1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * 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.
6 *
7 * Copyright (C) 2006 Intel Corp.
8 * Tom Long Nguyen (tom.l.nguyen@intel.com)
9 * Zhang Yanmin (yanmin.zhang@intel.com)
10 *
11 * (C) Copyright 2009 Hewlett-Packard Development Company, L.P.
12 * Andrew Patterson <andrew.patterson@hp.com>
13 */
14
15 #define pr_fmt(fmt) "AER: " fmt
16 #define dev_fmt pr_fmt
17
18 #include <linux/cper.h>
19 #include <linux/pci.h>
20 #include <linux/pci-acpi.h>
21 #include <linux/sched.h>
22 #include <linux/kernel.h>
23 #include <linux/errno.h>
24 #include <linux/pm.h>
25 #include <linux/init.h>
26 #include <linux/interrupt.h>
27 #include <linux/delay.h>
28 #include <linux/kfifo.h>
29 #include <linux/slab.h>
30 #include <acpi/apei.h>
31 #include <ras/ras_event.h>
32
33 #include "../pci.h"
34 #include "portdrv.h"
35
36 #define AER_ERROR_SOURCES_MAX 128
37
38 #define AER_MAX_TYPEOF_COR_ERRS 16 /* as per PCI_ERR_COR_STATUS */
39 #define AER_MAX_TYPEOF_UNCOR_ERRS 26 /* as per PCI_ERR_UNCOR_STATUS*/
40
41 struct aer_err_source {
42 unsigned int status;
43 unsigned int id;
44 };
45
46 struct aer_rpc {
47 struct pci_dev *rpd; /* Root Port device */
48 DECLARE_KFIFO(aer_fifo, struct aer_err_source, AER_ERROR_SOURCES_MAX);
49 };
50
51 /* AER stats for the device */
52 struct aer_stats {
53
54 /*
55 * Fields for all AER capable devices. They indicate the errors
56 * "as seen by this device". Note that this may mean that if an
57 * end point is causing problems, the AER counters may increment
58 * at its link partner (e.g. root port) because the errors will be
59 * "seen" by the link partner and not the the problematic end point
60 * itself (which may report all counters as 0 as it never saw any
61 * problems).
62 */
63 /* Counters for different type of correctable errors */
64 u64 dev_cor_errs[AER_MAX_TYPEOF_COR_ERRS];
65 /* Counters for different type of fatal uncorrectable errors */
66 u64 dev_fatal_errs[AER_MAX_TYPEOF_UNCOR_ERRS];
67 /* Counters for different type of nonfatal uncorrectable errors */
68 u64 dev_nonfatal_errs[AER_MAX_TYPEOF_UNCOR_ERRS];
69 /* Total number of ERR_COR sent by this device */
70 u64 dev_total_cor_errs;
71 /* Total number of ERR_FATAL sent by this device */
72 u64 dev_total_fatal_errs;
73 /* Total number of ERR_NONFATAL sent by this device */
74 u64 dev_total_nonfatal_errs;
75
76 /*
77 * Fields for Root ports & root complex event collectors only, these
78 * indicate the total number of ERR_COR, ERR_FATAL, and ERR_NONFATAL
79 * messages received by the root port / event collector, INCLUDING the
80 * ones that are generated internally (by the rootport itself)
81 */
82 u64 rootport_total_cor_errs;
83 u64 rootport_total_fatal_errs;
84 u64 rootport_total_nonfatal_errs;
85 };
86
87 #define AER_LOG_TLP_MASKS (PCI_ERR_UNC_POISON_TLP| \
88 PCI_ERR_UNC_ECRC| \
89 PCI_ERR_UNC_UNSUP| \
90 PCI_ERR_UNC_COMP_ABORT| \
91 PCI_ERR_UNC_UNX_COMP| \
92 PCI_ERR_UNC_MALF_TLP)
93
94 #define SYSTEM_ERROR_INTR_ON_MESG_MASK (PCI_EXP_RTCTL_SECEE| \
95 PCI_EXP_RTCTL_SENFEE| \
96 PCI_EXP_RTCTL_SEFEE)
97 #define ROOT_PORT_INTR_ON_MESG_MASK (PCI_ERR_ROOT_CMD_COR_EN| \
98 PCI_ERR_ROOT_CMD_NONFATAL_EN| \
99 PCI_ERR_ROOT_CMD_FATAL_EN)
100 #define ERR_COR_ID(d) (d & 0xffff)
101 #define ERR_UNCOR_ID(d) (d >> 16)
102
103 static int pcie_aer_disable;
104
pci_no_aer(void)105 void pci_no_aer(void)
106 {
107 pcie_aer_disable = 1;
108 }
109
pci_aer_available(void)110 bool pci_aer_available(void)
111 {
112 return !pcie_aer_disable && pci_msi_enabled();
113 }
114
115 #ifdef CONFIG_PCIE_ECRC
116
117 #define ECRC_POLICY_DEFAULT 0 /* ECRC set by BIOS */
118 #define ECRC_POLICY_OFF 1 /* ECRC off for performance */
119 #define ECRC_POLICY_ON 2 /* ECRC on for data integrity */
120
121 static int ecrc_policy = ECRC_POLICY_DEFAULT;
122
123 static const char * const ecrc_policy_str[] = {
124 [ECRC_POLICY_DEFAULT] = "bios",
125 [ECRC_POLICY_OFF] = "off",
126 [ECRC_POLICY_ON] = "on"
127 };
128
129 /**
130 * enable_ercr_checking - enable PCIe ECRC checking for a device
131 * @dev: the PCI device
132 *
133 * Returns 0 on success, or negative on failure.
