blob: ddb80f5d850b77231783b7f05c1d7adadd6f9aed [file] [log] [blame]
Linas Vepstas172ca922005-11-03 18:50:04 -06001/*
2 * eeh_event.c
3 *
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or
7 * (at your option) any later version.
8 *
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17 *
18 * Copyright (c) 2005 Linas Vepstas <linas@linas.org>
19 */
20
Linas Vepstasac325ac2006-04-18 21:05:21 -070021#include <linux/delay.h>
Linas Vepstas172ca922005-11-03 18:50:04 -060022#include <linux/list.h>
Linas Vepstas8c33fd12006-03-29 15:29:18 -060023#include <linux/mutex.h>
Linas Vepstas172ca922005-11-03 18:50:04 -060024#include <linux/pci.h>
Linas Vepstas8c33fd12006-03-29 15:29:18 -060025#include <linux/workqueue.h>
Linas Vepstas172ca922005-11-03 18:50:04 -060026#include <asm/eeh_event.h>
Linas Vepstas77bd7412005-11-03 18:52:49 -060027#include <asm/ppc-pci.h>
Linas Vepstas172ca922005-11-03 18:50:04 -060028
29/** Overview:
30 * EEH error states may be detected within exception handlers;
31 * however, the recovery processing needs to occur asynchronously
32 * in a normal kernel context and not an interrupt context.
33 * This pair of routines creates an event and queues it onto a
34 * work-queue, where a worker thread can drive recovery.
35 */
36
37/* EEH event workqueue setup. */
Ingo Molnar34af9462006-06-27 02:53:55 -070038static DEFINE_SPINLOCK(eeh_eventlist_lock);
Linas Vepstas172ca922005-11-03 18:50:04 -060039LIST_HEAD(eeh_eventlist);
David Howellsc4028952006-11-22 14:57:56 +000040static void eeh_thread_launcher(struct work_struct *);
41DECLARE_WORK(eeh_event_wq, eeh_thread_launcher);
Linas Vepstas172ca922005-11-03 18:50:04 -060042
Linas Vepstas8c33fd12006-03-29 15:29:18 -060043/* Serialize reset sequences for a given pci device */
44DEFINE_MUTEX(eeh_event_mutex);
45
Linas Vepstas172ca922005-11-03 18:50:04 -060046/**
Linas Vepstas8c33fd12006-03-29 15:29:18 -060047 * eeh_event_handler - dispatch EEH events.
Linas Vepstas172ca922005-11-03 18:50:04 -060048 * @dummy - unused
Linas Vepstas8c33fd12006-03-29 15:29:18 -060049 *
50 * The detection of a frozen slot can occur inside an interrupt,
51 * where it can be hard to do anything about it. The goal of this
52 * routine is to pull these detection events out of the context
53 * of the interrupt handler, and re-dispatch them for processing
54 * at a later time in a normal context.
Linas Vepstas172ca922005-11-03 18:50:04 -060055 */
56static int eeh_event_handler(void * dummy)
57{
58 unsigned long flags;
59 struct eeh_event *event;
Linas Vepstasac325ac2006-04-18 21:05:21 -070060 struct pci_dn *pdn;
Linas Vepstas172ca922005-11-03 18:50:04 -060061
62 daemonize ("eehd");
Linas Vepstasac325ac2006-04-18 21:05:21 -070063 set_current_state(TASK_INTERRUPTIBLE);
Linas Vepstas172ca922005-11-03 18:50:04 -060064
Linas Vepstasac325ac2006-04-18 21:05:21 -070065 spin_lock_irqsave(&eeh_eventlist_lock, flags);
66 event = NULL;
Linas Vepstas172ca922005-11-03 18:50:04 -060067
Linas Vepstasac325ac2006-04-18 21:05:21 -070068 /* Unqueue the event, get ready to process. */
69 if (!list_empty(&eeh_eventlist)) {
70 event = list_entry(eeh_eventlist.next, struct eeh_event, list);
71 list_del(&event->list);
72 }
73 spin_unlock_irqrestore(&eeh_eventlist_lock, flags);
Linas Vepstas77bd7412005-11-03 18:52:49 -060074
