2 * Copyright IBM Corporation 2001, 2005, 2006
3 * Copyright Dave Engebretsen & Todd Inglett 2001
4 * Copyright Linas Vepstas 2005, 2006
5 * Copyright 2001-2012 IBM Corporation.
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.
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.
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
21 * Please address comments and feedback to Linas Vepstas <linas@austin.ibm.com>
24 #include <linux/delay.h>
25 #include <linux/sched.h>
26 #include <linux/init.h>
27 #include <linux/list.h>
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>
33 #include <linux/export.h>
36 #include <linux/atomic.h>
38 #include <asm/eeh_event.h>
40 #include <asm/machdep.h>
41 #include <asm/ppc-pci.h>
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.
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
69 * Ideally, a PCI device driver, when suspecting that an isolation
70 * event has occurred (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.
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.
84 #define EEH_MAX_FAILS 2100000
86 /* Time to wait for a PCI slot to report status, in milliseconds */
87 #define PCI_BUS_RESET_WAIT_MSEC (60*1000)
89 /* Platform dependent EEH operations */
90 struct eeh_ops *eeh_ops = NULL;
92 int eeh_subsystem_enabled;
93 EXPORT_SYMBOL(eeh_subsystem_enabled);
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
106 /* Global EEH mutex */
107 DEFINE_MUTEX(eeh_mutex);
109 /* Lock to avoid races due to multiple reports of an error */
110 static DEFINE_RAW_SPINLOCK(confirm_error_lock);
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
116 #define EEH_PCI_REGS_LOG_LEN 4096
117 static unsigned char pci_regs_buf[EEH_PCI_REGS_LOG_LEN];
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
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 */
134 static struct eeh_stats eeh_stats;
136 #define IS_BRIDGE(class_code) (((class_code)<<16) == PCI_BASE_CLASS_BRIDGE)
139 * eeh_gather_pci_data - Copy assorted PCI config space registers to buff
140 * @edev: device to report data for
141 * @buf: point to buffer in which to log
142 * @len: amount of room in buffer
144 * This routine captures assorted PCI configuration space data,
145 * and puts them into a buffer for RTAS error logging.
147 static size_t eeh_gather_pci_data(struct eeh_dev *edev, char * buf, size_t len)
149 struct device_node *dn = eeh_dev_to_of_node(edev);
150 struct pci_dev *dev = eeh_dev_to_pci_dev(edev);
155 n += scnprintf(buf+n, len-n, "%s\n", dn->full_name);
156 printk(KERN_WARNING "EEH: of node=%s\n", dn->full_name);
158 eeh_ops->read_config(dn, PCI_VENDOR_ID, 4, &cfg);
159 n += scnprintf(buf+n, len-n, "dev/vend:%08x\n", cfg);
160 printk(KERN_WARNING "EEH: PCI device/vendor: %08x\n", cfg);
162 eeh_ops->read_config(dn, PCI_COMMAND, 4, &cfg);
163 n += scnprintf(buf+n, len-n, "cmd/stat:%x\n", cfg);
164 printk(KERN_WARNING "EEH: PCI cmd/status register: %08x\n", cfg);
167 printk(KERN_WARNING "EEH: no PCI device for this of node\n");
171 /* Gather bridge-specific registers */
172 if (dev->class >> 16 == PCI_BASE_CLASS_BRIDGE) {
173 eeh_ops->read_config(dn, PCI_SEC_STATUS, 2, &cfg);
174 n += scnprintf(buf+n, len-n, "sec stat:%x\n", cfg);
175 printk(KERN_WARNING "EEH: Bridge secondary status: %04x\n", cfg);
177 eeh_ops->read_config(dn, PCI_BRIDGE_CONTROL, 2, &cfg);
178 n += scnprintf(buf+n, len-n, "brdg ctl:%x\n", cfg);
179 printk(KERN_WARNING "EEH: Bridge control: %04x\n", cfg);
182 /* Dump out the PCI-X command and status regs */
183 cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
185 eeh_ops->read_config(dn, cap, 4, &cfg);
186 n += scnprintf(buf+n, len-n, "pcix-cmd:%x\n", cfg);
187 printk(KERN_WARNING "EEH: PCI-X cmd: %08x\n", cfg);
189 eeh_ops->read_config(dn, cap+4, 4, &cfg);
190 n += scnprintf(buf+n, len-n, "pcix-stat:%x\n", cfg);
191 printk(KERN_WARNING "EEH: PCI-X status: %08x\n", cfg);
194 /* If PCI-E capable, dump PCI-E cap 10, and the AER */
195 cap = pci_find_capability(dev, PCI_CAP_ID_EXP);
197 n += scnprintf(buf+n, len-n, "pci-e cap10:\n");
199 "EEH: PCI-E capabilities and status follow:\n");
201 for (i=0; i<=8; i++) {
202 eeh_ops->read_config(dn, cap+4*i, 4, &cfg);
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);
207 cap = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
209 n += scnprintf(buf+n, len-n, "pci-e AER:\n");
211 "EEH: PCI-E AER capability register set follows:\n");
213 for (i=0; i<14; i++) {
214 eeh_ops->read_config(dn, cap+4*i, 4, &cfg);
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);
225 * eeh_slot_error_detail - Generate combined log including driver log and error log
227 * @severity: temporary or permanent error log
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.
