xref: /linux/drivers/pci/probe.c (revision 995832b2cebe6969d1b42635db698803ee31294d)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * PCI detection and setup code
4  */
5 
6 #include <linux/array_size.h>
7 #include <linux/kernel.h>
8 #include <linux/delay.h>
9 #include <linux/init.h>
10 #include <linux/pci.h>
11 #include <linux/msi.h>
12 #include <linux/of_pci.h>
13 #include <linux/of_platform.h>
14 #include <linux/platform_device.h>
15 #include <linux/pci_hotplug.h>
16 #include <linux/slab.h>
17 #include <linux/sprintf.h>
18 #include <linux/module.h>
19 #include <linux/cpumask.h>
20 #include <linux/aer.h>
21 #include <linux/acpi.h>
22 #include <linux/hypervisor.h>
23 #include <linux/irqdomain.h>
24 #include <linux/pm_runtime.h>
25 #include <linux/bitfield.h>
26 #include <trace/events/pci.h>
27 #include "pci.h"
28 
29 static struct resource busn_resource = {
30 	.name	= "PCI busn",
31 	.start	= 0,
32 	.end	= 255,
33 	.flags	= IORESOURCE_BUS,
34 };
35 
36 /* Ugh.  Need to stop exporting this to modules. */
37 LIST_HEAD(pci_root_buses);
38 EXPORT_SYMBOL(pci_root_buses);
39 
40 static LIST_HEAD(pci_domain_busn_res_list);
41 
42 struct pci_domain_busn_res {
43 	struct list_head list;
44 	struct resource res;
45 	int domain_nr;
46 };
47 
48 static struct resource *get_pci_domain_busn_res(int domain_nr)
49 {
50 	struct pci_domain_busn_res *r;
51 
52 	list_for_each_entry(r, &pci_domain_busn_res_list, list)
53 		if (r->domain_nr == domain_nr)
54 			return &r->res;
55 
56 	r = kzalloc_obj(*r);
57 	if (!r)
58 		return NULL;
59 
60 	r->domain_nr = domain_nr;
61 	r->res.start = 0;
62 	r->res.end = 0xff;
63 	r->res.flags = IORESOURCE_BUS | IORESOURCE_PCI_FIXED;
64 
65 	list_add_tail(&r->list, &pci_domain_busn_res_list);
66 
67 	return &r->res;
68 }
69 
70 /*
71  * PCI Bus Class
72  */
73 static void release_pcibus_dev(struct device *dev)
74 {
75 	struct pci_bus *pci_bus = to_pci_bus(dev);
76 
77 	put_device(pci_bus->bridge);
78 	pci_bus_remove_resources(pci_bus);
79 	pci_release_bus_of_node(pci_bus);
80 	kfree(pci_bus);
81 }
82 
83 static const struct class pcibus_class = {
84 	.name		= "pci_bus",
85 	.dev_release	= &release_pcibus_dev,
86 	.dev_groups	= pcibus_groups,
87 };
88 
89 static int __init pcibus_class_init(void)
90 {
91 	return class_register(&pcibus_class);
92 }
93 postcore_initcall(pcibus_class_init);
94 
95 static u64 pci_size(u64 base, u64 maxbase, u64 mask)
96 {
97 	u64 size = mask & maxbase;	/* Find the significant bits */
98 	if (!size)
99 		return 0;
100 
101 	/*
102 	 * Get the lowest of them to find the decode size, and from that
103 	 * the extent.
104 	 */
105 	size = size & ~(size-1);
106 
107 	/*
108 	 * base == maxbase can be valid only if the BAR has already been
109 	 * programmed with all 1s.
110 	 */
111 	if (base == maxbase && ((base | (size - 1)) & mask) != mask)
112 		return 0;
113 
114 	return size;
115 }
116 
117 static inline unsigned long decode_bar(struct pci_dev *dev, u32 bar)
118 {
119 	u32 mem_type;
120 	unsigned long flags;
121 
122 	if ((bar & PCI_BASE_ADDRESS_SPACE) == PCI_BASE_ADDRESS_SPACE_IO) {
123 		flags = bar & ~PCI_BASE_ADDRESS_IO_MASK;
124 		flags |= IORESOURCE_IO;
125 		return flags;
126 	}
127 
128 	flags = bar & ~PCI_BASE_ADDRESS_MEM_MASK;
129 	flags |= IORESOURCE_MEM;
130 	if (flags & PCI_BASE_ADDRESS_MEM_PREFETCH)
131 		flags |= IORESOURCE_PREFETCH;
132 
133 	mem_type = bar & PCI_BASE_ADDRESS_MEM_TYPE_MASK;
134 	switch (mem_type) {
135 	case PCI_BASE_ADDRESS_MEM_TYPE_32:
136 		break;
137 	case PCI_BASE_ADDRESS_MEM_TYPE_1M:
138 		/* 1M mem BAR treated as 32-bit BAR */
139 		break;
140 	case PCI_BASE_ADDRESS_MEM_TYPE_64:
141 		flags |= IORESOURCE_MEM_64;
142 		break;
143 	default:
144 		/* mem unknown type treated as 32-bit BAR */
145 		break;
146 	}
147 	return flags;
148 }
149 
150 #define PCI_COMMAND_DECODE_ENABLE	(PCI_COMMAND_MEMORY | PCI_COMMAND_IO)
151 
152 /**
153  * __pci_size_bars - Read the raw BAR mask for a range of PCI BARs
154  * @dev: the PCI device
155  * @count: number of BARs to size
156  * @pos: starting config space position
157  * @sizes: array to store mask values
158  * @rom: indicate whether to use ROM mask, which avoids enabling ROM BARs
159  *
160  * Provided @sizes array must be sufficiently sized to store results for
161  * @count u32 BARs.  Caller is responsible for disabling decode to specified
162  * BAR range around calling this function.  This function is intended to avoid
163  * disabling decode around sizing each BAR individually, which can result in
164  * non-trivial overhead in virtualized environments with very large PCI BARs.
165  */
166 static void __pci_size_bars(struct pci_dev *dev, int count,
167 			    unsigned int pos, u32 *sizes, bool rom)
168 {
169 	u32 orig, mask = rom ? PCI_ROM_ADDRESS_MASK : ~0;
170 	int i;
171 
172 	for (i = 0; i < count; i++, pos += 4, sizes++) {
173 		pci_read_config_dword(dev, pos, &orig);
174 		pci_write_config_dword(dev, pos, mask);
175 		pci_read_config_dword(dev, pos, sizes);
176 		pci_write_config_dword(dev, pos, orig);
177 	}
178 }
179 
180 void __pci_size_stdbars(struct pci_dev *dev, int count,
181 			unsigned int pos, u32 *sizes)
182 {
183 	__pci_size_bars(dev, count, pos, sizes, false);
184 }
185 
186 static void __pci_size_rom(struct pci_dev *dev, unsigned int pos, u32 *sizes)
187 {
188 	__pci_size_bars(dev, 1, pos, sizes, true);
189 }
190 
191 /**
192  * __pci_read_base - Read a PCI BAR
193  * @dev: the PCI device
194  * @type: type of the BAR
195  * @res: resource buffer to be filled in
196  * @pos: BAR position in the config space
197  * @sizes: array of one or more pre-read BAR masks
198  *
199  * Returns 1 if the BAR is 64-bit, or 0 if 32-bit.
200  */
201 int __pci_read_base(struct pci_dev *dev, enum pci_bar_type type,
202 		    struct resource *res, unsigned int pos, u32 *sizes)
203 {
204 	u32 l = 0, sz;
205 	u64 l64, sz64, mask64;
206 	struct pci_bus_region region, inverted_region;
207 	const char *res_name = pci_resource_name(dev, res - dev->resource);
208 
209 	res->name = pci_name(dev);
210 
211 	pci_read_config_dword(dev, pos, &l);
212 	sz = sizes[0];
213 
214 	/*
215 	 * All bits set in sz means the device isn't working properly.
216 	 * If the BAR isn't implemented, all bits must be 0.  If it's a
217 	 * memory BAR or a ROM, bit 0 must be clear; if it's an io BAR, bit
218 	 * 1 must be clear.
219 	 */
220 	if (PCI_POSSIBLE_ERROR(sz))
221 		sz = 0;
222 
223 	/*
224 	 * I don't know how l can have all bits set.  Copied from old code.
225 	 * Maybe it fixes a bug on some ancient platform.
226 	 */
227 	if (PCI_POSSIBLE_ERROR(l))
228 		l = 0;
229 
230 	if (type == pci_bar_unknown) {
231 		res->flags = decode_bar(dev, l);
232 		res->flags |= IORESOURCE_SIZEALIGN;
233 		if (res->flags & IORESOURCE_IO) {
234 			l64 = l & PCI_BASE_ADDRESS_IO_MASK;
235 			sz64 = sz & PCI_BASE_ADDRESS_IO_MASK;
236 			mask64 = PCI_BASE_ADDRESS_IO_MASK & (u32)IO_SPACE_LIMIT;
237 		} else {
238 			l64 = l & PCI_BASE_ADDRESS_MEM_MASK;
239 			sz64 = sz & PCI_BASE_ADDRESS_MEM_MASK;
240 			mask64 = (u32)PCI_BASE_ADDRESS_MEM_MASK;
241 		}
242 	} else {
243 		if (l & PCI_ROM_ADDRESS_ENABLE)
244 			res->flags |= IORESOURCE_ROM_ENABLE;
245 		l64 = l & PCI_ROM_ADDRESS_MASK;
246 		sz64 = sz & PCI_ROM_ADDRESS_MASK;
247 		mask64 = PCI_ROM_ADDRESS_MASK;
248 	}
249 
250 	if (res->flags & IORESOURCE_MEM_64) {
251 		pci_read_config_dword(dev, pos + 4, &l);
252 		sz = sizes[1];
253 
254 		l64 |= ((u64)l << 32);
255 		sz64 |= ((u64)sz << 32);
256 		mask64 |= ((u64)~0 << 32);
257 	}
258 
259 	if (!sz64)
260 		goto fail;
261 
262 	sz64 = pci_size(l64, sz64, mask64);
263 	if (!sz64) {
264 		pci_info(dev, FW_BUG "%s: invalid; can't size\n", res_name);
265 		goto fail;
266 	}
267 
268 	if (res->flags & IORESOURCE_MEM_64) {
269 		if ((sizeof(pci_bus_addr_t) < 8 || sizeof(resource_size_t) < 8)
270 		    && sz64 > 0x100000000ULL) {
271 			res->flags |= IORESOURCE_UNSET | IORESOURCE_DISABLED;
272 			resource_set_range(res, 0, 0);
273 			pci_err(dev, "%s: can't handle BAR larger than 4GB (size %#010llx)\n",
274 				res_name, (unsigned long long)sz64);
275 			goto out;
276 		}
277 
278 		if ((sizeof(pci_bus_addr_t) < 8) && l) {
279 			/* Above 32-bit boundary; try to reallocate */
280 			res->flags |= IORESOURCE_UNSET;
281 			resource_set_range(res, 0, sz64);
282 			pci_info(dev, "%s: can't handle BAR above 4GB (bus address %#010llx)\n",
283 				 res_name, (unsigned long long)l64);
284 			goto out;
285 		}
286 	}
287 
288 	region.start = l64;
289 	region.end = l64 + sz64 - 1;
290 
291 	pcibios_bus_to_resource(dev->bus, res, &region);
292 	pcibios_resource_to_bus(dev->bus, &inverted_region, res);
293 
294 	/*
295 	 * If "A" is a BAR value (a bus address), "bus_to_resource(A)" is
296 	 * the corresponding resource address (the physical address used by
297 	 * the CPU.  Converting that resource address back to a bus address
298 	 * should yield the original BAR value:
299 	 *
300 	 *     resource_to_bus(bus_to_resource(A)) == A
301 	 *
302 	 * If it doesn't, CPU accesses to "bus_to_resource(A)" will not
303 	 * be claimed by the device.
304 	 */
305 	if (inverted_region.start != region.start) {
306 		res->flags |= IORESOURCE_UNSET;
307 		res->start = 0;
308 		res->end = region.end - region.start;
309 		pci_info(dev, "%s: initial BAR value %#010llx invalid\n",
310 			 res_name, (unsigned long long)region.start);
311 	}
312 
313 	goto out;
314 
315 
316 fail:
317 	res->flags = 0;
318 out:
319 	if (res->flags)
320 		pci_info(dev, "%s %pR\n", res_name, res);
321 
322 	return (res->flags & IORESOURCE_MEM_64) ? 1 : 0;
323 }
324 
325 static __always_inline void pci_read_bases(struct pci_dev *dev,
326 					   unsigned int howmany, int rom)
327 {
328 	u32 rombar, stdbars[PCI_STD_NUM_BARS];
329 	unsigned int pos, reg;
330 	u16 orig_cmd;
331 
332 	BUILD_BUG_ON(statically_true(howmany > PCI_STD_NUM_BARS));
333 
334 	if (dev->non_compliant_bars)
335 		return;
336 
337 	/* Per PCIe r4.0, sec 9.3.4.1.11, the VF BARs are all RO Zero */
338 	if (dev->is_virtfn)
339 		return;
340 
341 	/* No printks while decoding is disabled! */
342 	if (!dev->mmio_always_on) {
343 		pci_read_config_word(dev, PCI_COMMAND, &orig_cmd);
344 		if (orig_cmd & PCI_COMMAND_DECODE_ENABLE) {
345 			pci_write_config_word(dev, PCI_COMMAND,
346 				orig_cmd & ~PCI_COMMAND_DECODE_ENABLE);
347 		}
348 	}
349 
350 	__pci_size_stdbars(dev, howmany, PCI_BASE_ADDRESS_0, stdbars);
351 	if (rom)
352 		__pci_size_rom(dev, rom, &rombar);
353 
354 	if (!dev->mmio_always_on &&
355 	    (orig_cmd & PCI_COMMAND_DECODE_ENABLE))
356 		pci_write_config_word(dev, PCI_COMMAND, orig_cmd);
357 
358 	for (pos = 0; pos < howmany; pos++) {
359 		struct resource *res = &dev->resource[pos];
360 		reg = PCI_BASE_ADDRESS_0 + (pos << 2);
361 		pos += __pci_read_base(dev, pci_bar_unknown,
362 				       res, reg, &stdbars[pos]);
363 	}
364 
365 	if (rom) {
366 		struct resource *res = &dev->resource[PCI_ROM_RESOURCE];
367 		dev->rom_base_reg = rom;
368 		res->flags = IORESOURCE_MEM | IORESOURCE_PREFETCH |
369 				IORESOURCE_READONLY | IORESOURCE_SIZEALIGN;
370 		__pci_read_base(dev, pci_bar_mem32, res, rom, &rombar);
371 	}
372 }
373 
374 static void pci_read_bridge_io(struct pci_dev *dev, struct resource *res,
375 			       bool log)
376 {
377 	u8 io_base_lo, io_limit_lo;
378 	unsigned long io_mask, io_granularity, base, limit;
379 	struct pci_bus_region region;
380 
381 	if (!dev->io_window)
382 		return;
383 
384 	io_mask = PCI_IO_RANGE_MASK;
385 	io_granularity = 0x1000;
386 	if (dev->io_window_1k) {
387 		/* Support 1K I/O space granularity */
388 		io_mask = PCI_IO_1K_RANGE_MASK;
389 		io_granularity = 0x400;
390 	}
391 
392 	pci_read_config_byte(dev, PCI_IO_BASE, &io_base_lo);
393 	pci_read_config_byte(dev, PCI_IO_LIMIT, &io_limit_lo);
394 	base = (io_base_lo & io_mask) << 8;
395 	limit = (io_limit_lo & io_mask) << 8;
396 
397 	if ((io_base_lo & PCI_IO_RANGE_TYPE_MASK) == PCI_IO_RANGE_TYPE_32) {
398 		u16 io_base_hi, io_limit_hi;
399 
400 		pci_read_config_word(dev, PCI_IO_BASE_UPPER16, &io_base_hi);
401 		pci_read_config_word(dev, PCI_IO_LIMIT_UPPER16, &io_limit_hi);
402 		base |= ((unsigned long) io_base_hi << 16);
403 		limit |= ((unsigned long) io_limit_hi << 16);
404 	}
405 
406 	res->flags = (io_base_lo & PCI_IO_RANGE_TYPE_MASK) | IORESOURCE_IO;
407 
408 	if (base <= limit) {
409 		region.start = base;
410 		region.end = limit + io_granularity - 1;
411 		pcibios_bus_to_resource(dev->bus, res, &region);
412 		if (log)
413 			pci_info(dev, "  bridge window %pR\n", res);
414 	} else {
415 		resource_set_range(res, 0, 0);
416 		res->flags |= IORESOURCE_UNSET | IORESOURCE_DISABLED;
417 	}
418 }
419 
420 static void pci_read_bridge_mmio(struct pci_dev *dev, struct resource *res,
421 				 bool log)
422 {
423 	u16 mem_base_lo, mem_limit_lo;
424 	unsigned long base, limit;
425 	struct pci_bus_region region;
426 
427 	pci_read_config_word(dev, PCI_MEMORY_BASE, &mem_base_lo);
428 	pci_read_config_word(dev, PCI_MEMORY_LIMIT, &mem_limit_lo);
429 	base = ((unsigned long) mem_base_lo & PCI_MEMORY_RANGE_MASK) << 16;
430 	limit = ((unsigned long) mem_limit_lo & PCI_MEMORY_RANGE_MASK) << 16;
431 
432 	res->flags = (mem_base_lo & PCI_MEMORY_RANGE_TYPE_MASK) | IORESOURCE_MEM;
433 
434 	if (base <= limit) {
435 		region.start = base;
436 		region.end = limit + 0xfffff;
437 		pcibios_bus_to_resource(dev->bus, res, &region);
438 		if (log)
439 			pci_info(dev, "  bridge window %pR\n", res);
440 	} else {
441 		resource_set_range(res, 0, 0);
442 		res->flags |= IORESOURCE_UNSET | IORESOURCE_DISABLED;
443 	}
444 }
445 
446 static void pci_read_bridge_mmio_pref(struct pci_dev *dev, struct resource *res,
447 				      bool log)
448 {
449 	u16 mem_base_lo, mem_limit_lo;
450 	u64 base64, limit64;
451 	pci_bus_addr_t base, limit;
452 	struct pci_bus_region region;
453 
454 	if (!dev->pref_window)
455 		return;
456 
457 	pci_read_config_word(dev, PCI_PREF_MEMORY_BASE, &mem_base_lo);
458 	pci_read_config_word(dev, PCI_PREF_MEMORY_LIMIT, &mem_limit_lo);
459 	base64 = (mem_base_lo & PCI_PREF_RANGE_MASK) << 16;
460 	limit64 = (mem_limit_lo & PCI_PREF_RANGE_MASK) << 16;
461 
462 	if ((mem_base_lo & PCI_PREF_RANGE_TYPE_MASK) == PCI_PREF_RANGE_TYPE_64) {
463 		u32 mem_base_hi, mem_limit_hi;
464 
465 		pci_read_config_dword(dev, PCI_PREF_BASE_UPPER32, &mem_base_hi);
466 		pci_read_config_dword(dev, PCI_PREF_LIMIT_UPPER32, &mem_limit_hi);
467 
468 		/*
469 		 * Some bridges set the base > limit by default, and some
470 		 * (broken) BIOSes do not initialize them.  If we find
471 		 * this, just assume they are not being used.
