2 * APEI Error INJection support
4 * EINJ provides a hardware error injection mechanism, this is useful
5 * for debugging and testing of other APEI and RAS features.
7 * For more information about EINJ, please refer to ACPI Specification
8 * version 4.0, section 17.5.
10 * Copyright 2009-2010 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.
22 * You should have received a copy of the GNU General Public License
23 * along with this program; if not, write to the Free Software
24 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
27 #include <linux/kernel.h>
28 #include <linux/module.h>
29 #include <linux/init.h>
31 #include <linux/debugfs.h>
32 #include <linux/seq_file.h>
33 #include <linux/nmi.h>
34 #include <linux/delay.h>
35 #include <acpi/acpi.h>
37 #include "apei-internal.h"
39 #define EINJ_PFX "EINJ: "
41 #define SPIN_UNIT 100 /* 100ns */
42 /* Firmware should respond within 1 milliseconds */
43 #define FIRMWARE_TIMEOUT (1 * NSEC_PER_MSEC)
46 * ACPI version 5 provides a SET_ERROR_TYPE_WITH_ADDRESS action.
50 struct set_error_type_with_address {
56 u64 memory_address_range;
60 SETWA_FLAGS_APICID = 1,
62 SETWA_FLAGS_PCIE_SBDF = 4,
66 * Vendor extensions for platform specific operations
68 struct vendor_error_type_extension {
79 static u32 vendor_flags;
80 static struct debugfs_blob_wrapper vendor_blob;
81 static char vendor_dev[64];
84 * Some BIOSes allow parameters to the SET_ERROR_TYPE entries in the
85 * EINJ table through an unpublished extension. Use with caution as
86 * most will ignore the parameter and make their own choice of address
87 * for error injection. This extension is used only if
88 * param_extension module parameter is specified.
90 struct einj_parameter {
98 #define EINJ_OP_BUSY 0x1
99 #define EINJ_STATUS_SUCCESS 0x0
100 #define EINJ_STATUS_FAIL 0x1
101 #define EINJ_STATUS_INVAL 0x2
103 #define EINJ_TAB_ENTRY(tab) \
104 ((struct acpi_whea_header *)((char *)(tab) + \
105 sizeof(struct acpi_table_einj)))
107 static bool param_extension;
108 module_param(param_extension, bool, 0);
110 static struct acpi_table_einj *einj_tab;
112 static struct apei_resources einj_resources;
114 static struct apei_exec_ins_type einj_ins_type[] = {
115 [ACPI_EINJ_READ_REGISTER] = {
116 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
117 .run = apei_exec_read_register,
119 [ACPI_EINJ_READ_REGISTER_VALUE] = {
120 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
121 .run = apei_exec_read_register_value,
123 [ACPI_EINJ_WRITE_REGISTER] = {
124 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
125 .run = apei_exec_write_register,
127 [ACPI_EINJ_WRITE_REGISTER_VALUE] = {
128 .flags = APEI_EXEC_INS_ACCESS_REGISTER,
129 .run = apei_exec_write_register_value,
133 .run = apei_exec_noop,
138 * Prevent EINJ interpreter to run simultaneously, because the
139 * corresponding firmware implementation may not work properly when
140 * invoked simultaneously.
