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>
25 #include <linux/delay.h>
26 #include <linux/slab.h>
27 #include <linux/kfifo.h>
28 #include "aerdrv.h"
29
30 static bool forceload;
31 static bool nosourceid;
32 module_param(forceload, bool, 0);
33 module_param(nosourceid, bool, 0);
34
35 #define PCI_EXP_AER_FLAGS (PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE | \
36 PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE)
37
pci_enable_pcie_error_reporting(struct pci_dev * dev)38 int pci_enable_pcie_error_reporting(struct pci_dev *dev)
39 {
40 if (pcie_aer_get_firmware_first(dev))
41 return -EIO;
42
43 if (!pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR))
44 return -EIO;
45
46 return pcie_capability_set_word(dev, PCI_EXP_DEVCTL, PCI_EXP_AER_FLAGS);
47 }
48 EXPORT_SYMBOL_GPL(pci_enable_pcie_error_reporting);
49
pci_disable_pcie_error_reporting(struct pci_dev * dev)50 int pci_disable_pcie_error_reporting(struct pci_dev *dev)
51 {
52 if (pcie_aer_get_firmware_first(dev))
53 return -EIO;
54
55 return pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
56 PCI_EXP_AER_FLAGS);
57 }
58 EXPORT_SYMBOL_GPL(pci_disable_pcie_error_reporting);
59
pci_cleanup_aer_uncorrect_error_status(struct pci_dev * dev)60 int pci_cleanup_aer_uncorrect_error_status(struct pci_dev *dev)
61 {
62 int pos;
63 u32 status;
64
65 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
66 if (!pos)
67 return -EIO;
68
69 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
70 if (status)
71 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
72
73 return 0;
74 }
75 EXPORT_SYMBOL_GPL(pci_cleanup_aer_uncorrect_error_status);
76
pci_cleanup_aer_error_status_regs(struct pci_dev * dev)77 int pci_cleanup_aer_error_status_regs(struct pci_dev *dev)
78 {
79 int pos;
80 u32 status;
81 int port_type;
82
83 if (!pci_is_pcie(dev))
84 return -ENODEV;
85
86 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
87 if (!pos)
88 return -EIO;
89
90 port_type = pci_pcie_type(dev);
91 if (port_type == PCI_EXP_TYPE_ROOT_PORT) {
92 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, &status);
93 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, status);
94 }
95
96 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
97 pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS, status);
98
99 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
100 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
101
102 return 0;
103 }
104
105 /**
106 * add_error_device - list device to be handled
107 * @e_info: pointer to error info
108 * @dev: pointer to pci_dev to be added
109 */
add_error_device(struct aer_err_info * e_info,struct pci_dev * dev)110 static int add_error_device(struct aer_err_info *e_info, struct pci_dev *dev)
111 {
112 if (e_info->error_dev_num < AER_MAX_MULTI_ERR_DEVICES) {
113 e_info->dev[e_info->error_dev_num] = dev;
114 e_info->error_dev_num++;
115 return 0;
116 }
117 return -ENOSPC;
118 }
119
120 /**
121 * is_error_source - check whether the device is source of reported error
122 * @dev: pointer to pci_dev to be checked
123 * @e_info: pointer to reported error info
124 */
is_error_source(struct pci_dev * dev,struct aer_err_info * e_info)125 static bool is_error_source(struct pci_dev *dev, struct aer_err_info *e_info)
126 {
127 int pos;
128 u32 status, mask;
129 u16 reg16;
130
131 /*
132 * When bus id is equal to 0, it might be a bad id
133 * reported by root port.
134 */
135 if (!nosourceid && (PCI_BUS_NUM(e_info->id) != 0)) {
136 /* Device ID match? */
137 if (e_info->id == ((dev->bus->number << 8) | dev->devfn))
138 return true;
139
140 /* Continue id comparing if there is no multiple error */
141 if (!e_info->multi_error_valid)
142 return false;
143 }
144
145 /*
146 * When either
147 * 1) nosourceid==y;
148 * 2) bus id is equal to 0. Some ports might lose the bus
149 * id of error source id;
150 * 3) There are multiple errors and prior id comparing fails;
151 * We check AER status registers to find possible reporter.
