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 #include <linux/of.h>
17 #include <linux/of_fdt.h>
18 
19 #include <asm/page.h>
20 #include <asm/rtas.h>
21 #include <asm/fadump.h>
22 #include <asm/fadump-internal.h>
23 
24 #include "rtas-fadump.h"
25 
26 static struct rtas_fadump_mem_struct fdm;
27 static const struct rtas_fadump_mem_struct *fdm_active;
28 
29 static void rtas_fadump_update_config(struct fw_dump *fadump_conf,
30 				      const struct rtas_fadump_mem_struct *fdm)
31 {
32 	fadump_conf->boot_mem_dest_addr =
33 		be64_to_cpu(fdm->rmr_region.destination_address);
34 
35 	fadump_conf->fadumphdr_addr = (fadump_conf->boot_mem_dest_addr +
36 				       fadump_conf->boot_memory_size);
37 }
38 
39 /*
40  * This function is called in the capture kernel to get configuration details
41  * setup in the first kernel and passed to the f/w.
42  */
43 static void __init rtas_fadump_get_config(struct fw_dump *fadump_conf,
44 				   const struct rtas_fadump_mem_struct *fdm)
45 {
46 	fadump_conf->boot_mem_addr[0] =
47 		be64_to_cpu(fdm->rmr_region.source_address);
48 	fadump_conf->boot_mem_sz[0] = be64_to_cpu(fdm->rmr_region.source_len);
49 	fadump_conf->boot_memory_size = fadump_conf->boot_mem_sz[0];
50 
51 	fadump_conf->boot_mem_top = fadump_conf->boot_memory_size;
52 	fadump_conf->boot_mem_regs_cnt = 1;
53 
54 	/*
55 	 * Start address of reserve dump area (permanent reservation) for
56 	 * re-registering FADump after dump capture.
57 	 */
58 	fadump_conf->reserve_dump_area_start =
59 		be64_to_cpu(fdm->cpu_state_data.destination_address);
60 
61 	rtas_fadump_update_config(fadump_conf, fdm);
62 }
63 
64 static u64 rtas_fadump_init_mem_struct(struct fw_dump *fadump_conf)
65 {
66 	u64 addr = fadump_conf->reserve_dump_area_start;
67 
68 	memset(&fdm, 0, sizeof(struct rtas_fadump_mem_struct));
69 	addr = addr & PAGE_MASK;
70 
71 	fdm.header.dump_format_version = cpu_to_be32(0x00000001);
72 	fdm.header.dump_num_sections = cpu_to_be16(3);
73 	fdm.header.dump_status_flag = 0;
74 	fdm.header.offset_first_dump_section =
75 		cpu_to_be32((u32)offsetof(struct rtas_fadump_mem_struct,
76 					  cpu_state_data));
77 
78 	/*
79 	 * Fields for disk dump option.
80 	 * We are not using disk dump option, hence set these fields to 0.
81 	 */
82 	fdm.header.dd_block_size = 0;
83 	fdm.header.dd_block_offset = 0;
84 	fdm.header.dd_num_blocks = 0;
85 	fdm.header.dd_offset_disk_path = 0;
86 
87 	/* set 0 to disable an automatic dump-reboot. */
88 	fdm.header.max_time_auto = 0;
89 
90 	/* Kernel dump sections */
91 	/* cpu state data section. */
92 	fdm.cpu_state_data.request_flag =
93 		cpu_to_be32(RTAS_FADUMP_REQUEST_FLAG);
94 	fdm.cpu_state_data.source_data_type =
95 		cpu_to_be16(RTAS_FADUMP_CPU_STATE_DATA);
96 	fdm.cpu_state_data.source_address = 0;
97 	fdm.cpu_state_data.source_len =
98 		cpu_to_be64(fadump_conf->cpu_state_data_size);
99 	fdm.cpu_state_data.destination_address = cpu_to_be64(addr);
100 	addr += fadump_conf->cpu_state_data_size;
101 
102 	/* hpte region section */
103 	fdm.hpte_region.request_flag = cpu_to_be32(RTAS_FADUMP_REQUEST_FLAG);
104 	fdm.hpte_region.source_data_type =
105 		cpu_to_be16(RTAS_FADUMP_HPTE_REGION);
106 	fdm.hpte_region.source_address = 0;
107 	fdm.hpte_region.source_len =
108 		cpu_to_be64(fadump_conf->hpte_region_size);
109 	fdm.hpte_region.destination_address = cpu_to_be64(addr);
110 	addr += fadump_conf->hpte_region_size;
111 
112 	/*
113 	 * Align boot memory area destination address to page boundary to
114 	 * be able to mmap read this area in the vmcore.
