2
0
mirror of https://github.com/edk2-porting/linux-next.git synced 2024-12-26 14:14:01 +08:00
linux-next/drivers/char/ipmi/ipmi_watchdog.c

1254 lines
31 KiB
C
Raw Normal View History

/*
* ipmi_watchdog.c
*
* A watchdog timer based upon the IPMI interface.
*
* Author: MontaVista Software, Inc.
* Corey Minyard <minyard@mvista.com>
* source@mvista.com
*
* Copyright 2002 MontaVista Software Inc.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2 of the License, or (at your
* option) any later version.
*
*
* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
* OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
* ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
* TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
* USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* You should have received a copy of the GNU General Public License along
* with this program; if not, write to the Free Software Foundation, Inc.,
* 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#include <linux/module.h>
#include <linux/moduleparam.h>
#include <linux/ipmi.h>
#include <linux/ipmi_smi.h>
#include <linux/watchdog.h>
#include <linux/miscdevice.h>
#include <linux/init.h>
#include <linux/completion.h>
#include <linux/rwsem.h>
#include <linux/errno.h>
#include <asm/uaccess.h>
#include <linux/notifier.h>
#include <linux/nmi.h>
#include <linux/reboot.h>
#include <linux/wait.h>
#include <linux/poll.h>
#include <linux/string.h>
#include <linux/ctype.h>
#include <asm/atomic.h>
#ifdef CONFIG_X86_LOCAL_APIC
#include <asm/apic.h>
#endif
#define PFX "IPMI Watchdog: "
/*
* The IPMI command/response information for the watchdog timer.
*/
/* values for byte 1 of the set command, byte 2 of the get response. */
#define WDOG_DONT_LOG (1 << 7)
#define WDOG_DONT_STOP_ON_SET (1 << 6)
#define WDOG_SET_TIMER_USE(byte, use) \
byte = ((byte) & 0xf8) | ((use) & 0x7)
#define WDOG_GET_TIMER_USE(byte) ((byte) & 0x7)
#define WDOG_TIMER_USE_BIOS_FRB2 1
#define WDOG_TIMER_USE_BIOS_POST 2
#define WDOG_TIMER_USE_OS_LOAD 3
#define WDOG_TIMER_USE_SMS_OS 4
#define WDOG_TIMER_USE_OEM 5
/* values for byte 2 of the set command, byte 3 of the get response. */
#define WDOG_SET_PRETIMEOUT_ACT(byte, use) \
byte = ((byte) & 0x8f) | (((use) & 0x7) << 4)
#define WDOG_GET_PRETIMEOUT_ACT(byte) (((byte) >> 4) & 0x7)
#define WDOG_PRETIMEOUT_NONE 0
#define WDOG_PRETIMEOUT_SMI 1
#define WDOG_PRETIMEOUT_NMI 2
#define WDOG_PRETIMEOUT_MSG_INT 3
/* Operations that can be performed on a pretimout. */
#define WDOG_PREOP_NONE 0
#define WDOG_PREOP_PANIC 1
#define WDOG_PREOP_GIVE_DATA 2 /* Cause data to be available to
read. Doesn't work in NMI
mode. */
/* Actions to perform on a full timeout. */
#define WDOG_SET_TIMEOUT_ACT(byte, use) \
byte = ((byte) & 0xf8) | ((use) & 0x7)
#define WDOG_GET_TIMEOUT_ACT(byte) ((byte) & 0x7)
#define WDOG_TIMEOUT_NONE 0
#define WDOG_TIMEOUT_RESET 1
#define WDOG_TIMEOUT_POWER_DOWN 2
#define WDOG_TIMEOUT_POWER_CYCLE 3
/* Byte 3 of the get command, byte 4 of the get response is the
pre-timeout in seconds. */
/* Bits for setting byte 4 of the set command, byte 5 of the get response. */
#define WDOG_EXPIRE_CLEAR_BIOS_FRB2 (1 << 1)
#define WDOG_EXPIRE_CLEAR_BIOS_POST (1 << 2)
