Merge branches 'acpi-video', 'device-properties', 'pm-sleep' and 'pm-cpuidle'
[linux-2.6-block.git] / drivers / pci / pcie / aer / aerdrv_core.c
CommitLineData
6c2b374d
ZY
1/*
2 * drivers/pci/pcie/aer/aerdrv_core.c
3 *
4 * This file is subject to the terms and conditions of the GNU General Public
5 * License. See the file "COPYING" in the main directory of this archive
6 * for more details.
7 *
8 * This file implements the core part of PCI-Express AER. When an pci-express
9 * error is delivered, an error message will be collected and printed to
10 * console, then, an error recovery procedure will be executed by following
11 * the pci error recovery rules.
12 *
13 * Copyright (C) 2006 Intel Corp.
14 * Tom Long Nguyen (tom.l.nguyen@intel.com)
15 * Zhang Yanmin (yanmin.zhang@intel.com)
16 *
17 */
18
19#include <linux/module.h>
20#include <linux/pci.h>
21#include <linux/kernel.h>
22#include <linux/errno.h>
23#include <linux/pm.h>
24#include <linux/suspend.h>
6c2b374d 25#include <linux/delay.h>
5a0e3ad6 26#include <linux/slab.h>
0918472c 27#include <linux/kfifo.h>
6c2b374d
ZY
28#include "aerdrv.h"
29
90ab5ee9
RR
30static bool forceload;
31static bool nosourceid;
6c2b374d 32module_param(forceload, bool, 0);
28eb27cf 33module_param(nosourceid, bool, 0);
6c2b374d 34
43bd4ee8
JL
35#define PCI_EXP_AER_FLAGS (PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE | \
36 PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE)
37
6c2b374d
ZY
38int pci_enable_pcie_error_reporting(struct pci_dev *dev)
39{
affb72c3 40 if (pcie_aer_get_firmware_first(dev))
05843961
MD
41 return -EIO;
42
43bd4ee8 43 if (!pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR))
0927678f
JB
44 return -EIO;
45
43bd4ee8 46 return pcie_capability_set_word(dev, PCI_EXP_DEVCTL, PCI_EXP_AER_FLAGS);
6c2b374d 47}
c9a91883 48EXPORT_SYMBOL_GPL(pci_enable_pcie_error_reporting);
6c2b374d
ZY
49
50int pci_disable_pcie_error_reporting(struct pci_dev *dev)
51{
affb72c3 52 if (pcie_aer_get_firmware_first(dev))
05843961
MD
53 return -EIO;
54
43bd4ee8
JL
55 return pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
56 PCI_EXP_AER_FLAGS);
6c2b374d 57}
c9a91883 58EXPORT_SYMBOL_GPL(pci_disable_pcie_error_reporting);
6c2b374d
ZY
59
60int pci_cleanup_aer_uncorrect_error_status(struct pci_dev *dev)
61{
62 int pos;
6cdfd995 63 u32 status;
6c2b374d 64
0927678f 65 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
6c2b374d
ZY
66 if (!pos)
67 return -EIO;
68
69 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
6cdfd995
AP
70 if (status)
71 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
6c2b374d
ZY
72
73 return 0;
74}
c9a91883 75EXPORT_SYMBOL_GPL(pci_cleanup_aer_uncorrect_error_status);
6c2b374d 76
4a0c096e
HS
77/**
78 * add_error_device - list device to be handled
79 * @e_info: pointer to error info
80 * @dev: pointer to pci_dev to be added
81 */
3d5505c5
ZY
82static int add_error_device(struct aer_err_info *e_info, struct pci_dev *dev)
83{
84 if (e_info->error_dev_num < AER_MAX_MULTI_ERR_DEVICES) {
85 e_info->dev[e_info->error_dev_num] = dev;
86 e_info->error_dev_num++;
4a0c096e 87 return 0;
c9a91883 88 }
4a0c096e 89 return -ENOSPC;
3d5505c5
ZY
90}
91
c887275e
HS
92/**
93 * is_error_source - check whether the device is source of reported error
94 * @dev: pointer to pci_dev to be checked
95 * @e_info: pointer to reported error info
96 */
97static bool is_error_source(struct pci_dev *dev, struct aer_err_info *e_info)
28eb27cf
ZY
98{
99 int pos;
c887275e 100 u32 status, mask;
28eb27cf 101 u16 reg16;
28eb27cf
ZY
102
103 /*
104 * When bus id is equal to 0, it might be a bad id
105 * reported by root port.