472 		 */
473 		if (mem_base_hi <= mem_limit_hi) {
474 			base64 |= (u64) mem_base_hi << 32;
475 			limit64 |= (u64) mem_limit_hi << 32;
476 		}
477 	}
478 
479 	base = (pci_bus_addr_t) base64;
480 	limit = (pci_bus_addr_t) limit64;
481 
482 	if (base != base64) {
483 		pci_err(dev, "can't handle bridge window above 4GB (bus address %#010llx)\n",
484 			(unsigned long long) base64);
485 		return;
486 	}
487 
488 	res->flags = (mem_base_lo & PCI_PREF_RANGE_TYPE_MASK) | IORESOURCE_MEM |
489 		     IORESOURCE_PREFETCH;
490 	if (res->flags & PCI_PREF_RANGE_TYPE_64)
491 		res->flags |= IORESOURCE_MEM_64;
492 
493 	if (base <= limit) {
494 		region.start = base;
495 		region.end = limit + 0xfffff;
496 		pcibios_bus_to_resource(dev->bus, res, &region);
497 		if (log)
498 			pci_info(dev, "  bridge window %pR\n", res);
499 	} else {
500 		resource_set_range(res, 0, 0);
501 		res->flags |= IORESOURCE_UNSET | IORESOURCE_DISABLED;
502 	}
503 }
504 
505 static void pci_read_bridge_windows(struct pci_dev *bridge)
506 {
507 	u32 buses;
508 	u16 io;
509 	u32 pmem, tmp;
510 	struct resource res;
511 
512 	pci_read_config_dword(bridge, PCI_PRIMARY_BUS, &buses);
513 	res.flags = IORESOURCE_BUS;
514 	res.start = FIELD_GET(PCI_SECONDARY_BUS_MASK, buses);
515 	res.end = FIELD_GET(PCI_SUBORDINATE_BUS_MASK, buses);
516 	pci_info(bridge, "PCI bridge to %pR%s\n", &res,
517 		 bridge->transparent ? " (subtractive decode)" : "");
518 
519 	pci_read_config_word(bridge, PCI_IO_BASE, &io);
520 	if (!io) {
521 		pci_write_config_word(bridge, PCI_IO_BASE, 0xe0f0);
522 		pci_read_config_word(bridge, PCI_IO_BASE, &io);
523 		pci_write_config_word(bridge, PCI_IO_BASE, 0x0);
524 	}
525 	if (io) {
526 		bridge->io_window = 1;
527 		pci_read_bridge_io(bridge, &res, true);
528 	}
529 
530 	pci_read_bridge_mmio(bridge, &res, true);
531 
532 	/*
533 	 * DECchip 21050 pass 2 errata: the bridge may miss an address
534 	 * disconnect boundary by one PCI data phase.  Workaround: do not
535 	 * use prefetching on this device.
536 	 */
537 	if (bridge->vendor == PCI_VENDOR_ID_DEC && bridge->device == 0x0001)
538 		return;
539 
540 	pci_read_config_dword(bridge, PCI_PREF_MEMORY_BASE, &pmem);
541 	if (!pmem) {
542 		pci_write_config_dword(bridge, PCI_PREF_MEMORY_BASE,
543 					       0xffe0fff0);
544 		pci_read_config_dword(bridge, PCI_PREF_MEMORY_BASE, &pmem);
545 		pci_write_config_dword(bridge, PCI_PREF_MEMORY_BASE, 0x0);
546 	}
547 	if (!pmem)
548 		return;
549 
550 	bridge->pref_window = 1;
551 
552 	if ((pmem & PCI_PREF_RANGE_TYPE_MASK) == PCI_PREF_RANGE_TYPE_64) {
553 
554 		/*
555 		 * Bridge claims to have a 64-bit prefetchable memory
556 		 * window; verify that the upper bits are actually
557 		 * writable.
558 		 */
559 		pci_read_config_dword(bridge, PCI_PREF_BASE_UPPER32, &pmem);
560 		pci_write_config_dword(bridge, PCI_PREF_BASE_UPPER32,
561 				       0xffffffff);
562 		pci_read_config_dword(bridge, PCI_PREF_BASE_UPPER32, &tmp);
563 		pci_write_config_dword(bridge, PCI_PREF_BASE_UPPER32, pmem);
564 		if (tmp)
565 			bridge->pref_64_window = 1;
566 	}
567 
568 	pci_read_bridge_mmio_pref(bridge, &res, true);
569 }
570 
571 void pci_read_bridge_bases(struct pci_bus *child)
572 {
573 	struct pci_dev *dev = child->self;
574 	struct resource *res;
575 	int i;
576 
577 	if (pci_is_root_bus(child))	/* It's a host bus, nothing to read */
578 		return;
579 
580 	pci_info(dev, "PCI bridge to %pR%s\n",
581 		 &child->busn_res,
582 		 dev->transparent ? " (subtractive decode)" : "");
583 
584 	pci_bus_remove_resources(child);
585 	for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++)
586 		child->resource[i] = &dev->resource[PCI_BRIDGE_RESOURCES+i];
587 
588 	pci_read_bridge_io(child->self,
589 			   child->resource[PCI_BUS_BRIDGE_IO_WINDOW], false);
590 	pci_read_bridge_mmio(child->self,
591 			     child->resource[PCI_BUS_BRIDGE_MEM_WINDOW], false);
592 	pci_read_bridge_mmio_pref(child->self,
593 				  child->resource[PCI_BUS_BRIDGE_PREF_MEM_WINDOW],
594 				  false);
595 
596 	if (!dev->transparent)
597 		return;
598 
599 	pci_bus_for_each_resource(child->parent, res) {
600 		if (!res || !res->flags)
601 			continue;
602 
603 		pci_bus_add_resource(child, res);
604 		pci_info(dev, "  bridge window %pR (subtractive decode)\n", res);
605 	}
606 }
607 
608 static struct pci_bus *pci_alloc_bus(struct pci_bus *parent)
609 {
610 	struct pci_bus *b;
611 
612 	b = kzalloc_obj(*b);
613 	if (!b)
614 		return NULL;
615 
616 	INIT_LIST_HEAD(&b->node);
617 	INIT_LIST_HEAD(&b->children);
618 	INIT_LIST_HEAD(&b->devices);
619 	INIT_LIST_HEAD(&b->slots);
620 	INIT_LIST_HEAD(&b->resources);
621 	b->max_bus_speed = PCI_SPEED_UNKNOWN;
622 	b->cur_bus_speed = PCI_SPEED_UNKNOWN;
623 #ifdef CONFIG_PCI_DOMAINS_GENERIC
624 	if (parent)
625 		b->domain_nr = parent->domain_nr;
626 #endif
627 	return b;
628 }
629 
630 static void pci_release_host_bridge_dev(struct device *dev)
631 {
632 	struct pci_host_bridge *bridge = to_pci_host_bridge(dev);
633 
634 	if (bridge->release_fn)
635 		bridge->release_fn(bridge);
636 
637 	pci_free_resource_list(&bridge->windows);
638 	pci_free_resource_list(&bridge->dma_ranges);
639 
640 	/* Host bridges only have domain_nr set in the emulation case */
641 	if (bridge->domain_nr != PCI_DOMAIN_NR_NOT_SET)
642 		pci_bus_release_emul_domain_nr(bridge->domain_nr);
643 
644 	kfree(bridge);
645 }
646 
647 static const struct attribute_group *pci_host_bridge_groups[] = {
648 #ifdef CONFIG_PCI_IDE
649 	&pci_ide_attr_group,
650 #endif
651 	NULL
652 };
653 
654 static const struct device_type pci_host_bridge_type = {
655 	.groups = pci_host_bridge_groups,
656 	.release = pci_release_host_bridge_dev,
657 };
658 
659 static void pci_init_host_bridge(struct pci_host_bridge *bridge)
660 {
661 	INIT_LIST_HEAD(&bridge->windows);
662 	INIT_LIST_HEAD(&bridge->dma_ranges);
663 	INIT_LIST_HEAD(&bridge->ports);
664 
665 	/*
666 	 * We assume we can manage these PCIe features.  Some systems may
667 	 * reserve these for use by the platform itself, e.g., an ACPI BIOS
668 	 * may implement its own AER handling and use _OSC to prevent the
669 	 * OS from interfering.
670 	 */
671 	bridge->native_aer = 1;
672 	bridge->native_pcie_hotplug = 1;
673 	bridge->native_shpc_hotplug = 1;
674 	bridge->native_pme = 1;
675 	bridge->native_ltr = 1;
676 	bridge->native_dpc = 1;
677 	bridge->domain_nr = PCI_DOMAIN_NR_NOT_SET;
678 	bridge->native_cxl_error = 1;
679 	bridge->dev.type = &pci_host_bridge_type;
680 	pci_ide_init_host_bridge(bridge);
681 
682 	device_initialize(&bridge->dev);
683 }
684 
685 struct pci_host_bridge *pci_alloc_host_bridge(size_t priv)
686 {
687 	struct pci_host_bridge *bridge;
688 
689 	bridge = kzalloc(sizeof(*bridge) + priv, GFP_KERNEL);
690 	if (!bridge)
691 		return NULL;
692 
693 	pci_init_host_bridge(bridge);
694 
695 	return bridge;
696 }
697 EXPORT_SYMBOL(pci_alloc_host_bridge);
698 
699 static void devm_pci_alloc_host_bridge_release(void *data)
700 {
701 	pci_free_host_bridge(data);
702 }
703 
704 struct pci_host_bridge *devm_pci_alloc_host_bridge(struct device *dev,
705 						   size_t priv)
706 {
707 	int ret;
708 	struct pci_host_bridge *bridge;
709 
710 	bridge = pci_alloc_host_bridge(priv);
711 	if (!bridge)
712 		return NULL;
713 
714 	bridge->dev.parent = dev;
715 
716 	ret = devm_add_action_or_reset(dev, devm_pci_alloc_host_bridge_release,
717 				       bridge);
718 	if (ret)
719 		return NULL;
720 
721 	ret = devm_of_pci_bridge_init(dev, bridge);
722 	if (ret)
723 		return NULL;
724 
725 	return bridge;
726 }
727 EXPORT_SYMBOL(devm_pci_alloc_host_bridge);
728 
729 void pci_free_host_bridge(struct pci_host_bridge *bridge)
730 {
731 	put_device(&bridge->dev);
732 }
733 EXPORT_SYMBOL(pci_free_host_bridge);
734 
735 /* Indexed by PCI_X_SSTATUS_FREQ (secondary bus mode and frequency) */
736 static const unsigned char pcix_bus_speed[] = {
737 	PCI_SPEED_UNKNOWN,		/* 0 */
738 	PCI_SPEED_66MHz_PCIX,		/* 1 */
739 	PCI_SPEED_100MHz_PCIX,		/* 2 */
740 	PCI_SPEED_133MHz_PCIX,		/* 3 */
741 	PCI_SPEED_UNKNOWN,		/* 4 */
742 	PCI_SPEED_66MHz_PCIX_ECC,	/* 5 */
743 	PCI_SPEED_100MHz_PCIX_ECC,	/* 6 */
744 	PCI_SPEED_133MHz_PCIX_ECC,	/* 7 */
745 	PCI_SPEED_UNKNOWN,		/* 8 */
746 	PCI_SPEED_66MHz_PCIX_266,	/* 9 */
747 	PCI_SPEED_100MHz_PCIX_266,	/* A */
748 	PCI_SPEED_133MHz_PCIX_266,	/* B */
749 	PCI_SPEED_UNKNOWN,		/* C */
750 	PCI_SPEED_66MHz_PCIX_533,	/* D */
751 	PCI_SPEED_100MHz_PCIX_533,	/* E */
752 	PCI_SPEED_133MHz_PCIX_533	/* F */
753 };
754 
755 /* Indexed by PCI_EXP_LNKCAP_SLS, PCI_EXP_LNKSTA_CLS */
756 const unsigned char pcie_link_speed[] = {
757 	PCI_SPEED_UNKNOWN,		/* 0 */
758 	PCIE_SPEED_2_5GT,		/* 1 */
759 	PCIE_SPEED_5_0GT,		/* 2 */
760 	PCIE_SPEED_8_0GT,		/* 3 */
761 	PCIE_SPEED_16_0GT,		/* 4 */
762 	PCIE_SPEED_32_0GT,		/* 5 */
763 	PCIE_SPEED_64_0GT,		/* 6 */
764 	PCI_SPEED_UNKNOWN,		/* 7 */
765 	PCI_SPEED_UNKNOWN,		/* 8 */
766 	PCI_SPEED_UNKNOWN,		/* 9 */
767 	PCI_SPEED_UNKNOWN,		/* A */
768 	PCI_SPEED_UNKNOWN,		/* B */
769 	PCI_SPEED_UNKNOWN,		/* C */
770 	PCI_SPEED_UNKNOWN,		/* D */
771 	PCI_SPEED_UNKNOWN,		/* E */
772 	PCI_SPEED_UNKNOWN		/* F */
773 };
774 EXPORT_SYMBOL_GPL(pcie_link_speed);
775 
776 /**
777  * pcie_get_link_speed - Get speed value from PCIe generation number
778  * @speed: PCIe speed (1-based: 1 = 2.5GT, 2 = 5GT, ...)
779  *
780  * Returns the speed value (e.g., PCIE_SPEED_2_5GT) if @speed is valid,
781  * otherwise returns PCI_SPEED_UNKNOWN.
782  */
783 unsigned char pcie_get_link_speed(unsigned int speed)
784 {
785 	if (speed >= ARRAY_SIZE(pcie_link_speed))
786 		return PCI_SPEED_UNKNOWN;
787 
788 	return pcie_link_speed[speed];
789 }
790 EXPORT_SYMBOL_GPL(pcie_get_link_speed);
791 
792 const char *pci_speed_string(enum pci_bus_speed speed)
793 {
794 	/* Indexed by the pci_bus_speed enum */
795 	static const char *speed_strings[] = {
796 	    "33 MHz PCI",		/* 0x00 */
797 	    "66 MHz PCI",		/* 0x01 */
798 	    "66 MHz PCI-X",		/* 0x02 */
799 	    "100 MHz PCI-X",		/* 0x03 */
800 	    "133 MHz PCI-X",		/* 0x04 */
801 	    NULL,			/* 0x05 */
802 	    NULL,			/* 0x06 */
803 	    NULL,			/* 0x07 */
804 	    NULL,			/* 0x08 */
805 	    "66 MHz PCI-X 266",		/* 0x09 */
806 	    "100 MHz PCI-X 266",	/* 0x0a */
807 	    "133 MHz PCI-X 266",	/* 0x0b */
808 	    "Unknown AGP",		/* 0x0c */
809 	    "1x AGP",			/* 0x0d */
810 	    "2x AGP",			/* 0x0e */
811 	    "4x AGP",			/* 0x0f */
812 	    "8x AGP",			/* 0x10 */
813 	    "66 MHz PCI-X 533",		/* 0x11 */
814 	    "100 MHz PCI-X 533",	/* 0x12 */
815 	    "133 MHz PCI-X 533",	/* 0x13 */
816 	    "2.5 GT/s PCIe",		/* 0x14 */
817 	    "5.0 GT/s PCIe",		/* 0x15 */
818 	    "8.0 GT/s PCIe",		/* 0x16 */
819 	    "16.0 GT/s PCIe",		/* 0x17 */
820 	    "32.0 GT/s PCIe",		/* 0x18 */
821 	    "64.0 GT/s PCIe",		/* 0x19 */
822 	};
823 
824 	if (speed < ARRAY_SIZE(speed_strings))
825 		return speed_strings[speed];
826 	return "Unknown";
827 }
828 EXPORT_SYMBOL_GPL(pci_speed_string);
829 
830 void pcie_update_link_speed(struct pci_bus *bus,
831 			    enum pcie_link_change_reason reason)
832 {
833 	struct pci_dev *bridge = bus->self;
834 	u16 linksta, linksta2;
835 
836 	pcie_capability_read_word(bridge, PCI_EXP_LNKSTA, &linksta);
837 	pcie_capability_read_word(bridge, PCI_EXP_LNKSTA2, &linksta2);
838 
839 	__pcie_update_link_speed(bus, reason, linksta, linksta2);
840 }
841 EXPORT_SYMBOL_GPL(pcie_update_link_speed);
842 
843 static unsigned char agp_speeds[] = {
844 	AGP_UNKNOWN,
845 	AGP_1X,
846 	AGP_2X,
847 	AGP_4X,
848 	AGP_8X
849 };
850 
851 static enum pci_bus_speed agp_speed(int agp3, int agpstat)
852 {
853 	int index = 0;
854 
855 	if (agpstat & 4)
856 		index = 3;
857 	else if (agpstat & 2)
858 		index = 2;
859 	else if (agpstat & 1)
860 		index = 1;
861 	else
862 		goto out;
863 
864 	if (agp3) {
865 		index += 2;
866 		if (index == 5)
867 			index = 0;
868 	}
869 
870  out:
871 	return agp_speeds[index];
872 }
873 
874 static void pci_set_bus_speed(struct pci_bus *bus)
875 {
876 	struct pci_dev *bridge = bus->self;
877 	int pos;
878 
879 	pos = pci_find_capability(bridge, PCI_CAP_ID_AGP);
880 	if (!pos)
881 		pos = pci_find_capability(bridge, PCI_CAP_ID_AGP3);
882 	if (pos) {
883 		u32 agpstat, agpcmd;
884 
885 		pci_read_config_dword(bridge, pos + PCI_AGP_STATUS, &agpstat);
886 		bus->max_bus_speed = agp_speed(agpstat & 8, agpstat & 7);
887 
888 		pci_read_config_dword(bridge, pos + PCI_AGP_COMMAND, &agpcmd);
889 		bus->cur_bus_speed = agp_speed(agpstat & 8, agpcmd & 7);
890 	}
891 
892 	pos = pci_find_capability(bridge, PCI_CAP_ID_PCIX);
893 	if (pos) {
894 		u16 status;
895 		enum pci_bus_speed max;
896 
897 		pci_read_config_word(bridge, pos + PCI_X_BRIDGE_SSTATUS,
898 				     &status);
899 
900 		if (status & PCI_X_SSTATUS_533MHZ) {
901 			max = PCI_SPEED_133MHz_PCIX_533;
902 		} else if (status & PCI_X_SSTATUS_266MHZ) {
903 			max = PCI_SPEED_133MHz_PCIX_266;
904 		} else if (status & PCI_X_SSTATUS_133MHZ) {
905 			if ((status & PCI_X_SSTATUS_VERS) == PCI_X_SSTATUS_V2)
906 				max = PCI_SPEED_133MHz_PCIX_ECC;
907 			else
908 				max = PCI_SPEED_133MHz_PCIX;
909 		} else {
910 			max = PCI_SPEED_66MHz_PCIX;
911 		}
912 
913 		bus->max_bus_speed = max;
914 		bus->cur_bus_speed =
915 			pcix_bus_speed[FIELD_GET(PCI_X_SSTATUS_FREQ, status)];
916 
917 		return;
918 	}
919 
920 	if (pci_is_pcie(bridge)) {
921 		u32 linkcap;
922 
923 		pcie_capability_read_dword(bridge, PCI_EXP_LNKCAP, &linkcap);
924 		bus->max_bus_speed = pcie_link_speed[linkcap & PCI_EXP_LNKCAP_SLS];
925 
926 		pcie_update_link_speed(bus, PCIE_ADD_BUS);
927 	}
928 }
929 
930 static struct irq_domain *pci_host_bridge_msi_domain(struct pci_bus *bus)
931 {
932 	struct irq_domain *d;
933 
934 	/* If the host bridge driver sets a MSI domain of the bridge, use it */
935 	d = dev_get_msi_domain(bus->bridge);
936 
937 	/*
938 	 * Any firmware interface that can resolve the msi_domain
939 	 * should be called from here.
