xref: /linux/drivers/pci/probe.c (revision fc4fa6e112c0f999fab022a4eb7f6614bb47c7ab)
1 /*
2  * probe.c - PCI detection and setup code
3  */
4 
5 #include <linux/kernel.h>
6 #include <linux/delay.h>
7 #include <linux/init.h>
8 #include <linux/pci.h>
9 #include <linux/of_pci.h>
10 #include <linux/pci_hotplug.h>
11 #include <linux/slab.h>
12 #include <linux/module.h>
13 #include <linux/cpumask.h>
14 #include <linux/pci-aspm.h>
15 #include <linux/aer.h>
16 #include <asm-generic/pci-bridge.h>
17 #include "pci.h"
18 
19 #define CARDBUS_LATENCY_TIMER	176	/* secondary latency timer */
20 #define CARDBUS_RESERVE_BUSNR	3
21 
22 static struct resource busn_resource = {
23 	.name	= "PCI busn",
24 	.start	= 0,
25 	.end	= 255,
26 	.flags	= IORESOURCE_BUS,
27 };
28 
29 /* Ugh.  Need to stop exporting this to modules. */
30 LIST_HEAD(pci_root_buses);
31 EXPORT_SYMBOL(pci_root_buses);
32 
33 static LIST_HEAD(pci_domain_busn_res_list);
34 
35 struct pci_domain_busn_res {
36 	struct list_head list;
37 	struct resource res;
38 	int domain_nr;
39 };
40 
41 static struct resource *get_pci_domain_busn_res(int domain_nr)
42 {
43 	struct pci_domain_busn_res *r;
44 
45 	list_for_each_entry(r, &pci_domain_busn_res_list, list)
46 		if (r->domain_nr == domain_nr)
47 			return &r->res;
48 
49 	r = kzalloc(sizeof(*r), GFP_KERNEL);
50 	if (!r)
51 		return NULL;
52 
53 	r->domain_nr = domain_nr;
54 	r->res.start = 0;
55 	r->res.end = 0xff;
56 	r->res.flags = IORESOURCE_BUS | IORESOURCE_PCI_FIXED;
57 
58 	list_add_tail(&r->list, &pci_domain_busn_res_list);
59 
60 	return &r->res;
61 }
62 
63 static int find_anything(struct device *dev, void *data)
64 {
65 	return 1;
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_device(&pci_bus_type, NULL, NULL, find_anything);
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 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 	/* Get the lowest of them to find the decode size, and
117 	   from that the extent.  */
118 	size = (size & ~(size-1)) - 1;
119 
120 	/* base == maxbase can be valid only if the BAR has
121 	   already been programmed with all 1s.  */
122 	if (base == maxbase && ((base | size) & mask) != mask)
123 		return 0;
124 
125 	return size;
126 }
127 
128 static inline unsigned long decode_bar(struct pci_dev *dev, u32 bar)
129 {
130 	u32 mem_type;
131 	unsigned long flags;
132 
133 	if ((bar & PCI_BASE_ADDRESS_SPACE) == PCI_BASE_ADDRESS_SPACE_IO) {
134 		flags = bar & ~PCI_BASE_ADDRESS_IO_MASK;
135 		flags |= IORESOURCE_IO;
136 		return flags;
137 	}
138 
139 	flags = bar & ~PCI_BASE_ADDRESS_MEM_MASK;
140 	flags |= IORESOURCE_MEM;
141 	if (flags & PCI_BASE_ADDRESS_MEM_PREFETCH)
142 		flags |= IORESOURCE_PREFETCH;
143 
144 	mem_type = bar & PCI_BASE_ADDRESS_MEM_TYPE_MASK;
145 	switch (mem_type) {
146 	case PCI_BASE_ADDRESS_MEM_TYPE_32:
147 		break;
148 	case PCI_BASE_ADDRESS_MEM_TYPE_1M:
149 		/* 1M mem BAR treated as 32-bit BAR */
150 		break;
151 	case PCI_BASE_ADDRESS_MEM_TYPE_64:
152 		flags |= IORESOURCE_MEM_64;
153 		break;
154 	default:
155 		/* mem unknown type treated as 32-bit BAR */
156 		break;
157 	}
158 	return flags;
159 }
160 
161 #define PCI_COMMAND_DECODE_ENABLE	(PCI_COMMAND_MEMORY | PCI_COMMAND_IO)
162 
163 /**
164  * pci_read_base - read a PCI BAR
165  * @dev: the PCI device
166  * @type: type of the BAR
167  * @res: resource buffer to be filled in
168  * @pos: BAR position in the config space
169  *
170  * Returns 1 if the BAR is 64-bit, or 0 if 32-bit.
171  */
172 int __pci_read_base(struct pci_dev *dev, enum pci_bar_type type,
173 		    struct resource *res, unsigned int pos)
174 {
175 	u32 l, sz, mask;
176 	u64 l64, sz64, mask64;
177 	u16 orig_cmd;
178 	struct pci_bus_region region, inverted_region;
179 
180 	mask = type ? PCI_ROM_ADDRESS_MASK : ~0;
181 
182 	/* No printks while decoding is disabled! */
183 	if (!dev->mmio_always_on) {
184 		pci_read_config_word(dev, PCI_COMMAND, &orig_cmd);
185 		if (orig_cmd & PCI_COMMAND_DECODE_ENABLE) {
186 			pci_write_config_word(dev, PCI_COMMAND,
187 				orig_cmd & ~PCI_COMMAND_DECODE_ENABLE);
188 		}
189 	}
190 
191 	res->name = pci_name(dev);
192 
193 	pci_read_config_dword(dev, pos, &l);
194 	pci_write_config_dword(dev, pos, l | mask);
195 	pci_read_config_dword(dev, pos, &sz);
196 	pci_write_config_dword(dev, pos, l);
197 
198 	/*
199 	 * All bits set in sz means the device isn't working properly.
200 	 * If the BAR isn't implemented, all bits must be 0.  If it's a
201 	 * memory BAR or a ROM, bit 0 must be clear; if it's an io BAR, bit
202 	 * 1 must be clear.
203 	 */
204 	if (sz == 0xffffffff)
205 		sz = 0;
206 
207 	/*
208 	 * I don't know how l can have all bits set.  Copied from old code.
209 	 * Maybe it fixes a bug on some ancient platform.
210 	 */
211 	if (l == 0xffffffff)
212 		l = 0;
213 
214 	if (type == pci_bar_unknown) {
215 		res->flags = decode_bar(dev, l);
216 		res->flags |= IORESOURCE_SIZEALIGN;
217 		if (res->flags & IORESOURCE_IO) {
218 			l64 = l & PCI_BASE_ADDRESS_IO_MASK;
219 			sz64 = sz & PCI_BASE_ADDRESS_IO_MASK;
220 			mask64 = PCI_BASE_ADDRESS_IO_MASK & (u32)IO_SPACE_LIMIT;
221 		} else {
222 			l64 = l & PCI_BASE_ADDRESS_MEM_MASK;
223 			sz64 = sz & PCI_BASE_ADDRESS_MEM_MASK;
224 			mask64 = (u32)PCI_BASE_ADDRESS_MEM_MASK;
225 		}
226 	} else {
227 		res->flags |= (l & IORESOURCE_ROM_ENABLE);
228 		l64 = l & PCI_ROM_ADDRESS_MASK;
229 		sz64 = sz & PCI_ROM_ADDRESS_MASK;
230 		mask64 = (u32)PCI_ROM_ADDRESS_MASK;
231 	}
232 
233 	if (res->flags & IORESOURCE_MEM_64) {
234 		pci_read_config_dword(dev, pos + 4, &l);
235 		pci_write_config_dword(dev, pos + 4, ~0);
236 		pci_read_config_dword(dev, pos + 4, &sz);
237 		pci_write_config_dword(dev, pos + 4, l);
238 
239 		l64 |= ((u64)l << 32);
240 		sz64 |= ((u64)sz << 32);
241 		mask64 |= ((u64)~0 << 32);
242 	}
243 
244 	if (!dev->mmio_always_on && (orig_cmd & PCI_COMMAND_DECODE_ENABLE))
245 		pci_write_config_word(dev, PCI_COMMAND, orig_cmd);
246 
247 	if (!sz64)
248 		goto fail;
249 
250 	sz64 = pci_size(l64, sz64, mask64);
251 	if (!sz64) {
252 		dev_info(&dev->dev, FW_BUG "reg 0x%x: invalid BAR (can't size)\n",
253 			 pos);
254 		goto fail;
255 	}
256 
257 	if (res->flags & IORESOURCE_MEM_64) {
258 		if ((sizeof(pci_bus_addr_t) < 8 || sizeof(resource_size_t) < 8)
259 		    && sz64 > 0x100000000ULL) {
260 			res->flags |= IORESOURCE_UNSET | IORESOURCE_DISABLED;
261 			res->start = 0;
262 			res->end = 0;
263 			dev_err(&dev->dev, "reg 0x%x: can't handle BAR larger than 4GB (size %#010llx)\n",
264 				pos, (unsigned long long)sz64);
265 			goto out;
266 		}
267 
268 		if ((sizeof(pci_bus_addr_t) < 8) && l) {
269 			/* Above 32-bit boundary; try to reallocate */
270 			res->flags |= IORESOURCE_UNSET;
271 			res->start = 0;
272 			res->end = sz64;
273 			dev_info(&dev->dev, "reg 0x%x: can't handle BAR above 4GB (bus address %#010llx)\n",
274 				 pos, (unsigned long long)l64);
275 			goto out;
276 		}
277 	}
278 
279 	region.start = l64;
280 	region.end = l64 + sz64;
281 
282 	pcibios_bus_to_resource(dev->bus, res, &region);
283 	pcibios_resource_to_bus(dev->bus, &inverted_region, res);
284 
285 	/*
286 	 * If "A" is a BAR value (a bus address), "bus_to_resource(A)" is
287 	 * the corresponding resource address (the physical address used by
288 	 * the CPU.  Converting that resource address back to a bus address
289 	 * should yield the original BAR value:
290 	 *
291 	 *     resource_to_bus(bus_to_resource(A)) == A
292 	 *
293 	 * If it doesn't, CPU accesses to "bus_to_resource(A)" will not
294 	 * be claimed by the device.
