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
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 */
add_error_device(struct aer_err_info * e_info,struct pci_dev * dev)82 static 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++;
87 return 0;
88 }
89 return -ENOSPC;
90 }
91
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 */
is_error_source(struct pci_dev * dev,struct aer_err_info * e_info)97 static bool is_error_source(struct pci_dev *dev, struct aer_err_info *e_info)
98 {
99 int pos;
100 u32 status, mask;
101 u16 reg16;
102
103 /*
104 * When bus id is equal to 0, it might be a bad id
105 * reported by root port.
106 */
107 if (!nosourceid && (PCI_BUS_NUM(e_info->id) != 0)) {
108 /* Device ID match? */
109 if (e_info->id == ((dev->bus->number << 8) | dev->devfn))
110 return true;
111
112 /* Continue id comparing if there is no multiple error */
113 if (!e_info->multi_error_valid)
114 return false;
115 }
116
117 /*
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;
123 * We check AER status registers to find possible reporter.
124 */
125 if (atomic_read(&dev->enable_cnt) == 0)
126 return false;
127
128 /* Check if AER is enabled */
129 pcie_capability_read_word(dev, PCI_EXP_DEVCTL, ®16);
130 if (!(reg16 & PCI_EXP_AER_FLAGS))
131 return false;
132
133 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
134 if (!pos)
135 return false;
136
137 /* Check if error is recorded */
138 if (e_info->severity == AER_CORRECTABLE) {
139 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
140 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, &mask);
141 } else {
142 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
143 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, &mask);
144 }
145 if (status & ~mask)
146 return true;
147
148 return false;
149 }
150
find_device_iter(struct pci_dev * dev,void * data)151 static 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)) {
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 }
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;
168 }
169
170 /**
171 * find_source_device - search through device hierarchy for source device
172 * @parent: pointer to Root Port pci_dev data structure
173 * @e_info: including detailed error information such like id
174 *
175 * Return true if found.
176 *
177 * Invoked by DPC when error is detected at the Root Port.
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.
181 */
find_source_device(struct pci_dev * parent,struct aer_err_info * e_info)182 static bool find_source_device(struct pci_dev *parent,
183 struct aer_err_info *e_info)
184 {
185 struct pci_dev *dev = parent;
186 int result;
187
188 /* Must reset in this function */
189 e_info->error_dev_num = 0;
190
191 /* Is Root Port an agent that sends error message? */
192 result = find_device_iter(dev, e_info);
193 if (result)
194 return true;
195
196 pci_walk_bus(parent->subordinate, find_device_iter, e_info);
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;
205 }
206
report_error_detected(struct pci_dev * dev,void * data)207 static int report_error_detected(struct pci_dev *dev, void *data)
208 {
209 pci_ers_result_t vote;
210 const struct pci_error_handlers *err_handler;
211 struct aer_broadcast_data *result_data;
212 result_data = (struct aer_broadcast_data *) data;
213
214 device_lock(&dev->dev);
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 */
229 dev_printk(KERN_DEBUG, &dev->dev, "device has %s\n",
230 dev->driver ?
231 "no AER-aware driver" : "no driver");
232 }
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);
251 }
252
253 result_data->result = merge_result(result_data->result, vote);
254 device_unlock(&dev->dev);
255 return 0;
256 }
257
report_mmio_enabled(struct pci_dev * dev,void * data)258 static int report_mmio_enabled(struct pci_dev *dev, void *data)
259 {
260 pci_ers_result_t vote;
261 const struct pci_error_handlers *err_handler;
262 struct aer_broadcast_data *result_data;
263 result_data = (struct aer_broadcast_data *) data;
264
265 device_lock(&dev->dev);
266 if (!dev->driver ||
267 !dev->driver->err_handler ||
268 !dev->driver->err_handler->mmio_enabled)
269 goto out;
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);
274 out:
275 device_unlock(&dev->dev);
276 return 0;
277 }
278
report_slot_reset(struct pci_dev * dev,void * data)279 static int report_slot_reset(struct pci_dev *dev, void *data)
280 {
281 pci_ers_result_t vote;
282 const struct pci_error_handlers *err_handler;
283 struct aer_broadcast_data *result_data;
284 result_data = (struct aer_broadcast_data *) data;
285
286 device_lock(&dev->dev);
287 if (!dev->driver ||
288 !dev->driver->err_handler ||
289 !dev->driver->err_handler->slot_reset)
290 goto out;
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);
295 out:
296 device_unlock(&dev->dev);
297 return 0;
298 }
299
report_resume(struct pci_dev * dev,void * data)300 static int report_resume(struct pci_dev *dev, void *data)
301 {
302 const struct pci_error_handlers *err_handler;
303
304 device_lock(&dev->dev);
305 dev->error_state = pci_channel_io_normal;
306
307 if (!dev->driver ||
308 !dev->driver->err_handler ||
309 !dev->driver->err_handler->resume)
310 goto out;
311
312 err_handler = dev->driver->err_handler;
313 err_handler->resume(dev);
314 out:
315 device_unlock(&dev->dev);
316 return 0;
317 }
318
319 /**
320 * broadcast_error_message - handle message broadcast to downstream drivers
321 * @dev: pointer to from where in a hierarchy message is broadcasted down
322 * @state: error state
323 * @error_mesg: message to print
324 * @cb: callback to be broadcasted
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.
