[IA64-SGI] SGI Altix cross partition functionality [2/3]

This patch contains the communication module (XPC) for cross partition
communication on a partitioned SGI Altix.

Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

1 files changed, 1064 insertions, 0 deletions

diff --git a/arch/ia64/sn/kernel/xpc_main.c b/arch/ia64/sn/kernel/xpc_main.c
new file mode 100644
index 00000000000..177ddb748eb
--- /dev/null
+++ b/arch/ia64/sn/kernel/xpc_main.c
@@ -0,0 +1,1064 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (c) 2004-2005 Silicon Graphics, Inc.  All Rights Reserved.
+ */
+
+
+/*
+ * Cross Partition Communication (XPC) support - standard version.
+ *
+ *	XPC provides a message passing capability that crosses partition
+ *	boundaries. This module is made up of two parts:
+ *
+ *	    partition	This part detects the presence/absence of other
+ *			partitions. It provides a heartbeat and monitors
+ *			the heartbeats of other partitions.
+ *
+ *	    channel	This part manages the channels and sends/receives
+ *			messages across them to/from other partitions.
+ *
+ *	There are a couple of additional functions residing in XP, which
+ *	provide an interface to XPC for its users.
+ *
+ *
+ *	Caveats:
+ *
+ *	  . We currently have no way to determine which nasid an IPI came
+ *	    from. Thus, xpc_IPI_send() does a remote AMO write followed by
+ *	    an IPI. The AMO indicates where data is to be pulled from, so
+ *	    after the IPI arrives, the remote partition checks the AMO word.
+ *	    The IPI can actually arrive before the AMO however, so other code
+ *	    must periodically check for this case. Also, remote AMO operations
+ *	    do not reliably time out. Thus we do a remote PIO read solely to
+ *	    know whether the remote partition is down and whether we should
+ *	    stop sending IPIs to it. This remote PIO read operation is set up
+ *	    in a special nofault region so SAL knows to ignore (and cleanup)
+ *	    any errors due to the remote AMO write, PIO read, and/or PIO
+ *	    write operations.
+ *
+ *	    If/when new hardware solves this IPI problem, we should abandon
+ *	    the current approach.
+ *
+ */
+
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/sched.h>
+#include <linux/syscalls.h>
+#include <linux/cache.h>
+#include <linux/interrupt.h>
+#include <linux/slab.h>
+#include <asm/sn/intr.h>
+#include <asm/sn/sn_sal.h>
+#include <asm/uaccess.h>
+#include "xpc.h"
+
+
+/* define two XPC debug device structures to be used with dev_dbg() et al */
+
+struct device_driver xpc_dbg_name = {
+	.name = "xpc"
+};
+
+struct device xpc_part_dbg_subname = {
+	.bus_id = {0},		/* set to "part" at xpc_init() time */
+	.driver = &xpc_dbg_name
+};
+
+struct device xpc_chan_dbg_subname = {
+	.bus_id = {0},		/* set to "chan" at xpc_init() time */
+	.driver = &xpc_dbg_name
+};
+
+struct device *xpc_part = &xpc_part_dbg_subname;
+struct device *xpc_chan = &xpc_chan_dbg_subname;
+
+
+/* systune related variables for /proc/sys directories */
+
+static int xpc_hb_min = 1;
+static int xpc_hb_max = 10;
+
+static int xpc_hb_check_min = 10;
+static int xpc_hb_check_max = 120;
+
+static ctl_table xpc_sys_xpc_hb_dir[] = {
+	{
+		1,
+		"hb_interval",
+		&xpc_hb_interval,
+		sizeof(int),
+		0644,
+		NULL,
+		&proc_dointvec_minmax,
+		&sysctl_intvec,
+		NULL,
+		&xpc_hb_min, &xpc_hb_max
+	},
+	{
+		2,
+		"hb_check_interval",
+		&xpc_hb_check_interval,
+		sizeof(int),
+		0644,
+		NULL,
+		&proc_dointvec_minmax,
+		&sysctl_intvec,
+		NULL,
+		&xpc_hb_check_min, &xpc_hb_check_max
+	},
+	{0}
+};
+static ctl_table xpc_sys_xpc_dir[] = {
+	{
+		1,
+		"hb",
+		NULL,
+		0,
+		0555,
+		xpc_sys_xpc_hb_dir
+	},
+	{0}
+};
+static ctl_table xpc_sys_dir[] = {
+	{
+		1,
+		"xpc",
+		NULL,
+		0,
+		0555,
+		xpc_sys_xpc_dir
+	},
+	{0}
+};
+static struct ctl_table_header *xpc_sysctl;
+
+
+/* #of IRQs received */
+static atomic_t xpc_act_IRQ_rcvd;
+
+/* IRQ handler notifies this wait queue on receipt of an IRQ */
+static DECLARE_WAIT_QUEUE_HEAD(xpc_act_IRQ_wq);
+
+static unsigned long xpc_hb_check_timeout;
+
+/* xpc_hb_checker thread exited notification */
+static DECLARE_MUTEX_LOCKED(xpc_hb_checker_exited);
+
+/* xpc_discovery thread exited notification */
+static DECLARE_MUTEX_LOCKED(xpc_discovery_exited);
+
+
+static struct timer_list xpc_hb_timer;
+
+
+static void xpc_kthread_waitmsgs(struct xpc_partition *, struct xpc_channel *);
+
+
+/*
+ * Notify the heartbeat check thread that an IRQ has been received.
