xref: /openbmc/linux/drivers/pci/probe.c (revision 6774def6)
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/pci_hotplug.h>
10 #include <linux/slab.h>
11 #include <linux/module.h>
12 #include <linux/cpumask.h>
13 #include <linux/pci-aspm.h>
14 #include <asm-generic/pci-bridge.h>
15 #include "pci.h"
16 
17 #define CARDBUS_LATENCY_TIMER	176	/* secondary latency timer */
18 #define CARDBUS_RESERVE_BUSNR	3
19 
20 static struct resource busn_resource = {
21 	.name	= "PCI busn",
22 	.start	= 0,
23 	.end	= 255,
24 	.flags	= IORESOURCE_BUS,
25 };
26 
27 /* Ugh.  Need to stop exporting this to modules. */
28 LIST_HEAD(pci_root_buses);
29 EXPORT_SYMBOL(pci_root_buses);
30 
31 static LIST_HEAD(pci_domain_busn_res_list);
32 
33 struct pci_domain_busn_res {
34 	struct list_head list;
35 	struct resource res;
36 	int domain_nr;
37 };
38 
39 static struct resource *get_pci_domain_busn_res(int domain_nr)
40 {
41 	struct pci_domain_busn_res *r;
42 
43 	list_for_each_entry(r, &pci_domain_busn_res_list, list)
44 		if (r->domain_nr == domain_nr)
45 			return &r->res;
46 
47 	r = kzalloc(sizeof(*r), GFP_KERNEL);
48 	if (!r)
49 		return NULL;
50 
51 	r->domain_nr = domain_nr;
52 	r->res.start = 0;
53 	r->res.end = 0xff;
54 	r->res.flags = IORESOURCE_BUS | IORESOURCE_PCI_FIXED;
55 
56 	list_add_tail(&r->list, &pci_domain_busn_res_list);
57 
58 	return &r->res;
59 }
60 
61 static int find_anything(struct device *dev, void *data)
62 {
63 	return 1;
64 }
65 
66 /*
67  * Some device drivers need know if pci is initiated.
68  * Basically, we think pci is not initiated when there
69  * is no device to be found on the pci_bus_type.
70  */
71 int no_pci_devices(void)
72 {
73 	struct device *dev;
74 	int no_devices;
75 
76 	dev = bus_find_device(&pci_bus_type, NULL, NULL, find_anything);
77 	no_devices = (dev == NULL);
78 	put_device(dev);
79 	return no_devices;
80 }
81 EXPORT_SYMBOL(no_pci_devices);
82 
83 /*
84  * PCI Bus Class
85  */
86 static void release_pcibus_dev(struct device *dev)
87 {
88 	struct pci_bus *pci_bus = to_pci_bus(dev);
89 
90 	if (pci_bus->bridge)
91 		put_device(pci_bus->bridge);
92 	pci_bus_remove_resources(pci_bus);
93 	pci_release_bus_of_node(pci_bus);
94 	kfree(pci_bus);
95 }
96 
97 static struct class pcibus_class = {
98 	.name		= "pci_bus",
99 	.dev_release	= &release_pcibus_dev,
100 	.dev_groups	= pcibus_groups,
101 };
102 
103 static int __init pcibus_class_init(void)
104 {
105 	return class_register(&pcibus_class);
106 }
107 postcore_initcall(pcibus_class_init);
108 
109 static u64 pci_size(u64 base, u64 maxbase, u64 mask)
110 {
111 	u64 size = mask & maxbase;	/* Find the significant bits */
112 	if (!size)
113 		return 0;
114 
115 	/* Get the lowest of them to find the decode size, and
116 	   from that the extent.  */
117 	size = (size & ~(size-1)) - 1;
118 
119 	/* base == maxbase can be valid only if the BAR has
120 	   already been programmed with all 1s.  */
121 	if (base == maxbase && ((base | size) & mask) != mask)
122 		return 0;
123 
124 	return size;
125 }
126 
127 static inline unsigned long decode_bar(struct pci_dev *dev, u32 bar)
128 {
129 	u32 mem_type;
130 	unsigned long flags;
131 
132 	if ((bar & PCI_BASE_ADDRESS_SPACE) == PCI_BASE_ADDRESS_SPACE_IO) {
133 		flags = bar & ~PCI_BASE_ADDRESS_IO_MASK;
134 		flags |= IORESOURCE_IO;
135 		return flags;
136 	}
137 
138 	flags = bar & ~PCI_BASE_ADDRESS_MEM_MASK;
139 	flags |= IORESOURCE_MEM;
140 	if (flags & PCI_BASE_ADDRESS_MEM_PREFETCH)
141 		flags |= IORESOURCE_PREFETCH;
142 
143 	mem_type = bar & PCI_BASE_ADDRESS_MEM_TYPE_MASK;
144 	switch (mem_type) {
145 	case PCI_BASE_ADDRESS_MEM_TYPE_32:
146 		break;
147 	case PCI_BASE_ADDRESS_MEM_TYPE_1M:
148 		/* 1M mem BAR treated as 32-bit BAR */
149 		break;
150 	case PCI_BASE_ADDRESS_MEM_TYPE_64:
151 		flags |= IORESOURCE_MEM_64;
152 		break;
153 	default:
154 		/* mem unknown type treated as 32-bit BAR */
155 		break;
156 	}
157 	return flags;
158 }
159 
160 #define PCI_COMMAND_DECODE_ENABLE	(PCI_COMMAND_MEMORY | PCI_COMMAND_IO)
161 
162 /**
163  * pci_read_base - read a PCI BAR
164  * @dev: the PCI device
165  * @type: type of the BAR
166  * @res: resource buffer to be filled in
167  * @pos: BAR position in the config space
168  *
169  * Returns 1 if the BAR is 64-bit, or 0 if 32-bit.
170  */
171 int __pci_read_base(struct pci_dev *dev, enum pci_bar_type type,
172 		    struct resource *res, unsigned int pos)
173 {
174 	u32 l, sz, mask;
175 	u64 l64, sz64, mask64;
176 	u16 orig_cmd;
177 	struct pci_bus_region region, inverted_region;
178 	bool bar_too_big = false, bar_too_high = false, bar_invalid = false;
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 || sz == 0xffffffff)
205 		goto fail;
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 			l &= PCI_BASE_ADDRESS_IO_MASK;
219 			mask = PCI_BASE_ADDRESS_IO_MASK & (u32) IO_SPACE_LIMIT;
220 		} else {
221 			l &= PCI_BASE_ADDRESS_MEM_MASK;
222 			mask = (u32)PCI_BASE_ADDRESS_MEM_MASK;
223 		}
224 	} else {
225 		res->flags |= (l & IORESOURCE_ROM_ENABLE);
226 		l &= PCI_ROM_ADDRESS_MASK;
227 		mask = (u32)PCI_ROM_ADDRESS_MASK;
228 	}
229 
230 	if (res->flags & IORESOURCE_MEM_64) {
231 		l64 = l;
232 		sz64 = sz;
233 		mask64 = mask | (u64)~0 << 32;
234 
235 		pci_read_config_dword(dev, pos + 4, &l);
236 		pci_write_config_dword(dev, pos + 4, ~0);
237 		pci_read_config_dword(dev, pos + 4, &sz);
238 		pci_write_config_dword(dev, pos + 4, l);
239 
240 		l64 |= ((u64)l << 32);
241 		sz64 |= ((u64)sz << 32);
242 
243 		sz64 = pci_size(l64, sz64, mask64);
244 
245 		if (!sz64)
246 			goto fail;
247 
248 		if ((sizeof(dma_addr_t) < 8 || sizeof(resource_size_t) < 8) &&
249 		    sz64 > 0x100000000ULL) {
250 			res->flags |= IORESOURCE_UNSET | IORESOURCE_DISABLED;
251 			res->start = 0;
252 			res->end = 0;
253 			bar_too_big = true;
254 			goto out;
255 		}
256 
257 		if ((sizeof(dma_addr_t) < 8) && l) {
258 			/* Above 32-bit boundary; try to reallocate */
259 			res->flags |= IORESOURCE_UNSET;
260 			res->start = 0;
261 			res->end = sz64;
262 			bar_too_high = true;
263 			goto out;
264 		} else {
265 			region.start = l64;
266 			region.end = l64 + sz64;
267 		}
268 	} else {
269 		sz = pci_size(l, sz, mask);
270 
271 		if (!sz)
272 			goto fail;
273 
274 		region.start = l;
275 		region.end = l + sz;
276 	}
277 
278 	pcibios_bus_to_resource(dev->bus, res, &region);
279 	pcibios_resource_to_bus(dev->bus, &inverted_region, res);
280 
281 	/*
282 	 * If "A" is a BAR value (a bus address), "bus_to_resource(A)" is
283 	 * the corresponding resource address (the physical address used by
284 	 * the CPU.  Converting that resource address back to a bus address
285 	 * should yield the original BAR value:
286 	 *
287 	 *     resource_to_bus(bus_to_resource(A)) == A
288 	 *
289 	 * If it doesn't, CPU accesses to "bus_to_resource(A)" will not
290 	 * be claimed by the device.