#define WDOG_EXPIRE_CLEAR_OS_LOAD (1 << 3)
#define WDOG_EXPIRE_CLEAR_SMS_OS (1 << 4)
#define WDOG_EXPIRE_CLEAR_OEM (1 << 5)
/* Setting/getting the watchdog timer value. This is for bytes 5 and
6 (the timeout time) of the set command, and bytes 6 and 7 (the
timeout time) and 8 and 9 (the current countdown value) of the
response. The timeout value is given in seconds (in the command it
is 100ms intervals). */
#define WDOG_SET_TIMEOUT(byte1, byte2, val) \
(byte1) = (((val) * 10) & 0xff), (byte2) = (((val) * 10) >> 8)
#define WDOG_GET_TIMEOUT(byte1, byte2) \
(((byte1) | ((byte2) << 8)) / 10)
#define IPMI_WDOG_RESET_TIMER 0x22
#define IPMI_WDOG_SET_TIMER 0x24
#define IPMI_WDOG_GET_TIMER 0x25
/* These are here until the real ones get into the watchdog.h interface. */
#ifndef WDIOC_GETTIMEOUT
#define WDIOC_GETTIMEOUT _IOW(WATCHDOG_IOCTL_BASE, 20, int)
#endif
#ifndef WDIOC_SET_PRETIMEOUT
#define WDIOC_SET_PRETIMEOUT _IOW(WATCHDOG_IOCTL_BASE, 21, int)
#endif
#ifndef WDIOC_GET_PRETIMEOUT
#define WDIOC_GET_PRETIMEOUT _IOW(WATCHDOG_IOCTL_BASE, 22, int)
#endif
static int nowayout = WATCHDOG_NOWAYOUT;
static ipmi_user_t watchdog_user = NULL;
static int watchdog_ifnum;
/* Default the timeout to 10 seconds. */
static int timeout = 10;
/* The pre-timeout is disabled by default. */
static int pretimeout = 0;
/* Default action is to reset the board on a timeout. */
static unsigned char action_val = WDOG_TIMEOUT_RESET;
static char action[16] = "reset";
static unsigned char preaction_val = WDOG_PRETIMEOUT_NONE;
static char preaction[16] = "pre_none";
static unsigned char preop_val = WDOG_PREOP_NONE;
static char preop[16] = "preop_none";
static DEFINE_SPINLOCK(ipmi_read_lock);
static char data_to_read = 0;
static DECLARE_WAIT_QUEUE_HEAD(read_q);
static struct fasync_struct *fasync_q = NULL;
static char pretimeout_since_last_heartbeat = 0;
static char expect_close;
static int ifnum_to_use = -1;
static DECLARE_RWSEM(register_sem);
/* Parameters to ipmi_set_timeout */
#define IPMI_SET_TIMEOUT_NO_HB 0
#define IPMI_SET_TIMEOUT_HB_IF_NECESSARY 1
#define IPMI_SET_TIMEOUT_FORCE_HB 2
static int ipmi_set_timeout(int do_heartbeat);
static void ipmi_register_watchdog(int ipmi_intf);
static void ipmi_unregister_watchdog(int ipmi_intf);
/* If true, the driver will start running as soon as it is configured
and ready. */
static int start_now = 0;
static int set_param_int(const char *val, struct kernel_param *kp)
{
char *endp;
int l;
int rv = 0;
if (!val)
return -EINVAL;
l = simple_strtoul(val, &endp, 0);
if (endp == val)
return -EINVAL;
down_read(&register_sem);
*((int *)kp->arg) = l;
if (watchdog_user)
rv = ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY);
up_read(&register_sem);
return rv;
}
static int get_param_int(char *buffer, struct kernel_param *kp)
{
return sprintf(buffer, "%i", *((int *)kp->arg));
}
typedef int (*action_fn)(const char *intval, char *outval);
static int action_op(const char *inval, char *outval);
static int preaction_op(const char *inval, char *outval);
static int preop_op(const char *inval, char *outval);
static void check_parms(void);
static int set_param_str(const char *val, struct kernel_param *kp)
{
