Commit | Line | Data |
---|---|---|
77bd7415 LV |
1 | /* |
2 | * PCI Error Recovery Driver for RPA-compliant PPC64 platform. | |
3c8c90ab LV |
3 | * Copyright IBM Corp. 2004 2005 |
4 | * Copyright Linas Vepstas <linas@linas.org> 2004, 2005 | |
77bd7415 LV |
5 | * |
6 | * All rights reserved. | |
7 | * | |
8 | * This program is free software; you can redistribute it and/or modify | |
9 | * it under the terms of the GNU General Public License as published by | |
10 | * the Free Software Foundation; either version 2 of the License, or (at | |
11 | * your option) any later version. | |
12 | * | |
13 | * This program is distributed in the hope that it will be useful, but | |
14 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | * MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or | |
16 | * NON INFRINGEMENT. See the GNU General Public License for more | |
17 | * details. | |
18 | * | |
19 | * You should have received a copy of the GNU General Public License | |
20 | * along with this program; if not, write to the Free Software | |
21 | * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. | |
22 | * | |
3c8c90ab | 23 | * Send comments and feedback to Linas Vepstas <linas@austin.ibm.com> |
77bd7415 LV |
24 | */ |
25 | #include <linux/delay.h> | |
77bd7415 | 26 | #include <linux/interrupt.h> |
ac325acd | 27 | #include <linux/irq.h> |
77bd7415 LV |
28 | #include <linux/pci.h> |
29 | #include <asm/eeh.h> | |
30 | #include <asm/eeh_event.h> | |
31 | #include <asm/ppc-pci.h> | |
32 | #include <asm/pci-bridge.h> | |
33 | #include <asm/prom.h> | |
34 | #include <asm/rtas.h> | |
35 | ||
29f8bf1b GS |
36 | /** |
37 | * eeh_pcid_name - Retrieve name of PCI device driver | |
38 | * @pdev: PCI device | |
39 | * | |
40 | * This routine is used to retrieve the name of PCI device driver | |
41 | * if that's valid. | |
42 | */ | |
40a7cd92 | 43 | static inline const char *eeh_pcid_name(struct pci_dev *pdev) |
77bd7415 | 44 | { |
273d2803 | 45 | if (pdev && pdev->dev.driver) |
77bd7415 LV |
46 | return pdev->dev.driver->name; |
47 | return ""; | |
48 | } | |
49 | ||
dcfcfe75 | 50 | #if 0 |
8e01520c | 51 | static void print_device_node_tree(struct pci_dn *pdn, int dent) |
77bd7415 LV |
52 | { |
53 | int i; | |
8e01520c AM |
54 | struct device_node *pc; |
55 | ||
56 | if (!pdn) | |
57 | return; | |
58 | for (i = 0; i < dent; i++) | |
77bd7415 LV |
59 | printk(" "); |
60 | printk("dn=%s mode=%x \tcfg_addr=%x pe_addr=%x \tfull=%s\n", | |
61 | pdn->node->name, pdn->eeh_mode, pdn->eeh_config_addr, | |
62 | pdn->eeh_pe_config_addr, pdn->node->full_name); | |
63 | dent += 3; | |
8e01520c | 64 | pc = pdn->node->child; |
77bd7415 LV |
65 | while (pc) { |
66 | print_device_node_tree(PCI_DN(pc), dent); | |
67 | pc = pc->sibling; | |
68 | } | |
69 | } | |
70 | #endif | |
71 | ||
8535ef05 | 72 | /** |
29f8bf1b GS |
73 | * eeh_disable_irq - Disable interrupt for the recovering device |
74 | * @dev: PCI device | |
75 | * | |
