xref: /linux/drivers/acpi/pci_root.c (revision 4b132aacb0768ac1e652cf517097ea6f237214b9)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *  pci_root.c - ACPI PCI Root Bridge Driver ($Revision: 40 $)
4  *
5  *  Copyright (C) 2001, 2002 Andy Grover <andrew.grover@intel.com>
6  *  Copyright (C) 2001, 2002 Paul Diefenbaugh <paul.s.diefenbaugh@intel.com>
7  */
8 
9 #define pr_fmt(fmt) "ACPI: " fmt
10 
11 #include <linux/kernel.h>
12 #include <linux/module.h>
13 #include <linux/init.h>
14 #include <linux/types.h>
15 #include <linux/mutex.h>
16 #include <linux/pm.h>
17 #include <linux/pm_runtime.h>
18 #include <linux/pci.h>
19 #include <linux/pci-acpi.h>
20 #include <linux/dmar.h>
21 #include <linux/acpi.h>
22 #include <linux/slab.h>
23 #include <linux/dmi.h>
24 #include <linux/platform_data/x86/apple.h>
25 #include "internal.h"
26 
27 #define ACPI_PCI_ROOT_CLASS		"pci_bridge"
28 #define ACPI_PCI_ROOT_DEVICE_NAME	"PCI Root Bridge"
29 static int acpi_pci_root_add(struct acpi_device *device,
30 			     const struct acpi_device_id *not_used);
31 static void acpi_pci_root_remove(struct acpi_device *device);
32 
33 static int acpi_pci_root_scan_dependent(struct acpi_device *adev)
34 {
35 	acpiphp_check_host_bridge(adev);
36 	return 0;
37 }
38 
39 #define ACPI_PCIE_REQ_SUPPORT (OSC_PCI_EXT_CONFIG_SUPPORT \
40 				| OSC_PCI_ASPM_SUPPORT \
41 				| OSC_PCI_CLOCK_PM_SUPPORT \
42 				| OSC_PCI_MSI_SUPPORT)
43 
44 static const struct acpi_device_id root_device_ids[] = {
45 	{"PNP0A03", 0},
46 	{"", 0},
47 };
48 
49 static struct acpi_scan_handler pci_root_handler = {
50 	.ids = root_device_ids,
51 	.attach = acpi_pci_root_add,
52 	.detach = acpi_pci_root_remove,
53 	.hotplug = {
54 		.enabled = true,
55 		.scan_dependent = acpi_pci_root_scan_dependent,
56 	},
57 };
58 
59 /**
60  * acpi_is_root_bridge - determine whether an ACPI CA node is a PCI root bridge
61  * @handle:  the ACPI CA node in question.
62  *
63  * Note: we could make this API take a struct acpi_device * instead, but
64  * for now, it's more convenient to operate on an acpi_handle.
65  */
66 int acpi_is_root_bridge(acpi_handle handle)
67 {
68 	struct acpi_device *device = acpi_fetch_acpi_dev(handle);
69 	int ret;
70 
71 	if (!device)
72 		return 0;
73 
74 	ret = acpi_match_device_ids(device, root_device_ids);
75 	if (ret)
76 		return 0;
77 	else
78 		return 1;
79 }
80 EXPORT_SYMBOL_GPL(acpi_is_root_bridge);
81 
82 static acpi_status
83 get_root_bridge_busnr_callback(struct acpi_resource *resource, void *data)
84 {
85 	struct resource *res = data;
86 	struct acpi_resource_address64 address;
87 	acpi_status status;
88 
89 	status = acpi_resource_to_address64(resource, &address);
90 	if (ACPI_FAILURE(status))
91 		return AE_OK;
92 
93 	if ((address.address.address_length > 0) &&
94 	    (address.resource_type == ACPI_BUS_NUMBER_RANGE)) {
95 		res->start = address.address.minimum;
96 		res->end = address.address.minimum + address.address.address_length - 1;
97 	}
98 
99 	return AE_OK;
100 }
101 
102 static acpi_status try_get_root_bridge_busnr(acpi_handle handle,
103 					     struct resource *res)
104 {
105 	acpi_status status;
106 
107 	res->start = -1;
108 	status =
109 	    acpi_walk_resources(handle, METHOD_NAME__CRS,
110 				get_root_bridge_busnr_callback, res);
111 	if (ACPI_FAILURE(status))
112 		return status;
113 	if (res->start == -1)
114 		return AE_ERROR;
115 	return AE_OK;
116 }
117 
118 struct pci_osc_bit_struct {
119 	u32 bit;
120 	char *desc;
121 };
122 
123 static struct pci_osc_bit_struct pci_osc_support_bit[] = {
124 	{ OSC_PCI_EXT_CONFIG_SUPPORT, "ExtendedConfig" },
125 	{ OSC_PCI_ASPM_SUPPORT, "ASPM" },
126 	{ OSC_PCI_CLOCK_PM_SUPPORT, "ClockPM" },
127 	{ OSC_PCI_SEGMENT_GROUPS_SUPPORT, "Segments" },
128 	{ OSC_PCI_MSI_SUPPORT, "MSI" },
129 	{ OSC_PCI_EDR_SUPPORT, "EDR" },
130 	{ OSC_PCI_HPX_TYPE_3_SUPPORT, "HPX-Type3" },
131 };
132 
133 static struct pci_osc_bit_struct pci_osc_control_bit[] = {
134 	{ OSC_PCI_EXPRESS_NATIVE_HP_CONTROL, "PCIeHotplug" },
135 	{ OSC_PCI_SHPC_NATIVE_HP_CONTROL, "SHPCHotplug" },
136 	{ OSC_PCI_EXPRESS_PME_CONTROL, "PME" },
137 	{ OSC_PCI_EXPRESS_AER_CONTROL, "AER" },
138 	{ OSC_PCI_EXPRESS_CAPABILITY_CONTROL, "PCIeCapability" },
139 	{ OSC_PCI_EXPRESS_LTR_CONTROL, "LTR" },
140 	{ OSC_PCI_EXPRESS_DPC_CONTROL, "DPC" },
141 };
142 
143 static struct pci_osc_bit_struct cxl_osc_support_bit[] = {