134 */
enable_ecrc_checking(struct pci_dev * dev)135 static int enable_ecrc_checking(struct pci_dev *dev)
136 {
137 int pos;
138 u32 reg32;
139
140 if (!pci_is_pcie(dev))
141 return -ENODEV;
142
143 pos = dev->aer_cap;
144 if (!pos)
145 return -ENODEV;
146
147 pci_read_config_dword(dev, pos + PCI_ERR_CAP, ®32);
148 if (reg32 & PCI_ERR_CAP_ECRC_GENC)
149 reg32 |= PCI_ERR_CAP_ECRC_GENE;
150 if (reg32 & PCI_ERR_CAP_ECRC_CHKC)
151 reg32 |= PCI_ERR_CAP_ECRC_CHKE;
152 pci_write_config_dword(dev, pos + PCI_ERR_CAP, reg32);
153
154 return 0;
155 }
156
157 /**
158 * disable_ercr_checking - disables PCIe ECRC checking for a device
159 * @dev: the PCI device
160 *
161 * Returns 0 on success, or negative on failure.
162 */
disable_ecrc_checking(struct pci_dev * dev)163 static int disable_ecrc_checking(struct pci_dev *dev)
164 {
165 int pos;
166 u32 reg32;
167
168 if (!pci_is_pcie(dev))
169 return -ENODEV;
170
171 pos = dev->aer_cap;
172 if (!pos)
173 return -ENODEV;
174
175 pci_read_config_dword(dev, pos + PCI_ERR_CAP, ®32);
176 reg32 &= ~(PCI_ERR_CAP_ECRC_GENE | PCI_ERR_CAP_ECRC_CHKE);
177 pci_write_config_dword(dev, pos + PCI_ERR_CAP, reg32);
178
179 return 0;
180 }
181
182 /**
183 * pcie_set_ecrc_checking - set/unset PCIe ECRC checking for a device based on global policy
184 * @dev: the PCI device
185 */
pcie_set_ecrc_checking(struct pci_dev * dev)186 void pcie_set_ecrc_checking(struct pci_dev *dev)
187 {
188 switch (ecrc_policy) {
189 case ECRC_POLICY_DEFAULT:
190 return;
191 case ECRC_POLICY_OFF:
192 disable_ecrc_checking(dev);
193 break;
194 case ECRC_POLICY_ON:
195 enable_ecrc_checking(dev);
196 break;
197 default:
198 return;
199 }
200 }
201
202 /**
203 * pcie_ecrc_get_policy - parse kernel command-line ecrc option
204 */
pcie_ecrc_get_policy(char * str)205 void pcie_ecrc_get_policy(char *str)
206 {
207 int i;
208
209 i = match_string(ecrc_policy_str, ARRAY_SIZE(ecrc_policy_str), str);
210 if (i < 0)
211 return;
212
213 ecrc_policy = i;
214 }
215 #endif /* CONFIG_PCIE_ECRC */
216
217 #ifdef CONFIG_ACPI_APEI
hest_match_pci(struct acpi_hest_aer_common * p,struct pci_dev * pci)218 static 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
hest_match_type(struct acpi_hest_header * hest_hdr,struct pci_dev * dev)227 static 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
243 struct aer_hest_parse_info {
244 struct pci_dev *pci_dev;
245 int firmware_first;
246 };
247
hest_source_is_pcie_aer(struct acpi_hest_header * hest_hdr)248 static 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
aer_hest_parse(struct acpi_hest_header * hest_hdr,void * data)257 static 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
aer_set_firmware_first(struct pci_dev * pci_dev)289 static 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
pcie_aer_get_firmware_first(struct pci_dev * dev)306 int pcie_aer_get_firmware_first(struct pci_dev *dev)
307 {
308 if (!pci_is_pcie(dev))
309 return 0;
310
311 if (pcie_ports_native)
312 return 0;
313
314 if (!dev->__aer_firmware_first_valid)
315 aer_set_firmware_first(dev);
316 return dev->__aer_firmware_first;
317 }
318
319 static bool aer_firmware_first;
320
321 /**
322 * aer_acpi_firmware_first - Check if APEI should control AER.
