xref: /linux/drivers/pci/probe.c (revision 2eff01ee2881becc9daaa0d53477ec202136b1f4)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * PCI detection and setup code
4  */
5 
6 #include <linux/kernel.h>
7 #include <linux/delay.h>
8 #include <linux/init.h>
9 #include <linux/pci.h>
10 #include <linux/msi.h>
11 #include <linux/of_pci.h>
12 #include <linux/pci_hotplug.h>
13 #include <linux/slab.h>
14 #include <linux/module.h>
15 #include <linux/cpumask.h>
16 #include <linux/aer.h>
17 #include <linux/acpi.h>
18 #include <linux/hypervisor.h>
19 #include <linux/irqdomain.h>
20 #include <linux/pm_runtime.h>
21 #include <linux/bitfield.h>
22 #include "pci.h"
23 
24 #define CARDBUS_LATENCY_TIMER	176	/* secondary latency timer */
25 #define CARDBUS_RESERVE_BUSNR	3
26 
27 static struct resource busn_resource = {
28 	.name	= "PCI busn",
29 	.start	= 0,
30 	.end	= 255,
31 	.flags	= IORESOURCE_BUS,
32 };
33 
34 /* Ugh.  Need to stop exporting this to modules. */
35 LIST_HEAD(pci_root_buses);
36 EXPORT_SYMBOL(pci_root_buses);
37 
38 static LIST_HEAD(pci_domain_busn_res_list);
39 
40 struct pci_domain_busn_res {
41 	struct list_head list;
42 	struct resource res;
43 	int domain_nr;
44 };
45 
46 static struct resource *get_pci_domain_busn_res(int domain_nr)
47 {
48 	struct pci_domain_busn_res *r;
49 
50 	list_for_each_entry(r, &pci_domain_busn_res_list, list)
51 		if (r->domain_nr == domain_nr)
52 			return &r->res;
53 
54 	r = kzalloc(sizeof(*r), GFP_KERNEL);
55 	if (!r)
56 		return NULL;
57 
58 	r->domain_nr = domain_nr;
59 	r->res.start = 0;
60 	r->res.end = 0xff;
61 	r->res.flags = IORESOURCE_BUS | IORESOURCE_PCI_FIXED;
62 
63 	list_add_tail(&r->list, &pci_domain_busn_res_list);
64 
65 	return &r->res;
66 }
67 
68 /*
69  * Some device drivers need know if PCI is initiated.
70  * Basically, we think PCI is not initiated when there
71  * is no device to be found on the pci_bus_type.
72  */
73 int no_pci_devices(void)
74 {
75 	struct device *dev;
76 	int no_devices;
77 
78 	dev = bus_find_next_device(&pci_bus_type, NULL);
79 	no_devices = (dev == NULL);
80 	put_device(dev);
81 	return no_devices;
82 }
83 EXPORT_SYMBOL(no_pci_devices);
84 
85 /*
86  * PCI Bus Class
87  */
88 static void release_pcibus_dev(struct device *dev)
89 {
90 	struct pci_bus *pci_bus = to_pci_bus(dev);
91 
92 	put_device(pci_bus->bridge);
93 	pci_bus_remove_resources(pci_bus);
94 	pci_release_bus_of_node(pci_bus);
95 	kfree(pci_bus);
96 }
97 
98 static const struct class pcibus_class = {
99 	.name		= "pci_bus",
100 	.dev_release	= &release_pcibus_dev,
101 	.dev_groups	= pcibus_groups,
102 };
103 
104 static int __init pcibus_class_init(void)
105 {
106 	return class_register(&pcibus_class);
107 }
108 postcore_initcall(pcibus_class_init);
109 
110 static u64 pci_size(u64 base, u64 maxbase, u64 mask)
111 {
112 	u64 size = mask & maxbase;	/* Find the significant bits */
113 	if (!size)
114 		return 0;
115 
116 	/*
117 	 * Get the lowest of them to find the decode size, and from that
118 	 * the extent.
119 	 */
120 	size = size & ~(size-1);
121 
122 	/*
123 	 * base == maxbase can be valid only if the BAR has already been
124 	 * programmed with all 1s.
125 	 */
126 	if (base == maxbase && ((base | (size - 1)) & mask) != mask)
127 		return 0;
128 
129 	return size;
130 }
131 
132 static inline unsigned long decode_bar(struct pci_dev *dev, u32 bar)
133 {
134 	u32 mem_type;
135 	unsigned long flags;
136 
137 	if ((bar & PCI_BASE_ADDRESS_SPACE) == PCI_BASE_ADDRESS_SPACE_IO) {
138 		flags = bar & ~PCI_BASE_ADDRESS_IO_MASK;
139 		flags |= IORESOURCE_IO;
140 		return flags;
141 	}
142 
143 	flags = bar & ~PCI_BASE_ADDRESS_MEM_MASK;
144 	flags |= IORESOURCE_MEM;
145 	if (flags & PCI_BASE_ADDRESS_MEM_PREFETCH)
146 		flags |= IORESOURCE_PREFETCH;
147 
148 	mem_type = bar & PCI_BASE_ADDRESS_MEM_TYPE_MASK;
149 	switch (mem_type) {
150 	case PCI_BASE_ADDRESS_MEM_TYPE_32:
151 		break;
152 	case PCI_BASE_ADDRESS_MEM_TYPE_1M:
153 		/* 1M mem BAR treated as 32-bit BAR */
154 		break;
155 	case PCI_BASE_ADDRESS_MEM_TYPE_64:
156 		flags |= IORESOURCE_MEM_64;
157 		break;
158 	default:
159 		/* mem unknown type treated as 32-bit BAR */
160 		break;
161 	}
162 	return flags;
163 }
164 
165 #define PCI_COMMAND_DECODE_ENABLE	(PCI_COMMAND_MEMORY | PCI_COMMAND_IO)
166 
167 /**
168  * __pci_read_base - Read a PCI BAR
169  * @dev: the PCI device
170  * @type: type of the BAR
171  * @res: resource buffer to be filled in
172  * @pos: BAR position in the config space
173  *
174  * Returns 1 if the BAR is 64-bit, or 0 if 32-bit.
175  */
176 int __pci_read_base(struct pci_dev *dev, enum pci_bar_type type,
177 		    struct resource *res, unsigned int pos)
178 {
179 	u32 l = 0, sz = 0, mask;
180 	u64 l64, sz64, mask64;
181 	u16 orig_cmd;
182 	struct pci_bus_region region, inverted_region;
183 	const char *res_name = pci_resource_name(dev, res - dev->resource);
184 
185 	mask = type ? PCI_ROM_ADDRESS_MASK : ~0;
186 
187 	/* No printks while decoding is disabled! */
188 	if (!dev->mmio_always_on) {
189 		pci_read_config_word(dev, PCI_COMMAND, &orig_cmd);
190 		if (orig_cmd & PCI_COMMAND_DECODE_ENABLE) {
191 			pci_write_config_word(dev, PCI_COMMAND,
192 				orig_cmd & ~PCI_COMMAND_DECODE_ENABLE);
193 		}
194 	}
195 
196 	res->name = pci_name(dev);
197 
198 	pci_read_config_dword(dev, pos, &l);
199 	pci_write_config_dword(dev, pos, l | mask);
200 	pci_read_config_dword(dev, pos, &sz);
201 	pci_write_config_dword(dev, pos, l);
202 
203 	/*
204 	 * All bits set in sz means the device isn't working properly.
205 	 * If the BAR isn't implemented, all bits must be 0.  If it's a
206 	 * memory BAR or a ROM, bit 0 must be clear; if it's an io BAR, bit
207 	 * 1 must be clear.
208 	 */
209 	if (PCI_POSSIBLE_ERROR(sz))
210 		sz = 0;
211 
212 	/*
213 	 * I don't know how l can have all bits set.  Copied from old code.
214 	 * Maybe it fixes a bug on some ancient platform.
215 	 */
216 	if (PCI_POSSIBLE_ERROR(l))
217 		l = 0;
218 
219 	if (type == pci_bar_unknown) {
220 		res->flags = decode_bar(dev, l);
221 		res->flags |= IORESOURCE_SIZEALIGN;
222 		if (res->flags & IORESOURCE_IO) {
223 			l64 = l & PCI_BASE_ADDRESS_IO_MASK;
224 			sz64 = sz & PCI_BASE_ADDRESS_IO_MASK;
225 			mask64 = PCI_BASE_ADDRESS_IO_MASK & (u32)IO_SPACE_LIMIT;
226 		} else {
227 			l64 = l & PCI_BASE_ADDRESS_MEM_MASK;
228 			sz64 = sz & PCI_BASE_ADDRESS_MEM_MASK;
229 			mask64 = (u32)PCI_BASE_ADDRESS_MEM_MASK;
230 		}
231 	} else {
232 		if (l & PCI_ROM_ADDRESS_ENABLE)
233 			res->flags |= IORESOURCE_ROM_ENABLE;
234 		l64 = l & PCI_ROM_ADDRESS_MASK;
235 		sz64 = sz & PCI_ROM_ADDRESS_MASK;
236 		mask64 = PCI_ROM_ADDRESS_MASK;
237 	}
238 
239 	if (res->flags & IORESOURCE_MEM_64) {
240 		pci_read_config_dword(dev, pos + 4, &l);
241 		pci_write_config_dword(dev, pos + 4, ~0);
242 		pci_read_config_dword(dev, pos + 4, &sz);
243 		pci_write_config_dword(dev, pos + 4, l);
244 
245 		l64 |= ((u64)l << 32);
246 		sz64 |= ((u64)sz << 32);
247 		mask64 |= ((u64)~0 << 32);
248 	}
249 
250 	if (!dev->mmio_always_on && (orig_cmd & PCI_COMMAND_DECODE_ENABLE))
251 		pci_write_config_word(dev, PCI_COMMAND, orig_cmd);
252 
253 	if (!sz64)
254 		goto fail;
255 
256 	sz64 = pci_size(l64, sz64, mask64);
257 	if (!sz64) {
258 		pci_info(dev, FW_BUG "%s: invalid; can't size\n", res_name);
259 		goto fail;
260 	}
261 
262 	if (res->flags & IORESOURCE_MEM_64) {
263 		if ((sizeof(pci_bus_addr_t) < 8 || sizeof(resource_size_t) < 8)
264 		    && sz64 > 0x100000000ULL) {
265 			res->flags |= IORESOURCE_UNSET | IORESOURCE_DISABLED;
266 			res->start = 0;
267 			res->end = 0;
268 			pci_err(dev, "%s: can't handle BAR larger than 4GB (size %#010llx)\n",
269 				res_name, (unsigned long long)sz64);
270 			goto out;
271 		}
272 
273 		if ((sizeof(pci_bus_addr_t) < 8) && l) {
274 			/* Above 32-bit boundary; try to reallocate */
275 			res->flags |= IORESOURCE_UNSET;
276 			res->start = 0;
277 			res->end = sz64 - 1;
278 			pci_info(dev, "%s: can't handle BAR above 4GB (bus address %#010llx)\n",
279 				 res_name, (unsigned long long)l64);
280 			goto out;
281 		}
282 	}
283 
284 	region.start = l64;
285 	region.end = l64 + sz64 - 1;
286 
287 	pcibios_bus_to_resource(dev->bus, res, &region);
288 	pcibios_resource_to_bus(dev->bus, &inverted_region, res);
289 
290 	/*
291 	 * If "A" is a BAR value (a bus address), "bus_to_resource(A)" is
292 	 * the corresponding resource address (the physical address used by
293 	 * the CPU.  Converting that resource address back to a bus address
294 	 * should yield the original BAR value:
295 	 *
296 	 *     resource_to_bus(bus_to_resource(A)) == A
297 	 *
298 	 * If it doesn't, CPU accesses to "bus_to_resource(A)" will not
299 	 * be claimed by the device.
300 	 */
301 	if (inverted_region.start != region.start) {
302 		res->flags |= IORESOURCE_UNSET;
303 		res->start = 0;
304 		res->end = region.end - region.start;
305 		pci_info(dev, "%s: initial BAR value %#010llx invalid\n",
306 			 res_name, (unsigned long long)region.start);
307 	}
308 
309 	goto out;
310 
311 
312 fail:
313 	res->flags = 0;
314 out:
315 	if (res->flags)
316 		pci_info(dev, "%s %pR\n", res_name, res);
317 
318 	return (res->flags & IORESOURCE_MEM_64) ? 1 : 0;
319 }
320 
321 static void pci_read_bases(struct pci_dev *dev, unsigned int howmany, int rom)
322 {
323 	unsigned int pos, reg;
324 
325 	if (dev->non_compliant_bars)
326 		return;
327 
328 	/* Per PCIe r4.0, sec 9.3.4.1.11, the VF BARs are all RO Zero */
329 	if (dev->is_virtfn)
330 		return;
331 
332 	for (pos = 0; pos < howmany; pos++) {
333 		struct resource *res = &dev->resource[pos];
334 		reg = PCI_BASE_ADDRESS_0 + (pos << 2);
335 		pos += __pci_read_base(dev, pci_bar_unknown, res, reg);
336 	}
337 
338 	if (rom) {
339 		struct resource *res = &dev->resource[PCI_ROM_RESOURCE];
340 		dev->rom_base_reg = rom;
341 		res->flags = IORESOURCE_MEM | IORESOURCE_PREFETCH |
342 				IORESOURCE_READONLY | IORESOURCE_SIZEALIGN;
343 		__pci_read_base(dev, pci_bar_mem32, res, rom);
344 	}
345 }
346 
347 static void pci_read_bridge_io(struct pci_dev *dev, struct resource *res,
348 			       bool log)
349 {
350 	u8 io_base_lo, io_limit_lo;
351 	unsigned long io_mask, io_granularity, base, limit;
352 	struct pci_bus_region region;
353 
354 	io_mask = PCI_IO_RANGE_MASK;
355 	io_granularity = 0x1000;
356 	if (dev->io_window_1k) {
357 		/* Support 1K I/O space granularity */
358 		io_mask = PCI_IO_1K_RANGE_MASK;
359 		io_granularity = 0x400;
360 	}
361 
362 	pci_read_config_byte(dev, PCI_IO_BASE, &io_base_lo);
363 	pci_read_config_byte(dev, PCI_IO_LIMIT, &io_limit_lo);
364 	base = (io_base_lo & io_mask) << 8;
365 	limit = (io_limit_lo & io_mask) << 8;
366 
367 	if ((io_base_lo & PCI_IO_RANGE_TYPE_MASK) == PCI_IO_RANGE_TYPE_32) {
368 		u16 io_base_hi, io_limit_hi;
369 
370 		pci_read_config_word(dev, PCI_IO_BASE_UPPER16, &io_base_hi);
371 		pci_read_config_word(dev, PCI_IO_LIMIT_UPPER16, &io_limit_hi);
372 		base |= ((unsigned long) io_base_hi << 16);
373 		limit |= ((unsigned long) io_limit_hi << 16);
374 	}
375 
376 	if (base <= limit) {
377 		res->flags = (io_base_lo & PCI_IO_RANGE_TYPE_MASK) | IORESOURCE_IO;
378 		region.start = base;
379 		region.end = limit + io_granularity - 1;
380 		pcibios_bus_to_resource(dev->bus, res, &region);
381 		if (log)
382 			pci_info(dev, "  bridge window %pR\n", res);
383 	}
384 }
385 
386 static void pci_read_bridge_mmio(struct pci_dev *dev, struct resource *res,
387 				 bool log)
388 {
389 	u16 mem_base_lo, mem_limit_lo;
390 	unsigned long base, limit;
391 	struct pci_bus_region region;
392 
393 	pci_read_config_word(dev, PCI_MEMORY_BASE, &mem_base_lo);
394 	pci_read_config_word(dev, PCI_MEMORY_LIMIT, &mem_limit_lo);
395 	base = ((unsigned long) mem_base_lo & PCI_MEMORY_RANGE_MASK) << 16;
396 	limit = ((unsigned long) mem_limit_lo & PCI_MEMORY_RANGE_MASK) << 16;
397 	if (base <= limit) {
398 		res->flags = (mem_base_lo & PCI_MEMORY_RANGE_TYPE_MASK) | IORESOURCE_MEM;
399 		region.start = base;
400 		region.end = limit + 0xfffff;
401 		pcibios_bus_to_resource(dev->bus, res, &region);
402 		if (log)
403 			pci_info(dev, "  bridge window %pR\n", res);
404 	}
405 }
406 
407 static void pci_read_bridge_mmio_pref(struct pci_dev *dev, struct resource *res,
408 				      bool log)
409 {
410 	u16 mem_base_lo, mem_limit_lo;
411 	u64 base64, limit64;
412 	pci_bus_addr_t base, limit;
413 	struct pci_bus_region region;
414 
415 	pci_read_config_word(dev, PCI_PREF_MEMORY_BASE, &mem_base_lo);
416 	pci_read_config_word(dev, PCI_PREF_MEMORY_LIMIT, &mem_limit_lo);
417 	base64 = (mem_base_lo & PCI_PREF_RANGE_MASK) << 16;
418 	limit64 = (mem_limit_lo & PCI_PREF_RANGE_MASK) << 16;
419 
420 	if ((mem_base_lo & PCI_PREF_RANGE_TYPE_MASK) == PCI_PREF_RANGE_TYPE_64) {
421 		u32 mem_base_hi, mem_limit_hi;
422 
423 		pci_read_config_dword(dev, PCI_PREF_BASE_UPPER32, &mem_base_hi);
424 		pci_read_config_dword(dev, PCI_PREF_LIMIT_UPPER32, &mem_limit_hi);
425 
426 		/*
427 		 * Some bridges set the base > limit by default, and some
428 		 * (broken) BIOSes do not initialize them.  If we find
429 		 * this, just assume they are not being used.
