Commit c2896def authored by Linus Torvalds's avatar Linus Torvalds

Merge branch 'ipmi' (emailed ipmi fixes)

Merge ipmi fixes from Corey Minyard:
 "Things collected since last kernel release.

  Some of these are pretty important.  The first three are bug fixes.
  The next two are to hopefully make everyone happy about allowing
  ACPI to be on all the time and not have IPMI have an effect on the
  system when not in use.  The last is a little cleanup"

* emailed patches from Corey Minyard <cminyard@mvista.com>:
  ipmi: boolify some things
  ipmi: Turn off all activity on an idle ipmi interface
  ipmi: Turn off default probing of interfaces
  ipmi: Reset the KCS timeout when starting error recovery
  ipmi: Fix a race restarting the timer
  Char: ipmi_bt_sm, fix infinite loop
parents 88764e0a 7aefac26
......@@ -50,6 +50,18 @@ config IPMI_SI
Currently, only KCS and SMIC are supported. If
you are using IPMI, you should probably say "y" here.
config IPMI_SI_PROBE_DEFAULTS
bool 'Probe for all possible IPMI system interfaces by default'
default n
depends on IPMI_SI
help
Modern systems will usually expose IPMI interfaces via a discoverable
firmware mechanism such as ACPI or DMI. Older systems do not, and so
the driver is forced to probe hardware manually. This may cause boot
delays. Say "n" here to disable this manual probing. IPMI will then
only be available on older systems if the "ipmi_si_intf.trydefaults=1"
boot argument is passed.
config IPMI_WATCHDOG
tristate 'IPMI Watchdog Timer'
help
......
......@@ -352,7 +352,7 @@ static inline void write_all_bytes(struct si_sm_data *bt)
static inline int read_all_bytes(struct si_sm_data *bt)
{
unsigned char i;
unsigned int i;
/*
* length is "framing info", minimum = 4: NetFn, Seq, Cmd, cCode.
......
......@@ -251,8 +251,9 @@ static inline int check_obf(struct si_sm_data *kcs, unsigned char status,
if (!GET_STATUS_OBF(status)) {
kcs->obf_timeout -= time;
if (kcs->obf_timeout < 0) {
start_error_recovery(kcs, "OBF not ready in time");
return 1;
kcs->obf_timeout = OBF_RETRY_TIMEOUT;
start_error_recovery(kcs, "OBF not ready in time");
return 1;
}
return 0;
}
......
......@@ -55,6 +55,7 @@ static struct ipmi_recv_msg *ipmi_alloc_recv_msg(void);
static int ipmi_init_msghandler(void);
static void smi_recv_tasklet(unsigned long);
static void handle_new_recv_msgs(ipmi_smi_t intf);
static void need_waiter(ipmi_smi_t intf);
static int initialized;
......@@ -73,14 +74,28 @@ static struct proc_dir_entry *proc_ipmi_root;
*/
#define MAX_MSG_TIMEOUT 60000
/* Call every ~1000 ms. */
#define IPMI_TIMEOUT_TIME 1000
/* How many jiffies does it take to get to the timeout time. */
#define IPMI_TIMEOUT_JIFFIES ((IPMI_TIMEOUT_TIME * HZ) / 1000)
/*
* Request events from the queue every second (this is the number of
* IPMI_TIMEOUT_TIMES between event requests). Hopefully, in the
* future, IPMI will add a way to know immediately if an event is in
* the queue and this silliness can go away.
*/
#define IPMI_REQUEST_EV_TIME (1000 / (IPMI_TIMEOUT_TIME))
/*
* The main "user" data structure.