323 */
aer_acpi_firmware_first(void)324 bool 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
332 if (pcie_ports_native)
333 return false;
334
335 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
344 #define PCI_EXP_AER_FLAGS (PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE | \
345 PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE)
346
pci_enable_pcie_error_reporting(struct pci_dev * dev)347 int 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 }
357 EXPORT_SYMBOL_GPL(pci_enable_pcie_error_reporting);
358
pci_disable_pcie_error_reporting(struct pci_dev * dev)359 int 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 }
367 EXPORT_SYMBOL_GPL(pci_disable_pcie_error_reporting);
368
pci_aer_clear_device_status(struct pci_dev * dev)369 void 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
pci_cleanup_aer_uncorrect_error_status(struct pci_dev * dev)377 int pci_cleanup_aer_uncorrect_error_status(struct pci_dev *dev)
378 {
379 int pos;
380 u32 status, sev;
381
382 pos = dev->aer_cap;
383 if (!pos)
384 return -EIO;
385
386 if (pcie_aer_get_firmware_first(dev))
387 return -EIO;
388
389 /* Clear status bits for ERR_NONFATAL errors only */
390 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
391 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &sev);
392 status &= ~sev;
393 if (status)
394 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
395
396 return 0;
397 }
398 EXPORT_SYMBOL_GPL(pci_cleanup_aer_uncorrect_error_status);
399
pci_aer_clear_fatal_status(struct pci_dev * dev)400 void 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
409 if (pcie_aer_get_firmware_first(dev))
410 return;
411
412 /* 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
pci_cleanup_aer_error_status_regs(struct pci_dev * dev)420 int 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
433 if (pcie_aer_get_firmware_first(dev))
434 return -EIO;
435
436 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
pci_aer_init(struct pci_dev * dev)451 void pci_aer_init(struct pci_dev *dev)
452 {
453 dev->aer_cap = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
454
455 if (dev->aer_cap)
456 dev->aer_stats = kzalloc(sizeof(struct aer_stats), GFP_KERNEL);
457
458 pci_cleanup_aer_error_status_regs(dev);
459 }
460
pci_aer_exit(struct pci_dev * dev)461 void pci_aer_exit(struct pci_dev *dev)
462 {
463 kfree(dev->aer_stats);
464 dev->aer_stats = NULL;
465 }
466
467 #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 */
505 static const char *aer_error_severity_string[] = {
506 "Uncorrected (Non-Fatal)",
507 "Uncorrected (Fatal)",
508 "Corrected"
509 };
510
511 static const char *aer_error_layer[] = {
512 "Physical Layer",
513 "Data Link Layer",
514 "Transaction Layer"
515 };
516
517 static const char *aer_correctable_error_string[AER_MAX_TYPEOF_COR_ERRS] = {
518 "RxErr", /* Bit Position 0 */
519 NULL,
520 NULL,
521 NULL,
522 NULL,
523 NULL,
524 "BadTLP", /* Bit Position 6 */
525 "BadDLLP", /* Bit Position 7 */
526 "Rollover", /* Bit Position 8 */
527 NULL,
528 NULL,
529 NULL,
530 "Timeout", /* Bit Position 12 */
531 "NonFatalErr", /* Bit Position 13 */
532 "CorrIntErr", /* Bit Position 14 */
533 "HeaderOF", /* Bit Position 15 */
534 };
535
536 static const char *aer_uncorrectable_error_string[AER_MAX_TYPEOF_UNCOR_ERRS] = {
537 "Undefined", /* Bit Position 0 */
538 NULL,
539 NULL,
540 NULL,
541 "DLP", /* Bit Position 4 */
542 "SDES", /* Bit Position 5 */
543 NULL,
544 NULL,
545 NULL,
546 NULL,
547 NULL,
548 NULL,
549 "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 */
556 "ECRC", /* Bit Position 19 */
557 "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 */
563 };
564
565 static const char *aer_agent_string[] = {
566 "Receiver ID",
567 "Requester ID",
568 "Completer ID",
569 "Transmitter ID"
570 };
571
572 #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 } \
595 static DEVICE_ATTR_RO(name)
596
597 aer_stats_dev_attr(aer_dev_correctable, dev_cor_errs,
598 aer_correctable_error_string, "ERR_COR",
599 dev_total_cor_errs);
600 aer_stats_dev_attr(aer_dev_fatal, dev_fatal_errs,
601 aer_uncorrectable_error_string, "ERR_FATAL",
602 dev_total_fatal_errs);
603 aer_stats_dev_attr(aer_dev_nonfatal, dev_nonfatal_errs,
604 aer_uncorrectable_error_string, "ERR_NONFATAL",
605 dev_total_nonfatal_errs);
606
607 #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 } \
615 static DEVICE_ATTR_RO(name)
616
617 aer_stats_rootport_attr(aer_rootport_total_err_cor,
618 rootport_total_cor_errs);
