]>
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 | |
b07461a8 TI |
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 | ||
4a0c096e HS |
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 | */ | |
3d5505c5 ZY |
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++; | |
4a0c096e | 115 | return 0; |
c9a91883 | 116 | } |
4a0c096e | 117 | return -ENOSPC; |
3d5505c5 ZY |
118 | } |
119 | ||
c887275e HS |
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 | */ | |
125 | static bool is_error_source(struct pci_dev *dev, struct aer_err_info *e_info) | |
28eb27cf ZY |
126 | { |
127 | int pos; | |
c887275e | 128 | u32 status, mask; |
28eb27cf | 129 | u16 reg16; |
28eb27cf ZY |
130 | |
131 | /* | |
132 | * When bus id is equal to 0, it might be a bad id | |
133 | * reported by root port. | |
134 | */ | |
fff0ee36 | 135 | if (!nosourceid && (PCI_BUS_NUM(e_info->id) != 0)) { |
bd17d474 HS |
136 | /* Device ID match? */ |
137 | if (e_info->id == ((dev->bus->number << 8) | dev->devfn)) | |
c887275e | 138 | return true; |
3d5505c5 | 139 | |
c887275e | 140 | /* Continue id comparing if there is no multiple error */ |
273024de | 141 | if (!e_info->multi_error_valid) |
c887275e | 142 | return false; |
28eb27cf ZY |
143 | } |
144 | ||
145 | /* | |
3d5505c5 ZY |
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; | |
c887275e | 151 | * We check AER status registers to find possible reporter. |
28eb27cf ZY |
152 | */ |
153 | if (atomic_read(&dev->enable_cnt) == 0) | |
c887275e | 154 | return false; |
c887275e | 155 | |
28eb27cf | 156 | /* Check if AER is enabled */ |
43bd4ee8 JL |
157 | pcie_capability_read_word(dev, PCI_EXP_DEVCTL, ®16); |
158 | if (!(reg16 & PCI_EXP_AER_FLAGS)) | |
c887275e | 159 | return false; |
43bd4ee8 | 160 | |
28eb27cf ZY |
161 | pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR); |
162 | if (!pos) | |
c887275e | 163 | return false; |
28eb27cf | 164 | |
c887275e | 165 | /* Check if error is recorded */ |
28eb27cf | 166 | if (e_info->severity == AER_CORRECTABLE) { |
0d90c3ac HS |
167 | pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status); |
168 | pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, &mask); | |
28eb27cf | 169 | } else { |
0d90c3ac HS |
170 | pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status); |
171 | pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, &mask); | |
6c2b374d | 172 | } |
c887275e HS |
173 | if (status & ~mask) |
174 | return true; | |
6c2b374d | 175 | |
c887275e HS |
176 | return false; |
177 | } | |
3d5505c5 | 178 | |
c887275e HS |
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)) { | |
4a0c096e HS |
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 | } | |
c887275e HS |
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; | |
6c2b374d ZY |
196 | } |
197 | ||
198 | /** | |
199 | * find_source_device - search through device hierarchy for source device | |
d885c6b7 | 200 | * @parent: pointer to Root Port pci_dev data structure |
98ca3964 | 201 | * @e_info: including detailed error information such like id |
6c2b374d | 202 | * |
98ca3964 HS |
203 | * Return true if found. |
204 | * | |
205 | * Invoked by DPC when error is detected at the Root Port. | |
7c4ec94f HS |
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. | |
d885c6b7 | 209 | */ |
98ca3964 | 210 | static bool find_source_device(struct pci_dev *parent, |