115 	 */
116 	addr = PAGE_ALIGN(addr);
117 
118 	/* RMA region section */
119 	fdm.rmr_region.request_flag = cpu_to_be32(RTAS_FADUMP_REQUEST_FLAG);
120 	fdm.rmr_region.source_data_type =
121 		cpu_to_be16(RTAS_FADUMP_REAL_MODE_REGION);
122 	fdm.rmr_region.source_address = cpu_to_be64(0);
123 	fdm.rmr_region.source_len = cpu_to_be64(fadump_conf->boot_memory_size);
124 	fdm.rmr_region.destination_address = cpu_to_be64(addr);
125 	addr += fadump_conf->boot_memory_size;
126 
127 	rtas_fadump_update_config(fadump_conf, &fdm);
128 
129 	return addr;
130 }
131 
132 static u64 rtas_fadump_get_bootmem_min(void)
133 {
134 	return RTAS_FADUMP_MIN_BOOT_MEM;
135 }
136 
137 static int rtas_fadump_register(struct fw_dump *fadump_conf)
138 {
139 	unsigned int wait_time;
140 	int rc, err = -EIO;
141 
142 	/* TODO: Add upper time limit for the delay */
143 	do {
144 		rc =  rtas_call(fadump_conf->ibm_configure_kernel_dump, 3, 1,
145 				NULL, FADUMP_REGISTER, &fdm,
146 				sizeof(struct rtas_fadump_mem_struct));
147 
148 		wait_time = rtas_busy_delay_time(rc);
149 		if (wait_time)
150 			mdelay(wait_time);
151 
152 	} while (wait_time);
153 
154 	switch (rc) {
155 	case 0:
156 		pr_info("Registration is successful!\n");
157 		fadump_conf->dump_registered = 1;
158 		err = 0;
159 		break;
160 	case -1:
161 		pr_err("Failed to register. Hardware Error(%d).\n", rc);
162 		break;
163 	case -3:
164 		if (!is_fadump_boot_mem_contiguous())
165 			pr_err("Can't have holes in boot memory area.\n");
166 		else if (!is_fadump_reserved_mem_contiguous())
167 			pr_err("Can't have holes in reserved memory area.\n");
168 
169 		pr_err("Failed to register. Parameter Error(%d).\n", rc);
170 		err = -EINVAL;
171 		break;
172 	case -9:
173 		pr_err("Already registered!\n");
174 		fadump_conf->dump_registered = 1;
175 		err = -EEXIST;
176 		break;
177 	default:
178 		pr_err("Failed to register. Unknown Error(%d).\n", rc);
179 		break;
180 	}
181 
182 	return err;
183 }
184 
185 static int rtas_fadump_unregister(struct fw_dump *fadump_conf)
186 {
187 	unsigned int wait_time;
188 	int rc;
189 
190 	/* TODO: Add upper time limit for the delay */
191 	do {
192 		rc =  rtas_call(fadump_conf->ibm_configure_kernel_dump, 3, 1,
193 				NULL, FADUMP_UNREGISTER, &fdm,
194 				sizeof(struct rtas_fadump_mem_struct));
195 
196 		wait_time = rtas_busy_delay_time(rc);
197 		if (wait_time)
198 			mdelay(wait_time);
199 	} while (wait_time);
200 
201 	if (rc) {
202 		pr_err("Failed to un-register - unexpected error(%d).\n", rc);
203 		return -EIO;
204 	}
205 
206 	fadump_conf->dump_registered = 0;
207 	return 0;
208 }
209 
210 static int rtas_fadump_invalidate(struct fw_dump *fadump_conf)
