• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * eeh.c
3  * Copyright IBM Corporation 2001, 2005, 2006
4  * Copyright Dave Engebretsen & Todd Inglett 2001
5  * Copyright Linas Vepstas 2005, 2006
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License as published by
9  * the Free Software Foundation; either version 2 of the License, or
10  * (at your option) any later version.
11  *
12  * This program is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15  * GNU General Public License for more details.
16  *
17  * You should have received a copy of the GNU General Public License
18  * along with this program; if not, write to the Free Software
19  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
20  *
21  * Please address comments and feedback to Linas Vepstas <linas@austin.ibm.com>
22  */
23 
24 #undef DEBUG
25 
26 #include <linux/delay.h>
27 #include <linux/init.h>
28 #include <linux/list.h>
29 #include <linux/pci.h>
30 #include <linux/proc_fs.h>
31 #include <linux/rbtree.h>
32 #include <linux/seq_file.h>
33 #include <linux/spinlock.h>
34 #include <linux/of.h>
35 
36 #include <asm/atomic.h>
37 #include <asm/eeh.h>
38 #include <asm/eeh_event.h>
39 #include <asm/io.h>
40 #include <asm/machdep.h>
41 #include <asm/ppc-pci.h>
42 #include <asm/rtas.h>
43 
44 
45 /** Overview:
46  *  EEH, or "Extended Error Handling" is a PCI bridge technology for
47  *  dealing with PCI bus errors that can't be dealt with within the
48  *  usual PCI framework, except by check-stopping the CPU.  Systems
49  *  that are designed for high-availability/reliability cannot afford
50  *  to crash due to a "mere" PCI error, thus the need for EEH.
51  *  An EEH-capable bridge operates by converting a detected error
52  *  into a "slot freeze", taking the PCI adapter off-line, making
53  *  the slot behave, from the OS'es point of view, as if the slot
54  *  were "empty": all reads return 0xff's and all writes are silently
55  *  ignored.  EEH slot isolation events can be triggered by parity
56  *  errors on the address or data busses (e.g. during posted writes),
57  *  which in turn might be caused by low voltage on the bus, dust,
58  *  vibration, humidity, radioactivity or plain-old failed hardware.
59  *
60  *  Note, however, that one of the leading causes of EEH slot
61  *  freeze events are buggy device drivers, buggy device microcode,
62  *  or buggy device hardware.  This is because any attempt by the
63  *  device to bus-master data to a memory address that is not
64  *  assigned to the device will trigger a slot freeze.   (The idea
65  *  is to prevent devices-gone-wild from corrupting system memory).
66  *  Buggy hardware/drivers will have a miserable time co-existing
67  *  with EEH.
68  *
69  *  Ideally, a PCI device driver, when suspecting that an isolation
70  *  event has occured (e.g. by reading 0xff's), will then ask EEH
71  *  whether this is the case, and then take appropriate steps to
72  *  reset the PCI slot, the PCI device, and then resume operations.
73  *  However, until that day,  the checking is done here, with the
74  *  eeh_check_failure() routine embedded in the MMIO macros.  If
75  *  the slot is found to be isolated, an "EEH Event" is synthesized
76  *  and sent out for processing.
77  */
78 
79 /* If a device driver keeps reading an MMIO register in an interrupt
80  * handler after a slot isolation event, it might be broken.
81  * This sets the threshold for how many read attempts we allow
82  * before printing an error message.
83  */
84 #define EEH_MAX_FAILS	2100000
85 
86 /* Time to wait for a PCI slot to report status, in milliseconds */
87 #define PCI_BUS_RESET_WAIT_MSEC (60*1000)
88 
89 /* RTAS tokens */
90 static int ibm_set_eeh_option;
91 static int ibm_set_slot_reset;
92 static int ibm_read_slot_reset_state;
93 static int ibm_read_slot_reset_state2;
94 static int ibm_slot_error_detail;
95 static int ibm_get_config_addr_info;
96 static int ibm_get_config_addr_info2;
97 static int ibm_configure_bridge;
98 
99 int eeh_subsystem_enabled;
100 EXPORT_SYMBOL(eeh_subsystem_enabled);
101 
102 /* Lock to avoid races due to multiple reports of an error */
103 static DEFINE_SPINLOCK(confirm_error_lock);
104 
105 /* Buffer for reporting slot-error-detail rtas calls. Its here
106  * in BSS, and not dynamically alloced, so that it ends up in
107  * RMO where RTAS can access it.
108  */
109 static unsigned char slot_errbuf[RTAS_ERROR_LOG_MAX];
110 static DEFINE_SPINLOCK(slot_errbuf_lock);
111 static int eeh_error_buf_size;
112 
113 /* Buffer for reporting pci register dumps. Its here in BSS, and
114  * not dynamically alloced, so that it ends up in RMO where RTAS
115  * can access it.
116  */
117 #define EEH_PCI_REGS_LOG_LEN 4096
118 static unsigned char pci_regs_buf[EEH_PCI_REGS_LOG_LEN];
119 
120 /* System monitoring statistics */
121 static unsigned long no_device;
122 static unsigned long no_dn;
123 static unsigned long no_cfg_addr;
124 static unsigned long ignored_check;
125 static unsigned long total_mmio_ffs;
126 static unsigned long false_positives;
127 static unsigned long slot_resets;
128 
129 #define IS_BRIDGE(class_code) (((class_code)<<16) == PCI_BASE_CLASS_BRIDGE)
130 
131 /* --------------------------------------------------------------- */
132 /* Below lies the EEH event infrastructure */
133 
rtas_slot_error_detail(struct pci_dn * pdn,int severity,char * driver_log,size_t loglen)134 static void rtas_slot_error_detail(struct pci_dn *pdn, int severity,
135                                    char *driver_log, size_t loglen)
136 {
137 	int config_addr;
138 	unsigned long flags;
139 	int rc;
140 
141 	/* Log the error with the rtas logger */
142 	spin_lock_irqsave(&slot_errbuf_lock, flags);
143 	memset(slot_errbuf, 0, eeh_error_buf_size);
144 
145 	/* Use PE configuration address, if present */
146 	config_addr = pdn->eeh_config_addr;
147 	if (pdn->eeh_pe_config_addr)
148 		config_addr = pdn->eeh_pe_config_addr;
149 
150 	rc = rtas_call(ibm_slot_error_detail,
151 	               8, 1, NULL, config_addr,
152 	               BUID_HI(pdn->phb->buid),
153 	               BUID_LO(pdn->phb->buid),
154 	               virt_to_phys(driver_log), loglen,
155 	               virt_to_phys(slot_errbuf),
156 	               eeh_error_buf_size,
157 	               severity);
158 
159 	if (rc == 0)
160 		log_error(slot_errbuf, ERR_TYPE_RTAS_LOG, 0);
161 	spin_unlock_irqrestore(&slot_errbuf_lock, flags);
162 }
163 
164 /**
165  * gather_pci_data - copy assorted PCI config space registers to buff
166  * @pdn: device to report data for
167  * @buf: point to buffer in which to log
168  * @len: amount of room in buffer
169  *
170  * This routine captures assorted PCI configuration space data,
171  * and puts them into a buffer for RTAS error logging.
