Commit | Line | Data |
---|---|---|
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 |
30 | static bool forceload; |
31 | static bool nosourceid; | |
6c2b374d | 32 | module_param(forceload, bool, 0); |
28eb27cf | 33 | module_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 |
38 | int 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 | 48 | EXPORT_SYMBOL_GPL(pci_enable_pcie_error_reporting); |
6c2b374d ZY |
49 | |
50 | int 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 | 58 | EXPORT_SYMBOL_GPL(pci_disable_pcie_error_reporting); |
6c2b374d ZY |
59 | |
60 | int 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 | 75 | EXPORT_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 |
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++; | |
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 | */ | |
97 | static 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, ®16); |
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 |
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)) { | |
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 | 182 | static 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 | 207 | static 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 | 258 | static 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 |
274 | out: |
275 | device_unlock(&dev->dev); | |
70298c6e | 276 | return 0; |
6c2b374d ZY |
277 | } |
278 | ||
70298c6e | 279 | static 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 |
295 | out: |
296 | device_unlock(&dev->dev); | |
70298c6e | 297 | return 0; |
6c2b374d ZY |
298 | } |
299 | ||
70298c6e | 300 | static 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 |
314 | out: |
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 |
330 | static 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 | 376 | static 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 |
383 | static 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 | 400 | static 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 | 409 | static 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); | |
081d0fe0 BD |
428 | } else if (pci_pcie_type(udev) == PCI_EXP_TYPE_DOWNSTREAM || |
429 | pci_pcie_type(udev) == PCI_EXP_TYPE_ROOT_PORT) { | |
430 | status = default_reset_link(udev); | |
89713422 HS |
431 | } else { |
432 | dev_printk(KERN_DEBUG, &dev->dev, | |
433 | "no link-reset support at upstream device %s\n", | |
434 | pci_name(udev)); | |
435 | return PCI_ERS_RESULT_DISCONNECT; | |
6c2b374d ZY |
436 | } |
437 | ||
6c2b374d | 438 | if (status != PCI_ERS_RESULT_RECOVERED) { |
4f7ccf6a HS |
439 | dev_printk(KERN_DEBUG, &dev->dev, |
440 | "link reset at upstream device %s failed\n", | |
441 | pci_name(udev)); | |
6c2b374d ZY |
442 | return PCI_ERS_RESULT_DISCONNECT; |
443 | } | |
444 | ||
445 | return status; | |
446 | } | |
447 | ||
448 | /** | |
449 | * do_recovery - handle nonfatal/fatal error recovery process | |
6c2b374d ZY |
450 | * @dev: pointer to a pci_dev data structure of agent detecting an error |
451 | * @severity: error severity type | |
452 | * | |
453 | * Invoked when an error is nonfatal/fatal. Once being invoked, broadcast | |
454 | * error detected message to all downstream drivers within a hierarchy in | |
455 | * question and return the returned code. | |
d885c6b7 | 456 | */ |
0918472c | 457 | static void do_recovery(struct pci_dev *dev, int severity) |