action_fn fn = (action_fn) kp->arg;
int rv = 0;
char *dup, *s;
dup = kstrdup(val, GFP_KERNEL);
if (!dup)
return -ENOMEM;
s = strstrip(dup);
down_read(&register_sem);
rv = fn(s, NULL);
if (rv)
goto out_unlock;
check_parms();
if (watchdog_user)
rv = ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY);
out_unlock:
up_read(&register_sem);
kfree(dup);
return rv;
}
static int get_param_str(char *buffer, struct kernel_param *kp)
{
action_fn fn = (action_fn) kp->arg;
int rv;
rv = fn(NULL, buffer);
if (rv)
return rv;
return strlen(buffer);
}
static int set_param_wdog_ifnum(const char *val, struct kernel_param *kp)
{
int rv = param_set_int(val, kp);
if (rv)
return rv;
if ((ifnum_to_use < 0) || (ifnum_to_use == watchdog_ifnum))
return 0;
ipmi_unregister_watchdog(watchdog_ifnum);
ipmi_register_watchdog(ifnum_to_use);
return 0;
}
module_param_call(ifnum_to_use, set_param_wdog_ifnum, get_param_int,
&ifnum_to_use, 0644);
MODULE_PARM_DESC(ifnum_to_use, "The interface number to use for the watchdog "
"timer. Setting to -1 defaults to the first registered "
"interface");
module_param_call(timeout, set_param_int, get_param_int, &timeout, 0644);
MODULE_PARM_DESC(timeout, "Timeout value in seconds.");
module_param_call(pretimeout, set_param_int, get_param_int, &pretimeout, 0644);
MODULE_PARM_DESC(pretimeout, "Pretimeout value in seconds.");
module_param_call(action, set_param_str, get_param_str, action_op, 0644);
MODULE_PARM_DESC(action, "Timeout action. One of: "
"reset, none, power_cycle, power_off.");
module_param_call(preaction, set_param_str, get_param_str, preaction_op, 0644);
MODULE_PARM_DESC(preaction, "Pretimeout action. One of: "
"pre_none, pre_smi, pre_nmi, pre_int.");
module_param_call(preop, set_param_str, get_param_str, preop_op, 0644);
MODULE_PARM_DESC(preop, "Pretimeout driver operation. One of: "
"preop_none, preop_panic, preop_give_data.");
module_param(start_now, int, 0444);
MODULE_PARM_DESC(start_now, "Set to 1 to start the watchdog as"
"soon as the driver is loaded.");
module_param(nowayout, int, 0644);
MODULE_PARM_DESC(nowayout, "Watchdog cannot be stopped once started "
"(default=CONFIG_WATCHDOG_NOWAYOUT)");
/* Default state of the timer. */
static unsigned char ipmi_watchdog_state = WDOG_TIMEOUT_NONE;
/* If shutting down via IPMI, we ignore the heartbeat. */
static int ipmi_ignore_heartbeat = 0;
/* Is someone using the watchdog? Only one user is allowed. */
static unsigned long ipmi_wdog_open = 0;
/* If set to 1, the heartbeat command will set the state to reset and
start the timer. The timer doesn't normally run when the driver is
first opened until the heartbeat is set the first time, this
variable is used to accomplish this. */
static int ipmi_start_timer_on_heartbeat = 0;
/* IPMI version of the BMC. */
static unsigned char ipmi_version_major;
static unsigned char ipmi_version_minor;
/* If a pretimeout occurs, this is used to allow only one panic to happen. */
static atomic_t preop_panic_excl = ATOMIC_INIT(-1);
static int ipmi_heartbeat(void);
static void panic_halt_ipmi_heartbeat(void);
/* We use a mutex to make sure that only one thing can send a set
timeout at one time, because we only have one copy of the data.
