blob: 638d0cbc704eca04f125b145dfee46de248fbf4b [file] [log] [blame]
Bjorn Helgaase1e86ee2018-01-26 14:12:23 -06001// SPDX-License-Identifier: GPL-2.0
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08002/*
Bjorn Helgaasdf62ab52018-03-09 16:36:33 -06003 * Implement the AER root port service driver. The driver registers an IRQ
4 * handler. When a root port triggers an AER interrupt, the IRQ handler
5 * collects root port status and schedules work.
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08006 *
7 * Copyright (C) 2006 Intel Corp.
8 * Tom Long Nguyen (tom.l.nguyen@intel.com)
9 * Zhang Yanmin (yanmin.zhang@intel.com)
Bjorn Helgaas41cbc9e2018-06-08 08:40:00 -050010 *
11 * (C) Copyright 2009 Hewlett-Packard Development Company, L.P.
12 * Andrew Patterson <andrew.patterson@hp.com>
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080013 */
14
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -050015#include <linux/cper.h>
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080016#include <linux/pci.h>
Rafael J. Wysocki415e12b2011-01-07 00:55:09 +010017#include <linux/pci-acpi.h>
Alexey Dobriyand43c36d2009-10-07 17:09:06 +040018#include <linux/sched.h>
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080019#include <linux/kernel.h>
20#include <linux/errno.h>
21#include <linux/pm.h>
22#include <linux/init.h>
23#include <linux/interrupt.h>
24#include <linux/delay.h>
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -050025#include <linux/kfifo.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090026#include <linux/slab.h>
Bjorn Helgaas256a4592018-06-08 08:39:45 -050027#include <acpi/apei.h>
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -050028#include <ras/ras_event.h>
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080029
Bjorn Helgaas4696b822018-06-08 08:48:47 -050030#include "../pci.h"
31#include "portdrv.h"
Bjorn Helgaas23e672b2018-06-08 08:41:28 -050032
Keith Busch27c1ce82018-09-18 17:58:43 -060033#define AER_ERROR_SOURCES_MAX 128
Bjorn Helgaas23e672b2018-06-08 08:41:28 -050034
Rajat Jaindb89ccb2018-06-30 15:07:17 -050035#define AER_MAX_TYPEOF_COR_ERRS 16 /* as per PCI_ERR_COR_STATUS */
36#define AER_MAX_TYPEOF_UNCOR_ERRS 26 /* as per PCI_ERR_UNCOR_STATUS*/
37
Bjorn Helgaas23e672b2018-06-08 08:41:28 -050038struct aer_err_source {
39 unsigned int status;
40 unsigned int id;
41};
42
43struct aer_rpc {
44 struct pci_dev *rpd; /* Root Port device */
Keith Busch27c1ce82018-09-18 17:58:43 -060045 DECLARE_KFIFO(aer_fifo, struct aer_err_source, AER_ERROR_SOURCES_MAX);
Bjorn Helgaas23e672b2018-06-08 08:41:28 -050046};
47
Rajat Jaindb89ccb2018-06-30 15:07:17 -050048/* AER stats for the device */
49struct aer_stats {
50
51 /*
52 * Fields for all AER capable devices. They indicate the errors
53 * "as seen by this device". Note that this may mean that if an
54 * end point is causing problems, the AER counters may increment
55 * at its link partner (e.g. root port) because the errors will be
56 * "seen" by the link partner and not the the problematic end point
57 * itself (which may report all counters as 0 as it never saw any
58 * problems).
59 */
60 /* Counters for different type of correctable errors */
61 u64 dev_cor_errs[AER_MAX_TYPEOF_COR_ERRS];
62 /* Counters for different type of fatal uncorrectable errors */
63 u64 dev_fatal_errs[AER_MAX_TYPEOF_UNCOR_ERRS];
64 /* Counters for different type of nonfatal uncorrectable errors */
65 u64 dev_nonfatal_errs[AER_MAX_TYPEOF_UNCOR_ERRS];
66 /* Total number of ERR_COR sent by this device */
67 u64 dev_total_cor_errs;
68 /* Total number of ERR_FATAL sent by this device */
69 u64 dev_total_fatal_errs;
70 /* Total number of ERR_NONFATAL sent by this device */
71 u64 dev_total_nonfatal_errs;
72
73 /*
74 * Fields for Root ports & root complex event collectors only, these
75 * indicate the total number of ERR_COR, ERR_FATAL, and ERR_NONFATAL
76 * messages received by the root port / event collector, INCLUDING the
77 * ones that are generated internally (by the rootport itself)
78 */
79 u64 rootport_total_cor_errs;
80 u64 rootport_total_fatal_errs;
81 u64 rootport_total_nonfatal_errs;
82};
83
Bjorn Helgaas23e672b2018-06-08 08:41:28 -050084#define AER_LOG_TLP_MASKS (PCI_ERR_UNC_POISON_TLP| \
85 PCI_ERR_UNC_ECRC| \
86 PCI_ERR_UNC_UNSUP| \
87 PCI_ERR_UNC_COMP_ABORT| \
88 PCI_ERR_UNC_UNX_COMP| \
89 PCI_ERR_UNC_MALF_TLP)
90
91#define SYSTEM_ERROR_INTR_ON_MESG_MASK (PCI_EXP_RTCTL_SECEE| \
92 PCI_EXP_RTCTL_SENFEE| \
93 PCI_EXP_RTCTL_SEFEE)
94#define ROOT_PORT_INTR_ON_MESG_MASK (PCI_ERR_ROOT_CMD_COR_EN| \
95 PCI_ERR_ROOT_CMD_NONFATAL_EN| \
96 PCI_ERR_ROOT_CMD_FATAL_EN)
97#define ERR_COR_ID(d) (d & 0xffff)
98#define ERR_UNCOR_ID(d) (d >> 16)
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080099
Randy Dunlap7f785762007-10-05 13:17:58 -0700100static int pcie_aer_disable;
101
102void pci_no_aer(void)
103{
Bjorn Helgaas7ece1412016-09-06 16:24:37 -0500104 pcie_aer_disable = 1;
Randy Dunlap7f785762007-10-05 13:17:58 -0700105}
106
Rafael J. Wysockif1a7bfa2010-08-21 01:50:52 +0200107bool pci_aer_available(void)
108{
109 return !pcie_aer_disable && pci_msi_enabled();
110}
111
Bjorn Helgaas41cbc9e2018-06-08 08:40:00 -0500112#ifdef CONFIG_PCIE_ECRC
113
114#define ECRC_POLICY_DEFAULT 0 /* ECRC set by BIOS */
115#define ECRC_POLICY_OFF 1 /* ECRC off for performance */
116#define ECRC_POLICY_ON 2 /* ECRC on for data integrity */
117
118static int ecrc_policy = ECRC_POLICY_DEFAULT;
119
120static const char *ecrc_policy_str[] = {
121 [ECRC_POLICY_DEFAULT] = "bios",
122 [ECRC_POLICY_OFF] = "off",
123 [ECRC_POLICY_ON] = "on"
124};
125
126/**
127 * enable_ercr_checking - enable PCIe ECRC checking for a device
128 * @dev: the PCI device
129 *
130 * Returns 0 on success, or negative on failure.
131 */
132static int enable_ecrc_checking(struct pci_dev *dev)
133{
134 int pos;
135 u32 reg32;
136
137 if (!pci_is_pcie(dev))
138 return -ENODEV;
139
140 pos = dev->aer_cap;
141 if (!pos)
142 return -ENODEV;
143
144 pci_read_config_dword(dev, pos + PCI_ERR_CAP, &reg32);
145 if (reg32 & PCI_ERR_CAP_ECRC_GENC)
146 reg32 |= PCI_ERR_CAP_ECRC_GENE;
147 if (reg32 & PCI_ERR_CAP_ECRC_CHKC)
148 reg32 |= PCI_ERR_CAP_ECRC_CHKE;
149 pci_write_config_dword(dev, pos + PCI_ERR_CAP, reg32);
150
151 return 0;
152}
153
154/**
155 * disable_ercr_checking - disables PCIe ECRC checking for a device
156 * @dev: the PCI device
157 *
158 * Returns 0 on success, or negative on failure.