619 aer_stats_rootport_attr(aer_rootport_total_err_fatal,
620 rootport_total_fatal_errs);
621 aer_stats_rootport_attr(aer_rootport_total_err_nonfatal,
622 rootport_total_nonfatal_errs);
623
624 static 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,
628 &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,
631 NULL
632 };
633
aer_stats_attrs_are_visible(struct kobject * kobj,struct attribute * a,int n)634 static 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
643 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
649 return a->mode;
650 }
651
652 const struct attribute_group aer_stats_attr_group = {
653 .attrs = aer_stats_attrs,
654 .is_visible = aer_stats_attrs_are_visible,
655 };
656
pci_dev_aer_stats_incr(struct pci_dev * pdev,struct aer_err_info * info)657 static 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
pci_rootport_aer_stats_incr(struct pci_dev * pdev,struct aer_err_source * e_src)691 static 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
__print_tlp_header(struct pci_dev * dev,struct aer_header_log_regs * t)710 static 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
__aer_print_error(struct pci_dev * dev,struct aer_err_info * info)717 static 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 }
742 pci_dev_aer_stats_incr(dev, info);
743 }
744
aer_print_error(struct pci_dev * dev,struct aer_err_info * info)745 void aer_print_error(struct pci_dev *dev, struct aer_err_info *info)
746 {
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
772 out:
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
aer_print_port_info(struct pci_dev * dev,struct aer_err_info * info)780 static 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, "%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),
789 PCI_FUNC(devfn));
790 }
791
792 #ifdef CONFIG_ACPI_APEI_PCIEAER
cper_severity_to_aer(int cper_severity)793 int cper_severity_to_aer(int cper_severity)
794 {
795 switch (cper_severity) {
796 case CPER_SEV_RECOVERABLE:
797 return AER_NONFATAL;
798 case CPER_SEV_FATAL:
799 return AER_FATAL;
800 default:
801 return AER_CORRECTABLE;
802 }
803 }
804 EXPORT_SYMBOL_GPL(cper_severity_to_aer);
805
cper_print_aer(struct pci_dev * dev,int aer_severity,struct aer_capability_regs * aer)806 void cper_print_aer(struct pci_dev *dev, int aer_severity,
807 struct aer_capability_regs *aer)
808 {
809 int layer, agent, tlp_header_valid = 0;
810 u32 status, mask;
811 struct aer_err_info info;
812
813 if (aer_severity == AER_CORRECTABLE) {
814 status = aer->cor_status;
815 mask = aer->cor_mask;
816 } else {
817 status = aer->uncor_status;
818 mask = aer->uncor_mask;
819 tlp_header_valid = status & AER_LOG_TLP_MASKS;
820 }
821
822 layer = AER_GET_LAYER_ERROR(aer_severity, status);
823 agent = AER_GET_AGENT(aer_severity, status);
824
825 memset(&info, 0, sizeof(info));
826 info.severity = aer_severity;
827 info.status = status;
828 info.mask = mask;
829 info.first_error = PCI_ERR_CAP_FEP(aer->cap_control);
830
831 pci_err(dev, "aer_status: 0x%08x, aer_mask: 0x%08x\n", status, mask);
832 __aer_print_error(dev, &info);
833 pci_err(dev, "aer_layer=%s, aer_agent=%s\n",
834 aer_error_layer[layer], aer_agent_string[agent]);
835
836 if (aer_severity != AER_CORRECTABLE)
837 pci_err(dev, "aer_uncor_severity: 0x%08x\n",
838 aer->uncor_severity);
839
840 if (tlp_header_valid)
841 __print_tlp_header(dev, &aer->header_log);
842
843 trace_aer_event(dev_name(&dev->dev), (status & ~mask),
844 aer_severity, tlp_header_valid, &aer->header_log);
845 }
846 #endif
847
848 /**
849 * add_error_device - list device to be handled
850 * @e_info: pointer to error info
851 * @dev: pointer to pci_dev to be added
852 */
add_error_device(struct aer_err_info * e_info,struct pci_dev * dev)853 static int add_error_device(struct aer_err_info *e_info, struct pci_dev *dev)
854 {
855 if (e_info->error_dev_num < AER_MAX_MULTI_ERR_DEVICES) {
856 e_info->dev[e_info->error_dev_num] = pci_dev_get(dev);
857 e_info->error_dev_num++;
858 return 0;
859 }
860 return -ENOSPC;
861 }
862
863 /**
864 * is_error_source - check whether the device is source of reported error
865 * @dev: pointer to pci_dev to be checked
866 * @e_info: pointer to reported error info
867 */
is_error_source(struct pci_dev * dev,struct aer_err_info * e_info)868 static bool is_error_source(struct pci_dev *dev, struct aer_err_info *e_info)
869 {
870 int pos;
871 u32 status, mask;
872 u16 reg16;
873
874 /*
875 * When bus id is equal to 0, it might be a bad id
876 * reported by root port.