430 		 */
431 		if (mem_base_hi <= mem_limit_hi) {
432 			base64 |= (u64) mem_base_hi << 32;
433 			limit64 |= (u64) mem_limit_hi << 32;
434 		}
435 	}
436 
437 	base = (pci_bus_addr_t) base64;
438 	limit = (pci_bus_addr_t) limit64;
439 
440 	if (base != base64) {
441 		pci_err(dev, "can't handle bridge window above 4GB (bus address %#010llx)\n",
442 			(unsigned long long) base64);
443 		return;
444 	}
445 
446 	if (base <= limit) {
447 		res->flags = (mem_base_lo & PCI_PREF_RANGE_TYPE_MASK) |
448 					 IORESOURCE_MEM | IORESOURCE_PREFETCH;
449 		if (res->flags & PCI_PREF_RANGE_TYPE_64)
450 			res->flags |= IORESOURCE_MEM_64;
451 		region.start = base;
452 		region.end = limit + 0xfffff;
453 		pcibios_bus_to_resource(dev->bus, res, &region);
454 		if (log)
455 			pci_info(dev, "  bridge window %pR\n", res);
456 	}
457 }
458 
459 static void pci_read_bridge_windows(struct pci_dev *bridge)
460 {
461 	u32 buses;
462 	u16 io;
463 	u32 pmem, tmp;
464 	struct resource res;
465 
466 	pci_read_config_dword(bridge, PCI_PRIMARY_BUS, &buses);
467 	res.flags = IORESOURCE_BUS;
468 	res.start = (buses >> 8) & 0xff;
469 	res.end = (buses >> 16) & 0xff;
470 	pci_info(bridge, "PCI bridge to %pR%s\n", &res,
471 		 bridge->transparent ? " (subtractive decode)" : "");
472 
473 	pci_read_config_word(bridge, PCI_IO_BASE, &io);
474 	if (!io) {
475 		pci_write_config_word(bridge, PCI_IO_BASE, 0xe0f0);
476 		pci_read_config_word(bridge, PCI_IO_BASE, &io);
477 		pci_write_config_word(bridge, PCI_IO_BASE, 0x0);
478 	}
479 	if (io) {
480 		bridge->io_window = 1;
481 		pci_read_bridge_io(bridge, &res, true);
482 	}
483 
484 	pci_read_bridge_mmio(bridge, &res, true);
485 
486 	/*
487 	 * DECchip 21050 pass 2 errata: the bridge may miss an address
488 	 * disconnect boundary by one PCI data phase.  Workaround: do not
489 	 * use prefetching on this device.
490 	 */
491 	if (bridge->vendor == PCI_VENDOR_ID_DEC && bridge->device == 0x0001)
492 		return;
493 
494 	pci_read_config_dword(bridge, PCI_PREF_MEMORY_BASE, &pmem);
495 	if (!pmem) {
496 		pci_write_config_dword(bridge, PCI_PREF_MEMORY_BASE,
497 					       0xffe0fff0);
498 		pci_read_config_dword(bridge, PCI_PREF_MEMORY_BASE, &pmem);
499 		pci_write_config_dword(bridge, PCI_PREF_MEMORY_BASE, 0x0);
500 	}
501 	if (!pmem)
502 		return;
503 
504 	bridge->pref_window = 1;
505 
506 	if ((pmem & PCI_PREF_RANGE_TYPE_MASK) == PCI_PREF_RANGE_TYPE_64) {
507 
508 		/*
509 		 * Bridge claims to have a 64-bit prefetchable memory
510 		 * window; verify that the upper bits are actually
511 		 * writable.
512 		 */
513 		pci_read_config_dword(bridge, PCI_PREF_BASE_UPPER32, &pmem);
514 		pci_write_config_dword(bridge, PCI_PREF_BASE_UPPER32,
515 				       0xffffffff);
516 		pci_read_config_dword(bridge, PCI_PREF_BASE_UPPER32, &tmp);
517 		pci_write_config_dword(bridge, PCI_PREF_BASE_UPPER32, pmem);
518 		if (tmp)
519 			bridge->pref_64_window = 1;
520 	}
521 
522 	pci_read_bridge_mmio_pref(bridge, &res, true);
523 }
524 
525 void pci_read_bridge_bases(struct pci_bus *child)
526 {
527 	struct pci_dev *dev = child->self;
528 	struct resource *res;
529 	int i;
530 
531 	if (pci_is_root_bus(child))	/* It's a host bus, nothing to read */
532 		return;
533 
534 	pci_info(dev, "PCI bridge to %pR%s\n",
535 		 &child->busn_res,
536 		 dev->transparent ? " (subtractive decode)" : "");
537 
538 	pci_bus_remove_resources(child);
539 	for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++)
540 		child->resource[i] = &dev->resource[PCI_BRIDGE_RESOURCES+i];
541 
542 	pci_read_bridge_io(child->self, child->resource[0], false);
543 	pci_read_bridge_mmio(child->self, child->resource[1], false);
544 	pci_read_bridge_mmio_pref(child->self, child->resource[2], false);
545 
546 	if (!dev->transparent)
547 		return;
548 
549 	pci_bus_for_each_resource(child->parent, res) {
550 		if (!res || !res->flags)
551 			continue;
552 
553 		pci_bus_add_resource(child, res);
554 		pci_info(dev, "  bridge window %pR (subtractive decode)\n", res);
555 	}
556 }
557 
558 static struct pci_bus *pci_alloc_bus(struct pci_bus *parent)
559 {
560 	struct pci_bus *b;
561 
562 	b = kzalloc(sizeof(*b), GFP_KERNEL);
563 	if (!b)
564 		return NULL;
565 
566 	INIT_LIST_HEAD(&b->node);
567 	INIT_LIST_HEAD(&b->children);
568 	INIT_LIST_HEAD(&b->devices);
569 	INIT_LIST_HEAD(&b->slots);
570 	INIT_LIST_HEAD(&b->resources);
571 	b->max_bus_speed = PCI_SPEED_UNKNOWN;
572 	b->cur_bus_speed = PCI_SPEED_UNKNOWN;
573 #ifdef CONFIG_PCI_DOMAINS_GENERIC
574 	if (parent)
575 		b->domain_nr = parent->domain_nr;
576 #endif
577 	return b;
578 }
579 
580 static void pci_release_host_bridge_dev(struct device *dev)
581 {
582 	struct pci_host_bridge *bridge = to_pci_host_bridge(dev);
583 
584 	if (bridge->release_fn)
585 		bridge->release_fn(bridge);
586 
587 	pci_free_resource_list(&bridge->windows);
588 	pci_free_resource_list(&bridge->dma_ranges);
589 	kfree(bridge);
590 }
591 
592 static void pci_init_host_bridge(struct pci_host_bridge *bridge)
593 {
594 	INIT_LIST_HEAD(&bridge->windows);
595 	INIT_LIST_HEAD(&bridge->dma_ranges);
596 
597 	/*
598 	 * We assume we can manage these PCIe features.  Some systems may
599 	 * reserve these for use by the platform itself, e.g., an ACPI BIOS
600 	 * may implement its own AER handling and use _OSC to prevent the
601 	 * OS from interfering.
602 	 */
603 	bridge->native_aer = 1;
604 	bridge->native_pcie_hotplug = 1;
605 	bridge->native_shpc_hotplug = 1;
606 	bridge->native_pme = 1;
607 	bridge->native_ltr = 1;
608 	bridge->native_dpc = 1;
609 	bridge->domain_nr = PCI_DOMAIN_NR_NOT_SET;
610 	bridge->native_cxl_error = 1;
611 
612 	device_initialize(&bridge->dev);
613 }
614 
615 struct pci_host_bridge *pci_alloc_host_bridge(size_t priv)
616 {
617 	struct pci_host_bridge *bridge;
618 
619 	bridge = kzalloc(sizeof(*bridge) + priv, GFP_KERNEL);
620 	if (!bridge)
621 		return NULL;
622 
623 	pci_init_host_bridge(bridge);
624 	bridge->dev.release = pci_release_host_bridge_dev;
625 
626 	return bridge;
627 }
628 EXPORT_SYMBOL(pci_alloc_host_bridge);
629 
630 static void devm_pci_alloc_host_bridge_release(void *data)
631 {
632 	pci_free_host_bridge(data);
633 }
634 
635 struct pci_host_bridge *devm_pci_alloc_host_bridge(struct device *dev,
636 						   size_t priv)
637 {
638 	int ret;
639 	struct pci_host_bridge *bridge;
640 
641 	bridge = pci_alloc_host_bridge(priv);
642 	if (!bridge)
643 		return NULL;
644 
645 	bridge->dev.parent = dev;
646 
647 	ret = devm_add_action_or_reset(dev, devm_pci_alloc_host_bridge_release,
648 				       bridge);
649 	if (ret)
650 		return NULL;
651 
652 	ret = devm_of_pci_bridge_init(dev, bridge);
653 	if (ret)
654 		return NULL;
655 
656 	return bridge;
657 }
658 EXPORT_SYMBOL(devm_pci_alloc_host_bridge);
659 
660 void pci_free_host_bridge(struct pci_host_bridge *bridge)
661 {
662 	put_device(&bridge->dev);
663 }
664 EXPORT_SYMBOL(pci_free_host_bridge);
665 
666 /* Indexed by PCI_X_SSTATUS_FREQ (secondary bus mode and frequency) */
667 static const unsigned char pcix_bus_speed[] = {
668 	PCI_SPEED_UNKNOWN,		/* 0 */
669 	PCI_SPEED_66MHz_PCIX,		/* 1 */
670 	PCI_SPEED_100MHz_PCIX,		/* 2 */
671 	PCI_SPEED_133MHz_PCIX,		/* 3 */
672 	PCI_SPEED_UNKNOWN,		/* 4 */
673 	PCI_SPEED_66MHz_PCIX_ECC,	/* 5 */
674 	PCI_SPEED_100MHz_PCIX_ECC,	/* 6 */
675 	PCI_SPEED_133MHz_PCIX_ECC,	/* 7 */
676 	PCI_SPEED_UNKNOWN,		/* 8 */
677 	PCI_SPEED_66MHz_PCIX_266,	/* 9 */
678 	PCI_SPEED_100MHz_PCIX_266,	/* A */
679 	PCI_SPEED_133MHz_PCIX_266,	/* B */
680 	PCI_SPEED_UNKNOWN,		/* C */
681 	PCI_SPEED_66MHz_PCIX_533,	/* D */
682 	PCI_SPEED_100MHz_PCIX_533,	/* E */
683 	PCI_SPEED_133MHz_PCIX_533	/* F */
684 };
685 
686 /* Indexed by PCI_EXP_LNKCAP_SLS, PCI_EXP_LNKSTA_CLS */
687 const unsigned char pcie_link_speed[] = {
688 	PCI_SPEED_UNKNOWN,		/* 0 */
689 	PCIE_SPEED_2_5GT,		/* 1 */
690 	PCIE_SPEED_5_0GT,		/* 2 */
691 	PCIE_SPEED_8_0GT,		/* 3 */
692 	PCIE_SPEED_16_0GT,		/* 4 */
693 	PCIE_SPEED_32_0GT,		/* 5 */
694 	PCIE_SPEED_64_0GT,		/* 6 */
695 	PCI_SPEED_UNKNOWN,		/* 7 */
696 	PCI_SPEED_UNKNOWN,		/* 8 */
697 	PCI_SPEED_UNKNOWN,		/* 9 */
698 	PCI_SPEED_UNKNOWN,		/* A */
699 	PCI_SPEED_UNKNOWN,		/* B */
700 	PCI_SPEED_UNKNOWN,		/* C */
701 	PCI_SPEED_UNKNOWN,		/* D */
702 	PCI_SPEED_UNKNOWN,		/* E */
703 	PCI_SPEED_UNKNOWN		/* F */
704 };
705 EXPORT_SYMBOL_GPL(pcie_link_speed);
706 
707 const char *pci_speed_string(enum pci_bus_speed speed)
708 {
709 	/* Indexed by the pci_bus_speed enum */
710 	static const char *speed_strings[] = {
711 	    "33 MHz PCI",		/* 0x00 */
712 	    "66 MHz PCI",		/* 0x01 */
713 	    "66 MHz PCI-X",		/* 0x02 */
714 	    "100 MHz PCI-X",		/* 0x03 */
715 	    "133 MHz PCI-X",		/* 0x04 */
716 	    NULL,			/* 0x05 */
717 	    NULL,			/* 0x06 */
718 	    NULL,			/* 0x07 */
719 	    NULL,			/* 0x08 */
720 	    "66 MHz PCI-X 266",		/* 0x09 */
721 	    "100 MHz PCI-X 266",	/* 0x0a */
722 	    "133 MHz PCI-X 266",	/* 0x0b */
723 	    "Unknown AGP",		/* 0x0c */
724 	    "1x AGP",			/* 0x0d */
725 	    "2x AGP",			/* 0x0e */
726 	    "4x AGP",			/* 0x0f */
727 	    "8x AGP",			/* 0x10 */
728 	    "66 MHz PCI-X 533",		/* 0x11 */
729 	    "100 MHz PCI-X 533",	/* 0x12 */
730 	    "133 MHz PCI-X 533",	/* 0x13 */
731 	    "2.5 GT/s PCIe",		/* 0x14 */
732 	    "5.0 GT/s PCIe",		/* 0x15 */
733 	    "8.0 GT/s PCIe",		/* 0x16 */
734 	    "16.0 GT/s PCIe",		/* 0x17 */
735 	    "32.0 GT/s PCIe",		/* 0x18 */
736 	    "64.0 GT/s PCIe",		/* 0x19 */
737 	};
738 
739 	if (speed < ARRAY_SIZE(speed_strings))
740 		return speed_strings[speed];
741 	return "Unknown";
742 }
743 EXPORT_SYMBOL_GPL(pci_speed_string);
744 
745 void pcie_update_link_speed(struct pci_bus *bus)
746 {
747 	struct pci_dev *bridge = bus->self;
748 	u16 linksta;
749 
750 	pcie_capability_read_word(bridge, PCI_EXP_LNKSTA, &linksta);
751 	__pcie_update_link_speed(bus, linksta);
752 }
753 EXPORT_SYMBOL_GPL(pcie_update_link_speed);
754 
755 static unsigned char agp_speeds[] = {
756 	AGP_UNKNOWN,
757 	AGP_1X,
758 	AGP_2X,
759 	AGP_4X,
760 	AGP_8X
761 };
762 
763 static enum pci_bus_speed agp_speed(int agp3, int agpstat)
764 {
765 	int index = 0;
766 
767 	if (agpstat & 4)
768 		index = 3;
769 	else if (agpstat & 2)
770 		index = 2;
771 	else if (agpstat & 1)
772 		index = 1;
773 	else
774 		goto out;
775 
776 	if (agp3) {
777 		index += 2;
778 		if (index == 5)
779 			index = 0;
780 	}
781 
782  out:
783 	return agp_speeds[index];
784 }
785 
786 static void pci_set_bus_speed(struct pci_bus *bus)
787 {
788 	struct pci_dev *bridge = bus->self;
789 	int pos;
790 
791 	pos = pci_find_capability(bridge, PCI_CAP_ID_AGP);
792 	if (!pos)
793 		pos = pci_find_capability(bridge, PCI_CAP_ID_AGP3);
794 	if (pos) {
795 		u32 agpstat, agpcmd;
796 
797 		pci_read_config_dword(bridge, pos + PCI_AGP_STATUS, &agpstat);
798 		bus->max_bus_speed = agp_speed(agpstat & 8, agpstat & 7);
799 
800 		pci_read_config_dword(bridge, pos + PCI_AGP_COMMAND, &agpcmd);
801 		bus->cur_bus_speed = agp_speed(agpstat & 8, agpcmd & 7);
802 	}
803 
804 	pos = pci_find_capability(bridge, PCI_CAP_ID_PCIX);
805 	if (pos) {
806 		u16 status;
807 		enum pci_bus_speed max;
808 
809 		pci_read_config_word(bridge, pos + PCI_X_BRIDGE_SSTATUS,
810 				     &status);
811 
812 		if (status & PCI_X_SSTATUS_533MHZ) {
813 			max = PCI_SPEED_133MHz_PCIX_533;
814 		} else if (status & PCI_X_SSTATUS_266MHZ) {
815 			max = PCI_SPEED_133MHz_PCIX_266;
816 		} else if (status & PCI_X_SSTATUS_133MHZ) {
817 			if ((status & PCI_X_SSTATUS_VERS) == PCI_X_SSTATUS_V2)
818 				max = PCI_SPEED_133MHz_PCIX_ECC;
819 			else
820 				max = PCI_SPEED_133MHz_PCIX;
821 		} else {
822 			max = PCI_SPEED_66MHz_PCIX;
823 		}
824 
825 		bus->max_bus_speed = max;
826 		bus->cur_bus_speed =
827 			pcix_bus_speed[FIELD_GET(PCI_X_SSTATUS_FREQ, status)];
828 
829 		return;
830 	}
831 
832 	if (pci_is_pcie(bridge)) {
833 		u32 linkcap;
834 
835 		pcie_capability_read_dword(bridge, PCI_EXP_LNKCAP, &linkcap);
836 		bus->max_bus_speed = pcie_link_speed[linkcap & PCI_EXP_LNKCAP_SLS];
837 
838 		pcie_update_link_speed(bus);
839 	}
840 }
841 
842 static struct irq_domain *pci_host_bridge_msi_domain(struct pci_bus *bus)
843 {
844 	struct irq_domain *d;
845 
846 	/* If the host bridge driver sets a MSI domain of the bridge, use it */
847 	d = dev_get_msi_domain(bus->bridge);
848 
849 	/*
850 	 * Any firmware interface that can resolve the msi_domain
851 	 * should be called from here.
