]> pilppa.org Git - linux-2.6-omap-h63xx.git/blobdiff - drivers/misc/sgi-xp/xpc_sn2.c
Merge branch 'for_rmk' of git://git.mnementh.co.uk/linux-2.6-im into devel
[linux-2.6-omap-h63xx.git] / drivers / misc / sgi-xp / xpc_sn2.c
index 89c0bb9a27f57a8a9414e8fa53b495a3114c8dac..b4882ccf6344a69607ee8c62f1ff3c84c32ce894 100644 (file)
  *
  */
 
-#include <linux/kernel.h>
 #include <linux/delay.h>
 #include <asm/uncached.h>
+#include <asm/sn/mspec.h>
 #include <asm/sn/sn_sal.h>
 #include "xpc.h"
 
-static struct xpc_vars_sn2 *xpc_vars;  /* >>> Add _sn2 suffix? */
-static struct xpc_vars_part_sn2 *xpc_vars_part; /* >>> Add _sn2 suffix? */
+/*
+ * Define the number of u64s required to represent all the C-brick nasids
+ * as a bitmap.  The cross-partition kernel modules deal only with
+ * C-brick nasids, thus the need for bitmaps which don't account for
+ * odd-numbered (non C-brick) nasids.
+ */
+#define XPC_MAX_PHYSNODES_SN2  (MAX_NUMALINK_NODES / 2)
+#define XP_NASID_MASK_BYTES_SN2        ((XPC_MAX_PHYSNODES_SN2 + 7) / 8)
+#define XP_NASID_MASK_WORDS_SN2        ((XPC_MAX_PHYSNODES_SN2 + 63) / 64)
+
+/*
+ * Memory for XPC's amo variables is allocated by the MSPEC driver. These
+ * pages are located in the lowest granule. The lowest granule uses 4k pages
+ * for cached references and an alternate TLB handler to never provide a
+ * cacheable mapping for the entire region. This will prevent speculative
+ * reading of cached copies of our lines from being issued which will cause
+ * a PI FSB Protocol error to be generated by the SHUB. For XPC, we need 64
+ * amo variables (based on XP_MAX_NPARTITIONS_SN2) to identify the senders of
+ * NOTIFY IRQs, 128 amo variables (based on XP_NASID_MASK_WORDS_SN2) to identify
+ * the senders of ACTIVATE IRQs, 1 amo variable to identify which remote
+ * partitions (i.e., XPCs) consider themselves currently engaged with the
+ * local XPC and 1 amo variable to request partition deactivation.
+ */
+#define XPC_NOTIFY_IRQ_AMOS_SN2                0
+#define XPC_ACTIVATE_IRQ_AMOS_SN2      (XPC_NOTIFY_IRQ_AMOS_SN2 + \
+                                        XP_MAX_NPARTITIONS_SN2)
+#define XPC_ENGAGED_PARTITIONS_AMO_SN2 (XPC_ACTIVATE_IRQ_AMOS_SN2 + \
+                                        XP_NASID_MASK_WORDS_SN2)
+#define XPC_DEACTIVATE_REQUEST_AMO_SN2 (XPC_ENGAGED_PARTITIONS_AMO_SN2 + 1)
+
+/*
+ * Buffer used to store a local copy of portions of a remote partition's
+ * reserved page (either its header and part_nasids mask, or its vars).
+ */
+static void *xpc_remote_copy_buffer_base_sn2;
+static char *xpc_remote_copy_buffer_sn2;
+
+static struct xpc_vars_sn2 *xpc_vars_sn2;
+static struct xpc_vars_part_sn2 *xpc_vars_part_sn2;
+
+static int
+xpc_setup_partitions_sn_sn2(void)
+{
+       /* nothing needs to be done */
+       return 0;
+}
+
+/* SH_IPI_ACCESS shub register value on startup */
+static u64 xpc_sh1_IPI_access_sn2;
+static u64 xpc_sh2_IPI_access0_sn2;
+static u64 xpc_sh2_IPI_access1_sn2;
+static u64 xpc_sh2_IPI_access2_sn2;
+static u64 xpc_sh2_IPI_access3_sn2;
+
+/*
+ * Change protections to allow IPI operations.
+ */
+static void
+xpc_allow_IPI_ops_sn2(void)
+{
+       int node;
+       int nasid;
+
+       /* !!! The following should get moved into SAL. */
+       if (is_shub2()) {
+               xpc_sh2_IPI_access0_sn2 =
+                   (u64)HUB_L((u64 *)LOCAL_MMR_ADDR(SH2_IPI_ACCESS0));
+               xpc_sh2_IPI_access1_sn2 =
+                   (u64)HUB_L((u64 *)LOCAL_MMR_ADDR(SH2_IPI_ACCESS1));
+               xpc_sh2_IPI_access2_sn2 =
+                   (u64)HUB_L((u64 *)LOCAL_MMR_ADDR(SH2_IPI_ACCESS2));
+               xpc_sh2_IPI_access3_sn2 =
+                   (u64)HUB_L((u64 *)LOCAL_MMR_ADDR(SH2_IPI_ACCESS3));
+
+               for_each_online_node(node) {
+                       nasid = cnodeid_to_nasid(node);
+                       HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS0),
+                             -1UL);
+                       HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS1),
+                             -1UL);
+                       HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS2),
+                             -1UL);
+                       HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS3),
+                             -1UL);
+               }
+       } else {
+               xpc_sh1_IPI_access_sn2 =
+                   (u64)HUB_L((u64 *)LOCAL_MMR_ADDR(SH1_IPI_ACCESS));
+
+               for_each_online_node(node) {
+                       nasid = cnodeid_to_nasid(node);
+                       HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid, SH1_IPI_ACCESS),
+                             -1UL);
+               }
+       }
+}
 