+ */
+static irqreturn_t
+xpc_act_IRQ_handler(int irq, void *dev_id, struct pt_regs *regs)
+{
+	atomic_inc(&xpc_act_IRQ_rcvd);
+	wake_up_interruptible(&xpc_act_IRQ_wq);
+	return IRQ_HANDLED;
+}
+
+
+/*
+ * Timer to produce the heartbeat.  The timer structures function is
+ * already set when this is initially called.  A tunable is used to
+ * specify when the next timeout should occur.
+ */
+static void
+xpc_hb_beater(unsigned long dummy)
+{
+	xpc_vars->heartbeat++;
+
+	if (jiffies >= xpc_hb_check_timeout) {
+		wake_up_interruptible(&xpc_act_IRQ_wq);
+	}
+
+	xpc_hb_timer.expires = jiffies + (xpc_hb_interval * HZ);
+	add_timer(&xpc_hb_timer);
+}
+
+
+/*
+ * This thread is responsible for nearly all of the partition
+ * activation/deactivation.
+ */
+static int
+xpc_hb_checker(void *ignore)
+{
+	int last_IRQ_count = 0;
+	int new_IRQ_count;
+	int force_IRQ=0;
+
+
+	/* this thread was marked active by xpc_hb_init() */
+
+	daemonize(XPC_HB_CHECK_THREAD_NAME);
+
+	set_cpus_allowed(current, cpumask_of_cpu(XPC_HB_CHECK_CPU));
+
+	xpc_hb_check_timeout = jiffies + (xpc_hb_check_interval * HZ);
+
+	while (!(volatile int) xpc_exiting) {
+
+		/* wait for IRQ or timeout */
+		(void) wait_event_interruptible(xpc_act_IRQ_wq,
+			    (last_IRQ_count < atomic_read(&xpc_act_IRQ_rcvd) ||
+					jiffies >= xpc_hb_check_timeout ||
+						(volatile int) xpc_exiting));
+
+		dev_dbg(xpc_part, "woke up with %d ticks rem; %d IRQs have "
+			"been received\n",
+			(int) (xpc_hb_check_timeout - jiffies),
+			atomic_read(&xpc_act_IRQ_rcvd) - last_IRQ_count);
+
+
+		/* checking of remote heartbeats is skewed by IRQ handling */
+		if (jiffies >= xpc_hb_check_timeout) {
+			dev_dbg(xpc_part, "checking remote heartbeats\n");
+			xpc_check_remote_hb();
+
+			/*
+			 * We need to periodically recheck to ensure no
+			 * IPI/AMO pairs have been missed.  That check
+			 * must always reset xpc_hb_check_timeout.
+			 */
+			force_IRQ = 1;
+		}
+
+
+		new_IRQ_count = atomic_read(&xpc_act_IRQ_rcvd);
+		if (last_IRQ_count < new_IRQ_count || force_IRQ != 0) {
+			force_IRQ = 0;
+
+			dev_dbg(xpc_part, "found an IRQ to process; will be "
+				"resetting xpc_hb_check_timeout\n");
+
+			last_IRQ_count += xpc_identify_act_IRQ_sender();
+			if (last_IRQ_count < new_IRQ_count) {
+				/* retry once to help avoid missing AMO */
+				(void) xpc_identify_act_IRQ_sender();
+			}
+			last_IRQ_count = new_IRQ_count;
+
+			xpc_hb_check_timeout = jiffies +
+					   (xpc_hb_check_interval * HZ);
+		}
+	}
+
+	dev_dbg(xpc_part, "heartbeat checker is exiting\n");
+
+
+	/* mark this thread as inactive */
+	up(&xpc_hb_checker_exited);
+	return 0;
+}
+
+
+/*
+ * This thread will attempt to discover other partitions to activate
+ * based on info provided by SAL. This new thread is short lived and
+ * will exit once discovery is complete.