295 	 */
296 	if (inverted_region.start != region.start) {
297 		res->flags |= IORESOURCE_UNSET;
298 		res->start = 0;
299 		res->end = region.end - region.start;
300 		dev_info(&dev->dev, "reg 0x%x: initial BAR value %#010llx invalid\n",
301 			 pos, (unsigned long long)region.start);
302 	}
303 
304 	goto out;
305 
306 
307 fail:
308 	res->flags = 0;
309 out:
310 	if (res->flags)
311 		dev_printk(KERN_DEBUG, &dev->dev, "reg 0x%x: %pR\n", pos, res);
312 
313 	return (res->flags & IORESOURCE_MEM_64) ? 1 : 0;
314 }
315 
316 static void pci_read_bases(struct pci_dev *dev, unsigned int howmany, int rom)
317 {
318 	unsigned int pos, reg;
319 
320 	for (pos = 0; pos < howmany; pos++) {
321 		struct resource *res = &dev->resource[pos];
322 		reg = PCI_BASE_ADDRESS_0 + (pos << 2);
323 		pos += __pci_read_base(dev, pci_bar_unknown, res, reg);
324 	}
325 
326 	if (rom) {
327 		struct resource *res = &dev->resource[PCI_ROM_RESOURCE];
328 		dev->rom_base_reg = rom;
329 		res->flags = IORESOURCE_MEM | IORESOURCE_PREFETCH |
330 				IORESOURCE_READONLY | IORESOURCE_SIZEALIGN;
331 		__pci_read_base(dev, pci_bar_mem32, res, rom);
332 	}
333 }
334 
335 static void pci_read_bridge_io(struct pci_bus *child)
336 {
337 	struct pci_dev *dev = child->self;
338 	u8 io_base_lo, io_limit_lo;
339 	unsigned long io_mask, io_granularity, base, limit;
340 	struct pci_bus_region region;
341 	struct resource *res;
342 
343 	io_mask = PCI_IO_RANGE_MASK;
344 	io_granularity = 0x1000;
345 	if (dev->io_window_1k) {
346 		/* Support 1K I/O space granularity */
347 		io_mask = PCI_IO_1K_RANGE_MASK;
348 		io_granularity = 0x400;
349 	}
350 
351 	res = child->resource[0];
352 	pci_read_config_byte(dev, PCI_IO_BASE, &io_base_lo);
353 	pci_read_config_byte(dev, PCI_IO_LIMIT, &io_limit_lo);
354 	base = (io_base_lo & io_mask) << 8;
355 	limit = (io_limit_lo & io_mask) << 8;
356 
357 	if ((io_base_lo & PCI_IO_RANGE_TYPE_MASK) == PCI_IO_RANGE_TYPE_32) {
358 		u16 io_base_hi, io_limit_hi;
359 
360 		pci_read_config_word(dev, PCI_IO_BASE_UPPER16, &io_base_hi);
361 		pci_read_config_word(dev, PCI_IO_LIMIT_UPPER16, &io_limit_hi);
362 		base |= ((unsigned long) io_base_hi << 16);
363 		limit |= ((unsigned long) io_limit_hi << 16);
364 	}
365 
366 	if (base <= limit) {
367 		res->flags = (io_base_lo & PCI_IO_RANGE_TYPE_MASK) | IORESOURCE_IO;
368 		region.start = base;
369 		region.end = limit + io_granularity - 1;
370 		pcibios_bus_to_resource(dev->bus, res, &region);
371 		dev_printk(KERN_DEBUG, &dev->dev, "  bridge window %pR\n", res);
372 	}
373 }
374 
375 static void pci_read_bridge_mmio(struct pci_bus *child)
376 {
377 	struct pci_dev *dev = child->self;
378 	u16 mem_base_lo, mem_limit_lo;
379 	unsigned long base, limit;
380 	struct pci_bus_region region;
381 	struct resource *res;
382 
383 	res = child->resource[1];
384 	pci_read_config_word(dev, PCI_MEMORY_BASE, &mem_base_lo);
385 	pci_read_config_word(dev, PCI_MEMORY_LIMIT, &mem_limit_lo);
386 	base = ((unsigned long) mem_base_lo & PCI_MEMORY_RANGE_MASK) << 16;
387 	limit = ((unsigned long) mem_limit_lo & PCI_MEMORY_RANGE_MASK) << 16;
388 	if (base <= limit) {
389 		res->flags = (mem_base_lo & PCI_MEMORY_RANGE_TYPE_MASK) | IORESOURCE_MEM;
390 		region.start = base;
391 		region.end = limit + 0xfffff;
392 		pcibios_bus_to_resource(dev->bus, res, &region);
393 		dev_printk(KERN_DEBUG, &dev->dev, "  bridge window %pR\n", res);
394 	}
395 }
396 
397 static void pci_read_bridge_mmio_pref(struct pci_bus *child)
398 {
399 	struct pci_dev *dev = child->self;
400 	u16 mem_base_lo, mem_limit_lo;
401 	u64 base64, limit64;
402 	pci_bus_addr_t base, limit;
403 	struct pci_bus_region region;
404 	struct resource *res;
405 
406 	res = child->resource[2];
407 	pci_read_config_word(dev, PCI_PREF_MEMORY_BASE, &mem_base_lo);
408 	pci_read_config_word(dev, PCI_PREF_MEMORY_LIMIT, &mem_limit_lo);
409 	base64 = (mem_base_lo & PCI_PREF_RANGE_MASK) << 16;
410 	limit64 = (mem_limit_lo & PCI_PREF_RANGE_MASK) << 16;
411 
412 	if ((mem_base_lo & PCI_PREF_RANGE_TYPE_MASK) == PCI_PREF_RANGE_TYPE_64) {
413 		u32 mem_base_hi, mem_limit_hi;
414 
415 		pci_read_config_dword(dev, PCI_PREF_BASE_UPPER32, &mem_base_hi);
416 		pci_read_config_dword(dev, PCI_PREF_LIMIT_UPPER32, &mem_limit_hi);
417 
418 		/*
419 		 * Some bridges set the base > limit by default, and some
420 		 * (broken) BIOSes do not initialize them.  If we find
421 		 * this, just assume they are not being used.
422 		 */
423 		if (mem_base_hi <= mem_limit_hi) {
424 			base64 |= (u64) mem_base_hi << 32;
425 			limit64 |= (u64) mem_limit_hi << 32;
426 		}
427 	}
428 
429 	base = (pci_bus_addr_t) base64;
430 	limit = (pci_bus_addr_t) limit64;
431 
432 	if (base != base64) {
433 		dev_err(&dev->dev, "can't handle bridge window above 4GB (bus address %#010llx)\n",
434 			(unsigned long long) base64);
435 		return;
436 	}
437 
438 	if (base <= limit) {
439 		res->flags = (mem_base_lo & PCI_PREF_RANGE_TYPE_MASK) |
440 					 IORESOURCE_MEM | IORESOURCE_PREFETCH;
441 		if (res->flags & PCI_PREF_RANGE_TYPE_64)
442 			res->flags |= IORESOURCE_MEM_64;
443 		region.start = base;
444 		region.end = limit + 0xfffff;
445 		pcibios_bus_to_resource(dev->bus, res, &region);
446 		dev_printk(KERN_DEBUG, &dev->dev, "  bridge window %pR\n", res);
447 	}
448 }
449 
450 void pci_read_bridge_bases(struct pci_bus *child)
451 {
452 	struct pci_dev *dev = child->self;
453 	struct resource *res;
454 	int i;
455 
456 	if (pci_is_root_bus(child))	/* It's a host bus, nothing to read */
457 		return;
458 
459 	dev_info(&dev->dev, "PCI bridge to %pR%s\n",
460 		 &child->busn_res,
461 		 dev->transparent ? " (subtractive decode)" : "");
462 
463 	pci_bus_remove_resources(child);
464 	for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++)
465 		child->resource[i] = &dev->resource[PCI_BRIDGE_RESOURCES+i];
466 
467 	pci_read_bridge_io(child);
468 	pci_read_bridge_mmio(child);
469 	pci_read_bridge_mmio_pref(child);
470 
471 	if (dev->transparent) {
472 		pci_bus_for_each_resource(child->parent, res, i) {
473 			if (res && res->flags) {
474 				pci_bus_add_resource(child, res,
475 						     PCI_SUBTRACTIVE_DECODE);
476 				dev_printk(KERN_DEBUG, &dev->dev,
477 					   "  bridge window %pR (subtractive decode)\n",
478 					   res);
479 			}
480 		}
481 	}
482 }
483 
484 static struct pci_bus *pci_alloc_bus(struct pci_bus *parent)
485 {
486 	struct pci_bus *b;
487 
488 	b = kzalloc(sizeof(*b), GFP_KERNEL);
489 	if (!b)
490 		return NULL;
491 
492 	INIT_LIST_HEAD(&b->node);
493 	INIT_LIST_HEAD(&b->children);
494 	INIT_LIST_HEAD(&b->devices);
495 	INIT_LIST_HEAD(&b->slots);
496 	INIT_LIST_HEAD(&b->resources);
497 	b->max_bus_speed = PCI_SPEED_UNKNOWN;
498 	b->cur_bus_speed = PCI_SPEED_UNKNOWN;
499 #ifdef CONFIG_PCI_DOMAINS_GENERIC
500 	if (parent)
501 		b->domain_nr = parent->domain_nr;
502 #endif
503 	return b;
504 }
505 
506 static void pci_release_host_bridge_dev(struct device *dev)
507 {
508 	struct pci_host_bridge *bridge = to_pci_host_bridge(dev);
509 
510 	if (bridge->release_fn)
511 		bridge->release_fn(bridge);
512 
513 	pci_free_resource_list(&bridge->windows);
514 
515 	kfree(bridge);
516 }
517 
518 static struct pci_host_bridge *pci_alloc_host_bridge(struct pci_bus *b)
519 {
520 	struct pci_host_bridge *bridge;
521 
522 	bridge = kzalloc(sizeof(*bridge), GFP_KERNEL);
523 	if (!bridge)
524 		return NULL;
525 
526 	INIT_LIST_HEAD(&bridge->windows);
527 	bridge->bus = b;
528 	return bridge;
529 }
530 
531 static const unsigned char pcix_bus_speed[] = {
532 	PCI_SPEED_UNKNOWN,		/* 0 */
533 	PCI_SPEED_66MHz_PCIX,		/* 1 */
534 	PCI_SPEED_100MHz_PCIX,		/* 2 */
535 	PCI_SPEED_133MHz_PCIX,		/* 3 */
536 	PCI_SPEED_UNKNOWN,		/* 4 */
537 	PCI_SPEED_66MHz_PCIX_ECC,	/* 5 */
538 	PCI_SPEED_100MHz_PCIX_ECC,	/* 6 */
539 	PCI_SPEED_133MHz_PCIX_ECC,	/* 7 */
540 	PCI_SPEED_UNKNOWN,		/* 8 */
541 	PCI_SPEED_66MHz_PCIX_266,	/* 9 */
542 	PCI_SPEED_100MHz_PCIX_266,	/* A */
543 	PCI_SPEED_133MHz_PCIX_266,	/* B */
544 	PCI_SPEED_UNKNOWN,		/* C */
545 	PCI_SPEED_66MHz_PCIX_533,	/* D */
546 	PCI_SPEED_100MHz_PCIX_533,	/* E */
547 	PCI_SPEED_133MHz_PCIX_533	/* F */
548 };
549 
550 const unsigned char pcie_link_speed[] = {
551 	PCI_SPEED_UNKNOWN,		/* 0 */
552 	PCIE_SPEED_2_5GT,		/* 1 */
553 	PCIE_SPEED_5_0GT,		/* 2 */
554 	PCIE_SPEED_8_0GT,		/* 3 */
555 	PCI_SPEED_UNKNOWN,		/* 4 */
556 	PCI_SPEED_UNKNOWN,		/* 5 */
557 	PCI_SPEED_UNKNOWN,		/* 6 */
558 	PCI_SPEED_UNKNOWN,		/* 7 */
559 	PCI_SPEED_UNKNOWN,		/* 8 */
560 	PCI_SPEED_UNKNOWN,		/* 9 */
561 	PCI_SPEED_UNKNOWN,		/* A */
562 	PCI_SPEED_UNKNOWN,		/* B */
563 	PCI_SPEED_UNKNOWN,		/* C */
564 	PCI_SPEED_UNKNOWN,		/* D */
565 	PCI_SPEED_UNKNOWN,		/* E */
566 	PCI_SPEED_UNKNOWN		/* F */
567 };
568 
569 void pcie_update_link_speed(struct pci_bus *bus, u16 linksta)
570 {
571 	bus->cur_bus_speed = pcie_link_speed[linksta & PCI_EXP_LNKSTA_CLS];
572 }
573 EXPORT_SYMBOL_GPL(pcie_update_link_speed);
574 
575 static unsigned char agp_speeds[] = {
576 	AGP_UNKNOWN,
577 	AGP_1X,
578 	AGP_2X,
579 	AGP_4X,
580 	AGP_8X
581 };
582 
583 static enum pci_bus_speed agp_speed(int agp3, int agpstat)
584 {
585 	int index = 0;
586 
587 	if (agpstat & 4)
588 		index = 3;
589 	else if (agpstat & 2)
590 		index = 2;
591 	else if (agpstat & 1)
592 		index = 1;
593 	else
594 		goto out;
595 
596 	if (agp3) {
597 		index += 2;
598 		if (index == 5)
599 			index = 0;
600 	}
601 
602  out:
603 	return agp_speeds[index];
604 }
605 
606 static void pci_set_bus_speed(struct pci_bus *bus)
607 {
608 	struct pci_dev *bridge = bus->self;
609 	int pos;
610 
611 	pos = pci_find_capability(bridge, PCI_CAP_ID_AGP);
612 	if (!pos)
613 		pos = pci_find_capability(bridge, PCI_CAP_ID_AGP3);
614 	if (pos) {
615 		u32 agpstat, agpcmd;
616 
617 		pci_read_config_dword(bridge, pos + PCI_AGP_STATUS, &agpstat);
618 		bus->max_bus_speed = agp_speed(agpstat & 8, agpstat & 7);
619 
620 		pci_read_config_dword(bridge, pos + PCI_AGP_COMMAND, &agpcmd);
621 		bus->cur_bus_speed = agp_speed(agpstat & 8, agpcmd & 7);
622 	}
623 
624 	pos = pci_find_capability(bridge, PCI_CAP_ID_PCIX);
625 	if (pos) {
626 		u16 status;
627 		enum pci_bus_speed max;
628 
629 		pci_read_config_word(bridge, pos + PCI_X_BRIDGE_SSTATUS,
630 				     &status);
631 
632 		if (status & PCI_X_SSTATUS_533MHZ) {
633 			max = PCI_SPEED_133MHz_PCIX_533;
634 		} else if (status & PCI_X_SSTATUS_266MHZ) {
635 			max = PCI_SPEED_133MHz_PCIX_266;
636 		} else if (status & PCI_X_SSTATUS_133MHZ) {
637 			if ((status & PCI_X_SSTATUS_VERS) == PCI_X_SSTATUS_V2)
638 				max = PCI_SPEED_133MHz_PCIX_ECC;
639 			else
640 				max = PCI_SPEED_133MHz_PCIX;
641 		} else {
642 			max = PCI_SPEED_66MHz_PCIX;
643 		}
644 
645 		bus->max_bus_speed = max;
646 		bus->cur_bus_speed = pcix_bus_speed[
647 			(status & PCI_X_SSTATUS_FREQ) >> 6];
648 
649 		return;
650 	}
651 
652 	if (pci_is_pcie(bridge)) {
653 		u32 linkcap;
654 		u16 linksta;
655 
656 		pcie_capability_read_dword(bridge, PCI_EXP_LNKCAP, &linkcap);
657 		bus->max_bus_speed = pcie_link_speed[linkcap & PCI_EXP_LNKCAP_SLS];
658 
659 		pcie_capability_read_word(bridge, PCI_EXP_LNKSTA, &linksta);
660 		pcie_update_link_speed(bus, linksta);
661 	}
662 }
663 
664 static struct irq_domain *pci_host_bridge_msi_domain(struct pci_bus *bus)
665 {
666 	struct irq_domain *d;
667 
668 	/*
669 	 * Any firmware interface that can resolve the msi_domain
670 	 * should be called from here.