152 */
153 if (atomic_read(&dev->enable_cnt) == 0)
154 return false;
155
156 /* Check if AER is enabled */
157 pcie_capability_read_word(dev, PCI_EXP_DEVCTL, ®16);
158 if (!(reg16 & PCI_EXP_AER_FLAGS))
159 return false;
160
161 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
162 if (!pos)
163 return false;
164
165 /* Check if error is recorded */
166 if (e_info->severity == AER_CORRECTABLE) {
167 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
168 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, &mask);
169 } else {
170 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
171 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, &mask);
172 }
173 if (status & ~mask)
174 return true;
175
176 return false;
177 }
178
find_device_iter(struct pci_dev * dev,void * data)179 static int find_device_iter(struct pci_dev *dev, void *data)
180 {
181 struct aer_err_info *e_info = (struct aer_err_info *)data;
182
183 if (is_error_source(dev, e_info)) {
184 /* List this device */
185 if (add_error_device(e_info, dev)) {
186 /* We cannot handle more... Stop iteration */
187 /* TODO: Should print error message here? */
188 return 1;
189 }
190
191 /* If there is only a single error, stop iteration */
192 if (!e_info->multi_error_valid)
193 return 1;
194 }
195 return 0;
196 }
197
198 /**
199 * find_source_device - search through device hierarchy for source device
200 * @parent: pointer to Root Port pci_dev data structure
201 * @e_info: including detailed error information such like id
202 *
203 * Return true if found.
204 *
205 * Invoked by DPC when error is detected at the Root Port.
206 * Caller of this function must set id, severity, and multi_error_valid of
207 * struct aer_err_info pointed by @e_info properly. This function must fill
208 * e_info->error_dev_num and e_info->dev[], based on the given information.
209 */
find_source_device(struct pci_dev * parent,struct aer_err_info * e_info)210 static bool find_source_device(struct pci_dev *parent,
211 struct aer_err_info *e_info)
212 {
213 struct pci_dev *dev = parent;
214 int result;
215
216 /* Must reset in this function */
217 e_info->error_dev_num = 0;
218
219 /* Is Root Port an agent that sends error message? */
220 result = find_device_iter(dev, e_info);
221 if (result)
222 return true;
223
224 pci_walk_bus(parent->subordinate, find_device_iter, e_info);
225
226 if (!e_info->error_dev_num) {
227 dev_printk(KERN_DEBUG, &parent->dev,
228 "can't find device of ID%04x\n",
229 e_info->id);
230 return false;
231 }
232 return true;
233 }
234
report_error_detected(struct pci_dev * dev,void * data)235 static int report_error_detected(struct pci_dev *dev, void *data)
236 {
237 pci_ers_result_t vote;
238 const struct pci_error_handlers *err_handler;
239 struct aer_broadcast_data *result_data;
240 result_data = (struct aer_broadcast_data *) data;
241
242 device_lock(&dev->dev);
243 dev->error_state = result_data->state;
244
245 if (!dev->driver ||
246 !dev->driver->err_handler ||
247 !dev->driver->err_handler->error_detected) {
248 if (result_data->state == pci_channel_io_frozen &&
249 !(dev->hdr_type & PCI_HEADER_TYPE_BRIDGE)) {
250 /*
251 * In case of fatal recovery, if one of down-
252 * stream device has no driver. We might be
253 * unable to recover because a later insmod
254 * of a driver for this device is unaware of
255 * its hw state.
256 */
257 dev_printk(KERN_DEBUG, &dev->dev, "device has %s\n",
258 dev->driver ?
259 "no AER-aware driver" : "no driver");
260 }
261
262 /*
263 * If there's any device in the subtree that does not
264 * have an error_detected callback, returning
265 * PCI_ERS_RESULT_NO_AER_DRIVER prevents calling of
266 * the subsequent mmio_enabled/slot_reset/resume
267 * callbacks of "any" device in the subtree. All the
268 * devices in the subtree are left in the error state
269 * without recovery.