The mutex is claimed when the set_timeout is sent and freed
when both messages are free. */
static atomic_t set_timeout_tofree = ATOMIC_INIT(0);
static DEFINE_MUTEX(set_timeout_lock);
static DECLARE_COMPLETION(set_timeout_wait);
static void set_timeout_free_smi(struct ipmi_smi_msg *msg)
{
if (atomic_dec_and_test(&set_timeout_tofree))
complete(&set_timeout_wait);
}
static void set_timeout_free_recv(struct ipmi_recv_msg *msg)
{
if (atomic_dec_and_test(&set_timeout_tofree))
complete(&set_timeout_wait);
}
static struct ipmi_smi_msg set_timeout_smi_msg =
{
.done = set_timeout_free_smi
};
static struct ipmi_recv_msg set_timeout_recv_msg =
{
.done = set_timeout_free_recv
};
static int i_ipmi_set_timeout(struct ipmi_smi_msg *smi_msg,
struct ipmi_recv_msg *recv_msg,
int *send_heartbeat_now)
{
struct kernel_ipmi_msg msg;
unsigned char data[6];
int rv;
struct ipmi_system_interface_addr addr;
int hbnow = 0;
data[0] = 0;
WDOG_SET_TIMER_USE(data[0], WDOG_TIMER_USE_SMS_OS);
if ((ipmi_version_major > 1)
|| ((ipmi_version_major == 1) && (ipmi_version_minor >= 5)))
{
/* This is an IPMI 1.5-only feature. */
data[0] |= WDOG_DONT_STOP_ON_SET;
} else if (ipmi_watchdog_state != WDOG_TIMEOUT_NONE) {
/* In ipmi 1.0, setting the timer stops the watchdog, we
need to start it back up again. */
hbnow = 1;
}
data[1] = 0;
WDOG_SET_TIMEOUT_ACT(data[1], ipmi_watchdog_state);
if ((pretimeout > 0) && (ipmi_watchdog_state != WDOG_TIMEOUT_NONE)) {
WDOG_SET_PRETIMEOUT_ACT(data[1], preaction_val);
data[2] = pretimeout;
} else {
WDOG_SET_PRETIMEOUT_ACT(data[1], WDOG_PRETIMEOUT_NONE);
data[2] = 0; /* No pretimeout. */
}
data[3] = 0;
WDOG_SET_TIMEOUT(data[4], data[5], timeout);
addr.addr_type = IPMI_SYSTEM_INTERFACE_ADDR_TYPE;
addr.channel = IPMI_BMC_CHANNEL;
addr.lun = 0;
msg.netfn = 0x06;
msg.cmd = IPMI_WDOG_SET_TIMER;
msg.data = data;
msg.data_len = sizeof(data);
rv = ipmi_request_supply_msgs(watchdog_user,
(struct ipmi_addr *) &addr,
0,
&msg,
NULL,
smi_msg,
recv_msg,
1);
if (rv) {
printk(KERN_WARNING PFX "set timeout error: %d\n",
rv);
}
if (send_heartbeat_now)
*send_heartbeat_now = hbnow;
return rv;
}
static int ipmi_set_timeout(int do_heartbeat)
{
int send_heartbeat_now;
int rv;
/* We can only send one of these at a time. */
mutex_lock(&set_timeout_lock);
atomic_set(&set_timeout_tofree, 2);
rv = i_ipmi_set_timeout(&set_timeout_smi_msg,
&set_timeout_recv_msg,
&send_heartbeat_now);
if (rv) {
mutex_unlock(&set_timeout_lock);
goto out;
}
wait_for_completion(&set_timeout_wait);
if ((do_heartbeat == IPMI_SET_TIMEOUT_FORCE_HB)
|| ((send_heartbeat_now)
&& (do_heartbeat == IPMI_SET_TIMEOUT_HB_IF_NECESSARY)))
{
rv = ipmi_heartbeat();
}
mutex_unlock(&set_timeout_lock);
out:
return rv;
}
static void dummy_smi_free(struct ipmi_smi_msg *msg)
{
}
static void dummy_recv_free(struct ipmi_recv_msg *msg)
{
}
static struct ipmi_smi_msg panic_halt_smi_msg =
{
.done = dummy_smi_free
};
static struct ipmi_recv_msg panic_halt_recv_msg =
{
.done = dummy_recv_free
};
/* Special call, doesn't claim any locks. This is only to be called
at panic or halt time, in run-to-completion mode, when the caller
is the only CPU and the only thing that will be going is these IPMI
calls. */
static void panic_halt_ipmi_set_timeout(void)
{
int send_heartbeat_now;
int rv;
rv = i_ipmi_set_timeout(&panic_halt_smi_msg,
&panic_halt_recv_msg,
&send_heartbeat_now);
if (!rv) {
if (send_heartbeat_now)
panic_halt_ipmi_heartbeat();
}
}
/* We use a semaphore to make sure that only one thing can send a
heartbeat at one time, because we only have one copy of the data.
