xref: /openbmc/linux/drivers/pci/probe.c (revision fe7498ef)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * PCI detection and setup code
4  */
5 
6 #include <linux/kernel.h>
7 #include <linux/delay.h>
8 #include <linux/init.h>
9 #include <linux/pci.h>
10 #include <linux/msi.h>
11 #include <linux/of_device.h>
12 #include <linux/of_pci.h>
13 #include <linux/pci_hotplug.h>
14 #include <linux/slab.h>
15 #include <linux/module.h>
16 #include <linux/cpumask.h>
17 #include <linux/aer.h>
18 #include <linux/acpi.h>
19 #include <linux/hypervisor.h>
20 #include <linux/irqdomain.h>
21 #include <linux/pm_runtime.h>
22 #include <linux/bitfield.h>
23 #include "pci.h"
24 
25 #define CARDBUS_LATENCY_TIMER	176	/* secondary latency timer */
26 #define CARDBUS_RESERVE_BUSNR	3
27 
28 static struct resource busn_resource = {
29 	.name	= "PCI busn",
30 	.start	= 0,
31 	.end	= 255,
32 	.flags	= IORESOURCE_BUS,
33 };
34 
35 /* Ugh.  Need to stop exporting this to modules. */
36 LIST_HEAD(pci_root_buses);
37 EXPORT_SYMBOL(pci_root_buses);
38 
39 static LIST_HEAD(pci_domain_busn_res_list);
40 
41 struct pci_domain_busn_res {
42 	struct list_head list;
43 	struct resource res;
44 	int domain_nr;
45 };
46 
47 static struct resource *get_pci_domain_busn_res(int domain_nr)
48 {
49 	struct pci_domain_busn_res *r;
50 
51 	list_for_each_entry(r, &pci_domain_busn_res_list, list)
52 		if (r->domain_nr == domain_nr)
53 			return &r->res;
54 
55 	r = kzalloc(sizeof(*r), GFP_KERNEL);
56 	if (!r)
57 		return NULL;
58 
59 	r->domain_nr = domain_nr;
60 	r->res.start = 0;
61 	r->res.end = 0xff;
62 	r->res.flags = IORESOURCE_BUS | IORESOURCE_PCI_FIXED;
63 
64 	list_add_tail(&r->list, &pci_domain_busn_res_list);
65 
66 	return &r->res;
67 }
68 
69 /*
70  * Some device drivers need know if PCI is initiated.
71  * Basically, we think PCI is not initiated when there
72  * is no device to be found on the pci_bus_type.
73  */
74 int no_pci_devices(void)
75 {
76 	struct device *dev;
77 	int no_devices;
78 
79 	dev = bus_find_next_device(&pci_bus_type, NULL);
80 	no_devices = (dev == NULL);
81 	put_device(dev);
82 	return no_devices;
83 }
84 EXPORT_SYMBOL(no_pci_devices);
85 
86 /*
87  * PCI Bus Class
88  */
89 static void release_pcibus_dev(struct device *dev)
90 {
91 	struct pci_bus *pci_bus = to_pci_bus(dev);
92 
93 	put_device(pci_bus->bridge);
94 	pci_bus_remove_resources(pci_bus);
95 	pci_release_bus_of_node(pci_bus);
96 	kfree(pci_bus);
97 }
98 
99 static struct class pcibus_class = {
100 	.name		= "pci_bus",
101 	.dev_release	= &release_pcibus_dev,
102 	.dev_groups	= pcibus_groups,
103 };
104 
105 static int __init pcibus_class_init(void)
106 {
107 	return class_register(&pcibus_class);
108 }
109 postcore_initcall(pcibus_class_init);
110 
111 static u64 pci_size(u64 base, u64 maxbase, u64 mask)
112 {
113 	u64 size = mask & maxbase;	/* Find the significant bits */
114 	if (!size)
115 		return 0;
116 
117 	/*
118 	 * Get the lowest of them to find the decode size, and from that
119 	 * the extent.
120 	 */
121 	size = size & ~(size-1);
122 
123 	/*
124 	 * base == maxbase can be valid only if the BAR has already been
125 	 * programmed with all 1s.
126 	 */
127 	if (base == maxbase && ((base | (size - 1)) & mask) != mask)
128 		return 0;
129 
130 	return size;
131 }
132 
133 static inline unsigned long decode_bar(struct pci_dev *dev, u32 bar)
134 {
135 	u32 mem_type;
136 	unsigned long flags;
137 
138 	if ((bar & PCI_BASE_ADDRESS_SPACE) == PCI_BASE_ADDRESS_SPACE_IO) {
139 		flags = bar & ~PCI_BASE_ADDRESS_IO_MASK;
140 		flags |= IORESOURCE_IO;
141 		return flags;
142 	}
143 
144 	flags = bar & ~PCI_BASE_ADDRESS_MEM_MASK;
145 	flags |= IORESOURCE_MEM;
146 	if (flags & PCI_BASE_ADDRESS_MEM_PREFETCH)
147 		flags |= IORESOURCE_PREFETCH;
148 
149 	mem_type = bar & PCI_BASE_ADDRESS_MEM_TYPE_MASK;
150 	switch (mem_type) {
151 	case PCI_BASE_ADDRESS_MEM_TYPE_32:
152 		break;
153 	case PCI_BASE_ADDRESS_MEM_TYPE_1M:
154 		/* 1M mem BAR treated as 32-bit BAR */
155 		break;
156 	case PCI_BASE_ADDRESS_MEM_TYPE_64:
157 		flags |= IORESOURCE_MEM_64;
158 		break;
159 	default:
160 		/* mem unknown type treated as 32-bit BAR */
161 		break;
162 	}
163 	return flags;
164 }
165 
166 #define PCI_COMMAND_DECODE_ENABLE	(PCI_COMMAND_MEMORY | PCI_COMMAND_IO)
167 
168 /**
169  * __pci_read_base - Read a PCI BAR
170  * @dev: the PCI device
171  * @type: type of the BAR
172  * @res: resource buffer to be filled in
173  * @pos: BAR position in the config space
174  *
175  * Returns 1 if the BAR is 64-bit, or 0 if 32-bit.
176  */
177 int __pci_read_base(struct pci_dev *dev, enum pci_bar_type type,
178 		    struct resource *res, unsigned int pos)
179 {
180 	u32 l = 0, sz = 0, mask;
181 	u64 l64, sz64, mask64;
182 	u16 orig_cmd;
183 	struct pci_bus_region region, inverted_region;
184 
185 	mask = type ? PCI_ROM_ADDRESS_MASK : ~0;
186 
187 	/* No printks while decoding is disabled! */
188 	if (!dev->mmio_always_on) {
189 		pci_read_config_word(dev, PCI_COMMAND, &orig_cmd);
190 		if (orig_cmd & PCI_COMMAND_DECODE_ENABLE) {
191 			pci_write_config_word(dev, PCI_COMMAND,
192 				orig_cmd & ~PCI_COMMAND_DECODE_ENABLE);
193 		}
194 	}
195 
196 	res->name = pci_name(dev);
197 
198 	pci_read_config_dword(dev, pos, &l);
199 	pci_write_config_dword(dev, pos, l | mask);
200 	pci_read_config_dword(dev, pos, &sz);
201 	pci_write_config_dword(dev, pos, l);
202 
203 	/*
204 	 * All bits set in sz means the device isn't working properly.
205 	 * If the BAR isn't implemented, all bits must be 0.  If it's a
206 	 * memory BAR or a ROM, bit 0 must be clear; if it's an io BAR, bit
207 	 * 1 must be clear.
208 	 */
209 	if (sz == 0xffffffff)
210 		sz = 0;
211 
212 	/*
213 	 * I don't know how l can have all bits set.  Copied from old code.
214 	 * Maybe it fixes a bug on some ancient platform.
215 	 */
216 	if (l == 0xffffffff)
217 		l = 0;
218 
219 	if (type == pci_bar_unknown) {
220 		res->flags = decode_bar(dev, l);
221 		res->flags |= IORESOURCE_SIZEALIGN;
222 		if (res->flags & IORESOURCE_IO) {
223 			l64 = l & PCI_BASE_ADDRESS_IO_MASK;
224 			sz64 = sz & PCI_BASE_ADDRESS_IO_MASK;
225 			mask64 = PCI_BASE_ADDRESS_IO_MASK & (u32)IO_SPACE_LIMIT;
226 		} else {
227 			l64 = l & PCI_BASE_ADDRESS_MEM_MASK;
228 			sz64 = sz & PCI_BASE_ADDRESS_MEM_MASK;
229 			mask64 = (u32)PCI_BASE_ADDRESS_MEM_MASK;
230 		}
231 	} else {
232 		if (l & PCI_ROM_ADDRESS_ENABLE)
233 			res->flags |= IORESOURCE_ROM_ENABLE;
234 		l64 = l & PCI_ROM_ADDRESS_MASK;
235 		sz64 = sz & PCI_ROM_ADDRESS_MASK;
236 		mask64 = PCI_ROM_ADDRESS_MASK;
237 	}
238 
239 	if (res->flags & IORESOURCE_MEM_64) {
240 		pci_read_config_dword(dev, pos + 4, &l);
241 		pci_write_config_dword(dev, pos + 4, ~0);
242 		pci_read_config_dword(dev, pos + 4, &sz);
243 		pci_write_config_dword(dev, pos + 4, l);
244 
245 		l64 |= ((u64)l << 32);
246 		sz64 |= ((u64)sz << 32);
247 		mask64 |= ((u64)~0 << 32);
248 	}
249 
250 	if (!dev->mmio_always_on && (orig_cmd & PCI_COMMAND_DECODE_ENABLE))
251 		pci_write_config_word(dev, PCI_COMMAND, orig_cmd);
252 
253 	if (!sz64)
254 		goto fail;
255 
256 	sz64 = pci_size(l64, sz64, mask64);
257 	if (!sz64) {
258 		pci_info(dev, FW_BUG "reg 0x%x: invalid BAR (can't size)\n",
259 			 pos);
260 		goto fail;
261 	}
262 
263 	if (res->flags & IORESOURCE_MEM_64) {
264 		if ((sizeof(pci_bus_addr_t) < 8 || sizeof(resource_size_t) < 8)
265 		    && sz64 > 0x100000000ULL) {
266 			res->flags |= IORESOURCE_UNSET | IORESOURCE_DISABLED;
267 			res->start = 0;
268 			res->end = 0;
269 			pci_err(dev, "reg 0x%x: can't handle BAR larger than 4GB (size %#010llx)\n",
270 				pos, (unsigned long long)sz64);
271 			goto out;
272 		}
273 
274 		if ((sizeof(pci_bus_addr_t) < 8) && l) {
275 			/* Above 32-bit boundary; try to reallocate */
276 			res->flags |= IORESOURCE_UNSET;
277 			res->start = 0;
278 			res->end = sz64 - 1;
279 			pci_info(dev, "reg 0x%x: can't handle BAR above 4GB (bus address %#010llx)\n",
280 				 pos, (unsigned long long)l64);
281 			goto out;
282 		}
283 	}
284 
285 	region.start = l64;
286 	region.end = l64 + sz64 - 1;
287 
288 	pcibios_bus_to_resource(dev->bus, res, &region);
289 	pcibios_resource_to_bus(dev->bus, &inverted_region, res);
290 
291 	/*
292 	 * If "A" is a BAR value (a bus address), "bus_to_resource(A)" is
293 	 * the corresponding resource address (the physical address used by
294 	 * the CPU.  Converting that resource address back to a bus address
295 	 * should yield the original BAR value:
296 	 *
297 	 *     resource_to_bus(bus_to_resource(A)) == A
298 	 *
299 	 * If it doesn't, CPU accesses to "bus_to_resource(A)" will not
300 	 * be claimed by the device.
301 	 */
302 	if (inverted_region.start != region.start) {
303 		res->flags |= IORESOURCE_UNSET;
304 		res->start = 0;
305 		res->end = region.end - region.start;
306 		pci_info(dev, "reg 0x%x: initial BAR value %#010llx invalid\n",
307 			 pos, (unsigned long long)region.start);
308 	}
309 
310 	goto out;
311 
312 
313 fail:
314 	res->flags = 0;
315 out:
316 	if (res->flags)
317 		pci_info(dev, "reg 0x%x: %pR\n", pos, res);
318 
319 	return (res->flags & IORESOURCE_MEM_64) ? 1 : 0;
320 }
321 
322 static void pci_read_bases(struct pci_dev *dev, unsigned int howmany, int rom)
323 {
324 	unsigned int pos, reg;
325 
326 	if (dev->non_compliant_bars)
327 		return;
328 
329 	/* Per PCIe r4.0, sec 9.3.4.1.11, the VF BARs are all RO Zero */
330 	if (dev->is_virtfn)
331 		return;
332 
333 	for (pos = 0; pos < howmany; pos++) {
334 		struct resource *res = &dev->resource[pos];
335 		reg = PCI_BASE_ADDRESS_0 + (pos << 2);
336 		pos += __pci_read_base(dev, pci_bar_unknown, res, reg);
337 	}
338 
339 	if (rom) {
340 		struct resource *res = &dev->resource[PCI_ROM_RESOURCE];
341 		dev->rom_base_reg = rom;
342 		res->flags = IORESOURCE_MEM | IORESOURCE_PREFETCH |
343 				IORESOURCE_READONLY | IORESOURCE_SIZEALIGN;
344 		__pci_read_base(dev, pci_bar_mem32, res, rom);
345 	}
346 }
347 
348 static void pci_read_bridge_windows(struct pci_dev *bridge)
349 {
350 	u16 io;
351 	u32 pmem, tmp;
352 
353 	pci_read_config_word(bridge, PCI_IO_BASE, &io);
354 	if (!io) {
355 		pci_write_config_word(bridge, PCI_IO_BASE, 0xe0f0);
356 		pci_read_config_word(bridge, PCI_IO_BASE, &io);
357 		pci_write_config_word(bridge, PCI_IO_BASE, 0x0);
358 	}
359 	if (io)
360 		bridge->io_window = 1;
361 
362 	/*
363 	 * DECchip 21050 pass 2 errata: the bridge may miss an address
364 	 * disconnect boundary by one PCI data phase.  Workaround: do not
365 	 * use prefetching on this device.
366 	 */
367 	if (bridge->vendor == PCI_VENDOR_ID_DEC && bridge->device == 0x0001)
368 		return;
369 
370 	pci_read_config_dword(bridge, PCI_PREF_MEMORY_BASE, &pmem);
371 	if (!pmem) {
372 		pci_write_config_dword(bridge, PCI_PREF_MEMORY_BASE,
373 					       0xffe0fff0);
374 		pci_read_config_dword(bridge, PCI_PREF_MEMORY_BASE, &pmem);
375 		pci_write_config_dword(bridge, PCI_PREF_MEMORY_BASE, 0x0);
376 	}
377 	if (!pmem)
378 		return;
379 
380 	bridge->pref_window = 1;
381 
382 	if ((pmem & PCI_PREF_RANGE_TYPE_MASK) == PCI_PREF_RANGE_TYPE_64) {
383 
384 		/*
385 		 * Bridge claims to have a 64-bit prefetchable memory
386 		 * window; verify that the upper bits are actually
387 		 * writable.
