reset(slot, 0);
4442 pci_slot_restore(slot);
4446 EXPORT_SYMBOL_GPL(pci_reset_slot);
4449 * pci_try_reset_slot - Try to reset a PCI slot
4450 * @slot: PCI slot to reset
4452 * Same as above except return -EAGAIN if the slot cannot be locked
4454 int pci_try_reset_slot(struct pci_slot *slot)
4458 rc = pci_slot_reset(slot, 1);
4462 pci_slot_save_and_disable(slot);
4464 if (pci_slot_trylock(slot)) {
4466 rc = pci_reset_hotplug_slot(slot->hotplug, 0);
4467 pci_slot_unlock(slot);
4471 pci_slot_restore(slot);
4475 EXPORT_SYMBOL_GPL(pci_try_reset_slot);
4477 static int pci_bus_reset(struct pci_bus *bus, int probe)
4479 if (!bus->self || !pci_bus_resetable(bus))
4489 pci_reset_bridge_secondary_bus(bus->self);
4491 pci_bus_unlock(bus);
4497 * pci_probe_reset_bus - probe whether a PCI bus can be reset
4498 * @bus: PCI bus to probe
4500 * Return 0 if bus can be reset, negative if a bus reset is not supported.
4502 int pci_probe_reset_bus(struct pci_bus *bus)
4504 return pci_bus_reset(bus, 1);
4506 EXPORT_SYMBOL_GPL(pci_probe_reset_bus);
4509 * pci_reset_bus - reset a PCI bus
4510 * @bus: top level PCI bus to reset
4512 * Do a bus reset on the given bus and any subordinate buses, saving
4513 * and restoring state of all devices.
4515 * Return 0 on success, non-zero on error.
4517 int pci_reset_bus(struct pci_bus *bus)
4521 rc = pci_bus_reset(bus, 1);
4525 pci_bus_save_and_disable(bus);
4527 rc = pci_bus_reset(bus, 0);
4529 pci_bus_restore(bus);
4533 EXPORT_SYMBOL_GPL(pci_reset_bus);
4536 * pci_try_reset_bus - Try to reset a PCI bus
4537 * @bus: top level PCI bus to reset
4539 * Same as above except return -EAGAIN if the bus cannot be locked
4541 int pci_try_reset_bus(struct pci_bus *bus)
4545 rc = pci_bus_reset(bus, 1);
4549 pci_bus_save_and_disable(bus);
4551 if (pci_bus_trylock(bus)) {
4553 pci_reset_bridge_secondary_bus(bus->self);
4554 pci_bus_unlock(bus);
4558 pci_bus_restore(bus);
4562 EXPORT_SYMBOL_GPL(pci_try_reset_bus);
4565 * pcix_get_max_mmrbc - get PCI-X maximum designed memory read byte count
4566 * @dev: PCI device to query
4568 * Returns mmrbc: maximum designed memory read count in bytes
4569 * or appropriate error value.
4571 int pcix_get_max_mmrbc(struct pci_dev *dev)
4576 cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
4580 if (pci_read_config_dword(dev, cap + PCI_X_STATUS, &stat))
4583 return 512 << ((stat & PCI_X_STATUS_MAX_READ) >> 21);
4585 EXPORT_SYMBOL(pcix_get_max_mmrbc);
4588 * pcix_get_mmrbc - get PCI-X maximum memory read byte count
4589 * @dev: PCI device to query
4591 * Returns mmrbc: maximum memory read count in bytes
4592 * or appropriate error value.
4594 int pcix_get_mmrbc(struct pci_dev *dev)
4599 cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
4603 if (pci_read_config_word(dev, cap + PCI_X_CMD, &cmd))
4606 return 512 << ((cmd & PCI_X_CMD_MAX_READ) >> 2);
4608 EXPORT_SYMBOL(pcix_get_mmrbc);
4611 * pcix_set_mmrbc - set PCI-X maximum memory read byte count
4612 * @dev: PCI device to query
4613 * @mmrbc: maximum memory read count in bytes
4614 * valid values are 512, 1024, 2048, 4096
4616 * If possible sets maximum memory read byte count, some bridges have erratas
4617 * that prevent this.
