Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
3c8c90ab LV |
2 | * Copyright IBM Corporation 2001, 2005, 2006 |
3 | * Copyright Dave Engebretsen & Todd Inglett 2001 | |
4 | * Copyright Linas Vepstas 2005, 2006 | |
cb3bc9d0 | 5 | * Copyright 2001-2012 IBM Corporation. |
69376502 | 6 | * |
1da177e4 LT |
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. | |
69376502 | 11 | * |
1da177e4 LT |
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. | |
69376502 | 16 | * |
1da177e4 LT |
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 | |
3c8c90ab LV |
20 | * |
21 | * Please address comments and feedback to Linas Vepstas <linas@austin.ibm.com> | |
1da177e4 LT |
22 | */ |
23 | ||
6dee3fb9 | 24 | #include <linux/delay.h> |
cb3bc9d0 | 25 | #include <linux/sched.h> |
1da177e4 LT |
26 | #include <linux/init.h> |
27 | #include <linux/list.h> | |
1da177e4 LT |
28 | #include <linux/pci.h> |
29 | #include <linux/proc_fs.h> | |
30 | #include <linux/rbtree.h> | |
31 | #include <linux/seq_file.h> | |
32 | #include <linux/spinlock.h> | |
66b15db6 | 33 | #include <linux/export.h> |
acaa6176 SR |
34 | #include <linux/of.h> |
35 | ||
60063497 | 36 | #include <linux/atomic.h> |
1da177e4 | 37 | #include <asm/eeh.h> |
172ca926 | 38 | #include <asm/eeh_event.h> |
1da177e4 LT |
39 | #include <asm/io.h> |
40 | #include <asm/machdep.h> | |
172ca926 | 41 | #include <asm/ppc-pci.h> |
1da177e4 | 42 | #include <asm/rtas.h> |
1da177e4 | 43 | |
1da177e4 LT |
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), | |
69376502 LV |
57 | * which in turn might be caused by low voltage on the bus, dust, |
58 | * vibration, humidity, radioactivity or plain-old failed hardware. | |
1da177e4 LT |
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 | |
25985edc | 70 | * event has occurred (e.g. by reading 0xff's), will then ask EEH |
1da177e4 LT |
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 | ||
5c1344e9 | 79 | /* If a device driver keeps reading an MMIO register in an interrupt |
f36c5227 MM |
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. | |
1da177e4 | 83 | */ |
2fd30be8 | 84 | #define EEH_MAX_FAILS 2100000 |
1da177e4 | 85 | |
17213c3b | 86 | /* Time to wait for a PCI slot to report status, in milliseconds */ |
9c547768 LV |
87 | #define PCI_BUS_RESET_WAIT_MSEC (60*1000) |
88 | ||
aa1e6374 GS |
89 | /* Platform dependent EEH operations */ |
90 | struct eeh_ops *eeh_ops = NULL; | |
91 | ||
1e28a7dd DW |
92 | int eeh_subsystem_enabled; |
93 | EXPORT_SYMBOL(eeh_subsystem_enabled); | |
1da177e4 | 94 | |
d7bb8862 GS |
95 | /* |
96 | * EEH probe mode support. The intention is to support multiple | |
97 | * platforms for EEH. Some platforms like pSeries do PCI emunation | |
98 | * based on device tree. However, other platforms like powernv probe | |
99 | * PCI devices from hardware. The flag is used to distinguish that. | |
100 | * In addition, struct eeh_ops::probe would be invoked for particular | |
101 | * OF node or PCI device so that the corresponding PE would be created | |
102 | * there. | |
103 | */ | |
104 | int eeh_probe_mode; | |
105 | ||
646a8499 GS |
106 | /* Global EEH mutex */ |
107 | DEFINE_MUTEX(eeh_mutex); | |
108 | ||
fd761fd8 | 109 | /* Lock to avoid races due to multiple reports of an error */ |
4907581d | 110 | DEFINE_RAW_SPINLOCK(confirm_error_lock); |
fd761fd8 | 111 | |
17213c3b LV |
112 | /* Buffer for reporting pci register dumps. Its here in BSS, and |
113 | * not dynamically alloced, so that it ends up in RMO where RTAS | |
114 | * can access it. | |
115 | */ | |
d99bb1db LV |
116 | #define EEH_PCI_REGS_LOG_LEN 4096 |
117 | static unsigned char pci_regs_buf[EEH_PCI_REGS_LOG_LEN]; | |
118 | ||
e575f8db GS |
119 | /* |
120 | * The struct is used to maintain the EEH global statistic | |
121 | * information. Besides, the EEH global statistics will be | |
122 | * exported to user space through procfs | |
123 | */ | |
124 | struct eeh_stats { | |
125 | u64 no_device; /* PCI device not found */ | |
126 | u64 no_dn; /* OF node not found */ | |
127 | u64 no_cfg_addr; /* Config address not found */ | |
128 | u64 ignored_check; /* EEH check skipped */ | |
129 | u64 total_mmio_ffs; /* Total EEH checks */ | |
130 | u64 false_positives; /* Unnecessary EEH checks */ | |
131 | u64 slot_resets; /* PE reset */ | |
132 | }; | |
133 | ||
134 | static struct eeh_stats eeh_stats; | |
1da177e4 | 135 | |
7684b40c LV |
136 | #define IS_BRIDGE(class_code) (((class_code)<<16) == PCI_BASE_CLASS_BRIDGE) |
137 | ||
d99bb1db | 138 | /** |
cce4b2d2 | 139 | * eeh_gather_pci_data - Copy assorted PCI config space registers to buff |
f631acd3 | 140 | * @edev: device to report data for |
d99bb1db LV |
141 | * @buf: point to buffer in which to log |
142 | * @len: amount of room in buffer | |
143 | * | |
144 | * This routine captures assorted PCI configuration space data, | |
145 | * and puts them into a buffer for RTAS error logging. | |
146 | */ | |
f631acd3 | 147 | static size_t eeh_gather_pci_data(struct eeh_dev *edev, char * buf, size_t len) |
d99bb1db | 148 | { |
f631acd3 GS |
149 | struct device_node *dn = eeh_dev_to_of_node(edev); |