388 		 */
389 		pci_read_config_dword(bridge, PCI_PREF_BASE_UPPER32, &pmem);
390 		pci_write_config_dword(bridge, PCI_PREF_BASE_UPPER32,
391 				       0xffffffff);
392 		pci_read_config_dword(bridge, PCI_PREF_BASE_UPPER32, &tmp);
393 		pci_write_config_dword(bridge, PCI_PREF_BASE_UPPER32, pmem);
394 		if (tmp)
395 			bridge->pref_64_window = 1;
396 	}
397 }
398 
399 static void pci_read_bridge_io(struct pci_bus *child)
400 {
401 	struct pci_dev *dev = child->self;
402 	u8 io_base_lo, io_limit_lo;
403 	unsigned long io_mask, io_granularity, base, limit;
404 	struct pci_bus_region region;
405 	struct resource *res;
406 
407 	io_mask = PCI_IO_RANGE_MASK;
408 	io_granularity = 0x1000;
409 	if (dev->io_window_1k) {
410 		/* Support 1K I/O space granularity */
411 		io_mask = PCI_IO_1K_RANGE_MASK;
412 		io_granularity = 0x400;
413 	}
414 
415 	res = child->resource[0];
416 	pci_read_config_byte(dev, PCI_IO_BASE, &io_base_lo);
417 	pci_read_config_byte(dev, PCI_IO_LIMIT, &io_limit_lo);
418 	base = (io_base_lo & io_mask) << 8;
419 	limit = (io_limit_lo & io_mask) << 8;
420 
421 	if ((io_base_lo & PCI_IO_RANGE_TYPE_MASK) == PCI_IO_RANGE_TYPE_32) {
422 		u16 io_base_hi, io_limit_hi;
423 
424 		pci_read_config_word(dev, PCI_IO_BASE_UPPER16, &io_base_hi);
425 		pci_read_config_word(dev, PCI_IO_LIMIT_UPPER16, &io_limit_hi);
426 		base |= ((unsigned long) io_base_hi << 16);
427 		limit |= ((unsigned long) io_limit_hi << 16);
428 	}
429 
430 	if (base <= limit) {
431 		res->flags = (io_base_lo & PCI_IO_RANGE_TYPE_MASK) | IORESOURCE_IO;
432 		region.start = base;
433 		region.end = limit + io_granularity - 1;
434 		pcibios_bus_to_resource(dev->bus, res, &region);
435 		pci_info(dev, "  bridge window %pR\n", res);
436 	}
437 }
438 
439 static void pci_read_bridge_mmio(struct pci_bus *child)
440 {
441 	struct pci_dev *dev = child->self;
442 	u16 mem_base_lo, mem_limit_lo;
443 	unsigned long base, limit;
444 	struct pci_bus_region region;
445 	struct resource *res;
446 
447 	res = child->resource[1];
448 	pci_read_config_word(dev, PCI_MEMORY_BASE, &mem_base_lo);
449 	pci_read_config_word(dev, PCI_MEMORY_LIMIT, &mem_limit_lo);
450 	base = ((unsigned long) mem_base_lo & PCI_MEMORY_RANGE_MASK) << 16;
451 	limit = ((unsigned long) mem_limit_lo & PCI_MEMORY_RANGE_MASK) << 16;
452 	if (base <= limit) {
453 		res->flags = (mem_base_lo & PCI_MEMORY_RANGE_TYPE_MASK) | IORESOURCE_MEM;
454 		region.start = base;
455 		region.end = limit + 0xfffff;
456 		pcibios_bus_to_resource(dev->bus, res, &region);
457 		pci_info(dev, "  bridge window %pR\n", res);
458 	}
459 }
460 
461 static void pci_read_bridge_mmio_pref(struct pci_bus *child)
462 {
463 	struct pci_dev *dev = child->self;
464 	u16 mem_base_lo, mem_limit_lo;
465 	u64 base64, limit64;
466 	pci_bus_addr_t base, limit;
467 	struct pci_bus_region region;
468 	struct resource *res;
469 
470 	res = child->resource[2];
471 	pci_read_config_word(dev, PCI_PREF_MEMORY_BASE, &mem_base_lo);
472 	pci_read_config_word(dev, PCI_PREF_MEMORY_LIMIT, &mem_limit_lo);
473 	base64 = (mem_base_lo & PCI_PREF_RANGE_MASK) << 16;
474 	limit64 = (mem_limit_lo & PCI_PREF_RANGE_MASK) << 16;
475 
476 	if ((mem_base_lo & PCI_PREF_RANGE_TYPE_MASK) == PCI_PREF_RANGE_TYPE_64) {
477 		u32 mem_base_hi, mem_limit_hi;
478 
479 		pci_read_config_dword(dev, PCI_PREF_BASE_UPPER32, &mem_base_hi);
480 		pci_read_config_dword(dev, PCI_PREF_LIMIT_UPPER32, &mem_limit_hi);
481 
482 		/*
483 		 * Some bridges set the base > limit by default, and some
484 		 * (broken) BIOSes do not initialize them.  If we find
485 		 * this, just assume they are not being used.
486 		 */
487 		if (mem_base_hi <= mem_limit_hi) {
488 			base64 |= (u64) mem_base_hi << 32;
489 			limit64 |= (u64) mem_limit_hi << 32;
490 		}
491 	}
492 
493 	base = (pci_bus_addr_t) base64;
494 	limit = (pci_bus_addr_t) limit64;
495 
496 	if (base != base64) {
497 		pci_err(dev, "can't handle bridge window above 4GB (bus address %#010llx)\n",
498 			(unsigned long long) base64);
499 		return;
500 	}
501 
502 	if (base <= limit) {
503 		res->flags = (mem_base_lo & PCI_PREF_RANGE_TYPE_MASK) |
504 					 IORESOURCE_MEM | IORESOURCE_PREFETCH;
505 		if (res->flags & PCI_PREF_RANGE_TYPE_64)
506 			res->flags |= IORESOURCE_MEM_64;
507 		region.start = base;
508 		region.end = limit + 0xfffff;
509 		pcibios_bus_to_resource(dev->bus, res, &region);
510 		pci_info(dev, "  bridge window %pR\n", res);
511 	}
512 }
513 
514 void pci_read_bridge_bases(struct pci_bus *child)
515 {
516 	struct pci_dev *dev = child->self;
517 	struct resource *res;
518 	int i;
519 
520 	if (pci_is_root_bus(child))	/* It's a host bus, nothing to read */
521 		return;
522 
523 	pci_info(dev, "PCI bridge to %pR%s\n",
524 		 &child->busn_res,
525 		 dev->transparent ? " (subtractive decode)" : "");
526 
527 	pci_bus_remove_resources(child);
528 	for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++)
529 		child->resource[i] = &dev->resource[PCI_BRIDGE_RESOURCES+i];
530 
531 	pci_read_bridge_io(child);
532 	pci_read_bridge_mmio(child);
533 	pci_read_bridge_mmio_pref(child);
534 
535 	if (dev->transparent) {
536 		pci_bus_for_each_resource(child->parent, res, i) {
537 			if (res && res->flags) {
538 				pci_bus_add_resource(child, res,
539 						     PCI_SUBTRACTIVE_DECODE);
540 				pci_info(dev, "  bridge window %pR (subtractive decode)\n",
541 					   res);
542 			}
543 		}
544 	}
545 }
546 
547 static struct pci_bus *pci_alloc_bus(struct pci_bus *parent)
548 {
549 	struct pci_bus *b;
550 
551 	b = kzalloc(sizeof(*b), GFP_KERNEL);
552 	if (!b)
553 		return NULL;
554 
555 	INIT_LIST_HEAD(&b->node);
556 	INIT_LIST_HEAD(&b->children);
557 	INIT_LIST_HEAD(&b->devices);
558 	INIT_LIST_HEAD(&b->slots);
559 	INIT_LIST_HEAD(&b->resources);
560 	b->max_bus_speed = PCI_SPEED_UNKNOWN;
561 	b->cur_bus_speed = PCI_SPEED_UNKNOWN;
562 #ifdef CONFIG_PCI_DOMAINS_GENERIC
563 	if (parent)
564 		b->domain_nr = parent->domain_nr;
565 #endif
566 	return b;
567 }
568 
569 static void pci_release_host_bridge_dev(struct device *dev)
570 {
571 	struct pci_host_bridge *bridge = to_pci_host_bridge(dev);
572 
573 	if (bridge->release_fn)
574 		bridge->release_fn(bridge);
575 
576 	pci_free_resource_list(&bridge->windows);
577 	pci_free_resource_list(&bridge->dma_ranges);
578 	kfree(bridge);
579 }
580 
581 static void pci_init_host_bridge(struct pci_host_bridge *bridge)
582 {
583 	INIT_LIST_HEAD(&bridge->windows);
584 	INIT_LIST_HEAD(&bridge->dma_ranges);
585 
586 	/*
587 	 * We assume we can manage these PCIe features.  Some systems may
588 	 * reserve these for use by the platform itself, e.g., an ACPI BIOS
589 	 * may implement its own AER handling and use _OSC to prevent the
590 	 * OS from interfering.
591 	 */
592 	bridge->native_aer = 1;
593 	bridge->native_pcie_hotplug = 1;
594 	bridge->native_shpc_hotplug = 1;
595 	bridge->native_pme = 1;
596 	bridge->native_ltr = 1;
597 	bridge->native_dpc = 1;
598 	bridge->domain_nr = PCI_DOMAIN_NR_NOT_SET;
599 
600 	device_initialize(&bridge->dev);
601 }
602 
603 struct pci_host_bridge *pci_alloc_host_bridge(size_t priv)
604 {
605 	struct pci_host_bridge *bridge;
606 
607 	bridge = kzalloc(sizeof(*bridge) + priv, GFP_KERNEL);
608 	if (!bridge)
609 		return NULL;
610 
611 	pci_init_host_bridge(bridge);
612 	bridge->dev.release = pci_release_host_bridge_dev;
613 
614 	return bridge;
615 }
616 EXPORT_SYMBOL(pci_alloc_host_bridge);
617 
618 static void devm_pci_alloc_host_bridge_release(void *data)
619 {
620 	pci_free_host_bridge(data);
621 }
622 
623 struct pci_host_bridge *devm_pci_alloc_host_bridge(struct device *dev,
624 						   size_t priv)
625 {
626 	int ret;
627 	struct pci_host_bridge *bridge;
628 
629 	bridge = pci_alloc_host_bridge(priv);
630 	if (!bridge)
631 		return NULL;
632 
633 	bridge->dev.parent = dev;
634 
635 	ret = devm_add_action_or_reset(dev, devm_pci_alloc_host_bridge_release,
636 				       bridge);
637 	if (ret)
638 		return NULL;
639 
640 	ret = devm_of_pci_bridge_init(dev, bridge);
641 	if (ret)
642 		return NULL;
643 
644 	return bridge;
645 }
646 EXPORT_SYMBOL(devm_pci_alloc_host_bridge);
647 
648 void pci_free_host_bridge(struct pci_host_bridge *bridge)
649 {
650 	put_device(&bridge->dev);
651 }
652 EXPORT_SYMBOL(pci_free_host_bridge);
653 
654 /* Indexed by PCI_X_SSTATUS_FREQ (secondary bus mode and frequency) */
655 static const unsigned char pcix_bus_speed[] = {
656 	PCI_SPEED_UNKNOWN,		/* 0 */
657 	PCI_SPEED_66MHz_PCIX,		/* 1 */
658 	PCI_SPEED_100MHz_PCIX,		/* 2 */
659 	PCI_SPEED_133MHz_PCIX,		/* 3 */
660 	PCI_SPEED_UNKNOWN,		/* 4 */
661 	PCI_SPEED_66MHz_PCIX_ECC,	/* 5 */
662 	PCI_SPEED_100MHz_PCIX_ECC,	/* 6 */
663 	PCI_SPEED_133MHz_PCIX_ECC,	/* 7 */
664 	PCI_SPEED_UNKNOWN,		/* 8 */
665 	PCI_SPEED_66MHz_PCIX_266,	/* 9 */
666 	PCI_SPEED_100MHz_PCIX_266,	/* A */
667 	PCI_SPEED_133MHz_PCIX_266,	/* B */
668 	PCI_SPEED_UNKNOWN,		/* C */
669 	PCI_SPEED_66MHz_PCIX_533,	/* D */
670 	PCI_SPEED_100MHz_PCIX_533,	/* E */
671 	PCI_SPEED_133MHz_PCIX_533	/* F */
672 };
673 
674 /* Indexed by PCI_EXP_LNKCAP_SLS, PCI_EXP_LNKSTA_CLS */
675 const unsigned char pcie_link_speed[] = {
676 	PCI_SPEED_UNKNOWN,		/* 0 */
677 	PCIE_SPEED_2_5GT,		/* 1 */
678 	PCIE_SPEED_5_0GT,		/* 2 */
679 	PCIE_SPEED_8_0GT,		/* 3 */
680 	PCIE_SPEED_16_0GT,		/* 4 */
681 	PCIE_SPEED_32_0GT,		/* 5 */
682 	PCIE_SPEED_64_0GT,		/* 6 */
683 	PCI_SPEED_UNKNOWN,		/* 7 */
684 	PCI_SPEED_UNKNOWN,		/* 8 */
685 	PCI_SPEED_UNKNOWN,		/* 9 */
686 	PCI_SPEED_UNKNOWN,		/* A */
687 	PCI_SPEED_UNKNOWN,		/* B */
688 	PCI_SPEED_UNKNOWN,		/* C */
689 	PCI_SPEED_UNKNOWN,		/* D */
690 	PCI_SPEED_UNKNOWN,		/* E */
691 	PCI_SPEED_UNKNOWN		/* F */
692 };
693 EXPORT_SYMBOL_GPL(pcie_link_speed);
694 
695 const char *pci_speed_string(enum pci_bus_speed speed)
696 {
697 	/* Indexed by the pci_bus_speed enum */
698 	static const char *speed_strings[] = {
699 	    "33 MHz PCI",		/* 0x00 */
700 	    "66 MHz PCI",		/* 0x01 */
701 	    "66 MHz PCI-X",		/* 0x02 */
702 	    "100 MHz PCI-X",		/* 0x03 */
703 	    "133 MHz PCI-X",		/* 0x04 */
704 	    NULL,			/* 0x05 */
705 	    NULL,			/* 0x06 */
706 	    NULL,			/* 0x07 */
707 	    NULL,			/* 0x08 */
708 	    "66 MHz PCI-X 266",		/* 0x09 */
709 	    "100 MHz PCI-X 266",	/* 0x0a */
710 	    "133 MHz PCI-X 266",	/* 0x0b */
711 	    "Unknown AGP",		/* 0x0c */
712 	    "1x AGP",			/* 0x0d */
713 	    "2x AGP",			/* 0x0e */
714 	    "4x AGP",			/* 0x0f */
715 	    "8x AGP",			/* 0x10 */
716 	    "66 MHz PCI-X 533",		/* 0x11 */
717 	    "100 MHz PCI-X 533",	/* 0x12 */
718 	    "133 MHz PCI-X 533",	/* 0x13 */
719 	    "2.5 GT/s PCIe",		/* 0x14 */
720 	    "5.0 GT/s PCIe",		/* 0x15 */
721 	    "8.0 GT/s PCIe",		/* 0x16 */
722 	    "16.0 GT/s PCIe",		/* 0x17 */
723 	    "32.0 GT/s PCIe",		/* 0x18 */
724 	    "64.0 GT/s PCIe",		/* 0x19 */
725 	};
726 
727 	if (speed < ARRAY_SIZE(speed_strings))
728 		return speed_strings[speed];
729 	return "Unknown";
730 }
731 EXPORT_SYMBOL_GPL(pci_speed_string);
732 
733 void pcie_update_link_speed(struct pci_bus *bus, u16 linksta)
734 {
735 	bus->cur_bus_speed = pcie_link_speed[linksta & PCI_EXP_LNKSTA_CLS];
736 }
737 EXPORT_SYMBOL_GPL(pcie_update_link_speed);
738 
739 static unsigned char agp_speeds[] = {
740 	AGP_UNKNOWN,
741 	AGP_1X,
742 	AGP_2X,
743 	AGP_4X,
744 	AGP_8X
745 };
746 
747 static enum pci_bus_speed agp_speed(int agp3, int agpstat)
748 {
749 	int index = 0;
750 
751 	if (agpstat & 4)
752 		index = 3;
753 	else if (agpstat & 2)
754 		index = 2;
755 	else if (agpstat & 1)
756 		index = 1;
757 	else
758 		goto out;
759 
760 	if (agp3) {
761 		index += 2;
762 		if (index == 5)
763 			index = 0;
764 	}
765 
766  out:
767 	return agp_speeds[index];
768 }
769 
770 static void pci_set_bus_speed(struct pci_bus *bus)
771 {
772 	struct pci_dev *bridge = bus->self;
773 	int pos;
774 
775 	pos = pci_find_capability(bridge, PCI_CAP_ID_AGP);
776 	if (!pos)
777 		pos = pci_find_capability(bridge, PCI_CAP_ID_AGP3);
778 	if (pos) {
779 		u32 agpstat, agpcmd;
780 
781 		pci_read_config_dword(bridge, pos + PCI_AGP_STATUS, &agpstat);
782 		bus->max_bus_speed = agp_speed(agpstat & 8, agpstat & 7);
783 
784 		pci_read_config_dword(bridge, pos + PCI_AGP_COMMAND, &agpcmd);
785 		bus->cur_bus_speed = agp_speed(agpstat & 8, agpcmd & 7);
786 	}
787 
788 	pos = pci_find_capability(bridge, PCI_CAP_ID_PCIX);
789 	if (pos) {
790 		u16 status;
791 		enum pci_bus_speed max;
792 
793 		pci_read_config_word(bridge, pos + PCI_X_BRIDGE_SSTATUS,
794 				     &status);
795 
796 		if (status & PCI_X_SSTATUS_533MHZ) {
797 			max = PCI_SPEED_133MHz_PCIX_533;
798 		} else if (status & PCI_X_SSTATUS_266MHZ) {
799 			max = PCI_SPEED_133MHz_PCIX_266;
800 		} else if (status & PCI_X_SSTATUS_133MHZ) {
801 			if ((status & PCI_X_SSTATUS_VERS) == PCI_X_SSTATUS_V2)
802 				max = PCI_SPEED_133MHz_PCIX_ECC;
803 			else
804 				max = PCI_SPEED_133MHz_PCIX;
805 		} else {
806 			max = PCI_SPEED_66MHz_PCIX;
807 		}
808 
809 		bus->max_bus_speed = max;
810 		bus->cur_bus_speed = pcix_bus_speed[
811 			(status & PCI_X_SSTATUS_FREQ) >> 6];
812 
813 		return;
814 	}
815 
816 	if (pci_is_pcie(bridge)) {
817 		u32 linkcap;
818 		u16 linksta;
819 
820 		pcie_capability_read_dword(bridge, PCI_EXP_LNKCAP, &linkcap);
821 		bus->max_bus_speed = pcie_link_speed[linkcap & PCI_EXP_LNKCAP_SLS];
822 		bridge->link_active_reporting = !!(linkcap & PCI_EXP_LNKCAP_DLLLARC);
823 
824 		pcie_capability_read_word(bridge, PCI_EXP_LNKSTA, &linksta);
825 		pcie_update_link_speed(bus, linksta);
826 	}
827 }
828 
829 static struct irq_domain *pci_host_bridge_msi_domain(struct pci_bus *bus)
830 {
831 	struct irq_domain *d;
832 
833 	/* If the host bridge driver sets a MSI domain of the bridge, use it */
834 	d = dev_get_msi_domain(bus->bridge);
835 
836 	/*
837 	 * Any firmware interface that can resolve the msi_domain
838 	 * should be called from here.
