1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * Firmware-Assisted Dump support on POWER platform (OPAL).
4  *
5  * Copyright 2019, Hari Bathini, IBM Corporation.
6  */
7 
8 #define pr_fmt(fmt) "opal fadump: " fmt
9 
10 #include <linux/string.h>
11 #include <linux/seq_file.h>
12 #include <linux/of.h>
13 #include <linux/of_fdt.h>
14 #include <linux/libfdt.h>
15 #include <linux/mm.h>
16 #include <linux/crash_dump.h>
17 
18 #include <asm/page.h>
19 #include <asm/opal.h>
20 #include <asm/fadump-internal.h>
21 
22 #include "opal-fadump.h"
23 
24 
25 #ifdef CONFIG_PRESERVE_FA_DUMP
26 /*
27  * When dump is active but PRESERVE_FA_DUMP is enabled on the kernel,
28  * ensure crash data is preserved in hope that the subsequent memory
29  * preserving kernel boot is going to process this crash data.
30  */
31 void __init opal_fadump_dt_scan(struct fw_dump *fadump_conf, u64 node)
32 {
33 	const struct opal_fadump_mem_struct *opal_fdm_active;
34 	const __be32 *prop;
35 	unsigned long dn;
36 	u64 addr = 0;
37 	s64 ret;
38 
39 	dn = of_get_flat_dt_subnode_by_name(node, "dump");
40 	if (dn == -FDT_ERR_NOTFOUND)
41 		return;
42 
43 	/*
44 	 * Check if dump has been initiated on last reboot.
45 	 */
46 	prop = of_get_flat_dt_prop(dn, "mpipl-boot", NULL);
47 	if (!prop)
48 		return;
49 
50 	ret = opal_mpipl_query_tag(OPAL_MPIPL_TAG_KERNEL, &addr);
51 	if ((ret != OPAL_SUCCESS) || !addr) {
52 		pr_debug("Could not get Kernel metadata (%lld)\n", ret);
53 		return;
54 	}
55 
56 	/*
57 	 * Preserve memory only if kernel memory regions are registered
58 	 * with f/w for MPIPL.
59 	 */
60 	addr = be64_to_cpu(addr);
61 	pr_debug("Kernel metadata addr: %llx\n", addr);
62 	opal_fdm_active = (void *)addr;
63 	if (opal_fdm_active->registered_regions == 0)
64 		return;
65 
66 	ret = opal_mpipl_query_tag(OPAL_MPIPL_TAG_BOOT_MEM, &addr);
67 	if ((ret != OPAL_SUCCESS) || !addr) {
68 		pr_err("Failed to get boot memory tag (%lld)\n", ret);
69 		return;
70 	}
71 
72 	/*
73 	 * Memory below this address can be used for booting a
74 	 * capture kernel or petitboot kernel. Preserve everything
75 	 * above this address for processing crashdump.
76 	 */
77 	fadump_conf->boot_mem_top = be64_to_cpu(addr);
78 	pr_debug("Preserve everything above %llx\n", fadump_conf->boot_mem_top);
79 
80 	pr_info("Firmware-assisted dump is active.\n");
81 	fadump_conf->dump_active = 1;
82 }
83 
84 #else /* CONFIG_PRESERVE_FA_DUMP */
85 static const struct opal_fadump_mem_struct *opal_fdm_active;
86 static const struct opal_mpipl_fadump *opal_cpu_metadata;
87 static struct opal_fadump_mem_struct *opal_fdm;
88 
89 #ifdef CONFIG_OPAL_CORE
90 extern bool kernel_initiated;
91 #endif
92 
93 static int opal_fadump_unregister(struct fw_dump *fadump_conf);
94 
95 static void opal_fadump_update_config(struct fw_dump *fadump_conf,
96 				      const struct opal_fadump_mem_struct *fdm)
97 {
98 	pr_debug("Boot memory regions count: %d\n", fdm->region_cnt);
99 
100 	/*
101 	 * The destination address of the first boot memory region is the
102 	 * destination address of boot memory regions.
103 	 */
104 	fadump_conf->boot_mem_dest_addr = fdm->rgn[0].dest;
105 	pr_debug("Destination address of boot memory regions: %#016llx\n",
106 		 fadump_conf->boot_mem_dest_addr);
107 
108 	fadump_conf->fadumphdr_addr = fdm->fadumphdr_addr;
109 }
110 
111 /*
112  * This function is called in the capture kernel to get configuration details
113  * from metadata setup by the first kernel.
