xref: /linux/arch/powerpc/platforms/pseries/rtas-fadump.c (revision 26fbb4c8c7c3ee9a4c3b4de555a8587b5a19154e)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * Firmware-Assisted Dump support on POWERVM platform.
4  *
5  * Copyright 2011, Mahesh Salgaonkar, IBM Corporation.
6  * Copyright 2019, Hari Bathini, IBM Corporation.
7  */
8 
9 #define pr_fmt(fmt) "rtas fadump: " fmt
10 
11 #include <linux/string.h>
12 #include <linux/memblock.h>
13 #include <linux/delay.h>
14 #include <linux/seq_file.h>
15 #include <linux/crash_dump.h>
16 
17 #include <asm/page.h>
18 #include <asm/prom.h>
19 #include <asm/rtas.h>
20 #include <asm/fadump.h>
21 #include <asm/fadump-internal.h>
22 
23 #include "rtas-fadump.h"
24 
25 static struct rtas_fadump_mem_struct fdm;
26 static const struct rtas_fadump_mem_struct *fdm_active;
27 
28 static void rtas_fadump_update_config(struct fw_dump *fadump_conf,
29 				      const struct rtas_fadump_mem_struct *fdm)
30 {
31 	fadump_conf->boot_mem_dest_addr =
32 		be64_to_cpu(fdm->rmr_region.destination_address);
33 
34 	fadump_conf->fadumphdr_addr = (fadump_conf->boot_mem_dest_addr +
35 				       fadump_conf->boot_memory_size);
36 }
37 
38 /*
39  * This function is called in the capture kernel to get configuration details
40  * setup in the first kernel and passed to the f/w.
41  */
42 static void rtas_fadump_get_config(struct fw_dump *fadump_conf,
43 				   const struct rtas_fadump_mem_struct *fdm)
44 {
45 	fadump_conf->boot_mem_addr[0] =
46 		be64_to_cpu(fdm->rmr_region.source_address);
47 	fadump_conf->boot_mem_sz[0] = be64_to_cpu(fdm->rmr_region.source_len);
48 	fadump_conf->boot_memory_size = fadump_conf->boot_mem_sz[0];
49 
50 	fadump_conf->boot_mem_top = fadump_conf->boot_memory_size;
51 	fadump_conf->boot_mem_regs_cnt = 1;
52 
53 	/*
54 	 * Start address of reserve dump area (permanent reservation) for
55 	 * re-registering FADump after dump capture.
56 	 */
57 	fadump_conf->reserve_dump_area_start =
58 		be64_to_cpu(fdm->cpu_state_data.destination_address);
59 
60 	rtas_fadump_update_config(fadump_conf, fdm);
61 }
62 
63 static u64 rtas_fadump_init_mem_struct(struct fw_dump *fadump_conf)
64 {
65 	u64 addr = fadump_conf->reserve_dump_area_start;
66 
67 	memset(&fdm, 0, sizeof(struct rtas_fadump_mem_struct));
68 	addr = addr & PAGE_MASK;
69 
70 	fdm.header.dump_format_version = cpu_to_be32(0x00000001);
71 	fdm.header.dump_num_sections = cpu_to_be16(3);
72 	fdm.header.dump_status_flag = 0;
73 	fdm.header.offset_first_dump_section =
74 		cpu_to_be32((u32)offsetof(struct rtas_fadump_mem_struct,
75 					  cpu_state_data));
76 
77 	/*
78 	 * Fields for disk dump option.
79 	 * We are not using disk dump option, hence set these fields to 0.