270 */
271
272 if (!(dev->hdr_type & PCI_HEADER_TYPE_BRIDGE))
273 vote = PCI_ERS_RESULT_NO_AER_DRIVER;
274 else
275 vote = PCI_ERS_RESULT_NONE;
276 } else {
277 err_handler = dev->driver->err_handler;
278 vote = err_handler->error_detected(dev, result_data->state);
279 }
280
281 result_data->result = merge_result(result_data->result, vote);
282 device_unlock(&dev->dev);
283 return 0;
284 }
285
report_mmio_enabled(struct pci_dev * dev,void * data)286 static int report_mmio_enabled(struct pci_dev *dev, void *data)
287 {
288 pci_ers_result_t vote;
289 const struct pci_error_handlers *err_handler;
290 struct aer_broadcast_data *result_data;
291 result_data = (struct aer_broadcast_data *) data;
292
293 device_lock(&dev->dev);
294 if (!dev->driver ||
295 !dev->driver->err_handler ||
296 !dev->driver->err_handler->mmio_enabled)
297 goto out;
298
299 err_handler = dev->driver->err_handler;
300 vote = err_handler->mmio_enabled(dev);
301 result_data->result = merge_result(result_data->result, vote);
302 out:
303 device_unlock(&dev->dev);
304 return 0;
305 }
306
report_slot_reset(struct pci_dev * dev,void * data)307 static int report_slot_reset(struct pci_dev *dev, void *data)
308 {
309 pci_ers_result_t vote;
310 const struct pci_error_handlers *err_handler;
311 struct aer_broadcast_data *result_data;
312 result_data = (struct aer_broadcast_data *) data;
313
314 device_lock(&dev->dev);
315 if (!dev->driver ||
316 !dev->driver->err_handler ||
317 !dev->driver->err_handler->slot_reset)
318 goto out;
319
320 err_handler = dev->driver->err_handler;
321 vote = err_handler->slot_reset(dev);
322 result_data->result = merge_result(result_data->result, vote);
323 out:
324 device_unlock(&dev->dev);
325 return 0;
326 }
327
report_resume(struct pci_dev * dev,void * data)328 static int report_resume(struct pci_dev *dev, void *data)
329 {
330 const struct pci_error_handlers *err_handler;
331
332 device_lock(&dev->dev);
333 dev->error_state = pci_channel_io_normal;
334
335 if (!dev->driver ||
336 !dev->driver->err_handler ||
337 !dev->driver->err_handler->resume)
338 goto out;
339
340 err_handler = dev->driver->err_handler;
341 err_handler->resume(dev);
342 out:
343 device_unlock(&dev->dev);
344 return 0;
345 }
346
347 /**
348 * broadcast_error_message - handle message broadcast to downstream drivers
349 * @dev: pointer to from where in a hierarchy message is broadcasted down
350 * @state: error state
351 * @error_mesg: message to print
352 * @cb: callback to be broadcasted
353 *
354 * Invoked during error recovery process. Once being invoked, the content
355 * of error severity will be broadcasted to all downstream drivers in a
356 * hierarchy in question.
357 */
broadcast_error_message(struct pci_dev * dev,enum pci_channel_state state,char * error_mesg,int (* cb)(struct pci_dev *,void *))358 static pci_ers_result_t broadcast_error_message(struct pci_dev *dev,
359 enum pci_channel_state state,
360 char *error_mesg,
361 int (*cb)(struct pci_dev *, void *))
362 {
363 struct aer_broadcast_data result_data;
364
365 dev_printk(KERN_DEBUG, &dev->dev, "broadcast %s message\n", error_mesg);
366 result_data.state = state;
367 if (cb == report_error_detected)
368 result_data.result = PCI_ERS_RESULT_CAN_RECOVER;
369 else
370 result_data.result = PCI_ERS_RESULT_RECOVERED;
371
372 if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE) {
373 /*
374 * If the error is reported by a bridge, we think this error
375 * is related to the downstream link of the bridge, so we
376 * do error recovery on all subordinates of the bridge instead
377 * of the bridge and clear the error status of the bridge.
