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/smp_lock.h>
39 #include <linux/watchdog.h>
40 #include <linux/miscdevice.h>
41 #include <linux/init.h>
42 #include <linux/completion.h>
43 #include <linux/kdebug.h>
44 #include <linux/rwsem.h>
45 #include <linux/errno.h>
46 #include <asm/uaccess.h>
47 #include <linux/notifier.h>
48 #include <linux/nmi.h>
49 #include <linux/reboot.h>
50 #include <linux/wait.h>
51 #include <linux/poll.h>
52 #include <linux/string.h>
53 #include <linux/ctype.h>
54 #include <linux/delay.h>
55 #include <asm/atomic.h>
59 * This is ugly, but I've determined that x86 is the only architecture
60 * that can reasonably support the IPMI NMI watchdog timeout at this
61 * time. If another architecture adds this capability somehow, it
62 * will have to be a somewhat different mechanism and I have no idea
63 * how it will work. So in the unlikely event that another
64 * architecture supports this, we can figure out a good generic
65 * mechanism for it at that time.
67 #include <asm/kdebug.h>
71 #define PFX "IPMI Watchdog: "
74 * The IPMI command/response information for the watchdog timer.
77 /* values for byte 1 of the set command, byte 2 of the get response. */
78 #define WDOG_DONT_LOG (1 << 7)
79 #define WDOG_DONT_STOP_ON_SET (1 << 6)
80 #define WDOG_SET_TIMER_USE(byte, use) \
81 byte = ((byte) & 0xf8) | ((use) & 0x7)
82 #define WDOG_GET_TIMER_USE(byte) ((byte) & 0x7)
83 #define WDOG_TIMER_USE_BIOS_FRB2 1
84 #define WDOG_TIMER_USE_BIOS_POST 2
85 #define WDOG_TIMER_USE_OS_LOAD 3
86 #define WDOG_TIMER_USE_SMS_OS 4
87 #define WDOG_TIMER_USE_OEM 5
89 /* values for byte 2 of the set command, byte 3 of the get response. */
90 #define WDOG_SET_PRETIMEOUT_ACT(byte, use) \
91 byte = ((byte) & 0x8f) | (((use) & 0x7) << 4)
92 #define WDOG_GET_PRETIMEOUT_ACT(byte) (((byte) >> 4) & 0x7)
93 #define WDOG_PRETIMEOUT_NONE 0
94 #define WDOG_PRETIMEOUT_SMI 1
95 #define WDOG_PRETIMEOUT_NMI 2
96 #define WDOG_PRETIMEOUT_MSG_INT 3
98 /* Operations that can be performed on a pretimout. */
99 #define WDOG_PREOP_NONE 0
100 #define WDOG_PREOP_PANIC 1
101 /* Cause data to be available to read. Doesn't work in NMI mode. */
102 #define WDOG_PREOP_GIVE_DATA 2
104 /* Actions to perform on a full timeout. */
105 #define WDOG_SET_TIMEOUT_ACT(byte, use) \
106 byte = ((byte) & 0xf8) | ((use) & 0x7)
107 #define WDOG_GET_TIMEOUT_ACT(byte) ((byte) & 0x7)
108 #define WDOG_TIMEOUT_NONE 0
109 #define WDOG_TIMEOUT_RESET 1
110 #define WDOG_TIMEOUT_POWER_DOWN 2
111 #define WDOG_TIMEOUT_POWER_CYCLE 3
114 * Byte 3 of the get command, byte 4 of the get response is the
115 * pre-timeout in seconds.
118 /* Bits for setting byte 4 of the set command, byte 5 of the get response. */
119 #define WDOG_EXPIRE_CLEAR_BIOS_FRB2 (1 << 1)
120 #define WDOG_EXPIRE_CLEAR_BIOS_POST (1 << 2)
121 #define WDOG_EXPIRE_CLEAR_OS_LOAD (1 << 3)
122 #define WDOG_EXPIRE_CLEAR_SMS_OS (1 << 4)
123 #define WDOG_EXPIRE_CLEAR_OEM (1 << 5)
126 * Setting/getting the watchdog timer value. This is for bytes 5 and
127 * 6 (the timeout time) of the set command, and bytes 6 and 7 (the
128 * timeout time) and 8 and 9 (the current countdown value) of the
129 * response. The timeout value is given in seconds (in the command it
130 * is 100ms intervals).