6c2b374d ZY |
458 | { |
459 | pci_ers_result_t status, result = PCI_ERS_RESULT_RECOVERED; | |
460 | enum pci_channel_state state; | |
461 | ||
462 | if (severity == AER_FATAL) | |
463 | state = pci_channel_io_frozen; | |
464 | else | |
465 | state = pci_channel_io_normal; | |
466 | ||
467 | status = broadcast_error_message(dev, | |
468 | state, | |
469 | "error_detected", | |
470 | report_error_detected); | |
471 | ||
472 | if (severity == AER_FATAL) { | |
0918472c | 473 | result = reset_link(dev); |
17e21854 HS |
474 | if (result != PCI_ERS_RESULT_RECOVERED) |
475 | goto failed; | |
6c2b374d ZY |
476 | } |
477 | ||
478 | if (status == PCI_ERS_RESULT_CAN_RECOVER) | |
479 | status = broadcast_error_message(dev, | |
480 | state, | |
481 | "mmio_enabled", | |
482 | report_mmio_enabled); | |
483 | ||
484 | if (status == PCI_ERS_RESULT_NEED_RESET) { | |
485 | /* | |
486 | * TODO: Should call platform-specific | |
487 | * functions to reset slot before calling | |
488 | * drivers' slot_reset callbacks? | |
489 | */ | |
490 | status = broadcast_error_message(dev, | |
491 | state, | |
492 | "slot_reset", | |
493 | report_slot_reset); | |
494 | } | |
495 | ||
17e21854 HS |
496 | if (status != PCI_ERS_RESULT_RECOVERED) |
497 | goto failed; | |
498 | ||
499 | broadcast_error_message(dev, | |
6c2b374d ZY |
500 | state, |
501 | "resume", | |
502 | report_resume); | |
503 | ||
be5ac3d3 | 504 | dev_info(&dev->dev, "AER: Device recovery successful\n"); |
17e21854 HS |
505 | return; |
506 | ||
507 | failed: | |
508 | /* TODO: Should kernel panic here? */ | |
be5ac3d3 | 509 | dev_info(&dev->dev, "AER: Device recovery failed\n"); |
6c2b374d ZY |
510 | } |
511 | ||
512 | /** | |
513 | * handle_error_source - handle logging error into an event log | |
514 | * @aerdev: pointer to pcie_device data structure of the root port | |
515 | * @dev: pointer to pci_dev data structure of error source device | |
516 | * @info: comprehensive error information | |
517 | * | |
518 | * Invoked when an error being detected by Root Port. | |
d885c6b7 | 519 | */ |
c9a91883 | 520 | static void handle_error_source(struct pcie_device *aerdev, |
6c2b374d | 521 | struct pci_dev *dev, |
28eb27cf | 522 | struct aer_err_info *info) |
6c2b374d | 523 | { |
6c2b374d ZY |
524 | int pos; |
525 | ||
28eb27cf | 526 | if (info->severity == AER_CORRECTABLE) { |
6c2b374d | 527 | /* |
f7625980 | 528 | * Correctable error does not need software intervention. |
6c2b374d ZY |
529 | * No need to go through error recovery process. |
530 | */ | |
0927678f | 531 | pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR); |
6c2b374d ZY |
532 | if (pos) |
533 | pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS, | |
28eb27cf | 534 | info->status); |
17e21854 | 535 | } else |
0918472c | 536 | do_recovery(dev, info->severity); |
6c2b374d ZY |
537 | } |
538 | ||
0918472c HY |
539 | #ifdef CONFIG_ACPI_APEI_PCIEAER |
540 | static void aer_recover_work_func(struct work_struct *work); | |
541 | ||
542 | #define AER_RECOVER_RING_ORDER 4 | |
543 | #define AER_RECOVER_RING_SIZE (1 << AER_RECOVER_RING_ORDER) | |
544 | ||
3c78bc61 | 545 | struct aer_recover_entry { |
0918472c HY |
546 | u8 bus; |
547 | u8 devfn; | |
548 | u16 domain; | |
549 | int severity; | |
37448adf | 550 | struct aer_capability_regs *regs; |
0918472c HY |
551 | }; |
552 | ||
553 | static DEFINE_KFIFO(aer_recover_ring, struct aer_recover_entry, | |
