xref: /linux/drivers/pci/probe.c (revision a3a02a52bcfcbcc4a637d4b68bf1bc391c9fad02)
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 		pci_bus_for_each_resource(child->parent, res) {
548 			if (res && res->flags) {
549 				pci_bus_add_resource(child, res,
550 						     PCI_SUBTRACTIVE_DECODE);
551 				pci_info(dev, "  bridge window %pR (subtractive decode)\n",
552 					   res);
553 			}
554 		}
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, u16 linksta)
746 {
747 	bus->cur_bus_speed = pcie_link_speed[linksta & PCI_EXP_LNKSTA_CLS];
748 }
749 EXPORT_SYMBOL_GPL(pcie_update_link_speed);
750 
751 static unsigned char agp_speeds[] = {
752 	AGP_UNKNOWN,
753 	AGP_1X,
754 	AGP_2X,
755 	AGP_4X,
756 	AGP_8X
757 };
758 
759 static enum pci_bus_speed agp_speed(int agp3, int agpstat)
760 {
761 	int index = 0;
762 
763 	if (agpstat & 4)
764 		index = 3;
765 	else if (agpstat & 2)
766 		index = 2;
767 	else if (agpstat & 1)
768 		index = 1;
769 	else
770 		goto out;
771 
772 	if (agp3) {
773 		index += 2;
774 		if (index == 5)
775 			index = 0;
776 	}
777 
778  out:
779 	return agp_speeds[index];
780 }
781 
782 static void pci_set_bus_speed(struct pci_bus *bus)
783 {
784 	struct pci_dev *bridge = bus->self;
785 	int pos;
786 
787 	pos = pci_find_capability(bridge, PCI_CAP_ID_AGP);
788 	if (!pos)
789 		pos = pci_find_capability(bridge, PCI_CAP_ID_AGP3);
790 	if (pos) {
791 		u32 agpstat, agpcmd;
792 
793 		pci_read_config_dword(bridge, pos + PCI_AGP_STATUS, &agpstat);
794 		bus->max_bus_speed = agp_speed(agpstat & 8, agpstat & 7);
795 
796 		pci_read_config_dword(bridge, pos + PCI_AGP_COMMAND, &agpcmd);
797 		bus->cur_bus_speed = agp_speed(agpstat & 8, agpcmd & 7);
798 	}
799 
800 	pos = pci_find_capability(bridge, PCI_CAP_ID_PCIX);
801 	if (pos) {
802 		u16 status;
803 		enum pci_bus_speed max;
804 
805 		pci_read_config_word(bridge, pos + PCI_X_BRIDGE_SSTATUS,
806 				     &status);
807 
808 		if (status & PCI_X_SSTATUS_533MHZ) {
809 			max = PCI_SPEED_133MHz_PCIX_533;
810 		} else if (status & PCI_X_SSTATUS_266MHZ) {
811 			max = PCI_SPEED_133MHz_PCIX_266;
812 		} else if (status & PCI_X_SSTATUS_133MHZ) {
813 			if ((status & PCI_X_SSTATUS_VERS) == PCI_X_SSTATUS_V2)
814 				max = PCI_SPEED_133MHz_PCIX_ECC;
815 			else
816 				max = PCI_SPEED_133MHz_PCIX;
817 		} else {
818 			max = PCI_SPEED_66MHz_PCIX;
819 		}
820 
821 		bus->max_bus_speed = max;
822 		bus->cur_bus_speed =
823 			pcix_bus_speed[FIELD_GET(PCI_X_SSTATUS_FREQ, status)];
824 
825 		return;
826 	}
827 
828 	if (pci_is_pcie(bridge)) {
829 		u32 linkcap;
830 		u16 linksta;
831 
832 		pcie_capability_read_dword(bridge, PCI_EXP_LNKCAP, &linkcap);
833 		bus->max_bus_speed = pcie_link_speed[linkcap & PCI_EXP_LNKCAP_SLS];
834 
835 		pcie_capability_read_word(bridge, PCI_EXP_LNKSTA, &linksta);
836 		pcie_update_link_speed(bus, linksta);
837 	}
838 }
839 
840 static struct irq_domain *pci_host_bridge_msi_domain(struct pci_bus *bus)
841 {
842 	struct irq_domain *d;
843 
844 	/* If the host bridge driver sets a MSI domain of the bridge, use it */
845 	d = dev_get_msi_domain(bus->bridge);
846 
847 	/*
848 	 * Any firmware interface that can resolve the msi_domain
849 	 * should be called from here.
850 	 */
851 	if (!d)
852 		d = pci_host_bridge_of_msi_domain(bus);
853 	if (!d)
854 		d = pci_host_bridge_acpi_msi_domain(bus);
855 
856 	/*
857 	 * If no IRQ domain was found via the OF tree, try looking it up
858 	 * directly through the fwnode_handle.
859 	 */
860 	if (!d) {
861 		struct fwnode_handle *fwnode = pci_root_bus_fwnode(bus);
862 
863 		if (fwnode)
864 			d = irq_find_matching_fwnode(fwnode,
865 						     DOMAIN_BUS_PCI_MSI);
866 	}
867 
868 	return d;
869 }
870 
871 static void pci_set_bus_msi_domain(struct pci_bus *bus)
872 {
873 	struct irq_domain *d;
874 	struct pci_bus *b;
875 
876 	/*
877 	 * The bus can be a root bus, a subordinate bus, or a virtual bus
878 	 * created by an SR-IOV device.  Walk up to the first bridge device
879 	 * found or derive the domain from the host bridge.
880 	 */
881 	for (b = bus, d = NULL; !d && !pci_is_root_bus(b); b = b->parent) {
882 		if (b->self)
883 			d = dev_get_msi_domain(&b->self->dev);
884 	}
885 
886 	if (!d)
887 		d = pci_host_bridge_msi_domain(b);
888 
889 	dev_set_msi_domain(&bus->dev, d);
890 }
891 
892 static bool pci_preserve_config(struct pci_host_bridge *host_bridge)
893 {
894 	if (pci_acpi_preserve_config(host_bridge))
895 		return true;
896 
897 	if (host_bridge->dev.parent && host_bridge->dev.parent->of_node)
898 		return of_pci_preserve_config(host_bridge->dev.parent->of_node);
899 
900 	return false;
901 }
902 
903 static int pci_register_host_bridge(struct pci_host_bridge *bridge)
904 {
905 	struct device *parent = bridge->dev.parent;
906 	struct resource_entry *window, *next, *n;
907 	struct pci_bus *bus, *b;
908 	resource_size_t offset, next_offset;
909 	LIST_HEAD(resources);
910 	struct resource *res, *next_res;
911 	char addr[64], *fmt;
912 	const char *name;
913 	int err;
914 
915 	bus = pci_alloc_bus(NULL);
916 	if (!bus)
917 		return -ENOMEM;
918 
919 	bridge->bus = bus;
920 
921 	bus->sysdata = bridge->sysdata;
922 	bus->ops = bridge->ops;
923 	bus->number = bus->busn_res.start = bridge->busnr;
924 #ifdef CONFIG_PCI_DOMAINS_GENERIC
925 	if (bridge->domain_nr == PCI_DOMAIN_NR_NOT_SET)
926 		bus->domain_nr = pci_bus_find_domain_nr(bus, parent);
927 	else
928 		bus->domain_nr = bridge->domain_nr;
929 	if (bus->domain_nr < 0) {
930 		err = bus->domain_nr;
931 		goto free;
932 	}
933 #endif
934 
935 	b = pci_find_bus(pci_domain_nr(bus), bridge->busnr);
936 	if (b) {
937 		/* Ignore it if we already got here via a different bridge */
938 		dev_dbg(&b->dev, "bus already known\n");
939 		err = -EEXIST;
940 		goto free;
941 	}
942 
943 	dev_set_name(&bridge->dev, "pci%04x:%02x", pci_domain_nr(bus),
944 		     bridge->busnr);
945 
946 	err = pcibios_root_bridge_prepare(bridge);
947 	if (err)
948 		goto free;
949 
950 	/* Temporarily move resources off the list */
951 	list_splice_init(&bridge->windows, &resources);
952 	err = device_add(&bridge->dev);
953 	if (err) {
954 		put_device(&bridge->dev);
955 		goto free;
956 	}
957 	bus->bridge = get_device(&bridge->dev);
958 	device_enable_async_suspend(bus->bridge);
959 	pci_set_bus_of_node(bus);
960 	pci_set_bus_msi_domain(bus);
961 	if (bridge->msi_domain && !dev_get_msi_domain(&bus->dev) &&
962 	    !pci_host_of_has_msi_map(parent))
963 		bus->bus_flags |= PCI_BUS_FLAGS_NO_MSI;
964 
965 	if (!parent)
966 		set_dev_node(bus->bridge, pcibus_to_node(bus));
967 
968 	bus->dev.class = &pcibus_class;
969 	bus->dev.parent = bus->bridge;
970 
971 	dev_set_name(&bus->dev, "%04x:%02x", pci_domain_nr(bus), bus->number);
972 	name = dev_name(&bus->dev);
973 
974 	err = device_register(&bus->dev);
975 	if (err)
976 		goto unregister;
977 
978 	pcibios_add_bus(bus);
979 
980 	if (bus->ops->add_bus) {
981 		err = bus->ops->add_bus(bus);
982 		if (WARN_ON(err < 0))
983 			dev_err(&bus->dev, "failed to add bus: %d\n", err);
984 	}
985 
986 	/* Create legacy_io and legacy_mem files for this bus */
987 	pci_create_legacy_files(bus);
988 
989 	if (parent)
990 		dev_info(parent, "PCI host bridge to bus %s\n", name);
991 	else
992 		pr_info("PCI host bridge to bus %s\n", name);
993 
994 	if (nr_node_ids > 1 && pcibus_to_node(bus) == NUMA_NO_NODE)
995 		dev_warn(&bus->dev, "Unknown NUMA node; performance will be reduced\n");
996 
997 	/* Check if the boot configuration by FW needs to be preserved */
998 	bridge->preserve_config = pci_preserve_config(bridge);
999 
1000 	/* Coalesce contiguous windows */
1001 	resource_list_for_each_entry_safe(window, n, &resources) {
1002 		if (list_is_last(&window->node, &resources))
1003 			break;
1004 
1005 		next = list_next_entry(window, node);
1006 		offset = window->offset;
1007 		res = window->res;
1008 		next_offset = next->offset;
1009 		next_res = next->res;
1010 
1011 		if (res->flags != next_res->flags || offset != next_offset)
1012 			continue;
1013 
1014 		if (res->end + 1 == next_res->start) {
1015 			next_res->start = res->start;
1016 			res->flags = res->start = res->end = 0;
1017 		}
1018 	}
1019 
1020 	/* Add initial resources to the bus */
1021 	resource_list_for_each_entry_safe(window, n, &resources) {
1022 		offset = window->offset;
1023 		res = window->res;
1024 		if (!res->flags && !res->start && !res->end) {
1025 			release_resource(res);
1026 			resource_list_destroy_entry(window);
1027 			continue;
1028 		}
1029 
1030 		list_move_tail(&window->node, &bridge->windows);
1031 
1032 		if (res->flags & IORESOURCE_BUS)
1033 			pci_bus_insert_busn_res(bus, bus->number, res->end);
1034 		else
1035 			pci_bus_add_resource(bus, res, 0);
1036 
1037 		if (offset) {
1038 			if (resource_type(res) == IORESOURCE_IO)
1039 				fmt = " (bus address [%#06llx-%#06llx])";
1040 			else
1041 				fmt = " (bus address [%#010llx-%#010llx])";
1042 
1043 			snprintf(addr, sizeof(addr), fmt,
1044 				 (unsigned long long)(res->start - offset),
1045 				 (unsigned long long)(res->end - offset));
1046 		} else
1047 			addr[0] = '\0';
1048 
1049 		dev_info(&bus->dev, "root bus resource %pR%s\n", res, addr);
1050 	}
1051 
1052 	down_write(&pci_bus_sem);
1053 	list_add_tail(&bus->node, &pci_root_buses);
1054 	up_write(&pci_bus_sem);
1055 
1056 	return 0;
1057 
1058 unregister:
1059 	put_device(&bridge->dev);
1060 	device_del(&bridge->dev);
1061 
1062 free:
1063 #ifdef CONFIG_PCI_DOMAINS_GENERIC
1064 	pci_bus_release_domain_nr(bus, parent);
1065 #endif
1066 	kfree(bus);
1067 	return err;
1068 }
1069 
1070 static bool pci_bridge_child_ext_cfg_accessible(struct pci_dev *bridge)
1071 {
1072 	int pos;
1073 	u32 status;
1074 
1075 	/*
1076 	 * If extended config space isn't accessible on a bridge's primary
1077 	 * bus, we certainly can't access it on the secondary bus.
1078 	 */
1079 	if (bridge->bus->bus_flags & PCI_BUS_FLAGS_NO_EXTCFG)
1080 		return false;
1081 
1082 	/*
1083 	 * PCIe Root Ports and switch ports are PCIe on both sides, so if
1084 	 * extended config space is accessible on the primary, it's also
1085 	 * accessible on the secondary.
