mm, page_alloc: set alloc_flags only once in slowpath
[linux/fpc-iii.git] / drivers / acpi / apei / erst.c
blobf096ab3cb54d6f8b67fa1c7dee4cf0d4cc5179e2
1 /*
2 * APEI Error Record Serialization Table support
4 * ERST is a way provided by APEI to save and retrieve hardware error
5 * information to and from a persistent store.
7 * For more information about ERST, please refer to ACPI Specification
8 * version 4.0, section 17.4.
10 * Copyright 2010 Intel Corp.
11 * Author: Huang Ying <ying.huang@intel.com>
13 * This program is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU General Public License version
15 * 2 as published by the Free Software Foundation.
17 * This program is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20 * GNU General Public License for more details.
23 #include <linux/kernel.h>
24 #include <linux/module.h>
25 #include <linux/init.h>
26 #include <linux/delay.h>
27 #include <linux/io.h>
28 #include <linux/acpi.h>
29 #include <linux/uaccess.h>
30 #include <linux/cper.h>
31 #include <linux/nmi.h>
32 #include <linux/hardirq.h>
33 #include <linux/pstore.h>
34 #include <linux/vmalloc.h>
35 #include <linux/mm.h> /* kvfree() */
36 #include <acpi/apei.h>
38 #include "apei-internal.h"
40 #undef pr_fmt
41 #define pr_fmt(fmt) "ERST: " fmt
43 /* ERST command status */
44 #define ERST_STATUS_SUCCESS 0x0
45 #define ERST_STATUS_NOT_ENOUGH_SPACE 0x1
46 #define ERST_STATUS_HARDWARE_NOT_AVAILABLE 0x2
47 #define ERST_STATUS_FAILED 0x3
48 #define ERST_STATUS_RECORD_STORE_EMPTY 0x4
49 #define ERST_STATUS_RECORD_NOT_FOUND 0x5
51 #define ERST_TAB_ENTRY(tab) \
52 ((struct acpi_whea_header *)((char *)(tab) + \
53 sizeof(struct acpi_table_erst)))
55 #define SPIN_UNIT 100 /* 100ns */
56 /* Firmware should respond within 1 milliseconds */
57 #define FIRMWARE_TIMEOUT (1 * NSEC_PER_MSEC)
58 #define FIRMWARE_MAX_STALL 50 /* 50us */
60 int erst_disable;
61 EXPORT_SYMBOL_GPL(erst_disable);
63 static struct acpi_table_erst *erst_tab;
65 /* ERST Error Log Address Range atrributes */
66 #define ERST_RANGE_RESERVED 0x0001
67 #define ERST_RANGE_NVRAM 0x0002
68 #define ERST_RANGE_SLOW 0x0004
71 * ERST Error Log Address Range, used as buffer for reading/writing
72 * error records.
74 static struct erst_erange {
75 u64 base;
76 u64 size;
77 void __iomem *vaddr;
78 u32 attr;
79 } erst_erange;
82 * Prevent ERST interpreter to run simultaneously, because the
83 * corresponding firmware implementation may not work properly when
84 * invoked simultaneously.
86 * It is used to provide exclusive accessing for ERST Error Log
87 * Address Range too.
89 static DEFINE_RAW_SPINLOCK(erst_lock);
91 static inline int erst_errno(int command_status)
93 switch (command_status) {
94 case ERST_STATUS_SUCCESS:
95 return 0;
96 case ERST_STATUS_HARDWARE_NOT_AVAILABLE:
97 return -ENODEV;
98 case ERST_STATUS_NOT_ENOUGH_SPACE:
99 return -ENOSPC;
100 case ERST_STATUS_RECORD_STORE_EMPTY:
101 case ERST_STATUS_RECORD_NOT_FOUND:
102 return -ENOENT;
103 default:
104 return -EINVAL;
108 static int erst_timedout(u64 *t, u64 spin_unit)
110 if ((s64)*t < spin_unit) {
111 pr_warn(FW_WARN "Firmware does not respond in time.\n");
112 return 1;
114 *t -= spin_unit;
115 ndelay(spin_unit);
116 touch_nmi_watchdog();
117 return 0;
120 static int erst_exec_load_var1(struct apei_exec_context *ctx,
121 struct acpi_whea_header *entry)
123 return __apei_exec_read_register(entry, &ctx->var1);
126 static int erst_exec_load_var2(struct apei_exec_context *ctx,
127 struct acpi_whea_header *entry)
129 return __apei_exec_read_register(entry, &ctx->var2);
132 static int erst_exec_store_var1(struct apei_exec_context *ctx,
133 struct acpi_whea_header *entry)
135 return __apei_exec_write_register(entry, ctx->var1);
138 static int erst_exec_add(struct apei_exec_context *ctx,
139 struct acpi_whea_header *entry)
141 ctx->var1 += ctx->var2;
142 return 0;
145 static int erst_exec_subtract(struct apei_exec_context *ctx,
146 struct acpi_whea_header *entry)
148 ctx->var1 -= ctx->var2;
149 return 0;
152 static int erst_exec_add_value(struct apei_exec_context *ctx,
153 struct acpi_whea_header *entry)
155 int rc;
156 u64 val;
158 rc = __apei_exec_read_register(entry, &val);
159 if (rc)
160 return rc;
161 val += ctx->value;
162 rc = __apei_exec_write_register(entry, val);
163 return rc;
166 static int erst_exec_subtract_value(struct apei_exec_context *ctx,
167 struct acpi_whea_header *entry)
169 int rc;
170 u64 val;
172 rc = __apei_exec_read_register(entry, &val);
173 if (rc)
174 return rc;
175 val -= ctx->value;
176 rc = __apei_exec_write_register(entry, val);
177 return rc;