839 	 */
840 	if (!d)
841 		d = pci_host_bridge_of_msi_domain(bus);
842 	if (!d)
843 		d = pci_host_bridge_acpi_msi_domain(bus);
844 
845 #ifdef CONFIG_PCI_MSI_IRQ_DOMAIN
846 	/*
847 	 * If no IRQ domain was found via the OF tree, try looking it up
848 	 * directly through the fwnode_handle.
849 	 */
850 	if (!d) {
851 		struct fwnode_handle *fwnode = pci_root_bus_fwnode(bus);
852 
853 		if (fwnode)
854 			d = irq_find_matching_fwnode(fwnode,
855 						     DOMAIN_BUS_PCI_MSI);
856 	}
857 #endif
858 
859 	return d;
860 }
861 
862 static void pci_set_bus_msi_domain(struct pci_bus *bus)
863 {
864 	struct irq_domain *d;
865 	struct pci_bus *b;
866 
867 	/*
868 	 * The bus can be a root bus, a subordinate bus, or a virtual bus
869 	 * created by an SR-IOV device.  Walk up to the first bridge device
870 	 * found or derive the domain from the host bridge.
871 	 */
872 	for (b = bus, d = NULL; !d && !pci_is_root_bus(b); b = b->parent) {
873 		if (b->self)
874 			d = dev_get_msi_domain(&b->self->dev);
875 	}
876 
877 	if (!d)
878 		d = pci_host_bridge_msi_domain(b);
879 
880 	dev_set_msi_domain(&bus->dev, d);
881 }
882 
883 static int pci_register_host_bridge(struct pci_host_bridge *bridge)
884 {
885 	struct device *parent = bridge->dev.parent;
886 	struct resource_entry *window, *next, *n;
887 	struct pci_bus *bus, *b;
888 	resource_size_t offset, next_offset;
889 	LIST_HEAD(resources);
890 	struct resource *res, *next_res;
891 	char addr[64], *fmt;
892 	const char *name;
893 	int err;
894 
895 	bus = pci_alloc_bus(NULL);
896 	if (!bus)
897 		return -ENOMEM;
898 
899 	bridge->bus = bus;
900 
901 	bus->sysdata = bridge->sysdata;
902 	bus->ops = bridge->ops;
903 	bus->number = bus->busn_res.start = bridge->busnr;
904 #ifdef CONFIG_PCI_DOMAINS_GENERIC
905 	if (bridge->domain_nr == PCI_DOMAIN_NR_NOT_SET)
906 		bus->domain_nr = pci_bus_find_domain_nr(bus, parent);
907 	else
908 		bus->domain_nr = bridge->domain_nr;
909 #endif
910 
911 	b = pci_find_bus(pci_domain_nr(bus), bridge->busnr);
912 	if (b) {
913 		/* Ignore it if we already got here via a different bridge */
914 		dev_dbg(&b->dev, "bus already known\n");
915 		err = -EEXIST;
916 		goto free;
917 	}
918 
919 	dev_set_name(&bridge->dev, "pci%04x:%02x", pci_domain_nr(bus),
920 		     bridge->busnr);
921 
922 	err = pcibios_root_bridge_prepare(bridge);
923 	if (err)
924 		goto free;
925 
926 	/* Temporarily move resources off the list */
927 	list_splice_init(&bridge->windows, &resources);
928 	err = device_add(&bridge->dev);
929 	if (err) {
930 		put_device(&bridge->dev);
931 		goto free;
932 	}
933 	bus->bridge = get_device(&bridge->dev);
934 	device_enable_async_suspend(bus->bridge);
935 	pci_set_bus_of_node(bus);
936 	pci_set_bus_msi_domain(bus);
937 	if (bridge->msi_domain && !dev_get_msi_domain(&bus->dev) &&
938 	    !pci_host_of_has_msi_map(parent))
939 		bus->bus_flags |= PCI_BUS_FLAGS_NO_MSI;
940 
941 	if (!parent)
942 		set_dev_node(bus->bridge, pcibus_to_node(bus));
943 
944 	bus->dev.class = &pcibus_class;
945 	bus->dev.parent = bus->bridge;
946 
947 	dev_set_name(&bus->dev, "%04x:%02x", pci_domain_nr(bus), bus->number);
948 	name = dev_name(&bus->dev);
949 
950 	err = device_register(&bus->dev);
951 	if (err)
952 		goto unregister;
953 
954 	pcibios_add_bus(bus);
955 
956 	if (bus->ops->add_bus) {
957 		err = bus->ops->add_bus(bus);
958 		if (WARN_ON(err < 0))
959 			dev_err(&bus->dev, "failed to add bus: %d\n", err);
960 	}
961 
962 	/* Create legacy_io and legacy_mem files for this bus */
963 	pci_create_legacy_files(bus);
964 
965 	if (parent)
966 		dev_info(parent, "PCI host bridge to bus %s\n", name);
967 	else
968 		pr_info("PCI host bridge to bus %s\n", name);
969 
970 	if (nr_node_ids > 1 && pcibus_to_node(bus) == NUMA_NO_NODE)
971 		dev_warn(&bus->dev, "Unknown NUMA node; performance will be reduced\n");
972 
973 	/* Coalesce contiguous windows */
974 	resource_list_for_each_entry_safe(window, n, &resources) {
975 		if (list_is_last(&window->node, &resources))
976 			break;
977 
978 		next = list_next_entry(window, node);
979 		offset = window->offset;
980 		res = window->res;
981 		next_offset = next->offset;
982 		next_res = next->res;
983 
984 		if (res->flags != next_res->flags || offset != next_offset)
985 			continue;
986 
987 		if (res->end + 1 == next_res->start) {
988 			next_res->start = res->start;
989 			res->flags = res->start = res->end = 0;
990 		}
991 	}
992 
993 	/* Add initial resources to the bus */
994 	resource_list_for_each_entry_safe(window, n, &resources) {
995 		offset = window->offset;
996 		res = window->res;
997 		if (!res->end)
998 			continue;
999 
1000 		list_move_tail(&window->node, &bridge->windows);
1001 
1002 		if (res->flags & IORESOURCE_BUS)
1003 			pci_bus_insert_busn_res(bus, bus->number, res->end);
1004 		else
1005 			pci_bus_add_resource(bus, res, 0);
1006 
1007 		if (offset) {
1008 			if (resource_type(res) == IORESOURCE_IO)
1009 				fmt = " (bus address [%#06llx-%#06llx])";
1010 			else
1011 				fmt = " (bus address [%#010llx-%#010llx])";
1012 
1013 			snprintf(addr, sizeof(addr), fmt,
1014 				 (unsigned long long)(res->start - offset),
1015 				 (unsigned long long)(res->end - offset));
1016 		} else
1017 			addr[0] = '\0';
1018 
1019 		dev_info(&bus->dev, "root bus resource %pR%s\n", res, addr);
1020 	}
1021 
1022 	down_write(&pci_bus_sem);
1023 	list_add_tail(&bus->node, &pci_root_buses);
1024 	up_write(&pci_bus_sem);
1025 
1026 	return 0;
1027 
1028 unregister:
1029 	put_device(&bridge->dev);
1030 	device_del(&bridge->dev);
1031 
1032 free:
1033 	kfree(bus);
1034 	return err;
1035 }
1036 
1037 static bool pci_bridge_child_ext_cfg_accessible(struct pci_dev *bridge)
1038 {
1039 	int pos;
1040 	u32 status;
1041 
1042 	/*
1043 	 * If extended config space isn't accessible on a bridge's primary
1044 	 * bus, we certainly can't access it on the secondary bus.
1045 	 */
1046 	if (bridge->bus->bus_flags & PCI_BUS_FLAGS_NO_EXTCFG)
1047 		return false;
1048 
1049 	/*
1050 	 * PCIe Root Ports and switch ports are PCIe on both sides, so if
1051 	 * extended config space is accessible on the primary, it's also
1052 	 * accessible on the secondary.
1053 	 */
1054 	if (pci_is_pcie(bridge) &&
1055 	    (pci_pcie_type(bridge) == PCI_EXP_TYPE_ROOT_PORT ||
1056 	     pci_pcie_type(bridge) == PCI_EXP_TYPE_UPSTREAM ||
1057 	     pci_pcie_type(bridge) == PCI_EXP_TYPE_DOWNSTREAM))
1058 		return true;
1059 
1060 	/*
1061 	 * For the other bridge types:
1062 	 *   - PCI-to-PCI bridges
1063 	 *   - PCIe-to-PCI/PCI-X forward bridges
1064 	 *   - PCI/PCI-X-to-PCIe reverse bridges
1065 	 * extended config space on the secondary side is only accessible
1066 	 * if the bridge supports PCI-X Mode 2.
1067 	 */
1068 	pos = pci_find_capability(bridge, PCI_CAP_ID_PCIX);
1069 	if (!pos)
1070 		return false;
1071 
1072 	pci_read_config_dword(bridge, pos + PCI_X_STATUS, &status);
1073 	return status & (PCI_X_STATUS_266MHZ | PCI_X_STATUS_533MHZ);
1074 }
1075 
1076 static struct pci_bus *pci_alloc_child_bus(struct pci_bus *parent,
1077 					   struct pci_dev *bridge, int busnr)
1078 {
1079 	struct pci_bus *child;
1080 	struct pci_host_bridge *host;
1081 	int i;
1082 	int ret;
1083 
1084 	/* Allocate a new bus and inherit stuff from the parent */
1085 	child = pci_alloc_bus(parent);
1086 	if (!child)
1087 		return NULL;
1088 
1089 	child->parent = parent;
1090 	child->sysdata = parent->sysdata;
1091 	child->bus_flags = parent->bus_flags;
1092 
1093 	host = pci_find_host_bridge(parent);
1094 	if (host->child_ops)
1095 		child->ops = host->child_ops;
1096 	else
1097 		child->ops = parent->ops;
1098 
1099 	/*
1100 	 * Initialize some portions of the bus device, but don't register
1101 	 * it now as the parent is not properly set up yet.
1102 	 */
1103 	child->dev.class = &pcibus_class;
1104 	dev_set_name(&child->dev, "%04x:%02x", pci_domain_nr(child), busnr);
1105 
1106 	/* Set up the primary, secondary and subordinate bus numbers */
1107 	child->number = child->busn_res.start = busnr;
1108 	child->primary = parent->busn_res.start;
1109 	child->busn_res.end = 0xff;
1110 
1111 	if (!bridge) {
1112 		child->dev.parent = parent->bridge;
1113 		goto add_dev;
1114 	}
1115 
1116 	child->self = bridge;
1117 	child->bridge = get_device(&bridge->dev);
1118 	child->dev.parent = child->bridge;
1119 	pci_set_bus_of_node(child);
1120 	pci_set_bus_speed(child);
1121 
1122 	/*
1123 	 * Check whether extended config space is accessible on the child
1124 	 * bus.  Note that we currently assume it is always accessible on
1125 	 * the root bus.
1126 	 */
1127 	if (!pci_bridge_child_ext_cfg_accessible(bridge)) {
1128 		child->bus_flags |= PCI_BUS_FLAGS_NO_EXTCFG;
1129 		pci_info(child, "extended config space not accessible\n");
1130 	}
1131 
1132 	/* Set up default resource pointers and names */
1133 	for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++) {
1134 		child->resource[i] = &bridge->resource[PCI_BRIDGE_RESOURCES+i];
1135 		child->resource[i]->name = child->name;
1136 	}
1137 	bridge->subordinate = child;
1138 
1139 add_dev:
1140 	pci_set_bus_msi_domain(child);
1141 	ret = device_register(&child->dev);
1142 	WARN_ON(ret < 0);
1143 
1144 	pcibios_add_bus(child);
1145 
1146 	if (child->ops->add_bus) {
1147 		ret = child->ops->add_bus(child);
1148 		if (WARN_ON(ret < 0))
1149 			dev_err(&child->dev, "failed to add bus: %d\n", ret);
1150 	}
1151 
1152 	/* Create legacy_io and legacy_mem files for this bus */
1153 	pci_create_legacy_files(child);
1154 
1155 	return child;
1156 }
1157 
1158 struct pci_bus *pci_add_new_bus(struct pci_bus *parent, struct pci_dev *dev,
1159 				int busnr)
1160 {
1161 	struct pci_bus *child;
1162 
1163 	child = pci_alloc_child_bus(parent, dev, busnr);
1164 	if (child) {
1165 		down_write(&pci_bus_sem);
1166 		list_add_tail(&child->node, &parent->children);
1167 		up_write(&pci_bus_sem);
1168 	}
1169 	return child;
1170 }
1171 EXPORT_SYMBOL(pci_add_new_bus);
1172 
1173 static void pci_enable_crs(struct pci_dev *pdev)
1174 {
1175 	u16 root_cap = 0;
1176 
1177 	/* Enable CRS Software Visibility if supported */
1178 	pcie_capability_read_word(pdev, PCI_EXP_RTCAP, &root_cap);
1179 	if (root_cap & PCI_EXP_RTCAP_CRSVIS)
1180 		pcie_capability_set_word(pdev, PCI_EXP_RTCTL,
1181 					 PCI_EXP_RTCTL_CRSSVE);
1182 }
1183 
1184 static unsigned int pci_scan_child_bus_extend(struct pci_bus *bus,
1185 					      unsigned int available_buses);
1186 /**
1187  * pci_ea_fixed_busnrs() - Read fixed Secondary and Subordinate bus
1188  * numbers from EA capability.
1189  * @dev: Bridge
1190  * @sec: updated with secondary bus number from EA
1191  * @sub: updated with subordinate bus number from EA
1192  *
1193  * If @dev is a bridge with EA capability that specifies valid secondary
1194  * and subordinate bus numbers, return true with the bus numbers in @sec
1195  * and @sub.  Otherwise return false.
1196  */
1197 static bool pci_ea_fixed_busnrs(struct pci_dev *dev, u8 *sec, u8 *sub)
1198 {
1199 	int ea, offset;
1200 	u32 dw;
1201 	u8 ea_sec, ea_sub;
1202 
1203 	if (dev->hdr_type != PCI_HEADER_TYPE_BRIDGE)
1204 		return false;
1205 
1206 	/* find PCI EA capability in list */
1207 	ea = pci_find_capability(dev, PCI_CAP_ID_EA);
1208 	if (!ea)
1209 		return false;
1210 
1211 	offset = ea + PCI_EA_FIRST_ENT;
1212 	pci_read_config_dword(dev, offset, &dw);
1213 	ea_sec =  dw & PCI_EA_SEC_BUS_MASK;
1214 	ea_sub = (dw & PCI_EA_SUB_BUS_MASK) >> PCI_EA_SUB_BUS_SHIFT;
1215 	if (ea_sec  == 0 || ea_sub < ea_sec)
1216 		return false;
1217 
1218 	*sec = ea_sec;
1219 	*sub = ea_sub;
1220 	return true;
1221 }
1222 
1223 /*
1224  * pci_scan_bridge_extend() - Scan buses behind a bridge
1225  * @bus: Parent bus the bridge is on
1226  * @dev: Bridge itself
1227  * @max: Starting subordinate number of buses behind this bridge
1228  * @available_buses: Total number of buses available for this bridge and
1229  *		     the devices below. After the minimal bus space has
1230  *		     been allocated the remaining buses will be
1231  *		     distributed equally between hotplug-capable bridges.
