blob: 5ab14251839d0f06c395ac0dfc064c166cf3b75a [file] [log] [blame]
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001/*
2 * drivers/pci/pcie/aer/aerdrv_errprint.c
3 *
4 * This file is subject to the terms and conditions of the GNU General Public
5 * License. See the file "COPYING" in the main directory of this archive
6 * for more details.
7 *
8 * Format error messages and print them to console.
9 *
10 * Copyright (C) 2006 Intel Corp.
11 * Tom Long Nguyen (tom.l.nguyen@intel.com)
12 * Zhang Yanmin (yanmin.zhang@intel.com)
13 *
14 */
15
16#include <linux/module.h>
17#include <linux/pci.h>
18#include <linux/kernel.h>
19#include <linux/errno.h>
20#include <linux/pm.h>
21#include <linux/suspend.h>
Huang Yingc413d762011-02-21 13:54:43 +080022#include <linux/cper.h>
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080023
24#include "aerdrv.h"
25
Lance Ortiz1d521002013-01-03 15:34:08 -070026#define CREATE_TRACE_POINTS
27#include <trace/events/ras.h>
28
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080029#define AER_AGENT_RECEIVER 0
30#define AER_AGENT_REQUESTER 1
31#define AER_AGENT_COMPLETER 2
32#define AER_AGENT_TRANSMITTER 3
33
Hidetoshi Setof1585752009-09-07 17:08:59 +090034#define AER_AGENT_REQUESTER_MASK(t) ((t == AER_CORRECTABLE) ? \
35 0 : (PCI_ERR_UNC_COMP_TIME|PCI_ERR_UNC_UNSUP))
36#define AER_AGENT_COMPLETER_MASK(t) ((t == AER_CORRECTABLE) ? \
37 0 : PCI_ERR_UNC_COMP_ABORT)
38#define AER_AGENT_TRANSMITTER_MASK(t) ((t == AER_CORRECTABLE) ? \
39 (PCI_ERR_COR_REP_ROLL|PCI_ERR_COR_REP_TIMER) : 0)
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080040
41#define AER_GET_AGENT(t, e) \
Hidetoshi Setof1585752009-09-07 17:08:59 +090042 ((e & AER_AGENT_COMPLETER_MASK(t)) ? AER_AGENT_COMPLETER : \
43 (e & AER_AGENT_REQUESTER_MASK(t)) ? AER_AGENT_REQUESTER : \
44 (e & AER_AGENT_TRANSMITTER_MASK(t)) ? AER_AGENT_TRANSMITTER : \
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080045 AER_AGENT_RECEIVER)
46
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080047#define AER_PHYSICAL_LAYER_ERROR 0
48#define AER_DATA_LINK_LAYER_ERROR 1
49#define AER_TRANSACTION_LAYER_ERROR 2
50
Hidetoshi Setof1585752009-09-07 17:08:59 +090051#define AER_PHYSICAL_LAYER_ERROR_MASK(t) ((t == AER_CORRECTABLE) ? \
52 PCI_ERR_COR_RCVR : 0)
53#define AER_DATA_LINK_LAYER_ERROR_MASK(t) ((t == AER_CORRECTABLE) ? \
54 (PCI_ERR_COR_BAD_TLP| \
55 PCI_ERR_COR_BAD_DLLP| \
56 PCI_ERR_COR_REP_ROLL| \
57 PCI_ERR_COR_REP_TIMER) : PCI_ERR_UNC_DLP)
58
59#define AER_GET_LAYER_ERROR(t, e) \
60 ((e & AER_PHYSICAL_LAYER_ERROR_MASK(t)) ? AER_PHYSICAL_LAYER_ERROR : \
61 (e & AER_DATA_LINK_LAYER_ERROR_MASK(t)) ? AER_DATA_LINK_LAYER_ERROR : \
62 AER_TRANSACTION_LAYER_ERROR)
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080063
64/*
65 * AER error strings
66 */
Huang Yingb64a4412011-02-21 13:54:42 +080067static const char *aer_error_severity_string[] = {
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080068 "Uncorrected (Non-Fatal)",
69 "Uncorrected (Fatal)",
70 "Corrected"
71};
72
Huang Yingb64a4412011-02-21 13:54:42 +080073static const char *aer_error_layer[] = {