291 	 */
292 	if (inverted_region.start != region.start) {
293 		res->flags |= IORESOURCE_UNSET;
294 		res->start = 0;
295 		res->end = region.end - region.start;
296 		bar_invalid = true;
297 	}
298 
299 	goto out;
300 
301 
302 fail:
303 	res->flags = 0;
304 out:
305 	if (!dev->mmio_always_on &&
306 	    (orig_cmd & PCI_COMMAND_DECODE_ENABLE))
307 		pci_write_config_word(dev, PCI_COMMAND, orig_cmd);
308 
309 	if (bar_too_big)
310 		dev_err(&dev->dev, "reg 0x%x: can't handle BAR larger than 4GB (size %#010llx)\n",
311 			pos, (unsigned long long) sz64);
312 	if (bar_too_high)
313 		dev_info(&dev->dev, "reg 0x%x: can't handle BAR above 4G (bus address %#010llx)\n",
314 			 pos, (unsigned long long) l64);
315 	if (bar_invalid)
316 		dev_info(&dev->dev, "reg 0x%x: initial BAR value %#010llx invalid\n",
317 			 pos, (unsigned long long) region.start);
318 	if (res->flags)
319 		dev_printk(KERN_DEBUG, &dev->dev, "reg 0x%x: %pR\n", pos, res);
320 
321 	return (res->flags & IORESOURCE_MEM_64) ? 1 : 0;
322 }
323 
324 static void pci_read_bases(struct pci_dev *dev, unsigned int howmany, int rom)
325 {
326 	unsigned int pos, reg;
327 
328 	for (pos = 0; pos < howmany; pos++) {
329 		struct resource *res = &dev->resource[pos];
330 		reg = PCI_BASE_ADDRESS_0 + (pos << 2);
331 		pos += __pci_read_base(dev, pci_bar_unknown, res, reg);
332 	}
333 
334 	if (rom) {
335 		struct resource *res = &dev->resource[PCI_ROM_RESOURCE];
336 		dev->rom_base_reg = rom;
337 		res->flags = IORESOURCE_MEM | IORESOURCE_PREFETCH |
338 				IORESOURCE_READONLY | IORESOURCE_CACHEABLE |
339 				IORESOURCE_SIZEALIGN;
340 		__pci_read_base(dev, pci_bar_mem32, res, rom);
341 	}
342 }
343 
344 static void pci_read_bridge_io(struct pci_bus *child)
345 {
346 	struct pci_dev *dev = child->self;
347 	u8 io_base_lo, io_limit_lo;
348 	unsigned long io_mask, io_granularity, base, limit;
349 	struct pci_bus_region region;
350 	struct resource *res;
351 
352 	io_mask = PCI_IO_RANGE_MASK;
353 	io_granularity = 0x1000;
354 	if (dev->io_window_1k) {
355 		/* Support 1K I/O space granularity */
356 		io_mask = PCI_IO_1K_RANGE_MASK;
357 		io_granularity = 0x400;
358 	}
359 
360 	res = child->resource[0];
361 	pci_read_config_byte(dev, PCI_IO_BASE, &io_base_lo);
362 	pci_read_config_byte(dev, PCI_IO_LIMIT, &io_limit_lo);
363 	base = (io_base_lo & io_mask) << 8;
364 	limit = (io_limit_lo & io_mask) << 8;
365 
366 	if ((io_base_lo & PCI_IO_RANGE_TYPE_MASK) == PCI_IO_RANGE_TYPE_32) {
367 		u16 io_base_hi, io_limit_hi;
368 
369 		pci_read_config_word(dev, PCI_IO_BASE_UPPER16, &io_base_hi);
370 		pci_read_config_word(dev, PCI_IO_LIMIT_UPPER16, &io_limit_hi);
371 		base |= ((unsigned long) io_base_hi << 16);
372 		limit |= ((unsigned long) io_limit_hi << 16);
373 	}
374 
375 	if (base <= limit) {
376 		res->flags = (io_base_lo & PCI_IO_RANGE_TYPE_MASK) | IORESOURCE_IO;
377 		region.start = base;
378 		region.end = limit + io_granularity - 1;
379 		pcibios_bus_to_resource(dev->bus, res, &region);
380 		dev_printk(KERN_DEBUG, &dev->dev, "  bridge window %pR\n", res);
381 	}
382 }
383 
384 static void pci_read_bridge_mmio(struct pci_bus *child)
385 {
386 	struct pci_dev *dev = child->self;
387 	u16 mem_base_lo, mem_limit_lo;
388 	unsigned long base, limit;
389 	struct pci_bus_region region;
390 	struct resource *res;
391 
392 	res = child->resource[1];
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 		dev_printk(KERN_DEBUG, &dev->dev, "  bridge window %pR\n", res);
403 	}
404 }
405 
406 static void pci_read_bridge_mmio_pref(struct pci_bus *child)
407 {
408 	struct pci_dev *dev = child->self;
409 	u16 mem_base_lo, mem_limit_lo;
410 	unsigned long base, limit;
411 	struct pci_bus_region region;
412 	struct resource *res;
413 
414 	res = child->resource[2];
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 	base = ((unsigned long) mem_base_lo & PCI_PREF_RANGE_MASK) << 16;
418 	limit = ((unsigned long) 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 #if BITS_PER_LONG == 64
433 			base |= ((unsigned long) mem_base_hi) << 32;
434 			limit |= ((unsigned long) mem_limit_hi) << 32;
435 #else
436 			if (mem_base_hi || mem_limit_hi) {
437 				dev_err(&dev->dev, "can't handle 64-bit address space for bridge\n");
438 				return;
439 			}
440 #endif
441 		}
442 	}
443 	if (base <= limit) {
444 		res->flags = (mem_base_lo & PCI_PREF_RANGE_TYPE_MASK) |
445 					 IORESOURCE_MEM | IORESOURCE_PREFETCH;
446 		if (res->flags & PCI_PREF_RANGE_TYPE_64)
447 			res->flags |= IORESOURCE_MEM_64;
448 		region.start = base;
449 		region.end = limit + 0xfffff;
450 		pcibios_bus_to_resource(dev->bus, res, &region);
451 		dev_printk(KERN_DEBUG, &dev->dev, "  bridge window %pR\n", res);
452 	}
453 }
454 
455 void pci_read_bridge_bases(struct pci_bus *child)
456 {
457 	struct pci_dev *dev = child->self;
458 	struct resource *res;
459 	int i;
460 
461 	if (pci_is_root_bus(child))	/* It's a host bus, nothing to read */
462 		return;
463 
464 	dev_info(&dev->dev, "PCI bridge to %pR%s\n",
465 		 &child->busn_res,
466 		 dev->transparent ? " (subtractive decode)" : "");
467 
468 	pci_bus_remove_resources(child);
469 	for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++)
470 		child->resource[i] = &dev->resource[PCI_BRIDGE_RESOURCES+i];
471 
472 	pci_read_bridge_io(child);
473 	pci_read_bridge_mmio(child);
474 	pci_read_bridge_mmio_pref(child);
475 
476 	if (dev->transparent) {
477 		pci_bus_for_each_resource(child->parent, res, i) {
478 			if (res && res->flags) {
479 				pci_bus_add_resource(child, res,
480 						     PCI_SUBTRACTIVE_DECODE);
481 				dev_printk(KERN_DEBUG, &dev->dev,
482 					   "  bridge window %pR (subtractive decode)\n",
483 					   res);
484 			}
485 		}
486 	}
487 }
488 
489 static struct pci_bus *pci_alloc_bus(struct pci_bus *parent)
490 {
491 	struct pci_bus *b;
492 
493 	b = kzalloc(sizeof(*b), GFP_KERNEL);
494 	if (!b)
495 		return NULL;
496 
497 	INIT_LIST_HEAD(&b->node);
498 	INIT_LIST_HEAD(&b->children);
499 	INIT_LIST_HEAD(&b->devices);
500 	INIT_LIST_HEAD(&b->slots);
501 	INIT_LIST_HEAD(&b->resources);
502 	b->max_bus_speed = PCI_SPEED_UNKNOWN;
503 	b->cur_bus_speed = PCI_SPEED_UNKNOWN;
504 #ifdef CONFIG_PCI_DOMAINS_GENERIC
505 	if (parent)
506 		b->domain_nr = parent->domain_nr;
507 #endif
508 	return b;
509 }
510 
511 static void pci_release_host_bridge_dev(struct device *dev)
512 {
513 	struct pci_host_bridge *bridge = to_pci_host_bridge(dev);
514 
515 	if (bridge->release_fn)
516 		bridge->release_fn(bridge);
517 
518 	pci_free_resource_list(&bridge->windows);
519 
520 	kfree(bridge);
521 }
522 
523 static struct pci_host_bridge *pci_alloc_host_bridge(struct pci_bus *b)
524 {
525 	struct pci_host_bridge *bridge;
526 
527 	bridge = kzalloc(sizeof(*bridge), GFP_KERNEL);
528 	if (!bridge)
529 		return NULL;
530 
531 	INIT_LIST_HEAD(&bridge->windows);
532 	bridge->bus = b;
533 	return bridge;
534 }
535 
536 static const unsigned char pcix_bus_speed[] = {
537 	PCI_SPEED_UNKNOWN,		/* 0 */
538 	PCI_SPEED_66MHz_PCIX,		/* 1 */
539 	PCI_SPEED_100MHz_PCIX,		/* 2 */
540 	PCI_SPEED_133MHz_PCIX,		/* 3 */
541 	PCI_SPEED_UNKNOWN,		/* 4 */
542 	PCI_SPEED_66MHz_PCIX_ECC,	/* 5 */
543 	PCI_SPEED_100MHz_PCIX_ECC,	/* 6 */
544 	PCI_SPEED_133MHz_PCIX_ECC,	/* 7 */
545 	PCI_SPEED_UNKNOWN,		/* 8 */
546 	PCI_SPEED_66MHz_PCIX_266,	/* 9 */
547 	PCI_SPEED_100MHz_PCIX_266,	/* A */
548 	PCI_SPEED_133MHz_PCIX_266,	/* B */
549 	PCI_SPEED_UNKNOWN,		/* C */
550 	PCI_SPEED_66MHz_PCIX_533,	/* D */
551 	PCI_SPEED_100MHz_PCIX_533,	/* E */
552 	PCI_SPEED_133MHz_PCIX_533	/* F */
553 };
554 
555 const unsigned char pcie_link_speed[] = {
556 	PCI_SPEED_UNKNOWN,		/* 0 */
557 	PCIE_SPEED_2_5GT,		/* 1 */
558 	PCIE_SPEED_5_0GT,		/* 2 */
559 	PCIE_SPEED_8_0GT,		/* 3 */
560 	PCI_SPEED_UNKNOWN,		/* 4 */
561 	PCI_SPEED_UNKNOWN,		/* 5 */
562 	PCI_SPEED_UNKNOWN,		/* 6 */
563 	PCI_SPEED_UNKNOWN,		/* 7 */
564 	PCI_SPEED_UNKNOWN,		/* 8 */
565 	PCI_SPEED_UNKNOWN,		/* 9 */
566 	PCI_SPEED_UNKNOWN,		/* A */
567 	PCI_SPEED_UNKNOWN,		/* B */
568 	PCI_SPEED_UNKNOWN,		/* C */
569 	PCI_SPEED_UNKNOWN,		/* D */
570 	PCI_SPEED_UNKNOWN,		/* E */
571 	PCI_SPEED_UNKNOWN		/* F */
572 };
573 
574 void pcie_update_link_speed(struct pci_bus *bus, u16 linksta)
575 {
576 	bus->cur_bus_speed = pcie_link_speed[linksta & PCI_EXP_LNKSTA_CLS];
577 }
578 EXPORT_SYMBOL_GPL(pcie_update_link_speed);
579 
580 static unsigned char agp_speeds[] = {
581 	AGP_UNKNOWN,
582 	AGP_1X,
583 	AGP_2X,
584 	AGP_4X,
585 	AGP_8X
586 };
587 
588 static enum pci_bus_speed agp_speed(int agp3, int agpstat)
589 {
590 	int index = 0;
591 
592 	if (agpstat & 4)
593 		index = 3;
594 	else if (agpstat & 2)
595 		index = 2;
596 	else if (agpstat & 1)
597 		index = 1;
598 	else
599 		goto out;
600 
601 	if (agp3) {
602 		index += 2;
603 		if (index == 5)
604 			index = 0;
605 	}
606 
607  out:
608 	return agp_speeds[index];
609 }
610 
611 static void pci_set_bus_speed(struct pci_bus *bus)
612 {
613 	struct pci_dev *bridge = bus->self;
614 	int pos;
615 
616 	pos = pci_find_capability(bridge, PCI_CAP_ID_AGP);
617 	if (!pos)
618 		pos = pci_find_capability(bridge, PCI_CAP_ID_AGP3);
619 	if (pos) {
620 		u32 agpstat, agpcmd;
621 
622 		pci_read_config_dword(bridge, pos + PCI_AGP_STATUS, &agpstat);
623 		bus->max_bus_speed = agp_speed(agpstat & 8, agpstat & 7);
624 
625 		pci_read_config_dword(bridge, pos + PCI_AGP_COMMAND, &agpcmd);
626 		bus->cur_bus_speed = agp_speed(agpstat & 8, agpcmd & 7);
627 	}
628 
629 	pos = pci_find_capability(bridge, PCI_CAP_ID_PCIX);
630 	if (pos) {
631 		u16 status;
632 		enum pci_bus_speed max;
633 
634 		pci_read_config_word(bridge, pos + PCI_X_BRIDGE_SSTATUS,
635 				     &status);
636 
637 		if (status & PCI_X_SSTATUS_533MHZ) {
638 			max = PCI_SPEED_133MHz_PCIX_533;
639 		} else if (status & PCI_X_SSTATUS_266MHZ) {
640 			max = PCI_SPEED_133MHz_PCIX_266;
641 		} else if (status & PCI_X_SSTATUS_133MHZ) {
642 			if ((status & PCI_X_SSTATUS_VERS) == PCI_X_SSTATUS_V2)
643 				max = PCI_SPEED_133MHz_PCIX_ECC;
644 			else
645 				max = PCI_SPEED_133MHz_PCIX;
646 		} else {
647 			max = PCI_SPEED_66MHz_PCIX;
648 		}
649 
650 		bus->max_bus_speed = max;
651 		bus->cur_bus_speed = pcix_bus_speed[
652 			(status & PCI_X_SSTATUS_FREQ) >> 6];
653 
654 		return;
655 	}
656 
657 	if (pci_is_pcie(bridge)) {
658 		u32 linkcap;
659 		u16 linksta;
660 
661 		pcie_capability_read_dword(bridge, PCI_EXP_LNKCAP, &linkcap);
662 		bus->max_bus_speed = pcie_link_speed[linkcap & PCI_EXP_LNKCAP_SLS];
663 
664 		pcie_capability_read_word(bridge, PCI_EXP_LNKSTA, &linksta);
665 		pcie_update_link_speed(bus, linksta);
666 	}
667 }
668 
669 static struct pci_bus *pci_alloc_child_bus(struct pci_bus *parent,
670 					   struct pci_dev *bridge, int busnr)
671 {
672 	struct pci_bus *child;
673 	int i;
674 	int ret;
675 
676 	/*
677 	 * Allocate a new bus, and inherit stuff from the parent..