28eb27cf | 211 | struct aer_err_info *e_info) |
6c2b374d ZY |
212 | { |
213 | struct pci_dev *dev = parent; | |
28eb27cf | 214 | int result; |
6c2b374d | 215 | |
7c4ec94f HS |
216 | /* Must reset in this function */ |
217 | e_info->error_dev_num = 0; | |
218 | ||
6c2b374d | 219 | /* Is Root Port an agent that sends error message? */ |
28eb27cf ZY |
220 | result = find_device_iter(dev, e_info); |
221 | if (result) | |
98ca3964 | 222 | return true; |
6c2b374d | 223 | |
28eb27cf | 224 | pci_walk_bus(parent->subordinate, find_device_iter, e_info); |
98ca3964 HS |
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; | |
6c2b374d ZY |
233 | } |
234 | ||
70298c6e | 235 | static int report_error_detected(struct pci_dev *dev, void *data) |
6c2b374d ZY |
236 | { |
237 | pci_ers_result_t vote; | |
49453028 | 238 | const struct pci_error_handlers *err_handler; |
6c2b374d ZY |
239 | struct aer_broadcast_data *result_data; |
240 | result_data = (struct aer_broadcast_data *) data; | |
241 | ||
90b5c1d7 | 242 | device_lock(&dev->dev); |
6c2b374d ZY |
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 && | |
93de6901 | 249 | dev->hdr_type != PCI_HEADER_TYPE_BRIDGE) { |
6c2b374d ZY |
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 | */ | |
531f254e BH |
257 | dev_printk(KERN_DEBUG, &dev->dev, "device has %s\n", |
258 | dev->driver ? | |
259 | "no AER-aware driver" : "no driver"); | |
6c2b374d | 260 | } |
918b4053 VMP |
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 | ||
93de6901 | 272 | if (dev->hdr_type != PCI_HEADER_TYPE_BRIDGE) |
918b4053 VMP |
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); | |
6c2b374d ZY |
279 | } |
280 | ||
6c2b374d | 281 | result_data->result = merge_result(result_data->result, vote); |
90b5c1d7 | 282 | device_unlock(&dev->dev); |
70298c6e | 283 | return 0; |
6c2b374d ZY |
284 | } |
285 | ||
70298c6e | 286 | static int report_mmio_enabled(struct pci_dev *dev, void *data) |
6c2b374d ZY |
287 | { |
288 | pci_ers_result_t vote; | |
49453028 | 289 | const struct pci_error_handlers *err_handler; |
6c2b374d ZY |
290 | struct aer_broadcast_data *result_data; |
291 | result_data = (struct aer_broadcast_data *) data; | |
292 | ||
90b5c1d7 | 293 | device_lock(&dev->dev); |
6c2b374d ZY |
294 | if (!dev->driver || |
295 | !dev->driver->err_handler || | |
296 | !dev->driver->err_handler->mmio_enabled) | |
90b5c1d7 | 297 | goto out; |
6c2b374d ZY |
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); | |
90b5c1d7 HY |
302 | out: |
303 | device_unlock(&dev->dev); | |
70298c6e | 304 | return 0; |
6c2b374d ZY |
305 | } |
306 | ||
70298c6e | 307 | static int report_slot_reset(struct pci_dev *dev, void *data) |
6c2b374d ZY |
308 | { |
309 | pci_ers_result_t vote; | |
49453028 | 310 | const struct pci_error_handlers *err_handler; |
6c2b374d ZY |
311 | struct aer_broadcast_data *result_data; |
312 | result_data = (struct aer_broadcast_data *) data; | |
313 | ||
90b5c1d7 | 314 | device_lock(&dev->dev); |
6c2b374d ZY |
315 | if (!dev->driver || |
316 | !dev->driver->err_handler || | |
317 | !dev->driver->err_handler->slot_reset) | |
90b5c1d7 | 318 | goto out; |
6c2b374d ZY |
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); | |
90b5c1d7 HY |
323 | out: |
324 | device_unlock(&dev->dev); | |
70298c6e | 325 | return 0; |
6c2b374d ZY |
