/*
 * ipmi_si.c
 *
 * The interface to the IPMI driver for the system interfaces (KCS, SMIC,
 * BT).
 *
 * Author: MontaVista Software, Inc.
 *         Corey Minyard <minyard@mvista.com>
 *         source@mvista.com
 *
 * Copyright 2002 MontaVista Software Inc.
 * Copyright 2006 IBM Corp., Christian Krafft <krafft@de.ibm.com>
 *
 *  This program is free software; you can redistribute it and/or modify it
 *  under the terms of the GNU General Public License as published by the
 *  Free Software Foundation; either version 2 of the License, or (at your
 *  option) any later version.
 *
 *
 *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
 *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
 *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 *  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
 *  INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
 *  BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
 *  OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
 *  ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
 *  TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
 *  USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 *  You should have received a copy of the GNU General Public License along
 *  with this program; if not, write to the Free Software Foundation, Inc.,
 *  675 Mass Ave, Cambridge, MA 02139, USA.
 */

/*
 * This file holds the "policy" for the interface to the SMI state
 * machine.  It does the configuration, handles timers and interrupts,
 * and drives the real SMI state machine.
 */

#include <linux/module.h>
#include <linux/moduleparam.h>
#include <asm/system.h>
#include <linux/sched.h>
#include <linux/seq_file.h>
#include <linux/timer.h>
#include <linux/errno.h>
#include <linux/spinlock.h>
#include <linux/slab.h>
#include <linux/delay.h>
#include <linux/list.h>
#include <linux/pci.h>
#include <linux/ioport.h>
#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.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>
#include <linux/pnp.h>
#include <linux/of_device.h>
#include <linux/of_platform.h>
#include <linux/of_address.h>
#include <linux/of_irq.h>

#define PFX "ipmi_si: "

/* Measure times between events in the driver. */
#undef DEBUG_TIMING

/* Call every 10 ms. */
#define SI_TIMEOUT_TIME_USEC	10000
#define SI_USEC_PER_JIFFY	(1000000/HZ)
#define SI_TIMEOUT_JIFFIES	(SI_TIMEOUT_TIME_USEC/SI_USEC_PER_JIFFY)
#define SI_SHORT_TIMEOUT_USEC  250 /* .25ms when the SM request a
				      short timeout */

enum si_intf_state {
	SI_NORMAL,
	SI_GETTING_FLAGS,
	SI_GETTING_EVENTS,
	SI_CLEARING_FLAGS,
	SI_CLEARING_FLAGS_THEN_SET_IRQ,
	SI_GETTING_MESSAGES,
	SI_ENABLE_INTERRUPTS1,
	SI_ENABLE_INTERRUPTS2,
	SI_DISABLE_INTERRUPTS1,
	SI_DISABLE_INTERRUPTS2
	/* 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" };

static char *ipmi_addr_src_to_str[] = { NULL, "hotmod", "hardcoded", "SPMI",
					"ACPI", "SMBIOS", "PCI",
					"device-tree", "default" };

#define DEVICE_NAME "ipmi_si"

static struct platform_driver ipmi_driver;

/*
 * Indexes into stats[] in smi_info below.
 */
enum si_stat_indexes {
	/*
	 * Number of times the driver requested a timer while an operation
	 * was in progress.
	 */
	SI_STAT_short_timeouts = 0,

	/*
	 * Number of times the driver requested a timer while nothing was in
	 * progress.
	 */
	SI_STAT_long_timeouts,

	/* Number of times the interface was idle while being polled. */
	SI_STAT_idles,

	/* Number of interrupts the driver handled. */
	SI_STAT_interrupts,

	/* Number of time the driver got an ATTN from the hardware. */
	SI_STAT_attentions,

	/* Number of times the driver requested flags from the hardware. */
	SI_STAT_flag_fetches,

	/* Number of times the hardware didn't follow the state machine. */
	SI_STAT_hosed_count,

	/* Number of completed messages. */
	SI_STAT_complete_transactions,

	/* Number of IPMI events received from the hardware. */
	SI_STAT_events,

	/* Number of watchdog pretimeouts. */
	SI_STAT_watchdog_pretimeouts,

	/* Number of asyncronous messages received. */
	SI_STAT_incoming_messages,


	/* This *must* remain last, add new values above this. */
	SI_NUM_STATS
};

struct smi_info {
	int                    intf_num;
	ipmi_smi_t             intf;
	struct si_sm_data      *si_sm;
	struct si_sm_handlers  *handlers;
	enum si_type           si_type;
	spinlock_t             si_lock;
	spinlock_t             msg_lock;
	struct list_head       xmit_msgs;
	struct list_head       hp_xmit_msgs;
	struct ipmi_smi_msg    *curr_msg;
	enum si_intf_state     si_state;

