]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - drivers/edac/mce_amd_inj.c
Merge tag 'module-builtin_driver-v4.1-rc8' of git://git.kernel.org/pub/scm/linux...
[mirror_ubuntu-bionic-kernel.git] / drivers / edac / mce_amd_inj.c
CommitLineData
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 *
fd19fcd6 9 * Copyright (c) 2010-14: 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>
9cdeb404
BP
20#include <asm/mce.h>
21
47ca08a4 22#include "mce_amd.h"
9cdeb404 23
9cdeb404
BP
24/*
25 * Collect all the MCi_XXX settings
26 */
27static struct mce i_mce;
fd19fcd6 28static struct dentry *dfs_inj;
9cdeb404 29
685d46d7
AG
30static u8 n_banks;
31
0451d14d
AG
32#define MAX_FLAG_OPT_SIZE 3
33
34enum injection_type {
35 SW_INJ = 0, /* SW injection, simply decode the error */
36 HW_INJ, /* Trigger a #MC */
37 N_INJ_TYPES,
38};
39
40static const char * const flags_options[] = {
41 [SW_INJ] = "sw",
42 [HW_INJ] = "hw",
43 NULL
44};
45
46/* Set default injection to SW_INJ */
de277678 47static enum injection_type inj_type = SW_INJ;
0451d14d 48
fd19fcd6
BP
49#define MCE_INJECT_SET(reg) \
50static int inj_##reg##_set(void *data, u64 val) \
9cdeb404 51{ \
fd19fcd6 52 struct mce *m = (struct mce *)data; \
9cdeb404 53 \
fd19fcd6
BP
54 m->reg = val; \
55 return 0; \
9cdeb404
BP
56}
57
fd19fcd6
BP
58MCE_INJECT_SET(status);
59MCE_INJECT_SET(misc);
60MCE_INJECT_SET(addr);
9cdeb404 61
fd19fcd6
BP
62#define MCE_INJECT_GET(reg) \
63static int inj_##reg##_get(void *data, u64 *val) \
9cdeb404 64{ \
fd19fcd6
BP
65 struct mce *m = (struct mce *)data; \
66 \
67 *val = m->reg; \
68 return 0; \
9cdeb404
BP
69}
70
fd19fcd6
BP
71MCE_INJECT_GET(status);
72MCE_INJECT_GET(misc);
73MCE_INJECT_GET(addr);
9cdeb404 74
fd19fcd6
BP
75DEFINE_SIMPLE_ATTRIBUTE(status_fops, inj_status_get, inj_status_set, "%llx\n");
76DEFINE_SIMPLE_ATTRIBUTE(misc_fops, inj_misc_get, inj_misc_set, "%llx\n");
77DEFINE_SIMPLE_ATTRIBUTE(addr_fops, inj_addr_get, inj_addr_set, "%llx\n");
9cdeb404 78
21690934
BP
79/*
80 * Caller needs to be make sure this cpu doesn't disappear
81 * from under us, i.e.: get_cpu/put_cpu.
82 */
83static int toggle_hw_mce_inject(unsigned int cpu, bool enable)
84{
85 u32 l, h;
86 int err;
87
88 err = rdmsr_on_cpu(cpu, MSR_K7_HWCR, &l, &h);
89 if (err) {
90 pr_err("%s: error reading HWCR\n", __func__);
91 return err;
92 }
93
94 enable ? (l |= BIT(18)) : (l &= ~BIT(18));
95
96 err = wrmsr_on_cpu(cpu, MSR_K7_HWCR, l, h);
97 if (err)
98 pr_err("%s: error writing HWCR\n", __func__);
99
100 return err;
101}
102
0451d14d 103static int __set_inj(const char *buf)
b18f3864 104{
0451d14d
AG
105 int i;
106
107 for (i = 0; i < N_INJ_TYPES; i++) {
108 if (!strncmp(flags_options[i], buf, strlen(flags_options[i]))) {
109 inj_type = i;
110 return 0;
111 }
112 }
113 return -EINVAL;
114}
115
116static ssize_t flags_read(struct file *filp, char __user *ubuf,
117 size_t cnt, loff_t *ppos)
118{
119 char buf[MAX_FLAG_OPT_SIZE];
120 int n;
b18f3864 121
0451d14d 122 n = sprintf(buf, "%s\n", flags_options[inj_type]);
b18f3864 123
0451d14d 124 return simple_read_from_buffer(ubuf, cnt, ppos, buf, n);
b18f3864
BP
125}
126
0451d14d
AG
127static ssize_t flags_write(struct file *filp, const char __user *ubuf,
128 size_t cnt, loff_t *ppos)
b18f3864 129{
0451d14d
AG
130 char buf[MAX_FLAG_OPT_SIZE], *__buf;
131 int err;
132 size_t ret;
b18f3864 133
0451d14d
AG
134 if (cnt > MAX_FLAG_OPT_SIZE)
135 cnt = MAX_FLAG_OPT_SIZE;
136
137 ret = cnt;
138
139 if (copy_from_user(&buf, ubuf, cnt))
140 return -EFAULT;
141
142 buf[cnt - 1] = 0;
143
144 /* strip whitespace */
145 __buf = strstrip(buf);
146
147 err = __set_inj(__buf);
148 if (err) {
149 pr_err("%s: Invalid flags value: %s\n", __func__, __buf);
150 return err;
151 }
152
153 *ppos += ret;
154
155 return ret;
b18f3864
BP
156}
157
0451d14d
AG
158static const struct file_operations flags_fops = {
159 .read = flags_read,
160 .write = flags_write,
161 .llseek = generic_file_llseek,
162};
b18f3864
BP
163
164/*
165 * On which CPU to inject?