671 	 */
672 	d = pci_host_bridge_of_msi_domain(bus);
673 
674 	return d;
675 }
676 
677 static void pci_set_bus_msi_domain(struct pci_bus *bus)
678 {
679 	struct irq_domain *d;
680 	struct pci_bus *b;
681 
682 	/*
683 	 * The bus can be a root bus, a subordinate bus, or a virtual bus
684 	 * created by an SR-IOV device.  Walk up to the first bridge device
685 	 * found or derive the domain from the host bridge.
686 	 */
687 	for (b = bus, d = NULL; !d && !pci_is_root_bus(b); b = b->parent) {
688 		if (b->self)
689 			d = dev_get_msi_domain(&b->self->dev);
690 	}
691 
692 	if (!d)
693 		d = pci_host_bridge_msi_domain(b);
694 
695 	dev_set_msi_domain(&bus->dev, d);
696 }
697 
698 static struct pci_bus *pci_alloc_child_bus(struct pci_bus *parent,
699 					   struct pci_dev *bridge, int busnr)
700 {
701 	struct pci_bus *child;
702 	int i;
703 	int ret;
704 
705 	/*
706 	 * Allocate a new bus, and inherit stuff from the parent..
707 	 */
708 	child = pci_alloc_bus(parent);
709 	if (!child)
710 		return NULL;
711 
712 	child->parent = parent;
713 	child->ops = parent->ops;
714 	child->msi = parent->msi;
715 	child->sysdata = parent->sysdata;
716 	child->bus_flags = parent->bus_flags;
717 
718 	/* initialize some portions of the bus device, but don't register it
719 	 * now as the parent is not properly set up yet.
720 	 */
721 	child->dev.class = &pcibus_class;
722 	dev_set_name(&child->dev, "%04x:%02x", pci_domain_nr(child), busnr);
723 
724 	/*
725 	 * Set up the primary, secondary and subordinate
726 	 * bus numbers.
727 	 */
728 	child->number = child->busn_res.start = busnr;
729 	child->primary = parent->busn_res.start;
730 	child->busn_res.end = 0xff;
731 
732 	if (!bridge) {
733 		child->dev.parent = parent->bridge;
734 		goto add_dev;
735 	}
736 
737 	child->self = bridge;
738 	child->bridge = get_device(&bridge->dev);
739 	child->dev.parent = child->bridge;
740 	pci_set_bus_of_node(child);
741 	pci_set_bus_speed(child);
742 
743 	/* Set up default resource pointers and names.. */
744 	for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++) {
745 		child->resource[i] = &bridge->resource[PCI_BRIDGE_RESOURCES+i];
746 		child->resource[i]->name = child->name;
747 	}
748 	bridge->subordinate = child;
749 
750 add_dev:
751 	pci_set_bus_msi_domain(child);
752 	ret = device_register(&child->dev);
753 	WARN_ON(ret < 0);
754 
755 	pcibios_add_bus(child);
756 
757 	/* Create legacy_io and legacy_mem files for this bus */
758 	pci_create_legacy_files(child);
759 
760 	return child;
761 }
762 
763 struct pci_bus *pci_add_new_bus(struct pci_bus *parent, struct pci_dev *dev,
764 				int busnr)
765 {
766 	struct pci_bus *child;
767 
768 	child = pci_alloc_child_bus(parent, dev, busnr);
769 	if (child) {
770 		down_write(&pci_bus_sem);
771 		list_add_tail(&child->node, &parent->children);
772 		up_write(&pci_bus_sem);
773 	}
774 	return child;
775 }
776 EXPORT_SYMBOL(pci_add_new_bus);
777 
778 static void pci_enable_crs(struct pci_dev *pdev)
779 {
780 	u16 root_cap = 0;
781 
782 	/* Enable CRS Software Visibility if supported */
783 	pcie_capability_read_word(pdev, PCI_EXP_RTCAP, &root_cap);
784 	if (root_cap & PCI_EXP_RTCAP_CRSVIS)
785 		pcie_capability_set_word(pdev, PCI_EXP_RTCTL,
786 					 PCI_EXP_RTCTL_CRSSVE);
787 }
788 
789 /*
790  * If it's a bridge, configure it and scan the bus behind it.
791  * For CardBus bridges, we don't scan behind as the devices will
792  * be handled by the bridge driver itself.
793  *
794  * We need to process bridges in two passes -- first we scan those
795  * already configured by the BIOS and after we are done with all of
796  * them, we proceed to assigning numbers to the remaining buses in
797  * order to avoid overlaps between old and new bus numbers.
798  */
799 int pci_scan_bridge(struct pci_bus *bus, struct pci_dev *dev, int max, int pass)
800 {
801 	struct pci_bus *child;
802 	int is_cardbus = (dev->hdr_type == PCI_HEADER_TYPE_CARDBUS);
803 	u32 buses, i, j = 0;
804 	u16 bctl;
805 	u8 primary, secondary, subordinate;
806 	int broken = 0;
807 
808 	pci_read_config_dword(dev, PCI_PRIMARY_BUS, &buses);
809 	primary = buses & 0xFF;
810 	secondary = (buses >> 8) & 0xFF;
811 	subordinate = (buses >> 16) & 0xFF;
812 
813 	dev_dbg(&dev->dev, "scanning [bus %02x-%02x] behind bridge, pass %d\n",
814 		secondary, subordinate, pass);
815 
816 	if (!primary && (primary != bus->number) && secondary && subordinate) {
817 		dev_warn(&dev->dev, "Primary bus is hard wired to 0\n");
818 		primary = bus->number;
819 	}
820 
821 	/* Check if setup is sensible at all */
822 	if (!pass &&
823 	    (primary != bus->number || secondary <= bus->number ||
824 	     secondary > subordinate)) {
825 		dev_info(&dev->dev, "bridge configuration invalid ([bus %02x-%02x]), reconfiguring\n",
826 			 secondary, subordinate);
827 		broken = 1;
828 	}
829 
830 	/* Disable MasterAbortMode during probing to avoid reporting
831 	   of bus errors (in some architectures) */
832 	pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &bctl);
833 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL,
834 			      bctl & ~PCI_BRIDGE_CTL_MASTER_ABORT);
835 
836 	pci_enable_crs(dev);
837 
838 	if ((secondary || subordinate) && !pcibios_assign_all_busses() &&
839 	    !is_cardbus && !broken) {
840 		unsigned int cmax;
841 		/*
842 		 * Bus already configured by firmware, process it in the first
843 		 * pass and just note the configuration.
844 		 */
845 		if (pass)
846 			goto out;
847 
848 		/*
849 		 * The bus might already exist for two reasons: Either we are
850 		 * rescanning the bus or the bus is reachable through more than
851 		 * one bridge. The second case can happen with the i450NX
852 		 * chipset.
853 		 */
854 		child = pci_find_bus(pci_domain_nr(bus), secondary);
855 		if (!child) {
856 			child = pci_add_new_bus(bus, dev, secondary);
857 			if (!child)
858 				goto out;
859 			child->primary = primary;
860 			pci_bus_insert_busn_res(child, secondary, subordinate);
861 			child->bridge_ctl = bctl;
862 		}
863 
864 		cmax = pci_scan_child_bus(child);
865 		if (cmax > subordinate)
866 			dev_warn(&dev->dev, "bridge has subordinate %02x but max busn %02x\n",
867 				 subordinate, cmax);
868 		/* subordinate should equal child->busn_res.end */
869 		if (subordinate > max)
870 			max = subordinate;
871 	} else {
872 		/*
873 		 * We need to assign a number to this bus which we always
874 		 * do in the second pass.
875 		 */
876 		if (!pass) {
877 			if (pcibios_assign_all_busses() || broken || is_cardbus)
878 				/* Temporarily disable forwarding of the
879 				   configuration cycles on all bridges in
880 				   this bus segment to avoid possible
881 				   conflicts in the second pass between two
882 				   bridges programmed with overlapping
883 				   bus ranges. */
884 				pci_write_config_dword(dev, PCI_PRIMARY_BUS,
885 						       buses & ~0xffffff);
886 			goto out;
887 		}
888 
889 		/* Clear errors */
890 		pci_write_config_word(dev, PCI_STATUS, 0xffff);
891 
892 		/* Prevent assigning a bus number that already exists.