1232  * @pass: Either %0 (scan already configured bridges) or %1 (scan bridges
1233  *        that need to be reconfigured.
1234  *
1235  * If it's a bridge, configure it and scan the bus behind it.
1236  * For CardBus bridges, we don't scan behind as the devices will
1237  * be handled by the bridge driver itself.
1238  *
1239  * We need to process bridges in two passes -- first we scan those
1240  * already configured by the BIOS and after we are done with all of
1241  * them, we proceed to assigning numbers to the remaining buses in
1242  * order to avoid overlaps between old and new bus numbers.
1243  *
1244  * Return: New subordinate number covering all buses behind this bridge.
1245  */
1246 static int pci_scan_bridge_extend(struct pci_bus *bus, struct pci_dev *dev,
1247 				  int max, unsigned int available_buses,
1248 				  int pass)
1249 {
1250 	struct pci_bus *child;
1251 	int is_cardbus = (dev->hdr_type == PCI_HEADER_TYPE_CARDBUS);
1252 	u32 buses, i, j = 0;
1253 	u16 bctl;
1254 	u8 primary, secondary, subordinate;
1255 	int broken = 0;
1256 	bool fixed_buses;
1257 	u8 fixed_sec, fixed_sub;
1258 	int next_busnr;
1259 
1260 	/*
1261 	 * Make sure the bridge is powered on to be able to access config
1262 	 * space of devices below it.
1263 	 */
1264 	pm_runtime_get_sync(&dev->dev);
1265 
1266 	pci_read_config_dword(dev, PCI_PRIMARY_BUS, &buses);
1267 	primary = buses & 0xFF;
1268 	secondary = (buses >> 8) & 0xFF;
1269 	subordinate = (buses >> 16) & 0xFF;
1270 
1271 	pci_dbg(dev, "scanning [bus %02x-%02x] behind bridge, pass %d\n",
1272 		secondary, subordinate, pass);
1273 
1274 	if (!primary && (primary != bus->number) && secondary && subordinate) {
1275 		pci_warn(dev, "Primary bus is hard wired to 0\n");
1276 		primary = bus->number;
1277 	}
1278 
1279 	/* Check if setup is sensible at all */
1280 	if (!pass &&
1281 	    (primary != bus->number || secondary <= bus->number ||
1282 	     secondary > subordinate)) {
1283 		pci_info(dev, "bridge configuration invalid ([bus %02x-%02x]), reconfiguring\n",
1284 			 secondary, subordinate);
1285 		broken = 1;
1286 	}
1287 
1288 	/*
1289 	 * Disable Master-Abort Mode during probing to avoid reporting of
1290 	 * bus errors in some architectures.
1291 	 */
1292 	pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &bctl);
1293 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL,
1294 			      bctl & ~PCI_BRIDGE_CTL_MASTER_ABORT);
1295 
1296 	pci_enable_crs(dev);
1297 
1298 	if ((secondary || subordinate) && !pcibios_assign_all_busses() &&
1299 	    !is_cardbus && !broken) {
1300 		unsigned int cmax;
1301 
1302 		/*
1303 		 * Bus already configured by firmware, process it in the
1304 		 * first pass and just note the configuration.
1305 		 */
1306 		if (pass)
1307 			goto out;
1308 
1309 		/*
1310 		 * The bus might already exist for two reasons: Either we
1311 		 * are rescanning the bus or the bus is reachable through
1312 		 * more than one bridge. The second case can happen with
1313 		 * the i450NX chipset.
1314 		 */
1315 		child = pci_find_bus(pci_domain_nr(bus), secondary);
1316 		if (!child) {
1317 			child = pci_add_new_bus(bus, dev, secondary);
1318 			if (!child)
1319 				goto out;
1320 			child->primary = primary;
1321 			pci_bus_insert_busn_res(child, secondary, subordinate);
1322 			child->bridge_ctl = bctl;
1323 		}
1324 
1325 		cmax = pci_scan_child_bus(child);
1326 		if (cmax > subordinate)
1327 			pci_warn(dev, "bridge has subordinate %02x but max busn %02x\n",
1328 				 subordinate, cmax);
1329 
1330 		/* Subordinate should equal child->busn_res.end */
1331 		if (subordinate > max)
1332 			max = subordinate;
1333 	} else {
1334 
1335 		/*
1336 		 * We need to assign a number to this bus which we always
1337 		 * do in the second pass.
1338 		 */
1339 		if (!pass) {
1340 			if (pcibios_assign_all_busses() || broken || is_cardbus)
1341 
1342 				/*
1343 				 * Temporarily disable forwarding of the
1344 				 * configuration cycles on all bridges in
1345 				 * this bus segment to avoid possible
1346 				 * conflicts in the second pass between two
1347 				 * bridges programmed with overlapping bus
1348 				 * ranges.
1349 				 */
1350 				pci_write_config_dword(dev, PCI_PRIMARY_BUS,
1351 						       buses & ~0xffffff);
1352 			goto out;
1353 		}
1354 
1355 		/* Clear errors */
1356 		pci_write_config_word(dev, PCI_STATUS, 0xffff);
1357 
1358 		/* Read bus numbers from EA Capability (if present) */
1359 		fixed_buses = pci_ea_fixed_busnrs(dev, &fixed_sec, &fixed_sub);
1360 		if (fixed_buses)
1361 			next_busnr = fixed_sec;
1362 		else
1363 			next_busnr = max + 1;
1364 
1365 		/*
1366 		 * Prevent assigning a bus number that already exists.
1367 		 * This can happen when a bridge is hot-plugged, so in this
1368 		 * case we only re-scan this bus.
1369 		 */
1370 		child = pci_find_bus(pci_domain_nr(bus), next_busnr);
1371 		if (!child) {
1372 			child = pci_add_new_bus(bus, dev, next_busnr);
1373 			if (!child)
1374 				goto out;
1375 			pci_bus_insert_busn_res(child, next_busnr,
1376 						bus->busn_res.end);
1377 		}
1378 		max++;
1379 		if (available_buses)
1380 			available_buses--;
1381 
1382 		buses = (buses & 0xff000000)
1383 		      | ((unsigned int)(child->primary)     <<  0)
1384 		      | ((unsigned int)(child->busn_res.start)   <<  8)
1385 		      | ((unsigned int)(child->busn_res.end) << 16);
1386 
1387 		/*
1388 		 * yenta.c forces a secondary latency timer of 176.
1389 		 * Copy that behaviour here.
1390 		 */
1391 		if (is_cardbus) {
1392 			buses &= ~0xff000000;
1393 			buses |= CARDBUS_LATENCY_TIMER << 24;
1394 		}
1395 
1396 		/* We need to blast all three values with a single write */
1397 		pci_write_config_dword(dev, PCI_PRIMARY_BUS, buses);
1398 
1399 		if (!is_cardbus) {
1400 			child->bridge_ctl = bctl;
1401 			max = pci_scan_child_bus_extend(child, available_buses);
1402 		} else {
1403 
1404 			/*
1405 			 * For CardBus bridges, we leave 4 bus numbers as
1406 			 * cards with a PCI-to-PCI bridge can be inserted
1407 			 * later.
1408 			 */
1409 			for (i = 0; i < CARDBUS_RESERVE_BUSNR; i++) {
1410 				struct pci_bus *parent = bus;
1411 				if (pci_find_bus(pci_domain_nr(bus),
1412 							max+i+1))
1413 					break;
1414 				while (parent->parent) {
1415 					if ((!pcibios_assign_all_busses()) &&
1416 					    (parent->busn_res.end > max) &&
1417 					    (parent->busn_res.end <= max+i)) {
1418 						j = 1;
1419 					}
1420 					parent = parent->parent;
1421 				}
1422 				if (j) {
1423 
1424 					/*
1425 					 * Often, there are two CardBus
1426 					 * bridges -- try to leave one
1427 					 * valid bus number for each one.
1428 					 */
1429 					i /= 2;
1430 					break;
1431 				}
1432 			}
1433 			max += i;
1434 		}
1435 
1436 		/*
1437 		 * Set subordinate bus number to its real value.
1438 		 * If fixed subordinate bus number exists from EA
1439 		 * capability then use it.
1440 		 */
1441 		if (fixed_buses)
1442 			max = fixed_sub;
1443 		pci_bus_update_busn_res_end(child, max);
1444 		pci_write_config_byte(dev, PCI_SUBORDINATE_BUS, max);
1445 	}
1446 
1447 	sprintf(child->name,
1448 		(is_cardbus ? "PCI CardBus %04x:%02x" : "PCI Bus %04x:%02x"),
1449 		pci_domain_nr(bus), child->number);
1450 
1451 	/* Check that all devices are accessible */
1452 	while (bus->parent) {
1453 		if ((child->busn_res.end > bus->busn_res.end) ||
1454 		    (child->number > bus->busn_res.end) ||
1455 		    (child->number < bus->number) ||
1456 		    (child->busn_res.end < bus->number)) {
1457 			dev_info(&dev->dev, "devices behind bridge are unusable because %pR cannot be assigned for them\n",
1458 				 &child->busn_res);
1459 			break;
1460 		}
1461 		bus = bus->parent;
1462 	}
1463 
1464 out:
1465 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL, bctl);
1466 
1467 	pm_runtime_put(&dev->dev);
1468 
1469 	return max;
1470 }
1471 
1472 /*
1473  * pci_scan_bridge() - Scan buses behind a bridge
1474  * @bus: Parent bus the bridge is on
1475  * @dev: Bridge itself
1476  * @max: Starting subordinate number of buses behind this bridge
1477  * @pass: Either %0 (scan already configured bridges) or %1 (scan bridges
1478  *        that need to be reconfigured.
1479  *
1480  * If it's a bridge, configure it and scan the bus behind it.
1481  * For CardBus bridges, we don't scan behind as the devices will
1482  * be handled by the bridge driver itself.
1483  *
1484  * We need to process bridges in two passes -- first we scan those
1485  * already configured by the BIOS and after we are done with all of
1486  * them, we proceed to assigning numbers to the remaining buses in
1487  * order to avoid overlaps between old and new bus numbers.
1488  *
1489  * Return: New subordinate number covering all buses behind this bridge.
1490  */
1491 int pci_scan_bridge(struct pci_bus *bus, struct pci_dev *dev, int max, int pass)
1492 {
1493 	return pci_scan_bridge_extend(bus, dev, max, 0, pass);
1494 }
1495 EXPORT_SYMBOL(pci_scan_bridge);
1496 
1497 /*
1498  * Read interrupt line and base address registers.
1499  * The architecture-dependent code can tweak these, of course.
1500  */
1501 static void pci_read_irq(struct pci_dev *dev)
1502 {
1503 	unsigned char irq;
1504 
1505 	/* VFs are not allowed to use INTx, so skip the config reads */
1506 	if (dev->is_virtfn) {
1507 		dev->pin = 0;
1508 		dev->irq = 0;
1509 		return;
1510 	}
1511 
1512 	pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &irq);
1513 	dev->pin = irq;
1514 	if (irq)
1515 		pci_read_config_byte(dev, PCI_INTERRUPT_LINE, &irq);
1516 	dev->irq = irq;
1517 }
1518 
1519 void set_pcie_port_type(struct pci_dev *pdev)
1520 {
1521 	int pos;
1522 	u16 reg16;
1523 	int type;
1524 	struct pci_dev *parent;
1525 
1526 	pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
1527 	if (!pos)
1528 		return;
1529 
1530 	pdev->pcie_cap = pos;
1531 	pci_read_config_word(pdev, pos + PCI_EXP_FLAGS, &reg16);
1532 	pdev->pcie_flags_reg = reg16;
1533 	pci_read_config_dword(pdev, pos + PCI_EXP_DEVCAP, &pdev->devcap);
1534 	pdev->pcie_mpss = FIELD_GET(PCI_EXP_DEVCAP_PAYLOAD, pdev->devcap);
1535 
1536 	parent = pci_upstream_bridge(pdev);
1537 	if (!parent)
1538 		return;
1539 
1540 	/*
1541 	 * Some systems do not identify their upstream/downstream ports
1542 	 * correctly so detect impossible configurations here and correct
1543 	 * the port type accordingly.
1544 	 */
1545 	type = pci_pcie_type(pdev);
1546 	if (type == PCI_EXP_TYPE_DOWNSTREAM) {
1547 		/*
1548 		 * If pdev claims to be downstream port but the parent
1549 		 * device is also downstream port assume pdev is actually
1550 		 * upstream port.
1551 		 */
1552 		if (pcie_downstream_port(parent)) {
1553 			pci_info(pdev, "claims to be downstream port but is acting as upstream port, correcting type\n");
1554 			pdev->pcie_flags_reg &= ~PCI_EXP_FLAGS_TYPE;
1555 			pdev->pcie_flags_reg |= PCI_EXP_TYPE_UPSTREAM;
1556 		}
1557 	} else if (type == PCI_EXP_TYPE_UPSTREAM) {
1558 		/*
1559 		 * If pdev claims to be upstream port but the parent
1560 		 * device is also upstream port assume pdev is actually
1561 		 * downstream port.
1562 		 */
1563 		if (pci_pcie_type(parent) == PCI_EXP_TYPE_UPSTREAM) {
1564 			pci_info(pdev, "claims to be upstream port but is acting as downstream port, correcting type\n");
1565 			pdev->pcie_flags_reg &= ~PCI_EXP_FLAGS_TYPE;
1566 			pdev->pcie_flags_reg |= PCI_EXP_TYPE_DOWNSTREAM;
1567 		}
1568 	}
1569 }
1570 
1571 void set_pcie_hotplug_bridge(struct pci_dev *pdev)
1572 {
1573 	u32 reg32;
1574 
1575 	pcie_capability_read_dword(pdev, PCI_EXP_SLTCAP, &reg32);
1576 	if (reg32 & PCI_EXP_SLTCAP_HPC)
1577 		pdev->is_hotplug_bridge = 1;
1578 }
1579 
1580 static void set_pcie_thunderbolt(struct pci_dev *dev)
1581 {
1582 	int vsec = 0;
1583 	u32 header;
1584 
1585 	while ((vsec = pci_find_next_ext_capability(dev, vsec,
1586 						    PCI_EXT_CAP_ID_VNDR))) {
1587 		pci_read_config_dword(dev, vsec + PCI_VNDR_HEADER, &header);
1588 
1589 		/* Is the device part of a Thunderbolt controller? */
1590 		if (dev->vendor == PCI_VENDOR_ID_INTEL &&
1591 		    PCI_VNDR_HEADER_ID(header) == PCI_VSEC_ID_INTEL_TBT) {
1592 			dev->is_thunderbolt = 1;
1593 			return;
1594 		}
1595 	}
1596 }
1597 
1598 static void set_pcie_untrusted(struct pci_dev *dev)
1599 {
1600 	struct pci_dev *parent;
1601 
1602 	/*
1603 	 * If the upstream bridge is untrusted we treat this device
1604 	 * untrusted as well.
1605 	 */
1606 	parent = pci_upstream_bridge(dev);
1607 	if (parent && (parent->untrusted || parent->external_facing))
1608 		dev->untrusted = true;
1609 }
1610 
1611 static void pci_set_removable(struct pci_dev *dev)
1612 {
1613 	struct pci_dev *parent = pci_upstream_bridge(dev);
1614 
1615 	/*
1616 	 * We (only) consider everything downstream from an external_facing
1617 	 * device to be removable by the user. We're mainly concerned with
1618 	 * consumer platforms with user accessible thunderbolt ports that are
1619 	 * vulnerable to DMA attacks, and we expect those ports to be marked by
1620 	 * the firmware as external_facing. Devices in traditional hotplug
1621 	 * slots can technically be removed, but the expectation is that unless
1622 	 * the port is marked with external_facing, such devices are less
1623 	 * accessible to user / may not be removed by end user, and thus not
1624 	 * exposed as "removable" to userspace.
1625 	 */
1626 	if (parent &&
1627 	    (parent->external_facing || dev_is_removable(&parent->dev)))
1628 		dev_set_removable(&dev->dev, DEVICE_REMOVABLE);
1629 }
1630 
1631 /**
1632  * pci_ext_cfg_is_aliased - Is ext config space just an alias of std config?
1633  * @dev: PCI device
1634  *
1635  * PCI Express to PCI/PCI-X Bridge Specification, rev 1.0, 4.1.4 says that
1636  * when forwarding a type1 configuration request the bridge must check that
1637  * the extended register address field is zero.  The bridge is not permitted
1638  * to forward the transactions and must handle it as an Unsupported Request.
1639  * Some bridges do not follow this rule and simply drop the extended register
1640  * bits, resulting in the standard config space being aliased, every 256
1641  * bytes across the entire configuration space.  Test for this condition by
1642  * comparing the first dword of each potential alias to the vendor/device ID.