326 | } |
327 | ||
70298c6e | 328 | static int report_resume(struct pci_dev *dev, void *data) |
6c2b374d | 329 | { |
49453028 | 330 | const struct pci_error_handlers *err_handler; |
6c2b374d | 331 | |
90b5c1d7 | 332 | device_lock(&dev->dev); |
6c2b374d ZY |
333 | dev->error_state = pci_channel_io_normal; |
334 | ||
335 | if (!dev->driver || | |
336 | !dev->driver->err_handler || | |
b0b801dd | 337 | !dev->driver->err_handler->resume) |
90b5c1d7 | 338 | goto out; |
6c2b374d ZY |
339 | |
340 | err_handler = dev->driver->err_handler; | |
341 | err_handler->resume(dev); | |
90b5c1d7 HY |
342 | out: |
343 | device_unlock(&dev->dev); | |
70298c6e | 344 | return 0; |
6c2b374d ZY |
345 | } |
346 | ||
347 | /** | |
348 | * broadcast_error_message - handle message broadcast to downstream drivers | |
d885c6b7 | 349 | * @dev: pointer to from where in a hierarchy message is broadcasted down |
6c2b374d | 350 | * @state: error state |
d885c6b7 RD |
351 | * @error_mesg: message to print |
352 | * @cb: callback to be broadcasted | |
6c2b374d ZY |
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. | |
d885c6b7 | 357 | */ |
6c2b374d ZY |
358 | static pci_ers_result_t broadcast_error_message(struct pci_dev *dev, |
359 | enum pci_channel_state state, | |
360 | char *error_mesg, | |
70298c6e | 361 | int (*cb)(struct pci_dev *, void *)) |
6c2b374d ZY |
362 | { |
363 | struct aer_broadcast_data result_data; | |
364 | ||
531f254e | 365 | dev_printk(KERN_DEBUG, &dev->dev, "broadcast %s message\n", error_mesg); |
6c2b374d ZY |
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 | ||
93de6901 | 372 | if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) { |
6c2b374d ZY |
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 | } | |
c9a91883 | 386 | } else { |
6c2b374d ZY |
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 | ||
89713422 | 397 | /** |
081d0fe0 BD |
398 | * default_reset_link - default reset function |
399 | * @dev: pointer to pci_dev data structure | |
89713422 | 400 | * |
081d0fe0 BD |
401 | * Invoked when performing link reset on a Downstream Port or a |
402 | * Root Port with no aer driver. | |
89713422 | 403 | */ |
081d0fe0 | 404 | static pci_ers_result_t default_reset_link(struct pci_dev *dev) |
89713422 | 405 | { |
1b95ce8f | 406 | pci_reset_bridge_secondary_bus(dev); |
081d0fe0 | 407 | dev_printk(KERN_DEBUG, &dev->dev, "downstream link has been reset\n"); |
89713422 HS |
408 | return PCI_ERS_RESULT_RECOVERED; |
409 | } | |
410 | ||
6c2b374d ZY |
411 | static int find_aer_service_iter(struct device *device, void *data) |
412 | { | |
517cae38 | 413 | struct pcie_port_service_driver *service_driver, **drv; |
6c2b374d | 414 | |
517cae38 | 415 | drv = (struct pcie_port_service_driver **) data; |
6c2b374d | 416 | |
4f7ccf6a HS |
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) { | |
517cae38 | 420 | *drv = service_driver; |
4f7ccf6a | 421 | return 1; |
6c2b374d ZY |
422 | } |
423 | } | |
424 | ||
425 | return 0; | |
426 | } | |
427 | ||
517cae38 | 428 | static struct pcie_port_service_driver *find_aer_service(struct pci_dev *dev) |
6c2b374d | 429 | { |
517cae38 HS |
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; | |
6c2b374d ZY |
435 | } |
436 | ||
0918472c | 437 | static pci_ers_result_t reset_link(struct pci_dev *dev) |
6c2b374d ZY |
438 | { |
439 | struct pci_dev *udev; | |