80 	 */
81 	fdm.header.dd_block_size = 0;
82 	fdm.header.dd_block_offset = 0;
83 	fdm.header.dd_num_blocks = 0;
84 	fdm.header.dd_offset_disk_path = 0;
85 
86 	/* set 0 to disable an automatic dump-reboot. */
87 	fdm.header.max_time_auto = 0;
88 
89 	/* Kernel dump sections */
90 	/* cpu state data section. */
91 	fdm.cpu_state_data.request_flag =
92 		cpu_to_be32(RTAS_FADUMP_REQUEST_FLAG);
93 	fdm.cpu_state_data.source_data_type =
94 		cpu_to_be16(RTAS_FADUMP_CPU_STATE_DATA);
95 	fdm.cpu_state_data.source_address = 0;
96 	fdm.cpu_state_data.source_len =
97 		cpu_to_be64(fadump_conf->cpu_state_data_size);
98 	fdm.cpu_state_data.destination_address = cpu_to_be64(addr);
99 	addr += fadump_conf->cpu_state_data_size;
100 
101 	/* hpte region section */
102 	fdm.hpte_region.request_flag = cpu_to_be32(RTAS_FADUMP_REQUEST_FLAG);
103 	fdm.hpte_region.source_data_type =
104 		cpu_to_be16(RTAS_FADUMP_HPTE_REGION);
105 	fdm.hpte_region.source_address = 0;
106 	fdm.hpte_region.source_len =
107 		cpu_to_be64(fadump_conf->hpte_region_size);
108 	fdm.hpte_region.destination_address = cpu_to_be64(addr);
109 	addr += fadump_conf->hpte_region_size;
110 
111 	/* RMA region section */
112 	fdm.rmr_region.request_flag = cpu_to_be32(RTAS_FADUMP_REQUEST_FLAG);
113 	fdm.rmr_region.source_data_type =
114 		cpu_to_be16(RTAS_FADUMP_REAL_MODE_REGION);
115 	fdm.rmr_region.source_address = cpu_to_be64(0);
116 	fdm.rmr_region.source_len = cpu_to_be64(fadump_conf->boot_memory_size);
117 	fdm.rmr_region.destination_address = cpu_to_be64(addr);
118 	addr += fadump_conf->boot_memory_size;
119 
120 	rtas_fadump_update_config(fadump_conf, &fdm);
121 
122 	return addr;
123 }
124 
125 static u64 rtas_fadump_get_bootmem_min(void)
126 {
127 	return RTAS_FADUMP_MIN_BOOT_MEM;
128 }
129 
130 static int rtas_fadump_register(struct fw_dump *fadump_conf)
131 {
132 	unsigned int wait_time;
133 	int rc, err = -EIO;
134 
135 	/* TODO: Add upper time limit for the delay */
136 	do {
137 		rc =  rtas_call(fadump_conf->ibm_configure_kernel_dump, 3, 1,
138 				NULL, FADUMP_REGISTER, &fdm,
139 				sizeof(struct rtas_fadump_mem_struct));
140 
141 		wait_time = rtas_busy_delay_time(rc);
142 		if (wait_time)
143 			mdelay(wait_time);
144 
145 	} while (wait_time);
146 
147 	switch (rc) {
148 	case 0:
149 		pr_info("Registration is successful!\n");
150 		fadump_conf->dump_registered = 1;
151 		err = 0;
152 		break;
153 	case -1:
154 		pr_err("Failed to register. Hardware Error(%d).\n", rc);
155 		break;
156 	case -3:
157 		if (!is_fadump_boot_mem_contiguous())
158 			pr_err("Can't have holes in boot memory area.\n");
159 		else if (!is_fadump_reserved_mem_contiguous())
160 			pr_err("Can't have holes in reserved memory area.\n");
161 
162 		pr_err("Failed to register. Parameter Error(%d).\n", rc);
163 		err = -EINVAL;
164 		break;
165 	case -9:
166 		pr_err("Already registered!\n");
167 		fadump_conf->dump_registered = 1;
168 		err = -EEXIST;
169 		break;
170 	default:
171 		pr_err("Failed to register. Unknown Error(%d).\n", rc);
172 		break;
173 	}
174 
175 	return err;
176 }
177 
178 static int rtas_fadump_unregister(struct fw_dump *fadump_conf)
179 {
180 	unsigned int wait_time;
181 	int rc;
182 
