powerpc/eeh: Trace error based on PE from beginning
[linux-block.git] / arch / powerpc / platforms / pseries / eeh_pseries.c
CommitLineData
aa1e6374
GS
1/*
2 * The file intends to implement the platform dependent EEH operations on pseries.
3 * Actually, the pseries platform is built based on RTAS heavily. That means the
4 * pseries platform dependent EEH operations will be built on RTAS calls. The functions
5 * are devired from arch/powerpc/platforms/pseries/eeh.c and necessary cleanup has
6 * been done.
7 *
8 * Copyright Benjamin Herrenschmidt & Gavin Shan, IBM Corporation 2011.
9 * Copyright IBM Corporation 2001, 2005, 2006
10 * Copyright Dave Engebretsen & Todd Inglett 2001
11 * Copyright Linas Vepstas 2005, 2006
12 *
13 * This program is free software; you can redistribute it and/or modify
14 * it under the terms of the GNU General Public License as published by
15 * the Free Software Foundation; either version 2 of the License, or
16 * (at your option) any later version.
17 *
18 * This program is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21 * GNU General Public License for more details.
22 *
23 * You should have received a copy of the GNU General Public License
24 * along with this program; if not, write to the Free Software
25 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
26 */
27
28#include <linux/atomic.h>
29#include <linux/delay.h>
30#include <linux/export.h>
31#include <linux/init.h>
32#include <linux/list.h>
33#include <linux/of.h>
34#include <linux/pci.h>
35#include <linux/proc_fs.h>
36#include <linux/rbtree.h>
37#include <linux/sched.h>
38#include <linux/seq_file.h>
39#include <linux/spinlock.h>
40
41#include <asm/eeh.h>
42#include <asm/eeh_event.h>
43#include <asm/io.h>
44#include <asm/machdep.h>
45#include <asm/ppc-pci.h>
46#include <asm/rtas.h>
47
e2af155c
GS
48/* RTAS tokens */
49static int ibm_set_eeh_option;
50static int ibm_set_slot_reset;
51static int ibm_read_slot_reset_state;
52static int ibm_read_slot_reset_state2;
53static int ibm_slot_error_detail;
54static int ibm_get_config_addr_info;
55static int ibm_get_config_addr_info2;
56static int ibm_configure_bridge;
57static int ibm_configure_pe;
58
8d633291
GS
59/*
60 * Buffer for reporting slot-error-detail rtas calls. Its here
61 * in BSS, and not dynamically alloced, so that it ends up in
62 * RMO where RTAS can access it.
63 */
64static unsigned char slot_errbuf[RTAS_ERROR_LOG_MAX];
65static DEFINE_SPINLOCK(slot_errbuf_lock);
66static int eeh_error_buf_size;
67
aa1e6374
GS
68/**
69 * pseries_eeh_init - EEH platform dependent initialization
70 *
71 * EEH platform dependent initialization on pseries.