+ */
+static int
+xpc_initiate_discovery(void *ignore)
+{
+	daemonize(XPC_DISCOVERY_THREAD_NAME);
+
+	xpc_discovery();
+
+	dev_dbg(xpc_part, "discovery thread is exiting\n");
+
+	/* mark this thread as inactive */
+	up(&xpc_discovery_exited);
+	return 0;
+}
+
+
+/*
+ * Establish first contact with the remote partititon. This involves pulling
+ * the XPC per partition variables from the remote partition and waiting for
+ * the remote partition to pull ours.
+ */
+static enum xpc_retval
+xpc_make_first_contact(struct xpc_partition *part)
+{
+	enum xpc_retval ret;
+
+
+	while ((ret = xpc_pull_remote_vars_part(part)) != xpcSuccess) {
+		if (ret != xpcRetry) {
+			XPC_DEACTIVATE_PARTITION(part, ret);
+			return ret;
+		}
+
+		dev_dbg(xpc_chan, "waiting to make first contact with "
+			"partition %d\n", XPC_PARTID(part));
+
+		/* wait a 1/4 of a second or so */
+		set_current_state(TASK_INTERRUPTIBLE);
+		(void) schedule_timeout(0.25 * HZ);
+
+		if (part->act_state == XPC_P_DEACTIVATING) {
+			return part->reason;
+		}
+	}
+
+	return xpc_mark_partition_active(part);
+}
+
+
+/*
+ * The first kthread assigned to a newly activated partition is the one
+ * created by XPC HB with which it calls xpc_partition_up(). XPC hangs on to
+ * that kthread until the partition is brought down, at which time that kthread
+ * returns back to XPC HB. (The return of that kthread will signify to XPC HB
+ * that XPC has dismantled all communication infrastructure for the associated
+ * partition.) This kthread becomes the channel manager for that partition.
+ *
+ * Each active partition has a channel manager, who, besides connecting and
+ * disconnecting channels, will ensure that each of the partition's connected
+ * channels has the required number of assigned kthreads to get the work done.
+ */
+static void
+xpc_channel_mgr(struct xpc_partition *part)
+{
+	while (part->act_state != XPC_P_DEACTIVATING ||
+				atomic_read(&part->nchannels_active) > 0) {
+
+		xpc_process_channel_activity(part);
+
+
+		/*
+		 * Wait until we've been requested to activate kthreads or
+		 * all of the channel's message queues have been torn down or
+		 * a signal is pending.
+		 *
+		 * The channel_mgr_requests is set to 1 after being awakened,
+		 * This is done to prevent the channel mgr from making one pass
+		 * through the loop for each request, since he will
+		 * be servicing all the requests in one pass. The reason it's
+		 * set to 1 instead of 0 is so that other kthreads will know
+		 * that the channel mgr is running and won't bother trying to
+		 * wake him up.
+		 */
+		atomic_dec(&part->channel_mgr_requests);
+		(void) wait_event_interruptible(part->channel_mgr_wq,
+				(atomic_read(&part->channel_mgr_requests) > 0 ||
+				(volatile u64) part->local_IPI_amo != 0 ||
+				((volatile u8) part->act_state ==
+							XPC_P_DEACTIVATING &&
+				atomic_read(&part->nchannels_active) == 0)));
+		atomic_set(&part->channel_mgr_requests, 1);
+
+		// >>> Does it need to wakeup periodically as well? In case we
+		// >>> miscalculated the #of kthreads to wakeup or create?
+	}
+}
+
+
+/*
+ * When XPC HB determines that a partition has come up, it will create a new
+ * kthread and that kthread will call this function to attempt to set up the
+ * basic infrastructure used for Cross Partition Communication with the newly
+ * upped partition.
+ *
+ * The kthread that was created by XPC HB and which setup the XPC
+ * infrastructure will remain assigned to the partition until the partition
+ * goes down. At which time the kthread will teardown the XPC infrastructure
+ * and then exit.
+ *
+ * XPC HB will put the remote partition's XPC per partition specific variables
+ * physical address into xpc_partitions[partid].remote_vars_part_pa prior to
+ * calling xpc_partition_up().
+ */
+static void
+xpc_partition_up(struct xpc_partition *part)
+{
+	DBUG_ON(part->channels != NULL);
+
+	dev_dbg(xpc_chan, "activating partition %d\n", XPC_PARTID(part));
+
+	if (xpc_setup_infrastructure(part) != xpcSuccess) {
+		return;
+	}
+
+	/*
+	 * The kthread that XPC HB called us with will become the
+	 * channel manager for this partition. It will not return
+	 * back to XPC HB until the partition's XPC infrastructure
+	 * has been dismantled.