106 */
fff0ee36 107 if (!nosourceid && (PCI_BUS_NUM(e_info->id) != 0)) {
bd17d474
HS
108 /* Device ID match? */
109 if (e_info->id == ((dev->bus->number << 8) | dev->devfn))
c887275e 110 return true;
3d5505c5 111
c887275e 112 /* Continue id comparing if there is no multiple error */
273024de 113 if (!e_info->multi_error_valid)
c887275e 114 return false;
28eb27cf
ZY
115 }
116
117 /*
3d5505c5
ZY
118 * When either
119 * 1) nosourceid==y;
120 * 2) bus id is equal to 0. Some ports might lose the bus
121 * id of error source id;
122 * 3) There are multiple errors and prior id comparing fails;
c887275e 123 * We check AER status registers to find possible reporter.
28eb27cf
ZY
124 */
125 if (atomic_read(&dev->enable_cnt) == 0)
c887275e 126 return false;
c887275e 127
28eb27cf 128 /* Check if AER is enabled */
43bd4ee8
JL
129 pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &reg16);
130 if (!(reg16 & PCI_EXP_AER_FLAGS))
c887275e 131 return false;
43bd4ee8 132
28eb27cf
ZY
133 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
134 if (!pos)
c887275e 135 return false;
28eb27cf 136
c887275e 137 /* Check if error is recorded */
28eb27cf 138 if (e_info->severity == AER_CORRECTABLE) {
0d90c3ac
HS
139 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
140 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, &mask);
28eb27cf 141 } else {
0d90c3ac
HS
142 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
143 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, &mask);
6c2b374d 144 }
c887275e
HS
145 if (status & ~mask)
146 return true;
6c2b374d 147
c887275e
HS
148 return false;
149}
3d5505c5 150
c887275e
HS
151static int find_device_iter(struct pci_dev *dev, void *data)
152{
153 struct aer_err_info *e_info = (struct aer_err_info *)data;
154
155 if (is_error_source(dev, e_info)) {
4a0c096e
HS
156 /* List this device */
157 if (add_error_device(e_info, dev)) {
158 /* We cannot handle more... Stop iteration */
159 /* TODO: Should print error message here? */
160 return 1;
161 }
c887275e
HS
162
163 /* If there is only a single error, stop iteration */
164 if (!e_info->multi_error_valid)
165 return 1;
166 }
167 return 0;
6c2b374d
ZY
168}
169
170/**
171 * find_source_device - search through device hierarchy for source device
d885c6b7 172 * @parent: pointer to Root Port pci_dev data structure
98ca3964 173 * @e_info: including detailed error information such like id
6c2b374d 174 *
98ca3964
HS
175 * Return true if found.
176 *
177 * Invoked by DPC when error is detected at the Root Port.
7c4ec94f
HS
178 * Caller of this function must set id, severity, and multi_error_valid of
179 * struct aer_err_info pointed by @e_info properly. This function must fill
180 * e_info->error_dev_num and e_info->dev[], based on the given information.