144 	{ OSC_CXL_1_1_PORT_REG_ACCESS_SUPPORT, "CXL11PortRegAccess" },
145 	{ OSC_CXL_2_0_PORT_DEV_REG_ACCESS_SUPPORT, "CXL20PortDevRegAccess" },
146 	{ OSC_CXL_PROTOCOL_ERR_REPORTING_SUPPORT, "CXLProtocolErrorReporting" },
147 	{ OSC_CXL_NATIVE_HP_SUPPORT, "CXLNativeHotPlug" },
148 };
149 
150 static struct pci_osc_bit_struct cxl_osc_control_bit[] = {
151 	{ OSC_CXL_ERROR_REPORTING_CONTROL, "CXLMemErrorReporting" },
152 };
153 
154 static void decode_osc_bits(struct acpi_pci_root *root, char *msg, u32 word,
155 			    struct pci_osc_bit_struct *table, int size)
156 {
157 	char buf[80];
158 	int i, len = 0;
159 	struct pci_osc_bit_struct *entry;
160 
161 	buf[0] = '\0';
162 	for (i = 0, entry = table; i < size; i++, entry++)
163 		if (word & entry->bit)
164 			len += scnprintf(buf + len, sizeof(buf) - len, "%s%s",
165 					len ? " " : "", entry->desc);
166 
167 	dev_info(&root->device->dev, "_OSC: %s [%s]\n", msg, buf);
168 }
169 
170 static void decode_osc_support(struct acpi_pci_root *root, char *msg, u32 word)
171 {
172 	decode_osc_bits(root, msg, word, pci_osc_support_bit,
173 			ARRAY_SIZE(pci_osc_support_bit));
174 }
175 
176 static void decode_osc_control(struct acpi_pci_root *root, char *msg, u32 word)
177 {
178 	decode_osc_bits(root, msg, word, pci_osc_control_bit,
179 			ARRAY_SIZE(pci_osc_control_bit));
180 }
181 
182 static void decode_cxl_osc_support(struct acpi_pci_root *root, char *msg, u32 word)
183 {
184 	decode_osc_bits(root, msg, word, cxl_osc_support_bit,
185 			ARRAY_SIZE(cxl_osc_support_bit));
186 }
187 
188 static void decode_cxl_osc_control(struct acpi_pci_root *root, char *msg, u32 word)
189 {
190 	decode_osc_bits(root, msg, word, cxl_osc_control_bit,
191 			ARRAY_SIZE(cxl_osc_control_bit));
192 }
193 
194 static inline bool is_pcie(struct acpi_pci_root *root)
195 {
196 	return root->bridge_type == ACPI_BRIDGE_TYPE_PCIE;
197 }
198 
199 static inline bool is_cxl(struct acpi_pci_root *root)
200 {
201 	return root->bridge_type == ACPI_BRIDGE_TYPE_CXL;
202 }
203 
204 static u8 pci_osc_uuid_str[] = "33DB4D5B-1FF7-401C-9657-7441C03DD766";
205 static u8 cxl_osc_uuid_str[] = "68F2D50B-C469-4d8A-BD3D-941A103FD3FC";
206 
207 static char *to_uuid(struct acpi_pci_root *root)
208 {
209 	if (is_cxl(root))
210 		return cxl_osc_uuid_str;
211 	return pci_osc_uuid_str;
212 }
213 
214 static int cap_length(struct acpi_pci_root *root)
215 {
216 	if (is_cxl(root))
217 		return sizeof(u32) * OSC_CXL_CAPABILITY_DWORDS;
218 	return sizeof(u32) * OSC_PCI_CAPABILITY_DWORDS;
219 }
220 
221 static acpi_status acpi_pci_run_osc(struct acpi_pci_root *root,
222 				    const u32 *capbuf, u32 *pci_control,
223 				    u32 *cxl_control)
224 {
225 	struct acpi_osc_context context = {
226 		.uuid_str = to_uuid(root),
227 		.rev = 1,
228 		.cap.length = cap_length(root),
229 		.cap.pointer = (void *)capbuf,
230 	};
231 	acpi_status status;
232 
233 	status = acpi_run_osc(root->device->handle, &context);
234 	if (ACPI_SUCCESS(status)) {
235 		*pci_control = acpi_osc_ctx_get_pci_control(&context);
236 		if (is_cxl(root))
237 			*cxl_control = acpi_osc_ctx_get_cxl_control(&context);
238 		kfree(context.ret.pointer);
239 	}
240 	return status;
241 }
242 
243 static acpi_status acpi_pci_query_osc(struct acpi_pci_root *root, u32 support,
244 				      u32 *control, u32 cxl_support,
245 				      u32 *cxl_control)
246 {
247 	acpi_status status;
248 	u32 pci_result, cxl_result, capbuf[OSC_CXL_CAPABILITY_DWORDS];
249 
250 	support |= root->osc_support_set;
251 
252 	capbuf[OSC_QUERY_DWORD] = OSC_QUERY_ENABLE;
253 	capbuf[OSC_SUPPORT_DWORD] = support;
254 	capbuf[OSC_CONTROL_DWORD] = *control | root->osc_control_set;
255 
256 	if (is_cxl(root)) {
257 		cxl_support |= root->osc_ext_support_set;
258 		capbuf[OSC_EXT_SUPPORT_DWORD] = cxl_support;
259 		capbuf[OSC_EXT_CONTROL_DWORD] = *cxl_control | root->osc_ext_control_set;
260 	}
261 
262 retry:
263 	status = acpi_pci_run_osc(root, capbuf, &pci_result, &cxl_result);
264 	if (ACPI_SUCCESS(status)) {
265 		root->osc_support_set = support;
266 		*control = pci_result;
267 		if (is_cxl(root)) {
268 			root->osc_ext_support_set = cxl_support;
269 			*cxl_control = cxl_result;
270 		}
271 	} else if (is_cxl(root)) {
272 		/*
273 		 * CXL _OSC is optional on CXL 1.1 hosts. Fall back to PCIe _OSC
274 		 * upon any failure using CXL _OSC.