554 | AER_RECOVER_RING_SIZE); | |
555 | /* | |
556 | * Mutual exclusion for writers of aer_recover_ring, reader side don't | |
557 | * need lock, because there is only one reader and lock is not needed | |
558 | * between reader and writer. | |
559 | */ | |
560 | static DEFINE_SPINLOCK(aer_recover_ring_lock); | |
561 | static DECLARE_WORK(aer_recover_work, aer_recover_work_func); | |
562 | ||
563 | void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn, | |
37448adf | 564 | int severity, struct aer_capability_regs *aer_regs) |
0918472c HY |
565 | { |
566 | unsigned long flags; | |
567 | struct aer_recover_entry entry = { | |
568 | .bus = bus, | |
569 | .devfn = devfn, | |
570 | .domain = domain, | |
571 | .severity = severity, | |
37448adf | 572 | .regs = aer_regs, |
0918472c HY |
573 | }; |
574 | ||
575 | spin_lock_irqsave(&aer_recover_ring_lock, flags); | |
498d319b | 576 | if (kfifo_put(&aer_recover_ring, entry)) |
0918472c HY |
577 | schedule_work(&aer_recover_work); |
578 | else | |
579 | pr_err("AER recover: Buffer overflow when recovering AER for %04x:%02x:%02x:%x\n", | |
580 | domain, bus, PCI_SLOT(devfn), PCI_FUNC(devfn)); | |
581 | spin_unlock_irqrestore(&aer_recover_ring_lock, flags); | |
582 | } | |
583 | EXPORT_SYMBOL_GPL(aer_recover_queue); | |
584 | ||
585 | static void aer_recover_work_func(struct work_struct *work) | |
586 | { | |
587 | struct aer_recover_entry entry; | |
588 | struct pci_dev *pdev; | |
589 | ||
590 | while (kfifo_get(&aer_recover_ring, &entry)) { | |
591 | pdev = pci_get_domain_bus_and_slot(entry.domain, entry.bus, | |
592 | entry.devfn); | |
593 | if (!pdev) { | |
594 | pr_err("AER recover: Can not find pci_dev for %04x:%02x:%02x:%x\n", | |
595 | entry.domain, entry.bus, | |
596 | PCI_SLOT(entry.devfn), PCI_FUNC(entry.devfn)); | |
597 | continue; | |
598 | } | |
37448adf | 599 | cper_print_aer(pdev, entry.severity, entry.regs); |
0918472c | 600 | do_recovery(pdev, entry.severity); |
a82b6af3 | 601 | pci_dev_put(pdev); |
0918472c HY |
602 | } |
603 | } | |
604 | #endif | |
605 | ||
b1c089b7 HS |
606 | /** |
607 | * get_device_error_info - read error status from dev and store it to info | |
608 | * @dev: pointer to the device expected to have a error record | |
609 | * @info: pointer to structure to store the error record | |
610 | * | |
611 | * Return 1 on success, 0 on error. | |
7c4ec94f HS |
612 | * |
613 | * Note that @info is reused among all error devices. Clear fields properly. | |
b1c089b7 | 614 | */ |
6c2b374d ZY |
615 | static int get_device_error_info(struct pci_dev *dev, struct aer_err_info *info) |
616 | { | |
e7a0d92b | 617 | int pos, temp; |
6c2b374d | 618 | |
7c4ec94f | 619 | /* Must reset in this function */ |
1b4ffcf8 | 620 | info->status = 0; |
273024de | 621 | info->tlp_header_valid = 0; |
1b4ffcf8 | 622 | |
0927678f | 623 | pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR); |
6c2b374d ZY |
624 | |
625 | /* The device might not support AER */ | |
626 | if (!pos) | |
b1c089b7 | 627 | return 1; |
6c2b374d ZY |
628 | |
629 | if (info->severity == AER_CORRECTABLE) { | |
630 | pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, | |
631 | &info->status); | |
0d90c3ac HS |
632 | pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, |
633 | &info->mask); | |
634 | if (!(info->status & ~info->mask)) | |
b1c089b7 | 635 | return 0; |
