1 ======== 1 ========
2 hwpoison 2 hwpoison
3 ======== 3 ========
4 4
5 What is hwpoison? 5 What is hwpoison?
6 ================= 6 =================
7 7
8 Upcoming Intel CPUs have support for recoverin 8 Upcoming Intel CPUs have support for recovering from some memory errors
9 (``MCA recovery``). This requires the OS to de 9 (``MCA recovery``). This requires the OS to declare a page "poisoned",
10 kill the processes associated with it and avoi 10 kill the processes associated with it and avoid using it in the future.
11 11
12 This patchkit implements the necessary infrast 12 This patchkit implements the necessary infrastructure in the VM.
13 13
14 To quote the overview comment:: 14 To quote the overview comment::
15 15
16 High level machine check handler. Hand 16 High level machine check handler. Handles pages reported by the
17 hardware as being corrupted usually du 17 hardware as being corrupted usually due to a 2bit ECC memory or cache
18 failure. 18 failure.
19 19
20 This focusses on pages detected as cor 20 This focusses on pages detected as corrupted in the background.
21 When the current CPU tries to consume 21 When the current CPU tries to consume corruption the currently
22 running process can just be killed dir 22 running process can just be killed directly instead. This implies
23 that if the error cannot be handled fo 23 that if the error cannot be handled for some reason it's safe to
24 just ignore it because no corruption h 24 just ignore it because no corruption has been consumed yet. Instead
25 when that happens another machine chec 25 when that happens another machine check will happen.
26 26
27 Handles page cache pages in various st 27 Handles page cache pages in various states. The tricky part
28 here is that we can access any page as 28 here is that we can access any page asynchronous to other VM
29 users, because memory failures could h 29 users, because memory failures could happen anytime and anywhere,
30 possibly violating some of their assum 30 possibly violating some of their assumptions. This is why this code
31 has to be extremely careful. Generally 31 has to be extremely careful. Generally it tries to use normal locking
32 rules, as in get the standard locks, e 32 rules, as in get the standard locks, even if that means the
33 error handling takes potentially a lon 33 error handling takes potentially a long time.
34 34
35 Some of the operations here are somewh 35 Some of the operations here are somewhat inefficient and have non
36 linear algorithmic complexity, because 36 linear algorithmic complexity, because the data structures have not
37 been optimized for this case. This is 37 been optimized for this case. This is in particular the case
38 for the mapping from a vma to a proces 38 for the mapping from a vma to a process. Since this case is expected
39 to be rare we hope we can get away wit 39 to be rare we hope we can get away with this.
40 40
41 The code consists of a the high level handler 41 The code consists of a the high level handler in mm/memory-failure.c,
42 a new page poison bit and various checks in th 42 a new page poison bit and various checks in the VM to handle poisoned
43 pages. 43 pages.
44 44
45 The main target right now is KVM guests, but i 45 The main target right now is KVM guests, but it works for all kinds
46 of applications. KVM support requires a recent 46 of applications. KVM support requires a recent qemu-kvm release.
47 47
48 For the KVM use there was need for a new signa 48 For the KVM use there was need for a new signal type so that
49 KVM can inject the machine check into the gues 49 KVM can inject the machine check into the guest with the proper
50 address. This in theory allows other applicati 50 address. This in theory allows other applications to handle
51 memory failures too. The expectation is that m 51 memory failures too. The expectation is that most applications
52 won't do that, but some very specialized ones 52 won't do that, but some very specialized ones might.