 /*
- * The following set of macros and functions are used for the sending and
- * receiving of IPIs (also known as IRQs). There are two flavors of IPIs,
- * one that is associated with partition activity (SGI_XPC_ACTIVATE) and
- * the other that is associated with channel activity (SGI_XPC_NOTIFY).
+ * Restrict protections to disallow IPI operations.
+ */
+static void
+xpc_disallow_IPI_ops_sn2(void)
+{
+       int node;
+       int nasid;
+
+       /* !!! The following should get moved into SAL. */
+       if (is_shub2()) {
+               for_each_online_node(node) {
+                       nasid = cnodeid_to_nasid(node);
+                       HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS0),
+                             xpc_sh2_IPI_access0_sn2);
+                       HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS1),
+                             xpc_sh2_IPI_access1_sn2);
+                       HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS2),
+                             xpc_sh2_IPI_access2_sn2);
+                       HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS3),
+                             xpc_sh2_IPI_access3_sn2);
+               }
+       } else {
+               for_each_online_node(node) {
+                       nasid = cnodeid_to_nasid(node);
+                       HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid, SH1_IPI_ACCESS),
+                             xpc_sh1_IPI_access_sn2);
+               }
+       }
+}
+
+/*
+ * The following set of functions are used for the sending and receiving of
+ * IRQs (also known as IPIs). There are two flavors of IRQs, one that is
+ * associated with partition activity (SGI_XPC_ACTIVATE) and the other that
+ * is associated with channel activity (SGI_XPC_NOTIFY).
  */
 
 static u64
-xpc_IPI_receive_sn2(AMO_t *amo)
+xpc_receive_IRQ_amo_sn2(struct amo *amo)
 {
        return FETCHOP_LOAD_OP(TO_AMO((u64)&amo->variable), FETCHOP_CLEAR);
 }
 
 static enum xp_retval
-xpc_IPI_send_sn2(AMO_t *amo, u64 flag, int nasid, int phys_cpuid, int vector)
+xpc_send_IRQ_sn2(struct amo *amo, u64 flag, int nasid, int phys_cpuid,
+                int vector)
 {
        int ret = 0;
        unsigned long irq_flags;
@@ -50,111 +176,182 @@ xpc_IPI_send_sn2(AMO_t *amo, u64 flag, int nasid, int phys_cpuid, int vector)
         * We must always use the nofault function regardless of whether we
         * are on a Shub 1.1 system or a Shub 1.2 slice 0xc processor. If we
         * didn't, we'd never know that the other partition is down and would
-        * keep sending IPIs and AMOs to it until the heartbeat times out.
+        * keep sending IRQs and amos to it until the heartbeat times out.
         */
        ret = xp_nofault_PIOR((u64 *)GLOBAL_MMR_ADDR(NASID_GET(&amo->variable),
                                                     xp_nofault_PIOR_target));
 
        local_irq_restore(irq_flags);
 