166 */
167MCE_INJECT_GET(extcpu);
168
169static int inj_extcpu_set(void *data, u64 val)
170{
171 struct mce *m = (struct mce *)data;
172
173 if (val >= nr_cpu_ids || !cpu_online(val)) {
174 pr_err("%s: Invalid CPU: %llu\n", __func__, val);
175 return -EINVAL;
176 }
177 m->extcpu = val;
178 return 0;
179}
180
181DEFINE_SIMPLE_ATTRIBUTE(extcpu_fops, inj_extcpu_get, inj_extcpu_set, "%llu\n");
182
51756a50
BP
183static void trigger_mce(void *info)
184{
185 asm volatile("int $18");
186}
187
188static void do_inject(void)
189{
190 u64 mcg_status = 0;
191 unsigned int cpu = i_mce.extcpu;
192 u8 b = i_mce.bank;
193
cda9459d
BP
194 if (i_mce.misc)
195 i_mce.status |= MCI_STATUS_MISCV;
196
0451d14d 197 if (inj_type == SW_INJ) {
51756a50
BP
198 amd_decode_mce(NULL, 0, &i_mce);
199 return;
200 }
201
51756a50
BP
202 /* prep MCE global settings for the injection */
203 mcg_status = MCG_STATUS_MCIP | MCG_STATUS_EIPV;
204
205 if (!(i_mce.status & MCI_STATUS_PCC))
206 mcg_status |= MCG_STATUS_RIPV;
207
6d1e9bf5
BP
208 get_online_cpus();
209 if (!cpu_online(cpu))
210 goto err;
211
51756a50
BP
212 toggle_hw_mce_inject(cpu, true);
213
214 wrmsr_on_cpu(cpu, MSR_IA32_MCG_STATUS,
215 (u32)mcg_status, (u32)(mcg_status >> 32));
216
217 wrmsr_on_cpu(cpu, MSR_IA32_MCx_STATUS(b),
218 (u32)i_mce.status, (u32)(i_mce.status >> 32));
219
220 wrmsr_on_cpu(cpu, MSR_IA32_MCx_ADDR(b),
221 (u32)i_mce.addr, (u32)(i_mce.addr >> 32));
222
223 wrmsr_on_cpu(cpu, MSR_IA32_MCx_MISC(b),
224 (u32)i_mce.misc, (u32)(i_mce.misc >> 32));
225
226 toggle_hw_mce_inject(cpu, false);
227
228 smp_call_function_single(cpu, trigger_mce, NULL, 0);
229
230err:
231 put_online_cpus();
232
233}
234
9cdeb404
BP
235/*
236 * This denotes into which bank we're injecting and triggers
237 * the injection, at the same time.
238 */
fd19fcd6 239static int inj_bank_set(void *data, u64 val)
9cdeb404 240{
fd19fcd6 241 struct mce *m = (struct mce *)data;
9cdeb404 242
685d46d7
AG
243 if (val >= n_banks) {
244 pr_err("Non-existent MCE bank: %llu\n", val);
245 return -EINVAL;
fd19fcd6 246 }
9cdeb404 247
fd19fcd6 248 m->bank = val;
51756a50 249 do_inject();
9cdeb404 250
fd19fcd6 251 return 0;
9cdeb404
BP
252}
253
e7f2ea1d 254MCE_INJECT_GET(bank);
9cdeb404 255
fd19fcd6
BP
256DEFINE_SIMPLE_ATTRIBUTE(bank_fops, inj_bank_get, inj_bank_set, "%llu\n");
257
99e21fea 258static const char readme_msg[] =
f2f3dca1
BP
259"Description of the files and their usages:\n"
260"\n"
261"Note1: i refers to the bank number below.\n"
262"Note2: See respective BKDGs for the exact bit definitions of the files below\n"
263"as they mirror the hardware registers.\n"
264"\n"
265"status:\t Set MCi_STATUS: the bits in that MSR control the error type and\n"
266"\t attributes of the error which caused the MCE.\n"
267"\n"
268"misc:\t Set MCi_MISC: provide auxiliary info about the error. It is mostly\n"
269"\t used for error thresholding purposes and its validity is indicated by\n"
270"\t MCi_STATUS[MiscV].\n"
271"\n"
272"addr:\t Error address value to be written to MCi_ADDR. Log address information\n"
273"\t associated with the error.\n"
274"\n"
275"cpu:\t The CPU to inject the error on.\n"
276"\n"
277"bank:\t Specify the bank you want to inject the error into: the number of\n"
278"\t banks in a processor varies and is family/model-specific, therefore, the\n"