378 */
379 if (cb == report_error_detected)
380 dev->error_state = state;
381 pci_walk_bus(dev->subordinate, cb, &result_data);
382 if (cb == report_resume) {
383 pci_cleanup_aer_uncorrect_error_status(dev);
384 dev->error_state = pci_channel_io_normal;
385 }
386 } else {
387 /*
388 * If the error is reported by an end point, we think this
389 * error is related to the upstream link of the end point.
390 */
391 pci_walk_bus(dev->bus, cb, &result_data);
392 }
393
394 return result_data.result;
395 }
396
397 /**
398 * default_reset_link - default reset function
399 * @dev: pointer to pci_dev data structure
400 *
401 * Invoked when performing link reset on a Downstream Port or a
402 * Root Port with no aer driver.
403 */
default_reset_link(struct pci_dev * dev)404 static pci_ers_result_t default_reset_link(struct pci_dev *dev)
405 {
406 pci_reset_bridge_secondary_bus(dev);
407 dev_printk(KERN_DEBUG, &dev->dev, "downstream link has been reset\n");
408 return PCI_ERS_RESULT_RECOVERED;
409 }
410
find_aer_service_iter(struct device * device,void * data)411 static int find_aer_service_iter(struct device *device, void *data)
412 {
413 struct pcie_port_service_driver *service_driver, **drv;
414
415 drv = (struct pcie_port_service_driver **) data;
416
417 if (device->bus == &pcie_port_bus_type && device->driver) {
418 service_driver = to_service_driver(device->driver);
419 if (service_driver->service == PCIE_PORT_SERVICE_AER) {
420 *drv = service_driver;
421 return 1;
422 }
423 }
424
425 return 0;
426 }
427
find_aer_service(struct pci_dev * dev)428 static struct pcie_port_service_driver *find_aer_service(struct pci_dev *dev)
429 {
430 struct pcie_port_service_driver *drv = NULL;
431
432 device_for_each_child(&dev->dev, &drv, find_aer_service_iter);
433
434 return drv;
435 }
436
reset_link(struct pci_dev * dev)437 static pci_ers_result_t reset_link(struct pci_dev *dev)
438 {
439 struct pci_dev *udev;
440 pci_ers_result_t status;
441 struct pcie_port_service_driver *driver;
442
443 if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE) {
444 /* Reset this port for all subordinates */
445 udev = dev;
446 } else {
447 /* Reset the upstream component (likely downstream port) */
448 udev = dev->bus->self;
449 }
450
451 /* Use the aer driver of the component firstly */
452 driver = find_aer_service(udev);
453
454 if (driver && driver->reset_link) {
455 status = driver->reset_link(udev);
456 } else if (udev->has_secondary_link) {
457 status = default_reset_link(udev);
458 } else {
459 dev_printk(KERN_DEBUG, &dev->dev,
460 "no link-reset support at upstream device %s\n",
461 pci_name(udev));
462 return PCI_ERS_RESULT_DISCONNECT;
463 }
464
465 if (status != PCI_ERS_RESULT_RECOVERED) {
466 dev_printk(KERN_DEBUG, &dev->dev,
467 "link reset at upstream device %s failed\n",
468 pci_name(udev));
469 return PCI_ERS_RESULT_DISCONNECT;
470 }
471
472 return status;
473 }
474
475 /**
476 * do_recovery - handle nonfatal/fatal error recovery process
477 * @dev: pointer to a pci_dev data structure of agent detecting an error
478 * @severity: error severity type
479 *
480 * Invoked when an error is nonfatal/fatal. Once being invoked, broadcast
481 * error detected message to all downstream drivers within a hierarchy in
482 * question and return the returned code.