172  */
gather_pci_data(struct pci_dn * pdn,char * buf,size_t len)173 static size_t gather_pci_data(struct pci_dn *pdn, char * buf, size_t len)
174 {
175 	struct pci_dev *dev = pdn->pcidev;
176 	u32 cfg;
177 	int cap, i;
178 	int n = 0;
179 
180 	n += scnprintf(buf+n, len-n, "%s\n", pdn->node->full_name);
181 	printk(KERN_WARNING "EEH: of node=%s\n", pdn->node->full_name);
182 
183 	rtas_read_config(pdn, PCI_VENDOR_ID, 4, &cfg);
184 	n += scnprintf(buf+n, len-n, "dev/vend:%08x\n", cfg);
185 	printk(KERN_WARNING "EEH: PCI device/vendor: %08x\n", cfg);
186 
187 	rtas_read_config(pdn, PCI_COMMAND, 4, &cfg);
188 	n += scnprintf(buf+n, len-n, "cmd/stat:%x\n", cfg);
189 	printk(KERN_WARNING "EEH: PCI cmd/status register: %08x\n", cfg);
190 
191 	if (!dev) {
192 		printk(KERN_WARNING "EEH: no PCI device for this of node\n");
193 		return n;
194 	}
195 
196 	/* Gather bridge-specific registers */
197 	if (dev->class >> 16 == PCI_BASE_CLASS_BRIDGE) {
198 		rtas_read_config(pdn, PCI_SEC_STATUS, 2, &cfg);
199 		n += scnprintf(buf+n, len-n, "sec stat:%x\n", cfg);
200 		printk(KERN_WARNING "EEH: Bridge secondary status: %04x\n", cfg);
201 
202 		rtas_read_config(pdn, PCI_BRIDGE_CONTROL, 2, &cfg);
203 		n += scnprintf(buf+n, len-n, "brdg ctl:%x\n", cfg);
204 		printk(KERN_WARNING "EEH: Bridge control: %04x\n", cfg);
205 	}
206 
207 	/* Dump out the PCI-X command and status regs */
208 	cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
209 	if (cap) {
210 		rtas_read_config(pdn, cap, 4, &cfg);
211 		n += scnprintf(buf+n, len-n, "pcix-cmd:%x\n", cfg);
212 		printk(KERN_WARNING "EEH: PCI-X cmd: %08x\n", cfg);
213 
214 		rtas_read_config(pdn, cap+4, 4, &cfg);
215 		n += scnprintf(buf+n, len-n, "pcix-stat:%x\n", cfg);
216 		printk(KERN_WARNING "EEH: PCI-X status: %08x\n", cfg);
217 	}
218 
219 	/* If PCI-E capable, dump PCI-E cap 10, and the AER */
220 	cap = pci_find_capability(dev, PCI_CAP_ID_EXP);
221 	if (cap) {
222 		n += scnprintf(buf+n, len-n, "pci-e cap10:\n");
223 		printk(KERN_WARNING
224 		       "EEH: PCI-E capabilities and status follow:\n");
225 
226 		for (i=0; i<=8; i++) {
227 			rtas_read_config(pdn, cap+4*i, 4, &cfg);
228 			n += scnprintf(buf+n, len-n, "%02x:%x\n", 4*i, cfg);
229 			printk(KERN_WARNING "EEH: PCI-E %02x: %08x\n", i, cfg);
230 		}
231 
232 		cap = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
233 		if (cap) {
234 			n += scnprintf(buf+n, len-n, "pci-e AER:\n");
235 			printk(KERN_WARNING
236 			       "EEH: PCI-E AER capability register set follows:\n");
237 
238 			for (i=0; i<14; i++) {
239 				rtas_read_config(pdn, cap+4*i, 4, &cfg);
240 				n += scnprintf(buf+n, len-n, "%02x:%x\n", 4*i, cfg);
241 				printk(KERN_WARNING "EEH: PCI-E AER %02x: %08x\n", i, cfg);
242 			}
243 		}
244 	}
245 
246 	/* Gather status on devices under the bridge */
247 	if (dev->class >> 16 == PCI_BASE_CLASS_BRIDGE) {
248 		struct device_node *dn;
249 
250 		for_each_child_of_node(pdn->node, dn) {
251 			pdn = PCI_DN(dn);
252 			if (pdn)
253 				n += gather_pci_data(pdn, buf+n, len-n);
254 		}
255 	}
256 
257 	return n;
258 }
259 
eeh_slot_error_detail(struct pci_dn * pdn,int severity)260 void eeh_slot_error_detail(struct pci_dn *pdn, int severity)
261 {
262 	size_t loglen = 0;
263 	pci_regs_buf[0] = 0;
264 
265 	rtas_pci_enable(pdn, EEH_THAW_MMIO);
266 	loglen = gather_pci_data(pdn, pci_regs_buf, EEH_PCI_REGS_LOG_LEN);
267 
268 	rtas_slot_error_detail(pdn, severity, pci_regs_buf, loglen);
269 }
270 
271 /**
272  * read_slot_reset_state - Read the reset state of a device node's slot
273  * @dn: device node to read
274  * @rets: array to return results in
275  */
read_slot_reset_state(struct pci_dn * pdn,int rets[])276 static int read_slot_reset_state(struct pci_dn *pdn, int rets[])
277 {
278 	int token, outputs;
279 	int config_addr;
280 
281 	if (ibm_read_slot_reset_state2 != RTAS_UNKNOWN_SERVICE) {
282 		token = ibm_read_slot_reset_state2;
283 		outputs = 4;
284 	} else {
285 		token = ibm_read_slot_reset_state;
286 		rets[2] = 0; /* fake PE Unavailable info */
287 		outputs = 3;
288 	}
289 
290 	/* Use PE configuration address, if present */
291 	config_addr = pdn->eeh_config_addr;
292 	if (pdn->eeh_pe_config_addr)
293 		config_addr = pdn->eeh_pe_config_addr;
294 
295 	return rtas_call(token, 3, outputs, rets, config_addr,
296 			 BUID_HI(pdn->phb->buid), BUID_LO(pdn->phb->buid));
297 }
298 
299 /**
300  * eeh_wait_for_slot_status - returns error status of slot
301  * @pdn pci device node
302  * @max_wait_msecs maximum number to millisecs to wait
303  *
304  * Return negative value if a permanent error, else return
305  * Partition Endpoint (PE) status value.