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080074 "Physical Layer",
75 "Data Link Layer",
76 "Transaction Layer"
77};
Huang Yingb64a4412011-02-21 13:54:42 +080078
79static const char *aer_correctable_error_string[] = {
80 "Receiver Error", /* Bit Position 0 */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080081 NULL,
82 NULL,
83 NULL,
84 NULL,
85 NULL,
Huang Yingb64a4412011-02-21 13:54:42 +080086 "Bad TLP", /* Bit Position 6 */
87 "Bad DLLP", /* Bit Position 7 */
88 "RELAY_NUM Rollover", /* Bit Position 8 */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080089 NULL,
90 NULL,
91 NULL,
Huang Yingb64a4412011-02-21 13:54:42 +080092 "Replay Timer Timeout", /* Bit Position 12 */
93 "Advisory Non-Fatal", /* Bit Position 13 */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080094};
95
Huang Yingb64a4412011-02-21 13:54:42 +080096static const char *aer_uncorrectable_error_string[] = {
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080097 NULL,
98 NULL,
99 NULL,
100 NULL,
Huang Yingb64a4412011-02-21 13:54:42 +0800101 "Data Link Protocol", /* Bit Position 4 */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800102 NULL,
103 NULL,
104 NULL,
105 NULL,
106 NULL,
107 NULL,
108 NULL,
Huang Yingb64a4412011-02-21 13:54:42 +0800109 "Poisoned TLP", /* Bit Position 12 */
110 "Flow Control Protocol", /* Bit Position 13 */
111 "Completion Timeout", /* Bit Position 14 */
112 "Completer Abort", /* Bit Position 15 */
113 "Unexpected Completion", /* Bit Position 16 */
114 "Receiver Overflow", /* Bit Position 17 */
115 "Malformed TLP", /* Bit Position 18 */
116 "ECRC", /* Bit Position 19 */
117 "Unsupported Request", /* Bit Position 20 */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800118};
119
Huang Yingb64a4412011-02-21 13:54:42 +0800120static const char *aer_agent_string[] = {
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800121 "Receiver ID",
122 "Requester ID",
123 "Completer ID",
124 "Transmitter ID"
125};
126
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700127static void __aer_print_error(struct pci_dev *dev,
Huang Yingb64a4412011-02-21 13:54:42 +0800128 struct aer_err_info *info)
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800129{
Hidetoshi Seto0d90c3a2009-09-07 17:12:25 +0900130 int i, status;
Huang Yingb64a4412011-02-21 13:54:42 +0800131 const char *errmsg = NULL;
Hidetoshi Seto0d90c3a2009-09-07 17:12:25 +0900132 status = (info->status & ~info->mask);
133
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800134 for (i = 0; i < 32; i++) {
Hidetoshi Seto0d90c3a2009-09-07 17:12:25 +0900135 if (!(status & (1 << i)))
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800136 continue;
137
Hidetoshi Seto24dbb7b2009-09-07 17:11:29 +0900138 if (info->severity == AER_CORRECTABLE)
Huang Yingb64a4412011-02-21 13:54:42 +0800139 errmsg = i < ARRAY_SIZE(aer_correctable_error_string) ?
140 aer_correctable_error_string[i] : NULL;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800141 else
Huang Yingb64a4412011-02-21 13:54:42 +0800142 errmsg = i < ARRAY_SIZE(aer_uncorrectable_error_string) ?