76 | * This routine must be called when reporting temporary or permanent | |
77 | * error to the particular PCI device to disable interrupt of that | |
78 | * device. If the device has enabled MSI or MSI-X interrupt, we needn't | |
79 | * do real work because EEH should freeze DMA transfers for those PCI | |
80 | * devices encountering EEH errors, which includes MSI or MSI-X. | |
8535ef05 MM |
81 | */ |
82 | static void eeh_disable_irq(struct pci_dev *dev) | |
83 | { | |
40a7cd92 | 84 | struct eeh_dev *edev = pci_dev_to_eeh_dev(dev); |
8535ef05 MM |
85 | |
86 | /* Don't disable MSI and MSI-X interrupts. They are | |
87 | * effectively disabled by the DMA Stopped state | |
88 | * when an EEH error occurs. | |
29f8bf1b | 89 | */ |
8535ef05 MM |
90 | if (dev->msi_enabled || dev->msix_enabled) |
91 | return; | |
92 | ||
59e3f837 | 93 | if (!irq_has_action(dev->irq)) |
8535ef05 MM |
94 | return; |
95 | ||
dbbceee1 | 96 | edev->mode |= EEH_DEV_IRQ_DISABLED; |
8535ef05 MM |
97 | disable_irq_nosync(dev->irq); |
98 | } | |
99 | ||
100 | /** | |
29f8bf1b GS |
101 | * eeh_enable_irq - Enable interrupt for the recovering device |
102 | * @dev: PCI device | |
103 | * | |
104 | * This routine must be called to enable interrupt while failed | |
105 | * device could be resumed. | |
8535ef05 MM |
106 | */ |
107 | static void eeh_enable_irq(struct pci_dev *dev) | |
108 | { | |
40a7cd92 | 109 | struct eeh_dev *edev = pci_dev_to_eeh_dev(dev); |
8535ef05 | 110 | |
dbbceee1 GS |
111 | if ((edev->mode) & EEH_DEV_IRQ_DISABLED) { |
112 | edev->mode &= ~EEH_DEV_IRQ_DISABLED; | |
8535ef05 MM |
113 | enable_irq(dev->irq); |
114 | } | |
115 | } | |
116 | ||
cb5b5624 | 117 | /** |
29f8bf1b | 118 | * eeh_report_error - Report pci error to each device driver |
9b3c76f0 | 119 | * @data: eeh device |
29f8bf1b | 120 | * @userdata: return value |
cb5b5624 LV |
121 | * |
122 | * Report an EEH error to each device driver, collect up and | |
123 | * merge the device driver responses. Cumulative response | |
124 | * passed back in "userdata". | |
77bd7415 | 125 | */ |
9b3c76f0 | 126 | static void *eeh_report_error(void *data, void *userdata) |
77bd7415 | 127 | { |
9b3c76f0 GS |
128 | struct eeh_dev *edev = (struct eeh_dev *)data; |
129 | struct pci_dev *dev = eeh_dev_to_pci_dev(edev); | |
18eb3b39 | 130 | enum pci_ers_result rc, *res = userdata; |
77bd7415 LV |
131 | struct pci_driver *driver = dev->driver; |
132 | ||
9b3c76f0 GS |
133 | /* We might not have the associated PCI device, |
134 | * then we should continue for next one. | |
135 | */ | |
136 | if (!dev) return NULL; | |
137 | ||
77bd7415 LV |
138 | dev->error_state = pci_channel_io_frozen; |
139 | ||
140 | if (!driver) | |
9b3c76f0 | 141 | return NULL; |
77bd7415 | 142 | |
8535ef05 MM |
143 | eeh_disable_irq(dev); |
144 | ||
6a1ca373 LV |
145 | if (!driver->err_handler || |
146 | !driver->err_handler->error_detected) | |
9b3c76f0 | 147 | return NULL; |
77bd7415 | 148 | |
29f8bf1b | 149 | rc = driver->err_handler->error_detected(dev, pci_channel_io_frozen); |