	/*
	 * Used to handle the various types of I/O that can occur with
	 * IPMI
	 */
	struct si_sm_io io;
	int (*io_setup)(struct smi_info *info);
	void (*io_cleanup)(struct smi_info *info);
	int (*irq_setup)(struct smi_info *info);
	void (*irq_cleanup)(struct smi_info *info);
	unsigned int io_size;
	enum ipmi_addr_src addr_source; /* ACPI, PCI, SMBIOS, hardcode, etc. */
	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
	 * from the flags.
	 */
#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;

	/* Does the BMC have an event buffer? */
	char		    has_event_buffer;

	/*
	 * If set to true, this will request events the next time the
	 * state machine is idle.
	 */
	atomic_t            req_events;

	/*
	 * If true, run the state machine to completion on every send
	 * call.  Generally used after a panic to make sure stuff goes
	 * out.
	 */
	int                 run_to_completion;

	/* 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;

	/* The timer for this si. */
	struct timer_list   si_timer;

	/* The time (in jiffies) the last timeout occurred at. */
	unsigned long       last_timeout_jiffies;

	/* Used to gracefully stop the timer without race conditions. */
	atomic_t            stop_operation;

	/*
	 * The driver will disable interrupts when it gets into a
	 * situation where it cannot handle messages due to lack of
	 * memory.  Once that situation clears up, it will re-enable
	 * interrupts.
	 */
	int interrupt_disabled;

	/* 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. */
	atomic_t stats[SI_NUM_STATS];

	struct task_struct *thread;

	struct list_head link;
	union ipmi_smi_info_union addr_info;
};

#define smi_inc_stat(smi, stat) \
	atomic_inc(&(smi)->stats[SI_STAT_ ## stat])
#define smi_get_stat(smi, stat) \
	((unsigned int) atomic_read(&(smi)->stats[SI_STAT_ ## stat]))

#define SI_MAX_PARMS 4

static int force_kipmid[SI_MAX_PARMS];
static int num_force_kipmid;
#ifdef CONFIG_PCI
static int pci_registered;
#endif
#ifdef CONFIG_ACPI
static int pnp_registered;
#endif

static unsigned int kipmid_max_busy_us[SI_MAX_PARMS];
static int num_max_busy_us;

static int unload_when_empty = 1;

static int add_smi(struct smi_info *smi);
static int try_smi_init(struct smi_info *smi);
static void cleanup_one_si(struct smi_info *to_clean);
static void cleanup_ipmi_si(void);

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)
{
	/* Deliver the message to the upper layer with the lock
	   released. */

	if (smi_info->run_to_completion) {
		ipmi_smi_msg_received(smi_info->intf, msg);
	} else {
		spin_unlock(&(smi_info->si_lock));
		ipmi_smi_msg_received(smi_info->intf, msg);
		spin_lock(&(smi_info->si_lock));
	}
}

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 response */
	msg->rsp[0] = msg->data[0] | 4;
	msg->rsp[1] = msg->data[1];
	msg->rsp[2] = cCode;
	msg->rsp_size = 3;

	smi_info->curr_msg = NULL;
	deliver_recv_msg(smi_info, msg);
}

static enum si_sm_result start_next_msg(struct smi_info *smi_info)
{
	int              rv;
	struct list_head *entry = NULL;
#ifdef DEBUG_TIMING
	struct timeval t;
#endif

	/*
	 * No need to save flags, we aleady have interrupts off and we
	 * already hold the SMI lock.
	 */
	if (!smi_info->run_to_completion)
		spin_lock(&(smi_info->msg_lock));

	/* Pick the high priority queue first. */
	if (!list_empty(&(smi_info->hp_xmit_msgs))) {
		entry = smi_info->hp_xmit_msgs.next;
	} else if (!list_empty(&(smi_info->xmit_msgs))) {
		entry = smi_info->xmit_msgs.next;
	}

	if (!entry) {
		smi_info->curr_msg = NULL;
		rv = SI_SM_IDLE;
	} else {
		int err;

		list_del(entry);
		smi_info->curr_msg = list_entry(entry,
						struct ipmi_smi_msg,
						link);
#ifdef DEBUG_TIMING
		do_gettimeofday(&t);
		printk(KERN_DEBUG "**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, err);

		rv = SI_SM_CALL_WITHOUT_DELAY;
	}
 out:
	if (!smi_info->run_to_completion)
		spin_unlock(&(smi_info->msg_lock));

	return rv;
}

static void start_enable_irq(struct smi_info *smi_info)
{
	unsigned char msg[2];