+	 */
+
+	(void) xpc_part_ref(part);	/* this will always succeed */
+
+	if (xpc_make_first_contact(part) == xpcSuccess) {
+		xpc_channel_mgr(part);
+	}
+
+	xpc_part_deref(part);
+
+	xpc_teardown_infrastructure(part);
+}
+
+
+static int
+xpc_activating(void *__partid)
+{
+	partid_t partid = (u64) __partid;
+	struct xpc_partition *part = &xpc_partitions[partid];
+	unsigned long irq_flags;
+	struct sched_param param = { sched_priority: MAX_USER_RT_PRIO - 1 };
+	int ret;
+
+
+	DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS);
+
+	spin_lock_irqsave(&part->act_lock, irq_flags);
+
+	if (part->act_state == XPC_P_DEACTIVATING) {
+		part->act_state = XPC_P_INACTIVE;
+		spin_unlock_irqrestore(&part->act_lock, irq_flags);
+		part->remote_rp_pa = 0;
+		return 0;
+	}
+
+	/* indicate the thread is activating */
+	DBUG_ON(part->act_state != XPC_P_ACTIVATION_REQ);
+	part->act_state = XPC_P_ACTIVATING;
+
+	XPC_SET_REASON(part, 0, 0);
+	spin_unlock_irqrestore(&part->act_lock, irq_flags);
+
+	dev_dbg(xpc_part, "bringing partition %d up\n", partid);
+
+	daemonize("xpc%02d", partid);
+
+	/*
+	 * This thread needs to run at a realtime priority to prevent a
+	 * significant performance degradation.
+	 */
+	ret = sched_setscheduler(current, SCHED_FIFO, &param);
+	if (ret != 0) {
+		dev_warn(xpc_part, "unable to set pid %d to a realtime "
+			"priority, ret=%d\n", current->pid, ret);
+	}
+
+	/* allow this thread and its children to run on any CPU */
+	set_cpus_allowed(current, CPU_MASK_ALL);
+
+	/*
+	 * Register the remote partition's AMOs with SAL so it can handle
+	 * and cleanup errors within that address range should the remote
+	 * partition go down. We don't unregister this range because it is
+	 * difficult to tell when outstanding writes to the remote partition
+	 * are finished and thus when it is safe to unregister. This should
+	 * not result in wasted space in the SAL xp_addr_region table because
+	 * we should get the same page for remote_amos_page_pa after module
+	 * reloads and system reboots.
+	 */
+	if (sn_register_xp_addr_region(part->remote_amos_page_pa,
+							PAGE_SIZE, 1) < 0) {
+		dev_warn(xpc_part, "xpc_partition_up(%d) failed to register "
+			"xp_addr region\n", partid);
+
+		spin_lock_irqsave(&part->act_lock, irq_flags);
+		part->act_state = XPC_P_INACTIVE;
+		XPC_SET_REASON(part, xpcPhysAddrRegFailed, __LINE__);
+		spin_unlock_irqrestore(&part->act_lock, irq_flags);
+		part->remote_rp_pa = 0;
+		return 0;
+	}
+
+	XPC_ALLOW_HB(partid, xpc_vars);
+	xpc_IPI_send_activated(part);
+
+
+	/*
+	 * xpc_partition_up() holds this thread and marks this partition as
+	 * XPC_P_ACTIVE by calling xpc_hb_mark_active().
+	 */
+	(void) xpc_partition_up(part);
+
+	xpc_mark_partition_inactive(part);
+
+	if (part->reason == xpcReactivating) {
+		/* interrupting ourselves results in activating partition */
+		xpc_IPI_send_reactivate(part);
+	}
+
+	return 0;
+}
+
+
+void
+xpc_activate_partition(struct xpc_partition *part)
+{
+	partid_t partid = XPC_PARTID(part);
+	unsigned long irq_flags;
+	pid_t pid;
+
+
+	spin_lock_irqsave(&part->act_lock, irq_flags);
+
+	pid = kernel_thread(xpc_activating, (void *) ((u64) partid), 0);
+
+	DBUG_ON(part->act_state != XPC_P_INACTIVE);
+
+	if (pid > 0) {
+		part->act_state = XPC_P_ACTIVATION_REQ;
+		XPC_SET_REASON(part, xpcCloneKThread, __LINE__);
+	} else {
+		XPC_SET_REASON(part, xpcCloneKThreadFailed, __LINE__);
+	}
+
+	spin_unlock_irqrestore(&part->act_lock, irq_flags);
+}
+
+
+/*
+ * Handle the receipt of a SGI_XPC_NOTIFY IRQ by seeing whether the specified
+ * partition actually sent it. Since SGI_XPC_NOTIFY IRQs may be shared by more
+ * than one partition, we use an AMO_t structure per partition to indicate
+ * whether a partition has sent an IPI or not.  >>> If it has, then wake up the
+ * associated kthread to handle it.