d885c6b7 181 */
98ca3964 182static bool find_source_device(struct pci_dev *parent,
28eb27cf 183 struct aer_err_info *e_info)
6c2b374d
ZY
184{
185 struct pci_dev *dev = parent;
28eb27cf 186 int result;
6c2b374d 187
7c4ec94f
HS
188 /* Must reset in this function */
189 e_info->error_dev_num = 0;
190
6c2b374d 191 /* Is Root Port an agent that sends error message? */
28eb27cf
ZY
192 result = find_device_iter(dev, e_info);
193 if (result)
98ca3964 194 return true;
6c2b374d 195
28eb27cf 196 pci_walk_bus(parent->subordinate, find_device_iter, e_info);
98ca3964
HS
197
198 if (!e_info->error_dev_num) {
199 dev_printk(KERN_DEBUG, &parent->dev,
200 "can't find device of ID%04x\n",
201 e_info->id);
202 return false;
203 }
204 return true;
6c2b374d
ZY
205}
206
70298c6e 207static int report_error_detected(struct pci_dev *dev, void *data)
6c2b374d
ZY
208{
209 pci_ers_result_t vote;
49453028 210 const struct pci_error_handlers *err_handler;
6c2b374d
ZY
211 struct aer_broadcast_data *result_data;
212 result_data = (struct aer_broadcast_data *) data;
213
90b5c1d7 214 device_lock(&dev->dev);
6c2b374d
ZY
215 dev->error_state = result_data->state;
216
217 if (!dev->driver ||
218 !dev->driver->err_handler ||
219 !dev->driver->err_handler->error_detected) {
220 if (result_data->state == pci_channel_io_frozen &&
221 !(dev->hdr_type & PCI_HEADER_TYPE_BRIDGE)) {
222 /*
223 * In case of fatal recovery, if one of down-
224 * stream device has no driver. We might be
225 * unable to recover because a later insmod
226 * of a driver for this device is unaware of
227 * its hw state.
228 */
531f254e
BH
229 dev_printk(KERN_DEBUG, &dev->dev, "device has %s\n",
230 dev->driver ?
231 "no AER-aware driver" : "no driver");
6c2b374d 232 }
918b4053
VMP
233
234 /*
235 * If there's any device in the subtree that does not
236 * have an error_detected callback, returning
237 * PCI_ERS_RESULT_NO_AER_DRIVER prevents calling of
238 * the subsequent mmio_enabled/slot_reset/resume
239 * callbacks of "any" device in the subtree. All the
240 * devices in the subtree are left in the error state
241 * without recovery.
242 */
243
244 if (!(dev->hdr_type & PCI_HEADER_TYPE_BRIDGE))
245 vote = PCI_ERS_RESULT_NO_AER_DRIVER;
246 else
247 vote = PCI_ERS_RESULT_NONE;
248 } else {
249 err_handler = dev->driver->err_handler;
250 vote = err_handler->error_detected(dev, result_data->state);
6c2b374d
ZY
251 }
252
6c2b374d 253 result_data->result = merge_result(result_data->result, vote);
90b5c1d7 254 device_unlock(&dev->dev);
70298c6e 255 return 0;
6c2b374d
ZY
256}
257
70298c6e 258static int report_mmio_enabled(struct pci_dev *dev, void *data)
6c2b374d
ZY
259{
260 pci_ers_result_t vote;
49453028 261 const struct pci_error_handlers *err_handler;
6c2b374d
ZY
262 struct aer_broadcast_data *result_data;
263 result_data = (struct aer_broadcast_data *) data;
264
90b5c1d7 265 device_lock(&dev->dev);
6c2b374d
ZY
266 if (!dev->driver ||
267 !dev->driver->err_handler ||
268 !dev->driver->err_handler->mmio_enabled)
90b5c1d7 269 goto out;
6c2b374d
ZY
270
271 err_handler = dev->driver->err_handler;
272 vote = err_handler->mmio_enabled(dev);
273 result_data->result = merge_result(result_data->result, vote);
90b5c1d7
HY
274out:
275 device_unlock(&dev->dev);
70298c6e 276 return 0;
6c2b374d
ZY
277}
278
70298c6e 279static int report_slot_reset(struct pci_dev *dev, void *data)
6c2b374d
