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
{
77 static u32 vendor_flags
;
78 static struct debugfs_blob_wrapper vendor_blob
;
79 static char vendor_dev
[64];
82 * Some BIOSes allow parameters to the SET_ERROR_TYPE entries in the
83 * EINJ table through an unpublished extension. Use with caution as
84 * most will ignore the parameter and make their own choice of address
85 * for error injection. This extension is used only if
86 * param_extension module parameter is specified.
88 struct einj_parameter
{
96 #define EINJ_OP_BUSY 0x1
97 #define EINJ_STATUS_SUCCESS 0x0
98 #define EINJ_STATUS_FAIL 0x1
99 #define EINJ_STATUS_INVAL 0x2
101 #define EINJ_TAB_ENTRY(tab) \
102 ((struct acpi_whea_header *)((char *)(tab) + \
103 sizeof(struct acpi_table_einj)))
105 static bool param_extension
;
106 module_param(param_extension
, bool, 0);
108 static struct acpi_table_einj
*einj_tab
;
110 static struct apei_resources einj_resources
;
112 static struct apei_exec_ins_type einj_ins_type
[] = {
113 [ACPI_EINJ_READ_REGISTER
] = {
114 .flags
= APEI_EXEC_INS_ACCESS_REGISTER
,
115 .run
= apei_exec_read_register
,
117 [ACPI_EINJ_READ_REGISTER_VALUE
] = {
118 .flags
= APEI_EXEC_INS_ACCESS_REGISTER
,
119 .run
= apei_exec_read_register_value
,
121 [ACPI_EINJ_WRITE_REGISTER
] = {
122 .flags
= APEI_EXEC_INS_ACCESS_REGISTER
,
123 .run
= apei_exec_write_register
,
125 [ACPI_EINJ_WRITE_REGISTER_VALUE
] = {
126 .flags
= APEI_EXEC_INS_ACCESS_REGISTER
,
127 .run
= apei_exec_write_register_value
,
131 .run
= apei_exec_noop
,
136 * Prevent EINJ interpreter to run simultaneously, because the
137 * corresponding firmware implementation may not work properly when
138 * invoked simultaneously.
140 static DEFINE_MUTEX(einj_mutex
);
142 static void *einj_param
;
144 static void einj_exec_ctx_init(struct apei_exec_context
*ctx
)
146 apei_exec_ctx_init(ctx
, einj_ins_type
, ARRAY_SIZE(einj_ins_type
),
147 EINJ_TAB_ENTRY(einj_tab
), einj_tab
->entries
);
150 static int __einj_get_available_error_type(u32
*type
)
152 struct apei_exec_context ctx
;
155 einj_exec_ctx_init(&ctx
);
156 rc
= apei_exec_run(&ctx
, ACPI_EINJ_GET_ERROR_TYPE
);
159 *type
= apei_exec_ctx_get_output(&ctx
);
164 /* Get error injection capabilities of the platform */
165 static int einj_get_available_error_type(u32
*type
)
169 mutex_lock(&einj_mutex
);
170 rc
= __einj_get_available_error_type(type
);
171 mutex_unlock(&einj_mutex
);
176 static int einj_timedout(u64
*t
)
178 if ((s64
)*t
< SPIN_UNIT
) {
179 pr_warning(FW_WARN EINJ_PFX
180 "Firmware does not respond in time\n");
185 touch_nmi_watchdog();
189 static void check_vendor_extension(u64 paddr
,
190 struct set_error_type_with_address
*v5param
)
192 int offset
= v5param
->vendor_extension
;
193 struct vendor_error_type_extension
*v
;
198 v
= acpi_os_map_memory(paddr
+ offset
, sizeof(*v
));
202 sprintf(vendor_dev
, "%x:%x:%x.%x vendor_id=%x device_id=%x rev_id=%x\n",
203 sbdf
>> 24, (sbdf
>> 16) & 0xff,
204 (sbdf
>> 11) & 0x1f, (sbdf
>> 8) & 0x7,
205 v
->vendor_id
, v
->device_id
, v
->rev_id
);
206 acpi_os_unmap_memory(v
, sizeof(*v
));
209 static void *einj_get_parameter_address(void)
212 u64 paddrv4
= 0, paddrv5
= 0;
213 struct acpi_whea_header
*entry
;
215 entry
= EINJ_TAB_ENTRY(einj_tab
);
216 for (i
= 0; i
< einj_tab
->entries
; i
++) {
217 if (entry
->action
== ACPI_EINJ_SET_ERROR_TYPE
&&
218 entry
->instruction
== ACPI_EINJ_WRITE_REGISTER
&&
219 entry
->register_region
.space_id
==
220 ACPI_ADR_SPACE_SYSTEM_MEMORY
)
221 memcpy(&paddrv4
, &entry
->register_region
.address
,
223 if (entry
->action
== ACPI_EINJ_SET_ERROR_TYPE_WITH_ADDRESS
&&
224 entry
->instruction
== ACPI_EINJ_WRITE_REGISTER
&&
225 entry
->register_region