893 		 * This can happen when a bridge is hot-plugged, so in
894 		 * this case we only re-scan this bus. */
895 		child = pci_find_bus(pci_domain_nr(bus), max+1);
896 		if (!child) {
897 			child = pci_add_new_bus(bus, dev, max+1);
898 			if (!child)
899 				goto out;
900 			pci_bus_insert_busn_res(child, max+1, 0xff);
901 		}
902 		max++;
903 		buses = (buses & 0xff000000)
904 		      | ((unsigned int)(child->primary)     <<  0)
905 		      | ((unsigned int)(child->busn_res.start)   <<  8)
906 		      | ((unsigned int)(child->busn_res.end) << 16);
907 
908 		/*
909 		 * yenta.c forces a secondary latency timer of 176.
910 		 * Copy that behaviour here.
911 		 */
912 		if (is_cardbus) {
913 			buses &= ~0xff000000;
914 			buses |= CARDBUS_LATENCY_TIMER << 24;
915 		}
916 
917 		/*
918 		 * We need to blast all three values with a single write.
919 		 */
920 		pci_write_config_dword(dev, PCI_PRIMARY_BUS, buses);
921 
922 		if (!is_cardbus) {
923 			child->bridge_ctl = bctl;
924 			max = pci_scan_child_bus(child);
925 		} else {
926 			/*
927 			 * For CardBus bridges, we leave 4 bus numbers
928 			 * as cards with a PCI-to-PCI bridge can be
929 			 * inserted later.
930 			 */
931 			for (i = 0; i < CARDBUS_RESERVE_BUSNR; i++) {
932 				struct pci_bus *parent = bus;
933 				if (pci_find_bus(pci_domain_nr(bus),
934 							max+i+1))
935 					break;
936 				while (parent->parent) {
937 					if ((!pcibios_assign_all_busses()) &&
938 					    (parent->busn_res.end > max) &&
939 					    (parent->busn_res.end <= max+i)) {
940 						j = 1;
941 					}
942 					parent = parent->parent;
943 				}
944 				if (j) {
945 					/*
946 					 * Often, there are two cardbus bridges
947 					 * -- try to leave one valid bus number
948 					 * for each one.
949 					 */
950 					i /= 2;
951 					break;
952 				}
953 			}
954 			max += i;
955 		}
956 		/*
957 		 * Set the subordinate bus number to its real value.
958 		 */
959 		pci_bus_update_busn_res_end(child, max);
960 		pci_write_config_byte(dev, PCI_SUBORDINATE_BUS, max);
961 	}
962 
963 	sprintf(child->name,
964 		(is_cardbus ? "PCI CardBus %04x:%02x" : "PCI Bus %04x:%02x"),
965 		pci_domain_nr(bus), child->number);
966 
967 	/* Has only triggered on CardBus, fixup is in yenta_socket */
968 	while (bus->parent) {
969 		if ((child->busn_res.end > bus->busn_res.end) ||
970 		    (child->number > bus->busn_res.end) ||
971 		    (child->number < bus->number) ||
972 		    (child->busn_res.end < bus->number)) {
973 			dev_info(&child->dev, "%pR %s hidden behind%s bridge %s %pR\n",
974 				&child->busn_res,
975 				(bus->number > child->busn_res.end &&
976 				 bus->busn_res.end < child->number) ?
977 					"wholly" : "partially",
978 				bus->self->transparent ? " transparent" : "",
979 				dev_name(&bus->dev),
980 				&bus->busn_res);
981 		}
982 		bus = bus->parent;
983 	}
984 
985 out:
986 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL, bctl);
987 
988 	return max;
989 }
990 EXPORT_SYMBOL(pci_scan_bridge);
991 
992 /*
993  * Read interrupt line and base address registers.
994  * The architecture-dependent code can tweak these, of course.
995  */
996 static void pci_read_irq(struct pci_dev *dev)
997 {
998 	unsigned char irq;
999 
1000 	pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &irq);
1001 	dev->pin = irq;
1002 	if (irq)
1003 		pci_read_config_byte(dev, PCI_INTERRUPT_LINE, &irq);
1004 	dev->irq = irq;
1005 }
1006 
1007 void set_pcie_port_type(struct pci_dev *pdev)
1008 {
1009 	int pos;
1010 	u16 reg16;
1011 	int type;
1012 	struct pci_dev *parent;
1013 
1014 	pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
1015 	if (!pos)
1016 		return;
1017 	pdev->pcie_cap = pos;
1018 	pci_read_config_word(pdev, pos + PCI_EXP_FLAGS, &reg16);
1019 	pdev->pcie_flags_reg = reg16;
1020 	pci_read_config_word(pdev, pos + PCI_EXP_DEVCAP, &reg16);
1021 	pdev->pcie_mpss = reg16 & PCI_EXP_DEVCAP_PAYLOAD;
1022 
1023 	/*
1024 	 * A Root Port is always the upstream end of a Link.  No PCIe
1025 	 * component has two Links.  Two Links are connected by a Switch
1026 	 * that has a Port on each Link and internal logic to connect the
1027 	 * two Ports.
1028 	 */
1029 	type = pci_pcie_type(pdev);
1030 	if (type == PCI_EXP_TYPE_ROOT_PORT)
1031 		pdev->has_secondary_link = 1;
1032 	else if (type == PCI_EXP_TYPE_UPSTREAM ||
1033 		 type == PCI_EXP_TYPE_DOWNSTREAM) {
1034 		parent = pci_upstream_bridge(pdev);
1035 
1036 		/*
1037 		 * Usually there's an upstream device (Root Port or Switch
1038 		 * Downstream Port), but we can't assume one exists.
1039 		 */
1040 		if (parent && !parent->has_secondary_link)
1041 			pdev->has_secondary_link = 1;
1042 	}
1043 }
1044 
1045 void set_pcie_hotplug_bridge(struct pci_dev *pdev)
1046 {
1047 	u32 reg32;
1048 
1049 	pcie_capability_read_dword(pdev, PCI_EXP_SLTCAP, &reg32);
1050 	if (reg32 & PCI_EXP_SLTCAP_HPC)
1051 		pdev->is_hotplug_bridge = 1;
1052 }
1053 
1054 /**
1055  * pci_ext_cfg_is_aliased - is ext config space just an alias of std config?
1056  * @dev: PCI device
1057  *
1058  * PCI Express to PCI/PCI-X Bridge Specification, rev 1.0, 4.1.4 says that
1059  * when forwarding a type1 configuration request the bridge must check that
1060  * the extended register address field is zero.  The bridge is not permitted
1061  * to forward the transactions and must handle it as an Unsupported Request.
1062  * Some bridges do not follow this rule and simply drop the extended register
1063  * bits, resulting in the standard config space being aliased, every 256
1064  * bytes across the entire configuration space.  Test for this condition by
1065  * comparing the first dword of each potential alias to the vendor/device ID.
1066  * Known offenders:
1067  *   ASM1083/1085 PCIe-to-PCI Reversible Bridge (1b21:1080, rev 01 & 03)
1068  *   AMD/ATI SBx00 PCI to PCI Bridge (1002:4384, rev 40)
1069  */
1070 static bool pci_ext_cfg_is_aliased(struct pci_dev *dev)
1071 {
1072 #ifdef CONFIG_PCI_QUIRKS
1073 	int pos;
1074 	u32 header, tmp;
1075 
1076 	pci_read_config_dword(dev, PCI_VENDOR_ID, &header);
1077 
1078 	for (pos = PCI_CFG_SPACE_SIZE;
1079 	     pos < PCI_CFG_SPACE_EXP_SIZE; pos += PCI_CFG_SPACE_SIZE) {
1080 		if (pci_read_config_dword(dev, pos, &tmp) != PCIBIOS_SUCCESSFUL
1081 		    || header != tmp)
1082 			return false;
1083 	}
1084 
1085 	return true;
1086 #else
1087 	return false;
1088 #endif
1089 }
1090 
1091 /**
1092  * pci_cfg_space_size - get the configuration space size of the PCI device.
1093  * @dev: PCI device
1094  *
1095  * Regular PCI devices have 256 bytes, but PCI-X 2 and PCI Express devices
1096  * have 4096 bytes.  Even if the device is capable, that doesn't mean we can
1097  * access it.  Maybe we don't have a way to generate extended config space
1098  * accesses, or the device is behind a reverse Express bridge.  So we try
1099  * reading the dword at 0x100 which must either be 0 or a valid extended
1100  * capability header.
1101  */
1102 static int pci_cfg_space_size_ext(struct pci_dev *dev)
1103 {
1104 	u32 status;
1105 	int pos = PCI_CFG_SPACE_SIZE;
1106 
1107 	if (pci_read_config_dword(dev, pos, &status) != PCIBIOS_SUCCESSFUL)
1108 		goto fail;
1109 	if (status == 0xffffffff || pci_ext_cfg_is_aliased(dev))
1110 		goto fail;
1111 
1112 	return PCI_CFG_SPACE_EXP_SIZE;
1113 
1114  fail:
1115 	return PCI_CFG_SPACE_SIZE;
1116 }
1117 
1118 int pci_cfg_space_size(struct pci_dev *dev)
1119 {
1120 	int pos;
1121 	u32 status;
1122 	u16 class;
1123 
1124 	class = dev->class >> 8;
1125 	if (class == PCI_CLASS_BRIDGE_HOST)
1126 		return pci_cfg_space_size_ext(dev);
1127 
1128 	if (!pci_is_pcie(dev)) {
1129 		pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1130 		if (!pos)
1131 			goto fail;
1132 
1133 		pci_read_config_dword(dev, pos + PCI_X_STATUS, &status);
1134 		if (!(status & (PCI_X_STATUS_266MHZ | PCI_X_STATUS_533MHZ)))
1135 			goto fail;
1136 	}
1137 
1138 	return pci_cfg_space_size_ext(dev);
1139 
1140  fail:
1141 	return PCI_CFG_SPACE_SIZE;
1142 }
1143 
1144 #define LEGACY_IO_RESOURCE	(IORESOURCE_IO | IORESOURCE_PCI_FIXED)
1145 
1146 void pci_msi_setup_pci_dev(struct pci_dev *dev)
1147 {
1148 	/*
1149 	 * Disable the MSI hardware to avoid screaming interrupts
1150 	 * during boot.  This is the power on reset default so
1151 	 * usually this should be a noop.
1152 	 */
1153 	dev->msi_cap = pci_find_capability(dev, PCI_CAP_ID_MSI);
1154 	if (dev->msi_cap)
1155 		pci_msi_set_enable(dev, 0);
1156 
1157 	dev->msix_cap = pci_find_capability(dev, PCI_CAP_ID_MSIX);
1158 	if (dev->msix_cap)
1159 		pci_msix_clear_and_set_ctrl(dev, PCI_MSIX_FLAGS_ENABLE, 0);
1160 }
1161 
1162 /**
1163  * pci_setup_device - fill in class and map information of a device
1164  * @dev: the device structure to fill
1165  *
1166  * Initialize the device structure with information about the device's
1167  * vendor,class,memory and IO-space addresses,IRQ lines etc.
1168  * Called at initialisation of the PCI subsystem and by CardBus services.
1169  * Returns 0 on success and negative if unknown type of device (not normal,
1170  * bridge or CardBus).