ZY
280{
281 pci_ers_result_t vote;
49453028 282 const struct pci_error_handlers *err_handler;
6c2b374d
ZY
283 struct aer_broadcast_data *result_data;
284 result_data = (struct aer_broadcast_data *) data;
285
90b5c1d7 286 device_lock(&dev->dev);
6c2b374d
ZY
287 if (!dev->driver ||
288 !dev->driver->err_handler ||
289 !dev->driver->err_handler->slot_reset)
90b5c1d7 290 goto out;
6c2b374d
ZY
291
292 err_handler = dev->driver->err_handler;
293 vote = err_handler->slot_reset(dev);
294 result_data->result = merge_result(result_data->result, vote);
90b5c1d7
HY
295out:
296 device_unlock(&dev->dev);
70298c6e 297 return 0;
6c2b374d
ZY
298}
299
70298c6e 300static int report_resume(struct pci_dev *dev, void *data)
6c2b374d 301{
49453028 302 const struct pci_error_handlers *err_handler;
6c2b374d 303
90b5c1d7 304 device_lock(&dev->dev);
6c2b374d
ZY
305 dev->error_state = pci_channel_io_normal;
306
307 if (!dev->driver ||
308 !dev->driver->err_handler ||
b0b801dd 309 !dev->driver->err_handler->resume)
90b5c1d7 310 goto out;
6c2b374d
ZY
311
312 err_handler = dev->driver->err_handler;
313 err_handler->resume(dev);
90b5c1d7
HY
314out:
315 device_unlock(&dev->dev);
70298c6e 316 return 0;
6c2b374d
ZY
317}
318
319/**
320 * broadcast_error_message - handle message broadcast to downstream drivers
d885c6b7 321 * @dev: pointer to from where in a hierarchy message is broadcasted down
6c2b374d 322 * @state: error state
d885c6b7
RD
323 * @error_mesg: message to print
324 * @cb: callback to be broadcasted
6c2b374d
ZY
325 *
326 * Invoked during error recovery process. Once being invoked, the content
327 * of error severity will be broadcasted to all downstream drivers in a
328 * hierarchy in question.
d885c6b7 329 */
6c2b374d
ZY
330static pci_ers_result_t broadcast_error_message(struct pci_dev *dev,
331 enum pci_channel_state state,
332 char *error_mesg,
70298c6e 333 int (*cb)(struct pci_dev *, void *))
6c2b374d
ZY
334{
335 struct aer_broadcast_data result_data;
336
531f254e 337 dev_printk(KERN_DEBUG, &dev->dev, "broadcast %s message\n", error_mesg);
6c2b374d
ZY
338 result_data.state = state;
339 if (cb == report_error_detected)
340 result_data.result = PCI_ERS_RESULT_CAN_RECOVER;
341 else
342 result_data.result = PCI_ERS_RESULT_RECOVERED;
343
344 if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE) {
345 /*
346 * If the error is reported by a bridge, we think this error
347 * is related to the downstream link of the bridge, so we
348 * do error recovery on all subordinates of the bridge instead
349 * of the bridge and clear the error status of the bridge.
350 */
351 if (cb == report_error_detected)
352 dev->error_state = state;
353 pci_walk_bus(dev->subordinate, cb, &result_data);
354 if (cb == report_resume) {
355 pci_cleanup_aer_uncorrect_error_status(dev);
356 dev->error_state = pci_channel_io_normal;
357 }
c9a91883 358 } else {
6c2b374d
ZY
359 /*
360 * If the error is reported by an end point, we think this
361 * error is related to the upstream link of the end point.
362 */
363 pci_walk_bus(dev->bus, cb, &result_data);
364 }
365
366 return result_data.result;
367}
368
89713422 369/**
081d0fe0
BD
370 * default_reset_link - default reset function
371 * @dev: pointer to pci_dev data structure
89713422 372 *
081d0fe0
BD
373 * Invoked when performing link reset on a Downstream Port or a
374 * Root Port with no aer driver.