	/*
	 * If we are enabling interrupts, we have to tell the
	 * BMC to use them.
	 */
	msg[0] = (IPMI_NETFN_APP_REQUEST << 2);
	msg[1] = IPMI_GET_BMC_GLOBAL_ENABLES_CMD;

	smi_info->handlers->start_transaction(smi_info->si_sm, msg, 2);
	smi_info->si_state = SI_ENABLE_INTERRUPTS1;
}

static void start_disable_irq(struct smi_info *smi_info)
{
	unsigned char msg[2];

	msg[0] = (IPMI_NETFN_APP_REQUEST << 2);
	msg[1] = IPMI_GET_BMC_GLOBAL_ENABLES_CMD;

	smi_info->handlers->start_transaction(smi_info->si_sm, msg, 2);
	smi_info->si_state = SI_DISABLE_INTERRUPTS1;
}

static void start_clear_flags(struct smi_info *smi_info)
{
	unsigned char msg[3];

	/* Make sure the watchdog pre-timeout flag is not set at startup. */
	msg[0] = (IPMI_NETFN_APP_REQUEST << 2);
	msg[1] = IPMI_CLEAR_MSG_FLAGS_CMD;
	msg[2] = WDT_PRE_TIMEOUT_INT;

	smi_info->handlers->start_transaction(smi_info->si_sm, msg, 3);
	smi_info->si_state = SI_CLEARING_FLAGS;
}

/*
 * When we have a situtaion where we run out of memory and cannot
 * allocate messages, we just leave them in the BMC and run the system
 * polled until we can allocate some memory.  Once we have some
 * memory, we will re-enable the interrupt.
 */
static inline void disable_si_irq(struct smi_info *smi_info)
{
	if ((smi_info->irq) && (!smi_info->interrupt_disabled)) {
		start_disable_irq(smi_info);
		smi_info->interrupt_disabled = 1;
		if (!atomic_read(&smi_info->stop_operation))
			mod_timer(&smi_info->si_timer,
				  jiffies + SI_TIMEOUT_JIFFIES);
	}
}

static inline void enable_si_irq(struct smi_info *smi_info)
{
	if ((smi_info->irq) && (smi_info->interrupt_disabled)) {
		start_enable_irq(smi_info);
		smi_info->interrupt_disabled = 0;
	}
}

static void handle_flags(struct smi_info *smi_info)
{
 retry:
	if (smi_info->msg_flags & WDT_PRE_TIMEOUT_INT) {
		/* Watchdog pre-timeout */
		smi_inc_stat(smi_info, watchdog_pretimeouts);

		start_clear_flags(smi_info);
		smi_info->msg_flags &= ~WDT_PRE_TIMEOUT_INT;
		spin_unlock(&(smi_info->si_lock));
		ipmi_smi_watchdog_pretimeout(smi_info->intf);
		spin_lock(&(smi_info->si_lock));
	} else if (smi_info->msg_flags & RECEIVE_MSG_AVAIL) {
		/* Messages available. */
		smi_info->curr_msg = ipmi_alloc_smi_msg();
		if (!smi_info->curr_msg) {
			disable_si_irq(smi_info);
			smi_info->si_state = SI_NORMAL;
			return;
		}
		enable_si_irq(smi_info);

		smi_info->curr_msg->data[0] = (IPMI_NETFN_APP_REQUEST << 2);
		smi_info->curr_msg->data[1] = IPMI_GET_MSG_CMD;
		smi_info->curr_msg->data_size = 2;

		smi_info->handlers->start_transaction(
			smi_info->si_sm,
			smi_info->curr_msg->data,
			smi_info->curr_msg->data_size);
		smi_info->si_state = SI_GETTING_MESSAGES;
	} else if (smi_info->msg_flags & EVENT_MSG_BUFFER_FULL) {
		/* Events available. */
		smi_info->curr_msg = ipmi_alloc_smi_msg();
		if (!smi_info->curr_msg) {
			disable_si_irq(smi_info);
			smi_info->si_state = SI_NORMAL;
			return;
		}
		enable_si_irq(smi_info);

		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,
			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;
}

static void handle_transaction_done(struct smi_info *smi_info)
{
	struct ipmi_smi_msg *msg;
#ifdef DEBUG_TIMING
	struct timeval t;

	do_gettimeofday(&t);
	printk(KERN_DEBUG "**Done: %d.%9.9d\n", t.tv_sec, t.tv_usec);
#endif
	switch (smi_info->si_state) {
	case SI_NORMAL:
		if (!smi_info->curr_msg)
			break;

		smi_info->curr_msg->rsp_size
			= smi_info->handlers->get_result(
				smi_info->si_sm,
				smi_info->curr_msg->rsp,
				IPMI_MAX_MSG_LENGTH);

		/*
		 * Do this here becase deliver_recv_msg() releases the
		 * lock, and a new message can be put in during the
		 * time the lock is released.
		 */
		msg = smi_info->curr_msg;
		smi_info->curr_msg = NULL;
		deliver_recv_msg(smi_info, msg);
		break;

	case SI_GETTING_FLAGS:
	{
		unsigned char msg[4];
		unsigned int  len;