1171  */
1172 int pci_setup_device(struct pci_dev *dev)
1173 {
1174 	u32 class;
1175 	u8 hdr_type;
1176 	int pos = 0;
1177 	struct pci_bus_region region;
1178 	struct resource *res;
1179 
1180 	if (pci_read_config_byte(dev, PCI_HEADER_TYPE, &hdr_type))
1181 		return -EIO;
1182 
1183 	dev->sysdata = dev->bus->sysdata;
1184 	dev->dev.parent = dev->bus->bridge;
1185 	dev->dev.bus = &pci_bus_type;
1186 	dev->hdr_type = hdr_type & 0x7f;
1187 	dev->multifunction = !!(hdr_type & 0x80);
1188 	dev->error_state = pci_channel_io_normal;
1189 	set_pcie_port_type(dev);
1190 
1191 	pci_dev_assign_slot(dev);
1192 	/* Assume 32-bit PCI; let 64-bit PCI cards (which are far rarer)
1193 	   set this higher, assuming the system even supports it.  */
1194 	dev->dma_mask = 0xffffffff;
1195 
1196 	dev_set_name(&dev->dev, "%04x:%02x:%02x.%d", pci_domain_nr(dev->bus),
1197 		     dev->bus->number, PCI_SLOT(dev->devfn),
1198 		     PCI_FUNC(dev->devfn));
1199 
1200 	pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
1201 	dev->revision = class & 0xff;
1202 	dev->class = class >> 8;		    /* upper 3 bytes */
1203 
1204 	dev_printk(KERN_DEBUG, &dev->dev, "[%04x:%04x] type %02x class %#08x\n",
1205 		   dev->vendor, dev->device, dev->hdr_type, dev->class);
1206 
1207 	/* need to have dev->class ready */
1208 	dev->cfg_size = pci_cfg_space_size(dev);
1209 
1210 	/* "Unknown power state" */
1211 	dev->current_state = PCI_UNKNOWN;
1212 
1213 	pci_msi_setup_pci_dev(dev);
1214 
1215 	/* Early fixups, before probing the BARs */
1216 	pci_fixup_device(pci_fixup_early, dev);
1217 	/* device class may be changed after fixup */
1218 	class = dev->class >> 8;
1219 
1220 	switch (dev->hdr_type) {		    /* header type */
1221 	case PCI_HEADER_TYPE_NORMAL:		    /* standard header */
1222 		if (class == PCI_CLASS_BRIDGE_PCI)
1223 			goto bad;
1224 		pci_read_irq(dev);
1225 		pci_read_bases(dev, 6, PCI_ROM_ADDRESS);
1226 		pci_read_config_word(dev, PCI_SUBSYSTEM_VENDOR_ID, &dev->subsystem_vendor);
1227 		pci_read_config_word(dev, PCI_SUBSYSTEM_ID, &dev->subsystem_device);
1228 
1229 		/*
1230 		 * Do the ugly legacy mode stuff here rather than broken chip
1231 		 * quirk code. Legacy mode ATA controllers have fixed
1232 		 * addresses. These are not always echoed in BAR0-3, and
1233 		 * BAR0-3 in a few cases contain junk!
1234 		 */
1235 		if (class == PCI_CLASS_STORAGE_IDE) {
1236 			u8 progif;
1237 			pci_read_config_byte(dev, PCI_CLASS_PROG, &progif);
1238 			if ((progif & 1) == 0) {
1239 				region.start = 0x1F0;
1240 				region.end = 0x1F7;
1241 				res = &dev->resource[0];
1242 				res->flags = LEGACY_IO_RESOURCE;
1243 				pcibios_bus_to_resource(dev->bus, res, &region);
1244 				dev_info(&dev->dev, "legacy IDE quirk: reg 0x10: %pR\n",
1245 					 res);
1246 				region.start = 0x3F6;
1247 				region.end = 0x3F6;
1248 				res = &dev->resource[1];
1249 				res->flags = LEGACY_IO_RESOURCE;
1250 				pcibios_bus_to_resource(dev->bus, res, &region);
1251 				dev_info(&dev->dev, "legacy IDE quirk: reg 0x14: %pR\n",
1252 					 res);
1253 			}
1254 			if ((progif & 4) == 0) {
1255 				region.start = 0x170;
1256 				region.end = 0x177;
1257 				res = &dev->resource[2];
1258 				res->flags = LEGACY_IO_RESOURCE;
1259 				pcibios_bus_to_resource(dev->bus, res, &region);
1260 				dev_info(&dev->dev, "legacy IDE quirk: reg 0x18: %pR\n",
1261 					 res);
1262 				region.start = 0x376;
1263 				region.end = 0x376;
1264 				res = &dev->resource[3];
1265 				res->flags = LEGACY_IO_RESOURCE;
1266 				pcibios_bus_to_resource(dev->bus, res, &region);
1267 				dev_info(&dev->dev, "legacy IDE quirk: reg 0x1c: %pR\n",
1268 					 res);
1269 			}
1270 		}
1271 		break;
1272 
1273 	case PCI_HEADER_TYPE_BRIDGE:		    /* bridge header */
1274 		if (class != PCI_CLASS_BRIDGE_PCI)
1275 			goto bad;
1276 		/* The PCI-to-PCI bridge spec requires that subtractive
1277 		   decoding (i.e. transparent) bridge must have programming
1278 		   interface code of 0x01. */
1279 		pci_read_irq(dev);
1280 		dev->transparent = ((dev->class & 0xff) == 1);
1281 		pci_read_bases(dev, 2, PCI_ROM_ADDRESS1);
1282 		set_pcie_hotplug_bridge(dev);
1283 		pos = pci_find_capability(dev, PCI_CAP_ID_SSVID);
1284 		if (pos) {
1285 			pci_read_config_word(dev, pos + PCI_SSVID_VENDOR_ID, &dev->subsystem_vendor);
1286 			pci_read_config_word(dev, pos + PCI_SSVID_DEVICE_ID, &dev->subsystem_device);
1287 		}
1288 		break;
1289 
1290 	case PCI_HEADER_TYPE_CARDBUS:		    /* CardBus bridge header */
1291 		if (class != PCI_CLASS_BRIDGE_CARDBUS)
1292 			goto bad;
1293 		pci_read_irq(dev);
1294 		pci_read_bases(dev, 1, 0);
1295 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_VENDOR_ID, &dev->subsystem_vendor);
1296 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_ID, &dev->subsystem_device);
1297 		break;
1298 
1299 	default:				    /* unknown header */
1300 		dev_err(&dev->dev, "unknown header type %02x, ignoring device\n",
1301 			dev->hdr_type);
1302 		return -EIO;
1303 
1304 	bad:
1305 		dev_err(&dev->dev, "ignoring class %#08x (doesn't match header type %02x)\n",
1306 			dev->class, dev->hdr_type);
1307 		dev->class = PCI_CLASS_NOT_DEFINED << 8;
1308 	}
1309 
1310 	/* We found a fine healthy device, go go go... */
1311 	return 0;
1312 }
1313 
1314 static void pci_configure_mps(struct pci_dev *dev)
1315 {
1316 	struct pci_dev *bridge = pci_upstream_bridge(dev);
1317 	int mps, p_mps, rc;
1318 
1319 	if (!pci_is_pcie(dev) || !bridge || !pci_is_pcie(bridge))
1320 		return;
1321 
1322 	mps = pcie_get_mps(dev);
1323 	p_mps = pcie_get_mps(bridge);
1324 
1325 	if (mps == p_mps)
1326 		return;
1327 
1328 	if (pcie_bus_config == PCIE_BUS_TUNE_OFF) {
1329 		dev_warn(&dev->dev, "Max Payload Size %d, but upstream %s set to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
1330 			 mps, pci_name(bridge), p_mps);
1331 		return;
1332 	}
1333 
1334 	/*
1335 	 * Fancier MPS configuration is done later by
1336 	 * pcie_bus_configure_settings()
1337 	 */
1338 	if (pcie_bus_config != PCIE_BUS_DEFAULT)
1339 		return;
1340 
1341 	rc = pcie_set_mps(dev, p_mps);
1342 	if (rc) {
1343 		dev_warn(&dev->dev, "can't set Max Payload Size to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
1344 			 p_mps);
1345 		return;
1346 	}
1347 
1348 	dev_info(&dev->dev, "Max Payload Size set to %d (was %d, max %d)\n",
1349 		 p_mps, mps, 128 << dev->pcie_mpss);
1350 }
1351 
1352 static struct hpp_type0 pci_default_type0 = {
1353 	.revision = 1,
1354 	.cache_line_size = 8,
1355 	.latency_timer = 0x40,
1356 	.enable_serr = 0,
1357 	.enable_perr = 0,
1358 };
1359 
1360 static void program_hpp_type0(struct pci_dev *dev, struct hpp_type0 *hpp)
1361 {
1362 	u16 pci_cmd, pci_bctl;
1363 
1364 	if (!hpp)
1365 		hpp = &pci_default_type0;
1366 
1367 	if (hpp->revision > 1) {
1368 		dev_warn(&dev->dev,
1369 			 "PCI settings rev %d not supported; using defaults\n",
1370 			 hpp->revision);
1371 		hpp = &pci_default_type0;
1372 	}
1373 
1374 	pci_write_config_byte(dev, PCI_CACHE_LINE_SIZE, hpp->cache_line_size);
1375 	pci_write_config_byte(dev, PCI_LATENCY_TIMER, hpp->latency_timer);
1376 	pci_read_config_word(dev, PCI_COMMAND, &pci_cmd);
1377 	if (hpp->enable_serr)
1378 		pci_cmd |= PCI_COMMAND_SERR;
1379 	if (hpp->enable_perr)
1380 		pci_cmd |= PCI_COMMAND_PARITY;
1381 	pci_write_config_word(dev, PCI_COMMAND, pci_cmd);
1382 
1383 	/* Program bridge control value */
1384 	if ((dev->class >> 8) == PCI_CLASS_BRIDGE_PCI) {
1385 		pci_write_config_byte(dev, PCI_SEC_LATENCY_TIMER,
1386 				      hpp->latency_timer);
1387 		pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &pci_bctl);
1388 		if (hpp->enable_serr)
1389 			pci_bctl |= PCI_BRIDGE_CTL_SERR;
1390 		if (hpp->enable_perr)
1391 			pci_bctl |= PCI_BRIDGE_CTL_PARITY;
1392 		pci_write_config_word(dev, PCI_BRIDGE_CONTROL, pci_bctl);
1393 	}
1394 }
1395 
1396 static void program_hpp_type1(struct pci_dev *dev, struct hpp_type1 *hpp)
1397 {
1398 	if (hpp)
1399 		dev_warn(&dev->dev, "PCI-X settings not supported\n");
1400 }
1401 
1402 static void program_hpp_type2(struct pci_dev *dev, struct hpp_type2 *hpp)
1403 {
1404 	int pos;
1405 	u32 reg32;
1406 
1407 	if (!hpp)
1408 		return;
1409 
1410 	if (hpp->revision > 1) {
1411 		dev_warn(&dev->dev, "PCIe settings rev %d not supported\n",
1412 			 hpp->revision);
1413 		return;
1414 	}
1415 
1416 	/*
1417 	 * Don't allow _HPX to change MPS or MRRS settings.  We manage
1418 	 * those to make sure they're consistent with the rest of the
1419 	 * platform.
