ipmi_watchdog.c

h内核

	case WDIOC_SET_PRETIMEOUT:
		i = copy_from_user(&val, argp, sizeof(int));
		if (i)
			return -EFAULT;
		pretimeout = val;
		return ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY);

	case WDIOC_GET_PRETIMEOUT:
		i = copy_to_user(argp, &pretimeout, sizeof(pretimeout));
		if (i)
			return -EFAULT;
		return 0;

	case WDIOC_KEEPALIVE:
		return ipmi_heartbeat();

	case WDIOC_SETOPTIONS:
		i = copy_from_user(&val, argp, sizeof(int));
		if (i)
			return -EFAULT;
		if (val & WDIOS_DISABLECARD)
		{
			ipmi_watchdog_state = WDOG_TIMEOUT_NONE;
			ipmi_set_timeout(IPMI_SET_TIMEOUT_NO_HB);
			ipmi_start_timer_on_heartbeat = 0;
		}

		if (val & WDIOS_ENABLECARD)
		{
			ipmi_watchdog_state = action_val;
			ipmi_set_timeout(IPMI_SET_TIMEOUT_FORCE_HB);
		}
		return 0;

	case WDIOC_GETSTATUS:
		val = 0;
		i = copy_to_user(argp, &val, sizeof(val));
		if (i)
			return -EFAULT;
		return 0;

	default:
		return -ENOIOCTLCMD;
	}
}

static ssize_t ipmi_write(struct file *file,
			  const char  __user *buf,
			  size_t      len,
			  loff_t      *ppos)
{
	int rv;

	if (len) {
	    	if (!nowayout) {
		    	size_t i;

			/* In case it was set long ago */
			expect_close = 0;

    			for (i = 0; i != len; i++) {
				char c;

				if (get_user(c, buf + i))
					return -EFAULT;
				if (c == 'V')
					expect_close = 42;
			}
		}
		rv = ipmi_heartbeat();
		if (rv)
			return rv;
		return 1;
	}
	return 0;
}

static ssize_t ipmi_read(struct file *file,
			 char        __user *buf,
			 size_t      count,
			 loff_t      *ppos)
{
	int          rv = 0;
	wait_queue_t wait;

	if (count <= 0)
		return 0;

	/* Reading returns if the pretimeout has gone off, and it only does
	   it once per pretimeout. */
	spin_lock(&ipmi_read_lock);
	if (!data_to_read) {
		if (file->f_flags & O_NONBLOCK) {
			rv = -EAGAIN;
			goto out;
		}
		
		init_waitqueue_entry(&wait, current);
		add_wait_queue(&read_q, &wait);
		while (!data_to_read) {
			set_current_state(TASK_INTERRUPTIBLE);
			spin_unlock(&ipmi_read_lock);
			schedule();
			spin_lock(&ipmi_read_lock);
		}
		remove_wait_queue(&read_q, &wait);
	    
		if (signal_pending(current)) {
			rv = -ERESTARTSYS;
			goto out;
		}
	}
	data_to_read = 0;

 out:
	spin_unlock(&ipmi_read_lock);

	if (rv == 0) {
		if (copy_to_user(buf, &data_to_read, 1))
			rv = -EFAULT;
		else
			rv = 1;
	}

	return rv;
}

static int ipmi_open(struct inode *ino, struct file *filep)
{
        switch (iminor(ino))
        {
                case WATCHDOG_MINOR:
		    if(test_and_set_bit(0, &ipmi_wdog_open))
                        return -EBUSY;

		    /* Don't start the timer now, let it start on the
		       first heartbeat. */
		    ipmi_start_timer_on_heartbeat = 1;
                    return nonseekable_open(ino, filep);

                default:
                    return (-ENODEV);
        }
}

static unsigned int ipmi_poll(struct file *file, poll_table *wait)
{
	unsigned int mask = 0;
	
	poll_wait(file, &read_q, wait);

	spin_lock(&ipmi_read_lock);
	if (data_to_read)
		mask |= (POLLIN | POLLRDNORM);
	spin_unlock(&ipmi_read_lock);

	return mask;
}

static int ipmi_fasync(int fd, struct file *file, int on)
{
	int result;