72 */
73static int pseries_eeh_init(void)
74{
e2af155c
GS
75 /* figure out EEH RTAS function call tokens */
76 ibm_set_eeh_option = rtas_token("ibm,set-eeh-option");
77 ibm_set_slot_reset = rtas_token("ibm,set-slot-reset");
78 ibm_read_slot_reset_state2 = rtas_token("ibm,read-slot-reset-state2");
79 ibm_read_slot_reset_state = rtas_token("ibm,read-slot-reset-state");
80 ibm_slot_error_detail = rtas_token("ibm,slot-error-detail");
81 ibm_get_config_addr_info2 = rtas_token("ibm,get-config-addr-info2");
82 ibm_get_config_addr_info = rtas_token("ibm,get-config-addr-info");
83 ibm_configure_pe = rtas_token("ibm,configure-pe");
9c41ef08 84 ibm_configure_bridge = rtas_token("ibm,configure-bridge");
e2af155c
GS
85
86 /* necessary sanity check */
87 if (ibm_set_eeh_option == RTAS_UNKNOWN_SERVICE) {
88 pr_warning("%s: RTAS service <ibm,set-eeh-option> invalid\n",
89 __func__);
90 return -EINVAL;
91 } else if (ibm_set_slot_reset == RTAS_UNKNOWN_SERVICE) {
9c41ef08 92 pr_warning("%s: RTAS service <ibm,set-slot-reset> invalid\n",
e2af155c
GS
93 __func__);
94 return -EINVAL;
95 } else if (ibm_read_slot_reset_state2 == RTAS_UNKNOWN_SERVICE &&
96 ibm_read_slot_reset_state == RTAS_UNKNOWN_SERVICE) {
97 pr_warning("%s: RTAS service <ibm,read-slot-reset-state2> and "
98 "<ibm,read-slot-reset-state> invalid\n",
99 __func__);
100 return -EINVAL;
101 } else if (ibm_slot_error_detail == RTAS_UNKNOWN_SERVICE) {
102 pr_warning("%s: RTAS service <ibm,slot-error-detail> invalid\n",
103 __func__);
104 return -EINVAL;
105 } else if (ibm_get_config_addr_info2 == RTAS_UNKNOWN_SERVICE &&
106 ibm_get_config_addr_info == RTAS_UNKNOWN_SERVICE) {
107 pr_warning("%s: RTAS service <ibm,get-config-addr-info2> and "
108 "<ibm,get-config-addr-info> invalid\n",
109 __func__);
110 return -EINVAL;
111 } else if (ibm_configure_pe == RTAS_UNKNOWN_SERVICE &&
112 ibm_configure_bridge == RTAS_UNKNOWN_SERVICE) {
113 pr_warning("%s: RTAS service <ibm,configure-pe> and "
114 "<ibm,configure-bridge> invalid\n",
115 __func__);
116 return -EINVAL;
117 }
118
8d633291
GS
119 /* Initialize error log lock and size */
120 spin_lock_init(&slot_errbuf_lock);
121 eeh_error_buf_size = rtas_token("rtas-error-log-max");
122 if (eeh_error_buf_size == RTAS_UNKNOWN_SERVICE) {
123 pr_warning("%s: unknown EEH error log size\n",
124 __func__);
125 eeh_error_buf_size = 1024;
126 } else if (eeh_error_buf_size > RTAS_ERROR_LOG_MAX) {
127 pr_warning("%s: EEH error log size %d exceeds the maximal %d\n",
128 __func__, eeh_error_buf_size, RTAS_ERROR_LOG_MAX);
129 eeh_error_buf_size = RTAS_ERROR_LOG_MAX;
130 }
131
aa1e6374
GS
132 return 0;
133}
134
135/**
136 * pseries_eeh_set_option - Initialize EEH or MMIO/DMA reenable
137 * @dn: device node
138 * @option: operation to be issued
139 *
140 * The function is used to control the EEH functionality globally.
141 * Currently, following options are support according to PAPR:
142 * Enable EEH, Disable EEH, Enable MMIO and Enable DMA
143 */
144static int pseries_eeh_set_option(struct device_node *dn, int option)
145{
8fb8f709 146 int ret = 0;
54793d0e 147 struct eeh_dev *edev;
8fb8f709
GS
148 const u32 *reg;
149 int config_addr;
150
54793d0e 151 edev = of_node_to_eeh_dev(dn);
8fb8f709
GS
152
153 /*
154 * When we're enabling or disabling EEH functioality on
155 * the particular PE, the PE config address is possibly
156 * unavailable. Therefore, we have to figure it out from
157 * the FDT node.