1643  * Known offenders:
1644  *   ASM1083/1085 PCIe-to-PCI Reversible Bridge (1b21:1080, rev 01 & 03)
1645  *   AMD/ATI SBx00 PCI to PCI Bridge (1002:4384, rev 40)
1646  */
1647 static bool pci_ext_cfg_is_aliased(struct pci_dev *dev)
1648 {
1649 #ifdef CONFIG_PCI_QUIRKS
1650 	int pos;
1651 	u32 header, tmp;
1652 
1653 	pci_read_config_dword(dev, PCI_VENDOR_ID, &header);
1654 
1655 	for (pos = PCI_CFG_SPACE_SIZE;
1656 	     pos < PCI_CFG_SPACE_EXP_SIZE; pos += PCI_CFG_SPACE_SIZE) {
1657 		if (pci_read_config_dword(dev, pos, &tmp) != PCIBIOS_SUCCESSFUL
1658 		    || header != tmp)
1659 			return false;
1660 	}
1661 
1662 	return true;
1663 #else
1664 	return false;
1665 #endif
1666 }
1667 
1668 /**
1669  * pci_cfg_space_size_ext - Get the configuration space size of the PCI device
1670  * @dev: PCI device
1671  *
1672  * Regular PCI devices have 256 bytes, but PCI-X 2 and PCI Express devices
1673  * have 4096 bytes.  Even if the device is capable, that doesn't mean we can
1674  * access it.  Maybe we don't have a way to generate extended config space
1675  * accesses, or the device is behind a reverse Express bridge.  So we try
1676  * reading the dword at 0x100 which must either be 0 or a valid extended
1677  * capability header.
1678  */
1679 static int pci_cfg_space_size_ext(struct pci_dev *dev)
1680 {
1681 	u32 status;
1682 	int pos = PCI_CFG_SPACE_SIZE;
1683 
1684 	if (pci_read_config_dword(dev, pos, &status) != PCIBIOS_SUCCESSFUL)
1685 		return PCI_CFG_SPACE_SIZE;
1686 	if (status == 0xffffffff || pci_ext_cfg_is_aliased(dev))
1687 		return PCI_CFG_SPACE_SIZE;
1688 
1689 	return PCI_CFG_SPACE_EXP_SIZE;
1690 }
1691 
1692 int pci_cfg_space_size(struct pci_dev *dev)
1693 {
1694 	int pos;
1695 	u32 status;
1696 	u16 class;
1697 
1698 #ifdef CONFIG_PCI_IOV
1699 	/*
1700 	 * Per the SR-IOV specification (rev 1.1, sec 3.5), VFs are required to
1701 	 * implement a PCIe capability and therefore must implement extended
1702 	 * config space.  We can skip the NO_EXTCFG test below and the
1703 	 * reachability/aliasing test in pci_cfg_space_size_ext() by virtue of
1704 	 * the fact that the SR-IOV capability on the PF resides in extended
1705 	 * config space and must be accessible and non-aliased to have enabled
1706 	 * support for this VF.  This is a micro performance optimization for
1707 	 * systems supporting many VFs.
1708 	 */
1709 	if (dev->is_virtfn)
1710 		return PCI_CFG_SPACE_EXP_SIZE;
1711 #endif
1712 
1713 	if (dev->bus->bus_flags & PCI_BUS_FLAGS_NO_EXTCFG)
1714 		return PCI_CFG_SPACE_SIZE;
1715 
1716 	class = dev->class >> 8;
1717 	if (class == PCI_CLASS_BRIDGE_HOST)
1718 		return pci_cfg_space_size_ext(dev);
1719 
1720 	if (pci_is_pcie(dev))
1721 		return pci_cfg_space_size_ext(dev);
1722 
1723 	pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1724 	if (!pos)
1725 		return PCI_CFG_SPACE_SIZE;
1726 
1727 	pci_read_config_dword(dev, pos + PCI_X_STATUS, &status);
1728 	if (status & (PCI_X_STATUS_266MHZ | PCI_X_STATUS_533MHZ))
1729 		return pci_cfg_space_size_ext(dev);
1730 
1731 	return PCI_CFG_SPACE_SIZE;
1732 }
1733 
1734 static u32 pci_class(struct pci_dev *dev)
1735 {
1736 	u32 class;
1737 
1738 #ifdef CONFIG_PCI_IOV
1739 	if (dev->is_virtfn)
1740 		return dev->physfn->sriov->class;
1741 #endif
1742 	pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
1743 	return class;
1744 }
1745 
1746 static void pci_subsystem_ids(struct pci_dev *dev, u16 *vendor, u16 *device)
1747 {
1748 #ifdef CONFIG_PCI_IOV
1749 	if (dev->is_virtfn) {
1750 		*vendor = dev->physfn->sriov->subsystem_vendor;
1751 		*device = dev->physfn->sriov->subsystem_device;
1752 		return;
1753 	}
1754 #endif
1755 	pci_read_config_word(dev, PCI_SUBSYSTEM_VENDOR_ID, vendor);
1756 	pci_read_config_word(dev, PCI_SUBSYSTEM_ID, device);
1757 }
1758 
1759 static u8 pci_hdr_type(struct pci_dev *dev)
1760 {
1761 	u8 hdr_type;
1762 
1763 #ifdef CONFIG_PCI_IOV
1764 	if (dev->is_virtfn)
1765 		return dev->physfn->sriov->hdr_type;
1766 #endif
1767 	pci_read_config_byte(dev, PCI_HEADER_TYPE, &hdr_type);
1768 	return hdr_type;
1769 }
1770 
1771 #define LEGACY_IO_RESOURCE	(IORESOURCE_IO | IORESOURCE_PCI_FIXED)
1772 
1773 /**
1774  * pci_intx_mask_broken - Test PCI_COMMAND_INTX_DISABLE writability
1775  * @dev: PCI device
1776  *
1777  * Test whether PCI_COMMAND_INTX_DISABLE is writable for @dev.  Check this
1778  * at enumeration-time to avoid modifying PCI_COMMAND at run-time.
1779  */
1780 static int pci_intx_mask_broken(struct pci_dev *dev)
1781 {
1782 	u16 orig, toggle, new;
1783 
1784 	pci_read_config_word(dev, PCI_COMMAND, &orig);
1785 	toggle = orig ^ PCI_COMMAND_INTX_DISABLE;
1786 	pci_write_config_word(dev, PCI_COMMAND, toggle);
1787 	pci_read_config_word(dev, PCI_COMMAND, &new);
1788 
1789 	pci_write_config_word(dev, PCI_COMMAND, orig);
1790 
1791 	/*
1792 	 * PCI_COMMAND_INTX_DISABLE was reserved and read-only prior to PCI
1793 	 * r2.3, so strictly speaking, a device is not *broken* if it's not
1794 	 * writable.  But we'll live with the misnomer for now.
1795 	 */
1796 	if (new != toggle)
1797 		return 1;
1798 	return 0;
1799 }
1800 
1801 static void early_dump_pci_device(struct pci_dev *pdev)
1802 {
1803 	u32 value[256 / 4];
1804 	int i;
1805 
1806 	pci_info(pdev, "config space:\n");
1807 
1808 	for (i = 0; i < 256; i += 4)
1809 		pci_read_config_dword(pdev, i, &value[i / 4]);
1810 
1811 	print_hex_dump(KERN_INFO, "", DUMP_PREFIX_OFFSET, 16, 1,
1812 		       value, 256, false);
1813 }
1814 
1815 /**
1816  * pci_setup_device - Fill in class and map information of a device
1817  * @dev: the device structure to fill
1818  *
1819  * Initialize the device structure with information about the device's
1820  * vendor,class,memory and IO-space addresses, IRQ lines etc.
1821  * Called at initialisation of the PCI subsystem and by CardBus services.
1822  * Returns 0 on success and negative if unknown type of device (not normal,
1823  * bridge or CardBus).
1824  */
1825 int pci_setup_device(struct pci_dev *dev)
1826 {
1827 	u32 class;
1828 	u16 cmd;
1829 	u8 hdr_type;
1830 	int pos = 0;
1831 	struct pci_bus_region region;
1832 	struct resource *res;
1833 
1834 	hdr_type = pci_hdr_type(dev);
1835 
1836 	dev->sysdata = dev->bus->sysdata;
1837 	dev->dev.parent = dev->bus->bridge;
1838 	dev->dev.bus = &pci_bus_type;
1839 	dev->hdr_type = hdr_type & 0x7f;
1840 	dev->multifunction = !!(hdr_type & 0x80);
1841 	dev->error_state = pci_channel_io_normal;
1842 	set_pcie_port_type(dev);
1843 
1844 	pci_set_of_node(dev);
1845 	pci_set_acpi_fwnode(dev);
1846 
1847 	pci_dev_assign_slot(dev);
1848 
1849 	/*
1850 	 * Assume 32-bit PCI; let 64-bit PCI cards (which are far rarer)
1851 	 * set this higher, assuming the system even supports it.
1852 	 */
1853 	dev->dma_mask = 0xffffffff;
1854 
1855 	dev_set_name(&dev->dev, "%04x:%02x:%02x.%d", pci_domain_nr(dev->bus),
1856 		     dev->bus->number, PCI_SLOT(dev->devfn),
1857 		     PCI_FUNC(dev->devfn));
1858 
1859 	class = pci_class(dev);
1860 
1861 	dev->revision = class & 0xff;
1862 	dev->class = class >> 8;		    /* upper 3 bytes */
1863 
1864 	if (pci_early_dump)
1865 		early_dump_pci_device(dev);
1866 
1867 	/* Need to have dev->class ready */
1868 	dev->cfg_size = pci_cfg_space_size(dev);
1869 
1870 	/* Need to have dev->cfg_size ready */
1871 	set_pcie_thunderbolt(dev);
1872 
1873 	set_pcie_untrusted(dev);
1874 
1875 	/* "Unknown power state" */
1876 	dev->current_state = PCI_UNKNOWN;
1877 
1878 	/* Early fixups, before probing the BARs */
1879 	pci_fixup_device(pci_fixup_early, dev);
1880 
1881 	pci_set_removable(dev);
1882 
1883 	pci_info(dev, "[%04x:%04x] type %02x class %#08x\n",
1884 		 dev->vendor, dev->device, dev->hdr_type, dev->class);
1885 
1886 	/* Device class may be changed after fixup */
1887 	class = dev->class >> 8;
1888 
1889 	if (dev->non_compliant_bars && !dev->mmio_always_on) {
1890 		pci_read_config_word(dev, PCI_COMMAND, &cmd);
1891 		if (cmd & (PCI_COMMAND_IO | PCI_COMMAND_MEMORY)) {
1892 			pci_info(dev, "device has non-compliant BARs; disabling IO/MEM decoding\n");
1893 			cmd &= ~PCI_COMMAND_IO;
1894 			cmd &= ~PCI_COMMAND_MEMORY;
1895 			pci_write_config_word(dev, PCI_COMMAND, cmd);
1896 		}
1897 	}
1898 
1899 	dev->broken_intx_masking = pci_intx_mask_broken(dev);
1900 
1901 	switch (dev->hdr_type) {		    /* header type */
1902 	case PCI_HEADER_TYPE_NORMAL:		    /* standard header */
1903 		if (class == PCI_CLASS_BRIDGE_PCI)
1904 			goto bad;
1905 		pci_read_irq(dev);
1906 		pci_read_bases(dev, 6, PCI_ROM_ADDRESS);
1907 
1908 		pci_subsystem_ids(dev, &dev->subsystem_vendor, &dev->subsystem_device);
1909 
1910 		/*
1911 		 * Do the ugly legacy mode stuff here rather than broken chip
1912 		 * quirk code. Legacy mode ATA controllers have fixed
1913 		 * addresses. These are not always echoed in BAR0-3, and
1914 		 * BAR0-3 in a few cases contain junk!
1915 		 */
1916 		if (class == PCI_CLASS_STORAGE_IDE) {
1917 			u8 progif;
1918 			pci_read_config_byte(dev, PCI_CLASS_PROG, &progif);
1919 			if ((progif & 1) == 0) {
1920 				region.start = 0x1F0;
1921 				region.end = 0x1F7;
1922 				res = &dev->resource[0];
1923 				res->flags = LEGACY_IO_RESOURCE;
1924 				pcibios_bus_to_resource(dev->bus, res, &region);
1925 				pci_info(dev, "legacy IDE quirk: reg 0x10: %pR\n",
1926 					 res);
1927 				region.start = 0x3F6;
1928 				region.end = 0x3F6;
1929 				res = &dev->resource[1];
1930 				res->flags = LEGACY_IO_RESOURCE;
1931 				pcibios_bus_to_resource(dev->bus, res, &region);
1932 				pci_info(dev, "legacy IDE quirk: reg 0x14: %pR\n",
1933 					 res);
1934 			}
1935 			if ((progif & 4) == 0) {
1936 				region.start = 0x170;
1937 				region.end = 0x177;
1938 				res = &dev->resource[2];
1939 				res->flags = LEGACY_IO_RESOURCE;
1940 				pcibios_bus_to_resource(dev->bus, res, &region);
1941 				pci_info(dev, "legacy IDE quirk: reg 0x18: %pR\n",
1942 					 res);
1943 				region.start = 0x376;
1944 				region.end = 0x376;
1945 				res = &dev->resource[3];
1946 				res->flags = LEGACY_IO_RESOURCE;
1947 				pcibios_bus_to_resource(dev->bus, res, &region);
1948 				pci_info(dev, "legacy IDE quirk: reg 0x1c: %pR\n",
1949 					 res);
1950 			}
1951 		}
1952 		break;
1953 
1954 	case PCI_HEADER_TYPE_BRIDGE:		    /* bridge header */
1955 		/*
1956 		 * The PCI-to-PCI bridge spec requires that subtractive
1957 		 * decoding (i.e. transparent) bridge must have programming
1958 		 * interface code of 0x01.
1959 		 */
1960 		pci_read_irq(dev);
1961 		dev->transparent = ((dev->class & 0xff) == 1);
1962 		pci_read_bases(dev, 2, PCI_ROM_ADDRESS1);
1963 		pci_read_bridge_windows(dev);
1964 		set_pcie_hotplug_bridge(dev);
1965 		pos = pci_find_capability(dev, PCI_CAP_ID_SSVID);
1966 		if (pos) {
1967 			pci_read_config_word(dev, pos + PCI_SSVID_VENDOR_ID, &dev->subsystem_vendor);
1968 			pci_read_config_word(dev, pos + PCI_SSVID_DEVICE_ID, &dev->subsystem_device);
1969 		}
1970 		break;
1971 
1972 	case PCI_HEADER_TYPE_CARDBUS:		    /* CardBus bridge header */
1973 		if (class != PCI_CLASS_BRIDGE_CARDBUS)
1974 			goto bad;
1975 		pci_read_irq(dev);
1976 		pci_read_bases(dev, 1, 0);
1977 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_VENDOR_ID, &dev->subsystem_vendor);
1978 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_ID, &dev->subsystem_device);
1979 		break;
1980 
1981 	default:				    /* unknown header */
1982 		pci_err(dev, "unknown header type %02x, ignoring device\n",
1983 			dev->hdr_type);
1984 		pci_release_of_node(dev);
1985 		return -EIO;
1986 
1987 	bad:
1988 		pci_err(dev, "ignoring class %#08x (doesn't match header type %02x)\n",
1989 			dev->class, dev->hdr_type);
1990 		dev->class = PCI_CLASS_NOT_DEFINED << 8;
1991 	}
1992 
1993 	/* We found a fine healthy device, go go go... */
1994 	return 0;
1995 }
1996 
1997 static void pci_configure_mps(struct pci_dev *dev)
1998 {
1999 	struct pci_dev *bridge = pci_upstream_bridge(dev);
2000 	int mps, mpss, p_mps, rc;
2001 
2002 	if (!pci_is_pcie(dev))
2003 		return;
2004 
2005 	/* MPS and MRRS fields are of type 'RsvdP' for VFs, short-circuit out */
2006 	if (dev->is_virtfn)
2007 		return;
2008 
2009 	/*
2010 	 * For Root Complex Integrated Endpoints, program the maximum
2011 	 * supported value unless limited by the PCIE_BUS_PEER2PEER case.
