fedora core 6 1.2949 + vserver 2.2.0
[linux-2.6.git] / drivers / char / ipmi / ipmi_si_intf.c
index 34ceaee..f1afd26 100644 (file)
@@ -38,7 +38,6 @@
  * and drives the real SMI state machine.
  */
 
-#include <linux/config.h>
 #include <linux/module.h>
 #include <linux/moduleparam.h>
 #include <asm/system.h>
 #include <linux/list.h>
 #include <linux/pci.h>
 #include <linux/ioport.h>
-#include <linux/irq.h>
-#ifdef CONFIG_HIGH_RES_TIMERS
-#include <linux/hrtime.h>
-# if defined(schedule_next_int)
-/* Old high-res timer code, do translations. */
-#  define get_arch_cycles(a) quick_update_jiffies_sub(a)
-#  define arch_cycles_per_jiffy cycles_per_jiffies
-# endif
-static inline void add_usec_to_timer(struct timer_list *t, long v)
-{
-       t->sub_expires += nsec_to_arch_cycle(v * 1000);
-       while (t->sub_expires >= arch_cycles_per_jiffy)
-       {
-               t->expires++;
-               t->sub_expires -= arch_cycles_per_jiffy;
-       }
-}
-#endif
+#include <linux/notifier.h>
+#include <linux/mutex.h>
+#include <linux/kthread.h>
+#include <asm/irq.h>
 #include <linux/interrupt.h>
 #include <linux/rcupdate.h>
 #include <linux/ipmi_smi.h>
 #include <asm/io.h>
 #include "ipmi_si_sm.h"
 #include <linux/init.h>
+#include <linux/dmi.h>
+#include <linux/string.h>
+#include <linux/ctype.h>
 
-#define IPMI_SI_VERSION "v31"
+#define PFX "ipmi_si: "
 
 /* Measure times between events in the driver. */
 #undef DEBUG_TIMING
@@ -100,12 +88,27 @@ enum si_intf_state {
        /* FIXME - add watchdog stuff. */
 };
 
+/* Some BT-specific defines we need here. */
+#define IPMI_BT_INTMASK_REG            2
+#define IPMI_BT_INTMASK_CLEAR_IRQ_BIT  2
+#define IPMI_BT_INTMASK_ENABLE_IRQ_BIT 1
+
 enum si_type {
     SI_KCS, SI_SMIC, SI_BT
 };
+static char *si_to_str[] = { "kcs", "smic", "bt" };
+
+#define DEVICE_NAME "ipmi_si"
+
+static struct device_driver ipmi_driver =
+{
+       .name = DEVICE_NAME,
+       .bus = &platform_bus_type
+};
 
 struct smi_info
 {
+       int                    intf_num;
        ipmi_smi_t             intf;
        struct si_sm_data      *si_sm;
        struct si_sm_handlers  *handlers;
@@ -125,6 +128,15 @@ struct smi_info
        int (*irq_setup)(struct smi_info *info);
        void (*irq_cleanup)(struct smi_info *info);
        unsigned int io_size;
+       char *addr_source; /* ACPI, PCI, SMBIOS, hardcode, default. */
+       void (*addr_source_cleanup)(struct smi_info *info);
+       void *addr_source_data;
+
+       /* Per-OEM handler, called from handle_flags().
+          Returns 1 when handle_flags() needs to be re-run
+          or 0 indicating it set si_state itself.
+       */
+       int (*oem_data_avail_handler)(struct smi_info *smi_info);
 
        /* Flags from the last GET_MSG_FLAGS command, used when an ATTN
           is set to hold the flags until we are done handling everything
@@ -132,6 +144,12 @@ struct smi_info
 #define RECEIVE_MSG_AVAIL      0x01
 #define EVENT_MSG_BUFFER_FULL  0x02
 #define WDT_PRE_TIMEOUT_INT    0x08
+#define OEM0_DATA_AVAIL     0x20
+#define OEM1_DATA_AVAIL     0x40
+#define OEM2_DATA_AVAIL     0x80
+#define OEM_DATA_AVAIL      (OEM0_DATA_AVAIL | \
+                             OEM1_DATA_AVAIL | \
+                             OEM2_DATA_AVAIL)
        unsigned char       msg_flags;
 
        /* If set to true, this will request events the next time the
@@ -146,6 +164,11 @@ struct smi_info
        /* The I/O port of an SI interface. */
        int                 port;
 
+       /* The space between start addresses of the two ports.  For
+          instance, if the first port is 0xca2 and the spacing is 4, then
+          the second port is 0xca6. */
+       unsigned int        spacing;
+
        /* zero if no irq; */
        int                 irq;
 
@@ -156,8 +179,7 @@ struct smi_info
        unsigned long       last_timeout_jiffies;
 
        /* Used to gracefully stop the timer without race conditions. */
-       volatile int        stop_operation;
-       volatile int        timer_stopped;
+       atomic_t            stop_operation;
 
        /* The driver will disable interrupts when it gets into a
           situation where it cannot handle messages due to lack of
@@ -165,11 +187,19 @@ struct smi_info
           interrupts. */
        int interrupt_disabled;
 
-       unsigned char ipmi_si_dev_rev;
-       unsigned char ipmi_si_fw_rev_major;
-       unsigned char ipmi_si_fw_rev_minor;
-       unsigned char ipmi_version_major;
-       unsigned char ipmi_version_minor;
+       /* From the get device id response... */
+       struct ipmi_device_id device_id;
+
+       /* Driver model stuff. */
+       struct device *dev;
+       struct platform_device *pdev;
+
+        /* True if we allocated the device, false if it came from
+         * someplace else (like PCI). */
+       int dev_registered;
+
+       /* Slave address, could be reported from DMI. */
+       unsigned char slave_addr;
 
        /* Counters and things for the proc filesystem. */
        spinlock_t count_lock;
@@ -185,9 +215,27 @@ struct smi_info
        unsigned long events;
        unsigned long watchdog_pretimeouts;
        unsigned long incoming_messages;
+
+        struct task_struct *thread;
+
+       struct list_head link;
 };
 
-static void si_restart_short_timer(struct smi_info *smi_info);
+#define SI_MAX_PARMS 4
+
+static int force_kipmid[SI_MAX_PARMS];
+static int num_force_kipmid;
+
+static int unload_when_empty = 1;
+
+static int try_smi_init(struct smi_info *smi);
+static void cleanup_one_si(struct smi_info *to_clean);
+
+static ATOMIC_NOTIFIER_HEAD(xaction_notifier_list);
+static int register_xaction_notifier(struct notifier_block * nb)
+{
+       return atomic_notifier_chain_register(&xaction_notifier_list, nb);
+}
 
 static void deliver_recv_msg(struct smi_info *smi_info,
                             struct ipmi_smi_msg *msg)
@@ -199,14 +247,18 @@ static void deliver_recv_msg(struct smi_info *smi_info,
        spin_lock(&(smi_info->si_lock));
 }
 
-static void return_hosed_msg(struct smi_info *smi_info)
+static void return_hosed_msg(struct smi_info *smi_info, int cCode)
 {
        struct ipmi_smi_msg *msg = smi_info->curr_msg;
 
+       if (cCode < 0 || cCode > IPMI_ERR_UNSPECIFIED)
+               cCode = IPMI_ERR_UNSPECIFIED;
+       /* else use it as is */
+
        /* Make it a reponse */
        msg->rsp[0] = msg->data[0] | 4;
        msg->rsp[1] = msg->data[1];
-       msg->rsp[2] = 0xFF; /* Unknown error. */
+       msg->rsp[2] = cCode;
        msg->rsp_size = 3;
 
        smi_info->curr_msg = NULL;
@@ -226,9 +278,9 @@ static enum si_sm_result start_next_msg(struct smi_info *smi_info)
        spin_lock(&(smi_info->msg_lock));
 
        /* Pick the high priority queue first. */
-       if (! list_empty(&(smi_info->hp_xmit_msgs))) {
+       if (!list_empty(&(smi_info->hp_xmit_msgs))) {
                entry = smi_info->hp_xmit_msgs.next;
-       } else if (! list_empty(&(smi_info->xmit_msgs))) {
+       } else if (!list_empty(&(smi_info->xmit_msgs))) {
                entry = smi_info->xmit_msgs.next;
        }
 
@@ -246,16 +298,23 @@ static enum si_sm_result start_next_msg(struct smi_info *smi_info)
                do_gettimeofday(&t);
                printk("**Start2: %d.%9.9d\n", t.tv_sec, t.tv_usec);
 #endif
+               err = atomic_notifier_call_chain(&xaction_notifier_list,
+                               0, smi_info);
+               if (err & NOTIFY_STOP_MASK) {
+                       rv = SI_SM_CALL_WITHOUT_DELAY;
+                       goto out;
+               }
                err = smi_info->handlers->start_transaction(
                        smi_info->si_sm,
                        smi_info->curr_msg->data,
                        smi_info->curr_msg->data_size);
                if (err) {
-                       return_hosed_msg(smi_info);
+                       return_hosed_msg(smi_info, err);
                }
 
                rv = SI_SM_CALL_WITHOUT_DELAY;
        }
+       out:
        spin_unlock(&(smi_info->msg_lock));
 
        return rv;
@@ -309,6 +368,7 @@ static inline void enable_si_irq(struct smi_info *smi_info)
 
 static void handle_flags(struct smi_info *smi_info)
 {
+ retry:
        if (smi_info->msg_flags & WDT_PRE_TIMEOUT_INT) {
                /* Watchdog pre-timeout */
                spin_lock(&smi_info->count_lock);
@@ -358,6 +418,10 @@ static void handle_flags(struct smi_info *smi_info)
                        smi_info->curr_msg->data,
                        smi_info->curr_msg->data_size);
                smi_info->si_state = SI_GETTING_EVENTS;
+       } else if (smi_info->msg_flags & OEM_DATA_AVAIL &&
+                  smi_info->oem_data_avail_handler) {
+               if (smi_info->oem_data_avail_handler(smi_info))
+                       goto retry;
        } else {
                smi_info->si_state = SI_NORMAL;
        }
@@ -402,7 +466,7 @@ static void handle_transaction_done(struct smi_info *smi_info)
                        /* Error fetching flags, just give up for
                           now. */
                        smi_info->si_state = SI_NORMAL;
-               } else if (len < 3) {
+               } else if (len < 4) {
                        /* Hmm, no flags.  That's technically illegal, but
                           don't use uninitialized data. */
                        smi_info->si_state = SI_NORMAL;
@@ -452,14 +516,20 @@ static void handle_transaction_done(struct smi_info *smi_info)
 
                        /* Take off the event flag. */
                        smi_info->msg_flags &= ~EVENT_MSG_BUFFER_FULL;
+                       handle_flags(smi_info);
                } else {
                        spin_lock(&smi_info->count_lock);
                        smi_info->events++;
                        spin_unlock(&smi_info->count_lock);
 
+                       /* Do this before we deliver the message
+                          because delivering the message releases the
+                          lock and something else can mess with the
+                          state. */
+                       handle_flags(smi_info);
+
                        deliver_recv_msg(smi_info, msg);
                }
-               handle_flags(smi_info);
                break;
        }
 
@@ -482,14 +552,20 @@ static void handle_transaction_done(struct smi_info *smi_info)
 
                        /* Take off the msg flag. */
                        smi_info->msg_flags &= ~RECEIVE_MSG_AVAIL;
+                       handle_flags(smi_info);
                } else {
                        spin_lock(&smi_info->count_lock);
                        smi_info->incoming_messages++;
                        spin_unlock(&smi_info->count_lock);
 
+                       /* Do this before we deliver the message
+                          because delivering the message releases the
+                          lock and something else can mess with the
+                          state. */
+                       handle_flags(smi_info);
+
                        deliver_recv_msg(smi_info, msg);
                }
-               handle_flags(smi_info);
                break;
        }
 
@@ -568,14 +644,16 @@ static enum si_sm_result smi_event_handler(struct smi_info *smi_info,
                smi_info->hosed_count++;
                spin_unlock(&smi_info->count_lock);
 
+               /* Do the before return_hosed_msg, because that
+                  releases the lock. */
+               smi_info->si_state = SI_NORMAL;
                if (smi_info->curr_msg != NULL) {
                        /* If we were handling a user message, format
                            a response to send to the upper layer to
                            tell it about the error. */
-                       return_hosed_msg(smi_info);
+                       return_hosed_msg(smi_info, IPMI_ERR_UNSPECIFIED);
                }
                si_sm_result = smi_info->handlers->event(smi_info->si_sm, 0);
-               smi_info->si_state = SI_NORMAL;
        }
 
        /* We prefer handling attn over new messages. */
@@ -617,22 +695,24 @@ static enum si_sm_result smi_event_handler(struct smi_info *smi_info,
        {
                /* We are idle and the upper layer requested that I fetch
                   events, so do so. */
-               unsigned char msg[2];
+               atomic_set(&smi_info->req_events, 0);
 
-               spin_lock(&smi_info->count_lock);
-               smi_info->flag_fetches++;
-               spin_unlock(&smi_info->count_lock);
+               smi_info->curr_msg = ipmi_alloc_smi_msg();
+               if (!smi_info->curr_msg)
+                       goto out;
 
