xref: /linux/drivers/pci/pci.c (revision bd628c1bed7902ec1f24ba0fe70758949146abbe)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * PCI Bus Services, see include/linux/pci.h for further explanation.
4  *
5  * Copyright 1993 -- 1997 Drew Eckhardt, Frederic Potter,
6  * David Mosberger-Tang
7  *
8  * Copyright 1997 -- 2000 Martin Mares <mj@ucw.cz>
9  */
10 
11 #include <linux/acpi.h>
12 #include <linux/kernel.h>
13 #include <linux/delay.h>
14 #include <linux/dmi.h>
15 #include <linux/init.h>
16 #include <linux/of.h>
17 #include <linux/of_pci.h>
18 #include <linux/pci.h>
19 #include <linux/pm.h>
20 #include <linux/slab.h>
21 #include <linux/module.h>
22 #include <linux/spinlock.h>
23 #include <linux/string.h>
24 #include <linux/log2.h>
25 #include <linux/logic_pio.h>
26 #include <linux/pm_wakeup.h>
27 #include <linux/interrupt.h>
28 #include <linux/device.h>
29 #include <linux/pm_runtime.h>
30 #include <linux/pci_hotplug.h>
31 #include <linux/vmalloc.h>
32 #include <linux/pci-ats.h>
33 #include <asm/setup.h>
34 #include <asm/dma.h>
35 #include <linux/aer.h>
36 #include "pci.h"
37 
38 DEFINE_MUTEX(pci_slot_mutex);
39 
40 const char *pci_power_names[] = {
41 	"error", "D0", "D1", "D2", "D3hot", "D3cold", "unknown",
42 };
43 EXPORT_SYMBOL_GPL(pci_power_names);
44 
45 int isa_dma_bridge_buggy;
46 EXPORT_SYMBOL(isa_dma_bridge_buggy);
47 
48 int pci_pci_problems;
49 EXPORT_SYMBOL(pci_pci_problems);
50 
51 unsigned int pci_pm_d3_delay;
52 
53 static void pci_pme_list_scan(struct work_struct *work);
54 
55 static LIST_HEAD(pci_pme_list);
56 static DEFINE_MUTEX(pci_pme_list_mutex);
57 static DECLARE_DELAYED_WORK(pci_pme_work, pci_pme_list_scan);
58 
59 struct pci_pme_device {
60 	struct list_head list;
61 	struct pci_dev *dev;
62 };
63 
64 #define PME_TIMEOUT 1000 /* How long between PME checks */
65 
66 static void pci_dev_d3_sleep(struct pci_dev *dev)
67 {
68 	unsigned int delay = dev->d3_delay;
69 
70 	if (delay < pci_pm_d3_delay)
71 		delay = pci_pm_d3_delay;
72 
73 	if (delay)
74 		msleep(delay);
75 }
76 
77 #ifdef CONFIG_PCI_DOMAINS
78 int pci_domains_supported = 1;
79 #endif
80 
81 #define DEFAULT_CARDBUS_IO_SIZE		(256)
82 #define DEFAULT_CARDBUS_MEM_SIZE	(64*1024*1024)
83 /* pci=cbmemsize=nnM,cbiosize=nn can override this */
84 unsigned long pci_cardbus_io_size = DEFAULT_CARDBUS_IO_SIZE;
85 unsigned long pci_cardbus_mem_size = DEFAULT_CARDBUS_MEM_SIZE;
86 
87 #define DEFAULT_HOTPLUG_IO_SIZE		(256)
88 #define DEFAULT_HOTPLUG_MEM_SIZE	(2*1024*1024)
89 /* pci=hpmemsize=nnM,hpiosize=nn can override this */
90 unsigned long pci_hotplug_io_size  = DEFAULT_HOTPLUG_IO_SIZE;
91 unsigned long pci_hotplug_mem_size = DEFAULT_HOTPLUG_MEM_SIZE;
92 
93 #define DEFAULT_HOTPLUG_BUS_SIZE	1
94 unsigned long pci_hotplug_bus_size = DEFAULT_HOTPLUG_BUS_SIZE;
95 
96 enum pcie_bus_config_types pcie_bus_config = PCIE_BUS_DEFAULT;
97 
98 /*
99  * The default CLS is used if arch didn't set CLS explicitly and not
100  * all pci devices agree on the same value.  Arch can override either
101  * the dfl or actual value as it sees fit.  Don't forget this is
102  * measured in 32-bit words, not bytes.
103  */
104 u8 pci_dfl_cache_line_size = L1_CACHE_BYTES >> 2;
105 u8 pci_cache_line_size;
106 
107 /*
108  * If we set up a device for bus mastering, we need to check the latency
109  * timer as certain BIOSes forget to set it properly.
110  */
111 unsigned int pcibios_max_latency = 255;
112 
113 /* If set, the PCIe ARI capability will not be used. */
114 static bool pcie_ari_disabled;
115 
116 /* If set, the PCIe ATS capability will not be used. */
117 static bool pcie_ats_disabled;
118 
119 /* If set, the PCI config space of each device is printed during boot. */
120 bool pci_early_dump;
121 
122 bool pci_ats_disabled(void)
123 {
124 	return pcie_ats_disabled;
125 }
126 
127 /* Disable bridge_d3 for all PCIe ports */
128 static bool pci_bridge_d3_disable;
129 /* Force bridge_d3 for all PCIe ports */
130 static bool pci_bridge_d3_force;
131 
132 static int __init pcie_port_pm_setup(char *str)
133 {
134 	if (!strcmp(str, "off"))
135 		pci_bridge_d3_disable = true;
136 	else if (!strcmp(str, "force"))
137 		pci_bridge_d3_force = true;
138 	return 1;
139 }
140 __setup("pcie_port_pm=", pcie_port_pm_setup);
141 
142 /* Time to wait after a reset for device to become responsive */
143 #define PCIE_RESET_READY_POLL_MS 60000
144 
145 /**
146  * pci_bus_max_busnr - returns maximum PCI bus number of given bus' children
147  * @bus: pointer to PCI bus structure to search
148  *
149  * Given a PCI bus, returns the highest PCI bus number present in the set
150  * including the given PCI bus and its list of child PCI buses.
151  */
152 unsigned char pci_bus_max_busnr(struct pci_bus *bus)
153 {
154 	struct pci_bus *tmp;
155 	unsigned char max, n;
156 
157 	max = bus->busn_res.end;
158 	list_for_each_entry(tmp, &bus->children, node) {
159 		n = pci_bus_max_busnr(tmp);
160 		if (n > max)
161 			max = n;
162 	}
163 	return max;
164 }
165 EXPORT_SYMBOL_GPL(pci_bus_max_busnr);
166 
167 #ifdef CONFIG_HAS_IOMEM
168 void __iomem *pci_ioremap_bar(struct pci_dev *pdev, int bar)
169 {
170 	struct resource *res = &pdev->resource[bar];
171 
172 	/*
173 	 * Make sure the BAR is actually a memory resource, not an IO resource
174 	 */
175 	if (res->flags & IORESOURCE_UNSET || !(res->flags & IORESOURCE_MEM)) {
176 		pci_warn(pdev, "can't ioremap BAR %d: %pR\n", bar, res);
177 		return NULL;
178 	}
179 	return ioremap_nocache(res->start, resource_size(res));
180 }
181 EXPORT_SYMBOL_GPL(pci_ioremap_bar);
182 
183 void __iomem *pci_ioremap_wc_bar(struct pci_dev *pdev, int bar)
184 {
185 	/*
186 	 * Make sure the BAR is actually a memory resource, not an IO resource
187 	 */
188 	if (!(pci_resource_flags(pdev, bar) & IORESOURCE_MEM)) {
189 		WARN_ON(1);
190 		return NULL;
191 	}
192 	return ioremap_wc(pci_resource_start(pdev, bar),
193 			  pci_resource_len(pdev, bar));
194 }
195 EXPORT_SYMBOL_GPL(pci_ioremap_wc_bar);
196 #endif
197 
198 /**
199  * pci_dev_str_match_path - test if a path string matches a device
200  * @dev:    the PCI device to test
201  * @path:   string to match the device against
202  * @endptr: pointer to the string after the match
203  *
204  * Test if a string (typically from a kernel parameter) formatted as a
205  * path of device/function addresses matches a PCI device. The string must
206  * be of the form:
207  *
208  *   [<domain>:]<bus>:<device>.<func>[/<device>.<func>]*
209  *
210  * A path for a device can be obtained using 'lspci -t'.  Using a path
211  * is more robust against bus renumbering than using only a single bus,
212  * device and function address.
213  *
214  * Returns 1 if the string matches the device, 0 if it does not and
215  * a negative error code if it fails to parse the string.
216  */
217 static int pci_dev_str_match_path(struct pci_dev *dev, const char *path,
218 				  const char **endptr)
219 {
220 	int ret;
221 	int seg, bus, slot, func;
222 	char *wpath, *p;
223 	char end;
224 
225 	*endptr = strchrnul(path, ';');
226 
227 	wpath = kmemdup_nul(path, *endptr - path, GFP_KERNEL);
228 	if (!wpath)
229 		return -ENOMEM;
230 
231 	while (1) {
232 		p = strrchr(wpath, '/');
233 		if (!p)
234 			break;
235 		ret = sscanf(p, "/%x.%x%c", &slot, &func, &end);
236 		if (ret != 2) {
237 			ret = -EINVAL;
238 			goto free_and_exit;
239 		}
240 
241 		if (dev->devfn != PCI_DEVFN(slot, func)) {
242 			ret = 0;
243 			goto free_and_exit;
244 		}
245 
246 		/*
247 		 * Note: we don't need to get a reference to the upstream
248 		 * bridge because we hold a reference to the top level
249 		 * device which should hold a reference to the bridge,
250 		 * and so on.
251 		 */
252 		dev = pci_upstream_bridge(dev);
253 		if (!dev) {
254 			ret = 0;
255 			goto free_and_exit;
256 		}
257 
258 		*p = 0;
259 	}
260 
261 	ret = sscanf(wpath, "%x:%x:%x.%x%c", &seg, &bus, &slot,
262 		     &func, &end);
263 	if (ret != 4) {
264 		seg = 0;
265 		ret = sscanf(wpath, "%x:%x.%x%c", &bus, &slot, &func, &end);
266 		if (ret != 3) {
267 			ret = -EINVAL;
268 			goto free_and_exit;
269 		}
270 	}
271 
272 	ret = (seg == pci_domain_nr(dev->bus) &&
273 	       bus == dev->bus->number &&
274 	       dev->devfn == PCI_DEVFN(slot, func));
275 
276 free_and_exit:
277 	kfree(wpath);
278 	return ret;
279 }
280 
281 /**
282  * pci_dev_str_match - test if a string matches a device
283  * @dev:    the PCI device to test
284  * @p:      string to match the device against
285  * @endptr: pointer to the string after the match
286  *
287  * Test if a string (typically from a kernel parameter) matches a specified
288  * PCI device. The string may be of one of the following formats:
289  *
290  *   [<domain>:]<bus>:<device>.<func>[/<device>.<func>]*
291  *   pci:<vendor>:<device>[:<subvendor>:<subdevice>]
292  *
293  * The first format specifies a PCI bus/device/function address which
294  * may change if new hardware is inserted, if motherboard firmware changes,
295  * or due to changes caused in kernel parameters. If the domain is
296  * left unspecified, it is taken to be 0.  In order to be robust against
297  * bus renumbering issues, a path of PCI device/function numbers may be used
298  * to address the specific device.  The path for a device can be determined
299  * through the use of 'lspci -t'.
300  *
301  * The second format matches devices using IDs in the configuration
302  * space which may match multiple devices in the system. A value of 0
303  * for any field will match all devices. (Note: this differs from
304  * in-kernel code that uses PCI_ANY_ID which is ~0; this is for
305  * legacy reasons and convenience so users don't have to specify
306  * FFFFFFFFs on the command line.)
307  *
308  * Returns 1 if the string matches the device, 0 if it does not and
309  * a negative error code if the string cannot be parsed.
310  */
311 static int pci_dev_str_match(struct pci_dev *dev, const char *p,
312 			     const char **endptr)
313 {
314 	int ret;
315 	int count;
316 	unsigned short vendor, device, subsystem_vendor, subsystem_device;
317 
318 	if (strncmp(p, "pci:", 4) == 0) {
319 		/* PCI vendor/device (subvendor/subdevice) IDs are specified */
320 		p += 4;
321 		ret = sscanf(p, "%hx:%hx:%hx:%hx%n", &vendor, &device,
322 			     &subsystem_vendor, &subsystem_device, &count);
323 		if (ret != 4) {
324 			ret = sscanf(p, "%hx:%hx%n", &vendor, &device, &count);
325 			if (ret != 2)
326 				return -EINVAL;
327 
328 			subsystem_vendor = 0;
329 			subsystem_device = 0;
330 		}
331 
332 		p += count;
333 
334 		if ((!vendor || vendor == dev->vendor) &&
335 		    (!device || device == dev->device) &&
336 		    (!subsystem_vendor ||
337 			    subsystem_vendor == dev->subsystem_vendor) &&
338 		    (!subsystem_device ||
339 			    subsystem_device == dev->subsystem_device))
340 			goto found;
341 	} else {
342 		/*
343 		 * PCI Bus, Device, Function IDs are specified
344 		 *  (optionally, may include a path of devfns following it)
345 		 */
346 		ret = pci_dev_str_match_path(dev, p, &p);
347 		if (ret < 0)
348 			return ret;
349 		else if (ret)
350 			goto found;
351 	}
352 
353 	*endptr = p;
354 	return 0;
355 
356 found:
357 	*endptr = p;
358 	return 1;
359 }
360 
361 static int __pci_find_next_cap_ttl(struct pci_bus *bus, unsigned int devfn,
362 				   u8 pos, int cap, int *ttl)
363 {
364 	u8 id;
365 	u16 ent;
366 
367 	pci_bus_read_config_byte(bus, devfn, pos, &pos);
368 
369 	while ((*ttl)--) {
370 		if (pos < 0x40)
371 			break;
372 		pos &= ~3;
373 		pci_bus_read_config_word(bus, devfn, pos, &ent);
374 
375 		id = ent & 0xff;
376 		if (id == 0xff)
377 			break;
378 		if (id == cap)
379 			return pos;
380 		pos = (ent >> 8);
381 	}
382 	return 0;
383 }
384 
385 static int __pci_find_next_cap(struct pci_bus *bus, unsigned int devfn,
386 			       u8 pos, int cap)
387 {
388 	int ttl = PCI_FIND_CAP_TTL;
389 
390 	return __pci_find_next_cap_ttl(bus, devfn, pos, cap, &ttl);
391 }
392 
393 int pci_find_next_capability(struct pci_dev *dev, u8 pos, int cap)
394 {
395 	return __pci_find_next_cap(dev->bus, dev->devfn,
396 				   pos + PCI_CAP_LIST_NEXT, cap);
397 }
398 EXPORT_SYMBOL_GPL(pci_find_next_capability);
399 
400 static int __pci_bus_find_cap_start(struct pci_bus *bus,
401 				    unsigned int devfn, u8 hdr_type)
402 {
403 	u16 status;
404 
405 	pci_bus_read_config_word(bus, devfn, PCI_STATUS, &status);
406 	if (!(status & PCI_STATUS_CAP_LIST))
407 		return 0;
408 
409 	switch (hdr_type) {
410 	case PCI_HEADER_TYPE_NORMAL:
411 	case PCI_HEADER_TYPE_BRIDGE:
412 		return PCI_CAPABILITY_LIST;
413 	case PCI_HEADER_TYPE_CARDBUS:
414 		return PCI_CB_CAPABILITY_LIST;
415 	}
416 
417 	return 0;
418 }
419 
420 /**
421  * pci_find_capability - query for devices' capabilities
422  * @dev: PCI device to query
423  * @cap: capability code
424  *
425  * Tell if a device supports a given PCI capability.
426  * Returns the address of the requested capability structure within the
427  * device's PCI configuration space or 0 in case the device does not
428  * support it.  Possible values for @cap:
429  *
430  *  %PCI_CAP_ID_PM           Power Management
431  *  %PCI_CAP_ID_AGP          Accelerated Graphics Port
432  *  %PCI_CAP_ID_VPD          Vital Product Data
433  *  %PCI_CAP_ID_SLOTID       Slot Identification
434  *  %PCI_CAP_ID_MSI          Message Signalled Interrupts
435  *  %PCI_CAP_ID_CHSWP        CompactPCI HotSwap
436  *  %PCI_CAP_ID_PCIX         PCI-X
437  *  %PCI_CAP_ID_EXP          PCI Express
438  */
439 int pci_find_capability(struct pci_dev *dev, int cap)
440 {
441 	int pos;
442 
443 	pos = __pci_bus_find_cap_start(dev->bus, dev->devfn, dev->hdr_type);
444 	if (pos)
445 		pos = __pci_find_next_cap(dev->bus, dev->devfn, pos, cap);
446 
447 	return pos;
448 }
449 EXPORT_SYMBOL(pci_find_capability);
450 
451 /**
452  * pci_bus_find_capability - query for devices' capabilities
453  * @bus:   the PCI bus to query
454  * @devfn: PCI device to query
455  * @cap:   capability code
456  *
457  * Like pci_find_capability() but works for pci devices that do not have a
458  * pci_dev structure set up yet.
459  *
460  * Returns the address of the requested capability structure within the
461  * device's PCI configuration space or 0 in case the device does not
462  * support it.
463  */
464 int pci_bus_find_capability(struct pci_bus *bus, unsigned int devfn, int cap)
465 {
466 	int pos;
467 	u8 hdr_type;
468 
469 	pci_bus_read_config_byte(bus, devfn, PCI_HEADER_TYPE, &hdr_type);
470 
471 	pos = __pci_bus_find_cap_start(bus, devfn, hdr_type & 0x7f);
472 	if (pos)
473 		pos = __pci_find_next_cap(bus, devfn, pos, cap);
474 
475 	return pos;
476 }
477 EXPORT_SYMBOL(pci_bus_find_capability);
478 
479 /**
480  * pci_find_next_ext_capability - Find an extended capability
481  * @dev: PCI device to query
482  * @start: address at which to start looking (0 to start at beginning of list)
483  * @cap: capability code
484  *
485  * Returns the address of the next matching extended capability structure
486  * within the device's PCI configuration space or 0 if the device does
487  * not support it.  Some capabilities can occur several times, e.g., the
488  * vendor-specific capability, and this provides a way to find them all.
489  */
490 int pci_find_next_ext_capability(struct pci_dev *dev, int start, int cap)
491 {
492 	u32 header;
493 	int ttl;
494 	int pos = PCI_CFG_SPACE_SIZE;
495 
496 	/* minimum 8 bytes per capability */
497 	ttl = (PCI_CFG_SPACE_EXP_SIZE - PCI_CFG_SPACE_SIZE) / 8;
498 
499 	if (dev->cfg_size <= PCI_CFG_SPACE_SIZE)
500 		return 0;
501 
502 	if (start)
503 		pos = start;
504 
505 	if (pci_read_config_dword(dev, pos, &header) != PCIBIOS_SUCCESSFUL)
506 		return 0;
507 
508 	/*
509 	 * If we have no capabilities, this is indicated by cap ID,
510 	 * cap version and next pointer all being 0.
511 	 */
512 	if (header == 0)
513 		return 0;
514 
515 	while (ttl-- > 0) {
516 		if (PCI_EXT_CAP_ID(header) == cap && pos != start)
517 			return pos;
518 
519 		pos = PCI_EXT_CAP_NEXT(header);
520 		if (pos < PCI_CFG_SPACE_SIZE)
521 			break;
522 
523 		if (pci_read_config_dword(dev, pos, &header) != PCIBIOS_SUCCESSFUL)
524 			break;
525 	}
526 
527 	return 0;
528 }
529 EXPORT_SYMBOL_GPL(pci_find_next_ext_capability);
530 
531 /**
532  * pci_find_ext_capability - Find an extended capability
533  * @dev: PCI device to query
534  * @cap: capability code
535  *
536  * Returns the address of the requested extended capability structure
537  * within the device's PCI configuration space or 0 if the device does
538  * not support it.  Possible values for @cap:
539  *
540  *  %PCI_EXT_CAP_ID_ERR		Advanced Error Reporting
541  *  %PCI_EXT_CAP_ID_VC		Virtual Channel
542  *  %PCI_EXT_CAP_ID_DSN		Device Serial Number
543  *  %PCI_EXT_CAP_ID_PWR		Power Budgeting
544  */
545 int pci_find_ext_capability(struct pci_dev *dev, int cap)
546 {
547 	return pci_find_next_ext_capability(dev, 0, cap);
548 }
549 EXPORT_SYMBOL_GPL(pci_find_ext_capability);
550 
551 static int __pci_find_next_ht_cap(struct pci_dev *dev, int pos, int ht_cap)
552 {
553 	int rc, ttl = PCI_FIND_CAP_TTL;
554 	u8 cap, mask;
555 
556 	if (ht_cap == HT_CAPTYPE_SLAVE || ht_cap == HT_CAPTYPE_HOST)
557 		mask = HT_3BIT_CAP_MASK;
558 	else
559 		mask = HT_5BIT_CAP_MASK;
560 
561 	pos = __pci_find_next_cap_ttl(dev->bus, dev->devfn, pos,
562 				      PCI_CAP_ID_HT, &ttl);
563 	while (pos) {
564 		rc = pci_read_config_byte(dev, pos + 3, &cap);
565 		if (rc != PCIBIOS_SUCCESSFUL)
566 			return 0;
567 
568 		if ((cap & mask) == ht_cap)
569 			return pos;
570 
571 		pos = __pci_find_next_cap_ttl(dev->bus, dev->devfn,
572 					      pos + PCI_CAP_LIST_NEXT,
573 					      PCI_CAP_ID_HT, &ttl);
574 	}
575 
576 	return 0;
577 }
578 /**
579  * pci_find_next_ht_capability - query a device's Hypertransport capabilities
580  * @dev: PCI device to query
581  * @pos: Position from which to continue searching
582  * @ht_cap: Hypertransport capability code
583  *
584  * To be used in conjunction with pci_find_ht_capability() to search for
585  * all capabilities matching @ht_cap. @pos should always be a value returned
586  * from pci_find_ht_capability().
587  *
588  * NB. To be 100% safe against broken PCI devices, the caller should take
589  * steps to avoid an infinite loop.
590  */
591 int pci_find_next_ht_capability(struct pci_dev *dev, int pos, int ht_cap)
592 {
593 	return __pci_find_next_ht_cap(dev, pos + PCI_CAP_LIST_NEXT, ht_cap);
594 }
595 EXPORT_SYMBOL_GPL(pci_find_next_ht_capability);
596 
597 /**
598  * pci_find_ht_capability - query a device's Hypertransport capabilities
599  * @dev: PCI device to query
600  * @ht_cap: Hypertransport capability code
601  *
602  * Tell if a device supports a given Hypertransport capability.
603  * Returns an address within the device's PCI configuration space
604  * or 0 in case the device does not support the request capability.
605  * The address points to the PCI capability, of type PCI_CAP_ID_HT,
606  * which has a Hypertransport capability matching @ht_cap.
607  */
608 int pci_find_ht_capability(struct pci_dev *dev, int ht_cap)
609 {
610 	int pos;
611 
612 	pos = __pci_bus_find_cap_start(dev->bus, dev->devfn, dev->hdr_type);
613 	if (pos)
614 		pos = __pci_find_next_ht_cap(dev, pos, ht_cap);
615 
616 	return pos;
617 }
618 EXPORT_SYMBOL_GPL(pci_find_ht_capability);
619 
620 /**
621  * pci_find_parent_resource - return resource region of parent bus of given region
622  * @dev: PCI device structure contains resources to be searched
623  * @res: child resource record for which parent is sought
624  *
625  *  For given resource region of given device, return the resource
626  *  region of parent bus the given region is contained in.
627  */
628 struct resource *pci_find_parent_resource(const struct pci_dev *dev,
629 					  struct resource *res)
630 {
631 	const struct pci_bus *bus = dev->bus;
632 	struct resource *r;
633 	int i;
634 
635 	pci_bus_for_each_resource(bus, r, i) {
636 		if (!r)
637 			continue;
638 		if (resource_contains(r, res)) {
639 
640 			/*
641 			 * If the window is prefetchable but the BAR is
642 			 * not, the allocator made a mistake.
643 			 */
644 			if (r->flags & IORESOURCE_PREFETCH &&
645 			    !(res->flags & IORESOURCE_PREFETCH))
646 				return NULL;
647 
648 			/*
649 			 * If we're below a transparent bridge, there may
650 			 * be both a positively-decoded aperture and a
651 			 * subtractively-decoded region that contain the BAR.
652 			 * We want the positively-decoded one, so this depends
653 			 * on pci_bus_for_each_resource() giving us those
654 			 * first.
655 			 */
656 			return r;
657 		}
658 	}
659 	return NULL;
660 }
661 EXPORT_SYMBOL(pci_find_parent_resource);
662 
663 /**
664  * pci_find_resource - Return matching PCI device resource
665  * @dev: PCI device to query
666  * @res: Resource to look for
667  *
668  * Goes over standard PCI resources (BARs) and checks if the given resource
669  * is partially or fully contained in any of them. In that case the
670  * matching resource is returned, %NULL otherwise.
671  */
672 struct resource *pci_find_resource(struct pci_dev *dev, struct resource *res)
673 {
674 	int i;
675 
676 	for (i = 0; i < PCI_ROM_RESOURCE; i++) {
677 		struct resource *r = &dev->resource[i];
678 
679 		if (r->start && resource_contains(r, res))
680 			return r;
681 	}
682 
683 	return NULL;
684 }
685 EXPORT_SYMBOL(pci_find_resource);
686 
687 /**
688  * pci_find_pcie_root_port - return PCIe Root Port
689  * @dev: PCI device to query
690  *
691  * Traverse up the parent chain and return the PCIe Root Port PCI Device
692  * for a given PCI Device.
693  */
694 struct pci_dev *pci_find_pcie_root_port(struct pci_dev *dev)
695 {
696 	struct pci_dev *bridge, *highest_pcie_bridge = dev;
697 
698 	bridge = pci_upstream_bridge(dev);
699 	while (bridge && pci_is_pcie(bridge)) {
700 		highest_pcie_bridge = bridge;
701 		bridge = pci_upstream_bridge(bridge);
702 	}
703 
704 	if (pci_pcie_type(highest_pcie_bridge) != PCI_EXP_TYPE_ROOT_PORT)
705 		return NULL;
706 
707 	return highest_pcie_bridge;
708 }
709 EXPORT_SYMBOL(pci_find_pcie_root_port);
710 
711 /**
712  * pci_wait_for_pending - wait for @mask bit(s) to clear in status word @pos
713  * @dev: the PCI device to operate on
714  * @pos: config space offset of status word
715  * @mask: mask of bit(s) to care about in status word
716  *
717  * Return 1 when mask bit(s) in status word clear, 0 otherwise.
718  */
719 int pci_wait_for_pending(struct pci_dev *dev, int pos, u16 mask)
720 {
721 	int i;
722 
723 	/* Wait for Transaction Pending bit clean */
724 	for (i = 0; i < 4; i++) {
725 		u16 status;
726 		if (i)
727 			msleep((1 << (i - 1)) * 100);
728 
729 		pci_read_config_word(dev, pos, &status);
730 		if (!(status & mask))
731 			return 1;
732 	}
733 
734 	return 0;
735 }
736 
737 /**
738  * pci_restore_bars - restore a device's BAR values (e.g. after wake-up)
739  * @dev: PCI device to have its BARs restored
740  *
741  * Restore the BAR values for a given device, so as to make it
742  * accessible by its driver.
743  */
744 static void pci_restore_bars(struct pci_dev *dev)
745 {
746 	int i;
747 
748 	for (i = 0; i < PCI_BRIDGE_RESOURCES; i++)
749 		pci_update_resource(dev, i);
750 }
751 
752 static const struct pci_platform_pm_ops *pci_platform_pm;
753 
754 int pci_set_platform_pm(const struct pci_platform_pm_ops *ops)
755 {
756 	if (!ops->is_manageable || !ops->set_state  || !ops->get_state ||
757 	    !ops->choose_state  || !ops->set_wakeup || !ops->need_resume)
758 		return -EINVAL;
759 	pci_platform_pm = ops;
760 	return 0;
761 }
762 
763 static inline bool platform_pci_power_manageable(struct pci_dev *dev)
764 {
765 	return pci_platform_pm ? pci_platform_pm->is_manageable(dev) : false;
766 }
767 
768 static inline int platform_pci_set_power_state(struct pci_dev *dev,
769 					       pci_power_t t)
770 {
771 	return pci_platform_pm ? pci_platform_pm->set_state(dev, t) : -ENOSYS;
772 }
773 
774 static inline pci_power_t platform_pci_get_power_state(struct pci_dev *dev)
775 {
776 	return pci_platform_pm ? pci_platform_pm->get_state(dev) : PCI_UNKNOWN;
777 }
778 
779 static inline pci_power_t platform_pci_choose_state(struct pci_dev *dev)
780 {
781 	return pci_platform_pm ?
782 			pci_platform_pm->choose_state(dev) : PCI_POWER_ERROR;
783 }
784 
785 static inline int platform_pci_set_wakeup(struct pci_dev *dev, bool enable)
786 {
787 	return pci_platform_pm ?
788 			pci_platform_pm->set_wakeup(dev, enable) : -ENODEV;
789 }
790 
791 static inline bool platform_pci_need_resume(struct pci_dev *dev)
792 {
793 	return pci_platform_pm ? pci_platform_pm->need_resume(dev) : false;
794 }
795 
796 static inline bool platform_pci_bridge_d3(struct pci_dev *dev)
797 {
798 	return pci_platform_pm ? pci_platform_pm->bridge_d3(dev) : false;
799 }
800 
801 /**
802  * pci_raw_set_power_state - Use PCI PM registers to set the power state of
803  *                           given PCI device
804  * @dev: PCI device to handle.
805  * @state: PCI power state (D0, D1, D2, D3hot) to put the device into.