114  */
115 static void opal_fadump_get_config(struct fw_dump *fadump_conf,
116 				   const struct opal_fadump_mem_struct *fdm)
117 {
118 	unsigned long base, size, last_end, hole_size;
119 	int i;
120 
121 	if (!fadump_conf->dump_active)
122 		return;
123 
124 	last_end = 0;
125 	hole_size = 0;
126 	fadump_conf->boot_memory_size = 0;
127 
128 	pr_debug("Boot memory regions:\n");
129 	for (i = 0; i < fdm->region_cnt; i++) {
130 		base = fdm->rgn[i].src;
131 		size = fdm->rgn[i].size;
132 		pr_debug("\t[%03d] base: 0x%lx, size: 0x%lx\n", i, base, size);
133 
134 		fadump_conf->boot_mem_addr[i] = base;
135 		fadump_conf->boot_mem_sz[i] = size;
136 		fadump_conf->boot_memory_size += size;
137 		hole_size += (base - last_end);
138 
139 		last_end = base + size;
140 	}
141 
142 	/*
143 	 * Start address of reserve dump area (permanent reservation) for
144 	 * re-registering FADump after dump capture.
145 	 */
146 	fadump_conf->reserve_dump_area_start = fdm->rgn[0].dest;
147 
148 	/*
149 	 * Rarely, but it can so happen that system crashes before all
150 	 * boot memory regions are registered for MPIPL. In such
151 	 * cases, warn that the vmcore may not be accurate and proceed
152 	 * anyway as that is the best bet considering free pages, cache
153 	 * pages, user pages, etc are usually filtered out.
154 	 *
155 	 * Hope the memory that could not be preserved only has pages
156 	 * that are usually filtered out while saving the vmcore.
157 	 */
158 	if (fdm->region_cnt > fdm->registered_regions) {
159 		pr_warn("Not all memory regions were saved!!!\n");
160 		pr_warn("  Unsaved memory regions:\n");
161 		i = fdm->registered_regions;
162 		while (i < fdm->region_cnt) {
163 			pr_warn("\t[%03d] base: 0x%llx, size: 0x%llx\n",
164 				i, fdm->rgn[i].src, fdm->rgn[i].size);
165 			i++;
166 		}
167 
168 		pr_warn("If the unsaved regions only contain pages that are filtered out (eg. free/user pages), the vmcore should still be usable.\n");
169 		pr_warn("WARNING: If the unsaved regions contain kernel pages, the vmcore will be corrupted.\n");
170 	}
171 
172 	fadump_conf->boot_mem_top = (fadump_conf->boot_memory_size + hole_size);
173 	fadump_conf->boot_mem_regs_cnt = fdm->region_cnt;
174 	opal_fadump_update_config(fadump_conf, fdm);
175 }
176 
177 /* Initialize kernel metadata */
178 static void opal_fadump_init_metadata(struct opal_fadump_mem_struct *fdm)
179 {
180 	fdm->version = OPAL_FADUMP_VERSION;
181 	fdm->region_cnt = 0;
182 	fdm->registered_regions = 0;
183 	fdm->fadumphdr_addr = 0;
184 }
185 
186 static u64 opal_fadump_init_mem_struct(struct fw_dump *fadump_conf)
187 {
188 	u64 addr = fadump_conf->reserve_dump_area_start;
189 	int i;
190 
191 	opal_fdm = __va(fadump_conf->kernel_metadata);
192 	opal_fadump_init_metadata(opal_fdm);
193 
194 	/* Boot memory regions */
195 	for (i = 0; i < fadump_conf->boot_mem_regs_cnt; i++) {
196 		opal_fdm->rgn[i].src	= fadump_conf->boot_mem_addr[i];
197 		opal_fdm->rgn[i].dest	= addr;
198 		opal_fdm->rgn[i].size	= fadump_conf->boot_mem_sz[i];
199 
200 		opal_fdm->region_cnt++;
201 		addr += fadump_conf->boot_mem_sz[i];
202 	}
203 
204 	/*
205 	 * Kernel metadata is passed to f/w and retrieved in capture kerenl.
206 	 * So, use it to save fadump header address instead of calculating it.