142 static DEFINE_MUTEX(einj_mutex);
144 static void *einj_param;
146 static void einj_exec_ctx_init(struct apei_exec_context *ctx)
148 apei_exec_ctx_init(ctx, einj_ins_type, ARRAY_SIZE(einj_ins_type),
149 EINJ_TAB_ENTRY(einj_tab), einj_tab->entries);
152 static int __einj_get_available_error_type(u32 *type)
154 struct apei_exec_context ctx;
157 einj_exec_ctx_init(&ctx);
158 rc = apei_exec_run(&ctx, ACPI_EINJ_GET_ERROR_TYPE);
161 *type = apei_exec_ctx_get_output(&ctx);
166 /* Get error injection capabilities of the platform */
167 static int einj_get_available_error_type(u32 *type)
171 mutex_lock(&einj_mutex);
172 rc = __einj_get_available_error_type(type);
173 mutex_unlock(&einj_mutex);
178 static int einj_timedout(u64 *t)
180 if ((s64)*t < SPIN_UNIT) {
181 pr_warning(FW_WARN EINJ_PFX
182 "Firmware does not respond in time\n");
187 touch_nmi_watchdog();
191 static void check_vendor_extension(u64 paddr,
192 struct set_error_type_with_address *v5param)
194 int offset = v5param->vendor_extension;
195 struct vendor_error_type_extension *v;
200 v = acpi_os_map_memory(paddr + offset, sizeof(*v));
204 sprintf(vendor_dev, "%x:%x:%x.%x vendor_id=%x device_id=%x rev_id=%x\n",
205 sbdf >> 24, (sbdf >> 16) & 0xff,
206 (sbdf >> 11) & 0x1f, (sbdf >> 8) & 0x7,
207 v->vendor_id, v->device_id, v->rev_id);
208 acpi_os_unmap_memory(v, sizeof(*v));
211 static void *einj_get_parameter_address(void)
214 u64 paddrv4 = 0, paddrv5 = 0;
215 struct acpi_whea_header *entry;
217 entry = EINJ_TAB_ENTRY(einj_tab);
218 for (i = 0; i < einj_tab->entries; i++) {
219 if (entry->action == ACPI_EINJ_SET_ERROR_TYPE &&
220 entry->instruction == ACPI_EINJ_WRITE_REGISTER &&
221 entry->register_region.space_id ==
222 ACPI_ADR_SPACE_SYSTEM_MEMORY)
223 memcpy(&paddrv4, &entry->register_region.address,
225 if (entry->action == ACPI_EINJ_SET_ERROR_TYPE_WITH_ADDRESS &&
226 entry->instruction == ACPI_EINJ_WRITE_REGISTER &&
227 entry->register_region.space_id ==
228 ACPI_ADR_SPACE_SYSTEM_MEMORY)
229 memcpy(&paddrv5, &entry->register_region.address,
234 struct set_error_type_with_address *v5param;
236 v5param = acpi_os_map_memory(paddrv5, sizeof(*v5param));
239 check_vendor_extension(paddrv5, v5param);
243 if (param_extension && paddrv4) {
244 struct einj_parameter *v4param;
246 v4param = acpi_os_map_memory(paddrv4, sizeof(*v4param));
249 if (v4param->reserved1 || v4param->reserved2) {
250 acpi_os_unmap_memory(v4param, sizeof(*v4param));
259 /* do sanity check to trigger table */
260 static int einj_check_trigger_header(struct acpi_einj_trigger *trigger_tab)
262 if (trigger_tab->header_size != sizeof(struct acpi_einj_trigger))
264 if (trigger_tab->table_size > PAGE_SIZE ||
265 trigger_tab->table_size < trigger_tab->header_size)
267 if (trigger_tab->entry_count !=
268 (trigger_tab->table_size - trigger_tab->header_size) /
269 sizeof(struct acpi_einj_entry))
275 static struct acpi_generic_address *einj_get_trigger_parameter_region(
276 struct acpi_einj_trigger *trigger_tab, u64 param1, u64 param2)
279 struct acpi_whea_header *entry;
281 entry = (struct acpi_whea_header *)
282 ((char *)trigger_tab + sizeof(struct acpi_einj_trigger));
283 for (i = 0; i < trigger_tab->entry_count; i++) {
284 if (entry->action == ACPI_EINJ_TRIGGER_ERROR &&
285 entry->instruction == ACPI_EINJ_WRITE_REGISTER_VALUE &&
286 entry->register_region.space_id ==
287 ACPI_ADR_SPACE_SYSTEM_MEMORY &&
288 (entry->register_region.address & param2) == (param1 & param2))
289 return &entry->register_region;
295 /* Execute instructions in trigger error action table */
296 static int __einj_error_trigger(u64 trigger_paddr, u32 type,
297 u64 param1, u64 param2)
299 struct acpi_einj_trigger *trigger_tab = NULL;