211 {
212 	unsigned int wait_time;
213 	int rc;
214 
215 	/* TODO: Add upper time limit for the delay */
216 	do {
217 		rc =  rtas_call(fadump_conf->ibm_configure_kernel_dump, 3, 1,
218 				NULL, FADUMP_INVALIDATE, fdm_active,
219 				sizeof(struct rtas_fadump_mem_struct));
220 
221 		wait_time = rtas_busy_delay_time(rc);
222 		if (wait_time)
223 			mdelay(wait_time);
224 	} while (wait_time);
225 
226 	if (rc) {
227 		pr_err("Failed to invalidate - unexpected error (%d).\n", rc);
228 		return -EIO;
229 	}
230 
231 	fadump_conf->dump_active = 0;
232 	fdm_active = NULL;
233 	return 0;
234 }
235 
236 #define RTAS_FADUMP_GPR_MASK		0xffffff0000000000
237 static inline int rtas_fadump_gpr_index(u64 id)
238 {
239 	char str[3];
240 	int i = -1;
241 
242 	if ((id & RTAS_FADUMP_GPR_MASK) == fadump_str_to_u64("GPR")) {
243 		/* get the digits at the end */
244 		id &= ~RTAS_FADUMP_GPR_MASK;
245 		id >>= 24;
246 		str[2] = '\0';
247 		str[1] = id & 0xff;
248 		str[0] = (id >> 8) & 0xff;
249 		if (kstrtoint(str, 10, &i))
250 			i = -EINVAL;
251 		if (i > 31)
252 			i = -1;
253 	}
254 	return i;
255 }
256 
257 static void __init rtas_fadump_set_regval(struct pt_regs *regs, u64 reg_id, u64 reg_val)
258 {
259 	int i;
260 
261 	i = rtas_fadump_gpr_index(reg_id);
262 	if (i >= 0)
263 		regs->gpr[i] = (unsigned long)reg_val;
264 	else if (reg_id == fadump_str_to_u64("NIA"))
265 		regs->nip = (unsigned long)reg_val;
266 	else if (reg_id == fadump_str_to_u64("MSR"))
267 		regs->msr = (unsigned long)reg_val;
268 	else if (reg_id == fadump_str_to_u64("CTR"))
269 		regs->ctr = (unsigned long)reg_val;
270 	else if (reg_id == fadump_str_to_u64("LR"))
271 		regs->link = (unsigned long)reg_val;
272 	else if (reg_id == fadump_str_to_u64("XER"))
273 		regs->xer = (unsigned long)reg_val;
274 	else if (reg_id == fadump_str_to_u64("CR"))
275 		regs->ccr = (unsigned long)reg_val;
276 	else if (reg_id == fadump_str_to_u64("DAR"))
277 		regs->dar = (unsigned long)reg_val;
278 	else if (reg_id == fadump_str_to_u64("DSISR"))
279 		regs->dsisr = (unsigned long)reg_val;
280 }
281 
282 static struct rtas_fadump_reg_entry* __init
283 rtas_fadump_read_regs(struct rtas_fadump_reg_entry *reg_entry,
284 		      struct pt_regs *regs)
285 {
286 	memset(regs, 0, sizeof(struct pt_regs));
287 
288 	while (be64_to_cpu(reg_entry->reg_id) != fadump_str_to_u64("CPUEND")) {
289 		rtas_fadump_set_regval(regs, be64_to_cpu(reg_entry->reg_id),
290 				       be64_to_cpu(reg_entry->reg_value));
291 		reg_entry++;
292 	}
293 	reg_entry++;
294 	return reg_entry;
295 }
296 
297 /*
298  * Read CPU state dump data and convert it into ELF notes.