.space_id
==
226 ACPI_ADR_SPACE_SYSTEM_MEMORY
)
227 memcpy(&paddrv5
, &entry
->register_region
.address
,
232 struct set_error_type_with_address
*v5param
;
234 v5param
= acpi_os_map_memory(paddrv5
, sizeof(*v5param
));
237 check_vendor_extension(paddrv5
, v5param
);
242 struct einj_parameter
*v4param
;
244 v4param
= acpi_os_map_memory(paddrv4
, sizeof(*v4param
));
247 if (v4param
->reserved1
|| v4param
->reserved2
) {
248 acpi_os_unmap_memory(v4param
, sizeof(*v4param
));
257 /* do sanity check to trigger table */
258 static int einj_check_trigger_header(struct acpi_einj_trigger
*trigger_tab
)
260 if (trigger_tab
->header_size
!= sizeof(struct acpi_einj_trigger
))
262 if (trigger_tab
->table_size
> PAGE_SIZE
||
263 trigger_tab
->table_size
< trigger_tab
->header_size
)
265 if (trigger_tab
->entry_count
!=
266 (trigger_tab
->table_size
- trigger_tab
->header_size
) /
267 sizeof(struct acpi_einj_entry
))
273 static struct acpi_generic_address
*einj_get_trigger_parameter_region(
274 struct acpi_einj_trigger
*trigger_tab
, u64 param1
, u64 param2
)
277 struct acpi_whea_header
*entry
;
279 entry
= (struct acpi_whea_header
*)
280 ((char *)trigger_tab
+ sizeof(struct acpi_einj_trigger
));
281 for (i
= 0; i
< trigger_tab
->entry_count
; i
++) {
282 if (entry
->action
== ACPI_EINJ_TRIGGER_ERROR
&&
283 entry
->instruction
== ACPI_EINJ_WRITE_REGISTER_VALUE
&&
284 entry
->register_region
.space_id
==
285 ACPI_ADR_SPACE_SYSTEM_MEMORY
&&
286 (entry
->register_region
.address
& param2
) == (param1
& param2
))
287 return &entry
->register_region
;
293 /* Execute instructions in trigger error action table */
294 static int __einj_error_trigger(u64 trigger_paddr
, u32 type
,
295 u64 param1
, u64 param2
)
297 struct acpi_einj_trigger
*trigger_tab
= NULL
;
298 struct apei_exec_context trigger_ctx
;
299 struct apei_resources trigger_resources
;
300 struct acpi_whea_header
*trigger_entry
;
304 struct acpi_generic_address
*trigger_param_region
= NULL
;
306 r
= request_mem_region(trigger_paddr
, sizeof(*trigger_tab
),
307 "APEI EINJ Trigger Table");
310 "Can not request [mem %#010llx-%#010llx] for Trigger table\n",
311 (unsigned long long)trigger_paddr
,
312 (unsigned long long)trigger_paddr
+
313 sizeof(*trigger_tab
) - 1);
316 trigger_tab
= ioremap_cache(trigger_paddr
, sizeof(*trigger_tab
));
318 pr_err(EINJ_PFX
"Failed to map trigger table!\n");
321 rc
= einj_check_trigger_header(trigger_tab
);
323 pr_warning(FW_BUG EINJ_PFX
324 "The trigger error action table is invalid\n");
328 /* No action structures in the TRIGGER_ERROR table, nothing to do */
329 if (!trigger_tab
->entry_count
)
333 table_size
= trigger_tab
->table_size
;
334 r
= request_mem_region(trigger_paddr
+ sizeof(*trigger_tab
),
335 table_size
- sizeof(*trigger_tab
),
336 "APEI EINJ Trigger Table");
339 "Can not request [mem %#010llx-%#010llx] for Trigger Table Entry\n",
340 (unsigned long long)trigger_paddr
+ sizeof(*trigger_tab
),
341 (unsigned long long)trigger_paddr
+ table_size
- 1);
344 iounmap(trigger_tab
);
345 trigger_tab
= ioremap_cache(trigger_paddr
, table_size
);
347 pr_err(EINJ_PFX
"Failed to map trigger table!\n");
350 trigger_entry
= (struct acpi_whea_header
*)
351 ((char *)trigger_tab
+ sizeof(struct acpi_einj_trigger
));
352 apei_resources_init(&trigger_resources
);
353 apei_exec_ctx_init(&trigger_ctx
, einj_ins_type
,
354 ARRAY_SIZE(einj_ins_type
),
355 trigger_entry
, trigger_tab
->entry_count
);
356 rc
= apei_exec_collect_resources(&trigger_ctx
, &trigger_resources
);
359 rc
= apei_resources_sub(&trigger_resources
, &einj_resources
);
363 * Some firmware will access target address specified in
364 * param1 to trigger the error when injecting memory error.