275 		 */
276 		root->bridge_type = ACPI_BRIDGE_TYPE_PCIE;
277 		goto retry;
278 	}
279 	return status;
280 }
281 
282 struct acpi_pci_root *acpi_pci_find_root(acpi_handle handle)
283 {
284 	struct acpi_device *device = acpi_fetch_acpi_dev(handle);
285 	struct acpi_pci_root *root;
286 
287 	if (!device || acpi_match_device_ids(device, root_device_ids))
288 		return NULL;
289 
290 	root = acpi_driver_data(device);
291 
292 	return root;
293 }
294 EXPORT_SYMBOL_GPL(acpi_pci_find_root);
295 
296 /**
297  * acpi_get_pci_dev - convert ACPI CA handle to struct pci_dev
298  * @handle: the handle in question
299  *
300  * Given an ACPI CA handle, the desired PCI device is located in the
301  * list of PCI devices.
302  *
303  * If the device is found, its reference count is increased and this
304  * function returns a pointer to its data structure.  The caller must
305  * decrement the reference count by calling pci_dev_put().
306  * If no device is found, %NULL is returned.
307  */
308 struct pci_dev *acpi_get_pci_dev(acpi_handle handle)
309 {
310 	struct acpi_device *adev = acpi_fetch_acpi_dev(handle);
311 	struct acpi_device_physical_node *pn;
312 	struct pci_dev *pci_dev = NULL;
313 
314 	if (!adev)
315 		return NULL;
316 
317 	mutex_lock(&adev->physical_node_lock);
318 
319 	list_for_each_entry(pn, &adev->physical_node_list, node) {
320 		if (dev_is_pci(pn->dev)) {
321 			get_device(pn->dev);
322 			pci_dev = to_pci_dev(pn->dev);
323 			break;
324 		}
325 	}
326 
327 	mutex_unlock(&adev->physical_node_lock);
328 
329 	return pci_dev;
330 }
331 EXPORT_SYMBOL_GPL(acpi_get_pci_dev);
332 
333 /**
334  * acpi_pci_osc_control_set - Request control of PCI root _OSC features.
335  * @handle: ACPI handle of a PCI root bridge (or PCIe Root Complex).
336  * @mask: Mask of _OSC bits to request control of, place to store control mask.
337  * @support: _OSC supported capability.
338  * @cxl_mask: Mask of CXL _OSC control bits, place to store control mask.
339  * @cxl_support: CXL _OSC supported capability.
340  *
341  * Run _OSC query for @mask and if that is successful, compare the returned
342  * mask of control bits with @req.  If all of the @req bits are set in the
343  * returned mask, run _OSC request for it.
344  *
345  * The variable at the @mask address may be modified regardless of whether or
346  * not the function returns success.  On success it will contain the mask of
347  * _OSC bits the BIOS has granted control of, but its contents are meaningless
348  * on failure.
349  **/
350 static acpi_status acpi_pci_osc_control_set(acpi_handle handle, u32 *mask,
351 					    u32 support, u32 *cxl_mask,
352 					    u32 cxl_support)
353 {
354 	u32 req = OSC_PCI_EXPRESS_CAPABILITY_CONTROL;
355 	struct acpi_pci_root *root;
356 	acpi_status status;
357 	u32 ctrl, cxl_ctrl = 0, capbuf[OSC_CXL_CAPABILITY_DWORDS];
358 
359 	if (!mask)
360 		return AE_BAD_PARAMETER;
361 
362 	root = acpi_pci_find_root(handle);
363 	if (!root)
364 		return AE_NOT_EXIST;
365 
366 	ctrl   = *mask;
367 	*mask |= root->osc_control_set;
368 
369 	if (is_cxl(root)) {
370 		cxl_ctrl = *cxl_mask;
371 		*cxl_mask |= root->osc_ext_control_set;
372 	}
373 
374 	/* Need to check the available controls bits before requesting them. */
375 	do {
376 		u32 pci_missing = 0, cxl_missing = 0;
377 
378 		status = acpi_pci_query_osc(root, support, mask, cxl_support,
379 					    cxl_mask);
380 		if (ACPI_FAILURE(status))
381 			return status;
382 		if (is_cxl(root)) {
383 			if (ctrl == *mask && cxl_ctrl == *cxl_mask)
384 				break;
385 			pci_missing = ctrl & ~(*mask);
386 			cxl_missing = cxl_ctrl & ~(*cxl_mask);
387 		} else {
388 			if (ctrl == *mask)
389 				break;
390 			pci_missing = ctrl & ~(*mask);
391 		}
392 		if (pci_missing)
393 			decode_osc_control(root, "platform does not support",
394 					   pci_missing);
395 		if (cxl_missing)
396 			decode_cxl_osc_control(root, "CXL platform does not support",
397 					   cxl_missing);
398 		ctrl = *mask;
399 		cxl_ctrl = *cxl_mask;
400 	} while (*mask || *cxl_mask);
401 
402 	/* No need to request _OSC if the control was already granted. */
403 	if ((root->osc_control_set & ctrl) == ctrl &&
404 	    (root->osc_ext_control_set & cxl_ctrl) == cxl_ctrl)
405 		return AE_OK;
406 
407 	if ((ctrl & req) != req) {
408 		decode_osc_control(root, "not requesting control; platform does not support",