180 static int erst_exec_stall(struct apei_exec_context *ctx,
181 struct acpi_whea_header *entry)
183 u64 stall_time;
185 if (ctx->value > FIRMWARE_MAX_STALL) {
186 if (!in_nmi())
187 pr_warn(FW_WARN
188 "Too long stall time for stall instruction: 0x%llx.\n",
189 ctx->value);
190 stall_time = FIRMWARE_MAX_STALL;
191 } else
192 stall_time = ctx->value;
193 udelay(stall_time);
194 return 0;
197 static int erst_exec_stall_while_true(struct apei_exec_context *ctx,
198 struct acpi_whea_header *entry)
200 int rc;
201 u64 val;
202 u64 timeout = FIRMWARE_TIMEOUT;
203 u64 stall_time;
205 if (ctx->var1 > FIRMWARE_MAX_STALL) {
206 if (!in_nmi())
207 pr_warn(FW_WARN
208 "Too long stall time for stall while true instruction: 0x%llx.\n",
209 ctx->var1);
210 stall_time = FIRMWARE_MAX_STALL;
211 } else
212 stall_time = ctx->var1;
214 for (;;) {
215 rc = __apei_exec_read_register(entry, &val);
216 if (rc)
217 return rc;
218 if (val != ctx->value)
219 break;
220 if (erst_timedout(&timeout, stall_time * NSEC_PER_USEC))
221 return -EIO;
223 return 0;
226 static int erst_exec_skip_next_instruction_if_true(
227 struct apei_exec_context *ctx,
228 struct acpi_whea_header *entry)
230 int rc;
231 u64 val;
233 rc = __apei_exec_read_register(entry, &val);
234 if (rc)
235 return rc;
236 if (val == ctx->value) {
237 ctx->ip += 2;
238 return APEI_EXEC_SET_IP;
241 return 0;
244 static int erst_exec_goto(struct apei_exec_context *ctx,
245 struct acpi_whea_header *entry)
247 ctx->ip = ctx->value;
248 return APEI_EXEC_SET_IP;
251 static int erst_exec_set_src_address_base(struct apei_exec_context *ctx,
252 struct acpi_whea_header *entry)
254 return __apei_exec_read_register(entry, &ctx->src_base);
257 static int erst_exec_set_dst_address_base(struct apei_exec_context *ctx,
258 struct acpi_whea_header *entry)
260 return __apei_exec_read_register(entry, &ctx->dst_base);
263 static int erst_exec_move_data(struct apei_exec_context *ctx,
264 struct acpi_whea_header *entry)
266 int rc;
267 u64 offset;
268 void *src, *dst;
270 /* ioremap does not work in interrupt context */
271 if (in_interrupt()) {
272 pr_warn("MOVE_DATA can not be used in interrupt context.\n");
273 return -EBUSY;
276 rc = __apei_exec_read_register(entry, &offset);
277 if (rc)
278 return rc;
280 src = ioremap(ctx->src_base + offset, ctx->var2);
281 if (!src)
282 return -ENOMEM;
283 dst = ioremap(ctx->dst_base + offset, ctx->var2);
284 if (!dst) {
285 iounmap(src);
286 return -ENOMEM;
289 memmove(dst, src, ctx->var2);
291 iounmap(src);
292 iounmap(dst);
294 return 0;
297 static struct apei_exec_ins_type erst_ins_type[] = {
298 [ACPI_ERST_READ_REGISTER] = {
299 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
300 .run = apei_exec_read_register,
302 [ACPI_ERST_READ_REGISTER_VALUE] = {
303 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
304 .run = apei_exec_read_register_value,
306 [ACPI_ERST_WRITE_REGISTER] = {
307 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
308 .run = apei_exec_write_register,
310 [ACPI_ERST_WRITE_REGISTER_VALUE] = {
311 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
312 .run = apei_exec_write_register_value,
314 [ACPI_ERST_NOOP] = {
315 .flags = 0,
316 .run = apei_exec_noop,
318 [ACPI_ERST_LOAD_VAR1] = {
319 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
320 .run = erst_exec_load_var1,
322 [ACPI_ERST_LOAD_VAR2] = {
323 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
324 .run = erst_exec_load_var2,
326 [ACPI_ERST_STORE_VAR1] = {
327 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
328 .run = erst_exec_store_var1,
330 [ACPI_ERST_ADD] = {
331 .flags = 0,
332 .run = erst_exec_add,
334 [ACPI_ERST_SUBTRACT] = {
335 .flags = 0,
336 .run = erst_exec_subtract,
338 [ACPI_ERST_ADD_VALUE] = {
339 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
340 .run = erst_exec_add_value,
342 [ACPI_ERST_SUBTRACT_VALUE] = {
343 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
344 .run = erst_exec_subtract_value,
346 [ACPI_ERST_STALL] = {
347 .flags = 0,
348 .run = erst_exec_stall,
350 [ACPI_ERST_STALL_WHILE_TRUE] = {
351 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
352 .run = erst_exec_stall_while_true,
354 [ACPI_ERST_SKIP_NEXT_IF_TRUE] = {
355 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
356 .run = erst_exec_skip_next_instruction_if_true,
358 [ACPI_ERST_GOTO] = {
359 .flags = 0,
360 .run = erst_exec_goto,
362 [ACPI_ERST_SET_SRC_ADDRESS_BASE] = {
363 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
364 .run = erst_exec_set_src_address_base,
366 [ACPI_ERST_SET_DST_ADDRESS_BASE] = {