329 */
broadcast_error_message(struct pci_dev * dev,enum pci_channel_state state,char * error_mesg,int (* cb)(struct pci_dev *,void *))330 static pci_ers_result_t broadcast_error_message(struct pci_dev *dev,
331 enum pci_channel_state state,
332 char *error_mesg,
333 int (*cb)(struct pci_dev *, void *))
334 {
335 struct aer_broadcast_data result_data;
336
337 dev_printk(KERN_DEBUG, &dev->dev, "broadcast %s message\n", error_mesg);
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 }
358 } else {
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 if (state == pci_channel_io_normal)
364 /*
365 * the error is non fatal so the bus is ok, just invoke
366 * the callback for the function that logged the error.
367 */
368 cb(dev, &result_data);
369 else
370 pci_walk_bus(dev->bus, cb, &result_data);
371 }
372
373 return result_data.result;
374 }
375
376 /**
377 * default_reset_link - default reset function
378 * @dev: pointer to pci_dev data structure
379 *
380 * Invoked when performing link reset on a Downstream Port or a
381 * Root Port with no aer driver.
382 */
default_reset_link(struct pci_dev * dev)383 static pci_ers_result_t default_reset_link(struct pci_dev *dev)
384 {
385 pci_reset_bridge_secondary_bus(dev);
386 dev_printk(KERN_DEBUG, &dev->dev, "downstream link has been reset\n");
387 return PCI_ERS_RESULT_RECOVERED;
388 }
389
find_aer_service_iter(struct device * device,void * data)390 static int find_aer_service_iter(struct device *device, void *data)
391 {
392 struct pcie_port_service_driver *service_driver, **drv;
393
394 drv = (struct pcie_port_service_driver **) data;
395
396 if (device->bus == &pcie_port_bus_type && device->driver) {
397 service_driver = to_service_driver(device->driver);
398 if (service_driver->service == PCIE_PORT_SERVICE_AER) {
399 *drv = service_driver;
400 return 1;
401 }
402 }
403
404 return 0;
405 }
406
find_aer_service(struct pci_dev * dev)407 static struct pcie_port_service_driver *find_aer_service(struct pci_dev *dev)
408 {
409 struct pcie_port_service_driver *drv = NULL;
410
411 device_for_each_child(&dev->dev, &drv, find_aer_service_iter);
412
413 return drv;
414 }
415
reset_link(struct pci_dev * dev)416 static pci_ers_result_t reset_link(struct pci_dev *dev)
417 {
418 struct pci_dev *udev;
419 pci_ers_result_t status;
420 struct pcie_port_service_driver *driver;
421
422 if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE) {
423 /* Reset this port for all subordinates */
424 udev = dev;
425 } else {
426 /* Reset the upstream component (likely downstream port) */
427 udev = dev->bus->self;
428 }
429
430 /* Use the aer driver of the component firstly */
431 driver = find_aer_service(udev);
432
433 if (driver && driver->reset_link) {
434 status = driver->reset_link(udev);
435 } else if (pci_pcie_type(udev) == PCI_EXP_TYPE_DOWNSTREAM ||
436 pci_pcie_type(udev) == PCI_EXP_TYPE_ROOT_PORT) {
437 status = default_reset_link(udev);
438 } else {
439 dev_printk(KERN_DEBUG, &dev->dev,
440 "no link-reset support at upstream device %s\n",
441 pci_name(udev));
442 return PCI_ERS_RESULT_DISCONNECT;
443 }
444
445 if (status != PCI_ERS_RESULT_RECOVERED) {
446 dev_printk(KERN_DEBUG, &dev->dev,
447 "link reset at upstream device %s failed\n",
448 pci_name(udev));
449 return PCI_ERS_RESULT_DISCONNECT;
450 }
451
452 return status;
453 }
454
455 /**
456 * do_recovery - handle nonfatal/fatal error recovery process
457 * @dev: pointer to a pci_dev data structure of agent detecting an error
458 * @severity: error severity type
459 *
460 * Invoked when an error is nonfatal/fatal. Once being invoked, broadcast
461 * error detected message to all downstream drivers within a hierarchy in
462 * question and return the returned code.