852 	 */
853 	if (!d)
854 		d = pci_host_bridge_of_msi_domain(bus);
855 	if (!d)
856 		d = pci_host_bridge_acpi_msi_domain(bus);
857 
858 	/*
859 	 * If no IRQ domain was found via the OF tree, try looking it up
860 	 * directly through the fwnode_handle.
861 	 */
862 	if (!d) {
863 		struct fwnode_handle *fwnode = pci_root_bus_fwnode(bus);
864 
865 		if (fwnode)
866 			d = irq_find_matching_fwnode(fwnode,
867 						     DOMAIN_BUS_PCI_MSI);
868 	}
869 
870 	return d;
871 }
872 
873 static void pci_set_bus_msi_domain(struct pci_bus *bus)
874 {
875 	struct irq_domain *d;
876 	struct pci_bus *b;
877 
878 	/*
879 	 * The bus can be a root bus, a subordinate bus, or a virtual bus
880 	 * created by an SR-IOV device.  Walk up to the first bridge device
881 	 * found or derive the domain from the host bridge.
882 	 */
883 	for (b = bus, d = NULL; !d && !pci_is_root_bus(b); b = b->parent) {
884 		if (b->self)
885 			d = dev_get_msi_domain(&b->self->dev);
886 	}
887 
888 	if (!d)
889 		d = pci_host_bridge_msi_domain(b);
890 
891 	dev_set_msi_domain(&bus->dev, d);
892 }
893 
894 static bool pci_preserve_config(struct pci_host_bridge *host_bridge)
895 {
896 	if (pci_acpi_preserve_config(host_bridge))
897 		return true;
898 
899 	if (host_bridge->dev.parent && host_bridge->dev.parent->of_node)
900 		return of_pci_preserve_config(host_bridge->dev.parent->of_node);
901 
902 	return false;
903 }
904 
905 static int pci_register_host_bridge(struct pci_host_bridge *bridge)
906 {
907 	struct device *parent = bridge->dev.parent;
908 	struct resource_entry *window, *next, *n;
909 	struct pci_bus *bus, *b;
910 	resource_size_t offset, next_offset;
911 	LIST_HEAD(resources);
912 	struct resource *res, *next_res;
913 	char addr[64], *fmt;
914 	const char *name;
915 	int err;
916 
917 	bus = pci_alloc_bus(NULL);
918 	if (!bus)
919 		return -ENOMEM;
920 
921 	bridge->bus = bus;
922 
923 	bus->sysdata = bridge->sysdata;
924 	bus->ops = bridge->ops;
925 	bus->number = bus->busn_res.start = bridge->busnr;
926 #ifdef CONFIG_PCI_DOMAINS_GENERIC
927 	if (bridge->domain_nr == PCI_DOMAIN_NR_NOT_SET)
928 		bus->domain_nr = pci_bus_find_domain_nr(bus, parent);
929 	else
930 		bus->domain_nr = bridge->domain_nr;
931 	if (bus->domain_nr < 0) {
932 		err = bus->domain_nr;
933 		goto free;
934 	}
935 #endif
936 
937 	b = pci_find_bus(pci_domain_nr(bus), bridge->busnr);
938 	if (b) {
939 		/* Ignore it if we already got here via a different bridge */
940 		dev_dbg(&b->dev, "bus already known\n");
941 		err = -EEXIST;
942 		goto free;
943 	}
944 
945 	dev_set_name(&bridge->dev, "pci%04x:%02x", pci_domain_nr(bus),
946 		     bridge->busnr);
947 
948 	err = pcibios_root_bridge_prepare(bridge);
949 	if (err)
950 		goto free;
951 
952 	/* Temporarily move resources off the list */
953 	list_splice_init(&bridge->windows, &resources);
954 	err = device_add(&bridge->dev);
955 	if (err) {
956 		put_device(&bridge->dev);
957 		goto free;
958 	}
959 	bus->bridge = get_device(&bridge->dev);
960 	device_enable_async_suspend(bus->bridge);
961 	pci_set_bus_of_node(bus);
962 	pci_set_bus_msi_domain(bus);
963 	if (bridge->msi_domain && !dev_get_msi_domain(&bus->dev) &&
964 	    !pci_host_of_has_msi_map(parent))
965 		bus->bus_flags |= PCI_BUS_FLAGS_NO_MSI;
966 
967 	if (!parent)
968 		set_dev_node(bus->bridge, pcibus_to_node(bus));
969 
970 	bus->dev.class = &pcibus_class;
971 	bus->dev.parent = bus->bridge;
972 
973 	dev_set_name(&bus->dev, "%04x:%02x", pci_domain_nr(bus), bus->number);
974 	name = dev_name(&bus->dev);
975 
976 	err = device_register(&bus->dev);
977 	if (err)
978 		goto unregister;
979 
980 	pcibios_add_bus(bus);
981 
982 	if (bus->ops->add_bus) {
983 		err = bus->ops->add_bus(bus);
984 		if (WARN_ON(err < 0))
985 			dev_err(&bus->dev, "failed to add bus: %d\n", err);
986 	}
987 
988 	/* Create legacy_io and legacy_mem files for this bus */
989 	pci_create_legacy_files(bus);
990 
991 	if (parent)
992 		dev_info(parent, "PCI host bridge to bus %s\n", name);
993 	else
994 		pr_info("PCI host bridge to bus %s\n", name);
995 
996 	if (nr_node_ids > 1 && pcibus_to_node(bus) == NUMA_NO_NODE)
997 		dev_warn(&bus->dev, "Unknown NUMA node; performance will be reduced\n");
998 
999 	/* Check if the boot configuration by FW needs to be preserved */
1000 	bridge->preserve_config = pci_preserve_config(bridge);
1001 
1002 	/* Coalesce contiguous windows */
1003 	resource_list_for_each_entry_safe(window, n, &resources) {
1004 		if (list_is_last(&window->node, &resources))
1005 			break;
1006 
1007 		next = list_next_entry(window, node);
1008 		offset = window->offset;
1009 		res = window->res;
1010 		next_offset = next->offset;
1011 		next_res = next->res;
1012 
1013 		if (res->flags != next_res->flags || offset != next_offset)
1014 			continue;
1015 
1016 		if (res->end + 1 == next_res->start) {
1017 			next_res->start = res->start;
1018 			res->flags = res->start = res->end = 0;
1019 		}
1020 	}
1021 
1022 	/* Add initial resources to the bus */
1023 	resource_list_for_each_entry_safe(window, n, &resources) {
1024 		offset = window->offset;
1025 		res = window->res;
1026 		if (!res->flags && !res->start && !res->end) {
1027 			release_resource(res);
1028 			resource_list_destroy_entry(window);
1029 			continue;
1030 		}
1031 
1032 		list_move_tail(&window->node, &bridge->windows);
1033 
1034 		if (res->flags & IORESOURCE_BUS)
1035 			pci_bus_insert_busn_res(bus, bus->number, res->end);
1036 		else
1037 			pci_bus_add_resource(bus, res);
1038 
1039 		if (offset) {
1040 			if (resource_type(res) == IORESOURCE_IO)
1041 				fmt = " (bus address [%#06llx-%#06llx])";
1042 			else
1043 				fmt = " (bus address [%#010llx-%#010llx])";
1044 
1045 			snprintf(addr, sizeof(addr), fmt,
1046 				 (unsigned long long)(res->start - offset),
1047 				 (unsigned long long)(res->end - offset));
1048 		} else
1049 			addr[0] = '\0';
1050 
1051 		dev_info(&bus->dev, "root bus resource %pR%s\n", res, addr);
1052 	}
1053 
1054 	down_write(&pci_bus_sem);
1055 	list_add_tail(&bus->node, &pci_root_buses);
1056 	up_write(&pci_bus_sem);
1057 
1058 	return 0;
1059 
1060 unregister:
1061 	put_device(&bridge->dev);
1062 	device_del(&bridge->dev);
1063 
1064 free:
1065 #ifdef CONFIG_PCI_DOMAINS_GENERIC
1066 	pci_bus_release_domain_nr(parent, bus->domain_nr);
1067 #endif
1068 	kfree(bus);
1069 	return err;
1070 }
1071 
1072 static bool pci_bridge_child_ext_cfg_accessible(struct pci_dev *bridge)
1073 {
1074 	int pos;
1075 	u32 status;
1076 
1077 	/*
1078 	 * If extended config space isn't accessible on a bridge's primary
1079 	 * bus, we certainly can't access it on the secondary bus.
1080 	 */
1081 	if (bridge->bus->bus_flags & PCI_BUS_FLAGS_NO_EXTCFG)
1082 		return false;
1083 
1084 	/*
1085 	 * PCIe Root Ports and switch ports are PCIe on both sides, so if
1086 	 * extended config space is accessible on the primary, it's also
1087 	 * accessible on the secondary.
1088 	 */
1089 	if (pci_is_pcie(bridge) &&
1090 	    (pci_pcie_type(bridge) == PCI_EXP_TYPE_ROOT_PORT ||
1091 	     pci_pcie_type(bridge) == PCI_EXP_TYPE_UPSTREAM ||
1092 	     pci_pcie_type(bridge) == PCI_EXP_TYPE_DOWNSTREAM))
1093 		return true;
1094 
1095 	/*
1096 	 * For the other bridge types:
1097 	 *   - PCI-to-PCI bridges
1098 	 *   - PCIe-to-PCI/PCI-X forward bridges
1099 	 *   - PCI/PCI-X-to-PCIe reverse bridges
1100 	 * extended config space on the secondary side is only accessible
1101 	 * if the bridge supports PCI-X Mode 2.
1102 	 */
1103 	pos = pci_find_capability(bridge, PCI_CAP_ID_PCIX);
1104 	if (!pos)
1105 		return false;
1106 
1107 	pci_read_config_dword(bridge, pos + PCI_X_STATUS, &status);
1108 	return status & (PCI_X_STATUS_266MHZ | PCI_X_STATUS_533MHZ);
1109 }
1110 
1111 static struct pci_bus *pci_alloc_child_bus(struct pci_bus *parent,
1112 					   struct pci_dev *bridge, int busnr)
1113 {
1114 	struct pci_bus *child;
1115 	struct pci_host_bridge *host;
1116 	int i;
1117 	int ret;
1118 
1119 	/* Allocate a new bus and inherit stuff from the parent */
1120 	child = pci_alloc_bus(parent);
1121 	if (!child)
1122 		return NULL;
1123 
1124 	child->parent = parent;
1125 	child->sysdata = parent->sysdata;
1126 	child->bus_flags = parent->bus_flags;
1127 
1128 	host = pci_find_host_bridge(parent);
1129 	if (host->child_ops)
1130 		child->ops = host->child_ops;
1131 	else
1132 		child->ops = parent->ops;
1133 
1134 	/*
1135 	 * Initialize some portions of the bus device, but don't register
1136 	 * it now as the parent is not properly set up yet.
1137 	 */
1138 	child->dev.class = &pcibus_class;
1139 	dev_set_name(&child->dev, "%04x:%02x", pci_domain_nr(child), busnr);
1140 
1141 	/* Set up the primary, secondary and subordinate bus numbers */
1142 	child->number = child->busn_res.start = busnr;
1143 	child->primary = parent->busn_res.start;
1144 	child->busn_res.end = 0xff;
1145 
1146 	if (!bridge) {
1147 		child->dev.parent = parent->bridge;
1148 		goto add_dev;
1149 	}
1150 
1151 	child->self = bridge;
1152 	child->bridge = get_device(&bridge->dev);
1153 	child->dev.parent = child->bridge;
1154 	pci_set_bus_of_node(child);
1155 	pci_set_bus_speed(child);
1156 
1157 	/*
1158 	 * Check whether extended config space is accessible on the child
1159 	 * bus.  Note that we currently assume it is always accessible on
1160 	 * the root bus.
1161 	 */
1162 	if (!pci_bridge_child_ext_cfg_accessible(bridge)) {
1163 		child->bus_flags |= PCI_BUS_FLAGS_NO_EXTCFG;
1164 		pci_info(child, "extended config space not accessible\n");
1165 	}
1166 
1167 	/* Set up default resource pointers and names */
1168 	for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++) {
1169 		child->resource[i] = &bridge->resource[PCI_BRIDGE_RESOURCES+i];
1170 		child->resource[i]->name = child->name;
1171 	}
1172 	bridge->subordinate = child;
1173 
1174 add_dev:
1175 	pci_set_bus_msi_domain(child);
1176 	ret = device_register(&child->dev);
1177 	WARN_ON(ret < 0);
1178 
1179 	pcibios_add_bus(child);
1180 
1181 	if (child->ops->add_bus) {
1182 		ret = child->ops->add_bus(child);
1183 		if (WARN_ON(ret < 0))
1184 			dev_err(&child->dev, "failed to add bus: %d\n", ret);
1185 	}
1186 
1187 	/* Create legacy_io and legacy_mem files for this bus */
1188 	pci_create_legacy_files(child);
1189 
1190 	return child;
1191 }
1192 
1193 struct pci_bus *pci_add_new_bus(struct pci_bus *parent, struct pci_dev *dev,
1194 				int busnr)
1195 {
1196 	struct pci_bus *child;
1197 
1198 	child = pci_alloc_child_bus(parent, dev, busnr);
1199 	if (child) {
1200 		down_write(&pci_bus_sem);
1201 		list_add_tail(&child->node, &parent->children);
1202 		up_write(&pci_bus_sem);
1203 	}
1204 	return child;
1205 }
1206 EXPORT_SYMBOL(pci_add_new_bus);
1207 
1208 static void pci_enable_rrs_sv(struct pci_dev *pdev)
1209 {
1210 	u16 root_cap = 0;
1211 
1212 	/* Enable Configuration RRS Software Visibility if supported */
1213 	pcie_capability_read_word(pdev, PCI_EXP_RTCAP, &root_cap);
1214 	if (root_cap & PCI_EXP_RTCAP_RRS_SV) {
1215 		pcie_capability_set_word(pdev, PCI_EXP_RTCTL,
1216 					 PCI_EXP_RTCTL_RRS_SVE);
1217 		pdev->config_rrs_sv = 1;
1218 	}
1219 }
1220 
1221 static unsigned int pci_scan_child_bus_extend(struct pci_bus *bus,
1222 					      unsigned int available_buses);
1223 /**
1224  * pci_ea_fixed_busnrs() - Read fixed Secondary and Subordinate bus
1225  * numbers from EA capability.
1226  * @dev: Bridge
1227  * @sec: updated with secondary bus number from EA
1228  * @sub: updated with subordinate bus number from EA
1229  *
1230  * If @dev is a bridge with EA capability that specifies valid secondary
1231  * and subordinate bus numbers, return true with the bus numbers in @sec
1232  * and @sub.  Otherwise return false.
1233  */
1234 static bool pci_ea_fixed_busnrs(struct pci_dev *dev, u8 *sec, u8 *sub)
1235 {
1236 	int ea, offset;
1237 	u32 dw;
1238 	u8 ea_sec, ea_sub;
1239 
1240 	if (dev->hdr_type != PCI_HEADER_TYPE_BRIDGE)
1241 		return false;
1242 
1243 	/* find PCI EA capability in list */
1244 	ea = pci_find_capability(dev, PCI_CAP_ID_EA);
1245 	if (!ea)
1246 		return false;
1247 
1248 	offset = ea + PCI_EA_FIRST_ENT;
1249 	pci_read_config_dword(dev, offset, &dw);
1250 	ea_sec = FIELD_GET(PCI_EA_SEC_BUS_MASK, dw);
1251 	ea_sub = FIELD_GET(PCI_EA_SUB_BUS_MASK, dw);
1252 	if (ea_sec  == 0 || ea_sub < ea_sec)
1253 		return false;
1254 
1255 	*sec = ea_sec;
1256 	*sub = ea_sub;
1257 	return true;
1258 }
1259 
1260 /*
1261  * pci_scan_bridge_extend() - Scan buses behind a bridge
1262  * @bus: Parent bus the bridge is on
1263  * @dev: Bridge itself
1264  * @max: Starting subordinate number of buses behind this bridge
1265  * @available_buses: Total number of buses available for this bridge and
1266  *		     the devices below. After the minimal bus space has
1267  *		     been allocated the remaining buses will be
1268  *		     distributed equally between hotplug-capable bridges.