	result = fasync_helper(fd, file, on, &fasync_q);

	return (result);
}

static int ipmi_close(struct inode *ino, struct file *filep)
{
	if (iminor(ino)==WATCHDOG_MINOR)
	{
		if (expect_close == 42) {
			ipmi_watchdog_state = WDOG_TIMEOUT_NONE;
			ipmi_set_timeout(IPMI_SET_TIMEOUT_NO_HB);
			clear_bit(0, &ipmi_wdog_open);
		} else {
			printk(KERN_CRIT PFX "Unexpected close, not stopping watchdog!\n");
			ipmi_heartbeat();
		}
	}

	ipmi_fasync (-1, filep, 0);
	expect_close = 0;

	return 0;
}

static struct file_operations ipmi_wdog_fops = {
	.owner   = THIS_MODULE,
	.read    = ipmi_read,
	.poll    = ipmi_poll,
	.write   = ipmi_write,
	.ioctl   = ipmi_ioctl,
	.open    = ipmi_open,
	.release = ipmi_close,
	.fasync  = ipmi_fasync,
};

static struct miscdevice ipmi_wdog_miscdev = {
	.minor		= WATCHDOG_MINOR,
	.name		= "watchdog",
	.fops		= &ipmi_wdog_fops
};

static DECLARE_RWSEM(register_sem);

static void ipmi_wdog_msg_handler(struct ipmi_recv_msg *msg,
				  void                 *handler_data)
{
	if (msg->msg.data[0] != 0) {
		printk(KERN_ERR PFX "response: Error %x on cmd %x\n",
		       msg->msg.data[0],
		       msg->msg.cmd);
	}
	
	ipmi_free_recv_msg(msg);
}

static void ipmi_wdog_pretimeout_handler(void *handler_data)
{
	if (preaction_val != WDOG_PRETIMEOUT_NONE) {
		if (preop_val == WDOG_PREOP_PANIC)
			panic("Watchdog pre-timeout");
		else if (preop_val == WDOG_PREOP_GIVE_DATA) {
			spin_lock(&ipmi_read_lock);
			data_to_read = 1;
			wake_up_interruptible(&read_q);
			kill_fasync(&fasync_q, SIGIO, POLL_IN);

			spin_unlock(&ipmi_read_lock);
		}
	}

	/* On some machines, the heartbeat will give
	   an error and not work unless we re-enable
	   the timer.   So do so. */
	pretimeout_since_last_heartbeat = 1;
}

static struct ipmi_user_hndl ipmi_hndlrs =
{
	.ipmi_recv_hndl           = ipmi_wdog_msg_handler,
	.ipmi_watchdog_pretimeout = ipmi_wdog_pretimeout_handler
};

static void ipmi_register_watchdog(int ipmi_intf)
{
	int rv = -EBUSY;

	down_write(&register_sem);
	if (watchdog_user)
		goto out;

	rv = ipmi_create_user(ipmi_intf, &ipmi_hndlrs, NULL, &watchdog_user);
	if (rv < 0) {
		printk(KERN_CRIT PFX "Unable to register with ipmi\n");
		goto out;
	}

	ipmi_get_version(watchdog_user,
			 &ipmi_version_major,
			 &ipmi_version_minor);

	rv = misc_register(&ipmi_wdog_miscdev);
	if (rv < 0) {
		ipmi_destroy_user(watchdog_user);
		watchdog_user = NULL;
		printk(KERN_CRIT PFX "Unable to register misc device\n");
	}

 out:
	up_write(&register_sem);

	if ((start_now) && (rv == 0)) {
		/* Run from startup, so start the timer now. */
		start_now = 0; /* Disable this function after first startup. */
		ipmi_watchdog_state = action_val;
		ipmi_set_timeout(IPMI_SET_TIMEOUT_FORCE_HB);
		printk(KERN_INFO PFX "Starting now!\n");
	}
}

#ifdef HAVE_NMI_HANDLER
static int
ipmi_nmi(void *dev_id, struct pt_regs *regs, int cpu, int handled)
{
	/* If no one else handled the NMI, we assume it was the IPMI
           watchdog. */
	if ((!handled) && (preop_val == WDOG_PREOP_PANIC))
		panic(PFX "pre-timeout");