4619 int pcix_set_mmrbc(struct pci_dev *dev, int mmrbc)
4625 if (mmrbc < 512 || mmrbc > 4096 || !is_power_of_2(mmrbc))
4628 v = ffs(mmrbc) - 10;
4630 cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
4634 if (pci_read_config_dword(dev, cap + PCI_X_STATUS, &stat))
4637 if (v > (stat & PCI_X_STATUS_MAX_READ) >> 21)
4640 if (pci_read_config_word(dev, cap + PCI_X_CMD, &cmd))
4643 o = (cmd & PCI_X_CMD_MAX_READ) >> 2;
4645 if (v > o && (dev->bus->bus_flags & PCI_BUS_FLAGS_NO_MMRBC))
4648 cmd &= ~PCI_X_CMD_MAX_READ;
4650 if (pci_write_config_word(dev, cap + PCI_X_CMD, cmd))
4655 EXPORT_SYMBOL(pcix_set_mmrbc);
4658 * pcie_get_readrq - get PCI Express read request size
4659 * @dev: PCI device to query
4661 * Returns maximum memory read request in bytes
4662 * or appropriate error value.
4664 int pcie_get_readrq(struct pci_dev *dev)
4668 pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &ctl);
4670 return 128 << ((ctl & PCI_EXP_DEVCTL_READRQ) >> 12);
4672 EXPORT_SYMBOL(pcie_get_readrq);
4675 * pcie_set_readrq - set PCI Express maximum memory read request
4676 * @dev: PCI device to query
4677 * @rq: maximum memory read count in bytes
4678 * valid values are 128, 256, 512, 1024, 2048, 4096
4680 * If possible sets maximum memory read request in bytes
4682 int pcie_set_readrq(struct pci_dev *dev, int rq)
4686 if (rq < 128 || rq > 4096 || !is_power_of_2(rq))
4690 * If using the "performance" PCIe config, we clamp the
4691 * read rq size to the max packet size to prevent the
4692 * host bridge generating requests larger than we can
4695 if (pcie_bus_config == PCIE_BUS_PERFORMANCE) {
4696 int mps = pcie_get_mps(dev);
4702 v = (ffs(rq) - 8) << 12;
4704 return pcie_capability_clear_and_set_word(dev, PCI_EXP_DEVCTL,
4705 PCI_EXP_DEVCTL_READRQ, v);
4707 EXPORT_SYMBOL(pcie_set_readrq);
4710 * pcie_get_mps - get PCI Express maximum payload size
4711 * @dev: PCI device to query
4713 * Returns maximum payload size in bytes
4715 int pcie_get_mps(struct pci_dev *dev)
4719 pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &ctl);
4721 return 128 << ((ctl & PCI_EXP_DEVCTL_PAYLOAD) >> 5);
4723 EXPORT_SYMBOL(pcie_get_mps);
4726 * pcie_set_mps - set PCI Express maximum payload size
4727 * @dev: PCI device to query
4728 * @mps: maximum payload size in bytes
4729 * valid values are 128, 256, 512, 1024, 2048, 4096
4731 * If possible sets maximum payload size
4733 int pcie_set_mps(struct pci_dev *dev, int mps)
4737 if (mps < 128 || mps > 4096 || !is_power_of_2(mps))
4741 if (v > dev->pcie_mpss)
4745 return pcie_capability_clear_and_set_word(dev, PCI_EXP_DEVCTL,
4746 PCI_EXP_DEVCTL_PAYLOAD, v);
4748 EXPORT_SYMBOL(pcie_set_mps);
4751 * pcie_get_minimum_link - determine minimum link settings of a PCI device
4752 * @dev: PCI device to query
4753 * @speed: storage for minimum speed
4754 * @width: storage for minimum width
4756 * This function will walk up the PCI device chain and determine the minimum
4757 * link width and speed of the device.
4759 int pcie_get_minimum_link(struct pci_dev *dev, enum pci_bus_speed *speed,
4760 enum pcie_link_width *width)
4764 *speed = PCI_SPEED_UNKNOWN;
4765 *width = PCIE_LNK_WIDTH_UNKNOWN;
4769 enum pci_bus_speed next_speed;
4770 enum pcie_link_width next_width;
4772 ret = pcie_capability_read_word(dev, PCI_EXP_LNKSTA, &lnksta);
4776 next_speed = pcie_link_speed[lnksta & PCI_EXP_LNKSTA_CLS];
4777 next_width = (lnksta & PCI_EXP_LNKSTA_NLW) >>
4778 PCI_EXP_LNKSTA_NLW_SHIFT;
4780 if (next_speed < *speed)
4781 *speed = next_speed;
4783 if (next_width < *width)
4784 *width = next_width;
4786 dev = dev->bus->self;
4791 EXPORT_SYMBOL(pcie_get_minimum_link);
4794 * pci_select_bars - Make BAR mask from the type of resource
4795 * @dev: the PCI device for which BAR mask is made
4796 * @flags: resource type mask to be selected
4798 * This helper routine makes bar mask from the type of resource.
