4 * A watchdog timer based upon the IPMI interface.
6 * Author: MontaVista Software, Inc.
7 * Corey Minyard <minyard@mvista.com>
10 * Copyright 2002 MontaVista Software Inc.
12 * This program is free software; you can redistribute it and/or modify it
13 * under the terms of the GNU General Public License as published by the
14 * Free Software Foundation; either version 2 of the License, or (at your
15 * option) any later version.
18 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
19 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
20 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
23 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
24 * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
25 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
26 * TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
27 * USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29 * You should have received a copy of the GNU General Public License along
30 * with this program; if not, write to the Free Software Foundation, Inc.,
31 * 675 Mass Ave, Cambridge, MA 02139, USA.
34 #include <linux/module.h>
35 #include <linux/moduleparam.h>
36 #include <linux/ipmi.h>
37 #include <linux/ipmi_smi.h>
38 #include <linux/watchdog.h>
39 #include <linux/miscdevice.h>
40 #include <linux/init.h>
41 #include <linux/completion.h>
42 #include <linux/kdebug.h>
43 #include <linux/rwsem.h>
44 #include <linux/errno.h>
45 #include <asm/uaccess.h>
46 #include <linux/notifier.h>
47 #include <linux/nmi.h>
48 #include <linux/reboot.h>
49 #include <linux/wait.h>
50 #include <linux/poll.h>
51 #include <linux/string.h>
52 #include <linux/ctype.h>
53 #include <linux/delay.h>
54 #include <asm/atomic.h>
58 * This is ugly, but I've determined that x86 is the only architecture
59 * that can reasonably support the IPMI NMI watchdog timeout at this
60 * time. If another architecture adds this capability somehow, it
61 * will have to be a somewhat different mechanism and I have no idea
62 * how it will work. So in the unlikely event that another
63 * architecture supports this, we can figure out a good generic
64 * mechanism for it at that time.
66 #include <asm/kdebug.h>
70 #define PFX "IPMI Watchdog: "
73 * The IPMI command/response information for the watchdog timer.
76 /* values for byte 1 of the set command, byte 2 of the get response. */
77 #define WDOG_DONT_LOG (1 << 7)
78 #define WDOG_DONT_STOP_ON_SET (1 << 6)
79 #define WDOG_SET_TIMER_USE(byte, use) \
80 byte = ((byte) & 0xf8) | ((use) & 0x7)
81 #define WDOG_GET_TIMER_USE(byte) ((byte) & 0x7)
82 #define WDOG_TIMER_USE_BIOS_FRB2 1
83 #define WDOG_TIMER_USE_BIOS_POST 2
84 #define WDOG_TIMER_USE_OS_LOAD 3
85 #define WDOG_TIMER_USE_SMS_OS 4
86 #define WDOG_TIMER_USE_OEM 5
88 /* values for byte 2 of the set command, byte 3 of the get response. */
89 #define WDOG_SET_PRETIMEOUT_ACT(byte, use) \
90 byte = ((byte) & 0x8f) | (((use) & 0x7) << 4)
91 #define WDOG_GET_PRETIMEOUT_ACT(byte) (((byte) >> 4) & 0x7)
92 #define WDOG_PRETIMEOUT_NONE 0
93 #define WDOG_PRETIMEOUT_SMI 1
94 #define WDOG_PRETIMEOUT_NMI 2
95 #define WDOG_PRETIMEOUT_MSG_INT 3
97 /* Operations that can be performed on a pretimout. */
98 #define WDOG_PREOP_NONE 0
99 #define WDOG_PREOP_PANIC 1
100 /* Cause data to be available to read. Doesn't work in NMI mode. */
101 #define WDOG_PREOP_GIVE_DATA 2
103 /* Actions to perform on a full timeout. */
104 #define WDOG_SET_TIMEOUT_ACT(byte, use) \
105 byte = ((byte) & 0xf8) | ((use) & 0x7)
106 #define WDOG_GET_TIMEOUT_ACT(byte) ((byte) & 0x7)
107 #define WDOG_TIMEOUT_NONE 0
108 #define WDOG_TIMEOUT_RESET 1
109 #define WDOG_TIMEOUT_POWER_DOWN 2
110 #define WDOG_TIMEOUT_POWER_CYCLE 3
113 * Byte 3 of the get command, byte 4 of the get response is the
114 * pre-timeout in seconds.
117 /* Bits for setting byte 4 of the set command, byte 5 of the get response. */
118 #define WDOG_EXPIRE_CLEAR_BIOS_FRB2 (1 << 1)
119 #define WDOG_EXPIRE_CLEAR_BIOS_POST (1 << 2)
120 #define WDOG_EXPIRE_CLEAR_OS_LOAD (1 << 3)
121 #define WDOG_EXPIRE_CLEAR_SMS_OS (1 << 4)
122 #define WDOG_EXPIRE_CLEAR_OEM (1 << 5)
125 * Setting/getting the watchdog timer value. This is for bytes 5 and
126 * 6 (the timeout time) of the set command, and bytes 6 and 7 (the
127 * timeout time) and 8 and 9 (the current countdown value) of the
128 * response. The timeout value is given in seconds (in the command it
129 * is 100ms intervals).
