3 * Copyright 2008, 2009 Intel Corporation.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License
7 * as published by the Free Software Foundation; version 2
12 #include <linux/kernel.h>
13 #include <linux/seq_file.h>
14 #include <linux/init.h>
15 #include <linux/debugfs.h>
18 #include "mce-internal.h"
21 * Grade an mce by severity. In general the most severe ones are processed
22 * first. Since there are quite a lot of combinations test the bits in a
23 * table-driven way. The rules are simply processed in order, first
26 * Note this is only used for machine check exceptions, the corrected
27 * errors use much simpler rules. The exceptions still check for the corrected
28 * errors, but only to leave them alone for the CMCI handler (except for
32 enum context { IN_KERNEL = 1, IN_USER = 2 };
33 enum ser { SER_REQUIRED = 1, NO_SER = 2 };
34 enum exception { EXCP_CONTEXT = 1, NO_EXCP = 2 };
36 static struct severity {
40 unsigned char mcgmask;
43 unsigned char context;
45 unsigned char covered;
48 #define MCESEV(s, m, c...) { .sev = MCE_ ## s ## _SEVERITY, .msg = m, ## c }
49 #define KERNEL .context = IN_KERNEL
50 #define USER .context = IN_USER
51 #define SER .ser = SER_REQUIRED
52 #define NOSER .ser = NO_SER
53 #define EXCP .excp = EXCP_CONTEXT
54 #define NOEXCP .excp = NO_EXCP
55 #define BITCLR(x) .mask = x, .result = 0
56 #define BITSET(x) .mask = x, .result = x
57 #define MCGMASK(x, y) .mcgmask = x, .mcgres = y
58 #define MASK(x, y) .mask = x, .result = y
59 #define MCI_UC_S (MCI_STATUS_UC|MCI_STATUS_S)
60 #define MCI_UC_SAR (MCI_STATUS_UC|MCI_STATUS_S|MCI_STATUS_AR)
61 #define MCI_ADDR (MCI_STATUS_ADDRV|MCI_STATUS_MISCV)
65 BITCLR(MCI_STATUS_VAL)
69 EXCP, BITCLR(MCI_STATUS_EN)
72 PANIC, "Processor context corrupt",
73 BITSET(MCI_STATUS_PCC)
75 /* When MCIP is not set something is very confused */
77 PANIC, "MCIP not set in MCA handler",
78 EXCP, MCGMASK(MCG_STATUS_MCIP, 0)
80 /* Neither return not error IP -- no chance to recover -> PANIC */
82 PANIC, "Neither restart nor error IP",
83 EXCP, MCGMASK(MCG_STATUS_RIPV|MCG_STATUS_EIPV, 0)
86 PANIC, "In kernel and no restart IP",
87 EXCP, KERNEL, MCGMASK(MCG_STATUS_RIPV, 0)
90 DEFERRED, "Deferred error",
91 NOSER, MASK(MCI_STATUS_UC|MCI_STATUS_DEFERRED|MCI_STATUS_POISON, MCI_STATUS_DEFERRED)
94 KEEP, "Corrected error",
95 NOSER, BITCLR(MCI_STATUS_UC)
98 /* ignore OVER for UCNA */
100 UCNA, "Uncorrected no action required",
101 SER, MASK(MCI_UC_SAR, MCI_STATUS_UC)
104 PANIC, "Illegal combination (UCNA with AR=1)",
106 MASK(MCI_STATUS_OVER|MCI_UC_SAR, MCI_STATUS_UC|MCI_STATUS_AR)
109 KEEP, "Non signalled machine check",
110 SER, BITCLR(MCI_STATUS_S)
114 PANIC, "Action required with lost events",
115 SER, BITSET(MCI_STATUS_OVER|MCI_UC_SAR)
118 /* known AR MCACODs: */
119 #ifdef CONFIG_MEMORY_FAILURE
121 KEEP, "Action required but unaffected thread is continuable",
122 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCI_ADDR, MCI_UC_SAR|MCI_ADDR),
123 MCGMASK(MCG_STATUS_RIPV|MCG_STATUS_EIPV, MCG_STATUS_RIPV)
126 AR, "Action required: data load error in a user process",
127 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCI_ADDR|MCACOD, MCI_UC_SAR|MCI_ADDR|MCACOD_DATA),
131 AR, "Action required: instruction fetch error in a user process",
132 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCI_ADDR|MCACOD, MCI_UC_SAR|MCI_ADDR|MCACOD_INSTR),
137 PANIC, "Action required: unknown MCACOD",
138 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR, MCI_UC_SAR)
141 /* known AO MCACODs: */
143 AO, "Action optional: memory scrubbing error",
144 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCACOD_SCRUBMSK, MCI_UC_S|MCACOD_SCRUB)
147 AO, "Action optional: last level cache writeback error",
148 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCACOD, MCI_UC_S|MCACOD_L3WB)
151 SOME, "Action optional: unknown MCACOD",
152 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR, MCI_UC_S)
155 SOME, "Action optional with lost events",
156 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR, MCI_STATUS_OVER|MCI_UC_S)
160 PANIC, "Overflowed uncorrected",
161 BITSET(MCI_STATUS_OVER|MCI_STATUS_UC)
165 BITSET(MCI_STATUS_UC)
170 ) /* always matches. keep at end */
174 * If mcgstatus indicated that ip/cs on the stack were
175 * no good, then "m->cs" will be zero and we will have
176 * to assume the worst case (IN_KERNEL) as we actually
177 * have no idea what we were executing when the machine
179 * If we do have a good "m->cs" (or a faked one in the
180 * case we were executing in VM86 mode) we can use it to
181 * distinguish an exception taken in user from from one
182 * taken in the kernel.