89713422 375 */
081d0fe0 376static pci_ers_result_t default_reset_link(struct pci_dev *dev)
89713422 377{
1b95ce8f 378 pci_reset_bridge_secondary_bus(dev);
081d0fe0 379 dev_printk(KERN_DEBUG, &dev->dev, "downstream link has been reset\n");
89713422
HS
380 return PCI_ERS_RESULT_RECOVERED;
381}
382
6c2b374d
ZY
383static int find_aer_service_iter(struct device *device, void *data)
384{
517cae38 385 struct pcie_port_service_driver *service_driver, **drv;
6c2b374d 386
517cae38 387 drv = (struct pcie_port_service_driver **) data;
6c2b374d 388
4f7ccf6a
HS
389 if (device->bus == &pcie_port_bus_type && device->driver) {
390 service_driver = to_service_driver(device->driver);
391 if (service_driver->service == PCIE_PORT_SERVICE_AER) {
517cae38 392 *drv = service_driver;
4f7ccf6a 393 return 1;
6c2b374d
ZY
394 }
395 }
396
397 return 0;
398}
399
517cae38 400static struct pcie_port_service_driver *find_aer_service(struct pci_dev *dev)
6c2b374d 401{
517cae38
HS
402 struct pcie_port_service_driver *drv = NULL;
403
404 device_for_each_child(&dev->dev, &drv, find_aer_service_iter);
405
406 return drv;
6c2b374d
ZY
407}
408
0918472c 409static pci_ers_result_t reset_link(struct pci_dev *dev)
6c2b374d
ZY
410{
411 struct pci_dev *udev;
412 pci_ers_result_t status;
517cae38 413 struct pcie_port_service_driver *driver;
6c2b374d 414
89713422
HS
415 if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE) {
416 /* Reset this port for all subordinates */
6c2b374d 417 udev = dev;
89713422
HS
418 } else {
419 /* Reset the upstream component (likely downstream port) */
c9a91883 420 udev = dev->bus->self;
89713422 421 }
6c2b374d 422
517cae38
HS
423 /* Use the aer driver of the component firstly */
424 driver = find_aer_service(udev);
6c2b374d 425
89713422
HS
426 if (driver && driver->reset_link) {
427 status = driver->reset_link(udev);
777e61ea 428 } else if (udev->has_secondary_link) {
081d0fe0 429 status = default_reset_link(udev);
89713422
HS
430 } else {
431 dev_printk(KERN_DEBUG, &dev->dev,
432 "no link-reset support at upstream device %s\n",
433 pci_name(udev));
434 return PCI_ERS_RESULT_DISCONNECT;
6c2b374d
ZY
435 }
436
6c2b374d 437 if (status != PCI_ERS_RESULT_RECOVERED) {
4f7ccf6a
HS
438 dev_printk(KERN_DEBUG, &dev->dev,
439 "link reset at upstream device %s failed\n",
440 pci_name(udev));
6c2b374d
ZY
441 return PCI_ERS_RESULT_DISCONNECT;
442 }
443
444 return status;
445}
446
447/**
448 * do_recovery - handle nonfatal/fatal error recovery process
6c2b374d
ZY
449 * @dev: pointer to a pci_dev data structure of agent detecting an error
450 * @severity: error severity type
451 *
452 * Invoked when an error is nonfatal/fatal. Once being invoked, broadcast
453 * error detected message to all downstream drivers within a hierarchy in
454 * question and return the returned code.
d885c6b7 455 */
0918472c 456static void do_recovery(struct pci_dev *dev, int severity)
6c2b374d
ZY
457{
458 pci_ers_result_t status, result = PCI_ERS_RESULT_RECOVERED;
459 enum pci_channel_state state;
460
461 if (severity == AER_FATAL)
462 state = pci_channel_io_frozen;
463 else
464 state = pci_channel_io_normal;
465
466 status = broadcast_error_message(dev,
467 state,
468 "error_detected",
469 report_error_detected);
470
471 if (severity == AER_FATAL) {
0918472c 472 result = reset_link(dev);
17e21854
HS
473 if (result != PCI_ERS_RESULT_RECOVERED)
474 goto failed;
6c2b374d
ZY
475 }
476
477 if (status == PCI_ERS_RESULT_CAN_RECOVER)
478 status = broadcast_error_message(dev,
479 state,
480 "mmio_enabled",
481 report_mmio_enabled);
482
483 if (status == PCI_ERS_RESULT_NEED_RESET) {
484 /*
485 * TODO: Should call platform-specific
486 * functions to reset slot before calling
487 * drivers' slot_reset callbacks?