365 * This will cause resource conflict with regular memory. So
366 * remove it from trigger table resources.
368 if (param_extension
&& (type
& 0x0038) && param2
) {
369 struct apei_resources addr_resources
;
370 apei_resources_init(&addr_resources
);
371 trigger_param_region
= einj_get_trigger_parameter_region(
372 trigger_tab
, param1
, param2
);
373 if (trigger_param_region
) {
374 rc
= apei_resources_add(&addr_resources
,
375 trigger_param_region
->address
,
376 trigger_param_region
->bit_width
/8, true);
379 rc
= apei_resources_sub(&trigger_resources
,
382 apei_resources_fini(&addr_resources
);
386 rc
= apei_resources_request(&trigger_resources
, "APEI EINJ Trigger");
389 rc
= apei_exec_pre_map_gars(&trigger_ctx
);
393 rc
= apei_exec_run(&trigger_ctx
, ACPI_EINJ_TRIGGER_ERROR
);
395 apei_exec_post_unmap_gars(&trigger_ctx
);
397 apei_resources_release(&trigger_resources
);
399 apei_resources_fini(&trigger_resources
);
401 release_mem_region(trigger_paddr
+ sizeof(*trigger_tab
),
402 table_size
- sizeof(*trigger_tab
));
404 release_mem_region(trigger_paddr
, sizeof(*trigger_tab
));
407 iounmap(trigger_tab
);
412 static int __einj_error_inject(u32 type
, u64 param1
, u64 param2
)
414 struct apei_exec_context ctx
;
415 u64 val
, trigger_paddr
, timeout
= FIRMWARE_TIMEOUT
;
418 einj_exec_ctx_init(&ctx
);
420 rc
= apei_exec_run_optional(&ctx
, ACPI_EINJ_BEGIN_OPERATION
);
423 apei_exec_ctx_set_input(&ctx
, type
);
425 struct set_error_type_with_address
*v5param
= einj_param
;
427 v5param
->type
= type
;
428 if (type
& 0x80000000) {
429 switch (vendor_flags
) {
430 case SETWA_FLAGS_APICID
:
431 v5param
->apicid
= param1
;
433 case SETWA_FLAGS_MEM
:
434 v5param
->memory_address
= param1
;
435 v5param
->memory_address_range
= param2
;
437 case SETWA_FLAGS_PCIE_SBDF
:
438 v5param
->pcie_sbdf
= param1
;
441 v5param
->flags
= vendor_flags
;
444 case ACPI_EINJ_PROCESSOR_CORRECTABLE
:
445 case ACPI_EINJ_PROCESSOR_UNCORRECTABLE
:
446 case ACPI_EINJ_PROCESSOR_FATAL
:
447 v5param
->apicid
= param1
;
448 v5param
->flags
= SETWA_FLAGS_APICID
;
450 case ACPI_EINJ_MEMORY_CORRECTABLE
:
451 case ACPI_EINJ_MEMORY_UNCORRECTABLE
:
452 case ACPI_EINJ_MEMORY_FATAL
:
453 v5param
->memory_address
= param1
;
454 v5param
->memory_address_range
= param2
;
455 v5param
->flags
= SETWA_FLAGS_MEM
;
457 case ACPI_EINJ_PCIX_CORRECTABLE
:
458 case ACPI_EINJ_PCIX_UNCORRECTABLE
:
459 case ACPI_EINJ_PCIX_FATAL
:
460 v5param
->pcie_sbdf
= param1
;
461 v5param
->flags
= SETWA_FLAGS_PCIE_SBDF
;
466 rc
= apei_exec_run(&ctx
, ACPI_EINJ_SET_ERROR_TYPE
);
470 struct einj_parameter
*v4param
= einj_param
;
471 v4param
->param1
= param1
;
472 v4param
->param2
= param2
;
475 rc
= apei_exec_run(&ctx
, ACPI_EINJ_EXECUTE_OPERATION
);
479 rc
= apei_exec_run(&ctx
, ACPI_EINJ_CHECK_BUSY_STATUS
);
482 val
= apei_exec_ctx_get_output(&ctx
);
483 if (!(val
& EINJ_OP_BUSY
))
485 if (einj_timedout(&timeout
))
488 rc
= apei_exec_run(&ctx
, ACPI_EINJ_GET_COMMAND_STATUS
);
491 val
= apei_exec_ctx_get_output(&ctx
);
492 if (val
!= EINJ_STATUS_SUCCESS
)
495 rc
= apei_exec_run(&ctx
, ACPI_EINJ_GET_TRIGGER_TABLE
);
498 trigger_paddr