132 #define WDOG_SET_TIMEOUT(byte1, byte2, val) \
133 (byte1) = (((val) * 10) & 0xff), (byte2) = (((val) * 10) >> 8)
134 #define WDOG_GET_TIMEOUT(byte1, byte2) \
135 (((byte1) | ((byte2) << 8)) / 10)
137 #define IPMI_WDOG_RESET_TIMER 0x22
138 #define IPMI_WDOG_SET_TIMER 0x24
139 #define IPMI_WDOG_GET_TIMER 0x25
141 /* These are here until the real ones get into the watchdog.h interface. */
142 #ifndef WDIOC_GETTIMEOUT
143 #define WDIOC_GETTIMEOUT _IOW(WATCHDOG_IOCTL_BASE, 20, int)
145 #ifndef WDIOC_SET_PRETIMEOUT
146 #define WDIOC_SET_PRETIMEOUT _IOW(WATCHDOG_IOCTL_BASE, 21, int)
148 #ifndef WDIOC_GET_PRETIMEOUT
149 #define WDIOC_GET_PRETIMEOUT _IOW(WATCHDOG_IOCTL_BASE, 22, int)
152 static int nowayout
= WATCHDOG_NOWAYOUT
;
154 static ipmi_user_t watchdog_user
;
155 static int watchdog_ifnum
;
157 /* Default the timeout to 10 seconds. */
158 static int timeout
= 10;
160 /* The pre-timeout is disabled by default. */
161 static int pretimeout
;
163 /* Default action is to reset the board on a timeout. */
164 static unsigned char action_val
= WDOG_TIMEOUT_RESET
;
166 static char action
[16] = "reset";
168 static unsigned char preaction_val
= WDOG_PRETIMEOUT_NONE
;
170 static char preaction
[16] = "pre_none";
172 static unsigned char preop_val
= WDOG_PREOP_NONE
;
174 static char preop
[16] = "preop_none";
175 static DEFINE_SPINLOCK(ipmi_read_lock
);
176 static char data_to_read
;
177 static DECLARE_WAIT_QUEUE_HEAD(read_q
);
178 static struct fasync_struct
*fasync_q
;
179 static char pretimeout_since_last_heartbeat
;
180 static char expect_close
;
182 static int ifnum_to_use
= -1;
184 /* Parameters to ipmi_set_timeout */
185 #define IPMI_SET_TIMEOUT_NO_HB 0
186 #define IPMI_SET_TIMEOUT_HB_IF_NECESSARY 1
187 #define IPMI_SET_TIMEOUT_FORCE_HB 2
189 static int ipmi_set_timeout(int do_heartbeat
);
190 static void ipmi_register_watchdog(int ipmi_intf
);
191 static void ipmi_unregister_watchdog(int ipmi_intf
);
194 * If true, the driver will start running as soon as it is configured
197 static int start_now
;
199 static int set_param_int(const char *val
, struct kernel_param
*kp
)
207 l
= simple_strtoul(val
, &endp
, 0);
211 *((int *)kp
->arg
) = l
;
213 rv
= ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY
);
218 static int get_param_int(char *buffer
, struct kernel_param
*kp
)
220 return sprintf(buffer
, "%i", *((int *)kp
->arg
));
223 typedef int (*action_fn
)(const char *intval
, char *outval
);
225 static int action_op(const char *inval
, char *outval
);
226 static int preaction_op(const char *inval
, char *outval
);
227 static int preop_op(const char *inval
, char *outval
);
228 static void check_parms(void);
230 static int set_param_str(const char *val
, struct kernel_param
*kp
)
232 action_fn fn
= (action_fn
) kp
->arg
;
237 strncpy(valcp
, val
, 16);
248 rv
= ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY
);
254 static int get_param_str(char *buffer
, struct kernel_param
*kp
)
256 action_fn fn
= (action_fn
) kp
->arg
;
259 rv
= fn(NULL
, buffer
);
262 return strlen(buffer
);
266 static int set_param_wdog_ifnum(const char *val
, struct kernel_param
*kp
)
268 int rv
= param_set_int(val
, kp
);
271 if ((ifnum_to_use
< 0) || (ifnum_to_use
== watchdog_ifnum
))
274 ipmi_unregister_watchdog(watchdog_ifnum
);
275 ipmi_register_watchdog(ifnum_to_use
);
279 module_param_call(ifnum_to_use
, set_param_wdog_ifnum
, get_param_int
,
280 &ifnum_to_use
, 0644);
281 MODULE_PARM_DESC(ifnum_to_use
, "The interface number to use for the watchdog "
282 "timer. Setting to -1 defaults to the first registered "
285 module_param_call(timeout
, set_param_int
, get_param_int
, &timeout
, 0644);
286 MODULE_PARM_DESC(timeout
, "Timeout value in seconds.");
288 module_param_call(pretimeout
, set_param_int
, get_param_int
, &pretimeout
, 0644);
289 MODULE_PARM_DESC(pretimeout
, "Pretimeout value in seconds.");
291 module_param_call(action
, set_param_str
, get_param_str
, action_op
, 0644);
292 MODULE_PARM_DESC(action
, "Timeout action. One of: "
293 "reset, none, power_cycle, power_off.");
295 module_param_call(preaction
, set_param_str
, get_param_str
, preaction_op
, 0644);
296 MODULE_PARM_DESC(preaction
, "Pretimeout action. One of: "
297 "pre_none, pre_smi, pre_nmi, pre_int.");
299 module_param_call(preop
, set_param_str
, get_param_str
, preop_op
, 0644);
300 MODULE_PARM_DESC(preop
, "Pretimeout driver operation. One of: "
301 "preop_none, preop_panic, preop_give_data.");
303 module_param(start_now
, int, 0444);
304 MODULE_PARM_DESC(start_now
, "Set to 1 to start the watchdog as"
305 "soon as the driver is loaded.");
307 module_param(nowayout
, int, 0644);
308 MODULE_PARM_DESC(nowayout
, "Watchdog cannot be stopped once started "
309 "(default=CONFIG_WATCHDOG_NOWAYOUT)");
311 /* Default state of the timer. */
312 static unsigned char ipmi_watchdog_state
= WDOG_TIMEOUT_NONE
;
314 /* If shutting down via IPMI, we ignore the heartbeat. */
315 static int ipmi_ignore_heartbeat
;
317 /* Is someone using the watchdog? Only one user is allowed. */
318 static unsigned long ipmi_wdog_open
;
321 * If set to 1, the heartbeat command will set the state to reset and
322 * start the timer. The timer doesn't normally run when the driver is
323 * first opened until the heartbeat is set the first time, this
324 * variable is used to accomplish this.