2a50f144 LV |
150 | |
151 | /* A driver that needs a reset trumps all others */ | |
152 | if (rc == PCI_ERS_RESULT_NEED_RESET) *res = rc; | |
18eb3b39 | 153 | if (*res == PCI_ERS_RESULT_NONE) *res = rc; |
70298c6e | 154 | |
9b3c76f0 | 155 | return NULL; |
6a1ca373 LV |
156 | } |
157 | ||
158 | /** | |
29f8bf1b | 159 | * eeh_report_mmio_enabled - Tell drivers that MMIO has been enabled |
9b3c76f0 | 160 | * @data: eeh device |
29f8bf1b | 161 | * @userdata: return value |
6a1ca373 | 162 | * |
638799b3 LV |
163 | * Tells each device driver that IO ports, MMIO and config space I/O |
164 | * are now enabled. Collects up and merges the device driver responses. | |
165 | * Cumulative response passed back in "userdata". | |
6a1ca373 | 166 | */ |
9b3c76f0 | 167 | static void *eeh_report_mmio_enabled(void *data, void *userdata) |
6a1ca373 | 168 | { |
9b3c76f0 GS |
169 | struct eeh_dev *edev = (struct eeh_dev *)data; |
170 | struct pci_dev *dev = eeh_dev_to_pci_dev(edev); | |
6a1ca373 | 171 | enum pci_ers_result rc, *res = userdata; |
9b3c76f0 | 172 | struct pci_driver *driver; |
6a1ca373 | 173 | |
9b3c76f0 GS |
174 | if (!dev) return NULL; |
175 | ||
176 | if (!(driver = dev->driver) || | |
6a1ca373 LV |
177 | !driver->err_handler || |
178 | !driver->err_handler->mmio_enabled) | |
9b3c76f0 | 179 | return NULL; |
6a1ca373 | 180 | |
29f8bf1b | 181 | rc = driver->err_handler->mmio_enabled(dev); |
2a50f144 LV |
182 | |
183 | /* A driver that needs a reset trumps all others */ | |
184 | if (rc == PCI_ERS_RESULT_NEED_RESET) *res = rc; | |
6a1ca373 | 185 | if (*res == PCI_ERS_RESULT_NONE) *res = rc; |
70298c6e | 186 | |
9b3c76f0 | 187 | return NULL; |
77bd7415 LV |
188 | } |
189 | ||
cb5b5624 | 190 | /** |
29f8bf1b | 191 | * eeh_report_reset - Tell device that slot has been reset |
9b3c76f0 | 192 | * @data: eeh device |
29f8bf1b GS |
193 | * @userdata: return value |
194 | * | |
195 | * This routine must be called while EEH tries to reset particular | |
196 | * PCI device so that the associated PCI device driver could take | |
197 | * some actions, usually to save data the driver needs so that the | |
198 | * driver can work again while the device is recovered. | |
77bd7415 | 199 | */ |
9b3c76f0 | 200 | static void *eeh_report_reset(void *data, void *userdata) |
77bd7415 | 201 | { |
9b3c76f0 GS |
202 | struct eeh_dev *edev = (struct eeh_dev *)data; |
203 | struct pci_dev *dev = eeh_dev_to_pci_dev(edev); | |
6a1ca373 | 204 | enum pci_ers_result rc, *res = userdata; |
9b3c76f0 | 205 | struct pci_driver *driver; |
77bd7415 | 206 | |
9b3c76f0 GS |
207 | if (!dev || !(driver = dev->driver)) |
208 | return NULL; | |
77bd7415 | 209 | |
c58dc575 MM |
210 | dev->error_state = pci_channel_io_normal; |
211 | ||
8535ef05 MM |
212 | eeh_enable_irq(dev); |
213 | ||
6a1ca373 LV |
214 | if (!driver->err_handler || |
215 | !driver->err_handler->slot_reset) | |
9b3c76f0 | 216 | return NULL; |
77bd7415 | 217 | |