-               atomic_set(&smi_info->req_events, 0);
-               msg[0] = (IPMI_NETFN_APP_REQUEST << 2);
-               msg[1] = IPMI_GET_MSG_FLAGS_CMD;
+               smi_info->curr_msg->data[0] = (IPMI_NETFN_APP_REQUEST << 2);
+               smi_info->curr_msg->data[1] = IPMI_READ_EVENT_MSG_BUFFER_CMD;
+               smi_info->curr_msg->data_size = 2;
 
                smi_info->handlers->start_transaction(
-                       smi_info->si_sm, msg, 2);
-               smi_info->si_state = SI_GETTING_FLAGS;
+                       smi_info->si_sm,
+                       smi_info->curr_msg->data,
+                       smi_info->curr_msg->data_size);
+               smi_info->si_state = SI_GETTING_EVENTS;
                goto restart;
        }
-
+ out:
        return si_sm_result;
 }
 
@@ -647,6 +727,15 @@ static void sender(void                *send_info,
        struct timeval    t;
 #endif
 
+       if (atomic_read(&smi_info->stop_operation)) {
+               msg->rsp[0] = msg->data[0] | 4;
+               msg->rsp[1] = msg->data[1];
+               msg->rsp[2] = IPMI_ERR_UNSPECIFIED;
+               msg->rsp_size = 3;
+               deliver_recv_msg(smi_info, msg);
+               return;
+       }
+
        spin_lock_irqsave(&(smi_info->msg_lock), flags);
 #ifdef DEBUG_TIMING
        do_gettimeofday(&t);
@@ -686,7 +775,6 @@ static void sender(void                *send_info,
            && (smi_info->curr_msg == NULL))
        {
                start_next_msg(smi_info);
-               si_restart_short_timer(smi_info);
        }
        spin_unlock_irqrestore(&(smi_info->si_lock), flags);
 }
@@ -712,59 +800,66 @@ static void set_run_to_completion(void *send_info, int i_run_to_completion)
        spin_unlock_irqrestore(&(smi_info->si_lock), flags);
 }
 
-static void request_events(void *send_info)
+static int ipmi_thread(void *data)
 {
-       struct smi_info *smi_info = send_info;
+       struct smi_info *smi_info = data;
+       unsigned long flags;
+       enum si_sm_result smi_result;
 
-       atomic_set(&smi_info->req_events, 1);
+       set_user_nice(current, 19);
+       while (!kthread_should_stop()) {
+               spin_lock_irqsave(&(smi_info->si_lock), flags);
+               smi_result = smi_event_handler(smi_info, 0);
+               spin_unlock_irqrestore(&(smi_info->si_lock), flags);
+               if (smi_result == SI_SM_CALL_WITHOUT_DELAY) {
+                       /* do nothing */
+               }
+               else if (smi_result == SI_SM_CALL_WITH_DELAY)
+                       schedule();
+               else
+                       schedule_timeout_interruptible(1);
+       }
+       return 0;
 }
 
-static int initialized = 0;
 
-/* Must be called with interrupts off and with the si_lock held. */
-static void si_restart_short_timer(struct smi_info *smi_info)
+static void poll(void *send_info)
 {
-#if defined(CONFIG_HIGH_RES_TIMERS)
-       unsigned long flags;
-       unsigned long jiffies_now;
+       struct smi_info *smi_info = send_info;
 
-       if (del_timer(&(smi_info->si_timer))) {
-               /* If we don't delete the timer, then it will go off
-                  immediately, anyway.  So we only process if we
-                  actually delete the timer. */
+       /*
+        * Make sure there is some delay in the poll loop so we can
+        * drive time forward and timeout things.
+        */
+       udelay(10);
+       smi_event_handler(smi_info, 10);
+}
 
-               /* We already have irqsave on, so no need for it
-                   here. */
-               read_lock(&xtime_lock);
-               jiffies_now = jiffies;
-               smi_info->si_timer.expires = jiffies_now;
-               smi_info->si_timer.sub_expires = get_arch_cycles(jiffies_now);
+static void request_events(void *send_info)
+{
+       struct smi_info *smi_info = send_info;
 
-               add_usec_to_timer(&smi_info->si_timer, SI_SHORT_TIMEOUT_USEC);
+       if (atomic_read(&smi_info->stop_operation))
+               return;
 
-               add_timer(&(smi_info->si_timer));
-               spin_lock_irqsave(&smi_info->count_lock, flags);
-               smi_info->timeout_restarts++;
-               spin_unlock_irqrestore(&smi_info->count_lock, flags);
-       }
-#endif
+       atomic_set(&smi_info->req_events, 1);
 }
 
+static int initialized;
+
 static void smi_timeout(unsigned long data)
 {
        struct smi_info   *smi_info = (struct smi_info *) data;
        enum si_sm_result smi_result;
        unsigned long     flags;
        unsigned long     jiffies_now;
-       unsigned long     time_diff;
+       long              time_diff;
 #ifdef DEBUG_TIMING
        struct timeval    t;
 #endif
 
-       if (smi_info->stop_operation) {
-               smi_info->timer_stopped = 1;
+       if (atomic_read(&smi_info->stop_operation))
                return;
-       }
 
        spin_lock_irqsave(&(smi_info->si_lock), flags);
 #ifdef DEBUG_TIMING
@@ -772,7 +867,7 @@ static void smi_timeout(unsigned long data)
        printk("**Timer: %d.%9.9d\n", t.tv_sec, t.tv_usec);
 #endif
        jiffies_now = jiffies;
-       time_diff = ((jiffies_now - smi_info->last_timeout_jiffies)
+       time_diff = (((long)jiffies_now - (long)smi_info->last_timeout_jiffies)
                     * SI_USEC_PER_JIFFY);
        smi_result = smi_event_handler(smi_info, time_diff);
 
@@ -780,7 +875,7 @@ static void smi_timeout(unsigned long data)
 
        smi_info->last_timeout_jiffies = jiffies_now;
 
-       if ((smi_info->irq) && (! smi_info->interrupt_disabled)) {
+       if ((smi_info->irq) && (!smi_info->interrupt_disabled)) {
                /* Running with interrupts, only do long timeouts. */
                smi_info->si_timer.expires = jiffies + SI_TIMEOUT_JIFFIES;
                spin_lock_irqsave(&smi_info->count_lock, flags);
@@ -795,31 +890,19 @@ static void smi_timeout(unsigned long data)
                spin_lock_irqsave(&smi_info->count_lock, flags);
                smi_info->short_timeouts++;
                spin_unlock_irqrestore(&smi_info->count_lock, flags);
-#if defined(CONFIG_HIGH_RES_TIMERS)
-               read_lock(&xtime_lock);
-                smi_info->si_timer.expires = jiffies;
-                smi_info->si_timer.sub_expires
-                        = get_arch_cycles(smi_info->si_timer.expires);
-                read_unlock(&xtime_lock);
-               add_usec_to_timer(&smi_info->si_timer, SI_SHORT_TIMEOUT_USEC);
-#else
                smi_info->si_timer.expires = jiffies + 1;
-#endif
        } else {
                spin_lock_irqsave(&smi_info->count_lock, flags);
                smi_info->long_timeouts++;
                spin_unlock_irqrestore(&smi_info->count_lock, flags);
                smi_info->si_timer.expires = jiffies + SI_TIMEOUT_JIFFIES;
-#if defined(CONFIG_HIGH_RES_TIMERS)
-               smi_info->si_timer.sub_expires = 0;
-#endif
        }
 
  do_add_timer:
        add_timer(&(smi_info->si_timer));
 }
 
-static irqreturn_t si_irq_handler(int irq, void *data, struct pt_regs *regs)
+static irqreturn_t si_irq_handler(int irq, void *data)
 {
        struct smi_info *smi_info = data;
        unsigned long   flags;
@@ -833,7 +916,7 @@ static irqreturn_t si_irq_handler(int irq, void *data, struct pt_regs *regs)
        smi_info->interrupts++;
        spin_unlock(&smi_info->count_lock);
 
-       if (smi_info->stop_operation)
+       if (atomic_read(&smi_info->stop_operation))
                goto out;
 
 #ifdef DEBUG_TIMING
@@ -846,39 +929,113 @@ static irqreturn_t si_irq_handler(int irq, void *data, struct pt_regs *regs)
        return IRQ_HANDLED;
 }
 
+static irqreturn_t si_bt_irq_handler(int irq, void *data)
+{
+       struct smi_info *smi_info = data;
+       /* We need to clear the IRQ flag for the BT interface. */
+       smi_info->io.outputb(&smi_info->io, IPMI_BT_INTMASK_REG,
+                            IPMI_BT_INTMASK_CLEAR_IRQ_BIT
+                            | IPMI_BT_INTMASK_ENABLE_IRQ_BIT);
+       return si_irq_handler(irq, data);
+}
+
+static int smi_start_processing(void       *send_info,
+                               ipmi_smi_t intf)
+{
+       struct smi_info *new_smi = send_info;
+       int             enable = 0;
+
+       new_smi->intf = intf;
+
+       /* Set up the timer that drives the interface. */
+       setup_timer(&new_smi->si_timer, smi_timeout, (long)new_smi);
+       new_smi->last_timeout_jiffies = jiffies;
+       mod_timer(&new_smi->si_timer, jiffies + SI_TIMEOUT_JIFFIES);
+
+       /*
+        * Check if the user forcefully enabled the daemon.
+        */
+       if (new_smi->intf_num < num_force_kipmid)
+               enable = force_kipmid[new_smi->intf_num];
+       /*
+        * The BT interface is efficient enough to not need a thread,
+        * and there is no need for a thread if we have interrupts.
+        */
+       else if ((new_smi->si_type != SI_BT) && (!new_smi->irq))
+               enable = 1;
+
+       if (enable) {
+               new_smi->thread = kthread_run(ipmi_thread, new_smi,
+                                             "kipmi%d", new_smi->intf_num);
+               if (IS_ERR(new_smi->thread)) {
+                       printk(KERN_NOTICE "ipmi_si_intf: Could not start"
+                              " kernel thread due to error %ld, only using"
+                              " timers to drive the interface\n",
+                              PTR_ERR(new_smi->thread));
+                       new_smi->thread = NULL;
+               }
+       }
+
+       return 0;
+}
+
+static void set_maintenance_mode(void *send_info, int enable)
+{
+       struct smi_info   *smi_info = send_info;
+
+       if (!enable)
+               atomic_set(&smi_info->req_events, 0);
+}
+
 static struct ipmi_smi_handlers handlers =
 {
        .owner                  = THIS_MODULE,
+       .start_processing       = smi_start_processing,
        .sender                 = sender,
        .request_events         = request_events,
-       .set_run_to_completion  = set_run_to_completion
+       .set_maintenance_mode   = set_maintenance_mode,
+       .set_run_to_completion  = set_run_to_completion,
+       .poll                   = poll,
 };
 
 /* There can be 4 IO ports passed in (with or without IRQs), 4 addresses,
    a default IO port, and 1 ACPI/SPMI address.  That sets SI_MAX_DRIVERS */
 
-#define SI_MAX_PARMS 4
-#define SI_MAX_DRIVERS ((SI_MAX_PARMS * 2) + 2)
-static struct smi_info *smi_infos[SI_MAX_DRIVERS] =
-{ NULL, NULL, NULL, NULL };
-
-#define DEVICE_NAME "ipmi_si"
+static LIST_HEAD(smi_infos);
+static DEFINE_MUTEX(smi_infos_lock);
+static int smi_num; /* Used to sequence the SMIs */
 
-#define DEFAULT_KCS_IO_PORT 0xca2
-#define DEFAULT_SMIC_IO_PORT 0xca9
-#define DEFAULT_BT_IO_PORT   0xe4
+#define DEFAULT_REGSPACING     1
 
 static int           si_trydefaults = 1;
-static char          *si_type[SI_MAX_PARMS] = { NULL, NULL, NULL, NULL };
+static char          *si_type[SI_MAX_PARMS];
 #define MAX_SI_TYPE_STR 30
 static char          si_type_str[MAX_SI_TYPE_STR];
-static unsigned long addrs[SI_MAX_PARMS] = { 0, 0, 0, 0 };
-static int num_addrs = 0;
-static unsigned int  ports[SI_MAX_PARMS] = { 0, 0, 0, 0 };
-static int num_ports = 0;
-static int           irqs[SI_MAX_PARMS] = { 0, 0, 0, 0 };
-static int num_irqs = 0;
+static unsigned long addrs[SI_MAX_PARMS];
+static int num_addrs;
+static unsigned int  ports[SI_MAX_PARMS];
+static int num_ports;
+static int           irqs[SI_MAX_PARMS];
+static int num_irqs;
+static int           regspacings[SI_MAX_PARMS];
+static int num_regspacings;
+static int           regsizes[SI_MAX_PARMS];
+static int num_regsizes;
+static int           regshifts[SI_MAX_PARMS];
+static int num_regshifts;
+static int slave_addrs[SI_MAX_PARMS];
+static int num_slave_addrs;
+
+#define IPMI_IO_ADDR_SPACE  0
+#define IPMI_MEM_ADDR_SPACE 1
+static char *addr_space_to_str[] = { "i/o", "mem" };
 