234 void eeh_slot_error_detail(struct eeh_pe *pe, int severity)
237 struct eeh_dev *edev;
239 eeh_pci_enable(pe, EEH_OPT_THAW_MMIO);
240 eeh_ops->configure_bridge(pe);
241 eeh_pe_restore_bars(pe);
244 eeh_pe_for_each_dev(pe, edev) {
245 loglen += eeh_gather_pci_data(edev, pci_regs_buf,
246 EEH_PCI_REGS_LOG_LEN);
249 eeh_ops->get_log(pe, severity, pci_regs_buf, loglen);
253 * eeh_token_to_phys - Convert EEH address token to phys address
254 * @token: I/O token, should be address in the form 0xA....
256 * This routine should be called to convert virtual I/O address
259 static inline unsigned long eeh_token_to_phys(unsigned long token)
264 ptep = find_linux_pte(init_mm.pgd, token);
267 pa = pte_pfn(*ptep) << PAGE_SHIFT;
269 return pa | (token & (PAGE_SIZE-1));
273 * eeh_dn_check_failure - Check if all 1's data is due to EEH slot freeze
275 * @dev: pci device, if known
277 * Check for an EEH failure for the given device node. Call this
278 * routine if the result of a read was all 0xff's and you want to
279 * find out if this is due to an EEH slot freeze. This routine
280 * will query firmware for the EEH status.
282 * Returns 0 if there has not been an EEH error; otherwise returns
283 * a non-zero value and queues up a slot isolation event notification.
285 * It is safe to call this routine in an interrupt context.
287 int eeh_dn_check_failure(struct device_node *dn, struct pci_dev *dev)
292 struct eeh_dev *edev;
294 const char *location;
296 eeh_stats.total_mmio_ffs++;
298 if (!eeh_subsystem_enabled)
302 edev = of_node_to_eeh_dev(dn);
304 edev = pci_dev_to_eeh_dev(dev);
305 dn = pci_device_to_OF_node(dev);
312 /* Access to IO BARs might get this far and still not want checking. */
314 eeh_stats.ignored_check++;
315 pr_debug("EEH: Ignored check for %s %s\n",
316 eeh_pci_name(dev), dn->full_name);
320 if (!pe->addr && !pe->config_addr) {
321 eeh_stats.no_cfg_addr++;
325 /* If we already have a pending isolation event for this
326 * slot, we know it's bad already, we don't need to check.
327 * Do this checking under a lock; as multiple PCI devices
328 * in one slot might report errors simultaneously, and we
329 * only want one error recovery routine running.
331 raw_spin_lock_irqsave(&confirm_error_lock, flags);
333 if (pe->state & EEH_PE_ISOLATED) {
335 if (pe->check_count % EEH_MAX_FAILS == 0) {
336 location = of_get_property(dn, "ibm,loc-code", NULL);
337 printk(KERN_ERR "EEH: %d reads ignored for recovering device at "
338 "location=%s driver=%s pci addr=%s\n",
339 pe->check_count, location,
340 eeh_driver_name(dev), eeh_pci_name(dev));
341 printk(KERN_ERR "EEH: Might be infinite loop in %s driver\n",
342 eeh_driver_name(dev));
349 * Now test for an EEH failure. This is VERY expensive.