1269  * @pass: Either %0 (scan already configured bridges) or %1 (scan bridges
1270  *        that need to be reconfigured.
1271  *
1272  * If it's a bridge, configure it and scan the bus behind it.
1273  * For CardBus bridges, we don't scan behind as the devices will
1274  * be handled by the bridge driver itself.
1275  *
1276  * We need to process bridges in two passes -- first we scan those
1277  * already configured by the BIOS and after we are done with all of
1278  * them, we proceed to assigning numbers to the remaining buses in
1279  * order to avoid overlaps between old and new bus numbers.
1280  *
1281  * Return: New subordinate number covering all buses behind this bridge.
1282  */
1283 static int pci_scan_bridge_extend(struct pci_bus *bus, struct pci_dev *dev,
1284 				  int max, unsigned int available_buses,
1285 				  int pass)
1286 {
1287 	struct pci_bus *child;
1288 	int is_cardbus = (dev->hdr_type == PCI_HEADER_TYPE_CARDBUS);
1289 	u32 buses, i, j = 0;
1290 	u16 bctl;
1291 	u8 primary, secondary, subordinate;
1292 	int broken = 0;
1293 	bool fixed_buses;
1294 	u8 fixed_sec, fixed_sub;
1295 	int next_busnr;
1296 
1297 	/*
1298 	 * Make sure the bridge is powered on to be able to access config
1299 	 * space of devices below it.
1300 	 */
1301 	pm_runtime_get_sync(&dev->dev);
1302 
1303 	pci_read_config_dword(dev, PCI_PRIMARY_BUS, &buses);
1304 	primary = buses & 0xFF;
1305 	secondary = (buses >> 8) & 0xFF;
1306 	subordinate = (buses >> 16) & 0xFF;
1307 
1308 	pci_dbg(dev, "scanning [bus %02x-%02x] behind bridge, pass %d\n",
1309 		secondary, subordinate, pass);
1310 
1311 	if (!primary && (primary != bus->number) && secondary && subordinate) {
1312 		pci_warn(dev, "Primary bus is hard wired to 0\n");
1313 		primary = bus->number;
1314 	}
1315 
1316 	/* Check if setup is sensible at all */
1317 	if (!pass &&
1318 	    (primary != bus->number || secondary <= bus->number ||
1319 	     secondary > subordinate)) {
1320 		pci_info(dev, "bridge configuration invalid ([bus %02x-%02x]), reconfiguring\n",
1321 			 secondary, subordinate);
1322 		broken = 1;
1323 	}
1324 
1325 	/*
1326 	 * Disable Master-Abort Mode during probing to avoid reporting of
1327 	 * bus errors in some architectures.
1328 	 */
1329 	pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &bctl);
1330 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL,
1331 			      bctl & ~PCI_BRIDGE_CTL_MASTER_ABORT);
1332 
1333 	pci_enable_rrs_sv(dev);
1334 
1335 	if ((secondary || subordinate) && !pcibios_assign_all_busses() &&
1336 	    !is_cardbus && !broken) {
1337 		unsigned int cmax, buses;
1338 
1339 		/*
1340 		 * Bus already configured by firmware, process it in the
1341 		 * first pass and just note the configuration.
1342 		 */
1343 		if (pass)
1344 			goto out;
1345 
1346 		/*
1347 		 * The bus might already exist for two reasons: Either we
1348 		 * are rescanning the bus or the bus is reachable through
1349 		 * more than one bridge. The second case can happen with
1350 		 * the i450NX chipset.
1351 		 */
1352 		child = pci_find_bus(pci_domain_nr(bus), secondary);
1353 		if (!child) {
1354 			child = pci_add_new_bus(bus, dev, secondary);
1355 			if (!child)
1356 				goto out;
1357 			child->primary = primary;
1358 			pci_bus_insert_busn_res(child, secondary, subordinate);
1359 			child->bridge_ctl = bctl;
1360 		}
1361 
1362 		buses = subordinate - secondary;
1363 		cmax = pci_scan_child_bus_extend(child, buses);
1364 		if (cmax > subordinate)
1365 			pci_warn(dev, "bridge has subordinate %02x but max busn %02x\n",
1366 				 subordinate, cmax);
1367 
1368 		/* Subordinate should equal child->busn_res.end */
1369 		if (subordinate > max)
1370 			max = subordinate;
1371 	} else {
1372 
1373 		/*
1374 		 * We need to assign a number to this bus which we always
1375 		 * do in the second pass.
1376 		 */
1377 		if (!pass) {
1378 			if (pcibios_assign_all_busses() || broken || is_cardbus)
1379 
1380 				/*
1381 				 * Temporarily disable forwarding of the
1382 				 * configuration cycles on all bridges in
1383 				 * this bus segment to avoid possible
1384 				 * conflicts in the second pass between two
1385 				 * bridges programmed with overlapping bus
1386 				 * ranges.
1387 				 */
1388 				pci_write_config_dword(dev, PCI_PRIMARY_BUS,
1389 						       buses & ~0xffffff);
1390 			goto out;
1391 		}
1392 
1393 		/* Clear errors */
1394 		pci_write_config_word(dev, PCI_STATUS, 0xffff);
1395 
1396 		/* Read bus numbers from EA Capability (if present) */
1397 		fixed_buses = pci_ea_fixed_busnrs(dev, &fixed_sec, &fixed_sub);
1398 		if (fixed_buses)
1399 			next_busnr = fixed_sec;
1400 		else
1401 			next_busnr = max + 1;
1402 
1403 		/*
1404 		 * Prevent assigning a bus number that already exists.
1405 		 * This can happen when a bridge is hot-plugged, so in this
1406 		 * case we only re-scan this bus.
1407 		 */
1408 		child = pci_find_bus(pci_domain_nr(bus), next_busnr);
1409 		if (!child) {
1410 			child = pci_add_new_bus(bus, dev, next_busnr);
1411 			if (!child)
1412 				goto out;
1413 			pci_bus_insert_busn_res(child, next_busnr,
1414 						bus->busn_res.end);
1415 		}
1416 		max++;
1417 		if (available_buses)
1418 			available_buses--;
1419 
1420 		buses = (buses & 0xff000000)
1421 		      | ((unsigned int)(child->primary)     <<  0)
1422 		      | ((unsigned int)(child->busn_res.start)   <<  8)
1423 		      | ((unsigned int)(child->busn_res.end) << 16);
1424 
1425 		/*
1426 		 * yenta.c forces a secondary latency timer of 176.
1427 		 * Copy that behaviour here.
1428 		 */
1429 		if (is_cardbus) {
1430 			buses &= ~0xff000000;
1431 			buses |= CARDBUS_LATENCY_TIMER << 24;
1432 		}
1433 
1434 		/* We need to blast all three values with a single write */
1435 		pci_write_config_dword(dev, PCI_PRIMARY_BUS, buses);
1436 
1437 		if (!is_cardbus) {
1438 			child->bridge_ctl = bctl;
1439 			max = pci_scan_child_bus_extend(child, available_buses);
1440 		} else {
1441 
1442 			/*
1443 			 * For CardBus bridges, we leave 4 bus numbers as
1444 			 * cards with a PCI-to-PCI bridge can be inserted
1445 			 * later.
1446 			 */
1447 			for (i = 0; i < CARDBUS_RESERVE_BUSNR; i++) {
1448 				struct pci_bus *parent = bus;
1449 				if (pci_find_bus(pci_domain_nr(bus),
1450 							max+i+1))
1451 					break;
1452 				while (parent->parent) {
1453 					if ((!pcibios_assign_all_busses()) &&
1454 					    (parent->busn_res.end > max) &&
1455 					    (parent->busn_res.end <= max+i)) {
1456 						j = 1;
1457 					}
1458 					parent = parent->parent;
1459 				}
1460 				if (j) {
1461 
1462 					/*
1463 					 * Often, there are two CardBus
1464 					 * bridges -- try to leave one
1465 					 * valid bus number for each one.
1466 					 */
1467 					i /= 2;
1468 					break;
1469 				}
1470 			}
1471 			max += i;
1472 		}
1473 
1474 		/*
1475 		 * Set subordinate bus number to its real value.
1476 		 * If fixed subordinate bus number exists from EA
1477 		 * capability then use it.
1478 		 */
1479 		if (fixed_buses)
1480 			max = fixed_sub;
1481 		pci_bus_update_busn_res_end(child, max);
1482 		pci_write_config_byte(dev, PCI_SUBORDINATE_BUS, max);
1483 	}
1484 
1485 	sprintf(child->name,
1486 		(is_cardbus ? "PCI CardBus %04x:%02x" : "PCI Bus %04x:%02x"),
1487 		pci_domain_nr(bus), child->number);
1488 
1489 	/* Check that all devices are accessible */
1490 	while (bus->parent) {
1491 		if ((child->busn_res.end > bus->busn_res.end) ||
1492 		    (child->number > bus->busn_res.end) ||
1493 		    (child->number < bus->number) ||
1494 		    (child->busn_res.end < bus->number)) {
1495 			dev_info(&dev->dev, "devices behind bridge are unusable because %pR cannot be assigned for them\n",
1496 				 &child->busn_res);
1497 			break;
1498 		}
1499 		bus = bus->parent;
1500 	}
1501 
1502 out:
1503 	/* Clear errors in the Secondary Status Register */
1504 	pci_write_config_word(dev, PCI_SEC_STATUS, 0xffff);
1505 
1506 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL, bctl);
1507 
1508 	pm_runtime_put(&dev->dev);
1509 
1510 	return max;
1511 }
1512 
1513 /*
1514  * pci_scan_bridge() - Scan buses behind a bridge
1515  * @bus: Parent bus the bridge is on
1516  * @dev: Bridge itself
1517  * @max: Starting subordinate number of buses behind this bridge
1518  * @pass: Either %0 (scan already configured bridges) or %1 (scan bridges
1519  *        that need to be reconfigured.
1520  *
1521  * If it's a bridge, configure it and scan the bus behind it.
1522  * For CardBus bridges, we don't scan behind as the devices will
1523  * be handled by the bridge driver itself.
1524  *
1525  * We need to process bridges in two passes -- first we scan those
1526  * already configured by the BIOS and after we are done with all of
1527  * them, we proceed to assigning numbers to the remaining buses in
1528  * order to avoid overlaps between old and new bus numbers.
1529  *
1530  * Return: New subordinate number covering all buses behind this bridge.
1531  */
1532 int pci_scan_bridge(struct pci_bus *bus, struct pci_dev *dev, int max, int pass)
1533 {
1534 	return pci_scan_bridge_extend(bus, dev, max, 0, pass);
1535 }
1536 EXPORT_SYMBOL(pci_scan_bridge);
1537 
1538 /*
1539  * Read interrupt line and base address registers.
1540  * The architecture-dependent code can tweak these, of course.
1541  */
1542 static void pci_read_irq(struct pci_dev *dev)
1543 {
1544 	unsigned char irq;
1545 
1546 	/* VFs are not allowed to use INTx, so skip the config reads */
1547 	if (dev->is_virtfn) {
1548 		dev->pin = 0;
1549 		dev->irq = 0;
1550 		return;
1551 	}
1552 
1553 	pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &irq);
1554 	dev->pin = irq;
1555 	if (irq)
1556 		pci_read_config_byte(dev, PCI_INTERRUPT_LINE, &irq);
1557 	dev->irq = irq;
1558 }
1559 
1560 void set_pcie_port_type(struct pci_dev *pdev)
1561 {
1562 	int pos;
1563 	u16 reg16;
1564 	u32 reg32;
1565 	int type;
1566 	struct pci_dev *parent;
1567 
1568 	pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
1569 	if (!pos)
1570 		return;
1571 
1572 	pdev->pcie_cap = pos;
1573 	pci_read_config_word(pdev, pos + PCI_EXP_FLAGS, &reg16);
1574 	pdev->pcie_flags_reg = reg16;
1575 	pci_read_config_dword(pdev, pos + PCI_EXP_DEVCAP, &pdev->devcap);
1576 	pdev->pcie_mpss = FIELD_GET(PCI_EXP_DEVCAP_PAYLOAD, pdev->devcap);
1577 
1578 	pcie_capability_read_dword(pdev, PCI_EXP_LNKCAP, &reg32);
1579 	if (reg32 & PCI_EXP_LNKCAP_DLLLARC)
1580 		pdev->link_active_reporting = 1;
1581 
1582 	parent = pci_upstream_bridge(pdev);
1583 	if (!parent)
1584 		return;
1585 
1586 	/*
1587 	 * Some systems do not identify their upstream/downstream ports
1588 	 * correctly so detect impossible configurations here and correct
1589 	 * the port type accordingly.
1590 	 */
1591 	type = pci_pcie_type(pdev);
1592 	if (type == PCI_EXP_TYPE_DOWNSTREAM) {
1593 		/*
1594 		 * If pdev claims to be downstream port but the parent
1595 		 * device is also downstream port assume pdev is actually
1596 		 * upstream port.
1597 		 */
1598 		if (pcie_downstream_port(parent)) {
1599 			pci_info(pdev, "claims to be downstream port but is acting as upstream port, correcting type\n");
1600 			pdev->pcie_flags_reg &= ~PCI_EXP_FLAGS_TYPE;
1601 			pdev->pcie_flags_reg |= PCI_EXP_TYPE_UPSTREAM;
1602 		}
1603 	} else if (type == PCI_EXP_TYPE_UPSTREAM) {
1604 		/*
1605 		 * If pdev claims to be upstream port but the parent
1606 		 * device is also upstream port assume pdev is actually
1607 		 * downstream port.
1608 		 */
1609 		if (pci_pcie_type(parent) == PCI_EXP_TYPE_UPSTREAM) {
1610 			pci_info(pdev, "claims to be upstream port but is acting as downstream port, correcting type\n");
1611 			pdev->pcie_flags_reg &= ~PCI_EXP_FLAGS_TYPE;
1612 			pdev->pcie_flags_reg |= PCI_EXP_TYPE_DOWNSTREAM;
1613 		}
1614 	}
1615 }
1616 
1617 void set_pcie_hotplug_bridge(struct pci_dev *pdev)
1618 {
1619 	u32 reg32;
1620 
1621 	pcie_capability_read_dword(pdev, PCI_EXP_SLTCAP, &reg32);
1622 	if (reg32 & PCI_EXP_SLTCAP_HPC)
1623 		pdev->is_hotplug_bridge = 1;
1624 }
1625 
1626 static void set_pcie_thunderbolt(struct pci_dev *dev)
1627 {
1628 	u16 vsec;
1629 
1630 	/* Is the device part of a Thunderbolt controller? */
1631 	vsec = pci_find_vsec_capability(dev, PCI_VENDOR_ID_INTEL, PCI_VSEC_ID_INTEL_TBT);
1632 	if (vsec)
1633 		dev->is_thunderbolt = 1;
1634 }
1635 
1636 static void set_pcie_untrusted(struct pci_dev *dev)
1637 {
1638 	struct pci_dev *parent = pci_upstream_bridge(dev);
1639 
1640 	if (!parent)
1641 		return;
1642 	/*
1643 	 * If the upstream bridge is untrusted we treat this device as
1644 	 * untrusted as well.
1645 	 */
1646 	if (parent->untrusted) {
1647 		dev->untrusted = true;
1648 		return;
1649 	}
1650 
1651 	if (arch_pci_dev_is_removable(dev)) {
1652 		pci_dbg(dev, "marking as untrusted\n");
1653 		dev->untrusted = true;
1654 	}
1655 }
1656 
1657 static void pci_set_removable(struct pci_dev *dev)
1658 {
1659 	struct pci_dev *parent = pci_upstream_bridge(dev);
1660 
1661 	if (!parent)
1662 		return;
1663 	/*
1664 	 * We (only) consider everything tunneled below an external_facing
1665 	 * device to be removable by the user. We're mainly concerned with
1666 	 * consumer platforms with user accessible thunderbolt ports that are
1667 	 * vulnerable to DMA attacks, and we expect those ports to be marked by
1668 	 * the firmware as external_facing. Devices in traditional hotplug
1669 	 * slots can technically be removed, but the expectation is that unless
1670 	 * the port is marked with external_facing, such devices are less
1671 	 * accessible to user / may not be removed by end user, and thus not
1672 	 * exposed as "removable" to userspace.