440 | pci_ers_result_t status; | |
517cae38 | 441 | struct pcie_port_service_driver *driver; |
6c2b374d | 442 | |
93de6901 | 443 | if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) { |
89713422 | 444 | /* Reset this port for all subordinates */ |
6c2b374d | 445 | udev = dev; |
89713422 HS |
446 | } else { |
447 | /* Reset the upstream component (likely downstream port) */ | |
c9a91883 | 448 | udev = dev->bus->self; |
89713422 | 449 | } |
6c2b374d | 450 | |
517cae38 HS |
451 | /* Use the aer driver of the component firstly */ |
452 | driver = find_aer_service(udev); | |
6c2b374d | 453 | |
89713422 HS |
454 | if (driver && driver->reset_link) { |
455 | status = driver->reset_link(udev); | |
777e61ea | 456 | } else if (udev->has_secondary_link) { |
081d0fe0 | 457 | status = default_reset_link(udev); |
89713422 HS |
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; | |
6c2b374d ZY |
463 | } |
464 | ||
6c2b374d | 465 | if (status != PCI_ERS_RESULT_RECOVERED) { |
4f7ccf6a HS |
466 | dev_printk(KERN_DEBUG, &dev->dev, |
467 | "link reset at upstream device %s failed\n", | |
468 | pci_name(udev)); | |
6c2b374d ZY |
469 | return PCI_ERS_RESULT_DISCONNECT; |
470 | } | |
471 | ||
472 | return status; | |
473 | } | |
474 | ||
475 | /** | |
476 | * do_recovery - handle nonfatal/fatal error recovery process | |
6c2b374d ZY |
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. | |
d885c6b7 | 483 | */ |
0918472c | 484 | static void do_recovery(struct pci_dev *dev, int severity) |
6c2b374d ZY |
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) { | |
0918472c | 500 | result = reset_link(dev); |
17e21854 HS |
501 | if (result != PCI_ERS_RESULT_RECOVERED) |
502 | goto failed; | |
6c2b374d ZY |
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 | ||
17e21854 HS |
523 | if (status != PCI_ERS_RESULT_RECOVERED) |
524 | goto failed; | |
525 | ||
526 | broadcast_error_message(dev, | |
6c2b374d ZY |
527 | state, |
528 | "resume", | |
529 | report_resume); | |
530 | ||
be5ac3d3 | 531 | dev_info(&dev->dev, "AER: Device recovery successful\n"); |
17e21854 HS |
532 | return; |
533 | ||
534 | failed: | |
535 | /* TODO: Should kernel panic here? */ | |
be5ac3d3 | 536 | dev_info(&dev->dev, "AER: Device recovery failed\n"); |
6c2b374d ZY |
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. | |
d885c6b7 | 546 | */ |
c9a91883 | 547 | static void handle_error_source(struct pcie_device *aerdev, |
6c2b374d | 548 | struct pci_dev *dev, |
28eb27cf | 549 | struct aer_err_info *info) |
6c2b374d | 550 | { |
6c2b374d ZY |
551 | int pos; |
552 | ||
28eb27cf | 553 | if (info->severity == AER_CORRECTABLE) { |
6c2b374d | 554 | /* |
f7625980 | 555 | * Correctable error does not need software intervention. |
6c2b374d ZY |
556 | * No need to go through error recovery process. |
557 | */ | |
0927678f | 558 | pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR); |
6c2b374d ZY |
559 | if (pos) |
560 | pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS, | |
28eb27cf | 561 | info->status); |
17e21854 | 562 | } else |
0918472c | 563 | do_recovery(dev, info->severity); |
6c2b374d ZY |
564 | } |
565 | ||
0918472c HY |
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 | ||
3c78bc61 | 572 | struct aer_recover_entry { |
0918472c HY |
573 | u8 bus; |
574 | u8 devfn; | |
575 | u16 domain; | |
