arch/powerpc/kernel/eeh_event.c

   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 struct semaphore eeh_eventlist_sem;
  39 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 (down_interruptible(&eeh_eventlist_sem))
  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                         eeh_pe_state_mark(pe, EEH_PE_RECOVERING);
  77                         if (pe->type & EEH_PE_PHB)
  78                                 pr_info("EEH: Detected error on PHB#%d\n",
  79                                          pe->phb->global_number);
  80                         else
  81                                 pr_info("EEH: Detected PCI bus error on "
  82                                         "PHB#%d-PE#%x\n",
  83                                         pe->phb->global_number, pe->addr);
  84                         eeh_handle_event(pe);
  85                         eeh_pe_state_clear(pe, EEH_PE_RECOVERING);
  86                 } else {
  87                         eeh_handle_event(NULL);
  88                 }
  89
  90                 kfree(event);
  91         }
  92
  93         return 0;
  94 }
  95
  96 /**
  97  * eeh_event_init - Start kernel thread to handle EEH events
  98  *
  99  * This routine is called to start the kernel thread for processing
 100  * EEH event.
 101  */
 102 int eeh_event_init(void)
 103 {
 104         struct task_struct *t;
 105         int ret = 0;
 106
 107         /* Initialize semaphore */
 108         sema_init(&eeh_eventlist_sem, 0);
 109
 110         t = kthread_run(eeh_event_handler, NULL, "eehd");
 111         if (IS_ERR(t)) {
 112                 ret = PTR_ERR(t);
 113                 pr_err("%s: Failed to start EEH daemon (%d)\n",
 114                         __func__, ret);
 115                 return ret;
 116         }
 117
 118         return 0;
 119 }
 120
 121 /**
 122  * eeh_send_failure_event - Generate a PCI error event
 123  * @pe: EEH PE
 124  *
 125  * This routine can be called within an interrupt context;
 126  * the actual event will be delivered in a normal context
 127  * (from a workqueue).
 128  */
 129 int eeh_send_failure_event(struct eeh_pe *pe)
 130 {
 131         unsigned long flags;
 132         struct eeh_event *event;
 133
 134         event = kzalloc(sizeof(*event), GFP_ATOMIC);
 135         if (!event) {
 136                 pr_err("EEH: out of memory, event not handled\n");
 137                 return -ENOMEM;
 138         }
 139         event->pe = pe;
 140
 141         /* We may or may not be called in an interrupt context */
 142         spin_lock_irqsave(&eeh_eventlist_lock, flags);
 143         list_add(&event->list, &eeh_eventlist);
 144         spin_unlock_irqrestore(&eeh_eventlist_lock, flags);
 145
 146         /* For EEH deamon to knick in */
 147         up(&eeh_eventlist_sem);
 148
 149         return 0;
 150 }
 151
 152 /**
 153  * eeh_remove_event - Remove EEH event from the queue
 154  * @pe: Event binding to the PE
 155  * @force: Event will be removed unconditionally
 156  *
 157  * On PowerNV platform, we might have subsequent coming events
 158  * is part of the former one. For that case, those subsequent
 159  * coming events are totally duplicated and unnecessary, thus
 160  * they should be removed.
 161  */
 162 void eeh_remove_event(struct eeh_pe *pe, bool force)
 163 {
 164         unsigned long flags;
 165         struct eeh_event *event, *tmp;
 166
 167         /*
 168          * If we have NULL PE passed in, we have dead IOC
 169          * or we're sure we can report all existing errors
 170          * by the caller.
 171          *
 172          * With "force", the event with associated PE that
 173          * have been isolated, the event won't be removed
 174          * to avoid event lost.
 175          */
 176         spin_lock_irqsave(&eeh_eventlist_lock, flags);
 177         list_for_each_entry_safe(event, tmp, &eeh_eventlist, list) {
 178                 if (!force && event->pe &&
 179                     (event->pe->state & EEH_PE_ISOLATED))
 180                         continue;
 181
 182                 if (!pe) {
 183                         list_del(&event->list);
 184                         kfree(event);
 185                 } else if (pe->type & EEH_PE_PHB) {
 186                         if (event->pe && event->pe->phb == pe->phb) {
 187                                 list_del(&event->list);
 188                                 kfree(event);
 189                         }
 190                 } else if (event->pe == pe) {
 191                         list_del(&event->list);
 192                         kfree(event);
 193                 }
 194         }
 195         spin_unlock_irqrestore(&eeh_eventlist_lock, flags);
 196 }