877 */
878 if ((PCI_BUS_NUM(e_info->id) != 0) &&
879 !(dev->bus->bus_flags & PCI_BUS_FLAGS_NO_AERSID)) {
880 /* Device ID match? */
881 if (e_info->id == ((dev->bus->number << 8) | dev->devfn))
882 return true;
883
884 /* Continue id comparing if there is no multiple error */
885 if (!e_info->multi_error_valid)
886 return false;
887 }
888
889 /*
890 * When either
891 * 1) bus id is equal to 0. Some ports might lose the bus
892 * id of error source id;
893 * 2) bus flag PCI_BUS_FLAGS_NO_AERSID is set
894 * 3) There are multiple errors and prior ID comparing fails;
895 * We check AER status registers to find possible reporter.
896 */
897 if (atomic_read(&dev->enable_cnt) == 0)
898 return false;
899
900 /* Check if AER is enabled */
901 pcie_capability_read_word(dev, PCI_EXP_DEVCTL, ®16);
902 if (!(reg16 & PCI_EXP_AER_FLAGS))
903 return false;
904
905 pos = dev->aer_cap;
906 if (!pos)
907 return false;
908
909 /* Check if error is recorded */
910 if (e_info->severity == AER_CORRECTABLE) {
911 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
912 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, &mask);
913 } else {
914 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
915 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, &mask);
916 }
917 if (status & ~mask)
918 return true;
919
920 return false;
921 }
922
find_device_iter(struct pci_dev * dev,void * data)923 static int find_device_iter(struct pci_dev *dev, void *data)
924 {
925 struct aer_err_info *e_info = (struct aer_err_info *)data;
926
927 if (is_error_source(dev, e_info)) {
928 /* List this device */
929 if (add_error_device(e_info, dev)) {
930 /* We cannot handle more... Stop iteration */
931 /* TODO: Should print error message here? */
932 return 1;
933 }
934
935 /* If there is only a single error, stop iteration */
936 if (!e_info->multi_error_valid)
937 return 1;
938 }
939 return 0;
940 }
941
942 /**
943 * find_source_device - search through device hierarchy for source device
944 * @parent: pointer to Root Port pci_dev data structure
945 * @e_info: including detailed error information such like id
946 *
947 * Return true if found.
948 *
949 * Invoked by DPC when error is detected at the Root Port.
950 * Caller of this function must set id, severity, and multi_error_valid of
951 * struct aer_err_info pointed by @e_info properly. This function must fill
952 * e_info->error_dev_num and e_info->dev[], based on the given information.