306  *
307  * If @max_wait_msecs is positive, then this routine will
308  * sleep until a valid status can be obtained, or until
309  * the max allowed wait time is exceeded, in which case
310  * a -2 is returned.
311  */
312 int
eeh_wait_for_slot_status(struct pci_dn * pdn,int max_wait_msecs)313 eeh_wait_for_slot_status(struct pci_dn *pdn, int max_wait_msecs)
314 {
315 	int rc;
316 	int rets[3];
317 	int mwait;
318 
319 	while (1) {
320 		rc = read_slot_reset_state(pdn, rets);
321 		if (rc) return rc;
322 		if (rets[1] == 0) return -1;  /* EEH is not supported */
323 
324 		if (rets[0] != 5) return rets[0]; /* return actual status */
325 
326 		if (rets[2] == 0) return -1; /* permanently unavailable */
327 
328 		if (max_wait_msecs <= 0) break;
329 
330 		mwait = rets[2];
331 		if (mwait <= 0) {
332 			printk (KERN_WARNING
333 			        "EEH: Firmware returned bad wait value=%d\n", mwait);
334 			mwait = 1000;
335 		} else if (mwait > 300*1000) {
336 			printk (KERN_WARNING
337 			        "EEH: Firmware is taking too long, time=%d\n", mwait);
338 			mwait = 300*1000;
339 		}
340 		max_wait_msecs -= mwait;
341 		msleep (mwait);
342 	}
343 
344 	printk(KERN_WARNING "EEH: Timed out waiting for slot status\n");
345 	return -2;
346 }
347 
348 /**
349  * eeh_token_to_phys - convert EEH address token to phys address
350  * @token i/o token, should be address in the form 0xA....
351  */
eeh_token_to_phys(unsigned long token)352 static inline unsigned long eeh_token_to_phys(unsigned long token)
353 {
354 	pte_t *ptep;
355 	unsigned long pa;
356 
357 	ptep = find_linux_pte(init_mm.pgd, token);
358 	if (!ptep)
359 		return token;
360 	pa = pte_pfn(*ptep) << PAGE_SHIFT;
361 
362 	return pa | (token & (PAGE_SIZE-1));
363 }
364 
365 /**
366  * Return the "partitionable endpoint" (pe) under which this device lies
367  */
find_device_pe(struct device_node * dn)368 struct device_node * find_device_pe(struct device_node *dn)
369 {
370 	while ((dn->parent) && PCI_DN(dn->parent) &&
371 	      (PCI_DN(dn->parent)->eeh_mode & EEH_MODE_SUPPORTED)) {
372 		dn = dn->parent;
373 	}
374 	return dn;
375 }
376 
377 /** Mark all devices that are children of this device as failed.
378  *  Mark the device driver too, so that it can see the failure
379  *  immediately; this is critical, since some drivers poll
380  *  status registers in interrupts ... If a driver is polling,
381  *  and the slot is frozen, then the driver can deadlock in
382  *  an interrupt context, which is bad.
383  */
384 
__eeh_mark_slot(struct device_node * parent,int mode_flag)385 static void __eeh_mark_slot(struct device_node *parent, int mode_flag)
386 {
387 	struct device_node *dn;
388 
389 	for_each_child_of_node(parent, dn) {
390 		if (PCI_DN(dn)) {
391 			/* Mark the pci device driver too */
392 			struct pci_dev *dev = PCI_DN(dn)->pcidev;
393 
394 			PCI_DN(dn)->eeh_mode |= mode_flag;
395 
396 			if (dev && dev->driver)
397 				dev->error_state = pci_channel_io_frozen;
398 
399 			__eeh_mark_slot(dn, mode_flag);
400 		}
401 	}
402 }
403 
eeh_mark_slot(struct device_node * dn,int mode_flag)404 void eeh_mark_slot (struct device_node *dn, int mode_flag)
405 {
406 	struct pci_dev *dev;
407 	dn = find_device_pe (dn);
408 
409 	/* Back up one, since config addrs might be shared */
410 	if (!pcibios_find_pci_bus(dn) && PCI_DN(dn->parent))
411 		dn = dn->parent;
412 
413 	PCI_DN(dn)->eeh_mode |= mode_flag;
414 
415 	/* Mark the pci device too */
416 	dev = PCI_DN(dn)->pcidev;
417 	if (dev)
418 		dev->error_state = pci_channel_io_frozen;
419 
420 	__eeh_mark_slot(dn, mode_flag);
421 }
422 
__eeh_clear_slot(struct device_node * parent,int mode_flag)423 static void __eeh_clear_slot(struct device_node *parent, int mode_flag)
424 {
425 	struct device_node *dn;
426 
427 	for_each_child_of_node(parent, dn) {
428 		if (PCI_DN(dn)) {
429 			PCI_DN(dn)->eeh_mode &= ~mode_flag;
430 			PCI_DN(dn)->eeh_check_count = 0;
431 			__eeh_clear_slot(dn, mode_flag);
432 		}
433 	}
434 }
435 
eeh_clear_slot(struct device_node * dn,int mode_flag)436 void eeh_clear_slot (struct device_node *dn, int mode_flag)
437 {
438 	unsigned long flags;
439 	spin_lock_irqsave(&confirm_error_lock, flags);
440 
441 	dn = find_device_pe (dn);
442 
443 	/* Back up one, since config addrs might be shared */
444 	if (!pcibios_find_pci_bus(dn) && PCI_DN(dn->parent))
445 		dn = dn->parent;
446 
447 	PCI_DN(dn)->eeh_mode &= ~mode_flag;
448 	PCI_DN(dn)->eeh_check_count = 0;
449 	__eeh_clear_slot(dn, mode_flag);
450 	spin_unlock_irqrestore(&confirm_error_lock, flags);
451 }
452 
453 /**
454  * eeh_dn_check_failure - check if all 1's data is due to EEH slot freeze
455  * @dn device node
456  * @dev pci device, if known
457  *
458  * Check for an EEH failure for the given device node.  Call this
459  * routine if the result of a read was all 0xff's and you want to
460  * find out if this is due to an EEH slot freeze.  This routine
461  * will query firmware for the EEH status.
462  *
463  * Returns 0 if there has not been an EEH error; otherwise returns
464  * a non-zero value and queues up a slot isolation event notification.
465  *
466  * It is safe to call this routine in an interrupt context.