678 	 */
679 	child = pci_alloc_bus(parent);
680 	if (!child)
681 		return NULL;
682 
683 	child->parent = parent;
684 	child->ops = parent->ops;
685 	child->msi = parent->msi;
686 	child->sysdata = parent->sysdata;
687 	child->bus_flags = parent->bus_flags;
688 
689 	/* initialize some portions of the bus device, but don't register it
690 	 * now as the parent is not properly set up yet.
691 	 */
692 	child->dev.class = &pcibus_class;
693 	dev_set_name(&child->dev, "%04x:%02x", pci_domain_nr(child), busnr);
694 
695 	/*
696 	 * Set up the primary, secondary and subordinate
697 	 * bus numbers.
698 	 */
699 	child->number = child->busn_res.start = busnr;
700 	child->primary = parent->busn_res.start;
701 	child->busn_res.end = 0xff;
702 
703 	if (!bridge) {
704 		child->dev.parent = parent->bridge;
705 		goto add_dev;
706 	}
707 
708 	child->self = bridge;
709 	child->bridge = get_device(&bridge->dev);
710 	child->dev.parent = child->bridge;
711 	pci_set_bus_of_node(child);
712 	pci_set_bus_speed(child);
713 
714 	/* Set up default resource pointers and names.. */
715 	for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++) {
716 		child->resource[i] = &bridge->resource[PCI_BRIDGE_RESOURCES+i];
717 		child->resource[i]->name = child->name;
718 	}
719 	bridge->subordinate = child;
720 
721 add_dev:
722 	ret = device_register(&child->dev);
723 	WARN_ON(ret < 0);
724 
725 	pcibios_add_bus(child);
726 
727 	/* Create legacy_io and legacy_mem files for this bus */
728 	pci_create_legacy_files(child);
729 
730 	return child;
731 }
732 
733 struct pci_bus *pci_add_new_bus(struct pci_bus *parent, struct pci_dev *dev,
734 				int busnr)
735 {
736 	struct pci_bus *child;
737 
738 	child = pci_alloc_child_bus(parent, dev, busnr);
739 	if (child) {
740 		down_write(&pci_bus_sem);
741 		list_add_tail(&child->node, &parent->children);
742 		up_write(&pci_bus_sem);
743 	}
744 	return child;
745 }
746 EXPORT_SYMBOL(pci_add_new_bus);
747 
748 static void pci_enable_crs(struct pci_dev *pdev)
749 {
750 	u16 root_cap = 0;
751 
752 	/* Enable CRS Software Visibility if supported */
753 	pcie_capability_read_word(pdev, PCI_EXP_RTCAP, &root_cap);
754 	if (root_cap & PCI_EXP_RTCAP_CRSVIS)
755 		pcie_capability_set_word(pdev, PCI_EXP_RTCTL,
756 					 PCI_EXP_RTCTL_CRSSVE);
757 }
758 
759 /*
760  * If it's a bridge, configure it and scan the bus behind it.
761  * For CardBus bridges, we don't scan behind as the devices will
762  * be handled by the bridge driver itself.
763  *
764  * We need to process bridges in two passes -- first we scan those
765  * already configured by the BIOS and after we are done with all of
766  * them, we proceed to assigning numbers to the remaining buses in
767  * order to avoid overlaps between old and new bus numbers.
768  */
769 int pci_scan_bridge(struct pci_bus *bus, struct pci_dev *dev, int max, int pass)
770 {
771 	struct pci_bus *child;
772 	int is_cardbus = (dev->hdr_type == PCI_HEADER_TYPE_CARDBUS);
773 	u32 buses, i, j = 0;
774 	u16 bctl;
775 	u8 primary, secondary, subordinate;
776 	int broken = 0;
777 
778 	pci_read_config_dword(dev, PCI_PRIMARY_BUS, &buses);
779 	primary = buses & 0xFF;
780 	secondary = (buses >> 8) & 0xFF;
781 	subordinate = (buses >> 16) & 0xFF;
782 
783 	dev_dbg(&dev->dev, "scanning [bus %02x-%02x] behind bridge, pass %d\n",
784 		secondary, subordinate, pass);
785 
786 	if (!primary && (primary != bus->number) && secondary && subordinate) {
787 		dev_warn(&dev->dev, "Primary bus is hard wired to 0\n");
788 		primary = bus->number;
789 	}
790 
791 	/* Check if setup is sensible at all */
792 	if (!pass &&
793 	    (primary != bus->number || secondary <= bus->number ||
794 	     secondary > subordinate)) {
795 		dev_info(&dev->dev, "bridge configuration invalid ([bus %02x-%02x]), reconfiguring\n",
796 			 secondary, subordinate);
797 		broken = 1;
798 	}
799 
800 	/* Disable MasterAbortMode during probing to avoid reporting
801 	   of bus errors (in some architectures) */
802 	pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &bctl);
803 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL,
804 			      bctl & ~PCI_BRIDGE_CTL_MASTER_ABORT);
805 
806 	pci_enable_crs(dev);
807 
808 	if ((secondary || subordinate) && !pcibios_assign_all_busses() &&
809 	    !is_cardbus && !broken) {
810 		unsigned int cmax;
811 		/*
812 		 * Bus already configured by firmware, process it in the first
813 		 * pass and just note the configuration.
814 		 */
815 		if (pass)
816 			goto out;
817 
818 		/*
819 		 * The bus might already exist for two reasons: Either we are
820 		 * rescanning the bus or the bus is reachable through more than
821 		 * one bridge. The second case can happen with the i450NX
822 		 * chipset.
823 		 */
824 		child = pci_find_bus(pci_domain_nr(bus), secondary);
825 		if (!child) {
826 			child = pci_add_new_bus(bus, dev, secondary);
827 			if (!child)
828 				goto out;
829 			child->primary = primary;
830 			pci_bus_insert_busn_res(child, secondary, subordinate);
831 			child->bridge_ctl = bctl;
832 		}
833 
834 		cmax = pci_scan_child_bus(child);
835 		if (cmax > subordinate)
836 			dev_warn(&dev->dev, "bridge has subordinate %02x but max busn %02x\n",
837 				 subordinate, cmax);
838 		/* subordinate should equal child->busn_res.end */
839 		if (subordinate > max)
840 			max = subordinate;
841 	} else {
842 		/*
843 		 * We need to assign a number to this bus which we always
844 		 * do in the second pass.
845 		 */
846 		if (!pass) {
847 			if (pcibios_assign_all_busses() || broken || is_cardbus)
848 				/* Temporarily disable forwarding of the
849 				   configuration cycles on all bridges in
850 				   this bus segment to avoid possible
851 				   conflicts in the second pass between two
852 				   bridges programmed with overlapping
853 				   bus ranges. */
854 				pci_write_config_dword(dev, PCI_PRIMARY_BUS,
855 						       buses & ~0xffffff);
856 			goto out;
857 		}
858 
859 		/* Clear errors */
860 		pci_write_config_word(dev, PCI_STATUS, 0xffff);
861 
862 		/* Prevent assigning a bus number that already exists.
863 		 * This can happen when a bridge is hot-plugged, so in
864 		 * this case we only re-scan this bus. */
865 		child = pci_find_bus(pci_domain_nr(bus), max+1);
866 		if (!child) {
867 			child = pci_add_new_bus(bus, dev, max+1);
868 			if (!child)
869 				goto out;
870 			pci_bus_insert_busn_res(child, max+1, 0xff);
871 		}
872 		max++;
873 		buses = (buses & 0xff000000)
874 		      | ((unsigned int)(child->primary)     <<  0)
875 		      | ((unsigned int)(child->busn_res.start)   <<  8)
876 		      | ((unsigned int)(child->busn_res.end) << 16);
877 
878 		/*
879 		 * yenta.c forces a secondary latency timer of 176.