+ *
+ * All SGI_XPC_NOTIFY IRQs received by XPC are the result of IPIs sent by XPC
+ * running on other partitions.
+ *
+ * Noteworthy Arguments:
+ *
+ *	irq - Interrupt ReQuest number. NOT USED.
+ *
+ *	dev_id - partid of IPI's potential sender.
+ *
+ *	regs - processor's context before the processor entered
+ *	       interrupt code. NOT USED.
+ */
+irqreturn_t
+xpc_notify_IRQ_handler(int irq, void *dev_id, struct pt_regs *regs)
+{
+	partid_t partid = (partid_t) (u64) dev_id;
+	struct xpc_partition *part = &xpc_partitions[partid];
+
+
+	DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS);
+
+	if (xpc_part_ref(part)) {
+		xpc_check_for_channel_activity(part);
+
+		xpc_part_deref(part);
+	}
+	return IRQ_HANDLED;
+}
+
+
+/*
+ * Check to see if xpc_notify_IRQ_handler() dropped any IPIs on the floor
+ * because the write to their associated IPI amo completed after the IRQ/IPI
+ * was received.
+ */
+void
+xpc_dropped_IPI_check(struct xpc_partition *part)
+{
+	if (xpc_part_ref(part)) {
+		xpc_check_for_channel_activity(part);
+
+		part->dropped_IPI_timer.expires = jiffies +
+							XPC_P_DROPPED_IPI_WAIT;
+		add_timer(&part->dropped_IPI_timer);
+		xpc_part_deref(part);
+	}
+}
+
+
+void
+xpc_activate_kthreads(struct xpc_channel *ch, int needed)
+{
+	int idle = atomic_read(&ch->kthreads_idle);
+	int assigned = atomic_read(&ch->kthreads_assigned);
+	int wakeup;
+
+
+	DBUG_ON(needed <= 0);
+
+	if (idle > 0) {
+		wakeup = (needed > idle) ? idle : needed;
+		needed -= wakeup;
+
+		dev_dbg(xpc_chan, "wakeup %d idle kthreads, partid=%d, "
+			"channel=%d\n", wakeup, ch->partid, ch->number);
+
+		/* only wakeup the requested number of kthreads */
+		wake_up_nr(&ch->idle_wq, wakeup);
+	}
+
+	if (needed <= 0) {
+		return;
+	}
+
+	if (needed + assigned > ch->kthreads_assigned_limit) {
+		needed = ch->kthreads_assigned_limit - assigned;
+		// >>>should never be less than 0
+		if (needed <= 0) {
+			return;
+		}
+	}
+
+	dev_dbg(xpc_chan, "create %d new kthreads, partid=%d, channel=%d\n",
+		needed, ch->partid, ch->number);
+
+	xpc_create_kthreads(ch, needed);
+}
+
+
+/*
+ * This function is where XPC's kthreads wait for messages to deliver.
+ */
+static void
+xpc_kthread_waitmsgs(struct xpc_partition *part, struct xpc_channel *ch)
+{
+	do {
+		/* deliver messages to their intended recipients */
+
+		while ((volatile s64) ch->w_local_GP.get <
+				(volatile s64) ch->w_remote_GP.put &&
+					!((volatile u32) ch->flags &
+						XPC_C_DISCONNECTING)) {
+			xpc_deliver_msg(ch);
+		}
+
+		if (atomic_inc_return(&ch->kthreads_idle) >
+						ch->kthreads_idle_limit) {
+			/* too many idle kthreads on this channel */
+			atomic_dec(&ch->kthreads_idle);
+			break;
+		}
+
+		dev_dbg(xpc_chan, "idle kthread calling "
+			"wait_event_interruptible_exclusive()\n");
+
+		(void) wait_event_interruptible_exclusive(ch->idle_wq,
+				((volatile s64) ch->w_local_GP.get <
+					(volatile s64) ch->w_remote_GP.put ||
+				((volatile u32) ch->flags &
+						XPC_C_DISCONNECTING)));
+
+		atomic_dec(&ch->kthreads_idle);
+
+	} while (!((volatile u32) ch->flags & XPC_C_DISCONNECTING));
+}
+
+
+static int
+xpc_daemonize_kthread(void *args)
+{
+	partid_t partid = XPC_UNPACK_ARG1(args);
+	u16 ch_number = XPC_UNPACK_ARG2(args);
+	struct xpc_partition *part = &xpc_partitions[partid];
+	struct xpc_channel *ch;
+	int n_needed;
+
+
+	daemonize("xpc%02dc%d", partid, ch_number);
+
+	dev_dbg(xpc_chan, "kthread starting, partid=%d, channel=%d\n",
+		partid, ch_number);
+
+	ch = &part->channels[ch_number];
+
+	if (!(ch->flags & XPC_C_DISCONNECTING)) {
+		DBUG_ON(!(ch->flags & XPC_C_CONNECTED));
+
+		/* let registerer know that connection has been established */
+
+		if (atomic_read(&ch->kthreads_assigned) == 1) {
+			xpc_connected_callout(ch);
+
+			/*
+			 * It is possible that while the callout was being
+			 * made that the remote partition sent some messages.