207 	 */
208 	opal_fdm->fadumphdr_addr = (opal_fdm->rgn[0].dest +
209 				    fadump_conf->boot_memory_size);
210 
211 	opal_fadump_update_config(fadump_conf, opal_fdm);
212 
213 	return addr;
214 }
215 
216 static u64 opal_fadump_get_metadata_size(void)
217 {
218 	return PAGE_ALIGN(sizeof(struct opal_fadump_mem_struct));
219 }
220 
221 static int opal_fadump_setup_metadata(struct fw_dump *fadump_conf)
222 {
223 	int err = 0;
224 	s64 ret;
225 
226 	/*
227 	 * Use the last page(s) in FADump memory reservation for
228 	 * kernel metadata.
229 	 */
230 	fadump_conf->kernel_metadata = (fadump_conf->reserve_dump_area_start +
231 					fadump_conf->reserve_dump_area_size -
232 					opal_fadump_get_metadata_size());
233 	pr_info("Kernel metadata addr: %llx\n", fadump_conf->kernel_metadata);
234 
235 	/* Initialize kernel metadata before registering the address with f/w */
236 	opal_fdm = __va(fadump_conf->kernel_metadata);
237 	opal_fadump_init_metadata(opal_fdm);
238 
239 	/*
240 	 * Register metadata address with f/w. Can be retrieved in
241 	 * the capture kernel.
242 	 */
243 	ret = opal_mpipl_register_tag(OPAL_MPIPL_TAG_KERNEL,
244 				      fadump_conf->kernel_metadata);
245 	if (ret != OPAL_SUCCESS) {
246 		pr_err("Failed to set kernel metadata tag!\n");
247 		err = -EPERM;
248 	}
249 
250 	/*
251 	 * Register boot memory top address with f/w. Should be retrieved
252 	 * by a kernel that intends to preserve crash'ed kernel's memory.
253 	 */
254 	ret = opal_mpipl_register_tag(OPAL_MPIPL_TAG_BOOT_MEM,
255 				      fadump_conf->boot_mem_top);
256 	if (ret != OPAL_SUCCESS) {
257 		pr_err("Failed to set boot memory tag!\n");
258 		err = -EPERM;
259 	}
260 
261 	return err;
262 }
263 
264 static u64 opal_fadump_get_bootmem_min(void)
265 {
266 	return OPAL_FADUMP_MIN_BOOT_MEM;
267 }
268 
269 static int opal_fadump_register(struct fw_dump *fadump_conf)
270 {
271 	s64 rc = OPAL_PARAMETER;
272 	int i, err = -EIO;
273 
274 	for (i = 0; i < opal_fdm->region_cnt; i++) {
275 		rc = opal_mpipl_update(OPAL_MPIPL_ADD_RANGE,
276 				       opal_fdm->rgn[i].src,
277 				       opal_fdm->rgn[i].dest,
278 				       opal_fdm->rgn[i].size);
279 		if (rc != OPAL_SUCCESS)
280 			break;
281 
282 		opal_fdm->registered_regions++;
283 	}
284 
285 	switch (rc) {
286 	case OPAL_SUCCESS:
287 		pr_info("Registration is successful!\n");
288 		fadump_conf->dump_registered = 1;
289 		err = 0;
290 		break;
291 	case OPAL_RESOURCE:
292 		/* If MAX regions limit in f/w is hit, warn and proceed. */
293 		pr_warn("%d regions could not be registered for MPIPL as MAX limit is reached!\n",
294 			(opal_fdm->region_cnt - opal_fdm->registered_regions));
295 		fadump_conf->dump_registered = 1;
296 		err = 0;
297 		break;
298 	case OPAL_PARAMETER:
299 		pr_err("Failed to register. Parameter Error(%lld).\n", rc);
300 		break;
301 	case OPAL_HARDWARE:
302 		pr_err("Support not available.\n");
303 		fadump_conf->fadump_supported = 0;
304 		fadump_conf->fadump_enabled = 0;
305 		break;
306 	default:
307 		pr_err("Failed to register. Unknown Error(%lld).\n", rc);
308 		break;
309 	}
310 
311 	/*
312 	 * If some regions were registered before OPAL_MPIPL_ADD_RANGE
313 	 * OPAL call failed, unregister all regions.