350 * Note that the eeh_config_addr may be a parent device
351 * in the case of a device behind a bridge, or it may be
352 * function zero of a multi-function device.
353 * In any case they must share a common PHB.
355 ret = eeh_ops->get_state(pe, NULL);
357 /* Note that config-io to empty slots may fail;
358 * they are empty when they don't have children.
359 * We will punt with the following conditions: Failure to get
360 * PE's state, EEH not support and Permanently unavailable
361 * state, PE is in good state.
364 (ret == EEH_STATE_NOT_SUPPORT) ||
365 (ret & (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE)) ==
366 (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE)) {
367 eeh_stats.false_positives++;
368 pe->false_positives++;
373 eeh_stats.slot_resets++;
375 /* Avoid repeated reports of this failure, including problems
376 * with other functions on this device, and functions under
379 eeh_pe_state_mark(pe, EEH_PE_ISOLATED);
380 raw_spin_unlock_irqrestore(&confirm_error_lock, flags);
382 eeh_send_failure_event(pe);
384 /* Most EEH events are due to device driver bugs. Having
385 * a stack trace will help the device-driver authors figure
386 * out what happened. So print that out.
388 WARN(1, "EEH: failure detected\n");
392 raw_spin_unlock_irqrestore(&confirm_error_lock, flags);
396 EXPORT_SYMBOL_GPL(eeh_dn_check_failure);
399 * eeh_check_failure - Check if all 1's data is due to EEH slot freeze
400 * @token: I/O token, should be address in the form 0xA....
401 * @val: value, should be all 1's (XXX why do we need this arg??)
403 * Check for an EEH failure at the given token address. Call this
404 * routine if the result of a read was all 0xff's and you want to
405 * find out if this is due to an EEH slot freeze event. This routine
406 * will query firmware for the EEH status.
408 * Note this routine is safe to call in an interrupt context.
410 unsigned long eeh_check_failure(const volatile void __iomem *token, unsigned long val)
414 struct device_node *dn;
416 /* Finding the phys addr + pci device; this is pretty quick. */
417 addr = eeh_token_to_phys((unsigned long __force) token);
418 dev = pci_addr_cache_get_device(addr);
420 eeh_stats.no_device++;
424 dn = pci_device_to_OF_node(dev);
425 eeh_dn_check_failure(dn, dev);
431 EXPORT_SYMBOL(eeh_check_failure);
435 * eeh_pci_enable - Enable MMIO or DMA transfers for this slot
438 * This routine should be called to reenable frozen MMIO or DMA
439 * so that it would work correctly again. It's useful while doing
440 * recovery or log collection on the indicated device.
442 int eeh_pci_enable(struct eeh_pe *pe, int function)
446 rc = eeh_ops->set_option(pe, function);
448 pr_warning("%s: Unexpected state change %d on PHB#%d-PE#%x, err=%d\n",
449 __func__, function, pe->phb->global_number, pe->addr, rc);
451 rc = eeh_ops->wait_state(pe, PCI_BUS_RESET_WAIT_MSEC);
452 if (rc > 0 && (rc & EEH_STATE_MMIO_ENABLED) &&
453 (function == EEH_OPT_THAW_MMIO))
460 * pcibios_set_pcie_slot_reset - Set PCI-E reset state
461 * @dev: pci device struct
462 * @state: reset state to enter
467 int pcibios_set_pcie_reset_state(struct pci_dev *dev, enum pcie_reset_state state)
469 struct eeh_dev *edev = pci_dev_to_eeh_dev(dev);
470 struct eeh_pe *pe = edev->pe;
473 pr_err("%s: No PE found on PCI device %s\n",
474 __func__, pci_name(dev));
479 case pcie_deassert_reset:
480 eeh_ops->reset(pe, EEH_RESET_DEACTIVATE);
483 eeh_ops->reset(pe, EEH_RESET_HOT);
485 case pcie_warm_reset:
486 eeh_ops->reset(pe, EEH_RESET_FUNDAMENTAL);
496 * eeh_set_pe_freset - Check the required reset for the indicated device
498 * @flag: return value
500 * Each device might have its preferred reset type: fundamental or
501 * hot reset. The routine is used to collected the information for
502 * the indicated device and its children so that the bunch of the
503 * devices could be reset properly.