880 		 * Copy that behaviour here.
881 		 */
882 		if (is_cardbus) {
883 			buses &= ~0xff000000;
884 			buses |= CARDBUS_LATENCY_TIMER << 24;
885 		}
886 
887 		/*
888 		 * We need to blast all three values with a single write.
889 		 */
890 		pci_write_config_dword(dev, PCI_PRIMARY_BUS, buses);
891 
892 		if (!is_cardbus) {
893 			child->bridge_ctl = bctl;
894 			max = pci_scan_child_bus(child);
895 		} else {
896 			/*
897 			 * For CardBus bridges, we leave 4 bus numbers
898 			 * as cards with a PCI-to-PCI bridge can be
899 			 * inserted later.
900 			 */
901 			for (i = 0; i < CARDBUS_RESERVE_BUSNR; i++) {
902 				struct pci_bus *parent = bus;
903 				if (pci_find_bus(pci_domain_nr(bus),
904 							max+i+1))
905 					break;
906 				while (parent->parent) {
907 					if ((!pcibios_assign_all_busses()) &&
908 					    (parent->busn_res.end > max) &&
909 					    (parent->busn_res.end <= max+i)) {
910 						j = 1;
911 					}
912 					parent = parent->parent;
913 				}
914 				if (j) {
915 					/*
916 					 * Often, there are two cardbus bridges
917 					 * -- try to leave one valid bus number
918 					 * for each one.
919 					 */
920 					i /= 2;
921 					break;
922 				}
923 			}
924 			max += i;
925 		}
926 		/*
927 		 * Set the subordinate bus number to its real value.
928 		 */
929 		pci_bus_update_busn_res_end(child, max);
930 		pci_write_config_byte(dev, PCI_SUBORDINATE_BUS, max);
931 	}
932 
933 	sprintf(child->name,
934 		(is_cardbus ? "PCI CardBus %04x:%02x" : "PCI Bus %04x:%02x"),
935 		pci_domain_nr(bus), child->number);
936 
937 	/* Has only triggered on CardBus, fixup is in yenta_socket */
938 	while (bus->parent) {
939 		if ((child->busn_res.end > bus->busn_res.end) ||
940 		    (child->number > bus->busn_res.end) ||
941 		    (child->number < bus->number) ||
942 		    (child->busn_res.end < bus->number)) {
943 			dev_info(&child->dev, "%pR %s hidden behind%s bridge %s %pR\n",
944 				&child->busn_res,
945 				(bus->number > child->busn_res.end &&
946 				 bus->busn_res.end < child->number) ?
947 					"wholly" : "partially",
948 				bus->self->transparent ? " transparent" : "",
949 				dev_name(&bus->dev),
950 				&bus->busn_res);
951 		}
952 		bus = bus->parent;
953 	}
954 
955 out:
956 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL, bctl);
957 
958 	return max;
959 }
960 EXPORT_SYMBOL(pci_scan_bridge);
961 
962 /*
963  * Read interrupt line and base address registers.
964  * The architecture-dependent code can tweak these, of course.
965  */
966 static void pci_read_irq(struct pci_dev *dev)
967 {
968 	unsigned char irq;
969 
970 	pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &irq);
971 	dev->pin = irq;
972 	if (irq)
973 		pci_read_config_byte(dev, PCI_INTERRUPT_LINE, &irq);
974 	dev->irq = irq;
975 }
976 
977 void set_pcie_port_type(struct pci_dev *pdev)
978 {
979 	int pos;
980 	u16 reg16;
981 
982 	pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
983 	if (!pos)
984 		return;
985 	pdev->pcie_cap = pos;
986 	pci_read_config_word(pdev, pos + PCI_EXP_FLAGS, &reg16);
987 	pdev->pcie_flags_reg = reg16;
988 	pci_read_config_word(pdev, pos + PCI_EXP_DEVCAP, &reg16);
989 	pdev->pcie_mpss = reg16 & PCI_EXP_DEVCAP_PAYLOAD;
990 }
991 
992 void set_pcie_hotplug_bridge(struct pci_dev *pdev)
993 {
994 	u32 reg32;
995 
996 	pcie_capability_read_dword(pdev, PCI_EXP_SLTCAP, &reg32);
997 	if (reg32 & PCI_EXP_SLTCAP_HPC)
998 		pdev->is_hotplug_bridge = 1;
999 }
1000 
1001 /**
1002  * pci_ext_cfg_is_aliased - is ext config space just an alias of std config?
1003  * @dev: PCI device
1004  *
1005  * PCI Express to PCI/PCI-X Bridge Specification, rev 1.0, 4.1.4 says that
1006  * when forwarding a type1 configuration request the bridge must check that
1007  * the extended register address field is zero.  The bridge is not permitted
1008  * to forward the transactions and must handle it as an Unsupported Request.
1009  * Some bridges do not follow this rule and simply drop the extended register
1010  * bits, resulting in the standard config space being aliased, every 256
1011  * bytes across the entire configuration space.  Test for this condition by
1012  * comparing the first dword of each potential alias to the vendor/device ID.
1013  * Known offenders:
1014  *   ASM1083/1085 PCIe-to-PCI Reversible Bridge (1b21:1080, rev 01 & 03)
1015  *   AMD/ATI SBx00 PCI to PCI Bridge (1002:4384, rev 40)
1016  */
1017 static bool pci_ext_cfg_is_aliased(struct pci_dev *dev)
1018 {
1019 #ifdef CONFIG_PCI_QUIRKS
1020 	int pos;
1021 	u32 header, tmp;
1022 
1023 	pci_read_config_dword(dev, PCI_VENDOR_ID, &header);
1024 
1025 	for (pos = PCI_CFG_SPACE_SIZE;
1026 	     pos < PCI_CFG_SPACE_EXP_SIZE; pos += PCI_CFG_SPACE_SIZE) {
1027 		if (pci_read_config_dword(dev, pos, &tmp) != PCIBIOS_SUCCESSFUL
1028 		    || header != tmp)
1029 			return false;
1030 	}
1031 
1032 	return true;
1033 #else
1034 	return false;
1035 #endif
1036 }
1037 
1038 /**
1039  * pci_cfg_space_size - get the configuration space size of the PCI device.
1040  * @dev: PCI device
1041  *
1042  * Regular PCI devices have 256 bytes, but PCI-X 2 and PCI Express devices
1043  * have 4096 bytes.  Even if the device is capable, that doesn't mean we can
1044  * access it.  Maybe we don't have a way to generate extended config space
1045  * accesses, or the device is behind a reverse Express bridge.  So we try
1046  * reading the dword at 0x100 which must either be 0 or a valid extended
1047  * capability header.
1048  */
1049 static int pci_cfg_space_size_ext(struct pci_dev *dev)
1050 {
1051 	u32 status;
1052 	int pos = PCI_CFG_SPACE_SIZE;
1053 
1054 	if (pci_read_config_dword(dev, pos, &status) != PCIBIOS_SUCCESSFUL)
1055 		goto fail;
1056 	if (status == 0xffffffff || pci_ext_cfg_is_aliased(dev))
1057 		goto fail;
1058 
1059 	return PCI_CFG_SPACE_EXP_SIZE;
1060 
1061  fail:
1062 	return PCI_CFG_SPACE_SIZE;
1063 }
1064 
1065 int pci_cfg_space_size(struct pci_dev *dev)
1066 {
1067 	int pos;
1068 	u32 status;
1069 	u16 class;
1070 
1071 	class = dev->class >> 8;
1072 	if (class == PCI_CLASS_BRIDGE_HOST)
1073 		return pci_cfg_space_size_ext(dev);
1074 
1075 	if (!pci_is_pcie(dev)) {
1076 		pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1077 		if (!pos)
1078 			goto fail;
1079 
1080 		pci_read_config_dword(dev, pos + PCI_X_STATUS, &status);
1081 		if (!(status & (PCI_X_STATUS_266MHZ | PCI_X_STATUS_533MHZ)))
1082 			goto fail;
1083 	}
1084 
1085 	return pci_cfg_space_size_ext(dev);
1086 
1087  fail:
1088 	return PCI_CFG_SPACE_SIZE;
1089 }
1090 
1091 #define LEGACY_IO_RESOURCE	(IORESOURCE_IO | IORESOURCE_PCI_FIXED)
1092 
1093 /**
1094  * pci_setup_device - fill in class and map information of a device
1095  * @dev: the device structure to fill
1096  *
1097  * Initialize the device structure with information about the device's
1098  * vendor,class,memory and IO-space addresses,IRQ lines etc.
1099  * Called at initialisation of the PCI subsystem and by CardBus services.
1100  * Returns 0 on success and negative if unknown type of device (not normal,
1101  * bridge or CardBus).
1102  */
1103 int pci_setup_device(struct pci_dev *dev)
1104 {
1105 	u32 class;
1106 	u8 hdr_type;
1107 	struct pci_slot *slot;
1108 	int pos = 0;
1109 	struct pci_bus_region region;
1110 	struct resource *res;
1111 
1112 	if (pci_read_config_byte(dev, PCI_HEADER_TYPE, &hdr_type))
1113 		return -EIO;
1114 
1115 	dev->sysdata = dev->bus->sysdata;
1116 	dev->dev.parent = dev->bus->bridge;
1117 	dev->dev.bus = &pci_bus_type;
1118 	dev->hdr_type = hdr_type & 0x7f;
1119 	dev->multifunction = !!(hdr_type & 0x80);
1120 	dev->error_state = pci_channel_io_normal;
1121 	set_pcie_port_type(dev);
1122 
1123 	list_for_each_entry(slot, &dev->bus->slots, list)
1124 		if (PCI_SLOT(dev->devfn) == slot->number)
1125 			dev->slot = slot;
1126 
1127 	/* Assume 32-bit PCI; let 64-bit PCI cards (which are far rarer)
1128 	   set this higher, assuming the system even supports it.  */
1129 	dev->dma_mask = 0xffffffff;
1130 
1131 	dev_set_name(&dev->dev, "%04x:%02x:%02x.%d", pci_domain_nr(dev->bus),
1132 		     dev->bus->number, PCI_SLOT(dev->devfn),
1133 		     PCI_FUNC(dev->devfn));
1134 
1135 	pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
1136 	dev->revision = class & 0xff;
1137 	dev->class = class >> 8;		    /* upper 3 bytes */
1138 
1139 	dev_printk(KERN_DEBUG, &dev->dev, "[%04x:%04x] type %02x class %#08x\n",
1140 		   dev->vendor, dev->device, dev->hdr_type, dev->class);
1141 
1142 	/* need to have dev->class ready */
1143 	dev->cfg_size = pci_cfg_space_size(dev);
1144 
1145 	/* "Unknown power state" */
1146 	dev->current_state = PCI_UNKNOWN;
1147 
1148 	/* Early fixups, before probing the BARs */
1149 	pci_fixup_device(pci_fixup_early, dev);
1150 	/* device class may be changed after fixup */
1151 	class = dev->class >> 8;
1152 
1153 	switch (dev->hdr_type) {		    /* header type */
1154 	case PCI_HEADER_TYPE_NORMAL:		    /* standard header */
1155 		if (class == PCI_CLASS_BRIDGE_PCI)
1156 			goto bad;
1157 		pci_read_irq(dev);
1158 		pci_read_bases(dev, 6, PCI_ROM_ADDRESS);
1159 		pci_read_config_word(dev, PCI_SUBSYSTEM_VENDOR_ID, &dev->subsystem_vendor);
1160 		pci_read_config_word(dev, PCI_SUBSYSTEM_ID, &dev->subsystem_device);
1161 
1162 		/*
1163 		 * Do the ugly legacy mode stuff here rather than broken chip
1164 		 * quirk code. Legacy mode ATA controllers have fixed
1165 		 * addresses. These are not always echoed in BAR0-3, and
1166 		 * BAR0-3 in a few cases contain junk!