	/* On some machines, the heartbeat will give
	   an error and not work unless we re-enable
	   the timer.   So do so. */
	pretimeout_since_last_heartbeat = 1;

	return NOTIFY_DONE;
}

static struct nmi_handler ipmi_nmi_handler =
{
	.link     = LIST_HEAD_INIT(ipmi_nmi_handler.link),
	.dev_name = "ipmi_watchdog",
	.dev_id   = NULL,
	.handler  = ipmi_nmi,
	.priority = 0, /* Call us last. */
};
#endif

static int wdog_reboot_handler(struct notifier_block *this,
			       unsigned long         code,
			       void                  *unused)
{
	static int reboot_event_handled = 0;

	if ((watchdog_user) && (!reboot_event_handled)) {
		/* Make sure we only do this once. */
		reboot_event_handled = 1;

		if (code == SYS_DOWN || code == SYS_HALT) {
			/* Disable the WDT if we are shutting down. */
			ipmi_watchdog_state = WDOG_TIMEOUT_NONE;
			panic_halt_ipmi_set_timeout();
		} else {
			/* Set a long timer to let the reboot happens, but
			   reboot if it hangs. */
			timeout = 120;
			pretimeout = 0;
			ipmi_watchdog_state = WDOG_TIMEOUT_RESET;
			panic_halt_ipmi_set_timeout();
		}
	}
	return NOTIFY_OK;
}

static struct notifier_block wdog_reboot_notifier = {
	.notifier_call	= wdog_reboot_handler,
	.next		= NULL,
	.priority	= 0
};

static int wdog_panic_handler(struct notifier_block *this,
			      unsigned long         event,
			      void                  *unused)
{
	static int panic_event_handled = 0;

	/* On a panic, if we have a panic timeout, make sure that the thing
	   reboots, even if it hangs during that panic. */
	if (watchdog_user && !panic_event_handled) {
		/* Make sure the panic doesn't hang, and make sure we
		   do this only once. */
		panic_event_handled = 1;
	    
		timeout = 255;
		pretimeout = 0;
		ipmi_watchdog_state = WDOG_TIMEOUT_RESET;
		panic_halt_ipmi_set_timeout();
	}

	return NOTIFY_OK;
}

static struct notifier_block wdog_panic_notifier = {
	.notifier_call	= wdog_panic_handler,
	.next		= NULL,
	.priority	= 150	/* priority: INT_MAX >= x >= 0 */
};


static void ipmi_new_smi(int if_num)
{
	ipmi_register_watchdog(if_num);
}

static void ipmi_smi_gone(int if_num)
{
	/* This can never be called, because once the watchdog is
	   registered, the interface can't go away until the watchdog
	   is unregistered. */
}

static struct ipmi_smi_watcher smi_watcher =
{
	.owner    = THIS_MODULE,
	.new_smi  = ipmi_new_smi,
	.smi_gone = ipmi_smi_gone
};

static int __init ipmi_wdog_init(void)
{
	int rv;

	printk(KERN_INFO PFX "driver version "
	       IPMI_WATCHDOG_VERSION "\n");

	if (strcmp(action, "reset") == 0) {
		action_val = WDOG_TIMEOUT_RESET;
	} else if (strcmp(action, "none") == 0) {
		action_val = WDOG_TIMEOUT_NONE;
	} else if (strcmp(action, "power_cycle") == 0) {
		action_val = WDOG_TIMEOUT_POWER_CYCLE;
	} else if (strcmp(action, "power_off") == 0) {
		action_val = WDOG_TIMEOUT_POWER_DOWN;
	} else {
		action_val = WDOG_TIMEOUT_RESET;
		printk(KERN_INFO PFX "Unknown action '%s', defaulting to"
		       " reset\n", action);
	}