505 static void *eeh_set_dev_freset(void *data, void *flag)
508 unsigned int *freset = (unsigned int *)flag;
509 struct eeh_dev *edev = (struct eeh_dev *)data;
511 dev = eeh_dev_to_pci_dev(edev);
513 *freset |= dev->needs_freset;
519 * eeh_reset_pe_once - Assert the pci #RST line for 1/4 second
522 * Assert the PCI #RST line for 1/4 second.
524 static void eeh_reset_pe_once(struct eeh_pe *pe)
526 unsigned int freset = 0;
528 /* Determine type of EEH reset required for
529 * Partitionable Endpoint, a hot-reset (1)
530 * or a fundamental reset (3).
531 * A fundamental reset required by any device under
532 * Partitionable Endpoint trumps hot-reset.
534 eeh_pe_dev_traverse(pe, eeh_set_dev_freset, &freset);
537 eeh_ops->reset(pe, EEH_RESET_FUNDAMENTAL);
539 eeh_ops->reset(pe, EEH_RESET_HOT);
541 /* The PCI bus requires that the reset be held high for at least
542 * a 100 milliseconds. We wait a bit longer 'just in case'.
544 #define PCI_BUS_RST_HOLD_TIME_MSEC 250
545 msleep(PCI_BUS_RST_HOLD_TIME_MSEC);
547 /* We might get hit with another EEH freeze as soon as the
548 * pci slot reset line is dropped. Make sure we don't miss
549 * these, and clear the flag now.
551 eeh_pe_state_clear(pe, EEH_PE_ISOLATED);
553 eeh_ops->reset(pe, EEH_RESET_DEACTIVATE);
555 /* After a PCI slot has been reset, the PCI Express spec requires
556 * a 1.5 second idle time for the bus to stabilize, before starting
559 #define PCI_BUS_SETTLE_TIME_MSEC 1800
560 msleep(PCI_BUS_SETTLE_TIME_MSEC);
564 * eeh_reset_pe - Reset the indicated PE
567 * This routine should be called to reset indicated device, including
568 * PE. A PE might include multiple PCI devices and sometimes PCI bridges
569 * might be involved as well.
571 int eeh_reset_pe(struct eeh_pe *pe)
575 /* Take three shots at resetting the bus */
576 for (i=0; i<3; i++) {
577 eeh_reset_pe_once(pe);
579 rc = eeh_ops->wait_state(pe, PCI_BUS_RESET_WAIT_MSEC);
580 if (rc == (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE))
584 pr_err("%s: Unrecoverable slot failure on PHB#%d-PE#%x",
585 __func__, pe->phb->global_number, pe->addr);
588 pr_err("EEH: bus reset %d failed on PHB#%d-PE#%x, rc=%d\n",
589 i+1, pe->phb->global_number, pe->addr, rc);
596 * eeh_save_bars - Save device bars
597 * @edev: PCI device associated EEH device
599 * Save the values of the device bars. Unlike the restore
600 * routine, this routine is *not* recursive. This is because
601 * PCI devices are added individually; but, for the restore,
602 * an entire slot is reset at a time.
604 void eeh_save_bars(struct eeh_dev *edev)
607 struct device_node *dn;
611 dn = eeh_dev_to_of_node(edev);
613 for (i = 0; i < 16; i++)
614 eeh_ops->read_config(dn, i * 4, 4, &edev->config_space[i]);
618 * eeh_ops_register - Register platform dependent EEH operations
619 * @ops: platform dependent EEH operations
621 * Register the platform dependent EEH operation callback
622 * functions. The platform should call this function before
623 * any other EEH operations.