131 #define WDOG_SET_TIMEOUT(byte1, byte2, val) \
132 (byte1) = (((val) * 10) & 0xff), (byte2) = (((val) * 10) >> 8)
133 #define WDOG_GET_TIMEOUT(byte1, byte2) \
134 (((byte1) | ((byte2) << 8)) / 10)
136 #define IPMI_WDOG_RESET_TIMER 0x22
137 #define IPMI_WDOG_SET_TIMER 0x24
138 #define IPMI_WDOG_GET_TIMER 0x25
140 /* These are here until the real ones get into the watchdog.h interface. */
141 #ifndef WDIOC_GETTIMEOUT
142 #define WDIOC_GETTIMEOUT _IOW(WATCHDOG_IOCTL_BASE, 20, int)
144 #ifndef WDIOC_SET_PRETIMEOUT
145 #define WDIOC_SET_PRETIMEOUT _IOW(WATCHDOG_IOCTL_BASE, 21, int)
147 #ifndef WDIOC_GET_PRETIMEOUT
148 #define WDIOC_GET_PRETIMEOUT _IOW(WATCHDOG_IOCTL_BASE, 22, int)
151 static int nowayout
= WATCHDOG_NOWAYOUT
;
153 static ipmi_user_t watchdog_user
;
154 static int watchdog_ifnum
;
156 /* Default the timeout to 10 seconds. */
157 static int timeout
= 10;
159 /* The pre-timeout is disabled by default. */
160 static int pretimeout
;
162 /* Default action is to reset the board on a timeout. */
163 static unsigned char action_val
= WDOG_TIMEOUT_RESET
;
165 static char action
[16] = "reset";
167 static unsigned char preaction_val
= WDOG_PRETIMEOUT_NONE
;
169 static char preaction
[16] = "pre_none";
171 static unsigned char preop_val
= WDOG_PREOP_NONE
;
173 static char preop
[16] = "preop_none";
174 static DEFINE_SPINLOCK(ipmi_read_lock
);
175 static char data_to_read
;
176 static DECLARE_WAIT_QUEUE_HEAD(read_q
);
177 static struct fasync_struct
*fasync_q
;
178 static char pretimeout_since_last_heartbeat
;
179 static char expect_close
;
181 static int ifnum_to_use
= -1;
183 /* Parameters to ipmi_set_timeout */
184 #define IPMI_SET_TIMEOUT_NO_HB 0
185 #define IPMI_SET_TIMEOUT_HB_IF_NECESSARY 1
186 #define IPMI_SET_TIMEOUT_FORCE_HB 2
188 static int ipmi_set_timeout(int do_heartbeat
);
189 static void ipmi_register_watchdog(int ipmi_intf
);
190 static void ipmi_unregister_watchdog(int ipmi_intf
);
193 * If true, the driver will start running as soon as it is configured
196 static int start_now
;
198 static int set_param_int(const char *val
, struct kernel_param
*kp
)
206 l
= simple_strtoul(val
, &endp
, 0);
210 *((int *)kp
->arg
) = l
;
212 rv
= ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY
);
217 static int get_param_int(char *buffer
, struct kernel_param
*kp
)
219 return sprintf(buffer
, "%i", *((int *)kp
->arg
));
222 typedef int (*action_fn
)(const char *intval
, char *outval
);
224 static int action_op(const char *inval
, char *outval
);
225 static int preaction_op(const char *inval
, char *outval
);
226 static int preop_op(const char *inval
, char *outval
);
227 static void check_parms(void);
229 static int set_param_str(const char *val
, struct kernel_param
*kp
)
231 action_fn fn
= (action_fn
) kp
->arg
;
236 strncpy(valcp
, val
, 16);
247 rv
= ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY
);
253 static int get_param_str(char *buffer
, struct kernel_param
*kp
)
255 action_fn fn
= (action_fn
) kp
->arg
;
258 rv
= fn(NULL
, buffer
);
261 return strlen(buffer
);
265 static int set_param_wdog_ifnum(const char *val
, struct kernel_param
*kp
)
267 int rv
= param_set_int(val
, kp
);
270 if ((ifnum_to_use
< 0) || (ifnum_to_use
== watchdog_ifnum
))
273 ipmi_unregister_watchdog(watchdog_ifnum
);
274 ipmi_register_watchdog(ifnum_to_use
);
278 module_param_call(ifnum_to_use
, set_param_wdog_ifnum
, get_param_int
,
279 &ifnum_to_use
, 0644);
280 MODULE_PARM_DESC(ifnum_to_use
, "The interface number to use for the watchdog "
281 "timer. Setting to -1 defaults to the first registered "
284 module_param_call(timeout
, set_param_int
, get_param_int
, &timeout
, 0644);
285 MODULE_PARM_DESC(timeout
