Commit | Line | Data |
---|---|---|
9cdeb404 | 1 | /* |
fd19fcd6 BP |
2 | * A simple MCE injection facility for testing different aspects of the RAS |
3 | * code. This driver should be built as module so that it can be loaded | |
4 | * on production kernels for testing purposes. | |
9cdeb404 BP |
5 | * |
6 | * This file may be distributed under the terms of the GNU General Public | |
7 | * License version 2. | |
8 | * | |
6c36dfe9 | 9 | * Copyright (c) 2010-15: Borislav Petkov <bp@alien8.de> |
9cdeb404 BP |
10 | * Advanced Micro Devices Inc. |
11 | */ | |
12 | ||
13 | #include <linux/kobject.h> | |
fd19fcd6 | 14 | #include <linux/debugfs.h> |
51990e82 | 15 | #include <linux/device.h> |
80a2e2e3 | 16 | #include <linux/module.h> |
51756a50 | 17 | #include <linux/cpu.h> |
0451d14d AG |
18 | #include <linux/string.h> |
19 | #include <linux/uaccess.h> | |
a1300e50 | 20 | |
9cdeb404 | 21 | #include <asm/mce.h> |
a1300e50 | 22 | #include <asm/irq_vectors.h> |
9cdeb404 | 23 | |
6c36dfe9 | 24 | #include "../kernel/cpu/mcheck/mce-internal.h" |
9cdeb404 | 25 | |
9cdeb404 BP |
26 | /* |
27 | * Collect all the MCi_XXX settings | |
28 | */ | |
29 | static struct mce i_mce; | |
fd19fcd6 | 30 | static struct dentry *dfs_inj; |
9cdeb404 | 31 | |
685d46d7 AG |
32 | static u8 n_banks; |
33 | ||
0451d14d AG |
34 | #define MAX_FLAG_OPT_SIZE 3 |
35 | ||
36 | enum injection_type { | |
37 | SW_INJ = 0, /* SW injection, simply decode the error */ | |
38 | HW_INJ, /* Trigger a #MC */ | |
a1300e50 AG |
39 | DFR_INT_INJ, /* Trigger Deferred error interrupt */ |
40 | THR_INT_INJ, /* Trigger threshold interrupt */ | |
0451d14d AG |
41 | N_INJ_TYPES, |
42 | }; | |
43 | ||
44 | static const char * const flags_options[] = { | |
45 | [SW_INJ] = "sw", | |
46 | [HW_INJ] = "hw", | |
a1300e50 AG |
47 | [DFR_INT_INJ] = "df", |
48 | [THR_INT_INJ] = "th", | |
0451d14d AG |
49 | NULL |
50 | }; | |
51 | ||
52 | /* Set default injection to SW_INJ */ | |
de277678 | 53 | static enum injection_type inj_type = SW_INJ; |
0451d14d | 54 | |
fd19fcd6 BP |
55 | #define MCE_INJECT_SET(reg) \ |
56 | static int inj_##reg##_set(void *data, u64 val) \ | |
9cdeb404 | 57 | { \ |
fd19fcd6 | 58 | struct mce *m = (struct mce *)data; \ |
9cdeb404 | 59 | \ |
fd19fcd6 BP |
60 | m->reg = val; \ |
61 | return 0; \ | |
9cdeb404 BP |
62 | } |
63 | ||
fd19fcd6 BP |
64 | MCE_INJECT_SET(status); |
65 | MCE_INJECT_SET(misc); | |
66 | MCE_INJECT_SET(addr); | |
9cdeb404 | 67 | |
fd19fcd6 BP |
68 | #define MCE_INJECT_GET(reg) \ |
69 | static int inj_##reg##_get(void *data, u64 *val) \ | |
9cdeb404 | 70 | { \ |
fd19fcd6 BP |
71 | struct mce *m = (struct mce *)data; \ |
72 | \ | |
73 | *val = m->reg; \ | |
74 | return 0; \ | |
9cdeb404 BP |
75 | } |
76 | ||
fd19fcd6 BP |
77 | MCE_INJECT_GET(status); |
78 | MCE_INJECT_GET(misc); | |
79 | MCE_INJECT_GET(addr); | |
9cdeb404 | 80 | |
fd19fcd6 BP |
81 | DEFINE_SIMPLE_ATTRIBUTE(status_fops, inj_status_get, inj_status_set, "%llx\n"); |
82 | DEFINE_SIMPLE_ATTRIBUTE(misc_fops, inj_misc_get, inj_misc_set, "%llx\n"); | |
