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 };
35 static struct severity {
39 unsigned char mcgmask;
42 unsigned char context;
43 unsigned char covered;
46 #define MCESEV(s, m, c...) { .sev = MCE_ ## s ## _SEVERITY, .msg = m, ## c }
47 #define KERNEL .context = IN_KERNEL
48 #define USER .context = IN_USER
49 #define SER .ser = SER_REQUIRED
50 #define NOSER .ser = NO_SER
51 #define BITCLR(x) .mask = x, .result = 0
52 #define BITSET(x) .mask = x, .result = x
53 #define MCGMASK(x, y) .mcgmask = x, .mcgres = y
54 #define MASK(x, y) .mask = x, .result = y
55 #define MCI_UC_S (MCI_STATUS_UC|MCI_STATUS_S)
56 #define MCI_UC_SAR (MCI_STATUS_UC|MCI_STATUS_S|MCI_STATUS_AR)
57 #define MCI_ADDR (MCI_STATUS_ADDRV|MCI_STATUS_MISCV)
59 /* Architecturally defined codes from SDM Vol. 3B Chapter 15 */
60 #define MCACOD_SCRUB 0x00C0 /* 0xC0-0xCF Memory Scrubbing */
61 #define MCACOD_SCRUBMSK 0xfff0
62 #define MCACOD_L3WB 0x017A /* L3 Explicit Writeback */
63 #define MCACOD_DATA 0x0134 /* Data Load */
64 #define MCACOD_INSTR 0x0150 /* Instruction Fetch */
68 BITCLR(MCI_STATUS_VAL)
75 PANIC, "Processor context corrupt",
76 BITSET(MCI_STATUS_PCC)
78 /* When MCIP is not set something is very confused */
80 PANIC, "MCIP not set in MCA handler",
81 MCGMASK(MCG_STATUS_MCIP, 0)
83 /* Neither return not error IP -- no chance to recover -> PANIC */
85 PANIC, "Neither restart nor error IP",
86 MCGMASK(MCG_STATUS_RIPV|MCG_STATUS_EIPV, 0)
89 PANIC, "In kernel and no restart IP",
90 KERNEL, MCGMASK(MCG_STATUS_RIPV, 0)
93 KEEP, "Corrected error",
94 NOSER, BITCLR(MCI_STATUS_UC)
97 /* ignore OVER for UCNA */
99 KEEP, "Uncorrected no action required",
100 SER, MASK(MCI_UC_SAR, MCI_STATUS_UC)
103 PANIC, "Illegal combination (UCNA with AR=1)",
105 MASK(MCI_STATUS_OVER|MCI_UC_SAR, MCI_STATUS_UC|MCI_STATUS_AR)
108 KEEP, "Non signalled machine check",
109 SER, BITCLR(MCI_STATUS_S)
113 PANIC, "Action required with lost events",
114 SER, BITSET(MCI_STATUS_OVER|MCI_UC_SAR)
117 /* known AR MCACODs: */
118 #ifdef CONFIG_MEMORY_FAILURE
120 KEEP, "HT thread notices Action required: data load error",
121 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCI_ADDR|MCACOD, MCI_UC_SAR|MCI_ADDR|MCACOD_DATA),
122 MCGMASK(MCG_STATUS_EIPV, 0)
125 AR, "Action required: data load error",
126 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCI_ADDR|MCACOD, MCI_UC_SAR|MCI_ADDR|MCACOD_DATA),
131 PANIC, "Action required: unknown MCACOD",
132 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR, MCI_UC_SAR)
135 /* known AO MCACODs: */
137 AO, "Action optional: memory scrubbing error",
138 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCACOD_SCRUBMSK, MCI_UC_S|MCACOD_SCRUB)
141 AO, "Action optional: last level cache writeback error",
142 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR|MCACOD, MCI_UC_S|MCACOD_L3WB)
145 SOME, "Action optional: unknown MCACOD",
146 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR, MCI_UC_S)
149 SOME, "Action optional with lost events",
150 SER, MASK(MCI_STATUS_OVER|MCI_UC_SAR, MCI_STATUS_OVER|MCI_UC_S)
154 PANIC, "Overflowed uncorrected",
155 BITSET(MCI_STATUS_OVER|MCI_STATUS_UC)
159 BITSET(MCI_STATUS_UC)
164 ) /* always matches. keep at end */
168 * If the EIPV bit is set, it means the saved IP is the
169 * instruction which caused the MCE.
171 static int error_context(struct mce *m)
173 if (m->mcgstatus & MCG_STATUS_EIPV)
174 return (m->ip && (m->cs & 3) == 3) ? IN_USER : IN_KERNEL;
175 /* Unknown, assume kernel */
179 int mce_severity(struct mce *m, int tolerant, char **msg)
181 enum context ctx = error_context(m);
184 for (s = severities;; s++) {
185 if ((m->status & s->mask) != s->result)
187 if ((m->mcgstatus & s->mcgmask) != s->mcgres)
189 if (s->ser == SER_REQUIRED && !mce_ser)
191 if (s->ser == NO_SER && mce_ser)
193 if (s->context && ctx != s->context)
198 if (s->sev >= MCE_UC_SEVERITY && ctx == IN_KERNEL) {
199 if (panic_on_oops || tolerant < 1)
200 return MCE_PANIC_SEVERITY;
206 #ifdef CONFIG_DEBUG_FS
207 static void *s_start(struct seq_file *f, loff_t *pos)
209 if (*pos >= ARRAY_SIZE(severities))
211 return &severities[*pos];
214 static void *s_next(struct seq_file *f, void *data, loff_t *pos)
216 if (++(*pos) >= ARRAY_SIZE(severities))
218 return &severities[*pos];
221 static void s_stop(struct seq_file *f, void *data)
225 static int s_show(struct seq_file *f, void *data)
227 struct severity *ser = data;
228 seq_printf(f, "%d\t%s\n", ser->covered, ser->msg);
232 static const struct seq_operations severities_seq_ops = {
239 static int severities_coverage_open(struct inode *inode, struct file *file)
241 return seq_open(file, &severities_seq_ops);
244 static ssize_t severities_coverage_write(struct file *file,
245 const char __user *ubuf,
246 size_t count, loff_t *ppos)
249 for (i = 0; i < ARRAY_SIZE(severities); i++)
250 severities[i].covered = 0;
254 static const struct file_operations severities_coverage_fops = {
255 .open = severities_coverage_open,
256 .release = seq_release,
258 .write = severities_coverage_write,
262 static int __init severities_debugfs_init(void)
264 struct dentry *dmce, *fsev;
266 dmce = mce_get_debugfs_dir();
270 fsev = debugfs_create_file("severities-coverage", 0444, dmce, NULL,
271 &severities_coverage_fops);
280 late_initcall(severities_debugfs_init);
281 #endif /* CONFIG_DEBUG_FS */