, "Timeout value in seconds.");
287 module_param_call(pretimeout
, set_param_int
, get_param_int
, &pretimeout
, 0644);
288 MODULE_PARM_DESC(pretimeout
, "Pretimeout value in seconds.");
290 module_param_call(action
, set_param_str
, get_param_str
, action_op
, 0644);
291 MODULE_PARM_DESC(action
, "Timeout action. One of: "
292 "reset, none, power_cycle, power_off.");
294 module_param_call(preaction
, set_param_str
, get_param_str
, preaction_op
, 0644);
295 MODULE_PARM_DESC(preaction
, "Pretimeout action. One of: "
296 "pre_none, pre_smi, pre_nmi, pre_int.");
298 module_param_call(preop
, set_param_str
, get_param_str
, preop_op
, 0644);
299 MODULE_PARM_DESC(preop
, "Pretimeout driver operation. One of: "
300 "preop_none, preop_panic, preop_give_data.");
302 module_param(start_now
, int, 0444);
303 MODULE_PARM_DESC(start_now
, "Set to 1 to start the watchdog as"
304 "soon as the driver is loaded.");
306 module_param(nowayout
, int, 0644);
307 MODULE_PARM_DESC(nowayout
, "Watchdog cannot be stopped once started "
308 "(default=CONFIG_WATCHDOG_NOWAYOUT)");
310 /* Default state of the timer. */
311 static unsigned char ipmi_watchdog_state
= WDOG_TIMEOUT_NONE
;
313 /* If shutting down via IPMI, we ignore the heartbeat. */
314 static int ipmi_ignore_heartbeat
;
316 /* Is someone using the watchdog? Only one user is allowed. */
317 static unsigned long ipmi_wdog_open
;
320 * If set to 1, the heartbeat command will set the state to reset and
321 * start the timer. The timer doesn't normally run when the driver is
322 * first opened until the heartbeat is set the first time, this
323 * variable is used to accomplish this.
325 static int ipmi_start_timer_on_heartbeat
;
327 /* IPMI version of the BMC. */
328 static unsigned char ipmi_version_major
;
329 static unsigned char ipmi_version_minor
;
331 /* If a pretimeout occurs, this is used to allow only one panic to happen. */
332 static atomic_t preop_panic_excl
= ATOMIC_INIT(-1);
335 static int testing_nmi
;
336 static int nmi_handler_registered
;
339 static int ipmi_heartbeat(void);
342 * We use a mutex to make sure that only one thing can send a set
343 * timeout at one time, because we only have one copy of the data.
344 * The mutex is claimed when the set_timeout is sent and freed
345 * when both messages are free.
347 static atomic_t set_timeout_tofree
= ATOMIC_INIT(0);
348 static DEFINE_MUTEX(set_timeout_lock
);
349 static DECLARE_COMPLETION(set_timeout_wait
);
350 static void set_timeout_free_smi(struct ipmi_smi_msg
*msg
)
352 if (atomic_dec_and_test(&set_timeout_tofree
))
353 complete(&set_timeout_wait
);
355 static void set_timeout_free_recv(struct ipmi_recv_msg
*msg
)
357 if (atomic_dec_and_test(&set_timeout_tofree
))
358 complete(&set_timeout_wait
);
360 static struct ipmi_smi_msg set_timeout_smi_msg
= {
361 .done
= set_timeout_free_smi
363 static struct ipmi_recv_msg set_timeout_recv_msg
= {
364 .done
= set_timeout_free_recv
367 static int i_ipmi_set_timeout(struct ipmi_smi_msg
*smi_msg
,
368 struct ipmi_recv_msg
*recv_msg
,
369 int *send_heartbeat_now
)
371 struct kernel_ipmi_msg msg
;
372 unsigned char data
[6];
374 struct ipmi_system_interface_addr addr
;
378 /* These can be cleared as we are setting the timeout. */
379 pretimeout_since_last_heartbeat
= 0;
382 WDOG_SET_TIMER_USE(data
[0], WDOG_TIMER_USE_SMS_OS
);
384 if ((ipmi_version_major
> 1)
385 || ((ipmi_version_major
== 1) && (ipmi_version_minor
>= 5))) {
386 /* This is an IPMI 1.5-only feature. */
387 data
[0] |= WDOG_DONT_STOP_ON_SET
;
388 } else if (ipmi_watchdog_state
!= WDOG_TIMEOUT_NONE
) {
390 * In ipmi 1.0, setting the timer stops the watchdog, we
391 * need to start it back up again.