+static int hotmod_handler(const char *val, struct kernel_param *kp);
+
+module_param_call(hotmod, hotmod_handler, NULL, NULL, 0200);
+MODULE_PARM_DESC(hotmod, "Add and remove interfaces.  See"
+                " Documentation/IPMI.txt in the kernel sources for the"
+                " gory details.");
 
 module_param_named(trydefaults, si_trydefaults, bool, 0);
 MODULE_PARM_DESC(trydefaults, "Setting this to 'false' will disable the"
@@ -889,48 +1046,60 @@ MODULE_PARM_DESC(type, "Defines the type of each interface, each"
                 " interface separated by commas.  The types are 'kcs',"
                 " 'smic', and 'bt'.  For example si_type=kcs,bt will set"
                 " the first interface to kcs and the second to bt");
-module_param_array(addrs, long, num_addrs, 0);
+module_param_array(addrs, long, &num_addrs, 0);
 MODULE_PARM_DESC(addrs, "Sets the memory address of each interface, the"
                 " addresses separated by commas.  Only use if an interface"
                 " is in memory.  Otherwise, set it to zero or leave"
                 " it blank.");
-module_param_array(ports, int, num_ports, 0);
+module_param_array(ports, int, &num_ports, 0);
 MODULE_PARM_DESC(ports, "Sets the port address of each interface, the"
                 " addresses separated by commas.  Only use if an interface"
                 " is a port.  Otherwise, set it to zero or leave"
                 " it blank.");
-module_param_array(irqs, int, num_irqs, 0);
+module_param_array(irqs, int, &num_irqs, 0);
 MODULE_PARM_DESC(irqs, "Sets the interrupt of each interface, the"
                 " addresses separated by commas.  Only use if an interface"
                 " has an interrupt.  Otherwise, set it to zero or leave"
                 " it blank.");
+module_param_array(regspacings, int, &num_regspacings, 0);
+MODULE_PARM_DESC(regspacings, "The number of bytes between the start address"
+                " and each successive register used by the interface.  For"
+                " instance, if the start address is 0xca2 and the spacing"
+                " is 2, then the second address is at 0xca4.  Defaults"
+                " to 1.");
+module_param_array(regsizes, int, &num_regsizes, 0);
+MODULE_PARM_DESC(regsizes, "The size of the specific IPMI register in bytes."
+                " This should generally be 1, 2, 4, or 8 for an 8-bit,"
+                " 16-bit, 32-bit, or 64-bit register.  Use this if you"
+                " the 8-bit IPMI register has to be read from a larger"
+                " register.");
+module_param_array(regshifts, int, &num_regshifts, 0);
+MODULE_PARM_DESC(regshifts, "The amount to shift the data read from the."
+                " IPMI register, in bits.  For instance, if the data"
+                " is read from a 32-bit word and the IPMI data is in"
+                " bit 8-15, then the shift would be 8");
+module_param_array(slave_addrs, int, &num_slave_addrs, 0);
+MODULE_PARM_DESC(slave_addrs, "Set the default IPMB slave address for"
+                " the controller.  Normally this is 0x20, but can be"
+                " overridden by this parm.  This is an array indexed"
+                " by interface number.");
+module_param_array(force_kipmid, int, &num_force_kipmid, 0);
+MODULE_PARM_DESC(force_kipmid, "Force the kipmi daemon to be enabled (1) or"
+                " disabled(0).  Normally the IPMI driver auto-detects"
+                " this, but the value may be overridden by this parm.");
+module_param(unload_when_empty, int, 0);
+MODULE_PARM_DESC(unload_when_empty, "Unload the module if no interfaces are"
+                " specified or found, default is 1.  Setting to 0"
+                " is useful for hot add of devices using hotmod.");
 
 
-#if defined(CONFIG_ACPI_INTERPETER) || defined(CONFIG_X86) || defined(CONFIG_PCI)
-#define IPMI_MEM_ADDR_SPACE 1
-#define IPMI_IO_ADDR_SPACE  2
-static int is_new_interface(int intf, u8 addr_space, unsigned long base_addr)
+static void std_irq_cleanup(struct smi_info *info)
 {
-       int i;
-
-       for (i = 0; i < SI_MAX_PARMS; ++i) {
-               /* Don't check our address. */
-               if (i == intf)
-                       continue;
-               if (si_type[i] != NULL) {
-                       if ((addr_space == IPMI_MEM_ADDR_SPACE &&
-                            base_addr == addrs[i]) ||
-                           (addr_space == IPMI_IO_ADDR_SPACE &&
-                            base_addr == ports[i]))
-                               return 0;
-               }
-               else
-                       break;
-       }
-
-       return 1;
+       if (info->si_type == SI_BT)
+               /* Disable the interrupt in the BT interface. */
+               info->io.outputb(&info->io, IPMI_BT_INTMASK_REG, 0);
+       free_irq(info->irq, info);
 }
-#endif
 
 static int std_irq_setup(struct smi_info *info)
 {
@@ -939,11 +1108,22 @@ static int std_irq_setup(struct smi_info *info)
        if (!info->irq)
                return 0;
 
-       rv = request_irq(info->irq,
-                        si_irq_handler,
-                        SA_INTERRUPT,
-                        DEVICE_NAME,
-                        info);
+       if (info->si_type == SI_BT) {
+               rv = request_irq(info->irq,
+                                si_bt_irq_handler,
+                                IRQF_DISABLED,
+                                DEVICE_NAME,
+                                info);
+               if (!rv)
+                       /* Enable the interrupt in the BT interface. */
+                       info->io.outputb(&info->io, IPMI_BT_INTMASK_REG,
+                                        IPMI_BT_INTMASK_ENABLE_IRQ_BIT);
+       } else
+               rv = request_irq(info->irq,
+                                si_irq_handler,
+                                IRQF_DISABLED,
+                                DEVICE_NAME,
+                                info);
        if (rv) {
                printk(KERN_WARNING
                       "ipmi_si: %s unable to claim interrupt %d,"
@@ -951,195 +1131,578 @@ static int std_irq_setup(struct smi_info *info)
                       DEVICE_NAME, info->irq);
                info->irq = 0;
        } else {
+               info->irq_cleanup = std_irq_cleanup;
                printk("  Using irq %d\n", info->irq);
        }
 
        return rv;
 }
 
-static void std_irq_cleanup(struct smi_info *info)
+static unsigned char port_inb(struct si_sm_io *io, unsigned int offset)
 {
-       if (!info->irq)
-               return;
+       unsigned int addr = io->addr_data;
 
-       free_irq(info->irq, info);
+       return inb(addr + (offset * io->regspacing));
 }
 
-static unsigned char port_inb(struct si_sm_io *io, unsigned int offset)
+static void port_outb(struct si_sm_io *io, unsigned int offset,
+                     unsigned char b)
 {
-       unsigned int *addr = io->info;
+       unsigned int addr = io->addr_data;
 
-       return inb((*addr)+offset);
+       outb(b, addr + (offset * io->regspacing));
 }
 
-static void port_outb(struct si_sm_io *io, unsigned int offset,
+static unsigned char port_inw(struct si_sm_io *io, unsigned int offset)
+{
+       unsigned int addr = io->addr_data;
+
+       return (inw(addr + (offset * io->regspacing)) >> io->regshift) & 0xff;
+}
+
+static void port_outw(struct si_sm_io *io, unsigned int offset,
                      unsigned char b)
 {
-       unsigned int *addr = io->info;
+       unsigned int addr = io->addr_data;
 
-       outb(b, (*addr)+offset);
+       outw(b << io->regshift, addr + (offset * io->regspacing));
 }
 
-static int port_setup(struct smi_info *info)
+static unsigned char port_inl(struct si_sm_io *io, unsigned int offset)
 {
-       unsigned int *addr = info->io.info;
+       unsigned int addr = io->addr_data;
 
-       if (!addr || (!*addr))
-               return -ENODEV;
+       return (inl(addr + (offset * io->regspacing)) >> io->regshift) & 0xff;
+}
 
-       if (request_region(*addr, info->io_size, DEVICE_NAME) == NULL)
-               return -EIO;
-       return 0;
+static void port_outl(struct si_sm_io *io, unsigned int offset,
+                     unsigned char b)
+{
+       unsigned int addr = io->addr_data;
+
+       outl(b << io->regshift, addr+(offset * io->regspacing));
 }
 
 static void port_cleanup(struct smi_info *info)
 {
-       unsigned int *addr = info->io.info;
+       unsigned int addr = info->io.addr_data;
+       int          idx;
 
-       if (addr && (*addr))
-               release_region (*addr, info->io_size);
-       kfree(info);
+       if (addr) {
+               for (idx = 0; idx < info->io_size; idx++) {
+                       release_region(addr + idx * info->io.regspacing,
+                                      info->io.regsize);
+               }
+       }
 }
 
-static int try_init_port(int intf_num, struct smi_info **new_info)
+static int port_setup(struct smi_info *info)
 {
-       struct smi_info *info;
+       unsigned int addr = info->io.addr_data;
+       int          idx;
 
-       if (!ports[intf_num])
+       if (!addr)
                return -ENODEV;
 
-       if (!is_new_interface(intf_num, IPMI_IO_ADDR_SPACE,
-                             ports[intf_num]))
-               return -ENODEV;
+       info->io_cleanup = port_cleanup;
 
-       info = kmalloc(sizeof(*info), GFP_KERNEL);
-       if (!info) {
-               printk(KERN_ERR "ipmi_si: Could not allocate SI data (1)\n");
-               return -ENOMEM;
+       /* Figure out the actual inb/inw/inl/etc routine to use based
+          upon the register size. */
+       switch (info->io.regsize) {
+       case 1:
+               info->io.inputb = port_inb;
+               info->io.outputb = port_outb;
+               break;
+       case 2:
+               info->io.inputb = port_inw;
+               info->io.outputb = port_outw;
+               break;
+       case 4:
+               info->io.inputb = port_inl;
+               info->io.outputb = port_outl;
+               break;
+       default:
+               printk("ipmi_si: Invalid register size: %d\n",
+                      info->io.regsize);
+               return -EINVAL;
        }
-       memset(info, 0, sizeof(*info));
 
-       info->io_setup = port_setup;
-       info->io_cleanup = port_cleanup;
-       info->io.inputb = port_inb;
-       info->io.outputb = port_outb;
-       info->io.info = &(ports[intf_num]);
-       info->io.addr = NULL;
-       info->irq = 0;
-       info->irq_setup = NULL;
-       *new_info = info;
-
-       if (si_type[intf_num] == NULL)
-               si_type[intf_num] = "kcs";
-
-       printk("ipmi_si: Trying \"%s\" at I/O port 0x%x\n",
-              si_type[intf_num], ports[intf_num]);
+       /* Some BIOSes reserve disjoint I/O regions in their ACPI
+        * tables.  This causes problems when trying to register the
+        * entire I/O region.  Therefore we must register each I/O
+        * port separately.
+        */
+       for (idx = 0; idx < info->io_size; idx++) {
+               if (request_region(addr + idx * info->io.regspacing,
+                                  info->io.regsize, DEVICE_NAME) == NULL) {
+                       /* Undo allocations */
+                       while (idx--) {
+                               release_region(addr + idx * info->io.regspacing,
+                                              info->io.regsize);
+                       }
+                       return -EIO;
+               }
+       }
        return 0;
 }
 
-static unsigned char mem_inb(struct si_sm_io *io, unsigned int offset)
+static unsigned char intf_mem_inb(struct si_sm_io *io, unsigned int offset)
 {
-       return readb((io->addr)+offset);
+       return readb((io->addr)+(offset * io->regspacing));
 }
 
-static void mem_outb(struct si_sm_io *io, unsigned int offset,
+static void intf_mem_outb(struct si_sm_io *io, unsigned int offset,
                     unsigned char b)
 {
-       writeb(b, (io->addr)+offset);
+       writeb(b, (io->addr)+(offset * io->regspacing));
 }
 
-static int mem_setup(struct smi_info *info)
+static unsigned char intf_mem_inw(struct si_sm_io *io, unsigned int offset)
+{
+       return (readw((io->addr)+(offset * io->regspacing)) >> io->regshift)
+               & 0xff;
+}
+
+static void intf_mem_outw(struct si_sm_io *io, unsigned int offset,
+                    unsigned char b)
 {
-       unsigned long *addr = info->io.info;
+       writeb(b << io->regshift, (io->addr)+(offset * io->regspacing));
+}
 
-       if (!addr || (!*addr))
-               return -ENODEV;
+static unsigned char intf_mem_inl(struct si_sm_io *io, unsigned int offset)
+{
+       return (readl((io->addr)+(offset * io->regspacing)) >> io->regshift)
+               & 0xff;
+}
 
-       if (request_mem_region(*addr, info->io_size, DEVICE_NAME) == NULL)
-               return -EIO;
+static void intf_mem_outl(struct si_sm_io *io, unsigned int offset,
+                    unsigned char b)
+{
+       writel(b << io->regshift, (io->addr)+(offset * io->regspacing));
+}
 
-       info->io.addr = ioremap(*addr, info->io_size);
-       if (info->io.addr == NULL) {
-               release_mem_region(*addr, info->io_size);
-               return -EIO;
-       }
-       return 0;
+#ifdef readq
+static unsigned char mem_inq(struct si_sm_io *io, unsigned int offset)
+{
+       return (readq((io->addr)+(offset * io->regspacing)) >> io->regshift)
+               & 0xff;
 }
 
+static void mem_outq(struct si_sm_io *io, unsigned int offset,
+                    unsigned char b)
+{
+       writeq(b << io->regshift, (io->addr)+(offset * io->regspacing));
+}
+#endif
+
 static void mem_cleanup(struct smi_info *info)
 {
-       unsigned long *addr = info->io.info;
+       unsigned long addr = info->io.addr_data;
+       int           mapsize;
 
        if (info->io.addr) {
                iounmap(info->io.addr);
-               release_mem_region(*addr, info->io_size);
+
+               mapsize = ((info->io_size * info->io.regspacing)
+                          - (info->io.regspacing - info->io.regsize));
+
+               release_mem_region(addr, mapsize);
        }
-       kfree(info);
 }
 
-static int try_init_mem(int intf_num, struct smi_info **new_info)
+static int mem_setup(struct smi_info *info)
 {
-       struct smi_info *info;
+       unsigned long addr = info->io.addr_data;
+       int           mapsize;
 