326 static int ipmi_start_timer_on_heartbeat
;
328 /* IPMI version of the BMC. */
329 static unsigned char ipmi_version_major
;
330 static unsigned char ipmi_version_minor
;
332 /* If a pretimeout occurs, this is used to allow only one panic to happen. */
333 static atomic_t preop_panic_excl
= ATOMIC_INIT(-1);
336 static int testing_nmi
;
337 static int nmi_handler_registered
;
340 static int ipmi_heartbeat(void);
343 * We use a mutex to make sure that only one thing can send a set
344 * timeout at one time, because we only have one copy of the data.
345 * The mutex is claimed when the set_timeout is sent and freed
346 * when both messages are free.
348 static atomic_t set_timeout_tofree
= ATOMIC_INIT(0);
349 static DEFINE_MUTEX(set_timeout_lock
);
350 static DECLARE_COMPLETION(set_timeout_wait
);
351 static void set_timeout_free_smi(struct ipmi_smi_msg
*msg
)
353 if (atomic_dec_and_test(&set_timeout_tofree
))
354 complete(&set_timeout_wait
);
356 static void set_timeout_free_recv(struct ipmi_recv_msg
*msg
)
358 if (atomic_dec_and_test(&set_timeout_tofree
))
359 complete(&set_timeout_wait
);
361 static struct ipmi_smi_msg set_timeout_smi_msg
= {
362 .done
= set_timeout_free_smi
364 static struct ipmi_recv_msg set_timeout_recv_msg
= {
365 .done
= set_timeout_free_recv
368 static int i_ipmi_set_timeout(struct ipmi_smi_msg
*smi_msg
,
369 struct ipmi_recv_msg
*recv_msg
,
370 int *send_heartbeat_now
)
372 struct kernel_ipmi_msg msg
;
373 unsigned char data
[6];
375 struct ipmi_system_interface_addr addr
;
379 /* These can be cleared as we are setting the timeout. */
380 pretimeout_since_last_heartbeat
= 0;
383 WDOG_SET_TIMER_USE(data
[0], WDOG_TIMER_USE_SMS_OS
);
385 if ((ipmi_version_major
> 1)
386 || ((ipmi_version_major
== 1) && (ipmi_version_minor
>= 5))) {
387 /* This is an IPMI 1.5-only feature. */
388 data
[0] |= WDOG_DONT_STOP_ON_SET
;
389 } else if (ipmi_watchdog_state
!= WDOG_TIMEOUT_NONE
) {
391 * In ipmi 1.0, setting the timer stops the watchdog, we
392 * need to start it back up again.