467  */
eeh_dn_check_failure(struct device_node * dn,struct pci_dev * dev)468 int eeh_dn_check_failure(struct device_node *dn, struct pci_dev *dev)
469 {
470 	int ret;
471 	int rets[3];
472 	unsigned long flags;
473 	struct pci_dn *pdn;
474 	int rc = 0;
475 	const char *location;
476 
477 	total_mmio_ffs++;
478 
479 	if (!eeh_subsystem_enabled)
480 		return 0;
481 
482 	if (!dn) {
483 		no_dn++;
484 		return 0;
485 	}
486 	dn = find_device_pe(dn);
487 	pdn = PCI_DN(dn);
488 
489 	/* Access to IO BARs might get this far and still not want checking. */
490 	if (!(pdn->eeh_mode & EEH_MODE_SUPPORTED) ||
491 	    pdn->eeh_mode & EEH_MODE_NOCHECK) {
492 		ignored_check++;
493 		pr_debug("EEH: Ignored check (%x) for %s %s\n",
494 			 pdn->eeh_mode, pci_name (dev), dn->full_name);
495 		return 0;
496 	}
497 
498 	if (!pdn->eeh_config_addr && !pdn->eeh_pe_config_addr) {
499 		no_cfg_addr++;
500 		return 0;
501 	}
502 
503 	/* If we already have a pending isolation event for this
504 	 * slot, we know it's bad already, we don't need to check.
505 	 * Do this checking under a lock; as multiple PCI devices
506 	 * in one slot might report errors simultaneously, and we
507 	 * only want one error recovery routine running.
508 	 */
509 	spin_lock_irqsave(&confirm_error_lock, flags);
510 	rc = 1;
511 	if (pdn->eeh_mode & EEH_MODE_ISOLATED) {
512 		pdn->eeh_check_count ++;
513 		if (pdn->eeh_check_count % EEH_MAX_FAILS == 0) {
514 			location = of_get_property(dn, "ibm,loc-code", NULL);
515 			printk (KERN_ERR "EEH: %d reads ignored for recovering device at "
516 				"location=%s driver=%s pci addr=%s\n",
517 				pdn->eeh_check_count, location,
518 				dev->driver->name, pci_name(dev));
519 			printk (KERN_ERR "EEH: Might be infinite loop in %s driver\n",
520 				dev->driver->name);
521 			dump_stack();
522 		}
523 		goto dn_unlock;
524 	}
525 
526 	/*
527 	 * Now test for an EEH failure.  This is VERY expensive.
528 	 * Note that the eeh_config_addr may be a parent device
529 	 * in the case of a device behind a bridge, or it may be
530 	 * function zero of a multi-function device.
531 	 * In any case they must share a common PHB.
532 	 */
533 	ret = read_slot_reset_state(pdn, rets);
534 
535 	/* If the call to firmware failed, punt */
536 	if (ret != 0) {
537 		printk(KERN_WARNING "EEH: read_slot_reset_state() failed; rc=%d dn=%s\n",
538 		       ret, dn->full_name);
539 		false_positives++;
540 		pdn->eeh_false_positives ++;
541 		rc = 0;
542 		goto dn_unlock;
543 	}
544 
545 	/* Note that config-io to empty slots may fail;
546 	 * they are empty when they don't have children. */
547 	if ((rets[0] == 5) && (rets[2] == 0) && (dn->child == NULL)) {
548 		false_positives++;
549 		pdn->eeh_false_positives ++;
550 		rc = 0;
551 		goto dn_unlock;
552 	}
553 
554 	/* If EEH is not supported on this device, punt. */
555 	if (rets[1] != 1) {
556 		printk(KERN_WARNING "EEH: event on unsupported device, rc=%d dn=%s\n",
557 		       ret, dn->full_name);
558 		false_positives++;
559 		pdn->eeh_false_positives ++;
560 		rc = 0;
561 		goto dn_unlock;
562 	}
563 
564 	/* If not the kind of error we know about, punt. */
565 	if (rets[0] != 1 && rets[0] != 2 && rets[0] != 4 && rets[0] != 5) {
566 		false_positives++;
567 		pdn->eeh_false_positives ++;
568 		rc = 0;
569 		goto dn_unlock;
570 	}
571 
572 	slot_resets++;
573 
574 	/* Avoid repeated reports of this failure, including problems
575 	 * with other functions on this device, and functions under
576 	 * bridges. */
577 	eeh_mark_slot (dn, EEH_MODE_ISOLATED);
578 	spin_unlock_irqrestore(&confirm_error_lock, flags);
579 
580 	eeh_send_failure_event (dn, dev);
581 
582 	/* Most EEH events are due to device driver bugs.  Having
583 	 * a stack trace will help the device-driver authors figure
584 	 * out what happened.  So print that out. */
585 	dump_stack();
586 	return 1;
587 
588 dn_unlock:
589 	spin_unlock_irqrestore(&confirm_error_lock, flags);
590 	return rc;
591 }
592 
593 EXPORT_SYMBOL_GPL(eeh_dn_check_failure);
594 
595 /**
596  * eeh_check_failure - check if all 1's data is due to EEH slot freeze
597  * @token i/o token, should be address in the form 0xA....
598  * @val value, should be all 1's (XXX why do we need this arg??)
599  *
600  * Check for an EEH failure at the given token address.  Call this
601  * routine if the result of a read was all 0xff's and you want to
602  * find out if this is due to an EEH slot freeze event.  This routine
603  * will query firmware for the EEH status.
604  *
605  * Note this routine is safe to call in an interrupt context.
606  */
eeh_check_failure(const volatile void __iomem * token,unsigned long val)607 unsigned long eeh_check_failure(const volatile void __iomem *token, unsigned long val)
608 {
609 	unsigned long addr;
610 	struct pci_dev *dev;
611 	struct device_node *dn;
612 
613 	/* Finding the phys addr + pci device; this is pretty quick. */
614 	addr = eeh_token_to_phys((unsigned long __force) token);
615 	dev = pci_get_device_by_addr(addr);
616 	if (!dev) {
617 		no_device++;
618 		return val;
619 	}
620 
621 	dn = pci_device_to_OF_node(dev);
622 	eeh_dn_check_failure (dn, dev);
623 
624 	pci_dev_put(dev);
625 	return val;
626 }
627 
628 EXPORT_SYMBOL(eeh_check_failure);
629 
630 /* ------------------------------------------------------------- */
631 /* The code below deals with error recovery */
632 
633 /**
634  * rtas_pci_enable - enable MMIO or DMA transfers for this slot
635  * @pdn pci device node
636  */
637 
638 int
rtas_pci_enable(struct pci_dn * pdn,int function)639 rtas_pci_enable(struct pci_dn *pdn, int function)
640 {
641 	int config_addr;
642 	int rc;
643 
644 	/* Use PE configuration address, if present */
645 	config_addr = pdn->eeh_config_addr;
646 	if (pdn->eeh_pe_config_addr)
647 		config_addr = pdn->eeh_pe_config_addr;
648 
649 	rc = rtas_call(ibm_set_eeh_option, 4, 1, NULL,
650 	               config_addr,
651 	               BUID_HI(pdn->phb->buid),
652 	               BUID_LO(pdn->phb->buid),
653 		            function);
654 
655 	if (rc)
656 		printk(KERN_WARNING "EEH: Unexpected state change %d, err=%d dn=%s\n",
657 		        function, rc, pdn->node->full_name);
658 
659 	rc = eeh_wait_for_slot_status (pdn, PCI_BUS_RESET_WAIT_MSEC);
660 	if ((rc == 4) && (function == EEH_THAW_MMIO))
661 		return 0;
662 
663 	return rc;
664 }
665 
666 /**
667  * rtas_pci_slot_reset - raises/lowers the pci #RST line
668  * @pdn pci device node
669  * @state: 1/0 to raise/lower the #RST
670  *
671  * Clear the EEH-frozen condition on a slot.  This routine
672  * asserts the PCI #RST line if the 'state' argument is '1',
673  * and drops the #RST line if 'state is '0'.  This routine is
674  * safe to call in an interrupt context.