953 */
find_source_device(struct pci_dev * parent,struct aer_err_info * e_info)954 static bool find_source_device(struct pci_dev *parent,
955 struct aer_err_info *e_info)
956 {
957 struct pci_dev *dev = parent;
958 int result;
959
960 /* Must reset in this function */
961 e_info->error_dev_num = 0;
962
963 /* Is Root Port an agent that sends error message? */
964 result = find_device_iter(dev, e_info);
965 if (result)
966 return true;
967
968 pci_walk_bus(parent->subordinate, find_device_iter, e_info);
969
970 if (!e_info->error_dev_num) {
971 pci_info(parent, "can't find device of ID%04x\n", 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 */
handle_error_source(struct pci_dev * dev,struct aer_err_info * info)984 static 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);
997 pci_aer_clear_device_status(dev);
998 } else if (info->severity == AER_NONFATAL)
999 pcie_do_recovery(dev, pci_channel_io_normal,
1000 PCIE_PORT_SERVICE_AER);
1001 else if (info->severity == AER_FATAL)
1002 pcie_do_recovery(dev, pci_channel_io_frozen,
1003 PCIE_PORT_SERVICE_AER);
1004 pci_dev_put(dev);
1005 }
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
1012 struct aer_recover_entry {
1013 u8 bus;
1014 u8 devfn;
1015 u16 domain;
1016 int severity;
1017 struct aer_capability_regs *regs;
1018 };
1019
1020 static DEFINE_KFIFO(aer_recover_ring, struct aer_recover_entry,
1021 AER_RECOVER_RING_SIZE);
1022
aer_recover_work_func(struct work_struct * work)1023 static 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)
1039 pcie_do_recovery(pdev, pci_channel_io_normal,
1040 PCIE_PORT_SERVICE_AER);
1041 else if (entry.severity == AER_FATAL)
1042 pcie_do_recovery(pdev, pci_channel_io_frozen,
1043 PCIE_PORT_SERVICE_AER);
1044 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 */
1053 static DEFINE_SPINLOCK(aer_recover_ring_lock);
1054 static DECLARE_WORK(aer_recover_work, aer_recover_work_func);
1055
aer_recover_queue(int domain,unsigned int bus,unsigned int devfn,int severity,struct aer_capability_regs * aer_regs)1056 void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn,
1057 int severity, struct aer_capability_regs *aer_regs)
1058 {
1059 struct aer_recover_entry entry = {
1060 .bus = bus,
1061 .devfn = devfn,
1062 .domain = domain,
1063 .severity = severity,
1064 .regs = aer_regs,
1065 };
1066
1067 if (kfifo_in_spinlocked(&aer_recover_ring, &entry, 1,
1068 &aer_recover_ring_lock))
1069 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));
1073 }
1074 EXPORT_SYMBOL_GPL(aer_recover_queue);
1075 #endif
1076
1077 /**
1078 * aer_get_device_error_info - read error status from dev and store it to info
1079 * @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 */
aer_get_device_error_info(struct pci_dev * dev,struct aer_err_info * info)1086 int aer_get_device_error_info(struct pci_dev *dev, struct aer_err_info *info)
1087 {
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;
1107 } 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) {
1110
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
aer_process_err_devices(struct aer_err_info * e_info)1139 static 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++) {
1145 if (aer_get_device_error_info(e_info->dev[i], e_info))
1146 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++) {
1149 if (aer_get_device_error_info(e_info->dev[i], e_info))
1150 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 */
aer_isr_one_error(struct aer_rpc * rpc,struct aer_err_source * e_src)1159 static void aer_isr_one_error(struct aer_rpc *rpc,
1160 struct aer_err_source *e_src)
1161 {
1162 struct pci_dev *pdev = rpc->rpd;
1163 struct aer_err_info e_info;
1164
1165 pci_rootport_aer_stats_incr(pdev, e_src);
1166
1167 /*
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) {
1172 e_info.id = ERR_COR_ID(e_src->id);
1173 e_info.severity = AER_CORRECTABLE;
1174
1175 if (e_src->status & PCI_ERR_ROOT_MULTI_COR_RCV)
1176 e_info.multi_error_valid = 1;
1177 else
1178 e_info.multi_error_valid = 0;
1179 aer_print_port_info(pdev, &e_info);
1180
1181 if (find_source_device(pdev, &e_info))
1182 aer_process_err_devices(&e_info);
1183 }
1184
1185 if (e_src->status & PCI_ERR_ROOT_UNCOR_RCV) {
1186 e_info.id = ERR_UNCOR_ID(e_src->id);
1187
1188 if (e_src->status & PCI_ERR_ROOT_FATAL_RCV)
1189 e_info.severity = AER_FATAL;
1190 else
1191 e_info.severity = AER_NONFATAL;
1192
1193 if (e_src->status & PCI_ERR_ROOT_MULTI_UNCOR_RCV)
1194 e_info.multi_error_valid = 1;
1195 else