397 WDOG_SET_TIMEOUT_ACT(data
[1], ipmi_watchdog_state
);
398 if ((pretimeout
> 0) && (ipmi_watchdog_state
!= WDOG_TIMEOUT_NONE
)) {
399 WDOG_SET_PRETIMEOUT_ACT(data
[1], preaction_val
);
400 data
[2] = pretimeout
;
402 WDOG_SET_PRETIMEOUT_ACT(data
[1], WDOG_PRETIMEOUT_NONE
);
403 data
[2] = 0; /* No pretimeout. */
406 WDOG_SET_TIMEOUT(data
[4], data
[5], timeout
);
408 addr
.addr_type
= IPMI_SYSTEM_INTERFACE_ADDR_TYPE
;
409 addr
.channel
= IPMI_BMC_CHANNEL
;
413 msg
.cmd
= IPMI_WDOG_SET_TIMER
;
415 msg
.data_len
= sizeof(data
);
416 rv
= ipmi_request_supply_msgs(watchdog_user
,
417 (struct ipmi_addr
*) &addr
,
425 printk(KERN_WARNING PFX
"set timeout error: %d\n",
429 if (send_heartbeat_now
)
430 *send_heartbeat_now
= hbnow
;
435 static int ipmi_set_timeout(int do_heartbeat
)
437 int send_heartbeat_now
;
441 /* We can only send one of these at a time. */
442 mutex_lock(&set_timeout_lock
);
444 atomic_set(&set_timeout_tofree
, 2);
446 rv
= i_ipmi_set_timeout(&set_timeout_smi_msg
,
447 &set_timeout_recv_msg
,
448 &send_heartbeat_now
);
450 mutex_unlock(&set_timeout_lock
);
454 wait_for_completion(&set_timeout_wait
);
456 mutex_unlock(&set_timeout_lock
);
458 if ((do_heartbeat
== IPMI_SET_TIMEOUT_FORCE_HB
)
459 || ((send_heartbeat_now
)
460 && (do_heartbeat
== IPMI_SET_TIMEOUT_HB_IF_NECESSARY
)))
461 rv
= ipmi_heartbeat();
467 static atomic_t panic_done_count
= ATOMIC_INIT(0);
469 static void panic_smi_free(struct ipmi_smi_msg
*msg
)
471 atomic_dec(&panic_done_count
);
473 static void panic_recv_free(struct ipmi_recv_msg
*msg
)
475 atomic_dec(&panic_done_count
);
478 static struct ipmi_smi_msg panic_halt_heartbeat_smi_msg
= {
479 .done
= panic_smi_free
481 static struct ipmi_recv_msg panic_halt_heartbeat_recv_msg
= {
482 .done
= panic_recv_free
485 static void panic_halt_ipmi_heartbeat(void)
487 struct kernel_ipmi_msg msg
;
488 struct ipmi_system_interface_addr addr
;
492 * Don't reset the timer if we have the timer turned off, that
493 * re-enables the watchdog.
495 if (ipmi_watchdog_state
== WDOG_TIMEOUT_NONE
)
498 addr
.addr_type
= IPMI_SYSTEM_INTERFACE_ADDR_TYPE
;
499 addr
.channel
= IPMI_BMC_CHANNEL
;
503 msg
.cmd
= IPMI_WDOG_RESET_TIMER
;
506 rv
= ipmi_request_supply_msgs(watchdog_user
,
507 (struct ipmi_addr
*) &addr
,
511 &panic_halt_heartbeat_smi_msg
,
512 &panic_halt_heartbeat_recv_msg
,
515 atomic_add(2, &panic_done_count
);
518 static struct ipmi_smi_msg panic_halt_smi_msg
= {
519 .done
= panic_smi_free
521 static struct ipmi_recv_msg panic_halt_recv_msg
= {
522 .done
= panic_recv_free
526 * Special call, doesn't claim any locks. This is only to be called
527 * at panic or halt time, in run-to-completion mode, when the caller
528 * is the only CPU and the only thing that will be going is these IPMI
531 static void panic_halt_ipmi_set_timeout(void)
533 int send_heartbeat_now
;
536 /* Wait for the messages to be free. */
537 while (atomic_read(&panic_done_count
) != 0)
538 ipmi_poll_interface(watchdog_user
);
539 rv
= i_ipmi_set_timeout(&panic_halt_smi_msg
,
540 &panic_halt_recv_msg
,
541 &send_heartbeat_now
);
543 atomic_add(2, &panic_done_count
);
544 if (send_heartbeat_now
)
545 panic_halt_ipmi_heartbeat();
547 printk(KERN_WARNING PFX
548 "Unable to extend the watchdog timeout.");
549 while (atomic_read(&panic_done_count
) != 0)
550 ipmi_poll_interface(watchdog_user
);
554 * We use a mutex to make sure that only one thing can send a
555 * heartbeat at one time, because we only have one copy of the data.
556 * The semaphore is claimed when the set_timeout is sent and freed
557 * when both messages are free.
559 static atomic_t heartbeat_tofree
= ATOMIC_INIT(0);
560 static DEFINE_MUTEX(heartbeat_lock
);
561 static DECLARE_COMPLETION(heartbeat_wait
);
562 static void heartbeat_free_smi(struct ipmi_smi_msg
*msg
)
564 if (atomic_dec_and_test(&heartbeat_tofree
))
565 complete(&heartbeat_wait
);
567 static void heartbeat_free_recv(struct ipmi_recv_msg
*msg
)
569 if (atomic_dec_and_test(&heartbeat_tofree
))
570 complete(&heartbeat_wait
);
572 static struct ipmi_smi_msg heartbeat_smi_msg
= {
573 .done
= heartbeat_free_smi
575 static struct ipmi_recv_msg heartbeat_recv_msg
= {
576 .done
= heartbeat_free_recv
579 static int ipmi_heartbeat(void)
581 struct kernel_ipmi_msg msg
;
583 struct ipmi_system_interface_addr addr
;
585 if (ipmi_ignore_heartbeat
)
588 if (ipmi_start_timer_on_heartbeat
) {
589 ipmi_start_timer_on_heartbeat
= 0;
590 ipmi_watchdog_state
= action_val
;
591 return ipmi_set_timeout(IPMI_SET_TIMEOUT_FORCE_HB
);
592 } else if (pretimeout_since_last_heartbeat
) {
594 * A pretimeout occurred, make sure we set the timeout.