625 int __init eeh_ops_register(struct eeh_ops *ops)
628 pr_warning("%s: Invalid EEH ops name for %p\n",
633 if (eeh_ops && eeh_ops != ops) {
634 pr_warning("%s: EEH ops of platform %s already existing (%s)\n",
635 __func__, eeh_ops->name, ops->name);
645 * eeh_ops_unregister - Unreigster platform dependent EEH operations
646 * @name: name of EEH platform operations
648 * Unregister the platform dependent EEH operation callback
651 int __exit eeh_ops_unregister(const char *name)
653 if (!name || !strlen(name)) {
654 pr_warning("%s: Invalid EEH ops name\n",
659 if (eeh_ops && !strcmp(eeh_ops->name, name)) {
668 * eeh_init - EEH initialization
670 * Initialize EEH by trying to enable it for all of the adapters in the system.
671 * As a side effect we can determine here if eeh is supported at all.
672 * Note that we leave EEH on so failed config cycles won't cause a machine
673 * check. If a user turns off EEH for a particular adapter they are really
674 * telling Linux to ignore errors. Some hardware (e.g. POWER5) won't
675 * grant access to a slot if EEH isn't enabled, and so we always enable
676 * EEH for all slots/all devices.
678 * The eeh-force-off option disables EEH checking globally, for all slots.
679 * Even if force-off is set, the EEH hardware is still enabled, so that
680 * newer systems can boot.
682 static int __init eeh_init(void)
684 struct pci_controller *hose, *tmp;
685 struct device_node *phb;
688 /* call platform initialization function */
690 pr_warning("%s: Platform EEH operation not found\n",
693 } else if ((ret = eeh_ops->init())) {
694 pr_warning("%s: Failed to call platform init function (%d)\n",
699 raw_spin_lock_init(&confirm_error_lock);
701 /* Enable EEH for all adapters */
702 if (eeh_probe_mode_devtree()) {
703 list_for_each_entry_safe(hose, tmp,
704 &hose_list, list_node) {
706 traverse_pci_devices(phb, eeh_ops->of_probe, NULL);
710 if (eeh_subsystem_enabled)
711 pr_info("EEH: PCI Enhanced I/O Error Handling Enabled\n");
713 pr_warning("EEH: No capable adapters found\n");
718 core_initcall_sync(eeh_init);
721 * eeh_add_device_early - Enable EEH for the indicated device_node
722 * @dn: device node for which to set up EEH
724 * This routine must be used to perform EEH initialization for PCI
725 * devices that were added after system boot (e.g. hotplug, dlpar).
726 * This routine must be called before any i/o is performed to the
727 * adapter (inluding any config-space i/o).
728 * Whether this actually enables EEH or not for this device depends
729 * on the CEC architecture, type of the device, on earlier boot
730 * command-line arguments & etc.
732 static void eeh_add_device_early(struct device_node *dn)
734 struct pci_controller *phb;
736 if (!dn || !of_node_to_eeh_dev(dn))
738 phb = of_node_to_eeh_dev(dn)->phb;
740 /* USB Bus children of PCI devices will not have BUID's */
741 if (NULL == phb || 0 == phb->buid)
744 /* FIXME: hotplug support on POWERNV */
745 eeh_ops->of_probe(dn, NULL);
749 * eeh_add_device_tree_early - Enable EEH for the indicated device
752 * This routine must be used to perform EEH initialization for the
753 * indicated PCI device that was added after system boot (e.g.
756 void eeh_add_device_tree_early(struct device_node *dn)
758 struct device_node *sib;
760 for_each_child_of_node(dn, sib)
761 eeh_add_device_tree_early(sib);
762 eeh_add_device_early(dn);
764 EXPORT_SYMBOL_GPL(eeh_add_device_tree_early);
767 * eeh_add_device_late - Perform EEH initialization for the indicated pci device
768 * @dev: pci device for which to set up EEH
770 * This routine must be used to complete EEH initialization for PCI
771 * devices that were added after system boot (e.g. hotplug, dlpar).