1086 	 */
1087 	if (pci_is_pcie(bridge) &&
1088 	    (pci_pcie_type(bridge) == PCI_EXP_TYPE_ROOT_PORT ||
1089 	     pci_pcie_type(bridge) == PCI_EXP_TYPE_UPSTREAM ||
1090 	     pci_pcie_type(bridge) == PCI_EXP_TYPE_DOWNSTREAM))
1091 		return true;
1092 
1093 	/*
1094 	 * For the other bridge types:
1095 	 *   - PCI-to-PCI bridges
1096 	 *   - PCIe-to-PCI/PCI-X forward bridges
1097 	 *   - PCI/PCI-X-to-PCIe reverse bridges
1098 	 * extended config space on the secondary side is only accessible
1099 	 * if the bridge supports PCI-X Mode 2.
1100 	 */
1101 	pos = pci_find_capability(bridge, PCI_CAP_ID_PCIX);
1102 	if (!pos)
1103 		return false;
1104 
1105 	pci_read_config_dword(bridge, pos + PCI_X_STATUS, &status);
1106 	return status & (PCI_X_STATUS_266MHZ | PCI_X_STATUS_533MHZ);
1107 }
1108 
1109 static struct pci_bus *pci_alloc_child_bus(struct pci_bus *parent,
1110 					   struct pci_dev *bridge, int busnr)
1111 {
1112 	struct pci_bus *child;
1113 	struct pci_host_bridge *host;
1114 	int i;
1115 	int ret;
1116 
1117 	/* Allocate a new bus and inherit stuff from the parent */
1118 	child = pci_alloc_bus(parent);
1119 	if (!child)
1120 		return NULL;
1121 
1122 	child->parent = parent;
1123 	child->sysdata = parent->sysdata;
1124 	child->bus_flags = parent->bus_flags;
1125 
1126 	host = pci_find_host_bridge(parent);
1127 	if (host->child_ops)
1128 		child->ops = host->child_ops;
1129 	else
1130 		child->ops = parent->ops;
1131 
1132 	/*
1133 	 * Initialize some portions of the bus device, but don't register
1134 	 * it now as the parent is not properly set up yet.
1135 	 */
1136 	child->dev.class = &pcibus_class;
1137 	dev_set_name(&child->dev, "%04x:%02x", pci_domain_nr(child), busnr);
1138 
1139 	/* Set up the primary, secondary and subordinate bus numbers */
1140 	child->number = child->busn_res.start = busnr;
1141 	child->primary = parent->busn_res.start;
1142 	child->busn_res.end = 0xff;
1143 
1144 	if (!bridge) {
1145 		child->dev.parent = parent->bridge;
1146 		goto add_dev;
1147 	}
1148 
1149 	child->self = bridge;
1150 	child->bridge = get_device(&bridge->dev);
1151 	child->dev.parent = child->bridge;
1152 	pci_set_bus_of_node(child);
1153 	pci_set_bus_speed(child);
1154 
1155 	/*
1156 	 * Check whether extended config space is accessible on the child
1157 	 * bus.  Note that we currently assume it is always accessible on
1158 	 * the root bus.
1159 	 */
1160 	if (!pci_bridge_child_ext_cfg_accessible(bridge)) {
1161 		child->bus_flags |= PCI_BUS_FLAGS_NO_EXTCFG;
1162 		pci_info(child, "extended config space not accessible\n");
1163 	}
1164 
1165 	/* Set up default resource pointers and names */
1166 	for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++) {
1167 		child->resource[i] = &bridge->resource[PCI_BRIDGE_RESOURCES+i];
1168 		child->resource[i]->name = child->name;
1169 	}
1170 	bridge->subordinate = child;
1171 
1172 add_dev:
1173 	pci_set_bus_msi_domain(child);
1174 	ret = device_register(&child->dev);
1175 	WARN_ON(ret < 0);
1176 
1177 	pcibios_add_bus(child);
1178 
1179 	if (child->ops->add_bus) {
1180 		ret = child->ops->add_bus(child);
1181 		if (WARN_ON(ret < 0))
1182 			dev_err(&child->dev, "failed to add bus: %d\n", ret);
1183 	}
1184 
1185 	/* Create legacy_io and legacy_mem files for this bus */
1186 	pci_create_legacy_files(child);
1187 
1188 	return child;
1189 }
1190 
1191 struct pci_bus *pci_add_new_bus(struct pci_bus *parent, struct pci_dev *dev,
1192 				int busnr)
1193 {
1194 	struct pci_bus *child;
1195 
1196 	child = pci_alloc_child_bus(parent, dev, busnr);
1197 	if (child) {
1198 		down_write(&pci_bus_sem);
1199 		list_add_tail(&child->node, &parent->children);
1200 		up_write(&pci_bus_sem);
1201 	}
1202 	return child;
1203 }
1204 EXPORT_SYMBOL(pci_add_new_bus);
1205 
1206 static void pci_enable_crs(struct pci_dev *pdev)
1207 {
1208 	u16 root_cap = 0;
1209 
1210 	/* Enable CRS Software Visibility if supported */
1211 	pcie_capability_read_word(pdev, PCI_EXP_RTCAP, &root_cap);
1212 	if (root_cap & PCI_EXP_RTCAP_CRSVIS)
1213 		pcie_capability_set_word(pdev, PCI_EXP_RTCTL,
1214 					 PCI_EXP_RTCTL_CRSSVE);
1215 }
1216 
1217 static unsigned int pci_scan_child_bus_extend(struct pci_bus *bus,
1218 					      unsigned int available_buses);
1219 /**
1220  * pci_ea_fixed_busnrs() - Read fixed Secondary and Subordinate bus
1221  * numbers from EA capability.
1222  * @dev: Bridge
1223  * @sec: updated with secondary bus number from EA
1224  * @sub: updated with subordinate bus number from EA
1225  *
1226  * If @dev is a bridge with EA capability that specifies valid secondary
1227  * and subordinate bus numbers, return true with the bus numbers in @sec
1228  * and @sub.  Otherwise return false.
1229  */
1230 static bool pci_ea_fixed_busnrs(struct pci_dev *dev, u8 *sec, u8 *sub)
1231 {
1232 	int ea, offset;
1233 	u32 dw;
1234 	u8 ea_sec, ea_sub;
1235 
1236 	if (dev->hdr_type != PCI_HEADER_TYPE_BRIDGE)
1237 		return false;
1238 
1239 	/* find PCI EA capability in list */
1240 	ea = pci_find_capability(dev, PCI_CAP_ID_EA);
1241 	if (!ea)
1242 		return false;
1243 
1244 	offset = ea + PCI_EA_FIRST_ENT;
1245 	pci_read_config_dword(dev, offset, &dw);
1246 	ea_sec = FIELD_GET(PCI_EA_SEC_BUS_MASK, dw);
1247 	ea_sub = FIELD_GET(PCI_EA_SUB_BUS_MASK, dw);
1248 	if (ea_sec  == 0 || ea_sub < ea_sec)
1249 		return false;
1250 
1251 	*sec = ea_sec;
1252 	*sub = ea_sub;
1253 	return true;
1254 }
1255 
1256 /*
1257  * pci_scan_bridge_extend() - Scan buses behind a bridge
1258  * @bus: Parent bus the bridge is on
1259  * @dev: Bridge itself
1260  * @max: Starting subordinate number of buses behind this bridge
1261  * @available_buses: Total number of buses available for this bridge and
1262  *		     the devices below. After the minimal bus space has
1263  *		     been allocated the remaining buses will be
1264  *		     distributed equally between hotplug-capable bridges.
1265  * @pass: Either %0 (scan already configured bridges) or %1 (scan bridges
1266  *        that need to be reconfigured.
1267  *
1268  * If it's a bridge, configure it and scan the bus behind it.
1269  * For CardBus bridges, we don't scan behind as the devices will
1270  * be handled by the bridge driver itself.
1271  *
1272  * We need to process bridges in two passes -- first we scan those
1273  * already configured by the BIOS and after we are done with all of
1274  * them, we proceed to assigning numbers to the remaining buses in
1275  * order to avoid overlaps between old and new bus numbers.
1276  *
1277  * Return: New subordinate number covering all buses behind this bridge.
1278  */
1279 static int pci_scan_bridge_extend(struct pci_bus *bus, struct pci_dev *dev,
1280 				  int max, unsigned int available_buses,
1281 				  int pass)
1282 {
1283 	struct pci_bus *child;
1284 	int is_cardbus = (dev->hdr_type == PCI_HEADER_TYPE_CARDBUS);
1285 	u32 buses, i, j = 0;
1286 	u16 bctl;
1287 	u8 primary, secondary, subordinate;
1288 	int broken = 0;
1289 	bool fixed_buses;
1290 	u8 fixed_sec, fixed_sub;
1291 	int next_busnr;
1292 
1293 	/*
1294 	 * Make sure the bridge is powered on to be able to access config
1295 	 * space of devices below it.
1296 	 */
1297 	pm_runtime_get_sync(&dev->dev);
1298 
1299 	pci_read_config_dword(dev, PCI_PRIMARY_BUS, &buses);
1300 	primary = buses & 0xFF;
1301 	secondary = (buses >> 8) & 0xFF;
1302 	subordinate = (buses >> 16) & 0xFF;
1303 
1304 	pci_dbg(dev, "scanning [bus %02x-%02x] behind bridge, pass %d\n",
1305 		secondary, subordinate, pass);
1306 
1307 	if (!primary && (primary != bus->number) && secondary && subordinate) {
1308 		pci_warn(dev, "Primary bus is hard wired to 0\n");
1309 		primary = bus->number;
1310 	}
1311 
1312 	/* Check if setup is sensible at all */
1313 	if (!pass &&
1314 	    (primary != bus->number || secondary <= bus->number ||
1315 	     secondary > subordinate)) {
1316 		pci_info(dev, "bridge configuration invalid ([bus %02x-%02x]), reconfiguring\n",
1317 			 secondary, subordinate);
1318 		broken = 1;
1319 	}
1320 
1321 	/*
1322 	 * Disable Master-Abort Mode during probing to avoid reporting of
1323 	 * bus errors in some architectures.
1324 	 */
1325 	pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &bctl);
1326 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL,
1327 			      bctl & ~PCI_BRIDGE_CTL_MASTER_ABORT);
1328 
1329 	pci_enable_crs(dev);
1330 
1331 	if ((secondary || subordinate) && !pcibios_assign_all_busses() &&
1332 	    !is_cardbus && !broken) {
1333 		unsigned int cmax, buses;
1334 
1335 		/*
1336 		 * Bus already configured by firmware, process it in the
1337 		 * first pass and just note the configuration.
1338 		 */
1339 		if (pass)
1340 			goto out;
1341 
1342 		/*
1343 		 * The bus might already exist for two reasons: Either we
1344 		 * are rescanning the bus or the bus is reachable through
1345 		 * more than one bridge. The second case can happen with
1346 		 * the i450NX chipset.
1347 		 */
1348 		child = pci_find_bus(pci_domain_nr(bus), secondary);
1349 		if (!child) {
1350 			child = pci_add_new_bus(bus, dev, secondary);
1351 			if (!child)
1352 				goto out;
1353 			child->primary = primary;
1354 			pci_bus_insert_busn_res(child, secondary, subordinate);
1355 			child->bridge_ctl = bctl;
1356 		}
1357 
1358 		buses = subordinate - secondary;
1359 		cmax = pci_scan_child_bus_extend(child, buses);
1360 		if (cmax > subordinate)
1361 			pci_warn(dev, "bridge has subordinate %02x but max busn %02x\n",
1362 				 subordinate, cmax);
1363 
1364 		/* Subordinate should equal child->busn_res.end */
1365 		if (subordinate > max)
1366 			max = subordinate;
1367 	} else {
1368 
1369 		/*
1370 		 * We need to assign a number to this bus which we always
1371 		 * do in the second pass.
1372 		 */
1373 		if (!pass) {
1374 			if (pcibios_assign_all_busses() || broken || is_cardbus)
1375 
1376 				/*
1377 				 * Temporarily disable forwarding of the
1378 				 * configuration cycles on all bridges in
1379 				 * this bus segment to avoid possible
1380 				 * conflicts in the second pass between two
1381 				 * bridges programmed with overlapping bus
1382 				 * ranges.
1383 				 */
1384 				pci_write_config_dword(dev, PCI_PRIMARY_BUS,
1385 						       buses & ~0xffffff);
1386 			goto out;
1387 		}
1388 
1389 		/* Clear errors */
1390 		pci_write_config_word(dev, PCI_STATUS, 0xffff);
1391 
1392 		/* Read bus numbers from EA Capability (if present) */
1393 		fixed_buses = pci_ea_fixed_busnrs(dev, &fixed_sec, &fixed_sub);
1394 		if (fixed_buses)
1395 			next_busnr = fixed_sec;
1396 		else
1397 			next_busnr = max + 1;
1398 
1399 		/*
1400 		 * Prevent assigning a bus number that already exists.