4800 int pci_select_bars(struct pci_dev *dev, unsigned long flags)
4803 for (i = 0; i < PCI_NUM_RESOURCES; i++)
4804 if (pci_resource_flags(dev, i) & flags)
4808 EXPORT_SYMBOL(pci_select_bars);
4811 * pci_resource_bar - get position of the BAR associated with a resource
4812 * @dev: the PCI device
4813 * @resno: the resource number
4814 * @type: the BAR type to be filled in
4816 * Returns BAR position in config space, or 0 if the BAR is invalid.
4818 int pci_resource_bar(struct pci_dev *dev, int resno, enum pci_bar_type *type)
4822 if (resno < PCI_ROM_RESOURCE) {
4823 *type = pci_bar_unknown;
4824 return PCI_BASE_ADDRESS_0 + 4 * resno;
4825 } else if (resno == PCI_ROM_RESOURCE) {
4826 *type = pci_bar_mem32;
4827 return dev->rom_base_reg;
4828 } else if (resno < PCI_BRIDGE_RESOURCES) {
4829 /* device specific resource */
4830 *type = pci_bar_unknown;
4831 reg = pci_iov_resource_bar(dev, resno);
4836 dev_err(&dev->dev, "BAR %d: invalid resource\n", resno);
4840 /* Some architectures require additional programming to enable VGA */
4841 static arch_set_vga_state_t arch_set_vga_state;
4843 void __init pci_register_set_vga_state(arch_set_vga_state_t func)
4845 arch_set_vga_state = func; /* NULL disables */
4848 static int pci_set_vga_state_arch(struct pci_dev *dev, bool decode,
4849 unsigned int command_bits, u32 flags)
4851 if (arch_set_vga_state)
4852 return arch_set_vga_state(dev, decode, command_bits,
4858 * pci_set_vga_state - set VGA decode state on device and parents if requested
4859 * @dev: the PCI device
4860 * @decode: true = enable decoding, false = disable decoding
4861 * @command_bits: PCI_COMMAND_IO and/or PCI_COMMAND_MEMORY
4862 * @flags: traverse ancestors and change bridges
4863 * CHANGE_BRIDGE_ONLY / CHANGE_BRIDGE
4865 int pci_set_vga_state(struct pci_dev *dev, bool decode,
4866 unsigned int command_bits, u32 flags)
4868 struct pci_bus *bus;
4869 struct pci_dev *bridge;
4873 WARN_ON((flags & PCI_VGA_STATE_CHANGE_DECODES) && (command_bits & ~(PCI_COMMAND_IO|PCI_COMMAND_MEMORY)));
4875 /* ARCH specific VGA enables */
4876 rc = pci_set_vga_state_arch(dev, decode, command_bits, flags);
4880 if (flags & PCI_VGA_STATE_CHANGE_DECODES) {
4881 pci_read_config_word(dev, PCI_COMMAND, &cmd);
4883 cmd |= command_bits;
4885 cmd &= ~command_bits;
4886 pci_write_config_word(dev, PCI_COMMAND, cmd);
4889 if (!(flags & PCI_VGA_STATE_CHANGE_BRIDGE))
4896 pci_read_config_word(bridge, PCI_BRIDGE_CONTROL,
4899 cmd |= PCI_BRIDGE_CTL_VGA;
4901 cmd &= ~PCI_BRIDGE_CTL_VGA;
4902 pci_write_config_word(bridge, PCI_BRIDGE_CONTROL,
4911 * pci_add_dma_alias - Add a DMA devfn alias for a device
4912 * @dev: the PCI device for which alias is added
4913 * @devfn: alias slot and function
4915 * This helper encodes 8-bit devfn as bit number in dma_alias_mask.
4916 * It should be called early, preferably as PCI fixup header quirk.