= apei_exec_ctx_get_output(&ctx
);
499 rc
= __einj_error_trigger(trigger_paddr
, type
, param1
, param2
);
502 rc
= apei_exec_run_optional(&ctx
, ACPI_EINJ_END_OPERATION
);
507 /* Inject the specified hardware error */
508 static int einj_error_inject(u32 type
, u64 param1
, u64 param2
)
512 mutex_lock(&einj_mutex
);
513 rc
= __einj_error_inject(type
, param1
, param2
);
514 mutex_unlock(&einj_mutex
);
519 static u32 error_type
;
520 static u64 error_param1
;
521 static u64 error_param2
;
522 static struct dentry
*einj_debug_dir
;
524 static int available_error_type_show(struct seq_file
*m
, void *v
)
527 u32 available_error_type
= 0;
529 rc
= einj_get_available_error_type(&available_error_type
);
532 if (available_error_type
& 0x0001)
533 seq_printf(m
, "0x00000001\tProcessor Correctable\n");
534 if (available_error_type
& 0x0002)
535 seq_printf(m
, "0x00000002\tProcessor Uncorrectable non-fatal\n");
536 if (available_error_type
& 0x0004)
537 seq_printf(m
, "0x00000004\tProcessor Uncorrectable fatal\n");
538 if (available_error_type
& 0x0008)
539 seq_printf(m
, "0x00000008\tMemory Correctable\n");
540 if (available_error_type
& 0x0010)
541 seq_printf(m
, "0x00000010\tMemory Uncorrectable non-fatal\n");
542 if (available_error_type
& 0x0020)
543 seq_printf(m
, "0x00000020\tMemory Uncorrectable fatal\n");
544 if (available_error_type
& 0x0040)
545 seq_printf(m
, "0x00000040\tPCI Express Correctable\n");
546 if (available_error_type
& 0x0080)
547 seq_printf(m
, "0x00000080\tPCI Express Uncorrectable non-fatal\n");
548 if (available_error_type
& 0x0100)
549 seq_printf(m
, "0x00000100\tPCI Express Uncorrectable fatal\n");
550 if (available_error_type
& 0x0200)
551 seq_printf(m
, "0x00000200\tPlatform Correctable\n");
552 if (available_error_type
& 0x0400)
553 seq_printf(m
, "0x00000400\tPlatform Uncorrectable non-fatal\n");
554 if (available_error_type
& 0x0800)
555 seq_printf(m
, "0x00000800\tPlatform Uncorrectable fatal\n");
560 static int available_error_type_open(struct inode
*inode
, struct file
*file
)
562 return single_open(file
, available_error_type_show
, NULL
);
565 static const struct file_operations available_error_type_fops
= {
566 .open
= available_error_type_open
,
569 .release
= single_release
,
572 static int error_type_get(void *data
, u64
*val
)
579 static int error_type_set(void *data
, u64 val
)
582 u32 available_error_type
= 0;
586 * Vendor defined types have 0x80000000 bit set, and
587 * are not enumerated by ACPI_EINJ_GET_ERROR_TYPE
589 vendor
= val
& 0x80000000;
590 tval
= val
& 0x7fffffff;
592 /* Only one error type can be specified */
593 if (tval
& (tval
- 1))
596 rc
= einj_get_available_error_type(&available_error_type
);
599 if (!(val
& available_error_type
))
607 DEFINE_SIMPLE_ATTRIBUTE(error_type_fops
, error_type_get
,
608 error_type_set
, "0x%llx\n");
610 static int error_inject_set(void *data
, u64 val
)
615 return einj_error_inject(error_type
, error_param1
, error_param2
);
618 DEFINE_SIMPLE_ATTRIBUTE(error_inject_fops
, NULL
,
619 error_inject_set
, "%llu\n");
621 static int einj_check_table(struct acpi_table_einj