806  *
807  * RETURN VALUE:
808  * -EINVAL if the requested state is invalid.
809  * -EIO if device does not support PCI PM or its PM capabilities register has a
810  * wrong version, or device doesn't support the requested state.
811  * 0 if device already is in the requested state.
812  * 0 if device's power state has been successfully changed.
813  */
814 static int pci_raw_set_power_state(struct pci_dev *dev, pci_power_t state)
815 {
816 	u16 pmcsr;
817 	bool need_restore = false;
818 
819 	/* Check if we're already there */
820 	if (dev->current_state == state)
821 		return 0;
822 
823 	if (!dev->pm_cap)
824 		return -EIO;
825 
826 	if (state < PCI_D0 || state > PCI_D3hot)
827 		return -EINVAL;
828 
829 	/* Validate current state:
830 	 * Can enter D0 from any state, but if we can only go deeper
831 	 * to sleep if we're already in a low power state
832 	 */
833 	if (state != PCI_D0 && dev->current_state <= PCI_D3cold
834 	    && dev->current_state > state) {
835 		pci_err(dev, "invalid power transition (from state %d to %d)\n",
836 			dev->current_state, state);
837 		return -EINVAL;
838 	}
839 
840 	/* check if this device supports the desired state */
841 	if ((state == PCI_D1 && !dev->d1_support)
842 	   || (state == PCI_D2 && !dev->d2_support))
843 		return -EIO;
844 
845 	pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
846 
847 	/* If we're (effectively) in D3, force entire word to 0.
848 	 * This doesn't affect PME_Status, disables PME_En, and
849 	 * sets PowerState to 0.
850 	 */
851 	switch (dev->current_state) {
852 	case PCI_D0:
853 	case PCI_D1:
854 	case PCI_D2:
855 		pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
856 		pmcsr |= state;
857 		break;
858 	case PCI_D3hot:
859 	case PCI_D3cold:
860 	case PCI_UNKNOWN: /* Boot-up */
861 		if ((pmcsr & PCI_PM_CTRL_STATE_MASK) == PCI_D3hot
862 		 && !(pmcsr & PCI_PM_CTRL_NO_SOFT_RESET))
863 			need_restore = true;
864 		/* Fall-through: force to D0 */
865 	default:
866 		pmcsr = 0;
867 		break;
868 	}
869 
870 	/* enter specified state */
871 	pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, pmcsr);
872 
873 	/* Mandatory power management transition delays */
874 	/* see PCI PM 1.1 5.6.1 table 18 */
875 	if (state == PCI_D3hot || dev->current_state == PCI_D3hot)
876 		pci_dev_d3_sleep(dev);
877 	else if (state == PCI_D2 || dev->current_state == PCI_D2)
878 		udelay(PCI_PM_D2_DELAY);
879 
880 	pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
881 	dev->current_state = (pmcsr & PCI_PM_CTRL_STATE_MASK);
882 	if (dev->current_state != state && printk_ratelimit())
883 		pci_info(dev, "Refused to change power state, currently in D%d\n",
884 			 dev->current_state);
885 
886 	/*
887 	 * According to section 5.4.1 of the "PCI BUS POWER MANAGEMENT
888 	 * INTERFACE SPECIFICATION, REV. 1.2", a device transitioning
889 	 * from D3hot to D0 _may_ perform an internal reset, thereby
890 	 * going to "D0 Uninitialized" rather than "D0 Initialized".
891 	 * For example, at least some versions of the 3c905B and the
892 	 * 3c556B exhibit this behaviour.
893 	 *
894 	 * At least some laptop BIOSen (e.g. the Thinkpad T21) leave
895 	 * devices in a D3hot state at boot.  Consequently, we need to
896 	 * restore at least the BARs so that the device will be
897 	 * accessible to its driver.
898 	 */
899 	if (need_restore)
900 		pci_restore_bars(dev);
901 
902 	if (dev->bus->self)
903 		pcie_aspm_pm_state_change(dev->bus->self);
904 
905 	return 0;
906 }
907 
908 /**
909  * pci_update_current_state - Read power state of given device and cache it
910  * @dev: PCI device to handle.
911  * @state: State to cache in case the device doesn't have the PM capability
912  *
913  * The power state is read from the PMCSR register, which however is
914  * inaccessible in D3cold.  The platform firmware is therefore queried first
915  * to detect accessibility of the register.  In case the platform firmware
916  * reports an incorrect state or the device isn't power manageable by the
917  * platform at all, we try to detect D3cold by testing accessibility of the
918  * vendor ID in config space.
919  */
920 void pci_update_current_state(struct pci_dev *dev, pci_power_t state)
921 {
922 	if (platform_pci_get_power_state(dev) == PCI_D3cold ||
923 	    !pci_device_is_present(dev)) {
924 		dev->current_state = PCI_D3cold;
925 	} else if (dev->pm_cap) {
926 		u16 pmcsr;
927 
928 		pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
929 		dev->current_state = (pmcsr & PCI_PM_CTRL_STATE_MASK);
930 	} else {
931 		dev->current_state = state;
932 	}
933 }
934 
935 /**
936  * pci_power_up - Put the given device into D0 forcibly
937  * @dev: PCI device to power up
938  */
939 void pci_power_up(struct pci_dev *dev)
940 {
941 	if (platform_pci_power_manageable(dev))
942 		platform_pci_set_power_state(dev, PCI_D0);
943 
944 	pci_raw_set_power_state(dev, PCI_D0);
945 	pci_update_current_state(dev, PCI_D0);
946 }
947 
948 /**
949  * pci_platform_power_transition - Use platform to change device power state
950  * @dev: PCI device to handle.
951  * @state: State to put the device into.
952  */
953 static int pci_platform_power_transition(struct pci_dev *dev, pci_power_t state)
954 {
955 	int error;
956 
957 	if (platform_pci_power_manageable(dev)) {
958 		error = platform_pci_set_power_state(dev, state);
959 		if (!error)
960 			pci_update_current_state(dev, state);
961 	} else
962 		error = -ENODEV;
963 
964 	if (error && !dev->pm_cap) /* Fall back to PCI_D0 */
965 		dev->current_state = PCI_D0;
966 
967 	return error;
968 }
969 
970 /**
971  * pci_wakeup - Wake up a PCI device
972  * @pci_dev: Device to handle.
973  * @ign: ignored parameter
974  */
975 static int pci_wakeup(struct pci_dev *pci_dev, void *ign)
976 {
977 	pci_wakeup_event(pci_dev);
978 	pm_request_resume(&pci_dev->dev);
979 	return 0;
980 }
981 
982 /**
983  * pci_wakeup_bus - Walk given bus and wake up devices on it
984  * @bus: Top bus of the subtree to walk.
985  */
986 void pci_wakeup_bus(struct pci_bus *bus)
987 {
988 	if (bus)
989 		pci_walk_bus(bus, pci_wakeup, NULL);
990 }
991 
992 /**
993  * __pci_start_power_transition - Start power transition of a PCI device
994  * @dev: PCI device to handle.
995  * @state: State to put the device into.
996  */
997 static void __pci_start_power_transition(struct pci_dev *dev, pci_power_t state)
998 {
999 	if (state == PCI_D0) {
1000 		pci_platform_power_transition(dev, PCI_D0);
1001 		/*
1002 		 * Mandatory power management transition delays, see
1003 		 * PCI Express Base Specification Revision 2.0 Section
1004 		 * 6.6.1: Conventional Reset.  Do not delay for
1005 		 * devices powered on/off by corresponding bridge,
1006 		 * because have already delayed for the bridge.
1007 		 */
1008 		if (dev->runtime_d3cold) {
1009 			if (dev->d3cold_delay && !dev->imm_ready)
1010 				msleep(dev->d3cold_delay);
1011 			/*
1012 			 * When powering on a bridge from D3cold, the
1013 			 * whole hierarchy may be powered on into
1014 			 * D0uninitialized state, resume them to give
1015 			 * them a chance to suspend again
1016 			 */
1017 			pci_wakeup_bus(dev->subordinate);
1018 		}
1019 	}
1020 }
1021 
1022 /**
1023  * __pci_dev_set_current_state - Set current state of a PCI device
1024  * @dev: Device to handle
1025  * @data: pointer to state to be set
1026  */
1027 static int __pci_dev_set_current_state(struct pci_dev *dev, void *data)
1028 {
1029 	pci_power_t state = *(pci_power_t *)data;
1030 
1031 	dev->current_state = state;
1032 	return 0;
1033 }
1034 
1035 /**
1036  * pci_bus_set_current_state - Walk given bus and set current state of devices
1037  * @bus: Top bus of the subtree to walk.
1038  * @state: state to be set
1039  */
1040 void pci_bus_set_current_state(struct pci_bus *bus, pci_power_t state)
1041 {
1042 	if (bus)
1043 		pci_walk_bus(bus, __pci_dev_set_current_state, &state);
1044 }
1045 
1046 /**
1047  * __pci_complete_power_transition - Complete power transition of a PCI device
1048  * @dev: PCI device to handle.
1049  * @state: State to put the device into.
1050  *
1051  * This function should not be called directly by device drivers.
1052  */
1053 int __pci_complete_power_transition(struct pci_dev *dev, pci_power_t state)
1054 {
1055 	int ret;
1056 
1057 	if (state <= PCI_D0)
1058 		return -EINVAL;
1059 	ret = pci_platform_power_transition(dev, state);
1060 	/* Power off the bridge may power off the whole hierarchy */
1061 	if (!ret && state == PCI_D3cold)
1062 		pci_bus_set_current_state(dev->subordinate, PCI_D3cold);
1063 	return ret;
1064 }
1065 EXPORT_SYMBOL_GPL(__pci_complete_power_transition);
1066 
1067 /**
1068  * pci_set_power_state - Set the power state of a PCI device
1069  * @dev: PCI device to handle.
1070  * @state: PCI power state (D0, D1, D2, D3hot) to put the device into.
1071  *
1072  * Transition a device to a new power state, using the platform firmware and/or
1073  * the device's PCI PM registers.
1074  *
1075  * RETURN VALUE:
1076  * -EINVAL if the requested state is invalid.
1077  * -EIO if device does not support PCI PM or its PM capabilities register has a
1078  * wrong version, or device doesn't support the requested state.
1079  * 0 if the transition is to D1 or D2 but D1 and D2 are not supported.
1080  * 0 if device already is in the requested state.
1081  * 0 if the transition is to D3 but D3 is not supported.
1082  * 0 if device's power state has been successfully changed.
1083  */
1084 int pci_set_power_state(struct pci_dev *dev, pci_power_t state)
1085 {
1086 	int error;
1087 
1088 	/* bound the state we're entering */
1089 	if (state > PCI_D3cold)
1090 		state = PCI_D3cold;
1091 	else if (state < PCI_D0)
1092 		state = PCI_D0;
1093 	else if ((state == PCI_D1 || state == PCI_D2) && pci_no_d1d2(dev))
1094 		/*
1095 		 * If the device or the parent bridge do not support PCI PM,
1096 		 * ignore the request if we're doing anything other than putting
1097 		 * it into D0 (which would only happen on boot).
1098 		 */
1099 		return 0;
1100 
1101 	/* Check if we're already there */
1102 	if (dev->current_state == state)
1103 		return 0;
1104 
1105 	__pci_start_power_transition(dev, state);
1106 
1107 	/* This device is quirked not to be put into D3, so
1108 	   don't put it in D3 */
1109 	if (state >= PCI_D3hot && (dev->dev_flags & PCI_DEV_FLAGS_NO_D3))
1110 		return 0;
1111 
1112 	/*
1113 	 * To put device in D3cold, we put device into D3hot in native
1114 	 * way, then put device into D3cold with platform ops
1115 	 */
1116 	error = pci_raw_set_power_state(dev, state > PCI_D3hot ?
1117 					PCI_D3hot : state);
1118 
1119 	if (!__pci_complete_power_transition(dev, state))
1120 		error = 0;
1121 
1122 	return error;
1123 }
1124 EXPORT_SYMBOL(pci_set_power_state);
1125 
1126 /**
1127  * pci_choose_state - Choose the power state of a PCI device
1128  * @dev: PCI device to be suspended
1129  * @state: target sleep state for the whole system. This is the value
1130  *	that is passed to suspend() function.
1131  *
1132  * Returns PCI power state suitable for given device and given system
1133  * message.
1134  */
1135 
1136 pci_power_t pci_choose_state(struct pci_dev *dev, pm_message_t state)
1137 {
1138 	pci_power_t ret;
1139 
1140 	if (!dev->pm_cap)
1141 		return PCI_D0;
1142 
1143 	ret = platform_pci_choose_state(dev);
1144 	if (ret != PCI_POWER_ERROR)
1145 		return ret;
1146 
1147 	switch (state.event) {
1148 	case PM_EVENT_ON:
1149 		return PCI_D0;
1150 	case PM_EVENT_FREEZE:
1151 	case PM_EVENT_PRETHAW:
1152 		/* REVISIT both freeze and pre-thaw "should" use D0 */
1153 	case PM_EVENT_SUSPEND:
1154 	case PM_EVENT_HIBERNATE:
1155 		return PCI_D3hot;
1156 	default:
1157 		pci_info(dev, "unrecognized suspend event %d\n",
1158 			 state.event);
1159 		BUG();
1160 	}
1161 	return PCI_D0;
1162 }
1163 EXPORT_SYMBOL(pci_choose_state);
1164 
1165 #define PCI_EXP_SAVE_REGS	7
1166 
1167 static struct pci_cap_saved_state *_pci_find_saved_cap(struct pci_dev *pci_dev,
1168 						       u16 cap, bool extended)
1169 {
1170 	struct pci_cap_saved_state *tmp;
1171 
1172 	hlist_for_each_entry(tmp, &pci_dev->saved_cap_space, next) {
1173 		if (tmp->cap.cap_extended == extended && tmp->cap.cap_nr == cap)
1174 			return tmp;
1175 	}
1176 	return NULL;
1177 }
1178 
1179 struct pci_cap_saved_state *pci_find_saved_cap(struct pci_dev *dev, char cap)
1180 {
1181 	return _pci_find_saved_cap(dev, cap, false);
1182 }
1183 
1184 struct pci_cap_saved_state *pci_find_saved_ext_cap(struct pci_dev *dev, u16 cap)
1185 {
1186 	return _pci_find_saved_cap(dev, cap, true);
1187 }
1188 
1189 static int pci_save_pcie_state(struct pci_dev *dev)
1190 {
1191 	int i = 0;
1192 	struct pci_cap_saved_state *save_state;
1193 	u16 *cap;
1194 
1195 	if (!pci_is_pcie(dev))
1196 		return 0;
1197 
1198 	save_state = pci_find_saved_cap(dev, PCI_CAP_ID_EXP);
1199 	if (!save_state) {
1200 		pci_err(dev, "buffer not found in %s\n", __func__);
1201 		return -ENOMEM;
1202 	}
1203 
1204 	cap = (u16 *)&save_state->cap.data[0];
1205 	pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &cap[i++]);
1206 	pcie_capability_read_word(dev, PCI_EXP_LNKCTL, &cap[i++]);
1207 	pcie_capability_read_word(dev, PCI_EXP_SLTCTL, &cap[i++]);
1208 	pcie_capability_read_word(dev, PCI_EXP_RTCTL,  &cap[i++]);
1209 	pcie_capability_read_word(dev, PCI_EXP_DEVCTL2, &cap[i++]);
1210 	pcie_capability_read_word(dev, PCI_EXP_LNKCTL2, &cap[i++]);
1211 	pcie_capability_read_word(dev, PCI_EXP_SLTCTL2, &cap[i++]);
1212 
1213 	return 0;
1214 }
1215 
1216 static void pci_restore_pcie_state(struct pci_dev *dev)
1217 {
1218 	int i = 0;
1219 	struct pci_cap_saved_state *save_state;
1220 	u16 *cap;
1221 
1222 	save_state = pci_find_saved_cap(dev, PCI_CAP_ID_EXP);
1223 	if (!save_state)
1224 		return;
1225 
1226 	cap = (u16 *)&save_state->cap.data[0];
1227 	pcie_capability_write_word(dev, PCI_EXP_DEVCTL, cap[i++]);
1228 	pcie_capability_write_word(dev, PCI_EXP_LNKCTL, cap[i++]);
1229 	pcie_capability_write_word(dev, PCI_EXP_SLTCTL, cap[i++]);
1230 	pcie_capability_write_word(dev, PCI_EXP_RTCTL, cap[i++]);
1231 	pcie_capability_write_word(dev, PCI_EXP_DEVCTL2, cap[i++]);
1232 	pcie_capability_write_word(dev, PCI_EXP_LNKCTL2, cap[i++]);
1233 	pcie_capability_write_word(dev, PCI_EXP_SLTCTL2, cap[i++]);
1234 }
1235 
1236 
1237 static int pci_save_pcix_state(struct pci_dev *dev)
1238 {
1239 	int pos;
1240 	struct pci_cap_saved_state *save_state;
1241 
1242 	pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1243 	if (!pos)
1244 		return 0;
1245 
1246 	save_state = pci_find_saved_cap(dev, PCI_CAP_ID_PCIX);
1247 	if (!save_state) {
1248 		pci_err(dev, "buffer not found in %s\n", __func__);
1249 		return -ENOMEM;
1250 	}
1251 
1252 	pci_read_config_word(dev, pos + PCI_X_CMD,
1253 			     (u16 *)save_state->cap.data);
1254 
1255 	return 0;
1256 }
1257 
1258 static void pci_restore_pcix_state(struct pci_dev *dev)
1259 {
1260 	int i = 0, pos;
1261 	struct pci_cap_saved_state *save_state;
1262 	u16 *cap;
1263 
1264 	save_state = pci_find_saved_cap(dev, PCI_CAP_ID_PCIX);
1265 	pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1266 	if (!save_state || !pos)
1267 		return;
1268 	cap = (u16 *)&save_state->cap.data[0];
1269 
1270 	pci_write_config_word(dev, pos + PCI_X_CMD, cap[i++]);
1271 }
1272 
1273 
1274 /**
1275  * pci_save_state - save the PCI configuration space of a device before suspending
1276  * @dev: - PCI device that we're dealing with
1277  */
1278 int pci_save_state(struct pci_dev *dev)
1279 {
1280 	int i;
1281 	/* XXX: 100% dword access ok here? */
1282 	for (i = 0; i < 16; i++)
1283 		pci_read_config_dword(dev, i * 4, &dev->saved_config_space[i]);
1284 	dev->state_saved = true;
1285 
1286 	i = pci_save_pcie_state(dev);
1287 	if (i != 0)
1288 		return i;
1289 
1290 	i = pci_save_pcix_state(dev);
1291 	if (i != 0)
1292 		return i;
1293 
1294 	pci_save_dpc_state(dev);
1295 	return pci_save_vc_state(dev);
1296 }
1297 EXPORT_SYMBOL(pci_save_state);
1298 
1299 static void pci_restore_config_dword(struct pci_dev *pdev, int offset,
1300 				     u32 saved_val, int retry, bool force)
1301 {
1302 	u32 val;
1303 
1304 	pci_read_config_dword(pdev, offset, &val);
1305 	if (!force && val == saved_val)
1306 		return;
1307 
1308 	for (;;) {
1309 		pci_dbg(pdev, "restoring config space at offset %#x (was %#x, writing %#x)\n",
1310 			offset, val, saved_val);
1311 		pci_write_config_dword(pdev, offset, saved_val);
1312 		if (retry-- <= 0)
1313 			return;
1314 
1315 		pci_read_config_dword(pdev, offset, &val);
1316 		if (val == saved_val)
1317 			return;
1318 
1319 		mdelay(1);
1320 	}
1321 }
1322 
1323 static void pci_restore_config_space_range(struct pci_dev *pdev,
1324 					   int start, int end, int retry,
1325 					   bool force)
1326 {
1327 	int index;
1328 
1329 	for (index = end; index >= start; index--)
1330 		pci_restore_config_dword(pdev, 4 * index,
1331 					 pdev->saved_config_space[index],
1332 					 retry, force);
1333 }
1334 
1335 static void pci_restore_config_space(struct pci_dev *pdev)
1336 {
1337 	if (pdev->hdr_type == PCI_HEADER_TYPE_NORMAL) {
1338 		pci_restore_config_space_range(pdev, 10, 15, 0, false);
1339 		/* Restore BARs before the command register. */
1340 		pci_restore_config_space_range(pdev, 4, 9, 10, false);
1341 		pci_restore_config_space_range(pdev, 0, 3, 0, false);
1342 	} else if (pdev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
1343 		pci_restore_config_space_range(pdev, 12, 15, 0, false);
1344 
1345 		/*
1346 		 * Force rewriting of prefetch registers to avoid S3 resume
1347 		 * issues on Intel PCI bridges that occur when these
1348 		 * registers are not explicitly written.
1349 		 */
1350 		pci_restore_config_space_range(pdev, 9, 11, 0, true);
1351 		pci_restore_config_space_range(pdev, 0, 8, 0, false);
1352 	} else {
1353 		pci_restore_config_space_range(pdev, 0, 15, 0, false);
1354 	}
1355 }
1356 
1357 static void pci_restore_rebar_state(struct pci_dev *pdev)
1358 {
1359 	unsigned int pos, nbars, i;
1360 	u32 ctrl;
1361 
1362 	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_REBAR);
1363 	if (!pos)
1364 		return;
1365 
1366 	pci_read_config_dword(pdev, pos + PCI_REBAR_CTRL, &ctrl);
1367 	nbars = (ctrl & PCI_REBAR_CTRL_NBAR_MASK) >>
1368 		    PCI_REBAR_CTRL_NBAR_SHIFT;
1369 
1370 	for (i = 0; i < nbars; i++, pos += 8) {
1371 		struct resource *res;
1372 		int bar_idx, size;
1373 
1374 		pci_read_config_dword(pdev, pos + PCI_REBAR_CTRL, &ctrl);
1375 		bar_idx = ctrl & PCI_REBAR_CTRL_BAR_IDX;
1376 		res = pdev->resource + bar_idx;
1377 		size = order_base_2((resource_size(res) >> 20) | 1) - 1;
1378 		ctrl &= ~PCI_REBAR_CTRL_BAR_SIZE;
1379 		ctrl |= size << PCI_REBAR_CTRL_BAR_SHIFT;
1380 		pci_write_config_dword(pdev, pos + PCI_REBAR_CTRL, ctrl);
1381 	}
1382 }
1383 
1384 /**
1385  * pci_restore_state - Restore the saved state of a PCI device
1386  * @dev: - PCI device that we're dealing with
1387  */
1388 void pci_restore_state(struct pci_dev *dev)
1389 {
1390 	if (!dev->state_saved)
1391 		return;
1392 
1393 	/* PCI Express register must be restored first */
1394 	pci_restore_pcie_state(dev);
1395 	pci_restore_pasid_state(dev);
1396 	pci_restore_pri_state(dev);
1397 	pci_restore_ats_state(dev);
1398 	pci_restore_vc_state(dev);
1399 	pci_restore_rebar_state(dev);
1400 	pci_restore_dpc_state(dev);
1401 
1402 	pci_cleanup_aer_error_status_regs(dev);
1403 
1404 	pci_restore_config_space(dev);
1405 
1406 	pci_restore_pcix_state(dev);
1407 	pci_restore_msi_state(dev);
1408 
1409 	/* Restore ACS and IOV configuration state */
1410 	pci_enable_acs(dev);
1411 	pci_restore_iov_state(dev);
1412 
1413 	dev->state_saved = false;
1414 }
1415 EXPORT_SYMBOL(pci_restore_state);
1416 
1417 struct pci_saved_state {
1418 	u32 config_space[16];
1419 	struct pci_cap_saved_data cap[0];
1420 };
1421 
1422 /**
1423  * pci_store_saved_state - Allocate and return an opaque struct containing
1424  *			   the device saved state.
1425  * @dev: PCI device that we're dealing with
1426  *
1427  * Return NULL if no state or error.
1428  */
1429 struct pci_saved_state *pci_store_saved_state(struct pci_dev *dev)
1430 {
1431 	struct pci_saved_state *state;
1432 	struct pci_cap_saved_state *tmp;
1433 	struct pci_cap_saved_data *cap;
1434 	size_t size;
1435 
1436 	if (!dev->state_saved)
1437 		return NULL;
1438 
1439 	size = sizeof(*state) + sizeof(struct pci_cap_saved_data);
1440 
1441 	hlist_for_each_entry(tmp, &dev->saved_cap_space, next)
1442 		size += sizeof(struct pci_cap_saved_data) + tmp->cap.size;
1443 
1444 	state = kzalloc(size, GFP_KERNEL);
1445 	if (!state)
1446 		return NULL;
1447 
1448 	memcpy(state->config_space, dev->saved_config_space,
1449 	       sizeof(state->config_space));
1450 
1451 	cap = state->cap;
1452 	hlist_for_each_entry(tmp, &dev->saved_cap_space, next) {
1453 		size_t len = sizeof(struct pci_cap_saved_data) + tmp->cap.size;
1454 		memcpy(cap, &tmp->cap, len);
1455 		cap = (struct pci_cap_saved_data *)((u8 *)cap + len);
1456 	}
1457 	/* Empty cap_save terminates list */
1458 
1459 	return state;
1460 }
1461 EXPORT_SYMBOL_GPL(pci_store_saved_state);
1462 
1463 /**
1464  * pci_load_saved_state - Reload the provided save state into struct pci_dev.
1465  * @dev: PCI device that we're dealing with
1466  * @state: Saved state returned from pci_store_saved_state()
1467  */
1468 int pci_load_saved_state(struct pci_dev *dev,
1469 			 struct pci_saved_state *state)
1470 {
1471 	struct pci_cap_saved_data *cap;
1472 
1473 	dev->state_saved = false;
1474 
1475 	if (!state)
1476 		return 0;
1477 
1478 	memcpy(dev->saved_config_space, state->config_space,
1479 	       sizeof(state->config_space));
1480 
1481 	cap = state->cap;
1482 	while (cap->size) {
1483 		struct pci_cap_saved_state *tmp;
1484 
1485 		tmp = _pci_find_saved_cap(dev, cap->cap_nr, cap->cap_extended);
1486 		if (!tmp || tmp->cap.size != cap->size)
1487 			return -EINVAL;
1488 
1489 		memcpy(tmp->cap.data, cap->data, tmp->cap.size);
1490 		cap = (struct pci_cap_saved_data *)((u8 *)cap +
1491 		       sizeof(struct pci_cap_saved_data) + cap->size);
1492 	}
1493 
1494 	dev->state_saved = true;
1495 	return 0;
1496 }
1497 EXPORT_SYMBOL_GPL(pci_load_saved_state);
1498 
1499 /**
1500  * pci_load_and_free_saved_state - Reload the save state pointed to by state,
1501  *				   and free the memory allocated for it.
1502  * @dev: PCI device that we're dealing with
1503  * @state: Pointer to saved state returned from pci_store_saved_state()
1504  */
1505 int pci_load_and_free_saved_state(struct pci_dev *dev,
1506 				  struct pci_saved_state **state)
1507 {
1508 	int ret = pci_load_saved_state(dev, *state);
1509 	kfree(*state);
1510 	*state = NULL;
1511 	return ret;
1512 }
1513 EXPORT_SYMBOL_GPL(pci_load_and_free_saved_state);
1514 
1515 int __weak pcibios_enable_device(struct pci_dev *dev, int bars)
1516 {
1517 	return pci_enable_resources(dev, bars);
1518 }
1519 
1520 static int do_pci_enable_device(struct pci_dev *dev, int bars)
1521 {
1522 	int err;
1523 	struct pci_dev *bridge;
1524 	u16 cmd;
1525 	u8 pin;
1526 
1527 	err = pci_set_power_state(dev, PCI_D0);
1528 	if (err < 0 && err != -EIO)
1529 		return err;
1530 
1531 	bridge = pci_upstream_bridge(dev);
1532 	if (bridge)
1533 		pcie_aspm_powersave_config_link(bridge);
1534 
1535 	err = pcibios_enable_device(dev, bars);
1536 	if (err < 0)
1537 		return err;
1538 	pci_fixup_device(pci_fixup_enable, dev);
1539 
1540 	if (dev->msi_enabled || dev->msix_enabled)
1541 		return 0;
1542 
1543 	pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &pin);
1544 	if (pin) {
1545 		pci_read_config_word(dev, PCI_COMMAND, &cmd);
1546 		if (cmd & PCI_COMMAND_INTX_DISABLE)
1547 			pci_write_config_word(dev, PCI_COMMAND,
1548 					      cmd & ~PCI_COMMAND_INTX_DISABLE);
1549 	}
1550 
1551 	return 0;
1552 }
1553 
1554 /**
1555  * pci_reenable_device - Resume abandoned device
1556  * @dev: PCI device to be resumed
1557  *
1558  *  Note this function is a backend of pci_default_resume and is not supposed
1559  *  to be called by normal code, write proper resume handler and use it instead.
1560  */
1561 int pci_reenable_device(struct pci_dev *dev)
1562 {
1563 	if (pci_is_enabled(dev))
1564 		return do_pci_enable_device(dev, (1 << PCI_NUM_RESOURCES) - 1);
1565 	return 0;
1566 }
1567 EXPORT_SYMBOL(pci_reenable_device);
1568 
1569 static void pci_enable_bridge(struct pci_dev *dev)
1570 {
1571 	struct pci_dev *bridge;
1572 	int retval;
1573 
1574 	bridge = pci_upstream_bridge(dev);
1575 	if (bridge)
1576 		pci_enable_bridge(bridge);
1577 
1578 	if (pci_is_enabled(dev)) {
1579 		if (!dev->is_busmaster)
1580 			pci_set_master(dev);
1581 		return;
1582 	}
1583 
1584 	retval = pci_enable_device(dev);
1585 	if (retval)
1586 		pci_err(dev, "Error enabling bridge (%d), continuing\n",
1587 			retval);
1588 	pci_set_master(dev);
1589 }
1590 
1591 static int pci_enable_device_flags(struct pci_dev *dev, unsigned long flags)
1592 {
1593 	struct pci_dev *bridge;
1594 	int err;
1595 	int i, bars = 0;
1596 
1597 	/*
1598 	 * Power state could be unknown at this point, either due to a fresh
1599 	 * boot or a device removal call.  So get the current power state
1600 	 * so that things like MSI message writing will behave as expected
1601 	 * (e.g. if the device really is in D0 at enable time).