		/* We got the flags from the SMI, now handle them. */
		len = smi_info->handlers->get_result(smi_info->si_sm, msg, 4);
		if (msg[2] != 0) {
			/* Error fetching flags, just give up for now. */
			smi_info->si_state = SI_NORMAL;
		} else if (len < 4) {
			/*
			 * Hmm, no flags.  That's technically illegal, but
			 * don't use uninitialized data.
			 */
			smi_info->si_state = SI_NORMAL;
		} else {
			smi_info->msg_flags = msg[3];
			handle_flags(smi_info);
		}
		break;
	}

	case SI_CLEARING_FLAGS:
	case SI_CLEARING_FLAGS_THEN_SET_IRQ:
	{
		unsigned char msg[3];

		/* We cleared the flags. */
		smi_info->handlers->get_result(smi_info->si_sm, msg, 3);
		if (msg[2] != 0) {
			/* Error clearing flags */
			dev_warn(smi_info->dev,
				 "Error clearing flags: %2.2x\n", msg[2]);
		}
		if (smi_info->si_state == SI_CLEARING_FLAGS_THEN_SET_IRQ)
			start_enable_irq(smi_info);
		else
			smi_info->si_state = SI_NORMAL;
		break;
	}

	case SI_GETTING_EVENTS:
	{
		smi_info->curr_msg->rsp_size
			= smi_info->handlers->get_result(
				smi_info->si_sm,
				smi_info->curr_msg->rsp,
				IPMI_MAX_MSG_LENGTH);

		/*
		 * Do this here becase deliver_recv_msg() releases the
		 * lock, and a new message can be put in during the
		 * time the lock is released.
		 */
		msg = smi_info->curr_msg;
		smi_info->curr_msg = NULL;
		if (msg->rsp[2] != 0) {
			/* Error getting event, probably done. */
			msg->done(msg);

			/* Take off the event flag. */
			smi_info->msg_flags &= ~EVENT_MSG_BUFFER_FULL;
			handle_flags(smi_info);
		} else {
			smi_inc_stat(smi_info, events);

			/*
			 * 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);
		}
		break;
	}

	case SI_GETTING_MESSAGES:
	{
		smi_info->curr_msg->rsp_size
			= smi_info->handlers->get_result(
				smi_info->si_sm,
				smi_info->curr_msg->rsp,
				IPMI_MAX_MSG_LENGTH);

		/*
		 * Do this here becase deliver_recv_msg() releases the
		 * lock, and a new message can be put in during the
		 * time the lock is released.
		 */
		msg = smi_info->curr_msg;
		smi_info->curr_msg = NULL;
		if (msg->rsp[2] != 0) {
			/* Error getting event, probably done. */
			msg->done(msg);

			/* Take off the msg flag. */
			smi_info->msg_flags &= ~RECEIVE_MSG_AVAIL;
			handle_flags(smi_info);
		} else {
			smi_inc_stat(smi_info, incoming_messages);

			/*
			 * 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);
		}
		break;
	}

	case SI_ENABLE_INTERRUPTS1:
	{
		unsigned char msg[4];

		/* We got the flags from the SMI, now handle them. */
		smi_info->handlers->get_result(smi_info->si_sm, msg, 4);
		if (msg[2] != 0) {
			dev_warn(smi_info->dev, "Could not enable interrupts"
				 ", failed get, using polled mode.\n");
			smi_info->si_state = SI_NORMAL;
		} else {
			msg[0] = (IPMI_NETFN_APP_REQUEST << 2);
			msg[1] = IPMI_SET_BMC_GLOBAL_ENABLES_CMD;
			msg[2] = (msg[3] |
				  IPMI_BMC_RCV_MSG_INTR |
				  IPMI_BMC_EVT_MSG_INTR);
			smi_info->handlers->start_transaction(
				smi_info->si_sm, msg, 3);
			smi_info->si_state = SI_ENABLE_INTERRUPTS2;
		}
		break;
	}

	case SI_ENABLE_INTERRUPTS2:
	{
		unsigned char msg[4];

		/* We got the flags from the SMI, now handle them. */
		smi_info->handlers->get_result(smi_info->si_sm, msg, 4);
		if (msg[2] != 0)
			dev_warn(smi_info->dev, "Could not enable interrupts"
				 ", failed set, using polled mode.\n");
		else
			smi_info->interrupt_disabled = 0;
		smi_info->si_state = SI_NORMAL;
		break;
	}

	case SI_DISABLE_INTERRUPTS1:
	{
		unsigned char msg[4];