150 | struct pci_dev *dev = eeh_dev_to_pci_dev(edev); | |
d99bb1db | 151 | u32 cfg; |
fcf9892b | 152 | int cap, i; |
d99bb1db LV |
153 | int n = 0; |
154 | ||
f631acd3 GS |
155 | n += scnprintf(buf+n, len-n, "%s\n", dn->full_name); |
156 | printk(KERN_WARNING "EEH: of node=%s\n", dn->full_name); | |
fcf9892b | 157 | |
3780444c | 158 | eeh_ops->read_config(dn, PCI_VENDOR_ID, 4, &cfg); |
fcf9892b LV |
159 | n += scnprintf(buf+n, len-n, "dev/vend:%08x\n", cfg); |
160 | printk(KERN_WARNING "EEH: PCI device/vendor: %08x\n", cfg); | |
161 | ||
3780444c | 162 | eeh_ops->read_config(dn, PCI_COMMAND, 4, &cfg); |
d99bb1db | 163 | n += scnprintf(buf+n, len-n, "cmd/stat:%x\n", cfg); |
fcf9892b LV |
164 | printk(KERN_WARNING "EEH: PCI cmd/status register: %08x\n", cfg); |
165 | ||
b37ceefe LV |
166 | if (!dev) { |
167 | printk(KERN_WARNING "EEH: no PCI device for this of node\n"); | |
168 | return n; | |
169 | } | |
170 | ||
0b9369f4 LV |
171 | /* Gather bridge-specific registers */ |
172 | if (dev->class >> 16 == PCI_BASE_CLASS_BRIDGE) { | |
3780444c | 173 | eeh_ops->read_config(dn, PCI_SEC_STATUS, 2, &cfg); |
0b9369f4 LV |
174 | n += scnprintf(buf+n, len-n, "sec stat:%x\n", cfg); |
175 | printk(KERN_WARNING "EEH: Bridge secondary status: %04x\n", cfg); | |
176 | ||
3780444c | 177 | eeh_ops->read_config(dn, PCI_BRIDGE_CONTROL, 2, &cfg); |
0b9369f4 LV |
178 | n += scnprintf(buf+n, len-n, "brdg ctl:%x\n", cfg); |
179 | printk(KERN_WARNING "EEH: Bridge control: %04x\n", cfg); | |
180 | } | |
181 | ||
fcf9892b | 182 | /* Dump out the PCI-X command and status regs */ |
b37ceefe | 183 | cap = pci_find_capability(dev, PCI_CAP_ID_PCIX); |
fcf9892b | 184 | if (cap) { |
3780444c | 185 | eeh_ops->read_config(dn, cap, 4, &cfg); |
fcf9892b LV |
186 | n += scnprintf(buf+n, len-n, "pcix-cmd:%x\n", cfg); |
187 | printk(KERN_WARNING "EEH: PCI-X cmd: %08x\n", cfg); | |
188 | ||
3780444c | 189 | eeh_ops->read_config(dn, cap+4, 4, &cfg); |
fcf9892b LV |
190 | n += scnprintf(buf+n, len-n, "pcix-stat:%x\n", cfg); |
191 | printk(KERN_WARNING "EEH: PCI-X status: %08x\n", cfg); | |
192 | } | |
193 | ||
194 | /* If PCI-E capable, dump PCI-E cap 10, and the AER */ | |
b37ceefe | 195 | cap = pci_find_capability(dev, PCI_CAP_ID_EXP); |
fcf9892b LV |
196 | if (cap) { |
197 | n += scnprintf(buf+n, len-n, "pci-e cap10:\n"); | |
198 | printk(KERN_WARNING | |
199 | "EEH: PCI-E capabilities and status follow:\n"); | |
200 | ||
201 | for (i=0; i<=8; i++) { | |
3780444c | 202 | eeh_ops->read_config(dn, cap+4*i, 4, &cfg); |
fcf9892b LV |
203 | n += scnprintf(buf+n, len-n, "%02x:%x\n", 4*i, cfg); |
204 | printk(KERN_WARNING "EEH: PCI-E %02x: %08x\n", i, cfg); | |
205 | } | |
206 | ||
b37ceefe | 207 | cap = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR); |
fcf9892b LV |
208 | if (cap) { |
209 | n += scnprintf(buf+n, len-n, "pci-e AER:\n"); | |
210 | printk(KERN_WARNING | |
211 | "EEH: PCI-E AER capability register set follows:\n"); | |
212 | ||
213 | for (i=0; i<14; i++) { | |
3780444c | 214 | eeh_ops->read_config(dn, cap+4*i, 4, &cfg); |
fcf9892b LV |
215 | n += scnprintf(buf+n, len-n, "%02x:%x\n", 4*i, cfg); |
216 | printk(KERN_WARNING "EEH: PCI-E AER %02x: %08x\n", i, cfg); | |
217 | } | |
218 | } | |
219 | } | |
0b9369f4 | 220 | |
d99bb1db LV |
221 | return n; |
222 | } | |
223 | ||
cb3bc9d0 GS |
224 | /** |
225 | * eeh_slot_error_detail - Generate combined log including driver log and error log | |
ff477966 | 226 | * @pe: EEH PE |
cb3bc9d0 GS |
227 | * @severity: temporary or permanent error log |
228 | * | |
229 | * This routine should be called to generate the combined log, which | |
230 | * is comprised of driver log and error log. The driver log is figured | |
231 | * out from the config space of the corresponding PCI device, while | |
232 | * the error log is fetched through platform dependent function call. | |
233 | */ | |
ff477966 | 234 | void eeh_slot_error_detail(struct eeh_pe *pe, int severity) |
d99bb1db LV |
235 | { |
236 | size_t loglen = 0; | |
ff477966 | 237 | struct eeh_dev *edev; |
d99bb1db | 238 | |
ff477966 GS |
239 | eeh_pci_enable(pe, EEH_OPT_THAW_MMIO); |
240 | eeh_ops->configure_bridge(pe); | |
241 | eeh_pe_restore_bars(pe); | |
d99bb1db | 242 | |
ff477966 GS |
243 | pci_regs_buf[0] = 0; |
244 | eeh_pe_for_each_dev(pe, edev) { | |
245 | loglen += eeh_gather_pci_data(edev, pci_regs_buf, | |
246 | EEH_PCI_REGS_LOG_LEN); | |
247 | } | |
248 | ||
249 | eeh_ops->get_log(pe, severity, pci_regs_buf, loglen); | |
d99bb1db LV |
250 | } |
251 | ||
1da177e4 | 252 | /** |
cb3bc9d0 GS |
253 | * eeh_token_to_phys - Convert EEH address token to phys address |
254 | * @token: I/O token, should be address in the form 0xA.... | |
255 | * | |
256 | * This routine should be called to convert virtual I/O address | |
257 | * to physical one. | |
1da177e4 LT |
258 | */ |
259 | static inline unsigned long eeh_token_to_phys(unsigned long token) | |
260 | { | |
261 | pte_t *ptep; | |
262 | unsigned long pa; | |
263 | ||
20cee16c | 264 | ptep = find_linux_pte(init_mm.pgd, token); |
1da177e4 LT |
265 | if (!ptep) |
266 | return token; | |
267 | pa = pte_pfn(*ptep) << PAGE_SHIFT; | |
268 | ||