1420 	 */
1421 	hpp->pci_exp_devctl_and |= PCI_EXP_DEVCTL_PAYLOAD |
1422 				    PCI_EXP_DEVCTL_READRQ;
1423 	hpp->pci_exp_devctl_or &= ~(PCI_EXP_DEVCTL_PAYLOAD |
1424 				    PCI_EXP_DEVCTL_READRQ);
1425 
1426 	/* Initialize Device Control Register */
1427 	pcie_capability_clear_and_set_word(dev, PCI_EXP_DEVCTL,
1428 			~hpp->pci_exp_devctl_and, hpp->pci_exp_devctl_or);
1429 
1430 	/* Initialize Link Control Register */
1431 	if (pcie_cap_has_lnkctl(dev))
1432 		pcie_capability_clear_and_set_word(dev, PCI_EXP_LNKCTL,
1433 			~hpp->pci_exp_lnkctl_and, hpp->pci_exp_lnkctl_or);
1434 
1435 	/* Find Advanced Error Reporting Enhanced Capability */
1436 	pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
1437 	if (!pos)
1438 		return;
1439 
1440 	/* Initialize Uncorrectable Error Mask Register */
1441 	pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, &reg32);
1442 	reg32 = (reg32 & hpp->unc_err_mask_and) | hpp->unc_err_mask_or;
1443 	pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, reg32);
1444 
1445 	/* Initialize Uncorrectable Error Severity Register */
1446 	pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &reg32);
1447 	reg32 = (reg32 & hpp->unc_err_sever_and) | hpp->unc_err_sever_or;
1448 	pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, reg32);
1449 
1450 	/* Initialize Correctable Error Mask Register */
1451 	pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, &reg32);
1452 	reg32 = (reg32 & hpp->cor_err_mask_and) | hpp->cor_err_mask_or;
1453 	pci_write_config_dword(dev, pos + PCI_ERR_COR_MASK, reg32);
1454 
1455 	/* Initialize Advanced Error Capabilities and Control Register */
1456 	pci_read_config_dword(dev, pos + PCI_ERR_CAP, &reg32);
1457 	reg32 = (reg32 & hpp->adv_err_cap_and) | hpp->adv_err_cap_or;
1458 	pci_write_config_dword(dev, pos + PCI_ERR_CAP, reg32);
1459 
1460 	/*
1461 	 * FIXME: The following two registers are not supported yet.
1462 	 *
1463 	 *   o Secondary Uncorrectable Error Severity Register
1464 	 *   o Secondary Uncorrectable Error Mask Register
1465 	 */
1466 }
1467 
1468 static void pci_configure_device(struct pci_dev *dev)
1469 {
1470 	struct hotplug_params hpp;
1471 	int ret;
1472 
1473 	pci_configure_mps(dev);
1474 
1475 	memset(&hpp, 0, sizeof(hpp));
1476 	ret = pci_get_hp_params(dev, &hpp);
1477 	if (ret)
1478 		return;
1479 
1480 	program_hpp_type2(dev, hpp.t2);
1481 	program_hpp_type1(dev, hpp.t1);
1482 	program_hpp_type0(dev, hpp.t0);
1483 }
1484 
1485 static void pci_release_capabilities(struct pci_dev *dev)
1486 {
1487 	pci_vpd_release(dev);
1488 	pci_iov_release(dev);
1489 	pci_free_cap_save_buffers(dev);
1490 }
1491 
1492 /**
1493  * pci_release_dev - free a pci device structure when all users of it are finished.
1494  * @dev: device that's been disconnected
1495  *
1496  * Will be called only by the device core when all users of this pci device are
1497  * done.
1498  */
1499 static void pci_release_dev(struct device *dev)
1500 {
1501 	struct pci_dev *pci_dev;
1502 
1503 	pci_dev = to_pci_dev(dev);
1504 	pci_release_capabilities(pci_dev);
1505 	pci_release_of_node(pci_dev);
1506 	pcibios_release_device(pci_dev);
1507 	pci_bus_put(pci_dev->bus);
1508 	kfree(pci_dev->driver_override);
1509 	kfree(pci_dev);
1510 }
1511 
1512 struct pci_dev *pci_alloc_dev(struct pci_bus *bus)
1513 {
1514 	struct pci_dev *dev;
1515 
1516 	dev = kzalloc(sizeof(struct pci_dev), GFP_KERNEL);
1517 	if (!dev)
1518 		return NULL;
1519 
1520 	INIT_LIST_HEAD(&dev->bus_list);
1521 	dev->dev.type = &pci_dev_type;
1522 	dev->bus = pci_bus_get(bus);
1523 
1524 	return dev;
1525 }
1526 EXPORT_SYMBOL(pci_alloc_dev);
1527 
1528 bool pci_bus_read_dev_vendor_id(struct pci_bus *bus, int devfn, u32 *l,
1529 				int crs_timeout)
1530 {
1531 	int delay = 1;
1532 
1533 	if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
1534 		return false;
1535 
1536 	/* some broken boards return 0 or ~0 if a slot is empty: */
1537 	if (*l == 0xffffffff || *l == 0x00000000 ||
1538 	    *l == 0x0000ffff || *l == 0xffff0000)
1539 		return false;
1540 
1541 	/*
1542 	 * Configuration Request Retry Status.  Some root ports return the
1543 	 * actual device ID instead of the synthetic ID (0xFFFF) required
1544 	 * by the PCIe spec.  Ignore the device ID and only check for
1545 	 * (vendor id == 1).
1546 	 */
1547 	while ((*l & 0xffff) == 0x0001) {
1548 		if (!crs_timeout)
1549 			return false;
1550 
1551 		msleep(delay);
1552 		delay *= 2;
1553 		if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
1554 			return false;
1555 		/* Card hasn't responded in 60 seconds?  Must be stuck. */
1556 		if (delay > crs_timeout) {
1557 			printk(KERN_WARNING "pci %04x:%02x:%02x.%d: not responding\n",
1558 			       pci_domain_nr(bus), bus->number, PCI_SLOT(devfn),
1559 			       PCI_FUNC(devfn));
1560 			return false;
1561 		}
1562 	}
1563 
1564 	return true;
1565 }
1566 EXPORT_SYMBOL(pci_bus_read_dev_vendor_id);
1567 
1568 /*
1569  * Read the config data for a PCI device, sanity-check it
1570  * and fill in the dev structure...
1571  */
1572 static struct pci_dev *pci_scan_device(struct pci_bus *bus, int devfn)
1573 {
1574 	struct pci_dev *dev;
1575 	u32 l;
1576 
1577 	if (!pci_bus_read_dev_vendor_id(bus, devfn, &l, 60*1000))
1578 		return NULL;
1579 
1580 	dev = pci_alloc_dev(bus);
1581 	if (!dev)
1582 		return NULL;
1583 
1584 	dev->devfn = devfn;
1585 	dev->vendor = l & 0xffff;
1586 	dev->device = (l >> 16) & 0xffff;
1587 
1588 	pci_set_of_node(dev);
1589 
1590 	if (pci_setup_device(dev)) {
1591 		pci_bus_put(dev->bus);
1592 		kfree(dev);
1593 		return NULL;
1594 	}
1595 
1596 	return dev;
1597 }
1598 
1599 static void pci_init_capabilities(struct pci_dev *dev)
1600 {
1601 	/* Enhanced Allocation */
1602 	pci_ea_init(dev);
1603 
1604 	/* MSI/MSI-X list */
1605 	pci_msi_init_pci_dev(dev);
1606 
1607 	/* Buffers for saving PCIe and PCI-X capabilities */
1608 	pci_allocate_cap_save_buffers(dev);
1609 
1610 	/* Power Management */
1611 	pci_pm_init(dev);
1612 
1613 	/* Vital Product Data */
1614 	pci_vpd_pci22_init(dev);
1615 
1616 	/* Alternative Routing-ID Forwarding */
1617 	pci_configure_ari(dev);
1618 
1619 	/* Single Root I/O Virtualization */
1620 	pci_iov_init(dev);
1621 
1622 	/* Address Translation Services */
1623 	pci_ats_init(dev);
1624 
1625 	/* Enable ACS P2P upstream forwarding */
1626 	pci_enable_acs(dev);
1627 
1628 	pci_cleanup_aer_error_status_regs(dev);
1629 }
1630 
1631 /*
1632  * This is the equivalent of pci_host_bridge_msi_domain that acts on
1633  * devices. Firmware interfaces that can select the MSI domain on a
1634  * per-device basis should be called from here.
1635  */
1636 static struct irq_domain *pci_dev_msi_domain(struct pci_dev *dev)
1637 {
1638 	struct irq_domain *d;
1639 
1640 	/*
1641 	 * If a domain has been set through the pcibios_add_device
1642 	 * callback, then this is the one (platform code knows best).
1643 	 */
1644 	d = dev_get_msi_domain(&dev->dev);
1645 	if (d)
1646 		return d;
1647 
1648 	/*
1649 	 * Let's see if we have a firmware interface able to provide
1650 	 * the domain.
1651 	 */
1652 	d = pci_msi_get_device_domain(dev);
1653 	if (d)
1654 		return d;
1655 
1656 	return NULL;
1657 }
1658 
1659 static void pci_set_msi_domain(struct pci_dev *dev)
1660 {
1661 	struct irq_domain *d;
1662 
1663 	/*
1664 	 * If the platform or firmware interfaces cannot supply a
1665 	 * device-specific MSI domain, then inherit the default domain
1666 	 * from the host bridge itself.
1667 	 */
1668 	d = pci_dev_msi_domain(dev);
1669 	if (!d)
1670 		d = dev_get_msi_domain(&dev->bus->dev);
1671 
1672 	dev_set_msi_domain(&dev->dev, d);
1673 }
1674 
1675 void pci_device_add(struct pci_dev *dev, struct pci_bus *bus)
1676 {
1677 	int ret;
1678 
1679 	pci_configure_device(dev);
1680 
1681 	device_initialize(&dev->dev);
1682 	dev->dev.release = pci_release_dev;
1683 
1684 	set_dev_node(&dev->dev, pcibus_to_node(bus));
1685 	dev->dev.dma_mask = &dev->dma_mask;
1686 	dev->dev.dma_parms = &dev->dma_parms;
1687 	dev->dev.coherent_dma_mask = 0xffffffffull;
1688 	of_pci_dma_configure(dev);
1689 
1690 	pci_set_dma_max_seg_size(dev, 65536);
1691 	pci_set_dma_seg_boundary(dev, 0xffffffff);
1692 
1693 	/* Fix up broken headers */
1694 	pci_fixup_device(pci_fixup_header, dev);
1695 
1696 	/* moved out from quirk header fixup code */
1697 	pci_reassigndev_resource_alignment(dev);
1698 
1699 	/* Clear the state_saved flag. */
1700 	dev->state_saved = false;
1701 
1702 	/* Initialize various capabilities */
1703 	pci_init_capabilities(dev);
1704 
1705 	/*
1706 	 * Add the device to our list of discovered devices
1707 	 * and the bus list for fixup functions, etc.
1708 	 */
1709 	down_write(&pci_bus_sem);
1710 	list_add_tail(&dev->bus_list, &bus->devices);
1711 	up_write(&pci_bus_sem);
1712 
1713 	ret = pcibios_add_device(dev);
1714 	WARN_ON(ret < 0);
1715 
1716 	/* Setup MSI irq domain */
1717 	pci_set_msi_domain(dev);
1718 
1719 	/* Notifier could use PCI capabilities */
1720 	dev->match_driver = false;
1721 	ret = device_add(&dev->dev);
1722 	WARN_ON(ret < 0);
1723 }
1724 
1725 struct pci_dev *pci_scan_single_device(struct pci_bus *bus, int devfn)
1726 {
1727 	struct pci_dev *dev;
1728 
1729 	dev = pci_get_slot(bus, devfn);
1730 	if (dev) {
1731 		pci_dev_put(dev);
1732 		return dev;
1733 	}
1734 
1735 	dev = pci_scan_device(bus, devfn);
1736 	if (!dev)
1737 		return NULL;
1738 
1739 	pci_device_add(dev, bus);
1740 
1741 	return dev;
1742 }
1743 EXPORT_SYMBOL(pci_scan_single_device);
1744 
1745 static unsigned next_fn(struct pci_bus *bus, struct pci_dev *dev, unsigned fn)
1746 {
1747 	int pos;
1748 	u16 cap = 0;
1749 	unsigned next_fn;
1750 
1751 	if (pci_ari_enabled(bus)) {
1752 		if (!dev)
1753 			return 0;
1754 		pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ARI);
1755 		if (!pos)
1756 			return 0;
1757 
1758 		pci_read_config_word(dev, pos + PCI_ARI_CAP, &cap);
1759 		next_fn = PCI_ARI_CAP_NFN(cap);
1760 		if (next_fn <= fn)
1761 			return 0;	/* protect against malformed list */
1762 
1763 		return next_fn;
1764 	}
1765 
1766 	/* dev may be NULL for non-contiguous multifunction devices */
1767 	if (!dev || dev->multifunction)
1768 		return (fn + 1) % 8;
1769 
1770 	return 0;
1771 }
1772 
1773 static int only_one_child(struct pci_bus *bus)
1774 {
1775 	struct pci_dev *parent = bus->self;
1776 
1777 	if (!parent || !pci_is_pcie(parent))
1778 		return 0;
1779 	if (pci_pcie_type(parent) == PCI_EXP_TYPE_ROOT_PORT)
1780 		return 1;
1781 	if (parent->has_secondary_link &&
1782 	    !pci_has_flag(PCI_SCAN_ALL_PCIE_DEVS))
1783 		return 1;
1784 	return 0;
1785 }
1786 
1787 /**
1788  * pci_scan_slot - scan a PCI slot on a bus for devices.