398 WDOG_SET_TIMEOUT_ACT(data
[1], ipmi_watchdog_state
);
399 if ((pretimeout
> 0) && (ipmi_watchdog_state
!= WDOG_TIMEOUT_NONE
)) {
400 WDOG_SET_PRETIMEOUT_ACT(data
[1], preaction_val
);
401 data
[2] = pretimeout
;
403 WDOG_SET_PRETIMEOUT_ACT(data
[1], WDOG_PRETIMEOUT_NONE
);
404 data
[2] = 0; /* No pretimeout. */
407 WDOG_SET_TIMEOUT(data
[4], data
[5], timeout
);
409 addr
.addr_type
= IPMI_SYSTEM_INTERFACE_ADDR_TYPE
;
410 addr
.channel
= IPMI_BMC_CHANNEL
;
414 msg
.cmd
= IPMI_WDOG_SET_TIMER
;
416 msg
.data_len
= sizeof(data
);
417 rv
= ipmi_request_supply_msgs(watchdog_user
,
418 (struct ipmi_addr
*) &addr
,
426 printk(KERN_WARNING PFX
"set timeout error: %d\n",
430 if (send_heartbeat_now
)
431 *send_heartbeat_now
= hbnow
;
436 static int ipmi_set_timeout(int do_heartbeat
)
438 int send_heartbeat_now
;
442 /* We can only send one of these at a time. */
443 mutex_lock(&set_timeout_lock
);
445 atomic_set(&set_timeout_tofree
, 2);
447 rv
= i_ipmi_set_timeout(&set_timeout_smi_msg
,
448 &set_timeout_recv_msg
,
449 &send_heartbeat_now
);
451 mutex_unlock(&set_timeout_lock
);
455 wait_for_completion(&set_timeout_wait
);
457 mutex_unlock(&set_timeout_lock
);
459 if ((do_heartbeat
== IPMI_SET_TIMEOUT_FORCE_HB
)
460 || ((send_heartbeat_now
)
461 && (do_heartbeat
== IPMI_SET_TIMEOUT_HB_IF_NECESSARY
)))
462 rv
= ipmi_heartbeat();
468 static atomic_t panic_done_count
= ATOMIC_INIT(0);
470 static void panic_smi_free(struct ipmi_smi_msg
*msg
)
472 atomic_dec(&panic_done_count
);
474 static void panic_recv_free(struct ipmi_recv_msg
*msg
)
476 atomic_dec(&panic_done_count
);
479 static struct ipmi_smi_msg panic_halt_heartbeat_smi_msg
= {
480 .done
= panic_smi_free
482 static struct ipmi_recv_msg panic_halt_heartbeat_recv_msg
= {
483 .done
= panic_recv_free
486 static void panic_halt_ipmi_heartbeat(void)
488 struct kernel_ipmi_msg msg
;
489 struct ipmi_system_interface_addr addr
;
493 * Don't reset the timer if we have the timer turned off, that
494 * re-enables the watchdog.
496 if (ipmi_watchdog_state
== WDOG_TIMEOUT_NONE
)
499 addr
.addr_type
= IPMI_SYSTEM_INTERFACE_ADDR_TYPE
;
500 addr
.channel
= IPMI_BMC_CHANNEL
;
504 msg
.cmd
= IPMI_WDOG_RESET_TIMER
;
507 rv
= ipmi_request_supply_msgs(watchdog_user
,
508 (struct ipmi_addr
*) &addr
,
512 &panic_halt_heartbeat_smi_msg
,
513 &panic_halt_heartbeat_recv_msg
,
516 atomic_add(2, &panic_done_count
);
519 static struct ipmi_smi_msg panic_halt_smi_msg
= {
520 .done
= panic_smi_free
522 static struct ipmi_recv_msg panic_halt_recv_msg
= {
523 .done
= panic_recv_free
527 * Special call, doesn't claim any locks. This is only to be called
528 * at panic or halt time, in run-to-completion mode, when the caller
529 * is the only CPU and the only thing that will be going is these IPMI
532 static void panic_halt_ipmi_set_timeout(void)
534 int send_heartbeat_now
;
537 /* Wait for the messages to be free. */
538 while (atomic_read(&panic_done_count
) != 0)
539 ipmi_poll_interface(watchdog_user
);
540 rv
= i_ipmi_set_timeout(&panic_halt_smi_msg
,
541 &panic_halt_recv_msg
,
542 &send_heartbeat_now
);
544 atomic_add(2, &panic_done_count
);
545 if (send_heartbeat_now
)
546 panic_halt_ipmi_heartbeat();
548 printk(KERN_WARNING PFX
549 "Unable to extend the watchdog timeout.");
550 while (atomic_read(&panic_done_count
) != 0)
551 ipmi_poll_interface(watchdog_user
);
555 * We use a mutex to make sure that only one thing can send a
556 * heartbeat at one time, because we only have one copy of the data.
557 * The semaphore is claimed when the set_timeout is sent and freed
558 * when both messages are free.