The semaphore is claimed when the set_timeout is sent and freed
when both messages are free. */
static atomic_t heartbeat_tofree = ATOMIC_INIT(0);
static DEFINE_MUTEX(heartbeat_lock);
static DECLARE_COMPLETION(heartbeat_wait);
static void heartbeat_free_smi(struct ipmi_smi_msg *msg)
{
if (atomic_dec_and_test(&heartbeat_tofree))
complete(&heartbeat_wait);
}
static void heartbeat_free_recv(struct ipmi_recv_msg *msg)
{
if (atomic_dec_and_test(&heartbeat_tofree))
complete(&heartbeat_wait);
}
static struct ipmi_smi_msg heartbeat_smi_msg =
{
.done = heartbeat_free_smi
};
static struct ipmi_recv_msg heartbeat_recv_msg =
{
.done = heartbeat_free_recv
};
static struct ipmi_smi_msg panic_halt_heartbeat_smi_msg =
{
.done = dummy_smi_free
};
static struct ipmi_recv_msg panic_halt_heartbeat_recv_msg =
{
.done = dummy_recv_free
};
static int ipmi_heartbeat(void)
{
struct kernel_ipmi_msg msg;
int rv;
struct ipmi_system_interface_addr addr;
if (ipmi_ignore_heartbeat) {
return 0;
}
if (ipmi_start_timer_on_heartbeat) {
ipmi_start_timer_on_heartbeat = 0;
ipmi_watchdog_state = action_val;
return ipmi_set_timeout(IPMI_SET_TIMEOUT_FORCE_HB);
} else if (pretimeout_since_last_heartbeat) {
/* A pretimeout occurred, make sure we set the timeout.
We don't want to set the action, though, we want to
leave that alone (thus it can't be combined with the
above operation. */
pretimeout_since_last_heartbeat = 0;
return ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY);
}
mutex_lock(&heartbeat_lock);
atomic_set(&heartbeat_tofree, 2);
/* Don't reset the timer if we have the timer turned off, that
re-enables the watchdog. */
if (ipmi_watchdog_state == WDOG_TIMEOUT_NONE) {
mutex_unlock(&heartbeat_lock);
return 0;
}
addr.addr_type = IPMI_SYSTEM_INTERFACE_ADDR_TYPE;
addr.channel = IPMI_BMC_CHANNEL;
addr.lun = 0;
msg.netfn = 0x06;
msg.cmd = IPMI_WDOG_RESET_TIMER;
msg.data = NULL;
msg.data_len = 0;
rv = ipmi_request_supply_msgs(watchdog_user,
(struct ipmi_addr *) &addr,
0,
&msg,
NULL,
&heartbeat_smi_msg,
&heartbeat_recv_msg,
1);
if (rv) {
mutex_unlock(&heartbeat_lock);
printk(KERN_WARNING PFX "heartbeat failure: %d\n",
rv);
return rv;
}
/* Wait for the heartbeat to be sent. */
wait_for_completion(&heartbeat_wait);
if (heartbeat_recv_msg.msg.data[0] != 0) {
/* Got an error in the heartbeat response. It was already
reported in ipmi_wdog_msg_handler, but we should return
an error here. */
rv = -EINVAL;
}
mutex_unlock(&heartbeat_lock);
return rv;
}
static void panic_halt_ipmi_heartbeat(void)
{
struct kernel_ipmi_msg msg;
struct ipmi_system_interface_addr addr;
/* Don't reset the timer if we have the timer turned off, that
re-enables the watchdog. */
if (ipmi_watchdog_state == WDOG_TIMEOUT_NONE)
return;
addr.addr_type = IPMI_SYSTEM_INTERFACE_ADDR_TYPE;
addr.channel = IPMI_BMC_CHANNEL;
addr.lun = 0;
msg.netfn = 0x06;
msg.cmd = IPMI_WDOG_RESET_TIMER;
msg.data = NULL;
msg.data_len = 0;
ipmi_request_supply_msgs(watchdog_user,
(struct ipmi_addr *) &addr,
0,
&msg,
NULL,
&panic_halt_heartbeat_smi_msg,
&panic_halt_heartbeat_recv_msg,
1);
}
static struct watchdog_info ident =
{
.options = 0, /* WDIOF_SETTIMEOUT, */
.firmware_version = 1,
.identity = "IPMI"
};
static int ipmi_ioctl(struct inode *inode, struct file *file,