6a1ca373 | 218 | rc = driver->err_handler->slot_reset(dev); |
5794dbcb LV |
219 | if ((*res == PCI_ERS_RESULT_NONE) || |
220 | (*res == PCI_ERS_RESULT_RECOVERED)) *res = rc; | |
6a1ca373 LV |
221 | if (*res == PCI_ERS_RESULT_DISCONNECT && |
222 | rc == PCI_ERS_RESULT_NEED_RESET) *res = rc; | |
70298c6e | 223 | |
9b3c76f0 | 224 | return NULL; |
77bd7415 LV |
225 | } |
226 | ||
cb5b5624 | 227 | /** |
29f8bf1b | 228 | * eeh_report_resume - Tell device to resume normal operations |
9b3c76f0 | 229 | * @data: eeh device |
29f8bf1b GS |
230 | * @userdata: return value |
231 | * | |
232 | * This routine must be called to notify the device driver that it | |
233 | * could resume so that the device driver can do some initialization | |
234 | * to make the recovered device work again. | |
cb5b5624 | 235 | */ |
9b3c76f0 | 236 | static void *eeh_report_resume(void *data, void *userdata) |
77bd7415 | 237 | { |
9b3c76f0 GS |
238 | struct eeh_dev *edev = (struct eeh_dev *)data; |
239 | struct pci_dev *dev = eeh_dev_to_pci_dev(edev); | |
240 | struct pci_driver *driver; | |
241 | ||
242 | if (!dev) return NULL; | |
77bd7415 LV |
243 | |
244 | dev->error_state = pci_channel_io_normal; | |
245 | ||
9b3c76f0 GS |
246 | if (!(driver = dev->driver)) |
247 | return NULL; | |
d0e70341 | 248 | |
8535ef05 MM |
249 | eeh_enable_irq(dev); |
250 | ||
d0e70341 LV |
251 | if (!driver->err_handler || |
252 | !driver->err_handler->resume) | |
9b3c76f0 | 253 | return NULL; |
77bd7415 LV |
254 | |
255 | driver->err_handler->resume(dev); | |
70298c6e | 256 | |
9b3c76f0 | 257 | return NULL; |
77bd7415 LV |
258 | } |
259 | ||
cb5b5624 | 260 | /** |
29f8bf1b | 261 | * eeh_report_failure - Tell device driver that device is dead. |
9b3c76f0 | 262 | * @data: eeh device |
29f8bf1b | 263 | * @userdata: return value |
cb5b5624 LV |
264 | * |
265 | * This informs the device driver that the device is permanently | |
266 | * dead, and that no further recovery attempts will be made on it. | |
267 | */ | |
9b3c76f0 | 268 | static void *eeh_report_failure(void *data, void *userdata) |
77bd7415 | 269 | { |
9b3c76f0 GS |
270 | struct eeh_dev *edev = (struct eeh_dev *)data; |
271 | struct pci_dev *dev = eeh_dev_to_pci_dev(edev); | |
272 | struct pci_driver *driver; | |
273 | ||
274 | if (!dev) return NULL; | |
77bd7415 LV |
275 | |
276 | dev->error_state = pci_channel_io_perm_failure; | |
277 | ||
9b3c76f0 GS |
278 | if (!(driver = dev->driver)) |
279 | return NULL; | |
77bd7415 | 280 | |
8535ef05 MM |
281 | eeh_disable_irq(dev); |
282 | ||
283 | if (!driver->err_handler || | |
284 | !driver->err_handler->error_detected) | |
9b3c76f0 | 285 | return NULL; |
8535ef05 | 286 | |
77bd7415 | 287 | driver->err_handler->error_detected(dev, pci_channel_io_perm_failure); |
70298c6e | 288 | |
9b3c76f0 | 289 | return NULL; |
77bd7415 LV |
290 | } |
291 | ||
77bd7415 | 292 | /** |
29f8bf1b | 293 | * eeh_reset_device - Perform actual reset of a pci slot |
9b3c76f0 | 294 | * @pe: EEH PE |