299  * The CPU dump starts with magic number "REGSAVE". NumCpusOffset should be
300  * used to access the data to allow for additional fields to be added without
301  * affecting compatibility. Each list of registers for a CPU starts with
302  * "CPUSTRT" and ends with "CPUEND". Each register entry is of 16 bytes,
303  * 8 Byte ASCII identifier and 8 Byte register value. The register entry
304  * with identifier "CPUSTRT" and "CPUEND" contains 4 byte cpu id as part
305  * of register value. For more details refer to PAPR document.
306  *
307  * Only for the crashing cpu we ignore the CPU dump data and get exact
308  * state from fadump crash info structure populated by first kernel at the
309  * time of crash.
310  */
311 static int __init rtas_fadump_build_cpu_notes(struct fw_dump *fadump_conf)
312 {
313 	struct rtas_fadump_reg_save_area_header *reg_header;
314 	struct fadump_crash_info_header *fdh = NULL;
315 	struct rtas_fadump_reg_entry *reg_entry;
316 	u32 num_cpus, *note_buf;
317 	int i, rc = 0, cpu = 0;
318 	struct pt_regs regs;
319 	unsigned long addr;
320 	void *vaddr;
321 
322 	addr = be64_to_cpu(fdm_active->cpu_state_data.destination_address);
323 	vaddr = __va(addr);
324 
325 	reg_header = vaddr;
326 	if (be64_to_cpu(reg_header->magic_number) !=
327 	    fadump_str_to_u64("REGSAVE")) {
328 		pr_err("Unable to read register save area.\n");
329 		return -ENOENT;
330 	}
331 
332 	pr_debug("--------CPU State Data------------\n");
333 	pr_debug("Magic Number: %llx\n", be64_to_cpu(reg_header->magic_number));
334 	pr_debug("NumCpuOffset: %x\n", be32_to_cpu(reg_header->num_cpu_offset));
335 
336 	vaddr += be32_to_cpu(reg_header->num_cpu_offset);
337 	num_cpus = be32_to_cpu(*((__be32 *)(vaddr)));
338 	pr_debug("NumCpus     : %u\n", num_cpus);
339 	vaddr += sizeof(u32);
340 	reg_entry = (struct rtas_fadump_reg_entry *)vaddr;
341 
342 	rc = fadump_setup_cpu_notes_buf(num_cpus);
343 	if (rc != 0)
344 		return rc;
345 
346 	note_buf = (u32 *)fadump_conf->cpu_notes_buf_vaddr;
347 
348 	if (fadump_conf->fadumphdr_addr)
349 		fdh = __va(fadump_conf->fadumphdr_addr);
350 
351 	for (i = 0; i < num_cpus; i++) {
352 		if (be64_to_cpu(reg_entry->reg_id) !=
353 		    fadump_str_to_u64("CPUSTRT")) {
354 			pr_err("Unable to read CPU state data\n");
355 			rc = -ENOENT;
356 			goto error_out;
357 		}
358 		/* Lower 4 bytes of reg_value contains logical cpu id */
359 		cpu = (be64_to_cpu(reg_entry->reg_value) &
360 		       RTAS_FADUMP_CPU_ID_MASK);
361 		if (fdh && !cpumask_test_cpu(cpu, &fdh->cpu_mask)) {
362 			RTAS_FADUMP_SKIP_TO_NEXT_CPU(reg_entry);
363 			continue;
364 		}
365 		pr_debug("Reading register data for cpu %d...\n", cpu);
366 		if (fdh && fdh->crashing_cpu == cpu) {
367 			regs = fdh->regs;
368 			note_buf = fadump_regs_to_elf_notes(note_buf, &regs);
369 			RTAS_FADUMP_SKIP_TO_NEXT_CPU(reg_entry);
370 		} else {
371 			reg_entry++;
372 			reg_entry = rtas_fadump_read_regs(reg_entry, &regs);
373 			note_buf = fadump_regs_to_elf_notes(note_buf, &regs);
374 		}
375 	}
376 	final_note(note_buf);
377 
378 	if (fdh) {
379 		pr_debug("Updating elfcore header (%llx) with cpu notes\n",
380 			 fdh->elfcorehdr_addr);
381 		fadump_update_elfcore_header(__va(fdh->elfcorehdr_addr));
382 	}
383 	return 0;
384 
385 error_out:
386 	fadump_free_cpu_notes_buf();
387 	return rc;
388 
389 }
390 
391 /*
392  * Validate and process the dump data stored by firmware before exporting
393  * it through '/proc/vmcore'.