-       if (!addrs[intf_num])
+       if (!addr)
                return -ENODEV;
 
-       if (!is_new_interface(intf_num, IPMI_MEM_ADDR_SPACE,
-                             addrs[intf_num]))
-               return -ENODEV;
+       info->io_cleanup = mem_cleanup;
 
-       info = kmalloc(sizeof(*info), GFP_KERNEL);
-       if (!info) {
-               printk(KERN_ERR "ipmi_si: Could not allocate SI data (2)\n");
-               return -ENOMEM;
+       /* Figure out the actual readb/readw/readl/etc routine to use based
+          upon the register size. */
+       switch (info->io.regsize) {
+       case 1:
+               info->io.inputb = intf_mem_inb;
+               info->io.outputb = intf_mem_outb;
+               break;
+       case 2:
+               info->io.inputb = intf_mem_inw;
+               info->io.outputb = intf_mem_outw;
+               break;
+       case 4:
+               info->io.inputb = intf_mem_inl;
+               info->io.outputb = intf_mem_outl;
+               break;
+#ifdef readq
+       case 8:
+               info->io.inputb = mem_inq;
+               info->io.outputb = mem_outq;
+               break;
+#endif
+       default:
+               printk("ipmi_si: Invalid register size: %d\n",
+                      info->io.regsize);
+               return -EINVAL;
        }
-       memset(info, 0, sizeof(*info));
 
-       info->io_setup = mem_setup;
-       info->io_cleanup = mem_cleanup;
-       info->io.inputb = mem_inb;
-       info->io.outputb = mem_outb;
-       info->io.info = (void *) addrs[intf_num];
-       info->io.addr = NULL;
-       info->irq = 0;
-       info->irq_setup = NULL;
-       *new_info = info;
-
-       if (si_type[intf_num] == NULL)
-               si_type[intf_num] = "kcs";
-
-       printk("ipmi_si: Trying \"%s\" at memory address 0x%lx\n",
-              si_type[intf_num], addrs[intf_num]);
+       /* Calculate the total amount of memory to claim.  This is an
+        * unusual looking calculation, but it avoids claiming any
+        * more memory than it has to.  It will claim everything
+        * between the first address to the end of the last full
+        * register. */
+       mapsize = ((info->io_size * info->io.regspacing)
+                  - (info->io.regspacing - info->io.regsize));
+
+       if (request_mem_region(addr, mapsize, DEVICE_NAME) == NULL)
+               return -EIO;
+
+       info->io.addr = ioremap(addr, mapsize);
+       if (info->io.addr == NULL) {
+               release_mem_region(addr, mapsize);
+               return -EIO;
+       }
        return 0;
 }
 
+/*
+ * Parms come in as <op1>[:op2[:op3...]].  ops are:
+ *   add|remove,kcs|bt|smic,mem|i/o,<address>[,<opt1>[,<opt2>[,...]]]
+ * Options are:
+ *   rsp=<regspacing>
+ *   rsi=<regsize>
+ *   rsh=<regshift>
+ *   irq=<irq>
+ *   ipmb=<ipmb addr>
+ */
+enum hotmod_op { HM_ADD, HM_REMOVE };
+struct hotmod_vals {
+       char *name;
+       int  val;
+};
+static struct hotmod_vals hotmod_ops[] = {
+       { "add",        HM_ADD },
+       { "remove",     HM_REMOVE },
+       { NULL }
+};
+static struct hotmod_vals hotmod_si[] = {
+       { "kcs",        SI_KCS },
+       { "smic",       SI_SMIC },
+       { "bt",         SI_BT },
+       { NULL }
+};
+static struct hotmod_vals hotmod_as[] = {
+       { "mem",        IPMI_MEM_ADDR_SPACE },
+       { "i/o",        IPMI_IO_ADDR_SPACE },
+       { NULL }
+};
 
-#ifdef CONFIG_ACPI_INTERPRETER
+static int parse_str(struct hotmod_vals *v, int *val, char *name, char **curr)
+{
+       char *s;
+       int  i;
 
-#include <linux/acpi.h>
+       s = strchr(*curr, ',');
+       if (!s) {
+               printk(KERN_WARNING PFX "No hotmod %s given.\n", name);
+               return -EINVAL;
+       }
+       *s = '\0';
+       s++;
+       for (i = 0; hotmod_ops[i].name; i++) {
+               if (strcmp(*curr, v[i].name) == 0) {
+                       *val = v[i].val;
+                       *curr = s;
+                       return 0;
+               }
+       }
 
-/* Once we get an ACPI failure, we don't try any more, because we go
-   through the tables sequentially.  Once we don't find a table, there
-   are no more. */
-static int acpi_failure = 0;
+       printk(KERN_WARNING PFX "Invalid hotmod %s '%s'\n", name, *curr);
+       return -EINVAL;
+}
 
-/* For GPE-type interrupts. */
-void ipmi_acpi_gpe(void *context)
+static int check_hotmod_int_op(const char *curr, const char *option,
+                              const char *name, int *val)
 {
-       struct smi_info *smi_info = context;
-       unsigned long   flags;
-#ifdef DEBUG_TIMING
-       struct timeval t;
-#endif
+       char *n;
+
+       if (strcmp(curr, name) == 0) {
+               if (!option) {
+                       printk(KERN_WARNING PFX
+                              "No option given for '%s'\n",
+                              curr);
+                       return -EINVAL;
+               }
+               *val = simple_strtoul(option, &n, 0);
+               if ((*n != '\0') || (*option == '\0')) {
+                       printk(KERN_WARNING PFX
+                              "Bad option given for '%s'\n",
+                              curr);
+                       return -EINVAL;
+               }
+               return 1;
+       }
+       return 0;
+}
 
-       spin_lock_irqsave(&(smi_info->si_lock), flags);
+static int hotmod_handler(const char *val, struct kernel_param *kp)
+{
+       char *str = kstrdup(val, GFP_KERNEL);
+       int  rv;
+       char *next, *curr, *s, *n, *o;
+       enum hotmod_op op;
+       enum si_type si_type;
+       int  addr_space;
+       unsigned long addr;
+       int regspacing;
+       int regsize;
+       int regshift;
+       int irq;
+       int ipmb;
+       int ival;
+       int len;
+       struct smi_info *info;
 
-       spin_lock(&smi_info->count_lock);
-       smi_info->interrupts++;
-       spin_unlock(&smi_info->count_lock);
+       if (!str)
+               return -ENOMEM;
 
-       if (smi_info->stop_operation)
-               goto out;
+       /* Kill any trailing spaces, as we can get a "\n" from echo. */
+       len = strlen(str);
+       ival = len - 1;
+       while ((ival >= 0) && isspace(str[ival])) {
+               str[ival] = '\0';
+               ival--;
+       }
+
+       for (curr = str; curr; curr = next) {
+               regspacing = 1;
+               regsize = 1;
+               regshift = 0;
+               irq = 0;
+               ipmb = 0x20;
+
+               next = strchr(curr, ':');
+               if (next) {
+                       *next = '\0';
+                       next++;
+               }
+
+               rv = parse_str(hotmod_ops, &ival, "operation", &curr);
+               if (rv)
+                       break;
+               op = ival;
+
+               rv = parse_str(hotmod_si, &ival, "interface type", &curr);
+               if (rv)
+                       break;
+               si_type = ival;
+
+               rv = parse_str(hotmod_as, &addr_space, "address space", &curr);
+               if (rv)
+                       break;
+
+               s = strchr(curr, ',');
+               if (s) {
+                       *s = '\0';
+                       s++;
+               }
+               addr = simple_strtoul(curr, &n, 0);
+               if ((*n != '\0') || (*curr == '\0')) {
+                       printk(KERN_WARNING PFX "Invalid hotmod address"
+                              " '%s'\n", curr);
+                       break;
+               }
+
+               while (s) {
+                       curr = s;
+                       s = strchr(curr, ',');
+                       if (s) {
+                               *s = '\0';
+                               s++;
+                       }
+                       o = strchr(curr, '=');
+                       if (o) {
+                               *o = '\0';
+                               o++;
+                       }
+                       rv = check_hotmod_int_op(curr, o, "rsp", &regspacing);
+                       if (rv < 0)
+                               goto out;
+                       else if (rv)
+                               continue;
+                       rv = check_hotmod_int_op(curr, o, "rsi", &regsize);
+                       if (rv < 0)
+                               goto out;
+                       else if (rv)
+                               continue;
+                       rv = check_hotmod_int_op(curr, o, "rsh", &regshift);
+                       if (rv < 0)
+                               goto out;
+                       else if (rv)
+                               continue;
+                       rv = check_hotmod_int_op(curr, o, "irq", &irq);
+                       if (rv < 0)
+                               goto out;
+                       else if (rv)
+                               continue;
+                       rv = check_hotmod_int_op(curr, o, "ipmb", &ipmb);
+                       if (rv < 0)
+                               goto out;
+                       else if (rv)
+                               continue;
+
+                       rv = -EINVAL;
+                       printk(KERN_WARNING PFX
+                              "Invalid hotmod option '%s'\n",
+                              curr);
+                       goto out;
+               }
+
+               if (op == HM_ADD) {
+                       info = kzalloc(sizeof(*info), GFP_KERNEL);
+                       if (!info) {
+                               rv = -ENOMEM;
+                               goto out;
+                       }
+
+                       info->addr_source = "hotmod";
+                       info->si_type = si_type;
+                       info->io.addr_data = addr;
+                       info->io.addr_type = addr_space;
+                       if (addr_space == IPMI_MEM_ADDR_SPACE)
+                               info->io_setup = mem_setup;
+                       else
+                               info->io_setup = port_setup;
+
+                       info->io.addr = NULL;
+                       info->io.regspacing = regspacing;
+                       if (!info->io.regspacing)
+                               info->io.regspacing = DEFAULT_REGSPACING;
+                       info->io.regsize = regsize;
+                       if (!info->io.regsize)
+                               info->io.regsize = DEFAULT_REGSPACING;
+                       info->io.regshift = regshift;
+                       info->irq = irq;
+                       if (info->irq)
+                               info->irq_setup = std_irq_setup;
+                       info->slave_addr = ipmb;
+
+                       try_smi_init(info);
+               } else {
+                       /* remove */
+                       struct smi_info *e, *tmp_e;
+
+                       mutex_lock(&smi_infos_lock);
+                       list_for_each_entry_safe(e, tmp_e, &smi_infos, link) {
+                               if (e->io.addr_type != addr_space)
+                                       continue;
+                               if (e->si_type != si_type)
+                                       continue;
+                               if (e->io.addr_data == addr)
+                                       cleanup_one_si(e);
+                       }
+                       mutex_unlock(&smi_infos_lock);
+               }
+       }
+       rv = len;
+ out:
+       kfree(str);
+       return rv;
+}
+
+static __devinit void hardcode_find_bmc(void)
+{
+       int             i;
+       struct smi_info *info;
+
+       for (i = 0; i < SI_MAX_PARMS; i++) {
+               if (!ports[i] && !addrs[i])
+                       continue;
+
+               info = kzalloc(sizeof(*info), GFP_KERNEL);
+               if (!info)
+                       return;
+
+               info->addr_source = "hardcoded";
+
+               if (!si_type[i] || strcmp(si_type[i], "kcs") == 0) {
+                       info->si_type = SI_KCS;
+               } else if (strcmp(si_type[i], "smic") == 0) {
+                       info->si_type = SI_SMIC;
+               } else if (strcmp(si_type[i], "bt") == 0) {
+                       info->si_type = SI_BT;
+               } else {
+                       printk(KERN_WARNING
+                              "ipmi_si: Interface type specified "
+                              "for interface %d, was invalid: %s\n",
+                              i, si_type[i]);
+                       kfree(info);
+                       continue;
+               }
+
+               if (ports[i]) {
+                       /* An I/O port */
+                       info->io_setup = port_setup;
+                       info->io.addr_data = ports[i];
+                       info->io.addr_type = IPMI_IO_ADDR_SPACE;
+               } else if (addrs[i]) {
+                       /* A memory port */
+                       info->io_setup = mem_setup;
+                       info->io.addr_data = addrs[i];
+                       info->io.addr_type = IPMI_MEM_ADDR_SPACE;
+               } else {
+                       printk(KERN_WARNING
+                              "ipmi_si: Interface type specified "
+                              "for interface %d, "
+                              "but port and address were not set or "
+                              "set to zero.\n", i);
+                       kfree(info);
+                       continue;
+               }
+
+               info->io.addr = NULL;
+               info->io.regspacing = regspacings[i];
+               if (!info->io.regspacing)
+                       info->io.regspacing = DEFAULT_REGSPACING;
+               info->io.regsize = regsizes[i];
+               if (!info->io.regsize)
+                       info->io.regsize = DEFAULT_REGSPACING;
+               info->io.regshift = regshifts[i];
+               info->irq = irqs[i];
+               if (info->irq)
+                       info->irq_setup = std_irq_setup;
+
+               try_smi_init(info);
+       }
+}
+
+#ifdef CONFIG_ACPI
+
+#include <linux/acpi.h>
+
+/* Once we get an ACPI failure, we don't try any more, because we go
+   through the tables sequentially.  Once we don't find a table, there
+   are no more. */
+static int acpi_failure;
+
+/* For GPE-type interrupts. */
+static u32 ipmi_acpi_gpe(void *context)
+{
+       struct smi_info *smi_info = context;
+       unsigned long   flags;
+#ifdef DEBUG_TIMING
+       struct timeval t;
+#endif
+
+       spin_lock_irqsave(&(smi_info->si_lock), flags);
+
+       spin_lock(&smi_info->count_lock);
+       smi_info->interrupts++;
+       spin_unlock(&smi_info->count_lock);
+
+       if (atomic_read(&smi_info->stop_operation))
+               goto out;
 