1167 		 */
1168 		if (class == PCI_CLASS_STORAGE_IDE) {
1169 			u8 progif;
1170 			pci_read_config_byte(dev, PCI_CLASS_PROG, &progif);
1171 			if ((progif & 1) == 0) {
1172 				region.start = 0x1F0;
1173 				region.end = 0x1F7;
1174 				res = &dev->resource[0];
1175 				res->flags = LEGACY_IO_RESOURCE;
1176 				pcibios_bus_to_resource(dev->bus, res, &region);
1177 				dev_info(&dev->dev, "legacy IDE quirk: reg 0x10: %pR\n",
1178 					 res);
1179 				region.start = 0x3F6;
1180 				region.end = 0x3F6;
1181 				res = &dev->resource[1];
1182 				res->flags = LEGACY_IO_RESOURCE;
1183 				pcibios_bus_to_resource(dev->bus, res, &region);
1184 				dev_info(&dev->dev, "legacy IDE quirk: reg 0x14: %pR\n",
1185 					 res);
1186 			}
1187 			if ((progif & 4) == 0) {
1188 				region.start = 0x170;
1189 				region.end = 0x177;
1190 				res = &dev->resource[2];
1191 				res->flags = LEGACY_IO_RESOURCE;
1192 				pcibios_bus_to_resource(dev->bus, res, &region);
1193 				dev_info(&dev->dev, "legacy IDE quirk: reg 0x18: %pR\n",
1194 					 res);
1195 				region.start = 0x376;
1196 				region.end = 0x376;
1197 				res = &dev->resource[3];
1198 				res->flags = LEGACY_IO_RESOURCE;
1199 				pcibios_bus_to_resource(dev->bus, res, &region);
1200 				dev_info(&dev->dev, "legacy IDE quirk: reg 0x1c: %pR\n",
1201 					 res);
1202 			}
1203 		}
1204 		break;
1205 
1206 	case PCI_HEADER_TYPE_BRIDGE:		    /* bridge header */
1207 		if (class != PCI_CLASS_BRIDGE_PCI)
1208 			goto bad;
1209 		/* The PCI-to-PCI bridge spec requires that subtractive
1210 		   decoding (i.e. transparent) bridge must have programming
1211 		   interface code of 0x01. */
1212 		pci_read_irq(dev);
1213 		dev->transparent = ((dev->class & 0xff) == 1);
1214 		pci_read_bases(dev, 2, PCI_ROM_ADDRESS1);
1215 		set_pcie_hotplug_bridge(dev);
1216 		pos = pci_find_capability(dev, PCI_CAP_ID_SSVID);
1217 		if (pos) {
1218 			pci_read_config_word(dev, pos + PCI_SSVID_VENDOR_ID, &dev->subsystem_vendor);
1219 			pci_read_config_word(dev, pos + PCI_SSVID_DEVICE_ID, &dev->subsystem_device);
1220 		}
1221 		break;
1222 
1223 	case PCI_HEADER_TYPE_CARDBUS:		    /* CardBus bridge header */
1224 		if (class != PCI_CLASS_BRIDGE_CARDBUS)
1225 			goto bad;
1226 		pci_read_irq(dev);
1227 		pci_read_bases(dev, 1, 0);
1228 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_VENDOR_ID, &dev->subsystem_vendor);
1229 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_ID, &dev->subsystem_device);
1230 		break;
1231 
1232 	default:				    /* unknown header */
1233 		dev_err(&dev->dev, "unknown header type %02x, ignoring device\n",
1234 			dev->hdr_type);
1235 		return -EIO;
1236 
1237 	bad:
1238 		dev_err(&dev->dev, "ignoring class %#08x (doesn't match header type %02x)\n",
1239 			dev->class, dev->hdr_type);
1240 		dev->class = PCI_CLASS_NOT_DEFINED;
1241 	}
1242 
1243 	/* We found a fine healthy device, go go go... */
1244 	return 0;
1245 }
1246 
1247 static struct hpp_type0 pci_default_type0 = {
1248 	.revision = 1,
1249 	.cache_line_size = 8,
1250 	.latency_timer = 0x40,
1251 	.enable_serr = 0,
1252 	.enable_perr = 0,
1253 };
1254 
1255 static void program_hpp_type0(struct pci_dev *dev, struct hpp_type0 *hpp)
1256 {
1257 	u16 pci_cmd, pci_bctl;
1258 
1259 	if (!hpp)
1260 		hpp = &pci_default_type0;
1261 
1262 	if (hpp->revision > 1) {
1263 		dev_warn(&dev->dev,
1264 			 "PCI settings rev %d not supported; using defaults\n",
1265 			 hpp->revision);
1266 		hpp = &pci_default_type0;
1267 	}
1268 
1269 	pci_write_config_byte(dev, PCI_CACHE_LINE_SIZE, hpp->cache_line_size);
1270 	pci_write_config_byte(dev, PCI_LATENCY_TIMER, hpp->latency_timer);
1271 	pci_read_config_word(dev, PCI_COMMAND, &pci_cmd);
1272 	if (hpp->enable_serr)
1273 		pci_cmd |= PCI_COMMAND_SERR;
1274 	if (hpp->enable_perr)
1275 		pci_cmd |= PCI_COMMAND_PARITY;
1276 	pci_write_config_word(dev, PCI_COMMAND, pci_cmd);
1277 
1278 	/* Program bridge control value */
1279 	if ((dev->class >> 8) == PCI_CLASS_BRIDGE_PCI) {
1280 		pci_write_config_byte(dev, PCI_SEC_LATENCY_TIMER,
1281 				      hpp->latency_timer);
1282 		pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &pci_bctl);
1283 		if (hpp->enable_serr)
1284 			pci_bctl |= PCI_BRIDGE_CTL_SERR;
1285 		if (hpp->enable_perr)
1286 			pci_bctl |= PCI_BRIDGE_CTL_PARITY;
1287 		pci_write_config_word(dev, PCI_BRIDGE_CONTROL, pci_bctl);
1288 	}
1289 }
1290 
1291 static void program_hpp_type1(struct pci_dev *dev, struct hpp_type1 *hpp)
1292 {
1293 	if (hpp)
1294 		dev_warn(&dev->dev, "PCI-X settings not supported\n");
1295 }
1296 
1297 static void program_hpp_type2(struct pci_dev *dev, struct hpp_type2 *hpp)
1298 {
1299 	int pos;
1300 	u32 reg32;
1301 
1302 	if (!hpp)
1303 		return;
1304 
1305 	if (hpp->revision > 1) {
1306 		dev_warn(&dev->dev, "PCIe settings rev %d not supported\n",
1307 			 hpp->revision);
1308 		return;
1309 	}
1310 
1311 	/*
1312 	 * Don't allow _HPX to change MPS or MRRS settings.  We manage
1313 	 * those to make sure they're consistent with the rest of the
1314 	 * platform.
1315 	 */
1316 	hpp->pci_exp_devctl_and |= PCI_EXP_DEVCTL_PAYLOAD |
1317 				    PCI_EXP_DEVCTL_READRQ;
1318 	hpp->pci_exp_devctl_or &= ~(PCI_EXP_DEVCTL_PAYLOAD |
1319 				    PCI_EXP_DEVCTL_READRQ);
1320 
1321 	/* Initialize Device Control Register */
1322 	pcie_capability_clear_and_set_word(dev, PCI_EXP_DEVCTL,
1323 			~hpp->pci_exp_devctl_and, hpp->pci_exp_devctl_or);
1324 
1325 	/* Initialize Link Control Register */
1326 	if (dev->subordinate)
1327 		pcie_capability_clear_and_set_word(dev, PCI_EXP_LNKCTL,
1328 			~hpp->pci_exp_lnkctl_and, hpp->pci_exp_lnkctl_or);
1329 
1330 	/* Find Advanced Error Reporting Enhanced Capability */
1331 	pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
1332 	if (!pos)
1333 		return;
1334 
1335 	/* Initialize Uncorrectable Error Mask Register */
1336 	pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, &reg32);
1337 	reg32 = (reg32 & hpp->unc_err_mask_and) | hpp->unc_err_mask_or;
1338 	pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, reg32);
1339 
1340 	/* Initialize Uncorrectable Error Severity Register */
1341 	pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &reg32);
1342 	reg32 = (reg32 & hpp->unc_err_sever_and) | hpp->unc_err_sever_or;
1343 	pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, reg32);
1344 
1345 	/* Initialize Correctable Error Mask Register */
1346 	pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, &reg32);
1347 	reg32 = (reg32 & hpp->cor_err_mask_and) | hpp->cor_err_mask_or;
1348 	pci_write_config_dword(dev, pos + PCI_ERR_COR_MASK, reg32);
1349 
1350 	/* Initialize Advanced Error Capabilities and Control Register */
1351 	pci_read_config_dword(dev, pos + PCI_ERR_CAP, &reg32);
1352 	reg32 = (reg32 & hpp->adv_err_cap_and) | hpp->adv_err_cap_or;
1353 	pci_write_config_dword(dev, pos + PCI_ERR_CAP, reg32);
1354 
1355 	/*
1356 	 * FIXME: The following two registers are not supported yet.
1357 	 *
1358 	 *   o Secondary Uncorrectable Error Severity Register
1359 	 *   o Secondary Uncorrectable Error Mask Register
1360 	 */
1361 }
1362 
1363 static void pci_configure_device(struct pci_dev *dev)
1364 {
1365 	struct hotplug_params hpp;
1366 	int ret;
1367 
1368 	memset(&hpp, 0, sizeof(hpp));
1369 	ret = pci_get_hp_params(dev, &hpp);
1370 	if (ret)
1371 		return;
1372 
1373 	program_hpp_type2(dev, hpp.t2);
1374 	program_hpp_type1(dev, hpp.t1);
1375 	program_hpp_type0(dev, hpp.t0);
1376 }
1377 
1378 static void pci_release_capabilities(struct pci_dev *dev)
1379 {
1380 	pci_vpd_release(dev);
1381 	pci_iov_release(dev);
1382 	pci_free_cap_save_buffers(dev);
1383 }
1384 
1385 /**
1386  * pci_release_dev - free a pci device structure when all users of it are finished.