394  */
395 static int __init rtas_fadump_process(struct fw_dump *fadump_conf)
396 {
397 	struct fadump_crash_info_header *fdh;
398 	int rc = 0;
399 
400 	if (!fdm_active || !fadump_conf->fadumphdr_addr)
401 		return -EINVAL;
402 
403 	/* Check if the dump data is valid. */
404 	if ((be16_to_cpu(fdm_active->header.dump_status_flag) ==
405 			RTAS_FADUMP_ERROR_FLAG) ||
406 			(fdm_active->cpu_state_data.error_flags != 0) ||
407 			(fdm_active->rmr_region.error_flags != 0)) {
408 		pr_err("Dump taken by platform is not valid\n");
409 		return -EINVAL;
410 	}
411 	if ((fdm_active->rmr_region.bytes_dumped !=
412 			fdm_active->rmr_region.source_len) ||
413 			!fdm_active->cpu_state_data.bytes_dumped) {
414 		pr_err("Dump taken by platform is incomplete\n");
415 		return -EINVAL;
416 	}
417 
418 	/* Validate the fadump crash info header */
419 	fdh = __va(fadump_conf->fadumphdr_addr);
420 	if (fdh->magic_number != FADUMP_CRASH_INFO_MAGIC) {
421 		pr_err("Crash info header is not valid.\n");
422 		return -EINVAL;
423 	}
424 
425 	rc = rtas_fadump_build_cpu_notes(fadump_conf);
426 	if (rc)
427 		return rc;
428 
429 	/*
430 	 * We are done validating dump info and elfcore header is now ready
431 	 * to be exported. set elfcorehdr_addr so that vmcore module will
432 	 * export the elfcore header through '/proc/vmcore'.
433 	 */
434 	elfcorehdr_addr = fdh->elfcorehdr_addr;
435 
436 	return 0;
437 }
438 
439 static void rtas_fadump_region_show(struct fw_dump *fadump_conf,
440 				    struct seq_file *m)
441 {
442 	const struct rtas_fadump_section *cpu_data_section;
443 	const struct rtas_fadump_mem_struct *fdm_ptr;
444 
445 	if (fdm_active)
446 		fdm_ptr = fdm_active;
447 	else
448 		fdm_ptr = &fdm;
449 
450 	cpu_data_section = &(fdm_ptr->cpu_state_data);
451 	seq_printf(m, "CPU :[%#016llx-%#016llx] %#llx bytes, Dumped: %#llx\n",
452 		   be64_to_cpu(cpu_data_section->destination_address),
453 		   be64_to_cpu(cpu_data_section->destination_address) +
454 		   be64_to_cpu(cpu_data_section->source_len) - 1,
455 		   be64_to_cpu(cpu_data_section->source_len),
456 		   be64_to_cpu(cpu_data_section->bytes_dumped));
457 
458 	seq_printf(m, "HPTE:[%#016llx-%#016llx] %#llx bytes, Dumped: %#llx\n",
459 		   be64_to_cpu(fdm_ptr->hpte_region.destination_address),
460 		   be64_to_cpu(fdm_ptr->hpte_region.destination_address) +
461 		   be64_to_cpu(fdm_ptr->hpte_region.source_len) - 1,
462 		   be64_to_cpu(fdm_ptr->hpte_region.source_len),
463 		   be64_to_cpu(fdm_ptr->hpte_region.bytes_dumped));
464 
465 	seq_printf(m, "DUMP: Src: %#016llx, Dest: %#016llx, ",
466 		   be64_to_cpu(fdm_ptr->rmr_region.source_address),
467 		   be64_to_cpu(fdm_ptr->rmr_region.destination_address));
468 	seq_printf(m, "Size: %#llx, Dumped: %#llx bytes\n",
469 		   be64_to_cpu(fdm_ptr->rmr_region.source_len),
470 		   be64_to_cpu(fdm_ptr->rmr_region.bytes_dumped));
471 