1196 e_info.multi_error_valid = 0;
1197
1198 aer_print_port_info(pdev, &e_info);
1199
1200 if (find_source_device(pdev, &e_info))
1201 aer_process_err_devices(&e_info);
1202 }
1203 }
1204
1205 /**
1206 * 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 */
aer_isr(int irq,void * context)1211 static irqreturn_t aer_isr(int irq, void *context)
1212 {
1213 struct pcie_device *dev = (struct pcie_device *)context;
1214 struct aer_rpc *rpc = get_service_data(dev);
1215 struct aer_err_source uninitialized_var(e_src);
1216
1217 if (kfifo_is_empty(&rpc->aer_fifo))
1218 return IRQ_NONE;
1219
1220 while (kfifo_get(&rpc->aer_fifo, &e_src))
1221 aer_isr_one_error(rpc, &e_src);
1222 return IRQ_HANDLED;
1223 }
1224
1225 /**
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 */
aer_irq(int irq,void * context)1232 static irqreturn_t aer_irq(int irq, void *context)
1233 {
1234 struct pcie_device *pdev = (struct pcie_device *)context;
1235 struct aer_rpc *rpc = get_service_data(pdev);
1236 struct pci_dev *rp = rpc->rpd;
1237 struct aer_err_source e_src = {};
1238 int pos = rp->aer_cap;
1239
1240 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)))
1242 return IRQ_NONE;
1243
1244 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);
1246
1247 if (!kfifo_put(&rpc->aer_fifo, e_src))
1248 return IRQ_HANDLED;
1249
1250 return IRQ_WAKE_THREAD;
1251 }
1252
set_device_error_reporting(struct pci_dev * dev,void * data)1253 static int set_device_error_reporting(struct pci_dev *dev, void *data)
1254 {
1255 bool enable = *((bool *)data);
1256 int type = pci_pcie_type(dev);
1257
1258 if ((type == PCI_EXP_TYPE_ROOT_PORT) ||
1259 (type == PCI_EXP_TYPE_UPSTREAM) ||
1260 (type == PCI_EXP_TYPE_DOWNSTREAM)) {
1261 if (enable)
1262 pci_enable_pcie_error_reporting(dev);
1263 else
1264 pci_disable_pcie_error_reporting(dev);
1265 }
1266
1267 if (enable)
1268 pcie_set_ecrc_checking(dev);
1269
1270 return 0;
1271 }
1272
1273 /**
1274 * set_downstream_devices_error_reporting - enable/disable the error reporting bits on the root port and its downstream ports.
1275 * @dev: pointer to root port's pci_dev data structure
1276 * @enable: true = enable error reporting, false = disable error reporting.
1277 */
set_downstream_devices_error_reporting(struct pci_dev * dev,bool enable)1278 static void set_downstream_devices_error_reporting(struct pci_dev *dev,
1279 bool enable)
1280 {
1281 set_device_error_reporting(dev, &enable);
1282
1283 if (!dev->subordinate)
1284 return;
1285 pci_walk_bus(dev->subordinate, set_device_error_reporting, &enable);
1286 }
1287
1288 /**
1289 * aer_enable_rootport - enable Root Port's interrupts when receiving messages
1290 * @rpc: pointer to a Root Port data structure
1291 *
1292 * Invoked when PCIe bus loads AER service driver.
1293 */
aer_enable_rootport(struct aer_rpc * rpc)1294 static void aer_enable_rootport(struct aer_rpc *rpc)
1295 {
1296 struct pci_dev *pdev = rpc->rpd;
1297 int aer_pos;
1298 u16 reg16;
1299 u32 reg32;
1300
1301 /* Clear PCIe Capability's Device Status */
1302 pcie_capability_read_word(pdev, PCI_EXP_DEVSTA, ®16);
1303 pcie_capability_write_word(pdev, PCI_EXP_DEVSTA, reg16);
1304
1305 /* Disable system error generation in response to error messages */
1306 pcie_capability_clear_word(pdev, PCI_EXP_RTCTL,
1307 SYSTEM_ERROR_INTR_ON_MESG_MASK);
1308
1309 aer_pos = pdev->aer_cap;
1310 /* Clear error status */
1311 pci_read_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, ®32);
1312 pci_write_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, reg32);
1313 pci_read_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, ®32);
1314 pci_write_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, reg32);
1315 pci_read_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, ®32);
1316 pci_write_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, reg32);
1317
1318 /*
1319 * Enable error reporting for the root port device and downstream port
1320 * devices.
1321 */
1322 set_downstream_devices_error_reporting(pdev, true);
1323
1324 /* Enable Root Port's interrupt in response to error messages */
1325 pci_read_config_dword(pdev, aer_pos + PCI_ERR_ROOT_COMMAND, ®32);
1326 reg32 |= ROOT_PORT_INTR_ON_MESG_MASK;
1327 pci_write_config_dword(pdev, aer_pos + PCI_ERR_ROOT_COMMAND, reg32);
1328 }
1329
1330 /**
1331 * aer_disable_rootport - disable Root Port's interrupts when receiving messages
1332 * @rpc: pointer to a Root Port data structure
1333 *
1334 * Invoked when PCIe bus unloads AER service driver.