1401 		 * This can happen when a bridge is hot-plugged, so in this
1402 		 * case we only re-scan this bus.
1403 		 */
1404 		child = pci_find_bus(pci_domain_nr(bus), next_busnr);
1405 		if (!child) {
1406 			child = pci_add_new_bus(bus, dev, next_busnr);
1407 			if (!child)
1408 				goto out;
1409 			pci_bus_insert_busn_res(child, next_busnr,
1410 						bus->busn_res.end);
1411 		}
1412 		max++;
1413 		if (available_buses)
1414 			available_buses--;
1415 
1416 		buses = (buses & 0xff000000)
1417 		      | ((unsigned int)(child->primary)     <<  0)
1418 		      | ((unsigned int)(child->busn_res.start)   <<  8)
1419 		      | ((unsigned int)(child->busn_res.end) << 16);
1420 
1421 		/*
1422 		 * yenta.c forces a secondary latency timer of 176.
1423 		 * Copy that behaviour here.
1424 		 */
1425 		if (is_cardbus) {
1426 			buses &= ~0xff000000;
1427 			buses |= CARDBUS_LATENCY_TIMER << 24;
1428 		}
1429 
1430 		/* We need to blast all three values with a single write */
1431 		pci_write_config_dword(dev, PCI_PRIMARY_BUS, buses);
1432 
1433 		if (!is_cardbus) {
1434 			child->bridge_ctl = bctl;
1435 			max = pci_scan_child_bus_extend(child, available_buses);
1436 		} else {
1437 
1438 			/*
1439 			 * For CardBus bridges, we leave 4 bus numbers as
1440 			 * cards with a PCI-to-PCI bridge can be inserted
1441 			 * later.
1442 			 */
1443 			for (i = 0; i < CARDBUS_RESERVE_BUSNR; i++) {
1444 				struct pci_bus *parent = bus;
1445 				if (pci_find_bus(pci_domain_nr(bus),
1446 							max+i+1))
1447 					break;
1448 				while (parent->parent) {
1449 					if ((!pcibios_assign_all_busses()) &&
1450 					    (parent->busn_res.end > max) &&
1451 					    (parent->busn_res.end <= max+i)) {
1452 						j = 1;
1453 					}
1454 					parent = parent->parent;
1455 				}
1456 				if (j) {
1457 
1458 					/*
1459 					 * Often, there are two CardBus
1460 					 * bridges -- try to leave one
1461 					 * valid bus number for each one.
1462 					 */
1463 					i /= 2;
1464 					break;
1465 				}
1466 			}
1467 			max += i;
1468 		}
1469 
1470 		/*
1471 		 * Set subordinate bus number to its real value.
1472 		 * If fixed subordinate bus number exists from EA
1473 		 * capability then use it.
1474 		 */
1475 		if (fixed_buses)
1476 			max = fixed_sub;
1477 		pci_bus_update_busn_res_end(child, max);
1478 		pci_write_config_byte(dev, PCI_SUBORDINATE_BUS, max);
1479 	}
1480 
1481 	sprintf(child->name,
1482 		(is_cardbus ? "PCI CardBus %04x:%02x" : "PCI Bus %04x:%02x"),
1483 		pci_domain_nr(bus), child->number);
1484 
1485 	/* Check that all devices are accessible */
1486 	while (bus->parent) {
1487 		if ((child->busn_res.end > bus->busn_res.end) ||
1488 		    (child->number > bus->busn_res.end) ||
1489 		    (child->number < bus->number) ||
1490 		    (child->busn_res.end < bus->number)) {
1491 			dev_info(&dev->dev, "devices behind bridge are unusable because %pR cannot be assigned for them\n",
1492 				 &child->busn_res);
1493 			break;
1494 		}
1495 		bus = bus->parent;
1496 	}
1497 
1498 out:
1499 	/* Clear errors in the Secondary Status Register */
1500 	pci_write_config_word(dev, PCI_SEC_STATUS, 0xffff);
1501 
1502 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL, bctl);
1503 
1504 	pm_runtime_put(&dev->dev);
1505 
1506 	return max;
1507 }
1508 
1509 /*
1510  * pci_scan_bridge() - Scan buses behind a bridge
1511  * @bus: Parent bus the bridge is on
1512  * @dev: Bridge itself
1513  * @max: Starting subordinate number of buses behind this bridge
1514  * @pass: Either %0 (scan already configured bridges) or %1 (scan bridges
1515  *        that need to be reconfigured.
1516  *
1517  * If it's a bridge, configure it and scan the bus behind it.
1518  * For CardBus bridges, we don't scan behind as the devices will
1519  * be handled by the bridge driver itself.
1520  *
1521  * We need to process bridges in two passes -- first we scan those
1522  * already configured by the BIOS and after we are done with all of
1523  * them, we proceed to assigning numbers to the remaining buses in
1524  * order to avoid overlaps between old and new bus numbers.
1525  *
1526  * Return: New subordinate number covering all buses behind this bridge.
1527  */
1528 int pci_scan_bridge(struct pci_bus *bus, struct pci_dev *dev, int max, int pass)
1529 {
1530 	return pci_scan_bridge_extend(bus, dev, max, 0, pass);
1531 }
1532 EXPORT_SYMBOL(pci_scan_bridge);
1533 
1534 /*
1535  * Read interrupt line and base address registers.
1536  * The architecture-dependent code can tweak these, of course.
1537  */
1538 static void pci_read_irq(struct pci_dev *dev)
1539 {
1540 	unsigned char irq;
1541 
1542 	/* VFs are not allowed to use INTx, so skip the config reads */
1543 	if (dev->is_virtfn) {
1544 		dev->pin = 0;
1545 		dev->irq = 0;
1546 		return;
1547 	}
1548 
1549 	pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &irq);
1550 	dev->pin = irq;
1551 	if (irq)
1552 		pci_read_config_byte(dev, PCI_INTERRUPT_LINE, &irq);
1553 	dev->irq = irq;
1554 }
1555 
1556 void set_pcie_port_type(struct pci_dev *pdev)
1557 {
1558 	int pos;
1559 	u16 reg16;
1560 	u32 reg32;
1561 	int type;
1562 	struct pci_dev *parent;
1563 
1564 	pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
1565 	if (!pos)
1566 		return;
1567 
1568 	pdev->pcie_cap = pos;
1569 	pci_read_config_word(pdev, pos + PCI_EXP_FLAGS, &reg16);
1570 	pdev->pcie_flags_reg = reg16;
1571 	pci_read_config_dword(pdev, pos + PCI_EXP_DEVCAP, &pdev->devcap);
1572 	pdev->pcie_mpss = FIELD_GET(PCI_EXP_DEVCAP_PAYLOAD, pdev->devcap);
1573 
1574 	pcie_capability_read_dword(pdev, PCI_EXP_LNKCAP, &reg32);
1575 	if (reg32 & PCI_EXP_LNKCAP_DLLLARC)
1576 		pdev->link_active_reporting = 1;
1577 
1578 	parent = pci_upstream_bridge(pdev);
1579 	if (!parent)
1580 		return;
1581 
1582 	/*
1583 	 * Some systems do not identify their upstream/downstream ports
1584 	 * correctly so detect impossible configurations here and correct
1585 	 * the port type accordingly.
1586 	 */
1587 	type = pci_pcie_type(pdev);
1588 	if (type == PCI_EXP_TYPE_DOWNSTREAM) {
1589 		/*
1590 		 * If pdev claims to be downstream port but the parent
1591 		 * device is also downstream port assume pdev is actually
1592 		 * upstream port.
1593 		 */
1594 		if (pcie_downstream_port(parent)) {
1595 			pci_info(pdev, "claims to be downstream port but is acting as upstream port, correcting type\n");
1596 			pdev->pcie_flags_reg &= ~PCI_EXP_FLAGS_TYPE;
1597 			pdev->pcie_flags_reg |= PCI_EXP_TYPE_UPSTREAM;
1598 		}
1599 	} else if (type == PCI_EXP_TYPE_UPSTREAM) {
1600 		/*
1601 		 * If pdev claims to be upstream port but the parent
1602 		 * device is also upstream port assume pdev is actually
1603 		 * downstream port.
1604 		 */
1605 		if (pci_pcie_type(parent) == PCI_EXP_TYPE_UPSTREAM) {
1606 			pci_info(pdev, "claims to be upstream port but is acting as downstream port, correcting type\n");
1607 			pdev->pcie_flags_reg &= ~PCI_EXP_FLAGS_TYPE;
1608 			pdev->pcie_flags_reg |= PCI_EXP_TYPE_DOWNSTREAM;
1609 		}
1610 	}
1611 }
1612 
1613 void set_pcie_hotplug_bridge(struct pci_dev *pdev)
1614 {
1615 	u32 reg32;
1616 
1617 	pcie_capability_read_dword(pdev, PCI_EXP_SLTCAP, &reg32);
1618 	if (reg32 & PCI_EXP_SLTCAP_HPC)
1619 		pdev->is_hotplug_bridge = 1;
1620 }
1621 
1622 static void set_pcie_thunderbolt(struct pci_dev *dev)
1623 {
1624 	u16 vsec;
1625 
1626 	/* Is the device part of a Thunderbolt controller? */
1627 	vsec = pci_find_vsec_capability(dev, PCI_VENDOR_ID_INTEL, PCI_VSEC_ID_INTEL_TBT);
1628 	if (vsec)
1629 		dev->is_thunderbolt = 1;
1630 }
1631 
1632 static void set_pcie_untrusted(struct pci_dev *dev)
1633 {
1634 	struct pci_dev *parent;
1635 
1636 	/*
1637 	 * If the upstream bridge is untrusted we treat this device
1638 	 * untrusted as well.
1639 	 */
1640 	parent = pci_upstream_bridge(dev);
1641 	if (parent && (parent->untrusted || parent->external_facing))
1642 		dev->untrusted = true;
1643 }
1644 
1645 static void pci_set_removable(struct pci_dev *dev)
1646 {
1647 	struct pci_dev *parent = pci_upstream_bridge(dev);
1648 
1649 	/*
1650 	 * We (only) consider everything downstream from an external_facing
1651 	 * device to be removable by the user. We're mainly concerned with
1652 	 * consumer platforms with user accessible thunderbolt ports that are
1653 	 * vulnerable to DMA attacks, and we expect those ports to be marked by
1654 	 * the firmware as external_facing. Devices in traditional hotplug
1655 	 * slots can technically be removed, but the expectation is that unless
1656 	 * the port is marked with external_facing, such devices are less
1657 	 * accessible to user / may not be removed by end user, and thus not
1658 	 * exposed as "removable" to userspace.
1659 	 */
1660 	if (parent &&
1661 	    (parent->external_facing || dev_is_removable(&parent->dev)))
1662 		dev_set_removable(&dev->dev, DEVICE_REMOVABLE);
1663 }
1664 
1665 /**
1666  * pci_ext_cfg_is_aliased - Is ext config space just an alias of std config?
1667  * @dev: PCI device
1668  *
1669  * PCI Express to PCI/PCI-X Bridge Specification, rev 1.0, 4.1.4 says that
1670  * when forwarding a type1 configuration request the bridge must check that
1671  * the extended register address field is zero.  The bridge is not permitted
1672  * to forward the transactions and must handle it as an Unsupported Request.
1673  * Some bridges do not follow this rule and simply drop the extended register
1674  * bits, resulting in the standard config space being aliased, every 256
1675  * bytes across the entire configuration space.  Test for this condition by
1676  * comparing the first dword of each potential alias to the vendor/device ID.
1677  * Known offenders:
1678  *   ASM1083/1085 PCIe-to-PCI Reversible Bridge (1b21:1080, rev 01 & 03)
1679  *   AMD/ATI SBx00 PCI to PCI Bridge (1002:4384, rev 40)
1680  */
1681 static bool pci_ext_cfg_is_aliased(struct pci_dev *dev)
1682 {
1683 #ifdef CONFIG_PCI_QUIRKS
1684 	int pos, ret;
1685 	u32 header, tmp;
1686 
1687 	pci_read_config_dword(dev, PCI_VENDOR_ID, &header);
1688 
1689 	for (pos = PCI_CFG_SPACE_SIZE;
1690 	     pos < PCI_CFG_SPACE_EXP_SIZE; pos += PCI_CFG_SPACE_SIZE) {
1691 		ret = pci_read_config_dword(dev, pos, &tmp);
1692 		if ((ret != PCIBIOS_SUCCESSFUL) || (header != tmp))
1693 			return false;
1694 	}
1695 
1696 	return true;
1697 #else
1698 	return false;
1699 #endif
1700 }
1701 
1702 /**
1703  * pci_cfg_space_size_ext - Get the configuration space size of the PCI device
1704  * @dev: PCI device
1705  *
1706  * Regular PCI devices have 256 bytes, but PCI-X 2 and PCI Express devices
1707  * have 4096 bytes.  Even if the device is capable, that doesn't mean we can
1708  * access it.  Maybe we don't have a way to generate extended config space
1709  * accesses, or the device is behind a reverse Express bridge.  So we try
1710  * reading the dword at 0x100 which must either be 0 or a valid extended
1711  * capability header.