595 * We don't want to set the action, though, we want to
596 * leave that alone (thus it can't be combined with the
599 return ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY
);
602 mutex_lock(&heartbeat_lock
);
604 atomic_set(&heartbeat_tofree
, 2);
607 * Don't reset the timer if we have the timer turned off, that
608 * re-enables the watchdog.
610 if (ipmi_watchdog_state
== WDOG_TIMEOUT_NONE
) {
611 mutex_unlock(&heartbeat_lock
);
615 addr
.addr_type
= IPMI_SYSTEM_INTERFACE_ADDR_TYPE
;
616 addr
.channel
= IPMI_BMC_CHANNEL
;
620 msg
.cmd
= IPMI_WDOG_RESET_TIMER
;
623 rv
= ipmi_request_supply_msgs(watchdog_user
,
624 (struct ipmi_addr
*) &addr
,
632 mutex_unlock(&heartbeat_lock
);
633 printk(KERN_WARNING PFX
"heartbeat failure: %d\n",
638 /* Wait for the heartbeat to be sent. */
639 wait_for_completion(&heartbeat_wait
);
641 if (heartbeat_recv_msg
.msg
.data
[0] != 0) {
643 * Got an error in the heartbeat response. It was already
644 * reported in ipmi_wdog_msg_handler, but we should return
650 mutex_unlock(&heartbeat_lock
);
655 static struct watchdog_info ident
= {
656 .options
= 0, /* WDIOF_SETTIMEOUT, */
657 .firmware_version
= 1,
661 static int ipmi_ioctl(struct inode
*inode
, struct file
*file
,
662 unsigned int cmd
, unsigned long arg
)
664 void __user
*argp
= (void __user
*)arg
;
669 case WDIOC_GETSUPPORT
:
670 i
= copy_to_user(argp
, &ident
, sizeof(ident
));
671 return i
? -EFAULT
: 0;
673 case WDIOC_SETTIMEOUT
:
674 i
= copy_from_user(&val
, argp
, sizeof(int));
678 return ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY
);
680 case WDIOC_GETTIMEOUT
:
681 i
= copy_to_user(argp
, &timeout
, sizeof(timeout
));
686 case WDIOC_SET_PRETIMEOUT
:
687 case WDIOC_SETPRETIMEOUT
:
688 i
= copy_from_user(&val
, argp
, sizeof(int));
692 return ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY
);
694 case WDIOC_GET_PRETIMEOUT
:
695 case WDIOC_GETPRETIMEOUT
:
696 i
= copy_to_user(argp
, &pretimeout
, sizeof(pretimeout
));
701 case WDIOC_KEEPALIVE
:
702 return ipmi_heartbeat();
704 case WDIOC_SETOPTIONS
:
705 i
= copy_from_user(&val
, argp
, sizeof(int));
708 if (val
& WDIOS_DISABLECARD
) {
709 ipmi_watchdog_state
= WDOG_TIMEOUT_NONE
;
710 ipmi_set_timeout(IPMI_SET_TIMEOUT_NO_HB
);
711 ipmi_start_timer_on_heartbeat
= 0;
714 if (val
& WDIOS_ENABLECARD
) {
715 ipmi_watchdog_state
= action_val
;
716 ipmi_set_timeout(IPMI_SET_TIMEOUT_FORCE_HB
);
720 case WDIOC_GETSTATUS
:
722 i
= copy_to_user(argp
, &val
, sizeof(val
));
732 static ssize_t
ipmi_write(struct file
*file
,
733 const char __user
*buf
,
743 /* In case it was set long ago */
746 for (i
= 0; i
!= len
; i
++) {
749 if (get_user(c
, buf
+ i
))
755 rv
= ipmi_heartbeat();
763 static ssize_t
ipmi_read(struct file
*file
,
775 * Reading returns if the pretimeout has gone off, and it only does
776 * it once per pretimeout.