-       return ((ret == 0) ? xpSuccess : xpPioReadError);
+       return (ret == 0) ? xpSuccess : xpPioReadError;
 }
 
-static AMO_t *
-xpc_IPI_init_sn2(int index)
+static struct amo *
+xpc_init_IRQ_amo_sn2(int index)
 {
-       AMO_t *amo = xpc_vars->amos_page + index;
+       struct amo *amo = xpc_vars_sn2->amos_page + index;
 
-       (void)xpc_IPI_receive_sn2(amo); /* clear AMO variable */
+       (void)xpc_receive_IRQ_amo_sn2(amo);     /* clear amo variable */
        return amo;
 }
 
 /*
- * IPIs associated with SGI_XPC_ACTIVATE IRQ.
+ * Functions associated with SGI_XPC_ACTIVATE IRQ.
  */
 
 /*
- * Flag the appropriate AMO variable and send an IPI to the specified node.
+ * Notify the heartbeat check thread that an activate IRQ has been received.
  */
-static void
-xpc_activate_IRQ_send_sn2(u64 amos_page_pa, int from_nasid, int to_nasid,
-                     int to_phys_cpuid)
+static irqreturn_t
+xpc_handle_activate_IRQ_sn2(int irq, void *dev_id)
 {
-       int w_index = XPC_NASID_W_INDEX(from_nasid);
-       int b_index = XPC_NASID_B_INDEX(from_nasid);
-       AMO_t *amos = (AMO_t *)__va(amos_page_pa +
-                                   (XPC_ACTIVATE_IRQ_AMOS * sizeof(AMO_t)));
+       unsigned long irq_flags;
 
-       (void)xpc_IPI_send_sn2(&amos[w_index], (1UL << b_index), to_nasid,
-                              to_phys_cpuid, SGI_XPC_ACTIVATE);
+       spin_lock_irqsave(&xpc_activate_IRQ_rcvd_lock, irq_flags);
+       xpc_activate_IRQ_rcvd++;
+       spin_unlock_irqrestore(&xpc_activate_IRQ_rcvd_lock, irq_flags);
+
+       wake_up_interruptible(&xpc_activate_IRQ_wq);
+       return IRQ_HANDLED;
 }
 
+/*
+ * Flag the appropriate amo variable and send an IRQ to the specified node.
+ */
 static void
-xpc_activate_IRQ_send_local_sn2(int from_nasid)
+xpc_send_activate_IRQ_sn2(unsigned long amos_page_pa, int from_nasid,
+                         int to_nasid, int to_phys_cpuid)
 {
-       int w_index = XPC_NASID_W_INDEX(from_nasid);
-       int b_index = XPC_NASID_B_INDEX(from_nasid);
-       AMO_t *amos = (AMO_t *)__va(xpc_vars->amos_page_pa +
-                                   (XPC_ACTIVATE_IRQ_AMOS * sizeof(AMO_t)));
+       struct amo *amos = (struct amo *)__va(amos_page_pa +
+                                             (XPC_ACTIVATE_IRQ_AMOS_SN2 *
+                                             sizeof(struct amo)));
 
-       /* fake the sending and receipt of an activate IRQ from remote nasid */
-       FETCHOP_STORE_OP(TO_AMO((u64)&amos[w_index].variable), FETCHOP_OR,
-                        (1UL << b_index));
-       atomic_inc(&xpc_act_IRQ_rcvd);
-       wake_up_interruptible(&xpc_act_IRQ_wq);
+       (void)xpc_send_IRQ_sn2(&amos[BIT_WORD(from_nasid / 2)],
+                              BIT_MASK(from_nasid / 2), to_nasid,
+                              to_phys_cpuid, SGI_XPC_ACTIVATE);
 }
 
 static void
-xpc_IPI_send_local_activate_sn2(int from_nasid)
+xpc_send_local_activate_IRQ_sn2(int from_nasid)
 {
-       xpc_activate_IRQ_send_local_sn2(from_nasid);
-}
+       unsigned long irq_flags;
+       struct amo *amos = (struct amo *)__va(xpc_vars_sn2->amos_page_pa +
+                                             (XPC_ACTIVATE_IRQ_AMOS_SN2 *
+                                             sizeof(struct amo)));
 