159 */
160static int disable_ecrc_checking(struct pci_dev *dev)
161{
162 int pos;
163 u32 reg32;
164
165 if (!pci_is_pcie(dev))
166 return -ENODEV;
167
168 pos = dev->aer_cap;
169 if (!pos)
170 return -ENODEV;
171
172 pci_read_config_dword(dev, pos + PCI_ERR_CAP, &reg32);
173 reg32 &= ~(PCI_ERR_CAP_ECRC_GENE | PCI_ERR_CAP_ECRC_CHKE);
174 pci_write_config_dword(dev, pos + PCI_ERR_CAP, reg32);
175
176 return 0;
177}
178
179/**
180 * pcie_set_ecrc_checking - set/unset PCIe ECRC checking for a device based on global policy
181 * @dev: the PCI device
182 */
183void pcie_set_ecrc_checking(struct pci_dev *dev)
184{
185 switch (ecrc_policy) {
186 case ECRC_POLICY_DEFAULT:
187 return;
188 case ECRC_POLICY_OFF:
189 disable_ecrc_checking(dev);
190 break;
191 case ECRC_POLICY_ON:
192 enable_ecrc_checking(dev);
193 break;
194 default:
195 return;
196 }
197}
198
199/**
200 * pcie_ecrc_get_policy - parse kernel command-line ecrc option
201 */
202void pcie_ecrc_get_policy(char *str)
203{
204 int i;
205
206 for (i = 0; i < ARRAY_SIZE(ecrc_policy_str); i++)
207 if (!strncmp(str, ecrc_policy_str[i],
208 strlen(ecrc_policy_str[i])))
209 break;
210 if (i >= ARRAY_SIZE(ecrc_policy_str))
211 return;
212
213 ecrc_policy = i;
214}
215#endif /* CONFIG_PCIE_ECRC */
216
Bjorn Helgaas256a4592018-06-08 08:39:45 -0500217#ifdef CONFIG_ACPI_APEI
218static inline int hest_match_pci(struct acpi_hest_aer_common *p,
219 struct pci_dev *pci)
220{
221 return ACPI_HEST_SEGMENT(p->bus) == pci_domain_nr(pci->bus) &&
222 ACPI_HEST_BUS(p->bus) == pci->bus->number &&
223 p->device == PCI_SLOT(pci->devfn) &&
224 p->function == PCI_FUNC(pci->devfn);
225}
226
227static inline bool hest_match_type(struct acpi_hest_header *hest_hdr,
228 struct pci_dev *dev)
229{
230 u16 hest_type = hest_hdr->type;
231 u8 pcie_type = pci_pcie_type(dev);
232
233 if ((hest_type == ACPI_HEST_TYPE_AER_ROOT_PORT &&
234 pcie_type == PCI_EXP_TYPE_ROOT_PORT) ||
235 (hest_type == ACPI_HEST_TYPE_AER_ENDPOINT &&
236 pcie_type == PCI_EXP_TYPE_ENDPOINT) ||
237 (hest_type == ACPI_HEST_TYPE_AER_BRIDGE &&
238 (dev->class >> 16) == PCI_BASE_CLASS_BRIDGE))
239 return true;
240 return false;
241}
242
243struct aer_hest_parse_info {
244 struct pci_dev *pci_dev;
245 int firmware_first;
246};
247
248static int hest_source_is_pcie_aer(struct acpi_hest_header *hest_hdr)
249{
250 if (hest_hdr->type == ACPI_HEST_TYPE_AER_ROOT_PORT ||
251 hest_hdr->type == ACPI_HEST_TYPE_AER_ENDPOINT ||
252 hest_hdr->type == ACPI_HEST_TYPE_AER_BRIDGE)
253 return 1;
254 return 0;
255}
256
257static int aer_hest_parse(struct acpi_hest_header *hest_hdr, void *data)
258{
259 struct aer_hest_parse_info *info = data;
260 struct acpi_hest_aer_common *p;
261 int ff;
262
263 if (!hest_source_is_pcie_aer(hest_hdr))
264 return 0;
265
266 p = (struct acpi_hest_aer_common *)(hest_hdr + 1);
267 ff = !!(p->flags & ACPI_HEST_FIRMWARE_FIRST);
268
269 /*
270 * If no specific device is supplied, determine whether
271 * FIRMWARE_FIRST is set for *any* PCIe device.
272 */
273 if (!info->pci_dev) {
274 info->firmware_first |= ff;
275 return 0;
276 }
277
278 /* Otherwise, check the specific device */
279 if (p->flags & ACPI_HEST_GLOBAL) {
280 if (hest_match_type(hest_hdr, info->pci_dev))
281 info->firmware_first = ff;
282 } else
283 if (hest_match_pci(p, info->pci_dev))
284 info->firmware_first = ff;
285
286 return 0;
287}
288
289static void aer_set_firmware_first(struct pci_dev *pci_dev)
290{
291 int rc;
292 struct aer_hest_parse_info info = {
293 .pci_dev = pci_dev,
294 .firmware_first = 0,
295 };
296
297 rc = apei_hest_parse(aer_hest_parse, &info);
298
299 if (rc)
300 pci_dev->__aer_firmware_first = 0;
301 else
302 pci_dev->__aer_firmware_first = info.firmware_first;
303 pci_dev->__aer_firmware_first_valid = 1;
304}
305
306int pcie_aer_get_firmware_first(struct pci_dev *dev)
307{
308 if (!pci_is_pcie(dev))
309 return 0;
310
Alexandru Gagniuc7af02fc2018-07-03 18:27:43 -0500311 if (pcie_ports_native)
312 return 0;
313
Bjorn Helgaas256a4592018-06-08 08:39:45 -0500314 if (!dev->__aer_firmware_first_valid)
315 aer_set_firmware_first(dev);
316 return dev->__aer_firmware_first;
317}
318
319static bool aer_firmware_first;
320
321/**
322 * aer_acpi_firmware_first - Check if APEI should control AER.