940 	 */
941 	if (!d)
942 		d = pci_host_bridge_of_msi_domain(bus);
943 	if (!d)
944 		d = pci_host_bridge_acpi_msi_domain(bus);
945 
946 	/*
947 	 * If no IRQ domain was found via the OF tree, try looking it up
948 	 * directly through the fwnode_handle.
949 	 */
950 	if (!d) {
951 		struct fwnode_handle *fwnode = pci_root_bus_fwnode(bus);
952 
953 		if (fwnode)
954 			d = irq_find_matching_fwnode(fwnode,
955 						     DOMAIN_BUS_PCI_MSI);
956 	}
957 
958 	return d;
959 }
960 
961 static void pci_set_bus_msi_domain(struct pci_bus *bus)
962 {
963 	struct irq_domain *d;
964 	struct pci_bus *b;
965 
966 	/*
967 	 * The bus can be a root bus, a subordinate bus, or a virtual bus
968 	 * created by an SR-IOV device.  Walk up to the first bridge device
969 	 * found or derive the domain from the host bridge.
970 	 */
971 	for (b = bus, d = NULL; !d && !pci_is_root_bus(b); b = b->parent) {
972 		if (b->self)
973 			d = dev_get_msi_domain(&b->self->dev);
974 	}
975 
976 	if (!d)
977 		d = pci_host_bridge_msi_domain(b);
978 
979 	dev_set_msi_domain(&bus->dev, d);
980 }
981 
982 static bool pci_preserve_config(struct pci_host_bridge *host_bridge)
983 {
984 	if (pci_acpi_preserve_config(host_bridge))
985 		return true;
986 
987 	if (host_bridge->dev.parent && host_bridge->dev.parent->of_node)
988 		return of_pci_preserve_config(host_bridge->dev.parent->of_node);
989 
990 	return false;
991 }
992 
993 static int pci_register_host_bridge(struct pci_host_bridge *bridge)
994 {
995 	struct device *parent = bridge->dev.parent;
996 	struct resource_entry *window, *next, *n;
997 	struct pci_bus *bus, *b;
998 	resource_size_t offset, next_offset;
999 	LIST_HEAD(resources);
1000 	struct resource *res, *next_res;
1001 	bool bus_registered = false;
1002 	char addr[64], *fmt;
1003 	const char *name;
1004 	int err;
1005 
1006 	bus = pci_alloc_bus(NULL);
1007 	if (!bus)
1008 		return -ENOMEM;
1009 
1010 	bridge->bus = bus;
1011 
1012 	bus->sysdata = bridge->sysdata;
1013 	bus->ops = bridge->ops;
1014 	bus->number = bus->busn_res.start = bridge->busnr;
1015 #ifdef CONFIG_PCI_DOMAINS_GENERIC
1016 	if (bridge->domain_nr == PCI_DOMAIN_NR_NOT_SET)
1017 		bus->domain_nr = pci_bus_find_domain_nr(bus, parent);
1018 	else
1019 		bus->domain_nr = bridge->domain_nr;
1020 	if (bus->domain_nr < 0) {
1021 		err = bus->domain_nr;
1022 		goto free;
1023 	}
1024 #endif
1025 
1026 	b = pci_find_bus(pci_domain_nr(bus), bridge->busnr);
1027 	if (b) {
1028 		/* Ignore it if we already got here via a different bridge */
1029 		dev_dbg(&b->dev, "bus already known\n");
1030 		err = -EEXIST;
1031 		goto free;
1032 	}
1033 
1034 	dev_set_name(&bridge->dev, "pci%04x:%02x", pci_domain_nr(bus),
1035 		     bridge->busnr);
1036 
1037 	err = pcibios_root_bridge_prepare(bridge);
1038 	if (err)
1039 		goto free;
1040 
1041 	/* Temporarily move resources off the list */
1042 	list_splice_init(&bridge->windows, &resources);
1043 	err = device_add(&bridge->dev);
1044 	if (err)
1045 		goto free;
1046 
1047 	bus->bridge = get_device(&bridge->dev);
1048 	device_enable_async_suspend(bus->bridge);
1049 	pci_set_bus_of_node(bus);
1050 	pci_set_bus_msi_domain(bus);
1051 	if (bridge->msi_domain && !dev_get_msi_domain(&bus->dev) &&
1052 	    !pci_host_of_has_msi_map(parent))
1053 		bus->bus_flags |= PCI_BUS_FLAGS_NO_MSI;
1054 
1055 	if (!parent)
1056 		set_dev_node(bus->bridge, pcibus_to_node(bus));
1057 
1058 	bus->dev.class = &pcibus_class;
1059 	bus->dev.parent = bus->bridge;
1060 
1061 	dev_set_name(&bus->dev, "%04x:%02x", pci_domain_nr(bus), bus->number);
1062 	name = dev_name(&bus->dev);
1063 
1064 	err = device_register(&bus->dev);
1065 	bus_registered = true;
1066 	if (err)
1067 		goto unregister;
1068 
1069 	pcibios_add_bus(bus);
1070 
1071 	if (bus->ops->add_bus) {
1072 		err = bus->ops->add_bus(bus);
1073 		if (WARN_ON(err < 0))
1074 			dev_err(&bus->dev, "failed to add bus: %d\n", err);
1075 	}
1076 
1077 	if (parent)
1078 		dev_info(parent, "PCI host bridge to bus %s\n", name);
1079 	else
1080 		pr_info("PCI host bridge to bus %s\n", name);
1081 
1082 	if (nr_node_ids > 1 && pcibus_to_node(bus) == NUMA_NO_NODE)
1083 		dev_warn(&bus->dev, "Unknown NUMA node; performance will be reduced\n");
1084 
1085 	/* Check if the boot configuration by FW needs to be preserved */
1086 	bridge->preserve_config = pci_preserve_config(bridge);
1087 
1088 	/* Coalesce contiguous windows */
1089 	resource_list_for_each_entry_safe(window, n, &resources) {
1090 		if (list_is_last(&window->node, &resources))
1091 			break;
1092 
1093 		next = list_next_entry(window, node);
1094 		offset = window->offset;
1095 		res = window->res;
1096 		next_offset = next->offset;
1097 		next_res = next->res;
1098 
1099 		if (res->flags != next_res->flags || offset != next_offset)
1100 			continue;
1101 
1102 		if (res->end + 1 == next_res->start) {
1103 			next_res->start = res->start;
1104 			res->flags = res->start = res->end = 0;
1105 		}
1106 	}
1107 
1108 	/* Add initial resources to the bus */
1109 	resource_list_for_each_entry_safe(window, n, &resources) {
1110 		offset = window->offset;
1111 		res = window->res;
1112 		if (!res->flags && !res->start && !res->end) {
1113 			release_resource(res);
1114 			resource_list_destroy_entry(window);
1115 			continue;
1116 		}
1117 
1118 		list_move_tail(&window->node, &bridge->windows);
1119 
1120 		if (res->flags & IORESOURCE_BUS)
1121 			pci_bus_insert_busn_res(bus, bus->number, res->end);
1122 		else
1123 			pci_bus_add_resource(bus, res);
1124 
1125 		if (offset) {
1126 			if (resource_type(res) == IORESOURCE_IO)
1127 				fmt = " (bus address [%#06llx-%#06llx])";
1128 			else
1129 				fmt = " (bus address [%#010llx-%#010llx])";
1130 
1131 			snprintf(addr, sizeof(addr), fmt,
1132 				 (unsigned long long)(res->start - offset),
1133 				 (unsigned long long)(res->end - offset));
1134 		} else
1135 			addr[0] = '\0';
1136 
1137 		dev_info(&bus->dev, "root bus resource %pR%s\n", res, addr);
1138 	}
1139 
1140 	of_pci_make_host_bridge_node(bridge);
1141 
1142 	down_write(&pci_bus_sem);
1143 	list_add_tail(&bus->node, &pci_root_buses);
1144 	up_write(&pci_bus_sem);
1145 
1146 	return 0;
1147 
1148 unregister:
1149 	put_device(&bridge->dev);
1150 	device_del(&bridge->dev);
1151 free:
1152 #ifdef CONFIG_PCI_DOMAINS_GENERIC
1153 	if (bridge->domain_nr == PCI_DOMAIN_NR_NOT_SET)
1154 		pci_bus_release_domain_nr(parent, bus->domain_nr);
1155 #endif
1156 	if (bus_registered)
1157 		put_device(&bus->dev);
1158 	else
1159 		kfree(bus);
1160 
1161 	return err;
1162 }
1163 
1164 static bool pci_bridge_child_ext_cfg_accessible(struct pci_dev *bridge)
1165 {
1166 	int pos;
1167 	u32 status;
1168 
1169 	/*
1170 	 * If extended config space isn't accessible on a bridge's primary
1171 	 * bus, we certainly can't access it on the secondary bus.
1172 	 */
1173 	if (bridge->bus->bus_flags & PCI_BUS_FLAGS_NO_EXTCFG)
1174 		return false;
1175 
1176 	/*
1177 	 * PCIe Root Ports and switch ports are PCIe on both sides, so if
1178 	 * extended config space is accessible on the primary, it's also
1179 	 * accessible on the secondary.
1180 	 */
1181 	if (pci_is_pcie(bridge) &&
1182 	    (pci_pcie_type(bridge) == PCI_EXP_TYPE_ROOT_PORT ||
1183 	     pci_pcie_type(bridge) == PCI_EXP_TYPE_UPSTREAM ||
1184 	     pci_pcie_type(bridge) == PCI_EXP_TYPE_DOWNSTREAM))
1185 		return true;
1186 
1187 	/*
1188 	 * For the other bridge types:
1189 	 *   - PCI-to-PCI bridges
1190 	 *   - PCIe-to-PCI/PCI-X forward bridges
1191 	 *   - PCI/PCI-X-to-PCIe reverse bridges
1192 	 * extended config space on the secondary side is only accessible
1193 	 * if the bridge supports PCI-X Mode 2.
1194 	 */
1195 	pos = pci_find_capability(bridge, PCI_CAP_ID_PCIX);
1196 	if (!pos)
1197 		return false;
1198 
1199 	pci_read_config_dword(bridge, pos + PCI_X_STATUS, &status);
1200 	return status & (PCI_X_STATUS_266MHZ | PCI_X_STATUS_533MHZ);
1201 }
1202 
1203 static struct pci_bus *pci_alloc_child_bus(struct pci_bus *parent,
1204 					   struct pci_dev *bridge, int busnr)
1205 {
1206 	struct pci_bus *child;
1207 	struct pci_host_bridge *host;
1208 	int i;
1209 	int ret;
1210 
1211 	/* Allocate a new bus and inherit stuff from the parent */
1212 	child = pci_alloc_bus(parent);
1213 	if (!child)
1214 		return NULL;
1215 
1216 	child->parent = parent;
1217 	child->sysdata = parent->sysdata;
1218 	child->bus_flags = parent->bus_flags;
1219 
1220 	host = pci_find_host_bridge(parent);
1221 	if (host->child_ops)
1222 		child->ops = host->child_ops;
1223 	else
1224 		child->ops = parent->ops;
1225 
1226 	/*
1227 	 * Initialize some portions of the bus device, but don't register
1228 	 * it now as the parent is not properly set up yet.
1229 	 */
1230 	child->dev.class = &pcibus_class;
1231 	dev_set_name(&child->dev, "%04x:%02x", pci_domain_nr(child), busnr);
1232 
1233 	/* Set up the primary, secondary and subordinate bus numbers */
1234 	child->number = child->busn_res.start = busnr;
1235 	child->primary = parent->busn_res.start;
1236 	child->busn_res.end = 0xff;
1237 
1238 	if (!bridge) {
1239 		child->dev.parent = parent->bridge;
1240 		goto add_dev;
1241 	}
1242 
1243 	child->self = bridge;
1244 	child->bridge = get_device(&bridge->dev);
1245 	child->dev.parent = child->bridge;
1246 	pci_set_bus_of_node(child);
1247 	pci_set_bus_speed(child);
1248 
1249 	/*
1250 	 * Check whether extended config space is accessible on the child
1251 	 * bus.  Note that we currently assume it is always accessible on
1252 	 * the root bus.
1253 	 */
1254 	if (!pci_bridge_child_ext_cfg_accessible(bridge)) {
1255 		child->bus_flags |= PCI_BUS_FLAGS_NO_EXTCFG;
1256 		pci_info(child, "extended config space not accessible\n");
1257 	}
1258 
1259 	/* Set up default resource pointers and names */
1260 	for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++) {
1261 		child->resource[i] = &bridge->resource[PCI_BRIDGE_RESOURCES+i];
1262 		child->resource[i]->name = child->name;
1263 	}
1264 	bridge->subordinate = child;
1265 
1266 add_dev:
1267 	pci_set_bus_msi_domain(child);
1268 	ret = device_register(&child->dev);
1269 	if (WARN_ON(ret < 0)) {
1270 		put_device(&child->dev);
1271 		return NULL;
1272 	}
1273 
1274 	pcibios_add_bus(child);
1275 
1276 	if (child->ops->add_bus) {
1277 		ret = child->ops->add_bus(child);
1278 		if (WARN_ON(ret < 0))
1279 			dev_err(&child->dev, "failed to add bus: %d\n", ret);
1280 	}
1281 
1282 	return child;
1283 }
1284 
1285 struct pci_bus *pci_add_new_bus(struct pci_bus *parent, struct pci_dev *dev,
1286 				int busnr)
1287 {
1288 	struct pci_bus *child;
1289 
1290 	child = pci_alloc_child_bus(parent, dev, busnr);
1291 	if (child) {
1292 		down_write(&pci_bus_sem);
1293 		list_add_tail(&child->node, &parent->children);
1294 		up_write(&pci_bus_sem);
1295 	}
1296 	return child;
1297 }
1298 EXPORT_SYMBOL(pci_add_new_bus);
1299 
1300 static void pci_enable_rrs_sv(struct pci_dev *pdev)
1301 {
1302 	u16 root_cap = 0;
1303 
1304 	/* Enable Configuration RRS Software Visibility if supported */
1305 	pcie_capability_read_word(pdev, PCI_EXP_RTCAP, &root_cap);
1306 	if (root_cap & PCI_EXP_RTCAP_RRS_SV) {
1307 		pcie_capability_set_word(pdev, PCI_EXP_RTCTL,
1308 					 PCI_EXP_RTCTL_RRS_SVE);
1309 		pdev->config_rrs_sv = 1;
1310 	}
1311 }
1312 
1313 static unsigned int pci_scan_child_bus_extend(struct pci_bus *bus,
1314 					      unsigned int available_buses);
1315 
1316 void pbus_validate_busn(struct pci_bus *bus)
1317 {
1318 	struct pci_bus *upstream = bus->parent;
1319 	struct pci_dev *bridge = bus->self;
1320 
1321 	/* Check that all devices are accessible */
1322 	while (upstream->parent) {
1323 		if ((bus->busn_res.end > upstream->busn_res.end) ||
1324 		    (bus->number > upstream->busn_res.end) ||
1325 		    (bus->number < upstream->number) ||
1326 		    (bus->busn_res.end < upstream->number)) {
1327 			pci_info(bridge, "devices behind bridge are unusable because %pR cannot be assigned for them\n",
1328 				 &bus->busn_res);
1329 			break;
1330 		}
1331 		upstream = upstream->parent;
1332 	}
1333 }
1334 
1335 /**
1336  * pci_ea_fixed_busnrs() - Read fixed Secondary and Subordinate bus
1337  * numbers from EA capability.