409 				   req & ~(ctrl));
410 		return AE_SUPPORT;
411 	}
412 
413 	capbuf[OSC_QUERY_DWORD] = 0;
414 	capbuf[OSC_SUPPORT_DWORD] = root->osc_support_set;
415 	capbuf[OSC_CONTROL_DWORD] = ctrl;
416 	if (is_cxl(root)) {
417 		capbuf[OSC_EXT_SUPPORT_DWORD] = root->osc_ext_support_set;
418 		capbuf[OSC_EXT_CONTROL_DWORD] = cxl_ctrl;
419 	}
420 
421 	status = acpi_pci_run_osc(root, capbuf, mask, cxl_mask);
422 	if (ACPI_FAILURE(status))
423 		return status;
424 
425 	root->osc_control_set = *mask;
426 	root->osc_ext_control_set = *cxl_mask;
427 	return AE_OK;
428 }
429 
430 static u32 calculate_support(void)
431 {
432 	u32 support;
433 
434 	/*
435 	 * All supported architectures that use ACPI have support for
436 	 * PCI domains, so we indicate this in _OSC support capabilities.
437 	 */
438 	support = OSC_PCI_SEGMENT_GROUPS_SUPPORT;
439 	support |= OSC_PCI_HPX_TYPE_3_SUPPORT;
440 	if (pci_ext_cfg_avail())
441 		support |= OSC_PCI_EXT_CONFIG_SUPPORT;
442 	if (pcie_aspm_support_enabled())
443 		support |= OSC_PCI_ASPM_SUPPORT | OSC_PCI_CLOCK_PM_SUPPORT;
444 	if (pci_msi_enabled())
445 		support |= OSC_PCI_MSI_SUPPORT;
446 	if (IS_ENABLED(CONFIG_PCIE_EDR))
447 		support |= OSC_PCI_EDR_SUPPORT;
448 
449 	return support;
450 }
451 
452 /*
453  * Background on hotplug support, and making it depend on only
454  * CONFIG_HOTPLUG_PCI_PCIE vs. also considering CONFIG_MEMORY_HOTPLUG:
455  *
456  * CONFIG_ACPI_HOTPLUG_MEMORY does depend on CONFIG_MEMORY_HOTPLUG, but
457  * there is no existing _OSC for memory hotplug support. The reason is that
458  * ACPI memory hotplug requires the OS to acknowledge / coordinate with
459  * memory plug events via a scan handler. On the CXL side the equivalent
460  * would be if Linux supported the Mechanical Retention Lock [1], or
461  * otherwise had some coordination for the driver of a PCI device
462  * undergoing hotplug to be consulted on whether the hotplug should
463  * proceed or not.
464  *
465  * The concern is that if Linux says no to supporting CXL hotplug then
466  * the BIOS may say no to giving the OS hotplug control of any other PCIe
467  * device. So the question here is not whether hotplug is enabled, it's
468  * whether it is handled natively by the at all OS, and if
469  * CONFIG_HOTPLUG_PCI_PCIE is enabled then the answer is "yes".
470  *
471  * Otherwise, the plan for CXL coordinated remove, since the kernel does
472  * not support blocking hotplug, is to require the memory device to be
473  * disabled before hotplug is attempted. When CONFIG_MEMORY_HOTPLUG is
474  * disabled that step will fail and the remove attempt cancelled by the
475  * user. If that is not honored and the card is removed anyway then it
476  * does not matter if CONFIG_MEMORY_HOTPLUG is enabled or not, it will
477  * cause a crash and other badness.
478  *
479  * Therefore, just say yes to CXL hotplug and require removal to
480  * be coordinated by userspace unless and until the kernel grows better
481  * mechanisms for doing "managed" removal of devices in consultation with
482  * the driver.
483  *
484  * [1]: https://lore.kernel.org/all/20201122014203.4706-1-ashok.raj@intel.com/
485  */
486 static u32 calculate_cxl_support(void)
487 {
488 	u32 support;
489 
490 	support = OSC_CXL_2_0_PORT_DEV_REG_ACCESS_SUPPORT;
491 	support |= OSC_CXL_1_1_PORT_REG_ACCESS_SUPPORT;
492 	if (pci_aer_available())
493 		support |= OSC_CXL_PROTOCOL_ERR_REPORTING_SUPPORT;
494 	if (IS_ENABLED(CONFIG_HOTPLUG_PCI_PCIE))
495 		support |= OSC_CXL_NATIVE_HP_SUPPORT;
496 
497 	return support;
498 }
499 
500 static u32 calculate_control(void)
501 {
502 	u32 control;
503 
504 	control = OSC_PCI_EXPRESS_CAPABILITY_CONTROL
505 		| OSC_PCI_EXPRESS_PME_CONTROL;
506 
507 	if (IS_ENABLED(CONFIG_PCIEASPM))
508 		control |= OSC_PCI_EXPRESS_LTR_CONTROL;
509 
510 	if (IS_ENABLED(CONFIG_HOTPLUG_PCI_PCIE))
511 		control |= OSC_PCI_EXPRESS_NATIVE_HP_CONTROL;
512 
513 	if (IS_ENABLED(CONFIG_HOTPLUG_PCI_SHPC))
514 		control |= OSC_PCI_SHPC_NATIVE_HP_CONTROL;
515 
516 	if (pci_aer_available())
517 		control |= OSC_PCI_EXPRESS_AER_CONTROL;
518 
519 	/*
520 	 * Per the Downstream Port Containment Related Enhancements ECN to
521 	 * the PCI Firmware Spec, r3.2, sec 4.5.1, table 4-5,
522 	 * OSC_PCI_EXPRESS_DPC_CONTROL indicates the OS supports both DPC
523 	 * and EDR.