314 	 */
315 	if ((err < 0) && (opal_fdm->registered_regions > 0))
316 		opal_fadump_unregister(fadump_conf);
317 
318 	return err;
319 }
320 
321 static int opal_fadump_unregister(struct fw_dump *fadump_conf)
322 {
323 	s64 rc;
324 
325 	rc = opal_mpipl_update(OPAL_MPIPL_REMOVE_ALL, 0, 0, 0);
326 	if (rc) {
327 		pr_err("Failed to un-register - unexpected Error(%lld).\n", rc);
328 		return -EIO;
329 	}
330 
331 	opal_fdm->registered_regions = 0;
332 	fadump_conf->dump_registered = 0;
333 	return 0;
334 }
335 
336 static int opal_fadump_invalidate(struct fw_dump *fadump_conf)
337 {
338 	s64 rc;
339 
340 	rc = opal_mpipl_update(OPAL_MPIPL_FREE_PRESERVED_MEMORY, 0, 0, 0);
341 	if (rc) {
342 		pr_err("Failed to invalidate - unexpected Error(%lld).\n", rc);
343 		return -EIO;
344 	}
345 
346 	fadump_conf->dump_active = 0;
347 	opal_fdm_active = NULL;
348 	return 0;
349 }
350 
351 static void opal_fadump_cleanup(struct fw_dump *fadump_conf)
352 {
353 	s64 ret;
354 
355 	ret = opal_mpipl_register_tag(OPAL_MPIPL_TAG_KERNEL, 0);
356 	if (ret != OPAL_SUCCESS)
357 		pr_warn("Could not reset (%llu) kernel metadata tag!\n", ret);
358 }
359 
360 /*
361  * Verify if CPU state data is available. If available, do a bit of sanity
362  * checking before processing this data.
363  */
364 static bool __init is_opal_fadump_cpu_data_valid(struct fw_dump *fadump_conf)
365 {
366 	if (!opal_cpu_metadata)
367 		return false;
368 
369 	fadump_conf->cpu_state_data_version =
370 		be32_to_cpu(opal_cpu_metadata->cpu_data_version);
371 	fadump_conf->cpu_state_entry_size =
372 		be32_to_cpu(opal_cpu_metadata->cpu_data_size);
373 	fadump_conf->cpu_state_dest_vaddr =
374 		(u64)__va(be64_to_cpu(opal_cpu_metadata->region[0].dest));
375 	fadump_conf->cpu_state_data_size =
376 		be64_to_cpu(opal_cpu_metadata->region[0].size);
377 
378 	if (fadump_conf->cpu_state_data_version != HDAT_FADUMP_CPU_DATA_VER) {
379 		pr_warn("Supported CPU state data version: %u, found: %d!\n",
380 			HDAT_FADUMP_CPU_DATA_VER,
381 			fadump_conf->cpu_state_data_version);
382 		pr_warn("WARNING: F/W using newer CPU state data format!!\n");
383 	}
384 
385 	if ((fadump_conf->cpu_state_dest_vaddr == 0) ||
386 	    (fadump_conf->cpu_state_entry_size == 0) ||
387 	    (fadump_conf->cpu_state_entry_size >
388 	     fadump_conf->cpu_state_data_size)) {
389 		pr_err("CPU state data is invalid. Ignoring!\n");
390 		return false;
391 	}
392 
393 	return true;
394 }
395 
396 /*
397  * Convert CPU state data saved at the time of crash into ELF notes.
398  *
399  * While the crashing CPU's register data is saved by the kernel, CPU state
400  * data for all CPUs is saved by f/w. In CPU state data provided by f/w,
401  * each register entry is of 16 bytes, a numerical identifier along with
402  * a GPR/SPR flag in the first 8 bytes and the register value in the next
403  * 8 bytes. For more details refer to F/W documentation. If this data is
404  * missing or in unsupported format, append crashing CPU's register data
405  * saved by the kernel in the PT_NOTE, to have something to work with in
406  * the vmcore file.