-static void
-xpc_IPI_send_activated_sn2(struct xpc_partition *part)
-{
-       xpc_activate_IRQ_send_sn2(part->remote_amos_page_pa,
-                                 cnodeid_to_nasid(0), part->remote_act_nasid,
-                                 part->remote_act_phys_cpuid);
+       /* fake the sending and receipt of an activate IRQ from remote nasid */
+       FETCHOP_STORE_OP(TO_AMO((u64)&amos[BIT_WORD(from_nasid / 2)].variable),
+                        FETCHOP_OR, BIT_MASK(from_nasid / 2));
+
+       spin_lock_irqsave(&xpc_activate_IRQ_rcvd_lock, irq_flags);
+       xpc_activate_IRQ_rcvd++;
+       spin_unlock_irqrestore(&xpc_activate_IRQ_rcvd_lock, irq_flags);
+
+       wake_up_interruptible(&xpc_activate_IRQ_wq);
 }
 
+/*
+ * Functions associated with SGI_XPC_NOTIFY IRQ.
+ */
+
+/*
+ * Check to see if any chctl flags were sent from the specified partition.
+ */
 static void
-xpc_IPI_send_local_reactivate_sn2(int from_nasid)
+xpc_check_for_sent_chctl_flags_sn2(struct xpc_partition *part)
 {
-       xpc_activate_IRQ_send_local_sn2(from_nasid);
+       union xpc_channel_ctl_flags chctl;
+       unsigned long irq_flags;
+
+       chctl.all_flags = xpc_receive_IRQ_amo_sn2(part->sn.sn2.
+                                                 local_chctl_amo_va);
+       if (chctl.all_flags == 0)
+               return;
+
+       spin_lock_irqsave(&part->chctl_lock, irq_flags);
+       part->chctl.all_flags |= chctl.all_flags;
+       spin_unlock_irqrestore(&part->chctl_lock, irq_flags);
+
+       dev_dbg(xpc_chan, "received notify IRQ from partid=%d, chctl.all_flags="
+               "0x%lx\n", XPC_PARTID(part), chctl.all_flags);
+
+       xpc_wakeup_channel_mgr(part);
 }
 
-static void
-xpc_IPI_send_disengage_sn2(struct xpc_partition *part)
+/*
+ * 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 structure per partition to indicate
+ * whether a partition has sent an IRQ 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 IRQs sent by XPC
+ * running on other partitions.
+ *
+ * Noteworthy Arguments:
+ *
+ *     irq - Interrupt ReQuest number. NOT USED.
+ *
+ *     dev_id - partid of IRQ's potential sender.
+ */
+static irqreturn_t
+xpc_handle_notify_IRQ_sn2(int irq, void *dev_id)
 {
-       xpc_activate_IRQ_send_sn2(part->remote_amos_page_pa,
-                                 cnodeid_to_nasid(0), part->remote_act_nasid,
-                                 part->remote_act_phys_cpuid);
+       short partid = (short)(u64)dev_id;
+       struct xpc_partition *part = &xpc_partitions[partid];
+
+       DBUG_ON(partid < 0 || partid >= XP_MAX_NPARTITIONS_SN2);
+
+       if (xpc_part_ref(part)) {
+               xpc_check_for_sent_chctl_flags_sn2(part);
+
+               xpc_part_deref(part);
+       }
+       return IRQ_HANDLED;
 }
 
 /*
- * IPIs associated with SGI_XPC_NOTIFY IRQ.
+ * Check to see if xpc_handle_notify_IRQ_sn2() dropped any IRQs on the floor
+ * because the write to their associated amo variable completed after the IRQ
+ * was received.
  */
+static void
+xpc_check_for_dropped_notify_IRQ_sn2(struct xpc_partition *part)
+{
+       struct xpc_partition_sn2 *part_sn2 = &part->sn.sn2;
+
+       if (xpc_part_ref(part)) {
+               xpc_check_for_sent_chctl_flags_sn2(part);
+
+               part_sn2->dropped_notify_IRQ_timer.expires = jiffies +
+                   XPC_DROPPED_NOTIFY_IRQ_WAIT_INTERVAL;
+               add_timer(&part_sn2->dropped_notify_IRQ_timer);
+               xpc_part_deref(part);
+       }
+}
 