1602 	 */
1603 	if (dev->pm_cap) {
1604 		u16 pmcsr;
1605 		pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
1606 		dev->current_state = (pmcsr & PCI_PM_CTRL_STATE_MASK);
1607 	}
1608 
1609 	if (atomic_inc_return(&dev->enable_cnt) > 1)
1610 		return 0;		/* already enabled */
1611 
1612 	bridge = pci_upstream_bridge(dev);
1613 	if (bridge)
1614 		pci_enable_bridge(bridge);
1615 
1616 	/* only skip sriov related */
1617 	for (i = 0; i <= PCI_ROM_RESOURCE; i++)
1618 		if (dev->resource[i].flags & flags)
1619 			bars |= (1 << i);
1620 	for (i = PCI_BRIDGE_RESOURCES; i < DEVICE_COUNT_RESOURCE; i++)
1621 		if (dev->resource[i].flags & flags)
1622 			bars |= (1 << i);
1623 
1624 	err = do_pci_enable_device(dev, bars);
1625 	if (err < 0)
1626 		atomic_dec(&dev->enable_cnt);
1627 	return err;
1628 }
1629 
1630 /**
1631  * pci_enable_device_io - Initialize a device for use with IO space
1632  * @dev: PCI device to be initialized
1633  *
1634  *  Initialize device before it's used by a driver. Ask low-level code
1635  *  to enable I/O resources. Wake up the device if it was suspended.
1636  *  Beware, this function can fail.
1637  */
1638 int pci_enable_device_io(struct pci_dev *dev)
1639 {
1640 	return pci_enable_device_flags(dev, IORESOURCE_IO);
1641 }
1642 EXPORT_SYMBOL(pci_enable_device_io);
1643 
1644 /**
1645  * pci_enable_device_mem - Initialize a device for use with Memory space
1646  * @dev: PCI device to be initialized
1647  *
1648  *  Initialize device before it's used by a driver. Ask low-level code
1649  *  to enable Memory resources. Wake up the device if it was suspended.
1650  *  Beware, this function can fail.
1651  */
1652 int pci_enable_device_mem(struct pci_dev *dev)
1653 {
1654 	return pci_enable_device_flags(dev, IORESOURCE_MEM);
1655 }
1656 EXPORT_SYMBOL(pci_enable_device_mem);
1657 
1658 /**
1659  * pci_enable_device - Initialize device before it's used by a driver.
1660  * @dev: PCI device to be initialized
1661  *
1662  *  Initialize device before it's used by a driver. Ask low-level code
1663  *  to enable I/O and memory. Wake up the device if it was suspended.
1664  *  Beware, this function can fail.
1665  *
1666  *  Note we don't actually enable the device many times if we call
1667  *  this function repeatedly (we just increment the count).
1668  */
1669 int pci_enable_device(struct pci_dev *dev)
1670 {
1671 	return pci_enable_device_flags(dev, IORESOURCE_MEM | IORESOURCE_IO);
1672 }
1673 EXPORT_SYMBOL(pci_enable_device);
1674 
1675 /*
1676  * Managed PCI resources.  This manages device on/off, intx/msi/msix
1677  * on/off and BAR regions.  pci_dev itself records msi/msix status, so
1678  * there's no need to track it separately.  pci_devres is initialized
1679  * when a device is enabled using managed PCI device enable interface.
1680  */
1681 struct pci_devres {
1682 	unsigned int enabled:1;
1683 	unsigned int pinned:1;
1684 	unsigned int orig_intx:1;
1685 	unsigned int restore_intx:1;
1686 	unsigned int mwi:1;
1687 	u32 region_mask;
1688 };
1689 
1690 static void pcim_release(struct device *gendev, void *res)
1691 {
1692 	struct pci_dev *dev = to_pci_dev(gendev);
1693 	struct pci_devres *this = res;
1694 	int i;
1695 
1696 	if (dev->msi_enabled)
1697 		pci_disable_msi(dev);
1698 	if (dev->msix_enabled)
1699 		pci_disable_msix(dev);
1700 
1701 	for (i = 0; i < DEVICE_COUNT_RESOURCE; i++)
1702 		if (this->region_mask & (1 << i))
1703 			pci_release_region(dev, i);
1704 
1705 	if (this->mwi)
1706 		pci_clear_mwi(dev);
1707 
1708 	if (this->restore_intx)
1709 		pci_intx(dev, this->orig_intx);
1710 
1711 	if (this->enabled && !this->pinned)
1712 		pci_disable_device(dev);
1713 }
1714 
1715 static struct pci_devres *get_pci_dr(struct pci_dev *pdev)
1716 {
1717 	struct pci_devres *dr, *new_dr;
1718 
1719 	dr = devres_find(&pdev->dev, pcim_release, NULL, NULL);
1720 	if (dr)
1721 		return dr;
1722 
1723 	new_dr = devres_alloc(pcim_release, sizeof(*new_dr), GFP_KERNEL);
1724 	if (!new_dr)
1725 		return NULL;
1726 	return devres_get(&pdev->dev, new_dr, NULL, NULL);
1727 }
1728 
1729 static struct pci_devres *find_pci_dr(struct pci_dev *pdev)
1730 {
1731 	if (pci_is_managed(pdev))
1732 		return devres_find(&pdev->dev, pcim_release, NULL, NULL);
1733 	return NULL;
1734 }
1735 
1736 /**
1737  * pcim_enable_device - Managed pci_enable_device()
1738  * @pdev: PCI device to be initialized
1739  *
1740  * Managed pci_enable_device().
1741  */
1742 int pcim_enable_device(struct pci_dev *pdev)
1743 {
1744 	struct pci_devres *dr;
1745 	int rc;
1746 
1747 	dr = get_pci_dr(pdev);
1748 	if (unlikely(!dr))
1749 		return -ENOMEM;
1750 	if (dr->enabled)
1751 		return 0;
1752 
1753 	rc = pci_enable_device(pdev);
1754 	if (!rc) {
1755 		pdev->is_managed = 1;
1756 		dr->enabled = 1;
1757 	}
1758 	return rc;
1759 }
1760 EXPORT_SYMBOL(pcim_enable_device);
1761 
1762 /**
1763  * pcim_pin_device - Pin managed PCI device
1764  * @pdev: PCI device to pin
1765  *
1766  * Pin managed PCI device @pdev.  Pinned device won't be disabled on
1767  * driver detach.  @pdev must have been enabled with
1768  * pcim_enable_device().
1769  */
1770 void pcim_pin_device(struct pci_dev *pdev)
1771 {
1772 	struct pci_devres *dr;
1773 
1774 	dr = find_pci_dr(pdev);
1775 	WARN_ON(!dr || !dr->enabled);
1776 	if (dr)
1777 		dr->pinned = 1;
1778 }
1779 EXPORT_SYMBOL(pcim_pin_device);
1780 
1781 /*
1782  * pcibios_add_device - provide arch specific hooks when adding device dev
1783  * @dev: the PCI device being added
1784  *
1785  * Permits the platform to provide architecture specific functionality when
1786  * devices are added. This is the default implementation. Architecture
1787  * implementations can override this.
1788  */
1789 int __weak pcibios_add_device(struct pci_dev *dev)
1790 {
1791 	return 0;
1792 }
1793 
1794 /**
1795  * pcibios_release_device - provide arch specific hooks when releasing device dev
1796  * @dev: the PCI device being released
1797  *
1798  * Permits the platform to provide architecture specific functionality when
1799  * devices are released. This is the default implementation. Architecture
1800  * implementations can override this.
1801  */
1802 void __weak pcibios_release_device(struct pci_dev *dev) {}
1803 
1804 /**
1805  * pcibios_disable_device - disable arch specific PCI resources for device dev
1806  * @dev: the PCI device to disable
1807  *
1808  * Disables architecture specific PCI resources for the device. This
1809  * is the default implementation. Architecture implementations can
1810  * override this.
1811  */
1812 void __weak pcibios_disable_device(struct pci_dev *dev) {}
1813 
1814 /**
1815  * pcibios_penalize_isa_irq - penalize an ISA IRQ
1816  * @irq: ISA IRQ to penalize
1817  * @active: IRQ active or not
1818  *
1819  * Permits the platform to provide architecture-specific functionality when
1820  * penalizing ISA IRQs. This is the default implementation. Architecture
1821  * implementations can override this.
1822  */
1823 void __weak pcibios_penalize_isa_irq(int irq, int active) {}
1824 
1825 static void do_pci_disable_device(struct pci_dev *dev)
1826 {
1827 	u16 pci_command;
1828 
1829 	pci_read_config_word(dev, PCI_COMMAND, &pci_command);
1830 	if (pci_command & PCI_COMMAND_MASTER) {
1831 		pci_command &= ~PCI_COMMAND_MASTER;
1832 		pci_write_config_word(dev, PCI_COMMAND, pci_command);
1833 	}
1834 
1835 	pcibios_disable_device(dev);
1836 }
1837 
1838 /**
1839  * pci_disable_enabled_device - Disable device without updating enable_cnt
1840  * @dev: PCI device to disable
1841  *
1842  * NOTE: This function is a backend of PCI power management routines and is
1843  * not supposed to be called drivers.
1844  */
1845 void pci_disable_enabled_device(struct pci_dev *dev)
1846 {
1847 	if (pci_is_enabled(dev))
1848 		do_pci_disable_device(dev);
1849 }
1850 
1851 /**
1852  * pci_disable_device - Disable PCI device after use
1853  * @dev: PCI device to be disabled
1854  *
1855  * Signal to the system that the PCI device is not in use by the system
1856  * anymore.  This only involves disabling PCI bus-mastering, if active.
1857  *
1858  * Note we don't actually disable the device until all callers of
1859  * pci_enable_device() have called pci_disable_device().
1860  */
1861 void pci_disable_device(struct pci_dev *dev)
1862 {
1863 	struct pci_devres *dr;
1864 
1865 	dr = find_pci_dr(dev);
1866 	if (dr)
1867 		dr->enabled = 0;
1868 
1869 	dev_WARN_ONCE(&dev->dev, atomic_read(&dev->enable_cnt) <= 0,
1870 		      "disabling already-disabled device");
1871 
1872 	if (atomic_dec_return(&dev->enable_cnt) != 0)
1873 		return;
1874 
1875 	do_pci_disable_device(dev);
1876 
1877 	dev->is_busmaster = 0;
1878 }
1879 EXPORT_SYMBOL(pci_disable_device);
1880 
1881 /**
1882  * pcibios_set_pcie_reset_state - set reset state for device dev
1883  * @dev: the PCIe device reset
1884  * @state: Reset state to enter into
1885  *
1886  *
1887  * Sets the PCIe reset state for the device. This is the default
1888  * implementation. Architecture implementations can override this.
1889  */
1890 int __weak pcibios_set_pcie_reset_state(struct pci_dev *dev,
1891 					enum pcie_reset_state state)
1892 {
1893 	return -EINVAL;
1894 }
1895 
1896 /**
1897  * pci_set_pcie_reset_state - set reset state for device dev
1898  * @dev: the PCIe device reset
1899  * @state: Reset state to enter into
1900  *
1901  *
1902  * Sets the PCI reset state for the device.
1903  */
1904 int pci_set_pcie_reset_state(struct pci_dev *dev, enum pcie_reset_state state)
1905 {
1906 	return pcibios_set_pcie_reset_state(dev, state);
1907 }
1908 EXPORT_SYMBOL_GPL(pci_set_pcie_reset_state);
1909 
1910 /**
1911  * pcie_clear_root_pme_status - Clear root port PME interrupt status.
1912  * @dev: PCIe root port or event collector.
1913  */
1914 void pcie_clear_root_pme_status(struct pci_dev *dev)
1915 {
1916 	pcie_capability_set_dword(dev, PCI_EXP_RTSTA, PCI_EXP_RTSTA_PME);
1917 }
1918 
1919 /**
1920  * pci_check_pme_status - Check if given device has generated PME.
1921  * @dev: Device to check.
1922  *
1923  * Check the PME status of the device and if set, clear it and clear PME enable
1924  * (if set).  Return 'true' if PME status and PME enable were both set or
1925  * 'false' otherwise.
1926  */
1927 bool pci_check_pme_status(struct pci_dev *dev)
1928 {
1929 	int pmcsr_pos;
1930 	u16 pmcsr;
1931 	bool ret = false;
1932 
1933 	if (!dev->pm_cap)
1934 		return false;
1935 
1936 	pmcsr_pos = dev->pm_cap + PCI_PM_CTRL;
1937 	pci_read_config_word(dev, pmcsr_pos, &pmcsr);
1938 	if (!(pmcsr & PCI_PM_CTRL_PME_STATUS))
1939 		return false;
1940 
1941 	/* Clear PME status. */
1942 	pmcsr |= PCI_PM_CTRL_PME_STATUS;
1943 	if (pmcsr & PCI_PM_CTRL_PME_ENABLE) {
1944 		/* Disable PME to avoid interrupt flood. */
1945 		pmcsr &= ~PCI_PM_CTRL_PME_ENABLE;
1946 		ret = true;
1947 	}
1948 
1949 	pci_write_config_word(dev, pmcsr_pos, pmcsr);
1950 
1951 	return ret;
1952 }
1953 
1954 /**
1955  * pci_pme_wakeup - Wake up a PCI device if its PME Status bit is set.
1956  * @dev: Device to handle.
1957  * @pme_poll_reset: Whether or not to reset the device's pme_poll flag.
1958  *
1959  * Check if @dev has generated PME and queue a resume request for it in that
1960  * case.
1961  */
1962 static int pci_pme_wakeup(struct pci_dev *dev, void *pme_poll_reset)
1963 {
1964 	if (pme_poll_reset && dev->pme_poll)
1965 		dev->pme_poll = false;
1966 
1967 	if (pci_check_pme_status(dev)) {
1968 		pci_wakeup_event(dev);
1969 		pm_request_resume(&dev->dev);
1970 	}
1971 	return 0;
1972 }
1973 
1974 /**
1975  * pci_pme_wakeup_bus - Walk given bus and wake up devices on it, if necessary.
1976  * @bus: Top bus of the subtree to walk.
1977  */
1978 void pci_pme_wakeup_bus(struct pci_bus *bus)
1979 {
1980 	if (bus)
1981 		pci_walk_bus(bus, pci_pme_wakeup, (void *)true);
1982 }
1983 
1984 
1985 /**
1986  * pci_pme_capable - check the capability of PCI device to generate PME#
1987  * @dev: PCI device to handle.
1988  * @state: PCI state from which device will issue PME#.
1989  */
1990 bool pci_pme_capable(struct pci_dev *dev, pci_power_t state)
1991 {
1992 	if (!dev->pm_cap)
1993 		return false;
1994 
1995 	return !!(dev->pme_support & (1 << state));
1996 }
1997 EXPORT_SYMBOL(pci_pme_capable);
1998 
1999 static void pci_pme_list_scan(struct work_struct *work)
2000 {
2001 	struct pci_pme_device *pme_dev, *n;
2002 
2003 	mutex_lock(&pci_pme_list_mutex);
2004 	list_for_each_entry_safe(pme_dev, n, &pci_pme_list, list) {
2005 		if (pme_dev->dev->pme_poll) {
2006 			struct pci_dev *bridge;
2007 
2008 			bridge = pme_dev->dev->bus->self;
2009 			/*
2010 			 * If bridge is in low power state, the
2011 			 * configuration space of subordinate devices
2012 			 * may be not accessible
2013 			 */
2014 			if (bridge && bridge->current_state != PCI_D0)
2015 				continue;
2016 			pci_pme_wakeup(pme_dev->dev, NULL);
2017 		} else {
2018 			list_del(&pme_dev->list);
2019 			kfree(pme_dev);
2020 		}
2021 	}
2022 	if (!list_empty(&pci_pme_list))
2023 		queue_delayed_work(system_freezable_wq, &pci_pme_work,
2024 				   msecs_to_jiffies(PME_TIMEOUT));
2025 	mutex_unlock(&pci_pme_list_mutex);
2026 }
2027 
2028 static void __pci_pme_active(struct pci_dev *dev, bool enable)
2029 {
2030 	u16 pmcsr;
2031 
2032 	if (!dev->pme_support)
2033 		return;
2034 
2035 	pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
2036 	/* Clear PME_Status by writing 1 to it and enable PME# */
2037 	pmcsr |= PCI_PM_CTRL_PME_STATUS | PCI_PM_CTRL_PME_ENABLE;
2038 	if (!enable)
2039 		pmcsr &= ~PCI_PM_CTRL_PME_ENABLE;
2040 
2041 	pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, pmcsr);
2042 }
2043 
2044 /**
2045  * pci_pme_restore - Restore PME configuration after config space restore.
2046  * @dev: PCI device to update.
2047  */
2048 void pci_pme_restore(struct pci_dev *dev)
2049 {
2050 	u16 pmcsr;
2051 
2052 	if (!dev->pme_support)
2053 		return;
2054 
2055 	pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
2056 	if (dev->wakeup_prepared) {
2057 		pmcsr |= PCI_PM_CTRL_PME_ENABLE;
2058 		pmcsr &= ~PCI_PM_CTRL_PME_STATUS;
2059 	} else {
2060 		pmcsr &= ~PCI_PM_CTRL_PME_ENABLE;
2061 		pmcsr |= PCI_PM_CTRL_PME_STATUS;
2062 	}
2063 	pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, pmcsr);
2064 }
2065 
2066 /**
2067  * pci_pme_active - enable or disable PCI device's PME# function
2068  * @dev: PCI device to handle.
2069  * @enable: 'true' to enable PME# generation; 'false' to disable it.
2070  *
2071  * The caller must verify that the device is capable of generating PME# before
2072  * calling this function with @enable equal to 'true'.
2073  */
2074 void pci_pme_active(struct pci_dev *dev, bool enable)
2075 {
2076 	__pci_pme_active(dev, enable);
2077 
2078 	/*
2079 	 * PCI (as opposed to PCIe) PME requires that the device have
2080 	 * its PME# line hooked up correctly. Not all hardware vendors
2081 	 * do this, so the PME never gets delivered and the device
2082 	 * remains asleep. The easiest way around this is to
2083 	 * periodically walk the list of suspended devices and check
2084 	 * whether any have their PME flag set. The assumption is that
2085 	 * we'll wake up often enough anyway that this won't be a huge
2086 	 * hit, and the power savings from the devices will still be a
2087 	 * win.
2088 	 *
2089 	 * Although PCIe uses in-band PME message instead of PME# line
2090 	 * to report PME, PME does not work for some PCIe devices in
2091 	 * reality.  For example, there are devices that set their PME
2092 	 * status bits, but don't really bother to send a PME message;
2093 	 * there are PCI Express Root Ports that don't bother to
2094 	 * trigger interrupts when they receive PME messages from the
2095 	 * devices below.  So PME poll is used for PCIe devices too.
2096 	 */
2097 
2098 	if (dev->pme_poll) {
2099 		struct pci_pme_device *pme_dev;
2100 		if (enable) {
2101 			pme_dev = kmalloc(sizeof(struct pci_pme_device),
2102 					  GFP_KERNEL);
2103 			if (!pme_dev) {
2104 				pci_warn(dev, "can't enable PME#\n");
2105 				return;
2106 			}
2107 			pme_dev->dev = dev;
2108 			mutex_lock(&pci_pme_list_mutex);
2109 			list_add(&pme_dev->list, &pci_pme_list);
2110 			if (list_is_singular(&pci_pme_list))
2111 				queue_delayed_work(system_freezable_wq,
2112 						   &pci_pme_work,
2113 						   msecs_to_jiffies(PME_TIMEOUT));
2114 			mutex_unlock(&pci_pme_list_mutex);
2115 		} else {
2116 			mutex_lock(&pci_pme_list_mutex);
2117 			list_for_each_entry(pme_dev, &pci_pme_list, list) {
2118 				if (pme_dev->dev == dev) {
2119 					list_del(&pme_dev->list);
2120 					kfree(pme_dev);
2121 					break;
2122 				}
2123 			}
2124 			mutex_unlock(&pci_pme_list_mutex);
2125 		}
2126 	}
2127 
2128 	pci_dbg(dev, "PME# %s\n", enable ? "enabled" : "disabled");
2129 }
2130 EXPORT_SYMBOL(pci_pme_active);
2131 
2132 /**
2133  * __pci_enable_wake - enable PCI device as wakeup event source
2134  * @dev: PCI device affected
2135  * @state: PCI state from which device will issue wakeup events
2136  * @enable: True to enable event generation; false to disable
2137  *
2138  * This enables the device as a wakeup event source, or disables it.
2139  * When such events involves platform-specific hooks, those hooks are
2140  * called automatically by this routine.
2141  *
2142  * Devices with legacy power management (no standard PCI PM capabilities)
2143  * always require such platform hooks.
2144  *
2145  * RETURN VALUE:
2146  * 0 is returned on success
2147  * -EINVAL is returned if device is not supposed to wake up the system
2148  * Error code depending on the platform is returned if both the platform and
2149  * the native mechanism fail to enable the generation of wake-up events
2150  */
2151 static int __pci_enable_wake(struct pci_dev *dev, pci_power_t state, bool enable)
2152 {
2153 	int ret = 0;
2154 
2155 	/*
2156 	 * Bridges that are not power-manageable directly only signal
2157 	 * wakeup on behalf of subordinate devices which is set up
2158 	 * elsewhere, so skip them. However, bridges that are
2159 	 * power-manageable may signal wakeup for themselves (for example,
2160 	 * on a hotplug event) and they need to be covered here.
2161 	 */
2162 	if (!pci_power_manageable(dev))
2163 		return 0;
2164 
2165 	/* Don't do the same thing twice in a row for one device. */
2166 	if (!!enable == !!dev->wakeup_prepared)
2167 		return 0;
2168 
2169 	/*
2170 	 * According to "PCI System Architecture" 4th ed. by Tom Shanley & Don
2171 	 * Anderson we should be doing PME# wake enable followed by ACPI wake
2172 	 * enable.  To disable wake-up we call the platform first, for symmetry.
2173 	 */
2174 
2175 	if (enable) {
2176 		int error;
2177 
2178 		if (pci_pme_capable(dev, state))
2179 			pci_pme_active(dev, true);
2180 		else
2181 			ret = 1;
2182 		error = platform_pci_set_wakeup(dev, true);
2183 		if (ret)
2184 			ret = error;
2185 		if (!ret)
2186 			dev->wakeup_prepared = true;
2187 	} else {
2188 		platform_pci_set_wakeup(dev, false);
2189 		pci_pme_active(dev, false);
2190 		dev->wakeup_prepared = false;
2191 	}
2192 
2193 	return ret;
2194 }
2195 
2196 /**
2197  * pci_enable_wake - change wakeup settings for a PCI device
2198  * @pci_dev: Target device
2199  * @state: PCI state from which device will issue wakeup events
2200  * @enable: Whether or not to enable event generation
2201  *
2202  * If @enable is set, check device_may_wakeup() for the device before calling
2203  * __pci_enable_wake() for it.
2204  */
2205 int pci_enable_wake(struct pci_dev *pci_dev, pci_power_t state, bool enable)
2206 {
2207 	if (enable && !device_may_wakeup(&pci_dev->dev))
2208 		return -EINVAL;
2209 
2210 	return __pci_enable_wake(pci_dev, state, enable);
2211 }
2212 EXPORT_SYMBOL(pci_enable_wake);
2213 
2214 /**
2215  * pci_wake_from_d3 - enable/disable device to wake up from D3_hot or D3_cold
2216  * @dev: PCI device to prepare
2217  * @enable: True to enable wake-up event generation; false to disable
2218  *
2219  * Many drivers want the device to wake up the system from D3_hot or D3_cold
2220  * and this function allows them to set that up cleanly - pci_enable_wake()
2221  * should not be called twice in a row to enable wake-up due to PCI PM vs ACPI
2222  * ordering constraints.
2223  *
2224  * This function only returns error code if the device is not allowed to wake
2225  * up the system from sleep or it is not capable of generating PME# from both
2226  * D3_hot and D3_cold and the platform is unable to enable wake-up power for it.
2227  */
2228 int pci_wake_from_d3(struct pci_dev *dev, bool enable)
2229 {
2230 	return pci_pme_capable(dev, PCI_D3cold) ?
2231 			pci_enable_wake(dev, PCI_D3cold, enable) :
2232 			pci_enable_wake(dev, PCI_D3hot, enable);
2233 }
2234 EXPORT_SYMBOL(pci_wake_from_d3);
2235 
2236 /**
2237  * pci_target_state - find an appropriate low power state for a given PCI dev
2238  * @dev: PCI device
2239  * @wakeup: Whether or not wakeup functionality will be enabled for the device.
2240  *
2241  * Use underlying platform code to find a supported low power state for @dev.
2242  * If the platform can't manage @dev, return the deepest state from which it
2243  * can generate wake events, based on any available PME info.
2244  */
2245 static pci_power_t pci_target_state(struct pci_dev *dev, bool wakeup)
2246 {
2247 	pci_power_t target_state = PCI_D3hot;
2248 
2249 	if (platform_pci_power_manageable(dev)) {
2250 		/*
2251 		 * Call the platform to find the target state for the device.
2252 		 */
2253 		pci_power_t state = platform_pci_choose_state(dev);
2254 
2255 		switch (state) {
2256 		case PCI_POWER_ERROR:
2257 		case PCI_UNKNOWN:
2258 			break;
2259 		case PCI_D1:
2260 		case PCI_D2:
2261 			if (pci_no_d1d2(dev))
2262 				break;
2263 			/* else: fall through */
2264 		default:
2265 			target_state = state;
2266 		}
2267 
2268 		return target_state;
2269 	}
2270 
2271 	if (!dev->pm_cap)
2272 		target_state = PCI_D0;
2273 
2274 	/*
2275 	 * If the device is in D3cold even though it's not power-manageable by
2276 	 * the platform, it may have been powered down by non-standard means.
2277 	 * Best to let it slumber.
2278 	 */
2279 	if (dev->current_state == PCI_D3cold)
2280 		target_state = PCI_D3cold;
2281 
2282 	if (wakeup) {
2283 		/*
2284 		 * Find the deepest state from which the device can generate
2285 		 * PME#.
2286 		 */
2287 		if (dev->pme_support) {
2288 			while (target_state
2289 			      && !(dev->pme_support & (1 << target_state)))
2290 				target_state--;
2291 		}
2292 	}
2293 
2294 	return target_state;
2295 }
2296 
2297 /**
2298  * pci_prepare_to_sleep - prepare PCI device for system-wide transition into a sleep state
2299  * @dev: Device to handle.
2300  *
2301  * Choose the power state appropriate for the device depending on whether
2302  * it can wake up the system and/or is power manageable by the platform
2303  * (PCI_D3hot is the default) and put the device into that state.
2304  */
2305 int pci_prepare_to_sleep(struct pci_dev *dev)
2306 {
2307 	bool wakeup = device_may_wakeup(&dev->dev);
2308 	pci_power_t target_state = pci_target_state(dev, wakeup);
2309 	int error;
2310 
2311 	if (target_state == PCI_POWER_ERROR)
2312 		return -EIO;
2313 
2314 	pci_enable_wake(dev, target_state, wakeup);
2315 
2316 	error = pci_set_power_state(dev, target_state);
2317 
2318 	if (error)
2319 		pci_enable_wake(dev, target_state, false);
2320 
2321 	return error;
2322 }
2323 EXPORT_SYMBOL(pci_prepare_to_sleep);
2324 
2325 /**
2326  * pci_back_from_sleep - turn PCI device on during system-wide transition into working state
2327  * @dev: Device to handle.
2328  *
2329  * Disable device's system wake-up capability and put it into D0.
2330  */
2331 int pci_back_from_sleep(struct pci_dev *dev)
2332 {
2333 	pci_enable_wake(dev, PCI_D0, false);
2334 	return pci_set_power_state(dev, PCI_D0);
2335 }
2336 EXPORT_SYMBOL(pci_back_from_sleep);
2337 
2338 /**
2339  * pci_finish_runtime_suspend - Carry out PCI-specific part of runtime suspend.
2340  * @dev: PCI device being suspended.
2341  *
2342  * Prepare @dev to generate wake-up events at run time and put it into a low
2343  * power state.
2344  */
2345 int pci_finish_runtime_suspend(struct pci_dev *dev)
2346 {
2347 	pci_power_t target_state;
2348 	int error;
2349 
2350 	target_state = pci_target_state(dev, device_can_wakeup(&dev->dev));
2351 	if (target_state == PCI_POWER_ERROR)
2352 		return -EIO;
2353 
2354 	dev->runtime_d3cold = target_state == PCI_D3cold;
2355 
2356 	__pci_enable_wake(dev, target_state, pci_dev_run_wake(dev));
2357 
2358 	error = pci_set_power_state(dev, target_state);
2359 
2360 	if (error) {
2361 		pci_enable_wake(dev, target_state, false);
2362 		dev->runtime_d3cold = false;
2363 	}
2364 
2365 	return error;
2366 }
2367 
2368 /**
2369  * pci_dev_run_wake - Check if device can generate run-time wake-up events.
2370  * @dev: Device to check.
2371  *
2372  * Return true if the device itself is capable of generating wake-up events
2373  * (through the platform or using the native PCIe PME) or if the device supports
2374  * PME and one of its upstream bridges can generate wake-up events.