367 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
368 .run = erst_exec_set_dst_address_base,
370 [ACPI_ERST_MOVE_DATA] = {
371 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
372 .run = erst_exec_move_data,
376 static inline void erst_exec_ctx_init(struct apei_exec_context *ctx)
378 apei_exec_ctx_init(ctx, erst_ins_type, ARRAY_SIZE(erst_ins_type),
379 ERST_TAB_ENTRY(erst_tab), erst_tab->entries);
382 static int erst_get_erange(struct erst_erange *range)
384 struct apei_exec_context ctx;
385 int rc;
387 erst_exec_ctx_init(&ctx);
388 rc = apei_exec_run(&ctx, ACPI_ERST_GET_ERROR_RANGE);
389 if (rc)
390 return rc;
391 range->base = apei_exec_ctx_get_output(&ctx);
392 rc = apei_exec_run(&ctx, ACPI_ERST_GET_ERROR_LENGTH);
393 if (rc)
394 return rc;
395 range->size = apei_exec_ctx_get_output(&ctx);
396 rc = apei_exec_run(&ctx, ACPI_ERST_GET_ERROR_ATTRIBUTES);
397 if (rc)
398 return rc;
399 range->attr = apei_exec_ctx_get_output(&ctx);
401 return 0;
404 static ssize_t __erst_get_record_count(void)
406 struct apei_exec_context ctx;
407 int rc;
409 erst_exec_ctx_init(&ctx);
410 rc = apei_exec_run(&ctx, ACPI_ERST_GET_RECORD_COUNT);
411 if (rc)
412 return rc;
413 return apei_exec_ctx_get_output(&ctx);
416 ssize_t erst_get_record_count(void)
418 ssize_t count;
419 unsigned long flags;
421 if (erst_disable)
422 return -ENODEV;
424 raw_spin_lock_irqsave(&erst_lock, flags);
425 count = __erst_get_record_count();
426 raw_spin_unlock_irqrestore(&erst_lock, flags);
428 return count;
430 EXPORT_SYMBOL_GPL(erst_get_record_count);
432 #define ERST_RECORD_ID_CACHE_SIZE_MIN 16
433 #define ERST_RECORD_ID_CACHE_SIZE_MAX 1024
435 struct erst_record_id_cache {
436 struct mutex lock;
437 u64 *entries;
438 int len;
439 int size;
440 int refcount;
443 static struct erst_record_id_cache erst_record_id_cache = {
444 .lock = __MUTEX_INITIALIZER(erst_record_id_cache.lock),
445 .refcount = 0,
448 static int __erst_get_next_record_id(u64 *record_id)
450 struct apei_exec_context ctx;
451 int rc;
453 erst_exec_ctx_init(&ctx);
454 rc = apei_exec_run(&ctx, ACPI_ERST_GET_RECORD_ID);
455 if (rc)
456 return rc;
457 *record_id = apei_exec_ctx_get_output(&ctx);
459 return 0;
462 int erst_get_record_id_begin(int *pos)
464 int rc;
466 if (erst_disable)
467 return -ENODEV;
469 rc = mutex_lock_interruptible(&erst_record_id_cache.lock);
470 if (rc)
471 return rc;
472 erst_record_id_cache.refcount++;
473 mutex_unlock(&erst_record_id_cache.lock);
475 *pos = 0;
477 return 0;
479 EXPORT_SYMBOL_GPL(erst_get_record_id_begin);
481 /* erst_record_id_cache.lock must be held by caller */
482 static int __erst_record_id_cache_add_one(void)
484 u64 id, prev_id, first_id;
485 int i, rc;
486 u64 *entries;
487 unsigned long flags;
489 id = prev_id = first_id = APEI_ERST_INVALID_RECORD_ID;
490 retry:
491 raw_spin_lock_irqsave(&erst_lock, flags);
492 rc = __erst_get_next_record_id(&id);
493 raw_spin_unlock_irqrestore(&erst_lock, flags);
494 if (rc == -ENOENT)
495 return 0;
496 if (rc)
497 return rc;
498 if (id == APEI_ERST_INVALID_RECORD_ID)
499 return 0;
500 /* can not skip current ID, or loop back to first ID */
501 if (id == prev_id || id == first_id)
502 return 0;
503 if (first_id == APEI_ERST_INVALID_RECORD_ID)
504 first_id = id;
505 prev_id = id;
507 entries = erst_record_id_cache.entries;
508 for (i = 0; i < erst_record_id_cache.len; i++) {
509 if (entries[i] == id)
510 break;
512 /* record id already in cache, try next */
513 if (i < erst_record_id_cache.len)
514 goto retry;
515 if (erst_record_id_cache.len >= erst_record_id_cache.size) {
516 int new_size, alloc_size;
517 u64 *new_entries;
519 new_size = erst_record_id_cache.size * 2;
520 new_size = clamp_val(new_size, ERST_RECORD_ID_CACHE_SIZE_MIN,
521 ERST_RECORD_ID_CACHE_SIZE_MAX);
522 if (new_size <= erst_record_id_cache.size) {
523 if (printk_ratelimit())
524 pr_warn(FW_WARN "too many record IDs!\n");
525 return 0;
527 alloc_size = new_size * sizeof(entries[0]);
528 if (alloc_size < PAGE_SIZE)
529 new_entries = kmalloc(alloc_size, GFP_KERNEL);
530 else
531 new_entries = vmalloc(alloc_size);
532 if (!new_entries)
533 return -ENOMEM;
534 memcpy(new_entries, entries,
535 erst_record_id_cache.len * sizeof(entries[0]));
536 kvfree(entries);
537 erst_record_id_cache.entries = entries = new_entries;
538 erst_record_id_cache.size = new_size;
540 entries[i] = id;
541 erst_record_id_cache.len++;
543 return 1;
547 * Get the record ID of an existing error record on the persistent
548 * storage. If there is no error record on the persistent storage, the
549 * returned record_id is APEI_ERST_INVALID_RECORD_ID.