184 static int error_context(struct mce *m)
186 return ((m->cs & 3) == 3) ? IN_USER : IN_KERNEL;
190 * See AMD Error Scope Hierarchy table in a newer BKDG. For example
191 * 49125_15h_Models_30h-3Fh_BKDG.pdf, section "RAS Features"
193 static int mce_severity_amd(struct mce *m, int tolerant, char **msg, bool is_excp)
195 enum context ctx = error_context(m);
197 /* Processor Context Corrupt, no need to fumble too much, die! */
198 if (m->status & MCI_STATUS_PCC)
199 return MCE_PANIC_SEVERITY;
201 if (m->status & MCI_STATUS_UC) {
204 * On older systems where overflow_recov flag is not present, we
205 * should simply panic if an error overflow occurs. If
206 * overflow_recov flag is present and set, then software can try
207 * to at least kill process to prolong system operation.
209 if (mce_flags.overflow_recov) {
210 /* software can try to contain */
211 if (!(m->mcgstatus & MCG_STATUS_RIPV) && (ctx == IN_KERNEL))
212 return MCE_PANIC_SEVERITY;
214 /* kill current process */
215 return MCE_AR_SEVERITY;
217 /* at least one error was not logged */
218 if (m->status & MCI_STATUS_OVER)
219 return MCE_PANIC_SEVERITY;
223 * For any other case, return MCE_UC_SEVERITY so that we log the
224 * error and exit #MC handler.
226 return MCE_UC_SEVERITY;
230 * deferred error: poll handler catches these and adds to mce_ring so
231 * memory-failure can take recovery actions.
233 if (m->status & MCI_STATUS_DEFERRED)
234 return MCE_DEFERRED_SEVERITY;
237 * corrected error: poll handler catches these and passes responsibility
238 * of decoding the error to EDAC
240 return MCE_KEEP_SEVERITY;
243 static int mce_severity_intel(struct mce *m, int tolerant, char **msg, bool is_excp)
245 enum exception excp = (is_excp ? EXCP_CONTEXT : NO_EXCP);
246 enum context ctx = error_context(m);
249 for (s = severities;; s++) {
250 if ((m->status & s->mask) != s->result)
252 if ((m->mcgstatus & s->mcgmask) != s->mcgres)
254 if (s->ser == SER_REQUIRED && !mca_cfg.ser)
256 if (s->ser == NO_SER && mca_cfg.ser)
258 if (s->context && ctx != s->context)
260 if (s->excp && excp != s->excp)
265 if (s->sev >= MCE_UC_SEVERITY && ctx == IN_KERNEL) {
266 if (panic_on_oops || tolerant < 1)
267 return MCE_PANIC_SEVERITY;
273 /* Default to mce_severity_intel */
274 int (*mce_severity)(struct mce *m, int tolerant, char **msg, bool is_excp) =
277 void __init mcheck_vendor_init_severity(void)
279 if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
280 mce_severity = mce_severity_amd;
283 #ifdef CONFIG_DEBUG_FS
284 static void *s_start(struct seq_file *f, loff_t *pos)
286 if (*pos >= ARRAY_SIZE(severities))
288 return &severities[*pos];
291 static void *s_next(struct seq_file *f, void *data, loff_t *pos)
293 if (++(*pos) >= ARRAY_SIZE(severities))
295 return &severities[*pos];
298 static void s_stop(struct seq_file *f, void *data)
302 static int s_show(struct seq_file *f, void *data)
304 struct severity *ser = data;
305 seq_printf(f, "%d\t%s\n", ser->covered, ser->msg);
309 static const struct seq_operations severities_seq_ops = {
316 static int severities_coverage_open(struct inode *inode, struct file *file)
318 return seq_open(file, &severities_seq_ops);
321 static ssize_t severities_coverage_write(struct file *file,
322 const char __user *ubuf,
323 size_t count, loff_t *ppos)
326 for (i = 0; i < ARRAY_SIZE(severities); i++)
327 severities[i].covered = 0;
331 static const struct file_operations severities_coverage_fops = {
332 .open = severities_coverage_open,
333 .release = seq_release,
335 .write = severities_coverage_write,
339 static int __init severities_debugfs_init(void)
341 struct dentry *dmce, *fsev;
343 dmce = mce_get_debugfs_dir();
347 fsev = debugfs_create_file("severities-coverage", 0444, dmce, NULL,
348 &severities_coverage_fops);
357 late_initcall(severities_debugfs_init);
358 #endif /* CONFIG_DEBUG_FS */