*/
struct ipmi_user {
struct list_head link;
/* Set to "0" when the user is destroyed. */
int valid;
/* Set to false when the user is destroyed. */
bool valid;
struct kref refcount;
......@@ -92,7 +107,7 @@ struct ipmi_user {
ipmi_smi_t intf;
/* Does this interface receive IPMI events? */
int gets_events;
bool gets_events;
};
struct cmd_rcvr {
......@@ -383,6 +398,9 @@ struct ipmi_smi {
unsigned int waiting_events_count; /* How many events in queue? */
char delivering_events;
char event_msg_printed;
atomic_t event_waiters;
unsigned int ticks_to_req_ev;
int last_needs_timer;
/*
* The event receiver for my BMC, only really used at panic
......@@ -395,7 +413,7 @@ struct ipmi_smi {
/* For handling of maintenance mode. */
int maintenance_mode;
int maintenance_mode_enable;
bool maintenance_mode_enable;
int auto_maintenance_timeout;
spinlock_t maintenance_mode_lock; /* Used in a timer... */
......@@ -451,7 +469,6 @@ static DEFINE_MUTEX(ipmi_interfaces_mutex);
static LIST_HEAD(smi_watchers);
static DEFINE_MUTEX(smi_watchers_mutex);
#define ipmi_inc_stat(intf, stat) \
atomic_inc(&(intf)->stats[IPMI_STAT_ ## stat])
#define ipmi_get_stat(intf, stat) \
......@@ -772,6 +789,7 @@ static int intf_next_seq(ipmi_smi_t intf,
*seq = i;
*seqid = intf->seq_table[i].seqid;
intf->curr_seq = (i+1)%IPMI_IPMB_NUM_SEQ;
need_waiter(intf);
} else {
rv = -EAGAIN;
}
......@@ -941,7 +959,7 @@ int ipmi_create_user(unsigned int if_num,
new_user->handler = handler;
new_user->handler_data = handler_data;
new_user->intf = intf;
new_user->gets_events = 0;
new_user->gets_events = false;
if (!try_module_get(intf->handlers->owner)) {
rv = -ENODEV;
......@@ -962,10 +980,15 @@ int ipmi_create_user(unsigned int if_num,
*/
mutex_unlock(&ipmi_interfaces_mutex);
new_user->valid = 1;
new_user->valid = true;
spin_lock_irqsave(&intf->seq_lock, flags);
list_add_rcu(&new_user->link, &intf->users);
spin_unlock_irqrestore(&intf->seq_lock, flags);
if (handler->ipmi_watchdog_pretimeout) {
/* User wants pretimeouts, so make sure to watch for them. */
if (atomic_inc_return(&intf->event_waiters) == 1)
need_waiter(intf);
}
*user = new_user;
return 0;
......@@ -1019,7 +1042,13 @@ int ipmi_destroy_user(ipmi_user_t user)
struct cmd_rcvr *rcvr;
struct cmd_rcvr *rcvrs = NULL;
user->valid = 0;
user->valid = false;
if (user->handler->ipmi_watchdog_pretimeout)
atomic_dec(&intf->event_waiters);
if (user->gets_events)
atomic_dec(&intf->event_waiters);
/* Remove the user from the interface's sequence table. */
spin_lock_irqsave(&intf->seq_lock, flags);
......@@ -1155,25 +1184,23 @@ int ipmi_set_maintenance_mode(ipmi_user_t user, int mode)
if (intf->maintenance_mode != mode) {
switch (mode) {
case IPMI_MAINTENANCE_MODE_AUTO:
intf->maintenance_mode = mode;
intf->maintenance_mode_enable
= (intf->auto_maintenance_timeout > 0);
break;
case IPMI_MAINTENANCE_MODE_OFF:
intf->maintenance_mode = mode;
intf->maintenance_mode_enable = 0;
intf->maintenance_mode_enable = false;
break;
case IPMI_MAINTENANCE_MODE_ON:
intf->maintenance_mode = mode;
intf->maintenance_mode_enable = 1;
intf->maintenance_mode_enable = true;
break;
default:
rv = -EINVAL;
goto out_unlock;
}
intf->maintenance_mode = mode;
maintenance_mode_update(intf);
}
......@@ -1184,7 +1211,7 @@ int ipmi_set_maintenance_mode(ipmi_user_t user, int mode)
}
EXPORT_SYMBOL(ipmi_set_maintenance_mode);
int ipmi_set_gets_events(ipmi_user_t user, int val)
int ipmi_set_gets_events(ipmi_user_t user, bool val)
{
unsigned long flags;
ipmi_smi_t intf = user->intf;
......@@ -1194,8 +1221,18 @@ int ipmi_set_gets_events(ipmi_user_t user, int val)
INIT_LIST_HEAD(&msgs);
spin_lock_irqsave(&intf->events_lock, flags);
if (user->gets_events == val)
goto out;
user->gets_events = val;
if (val) {
if (atomic_inc_return(&intf->event_waiters) == 1)
need_waiter(intf);
} else {
atomic_dec(&intf->event_waiters);
}
if (intf->delivering_events)
/*
* Another thread is delivering events for this, so
......@@ -1289,6 +1326,9 @@ int ipmi_register_for_cmd(ipmi_user_t user,
goto out_unlock;
}
if (atomic_inc_return(&intf->event_waiters) == 1)
need_waiter(intf);
list_add_rcu(&rcvr->link, &intf->cmd_rcvrs);
out_unlock:
......@@ -1330,6 +1370,7 @@ int ipmi_unregister_for_cmd(ipmi_user_t user,
mutex_unlock(&intf->cmd_rcvrs_mutex);
synchronize_rcu();
while (rcvrs) {
atomic_dec(&intf->event_waiters);
rcvr = rcvrs;
rcvrs = rcvr->next;
kfree(rcvr);
......@@ -1535,7 +1576,7 @@ static int i_ipmi_request(ipmi_user_t user,
= IPMI_MAINTENANCE_MODE_TIMEOUT;
if (!intf->maintenance_mode
&& !intf->maintenance_mode_enable) {
intf->maintenance_mode_enable = 1;
intf->maintenance_mode_enable = true;
maintenance_mode_update(intf);
}
spin_unlock_irqrestore(&intf->maintenance_mode_lock,
......@@ -2876,6 +2917,8 @@ int ipmi_register_smi(struct ipmi_smi_handlers *handlers,
(unsigned long) intf);
atomic_set(&intf->watchdog_pretimeouts_to_deliver, 0);
spin_lock_init(&intf->events_lock);
atomic_set(&intf->event_waiters, 0);
intf->ticks_to_req_ev = IPMI_REQUEST_EV_TIME;
INIT_LIST_HEAD(&intf->waiting_events);
intf->waiting_events_count = 0;
mutex_init(&intf->cmd_rcvrs_mutex);
......@@ -3965,7 +4008,8 @@ smi_from_recv_msg(ipmi_smi_t intf, struct ipmi_recv_msg *recv_msg,
static void check_msg_timeout(ipmi_smi_t intf, struct seq_table *ent,
struct list_head *timeouts, long timeout_period,
int slot, unsigned long *flags)
int slot, unsigned long *flags,
unsigned int *waiting_msgs)
{
struct ipmi_recv_msg *msg;
struct ipmi_smi_handlers *handlers;
......@@ -3977,8 +4021,10 @@ static void check_msg_timeout(ipmi_smi_t intf, struct seq_table *ent,
return;
ent->timeout -= timeout_period;
if (ent->timeout > 0)
if (ent->timeout > 0) {
(*waiting_msgs)++;
return;
}
if (ent->retries_left == 0) {
/* The message has used all its retries. */
......@@ -3995,6 +4041,8 @@ static void check_msg_timeout(ipmi_smi_t intf, struct seq_table *ent,
struct ipmi_smi_msg *smi_msg;
/* More retries, send again. */
(*waiting_msgs)++;
/*
* Start with the max timer, set to normal timer after
* the message is sent.