29f8bf1b | 295 | * @bus: PCI bus corresponding to the isolcated slot |
77bd7415 | 296 | * |
29f8bf1b GS |
297 | * This routine must be called to do reset on the indicated PE. |
298 | * During the reset, udev might be invoked because those affected | |
299 | * PCI devices will be removed and then added. | |
77bd7415 | 300 | */ |
9b3c76f0 | 301 | static int eeh_reset_device(struct eeh_pe *pe, struct pci_bus *bus) |
77bd7415 | 302 | { |
42405456 LV |
303 | int cnt, rc; |
304 | ||
305 | /* pcibios will clear the counter; save the value */ | |
9b3c76f0 | 306 | cnt = pe->freeze_count; |
42405456 | 307 | |
77bd7415 LV |
308 | if (bus) |
309 | pcibios_remove_pci_devices(bus); | |
310 | ||
311 | /* Reset the pci controller. (Asserts RST#; resets config space). | |
b6495c0c | 312 | * Reconfigure bridges and devices. Don't try to bring the system |
29f8bf1b GS |
313 | * up if the reset failed for some reason. |
314 | */ | |
9b3c76f0 | 315 | rc = eeh_reset_pe(pe); |
b6495c0c LV |
316 | if (rc) |
317 | return rc; | |
77bd7415 | 318 | |
9b3c76f0 GS |
319 | /* Restore PE */ |
320 | eeh_ops->configure_bridge(pe); | |
321 | eeh_pe_restore_bars(pe); | |
77bd7415 LV |
322 | |
323 | /* Give the system 5 seconds to finish running the user-space | |
324 | * hotplug shutdown scripts, e.g. ifdown for ethernet. Yes, | |
325 | * this is a hack, but if we don't do this, and try to bring | |
326 | * the device up before the scripts have taken it down, | |
327 | * potentially weird things happen. | |
328 | */ | |
329 | if (bus) { | |
29f8bf1b | 330 | ssleep(5); |
77bd7415 LV |
331 | pcibios_add_pci_devices(bus); |
332 | } | |
9b3c76f0 | 333 | pe->freeze_count = cnt; |
b6495c0c LV |
334 | |
335 | return 0; | |
77bd7415 LV |
336 | } |
337 | ||
338 | /* The longest amount of time to wait for a pci device | |
339 | * to come back on line, in seconds. | |
340 | */ | |
fa1be476 | 341 | #define MAX_WAIT_FOR_RECOVERY 150 |
77bd7415 | 342 | |
29f8bf1b GS |
343 | /** |
344 | * eeh_handle_event - Reset a PCI device after hard lockup. | |
9b3c76f0 | 345 | * @pe: EEH PE |
29f8bf1b GS |
346 | * |
347 | * While PHB detects address or data parity errors on particular PCI | |
348 | * slot, the associated PE will be frozen. Besides, DMA's occurring | |
349 | * to wild addresses (which usually happen due to bugs in device | |
350 | * drivers or in PCI adapter firmware) can cause EEH error. #SERR, | |
351 | * #PERR or other misc PCI-related errors also can trigger EEH errors. | |
352 | * | |
353 | * Recovery process consists of unplugging the device driver (which | |
354 | * generated hotplug events to userspace), then issuing a PCI #RST to | |
355 | * the device, then reconfiguring the PCI config space for all bridges | |
356 | * & devices under this slot, and then finally restarting the device | |
357 | * drivers (which cause a second set of hotplug events to go out to | |
358 | * userspace). | |
359 | */ | |
9b3c76f0 | 360 | void eeh_handle_event(struct eeh_pe *pe) |
77bd7415 | 361 | { |