53 53
54 Failure recovery modes 54 Failure recovery modes
55 ====================== 55 ======================
56 56
57 There are two (actually three) modes memory fa 57 There are two (actually three) modes memory failure recovery can be in:
58 58
59 vm.memory_failure_recovery sysctl set to zero: 59 vm.memory_failure_recovery sysctl set to zero:
60 All memory failures cause a panic. Do 60 All memory failures cause a panic. Do not attempt recovery.
61 61
62 early kill 62 early kill
63 (can be controlled globally and per pr 63 (can be controlled globally and per process)
64 Send SIGBUS to the application as soon 64 Send SIGBUS to the application as soon as the error is detected
65 This allows applications who can proce 65 This allows applications who can process memory errors in a gentle
66 way (e.g. drop affected object) 66 way (e.g. drop affected object)
67 This is the mode used by KVM qemu. 67 This is the mode used by KVM qemu.
68 68
69 late kill 69 late kill
70 Send SIGBUS when the application runs 70 Send SIGBUS when the application runs into the corrupted page.
71 This is best for memory error unaware 71 This is best for memory error unaware applications and default
72 Note some pages are always handled as 72 Note some pages are always handled as late kill.
73 73
74 User control 74 User control
75 ============ 75 ============
76 76
77 vm.memory_failure_recovery 77 vm.memory_failure_recovery
78 See sysctl.txt 78 See sysctl.txt
79 79
80 vm.memory_failure_early_kill 80 vm.memory_failure_early_kill
81 Enable early kill mode globally 81 Enable early kill mode globally
82 82
83 PR_MCE_KILL 83 PR_MCE_KILL
84 Set early/late kill mode/revert to sys 84 Set early/late kill mode/revert to system default
85 85
86 arg1: PR_MCE_KILL_CLEAR: 86 arg1: PR_MCE_KILL_CLEAR:
87 Revert to system default 87 Revert to system default
88 arg1: PR_MCE_KILL_SET: 88 arg1: PR_MCE_KILL_SET:
89 arg2 defines thread specific m 89 arg2 defines thread specific mode
90 90
91 PR_MCE_KILL_EARLY: 91 PR_MCE_KILL_EARLY:
92 Early kill 92 Early kill
93 PR_MCE_KILL_LATE: 93 PR_MCE_KILL_LATE:
94 Late kill 94 Late kill
95 PR_MCE_KILL_DEFAULT 95 PR_MCE_KILL_DEFAULT
96 Use system global defa 96 Use system global default
97 97
98 Note that if you want to have a dedica 98 Note that if you want to have a dedicated thread which handles
99 the SIGBUS(BUS_MCEERR_AO) on behalf of 99 the SIGBUS(BUS_MCEERR_AO) on behalf of the process, you should
100 call prctl(PR_MCE_KILL_EARLY) on the d 100 call prctl(PR_MCE_KILL_EARLY) on the designated thread. Otherwise,
101 the SIGBUS is sent to the main thread. 101 the SIGBUS is sent to the main thread.
102 102
103 PR_MCE_KILL_GET 103 PR_MCE_KILL_GET
104 return current mode 104 return current mode
105 105
106 Testing 106 Testing
107 ======= 107 =======
108 108
109 * madvise(MADV_HWPOISON, ....) (as root) - Poi 109 * madvise(MADV_HWPOISON, ....) (as root) - Poison a page in the
110 process for testing 110 process for testing
111 111
112 * hwpoison-inject module through debugfs ``/sy 112 * hwpoison-inject module through debugfs ``/sys/kernel/debug/hwpoison/``
113 113
114 corrupt-pfn 114 corrupt-pfn
115 Inject hwpoison fault at PFN echoed in 115 Inject hwpoison fault at PFN echoed into this file. This does
116 some early filtering to avoid corrupte 116 some early filtering to avoid corrupted unintended pages in test suites.
117 117
118 unpoison-pfn 118 unpoison-pfn
119 Software-unpoison page at PFN echoed i 119 Software-unpoison page at PFN echoed into this file. This way
120 a page can be reused again. This only 120 a page can be reused again. This only works for Linux
121 injected failures, not for real memory 121 injected failures, not for real memory failures. Once any hardware
122 memory failure happens, this feature i 122 memory failure happens, this feature is disabled.