 #ifdef DEBUG_TIMING
        do_gettimeofday(&t);
@@ -1148,6 +1711,16 @@ void ipmi_acpi_gpe(void *context)
        smi_event_handler(smi_info, 0);
  out:
        spin_unlock_irqrestore(&(smi_info->si_lock), flags);
+
+       return ACPI_INTERRUPT_HANDLED;
+}
+
+static void acpi_gpe_irq_cleanup(struct smi_info *info)
+{
+       if (!info->irq)
+               return;
+
+       acpi_remove_gpe_handler(NULL, info->irq, &ipmi_acpi_gpe);
 }
 
 static int acpi_gpe_irq_setup(struct smi_info *info)
@@ -1161,7 +1734,7 @@ static int acpi_gpe_irq_setup(struct smi_info *info)
        status = acpi_install_gpe_handler(NULL,
                                          info->irq,
                                          ACPI_GPE_LEVEL_TRIGGERED,
-                                         ipmi_acpi_gpe,
+                                         &ipmi_acpi_gpe,
                                          info);
        if (status != AE_OK) {
                printk(KERN_WARNING
@@ -1171,18 +1744,10 @@ static int acpi_gpe_irq_setup(struct smi_info *info)
                info->irq = 0;
                return -EINVAL;
        } else {
+               info->irq_cleanup = acpi_gpe_irq_cleanup;
                printk("  Using ACPI GPE %d\n", info->irq);
                return 0;
        }
-
-}
-
-static void acpi_gpe_irq_cleanup(struct smi_info *info)
-{
-       if (!info->irq)
-               return;
-
-       acpi_remove_gpe_handler(NULL, info->irq, ipmi_acpi_gpe);
 }
 
 /*
@@ -1227,25 +1792,11 @@ struct SPMITable {
        s8      spmi_id[1]; /* A '\0' terminated array starts here. */
 };
 
-static int try_init_acpi(int intf_num, struct smi_info **new_info)
+static __devinit int try_init_acpi(struct SPMITable *spmi)
 {
        struct smi_info  *info;
-       acpi_status      status;
-       struct SPMITable *spmi;
-       char             *io_type;
        u8               addr_space;
 
-       if (acpi_failure)
-               return -ENODEV;
-
-       status = acpi_get_firmware_table("SPMI", intf_num+1,
-                                        ACPI_LOGICAL_ADDRESSING,
-                                        (struct acpi_table_header **) &spmi);
-       if (status != AE_OK) {
-               acpi_failure = 1;
-               return -ENODEV;
-       }
-
        if (spmi->IPMIlegacy != 1) {
            printk(KERN_INFO "IPMI: Bad SPMI legacy %d\n", spmi->IPMIlegacy);
            return -ENODEV;
@@ -1255,376 +1806,381 @@ static int try_init_acpi(int intf_num, struct smi_info **new_info)
                addr_space = IPMI_MEM_ADDR_SPACE;
        else
                addr_space = IPMI_IO_ADDR_SPACE;
-       if (!is_new_interface(-1, addr_space, spmi->addr.address))
-               return -ENODEV;
+
+       info = kzalloc(sizeof(*info), GFP_KERNEL);
+       if (!info) {
+               printk(KERN_ERR "ipmi_si: Could not allocate SI data (3)\n");
+               return -ENOMEM;
+       }
+
+       info->addr_source = "ACPI";
 
        /* Figure out the interface type. */
        switch (spmi->InterfaceType)
        {
        case 1: /* KCS */
-               si_type[intf_num] = "kcs";
+               info->si_type = SI_KCS;
                break;
-
        case 2: /* SMIC */
-               si_type[intf_num] = "smic";
+               info->si_type = SI_SMIC;
                break;
-
        case 3: /* BT */
-               si_type[intf_num] = "bt";
+               info->si_type = SI_BT;
                break;
-
        default:
                printk(KERN_INFO "ipmi_si: Unknown ACPI/SPMI SI type %d\n",
                        spmi->InterfaceType);
+               kfree(info);
                return -EIO;
        }
 
-       info = kmalloc(sizeof(*info), GFP_KERNEL);
-       if (!info) {
-               printk(KERN_ERR "ipmi_si: Could not allocate SI data (3)\n");
-               return -ENOMEM;
-       }
-       memset(info, 0, sizeof(*info));
-
        if (spmi->InterruptType & 1) {
                /* We've got a GPE interrupt. */
                info->irq = spmi->GPE;
                info->irq_setup = acpi_gpe_irq_setup;
-               info->irq_cleanup = acpi_gpe_irq_cleanup;
        } else if (spmi->InterruptType & 2) {
                /* We've got an APIC/SAPIC interrupt. */
                info->irq = spmi->GlobalSystemInterrupt;
                info->irq_setup = std_irq_setup;
-               info->irq_cleanup = std_irq_cleanup;
        } else {
                /* Use the default interrupt setting. */
                info->irq = 0;
                info->irq_setup = NULL;
        }
 
+       if (spmi->addr.register_bit_width) {
+               /* A (hopefully) properly formed register bit width. */
+               info->io.regspacing = spmi->addr.register_bit_width / 8;
+       } else {
+               info->io.regspacing = DEFAULT_REGSPACING;
+       }
+       info->io.regsize = info->io.regspacing;
+       info->io.regshift = spmi->addr.register_bit_offset;
+
        if (spmi->addr.address_space_id == ACPI_ADR_SPACE_SYSTEM_MEMORY) {
-               io_type = "memory";
                info->io_setup = mem_setup;
-               info->io_cleanup = mem_cleanup;
-               addrs[intf_num] = spmi->addr.address;
-               info->io.inputb = mem_inb;
-               info->io.outputb = mem_outb;
-               info->io.info = &(addrs[intf_num]);
+               info->io.addr_type = IPMI_IO_ADDR_SPACE;
        } else if (spmi->addr.address_space_id == ACPI_ADR_SPACE_SYSTEM_IO) {
-               io_type = "I/O";
                info->io_setup = port_setup;
-               info->io_cleanup = port_cleanup;
-               ports[intf_num] = spmi->addr.address;
-               info->io.inputb = port_inb;
-               info->io.outputb = port_outb;
-               info->io.info = &(ports[intf_num]);
+               info->io.addr_type = IPMI_MEM_ADDR_SPACE;
        } else {
                kfree(info);
                printk("ipmi_si: Unknown ACPI I/O Address type\n");
                return -EIO;
        }
+       info->io.addr_data = spmi->addr.address;
 
-       *new_info = info;
+       try_smi_init(info);
 
-       printk("ipmi_si: ACPI/SPMI specifies \"%s\" %s SI @ 0x%lx\n",
-              si_type[intf_num], io_type, (unsigned long) spmi->addr.address);
        return 0;
 }
-#endif
 
-#ifdef CONFIG_X86
+static __devinit void acpi_find_bmc(void)
+{
+       acpi_status      status;
+       struct SPMITable *spmi;
+       int              i;
+
+       if (acpi_disabled)
+               return;
+
+       if (acpi_failure)
+               return;
 
-typedef struct dmi_ipmi_data
+       for (i = 0; ; i++) {
+               status = acpi_get_firmware_table("SPMI", i+1,
+                                                ACPI_LOGICAL_ADDRESSING,
+                                                (struct acpi_table_header **)
+                                                &spmi);
+               if (status != AE_OK)
+                       return;
+
+               try_init_acpi(spmi);
+       }
+}
+#endif
+
+#ifdef CONFIG_DMI
+struct dmi_ipmi_data
 {
        u8              type;
        u8              addr_space;
        unsigned long   base_addr;
        u8              irq;
-}dmi_ipmi_data_t;
-
-typedef struct dmi_header
-{
-       u8      type;
-       u8      length;
-       u16     handle;
-}dmi_header_t;
+       u8              offset;
+       u8              slave_addr;
+};
 
-static int decode_dmi(dmi_header_t *dm, dmi_ipmi_data_t *ipmi_data)
+static int __devinit decode_dmi(struct dmi_header *dm,
+                               struct dmi_ipmi_data *dmi)
 {
-       u8              *data = (u8 *)dm;
+       u8              *data = (u8 *)dm;
        unsigned long   base_addr;
+       u8              reg_spacing;
+       u8              len = dm->length;
 
-       ipmi_data->type = data[0x04];
+       dmi->type = data[4];
 
-       memcpy(&base_addr,&data[0x08],sizeof(unsigned long));
-       if (base_addr & 1) {
-               /* I/O */
-               base_addr &= 0xFFFE;
-               ipmi_data->addr_space = IPMI_IO_ADDR_SPACE;
-       }
-       else {
-               /* Memory */
-               ipmi_data->addr_space = IPMI_MEM_ADDR_SPACE;
+       memcpy(&base_addr, data+8, sizeof(unsigned long));
+       if (len >= 0x11) {
+               if (base_addr & 1) {
+                       /* I/O */
+                       base_addr &= 0xFFFE;
+                       dmi->addr_space = IPMI_IO_ADDR_SPACE;
+               }
+               else {
+                       /* Memory */
+                       dmi->addr_space = IPMI_MEM_ADDR_SPACE;
+               }
+               /* If bit 4 of byte 0x10 is set, then the lsb for the address
+                  is odd. */
+               dmi->base_addr = base_addr | ((data[0x10] & 0x10) >> 4);
+
+               dmi->irq = data[0x11];
+
+               /* The top two bits of byte 0x10 hold the register spacing. */
+               reg_spacing = (data[0x10] & 0xC0) >> 6;
+               switch(reg_spacing){
+               case 0x00: /* Byte boundaries */
+                   dmi->offset = 1;
+                   break;
+               case 0x01: /* 32-bit boundaries */
+                   dmi->offset = 4;
+                   break;
+               case 0x02: /* 16-byte boundaries */
+                   dmi->offset = 16;
+                   break;
+               default:
+                   /* Some other interface, just ignore it. */
+                   return -EIO;
+               }
+       } else {
+               /* Old DMI spec. */
+               /* Note that technically, the lower bit of the base
+                * address should be 1 if the address is I/O and 0 if
+                * the address is in memory.  So many systems get that
+                * wrong (and all that I have seen are I/O) so we just
+                * ignore that bit and assume I/O.  Systems that use
+                * memory should use the newer spec, anyway. */
+               dmi->base_addr = base_addr & 0xfffe;
+               dmi->addr_space = IPMI_IO_ADDR_SPACE;
+               dmi->offset = 1;
        }
 
-       ipmi_data->base_addr = base_addr;
-       ipmi_data->irq = data[0x11];
-
-       if (is_new_interface(-1, ipmi_data->addr_space,ipmi_data->base_addr))
-           return 0;
-
-       memset(ipmi_data,0,sizeof(dmi_ipmi_data_t));
+       dmi->slave_addr = data[6];
 
-       return -1;
+       return 0;
 }
 
-static int dmi_table(u32 base, int len, int num,
-       dmi_ipmi_data_t *ipmi_data)
+static __devinit void try_init_dmi(struct dmi_ipmi_data *ipmi_data)
 {
-       u8                *buf;
-       struct dmi_header *dm;
-       u8                *data;
-       int               i=1;
-       int               status=-1;
-
-       buf = ioremap(base, len);
-       if(buf==NULL)
-               return -1;
-
-       data = buf;
-
-       while(i<num && (data - buf) < len)
-       {
-               dm=(dmi_header_t *)data;
+       struct smi_info *info;
 
-               if((data-buf+dm->length) >= len)
-                       break;
+       info = kzalloc(sizeof(*info), GFP_KERNEL);
+       if (!info) {
+               printk(KERN_ERR
+                      "ipmi_si: Could not allocate SI data\n");
+               return;
+       }
 
-               if (dm->type == 38) {
-                       if (decode_dmi(dm, ipmi_data) == 0) {
-                               status = 0;
-                               break;
-                       }
-               }
+       info->addr_source = "SMBIOS";
 
-               data+=dm->length;
-               while((data-buf) < len && (*data || data[1]))
-                       data++;
-               data+=2;
-               i++;
+       switch (ipmi_data->type) {
+       case 0x01: /* KCS */
+               info->si_type = SI_KCS;
+               break;
+       case 0x02: /* SMIC */
+               info->si_type = SI_SMIC;
+               break;
+       case 0x03: /* BT */
+               info->si_type = SI_BT;
+               break;
+       default:
+               return;
        }
-       iounmap(buf);
-
-       return status;
-}
 
-inline static int dmi_checksum(u8 *buf)
-{
-       u8   sum=0;
-       int  a;
+       switch (ipmi_data->addr_space) {
+       case IPMI_MEM_ADDR_SPACE:
+               info->io_setup = mem_setup;
+               info->io.addr_type = IPMI_MEM_ADDR_SPACE;
+               break;
 
-       for(a=0; a<15; a++)
-               sum+=buf[a];
-       return (sum==0);
-}
+       case IPMI_IO_ADDR_SPACE:
+               info->io_setup = port_setup;
+               info->io.addr_type = IPMI_IO_ADDR_SPACE;
+               break;
 
-static int dmi_iterator(dmi_ipmi_data_t *ipmi_data)
-{
-       u8   buf[15];
-       u32  fp=0xF0000;
+       default:
+               kfree(info);
+               printk(KERN_WARNING
+                      "ipmi_si: Unknown SMBIOS I/O Address type: %d.\n",
+                      ipmi_data->addr_space);
+               return;
+       }
+       info->io.addr_data = ipmi_data->base_addr;
 