1673 	 */
1674 	if (dev_is_removable(&parent->dev)) {
1675 		dev_set_removable(&dev->dev, DEVICE_REMOVABLE);
1676 		return;
1677 	}
1678 
1679 	if (arch_pci_dev_is_removable(dev)) {
1680 		pci_dbg(dev, "marking as removable\n");
1681 		dev_set_removable(&dev->dev, DEVICE_REMOVABLE);
1682 	}
1683 }
1684 
1685 /**
1686  * pci_ext_cfg_is_aliased - Is ext config space just an alias of std config?
1687  * @dev: PCI device
1688  *
1689  * PCI Express to PCI/PCI-X Bridge Specification, rev 1.0, 4.1.4 says that
1690  * when forwarding a type1 configuration request the bridge must check that
1691  * the extended register address field is zero.  The bridge is not permitted
1692  * to forward the transactions and must handle it as an Unsupported Request.
1693  * Some bridges do not follow this rule and simply drop the extended register
1694  * bits, resulting in the standard config space being aliased, every 256
1695  * bytes across the entire configuration space.  Test for this condition by
1696  * comparing the first dword of each potential alias to the vendor/device ID.
1697  * Known offenders:
1698  *   ASM1083/1085 PCIe-to-PCI Reversible Bridge (1b21:1080, rev 01 & 03)
1699  *   AMD/ATI SBx00 PCI to PCI Bridge (1002:4384, rev 40)
1700  */
1701 static bool pci_ext_cfg_is_aliased(struct pci_dev *dev)
1702 {
1703 #ifdef CONFIG_PCI_QUIRKS
1704 	int pos, ret;
1705 	u32 header, tmp;
1706 
1707 	pci_read_config_dword(dev, PCI_VENDOR_ID, &header);
1708 
1709 	for (pos = PCI_CFG_SPACE_SIZE;
1710 	     pos < PCI_CFG_SPACE_EXP_SIZE; pos += PCI_CFG_SPACE_SIZE) {
1711 		ret = pci_read_config_dword(dev, pos, &tmp);
1712 		if ((ret != PCIBIOS_SUCCESSFUL) || (header != tmp))
1713 			return false;
1714 	}
1715 
1716 	return true;
1717 #else
1718 	return false;
1719 #endif
1720 }
1721 
1722 /**
1723  * pci_cfg_space_size_ext - Get the configuration space size of the PCI device
1724  * @dev: PCI device
1725  *
1726  * Regular PCI devices have 256 bytes, but PCI-X 2 and PCI Express devices
1727  * have 4096 bytes.  Even if the device is capable, that doesn't mean we can
1728  * access it.  Maybe we don't have a way to generate extended config space
1729  * accesses, or the device is behind a reverse Express bridge.  So we try
1730  * reading the dword at 0x100 which must either be 0 or a valid extended
1731  * capability header.
1732  */
1733 static int pci_cfg_space_size_ext(struct pci_dev *dev)
1734 {
1735 	u32 status;
1736 	int pos = PCI_CFG_SPACE_SIZE;
1737 
1738 	if (pci_read_config_dword(dev, pos, &status) != PCIBIOS_SUCCESSFUL)
1739 		return PCI_CFG_SPACE_SIZE;
1740 	if (PCI_POSSIBLE_ERROR(status) || pci_ext_cfg_is_aliased(dev))
1741 		return PCI_CFG_SPACE_SIZE;
1742 
1743 	return PCI_CFG_SPACE_EXP_SIZE;
1744 }
1745 
1746 int pci_cfg_space_size(struct pci_dev *dev)
1747 {
1748 	int pos;
1749 	u32 status;
1750 	u16 class;
1751 
1752 #ifdef CONFIG_PCI_IOV
1753 	/*
1754 	 * Per the SR-IOV specification (rev 1.1, sec 3.5), VFs are required to
1755 	 * implement a PCIe capability and therefore must implement extended
1756 	 * config space.  We can skip the NO_EXTCFG test below and the
1757 	 * reachability/aliasing test in pci_cfg_space_size_ext() by virtue of
1758 	 * the fact that the SR-IOV capability on the PF resides in extended
1759 	 * config space and must be accessible and non-aliased to have enabled
1760 	 * support for this VF.  This is a micro performance optimization for
1761 	 * systems supporting many VFs.
1762 	 */
1763 	if (dev->is_virtfn)
1764 		return PCI_CFG_SPACE_EXP_SIZE;
1765 #endif
1766 
1767 	if (dev->bus->bus_flags & PCI_BUS_FLAGS_NO_EXTCFG)
1768 		return PCI_CFG_SPACE_SIZE;
1769 
1770 	class = dev->class >> 8;
1771 	if (class == PCI_CLASS_BRIDGE_HOST)
1772 		return pci_cfg_space_size_ext(dev);
1773 
1774 	if (pci_is_pcie(dev))
1775 		return pci_cfg_space_size_ext(dev);
1776 
1777 	pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1778 	if (!pos)
1779 		return PCI_CFG_SPACE_SIZE;
1780 
1781 	pci_read_config_dword(dev, pos + PCI_X_STATUS, &status);
1782 	if (status & (PCI_X_STATUS_266MHZ | PCI_X_STATUS_533MHZ))
1783 		return pci_cfg_space_size_ext(dev);
1784 
1785 	return PCI_CFG_SPACE_SIZE;
1786 }
1787 
1788 static u32 pci_class(struct pci_dev *dev)
1789 {
1790 	u32 class;
1791 
1792 #ifdef CONFIG_PCI_IOV
1793 	if (dev->is_virtfn)
1794 		return dev->physfn->sriov->class;
1795 #endif
1796 	pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
1797 	return class;
1798 }
1799 
1800 static void pci_subsystem_ids(struct pci_dev *dev, u16 *vendor, u16 *device)
1801 {
1802 #ifdef CONFIG_PCI_IOV
1803 	if (dev->is_virtfn) {
1804 		*vendor = dev->physfn->sriov->subsystem_vendor;
1805 		*device = dev->physfn->sriov->subsystem_device;
1806 		return;
1807 	}
1808 #endif
1809 	pci_read_config_word(dev, PCI_SUBSYSTEM_VENDOR_ID, vendor);
1810 	pci_read_config_word(dev, PCI_SUBSYSTEM_ID, device);
1811 }
1812 
1813 static u8 pci_hdr_type(struct pci_dev *dev)
1814 {
1815 	u8 hdr_type;
1816 
1817 #ifdef CONFIG_PCI_IOV
1818 	if (dev->is_virtfn)
1819 		return dev->physfn->sriov->hdr_type;
1820 #endif
1821 	pci_read_config_byte(dev, PCI_HEADER_TYPE, &hdr_type);
1822 	return hdr_type;
1823 }
1824 
1825 #define LEGACY_IO_RESOURCE	(IORESOURCE_IO | IORESOURCE_PCI_FIXED)
1826 
1827 /**
1828  * pci_intx_mask_broken - Test PCI_COMMAND_INTX_DISABLE writability
1829  * @dev: PCI device
1830  *
1831  * Test whether PCI_COMMAND_INTX_DISABLE is writable for @dev.  Check this
1832  * at enumeration-time to avoid modifying PCI_COMMAND at run-time.
1833  */
1834 static int pci_intx_mask_broken(struct pci_dev *dev)
1835 {
1836 	u16 orig, toggle, new;
1837 
1838 	pci_read_config_word(dev, PCI_COMMAND, &orig);
1839 	toggle = orig ^ PCI_COMMAND_INTX_DISABLE;
1840 	pci_write_config_word(dev, PCI_COMMAND, toggle);
1841 	pci_read_config_word(dev, PCI_COMMAND, &new);
1842 
1843 	pci_write_config_word(dev, PCI_COMMAND, orig);
1844 
1845 	/*
1846 	 * PCI_COMMAND_INTX_DISABLE was reserved and read-only prior to PCI
1847 	 * r2.3, so strictly speaking, a device is not *broken* if it's not
1848 	 * writable.  But we'll live with the misnomer for now.
1849 	 */
1850 	if (new != toggle)
1851 		return 1;
1852 	return 0;
1853 }
1854 
1855 static void early_dump_pci_device(struct pci_dev *pdev)
1856 {
1857 	u32 value[256 / 4];
1858 	int i;
1859 
1860 	pci_info(pdev, "config space:\n");
1861 
1862 	for (i = 0; i < 256; i += 4)
1863 		pci_read_config_dword(pdev, i, &value[i / 4]);
1864 
1865 	print_hex_dump(KERN_INFO, "", DUMP_PREFIX_OFFSET, 16, 1,
1866 		       value, 256, false);
1867 }
1868 
1869 static const char *pci_type_str(struct pci_dev *dev)
1870 {
1871 	static const char * const str[] = {
1872 		"PCIe Endpoint",
1873 		"PCIe Legacy Endpoint",
1874 		"PCIe unknown",
1875 		"PCIe unknown",
1876 		"PCIe Root Port",
1877 		"PCIe Switch Upstream Port",
1878 		"PCIe Switch Downstream Port",
1879 		"PCIe to PCI/PCI-X bridge",
1880 		"PCI/PCI-X to PCIe bridge",
1881 		"PCIe Root Complex Integrated Endpoint",
1882 		"PCIe Root Complex Event Collector",
1883 	};
1884 	int type;
1885 
1886 	if (pci_is_pcie(dev)) {
1887 		type = pci_pcie_type(dev);
1888 		if (type < ARRAY_SIZE(str))
1889 			return str[type];
1890 
1891 		return "PCIe unknown";
1892 	}
1893 
1894 	switch (dev->hdr_type) {
1895 	case PCI_HEADER_TYPE_NORMAL:
1896 		return "conventional PCI endpoint";
1897 	case PCI_HEADER_TYPE_BRIDGE:
1898 		return "conventional PCI bridge";
1899 	case PCI_HEADER_TYPE_CARDBUS:
1900 		return "CardBus bridge";
1901 	default:
1902 		return "conventional PCI";
1903 	}
1904 }
1905 
1906 /**
1907  * pci_setup_device - Fill in class and map information of a device
1908  * @dev: the device structure to fill
1909  *
1910  * Initialize the device structure with information about the device's
1911  * vendor,class,memory and IO-space addresses, IRQ lines etc.
1912  * Called at initialisation of the PCI subsystem and by CardBus services.
1913  * Returns 0 on success and negative if unknown type of device (not normal,
1914  * bridge or CardBus).
1915  */
1916 int pci_setup_device(struct pci_dev *dev)
1917 {
1918 	u32 class;
1919 	u16 cmd;
1920 	u8 hdr_type;
1921 	int err, pos = 0;
1922 	struct pci_bus_region region;
1923 	struct resource *res;
1924 
1925 	hdr_type = pci_hdr_type(dev);
1926 
1927 	dev->sysdata = dev->bus->sysdata;
1928 	dev->dev.parent = dev->bus->bridge;
1929 	dev->dev.bus = &pci_bus_type;
1930 	dev->hdr_type = hdr_type & 0x7f;
1931 	dev->multifunction = !!(hdr_type & 0x80);
1932 	dev->error_state = pci_channel_io_normal;
1933 	set_pcie_port_type(dev);
1934 
1935 	err = pci_set_of_node(dev);
1936 	if (err)
1937 		return err;
1938 	pci_set_acpi_fwnode(dev);
1939 
1940 	pci_dev_assign_slot(dev);
1941 
1942 	/*
1943 	 * Assume 32-bit PCI; let 64-bit PCI cards (which are far rarer)
1944 	 * set this higher, assuming the system even supports it.
1945 	 */
1946 	dev->dma_mask = 0xffffffff;
1947 
1948 	dev_set_name(&dev->dev, "%04x:%02x:%02x.%d", pci_domain_nr(dev->bus),
1949 		     dev->bus->number, PCI_SLOT(dev->devfn),
1950 		     PCI_FUNC(dev->devfn));
1951 
1952 	class = pci_class(dev);
1953 
1954 	dev->revision = class & 0xff;
1955 	dev->class = class >> 8;		    /* upper 3 bytes */
1956 
1957 	if (pci_early_dump)
1958 		early_dump_pci_device(dev);
1959 
1960 	/* Need to have dev->class ready */
1961 	dev->cfg_size = pci_cfg_space_size(dev);
1962 
1963 	/* Need to have dev->cfg_size ready */
1964 	set_pcie_thunderbolt(dev);
1965 
1966 	set_pcie_untrusted(dev);
1967 
1968 	if (pci_is_pcie(dev))
1969 		dev->supported_speeds = pcie_get_supported_speeds(dev);
1970 
1971 	/* "Unknown power state" */
1972 	dev->current_state = PCI_UNKNOWN;
1973 
1974 	/* Early fixups, before probing the BARs */
1975 	pci_fixup_device(pci_fixup_early, dev);
1976 
1977 	pci_set_removable(dev);
1978 
1979 	pci_info(dev, "[%04x:%04x] type %02x class %#08x %s\n",
1980 		 dev->vendor, dev->device, dev->hdr_type, dev->class,
1981 		 pci_type_str(dev));
1982 
1983 	/* Device class may be changed after fixup */
1984 	class = dev->class >> 8;
1985 
1986 	if (dev->non_compliant_bars && !dev->mmio_always_on) {
1987 		pci_read_config_word(dev, PCI_COMMAND, &cmd);
1988 		if (cmd & (PCI_COMMAND_IO | PCI_COMMAND_MEMORY)) {
1989 			pci_info(dev, "device has non-compliant BARs; disabling IO/MEM decoding\n");
1990 			cmd &= ~PCI_COMMAND_IO;
1991 			cmd &= ~PCI_COMMAND_MEMORY;
1992 			pci_write_config_word(dev, PCI_COMMAND, cmd);
1993 		}
1994 	}
1995 
1996 	dev->broken_intx_masking = pci_intx_mask_broken(dev);
1997 
1998 	switch (dev->hdr_type) {		    /* header type */
1999 	case PCI_HEADER_TYPE_NORMAL:		    /* standard header */
2000 		if (class == PCI_CLASS_BRIDGE_PCI)
2001 			goto bad;
2002 		pci_read_irq(dev);
2003 		pci_read_bases(dev, 6, PCI_ROM_ADDRESS);
2004 
2005 		pci_subsystem_ids(dev, &dev->subsystem_vendor, &dev->subsystem_device);
2006 
2007 		/*
2008 		 * Do the ugly legacy mode stuff here rather than broken chip
2009 		 * quirk code. Legacy mode ATA controllers have fixed
2010 		 * addresses. These are not always echoed in BAR0-3, and
2011 		 * BAR0-3 in a few cases contain junk!
2012 		 */
2013 		if (class == PCI_CLASS_STORAGE_IDE) {
2014 			u8 progif;
2015 			pci_read_config_byte(dev, PCI_CLASS_PROG, &progif);
2016 			if ((progif & 1) == 0) {
2017 				region.start = 0x1F0;
2018 				region.end = 0x1F7;
2019 				res = &dev->resource[0];
2020 				res->flags = LEGACY_IO_RESOURCE;
2021 				pcibios_bus_to_resource(dev->bus, res, &region);
2022 				pci_info(dev, "BAR 0 %pR: legacy IDE quirk\n",
2023 					 res);
2024 				region.start = 0x3F6;
2025 				region.end = 0x3F6;
2026 				res = &dev->resource[1];
2027 				res->flags = LEGACY_IO_RESOURCE;
2028 				pcibios_bus_to_resource(dev->bus, res, &region);
2029 				pci_info(dev, "BAR 1 %pR: legacy IDE quirk\n",
2030 					 res);
2031 			}
2032 			if ((progif & 4) == 0) {
2033 				region.start = 0x170;
2034 				region.end = 0x177;
2035 				res = &dev->resource[2];
2036 				res->flags = LEGACY_IO_RESOURCE;
2037 				pcibios_bus_to_resource(dev->bus, res, &region);
2038 				pci_info(dev, "BAR 2 %pR: legacy IDE quirk\n",
2039 					 res);
2040 				region.start = 0x376;
2041 				region.end = 0x376;
2042 				res = &dev->resource[3];
2043 				res->flags = LEGACY_IO_RESOURCE;
2044 				pcibios_bus_to_resource(dev->bus, res, &region);
2045 				pci_info(dev, "BAR 3 %pR: legacy IDE quirk\n",
2046 					 res);
2047 			}
2048 		}
2049 		break;
2050 
2051 	case PCI_HEADER_TYPE_BRIDGE:		    /* bridge header */
2052 		/*
2053 		 * The PCI-to-PCI bridge spec requires that subtractive
2054 		 * decoding (i.e. transparent) bridge must have programming
2055 		 * interface code of 0x01.