551 int erst_get_record_id_next(int *pos, u64 *record_id)
553 int rc = 0;
554 u64 *entries;
556 if (erst_disable)
557 return -ENODEV;
559 /* must be enclosed by erst_get_record_id_begin/end */
560 BUG_ON(!erst_record_id_cache.refcount);
561 BUG_ON(*pos < 0 || *pos > erst_record_id_cache.len);
563 mutex_lock(&erst_record_id_cache.lock);
564 entries = erst_record_id_cache.entries;
565 for (; *pos < erst_record_id_cache.len; (*pos)++)
566 if (entries[*pos] != APEI_ERST_INVALID_RECORD_ID)
567 break;
568 /* found next record id in cache */
569 if (*pos < erst_record_id_cache.len) {
570 *record_id = entries[*pos];
571 (*pos)++;
572 goto out_unlock;
575 /* Try to add one more record ID to cache */
576 rc = __erst_record_id_cache_add_one();
577 if (rc < 0)
578 goto out_unlock;
579 /* successfully add one new ID */
580 if (rc == 1) {
581 *record_id = erst_record_id_cache.entries[*pos];
582 (*pos)++;
583 rc = 0;
584 } else {
585 *pos = -1;
586 *record_id = APEI_ERST_INVALID_RECORD_ID;
588 out_unlock:
589 mutex_unlock(&erst_record_id_cache.lock);
591 return rc;
593 EXPORT_SYMBOL_GPL(erst_get_record_id_next);
595 /* erst_record_id_cache.lock must be held by caller */
596 static void __erst_record_id_cache_compact(void)
598 int i, wpos = 0;
599 u64 *entries;
601 if (erst_record_id_cache.refcount)
602 return;
604 entries = erst_record_id_cache.entries;
605 for (i = 0; i < erst_record_id_cache.len; i++) {
606 if (entries[i] == APEI_ERST_INVALID_RECORD_ID)
607 continue;
608 if (wpos != i)
609 entries[wpos] = entries[i];
610 wpos++;
612 erst_record_id_cache.len = wpos;
615 void erst_get_record_id_end(void)
618 * erst_disable != 0 should be detected by invoker via the
619 * return value of erst_get_record_id_begin/next, so this
620 * function should not be called for erst_disable != 0.