......@@ -4040,117 +4088,118 @@ static void check_msg_timeout(ipmi_smi_t intf, struct seq_table *ent,
}
}
static void ipmi_timeout_handler(long timeout_period)
static unsigned int ipmi_timeout_handler(ipmi_smi_t intf, long timeout_period)
{
ipmi_smi_t intf;
struct list_head timeouts;
struct ipmi_recv_msg *msg, *msg2;
unsigned long flags;
int i;
unsigned int waiting_msgs = 0;
rcu_read_lock();
list_for_each_entry_rcu(intf, &ipmi_interfaces, link) {
tasklet_schedule(&intf->recv_tasklet);
/*
* Go through the seq table and find any messages that
* have timed out, putting them in the timeouts
* list.
*/
INIT_LIST_HEAD(&timeouts);
spin_lock_irqsave(&intf->seq_lock, flags);
for (i = 0; i < IPMI_IPMB_NUM_SEQ; i++)
check_msg_timeout(intf, &(intf->seq_table[i]),
&timeouts, timeout_period, i,
&flags);
spin_unlock_irqrestore(&intf->seq_lock, flags);
/*
* Go through the seq table and find any messages that
* have timed out, putting them in the timeouts
* list.
*/
INIT_LIST_HEAD(&timeouts);
spin_lock_irqsave(&intf->seq_lock, flags);
for (i = 0; i < IPMI_IPMB_NUM_SEQ; i++)
check_msg_timeout(intf, &(intf->seq_table[i]),
&timeouts, timeout_period, i,
&flags, &waiting_msgs);
spin_unlock_irqrestore(&intf->seq_lock, flags);
list_for_each_entry_safe(msg, msg2, &timeouts, link)
deliver_err_response(msg, IPMI_TIMEOUT_COMPLETION_CODE);
list_for_each_entry_safe(msg, msg2, &timeouts, link)
deliver_err_response(msg, IPMI_TIMEOUT_COMPLETION_CODE);
/*
* Maintenance mode handling. Check the timeout
* optimistically before we claim the lock. It may
* mean a timeout gets missed occasionally, but that
* only means the timeout gets extended by one period
* in that case. No big deal, and it avoids the lock
* most of the time.
*/
/*
* Maintenance mode handling. Check the timeout
* optimistically before we claim the lock. It may
* mean a timeout gets missed occasionally, but that
* only means the timeout gets extended by one period
* in that case. No big deal, and it avoids the lock
* most of the time.
*/
if (intf->auto_maintenance_timeout > 0) {
spin_lock_irqsave(&intf->maintenance_mode_lock, flags);
if (intf->auto_maintenance_timeout > 0) {
spin_lock_irqsave(&intf->maintenance_mode_lock, flags);
if (intf->auto_maintenance_timeout > 0) {
intf->auto_maintenance_timeout
-= timeout_period;
if (!intf->maintenance_mode
&& (intf->auto_maintenance_timeout <= 0)) {
intf->maintenance_mode_enable = 0;
maintenance_mode_update(intf);
}
intf->auto_maintenance_timeout
-= timeout_period;
if (!intf->maintenance_mode
&& (intf->auto_maintenance_timeout <= 0)) {
intf->maintenance_mode_enable = false;
maintenance_mode_update(intf);
}
spin_unlock_irqrestore(&intf->maintenance_mode_lock,
flags);
}
spin_unlock_irqrestore(&intf->maintenance_mode_lock,
flags);
}
rcu_read_unlock();
tasklet_schedule(&intf->recv_tasklet);
return waiting_msgs;
}
static void ipmi_request_event(void)
static void ipmi_request_event(ipmi_smi_t intf)
{
ipmi_smi_t intf;
struct ipmi_smi_handlers *handlers;
rcu_read_lock();
/*
* Called from the timer, no need to check if handlers is
* valid.