2056 		 */
2057 		pci_read_irq(dev);
2058 		dev->transparent = ((dev->class & 0xff) == 1);
2059 		pci_read_bases(dev, 2, PCI_ROM_ADDRESS1);
2060 		pci_read_bridge_windows(dev);
2061 		set_pcie_hotplug_bridge(dev);
2062 		pos = pci_find_capability(dev, PCI_CAP_ID_SSVID);
2063 		if (pos) {
2064 			pci_read_config_word(dev, pos + PCI_SSVID_VENDOR_ID, &dev->subsystem_vendor);
2065 			pci_read_config_word(dev, pos + PCI_SSVID_DEVICE_ID, &dev->subsystem_device);
2066 		}
2067 		break;
2068 
2069 	case PCI_HEADER_TYPE_CARDBUS:		    /* CardBus bridge header */
2070 		if (class != PCI_CLASS_BRIDGE_CARDBUS)
2071 			goto bad;
2072 		pci_read_irq(dev);
2073 		pci_read_bases(dev, 1, 0);
2074 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_VENDOR_ID, &dev->subsystem_vendor);
2075 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_ID, &dev->subsystem_device);
2076 		break;
2077 
2078 	default:				    /* unknown header */
2079 		pci_err(dev, "unknown header type %02x, ignoring device\n",
2080 			dev->hdr_type);
2081 		pci_release_of_node(dev);
2082 		return -EIO;
2083 
2084 	bad:
2085 		pci_err(dev, "ignoring class %#08x (doesn't match header type %02x)\n",
2086 			dev->class, dev->hdr_type);
2087 		dev->class = PCI_CLASS_NOT_DEFINED << 8;
2088 	}
2089 
2090 	/* We found a fine healthy device, go go go... */
2091 	return 0;
2092 }
2093 
2094 static void pci_configure_mps(struct pci_dev *dev)
2095 {
2096 	struct pci_dev *bridge = pci_upstream_bridge(dev);
2097 	int mps, mpss, p_mps, rc;
2098 
2099 	if (!pci_is_pcie(dev))
2100 		return;
2101 
2102 	/* MPS and MRRS fields are of type 'RsvdP' for VFs, short-circuit out */
2103 	if (dev->is_virtfn)
2104 		return;
2105 
2106 	/*
2107 	 * For Root Complex Integrated Endpoints, program the maximum
2108 	 * supported value unless limited by the PCIE_BUS_PEER2PEER case.
2109 	 */
2110 	if (pci_pcie_type(dev) == PCI_EXP_TYPE_RC_END) {
2111 		if (pcie_bus_config == PCIE_BUS_PEER2PEER)
2112 			mps = 128;
2113 		else
2114 			mps = 128 << dev->pcie_mpss;
2115 		rc = pcie_set_mps(dev, mps);
2116 		if (rc) {
2117 			pci_warn(dev, "can't set Max Payload Size to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
2118 				 mps);
2119 		}
2120 		return;
2121 	}
2122 
2123 	if (!bridge || !pci_is_pcie(bridge))
2124 		return;
2125 
2126 	mps = pcie_get_mps(dev);
2127 	p_mps = pcie_get_mps(bridge);
2128 
2129 	if (mps == p_mps)
2130 		return;
2131 
2132 	if (pcie_bus_config == PCIE_BUS_TUNE_OFF) {
2133 		pci_warn(dev, "Max Payload Size %d, but upstream %s set to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
2134 			 mps, pci_name(bridge), p_mps);
2135 		return;
2136 	}
2137 
2138 	/*
2139 	 * Fancier MPS configuration is done later by
2140 	 * pcie_bus_configure_settings()
2141 	 */
2142 	if (pcie_bus_config != PCIE_BUS_DEFAULT)
2143 		return;
2144 
2145 	mpss = 128 << dev->pcie_mpss;
2146 	if (mpss < p_mps && pci_pcie_type(bridge) == PCI_EXP_TYPE_ROOT_PORT) {
2147 		pcie_set_mps(bridge, mpss);
2148 		pci_info(dev, "Upstream bridge's Max Payload Size set to %d (was %d, max %d)\n",
2149 			 mpss, p_mps, 128 << bridge->pcie_mpss);
2150 		p_mps = pcie_get_mps(bridge);
2151 	}
2152 
2153 	rc = pcie_set_mps(dev, p_mps);
2154 	if (rc) {
2155 		pci_warn(dev, "can't set Max Payload Size to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
2156 			 p_mps);
2157 		return;
2158 	}
2159 
2160 	pci_info(dev, "Max Payload Size set to %d (was %d, max %d)\n",
2161 		 p_mps, mps, mpss);
2162 }
2163 
2164 int pci_configure_extended_tags(struct pci_dev *dev, void *ign)
2165 {
2166 	struct pci_host_bridge *host;
2167 	u32 cap;
2168 	u16 ctl;
2169 	int ret;
2170 
2171 	if (!pci_is_pcie(dev))
2172 		return 0;
2173 
2174 	ret = pcie_capability_read_dword(dev, PCI_EXP_DEVCAP, &cap);
2175 	if (ret)
2176 		return 0;
2177 
2178 	if (!(cap & PCI_EXP_DEVCAP_EXT_TAG))
2179 		return 0;
2180 
2181 	ret = pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &ctl);
2182 	if (ret)
2183 		return 0;
2184 
2185 	host = pci_find_host_bridge(dev->bus);
2186 	if (!host)
2187 		return 0;
2188 
2189 	/*
2190 	 * If some device in the hierarchy doesn't handle Extended Tags
2191 	 * correctly, make sure they're disabled.
2192 	 */
2193 	if (host->no_ext_tags) {
2194 		if (ctl & PCI_EXP_DEVCTL_EXT_TAG) {
2195 			pci_info(dev, "disabling Extended Tags\n");
2196 			pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
2197 						   PCI_EXP_DEVCTL_EXT_TAG);
2198 		}
2199 		return 0;
2200 	}
2201 
2202 	if (!(ctl & PCI_EXP_DEVCTL_EXT_TAG)) {
2203 		pci_info(dev, "enabling Extended Tags\n");
2204 		pcie_capability_set_word(dev, PCI_EXP_DEVCTL,
2205 					 PCI_EXP_DEVCTL_EXT_TAG);
2206 	}
2207 	return 0;
2208 }
2209 
2210 /**
2211  * pcie_relaxed_ordering_enabled - Probe for PCIe relaxed ordering enable
2212  * @dev: PCI device to query
2213  *
2214  * Returns true if the device has enabled relaxed ordering attribute.
2215  */
2216 bool pcie_relaxed_ordering_enabled(struct pci_dev *dev)
2217 {
2218 	u16 v;
2219 
2220 	pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &v);
2221 
2222 	return !!(v & PCI_EXP_DEVCTL_RELAX_EN);
2223 }
2224 EXPORT_SYMBOL(pcie_relaxed_ordering_enabled);
2225 
2226 static void pci_configure_relaxed_ordering(struct pci_dev *dev)
2227 {
2228 	struct pci_dev *root;
2229 
2230 	/* PCI_EXP_DEVCTL_RELAX_EN is RsvdP in VFs */
2231 	if (dev->is_virtfn)
2232 		return;
2233 
2234 	if (!pcie_relaxed_ordering_enabled(dev))
2235 		return;
2236 
2237 	/*
2238 	 * For now, we only deal with Relaxed Ordering issues with Root
2239 	 * Ports. Peer-to-Peer DMA is another can of worms.
2240 	 */
2241 	root = pcie_find_root_port(dev);
2242 	if (!root)
2243 		return;
2244 
2245 	if (root->dev_flags & PCI_DEV_FLAGS_NO_RELAXED_ORDERING) {
2246 		pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
2247 					   PCI_EXP_DEVCTL_RELAX_EN);
2248 		pci_info(dev, "Relaxed Ordering disabled because the Root Port didn't support it\n");
2249 	}
2250 }
2251 
2252 static void pci_configure_eetlp_prefix(struct pci_dev *dev)
2253 {
2254 #ifdef CONFIG_PCI_PASID
2255 	struct pci_dev *bridge;
2256 	int pcie_type;
2257 	u32 cap;
2258 
2259 	if (!pci_is_pcie(dev))
2260 		return;
2261 
2262 	pcie_capability_read_dword(dev, PCI_EXP_DEVCAP2, &cap);
2263 	if (!(cap & PCI_EXP_DEVCAP2_EE_PREFIX))
2264 		return;
2265 
2266 	pcie_type = pci_pcie_type(dev);
2267 	if (pcie_type == PCI_EXP_TYPE_ROOT_PORT ||
2268 	    pcie_type == PCI_EXP_TYPE_RC_END)
2269 		dev->eetlp_prefix_path = 1;
2270 	else {
2271 		bridge = pci_upstream_bridge(dev);
2272 		if (bridge && bridge->eetlp_prefix_path)
2273 			dev->eetlp_prefix_path = 1;
2274 	}
2275 #endif
2276 }
2277 
2278 static void pci_configure_serr(struct pci_dev *dev)
2279 {
2280 	u16 control;
2281 
2282 	if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
2283 
2284 		/*
2285 		 * A bridge will not forward ERR_ messages coming from an
2286 		 * endpoint unless SERR# forwarding is enabled.
2287 		 */
2288 		pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &control);
2289 		if (!(control & PCI_BRIDGE_CTL_SERR)) {
2290 			control |= PCI_BRIDGE_CTL_SERR;
2291 			pci_write_config_word(dev, PCI_BRIDGE_CONTROL, control);
2292 		}
2293 	}
2294 }
2295 
2296 static void pci_configure_device(struct pci_dev *dev)
2297 {
2298 	pci_configure_mps(dev);
2299 	pci_configure_extended_tags(dev, NULL);
2300 	pci_configure_relaxed_ordering(dev);
2301 	pci_configure_ltr(dev);
2302 	pci_configure_aspm_l1ss(dev);
2303 	pci_configure_eetlp_prefix(dev);
2304 	pci_configure_serr(dev);
2305 
2306 	pci_acpi_program_hp_params(dev);
2307 }
2308 
2309 static void pci_release_capabilities(struct pci_dev *dev)
2310 {
2311 	pci_aer_exit(dev);
2312 	pci_rcec_exit(dev);
2313 	pci_iov_release(dev);
2314 	pci_free_cap_save_buffers(dev);
2315 }
2316 
2317 /**
2318  * pci_release_dev - Free a PCI device structure when all users of it are
2319  *		     finished
2320  * @dev: device that's been disconnected
2321  *
2322  * Will be called only by the device core when all users of this PCI device are
2323  * done.
2324  */
2325 static void pci_release_dev(struct device *dev)
2326 {
2327 	struct pci_dev *pci_dev;
2328 
2329 	pci_dev = to_pci_dev(dev);
2330 	pci_release_capabilities(pci_dev);
2331 	pci_release_of_node(pci_dev);
2332 	pcibios_release_device(pci_dev);
2333 	pci_bus_put(pci_dev->bus);
2334 	kfree(pci_dev->driver_override);
2335 	bitmap_free(pci_dev->dma_alias_mask);
2336 	dev_dbg(dev, "device released\n");
2337 	kfree(pci_dev);
2338 }
2339 
2340 static const struct device_type pci_dev_type = {
2341 	.groups = pci_dev_attr_groups,
2342 };
2343 
2344 struct pci_dev *pci_alloc_dev(struct pci_bus *bus)
2345 {
2346 	struct pci_dev *dev;
2347 
2348 	dev = kzalloc(sizeof(struct pci_dev), GFP_KERNEL);
2349 	if (!dev)
2350 		return NULL;
2351 
2352 	INIT_LIST_HEAD(&dev->bus_list);
2353 	dev->dev.type = &pci_dev_type;
2354 	dev->bus = pci_bus_get(bus);
2355 	dev->driver_exclusive_resource = (struct resource) {
2356 		.name = "PCI Exclusive",
2357 		.start = 0,
2358 		.end = -1,
2359 	};
2360 
2361 	spin_lock_init(&dev->pcie_cap_lock);
2362 #ifdef CONFIG_PCI_MSI
2363 	raw_spin_lock_init(&dev->msi_lock);
2364 #endif
2365 	return dev;
2366 }
2367 EXPORT_SYMBOL(pci_alloc_dev);
2368 
2369 static bool pci_bus_wait_rrs(struct pci_bus *bus, int devfn, u32 *l,
2370 			     int timeout)
2371 {
2372 	int delay = 1;
2373 
2374 	if (!pci_bus_rrs_vendor_id(*l))
2375 		return true;	/* not a Configuration RRS completion */
2376 
2377 	if (!timeout)
2378 		return false;	/* RRS, but caller doesn't want to wait */
2379 
2380 	/*
2381 	 * We got the reserved Vendor ID that indicates a completion with
2382 	 * Configuration Request Retry Status (RRS).  Retry until we get a
2383 	 * valid Vendor ID or we time out.
2384 	 */
2385 	while (pci_bus_rrs_vendor_id(*l)) {
2386 		if (delay > timeout) {
2387 			pr_warn("pci %04x:%02x:%02x.%d: not ready after %dms; giving up\n",
2388 				pci_domain_nr(bus), bus->number,
2389 				PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2390 
2391 			return false;
2392 		}
2393 		if (delay >= 1000)
2394 			pr_info("pci %04x:%02x:%02x.%d: not ready after %dms; waiting\n",
2395 				pci_domain_nr(bus), bus->number,
2396 				PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2397 
2398 		msleep(delay);
2399 		delay *= 2;
2400 
2401 		if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
2402 			return false;
2403 	}
2404 
2405 	if (delay >= 1000)
2406 		pr_info("pci %04x:%02x:%02x.%d: ready after %dms\n",
2407 			pci_domain_nr(bus), bus->number,
2408 			PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2409 
2410 	return true;
2411 }
2412 
2413 bool pci_bus_generic_read_dev_vendor_id(struct pci_bus *bus, int devfn, u32 *l,
2414 					int timeout)
2415 {
2416 	if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
2417 		return false;
2418 
2419 	/* Some broken boards return 0 or ~0 (PCI_ERROR_RESPONSE) if a slot is empty: */
2420 	if (PCI_POSSIBLE_ERROR(*l) || *l == 0x00000000 ||
2421 	    *l == 0x0000ffff || *l == 0xffff0000)
2422 		return false;
2423 
2424 	if (pci_bus_rrs_vendor_id(*l))
2425 		return pci_bus_wait_rrs(bus, devfn, l, timeout);
2426 
2427 	return true;
2428 }
2429 
2430 bool pci_bus_read_dev_vendor_id(struct pci_bus *bus, int devfn, u32 *l,
2431 				int timeout)
2432 {
2433 #ifdef CONFIG_PCI_QUIRKS
2434 	struct pci_dev *bridge = bus->self;
2435 
2436 	/*
2437 	 * Certain IDT switches have an issue where they improperly trigger
2438 	 * ACS Source Validation errors on completions for config reads.
2439 	 */
2440 	if (bridge && bridge->vendor == PCI_VENDOR_ID_IDT &&
2441 	    bridge->device == 0x80b5)
2442 		return pci_idt_bus_quirk(bus, devfn, l, timeout);
2443 #endif
2444 
2445 	return pci_bus_generic_read_dev_vendor_id(bus, devfn, l, timeout);
2446 }
2447 EXPORT_SYMBOL(pci_bus_read_dev_vendor_id);
2448 
2449 /*
2450  * Read the config data for a PCI device, sanity-check it,
2451  * and fill in the dev structure.
2452  */
2453 static struct pci_dev *pci_scan_device(struct pci_bus *bus, int devfn)
2454 {
2455 	struct pci_dev *dev;
2456 	u32 l;
2457 
2458 	if (!pci_bus_read_dev_vendor_id(bus, devfn, &l, 60*1000))
2459 		return NULL;
2460 
2461 	dev = pci_alloc_dev(bus);
2462 	if (!dev)
2463 		return NULL;
2464 
2465 	dev->devfn = devfn;
2466 	dev->vendor = l & 0xffff;
2467 	dev->device = (l >> 16) & 0xffff;
2468 
2469 	if (pci_setup_device(dev)) {
2470 		pci_bus_put(dev->bus);
2471 		kfree(dev);
2472 		return NULL;
2473 	}
2474 
2475 	return dev;
2476 }
2477 
2478 void pcie_report_downtraining(struct pci_dev *dev)
2479 {
2480 	if (!pci_is_pcie(dev))
2481 		return;
2482 
2483 	/* Look from the device up to avoid downstream ports with no devices */
2484 	if ((pci_pcie_type(dev) != PCI_EXP_TYPE_ENDPOINT) &&
2485 	    (pci_pcie_type(dev) != PCI_EXP_TYPE_LEG_END) &&
2486 	    (pci_pcie_type(dev) != PCI_EXP_TYPE_UPSTREAM))
2487 		return;
2488 
2489 	/* Multi-function PCIe devices share the same link/status */
2490 	if (PCI_FUNC(dev->devfn) != 0 || dev->is_virtfn)
2491 		return;
2492 
2493 	/* Print link status only if the device is constrained by the fabric */
2494 	__pcie_print_link_status(dev, false);
2495 }
2496 
2497 static void pci_init_capabilities(struct pci_dev *dev)
2498 {
2499 	pci_ea_init(dev);		/* Enhanced Allocation */
2500 	pci_msi_init(dev);		/* Disable MSI */
2501 	pci_msix_init(dev);		/* Disable MSI-X */
2502 
2503 	/* Buffers for saving PCIe and PCI-X capabilities */
2504 	pci_allocate_cap_save_buffers(dev);
2505 
2506 	pci_pm_init(dev);		/* Power Management */
2507 	pci_vpd_init(dev);		/* Vital Product Data */
2508 	pci_configure_ari(dev);		/* Alternative Routing-ID Forwarding */
2509 	pci_iov_init(dev);		/* Single Root I/O Virtualization */
2510 	pci_ats_init(dev);		/* Address Translation Services */
2511 	pci_pri_init(dev);		/* Page Request Interface */
2512 	pci_pasid_init(dev);		/* Process Address Space ID */
2513 	pci_acs_init(dev);		/* Access Control Services */
2514 	pci_ptm_init(dev);		/* Precision Time Measurement */
2515 	pci_aer_init(dev);		/* Advanced Error Reporting */
2516 	pci_dpc_init(dev);		/* Downstream Port Containment */
2517 	pci_rcec_init(dev);		/* Root Complex Event Collector */
2518 	pci_doe_init(dev);		/* Data Object Exchange */
2519 	pci_tph_init(dev);		/* TLP Processing Hints */
2520 
2521 	pcie_report_downtraining(dev);
2522 	pci_init_reset_methods(dev);
2523 }
2524 
2525 /*
2526  * This is the equivalent of pci_host_bridge_msi_domain() that acts on
2527  * devices. Firmware interfaces that can select the MSI domain on a
2528  * per-device basis should be called from here.