323 */
324bool aer_acpi_firmware_first(void)
325{
326 static bool parsed = false;
327 struct aer_hest_parse_info info = {
328 .pci_dev = NULL, /* Check all PCIe devices */
329 .firmware_first = 0,
330 };
331
Alexandru Gagniuc7af02fc2018-07-03 18:27:43 -0500332 if (pcie_ports_native)
333 return false;
334
Bjorn Helgaas256a4592018-06-08 08:39:45 -0500335 if (!parsed) {
336 apei_hest_parse(aer_hest_parse, &info);
337 aer_firmware_first = info.firmware_first;
338 parsed = true;
339 }
340 return aer_firmware_first;
341}
342#endif
343
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500344#define PCI_EXP_AER_FLAGS (PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE | \
345 PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE)
346
347int pci_enable_pcie_error_reporting(struct pci_dev *dev)
348{
349 if (pcie_aer_get_firmware_first(dev))
350 return -EIO;
351
352 if (!dev->aer_cap)
353 return -EIO;
354
355 return pcie_capability_set_word(dev, PCI_EXP_DEVCTL, PCI_EXP_AER_FLAGS);
356}
357EXPORT_SYMBOL_GPL(pci_enable_pcie_error_reporting);
358
359int pci_disable_pcie_error_reporting(struct pci_dev *dev)
360{
361 if (pcie_aer_get_firmware_first(dev))
362 return -EIO;
363
364 return pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
365 PCI_EXP_AER_FLAGS);
366}
367EXPORT_SYMBOL_GPL(pci_disable_pcie_error_reporting);
368
Oza Pawandeepec752f52018-07-19 17:58:09 -0500369void pci_aer_clear_device_status(struct pci_dev *dev)
370{
371 u16 sta;
372
373 pcie_capability_read_word(dev, PCI_EXP_DEVSTA, &sta);
374 pcie_capability_write_word(dev, PCI_EXP_DEVSTA, sta);
375}
376
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500377int pci_cleanup_aer_uncorrect_error_status(struct pci_dev *dev)
378{
379 int pos;
Oza Pawandeepe7b0b842018-07-19 17:58:05 -0500380 u32 status, sev;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500381
382 pos = dev->aer_cap;
383 if (!pos)
384 return -EIO;
385
Alexandru Gagniuc45687f92018-07-17 10:31:23 -0500386 if (pcie_aer_get_firmware_first(dev))
387 return -EIO;
388
Oza Pawandeepe7b0b842018-07-19 17:58:05 -0500389 /* Clear status bits for ERR_NONFATAL errors only */
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500390 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
Oza Pawandeepe7b0b842018-07-19 17:58:05 -0500391 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &sev);
392 status &= ~sev;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500393 if (status)
394 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
395
396 return 0;
397}
398EXPORT_SYMBOL_GPL(pci_cleanup_aer_uncorrect_error_status);
399
Bjorn Helgaas7ab92e82018-07-19 17:55:58 -0500400void pci_aer_clear_fatal_status(struct pci_dev *dev)
401{
402 int pos;
403 u32 status, sev;
404
405 pos = dev->aer_cap;
406 if (!pos)
407 return;
408
Alexandru Gagniuc45687f92018-07-17 10:31:23 -0500409 if (pcie_aer_get_firmware_first(dev))
410 return;
411
Bjorn Helgaas7ab92e82018-07-19 17:55:58 -0500412 /* Clear status bits for ERR_FATAL errors only */
413 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
414 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &sev);
415 status &= sev;
416 if (status)
417 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
418}
419
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500420int pci_cleanup_aer_error_status_regs(struct pci_dev *dev)
421{
422 int pos;
423 u32 status;
424 int port_type;
425
426 if (!pci_is_pcie(dev))
427 return -ENODEV;
428
429 pos = dev->aer_cap;
430 if (!pos)
431 return -EIO;
432
Alexandru Gagniuc45687f92018-07-17 10:31:23 -0500433 if (pcie_aer_get_firmware_first(dev))
434 return -EIO;
435
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500436 port_type = pci_pcie_type(dev);
437 if (port_type == PCI_EXP_TYPE_ROOT_PORT) {
438 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, &status);
439 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, status);
440 }
441
442 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
443 pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS, status);
444
445 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
446 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
447
448 return 0;
449}
450
Rajat Jain60ed9822018-06-21 16:48:26 -0700451void pci_aer_init(struct pci_dev *dev)
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500452{
453 dev->aer_cap = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
Rajat Jaindb89ccb2018-06-30 15:07:17 -0500454
455 if (dev->aer_cap)
456 dev->aer_stats = kzalloc(sizeof(struct aer_stats), GFP_KERNEL);
457
Rajat Jain60ed9822018-06-21 16:48:26 -0700458 pci_cleanup_aer_error_status_regs(dev);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500459}
460
Rajat Jaindb89ccb2018-06-30 15:07:17 -0500461void pci_aer_exit(struct pci_dev *dev)
462{
463 kfree(dev->aer_stats);
464 dev->aer_stats = NULL;
465}
466
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500467#define AER_AGENT_RECEIVER 0
468#define AER_AGENT_REQUESTER 1
469#define AER_AGENT_COMPLETER 2
470#define AER_AGENT_TRANSMITTER 3
471
472#define AER_AGENT_REQUESTER_MASK(t) ((t == AER_CORRECTABLE) ? \
473 0 : (PCI_ERR_UNC_COMP_TIME|PCI_ERR_UNC_UNSUP))
474#define AER_AGENT_COMPLETER_MASK(t) ((t == AER_CORRECTABLE) ? \
475 0 : PCI_ERR_UNC_COMP_ABORT)
476#define AER_AGENT_TRANSMITTER_MASK(t) ((t == AER_CORRECTABLE) ? \
477 (PCI_ERR_COR_REP_ROLL|PCI_ERR_COR_REP_TIMER) : 0)
478
479#define AER_GET_AGENT(t, e) \
480 ((e & AER_AGENT_COMPLETER_MASK(t)) ? AER_AGENT_COMPLETER : \
481 (e & AER_AGENT_REQUESTER_MASK(t)) ? AER_AGENT_REQUESTER : \
482 (e & AER_AGENT_TRANSMITTER_MASK(t)) ? AER_AGENT_TRANSMITTER : \
483 AER_AGENT_RECEIVER)
484
485#define AER_PHYSICAL_LAYER_ERROR 0
486#define AER_DATA_LINK_LAYER_ERROR 1
487#define AER_TRANSACTION_LAYER_ERROR 2
488
489#define AER_PHYSICAL_LAYER_ERROR_MASK(t) ((t == AER_CORRECTABLE) ? \
490 PCI_ERR_COR_RCVR : 0)
491#define AER_DATA_LINK_LAYER_ERROR_MASK(t) ((t == AER_CORRECTABLE) ? \
492 (PCI_ERR_COR_BAD_TLP| \
493 PCI_ERR_COR_BAD_DLLP| \
494 PCI_ERR_COR_REP_ROLL| \
495 PCI_ERR_COR_REP_TIMER) : PCI_ERR_UNC_DLP)
496
497#define AER_GET_LAYER_ERROR(t, e) \
498 ((e & AER_PHYSICAL_LAYER_ERROR_MASK(t)) ? AER_PHYSICAL_LAYER_ERROR : \
499 (e & AER_DATA_LINK_LAYER_ERROR_MASK(t)) ? AER_DATA_LINK_LAYER_ERROR : \
500 AER_TRANSACTION_LAYER_ERROR)
501
502/*
503 * AER error strings
504 */
505static const char *aer_error_severity_string[] = {
506 "Uncorrected (Non-Fatal)",
507 "Uncorrected (Fatal)",
508 "Corrected"
509};
510
511static const char *aer_error_layer[] = {
512 "Physical Layer",
513 "Data Link Layer",
514 "Transaction Layer"
515};
516
Rajat Jaindb89ccb2018-06-30 15:07:17 -0500517static const char *aer_correctable_error_string[AER_MAX_TYPEOF_COR_ERRS] = {
Tyler Baicarbd237802018-06-26 11:44:15 -0400518 "RxErr", /* Bit Position 0 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500519 NULL,
520 NULL,
521 NULL,
522 NULL,
523 NULL,
Tyler Baicarbd237802018-06-26 11:44:15 -0400524 "BadTLP", /* Bit Position 6 */