675  *
676  */
677 
678 static void
rtas_pci_slot_reset(struct pci_dn * pdn,int state)679 rtas_pci_slot_reset(struct pci_dn *pdn, int state)
680 {
681 	int config_addr;
682 	int rc;
683 
684 	BUG_ON (pdn==NULL);
685 
686 	if (!pdn->phb) {
687 		printk (KERN_WARNING "EEH: in slot reset, device node %s has no phb\n",
688 		        pdn->node->full_name);
689 		return;
690 	}
691 
692 	/* Use PE configuration address, if present */
693 	config_addr = pdn->eeh_config_addr;
694 	if (pdn->eeh_pe_config_addr)
695 		config_addr = pdn->eeh_pe_config_addr;
696 
697 	rc = rtas_call(ibm_set_slot_reset,4,1, NULL,
698 	               config_addr,
699 	               BUID_HI(pdn->phb->buid),
700 	               BUID_LO(pdn->phb->buid),
701 	               state);
702 	if (rc)
703 		printk (KERN_WARNING "EEH: Unable to reset the failed slot,"
704 		        " (%d) #RST=%d dn=%s\n",
705 		        rc, state, pdn->node->full_name);
706 }
707 
708 /**
709  * pcibios_set_pcie_slot_reset - Set PCI-E reset state
710  * @dev:	pci device struct
711  * @state:	reset state to enter
712  *
713  * Return value:
714  * 	0 if success
715  **/
pcibios_set_pcie_reset_state(struct pci_dev * dev,enum pcie_reset_state state)716 int pcibios_set_pcie_reset_state(struct pci_dev *dev, enum pcie_reset_state state)
717 {
718 	struct device_node *dn = pci_device_to_OF_node(dev);
719 	struct pci_dn *pdn = PCI_DN(dn);
720 
721 	switch (state) {
722 	case pcie_deassert_reset:
723 		rtas_pci_slot_reset(pdn, 0);
724 		break;
725 	case pcie_hot_reset:
726 		rtas_pci_slot_reset(pdn, 1);
727 		break;
728 	case pcie_warm_reset:
729 		rtas_pci_slot_reset(pdn, 3);
730 		break;
731 	default:
732 		return -EINVAL;
733 	};
734 
735 	return 0;
736 }
737 
738 /**
739  * rtas_set_slot_reset -- assert the pci #RST line for 1/4 second
740  * @pdn: pci device node to be reset.
741  *
742  *  Return 0 if success, else a non-zero value.
743  */
744 
__rtas_set_slot_reset(struct pci_dn * pdn)745 static void __rtas_set_slot_reset(struct pci_dn *pdn)
746 {
747 	rtas_pci_slot_reset (pdn, 1);
748 
749 	/* The PCI bus requires that the reset be held high for at least
750 	 * a 100 milliseconds. We wait a bit longer 'just in case'.  */
751 
752 #define PCI_BUS_RST_HOLD_TIME_MSEC 250
753 	msleep (PCI_BUS_RST_HOLD_TIME_MSEC);
754 
755 	/* We might get hit with another EEH freeze as soon as the
756 	 * pci slot reset line is dropped. Make sure we don't miss
757 	 * these, and clear the flag now. */
758 	eeh_clear_slot (pdn->node, EEH_MODE_ISOLATED);
759 
760 	rtas_pci_slot_reset (pdn, 0);
761 
762 	/* After a PCI slot has been reset, the PCI Express spec requires
763 	 * a 1.5 second idle time for the bus to stabilize, before starting
764 	 * up traffic. */
765 #define PCI_BUS_SETTLE_TIME_MSEC 1800
766 	msleep (PCI_BUS_SETTLE_TIME_MSEC);
767 }
768 
rtas_set_slot_reset(struct pci_dn * pdn)769 int rtas_set_slot_reset(struct pci_dn *pdn)
770 {
771 	int i, rc;
772 
773 	/* Take three shots at resetting the bus */
774 	for (i=0; i<3; i++) {
775 		__rtas_set_slot_reset(pdn);
776 
777 		rc = eeh_wait_for_slot_status(pdn, PCI_BUS_RESET_WAIT_MSEC);
778 		if (rc == 0)
779 			return 0;
780 
781 		if (rc < 0) {
782 			printk(KERN_ERR "EEH: unrecoverable slot failure %s\n",
783 			       pdn->node->full_name);
784 			return -1;
785 		}
786 		printk(KERN_ERR "EEH: bus reset %d failed on slot %s, rc=%d\n",
787 		       i+1, pdn->node->full_name, rc);
788 	}
789 
790 	return -1;
791 }
792 
793 /* ------------------------------------------------------- */
794 /** Save and restore of PCI BARs
795  *
796  * Although firmware will set up BARs during boot, it doesn't
797  * set up device BAR's after a device reset, although it will,
798  * if requested, set up bridge configuration. Thus, we need to
799  * configure the PCI devices ourselves.
800  */
801 
802 /**
803  * __restore_bars - Restore the Base Address Registers
804  * @pdn: pci device node
805  *
806  * Loads the PCI configuration space base address registers,
807  * the expansion ROM base address, the latency timer, and etc.
808  * from the saved values in the device node.
