blob: a58b63da1a366ea5dddc8af574779ea68773610a [file] [log] [blame]
Huang Yingea8c0712010-05-18 14:35:15 +08001 APEI Error INJection
2 ~~~~~~~~~~~~~~~~~~~~
3
4EINJ provides a hardware error injection mechanism
5It is very useful for debugging and testing of other APEI and RAS features.
6
7To use EINJ, make sure the following are enabled in your kernel
8configuration:
9
10CONFIG_DEBUG_FS
11CONFIG_ACPI_APEI
12CONFIG_ACPI_APEI_EINJ
13
14The user interface of EINJ is debug file system, under the
15directory apei/einj. The following files are provided.
16
17- available_error_type
18 Reading this file returns the error injection capability of the
19 platform, that is, which error types are supported. The error type
20 definition is as follow, the left field is the error type value, the
21 right field is error description.
22
23 0x00000001 Processor Correctable
24 0x00000002 Processor Uncorrectable non-fatal
25 0x00000004 Processor Uncorrectable fatal
26 0x00000008 Memory Correctable
27 0x00000010 Memory Uncorrectable non-fatal
28 0x00000020 Memory Uncorrectable fatal
29 0x00000040 PCI Express Correctable
30 0x00000080 PCI Express Uncorrectable fatal
31 0x00000100 PCI Express Uncorrectable non-fatal
32 0x00000200 Platform Correctable
33 0x00000400 Platform Uncorrectable non-fatal
34 0x00000800 Platform Uncorrectable fatal
35
36 The format of file contents are as above, except there are only the
37 available error type lines.
38
39- error_type
40 This file is used to set the error type value. The error type value
41 is defined in "available_error_type" description.
42
43- error_inject
44 Write any integer to this file to trigger the error
45 injection. Before this, please specify all necessary error
46 parameters.
47
Huang Ying6e320ec2010-05-18 14:35:24 +080048- param1
49 This file is used to set the first error parameter value. Effect of
Chen Gongace36472013-06-06 15:28:11 -070050 parameter depends on error_type specified. For example, if error
51 type is memory related type, the param1 should be a valid physical
52 memory address.
Huang Ying6e320ec2010-05-18 14:35:24 +080053
54- param2
55 This file is used to set the second error parameter value. Effect of
Chen Gongace36472013-06-06 15:28:11 -070056 parameter depends on error_type specified. For example, if error
57 type is memory related type, the param2 should be a physical memory
58 address mask. Linux requires page or narrower granularity, say,
59 0xfffffffffffff000.
Huang Yingc3e60882011-07-20 16:09:29 +080060
Chen Gong6ef19ab2012-03-15 16:53:37 +080061- notrigger
62 The EINJ mechanism is a two step process. First inject the error, then
63 perform some actions to trigger it. Setting "notrigger" to 1 skips the
64 trigger phase, which *may* allow the user to cause the error in some other
65 context by a simple access to the cpu, memory location, or device that is
66 the target of the error injection. Whether this actually works depends
67 on what operations the BIOS actually includes in the trigger phase.
68
Tony Luckc130bd62012-01-17 12:10:16 -080069BIOS versions based in the ACPI 4.0 specification have limited options
70to control where the errors are injected. Your BIOS may support an
71extension (enabled with the param_extension=1 module parameter, or
72boot command line einj.param_extension=1). This allows the address
73and mask for memory injections to be specified by the param1 and
74param2 files in apei/einj.
75
76BIOS versions using the ACPI 5.0 specification have more control over
77the target of the injection. For processor related errors (type 0x1,
780x2 and 0x4) the APICID of the target should be provided using the
79param1 file in apei/einj. For memory errors (type 0x8, 0x10 and 0x20)
80the address is set using param1 with a mask in param2 (0x0 is equivalent
81to all ones). For PCI express errors (type 0x40, 0x80 and 0x100) the
82segment, bus, device and function are specified using param1:
83
84 31 24 23 16 15 11 10 8 7 0
85 +-------------------------------------------------+
86 | segment | bus | device | function | reserved |
87 +-------------------------------------------------+
88
89An ACPI 5.0 BIOS may also allow vendor specific errors to be injected.
90In this case a file named vendor will contain identifying information
91from the BIOS that hopefully will allow an application wishing to use
92the vendor specific extension to tell that they are running on a BIOS
93that supports it. All vendor extensions have the 0x80000000 bit set in
94error_type. A file vendor_flags controls the interpretation of param1
95and param2 (1 = PROCESSOR, 2 = MEMORY, 4 = PCI). See your BIOS vendor
96documentation for details (and expect changes to this API if vendors
97creativity in using this feature expands beyond our expectations).
98
99Example:
100# cd /sys/kernel/debug/apei/einj
101# cat available_error_type # See which errors can be injected
1020x00000002 Processor Uncorrectable non-fatal
1030x00000008 Memory Correctable
1040x00000010 Memory Uncorrectable non-fatal
105# echo 0x12345000 > param1 # Set memory address for injection
106# echo 0xfffffffffffff000 > param2 # Mask - anywhere in this page
107# echo 0x8 > error_type # Choose correctable memory error
108# echo 1 > error_inject # Inject now
109
Huang Ying6e320ec2010-05-18 14:35:24 +0800110
Huang Yingea8c0712010-05-18 14:35:15 +0800111For more information about EINJ, please refer to ACPI specification
Tony Luckc130bd62012-01-17 12:10:16 -0800112version 4.0, section 17.5 and ACPI 5.0, section 18.6.