525 "BadDLLP", /* Bit Position 7 */
526 "Rollover", /* Bit Position 8 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500527 NULL,
528 NULL,
529 NULL,
Tyler Baicarbd237802018-06-26 11:44:15 -0400530 "Timeout", /* Bit Position 12 */
531 "NonFatalErr", /* Bit Position 13 */
532 "CorrIntErr", /* Bit Position 14 */
533 "HeaderOF", /* Bit Position 15 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500534};
535
Rajat Jaindb89ccb2018-06-30 15:07:17 -0500536static const char *aer_uncorrectable_error_string[AER_MAX_TYPEOF_UNCOR_ERRS] = {
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500537 "Undefined", /* Bit Position 0 */
538 NULL,
539 NULL,
540 NULL,
Tyler Baicarbd237802018-06-26 11:44:15 -0400541 "DLP", /* Bit Position 4 */
542 "SDES", /* Bit Position 5 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500543 NULL,
544 NULL,
545 NULL,
546 NULL,
547 NULL,
548 NULL,
Tyler Baicarbd237802018-06-26 11:44:15 -0400549 "TLP", /* Bit Position 12 */
550 "FCP", /* Bit Position 13 */
551 "CmpltTO", /* Bit Position 14 */
552 "CmpltAbrt", /* Bit Position 15 */
553 "UnxCmplt", /* Bit Position 16 */
554 "RxOF", /* Bit Position 17 */
555 "MalfTLP", /* Bit Position 18 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500556 "ECRC", /* Bit Position 19 */
Tyler Baicarbd237802018-06-26 11:44:15 -0400557 "UnsupReq", /* Bit Position 20 */
558 "ACSViol", /* Bit Position 21 */
559 "UncorrIntErr", /* Bit Position 22 */
560 "BlockedTLP", /* Bit Position 23 */
561 "AtomicOpBlocked", /* Bit Position 24 */
562 "TLPBlockedErr", /* Bit Position 25 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500563};
564
565static const char *aer_agent_string[] = {
566 "Receiver ID",
567 "Requester ID",
568 "Completer ID",
569 "Transmitter ID"
570};
571
Rajat Jain81aa5202018-06-21 16:48:28 -0700572#define aer_stats_dev_attr(name, stats_array, strings_array, \
573 total_string, total_field) \
574 static ssize_t \
575 name##_show(struct device *dev, struct device_attribute *attr, \
576 char *buf) \
577{ \
578 unsigned int i; \
579 char *str = buf; \
580 struct pci_dev *pdev = to_pci_dev(dev); \
581 u64 *stats = pdev->aer_stats->stats_array; \
582 \
583 for (i = 0; i < ARRAY_SIZE(strings_array); i++) { \
584 if (strings_array[i]) \
585 str += sprintf(str, "%s %llu\n", \
586 strings_array[i], stats[i]); \
587 else if (stats[i]) \
588 str += sprintf(str, #stats_array "_bit[%d] %llu\n",\
589 i, stats[i]); \
590 } \
591 str += sprintf(str, "TOTAL_%s %llu\n", total_string, \
592 pdev->aer_stats->total_field); \
593 return str-buf; \
594} \
595static DEVICE_ATTR_RO(name)
596
597aer_stats_dev_attr(aer_dev_correctable, dev_cor_errs,
598 aer_correctable_error_string, "ERR_COR",
599 dev_total_cor_errs);
600aer_stats_dev_attr(aer_dev_fatal, dev_fatal_errs,
601 aer_uncorrectable_error_string, "ERR_FATAL",
602 dev_total_fatal_errs);
603aer_stats_dev_attr(aer_dev_nonfatal, dev_nonfatal_errs,
604 aer_uncorrectable_error_string, "ERR_NONFATAL",
605 dev_total_nonfatal_errs);
606
Rajat Jain12833012018-06-21 16:48:29 -0700607#define aer_stats_rootport_attr(name, field) \
608 static ssize_t \
609 name##_show(struct device *dev, struct device_attribute *attr, \
610 char *buf) \
611{ \
612 struct pci_dev *pdev = to_pci_dev(dev); \
613 return sprintf(buf, "%llu\n", pdev->aer_stats->field); \
614} \
615static DEVICE_ATTR_RO(name)
616
617aer_stats_rootport_attr(aer_rootport_total_err_cor,
618 rootport_total_cor_errs);
619aer_stats_rootport_attr(aer_rootport_total_err_fatal,
620 rootport_total_fatal_errs);
621aer_stats_rootport_attr(aer_rootport_total_err_nonfatal,
622 rootport_total_nonfatal_errs);
623
Rajat Jain81aa5202018-06-21 16:48:28 -0700624static struct attribute *aer_stats_attrs[] __ro_after_init = {
625 &dev_attr_aer_dev_correctable.attr,
626 &dev_attr_aer_dev_fatal.attr,
627 &dev_attr_aer_dev_nonfatal.attr,
Rajat Jain12833012018-06-21 16:48:29 -0700628 &dev_attr_aer_rootport_total_err_cor.attr,
629 &dev_attr_aer_rootport_total_err_fatal.attr,
630 &dev_attr_aer_rootport_total_err_nonfatal.attr,
Rajat Jain81aa5202018-06-21 16:48:28 -0700631 NULL
632};
633
634static umode_t aer_stats_attrs_are_visible(struct kobject *kobj,
635 struct attribute *a, int n)
636{
637 struct device *dev = kobj_to_dev(kobj);
638 struct pci_dev *pdev = to_pci_dev(dev);
639
640 if (!pdev->aer_stats)
641 return 0;
642
Rajat Jain12833012018-06-21 16:48:29 -0700643 if ((a == &dev_attr_aer_rootport_total_err_cor.attr ||
644 a == &dev_attr_aer_rootport_total_err_fatal.attr ||
645 a == &dev_attr_aer_rootport_total_err_nonfatal.attr) &&
646 pci_pcie_type(pdev) != PCI_EXP_TYPE_ROOT_PORT)
647 return 0;
648
Rajat Jain81aa5202018-06-21 16:48:28 -0700649 return a->mode;
650}
651
652const struct attribute_group aer_stats_attr_group = {
653 .attrs = aer_stats_attrs,
654 .is_visible = aer_stats_attrs_are_visible,
655};
656
657static void pci_dev_aer_stats_incr(struct pci_dev *pdev,
658 struct aer_err_info *info)
659{
660 int status, i, max = -1;
661 u64 *counter = NULL;
662 struct aer_stats *aer_stats = pdev->aer_stats;
663
664 if (!aer_stats)
665 return;
666
667 switch (info->severity) {
668 case AER_CORRECTABLE:
669 aer_stats->dev_total_cor_errs++;
670 counter = &aer_stats->dev_cor_errs[0];
671 max = AER_MAX_TYPEOF_COR_ERRS;
672 break;
673 case AER_NONFATAL:
674 aer_stats->dev_total_nonfatal_errs++;
675 counter = &aer_stats->dev_nonfatal_errs[0];
676 max = AER_MAX_TYPEOF_UNCOR_ERRS;
677 break;
678 case AER_FATAL:
679 aer_stats->dev_total_fatal_errs++;
680 counter = &aer_stats->dev_fatal_errs[0];
681 max = AER_MAX_TYPEOF_UNCOR_ERRS;
682 break;
683 }
684
685 status = (info->status & ~info->mask);
686 for (i = 0; i < max; i++)
687 if (status & (1 << i))
688 counter[i]++;
689}
690
Rajat Jain12833012018-06-21 16:48:29 -0700691static void pci_rootport_aer_stats_incr(struct pci_dev *pdev,
692 struct aer_err_source *e_src)
693{
694 struct aer_stats *aer_stats = pdev->aer_stats;
695
696 if (!aer_stats)
697 return;
698
699 if (e_src->status & PCI_ERR_ROOT_COR_RCV)
700 aer_stats->rootport_total_cor_errs++;
701
702 if (e_src->status & PCI_ERR_ROOT_UNCOR_RCV) {
703 if (e_src->status & PCI_ERR_ROOT_FATAL_RCV)
704 aer_stats->rootport_total_fatal_errs++;
705 else
706 aer_stats->rootport_total_nonfatal_errs++;
707 }
708}
709
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500710static void __print_tlp_header(struct pci_dev *dev,
711 struct aer_header_log_regs *t)
712{
713 pci_err(dev, " TLP Header: %08x %08x %08x %08x\n",
714 t->dw0, t->dw1, t->dw2, t->dw3);
715}
716
717static void __aer_print_error(struct pci_dev *dev,
718 struct aer_err_info *info)
719{
720 int i, status;
721 const char *errmsg = NULL;
722 status = (info->status & ~info->mask);
723
724 for (i = 0; i < 32; i++) {
725 if (!(status & (1 << i)))
726 continue;
727
728 if (info->severity == AER_CORRECTABLE)
729 errmsg = i < ARRAY_SIZE(aer_correctable_error_string) ?
730 aer_correctable_error_string[i] : NULL;
731 else
732 errmsg = i < ARRAY_SIZE(aer_uncorrectable_error_string) ?