-#ifdef CONFIG_SIMNOW
-       return -1;
-#endif
+       info->io.regspacing = ipmi_data->offset;
+       if (!info->io.regspacing)
+               info->io.regspacing = DEFAULT_REGSPACING;
+       info->io.regsize = DEFAULT_REGSPACING;
+       info->io.regshift = 0;
 
-       while(fp < 0xFFFFF)
-       {
-               isa_memcpy_fromio(buf, fp, 15);
-               if(memcmp(buf, "_DMI_", 5)==0 && dmi_checksum(buf))
-               {
-                       u16 num=buf[13]<<8|buf[12];
-                       u16 len=buf[7]<<8|buf[6];
-                       u32 base=buf[11]<<24|buf[10]<<16|buf[9]<<8|buf[8];
+       info->slave_addr = ipmi_data->slave_addr;
 
-                       if(dmi_table(base, len, num, ipmi_data) == 0)
-                               return 0;
-               }
-               fp+=16;
-       }
+       info->irq = ipmi_data->irq;
+       if (info->irq)
+               info->irq_setup = std_irq_setup;
 
-       return -1;
+       try_smi_init(info);
 }
 
-static int try_init_smbios(int intf_num, struct smi_info **new_info)
+static void __devinit dmi_find_bmc(void)
 {
-       struct smi_info   *info;
-       dmi_ipmi_data_t   ipmi_data;
-       char              *io_type;
-       int               status;
-
-       status = dmi_iterator(&ipmi_data);
-
-       if (status < 0)
-               return -ENODEV;
-
-       switch(ipmi_data.type) {
-               case 0x01: /* KCS */
-                       si_type[intf_num] = "kcs";
-                       break;
-               case 0x02: /* SMIC */
-                       si_type[intf_num] = "smic";
-                       break;
-               case 0x03: /* BT */
-                       si_type[intf_num] = "bt";
-                       break;
-               default:
-                       printk("ipmi_si: Unknown SMBIOS SI type.\n");
-                       return -EIO;
-       }
-
-       info = kmalloc(sizeof(*info), GFP_KERNEL);
-       if (!info) {
-               printk(KERN_ERR "ipmi_si: Could not allocate SI data (4)\n");
-               return -ENOMEM;
-       }
-       memset(info, 0, sizeof(*info));
+       struct dmi_device    *dev = NULL;
+       struct dmi_ipmi_data data;
+       int                  rv;
 
-       if (ipmi_data.addr_space == 1) {
-               io_type = "memory";
-               info->io_setup = mem_setup;
-               info->io_cleanup = mem_cleanup;
-               addrs[intf_num] = ipmi_data.base_addr;
-               info->io.inputb = mem_inb;
-               info->io.outputb = mem_outb;
-               info->io.info = &(addrs[intf_num]);
-       } else if (ipmi_data.addr_space == 2) {
-               io_type = "I/O";
-               info->io_setup = port_setup;
-               info->io_cleanup = port_cleanup;
-               ports[intf_num] = ipmi_data.base_addr;
-               info->io.inputb = port_inb;
-               info->io.outputb = port_outb;
-               info->io.info = &(ports[intf_num]);
-       } else {
-               kfree(info);
-               printk("ipmi_si: Unknown SMBIOS I/O Address type.\n");
-               return -EIO;
+       while ((dev = dmi_find_device(DMI_DEV_TYPE_IPMI, NULL, dev))) {
+               memset(&data, 0, sizeof(data));
+               rv = decode_dmi((struct dmi_header *) dev->device_data, &data);
+               if (!rv)
+                       try_init_dmi(&data);
        }
-
-       irqs[intf_num] = ipmi_data.irq;
-
-       *new_info = info;
-
-       printk("ipmi_si: Found SMBIOS-specified state machine at %s"
-              " address 0x%lx\n",
-              io_type, (unsigned long)ipmi_data.base_addr);
-       return 0;
 }
-#endif /* CONFIG_X86 */
+#endif /* CONFIG_DMI */
 
 #ifdef CONFIG_PCI
 
-#define PCI_ERMC_CLASSCODE  0x0C0700
+#define PCI_ERMC_CLASSCODE             0x0C0700
+#define PCI_ERMC_CLASSCODE_MASK                0xffffff00
+#define PCI_ERMC_CLASSCODE_TYPE_MASK   0xff
+#define PCI_ERMC_CLASSCODE_TYPE_SMIC   0x00
+#define PCI_ERMC_CLASSCODE_TYPE_KCS    0x01
+#define PCI_ERMC_CLASSCODE_TYPE_BT     0x02
+
 #define PCI_HP_VENDOR_ID    0x103C
 #define PCI_MMC_DEVICE_ID   0x121A
 #define PCI_MMC_ADDR_CW     0x10
 
-/* Avoid more than one attempt to probe pci smic. */
-static int pci_smic_checked = 0;
+static void ipmi_pci_cleanup(struct smi_info *info)
+{
+       struct pci_dev *pdev = info->addr_source_data;
+
+       pci_disable_device(pdev);
+}
 
-static int find_pci_smic(int intf_num, struct smi_info **new_info)
+static int __devinit ipmi_pci_probe(struct pci_dev *pdev,
+                                   const struct pci_device_id *ent)
 {
-       struct smi_info  *info;
-       int              error;
-       struct pci_dev   *pci_dev = NULL;
-       u16              base_addr;
-       int              fe_rmc = 0;
+       int rv;
+       int class_type = pdev->class & PCI_ERMC_CLASSCODE_TYPE_MASK;
+       struct smi_info *info;
+       int first_reg_offset = 0;
 
-       if (pci_smic_checked)
-               return -ENODEV;
+       info = kzalloc(sizeof(*info), GFP_KERNEL);
+       if (!info)
+               return -ENOMEM;
 
-       pci_smic_checked = 1;
+       info->addr_source = "PCI";
 
-       if ((pci_dev = pci_find_device(PCI_HP_VENDOR_ID, PCI_MMC_DEVICE_ID,
-                                      NULL)))
-               ;
-       else if ((pci_dev = pci_find_class(PCI_ERMC_CLASSCODE, NULL)) &&
-                pci_dev->subsystem_vendor == PCI_HP_VENDOR_ID)
-               fe_rmc = 1;
-       else
-               return -ENODEV;
+       switch (class_type) {
+       case PCI_ERMC_CLASSCODE_TYPE_SMIC:
+               info->si_type = SI_SMIC;
+               break;
 
-       error = pci_read_config_word(pci_dev, PCI_MMC_ADDR_CW, &base_addr);
-       if (error)
-       {
-               printk(KERN_ERR
-                      "ipmi_si: pci_read_config_word() failed (%d).\n",
-                      error);
-               return -ENODEV;
+       case PCI_ERMC_CLASSCODE_TYPE_KCS:
+               info->si_type = SI_KCS;
+               break;
+
+       case PCI_ERMC_CLASSCODE_TYPE_BT:
+               info->si_type = SI_BT;
+               break;
+
+       default:
+               kfree(info);
+               printk(KERN_INFO "ipmi_si: %s: Unknown IPMI type: %d\n",
+                      pci_name(pdev), class_type);
+               return -ENOMEM;
        }
 
-       /* Bit 0: 1 specifies programmed I/O, 0 specifies memory mapped I/O */
-       if (!(base_addr & 0x0001))
-       {
-               printk(KERN_ERR
-                      "ipmi_si: memory mapped I/O not supported for PCI"
-                      " smic.\n");
-               return -ENODEV;
+       rv = pci_enable_device(pdev);
+       if (rv) {
+               printk(KERN_ERR "ipmi_si: %s: couldn't enable PCI device\n",
+                      pci_name(pdev));
+               kfree(info);
+               return rv;
        }
 
-       base_addr &= 0xFFFE;
-       if (!fe_rmc)
-               /* Data register starts at base address + 1 in eRMC */
-               ++base_addr;
+       info->addr_source_cleanup = ipmi_pci_cleanup;
+       info->addr_source_data = pdev;
 
-       if (!is_new_interface(-1, IPMI_IO_ADDR_SPACE, base_addr))
-           return -ENODEV;
+       if (pdev->subsystem_vendor == PCI_HP_VENDOR_ID)
+               first_reg_offset = 1;
 
-       info = kmalloc(sizeof(*info), GFP_KERNEL);
-       if (!info) {
-               printk(KERN_ERR "ipmi_si: Could not allocate SI data (5)\n");
-               return -ENOMEM;
+       if (pci_resource_flags(pdev, 0) & IORESOURCE_IO) {
+               info->io_setup = port_setup;
+               info->io.addr_type = IPMI_IO_ADDR_SPACE;
+       } else {
+               info->io_setup = mem_setup;
+               info->io.addr_type = IPMI_MEM_ADDR_SPACE;
        }
-       memset(info, 0, sizeof(*info));
+       info->io.addr_data = pci_resource_start(pdev, 0);
 
-       info->io_setup = port_setup;
-       info->io_cleanup = port_cleanup;
-       ports[intf_num] = base_addr;
-       info->io.inputb = port_inb;
-       info->io.outputb = port_outb;
-       info->io.info = &(ports[intf_num]);
+       info->io.regspacing = DEFAULT_REGSPACING;
+       info->io.regsize = DEFAULT_REGSPACING;
+       info->io.regshift = 0;
 
-       *new_info = info;
+       info->irq = pdev->irq;
+       if (info->irq)
+               info->irq_setup = std_irq_setup;
 
-       irqs[intf_num] = pci_dev->irq;
-       si_type[intf_num] = "smic";
+       info->dev = &pdev->dev;
 
-       printk("ipmi_si: Found PCI SMIC at I/O address 0x%lx\n",
-               (long unsigned int) base_addr);
+       return try_smi_init(info);
+}
+
+static void __devexit ipmi_pci_remove(struct pci_dev *pdev)
+{
+}
 
+#ifdef CONFIG_PM
+static int ipmi_pci_suspend(struct pci_dev *pdev, pm_message_t state)
+{
        return 0;
 }
-#endif /* CONFIG_PCI */
 
-static int try_init_plug_and_play(int intf_num, struct smi_info **new_info)
+static int ipmi_pci_resume(struct pci_dev *pdev)
 {
-#ifdef CONFIG_PCI
-       if (find_pci_smic(intf_num, new_info)==0)
-               return 0;
+       return 0;
+}
 #endif
-       /* Include other methods here. */
 
-       return -ENODEV;
-}
+static struct pci_device_id ipmi_pci_devices[] = {
+       { PCI_DEVICE(PCI_HP_VENDOR_ID, PCI_MMC_DEVICE_ID) },
+       { PCI_DEVICE_CLASS(PCI_ERMC_CLASSCODE, PCI_ERMC_CLASSCODE_MASK) }
+};
+MODULE_DEVICE_TABLE(pci, ipmi_pci_devices);
+
+static struct pci_driver ipmi_pci_driver = {
+        .name =         DEVICE_NAME,
+        .id_table =     ipmi_pci_devices,
+        .probe =        ipmi_pci_probe,
+        .remove =       __devexit_p(ipmi_pci_remove),
+#ifdef CONFIG_PM
+        .suspend =      ipmi_pci_suspend,
+        .resume =       ipmi_pci_resume,
+#endif
+};
+#endif /* CONFIG_PCI */
 
 
 static int try_get_dev_id(struct smi_info *smi_info)
 {
-       unsigned char      msg[2];
-       unsigned char      *resp;
-       unsigned long      resp_len;
-       enum si_sm_result smi_result;
-       int               rv = 0;
+       unsigned char         msg[2];
+       unsigned char         *resp;
+       unsigned long         resp_len;
+       enum si_sm_result     smi_result;
+       int                   rv = 0;
 
        resp = kmalloc(IPMI_MAX_MSG_LENGTH, GFP_KERNEL);
        if (!resp)
@@ -1639,9 +2195,9 @@ static int try_get_dev_id(struct smi_info *smi_info)
        smi_result = smi_info->handlers->event(smi_info->si_sm, 0);
        for (;;)
        {
-               if (smi_result == SI_SM_CALL_WITH_DELAY) {
-                       set_current_state(TASK_UNINTERRUPTIBLE);
-                       schedule_timeout(1);
+               if (smi_result == SI_SM_CALL_WITH_DELAY ||
+                   smi_result == SI_SM_CALL_WITH_TICK_DELAY) {
+                       schedule_timeout_uninterruptible(1);
                        smi_result = smi_info->handlers->event(
                                smi_info->si_sm, 100);
                }
@@ -1663,7 +2219,7 @@ static int try_get_dev_id(struct smi_info *smi_info)
        /* Otherwise, we got some data. */
        resp_len = smi_info->handlers->get_result(smi_info->si_sm,
                                                  resp, IPMI_MAX_MSG_LENGTH);
-       if (resp_len < 6) {
+       if (resp_len < 14) {
                /* That's odd, it should be longer. */
                rv = -EINVAL;
                goto out;
@@ -1676,11 +2232,7 @@ static int try_get_dev_id(struct smi_info *smi_info)
        }
 
        /* Record info from the get device id, in case we need it. */
-       smi_info->ipmi_si_dev_rev = resp[4] & 0xf;
-       smi_info->ipmi_si_fw_rev_major = resp[5] & 0x7f;
-       smi_info->ipmi_si_fw_rev_minor = resp[6];
-       smi_info->ipmi_version_major = resp[7] & 0xf;
-       smi_info->ipmi_version_minor = resp[7] >> 4;
+       ipmi_demangle_device_id(resp+3, resp_len-3, &smi_info->device_id);
 
  out:
        kfree(resp);
@@ -1690,19 +2242,9 @@ static int try_get_dev_id(struct smi_info *smi_info)
 static int type_file_read_proc(char *page, char **start, off_t off,
                               int count, int *eof, void *data)
 {
-       char            *out = (char *) page;
        struct smi_info *smi = data;
 