1387  * @dev: device that's been disconnected
1388  *
1389  * Will be called only by the device core when all users of this pci device are
1390  * done.
1391  */
1392 static void pci_release_dev(struct device *dev)
1393 {
1394 	struct pci_dev *pci_dev;
1395 
1396 	pci_dev = to_pci_dev(dev);
1397 	pci_release_capabilities(pci_dev);
1398 	pci_release_of_node(pci_dev);
1399 	pcibios_release_device(pci_dev);
1400 	pci_bus_put(pci_dev->bus);
1401 	kfree(pci_dev->driver_override);
1402 	kfree(pci_dev);
1403 }
1404 
1405 struct pci_dev *pci_alloc_dev(struct pci_bus *bus)
1406 {
1407 	struct pci_dev *dev;
1408 
1409 	dev = kzalloc(sizeof(struct pci_dev), GFP_KERNEL);
1410 	if (!dev)
1411 		return NULL;
1412 
1413 	INIT_LIST_HEAD(&dev->bus_list);
1414 	dev->dev.type = &pci_dev_type;
1415 	dev->bus = pci_bus_get(bus);
1416 
1417 	return dev;
1418 }
1419 EXPORT_SYMBOL(pci_alloc_dev);
1420 
1421 bool pci_bus_read_dev_vendor_id(struct pci_bus *bus, int devfn, u32 *l,
1422 				int crs_timeout)
1423 {
1424 	int delay = 1;
1425 
1426 	if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
1427 		return false;
1428 
1429 	/* some broken boards return 0 or ~0 if a slot is empty: */
1430 	if (*l == 0xffffffff || *l == 0x00000000 ||
1431 	    *l == 0x0000ffff || *l == 0xffff0000)
1432 		return false;
1433 
1434 	/*
1435 	 * Configuration Request Retry Status.  Some root ports return the
1436 	 * actual device ID instead of the synthetic ID (0xFFFF) required
1437 	 * by the PCIe spec.  Ignore the device ID and only check for
1438 	 * (vendor id == 1).
1439 	 */
1440 	while ((*l & 0xffff) == 0x0001) {
1441 		if (!crs_timeout)
1442 			return false;
1443 
1444 		msleep(delay);
1445 		delay *= 2;
1446 		if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
1447 			return false;
1448 		/* Card hasn't responded in 60 seconds?  Must be stuck. */
1449 		if (delay > crs_timeout) {
1450 			printk(KERN_WARNING "pci %04x:%02x:%02x.%d: not responding\n",
1451 			       pci_domain_nr(bus), bus->number, PCI_SLOT(devfn),
1452 			       PCI_FUNC(devfn));
1453 			return false;
1454 		}
1455 	}
1456 
1457 	return true;
1458 }
1459 EXPORT_SYMBOL(pci_bus_read_dev_vendor_id);
1460 
1461 /*
1462  * Read the config data for a PCI device, sanity-check it
1463  * and fill in the dev structure...
1464  */
1465 static struct pci_dev *pci_scan_device(struct pci_bus *bus, int devfn)
1466 {
1467 	struct pci_dev *dev;
1468 	u32 l;
1469 
1470 	if (!pci_bus_read_dev_vendor_id(bus, devfn, &l, 60*1000))
1471 		return NULL;
1472 
1473 	dev = pci_alloc_dev(bus);
1474 	if (!dev)
1475 		return NULL;
1476 
1477 	dev->devfn = devfn;
1478 	dev->vendor = l & 0xffff;
1479 	dev->device = (l >> 16) & 0xffff;
1480 
1481 	pci_set_of_node(dev);
1482 
1483 	if (pci_setup_device(dev)) {
1484 		pci_bus_put(dev->bus);
1485 		kfree(dev);
1486 		return NULL;
1487 	}
1488 
1489 	return dev;
1490 }
1491 
1492 static void pci_init_capabilities(struct pci_dev *dev)
1493 {
1494 	/* MSI/MSI-X list */
1495 	pci_msi_init_pci_dev(dev);
1496 
1497 	/* Buffers for saving PCIe and PCI-X capabilities */
1498 	pci_allocate_cap_save_buffers(dev);
1499 
1500 	/* Power Management */
1501 	pci_pm_init(dev);
1502 
1503 	/* Vital Product Data */
1504 	pci_vpd_pci22_init(dev);
1505 
1506 	/* Alternative Routing-ID Forwarding */
1507 	pci_configure_ari(dev);
1508 
1509 	/* Single Root I/O Virtualization */
1510 	pci_iov_init(dev);
1511 
1512 	/* Enable ACS P2P upstream forwarding */
1513 	pci_enable_acs(dev);
1514 }
1515 
1516 void pci_device_add(struct pci_dev *dev, struct pci_bus *bus)
1517 {
1518 	int ret;
1519 
1520 	pci_configure_device(dev);
1521 
1522 	device_initialize(&dev->dev);
1523 	dev->dev.release = pci_release_dev;
1524 
1525 	set_dev_node(&dev->dev, pcibus_to_node(bus));
1526 	dev->dev.dma_mask = &dev->dma_mask;
1527 	dev->dev.dma_parms = &dev->dma_parms;
1528 	dev->dev.coherent_dma_mask = 0xffffffffull;
1529 
1530 	pci_set_dma_max_seg_size(dev, 65536);
1531 	pci_set_dma_seg_boundary(dev, 0xffffffff);
1532 
1533 	/* Fix up broken headers */
1534 	pci_fixup_device(pci_fixup_header, dev);
1535 
1536 	/* moved out from quirk header fixup code */
1537 	pci_reassigndev_resource_alignment(dev);
1538 
1539 	/* Clear the state_saved flag. */
1540 	dev->state_saved = false;
1541 
1542 	/* Initialize various capabilities */
1543 	pci_init_capabilities(dev);
1544 
1545 	/*
1546 	 * Add the device to our list of discovered devices
1547 	 * and the bus list for fixup functions, etc.
1548 	 */
1549 	down_write(&pci_bus_sem);
1550 	list_add_tail(&dev->bus_list, &bus->devices);
1551 	up_write(&pci_bus_sem);
1552 
1553 	ret = pcibios_add_device(dev);
1554 	WARN_ON(ret < 0);
1555 
1556 	/* Notifier could use PCI capabilities */
1557 	dev->match_driver = false;
1558 	ret = device_add(&dev->dev);
1559 	WARN_ON(ret < 0);
1560 }
1561 
1562 struct pci_dev *pci_scan_single_device(struct pci_bus *bus, int devfn)
1563 {
1564 	struct pci_dev *dev;
1565 
1566 	dev = pci_get_slot(bus, devfn);
1567 	if (dev) {
1568 		pci_dev_put(dev);
1569 		return dev;
1570 	}
1571 
1572 	dev = pci_scan_device(bus, devfn);
1573 	if (!dev)
1574 		return NULL;
1575 
1576 	pci_device_add(dev, bus);
1577 
1578 	return dev;
1579 }
1580 EXPORT_SYMBOL(pci_scan_single_device);
1581 
1582 static unsigned next_fn(struct pci_bus *bus, struct pci_dev *dev, unsigned fn)
1583 {
1584 	int pos;
1585 	u16 cap = 0;
1586 	unsigned next_fn;
1587 
1588 	if (pci_ari_enabled(bus)) {
1589 		if (!dev)
1590 			return 0;
1591 		pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ARI);
1592 		if (!pos)
1593 			return 0;
1594 
1595 		pci_read_config_word(dev, pos + PCI_ARI_CAP, &cap);
1596 		next_fn = PCI_ARI_CAP_NFN(cap);
1597 		if (next_fn <= fn)
1598 			return 0;	/* protect against malformed list */
1599 
1600 		return next_fn;
1601 	}
1602 
1603 	/* dev may be NULL for non-contiguous multifunction devices */
1604 	if (!dev || dev->multifunction)
1605 		return (fn + 1) % 8;
1606 
1607 	return 0;
1608 }
1609 
1610 static int only_one_child(struct pci_bus *bus)
1611 {
1612 	struct pci_dev *parent = bus->self;
1613 
1614 	if (!parent || !pci_is_pcie(parent))
1615 		return 0;
1616 	if (pci_pcie_type(parent) == PCI_EXP_TYPE_ROOT_PORT)
1617 		return 1;
1618 	if (pci_pcie_type(parent) == PCI_EXP_TYPE_DOWNSTREAM &&
1619 	    !pci_has_flag(PCI_SCAN_ALL_PCIE_DEVS))
1620 		return 1;
1621 	return 0;
1622 }
1623 
1624 /**
1625  * pci_scan_slot - scan a PCI slot on a bus for devices.
1626  * @bus: PCI bus to scan
1627  * @devfn: slot number to scan (must have zero function.)
1628  *
1629  * Scan a PCI slot on the specified PCI bus for devices, adding
1630  * discovered devices to the @bus->devices list.  New devices
1631  * will not have is_added set.
1632  *
1633  * Returns the number of new devices found.
1634  */
1635 int pci_scan_slot(struct pci_bus *bus, int devfn)
1636 {
1637 	unsigned fn, nr = 0;
1638 	struct pci_dev *dev;
1639 
1640 	if (only_one_child(bus) && (devfn > 0))
1641 		return 0; /* Already scanned the entire slot */
1642 
1643 	dev = pci_scan_single_device(bus, devfn);
1644 	if (!dev)
1645 		return 0;
1646 	if (!dev->is_added)
1647 		nr++;
1648 
1649 	for (fn = next_fn(bus, dev, 0); fn > 0; fn = next_fn(bus, dev, fn)) {
1650 		dev = pci_scan_single_device(bus, devfn + fn);
1651 		if (dev) {
1652 			if (!dev->is_added)
1653 				nr++;
1654 			dev->multifunction = 1;
1655 		}
1656 	}
1657 
1658 	/* only one slot has pcie device */
1659 	if (bus->self && nr)
1660 		pcie_aspm_init_link_state(bus->self);
1661 
1662 	return nr;
1663 }
1664 EXPORT_SYMBOL(pci_scan_slot);
1665 
1666 static int pcie_find_smpss(struct pci_dev *dev, void *data)
1667 {
1668 	u8 *smpss = data;
1669 
1670 	if (!pci_is_pcie(dev))
1671 		return 0;
1672 
1673 	/*
1674 	 * We don't have a way to change MPS settings on devices that have
1675 	 * drivers attached.  A hot-added device might support only the minimum
1676 	 * MPS setting (MPS=128).  Therefore, if the fabric contains a bridge
1677 	 * where devices may be hot-added, we limit the fabric MPS to 128 so
1678 	 * hot-added devices will work correctly.