1712  */
1713 static int pci_cfg_space_size_ext(struct pci_dev *dev)
1714 {
1715 	u32 status;
1716 	int pos = PCI_CFG_SPACE_SIZE;
1717 
1718 	if (pci_read_config_dword(dev, pos, &status) != PCIBIOS_SUCCESSFUL)
1719 		return PCI_CFG_SPACE_SIZE;
1720 	if (PCI_POSSIBLE_ERROR(status) || pci_ext_cfg_is_aliased(dev))
1721 		return PCI_CFG_SPACE_SIZE;
1722 
1723 	return PCI_CFG_SPACE_EXP_SIZE;
1724 }
1725 
1726 int pci_cfg_space_size(struct pci_dev *dev)
1727 {
1728 	int pos;
1729 	u32 status;
1730 	u16 class;
1731 
1732 #ifdef CONFIG_PCI_IOV
1733 	/*
1734 	 * Per the SR-IOV specification (rev 1.1, sec 3.5), VFs are required to
1735 	 * implement a PCIe capability and therefore must implement extended
1736 	 * config space.  We can skip the NO_EXTCFG test below and the
1737 	 * reachability/aliasing test in pci_cfg_space_size_ext() by virtue of
1738 	 * the fact that the SR-IOV capability on the PF resides in extended
1739 	 * config space and must be accessible and non-aliased to have enabled
1740 	 * support for this VF.  This is a micro performance optimization for
1741 	 * systems supporting many VFs.
1742 	 */
1743 	if (dev->is_virtfn)
1744 		return PCI_CFG_SPACE_EXP_SIZE;
1745 #endif
1746 
1747 	if (dev->bus->bus_flags & PCI_BUS_FLAGS_NO_EXTCFG)
1748 		return PCI_CFG_SPACE_SIZE;
1749 
1750 	class = dev->class >> 8;
1751 	if (class == PCI_CLASS_BRIDGE_HOST)
1752 		return pci_cfg_space_size_ext(dev);
1753 
1754 	if (pci_is_pcie(dev))
1755 		return pci_cfg_space_size_ext(dev);
1756 
1757 	pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1758 	if (!pos)
1759 		return PCI_CFG_SPACE_SIZE;
1760 
1761 	pci_read_config_dword(dev, pos + PCI_X_STATUS, &status);
1762 	if (status & (PCI_X_STATUS_266MHZ | PCI_X_STATUS_533MHZ))
1763 		return pci_cfg_space_size_ext(dev);
1764 
1765 	return PCI_CFG_SPACE_SIZE;
1766 }
1767 
1768 static u32 pci_class(struct pci_dev *dev)
1769 {
1770 	u32 class;
1771 
1772 #ifdef CONFIG_PCI_IOV
1773 	if (dev->is_virtfn)
1774 		return dev->physfn->sriov->class;
1775 #endif
1776 	pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
1777 	return class;
1778 }
1779 
1780 static void pci_subsystem_ids(struct pci_dev *dev, u16 *vendor, u16 *device)
1781 {
1782 #ifdef CONFIG_PCI_IOV
1783 	if (dev->is_virtfn) {
1784 		*vendor = dev->physfn->sriov->subsystem_vendor;
1785 		*device = dev->physfn->sriov->subsystem_device;
1786 		return;
1787 	}
1788 #endif
1789 	pci_read_config_word(dev, PCI_SUBSYSTEM_VENDOR_ID, vendor);
1790 	pci_read_config_word(dev, PCI_SUBSYSTEM_ID, device);
1791 }
1792 
1793 static u8 pci_hdr_type(struct pci_dev *dev)
1794 {
1795 	u8 hdr_type;
1796 
1797 #ifdef CONFIG_PCI_IOV
1798 	if (dev->is_virtfn)
1799 		return dev->physfn->sriov->hdr_type;
1800 #endif
1801 	pci_read_config_byte(dev, PCI_HEADER_TYPE, &hdr_type);
1802 	return hdr_type;
1803 }
1804 
1805 #define LEGACY_IO_RESOURCE	(IORESOURCE_IO | IORESOURCE_PCI_FIXED)
1806 
1807 /**
1808  * pci_intx_mask_broken - Test PCI_COMMAND_INTX_DISABLE writability
1809  * @dev: PCI device
1810  *
1811  * Test whether PCI_COMMAND_INTX_DISABLE is writable for @dev.  Check this
1812  * at enumeration-time to avoid modifying PCI_COMMAND at run-time.
1813  */
1814 static int pci_intx_mask_broken(struct pci_dev *dev)
1815 {
1816 	u16 orig, toggle, new;
1817 
1818 	pci_read_config_word(dev, PCI_COMMAND, &orig);
1819 	toggle = orig ^ PCI_COMMAND_INTX_DISABLE;
1820 	pci_write_config_word(dev, PCI_COMMAND, toggle);
1821 	pci_read_config_word(dev, PCI_COMMAND, &new);
1822 
1823 	pci_write_config_word(dev, PCI_COMMAND, orig);
1824 
1825 	/*
1826 	 * PCI_COMMAND_INTX_DISABLE was reserved and read-only prior to PCI
1827 	 * r2.3, so strictly speaking, a device is not *broken* if it's not
1828 	 * writable.  But we'll live with the misnomer for now.
1829 	 */
1830 	if (new != toggle)
1831 		return 1;
1832 	return 0;
1833 }
1834 
1835 static void early_dump_pci_device(struct pci_dev *pdev)
1836 {
1837 	u32 value[256 / 4];
1838 	int i;
1839 
1840 	pci_info(pdev, "config space:\n");
1841 
1842 	for (i = 0; i < 256; i += 4)
1843 		pci_read_config_dword(pdev, i, &value[i / 4]);
1844 
1845 	print_hex_dump(KERN_INFO, "", DUMP_PREFIX_OFFSET, 16, 1,
1846 		       value, 256, false);
1847 }
1848 
1849 static const char *pci_type_str(struct pci_dev *dev)
1850 {
1851 	static const char * const str[] = {
1852 		"PCIe Endpoint",
1853 		"PCIe Legacy Endpoint",
1854 		"PCIe unknown",
1855 		"PCIe unknown",
1856 		"PCIe Root Port",
1857 		"PCIe Switch Upstream Port",
1858 		"PCIe Switch Downstream Port",
1859 		"PCIe to PCI/PCI-X bridge",
1860 		"PCI/PCI-X to PCIe bridge",
1861 		"PCIe Root Complex Integrated Endpoint",
1862 		"PCIe Root Complex Event Collector",
1863 	};
1864 	int type;
1865 
1866 	if (pci_is_pcie(dev)) {
1867 		type = pci_pcie_type(dev);
1868 		if (type < ARRAY_SIZE(str))
1869 			return str[type];
1870 
1871 		return "PCIe unknown";
1872 	}
1873 
1874 	switch (dev->hdr_type) {
1875 	case PCI_HEADER_TYPE_NORMAL:
1876 		return "conventional PCI endpoint";
1877 	case PCI_HEADER_TYPE_BRIDGE:
1878 		return "conventional PCI bridge";
1879 	case PCI_HEADER_TYPE_CARDBUS:
1880 		return "CardBus bridge";
1881 	default:
1882 		return "conventional PCI";
1883 	}
1884 }
1885 
1886 /**
1887  * pci_setup_device - Fill in class and map information of a device
1888  * @dev: the device structure to fill
1889  *
1890  * Initialize the device structure with information about the device's
1891  * vendor,class,memory and IO-space addresses, IRQ lines etc.
1892  * Called at initialisation of the PCI subsystem and by CardBus services.
1893  * Returns 0 on success and negative if unknown type of device (not normal,
1894  * bridge or CardBus).
1895  */
1896 int pci_setup_device(struct pci_dev *dev)
1897 {
1898 	u32 class;
1899 	u16 cmd;
1900 	u8 hdr_type;
1901 	int err, pos = 0;
1902 	struct pci_bus_region region;
1903 	struct resource *res;
1904 
1905 	hdr_type = pci_hdr_type(dev);
1906 
1907 	dev->sysdata = dev->bus->sysdata;
1908 	dev->dev.parent = dev->bus->bridge;
1909 	dev->dev.bus = &pci_bus_type;
1910 	dev->hdr_type = hdr_type & 0x7f;
1911 	dev->multifunction = !!(hdr_type & 0x80);
1912 	dev->error_state = pci_channel_io_normal;
1913 	set_pcie_port_type(dev);
1914 
1915 	err = pci_set_of_node(dev);
1916 	if (err)
1917 		return err;
1918 	pci_set_acpi_fwnode(dev);
1919 
1920 	pci_dev_assign_slot(dev);
1921 
1922 	/*
1923 	 * Assume 32-bit PCI; let 64-bit PCI cards (which are far rarer)
1924 	 * set this higher, assuming the system even supports it.
1925 	 */
1926 	dev->dma_mask = 0xffffffff;
1927 
1928 	dev_set_name(&dev->dev, "%04x:%02x:%02x.%d", pci_domain_nr(dev->bus),
1929 		     dev->bus->number, PCI_SLOT(dev->devfn),
1930 		     PCI_FUNC(dev->devfn));
1931 
1932 	class = pci_class(dev);
1933 
1934 	dev->revision = class & 0xff;
1935 	dev->class = class >> 8;		    /* upper 3 bytes */
1936 
1937 	if (pci_early_dump)
1938 		early_dump_pci_device(dev);
1939 
1940 	/* Need to have dev->class ready */
1941 	dev->cfg_size = pci_cfg_space_size(dev);
1942 
1943 	/* Need to have dev->cfg_size ready */
1944 	set_pcie_thunderbolt(dev);
1945 
1946 	set_pcie_untrusted(dev);
1947 
1948 	/* "Unknown power state" */
1949 	dev->current_state = PCI_UNKNOWN;
1950 
1951 	/* Early fixups, before probing the BARs */
1952 	pci_fixup_device(pci_fixup_early, dev);
1953 
1954 	pci_set_removable(dev);
1955 
1956 	pci_info(dev, "[%04x:%04x] type %02x class %#08x %s\n",
1957 		 dev->vendor, dev->device, dev->hdr_type, dev->class,
1958 		 pci_type_str(dev));
1959 
1960 	/* Device class may be changed after fixup */
1961 	class = dev->class >> 8;
1962 
1963 	if (dev->non_compliant_bars && !dev->mmio_always_on) {
1964 		pci_read_config_word(dev, PCI_COMMAND, &cmd);
1965 		if (cmd & (PCI_COMMAND_IO | PCI_COMMAND_MEMORY)) {
1966 			pci_info(dev, "device has non-compliant BARs; disabling IO/MEM decoding\n");
1967 			cmd &= ~PCI_COMMAND_IO;
1968 			cmd &= ~PCI_COMMAND_MEMORY;
1969 			pci_write_config_word(dev, PCI_COMMAND, cmd);
1970 		}
1971 	}
1972 
1973 	dev->broken_intx_masking = pci_intx_mask_broken(dev);
1974 
1975 	switch (dev->hdr_type) {		    /* header type */
1976 	case PCI_HEADER_TYPE_NORMAL:		    /* standard header */
1977 		if (class == PCI_CLASS_BRIDGE_PCI)
1978 			goto bad;
1979 		pci_read_irq(dev);
1980 		pci_read_bases(dev, 6, PCI_ROM_ADDRESS);
1981 
1982 		pci_subsystem_ids(dev, &dev->subsystem_vendor, &dev->subsystem_device);
1983 
1984 		/*
1985 		 * Do the ugly legacy mode stuff here rather than broken chip
1986 		 * quirk code. Legacy mode ATA controllers have fixed
1987 		 * addresses. These are not always echoed in BAR0-3, and
1988 		 * BAR0-3 in a few cases contain junk!