778 spin_lock(&ipmi_read_lock
);
780 if (file
->f_flags
& O_NONBLOCK
) {
785 init_waitqueue_entry(&wait
, current
);
786 add_wait_queue(&read_q
, &wait
);
787 while (!data_to_read
) {
788 set_current_state(TASK_INTERRUPTIBLE
);
789 spin_unlock(&ipmi_read_lock
);
791 spin_lock(&ipmi_read_lock
);
793 remove_wait_queue(&read_q
, &wait
);
795 if (signal_pending(current
)) {
803 spin_unlock(&ipmi_read_lock
);
806 if (copy_to_user(buf
, &data_to_read
, 1))
815 static int ipmi_open(struct inode
*ino
, struct file
*filep
)
817 switch (iminor(ino
)) {
819 if (test_and_set_bit(0, &ipmi_wdog_open
))
823 * Don't start the timer now, let it start on the
826 ipmi_start_timer_on_heartbeat
= 1;
827 return nonseekable_open(ino
, filep
);
834 static unsigned int ipmi_poll(struct file
*file
, poll_table
*wait
)
836 unsigned int mask
= 0;
838 poll_wait(file
, &read_q
, wait
);
840 spin_lock(&ipmi_read_lock
);
842 mask
|= (POLLIN
| POLLRDNORM
);
843 spin_unlock(&ipmi_read_lock
);
848 static int ipmi_fasync(int fd
, struct file
*file
, int on
)
852 result
= fasync_helper(fd
, file
, on
, &fasync_q
);
857 static int ipmi_close(struct inode
*ino
, struct file
*filep
)
859 if (iminor(ino
) == WATCHDOG_MINOR
) {
860 if (expect_close
== 42) {
861 ipmi_watchdog_state
= WDOG_TIMEOUT_NONE
;
862 ipmi_set_timeout(IPMI_SET_TIMEOUT_NO_HB
);
865 "Unexpected close, not stopping watchdog!\n");
868 clear_bit(0, &ipmi_wdog_open
);
871 ipmi_fasync(-1, filep
, 0);
877 static const struct file_operations ipmi_wdog_fops
= {
878 .owner
= THIS_MODULE
,
884 .release
= ipmi_close
,
885 .fasync
= ipmi_fasync
,
888 static struct miscdevice ipmi_wdog_miscdev
= {
889 .minor
= WATCHDOG_MINOR
,
891 .fops
= &ipmi_wdog_fops
894 static void ipmi_wdog_msg_handler(struct ipmi_recv_msg
*msg
,
897 if (msg
->msg
.data
[0] != 0) {
898 printk(KERN_ERR PFX
"response: Error %x on cmd %x\n",
903 ipmi_free_recv_msg(msg
);
906 static void ipmi_wdog_pretimeout_handler(void *handler_data
)
908 if (preaction_val
!= WDOG_PRETIMEOUT_NONE
) {
909 if (preop_val
== WDOG_PREOP_PANIC
) {
910 if (atomic_inc_and_test(&preop_panic_excl
))
911 panic("Watchdog pre-timeout");
912 } else if (preop_val
== WDOG_PREOP_GIVE_DATA
) {
913 spin_lock(&ipmi_read_lock
);
915 wake_up_interruptible(&read_q
);
916 kill_fasync(&fasync_q
, SIGIO
, POLL_IN
);
918 spin_unlock(&ipmi_read_lock
);
923 * On some machines, the heartbeat will give an error and not
924 * work unless we re-enable the timer. So do so.
926 pretimeout_since_last_heartbeat
= 1;
929 static struct ipmi_user_hndl ipmi_hndlrs
= {
930 .ipmi_recv_hndl
= ipmi_wdog_msg_handler
,
931 .ipmi_watchdog_pretimeout
= ipmi_wdog_pretimeout_handler
934 static void ipmi_register_watchdog(int ipmi_intf
)
941 if ((ifnum_to_use
>= 0) && (ifnum_to_use
!= ipmi_intf
))
944 watchdog_ifnum
= ipmi_intf
;
946 rv
= ipmi_create_user(ipmi_intf
, &ipmi_hndlrs
, NULL
, &watchdog_user
);
948 printk(KERN_CRIT PFX
"Unable to register with ipmi\n");
952 ipmi_get_version(watchdog_user
,
954 &ipmi_version_minor
);
956 rv
= misc_register(&ipmi_wdog_miscdev
);
958 ipmi_destroy_user(watchdog_user
);
959 watchdog_user
= NULL
;
960 printk(KERN_CRIT PFX
"Unable to register misc device\n");
964 if (nmi_handler_registered
) {
965 int old_pretimeout
= pretimeout
;
966 int old_timeout
= timeout
;
967 int old_preop_val
= preop_val
;
970 * Set the pretimeout to go off in a second and give
971 * ourselves plenty of time to stop the timer.