77bd7415 | 362 | struct pci_bus *frozen_bus; |
b6495c0c | 363 | int rc = 0; |
18eb3b39 | 364 | enum pci_ers_result result = PCI_ERS_RESULT_NONE; |
77bd7415 | 365 | |
9b3c76f0 | 366 | frozen_bus = eeh_pe_bus_get(pe); |
77bd7415 | 367 | if (!frozen_bus) { |
9b3c76f0 GS |
368 | pr_err("%s: Cannot find PCI bus for PHB#%d-PE#%x\n", |
369 | __func__, pe->phb->global_number, pe->addr); | |
370 | return; | |
77bd7415 LV |
371 | } |
372 | ||
9b3c76f0 GS |
373 | pe->freeze_count++; |
374 | if (pe->freeze_count > EEH_MAX_ALLOWED_FREEZES) | |
8df83028 | 375 | goto excess_failures; |
9b3c76f0 GS |
376 | pr_warning("EEH: This PCI device has failed %d times in the last hour\n", |
377 | pe->freeze_count); | |
77bd7415 LV |
378 | |
379 | /* Walk the various device drivers attached to this slot through | |
380 | * a reset sequence, giving each an opportunity to do what it needs | |
381 | * to accomplish the reset. Each child gets a report of the | |
382 | * status ... if any child can't handle the reset, then the entire | |
383 | * slot is dlpar removed and added. | |
384 | */ | |
9b3c76f0 | 385 | eeh_pe_dev_traverse(pe, eeh_report_error, &result); |
77bd7415 | 386 | |
5f1a7c81 | 387 | /* Get the current PCI slot state. This can take a long time, |
29f8bf1b GS |
388 | * sometimes over 3 seconds for certain systems. |
389 | */ | |
9b3c76f0 | 390 | rc = eeh_ops->wait_state(pe, MAX_WAIT_FOR_RECOVERY*1000); |
eb594a47 | 391 | if (rc < 0 || rc == EEH_STATE_NOT_SUPPORT) { |
5f1a7c81 LV |
392 | printk(KERN_WARNING "EEH: Permanent failure\n"); |
393 | goto hard_fail; | |
394 | } | |
395 | ||
ede8ca26 LV |
396 | /* Since rtas may enable MMIO when posting the error log, |
397 | * don't post the error log until after all dev drivers | |
17213c3b LV |
398 | * have been informed. |
399 | */ | |
9b3c76f0 | 400 | eeh_slot_error_detail(pe, EEH_LOG_TEMP); |
ede8ca26 | 401 | |
77bd7415 LV |
402 | /* If all device drivers were EEH-unaware, then shut |
403 | * down all of the device drivers, and hope they | |
404 | * go down willingly, without panicing the system. | |
405 | */ | |
18eb3b39 | 406 | if (result == PCI_ERS_RESULT_NONE) { |
9b3c76f0 | 407 | rc = eeh_reset_device(pe, frozen_bus); |
e0f90b64 LV |
408 | if (rc) { |
409 | printk(KERN_WARNING "EEH: Unable to reset, rc=%d\n", rc); | |
b6495c0c | 410 | goto hard_fail; |
e0f90b64 | 411 | } |
77bd7415 LV |
412 | } |
413 | ||
6a1ca373 LV |
414 | /* If all devices reported they can proceed, then re-enable MMIO */ |
415 | if (result == PCI_ERS_RESULT_CAN_RECOVER) { | |
9b3c76f0 | 416 | rc = eeh_pci_enable(pe, EEH_OPT_THAW_MMIO); |
6a1ca373 | 417 | |
fa1be476 LV |
418 | if (rc < 0) |
419 | goto hard_fail; | |
6a1ca373 LV |
420 | if (rc) { |
421 | result = PCI_ERS_RESULT_NEED_RESET; | |
422 | } else { | |
423 | result = PCI_ERS_RESULT_NONE; | |
9b3c76f0 | 424 | eeh_pe_dev_traverse(pe, eeh_report_mmio_enabled, &result); |
6a1ca373 | 425 | } |
77bd7415 LV |
426 | } |
427 | ||