622 BUG_ON(erst_disable);
624 mutex_lock(&erst_record_id_cache.lock);
625 erst_record_id_cache.refcount--;
626 BUG_ON(erst_record_id_cache.refcount < 0);
627 __erst_record_id_cache_compact();
628 mutex_unlock(&erst_record_id_cache.lock);
630 EXPORT_SYMBOL_GPL(erst_get_record_id_end);
632 static int __erst_write_to_storage(u64 offset)
634 struct apei_exec_context ctx;
635 u64 timeout = FIRMWARE_TIMEOUT;
636 u64 val;
637 int rc;
639 erst_exec_ctx_init(&ctx);
640 rc = apei_exec_run_optional(&ctx, ACPI_ERST_BEGIN_WRITE);
641 if (rc)
642 return rc;
643 apei_exec_ctx_set_input(&ctx, offset);
644 rc = apei_exec_run(&ctx, ACPI_ERST_SET_RECORD_OFFSET);
645 if (rc)
646 return rc;
647 rc = apei_exec_run(&ctx, ACPI_ERST_EXECUTE_OPERATION);
648 if (rc)
649 return rc;
650 for (;;) {
651 rc = apei_exec_run(&ctx, ACPI_ERST_CHECK_BUSY_STATUS);
652 if (rc)
653 return rc;
654 val = apei_exec_ctx_get_output(&ctx);
655 if (!val)
656 break;
657 if (erst_timedout(&timeout, SPIN_UNIT))
658 return -EIO;
660 rc = apei_exec_run(&ctx, ACPI_ERST_GET_COMMAND_STATUS);
661 if (rc)
662 return rc;
663 val = apei_exec_ctx_get_output(&ctx);
664 rc = apei_exec_run_optional(&ctx, ACPI_ERST_END);
665 if (rc)
666 return rc;
668 return erst_errno(val);
671 static int __erst_read_from_storage(u64 record_id, u64 offset)
673 struct apei_exec_context ctx;
674 u64 timeout = FIRMWARE_TIMEOUT;
675 u64 val;
676 int rc;
678 erst_exec_ctx_init(&ctx);
679 rc = apei_exec_run_optional(&ctx, ACPI_ERST_BEGIN_READ);
680 if (rc)
681 return rc;
682 apei_exec_ctx_set_input(&ctx, offset);
683 rc = apei_exec_run(&ctx, ACPI_ERST_SET_RECORD_OFFSET);
684 if (rc)
685 return rc;
686 apei_exec_ctx_set_input(&ctx, record_id);
687 rc = apei_exec_run(&ctx, ACPI_ERST_SET_RECORD_ID);
688 if (rc)
689 return rc;
690 rc = apei_exec_run(&ctx, ACPI_ERST_EXECUTE_OPERATION);
691 if (rc)
692 return rc;
693 for (;;) {
694 rc = apei_exec_run(&ctx, ACPI_ERST_CHECK_BUSY_STATUS);
695 if (rc)
696 return rc;
697 val = apei_exec_ctx_get_output(&ctx);
698 if (!val)
699 break;
700 if (erst_timedout(&timeout, SPIN_UNIT))
701 return -EIO;
703 rc = apei_exec_run(&ctx, ACPI_ERST_GET_COMMAND_STATUS);
704 if (rc)
705 return rc;
706 val = apei_exec_ctx_get_output(&ctx);
707 rc = apei_exec_run_optional(&ctx, ACPI_ERST_END);
708 if (rc)
709 return rc;
711 return erst_errno(val);
714 static int __erst_clear_from_storage(u64 record_id)
716 struct apei_exec_context ctx;
717 u64 timeout = FIRMWARE_TIMEOUT;
718 u64 val;
719 int rc;
721 erst_exec_ctx_init(&ctx);
722 rc = apei_exec_run_optional(&ctx, ACPI_ERST_BEGIN_CLEAR);
723 if (rc)
724 return rc;
725 apei_exec_ctx_set_input(&ctx, record_id);
726 rc = apei_exec_run(&ctx, ACPI_ERST_SET_RECORD_ID);
727 if (rc)
728 return rc;
729 rc = apei_exec_run(&ctx, ACPI_ERST_EXECUTE_OPERATION);
730 if (rc)
731 return rc;
732 for (;;) {
733 rc = apei_exec_run(&ctx, ACPI_ERST_CHECK_BUSY_STATUS);
734 if (rc)
735 return rc;
736 val = apei_exec_ctx_get_output(&ctx);
737 if (!val)
738 break;
739 if (erst_timedout(&timeout, SPIN_UNIT))
740 return -EIO;
742 rc = apei_exec_run(&ctx, ACPI_ERST_GET_COMMAND_STATUS);
743 if (rc)
744 return rc;
745 val = apei_exec_ctx_get_output(&ctx);
746 rc = apei_exec_run_optional(&ctx, ACPI_ERST_END);
747 if (rc)
748 return rc;
750 return erst_errno(val);
753 /* NVRAM ERST Error Log Address Range is not supported yet */
754 static void pr_unimpl_nvram(void)
756 if (printk_ratelimit())
757 pr_warn("NVRAM ERST Log Address Range not implemented yet.\n");
760 static int __erst_write_to_nvram(const struct cper_record_header *record)
762 /* do not print message, because printk is not safe for NMI */
763 return -ENOSYS;
766 static int __erst_read_to_erange_from_nvram(u64 record_id, u64 *offset)
768 pr_unimpl_nvram();
769 return -ENOSYS;
772 static int __erst_clear_from_nvram(u64 record_id)
774 pr_unimpl_nvram();
775 return -ENOSYS;
778 int erst_write(const struct cper_record_header *record)
780 int rc;
781 unsigned long flags;
782 struct cper_record_header *rcd_erange;
784 if (erst_disable)
785 return -ENODEV;