809  */
__restore_bars(struct pci_dn * pdn)810 static inline void __restore_bars (struct pci_dn *pdn)
811 {
812 	int i;
813 	u32 cmd;
814 
815 	if (NULL==pdn->phb) return;
816 	for (i=4; i<10; i++) {
817 		rtas_write_config(pdn, i*4, 4, pdn->config_space[i]);
818 	}
819 
820 	/* 12 == Expansion ROM Address */
821 	rtas_write_config(pdn, 12*4, 4, pdn->config_space[12]);
822 
823 #define BYTE_SWAP(OFF) (8*((OFF)/4)+3-(OFF))
824 #define SAVED_BYTE(OFF) (((u8 *)(pdn->config_space))[BYTE_SWAP(OFF)])
825 
826 	rtas_write_config (pdn, PCI_CACHE_LINE_SIZE, 1,
827 	            SAVED_BYTE(PCI_CACHE_LINE_SIZE));
828 
829 	rtas_write_config (pdn, PCI_LATENCY_TIMER, 1,
830 	            SAVED_BYTE(PCI_LATENCY_TIMER));
831 
832 	/* max latency, min grant, interrupt pin and line */
833 	rtas_write_config(pdn, 15*4, 4, pdn->config_space[15]);
834 
835 	/* Restore PERR & SERR bits, some devices require it,
836 	   don't touch the other command bits */
837 	rtas_read_config(pdn, PCI_COMMAND, 4, &cmd);
838 	if (pdn->config_space[1] & PCI_COMMAND_PARITY)
839 		cmd |= PCI_COMMAND_PARITY;
840 	else
841 		cmd &= ~PCI_COMMAND_PARITY;
842 	if (pdn->config_space[1] & PCI_COMMAND_SERR)
843 		cmd |= PCI_COMMAND_SERR;
844 	else
845 		cmd &= ~PCI_COMMAND_SERR;
846 	rtas_write_config(pdn, PCI_COMMAND, 4, cmd);
847 }
848 
849 /**
850  * eeh_restore_bars - restore the PCI config space info
851  *
852  * This routine performs a recursive walk to the children
853  * of this device as well.
854  */
eeh_restore_bars(struct pci_dn * pdn)855 void eeh_restore_bars(struct pci_dn *pdn)
856 {
857 	struct device_node *dn;
858 	if (!pdn)
859 		return;
860 
861 	if ((pdn->eeh_mode & EEH_MODE_SUPPORTED) && !IS_BRIDGE(pdn->class_code))
862 		__restore_bars (pdn);
863 
864 	for_each_child_of_node(pdn->node, dn)
865 		eeh_restore_bars (PCI_DN(dn));
866 }
867 
868 /**
869  * eeh_save_bars - save device bars
870  *
871  * Save the values of the device bars. Unlike the restore
872  * routine, this routine is *not* recursive. This is because
873  * PCI devices are added individuallly; but, for the restore,
874  * an entire slot is reset at a time.
875  */
eeh_save_bars(struct pci_dn * pdn)876 static void eeh_save_bars(struct pci_dn *pdn)
877 {
878 	int i;
879 
880 	if (!pdn )
881 		return;
882 
883 	for (i = 0; i < 16; i++)
884 		rtas_read_config(pdn, i * 4, 4, &pdn->config_space[i]);
885 }
886 
887 void
rtas_configure_bridge(struct pci_dn * pdn)888 rtas_configure_bridge(struct pci_dn *pdn)
889 {
890 	int config_addr;
891 	int rc;
892 
893 	/* Use PE configuration address, if present */
894 	config_addr = pdn->eeh_config_addr;
895 	if (pdn->eeh_pe_config_addr)
896 		config_addr = pdn->eeh_pe_config_addr;
897 
898 	rc = rtas_call(ibm_configure_bridge,3,1, NULL,
899 	               config_addr,
900 	               BUID_HI(pdn->phb->buid),
901 	               BUID_LO(pdn->phb->buid));
902 	if (rc) {
903 		printk (KERN_WARNING "EEH: Unable to configure device bridge (%d) for %s\n",
904 		        rc, pdn->node->full_name);
905 	}
906 }
907 
908 /* ------------------------------------------------------------- */
909 /* The code below deals with enabling EEH for devices during  the
910  * early boot sequence.  EEH must be enabled before any PCI probing
911  * can be done.
912  */
913 
914 #define EEH_ENABLE 1
915 
916 struct eeh_early_enable_info {
917 	unsigned int buid_hi;
918 	unsigned int buid_lo;
919 };
920 
get_pe_addr(int config_addr,struct eeh_early_enable_info * info)921 static int get_pe_addr (int config_addr,
922                         struct eeh_early_enable_info *info)
923 {
924 	unsigned int rets[3];
925 	int ret;
926 
927 	/* Use latest config-addr token on power6 */
928 	if (ibm_get_config_addr_info2 != RTAS_UNKNOWN_SERVICE) {
929 		/* Make sure we have a PE in hand */
930 		ret = rtas_call (ibm_get_config_addr_info2, 4, 2, rets,
931 			config_addr, info->buid_hi, info->buid_lo, 1);
932 		if (ret || (rets[0]==0))
933 			return 0;
934 
935 		ret = rtas_call (ibm_get_config_addr_info2, 4, 2, rets,
936 			config_addr, info->buid_hi, info->buid_lo, 0);
937 		if (ret)
938 			return 0;
939 		return rets[0];
940 	}
941 
942 	/* Use older config-addr token on power5 */
943 	if (ibm_get_config_addr_info != RTAS_UNKNOWN_SERVICE) {
944 		ret = rtas_call (ibm_get_config_addr_info, 4, 2, rets,
945 			config_addr, info->buid_hi, info->buid_lo, 0);
946 		if (ret)
947 			return 0;
948 		return rets[0];
949 	}
950 	return 0;
951 }
952 
953 /* Enable eeh for the given device node. */
early_enable_eeh(struct device_node * dn,void * data)954 static void *early_enable_eeh(struct device_node *dn, void *data)
955 {
956 	unsigned int rets[3];
957 	struct eeh_early_enable_info *info = data;
958 	int ret;
959 	const u32 *class_code = of_get_property(dn, "class-code", NULL);
960 	const u32 *vendor_id = of_get_property(dn, "vendor-id", NULL);
961 	const u32 *device_id = of_get_property(dn, "device-id", NULL);
962 	const u32 *regs;
963 	int enable;
964 	struct pci_dn *pdn = PCI_DN(dn);
965 
966 	pdn->class_code = 0;
967 	pdn->eeh_mode = 0;
968 	pdn->eeh_check_count = 0;
969 	pdn->eeh_freeze_count = 0;
970 	pdn->eeh_false_positives = 0;
971 
972 	if (!of_device_is_available(dn))
973 		return NULL;
974 
975 	/* Ignore bad nodes. */
976 	if (!class_code || !vendor_id || !device_id)
977 		return NULL;
978 
979 	/* There is nothing to check on PCI to ISA bridges */