143 aer_uncorrectable_error_string[i] : NULL;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800144
Hidetoshi Seto24dbb7b2009-09-07 17:11:29 +0900145 if (errmsg)
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700146 dev_err(&dev->dev, " [%2d] %-22s%s\n", i, errmsg,
Hidetoshi Seto79e4b892009-09-07 17:16:45 +0900147 info->first_error == i ? " (First)" : "");
Hidetoshi Seto24dbb7b2009-09-07 17:11:29 +0900148 else
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700149 dev_err(&dev->dev, " [%2d] Unknown Error Bit%s\n",
150 i, info->first_error == i ? " (First)" : "");
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800151 }
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800152}
153
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800154void aer_print_error(struct pci_dev *dev, struct aer_err_info *info)
155{
Hidetoshi Seto0d465f22009-09-07 17:10:40 +0900156 int id = ((dev->bus->number << 8) | dev->devfn);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800157
Hidetoshi Setoc9a91882009-09-07 17:07:29 +0900158 if (info->status == 0) {
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700159 dev_err(&dev->dev,
160 "PCIe Bus Error: severity=%s, type=Unaccessible, "
161 "id=%04x(Unregistered Agent ID)\n",
Hidetoshi Seto79e4b892009-09-07 17:16:45 +0900162 aer_error_severity_string[info->severity], id);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800163 } else {
Hidetoshi Seto79e4b892009-09-07 17:16:45 +0900164 int layer, agent;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800165
Hidetoshi Seto79e4b892009-09-07 17:16:45 +0900166 layer = AER_GET_LAYER_ERROR(info->severity, info->status);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800167 agent = AER_GET_AGENT(info->severity, info->status);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800168
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700169 dev_err(&dev->dev,
170 "PCIe Bus Error: severity=%s, type=%s, id=%04x(%s)\n",
171 aer_error_severity_string[info->severity],
Hidetoshi Seto79e4b892009-09-07 17:16:45 +0900172 aer_error_layer[layer], id, aer_agent_string[agent]);
173
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700174 dev_err(&dev->dev,
175 " device [%04x:%04x] error status/mask=%08x/%08x\n",
176 dev->vendor, dev->device,
Huang Yingb64a4412011-02-21 13:54:42 +0800177 info->status, info->mask);
Hidetoshi Seto79e4b892009-09-07 17:16:45 +0900178
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700179 __aer_print_error(dev, info);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800180
Hidetoshi Seto273024d2009-09-07 17:16:20 +0900181 if (info->tlp_header_valid) {
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800182 unsigned char *tlp = (unsigned char *) &info->tlp;
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700183 dev_err(&dev->dev, " TLP Header:"
Hidetoshi Seto79e4b892009-09-07 17:16:45 +0900184 " %02x%02x%02x%02x %02x%02x%02x%02x"
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800185 " %02x%02x%02x%02x %02x%02x%02x%02x\n",
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700186 *(tlp + 3), *(tlp + 2), *(tlp + 1), *tlp,
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800187 *(tlp + 7), *(tlp + 6), *(tlp + 5), *(tlp + 4),
188 *(tlp + 11), *(tlp + 10), *(tlp + 9),
189 *(tlp + 8), *(tlp + 15), *(tlp + 14),
190 *(tlp + 13), *(tlp + 12));
191 }
192 }
Hidetoshi Seto0d465f22009-09-07 17:10:40 +0900193
194 if (info->id && info->error_dev_num > 1 && info->id == id)
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700195 dev_err(&dev->dev,