787 if (memcmp(record->signature, CPER_SIG_RECORD, CPER_SIG_SIZE))
788 return -EINVAL;
790 if (erst_erange.attr & ERST_RANGE_NVRAM) {
791 if (!raw_spin_trylock_irqsave(&erst_lock, flags))
792 return -EBUSY;
793 rc = __erst_write_to_nvram(record);
794 raw_spin_unlock_irqrestore(&erst_lock, flags);
795 return rc;
798 if (record->record_length > erst_erange.size)
799 return -EINVAL;
801 if (!raw_spin_trylock_irqsave(&erst_lock, flags))
802 return -EBUSY;
803 memcpy(erst_erange.vaddr, record, record->record_length);
804 rcd_erange = erst_erange.vaddr;
805 /* signature for serialization system */
806 memcpy(&rcd_erange->persistence_information, "ER", 2);
808 rc = __erst_write_to_storage(0);
809 raw_spin_unlock_irqrestore(&erst_lock, flags);
811 return rc;
813 EXPORT_SYMBOL_GPL(erst_write);
815 static int __erst_read_to_erange(u64 record_id, u64 *offset)
817 int rc;
819 if (erst_erange.attr & ERST_RANGE_NVRAM)
820 return __erst_read_to_erange_from_nvram(
821 record_id, offset);
823 rc = __erst_read_from_storage(record_id, 0);
824 if (rc)
825 return rc;
826 *offset = 0;
828 return 0;
831 static ssize_t __erst_read(u64 record_id, struct cper_record_header *record,
832 size_t buflen)
834 int rc;
835 u64 offset, len = 0;
836 struct cper_record_header *rcd_tmp;
838 rc = __erst_read_to_erange(record_id, &offset);
839 if (rc)
840 return rc;
841 rcd_tmp = erst_erange.vaddr + offset;
842 len = rcd_tmp->record_length;
843 if (len <= buflen)
844 memcpy(record, rcd_tmp, len);
846 return len;
850 * If return value > buflen, the buffer size is not big enough,
851 * else if return value < 0, something goes wrong,
852 * else everything is OK, and return value is record length
854 ssize_t erst_read(u64 record_id, struct cper_record_header *record,
855 size_t buflen)
857 ssize_t len;
858 unsigned long flags;
860 if (erst_disable)
861 return -ENODEV;
863 raw_spin_lock_irqsave(&erst_lock, flags);
864 len = __erst_read(record_id, record, buflen);
865 raw_spin_unlock_irqrestore(&erst_lock, flags);
866 return len;
868 EXPORT_SYMBOL_GPL(erst_read);
870 int erst_clear(u64 record_id)
872 int rc, i;
873 unsigned long flags;
874 u64 *entries;
876 if (erst_disable)
877 return -ENODEV;
879 rc = mutex_lock_interruptible(&erst_record_id_cache.lock);
880 if (rc)
881 return rc;
882 raw_spin_lock_irqsave(&erst_lock, flags);
883 if (erst_erange.attr & ERST_RANGE_NVRAM)
884 rc = __erst_clear_from_nvram(record_id);
885 else
886 rc = __erst_clear_from_storage(record_id);
887 raw_spin_unlock_irqrestore(&erst_lock, flags);
888 if (rc)
889 goto out;
890 entries = erst_record_id_cache.entries;
891 for (i = 0; i < erst_record_id_cache.len; i++) {
892 if (entries[i] == record_id)
893 entries[i] = APEI_ERST_INVALID_RECORD_ID;
895 __erst_record_id_cache_compact();
896 out:
897 mutex_unlock(&erst_record_id_cache.lock);
898 return rc;
900 EXPORT_SYMBOL_GPL(erst_clear);
902 static int __init setup_erst_disable(char *str)
904 erst_disable = 1;
905 return 0;
908 __setup("erst_disable", setup_erst_disable);
910 static int erst_check_table(struct acpi_table_erst *erst_tab)
912 if ((erst_tab->header_length !=
913 (sizeof(struct acpi_table_erst) - sizeof(erst_tab->header)))
914 && (erst_tab->header_length != sizeof(struct acpi_table_erst)))
915 return -EINVAL;
916 if (erst_tab->header.length < sizeof(struct acpi_table_erst))
917 return -EINVAL;
918 if (erst_tab->entries !=
919 (erst_tab->header.length - sizeof(struct acpi_table_erst)) /
920 sizeof(struct acpi_erst_entry))
921 return -EINVAL;
923 return 0;
926 static int erst_open_pstore(struct pstore_info *psi);
927 static int erst_close_pstore(struct pstore_info *psi);
928 static ssize_t erst_reader(u64 *id, enum pstore_type_id *type, int *count,
929 struct timespec *time, char **buf,
930 bool *compressed, ssize_t *ecc_notice_size,
931 struct pstore_info *psi);
932 static int erst_writer(enum pstore_type_id type, enum kmsg_dump_reason reason,
933 u64 *id, unsigned int part, int count, bool compressed,
934 size_t size, struct pstore_info *psi);
935 static int erst_clearer(enum pstore_type_id type, u64 id, int count,