407  */
408 static int __init
409 opal_fadump_build_cpu_notes(struct fw_dump *fadump_conf,
410 			    struct fadump_crash_info_header *fdh)
411 {
412 	u32 thread_pir, size_per_thread, regs_offset, regs_cnt, reg_esize;
413 	struct hdat_fadump_thread_hdr *thdr;
414 	bool is_cpu_data_valid = false;
415 	u32 num_cpus = 1, *note_buf;
416 	struct pt_regs regs;
417 	char *bufp;
418 	int rc, i;
419 
420 	if (is_opal_fadump_cpu_data_valid(fadump_conf)) {
421 		size_per_thread = fadump_conf->cpu_state_entry_size;
422 		num_cpus = (fadump_conf->cpu_state_data_size / size_per_thread);
423 		bufp = __va(fadump_conf->cpu_state_dest_vaddr);
424 		is_cpu_data_valid = true;
425 	}
426 
427 	rc = fadump_setup_cpu_notes_buf(num_cpus);
428 	if (rc != 0)
429 		return rc;
430 
431 	note_buf = (u32 *)fadump_conf->cpu_notes_buf_vaddr;
432 	if (!is_cpu_data_valid)
433 		goto out;
434 
435 	/*
436 	 * Offset for register entries, entry size and registers count is
437 	 * duplicated in every thread header in keeping with HDAT format.
438 	 * Use these values from the first thread header.
439 	 */
440 	thdr = (struct hdat_fadump_thread_hdr *)bufp;
441 	regs_offset = (offsetof(struct hdat_fadump_thread_hdr, offset) +
442 		       be32_to_cpu(thdr->offset));
443 	reg_esize = be32_to_cpu(thdr->esize);
444 	regs_cnt  = be32_to_cpu(thdr->ecnt);
445 
446 	pr_debug("--------CPU State Data------------\n");
447 	pr_debug("NumCpus     : %u\n", num_cpus);
448 	pr_debug("\tOffset: %u, Entry size: %u, Cnt: %u\n",
449 		 regs_offset, reg_esize, regs_cnt);
450 
451 	for (i = 0; i < num_cpus; i++, bufp += size_per_thread) {
452 		thdr = (struct hdat_fadump_thread_hdr *)bufp;
453 
454 		thread_pir = be32_to_cpu(thdr->pir);
455 		pr_debug("[%04d] PIR: 0x%x, core state: 0x%02x\n",
456 			 i, thread_pir, thdr->core_state);
457 
458 		/*
459 		 * If this is kernel initiated crash, crashing_cpu would be set
460 		 * appropriately and register data of the crashing CPU saved by
461 		 * crashing kernel. Add this saved register data of crashing CPU
462 		 * to elf notes and populate the pt_regs for the remaining CPUs
463 		 * from register state data provided by firmware.
464 		 */
465 		if (fdh->crashing_cpu == thread_pir) {
466 			note_buf = fadump_regs_to_elf_notes(note_buf,
467 							    &fdh->regs);
468 			pr_debug("Crashing CPU PIR: 0x%x - R1 : 0x%lx, NIP : 0x%lx\n",
469 				 fdh->crashing_cpu, fdh->regs.gpr[1],
470 				 fdh->regs.nip);
471 			continue;
472 		}
473 
474 		/*
475 		 * Register state data of MAX cores is provided by firmware,
476 		 * but some of this cores may not be active. So, while
477 		 * processing register state data, check core state and
478 		 * skip threads that belong to inactive cores.
479 		 */
480 		if (thdr->core_state == HDAT_FADUMP_CORE_INACTIVE)
481 			continue;
482 
483 		opal_fadump_read_regs((bufp + regs_offset), regs_cnt,
484 				      reg_esize, true, &regs);
485 		note_buf = fadump_regs_to_elf_notes(note_buf, &regs);
486 		pr_debug("CPU PIR: 0x%x - R1 : 0x%lx, NIP : 0x%lx\n",
487 			 thread_pir, regs.gpr[1], regs.nip);
488 	}
489 
490 out:
491 	/*
492 	 * CPU state data is invalid/unsupported. Try appending crashing CPU's
493 	 * register data, if it is saved by the kernel.