576 | int severity; | |
37448adf | 577 | struct aer_capability_regs *regs; |
0918472c HY |
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 | ||
590 | void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn, | |
37448adf | 591 | int severity, struct aer_capability_regs *aer_regs) |
0918472c HY |
592 | { |
593 | unsigned long flags; | |
594 | struct aer_recover_entry entry = { | |
595 | .bus = bus, | |
596 | .devfn = devfn, | |
597 | .domain = domain, | |
598 | .severity = severity, | |
37448adf | 599 | .regs = aer_regs, |
0918472c HY |
600 | }; |
601 | ||
602 | spin_lock_irqsave(&aer_recover_ring_lock, flags); | |
498d319b | 603 | if (kfifo_put(&aer_recover_ring, entry)) |
0918472c HY |
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 | ||
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 | } | |
37448adf | 626 | cper_print_aer(pdev, entry.severity, entry.regs); |
0918472c | 627 | do_recovery(pdev, entry.severity); |
a82b6af3 | 628 | pci_dev_put(pdev); |
0918472c HY |
629 | } |
630 | } | |
631 | #endif | |
632 | ||
b1c089b7 HS |
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. | |
7c4ec94f HS |
639 | * |
640 | * Note that @info is reused among all error devices. Clear fields properly. | |
b1c089b7 | 641 | */ |
6c2b374d ZY |
642 | static int get_device_error_info(struct pci_dev *dev, struct aer_err_info *info) |
643 | { | |
e7a0d92b | 644 | int pos, temp; |
6c2b374d | 645 | |
7c4ec94f | 646 | /* Must reset in this function */ |
1b4ffcf8 | 647 | info->status = 0; |
273024de | 648 | info->tlp_header_valid = 0; |
1b4ffcf8 | 649 | |
0927678f | 650 | pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR); |
6c2b374d ZY |
651 | |
652 | /* The device might not support AER */ | |
653 | if (!pos) | |
b1c089b7 | 654 | return 1; |
6c2b374d ZY |
655 | |
656 | if (info->severity == AER_CORRECTABLE) { | |
657 | pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, | |
658 | &info->status); | |
0d90c3ac HS |
659 | pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, |
660 | &info->mask); | |
661 | if (!(info->status & ~info->mask)) | |
b1c089b7 | 662 | return 0; |
93de6901 | 663 | } else if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE || |
6c2b374d ZY |
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); | |
0d90c3ac HS |
669 | pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, |
670 | &info->mask); | |
671 | if (!(info->status & ~info->mask)) | |
b1c089b7 | 672 | return 0; |
6c2b374d | 673 | |
e7a0d92b HS |
674 | /* Get First Error Pointer */ |
675 | pci_read_config_dword(dev, pos + PCI_ERR_CAP, &temp); | |
273024de | 676 | info->first_error = PCI_ERR_CAP_FEP(temp); |
e7a0d92b | 677 | |
6c2b374d | 678 | if (info->status & AER_LOG_TLP_MASKS) { |
273024de | 679 | info->tlp_header_valid = 1; |
6c2b374d ZY |
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 | ||
b1c089b7 | 691 | return 1; |
6c2b374d ZY |
692 | } |
693 | ||
3d5505c5 ZY |
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 | ||
b1c089b7 | 699 | /* Report all before handle them, not to lost records by reset etc. */ |
3d5505c5 | 700 | for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) { |
b1c089b7 | 701 | if (get_device_error_info(e_info->dev[i], e_info)) |
3d5505c5 | 702 | aer_print_error(e_info->dev[i], e_info); |
b1c089b7 HS |
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); | |