1338  * @dev: Bridge
1339  * @sec: updated with secondary bus number from EA
1340  * @sub: updated with subordinate bus number from EA
1341  *
1342  * If @dev is a bridge with EA capability that specifies valid secondary
1343  * and subordinate bus numbers, return true with the bus numbers in @sec
1344  * and @sub.  Otherwise return false.
1345  */
1346 bool pci_ea_fixed_busnrs(struct pci_dev *dev, u8 *sec, u8 *sub)
1347 {
1348 	int ea, offset;
1349 	u32 dw;
1350 	u8 ea_sec, ea_sub;
1351 
1352 	if (dev->hdr_type != PCI_HEADER_TYPE_BRIDGE)
1353 		return false;
1354 
1355 	/* find PCI EA capability in list */
1356 	ea = pci_find_capability(dev, PCI_CAP_ID_EA);
1357 	if (!ea)
1358 		return false;
1359 
1360 	offset = ea + PCI_EA_FIRST_ENT;
1361 	pci_read_config_dword(dev, offset, &dw);
1362 	ea_sec = FIELD_GET(PCI_EA_SEC_BUS_MASK, dw);
1363 	ea_sub = FIELD_GET(PCI_EA_SUB_BUS_MASK, dw);
1364 	if (ea_sec  == 0 || ea_sub < ea_sec)
1365 		return false;
1366 
1367 	*sec = ea_sec;
1368 	*sub = ea_sub;
1369 	return true;
1370 }
1371 
1372 /*
1373  * pci_scan_bridge_extend() - Scan buses behind a bridge
1374  * @bus: Parent bus the bridge is on
1375  * @dev: Bridge itself
1376  * @max: Starting subordinate number of buses behind this bridge
1377  * @available_buses: Total number of buses available for this bridge and
1378  *		     the devices below. After the minimal bus space has
1379  *		     been allocated the remaining buses will be
1380  *		     distributed equally between hotplug-capable bridges.
1381  * @pass: Either %0 (scan already configured bridges) or %1 (scan bridges
1382  *        that need to be reconfigured.
1383  *
1384  * If it's a bridge, configure it and scan the bus behind it.
1385  * For CardBus bridges, we don't scan behind as the devices will
1386  * be handled by the bridge driver itself.
1387  *
1388  * We need to process bridges in two passes -- first we scan those
1389  * already configured by the BIOS and after we are done with all of
1390  * them, we proceed to assigning numbers to the remaining buses in
1391  * order to avoid overlaps between old and new bus numbers.
1392  *
1393  * Return: New subordinate number covering all buses behind this bridge.
1394  */
1395 static int pci_scan_bridge_extend(struct pci_bus *bus, struct pci_dev *dev,
1396 				  int max, unsigned int available_buses,
1397 				  int pass)
1398 {
1399 	struct pci_bus *child;
1400 	u32 buses;
1401 	u16 bctl;
1402 	u8 primary, secondary, subordinate;
1403 	int broken = 0;
1404 	bool fixed_buses;
1405 	u8 fixed_sec, fixed_sub;
1406 	int next_busnr;
1407 
1408 	/*
1409 	 * Make sure the bridge is powered on to be able to access config
1410 	 * space of devices below it.
1411 	 */
1412 	pm_runtime_get_sync(&dev->dev);
1413 
1414 	pci_read_config_dword(dev, PCI_PRIMARY_BUS, &buses);
1415 	primary = FIELD_GET(PCI_PRIMARY_BUS_MASK, buses);
1416 	secondary = FIELD_GET(PCI_SECONDARY_BUS_MASK, buses);
1417 	subordinate = FIELD_GET(PCI_SUBORDINATE_BUS_MASK, buses);
1418 
1419 	pci_dbg(dev, "scanning [bus %02x-%02x] behind bridge, pass %d\n",
1420 		secondary, subordinate, pass);
1421 
1422 	if (!primary && (primary != bus->number) && secondary && subordinate) {
1423 		pci_warn(dev, "Primary bus is hard wired to 0\n");
1424 		primary = bus->number;
1425 	}
1426 
1427 	/* Check if setup is sensible at all */
1428 	if (!pass &&
1429 	    (primary != bus->number || secondary <= bus->number ||
1430 	     secondary > subordinate)) {
1431 		pci_info(dev, "bridge configuration invalid ([bus %02x-%02x]), reconfiguring\n",
1432 			 secondary, subordinate);
1433 		broken = 1;
1434 	}
1435 
1436 	/*
1437 	 * Disable Master-Abort Mode during probing to avoid reporting of
1438 	 * bus errors in some architectures.
1439 	 */
1440 	pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &bctl);
1441 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL,
1442 			      bctl & ~PCI_BRIDGE_CTL_MASTER_ABORT);
1443 
1444 	if (pci_is_cardbus_bridge(dev)) {
1445 		max = pci_cardbus_scan_bridge_extend(bus, dev, buses, max,
1446 						     available_buses,
1447 						     pass);
1448 		goto out;
1449 	}
1450 
1451 	if ((secondary || subordinate) &&
1452 	    !pcibios_assign_all_busses() && !broken) {
1453 		unsigned int cmax, buses;
1454 
1455 		/*
1456 		 * Bus already configured by firmware, process it in the
1457 		 * first pass and just note the configuration.
1458 		 */
1459 		if (pass)
1460 			goto out;
1461 
1462 		/*
1463 		 * The bus might already exist for two reasons: Either we
1464 		 * are rescanning the bus or the bus is reachable through
1465 		 * more than one bridge. The second case can happen with
1466 		 * the i450NX chipset.
1467 		 */
1468 		child = pci_find_bus(pci_domain_nr(bus), secondary);
1469 		if (!child) {
1470 			child = pci_add_new_bus(bus, dev, secondary);
1471 			if (!child)
1472 				goto out;
1473 			child->primary = primary;
1474 			pci_bus_insert_busn_res(child, secondary, subordinate);
1475 			child->bridge_ctl = bctl;
1476 		}
1477 
1478 		buses = subordinate - secondary;
1479 		cmax = pci_scan_child_bus_extend(child, buses);
1480 		if (cmax > subordinate)
1481 			pci_warn(dev, "bridge has subordinate %02x but max busn %02x\n",
1482 				 subordinate, cmax);
1483 
1484 		/* Subordinate should equal child->busn_res.end */
1485 		if (subordinate > max)
1486 			max = subordinate;
1487 	} else {
1488 
1489 		/*
1490 		 * We need to assign a number to this bus which we always
1491 		 * do in the second pass.
1492 		 */
1493 		if (!pass) {
1494 			if (pcibios_assign_all_busses() || broken)
1495 
1496 				/*
1497 				 * Temporarily disable forwarding of the
1498 				 * configuration cycles on all bridges in
1499 				 * this bus segment to avoid possible
1500 				 * conflicts in the second pass between two
1501 				 * bridges programmed with overlapping bus
1502 				 * ranges.
1503 				 */
1504 				pci_write_config_dword(dev, PCI_PRIMARY_BUS,
1505 						       buses & PCI_SEC_LATENCY_TIMER_MASK);
1506 			goto out;
1507 		}
1508 
1509 		/* Clear errors */
1510 		pci_write_config_word(dev, PCI_STATUS, 0xffff);
1511 
1512 		/* Read bus numbers from EA Capability (if present) */
1513 		fixed_buses = pci_ea_fixed_busnrs(dev, &fixed_sec, &fixed_sub);
1514 		if (fixed_buses)
1515 			next_busnr = fixed_sec;
1516 		else
1517 			next_busnr = max + 1;
1518 
1519 		/*
1520 		 * Prevent assigning a bus number that already exists.
1521 		 * This can happen when a bridge is hot-plugged, so in this
1522 		 * case we only re-scan this bus.
1523 		 */
1524 		child = pci_find_bus(pci_domain_nr(bus), next_busnr);
1525 		if (!child) {
1526 			child = pci_add_new_bus(bus, dev, next_busnr);
1527 			if (!child)
1528 				goto out;
1529 			pci_bus_insert_busn_res(child, next_busnr,
1530 						bus->busn_res.end);
1531 		}
1532 		max++;
1533 		if (available_buses)
1534 			available_buses--;
1535 
1536 		buses = (buses & PCI_SEC_LATENCY_TIMER_MASK) |
1537 			FIELD_PREP(PCI_PRIMARY_BUS_MASK, child->primary) |
1538 			FIELD_PREP(PCI_SECONDARY_BUS_MASK, child->busn_res.start) |
1539 			FIELD_PREP(PCI_SUBORDINATE_BUS_MASK, child->busn_res.end);
1540 
1541 		/* We need to blast all three values with a single write */
1542 		pci_write_config_dword(dev, PCI_PRIMARY_BUS, buses);
1543 
1544 		child->bridge_ctl = bctl;
1545 		max = pci_scan_child_bus_extend(child, available_buses);
1546 
1547 		/*
1548 		 * Set subordinate bus number to its real value.
1549 		 * If fixed subordinate bus number exists from EA
1550 		 * capability then use it.
1551 		 */
1552 		if (fixed_buses)
1553 			max = fixed_sub;
1554 		pci_bus_update_busn_res_end(child, max);
1555 		pci_write_config_byte(dev, PCI_SUBORDINATE_BUS, max);
1556 	}
1557 	scnprintf(child->name, sizeof(child->name), "PCI Bus %04x:%02x",
1558 		  pci_domain_nr(bus), child->number);
1559 
1560 	pbus_validate_busn(child);
1561 
1562 out:
1563 	/* Clear errors in the Secondary Status Register */
1564 	pci_write_config_word(dev, PCI_SEC_STATUS, 0xffff);
1565 
1566 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL, bctl);
1567 
1568 	pm_runtime_put(&dev->dev);
1569 
1570 	return max;
1571 }
1572 
1573 /*
1574  * pci_scan_bridge() - Scan buses behind a bridge
1575  * @bus: Parent bus the bridge is on
1576  * @dev: Bridge itself
1577  * @max: Starting subordinate number of buses behind this bridge
1578  * @pass: Either %0 (scan already configured bridges) or %1 (scan bridges
1579  *        that need to be reconfigured.
1580  *
1581  * If it's a bridge, configure it and scan the bus behind it.
1582  * For CardBus bridges, we don't scan behind as the devices will
1583  * be handled by the bridge driver itself.
1584  *
1585  * We need to process bridges in two passes -- first we scan those
1586  * already configured by the BIOS and after we are done with all of
1587  * them, we proceed to assigning numbers to the remaining buses in
1588  * order to avoid overlaps between old and new bus numbers.
1589  *
1590  * Return: New subordinate number covering all buses behind this bridge.
1591  */
1592 int pci_scan_bridge(struct pci_bus *bus, struct pci_dev *dev, int max, int pass)
1593 {
1594 	return pci_scan_bridge_extend(bus, dev, max, 0, pass);
1595 }
1596 EXPORT_SYMBOL(pci_scan_bridge);
1597 
1598 /*
1599  * Read interrupt line and base address registers.
1600  * The architecture-dependent code can tweak these, of course.
1601  */
1602 static void pci_read_irq(struct pci_dev *dev)
1603 {
1604 	unsigned char irq;
1605 
1606 	/* VFs are not allowed to use INTx, so skip the config reads */
1607 	if (dev->is_virtfn) {
1608 		dev->pin = 0;
1609 		dev->irq = 0;
1610 		return;
1611 	}
1612 
1613 	pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &irq);
1614 	dev->pin = irq;
1615 	if (irq)
1616 		pci_read_config_byte(dev, PCI_INTERRUPT_LINE, &irq);
1617 	dev->irq = irq;
1618 }
1619 
1620 void set_pcie_port_type(struct pci_dev *pdev)
1621 {
1622 	int pos;
1623 	u16 reg16;
1624 	u32 reg32;
1625 	int type;
1626 	struct pci_dev *parent;
1627 
1628 	pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
1629 	if (!pos)
1630 		return;
1631 
1632 	pdev->pcie_cap = pos;
1633 	pci_read_config_word(pdev, pos + PCI_EXP_FLAGS, &reg16);
1634 	pdev->pcie_flags_reg = reg16;
1635 
1636 	type = pci_pcie_type(pdev);
1637 	if (type == PCI_EXP_TYPE_ROOT_PORT)
1638 		pci_enable_rrs_sv(pdev);
1639 
1640 	pci_read_config_dword(pdev, pos + PCI_EXP_DEVCAP, &pdev->devcap);
1641 	pdev->pcie_mpss = FIELD_GET(PCI_EXP_DEVCAP_PAYLOAD, pdev->devcap);
1642 
1643 	pcie_capability_read_dword(pdev, PCI_EXP_LNKCAP, &reg32);
1644 	if (reg32 & PCI_EXP_LNKCAP_DLLLARC)
1645 		pdev->link_active_reporting = 1;
1646 
1647 #ifdef CONFIG_PCIEASPM
1648 	if (reg32 & PCI_EXP_LNKCAP_ASPM_L0S)
1649 		pdev->aspm_l0s_support = 1;
1650 	if (reg32 & PCI_EXP_LNKCAP_ASPM_L1)
1651 		pdev->aspm_l1_support = 1;
1652 #endif
1653 
1654 	parent = pci_upstream_bridge(pdev);
1655 	if (!parent)
1656 		return;
1657 
1658 	/*
1659 	 * Some systems do not identify their upstream/downstream ports
1660 	 * correctly so detect impossible configurations here and correct
1661 	 * the port type accordingly.
1662 	 */
1663 	if (type == PCI_EXP_TYPE_DOWNSTREAM) {
1664 		/*
1665 		 * If pdev claims to be downstream port but the parent
1666 		 * device is also downstream port assume pdev is actually
1667 		 * upstream port.
1668 		 */
1669 		if (pcie_downstream_port(parent)) {
1670 			pci_info(pdev, "claims to be downstream port but is acting as upstream port, correcting type\n");
1671 			pdev->pcie_flags_reg &= ~PCI_EXP_FLAGS_TYPE;
1672 			pdev->pcie_flags_reg |= PCI_EXP_TYPE_UPSTREAM;
1673 		}
1674 	} else if (type == PCI_EXP_TYPE_UPSTREAM) {
1675 		/*
1676 		 * If pdev claims to be upstream port but the parent
1677 		 * device is also upstream port assume pdev is actually
1678 		 * downstream port.
1679 		 */
1680 		if (pci_pcie_type(parent) == PCI_EXP_TYPE_UPSTREAM) {
1681 			pci_info(pdev, "claims to be upstream port but is acting as downstream port, correcting type\n");
1682 			pdev->pcie_flags_reg &= ~PCI_EXP_FLAGS_TYPE;
1683 			pdev->pcie_flags_reg |= PCI_EXP_TYPE_DOWNSTREAM;
1684 		}
1685 	}
1686 }
1687 
1688 void set_pcie_hotplug_bridge(struct pci_dev *pdev)
1689 {
1690 	u32 reg32;
1691 
1692 	pcie_capability_read_dword(pdev, PCI_EXP_SLTCAP, &reg32);
1693 	if (reg32 & PCI_EXP_SLTCAP_HPC)
1694 		pdev->is_hotplug_bridge = pdev->is_pciehp = 1;
1695 }
1696 
1697 static void set_pcie_thunderbolt(struct pci_dev *dev)
1698 {
1699 	u16 vsec;
1700 
1701 	/* Is the device part of a Thunderbolt controller? */
1702 	vsec = pci_find_vsec_capability(dev, PCI_VENDOR_ID_INTEL, PCI_VSEC_ID_INTEL_TBT);
1703 	if (vsec)
1704 		dev->is_thunderbolt = 1;
1705 }
1706 
1707 static void set_pcie_cxl(struct pci_dev *dev)
1708 {
1709 	struct pci_dev *bridge;
1710 	u16 dvsec, cap;
1711 
1712 	if (!pci_is_pcie(dev))
1713 		return;
1714 
1715 	/*
1716 	 * Update parent's CXL state because alternate protocol training
1717 	 * may have changed
1718 	 */
1719 	bridge = pci_upstream_bridge(dev);
1720 	if (bridge)
1721 		set_pcie_cxl(bridge);
1722 
1723 	dvsec = pci_find_dvsec_capability(dev, PCI_VENDOR_ID_CXL,
1724 					  PCI_DVSEC_CXL_FLEXBUS_PORT);
1725 	if (!dvsec)
1726 		return;
1727 
1728 	pci_read_config_word(dev, dvsec + PCI_DVSEC_CXL_FLEXBUS_PORT_STATUS,
1729 			     &cap);
1730 
1731 	dev->is_cxl = FIELD_GET(PCI_DVSEC_CXL_FLEXBUS_PORT_STATUS_CACHE, cap) ||
1732 		FIELD_GET(PCI_DVSEC_CXL_FLEXBUS_PORT_STATUS_MEM, cap);
1733 
1734 }
1735 
1736 static void set_pcie_untrusted(struct pci_dev *dev)
1737 {
1738 	struct pci_dev *parent = pci_upstream_bridge(dev);
1739 
1740 	if (!parent)
1741 		return;
1742 	/*
1743 	 * If the upstream bridge is untrusted we treat this device as
1744 	 * untrusted as well.