-       switch (smi->si_type) {
-           case SI_KCS:
-               return sprintf(out, "kcs\n");
-           case SI_SMIC:
-               return sprintf(out, "smic\n");
-           case SI_BT:
-               return sprintf(out, "bt\n");
-           default:
-               return 0;
-       }
+       return sprintf(page, "%s\n", si_to_str[smi->si_type]);
 }
 
 static int stat_file_read_proc(char *page, char **start, off_t off,
@@ -1738,69 +2280,285 @@ static int stat_file_read_proc(char *page, char **start, off_t off,
        out += sprintf(out, "incoming_messages:     %ld\n",
                       smi->incoming_messages);
 
-       return (out - ((char *) page));
+       return out - page;
 }
 
-/* Returns 0 if initialized, or negative on an error. */
-static int init_one_smi(int intf_num, struct smi_info **smi)
+static int param_read_proc(char *page, char **start, off_t off,
+                          int count, int *eof, void *data)
 {
-       int             rv;
-       struct smi_info *new_smi;
+       struct smi_info *smi = data;
 
+       return sprintf(page,
+                      "%s,%s,0x%lx,rsp=%d,rsi=%d,rsh=%d,irq=%d,ipmb=%d\n",
+                      si_to_str[smi->si_type],
+                      addr_space_to_str[smi->io.addr_type],
+                      smi->io.addr_data,
+                      smi->io.regspacing,
+                      smi->io.regsize,
+                      smi->io.regshift,
+                      smi->irq,
+                      smi->slave_addr);
+}
 
-       rv = try_init_mem(intf_num, &new_smi);
-       if (rv)
-               rv = try_init_port(intf_num, &new_smi);
-#ifdef CONFIG_ACPI_INTERPRETER
-       if ((rv) && (si_trydefaults)) {
-               rv = try_init_acpi(intf_num, &new_smi);
-       }
-#endif
-#ifdef CONFIG_X86
-       if ((rv) && (si_trydefaults)) {
-               rv = try_init_smbios(intf_num, &new_smi);
-        }
-#endif
-       if ((rv) && (si_trydefaults)) {
-               rv = try_init_plug_and_play(intf_num, &new_smi);
+/*
+ * oem_data_avail_to_receive_msg_avail
+ * @info - smi_info structure with msg_flags set
+ *
+ * Converts flags from OEM_DATA_AVAIL to RECEIVE_MSG_AVAIL
+ * Returns 1 indicating need to re-run handle_flags().
+ */
+static int oem_data_avail_to_receive_msg_avail(struct smi_info *smi_info)
+{
+       smi_info->msg_flags = ((smi_info->msg_flags & ~OEM_DATA_AVAIL) |
+                               RECEIVE_MSG_AVAIL);
+       return 1;
+}
+
+/*
+ * setup_dell_poweredge_oem_data_handler
+ * @info - smi_info.device_id must be populated
+ *
+ * Systems that match, but have firmware version < 1.40 may assert
+ * OEM0_DATA_AVAIL on their own, without being told via Set Flags that
+ * it's safe to do so.  Such systems will de-assert OEM1_DATA_AVAIL
+ * upon receipt of IPMI_GET_MSG_CMD, so we should treat these flags
+ * as RECEIVE_MSG_AVAIL instead.
+ *
+ * As Dell has no plans to release IPMI 1.5 firmware that *ever*
+ * assert the OEM[012] bits, and if it did, the driver would have to
+ * change to handle that properly, we don't actually check for the
+ * firmware version.
+ * Device ID = 0x20                BMC on PowerEdge 8G servers
+ * Device Revision = 0x80
+ * Firmware Revision1 = 0x01       BMC version 1.40
+ * Firmware Revision2 = 0x40       BCD encoded
+ * IPMI Version = 0x51             IPMI 1.5
+ * Manufacturer ID = A2 02 00      Dell IANA
+ *
+ * Additionally, PowerEdge systems with IPMI < 1.5 may also assert
+ * OEM0_DATA_AVAIL and needs to be treated as RECEIVE_MSG_AVAIL.
+ *
+ */
+#define DELL_POWEREDGE_8G_BMC_DEVICE_ID  0x20
+#define DELL_POWEREDGE_8G_BMC_DEVICE_REV 0x80
+#define DELL_POWEREDGE_8G_BMC_IPMI_VERSION 0x51
+#define DELL_IANA_MFR_ID 0x0002a2
+static void setup_dell_poweredge_oem_data_handler(struct smi_info *smi_info)
+{
+       struct ipmi_device_id *id = &smi_info->device_id;
+       if (id->manufacturer_id == DELL_IANA_MFR_ID) {
+               if (id->device_id       == DELL_POWEREDGE_8G_BMC_DEVICE_ID  &&
+                   id->device_revision == DELL_POWEREDGE_8G_BMC_DEVICE_REV &&
+                   id->ipmi_version   == DELL_POWEREDGE_8G_BMC_IPMI_VERSION) {
+                       smi_info->oem_data_avail_handler =
+                               oem_data_avail_to_receive_msg_avail;
+               }
+               else if (ipmi_version_major(id) < 1 ||
+                        (ipmi_version_major(id) == 1 &&
+                         ipmi_version_minor(id) < 5)) {
+                       smi_info->oem_data_avail_handler =
+                               oem_data_avail_to_receive_msg_avail;
+               }
        }
+}
 
+#define CANNOT_RETURN_REQUESTED_LENGTH 0xCA
+static void return_hosed_msg_badsize(struct smi_info *smi_info)
+{
+       struct ipmi_smi_msg *msg = smi_info->curr_msg;
 
-       if (rv)
-               return rv;
+       /* Make it a reponse */
+       msg->rsp[0] = msg->data[0] | 4;
+       msg->rsp[1] = msg->data[1];
+       msg->rsp[2] = CANNOT_RETURN_REQUESTED_LENGTH;
+       msg->rsp_size = 3;
+       smi_info->curr_msg = NULL;
+       deliver_recv_msg(smi_info, msg);
+}
 
-       /* So we know not to free it unless we have allocated one. */
-       new_smi->intf = NULL;
-       new_smi->si_sm = NULL;
-       new_smi->handlers = 0;
+/*
+ * dell_poweredge_bt_xaction_handler
+ * @info - smi_info.device_id must be populated
+ *
+ * Dell PowerEdge servers with the BT interface (x6xx and 1750) will
+ * not respond to a Get SDR command if the length of the data
+ * requested is exactly 0x3A, which leads to command timeouts and no
+ * data returned.  This intercepts such commands, and causes userspace
+ * callers to try again with a different-sized buffer, which succeeds.
+ */
+
+#define STORAGE_NETFN 0x0A
+#define STORAGE_CMD_GET_SDR 0x23
+static int dell_poweredge_bt_xaction_handler(struct notifier_block *self,
+                                            unsigned long unused,
+                                            void *in)
+{
+       struct smi_info *smi_info = in;
+       unsigned char *data = smi_info->curr_msg->data;
+       unsigned int size   = smi_info->curr_msg->data_size;
+       if (size >= 8 &&
+           (data[0]>>2) == STORAGE_NETFN &&
+           data[1] == STORAGE_CMD_GET_SDR &&
+           data[7] == 0x3A) {
+               return_hosed_msg_badsize(smi_info);
+               return NOTIFY_STOP;
+       }
+       return NOTIFY_DONE;
+}
 
-       if (!new_smi->irq_setup) {
-               new_smi->irq = irqs[intf_num];
-               new_smi->irq_setup = std_irq_setup;
-               new_smi->irq_cleanup = std_irq_cleanup;
+static struct notifier_block dell_poweredge_bt_xaction_notifier = {
+       .notifier_call  = dell_poweredge_bt_xaction_handler,
+};
+
+/*
+ * setup_dell_poweredge_bt_xaction_handler
+ * @info - smi_info.device_id must be filled in already
+ *
+ * Fills in smi_info.device_id.start_transaction_pre_hook
+ * when we know what function to use there.
+ */
+static void
+setup_dell_poweredge_bt_xaction_handler(struct smi_info *smi_info)
+{
+       struct ipmi_device_id *id = &smi_info->device_id;
+       if (id->manufacturer_id == DELL_IANA_MFR_ID &&
+           smi_info->si_type == SI_BT)
+               register_xaction_notifier(&dell_poweredge_bt_xaction_notifier);
+}
+
+/*
+ * setup_oem_data_handler
+ * @info - smi_info.device_id must be filled in already
+ *
+ * Fills in smi_info.device_id.oem_data_available_handler
+ * when we know what function to use there.
+ */
+
+static void setup_oem_data_handler(struct smi_info *smi_info)
+{
+       setup_dell_poweredge_oem_data_handler(smi_info);
+}
+
+static void setup_xaction_handlers(struct smi_info *smi_info)
+{
+       setup_dell_poweredge_bt_xaction_handler(smi_info);
+}
+
+static inline void wait_for_timer_and_thread(struct smi_info *smi_info)
+{
+       if (smi_info->intf) {
+               /* The timer and thread are only running if the
+                  interface has been started up and registered. */
+               if (smi_info->thread != NULL)
+                       kthread_stop(smi_info->thread);
+               del_timer_sync(&smi_info->si_timer);
        }
+}
 
-       /* Default to KCS if no type is specified. */
-       if (si_type[intf_num] == NULL) {
-               if (si_trydefaults)
-                       si_type[intf_num] = "kcs";
-               else {
-                       rv = -EINVAL;
-                       goto out_err;
+static __devinitdata struct ipmi_default_vals
+{
+       int type;
+       int port;
+} ipmi_defaults[] =
+{
+       { .type = SI_KCS, .port = 0xca2 },
+       { .type = SI_SMIC, .port = 0xca9 },
+       { .type = SI_BT, .port = 0xe4 },
+       { .port = 0 }
+};
+
+static __devinit void default_find_bmc(void)
+{
+       struct smi_info *info;
+       int             i;
+
+       for (i = 0; ; i++) {
+               if (!ipmi_defaults[i].port)
+                       break;
+
+               info = kzalloc(sizeof(*info), GFP_KERNEL);
+               if (!info)
+                       return;
+
+               info->addr_source = NULL;
+
+               info->si_type = ipmi_defaults[i].type;
+               info->io_setup = port_setup;
+               info->io.addr_data = ipmi_defaults[i].port;
+               info->io.addr_type = IPMI_IO_ADDR_SPACE;
+
+               info->io.addr = NULL;
+               info->io.regspacing = DEFAULT_REGSPACING;
+               info->io.regsize = DEFAULT_REGSPACING;
+               info->io.regshift = 0;
+
+               if (try_smi_init(info) == 0) {
+                       /* Found one... */
+                       printk(KERN_INFO "ipmi_si: Found default %s state"
+                              " machine at %s address 0x%lx\n",
+                              si_to_str[info->si_type],
+                              addr_space_to_str[info->io.addr_type],
+                              info->io.addr_data);
+                       return;
                }
        }
+}
+
+static int is_new_interface(struct smi_info *info)
+{
+       struct smi_info *e;
+
+       list_for_each_entry(e, &smi_infos, link) {
+               if (e->io.addr_type != info->io.addr_type)
+                       continue;
+               if (e->io.addr_data == info->io.addr_data)
+                       return 0;
+       }
 
-       /* Set up the state machine to use. */
-       if (strcmp(si_type[intf_num], "kcs") == 0) {
+       return 1;
+}
+
+static int try_smi_init(struct smi_info *new_smi)
+{
+       int rv;
+
+       if (new_smi->addr_source) {
+               printk(KERN_INFO "ipmi_si: Trying %s-specified %s state"
+                      " machine at %s address 0x%lx, slave address 0x%x,"
+                      " irq %d\n",
+                      new_smi->addr_source,
+                      si_to_str[new_smi->si_type],
+                      addr_space_to_str[new_smi->io.addr_type],
+                      new_smi->io.addr_data,
+                      new_smi->slave_addr, new_smi->irq);
+       }
+
+       mutex_lock(&smi_infos_lock);
+       if (!is_new_interface(new_smi)) {
+               printk(KERN_WARNING "ipmi_si: duplicate interface\n");
+               rv = -EBUSY;
+               goto out_err;
+       }
+
+       /* So we know not to free it unless we have allocated one. */
+       new_smi->intf = NULL;
+       new_smi->si_sm = NULL;
+       new_smi->handlers = NULL;
+
+       switch (new_smi->si_type) {
+       case SI_KCS:
                new_smi->handlers = &kcs_smi_handlers;
-               new_smi->si_type = SI_KCS;
-       } else if (strcmp(si_type[intf_num], "smic") == 0) {
+               break;
+
+       case SI_SMIC:
                new_smi->handlers = &smic_smi_handlers;
-               new_smi->si_type = SI_SMIC;
-       } else if (strcmp(si_type[intf_num], "bt") == 0) {
+               break;
+
+       case SI_BT:
                new_smi->handlers = &bt_smi_handlers;
-               new_smi->si_type = SI_BT;
-       } else {
+               break;
+
+       default:
                /* No support for anything else yet. */
                rv = -EIO;
                goto out_err;
@@ -1829,18 +2587,29 @@ static int init_one_smi(int intf_num, struct smi_info **smi)
 
        /* Do low-level detection first. */
        if (new_smi->handlers->detect(new_smi->si_sm)) {
+               if (new_smi->addr_source)
+                       printk(KERN_INFO "ipmi_si: Interface detection"
+                              " failed\n");
                rv = -ENODEV;
                goto out_err;
        }
 