980 	if (dn->type && !strcmp(dn->type, "isa")) {
981 		pdn->eeh_mode |= EEH_MODE_NOCHECK;
982 		return NULL;
983 	}
984 	pdn->class_code = *class_code;
985 
986 	/* Ok... see if this device supports EEH.  Some do, some don't,
987 	 * and the only way to find out is to check each and every one. */
988 	regs = of_get_property(dn, "reg", NULL);
989 	if (regs) {
990 		/* First register entry is addr (00BBSS00)  */
991 		/* Try to enable eeh */
992 		ret = rtas_call(ibm_set_eeh_option, 4, 1, NULL,
993 		                regs[0], info->buid_hi, info->buid_lo,
994 		                EEH_ENABLE);
995 
996 		enable = 0;
997 		if (ret == 0) {
998 			pdn->eeh_config_addr = regs[0];
999 
1000 			/* If the newer, better, ibm,get-config-addr-info is supported,
1001 			 * then use that instead. */
1002 			pdn->eeh_pe_config_addr = get_pe_addr(pdn->eeh_config_addr, info);
1003 
1004 			/* Some older systems (Power4) allow the
1005 			 * ibm,set-eeh-option call to succeed even on nodes
1006 			 * where EEH is not supported. Verify support
1007 			 * explicitly. */
1008 			ret = read_slot_reset_state(pdn, rets);
1009 			if ((ret == 0) && (rets[1] == 1))
1010 				enable = 1;
1011 		}
1012 
1013 		if (enable) {
1014 			eeh_subsystem_enabled = 1;
1015 			pdn->eeh_mode |= EEH_MODE_SUPPORTED;
1016 
1017 			pr_debug("EEH: %s: eeh enabled, config=%x pe_config=%x\n",
1018 				 dn->full_name, pdn->eeh_config_addr,
1019 				 pdn->eeh_pe_config_addr);
1020 		} else {
1021 
1022 			/* This device doesn't support EEH, but it may have an
1023 			 * EEH parent, in which case we mark it as supported. */
1024 			if (dn->parent && PCI_DN(dn->parent)
1025 			    && (PCI_DN(dn->parent)->eeh_mode & EEH_MODE_SUPPORTED)) {
1026 				/* Parent supports EEH. */
1027 				pdn->eeh_mode |= EEH_MODE_SUPPORTED;
1028 				pdn->eeh_config_addr = PCI_DN(dn->parent)->eeh_config_addr;
1029 				return NULL;
1030 			}
1031 		}
1032 	} else {
1033 		printk(KERN_WARNING "EEH: %s: unable to get reg property.\n",
1034 		       dn->full_name);
1035 	}
1036 
1037 	eeh_save_bars(pdn);
1038 	return NULL;
1039 }
1040 
1041 /*
1042  * Initialize EEH by trying to enable it for all of the adapters in the system.
1043  * As a side effect we can determine here if eeh is supported at all.
1044  * Note that we leave EEH on so failed config cycles won't cause a machine
1045  * check.  If a user turns off EEH for a particular adapter they are really
1046  * telling Linux to ignore errors.  Some hardware (e.g. POWER5) won't
1047  * grant access to a slot if EEH isn't enabled, and so we always enable
1048  * EEH for all slots/all devices.
1049  *
1050  * The eeh-force-off option disables EEH checking globally, for all slots.
1051  * Even if force-off is set, the EEH hardware is still enabled, so that
1052  * newer systems can boot.
1053  */
eeh_init(void)1054 void __init eeh_init(void)
1055 {
1056 	struct device_node *phb, *np;
1057 	struct eeh_early_enable_info info;
1058 
1059 	spin_lock_init(&confirm_error_lock);
1060 	spin_lock_init(&slot_errbuf_lock);
1061 
1062 	np = of_find_node_by_path("/rtas");
1063 	if (np == NULL)
1064 		return;
1065 
1066 	ibm_set_eeh_option = rtas_token("ibm,set-eeh-option");
1067 	ibm_set_slot_reset = rtas_token("ibm,set-slot-reset");
1068 	ibm_read_slot_reset_state2 = rtas_token("ibm,read-slot-reset-state2");
1069 	ibm_read_slot_reset_state = rtas_token("ibm,read-slot-reset-state");
1070 	ibm_slot_error_detail = rtas_token("ibm,slot-error-detail");
1071 	ibm_get_config_addr_info = rtas_token("ibm,get-config-addr-info");
1072 	ibm_get_config_addr_info2 = rtas_token("ibm,get-config-addr-info2");
1073 	ibm_configure_bridge = rtas_token ("ibm,configure-bridge");
1074 
1075 	if (ibm_set_eeh_option == RTAS_UNKNOWN_SERVICE)
1076 		return;
1077 
1078 	eeh_error_buf_size = rtas_token("rtas-error-log-max");
1079 	if (eeh_error_buf_size == RTAS_UNKNOWN_SERVICE) {
1080 		eeh_error_buf_size = 1024;
1081 	}
1082 	if (eeh_error_buf_size > RTAS_ERROR_LOG_MAX) {
1083 		printk(KERN_WARNING "EEH: rtas-error-log-max is bigger than allocated "
1084 		      "buffer ! (%d vs %d)", eeh_error_buf_size, RTAS_ERROR_LOG_MAX);
1085 		eeh_error_buf_size = RTAS_ERROR_LOG_MAX;
1086 	}
1087 
1088 	/* Enable EEH for all adapters.  Note that eeh requires buid's */
1089 	for (phb = of_find_node_by_name(NULL, "pci"); phb;
1090 	     phb = of_find_node_by_name(phb, "pci")) {
1091 		unsigned long buid;
1092 
1093 		buid = get_phb_buid(phb);
1094 		if (buid == 0 || PCI_DN(phb) == NULL)
1095 			continue;
1096 
1097 		info.buid_lo = BUID_LO(buid);
1098 		info.buid_hi = BUID_HI(buid);
1099 		traverse_pci_devices(phb, early_enable_eeh, &info);
1100 	}
1101 
1102 	if (eeh_subsystem_enabled)
1103 		printk(KERN_INFO "EEH: PCI Enhanced I/O Error Handling Enabled\n");
1104 	else
1105 		printk(KERN_WARNING "EEH: No capable adapters found\n");
1106 }
1107 
1108 /**
1109  * eeh_add_device_early - enable EEH for the indicated device_node
1110  * @dn: device node for which to set up EEH
1111  *
1112  * This routine must be used to perform EEH initialization for PCI
1113  * devices that were added after system boot (e.g. hotplug, dlpar).
1114  * This routine must be called before any i/o is performed to the
1115  * adapter (inluding any config-space i/o).
1116  * Whether this actually enables EEH or not for this device depends
1117  * on the CEC architecture, type of the device, on earlier boot
1118  * command-line arguments & etc.