183 	/* TODO: Add upper time limit for the delay */
184 	do {
185 		rc =  rtas_call(fadump_conf->ibm_configure_kernel_dump, 3, 1,
186 				NULL, FADUMP_UNREGISTER, &fdm,
187 				sizeof(struct rtas_fadump_mem_struct));
188 
189 		wait_time = rtas_busy_delay_time(rc);
190 		if (wait_time)
191 			mdelay(wait_time);
192 	} while (wait_time);
193 
194 	if (rc) {
195 		pr_err("Failed to un-register - unexpected error(%d).\n", rc);
196 		return -EIO;
197 	}
198 
199 	fadump_conf->dump_registered = 0;
200 	return 0;
201 }
202 
203 static int rtas_fadump_invalidate(struct fw_dump *fadump_conf)
204 {
205 	unsigned int wait_time;
206 	int rc;
207 
208 	/* TODO: Add upper time limit for the delay */
209 	do {
210 		rc =  rtas_call(fadump_conf->ibm_configure_kernel_dump, 3, 1,
211 				NULL, FADUMP_INVALIDATE, fdm_active,
212 				sizeof(struct rtas_fadump_mem_struct));
213 
214 		wait_time = rtas_busy_delay_time(rc);
215 		if (wait_time)
216 			mdelay(wait_time);
217 	} while (wait_time);
218 
219 	if (rc) {
220 		pr_err("Failed to invalidate - unexpected error (%d).\n", rc);
221 		return -EIO;
222 	}
223 
224 	fadump_conf->dump_active = 0;
225 	fdm_active = NULL;
226 	return 0;
227 }
228 
229 #define RTAS_FADUMP_GPR_MASK		0xffffff0000000000
230 static inline int rtas_fadump_gpr_index(u64 id)
231 {
232 	char str[3];
233 	int i = -1;
234 
235 	if ((id & RTAS_FADUMP_GPR_MASK) == fadump_str_to_u64("GPR")) {
236 		/* get the digits at the end */
237 		id &= ~RTAS_FADUMP_GPR_MASK;
238 		id >>= 24;
239 		str[2] = '\0';
240 		str[1] = id & 0xff;
241 		str[0] = (id >> 8) & 0xff;
242 		if (kstrtoint(str, 10, &i))
243 			i = -EINVAL;
244 		if (i > 31)
245 			i = -1;
246 	}
247 	return i;
248 }
249 
250 void rtas_fadump_set_regval(struct pt_regs *regs, u64 reg_id, u64 reg_val)
251 {
252 	int i;
253 
254 	i = rtas_fadump_gpr_index(reg_id);
255 	if (i >= 0)
256 		regs->gpr[i] = (unsigned long)reg_val;
257 	else if (reg_id == fadump_str_to_u64("NIA"))
258 		regs->nip = (unsigned long)reg_val;
259 	else if (reg_id == fadump_str_to_u64("MSR"))
260 		regs->msr = (unsigned long)reg_val;
261 	else if (reg_id == fadump_str_to_u64("CTR"))
262 		regs->ctr = (unsigned long)reg_val;
263 	else if (reg_id == fadump_str_to_u64("LR"))
264 		regs->link = (unsigned long)reg_val;
265 	else if (reg_id == fadump_str_to_u64("XER"))
266 		regs->xer = (unsigned long)reg_val;
267 	else if (reg_id == fadump_str_to_u64("CR"))
268 		regs->ccr = (unsigned long)reg_val;
269 	else if (reg_id == fadump_str_to_u64("DAR"))
270 		regs->dar = (unsigned long)reg_val;
271 	else if (reg_id == fadump_str_to_u64("DSISR"))
272 		regs->dsisr = (unsigned long)reg_val;
273 }
274 
275 static struct rtas_fadump_reg_entry*
276 rtas_fadump_read_regs(struct rtas_fadump_reg_entry *reg_entry,
277 		      struct pt_regs *regs)
278 {
279 	memset(regs, 0, sizeof(struct pt_regs));
280 
281 	while (be64_to_cpu(reg_entry->reg_id) != fadump_str_to_u64("CPUEND")) {
282 		rtas_fadump_set_regval(regs, be64_to_cpu(reg_entry->reg_id),
283 				       be64_to_cpu(reg_entry->reg_value));
284 		reg_entry++;
285 	}
286 	reg_entry++;
287 	return reg_entry;
288 }
289 
290 /*
291  * Read CPU state dump data and convert it into ELF notes.