733 aer_uncorrectable_error_string[i] : NULL;
734
735 if (errmsg)
736 pci_err(dev, " [%2d] %-22s%s\n", i, errmsg,
737 info->first_error == i ? " (First)" : "");
738 else
739 pci_err(dev, " [%2d] Unknown Error Bit%s\n",
740 i, info->first_error == i ? " (First)" : "");
741 }
Rajat Jain81aa5202018-06-21 16:48:28 -0700742 pci_dev_aer_stats_incr(dev, info);
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500743}
744
Keith Busch1e451162018-07-19 16:16:55 -0500745void aer_print_error(struct pci_dev *dev, struct aer_err_info *info)
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500746{
747 int layer, agent;
748 int id = ((dev->bus->number << 8) | dev->devfn);
749
750 if (!info->status) {
751 pci_err(dev, "PCIe Bus Error: severity=%s, type=Inaccessible, (Unregistered Agent ID)\n",
752 aer_error_severity_string[info->severity]);
753 goto out;
754 }
755
756 layer = AER_GET_LAYER_ERROR(info->severity, info->status);
757 agent = AER_GET_AGENT(info->severity, info->status);
758
759 pci_err(dev, "PCIe Bus Error: severity=%s, type=%s, (%s)\n",
760 aer_error_severity_string[info->severity],
761 aer_error_layer[layer], aer_agent_string[agent]);
762
763 pci_err(dev, " device [%04x:%04x] error status/mask=%08x/%08x\n",
764 dev->vendor, dev->device,
765 info->status, info->mask);
766
767 __aer_print_error(dev, info);
768
769 if (info->tlp_header_valid)
770 __print_tlp_header(dev, &info->tlp);
771
772out:
773 if (info->id && info->error_dev_num > 1 && info->id == id)
774 pci_err(dev, " Error of this Agent is reported first\n");
775
776 trace_aer_event(dev_name(&dev->dev), (info->status & ~info->mask),
777 info->severity, info->tlp_header_valid, &info->tlp);
778}
779
780static void aer_print_port_info(struct pci_dev *dev, struct aer_err_info *info)
781{
782 u8 bus = info->id >> 8;
783 u8 devfn = info->id & 0xff;
784
785 pci_info(dev, "AER: %s%s error received: %04x:%02x:%02x.%d\n",
786 info->multi_error_valid ? "Multiple " : "",
787 aer_error_severity_string[info->severity],
788 pci_domain_nr(dev->bus), bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
789}
790
791#ifdef CONFIG_ACPI_APEI_PCIEAER
792int cper_severity_to_aer(int cper_severity)
793{
794 switch (cper_severity) {
795 case CPER_SEV_RECOVERABLE:
796 return AER_NONFATAL;
797 case CPER_SEV_FATAL:
798 return AER_FATAL;
799 default:
800 return AER_CORRECTABLE;
801 }
802}
803EXPORT_SYMBOL_GPL(cper_severity_to_aer);
804
805void cper_print_aer(struct pci_dev *dev, int aer_severity,
806 struct aer_capability_regs *aer)
807{
808 int layer, agent, tlp_header_valid = 0;
809 u32 status, mask;
810 struct aer_err_info info;
811
812 if (aer_severity == AER_CORRECTABLE) {
813 status = aer->cor_status;
814 mask = aer->cor_mask;
815 } else {
816 status = aer->uncor_status;
817 mask = aer->uncor_mask;
818 tlp_header_valid = status & AER_LOG_TLP_MASKS;
819 }
820
821 layer = AER_GET_LAYER_ERROR(aer_severity, status);
822 agent = AER_GET_AGENT(aer_severity, status);
823
824 memset(&info, 0, sizeof(info));
825 info.severity = aer_severity;
826 info.status = status;
827 info.mask = mask;
828 info.first_error = PCI_ERR_CAP_FEP(aer->cap_control);
829
830 pci_err(dev, "aer_status: 0x%08x, aer_mask: 0x%08x\n", status, mask);
831 __aer_print_error(dev, &info);
832 pci_err(dev, "aer_layer=%s, aer_agent=%s\n",
833 aer_error_layer[layer], aer_agent_string[agent]);
834
835 if (aer_severity != AER_CORRECTABLE)
836 pci_err(dev, "aer_uncor_severity: 0x%08x\n",
837 aer->uncor_severity);
838
839 if (tlp_header_valid)
840 __print_tlp_header(dev, &aer->header_log);
841
842 trace_aer_event(dev_name(&dev->dev), (status & ~mask),
843 aer_severity, tlp_header_valid, &aer->header_log);
844}
845#endif
846
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500847/**
848 * add_error_device - list device to be handled
849 * @e_info: pointer to error info
850 * @dev: pointer to pci_dev to be added
851 */
852static int add_error_device(struct aer_err_info *e_info, struct pci_dev *dev)
853{
854 if (e_info->error_dev_num < AER_MAX_MULTI_ERR_DEVICES) {
Keith Busch60271ab02018-09-20 10:27:09 -0600855 e_info->dev[e_info->error_dev_num] = pci_dev_get(dev);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500856 e_info->error_dev_num++;
857 return 0;
858 }
859 return -ENOSPC;
860}
861
862/**
863 * is_error_source - check whether the device is source of reported error
864 * @dev: pointer to pci_dev to be checked
865 * @e_info: pointer to reported error info
866 */
867static bool is_error_source(struct pci_dev *dev, struct aer_err_info *e_info)
868{
869 int pos;
870 u32 status, mask;
871 u16 reg16;
872
873 /*
874 * When bus id is equal to 0, it might be a bad id
875 * reported by root port.
876 */
877 if ((PCI_BUS_NUM(e_info->id) != 0) &&
878 !(dev->bus->bus_flags & PCI_BUS_FLAGS_NO_AERSID)) {
879 /* Device ID match? */
880 if (e_info->id == ((dev->bus->number << 8) | dev->devfn))
881 return true;
882
883 /* Continue id comparing if there is no multiple error */
884 if (!e_info->multi_error_valid)
885 return false;
886 }
887
888 /*
889 * When either
890 * 1) bus id is equal to 0. Some ports might lose the bus
891 * id of error source id;
892 * 2) bus flag PCI_BUS_FLAGS_NO_AERSID is set
893 * 3) There are multiple errors and prior ID comparing fails;
894 * We check AER status registers to find possible reporter.
895 */
896 if (atomic_read(&dev->enable_cnt) == 0)
897 return false;
898
899 /* Check if AER is enabled */
900 pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &reg16);
901 if (!(reg16 & PCI_EXP_AER_FLAGS))
902 return false;
903
904 pos = dev->aer_cap;
905 if (!pos)
906 return false;
907
908 /* Check if error is recorded */
909 if (e_info->severity == AER_CORRECTABLE) {
910 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
911 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, &mask);
912 } else {
913 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
914 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, &mask);
915 }
916 if (status & ~mask)
917 return true;
918
919 return false;
920}
921
922static int find_device_iter(struct pci_dev *dev, void *data)
923{
924 struct aer_err_info *e_info = (struct aer_err_info *)data;
925
926 if (is_error_source(dev, e_info)) {
927 /* List this device */
928 if (add_error_device(e_info, dev)) {
929 /* We cannot handle more... Stop iteration */
930 /* TODO: Should print error message here? */
931 return 1;
932 }
933
934 /* If there is only a single error, stop iteration */
935 if (!e_info->multi_error_valid)
936 return 1;
937 }
938 return 0;
939}
940
941/**
942 * find_source_device - search through device hierarchy for source device
943 * @parent: pointer to Root Port pci_dev data structure
944 * @e_info: including detailed error information such like id
945 *
946 * Return true if found.
947 *
948 * Invoked by DPC when error is detected at the Root Port.
949 * Caller of this function must set id, severity, and multi_error_valid of
950 * struct aer_err_info pointed by @e_info properly. This function must fill
951 * e_info->error_dev_num and e_info->dev[], based on the given information.
952 */
953static bool find_source_device(struct pci_dev *parent,
954 struct aer_err_info *e_info)
955{
956 struct pci_dev *dev = parent;
957 int result;
958
959 /* Must reset in this function */
960 e_info->error_dev_num = 0;
961
962 /* Is Root Port an agent that sends error message? */
963 result = find_device_iter(dev, e_info);
964 if (result)
965 return true;
966
967 pci_walk_bus(parent->subordinate, find_device_iter, e_info);
968
969 if (!e_info->error_dev_num) {
970 pci_printk(KERN_DEBUG, parent, "can't find device of ID%04x\n",
971 e_info->id);
972 return false;
973 }
974 return true;
975}
976
977/**
978 * handle_error_source - handle logging error into an event log
979 * @dev: pointer to pci_dev data structure of error source device
980 * @info: comprehensive error information
981 *
982 * Invoked when an error being detected by Root Port.