494 	 */
495 	if (fadump_conf->cpu_notes_buf_vaddr == (u64)note_buf) {
496 		if (fdh->crashing_cpu == FADUMP_CPU_UNKNOWN) {
497 			fadump_free_cpu_notes_buf();
498 			return -ENODEV;
499 		}
500 
501 		pr_warn("WARNING: appending only crashing CPU's register data\n");
502 		note_buf = fadump_regs_to_elf_notes(note_buf, &(fdh->regs));
503 	}
504 
505 	final_note(note_buf);
506 
507 	pr_debug("Updating elfcore header (%llx) with cpu notes\n",
508 		 fdh->elfcorehdr_addr);
509 	fadump_update_elfcore_header(__va(fdh->elfcorehdr_addr));
510 	return 0;
511 }
512 
513 static int __init opal_fadump_process(struct fw_dump *fadump_conf)
514 {
515 	struct fadump_crash_info_header *fdh;
516 	int rc = -EINVAL;
517 
518 	if (!opal_fdm_active || !fadump_conf->fadumphdr_addr)
519 		return rc;
520 
521 	/* Validate the fadump crash info header */
522 	fdh = __va(fadump_conf->fadumphdr_addr);
523 	if (fdh->magic_number != FADUMP_CRASH_INFO_MAGIC) {
524 		pr_err("Crash info header is not valid.\n");
525 		return rc;
526 	}
527 
528 #ifdef CONFIG_OPAL_CORE
529 	/*
530 	 * If this is a kernel initiated crash, crashing_cpu would be set
531 	 * appropriately and register data of the crashing CPU saved by
532 	 * crashing kernel. Add this saved register data of crashing CPU
533 	 * to elf notes and populate the pt_regs for the remaining CPUs
534 	 * from register state data provided by firmware.
535 	 */
536 	if (fdh->crashing_cpu != FADUMP_CPU_UNKNOWN)
537 		kernel_initiated = true;
538 #endif
539 
540 	rc = opal_fadump_build_cpu_notes(fadump_conf, fdh);
541 	if (rc)
542 		return rc;
543 
544 	/*
545 	 * We are done validating dump info and elfcore header is now ready
546 	 * to be exported. set elfcorehdr_addr so that vmcore module will
547 	 * export the elfcore header through '/proc/vmcore'.
548 	 */
549 	elfcorehdr_addr = fdh->elfcorehdr_addr;
550 
551 	return rc;
552 }
553 
554 static void opal_fadump_region_show(struct fw_dump *fadump_conf,
555 				    struct seq_file *m)
556 {
557 	const struct opal_fadump_mem_struct *fdm_ptr;
558 	u64 dumped_bytes = 0;
559 	int i;
560 
561 	if (fadump_conf->dump_active)
562 		fdm_ptr = opal_fdm_active;
563 	else
564 		fdm_ptr = opal_fdm;
565 
566 	for (i = 0; i < fdm_ptr->region_cnt; i++) {
567 		/*
568 		 * Only regions that are registered for MPIPL
569 		 * would have dump data.
570 		 */
571 		if ((fadump_conf->dump_active) &&
572 		    (i < fdm_ptr->registered_regions))
573 			dumped_bytes = fdm_ptr->rgn[i].size;
574 
575 		seq_printf(m, "DUMP: Src: %#016llx, Dest: %#016llx, ",
576 			   fdm_ptr->rgn[i].src, fdm_ptr->rgn[i].dest);
577 		seq_printf(m, "Size: %#llx, Dumped: %#llx bytes\n",
578 			   fdm_ptr->rgn[i].size, dumped_bytes);
579 	}
580 
581 	/* Dump is active. Show reserved area start address. */
582 	if (fadump_conf->dump_active) {
583 		seq_printf(m, "\nMemory above %#016lx is reserved for saving crash dump\n",
584 			   fadump_conf->reserve_dump_area_start);
585 	}
586 }
587 
588 static void opal_fadump_trigger(struct fadump_crash_info_header *fdh,
589 				const char *msg)
590 {
591 	int rc;
592 
593 	/*
594 	 * Unlike on pSeries platform, logical CPU number is not provided
595 	 * with architected register state data. So, store the crashing
596 	 * CPU's PIR instead to plug the appropriate register data for
597 	 * crashing CPU in the vmcore file.
598 	 */
599 	fdh->crashing_cpu = (u32)mfspr(SPRN_PIR);
600 
601 	rc = opal_cec_reboot2(OPAL_REBOOT_MPIPL, msg);
602 	if (rc == OPAL_UNSUPPORTED) {
603 		pr_emerg("Reboot type %d not supported.\n",
604 			 OPAL_REBOOT_MPIPL);
605 	} else if (rc == OPAL_HARDWARE)
606 		pr_emerg("No backend support for MPIPL!\n");
607 }
608 
609 static struct fadump_ops opal_fadump_ops = {
610 	.fadump_init_mem_struct		= opal_fadump_init_mem_struct,
611 	.fadump_get_metadata_size	= opal_fadump_get_metadata_size,
612 	.fadump_setup_metadata		= opal_fadump_setup_metadata,
613 	.fadump_get_bootmem_min		= opal_fadump_get_bootmem_min,
614 	.fadump_register		= opal_fadump_register,
615 	.fadump_unregister		= opal_fadump_unregister,
616 	.fadump_invalidate		= opal_fadump_invalidate,
617 	.fadump_cleanup			= opal_fadump_cleanup,
618 	.fadump_process			= opal_fadump_process,
619 	.fadump_region_show		= opal_fadump_region_show,
620 	.fadump_trigger			= opal_fadump_trigger,
621 };
622 
623 void __init opal_fadump_dt_scan(struct fw_dump *fadump_conf, u64 node)
624 {
625 	const __be32 *prop;
626 	unsigned long dn;
627 	u64 addr = 0;
628 	int i, len;
629 	s64 ret;
630 
631 	/*
632 	 * Check if Firmware-Assisted Dump is supported. if yes, check
633 	 * if dump has been initiated on last reboot.