1989 		 */
1990 		if (class == PCI_CLASS_STORAGE_IDE) {
1991 			u8 progif;
1992 			pci_read_config_byte(dev, PCI_CLASS_PROG, &progif);
1993 			if ((progif & 1) == 0) {
1994 				region.start = 0x1F0;
1995 				region.end = 0x1F7;
1996 				res = &dev->resource[0];
1997 				res->flags = LEGACY_IO_RESOURCE;
1998 				pcibios_bus_to_resource(dev->bus, res, &region);
1999 				pci_info(dev, "BAR 0 %pR: legacy IDE quirk\n",
2000 					 res);
2001 				region.start = 0x3F6;
2002 				region.end = 0x3F6;
2003 				res = &dev->resource[1];
2004 				res->flags = LEGACY_IO_RESOURCE;
2005 				pcibios_bus_to_resource(dev->bus, res, &region);
2006 				pci_info(dev, "BAR 1 %pR: legacy IDE quirk\n",
2007 					 res);
2008 			}
2009 			if ((progif & 4) == 0) {
2010 				region.start = 0x170;
2011 				region.end = 0x177;
2012 				res = &dev->resource[2];
2013 				res->flags = LEGACY_IO_RESOURCE;
2014 				pcibios_bus_to_resource(dev->bus, res, &region);
2015 				pci_info(dev, "BAR 2 %pR: legacy IDE quirk\n",
2016 					 res);
2017 				region.start = 0x376;
2018 				region.end = 0x376;
2019 				res = &dev->resource[3];
2020 				res->flags = LEGACY_IO_RESOURCE;
2021 				pcibios_bus_to_resource(dev->bus, res, &region);
2022 				pci_info(dev, "BAR 3 %pR: legacy IDE quirk\n",
2023 					 res);
2024 			}
2025 		}
2026 		break;
2027 
2028 	case PCI_HEADER_TYPE_BRIDGE:		    /* bridge header */
2029 		/*
2030 		 * The PCI-to-PCI bridge spec requires that subtractive
2031 		 * decoding (i.e. transparent) bridge must have programming
2032 		 * interface code of 0x01.
2033 		 */
2034 		pci_read_irq(dev);
2035 		dev->transparent = ((dev->class & 0xff) == 1);
2036 		pci_read_bases(dev, 2, PCI_ROM_ADDRESS1);
2037 		pci_read_bridge_windows(dev);
2038 		set_pcie_hotplug_bridge(dev);
2039 		pos = pci_find_capability(dev, PCI_CAP_ID_SSVID);
2040 		if (pos) {
2041 			pci_read_config_word(dev, pos + PCI_SSVID_VENDOR_ID, &dev->subsystem_vendor);
2042 			pci_read_config_word(dev, pos + PCI_SSVID_DEVICE_ID, &dev->subsystem_device);
2043 		}
2044 		break;
2045 
2046 	case PCI_HEADER_TYPE_CARDBUS:		    /* CardBus bridge header */
2047 		if (class != PCI_CLASS_BRIDGE_CARDBUS)
2048 			goto bad;
2049 		pci_read_irq(dev);
2050 		pci_read_bases(dev, 1, 0);
2051 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_VENDOR_ID, &dev->subsystem_vendor);
2052 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_ID, &dev->subsystem_device);
2053 		break;
2054 
2055 	default:				    /* unknown header */
2056 		pci_err(dev, "unknown header type %02x, ignoring device\n",
2057 			dev->hdr_type);
2058 		pci_release_of_node(dev);
2059 		return -EIO;
2060 
2061 	bad:
2062 		pci_err(dev, "ignoring class %#08x (doesn't match header type %02x)\n",
2063 			dev->class, dev->hdr_type);
2064 		dev->class = PCI_CLASS_NOT_DEFINED << 8;
2065 	}
2066 
2067 	/* We found a fine healthy device, go go go... */
2068 	return 0;
2069 }
2070 
2071 static void pci_configure_mps(struct pci_dev *dev)
2072 {
2073 	struct pci_dev *bridge = pci_upstream_bridge(dev);
2074 	int mps, mpss, p_mps, rc;
2075 
2076 	if (!pci_is_pcie(dev))
2077 		return;
2078 
2079 	/* MPS and MRRS fields are of type 'RsvdP' for VFs, short-circuit out */
2080 	if (dev->is_virtfn)
2081 		return;
2082 
2083 	/*
2084 	 * For Root Complex Integrated Endpoints, program the maximum
2085 	 * supported value unless limited by the PCIE_BUS_PEER2PEER case.
2086 	 */
2087 	if (pci_pcie_type(dev) == PCI_EXP_TYPE_RC_END) {
2088 		if (pcie_bus_config == PCIE_BUS_PEER2PEER)
2089 			mps = 128;
2090 		else
2091 			mps = 128 << dev->pcie_mpss;
2092 		rc = pcie_set_mps(dev, mps);
2093 		if (rc) {
2094 			pci_warn(dev, "can't set Max Payload Size to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
2095 				 mps);
2096 		}
2097 		return;
2098 	}
2099 
2100 	if (!bridge || !pci_is_pcie(bridge))
2101 		return;
2102 
2103 	mps = pcie_get_mps(dev);
2104 	p_mps = pcie_get_mps(bridge);
2105 
2106 	if (mps == p_mps)
2107 		return;
2108 
2109 	if (pcie_bus_config == PCIE_BUS_TUNE_OFF) {
2110 		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",
2111 			 mps, pci_name(bridge), p_mps);
2112 		return;
2113 	}
2114 
2115 	/*
2116 	 * Fancier MPS configuration is done later by
2117 	 * pcie_bus_configure_settings()
2118 	 */
2119 	if (pcie_bus_config != PCIE_BUS_DEFAULT)
2120 		return;
2121 
2122 	mpss = 128 << dev->pcie_mpss;
2123 	if (mpss < p_mps && pci_pcie_type(bridge) == PCI_EXP_TYPE_ROOT_PORT) {
2124 		pcie_set_mps(bridge, mpss);
2125 		pci_info(dev, "Upstream bridge's Max Payload Size set to %d (was %d, max %d)\n",
2126 			 mpss, p_mps, 128 << bridge->pcie_mpss);
2127 		p_mps = pcie_get_mps(bridge);
2128 	}
2129 
2130 	rc = pcie_set_mps(dev, p_mps);
2131 	if (rc) {
2132 		pci_warn(dev, "can't set Max Payload Size to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
2133 			 p_mps);
2134 		return;
2135 	}
2136 
2137 	pci_info(dev, "Max Payload Size set to %d (was %d, max %d)\n",
2138 		 p_mps, mps, mpss);
2139 }
2140 
2141 int pci_configure_extended_tags(struct pci_dev *dev, void *ign)
2142 {
2143 	struct pci_host_bridge *host;
2144 	u32 cap;
2145 	u16 ctl;
2146 	int ret;
2147 
2148 	if (!pci_is_pcie(dev))
2149 		return 0;
2150 
2151 	ret = pcie_capability_read_dword(dev, PCI_EXP_DEVCAP, &cap);
2152 	if (ret)
2153 		return 0;
2154 
2155 	if (!(cap & PCI_EXP_DEVCAP_EXT_TAG))
2156 		return 0;
2157 
2158 	ret = pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &ctl);
2159 	if (ret)
2160 		return 0;
2161 
2162 	host = pci_find_host_bridge(dev->bus);
2163 	if (!host)
2164 		return 0;
2165 
2166 	/*
2167 	 * If some device in the hierarchy doesn't handle Extended Tags
2168 	 * correctly, make sure they're disabled.
2169 	 */
2170 	if (host->no_ext_tags) {
2171 		if (ctl & PCI_EXP_DEVCTL_EXT_TAG) {
2172 			pci_info(dev, "disabling Extended Tags\n");
2173 			pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
2174 						   PCI_EXP_DEVCTL_EXT_TAG);
2175 		}
2176 		return 0;
2177 	}
2178 
2179 	if (!(ctl & PCI_EXP_DEVCTL_EXT_TAG)) {
2180 		pci_info(dev, "enabling Extended Tags\n");
2181 		pcie_capability_set_word(dev, PCI_EXP_DEVCTL,
2182 					 PCI_EXP_DEVCTL_EXT_TAG);
2183 	}
2184 	return 0;
2185 }
2186 
2187 /**
2188  * pcie_relaxed_ordering_enabled - Probe for PCIe relaxed ordering enable
2189  * @dev: PCI device to query
2190  *
2191  * Returns true if the device has enabled relaxed ordering attribute.
2192  */
2193 bool pcie_relaxed_ordering_enabled(struct pci_dev *dev)
2194 {
2195 	u16 v;
2196 
2197 	pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &v);
2198 
2199 	return !!(v & PCI_EXP_DEVCTL_RELAX_EN);
2200 }
2201 EXPORT_SYMBOL(pcie_relaxed_ordering_enabled);
2202 
2203 static void pci_configure_relaxed_ordering(struct pci_dev *dev)
2204 {
2205 	struct pci_dev *root;
2206 
2207 	/* PCI_EXP_DEVCTL_RELAX_EN is RsvdP in VFs */
2208 	if (dev->is_virtfn)
2209 		return;
2210 
2211 	if (!pcie_relaxed_ordering_enabled(dev))
2212 		return;
2213 
2214 	/*
2215 	 * For now, we only deal with Relaxed Ordering issues with Root
2216 	 * Ports. Peer-to-Peer DMA is another can of worms.
2217 	 */
2218 	root = pcie_find_root_port(dev);
2219 	if (!root)
2220 		return;
2221 
2222 	if (root->dev_flags & PCI_DEV_FLAGS_NO_RELAXED_ORDERING) {
2223 		pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
2224 					   PCI_EXP_DEVCTL_RELAX_EN);
2225 		pci_info(dev, "Relaxed Ordering disabled because the Root Port didn't support it\n");
2226 	}
2227 }
2228 
2229 static void pci_configure_eetlp_prefix(struct pci_dev *dev)
2230 {
2231 #ifdef CONFIG_PCI_PASID
2232 	struct pci_dev *bridge;
2233 	int pcie_type;
2234 	u32 cap;
2235 
2236 	if (!pci_is_pcie(dev))
2237 		return;
2238 
2239 	pcie_capability_read_dword(dev, PCI_EXP_DEVCAP2, &cap);
2240 	if (!(cap & PCI_EXP_DEVCAP2_EE_PREFIX))
2241 		return;
2242 
2243 	pcie_type = pci_pcie_type(dev);
2244 	if (pcie_type == PCI_EXP_TYPE_ROOT_PORT ||
2245 	    pcie_type == PCI_EXP_TYPE_RC_END)
2246 		dev->eetlp_prefix_path = 1;
2247 	else {
2248 		bridge = pci_upstream_bridge(dev);
2249 		if (bridge && bridge->eetlp_prefix_path)
2250 			dev->eetlp_prefix_path = 1;
2251 	}
2252 #endif
2253 }
2254 
2255 static void pci_configure_serr(struct pci_dev *dev)
2256 {
2257 	u16 control;
2258 
2259 	if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
2260 
2261 		/*
2262 		 * A bridge will not forward ERR_ messages coming from an
2263 		 * endpoint unless SERR# forwarding is enabled.
2264 		 */
2265 		pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &control);
2266 		if (!(control & PCI_BRIDGE_CTL_SERR)) {
2267 			control |= PCI_BRIDGE_CTL_SERR;
2268 			pci_write_config_word(dev, PCI_BRIDGE_CONTROL, control);
2269 		}
2270 	}
2271 }
2272 
2273 static void pci_configure_device(struct pci_dev *dev)
2274 {
2275 	pci_configure_mps(dev);
2276 	pci_configure_extended_tags(dev, NULL);
2277 	pci_configure_relaxed_ordering(dev);
2278 	pci_configure_ltr(dev);
2279 	pci_configure_aspm_l1ss(dev);
2280 	pci_configure_eetlp_prefix(dev);
2281 	pci_configure_serr(dev);
2282 
2283 	pci_acpi_program_hp_params(dev);
2284 }
2285 
2286 static void pci_release_capabilities(struct pci_dev *dev)
2287 {
2288 	pci_aer_exit(dev);
2289 	pci_rcec_exit(dev);
2290 	pci_iov_release(dev);
2291 	pci_free_cap_save_buffers(dev);
2292 }
2293 
2294 /**
2295  * pci_release_dev - Free a PCI device structure when all users of it are
2296  *		     finished
2297  * @dev: device that's been disconnected
2298  *
2299  * Will be called only by the device core when all users of this PCI device are
2300  * done.
2301  */
2302 static void pci_release_dev(struct device *dev)
2303 {
2304 	struct pci_dev *pci_dev;
2305 
2306 	pci_dev = to_pci_dev(dev);
2307 	pci_release_capabilities(pci_dev);
2308 	pci_release_of_node(pci_dev);
2309 	pcibios_release_device(pci_dev);
2310 	pci_bus_put(pci_dev->bus);
2311 	kfree(pci_dev->driver_override);
2312 	bitmap_free(pci_dev->dma_alias_mask);
2313 	dev_dbg(dev, "device released\n");
2314 	kfree(pci_dev);
2315 }
2316 
2317 static const struct device_type pci_dev_type = {
2318 	.groups = pci_dev_attr_groups,
2319 };
2320 
2321 struct pci_dev *pci_alloc_dev(struct pci_bus *bus)
2322 {
2323 	struct pci_dev *dev;
2324 
2325 	dev = kzalloc(sizeof(struct pci_dev), GFP_KERNEL);
2326 	if (!dev)
2327 		return NULL;
2328 
2329 	INIT_LIST_HEAD(&dev->bus_list);
2330 	dev->dev.type = &pci_dev_type;
2331 	dev->bus = pci_bus_get(bus);
2332 	dev->driver_exclusive_resource = (struct resource) {
2333 		.name = "PCI Exclusive",
2334 		.start = 0,
2335 		.end = -1,
2336 	};
2337 
2338 	spin_lock_init(&dev->pcie_cap_lock);
2339 #ifdef CONFIG_PCI_MSI
2340 	raw_spin_lock_init(&dev->msi_lock);
2341 #endif
2342 	return dev;
2343 }
2344 EXPORT_SYMBOL(pci_alloc_dev);
2345 
2346 static bool pci_bus_crs_vendor_id(u32 l)
2347 {
2348 	return (l & 0xffff) == PCI_VENDOR_ID_PCI_SIG;
2349 }
2350 
2351 static bool pci_bus_wait_crs(struct pci_bus *bus, int devfn, u32 *l,
2352 			     int timeout)
2353 {
2354 	int delay = 1;
2355 
2356 	if (!pci_bus_crs_vendor_id(*l))
2357 		return true;	/* not a CRS completion */
2358 
2359 	if (!timeout)
2360 		return false;	/* CRS, but caller doesn't want to wait */
2361 
2362 	/*
2363 	 * We got the reserved Vendor ID that indicates a completion with
2364 	 * Configuration Request Retry Status (CRS).  Retry until we get a
2365 	 * valid Vendor ID or we time out.