+			 * If that is the case, we may need to activate
+			 * additional kthreads to help deliver them. We only
+			 * need one less than total #of messages to deliver.
+			 */
+			n_needed = ch->w_remote_GP.put - ch->w_local_GP.get - 1;
+			if (n_needed > 0 &&
+					!(ch->flags & XPC_C_DISCONNECTING)) {
+				xpc_activate_kthreads(ch, n_needed);
+			}
+		}
+
+		xpc_kthread_waitmsgs(part, ch);
+	}
+
+	if (atomic_dec_return(&ch->kthreads_assigned) == 0 &&
+			((ch->flags & XPC_C_CONNECTCALLOUT) ||
+				(ch->reason != xpcUnregistering &&
+					ch->reason != xpcOtherUnregistering))) {
+		xpc_disconnected_callout(ch);
+	}
+
+
+	xpc_msgqueue_deref(ch);
+
+	dev_dbg(xpc_chan, "kthread exiting, partid=%d, channel=%d\n",
+		partid, ch_number);
+
+	xpc_part_deref(part);
+	return 0;
+}
+
+
+/*
+ * For each partition that XPC has established communications with, there is
+ * a minimum of one kernel thread assigned to perform any operation that
+ * may potentially sleep or block (basically the callouts to the asynchronous
+ * functions registered via xpc_connect()).
+ *
+ * Additional kthreads are created and destroyed by XPC as the workload
+ * demands.
+ *
+ * A kthread is assigned to one of the active channels that exists for a given
+ * partition.
+ */
+void
+xpc_create_kthreads(struct xpc_channel *ch, int needed)
+{
+	unsigned long irq_flags;
+	pid_t pid;
+	u64 args = XPC_PACK_ARGS(ch->partid, ch->number);
+
+
+	while (needed-- > 0) {
+		pid = kernel_thread(xpc_daemonize_kthread, (void *) args, 0);
+		if (pid < 0) {
+			/* the fork failed */
+
+			if (atomic_read(&ch->kthreads_assigned) <
+						ch->kthreads_idle_limit) {
+				/*
+				 * Flag this as an error only if we have an
+				 * insufficient #of kthreads for the channel
+				 * to function.
+				 *
+				 * No xpc_msgqueue_ref() is needed here since
+				 * the channel mgr is doing this.
+				 */
+				spin_lock_irqsave(&ch->lock, irq_flags);
+				XPC_DISCONNECT_CHANNEL(ch, xpcLackOfResources,
+								&irq_flags);
+				spin_unlock_irqrestore(&ch->lock, irq_flags);
+			}
+			break;
+		}
+
+		/*
+		 * The following is done on behalf of the newly created
+		 * kthread. That kthread is responsible for doing the
+		 * counterpart to the following before it exits.
+		 */
+		(void) xpc_part_ref(&xpc_partitions[ch->partid]);
+		xpc_msgqueue_ref(ch);
+		atomic_inc(&ch->kthreads_assigned);
+		ch->kthreads_created++;	// >>> temporary debug only!!!
+	}
+}
+
+
+void
+xpc_disconnect_wait(int ch_number)
+{
+	partid_t partid;
+	struct xpc_partition *part;
+	struct xpc_channel *ch;
+
+
+	/* now wait for all callouts to the caller's function to cease */
+	for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
+		part = &xpc_partitions[partid];
+
+		if (xpc_part_ref(part)) {
+			ch = &part->channels[ch_number];
+
+// >>> how do we keep from falling into the window between our check and going
+// >>> down and coming back up where sema is re-inited?