463 */
do_recovery(struct pci_dev * dev,int severity)464 static void do_recovery(struct pci_dev *dev, int severity)
465 {
466 pci_ers_result_t status, result = PCI_ERS_RESULT_RECOVERED;
467 enum pci_channel_state state;
468
469 if (severity == AER_FATAL)
470 state = pci_channel_io_frozen;
471 else
472 state = pci_channel_io_normal;
473
474 status = broadcast_error_message(dev,
475 state,
476 "error_detected",
477 report_error_detected);
478
479 if (severity == AER_FATAL) {
480 result = reset_link(dev);
481 if (result != PCI_ERS_RESULT_RECOVERED)
482 goto failed;
483 }
484
485 if (status == PCI_ERS_RESULT_CAN_RECOVER)
486 status = broadcast_error_message(dev,
487 state,
488 "mmio_enabled",
489 report_mmio_enabled);
490
491 if (status == PCI_ERS_RESULT_NEED_RESET) {
492 /*
493 * TODO: Should call platform-specific
494 * functions to reset slot before calling
495 * drivers' slot_reset callbacks?
496 */
497 status = broadcast_error_message(dev,
498 state,
499 "slot_reset",
500 report_slot_reset);
501 }
502
503 if (status != PCI_ERS_RESULT_RECOVERED)
504 goto failed;
505
506 broadcast_error_message(dev,
507 state,
508 "resume",
509 report_resume);
510
511 dev_info(&dev->dev, "AER: Device recovery successful\n");
512 return;
513
514 failed:
515 /* TODO: Should kernel panic here? */
516 dev_info(&dev->dev, "AER: Device recovery failed\n");
517 }
518
519 /**
520 * handle_error_source - handle logging error into an event log
521 * @aerdev: pointer to pcie_device data structure of the root port
522 * @dev: pointer to pci_dev data structure of error source device
523 * @info: comprehensive error information
524 *
525 * Invoked when an error being detected by Root Port.
526 */
handle_error_source(struct pcie_device * aerdev,struct pci_dev * dev,struct aer_err_info * info)527 static void handle_error_source(struct pcie_device *aerdev,
528 struct pci_dev *dev,
529 struct aer_err_info *info)
530 {
531 int pos;
532
533 if (info->severity == AER_CORRECTABLE) {
534 /*
535 * Correctable error does not need software intervention.
536 * No need to go through error recovery process.
537 */
538 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
539 if (pos)
540 pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS,
541 info->status);
542 } else
543 do_recovery(dev, info->severity);
544 }
545
546 #ifdef CONFIG_ACPI_APEI_PCIEAER
547 static void aer_recover_work_func(struct work_struct *work);
548
549 #define AER_RECOVER_RING_ORDER 4
550 #define AER_RECOVER_RING_SIZE (1 << AER_RECOVER_RING_ORDER)
551
552 struct aer_recover_entry {
553 u8 bus;
554 u8 devfn;
555 u16 domain;
556 int severity;
557 struct aer_capability_regs *regs;
558 };
559
560 static DEFINE_KFIFO(aer_recover_ring, struct aer_recover_entry,
561 AER_RECOVER_RING_SIZE);
562 /*
563 * Mutual exclusion for writers of aer_recover_ring, reader side don't
564 * need lock, because there is only one reader and lock is not needed
565 * between reader and writer.