2375  */
2376 bool pci_dev_run_wake(struct pci_dev *dev)
2377 {
2378 	struct pci_bus *bus = dev->bus;
2379 
2380 	if (!dev->pme_support)
2381 		return false;
2382 
2383 	/* PME-capable in principle, but not from the target power state */
2384 	if (!pci_pme_capable(dev, pci_target_state(dev, true)))
2385 		return false;
2386 
2387 	if (device_can_wakeup(&dev->dev))
2388 		return true;
2389 
2390 	while (bus->parent) {
2391 		struct pci_dev *bridge = bus->self;
2392 
2393 		if (device_can_wakeup(&bridge->dev))
2394 			return true;
2395 
2396 		bus = bus->parent;
2397 	}
2398 
2399 	/* We have reached the root bus. */
2400 	if (bus->bridge)
2401 		return device_can_wakeup(bus->bridge);
2402 
2403 	return false;
2404 }
2405 EXPORT_SYMBOL_GPL(pci_dev_run_wake);
2406 
2407 /**
2408  * pci_dev_keep_suspended - Check if the device can stay in the suspended state.
2409  * @pci_dev: Device to check.
2410  *
2411  * Return 'true' if the device is runtime-suspended, it doesn't have to be
2412  * reconfigured due to wakeup settings difference between system and runtime
2413  * suspend and the current power state of it is suitable for the upcoming
2414  * (system) transition.
2415  *
2416  * If the device is not configured for system wakeup, disable PME for it before
2417  * returning 'true' to prevent it from waking up the system unnecessarily.
2418  */
2419 bool pci_dev_keep_suspended(struct pci_dev *pci_dev)
2420 {
2421 	struct device *dev = &pci_dev->dev;
2422 	bool wakeup = device_may_wakeup(dev);
2423 
2424 	if (!pm_runtime_suspended(dev)
2425 	    || pci_target_state(pci_dev, wakeup) != pci_dev->current_state
2426 	    || platform_pci_need_resume(pci_dev))
2427 		return false;
2428 
2429 	/*
2430 	 * At this point the device is good to go unless it's been configured
2431 	 * to generate PME at the runtime suspend time, but it is not supposed
2432 	 * to wake up the system.  In that case, simply disable PME for it
2433 	 * (it will have to be re-enabled on exit from system resume).
2434 	 *
2435 	 * If the device's power state is D3cold and the platform check above
2436 	 * hasn't triggered, the device's configuration is suitable and we don't
2437 	 * need to manipulate it at all.
2438 	 */
2439 	spin_lock_irq(&dev->power.lock);
2440 
2441 	if (pm_runtime_suspended(dev) && pci_dev->current_state < PCI_D3cold &&
2442 	    !wakeup)
2443 		__pci_pme_active(pci_dev, false);
2444 
2445 	spin_unlock_irq(&dev->power.lock);
2446 	return true;
2447 }
2448 
2449 /**
2450  * pci_dev_complete_resume - Finalize resume from system sleep for a device.
2451  * @pci_dev: Device to handle.
2452  *
2453  * If the device is runtime suspended and wakeup-capable, enable PME for it as
2454  * it might have been disabled during the prepare phase of system suspend if
2455  * the device was not configured for system wakeup.
2456  */
2457 void pci_dev_complete_resume(struct pci_dev *pci_dev)
2458 {
2459 	struct device *dev = &pci_dev->dev;
2460 
2461 	if (!pci_dev_run_wake(pci_dev))
2462 		return;
2463 
2464 	spin_lock_irq(&dev->power.lock);
2465 
2466 	if (pm_runtime_suspended(dev) && pci_dev->current_state < PCI_D3cold)
2467 		__pci_pme_active(pci_dev, true);
2468 
2469 	spin_unlock_irq(&dev->power.lock);
2470 }
2471 
2472 void pci_config_pm_runtime_get(struct pci_dev *pdev)
2473 {
2474 	struct device *dev = &pdev->dev;
2475 	struct device *parent = dev->parent;
2476 
2477 	if (parent)
2478 		pm_runtime_get_sync(parent);
2479 	pm_runtime_get_noresume(dev);
2480 	/*
2481 	 * pdev->current_state is set to PCI_D3cold during suspending,
2482 	 * so wait until suspending completes
2483 	 */
2484 	pm_runtime_barrier(dev);
2485 	/*
2486 	 * Only need to resume devices in D3cold, because config
2487 	 * registers are still accessible for devices suspended but
2488 	 * not in D3cold.
2489 	 */
2490 	if (pdev->current_state == PCI_D3cold)
2491 		pm_runtime_resume(dev);
2492 }
2493 
2494 void pci_config_pm_runtime_put(struct pci_dev *pdev)
2495 {
2496 	struct device *dev = &pdev->dev;
2497 	struct device *parent = dev->parent;
2498 
2499 	pm_runtime_put(dev);
2500 	if (parent)
2501 		pm_runtime_put_sync(parent);
2502 }
2503 
2504 /**
2505  * pci_bridge_d3_possible - Is it possible to put the bridge into D3
2506  * @bridge: Bridge to check
2507  *
2508  * This function checks if it is possible to move the bridge to D3.
2509  * Currently we only allow D3 for recent enough PCIe ports and Thunderbolt.
2510  */
2511 bool pci_bridge_d3_possible(struct pci_dev *bridge)
2512 {
2513 	if (!pci_is_pcie(bridge))
2514 		return false;
2515 
2516 	switch (pci_pcie_type(bridge)) {
2517 	case PCI_EXP_TYPE_ROOT_PORT:
2518 	case PCI_EXP_TYPE_UPSTREAM:
2519 	case PCI_EXP_TYPE_DOWNSTREAM:
2520 		if (pci_bridge_d3_disable)
2521 			return false;
2522 
2523 		/*
2524 		 * Hotplug ports handled by firmware in System Management Mode
2525 		 * may not be put into D3 by the OS (Thunderbolt on non-Macs).
2526 		 */
2527 		if (bridge->is_hotplug_bridge && !pciehp_is_native(bridge))
2528 			return false;
2529 
2530 		if (pci_bridge_d3_force)
2531 			return true;
2532 
2533 		/* Even the oldest 2010 Thunderbolt controller supports D3. */
2534 		if (bridge->is_thunderbolt)
2535 			return true;
2536 
2537 		/* Platform might know better if the bridge supports D3 */
2538 		if (platform_pci_bridge_d3(bridge))
2539 			return true;
2540 
2541 		/*
2542 		 * Hotplug ports handled natively by the OS were not validated
2543 		 * by vendors for runtime D3 at least until 2018 because there
2544 		 * was no OS support.
2545 		 */
2546 		if (bridge->is_hotplug_bridge)
2547 			return false;
2548 
2549 		/*
2550 		 * It should be safe to put PCIe ports from 2015 or newer
2551 		 * to D3.
2552 		 */
2553 		if (dmi_get_bios_year() >= 2015)
2554 			return true;
2555 		break;
2556 	}
2557 
2558 	return false;
2559 }
2560 
2561 static int pci_dev_check_d3cold(struct pci_dev *dev, void *data)
2562 {
2563 	bool *d3cold_ok = data;
2564 
2565 	if (/* The device needs to be allowed to go D3cold ... */
2566 	    dev->no_d3cold || !dev->d3cold_allowed ||
2567 
2568 	    /* ... and if it is wakeup capable to do so from D3cold. */
2569 	    (device_may_wakeup(&dev->dev) &&
2570 	     !pci_pme_capable(dev, PCI_D3cold)) ||
2571 
2572 	    /* If it is a bridge it must be allowed to go to D3. */
2573 	    !pci_power_manageable(dev))
2574 
2575 		*d3cold_ok = false;
2576 
2577 	return !*d3cold_ok;
2578 }
2579 
2580 /*
2581  * pci_bridge_d3_update - Update bridge D3 capabilities
2582  * @dev: PCI device which is changed
2583  *
2584  * Update upstream bridge PM capabilities accordingly depending on if the
2585  * device PM configuration was changed or the device is being removed.  The
2586  * change is also propagated upstream.
2587  */
2588 void pci_bridge_d3_update(struct pci_dev *dev)
2589 {
2590 	bool remove = !device_is_registered(&dev->dev);
2591 	struct pci_dev *bridge;
2592 	bool d3cold_ok = true;
2593 
2594 	bridge = pci_upstream_bridge(dev);
2595 	if (!bridge || !pci_bridge_d3_possible(bridge))
2596 		return;
2597 
2598 	/*
2599 	 * If D3 is currently allowed for the bridge, removing one of its
2600 	 * children won't change that.
2601 	 */
2602 	if (remove && bridge->bridge_d3)
2603 		return;
2604 
2605 	/*
2606 	 * If D3 is currently allowed for the bridge and a child is added or
2607 	 * changed, disallowance of D3 can only be caused by that child, so
2608 	 * we only need to check that single device, not any of its siblings.
2609 	 *
2610 	 * If D3 is currently not allowed for the bridge, checking the device
2611 	 * first may allow us to skip checking its siblings.
2612 	 */
2613 	if (!remove)
2614 		pci_dev_check_d3cold(dev, &d3cold_ok);
2615 
2616 	/*
2617 	 * If D3 is currently not allowed for the bridge, this may be caused
2618 	 * either by the device being changed/removed or any of its siblings,
2619 	 * so we need to go through all children to find out if one of them
2620 	 * continues to block D3.
2621 	 */
2622 	if (d3cold_ok && !bridge->bridge_d3)
2623 		pci_walk_bus(bridge->subordinate, pci_dev_check_d3cold,
2624 			     &d3cold_ok);
2625 
2626 	if (bridge->bridge_d3 != d3cold_ok) {
2627 		bridge->bridge_d3 = d3cold_ok;
2628 		/* Propagate change to upstream bridges */
2629 		pci_bridge_d3_update(bridge);
2630 	}
2631 }
2632 
2633 /**
2634  * pci_d3cold_enable - Enable D3cold for device
2635  * @dev: PCI device to handle
2636  *
2637  * This function can be used in drivers to enable D3cold from the device
2638  * they handle.  It also updates upstream PCI bridge PM capabilities
2639  * accordingly.
2640  */
2641 void pci_d3cold_enable(struct pci_dev *dev)
2642 {
2643 	if (dev->no_d3cold) {
2644 		dev->no_d3cold = false;
2645 		pci_bridge_d3_update(dev);
2646 	}
2647 }
2648 EXPORT_SYMBOL_GPL(pci_d3cold_enable);
2649 
2650 /**
2651  * pci_d3cold_disable - Disable D3cold for device
2652  * @dev: PCI device to handle
2653  *
2654  * This function can be used in drivers to disable D3cold from the device
2655  * they handle.  It also updates upstream PCI bridge PM capabilities
2656  * accordingly.
2657  */
2658 void pci_d3cold_disable(struct pci_dev *dev)
2659 {
2660 	if (!dev->no_d3cold) {
2661 		dev->no_d3cold = true;
2662 		pci_bridge_d3_update(dev);
2663 	}
2664 }
2665 EXPORT_SYMBOL_GPL(pci_d3cold_disable);
2666 
2667 /**
2668  * pci_pm_init - Initialize PM functions of given PCI device
2669  * @dev: PCI device to handle.
2670  */
2671 void pci_pm_init(struct pci_dev *dev)
2672 {
2673 	int pm;
2674 	u16 status;
2675 	u16 pmc;
2676 
2677 	pm_runtime_forbid(&dev->dev);
2678 	pm_runtime_set_active(&dev->dev);
2679 	pm_runtime_enable(&dev->dev);
2680 	device_enable_async_suspend(&dev->dev);
2681 	dev->wakeup_prepared = false;
2682 
2683 	dev->pm_cap = 0;
2684 	dev->pme_support = 0;
2685 
2686 	/* find PCI PM capability in list */
2687 	pm = pci_find_capability(dev, PCI_CAP_ID_PM);
2688 	if (!pm)
2689 		return;
2690 	/* Check device's ability to generate PME# */
2691 	pci_read_config_word(dev, pm + PCI_PM_PMC, &pmc);
2692 
2693 	if ((pmc & PCI_PM_CAP_VER_MASK) > 3) {
2694 		pci_err(dev, "unsupported PM cap regs version (%u)\n",
2695 			pmc & PCI_PM_CAP_VER_MASK);
2696 		return;
2697 	}
2698 
2699 	dev->pm_cap = pm;
2700 	dev->d3_delay = PCI_PM_D3_WAIT;
2701 	dev->d3cold_delay = PCI_PM_D3COLD_WAIT;
2702 	dev->bridge_d3 = pci_bridge_d3_possible(dev);
2703 	dev->d3cold_allowed = true;
2704 
2705 	dev->d1_support = false;
2706 	dev->d2_support = false;
2707 	if (!pci_no_d1d2(dev)) {
2708 		if (pmc & PCI_PM_CAP_D1)
2709 			dev->d1_support = true;
2710 		if (pmc & PCI_PM_CAP_D2)
2711 			dev->d2_support = true;
2712 
2713 		if (dev->d1_support || dev->d2_support)
2714 			pci_printk(KERN_DEBUG, dev, "supports%s%s\n",
2715 				   dev->d1_support ? " D1" : "",
2716 				   dev->d2_support ? " D2" : "");
2717 	}
2718 
2719 	pmc &= PCI_PM_CAP_PME_MASK;
2720 	if (pmc) {
2721 		pci_printk(KERN_DEBUG, dev, "PME# supported from%s%s%s%s%s\n",
2722 			 (pmc & PCI_PM_CAP_PME_D0) ? " D0" : "",
2723 			 (pmc & PCI_PM_CAP_PME_D1) ? " D1" : "",
2724 			 (pmc & PCI_PM_CAP_PME_D2) ? " D2" : "",
2725 			 (pmc & PCI_PM_CAP_PME_D3) ? " D3hot" : "",
2726 			 (pmc & PCI_PM_CAP_PME_D3cold) ? " D3cold" : "");
2727 		dev->pme_support = pmc >> PCI_PM_CAP_PME_SHIFT;
2728 		dev->pme_poll = true;
2729 		/*
2730 		 * Make device's PM flags reflect the wake-up capability, but
2731 		 * let the user space enable it to wake up the system as needed.
2732 		 */
2733 		device_set_wakeup_capable(&dev->dev, true);
2734 		/* Disable the PME# generation functionality */
2735 		pci_pme_active(dev, false);
2736 	}
2737 
2738 	pci_read_config_word(dev, PCI_STATUS, &status);
2739 	if (status & PCI_STATUS_IMM_READY)
2740 		dev->imm_ready = 1;
2741 }
2742 
2743 static unsigned long pci_ea_flags(struct pci_dev *dev, u8 prop)
2744 {
2745 	unsigned long flags = IORESOURCE_PCI_FIXED | IORESOURCE_PCI_EA_BEI;
2746 
2747 	switch (prop) {
2748 	case PCI_EA_P_MEM:
2749 	case PCI_EA_P_VF_MEM:
2750 		flags |= IORESOURCE_MEM;
2751 		break;
2752 	case PCI_EA_P_MEM_PREFETCH:
2753 	case PCI_EA_P_VF_MEM_PREFETCH:
2754 		flags |= IORESOURCE_MEM | IORESOURCE_PREFETCH;
2755 		break;
2756 	case PCI_EA_P_IO:
2757 		flags |= IORESOURCE_IO;
2758 		break;
2759 	default:
2760 		return 0;
2761 	}
2762 
2763 	return flags;
2764 }
2765 
2766 static struct resource *pci_ea_get_resource(struct pci_dev *dev, u8 bei,
2767 					    u8 prop)
2768 {
2769 	if (bei <= PCI_EA_BEI_BAR5 && prop <= PCI_EA_P_IO)
2770 		return &dev->resource[bei];
2771 #ifdef CONFIG_PCI_IOV
2772 	else if (bei >= PCI_EA_BEI_VF_BAR0 && bei <= PCI_EA_BEI_VF_BAR5 &&
2773 		 (prop == PCI_EA_P_VF_MEM || prop == PCI_EA_P_VF_MEM_PREFETCH))
2774 		return &dev->resource[PCI_IOV_RESOURCES +
2775 				      bei - PCI_EA_BEI_VF_BAR0];
2776 #endif
2777 	else if (bei == PCI_EA_BEI_ROM)
2778 		return &dev->resource[PCI_ROM_RESOURCE];
2779 	else
2780 		return NULL;
2781 }
2782 
2783 /* Read an Enhanced Allocation (EA) entry */
2784 static int pci_ea_read(struct pci_dev *dev, int offset)
2785 {
2786 	struct resource *res;
2787 	int ent_size, ent_offset = offset;
2788 	resource_size_t start, end;
2789 	unsigned long flags;
2790 	u32 dw0, bei, base, max_offset;
2791 	u8 prop;
2792 	bool support_64 = (sizeof(resource_size_t) >= 8);
2793 
2794 	pci_read_config_dword(dev, ent_offset, &dw0);
2795 	ent_offset += 4;
2796 
2797 	/* Entry size field indicates DWORDs after 1st */
2798 	ent_size = ((dw0 & PCI_EA_ES) + 1) << 2;
2799 
2800 	if (!(dw0 & PCI_EA_ENABLE)) /* Entry not enabled */
2801 		goto out;
2802 
2803 	bei = (dw0 & PCI_EA_BEI) >> 4;
2804 	prop = (dw0 & PCI_EA_PP) >> 8;
2805 
2806 	/*
2807 	 * If the Property is in the reserved range, try the Secondary
2808 	 * Property instead.
2809 	 */
2810 	if (prop > PCI_EA_P_BRIDGE_IO && prop < PCI_EA_P_MEM_RESERVED)
2811 		prop = (dw0 & PCI_EA_SP) >> 16;
2812 	if (prop > PCI_EA_P_BRIDGE_IO)
2813 		goto out;
2814 
2815 	res = pci_ea_get_resource(dev, bei, prop);
2816 	if (!res) {
2817 		pci_err(dev, "Unsupported EA entry BEI: %u\n", bei);
2818 		goto out;
2819 	}
2820 
2821 	flags = pci_ea_flags(dev, prop);
2822 	if (!flags) {
2823 		pci_err(dev, "Unsupported EA properties: %#x\n", prop);
2824 		goto out;
2825 	}
2826 
2827 	/* Read Base */
2828 	pci_read_config_dword(dev, ent_offset, &base);
2829 	start = (base & PCI_EA_FIELD_MASK);
2830 	ent_offset += 4;
2831 
2832 	/* Read MaxOffset */
2833 	pci_read_config_dword(dev, ent_offset, &max_offset);
2834 	ent_offset += 4;
2835 
2836 	/* Read Base MSBs (if 64-bit entry) */
2837 	if (base & PCI_EA_IS_64) {
2838 		u32 base_upper;
2839 
2840 		pci_read_config_dword(dev, ent_offset, &base_upper);
2841 		ent_offset += 4;
2842 
2843 		flags |= IORESOURCE_MEM_64;
2844 
2845 		/* entry starts above 32-bit boundary, can't use */
2846 		if (!support_64 && base_upper)
2847 			goto out;
2848 
2849 		if (support_64)
2850 			start |= ((u64)base_upper << 32);
2851 	}
2852 
2853 	end = start + (max_offset | 0x03);
2854 
2855 	/* Read MaxOffset MSBs (if 64-bit entry) */
2856 	if (max_offset & PCI_EA_IS_64) {
2857 		u32 max_offset_upper;
2858 
2859 		pci_read_config_dword(dev, ent_offset, &max_offset_upper);
2860 		ent_offset += 4;
2861 
2862 		flags |= IORESOURCE_MEM_64;
2863 
2864 		/* entry too big, can't use */
2865 		if (!support_64 && max_offset_upper)
2866 			goto out;
2867 
2868 		if (support_64)
2869 			end += ((u64)max_offset_upper << 32);
2870 	}
2871 
2872 	if (end < start) {
2873 		pci_err(dev, "EA Entry crosses address boundary\n");
2874 		goto out;
2875 	}
2876 
2877 	if (ent_size != ent_offset - offset) {
2878 		pci_err(dev, "EA Entry Size (%d) does not match length read (%d)\n",
2879 			ent_size, ent_offset - offset);
2880 		goto out;
2881 	}
2882 
2883 	res->name = pci_name(dev);
2884 	res->start = start;
2885 	res->end = end;
2886 	res->flags = flags;
2887 
2888 	if (bei <= PCI_EA_BEI_BAR5)
2889 		pci_printk(KERN_DEBUG, dev, "BAR %d: %pR (from Enhanced Allocation, properties %#02x)\n",
2890 			   bei, res, prop);
2891 	else if (bei == PCI_EA_BEI_ROM)
2892 		pci_printk(KERN_DEBUG, dev, "ROM: %pR (from Enhanced Allocation, properties %#02x)\n",
2893 			   res, prop);
2894 	else if (bei >= PCI_EA_BEI_VF_BAR0 && bei <= PCI_EA_BEI_VF_BAR5)
2895 		pci_printk(KERN_DEBUG, dev, "VF BAR %d: %pR (from Enhanced Allocation, properties %#02x)\n",
2896 			   bei - PCI_EA_BEI_VF_BAR0, res, prop);
2897 	else
2898 		pci_printk(KERN_DEBUG, dev, "BEI %d res: %pR (from Enhanced Allocation, properties %#02x)\n",
2899 			   bei, res, prop);
2900 
2901 out:
2902 	return offset + ent_size;
2903 }
2904 
2905 /* Enhanced Allocation Initialization */
2906 void pci_ea_init(struct pci_dev *dev)
2907 {
2908 	int ea;
2909 	u8 num_ent;
2910 	int offset;
2911 	int i;
2912 
2913 	/* find PCI EA capability in list */
2914 	ea = pci_find_capability(dev, PCI_CAP_ID_EA);
2915 	if (!ea)
2916 		return;
2917 
2918 	/* determine the number of entries */
2919 	pci_bus_read_config_byte(dev->bus, dev->devfn, ea + PCI_EA_NUM_ENT,
2920 					&num_ent);
2921 	num_ent &= PCI_EA_NUM_ENT_MASK;
2922 
2923 	offset = ea + PCI_EA_FIRST_ENT;
2924 
2925 	/* Skip DWORD 2 for type 1 functions */
2926 	if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE)
2927 		offset += 4;
2928 
2929 	/* parse each EA entry */
2930 	for (i = 0; i < num_ent; ++i)
2931 		offset = pci_ea_read(dev, offset);
2932 }
2933 
2934 static void pci_add_saved_cap(struct pci_dev *pci_dev,
2935 	struct pci_cap_saved_state *new_cap)
2936 {
2937 	hlist_add_head(&new_cap->next, &pci_dev->saved_cap_space);
2938 }
2939 
2940 /**
2941  * _pci_add_cap_save_buffer - allocate buffer for saving given
2942  *                            capability registers
2943  * @dev: the PCI device
2944  * @cap: the capability to allocate the buffer for
2945  * @extended: Standard or Extended capability ID
2946  * @size: requested size of the buffer
2947  */
2948 static int _pci_add_cap_save_buffer(struct pci_dev *dev, u16 cap,
2949 				    bool extended, unsigned int size)
2950 {
2951 	int pos;
2952 	struct pci_cap_saved_state *save_state;
2953 
2954 	if (extended)
2955 		pos = pci_find_ext_capability(dev, cap);
2956 	else
2957 		pos = pci_find_capability(dev, cap);
2958 
2959 	if (!pos)
2960 		return 0;
2961 
2962 	save_state = kzalloc(sizeof(*save_state) + size, GFP_KERNEL);
2963 	if (!save_state)
2964 		return -ENOMEM;
2965 
2966 	save_state->cap.cap_nr = cap;
2967 	save_state->cap.cap_extended = extended;
2968 	save_state->cap.size = size;
2969 	pci_add_saved_cap(dev, save_state);
2970 
2971 	return 0;
2972 }
2973 
2974 int pci_add_cap_save_buffer(struct pci_dev *dev, char cap, unsigned int size)
2975 {
2976 	return _pci_add_cap_save_buffer(dev, cap, false, size);
2977 }
2978 
2979 int pci_add_ext_cap_save_buffer(struct pci_dev *dev, u16 cap, unsigned int size)
2980 {
2981 	return _pci_add_cap_save_buffer(dev, cap, true, size);
2982 }
2983 
2984 /**
2985  * pci_allocate_cap_save_buffers - allocate buffers for saving capabilities
2986  * @dev: the PCI device
2987  */
2988 void pci_allocate_cap_save_buffers(struct pci_dev *dev)
2989 {
2990 	int error;
2991 
2992 	error = pci_add_cap_save_buffer(dev, PCI_CAP_ID_EXP,
2993 					PCI_EXP_SAVE_REGS * sizeof(u16));
2994 	if (error)
2995 		pci_err(dev, "unable to preallocate PCI Express save buffer\n");
2996 
2997 	error = pci_add_cap_save_buffer(dev, PCI_CAP_ID_PCIX, sizeof(u16));
2998 	if (error)
2999 		pci_err(dev, "unable to preallocate PCI-X save buffer\n");
3000 
3001 	pci_allocate_vc_save_buffers(dev);
3002 }
3003 
3004 void pci_free_cap_save_buffers(struct pci_dev *dev)
3005 {
3006 	struct pci_cap_saved_state *tmp;
3007 	struct hlist_node *n;
3008 
3009 	hlist_for_each_entry_safe(tmp, n, &dev->saved_cap_space, next)
3010 		kfree(tmp);
3011 }
3012 
3013 /**
3014  * pci_configure_ari - enable or disable ARI forwarding
3015  * @dev: the PCI device
3016  *
3017  * If @dev and its upstream bridge both support ARI, enable ARI in the
3018  * bridge.  Otherwise, disable ARI in the bridge.
3019  */
3020 void pci_configure_ari(struct pci_dev *dev)
3021 {
3022 	u32 cap;
3023 	struct pci_dev *bridge;
3024 
3025 	if (pcie_ari_disabled || !pci_is_pcie(dev) || dev->devfn)
3026 		return;
3027 
3028 	bridge = dev->bus->self;
3029 	if (!bridge)
3030 		return;
3031 
3032 	pcie_capability_read_dword(bridge, PCI_EXP_DEVCAP2, &cap);
3033 	if (!(cap & PCI_EXP_DEVCAP2_ARI))
3034 		return;
3035 
3036 	if (pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ARI)) {
3037 		pcie_capability_set_word(bridge, PCI_EXP_DEVCTL2,
3038 					 PCI_EXP_DEVCTL2_ARI);
3039 		bridge->ari_enabled = 1;
3040 	} else {
3041 		pcie_capability_clear_word(bridge, PCI_EXP_DEVCTL2,
3042 					   PCI_EXP_DEVCTL2_ARI);
3043 		bridge->ari_enabled = 0;
3044 	}
3045 }
3046 
3047 static int pci_acs_enable;
3048 
3049 /**
3050  * pci_request_acs - ask for ACS to be enabled if supported
3051  */
3052 void pci_request_acs(void)
3053 {
3054 	pci_acs_enable = 1;
3055 }
3056 
3057 static const char *disable_acs_redir_param;
3058 
3059 /**
3060  * pci_disable_acs_redir - disable ACS redirect capabilities
3061  * @dev: the PCI device
3062  *
3063  * For only devices specified in the disable_acs_redir parameter.
3064  */
3065 static void pci_disable_acs_redir(struct pci_dev *dev)
3066 {
3067 	int ret = 0;
3068 	const char *p;
3069 	int pos;
3070 	u16 ctrl;
3071 
3072 	if (!disable_acs_redir_param)
3073 		return;
3074 
3075 	p = disable_acs_redir_param;
3076 	while (*p) {
3077 		ret = pci_dev_str_match(dev, p, &p);
3078 		if (ret < 0) {
3079 			pr_info_once("PCI: Can't parse disable_acs_redir parameter: %s\n",
3080 				     disable_acs_redir_param);
3081 
3082 			break;
3083 		} else if (ret == 1) {
3084 			/* Found a match */
3085 			break;
3086 		}
3087 
3088 		if (*p != ';' && *p != ',') {
3089 			/* End of param or invalid format */
3090 			break;
3091 		}
3092 		p++;
3093 	}
3094 
3095 	if (ret != 1)
3096 		return;
3097 
3098 	if (!pci_dev_specific_disable_acs_redir(dev))
3099 		return;
3100 
3101 	pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ACS);
3102 	if (!pos) {
3103 		pci_warn(dev, "cannot disable ACS redirect for this hardware as it does not have ACS capabilities\n");
3104 		return;
3105 	}
3106 
3107 	pci_read_config_word(dev, pos + PCI_ACS_CTRL, &ctrl);
3108 
3109 	/* P2P Request & Completion Redirect */
3110 	ctrl &= ~(PCI_ACS_RR | PCI_ACS_CR | PCI_ACS_EC);
3111 
3112 	pci_write_config_word(dev, pos + PCI_ACS_CTRL, ctrl);
3113 
3114 	pci_info(dev, "disabled ACS redirect\n");
3115 }
3116 
3117 /**
3118  * pci_std_enable_acs - enable ACS on devices using standard ACS capabilites
3119  * @dev: the PCI device
3120  */
3121 static void pci_std_enable_acs(struct pci_dev *dev)
3122 {
3123 	int pos;
3124 	u16 cap;
3125 	u16 ctrl;
3126 
3127 	pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ACS);
3128 	if (!pos)
3129 		return;
3130 
3131 	pci_read_config_word(dev, pos + PCI_ACS_CAP, &cap);
3132 	pci_read_config_word(dev, pos + PCI_ACS_CTRL, &ctrl);
3133 
3134 	/* Source Validation */
3135 	ctrl |= (cap & PCI_ACS_SV);
3136 
3137 	/* P2P Request Redirect */
3138 	ctrl |= (cap & PCI_ACS_RR);
3139 
3140 	/* P2P Completion Redirect */
3141 	ctrl |= (cap & PCI_ACS_CR);
3142 
3143 	/* Upstream Forwarding */
3144 	ctrl |= (cap & PCI_ACS_UF);
3145 
3146 	pci_write_config_word(dev, pos + PCI_ACS_CTRL, ctrl);
3147 }
3148 
3149 /**
3150  * pci_enable_acs - enable ACS if hardware support it
3151  * @dev: the PCI device
3152  */
3153 void pci_enable_acs(struct pci_dev *dev)
3154 {
3155 	if (!pci_acs_enable)
3156 		goto disable_acs_redir;
3157 
3158 	if (!pci_dev_specific_enable_acs(dev))
3159 		goto disable_acs_redir;
3160 
3161 	pci_std_enable_acs(dev);
3162 
3163 disable_acs_redir:
3164 	/*
3165 	 * Note: pci_disable_acs_redir() must be called even if ACS was not
3166 	 * enabled by the kernel because it may have been enabled by
3167 	 * platform firmware.  So if we are told to disable it, we should
3168 	 * always disable it after setting the kernel's default
3169 	 * preferences.
