arch/powerpc/platforms/pseries/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/mutex.h>
  22 #include <linux/sched.h>
  23 #include <linux/pci.h>
  24 #include <linux/slab.h>
  25 #include <linux/workqueue.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 /* EEH event workqueue setup. */
  38 static DEFINE_SPINLOCK(eeh_eventlist_lock);
  39 LIST_HEAD(eeh_eventlist);
  40 static void eeh_thread_launcher(struct work_struct *);
  41 DECLARE_WORK(eeh_event_wq, eeh_thread_launcher);
  42
  43 /* Serialize reset sequences for a given pci device */
  44 DEFINE_MUTEX(eeh_event_mutex);
  45
  46 /**
  47  * eeh_event_handler - Dispatch EEH events.
  48  * @dummy - unused
  49  *
  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.
  55  */
  56 static int eeh_event_handler(void * dummy)
  57 {
  58         unsigned long flags;
  59         struct eeh_event *event;
  60         struct eeh_dev *edev;
  61
  62         set_task_comm(current, "eehd");
  63
  64         spin_lock_irqsave(&eeh_eventlist_lock, flags);
  65         event = NULL;
  66
  67         /* Unqueue the event, get ready to process. */
  68         if (!list_empty(&eeh_eventlist)) {
  69                 event = list_entry(eeh_eventlist.next, struct eeh_event, list);
  70                 list_del(&event->list);
  71         }
  72         spin_unlock_irqrestore(&eeh_eventlist_lock, flags);
  73
  74         if (event == NULL)
  75                 return 0;
  76
  77         /* Serialize processing of EEH events */
  78         mutex_lock(&eeh_event_mutex);
  79         edev = event->edev;
  80         eeh_mark_slot(eeh_dev_to_of_node(edev), EEH_MODE_RECOVERING);
  81
  82         printk(KERN_INFO "EEH: Detected PCI bus error on device %s\n",
  83                eeh_pci_name(edev->pdev));
  84
  85         set_current_state(TASK_INTERRUPTIBLE);  /* Don't add to load average */
  86         edev = handle_eeh_events(event);
  87
  88         if (edev) {
  89                 eeh_clear_slot(eeh_dev_to_of_node(edev), EEH_MODE_RECOVERING);
  90                 pci_dev_put(edev->pdev);
  91         }
  92
  93         kfree(event);
  94         mutex_unlock(&eeh_event_mutex);
  95
  96         /* If there are no new errors after an hour, clear the counter. */
  97         if (edev && edev->freeze_count>0) {
  98                 msleep_interruptible(3600*1000);
  99                 if (edev->freeze_count>0)
 100                         edev->freeze_count--;
 101
 102         }
 103
 104         return 0;
 105 }
 106
 107 /**
 108  * eeh_thread_launcher - Start kernel thread to handle EEH events
 109  * @dummy - unused
 110  *
 111  * This routine is called to start the kernel thread for processing
 112  * EEH event.
 113  */
 114 static void eeh_thread_launcher(struct work_struct *dummy)
 115 {
 116         if (kernel_thread(eeh_event_handler, NULL, CLONE_KERNEL) < 0)
 117                 printk(KERN_ERR "Failed to start EEH daemon\n");
 118 }
 119
 120 /**
 121  * eeh_send_failure_event - Generate a PCI error event
 122  * @edev: EEH device
 123  *
 124  * This routine can be called within an interrupt context;
 125  * the actual event will be delivered in a normal context
 126  * (from a workqueue).
 127  */
 128 int eeh_send_failure_event(struct eeh_dev *edev)
 129 {
 130         unsigned long flags;
 131         struct eeh_event *event;
 132         struct device_node *dn = eeh_dev_to_of_node(edev);
 133         const char *location;
 134
 135         if (!mem_init_done) {
 136                 printk(KERN_ERR "EEH: event during early boot not handled\n");
 137                 location = of_get_property(dn, "ibm,loc-code", NULL);
 138                 printk(KERN_ERR "EEH: device node = %s\n", dn->full_name);
 139                 printk(KERN_ERR "EEH: PCI location = %s\n", location);
 140                 return 1;
 141         }
 142         event = kmalloc(sizeof(*event), GFP_ATOMIC);
 143         if (event == NULL) {
 144                 printk(KERN_ERR "EEH: out of memory, event not handled\n");
 145                 return 1;
 146         }
 147
 148         if (edev->pdev)
 149                 pci_dev_get(edev->pdev);
 150
 151         event->edev = edev;
 152
 153         /* We may or may not be called in an interrupt context */
 154         spin_lock_irqsave(&eeh_eventlist_lock, flags);
 155         list_add(&event->list, &eeh_eventlist);
 156         spin_unlock_irqrestore(&eeh_eventlist_lock, flags);
 157
 158         schedule_work(&eeh_event_wq);
 159
 160         return 0;
 161 }