483 */
do_recovery(struct pci_dev * dev,int severity)484 static void do_recovery(struct pci_dev *dev, int severity)
485 {
486 pci_ers_result_t status, result = PCI_ERS_RESULT_RECOVERED;
487 enum pci_channel_state state;
488
489 if (severity == AER_FATAL)
490 state = pci_channel_io_frozen;
491 else
492 state = pci_channel_io_normal;
493
494 status = broadcast_error_message(dev,
495 state,
496 "error_detected",
497 report_error_detected);
498
499 if (severity == AER_FATAL) {
500 result = reset_link(dev);
501 if (result != PCI_ERS_RESULT_RECOVERED)
502 goto failed;
503 }
504
505 if (status == PCI_ERS_RESULT_CAN_RECOVER)
506 status = broadcast_error_message(dev,
507 state,
508 "mmio_enabled",
509 report_mmio_enabled);
510
511 if (status == PCI_ERS_RESULT_NEED_RESET) {
512 /*
513 * TODO: Should call platform-specific
514 * functions to reset slot before calling
515 * drivers' slot_reset callbacks?
516 */
517 status = broadcast_error_message(dev,
518 state,
519 "slot_reset",
520 report_slot_reset);
521 }
522
523 if (status != PCI_ERS_RESULT_RECOVERED)
524 goto failed;
525
526 broadcast_error_message(dev,
527 state,
528 "resume",
529 report_resume);
530
531 dev_info(&dev->dev, "AER: Device recovery successful\n");
532 return;
533
534 failed:
535 /* TODO: Should kernel panic here? */
536 dev_info(&dev->dev, "AER: Device recovery failed\n");
537 }
538
539 /**
540 * handle_error_source - handle logging error into an event log
541 * @aerdev: pointer to pcie_device data structure of the root port
542 * @dev: pointer to pci_dev data structure of error source device
543 * @info: comprehensive error information
544 *
545 * Invoked when an error being detected by Root Port.
546 */
handle_error_source(struct pcie_device * aerdev,struct pci_dev * dev,struct aer_err_info * info)547 static void handle_error_source(struct pcie_device *aerdev,
548 struct pci_dev *dev,
549 struct aer_err_info *info)
550 {
551 int pos;
552
553 if (info->severity == AER_CORRECTABLE) {
554 /*
555 * Correctable error does not need software intervention.
556 * No need to go through error recovery process.
557 */
558 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
559 if (pos)
560 pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS,
561 info->status);
562 } else
563 do_recovery(dev, info->severity);
564 }
565
566 #ifdef CONFIG_ACPI_APEI_PCIEAER
567 static void aer_recover_work_func(struct work_struct *work);
568
569 #define AER_RECOVER_RING_ORDER 4
570 #define AER_RECOVER_RING_SIZE (1 << AER_RECOVER_RING_ORDER)
571
572 struct aer_recover_entry {
573 u8 bus;
574 u8 devfn;
575 u16 domain;
576 int severity;
577 struct aer_capability_regs *regs;
578 };
579
580 static DEFINE_KFIFO(aer_recover_ring, struct aer_recover_entry,
581 AER_RECOVER_RING_SIZE);
582 /*
583 * Mutual exclusion for writers of aer_recover_ring, reader side don't
584 * need lock, because there is only one reader and lock is not needed
585 * between reader and writer.
586 */
587 static DEFINE_SPINLOCK(aer_recover_ring_lock);
588 static DECLARE_WORK(aer_recover_work, aer_recover_work_func);
589
aer_recover_queue(int domain,unsigned int bus,unsigned int devfn,int severity,struct aer_capability_regs * aer_regs)590 void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn,
591 int severity, struct aer_capability_regs *aer_regs)
592 {
593 unsigned long flags;
594 struct aer_recover_entry entry = {
595 .bus = bus,
596 .devfn = devfn,
597 .domain = domain,
598 .severity = severity,
599 .regs = aer_regs,
600 };
601
602 spin_lock_irqsave(&aer_recover_ring_lock, flags);
603 if (kfifo_put(&aer_recover_ring, entry))
604 schedule_work(&aer_recover_work);
605 else
606 pr_err("AER recover: Buffer overflow when recovering AER for %04x:%02x:%02x:%x\n",
607 domain, bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
608 spin_unlock_irqrestore(&aer_recover_ring_lock, flags);
609 }
610 EXPORT_SYMBOL_GPL(aer_recover_queue);
611
aer_recover_work_func(struct work_struct * work)612 static void aer_recover_work_func(struct work_struct *work)
613 {
614 struct aer_recover_entry entry;
615 struct pci_dev *pdev;
616
617 while (kfifo_get(&aer_recover_ring, &entry)) {
618 pdev = pci_get_domain_bus_and_slot(entry.domain, entry.bus,
619 entry.devfn);
620 if (!pdev) {
621 pr_err("AER recover: Can not find pci_dev for %04x:%02x:%02x:%x\n",
622 entry.domain, entry.bus,
623 PCI_SLOT(entry.devfn), PCI_FUNC(entry.devfn));
624 continue;
625 }
626 cper_print_aer(pdev, entry.severity, entry.regs);
627 do_recovery(pdev, entry.severity);
628 pci_dev_put(pdev);
629 }
630 }
631 #endif
632
633 /**
634 * get_device_error_info - read error status from dev and store it to info
635 * @dev: pointer to the device expected to have a error record
636 * @info: pointer to structure to store the error record
637 *
638 * Return 1 on success, 0 on error.