3170 	 */
3171 	pci_disable_acs_redir(dev);
3172 }
3173 
3174 static bool pci_acs_flags_enabled(struct pci_dev *pdev, u16 acs_flags)
3175 {
3176 	int pos;
3177 	u16 cap, ctrl;
3178 
3179 	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ACS);
3180 	if (!pos)
3181 		return false;
3182 
3183 	/*
3184 	 * Except for egress control, capabilities are either required
3185 	 * or only required if controllable.  Features missing from the
3186 	 * capability field can therefore be assumed as hard-wired enabled.
3187 	 */
3188 	pci_read_config_word(pdev, pos + PCI_ACS_CAP, &cap);
3189 	acs_flags &= (cap | PCI_ACS_EC);
3190 
3191 	pci_read_config_word(pdev, pos + PCI_ACS_CTRL, &ctrl);
3192 	return (ctrl & acs_flags) == acs_flags;
3193 }
3194 
3195 /**
3196  * pci_acs_enabled - test ACS against required flags for a given device
3197  * @pdev: device to test
3198  * @acs_flags: required PCI ACS flags
3199  *
3200  * Return true if the device supports the provided flags.  Automatically
3201  * filters out flags that are not implemented on multifunction devices.
3202  *
3203  * Note that this interface checks the effective ACS capabilities of the
3204  * device rather than the actual capabilities.  For instance, most single
3205  * function endpoints are not required to support ACS because they have no
3206  * opportunity for peer-to-peer access.  We therefore return 'true'
3207  * regardless of whether the device exposes an ACS capability.  This makes
3208  * it much easier for callers of this function to ignore the actual type
3209  * or topology of the device when testing ACS support.
3210  */
3211 bool pci_acs_enabled(struct pci_dev *pdev, u16 acs_flags)
3212 {
3213 	int ret;
3214 
3215 	ret = pci_dev_specific_acs_enabled(pdev, acs_flags);
3216 	if (ret >= 0)
3217 		return ret > 0;
3218 
3219 	/*
3220 	 * Conventional PCI and PCI-X devices never support ACS, either
3221 	 * effectively or actually.  The shared bus topology implies that
3222 	 * any device on the bus can receive or snoop DMA.
3223 	 */
3224 	if (!pci_is_pcie(pdev))
3225 		return false;
3226 
3227 	switch (pci_pcie_type(pdev)) {
3228 	/*
3229 	 * PCI/X-to-PCIe bridges are not specifically mentioned by the spec,
3230 	 * but since their primary interface is PCI/X, we conservatively
3231 	 * handle them as we would a non-PCIe device.
3232 	 */
3233 	case PCI_EXP_TYPE_PCIE_BRIDGE:
3234 	/*
3235 	 * PCIe 3.0, 6.12.1 excludes ACS on these devices.  "ACS is never
3236 	 * applicable... must never implement an ACS Extended Capability...".
3237 	 * This seems arbitrary, but we take a conservative interpretation
3238 	 * of this statement.
3239 	 */
3240 	case PCI_EXP_TYPE_PCI_BRIDGE:
3241 	case PCI_EXP_TYPE_RC_EC:
3242 		return false;
3243 	/*
3244 	 * PCIe 3.0, 6.12.1.1 specifies that downstream and root ports should
3245 	 * implement ACS in order to indicate their peer-to-peer capabilities,
3246 	 * regardless of whether they are single- or multi-function devices.
3247 	 */
3248 	case PCI_EXP_TYPE_DOWNSTREAM:
3249 	case PCI_EXP_TYPE_ROOT_PORT:
3250 		return pci_acs_flags_enabled(pdev, acs_flags);
3251 	/*
3252 	 * PCIe 3.0, 6.12.1.2 specifies ACS capabilities that should be
3253 	 * implemented by the remaining PCIe types to indicate peer-to-peer
3254 	 * capabilities, but only when they are part of a multifunction
3255 	 * device.  The footnote for section 6.12 indicates the specific
3256 	 * PCIe types included here.
3257 	 */
3258 	case PCI_EXP_TYPE_ENDPOINT:
3259 	case PCI_EXP_TYPE_UPSTREAM:
3260 	case PCI_EXP_TYPE_LEG_END:
3261 	case PCI_EXP_TYPE_RC_END:
3262 		if (!pdev->multifunction)
3263 			break;
3264 
3265 		return pci_acs_flags_enabled(pdev, acs_flags);
3266 	}
3267 
3268 	/*
3269 	 * PCIe 3.0, 6.12.1.3 specifies no ACS capabilities are applicable
3270 	 * to single function devices with the exception of downstream ports.
3271 	 */
3272 	return true;
3273 }
3274 
3275 /**
3276  * pci_acs_path_enable - test ACS flags from start to end in a hierarchy
3277  * @start: starting downstream device
3278  * @end: ending upstream device or NULL to search to the root bus
3279  * @acs_flags: required flags
3280  *
3281  * Walk up a device tree from start to end testing PCI ACS support.  If
3282  * any step along the way does not support the required flags, return false.
3283  */
3284 bool pci_acs_path_enabled(struct pci_dev *start,
3285 			  struct pci_dev *end, u16 acs_flags)
3286 {
3287 	struct pci_dev *pdev, *parent = start;
3288 
3289 	do {
3290 		pdev = parent;
3291 
3292 		if (!pci_acs_enabled(pdev, acs_flags))
3293 			return false;
3294 
3295 		if (pci_is_root_bus(pdev->bus))
3296 			return (end == NULL);
3297 
3298 		parent = pdev->bus->self;
3299 	} while (pdev != end);
3300 
3301 	return true;
3302 }
3303 
3304 /**
3305  * pci_rebar_find_pos - find position of resize ctrl reg for BAR
3306  * @pdev: PCI device
3307  * @bar: BAR to find
3308  *
3309  * Helper to find the position of the ctrl register for a BAR.
3310  * Returns -ENOTSUPP if resizable BARs are not supported at all.
3311  * Returns -ENOENT if no ctrl register for the BAR could be found.
3312  */
3313 static int pci_rebar_find_pos(struct pci_dev *pdev, int bar)
3314 {
3315 	unsigned int pos, nbars, i;
3316 	u32 ctrl;
3317 
3318 	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_REBAR);
3319 	if (!pos)
3320 		return -ENOTSUPP;
3321 
3322 	pci_read_config_dword(pdev, pos + PCI_REBAR_CTRL, &ctrl);
3323 	nbars = (ctrl & PCI_REBAR_CTRL_NBAR_MASK) >>
3324 		    PCI_REBAR_CTRL_NBAR_SHIFT;
3325 
3326 	for (i = 0; i < nbars; i++, pos += 8) {
3327 		int bar_idx;
3328 
3329 		pci_read_config_dword(pdev, pos + PCI_REBAR_CTRL, &ctrl);
3330 		bar_idx = ctrl & PCI_REBAR_CTRL_BAR_IDX;
3331 		if (bar_idx == bar)
3332 			return pos;
3333 	}
3334 
3335 	return -ENOENT;
3336 }
3337 
3338 /**
3339  * pci_rebar_get_possible_sizes - get possible sizes for BAR
3340  * @pdev: PCI device
3341  * @bar: BAR to query
3342  *
3343  * Get the possible sizes of a resizable BAR as bitmask defined in the spec
3344  * (bit 0=1MB, bit 19=512GB). Returns 0 if BAR isn't resizable.
3345  */
3346 u32 pci_rebar_get_possible_sizes(struct pci_dev *pdev, int bar)
3347 {
3348 	int pos;
3349 	u32 cap;
3350 
3351 	pos = pci_rebar_find_pos(pdev, bar);
3352 	if (pos < 0)
3353 		return 0;
3354 
3355 	pci_read_config_dword(pdev, pos + PCI_REBAR_CAP, &cap);
3356 	return (cap & PCI_REBAR_CAP_SIZES) >> 4;
3357 }
3358 
3359 /**
3360  * pci_rebar_get_current_size - get the current size of a BAR
3361  * @pdev: PCI device
3362  * @bar: BAR to set size to
3363  *
3364  * Read the size of a BAR from the resizable BAR config.
3365  * Returns size if found or negative error code.
3366  */
3367 int pci_rebar_get_current_size(struct pci_dev *pdev, int bar)
3368 {
3369 	int pos;
3370 	u32 ctrl;
3371 
3372 	pos = pci_rebar_find_pos(pdev, bar);
3373 	if (pos < 0)
3374 		return pos;
3375 
3376 	pci_read_config_dword(pdev, pos + PCI_REBAR_CTRL, &ctrl);
3377 	return (ctrl & PCI_REBAR_CTRL_BAR_SIZE) >> PCI_REBAR_CTRL_BAR_SHIFT;
3378 }
3379 
3380 /**
3381  * pci_rebar_set_size - set a new size for a BAR
3382  * @pdev: PCI device
3383  * @bar: BAR to set size to
3384  * @size: new size as defined in the spec (0=1MB, 19=512GB)
3385  *
3386  * Set the new size of a BAR as defined in the spec.
3387  * Returns zero if resizing was successful, error code otherwise.
3388  */
3389 int pci_rebar_set_size(struct pci_dev *pdev, int bar, int size)
3390 {
3391 	int pos;
3392 	u32 ctrl;
3393 
3394 	pos = pci_rebar_find_pos(pdev, bar);
3395 	if (pos < 0)
3396 		return pos;
3397 
3398 	pci_read_config_dword(pdev, pos + PCI_REBAR_CTRL, &ctrl);
3399 	ctrl &= ~PCI_REBAR_CTRL_BAR_SIZE;
3400 	ctrl |= size << PCI_REBAR_CTRL_BAR_SHIFT;
3401 	pci_write_config_dword(pdev, pos + PCI_REBAR_CTRL, ctrl);
3402 	return 0;
3403 }
3404 
3405 /**
3406  * pci_enable_atomic_ops_to_root - enable AtomicOp requests to root port
3407  * @dev: the PCI device
3408  * @cap_mask: mask of desired AtomicOp sizes, including one or more of:
3409  *	PCI_EXP_DEVCAP2_ATOMIC_COMP32
3410  *	PCI_EXP_DEVCAP2_ATOMIC_COMP64
3411  *	PCI_EXP_DEVCAP2_ATOMIC_COMP128
3412  *
3413  * Return 0 if all upstream bridges support AtomicOp routing, egress
3414  * blocking is disabled on all upstream ports, and the root port supports
3415  * the requested completion capabilities (32-bit, 64-bit and/or 128-bit
3416  * AtomicOp completion), or negative otherwise.
3417  */
3418 int pci_enable_atomic_ops_to_root(struct pci_dev *dev, u32 cap_mask)
3419 {
3420 	struct pci_bus *bus = dev->bus;
3421 	struct pci_dev *bridge;
3422 	u32 cap, ctl2;
3423 
3424 	if (!pci_is_pcie(dev))
3425 		return -EINVAL;
3426 
3427 	/*
3428 	 * Per PCIe r4.0, sec 6.15, endpoints and root ports may be
3429 	 * AtomicOp requesters.  For now, we only support endpoints as
3430 	 * requesters and root ports as completers.  No endpoints as
3431 	 * completers, and no peer-to-peer.
3432 	 */
3433 
3434 	switch (pci_pcie_type(dev)) {
3435 	case PCI_EXP_TYPE_ENDPOINT:
3436 	case PCI_EXP_TYPE_LEG_END:
3437 	case PCI_EXP_TYPE_RC_END:
3438 		break;
3439 	default:
3440 		return -EINVAL;
3441 	}
3442 
3443 	while (bus->parent) {
3444 		bridge = bus->self;
3445 
3446 		pcie_capability_read_dword(bridge, PCI_EXP_DEVCAP2, &cap);
3447 
3448 		switch (pci_pcie_type(bridge)) {
3449 		/* Ensure switch ports support AtomicOp routing */
3450 		case PCI_EXP_TYPE_UPSTREAM:
3451 		case PCI_EXP_TYPE_DOWNSTREAM:
3452 			if (!(cap & PCI_EXP_DEVCAP2_ATOMIC_ROUTE))
3453 				return -EINVAL;
3454 			break;
3455 
3456 		/* Ensure root port supports all the sizes we care about */
3457 		case PCI_EXP_TYPE_ROOT_PORT:
3458 			if ((cap & cap_mask) != cap_mask)
3459 				return -EINVAL;
3460 			break;
3461 		}
3462 
3463 		/* Ensure upstream ports don't block AtomicOps on egress */
3464 		if (!bridge->has_secondary_link) {
3465 			pcie_capability_read_dword(bridge, PCI_EXP_DEVCTL2,
3466 						   &ctl2);
3467 			if (ctl2 & PCI_EXP_DEVCTL2_ATOMIC_EGRESS_BLOCK)
3468 				return -EINVAL;
3469 		}
3470 
3471 		bus = bus->parent;
3472 	}
3473 
3474 	pcie_capability_set_word(dev, PCI_EXP_DEVCTL2,
3475 				 PCI_EXP_DEVCTL2_ATOMIC_REQ);
3476 	return 0;
3477 }
3478 EXPORT_SYMBOL(pci_enable_atomic_ops_to_root);
3479 
3480 /**
3481  * pci_swizzle_interrupt_pin - swizzle INTx for device behind bridge
3482  * @dev: the PCI device
3483  * @pin: the INTx pin (1=INTA, 2=INTB, 3=INTC, 4=INTD)
3484  *
3485  * Perform INTx swizzling for a device behind one level of bridge.  This is
3486  * required by section 9.1 of the PCI-to-PCI bridge specification for devices
3487  * behind bridges on add-in cards.  For devices with ARI enabled, the slot
3488  * number is always 0 (see the Implementation Note in section 2.2.8.1 of
3489  * the PCI Express Base Specification, Revision 2.1)
3490  */
3491 u8 pci_swizzle_interrupt_pin(const struct pci_dev *dev, u8 pin)
3492 {
3493 	int slot;
3494 
3495 	if (pci_ari_enabled(dev->bus))
3496 		slot = 0;
3497 	else
3498 		slot = PCI_SLOT(dev->devfn);
3499 
3500 	return (((pin - 1) + slot) % 4) + 1;
3501 }
3502 
3503 int pci_get_interrupt_pin(struct pci_dev *dev, struct pci_dev **bridge)
3504 {
3505 	u8 pin;
3506 
3507 	pin = dev->pin;
3508 	if (!pin)
3509 		return -1;
3510 
3511 	while (!pci_is_root_bus(dev->bus)) {
3512 		pin = pci_swizzle_interrupt_pin(dev, pin);
3513 		dev = dev->bus->self;
3514 	}
3515 	*bridge = dev;
3516 	return pin;
3517 }
3518 
3519 /**
3520  * pci_common_swizzle - swizzle INTx all the way to root bridge
3521  * @dev: the PCI device
3522  * @pinp: pointer to the INTx pin value (1=INTA, 2=INTB, 3=INTD, 4=INTD)
3523  *
3524  * Perform INTx swizzling for a device.  This traverses through all PCI-to-PCI
3525  * bridges all the way up to a PCI root bus.
3526  */
3527 u8 pci_common_swizzle(struct pci_dev *dev, u8 *pinp)
3528 {
3529 	u8 pin = *pinp;
3530 
3531 	while (!pci_is_root_bus(dev->bus)) {
3532 		pin = pci_swizzle_interrupt_pin(dev, pin);
3533 		dev = dev->bus->self;
3534 	}
3535 	*pinp = pin;
3536 	return PCI_SLOT(dev->devfn);
3537 }
3538 EXPORT_SYMBOL_GPL(pci_common_swizzle);
3539 
3540 /**
3541  *	pci_release_region - Release a PCI bar
3542  *	@pdev: PCI device whose resources were previously reserved by pci_request_region
3543  *	@bar: BAR to release
3544  *
3545  *	Releases the PCI I/O and memory resources previously reserved by a
3546  *	successful call to pci_request_region.  Call this function only
3547  *	after all use of the PCI regions has ceased.
3548  */
3549 void pci_release_region(struct pci_dev *pdev, int bar)
3550 {
3551 	struct pci_devres *dr;
3552 
3553 	if (pci_resource_len(pdev, bar) == 0)
3554 		return;
3555 	if (pci_resource_flags(pdev, bar) & IORESOURCE_IO)
3556 		release_region(pci_resource_start(pdev, bar),
3557 				pci_resource_len(pdev, bar));
3558 	else if (pci_resource_flags(pdev, bar) & IORESOURCE_MEM)
3559 		release_mem_region(pci_resource_start(pdev, bar),
3560 				pci_resource_len(pdev, bar));
3561 
3562 	dr = find_pci_dr(pdev);
3563 	if (dr)
3564 		dr->region_mask &= ~(1 << bar);
3565 }
3566 EXPORT_SYMBOL(pci_release_region);
3567 
3568 /**
3569  *	__pci_request_region - Reserved PCI I/O and memory resource
3570  *	@pdev: PCI device whose resources are to be reserved
3571  *	@bar: BAR to be reserved
3572  *	@res_name: Name to be associated with resource.
3573  *	@exclusive: whether the region access is exclusive or not
3574  *
3575  *	Mark the PCI region associated with PCI device @pdev BR @bar as
3576  *	being reserved by owner @res_name.  Do not access any
3577  *	address inside the PCI regions unless this call returns
3578  *	successfully.
3579  *
3580  *	If @exclusive is set, then the region is marked so that userspace
3581  *	is explicitly not allowed to map the resource via /dev/mem or
3582  *	sysfs MMIO access.
3583  *
3584  *	Returns 0 on success, or %EBUSY on error.  A warning
3585  *	message is also printed on failure.
3586  */
3587 static int __pci_request_region(struct pci_dev *pdev, int bar,
3588 				const char *res_name, int exclusive)
3589 {
3590 	struct pci_devres *dr;
3591 
3592 	if (pci_resource_len(pdev, bar) == 0)
3593 		return 0;
3594 
3595 	if (pci_resource_flags(pdev, bar) & IORESOURCE_IO) {
3596 		if (!request_region(pci_resource_start(pdev, bar),
3597 			    pci_resource_len(pdev, bar), res_name))
3598 			goto err_out;
3599 	} else if (pci_resource_flags(pdev, bar) & IORESOURCE_MEM) {
3600 		if (!__request_mem_region(pci_resource_start(pdev, bar),
3601 					pci_resource_len(pdev, bar), res_name,
3602 					exclusive))
3603 			goto err_out;
3604 	}
3605 
3606 	dr = find_pci_dr(pdev);
3607 	if (dr)
3608 		dr->region_mask |= 1 << bar;
3609 
3610 	return 0;
3611 
3612 err_out:
3613 	pci_warn(pdev, "BAR %d: can't reserve %pR\n", bar,
3614 		 &pdev->resource[bar]);
3615 	return -EBUSY;
3616 }
3617 
3618 /**
3619  *	pci_request_region - Reserve PCI I/O and memory resource
3620  *	@pdev: PCI device whose resources are to be reserved
3621  *	@bar: BAR to be reserved
3622  *	@res_name: Name to be associated with resource
3623  *
3624  *	Mark the PCI region associated with PCI device @pdev BAR @bar as
3625  *	being reserved by owner @res_name.  Do not access any
3626  *	address inside the PCI regions unless this call returns
3627  *	successfully.
3628  *
3629  *	Returns 0 on success, or %EBUSY on error.  A warning
3630  *	message is also printed on failure.
3631  */
3632 int pci_request_region(struct pci_dev *pdev, int bar, const char *res_name)
3633 {
3634 	return __pci_request_region(pdev, bar, res_name, 0);
3635 }
3636 EXPORT_SYMBOL(pci_request_region);
3637 
3638 /**
3639  *	pci_request_region_exclusive - Reserved PCI I/O and memory resource
3640  *	@pdev: PCI device whose resources are to be reserved
3641  *	@bar: BAR to be reserved
3642  *	@res_name: Name to be associated with resource.
3643  *
3644  *	Mark the PCI region associated with PCI device @pdev BR @bar as
3645  *	being reserved by owner @res_name.  Do not access any
3646  *	address inside the PCI regions unless this call returns
3647  *	successfully.
3648  *
3649  *	Returns 0 on success, or %EBUSY on error.  A warning
3650  *	message is also printed on failure.
3651  *
3652  *	The key difference that _exclusive makes it that userspace is
3653  *	explicitly not allowed to map the resource via /dev/mem or
3654  *	sysfs.
3655  */
3656 int pci_request_region_exclusive(struct pci_dev *pdev, int bar,
3657 				 const char *res_name)
3658 {
3659 	return __pci_request_region(pdev, bar, res_name, IORESOURCE_EXCLUSIVE);
3660 }
3661 EXPORT_SYMBOL(pci_request_region_exclusive);
3662 
3663 /**
3664  * pci_release_selected_regions - Release selected PCI I/O and memory resources
3665  * @pdev: PCI device whose resources were previously reserved
3666  * @bars: Bitmask of BARs to be released
3667  *
3668  * Release selected PCI I/O and memory resources previously reserved.
3669  * Call this function only after all use of the PCI regions has ceased.
3670  */
3671 void pci_release_selected_regions(struct pci_dev *pdev, int bars)
3672 {
3673 	int i;
3674 
3675 	for (i = 0; i < 6; i++)
3676 		if (bars & (1 << i))
3677 			pci_release_region(pdev, i);
3678 }
3679 EXPORT_SYMBOL(pci_release_selected_regions);
3680 
3681 static int __pci_request_selected_regions(struct pci_dev *pdev, int bars,
3682 					  const char *res_name, int excl)
3683 {
3684 	int i;
3685 
3686 	for (i = 0; i < 6; i++)
3687 		if (bars & (1 << i))
3688 			if (__pci_request_region(pdev, i, res_name, excl))
3689 				goto err_out;
3690 	return 0;
3691 
3692 err_out:
3693 	while (--i >= 0)
3694 		if (bars & (1 << i))
3695 			pci_release_region(pdev, i);
3696 
3697 	return -EBUSY;
3698 }
3699 
3700 
3701 /**
3702  * pci_request_selected_regions - Reserve selected PCI I/O and memory resources
3703  * @pdev: PCI device whose resources are to be reserved
3704  * @bars: Bitmask of BARs to be requested
3705  * @res_name: Name to be associated with resource
3706  */
3707 int pci_request_selected_regions(struct pci_dev *pdev, int bars,
3708 				 const char *res_name)
3709 {
3710 	return __pci_request_selected_regions(pdev, bars, res_name, 0);
3711 }
3712 EXPORT_SYMBOL(pci_request_selected_regions);
3713 
3714 int pci_request_selected_regions_exclusive(struct pci_dev *pdev, int bars,
3715 					   const char *res_name)
3716 {
3717 	return __pci_request_selected_regions(pdev, bars, res_name,
3718 			IORESOURCE_EXCLUSIVE);
3719 }
3720 EXPORT_SYMBOL(pci_request_selected_regions_exclusive);
3721 
3722 /**
3723  *	pci_release_regions - Release reserved PCI I/O and memory resources
3724  *	@pdev: PCI device whose resources were previously reserved by pci_request_regions
3725  *
3726  *	Releases all PCI I/O and memory resources previously reserved by a
3727  *	successful call to pci_request_regions.  Call this function only
3728  *	after all use of the PCI regions has ceased.
3729  */
3730 
3731 void pci_release_regions(struct pci_dev *pdev)
3732 {
3733 	pci_release_selected_regions(pdev, (1 << 6) - 1);
3734 }
3735 EXPORT_SYMBOL(pci_release_regions);
3736 
3737 /**
3738  *	pci_request_regions - Reserved PCI I/O and memory resources
3739  *	@pdev: PCI device whose resources are to be reserved
3740  *	@res_name: Name to be associated with resource.
3741  *
3742  *	Mark all PCI regions associated with PCI device @pdev as
3743  *	being reserved by owner @res_name.  Do not access any
3744  *	address inside the PCI regions unless this call returns
3745  *	successfully.
3746  *
3747  *	Returns 0 on success, or %EBUSY on error.  A warning
3748  *	message is also printed on failure.
3749  */
3750 int pci_request_regions(struct pci_dev *pdev, const char *res_name)
3751 {
3752 	return pci_request_selected_regions(pdev, ((1 << 6) - 1), res_name);
3753 }
3754 EXPORT_SYMBOL(pci_request_regions);
3755 
3756 /**
3757  *	pci_request_regions_exclusive - Reserved PCI I/O and memory resources
3758  *	@pdev: PCI device whose resources are to be reserved
3759  *	@res_name: Name to be associated with resource.
3760  *
3761  *	Mark all PCI regions associated with PCI device @pdev as
3762  *	being reserved by owner @res_name.  Do not access any
3763  *	address inside the PCI regions unless this call returns
3764  *	successfully.
3765  *
3766  *	pci_request_regions_exclusive() will mark the region so that
3767  *	/dev/mem and the sysfs MMIO access will not be allowed.
3768  *
3769  *	Returns 0 on success, or %EBUSY on error.  A warning
3770  *	message is also printed on failure.
3771  */
3772 int pci_request_regions_exclusive(struct pci_dev *pdev, const char *res_name)
3773 {
3774 	return pci_request_selected_regions_exclusive(pdev,
3775 					((1 << 6) - 1), res_name);
3776 }
3777 EXPORT_SYMBOL(pci_request_regions_exclusive);
3778 
3779 /*
3780  * Record the PCI IO range (expressed as CPU physical address + size).
3781  * Return a negative value if an error has occured, zero otherwise
3782  */
3783 int pci_register_io_range(struct fwnode_handle *fwnode, phys_addr_t addr,
3784 			resource_size_t	size)
3785 {
3786 	int ret = 0;
3787 #ifdef PCI_IOBASE
3788 	struct logic_pio_hwaddr *range;
3789 
3790 	if (!size || addr + size < addr)
3791 		return -EINVAL;
3792 
3793 	range = kzalloc(sizeof(*range), GFP_ATOMIC);
3794 	if (!range)
3795 		return -ENOMEM;
3796 
3797 	range->fwnode = fwnode;
3798 	range->size = size;
3799 	range->hw_start = addr;
3800 	range->flags = LOGIC_PIO_CPU_MMIO;
3801 
3802 	ret = logic_pio_register_range(range);
3803 	if (ret)
3804 		kfree(range);
3805 #endif
3806 
3807 	return ret;
3808 }
3809 
3810 phys_addr_t pci_pio_to_address(unsigned long pio)
3811 {
3812 	phys_addr_t address = (phys_addr_t)OF_BAD_ADDR;
3813 
3814 #ifdef PCI_IOBASE
3815 	if (pio >= MMIO_UPPER_LIMIT)
3816 		return address;
3817 
3818 	address = logic_pio_to_hwaddr(pio);
3819 #endif
3820 
3821 	return address;
3822 }
3823 
3824 unsigned long __weak pci_address_to_pio(phys_addr_t address)
3825 {
3826 #ifdef PCI_IOBASE
3827 	return logic_pio_trans_cpuaddr(address);
3828 #else
3829 	if (address > IO_SPACE_LIMIT)
3830 		return (unsigned long)-1;
3831 
3832 	return (unsigned long) address;
3833 #endif
3834 }
3835 
3836 /**
3837  *	pci_remap_iospace - Remap the memory mapped I/O space
3838  *	@res: Resource describing the I/O space
3839  *	@phys_addr: physical address of range to be mapped
3840  *
3841  *	Remap the memory mapped I/O space described by the @res
3842  *	and the CPU physical address @phys_addr into virtual address space.
3843  *	Only architectures that have memory mapped IO functions defined
3844  *	(and the PCI_IOBASE value defined) should call this function.
3845  */
3846 int pci_remap_iospace(const struct resource *res, phys_addr_t phys_addr)
3847 {
3848 #if defined(PCI_IOBASE) && defined(CONFIG_MMU)
3849 	unsigned long vaddr = (unsigned long)PCI_IOBASE + res->start;
3850 
3851 	if (!(res->flags & IORESOURCE_IO))
3852 		return -EINVAL;
3853 
3854 	if (res->end > IO_SPACE_LIMIT)
3855 		return -EINVAL;
3856 
3857 	return ioremap_page_range(vaddr, vaddr + resource_size(res), phys_addr,
3858 				  pgprot_device(PAGE_KERNEL));
3859 #else
3860 	/* this architecture does not have memory mapped I/O space,
3861 	   so this function should never be called */
3862 	WARN_ONCE(1, "This architecture does not support memory mapped I/O\n");
3863 	return -ENODEV;
3864 #endif
3865 }
3866 EXPORT_SYMBOL(pci_remap_iospace);
3867 
3868 /**
3869  *	pci_unmap_iospace - Unmap the memory mapped I/O space
3870  *	@res: resource to be unmapped
3871  *
3872  *	Unmap the CPU virtual address @res from virtual address space.
3873  *	Only architectures that have memory mapped IO functions defined
3874  *	(and the PCI_IOBASE value defined) should call this function.
3875  */
3876 void pci_unmap_iospace(struct resource *res)
3877 {
3878 #if defined(PCI_IOBASE) && defined(CONFIG_MMU)
3879 	unsigned long vaddr = (unsigned long)PCI_IOBASE + res->start;
3880 
3881 	unmap_kernel_range(vaddr, resource_size(res));
3882 #endif
3883 }
3884 EXPORT_SYMBOL(pci_unmap_iospace);
3885 
3886 static void devm_pci_unmap_iospace(struct device *dev, void *ptr)
3887 {
3888 	struct resource **res = ptr;
3889 
3890 	pci_unmap_iospace(*res);
3891 }
3892 
3893 /**
3894  * devm_pci_remap_iospace - Managed pci_remap_iospace()
3895  * @dev: Generic device to remap IO address for
3896  * @res: Resource describing the I/O space
3897  * @phys_addr: physical address of range to be mapped
3898  *
3899  * Managed pci_remap_iospace().  Map is automatically unmapped on driver
3900  * detach.