6c2b374d ZY |
636 | } else if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE || |
637 | info->severity == AER_NONFATAL) { | |
638 | ||
639 | /* Link is still healthy for IO reads */ | |
640 | pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, | |
641 | &info->status); | |
0d90c3ac HS |
642 | pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, |
643 | &info->mask); | |
644 | if (!(info->status & ~info->mask)) | |
b1c089b7 | 645 | return 0; |
6c2b374d | 646 | |
e7a0d92b HS |
647 | /* Get First Error Pointer */ |
648 | pci_read_config_dword(dev, pos + PCI_ERR_CAP, &temp); | |
273024de | 649 | info->first_error = PCI_ERR_CAP_FEP(temp); |
e7a0d92b | 650 | |
6c2b374d | 651 | if (info->status & AER_LOG_TLP_MASKS) { |
273024de | 652 | info->tlp_header_valid = 1; |
6c2b374d ZY |
653 | pci_read_config_dword(dev, |
654 | pos + PCI_ERR_HEADER_LOG, &info->tlp.dw0); | |
655 | pci_read_config_dword(dev, | |
656 | pos + PCI_ERR_HEADER_LOG + 4, &info->tlp.dw1); | |
657 | pci_read_config_dword(dev, | |
658 | pos + PCI_ERR_HEADER_LOG + 8, &info->tlp.dw2); | |
659 | pci_read_config_dword(dev, | |
660 | pos + PCI_ERR_HEADER_LOG + 12, &info->tlp.dw3); | |
661 | } | |
662 | } | |
663 | ||
b1c089b7 | 664 | return 1; |
6c2b374d ZY |
665 | } |
666 | ||
3d5505c5 ZY |
667 | static inline void aer_process_err_devices(struct pcie_device *p_device, |
668 | struct aer_err_info *e_info) | |
669 | { | |
670 | int i; | |
671 | ||
b1c089b7 | 672 | /* Report all before handle them, not to lost records by reset etc. */ |
3d5505c5 | 673 | for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) { |
b1c089b7 | 674 | if (get_device_error_info(e_info->dev[i], e_info)) |
3d5505c5 | 675 | aer_print_error(e_info->dev[i], e_info); |
b1c089b7 HS |
676 | } |
677 | for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) { | |
678 | if (get_device_error_info(e_info->dev[i], e_info)) | |
679 | handle_error_source(p_device, e_info->dev[i], e_info); | |
3d5505c5 ZY |
680 | } |
681 | } | |
682 | ||
6c2b374d ZY |
683 | /** |
684 | * aer_isr_one_error - consume an error detected by root port | |
685 | * @p_device: pointer to error root port service device | |
686 | * @e_src: pointer to an error source | |
d885c6b7 | 687 | */ |
6c2b374d ZY |
688 | static void aer_isr_one_error(struct pcie_device *p_device, |
689 | struct aer_err_source *e_src) | |
690 | { | |
28eb27cf | 691 | struct aer_err_info *e_info; |
28eb27cf ZY |
692 | |
693 | /* struct aer_err_info might be big, so we allocate it with slab */ | |
694 | e_info = kmalloc(sizeof(struct aer_err_info), GFP_KERNEL); | |
7c4ec94f | 695 | if (!e_info) { |
28eb27cf ZY |
696 | dev_printk(KERN_DEBUG, &p_device->port->dev, |
697 | "Can't allocate mem when processing AER errors\n"); | |
698 | return; | |
699 | } | |
6c2b374d ZY |
700 | |
701 | /* | |
702 | * There is a possibility that both correctable error and | |
703 | * uncorrectable error being logged. Report correctable error first. | |
704 | */ | |
7c4ec94f HS |
705 | if (e_src->status & PCI_ERR_ROOT_COR_RCV) { |
706 | e_info->id = ERR_COR_ID(e_src->id); | |
707 | e_info->severity = AER_CORRECTABLE; | |
708 | ||
709 | if (e_src->status & PCI_ERR_ROOT_MULTI_COR_RCV) | |
710 | e_info->multi_error_valid = 1; | |
711 | else | |
712 | e_info->multi_error_valid = 0; | |
713 | ||
714 | aer_print_port_info(p_device->port, e_info); | |
715 | ||