936 struct timespec time, struct pstore_info *psi);
938 static struct pstore_info erst_info = {
939 .owner = THIS_MODULE,
940 .name = "erst",
941 .flags = PSTORE_FLAGS_FRAGILE,
942 .open = erst_open_pstore,
943 .close = erst_close_pstore,
944 .read = erst_reader,
945 .write = erst_writer,
946 .erase = erst_clearer
949 #define CPER_CREATOR_PSTORE \
950 UUID_LE(0x75a574e3, 0x5052, 0x4b29, 0x8a, 0x8e, 0xbe, 0x2c, \
951 0x64, 0x90, 0xb8, 0x9d)
952 #define CPER_SECTION_TYPE_DMESG \
953 UUID_LE(0xc197e04e, 0xd545, 0x4a70, 0x9c, 0x17, 0xa5, 0x54, \
954 0x94, 0x19, 0xeb, 0x12)
955 #define CPER_SECTION_TYPE_DMESG_Z \
956 UUID_LE(0x4f118707, 0x04dd, 0x4055, 0xb5, 0xdd, 0x95, 0x6d, \
957 0x34, 0xdd, 0xfa, 0xc6)
958 #define CPER_SECTION_TYPE_MCE \
959 UUID_LE(0xfe08ffbe, 0x95e4, 0x4be7, 0xbc, 0x73, 0x40, 0x96, \
960 0x04, 0x4a, 0x38, 0xfc)
962 struct cper_pstore_record {
963 struct cper_record_header hdr;
964 struct cper_section_descriptor sec_hdr;
965 char data[];
966 } __packed;
968 static int reader_pos;
970 static int erst_open_pstore(struct pstore_info *psi)
972 int rc;
974 if (erst_disable)
975 return -ENODEV;
977 rc = erst_get_record_id_begin(&reader_pos);
979 return rc;
982 static int erst_close_pstore(struct pstore_info *psi)
984 erst_get_record_id_end();
986 return 0;
989 static ssize_t erst_reader(u64 *id, enum pstore_type_id *type, int *count,
990 struct timespec *time, char **buf,
991 bool *compressed, ssize_t *ecc_notice_size,
992 struct pstore_info *psi)
994 int rc;
995 ssize_t len = 0;
996 u64 record_id;
997 struct cper_pstore_record *rcd;
998 size_t rcd_len = sizeof(*rcd) + erst_info.bufsize;
1000 if (erst_disable)
1001 return -ENODEV;
1003 rcd = kmalloc(rcd_len, GFP_KERNEL);
1004 if (!rcd) {
1005 rc = -ENOMEM;
1006 goto out;
1008 skip:
1009 rc = erst_get_record_id_next(&reader_pos, &record_id);
1010 if (rc)
1011 goto out;
1013 /* no more record */
1014 if (record_id == APEI_ERST_INVALID_RECORD_ID) {
1015 rc = -EINVAL;
1016 goto out;
1019 len = erst_read(record_id, &rcd->hdr, rcd_len);
1020 /* The record may be cleared by others, try read next record */
1021 if (len == -ENOENT)
1022 goto skip;
1023 else if (len < sizeof(*rcd)) {
1024 rc = -EIO;
1025 goto out;
1027 if (uuid_le_cmp(rcd->hdr.creator_id, CPER_CREATOR_PSTORE) != 0)
1028 goto skip;
1030 *buf = kmalloc(len, GFP_KERNEL);
1031 if (*buf == NULL) {
1032 rc = -ENOMEM;
1033 goto out;
1035 memcpy(*buf, rcd->data, len - sizeof(*rcd));
1036 *id = record_id;
1037 *compressed = false;
1038 *ecc_notice_size = 0;
1039 if (uuid_le_cmp(rcd->sec_hdr.section_type,
1040 CPER_SECTION_TYPE_DMESG_Z) == 0) {
1041 *type = PSTORE_TYPE_DMESG;
1042 *compressed = true;
1043 } else if (uuid_le_cmp(rcd->sec_hdr.section_type,
1044 CPER_SECTION_TYPE_DMESG) == 0)
1045 *type = PSTORE_TYPE_DMESG;
1046 else if (uuid_le_cmp(rcd->sec_hdr.section_type,
1047 CPER_SECTION_TYPE_MCE) == 0)
1048 *type = PSTORE_TYPE_MCE;
1049 else
1050 *type = PSTORE_TYPE_UNKNOWN;
1052 if (rcd->hdr.validation_bits & CPER_VALID_TIMESTAMP)
1053 time->tv_sec = rcd->hdr.timestamp;
1054 else
1055 time->tv_sec = 0;
1056 time->tv_nsec = 0;
1058 out:
1059 kfree(rcd);
1060 return (rc < 0) ? rc : (len - sizeof(*rcd));
1063 static int erst_writer(enum pstore_type_id type, enum kmsg_dump_reason reason,
1064 u64 *id, unsigned int part, int count, bool compressed,
1065 size_t size, struct pstore_info *psi)
1067 struct cper_pstore_record *rcd = (struct cper_pstore_record *)
1068 (erst_info.buf - sizeof(*rcd));
1069 int ret;
1071 memset(rcd, 0, sizeof(*rcd));
1072 memcpy(rcd->hdr.signature, CPER_SIG_RECORD, CPER_SIG_SIZE);
1073 rcd->hdr.revision = CPER_RECORD_REV;
1074 rcd->hdr.signature_end = CPER_SIG_END;
1075 rcd->hdr.section_count = 1;
1076 rcd->hdr.error_severity = CPER_SEV_FATAL;
1077 /* timestamp valid. platform_id, partition_id are invalid */
1078 rcd->hdr.validation_bits = CPER_VALID_TIMESTAMP;
1079 rcd->hdr.timestamp = get_seconds();
1080 rcd->hdr.record_length = sizeof(*rcd) + size;
1081 rcd->hdr.creator_id = CPER_CREATOR_PSTORE;