983 */
984static void handle_error_source(struct pci_dev *dev, struct aer_err_info *info)
985{
986 int pos;
987
988 if (info->severity == AER_CORRECTABLE) {
989 /*
990 * Correctable error does not need software intervention.
991 * No need to go through error recovery process.
992 */
993 pos = dev->aer_cap;
994 if (pos)
995 pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS,
996 info->status);
Oza Pawandeep10d790d92018-07-19 17:58:09 -0500997 pci_aer_clear_device_status(dev);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500998 } else if (info->severity == AER_NONFATAL)
Keith Buschbdb5ac852018-09-20 10:27:12 -0600999 pcie_do_recovery(dev, pci_channel_io_normal,
1000 PCIE_PORT_SERVICE_AER);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001001 else if (info->severity == AER_FATAL)
Keith Buschbdb5ac852018-09-20 10:27:12 -06001002 pcie_do_recovery(dev, pci_channel_io_frozen,
1003 PCIE_PORT_SERVICE_AER);
Keith Busch60271ab02018-09-20 10:27:09 -06001004 pci_dev_put(dev);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001005}
1006
1007#ifdef CONFIG_ACPI_APEI_PCIEAER
1008
1009#define AER_RECOVER_RING_ORDER 4
1010#define AER_RECOVER_RING_SIZE (1 << AER_RECOVER_RING_ORDER)
1011
1012struct aer_recover_entry {
1013 u8 bus;
1014 u8 devfn;
1015 u16 domain;
1016 int severity;
1017 struct aer_capability_regs *regs;
1018};
1019
1020static DEFINE_KFIFO(aer_recover_ring, struct aer_recover_entry,
1021 AER_RECOVER_RING_SIZE);
1022
1023static void aer_recover_work_func(struct work_struct *work)
1024{
1025 struct aer_recover_entry entry;
1026 struct pci_dev *pdev;
1027
1028 while (kfifo_get(&aer_recover_ring, &entry)) {
1029 pdev = pci_get_domain_bus_and_slot(entry.domain, entry.bus,
1030 entry.devfn);
1031 if (!pdev) {
1032 pr_err("AER recover: Can not find pci_dev for %04x:%02x:%02x:%x\n",
1033 entry.domain, entry.bus,
1034 PCI_SLOT(entry.devfn), PCI_FUNC(entry.devfn));
1035 continue;
1036 }
1037 cper_print_aer(pdev, entry.severity, entry.regs);
1038 if (entry.severity == AER_NONFATAL)
Keith Buschbdb5ac852018-09-20 10:27:12 -06001039 pcie_do_recovery(pdev, pci_channel_io_normal,
1040 PCIE_PORT_SERVICE_AER);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001041 else if (entry.severity == AER_FATAL)
Keith Buschbdb5ac852018-09-20 10:27:12 -06001042 pcie_do_recovery(pdev, pci_channel_io_frozen,
1043 PCIE_PORT_SERVICE_AER);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001044 pci_dev_put(pdev);
1045 }
1046}
1047
1048/*
1049 * Mutual exclusion for writers of aer_recover_ring, reader side don't
1050 * need lock, because there is only one reader and lock is not needed
1051 * between reader and writer.
1052 */
1053static DEFINE_SPINLOCK(aer_recover_ring_lock);
1054static DECLARE_WORK(aer_recover_work, aer_recover_work_func);
1055
1056void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn,
1057 int severity, struct aer_capability_regs *aer_regs)
1058{
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001059 struct aer_recover_entry entry = {
1060 .bus = bus,
1061 .devfn = devfn,
1062 .domain = domain,
1063 .severity = severity,
1064 .regs = aer_regs,
1065 };
1066
Keith Buschecae65e2018-09-18 17:58:44 -06001067 if (kfifo_in_spinlocked(&aer_recover_ring, &entry, sizeof(entry),
1068 &aer_recover_ring_lock))
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001069 schedule_work(&aer_recover_work);
1070 else
1071 pr_err("AER recover: Buffer overflow when recovering AER for %04x:%02x:%02x:%x\n",
1072 domain, bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001073}
1074EXPORT_SYMBOL_GPL(aer_recover_queue);
1075#endif
1076
1077/**
Keith Busch1e451162018-07-19 16:16:55 -05001078 * aer_get_device_error_info - read error status from dev and store it to info
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001079 * @dev: pointer to the device expected to have a error record
1080 * @info: pointer to structure to store the error record
1081 *
1082 * Return 1 on success, 0 on error.
1083 *
1084 * Note that @info is reused among all error devices. Clear fields properly.
1085 */
Keith Busch1e451162018-07-19 16:16:55 -05001086int aer_get_device_error_info(struct pci_dev *dev, struct aer_err_info *info)
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001087{
1088 int pos, temp;
1089
1090 /* Must reset in this function */
1091 info->status = 0;
1092 info->tlp_header_valid = 0;
1093
1094 pos = dev->aer_cap;
1095
1096 /* The device might not support AER */
1097 if (!pos)
1098 return 0;
1099
1100 if (info->severity == AER_CORRECTABLE) {
1101 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS,
1102 &info->status);
1103 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK,
1104 &info->mask);
1105 if (!(info->status & ~info->mask))
1106 return 0;
Keith Busch9d938ea2018-09-20 10:27:10 -06001107 } else if (pci_pcie_type(dev) == PCI_EXP_TYPE_ROOT_PORT ||
1108 pci_pcie_type(dev) == PCI_EXP_TYPE_DOWNSTREAM ||
1109 info->severity == AER_NONFATAL) {
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001110
1111 /* Link is still healthy for IO reads */
1112 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS,
1113 &info->status);
1114 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK,
1115 &info->mask);
1116 if (!(info->status & ~info->mask))
1117 return 0;
1118
1119 /* Get First Error Pointer */
1120 pci_read_config_dword(dev, pos + PCI_ERR_CAP, &temp);
1121 info->first_error = PCI_ERR_CAP_FEP(temp);
1122
1123 if (info->status & AER_LOG_TLP_MASKS) {
1124 info->tlp_header_valid = 1;
1125 pci_read_config_dword(dev,
1126 pos + PCI_ERR_HEADER_LOG, &info->tlp.dw0);
1127 pci_read_config_dword(dev,
1128 pos + PCI_ERR_HEADER_LOG + 4, &info->tlp.dw1);
1129 pci_read_config_dword(dev,
1130 pos + PCI_ERR_HEADER_LOG + 8, &info->tlp.dw2);
1131 pci_read_config_dword(dev,
1132 pos + PCI_ERR_HEADER_LOG + 12, &info->tlp.dw3);
1133 }
1134 }
1135
1136 return 1;
1137}
1138
1139static inline void aer_process_err_devices(struct aer_err_info *e_info)
1140{
1141 int i;
1142
1143 /* Report all before handle them, not to lost records by reset etc. */
1144 for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
Keith Busch1e451162018-07-19 16:16:55 -05001145 if (aer_get_device_error_info(e_info->dev[i], e_info))
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001146 aer_print_error(e_info->dev[i], e_info);
1147 }
1148 for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
Keith Busch1e451162018-07-19 16:16:55 -05001149 if (aer_get_device_error_info(e_info->dev[i], e_info))
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001150 handle_error_source(e_info->dev[i], e_info);
1151 }
1152}
1153
1154/**
1155 * aer_isr_one_error - consume an error detected by root port
1156 * @rpc: pointer to the root port which holds an error
1157 * @e_src: pointer to an error source
1158 */
1159static void aer_isr_one_error(struct aer_rpc *rpc,
1160 struct aer_err_source *e_src)
1161{
1162 struct pci_dev *pdev = rpc->rpd;
Keith Buschfcd4d362018-09-18 17:58:42 -06001163 struct aer_err_info e_info;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001164
Rajat Jain12833012018-06-21 16:48:29 -07001165 pci_rootport_aer_stats_incr(pdev, e_src);
1166
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001167 /*
1168 * There is a possibility that both correctable error and
1169 * uncorrectable error being logged. Report correctable error first.