524 	 */
525 	if (IS_ENABLED(CONFIG_PCIE_DPC) && IS_ENABLED(CONFIG_PCIE_EDR))
526 		control |= OSC_PCI_EXPRESS_DPC_CONTROL;
527 
528 	return control;
529 }
530 
531 static u32 calculate_cxl_control(void)
532 {
533 	u32 control = 0;
534 
535 	if (IS_ENABLED(CONFIG_MEMORY_FAILURE))
536 		control |= OSC_CXL_ERROR_REPORTING_CONTROL;
537 
538 	return control;
539 }
540 
541 static bool os_control_query_checks(struct acpi_pci_root *root, u32 support)
542 {
543 	struct acpi_device *device = root->device;
544 
545 	if (pcie_ports_disabled) {
546 		dev_info(&device->dev, "PCIe port services disabled; not requesting _OSC control\n");
547 		return false;
548 	}
549 
550 	if ((support & ACPI_PCIE_REQ_SUPPORT) != ACPI_PCIE_REQ_SUPPORT) {
551 		decode_osc_support(root, "not requesting OS control; OS requires",
552 				   ACPI_PCIE_REQ_SUPPORT);
553 		return false;
554 	}
555 
556 	return true;
557 }
558 
559 static void negotiate_os_control(struct acpi_pci_root *root, int *no_aspm)
560 {
561 	u32 support, control = 0, requested = 0;
562 	u32 cxl_support = 0, cxl_control = 0, cxl_requested = 0;
563 	acpi_status status;
564 	struct acpi_device *device = root->device;
565 	acpi_handle handle = device->handle;
566 
567 	/*
568 	 * Apple always return failure on _OSC calls when _OSI("Darwin") has
569 	 * been called successfully. We know the feature set supported by the
570 	 * platform, so avoid calling _OSC at all
571 	 */
572 	if (x86_apple_machine) {
573 		root->osc_control_set = ~OSC_PCI_EXPRESS_PME_CONTROL;
574 		decode_osc_control(root, "OS assumes control of",
575 				   root->osc_control_set);
576 		return;
577 	}
578 
579 	support = calculate_support();
580 
581 	decode_osc_support(root, "OS supports", support);
582 
583 	if (os_control_query_checks(root, support))
584 		requested = control = calculate_control();
585 
586 	if (is_cxl(root)) {
587 		cxl_support = calculate_cxl_support();
588 		decode_cxl_osc_support(root, "OS supports", cxl_support);
589 		cxl_requested = cxl_control = calculate_cxl_control();
590 	}
591 
592 	status = acpi_pci_osc_control_set(handle, &control, support,
593 					  &cxl_control, cxl_support);
594 	if (ACPI_SUCCESS(status)) {
595 		if (control)
596 			decode_osc_control(root, "OS now controls", control);
597 		if (cxl_control)
598 			decode_cxl_osc_control(root, "OS now controls",
599 					   cxl_control);
600 
601 		if (acpi_gbl_FADT.boot_flags & ACPI_FADT_NO_ASPM) {
602 			/*
603 			 * We have ASPM control, but the FADT indicates that
604 			 * it's unsupported. Leave existing configuration
605 			 * intact and prevent the OS from touching it.
606 			 */
607 			dev_info(&device->dev, "FADT indicates ASPM is unsupported, using BIOS configuration\n");
608 			*no_aspm = 1;
609 		}
610 	} else {
611 		/*
612 		 * We want to disable ASPM here, but aspm_disabled
613 		 * needs to remain in its state from boot so that we
614 		 * properly handle PCIe 1.1 devices.  So we set this
615 		 * flag here, to defer the action until after the ACPI
616 		 * root scan.