1789  * @bus: PCI bus to scan
1790  * @devfn: slot number to scan (must have zero function.)
1791  *
1792  * Scan a PCI slot on the specified PCI bus for devices, adding
1793  * discovered devices to the @bus->devices list.  New devices
1794  * will not have is_added set.
1795  *
1796  * Returns the number of new devices found.
1797  */
1798 int pci_scan_slot(struct pci_bus *bus, int devfn)
1799 {
1800 	unsigned fn, nr = 0;
1801 	struct pci_dev *dev;
1802 
1803 	if (only_one_child(bus) && (devfn > 0))
1804 		return 0; /* Already scanned the entire slot */
1805 
1806 	dev = pci_scan_single_device(bus, devfn);
1807 	if (!dev)
1808 		return 0;
1809 	if (!dev->is_added)
1810 		nr++;
1811 
1812 	for (fn = next_fn(bus, dev, 0); fn > 0; fn = next_fn(bus, dev, fn)) {
1813 		dev = pci_scan_single_device(bus, devfn + fn);
1814 		if (dev) {
1815 			if (!dev->is_added)
1816 				nr++;
1817 			dev->multifunction = 1;
1818 		}
1819 	}
1820 
1821 	/* only one slot has pcie device */
1822 	if (bus->self && nr)
1823 		pcie_aspm_init_link_state(bus->self);
1824 
1825 	return nr;
1826 }
1827 EXPORT_SYMBOL(pci_scan_slot);
1828 
1829 static int pcie_find_smpss(struct pci_dev *dev, void *data)
1830 {
1831 	u8 *smpss = data;
1832 
1833 	if (!pci_is_pcie(dev))
1834 		return 0;
1835 
1836 	/*
1837 	 * We don't have a way to change MPS settings on devices that have
1838 	 * drivers attached.  A hot-added device might support only the minimum
1839 	 * MPS setting (MPS=128).  Therefore, if the fabric contains a bridge
1840 	 * where devices may be hot-added, we limit the fabric MPS to 128 so
1841 	 * hot-added devices will work correctly.
1842 	 *
1843 	 * However, if we hot-add a device to a slot directly below a Root
1844 	 * Port, it's impossible for there to be other existing devices below
1845 	 * the port.  We don't limit the MPS in this case because we can
1846 	 * reconfigure MPS on both the Root Port and the hot-added device,
1847 	 * and there are no other devices involved.
1848 	 *
1849 	 * Note that this PCIE_BUS_SAFE path assumes no peer-to-peer DMA.
1850 	 */
1851 	if (dev->is_hotplug_bridge &&
1852 	    pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT)
1853 		*smpss = 0;
1854 
1855 	if (*smpss > dev->pcie_mpss)
1856 		*smpss = dev->pcie_mpss;
1857 
1858 	return 0;
1859 }
1860 
1861 static void pcie_write_mps(struct pci_dev *dev, int mps)
1862 {
1863 	int rc;
1864 
1865 	if (pcie_bus_config == PCIE_BUS_PERFORMANCE) {
1866 		mps = 128 << dev->pcie_mpss;
1867 
1868 		if (pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT &&
1869 		    dev->bus->self)
1870 			/* For "Performance", the assumption is made that
1871 			 * downstream communication will never be larger than
1872 			 * the MRRS.  So, the MPS only needs to be configured
1873 			 * for the upstream communication.  This being the case,
1874 			 * walk from the top down and set the MPS of the child
1875 			 * to that of the parent bus.
1876 			 *
1877 			 * Configure the device MPS with the smaller of the
1878 			 * device MPSS or the bridge MPS (which is assumed to be
1879 			 * properly configured at this point to the largest
1880 			 * allowable MPS based on its parent bus).
1881 			 */
1882 			mps = min(mps, pcie_get_mps(dev->bus->self));
1883 	}
1884 
1885 	rc = pcie_set_mps(dev, mps);
1886 	if (rc)
1887 		dev_err(&dev->dev, "Failed attempting to set the MPS\n");
1888 }
1889 
1890 static void pcie_write_mrrs(struct pci_dev *dev)
1891 {
1892 	int rc, mrrs;
1893 
1894 	/* In the "safe" case, do not configure the MRRS.  There appear to be
1895 	 * issues with setting MRRS to 0 on a number of devices.
1896 	 */
1897 	if (pcie_bus_config != PCIE_BUS_PERFORMANCE)
1898 		return;
1899 
1900 	/* For Max performance, the MRRS must be set to the largest supported
1901 	 * value.  However, it cannot be configured larger than the MPS the
1902 	 * device or the bus can support.  This should already be properly
1903 	 * configured by a prior call to pcie_write_mps.
1904 	 */
1905 	mrrs = pcie_get_mps(dev);
1906 
1907 	/* MRRS is a R/W register.  Invalid values can be written, but a
1908 	 * subsequent read will verify if the value is acceptable or not.
1909 	 * If the MRRS value provided is not acceptable (e.g., too large),
1910 	 * shrink the value until it is acceptable to the HW.
1911 	 */
1912 	while (mrrs != pcie_get_readrq(dev) && mrrs >= 128) {
1913 		rc = pcie_set_readrq(dev, mrrs);
1914 		if (!rc)
1915 			break;
1916 
1917 		dev_warn(&dev->dev, "Failed attempting to set the MRRS\n");
1918 		mrrs /= 2;
1919 	}
1920 
1921 	if (mrrs < 128)
1922 		dev_err(&dev->dev, "MRRS was unable to be configured with a safe value.  If problems are experienced, try running with pci=pcie_bus_safe\n");
1923 }
1924 
1925 static int pcie_bus_configure_set(struct pci_dev *dev, void *data)
1926 {
1927 	int mps, orig_mps;
1928 
1929 	if (!pci_is_pcie(dev))
1930 		return 0;
1931 
1932 	if (pcie_bus_config == PCIE_BUS_TUNE_OFF ||
1933 	    pcie_bus_config == PCIE_BUS_DEFAULT)
1934 		return 0;
1935 
1936 	mps = 128 << *(u8 *)data;
1937 	orig_mps = pcie_get_mps(dev);
1938 
1939 	pcie_write_mps(dev, mps);
1940 	pcie_write_mrrs(dev);
1941 
1942 	dev_info(&dev->dev, "Max Payload Size set to %4d/%4d (was %4d), Max Read Rq %4d\n",
1943 		 pcie_get_mps(dev), 128 << dev->pcie_mpss,
1944 		 orig_mps, pcie_get_readrq(dev));
1945 
1946 	return 0;
1947 }
1948 
1949 /* pcie_bus_configure_settings requires that pci_walk_bus work in a top-down,
1950  * parents then children fashion.  If this changes, then this code will not
1951  * work as designed.
1952  */
1953 void pcie_bus_configure_settings(struct pci_bus *bus)
1954 {
1955 	u8 smpss = 0;
1956 
1957 	if (!bus->self)
1958 		return;
1959 
1960 	if (!pci_is_pcie(bus->self))
1961 		return;
1962 
1963 	/* FIXME - Peer to peer DMA is possible, though the endpoint would need
1964 	 * to be aware of the MPS of the destination.  To work around this,
1965 	 * simply force the MPS of the entire system to the smallest possible.
1966 	 */
1967 	if (pcie_bus_config == PCIE_BUS_PEER2PEER)
1968 		smpss = 0;
1969 
1970 	if (pcie_bus_config == PCIE_BUS_SAFE) {
1971 		smpss = bus->self->pcie_mpss;
1972 
1973 		pcie_find_smpss(bus->self, &smpss);
1974 		pci_walk_bus(bus, pcie_find_smpss, &smpss);
1975 	}
1976 
1977 	pcie_bus_configure_set(bus->self, &smpss);
1978 	pci_walk_bus(bus, pcie_bus_configure_set, &smpss);
1979 }
1980 EXPORT_SYMBOL_GPL(pcie_bus_configure_settings);
1981 
1982 unsigned int pci_scan_child_bus(struct pci_bus *bus)
1983 {
1984 	unsigned int devfn, pass, max = bus->busn_res.start;
1985 	struct pci_dev *dev;
1986 
1987 	dev_dbg(&bus->dev, "scanning bus\n");
1988 
1989 	/* Go find them, Rover! */
1990 	for (devfn = 0; devfn < 0x100; devfn += 8)
1991 		pci_scan_slot(bus, devfn);
1992 
1993 	/* Reserve buses for SR-IOV capability. */
1994 	max += pci_iov_bus_range(bus);
1995 
1996 	/*
1997 	 * After performing arch-dependent fixup of the bus, look behind
1998 	 * all PCI-to-PCI bridges on this bus.
1999 	 */
2000 	if (!bus->is_added) {
2001 		dev_dbg(&bus->dev, "fixups for bus\n");
2002 		pcibios_fixup_bus(bus);
2003 		bus->is_added = 1;
2004 	}
2005 
2006 	for (pass = 0; pass < 2; pass++)
2007 		list_for_each_entry(dev, &bus->devices, bus_list) {
2008 			if (pci_is_bridge(dev))
2009 				max = pci_scan_bridge(bus, dev, max, pass);
2010 		}
2011 
2012 	/*
2013 	 * We've scanned the bus and so we know all about what's on
2014 	 * the other side of any bridges that may be on this bus plus
2015 	 * any devices.
2016 	 *
2017 	 * Return how far we've got finding sub-buses.
2018 	 */
2019 	dev_dbg(&bus->dev, "bus scan returning with max=%02x\n", max);
2020 	return max;
2021 }
2022 EXPORT_SYMBOL_GPL(pci_scan_child_bus);
2023 
2024 /**
2025  * pcibios_root_bridge_prepare - Platform-specific host bridge setup.
2026  * @bridge: Host bridge to set up.
2027  *
2028  * Default empty implementation.  Replace with an architecture-specific setup
2029  * routine, if necessary.