292  * The CPU dump starts with magic number "REGSAVE". NumCpusOffset should be
293  * used to access the data to allow for additional fields to be added without
294  * affecting compatibility. Each list of registers for a CPU starts with
295  * "CPUSTRT" and ends with "CPUEND". Each register entry is of 16 bytes,
296  * 8 Byte ASCII identifier and 8 Byte register value. The register entry
297  * with identifier "CPUSTRT" and "CPUEND" contains 4 byte cpu id as part
298  * of register value. For more details refer to PAPR document.
299  *
300  * Only for the crashing cpu we ignore the CPU dump data and get exact
301  * state from fadump crash info structure populated by first kernel at the
302  * time of crash.
303  */
304 static int __init rtas_fadump_build_cpu_notes(struct fw_dump *fadump_conf)
305 {
306 	struct rtas_fadump_reg_save_area_header *reg_header;
307 	struct fadump_crash_info_header *fdh = NULL;
308 	struct rtas_fadump_reg_entry *reg_entry;
309 	u32 num_cpus, *note_buf;
310 	int i, rc = 0, cpu = 0;
311 	struct pt_regs regs;
312 	unsigned long addr;
313 	void *vaddr;
314 
315 	addr = be64_to_cpu(fdm_active->cpu_state_data.destination_address);
316 	vaddr = __va(addr);
317 
318 	reg_header = vaddr;
319 	if (be64_to_cpu(reg_header->magic_number) !=
320 	    fadump_str_to_u64("REGSAVE")) {
321 		pr_err("Unable to read register save area.\n");
322 		return -ENOENT;
323 	}
324 
325 	pr_debug("--------CPU State Data------------\n");
326 	pr_debug("Magic Number: %llx\n", be64_to_cpu(reg_header->magic_number));
327 	pr_debug("NumCpuOffset: %x\n", be32_to_cpu(reg_header->num_cpu_offset));
328 
329 	vaddr += be32_to_cpu(reg_header->num_cpu_offset);
330 	num_cpus = be32_to_cpu(*((__be32 *)(vaddr)));
331 	pr_debug("NumCpus     : %u\n", num_cpus);
332 	vaddr += sizeof(u32);
333 	reg_entry = (struct rtas_fadump_reg_entry *)vaddr;
334 
335 	rc = fadump_setup_cpu_notes_buf(num_cpus);
336 	if (rc != 0)
337 		return rc;
338 
339 	note_buf = (u32 *)fadump_conf->cpu_notes_buf_vaddr;
340 
341 	if (fadump_conf->fadumphdr_addr)
342 		fdh = __va(fadump_conf->fadumphdr_addr);
343 
344 	for (i = 0; i < num_cpus; i++) {
345 		if (be64_to_cpu(reg_entry->reg_id) !=
346 		    fadump_str_to_u64("CPUSTRT")) {
347 			pr_err("Unable to read CPU state data\n");
348 			rc = -ENOENT;
349 			goto error_out;
350 		}
351 		/* Lower 4 bytes of reg_value contains logical cpu id */
352 		cpu = (be64_to_cpu(reg_entry->reg_value) &
353 		       RTAS_FADUMP_CPU_ID_MASK);
354 		if (fdh && !cpumask_test_cpu(cpu, &fdh->online_mask)) {
355 			RTAS_FADUMP_SKIP_TO_NEXT_CPU(reg_entry);
356 			continue;
357 		}
358 		pr_debug("Reading register data for cpu %d...\n", cpu);
359 		if (fdh && fdh->crashing_cpu == cpu) {
360 			regs = fdh->regs;
361 			note_buf = fadump_regs_to_elf_notes(note_buf, &regs);
362 			RTAS_FADUMP_SKIP_TO_NEXT_CPU(reg_entry);
363 		} else {
364 			reg_entry++;
365 			reg_entry = rtas_fadump_read_regs(reg_entry, &regs);
366 			note_buf = fadump_regs_to_elf_notes(note_buf, &regs);
367 		}
368 	}
369 	final_note(note_buf);
370 
371 	if (fdh) {
372 		pr_debug("Updating elfcore header (%llx) with cpu notes\n",
373 			 fdh->elfcorehdr_addr);
374 		fadump_update_elfcore_header(__va(fdh->elfcorehdr_addr));
375 	}
376 	return 0;
377 
378 error_out:
379 	fadump_free_cpu_notes_buf();
380 	return rc;
381 
382 }
383 
384 /*
385  * Validate and process the dump data stored by firmware before exporting
386  * it through '/proc/vmcore'.