973 ipmi_watchdog_state
= WDOG_TIMEOUT_RESET
;
974 preop_val
= WDOG_PREOP_NONE
; /* Make sure nothing happens */
980 rv
= ipmi_set_timeout(IPMI_SET_TIMEOUT_FORCE_HB
);
982 printk(KERN_WARNING PFX
"Error starting timer to"
983 " test NMI: 0x%x. The NMI pretimeout will"
984 " likely not work\n", rv
);
991 if (testing_nmi
!= 2) {
992 printk(KERN_WARNING PFX
"IPMI NMI didn't seem to"
993 " occur. The NMI pretimeout will"
994 " likely not work\n");
998 preop_val
= old_preop_val
;
999 pretimeout
= old_pretimeout
;
1000 timeout
= old_timeout
;
1005 if ((start_now
) && (rv
== 0)) {
1006 /* Run from startup, so start the timer now. */
1007 start_now
= 0; /* Disable this function after first startup. */
1008 ipmi_watchdog_state
= action_val
;
1009 ipmi_set_timeout(IPMI_SET_TIMEOUT_FORCE_HB
);
1010 printk(KERN_INFO PFX
"Starting now!\n");
1012 /* Stop the timer now. */
1013 ipmi_watchdog_state
= WDOG_TIMEOUT_NONE
;
1014 ipmi_set_timeout(IPMI_SET_TIMEOUT_NO_HB
);
1018 static void ipmi_unregister_watchdog(int ipmi_intf
)
1025 if (watchdog_ifnum
!= ipmi_intf
)
1028 /* Make sure no one can call us any more. */
1029 misc_deregister(&ipmi_wdog_miscdev
);
1032 * Wait to make sure the message makes it out. The lower layer has
1033 * pointers to our buffers, we want to make sure they are done before
1034 * we release our memory.
1036 while (atomic_read(&set_timeout_tofree
))
1037 schedule_timeout_uninterruptible(1);
1039 /* Disconnect from IPMI. */
1040 rv
= ipmi_destroy_user(watchdog_user
);
1042 printk(KERN_WARNING PFX
"error unlinking from IPMI: %d\n",
1045 watchdog_user
= NULL
;
1053 ipmi_nmi(struct notifier_block
*self
, unsigned long val
, void *data
)
1055 struct die_args
*args
= data
;
1060 /* Hack, if it's a memory or I/O error, ignore it. */
1061 if (args
->err
& 0xc0)
1065 * If we get here, it's an NMI that's not a memory or I/O
1066 * error. We can't truly tell if it's from IPMI or not
1067 * without sending a message, and sending a message is almost
1068 * impossible because of locking.
1076 /* If we are not expecting a timeout, ignore it. */
1077 if (ipmi_watchdog_state
== WDOG_TIMEOUT_NONE
)
1080 if (preaction_val
!= WDOG_PRETIMEOUT_NMI
)
1084 * If no one else handled the NMI, we assume it was the IPMI
1087 if (preop_val
== WDOG_PREOP_PANIC
) {
1088 /* On some machines, the heartbeat will give
1089 an error and not work unless we re-enable
1090 the timer. So do so. */
1091 pretimeout_since_last_heartbeat
= 1;
1092 if (atomic_inc_and_test(&preop_panic_excl
))
1093 panic(PFX
"pre-timeout");
1099 static struct notifier_block ipmi_nmi_handler
= {
1100 .notifier_call
= ipmi_nmi
1104 static int wdog_reboot_handler(struct notifier_block
*this,
1108 static int reboot_event_handled
;
1110 if ((watchdog_user
) && (!reboot_event_handled
)) {
1111 /* Make sure we only do this once. */
1112 reboot_event_handled
= 1;
1114 if (code
== SYS_POWER_OFF
|| code
== SYS_HALT
) {
1115 /* Disable the WDT if we are shutting down. */
1116 ipmi_watchdog_state
= WDOG_TIMEOUT_NONE
;
1117 panic_halt_ipmi_set_timeout();
1118 } else if (ipmi_watchdog_state
!= WDOG_TIMEOUT_NONE
) {
1119 /* Set a long timer to let the reboot happens, but
1120 reboot if it hangs, but only if the watchdog
1121 timer was already running. */
1124 ipmi_watchdog_state
= WDOG_TIMEOUT_RESET
;
1125 panic_halt_ipmi_set_timeout();
1131 static struct notifier_block wdog_reboot_notifier
= {
1132 .notifier_call
= wdog_reboot_handler
,
1137 static int wdog_panic_handler(struct notifier_block
*this,
1138 unsigned long event
,
1141 static int panic_event_handled
;
1143 /* On a panic, if we have a panic timeout, make sure to extend
1144 the watchdog timer to a reasonable value to complete the
1145 panic, if the watchdog timer is running. Plus the
1146 pretimeout is meaningless at panic time. */
1147 if (watchdog_user
&& !panic_event_handled
&&
1148 ipmi_watchdog_state
!= WDOG_TIMEOUT_NONE
) {
1149 /* Make sure we do this only once. */
1150 panic_event_handled
= 1;
1154 panic_halt_ipmi_set_timeout();
1160 static struct notifier_block wdog_panic_notifier
= {