1335 */
aer_disable_rootport(struct aer_rpc * rpc)1336 static void aer_disable_rootport(struct aer_rpc *rpc)
1337 {
1338 struct pci_dev *pdev = rpc->rpd;
1339 u32 reg32;
1340 int pos;
1341
1342 /*
1343 * Disable error reporting for the root port device and downstream port
1344 * devices.
1345 */
1346 set_downstream_devices_error_reporting(pdev, false);
1347
1348 pos = pdev->aer_cap;
1349 /* Disable Root's interrupt in response to error messages */
1350 pci_read_config_dword(pdev, pos + PCI_ERR_ROOT_COMMAND, ®32);
1351 reg32 &= ~ROOT_PORT_INTR_ON_MESG_MASK;
1352 pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_COMMAND, reg32);
1353
1354 /* Clear Root's error status reg */
1355 pci_read_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, ®32);
1356 pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, reg32);
1357 }
1358
1359 /**
1360 * aer_remove - clean up resources
1361 * @dev: pointer to the pcie_dev data structure
1362 *
1363 * Invoked when PCI Express bus unloads or AER probe fails.
1364 */
aer_remove(struct pcie_device * dev)1365 static void aer_remove(struct pcie_device *dev)
1366 {
1367 struct aer_rpc *rpc = get_service_data(dev);
1368
1369 aer_disable_rootport(rpc);
1370 }
1371
1372 /**
1373 * aer_probe - initialize resources
1374 * @dev: pointer to the pcie_dev data structure
1375 *
1376 * Invoked when PCI Express bus loads AER service driver.
1377 */
aer_probe(struct pcie_device * dev)1378 static int aer_probe(struct pcie_device *dev)
1379 {
1380 int status;
1381 struct aer_rpc *rpc;
1382 struct device *device = &dev->device;
1383 struct pci_dev *port = dev->port;
1384
1385 rpc = devm_kzalloc(device, sizeof(struct aer_rpc), GFP_KERNEL);
1386 if (!rpc)
1387 return -ENOMEM;
1388
1389 rpc->rpd = port;
1390 set_service_data(dev, rpc);
1391
1392 status = devm_request_threaded_irq(device, dev->irq, aer_irq, aer_isr,
1393 IRQF_SHARED, "aerdrv", dev);
1394 if (status) {
1395 pci_err(port, "request AER IRQ %d failed\n", dev->irq);
1396 return status;
1397 }
1398
1399 aer_enable_rootport(rpc);
1400 pci_info(port, "enabled with IRQ %d\n", dev->irq);
1401 return 0;
1402 }
1403
1404 /**
1405 * aer_root_reset - reset link on Root Port
1406 * @dev: pointer to Root Port's pci_dev data structure
1407 *
1408 * Invoked by Port Bus driver when performing link reset at Root Port.
1409 */
aer_root_reset(struct pci_dev * dev)1410 static pci_ers_result_t aer_root_reset(struct pci_dev *dev)
1411 {
1412 u32 reg32;
1413 int pos;
1414 int rc;
1415
1416 pos = dev->aer_cap;
1417
1418 /* Disable Root's interrupt in response to error messages */
1419 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, ®32);
1420 reg32 &= ~ROOT_PORT_INTR_ON_MESG_MASK;
1421 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, reg32);
1422
1423 rc = pci_bus_error_reset(dev);
1424 pci_info(dev, "Root Port link has been reset\n");
1425
1426 /* Clear Root Error Status */
1427 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, ®32);
1428 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, reg32);
1429
1430 /* Enable Root Port's interrupt in response to error messages */
1431 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, ®32);
1432 reg32 |= ROOT_PORT_INTR_ON_MESG_MASK;
1433 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, reg32);
1434
1435 return rc ? PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_RECOVERED;
1436 }
1437
1438 static struct pcie_port_service_driver aerdriver = {
1439 .name = "aer",
1440 .port_type = PCI_EXP_TYPE_ROOT_PORT,
1441 .service = PCIE_PORT_SERVICE_AER,
1442
1443 .probe = aer_probe,
1444 .remove = aer_remove,
1445 .reset_link = aer_root_reset,
1446 };
1447
1448 /**
1449 * aer_service_init - register AER root service driver
1450 *
1451 * Invoked when AER root service driver is loaded.
1452 */
pcie_aer_init(void)1453 int __init pcie_aer_init(void)
1454 {
1455 if (!pci_aer_available() || aer_acpi_firmware_first())
1456 return -ENXIO;
1457 return pcie_port_service_register(&aerdriver);
1458 }
1459