1745 	 */
1746 	if (parent->untrusted) {
1747 		dev->untrusted = true;
1748 		return;
1749 	}
1750 
1751 	if (arch_pci_dev_is_removable(dev)) {
1752 		pci_dbg(dev, "marking as untrusted\n");
1753 		dev->untrusted = true;
1754 	}
1755 }
1756 
1757 static void pci_set_removable(struct pci_dev *dev)
1758 {
1759 	struct pci_dev *parent = pci_upstream_bridge(dev);
1760 
1761 	if (!parent)
1762 		return;
1763 	/*
1764 	 * We (only) consider everything tunneled below an external_facing
1765 	 * device to be removable by the user. We're mainly concerned with
1766 	 * consumer platforms with user accessible thunderbolt ports that are
1767 	 * vulnerable to DMA attacks, and we expect those ports to be marked by
1768 	 * the firmware as external_facing. Devices in traditional hotplug
1769 	 * slots can technically be removed, but the expectation is that unless
1770 	 * the port is marked with external_facing, such devices are less
1771 	 * accessible to user / may not be removed by end user, and thus not
1772 	 * exposed as "removable" to userspace.
1773 	 */
1774 	if (dev_is_removable(&parent->dev)) {
1775 		dev_set_removable(&dev->dev, DEVICE_REMOVABLE);
1776 		return;
1777 	}
1778 
1779 	if (arch_pci_dev_is_removable(dev)) {
1780 		pci_dbg(dev, "marking as removable\n");
1781 		dev_set_removable(&dev->dev, DEVICE_REMOVABLE);
1782 	}
1783 }
1784 
1785 /**
1786  * pci_ext_cfg_is_aliased - Is ext config space just an alias of std config?
1787  * @dev: PCI device
1788  *
1789  * PCI Express to PCI/PCI-X Bridge Specification, rev 1.0, 4.1.4 says that
1790  * when forwarding a type1 configuration request the bridge must check that
1791  * the extended register address field is zero.  The bridge is not permitted
1792  * to forward the transactions and must handle it as an Unsupported Request.
1793  * Some bridges do not follow this rule and simply drop the extended register
1794  * bits, resulting in the standard config space being aliased, every 256
1795  * bytes across the entire configuration space.  Test for this condition by
1796  * comparing the first dword of each potential alias to the vendor/device ID.
1797  * Known offenders:
1798  *   ASM1083/1085 PCIe-to-PCI Reversible Bridge (1b21:1080, rev 01 & 03)
1799  *   AMD/ATI SBx00 PCI to PCI Bridge (1002:4384, rev 40)
1800  */
1801 static bool pci_ext_cfg_is_aliased(struct pci_dev *dev)
1802 {
1803 #ifdef CONFIG_PCI_QUIRKS
1804 	int pos, ret;
1805 	u32 header, tmp;
1806 
1807 	pci_read_config_dword(dev, PCI_VENDOR_ID, &header);
1808 
1809 	for (pos = PCI_CFG_SPACE_SIZE;
1810 	     pos < PCI_CFG_SPACE_EXP_SIZE; pos += PCI_CFG_SPACE_SIZE) {
1811 		ret = pci_read_config_dword(dev, pos, &tmp);
1812 		if ((ret != PCIBIOS_SUCCESSFUL) || (header != tmp))
1813 			return false;
1814 	}
1815 
1816 	return true;
1817 #else
1818 	return false;
1819 #endif
1820 }
1821 
1822 /**
1823  * pci_cfg_space_size_ext - Get the configuration space size of the PCI device
1824  * @dev: PCI device
1825  *
1826  * Regular PCI devices have 256 bytes, but PCI-X 2 and PCI Express devices
1827  * have 4096 bytes.  Even if the device is capable, that doesn't mean we can
1828  * access it.  Maybe we don't have a way to generate extended config space
1829  * accesses, or the device is behind a reverse Express bridge.  So we try
1830  * reading the dword at 0x100 which must either be 0 or a valid extended
1831  * capability header.
1832  */
1833 static int pci_cfg_space_size_ext(struct pci_dev *dev)
1834 {
1835 	u32 status;
1836 	int pos = PCI_CFG_SPACE_SIZE;
1837 
1838 	if (pci_read_config_dword(dev, pos, &status) != PCIBIOS_SUCCESSFUL)
1839 		return PCI_CFG_SPACE_SIZE;
1840 	if (PCI_POSSIBLE_ERROR(status) || pci_ext_cfg_is_aliased(dev))
1841 		return PCI_CFG_SPACE_SIZE;
1842 
1843 	return PCI_CFG_SPACE_EXP_SIZE;
1844 }
1845 
1846 int pci_cfg_space_size(struct pci_dev *dev)
1847 {
1848 	int pos;
1849 	u32 status;
1850 	u16 class;
1851 
1852 #ifdef CONFIG_PCI_IOV
1853 	/*
1854 	 * Per the SR-IOV specification (rev 1.1, sec 3.5), VFs are required to
1855 	 * implement a PCIe capability and therefore must implement extended
1856 	 * config space.  We can skip the NO_EXTCFG test below and the
1857 	 * reachability/aliasing test in pci_cfg_space_size_ext() by virtue of
1858 	 * the fact that the SR-IOV capability on the PF resides in extended
1859 	 * config space and must be accessible and non-aliased to have enabled
1860 	 * support for this VF.  This is a micro performance optimization for
1861 	 * systems supporting many VFs.
1862 	 */
1863 	if (dev->is_virtfn)
1864 		return PCI_CFG_SPACE_EXP_SIZE;
1865 #endif
1866 
1867 	if (dev->bus->bus_flags & PCI_BUS_FLAGS_NO_EXTCFG)
1868 		return PCI_CFG_SPACE_SIZE;
1869 
1870 	class = dev->class >> 8;
1871 	if (class == PCI_CLASS_BRIDGE_HOST)
1872 		return pci_cfg_space_size_ext(dev);
1873 
1874 	if (pci_is_pcie(dev))
1875 		return pci_cfg_space_size_ext(dev);
1876 
1877 	pos = pci_find_capability(dev, PCI_CAP_ID_PCIX);
1878 	if (!pos)
1879 		return PCI_CFG_SPACE_SIZE;
1880 
1881 	pci_read_config_dword(dev, pos + PCI_X_STATUS, &status);
1882 	if (status & (PCI_X_STATUS_266MHZ | PCI_X_STATUS_533MHZ))
1883 		return pci_cfg_space_size_ext(dev);
1884 
1885 	return PCI_CFG_SPACE_SIZE;
1886 }
1887 
1888 static u32 pci_class(struct pci_dev *dev)
1889 {
1890 	u32 class;
1891 
1892 #ifdef CONFIG_PCI_IOV
1893 	if (dev->is_virtfn)
1894 		return dev->physfn->sriov->class;
1895 #endif
1896 	pci_read_config_dword(dev, PCI_CLASS_REVISION, &class);
1897 	return class;
1898 }
1899 
1900 static void pci_subsystem_ids(struct pci_dev *dev, u16 *vendor, u16 *device)
1901 {
1902 #ifdef CONFIG_PCI_IOV
1903 	if (dev->is_virtfn) {
1904 		*vendor = dev->physfn->sriov->subsystem_vendor;
1905 		*device = dev->physfn->sriov->subsystem_device;
1906 		return;
1907 	}
1908 #endif
1909 	pci_read_config_word(dev, PCI_SUBSYSTEM_VENDOR_ID, vendor);
1910 	pci_read_config_word(dev, PCI_SUBSYSTEM_ID, device);
1911 }
1912 
1913 static u8 pci_hdr_type(struct pci_dev *dev)
1914 {
1915 	u8 hdr_type;
1916 
1917 #ifdef CONFIG_PCI_IOV
1918 	if (dev->is_virtfn)
1919 		return dev->physfn->sriov->hdr_type;
1920 #endif
1921 	pci_read_config_byte(dev, PCI_HEADER_TYPE, &hdr_type);
1922 	return hdr_type;
1923 }
1924 
1925 #define LEGACY_IO_RESOURCE	(IORESOURCE_IO | IORESOURCE_PCI_FIXED)
1926 
1927 /**
1928  * pci_intx_mask_broken - Test PCI_COMMAND_INTX_DISABLE writability
1929  * @dev: PCI device
1930  *
1931  * Test whether PCI_COMMAND_INTX_DISABLE is writable for @dev.  Check this
1932  * at enumeration-time to avoid modifying PCI_COMMAND at run-time.
1933  */
1934 static int pci_intx_mask_broken(struct pci_dev *dev)
1935 {
1936 	u16 orig, toggle, new;
1937 
1938 	pci_read_config_word(dev, PCI_COMMAND, &orig);
1939 	toggle = orig ^ PCI_COMMAND_INTX_DISABLE;
1940 	pci_write_config_word(dev, PCI_COMMAND, toggle);
1941 	pci_read_config_word(dev, PCI_COMMAND, &new);
1942 
1943 	pci_write_config_word(dev, PCI_COMMAND, orig);
1944 
1945 	/*
1946 	 * PCI_COMMAND_INTX_DISABLE was reserved and read-only prior to PCI
1947 	 * r2.3, so strictly speaking, a device is not *broken* if it's not
1948 	 * writable.  But we'll live with the misnomer for now.
1949 	 */
1950 	if (new != toggle)
1951 		return 1;
1952 	return 0;
1953 }
1954 
1955 static void early_dump_pci_device(struct pci_dev *pdev)
1956 {
1957 	u32 value[PCI_CFG_SPACE_SIZE / sizeof(u32)];
1958 	int i;
1959 
1960 	pci_info(pdev, "config space:\n");
1961 
1962 	for (i = 0; i < ARRAY_SIZE(value); i++)
1963 		pci_read_config_dword(pdev, i * sizeof(u32), &value[i]);
1964 
1965 	print_hex_dump(KERN_INFO, "", DUMP_PREFIX_OFFSET, 16, 1,
1966 		       value, ARRAY_SIZE(value) * sizeof(u32), false);
1967 }
1968 
1969 static const char *pci_type_str(struct pci_dev *dev)
1970 {
1971 	static const char * const str[] = {
1972 		"PCIe Endpoint",
1973 		"PCIe Legacy Endpoint",
1974 		"PCIe unknown",
1975 		"PCIe unknown",
1976 		"PCIe Root Port",
1977 		"PCIe Switch Upstream Port",
1978 		"PCIe Switch Downstream Port",
1979 		"PCIe to PCI/PCI-X bridge",
1980 		"PCI/PCI-X to PCIe bridge",
1981 		"PCIe Root Complex Integrated Endpoint",
1982 		"PCIe Root Complex Event Collector",
1983 	};
1984 	int type;
1985 
1986 	if (pci_is_pcie(dev)) {
1987 		type = pci_pcie_type(dev);
1988 		if (type < ARRAY_SIZE(str))
1989 			return str[type];
1990 
1991 		return "PCIe unknown";
1992 	}
1993 
1994 	switch (dev->hdr_type) {
1995 	case PCI_HEADER_TYPE_NORMAL:
1996 		return "conventional PCI endpoint";
1997 	case PCI_HEADER_TYPE_BRIDGE:
1998 		return "conventional PCI bridge";
1999 	case PCI_HEADER_TYPE_CARDBUS:
2000 		return "CardBus bridge";
2001 	default:
2002 		return "conventional PCI";
2003 	}
2004 }
2005 
2006 /**
2007  * pci_setup_device - Fill in class and map information of a device
2008  * @dev: the device structure to fill
2009  *
2010  * Initialize the device structure with information about the device's
2011  * vendor,class,memory and IO-space addresses, IRQ lines etc.
2012  * Called at initialisation of the PCI subsystem and by CardBus services.
2013  * Returns 0 on success and negative if unknown type of device (not normal,
2014  * bridge or CardBus).
2015  */
2016 int pci_setup_device(struct pci_dev *dev)
2017 {
2018 	u32 class;
2019 	u16 cmd;
2020 	u8 hdr_type;
2021 	int err, pos = 0;
2022 	struct pci_bus_region region;
2023 	struct resource *res;
2024 
2025 	hdr_type = pci_hdr_type(dev);
2026 
2027 	dev->sysdata = dev->bus->sysdata;
2028 	dev->dev.parent = dev->bus->bridge;
2029 	dev->dev.bus = &pci_bus_type;
2030 	dev->hdr_type = FIELD_GET(PCI_HEADER_TYPE_MASK, hdr_type);
2031 	dev->multifunction = FIELD_GET(PCI_HEADER_TYPE_MFD, hdr_type);
2032 	dev->error_state = pci_channel_io_normal;
2033 	set_pcie_port_type(dev);
2034 
2035 	err = pci_set_of_node(dev);
2036 	if (err)
2037 		return err;
2038 	pci_set_acpi_fwnode(dev);
2039 
2040 	pci_dev_assign_slot(dev);
2041 
2042 	/*
2043 	 * Assume 32-bit PCI; let 64-bit PCI cards (which are far rarer)
2044 	 * set this higher, assuming the system even supports it.
2045 	 */
2046 	dev->dma_mask = 0xffffffff;
2047 
2048 	/*
2049 	 * Assume 64-bit addresses for MSI initially. Will be changed to 32-bit
2050 	 * if MSI (rather than MSI-X) capability does not have
2051 	 * PCI_MSI_FLAGS_64BIT. Can also be overridden by driver.
2052 	 */
2053 	dev->msi_addr_mask = DMA_BIT_MASK(64);
2054 
2055 	dev_set_name(&dev->dev, "%04x:%02x:%02x.%d", pci_domain_nr(dev->bus),
2056 		     dev->bus->number, PCI_SLOT(dev->devfn),
2057 		     PCI_FUNC(dev->devfn));
2058 
2059 	class = pci_class(dev);
2060 
2061 	dev->revision = class & 0xff;
2062 	dev->class = class >> 8;		    /* upper 3 bytes */
2063 
2064 	if (pci_early_dump)
2065 		early_dump_pci_device(dev);
2066 
2067 	/* Need to have dev->class ready */
2068 	dev->cfg_size = pci_cfg_space_size(dev);
2069 
2070 	/* Need to have dev->cfg_size ready */
2071 	set_pcie_thunderbolt(dev);
2072 
2073 	set_pcie_cxl(dev);
2074 
2075 	set_pcie_untrusted(dev);
2076 
2077 	if (pci_is_pcie(dev))
2078 		dev->supported_speeds = pcie_get_supported_speeds(dev);
2079 
2080 	/* "Unknown power state" */
2081 	dev->current_state = PCI_UNKNOWN;
2082 
2083 	/* Early fixups, before probing the BARs */
2084 	pci_fixup_device(pci_fixup_early, dev);
2085 
2086 	pci_set_removable(dev);
2087 
2088 	pci_info(dev, "[%04x:%04x] type %02x class %#08x %s\n",
2089 		 dev->vendor, dev->device, dev->hdr_type, dev->class,
2090 		 pci_type_str(dev));
2091 
2092 	/* Device class may be changed after fixup */
2093 	class = dev->class >> 8;
2094 
2095 	if (dev->non_compliant_bars && !dev->mmio_always_on) {
2096 		pci_read_config_word(dev, PCI_COMMAND, &cmd);
2097 		if (cmd & (PCI_COMMAND_IO | PCI_COMMAND_MEMORY)) {
2098 			pci_info(dev, "device has non-compliant BARs; disabling IO/MEM decoding\n");
2099 			cmd &= ~PCI_COMMAND_IO;
2100 			cmd &= ~PCI_COMMAND_MEMORY;
2101 			pci_write_config_word(dev, PCI_COMMAND, cmd);
2102 		}
2103 	}
2104 
2105 	dev->broken_intx_masking = pci_intx_mask_broken(dev);
2106 
2107 	switch (dev->hdr_type) {		    /* header type */
2108 	case PCI_HEADER_TYPE_NORMAL:		    /* standard header */
2109 		if (class == PCI_CLASS_BRIDGE_PCI)
2110 			goto bad;
2111 		pci_read_irq(dev);
2112 		pci_read_bases(dev, PCI_STD_NUM_BARS, PCI_ROM_ADDRESS);
2113 
2114 		pci_subsystem_ids(dev, &dev->subsystem_vendor, &dev->subsystem_device);
2115 
2116 		/*
2117 		 * Do the ugly legacy mode stuff here rather than broken chip
2118 		 * quirk code. Legacy mode ATA controllers have fixed
2119 		 * addresses. These are not always echoed in BAR0-3, and
2120 		 * BAR0-3 in a few cases contain junk!
2121 		 */
2122 		if (class == PCI_CLASS_STORAGE_IDE) {
2123 			u8 progif;
2124 			pci_read_config_byte(dev, PCI_CLASS_PROG, &progif);
2125 			if ((progif & 1) == 0) {
2126 				region.start = 0x1F0;
2127 				region.end = 0x1F7;
2128 				res = &dev->resource[0];
2129 				res->flags = LEGACY_IO_RESOURCE;
2130 				pcibios_bus_to_resource(dev->bus, res, &region);
2131 				pci_info(dev, "BAR 0 %pR: legacy IDE quirk\n",
2132 					 res);
2133 				region.start = 0x3F6;
2134 				region.end = 0x3F6;
2135 				res = &dev->resource[1];
2136 				res->flags = LEGACY_IO_RESOURCE;
2137 				pcibios_bus_to_resource(dev->bus, res, &region);
2138 				pci_info(dev, "BAR 1 %pR: legacy IDE quirk\n",
2139 					 res);
2140 			}
2141 			if ((progif & 4) == 0) {
2142 				region.start = 0x170;
2143 				region.end = 0x177;
2144 				res = &dev->resource[2];
2145 				res->flags = LEGACY_IO_RESOURCE;
2146 				pcibios_bus_to_resource(dev->bus, res, &region);
2147 				pci_info(dev, "BAR 2 %pR: legacy IDE quirk\n",
2148 					 res);
2149 				region.start = 0x376;
2150 				region.end = 0x376;
2151 				res = &dev->resource[3];
2152 				res->flags = LEGACY_IO_RESOURCE;
2153 				pcibios_bus_to_resource(dev->bus, res, &region);
2154 				pci_info(dev, "BAR 3 %pR: legacy IDE quirk\n",
2155 					 res);
2156 			}
2157 		}
2158 		break;
2159 
2160 	case PCI_HEADER_TYPE_BRIDGE:		    /* bridge header */
2161 		/*
2162 		 * The PCI-to-PCI bridge spec requires that subtractive
2163 		 * decoding (i.e. transparent) bridge must have programming
2164 		 * interface code of 0x01.