196 " Error of this Agent(%04x) is reported first\n",
197 id);
Lance Ortiz1d521002013-01-03 15:34:08 -0700198 trace_aer_event(dev_name(&dev->dev), (info->status & ~info->mask),
199 info->severity);
Hidetoshi Seto79e4b892009-09-07 17:16:45 +0900200}
201
202void aer_print_port_info(struct pci_dev *dev, struct aer_err_info *info)
203{
204 dev_info(&dev->dev, "AER: %s%s error received: id=%04x\n",
205 info->multi_error_valid ? "Multiple " : "",
206 aer_error_severity_string[info->severity], info->id);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800207}
Huang Yingc413d762011-02-21 13:54:43 +0800208
209#ifdef CONFIG_ACPI_APEI_PCIEAER
Huang Ying09184722011-05-17 16:08:37 +0800210int cper_severity_to_aer(int cper_severity)
Huang Yingc413d762011-02-21 13:54:43 +0800211{
212 switch (cper_severity) {
213 case CPER_SEV_RECOVERABLE:
214 return AER_NONFATAL;
215 case CPER_SEV_FATAL:
216 return AER_FATAL;
217 default:
218 return AER_CORRECTABLE;
219 }
220}
Huang Ying09184722011-05-17 16:08:37 +0800221EXPORT_SYMBOL_GPL(cper_severity_to_aer);
Huang Yingc413d762011-02-21 13:54:43 +0800222
Lance Ortiz1d521002013-01-03 15:34:08 -0700223void cper_print_aer(const char *prefix, struct pci_dev *dev, int cper_severity,
Huang Yingc413d762011-02-21 13:54:43 +0800224 struct aer_capability_regs *aer)
225{
226 int aer_severity, layer, agent, status_strs_size, tlp_header_valid = 0;
227 u32 status, mask;
228 const char **status_strs;
229
230 aer_severity = cper_severity_to_aer(cper_severity);
231 if (aer_severity == AER_CORRECTABLE) {
232 status = aer->cor_status;
233 mask = aer->cor_mask;
234 status_strs = aer_correctable_error_string;
235 status_strs_size = ARRAY_SIZE(aer_correctable_error_string);
236 } else {
237 status = aer->uncor_status;
238 mask = aer->uncor_mask;
239 status_strs = aer_uncorrectable_error_string;
240 status_strs_size = ARRAY_SIZE(aer_uncorrectable_error_string);
241 tlp_header_valid = status & AER_LOG_TLP_MASKS;
242 }
243 layer = AER_GET_LAYER_ERROR(aer_severity, status);
244 agent = AER_GET_AGENT(aer_severity, status);
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700245 dev_err(&dev->dev, "aer_status: 0x%08x, aer_mask: 0x%08x\n",
246 status, mask);
Huang Yingc413d762011-02-21 13:54:43 +0800247 cper_print_bits(prefix, status, status_strs, status_strs_size);
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700248 dev_err(&dev->dev, "aer_layer=%s, aer_agent=%s\n",
Huang Yingc413d762011-02-21 13:54:43 +0800249 aer_error_layer[layer], aer_agent_string[agent]);
250 if (aer_severity != AER_CORRECTABLE)
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700251 dev_err(&dev->dev, "aer_uncor_severity: 0x%08x\n",
252 aer->uncor_severity);
Huang Yingc413d762011-02-21 13:54:43 +0800253 if (tlp_header_valid) {
254 const unsigned char *tlp;
255 tlp = (const unsigned char *)&aer->header_log;
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700256 dev_err(&dev->dev, "aer_tlp_header:"
Huang Yingc413d762011-02-21 13:54:43 +0800257 " %02x%02x%02x%02x %02x%02x%02x%02x"
258 " %02x%02x%02x%02x %02x%02x%02x%02x\n",
Lance Ortiz2cced2d2013-01-03 15:34:15 -0700259 *(tlp + 3), *(tlp + 2), *(tlp + 1), *tlp,
Huang Yingc413d762011-02-21 13:54:43 +0800260 *(tlp + 7), *(tlp + 6), *(tlp + 5), *(tlp + 4),
261 *(tlp + 11), *(tlp + 10), *(tlp + 9),
262 *(tlp + 8), *(tlp + 15), *(tlp + 14),
263 *(tlp + 13), *(tlp + 12));
264 }
Lance Ortiz1d521002013-01-03 15:34:08 -0700265 trace_aer_event(dev_name(&dev->dev), (status & ~mask),
266 aer_severity);
Huang Yingc413d762011-02-21 13:54:43 +0800267}
268#endif