716 | if (find_source_device(p_device->port, e_info)) | |
717 | aer_process_err_devices(p_device, e_info); | |
718 | } | |
719 | ||
720 | if (e_src->status & PCI_ERR_ROOT_UNCOR_RCV) { | |
721 | e_info->id = ERR_UNCOR_ID(e_src->id); | |
722 | ||
723 | if (e_src->status & PCI_ERR_ROOT_FATAL_RCV) | |
724 | e_info->severity = AER_FATAL; | |
725 | else | |
726 | e_info->severity = AER_NONFATAL; | |
727 | ||
728 | if (e_src->status & PCI_ERR_ROOT_MULTI_UNCOR_RCV) | |
273024de | 729 | e_info->multi_error_valid = 1; |
7c4ec94f HS |
730 | else |
731 | e_info->multi_error_valid = 0; | |
28eb27cf | 732 | |
79e4b89b HS |
733 | aer_print_port_info(p_device->port, e_info); |
734 | ||
98ca3964 HS |
735 | if (find_source_device(p_device->port, e_info)) |
736 | aer_process_err_devices(p_device, e_info); | |
6c2b374d | 737 | } |
28eb27cf ZY |
738 | |
739 | kfree(e_info); | |
6c2b374d ZY |
740 | } |
741 | ||
88da13bf HS |
742 | /** |
743 | * get_e_source - retrieve an error source | |
744 | * @rpc: pointer to the root port which holds an error | |
745 | * @e_src: pointer to store retrieved error source | |
746 | * | |
747 | * Return 1 if an error source is retrieved, otherwise 0. | |
748 | * | |
749 | * Invoked by DPC handler to consume an error. | |
750 | */ | |
751 | static int get_e_source(struct aer_rpc *rpc, struct aer_err_source *e_src) | |
752 | { | |
753 | unsigned long flags; | |
88da13bf HS |
754 | |
755 | /* Lock access to Root error producer/consumer index */ | |
756 | spin_lock_irqsave(&rpc->e_lock, flags); | |
f6735590 LT |
757 | if (rpc->prod_idx == rpc->cons_idx) { |
758 | spin_unlock_irqrestore(&rpc->e_lock, flags); | |
759 | return 0; | |
88da13bf | 760 | } |
f6735590 LT |
761 | |
762 | *e_src = rpc->e_sources[rpc->cons_idx]; | |
763 | rpc->cons_idx++; | |
764 | if (rpc->cons_idx == AER_ERROR_SOURCES_MAX) | |
765 | rpc->cons_idx = 0; | |
88da13bf HS |
766 | spin_unlock_irqrestore(&rpc->e_lock, flags); |
767 | ||
f6735590 | 768 | return 1; |
88da13bf HS |
769 | } |
770 | ||
6c2b374d ZY |
771 | /** |
772 | * aer_isr - consume errors detected by root port | |
65f27f38 | 773 | * @work: definition of this work item |
6c2b374d ZY |
774 | * |
775 | * Invoked, as DPC, when root port records new detected error | |
d885c6b7 | 776 | */ |
65f27f38 | 777 | void aer_isr(struct work_struct *work) |
6c2b374d | 778 | { |
65f27f38 DH |
779 | struct aer_rpc *rpc = container_of(work, struct aer_rpc, dpc_handler); |
780 | struct pcie_device *p_device = rpc->rpd; | |
50c1126e | 781 | struct aer_err_source uninitialized_var(e_src); |
6c2b374d ZY |
782 | |
783 | mutex_lock(&rpc->rpc_mutex); | |
88da13bf HS |
784 | while (get_e_source(rpc, &e_src)) |
785 | aer_isr_one_error(p_device, &e_src); | |
6c2b374d ZY |
786 | mutex_unlock(&rpc->rpc_mutex); |
787 | ||
788 | wake_up(&rpc->wait_release); | |
789 | } | |
790 | ||
6c2b374d ZY |
791 | /** |
792 | * aer_init - provide AER initialization | |
793 | * @dev: pointer to AER pcie device | |
794 | * | |
795 | * Invoked when AER service driver is loaded. | |
d885c6b7 | 796 | */ |
6c2b374d ZY |
797 | int aer_init(struct pcie_device *dev) |
798 | { | |
05843961 MD |
799 | if (forceload) { |
800 | dev_printk(KERN_DEBUG, &dev->device, | |
801 | "aerdrv forceload requested.\n"); | |
affb72c3 | 802 | pcie_aer_force_firmware_first(dev->port, 0); |
05843961 | 803 | } |
28eb5f27 | 804 | return 0; |
6c2b374d | 805 | } |