1119  */
eeh_add_device_early(struct device_node * dn)1120 static void eeh_add_device_early(struct device_node *dn)
1121 {
1122 	struct pci_controller *phb;
1123 	struct eeh_early_enable_info info;
1124 
1125 	if (!dn || !PCI_DN(dn))
1126 		return;
1127 	phb = PCI_DN(dn)->phb;
1128 
1129 	/* USB Bus children of PCI devices will not have BUID's */
1130 	if (NULL == phb || 0 == phb->buid)
1131 		return;
1132 
1133 	info.buid_hi = BUID_HI(phb->buid);
1134 	info.buid_lo = BUID_LO(phb->buid);
1135 	early_enable_eeh(dn, &info);
1136 }
1137 
eeh_add_device_tree_early(struct device_node * dn)1138 void eeh_add_device_tree_early(struct device_node *dn)
1139 {
1140 	struct device_node *sib;
1141 
1142 	for_each_child_of_node(dn, sib)
1143 		eeh_add_device_tree_early(sib);
1144 	eeh_add_device_early(dn);
1145 }
1146 EXPORT_SYMBOL_GPL(eeh_add_device_tree_early);
1147 
1148 /**
1149  * eeh_add_device_late - perform EEH initialization for the indicated pci device
1150  * @dev: pci device for which to set up EEH
1151  *
1152  * This routine must be used to complete EEH initialization for PCI
1153  * devices that were added after system boot (e.g. hotplug, dlpar).
1154  */
eeh_add_device_late(struct pci_dev * dev)1155 static void eeh_add_device_late(struct pci_dev *dev)
1156 {
1157 	struct device_node *dn;
1158 	struct pci_dn *pdn;
1159 
1160 	if (!dev || !eeh_subsystem_enabled)
1161 		return;
1162 
1163 	pr_debug("EEH: Adding device %s\n", pci_name(dev));
1164 
1165 	dn = pci_device_to_OF_node(dev);
1166 	pdn = PCI_DN(dn);
1167 	if (pdn->pcidev == dev) {
1168 		pr_debug("EEH: Already referenced !\n");
1169 		return;
1170 	}
1171 	WARN_ON(pdn->pcidev);
1172 
1173 	pci_dev_get (dev);
1174 	pdn->pcidev = dev;
1175 
1176 	pci_addr_cache_insert_device(dev);
1177 	eeh_sysfs_add_device(dev);
1178 }
1179 
eeh_add_device_tree_late(struct pci_bus * bus)1180 void eeh_add_device_tree_late(struct pci_bus *bus)
1181 {
1182 	struct pci_dev *dev;
1183 
1184 	list_for_each_entry(dev, &bus->devices, bus_list) {
1185  		eeh_add_device_late(dev);
1186  		if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
1187  			struct pci_bus *subbus = dev->subordinate;
1188  			if (subbus)
1189  				eeh_add_device_tree_late(subbus);
1190  		}
1191 	}
1192 }
1193 EXPORT_SYMBOL_GPL(eeh_add_device_tree_late);
1194 
1195 /**
1196  * eeh_remove_device - undo EEH setup for the indicated pci device
1197  * @dev: pci device to be removed
1198  *
1199  * This routine should be called when a device is removed from
1200  * a running system (e.g. by hotplug or dlpar).  It unregisters
1201  * the PCI device from the EEH subsystem.  I/O errors affecting
1202  * this device will no longer be detected after this call; thus,
1203  * i/o errors affecting this slot may leave this device unusable.
1204  */
eeh_remove_device(struct pci_dev * dev)1205 static void eeh_remove_device(struct pci_dev *dev)
1206 {
1207 	struct device_node *dn;
1208 	if (!dev || !eeh_subsystem_enabled)
1209 		return;
1210 
1211 	/* Unregister the device with the EEH/PCI address search system */
1212 	pr_debug("EEH: Removing device %s\n", pci_name(dev));
1213 
1214 	dn = pci_device_to_OF_node(dev);
1215 	if (PCI_DN(dn)->pcidev == NULL) {
1216 		pr_debug("EEH: Not referenced !\n");
1217 		return;
1218 	}
1219 	PCI_DN(dn)->pcidev = NULL;
1220 	pci_dev_put (dev);
1221 
1222 	pci_addr_cache_remove_device(dev);
1223 	eeh_sysfs_remove_device(dev);
1224 }
1225 
eeh_remove_bus_device(struct pci_dev * dev)1226 void eeh_remove_bus_device(struct pci_dev *dev)
1227 {
1228 	struct pci_bus *bus = dev->subordinate;
1229 	struct pci_dev *child, *tmp;
1230 
1231 	eeh_remove_device(dev);
1232 
1233 	if (bus && dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
1234 		list_for_each_entry_safe(child, tmp, &bus->devices, bus_list)
1235 			 eeh_remove_bus_device(child);
1236 	}
1237 }
1238 EXPORT_SYMBOL_GPL(eeh_remove_bus_device);
1239 
proc_eeh_show(struct seq_file * m,void * v)1240 static int proc_eeh_show(struct seq_file *m, void *v)
1241 {
1242 	if (0 == eeh_subsystem_enabled) {
1243 		seq_printf(m, "EEH Subsystem is globally disabled\n");
1244 		seq_printf(m, "eeh_total_mmio_ffs=%ld\n", total_mmio_ffs);
1245 	} else {
1246 		seq_printf(m, "EEH Subsystem is enabled\n");
1247 		seq_printf(m,
1248 				"no device=%ld\n"
1249 				"no device node=%ld\n"
1250 				"no config address=%ld\n"
1251 				"check not wanted=%ld\n"
1252 				"eeh_total_mmio_ffs=%ld\n"
1253 				"eeh_false_positives=%ld\n"
1254 				"eeh_slot_resets=%ld\n",
1255 				no_device, no_dn, no_cfg_addr,
1256 				ignored_check, total_mmio_ffs,
1257 				false_positives,
1258 				slot_resets);
1259 	}
1260 
1261 	return 0;
1262 }
1263 
proc_eeh_open(struct inode * inode,struct file * file)1264 static int proc_eeh_open(struct inode *inode, struct file *file)
1265 {
1266 	return single_open(file, proc_eeh_show, NULL);
1267 }
1268 
1269 static const struct file_operations proc_eeh_operations = {
1270 	.open      = proc_eeh_open,
1271 	.read      = seq_read,
1272 	.llseek    = seq_lseek,
1273 	.release   = single_release,
1274 };
1275 
eeh_init_proc(void)1276 static int __init eeh_init_proc(void)
1277 {
1278 	if (machine_is(pseries))
1279 		proc_create("ppc64/eeh", 0, NULL, &proc_eeh_operations);
1280 	return 0;
1281 }
1282 __initcall(eeh_init_proc);
1283