83 | DEFINE_SIMPLE_ATTRIBUTE(addr_fops, inj_addr_get, inj_addr_set, "%llx\n"); | |
9cdeb404 | 84 | |
21690934 BP |
85 | /* |
86 | * Caller needs to be make sure this cpu doesn't disappear | |
87 | * from under us, i.e.: get_cpu/put_cpu. | |
88 | */ | |
89 | static int toggle_hw_mce_inject(unsigned int cpu, bool enable) | |
90 | { | |
91 | u32 l, h; | |
92 | int err; | |
93 | ||
94 | err = rdmsr_on_cpu(cpu, MSR_K7_HWCR, &l, &h); | |
95 | if (err) { | |
96 | pr_err("%s: error reading HWCR\n", __func__); | |
97 | return err; | |
98 | } | |
99 | ||
100 | enable ? (l |= BIT(18)) : (l &= ~BIT(18)); | |
101 | ||
102 | err = wrmsr_on_cpu(cpu, MSR_K7_HWCR, l, h); | |
103 | if (err) | |
104 | pr_err("%s: error writing HWCR\n", __func__); | |
105 | ||
106 | return err; | |
107 | } | |
108 | ||
0451d14d | 109 | static int __set_inj(const char *buf) |
b18f3864 | 110 | { |
0451d14d AG |
111 | int i; |
112 | ||
113 | for (i = 0; i < N_INJ_TYPES; i++) { | |
114 | if (!strncmp(flags_options[i], buf, strlen(flags_options[i]))) { | |
115 | inj_type = i; | |
116 | return 0; | |
117 | } | |
118 | } | |
119 | return -EINVAL; | |
120 | } | |
121 | ||
122 | static ssize_t flags_read(struct file *filp, char __user *ubuf, | |
123 | size_t cnt, loff_t *ppos) | |
124 | { | |
125 | char buf[MAX_FLAG_OPT_SIZE]; | |
126 | int n; | |
b18f3864 | 127 | |
0451d14d | 128 | n = sprintf(buf, "%s\n", flags_options[inj_type]); |
b18f3864 | 129 | |
0451d14d | 130 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, n); |
b18f3864 BP |
131 | } |
132 | ||
0451d14d AG |
133 | static ssize_t flags_write(struct file *filp, const char __user *ubuf, |
134 | size_t cnt, loff_t *ppos) | |
b18f3864 | 135 | { |
0451d14d AG |
136 | char buf[MAX_FLAG_OPT_SIZE], *__buf; |
137 | int err; | |
b18f3864 | 138 | |
0451d14d | 139 | if (cnt > MAX_FLAG_OPT_SIZE) |
85c9306d | 140 | return -EINVAL; |
0451d14d AG |
141 | |
142 | if (copy_from_user(&buf, ubuf, cnt)) | |
143 | return -EFAULT; | |
144 | ||
145 | buf[cnt - 1] = 0; | |
146 | ||
147 | /* strip whitespace */ | |
148 | __buf = strstrip(buf); | |
149 | ||
150 | err = __set_inj(__buf); | |
151 | if (err) { | |
152 | pr_err("%s: Invalid flags value: %s\n", __func__, __buf); | |
153 | return err; | |
154 | } | |
155 | ||
85c9306d | 156 | *ppos += cnt; |
0451d14d | 157 | |
85c9306d | 158 | return cnt; |
b18f3864 BP |
159 | } |
160 | ||
0451d14d AG |
161 | static const struct file_operations flags_fops = { |
162 | .read = flags_read, | |
163 | .write = flags_write, | |
164 | .llseek = generic_file_llseek, | |
165 | }; | |
b18f3864 BP |
166 | |
167 | /* | |
168 | * On which CPU to inject? | |
169 | */ | |
170 | MCE_INJECT_GET(extcpu); | |
171 | ||
172 | static int inj_extcpu_set(void *data, u64 val) | |
173 | { | |
174 | struct mce *m = (struct mce *)data; | |
175 | ||
176 | if (val >= nr_cpu_ids || !cpu_online(val)) { | |
177 | pr_err("%s: Invalid CPU: %llu\n", __func__, val); | |
178 | return -EINVAL; | |
179 | } | |
180 | m->extcpu = val; | |