617 		 */
618 		*no_aspm = 1;
619 
620 		/* _OSC is optional for PCI host bridges */
621 		if (status == AE_NOT_FOUND && !is_pcie(root))
622 			return;
623 
624 		if (control) {
625 			decode_osc_control(root, "OS requested", requested);
626 			decode_osc_control(root, "platform willing to grant", control);
627 		}
628 		if (cxl_control) {
629 			decode_cxl_osc_control(root, "OS requested", cxl_requested);
630 			decode_cxl_osc_control(root, "platform willing to grant",
631 					   cxl_control);
632 		}
633 
634 		dev_info(&device->dev, "_OSC: platform retains control of PCIe features (%s)\n",
635 			 acpi_format_exception(status));
636 	}
637 }
638 
639 static int acpi_pci_root_add(struct acpi_device *device,
640 			     const struct acpi_device_id *not_used)
641 {
642 	unsigned long long segment, bus;
643 	acpi_status status;
644 	int result;
645 	struct acpi_pci_root *root;
646 	acpi_handle handle = device->handle;
647 	int no_aspm = 0;
648 	bool hotadd = system_state == SYSTEM_RUNNING;
649 	const char *acpi_hid;
650 
651 	root = kzalloc(sizeof(struct acpi_pci_root), GFP_KERNEL);
652 	if (!root)
653 		return -ENOMEM;
654 
655 	segment = 0;
656 	status = acpi_evaluate_integer(handle, METHOD_NAME__SEG, NULL,
657 				       &segment);
658 	if (ACPI_FAILURE(status) && status != AE_NOT_FOUND) {
659 		dev_err(&device->dev,  "can't evaluate _SEG\n");
660 		result = -ENODEV;
661 		goto end;
662 	}
663 
664 	/* Check _CRS first, then _BBN.  If no _BBN, default to zero. */
665 	root->secondary.flags = IORESOURCE_BUS;
666 	status = try_get_root_bridge_busnr(handle, &root->secondary);
667 	if (ACPI_FAILURE(status)) {
668 		/*
669 		 * We need both the start and end of the downstream bus range
670 		 * to interpret _CBA (MMCONFIG base address), so it really is
671 		 * supposed to be in _CRS.  If we don't find it there, all we
672 		 * can do is assume [_BBN-0xFF] or [0-0xFF].
673 		 */
674 		root->secondary.end = 0xFF;
675 		dev_warn(&device->dev,
676 			 FW_BUG "no secondary bus range in _CRS\n");
677 		status = acpi_evaluate_integer(handle, METHOD_NAME__BBN,
678 					       NULL, &bus);
679 		if (ACPI_SUCCESS(status))
680 			root->secondary.start = bus;
681 		else if (status == AE_NOT_FOUND)
682 			root->secondary.start = 0;
683 		else {
684 			dev_err(&device->dev, "can't evaluate _BBN\n");
685 			result = -ENODEV;
686 			goto end;
687 		}
688 	}
689 
690 	root->device = device;
691 	root->segment = segment & 0xFFFF;
692 	strcpy(acpi_device_name(device), ACPI_PCI_ROOT_DEVICE_NAME);
693 	strcpy(acpi_device_class(device), ACPI_PCI_ROOT_CLASS);
694 	device->driver_data = root;
695 
696 	if (hotadd && dmar_device_add(handle)) {
697 		result = -ENXIO;
698 		goto end;
699 	}
700 
701 	pr_info("%s [%s] (domain %04x %pR)\n",
702 	       acpi_device_name(device), acpi_device_bid(device),
703 	       root->segment, &root->secondary);
704 
705 	root->mcfg_addr = acpi_pci_root_get_mcfg_addr(handle);
706 
707 	acpi_hid = acpi_device_hid(root->device);
708 	if (strcmp(acpi_hid, "PNP0A08") == 0)
709 		root->bridge_type = ACPI_BRIDGE_TYPE_PCIE;
710 	else if (strcmp(acpi_hid, "ACPI0016") == 0)
711 		root->bridge_type = ACPI_BRIDGE_TYPE_CXL;
712 	else
713 		dev_dbg(&device->dev, "Assuming non-PCIe host bridge\n");
714 
715 	negotiate_os_control(root, &no_aspm);
716 
717 	/*
718 	 * TBD: Need PCI interface for enumeration/configuration of roots.
719 	 */
720 
721 	/*
722 	 * Scan the Root Bridge
723 	 * --------------------
724 	 * Must do this prior to any attempt to bind the root device, as the
725 	 * PCI namespace does not get created until this call is made (and
726 	 * thus the root bridge's pci_dev does not exist).
727 	 */
728 	root->bus = pci_acpi_scan_root(root);
729 	if (!root->bus) {
730 		dev_err(&device->dev,
731 			"Bus %04x:%02x not present in PCI namespace\n",
732 			root->segment, (unsigned int)root->secondary.start);
733 		device->driver_data = NULL;
734 		result = -ENODEV;
735 		goto remove_dmar;
736 	}
737 
738 	if (no_aspm)
739 		pcie_no_aspm();
740 
741 	pci_acpi_add_bus_pm_notifier(device);
742 	device_set_wakeup_capable(root->bus->bridge, device->wakeup.flags.valid);
743 
744 	if (hotadd) {
745 		pcibios_resource_survey_bus(root->bus);
746 		pci_assign_unassigned_root_bus_resources(root->bus);
747 		/*
748 		 * This is only called for the hotadd case. For the boot-time
749 		 * case, we need to wait until after PCI initialization in
750 		 * order to deal with IOAPICs mapped in on a PCI BAR.
751 		 *
752 		 * This is currently x86-specific, because acpi_ioapic_add()
753 		 * is an empty function without CONFIG_ACPI_HOTPLUG_IOAPIC.
754 		 * And CONFIG_ACPI_HOTPLUG_IOAPIC depends on CONFIG_X86_IO_APIC
755 		 * (see drivers/acpi/Kconfig).
756 		 */
757 		acpi_ioapic_add(root->device->handle);
758 	}
759 
760 	pci_lock_rescan_remove();
761 	pci_bus_add_devices(root->bus);
762 	pci_unlock_rescan_remove();
763 	return 1;
764 
765 remove_dmar:
766 	if (hotadd)
767 		dmar_device_remove(handle);
768 end:
769 	kfree(root);
770 	return result;
771 }
772 
773 static void acpi_pci_root_remove(struct acpi_device *device)
774 {
775 	struct acpi_pci_root *root = acpi_driver_data(device);
776 
777 	pci_lock_rescan_remove();
778 
779 	pci_stop_root_bus(root->bus);
780 
781 	pci_ioapic_remove(root);
782 	device_set_wakeup_capable(root->bus->bridge, false);
783 	pci_acpi_remove_bus_pm_notifier(device);
784 
785 	pci_remove_root_bus(root->bus);
786 	WARN_ON(acpi_ioapic_remove(root));
787 
788 	dmar_device_remove(device->handle);
789 
790 	pci_unlock_rescan_remove();
791 
792 	kfree(root);
793 }
794 
795 /*
796  * Following code to support acpi_pci_root_create() is copied from
797  * arch/x86/pci/acpi.c and modified so it could be reused by x86, IA64
798  * and ARM64.