2165 		 */
2166 		pci_read_irq(dev);
2167 		dev->transparent = ((dev->class & 0xff) == 1);
2168 		pci_read_bases(dev, 2, PCI_ROM_ADDRESS1);
2169 		pci_read_bridge_windows(dev);
2170 		set_pcie_hotplug_bridge(dev);
2171 		pos = pci_find_capability(dev, PCI_CAP_ID_SSVID);
2172 		if (pos) {
2173 			pci_read_config_word(dev, pos + PCI_SSVID_VENDOR_ID, &dev->subsystem_vendor);
2174 			pci_read_config_word(dev, pos + PCI_SSVID_DEVICE_ID, &dev->subsystem_device);
2175 		}
2176 		break;
2177 
2178 	case PCI_HEADER_TYPE_CARDBUS:		    /* CardBus bridge header */
2179 		if (class != PCI_CLASS_BRIDGE_CARDBUS)
2180 			goto bad;
2181 		pci_read_irq(dev);
2182 		pci_read_bases(dev, 1, 0);
2183 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_VENDOR_ID, &dev->subsystem_vendor);
2184 		pci_read_config_word(dev, PCI_CB_SUBSYSTEM_ID, &dev->subsystem_device);
2185 		break;
2186 
2187 	default:				    /* unknown header */
2188 		pci_err(dev, "unknown header type %02x, ignoring device\n",
2189 			dev->hdr_type);
2190 		pci_release_of_node(dev);
2191 		return -EIO;
2192 
2193 	bad:
2194 		pci_err(dev, "ignoring class %#08x (doesn't match header type %02x)\n",
2195 			dev->class, dev->hdr_type);
2196 		dev->class = PCI_CLASS_NOT_DEFINED << 8;
2197 	}
2198 
2199 	/* We found a fine healthy device, go go go... */
2200 	return 0;
2201 }
2202 
2203 static void pci_configure_mps(struct pci_dev *dev)
2204 {
2205 	struct pci_dev *bridge = pci_upstream_bridge(dev);
2206 	int mps, mpss, p_mps, rc;
2207 
2208 	if (!pci_is_pcie(dev))
2209 		return;
2210 
2211 	/* MPS and MRRS fields are of type 'RsvdP' for VFs, short-circuit out */
2212 	if (dev->is_virtfn)
2213 		return;
2214 
2215 	/*
2216 	 * For Root Complex Integrated Endpoints, program the maximum
2217 	 * supported value unless limited by the PCIE_BUS_PEER2PEER case.
2218 	 */
2219 	if (pci_pcie_type(dev) == PCI_EXP_TYPE_RC_END) {
2220 		if (pcie_bus_config == PCIE_BUS_PEER2PEER)
2221 			mps = 128;
2222 		else
2223 			mps = 128 << dev->pcie_mpss;
2224 		rc = pcie_set_mps(dev, mps);
2225 		if (rc) {
2226 			pci_warn(dev, "can't set Max Payload Size to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
2227 				 mps);
2228 		}
2229 		return;
2230 	}
2231 
2232 	if (!bridge || !pci_is_pcie(bridge))
2233 		return;
2234 
2235 	mps = pcie_get_mps(dev);
2236 	p_mps = pcie_get_mps(bridge);
2237 
2238 	if (mps == p_mps)
2239 		return;
2240 
2241 	if (pcie_bus_config == PCIE_BUS_TUNE_OFF) {
2242 		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",
2243 			 mps, pci_name(bridge), p_mps);
2244 		return;
2245 	}
2246 
2247 	/*
2248 	 * Fancier MPS configuration is done later by
2249 	 * pcie_bus_configure_settings()
2250 	 */
2251 	if (pcie_bus_config != PCIE_BUS_DEFAULT)
2252 		return;
2253 
2254 	mpss = 128 << dev->pcie_mpss;
2255 	if (mpss < p_mps && pci_pcie_type(bridge) == PCI_EXP_TYPE_ROOT_PORT) {
2256 		pcie_set_mps(bridge, mpss);
2257 		pci_info(dev, "Upstream bridge's Max Payload Size set to %d (was %d, max %d)\n",
2258 			 mpss, p_mps, 128 << bridge->pcie_mpss);
2259 		p_mps = pcie_get_mps(bridge);
2260 	}
2261 
2262 	rc = pcie_set_mps(dev, p_mps);
2263 	if (rc) {
2264 		pci_warn(dev, "can't set Max Payload Size to %d; if necessary, use \"pci=pcie_bus_safe\" and report a bug\n",
2265 			 p_mps);
2266 		return;
2267 	}
2268 
2269 	pci_info(dev, "Max Payload Size set to %d (was %d, max %d)\n",
2270 		 p_mps, mps, mpss);
2271 }
2272 
2273 int pci_configure_extended_tags(struct pci_dev *dev, void *ign)
2274 {
2275 	struct pci_host_bridge *host;
2276 	u32 cap;
2277 	u16 ctl;
2278 	int ret;
2279 
2280 	/* PCI_EXP_DEVCTL_EXT_TAG is RsvdP in VFs */
2281 	if (!pci_is_pcie(dev) || dev->is_virtfn)
2282 		return 0;
2283 
2284 	ret = pcie_capability_read_dword(dev, PCI_EXP_DEVCAP, &cap);
2285 	if (ret)
2286 		return 0;
2287 
2288 	if (!(cap & PCI_EXP_DEVCAP_EXT_TAG))
2289 		return 0;
2290 
2291 	ret = pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &ctl);
2292 	if (ret)
2293 		return 0;
2294 
2295 	host = pci_find_host_bridge(dev->bus);
2296 	if (!host)
2297 		return 0;
2298 
2299 	/*
2300 	 * If some device in the hierarchy doesn't handle Extended Tags
2301 	 * correctly, make sure they're disabled.
2302 	 */
2303 	if (host->no_ext_tags) {
2304 		if (ctl & PCI_EXP_DEVCTL_EXT_TAG) {
2305 			pci_info(dev, "disabling Extended Tags\n");
2306 			pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
2307 						   PCI_EXP_DEVCTL_EXT_TAG);
2308 		}
2309 		return 0;
2310 	}
2311 
2312 	if (!(ctl & PCI_EXP_DEVCTL_EXT_TAG)) {
2313 		pci_info(dev, "enabling Extended Tags\n");
2314 		pcie_capability_set_word(dev, PCI_EXP_DEVCTL,
2315 					 PCI_EXP_DEVCTL_EXT_TAG);
2316 	}
2317 	return 0;
2318 }
2319 
2320 static void pci_dev3_init(struct pci_dev *pdev)
2321 {
2322 	u16 cap = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_DEV3);
2323 	u32 val = 0;
2324 
2325 	if (!cap)
2326 		return;
2327 	pci_read_config_dword(pdev, cap + PCI_DEV3_STA, &val);
2328 	pdev->fm_enabled = !!(val & PCI_DEV3_STA_SEGMENT);
2329 }
2330 
2331 /**
2332  * pcie_relaxed_ordering_enabled - Probe for PCIe relaxed ordering enable
2333  * @dev: PCI device to query
2334  *
2335  * Returns true if the device has enabled relaxed ordering attribute.
2336  */
2337 bool pcie_relaxed_ordering_enabled(struct pci_dev *dev)
2338 {
2339 	u16 v;
2340 
2341 	pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &v);
2342 
2343 	return !!(v & PCI_EXP_DEVCTL_RELAX_EN);
2344 }
2345 EXPORT_SYMBOL(pcie_relaxed_ordering_enabled);
2346 
2347 static void pci_configure_relaxed_ordering(struct pci_dev *dev)
2348 {
2349 	struct pci_dev *root;
2350 
2351 	/* PCI_EXP_DEVCTL_RELAX_EN is RsvdP in VFs */
2352 	if (dev->is_virtfn)
2353 		return;
2354 
2355 	if (!pcie_relaxed_ordering_enabled(dev))
2356 		return;
2357 
2358 	/*
2359 	 * For now, we only deal with Relaxed Ordering issues with Root
2360 	 * Ports. Peer-to-Peer DMA is another can of worms.
2361 	 */
2362 	root = pcie_find_root_port(dev);
2363 	if (!root)
2364 		return;
2365 
2366 	if (root->dev_flags & PCI_DEV_FLAGS_NO_RELAXED_ORDERING) {
2367 		pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
2368 					   PCI_EXP_DEVCTL_RELAX_EN);
2369 		pci_info(dev, "Relaxed Ordering disabled because the Root Port didn't support it\n");
2370 	}
2371 }
2372 
2373 static void pci_configure_eetlp_prefix(struct pci_dev *dev)
2374 {
2375 	struct pci_dev *bridge;
2376 	unsigned int eetlp_max;
2377 	int pcie_type;
2378 	u32 cap;
2379 
2380 	if (!pci_is_pcie(dev))
2381 		return;
2382 
2383 	pcie_capability_read_dword(dev, PCI_EXP_DEVCAP2, &cap);
2384 	if (!(cap & PCI_EXP_DEVCAP2_EE_PREFIX))
2385 		return;
2386 
2387 	pcie_type = pci_pcie_type(dev);
2388 
2389 	eetlp_max = FIELD_GET(PCI_EXP_DEVCAP2_EE_PREFIX_MAX, cap);
2390 	/* 00b means 4 */
2391 	eetlp_max = eetlp_max ?: 4;
2392 
2393 	if (pcie_type == PCI_EXP_TYPE_ROOT_PORT ||
2394 	    pcie_type == PCI_EXP_TYPE_RC_END)
2395 		dev->eetlp_prefix_max = eetlp_max;
2396 	else {
2397 		bridge = pci_upstream_bridge(dev);
2398 		if (bridge && bridge->eetlp_prefix_max)
2399 			dev->eetlp_prefix_max = eetlp_max;
2400 	}
2401 }
2402 
2403 static void pci_configure_serr(struct pci_dev *dev)
2404 {
2405 	u16 control;
2406 
2407 	if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
2408 
2409 		/*
2410 		 * A bridge will not forward ERR_ messages coming from an
2411 		 * endpoint unless SERR# forwarding is enabled.
2412 		 */
2413 		pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &control);
2414 		if (!(control & PCI_BRIDGE_CTL_SERR)) {
2415 			control |= PCI_BRIDGE_CTL_SERR;
2416 			pci_write_config_word(dev, PCI_BRIDGE_CONTROL, control);
2417 		}
2418 	}
2419 }
2420 
2421 static void pci_configure_rcb(struct pci_dev *dev)
2422 {
2423 	struct pci_dev *rp;
2424 	u16 rp_lnkctl;
2425 
2426 	/*
2427 	 * Per PCIe r7.0, sec 7.5.3.7, RCB is only meaningful in Root Ports
2428 	 * (where it is read-only), Endpoints, and Bridges.  It may only be
2429 	 * set for Endpoints and Bridges if it is set in the Root Port. For
2430 	 * Endpoints, it is 'RsvdP' for Virtual Functions.
2431 	 */
2432 	if (!pci_is_pcie(dev) ||
2433 	    pci_pcie_type(dev) == PCI_EXP_TYPE_ROOT_PORT ||
2434 	    pci_pcie_type(dev) == PCI_EXP_TYPE_UPSTREAM ||
2435 	    pci_pcie_type(dev) == PCI_EXP_TYPE_DOWNSTREAM ||
2436 	    pci_pcie_type(dev) == PCI_EXP_TYPE_RC_EC ||
2437 	    dev->is_virtfn)
2438 		return;
2439 
2440 	/* Root Port often not visible to virtualized guests */
2441 	rp = pcie_find_root_port(dev);
2442 	if (!rp)
2443 		return;
2444 
2445 	pcie_capability_read_word(rp, PCI_EXP_LNKCTL, &rp_lnkctl);
2446 	pcie_capability_clear_and_set_word(dev, PCI_EXP_LNKCTL,
2447 					   PCI_EXP_LNKCTL_RCB,
2448 					   (rp_lnkctl & PCI_EXP_LNKCTL_RCB) ?
2449 					   PCI_EXP_LNKCTL_RCB : 0);
2450 }
2451 
2452 static void pci_configure_device(struct pci_dev *dev)
2453 {
2454 	pci_configure_mps(dev);
2455 	pci_configure_extended_tags(dev, NULL);
2456 	pci_configure_relaxed_ordering(dev);
2457 	pci_configure_ltr(dev);
2458 	pci_configure_aspm_l1ss(dev);
2459 	pci_configure_eetlp_prefix(dev);
2460 	pci_configure_serr(dev);
2461 	pci_configure_rcb(dev);
2462 
2463 	pci_acpi_program_hp_params(dev);
2464 }
2465 
2466 static void pci_release_capabilities(struct pci_dev *dev)
2467 {
2468 	pci_aer_exit(dev);
2469 	pci_rcec_exit(dev);
2470 	pci_iov_release(dev);
2471 	pci_free_cap_save_buffers(dev);
2472 }
2473 
2474 /**
2475  * pci_release_dev - Free a PCI device structure when all users of it are
2476  *		     finished
2477  * @dev: device that's been disconnected
2478  *
2479  * Will be called only by the device core when all users of this PCI device are
2480  * done.
2481  */
2482 static void pci_release_dev(struct device *dev)
2483 {
2484 	struct pci_dev *pci_dev;
2485 
2486 	pci_dev = to_pci_dev(dev);
2487 	pci_release_capabilities(pci_dev);
2488 	pci_release_of_node(pci_dev);
2489 	pcibios_release_device(pci_dev);
2490 	pci_bus_put(pci_dev->bus);
2491 	bitmap_free(pci_dev->dma_alias_mask);
2492 	dev_dbg(dev, "device released\n");
2493 	kfree(pci_dev);
2494 }
2495 
2496 static const struct device_type pci_dev_type = {
2497 	.groups = pci_dev_attr_groups,
2498 };
2499 
2500 struct pci_dev *pci_alloc_dev(struct pci_bus *bus)
2501 {
2502 	struct pci_dev *dev;
2503 
2504 	dev = kzalloc_obj(struct pci_dev);
2505 	if (!dev)
2506 		return NULL;
2507 
2508 	INIT_LIST_HEAD(&dev->bus_list);
2509 	dev->dev.type = &pci_dev_type;
2510 	dev->bus = pci_bus_get(bus);
2511 	dev->driver_exclusive_resource = (struct resource) {
2512 		.name = "PCI Exclusive",
2513 		.start = 0,
2514 		.end = -1,
2515 	};
2516 
2517 	spin_lock_init(&dev->pcie_cap_lock);
2518 #ifdef CONFIG_PCI_MSI
2519 	raw_spin_lock_init(&dev->msi_lock);
2520 #endif
2521 	return dev;
2522 }
2523 EXPORT_SYMBOL(pci_alloc_dev);
2524 
2525 static bool pci_bus_wait_rrs(struct pci_bus *bus, int devfn, u32 *l,
2526 			     int timeout)
2527 {
2528 	int delay = 1;
2529 
2530 	if (!pci_bus_rrs_vendor_id(*l))
2531 		return true;	/* not a Configuration RRS completion */
2532 
2533 	if (!timeout)
2534 		return false;	/* RRS, but caller doesn't want to wait */
2535 
2536 	/*
2537 	 * We got the reserved Vendor ID that indicates a completion with
2538 	 * Configuration Request Retry Status (RRS).  Retry until we get a
2539 	 * valid Vendor ID or we time out.
2540 	 */
2541 	while (pci_bus_rrs_vendor_id(*l)) {
2542 		if (delay > timeout) {
2543 			pr_warn("pci %04x:%02x:%02x.%d: not ready after %dms; giving up\n",
2544 				pci_domain_nr(bus), bus->number,
2545 				PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2546 
2547 			return false;
2548 		}
2549 		if (delay >= 1000)
2550 			pr_info("pci %04x:%02x:%02x.%d: not ready after %dms; waiting\n",
2551 				pci_domain_nr(bus), bus->number,
2552 				PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2553 
2554 		msleep(delay);
2555 		delay *= 2;
2556 
2557 		if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
2558 			return false;
2559 	}
2560 
2561 	if (delay >= 1000)
2562 		pr_info("pci %04x:%02x:%02x.%d: ready after %dms\n",
2563 			pci_domain_nr(bus), bus->number,
2564 			PCI_SLOT(devfn), PCI_FUNC(devfn), delay - 1);
2565 
2566 	return true;
2567 }
2568 
2569 bool pci_bus_generic_read_dev_vendor_id(struct pci_bus *bus, int devfn, u32 *l,
2570 					int timeout)
2571 {
2572 	if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, l))
2573 		return false;
2574 
2575 	/* Some broken boards return 0 or ~0 (PCI_ERROR_RESPONSE) if a slot is empty: */
2576 	if (PCI_POSSIBLE_ERROR(*l) || *l == 0x00000000 ||
2577 	    *l == 0x0000ffff || *l == 0xffff0000)
2578 		return false;
2579 
2580 	if (pci_bus_rrs_vendor_id(*l))
2581 		return pci_bus_wait_rrs(bus, devfn, l, timeout);
2582 
2583 	return true;
2584 }
2585 
2586 bool pci_bus_read_dev_vendor_id(struct pci_bus *bus, int devfn, u32 *l,
2587 				int timeout)
2588 {
2589 	return pci_bus_generic_read_dev_vendor_id(bus, devfn, l, timeout);
2590 }
2591 EXPORT_SYMBOL(pci_bus_read_dev_vendor_id);
2592 
2593 /*
2594  * Read the config data for a PCI device, sanity-check it,
2595  * and fill in the dev structure.