3901  */
3902 int devm_pci_remap_iospace(struct device *dev, const struct resource *res,
3903 			   phys_addr_t phys_addr)
3904 {
3905 	const struct resource **ptr;
3906 	int error;
3907 
3908 	ptr = devres_alloc(devm_pci_unmap_iospace, sizeof(*ptr), GFP_KERNEL);
3909 	if (!ptr)
3910 		return -ENOMEM;
3911 
3912 	error = pci_remap_iospace(res, phys_addr);
3913 	if (error) {
3914 		devres_free(ptr);
3915 	} else	{
3916 		*ptr = res;
3917 		devres_add(dev, ptr);
3918 	}
3919 
3920 	return error;
3921 }
3922 EXPORT_SYMBOL(devm_pci_remap_iospace);
3923 
3924 /**
3925  * devm_pci_remap_cfgspace - Managed pci_remap_cfgspace()
3926  * @dev: Generic device to remap IO address for
3927  * @offset: Resource address to map
3928  * @size: Size of map
3929  *
3930  * Managed pci_remap_cfgspace().  Map is automatically unmapped on driver
3931  * detach.
3932  */
3933 void __iomem *devm_pci_remap_cfgspace(struct device *dev,
3934 				      resource_size_t offset,
3935 				      resource_size_t size)
3936 {
3937 	void __iomem **ptr, *addr;
3938 
3939 	ptr = devres_alloc(devm_ioremap_release, sizeof(*ptr), GFP_KERNEL);
3940 	if (!ptr)
3941 		return NULL;
3942 
3943 	addr = pci_remap_cfgspace(offset, size);
3944 	if (addr) {
3945 		*ptr = addr;
3946 		devres_add(dev, ptr);
3947 	} else
3948 		devres_free(ptr);
3949 
3950 	return addr;
3951 }
3952 EXPORT_SYMBOL(devm_pci_remap_cfgspace);
3953 
3954 /**
3955  * devm_pci_remap_cfg_resource - check, request region and ioremap cfg resource
3956  * @dev: generic device to handle the resource for
3957  * @res: configuration space resource to be handled
3958  *
3959  * Checks that a resource is a valid memory region, requests the memory
3960  * region and ioremaps with pci_remap_cfgspace() API that ensures the
3961  * proper PCI configuration space memory attributes are guaranteed.
3962  *
3963  * All operations are managed and will be undone on driver detach.
3964  *
3965  * Returns a pointer to the remapped memory or an ERR_PTR() encoded error code
3966  * on failure. Usage example::
3967  *
3968  *	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
3969  *	base = devm_pci_remap_cfg_resource(&pdev->dev, res);
3970  *	if (IS_ERR(base))
3971  *		return PTR_ERR(base);
3972  */
3973 void __iomem *devm_pci_remap_cfg_resource(struct device *dev,
3974 					  struct resource *res)
3975 {
3976 	resource_size_t size;
3977 	const char *name;
3978 	void __iomem *dest_ptr;
3979 
3980 	BUG_ON(!dev);
3981 
3982 	if (!res || resource_type(res) != IORESOURCE_MEM) {
3983 		dev_err(dev, "invalid resource\n");
3984 		return IOMEM_ERR_PTR(-EINVAL);
3985 	}
3986 
3987 	size = resource_size(res);
3988 	name = res->name ?: dev_name(dev);
3989 
3990 	if (!devm_request_mem_region(dev, res->start, size, name)) {
3991 		dev_err(dev, "can't request region for resource %pR\n", res);
3992 		return IOMEM_ERR_PTR(-EBUSY);
3993 	}
3994 
3995 	dest_ptr = devm_pci_remap_cfgspace(dev, res->start, size);
3996 	if (!dest_ptr) {
3997 		dev_err(dev, "ioremap failed for resource %pR\n", res);
3998 		devm_release_mem_region(dev, res->start, size);
3999 		dest_ptr = IOMEM_ERR_PTR(-ENOMEM);
4000 	}
4001 
4002 	return dest_ptr;
4003 }
4004 EXPORT_SYMBOL(devm_pci_remap_cfg_resource);
4005 
4006 static void __pci_set_master(struct pci_dev *dev, bool enable)
4007 {
4008 	u16 old_cmd, cmd;
4009 
4010 	pci_read_config_word(dev, PCI_COMMAND, &old_cmd);
4011 	if (enable)
4012 		cmd = old_cmd | PCI_COMMAND_MASTER;
4013 	else
4014 		cmd = old_cmd & ~PCI_COMMAND_MASTER;
4015 	if (cmd != old_cmd) {
4016 		pci_dbg(dev, "%s bus mastering\n",
4017 			enable ? "enabling" : "disabling");
4018 		pci_write_config_word(dev, PCI_COMMAND, cmd);
4019 	}
4020 	dev->is_busmaster = enable;
4021 }
4022 
4023 /**
4024  * pcibios_setup - process "pci=" kernel boot arguments
4025  * @str: string used to pass in "pci=" kernel boot arguments
4026  *
4027  * Process kernel boot arguments.  This is the default implementation.
4028  * Architecture specific implementations can override this as necessary.
4029  */
4030 char * __weak __init pcibios_setup(char *str)
4031 {
4032 	return str;
4033 }
4034 
4035 /**
4036  * pcibios_set_master - enable PCI bus-mastering for device dev
4037  * @dev: the PCI device to enable
4038  *
4039  * Enables PCI bus-mastering for the device.  This is the default
4040  * implementation.  Architecture specific implementations can override
4041  * this if necessary.
4042  */
4043 void __weak pcibios_set_master(struct pci_dev *dev)
4044 {
4045 	u8 lat;
4046 
4047 	/* The latency timer doesn't apply to PCIe (either Type 0 or Type 1) */
4048 	if (pci_is_pcie(dev))
4049 		return;
4050 
4051 	pci_read_config_byte(dev, PCI_LATENCY_TIMER, &lat);
4052 	if (lat < 16)
4053 		lat = (64 <= pcibios_max_latency) ? 64 : pcibios_max_latency;
4054 	else if (lat > pcibios_max_latency)
4055 		lat = pcibios_max_latency;
4056 	else
4057 		return;
4058 
4059 	pci_write_config_byte(dev, PCI_LATENCY_TIMER, lat);
4060 }
4061 
4062 /**
4063  * pci_set_master - enables bus-mastering for device dev
4064  * @dev: the PCI device to enable
4065  *
4066  * Enables bus-mastering on the device and calls pcibios_set_master()
4067  * to do the needed arch specific settings.
4068  */
4069 void pci_set_master(struct pci_dev *dev)
4070 {
4071 	__pci_set_master(dev, true);
4072 	pcibios_set_master(dev);
4073 }
4074 EXPORT_SYMBOL(pci_set_master);
4075 
4076 /**
4077  * pci_clear_master - disables bus-mastering for device dev
4078  * @dev: the PCI device to disable
4079  */
4080 void pci_clear_master(struct pci_dev *dev)
4081 {
4082 	__pci_set_master(dev, false);
4083 }
4084 EXPORT_SYMBOL(pci_clear_master);
4085 
4086 /**
4087  * pci_set_cacheline_size - ensure the CACHE_LINE_SIZE register is programmed
4088  * @dev: the PCI device for which MWI is to be enabled
4089  *
4090  * Helper function for pci_set_mwi.
4091  * Originally copied from drivers/net/acenic.c.
4092  * Copyright 1998-2001 by Jes Sorensen, <jes@trained-monkey.org>.
4093  *
4094  * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
4095  */
4096 int pci_set_cacheline_size(struct pci_dev *dev)
4097 {
4098 	u8 cacheline_size;
4099 
4100 	if (!pci_cache_line_size)
4101 		return -EINVAL;
4102 
4103 	/* Validate current setting: the PCI_CACHE_LINE_SIZE must be
4104 	   equal to or multiple of the right value. */
4105 	pci_read_config_byte(dev, PCI_CACHE_LINE_SIZE, &cacheline_size);
4106 	if (cacheline_size >= pci_cache_line_size &&
4107 	    (cacheline_size % pci_cache_line_size) == 0)
4108 		return 0;
4109 
4110 	/* Write the correct value. */
4111 	pci_write_config_byte(dev, PCI_CACHE_LINE_SIZE, pci_cache_line_size);
4112 	/* Read it back. */
4113 	pci_read_config_byte(dev, PCI_CACHE_LINE_SIZE, &cacheline_size);
4114 	if (cacheline_size == pci_cache_line_size)
4115 		return 0;
4116 
4117 	pci_printk(KERN_DEBUG, dev, "cache line size of %d is not supported\n",
4118 		   pci_cache_line_size << 2);
4119 
4120 	return -EINVAL;
4121 }
4122 EXPORT_SYMBOL_GPL(pci_set_cacheline_size);
4123 
4124 /**
4125  * pci_set_mwi - enables memory-write-invalidate PCI transaction
4126  * @dev: the PCI device for which MWI is enabled
4127  *
4128  * Enables the Memory-Write-Invalidate transaction in %PCI_COMMAND.
4129  *
4130  * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
4131  */
4132 int pci_set_mwi(struct pci_dev *dev)
4133 {
4134 #ifdef PCI_DISABLE_MWI
4135 	return 0;
4136 #else
4137 	int rc;
4138 	u16 cmd;
4139 
4140 	rc = pci_set_cacheline_size(dev);
4141 	if (rc)
4142 		return rc;
4143 
4144 	pci_read_config_word(dev, PCI_COMMAND, &cmd);
4145 	if (!(cmd & PCI_COMMAND_INVALIDATE)) {
4146 		pci_dbg(dev, "enabling Mem-Wr-Inval\n");
4147 		cmd |= PCI_COMMAND_INVALIDATE;
4148 		pci_write_config_word(dev, PCI_COMMAND, cmd);
4149 	}
4150 	return 0;
4151 #endif
4152 }
4153 EXPORT_SYMBOL(pci_set_mwi);
4154 
4155 /**
4156  * pcim_set_mwi - a device-managed pci_set_mwi()
4157  * @dev: the PCI device for which MWI is enabled
4158  *
4159  * Managed pci_set_mwi().
4160  *
4161  * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
4162  */
4163 int pcim_set_mwi(struct pci_dev *dev)
4164 {
4165 	struct pci_devres *dr;
4166 
4167 	dr = find_pci_dr(dev);
4168 	if (!dr)
4169 		return -ENOMEM;
4170 
4171 	dr->mwi = 1;
4172 	return pci_set_mwi(dev);
4173 }
4174 EXPORT_SYMBOL(pcim_set_mwi);
4175 
4176 /**
4177  * pci_try_set_mwi - enables memory-write-invalidate PCI transaction
4178  * @dev: the PCI device for which MWI is enabled
4179  *
4180  * Enables the Memory-Write-Invalidate transaction in %PCI_COMMAND.
4181  * Callers are not required to check the return value.
4182  *
4183  * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
4184  */
4185 int pci_try_set_mwi(struct pci_dev *dev)
4186 {
4187 #ifdef PCI_DISABLE_MWI
4188 	return 0;
4189 #else
4190 	return pci_set_mwi(dev);
4191 #endif
4192 }
4193 EXPORT_SYMBOL(pci_try_set_mwi);
4194 
4195 /**
4196  * pci_clear_mwi - disables Memory-Write-Invalidate for device dev
4197  * @dev: the PCI device to disable
4198  *
4199  * Disables PCI Memory-Write-Invalidate transaction on the device
4200  */
4201 void pci_clear_mwi(struct pci_dev *dev)
4202 {
4203 #ifndef PCI_DISABLE_MWI
4204 	u16 cmd;
4205 
4206 	pci_read_config_word(dev, PCI_COMMAND, &cmd);
4207 	if (cmd & PCI_COMMAND_INVALIDATE) {
4208 		cmd &= ~PCI_COMMAND_INVALIDATE;
4209 		pci_write_config_word(dev, PCI_COMMAND, cmd);
4210 	}
4211 #endif
4212 }
4213 EXPORT_SYMBOL(pci_clear_mwi);
4214 
4215 /**
4216  * pci_intx - enables/disables PCI INTx for device dev
4217  * @pdev: the PCI device to operate on
4218  * @enable: boolean: whether to enable or disable PCI INTx
4219  *
4220  * Enables/disables PCI INTx for device dev
4221  */
4222 void pci_intx(struct pci_dev *pdev, int enable)
4223 {
4224 	u16 pci_command, new;
4225 
4226 	pci_read_config_word(pdev, PCI_COMMAND, &pci_command);
4227 
4228 	if (enable)
4229 		new = pci_command & ~PCI_COMMAND_INTX_DISABLE;
4230 	else
4231 		new = pci_command | PCI_COMMAND_INTX_DISABLE;
4232 
4233 	if (new != pci_command) {
4234 		struct pci_devres *dr;
4235 
4236 		pci_write_config_word(pdev, PCI_COMMAND, new);
4237 
4238 		dr = find_pci_dr(pdev);
4239 		if (dr && !dr->restore_intx) {
4240 			dr->restore_intx = 1;
4241 			dr->orig_intx = !enable;
4242 		}
4243 	}
4244 }
4245 EXPORT_SYMBOL_GPL(pci_intx);
4246 
4247 static bool pci_check_and_set_intx_mask(struct pci_dev *dev, bool mask)
4248 {
4249 	struct pci_bus *bus = dev->bus;
4250 	bool mask_updated = true;
4251 	u32 cmd_status_dword;
4252 	u16 origcmd, newcmd;
4253 	unsigned long flags;
4254 	bool irq_pending;
4255 
4256 	/*
4257 	 * We do a single dword read to retrieve both command and status.
4258 	 * Document assumptions that make this possible.
4259 	 */
4260 	BUILD_BUG_ON(PCI_COMMAND % 4);
4261 	BUILD_BUG_ON(PCI_COMMAND + 2 != PCI_STATUS);
4262 
4263 	raw_spin_lock_irqsave(&pci_lock, flags);
4264 
4265 	bus->ops->read(bus, dev->devfn, PCI_COMMAND, 4, &cmd_status_dword);
4266 
4267 	irq_pending = (cmd_status_dword >> 16) & PCI_STATUS_INTERRUPT;
4268 
4269 	/*
4270 	 * Check interrupt status register to see whether our device
4271 	 * triggered the interrupt (when masking) or the next IRQ is
4272 	 * already pending (when unmasking).
4273 	 */
4274 	if (mask != irq_pending) {
4275 		mask_updated = false;
4276 		goto done;
4277 	}
4278 
4279 	origcmd = cmd_status_dword;
4280 	newcmd = origcmd & ~PCI_COMMAND_INTX_DISABLE;
4281 	if (mask)
4282 		newcmd |= PCI_COMMAND_INTX_DISABLE;
4283 	if (newcmd != origcmd)
4284 		bus->ops->write(bus, dev->devfn, PCI_COMMAND, 2, newcmd);
4285 
4286 done:
4287 	raw_spin_unlock_irqrestore(&pci_lock, flags);
4288 
4289 	return mask_updated;
4290 }
4291 
4292 /**
4293  * pci_check_and_mask_intx - mask INTx on pending interrupt
4294  * @dev: the PCI device to operate on
4295  *
4296  * Check if the device dev has its INTx line asserted, mask it and
4297  * return true in that case. False is returned if no interrupt was
4298  * pending.
4299  */
4300 bool pci_check_and_mask_intx(struct pci_dev *dev)
4301 {
4302 	return pci_check_and_set_intx_mask(dev, true);
4303 }
4304 EXPORT_SYMBOL_GPL(pci_check_and_mask_intx);
4305 
4306 /**
4307  * pci_check_and_unmask_intx - unmask INTx if no interrupt is pending
4308  * @dev: the PCI device to operate on
4309  *
4310  * Check if the device dev has its INTx line asserted, unmask it if not
4311  * and return true. False is returned and the mask remains active if
4312  * there was still an interrupt pending.
4313  */
4314 bool pci_check_and_unmask_intx(struct pci_dev *dev)
4315 {
4316 	return pci_check_and_set_intx_mask(dev, false);
4317 }
4318 EXPORT_SYMBOL_GPL(pci_check_and_unmask_intx);
4319 
4320 /**
4321  * pci_wait_for_pending_transaction - waits for pending transaction
4322  * @dev: the PCI device to operate on
4323  *
4324  * Return 0 if transaction is pending 1 otherwise.
4325  */
4326 int pci_wait_for_pending_transaction(struct pci_dev *dev)
4327 {
4328 	if (!pci_is_pcie(dev))
4329 		return 1;
4330 
4331 	return pci_wait_for_pending(dev, pci_pcie_cap(dev) + PCI_EXP_DEVSTA,
4332 				    PCI_EXP_DEVSTA_TRPND);
4333 }
4334 EXPORT_SYMBOL(pci_wait_for_pending_transaction);
4335 
4336 static int pci_dev_wait(struct pci_dev *dev, char *reset_type, int timeout)
4337 {
4338 	int delay = 1;
4339 	u32 id;
4340 
4341 	/*
4342 	 * After reset, the device should not silently discard config
4343 	 * requests, but it may still indicate that it needs more time by
4344 	 * responding to them with CRS completions.  The Root Port will
4345 	 * generally synthesize ~0 data to complete the read (except when
4346 	 * CRS SV is enabled and the read was for the Vendor ID; in that
4347 	 * case it synthesizes 0x0001 data).
4348 	 *
4349 	 * Wait for the device to return a non-CRS completion.  Read the
4350 	 * Command register instead of Vendor ID so we don't have to
4351 	 * contend with the CRS SV value.
4352 	 */
4353 	pci_read_config_dword(dev, PCI_COMMAND, &id);
4354 	while (id == ~0) {
4355 		if (delay > timeout) {
4356 			pci_warn(dev, "not ready %dms after %s; giving up\n",
4357 				 delay - 1, reset_type);
4358 			return -ENOTTY;
4359 		}
4360 
4361 		if (delay > 1000)
4362 			pci_info(dev, "not ready %dms after %s; waiting\n",
4363 				 delay - 1, reset_type);
4364 
4365 		msleep(delay);
4366 		delay *= 2;
4367 		pci_read_config_dword(dev, PCI_COMMAND, &id);
4368 	}
4369 
4370 	if (delay > 1000)
4371 		pci_info(dev, "ready %dms after %s\n", delay - 1,
4372 			 reset_type);
4373 
4374 	return 0;
4375 }
4376 
4377 /**
4378  * pcie_has_flr - check if a device supports function level resets
4379  * @dev:	device to check
4380  *
4381  * Returns true if the device advertises support for PCIe function level
4382  * resets.
4383  */
4384 bool pcie_has_flr(struct pci_dev *dev)
4385 {
4386 	u32 cap;
4387 
4388 	if (dev->dev_flags & PCI_DEV_FLAGS_NO_FLR_RESET)
4389 		return false;
4390 
4391 	pcie_capability_read_dword(dev, PCI_EXP_DEVCAP, &cap);
4392 	return cap & PCI_EXP_DEVCAP_FLR;
4393 }
4394 EXPORT_SYMBOL_GPL(pcie_has_flr);
4395 
4396 /**
4397  * pcie_flr - initiate a PCIe function level reset
4398  * @dev:	device to reset
4399  *
4400  * Initiate a function level reset on @dev.  The caller should ensure the
4401  * device supports FLR before calling this function, e.g. by using the
4402  * pcie_has_flr() helper.
4403  */
4404 int pcie_flr(struct pci_dev *dev)
4405 {
4406 	if (!pci_wait_for_pending_transaction(dev))
4407 		pci_err(dev, "timed out waiting for pending transaction; performing function level reset anyway\n");
4408 
4409 	pcie_capability_set_word(dev, PCI_EXP_DEVCTL, PCI_EXP_DEVCTL_BCR_FLR);
4410 
4411 	if (dev->imm_ready)
4412 		return 0;
4413 
4414 	/*
4415 	 * Per PCIe r4.0, sec 6.6.2, a device must complete an FLR within
4416 	 * 100ms, but may silently discard requests while the FLR is in
4417 	 * progress.  Wait 100ms before trying to access the device.
4418 	 */
4419 	msleep(100);
4420 
4421 	return pci_dev_wait(dev, "FLR", PCIE_RESET_READY_POLL_MS);
4422 }
4423 EXPORT_SYMBOL_GPL(pcie_flr);
4424 
4425 static int pci_af_flr(struct pci_dev *dev, int probe)
4426 {
4427 	int pos;
4428 	u8 cap;
4429 
4430 	pos = pci_find_capability(dev, PCI_CAP_ID_AF);
4431 	if (!pos)
4432 		return -ENOTTY;
4433 
4434 	if (dev->dev_flags & PCI_DEV_FLAGS_NO_FLR_RESET)
4435 		return -ENOTTY;
4436 
4437 	pci_read_config_byte(dev, pos + PCI_AF_CAP, &cap);
4438 	if (!(cap & PCI_AF_CAP_TP) || !(cap & PCI_AF_CAP_FLR))
4439 		return -ENOTTY;
4440 
4441 	if (probe)
4442 		return 0;
4443 
4444 	/*
4445 	 * Wait for Transaction Pending bit to clear.  A word-aligned test
4446 	 * is used, so we use the conrol offset rather than status and shift
4447 	 * the test bit to match.
4448 	 */
4449 	if (!pci_wait_for_pending(dev, pos + PCI_AF_CTRL,
4450 				 PCI_AF_STATUS_TP << 8))
4451 		pci_err(dev, "timed out waiting for pending transaction; performing AF function level reset anyway\n");
4452 
4453 	pci_write_config_byte(dev, pos + PCI_AF_CTRL, PCI_AF_CTRL_FLR);
4454 
4455 	if (dev->imm_ready)
4456 		return 0;
4457 
4458 	/*
4459 	 * Per Advanced Capabilities for Conventional PCI ECN, 13 April 2006,
4460 	 * updated 27 July 2006; a device must complete an FLR within
4461 	 * 100ms, but may silently discard requests while the FLR is in
4462 	 * progress.  Wait 100ms before trying to access the device.
4463 	 */
4464 	msleep(100);
4465 
4466 	return pci_dev_wait(dev, "AF_FLR", PCIE_RESET_READY_POLL_MS);
4467 }
4468 
4469 /**
4470  * pci_pm_reset - Put device into PCI_D3 and back into PCI_D0.
4471  * @dev: Device to reset.
4472  * @probe: If set, only check if the device can be reset this way.
4473  *
4474  * If @dev supports native PCI PM and its PCI_PM_CTRL_NO_SOFT_RESET flag is
4475  * unset, it will be reinitialized internally when going from PCI_D3hot to
4476  * PCI_D0.  If that's the case and the device is not in a low-power state
4477  * already, force it into PCI_D3hot and back to PCI_D0, causing it to be reset.
4478  *
4479  * NOTE: This causes the caller to sleep for twice the device power transition
4480  * cooldown period, which for the D0->D3hot and D3hot->D0 transitions is 10 ms
4481  * by default (i.e. unless the @dev's d3_delay field has a different value).
4482  * Moreover, only devices in D0 can be reset by this function.
4483  */
4484 static int pci_pm_reset(struct pci_dev *dev, int probe)
4485 {
4486 	u16 csr;
4487 
4488 	if (!dev->pm_cap || dev->dev_flags & PCI_DEV_FLAGS_NO_PM_RESET)
4489 		return -ENOTTY;
4490 
4491 	pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &csr);
4492 	if (csr & PCI_PM_CTRL_NO_SOFT_RESET)
4493 		return -ENOTTY;
4494 
4495 	if (probe)
4496 		return 0;
4497 
4498 	if (dev->current_state != PCI_D0)
4499 		return -EINVAL;
4500 
4501 	csr &= ~PCI_PM_CTRL_STATE_MASK;
4502 	csr |= PCI_D3hot;
4503 	pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, csr);
4504 	pci_dev_d3_sleep(dev);
4505 
4506 	csr &= ~PCI_PM_CTRL_STATE_MASK;
4507 	csr |= PCI_D0;
4508 	pci_write_config_word(dev, dev->pm_cap + PCI_PM_CTRL, csr);
4509 	pci_dev_d3_sleep(dev);
4510 
4511 	return pci_dev_wait(dev, "PM D3->D0", PCIE_RESET_READY_POLL_MS);
4512 }
4513 /**
4514  * pcie_wait_for_link - Wait until link is active or inactive
4515  * @pdev: Bridge device
4516  * @active: waiting for active or inactive?
4517  *
4518  * Use this to wait till link becomes active or inactive.
4519  */
4520 bool pcie_wait_for_link(struct pci_dev *pdev, bool active)
4521 {
4522 	int timeout = 1000;
4523 	bool ret;
4524 	u16 lnk_status;
4525 
4526 	/*
4527 	 * Some controllers might not implement link active reporting. In this
4528 	 * case, we wait for 1000 + 100 ms.
4529 	 */
4530 	if (!pdev->link_active_reporting) {
4531 		msleep(1100);
4532 		return true;
4533 	}
4534 
4535 	/*
4536 	 * PCIe r4.0 sec 6.6.1, a component must enter LTSSM Detect within 20ms,
4537 	 * after which we should expect an link active if the reset was
4538 	 * successful. If so, software must wait a minimum 100ms before sending
4539 	 * configuration requests to devices downstream this port.
4540 	 *
4541 	 * If the link fails to activate, either the device was physically
4542 	 * removed or the link is permanently failed.
4543 	 */
4544 	if (active)
4545 		msleep(20);
4546 	for (;;) {
4547 		pcie_capability_read_word(pdev, PCI_EXP_LNKSTA, &lnk_status);
4548 		ret = !!(lnk_status & PCI_EXP_LNKSTA_DLLLA);
4549 		if (ret == active)
4550 			break;
4551 		if (timeout <= 0)
4552 			break;
4553 		msleep(10);
4554 		timeout -= 10;
4555 	}
4556 	if (active && ret)
4557 		msleep(100);
4558 	else if (ret != active)
4559 		pci_info(pdev, "Data Link Layer Link Active not %s in 1000 msec\n",
4560 			active ? "set" : "cleared");
4561 	return ret == active;
4562 }
4563 
4564 void pci_reset_secondary_bus(struct pci_dev *dev)
4565 {
4566 	u16 ctrl;
4567 
4568 	pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &ctrl);
4569 	ctrl |= PCI_BRIDGE_CTL_BUS_RESET;
4570 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL, ctrl);
4571 
4572 	/*
4573 	 * PCI spec v3.0 7.6.4.2 requires minimum Trst of 1ms.  Double
4574 	 * this to 2ms to ensure that we meet the minimum requirement.
4575 	 */
4576 	msleep(2);
4577 
4578 	ctrl &= ~PCI_BRIDGE_CTL_BUS_RESET;
4579 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL, ctrl);
4580 
4581 	/*
4582 	 * Trhfa for conventional PCI is 2^25 clock cycles.
4583 	 * Assuming a minimum 33MHz clock this results in a 1s
4584 	 * delay before we can consider subordinate devices to
4585 	 * be re-initialized.  PCIe has some ways to shorten this,
4586 	 * but we don't make use of them yet.
4587 	 */
4588 	ssleep(1);
4589 }
4590 
4591 void __weak pcibios_reset_secondary_bus(struct pci_dev *dev)
4592 {
4593 	pci_reset_secondary_bus(dev);
4594 }
4595 
4596 /**
4597  * pci_bridge_secondary_bus_reset - Reset the secondary bus on a PCI bridge.
4598  * @dev: Bridge device
4599  *
4600  * Use the bridge control register to assert reset on the secondary bus.
4601  * Devices on the secondary bus are left in power-on state.
4602  */
4603 int pci_bridge_secondary_bus_reset(struct pci_dev *dev)
4604 {
4605 	pcibios_reset_secondary_bus(dev);
4606 
4607 	return pci_dev_wait(dev, "bus reset", PCIE_RESET_READY_POLL_MS);
4608 }
4609 EXPORT_SYMBOL_GPL(pci_bridge_secondary_bus_reset);
4610 
4611 static int pci_parent_bus_reset(struct pci_dev *dev, int probe)
4612 {
4613 	struct pci_dev *pdev;
4614 
4615 	if (pci_is_root_bus(dev->bus) || dev->subordinate ||
4616 	    !dev->bus->self || dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET)
4617 		return -ENOTTY;
4618 
4619 	list_for_each_entry(pdev, &dev->bus->devices, bus_list)
4620 		if (pdev != dev)
4621 			return -ENOTTY;
4622 
4623 	if (probe)
4624 		return 0;
4625 
4626 	return pci_bridge_secondary_bus_reset(dev->bus->self);
4627 }
4628 
4629 static int pci_reset_hotplug_slot(struct hotplug_slot *hotplug, int probe)
4630 {
4631 	int rc = -ENOTTY;
4632 
4633 	if (!hotplug || !try_module_get(hotplug->owner))
4634 		return rc;
4635 
4636 	if (hotplug->ops->reset_slot)
4637 		rc = hotplug->ops->reset_slot(hotplug, probe);
4638 
4639 	module_put(hotplug->owner);
4640 
4641 	return rc;
4642 }
4643 
4644 static int pci_dev_reset_slot_function(struct pci_dev *dev, int probe)
4645 {
4646 	struct pci_dev *pdev;
4647 
4648 	if (dev->subordinate || !dev->slot ||
4649 	    dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET)
4650 		return -ENOTTY;
4651 
4652 	list_for_each_entry(pdev, &dev->bus->devices, bus_list)
4653 		if (pdev != dev && pdev->slot == dev->slot)
4654 			return -ENOTTY;
4655 
4656 	return pci_reset_hotplug_slot(dev->slot->hotplug, probe);
4657 }
4658 
4659 static void pci_dev_lock(struct pci_dev *dev)
4660 {
4661 	pci_cfg_access_lock(dev);
4662 	/* block PM suspend, driver probe, etc. */
4663 	device_lock(&dev->dev);
4664 }
4665 
4666 /* Return 1 on successful lock, 0 on contention */
4667 static int pci_dev_trylock(struct pci_dev *dev)
4668 {
4669 	if (pci_cfg_access_trylock(dev)) {
4670 		if (device_trylock(&dev->dev))
4671 			return 1;
4672 		pci_cfg_access_unlock(dev);
4673 	}
4674 
4675 	return 0;
4676 }
4677 
4678 static void pci_dev_unlock(struct pci_dev *dev)
4679 {
4680 	device_unlock(&dev->dev);
4681 	pci_cfg_access_unlock(dev);
4682 }
4683 
4684 static void pci_dev_save_and_disable(struct pci_dev *dev)
4685 {
4686 	const struct pci_error_handlers *err_handler =
4687 			dev->driver ? dev->driver->err_handler : NULL;
4688 
4689 	/*
4690 	 * dev->driver->err_handler->reset_prepare() is protected against
4691 	 * races with ->remove() by the device lock, which must be held by
4692 	 * the caller.