181 | return 0; | |
182 | } | |
183 | ||
184 | DEFINE_SIMPLE_ATTRIBUTE(extcpu_fops, inj_extcpu_get, inj_extcpu_set, "%llu\n"); | |
185 | ||
51756a50 BP |
186 | static void trigger_mce(void *info) |
187 | { | |
188 | asm volatile("int $18"); | |
189 | } | |
190 | ||
a1300e50 AG |
191 | static void trigger_dfr_int(void *info) |
192 | { | |
193 | asm volatile("int %0" :: "i" (DEFERRED_ERROR_VECTOR)); | |
194 | } | |
195 | ||
196 | static void trigger_thr_int(void *info) | |
197 | { | |
198 | asm volatile("int %0" :: "i" (THRESHOLD_APIC_VECTOR)); | |
199 | } | |
200 | ||
51756a50 BP |
201 | static void do_inject(void) |
202 | { | |
203 | u64 mcg_status = 0; | |
204 | unsigned int cpu = i_mce.extcpu; | |
205 | u8 b = i_mce.bank; | |
206 | ||
cda9459d BP |
207 | if (i_mce.misc) |
208 | i_mce.status |= MCI_STATUS_MISCV; | |
209 | ||
0451d14d | 210 | if (inj_type == SW_INJ) { |
6c36dfe9 | 211 | mce_inject_log(&i_mce); |
51756a50 BP |
212 | return; |
213 | } | |
214 | ||
51756a50 BP |
215 | /* prep MCE global settings for the injection */ |
216 | mcg_status = MCG_STATUS_MCIP | MCG_STATUS_EIPV; | |
217 | ||
218 | if (!(i_mce.status & MCI_STATUS_PCC)) | |
219 | mcg_status |= MCG_STATUS_RIPV; | |
220 | ||
a1300e50 AG |
221 | /* |
222 | * Ensure necessary status bits for deferred errors: | |
223 | * - MCx_STATUS[Deferred]: make sure it is a deferred error | |
224 | * - MCx_STATUS[UC] cleared: deferred errors are _not_ UC | |
225 | */ | |
226 | if (inj_type == DFR_INT_INJ) { | |
227 | i_mce.status |= MCI_STATUS_DEFERRED; | |
228 | i_mce.status |= (i_mce.status & ~MCI_STATUS_UC); | |
229 | } | |
230 | ||
6d1e9bf5 BP |
231 | get_online_cpus(); |
232 | if (!cpu_online(cpu)) | |
233 | goto err; | |
234 | ||
51756a50 BP |
235 | toggle_hw_mce_inject(cpu, true); |
236 | ||
237 | wrmsr_on_cpu(cpu, MSR_IA32_MCG_STATUS, | |
238 | (u32)mcg_status, (u32)(mcg_status >> 32)); | |
239 | ||
240 | wrmsr_on_cpu(cpu, MSR_IA32_MCx_STATUS(b), | |
241 | (u32)i_mce.status, (u32)(i_mce.status >> 32)); | |
242 | ||
243 | wrmsr_on_cpu(cpu, MSR_IA32_MCx_ADDR(b), | |
244 | (u32)i_mce.addr, (u32)(i_mce.addr >> 32)); | |
245 | ||
246 | wrmsr_on_cpu(cpu, MSR_IA32_MCx_MISC(b), | |
247 | (u32)i_mce.misc, (u32)(i_mce.misc >> 32)); | |
248 | ||
249 | toggle_hw_mce_inject(cpu, false); | |
250 | ||
a1300e50 AG |
251 | switch (inj_type) { |
252 | case DFR_INT_INJ: | |
253 | smp_call_function_single(cpu, trigger_dfr_int, NULL, 0); | |
254 | break; | |
255 | case THR_INT_INJ: | |
256 | smp_call_function_single(cpu, trigger_thr_int, NULL, 0); | |
257 | break; | |
258 | default: | |
259 | smp_call_function_single(cpu, trigger_mce, NULL, 0); | |
260 | } | |
51756a50 BP |
261 | |
262 | err: | |
263 | put_online_cpus(); | |
264 | ||
265 | } | |
266 | ||
9cdeb404 BP |
267 | /* |
268 | * This denotes into which bank we're injecting and triggers | |
269 | * the injection, at the same time. | |
270 | */ | |
fd19fcd6 | 271 | static int inj_bank_set(void *data, u64 val) |
9cdeb404 | 272 | { |