3d5505c5 ZY |
707 | } |
708 | } | |
709 | ||
6c2b374d ZY |
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 | |
d885c6b7 | 714 | */ |
6c2b374d ZY |
715 | static void aer_isr_one_error(struct pcie_device *p_device, |
716 | struct aer_err_source *e_src) | |
717 | { | |
28eb27cf | 718 | struct aer_err_info *e_info; |
28eb27cf ZY |
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); | |
7c4ec94f | 722 | if (!e_info) { |
28eb27cf ZY |
723 | dev_printk(KERN_DEBUG, &p_device->port->dev, |
724 | "Can't allocate mem when processing AER errors\n"); | |
725 | return; | |
726 | } | |
6c2b374d ZY |
727 | |
728 | /* | |
729 | * There is a possibility that both correctable error and | |
730 | * uncorrectable error being logged. Report correctable error first. | |
731 | */ | |
7c4ec94f HS |
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) | |
273024de | 756 | e_info->multi_error_valid = 1; |
7c4ec94f HS |
757 | else |
758 | e_info->multi_error_valid = 0; | |
28eb27cf | 759 | |
79e4b89b HS |
760 | aer_print_port_info(p_device->port, e_info); |
761 | ||
98ca3964 HS |
762 | if (find_source_device(p_device->port, e_info)) |
763 | aer_process_err_devices(p_device, e_info); | |
6c2b374d | 764 | } |
28eb27cf ZY |
765 | |
766 | kfree(e_info); | |
6c2b374d ZY |
767 | } |
768 | ||
88da13bf HS |
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 | */ | |
778 | static int get_e_source(struct aer_rpc *rpc, struct aer_err_source *e_src) | |
779 | { | |
780 | unsigned long flags; | |
88da13bf HS |
781 | |
782 | /* Lock access to Root error producer/consumer index */ | |
783 | spin_lock_irqsave(&rpc->e_lock, flags); | |
f6735590 LT |
784 | if (rpc->prod_idx == rpc->cons_idx) { |
785 | spin_unlock_irqrestore(&rpc->e_lock, flags); | |
786 | return 0; | |
88da13bf | 787 | } |
f6735590 LT |
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; | |
88da13bf HS |
793 | spin_unlock_irqrestore(&rpc->e_lock, flags); |
794 | ||
f6735590 | 795 | return 1; |
88da13bf HS |
796 | } |
797 | ||
6c2b374d ZY |
798 | /** |
799 | * aer_isr - consume errors detected by root port | |
65f27f38 | 800 | * @work: definition of this work item |
6c2b374d ZY |
801 | * |
802 | * Invoked, as DPC, when root port records new detected error | |
d885c6b7 | 803 | */ |
65f27f38 | 804 | void aer_isr(struct work_struct *work) |
6c2b374d | 805 | { |
65f27f38 DH |
806 | struct aer_rpc *rpc = container_of(work, struct aer_rpc, dpc_handler); |
807 | struct pcie_device *p_device = rpc->rpd; | |
50c1126e | 808 | struct aer_err_source uninitialized_var(e_src); |
6c2b374d ZY |
809 | |
810 | mutex_lock(&rpc->rpc_mutex); | |
88da13bf HS |
811 | while (get_e_source(rpc, &e_src)) |
812 | aer_isr_one_error(p_device, &e_src); | |
6c2b374d | 813 | mutex_unlock(&rpc->rpc_mutex); |
6c2b374d ZY |
814 | } |
815 | ||
6c2b374d ZY |
816 | /** |
817 | * aer_init - provide AER initialization | |
818 | * @dev: pointer to AER pcie device | |
819 | * | |
820 | * Invoked when AER service driver is loaded. | |
d885c6b7 | 821 | */ |
6c2b374d ZY |
822 | int aer_init(struct pcie_device *dev) |
823 | { | |
05843961 MD |
824 | if (forceload) { |
825 | dev_printk(KERN_DEBUG, &dev->device, | |
826 | "aerdrv forceload requested.\n"); | |
affb72c3 | 827 | pcie_aer_force_firmware_first(dev->port, 0); |
05843961 | 828 | } |
28eb5f27 | 829 | return 0; |
6c2b374d | 830 | } |