2012 	 */
2013 	if (pci_pcie_type(dev) == PCI_EXP_TYPE_RC_END) {
2014 		if (pcie_bus_config == PCIE_BUS_PEER2PEER)
2015 			mps = 128;
2016 		else
2017 			mps = 128 << dev->pcie_mpss;
2018 		rc = pcie_set_mps(dev, mps);
2019 		if (rc) {
2020 			pci_warn(dev, "can't set Max Payload Size to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
2021 				 mps);
2022 		}
2023 		return;
2024 	}
2025 
2026 	if (!bridge || !pci_is_pcie(bridge))
2027 		return;
2028 
2029 	mps = pcie_get_mps(dev);
2030 	p_mps = pcie_get_mps(bridge);
2031 
2032 	if (mps == p_mps)
2033 		return;
2034 
2035 	if (pcie_bus_config == PCIE_BUS_TUNE_OFF) {
2036 		pci_warn(dev, "Max Payload Size %d, but upstream %s set to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
2037 			 mps, pci_name(bridge), p_mps);
2038 		return;
2039 	}
2040 
2041 	/*
2042 	 * Fancier MPS configuration is done later by
2043 	 * pcie_bus_configure_settings()
2044 	 */
2045 	if (pcie_bus_config != PCIE_BUS_DEFAULT)
2046 		return;
2047 
2048 	mpss = 128 << dev->pcie_mpss;
2049 	if (mpss < p_mps && pci_pcie_type(bridge) == PCI_EXP_TYPE_ROOT_PORT) {
2050 		pcie_set_mps(bridge, mpss);
2051 		pci_info(dev, "Upstream bridge's Max Payload Size set to %d (was %d, max %d)\n",
2052 			 mpss, p_mps, 128 << bridge->pcie_mpss);
2053 		p_mps = pcie_get_mps(bridge);
2054 	}
2055 
2056 	rc = pcie_set_mps(dev, p_mps);
2057 	if (rc) {
2058 		pci_warn(dev, "can't set Max Payload Size to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
2059 			 p_mps);
2060 		return;
2061 	}
2062 
2063 	pci_info(dev, "Max Payload Size set to %d (was %d, max %d)\n",
2064 		 p_mps, mps, mpss);
2065 }
2066 
2067 int pci_configure_extended_tags(struct pci_dev *dev, void *ign)
2068 {
2069 	struct pci_host_bridge *host;
2070 	u32 cap;
2071 	u16 ctl;
2072 	int ret;
2073 
2074 	if (!pci_is_pcie(dev))
2075 		return 0;
2076 
2077 	ret = pcie_capability_read_dword(dev, PCI_EXP_DEVCAP, &cap);
2078 	if (ret)
2079 		return 0;
2080 
2081 	if (!(cap & PCI_EXP_DEVCAP_EXT_TAG))
2082 		return 0;
2083 
2084 	ret = pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &ctl);
2085 	if (ret)
2086 		return 0;
2087 
2088 	host = pci_find_host_bridge(dev->bus);
2089 	if (!host)
2090 		return 0;
2091 
2092 	/*
2093 	 * If some device in the hierarchy doesn't handle Extended Tags
2094 	 * correctly, make sure they're disabled.
2095 	 */
2096 	if (host->no_ext_tags) {
2097 		if (ctl & PCI_EXP_DEVCTL_EXT_TAG) {
2098 			pci_info(dev, "disabling Extended Tags\n");
2099 			pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
2100 						   PCI_EXP_DEVCTL_EXT_TAG);
2101 		}
2102 		return 0;
2103 	}
2104 
2105 	if (!(ctl & PCI_EXP_DEVCTL_EXT_TAG)) {
2106 		pci_info(dev, "enabling Extended Tags\n");
2107 		pcie_capability_set_word(dev, PCI_EXP_DEVCTL,
2108 					 PCI_EXP_DEVCTL_EXT_TAG);
2109 	}
2110 	return 0;
2111 }
2112 
2113 /**
2114  * pcie_relaxed_ordering_enabled - Probe for PCIe relaxed ordering enable
2115  * @dev: PCI device to query
2116  *
2117  * Returns true if the device has enabled relaxed ordering attribute.
2118  */
2119 bool pcie_relaxed_ordering_enabled(struct pci_dev *dev)
2120 {
2121 	u16 v;
2122 
2123 	pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &v);
2124 
2125 	return !!(v & PCI_EXP_DEVCTL_RELAX_EN);
2126 }
2127 EXPORT_SYMBOL(pcie_relaxed_ordering_enabled);
2128 
2129 static void pci_configure_relaxed_ordering(struct pci_dev *dev)
2130 {
2131 	struct pci_dev *root;
2132 
2133 	/* PCI_EXP_DEVICE_RELAX_EN is RsvdP in VFs */
2134 	if (dev->is_virtfn)
2135 		return;
2136 
2137 	if (!pcie_relaxed_ordering_enabled(dev))
2138 		return;
2139 
2140 	/*
2141 	 * For now, we only deal with Relaxed Ordering issues with Root
2142 	 * Ports. Peer-to-Peer DMA is another can of worms.
2143 	 */
2144 	root = pcie_find_root_port(dev);
2145 	if (!root)
2146 		return;
2147 
2148 	if (root->dev_flags & PCI_DEV_FLAGS_NO_RELAXED_ORDERING) {
2149 		pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
2150 					   PCI_EXP_DEVCTL_RELAX_EN);
2151 		pci_info(dev, "Relaxed Ordering disabled because the Root Port didn't support it\n");
2152 	}
2153 }
2154 
2155 static void pci_configure_ltr(struct pci_dev *dev)
2156 {
2157 #ifdef CONFIG_PCIEASPM
2158 	struct pci_host_bridge *host = pci_find_host_bridge(dev->bus);
2159 	struct pci_dev *bridge;
2160 	u32 cap, ctl;
2161 
2162 	if (!pci_is_pcie(dev))
2163 		return;
2164 
2165 	/* Read L1 PM substate capabilities */
2166 	dev->l1ss = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_L1SS);
2167 
2168 	pcie_capability_read_dword(dev, PCI_EXP_DEVCAP2, &cap);
2169 	if (!(cap & PCI_EXP_DEVCAP2_LTR))
2170 		return;
2171 
2172 	pcie_capability_read_dword(dev, PCI_EXP_DEVCTL2, &ctl);
2173 	if (ctl & PCI_EXP_DEVCTL2_LTR_EN) {
2174 		if (pci_pcie_type(dev) == PCI_EXP_TYPE_ROOT_PORT) {
2175 			dev->ltr_path = 1;
2176 			return;
2177 		}
2178 
2179 		bridge = pci_upstream_bridge(dev);
2180 		if (bridge && bridge->ltr_path)
2181 			dev->ltr_path = 1;
2182 
2183 		return;
2184 	}
2185 
2186 	if (!host->native_ltr)
2187 		return;
2188 
2189 	/*
2190 	 * Software must not enable LTR in an Endpoint unless the Root
2191 	 * Complex and all intermediate Switches indicate support for LTR.
2192 	 * PCIe r4.0, sec 6.18.
2193 	 */
2194 	if (pci_pcie_type(dev) == PCI_EXP_TYPE_ROOT_PORT) {
2195 		pcie_capability_set_word(dev, PCI_EXP_DEVCTL2,
2196 					 PCI_EXP_DEVCTL2_LTR_EN);
2197 		dev->ltr_path = 1;
2198 		return;
2199 	}
2200 
2201 	/*
2202 	 * If we're configuring a hot-added device, LTR was likely
2203 	 * disabled in the upstream bridge, so re-enable it before enabling
2204 	 * it in the new device.
2205 	 */
2206 	bridge = pci_upstream_bridge(dev);
2207 	if (bridge && bridge->ltr_path) {
2208 		pci_bridge_reconfigure_ltr(dev);
2209 		pcie_capability_set_word(dev, PCI_EXP_DEVCTL2,
2210 					 PCI_EXP_DEVCTL2_LTR_EN);
2211 		dev->ltr_path = 1;
2212 	}
2213 #endif
2214 }
2215 
2216 static void pci_configure_eetlp_prefix(struct pci_dev *dev)
2217 {
2218 #ifdef CONFIG_PCI_PASID
2219 	struct pci_dev *bridge;
2220 	int pcie_type;
2221 	u32 cap;
2222 
2223 	if (!pci_is_pcie(dev))
2224 		return;
2225 
2226 	pcie_capability_read_dword(dev, PCI_EXP_DEVCAP2, &cap);
2227 	if (!(cap & PCI_EXP_DEVCAP2_EE_PREFIX))
2228 		return;
2229 
2230 	pcie_type = pci_pcie_type(dev);
2231 	if (pcie_type == PCI_EXP_TYPE_ROOT_PORT ||
2232 	    pcie_type == PCI_EXP_TYPE_RC_END)
2233 		dev->eetlp_prefix_path = 1;
2234 	else {
2235 		bridge = pci_upstream_bridge(dev);
2236 		if (bridge && bridge->eetlp_prefix_path)
2237 			dev->eetlp_prefix_path = 1;
2238 	}
2239 #endif
2240 }
2241 
2242 static void pci_configure_serr(struct pci_dev *dev)
2243 {
2244 	u16 control;
2245 
2246 	if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
2247 
2248 		/*
2249 		 * A bridge will not forward ERR_ messages coming from an
2250 		 * endpoint unless SERR# forwarding is enabled.
2251 		 */
2252 		pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &control);
2253 		if (!(control & PCI_BRIDGE_CTL_SERR)) {
2254 			control |= PCI_BRIDGE_CTL_SERR;
2255 			pci_write_config_word(dev, PCI_BRIDGE_CONTROL, control);
2256 		}
2257 	}
2258 }
2259 
2260 static void pci_configure_device(struct pci_dev *dev)
2261 {
2262 	pci_configure_mps(dev);
2263 	pci_configure_extended_tags(dev, NULL);
2264 	pci_configure_relaxed_ordering(dev);
2265 	pci_configure_ltr(dev);
2266 	pci_configure_eetlp_prefix(dev);
2267 	pci_configure_serr(dev);
2268 
2269 	pci_acpi_program_hp_params(dev);
2270 }
2271 
2272 static void pci_release_capabilities(struct pci_dev *dev)
2273 {
2274 	pci_aer_exit(dev);
2275 	pci_rcec_exit(dev);
2276 	pci_iov_release(dev);
2277 	pci_free_cap_save_buffers(dev);
2278 }
2279 
2280 /**
2281  * pci_release_dev - Free a PCI device structure when all users of it are
2282  *		     finished
2283  * @dev: device that's been disconnected
2284  *
2285  * Will be called only by the device core when all users of this PCI device are
2286  * done.
2287  */
2288 static void pci_release_dev(struct device *dev)
2289 {
2290 	struct pci_dev *pci_dev;
2291 
2292 	pci_dev = to_pci_dev(dev);
2293 	pci_release_capabilities(pci_dev);
2294 	pci_release_of_node(pci_dev);
2295 	pcibios_release_device(pci_dev);
2296 	pci_bus_put(pci_dev->bus);
2297 	kfree(pci_dev->driver_override);
2298 	bitmap_free(pci_dev->dma_alias_mask);
2299 	dev_dbg(dev, "device released\n");
2300 	kfree(pci_dev);
2301 }
2302 
2303 struct pci_dev *pci_alloc_dev(struct pci_bus *bus)
2304 {
2305 	struct pci_dev *dev;
2306 
2307 	dev = kzalloc(sizeof(struct pci_dev), GFP_KERNEL);
2308 	if (!dev)
2309 		return NULL;
2310 
2311 	INIT_LIST_HEAD(&dev->bus_list);
2312 	dev->dev.type = &pci_dev_type;
2313 	dev->bus = pci_bus_get(bus);
2314 
2315 	return dev;
2316 }
2317 EXPORT_SYMBOL(pci_alloc_dev);
2318 
2319 static bool pci_bus_crs_vendor_id(u32 l)
2320 {
2321 	return (l & 0xffff) == 0x0001;
2322 }
2323 
2324 static bool pci_bus_wait_crs(struct pci_bus *bus, int devfn, u32 *l,
2325 			     int timeout)
2326 {
2327 	int delay = 1;
2328 
2329 	if (!pci_bus_crs_vendor_id(*l))
2330 		return true;	/* not a CRS completion */
2331 
2332 	if (!timeout)
2333 		return false;	/* CRS, but caller doesn't want to wait */
2334 
2335 	/*
2336 	 * We got the reserved Vendor ID that indicates a completion with
2337 	 * Configuration Request Retry Status (CRS).  Retry until we get a
2338 	 * valid Vendor ID or we time out.
2339 	 */
2340 	while (pci_bus_crs_vendor_id(*l)) {
2341 		if (delay > timeout) {
2342 			pr_warn("pci %04x:%02x:%02x.%d: not ready after %dms; giving up\n",
2343 				pci_domain_nr(bus), bus->number,
2344 				PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2345 
2346 			return false;
2347 		}
2348 		if (delay >= 1000)
2349 			pr_info("pci %04x:%02x:%02x.%d: not ready after %dms; waiting\n",
2350 				pci_domain_nr(bus), bus->number,
2351 				PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2352 
2353 		msleep(delay);
2354 		delay *= 2;
2355 
2356 		if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
2357 			return false;
2358 	}
2359 
2360 	if (delay >= 1000)
2361 		pr_info("pci %04x:%02x:%02x.%d: ready after %dms\n",
2362 			pci_domain_nr(bus), bus->number,
2363 			PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2364 
2365 	return true;
2366 }
2367 
2368 bool pci_bus_generic_read_dev_vendor_id(struct pci_bus *bus, int devfn, u32 *l,
2369 					int timeout)
2370 {
2371 	if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
2372 		return false;
2373 
2374 	/* Some broken boards return 0 or ~0 if a slot is empty: */
2375 	if (*l == 0xffffffff || *l == 0x00000000 ||
2376 	    *l == 0x0000ffff || *l == 0xffff0000)
2377 		return false;
2378 
2379 	if (pci_bus_crs_vendor_id(*l))
2380 		return pci_bus_wait_crs(bus, devfn, l, timeout);
2381 
2382 	return true;
2383 }
2384 
2385 bool pci_bus_read_dev_vendor_id(struct pci_bus *bus, int devfn, u32 *l,
2386 				int timeout)
2387 {
2388 #ifdef CONFIG_PCI_QUIRKS
2389 	struct pci_dev *bridge = bus->self;
2390 
2391 	/*
2392 	 * Certain IDT switches have an issue where they improperly trigger
2393 	 * ACS Source Validation errors on completions for config reads.
2394 	 */
2395 	if (bridge && bridge->vendor == PCI_VENDOR_ID_IDT &&
2396 	    bridge->device == 0x80b5)
2397 		return pci_idt_bus_quirk(bus, devfn, l, timeout);
2398 #endif
2399 
2400 	return pci_bus_generic_read_dev_vendor_id(bus, devfn, l, timeout);
2401 }
2402 EXPORT_SYMBOL(pci_bus_read_dev_vendor_id);
2403 
2404 /*
2405  * Read the config data for a PCI device, sanity-check it,
2406  * and fill in the dev structure.
2407  */
2408 static struct pci_dev *pci_scan_device(struct pci_bus *bus, int devfn)
2409 {
2410 	struct pci_dev *dev;
2411 	u32 l;
2412 
2413 	if (!pci_bus_read_dev_vendor_id(bus, devfn, &l, 60*1000))
2414 		return NULL;
2415 
2416 	dev = pci_alloc_dev(bus);
2417 	if (!dev)
2418 		return NULL;
2419 
2420 	dev->devfn = devfn;
2421 	dev->vendor = l & 0xffff;
2422 	dev->device = (l >> 16) & 0xffff;
2423 
2424 	if (pci_setup_device(dev)) {
2425 		pci_bus_put(dev->bus);
2426 		kfree(dev);
2427 		return NULL;
2428 	}
2429 
2430 	return dev;
2431 }
2432 
2433 void pcie_report_downtraining(struct pci_dev *dev)
2434 {
2435 	if (!pci_is_pcie(dev))
2436 		return;
2437 
2438 	/* Look from the device up to avoid downstream ports with no devices */
2439 	if ((pci_pcie_type(dev) != PCI_EXP_TYPE_ENDPOINT) &&
2440 	    (pci_pcie_type(dev) != PCI_EXP_TYPE_LEG_END) &&
2441 	    (pci_pcie_type(dev) != PCI_EXP_TYPE_UPSTREAM))
2442 		return;
2443 
2444 	/* Multi-function PCIe devices share the same link/status */
2445 	if (PCI_FUNC(dev->devfn) != 0 || dev->is_virtfn)
2446 		return;
2447 
2448 	/* Print link status only if the device is constrained by the fabric */
2449 	__pcie_print_link_status(dev, false);
2450 }
2451 
2452 static void pci_init_capabilities(struct pci_dev *dev)
2453 {
2454 	pci_ea_init(dev);		/* Enhanced Allocation */
2455 	pci_msi_init(dev);		/* Disable MSI */
2456 	pci_msix_init(dev);		/* Disable MSI-X */
2457 
2458 	/* Buffers for saving PCIe and PCI-X capabilities */
2459 	pci_allocate_cap_save_buffers(dev);
2460 
2461 	pci_pm_init(dev);		/* Power Management */
2462 	pci_vpd_init(dev);		/* Vital Product Data */
2463 	pci_configure_ari(dev);		/* Alternative Routing-ID Forwarding */
2464 	pci_iov_init(dev);		/* Single Root I/O Virtualization */
2465 	pci_ats_init(dev);		/* Address Translation Services */
2466 	pci_pri_init(dev);		/* Page Request Interface */
2467 	pci_pasid_init(dev);		/* Process Address Space ID */
2468 	pci_acs_init(dev);		/* Access Control Services */
2469 	pci_ptm_init(dev);		/* Precision Time Measurement */
2470 	pci_aer_init(dev);		/* Advanced Error Reporting */
2471 	pci_dpc_init(dev);		/* Downstream Port Containment */
2472 	pci_rcec_init(dev);		/* Root Complex Event Collector */
2473 
2474 	pcie_report_downtraining(dev);
2475 	pci_init_reset_methods(dev);
2476 }
2477 
2478 /*
2479  * This is the equivalent of pci_host_bridge_msi_domain() that acts on
2480  * devices. Firmware interfaces that can select the MSI domain on a
2481  * per-device basis should be called from here.