*/
list_for_each_entry_rcu(intf, &ipmi_interfaces, link) {
/* No event requests when in maintenance mode. */
if (intf->maintenance_mode_enable)
continue;
/* No event requests when in maintenance mode. */
if (intf->maintenance_mode_enable)
return;
handlers = intf->handlers;
if (handlers)
handlers->request_events(intf->send_info);
}
rcu_read_unlock();
handlers = intf->handlers;
if (handlers)
handlers->request_events(intf->send_info);
}
static struct timer_list ipmi_timer;
/* Call every ~1000 ms. */
#define IPMI_TIMEOUT_TIME 1000
/* How many jiffies does it take to get to the timeout time. */
#define IPMI_TIMEOUT_JIFFIES ((IPMI_TIMEOUT_TIME * HZ) / 1000)
/*
* Request events from the queue every second (this is the number of
* IPMI_TIMEOUT_TIMES between event requests). Hopefully, in the
* future, IPMI will add a way to know immediately if an event is in
* the queue and this silliness can go away.
*/
#define IPMI_REQUEST_EV_TIME (1000 / (IPMI_TIMEOUT_TIME))
static atomic_t stop_operation;
static unsigned int ticks_to_req_ev = IPMI_REQUEST_EV_TIME;
static void ipmi_timeout(unsigned long data)
{
ipmi_smi_t intf;
int nt = 0;
if (atomic_read(&stop_operation))
return;
ticks_to_req_ev--;
if (ticks_to_req_ev == 0) {
ipmi_request_event();
ticks_to_req_ev = IPMI_REQUEST_EV_TIME;
}
rcu_read_lock();
list_for_each_entry_rcu(intf, &ipmi_interfaces, link) {
int lnt = 0;
if (atomic_read(&intf->event_waiters)) {
intf->ticks_to_req_ev--;
if (intf->ticks_to_req_ev == 0) {
ipmi_request_event(intf);
intf->ticks_to_req_ev = IPMI_REQUEST_EV_TIME;
}
lnt++;
}
ipmi_timeout_handler(IPMI_TIMEOUT_TIME);
lnt += ipmi_timeout_handler(intf, IPMI_TIMEOUT_TIME);
mod_timer(&ipmi_timer, jiffies + IPMI_TIMEOUT_JIFFIES);
lnt = !!lnt;
if (lnt != intf->last_needs_timer &&
intf->handlers->set_need_watch)
intf->handlers->set_need_watch(intf->send_info, lnt);
intf->last_needs_timer = lnt;
nt += lnt;
}
rcu_read_unlock();
if (nt)
mod_timer(&ipmi_timer, jiffies + IPMI_TIMEOUT_JIFFIES);
}
static void need_waiter(ipmi_smi_t intf)
{
/* Racy, but worst case we start the timer twice. */
if (!timer_pending(&ipmi_timer))
mod_timer(&ipmi_timer, jiffies + IPMI_TIMEOUT_JIFFIES);
}
static atomic_t smi_msg_inuse_count = ATOMIC_INIT(0);
static atomic_t recv_msg_inuse_count = ATOMIC_INIT(0);
......
......@@ -217,7 +217,7 @@ struct smi_info {
unsigned char msg_flags;
/* Does the BMC have an event buffer? */
char has_event_buffer;
bool has_event_buffer;
/*
* If set to true, this will request events the next time the
......@@ -230,7 +230,7 @@ struct smi_info {
* call. Generally used after a panic to make sure stuff goes
* out.
*/
int run_to_completion;
bool run_to_completion;
/* The I/O port of an SI interface. */
int port;
......@@ -248,19 +248,25 @@ struct smi_info {
/* The timer for this si. */
struct timer_list si_timer;
/* This flag is set, if the timer is running (timer_pending() isn't enough) */
bool timer_running;
/* The time (in jiffies) the last timeout occurred at. */
unsigned long last_timeout_jiffies;
/* Used to gracefully stop the timer without race conditions. */
atomic_t stop_operation;
/* Are we waiting for the events, pretimeouts, received msgs? */
atomic_t need_watch;
/*
* The driver will disable interrupts when it gets into a
* situation where it cannot handle messages due to lack of
* memory. Once that situation clears up, it will re-enable
* interrupts.