		/* We got the flags from the SMI, now handle them. */
		smi_info->handlers->get_result(smi_info->si_sm, msg, 4);
		if (msg[2] != 0) {
			dev_warn(smi_info->dev, "Could not disable interrupts"
				 ", failed get.\n");
			smi_info->si_state = SI_NORMAL;
		} else {
			msg[0] = (IPMI_NETFN_APP_REQUEST << 2);
			msg[1] = IPMI_SET_BMC_GLOBAL_ENABLES_CMD;
			msg[2] = (msg[3] &
				  ~(IPMI_BMC_RCV_MSG_INTR |
				    IPMI_BMC_EVT_MSG_INTR));
			smi_info->handlers->start_transaction(
				smi_info->si_sm, msg, 3);
			smi_info->si_state = SI_DISABLE_INTERRUPTS2;
		}
		break;
	}

	case SI_DISABLE_INTERRUPTS2:
	{
		unsigned char msg[4];

		/* We got the flags from the SMI, now handle them. */
		smi_info->handlers->get_result(smi_info->si_sm, msg, 4);
		if (msg[2] != 0) {
			dev_warn(smi_info->dev, "Could not disable interrupts"
				 ", failed set.\n");
		}
		smi_info->si_state = SI_NORMAL;
		break;
	}
	}
}

/*
 * Called on timeouts and events.  Timeouts should pass the elapsed
 * time, interrupts should pass in zero.  Must be called with
 * si_lock held and interrupts disabled.
 */
static enum si_sm_result smi_event_handler(struct smi_info *smi_info,
					   int time)
{
	enum si_sm_result si_sm_result;

 restart:
	/*
	 * There used to be a loop here that waited a little while
	 * (around 25us) before giving up.  That turned out to be
	 * pointless, the minimum delays I was seeing were in the 300us
	 * range, which is far too long to wait in an interrupt.  So
	 * we just run until the state machine tells us something
	 * happened or it needs a delay.
	 */
	si_sm_result = smi_info->handlers->event(smi_info->si_sm, time);
	time = 0;
	while (si_sm_result == SI_SM_CALL_WITHOUT_DELAY)
		si_sm_result = smi_info->handlers->event(smi_info->si_sm, 0);

	if (si_sm_result == SI_SM_TRANSACTION_COMPLETE) {
		smi_inc_stat(smi_info, complete_transactions);

		handle_transaction_done(smi_info);
		si_sm_result = smi_info->handlers->event(smi_info->si_sm, 0);
	} else if (si_sm_result == SI_SM_HOSED) {
		smi_inc_stat(smi_info, hosed_count);

		/*
		 * 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, IPMI_ERR_UNSPECIFIED);
		}
		si_sm_result = smi_info->handlers->event(smi_info->si_sm, 0);
	}

	/*
	 * We prefer handling attn over new messages.  But don't do
	 * this if there is not yet an upper layer to handle anything.
	 */
	if (likely(smi_info->intf) && si_sm_result == SI_SM_ATTN) {
		unsigned char msg[2];

		smi_inc_stat(smi_info, attentions);

		/*
		 * Got a attn, send down a get message flags to see
		 * what's causing it.  It would be better to handle
		 * this in the upper layer, but due to the way
		 * interrupts work with the SMI, that's not really
		 * possible.
		 */
		msg[0] = (IPMI_NETFN_APP_REQUEST << 2);
		msg[1] = IPMI_GET_MSG_FLAGS_CMD;

		smi_info->handlers->start_transaction(
			smi_info->si_sm, msg, 2);
		smi_info->si_state = SI_GETTING_FLAGS;
		goto restart;
	}

	/* If we are currently idle, try to start the next message. */
	if (si_sm_result == SI_SM_IDLE) {
		smi_inc_stat(smi_info, idles);

		si_sm_result = start_next_msg(smi_info);
		if (si_sm_result != SI_SM_IDLE)
			goto restart;
	}

	if ((si_sm_result == SI_SM_IDLE)
	    && (atomic_read(&smi_info->req_events))) {
		/*
		 * We are idle and the upper layer requested that I fetch
		 * events, so do so.
		 */
		atomic_set(&smi_info->req_events, 0);

		smi_info->curr_msg = ipmi_alloc_smi_msg();
		if (!smi_info->curr_msg)
			goto out;

		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,
			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;
}

static void sender(void                *send_info,
		   struct ipmi_smi_msg *msg,
		   int                 priority)
{
	struct smi_info   *smi_info = send_info;
	enum si_sm_result result;
	unsigned long     flags;
#ifdef DEBUG_TIMING
	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;
	}

#ifdef DEBUG_TIMING
	do_gettimeofday(&t);
	printk("**Enqueue: %d.%9.9d\n", t.tv_sec, t.tv_usec);
#endif