 /*
- * Send an IPI to the remote partition that is associated with the
+ * Send a notify IRQ to the remote partition that is associated with the
  * specified channel.
  */
 static void
-xpc_notify_IRQ_send_sn2(struct xpc_channel *ch, u8 ipi_flag,
-                       char *ipi_flag_string, unsigned long *irq_flags)
+xpc_send_notify_IRQ_sn2(struct xpc_channel *ch, u8 chctl_flag,
+                       char *chctl_flag_string, unsigned long *irq_flags)
 {
        struct xpc_partition *part = &xpc_partitions[ch->partid];
+       struct xpc_partition_sn2 *part_sn2 = &part->sn.sn2;
+       union xpc_channel_ctl_flags chctl = { 0 };
        enum xp_retval ret;
 
-       if (likely(part->act_state != XPC_P_DEACTIVATING)) {
-               ret = xpc_IPI_send_sn2(part->remote_IPI_amo_va,
-                                      (u64)ipi_flag << (ch->number * 8),
-                                      part->remote_IPI_nasid,
-                                      part->remote_IPI_phys_cpuid,
+       if (likely(part->act_state != XPC_P_AS_DEACTIVATING)) {
+               chctl.flags[ch->number] = chctl_flag;
+               ret = xpc_send_IRQ_sn2(part_sn2->remote_chctl_amo_va,
+                                      chctl.all_flags,
+                                      part_sn2->notify_IRQ_nasid,
+                                      part_sn2->notify_IRQ_phys_cpuid,
                                       SGI_XPC_NOTIFY);
                dev_dbg(xpc_chan, "%s sent to partid=%d, channel=%d, ret=%d\n",
-                       ipi_flag_string, ch->partid, ch->number, ret);
+                       chctl_flag_string, ch->partid, ch->number, ret);
                if (unlikely(ret != xpSuccess)) {
                        if (irq_flags != NULL)
                                spin_unlock_irqrestore(&ch->lock, *irq_flags);
@@ -165,75 +362,85 @@ xpc_notify_IRQ_send_sn2(struct xpc_channel *ch, u8 ipi_flag,
        }
 }
 
-#define XPC_NOTIFY_IRQ_SEND_SN2(_ch, _ipi_f, _irq_f) \
-               xpc_notify_IRQ_send_sn2(_ch, _ipi_f, #_ipi_f, _irq_f)
+#define XPC_SEND_NOTIFY_IRQ_SN2(_ch, _ipi_f, _irq_f) \
+               xpc_send_notify_IRQ_sn2(_ch, _ipi_f, #_ipi_f, _irq_f)
 
 /*
  * Make it look like the remote partition, which is associated with the
- * specified channel, sent us an IPI. This faked IPI will be handled
- * by xpc_dropped_IPI_check().
+ * specified channel, sent us a notify IRQ. This faked IRQ will be handled
+ * by xpc_check_for_dropped_notify_IRQ_sn2().
  */
 static void
-xpc_notify_IRQ_send_local_sn2(struct xpc_channel *ch, u8 ipi_flag,
-                             char *ipi_flag_string)
+xpc_send_local_notify_IRQ_sn2(struct xpc_channel *ch, u8 chctl_flag,
+                             char *chctl_flag_string)
 {
        struct xpc_partition *part = &xpc_partitions[ch->partid];
+       union xpc_channel_ctl_flags chctl = { 0 };
 
-       FETCHOP_STORE_OP(TO_AMO((u64)&part->local_IPI_amo_va->variable),
-                        FETCHOP_OR, ((u64)ipi_flag << (ch->number * 8)));
+       chctl.flags[ch->number] = chctl_flag;
+       FETCHOP_STORE_OP(TO_AMO((u64)&part->sn.sn2.local_chctl_amo_va->
+                               variable), FETCHOP_OR, chctl.all_flags);
        dev_dbg(xpc_chan, "%s sent local from partid=%d, channel=%d\n",
-               ipi_flag_string, ch->partid, ch->number);
+               chctl_flag_string, ch->partid, ch->number);
 }
 