2366 	 */
2367 	while (pci_bus_crs_vendor_id(*l)) {
2368 		if (delay > timeout) {
2369 			pr_warn("pci %04x:%02x:%02x.%d: not ready after %dms; giving up\n",
2370 				pci_domain_nr(bus), bus->number,
2371 				PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2372 
2373 			return false;
2374 		}
2375 		if (delay >= 1000)
2376 			pr_info("pci %04x:%02x:%02x.%d: not ready after %dms; waiting\n",
2377 				pci_domain_nr(bus), bus->number,
2378 				PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2379 
2380 		msleep(delay);
2381 		delay *= 2;
2382 
2383 		if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
2384 			return false;
2385 	}
2386 
2387 	if (delay >= 1000)
2388 		pr_info("pci %04x:%02x:%02x.%d: ready after %dms\n",
2389 			pci_domain_nr(bus), bus->number,
2390 			PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2391 
2392 	return true;
2393 }
2394 
2395 bool pci_bus_generic_read_dev_vendor_id(struct pci_bus *bus, int devfn, u32 *l,
2396 					int timeout)
2397 {
2398 	if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
2399 		return false;
2400 
2401 	/* Some broken boards return 0 or ~0 (PCI_ERROR_RESPONSE) if a slot is empty: */
2402 	if (PCI_POSSIBLE_ERROR(*l) || *l == 0x00000000 ||
2403 	    *l == 0x0000ffff || *l == 0xffff0000)
2404 		return false;
2405 
2406 	if (pci_bus_crs_vendor_id(*l))
2407 		return pci_bus_wait_crs(bus, devfn, l, timeout);
2408 
2409 	return true;
2410 }
2411 
2412 bool pci_bus_read_dev_vendor_id(struct pci_bus *bus, int devfn, u32 *l,
2413 				int timeout)
2414 {
2415 #ifdef CONFIG_PCI_QUIRKS
2416 	struct pci_dev *bridge = bus->self;
2417 
2418 	/*
2419 	 * Certain IDT switches have an issue where they improperly trigger
2420 	 * ACS Source Validation errors on completions for config reads.
2421 	 */
2422 	if (bridge && bridge->vendor == PCI_VENDOR_ID_IDT &&
2423 	    bridge->device == 0x80b5)
2424 		return pci_idt_bus_quirk(bus, devfn, l, timeout);
2425 #endif
2426 
2427 	return pci_bus_generic_read_dev_vendor_id(bus, devfn, l, timeout);
2428 }
2429 EXPORT_SYMBOL(pci_bus_read_dev_vendor_id);
2430 
2431 /*
2432  * Read the config data for a PCI device, sanity-check it,
2433  * and fill in the dev structure.
2434  */
2435 static struct pci_dev *pci_scan_device(struct pci_bus *bus, int devfn)
2436 {
2437 	struct pci_dev *dev;
2438 	u32 l;
2439 
2440 	if (!pci_bus_read_dev_vendor_id(bus, devfn, &l, 60*1000))
2441 		return NULL;
2442 
2443 	dev = pci_alloc_dev(bus);
2444 	if (!dev)
2445 		return NULL;
2446 
2447 	dev->devfn = devfn;
2448 	dev->vendor = l & 0xffff;
2449 	dev->device = (l >> 16) & 0xffff;
2450 
2451 	if (pci_setup_device(dev)) {
2452 		pci_bus_put(dev->bus);
2453 		kfree(dev);
2454 		return NULL;
2455 	}
2456 
2457 	return dev;
2458 }
2459 
2460 void pcie_report_downtraining(struct pci_dev *dev)
2461 {
2462 	if (!pci_is_pcie(dev))
2463 		return;
2464 
2465 	/* Look from the device up to avoid downstream ports with no devices */
2466 	if ((pci_pcie_type(dev) != PCI_EXP_TYPE_ENDPOINT) &&
2467 	    (pci_pcie_type(dev) != PCI_EXP_TYPE_LEG_END) &&
2468 	    (pci_pcie_type(dev) != PCI_EXP_TYPE_UPSTREAM))
2469 		return;
2470 
2471 	/* Multi-function PCIe devices share the same link/status */
2472 	if (PCI_FUNC(dev->devfn) != 0 || dev->is_virtfn)
2473 		return;
2474 
2475 	/* Print link status only if the device is constrained by the fabric */
2476 	__pcie_print_link_status(dev, false);
2477 }
2478 
2479 static void pci_init_capabilities(struct pci_dev *dev)
2480 {
2481 	pci_ea_init(dev);		/* Enhanced Allocation */
2482 	pci_msi_init(dev);		/* Disable MSI */
2483 	pci_msix_init(dev);		/* Disable MSI-X */
2484 
2485 	/* Buffers for saving PCIe and PCI-X capabilities */
2486 	pci_allocate_cap_save_buffers(dev);
2487 
2488 	pci_pm_init(dev);		/* Power Management */
2489 	pci_vpd_init(dev);		/* Vital Product Data */
2490 	pci_configure_ari(dev);		/* Alternative Routing-ID Forwarding */
2491 	pci_iov_init(dev);		/* Single Root I/O Virtualization */
2492 	pci_ats_init(dev);		/* Address Translation Services */
2493 	pci_pri_init(dev);		/* Page Request Interface */
2494 	pci_pasid_init(dev);		/* Process Address Space ID */
2495 	pci_acs_init(dev);		/* Access Control Services */
2496 	pci_ptm_init(dev);		/* Precision Time Measurement */
2497 	pci_aer_init(dev);		/* Advanced Error Reporting */
2498 	pci_dpc_init(dev);		/* Downstream Port Containment */
2499 	pci_rcec_init(dev);		/* Root Complex Event Collector */
2500 	pci_doe_init(dev);		/* Data Object Exchange */
2501 
2502 	pcie_report_downtraining(dev);
2503 	pci_init_reset_methods(dev);
2504 }
2505 
2506 /*
2507  * This is the equivalent of pci_host_bridge_msi_domain() that acts on
2508  * devices. Firmware interfaces that can select the MSI domain on a
2509  * per-device basis should be called from here.
2510  */
2511 static struct irq_domain *pci_dev_msi_domain(struct pci_dev *dev)
2512 {
2513 	struct irq_domain *d;
2514 
2515 	/*
2516 	 * If a domain has been set through the pcibios_device_add()
2517 	 * callback, then this is the one (platform code knows best).
2518 	 */
2519 	d = dev_get_msi_domain(&dev->dev);
2520 	if (d)
2521 		return d;
2522 
2523 	/*
2524 	 * Let's see if we have a firmware interface able to provide
2525 	 * the domain.
2526 	 */
2527 	d = pci_msi_get_device_domain(dev);
2528 	if (d)
2529 		return d;
2530 
2531 	return NULL;
2532 }
2533 
2534 static void pci_set_msi_domain(struct pci_dev *dev)
2535 {
2536 	struct irq_domain *d;
2537 
2538 	/*
2539 	 * If the platform or firmware interfaces cannot supply a
2540 	 * device-specific MSI domain, then inherit the default domain
2541 	 * from the host bridge itself.
2542 	 */
2543 	d = pci_dev_msi_domain(dev);
2544 	if (!d)
2545 		d = dev_get_msi_domain(&dev->bus->dev);
2546 
2547 	dev_set_msi_domain(&dev->dev, d);
2548 }
2549 
2550 void pci_device_add(struct pci_dev *dev, struct pci_bus *bus)
2551 {
2552 	int ret;
2553 
2554 	pci_configure_device(dev);
2555 
2556 	device_initialize(&dev->dev);
2557 	dev->dev.release = pci_release_dev;
2558 
2559 	set_dev_node(&dev->dev, pcibus_to_node(bus));
2560 	dev->dev.dma_mask = &dev->dma_mask;
2561 	dev->dev.dma_parms = &dev->dma_parms;
2562 	dev->dev.coherent_dma_mask = 0xffffffffull;
2563 
2564 	dma_set_max_seg_size(&dev->dev, 65536);
2565 	dma_set_seg_boundary(&dev->dev, 0xffffffff);
2566 
2567 	pcie_failed_link_retrain(dev);
2568 
2569 	/* Fix up broken headers */
2570 	pci_fixup_device(pci_fixup_header, dev);
2571 
2572 	pci_reassigndev_resource_alignment(dev);
2573 
2574 	dev->state_saved = false;
2575 
2576 	pci_init_capabilities(dev);
2577 
2578 	/*
2579 	 * Add the device to our list of discovered devices
2580 	 * and the bus list for fixup functions, etc.
2581 	 */
2582 	down_write(&pci_bus_sem);
2583 	list_add_tail(&dev->bus_list, &bus->devices);
2584 	up_write(&pci_bus_sem);
2585 
2586 	ret = pcibios_device_add(dev);
2587 	WARN_ON(ret < 0);
2588 
2589 	/* Set up MSI IRQ domain */
2590 	pci_set_msi_domain(dev);
2591 
2592 	/* Notifier could use PCI capabilities */
2593 	dev->match_driver = false;
2594 	ret = device_add(&dev->dev);
2595 	WARN_ON(ret < 0);
2596 }
2597 
2598 struct pci_dev *pci_scan_single_device(struct pci_bus *bus, int devfn)
2599 {
2600 	struct pci_dev *dev;
2601 
2602 	dev = pci_get_slot(bus, devfn);
2603 	if (dev) {
2604 		pci_dev_put(dev);
2605 		return dev;
2606 	}
2607 
2608 	dev = pci_scan_device(bus, devfn);
2609 	if (!dev)
2610 		return NULL;
2611 
2612 	pci_device_add(dev, bus);
2613 
2614 	return dev;
2615 }
2616 EXPORT_SYMBOL(pci_scan_single_device);
2617 
2618 static int next_ari_fn(struct pci_bus *bus, struct pci_dev *dev, int fn)
2619 {
2620 	int pos;
2621 	u16 cap = 0;
2622 	unsigned int next_fn;
2623 
2624 	if (!dev)
2625 		return -ENODEV;
2626 
2627 	pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ARI);
2628 	if (!pos)
2629 		return -ENODEV;
2630 
2631 	pci_read_config_word(dev, pos + PCI_ARI_CAP, &cap);
2632 	next_fn = PCI_ARI_CAP_NFN(cap);
2633 	if (next_fn <= fn)
2634 		return -ENODEV;	/* protect against malformed list */
2635 
2636 	return next_fn;
2637 }
2638 
2639 static int next_fn(struct pci_bus *bus, struct pci_dev *dev, int fn)
2640 {
2641 	if (pci_ari_enabled(bus))
2642 		return next_ari_fn(bus, dev, fn);
2643 
2644 	if (fn >= 7)
2645 		return -ENODEV;
2646 	/* only multifunction devices may have more functions */
2647 	if (dev && !dev->multifunction)
2648 		return -ENODEV;
2649 
2650 	return fn + 1;
2651 }
2652 
2653 static int only_one_child(struct pci_bus *bus)
2654 {
2655 	struct pci_dev *bridge = bus->self;
2656 
2657 	/*
2658 	 * Systems with unusual topologies set PCI_SCAN_ALL_PCIE_DEVS so
2659 	 * we scan for all possible devices, not just Device 0.
2660 	 */
2661 	if (pci_has_flag(PCI_SCAN_ALL_PCIE_DEVS))
2662 		return 0;
2663 
2664 	/*
2665 	 * A PCIe Downstream Port normally leads to a Link with only Device
2666 	 * 0 on it (PCIe spec r3.1, sec 7.3.1).  As an optimization, scan
2667 	 * only for Device 0 in that situation.
2668 	 */
2669 	if (bridge && pci_is_pcie(bridge) && pcie_downstream_port(bridge))
2670 		return 1;
2671 
2672 	return 0;
2673 }
2674 
2675 /**
2676  * pci_scan_slot - Scan a PCI slot on a bus for devices
2677  * @bus: PCI bus to scan
2678  * @devfn: slot number to scan (must have zero function)
2679  *
2680  * Scan a PCI slot on the specified PCI bus for devices, adding
2681  * discovered devices to the @bus->devices list.  New devices
2682  * will not have is_added set.