1082 rcd->hdr.notification_type = CPER_NOTIFY_MCE;
1083 rcd->hdr.record_id = cper_next_record_id();
1084 rcd->hdr.flags = CPER_HW_ERROR_FLAGS_PREVERR;
1086 rcd->sec_hdr.section_offset = sizeof(*rcd);
1087 rcd->sec_hdr.section_length = size;
1088 rcd->sec_hdr.revision = CPER_SEC_REV;
1089 /* fru_id and fru_text is invalid */
1090 rcd->sec_hdr.validation_bits = 0;
1091 rcd->sec_hdr.flags = CPER_SEC_PRIMARY;
1092 switch (type) {
1093 case PSTORE_TYPE_DMESG:
1094 if (compressed)
1095 rcd->sec_hdr.section_type = CPER_SECTION_TYPE_DMESG_Z;
1096 else
1097 rcd->sec_hdr.section_type = CPER_SECTION_TYPE_DMESG;
1098 break;
1099 case PSTORE_TYPE_MCE:
1100 rcd->sec_hdr.section_type = CPER_SECTION_TYPE_MCE;
1101 break;
1102 default:
1103 return -EINVAL;
1105 rcd->sec_hdr.section_severity = CPER_SEV_FATAL;
1107 ret = erst_write(&rcd->hdr);
1108 *id = rcd->hdr.record_id;
1110 return ret;
1113 static int erst_clearer(enum pstore_type_id type, u64 id, int count,
1114 struct timespec time, struct pstore_info *psi)
1116 return erst_clear(id);
1119 static int __init erst_init(void)
1121 int rc = 0;
1122 acpi_status status;
1123 struct apei_exec_context ctx;
1124 struct apei_resources erst_resources;
1125 struct resource *r;
1126 char *buf;
1128 if (acpi_disabled)
1129 goto err;
1131 if (erst_disable) {
1132 pr_info(
1133 "Error Record Serialization Table (ERST) support is disabled.\n");
1134 goto err;
1137 status = acpi_get_table(ACPI_SIG_ERST, 0,
1138 (struct acpi_table_header **)&erst_tab);
1139 if (status == AE_NOT_FOUND)
1140 goto err;
1141 else if (ACPI_FAILURE(status)) {
1142 const char *msg = acpi_format_exception(status);
1143 pr_err("Failed to get table, %s\n", msg);
1144 rc = -EINVAL;
1145 goto err;
1148 rc = erst_check_table(erst_tab);
1149 if (rc) {
1150 pr_err(FW_BUG "ERST table is invalid.\n");
1151 goto err;
1154 apei_resources_init(&erst_resources);
1155 erst_exec_ctx_init(&ctx);
1156 rc = apei_exec_collect_resources(&ctx, &erst_resources);
1157 if (rc)
1158 goto err_fini;
1159 rc = apei_resources_request(&erst_resources, "APEI ERST");
1160 if (rc)
1161 goto err_fini;
1162 rc = apei_exec_pre_map_gars(&ctx);
1163 if (rc)
1164 goto err_release;
1165 rc = erst_get_erange(&erst_erange);
1166 if (rc) {
1167 if (rc == -ENODEV)
1168 pr_info(
1169 "The corresponding hardware device or firmware implementation "
1170 "is not available.\n");
1171 else
1172 pr_err("Failed to get Error Log Address Range.\n");
1173 goto err_unmap_reg;
1176 r = request_mem_region(erst_erange.base, erst_erange.size, "APEI ERST");
1177 if (!r) {
1178 pr_err("Can not request [mem %#010llx-%#010llx] for ERST.\n",
1179 (unsigned long long)erst_erange.base,
1180 (unsigned long long)erst_erange.base + erst_erange.size - 1);
1181 rc = -EIO;
1182 goto err_unmap_reg;
1184 rc = -ENOMEM;
1185 erst_erange.vaddr = ioremap_cache(erst_erange.base,
1186 erst_erange.size);
1187 if (!erst_erange.vaddr)
1188 goto err_release_erange;
1190 pr_info(
1191 "Error Record Serialization Table (ERST) support is initialized.\n");
1193 buf = kmalloc(erst_erange.size, GFP_KERNEL);
1194 spin_lock_init(&erst_info.buf_lock);
1195 if (buf) {
1196 erst_info.buf = buf + sizeof(struct cper_pstore_record);
1197 erst_info.bufsize = erst_erange.size -
1198 sizeof(struct cper_pstore_record);
1199 rc = pstore_register(&erst_info);
1200 if (rc) {
1201 if (rc != -EPERM)
1202 pr_info(
1203 "Could not register with persistent store.\n");
1204 erst_info.buf = NULL;
1205 erst_info.bufsize = 0;
1206 kfree(buf);
1208 } else
1209 pr_err(
1210 "Failed to allocate %lld bytes for persistent store error log.\n",
1211 erst_erange.size);
1213 /* Cleanup ERST Resources */
1214 apei_resources_fini(&erst_resources);
1216 return 0;
1218 err_release_erange:
1219 release_mem_region(erst_erange.base, erst_erange.size);
1220 err_unmap_reg:
1221 apei_exec_post_unmap_gars(&ctx);
1222 err_release:
1223 apei_resources_release(&erst_resources);
1224 err_fini:
1225 apei_resources_fini(&erst_resources);
1226 err:
1227 erst_disable = 1;
1228 return rc;
1231 device_initcall(erst_init);