*/
int interrupt_disabled;
bool interrupt_disabled;
/* From the get device id response... */
struct ipmi_device_id device_id;
......@@ -273,7 +279,7 @@ struct smi_info {
* True if we allocated the device, false if it came from
* someplace else (like PCI).
*/
int dev_registered;
bool dev_registered;
/* Slave address, could be reported from DMI. */
unsigned char slave_addr;
......@@ -297,19 +303,19 @@ struct smi_info {
static int force_kipmid[SI_MAX_PARMS];
static int num_force_kipmid;
#ifdef CONFIG_PCI
static int pci_registered;
static bool pci_registered;
#endif
#ifdef CONFIG_ACPI
static int pnp_registered;
static bool pnp_registered;
#endif
#ifdef CONFIG_PARISC
static int parisc_registered;
static bool parisc_registered;
#endif
static unsigned int kipmid_max_busy_us[SI_MAX_PARMS];
static int num_max_busy_us;
static int unload_when_empty = 1;
static bool unload_when_empty = true;
static int add_smi(struct smi_info *smi);
static int try_smi_init(struct smi_info *smi);
......@@ -434,6 +440,13 @@ static void start_clear_flags(struct smi_info *smi_info)
smi_info->si_state = SI_CLEARING_FLAGS;
}
static void smi_mod_timer(struct smi_info *smi_info, unsigned long new_val)
{
smi_info->last_timeout_jiffies = jiffies;
mod_timer(&smi_info->si_timer, new_val);
smi_info->timer_running = true;
}
/*
* When we have a situtaion where we run out of memory and cannot
* allocate messages, we just leave them in the BMC and run the system
......@@ -444,10 +457,9 @@ static inline void disable_si_irq(struct smi_info *smi_info)
{
if ((smi_info->irq) && (!smi_info->interrupt_disabled)) {
start_disable_irq(smi_info);
smi_info->interrupt_disabled = 1;
smi_info->interrupt_disabled = true;
if (!atomic_read(&smi_info->stop_operation))
mod_timer(&smi_info->si_timer,
jiffies + SI_TIMEOUT_JIFFIES);
smi_mod_timer(smi_info, jiffies + SI_TIMEOUT_JIFFIES);
}
}
......@@ -455,7 +467,7 @@ static inline void enable_si_irq(struct smi_info *smi_info)
{
if ((smi_info->irq) && (smi_info->interrupt_disabled)) {
start_enable_irq(smi_info);
smi_info->interrupt_disabled = 0;
smi_info->interrupt_disabled = false;
}
}
......@@ -700,7 +712,7 @@ static void handle_transaction_done(struct smi_info *smi_info)
dev_warn(smi_info->dev,
"Maybe ok, but ipmi might run very slowly.\n");
} else
smi_info->interrupt_disabled = 0;
smi_info->interrupt_disabled = false;
smi_info->si_state = SI_NORMAL;
break;
}
......@@ -853,6 +865,19 @@ static enum si_sm_result smi_event_handler(struct smi_info *smi_info,
return si_sm_result;
}
static void check_start_timer_thread(struct smi_info *smi_info)
{
if (smi_info->si_state == SI_NORMAL && smi_info->curr_msg == NULL) {
smi_mod_timer(smi_info, jiffies + SI_TIMEOUT_JIFFIES);
if (smi_info->thread)
wake_up_process(smi_info->thread);
start_next_msg(smi_info);
smi_event_handler(smi_info, 0);
}
}
static void sender(void *send_info,
struct ipmi_smi_msg *msg,
int priority)
......@@ -906,27 +931,11 @@ static void sender(void *send_info,
else
list_add_tail(&msg->link, &smi_info->xmit_msgs);
if (smi_info->si_state == SI_NORMAL && smi_info->curr_msg == NULL) {
/*
* last_timeout_jiffies is updated here to avoid
* smi_timeout() handler passing very large time_diff
* value to smi_event_handler() that causes