2683  *
2684  * Returns the number of new devices found.
2685  */
2686 int pci_scan_slot(struct pci_bus *bus, int devfn)
2687 {
2688 	struct pci_dev *dev;
2689 	int fn = 0, nr = 0;
2690 
2691 	if (only_one_child(bus) && (devfn > 0))
2692 		return 0; /* Already scanned the entire slot */
2693 
2694 	do {
2695 		dev = pci_scan_single_device(bus, devfn + fn);
2696 		if (dev) {
2697 			if (!pci_dev_is_added(dev))
2698 				nr++;
2699 			if (fn > 0)
2700 				dev->multifunction = 1;
2701 		} else if (fn == 0) {
2702 			/*
2703 			 * Function 0 is required unless we are running on
2704 			 * a hypervisor that passes through individual PCI
2705 			 * functions.
2706 			 */
2707 			if (!hypervisor_isolated_pci_functions())
2708 				break;
2709 		}
2710 		fn = next_fn(bus, dev, fn);
2711 	} while (fn >= 0);
2712 
2713 	/* Only one slot has PCIe device */
2714 	if (bus->self && nr)
2715 		pcie_aspm_init_link_state(bus->self);
2716 
2717 	return nr;
2718 }
2719 EXPORT_SYMBOL(pci_scan_slot);
2720 
2721 static int pcie_find_smpss(struct pci_dev *dev, void *data)
2722 {
2723 	u8 *smpss = data;
2724 
2725 	if (!pci_is_pcie(dev))
2726 		return 0;
2727 
2728 	/*
2729 	 * We don't have a way to change MPS settings on devices that have
2730 	 * drivers attached.  A hot-added device might support only the minimum
2731 	 * MPS setting (MPS=128).  Therefore, if the fabric contains a bridge
2732 	 * where devices may be hot-added, we limit the fabric MPS to 128 so
2733 	 * hot-added devices will work correctly.
2734 	 *
2735 	 * However, if we hot-add a device to a slot directly below a Root
2736 	 * Port, it's impossible for there to be other existing devices below
2737 	 * the port.  We don't limit the MPS in this case because we can
2738 	 * reconfigure MPS on both the Root Port and the hot-added device,
2739 	 * and there are no other devices involved.
2740 	 *
2741 	 * Note that this PCIE_BUS_SAFE path assumes no peer-to-peer DMA.
2742 	 */
2743 	if (dev->is_hotplug_bridge &&
2744 	    pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT)
2745 		*smpss = 0;
2746 
2747 	if (*smpss > dev->pcie_mpss)
2748 		*smpss = dev->pcie_mpss;
2749 
2750 	return 0;
2751 }
2752 
2753 static void pcie_write_mps(struct pci_dev *dev, int mps)
2754 {
2755 	int rc;
2756 
2757 	if (pcie_bus_config == PCIE_BUS_PERFORMANCE) {
2758 		mps = 128 << dev->pcie_mpss;
2759 
2760 		if (pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT &&
2761 		    dev->bus->self)
2762 
2763 			/*
2764 			 * For "Performance", the assumption is made that
2765 			 * downstream communication will never be larger than
2766 			 * the MRRS.  So, the MPS only needs to be configured
2767 			 * for the upstream communication.  This being the case,
2768 			 * walk from the top down and set the MPS of the child
2769 			 * to that of the parent bus.
2770 			 *
2771 			 * Configure the device MPS with the smaller of the
2772 			 * device MPSS or the bridge MPS (which is assumed to be
2773 			 * properly configured at this point to the largest
2774 			 * allowable MPS based on its parent bus).
2775 			 */
2776 			mps = min(mps, pcie_get_mps(dev->bus->self));
2777 	}
2778 
2779 	rc = pcie_set_mps(dev, mps);
2780 	if (rc)
2781 		pci_err(dev, "Failed attempting to set the MPS\n");
2782 }
2783 
2784 static void pcie_write_mrrs(struct pci_dev *dev)
2785 {
2786 	int rc, mrrs;
2787 
2788 	/*
2789 	 * In the "safe" case, do not configure the MRRS.  There appear to be
2790 	 * issues with setting MRRS to 0 on a number of devices.
2791 	 */
2792 	if (pcie_bus_config != PCIE_BUS_PERFORMANCE)
2793 		return;
2794 
2795 	/*
2796 	 * For max performance, the MRRS must be set to the largest supported
2797 	 * value.  However, it cannot be configured larger than the MPS the
2798 	 * device or the bus can support.  This should already be properly
2799 	 * configured by a prior call to pcie_write_mps().
2800 	 */
2801 	mrrs = pcie_get_mps(dev);
2802 
2803 	/*
2804 	 * MRRS is a R/W register.  Invalid values can be written, but a
2805 	 * subsequent read will verify if the value is acceptable or not.
2806 	 * If the MRRS value provided is not acceptable (e.g., too large),
2807 	 * shrink the value until it is acceptable to the HW.
2808 	 */
2809 	while (mrrs != pcie_get_readrq(dev) && mrrs >= 128) {
2810 		rc = pcie_set_readrq(dev, mrrs);
2811 		if (!rc)
2812 			break;
2813 
2814 		pci_warn(dev, "Failed attempting to set the MRRS\n");
2815 		mrrs /= 2;
2816 	}
2817 
2818 	if (mrrs < 128)
2819 		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");
2820 }
2821 
2822 static int pcie_bus_configure_set(struct pci_dev *dev, void *data)
2823 {
2824 	int mps, orig_mps;
2825 
2826 	if (!pci_is_pcie(dev))
2827 		return 0;
2828 
2829 	if (pcie_bus_config == PCIE_BUS_TUNE_OFF ||
2830 	    pcie_bus_config == PCIE_BUS_DEFAULT)
2831 		return 0;
2832 
2833 	mps = 128 << *(u8 *)data;
2834 	orig_mps = pcie_get_mps(dev);
2835 
2836 	pcie_write_mps(dev, mps);
2837 	pcie_write_mrrs(dev);
2838 
2839 	pci_info(dev, "Max Payload Size set to %4d/%4d (was %4d), Max Read Rq %4d\n",
2840 		 pcie_get_mps(dev), 128 << dev->pcie_mpss,
2841 		 orig_mps, pcie_get_readrq(dev));
2842 
2843 	return 0;
2844 }
2845 
2846 /*
2847  * pcie_bus_configure_settings() requires that pci_walk_bus work in a top-down,
2848  * parents then children fashion.  If this changes, then this code will not
2849  * work as designed.
2850  */
2851 void pcie_bus_configure_settings(struct pci_bus *bus)
2852 {
2853 	u8 smpss = 0;
2854 
2855 	if (!bus->self)
2856 		return;
2857 
2858 	if (!pci_is_pcie(bus->self))
2859 		return;
2860 
2861 	/*
2862 	 * FIXME - Peer to peer DMA is possible, though the endpoint would need
2863 	 * to be aware of the MPS of the destination.  To work around this,
2864 	 * simply force the MPS of the entire system to the smallest possible.
2865 	 */
2866 	if (pcie_bus_config == PCIE_BUS_PEER2PEER)
2867 		smpss = 0;
2868 
2869 	if (pcie_bus_config == PCIE_BUS_SAFE) {
2870 		smpss = bus->self->pcie_mpss;
2871 
2872 		pcie_find_smpss(bus->self, &smpss);
2873 		pci_walk_bus(bus, pcie_find_smpss, &smpss);
2874 	}
2875 
2876 	pcie_bus_configure_set(bus->self, &smpss);
2877 	pci_walk_bus(bus, pcie_bus_configure_set, &smpss);
2878 }
2879 EXPORT_SYMBOL_GPL(pcie_bus_configure_settings);
2880 
2881 /*
2882  * Called after each bus is probed, but before its children are examined.  This
2883  * is marked as __weak because multiple architectures define it.
2884  */
2885 void __weak pcibios_fixup_bus(struct pci_bus *bus)
2886 {
2887        /* nothing to do, expected to be removed in the future */
2888 }
2889 
2890 /**
2891  * pci_scan_child_bus_extend() - Scan devices below a bus
2892  * @bus: Bus to scan for devices
2893  * @available_buses: Total number of buses available (%0 does not try to
2894  *		     extend beyond the minimal)
2895  *
2896  * Scans devices below @bus including subordinate buses. Returns new
2897  * subordinate number including all the found devices. Passing
2898  * @available_buses causes the remaining bus space to be distributed
2899  * equally between hotplug-capable bridges to allow future extension of the
2900  * hierarchy.
2901  */
2902 static unsigned int pci_scan_child_bus_extend(struct pci_bus *bus,
2903 					      unsigned int available_buses)
2904 {
2905 	unsigned int used_buses, normal_bridges = 0, hotplug_bridges = 0;
2906 	unsigned int start = bus->busn_res.start;
2907 	unsigned int devfn, cmax, max = start;
2908 	struct pci_dev *dev;
2909 
2910 	dev_dbg(&bus->dev, "scanning bus\n");
2911 
2912 	/* Go find them, Rover! */
2913 	for (devfn = 0; devfn < 256; devfn += 8)
2914 		pci_scan_slot(bus, devfn);
2915 
2916 	/* Reserve buses for SR-IOV capability */
2917 	used_buses = pci_iov_bus_range(bus);
2918 	max += used_buses;
2919 
2920 	/*
2921 	 * After performing arch-dependent fixup of the bus, look behind
2922 	 * all PCI-to-PCI bridges on this bus.
2923 	 */
2924 	if (!bus->is_added) {
2925 		dev_dbg(&bus->dev, "fixups for bus\n");
2926 		pcibios_fixup_bus(bus);
2927 		bus->is_added = 1;
2928 	}
2929 
2930 	/*
2931 	 * Calculate how many hotplug bridges and normal bridges there
2932 	 * are on this bus. We will distribute the additional available
2933 	 * buses between hotplug bridges.
2934 	 */
2935 	for_each_pci_bridge(dev, bus) {
2936 		if (dev->is_hotplug_bridge)
2937 			hotplug_bridges++;
2938 		else
2939 			normal_bridges++;
2940 	}
2941 
2942 	/*
2943 	 * Scan bridges that are already configured. We don't touch them
2944 	 * unless they are misconfigured (which will be done in the second
2945 	 * scan below).
2946 	 */
2947 	for_each_pci_bridge(dev, bus) {
2948 		cmax = max;
2949 		max = pci_scan_bridge_extend(bus, dev, max, 0, 0);
2950 
2951 		/*
2952 		 * Reserve one bus for each bridge now to avoid extending
2953 		 * hotplug bridges too much during the second scan below.
2954 		 */
2955 		used_buses++;
2956 		if (max - cmax > 1)
2957 			used_buses += max - cmax - 1;
2958 	}
2959 
2960 	/* Scan bridges that need to be reconfigured */
2961 	for_each_pci_bridge(dev, bus) {
2962 		unsigned int buses = 0;
2963 
2964 		if (!hotplug_bridges && normal_bridges == 1) {
2965 			/*
2966 			 * There is only one bridge on the bus (upstream
2967 			 * port) so it gets all available buses which it
2968 			 * can then distribute to the possible hotplug
2969 			 * bridges below.
2970 			 */
2971 			buses = available_buses;
2972 		} else if (dev->is_hotplug_bridge) {
2973 			/*
2974 			 * Distribute the extra buses between hotplug
2975 			 * bridges if any.
2976 			 */
2977 			buses = available_buses / hotplug_bridges;
2978 			buses = min(buses, available_buses - used_buses + 1);
2979 		}
2980 
2981 		cmax = max;
2982 		max = pci_scan_bridge_extend(bus, dev, cmax, buses, 1);
2983 		/* One bus is already accounted so don't add it again */
2984 		if (max - cmax > 1)
2985 			used_buses += max - cmax - 1;
2986 	}
2987 
2988 	/*
2989 	 * Make sure a hotplug bridge has at least the minimum requested
2990 	 * number of buses but allow it to grow up to the maximum available
2991 	 * bus number if there is room.
2992 	 */
2993 	if (bus->self && bus->self->is_hotplug_bridge) {
2994 		used_buses = max_t(unsigned int, available_buses,
2995 				   pci_hotplug_bus_size - 1);
2996 		if (max - start < used_buses) {
2997 			max = start + used_buses;
2998 
2999 			/* Do not allocate more buses than we have room left */
3000 			if (max > bus->busn_res.end)
3001 				max = bus->busn_res.end;
3002 
3003 			dev_dbg(&bus->dev, "%pR extended by %#02x\n",
3004 				&bus->busn_res, max - start);
3005 		}
3006 	}
3007 
3008 	/*
3009 	 * We've scanned the bus and so we know all about what's on
3010 	 * the other side of any bridges that may be on this bus plus
3011 	 * any devices.
3012 	 *
3013 	 * Return how far we've got finding sub-buses.