488 */
489 status = broadcast_error_message(dev,
490 state,
491 "slot_reset",
492 report_slot_reset);
493 }
494
17e21854
HS
495 if (status != PCI_ERS_RESULT_RECOVERED)
496 goto failed;
497
498 broadcast_error_message(dev,
6c2b374d
ZY
499 state,
500 "resume",
501 report_resume);
502
be5ac3d3 503 dev_info(&dev->dev, "AER: Device recovery successful\n");
17e21854
HS
504 return;
505
506failed:
507 /* TODO: Should kernel panic here? */
be5ac3d3 508 dev_info(&dev->dev, "AER: Device recovery failed\n");
6c2b374d
ZY
509}
510
511/**
512 * handle_error_source - handle logging error into an event log
513 * @aerdev: pointer to pcie_device data structure of the root port
514 * @dev: pointer to pci_dev data structure of error source device
515 * @info: comprehensive error information
516 *
517 * Invoked when an error being detected by Root Port.
d885c6b7 518 */
c9a91883 519static void handle_error_source(struct pcie_device *aerdev,
6c2b374d 520 struct pci_dev *dev,
28eb27cf 521 struct aer_err_info *info)
6c2b374d 522{
6c2b374d
ZY
523 int pos;
524
28eb27cf 525 if (info->severity == AER_CORRECTABLE) {
6c2b374d 526 /*
f7625980 527 * Correctable error does not need software intervention.
6c2b374d
ZY
528 * No need to go through error recovery process.
529 */
0927678f 530 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
6c2b374d
ZY
531 if (pos)
532 pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS,
28eb27cf 533 info->status);
17e21854 534 } else
0918472c 535 do_recovery(dev, info->severity);
6c2b374d
ZY
536}
537
0918472c
HY
538#ifdef CONFIG_ACPI_APEI_PCIEAER
539static void aer_recover_work_func(struct work_struct *work);
540
541#define AER_RECOVER_RING_ORDER 4
542#define AER_RECOVER_RING_SIZE (1 << AER_RECOVER_RING_ORDER)
543
3c78bc61 544struct aer_recover_entry {
0918472c
HY
545 u8 bus;
546 u8 devfn;
547 u16 domain;
548 int severity;
37448adf 549 struct aer_capability_regs *regs;
0918472c
HY
550};
551
552static DEFINE_KFIFO(aer_recover_ring, struct aer_recover_entry,
553 AER_RECOVER_RING_SIZE);
554/*
555 * Mutual exclusion for writers of aer_recover_ring, reader side don't
556 * need lock, because there is only one reader and lock is not needed
557 * between reader and writer.
558 */
559static DEFINE_SPINLOCK(aer_recover_ring_lock);
560static DECLARE_WORK(aer_recover_work, aer_recover_work_func);
561
562void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn,
37448adf 563 int severity, struct aer_capability_regs *aer_regs)
0918472c
HY
564{
565 unsigned long flags;
566 struct aer_recover_entry entry = {
567 .bus = bus,
568 .devfn = devfn,
569 .domain = domain,
570 .severity = severity,
37448adf 571 .regs = aer_regs,
0918472c
HY
572 };
573
574 spin_lock_irqsave(&aer_recover_ring_lock, flags);
498d319b 575 if (kfifo_put(&aer_recover_ring, entry))
0918472c
HY
576 schedule_work(&aer_recover_work);
577 else
578 pr_err("AER recover: Buffer overflow when recovering AER for %04x:%02x:%02x:%x\n",
579 domain, bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
580 spin_unlock_irqrestore(&aer_recover_ring_lock, flags);
581}
582EXPORT_SYMBOL_GPL(aer_recover_queue);
583
584static void aer_recover_work_func(struct work_struct *work)
585{
586 struct aer_recover_entry entry;
587 struct pci_dev *pdev;
588
589 while (kfifo_get(&aer_recover_ring, &entry)) {
590 pdev = pci_get_domain_bus_and_slot(entry.domain, entry.bus,
591 entry.devfn);
592 if (!pdev) {
593 pr_err("AER recover: Can not find pci_dev for %04x:%02x:%02x:%x\n",
594 entry.domain, entry.bus,
595 PCI_SLOT(entry.devfn), PCI_FUNC(entry.devfn));
596 continue;
597 }
37448adf 598 cper_print_aer(pdev, entry.severity, entry.regs);
0918472c 599 do_recovery(pdev, entry.severity);
a82b6af3 600 pci_dev_put(pdev);
0918472c
HY
601 }
602}
603#endif
604
b1c089b7
HS
605/**
606 * get_device_error_info - read error status from dev and store it to info
607 * @dev: pointer to the device expected to have a error record
608 * @info: pointer to structure to store the error record
609 *
610 * Return 1 on success, 0 on error.