4918 void pci_add_dma_alias(struct pci_dev *dev, u8 devfn)
4920 if (!dev->dma_alias_mask)
4921 dev->dma_alias_mask = kcalloc(BITS_TO_LONGS(U8_MAX),
4922 sizeof(long), GFP_KERNEL);
4923 if (!dev->dma_alias_mask) {
4924 dev_warn(&dev->dev, "Unable to allocate DMA alias mask\n");
4928 set_bit(devfn, dev->dma_alias_mask);
4929 dev_info(&dev->dev, "Enabling fixed DMA alias to %02x.%d\n",
4930 PCI_SLOT(devfn), PCI_FUNC(devfn));
4933 bool pci_devs_are_dma_aliases(struct pci_dev *dev1, struct pci_dev *dev2)
4935 return (dev1->dma_alias_mask &&
4936 test_bit(dev2->devfn, dev1->dma_alias_mask)) ||
4937 (dev2->dma_alias_mask &&
4938 test_bit(dev1->devfn, dev2->dma_alias_mask));
4941 bool pci_device_is_present(struct pci_dev *pdev)
4945 return pci_bus_read_dev_vendor_id(pdev->bus, pdev->devfn, &v, 0);
4947 EXPORT_SYMBOL_GPL(pci_device_is_present);
4949 void pci_ignore_hotplug(struct pci_dev *dev)
4951 struct pci_dev *bridge = dev->bus->self;
4953 dev->ignore_hotplug = 1;
4954 /* Propagate the "ignore hotplug" setting to the parent bridge. */
4956 bridge->ignore_hotplug = 1;
4958 EXPORT_SYMBOL_GPL(pci_ignore_hotplug);
4960 #define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
4961 static char resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
4962 static DEFINE_SPINLOCK(resource_alignment_lock);
4965 * pci_specified_resource_alignment - get resource alignment specified by user.
4966 * @dev: the PCI device to get
4968 * RETURNS: Resource alignment if it is specified.
4969 * Zero if it is not specified.
4971 static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev)
4973 int seg, bus, slot, func, align_order, count;
4974 unsigned short vendor, device, subsystem_vendor, subsystem_device;
4975 resource_size_t align = 0;
4978 spin_lock(&resource_alignment_lock);
4979 p = resource_alignment_param;
4982 if (pci_has_flag(PCI_PROBE_ONLY)) {
4983 pr_info_once("PCI: Ignoring requested alignments (PCI_PROBE_ONLY)\n");
4989 if (sscanf(p, "%d%n", &align_order, &count) == 1 &&
4995 if (strncmp(p, "pci:", 4) == 0) {
4996 /* PCI vendor/device (subvendor/subdevice) ids are specified */
4998 if (sscanf(p, "%hx:%hx:%hx:%hx%n",
4999 &vendor, &device, &subsystem_vendor, &subsystem_device, &count) != 4) {
5000 if (sscanf(p, "%hx:%hx%n", &vendor, &device, &count) != 2) {
5001 printk(KERN_ERR "PCI: Can't parse resource_alignment parameter: pci:%s\n",
5005 subsystem_vendor = subsystem_device = 0;
5008 if ((!vendor || (vendor == dev->vendor)) &&
5009 (!device || (device == dev->device)) &&
5010 (!subsystem_vendor || (subsystem_vendor == dev->subsystem_vendor)) &&
5011 (!subsystem_device || (subsystem_device == dev->subsystem_device))) {
5012 if (align_order == -1)
5015 align = 1 << align_order;
5021 if (sscanf(p, "%x:%x:%x.%x%n",
5022 &seg, &bus, &slot, &func, &count) != 4) {
5024 if (sscanf(p, "%x:%x.%x%n",
5025 &bus, &slot, &func, &count) != 3) {
5026 /* Invalid format */
5027 printk(KERN_ERR "PCI: Can't parse resource_alignment parameter: %s\n",
5033 if (seg == pci_domain_nr(dev->bus) &&
5034 bus == dev->bus->number &&
5035 slot == PCI_SLOT(dev->devfn) &&
5036 func == PCI_FUNC(dev->devfn)) {
5037 if (align_order == -1)
5040 align = 1 << align_order;
5045 if (*p != ';' && *p != ',') {
5046 /* End of param or invalid format */
5052 spin_unlock(&resource_alignment_lock);
5057 * This function disables memory decoding and releases memory resources
5058 * of the device specified by kernel's boot parameter 'pci=resource_alignment='.
5059 * It also rounds up size to specified alignment.
5060 * Later on, the kernel will assign page-aligned memory resource back
5063 void pci_reassigndev_resource_alignment(struct pci_dev *dev)
5067 resource_size_t align, size;
5071 * VF BARs are read-only zero according to SR-IOV spec r1.1, sec
5072 * 3.4.1.11. Their resources are allocated from the space
5073 * described by the VF BARx register in the PF's SR-IOV capability.