1679 	 *
1680 	 * However, if we hot-add a device to a slot directly below a Root
1681 	 * Port, it's impossible for there to be other existing devices below
1682 	 * the port.  We don't limit the MPS in this case because we can
1683 	 * reconfigure MPS on both the Root Port and the hot-added device,
1684 	 * and there are no other devices involved.
1685 	 *
1686 	 * Note that this PCIE_BUS_SAFE path assumes no peer-to-peer DMA.
1687 	 */
1688 	if (dev->is_hotplug_bridge &&
1689 	    pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT)
1690 		*smpss = 0;
1691 
1692 	if (*smpss > dev->pcie_mpss)
1693 		*smpss = dev->pcie_mpss;
1694 
1695 	return 0;
1696 }
1697 
1698 static void pcie_write_mps(struct pci_dev *dev, int mps)
1699 {
1700 	int rc;
1701 
1702 	if (pcie_bus_config == PCIE_BUS_PERFORMANCE) {
1703 		mps = 128 << dev->pcie_mpss;
1704 
1705 		if (pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT &&
1706 		    dev->bus->self)
1707 			/* For "Performance", the assumption is made that
1708 			 * downstream communication will never be larger than
1709 			 * the MRRS.  So, the MPS only needs to be configured
1710 			 * for the upstream communication.  This being the case,
1711 			 * walk from the top down and set the MPS of the child
1712 			 * to that of the parent bus.
1713 			 *
1714 			 * Configure the device MPS with the smaller of the
1715 			 * device MPSS or the bridge MPS (which is assumed to be
1716 			 * properly configured at this point to the largest
1717 			 * allowable MPS based on its parent bus).
1718 			 */
1719 			mps = min(mps, pcie_get_mps(dev->bus->self));
1720 	}
1721 
1722 	rc = pcie_set_mps(dev, mps);
1723 	if (rc)
1724 		dev_err(&dev->dev, "Failed attempting to set the MPS\n");
1725 }
1726 
1727 static void pcie_write_mrrs(struct pci_dev *dev)
1728 {
1729 	int rc, mrrs;
1730 
1731 	/* In the "safe" case, do not configure the MRRS.  There appear to be
1732 	 * issues with setting MRRS to 0 on a number of devices.
1733 	 */
1734 	if (pcie_bus_config != PCIE_BUS_PERFORMANCE)
1735 		return;
1736 
1737 	/* For Max performance, the MRRS must be set to the largest supported
1738 	 * value.  However, it cannot be configured larger than the MPS the
1739 	 * device or the bus can support.  This should already be properly
1740 	 * configured by a prior call to pcie_write_mps.
1741 	 */
1742 	mrrs = pcie_get_mps(dev);
1743 
1744 	/* MRRS is a R/W register.  Invalid values can be written, but a
1745 	 * subsequent read will verify if the value is acceptable or not.
1746 	 * If the MRRS value provided is not acceptable (e.g., too large),
1747 	 * shrink the value until it is acceptable to the HW.
1748 	 */
1749 	while (mrrs != pcie_get_readrq(dev) && mrrs >= 128) {
1750 		rc = pcie_set_readrq(dev, mrrs);
1751 		if (!rc)
1752 			break;
1753 
1754 		dev_warn(&dev->dev, "Failed attempting to set the MRRS\n");
1755 		mrrs /= 2;
1756 	}
1757 
1758 	if (mrrs < 128)
1759 		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");
1760 }
1761 
1762 static void pcie_bus_detect_mps(struct pci_dev *dev)
1763 {
1764 	struct pci_dev *bridge = dev->bus->self;
1765 	int mps, p_mps;
1766 
1767 	if (!bridge)
1768 		return;
1769 
1770 	mps = pcie_get_mps(dev);
1771 	p_mps = pcie_get_mps(bridge);
1772 
1773 	if (mps != p_mps)
1774 		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",
1775 			 mps, pci_name(bridge), p_mps);
1776 }
1777 
1778 static int pcie_bus_configure_set(struct pci_dev *dev, void *data)
1779 {
1780 	int mps, orig_mps;
1781 
1782 	if (!pci_is_pcie(dev))
1783 		return 0;
1784 
1785 	if (pcie_bus_config == PCIE_BUS_TUNE_OFF) {
1786 		pcie_bus_detect_mps(dev);
1787 		return 0;
1788 	}
1789 
1790 	mps = 128 << *(u8 *)data;
1791 	orig_mps = pcie_get_mps(dev);
1792 
1793 	pcie_write_mps(dev, mps);
1794 	pcie_write_mrrs(dev);
1795 
1796 	dev_info(&dev->dev, "Max Payload Size set to %4d/%4d (was %4d), Max Read Rq %4d\n",
1797 		 pcie_get_mps(dev), 128 << dev->pcie_mpss,
1798 		 orig_mps, pcie_get_readrq(dev));
1799 
1800 	return 0;
1801 }
1802 
1803 /* pcie_bus_configure_settings requires that pci_walk_bus work in a top-down,
1804  * parents then children fashion.  If this changes, then this code will not
1805  * work as designed.
1806  */
1807 void pcie_bus_configure_settings(struct pci_bus *bus)
1808 {
1809 	u8 smpss = 0;
1810 
1811 	if (!bus->self)
1812 		return;
1813 
1814 	if (!pci_is_pcie(bus->self))
1815 		return;
1816 
1817 	/* FIXME - Peer to peer DMA is possible, though the endpoint would need
1818 	 * to be aware of the MPS of the destination.  To work around this,
1819 	 * simply force the MPS of the entire system to the smallest possible.
1820 	 */
1821 	if (pcie_bus_config == PCIE_BUS_PEER2PEER)
1822 		smpss = 0;
1823 
1824 	if (pcie_bus_config == PCIE_BUS_SAFE) {
1825 		smpss = bus->self->pcie_mpss;
1826 
1827 		pcie_find_smpss(bus->self, &smpss);
1828 		pci_walk_bus(bus, pcie_find_smpss, &smpss);
1829 	}
1830 
1831 	pcie_bus_configure_set(bus->self, &smpss);
1832 	pci_walk_bus(bus, pcie_bus_configure_set, &smpss);
1833 }
1834 EXPORT_SYMBOL_GPL(pcie_bus_configure_settings);
1835 
1836 unsigned int pci_scan_child_bus(struct pci_bus *bus)
1837 {
1838 	unsigned int devfn, pass, max = bus->busn_res.start;
1839 	struct pci_dev *dev;
1840 
1841 	dev_dbg(&bus->dev, "scanning bus\n");
1842 
1843 	/* Go find them, Rover! */
1844 	for (devfn = 0; devfn < 0x100; devfn += 8)
1845 		pci_scan_slot(bus, devfn);
1846 
1847 	/* Reserve buses for SR-IOV capability. */
1848 	max += pci_iov_bus_range(bus);
1849 
1850 	/*
1851 	 * After performing arch-dependent fixup of the bus, look behind
1852 	 * all PCI-to-PCI bridges on this bus.
1853 	 */
1854 	if (!bus->is_added) {
1855 		dev_dbg(&bus->dev, "fixups for bus\n");
1856 		pcibios_fixup_bus(bus);
1857 		bus->is_added = 1;
1858 	}
1859 
1860 	for (pass = 0; pass < 2; pass++)
1861 		list_for_each_entry(dev, &bus->devices, bus_list) {
1862 			if (pci_is_bridge(dev))
1863 				max = pci_scan_bridge(bus, dev, max, pass);
1864 		}
1865 
1866 	/*
1867 	 * We've scanned the bus and so we know all about what's on
1868 	 * the other side of any bridges that may be on this bus plus
1869 	 * any devices.
1870 	 *
1871 	 * Return how far we've got finding sub-buses.
1872 	 */
1873 	dev_dbg(&bus->dev, "bus scan returning with max=%02x\n", max);
1874 	return max;
1875 }
1876 EXPORT_SYMBOL_GPL(pci_scan_child_bus);
1877 
1878 /**
1879  * pcibios_root_bridge_prepare - Platform-specific host bridge setup.
1880  * @bridge: Host bridge to set up.
1881  *
1882  * Default empty implementation.  Replace with an architecture-specific setup
1883  * routine, if necessary.