566 */
567 static DEFINE_SPINLOCK(aer_recover_ring_lock);
568 static DECLARE_WORK(aer_recover_work, aer_recover_work_func);
569
aer_recover_queue(int domain,unsigned int bus,unsigned int devfn,int severity,struct aer_capability_regs * aer_regs)570 void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn,
571 int severity, struct aer_capability_regs *aer_regs)
572 {
573 unsigned long flags;
574 struct aer_recover_entry entry = {
575 .bus = bus,
576 .devfn = devfn,
577 .domain = domain,
578 .severity = severity,
579 .regs = aer_regs,
580 };
581
582 spin_lock_irqsave(&aer_recover_ring_lock, flags);
583 if (kfifo_put(&aer_recover_ring, entry))
584 schedule_work(&aer_recover_work);
585 else
586 pr_err("AER recover: Buffer overflow when recovering AER for %04x:%02x:%02x:%x\n",
587 domain, bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
588 spin_unlock_irqrestore(&aer_recover_ring_lock, flags);
589 }
590 EXPORT_SYMBOL_GPL(aer_recover_queue);
591
aer_recover_work_func(struct work_struct * work)592 static void aer_recover_work_func(struct work_struct *work)
593 {
594 struct aer_recover_entry entry;
595 struct pci_dev *pdev;
596
597 while (kfifo_get(&aer_recover_ring, &entry)) {
598 pdev = pci_get_domain_bus_and_slot(entry.domain, entry.bus,
599 entry.devfn);
600 if (!pdev) {
601 pr_err("AER recover: Can not find pci_dev for %04x:%02x:%02x:%x\n",
602 entry.domain, entry.bus,
603 PCI_SLOT(entry.devfn), PCI_FUNC(entry.devfn));
604 continue;
605 }
606 cper_print_aer(pdev, entry.severity, entry.regs);
607 do_recovery(pdev, entry.severity);
608 pci_dev_put(pdev);
609 }
610 }
611 #endif
612
613 /**
614 * get_device_error_info - read error status from dev and store it to info
615 * @dev: pointer to the device expected to have a error record
616 * @info: pointer to structure to store the error record
617 *
618 * Return 1 on success, 0 on error.
619 *
620 * Note that @info is reused among all error devices. Clear fields properly.
621 */
get_device_error_info(struct pci_dev * dev,struct aer_err_info * info)622 static int get_device_error_info(struct pci_dev *dev, struct aer_err_info *info)
623 {
624 int pos, temp;
625
626 /* Must reset in this function */
627 info->status = 0;
628 info->tlp_header_valid = 0;
629
630 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
631
632 /* The device might not support AER */
633 if (!pos)
634 return 1;
635
636 if (info->severity == AER_CORRECTABLE) {
637 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS,
638 &info->status);
639 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK,
640 &info->mask);
641 if (!(info->status & ~info->mask))
642 return 0;
643 } else if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE ||
644 info->severity == AER_NONFATAL) {
645
646 /* Link is still healthy for IO reads */
647 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS,
648 &info->status);
649 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK,
650 &info->mask);
651 if (!(info->status & ~info->mask))
652 return 0;
653
654 /* Get First Error Pointer */
655 pci_read_config_dword(dev, pos + PCI_ERR_CAP, &temp);
656 info->first_error = PCI_ERR_CAP_FEP(temp);
657
658 if (info->status & AER_LOG_TLP_MASKS) {
659 info->tlp_header_valid = 1;
660 pci_read_config_dword(dev,
661 pos + PCI_ERR_HEADER_LOG, &info->tlp.dw0);
662 pci_read_config_dword(dev,
663 pos + PCI_ERR_HEADER_LOG + 4, &info->tlp.dw1);
664 pci_read_config_dword(dev,
665 pos + PCI_ERR_HEADER_LOG + 8, &info->tlp.dw2);
666 pci_read_config_dword(dev,
667 pos + PCI_ERR_HEADER_LOG + 12, &info->tlp.dw3);
668 }
669 }
670
671 return 1;
672 }
673
aer_process_err_devices(struct pcie_device * p_device,struct aer_err_info * e_info)674 static inline void aer_process_err_devices(struct pcie_device *p_device,
675 struct aer_err_info *e_info)
676 {
677 int i;
678
679 /* Report all before handle them, not to lost records by reset etc. */
680 for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
681 if (get_device_error_info(e_info->dev[i], e_info))
682 aer_print_error(e_info->dev[i], e_info);
683 }
684 for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
685 if (get_device_error_info(e_info->dev[i], e_info))
686 handle_error_source(p_device, e_info->dev[i], e_info);
687 }
688 }
689
690 /**
691 * aer_isr_one_error - consume an error detected by root port
692 * @p_device: pointer to error root port service device
693 * @e_src: pointer to an error source
694 */
aer_isr_one_error(struct pcie_device * p_device,struct aer_err_source * e_src)695 static void aer_isr_one_error(struct pcie_device *p_device,
696 struct aer_err_source *e_src)
697 {
698 struct aer_err_info *e_info;
699
700 /* struct aer_err_info might be big, so we allocate it with slab */
701 e_info = kmalloc(sizeof(struct aer_err_info), GFP_KERNEL);
702 if (!e_info) {
703 dev_printk(KERN_DEBUG, &p_device->port->dev,
704 "Can't allocate mem when processing AER errors\n");
705 return;
706 }
707
708 /*
709 * There is a possibility that both correctable error and
710 * uncorrectable error being logged. Report correctable error first.