5074 * We can't influence their alignment here.
5079 /* check if specified PCI is target device to reassign */
5080 align = pci_specified_resource_alignment(dev);
5084 if (dev->hdr_type == PCI_HEADER_TYPE_NORMAL &&
5085 (dev->class >> 8) == PCI_CLASS_BRIDGE_HOST) {
5087 "Can't reassign resources to host bridge.\n");
5092 "Disabling memory decoding and releasing memory resources.\n");
5093 pci_read_config_word(dev, PCI_COMMAND, &command);
5094 command &= ~PCI_COMMAND_MEMORY;
5095 pci_write_config_word(dev, PCI_COMMAND, command);
5097 for (i = 0; i < PCI_BRIDGE_RESOURCES; i++) {
5098 r = &dev->resource[i];
5099 if (!(r->flags & IORESOURCE_MEM))
5101 if (r->flags & IORESOURCE_PCI_FIXED) {
5102 dev_info(&dev->dev, "Ignoring requested alignment for BAR%d: %pR\n",
5107 size = resource_size(r);
5111 "Rounding up size of resource #%d to %#llx.\n",
5112 i, (unsigned long long)size);
5114 r->flags |= IORESOURCE_UNSET;
5118 /* Need to disable bridge's resource window,
5119 * to enable the kernel to reassign new resource
5122 if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE &&
5123 (dev->class >> 8) == PCI_CLASS_BRIDGE_PCI) {
5124 for (i = PCI_BRIDGE_RESOURCES; i < PCI_NUM_RESOURCES; i++) {
5125 r = &dev->resource[i];
5126 if (!(r->flags & IORESOURCE_MEM))
5128 r->flags |= IORESOURCE_UNSET;
5129 r->end = resource_size(r) - 1;
5132 pci_disable_bridge_window(dev);
5136 static ssize_t pci_set_resource_alignment_param(const char *buf, size_t count)
5138 if (count > RESOURCE_ALIGNMENT_PARAM_SIZE - 1)
5139 count = RESOURCE_ALIGNMENT_PARAM_SIZE - 1;
5140 spin_lock(&resource_alignment_lock);
5141 strncpy(resource_alignment_param, buf, count);
5142 resource_alignment_param[count] = '\0';
5143 spin_unlock(&resource_alignment_lock);
5147 static ssize_t pci_get_resource_alignment_param(char *buf, size_t size)
5150 spin_lock(&resource_alignment_lock);
5151 count = snprintf(buf, size, "%s", resource_alignment_param);
5152 spin_unlock(&resource_alignment_lock);
5156 static ssize_t pci_resource_alignment_show(struct bus_type *bus, char *buf)
5158 return pci_get_resource_alignment_param(buf, PAGE_SIZE);
5161 static ssize_t pci_resource_alignment_store(struct bus_type *bus,
5162 const char *buf, size_t count)
5164 return pci_set_resource_alignment_param(buf, count);
5167 static BUS_ATTR(resource_alignment, 0644, pci_resource_alignment_show,
5168 pci_resource_alignment_store);
5170 static int __init pci_resource_alignment_sysfs_init(void)
5172 return bus_create_file(&pci_bus_type,
5173 &bus_attr_resource_alignment);
5175 late_initcall(pci_resource_alignment_sysfs_init);
5177 static void pci_no_domains(void)
5179 #ifdef CONFIG_PCI_DOMAINS
5180 pci_domains_supported = 0;
5184 #ifdef CONFIG_PCI_DOMAINS
5185 static atomic_t __domain_nr = ATOMIC_INIT(-1);
5187 int pci_get_new_domain_nr(void)
5189 return atomic_inc_return(&__domain_nr);
5192 #ifdef CONFIG_PCI_DOMAINS_GENERIC
5193 static int of_pci_bus_find_domain_nr(struct device *parent)
5195 static int use_dt_domains = -1;
5199 domain = of_get_pci_domain_nr(parent->of_node);
5201 * Check DT domain and use_dt_domains values.