269 | return pa | (token & (PAGE_SIZE-1)); | |
270 | } | |
271 | ||
b95cd2cd GS |
272 | /* |
273 | * On PowerNV platform, we might already have fenced PHB there. | |
274 | * For that case, it's meaningless to recover frozen PE. Intead, | |
275 | * We have to handle fenced PHB firstly. | |
276 | */ | |
277 | static int eeh_phb_check_failure(struct eeh_pe *pe) | |
278 | { | |
279 | struct eeh_pe *phb_pe; | |
280 | unsigned long flags; | |
281 | int ret; | |
282 | ||
283 | if (!eeh_probe_mode_dev()) | |
284 | return -EPERM; | |
285 | ||
286 | /* Find the PHB PE */ | |
287 | phb_pe = eeh_phb_pe_get(pe->phb); | |
288 | if (!phb_pe) { | |
289 | pr_warning("%s Can't find PE for PHB#%d\n", | |
290 | __func__, pe->phb->global_number); | |
291 | return -EEXIST; | |
292 | } | |
293 | ||
294 | /* If the PHB has been in problematic state */ | |
295 | eeh_serialize_lock(&flags); | |
296 | if (phb_pe->state & (EEH_PE_ISOLATED | EEH_PE_PHB_DEAD)) { | |
297 | ret = 0; | |
298 | goto out; | |
299 | } | |
300 | ||
301 | /* Check PHB state */ | |
302 | ret = eeh_ops->get_state(phb_pe, NULL); | |
303 | if ((ret < 0) || | |
304 | (ret == EEH_STATE_NOT_SUPPORT) || | |
305 | (ret & (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE)) == | |
306 | (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE)) { | |
307 | ret = 0; | |
308 | goto out; | |
309 | } | |
310 | ||
311 | /* Isolate the PHB and send event */ | |
312 | eeh_pe_state_mark(phb_pe, EEH_PE_ISOLATED); | |
313 | eeh_serialize_unlock(flags); | |
314 | eeh_send_failure_event(phb_pe); | |
315 | ||
316 | WARN(1, "EEH: PHB failure detected\n"); | |
317 | ||
318 | return 1; | |
319 | out: | |
320 | eeh_serialize_unlock(flags); | |
321 | return ret; | |
322 | } | |
323 | ||
1da177e4 | 324 | /** |
f8f7d63f GS |
325 | * eeh_dev_check_failure - Check if all 1's data is due to EEH slot freeze |
326 | * @edev: eeh device | |
1da177e4 LT |
327 | * |
328 | * Check for an EEH failure for the given device node. Call this | |
329 | * routine if the result of a read was all 0xff's and you want to | |
330 | * find out if this is due to an EEH slot freeze. This routine | |
331 | * will query firmware for the EEH status. | |
332 | * | |
333 | * Returns 0 if there has not been an EEH error; otherwise returns | |
69376502 | 334 | * a non-zero value and queues up a slot isolation event notification. |
1da177e4 LT |
335 | * |
336 | * It is safe to call this routine in an interrupt context. | |
337 | */ | |
f8f7d63f | 338 | int eeh_dev_check_failure(struct eeh_dev *edev) |
1da177e4 LT |
339 | { |
340 | int ret; | |
1da177e4 | 341 | unsigned long flags; |
f8f7d63f GS |
342 | struct device_node *dn; |
343 | struct pci_dev *dev; | |
66523d9f | 344 | struct eeh_pe *pe; |
fd761fd8 | 345 | int rc = 0; |
f36c5227 | 346 | const char *location; |
1da177e4 | 347 | |
e575f8db | 348 | eeh_stats.total_mmio_ffs++; |
1da177e4 LT |
349 | |
350 | if (!eeh_subsystem_enabled) | |
351 | return 0; | |
352 | ||
f8f7d63f | 353 | if (!edev) { |
e575f8db | 354 | eeh_stats.no_dn++; |
1da177e4 | 355 | return 0; |
177bc936 | 356 | } |
f8f7d63f GS |
357 | dn = eeh_dev_to_of_node(edev); |
358 | dev = eeh_dev_to_pci_dev(edev); | |
66523d9f | 359 | pe = edev->pe; |
1da177e4 LT |
360 | |
361 | /* Access to IO BARs might get this far and still not want checking. */ | |
66523d9f | 362 | if (!pe) { |
e575f8db | 363 | eeh_stats.ignored_check++; |
66523d9f GS |
364 | pr_debug("EEH: Ignored check for %s %s\n", |
365 | eeh_pci_name(dev), dn->full_name); | |
1da177e4 LT |
366 | return 0; |
367 | } | |
368 | ||
66523d9f | 369 | if (!pe->addr && !pe->config_addr) { |
e575f8db | 370 | eeh_stats.no_cfg_addr++; |
1da177e4 LT |
371 | return 0; |
372 | } | |
373 | ||
b95cd2cd GS |
374 | /* |
375 | * On PowerNV platform, we might already have fenced PHB | |
376 | * there and we need take care of that firstly. | |
377 | */ | |
378 | ret = eeh_phb_check_failure(pe); | |
379 | if (ret > 0) | |
380 | return ret; | |
381 | ||
fd761fd8 LV |
382 | /* If we already have a pending isolation event for this |
383 | * slot, we know it's bad already, we don't need to check. | |
384 | * Do this checking under a lock; as multiple PCI devices | |
385 | * in one slot might report errors simultaneously, and we | |
386 | * only want one error recovery routine running. | |
1da177e4 | 387 | */ |
4907581d | 388 | eeh_serialize_lock(&flags); |
fd761fd8 | 389 | rc = 1; |
66523d9f GS |
390 | if (pe->state & EEH_PE_ISOLATED) { |
391 | pe->check_count++; | |
392 | if (pe->check_count % EEH_MAX_FAILS == 0) { | |
f36c5227 | 393 | location = of_get_property(dn, "ibm,loc-code", NULL); |
cb3bc9d0 | 394 | printk(KERN_ERR "EEH: %d reads ignored for recovering device at " |
f36c5227 | 395 | "location=%s driver=%s pci addr=%s\n", |
66523d9f | 396 | pe->check_count, location, |
778a785f | 397 | eeh_driver_name(dev), eeh_pci_name(dev)); |
cb3bc9d0 | 398 | printk(KERN_ERR "EEH: Might be infinite loop in %s driver\n", |
778a785f | 399 | eeh_driver_name(dev)); |
5c1344e9 | 400 | dump_stack(); |
1da177e4 | 401 | } |
fd761fd8 | 402 | goto dn_unlock; |
1da177e4 LT |
403 | } |
404 | ||
405 | /* | |