799  */
800 static void acpi_pci_root_validate_resources(struct device *dev,
801 					     struct list_head *resources,
802 					     unsigned long type)
803 {
804 	LIST_HEAD(list);
805 	struct resource *res1, *res2, *root = NULL;
806 	struct resource_entry *tmp, *entry, *entry2;
807 
808 	BUG_ON((type & (IORESOURCE_MEM | IORESOURCE_IO)) == 0);
809 	root = (type & IORESOURCE_MEM) ? &iomem_resource : &ioport_resource;
810 
811 	list_splice_init(resources, &list);
812 	resource_list_for_each_entry_safe(entry, tmp, &list) {
813 		bool free = false;
814 		resource_size_t end;
815 
816 		res1 = entry->res;
817 		if (!(res1->flags & type))
818 			goto next;
819 
820 		/* Exclude non-addressable range or non-addressable portion */
821 		end = min(res1->end, root->end);
822 		if (end <= res1->start) {
823 			dev_info(dev, "host bridge window %pR (ignored, not CPU addressable)\n",
824 				 res1);
825 			free = true;
826 			goto next;
827 		} else if (res1->end != end) {
828 			dev_info(dev, "host bridge window %pR ([%#llx-%#llx] ignored, not CPU addressable)\n",
829 				 res1, (unsigned long long)end + 1,
830 				 (unsigned long long)res1->end);
831 			res1->end = end;
832 		}
833 
834 		resource_list_for_each_entry(entry2, resources) {
835 			res2 = entry2->res;
836 			if (!(res2->flags & type))
837 				continue;
838 
839 			/*
840 			 * I don't like throwing away windows because then
841 			 * our resources no longer match the ACPI _CRS, but
842 			 * the kernel resource tree doesn't allow overlaps.
843 			 */
844 			if (resource_union(res1, res2, res2)) {
845 				dev_info(dev, "host bridge window expanded to %pR; %pR ignored\n",
846 					 res2, res1);
847 				free = true;
848 				goto next;
849 			}
850 		}
851 
852 next:
853 		resource_list_del(entry);
854 		if (free)
855 			resource_list_free_entry(entry);
856 		else
857 			resource_list_add_tail(entry, resources);
858 	}
859 }
860 
861 static void acpi_pci_root_remap_iospace(struct fwnode_handle *fwnode,
862 			struct resource_entry *entry)
863 {
864 #ifdef PCI_IOBASE
865 	struct resource *res = entry->res;
866 	resource_size_t cpu_addr = res->start;
867 	resource_size_t pci_addr = cpu_addr - entry->offset;
868 	resource_size_t length = resource_size(res);
869 	unsigned long port;
870 
871 	if (pci_register_io_range(fwnode, cpu_addr, length))
872 		goto err;
873 
874 	port = pci_address_to_pio(cpu_addr);
875 	if (port == (unsigned long)-1)
876 		goto err;
877 
878 	res->start = port;
879 	res->end = port + length - 1;
880 	entry->offset = port - pci_addr;
881 
882 	if (pci_remap_iospace(res, cpu_addr) < 0)
883 		goto err;
884 
885 	pr_info("Remapped I/O %pa to %pR\n", &cpu_addr, res);
886 	return;
887 err:
888 	res->flags |= IORESOURCE_DISABLED;
889 #endif
890 }
891 
892 int acpi_pci_probe_root_resources(struct acpi_pci_root_info *info)
893 {
894 	int ret;
895 	struct list_head *list = &info->resources;
896 	struct acpi_device *device = info->bridge;
897 	struct resource_entry *entry, *tmp;
898 	unsigned long flags;
899 
900 	flags = IORESOURCE_IO | IORESOURCE_MEM | IORESOURCE_MEM_8AND16BIT;
901 	ret = acpi_dev_get_resources(device, list,
902 				     acpi_dev_filter_resource_type_cb,
903 				     (void *)flags);
904 	if (ret < 0)
905 		dev_warn(&device->dev,
906 			 "failed to parse _CRS method, error code %d\n", ret);
907 	else if (ret == 0)
908 		dev_dbg(&device->dev,
909 			"no IO and memory resources present in _CRS\n");
910 	else {
911 		resource_list_for_each_entry_safe(entry, tmp, list) {
912 			if (entry->res->flags & IORESOURCE_IO)
913 				acpi_pci_root_remap_iospace(&device->fwnode,
914 						entry);
915 
916 			if (entry->res->flags & IORESOURCE_DISABLED)
917 				resource_list_destroy_entry(entry);
918 			else
919 				entry->res->name = info->name;
920 		}
921 		acpi_pci_root_validate_resources(&device->dev, list,
922 						 IORESOURCE_MEM);
923 		acpi_pci_root_validate_resources(&device->dev, list,
924 						 IORESOURCE_IO);
925 	}
926 
927 	return ret;
928 }
929 
930 static void pci_acpi_root_add_resources(struct acpi_pci_root_info *info)
931 {
932 	struct resource_entry *entry, *tmp;
933 	struct resource *res, *conflict, *root = NULL;
934 
935 	resource_list_for_each_entry_safe(entry, tmp, &info->resources) {
936 		res = entry->res;
937 		if (res->flags & IORESOURCE_MEM)
938 			root = &iomem_resource;
939 		else if (res->flags & IORESOURCE_IO)
940 			root = &ioport_resource;
941 		else
942 			continue;
943 
944 		/*
945 		 * Some legacy x86 host bridge drivers use iomem_resource and
946 		 * ioport_resource as default resource pool, skip it.