639 *
640 * Note that @info is reused among all error devices. Clear fields properly.
641 */
get_device_error_info(struct pci_dev * dev,struct aer_err_info * info)642 static int get_device_error_info(struct pci_dev *dev, struct aer_err_info *info)
643 {
644 int pos, temp;
645
646 /* Must reset in this function */
647 info->status = 0;
648 info->tlp_header_valid = 0;
649
650 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
651
652 /* The device might not support AER */
653 if (!pos)
654 return 1;
655
656 if (info->severity == AER_CORRECTABLE) {
657 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS,
658 &info->status);
659 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK,
660 &info->mask);
661 if (!(info->status & ~info->mask))
662 return 0;
663 } else if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE ||
664 info->severity == AER_NONFATAL) {
665
666 /* Link is still healthy for IO reads */
667 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS,
668 &info->status);
669 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK,
670 &info->mask);
671 if (!(info->status & ~info->mask))
672 return 0;
673
674 /* Get First Error Pointer */
675 pci_read_config_dword(dev, pos + PCI_ERR_CAP, &temp);
676 info->first_error = PCI_ERR_CAP_FEP(temp);
677
678 if (info->status & AER_LOG_TLP_MASKS) {
679 info->tlp_header_valid = 1;
680 pci_read_config_dword(dev,
681 pos + PCI_ERR_HEADER_LOG, &info->tlp.dw0);
682 pci_read_config_dword(dev,
683 pos + PCI_ERR_HEADER_LOG + 4, &info->tlp.dw1);
684 pci_read_config_dword(dev,
685 pos + PCI_ERR_HEADER_LOG + 8, &info->tlp.dw2);
686 pci_read_config_dword(dev,
687 pos + PCI_ERR_HEADER_LOG + 12, &info->tlp.dw3);
688 }
689 }
690
691 return 1;
692 }
693
aer_process_err_devices(struct pcie_device * p_device,struct aer_err_info * e_info)694 static inline void aer_process_err_devices(struct pcie_device *p_device,
695 struct aer_err_info *e_info)
696 {
697 int i;
698
699 /* Report all before handle them, not to lost records by reset etc. */
700 for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
701 if (get_device_error_info(e_info->dev[i], e_info))
702 aer_print_error(e_info->dev[i], e_info);
703 }
704 for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
705 if (get_device_error_info(e_info->dev[i], e_info))
706 handle_error_source(p_device, e_info->dev[i], e_info);
707 }
708 }
709
710 /**
711 * aer_isr_one_error - consume an error detected by root port
712 * @p_device: pointer to error root port service device
713 * @e_src: pointer to an error source
714 */
aer_isr_one_error(struct pcie_device * p_device,struct aer_err_source * e_src)715 static void aer_isr_one_error(struct pcie_device *p_device,
716 struct aer_err_source *e_src)
717 {
718 struct aer_err_info *e_info;
719
720 /* struct aer_err_info might be big, so we allocate it with slab */
721 e_info = kmalloc(sizeof(struct aer_err_info), GFP_KERNEL);
722 if (!e_info) {
723 dev_printk(KERN_DEBUG, &p_device->port->dev,
724 "Can't allocate mem when processing AER errors\n");
725 return;
726 }
727
728 /*
729 * There is a possibility that both correctable error and
730 * uncorrectable error being logged. Report correctable error first.