406 | * Now test for an EEH failure. This is VERY expensive. | |
407 | * Note that the eeh_config_addr may be a parent device | |
408 | * in the case of a device behind a bridge, or it may be | |
409 | * function zero of a multi-function device. | |
410 | * In any case they must share a common PHB. | |
411 | */ | |
66523d9f | 412 | ret = eeh_ops->get_state(pe, NULL); |
76e6faf7 | 413 | |
39d16e29 | 414 | /* Note that config-io to empty slots may fail; |
cb3bc9d0 | 415 | * they are empty when they don't have children. |
eb594a47 GS |
416 | * We will punt with the following conditions: Failure to get |
417 | * PE's state, EEH not support and Permanently unavailable | |
418 | * state, PE is in good state. | |
cb3bc9d0 | 419 | */ |
eb594a47 GS |
420 | if ((ret < 0) || |
421 | (ret == EEH_STATE_NOT_SUPPORT) || | |
422 | (ret & (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE)) == | |
423 | (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE)) { | |
e575f8db | 424 | eeh_stats.false_positives++; |
66523d9f | 425 | pe->false_positives++; |
fd761fd8 LV |
426 | rc = 0; |
427 | goto dn_unlock; | |
76e6faf7 LV |
428 | } |
429 | ||
e575f8db | 430 | eeh_stats.slot_resets++; |
a84f273c | 431 | |
fd761fd8 LV |
432 | /* Avoid repeated reports of this failure, including problems |
433 | * with other functions on this device, and functions under | |
cb3bc9d0 GS |
434 | * bridges. |
435 | */ | |
66523d9f | 436 | eeh_pe_state_mark(pe, EEH_PE_ISOLATED); |
4907581d | 437 | eeh_serialize_unlock(flags); |
1da177e4 | 438 | |
66523d9f | 439 | eeh_send_failure_event(pe); |
77bd7415 | 440 | |
1da177e4 LT |
441 | /* Most EEH events are due to device driver bugs. Having |
442 | * a stack trace will help the device-driver authors figure | |
cb3bc9d0 GS |
443 | * out what happened. So print that out. |
444 | */ | |
14fb1fa6 | 445 | WARN(1, "EEH: failure detected\n"); |
fd761fd8 LV |
446 | return 1; |
447 | ||
448 | dn_unlock: | |
4907581d | 449 | eeh_serialize_unlock(flags); |
fd761fd8 | 450 | return rc; |
1da177e4 LT |
451 | } |
452 | ||
f8f7d63f | 453 | EXPORT_SYMBOL_GPL(eeh_dev_check_failure); |
1da177e4 LT |
454 | |
455 | /** | |
cb3bc9d0 GS |
456 | * eeh_check_failure - Check if all 1's data is due to EEH slot freeze |
457 | * @token: I/O token, should be address in the form 0xA.... | |
458 | * @val: value, should be all 1's (XXX why do we need this arg??) | |
1da177e4 | 459 | * |
1da177e4 LT |
460 | * Check for an EEH failure at the given token address. Call this |
461 | * routine if the result of a read was all 0xff's and you want to | |
462 | * find out if this is due to an EEH slot freeze event. This routine | |
463 | * will query firmware for the EEH status. | |
464 | * | |
465 | * Note this routine is safe to call in an interrupt context. | |
466 | */ | |
467 | unsigned long eeh_check_failure(const volatile void __iomem *token, unsigned long val) | |
468 | { | |
469 | unsigned long addr; | |
f8f7d63f | 470 | struct eeh_dev *edev; |
1da177e4 LT |
471 | |
472 | /* Finding the phys addr + pci device; this is pretty quick. */ | |
473 | addr = eeh_token_to_phys((unsigned long __force) token); | |
3ab96a02 | 474 | edev = eeh_addr_cache_get_dev(addr); |
f8f7d63f | 475 | if (!edev) { |
e575f8db | 476 | eeh_stats.no_device++; |
1da177e4 | 477 | return val; |
177bc936 | 478 | } |
1da177e4 | 479 | |
f8f7d63f | 480 | eeh_dev_check_failure(edev); |
1da177e4 | 481 | |
f8f7d63f | 482 | pci_dev_put(eeh_dev_to_pci_dev(edev)); |
1da177e4 LT |
483 | return val; |
484 | } | |
485 | ||
486 | EXPORT_SYMBOL(eeh_check_failure); | |
487 | ||
6dee3fb9 | 488 | |
47b5c838 | 489 | /** |
cce4b2d2 | 490 | * eeh_pci_enable - Enable MMIO or DMA transfers for this slot |
ff477966 | 491 | * @pe: EEH PE |
cb3bc9d0 GS |
492 | * |
493 | * This routine should be called to reenable frozen MMIO or DMA | |
494 | * so that it would work correctly again. It's useful while doing | |
495 | * recovery or log collection on the indicated device. | |
47b5c838 | 496 | */ |
ff477966 | 497 | int eeh_pci_enable(struct eeh_pe *pe, int function) |
47b5c838 | 498 | { |
47b5c838 LV |
499 | int rc; |
500 | ||
ff477966 | 501 | rc = eeh_ops->set_option(pe, function); |
47b5c838 | 502 | if (rc) |
ff477966 GS |
503 | pr_warning("%s: Unexpected state change %d on PHB#%d-PE#%x, err=%d\n", |
504 | __func__, function, pe->phb->global_number, pe->addr, rc); | |
47b5c838 | 505 | |
ff477966 | 506 | rc = eeh_ops->wait_state(pe, PCI_BUS_RESET_WAIT_MSEC); |
eb594a47 GS |
507 | if (rc > 0 && (rc & EEH_STATE_MMIO_ENABLED) && |
508 | (function == EEH_OPT_THAW_MMIO)) | |
fa1be476 LV |
509 | return 0; |
510 | ||
47b5c838 LV |
511 | return rc; |
512 | } | |
513 | ||
00c2ae35 BK |
514 | /** |
515 | * pcibios_set_pcie_slot_reset - Set PCI-E reset state | |
cb3bc9d0 GS |
516 | * @dev: pci device struct |
517 | * @state: reset state to enter | |
00c2ae35 BK |
518 | * |
519 | * Return value: | |
520 | * 0 if success | |
cb3bc9d0 | 521 | */ |
00c2ae35 BK |
522 | int pcibios_set_pcie_reset_state(struct pci_dev *dev, enum pcie_reset_state state) |
523 | { | |
c270a24c GS |
524 | struct eeh_dev *edev = pci_dev_to_eeh_dev(dev); |
525 | struct eeh_pe *pe = edev->pe; | |