773 static void eeh_add_device_late(struct pci_dev *dev)
775 struct device_node *dn;
776 struct eeh_dev *edev;
778 if (!dev || !eeh_subsystem_enabled)
781 pr_debug("EEH: Adding device %s\n", pci_name(dev));
783 dn = pci_device_to_OF_node(dev);
784 edev = of_node_to_eeh_dev(dn);
785 if (edev->pdev == dev) {
786 pr_debug("EEH: Already referenced !\n");
793 dev->dev.archdata.edev = edev;
795 pci_addr_cache_insert_device(dev);
796 eeh_sysfs_add_device(dev);
800 * eeh_add_device_tree_late - Perform EEH initialization for the indicated PCI bus
803 * This routine must be used to perform EEH initialization for PCI
804 * devices which are attached to the indicated PCI bus. The PCI bus
805 * is added after system boot through hotplug or dlpar.
807 void eeh_add_device_tree_late(struct pci_bus *bus)
811 list_for_each_entry(dev, &bus->devices, bus_list) {
812 eeh_add_device_late(dev);
813 if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
814 struct pci_bus *subbus = dev->subordinate;
816 eeh_add_device_tree_late(subbus);
820 EXPORT_SYMBOL_GPL(eeh_add_device_tree_late);
823 * eeh_remove_device - Undo EEH setup for the indicated pci device
824 * @dev: pci device to be removed
826 * This routine should be called when a device is removed from
827 * a running system (e.g. by hotplug or dlpar). It unregisters
828 * the PCI device from the EEH subsystem. I/O errors affecting
829 * this device will no longer be detected after this call; thus,
830 * i/o errors affecting this slot may leave this device unusable.
832 static void eeh_remove_device(struct pci_dev *dev)
834 struct eeh_dev *edev;
836 if (!dev || !eeh_subsystem_enabled)
838 edev = pci_dev_to_eeh_dev(dev);
840 /* Unregister the device with the EEH/PCI address search system */
841 pr_debug("EEH: Removing device %s\n", pci_name(dev));
843 if (!edev || !edev->pdev) {
844 pr_debug("EEH: Not referenced !\n");
848 dev->dev.archdata.edev = NULL;
851 eeh_rmv_from_parent_pe(edev);
852 pci_addr_cache_remove_device(dev);
853 eeh_sysfs_remove_device(dev);
857 * eeh_remove_bus_device - Undo EEH setup for the indicated PCI device
860 * This routine must be called when a device is removed from the
861 * running system through hotplug or dlpar. The corresponding
862 * PCI address cache will be removed.
864 void eeh_remove_bus_device(struct pci_dev *dev)
866 struct pci_bus *bus = dev->subordinate;
867 struct pci_dev *child, *tmp;
869 eeh_remove_device(dev);
871 if (bus && dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
872 list_for_each_entry_safe(child, tmp, &bus->devices, bus_list)
873 eeh_remove_bus_device(child);
876 EXPORT_SYMBOL_GPL(eeh_remove_bus_device);
878 static int proc_eeh_show(struct seq_file *m, void *v)
880 if (0 == eeh_subsystem_enabled) {
881 seq_printf(m, "EEH Subsystem is globally disabled\n");
882 seq_printf(m, "eeh_total_mmio_ffs=%llu\n", eeh_stats.total_mmio_ffs);
884 seq_printf(m, "EEH Subsystem is enabled\n");
887 "no device node=%llu\n"
888 "no config address=%llu\n"
889 "check not wanted=%llu\n"
890 "eeh_total_mmio_ffs=%llu\n"
891 "eeh_false_positives=%llu\n"
892 "eeh_slot_resets=%llu\n",
895 eeh_stats.no_cfg_addr,
896 eeh_stats.ignored_check,
897 eeh_stats.total_mmio_ffs,
898 eeh_stats.false_positives,
899 eeh_stats.slot_resets);
905 static int proc_eeh_open(struct inode *inode, struct file *file)
907 return single_open(file, proc_eeh_show, NULL);
910 static const struct file_operations proc_eeh_operations = {
911 .open = proc_eeh_open,
914 .release = single_release,
917 static int __init eeh_init_proc(void)
919 if (machine_is(pseries))
920 proc_create("powerpc/eeh", 0, NULL, &proc_eeh_operations);
923 __initcall(eeh_init_proc);