4693 	 */
4694 	if (err_handler && err_handler->reset_prepare)
4695 		err_handler->reset_prepare(dev);
4696 
4697 	/*
4698 	 * Wake-up device prior to save.  PM registers default to D0 after
4699 	 * reset and a simple register restore doesn't reliably return
4700 	 * to a non-D0 state anyway.
4701 	 */
4702 	pci_set_power_state(dev, PCI_D0);
4703 
4704 	pci_save_state(dev);
4705 	/*
4706 	 * Disable the device by clearing the Command register, except for
4707 	 * INTx-disable which is set.  This not only disables MMIO and I/O port
4708 	 * BARs, but also prevents the device from being Bus Master, preventing
4709 	 * DMA from the device including MSI/MSI-X interrupts.  For PCI 2.3
4710 	 * compliant devices, INTx-disable prevents legacy interrupts.
4711 	 */
4712 	pci_write_config_word(dev, PCI_COMMAND, PCI_COMMAND_INTX_DISABLE);
4713 }
4714 
4715 static void pci_dev_restore(struct pci_dev *dev)
4716 {
4717 	const struct pci_error_handlers *err_handler =
4718 			dev->driver ? dev->driver->err_handler : NULL;
4719 
4720 	pci_restore_state(dev);
4721 
4722 	/*
4723 	 * dev->driver->err_handler->reset_done() is protected against
4724 	 * races with ->remove() by the device lock, which must be held by
4725 	 * the caller.
4726 	 */
4727 	if (err_handler && err_handler->reset_done)
4728 		err_handler->reset_done(dev);
4729 }
4730 
4731 /**
4732  * __pci_reset_function_locked - reset a PCI device function while holding
4733  * the @dev mutex lock.
4734  * @dev: PCI device to reset
4735  *
4736  * Some devices allow an individual function to be reset without affecting
4737  * other functions in the same device.  The PCI device must be responsive
4738  * to PCI config space in order to use this function.
4739  *
4740  * The device function is presumed to be unused and the caller is holding
4741  * the device mutex lock when this function is called.
4742  * Resetting the device will make the contents of PCI configuration space
4743  * random, so any caller of this must be prepared to reinitialise the
4744  * device including MSI, bus mastering, BARs, decoding IO and memory spaces,
4745  * etc.
4746  *
4747  * Returns 0 if the device function was successfully reset or negative if the
4748  * device doesn't support resetting a single function.
4749  */
4750 int __pci_reset_function_locked(struct pci_dev *dev)
4751 {
4752 	int rc;
4753 
4754 	might_sleep();
4755 
4756 	/*
4757 	 * A reset method returns -ENOTTY if it doesn't support this device
4758 	 * and we should try the next method.
4759 	 *
4760 	 * If it returns 0 (success), we're finished.  If it returns any
4761 	 * other error, we're also finished: this indicates that further
4762 	 * reset mechanisms might be broken on the device.
4763 	 */
4764 	rc = pci_dev_specific_reset(dev, 0);
4765 	if (rc != -ENOTTY)
4766 		return rc;
4767 	if (pcie_has_flr(dev)) {
4768 		rc = pcie_flr(dev);
4769 		if (rc != -ENOTTY)
4770 			return rc;
4771 	}
4772 	rc = pci_af_flr(dev, 0);
4773 	if (rc != -ENOTTY)
4774 		return rc;
4775 	rc = pci_pm_reset(dev, 0);
4776 	if (rc != -ENOTTY)
4777 		return rc;
4778 	rc = pci_dev_reset_slot_function(dev, 0);
4779 	if (rc != -ENOTTY)
4780 		return rc;
4781 	return pci_parent_bus_reset(dev, 0);
4782 }
4783 EXPORT_SYMBOL_GPL(__pci_reset_function_locked);
4784 
4785 /**
4786  * pci_probe_reset_function - check whether the device can be safely reset
4787  * @dev: PCI device to reset
4788  *
4789  * Some devices allow an individual function to be reset without affecting
4790  * other functions in the same device.  The PCI device must be responsive
4791  * to PCI config space in order to use this function.
4792  *
4793  * Returns 0 if the device function can be reset or negative if the
4794  * device doesn't support resetting a single function.
4795  */
4796 int pci_probe_reset_function(struct pci_dev *dev)
4797 {
4798 	int rc;
4799 
4800 	might_sleep();
4801 
4802 	rc = pci_dev_specific_reset(dev, 1);
4803 	if (rc != -ENOTTY)
4804 		return rc;
4805 	if (pcie_has_flr(dev))
4806 		return 0;
4807 	rc = pci_af_flr(dev, 1);
4808 	if (rc != -ENOTTY)
4809 		return rc;
4810 	rc = pci_pm_reset(dev, 1);
4811 	if (rc != -ENOTTY)
4812 		return rc;
4813 	rc = pci_dev_reset_slot_function(dev, 1);
4814 	if (rc != -ENOTTY)
4815 		return rc;
4816 
4817 	return pci_parent_bus_reset(dev, 1);
4818 }
4819 
4820 /**
4821  * pci_reset_function - quiesce and reset a PCI device function
4822  * @dev: PCI device to reset
4823  *
4824  * Some devices allow an individual function to be reset without affecting
4825  * other functions in the same device.  The PCI device must be responsive
4826  * to PCI config space in order to use this function.
4827  *
4828  * This function does not just reset the PCI portion of a device, but
4829  * clears all the state associated with the device.  This function differs
4830  * from __pci_reset_function_locked() in that it saves and restores device state
4831  * over the reset and takes the PCI device lock.
4832  *
4833  * Returns 0 if the device function was successfully reset or negative if the
4834  * device doesn't support resetting a single function.
4835  */
4836 int pci_reset_function(struct pci_dev *dev)
4837 {
4838 	int rc;
4839 
4840 	if (!dev->reset_fn)
4841 		return -ENOTTY;
4842 
4843 	pci_dev_lock(dev);
4844 	pci_dev_save_and_disable(dev);
4845 
4846 	rc = __pci_reset_function_locked(dev);
4847 
4848 	pci_dev_restore(dev);
4849 	pci_dev_unlock(dev);
4850 
4851 	return rc;
4852 }
4853 EXPORT_SYMBOL_GPL(pci_reset_function);
4854 
4855 /**
4856  * pci_reset_function_locked - quiesce and reset a PCI device function
4857  * @dev: PCI device to reset
4858  *
4859  * Some devices allow an individual function to be reset without affecting
4860  * other functions in the same device.  The PCI device must be responsive
4861  * to PCI config space in order to use this function.
4862  *
4863  * This function does not just reset the PCI portion of a device, but
4864  * clears all the state associated with the device.  This function differs
4865  * from __pci_reset_function_locked() in that it saves and restores device state
4866  * over the reset.  It also differs from pci_reset_function() in that it
4867  * requires the PCI device lock to be held.
4868  *
4869  * Returns 0 if the device function was successfully reset or negative if the
4870  * device doesn't support resetting a single function.
4871  */
4872 int pci_reset_function_locked(struct pci_dev *dev)
4873 {
4874 	int rc;
4875 
4876 	if (!dev->reset_fn)
4877 		return -ENOTTY;
4878 
4879 	pci_dev_save_and_disable(dev);
4880 
4881 	rc = __pci_reset_function_locked(dev);
4882 
4883 	pci_dev_restore(dev);
4884 
4885 	return rc;
4886 }
4887 EXPORT_SYMBOL_GPL(pci_reset_function_locked);
4888 
4889 /**
4890  * pci_try_reset_function - quiesce and reset a PCI device function
4891  * @dev: PCI device to reset
4892  *
4893  * Same as above, except return -EAGAIN if unable to lock device.
4894  */
4895 int pci_try_reset_function(struct pci_dev *dev)
4896 {
4897 	int rc;
4898 
4899 	if (!dev->reset_fn)
4900 		return -ENOTTY;
4901 
4902 	if (!pci_dev_trylock(dev))
4903 		return -EAGAIN;
4904 
4905 	pci_dev_save_and_disable(dev);
4906 	rc = __pci_reset_function_locked(dev);
4907 	pci_dev_restore(dev);
4908 	pci_dev_unlock(dev);
4909 
4910 	return rc;
4911 }
4912 EXPORT_SYMBOL_GPL(pci_try_reset_function);
4913 
4914 /* Do any devices on or below this bus prevent a bus reset? */
4915 static bool pci_bus_resetable(struct pci_bus *bus)
4916 {
4917 	struct pci_dev *dev;
4918 
4919 
4920 	if (bus->self && (bus->self->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET))
4921 		return false;
4922 
4923 	list_for_each_entry(dev, &bus->devices, bus_list) {
4924 		if (dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET ||
4925 		    (dev->subordinate && !pci_bus_resetable(dev->subordinate)))
4926 			return false;
4927 	}
4928 
4929 	return true;
4930 }
4931 
4932 /* Lock devices from the top of the tree down */
4933 static void pci_bus_lock(struct pci_bus *bus)
4934 {
4935 	struct pci_dev *dev;
4936 
4937 	list_for_each_entry(dev, &bus->devices, bus_list) {
4938 		pci_dev_lock(dev);
4939 		if (dev->subordinate)
4940 			pci_bus_lock(dev->subordinate);
4941 	}
4942 }
4943 
4944 /* Unlock devices from the bottom of the tree up */
4945 static void pci_bus_unlock(struct pci_bus *bus)
4946 {
4947 	struct pci_dev *dev;
4948 
4949 	list_for_each_entry(dev, &bus->devices, bus_list) {
4950 		if (dev->subordinate)
4951 			pci_bus_unlock(dev->subordinate);
4952 		pci_dev_unlock(dev);
4953 	}
4954 }
4955 
4956 /* Return 1 on successful lock, 0 on contention */
4957 static int pci_bus_trylock(struct pci_bus *bus)
4958 {
4959 	struct pci_dev *dev;
4960 
4961 	list_for_each_entry(dev, &bus->devices, bus_list) {
4962 		if (!pci_dev_trylock(dev))
4963 			goto unlock;
4964 		if (dev->subordinate) {
4965 			if (!pci_bus_trylock(dev->subordinate)) {
4966 				pci_dev_unlock(dev);
4967 				goto unlock;
4968 			}
4969 		}
4970 	}
4971 	return 1;
4972 
4973 unlock:
4974 	list_for_each_entry_continue_reverse(dev, &bus->devices, bus_list) {
4975 		if (dev->subordinate)
4976 			pci_bus_unlock(dev->subordinate);
4977 		pci_dev_unlock(dev);
4978 	}
4979 	return 0;
4980 }
4981 
4982 /* Do any devices on or below this slot prevent a bus reset? */
4983 static bool pci_slot_resetable(struct pci_slot *slot)
4984 {
4985 	struct pci_dev *dev;
4986 
4987 	if (slot->bus->self &&
4988 	    (slot->bus->self->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET))
4989 		return false;
4990 
4991 	list_for_each_entry(dev, &slot->bus->devices, bus_list) {
4992 		if (!dev->slot || dev->slot != slot)
4993 			continue;
4994 		if (dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET ||
4995 		    (dev->subordinate && !pci_bus_resetable(dev->subordinate)))
4996 			return false;
4997 	}
4998 
4999 	return true;
5000 }
5001 
5002 /* Lock devices from the top of the tree down */
5003 static void pci_slot_lock(struct pci_slot *slot)
5004 {
5005 	struct pci_dev *dev;
5006 
5007 	list_for_each_entry(dev, &slot->bus->devices, bus_list) {
5008 		if (!dev->slot || dev->slot != slot)
5009 			continue;
5010 		pci_dev_lock(dev);
5011 		if (dev->subordinate)
5012 			pci_bus_lock(dev->subordinate);
5013 	}
5014 }
5015 
5016 /* Unlock devices from the bottom of the tree up */
5017 static void pci_slot_unlock(struct pci_slot *slot)
5018 {
5019 	struct pci_dev *dev;
5020 
5021 	list_for_each_entry(dev, &slot->bus->devices, bus_list) {
5022 		if (!dev->slot || dev->slot != slot)
5023 			continue;
5024 		if (dev->subordinate)
5025 			pci_bus_unlock(dev->subordinate);
5026 		pci_dev_unlock(dev);
5027 	}
5028 }
5029 
5030 /* Return 1 on successful lock, 0 on contention */
5031 static int pci_slot_trylock(struct pci_slot *slot)
5032 {
5033 	struct pci_dev *dev;
5034 
5035 	list_for_each_entry(dev, &slot->bus->devices, bus_list) {
5036 		if (!dev->slot || dev->slot != slot)
5037 			continue;
5038 		if (!pci_dev_trylock(dev))
5039 			goto unlock;
5040 		if (dev->subordinate) {
5041 			if (!pci_bus_trylock(dev->subordinate)) {
5042 				pci_dev_unlock(dev);
5043 				goto unlock;
5044 			}
5045 		}
5046 	}
5047 	return 1;
5048 
5049 unlock:
5050 	list_for_each_entry_continue_reverse(dev,
5051 					     &slot->bus->devices, bus_list) {
5052 		if (!dev->slot || dev->slot != slot)
5053 			continue;
5054 		if (dev->subordinate)
5055 			pci_bus_unlock(dev->subordinate);
5056 		pci_dev_unlock(dev);
5057 	}
5058 	return 0;
5059 }
5060 
5061 /* Save and disable devices from the top of the tree down */
5062 static void pci_bus_save_and_disable(struct pci_bus *bus)
5063 {
5064 	struct pci_dev *dev;
5065 
5066 	list_for_each_entry(dev, &bus->devices, bus_list) {
5067 		pci_dev_lock(dev);
5068 		pci_dev_save_and_disable(dev);
5069 		pci_dev_unlock(dev);
5070 		if (dev->subordinate)
5071 			pci_bus_save_and_disable(dev->subordinate);
5072 	}
5073 }
5074 
5075 /*
5076  * Restore devices from top of the tree down - parent bridges need to be
5077  * restored before we can get to subordinate devices.
5078  */
5079 static void pci_bus_restore(struct pci_bus *bus)
5080 {
5081 	struct pci_dev *dev;
5082 
5083 	list_for_each_entry(dev, &bus->devices, bus_list) {
5084 		pci_dev_lock(dev);
5085 		pci_dev_restore(dev);
5086 		pci_dev_unlock(dev);
5087 		if (dev->subordinate)
5088 			pci_bus_restore(dev->subordinate);
5089 	}
5090 }
5091 
5092 /* Save and disable devices from the top of the tree down */
5093 static void pci_slot_save_and_disable(struct pci_slot *slot)
5094 {
5095 	struct pci_dev *dev;
5096 
5097 	list_for_each_entry(dev, &slot->bus->devices, bus_list) {
5098 		if (!dev->slot || dev->slot != slot)
5099 			continue;
5100 		pci_dev_save_and_disable(dev);
5101 		if (dev->subordinate)
5102 			pci_bus_save_and_disable(dev->subordinate);
5103 	}
5104 }
5105 
5106 /*
5107  * Restore devices from top of the tree down - parent bridges need to be
5108  * restored before we can get to subordinate devices.
5109  */
5110 static void pci_slot_restore(struct pci_slot *slot)
5111 {
5112 	struct pci_dev *dev;
5113 
5114 	list_for_each_entry(dev, &slot->bus->devices, bus_list) {
5115 		if (!dev->slot || dev->slot != slot)
5116 			continue;
5117 		pci_dev_lock(dev);
5118 		pci_dev_restore(dev);
5119 		pci_dev_unlock(dev);
5120 		if (dev->subordinate)
5121 			pci_bus_restore(dev->subordinate);
5122 	}
5123 }
5124 
5125 static int pci_slot_reset(struct pci_slot *slot, int probe)
5126 {
5127 	int rc;
5128 
5129 	if (!slot || !pci_slot_resetable(slot))
5130 		return -ENOTTY;
5131 
5132 	if (!probe)
5133 		pci_slot_lock(slot);
5134 
5135 	might_sleep();
5136 
5137 	rc = pci_reset_hotplug_slot(slot->hotplug, probe);
5138 
5139 	if (!probe)
5140 		pci_slot_unlock(slot);
5141 
5142 	return rc;
5143 }
5144 
5145 /**
5146  * pci_probe_reset_slot - probe whether a PCI slot can be reset
5147  * @slot: PCI slot to probe
5148  *
5149  * Return 0 if slot can be reset, negative if a slot reset is not supported.
5150  */
5151 int pci_probe_reset_slot(struct pci_slot *slot)
5152 {
5153 	return pci_slot_reset(slot, 1);
5154 }
5155 EXPORT_SYMBOL_GPL(pci_probe_reset_slot);
5156 
5157 /**
5158  * __pci_reset_slot - Try to reset a PCI slot
5159  * @slot: PCI slot to reset
5160  *
5161  * A PCI bus may host multiple slots, each slot may support a reset mechanism
5162  * independent of other slots.  For instance, some slots may support slot power
5163  * control.  In the case of a 1:1 bus to slot architecture, this function may
5164  * wrap the bus reset to avoid spurious slot related events such as hotplug.
5165  * Generally a slot reset should be attempted before a bus reset.  All of the
5166  * function of the slot and any subordinate buses behind the slot are reset
5167  * through this function.  PCI config space of all devices in the slot and
5168  * behind the slot is saved before and restored after reset.
5169  *
5170  * Same as above except return -EAGAIN if the slot cannot be locked
5171  */
5172 static int __pci_reset_slot(struct pci_slot *slot)
5173 {
5174 	int rc;
5175 
5176 	rc = pci_slot_reset(slot, 1);
5177 	if (rc)
5178 		return rc;
5179 
5180 	pci_slot_save_and_disable(slot);
5181 
5182 	if (pci_slot_trylock(slot)) {
5183 		might_sleep();
5184 		rc = pci_reset_hotplug_slot(slot->hotplug, 0);
5185 		pci_slot_unlock(slot);
5186 	} else
5187 		rc = -EAGAIN;
5188 
5189 	pci_slot_restore(slot);
5190 
5191 	return rc;
5192 }
5193 
5194 static int pci_bus_reset(struct pci_bus *bus, int probe)
5195 {
5196 	int ret;
5197 
5198 	if (!bus->self || !pci_bus_resetable(bus))
5199 		return -ENOTTY;
5200 
5201 	if (probe)
5202 		return 0;
5203 
5204 	pci_bus_lock(bus);
5205 
5206 	might_sleep();
5207 
5208 	ret = pci_bridge_secondary_bus_reset(bus->self);
5209 
5210 	pci_bus_unlock(bus);
5211 
5212 	return ret;
5213 }
5214 
5215 /**
5216  * pci_bus_error_reset - reset the bridge's subordinate bus
5217  * @bridge: The parent device that connects to the bus to reset
5218  *
5219  * This function will first try to reset the slots on this bus if the method is
5220  * available. If slot reset fails or is not available, this will fall back to a
5221  * secondary bus reset.
5222  */
5223 int pci_bus_error_reset(struct pci_dev *bridge)
5224 {
5225 	struct pci_bus *bus = bridge->subordinate;
5226 	struct pci_slot *slot;
5227 
5228 	if (!bus)
5229 		return -ENOTTY;
5230 
5231 	mutex_lock(&pci_slot_mutex);
5232 	if (list_empty(&bus->slots))
5233 		goto bus_reset;
5234 
5235 	list_for_each_entry(slot, &bus->slots, list)
5236 		if (pci_probe_reset_slot(slot))
5237 			goto bus_reset;
5238 
5239 	list_for_each_entry(slot, &bus->slots, list)
5240 		if (pci_slot_reset(slot, 0))
5241 			goto bus_reset;
5242 
5243 	mutex_unlock(&pci_slot_mutex);
5244 	return 0;
5245 bus_reset:
5246 	mutex_unlock(&pci_slot_mutex);
5247 	return pci_bus_reset(bridge->subordinate, 0);
5248 }
5249 
5250 /**
5251  * pci_probe_reset_bus - probe whether a PCI bus can be reset
5252  * @bus: PCI bus to probe
5253  *
5254  * Return 0 if bus can be reset, negative if a bus reset is not supported.
5255  */
5256 int pci_probe_reset_bus(struct pci_bus *bus)
5257 {
5258 	return pci_bus_reset(bus, 1);
5259 }
5260 EXPORT_SYMBOL_GPL(pci_probe_reset_bus);
5261 
5262 /**
5263  * __pci_reset_bus - Try to reset a PCI bus
5264  * @bus: top level PCI bus to reset
5265  *
5266  * Same as above except return -EAGAIN if the bus cannot be locked
5267  */
5268 static int __pci_reset_bus(struct pci_bus *bus)
5269 {
5270 	int rc;
5271 
5272 	rc = pci_bus_reset(bus, 1);
5273 	if (rc)
5274 		return rc;
5275 
5276 	pci_bus_save_and_disable(bus);
5277 
5278 	if (pci_bus_trylock(bus)) {
5279 		might_sleep();
5280 		rc = pci_bridge_secondary_bus_reset(bus->self);
5281 		pci_bus_unlock(bus);
5282 	} else
5283 		rc = -EAGAIN;
5284 
5285 	pci_bus_restore(bus);
5286 
5287 	return rc;
5288 }
5289 
5290 /**
5291  * pci_reset_bus - Try to reset a PCI bus
5292  * @pdev: top level PCI device to reset via slot/bus
5293  *
5294  * Same as above except return -EAGAIN if the bus cannot be locked
5295  */
5296 int pci_reset_bus(struct pci_dev *pdev)
5297 {
5298 	return (!pci_probe_reset_slot(pdev->slot)) ?
5299 	    __pci_reset_slot(pdev->slot) : __pci_reset_bus(pdev->bus);
5300 }
5301 EXPORT_SYMBOL_GPL(pci_reset_bus);
5302 
5303 /**
5304  * pcix_get_max_mmrbc - get PCI-X maximum designed memory read byte count
5305  * @dev: PCI device to query
5306  *
5307  * Returns mmrbc: maximum designed memory read count in bytes
5308  *    or appropriate error value.
5309  */
5310 int pcix_get_max_mmrbc(struct pci_dev *dev)
5311 {
5312 	int cap;
5313 	u32 stat;
5314 
5315 	cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
5316 	if (!cap)
5317 		return -EINVAL;
5318 
5319 	if (pci_read_config_dword(dev, cap + PCI_X_STATUS, &stat))
5320 		return -EINVAL;
5321 
5322 	return 512 << ((stat & PCI_X_STATUS_MAX_READ) >> 21);
5323 }
5324 EXPORT_SYMBOL(pcix_get_max_mmrbc);
5325 
5326 /**
5327  * pcix_get_mmrbc - get PCI-X maximum memory read byte count
5328  * @dev: PCI device to query
5329  *
5330  * Returns mmrbc: maximum memory read count in bytes
5331  *    or appropriate error value.
5332  */
5333 int pcix_get_mmrbc(struct pci_dev *dev)
5334 {
5335 	int cap;
5336 	u16 cmd;
5337 
5338 	cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
5339 	if (!cap)
5340 		return -EINVAL;
5341 
5342 	if (pci_read_config_word(dev, cap + PCI_X_CMD, &cmd))
5343 		return -EINVAL;
5344 
5345 	return 512 << ((cmd & PCI_X_CMD_MAX_READ) >> 2);
5346 }
5347 EXPORT_SYMBOL(pcix_get_mmrbc);
5348 
5349 /**
5350  * pcix_set_mmrbc - set PCI-X maximum memory read byte count
5351  * @dev: PCI device to query
5352  * @mmrbc: maximum memory read count in bytes
5353  *    valid values are 512, 1024, 2048, 4096
5354  *
5355  * If possible sets maximum memory read byte count, some bridges have erratas
5356  * that prevent this.
5357  */
5358 int pcix_set_mmrbc(struct pci_dev *dev, int mmrbc)
5359 {
5360 	int cap;
5361 	u32 stat, v, o;
5362 	u16 cmd;
5363 
5364 	if (mmrbc < 512 || mmrbc > 4096 || !is_power_of_2(mmrbc))
5365 		return -EINVAL;
5366 
5367 	v = ffs(mmrbc) - 10;
5368 
5369 	cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
5370 	if (!cap)
5371 		return -EINVAL;
5372 
5373 	if (pci_read_config_dword(dev, cap + PCI_X_STATUS, &stat))
5374 		return -EINVAL;
5375 
5376 	if (v > (stat & PCI_X_STATUS_MAX_READ) >> 21)
5377 		return -E2BIG;
5378 
5379 	if (pci_read_config_word(dev, cap + PCI_X_CMD, &cmd))
5380 		return -EINVAL;
5381 
5382 	o = (cmd & PCI_X_CMD_MAX_READ) >> 2;
5383 	if (o != v) {
5384 		if (v > o && (dev->bus->bus_flags & PCI_BUS_FLAGS_NO_MMRBC))
5385 			return -EIO;
5386 
5387 		cmd &= ~PCI_X_CMD_MAX_READ;
5388 		cmd |= v << 2;
5389 		if (pci_write_config_word(dev, cap + PCI_X_CMD, cmd))
5390 			return -EIO;
5391 	}
5392 	return 0;
5393 }
5394 EXPORT_SYMBOL(pcix_set_mmrbc);
5395 
5396 /**
5397  * pcie_get_readrq - get PCI Express read request size
5398  * @dev: PCI device to query
5399  *
5400  * Returns maximum memory read request in bytes
5401  *    or appropriate error value.
5402  */
5403 int pcie_get_readrq(struct pci_dev *dev)
5404 {
5405 	u16 ctl;
5406 
5407 	pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &ctl);
5408 
5409 	return 128 << ((ctl & PCI_EXP_DEVCTL_READRQ) >> 12);
5410 }
5411 EXPORT_SYMBOL(pcie_get_readrq);
5412 
5413 /**
5414  * pcie_set_readrq - set PCI Express maximum memory read request
5415  * @dev: PCI device to query
5416  * @rq: maximum memory read count in bytes
5417  *    valid values are 128, 256, 512, 1024, 2048, 4096
5418  *
5419  * If possible sets maximum memory read request in bytes
5420  */
5421 int pcie_set_readrq(struct pci_dev *dev, int rq)
5422 {
5423 	u16 v;
5424 
5425 	if (rq < 128 || rq > 4096 || !is_power_of_2(rq))
5426 		return -EINVAL;
5427 
5428 	/*
5429 	 * If using the "performance" PCIe config, we clamp the
5430 	 * read rq size to the max packet size to prevent the
5431 	 * host bridge generating requests larger than we can
5432 	 * cope with
5433 	 */
5434 	if (pcie_bus_config == PCIE_BUS_PERFORMANCE) {
5435 		int mps = pcie_get_mps(dev);
5436 
5437 		if (mps < rq)
5438 			rq = mps;
5439 	}
5440 
5441 	v = (ffs(rq) - 8) << 12;
5442 
5443 	return pcie_capability_clear_and_set_word(dev, PCI_EXP_DEVCTL,
5444 						  PCI_EXP_DEVCTL_READRQ, v);
5445 }
5446 EXPORT_SYMBOL(pcie_set_readrq);
5447 
5448 /**
5449  * pcie_get_mps - get PCI Express maximum payload size
5450  * @dev: PCI device to query
5451  *
5452  * Returns maximum payload size in bytes
5453  */
5454 int pcie_get_mps(struct pci_dev *dev)
5455 {
5456 	u16 ctl;
5457 
5458 	pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &ctl);
5459 
5460 	return 128 << ((ctl & PCI_EXP_DEVCTL_PAYLOAD) >> 5);
5461 }
5462 EXPORT_SYMBOL(pcie_get_mps);
5463 
5464 /**
5465  * pcie_set_mps - set PCI Express maximum payload size
5466  * @dev: PCI device to query
5467  * @mps: maximum payload size in bytes
5468  *    valid values are 128, 256, 512, 1024, 2048, 4096
5469  *
5470  * If possible sets maximum payload size
5471  */
5472 int pcie_set_mps(struct pci_dev *dev, int mps)
5473 {
5474 	u16 v;
5475 
5476 	if (mps < 128 || mps > 4096 || !is_power_of_2(mps))
5477 		return -EINVAL;
5478 
5479 	v = ffs(mps) - 8;
5480 	if (v > dev->pcie_mpss)
5481 		return -EINVAL;
5482 	v <<= 5;
5483 
5484 	return pcie_capability_clear_and_set_word(dev, PCI_EXP_DEVCTL,
5485 						  PCI_EXP_DEVCTL_PAYLOAD, v);
5486 }
5487 EXPORT_SYMBOL(pcie_set_mps);
5488 
5489 /**
5490  * pcie_bandwidth_available - determine minimum link settings of a PCIe
5491  *			      device and its bandwidth limitation
5492  * @dev: PCI device to query
5493  * @limiting_dev: storage for device causing the bandwidth limitation
5494  * @speed: storage for speed of limiting device
5495  * @width: storage for width of limiting device
5496  *
5497  * Walk up the PCI device chain and find the point where the minimum
5498  * bandwidth is available.  Return the bandwidth available there and (if
5499  * limiting_dev, speed, and width pointers are supplied) information about
5500  * that point.  The bandwidth returned is in Mb/s, i.e., megabits/second of
5501  * raw bandwidth.