+			if (ch->flags & XPC_C_SETUP) {
+				(void) down(&ch->teardown_sema);
+			}
+
+			xpc_part_deref(part);
+		}
+	}
+}
+
+
+static void
+xpc_do_exit(void)
+{
+	partid_t partid;
+	int active_part_count;
+	struct xpc_partition *part;
+
+
+	/* now it's time to eliminate our heartbeat */
+	del_timer_sync(&xpc_hb_timer);
+	xpc_vars->heartbeating_to_mask = 0;
+
+	/* indicate to others that our reserved page is uninitialized */
+	xpc_rsvd_page->vars_pa = 0;
+
+	/*
+	 * Ignore all incoming interrupts. Without interupts the heartbeat
+	 * checker won't activate any new partitions that may come up.
+	 */
+	free_irq(SGI_XPC_ACTIVATE, NULL);
+
+	/*
+	 * Cause the heartbeat checker and the discovery threads to exit.
+	 * We don't want them attempting to activate new partitions as we
+	 * try to deactivate the existing ones.
+	 */
+	xpc_exiting = 1;
+	wake_up_interruptible(&xpc_act_IRQ_wq);
+
+	/* wait for the heartbeat checker thread to mark itself inactive */
+	down(&xpc_hb_checker_exited);
+
+	/* wait for the discovery thread to mark itself inactive */
+	down(&xpc_discovery_exited);
+
+
+	set_current_state(TASK_INTERRUPTIBLE);
+	schedule_timeout(0.3 * HZ);
+	set_current_state(TASK_RUNNING);
+
+
+	/* wait for all partitions to become inactive */
+
+	do {
+		active_part_count = 0;
+
+		for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
+			part = &xpc_partitions[partid];
+			if (part->act_state != XPC_P_INACTIVE) {
+				active_part_count++;
+
+				XPC_DEACTIVATE_PARTITION(part, xpcUnloading);
+			}
+		}
+
+		if (active_part_count) {
+			set_current_state(TASK_INTERRUPTIBLE);
+			schedule_timeout(0.3 * HZ);
+			set_current_state(TASK_RUNNING);
+		}
+
+	} while (active_part_count > 0);
+
+
+	/* close down protections for IPI operations */
+	xpc_restrict_IPI_ops();
+
+
+	/* clear the interface to XPC's functions */
+	xpc_clear_interface();
+
+	if (xpc_sysctl) {
+		unregister_sysctl_table(xpc_sysctl);
+	}
+}
+
+
+int __init
+xpc_init(void)
+{
+	int ret;
+	partid_t partid;
+	struct xpc_partition *part;
+	pid_t pid;
+
+
+	/*
+	 * xpc_remote_copy_buffer is used as a temporary buffer for bte_copy'ng
+	 * both a partition's reserved page and its XPC variables. Its size was
+	 * based on the size of a reserved page. So we need to ensure that the
+	 * XPC variables will fit as well.
+	 */
+	if (XPC_VARS_ALIGNED_SIZE > XPC_RSVD_PAGE_ALIGNED_SIZE) {
+		dev_err(xpc_part, "xpc_remote_copy_buffer is not big enough\n");
+		return -EPERM;
+	}
+	DBUG_ON((u64) xpc_remote_copy_buffer !=
+				L1_CACHE_ALIGN((u64) xpc_remote_copy_buffer));
+
+	snprintf(xpc_part->bus_id, BUS_ID_SIZE, "part");
+	snprintf(xpc_chan->bus_id, BUS_ID_SIZE, "chan");
+
+	xpc_sysctl = register_sysctl_table(xpc_sys_dir, 1);
+
+	/*
+	 * The first few fields of each entry of xpc_partitions[] need to
+	 * be initialized now so that calls to xpc_connect() and
+	 * xpc_disconnect() can be made prior to the activation of any remote
+	 * partition. NOTE THAT NONE OF THE OTHER FIELDS BELONGING TO THESE
+	 * ENTRIES ARE MEANINGFUL UNTIL AFTER AN ENTRY'S CORRESPONDING
+	 * PARTITION HAS BEEN ACTIVATED.
+	 */
+	for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
+		part = &xpc_partitions[partid];
+
+		DBUG_ON((u64) part != L1_CACHE_ALIGN((u64) part));
+
+		part->act_IRQ_rcvd = 0;
+		spin_lock_init(&part->act_lock);
+		part->act_state = XPC_P_INACTIVE;
+		XPC_SET_REASON(part, 0, 0);
+		part->setup_state = XPC_P_UNSET;
+		init_waitqueue_head(&part->teardown_wq);
+		atomic_set(&part->references, 0);
+	}
+
+	/*
+	 * Open up protections for IPI operations (and AMO operations on
+	 * Shub 1.1 systems).