560 static atomic_t heartbeat_tofree
= ATOMIC_INIT(0);
561 static DEFINE_MUTEX(heartbeat_lock
);
562 static DECLARE_COMPLETION(heartbeat_wait
);
563 static void heartbeat_free_smi(struct ipmi_smi_msg
*msg
)
565 if (atomic_dec_and_test(&heartbeat_tofree
))
566 complete(&heartbeat_wait
);
568 static void heartbeat_free_recv(struct ipmi_recv_msg
*msg
)
570 if (atomic_dec_and_test(&heartbeat_tofree
))
571 complete(&heartbeat_wait
);
573 static struct ipmi_smi_msg heartbeat_smi_msg
= {
574 .done
= heartbeat_free_smi
576 static struct ipmi_recv_msg heartbeat_recv_msg
= {
577 .done
= heartbeat_free_recv
580 static int ipmi_heartbeat(void)
582 struct kernel_ipmi_msg msg
;
584 struct ipmi_system_interface_addr addr
;
586 if (ipmi_ignore_heartbeat
)
589 if (ipmi_start_timer_on_heartbeat
) {
590 ipmi_start_timer_on_heartbeat
= 0;
591 ipmi_watchdog_state
= action_val
;
592 return ipmi_set_timeout(IPMI_SET_TIMEOUT_FORCE_HB
);
593 } else if (pretimeout_since_last_heartbeat
) {
595 * A pretimeout occurred, make sure we set the timeout.
596 * We don't want to set the action, though, we want to
597 * leave that alone (thus it can't be combined with the
600 return ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY
);
603 mutex_lock(&heartbeat_lock
);
605 atomic_set(&heartbeat_tofree
, 2);
608 * Don't reset the timer if we have the timer turned off, that
609 * re-enables the watchdog.
611 if (ipmi_watchdog_state
== WDOG_TIMEOUT_NONE
) {
612 mutex_unlock(&heartbeat_lock
);
616 addr
.addr_type
= IPMI_SYSTEM_INTERFACE_ADDR_TYPE
;
617 addr
.channel
= IPMI_BMC_CHANNEL
;
621 msg
.cmd
= IPMI_WDOG_RESET_TIMER
;
624 rv
= ipmi_request_supply_msgs(watchdog_user
,
625 (struct ipmi_addr
*) &addr
,
633 mutex_unlock(&heartbeat_lock
);
634 printk(KERN_WARNING PFX
"heartbeat failure: %d\n",
639 /* Wait for the heartbeat to be sent. */
640 wait_for_completion(&heartbeat_wait
);
642 if (heartbeat_recv_msg
.msg
.data
[0] != 0) {
644 * Got an error in the heartbeat response. It was already
645 * reported in ipmi_wdog_msg_handler, but we should return
651 mutex_unlock(&heartbeat_lock
);
656 static struct watchdog_info ident
= {
657 .options
= 0, /* WDIOF_SETTIMEOUT, */
658 .firmware_version
= 1,
662 static int ipmi_ioctl(struct inode
*inode
, struct file
*file
,
663 unsigned int cmd
, unsigned long arg
)
665 void __user
*argp
= (void __user
*)arg
;
670 case WDIOC_GETSUPPORT
:
671 i
= copy_to_user(argp
, &ident
, sizeof(ident
));
672 return i
? -EFAULT
: 0;
674 case WDIOC_SETTIMEOUT
:
675 i
= copy_from_user(&val
, argp
, sizeof(int));
679 return ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY
);
681 case WDIOC_GETTIMEOUT
:
682 i
= copy_to_user(argp
, &timeout
, sizeof(timeout
));
687 case WDIOC_SET_PRETIMEOUT
:
688 case WDIOC_SETPRETIMEOUT
:
689 i
= copy_from_user(&val
, argp
, sizeof(int));
693 return ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY
);
695 case WDIOC_GET_PRETIMEOUT
:
696 case WDIOC_GETPRETIMEOUT
:
697 i
= copy_to_user(argp
, &pretimeout
, sizeof(pretimeout
));
702 case WDIOC_KEEPALIVE
:
703 return ipmi_heartbeat();
705 case WDIOC_SETOPTIONS
:
706 i
= copy_from_user(&val
, argp
, sizeof(int));
709 if (val
& WDIOS_DISABLECARD
) {
710 ipmi_watchdog_state
= WDOG_TIMEOUT_NONE
;
711 ipmi_set_timeout(IPMI_SET_TIMEOUT_NO_HB
);
712 ipmi_start_timer_on_heartbeat
= 0;
715 if (val
& WDIOS_ENABLECARD
) {
716 ipmi_watchdog_state
= action_val
;
717 ipmi_set_timeout(IPMI_SET_TIMEOUT_FORCE_HB
);
721 case WDIOC_GETSTATUS
:
723 i
= copy_to_user(argp
, &val
, sizeof(val
));
733 static ssize_t
ipmi_write(struct file
*file
,
734 const char __user
*buf
,
744 /* In case it was set long ago */
747 for (i
= 0; i
!= len
; i
++) {
750 if (get_user(c
, buf
+ i
))
756 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
))
825 * Don't start the timer now, let it start on the
828 ipmi_start_timer_on_heartbeat
= 1;
829 return nonseekable_open(ino
, filep
);
836 static unsigned int ipmi_poll(struct file