1884  */
1885 int __weak pcibios_root_bridge_prepare(struct pci_host_bridge *bridge)
1886 {
1887 	return 0;
1888 }
1889 
1890 void __weak pcibios_add_bus(struct pci_bus *bus)
1891 {
1892 }
1893 
1894 void __weak pcibios_remove_bus(struct pci_bus *bus)
1895 {
1896 }
1897 
1898 struct pci_bus *pci_create_root_bus(struct device *parent, int bus,
1899 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
1900 {
1901 	int error;
1902 	struct pci_host_bridge *bridge;
1903 	struct pci_bus *b, *b2;
1904 	struct pci_host_bridge_window *window, *n;
1905 	struct resource *res;
1906 	resource_size_t offset;
1907 	char bus_addr[64];
1908 	char *fmt;
1909 
1910 	b = pci_alloc_bus(NULL);
1911 	if (!b)
1912 		return NULL;
1913 
1914 	b->sysdata = sysdata;
1915 	b->ops = ops;
1916 	b->number = b->busn_res.start = bus;
1917 	pci_bus_assign_domain_nr(b, parent);
1918 	b2 = pci_find_bus(pci_domain_nr(b), bus);
1919 	if (b2) {
1920 		/* If we already got to this bus through a different bridge, ignore it */
1921 		dev_dbg(&b2->dev, "bus already known\n");
1922 		goto err_out;
1923 	}
1924 
1925 	bridge = pci_alloc_host_bridge(b);
1926 	if (!bridge)
1927 		goto err_out;
1928 
1929 	bridge->dev.parent = parent;
1930 	bridge->dev.release = pci_release_host_bridge_dev;
1931 	dev_set_name(&bridge->dev, "pci%04x:%02x", pci_domain_nr(b), bus);
1932 	error = pcibios_root_bridge_prepare(bridge);
1933 	if (error) {
1934 		kfree(bridge);
1935 		goto err_out;
1936 	}
1937 
1938 	error = device_register(&bridge->dev);
1939 	if (error) {
1940 		put_device(&bridge->dev);
1941 		goto err_out;
1942 	}
1943 	b->bridge = get_device(&bridge->dev);
1944 	device_enable_async_suspend(b->bridge);
1945 	pci_set_bus_of_node(b);
1946 
1947 	if (!parent)
1948 		set_dev_node(b->bridge, pcibus_to_node(b));
1949 
1950 	b->dev.class = &pcibus_class;
1951 	b->dev.parent = b->bridge;
1952 	dev_set_name(&b->dev, "%04x:%02x", pci_domain_nr(b), bus);
1953 	error = device_register(&b->dev);
1954 	if (error)
1955 		goto class_dev_reg_err;
1956 
1957 	pcibios_add_bus(b);
1958 
1959 	/* Create legacy_io and legacy_mem files for this bus */
1960 	pci_create_legacy_files(b);
1961 
1962 	if (parent)
1963 		dev_info(parent, "PCI host bridge to bus %s\n", dev_name(&b->dev));
1964 	else
1965 		printk(KERN_INFO "PCI host bridge to bus %s\n", dev_name(&b->dev));
1966 
1967 	/* Add initial resources to the bus */
1968 	list_for_each_entry_safe(window, n, resources, list) {
1969 		list_move_tail(&window->list, &bridge->windows);
1970 		res = window->res;
1971 		offset = window->offset;
1972 		if (res->flags & IORESOURCE_BUS)
1973 			pci_bus_insert_busn_res(b, bus, res->end);
1974 		else
1975 			pci_bus_add_resource(b, res, 0);
1976 		if (offset) {
1977 			if (resource_type(res) == IORESOURCE_IO)
1978 				fmt = " (bus address [%#06llx-%#06llx])";
1979 			else
1980 				fmt = " (bus address [%#010llx-%#010llx])";
1981 			snprintf(bus_addr, sizeof(bus_addr), fmt,
1982 				 (unsigned long long) (res->start - offset),
1983 				 (unsigned long long) (res->end - offset));
1984 		} else
1985 			bus_addr[0] = '\0';
1986 		dev_info(&b->dev, "root bus resource %pR%s\n", res, bus_addr);
1987 	}
1988 
1989 	down_write(&pci_bus_sem);
1990 	list_add_tail(&b->node, &pci_root_buses);
1991 	up_write(&pci_bus_sem);
1992 
1993 	return b;
1994 
1995 class_dev_reg_err:
1996 	put_device(&bridge->dev);
1997 	device_unregister(&bridge->dev);
1998 err_out:
1999 	kfree(b);
2000 	return NULL;
2001 }
2002 
2003 int pci_bus_insert_busn_res(struct pci_bus *b, int bus, int bus_max)
2004 {
2005 	struct resource *res = &b->busn_res;
2006 	struct resource *parent_res, *conflict;
2007 
2008 	res->start = bus;
2009 	res->end = bus_max;
2010 	res->flags = IORESOURCE_BUS;
2011 
2012 	if (!pci_is_root_bus(b))
2013 		parent_res = &b->parent->busn_res;
2014 	else {
2015 		parent_res = get_pci_domain_busn_res(pci_domain_nr(b));
2016 		res->flags |= IORESOURCE_PCI_FIXED;
2017 	}
2018 
2019 	conflict = request_resource_conflict(parent_res, res);
2020 
2021 	if (conflict)
2022 		dev_printk(KERN_DEBUG, &b->dev,
2023 			   "busn_res: can not insert %pR under %s%pR (conflicts with %s %pR)\n",
2024 			    res, pci_is_root_bus(b) ? "domain " : "",
2025 			    parent_res, conflict->name, conflict);
2026 
2027 	return conflict == NULL;
2028 }
2029 
2030 int pci_bus_update_busn_res_end(struct pci_bus *b, int bus_max)
2031 {
2032 	struct resource *res = &b->busn_res;
2033 	struct resource old_res = *res;
2034 	resource_size_t size;
2035 	int ret;
2036 
2037 	if (res->start > bus_max)
2038 		return -EINVAL;
2039 
2040 	size = bus_max - res->start + 1;
2041 	ret = adjust_resource(res, res->start, size);
2042 	dev_printk(KERN_DEBUG, &b->dev,
2043 			"busn_res: %pR end %s updated to %02x\n",
2044 			&old_res, ret ? "can not be" : "is", bus_max);
2045 
2046 	if (!ret && !res->parent)
2047 		pci_bus_insert_busn_res(b, res->start, res->end);
2048 
2049 	return ret;
2050 }
2051 
2052 void pci_bus_release_busn_res(struct pci_bus *b)
2053 {
2054 	struct resource *res = &b->busn_res;
2055 	int ret;
2056 
2057 	if (!res->flags || !res->parent)
2058 		return;
2059 
2060 	ret = release_resource(res);
2061 	dev_printk(KERN_DEBUG, &b->dev,
2062 			"busn_res: %pR %s released\n",
2063 			res, ret ? "can not be" : "is");
2064 }
2065 
2066 struct pci_bus *pci_scan_root_bus(struct device *parent, int bus,
2067 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
2068 {
2069 	struct pci_host_bridge_window *window;
2070 	bool found = false;
2071 	struct pci_bus *b;
2072 	int max;
2073 
2074 	list_for_each_entry(window, resources, list)
2075 		if (window->res->flags & IORESOURCE_BUS) {
2076 			found = true;
2077 			break;
2078 		}
2079 
2080 	b = pci_create_root_bus(parent, bus, ops, sysdata, resources);
2081 	if (!b)
2082 		return NULL;
2083 
2084 	if (!found) {
2085 		dev_info(&b->dev,
2086 		 "No busn resource found for root bus, will use [bus %02x-ff]\n",
2087 			bus);
2088 		pci_bus_insert_busn_res(b, bus, 255);
2089 	}
2090 
2091 	max = pci_scan_child_bus(b);
2092 
2093 	if (!found)
2094 		pci_bus_update_busn_res_end(b, max);
2095 
2096 	pci_bus_add_devices(b);
2097 	return b;
2098 }
2099 EXPORT_SYMBOL(pci_scan_root_bus);
2100 
2101 /* Deprecated; use pci_scan_root_bus() instead */
2102 struct pci_bus *pci_scan_bus_parented(struct device *parent,
2103 		int bus, struct pci_ops *ops, void *sysdata)
2104 {
2105 	LIST_HEAD(resources);
2106 	struct pci_bus *b;
2107 
2108 	pci_add_resource(&resources, &ioport_resource);
2109 	pci_add_resource(&resources, &iomem_resource);
2110 	pci_add_resource(&resources, &busn_resource);
2111 	b = pci_create_root_bus(parent, bus, ops, sysdata, &resources);
2112 	if (b)
2113 		pci_scan_child_bus(b);
2114 	else
2115 		pci_free_resource_list(&resources);
2116 	return b;
2117 }
2118 EXPORT_SYMBOL(pci_scan_bus_parented);
2119 
2120 struct pci_bus *pci_scan_bus(int bus, struct pci_ops *ops,
2121 					void *sysdata)
2122 {
2123 	LIST_HEAD(resources);
2124 	struct pci_bus *b;
2125 
2126 	pci_add_resource(&resources, &ioport_resource);
2127 	pci_add_resource(&resources, &iomem_resource);
2128 	pci_add_resource(&resources, &busn_resource);
2129 	b = pci_create_root_bus(NULL, bus, ops, sysdata, &resources);
2130 	if (b) {
2131 		pci_scan_child_bus(b);
2132 		pci_bus_add_devices(b);
2133 	} else {
2134 		pci_free_resource_list(&resources);
2135 	}
2136 	return b;
2137 }
2138 EXPORT_SYMBOL(pci_scan_bus);
2139 
2140 /**
2141  * pci_rescan_bus_bridge_resize - scan a PCI bus for devices.
2142  * @bridge: PCI bridge for the bus to scan
2143  *
2144  * Scan a PCI bus and child buses for new devices, add them,
2145  * and enable them, resizing bridge mmio/io resource if necessary
2146  * and possible.  The caller must ensure the child devices are already
2147  * removed for resizing to occur.
2148  *
2149  * Returns the max number of subordinate bus discovered.
2150  */
2151 unsigned int pci_rescan_bus_bridge_resize(struct pci_dev *bridge)
2152 {
2153 	unsigned int max;
2154 	struct pci_bus *bus = bridge->subordinate;
2155 
2156 	max = pci_scan_child_bus(bus);
2157 
2158 	pci_assign_unassigned_bridge_resources(bridge);
2159 
2160 	pci_bus_add_devices(bus);
2161 
2162 	return max;
2163 }
2164 
2165 /**
2166  * pci_rescan_bus - scan a PCI bus for devices.
2167  * @bus: PCI bus to scan
2168  *
2169  * Scan a PCI bus and child buses for new devices, adds them,
2170  * and enables them.
2171  *
2172  * Returns the max number of subordinate bus discovered.
2173  */
2174 unsigned int pci_rescan_bus(struct pci_bus *bus)
2175 {
2176 	unsigned int max;
2177 
2178 	max = pci_scan_child_bus(bus);
2179 	pci_assign_unassigned_bus_resources(bus);
2180 	pci_bus_add_devices(bus);
2181 
2182 	return max;
2183 }
2184 EXPORT_SYMBOL_GPL(pci_rescan_bus);
2185 
2186 /*
2187  * pci_rescan_bus(), pci_rescan_bus_bridge_resize() and PCI device removal
2188  * routines should always be executed under this mutex.
2189  */
2190 static DEFINE_MUTEX(pci_rescan_remove_lock);
2191 
2192 void pci_lock_rescan_remove(void)
2193 {
2194 	mutex_lock(&pci_rescan_remove_lock);
2195 }
2196 EXPORT_SYMBOL_GPL(pci_lock_rescan_remove);
2197 
2198 void pci_unlock_rescan_remove(void)
2199 {
2200 	mutex_unlock(&pci_rescan_remove_lock);
2201 }
2202 EXPORT_SYMBOL_GPL(pci_unlock_rescan_remove);
2203 
2204 static int __init pci_sort_bf_cmp(const struct device *d_a,
2205 				  const struct device *d_b)
2206 {
2207 	const struct pci_dev *a = to_pci_dev(d_a);
2208 	const struct pci_dev *b = to_pci_dev(d_b);
2209 
2210 	if      (pci_domain_nr(a->bus) < pci_domain_nr(b->bus)) return -1;
2211 	else if (pci_domain_nr(a->bus) > pci_domain_nr(b->bus)) return  1;
2212 
2213 	if      (a->bus->number < b->bus->number) return -1;
2214 	else if (a->bus->number > b->bus->number) return  1;
2215 
2216 	if      (a->devfn < b->devfn) return -1;
2217 	else if (a->devfn > b->devfn) return  1;
2218 
2219 	return 0;
2220 }
2221 
2222 void __init pci_sort_breadthfirst(void)
2223 {
2224 	bus_sort_breadthfirst(&pci_bus_type, &pci_sort_bf_cmp);
2225 }
2226