158 */
159 switch (option) {
160 case EEH_OPT_DISABLE:
161 case EEH_OPT_ENABLE:
162 reg = of_get_property(dn, "reg", NULL);
163 config_addr = reg[0];
164 break;
165
166 case EEH_OPT_THAW_MMIO:
167 case EEH_OPT_THAW_DMA:
54793d0e
GS
168 config_addr = edev->config_addr;
169 if (edev->pe_config_addr)
170 config_addr = edev->pe_config_addr;
8fb8f709
GS
171 break;
172
173 default:
174 pr_err("%s: Invalid option %d\n",
175 __func__, option);
176 return -EINVAL;
177 }
178
179 ret = rtas_call(ibm_set_eeh_option, 4, 1, NULL,
54793d0e
GS
180 config_addr, BUID_HI(edev->phb->buid),
181 BUID_LO(edev->phb->buid), option);
8fb8f709
GS
182
183 return ret;
aa1e6374
GS
184}
185
186/**
187 * pseries_eeh_get_pe_addr - Retrieve PE address
188 * @dn: device node
189 *
190 * Retrieve the assocated PE address. Actually, there're 2 RTAS
191 * function calls dedicated for the purpose. We need implement
192 * it through the new function and then the old one. Besides,
193 * you should make sure the config address is figured out from
194 * FDT node before calling the function.
195 *
196 * It's notable that zero'ed return value means invalid PE config
197 * address.
198 */
199static int pseries_eeh_get_pe_addr(struct device_node *dn)
200{
54793d0e 201 struct eeh_dev *edev;
c8c29b38
GS
202 int ret = 0;
203 int rets[3];
204
54793d0e 205 edev = of_node_to_eeh_dev(dn);
c8c29b38
GS
206
207 if (ibm_get_config_addr_info2 != RTAS_UNKNOWN_SERVICE) {
208 /*
209 * First of all, we need to make sure there has one PE
210 * associated with the device. Otherwise, PE address is
211 * meaningless.
212 */
213 ret = rtas_call(ibm_get_config_addr_info2, 4, 2, rets,
54793d0e
GS
214 edev->config_addr, BUID_HI(edev->phb->buid),
215 BUID_LO(edev->phb->buid), 1);
c8c29b38
GS
216 if (ret || (rets[0] == 0))
217 return 0;
218
219 /* Retrieve the associated PE config address */
220 ret = rtas_call(ibm_get_config_addr_info2, 4, 2, rets,
54793d0e
GS
221 edev->config_addr, BUID_HI(edev->phb->buid),
222 BUID_LO(edev->phb->buid), 0);
c8c29b38
GS
223 if (ret) {
224 pr_warning("%s: Failed to get PE address for %s\n",
225 __func__, dn->full_name);
226 return 0;
227 }
228
229 return rets[0];
230 }
231
232 if (ibm_get_config_addr_info != RTAS_UNKNOWN_SERVICE) {
233 ret = rtas_call(ibm_get_config_addr_info, 4, 2, rets,
54793d0e
GS
234 edev->config_addr, BUID_HI(edev->phb->buid),
235 BUID_LO(edev->phb->buid), 0);
c8c29b38
GS
236 if (ret) {
237 pr_warning("%s: Failed to get PE address for %s\n",
238 __func__, dn->full_name);
239 return 0;
240 }
241
242 return rets[0];
243 }
244
245 return ret;
aa1e6374
GS
246}
247
248/**
249 * pseries_eeh_get_state - Retrieve PE state
250 * @dn: PE associated device node
251 * @state: return value
252 *
253 * Retrieve the state of the specified PE. On RTAS compliant
254 * pseries platform, there already has one dedicated RTAS function
255 * for the purpose. It's notable that the associated PE config address
256 * might be ready when calling the function. Therefore, endeavour to
257 * use the PE config address if possible. Further more, there're 2
258 * RTAS calls for the purpose, we need to try the new one and back
259 * to the old one if the new one couldn't work properly.
