1 /* 2 * This program is free software; you can redistribute it and/or modify 3 * it under the terms of the GNU General Public License as published by 4 * the Free Software Foundation; either version 2 of the License, or 5 * (at your option) any later version. 6 * 7 * This program is distributed in the hope that it will be useful, 8 * but WITHOUT ANY WARRANTY; without even the implied warranty of 9 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 10 * GNU General Public License for more details. 11 * 12 * You should have received a copy of the GNU General Public License 13 * along with this program; if not, write to the Free Software 14 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA 15 * 16 * Copyright (c) 2005 Linas Vepstas <linas@linas.org> 17 */ 18 19 #include <linux/delay.h> 20 #include <linux/list.h> 21 #include <linux/sched.h> 22 #include <linux/semaphore.h> 23 #include <linux/pci.h> 24 #include <linux/slab.h> 25 #include <linux/kthread.h> 26 #include <asm/eeh_event.h> 27 #include <asm/ppc-pci.h> 28 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 static DEFINE_SPINLOCK(eeh_eventlist_lock); 38 static DECLARE_COMPLETION(eeh_eventlist_event); 39 static LIST_HEAD(eeh_eventlist); 40 41 /** 42 * eeh_event_handler - Dispatch EEH events. 43 * @dummy - unused 44 * 45 * The detection of a frozen slot can occur inside an interrupt, 46 * where it can be hard to do anything about it. The goal of this 47 * routine is to pull these detection events out of the context 48 * of the interrupt handler, and re-dispatch them for processing 49 * at a later time in a normal context. 50 */ 51 static int eeh_event_handler(void * dummy) 52 { 53 unsigned long flags; 54 struct eeh_event *event; 55 struct eeh_pe *pe; 56 57 while (!kthread_should_stop()) { 58 if (wait_for_completion_interruptible(&eeh_eventlist_event)) 59 break; 60 61 /* Fetch EEH event from the queue */ 62 spin_lock_irqsave(&eeh_eventlist_lock, flags); 63 event = NULL; 64 if (!list_empty(&eeh_eventlist)) { 65 event = list_entry(eeh_eventlist.next, 66 struct eeh_event, list); 67 list_del(&event->list); 68 } 69 spin_unlock_irqrestore(&eeh_eventlist_lock, flags); 70 if (!event) 71 continue; 72 73 /* We might have event without binding PE */ 74 pe = event->pe; 75 if (pe) { 76 if (pe->type & EEH_PE_PHB) 77 pr_info("EEH: Detected error on PHB#%x\n", 78 pe->phb->global_number); 79 else 80 pr_info("EEH: Detected PCI bus error on " 81 "PHB#%x-PE#%x\n", 82 pe->phb->global_number, pe->addr); 83 eeh_handle_normal_event(pe); 84 } else { 85 eeh_handle_special_event(); 86 } 87 88 kfree(event); 89 } 90 91 return 0; 92 } 93 94 /** 95 * eeh_event_init - Start kernel thread to handle EEH events 96 * 97 * This routine is called to start the kernel thread for processing 98 * EEH event. 99 */ 100 int eeh_event_init(void) 101 { 102 struct task_struct *t; 103 int ret = 0; 104 105 t = kthread_run(eeh_event_handler, NULL, "eehd"); 106 if (IS_ERR(t)) { 107 ret = PTR_ERR(t); 108 pr_err("%s: Failed to start EEH daemon (%d)\n", 109 __func__, ret); 110 return ret; 111 } 112 113 return 0; 114 } 115 116 /** 117 * eeh_send_failure_event - Generate a PCI error event 118 * @pe: EEH PE 119 * 120 * This routine can be called within an interrupt context; 121 * the actual event will be delivered in a normal context 122 * (from a workqueue). 123 */ 124 int __eeh_send_failure_event(struct eeh_pe *pe) 125 { 126 unsigned long flags; 127 struct eeh_event *event; 128 129 event = kzalloc(sizeof(*event), GFP_ATOMIC); 130 if (!event) { 131 pr_err("EEH: out of memory, event not handled\n"); 132 return -ENOMEM; 133 } 134 event->pe = pe; 135 136 /* We may or may not be called in an interrupt context */ 137 spin_lock_irqsave(&eeh_eventlist_lock, flags); 138 list_add(&event->list, &eeh_eventlist); 139 spin_unlock_irqrestore(&eeh_eventlist_lock, flags); 140 141 /* For EEH deamon to knick in */ 142 complete(&eeh_eventlist_event); 143 144 return 0; 145 } 146 147 int eeh_send_failure_event(struct eeh_pe *pe) 148 { 149 /* 150 * If we've manually supressed recovery events via debugfs 151 * then just drop it on the floor. 152 */ 153 if (eeh_debugfs_no_recover) { 154 pr_err("EEH: Event dropped due to no_recover setting\n"); 155 return 0; 156 } 157 158 return __eeh_send_failure_event(pe); 159 } 160 161 /** 162 * eeh_remove_event - Remove EEH event from the queue 163 * @pe: Event binding to the PE 164 * @force: Event will be removed unconditionally 165 * 166 * On PowerNV platform, we might have subsequent coming events 167 * is part of the former one. For that case, those subsequent 168 * coming events are totally duplicated and unnecessary, thus 169 * they should be removed. 170 */ 171 void eeh_remove_event(struct eeh_pe *pe, bool force) 172 { 173 unsigned long flags; 174 struct eeh_event *event, *tmp; 175 176 /* 177 * If we have NULL PE passed in, we have dead IOC 178 * or we're sure we can report all existing errors 179 * by the caller. 180 * 181 * With "force", the event with associated PE that 182 * have been isolated, the event won't be removed 183 * to avoid event lost. 184 */ 185 spin_lock_irqsave(&eeh_eventlist_lock, flags); 186 list_for_each_entry_safe(event, tmp, &eeh_eventlist, list) { 187 if (!force && event->pe && 188 (event->pe->state & EEH_PE_ISOLATED)) 189 continue; 190 191 if (!pe) { 192 list_del(&event->list); 193 kfree(event); 194 } else if (pe->type & EEH_PE_PHB) { 195 if (event->pe && event->pe->phb == pe->phb) { 196 list_del(&event->list); 197 kfree(event); 198 } 199 } else if (event->pe == pe) { 200 list_del(&event->list); 201 kfree(event); 202 } 203 } 204 spin_unlock_irqrestore(&eeh_eventlist_lock, flags); 205 } 206