5502  */
5503 u32 pcie_bandwidth_available(struct pci_dev *dev, struct pci_dev **limiting_dev,
5504 			     enum pci_bus_speed *speed,
5505 			     enum pcie_link_width *width)
5506 {
5507 	u16 lnksta;
5508 	enum pci_bus_speed next_speed;
5509 	enum pcie_link_width next_width;
5510 	u32 bw, next_bw;
5511 
5512 	if (speed)
5513 		*speed = PCI_SPEED_UNKNOWN;
5514 	if (width)
5515 		*width = PCIE_LNK_WIDTH_UNKNOWN;
5516 
5517 	bw = 0;
5518 
5519 	while (dev) {
5520 		pcie_capability_read_word(dev, PCI_EXP_LNKSTA, &lnksta);
5521 
5522 		next_speed = pcie_link_speed[lnksta & PCI_EXP_LNKSTA_CLS];
5523 		next_width = (lnksta & PCI_EXP_LNKSTA_NLW) >>
5524 			PCI_EXP_LNKSTA_NLW_SHIFT;
5525 
5526 		next_bw = next_width * PCIE_SPEED2MBS_ENC(next_speed);
5527 
5528 		/* Check if current device limits the total bandwidth */
5529 		if (!bw || next_bw <= bw) {
5530 			bw = next_bw;
5531 
5532 			if (limiting_dev)
5533 				*limiting_dev = dev;
5534 			if (speed)
5535 				*speed = next_speed;
5536 			if (width)
5537 				*width = next_width;
5538 		}
5539 
5540 		dev = pci_upstream_bridge(dev);
5541 	}
5542 
5543 	return bw;
5544 }
5545 EXPORT_SYMBOL(pcie_bandwidth_available);
5546 
5547 /**
5548  * pcie_get_speed_cap - query for the PCI device's link speed capability
5549  * @dev: PCI device to query
5550  *
5551  * Query the PCI device speed capability.  Return the maximum link speed
5552  * supported by the device.
5553  */
5554 enum pci_bus_speed pcie_get_speed_cap(struct pci_dev *dev)
5555 {
5556 	u32 lnkcap2, lnkcap;
5557 
5558 	/*
5559 	 * Link Capabilities 2 was added in PCIe r3.0, sec 7.8.18.  The
5560 	 * implementation note there recommends using the Supported Link
5561 	 * Speeds Vector in Link Capabilities 2 when supported.
5562 	 *
5563 	 * Without Link Capabilities 2, i.e., prior to PCIe r3.0, software
5564 	 * should use the Supported Link Speeds field in Link Capabilities,
5565 	 * where only 2.5 GT/s and 5.0 GT/s speeds were defined.
5566 	 */
5567 	pcie_capability_read_dword(dev, PCI_EXP_LNKCAP2, &lnkcap2);
5568 	if (lnkcap2) { /* PCIe r3.0-compliant */
5569 		if (lnkcap2 & PCI_EXP_LNKCAP2_SLS_16_0GB)
5570 			return PCIE_SPEED_16_0GT;
5571 		else if (lnkcap2 & PCI_EXP_LNKCAP2_SLS_8_0GB)
5572 			return PCIE_SPEED_8_0GT;
5573 		else if (lnkcap2 & PCI_EXP_LNKCAP2_SLS_5_0GB)
5574 			return PCIE_SPEED_5_0GT;
5575 		else if (lnkcap2 & PCI_EXP_LNKCAP2_SLS_2_5GB)
5576 			return PCIE_SPEED_2_5GT;
5577 		return PCI_SPEED_UNKNOWN;
5578 	}
5579 
5580 	pcie_capability_read_dword(dev, PCI_EXP_LNKCAP, &lnkcap);
5581 	if ((lnkcap & PCI_EXP_LNKCAP_SLS) == PCI_EXP_LNKCAP_SLS_5_0GB)
5582 		return PCIE_SPEED_5_0GT;
5583 	else if ((lnkcap & PCI_EXP_LNKCAP_SLS) == PCI_EXP_LNKCAP_SLS_2_5GB)
5584 		return PCIE_SPEED_2_5GT;
5585 
5586 	return PCI_SPEED_UNKNOWN;
5587 }
5588 EXPORT_SYMBOL(pcie_get_speed_cap);
5589 
5590 /**
5591  * pcie_get_width_cap - query for the PCI device's link width capability
5592  * @dev: PCI device to query
5593  *
5594  * Query the PCI device width capability.  Return the maximum link width
5595  * supported by the device.
5596  */
5597 enum pcie_link_width pcie_get_width_cap(struct pci_dev *dev)
5598 {
5599 	u32 lnkcap;
5600 
5601 	pcie_capability_read_dword(dev, PCI_EXP_LNKCAP, &lnkcap);
5602 	if (lnkcap)
5603 		return (lnkcap & PCI_EXP_LNKCAP_MLW) >> 4;
5604 
5605 	return PCIE_LNK_WIDTH_UNKNOWN;
5606 }
5607 EXPORT_SYMBOL(pcie_get_width_cap);
5608 
5609 /**
5610  * pcie_bandwidth_capable - calculate a PCI device's link bandwidth capability
5611  * @dev: PCI device
5612  * @speed: storage for link speed
5613  * @width: storage for link width
5614  *
5615  * Calculate a PCI device's link bandwidth by querying for its link speed
5616  * and width, multiplying them, and applying encoding overhead.  The result
5617  * is in Mb/s, i.e., megabits/second of raw bandwidth.
5618  */
5619 u32 pcie_bandwidth_capable(struct pci_dev *dev, enum pci_bus_speed *speed,
5620 			   enum pcie_link_width *width)
5621 {
5622 	*speed = pcie_get_speed_cap(dev);
5623 	*width = pcie_get_width_cap(dev);
5624 
5625 	if (*speed == PCI_SPEED_UNKNOWN || *width == PCIE_LNK_WIDTH_UNKNOWN)
5626 		return 0;
5627 
5628 	return *width * PCIE_SPEED2MBS_ENC(*speed);
5629 }
5630 
5631 /**
5632  * __pcie_print_link_status - Report the PCI device's link speed and width
5633  * @dev: PCI device to query
5634  * @verbose: Print info even when enough bandwidth is available
5635  *
5636  * If the available bandwidth at the device is less than the device is
5637  * capable of, report the device's maximum possible bandwidth and the
5638  * upstream link that limits its performance.  If @verbose, always print
5639  * the available bandwidth, even if the device isn't constrained.
5640  */
5641 void __pcie_print_link_status(struct pci_dev *dev, bool verbose)
5642 {
5643 	enum pcie_link_width width, width_cap;
5644 	enum pci_bus_speed speed, speed_cap;
5645 	struct pci_dev *limiting_dev = NULL;
5646 	u32 bw_avail, bw_cap;
5647 
5648 	bw_cap = pcie_bandwidth_capable(dev, &speed_cap, &width_cap);
5649 	bw_avail = pcie_bandwidth_available(dev, &limiting_dev, &speed, &width);
5650 
5651 	if (bw_avail >= bw_cap && verbose)
5652 		pci_info(dev, "%u.%03u Gb/s available PCIe bandwidth (%s x%d link)\n",
5653 			 bw_cap / 1000, bw_cap % 1000,
5654 			 PCIE_SPEED2STR(speed_cap), width_cap);
5655 	else if (bw_avail < bw_cap)
5656 		pci_info(dev, "%u.%03u Gb/s available PCIe bandwidth, limited by %s x%d link at %s (capable of %u.%03u Gb/s with %s x%d link)\n",
5657 			 bw_avail / 1000, bw_avail % 1000,
5658 			 PCIE_SPEED2STR(speed), width,
5659 			 limiting_dev ? pci_name(limiting_dev) : "<unknown>",
5660 			 bw_cap / 1000, bw_cap % 1000,
5661 			 PCIE_SPEED2STR(speed_cap), width_cap);
5662 }
5663 
5664 /**
5665  * pcie_print_link_status - Report the PCI device's link speed and width
5666  * @dev: PCI device to query
5667  *
5668  * Report the available bandwidth at the device.
5669  */
5670 void pcie_print_link_status(struct pci_dev *dev)
5671 {
5672 	__pcie_print_link_status(dev, true);
5673 }
5674 EXPORT_SYMBOL(pcie_print_link_status);
5675 
5676 /**
5677  * pci_select_bars - Make BAR mask from the type of resource
5678  * @dev: the PCI device for which BAR mask is made
5679  * @flags: resource type mask to be selected
5680  *
5681  * This helper routine makes bar mask from the type of resource.
5682  */
5683 int pci_select_bars(struct pci_dev *dev, unsigned long flags)
5684 {
5685 	int i, bars = 0;
5686 	for (i = 0; i < PCI_NUM_RESOURCES; i++)
5687 		if (pci_resource_flags(dev, i) & flags)
5688 			bars |= (1 << i);
5689 	return bars;
5690 }
5691 EXPORT_SYMBOL(pci_select_bars);
5692 
5693 /* Some architectures require additional programming to enable VGA */
5694 static arch_set_vga_state_t arch_set_vga_state;
5695 
5696 void __init pci_register_set_vga_state(arch_set_vga_state_t func)
5697 {
5698 	arch_set_vga_state = func;	/* NULL disables */
5699 }
5700 
5701 static int pci_set_vga_state_arch(struct pci_dev *dev, bool decode,
5702 				  unsigned int command_bits, u32 flags)
5703 {
5704 	if (arch_set_vga_state)
5705 		return arch_set_vga_state(dev, decode, command_bits,
5706 						flags);
5707 	return 0;
5708 }
5709 
5710 /**
5711  * pci_set_vga_state - set VGA decode state on device and parents if requested
5712  * @dev: the PCI device
5713  * @decode: true = enable decoding, false = disable decoding
5714  * @command_bits: PCI_COMMAND_IO and/or PCI_COMMAND_MEMORY
5715  * @flags: traverse ancestors and change bridges
5716  * CHANGE_BRIDGE_ONLY / CHANGE_BRIDGE
5717  */
5718 int pci_set_vga_state(struct pci_dev *dev, bool decode,
5719 		      unsigned int command_bits, u32 flags)
5720 {
5721 	struct pci_bus *bus;
5722 	struct pci_dev *bridge;
5723 	u16 cmd;
5724 	int rc;
5725 
5726 	WARN_ON((flags & PCI_VGA_STATE_CHANGE_DECODES) && (command_bits & ~(PCI_COMMAND_IO|PCI_COMMAND_MEMORY)));
5727 
5728 	/* ARCH specific VGA enables */
5729 	rc = pci_set_vga_state_arch(dev, decode, command_bits, flags);
5730 	if (rc)
5731 		return rc;
5732 
5733 	if (flags & PCI_VGA_STATE_CHANGE_DECODES) {
5734 		pci_read_config_word(dev, PCI_COMMAND, &cmd);
5735 		if (decode == true)
5736 			cmd |= command_bits;
5737 		else
5738 			cmd &= ~command_bits;
5739 		pci_write_config_word(dev, PCI_COMMAND, cmd);
5740 	}
5741 
5742 	if (!(flags & PCI_VGA_STATE_CHANGE_BRIDGE))
5743 		return 0;
5744 
5745 	bus = dev->bus;
5746 	while (bus) {
5747 		bridge = bus->self;
5748 		if (bridge) {
5749 			pci_read_config_word(bridge, PCI_BRIDGE_CONTROL,
5750 					     &cmd);
5751 			if (decode == true)
5752 				cmd |= PCI_BRIDGE_CTL_VGA;
5753 			else
5754 				cmd &= ~PCI_BRIDGE_CTL_VGA;
5755 			pci_write_config_word(bridge, PCI_BRIDGE_CONTROL,
5756 					      cmd);
5757 		}
5758 		bus = bus->parent;
5759 	}
5760 	return 0;
5761 }
5762 
5763 /**
5764  * pci_add_dma_alias - Add a DMA devfn alias for a device
5765  * @dev: the PCI device for which alias is added
5766  * @devfn: alias slot and function
5767  *
5768  * This helper encodes an 8-bit devfn as a bit number in dma_alias_mask
5769  * which is used to program permissible bus-devfn source addresses for DMA
5770  * requests in an IOMMU.  These aliases factor into IOMMU group creation
5771  * and are useful for devices generating DMA requests beyond or different
5772  * from their logical bus-devfn.  Examples include device quirks where the
5773  * device simply uses the wrong devfn, as well as non-transparent bridges
5774  * where the alias may be a proxy for devices in another domain.
5775  *
5776  * IOMMU group creation is performed during device discovery or addition,
5777  * prior to any potential DMA mapping and therefore prior to driver probing
5778  * (especially for userspace assigned devices where IOMMU group definition
5779  * cannot be left as a userspace activity).  DMA aliases should therefore
5780  * be configured via quirks, such as the PCI fixup header quirk.
5781  */
5782 void pci_add_dma_alias(struct pci_dev *dev, u8 devfn)
5783 {
5784 	if (!dev->dma_alias_mask)
5785 		dev->dma_alias_mask = bitmap_zalloc(U8_MAX, GFP_KERNEL);
5786 	if (!dev->dma_alias_mask) {
5787 		pci_warn(dev, "Unable to allocate DMA alias mask\n");
5788 		return;
5789 	}
5790 
5791 	set_bit(devfn, dev->dma_alias_mask);
5792 	pci_info(dev, "Enabling fixed DMA alias to %02x.%d\n",
5793 		 PCI_SLOT(devfn), PCI_FUNC(devfn));
5794 }
5795 
5796 bool pci_devs_are_dma_aliases(struct pci_dev *dev1, struct pci_dev *dev2)
5797 {
5798 	return (dev1->dma_alias_mask &&
5799 		test_bit(dev2->devfn, dev1->dma_alias_mask)) ||
5800 	       (dev2->dma_alias_mask &&
5801 		test_bit(dev1->devfn, dev2->dma_alias_mask));
5802 }
5803 
5804 bool pci_device_is_present(struct pci_dev *pdev)
5805 {
5806 	u32 v;
5807 
5808 	if (pci_dev_is_disconnected(pdev))
5809 		return false;
5810 	return pci_bus_read_dev_vendor_id(pdev->bus, pdev->devfn, &v, 0);
5811 }
5812 EXPORT_SYMBOL_GPL(pci_device_is_present);
5813 
5814 void pci_ignore_hotplug(struct pci_dev *dev)
5815 {
5816 	struct pci_dev *bridge = dev->bus->self;
5817 
5818 	dev->ignore_hotplug = 1;
5819 	/* Propagate the "ignore hotplug" setting to the parent bridge. */
5820 	if (bridge)
5821 		bridge->ignore_hotplug = 1;
5822 }
5823 EXPORT_SYMBOL_GPL(pci_ignore_hotplug);
5824 
5825 resource_size_t __weak pcibios_default_alignment(void)
5826 {
5827 	return 0;
5828 }
5829 
5830 #define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
5831 static char resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
5832 static DEFINE_SPINLOCK(resource_alignment_lock);
5833 
5834 /**
5835  * pci_specified_resource_alignment - get resource alignment specified by user.
5836  * @dev: the PCI device to get
5837  * @resize: whether or not to change resources' size when reassigning alignment
5838  *
5839  * RETURNS: Resource alignment if it is specified.
5840  *          Zero if it is not specified.
5841  */
5842 static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
5843 							bool *resize)
5844 {
5845 	int align_order, count;
5846 	resource_size_t align = pcibios_default_alignment();
5847 	const char *p;
5848 	int ret;
5849 
5850 	spin_lock(&resource_alignment_lock);
5851 	p = resource_alignment_param;
5852 	if (!*p && !align)
5853 		goto out;
5854 	if (pci_has_flag(PCI_PROBE_ONLY)) {
5855 		align = 0;
5856 		pr_info_once("PCI: Ignoring requested alignments (PCI_PROBE_ONLY)\n");
5857 		goto out;
5858 	}
5859 
5860 	while (*p) {
5861 		count = 0;
5862 		if (sscanf(p, "%d%n", &align_order, &count) == 1 &&
5863 							p[count] == '@') {
5864 			p += count + 1;
5865 		} else {
5866 			align_order = -1;
5867 		}
5868 
5869 		ret = pci_dev_str_match(dev, p, &p);
5870 		if (ret == 1) {
5871 			*resize = true;
5872 			if (align_order == -1)
5873 				align = PAGE_SIZE;
5874 			else
5875 				align = 1 << align_order;
5876 			break;
5877 		} else if (ret < 0) {
5878 			pr_err("PCI: Can't parse resource_alignment parameter: %s\n",
5879 			       p);
5880 			break;
5881 		}
5882 
5883 		if (*p != ';' && *p != ',') {
5884 			/* End of param or invalid format */
5885 			break;
5886 		}
5887 		p++;
5888 	}
5889 out:
5890 	spin_unlock(&resource_alignment_lock);
5891 	return align;
5892 }
5893 
5894 static void pci_request_resource_alignment(struct pci_dev *dev, int bar,
5895 					   resource_size_t align, bool resize)
5896 {
5897 	struct resource *r = &dev->resource[bar];
5898 	resource_size_t size;
5899 
5900 	if (!(r->flags & IORESOURCE_MEM))
5901 		return;
5902 
5903 	if (r->flags & IORESOURCE_PCI_FIXED) {
5904 		pci_info(dev, "BAR%d %pR: ignoring requested alignment %#llx\n",
5905 			 bar, r, (unsigned long long)align);
5906 		return;
5907 	}
5908 
5909 	size = resource_size(r);
5910 	if (size >= align)
5911 		return;
5912 
5913 	/*
5914 	 * Increase the alignment of the resource.  There are two ways we
5915 	 * can do this:
5916 	 *
5917 	 * 1) Increase the size of the resource.  BARs are aligned on their
5918 	 *    size, so when we reallocate space for this resource, we'll
5919 	 *    allocate it with the larger alignment.  This also prevents
5920 	 *    assignment of any other BARs inside the alignment region, so
5921 	 *    if we're requesting page alignment, this means no other BARs
5922 	 *    will share the page.
5923 	 *
5924 	 *    The disadvantage is that this makes the resource larger than
5925 	 *    the hardware BAR, which may break drivers that compute things
5926 	 *    based on the resource size, e.g., to find registers at a
5927 	 *    fixed offset before the end of the BAR.
5928 	 *
5929 	 * 2) Retain the resource size, but use IORESOURCE_STARTALIGN and
5930 	 *    set r->start to the desired alignment.  By itself this
5931 	 *    doesn't prevent other BARs being put inside the alignment
5932 	 *    region, but if we realign *every* resource of every device in
5933 	 *    the system, none of them will share an alignment region.
5934 	 *
5935 	 * When the user has requested alignment for only some devices via
5936 	 * the "pci=resource_alignment" argument, "resize" is true and we
5937 	 * use the first method.  Otherwise we assume we're aligning all
5938 	 * devices and we use the second.
5939 	 */
5940 
5941 	pci_info(dev, "BAR%d %pR: requesting alignment to %#llx\n",
5942 		 bar, r, (unsigned long long)align);
5943 
5944 	if (resize) {
5945 		r->start = 0;
5946 		r->end = align - 1;
5947 	} else {
5948 		r->flags &= ~IORESOURCE_SIZEALIGN;
5949 		r->flags |= IORESOURCE_STARTALIGN;
5950 		r->start = align;
5951 		r->end = r->start + size - 1;
5952 	}
5953 	r->flags |= IORESOURCE_UNSET;
5954 }
5955 
5956 /*
5957  * This function disables memory decoding and releases memory resources
5958  * of the device specified by kernel's boot parameter 'pci=resource_alignment='.
5959  * It also rounds up size to specified alignment.
5960  * Later on, the kernel will assign page-aligned memory resource back
5961  * to the device.
5962  */
5963 void pci_reassigndev_resource_alignment(struct pci_dev *dev)
5964 {
5965 	int i;
5966 	struct resource *r;
5967 	resource_size_t align;
5968 	u16 command;
5969 	bool resize = false;
5970 
5971 	/*
5972 	 * VF BARs are read-only zero according to SR-IOV spec r1.1, sec
5973 	 * 3.4.1.11.  Their resources are allocated from the space
5974 	 * described by the VF BARx register in the PF's SR-IOV capability.
5975 	 * We can't influence their alignment here.
5976 	 */
5977 	if (dev->is_virtfn)
5978 		return;
5979 
5980 	/* check if specified PCI is target device to reassign */
5981 	align = pci_specified_resource_alignment(dev, &resize);
5982 	if (!align)
5983 		return;
5984 
5985 	if (dev->hdr_type == PCI_HEADER_TYPE_NORMAL &&
5986 	    (dev->class >> 8) == PCI_CLASS_BRIDGE_HOST) {
5987 		pci_warn(dev, "Can't reassign resources to host bridge\n");
5988 		return;
5989 	}
5990 
5991 	pci_read_config_word(dev, PCI_COMMAND, &command);
5992 	command &= ~PCI_COMMAND_MEMORY;
5993 	pci_write_config_word(dev, PCI_COMMAND, command);
5994 
5995 	for (i = 0; i <= PCI_ROM_RESOURCE; i++)
5996 		pci_request_resource_alignment(dev, i, align, resize);
5997 
5998 	/*
5999 	 * Need to disable bridge's resource window,
6000 	 * to enable the kernel to reassign new resource
6001 	 * window later on.
6002 	 */
6003 	if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE &&
6004 	    (dev->class >> 8) == PCI_CLASS_BRIDGE_PCI) {
6005 		for (i = PCI_BRIDGE_RESOURCES; i < PCI_NUM_RESOURCES; i++) {
6006 			r = &dev->resource[i];
6007 			if (!(r->flags & IORESOURCE_MEM))
6008 				continue;
6009 			r->flags |= IORESOURCE_UNSET;
6010 			r->end = resource_size(r) - 1;
6011 			r->start = 0;
6012 		}
6013 		pci_disable_bridge_window(dev);
6014 	}
6015 }
6016 
6017 static ssize_t pci_set_resource_alignment_param(const char *buf, size_t count)
6018 {
6019 	if (count > RESOURCE_ALIGNMENT_PARAM_SIZE - 1)
6020 		count = RESOURCE_ALIGNMENT_PARAM_SIZE - 1;
6021 	spin_lock(&resource_alignment_lock);
6022 	strncpy(resource_alignment_param, buf, count);
6023 	resource_alignment_param[count] = '\0';
6024 	spin_unlock(&resource_alignment_lock);
6025 	return count;
6026 }
6027 
6028 static ssize_t pci_get_resource_alignment_param(char *buf, size_t size)
6029 {
6030 	size_t count;
6031 	spin_lock(&resource_alignment_lock);
6032 	count = snprintf(buf, size, "%s", resource_alignment_param);
6033 	spin_unlock(&resource_alignment_lock);
6034 	return count;
6035 }
6036 
6037 static ssize_t pci_resource_alignment_show(struct bus_type *bus, char *buf)
6038 {
6039 	return pci_get_resource_alignment_param(buf, PAGE_SIZE);
6040 }
6041 
6042 static ssize_t pci_resource_alignment_store(struct bus_type *bus,
6043 					const char *buf, size_t count)
6044 {
6045 	return pci_set_resource_alignment_param(buf, count);
6046 }
6047 
6048 static BUS_ATTR(resource_alignment, 0644, pci_resource_alignment_show,
6049 					pci_resource_alignment_store);
6050 
6051 static int __init pci_resource_alignment_sysfs_init(void)
6052 {
6053 	return bus_create_file(&pci_bus_type,
6054 					&bus_attr_resource_alignment);
6055 }
6056 late_initcall(pci_resource_alignment_sysfs_init);
6057 
6058 static void pci_no_domains(void)
6059 {
6060 #ifdef CONFIG_PCI_DOMAINS
6061 	pci_domains_supported = 0;
6062 #endif
6063 }
6064 
6065 #ifdef CONFIG_PCI_DOMAINS_GENERIC
6066 static atomic_t __domain_nr = ATOMIC_INIT(-1);
6067 
6068 static int pci_get_new_domain_nr(void)
6069 {
6070 	return atomic_inc_return(&__domain_nr);
6071 }
6072 
6073 static int of_pci_bus_find_domain_nr(struct device *parent)
6074 {
6075 	static int use_dt_domains = -1;
6076 	int domain = -1;
6077 
6078 	if (parent)
6079 		domain = of_get_pci_domain_nr(parent->of_node);
6080 	/*
6081 	 * Check DT domain and use_dt_domains values.
6082 	 *
6083 	 * If DT domain property is valid (domain >= 0) and
6084 	 * use_dt_domains != 0, the DT assignment is valid since this means
6085 	 * we have not previously allocated a domain number by using
6086 	 * pci_get_new_domain_nr(); we should also update use_dt_domains to
6087 	 * 1, to indicate that we have just assigned a domain number from
6088 	 * DT.
6089 	 *
6090 	 * If DT domain property value is not valid (ie domain < 0), and we
6091 	 * have not previously assigned a domain number from DT
6092 	 * (use_dt_domains != 1) we should assign a domain number by
6093 	 * using the:
6094 	 *
6095 	 * pci_get_new_domain_nr()
6096 	 *
6097 	 * API and update the use_dt_domains value to keep track of method we
6098 	 * are using to assign domain numbers (use_dt_domains = 0).
6099 	 *
6100 	 * All other combinations imply we have a platform that is trying
6101 	 * to mix domain numbers obtained from DT and pci_get_new_domain_nr(),
6102 	 * which is a recipe for domain mishandling and it is prevented by
6103 	 * invalidating the domain value (domain = -1) and printing a
6104 	 * corresponding error.
6105 	 */
6106 	if (domain >= 0 && use_dt_domains) {
6107 		use_dt_domains = 1;
6108 	} else if (domain < 0 && use_dt_domains != 1) {
6109 		use_dt_domains = 0;
6110 		domain = pci_get_new_domain_nr();
6111 	} else {
6112 		if (parent)
6113 			pr_err("Node %pOF has ", parent->of_node);
6114 		pr_err("Inconsistent \"linux,pci-domain\" property in DT\n");
6115 		domain = -1;
6116 	}
6117 
6118 	return domain;
6119 }
6120 
6121 int pci_bus_find_domain_nr(struct pci_bus *bus, struct device *parent)
6122 {
6123 	return acpi_disabled ? of_pci_bus_find_domain_nr(parent) :
6124 			       acpi_pci_bus_find_domain_nr(bus);
6125 }
6126 #endif
6127 
6128 /**
6129  * pci_ext_cfg_avail - can we access extended PCI config space?
6130  *
6131  * Returns 1 if we can access PCI extended config space (offsets
6132  * greater than 0xff). This is the default implementation. Architecture
6133  * implementations can override this.
6134  */
6135 int __weak pci_ext_cfg_avail(void)
6136 {
6137 	return 1;
6138 }
6139 
6140 void __weak pci_fixup_cardbus(struct pci_bus *bus)
6141 {
6142 }
6143 EXPORT_SYMBOL(pci_fixup_cardbus);
6144 
6145 static int __init pci_setup(char *str)
6146 {
6147 	while (str) {
6148 		char *k = strchr(str, ',');
6149 		if (k)
6150 			*k++ = 0;
6151 		if (*str && (str = pcibios_setup(str)) && *str) {
6152 			if (!strcmp(str, "nomsi")) {
6153 				pci_no_msi();
6154 			} else if (!strncmp(str, "noats", 5)) {
6155 				pr_info("PCIe: ATS is disabled\n");
6156 				pcie_ats_disabled = true;
6157 			} else if (!strcmp(str, "noaer")) {
6158 				pci_no_aer();
6159 			} else if (!strcmp(str, "earlydump")) {
6160 				pci_early_dump = true;
6161 			} else if (!strncmp(str, "realloc=", 8)) {
6162 				pci_realloc_get_opt(str + 8);
6163 			} else if (!strncmp(str, "realloc", 7)) {
6164 				pci_realloc_get_opt("on");
6165 			} else if (!strcmp(str, "nodomains")) {
6166 				pci_no_domains();
6167 			} else if (!strncmp(str, "noari", 5)) {
6168 				pcie_ari_disabled = true;
6169 			} else if (!strncmp(str, "cbiosize=", 9)) {
6170 				pci_cardbus_io_size = memparse(str + 9, &str);
6171 			} else if (!strncmp(str, "cbmemsize=", 10)) {
6172 				pci_cardbus_mem_size = memparse(str + 10, &str);
6173 			} else if (!strncmp(str, "resource_alignment=", 19)) {
6174 				pci_set_resource_alignment_param(str + 19,
6175 							strlen(str + 19));
6176 			} else if (!strncmp(str, "ecrc=", 5)) {
6177 				pcie_ecrc_get_policy(str + 5);
6178 			} else if (!strncmp(str, "hpiosize=", 9)) {
6179 				pci_hotplug_io_size = memparse(str + 9, &str);
6180 			} else if (!strncmp(str, "hpmemsize=", 10)) {
6181 				pci_hotplug_mem_size = memparse(str + 10, &str);
6182 			} else if (!strncmp(str, "hpbussize=", 10)) {
6183 				pci_hotplug_bus_size =
6184 					simple_strtoul(str + 10, &str, 0);
6185 				if (pci_hotplug_bus_size > 0xff)
6186 					pci_hotplug_bus_size = DEFAULT_HOTPLUG_BUS_SIZE;
6187 			} else if (!strncmp(str, "pcie_bus_tune_off", 17)) {
6188 				pcie_bus_config = PCIE_BUS_TUNE_OFF;
6189 			} else if (!strncmp(str, "pcie_bus_safe", 13)) {
6190 				pcie_bus_config = PCIE_BUS_SAFE;
6191 			} else if (!strncmp(str, "pcie_bus_perf", 13)) {
6192 				pcie_bus_config = PCIE_BUS_PERFORMANCE;
6193 			} else if (!strncmp(str, "pcie_bus_peer2peer", 18)) {
6194 				pcie_bus_config = PCIE_BUS_PEER2PEER;
6195 			} else if (!strncmp(str, "pcie_scan_all", 13)) {
6196 				pci_add_flags(PCI_SCAN_ALL_PCIE_DEVS);
6197 			} else if (!strncmp(str, "disable_acs_redir=", 18)) {
6198 				disable_acs_redir_param = str + 18;
6199 			} else {
6200 				printk(KERN_ERR "PCI: Unknown option `%s'\n",
6201 						str);
6202 			}
6203 		}
6204 		str = k;
6205 	}
6206 	return 0;
6207 }
6208 early_param("pci", pci_setup);
6209