260 */
261static int pseries_eeh_get_state(struct device_node *dn, int *state)
262{
54793d0e 263 struct eeh_dev *edev;
eb594a47
GS
264 int config_addr;
265 int ret;
266 int rets[4];
267 int result;
268
269 /* Figure out PE config address if possible */
54793d0e
GS
270 edev = of_node_to_eeh_dev(dn);
271 config_addr = edev->config_addr;
272 if (edev->pe_config_addr)
273 config_addr = edev->pe_config_addr;
eb594a47
GS
274
275 if (ibm_read_slot_reset_state2 != RTAS_UNKNOWN_SERVICE) {
276 ret = rtas_call(ibm_read_slot_reset_state2, 3, 4, rets,
54793d0e
GS
277 config_addr, BUID_HI(edev->phb->buid),
278 BUID_LO(edev->phb->buid));
eb594a47
GS
279 } else if (ibm_read_slot_reset_state != RTAS_UNKNOWN_SERVICE) {
280 /* Fake PE unavailable info */
281 rets[2] = 0;
282 ret = rtas_call(ibm_read_slot_reset_state, 3, 3, rets,
54793d0e
GS
283 config_addr, BUID_HI(edev->phb->buid),
284 BUID_LO(edev->phb->buid));
eb594a47
GS
285 } else {
286 return EEH_STATE_NOT_SUPPORT;
287 }
288
289 if (ret)
290 return ret;
291
292 /* Parse the result out */
293 result = 0;
294 if (rets[1]) {
295 switch(rets[0]) {
296 case 0:
297 result &= ~EEH_STATE_RESET_ACTIVE;
298 result |= EEH_STATE_MMIO_ACTIVE;
299 result |= EEH_STATE_DMA_ACTIVE;
300 break;
301 case 1:
302 result |= EEH_STATE_RESET_ACTIVE;
303 result |= EEH_STATE_MMIO_ACTIVE;
304 result |= EEH_STATE_DMA_ACTIVE;
305 break;
306 case 2:
307 result &= ~EEH_STATE_RESET_ACTIVE;
308 result &= ~EEH_STATE_MMIO_ACTIVE;
309 result &= ~EEH_STATE_DMA_ACTIVE;
310 break;
311 case 4:
312 result &= ~EEH_STATE_RESET_ACTIVE;
313 result &= ~EEH_STATE_MMIO_ACTIVE;
314 result &= ~EEH_STATE_DMA_ACTIVE;
315 result |= EEH_STATE_MMIO_ENABLED;
316 break;
317 case 5:
318 if (rets[2]) {
319 if (state) *state = rets[2];
320 result = EEH_STATE_UNAVAILABLE;
321 } else {
322 result = EEH_STATE_NOT_SUPPORT;
323 }
324 default:
325 result = EEH_STATE_NOT_SUPPORT;
326 }
327 } else {
328 result = EEH_STATE_NOT_SUPPORT;
329 }
330
331 return result;
aa1e6374
GS
332}
333
334/**
335 * pseries_eeh_reset - Reset the specified PE
336 * @dn: PE associated device node
337 * @option: reset option
338 *
339 * Reset the specified PE
340 */
341static int pseries_eeh_reset(struct device_node *dn, int option)
342{
54793d0e 343 struct eeh_dev *edev;
2652481f
GS
344 int config_addr;
345 int ret;
346
347 /* Figure out PE address */
54793d0e
GS
348 edev = of_node_to_eeh_dev(dn);
349 config_addr = edev->config_addr;
350 if (edev->pe_config_addr)
351 config_addr = edev->pe_config_addr;
2652481f
GS
352
353 /* Reset PE through RTAS call */
354 ret = rtas_call(ibm_set_slot_reset, 4, 1, NULL,
54793d0e
GS
355 config_addr, BUID_HI(edev->phb->buid),
356 BUID_LO(edev->phb->buid), option);
2652481f
GS
357
358 /* If fundamental-reset not supported, try hot-reset */
359 if (option == EEH_RESET_FUNDAMENTAL &&
360 ret == -8) {
361 ret = rtas_call(ibm_set_slot_reset, 4, 1, NULL,
54793d0e
GS
362 config_addr, BUID_HI(edev->phb->buid),
363 BUID_LO(edev->phb->buid), EEH_RESET_HOT);
2652481f
GS
364 }
365
366 return ret;
aa1e6374
GS
367}
368
369/**
370 * pseries_eeh_wait_state - Wait for PE state
371 * @dn: PE associated device node
372 * @max_wait: maximal period in microsecond
373 *
374 * Wait for the state of associated PE. It might take some time
375 * to retrieve the PE's state.