1170 */
1171 if (e_src->status & PCI_ERR_ROOT_COR_RCV) {
Keith Buschfcd4d362018-09-18 17:58:42 -06001172 e_info.id = ERR_COR_ID(e_src->id);
1173 e_info.severity = AER_CORRECTABLE;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001174
1175 if (e_src->status & PCI_ERR_ROOT_MULTI_COR_RCV)
Keith Buschfcd4d362018-09-18 17:58:42 -06001176 e_info.multi_error_valid = 1;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001177 else
Keith Buschfcd4d362018-09-18 17:58:42 -06001178 e_info.multi_error_valid = 0;
1179 aer_print_port_info(pdev, &e_info);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001180
Keith Buschfcd4d362018-09-18 17:58:42 -06001181 if (find_source_device(pdev, &e_info))
1182 aer_process_err_devices(&e_info);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001183 }
1184
1185 if (e_src->status & PCI_ERR_ROOT_UNCOR_RCV) {
Keith Buschfcd4d362018-09-18 17:58:42 -06001186 e_info.id = ERR_UNCOR_ID(e_src->id);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001187
1188 if (e_src->status & PCI_ERR_ROOT_FATAL_RCV)
Keith Buschfcd4d362018-09-18 17:58:42 -06001189 e_info.severity = AER_FATAL;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001190 else
Keith Buschfcd4d362018-09-18 17:58:42 -06001191 e_info.severity = AER_NONFATAL;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001192
1193 if (e_src->status & PCI_ERR_ROOT_MULTI_UNCOR_RCV)
Keith Buschfcd4d362018-09-18 17:58:42 -06001194 e_info.multi_error_valid = 1;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001195 else
Keith Buschfcd4d362018-09-18 17:58:42 -06001196 e_info.multi_error_valid = 0;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001197
Keith Buschfcd4d362018-09-18 17:58:42 -06001198 aer_print_port_info(pdev, &e_info);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001199
Keith Buschfcd4d362018-09-18 17:58:42 -06001200 if (find_source_device(pdev, &e_info))
1201 aer_process_err_devices(&e_info);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001202 }
1203}
1204
1205/**
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001206 * aer_isr - consume errors detected by root port
1207 * @work: definition of this work item
1208 *
1209 * Invoked, as DPC, when root port records new detected error
1210 */
Keith Busch6200cc52018-09-18 17:58:46 -06001211static irqreturn_t aer_isr(int irq, void *context)
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001212{
Keith Busch6200cc52018-09-18 17:58:46 -06001213 struct pcie_device *dev = (struct pcie_device *)context;
1214 struct aer_rpc *rpc = get_service_data(dev);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001215 struct aer_err_source uninitialized_var(e_src);
1216
Keith Busch6200cc52018-09-18 17:58:46 -06001217 if (kfifo_is_empty(&rpc->aer_fifo))
1218 return IRQ_NONE;
1219
Keith Busch27c1ce82018-09-18 17:58:43 -06001220 while (kfifo_get(&rpc->aer_fifo, &e_src))
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001221 aer_isr_one_error(rpc, &e_src);
Keith Busch6200cc52018-09-18 17:58:46 -06001222 return IRQ_HANDLED;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001223}
1224
Bjorn Helgaas3c43a642018-06-08 08:31:57 -05001225/**
1226 * aer_irq - Root Port's ISR
1227 * @irq: IRQ assigned to Root Port
1228 * @context: pointer to Root Port data structure
1229 *
1230 * Invoked when Root Port detects AER messages.
1231 */
1232irqreturn_t aer_irq(int irq, void *context)
1233{
Bjorn Helgaas3c43a642018-06-08 08:31:57 -05001234 struct pcie_device *pdev = (struct pcie_device *)context;
1235 struct aer_rpc *rpc = get_service_data(pdev);
Keith Busch27c1ce82018-09-18 17:58:43 -06001236 struct pci_dev *rp = rpc->rpd;
1237 struct aer_err_source e_src = {};
1238 int pos = rp->aer_cap;
Bjorn Helgaas3c43a642018-06-08 08:31:57 -05001239
Keith Busch27c1ce82018-09-18 17:58:43 -06001240 pci_read_config_dword(rp, pos + PCI_ERR_ROOT_STATUS, &e_src.status);
1241 if (!(e_src.status & (PCI_ERR_ROOT_UNCOR_RCV|PCI_ERR_ROOT_COR_RCV)))
Bjorn Helgaas3c43a642018-06-08 08:31:57 -05001242 return IRQ_NONE;
Bjorn Helgaas3c43a642018-06-08 08:31:57 -05001243
Keith Busch27c1ce82018-09-18 17:58:43 -06001244 pci_read_config_dword(rp, pos + PCI_ERR_ROOT_ERR_SRC, &e_src.id);
1245 pci_write_config_dword(rp, pos + PCI_ERR_ROOT_STATUS, e_src.status);
Bjorn Helgaas3c43a642018-06-08 08:31:57 -05001246
Keith Busch27c1ce82018-09-18 17:58:43 -06001247 if (!kfifo_put(&rpc->aer_fifo, e_src))
Bjorn Helgaas3c43a642018-06-08 08:31:57 -05001248 return IRQ_HANDLED;
Bjorn Helgaas3c43a642018-06-08 08:31:57 -05001249
Keith Busch6200cc52018-09-18 17:58:46 -06001250 return IRQ_WAKE_THREAD;
Bjorn Helgaas3c43a642018-06-08 08:31:57 -05001251}
1252EXPORT_SYMBOL_GPL(aer_irq);
1253
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001254static int set_device_error_reporting(struct pci_dev *dev, void *data)
1255{
1256 bool enable = *((bool *)data);
Yijing Wang62f87c02012-07-24 17:20:03 +08001257 int type = pci_pcie_type(dev);
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001258
Yijing Wang62f87c02012-07-24 17:20:03 +08001259 if ((type == PCI_EXP_TYPE_ROOT_PORT) ||
1260 (type == PCI_EXP_TYPE_UPSTREAM) ||
1261 (type == PCI_EXP_TYPE_DOWNSTREAM)) {
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001262 if (enable)
1263 pci_enable_pcie_error_reporting(dev);
1264 else
1265 pci_disable_pcie_error_reporting(dev);
1266 }
1267
1268 if (enable)
1269 pcie_set_ecrc_checking(dev);
1270
1271 return 0;
1272}
1273
1274/**
1275 * set_downstream_devices_error_reporting - enable/disable the error reporting bits on the root port and its downstream ports.
1276 * @dev: pointer to root port's pci_dev data structure
1277 * @enable: true = enable error reporting, false = disable error reporting.
1278 */
1279static void set_downstream_devices_error_reporting(struct pci_dev *dev,
1280 bool enable)
1281{
1282 set_device_error_reporting(dev, &enable);
1283
1284 if (!dev->subordinate)
1285 return;
1286 pci_walk_bus(dev->subordinate, set_device_error_reporting, &enable);
1287}
1288
1289/**
1290 * aer_enable_rootport - enable Root Port's interrupts when receiving messages
1291 * @rpc: pointer to a Root Port data structure
1292 *
1293 * Invoked when PCIe bus loads AER service driver.
1294 */
1295static void aer_enable_rootport(struct aer_rpc *rpc)
1296{
Keith Busche13d17f2018-04-09 16:04:42 -06001297 struct pci_dev *pdev = rpc->rpd;
Jiang Liu43bd4ee2012-07-24 17:20:11 +08001298 int aer_pos;
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001299 u16 reg16;
1300 u32 reg32;
1301
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001302 /* Clear PCIe Capability's Device Status */
Jiang Liu43bd4ee2012-07-24 17:20:11 +08001303 pcie_capability_read_word(pdev, PCI_EXP_DEVSTA, &reg16);
1304 pcie_capability_write_word(pdev, PCI_EXP_DEVSTA, reg16);
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001305
1306 /* Disable system error generation in response to error messages */
Jiang Liu43bd4ee2012-07-24 17:20:11 +08001307 pcie_capability_clear_word(pdev, PCI_EXP_RTCTL,
1308 SYSTEM_ERROR_INTR_ON_MESG_MASK);
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001309
Keith Busch66b80802016-09-27 16:23:34 -04001310 aer_pos = pdev->aer_cap;
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001311 /* Clear error status */
1312 pci_read_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, &reg32);
1313 pci_write_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, reg32);
1314 pci_read_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, &reg32);
1315 pci_write_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, reg32);
1316 pci_read_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, &reg32);
1317 pci_write_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, reg32);
1318
1319 /*
1320 * Enable error reporting for the root port device and downstream port
1321 * devices.