2529  */
2530 static struct irq_domain *pci_dev_msi_domain(struct pci_dev *dev)
2531 {
2532 	struct irq_domain *d;
2533 
2534 	/*
2535 	 * If a domain has been set through the pcibios_device_add()
2536 	 * callback, then this is the one (platform code knows best).
2537 	 */
2538 	d = dev_get_msi_domain(&dev->dev);
2539 	if (d)
2540 		return d;
2541 
2542 	/*
2543 	 * Let's see if we have a firmware interface able to provide
2544 	 * the domain.
2545 	 */
2546 	d = pci_msi_get_device_domain(dev);
2547 	if (d)
2548 		return d;
2549 
2550 	return NULL;
2551 }
2552 
2553 static void pci_set_msi_domain(struct pci_dev *dev)
2554 {
2555 	struct irq_domain *d;
2556 
2557 	/*
2558 	 * If the platform or firmware interfaces cannot supply a
2559 	 * device-specific MSI domain, then inherit the default domain
2560 	 * from the host bridge itself.
2561 	 */
2562 	d = pci_dev_msi_domain(dev);
2563 	if (!d)
2564 		d = dev_get_msi_domain(&dev->bus->dev);
2565 
2566 	dev_set_msi_domain(&dev->dev, d);
2567 }
2568 
2569 void pci_device_add(struct pci_dev *dev, struct pci_bus *bus)
2570 {
2571 	int ret;
2572 
2573 	pci_configure_device(dev);
2574 
2575 	device_initialize(&dev->dev);
2576 	dev->dev.release = pci_release_dev;
2577 
2578 	set_dev_node(&dev->dev, pcibus_to_node(bus));
2579 	dev->dev.dma_mask = &dev->dma_mask;
2580 	dev->dev.dma_parms = &dev->dma_parms;
2581 	dev->dev.coherent_dma_mask = 0xffffffffull;
2582 
2583 	dma_set_max_seg_size(&dev->dev, 65536);
2584 	dma_set_seg_boundary(&dev->dev, 0xffffffff);
2585 
2586 	pcie_failed_link_retrain(dev);
2587 
2588 	/* Fix up broken headers */
2589 	pci_fixup_device(pci_fixup_header, dev);
2590 
2591 	pci_reassigndev_resource_alignment(dev);
2592 
2593 	dev->state_saved = false;
2594 
2595 	pci_init_capabilities(dev);
2596 
2597 	/*
2598 	 * Add the device to our list of discovered devices
2599 	 * and the bus list for fixup functions, etc.
2600 	 */
2601 	down_write(&pci_bus_sem);
2602 	list_add_tail(&dev->bus_list, &bus->devices);
2603 	up_write(&pci_bus_sem);
2604 
2605 	ret = pcibios_device_add(dev);
2606 	WARN_ON(ret < 0);
2607 
2608 	/* Set up MSI IRQ domain */
2609 	pci_set_msi_domain(dev);
2610 
2611 	/* Notifier could use PCI capabilities */
2612 	dev->match_driver = false;
2613 	ret = device_add(&dev->dev);
2614 	WARN_ON(ret < 0);
2615 
2616 	pci_npem_create(dev);
2617 }
2618 
2619 struct pci_dev *pci_scan_single_device(struct pci_bus *bus, int devfn)
2620 {
2621 	struct pci_dev *dev;
2622 
2623 	dev = pci_get_slot(bus, devfn);
2624 	if (dev) {
2625 		pci_dev_put(dev);
2626 		return dev;
2627 	}
2628 
2629 	dev = pci_scan_device(bus, devfn);
2630 	if (!dev)
2631 		return NULL;
2632 
2633 	pci_device_add(dev, bus);
2634 
2635 	return dev;
2636 }
2637 EXPORT_SYMBOL(pci_scan_single_device);
2638 
2639 static int next_ari_fn(struct pci_bus *bus, struct pci_dev *dev, int fn)
2640 {
2641 	int pos;
2642 	u16 cap = 0;
2643 	unsigned int next_fn;
2644 
2645 	if (!dev)
2646 		return -ENODEV;
2647 
2648 	pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ARI);
2649 	if (!pos)
2650 		return -ENODEV;
2651 
2652 	pci_read_config_word(dev, pos + PCI_ARI_CAP, &cap);
2653 	next_fn = PCI_ARI_CAP_NFN(cap);
2654 	if (next_fn <= fn)
2655 		return -ENODEV;	/* protect against malformed list */
2656 
2657 	return next_fn;
2658 }
2659 
2660 static int next_fn(struct pci_bus *bus, struct pci_dev *dev, int fn)
2661 {
2662 	if (pci_ari_enabled(bus))
2663 		return next_ari_fn(bus, dev, fn);
2664 
2665 	if (fn >= 7)
2666 		return -ENODEV;
2667 	/* only multifunction devices may have more functions */
2668 	if (dev && !dev->multifunction)
2669 		return -ENODEV;
2670 
2671 	return fn + 1;
2672 }
2673 
2674 static int only_one_child(struct pci_bus *bus)
2675 {
2676 	struct pci_dev *bridge = bus->self;
2677 
2678 	/*
2679 	 * Systems with unusual topologies set PCI_SCAN_ALL_PCIE_DEVS so
2680 	 * we scan for all possible devices, not just Device 0.
2681 	 */
2682 	if (pci_has_flag(PCI_SCAN_ALL_PCIE_DEVS))
2683 		return 0;
2684 
2685 	/*
2686 	 * A PCIe Downstream Port normally leads to a Link with only Device
2687 	 * 0 on it (PCIe spec r3.1, sec 7.3.1).  As an optimization, scan
2688 	 * only for Device 0 in that situation.
2689 	 */
2690 	if (bridge && pci_is_pcie(bridge) && pcie_downstream_port(bridge))
2691 		return 1;
2692 
2693 	return 0;
2694 }
2695 
2696 /**
2697  * pci_scan_slot - Scan a PCI slot on a bus for devices
2698  * @bus: PCI bus to scan
2699  * @devfn: slot number to scan (must have zero function)
2700  *
2701  * Scan a PCI slot on the specified PCI bus for devices, adding
2702  * discovered devices to the @bus->devices list.  New devices
2703  * will not have is_added set.
2704  *
2705  * Returns the number of new devices found.
2706  */
2707 int pci_scan_slot(struct pci_bus *bus, int devfn)
2708 {
2709 	struct pci_dev *dev;
2710 	int fn = 0, nr = 0;
2711 
2712 	if (only_one_child(bus) && (devfn > 0))
2713 		return 0; /* Already scanned the entire slot */
2714 
2715 	do {
2716 		dev = pci_scan_single_device(bus, devfn + fn);
2717 		if (dev) {
2718 			if (!pci_dev_is_added(dev))
2719 				nr++;
2720 			if (fn > 0)
2721 				dev->multifunction = 1;
2722 		} else if (fn == 0) {
2723 			/*
2724 			 * Function 0 is required unless we are running on
2725 			 * a hypervisor that passes through individual PCI
2726 			 * functions.
2727 			 */
2728 			if (!hypervisor_isolated_pci_functions())
2729 				break;
2730 		}
2731 		fn = next_fn(bus, dev, fn);
2732 	} while (fn >= 0);
2733 
2734 	/* Only one slot has PCIe device */
2735 	if (bus->self && nr)
2736 		pcie_aspm_init_link_state(bus->self);
2737 
2738 	return nr;
2739 }
2740 EXPORT_SYMBOL(pci_scan_slot);
2741 
2742 static int pcie_find_smpss(struct pci_dev *dev, void *data)
2743 {
2744 	u8 *smpss = data;
2745 
2746 	if (!pci_is_pcie(dev))
2747 		return 0;
2748 
2749 	/*
2750 	 * We don't have a way to change MPS settings on devices that have
2751 	 * drivers attached.  A hot-added device might support only the minimum
2752 	 * MPS setting (MPS=128).  Therefore, if the fabric contains a bridge
2753 	 * where devices may be hot-added, we limit the fabric MPS to 128 so
2754 	 * hot-added devices will work correctly.
2755 	 *
2756 	 * However, if we hot-add a device to a slot directly below a Root
2757 	 * Port, it's impossible for there to be other existing devices below
2758 	 * the port.  We don't limit the MPS in this case because we can
2759 	 * reconfigure MPS on both the Root Port and the hot-added device,
2760 	 * and there are no other devices involved.
2761 	 *
2762 	 * Note that this PCIE_BUS_SAFE path assumes no peer-to-peer DMA.
2763 	 */
2764 	if (dev->is_hotplug_bridge &&
2765 	    pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT)
2766 		*smpss = 0;
2767 
2768 	if (*smpss > dev->pcie_mpss)
2769 		*smpss = dev->pcie_mpss;
2770 
2771 	return 0;
2772 }
2773 
2774 static void pcie_write_mps(struct pci_dev *dev, int mps)
2775 {
2776 	int rc;
2777 
2778 	if (pcie_bus_config == PCIE_BUS_PERFORMANCE) {
2779 		mps = 128 << dev->pcie_mpss;
2780 
2781 		if (pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT &&
2782 		    dev->bus->self)
2783 
2784 			/*
2785 			 * For "Performance", the assumption is made that
2786 			 * downstream communication will never be larger than
2787 			 * the MRRS.  So, the MPS only needs to be configured
2788 			 * for the upstream communication.  This being the case,
2789 			 * walk from the top down and set the MPS of the child
2790 			 * to that of the parent bus.
2791 			 *
2792 			 * Configure the device MPS with the smaller of the
2793 			 * device MPSS or the bridge MPS (which is assumed to be
2794 			 * properly configured at this point to the largest
2795 			 * allowable MPS based on its parent bus).
2796 			 */
2797 			mps = min(mps, pcie_get_mps(dev->bus->self));
2798 	}
2799 
2800 	rc = pcie_set_mps(dev, mps);
2801 	if (rc)
2802 		pci_err(dev, "Failed attempting to set the MPS\n");
2803 }
2804 
2805 static void pcie_write_mrrs(struct pci_dev *dev)
2806 {
2807 	int rc, mrrs;
2808 
2809 	/*
2810 	 * In the "safe" case, do not configure the MRRS.  There appear to be
2811 	 * issues with setting MRRS to 0 on a number of devices.
2812 	 */
2813 	if (pcie_bus_config != PCIE_BUS_PERFORMANCE)
2814 		return;
2815 
2816 	/*
2817 	 * For max performance, the MRRS must be set to the largest supported
2818 	 * value.  However, it cannot be configured larger than the MPS the
2819 	 * device or the bus can support.  This should already be properly
2820 	 * configured by a prior call to pcie_write_mps().
2821 	 */
2822 	mrrs = pcie_get_mps(dev);
2823 
2824 	/*
2825 	 * MRRS is a R/W register.  Invalid values can be written, but a
2826 	 * subsequent read will verify if the value is acceptable or not.
2827 	 * If the MRRS value provided is not acceptable (e.g., too large),
2828 	 * shrink the value until it is acceptable to the HW.
2829 	 */
2830 	while (mrrs != pcie_get_readrq(dev) && mrrs >= 128) {
2831 		rc = pcie_set_readrq(dev, mrrs);
2832 		if (!rc)
2833 			break;
2834 
2835 		pci_warn(dev, "Failed attempting to set the MRRS\n");
2836 		mrrs /= 2;
2837 	}
2838 
2839 	if (mrrs < 128)
2840 		pci_err(dev, "MRRS was unable to be configured with a safe value.  If problems are experienced, try running with pci=pcie_bus_safe\n");
2841 }
2842 
2843 static int pcie_bus_configure_set(struct pci_dev *dev, void *data)
2844 {
2845 	int mps, orig_mps;
2846 
2847 	if (!pci_is_pcie(dev))
2848 		return 0;
2849 
2850 	if (pcie_bus_config == PCIE_BUS_TUNE_OFF ||
2851 	    pcie_bus_config == PCIE_BUS_DEFAULT)
2852 		return 0;
2853 
2854 	mps = 128 << *(u8 *)data;
2855 	orig_mps = pcie_get_mps(dev);
2856 
2857 	pcie_write_mps(dev, mps);
2858 	pcie_write_mrrs(dev);
2859 
2860 	pci_info(dev, "Max Payload Size set to %4d/%4d (was %4d), Max Read Rq %4d\n",
2861 		 pcie_get_mps(dev), 128 << dev->pcie_mpss,
2862 		 orig_mps, pcie_get_readrq(dev));
2863 
2864 	return 0;
2865 }
2866 
2867 /*
2868  * pcie_bus_configure_settings() requires that pci_walk_bus work in a top-down,
2869  * parents then children fashion.  If this changes, then this code will not
2870  * work as designed.
2871  */
2872 void pcie_bus_configure_settings(struct pci_bus *bus)
2873 {
2874 	u8 smpss = 0;
2875 
2876 	if (!bus->self)
2877 		return;
2878 
2879 	if (!pci_is_pcie(bus->self))
2880 		return;
2881 
2882 	/*
2883 	 * FIXME - Peer to peer DMA is possible, though the endpoint would need
2884 	 * to be aware of the MPS of the destination.  To work around this,
2885 	 * simply force the MPS of the entire system to the smallest possible.
2886 	 */
2887 	if (pcie_bus_config == PCIE_BUS_PEER2PEER)
2888 		smpss = 0;
2889 
2890 	if (pcie_bus_config == PCIE_BUS_SAFE) {
2891 		smpss = bus->self->pcie_mpss;
2892 
2893 		pcie_find_smpss(bus->self, &smpss);
2894 		pci_walk_bus(bus, pcie_find_smpss, &smpss);
2895 	}
2896 
2897 	pcie_bus_configure_set(bus->self, &smpss);
2898 	pci_walk_bus(bus, pcie_bus_configure_set, &smpss);
2899 }
2900 EXPORT_SYMBOL_GPL(pcie_bus_configure_settings);
2901 
2902 /*
2903  * Called after each bus is probed, but before its children are examined.  This
2904  * is marked as __weak because multiple architectures define it.
2905  */
2906 void __weak pcibios_fixup_bus(struct pci_bus *bus)
2907 {
2908        /* nothing to do, expected to be removed in the future */
2909 }
2910 
2911 /**
2912  * pci_scan_child_bus_extend() - Scan devices below a bus
2913  * @bus: Bus to scan for devices
2914  * @available_buses: Total number of buses available (%0 does not try to
2915  *		     extend beyond the minimal)
2916  *
2917  * Scans devices below @bus including subordinate buses. Returns new
2918  * subordinate number including all the found devices. Passing
2919  * @available_buses causes the remaining bus space to be distributed
2920  * equally between hotplug-capable bridges to allow future extension of the
2921  * hierarchy.
2922  */
2923 static unsigned int pci_scan_child_bus_extend(struct pci_bus *bus,
2924 					      unsigned int available_buses)
2925 {
2926 	unsigned int used_buses, normal_bridges = 0, hotplug_bridges = 0;
2927 	unsigned int start = bus->busn_res.start;
2928 	unsigned int devfn, cmax, max = start;
2929 	struct pci_dev *dev;
2930 
2931 	dev_dbg(&bus->dev, "scanning bus\n");
2932 
2933 	/* Go find them, Rover! */
2934 	for (devfn = 0; devfn < 256; devfn += 8)
2935 		pci_scan_slot(bus, devfn);
2936 
2937 	/* Reserve buses for SR-IOV capability */
2938 	used_buses = pci_iov_bus_range(bus);
2939 	max += used_buses;
2940 
2941 	/*
2942 	 * After performing arch-dependent fixup of the bus, look behind
2943 	 * all PCI-to-PCI bridges on this bus.
2944 	 */
2945 	if (!bus->is_added) {
2946 		dev_dbg(&bus->dev, "fixups for bus\n");
2947 		pcibios_fixup_bus(bus);
2948 		bus->is_added = 1;
2949 	}
2950 
2951 	/*
2952 	 * Calculate how many hotplug bridges and normal bridges there
2953 	 * are on this bus. We will distribute the additional available
2954 	 * buses between hotplug bridges.