472 	/* Dump is active. Show preserved area start address. */
473 	if (fdm_active) {
474 		seq_printf(m, "\nMemory above %#016llx is reserved for saving crash dump\n",
475 			   fadump_conf->boot_mem_top);
476 	}
477 }
478 
479 static void rtas_fadump_trigger(struct fadump_crash_info_header *fdh,
480 				const char *msg)
481 {
482 	/* Call ibm,os-term rtas call to trigger firmware assisted dump */
483 	rtas_os_term((char *)msg);
484 }
485 
486 static struct fadump_ops rtas_fadump_ops = {
487 	.fadump_init_mem_struct		= rtas_fadump_init_mem_struct,
488 	.fadump_get_bootmem_min		= rtas_fadump_get_bootmem_min,
489 	.fadump_register		= rtas_fadump_register,
490 	.fadump_unregister		= rtas_fadump_unregister,
491 	.fadump_invalidate		= rtas_fadump_invalidate,
492 	.fadump_process			= rtas_fadump_process,
493 	.fadump_region_show		= rtas_fadump_region_show,
494 	.fadump_trigger			= rtas_fadump_trigger,
495 };
496 
497 void __init rtas_fadump_dt_scan(struct fw_dump *fadump_conf, u64 node)
498 {
499 	int i, size, num_sections;
500 	const __be32 *sections;
501 	const __be32 *token;
502 
503 	/*
504 	 * Check if Firmware Assisted dump is supported. if yes, check
505 	 * if dump has been initiated on last reboot.
506 	 */
507 	token = of_get_flat_dt_prop(node, "ibm,configure-kernel-dump", NULL);
508 	if (!token)
509 		return;
510 
511 	fadump_conf->ibm_configure_kernel_dump = be32_to_cpu(*token);
512 	fadump_conf->ops		= &rtas_fadump_ops;
513 	fadump_conf->fadump_supported	= 1;
514 
515 	/* Firmware supports 64-bit value for size, align it to pagesize. */
516 	fadump_conf->max_copy_size = ALIGN_DOWN(U64_MAX, PAGE_SIZE);
517 
518 	/*
519 	 * The 'ibm,kernel-dump' rtas node is present only if there is
520 	 * dump data waiting for us.
521 	 */
522 	fdm_active = of_get_flat_dt_prop(node, "ibm,kernel-dump", NULL);
523 	if (fdm_active) {
524 		pr_info("Firmware-assisted dump is active.\n");
525 		fadump_conf->dump_active = 1;
526 		rtas_fadump_get_config(fadump_conf, (void *)__pa(fdm_active));
527 	}
528 
529 	/* Get the sizes required to store dump data for the firmware provided
530 	 * dump sections.
531 	 * For each dump section type supported, a 32bit cell which defines
532 	 * the ID of a supported section followed by two 32 bit cells which
533 	 * gives the size of the section in bytes.
534 	 */
535 	sections = of_get_flat_dt_prop(node, "ibm,configure-kernel-dump-sizes",
536 					&size);
537 
538 	if (!sections)
539 		return;
540 
541 	num_sections = size / (3 * sizeof(u32));
542 
543 	for (i = 0; i < num_sections; i++, sections += 3) {
544 		u32 type = (u32)of_read_number(sections, 1);
545 
546 		switch (type) {
547 		case RTAS_FADUMP_CPU_STATE_DATA:
548 			fadump_conf->cpu_state_data_size =
549 					of_read_ulong(&sections[1], 2);
550 			break;
551 		case RTAS_FADUMP_HPTE_REGION:
552 			fadump_conf->hpte_region_size =
553 					of_read_ulong(&sections[1], 2);
554 			break;
555 		}
556 	}
557 }
558