731 */
732 if (e_src->status & PCI_ERR_ROOT_COR_RCV) {
733 e_info->id = ERR_COR_ID(e_src->id);
734 e_info->severity = AER_CORRECTABLE;
735
736 if (e_src->status & PCI_ERR_ROOT_MULTI_COR_RCV)
737 e_info->multi_error_valid = 1;
738 else
739 e_info->multi_error_valid = 0;
740
741 aer_print_port_info(p_device->port, e_info);
742
743 if (find_source_device(p_device->port, e_info))
744 aer_process_err_devices(p_device, e_info);
745 }
746
747 if (e_src->status & PCI_ERR_ROOT_UNCOR_RCV) {
748 e_info->id = ERR_UNCOR_ID(e_src->id);
749
750 if (e_src->status & PCI_ERR_ROOT_FATAL_RCV)
751 e_info->severity = AER_FATAL;
752 else
753 e_info->severity = AER_NONFATAL;
754
755 if (e_src->status & PCI_ERR_ROOT_MULTI_UNCOR_RCV)
756 e_info->multi_error_valid = 1;
757 else
758 e_info->multi_error_valid = 0;
759
760 aer_print_port_info(p_device->port, e_info);
761
762 if (find_source_device(p_device->port, e_info))
763 aer_process_err_devices(p_device, e_info);
764 }
765
766 kfree(e_info);
767 }
768
769 /**
770 * get_e_source - retrieve an error source
771 * @rpc: pointer to the root port which holds an error
772 * @e_src: pointer to store retrieved error source
773 *
774 * Return 1 if an error source is retrieved, otherwise 0.
775 *
776 * Invoked by DPC handler to consume an error.
777 */
get_e_source(struct aer_rpc * rpc,struct aer_err_source * e_src)778 static int get_e_source(struct aer_rpc *rpc, struct aer_err_source *e_src)
779 {
780 unsigned long flags;
781
782 /* Lock access to Root error producer/consumer index */
783 spin_lock_irqsave(&rpc->e_lock, flags);
784 if (rpc->prod_idx == rpc->cons_idx) {
785 spin_unlock_irqrestore(&rpc->e_lock, flags);
786 return 0;
787 }
788
789 *e_src = rpc->e_sources[rpc->cons_idx];
790 rpc->cons_idx++;
791 if (rpc->cons_idx == AER_ERROR_SOURCES_MAX)
792 rpc->cons_idx = 0;
793 spin_unlock_irqrestore(&rpc->e_lock, flags);
794
795 return 1;
796 }
797
798 /**
799 * aer_isr - consume errors detected by root port
800 * @work: definition of this work item
801 *
802 * Invoked, as DPC, when root port records new detected error
803 */
aer_isr(struct work_struct * work)804 void aer_isr(struct work_struct *work)
805 {
806 struct aer_rpc *rpc = container_of(work, struct aer_rpc, dpc_handler);
807 struct pcie_device *p_device = rpc->rpd;
808 struct aer_err_source uninitialized_var(e_src);
809
810 mutex_lock(&rpc->rpc_mutex);
811 while (get_e_source(rpc, &e_src))
812 aer_isr_one_error(p_device, &e_src);
813 mutex_unlock(&rpc->rpc_mutex);
814 }
815
816 /**
817 * aer_init - provide AER initialization
818 * @dev: pointer to AER pcie device
819 *
820 * Invoked when AER service driver is loaded.
821 */
aer_init(struct pcie_device * dev)822 int aer_init(struct pcie_device *dev)
823 {
824 if (forceload) {
825 dev_printk(KERN_DEBUG, &dev->device,
826 "aerdrv forceload requested.\n");
827 pcie_aer_force_firmware_first(dev->port, 0);
828 }
829 return 0;
830 }
831