*file
, poll_table
*wait
)
838 unsigned int mask
= 0;
840 poll_wait(file
, &read_q
, wait
);
842 spin_lock(&ipmi_read_lock
);
844 mask
|= (POLLIN
| POLLRDNORM
);
845 spin_unlock(&ipmi_read_lock
);
850 static int ipmi_fasync(int fd
, struct file
*file
, int on
)
854 result
= fasync_helper(fd
, file
, on
, &fasync_q
);
859 static int ipmi_close(struct inode
*ino
, struct file
*filep
)
861 if (iminor(ino
) == WATCHDOG_MINOR
) {
862 if (expect_close
== 42) {
863 ipmi_watchdog_state
= WDOG_TIMEOUT_NONE
;
864 ipmi_set_timeout(IPMI_SET_TIMEOUT_NO_HB
);
867 "Unexpected close, not stopping watchdog!\n");
870 clear_bit(0, &ipmi_wdog_open
);
878 static const struct file_operations ipmi_wdog_fops
= {
879 .owner
= THIS_MODULE
,
885 .release
= ipmi_close
,
886 .fasync
= ipmi_fasync
,
889 static struct miscdevice ipmi_wdog_miscdev
= {
890 .minor
= WATCHDOG_MINOR
,
892 .fops
= &ipmi_wdog_fops
895 static void ipmi_wdog_msg_handler(struct ipmi_recv_msg
*msg
,
898 if (msg
->msg
.data
[0] != 0) {
899 printk(KERN_ERR PFX
"response: Error %x on cmd %x\n",
904 ipmi_free_recv_msg(msg
);
907 static void ipmi_wdog_pretimeout_handler(void *handler_data
)
909 if (preaction_val
!= WDOG_PRETIMEOUT_NONE
) {
910 if (preop_val
== WDOG_PREOP_PANIC
) {
911 if (atomic_inc_and_test(&preop_panic_excl
))
912 panic("Watchdog pre-timeout");
913 } else if (preop_val
== WDOG_PREOP_GIVE_DATA
) {
914 spin_lock(&ipmi_read_lock
);
916 wake_up_interruptible(&read_q
);
917 kill_fasync(&fasync_q
, SIGIO
, POLL_IN
);
919 spin_unlock(&ipmi_read_lock
);
924 * On some machines, the heartbeat will give an error and not
925 * work unless we re-enable the timer. So do so.
927 pretimeout_since_last_heartbeat
= 1;
930 static struct ipmi_user_hndl ipmi_hndlrs
= {
931 .ipmi_recv_hndl
= ipmi_wdog_msg_handler
,
932 .ipmi_watchdog_pretimeout
= ipmi_wdog_pretimeout_handler
935 static void ipmi_register_watchdog(int ipmi_intf
)
942 if ((ifnum_to_use
>= 0) && (ifnum_to_use
!= ipmi_intf
))
945 watchdog_ifnum
= ipmi_intf
;
947 rv
= ipmi_create_user(ipmi_intf
, &ipmi_hndlrs
, NULL
, &watchdog_user
);
949 printk(KERN_CRIT PFX
"Unable to register with ipmi\n");
953 ipmi_get_version(watchdog_user
,
955 &ipmi_version_minor
);
957 rv
= misc_register(&ipmi_wdog_miscdev
);
959 ipmi_destroy_user(watchdog_user
);
960 watchdog_user
= NULL
;
961 printk(KERN_CRIT PFX
"Unable to register misc device\n");
965 if (nmi_handler_registered
) {
966 int old_pretimeout
= pretimeout
;
967 int old_timeout
= timeout
;
968 int old_preop_val
= preop_val
;
971 * Set the pretimeout to go off in a second and give
972 * ourselves plenty of time to stop the timer.
974 ipmi_watchdog_state
= WDOG_TIMEOUT_RESET
;
975 preop_val
= WDOG_PREOP_NONE
; /* Make sure nothing happens */
981 rv
= ipmi_set_timeout(IPMI_SET_TIMEOUT_FORCE_HB
);
983 printk(KERN_WARNING PFX
"Error starting timer to"
984 " test NMI: 0x%x. The NMI pretimeout will"
985 " likely not work\n", rv
);
992 if (testing_nmi
!= 2) {
993 printk(KERN_WARNING PFX
"IPMI NMI didn't seem to"
994 " occur. The NMI pretimeout will"
995 " likely not work\n");
999 preop_val
= old_preop_val
;
1000 pretimeout
= old_pretimeout
;
1001 timeout
= old_timeout
;
1006 if ((start_now
) && (rv
== 0)) {
1007 /* Run from startup, so start the timer now. */
1008 start_now
= 0; /* Disable this function after first startup. */
1009 ipmi_watchdog_state
= action_val
;
1010 ipmi_set_timeout(IPMI_SET_TIMEOUT_FORCE_HB
);
1011 printk(KERN_INFO PFX
"Starting now!\n");
1013 /* Stop the timer now. */
1014 ipmi_watchdog_state
= WDOG_TIMEOUT_NONE
;
1015 ipmi_set_timeout(IPMI_SET_TIMEOUT_NO_HB
);
1019 static void ipmi_unregister_watchdog(int ipmi_intf
)
1026 if (watchdog_ifnum
!= ipmi_intf
)
1029 /* Make sure no one can call us any more. */
1030 misc_deregister(&ipmi_wdog_miscdev
);
1033 * Wait to make sure the message makes it out. The lower layer has
1034 * pointers to our buffers, we want to make sure they are done before
1035 * we release our memory.