7c4ec94f
HS
611 *
612 * Note that @info is reused among all error devices. Clear fields properly.
b1c089b7 613 */
6c2b374d
ZY
614static int get_device_error_info(struct pci_dev *dev, struct aer_err_info *info)
615{
e7a0d92b 616 int pos, temp;
6c2b374d 617
7c4ec94f 618 /* Must reset in this function */
1b4ffcf8 619 info->status = 0;
273024de 620 info->tlp_header_valid = 0;
1b4ffcf8 621
0927678f 622 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
6c2b374d
ZY
623
624 /* The device might not support AER */
625 if (!pos)
b1c089b7 626 return 1;
6c2b374d
ZY
627
628 if (info->severity == AER_CORRECTABLE) {
629 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS,
630 &info->status);
0d90c3ac
HS
631 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK,
632 &info->mask);
633 if (!(info->status & ~info->mask))
b1c089b7 634 return 0;
6c2b374d
ZY
635 } else if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE ||
636 info->severity == AER_NONFATAL) {
637
638 /* Link is still healthy for IO reads */
639 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS,
640 &info->status);
0d90c3ac
HS
641 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK,
642 &info->mask);
643 if (!(info->status & ~info->mask))
b1c089b7 644 return 0;
6c2b374d 645
e7a0d92b
HS
646 /* Get First Error Pointer */
647 pci_read_config_dword(dev, pos + PCI_ERR_CAP, &temp);
273024de 648 info->first_error = PCI_ERR_CAP_FEP(temp);
e7a0d92b 649
6c2b374d 650 if (info->status & AER_LOG_TLP_MASKS) {
273024de 651 info->tlp_header_valid = 1;
6c2b374d
ZY
652 pci_read_config_dword(dev,
653 pos + PCI_ERR_HEADER_LOG, &info->tlp.dw0);
654 pci_read_config_dword(dev,
655 pos + PCI_ERR_HEADER_LOG + 4, &info->tlp.dw1);
656 pci_read_config_dword(dev,
657 pos + PCI_ERR_HEADER_LOG + 8, &info->tlp.dw2);
658 pci_read_config_dword(dev,
659 pos + PCI_ERR_HEADER_LOG + 12, &info->tlp.dw3);
660 }
661 }
662
b1c089b7 663 return 1;
6c2b374d
ZY
664}
665
3d5505c5
ZY
666static inline void aer_process_err_devices(struct pcie_device *p_device,
667 struct aer_err_info *e_info)
668{
669 int i;
670
b1c089b7 671 /* Report all before handle them, not to lost records by reset etc. */
3d5505c5 672 for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
b1c089b7 673 if (get_device_error_info(e_info->dev[i], e_info))
3d5505c5 674 aer_print_error(e_info->dev[i], e_info);
b1c089b7
HS
675 }
676 for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
677 if (get_device_error_info(e_info->dev[i], e_info))
678 handle_error_source(p_device, e_info->dev[i], e_info);
3d5505c5
ZY
679 }
680}
681
6c2b374d
ZY
682/**
683 * aer_isr_one_error - consume an error detected by root port
684 * @p_device: pointer to error root port service device
685 * @e_src: pointer to an error source
d885c6b7 686 */
6c2b374d
ZY
687static void aer_isr_one_error(struct pcie_device *p_device,
688 struct aer_err_source *e_src)
689{
28eb27cf 690 struct aer_err_info *e_info;
28eb27cf
ZY
691
692 /* struct aer_err_info might be big, so we allocate it with slab */
693 e_info = kmalloc(sizeof(struct aer_err_info), GFP_KERNEL);
7c4ec94f 694 if (!e_info) {
28eb27cf
ZY
695 dev_printk(KERN_DEBUG, &p_device->port->dev,
696 "Can't allocate mem when processing AER errors\n");
697 return;
698 }
6c2b374d
ZY
699
700 /*
701 * There is a possibility that both correctable error and
702 * uncorrectable error being logged. Report correctable error first.