1322 */
1323 set_downstream_devices_error_reporting(pdev, true);
1324
1325 /* Enable Root Port's interrupt in response to error messages */
1326 pci_read_config_dword(pdev, aer_pos + PCI_ERR_ROOT_COMMAND, &reg32);
1327 reg32 |= ROOT_PORT_INTR_ON_MESG_MASK;
1328 pci_write_config_dword(pdev, aer_pos + PCI_ERR_ROOT_COMMAND, reg32);
1329}
1330
1331/**
1332 * aer_disable_rootport - disable Root Port's interrupts when receiving messages
1333 * @rpc: pointer to a Root Port data structure
1334 *
1335 * Invoked when PCIe bus unloads AER service driver.
1336 */
1337static void aer_disable_rootport(struct aer_rpc *rpc)
1338{
Keith Busche13d17f2018-04-09 16:04:42 -06001339 struct pci_dev *pdev = rpc->rpd;
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001340 u32 reg32;
1341 int pos;
1342
1343 /*
1344 * Disable error reporting for the root port device and downstream port
1345 * devices.
1346 */
1347 set_downstream_devices_error_reporting(pdev, false);
1348
Keith Busch66b80802016-09-27 16:23:34 -04001349 pos = pdev->aer_cap;
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001350 /* Disable Root's interrupt in response to error messages */
1351 pci_read_config_dword(pdev, pos + PCI_ERR_ROOT_COMMAND, &reg32);
1352 reg32 &= ~ROOT_PORT_INTR_ON_MESG_MASK;
1353 pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_COMMAND, reg32);
1354
1355 /* Clear Root's error status reg */
1356 pci_read_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, &reg32);
1357 pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, reg32);
1358}
1359
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001360/**
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001361 * aer_remove - clean up resources
1362 * @dev: pointer to the pcie_dev data structure
1363 *
1364 * Invoked when PCI Express bus unloads or AER probe fails.
Hidetoshi Setof6d37802010-04-15 13:22:11 +09001365 */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001366static void aer_remove(struct pcie_device *dev)
1367{
1368 struct aer_rpc *rpc = get_service_data(dev);
1369
1370 if (rpc) {
Hidetoshi Seto460d2982010-04-15 13:10:03 +09001371 aer_disable_rootport(rpc);
1372 kfree(rpc);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001373 set_service_data(dev, NULL);
1374 }
1375}
1376
1377/**
1378 * aer_probe - initialize resources
1379 * @dev: pointer to the pcie_dev data structure
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001380 *
1381 * Invoked when PCI Express bus loads AER service driver.
Hidetoshi Setof6d37802010-04-15 13:22:11 +09001382 */
Bill Pemberton15856ad2012-11-21 15:35:00 -05001383static int aer_probe(struct pcie_device *dev)
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001384{
1385 int status;
1386 struct aer_rpc *rpc;
Bjorn Helgaas576700b2016-11-21 15:24:25 -06001387 struct device *device = &dev->port->dev;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001388
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001389 /* Alloc rpc data structure */
Keith Busch6200cc52018-09-18 17:58:46 -06001390 rpc = kzalloc(sizeof(struct aer_rpc), GFP_KERNEL);
Hidetoshi Setoc9a91882009-09-07 17:07:29 +09001391 if (!rpc) {
Bjorn Helgaas576700b2016-11-21 15:24:25 -06001392 dev_printk(KERN_DEBUG, device, "alloc AER rpc failed\n");
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001393 return -ENOMEM;
1394 }
Keith Busch6200cc52018-09-18 17:58:46 -06001395 rpc->rpd = dev->port;
1396 set_service_data(dev, rpc);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001397
1398 /* Request IRQ ISR */
Keith Busch6200cc52018-09-18 17:58:46 -06001399 status = request_threaded_irq(dev->irq, aer_irq, aer_isr,
1400 IRQF_SHARED, "aerdrv", dev);
Hidetoshi Setoc9a91882009-09-07 17:07:29 +09001401 if (status) {
Bjorn Helgaas576700b2016-11-21 15:24:25 -06001402 dev_printk(KERN_DEBUG, device, "request AER IRQ %d failed\n",
1403 dev->irq);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001404 aer_remove(dev);
1405 return status;
1406 }
1407
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001408 aer_enable_rootport(rpc);
Bjorn Helgaas68a55ae2016-11-21 15:34:02 -06001409 dev_info(device, "AER enabled with IRQ %d\n", dev->irq);
1410 return 0;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001411}
1412
1413/**
1414 * aer_root_reset - reset link on Root Port
1415 * @dev: pointer to Root Port's pci_dev data structure
1416 *
1417 * Invoked by Port Bus driver when performing link reset at Root Port.
Hidetoshi Setof6d37802010-04-15 13:22:11 +09001418 */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001419static pci_ers_result_t aer_root_reset(struct pci_dev *dev)
1420{
Hidetoshi Setoc6d34ed2010-04-15 13:09:13 +09001421 u32 reg32;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001422 int pos;
Sinan Kaya18426232018-07-19 18:04:09 -05001423 int rc;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001424
Keith Busch66b80802016-09-27 16:23:34 -04001425 pos = dev->aer_cap;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001426
1427 /* Disable Root's interrupt in response to error messages */
Hidetoshi Setoc6d34ed2010-04-15 13:09:13 +09001428 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, &reg32);
1429 reg32 &= ~ROOT_PORT_INTR_ON_MESG_MASK;
1430 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, reg32);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001431
Keith Buschc4eed622018-09-20 10:27:11 -06001432 rc = pci_bus_error_reset(dev);
Frederick Lawler7506dc72018-01-18 12:55:24 -06001433 pci_printk(KERN_DEBUG, dev, "Root Port link has been reset\n");
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001434
Hidetoshi Setoc6d34ed2010-04-15 13:09:13 +09001435 /* Clear Root Error Status */
1436 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, &reg32);
1437 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, reg32);
1438
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001439 /* Enable Root Port's interrupt in response to error messages */
Hidetoshi Setoc6d34ed2010-04-15 13:09:13 +09001440 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, &reg32);
1441 reg32 |= ROOT_PORT_INTR_ON_MESG_MASK;
1442 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, reg32);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001443
Sinan Kaya18426232018-07-19 18:04:09 -05001444 return rc ? PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_RECOVERED;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001445}
1446
Bjorn Helgaas0054ca82018-06-08 08:31:42 -05001447static struct pcie_port_service_driver aerdriver = {
1448 .name = "aer",
1449 .port_type = PCI_EXP_TYPE_ROOT_PORT,
1450 .service = PCIE_PORT_SERVICE_AER,
1451
1452 .probe = aer_probe,
1453 .remove = aer_remove,
Bjorn Helgaas0054ca82018-06-08 08:31:42 -05001454 .reset_link = aer_root_reset,
1455};
1456
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001457/**
1458 * aer_service_init - register AER root service driver
1459 *
1460 * Invoked when AER root service driver is loaded.
Hidetoshi Setof6d37802010-04-15 13:22:11 +09001461 */
Keith Buschc29de842018-09-20 10:27:06 -06001462int __init pcie_aer_init(void)
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001463{
Rafael J. Wysockib22c3d82010-09-20 18:50:00 +02001464 if (!pci_aer_available() || aer_acpi_firmware_first())
Andi Kleen3e77a3f2009-09-16 22:40:22 +02001465 return -ENXIO;
Sam Ravnborgc1996c22007-02-27 10:22:00 +01001466 return pcie_port_service_register(&aerdriver);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001467}