634 	 */
635 	dn = of_get_flat_dt_subnode_by_name(node, "dump");
636 	if (dn == -FDT_ERR_NOTFOUND) {
637 		pr_debug("FADump support is missing!\n");
638 		return;
639 	}
640 
641 	if (!of_flat_dt_is_compatible(dn, "ibm,opal-dump")) {
642 		pr_err("Support missing for this f/w version!\n");
643 		return;
644 	}
645 
646 	prop = of_get_flat_dt_prop(dn, "fw-load-area", &len);
647 	if (prop) {
648 		/*
649 		 * Each f/w load area is an (address,size) pair,
650 		 * 2 cells each, totalling 4 cells per range.
651 		 */
652 		for (i = 0; i < len / (sizeof(*prop) * 4); i++) {
653 			u64 base, end;
654 
655 			base = of_read_number(prop + (i * 4) + 0, 2);
656 			end = base;
657 			end += of_read_number(prop + (i * 4) + 2, 2);
658 			if (end > OPAL_FADUMP_MIN_BOOT_MEM) {
659 				pr_err("F/W load area: 0x%llx-0x%llx\n",
660 				       base, end);
661 				pr_err("F/W version not supported!\n");
662 				return;
663 			}
664 		}
665 	}
666 
667 	fadump_conf->ops		= &opal_fadump_ops;
668 	fadump_conf->fadump_supported	= 1;
669 
670 	/*
671 	 * Firmware supports 32-bit field for size. Align it to PAGE_SIZE
672 	 * and request firmware to copy multiple kernel boot memory regions.
673 	 */
674 	fadump_conf->max_copy_size = ALIGN_DOWN(U32_MAX, PAGE_SIZE);
675 
676 	/*
677 	 * Check if dump has been initiated on last reboot.
678 	 */
679 	prop = of_get_flat_dt_prop(dn, "mpipl-boot", NULL);
680 	if (!prop)
681 		return;
682 
683 	ret = opal_mpipl_query_tag(OPAL_MPIPL_TAG_KERNEL, &addr);
684 	if ((ret != OPAL_SUCCESS) || !addr) {
685 		pr_err("Failed to get Kernel metadata (%lld)\n", ret);
686 		return;
687 	}
688 
689 	addr = be64_to_cpu(addr);
690 	pr_debug("Kernel metadata addr: %llx\n", addr);
691 
692 	opal_fdm_active = __va(addr);
693 	if (opal_fdm_active->version != OPAL_FADUMP_VERSION) {
694 		pr_warn("Supported kernel metadata version: %u, found: %d!\n",
695 			OPAL_FADUMP_VERSION, opal_fdm_active->version);
696 		pr_warn("WARNING: Kernel metadata format mismatch identified! Core file maybe corrupted..\n");
697 	}
698 
699 	/* Kernel regions not registered with f/w for MPIPL */
700 	if (opal_fdm_active->registered_regions == 0) {
701 		opal_fdm_active = NULL;
702 		return;
703 	}
704 
705 	ret = opal_mpipl_query_tag(OPAL_MPIPL_TAG_CPU, &addr);
706 	if (addr) {
707 		addr = be64_to_cpu(addr);
708 		pr_debug("CPU metadata addr: %llx\n", addr);
709 		opal_cpu_metadata = __va(addr);
710 	}
711 
712 	pr_info("Firmware-assisted dump is active.\n");
713 	fadump_conf->dump_active = 1;
714 	opal_fadump_get_config(fadump_conf, opal_fdm_active);
715 }
716 #endif /* !CONFIG_PRESERVE_FA_DUMP */
717