2482  */
2483 static struct irq_domain *pci_dev_msi_domain(struct pci_dev *dev)
2484 {
2485 	struct irq_domain *d;
2486 
2487 	/*
2488 	 * If a domain has been set through the pcibios_device_add()
2489 	 * callback, then this is the one (platform code knows best).
2490 	 */
2491 	d = dev_get_msi_domain(&dev->dev);
2492 	if (d)
2493 		return d;
2494 
2495 	/*
2496 	 * Let's see if we have a firmware interface able to provide
2497 	 * the domain.
2498 	 */
2499 	d = pci_msi_get_device_domain(dev);
2500 	if (d)
2501 		return d;
2502 
2503 	return NULL;
2504 }
2505 
2506 static void pci_set_msi_domain(struct pci_dev *dev)
2507 {
2508 	struct irq_domain *d;
2509 
2510 	/*
2511 	 * If the platform or firmware interfaces cannot supply a
2512 	 * device-specific MSI domain, then inherit the default domain
2513 	 * from the host bridge itself.
2514 	 */
2515 	d = pci_dev_msi_domain(dev);
2516 	if (!d)
2517 		d = dev_get_msi_domain(&dev->bus->dev);
2518 
2519 	dev_set_msi_domain(&dev->dev, d);
2520 }
2521 
2522 void pci_device_add(struct pci_dev *dev, struct pci_bus *bus)
2523 {
2524 	int ret;
2525 
2526 	pci_configure_device(dev);
2527 
2528 	device_initialize(&dev->dev);
2529 	dev->dev.release = pci_release_dev;
2530 
2531 	set_dev_node(&dev->dev, pcibus_to_node(bus));
2532 	dev->dev.dma_mask = &dev->dma_mask;
2533 	dev->dev.dma_parms = &dev->dma_parms;
2534 	dev->dev.coherent_dma_mask = 0xffffffffull;
2535 
2536 	dma_set_max_seg_size(&dev->dev, 65536);
2537 	dma_set_seg_boundary(&dev->dev, 0xffffffff);
2538 
2539 	/* Fix up broken headers */
2540 	pci_fixup_device(pci_fixup_header, dev);
2541 
2542 	pci_reassigndev_resource_alignment(dev);
2543 
2544 	dev->state_saved = false;
2545 
2546 	pci_init_capabilities(dev);
2547 
2548 	/*
2549 	 * Add the device to our list of discovered devices
2550 	 * and the bus list for fixup functions, etc.
2551 	 */
2552 	down_write(&pci_bus_sem);
2553 	list_add_tail(&dev->bus_list, &bus->devices);
2554 	up_write(&pci_bus_sem);
2555 
2556 	ret = pcibios_device_add(dev);
2557 	WARN_ON(ret < 0);
2558 
2559 	/* Set up MSI IRQ domain */
2560 	pci_set_msi_domain(dev);
2561 
2562 	/* Notifier could use PCI capabilities */
2563 	dev->match_driver = false;
2564 	ret = device_add(&dev->dev);
2565 	WARN_ON(ret < 0);
2566 }
2567 
2568 struct pci_dev *pci_scan_single_device(struct pci_bus *bus, int devfn)
2569 {
2570 	struct pci_dev *dev;
2571 
2572 	dev = pci_get_slot(bus, devfn);
2573 	if (dev) {
2574 		pci_dev_put(dev);
2575 		return dev;
2576 	}
2577 
2578 	dev = pci_scan_device(bus, devfn);
2579 	if (!dev)
2580 		return NULL;
2581 
2582 	pci_device_add(dev, bus);
2583 
2584 	return dev;
2585 }
2586 EXPORT_SYMBOL(pci_scan_single_device);
2587 
2588 static unsigned int next_fn(struct pci_bus *bus, struct pci_dev *dev,
2589 			    unsigned int fn)
2590 {
2591 	int pos;
2592 	u16 cap = 0;
2593 	unsigned int next_fn;
2594 
2595 	if (pci_ari_enabled(bus)) {
2596 		if (!dev)
2597 			return 0;
2598 		pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ARI);
2599 		if (!pos)
2600 			return 0;
2601 
2602 		pci_read_config_word(dev, pos + PCI_ARI_CAP, &cap);
2603 		next_fn = PCI_ARI_CAP_NFN(cap);
2604 		if (next_fn <= fn)
2605 			return 0;	/* protect against malformed list */
2606 
2607 		return next_fn;
2608 	}
2609 
2610 	/* dev may be NULL for non-contiguous multifunction devices */
2611 	if (!dev || dev->multifunction)
2612 		return (fn + 1) % 8;
2613 
2614 	return 0;
2615 }
2616 
2617 static int only_one_child(struct pci_bus *bus)
2618 {
2619 	struct pci_dev *bridge = bus->self;
2620 
2621 	/*
2622 	 * Systems with unusual topologies set PCI_SCAN_ALL_PCIE_DEVS so
2623 	 * we scan for all possible devices, not just Device 0.
2624 	 */
2625 	if (pci_has_flag(PCI_SCAN_ALL_PCIE_DEVS))
2626 		return 0;
2627 
2628 	/*
2629 	 * A PCIe Downstream Port normally leads to a Link with only Device
2630 	 * 0 on it (PCIe spec r3.1, sec 7.3.1).  As an optimization, scan
2631 	 * only for Device 0 in that situation.
2632 	 */
2633 	if (bridge && pci_is_pcie(bridge) && pcie_downstream_port(bridge))
2634 		return 1;
2635 
2636 	return 0;
2637 }
2638 
2639 /**
2640  * pci_scan_slot - Scan a PCI slot on a bus for devices
2641  * @bus: PCI bus to scan
2642  * @devfn: slot number to scan (must have zero function)
2643  *
2644  * Scan a PCI slot on the specified PCI bus for devices, adding
2645  * discovered devices to the @bus->devices list.  New devices
2646  * will not have is_added set.
2647  *
2648  * Returns the number of new devices found.
2649  */
2650 int pci_scan_slot(struct pci_bus *bus, int devfn)
2651 {
2652 	unsigned int fn, nr = 0;
2653 	struct pci_dev *dev;
2654 
2655 	if (only_one_child(bus) && (devfn > 0))
2656 		return 0; /* Already scanned the entire slot */
2657 
2658 	dev = pci_scan_single_device(bus, devfn);
2659 	if (!dev)
2660 		return 0;
2661 	if (!pci_dev_is_added(dev))
2662 		nr++;
2663 
2664 	for (fn = next_fn(bus, dev, 0); fn > 0; fn = next_fn(bus, dev, fn)) {
2665 		dev = pci_scan_single_device(bus, devfn + fn);
2666 		if (dev) {
2667 			if (!pci_dev_is_added(dev))
2668 				nr++;
2669 			dev->multifunction = 1;
2670 		}
2671 	}
2672 
2673 	/* Only one slot has PCIe device */
2674 	if (bus->self && nr)
2675 		pcie_aspm_init_link_state(bus->self);
2676 
2677 	return nr;
2678 }
2679 EXPORT_SYMBOL(pci_scan_slot);
2680 
2681 static int pcie_find_smpss(struct pci_dev *dev, void *data)
2682 {
2683 	u8 *smpss = data;
2684 
2685 	if (!pci_is_pcie(dev))
2686 		return 0;
2687 
2688 	/*
2689 	 * We don't have a way to change MPS settings on devices that have
2690 	 * drivers attached.  A hot-added device might support only the minimum
2691 	 * MPS setting (MPS=128).  Therefore, if the fabric contains a bridge
2692 	 * where devices may be hot-added, we limit the fabric MPS to 128 so
2693 	 * hot-added devices will work correctly.
2694 	 *
2695 	 * However, if we hot-add a device to a slot directly below a Root
2696 	 * Port, it's impossible for there to be other existing devices below
2697 	 * the port.  We don't limit the MPS in this case because we can
2698 	 * reconfigure MPS on both the Root Port and the hot-added device,
2699 	 * and there are no other devices involved.
2700 	 *
2701 	 * Note that this PCIE_BUS_SAFE path assumes no peer-to-peer DMA.
2702 	 */
2703 	if (dev->is_hotplug_bridge &&
2704 	    pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT)
2705 		*smpss = 0;
2706 
2707 	if (*smpss > dev->pcie_mpss)
2708 		*smpss = dev->pcie_mpss;
2709 
2710 	return 0;
2711 }
2712 
2713 static void pcie_write_mps(struct pci_dev *dev, int mps)
2714 {
2715 	int rc;
2716 
2717 	if (pcie_bus_config == PCIE_BUS_PERFORMANCE) {
2718 		mps = 128 << dev->pcie_mpss;
2719 
2720 		if (pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT &&
2721 		    dev->bus->self)
2722 
2723 			/*
2724 			 * For "Performance", the assumption is made that
2725 			 * downstream communication will never be larger than
2726 			 * the MRRS.  So, the MPS only needs to be configured
2727 			 * for the upstream communication.  This being the case,
2728 			 * walk from the top down and set the MPS of the child
2729 			 * to that of the parent bus.
2730 			 *
2731 			 * Configure the device MPS with the smaller of the
2732 			 * device MPSS or the bridge MPS (which is assumed to be
2733 			 * properly configured at this point to the largest
2734 			 * allowable MPS based on its parent bus).
2735 			 */
2736 			mps = min(mps, pcie_get_mps(dev->bus->self));
2737 	}
2738 
2739 	rc = pcie_set_mps(dev, mps);
2740 	if (rc)
2741 		pci_err(dev, "Failed attempting to set the MPS\n");
2742 }
2743 
2744 static void pcie_write_mrrs(struct pci_dev *dev)
2745 {
2746 	int rc, mrrs;
2747 
2748 	/*
2749 	 * In the "safe" case, do not configure the MRRS.  There appear to be
2750 	 * issues with setting MRRS to 0 on a number of devices.
2751 	 */
2752 	if (pcie_bus_config != PCIE_BUS_PERFORMANCE)
2753 		return;
2754 
2755 	/*
2756 	 * For max performance, the MRRS must be set to the largest supported
2757 	 * value.  However, it cannot be configured larger than the MPS the
2758 	 * device or the bus can support.  This should already be properly
2759 	 * configured by a prior call to pcie_write_mps().
2760 	 */
2761 	mrrs = pcie_get_mps(dev);
2762 
2763 	/*
2764 	 * MRRS is a R/W register.  Invalid values can be written, but a
2765 	 * subsequent read will verify if the value is acceptable or not.
2766 	 * If the MRRS value provided is not acceptable (e.g., too large),
2767 	 * shrink the value until it is acceptable to the HW.
2768 	 */
2769 	while (mrrs != pcie_get_readrq(dev) && mrrs >= 128) {
2770 		rc = pcie_set_readrq(dev, mrrs);
2771 		if (!rc)
2772 			break;
2773 
2774 		pci_warn(dev, "Failed attempting to set the MRRS\n");
2775 		mrrs /= 2;
2776 	}
2777 
2778 	if (mrrs < 128)
2779 		pci_err(dev, "MRRS was unable to be configured with a safe value.  If problems are experienced, try running with pci=pcie_bus_safe\n");
2780 }
2781 
2782 static int pcie_bus_configure_set(struct pci_dev *dev, void *data)
2783 {
2784 	int mps, orig_mps;
2785 
2786 	if (!pci_is_pcie(dev))
2787 		return 0;
2788 
2789 	if (pcie_bus_config == PCIE_BUS_TUNE_OFF ||
2790 	    pcie_bus_config == PCIE_BUS_DEFAULT)
2791 		return 0;
2792 
2793 	mps = 128 << *(u8 *)data;
2794 	orig_mps = pcie_get_mps(dev);
2795 
2796 	pcie_write_mps(dev, mps);
2797 	pcie_write_mrrs(dev);
2798 
2799 	pci_info(dev, "Max Payload Size set to %4d/%4d (was %4d), Max Read Rq %4d\n",
2800 		 pcie_get_mps(dev), 128 << dev->pcie_mpss,
2801 		 orig_mps, pcie_get_readrq(dev));
2802 
2803 	return 0;
2804 }
2805 
2806 /*
2807  * pcie_bus_configure_settings() requires that pci_walk_bus work in a top-down,
2808  * parents then children fashion.  If this changes, then this code will not
2809  * work as designed.
2810  */
2811 void pcie_bus_configure_settings(struct pci_bus *bus)
2812 {
2813 	u8 smpss = 0;
2814 
2815 	if (!bus->self)
2816 		return;
2817 
2818 	if (!pci_is_pcie(bus->self))
2819 		return;
2820 
2821 	/*
2822 	 * FIXME - Peer to peer DMA is possible, though the endpoint would need
2823 	 * to be aware of the MPS of the destination.  To work around this,
2824 	 * simply force the MPS of the entire system to the smallest possible.
2825 	 */
2826 	if (pcie_bus_config == PCIE_BUS_PEER2PEER)
2827 		smpss = 0;
2828 
2829 	if (pcie_bus_config == PCIE_BUS_SAFE) {
2830 		smpss = bus->self->pcie_mpss;
2831 
2832 		pcie_find_smpss(bus->self, &smpss);
2833 		pci_walk_bus(bus, pcie_find_smpss, &smpss);
2834 	}
2835 
2836 	pcie_bus_configure_set(bus->self, &smpss);
2837 	pci_walk_bus(bus, pcie_bus_configure_set, &smpss);
2838 }
2839 EXPORT_SYMBOL_GPL(pcie_bus_configure_settings);
2840 
2841 /*
2842  * Called after each bus is probed, but before its children are examined.  This
2843  * is marked as __weak because multiple architectures define it.
2844  */
2845 void __weak pcibios_fixup_bus(struct pci_bus *bus)
2846 {
2847        /* nothing to do, expected to be removed in the future */
2848 }
2849 
2850 /**
2851  * pci_scan_child_bus_extend() - Scan devices below a bus
2852  * @bus: Bus to scan for devices
2853  * @available_buses: Total number of buses available (%0 does not try to
2854  *		     extend beyond the minimal)
2855  *
2856  * Scans devices below @bus including subordinate buses. Returns new
2857  * subordinate number including all the found devices. Passing
2858  * @available_buses causes the remaining bus space to be distributed
2859  * equally between hotplug-capable bridges to allow future extension of the
2860  * hierarchy.
2861  */
2862 static unsigned int pci_scan_child_bus_extend(struct pci_bus *bus,
2863 					      unsigned int available_buses)
2864 {
2865 	unsigned int used_buses, normal_bridges = 0, hotplug_bridges = 0;
2866 	unsigned int start = bus->busn_res.start;
2867 	unsigned int devfn, fn, cmax, max = start;
2868 	struct pci_dev *dev;
2869 	int nr_devs;
2870 
2871 	dev_dbg(&bus->dev, "scanning bus\n");
2872 
2873 	/* Go find them, Rover! */
2874 	for (devfn = 0; devfn < 256; devfn += 8) {
2875 		nr_devs = pci_scan_slot(bus, devfn);
2876 
2877 		/*
2878 		 * The Jailhouse hypervisor may pass individual functions of a
2879 		 * multi-function device to a guest without passing function 0.
2880 		 * Look for them as well.
2881 		 */
2882 		if (jailhouse_paravirt() && nr_devs == 0) {
2883 			for (fn = 1; fn < 8; fn++) {
2884 				dev = pci_scan_single_device(bus, devfn + fn);
2885 				if (dev)
2886 					dev->multifunction = 1;
2887 			}
2888 		}
2889 	}
2890 
2891 	/* Reserve buses for SR-IOV capability */
2892 	used_buses = pci_iov_bus_range(bus);
2893 	max += used_buses;
2894 
2895 	/*
2896 	 * After performing arch-dependent fixup of the bus, look behind
2897 	 * all PCI-to-PCI bridges on this bus.