947 		 */
948 		if (res == root)
949 			continue;
950 
951 		conflict = insert_resource_conflict(root, res);
952 		if (conflict) {
953 			dev_info(&info->bridge->dev,
954 				 "ignoring host bridge window %pR (conflicts with %s %pR)\n",
955 				 res, conflict->name, conflict);
956 			resource_list_destroy_entry(entry);
957 		}
958 	}
959 }
960 
961 static void __acpi_pci_root_release_info(struct acpi_pci_root_info *info)
962 {
963 	struct resource *res;
964 	struct resource_entry *entry, *tmp;
965 
966 	if (!info)
967 		return;
968 
969 	resource_list_for_each_entry_safe(entry, tmp, &info->resources) {
970 		res = entry->res;
971 		if (res->parent &&
972 		    (res->flags & (IORESOURCE_MEM | IORESOURCE_IO)))
973 			release_resource(res);
974 		resource_list_destroy_entry(entry);
975 	}
976 
977 	info->ops->release_info(info);
978 }
979 
980 static void acpi_pci_root_release_info(struct pci_host_bridge *bridge)
981 {
982 	struct resource *res;
983 	struct resource_entry *entry;
984 
985 	resource_list_for_each_entry(entry, &bridge->windows) {
986 		res = entry->res;
987 		if (res->flags & IORESOURCE_IO)
988 			pci_unmap_iospace(res);
989 		if (res->parent &&
990 		    (res->flags & (IORESOURCE_MEM | IORESOURCE_IO)))
991 			release_resource(res);
992 	}
993 	__acpi_pci_root_release_info(bridge->release_data);
994 }
995 
996 struct pci_bus *acpi_pci_root_create(struct acpi_pci_root *root,
997 				     struct acpi_pci_root_ops *ops,
998 				     struct acpi_pci_root_info *info,
999 				     void *sysdata)
1000 {
1001 	int ret, busnum = root->secondary.start;
1002 	struct acpi_device *device = root->device;
1003 	int node = acpi_get_node(device->handle);
1004 	struct pci_bus *bus;
1005 	struct pci_host_bridge *host_bridge;
1006 
1007 	info->root = root;
1008 	info->bridge = device;
1009 	info->ops = ops;
1010 	INIT_LIST_HEAD(&info->resources);
1011 	snprintf(info->name, sizeof(info->name), "PCI Bus %04x:%02x",
1012 		 root->segment, busnum);
1013 
1014 	if (ops->init_info && ops->init_info(info))
1015 		goto out_release_info;
1016 	if (ops->prepare_resources)
1017 		ret = ops->prepare_resources(info);
1018 	else
1019 		ret = acpi_pci_probe_root_resources(info);
1020 	if (ret < 0)
1021 		goto out_release_info;
1022 
1023 	pci_acpi_root_add_resources(info);
1024 	pci_add_resource(&info->resources, &root->secondary);
1025 	bus = pci_create_root_bus(NULL, busnum, ops->pci_ops,
1026 				  sysdata, &info->resources);
1027 	if (!bus)
1028 		goto out_release_info;
1029 
1030 	host_bridge = to_pci_host_bridge(bus->bridge);
1031 	if (!(root->osc_control_set & OSC_PCI_EXPRESS_NATIVE_HP_CONTROL))
1032 		host_bridge->native_pcie_hotplug = 0;
1033 	if (!(root->osc_control_set & OSC_PCI_SHPC_NATIVE_HP_CONTROL))
1034 		host_bridge->native_shpc_hotplug = 0;
1035 	if (!(root->osc_control_set & OSC_PCI_EXPRESS_AER_CONTROL))
1036 		host_bridge->native_aer = 0;
1037 	if (!(root->osc_control_set & OSC_PCI_EXPRESS_PME_CONTROL))
1038 		host_bridge->native_pme = 0;
1039 	if (!(root->osc_control_set & OSC_PCI_EXPRESS_LTR_CONTROL))
1040 		host_bridge->native_ltr = 0;
1041 	if (!(root->osc_control_set & OSC_PCI_EXPRESS_DPC_CONTROL))
1042 		host_bridge->native_dpc = 0;
1043 
1044 	if (!(root->osc_ext_control_set & OSC_CXL_ERROR_REPORTING_CONTROL))
1045 		host_bridge->native_cxl_error = 0;
1046 
1047 	acpi_dev_power_up_children_with_adr(device);
1048 
1049 	pci_scan_child_bus(bus);
1050 	pci_set_host_bridge_release(host_bridge, acpi_pci_root_release_info,
1051 				    info);
1052 	if (node != NUMA_NO_NODE)
1053 		dev_printk(KERN_DEBUG, &bus->dev, "on NUMA node %d\n", node);
1054 	return bus;
1055 
1056 out_release_info:
1057 	__acpi_pci_root_release_info(info);
1058 	return NULL;
1059 }
1060 
1061 void __init acpi_pci_root_init(void)
1062 {
1063 	if (acpi_pci_disabled)
1064 		return;
1065 
1066 	pci_acpi_crs_quirks();
1067 	acpi_scan_add_handler_with_hotplug(&pci_root_handler, "pci_root");
1068 }
1069