1037 while (atomic_read(&set_timeout_tofree
))
1038 schedule_timeout_uninterruptible(1);
1040 /* Disconnect from IPMI. */
1041 rv
= ipmi_destroy_user(watchdog_user
);
1043 printk(KERN_WARNING PFX
"error unlinking from IPMI: %d\n",
1046 watchdog_user
= NULL
;
1054 ipmi_nmi(struct notifier_block
*self
, unsigned long val
, void *data
)
1056 struct die_args
*args
= data
;
1061 /* Hack, if it's a memory or I/O error, ignore it. */
1062 if (args
->err
& 0xc0)
1066 * If we get here, it's an NMI that's not a memory or I/O
1067 * error. We can't truly tell if it's from IPMI or not
1068 * without sending a message, and sending a message is almost
1069 * impossible because of locking.
1077 /* If we are not expecting a timeout, ignore it. */
1078 if (ipmi_watchdog_state
== WDOG_TIMEOUT_NONE
)
1081 if (preaction_val
!= WDOG_PRETIMEOUT_NMI
)
1085 * If no one else handled the NMI, we assume it was the IPMI
1088 if (preop_val
== WDOG_PREOP_PANIC
) {
1089 /* On some machines, the heartbeat will give
1090 an error and not work unless we re-enable
1091 the timer. So do so. */
1092 pretimeout_since_last_heartbeat
= 1;
1093 if (atomic_inc_and_test(&preop_panic_excl
))
1094 panic(PFX
"pre-timeout");
1100 static struct notifier_block ipmi_nmi_handler
= {
1101 .notifier_call
= ipmi_nmi
1105 static int wdog_reboot_handler(struct notifier_block
*this,
1109 static int reboot_event_handled
;
1111 if ((watchdog_user
) && (!reboot_event_handled
)) {
1112 /* Make sure we only do this once. */
1113 reboot_event_handled
= 1;
1115 if (code
== SYS_POWER_OFF
|| code
== SYS_HALT
) {
1116 /* Disable the WDT if we are shutting down. */
1117 ipmi_watchdog_state
= WDOG_TIMEOUT_NONE
;
1118 panic_halt_ipmi_set_timeout();
1119 } else if (ipmi_watchdog_state
!= WDOG_TIMEOUT_NONE
) {
1120 /* Set a long timer to let the reboot happens, but
1121 reboot if it hangs, but only if the watchdog
1122 timer was already running. */
1125 ipmi_watchdog_state
= WDOG_TIMEOUT_RESET
;
1126 panic_halt_ipmi_set_timeout();
1132 static struct notifier_block wdog_reboot_notifier
= {
1133 .notifier_call
= wdog_reboot_handler
,
1138 static int wdog_panic_handler(struct notifier_block
*this,
1139 unsigned long event
,
1142 static int panic_event_handled
;
1144 /* On a panic, if we have a panic timeout, make sure to extend
1145 the watchdog timer to a reasonable value to complete the
1146 panic, if the watchdog timer is running. Plus the
1147 pretimeout is meaningless at panic time. */
1148 if (watchdog_user
&& !panic_event_handled
&&
1149 ipmi_watchdog_state
!= WDOG_TIMEOUT_NONE
) {
1150 /* Make sure we do this only once. */
1151 panic_event_handled
= 1;
1155 panic_halt_ipmi_set_timeout();
1161 static struct notifier_block wdog_panic_notifier
= {
1162 .notifier_call
= wdog_panic_handler
,
1164 .priority
= 150 /* priority: INT_MAX >= x >= 0 */
1168 static void ipmi_new_smi(int if_num
, struct device
*device
)
1170 ipmi_register_watchdog(if_num
);
1173 static void ipmi_smi_gone(int if_num
)
1175 ipmi_unregister_watchdog(if_num
);
1178 static struct ipmi_smi_watcher smi_watcher
= {
1179 .owner
= THIS_MODULE
,
1180 .new_smi
= ipmi_new_smi
,
1181 .smi_gone
= ipmi_smi_gone
1184 static int action_op(const char *inval
, char *outval
)
1187 strcpy(outval
, action
);
1192 if (strcmp(inval
, "reset") == 0)
1193 action_val
= WDOG_TIMEOUT_RESET
;
1194 else if (strcmp(inval
, "none") == 0)
1195 action_val
= WDOG_TIMEOUT_NONE