unsigned int cmd, unsigned long arg)
{
void __user *argp = (void __user *)arg;
int i;
int val;
switch(cmd) {
case WDIOC_GETSUPPORT:
i = copy_to_user(argp, &ident, sizeof(ident));
return i ? -EFAULT : 0;
case WDIOC_SETTIMEOUT:
i = copy_from_user(&val, argp, sizeof(int));
if (i)
return -EFAULT;
timeout = val;
return ipmi_set_timeout(IPMI_SET_TIMEOUT_HB_IF_NECESSARY);
case WDIOC_GETTIMEOUT:
i = copy_to_user(argp, &timeout, sizeof(timeout));
if (i)
return -EFAULT;
return 0;
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);
} else {
printk(KERN_CRIT PFX
"Unexpected close, not stopping watchdog!\n");
ipmi_heartbeat();
}
clear_bit(0, &ipmi_wdog_open);
}
ipmi_fasync (-1, filep, 0);
expect_close = 0;
return 0;
}
static const 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 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) {
if (atomic_inc_and_test(&preop_panic_excl))
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;
if ((ifnum_to_use >= 0) && (ifnum_to_use != ipmi_intf))
goto out;
watchdog_ifnum = ipmi_intf;
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");
}
}
static void ipmi_unregister_watchdog(int ipmi_intf)
{
int rv;
down_write(&register_sem);
if (!watchdog_user)
goto out;
if (watchdog_ifnum != ipmi_intf)
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))
schedule_timeout_uninterruptible(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);
}
#ifdef HAVE_NMI_HANDLER
static int
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 21:55:46 +08:00
ipmi_nmi(void *dev_id, int cpu, int handled)
{
/* If we are not expecting a timeout, ignore it. */
if (ipmi_watchdog_state == WDOG_TIMEOUT_NONE)
return NOTIFY_DONE;
/* If no one else handled the NMI, we assume it was the IPMI
watchdog. */
if ((!handled) && (preop_val == WDOG_PREOP_PANIC)) {
/* 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;
if (atomic_inc_and_test(&preop_panic_excl))
panic(PFX "pre-timeout");
}
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. */
};
int nmi_handler_registered;
#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 if (ipmi_watchdog_state != WDOG_TIMEOUT_NONE) {
/* Set a long timer to let the reboot happens, but
reboot if it hangs, but only if the watchdog
timer was already running. */
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 to extend
the watchdog timer to a reasonable value to complete the
panic, if the watchdog timer is running. Plus the
pretimeout is meaningless at panic time. */
if (watchdog_user && !panic_event_handled &&
ipmi_watchdog_state != WDOG_TIMEOUT_NONE) {
/* Make sure we do this only once. */
panic_event_handled = 1;
timeout = 255;
pretimeout = 0;
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, struct device *device)
{
ipmi_register_watchdog(if_num);
}
static void ipmi_smi_gone(int if_num)
{
ipmi_unregister_watchdog(if_num);
}
static struct ipmi_smi_watcher smi_watcher =
{
.owner = THIS_MODULE,
.new_smi = ipmi_new_smi,
.smi_gone = ipmi_smi_gone
};
static int action_op(const char *inval, char *outval)
{
if (outval)
strcpy(outval, action);
if (!inval)
return 0;
if (strcmp(inval, "reset") == 0)
action_val = WDOG_TIMEOUT_RESET;
else if (strcmp(inval, "none") == 0)
action_val = WDOG_TIMEOUT_NONE;
else if (strcmp(inval, "power_cycle") == 0)