526 | ||
527 | if (!pe) { | |
528 | pr_err("%s: No PE found on PCI device %s\n", | |
529 | __func__, pci_name(dev)); | |
530 | return -EINVAL; | |
531 | } | |
00c2ae35 BK |
532 | |
533 | switch (state) { | |
534 | case pcie_deassert_reset: | |
c270a24c | 535 | eeh_ops->reset(pe, EEH_RESET_DEACTIVATE); |
00c2ae35 BK |
536 | break; |
537 | case pcie_hot_reset: | |
c270a24c | 538 | eeh_ops->reset(pe, EEH_RESET_HOT); |
00c2ae35 BK |
539 | break; |
540 | case pcie_warm_reset: | |
c270a24c | 541 | eeh_ops->reset(pe, EEH_RESET_FUNDAMENTAL); |
00c2ae35 BK |
542 | break; |
543 | default: | |
544 | return -EINVAL; | |
545 | }; | |
546 | ||
547 | return 0; | |
548 | } | |
549 | ||
cb5b5624 | 550 | /** |
c270a24c GS |
551 | * eeh_set_pe_freset - Check the required reset for the indicated device |
552 | * @data: EEH device | |
553 | * @flag: return value | |
cb3bc9d0 GS |
554 | * |
555 | * Each device might have its preferred reset type: fundamental or | |
556 | * hot reset. The routine is used to collected the information for | |
557 | * the indicated device and its children so that the bunch of the | |
558 | * devices could be reset properly. | |
559 | */ | |
c270a24c | 560 | static void *eeh_set_dev_freset(void *data, void *flag) |
cb3bc9d0 GS |
561 | { |
562 | struct pci_dev *dev; | |
c270a24c GS |
563 | unsigned int *freset = (unsigned int *)flag; |
564 | struct eeh_dev *edev = (struct eeh_dev *)data; | |
6dee3fb9 | 565 | |
c270a24c | 566 | dev = eeh_dev_to_pci_dev(edev); |
cb3bc9d0 GS |
567 | if (dev) |
568 | *freset |= dev->needs_freset; | |
569 | ||
c270a24c | 570 | return NULL; |
cb3bc9d0 GS |
571 | } |
572 | ||
573 | /** | |
cce4b2d2 | 574 | * eeh_reset_pe_once - Assert the pci #RST line for 1/4 second |
c270a24c | 575 | * @pe: EEH PE |
cb3bc9d0 GS |
576 | * |
577 | * Assert the PCI #RST line for 1/4 second. | |
578 | */ | |
c270a24c | 579 | static void eeh_reset_pe_once(struct eeh_pe *pe) |
6dee3fb9 | 580 | { |
308fc4f8 | 581 | unsigned int freset = 0; |
6e19314c | 582 | |
308fc4f8 RL |
583 | /* Determine type of EEH reset required for |
584 | * Partitionable Endpoint, a hot-reset (1) | |
585 | * or a fundamental reset (3). | |
586 | * A fundamental reset required by any device under | |
587 | * Partitionable Endpoint trumps hot-reset. | |
a84f273c | 588 | */ |
c270a24c | 589 | eeh_pe_dev_traverse(pe, eeh_set_dev_freset, &freset); |
308fc4f8 RL |
590 | |
591 | if (freset) | |
c270a24c | 592 | eeh_ops->reset(pe, EEH_RESET_FUNDAMENTAL); |
6e19314c | 593 | else |
c270a24c | 594 | eeh_ops->reset(pe, EEH_RESET_HOT); |
6dee3fb9 LV |
595 | |
596 | /* The PCI bus requires that the reset be held high for at least | |
cb3bc9d0 GS |
597 | * a 100 milliseconds. We wait a bit longer 'just in case'. |
598 | */ | |
6dee3fb9 | 599 | #define PCI_BUS_RST_HOLD_TIME_MSEC 250 |
cb3bc9d0 | 600 | msleep(PCI_BUS_RST_HOLD_TIME_MSEC); |
a84f273c GS |
601 | |
602 | /* We might get hit with another EEH freeze as soon as the | |
d9564ad1 | 603 | * pci slot reset line is dropped. Make sure we don't miss |
cb3bc9d0 GS |
604 | * these, and clear the flag now. |
605 | */ | |
dbbceee1 | 606 | eeh_pe_state_clear(pe, EEH_PE_ISOLATED); |
d9564ad1 | 607 | |
c270a24c | 608 | eeh_ops->reset(pe, EEH_RESET_DEACTIVATE); |
6dee3fb9 LV |
609 | |
610 | /* After a PCI slot has been reset, the PCI Express spec requires | |
611 | * a 1.5 second idle time for the bus to stabilize, before starting | |
cb3bc9d0 GS |
612 | * up traffic. |
613 | */ | |
6dee3fb9 | 614 | #define PCI_BUS_SETTLE_TIME_MSEC 1800 |
cb3bc9d0 | 615 | msleep(PCI_BUS_SETTLE_TIME_MSEC); |
e1029263 LV |
616 | } |
617 | ||
cb3bc9d0 | 618 | /** |
cce4b2d2 | 619 | * eeh_reset_pe - Reset the indicated PE |
c270a24c | 620 | * @pe: EEH PE |
cb3bc9d0 GS |
621 | * |
622 | * This routine should be called to reset indicated device, including | |
623 | * PE. A PE might include multiple PCI devices and sometimes PCI bridges | |
624 | * might be involved as well. | |
625 | */ | |
c270a24c | 626 | int eeh_reset_pe(struct eeh_pe *pe) |
e1029263 | 627 | { |
326a98ea | 628 | int flags = (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE); |
e1029263 LV |
629 | int i, rc; |
630 | ||
9c547768 LV |
631 | /* Take three shots at resetting the bus */ |
632 | for (i=0; i<3; i++) { | |
c270a24c | 633 | eeh_reset_pe_once(pe); |
6dee3fb9 | 634 | |
c270a24c | 635 | rc = eeh_ops->wait_state(pe, PCI_BUS_RESET_WAIT_MSEC); |
326a98ea | 636 | if ((rc & flags) == flags) |
b6495c0c | 637 | return 0; |
e1029263 | 638 | |
e1029263 | 639 | if (rc < 0) { |
c270a24c GS |
640 | pr_err("%s: Unrecoverable slot failure on PHB#%d-PE#%x", |
641 | __func__, pe->phb->global_number, pe->addr); | |
b6495c0c | 642 | return -1; |
e1029263 | 643 | } |
c270a24c GS |
644 | pr_err("EEH: bus reset %d failed on PHB#%d-PE#%x, rc=%d\n", |
645 | i+1, pe->phb->global_number, pe->addr, rc); | |
6dee3fb9 | 646 | } |
b6495c0c | 647 | |
9c547768 | 648 | return -1; |
6dee3fb9 LV |
649 | } |
650 | ||
8b553f32 | 651 | /** |
cb3bc9d0 | 652 | * eeh_save_bars - Save device bars |
f631acd3 | 653 | * @edev: PCI device associated EEH device |
8b553f32 LV |
654 | * |