711 */
712 if (e_src->status & PCI_ERR_ROOT_COR_RCV) {
713 e_info->id = ERR_COR_ID(e_src->id);
714 e_info->severity = AER_CORRECTABLE;
715
716 if (e_src->status & PCI_ERR_ROOT_MULTI_COR_RCV)
717 e_info->multi_error_valid = 1;
718 else
719 e_info->multi_error_valid = 0;
720
721 aer_print_port_info(p_device->port, e_info);
722
723 if (find_source_device(p_device->port, e_info))
724 aer_process_err_devices(p_device, e_info);
725 }
726
727 if (e_src->status & PCI_ERR_ROOT_UNCOR_RCV) {
728 e_info->id = ERR_UNCOR_ID(e_src->id);
729
730 if (e_src->status & PCI_ERR_ROOT_FATAL_RCV)
731 e_info->severity = AER_FATAL;
732 else
733 e_info->severity = AER_NONFATAL;
734
735 if (e_src->status & PCI_ERR_ROOT_MULTI_UNCOR_RCV)
736 e_info->multi_error_valid = 1;
737 else
738 e_info->multi_error_valid = 0;
739
740 aer_print_port_info(p_device->port, e_info);
741
742 if (find_source_device(p_device->port, e_info))
743 aer_process_err_devices(p_device, e_info);
744 }
745
746 kfree(e_info);
747 }
748
749 /**
750 * get_e_source - retrieve an error source
751 * @rpc: pointer to the root port which holds an error
752 * @e_src: pointer to store retrieved error source
753 *
754 * Return 1 if an error source is retrieved, otherwise 0.
755 *
756 * Invoked by DPC handler to consume an error.
757 */
get_e_source(struct aer_rpc * rpc,struct aer_err_source * e_src)758 static int get_e_source(struct aer_rpc *rpc, struct aer_err_source *e_src)
759 {
760 unsigned long flags;
761
762 /* Lock access to Root error producer/consumer index */
763 spin_lock_irqsave(&rpc->e_lock, flags);
764 if (rpc->prod_idx == rpc->cons_idx) {
765 spin_unlock_irqrestore(&rpc->e_lock, flags);
766 return 0;
767 }
768
769 *e_src = rpc->e_sources[rpc->cons_idx];
770 rpc->cons_idx++;
771 if (rpc->cons_idx == AER_ERROR_SOURCES_MAX)
772 rpc->cons_idx = 0;
773 spin_unlock_irqrestore(&rpc->e_lock, flags);
774
775 return 1;
776 }
777
778 /**
779 * aer_isr - consume errors detected by root port
780 * @work: definition of this work item
781 *
782 * Invoked, as DPC, when root port records new detected error
783 */
aer_isr(struct work_struct * work)784 void aer_isr(struct work_struct *work)
785 {
786 struct aer_rpc *rpc = container_of(work, struct aer_rpc, dpc_handler);
787 struct pcie_device *p_device = rpc->rpd;
788 struct aer_err_source uninitialized_var(e_src);
789
790 mutex_lock(&rpc->rpc_mutex);
791 while (get_e_source(rpc, &e_src))
792 aer_isr_one_error(p_device, &e_src);
793 mutex_unlock(&rpc->rpc_mutex);
794 }
795
796 /**
797 * aer_init - provide AER initialization
798 * @dev: pointer to AER pcie device
799 *
800 * Invoked when AER service driver is loaded.
801 */
aer_init(struct pcie_device * dev)802 int aer_init(struct pcie_device *dev)
803 {
804 if (forceload) {
805 dev_printk(KERN_DEBUG, &dev->device,
806 "aerdrv forceload requested.\n");
807 pcie_aer_force_firmware_first(dev->port, 0);
808 }
809 return 0;
810 }
811