2955 	 */
2956 	for_each_pci_bridge(dev, bus) {
2957 		if (dev->is_hotplug_bridge)
2958 			hotplug_bridges++;
2959 		else
2960 			normal_bridges++;
2961 	}
2962 
2963 	/*
2964 	 * Scan bridges that are already configured. We don't touch them
2965 	 * unless they are misconfigured (which will be done in the second
2966 	 * scan below).
2967 	 */
2968 	for_each_pci_bridge(dev, bus) {
2969 		cmax = max;
2970 		max = pci_scan_bridge_extend(bus, dev, max, 0, 0);
2971 
2972 		/*
2973 		 * Reserve one bus for each bridge now to avoid extending
2974 		 * hotplug bridges too much during the second scan below.
2975 		 */
2976 		used_buses++;
2977 		if (max - cmax > 1)
2978 			used_buses += max - cmax - 1;
2979 	}
2980 
2981 	/* Scan bridges that need to be reconfigured */
2982 	for_each_pci_bridge(dev, bus) {
2983 		unsigned int buses = 0;
2984 
2985 		if (!hotplug_bridges && normal_bridges == 1) {
2986 			/*
2987 			 * There is only one bridge on the bus (upstream
2988 			 * port) so it gets all available buses which it
2989 			 * can then distribute to the possible hotplug
2990 			 * bridges below.
2991 			 */
2992 			buses = available_buses;
2993 		} else if (dev->is_hotplug_bridge) {
2994 			/*
2995 			 * Distribute the extra buses between hotplug
2996 			 * bridges if any.
2997 			 */
2998 			buses = available_buses / hotplug_bridges;
2999 			buses = min(buses, available_buses - used_buses + 1);
3000 		}
3001 
3002 		cmax = max;
3003 		max = pci_scan_bridge_extend(bus, dev, cmax, buses, 1);
3004 		/* One bus is already accounted so don't add it again */
3005 		if (max - cmax > 1)
3006 			used_buses += max - cmax - 1;
3007 	}
3008 
3009 	/*
3010 	 * Make sure a hotplug bridge has at least the minimum requested
3011 	 * number of buses but allow it to grow up to the maximum available
3012 	 * bus number if there is room.
3013 	 */
3014 	if (bus->self && bus->self->is_hotplug_bridge) {
3015 		used_buses = max_t(unsigned int, available_buses,
3016 				   pci_hotplug_bus_size - 1);
3017 		if (max - start < used_buses) {
3018 			max = start + used_buses;
3019 
3020 			/* Do not allocate more buses than we have room left */
3021 			if (max > bus->busn_res.end)
3022 				max = bus->busn_res.end;
3023 
3024 			dev_dbg(&bus->dev, "%pR extended by %#02x\n",
3025 				&bus->busn_res, max - start);
3026 		}
3027 	}
3028 
3029 	/*
3030 	 * We've scanned the bus and so we know all about what's on
3031 	 * the other side of any bridges that may be on this bus plus
3032 	 * any devices.
3033 	 *
3034 	 * Return how far we've got finding sub-buses.
3035 	 */
3036 	dev_dbg(&bus->dev, "bus scan returning with max=%02x\n", max);
3037 	return max;
3038 }
3039 
3040 /**
3041  * pci_scan_child_bus() - Scan devices below a bus
3042  * @bus: Bus to scan for devices
3043  *
3044  * Scans devices below @bus including subordinate buses. Returns new
3045  * subordinate number including all the found devices.
3046  */
3047 unsigned int pci_scan_child_bus(struct pci_bus *bus)
3048 {
3049 	return pci_scan_child_bus_extend(bus, 0);
3050 }
3051 EXPORT_SYMBOL_GPL(pci_scan_child_bus);
3052 
3053 /**
3054  * pcibios_root_bridge_prepare - Platform-specific host bridge setup
3055  * @bridge: Host bridge to set up
3056  *
3057  * Default empty implementation.  Replace with an architecture-specific setup
3058  * routine, if necessary.
3059  */
3060 int __weak pcibios_root_bridge_prepare(struct pci_host_bridge *bridge)
3061 {
3062 	return 0;
3063 }
3064 
3065 void __weak pcibios_add_bus(struct pci_bus *bus)
3066 {
3067 }
3068 
3069 void __weak pcibios_remove_bus(struct pci_bus *bus)
3070 {
3071 }
3072 
3073 struct pci_bus *pci_create_root_bus(struct device *parent, int bus,
3074 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
3075 {
3076 	int error;
3077 	struct pci_host_bridge *bridge;
3078 
3079 	bridge = pci_alloc_host_bridge(0);
3080 	if (!bridge)
3081 		return NULL;
3082 
3083 	bridge->dev.parent = parent;
3084 
3085 	list_splice_init(resources, &bridge->windows);
3086 	bridge->sysdata = sysdata;
3087 	bridge->busnr = bus;
3088 	bridge->ops = ops;
3089 
3090 	error = pci_register_host_bridge(bridge);
3091 	if (error < 0)
3092 		goto err_out;
3093 
3094 	return bridge->bus;
3095 
3096 err_out:
3097 	put_device(&bridge->dev);
3098 	return NULL;
3099 }
3100 EXPORT_SYMBOL_GPL(pci_create_root_bus);
3101 
3102 int pci_host_probe(struct pci_host_bridge *bridge)
3103 {
3104 	struct pci_bus *bus, *child;
3105 	int ret;
3106 
3107 	pci_lock_rescan_remove();
3108 	ret = pci_scan_root_bus_bridge(bridge);
3109 	pci_unlock_rescan_remove();
3110 	if (ret < 0) {
3111 		dev_err(bridge->dev.parent, "Scanning root bridge failed");
3112 		return ret;
3113 	}
3114 
3115 	bus = bridge->bus;
3116 
3117 	/* If we must preserve the resource configuration, claim now */
3118 	if (bridge->preserve_config)
3119 		pci_bus_claim_resources(bus);
3120 
3121 	/*
3122 	 * Assign whatever was left unassigned. If we didn't claim above,
3123 	 * this will reassign everything.
3124 	 */
3125 	pci_assign_unassigned_root_bus_resources(bus);
3126 
3127 	list_for_each_entry(child, &bus->children, node)
3128 		pcie_bus_configure_settings(child);
3129 
3130 	pci_lock_rescan_remove();
3131 	pci_bus_add_devices(bus);
3132 	pci_unlock_rescan_remove();
3133 
3134 	/*
3135 	 * Ensure pm_runtime_enable() is called for the controller drivers
3136 	 * before calling pci_host_probe(). The PM framework expects that
3137 	 * if the parent device supports runtime PM, it will be enabled
3138 	 * before child runtime PM is enabled.
3139 	 */
3140 	pm_runtime_set_active(&bridge->dev);
3141 	pm_runtime_no_callbacks(&bridge->dev);
3142 	devm_pm_runtime_enable(&bridge->dev);
3143 
3144 	return 0;
3145 }
3146 EXPORT_SYMBOL_GPL(pci_host_probe);
3147 
3148 int pci_bus_insert_busn_res(struct pci_bus *b, int bus, int bus_max)
3149 {
3150 	struct resource *res = &b->busn_res;
3151 	struct resource *parent_res, *conflict;
3152 
3153 	res->start = bus;
3154 	res->end = bus_max;
3155 	res->flags = IORESOURCE_BUS;
3156 
3157 	if (!pci_is_root_bus(b))
3158 		parent_res = &b->parent->busn_res;
3159 	else {
3160 		parent_res = get_pci_domain_busn_res(pci_domain_nr(b));
3161 		res->flags |= IORESOURCE_PCI_FIXED;
3162 	}
3163 
3164 	conflict = request_resource_conflict(parent_res, res);
3165 
3166 	if (conflict)
3167 		dev_info(&b->dev,
3168 			   "busn_res: can not insert %pR under %s%pR (conflicts with %s %pR)\n",
3169 			    res, pci_is_root_bus(b) ? "domain " : "",
3170 			    parent_res, conflict->name, conflict);
3171 
3172 	return conflict == NULL;
3173 }
3174 
3175 int pci_bus_update_busn_res_end(struct pci_bus *b, int bus_max)
3176 {
3177 	struct resource *res = &b->busn_res;
3178 	struct resource old_res = *res;
3179 	resource_size_t size;
3180 	int ret;
3181 
3182 	if (res->start > bus_max)
3183 		return -EINVAL;
3184 
3185 	size = bus_max - res->start + 1;
3186 	ret = adjust_resource(res, res->start, size);
3187 	dev_info(&b->dev, "busn_res: %pR end %s updated to %02x\n",
3188 			&old_res, ret ? "can not be" : "is", bus_max);
3189 
3190 	if (!ret && !res->parent)
3191 		pci_bus_insert_busn_res(b, res->start, res->end);
3192 
3193 	return ret;
3194 }
3195 
3196 void pci_bus_release_busn_res(struct pci_bus *b)
3197 {
3198 	struct resource *res = &b->busn_res;
3199 	int ret;
3200 
3201 	if (!res->flags || !res->parent)
3202 		return;
3203 
3204 	ret = release_resource(res);
3205 	dev_info(&b->dev, "busn_res: %pR %s released\n",
3206 			res, ret ? "can not be" : "is");
3207 }
3208 
3209 int pci_scan_root_bus_bridge(struct pci_host_bridge *bridge)
3210 {
3211 	struct resource_entry *window;
3212 	bool found = false;
3213 	struct pci_bus *b;
3214 	int max, bus, ret;
3215 
3216 	if (!bridge)
3217 		return -EINVAL;
3218 
3219 	resource_list_for_each_entry(window, &bridge->windows)
3220 		if (window->res->flags & IORESOURCE_BUS) {
3221 			bridge->busnr = window->res->start;
3222 			found = true;
3223 			break;
3224 		}
3225 
3226 	ret = pci_register_host_bridge(bridge);
3227 	if (ret < 0)
3228 		return ret;
3229 
3230 	b = bridge->bus;
3231 	bus = bridge->busnr;
3232 
3233 	if (!found) {
3234 		dev_info(&b->dev,
3235 		 "No busn resource found for root bus, will use [bus %02x-ff]\n",
3236 			bus);
3237 		pci_bus_insert_busn_res(b, bus, 255);
3238 	}
3239 
3240 	max = pci_scan_child_bus(b);
3241 
3242 	if (!found)
3243 		pci_bus_update_busn_res_end(b, max);
3244 
3245 	return 0;
3246 }
3247 EXPORT_SYMBOL(pci_scan_root_bus_bridge);
3248 
3249 struct pci_bus *pci_scan_root_bus(struct device *parent, int bus,
3250 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
3251 {
3252 	struct resource_entry *window;
3253 	bool found = false;
3254 	struct pci_bus *b;
3255 	int max;
3256 
3257 	resource_list_for_each_entry(window, resources)
3258 		if (window->res->flags & IORESOURCE_BUS) {
3259 			found = true;
3260 			break;
3261 		}
3262 
3263 	b = pci_create_root_bus(parent, bus, ops, sysdata, resources);
3264 	if (!b)
3265 		return NULL;
3266 
3267 	if (!found) {
3268 		dev_info(&b->dev,
3269 		 "No busn resource found for root bus, will use [bus %02x-ff]\n",
3270 			bus);
3271 		pci_bus_insert_busn_res(b, bus, 255);
3272 	}
3273 
3274 	max = pci_scan_child_bus(b);
3275 
3276 	if (!found)
3277 		pci_bus_update_busn_res_end(b, max);
3278 
3279 	return b;
3280 }
3281 EXPORT_SYMBOL(pci_scan_root_bus);
3282 
3283 struct pci_bus *pci_scan_bus(int bus, struct pci_ops *ops,
3284 					void *sysdata)
3285 {
3286 	LIST_HEAD(resources);
3287 	struct pci_bus *b;
3288 
3289 	pci_add_resource(&resources, &ioport_resource);
3290 	pci_add_resource(&resources, &iomem_resource);
3291 	pci_add_resource(&resources, &busn_resource);
3292 	b = pci_create_root_bus(NULL, bus, ops, sysdata, &resources);
3293 	if (b) {
3294 		pci_scan_child_bus(b);
3295 	} else {
3296 		pci_free_resource_list(&resources);
3297 	}
3298 	return b;
3299 }
3300 EXPORT_SYMBOL(pci_scan_bus);
3301 
3302 /**
3303  * pci_rescan_bus_bridge_resize - Scan a PCI bus for devices
3304  * @bridge: PCI bridge for the bus to scan
3305  *
3306  * Scan a PCI bus and child buses for new devices, add them,
3307  * and enable them, resizing bridge mmio/io resource if necessary
3308  * and possible.  The caller must ensure the child devices are already
3309  * removed for resizing to occur.
3310  *
3311  * Returns the max number of subordinate bus discovered.
3312  */
3313 unsigned int pci_rescan_bus_bridge_resize(struct pci_dev *bridge)
3314 {
3315 	unsigned int max;
3316 	struct pci_bus *bus = bridge->subordinate;
3317 
3318 	max = pci_scan_child_bus(bus);
3319 
3320 	pci_assign_unassigned_bridge_resources(bridge);
3321 
3322 	pci_bus_add_devices(bus);
3323 
3324 	return max;
3325 }
3326 
3327 /**
3328  * pci_rescan_bus - Scan a PCI bus for devices
3329  * @bus: PCI bus to scan
3330  *
3331  * Scan a PCI bus and child buses for new devices, add them,
3332  * and enable them.
3333  *
3334  * Returns the max number of subordinate bus discovered.
3335  */
3336 unsigned int pci_rescan_bus(struct pci_bus *bus)
3337 {
3338 	unsigned int max;
3339 
3340 	max = pci_scan_child_bus(bus);
3341 	pci_assign_unassigned_bus_resources(bus);
3342 	pci_bus_add_devices(bus);
3343 
3344 	return max;
3345 }
3346 EXPORT_SYMBOL_GPL(pci_rescan_bus);
3347 
3348 /*
3349  * pci_rescan_bus(), pci_rescan_bus_bridge_resize() and PCI device removal
3350  * routines should always be executed under this mutex.
3351  */
3352 static DEFINE_MUTEX(pci_rescan_remove_lock);
3353 
3354 void pci_lock_rescan_remove(void)
3355 {
3356 	mutex_lock(&pci_rescan_remove_lock);
3357 }
3358 EXPORT_SYMBOL_GPL(pci_lock_rescan_remove);
3359 
3360 void pci_unlock_rescan_remove(void)
3361 {
3362 	mutex_unlock(&pci_rescan_remove_lock);
3363 }
3364 EXPORT_SYMBOL_GPL(pci_unlock_rescan_remove);
3365 
3366 static int __init pci_sort_bf_cmp(const struct device *d_a,
3367 				  const struct device *d_b)
3368 {
3369 	const struct pci_dev *a = to_pci_dev(d_a);
3370 	const struct pci_dev *b = to_pci_dev(d_b);
3371 
3372 	if      (pci_domain_nr(a->bus) < pci_domain_nr(b->bus)) return -1;
3373 	else if (pci_domain_nr(a->bus) > pci_domain_nr(b->bus)) return  1;
3374 
3375 	if      (a->bus->number < b->bus->number) return -1;
3376 	else if (a->bus->number > b->bus->number) return  1;
3377 
3378 	if      (a->devfn < b->devfn) return -1;
3379 	else if (a->devfn > b->devfn) return  1;
3380 
3381 	return 0;
3382 }
3383 
3384 void __init pci_sort_breadthfirst(void)
3385 {
3386 	bus_sort_breadthfirst(&pci_bus_type, &pci_sort_bf_cmp);
3387 }
3388 
3389 int pci_hp_add_bridge(struct pci_dev *dev)
3390 {
3391 	struct pci_bus *parent = dev->bus;
3392 	int busnr, start = parent->busn_res.start;
3393 	unsigned int available_buses = 0;
3394 	int end = parent->busn_res.end;
3395 
3396 	for (busnr = start; busnr <= end; busnr++) {
3397 		if (!pci_find_bus(pci_domain_nr(parent), busnr))
3398 			break;
3399 	}
3400 	if (busnr-- > end) {
3401 		pci_err(dev, "No bus number available for hot-added bridge\n");
3402 		return -1;
3403 	}
3404 
3405 	/* Scan bridges that are already configured */
3406 	busnr = pci_scan_bridge(parent, dev, busnr, 0);
3407 
3408 	/*
3409 	 * Distribute the available bus numbers between hotplug-capable
3410 	 * bridges to make extending the chain later possible.
3411 	 */
3412 	available_buses = end - busnr;
3413 
3414 	/* Scan bridges that need to be reconfigured */
3415 	pci_scan_bridge_extend(parent, dev, busnr, available_buses, 1);
3416 
3417 	if (!dev->subordinate)
3418 		return -1;
3419 
3420 	return 0;
3421 }
3422 EXPORT_SYMBOL_GPL(pci_hp_add_bridge);
3423