376 */
377static int pseries_eeh_wait_state(struct device_node *dn, int max_wait)
378{
b0e5f742
GS
379 int ret;
380 int mwait;
381
382 /*
383 * According to PAPR, the state of PE might be temporarily
384 * unavailable. Under the circumstance, we have to wait
385 * for indicated time determined by firmware. The maximal
386 * wait time is 5 minutes, which is acquired from the original
387 * EEH implementation. Also, the original implementation
388 * also defined the minimal wait time as 1 second.
389 */
390#define EEH_STATE_MIN_WAIT_TIME (1000)
391#define EEH_STATE_MAX_WAIT_TIME (300 * 1000)
392
393 while (1) {
394 ret = pseries_eeh_get_state(dn, &mwait);
395
396 /*
397 * If the PE's state is temporarily unavailable,
398 * we have to wait for the specified time. Otherwise,
399 * the PE's state will be returned immediately.
400 */
401 if (ret != EEH_STATE_UNAVAILABLE)
402 return ret;
403
404 if (max_wait <= 0) {
405 pr_warning("%s: Timeout when getting PE's state (%d)\n",
406 __func__, max_wait);
407 return EEH_STATE_NOT_SUPPORT;
408 }
409
410 if (mwait <= 0) {
411 pr_warning("%s: Firmware returned bad wait value %d\n",
412 __func__, mwait);
413 mwait = EEH_STATE_MIN_WAIT_TIME;
414 } else if (mwait > EEH_STATE_MAX_WAIT_TIME) {
415 pr_warning("%s: Firmware returned too long wait value %d\n",
416 __func__, mwait);
417 mwait = EEH_STATE_MAX_WAIT_TIME;
418 }
419
420 max_wait -= mwait;
421 msleep(mwait);
422 }
423
424 return EEH_STATE_NOT_SUPPORT;
aa1e6374
GS
425}
426
427/**
428 * pseries_eeh_get_log - Retrieve error log
429 * @dn: device node
430 * @severity: temporary or permanent error log
431 * @drv_log: driver log to be combined with retrieved error log
432 * @len: length of driver log
433 *
434 * Retrieve the temporary or permanent error from the PE.
435 * Actually, the error will be retrieved through the dedicated
436 * RTAS call.
437 */
438static int pseries_eeh_get_log(struct device_node *dn, int severity, char *drv_log, unsigned long len)
439{
54793d0e 440 struct eeh_dev *edev;
8d633291
GS
441 int config_addr;
442 unsigned long flags;
443 int ret;
444
54793d0e 445 edev = of_node_to_eeh_dev(dn);
8d633291
GS
446 spin_lock_irqsave(&slot_errbuf_lock, flags);
447 memset(slot_errbuf, 0, eeh_error_buf_size);
448
449 /* Figure out the PE address */
54793d0e
GS
450 config_addr = edev->config_addr;
451 if (edev->pe_config_addr)
452 config_addr = edev->pe_config_addr;
8d633291
GS
453
454 ret = rtas_call(ibm_slot_error_detail, 8, 1, NULL, config_addr,
54793d0e 455 BUID_HI(edev->phb->buid), BUID_LO(edev->phb->buid),
8d633291
GS
456 virt_to_phys(drv_log), len,
457 virt_to_phys(slot_errbuf), eeh_error_buf_size,
458 severity);
459 if (!ret)
460 log_error(slot_errbuf, ERR_TYPE_RTAS_LOG, 0);
461 spin_unlock_irqrestore(&slot_errbuf_lock, flags);
462
463 return ret;
aa1e6374
GS
464}
465
466/**
467 * pseries_eeh_configure_bridge - Configure PCI bridges in the indicated PE
468 * @dn: PE associated device node
469 *
470 * The function will be called to reconfigure the bridges included
471 * in the specified PE so that the mulfunctional PE would be recovered
472 * again.