2596  */
2597 static struct pci_dev *pci_scan_device(struct pci_bus *bus, int devfn)
2598 {
2599 	struct pci_dev *dev;
2600 	u32 l;
2601 
2602 	if (!pci_bus_read_dev_vendor_id(bus, devfn, &l, 60*1000))
2603 		return NULL;
2604 
2605 	dev = pci_alloc_dev(bus);
2606 	if (!dev)
2607 		return NULL;
2608 
2609 	dev->devfn = devfn;
2610 	dev->vendor = l & 0xffff;
2611 	dev->device = (l >> 16) & 0xffff;
2612 
2613 	if (pci_setup_device(dev)) {
2614 		pci_bus_put(dev->bus);
2615 		kfree(dev);
2616 		return NULL;
2617 	}
2618 
2619 	return dev;
2620 }
2621 
2622 void pcie_report_downtraining(struct pci_dev *dev)
2623 {
2624 	if (!pci_is_pcie(dev))
2625 		return;
2626 
2627 	/* Look from the device up to avoid downstream ports with no devices */
2628 	if ((pci_pcie_type(dev) != PCI_EXP_TYPE_ENDPOINT) &&
2629 	    (pci_pcie_type(dev) != PCI_EXP_TYPE_LEG_END) &&
2630 	    (pci_pcie_type(dev) != PCI_EXP_TYPE_UPSTREAM))
2631 		return;
2632 
2633 	/* Multi-function PCIe devices share the same link/status */
2634 	if (PCI_FUNC(dev->devfn) != 0 || dev->is_virtfn)
2635 		return;
2636 
2637 	/* Print link status only if the device is constrained by the fabric */
2638 	__pcie_print_link_status(dev, false);
2639 }
2640 
2641 static void pci_imm_ready_init(struct pci_dev *dev)
2642 {
2643 	u16 status;
2644 
2645 	pci_read_config_word(dev, PCI_STATUS, &status);
2646 	if (status & PCI_STATUS_IMM_READY)
2647 		dev->imm_ready = 1;
2648 }
2649 
2650 static void pci_init_capabilities(struct pci_dev *dev)
2651 {
2652 	pci_ea_init(dev);		/* Enhanced Allocation */
2653 	pci_msi_init(dev);		/* Disable MSI */
2654 	pci_msix_init(dev);		/* Disable MSI-X */
2655 
2656 	/* Buffers for saving PCIe and PCI-X capabilities */
2657 	pci_allocate_cap_save_buffers(dev);
2658 
2659 	pci_imm_ready_init(dev);	/* Immediate Readiness */
2660 	pci_pm_init(dev);		/* Power Management */
2661 	pci_vpd_init(dev);		/* Vital Product Data */
2662 	pci_configure_ari(dev);		/* Alternative Routing-ID Forwarding */
2663 	pci_iov_init(dev);		/* Single Root I/O Virtualization */
2664 	pci_ats_init(dev);		/* Address Translation Services */
2665 	pci_pri_init(dev);		/* Page Request Interface */
2666 	pci_pasid_init(dev);		/* Process Address Space ID */
2667 	pci_acs_init(dev);		/* Access Control Services */
2668 	pci_ptm_init(dev);		/* Precision Time Measurement */
2669 	pci_aer_init(dev);		/* Advanced Error Reporting */
2670 	pci_dpc_init(dev);		/* Downstream Port Containment */
2671 	pci_rcec_init(dev);		/* Root Complex Event Collector */
2672 	pci_doe_init(dev);		/* Data Object Exchange */
2673 	pci_tph_init(dev);		/* TLP Processing Hints */
2674 	pci_rebar_init(dev);		/* Resizable BAR */
2675 	pci_dev3_init(dev);		/* Device 3 capabilities */
2676 	pci_ide_init(dev);		/* Link Integrity and Data Encryption */
2677 
2678 	pcie_report_downtraining(dev);
2679 	pci_init_reset_methods(dev);
2680 }
2681 
2682 /*
2683  * This is the equivalent of pci_host_bridge_msi_domain() that acts on
2684  * devices. Firmware interfaces that can select the MSI domain on a
2685  * per-device basis should be called from here.
2686  */
2687 static struct irq_domain *pci_dev_msi_domain(struct pci_dev *dev)
2688 {
2689 	struct irq_domain *d;
2690 
2691 	/*
2692 	 * If a domain has been set through the pcibios_device_add()
2693 	 * callback, then this is the one (platform code knows best).
2694 	 */
2695 	d = dev_get_msi_domain(&dev->dev);
2696 	if (d)
2697 		return d;
2698 
2699 	/*
2700 	 * Let's see if we have a firmware interface able to provide
2701 	 * the domain.
2702 	 */
2703 	d = pci_msi_get_device_domain(dev);
2704 	if (d)
2705 		return d;
2706 
2707 	return NULL;
2708 }
2709 
2710 static void pci_set_msi_domain(struct pci_dev *dev)
2711 {
2712 	struct irq_domain *d;
2713 
2714 	/*
2715 	 * If the platform or firmware interfaces cannot supply a
2716 	 * device-specific MSI domain, then inherit the default domain
2717 	 * from the host bridge itself.
2718 	 */
2719 	d = pci_dev_msi_domain(dev);
2720 	if (!d)
2721 		d = dev_get_msi_domain(&dev->bus->dev);
2722 
2723 	dev_set_msi_domain(&dev->dev, d);
2724 }
2725 
2726 void pci_device_add(struct pci_dev *dev, struct pci_bus *bus)
2727 {
2728 	int ret;
2729 
2730 	pci_configure_device(dev);
2731 
2732 	device_initialize(&dev->dev);
2733 	dev->dev.release = pci_release_dev;
2734 
2735 	set_dev_node(&dev->dev, pcibus_to_node(bus));
2736 	dev->dev.dma_mask = &dev->dma_mask;
2737 	dev->dev.dma_parms = &dev->dma_parms;
2738 	dev->dev.coherent_dma_mask = 0xffffffffull;
2739 
2740 	dma_set_max_seg_size(&dev->dev, 65536);
2741 	dma_set_seg_boundary(&dev->dev, 0xffffffff);
2742 
2743 	pcie_failed_link_retrain(dev);
2744 
2745 	/* Fix up broken headers */
2746 	pci_fixup_device(pci_fixup_header, dev);
2747 
2748 	pci_reassigndev_resource_alignment(dev);
2749 
2750 	pci_init_capabilities(dev);
2751 
2752 	/*
2753 	 * Add the device to our list of discovered devices
2754 	 * and the bus list for fixup functions, etc.
2755 	 */
2756 	down_write(&pci_bus_sem);
2757 	list_add_tail(&dev->bus_list, &bus->devices);
2758 	up_write(&pci_bus_sem);
2759 
2760 	ret = pcibios_device_add(dev);
2761 	WARN_ON(ret < 0);
2762 
2763 	/* Set up MSI IRQ domain */
2764 	pci_set_msi_domain(dev);
2765 
2766 	/* Notifier could use PCI capabilities */
2767 	ret = device_add(&dev->dev);
2768 	WARN_ON(ret < 0);
2769 
2770 	/* Establish pdev->tsm for newly added (e.g. new SR-IOV VFs) */
2771 	pci_tsm_init(dev);
2772 
2773 	pci_npem_create(dev);
2774 
2775 	pci_doe_sysfs_init(dev);
2776 }
2777 
2778 struct pci_dev *pci_scan_single_device(struct pci_bus *bus, int devfn)
2779 {
2780 	struct pci_dev *dev;
2781 
2782 	dev = pci_get_slot(bus, devfn);
2783 	if (dev) {
2784 		pci_dev_put(dev);
2785 		return dev;
2786 	}
2787 
2788 	dev = pci_scan_device(bus, devfn);
2789 	if (!dev)
2790 		return NULL;
2791 
2792 	pci_device_add(dev, bus);
2793 
2794 	return dev;
2795 }
2796 EXPORT_SYMBOL(pci_scan_single_device);
2797 
2798 static int next_ari_fn(struct pci_bus *bus, struct pci_dev *dev, int fn)
2799 {
2800 	int pos;
2801 	u16 cap = 0;
2802 	unsigned int next_fn;
2803 
2804 	if (!dev)
2805 		return -ENODEV;
2806 
2807 	pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ARI);
2808 	if (!pos)
2809 		return -ENODEV;
2810 
2811 	pci_read_config_word(dev, pos + PCI_ARI_CAP, &cap);
2812 	next_fn = PCI_ARI_CAP_NFN(cap);
2813 	if (next_fn <= fn)
2814 		return -ENODEV;	/* protect against malformed list */
2815 
2816 	return next_fn;
2817 }
2818 
2819 static int next_fn(struct pci_bus *bus, struct pci_dev *dev, int fn)
2820 {
2821 	if (pci_ari_enabled(bus))
2822 		return next_ari_fn(bus, dev, fn);
2823 
2824 	if (fn >= 7)
2825 		return -ENODEV;
2826 	/* only multifunction devices may have more functions */
2827 	if (dev && !dev->multifunction)
2828 		return -ENODEV;
2829 
2830 	return fn + 1;
2831 }
2832 
2833 static int only_one_child(struct pci_bus *bus)
2834 {
2835 	struct pci_dev *bridge = bus->self;
2836 
2837 	/*
2838 	 * Systems with unusual topologies set PCI_SCAN_ALL_PCIE_DEVS so
2839 	 * we scan for all possible devices, not just Device 0.
2840 	 */
2841 	if (pci_has_flag(PCI_SCAN_ALL_PCIE_DEVS))
2842 		return 0;
2843 
2844 	/*
2845 	 * A PCIe Downstream Port normally leads to a Link with only Device
2846 	 * 0 on it (PCIe spec r3.1, sec 7.3.1).  As an optimization, scan
2847 	 * only for Device 0 in that situation.
2848 	 */
2849 	if (bridge && pci_is_pcie(bridge) && pcie_downstream_port(bridge))
2850 		return 1;
2851 
2852 	return 0;
2853 }
2854 
2855 /**
2856  * pci_scan_slot - Scan a PCI slot on a bus for devices
2857  * @bus: PCI bus to scan
2858  * @devfn: slot number to scan (must have zero function)
2859  *
2860  * Scan a PCI slot on the specified PCI bus for devices, adding
2861  * discovered devices to the @bus->devices list.  New devices
2862  * will not have is_added set.
2863  *
2864  * Returns the number of new devices found.
2865  */
2866 int pci_scan_slot(struct pci_bus *bus, int devfn)
2867 {
2868 	struct pci_dev *dev;
2869 	int fn = 0, nr = 0;
2870 
2871 	if (only_one_child(bus) && (devfn > 0))
2872 		return 0; /* Already scanned the entire slot */
2873 
2874 	do {
2875 		dev = pci_scan_single_device(bus, devfn + fn);
2876 		if (dev) {
2877 			if (!pci_dev_is_added(dev))
2878 				nr++;
2879 			if (fn > 0)
2880 				dev->multifunction = 1;
2881 		} else if (fn == 0) {
2882 			/*
2883 			 * Function 0 is required unless we are running on
2884 			 * a hypervisor that passes through individual PCI
2885 			 * functions.
2886 			 */
2887 			if (!hypervisor_isolated_pci_functions())
2888 				break;
2889 		}
2890 		fn = next_fn(bus, dev, fn);
2891 	} while (fn >= 0);
2892 
2893 	/* Only one slot has PCIe device */
2894 	if (bus->self && nr)
2895 		pcie_aspm_init_link_state(bus->self);
2896 
2897 	return nr;
2898 }
2899 EXPORT_SYMBOL(pci_scan_slot);
2900 
2901 static int pcie_find_smpss(struct pci_dev *dev, void *data)
2902 {
2903 	u8 *smpss = data;
2904 
2905 	if (!pci_is_pcie(dev))
2906 		return 0;
2907 
2908 	/*
2909 	 * We don't have a way to change MPS settings on devices that have
2910 	 * drivers attached.  A hot-added device might support only the minimum
2911 	 * MPS setting (MPS=128).  Therefore, if the fabric contains a bridge
2912 	 * where devices may be hot-added, we limit the fabric MPS to 128 so
2913 	 * hot-added devices will work correctly.
2914 	 *
2915 	 * However, if we hot-add a device to a slot directly below a Root
2916 	 * Port, it's impossible for there to be other existing devices below
2917 	 * the port.  We don't limit the MPS in this case because we can
2918 	 * reconfigure MPS on both the Root Port and the hot-added device,
2919 	 * and there are no other devices involved.
2920 	 *
2921 	 * Note that this PCIE_BUS_SAFE path assumes no peer-to-peer DMA.
2922 	 */
2923 	if (dev->is_hotplug_bridge &&
2924 	    pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT)
2925 		*smpss = 0;
2926 
2927 	if (*smpss > dev->pcie_mpss)
2928 		*smpss = dev->pcie_mpss;
2929 
2930 	return 0;
2931 }
2932 
2933 static void pcie_write_mps(struct pci_dev *dev, int mps)
2934 {
2935 	int rc;
2936 
2937 	if (pcie_bus_config == PCIE_BUS_PERFORMANCE) {
2938 		mps = 128 << dev->pcie_mpss;
2939 
2940 		if (pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT &&
2941 		    dev->bus->self)
2942 
2943 			/*
2944 			 * For "Performance", the assumption is made that
2945 			 * downstream communication will never be larger than
2946 			 * the MRRS.  So, the MPS only needs to be configured
2947 			 * for the upstream communication.  This being the case,
2948 			 * walk from the top down and set the MPS of the child
2949 			 * to that of the parent bus.
2950 			 *
2951 			 * Configure the device MPS with the smaller of the
2952 			 * device MPSS or the bridge MPS (which is assumed to be
2953 			 * properly configured at this point to the largest
2954 			 * allowable MPS based on its parent bus).
2955 			 */
2956 			mps = min(mps, pcie_get_mps(dev->bus->self));
2957 	}
2958 
2959 	rc = pcie_set_mps(dev, mps);
2960 	if (rc)
2961 		pci_err(dev, "Failed attempting to set the MPS\n");
2962 }
2963 
2964 static void pcie_write_mrrs(struct pci_dev *dev)
2965 {
2966 	int rc, mrrs;
2967 
2968 	/*
2969 	 * In the "safe" case, do not configure the MRRS.  There appear to be
2970 	 * issues with setting MRRS to 0 on a number of devices.
2971 	 */
2972 	if (pcie_bus_config != PCIE_BUS_PERFORMANCE)
2973 		return;
2974 
2975 	/*
2976 	 * For max performance, the MRRS must be set to the largest supported
2977 	 * value.  However, it cannot be configured larger than the MPS the
2978 	 * device or the bus can support.  This should already be properly
2979 	 * configured by a prior call to pcie_write_mps().
2980 	 */
2981 	mrrs = pcie_get_mps(dev);
2982 
2983 	/*
2984 	 * MRRS is a R/W register.  Invalid values can be written, but a
2985 	 * subsequent read will verify if the value is acceptable or not.
2986 	 * If the MRRS value provided is not acceptable (e.g., too large),
2987 	 * shrink the value until it is acceptable to the HW.
2988 	 */
2989 	while (mrrs != pcie_get_readrq(dev) && mrrs >= 128) {
2990 		rc = pcie_set_readrq(dev, mrrs);
2991 		if (!rc)
2992 			break;
2993 
2994 		pci_warn(dev, "Failed attempting to set the MRRS\n");
2995 		mrrs /= 2;
2996 	}
2997 
2998 	if (mrrs < 128)
2999 		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");
3000 }
3001 
3002 static int pcie_bus_configure_set(struct pci_dev *dev, void *data)
3003 {
3004 	int mps, orig_mps;
3005 
3006 	if (!pci_is_pcie(dev))
3007 		return 0;
3008 
3009 	if (pcie_bus_config == PCIE_BUS_TUNE_OFF ||
3010 	    pcie_bus_config == PCIE_BUS_DEFAULT)
3011 		return 0;
3012 
3013 	mps = 128 << *(u8 *)data;
3014 	orig_mps = pcie_get_mps(dev);
3015 
3016 	pcie_write_mps(dev, mps);
3017 	pcie_write_mrrs(dev);
3018 
3019 	pci_info(dev, "Max Payload Size set to %4d/%4d (was %4d), Max Read Rq %4d\n",
3020 		 pcie_get_mps(dev), 128 << dev->pcie_mpss,
3021 		 orig_mps, pcie_get_readrq(dev));
3022 
3023 	return 0;
3024 }
3025 
3026 /*
3027  * pcie_bus_configure_settings() requires that pci_walk_bus work in a top-down,
3028  * parents then children fashion.  If this changes, then this code will not
3029  * work as designed.
3030  */
3031 void pcie_bus_configure_settings(struct pci_bus *bus)
3032 {
3033 	u8 smpss = 0;
3034 
3035 	if (!bus->self)
3036 		return;
3037 
3038 	if (!pci_is_pcie(bus->self))
3039 		return;
3040 
3041 	/*
3042 	 * FIXME - Peer to peer DMA is possible, though the endpoint would need
3043 	 * to be aware of the MPS of the destination.  To work around this,
3044 	 * simply force the MPS of the entire system to the smallest possible.
3045 	 */
3046 	if (pcie_bus_config == PCIE_BUS_PEER2PEER)
3047 		smpss = 0;
3048 
3049 	if (pcie_bus_config == PCIE_BUS_SAFE) {
3050 		smpss = bus->self->pcie_mpss;
3051 
3052 		pcie_find_smpss(bus->self, &smpss);
3053 		pci_walk_bus(bus, pcie_find_smpss, &smpss);
3054 	}
3055 
3056 	pcie_bus_configure_set(bus->self, &smpss);
3057 	pci_walk_bus(bus, pcie_bus_configure_set, &smpss);
3058 }
3059 EXPORT_SYMBOL_GPL(pcie_bus_configure_settings);
3060 
3061 /*
3062  * Called after each bus is probed, but before its children are examined.  This
3063  * is marked as __weak because multiple architectures define it.
3064  */
3065 void __weak pcibios_fixup_bus(struct pci_bus *bus)
3066 {
3067        /* nothing to do, expected to be removed in the future */
3068 }
3069 
3070 /**
3071  * pci_scan_child_bus_extend() - Scan devices below a bus
3072  * @bus: Bus to scan for devices
3073  * @available_buses: Total number of buses available (%0 does not try to
3074  *		     extend beyond the minimal)
3075  *
3076  * Scans devices below @bus including subordinate buses. Returns new
3077  * subordinate number including all the found devices. Passing
3078  * @available_buses causes the remaining bus space to be distributed
3079  * equally between hotplug-capable bridges to allow future extension of the
3080  * hierarchy.