279"\t supplied value is sanity-checked. Setting the bank value also triggers the\n"
280"\t injection.\n"
281"\n"
282"flags:\t Injection type to be performed. Writing to this file will trigger a\n"
283"\t real machine check, an APIC interrupt or invoke the error decoder routines\n"
284"\t for AMD processors.\n"
285"\n"
286"\t Allowed error injection types:\n"
287"\t - \"sw\": Software error injection. Decode error to a human-readable \n"
288"\t format only. Safe to use.\n"
289"\t - \"hw\": Hardware error injection. Causes the #MC exception handler to \n"
290"\t handle the error. Be warned: might cause system panic if MCi_STATUS[PCC] \n"
291"\t is set. Therefore, consider setting (debugfs_mountpoint)/mce/fake_panic \n"
292"\t before injecting.\n"
293"\n";
99e21fea
AG
294
295static ssize_t
296inj_readme_read(struct file *filp, char __user *ubuf,
297 size_t cnt, loff_t *ppos)
298{
299 return simple_read_from_buffer(ubuf, cnt, ppos,
300 readme_msg, strlen(readme_msg));
301}
302
303static const struct file_operations readme_fops = {
304 .read = inj_readme_read,
305};
306
8c2b117f 307static struct dfs_node {
fd19fcd6
BP
308 char *name;
309 struct dentry *d;
310 const struct file_operations *fops;
4c6034e8 311 umode_t perm;
fd19fcd6 312} dfs_fls[] = {
4c6034e8
AG
313 { .name = "status", .fops = &status_fops, .perm = S_IRUSR | S_IWUSR },
314 { .name = "misc", .fops = &misc_fops, .perm = S_IRUSR | S_IWUSR },
315 { .name = "addr", .fops = &addr_fops, .perm = S_IRUSR | S_IWUSR },
316 { .name = "bank", .fops = &bank_fops, .perm = S_IRUSR | S_IWUSR },
317 { .name = "flags", .fops = &flags_fops, .perm = S_IRUSR | S_IWUSR },
318 { .name = "cpu", .fops = &extcpu_fops, .perm = S_IRUSR | S_IWUSR },
99e21fea 319 { .name = "README", .fops = &readme_fops, .perm = S_IRUSR | S_IRGRP | S_IROTH },
9cdeb404
BP
320};
321
fd19fcd6 322static int __init init_mce_inject(void)
9cdeb404 323{
fd19fcd6 324 int i;
685d46d7
AG
325 u64 cap;
326
327 rdmsrl(MSR_IA32_MCG_CAP, cap);
328 n_banks = cap & MCG_BANKCNT_MASK;
9cdeb404 329
fd19fcd6
BP
330 dfs_inj = debugfs_create_dir("mce-inject", NULL);
331 if (!dfs_inj)
9cdeb404
BP
332 return -EINVAL;
333
fd19fcd6
BP
334 for (i = 0; i < ARRAY_SIZE(dfs_fls); i++) {
335 dfs_fls[i].d = debugfs_create_file(dfs_fls[i].name,
4c6034e8 336 dfs_fls[i].perm,
fd19fcd6
BP
337 dfs_inj,
338 &i_mce,
339 dfs_fls[i].fops);
9cdeb404 340
fd19fcd6
BP
341 if (!dfs_fls[i].d)
342 goto err_dfs_add;
9cdeb404 343 }
fd19fcd6 344
9cdeb404
BP
345 return 0;
346
fd19fcd6 347err_dfs_add:
df4b2a30 348 while (--i >= 0)
fd19fcd6 349 debugfs_remove(dfs_fls[i].d);
9cdeb404 350
fd19fcd6
BP
351 debugfs_remove(dfs_inj);
352 dfs_inj = NULL;
9cdeb404 353
fd19fcd6 354 return -ENOMEM;
9cdeb404
BP
355}
356
fd19fcd6 357static void __exit exit_mce_inject(void)
9cdeb404
BP
358{
359 int i;
360
fd19fcd6
BP
361 for (i = 0; i < ARRAY_SIZE(dfs_fls); i++)
362 debugfs_remove(dfs_fls[i].d);
9cdeb404 363
fd19fcd6 364 memset(&dfs_fls, 0, sizeof(dfs_fls));
9cdeb404 365
fd19fcd6
BP
366 debugfs_remove(dfs_inj);
367 dfs_inj = NULL;
9cdeb404 368}
fd19fcd6
BP
369module_init(init_mce_inject);
370module_exit(exit_mce_inject);
9cdeb404
BP
371
372MODULE_LICENSE("GPL");
43aff26c 373MODULE_AUTHOR("Borislav Petkov <bp@alien8.de>");
9cdeb404 374MODULE_AUTHOR("AMD Inc.");
fd19fcd6 375MODULE_DESCRIPTION("MCE injection facility for RAS testing");