	/*
	 * last_timeout_jiffies is updated here to avoid
	 * smi_timeout() handler passing very large time_diff
	 * value to smi_event_handler() that causes
	 * the send command to abort.
	 */
	smi_info->last_timeout_jiffies = jiffies;

	mod_timer(&smi_info->si_timer, jiffies + SI_TIMEOUT_JIFFIES);

	if (smi_info->thread)
		wake_up_process(smi_info->thread);

	if (smi_info->run_to_completion) {
		/*
		 * If we are running to completion, then throw it in
		 * the list and run transactions until everything is
		 * clear.  Priority doesn't matter here.
		 */

		/*
		 * Run to completion means we are single-threaded, no
		 * need for locks.
		 */
		list_add_tail(&(msg->link), &(smi_info->xmit_msgs));

		result = smi_event_handler(smi_info, 0);
		while (result != SI_SM_IDLE) {
			udelay(SI_SHORT_TIMEOUT_USEC);
			result = smi_event_handler(smi_info,
						   SI_SHORT_TIMEOUT_USEC);
		}
		return;
	}

	spin_lock_irqsave(&smi_info->msg_lock, flags);
	if (priority > 0)
		list_add_tail(&msg->link, &smi_info->hp_xmit_msgs);
	else
		list_add_tail(&msg->link, &smi_info->xmit_msgs);
	spin_unlock_irqrestore(&smi_info->msg_lock, flags);

	spin_lock_irqsave(&smi_info->si_lock, flags);
	if (smi_info->si_state == SI_NORMAL && smi_info->curr_msg == NULL)
		start_next_msg(smi_info);
	spin_unlock_irqrestore(&smi_info->si_lock, flags);
}

static void set_run_to_completion(void *send_info, int i_run_to_completion)
{
	struct smi_info   *smi_info = send_info;
	enum si_sm_result result;

	smi_info->run_to_completion = i_run_to_completion;
	if (i_run_to_completion) {
		result = smi_event_handler(smi_info, 0);
		while (result != SI_SM_IDLE) {
			udelay(SI_SHORT_TIMEOUT_USEC);
			result = smi_event_handler(smi_info,
						   SI_SHORT_TIMEOUT_USEC);
		}
	}
}

/*
 * Use -1 in the nsec value of the busy waiting timespec to tell that
 * we are spinning in kipmid looking for something and not delaying
 * between checks
 */
static inline void ipmi_si_set_not_busy(struct timespec *ts)
{
	ts->tv_nsec = -1;
}
static inline int ipmi_si_is_busy(struct timespec *ts)
{
	return ts->tv_nsec != -1;
}

static int ipmi_thread_busy_wait(enum si_sm_result smi_result,
				 const struct smi_info *smi_info,
				 struct timespec *busy_until)
{
	unsigned int max_busy_us = 0;

	if (smi_info->intf_num < num_max_busy_us)
		max_busy_us = kipmid_max_busy_us[smi_info->intf_num];
	if (max_busy_us == 0 || smi_result != SI_SM_CALL_WITH_DELAY)
		ipmi_si_set_not_busy(busy_until);
	else if (!ipmi_si_is_busy(busy_until)) {
		getnstimeofday(busy_until);
		timespec_add_ns(busy_until, max_busy_us*NSEC_PER_USEC);
	} else {
		struct timespec now;
		getnstimeofday(&now);
		if (unlikely(timespec_compare(&now, busy_until) > 0)) {
			ipmi_si_set_not_busy(busy_until);
			return 0;
		}
	}
	return 1;
}


/*
 * A busy-waiting loop for speeding up IPMI operation.
 *
 * Lousy hardware makes this hard.  This is only enabled for systems
 * that are not BT and do not have interrupts.  It starts spinning
 * when an operation is complete or until max_busy tells it to stop
 * (if that is enabled).  See the paragraph on kimid_max_busy_us in
 * Documentation/IPMI.txt for details.
 */
static int ipmi_thread(void *data)
{
	struct smi_info *smi_info = data;
	unsigned long flags;
	enum si_sm_result smi_result;
	struct timespec busy_until;

	ipmi_si_set_not_busy(&busy_until);
	set_user_nice(current, 19);
	while (!kthread_should_stop()) {
		int busy_wait;

		spin_lock_irqsave(&(smi_info->si_lock), flags);
		smi_result = smi_event_handler(smi_info, 0);
		spin_unlock_irqrestore(&(smi_info->si_lock), flags);
		busy_wait = ipmi_thread_busy_wait(smi_result, smi_info,
						  &busy_until);
		if (smi_result == SI_SM_CALL_WITHOUT_DELAY)
			; /* do nothing */
		else if (smi_result == SI_SM_CALL_WITH_DELAY && busy_wait)
			schedule();
		else if (smi_result == SI_SM_IDLE)
			schedule_timeout_interruptible(100);
		else
			schedule_timeout_interruptible(1);
	}
	return 0;
}


static void poll(void *send_info)
{
	struct smi_info *smi_info = send_info;
	unsigned long flags;