+	 */
+	xpc_allow_IPI_ops();
+
+	/*
+	 * Interrupts being processed will increment this atomic variable and
+	 * awaken the heartbeat thread which will process the interrupts.
+	 */
+	atomic_set(&xpc_act_IRQ_rcvd, 0);
+
+	/*
+	 * This is safe to do before the xpc_hb_checker thread has started
+	 * because the handler releases a wait queue.  If an interrupt is
+	 * received before the thread is waiting, it will not go to sleep,
+	 * but rather immediately process the interrupt.
+	 */
+	ret = request_irq(SGI_XPC_ACTIVATE, xpc_act_IRQ_handler, 0,
+							"xpc hb", NULL);
+	if (ret != 0) {
+		dev_err(xpc_part, "can't register ACTIVATE IRQ handler, "
+			"errno=%d\n", -ret);
+
+		xpc_restrict_IPI_ops();
+
+		if (xpc_sysctl) {
+			unregister_sysctl_table(xpc_sysctl);
+		}
+		return -EBUSY;
+	}
+
+	/*
+	 * Fill the partition reserved page with the information needed by
+	 * other partitions to discover we are alive and establish initial
+	 * communications.
+	 */
+	xpc_rsvd_page = xpc_rsvd_page_init();
+	if (xpc_rsvd_page == NULL) {
+		dev_err(xpc_part, "could not setup our reserved page\n");
+
+		free_irq(SGI_XPC_ACTIVATE, NULL);
+		xpc_restrict_IPI_ops();
+
+		if (xpc_sysctl) {
+			unregister_sysctl_table(xpc_sysctl);
+		}
+		return -EBUSY;
+	}
+
+
+	/*
+	 * Set the beating to other partitions into motion.  This is
+	 * the last requirement for other partitions' discovery to
+	 * initiate communications with us.
+	 */
+	init_timer(&xpc_hb_timer);
+	xpc_hb_timer.function = xpc_hb_beater;
+	xpc_hb_beater(0);
+
+
+	/*
+	 * The real work-horse behind xpc.  This processes incoming
+	 * interrupts and monitors remote heartbeats.
+	 */
+	pid = kernel_thread(xpc_hb_checker, NULL, 0);
+	if (pid < 0) {
+		dev_err(xpc_part, "failed while forking hb check thread\n");
+
+		/* indicate to others that our reserved page is uninitialized */
+		xpc_rsvd_page->vars_pa = 0;
+
+		del_timer_sync(&xpc_hb_timer);
+		free_irq(SGI_XPC_ACTIVATE, NULL);
+		xpc_restrict_IPI_ops();
+
+		if (xpc_sysctl) {
+			unregister_sysctl_table(xpc_sysctl);
+		}
+		return -EBUSY;
+	}
+
+
+	/*
+	 * Startup a thread that will attempt to discover other partitions to
+	 * activate based on info provided by SAL. This new thread is short
+	 * lived and will exit once discovery is complete.
+	 */
+	pid = kernel_thread(xpc_initiate_discovery, NULL, 0);
+	if (pid < 0) {
+		dev_err(xpc_part, "failed while forking discovery thread\n");
+
+		/* mark this new thread as a non-starter */
+		up(&xpc_discovery_exited);
+
+		xpc_do_exit();
+		return -EBUSY;
+	}
+
+
+	/* set the interface to point at XPC's functions */
+	xpc_set_interface(xpc_initiate_connect, xpc_initiate_disconnect,
+			  xpc_initiate_allocate, xpc_initiate_send,
+			  xpc_initiate_send_notify, xpc_initiate_received,
+			  xpc_initiate_partid_to_nasids);
+
+	return 0;
+}
+module_init(xpc_init);
+
+
+void __exit
+xpc_exit(void)
+{
+	xpc_do_exit();
+}
+module_exit(xpc_exit);
+
+
+MODULE_AUTHOR("Silicon Graphics, Inc.");
+MODULE_DESCRIPTION("Cross Partition Communication (XPC) support");
+MODULE_LICENSE("GPL");
+
+module_param(xpc_hb_interval, int, 0);
+MODULE_PARM_DESC(xpc_hb_interval, "Number of seconds between "
+		"heartbeat increments.");
+
+module_param(xpc_hb_check_interval, int, 0);
+MODULE_PARM_DESC(xpc_hb_check_interval, "Number of seconds between "
+		"heartbeat checks.");
+