387  */
388 static int __init rtas_fadump_process(struct fw_dump *fadump_conf)
389 {
390 	struct fadump_crash_info_header *fdh;
391 	int rc = 0;
392 
393 	if (!fdm_active || !fadump_conf->fadumphdr_addr)
394 		return -EINVAL;
395 
396 	/* Check if the dump data is valid. */
397 	if ((be16_to_cpu(fdm_active->header.dump_status_flag) ==
398 			RTAS_FADUMP_ERROR_FLAG) ||
399 			(fdm_active->cpu_state_data.error_flags != 0) ||
400 			(fdm_active->rmr_region.error_flags != 0)) {
401 		pr_err("Dump taken by platform is not valid\n");
402 		return -EINVAL;
403 	}
404 	if ((fdm_active->rmr_region.bytes_dumped !=
405 			fdm_active->rmr_region.source_len) ||
406 			!fdm_active->cpu_state_data.bytes_dumped) {
407 		pr_err("Dump taken by platform is incomplete\n");
408 		return -EINVAL;
409 	}
410 
411 	/* Validate the fadump crash info header */
412 	fdh = __va(fadump_conf->fadumphdr_addr);
413 	if (fdh->magic_number != FADUMP_CRASH_INFO_MAGIC) {
414 		pr_err("Crash info header is not valid.\n");
415 		return -EINVAL;
416 	}
417 
418 	rc = rtas_fadump_build_cpu_notes(fadump_conf);
419 	if (rc)
420 		return rc;
421 
422 	/*
423 	 * We are done validating dump info and elfcore header is now ready
424 	 * to be exported. set elfcorehdr_addr so that vmcore module will
425 	 * export the elfcore header through '/proc/vmcore'.
426 	 */
427 	elfcorehdr_addr = fdh->elfcorehdr_addr;
428 
429 	return 0;
430 }
431 
432 static void rtas_fadump_region_show(struct fw_dump *fadump_conf,
433 				    struct seq_file *m)
434 {
435 	const struct rtas_fadump_section *cpu_data_section;
436 	const struct rtas_fadump_mem_struct *fdm_ptr;
437 
438 	if (fdm_active)
439 		fdm_ptr = fdm_active;
440 	else
441 		fdm_ptr = &fdm;
442 
443 	cpu_data_section = &(fdm_ptr->cpu_state_data);
444 	seq_printf(m, "CPU :[%#016llx-%#016llx] %#llx bytes, Dumped: %#llx\n",
445 		   be64_to_cpu(cpu_data_section->destination_address),
446 		   be64_to_cpu(cpu_data_section->destination_address) +
447 		   be64_to_cpu(cpu_data_section->source_len) - 1,
448 		   be64_to_cpu(cpu_data_section->source_len),
449 		   be64_to_cpu(cpu_data_section->bytes_dumped));
450 
451 	seq_printf(m, "HPTE:[%#016llx-%#016llx] %#llx bytes, Dumped: %#llx\n",
452 		   be64_to_cpu(fdm_ptr->hpte_region.destination_address),
453 		   be64_to_cpu(fdm_ptr->hpte_region.destination_address) +
454 		   be64_to_cpu(fdm_ptr->hpte_region.source_len) - 1,
455 		   be64_to_cpu(fdm_ptr->hpte_region.source_len),
456 		   be64_to_cpu(fdm_ptr->hpte_region.bytes_dumped));
457 
458 	seq_printf(m, "DUMP: Src: %#016llx, Dest: %#016llx, ",
459 		   be64_to_cpu(fdm_ptr->rmr_region.source_address),
460 		   be64_to_cpu(fdm_ptr->rmr_region.destination_address));
461 	seq_printf(m, "Size: %#llx, Dumped: %#llx bytes\n",
462 		   be64_to_cpu(fdm_ptr->rmr_region.source_len),
463 		   be64_to_cpu(fdm_ptr->rmr_region.bytes_dumped));
464 
465 	/* Dump is active. Show reserved area start address. */