	if (strcmp(preaction, "pre_none") == 0) {
		preaction_val = WDOG_PRETIMEOUT_NONE;
	} else if (strcmp(preaction, "pre_smi") == 0) {
		preaction_val = WDOG_PRETIMEOUT_SMI;
#ifdef HAVE_NMI_HANDLER
	} else if (strcmp(preaction, "pre_nmi") == 0) {
		preaction_val = WDOG_PRETIMEOUT_NMI;
#endif
	} else if (strcmp(preaction, "pre_int") == 0) {
		preaction_val = WDOG_PRETIMEOUT_MSG_INT;
	} else {
		preaction_val = WDOG_PRETIMEOUT_NONE;
		printk(KERN_INFO PFX "Unknown preaction '%s', defaulting to"
		       " none\n", preaction);
	}

	if (strcmp(preop, "preop_none") == 0) {
		preop_val = WDOG_PREOP_NONE;
	} else if (strcmp(preop, "preop_panic") == 0) {
		preop_val = WDOG_PREOP_PANIC;
	} else if (strcmp(preop, "preop_give_data") == 0) {
		preop_val = WDOG_PREOP_GIVE_DATA;
	} else {
		preop_val = WDOG_PREOP_NONE;
		printk(KERN_INFO PFX "Unknown preop '%s', defaulting to"
		       " none\n", preop);
	}

#ifdef HAVE_NMI_HANDLER
	if (preaction_val == WDOG_PRETIMEOUT_NMI) {
		if (preop_val == WDOG_PREOP_GIVE_DATA) {
			printk(KERN_WARNING PFX "Pretimeout op is to give data"
			       " but NMI pretimeout is enabled, setting"
			       " pretimeout op to none\n");
			preop_val = WDOG_PREOP_NONE;
		}
#ifdef CONFIG_X86_LOCAL_APIC
		if (nmi_watchdog == NMI_IO_APIC) {
			printk(KERN_WARNING PFX "nmi_watchdog is set to IO APIC"
			       " mode (value is %d), that is incompatible"
			       " with using NMI in the IPMI watchdog."
			       " Disabling IPMI nmi pretimeout.\n",
			       nmi_watchdog);
			preaction_val = WDOG_PRETIMEOUT_NONE;
		} else {
#endif
		rv = request_nmi(&ipmi_nmi_handler);
		if (rv) {
			printk(KERN_WARNING PFX "Can't register nmi handler\n");
			return rv;
		}
#ifdef CONFIG_X86_LOCAL_APIC
		}
#endif
	}
#endif

	rv = ipmi_smi_watcher_register(&smi_watcher);
	if (rv) {
#ifdef HAVE_NMI_HANDLER
		if (preaction_val == WDOG_PRETIMEOUT_NMI)
			release_nmi(&ipmi_nmi_handler);
#endif
		printk(KERN_WARNING PFX "can't register smi watcher\n");
		return rv;
	}

	register_reboot_notifier(&wdog_reboot_notifier);
	notifier_chain_register(&panic_notifier_list, &wdog_panic_notifier);

	return 0;
}

static __exit void ipmi_unregister_watchdog(void)
{
	int rv;

	down_write(&register_sem);

#ifdef HAVE_NMI_HANDLER
	if (preaction_val == WDOG_PRETIMEOUT_NMI)
		release_nmi(&ipmi_nmi_handler);
#endif

	notifier_chain_unregister(&panic_notifier_list, &wdog_panic_notifier);
	unregister_reboot_notifier(&wdog_reboot_notifier);

	if (! watchdog_user)
		goto out;

	/* Make sure no one can call us any more. */
	misc_deregister(&ipmi_wdog_miscdev);

	/* Wait to make sure the message makes it out.  The lower layer has
	   pointers to our buffers, we want to make sure they are done before
	   we release our memory. */
	while (atomic_read(&set_timeout_tofree)) {
		set_current_state(TASK_UNINTERRUPTIBLE);
		schedule_timeout(1);
	}

	/* Disconnect from IPMI. */
	rv = ipmi_destroy_user(watchdog_user);
	if (rv) {
		printk(KERN_WARNING PFX "error unlinking from IPMI: %d\n",
		       rv);
	}
	watchdog_user = NULL;

 out:
	up_write(&register_sem);
}

static void __exit ipmi_wdog_exit(void)
{
	ipmi_smi_watcher_unregister(&smi_watcher);
	ipmi_unregister_watchdog();
}
module_exit(ipmi_wdog_exit);
module_init(ipmi_wdog_init);
MODULE_LICENSE("GPL");