473 */
474static int pseries_eeh_configure_bridge(struct device_node *dn)
475{
54793d0e 476 struct eeh_dev *edev;
1823fbf1
GS
477 int config_addr;
478 int ret;
479
480 /* Figure out the PE address */
54793d0e
GS
481 edev = of_node_to_eeh_dev(dn);
482 config_addr = edev->config_addr;
483 if (edev->pe_config_addr)
484 config_addr = edev->pe_config_addr;
1823fbf1
GS
485
486 /* Use new configure-pe function, if supported */
487 if (ibm_configure_pe != RTAS_UNKNOWN_SERVICE) {
488 ret = rtas_call(ibm_configure_pe, 3, 1, NULL,
54793d0e
GS
489 config_addr, BUID_HI(edev->phb->buid),
490 BUID_LO(edev->phb->buid));
1823fbf1
GS
491 } else if (ibm_configure_bridge != RTAS_UNKNOWN_SERVICE) {
492 ret = rtas_call(ibm_configure_bridge, 3, 1, NULL,
54793d0e
GS
493 config_addr, BUID_HI(edev->phb->buid),
494 BUID_LO(edev->phb->buid));
1823fbf1
GS
495 } else {
496 return -EFAULT;
497 }
498
499 if (ret)
500 pr_warning("%s: Unable to configure bridge %d for %s\n",
501 __func__, ret, dn->full_name);
502
503 return ret;
aa1e6374
GS
504}
505
3780444c
GS
506/**
507 * pseries_eeh_read_config - Read PCI config space
508 * @dn: device node
509 * @where: PCI address
510 * @size: size to read
511 * @val: return value
512 *
513 * Read config space from the speicifed device
514 */
515static int pseries_eeh_read_config(struct device_node *dn, int where, int size, u32 *val)
516{
517 struct pci_dn *pdn;
518
519 pdn = PCI_DN(dn);
520
521 return rtas_read_config(pdn, where, size, val);
522}
523
524/**
525 * pseries_eeh_write_config - Write PCI config space
526 * @dn: device node
527 * @where: PCI address
528 * @size: size to write
529 * @val: value to be written
530 *
531 * Write config space to the specified device
532 */
533static int pseries_eeh_write_config(struct device_node *dn, int where, int size, u32 val)
534{
535 struct pci_dn *pdn;
536
537 pdn = PCI_DN(dn);
538
539 return rtas_write_config(pdn, where, size, val);
540}
541
aa1e6374
GS
542static struct eeh_ops pseries_eeh_ops = {
543 .name = "pseries",
544 .init = pseries_eeh_init,
545 .set_option = pseries_eeh_set_option,
546 .get_pe_addr = pseries_eeh_get_pe_addr,
547 .get_state = pseries_eeh_get_state,
548 .reset = pseries_eeh_reset,
549 .wait_state = pseries_eeh_wait_state,
550 .get_log = pseries_eeh_get_log,
3780444c
GS
551 .configure_bridge = pseries_eeh_configure_bridge,
552 .read_config = pseries_eeh_read_config,
553 .write_config = pseries_eeh_write_config
aa1e6374
GS
554};
555
556/**
557 * eeh_pseries_init - Register platform dependent EEH operations
558 *
559 * EEH initialization on pseries platform. This function should be
560 * called before any EEH related functions.
561 */
35e5cfe2 562static int __init eeh_pseries_init(void)
aa1e6374 563{
3ea1ae98
GS
564 int ret = -EINVAL;
565
566 if (!machine_is(pseries))
567 return ret;
568
569 ret = eeh_ops_register(&pseries_eeh_ops);
570 if (!ret)
571 pr_info("EEH: pSeries platform initialized\n");
572 else
573 pr_info("EEH: pSeries platform initialization failure (%d)\n",
574 ret);
575
576 return ret;
aa1e6374 577}
35e5cfe2
GS
578
579early_initcall(eeh_pseries_init);