300 struct apei_exec_context trigger_ctx;
301 struct apei_resources trigger_resources;
302 struct acpi_whea_header *trigger_entry;
306 struct acpi_generic_address *trigger_param_region = NULL;
308 r = request_mem_region(trigger_paddr, sizeof(*trigger_tab),
309 "APEI EINJ Trigger Table");
312 "Can not request [mem %#010llx-%#010llx] for Trigger table\n",
313 (unsigned long long)trigger_paddr,
314 (unsigned long long)trigger_paddr +
315 sizeof(*trigger_tab) - 1);
318 trigger_tab = ioremap_cache(trigger_paddr, sizeof(*trigger_tab));
320 pr_err(EINJ_PFX "Failed to map trigger table!\n");
323 rc = einj_check_trigger_header(trigger_tab);
325 pr_warning(FW_BUG EINJ_PFX
326 "The trigger error action table is invalid\n");
330 /* No action structures in the TRIGGER_ERROR table, nothing to do */
331 if (!trigger_tab->entry_count)
335 table_size = trigger_tab->table_size;
336 r = request_mem_region(trigger_paddr + sizeof(*trigger_tab),
337 table_size - sizeof(*trigger_tab),
338 "APEI EINJ Trigger Table");
341 "Can not request [mem %#010llx-%#010llx] for Trigger Table Entry\n",
342 (unsigned long long)trigger_paddr + sizeof(*trigger_tab),
343 (unsigned long long)trigger_paddr + table_size - 1);
346 iounmap(trigger_tab);
347 trigger_tab = ioremap_cache(trigger_paddr, table_size);
349 pr_err(EINJ_PFX "Failed to map trigger table!\n");
352 trigger_entry = (struct acpi_whea_header *)
353 ((char *)trigger_tab + sizeof(struct acpi_einj_trigger));
354 apei_resources_init(&trigger_resources);
355 apei_exec_ctx_init(&trigger_ctx, einj_ins_type,
356 ARRAY_SIZE(einj_ins_type),
357 trigger_entry, trigger_tab->entry_count);
358 rc = apei_exec_collect_resources(&trigger_ctx, &trigger_resources);
361 rc = apei_resources_sub(&trigger_resources, &einj_resources);
365 * Some firmware will access target address specified in
366 * param1 to trigger the error when injecting memory error.
367 * This will cause resource conflict with regular memory. So
368 * remove it from trigger table resources.
370 if ((param_extension || acpi5) && (type & 0x0038) && param2) {
371 struct apei_resources addr_resources;
372 apei_resources_init(&addr_resources);
373 trigger_param_region = einj_get_trigger_parameter_region(
374 trigger_tab, param1, param2);
375 if (trigger_param_region) {
376 rc = apei_resources_add(&addr_resources,
377 trigger_param_region->address,
378 trigger_param_region->bit_width/8, true);
381 rc = apei_resources_sub(&trigger_resources,
384 apei_resources_fini(&addr_resources);
388 rc = apei_resources_request(&trigger_resources, "APEI EINJ Trigger");
391 rc = apei_exec_pre_map_gars(&trigger_ctx);
395 rc = apei_exec_run(&trigger_ctx, ACPI_EINJ_TRIGGER_ERROR);
397 apei_exec_post_unmap_gars(&trigger_ctx);
399 apei_resources_release(&trigger_resources);
401 apei_resources_fini(&trigger_resources);
403 release_mem_region(trigger_paddr + sizeof(*trigger_tab),
404 table_size - sizeof(*trigger_tab));
406 release_mem_region(trigger_paddr, sizeof(*trigger_tab));
409 iounmap(trigger_tab);
414 static int __einj_error_inject(u32 type, u64 param1, u64 param2)
416 struct apei_exec_context ctx;
417 u64 val, trigger_paddr, timeout = FIRMWARE_TIMEOUT;
420 einj_exec_ctx_init(&ctx);
422 rc = apei_exec_run_optional(&ctx, ACPI_EINJ_BEGIN_OPERATION);
425 apei_exec_ctx_set_input(&ctx, type);
427 struct set_error_type_with_address *v5param = einj_param;
429 v5param->type = type;
430 if (type & 0x80000000) {
431 switch (vendor_flags) {
432 case SETWA_FLAGS_APICID:
433 v5param->apicid = param1;
435 case SETWA_FLAGS_MEM:
436 v5param->memory_address = param1;
437 v5param->memory_address_range = param2;
439 case SETWA_FLAGS_PCIE_SBDF:
440 v5param->pcie_sbdf = param1;
443 v5param->flags = vendor_flags;