3014 	 */
3015 	dev_dbg(&bus->dev, "bus scan returning with max=%02x\n", max);
3016 	return max;
3017 }
3018 
3019 /**
3020  * pci_scan_child_bus() - Scan devices below a bus
3021  * @bus: Bus to scan for devices
3022  *
3023  * Scans devices below @bus including subordinate buses. Returns new
3024  * subordinate number including all the found devices.
3025  */
3026 unsigned int pci_scan_child_bus(struct pci_bus *bus)
3027 {
3028 	return pci_scan_child_bus_extend(bus, 0);
3029 }
3030 EXPORT_SYMBOL_GPL(pci_scan_child_bus);
3031 
3032 /**
3033  * pcibios_root_bridge_prepare - Platform-specific host bridge setup
3034  * @bridge: Host bridge to set up
3035  *
3036  * Default empty implementation.  Replace with an architecture-specific setup
3037  * routine, if necessary.
3038  */
3039 int __weak pcibios_root_bridge_prepare(struct pci_host_bridge *bridge)
3040 {
3041 	return 0;
3042 }
3043 
3044 void __weak pcibios_add_bus(struct pci_bus *bus)
3045 {
3046 }
3047 
3048 void __weak pcibios_remove_bus(struct pci_bus *bus)
3049 {
3050 }
3051 
3052 struct pci_bus *pci_create_root_bus(struct device *parent, int bus,
3053 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
3054 {
3055 	int error;
3056 	struct pci_host_bridge *bridge;
3057 
3058 	bridge = pci_alloc_host_bridge(0);
3059 	if (!bridge)
3060 		return NULL;
3061 
3062 	bridge->dev.parent = parent;
3063 
3064 	list_splice_init(resources, &bridge->windows);
3065 	bridge->sysdata = sysdata;
3066 	bridge->busnr = bus;
3067 	bridge->ops = ops;
3068 
3069 	error = pci_register_host_bridge(bridge);
3070 	if (error < 0)
3071 		goto err_out;
3072 
3073 	return bridge->bus;
3074 
3075 err_out:
3076 	put_device(&bridge->dev);
3077 	return NULL;
3078 }
3079 EXPORT_SYMBOL_GPL(pci_create_root_bus);
3080 
3081 int pci_host_probe(struct pci_host_bridge *bridge)
3082 {
3083 	struct pci_bus *bus, *child;
3084 	int ret;
3085 
3086 	pci_lock_rescan_remove();
3087 	ret = pci_scan_root_bus_bridge(bridge);
3088 	pci_unlock_rescan_remove();
3089 	if (ret < 0) {
3090 		dev_err(bridge->dev.parent, "Scanning root bridge failed");
3091 		return ret;
3092 	}
3093 
3094 	bus = bridge->bus;
3095 
3096 	/* If we must preserve the resource configuration, claim now */
3097 	if (bridge->preserve_config)
3098 		pci_bus_claim_resources(bus);
3099 
3100 	/*
3101 	 * Assign whatever was left unassigned. If we didn't claim above,
3102 	 * this will reassign everything.
3103 	 */
3104 	pci_assign_unassigned_root_bus_resources(bus);
3105 
3106 	list_for_each_entry(child, &bus->children, node)
3107 		pcie_bus_configure_settings(child);
3108 
3109 	pci_bus_add_devices(bus);
3110 	return 0;
3111 }
3112 EXPORT_SYMBOL_GPL(pci_host_probe);
3113 
3114 int pci_bus_insert_busn_res(struct pci_bus *b, int bus, int bus_max)
3115 {
3116 	struct resource *res = &b->busn_res;
3117 	struct resource *parent_res, *conflict;
3118 
3119 	res->start = bus;
3120 	res->end = bus_max;
3121 	res->flags = IORESOURCE_BUS;
3122 
3123 	if (!pci_is_root_bus(b))
3124 		parent_res = &b->parent->busn_res;
3125 	else {
3126 		parent_res = get_pci_domain_busn_res(pci_domain_nr(b));
3127 		res->flags |= IORESOURCE_PCI_FIXED;
3128 	}
3129 
3130 	conflict = request_resource_conflict(parent_res, res);
3131 
3132 	if (conflict)
3133 		dev_info(&b->dev,
3134 			   "busn_res: can not insert %pR under %s%pR (conflicts with %s %pR)\n",
3135 			    res, pci_is_root_bus(b) ? "domain " : "",
3136 			    parent_res, conflict->name, conflict);
3137 
3138 	return conflict == NULL;
3139 }
3140 
3141 int pci_bus_update_busn_res_end(struct pci_bus *b, int bus_max)
3142 {
3143 	struct resource *res = &b->busn_res;
3144 	struct resource old_res = *res;
3145 	resource_size_t size;
3146 	int ret;
3147 
3148 	if (res->start > bus_max)
3149 		return -EINVAL;
3150 
3151 	size = bus_max - res->start + 1;
3152 	ret = adjust_resource(res, res->start, size);
3153 	dev_info(&b->dev, "busn_res: %pR end %s updated to %02x\n",
3154 			&old_res, ret ? "can not be" : "is", bus_max);
3155 
3156 	if (!ret && !res->parent)
3157 		pci_bus_insert_busn_res(b, res->start, res->end);
3158 
3159 	return ret;
3160 }
3161 
3162 void pci_bus_release_busn_res(struct pci_bus *b)
3163 {
3164 	struct resource *res = &b->busn_res;
3165 	int ret;
3166 
3167 	if (!res->flags || !res->parent)
3168 		return;
3169 
3170 	ret = release_resource(res);
3171 	dev_info(&b->dev, "busn_res: %pR %s released\n",
3172 			res, ret ? "can not be" : "is");
3173 }
3174 
3175 int pci_scan_root_bus_bridge(struct pci_host_bridge *bridge)
3176 {
3177 	struct resource_entry *window;
3178 	bool found = false;
3179 	struct pci_bus *b;
3180 	int max, bus, ret;
3181 
3182 	if (!bridge)
3183 		return -EINVAL;
3184 
3185 	resource_list_for_each_entry(window, &bridge->windows)
3186 		if (window->res->flags & IORESOURCE_BUS) {
3187 			bridge->busnr = window->res->start;
3188 			found = true;
3189 			break;
3190 		}
3191 
3192 	ret = pci_register_host_bridge(bridge);
3193 	if (ret < 0)
3194 		return ret;
3195 
3196 	b = bridge->bus;
3197 	bus = bridge->busnr;
3198 
3199 	if (!found) {
3200 		dev_info(&b->dev,
3201 		 "No busn resource found for root bus, will use [bus %02x-ff]\n",
3202 			bus);
3203 		pci_bus_insert_busn_res(b, bus, 255);
3204 	}
3205 
3206 	max = pci_scan_child_bus(b);
3207 
3208 	if (!found)
3209 		pci_bus_update_busn_res_end(b, max);
3210 
3211 	return 0;
3212 }
3213 EXPORT_SYMBOL(pci_scan_root_bus_bridge);
3214 
3215 struct pci_bus *pci_scan_root_bus(struct device *parent, int bus,
3216 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
3217 {
3218 	struct resource_entry *window;
3219 	bool found = false;
3220 	struct pci_bus *b;
3221 	int max;
3222 
3223 	resource_list_for_each_entry(window, resources)
3224 		if (window->res->flags & IORESOURCE_BUS) {
3225 			found = true;
3226 			break;
3227 		}
3228 
3229 	b = pci_create_root_bus(parent, bus, ops, sysdata, resources);
3230 	if (!b)
3231 		return NULL;
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 b;
3246 }
3247 EXPORT_SYMBOL(pci_scan_root_bus);
3248 
3249 struct pci_bus *pci_scan_bus(int bus, struct pci_ops *ops,
3250 					void *sysdata)
3251 {
3252 	LIST_HEAD(resources);
3253 	struct pci_bus *b;
3254 
3255 	pci_add_resource(&resources, &ioport_resource);
3256 	pci_add_resource(&resources, &iomem_resource);
3257 	pci_add_resource(&resources, &busn_resource);
3258 	b = pci_create_root_bus(NULL, bus, ops, sysdata, &resources);
3259 	if (b) {
3260 		pci_scan_child_bus(b);
3261 	} else {
3262 		pci_free_resource_list(&resources);
3263 	}
3264 	return b;
3265 }
3266 EXPORT_SYMBOL(pci_scan_bus);
3267 
3268 /**
3269  * pci_rescan_bus_bridge_resize - Scan a PCI bus for devices
3270  * @bridge: PCI bridge for the bus to scan
3271  *
3272  * Scan a PCI bus and child buses for new devices, add them,
3273  * and enable them, resizing bridge mmio/io resource if necessary
3274  * and possible.  The caller must ensure the child devices are already
3275  * removed for resizing to occur.
3276  *
3277  * Returns the max number of subordinate bus discovered.
3278  */
3279 unsigned int pci_rescan_bus_bridge_resize(struct pci_dev *bridge)
3280 {
3281 	unsigned int max;
3282 	struct pci_bus *bus = bridge->subordinate;
3283 
3284 	max = pci_scan_child_bus(bus);
3285 
3286 	pci_assign_unassigned_bridge_resources(bridge);
3287 
3288 	pci_bus_add_devices(bus);
3289 
3290 	return max;
3291 }
3292 
3293 /**
3294  * pci_rescan_bus - Scan a PCI bus for devices
3295  * @bus: PCI bus to scan
3296  *
3297  * Scan a PCI bus and child buses for new devices, add them,
3298  * and enable them.
3299  *
3300  * Returns the max number of subordinate bus discovered.
3301  */
3302 unsigned int pci_rescan_bus(struct pci_bus *bus)
3303 {
3304 	unsigned int max;
3305 
3306 	max = pci_scan_child_bus(bus);
3307 	pci_assign_unassigned_bus_resources(bus);
3308 	pci_bus_add_devices(bus);
3309 
3310 	return max;
3311 }
3312 EXPORT_SYMBOL_GPL(pci_rescan_bus);
3313 
3314 /*
3315  * pci_rescan_bus(), pci_rescan_bus_bridge_resize() and PCI device removal
3316  * routines should always be executed under this mutex.
3317  */
3318 static DEFINE_MUTEX(pci_rescan_remove_lock);
3319 
3320 void pci_lock_rescan_remove(void)
3321 {
3322 	mutex_lock(&pci_rescan_remove_lock);
3323 }
3324 EXPORT_SYMBOL_GPL(pci_lock_rescan_remove);
3325 
3326 void pci_unlock_rescan_remove(void)
3327 {
3328 	mutex_unlock(&pci_rescan_remove_lock);
3329 }
3330 EXPORT_SYMBOL_GPL(pci_unlock_rescan_remove);
3331 
3332 static int __init pci_sort_bf_cmp(const struct device *d_a,
3333 				  const struct device *d_b)
3334 {
3335 	const struct pci_dev *a = to_pci_dev(d_a);
3336 	const struct pci_dev *b = to_pci_dev(d_b);
3337 
3338 	if      (pci_domain_nr(a->bus) < pci_domain_nr(b->bus)) return -1;
3339 	else if (pci_domain_nr(a->bus) > pci_domain_nr(b->bus)) return  1;
3340 
3341 	if      (a->bus->number < b->bus->number) return -1;
3342 	else if (a->bus->number > b->bus->number) return  1;
3343 
3344 	if      (a->devfn < b->devfn) return -1;
3345 	else if (a->devfn > b->devfn) return  1;
3346 
3347 	return 0;
3348 }
3349 
3350 void __init pci_sort_breadthfirst(void)
3351 {
3352 	bus_sort_breadthfirst(&pci_bus_type, &pci_sort_bf_cmp);
3353 }
3354 
3355 int pci_hp_add_bridge(struct pci_dev *dev)
3356 {
3357 	struct pci_bus *parent = dev->bus;
3358 	int busnr, start = parent->busn_res.start;
3359 	unsigned int available_buses = 0;
3360 	int end = parent->busn_res.end;
3361 
3362 	for (busnr = start; busnr <= end; busnr++) {
3363 		if (!pci_find_bus(pci_domain_nr(parent), busnr))
3364 			break;
3365 	}
3366 	if (busnr-- > end) {
3367 		pci_err(dev, "No bus number available for hot-added bridge\n");
3368 		return -1;
3369 	}
3370 
3371 	/* Scan bridges that are already configured */
3372 	busnr = pci_scan_bridge(parent, dev, busnr, 0);
3373 
3374 	/*
3375 	 * Distribute the available bus numbers between hotplug-capable
3376 	 * bridges to make extending the chain later possible.
3377 	 */
3378 	available_buses = end - busnr;
3379 
3380 	/* Scan bridges that need to be reconfigured */
3381 	pci_scan_bridge_extend(parent, dev, busnr, available_buses, 1);
3382 
3383 	if (!dev->subordinate)
3384 		return -1;
3385 
3386 	return 0;
3387 }
3388 EXPORT_SYMBOL_GPL(pci_hp_add_bridge);
3389