*einj_tab
)
623 if ((einj_tab
->header_length
!=
624 (sizeof(struct acpi_table_einj
) - sizeof(einj_tab
->header
)))
625 && (einj_tab
->header_length
!= sizeof(struct acpi_table_einj
)))
627 if (einj_tab
->header
.length
< sizeof(struct acpi_table_einj
))
629 if (einj_tab
->entries
!=
630 (einj_tab
->header
.length
- sizeof(struct acpi_table_einj
)) /
631 sizeof(struct acpi_einj_entry
))
637 static int __init
einj_init(void)
641 struct dentry
*fentry
;
642 struct apei_exec_context ctx
;
647 status
= acpi_get_table(ACPI_SIG_EINJ
, 0,
648 (struct acpi_table_header
**)&einj_tab
);
649 if (status
== AE_NOT_FOUND
)
651 else if (ACPI_FAILURE(status
)) {
652 const char *msg
= acpi_format_exception(status
);
653 pr_err(EINJ_PFX
"Failed to get table, %s\n", msg
);
657 rc
= einj_check_table(einj_tab
);
659 pr_warning(FW_BUG EINJ_PFX
"EINJ table is invalid\n");
664 einj_debug_dir
= debugfs_create_dir("einj", apei_get_debugfs_dir());
667 fentry
= debugfs_create_file("available_error_type", S_IRUSR
,
668 einj_debug_dir
, NULL
,
669 &available_error_type_fops
);
672 fentry
= debugfs_create_file("error_type", S_IRUSR
| S_IWUSR
,
673 einj_debug_dir
, NULL
, &error_type_fops
);
676 fentry
= debugfs_create_file("error_inject", S_IWUSR
,
677 einj_debug_dir
, NULL
, &error_inject_fops
);
681 apei_resources_init(&einj_resources
);
682 einj_exec_ctx_init(&ctx
);
683 rc
= apei_exec_collect_resources(&ctx
, &einj_resources
);
686 rc
= apei_resources_request(&einj_resources
, "APEI EINJ");
689 rc
= apei_exec_pre_map_gars(&ctx
);
693 einj_param
= einj_get_parameter_address();
694 if ((param_extension
|| acpi5
) && einj_param
) {
695 fentry
= debugfs_create_x64("param1", S_IRUSR
| S_IWUSR
,
696 einj_debug_dir
, &error_param1
);
699 fentry
= debugfs_create_x64("param2", S_IRUSR
| S_IWUSR
,
700 einj_debug_dir
, &error_param2
);
706 vendor_blob
.data
= vendor_dev
;
707 vendor_blob
.size
= strlen(vendor_dev
);
708 fentry
= debugfs_create_blob("vendor", S_IRUSR
,
709 einj_debug_dir
, &vendor_blob
);
712 fentry
= debugfs_create_x32("vendor_flags", S_IRUSR
| S_IWUSR
,
713 einj_debug_dir
, &vendor_flags
);
718 pr_info(EINJ_PFX
"Error INJection is initialized.\n");
724 acpi_size size
= (acpi5
) ?
725 sizeof(struct set_error_type_with_address
) :
726 sizeof(struct einj_parameter
);
728 acpi_os_unmap_memory(einj_param
, size
);
730 apei_exec_post_unmap_gars(&ctx
);
732 apei_resources_release(&einj_resources
);
734 apei_resources_fini(&einj_resources
);
736 debugfs_remove_recursive(einj_debug_dir
);
741 static void __exit
einj_exit(void)
743 struct apei_exec_context ctx
;
746 acpi_size size
= (acpi5
) ?
747 sizeof(struct set_error_type_with_address
) :
748 sizeof(struct einj_parameter
);
750 acpi_os_unmap_memory(einj_param
, size
);
752 einj_exec_ctx_init(&ctx
);
753 apei_exec_post_unmap_gars(&ctx
);
754 apei_resources_release(&einj_resources
);
755 apei_resources_fini(&einj_resources
);
756 debugfs_remove_recursive(einj_debug_dir
);
759 module_init(einj_init
);
760 module_exit(einj_exit
);
762 MODULE_AUTHOR("Huang Ying");
763 MODULE_DESCRIPTION("APEI Error INJection support");
764 MODULE_LICENSE("GPL");