446 case ACPI_EINJ_PROCESSOR_CORRECTABLE:
447 case ACPI_EINJ_PROCESSOR_UNCORRECTABLE:
448 case ACPI_EINJ_PROCESSOR_FATAL:
449 v5param->apicid = param1;
450 v5param->flags = SETWA_FLAGS_APICID;
452 case ACPI_EINJ_MEMORY_CORRECTABLE:
453 case ACPI_EINJ_MEMORY_UNCORRECTABLE:
454 case ACPI_EINJ_MEMORY_FATAL:
455 v5param->memory_address = param1;
456 v5param->memory_address_range = param2;
457 v5param->flags = SETWA_FLAGS_MEM;
459 case ACPI_EINJ_PCIX_CORRECTABLE:
460 case ACPI_EINJ_PCIX_UNCORRECTABLE:
461 case ACPI_EINJ_PCIX_FATAL:
462 v5param->pcie_sbdf = param1;
463 v5param->flags = SETWA_FLAGS_PCIE_SBDF;
468 rc = apei_exec_run(&ctx, ACPI_EINJ_SET_ERROR_TYPE);
472 struct einj_parameter *v4param = einj_param;
473 v4param->param1 = param1;
474 v4param->param2 = param2;
477 rc = apei_exec_run(&ctx, ACPI_EINJ_EXECUTE_OPERATION);
481 rc = apei_exec_run(&ctx, ACPI_EINJ_CHECK_BUSY_STATUS);
484 val = apei_exec_ctx_get_output(&ctx);
485 if (!(val & EINJ_OP_BUSY))
487 if (einj_timedout(&timeout))
490 rc = apei_exec_run(&ctx, ACPI_EINJ_GET_COMMAND_STATUS);
493 val = apei_exec_ctx_get_output(&ctx);
494 if (val != EINJ_STATUS_SUCCESS)
497 rc = apei_exec_run(&ctx, ACPI_EINJ_GET_TRIGGER_TABLE);
500 trigger_paddr = apei_exec_ctx_get_output(&ctx);
501 if (notrigger == 0) {
502 rc = __einj_error_trigger(trigger_paddr, type, param1, param2);
506 rc = apei_exec_run_optional(&ctx, ACPI_EINJ_END_OPERATION);
511 /* Inject the specified hardware error */
512 static int einj_error_inject(u32 type, u64 param1, u64 param2)
516 mutex_lock(&einj_mutex);
517 rc = __einj_error_inject(type, param1, param2);
518 mutex_unlock(&einj_mutex);
523 static u32 error_type;
524 static u64 error_param1;
525 static u64 error_param2;
526 static struct dentry *einj_debug_dir;
528 static int available_error_type_show(struct seq_file *m, void *v)
531 u32 available_error_type = 0;
533 rc = einj_get_available_error_type(&available_error_type);
536 if (available_error_type & 0x0001)
537 seq_printf(m, "0x00000001\tProcessor Correctable\n");
538 if (available_error_type & 0x0002)
539 seq_printf(m, "0x00000002\tProcessor Uncorrectable non-fatal\n");
540 if (available_error_type & 0x0004)
541 seq_printf(m, "0x00000004\tProcessor Uncorrectable fatal\n");
542 if (available_error_type & 0x0008)
543 seq_printf(m, "0x00000008\tMemory Correctable\n");
544 if (available_error_type & 0x0010)
545 seq_printf(m, "0x00000010\tMemory Uncorrectable non-fatal\n");
546 if (available_error_type & 0x0020)
547 seq_printf(m, "0x00000020\tMemory Uncorrectable fatal\n");
548 if (available_error_type & 0x0040)
549 seq_printf(m, "0x00000040\tPCI Express Correctable\n");
550 if (available_error_type & 0x0080)
551 seq_printf(m, "0x00000080\tPCI Express Uncorrectable non-fatal\n");
552 if (available_error_type & 0x0100)
553 seq_printf(m, "0x00000100\tPCI Express Uncorrectable fatal\n");
554 if (available_error_type & 0x0200)
555 seq_printf(m, "0x00000200\tPlatform Correctable\n");
556 if (available_error_type & 0x0400)
557 seq_printf(m, "0x00000400\tPlatform Uncorrectable non-fatal\n");
558 if (available_error_type & 0x0800)
559 seq_printf(m, "0x00000800\tPlatform Uncorrectable fatal\n");
564 static int available_error_type_open(struct inode *inode, struct file *file)
566 return single_open(file, available_error_type_show, NULL);
569 static const struct file_operations available_error_type_fops = {
570 .open = available_error_type_open,
573 .release = single_release,
576 static int error_type_get(void *data, u64 *val)
583 static int error_type_set(void *data, u64 val)
586 u32 available_error_type = 0;
590 * Vendor defined types have 0x80000000 bit set, and
591 * are not enumerated by ACPI_EINJ_GET_ERROR_TYPE
593 vendor = val & 0x80000000;