1161 .notifier_call
= wdog_panic_handler
,
1163 .priority
= 150 /* priority: INT_MAX >= x >= 0 */
1167 static void ipmi_new_smi(int if_num
, struct device
*device
)
1169 ipmi_register_watchdog(if_num
);
1172 static void ipmi_smi_gone(int if_num
)
1174 ipmi_unregister_watchdog(if_num
);
1177 static struct ipmi_smi_watcher smi_watcher
= {
1178 .owner
= THIS_MODULE
,
1179 .new_smi
= ipmi_new_smi
,
1180 .smi_gone
= ipmi_smi_gone
1183 static int action_op(const char *inval
, char *outval
)
1186 strcpy(outval
, action
);
1191 if (strcmp(inval
, "reset") == 0)
1192 action_val
= WDOG_TIMEOUT_RESET
;
1193 else if (strcmp(inval
, "none") == 0)
1194 action_val
= WDOG_TIMEOUT_NONE
;
1195 else if (strcmp(inval
, "power_cycle") == 0)
1196 action_val
= WDOG_TIMEOUT_POWER_CYCLE
;
1197 else if (strcmp(inval
, "power_off") == 0)
1198 action_val
= WDOG_TIMEOUT_POWER_DOWN
;
1201 strcpy(action
, inval
);
1205 static int preaction_op(const char *inval
, char *outval
)
1208 strcpy(outval
, preaction
);
1213 if (strcmp(inval
, "pre_none") == 0)
1214 preaction_val
= WDOG_PRETIMEOUT_NONE
;
1215 else if (strcmp(inval
, "pre_smi") == 0)
1216 preaction_val
= WDOG_PRETIMEOUT_SMI
;
1218 else if (strcmp(inval
, "pre_nmi") == 0)
1219 preaction_val
= WDOG_PRETIMEOUT_NMI
;
1221 else if (strcmp(inval
, "pre_int") == 0)
1222 preaction_val
= WDOG_PRETIMEOUT_MSG_INT
;
1225 strcpy(preaction
, inval
);
1229 static int preop_op(const char *inval
, char *outval
)
1232 strcpy(outval
, preop
);
1237 if (strcmp(inval
, "preop_none") == 0)
1238 preop_val
= WDOG_PREOP_NONE
;
1239 else if (strcmp(inval
, "preop_panic") == 0)
1240 preop_val
= WDOG_PREOP_PANIC
;
1241 else if (strcmp(inval
, "preop_give_data") == 0)
1242 preop_val
= WDOG_PREOP_GIVE_DATA
;
1245 strcpy(preop
, inval
);
1249 static void check_parms(void)
1255 if (preaction_val
== WDOG_PRETIMEOUT_NMI
) {
1257 if (preop_val
== WDOG_PREOP_GIVE_DATA
) {
1258 printk(KERN_WARNING PFX
"Pretimeout op is to give data"
1259 " but NMI pretimeout is enabled, setting"
1260 " pretimeout op to none\n");
1261 preop_op("preop_none", NULL
);
1265 if (do_nmi
&& !nmi_handler_registered
) {
1266 rv
= register_die_notifier(&ipmi_nmi_handler
);
1268 printk(KERN_WARNING PFX
1269 "Can't register nmi handler\n");
1272 nmi_handler_registered
= 1;
1273 } else if (!do_nmi
&& nmi_handler_registered
) {
1274 unregister_die_notifier(&ipmi_nmi_handler
);
1275 nmi_handler_registered
= 0;
1280 static int __init
ipmi_wdog_init(void)
1284 if (action_op(action
, NULL
)) {
1285 action_op("reset", NULL
);
1286 printk(KERN_INFO PFX
"Unknown action '%s', defaulting to"
1287 " reset\n", action
);
1290 if (preaction_op(preaction
, NULL
)) {
1291 preaction_op("pre_none", NULL
);
1292 printk(KERN_INFO PFX
"Unknown preaction '%s', defaulting to"
1293 " none\n", preaction
);
1296 if (preop_op(preop
, NULL
)) {
1297 preop_op("preop_none", NULL
);
1298 printk(KERN_INFO PFX
"Unknown preop '%s', defaulting to"
1304 register_reboot_notifier(&wdog_reboot_notifier
);
1305 atomic_notifier_chain_register(&panic_notifier_list
,
1306 &wdog_panic_notifier
);
1308 rv
= ipmi_smi_watcher_register(&smi_watcher
);
1311 if (nmi_handler_registered
)
1312 unregister_die_notifier(&ipmi_nmi_handler
);
1314 atomic_notifier_chain_unregister(&panic_notifier_list
,
1315 &wdog_panic_notifier
);
1316 unregister_reboot_notifier(&wdog_reboot_notifier
);
1317 printk(KERN_WARNING PFX
"can't register smi watcher\n");
1321 printk(KERN_INFO PFX
"driver initialized\n");
1326 static void __exit
ipmi_wdog_exit(void)
1328 ipmi_smi_watcher_unregister(&smi_watcher
);
1329 ipmi_unregister_watchdog(watchdog_ifnum
);
1332 if (nmi_handler_registered
)
1333 unregister_die_notifier(&ipmi_nmi_handler
);
1336 atomic_notifier_chain_unregister(&panic_notifier_list
,
1337 &wdog_panic_notifier
);
1338 unregister_reboot_notifier(&wdog_reboot_notifier
);
1340 module_exit(ipmi_wdog_exit
);
1341 module_init(ipmi_wdog_init
);
1342 MODULE_LICENSE("GPL");
1343 MODULE_AUTHOR("Corey Minyard <minyard@mvista.com>");
1344 MODULE_DESCRIPTION("watchdog timer based upon the IPMI interface.");