	/*
	 * Make sure there is some delay in the poll loop so we can
	 * drive time forward and timeout things.
	 */
	udelay(10);
	spin_lock_irqsave(&smi_info->si_lock, flags);
	smi_event_handler(smi_info, 10);
	spin_unlock_irqrestore(&smi_info->si_lock, flags);
}

static void request_events(void *send_info)
{
	struct smi_info *smi_info = send_info;

	if (atomic_read(&smi_info->stop_operation) ||
				!smi_info->has_event_buffer)
		return;

	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;
	long              time_diff;
	long		  timeout;
#ifdef DEBUG_TIMING
	struct timeval    t;
#endif

	spin_lock_irqsave(&(smi_info->si_lock), flags);
#ifdef DEBUG_TIMING
	do_gettimeofday(&t);
	printk(KERN_DEBUG "**Timer: %d.%9.9d\n", t.tv_sec, t.tv_usec);
#endif
	jiffies_now = 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);

	spin_unlock_irqrestore(&(smi_info->si_lock), flags);

	smi_info->last_timeout_jiffies = jiffies_now;

	if ((smi_info->irq) && (!smi_info->interrupt_disabled)) {
		/* Running with interrupts, only do long timeouts. */
		timeout = jiffies + SI_TIMEOUT_JIFFIES;
		smi_inc_stat(smi_info, long_timeouts);
		goto do_mod_timer;
	}

	/*
	 * If the state machine asks for a short delay, then shorten
	 * the timer timeout.
	 */
	if (smi_result == SI_SM_CALL_WITH_DELAY) {
		smi_inc_stat(smi_info, short_timeouts);
		timeout = jiffies + 1;
	} else {
		smi_inc_stat(smi_info, long_timeouts);
		timeout = jiffies + SI_TIMEOUT_JIFFIES;
	}

 do_mod_timer:
	if (smi_result != SI_SM_IDLE)
		mod_timer(&(smi_info->si_timer), timeout);
}

static irqreturn_t si_irq_handler(int irq, void *data)
{
	struct smi_info *smi_info = data;
	unsigned long   flags;
#ifdef DEBUG_TIMING
	struct timeval  t;
#endif

	spin_lock_irqsave(&(smi_info->si_lock), flags);

	smi_inc_stat(smi_info, interrupts);

#ifdef DEBUG_TIMING
	do_gettimeofday(&t);
	printk(KERN_DEBUG "**Interrupt: %d.%9.9d\n", t.tv_sec, t.tv_usec);
#endif
	smi_event_handler(smi_info, 0);
	spin_unlock_irqrestore(&(smi_info->si_lock), flags);
	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;

	/* Try to claim any interrupts. */
	if (new_smi->irq_setup)
		new_smi->irq_setup(new_smi);

	/* 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)) {
			dev_notice(new_smi->dev, "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 int get_smi_info(void *send_info, struct ipmi_smi_info *data)
{
	struct smi_info *smi = send_info;

	data->addr_src = smi->addr_source;
	data->dev = smi->dev;
	data->addr_info = smi->addr_info;
	get_device(smi->dev);

	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,
	.get_smi_info		= get_smi_info,
	.sender			= sender,
	.request_events		= request_events,
	.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.
 */

static LIST_HEAD(smi_infos);
static DEFINE_MUTEX(smi_infos_lock);
static int smi_num; /* Used to sequence the SMIs */

#define DEFAULT_REGSPACING	1
#define DEFAULT_REGSIZE		1

static int           si_trydefaults = 1;
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];
static unsigned int num_addrs;
static unsigned int  ports[SI_MAX_PARMS];
static unsigned int num_ports;
static int           irqs[SI_MAX_PARMS];
static unsigned int num_irqs;
static int           regspacings[SI_MAX_PARMS];
static unsigned int num_regspacings;
static int           regsizes[SI_MAX_PARMS];
static unsigned int num_regsizes;
static int           regshifts[SI_MAX_PARMS];
static unsigned int num_regshifts;
static int slave_addrs[SI_MAX_PARMS]; /* Leaving 0 chooses the default value */
static unsigned 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"
		 " default scan of the KCS and SMIC interface at the standard"
		 " address");
module_param_string(type, si_type_str, MAX_SI_TYPE_STR, 0);
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, ulong, &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, uint, &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_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.");
module_param_array(kipmid_max_busy_us, uint, &num_max_busy_us, 0644);
MODULE_PARM_DESC(kipmid_max_busy_us,
		 "Max time (in microseconds) to busy-wait for IPMI data before"
		 " sleeping. 0 (default) means to wait forever. Set to 100-500"
		 " if kipmid is using up a lot of CPU time.");


static void std_irq_cleanup(struct smi_info *info)
{
	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);
}

static int std_irq_setup(struct smi_info *info)
{
	int rv;