5203 * If DT domain property is valid (domain >= 0) and
5204 * use_dt_domains != 0, the DT assignment is valid since this means
5205 * we have not previously allocated a domain number by using
5206 * pci_get_new_domain_nr(); we should also update use_dt_domains to
5207 * 1, to indicate that we have just assigned a domain number from
5210 * If DT domain property value is not valid (ie domain < 0), and we
5211 * have not previously assigned a domain number from DT
5212 * (use_dt_domains != 1) we should assign a domain number by
5215 * pci_get_new_domain_nr()
5217 * API and update the use_dt_domains value to keep track of method we
5218 * are using to assign domain numbers (use_dt_domains = 0).
5220 * All other combinations imply we have a platform that is trying
5221 * to mix domain numbers obtained from DT and pci_get_new_domain_nr(),
5222 * which is a recipe for domain mishandling and it is prevented by
5223 * invalidating the domain value (domain = -1) and printing a
5224 * corresponding error.
5226 if (domain >= 0 && use_dt_domains) {
5228 } else if (domain < 0 && use_dt_domains != 1) {
5230 domain = pci_get_new_domain_nr();
5232 dev_err(parent, "Node %s has inconsistent \"linux,pci-domain\" property in DT\n",
5233 parent->of_node->full_name);
5240 int pci_bus_find_domain_nr(struct pci_bus *bus, struct device *parent)
5242 return acpi_disabled ? of_pci_bus_find_domain_nr(parent) :
5243 acpi_pci_bus_find_domain_nr(bus);
5249 * pci_ext_cfg_avail - can we access extended PCI config space?
5251 * Returns 1 if we can access PCI extended config space (offsets
5252 * greater than 0xff). This is the default implementation. Architecture
5253 * implementations can override this.
5255 int __weak pci_ext_cfg_avail(void)
5260 void __weak pci_fixup_cardbus(struct pci_bus *bus)
5263 EXPORT_SYMBOL(pci_fixup_cardbus);
5265 static int __init pci_setup(char *str)
5268 char *k = strchr(str, ',');
5271 if (*str && (str = pcibios_setup(str)) && *str) {
5272 if (!strcmp(str, "nomsi")) {
5274 } else if (!strcmp(str, "noaer")) {
5276 } else if (!strncmp(str, "realloc=", 8)) {
5277 pci_realloc_get_opt(str + 8);
5278 } else if (!strncmp(str, "realloc", 7)) {
5279 pci_realloc_get_opt("on");
5280 } else if (!strcmp(str, "nodomains")) {
5282 } else if (!strncmp(str, "noari", 5)) {
5283 pcie_ari_disabled = true;
5284 } else if (!strncmp(str, "cbiosize=", 9)) {
5285 pci_cardbus_io_size = memparse(str + 9, &str);
5286 } else if (!strncmp(str, "cbmemsize=", 10)) {
5287 pci_cardbus_mem_size = memparse(str + 10, &str);
5288 } else if (!strncmp(str, "resource_alignment=", 19)) {
5289 pci_set_resource_alignment_param(str + 19,
5291 } else if (!strncmp(str, "ecrc=", 5)) {
5292 pcie_ecrc_get_policy(str + 5);
5293 } else if (!strncmp(str, "hpiosize=", 9)) {
5294 pci_hotplug_io_size = memparse(str + 9, &str);
5295 } else if (!strncmp(str, "hpmemsize=", 10)) {
5296 pci_hotplug_mem_size = memparse(str + 10, &str);
5297 } else if (!strncmp(str, "hpbussize=", 10)) {
5298 pci_hotplug_bus_size =
5299 simple_strtoul(str + 10, &str, 0);
5300 if (pci_hotplug_bus_size > 0xff)
5301 pci_hotplug_bus_size = DEFAULT_HOTPLUG_BUS_SIZE;
5302 } else if (!strncmp(str, "pcie_bus_tune_off", 17)) {
5303 pcie_bus_config = PCIE_BUS_TUNE_OFF;
5304 } else if (!strncmp(str, "pcie_bus_safe", 13)) {
5305 pcie_bus_config = PCIE_BUS_SAFE;
5306 } else if (!strncmp(str, "pcie_bus_perf", 13)) {
5307 pcie_bus_config = PCIE_BUS_PERFORMANCE;
5308 } else if (!strncmp(str, "pcie_bus_peer2peer", 18)) {
5309 pcie_bus_config = PCIE_BUS_PEER2PEER;
5310 } else if (!strncmp(str, "pcie_scan_all", 13)) {
5311 pci_add_flags(PCI_SCAN_ALL_PCIE_DEVS);
5313 printk(KERN_ERR "PCI: Unknown option `%s'\n",
5321 early_param("pci", pci_setup);