Linas Vepstasac325ac2006-04-18 21:05:21 -070075 if (event == NULL)
76 return 0;
Linas Vepstas8c33fd12006-03-29 15:29:18 -060077
Linas Vepstasac325ac2006-04-18 21:05:21 -070078 /* Serialize processing of EEH events */
79 mutex_lock(&eeh_event_mutex);
80 eeh_mark_slot(event->dn, EEH_MODE_RECOVERING);
Linas Vepstas172ca922005-11-03 18:50:04 -060081
Linas Vepstasac325ac2006-04-18 21:05:21 -070082 printk(KERN_INFO "EEH: Detected PCI bus error on device %s\n",
83 pci_name(event->dev));
Linas Vepstas8c33fd12006-03-29 15:29:18 -060084
Linas Vepstasac325ac2006-04-18 21:05:21 -070085 pdn = handle_eeh_events(event);
Linas Vepstas172ca922005-11-03 18:50:04 -060086
Linas Vepstasac325ac2006-04-18 21:05:21 -070087 eeh_clear_slot(event->dn, EEH_MODE_RECOVERING);
88 pci_dev_put(event->dev);
89 kfree(event);
90 mutex_unlock(&eeh_event_mutex);
Linas Vepstas172ca922005-11-03 18:50:04 -060091
Linas Vepstasac325ac2006-04-18 21:05:21 -070092 /* If there are no new errors after an hour, clear the counter. */
93 if (pdn && pdn->eeh_freeze_count>0) {
94 msleep_interruptible (3600*1000);
95 if (pdn->eeh_freeze_count>0)
96 pdn->eeh_freeze_count--;
Linas Vepstas172ca922005-11-03 18:50:04 -060097 }
98
99 return 0;
100}
101
102/**
103 * eeh_thread_launcher
Linas Vepstas172ca922005-11-03 18:50:04 -0600104 * @dummy - unused
105 */
David Howellsc4028952006-11-22 14:57:56 +0000106static void eeh_thread_launcher(struct work_struct *dummy)
Linas Vepstas172ca922005-11-03 18:50:04 -0600107{
108 if (kernel_thread(eeh_event_handler, NULL, CLONE_KERNEL) < 0)
109 printk(KERN_ERR "Failed to start EEH daemon\n");
110}
111
112/**
113 * eeh_send_failure_event - generate a PCI error event
114 * @dev pci device
115 *
116 * This routine can be called within an interrupt context;
117 * the actual event will be delivered in a normal context
118 * (from a workqueue).
119 */
120int eeh_send_failure_event (struct device_node *dn,
Linas Vepstasd0ab95c2007-03-19 14:59:10 -0500121 struct pci_dev *dev)
Linas Vepstas172ca922005-11-03 18:50:04 -0600122{
123 unsigned long flags;
124 struct eeh_event *event;
Jeremy Kerr954a46e2006-07-12 15:39:43 +1000125 const char *location;
Linas Vepstas172ca922005-11-03 18:50:04 -0600126
Linas Vepstas054d8ff2006-04-27 02:31:20 -0700127 if (!mem_init_done) {
128 printk(KERN_ERR "EEH: event during early boot not handled\n");
Stephen Rothwelle2eb6392007-04-03 22:26:41 +1000129 location = of_get_property(dn, "ibm,loc-code", NULL);
Linas Vepstas054d8ff2006-04-27 02:31:20 -0700130 printk(KERN_ERR "EEH: device node = %s\n", dn->full_name);
131 printk(KERN_ERR "EEH: PCI location = %s\n", location);
132 return 1;
133 }
Linas Vepstas172ca922005-11-03 18:50:04 -0600134 event = kmalloc(sizeof(*event), GFP_ATOMIC);
135 if (event == NULL) {
136 printk (KERN_ERR "EEH: out of memory, event not handled\n");
137 return 1;
138 }
139
140 if (dev)
141 pci_dev_get(dev);
142
143 event->dn = dn;
144 event->dev = dev;
Linas Vepstas172ca922005-11-03 18:50:04 -0600145
146 /* We may or may not be called in an interrupt context */
147 spin_lock_irqsave(&eeh_eventlist_lock, flags);
148 list_add(&event->list, &eeh_eventlist);
149 spin_unlock_irqrestore(&eeh_eventlist_lock, flags);
150
151 schedule_work(&eeh_event_wq);
152
153 return 0;
154}
155
156/********************** END OF FILE ******************************/