action_val = WDOG_TIMEOUT_POWER_CYCLE;
else if (strcmp(inval, "power_off") == 0)
action_val = WDOG_TIMEOUT_POWER_DOWN;
else
return -EINVAL;
strcpy(action, inval);
return 0;
}
static int preaction_op(const char *inval, char *outval)
{
if (outval)
strcpy(outval, preaction);
if (!inval)
return 0;
if (strcmp(inval, "pre_none") == 0)
preaction_val = WDOG_PRETIMEOUT_NONE;
else if (strcmp(inval, "pre_smi") == 0)
preaction_val = WDOG_PRETIMEOUT_SMI;
#ifdef HAVE_NMI_HANDLER
else if (strcmp(inval, "pre_nmi") == 0)
preaction_val = WDOG_PRETIMEOUT_NMI;
#endif
else if (strcmp(inval, "pre_int") == 0)
preaction_val = WDOG_PRETIMEOUT_MSG_INT;
else
return -EINVAL;
strcpy(preaction, inval);
return 0;
}
static int preop_op(const char *inval, char *outval)
{
if (outval)
strcpy(outval, preop);
if (!inval)
return 0;
if (strcmp(inval, "preop_none") == 0)
preop_val = WDOG_PREOP_NONE;
else if (strcmp(inval, "preop_panic") == 0)
preop_val = WDOG_PREOP_PANIC;
else if (strcmp(inval, "preop_give_data") == 0)
preop_val = WDOG_PREOP_GIVE_DATA;
else
return -EINVAL;
strcpy(preop, inval);
return 0;
}
static void check_parms(void)
{
#ifdef HAVE_NMI_HANDLER
int do_nmi = 0;
int rv;
if (preaction_val == WDOG_PRETIMEOUT_NMI) {
do_nmi = 1;
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_op("preop_none", NULL);
do_nmi = 0;
}
#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;
do_nmi = 0;
}
#endif
}
if (do_nmi && !nmi_handler_registered) {
rv = request_nmi(&ipmi_nmi_handler);
if (rv) {
printk(KERN_WARNING PFX
"Can't register nmi handler\n");
return;
} else
nmi_handler_registered = 1;
} else if (!do_nmi && nmi_handler_registered) {
release_nmi(&ipmi_nmi_handler);
nmi_handler_registered = 0;
}
#endif
}
static int __init ipmi_wdog_init(void)
{
int rv;
if (action_op(action, NULL)) {
action_op("reset", NULL);
printk(KERN_INFO PFX "Unknown action '%s', defaulting to"
" reset\n", action);
}
if (preaction_op(preaction, NULL)) {
preaction_op("pre_none", NULL);
printk(KERN_INFO PFX "Unknown preaction '%s', defaulting to"
" none\n", preaction);
}
if (preop_op(preop, NULL)) {
preop_op("preop_none", NULL);
printk(KERN_INFO PFX "Unknown preop '%s', defaulting to"
" none\n", preop);
}
check_parms();
register_reboot_notifier(&wdog_reboot_notifier);
atomic_notifier_chain_register(&panic_notifier_list,
&wdog_panic_notifier);
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
atomic_notifier_chain_unregister(&panic_notifier_list,
&wdog_panic_notifier);
unregister_reboot_notifier(&wdog_reboot_notifier);
printk(KERN_WARNING PFX "can't register smi watcher\n");
return rv;
}
printk(KERN_INFO PFX "driver initialized\n");
return 0;
}
static void __exit ipmi_wdog_exit(void)
{
ipmi_smi_watcher_unregister(&smi_watcher);
ipmi_unregister_watchdog(watchdog_ifnum);
#ifdef HAVE_NMI_HANDLER
if (nmi_handler_registered)
release_nmi(&ipmi_nmi_handler);
#endif
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
atomic_notifier_chain_unregister(&panic_notifier_list,
&wdog_panic_notifier);
unregister_reboot_notifier(&wdog_reboot_notifier);
}
module_exit(ipmi_wdog_exit);
module_init(ipmi_wdog_init);
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Corey Minyard <minyard@mvista.com>");
MODULE_DESCRIPTION("watchdog timer based upon the IPMI interface.");