        /* Attempt a get device id command.  If it fails, we probably
-           don't have a SMI here. */
+           don't have a BMC here. */
        rv = try_get_dev_id(new_smi);
-       if (rv)
+       if (rv) {
+               if (new_smi->addr_source)
+                       printk(KERN_INFO "ipmi_si: There appears to be no BMC"
+                              " at this location\n");
                goto out_err;
+       }
+
+       setup_oem_data_handler(new_smi);
+       setup_xaction_handlers(new_smi);
 
        /* Try to claim any interrupts. */
-       new_smi->irq_setup(new_smi);
+       if (new_smi->irq_setup)
+               new_smi->irq_setup(new_smi);
 
        INIT_LIST_HEAD(&(new_smi->xmit_msgs));
        INIT_LIST_HEAD(&(new_smi->hp_xmit_msgs));
@@ -1848,16 +2617,55 @@ static int init_one_smi(int intf_num, struct smi_info **smi)
        atomic_set(&new_smi->req_events, 0);
        new_smi->run_to_completion = 0;
 
+       new_smi->interrupt_disabled = 0;
+       atomic_set(&new_smi->stop_operation, 0);
+       new_smi->intf_num = smi_num;
+       smi_num++;
+
+       /* Start clearing the flags before we enable interrupts or the
+          timer to avoid racing with the timer. */
+       start_clear_flags(new_smi);
+       /* IRQ is defined to be set when non-zero. */
+       if (new_smi->irq)
+               new_smi->si_state = SI_CLEARING_FLAGS_THEN_SET_IRQ;
+
+       if (!new_smi->dev) {
+               /* If we don't already have a device from something
+                * else (like PCI), then register a new one. */
+               new_smi->pdev = platform_device_alloc("ipmi_si",
+                                                     new_smi->intf_num);
+               if (rv) {
+                       printk(KERN_ERR
+                              "ipmi_si_intf:"
+                              " Unable to allocate platform device\n");
+                       goto out_err;
+               }
+               new_smi->dev = &new_smi->pdev->dev;
+               new_smi->dev->driver = &ipmi_driver;
+
+               rv = platform_device_add(new_smi->pdev);
+               if (rv) {
+                       printk(KERN_ERR
+                              "ipmi_si_intf:"
+                              " Unable to register system interface device:"
+                              " %d\n",
+                              rv);
+                       goto out_err;
+               }
+               new_smi->dev_registered = 1;
+       }
+
        rv = ipmi_register_smi(&handlers,
                               new_smi,
-                              new_smi->ipmi_version_major,
-                              new_smi->ipmi_version_minor,
-                              &(new_smi->intf));
+                              &new_smi->device_id,
+                              new_smi->dev,
+                              "bmc",
+                              new_smi->slave_addr);
        if (rv) {
                printk(KERN_ERR
                       "ipmi_si: Unable to register device: error %d\n",
                       rv);
-               goto out_err;
+               goto out_err_stop_timer;
        }
 
        rv = ipmi_smi_add_proc_entry(new_smi->intf, "type",
@@ -1867,7 +2675,7 @@ static int init_one_smi(int intf_num, struct smi_info **smi)
                printk(KERN_ERR
                       "ipmi_si: Unable to create proc entry: %d\n",
                       rv);
-               goto out_err;
+               goto out_err_stop_timer;
        }
 
        rv = ipmi_smi_add_proc_entry(new_smi->intf, "si_stats",
@@ -1877,61 +2685,87 @@ static int init_one_smi(int intf_num, struct smi_info **smi)
                printk(KERN_ERR
                       "ipmi_si: Unable to create proc entry: %d\n",
                       rv);
-               goto out_err;
+               goto out_err_stop_timer;
        }
 
-       start_clear_flags(new_smi);
-
-       /* IRQ is defined to be set when non-zero. */
-       if (new_smi->irq)
-               new_smi->si_state = SI_CLEARING_FLAGS_THEN_SET_IRQ;
-
-       new_smi->interrupt_disabled = 0;
-       new_smi->timer_stopped = 0;
-       new_smi->stop_operation = 0;
+       rv = ipmi_smi_add_proc_entry(new_smi->intf, "params",
+                                    param_read_proc, NULL,
+                                    new_smi, THIS_MODULE);
+       if (rv) {
+               printk(KERN_ERR
+                      "ipmi_si: Unable to create proc entry: %d\n",
+                      rv);
+               goto out_err_stop_timer;
+       }
 
-       init_timer(&(new_smi->si_timer));
-       new_smi->si_timer.data = (long) new_smi;
-       new_smi->si_timer.function = smi_timeout;
-       new_smi->last_timeout_jiffies = jiffies;
-       new_smi->si_timer.expires = jiffies + SI_TIMEOUT_JIFFIES;
-       add_timer(&(new_smi->si_timer));
+       list_add_tail(&new_smi->link, &smi_infos);
 
-       *smi = new_smi;
+       mutex_unlock(&smi_infos_lock);
 
-       printk(" IPMI %s interface initialized\n", si_type[intf_num]);
+       printk(" IPMI %s interface initialized\n",si_to_str[new_smi->si_type]);
 
        return 0;
 
+ out_err_stop_timer:
+       atomic_inc(&new_smi->stop_operation);
+       wait_for_timer_and_thread(new_smi);
+
  out_err:
        if (new_smi->intf)
                ipmi_unregister_smi(new_smi->intf);
 
-       new_smi->irq_cleanup(new_smi);
+       if (new_smi->irq_cleanup)
+               new_smi->irq_cleanup(new_smi);
+
+       /* Wait until we know that we are out of any interrupt
+          handlers might have been running before we freed the
+          interrupt. */
+       synchronize_sched();
+
        if (new_smi->si_sm) {
                if (new_smi->handlers)
                        new_smi->handlers->cleanup(new_smi->si_sm);
                kfree(new_smi->si_sm);
        }
-       new_smi->io_cleanup(new_smi);
+       if (new_smi->addr_source_cleanup)
+               new_smi->addr_source_cleanup(new_smi);
+       if (new_smi->io_cleanup)
+               new_smi->io_cleanup(new_smi);
+
+       if (new_smi->dev_registered)
+               platform_device_unregister(new_smi->pdev);
+
+       kfree(new_smi);
+
+       mutex_unlock(&smi_infos_lock);
+
        return rv;
 }
 
-static __init int init_ipmi_si(void)
+static __devinit int init_ipmi_si(void)
 {
-       int  rv = 0;
-       int  pos = 0;
        int  i;
        char *str;
+       int  rv;
 
        if (initialized)
                return 0;
        initialized = 1;
 
+       /* Register the device drivers. */
+       rv = driver_register(&ipmi_driver);
+       if (rv) {
+               printk(KERN_ERR
+                      "init_ipmi_si: Unable to register driver: %d\n",
+                      rv);
+               return rv;
+       }
+
+
        /* Parse out the si_type string into its components. */
        str = si_type_str;
        if (*str != '\0') {
-               for (i=0; (i<SI_MAX_PARMS) && (*str != '\0'); i++) {
+               for (i = 0; (i < SI_MAX_PARMS) && (*str != '\0'); i++) {
                        si_type[i] = str;
                        str = strchr(str, ',');
                        if (str) {
@@ -1943,70 +2777,73 @@ static __init int init_ipmi_si(void)
                }
        }
 
-       printk(KERN_INFO "IPMI System Interface driver version "
-              IPMI_SI_VERSION);
-       if (kcs_smi_handlers.version)
-               printk(", KCS version %s", kcs_smi_handlers.version);
-       if (smic_smi_handlers.version)
-               printk(", SMIC version %s", smic_smi_handlers.version);
-       if (bt_smi_handlers.version)
-               printk(", BT version %s", bt_smi_handlers.version);
-       printk("\n");
-
-       rv = init_one_smi(0, &(smi_infos[pos]));
-       if (rv && !ports[0] && si_trydefaults) {
-               /* If we are trying defaults and the initial port is
-                   not set, then set it. */
-               si_type[0] = "kcs";
-               ports[0] = DEFAULT_KCS_IO_PORT;
-               rv = init_one_smi(0, &(smi_infos[pos]));
-               if (rv) {
-                       /* No KCS - try SMIC */
-                       si_type[0] = "smic";
-                       ports[0] = DEFAULT_SMIC_IO_PORT;
-                       rv = init_one_smi(0, &(smi_infos[pos]));
-               }
-               if (rv) {
-                       /* No SMIC - try BT */
-                       si_type[0] = "bt";
-                       ports[0] = DEFAULT_BT_IO_PORT;
-                       rv = init_one_smi(0, &(smi_infos[pos]));
-               }
+       printk(KERN_INFO "IPMI System Interface driver.\n");
+
+       hardcode_find_bmc();
+
+#ifdef CONFIG_DMI
+       dmi_find_bmc();
+#endif
+
+#ifdef CONFIG_ACPI
+       acpi_find_bmc();
+#endif
+
+#ifdef CONFIG_PCI
+       rv = pci_register_driver(&ipmi_pci_driver);
+       if (rv){
+               printk(KERN_ERR
+                      "init_ipmi_si: Unable to register PCI driver: %d\n",
+                      rv);
        }
-       if (rv == 0)
-               pos++;
+#endif
 
-       for (i=1; i < SI_MAX_PARMS; i++) {
-               rv = init_one_smi(i, &(smi_infos[pos]));
-               if (rv == 0)
-                       pos++;
+       if (si_trydefaults) {
+               mutex_lock(&smi_infos_lock);
+               if (list_empty(&smi_infos)) {
+                       /* No BMC was found, try defaults. */
+                       mutex_unlock(&smi_infos_lock);
+                       default_find_bmc();
+               } else {
+                       mutex_unlock(&smi_infos_lock);
+               }
        }
 
-       if (smi_infos[0] == NULL) {
+       mutex_lock(&smi_infos_lock);
+       if (unload_when_empty && list_empty(&smi_infos)) {
+               mutex_unlock(&smi_infos_lock);
+#ifdef CONFIG_PCI
+               pci_unregister_driver(&ipmi_pci_driver);
+#endif
+               driver_unregister(&ipmi_driver);
                printk("ipmi_si: Unable to find any System Interface(s)\n");
                return -ENODEV;
+       } else {
+               mutex_unlock(&smi_infos_lock);
+               return 0;
        }
-
-       return 0;
 }
 module_init(init_ipmi_si);
 
-void __exit cleanup_one_si(struct smi_info *to_clean)
+static void cleanup_one_si(struct smi_info *to_clean)
 {
        int           rv;
        unsigned long flags;
 
-       if (! to_clean)
+       if (!to_clean)
                return;
 
+       list_del(&to_clean->link);
+
        /* Tell the timer and interrupt handlers that we are shutting
           down. */
        spin_lock_irqsave(&(to_clean->si_lock), flags);
        spin_lock(&(to_clean->msg_lock));
 
-       to_clean->stop_operation = 1;
+       atomic_inc(&to_clean->stop_operation);
 
-       to_clean->irq_cleanup(to_clean);
+       if (to_clean->irq_cleanup)
+               to_clean->irq_cleanup(to_clean);
 
        spin_unlock(&(to_clean->msg_lock));
        spin_unlock_irqrestore(&(to_clean->si_lock), flags);
@@ -2014,13 +2851,15 @@ void __exit cleanup_one_si(struct smi_info *to_clean)
        /* Wait until we know that we are out of any interrupt
           handlers might have been running before we freed the
           interrupt. */
-       synchronize_kernel();
+       synchronize_sched();
+
+       wait_for_timer_and_thread(to_clean);
 
-       /* Wait for the timer to stop.  This avoids problems with race
-          conditions removing the timer here. */
-       while (!to_clean->timer_stopped) {
-               set_current_state(TASK_UNINTERRUPTIBLE);
-               schedule_timeout(1);
+       /* Interrupts and timeouts are stopped, now make sure the
+          interface is in a clean state. */
+       while (to_clean->curr_msg || (to_clean->si_state != SI_NORMAL)) {
+               poll(to_clean);
+               schedule_timeout_uninterruptible(1);
        }
 
        rv = ipmi_unregister_smi(to_clean->intf);
@@ -2034,20 +2873,37 @@ void __exit cleanup_one_si(struct smi_info *to_clean)
 
        kfree(to_clean->si_sm);
 
-       to_clean->io_cleanup(to_clean);
+       if (to_clean->addr_source_cleanup)
+               to_clean->addr_source_cleanup(to_clean);
+       if (to_clean->io_cleanup)
+               to_clean->io_cleanup(to_clean);
+
+       if (to_clean->dev_registered)
+               platform_device_unregister(to_clean->pdev);
+
+       kfree(to_clean);
 }
 
 static __exit void cleanup_ipmi_si(void)
 {
-       int i;
+       struct smi_info *e, *tmp_e;
 
        if (!initialized)
                return;
 
-       for (i=0; i<SI_MAX_DRIVERS; i++) {
-               cleanup_one_si(smi_infos[i]);
-       }
+#ifdef CONFIG_PCI
+       pci_unregister_driver(&ipmi_pci_driver);
+#endif
+
+       mutex_lock(&smi_infos_lock);
+       list_for_each_entry_safe(e, tmp_e, &smi_infos, link)
+               cleanup_one_si(e);
+       mutex_unlock(&smi_infos_lock);
+
+       driver_unregister(&ipmi_driver);
 }
 module_exit(cleanup_ipmi_si);
 
 MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Corey Minyard <minyard@mvista.com>");
+MODULE_DESCRIPTION("Interface to the IPMI driver for the KCS, SMIC, and BT system interfaces.");