2030  */
2031 int __weak pcibios_root_bridge_prepare(struct pci_host_bridge *bridge)
2032 {
2033 	return 0;
2034 }
2035 
2036 void __weak pcibios_add_bus(struct pci_bus *bus)
2037 {
2038 }
2039 
2040 void __weak pcibios_remove_bus(struct pci_bus *bus)
2041 {
2042 }
2043 
2044 struct pci_bus *pci_create_root_bus(struct device *parent, int bus,
2045 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
2046 {
2047 	int error;
2048 	struct pci_host_bridge *bridge;
2049 	struct pci_bus *b, *b2;
2050 	struct resource_entry *window, *n;
2051 	struct resource *res;
2052 	resource_size_t offset;
2053 	char bus_addr[64];
2054 	char *fmt;
2055 
2056 	b = pci_alloc_bus(NULL);
2057 	if (!b)
2058 		return NULL;
2059 
2060 	b->sysdata = sysdata;
2061 	b->ops = ops;
2062 	b->number = b->busn_res.start = bus;
2063 	pci_bus_assign_domain_nr(b, parent);
2064 	b2 = pci_find_bus(pci_domain_nr(b), bus);
2065 	if (b2) {
2066 		/* If we already got to this bus through a different bridge, ignore it */
2067 		dev_dbg(&b2->dev, "bus already known\n");
2068 		goto err_out;
2069 	}
2070 
2071 	bridge = pci_alloc_host_bridge(b);
2072 	if (!bridge)
2073 		goto err_out;
2074 
2075 	bridge->dev.parent = parent;
2076 	bridge->dev.release = pci_release_host_bridge_dev;
2077 	dev_set_name(&bridge->dev, "pci%04x:%02x", pci_domain_nr(b), bus);
2078 	error = pcibios_root_bridge_prepare(bridge);
2079 	if (error) {
2080 		kfree(bridge);
2081 		goto err_out;
2082 	}
2083 
2084 	error = device_register(&bridge->dev);
2085 	if (error) {
2086 		put_device(&bridge->dev);
2087 		goto err_out;
2088 	}
2089 	b->bridge = get_device(&bridge->dev);
2090 	device_enable_async_suspend(b->bridge);
2091 	pci_set_bus_of_node(b);
2092 	pci_set_bus_msi_domain(b);
2093 
2094 	if (!parent)
2095 		set_dev_node(b->bridge, pcibus_to_node(b));
2096 
2097 	b->dev.class = &pcibus_class;
2098 	b->dev.parent = b->bridge;
2099 	dev_set_name(&b->dev, "%04x:%02x", pci_domain_nr(b), bus);
2100 	error = device_register(&b->dev);
2101 	if (error)
2102 		goto class_dev_reg_err;
2103 
2104 	pcibios_add_bus(b);
2105 
2106 	/* Create legacy_io and legacy_mem files for this bus */
2107 	pci_create_legacy_files(b);
2108 
2109 	if (parent)
2110 		dev_info(parent, "PCI host bridge to bus %s\n", dev_name(&b->dev));
2111 	else
2112 		printk(KERN_INFO "PCI host bridge to bus %s\n", dev_name(&b->dev));
2113 
2114 	/* Add initial resources to the bus */
2115 	resource_list_for_each_entry_safe(window, n, resources) {
2116 		list_move_tail(&window->node, &bridge->windows);
2117 		res = window->res;
2118 		offset = window->offset;
2119 		if (res->flags & IORESOURCE_BUS)
2120 			pci_bus_insert_busn_res(b, bus, res->end);
2121 		else
2122 			pci_bus_add_resource(b, res, 0);
2123 		if (offset) {
2124 			if (resource_type(res) == IORESOURCE_IO)
2125 				fmt = " (bus address [%#06llx-%#06llx])";
2126 			else
2127 				fmt = " (bus address [%#010llx-%#010llx])";
2128 			snprintf(bus_addr, sizeof(bus_addr), fmt,
2129 				 (unsigned long long) (res->start - offset),
2130 				 (unsigned long long) (res->end - offset));
2131 		} else
2132 			bus_addr[0] = '\0';
2133 		dev_info(&b->dev, "root bus resource %pR%s\n", res, bus_addr);
2134 	}
2135 
2136 	down_write(&pci_bus_sem);
2137 	list_add_tail(&b->node, &pci_root_buses);
2138 	up_write(&pci_bus_sem);
2139 
2140 	return b;
2141 
2142 class_dev_reg_err:
2143 	put_device(&bridge->dev);
2144 	device_unregister(&bridge->dev);
2145 err_out:
2146 	kfree(b);
2147 	return NULL;
2148 }
2149 EXPORT_SYMBOL_GPL(pci_create_root_bus);
2150 
2151 int pci_bus_insert_busn_res(struct pci_bus *b, int bus, int bus_max)
2152 {
2153 	struct resource *res = &b->busn_res;
2154 	struct resource *parent_res, *conflict;
2155 
2156 	res->start = bus;
2157 	res->end = bus_max;
2158 	res->flags = IORESOURCE_BUS;
2159 
2160 	if (!pci_is_root_bus(b))
2161 		parent_res = &b->parent->busn_res;
2162 	else {
2163 		parent_res = get_pci_domain_busn_res(pci_domain_nr(b));
2164 		res->flags |= IORESOURCE_PCI_FIXED;
2165 	}
2166 
2167 	conflict = request_resource_conflict(parent_res, res);
2168 
2169 	if (conflict)
2170 		dev_printk(KERN_DEBUG, &b->dev,
2171 			   "busn_res: can not insert %pR under %s%pR (conflicts with %s %pR)\n",
2172 			    res, pci_is_root_bus(b) ? "domain " : "",
2173 			    parent_res, conflict->name, conflict);
2174 
2175 	return conflict == NULL;
2176 }
2177 
2178 int pci_bus_update_busn_res_end(struct pci_bus *b, int bus_max)
2179 {
2180 	struct resource *res = &b->busn_res;
2181 	struct resource old_res = *res;
2182 	resource_size_t size;
2183 	int ret;
2184 
2185 	if (res->start > bus_max)
2186 		return -EINVAL;
2187 
2188 	size = bus_max - res->start + 1;
2189 	ret = adjust_resource(res, res->start, size);
2190 	dev_printk(KERN_DEBUG, &b->dev,
2191 			"busn_res: %pR end %s updated to %02x\n",
2192 			&old_res, ret ? "can not be" : "is", bus_max);
2193 
2194 	if (!ret && !res->parent)
2195 		pci_bus_insert_busn_res(b, res->start, res->end);
2196 
2197 	return ret;
2198 }
2199 
2200 void pci_bus_release_busn_res(struct pci_bus *b)
2201 {
2202 	struct resource *res = &b->busn_res;
2203 	int ret;
2204 
2205 	if (!res->flags || !res->parent)
2206 		return;
2207 
2208 	ret = release_resource(res);
2209 	dev_printk(KERN_DEBUG, &b->dev,
2210 			"busn_res: %pR %s released\n",
2211 			res, ret ? "can not be" : "is");
2212 }
2213 
2214 struct pci_bus *pci_scan_root_bus_msi(struct device *parent, int bus,
2215 		struct pci_ops *ops, void *sysdata,
2216 		struct list_head *resources, struct msi_controller *msi)
2217 {
2218 	struct resource_entry *window;
2219 	bool found = false;
2220 	struct pci_bus *b;
2221 	int max;
2222 
2223 	resource_list_for_each_entry(window, resources)
2224 		if (window->res->flags & IORESOURCE_BUS) {
2225 			found = true;
2226 			break;
2227 		}
2228 
2229 	b = pci_create_root_bus(parent, bus, ops, sysdata, resources);
2230 	if (!b)
2231 		return NULL;
2232 
2233 	b->msi = msi;
2234 
2235 	if (!found) {
2236 		dev_info(&b->dev,
2237 		 "No busn resource found for root bus, will use [bus %02x-ff]\n",
2238 			bus);
2239 		pci_bus_insert_busn_res(b, bus, 255);
2240 	}
2241 
2242 	max = pci_scan_child_bus(b);
2243 
2244 	if (!found)
2245 		pci_bus_update_busn_res_end(b, max);
2246 
2247 	return b;
2248 }
2249 
2250 struct pci_bus *pci_scan_root_bus(struct device *parent, int bus,
2251 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
2252 {
2253 	return pci_scan_root_bus_msi(parent, bus, ops, sysdata, resources,
2254 				     NULL);
2255 }
2256 EXPORT_SYMBOL(pci_scan_root_bus);
2257 
2258 struct pci_bus *pci_scan_bus(int bus, struct pci_ops *ops,
2259 					void *sysdata)
2260 {
2261 	LIST_HEAD(resources);
2262 	struct pci_bus *b;
2263 
2264 	pci_add_resource(&resources, &ioport_resource);
2265 	pci_add_resource(&resources, &iomem_resource);
2266 	pci_add_resource(&resources, &busn_resource);
2267 	b = pci_create_root_bus(NULL, bus, ops, sysdata, &resources);
2268 	if (b) {
2269 		pci_scan_child_bus(b);
2270 	} else {
2271 		pci_free_resource_list(&resources);
2272 	}
2273 	return b;
2274 }
2275 EXPORT_SYMBOL(pci_scan_bus);
2276 
2277 /**
2278  * pci_rescan_bus_bridge_resize - scan a PCI bus for devices.
2279  * @bridge: PCI bridge for the bus to scan
2280  *
2281  * Scan a PCI bus and child buses for new devices, add them,
2282  * and enable them, resizing bridge mmio/io resource if necessary
2283  * and possible.  The caller must ensure the child devices are already
2284  * removed for resizing to occur.
2285  *
2286  * Returns the max number of subordinate bus discovered.
2287  */
2288 unsigned int pci_rescan_bus_bridge_resize(struct pci_dev *bridge)
2289 {
2290 	unsigned int max;
2291 	struct pci_bus *bus = bridge->subordinate;
2292 
2293 	max = pci_scan_child_bus(bus);
2294 
2295 	pci_assign_unassigned_bridge_resources(bridge);
2296 
2297 	pci_bus_add_devices(bus);
2298 
2299 	return max;
2300 }
2301 
2302 /**
2303  * pci_rescan_bus - scan a PCI bus for devices.
2304  * @bus: PCI bus to scan
2305  *
2306  * Scan a PCI bus and child buses for new devices, adds them,
2307  * and enables them.
2308  *
2309  * Returns the max number of subordinate bus discovered.
2310  */
2311 unsigned int pci_rescan_bus(struct pci_bus *bus)
2312 {
2313 	unsigned int max;
2314 
2315 	max = pci_scan_child_bus(bus);
2316 	pci_assign_unassigned_bus_resources(bus);
2317 	pci_bus_add_devices(bus);
2318 
2319 	return max;
2320 }
2321 EXPORT_SYMBOL_GPL(pci_rescan_bus);
2322 
2323 /*
2324  * pci_rescan_bus(), pci_rescan_bus_bridge_resize() and PCI device removal
2325  * routines should always be executed under this mutex.
2326  */
2327 static DEFINE_MUTEX(pci_rescan_remove_lock);
2328 
2329 void pci_lock_rescan_remove(void)
2330 {
2331 	mutex_lock(&pci_rescan_remove_lock);
2332 }
2333 EXPORT_SYMBOL_GPL(pci_lock_rescan_remove);
2334 
2335 void pci_unlock_rescan_remove(void)
2336 {
2337 	mutex_unlock(&pci_rescan_remove_lock);
2338 }
2339 EXPORT_SYMBOL_GPL(pci_unlock_rescan_remove);
2340 
2341 static int __init pci_sort_bf_cmp(const struct device *d_a,
2342 				  const struct device *d_b)
2343 {
2344 	const struct pci_dev *a = to_pci_dev(d_a);
2345 	const struct pci_dev *b = to_pci_dev(d_b);
2346 
2347 	if      (pci_domain_nr(a->bus) < pci_domain_nr(b->bus)) return -1;
2348 	else if (pci_domain_nr(a->bus) > pci_domain_nr(b->bus)) return  1;
2349 
2350 	if      (a->bus->number < b->bus->number) return -1;
2351 	else if (a->bus->number > b->bus->number) return  1;
2352 
2353 	if      (a->devfn < b->devfn) return -1;
2354 	else if (a->devfn > b->devfn) return  1;
2355 
2356 	return 0;
2357 }
2358 
2359 void __init pci_sort_breadthfirst(void)
2360 {
2361 	bus_sort_breadthfirst(&pci_bus_type, &pci_sort_bf_cmp);
2362 }
2363