3081  */
3082 static unsigned int pci_scan_child_bus_extend(struct pci_bus *bus,
3083 					      unsigned int available_buses)
3084 {
3085 	unsigned int used_buses, normal_bridges = 0, hotplug_bridges = 0;
3086 	unsigned int start = bus->busn_res.start;
3087 	unsigned int devnr, cmax, max = start;
3088 	struct pci_dev *dev;
3089 
3090 	dev_dbg(&bus->dev, "scanning bus\n");
3091 
3092 	/* Go find them, Rover! */
3093 	for (devnr = 0; devnr < PCI_MAX_NR_DEVS; devnr++)
3094 		pci_scan_slot(bus, PCI_DEVFN(devnr, 0));
3095 
3096 	/* Reserve buses for SR-IOV capability */
3097 	used_buses = pci_iov_bus_range(bus);
3098 	max += used_buses;
3099 
3100 	/*
3101 	 * After performing arch-dependent fixup of the bus, look behind
3102 	 * all PCI-to-PCI bridges on this bus.
3103 	 */
3104 	if (!bus->is_added) {
3105 		dev_dbg(&bus->dev, "fixups for bus\n");
3106 		pcibios_fixup_bus(bus);
3107 		bus->is_added = 1;
3108 	}
3109 
3110 	/*
3111 	 * Calculate how many hotplug bridges and normal bridges there
3112 	 * are on this bus. We will distribute the additional available
3113 	 * buses between hotplug bridges.
3114 	 */
3115 	for_each_pci_bridge(dev, bus) {
3116 		if (dev->is_hotplug_bridge)
3117 			hotplug_bridges++;
3118 		else
3119 			normal_bridges++;
3120 	}
3121 
3122 	/*
3123 	 * Scan bridges that are already configured. We don't touch them
3124 	 * unless they are misconfigured (which will be done in the second
3125 	 * scan below).
3126 	 */
3127 	for_each_pci_bridge(dev, bus) {
3128 		cmax = max;
3129 		max = pci_scan_bridge_extend(bus, dev, max, 0, 0);
3130 
3131 		/*
3132 		 * Reserve one bus for each bridge now to avoid extending
3133 		 * hotplug bridges too much during the second scan below.
3134 		 */
3135 		used_buses++;
3136 		if (max - cmax > 1)
3137 			used_buses += max - cmax - 1;
3138 	}
3139 
3140 	/* Scan bridges that need to be reconfigured */
3141 	for_each_pci_bridge(dev, bus) {
3142 		unsigned int buses = 0;
3143 
3144 		if (!hotplug_bridges && normal_bridges == 1) {
3145 			/*
3146 			 * There is only one bridge on the bus (upstream
3147 			 * port) so it gets all available buses which it
3148 			 * can then distribute to the possible hotplug
3149 			 * bridges below.
3150 			 */
3151 			buses = available_buses;
3152 		} else if (dev->is_hotplug_bridge) {
3153 			/*
3154 			 * Distribute the extra buses between hotplug
3155 			 * bridges if any.
3156 			 */
3157 			buses = available_buses / hotplug_bridges;
3158 			buses = min(buses, available_buses - used_buses + 1);
3159 		}
3160 
3161 		cmax = max;
3162 		max = pci_scan_bridge_extend(bus, dev, cmax, buses, 1);
3163 		/* One bus is already accounted so don't add it again */
3164 		if (max - cmax > 1)
3165 			used_buses += max - cmax - 1;
3166 	}
3167 
3168 	/*
3169 	 * Make sure a hotplug bridge has at least the minimum requested
3170 	 * number of buses but allow it to grow up to the maximum available
3171 	 * bus number if there is room.
3172 	 */
3173 	if (bus->self && bus->self->is_hotplug_bridge) {
3174 		used_buses = max(available_buses, pci_hotplug_bus_size - 1);
3175 		if (max - start < used_buses) {
3176 			max = start + used_buses;
3177 
3178 			/* Do not allocate more buses than we have room left */
3179 			if (max > bus->busn_res.end)
3180 				max = bus->busn_res.end;
3181 
3182 			dev_dbg(&bus->dev, "%pR extended by %#02x\n",
3183 				&bus->busn_res, max - start);
3184 		}
3185 	}
3186 
3187 	/*
3188 	 * We've scanned the bus and so we know all about what's on
3189 	 * the other side of any bridges that may be on this bus plus
3190 	 * any devices.
3191 	 *
3192 	 * Return how far we've got finding sub-buses.
3193 	 */
3194 	dev_dbg(&bus->dev, "bus scan returning with max=%02x\n", max);
3195 	return max;
3196 }
3197 
3198 /**
3199  * pci_scan_child_bus() - Scan devices below a bus
3200  * @bus: Bus to scan for devices
3201  *
3202  * Scans devices below @bus including subordinate buses. Returns new
3203  * subordinate number including all the found devices.
3204  */
3205 unsigned int pci_scan_child_bus(struct pci_bus *bus)
3206 {
3207 	return pci_scan_child_bus_extend(bus, 0);
3208 }
3209 EXPORT_SYMBOL_GPL(pci_scan_child_bus);
3210 
3211 /**
3212  * pcibios_root_bridge_prepare - Platform-specific host bridge setup
3213  * @bridge: Host bridge to set up
3214  *
3215  * Default empty implementation.  Replace with an architecture-specific setup
3216  * routine, if necessary.
3217  */
3218 int __weak pcibios_root_bridge_prepare(struct pci_host_bridge *bridge)
3219 {
3220 	return 0;
3221 }
3222 
3223 void __weak pcibios_add_bus(struct pci_bus *bus)
3224 {
3225 }
3226 
3227 void __weak pcibios_remove_bus(struct pci_bus *bus)
3228 {
3229 }
3230 
3231 struct pci_bus *pci_create_root_bus(struct device *parent, int bus,
3232 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
3233 {
3234 	int error;
3235 	struct pci_host_bridge *bridge;
3236 
3237 	bridge = pci_alloc_host_bridge(0);
3238 	if (!bridge)
3239 		return NULL;
3240 
3241 	bridge->dev.parent = parent;
3242 
3243 	list_splice_init(resources, &bridge->windows);
3244 	bridge->sysdata = sysdata;
3245 	bridge->busnr = bus;
3246 	bridge->ops = ops;
3247 
3248 	error = pci_register_host_bridge(bridge);
3249 	if (error < 0)
3250 		goto err_out;
3251 
3252 	return bridge->bus;
3253 
3254 err_out:
3255 	put_device(&bridge->dev);
3256 	return NULL;
3257 }
3258 EXPORT_SYMBOL_GPL(pci_create_root_bus);
3259 
3260 int pci_host_probe(struct pci_host_bridge *bridge)
3261 {
3262 	struct pci_bus *bus, *child;
3263 	int ret;
3264 
3265 	pci_lock_rescan_remove();
3266 	ret = pci_scan_root_bus_bridge(bridge);
3267 	pci_unlock_rescan_remove();
3268 	if (ret < 0) {
3269 		dev_err(bridge->dev.parent, "Scanning root bridge failed");
3270 		return ret;
3271 	}
3272 
3273 	bus = bridge->bus;
3274 
3275 	/* If we must preserve the resource configuration, claim now */
3276 	if (bridge->preserve_config)
3277 		pci_bus_claim_resources(bus);
3278 
3279 	/*
3280 	 * Assign whatever was left unassigned. If we didn't claim above,
3281 	 * this will reassign everything.
3282 	 */
3283 	pci_assign_unassigned_root_bus_resources(bus);
3284 
3285 	list_for_each_entry(child, &bus->children, node)
3286 		pcie_bus_configure_settings(child);
3287 
3288 	pci_lock_rescan_remove();
3289 	pci_bus_add_devices(bus);
3290 	pci_unlock_rescan_remove();
3291 
3292 	/*
3293 	 * Ensure pm_runtime_enable() is called for the controller drivers
3294 	 * before calling pci_host_probe(). The PM framework expects that
3295 	 * if the parent device supports runtime PM, it will be enabled
3296 	 * before child runtime PM is enabled.
3297 	 */
3298 	pm_runtime_set_active(&bridge->dev);
3299 	pm_runtime_no_callbacks(&bridge->dev);
3300 	devm_pm_runtime_enable(&bridge->dev);
3301 
3302 	return 0;
3303 }
3304 EXPORT_SYMBOL_GPL(pci_host_probe);
3305 
3306 int pci_bus_insert_busn_res(struct pci_bus *b, int bus, int bus_max)
3307 {
3308 	struct resource *res = &b->busn_res;
3309 	struct resource *parent_res, *conflict;
3310 
3311 	res->start = bus;
3312 	res->end = bus_max;
3313 	res->flags = IORESOURCE_BUS;
3314 
3315 	if (!pci_is_root_bus(b))
3316 		parent_res = &b->parent->busn_res;
3317 	else {
3318 		parent_res = get_pci_domain_busn_res(pci_domain_nr(b));
3319 		res->flags |= IORESOURCE_PCI_FIXED;
3320 	}
3321 
3322 	conflict = request_resource_conflict(parent_res, res);
3323 
3324 	if (conflict)
3325 		dev_info(&b->dev,
3326 			   "busn_res: can not insert %pR under %s%pR (conflicts with %s %pR)\n",
3327 			    res, pci_is_root_bus(b) ? "domain " : "",
3328 			    parent_res, conflict->name, conflict);
3329 
3330 	return conflict == NULL;
3331 }
3332 
3333 int pci_bus_update_busn_res_end(struct pci_bus *b, int bus_max)
3334 {
3335 	struct resource *res = &b->busn_res;
3336 	struct resource old_res = *res;
3337 	resource_size_t size;
3338 	int ret;
3339 
3340 	if (res->start > bus_max)
3341 		return -EINVAL;
3342 
3343 	size = bus_max - res->start + 1;
3344 	ret = adjust_resource(res, res->start, size);
3345 	dev_info(&b->dev, "busn_res: %pR end %s updated to %02x\n",
3346 			&old_res, ret ? "can not be" : "is", bus_max);
3347 
3348 	if (!ret && !res->parent)
3349 		pci_bus_insert_busn_res(b, res->start, res->end);
3350 
3351 	return ret;
3352 }
3353 
3354 void pci_bus_release_busn_res(struct pci_bus *b)
3355 {
3356 	struct resource *res = &b->busn_res;
3357 	int ret;
3358 
3359 	if (!res->flags || !res->parent)
3360 		return;
3361 
3362 	ret = release_resource(res);
3363 	dev_info(&b->dev, "busn_res: %pR %s released\n",
3364 			res, ret ? "can not be" : "is");
3365 }
3366 
3367 int pci_scan_root_bus_bridge(struct pci_host_bridge *bridge)
3368 {
3369 	struct resource_entry *window;
3370 	bool found = false;
3371 	struct pci_bus *b;
3372 	int max, bus, ret;
3373 
3374 	if (!bridge)
3375 		return -EINVAL;
3376 
3377 	resource_list_for_each_entry(window, &bridge->windows)
3378 		if (window->res->flags & IORESOURCE_BUS) {
3379 			bridge->busnr = window->res->start;
3380 			found = true;
3381 			break;
3382 		}
3383 
3384 	ret = pci_register_host_bridge(bridge);
3385 	if (ret < 0)
3386 		return ret;
3387 
3388 	b = bridge->bus;
3389 	bus = bridge->busnr;
3390 
3391 	if (!found) {
3392 		dev_info(&b->dev,
3393 		 "No busn resource found for root bus, will use [bus %02x-ff]\n",
3394 			bus);
3395 		pci_bus_insert_busn_res(b, bus, 255);
3396 	}
3397 
3398 	max = pci_scan_child_bus(b);
3399 
3400 	if (!found)
3401 		pci_bus_update_busn_res_end(b, max);
3402 
3403 	return 0;
3404 }
3405 EXPORT_SYMBOL(pci_scan_root_bus_bridge);
3406 
3407 struct pci_bus *pci_scan_root_bus(struct device *parent, int bus,
3408 		struct pci_ops *ops, void *sysdata, struct list_head *resources)
3409 {
3410 	struct resource_entry *window;
3411 	bool found = false;
3412 	struct pci_bus *b;
3413 	int max;
3414 
3415 	resource_list_for_each_entry(window, resources)
3416 		if (window->res->flags & IORESOURCE_BUS) {
3417 			found = true;
3418 			break;
3419 		}
3420 
3421 	b = pci_create_root_bus(parent, bus, ops, sysdata, resources);
3422 	if (!b)
3423 		return NULL;
3424 
3425 	if (!found) {
3426 		dev_info(&b->dev,
3427 		 "No busn resource found for root bus, will use [bus %02x-ff]\n",
3428 			bus);
3429 		pci_bus_insert_busn_res(b, bus, 255);
3430 	}
3431 
3432 	max = pci_scan_child_bus(b);
3433 
3434 	if (!found)
3435 		pci_bus_update_busn_res_end(b, max);
3436 
3437 	return b;
3438 }
3439 EXPORT_SYMBOL(pci_scan_root_bus);
3440 
3441 struct pci_bus *pci_scan_bus(int bus, struct pci_ops *ops,
3442 					void *sysdata)
3443 {
3444 	LIST_HEAD(resources);
3445 	struct pci_bus *b;
3446 
3447 	pci_add_resource(&resources, &ioport_resource);
3448 	pci_add_resource(&resources, &iomem_resource);
3449 	pci_add_resource(&resources, &busn_resource);
3450 	b = pci_create_root_bus(NULL, bus, ops, sysdata, &resources);
3451 	if (b) {
3452 		pci_scan_child_bus(b);
3453 	} else {
3454 		pci_free_resource_list(&resources);
3455 	}
3456 	return b;
3457 }
3458 EXPORT_SYMBOL(pci_scan_bus);
3459 
3460 /**
3461  * pci_rescan_bus_bridge_resize - Scan a PCI bus for devices
3462  * @bridge: PCI bridge for the bus to scan
3463  *
3464  * Scan a PCI bus and child buses for new devices, add them,
3465  * and enable them, resizing bridge mmio/io resource if necessary
3466  * and possible.  The caller must ensure the child devices are already
3467  * removed for resizing to occur.
3468  *
3469  * Returns the max number of subordinate bus discovered.
3470  */
3471 unsigned int pci_rescan_bus_bridge_resize(struct pci_dev *bridge)
3472 {
3473 	unsigned int max;
3474 	struct pci_bus *bus = bridge->subordinate;
3475 
3476 	max = pci_scan_child_bus(bus);
3477 
3478 	pci_assign_unassigned_bridge_resources(bridge);
3479 
3480 	pci_bus_add_devices(bus);
3481 
3482 	return max;
3483 }
3484 
3485 /**
3486  * pci_rescan_bus - Scan a PCI bus for devices
3487  * @bus: PCI bus to scan
3488  *
3489  * Scan a PCI bus and child buses for new devices, add them,
3490  * and enable them.
3491  *
3492  * Returns the max number of subordinate bus discovered.
3493  */
3494 unsigned int pci_rescan_bus(struct pci_bus *bus)
3495 {
3496 	unsigned int max;
3497 
3498 	max = pci_scan_child_bus(bus);
3499 	pci_assign_unassigned_bus_resources(bus);
3500 	pci_bus_add_devices(bus);
3501 
3502 	return max;
3503 }
3504 EXPORT_SYMBOL_GPL(pci_rescan_bus);
3505 
3506 /*
3507  * pci_rescan_bus(), pci_rescan_bus_bridge_resize() and PCI device removal
3508  * routines should always be executed under this mutex.
3509  */
3510 DEFINE_MUTEX(pci_rescan_remove_lock);
3511 
3512 void pci_lock_rescan_remove(void)
3513 {
3514 	mutex_lock(&pci_rescan_remove_lock);
3515 }
3516 EXPORT_SYMBOL_GPL(pci_lock_rescan_remove);
3517 
3518 void pci_unlock_rescan_remove(void)
3519 {
3520 	mutex_unlock(&pci_rescan_remove_lock);
3521 }
3522 EXPORT_SYMBOL_GPL(pci_unlock_rescan_remove);
3523 
3524 static int __init pci_sort_bf_cmp(const struct device *d_a,
3525 				  const struct device *d_b)
3526 {
3527 	const struct pci_dev *a = to_pci_dev(d_a);
3528 	const struct pci_dev *b = to_pci_dev(d_b);
3529 
3530 	if      (pci_domain_nr(a->bus) < pci_domain_nr(b->bus)) return -1;
3531 	else if (pci_domain_nr(a->bus) > pci_domain_nr(b->bus)) return  1;
3532 
3533 	if      (a->bus->number < b->bus->number) return -1;
3534 	else if (a->bus->number > b->bus->number) return  1;
3535 
3536 	if      (a->devfn < b->devfn) return -1;
3537 	else if (a->devfn > b->devfn) return  1;
3538 
3539 	return 0;
3540 }
3541 
3542 void __init pci_sort_breadthfirst(void)
3543 {
3544 	bus_sort_breadthfirst(&pci_bus_type, &pci_sort_bf_cmp);
3545 }
3546 
3547 int pci_hp_add_bridge(struct pci_dev *dev)
3548 {
3549 	struct pci_bus *parent = dev->bus;
3550 	int busnr, start = parent->busn_res.start;
3551 	unsigned int available_buses = 0;
3552 	int end = parent->busn_res.end;
3553 
3554 	for (busnr = start; busnr <= end; busnr++) {
3555 		if (!pci_find_bus(pci_domain_nr(parent), busnr))
3556 			break;
3557 	}
3558 	if (busnr-- > end) {
3559 		pci_err(dev, "No bus number available for hot-added bridge\n");
3560 		return -1;
3561 	}
3562 
3563 	/* Scan bridges that are already configured */
3564 	busnr = pci_scan_bridge(parent, dev, busnr, 0);
3565 
3566 	/*
3567 	 * Distribute the available bus numbers between hotplug-capable
3568 	 * bridges to make extending the chain later possible.
3569 	 */
3570 	available_buses = end - busnr;
3571 
3572 	/* Scan bridges that need to be reconfigured */
3573 	pci_scan_bridge_extend(parent, dev, busnr, available_buses, 1);
3574 
3575 	if (!dev->subordinate)
3576 		return -1;
3577 
3578 	return 0;
3579 }
3580 EXPORT_SYMBOL_GPL(pci_hp_add_bridge);
3581