fd19fcd6 | 273 | struct mce *m = (struct mce *)data; |
9cdeb404 | 274 | |
685d46d7 AG |
275 | if (val >= n_banks) { |
276 | pr_err("Non-existent MCE bank: %llu\n", val); | |
277 | return -EINVAL; | |
fd19fcd6 | 278 | } |
9cdeb404 | 279 | |
fd19fcd6 | 280 | m->bank = val; |
51756a50 | 281 | do_inject(); |
9cdeb404 | 282 | |
fd19fcd6 | 283 | return 0; |
9cdeb404 BP |
284 | } |
285 | ||
e7f2ea1d | 286 | MCE_INJECT_GET(bank); |
9cdeb404 | 287 | |
fd19fcd6 BP |
288 | DEFINE_SIMPLE_ATTRIBUTE(bank_fops, inj_bank_get, inj_bank_set, "%llu\n"); |
289 | ||
99e21fea | 290 | static const char readme_msg[] = |
f2f3dca1 BP |
291 | "Description of the files and their usages:\n" |
292 | "\n" | |
293 | "Note1: i refers to the bank number below.\n" | |
294 | "Note2: See respective BKDGs for the exact bit definitions of the files below\n" | |
295 | "as they mirror the hardware registers.\n" | |
296 | "\n" | |
297 | "status:\t Set MCi_STATUS: the bits in that MSR control the error type and\n" | |
298 | "\t attributes of the error which caused the MCE.\n" | |
299 | "\n" | |
300 | "misc:\t Set MCi_MISC: provide auxiliary info about the error. It is mostly\n" | |
301 | "\t used for error thresholding purposes and its validity is indicated by\n" | |
302 | "\t MCi_STATUS[MiscV].\n" | |
303 | "\n" | |
304 | "addr:\t Error address value to be written to MCi_ADDR. Log address information\n" | |
305 | "\t associated with the error.\n" | |
306 | "\n" | |
307 | "cpu:\t The CPU to inject the error on.\n" | |
308 | "\n" | |
309 | "bank:\t Specify the bank you want to inject the error into: the number of\n" | |
310 | "\t banks in a processor varies and is family/model-specific, therefore, the\n" | |
311 | "\t supplied value is sanity-checked. Setting the bank value also triggers the\n" | |
312 | "\t injection.\n" | |
313 | "\n" | |
314 | "flags:\t Injection type to be performed. Writing to this file will trigger a\n" | |
315 | "\t real machine check, an APIC interrupt or invoke the error decoder routines\n" | |
316 | "\t for AMD processors.\n" | |
317 | "\n" | |
318 | "\t Allowed error injection types:\n" | |
319 | "\t - \"sw\": Software error injection. Decode error to a human-readable \n" | |
320 | "\t format only. Safe to use.\n" | |
321 | "\t - \"hw\": Hardware error injection. Causes the #MC exception handler to \n" | |
322 | "\t handle the error. Be warned: might cause system panic if MCi_STATUS[PCC] \n" | |
323 | "\t is set. Therefore, consider setting (debugfs_mountpoint)/mce/fake_panic \n" | |
324 | "\t before injecting.\n" | |
a1300e50 AG |
325 | "\t - \"df\": Trigger APIC interrupt for Deferred error. Causes deferred \n" |
326 | "\t error APIC interrupt handler to handle the error if the feature is \n" | |
327 | "\t is present in hardware. \n" | |
328 | "\t - \"th\": Trigger APIC interrupt for Threshold errors. Causes threshold \n" | |
329 | "\t APIC interrupt handler to handle the error. \n" | |
f2f3dca1 | 330 | "\n"; |
99e21fea AG |
331 | |
332 | static ssize_t | |