2898 	 */
2899 	if (!bus->is_added) {
2900 		dev_dbg(&bus->dev, "fixups for bus\n");
2901 		pcibios_fixup_bus(bus);
2902 		bus->is_added = 1;
2903 	}
2904 
2905 	/*
2906 	 * Calculate how many hotplug bridges and normal bridges there
2907 	 * are on this bus. We will distribute the additional available
2908 	 * buses between hotplug bridges.
2909 	 */
2910 	for_each_pci_bridge(dev, bus) {
2911 		if (dev->is_hotplug_bridge)
2912 			hotplug_bridges++;
2913 		else
2914 			normal_bridges++;
2915 	}
2916 
2917 	/*
2918 	 * Scan bridges that are already configured. We don't touch them
2919 	 * unless they are misconfigured (which will be done in the second
2920 	 * scan below).
2921 	 */
2922 	for_each_pci_bridge(dev, bus) {
2923 		cmax = max;
2924 		max = pci_scan_bridge_extend(bus, dev, max, 0, 0);
2925 
2926 		/*
2927 		 * Reserve one bus for each bridge now to avoid extending
2928 		 * hotplug bridges too much during the second scan below.
2929 		 */
2930 		used_buses++;
2931 		if (cmax - max > 1)
2932 			used_buses += cmax - max - 1;
2933 	}
2934 
2935 	/* Scan bridges that need to be reconfigured */
2936 	for_each_pci_bridge(dev, bus) {
2937 		unsigned int buses = 0;
2938 
2939 		if (!hotplug_bridges && normal_bridges == 1) {
2940 
2941 			/*
2942 			 * There is only one bridge on the bus (upstream
2943 			 * port) so it gets all available buses which it
2944 			 * can then distribute to the possible hotplug
2945 			 * bridges below.
2946 			 */
2947 			buses = available_buses;
2948 		} else if (dev->is_hotplug_bridge) {
2949 
2950 			/*
2951 			 * Distribute the extra buses between hotplug
2952 			 * bridges if any.
2953 			 */
2954 			buses = available_buses / hotplug_bridges;
2955 			buses = min(buses, available_buses - used_buses + 1);
2956 		}
2957 
2958 		cmax = max;
2959 		max = pci_scan_bridge_extend(bus, dev, cmax, buses, 1);
2960 		/* One bus is already accounted so don't add it again */
2961 		if (max - cmax > 1)
2962 			used_buses += max - cmax - 1;
2963 	}
2964 
2965 	/*
2966 	 * Make sure a hotplug bridge has at least the minimum requested
2967 	 * number of buses but allow it to grow up to the maximum available
2968 	 * bus number of there is room.
2969 	 */
2970 	if (bus->self && bus->self->is_hotplug_bridge) {
2971 		used_buses = max_t(unsigned int, available_buses,
2972 				   pci_hotplug_bus_size - 1);
2973 		if (max - start < used_buses) {
2974 			max = start + used_buses;
2975 
2976 			/* Do not allocate more buses than we have room left */
2977 			if (max > bus->busn_res.end)
2978 				max = bus->busn_res.end;
2979 
2980 			dev_dbg(&bus->dev, "%pR extended by %#02x\n",
2981 				&bus->busn_res, max - start);
2982 		}
2983 	}
2984 
2985 	/*
2986 	 * We've scanned the bus and so we know all about what's on
2987 	 * the other side of any bridges that may be on this bus plus
2988 	 * any devices.
2989 	 *
2990 	 * Return how far we've got finding sub-buses.
2991 	 */
2992 	dev_dbg(&bus->dev, "bus scan returning with max=%02x\n", max);
2993 	return max;
2994 }
2995 
2996 /**
2997  * pci_scan_child_bus() - Scan devices below a bus
2998  * @bus: Bus to scan for devices
2999  *
3000  * Scans devices below @bus including subordinate buses. Returns new
3001  * subordinate number including all the found devices.
3002  */
3003 unsigned int pci_scan_child_bus(struct pci_bus *bus)
3004 {
3005 	return pci_scan_child_bus_extend(bus, 0);
3006 }
3007 EXPORT_SYMBOL_GPL(pci_scan_child_bus);
3008 
3009 /**
3010  * pcibios_root_bridge_prepare - Platform-specific host bridge setup
3011  * @bridge: Host bridge to set up
3012  *
3013  * Default empty implementation.  Replace with an architecture-specific setup
3014  * routine, if necessary.
3015  */
3016 int __weak pcibios_root_bridge_prepare(struct pci_host_bridge *bridge)
3017 {
3018 	return 0;
3019 }
3020 
3021 void __weak pcibios_add_bus(struct pci_bus *bus)
3022 {
3023 }
3024 
3025 void __weak pcibios_remove_bus(struct pci_bus *bus)
3026 {
3027 }
3028 
3029 struct pci_bus *pci_create_root_bus(struct device *parent, int bus,
3030 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
3031 {
3032 	int error;
3033 	struct pci_host_bridge *bridge;
3034 
3035 	bridge = pci_alloc_host_bridge(0);
3036 	if (!bridge)
3037 		return NULL;
3038 
3039 	bridge->dev.parent = parent;
3040 
3041 	list_splice_init(resources, &bridge->windows);
3042 	bridge->sysdata = sysdata;
3043 	bridge->busnr = bus;
3044 	bridge->ops = ops;
3045 
3046 	error = pci_register_host_bridge(bridge);
3047 	if (error < 0)
3048 		goto err_out;
3049 
3050 	return bridge->bus;
3051 
3052 err_out:
3053 	put_device(&bridge->dev);
3054 	return NULL;
3055 }
3056 EXPORT_SYMBOL_GPL(pci_create_root_bus);
3057 
3058 int pci_host_probe(struct pci_host_bridge *bridge)
3059 {
3060 	struct pci_bus *bus, *child;
3061 	int ret;
3062 
3063 	ret = pci_scan_root_bus_bridge(bridge);
3064 	if (ret < 0) {
3065 		dev_err(bridge->dev.parent, "Scanning root bridge failed");
3066 		return ret;
3067 	}
3068 
3069 	bus = bridge->bus;
3070 
3071 	/*
3072 	 * We insert PCI resources into the iomem_resource and
3073 	 * ioport_resource trees in either pci_bus_claim_resources()
3074 	 * or pci_bus_assign_resources().
3075 	 */
3076 	if (pci_has_flag(PCI_PROBE_ONLY)) {
3077 		pci_bus_claim_resources(bus);
3078 	} else {
3079 		pci_bus_size_bridges(bus);
3080 		pci_bus_assign_resources(bus);
3081 
3082 		list_for_each_entry(child, &bus->children, node)
3083 			pcie_bus_configure_settings(child);
3084 	}
3085 
3086 	pci_bus_add_devices(bus);
3087 	return 0;
3088 }
3089 EXPORT_SYMBOL_GPL(pci_host_probe);
3090 
3091 int pci_bus_insert_busn_res(struct pci_bus *b, int bus, int bus_max)
3092 {
3093 	struct resource *res = &b->busn_res;
3094 	struct resource *parent_res, *conflict;
3095 
3096 	res->start = bus;
3097 	res->end = bus_max;
3098 	res->flags = IORESOURCE_BUS;
3099 
3100 	if (!pci_is_root_bus(b))
3101 		parent_res = &b->parent->busn_res;
3102 	else {
3103 		parent_res = get_pci_domain_busn_res(pci_domain_nr(b));
3104 		res->flags |= IORESOURCE_PCI_FIXED;
3105 	}
3106 
3107 	conflict = request_resource_conflict(parent_res, res);
3108 
3109 	if (conflict)
3110 		dev_info(&b->dev,
3111 			   "busn_res: can not insert %pR under %s%pR (conflicts with %s %pR)\n",
3112 			    res, pci_is_root_bus(b) ? "domain " : "",
3113 			    parent_res, conflict->name, conflict);
3114 
3115 	return conflict == NULL;
3116 }
3117 
3118 int pci_bus_update_busn_res_end(struct pci_bus *b, int bus_max)
3119 {
3120 	struct resource *res = &b->busn_res;
3121 	struct resource old_res = *res;
3122 	resource_size_t size;
3123 	int ret;
3124 
3125 	if (res->start > bus_max)
3126 		return -EINVAL;
3127 
3128 	size = bus_max - res->start + 1;
3129 	ret = adjust_resource(res, res->start, size);
3130 	dev_info(&b->dev, "busn_res: %pR end %s updated to %02x\n",
3131 			&old_res, ret ? "can not be" : "is", bus_max);
3132 
3133 	if (!ret && !res->parent)
3134 		pci_bus_insert_busn_res(b, res->start, res->end);
3135 
3136 	return ret;
3137 }
3138 
3139 void pci_bus_release_busn_res(struct pci_bus *b)
3140 {
3141 	struct resource *res = &b->busn_res;
3142 	int ret;
3143 
3144 	if (!res->flags || !res->parent)
3145 		return;
3146 
3147 	ret = release_resource(res);
3148 	dev_info(&b->dev, "busn_res: %pR %s released\n",
3149 			res, ret ? "can not be" : "is");
3150 }
3151 
3152 int pci_scan_root_bus_bridge(struct pci_host_bridge *bridge)
3153 {
3154 	struct resource_entry *window;
3155 	bool found = false;
3156 	struct pci_bus *b;
3157 	int max, bus, ret;
3158 
3159 	if (!bridge)
3160 		return -EINVAL;
3161 
3162 	resource_list_for_each_entry(window, &bridge->windows)
3163 		if (window->res->flags & IORESOURCE_BUS) {
3164 			bridge->busnr = window->res->start;
3165 			found = true;
3166 			break;
3167 		}
3168 
3169 	ret = pci_register_host_bridge(bridge);
3170 	if (ret < 0)
3171 		return ret;
3172 
3173 	b = bridge->bus;
3174 	bus = bridge->busnr;
3175 
3176 	if (!found) {
3177 		dev_info(&b->dev,
3178 		 "No busn resource found for root bus, will use [bus %02x-ff]\n",
3179 			bus);
3180 		pci_bus_insert_busn_res(b, bus, 255);
3181 	}
3182 
3183 	max = pci_scan_child_bus(b);
3184 
3185 	if (!found)
3186 		pci_bus_update_busn_res_end(b, max);
3187 
3188 	return 0;
3189 }
3190 EXPORT_SYMBOL(pci_scan_root_bus_bridge);
3191 
3192 struct pci_bus *pci_scan_root_bus(struct device *parent, int bus,
3193 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
3194 {
3195 	struct resource_entry *window;
3196 	bool found = false;
3197 	struct pci_bus *b;
3198 	int max;
3199 
3200 	resource_list_for_each_entry(window, resources)
3201 		if (window->res->flags & IORESOURCE_BUS) {
3202 			found = true;
3203 			break;
3204 		}
3205 
3206 	b = pci_create_root_bus(parent, bus, ops, sysdata, resources);
3207 	if (!b)
3208 		return NULL;
3209 
3210 	if (!found) {
3211 		dev_info(&b->dev,
3212 		 "No busn resource found for root bus, will use [bus %02x-ff]\n",
3213 			bus);
3214 		pci_bus_insert_busn_res(b, bus, 255);
3215 	}
3216 
3217 	max = pci_scan_child_bus(b);
3218 
3219 	if (!found)
3220 		pci_bus_update_busn_res_end(b, max);
3221 
3222 	return b;
3223 }
3224 EXPORT_SYMBOL(pci_scan_root_bus);
3225 
3226 struct pci_bus *pci_scan_bus(int bus, struct pci_ops *ops,
3227 					void *sysdata)
3228 {
3229 	LIST_HEAD(resources);
3230 	struct pci_bus *b;
3231 
3232 	pci_add_resource(&resources, &ioport_resource);
3233 	pci_add_resource(&resources, &iomem_resource);
3234 	pci_add_resource(&resources, &busn_resource);
3235 	b = pci_create_root_bus(NULL, bus, ops, sysdata, &resources);
3236 	if (b) {
3237 		pci_scan_child_bus(b);
3238 	} else {
3239 		pci_free_resource_list(&resources);
3240 	}
3241 	return b;
3242 }
3243 EXPORT_SYMBOL(pci_scan_bus);
3244 
3245 /**
3246  * pci_rescan_bus_bridge_resize - Scan a PCI bus for devices
3247  * @bridge: PCI bridge for the bus to scan
3248  *
3249  * Scan a PCI bus and child buses for new devices, add them,
3250  * and enable them, resizing bridge mmio/io resource if necessary
3251  * and possible.  The caller must ensure the child devices are already
3252  * removed for resizing to occur.
3253  *
3254  * Returns the max number of subordinate bus discovered.
3255  */
3256 unsigned int pci_rescan_bus_bridge_resize(struct pci_dev *bridge)
3257 {
3258 	unsigned int max;
3259 	struct pci_bus *bus = bridge->subordinate;
3260 
3261 	max = pci_scan_child_bus(bus);
3262 
3263 	pci_assign_unassigned_bridge_resources(bridge);
3264 
3265 	pci_bus_add_devices(bus);
3266 
3267 	return max;
3268 }
3269 
3270 /**
3271  * pci_rescan_bus - Scan a PCI bus for devices
3272  * @bus: PCI bus to scan
3273  *
3274  * Scan a PCI bus and child buses for new devices, add them,
3275  * and enable them.
3276  *
3277  * Returns the max number of subordinate bus discovered.
3278  */
3279 unsigned int pci_rescan_bus(struct pci_bus *bus)
3280 {
3281 	unsigned int max;
3282 
3283 	max = pci_scan_child_bus(bus);
3284 	pci_assign_unassigned_bus_resources(bus);
3285 	pci_bus_add_devices(bus);
3286 
3287 	return max;
3288 }
3289 EXPORT_SYMBOL_GPL(pci_rescan_bus);
3290 
3291 /*
3292  * pci_rescan_bus(), pci_rescan_bus_bridge_resize() and PCI device removal
3293  * routines should always be executed under this mutex.
3294  */
3295 static DEFINE_MUTEX(pci_rescan_remove_lock);
3296 
3297 void pci_lock_rescan_remove(void)
3298 {
3299 	mutex_lock(&pci_rescan_remove_lock);
3300 }
3301 EXPORT_SYMBOL_GPL(pci_lock_rescan_remove);
3302 
3303 void pci_unlock_rescan_remove(void)
3304 {
3305 	mutex_unlock(&pci_rescan_remove_lock);
3306 }
3307 EXPORT_SYMBOL_GPL(pci_unlock_rescan_remove);
3308 
3309 static int __init pci_sort_bf_cmp(const struct device *d_a,
3310 				  const struct device *d_b)
3311 {
3312 	const struct pci_dev *a = to_pci_dev(d_a);
3313 	const struct pci_dev *b = to_pci_dev(d_b);
3314 
3315 	if      (pci_domain_nr(a->bus) < pci_domain_nr(b->bus)) return -1;
3316 	else if (pci_domain_nr(a->bus) > pci_domain_nr(b->bus)) return  1;
3317 
3318 	if      (a->bus->number < b->bus->number) return -1;
3319 	else if (a->bus->number > b->bus->number) return  1;
3320 
3321 	if      (a->devfn < b->devfn) return -1;
3322 	else if (a->devfn > b->devfn) return  1;
3323 
3324 	return 0;
3325 }
3326 
3327 void __init pci_sort_breadthfirst(void)
3328 {
3329 	bus_sort_breadthfirst(&pci_bus_type, &pci_sort_bf_cmp);
3330 }
3331 
3332 int pci_hp_add_bridge(struct pci_dev *dev)
3333 {
3334 	struct pci_bus *parent = dev->bus;
3335 	int busnr, start = parent->busn_res.start;
3336 	unsigned int available_buses = 0;
3337 	int end = parent->busn_res.end;
3338 
3339 	for (busnr = start; busnr <= end; busnr++) {
3340 		if (!pci_find_bus(pci_domain_nr(parent), busnr))
3341 			break;
3342 	}
3343 	if (busnr-- > end) {
3344 		pci_err(dev, "No bus number available for hot-added bridge\n");
3345 		return -1;
3346 	}
3347 
3348 	/* Scan bridges that are already configured */
3349 	busnr = pci_scan_bridge(parent, dev, busnr, 0);
3350 
3351 	/*
3352 	 * Distribute the available bus numbers between hotplug-capable
3353 	 * bridges to make extending the chain later possible.
3354 	 */
3355 	available_buses = end - busnr;
3356 
3357 	/* Scan bridges that need to be reconfigured */
3358 	pci_scan_bridge_extend(parent, dev, busnr, available_buses, 1);
3359 
3360 	if (!dev->subordinate)
3361 		return -1;
3362 
3363 	return 0;
3364 }
3365 EXPORT_SYMBOL_GPL(pci_hp_add_bridge);
3366