	if (!info->irq)
		return 0;

	if (info->si_type == SI_BT) {
		rv = request_irq(info->irq,
				 si_bt_irq_handler,
				 IRQF_SHARED | 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_SHARED | IRQF_DISABLED,
				 DEVICE_NAME,
				 info);
	if (rv) {
		dev_warn(info->dev, "%s unable to claim interrupt %d,"
			 " running polled\n",
			 DEVICE_NAME, info->irq);
		info->irq = 0;
	} else {
		info->irq_cleanup = std_irq_cleanup;
		dev_info(info->dev, "Using irq %d\n", info->irq);
	}

	return rv;
}

static unsigned char port_inb(struct si_sm_io *io, unsigned int offset)
{
	unsigned int addr = io->addr_data;

	return inb(addr + (offset * io->regspacing));
}

static void port_outb(struct si_sm_io *io, unsigned int offset,
		      unsigned char b)
{
	unsigned int addr = io->addr_data;

	outb(b, addr + (offset * io->regspacing));
}

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->addr_data;

	outw(b << io->regshift, addr + (offset * io->regspacing));
}

static unsigned char port_inl(struct si_sm_io *io, unsigned int offset)
{
	unsigned int addr = io->addr_data;

	return (inl(addr + (offset * io->regspacing)) >> io->regshift) & 0xff;
}

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.addr_data;
	int          idx;

	if (addr) {
		for (idx = 0; idx < info->io_size; idx++)
			release_region(addr + idx * info->io.regspacing,
				       info->io.regsize);
	}
}

static int port_setup(struct smi_info *info)
{
	unsigned int addr = info->io.addr_data;
	int          idx;

	if (!addr)
		return -ENODEV;

	info->io_cleanup = port_cleanup;

	/*
	 * 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:
		dev_warn(info->dev, "Invalid register size: %d\n",
			 info->io.regsize);
		return -EINVAL;
	}

	/*
	 * 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 intf_mem_inb(struct si_sm_io *io, unsigned int offset)
{
	return readb((io->addr)+(offset * io->regspacing));
}

static void intf_mem_outb(struct si_sm_io *io, unsigned int offset,
		     unsigned char b)
{
	writeb(b, (io->addr)+(offset * io->regspacing));
}

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)
{
	writeb(b << io->regshift, (io->addr)+(offset * io->regspacing));
}

static unsigned char intf_mem_inl(struct si_sm_io *io, unsigned int offset)
{
	return (readl((io->addr)+(offset * io->regspacing)) >> io->regshift)
		& 0xff;
}

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));
}

#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.addr_data;
	int           mapsize;

	if (info->io.addr) {
		iounmap(info->io.addr);

		mapsize = ((info->io_size * info->io.regspacing)
			   - (info->io.regspacing - info->io.regsize));

		release_mem_region(addr, mapsize);
	}
}

static int mem_setup(struct smi_info *info)
{
	unsigned long addr = info->io.addr_data;
	int           mapsize;

	if (!addr)
		return -ENODEV;

	info->io_cleanup = mem_cleanup;

	/*
	 * 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:
		dev_warn(info->dev, "Invalid register size: %d\n",
			 info->io.regsize);
		return -EINVAL;
	}

	/*
	 * 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 }
};

static int parse_str(struct hotmod_vals *v, int *val, char *name, char **curr)
{
	char *s;
	int  i;

	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;
		}
	}

	printk(KERN_WARNING PFX "Invalid hotmod %s '%s'\n", name, *curr);
	return -EINVAL;
}

static int check_hotmod_int_op(const char *curr, const char *option,
			       const char *name, int *val)
{
	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;
}

static struct smi_info *smi_info_alloc(void)
{
	struct smi_info *info = kzalloc(sizeof(*info), GFP_KERNEL);

	if (info) {
		spin_lock_init(&info->si_lock);
		spin_lock_init(&info->msg_lock);
	}
	return info;
}

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;

	if (!str)
		return -ENOMEM;

	/* 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 = 0; /* Choose the default if not specified */

		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 = smi_info_alloc();
			if (!info) {
				rv = -ENOMEM;
				goto out;
			}

			info->addr_source = SI_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;

			if (!add_smi(info)) {
				if (try_smi_init(info))
					cleanup_one_si(info);
			} else {
				kfree(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 int __devinit hardcode_find_bmc(void)
{
	int ret = -ENODEV;
	int             i;
	struct smi_info *info;

	for (i = 0; i < SI_MAX_PARMS; i++) {
		if (!ports[i] && !addrs[i])
			continue;

		info = smi_info_alloc();
		if (!info)
			return -ENOMEM;

		info->addr_source = SI_HARDCODED;
		printk(KERN_INFO PFX "probing via hardcoded address\n");

		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 PFX "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;
		} e…