333 | inj_readme_read(struct file *filp, char __user *ubuf, | |
334 | size_t cnt, loff_t *ppos) | |
335 | { | |
336 | return simple_read_from_buffer(ubuf, cnt, ppos, | |
337 | readme_msg, strlen(readme_msg)); | |
338 | } | |
339 | ||
340 | static const struct file_operations readme_fops = { | |
341 | .read = inj_readme_read, | |
342 | }; | |
343 | ||
8c2b117f | 344 | static struct dfs_node { |
fd19fcd6 BP |
345 | char *name; |
346 | struct dentry *d; | |
347 | const struct file_operations *fops; | |
4c6034e8 | 348 | umode_t perm; |
fd19fcd6 | 349 | } dfs_fls[] = { |
4c6034e8 AG |
350 | { .name = "status", .fops = &status_fops, .perm = S_IRUSR | S_IWUSR }, |
351 | { .name = "misc", .fops = &misc_fops, .perm = S_IRUSR | S_IWUSR }, | |
352 | { .name = "addr", .fops = &addr_fops, .perm = S_IRUSR | S_IWUSR }, | |
353 | { .name = "bank", .fops = &bank_fops, .perm = S_IRUSR | S_IWUSR }, | |
354 | { .name = "flags", .fops = &flags_fops, .perm = S_IRUSR | S_IWUSR }, | |
355 | { .name = "cpu", .fops = &extcpu_fops, .perm = S_IRUSR | S_IWUSR }, | |
99e21fea | 356 | { .name = "README", .fops = &readme_fops, .perm = S_IRUSR | S_IRGRP | S_IROTH }, |
9cdeb404 BP |
357 | }; |
358 | ||
fd19fcd6 | 359 | static int __init init_mce_inject(void) |
9cdeb404 | 360 | { |
fd19fcd6 | 361 | int i; |
685d46d7 AG |
362 | u64 cap; |
363 | ||
364 | rdmsrl(MSR_IA32_MCG_CAP, cap); | |
365 | n_banks = cap & MCG_BANKCNT_MASK; | |
9cdeb404 | 366 | |
fd19fcd6 BP |
367 | dfs_inj = debugfs_create_dir("mce-inject", NULL); |
368 | if (!dfs_inj) | |
9cdeb404 BP |
369 | return -EINVAL; |
370 | ||
fd19fcd6 BP |
371 | for (i = 0; i < ARRAY_SIZE(dfs_fls); i++) { |
372 | dfs_fls[i].d = debugfs_create_file(dfs_fls[i].name, | |
4c6034e8 | 373 | dfs_fls[i].perm, |
fd19fcd6 BP |
374 | dfs_inj, |
375 | &i_mce, | |
376 | dfs_fls[i].fops); | |
9cdeb404 | 377 | |
fd19fcd6 BP |
378 | if (!dfs_fls[i].d) |
379 | goto err_dfs_add; | |
9cdeb404 | 380 | } |
fd19fcd6 | 381 | |
9cdeb404 BP |
382 | return 0; |
383 | ||
fd19fcd6 | 384 | err_dfs_add: |
df4b2a30 | 385 | while (--i >= 0) |
fd19fcd6 | 386 | debugfs_remove(dfs_fls[i].d); |
9cdeb404 | 387 | |
fd19fcd6 BP |
388 | debugfs_remove(dfs_inj); |
389 | dfs_inj = NULL; | |
9cdeb404 | 390 | |
fd19fcd6 | 391 | return -ENOMEM; |
9cdeb404 BP |
392 | } |
393 | ||
fd19fcd6 | 394 | static void __exit exit_mce_inject(void) |
9cdeb404 BP |
395 | { |
396 | int i; | |
397 | ||
fd19fcd6 BP |
398 | for (i = 0; i < ARRAY_SIZE(dfs_fls); i++) |
399 | debugfs_remove(dfs_fls[i].d); | |
9cdeb404 | 400 | |
fd19fcd6 | 401 | memset(&dfs_fls, 0, sizeof(dfs_fls)); |
9cdeb404 | 402 | |
fd19fcd6 BP |
403 | debugfs_remove(dfs_inj); |
404 | dfs_inj = NULL; | |
9cdeb404 | 405 | } |
fd19fcd6 BP |
406 | module_init(init_mce_inject); |
407 | module_exit(exit_mce_inject); | |
9cdeb404 BP |
408 | |
409 | MODULE_LICENSE("GPL"); | |
43aff26c | 410 | MODULE_AUTHOR("Borislav Petkov <bp@alien8.de>"); |
9cdeb404 | 411 | MODULE_AUTHOR("AMD Inc."); |
fd19fcd6 | 412 | MODULE_DESCRIPTION("MCE injection facility for RAS testing"); |