655 | * Save the values of the device bars. Unlike the restore | |
656 | * routine, this routine is *not* recursive. This is because | |
31116f0b | 657 | * PCI devices are added individually; but, for the restore, |
8b553f32 LV |
658 | * an entire slot is reset at a time. |
659 | */ | |
d7bb8862 | 660 | void eeh_save_bars(struct eeh_dev *edev) |
8b553f32 LV |
661 | { |
662 | int i; | |
f631acd3 | 663 | struct device_node *dn; |
8b553f32 | 664 | |
f631acd3 | 665 | if (!edev) |
8b553f32 | 666 | return; |
f631acd3 | 667 | dn = eeh_dev_to_of_node(edev); |
a84f273c | 668 | |
8b553f32 | 669 | for (i = 0; i < 16; i++) |
3780444c | 670 | eeh_ops->read_config(dn, i * 4, 4, &edev->config_space[i]); |
8b553f32 LV |
671 | } |
672 | ||
aa1e6374 GS |
673 | /** |
674 | * eeh_ops_register - Register platform dependent EEH operations | |
675 | * @ops: platform dependent EEH operations | |
676 | * | |
677 | * Register the platform dependent EEH operation callback | |
678 | * functions. The platform should call this function before | |
679 | * any other EEH operations. | |
680 | */ | |
681 | int __init eeh_ops_register(struct eeh_ops *ops) | |
682 | { | |
683 | if (!ops->name) { | |
684 | pr_warning("%s: Invalid EEH ops name for %p\n", | |
685 | __func__, ops); | |
686 | return -EINVAL; | |
687 | } | |
688 | ||
689 | if (eeh_ops && eeh_ops != ops) { | |
690 | pr_warning("%s: EEH ops of platform %s already existing (%s)\n", | |
691 | __func__, eeh_ops->name, ops->name); | |
692 | return -EEXIST; | |
693 | } | |
694 | ||
695 | eeh_ops = ops; | |
696 | ||
697 | return 0; | |
698 | } | |
699 | ||
700 | /** | |
701 | * eeh_ops_unregister - Unreigster platform dependent EEH operations | |
702 | * @name: name of EEH platform operations | |
703 | * | |
704 | * Unregister the platform dependent EEH operation callback | |
705 | * functions. | |
706 | */ | |
707 | int __exit eeh_ops_unregister(const char *name) | |
708 | { | |
709 | if (!name || !strlen(name)) { | |
710 | pr_warning("%s: Invalid EEH ops name\n", | |
711 | __func__); | |
712 | return -EINVAL; | |
713 | } | |
714 | ||
715 | if (eeh_ops && !strcmp(eeh_ops->name, name)) { | |
716 | eeh_ops = NULL; | |
717 | return 0; | |
718 | } | |
719 | ||
720 | return -EEXIST; | |
721 | } | |
722 | ||
cb3bc9d0 GS |
723 | /** |
724 | * eeh_init - EEH initialization | |
725 | * | |
1da177e4 LT |
726 | * Initialize EEH by trying to enable it for all of the adapters in the system. |
727 | * As a side effect we can determine here if eeh is supported at all. | |
728 | * Note that we leave EEH on so failed config cycles won't cause a machine | |
729 | * check. If a user turns off EEH for a particular adapter they are really | |
730 | * telling Linux to ignore errors. Some hardware (e.g. POWER5) won't | |
731 | * grant access to a slot if EEH isn't enabled, and so we always enable | |
732 | * EEH for all slots/all devices. | |
733 | * | |
734 | * The eeh-force-off option disables EEH checking globally, for all slots. | |
735 | * Even if force-off is set, the EEH hardware is still enabled, so that | |
736 | * newer systems can boot. | |
737 | */ | |
51fb5f56 | 738 | int __init eeh_init(void) |
1da177e4 | 739 | { |
1a5c2e63 GS |
740 | struct pci_controller *hose, *tmp; |
741 | struct device_node *phb; | |
51fb5f56 GS |
742 | static int cnt = 0; |
743 | int ret = 0; | |
744 | ||
745 | /* | |
746 | * We have to delay the initialization on PowerNV after | |
747 | * the PCI hierarchy tree has been built because the PEs | |
748 | * are figured out based on PCI devices instead of device | |
749 | * tree nodes | |
750 | */ | |
751 | if (machine_is(powernv) && cnt++ <= 0) | |
752 | return ret; | |
e2af155c GS |
753 | |
754 | /* call platform initialization function */ | |
755 | if (!eeh_ops) { | |
756 | pr_warning("%s: Platform EEH operation not found\n", | |
757 | __func__); | |
35e5cfe2 | 758 | return -EEXIST; |
e2af155c GS |
759 | } else if ((ret = eeh_ops->init())) { |
760 | pr_warning("%s: Failed to call platform init function (%d)\n", | |
761 | __func__, ret); | |
35e5cfe2 | 762 | return ret; |
e2af155c | 763 | } |
1da177e4 | 764 | |
c8608558 GS |
765 | /* Initialize EEH event */ |
766 | ret = eeh_event_init(); | |
767 | if (ret) | |
768 | return ret; | |
769 | ||
1a5c2e63 | 770 | /* Enable EEH for all adapters */ |
d7bb8862 GS |
771 | if (eeh_probe_mode_devtree()) { |
772 | list_for_each_entry_safe(hose, tmp, | |
773 | &hose_list, list_node) { | |
774 | phb = hose->dn; | |
775 | traverse_pci_devices(phb, eeh_ops->of_probe, NULL); | |
776 | } | |
51fb5f56 GS |
777 | } else if (eeh_probe_mode_dev()) { |
778 | list_for_each_entry_safe(hose, tmp, | |
779 | &hose_list, list_node) | |
780 | pci_walk_bus(hose->bus, eeh_ops->dev_probe, NULL); | |
781 | } else { | |
782 | pr_warning("%s: Invalid probe mode %d\n", | |
783 | __func__, eeh_probe_mode); | |
784 | return -EINVAL; | |
1da177e4 LT |
785 | } |
786 | ||
21fd21f5 GS |
787 | /* |
788 | * Call platform post-initialization. Actually, It's good chance | |
789 | * to inform platform that EEH is ready to supply service if the | |
790 | * I/O cache stuff has been built up. | |
791 | */ | |
792 | if (eeh_ops->post_init) { | |