594 tval = val & 0x7fffffff;
596 /* Only one error type can be specified */
597 if (tval & (tval - 1))
600 rc = einj_get_available_error_type(&available_error_type);
603 if (!(val & available_error_type))
611 DEFINE_SIMPLE_ATTRIBUTE(error_type_fops, error_type_get,
612 error_type_set, "0x%llx\n");
614 static int error_inject_set(void *data, u64 val)
619 return einj_error_inject(error_type, error_param1, error_param2);
622 DEFINE_SIMPLE_ATTRIBUTE(error_inject_fops, NULL,
623 error_inject_set, "%llu\n");
625 static int einj_check_table(struct acpi_table_einj *einj_tab)
627 if ((einj_tab->header_length !=
628 (sizeof(struct acpi_table_einj) - sizeof(einj_tab->header)))
629 && (einj_tab->header_length != sizeof(struct acpi_table_einj)))
631 if (einj_tab->header.length < sizeof(struct acpi_table_einj))
633 if (einj_tab->entries !=
634 (einj_tab->header.length - sizeof(struct acpi_table_einj)) /
635 sizeof(struct acpi_einj_entry))
641 static int __init einj_init(void)
645 struct dentry *fentry;
646 struct apei_exec_context ctx;
651 status = acpi_get_table(ACPI_SIG_EINJ, 0,
652 (struct acpi_table_header **)&einj_tab);
653 if (status == AE_NOT_FOUND)
655 else if (ACPI_FAILURE(status)) {
656 const char *msg = acpi_format_exception(status);
657 pr_err(EINJ_PFX "Failed to get table, %s\n", msg);
661 rc = einj_check_table(einj_tab);
663 pr_warning(FW_BUG EINJ_PFX "EINJ table is invalid\n");
668 einj_debug_dir = debugfs_create_dir("einj", apei_get_debugfs_dir());
671 fentry = debugfs_create_file("available_error_type", S_IRUSR,
672 einj_debug_dir, NULL,
673 &available_error_type_fops);
676 fentry = debugfs_create_file("error_type", S_IRUSR | S_IWUSR,
677 einj_debug_dir, NULL, &error_type_fops);
680 fentry = debugfs_create_file("error_inject", S_IWUSR,
681 einj_debug_dir, NULL, &error_inject_fops);
685 apei_resources_init(&einj_resources);
686 einj_exec_ctx_init(&ctx);
687 rc = apei_exec_collect_resources(&ctx, &einj_resources);
690 rc = apei_resources_request(&einj_resources, "APEI EINJ");
693 rc = apei_exec_pre_map_gars(&ctx);
697 einj_param = einj_get_parameter_address();
698 if ((param_extension || acpi5) && einj_param) {
699 fentry = debugfs_create_x64("param1", S_IRUSR | S_IWUSR,
700 einj_debug_dir, &error_param1);
703 fentry = debugfs_create_x64("param2", S_IRUSR | S_IWUSR,
704 einj_debug_dir, &error_param2);
708 fentry = debugfs_create_x32("notrigger", S_IRUSR | S_IWUSR,
709 einj_debug_dir, ¬rigger);
715 vendor_blob.data = vendor_dev;
716 vendor_blob.size = strlen(vendor_dev);
717 fentry = debugfs_create_blob("vendor", S_IRUSR,
718 einj_debug_dir, &vendor_blob);
721 fentry = debugfs_create_x32("vendor_flags", S_IRUSR | S_IWUSR,
722 einj_debug_dir, &vendor_flags);
727 pr_info(EINJ_PFX "Error INJection is initialized.\n");
733 acpi_size size = (acpi5) ?
734 sizeof(struct set_error_type_with_address) :
735 sizeof(struct einj_parameter);
737 acpi_os_unmap_memory(einj_param, size);
739 apei_exec_post_unmap_gars(&ctx);
741 apei_resources_release(&einj_resources);
743 apei_resources_fini(&einj_resources);
745 debugfs_remove_recursive(einj_debug_dir);
750 static void __exit einj_exit(void)
752 struct apei_exec_context ctx;
755 acpi_size size = (acpi5) ?
756 sizeof(struct set_error_type_with_address) :
757 sizeof(struct einj_parameter);
759 acpi_os_unmap_memory(einj_param, size);
761 einj_exec_ctx_init(&ctx);
762 apei_exec_post_unmap_gars(&ctx);
763 apei_resources_release(&einj_resources);
764 apei_resources_fini(&einj_resources);
765 debugfs_remove_recursive(einj_debug_dir);
768 module_init(einj_init);
769 module_exit(einj_exit);
771 MODULE_AUTHOR("Huang Ying");
772 MODULE_DESCRIPTION("APEI Error INJection support");
773 MODULE_LICENSE("GPL");