;
1196 else if (strcmp(inval
, "power_cycle") == 0)
1197 action_val
= WDOG_TIMEOUT_POWER_CYCLE
;
1198 else if (strcmp(inval
, "power_off") == 0)
1199 action_val
= WDOG_TIMEOUT_POWER_DOWN
;
1202 strcpy(action
, inval
);
1206 static int preaction_op(const char *inval
, char *outval
)
1209 strcpy(outval
, preaction
);
1214 if (strcmp(inval
, "pre_none") == 0)
1215 preaction_val
= WDOG_PRETIMEOUT_NONE
;
1216 else if (strcmp(inval
, "pre_smi") == 0)
1217 preaction_val
= WDOG_PRETIMEOUT_SMI
;
1219 else if (strcmp(inval
, "pre_nmi") == 0)
1220 preaction_val
= WDOG_PRETIMEOUT_NMI
;
1222 else if (strcmp(inval
, "pre_int") == 0)
1223 preaction_val
= WDOG_PRETIMEOUT_MSG_INT
;
1226 strcpy(preaction
, inval
);
1230 static int preop_op(const char *inval
, char *outval
)
1233 strcpy(outval
, preop
);
1238 if (strcmp(inval
, "preop_none") == 0)
1239 preop_val
= WDOG_PREOP_NONE
;
1240 else if (strcmp(inval
, "preop_panic") == 0)
1241 preop_val
= WDOG_PREOP_PANIC
;
1242 else if (strcmp(inval
, "preop_give_data") == 0)
1243 preop_val
= WDOG_PREOP_GIVE_DATA
;
1246 strcpy(preop
, inval
);
1250 static void check_parms(void)
1256 if (preaction_val
== WDOG_PRETIMEOUT_NMI
) {
1258 if (preop_val
== WDOG_PREOP_GIVE_DATA
) {
1259 printk(KERN_WARNING PFX
"Pretimeout op is to give data"
1260 " but NMI pretimeout is enabled, setting"
1261 " pretimeout op to none\n");
1262 preop_op("preop_none", NULL
);
1266 if (do_nmi
&& !nmi_handler_registered
) {
1267 rv
= register_die_notifier(&ipmi_nmi_handler
);
1269 printk(KERN_WARNING PFX
1270 "Can't register nmi handler\n");
1273 nmi_handler_registered
= 1;
1274 } else if (!do_nmi
&& nmi_handler_registered
) {
1275 unregister_die_notifier(&ipmi_nmi_handler
);
1276 nmi_handler_registered
= 0;
1281 static int __init
ipmi_wdog_init(void)
1285 if (action_op(action
, NULL
)) {
1286 action_op("reset", NULL
);
1287 printk(KERN_INFO PFX
"Unknown action '%s', defaulting to"
1288 " reset\n", action
);
1291 if (preaction_op(preaction
, NULL
)) {
1292 preaction_op("pre_none", NULL
);
1293 printk(KERN_INFO PFX
"Unknown preaction '%s', defaulting to"
1294 " none\n", preaction
);
1297 if (preop_op(preop
, NULL
)) {
1298 preop_op("preop_none", NULL
);
1299 printk(KERN_INFO PFX
"Unknown preop '%s', defaulting to"
1305 register_reboot_notifier(&wdog_reboot_notifier
);
1306 atomic_notifier_chain_register(&panic_notifier_list
,
1307 &wdog_panic_notifier
);
1309 rv
= ipmi_smi_watcher_register(&smi_watcher
);
1312 if (nmi_handler_registered
)
1313 unregister_die_notifier(&ipmi_nmi_handler
);
1315 atomic_notifier_chain_unregister(&panic_notifier_list
,
1316 &wdog_panic_notifier
);
1317 unregister_reboot_notifier(&wdog_reboot_notifier
);
1318 printk(KERN_WARNING PFX
"can't register smi watcher\n");
1322 printk(KERN_INFO PFX
"driver initialized\n");
1327 static void __exit
ipmi_wdog_exit(void)
1329 ipmi_smi_watcher_unregister(&smi_watcher
);
1330 ipmi_unregister_watchdog(watchdog_ifnum
);
1333 if (nmi_handler_registered
)
1334 unregister_die_notifier(&ipmi_nmi_handler
);
1337 atomic_notifier_chain_unregister(&panic_notifier_list
,
1338 &wdog_panic_notifier
);
1339 unregister_reboot_notifier(&wdog_reboot_notifier
);
1341 module_exit(ipmi_wdog_exit
);
1342 module_init(ipmi_wdog_init
);
1343 MODULE_LICENSE("GPL");
1344 MODULE_AUTHOR("Corey Minyard <minyard@mvista.com>");
1345 MODULE_DESCRIPTION("watchdog timer based upon the IPMI interface.");