793 | ret = eeh_ops->post_init(); | |
794 | if (ret) | |
795 | return ret; | |
796 | } | |
797 | ||
1da177e4 | 798 | if (eeh_subsystem_enabled) |
d7bb8862 | 799 | pr_info("EEH: PCI Enhanced I/O Error Handling Enabled\n"); |
1da177e4 | 800 | else |
d7bb8862 | 801 | pr_warning("EEH: No capable adapters found\n"); |
35e5cfe2 GS |
802 | |
803 | return ret; | |
1da177e4 LT |
804 | } |
805 | ||
35e5cfe2 GS |
806 | core_initcall_sync(eeh_init); |
807 | ||
1da177e4 | 808 | /** |
cb3bc9d0 | 809 | * eeh_add_device_early - Enable EEH for the indicated device_node |
1da177e4 LT |
810 | * @dn: device node for which to set up EEH |
811 | * | |
812 | * This routine must be used to perform EEH initialization for PCI | |
813 | * devices that were added after system boot (e.g. hotplug, dlpar). | |
814 | * This routine must be called before any i/o is performed to the | |
815 | * adapter (inluding any config-space i/o). | |
816 | * Whether this actually enables EEH or not for this device depends | |
817 | * on the CEC architecture, type of the device, on earlier boot | |
818 | * command-line arguments & etc. | |
819 | */ | |
794e085e | 820 | static void eeh_add_device_early(struct device_node *dn) |
1da177e4 LT |
821 | { |
822 | struct pci_controller *phb; | |
1da177e4 | 823 | |
26a74850 GS |
824 | /* |
825 | * If we're doing EEH probe based on PCI device, we | |
826 | * would delay the probe until late stage because | |
827 | * the PCI device isn't available this moment. | |
828 | */ | |
829 | if (!eeh_probe_mode_devtree()) | |
830 | return; | |
831 | ||
1e38b714 | 832 | if (!of_node_to_eeh_dev(dn)) |
1da177e4 | 833 | return; |
f631acd3 | 834 | phb = of_node_to_eeh_dev(dn)->phb; |
f751f841 LV |
835 | |
836 | /* USB Bus children of PCI devices will not have BUID's */ | |
837 | if (NULL == phb || 0 == phb->buid) | |
1da177e4 | 838 | return; |
1da177e4 | 839 | |
d7bb8862 | 840 | eeh_ops->of_probe(dn, NULL); |
1da177e4 | 841 | } |
1da177e4 | 842 | |
cb3bc9d0 GS |
843 | /** |
844 | * eeh_add_device_tree_early - Enable EEH for the indicated device | |
845 | * @dn: device node | |
846 | * | |
847 | * This routine must be used to perform EEH initialization for the | |
848 | * indicated PCI device that was added after system boot (e.g. | |
849 | * hotplug, dlpar). | |
850 | */ | |
e2a296ee LV |
851 | void eeh_add_device_tree_early(struct device_node *dn) |
852 | { | |
853 | struct device_node *sib; | |
acaa6176 SR |
854 | |
855 | for_each_child_of_node(dn, sib) | |
e2a296ee LV |
856 | eeh_add_device_tree_early(sib); |
857 | eeh_add_device_early(dn); | |
858 | } | |
859 | EXPORT_SYMBOL_GPL(eeh_add_device_tree_early); | |
860 | ||
1da177e4 | 861 | /** |
cb3bc9d0 | 862 | * eeh_add_device_late - Perform EEH initialization for the indicated pci device |
1da177e4 LT |
863 | * @dev: pci device for which to set up EEH |
864 | * | |
865 | * This routine must be used to complete EEH initialization for PCI | |
866 | * devices that were added after system boot (e.g. hotplug, dlpar). | |
867 | */ | |
794e085e | 868 | static void eeh_add_device_late(struct pci_dev *dev) |
1da177e4 | 869 | { |
56b0fca3 | 870 | struct device_node *dn; |
f631acd3 | 871 | struct eeh_dev *edev; |
56b0fca3 | 872 | |
1da177e4 LT |
873 | if (!dev || !eeh_subsystem_enabled) |
874 | return; | |
875 | ||
57b066ff | 876 | pr_debug("EEH: Adding device %s\n", pci_name(dev)); |
1da177e4 | 877 | |
56b0fca3 | 878 | dn = pci_device_to_OF_node(dev); |
2ef822c5 | 879 | edev = of_node_to_eeh_dev(dn); |
f631acd3 | 880 | if (edev->pdev == dev) { |
57b066ff BH |
881 | pr_debug("EEH: Already referenced !\n"); |
882 | return; | |
883 | } | |
f631acd3 | 884 | WARN_ON(edev->pdev); |
57b066ff | 885 | |
cb3bc9d0 | 886 | pci_dev_get(dev); |
f631acd3 GS |
887 | edev->pdev = dev; |
888 | dev->dev.archdata.edev = edev; | |
56b0fca3 | 889 | |
26a74850 GS |
890 | /* |
891 | * We have to do the EEH probe here because the PCI device | |
892 | * hasn't been created yet in the early stage. | |
893 | */ | |
894 | if (eeh_probe_mode_dev()) | |
895 | eeh_ops->dev_probe(dev, NULL); | |
896 | ||
3ab96a02 | 897 | eeh_addr_cache_insert_dev(dev); |
1da177e4 | 898 | } |
794e085e | 899 | |
cb3bc9d0 GS |
900 | /** |
901 | * eeh_add_device_tree_late - Perform EEH initialization for the indicated PCI bus | |
902 | * @bus: PCI bus | |
903 | * | |
904 | * This routine must be used to perform EEH initialization for PCI | |
905 | * devices which are attached to the indicated PCI bus. The PCI bus | |
906 | * is added after system boot through hotplug or dlpar. | |
907 | */ | |
794e085e NF |
908 | void eeh_add_device_tree_late(struct pci_bus *bus) |
909 | { | |
910 | struct pci_dev *dev; | |
911 | ||
912 | list_for_each_entry(dev, &bus->devices, bus_list) { | |
a84f273c GS |
913 | eeh_add_device_late(dev); |
914 | if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) { | |
915 | struct pci_bus *subbus = dev->subordinate; | |
916 | if (subbus) | |
917 | eeh_add_device_tree_late(subbus); | |
918 | } | |
794e085e NF |
919 | } |
920 | } | |
921 | EXPORT_SYMBOL_GPL(eeh_add_device_tree_late); | |
1da177e4 | 922 | |
6a040ce7 TLSC |
923 | /** |