6a1ca373 | 428 | /* If all devices reported they can proceed, then re-enable DMA */ |
18eb3b39 | 429 | if (result == PCI_ERS_RESULT_CAN_RECOVER) { |
9b3c76f0 | 430 | rc = eeh_pci_enable(pe, EEH_OPT_THAW_DMA); |
6a1ca373 | 431 | |
fa1be476 LV |
432 | if (rc < 0) |
433 | goto hard_fail; | |
6a1ca373 LV |
434 | if (rc) |
435 | result = PCI_ERS_RESULT_NEED_RESET; | |
d0e70341 LV |
436 | else |
437 | result = PCI_ERS_RESULT_RECOVERED; | |
6a1ca373 LV |
438 | } |
439 | ||
440 | /* If any device has a hard failure, then shut off everything. */ | |
e0f90b64 LV |
441 | if (result == PCI_ERS_RESULT_DISCONNECT) { |
442 | printk(KERN_WARNING "EEH: Device driver gave up\n"); | |
6a1ca373 | 443 | goto hard_fail; |
e0f90b64 | 444 | } |
6a1ca373 LV |
445 | |
446 | /* If any device called out for a reset, then reset the slot */ | |
447 | if (result == PCI_ERS_RESULT_NEED_RESET) { | |
9b3c76f0 | 448 | rc = eeh_reset_device(pe, NULL); |
e0f90b64 LV |
449 | if (rc) { |
450 | printk(KERN_WARNING "EEH: Cannot reset, rc=%d\n", rc); | |
b6495c0c | 451 | goto hard_fail; |
e0f90b64 | 452 | } |
6a1ca373 | 453 | result = PCI_ERS_RESULT_NONE; |
9b3c76f0 | 454 | eeh_pe_dev_traverse(pe, eeh_report_reset, &result); |
77bd7415 LV |
455 | } |
456 | ||
6a1ca373 | 457 | /* All devices should claim they have recovered by now. */ |
90fdd613 LV |
458 | if ((result != PCI_ERS_RESULT_RECOVERED) && |
459 | (result != PCI_ERS_RESULT_NONE)) { | |
e0f90b64 | 460 | printk(KERN_WARNING "EEH: Not recovered\n"); |
6a1ca373 | 461 | goto hard_fail; |
e0f90b64 | 462 | } |
6a1ca373 | 463 | |
77bd7415 | 464 | /* Tell all device drivers that they can resume operations */ |
9b3c76f0 | 465 | eeh_pe_dev_traverse(pe, eeh_report_resume, NULL); |
b6495c0c | 466 | |
9b3c76f0 | 467 | return; |
b6495c0c | 468 | |
8df83028 | 469 | excess_failures: |
b6495c0c LV |
470 | /* |
471 | * About 90% of all real-life EEH failures in the field | |
472 | * are due to poorly seated PCI cards. Only 10% or so are | |
473 | * due to actual, failed cards. | |
474 | */ | |
9b3c76f0 GS |
475 | pr_err("EEH: PHB#%d-PE#%x has failed %d times in the\n" |
476 | "last hour and has been permanently disabled.\n" | |
477 | "Please try reseating or replacing it.\n", | |
478 | pe->phb->global_number, pe->addr, | |
479 | pe->freeze_count); | |
8df83028 LV |
480 | goto perm_error; |
481 | ||
482 | hard_fail: | |
9b3c76f0 GS |
483 | pr_err("EEH: Unable to recover from failure from PHB#%d-PE#%x.\n" |
484 | "Please try reseating or replacing it\n", | |
485 | pe->phb->global_number, pe->addr); | |
b6495c0c | 486 | |
8df83028 | 487 | perm_error: |
9b3c76f0 | 488 | eeh_slot_error_detail(pe, EEH_LOG_PERM); |
b6495c0c LV |
489 | |
490 | /* Notify all devices that they're about to go down. */ | |
9b3c76f0 | 491 | eeh_pe_dev_traverse(pe, eeh_report_failure, NULL); |
b6495c0c LV |
492 | |
493 | /* Shut down the device drivers for good. */ | |
9b3c76f0 GS |
494 | if (frozen_bus) |
495 | pcibios_remove_pci_devices(frozen_bus); | |
77bd7415 LV |
496 | } |
497 |