703 */
7c4ec94f
HS
704 if (e_src->status & PCI_ERR_ROOT_COR_RCV) {
705 e_info->id = ERR_COR_ID(e_src->id);
706 e_info->severity = AER_CORRECTABLE;
707
708 if (e_src->status & PCI_ERR_ROOT_MULTI_COR_RCV)
709 e_info->multi_error_valid = 1;
710 else
711 e_info->multi_error_valid = 0;
712
713 aer_print_port_info(p_device->port, e_info);
714
715 if (find_source_device(p_device->port, e_info))
716 aer_process_err_devices(p_device, e_info);
717 }
718
719 if (e_src->status & PCI_ERR_ROOT_UNCOR_RCV) {
720 e_info->id = ERR_UNCOR_ID(e_src->id);
721
722 if (e_src->status & PCI_ERR_ROOT_FATAL_RCV)
723 e_info->severity = AER_FATAL;
724 else
725 e_info->severity = AER_NONFATAL;
726
727 if (e_src->status & PCI_ERR_ROOT_MULTI_UNCOR_RCV)
273024de 728 e_info->multi_error_valid = 1;
7c4ec94f
HS
729 else
730 e_info->multi_error_valid = 0;
28eb27cf 731
79e4b89b
HS
732 aer_print_port_info(p_device->port, e_info);
733
98ca3964
HS
734 if (find_source_device(p_device->port, e_info))
735 aer_process_err_devices(p_device, e_info);
6c2b374d 736 }
28eb27cf
ZY
737
738 kfree(e_info);
6c2b374d
ZY
739}
740
88da13bf
HS
741/**
742 * get_e_source - retrieve an error source
743 * @rpc: pointer to the root port which holds an error
744 * @e_src: pointer to store retrieved error source
745 *
746 * Return 1 if an error source is retrieved, otherwise 0.
747 *
748 * Invoked by DPC handler to consume an error.
749 */
750static int get_e_source(struct aer_rpc *rpc, struct aer_err_source *e_src)
751{
752 unsigned long flags;
88da13bf
HS
753
754 /* Lock access to Root error producer/consumer index */
755 spin_lock_irqsave(&rpc->e_lock, flags);
f6735590
LT
756 if (rpc->prod_idx == rpc->cons_idx) {
757 spin_unlock_irqrestore(&rpc->e_lock, flags);
758 return 0;
88da13bf 759 }
f6735590
LT
760
761 *e_src = rpc->e_sources[rpc->cons_idx];
762 rpc->cons_idx++;
763 if (rpc->cons_idx == AER_ERROR_SOURCES_MAX)
764 rpc->cons_idx = 0;
88da13bf
HS
765 spin_unlock_irqrestore(&rpc->e_lock, flags);
766
f6735590 767 return 1;
88da13bf
HS
768}
769
6c2b374d
ZY
770/**
771 * aer_isr - consume errors detected by root port
65f27f38 772 * @work: definition of this work item
6c2b374d
ZY
773 *
774 * Invoked, as DPC, when root port records new detected error
d885c6b7 775 */
65f27f38 776void aer_isr(struct work_struct *work)
6c2b374d 777{
65f27f38
DH
778 struct aer_rpc *rpc = container_of(work, struct aer_rpc, dpc_handler);
779 struct pcie_device *p_device = rpc->rpd;
50c1126e 780 struct aer_err_source uninitialized_var(e_src);
6c2b374d
ZY
781
782 mutex_lock(&rpc->rpc_mutex);
88da13bf
HS
783 while (get_e_source(rpc, &e_src))
784 aer_isr_one_error(p_device, &e_src);
6c2b374d
ZY
785 mutex_unlock(&rpc->rpc_mutex);
786
787 wake_up(&rpc->wait_release);
788}
789
6c2b374d
ZY
790/**
791 * aer_init - provide AER initialization
792 * @dev: pointer to AER pcie device
793 *
794 * Invoked when AER service driver is loaded.
d885c6b7 795 */
6c2b374d
ZY
796int aer_init(struct pcie_device *dev)
797{
05843961
MD
798 if (forceload) {
799 dev_printk(KERN_DEBUG, &dev->device,
800 "aerdrv forceload requested.\n");
affb72c3 801 pcie_aer_force_firmware_first(dev->port, 0);
05843961 802 }
28eb5f27 803 return 0;
6c2b374d 804}