924 | * eeh_add_sysfs_files - Add EEH sysfs files for the indicated PCI bus | |
925 | * @bus: PCI bus | |
926 | * | |
927 | * This routine must be used to add EEH sysfs files for PCI | |
928 | * devices which are attached to the indicated PCI bus. The PCI bus | |
929 | * is added after system boot through hotplug or dlpar. | |
930 | */ | |
931 | void eeh_add_sysfs_files(struct pci_bus *bus) | |
932 | { | |
933 | struct pci_dev *dev; | |
934 | ||
935 | list_for_each_entry(dev, &bus->devices, bus_list) { | |
936 | eeh_sysfs_add_device(dev); | |
937 | if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) { | |
938 | struct pci_bus *subbus = dev->subordinate; | |
939 | if (subbus) | |
940 | eeh_add_sysfs_files(subbus); | |
941 | } | |
942 | } | |
943 | } | |
944 | EXPORT_SYMBOL_GPL(eeh_add_sysfs_files); | |
945 | ||
1da177e4 | 946 | /** |
cb3bc9d0 | 947 | * eeh_remove_device - Undo EEH setup for the indicated pci device |
1da177e4 | 948 | * @dev: pci device to be removed |
20ee6a97 | 949 | * @purge_pe: remove the PE or not |
1da177e4 | 950 | * |
794e085e NF |
951 | * This routine should be called when a device is removed from |
952 | * a running system (e.g. by hotplug or dlpar). It unregisters | |
953 | * the PCI device from the EEH subsystem. I/O errors affecting | |
954 | * this device will no longer be detected after this call; thus, | |
955 | * i/o errors affecting this slot may leave this device unusable. | |
1da177e4 | 956 | */ |
20ee6a97 | 957 | static void eeh_remove_device(struct pci_dev *dev, int purge_pe) |
1da177e4 | 958 | { |
f631acd3 GS |
959 | struct eeh_dev *edev; |
960 | ||
1da177e4 LT |
961 | if (!dev || !eeh_subsystem_enabled) |
962 | return; | |
f631acd3 | 963 | edev = pci_dev_to_eeh_dev(dev); |
1da177e4 LT |
964 | |
965 | /* Unregister the device with the EEH/PCI address search system */ | |
57b066ff | 966 | pr_debug("EEH: Removing device %s\n", pci_name(dev)); |
56b0fca3 | 967 | |
f631acd3 | 968 | if (!edev || !edev->pdev) { |
57b066ff BH |
969 | pr_debug("EEH: Not referenced !\n"); |
970 | return; | |
b055a9e1 | 971 | } |
f631acd3 GS |
972 | edev->pdev = NULL; |
973 | dev->dev.archdata.edev = NULL; | |
cb3bc9d0 | 974 | pci_dev_put(dev); |
57b066ff | 975 | |
20ee6a97 | 976 | eeh_rmv_from_parent_pe(edev, purge_pe); |
3ab96a02 | 977 | eeh_addr_cache_rmv_dev(dev); |
57b066ff | 978 | eeh_sysfs_remove_device(dev); |
1da177e4 | 979 | } |
1da177e4 | 980 | |
cb3bc9d0 GS |
981 | /** |
982 | * eeh_remove_bus_device - Undo EEH setup for the indicated PCI device | |
983 | * @dev: PCI device | |
20ee6a97 | 984 | * @purge_pe: remove the corresponding PE or not |
cb3bc9d0 GS |
985 | * |
986 | * This routine must be called when a device is removed from the | |
987 | * running system through hotplug or dlpar. The corresponding | |
988 | * PCI address cache will be removed. | |
989 | */ | |
20ee6a97 | 990 | void eeh_remove_bus_device(struct pci_dev *dev, int purge_pe) |
e2a296ee | 991 | { |
794e085e NF |
992 | struct pci_bus *bus = dev->subordinate; |
993 | struct pci_dev *child, *tmp; | |
994 | ||
20ee6a97 | 995 | eeh_remove_device(dev, purge_pe); |
794e085e NF |
996 | |
997 | if (bus && dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) { | |
998 | list_for_each_entry_safe(child, tmp, &bus->devices, bus_list) | |
20ee6a97 | 999 | eeh_remove_bus_device(child, purge_pe); |
e2a296ee LV |
1000 | } |
1001 | } | |
1002 | EXPORT_SYMBOL_GPL(eeh_remove_bus_device); | |
1003 | ||
1da177e4 LT |
1004 | static int proc_eeh_show(struct seq_file *m, void *v) |
1005 | { | |
1da177e4 LT |
1006 | if (0 == eeh_subsystem_enabled) { |
1007 | seq_printf(m, "EEH Subsystem is globally disabled\n"); | |
e575f8db | 1008 | seq_printf(m, "eeh_total_mmio_ffs=%llu\n", eeh_stats.total_mmio_ffs); |
1da177e4 LT |
1009 | } else { |
1010 | seq_printf(m, "EEH Subsystem is enabled\n"); | |
177bc936 | 1011 | seq_printf(m, |
e575f8db GS |
1012 | "no device=%llu\n" |
1013 | "no device node=%llu\n" | |
1014 | "no config address=%llu\n" | |
1015 | "check not wanted=%llu\n" | |
1016 | "eeh_total_mmio_ffs=%llu\n" | |
1017 | "eeh_false_positives=%llu\n" | |
1018 | "eeh_slot_resets=%llu\n", | |
1019 | eeh_stats.no_device, | |
1020 | eeh_stats.no_dn, | |
1021 | eeh_stats.no_cfg_addr, | |
1022 | eeh_stats.ignored_check, | |
1023 | eeh_stats.total_mmio_ffs, | |
1024 | eeh_stats.false_positives, | |
1025 | eeh_stats.slot_resets); | |
1da177e4 LT |
1026 | } |
1027 | ||
1028 | return 0; | |
1029 | } | |
1030 | ||
1031 | static int proc_eeh_open(struct inode *inode, struct file *file) | |
1032 | { | |
1033 | return single_open(file, proc_eeh_show, NULL); | |
1034 | } | |
1035 | ||
5dfe4c96 | 1036 | static const struct file_operations proc_eeh_operations = { |
1da177e4 LT |
1037 | .open = proc_eeh_open, |
1038 | .read = seq_read, | |
1039 | .llseek = seq_lseek, | |
1040 | .release = single_release, | |
1041 | }; | |
1042 | ||
1043 | static int __init eeh_init_proc(void) | |
1044 | { | |
66747138 | 1045 | if (machine_is(pseries)) |
8feaa434 | 1046 | proc_create("powerpc/eeh", 0, NULL, &proc_eeh_operations); |
1da177e4 LT |
1047 | return 0; |
1048 | } | |
1049 | __initcall(eeh_init_proc); |