123 123
124 Note these injection interfaces are not stab 124 Note these injection interfaces are not stable and might change between
125 kernel versions 125 kernel versions
126 126
127 corrupt-filter-dev-major, corrupt-filter-dev 127 corrupt-filter-dev-major, corrupt-filter-dev-minor
128 Only handle memory failures to pages a 128 Only handle memory failures to pages associated with the file
129 system defined by block device major/m 129 system defined by block device major/minor. -1U is the
130 wildcard value. This should be only u 130 wildcard value. This should be only used for testing with
131 artificial injection. 131 artificial injection.
132 132
133 corrupt-filter-memcg 133 corrupt-filter-memcg
134 Limit injection to pages owned by memg 134 Limit injection to pages owned by memgroup. Specified by inode
135 number of the memcg. 135 number of the memcg.
136 136
137 Example:: 137 Example::
138 138
139 mkdir /sys/fs/cgroup/mem/hwpoi 139 mkdir /sys/fs/cgroup/mem/hwpoison
140 140
141 usemem -m 100 -s 1000 & 141 usemem -m 100 -s 1000 &
142 echo `jobs -p` > /sys/fs/cgrou 142 echo `jobs -p` > /sys/fs/cgroup/mem/hwpoison/tasks
143 143
144 memcg_ino=$(ls -id /sys/fs/cgr 144 memcg_ino=$(ls -id /sys/fs/cgroup/mem/hwpoison | cut -f1 -d' ')
145 echo $memcg_ino > /debug/hwpoi 145 echo $memcg_ino > /debug/hwpoison/corrupt-filter-memcg
146 146
147 page-types -p `pidof init` - 147 page-types -p `pidof init` --hwpoison # shall do nothing
148 page-types -p `pidof usemem` - 148 page-types -p `pidof usemem` --hwpoison # poison its pages
149 149
150 corrupt-filter-flags-mask, corrupt-filter-fl 150 corrupt-filter-flags-mask, corrupt-filter-flags-value
151 When specified, only poison pages if ( 151 When specified, only poison pages if ((page_flags & mask) ==
152 value). This allows stress testing of 152 value). This allows stress testing of many kinds of
153 pages. The page_flags are the same as 153 pages. The page_flags are the same as in /proc/kpageflags. The
154 flag bits are defined in include/linux 154 flag bits are defined in include/linux/kernel-page-flags.h and
155 documented in Documentation/admin-guid 155 documented in Documentation/admin-guide/mm/pagemap.rst
156 156
157 * Architecture specific MCE injector 157 * Architecture specific MCE injector
158 158
159 x86 has mce-inject, mce-test 159 x86 has mce-inject, mce-test
160 160
161 Some portable hwpoison test programs in mce- 161 Some portable hwpoison test programs in mce-test, see below.
162 162
163 References 163 References
164 ========== 164 ==========
165 165
166 http://halobates.de/mce-lc09-2.pdf 166 http://halobates.de/mce-lc09-2.pdf
167 Overview presentation from LinuxCon 09 167 Overview presentation from LinuxCon 09
168 168
169 git://git.kernel.org/pub/scm/utils/cpu/mce/mce 169 git://git.kernel.org/pub/scm/utils/cpu/mce/mce-test.git
170 Test suite (hwpoison specific portable 170 Test suite (hwpoison specific portable tests in tsrc)
171 171
172 git://git.kernel.org/pub/scm/utils/cpu/mce/mce 172 git://git.kernel.org/pub/scm/utils/cpu/mce/mce-inject.git
173 x86 specific injector 173 x86 specific injector
174 174
175 175
176 Limitations 176 Limitations
177 =========== 177 ===========
178 - Not all page types are supported and never w 178 - Not all page types are supported and never will. Most kernel internal
179 objects cannot be recovered, only LRU pages 179 objects cannot be recovered, only LRU pages for now.
180 180
181 --- 181 ---
182 Andi Kleen, Oct 2009 182 Andi Kleen, Oct 2009
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.