466 	if (fdm_active) {
467 		seq_printf(m, "\nMemory above %#016lx is reserved for saving crash dump\n",
468 			   fadump_conf->reserve_dump_area_start);
469 	}
470 }
471 
472 static void rtas_fadump_trigger(struct fadump_crash_info_header *fdh,
473 				const char *msg)
474 {
475 	/* Call ibm,os-term rtas call to trigger firmware assisted dump */
476 	rtas_os_term((char *)msg);
477 }
478 
479 static struct fadump_ops rtas_fadump_ops = {
480 	.fadump_init_mem_struct		= rtas_fadump_init_mem_struct,
481 	.fadump_get_bootmem_min		= rtas_fadump_get_bootmem_min,
482 	.fadump_register		= rtas_fadump_register,
483 	.fadump_unregister		= rtas_fadump_unregister,
484 	.fadump_invalidate		= rtas_fadump_invalidate,
485 	.fadump_process			= rtas_fadump_process,
486 	.fadump_region_show		= rtas_fadump_region_show,
487 	.fadump_trigger			= rtas_fadump_trigger,
488 };
489 
490 void __init rtas_fadump_dt_scan(struct fw_dump *fadump_conf, u64 node)
491 {
492 	int i, size, num_sections;
493 	const __be32 *sections;
494 	const __be32 *token;
495 
496 	/*
497 	 * Check if Firmware Assisted dump is supported. if yes, check
498 	 * if dump has been initiated on last reboot.
499 	 */
500 	token = of_get_flat_dt_prop(node, "ibm,configure-kernel-dump", NULL);
501 	if (!token)
502 		return;
503 
504 	fadump_conf->ibm_configure_kernel_dump = be32_to_cpu(*token);
505 	fadump_conf->ops		= &rtas_fadump_ops;
506 	fadump_conf->fadump_supported	= 1;
507 
508 	/* Firmware supports 64-bit value for size, align it to pagesize. */
509 	fadump_conf->max_copy_size = ALIGN_DOWN(U64_MAX, PAGE_SIZE);
510 
511 	/*
512 	 * The 'ibm,kernel-dump' rtas node is present only if there is
513 	 * dump data waiting for us.
514 	 */
515 	fdm_active = of_get_flat_dt_prop(node, "ibm,kernel-dump", NULL);
516 	if (fdm_active) {
517 		pr_info("Firmware-assisted dump is active.\n");
518 		fadump_conf->dump_active = 1;
519 		rtas_fadump_get_config(fadump_conf, (void *)__pa(fdm_active));
520 	}
521 
522 	/* Get the sizes required to store dump data for the firmware provided
523 	 * dump sections.
524 	 * For each dump section type supported, a 32bit cell which defines
525 	 * the ID of a supported section followed by two 32 bit cells which
526 	 * gives the size of the section in bytes.
527 	 */
528 	sections = of_get_flat_dt_prop(node, "ibm,configure-kernel-dump-sizes",
529 					&size);
530 
531 	if (!sections)
532 		return;
533 
534 	num_sections = size / (3 * sizeof(u32));
535 
536 	for (i = 0; i < num_sections; i++, sections += 3) {
537 		u32 type = (u32)of_read_number(sections, 1);
538 
539 		switch (type) {
540 		case RTAS_FADUMP_CPU_STATE_DATA:
541 			fadump_conf->cpu_state_data_size =
542 					of_read_ulong(&sections[1], 2);
543 			break;
544 		case RTAS_FADUMP_HPTE_REGION:
545 			fadump_conf->hpte_region_size =
546 					of_read_ulong(&sections[1], 2);
547 			break;
548 		}
549 	}
550 }
551