-#define XPC_NOTIFY_IRQ_SEND_LOCAL_SN2(_ch, _ipi_f) \
-               xpc_notify_IRQ_send_local_sn2(_ch, _ipi_f, #_ipi_f)
+#define XPC_SEND_LOCAL_NOTIFY_IRQ_SN2(_ch, _ipi_f) \
+               xpc_send_local_notify_IRQ_sn2(_ch, _ipi_f, #_ipi_f)
 
 static void
-xpc_IPI_send_closerequest_sn2(struct xpc_channel *ch, unsigned long *irq_flags)
+xpc_send_chctl_closerequest_sn2(struct xpc_channel *ch,
+                               unsigned long *irq_flags)
 {
-       struct xpc_openclose_args *args = ch->local_openclose_args;
+       struct xpc_openclose_args *args = ch->sn.sn2.local_openclose_args;
 
        args->reason = ch->reason;
-       XPC_NOTIFY_IRQ_SEND_SN2(ch, XPC_IPI_CLOSEREQUEST, irq_flags);
+       XPC_SEND_NOTIFY_IRQ_SN2(ch, XPC_CHCTL_CLOSEREQUEST, irq_flags);
 }
 
 static void
-xpc_IPI_send_closereply_sn2(struct xpc_channel *ch, unsigned long *irq_flags)
+xpc_send_chctl_closereply_sn2(struct xpc_channel *ch, unsigned long *irq_flags)
 {
-       XPC_NOTIFY_IRQ_SEND_SN2(ch, XPC_IPI_CLOSEREPLY, irq_flags);
+       XPC_SEND_NOTIFY_IRQ_SN2(ch, XPC_CHCTL_CLOSEREPLY, irq_flags);
 }
 
 static void
-xpc_IPI_send_openrequest_sn2(struct xpc_channel *ch, unsigned long *irq_flags)
+xpc_send_chctl_openrequest_sn2(struct xpc_channel *ch, unsigned long *irq_flags)
 {
-       struct xpc_openclose_args *args = ch->local_openclose_args;
+       struct xpc_openclose_args *args = ch->sn.sn2.local_openclose_args;
 
-       args->msg_size = ch->msg_size;
+       args->entry_size = ch->entry_size;
        args->local_nentries = ch->local_nentries;
-       XPC_NOTIFY_IRQ_SEND_SN2(ch, XPC_IPI_OPENREQUEST, irq_flags);
+       XPC_SEND_NOTIFY_IRQ_SN2(ch, XPC_CHCTL_OPENREQUEST, irq_flags);
 }
 
 static void
-xpc_IPI_send_openreply_sn2(struct xpc_channel *ch, unsigned long *irq_flags)
+xpc_send_chctl_openreply_sn2(struct xpc_channel *ch, unsigned long *irq_flags)
 {
-       struct xpc_openclose_args *args = ch->local_openclose_args;
+       struct xpc_openclose_args *args = ch->sn.sn2.local_openclose_args;
 
        args->remote_nentries = ch->remote_nentries;
        args->local_nentries = ch->local_nentries;
-       args->local_msgqueue_pa = __pa(ch->local_msgqueue);
-       XPC_NOTIFY_IRQ_SEND_SN2(ch, XPC_IPI_OPENREPLY, irq_flags);
+       args->local_msgqueue_pa = xp_pa(ch->sn.sn2.local_msgqueue);
+       XPC_SEND_NOTIFY_IRQ_SN2(ch, XPC_CHCTL_OPENREPLY, irq_flags);
+}
+
+static void
+xpc_send_chctl_msgrequest_sn2(struct xpc_channel *ch)
+{
+       XPC_SEND_NOTIFY_IRQ_SN2(ch, XPC_CHCTL_MSGREQUEST, NULL);
 }
 
 static void
-xpc_IPI_send_msgrequest_sn2(struct xpc_channel *ch)
+xpc_send_chctl_local_msgrequest_sn2(struct xpc_channel *ch)
 {
-       XPC_NOTIFY_IRQ_SEND_SN2(ch, XPC_IPI_MSGREQUEST, NULL);
+       XPC_SEND_LOCAL_NOTIFY_IRQ_SN2(ch, XPC_CHCTL_MSGREQUEST);
 }
 
 static void
-xpc_IPI_send_local_msgrequest_sn2(struct xpc_channel *ch)
+xpc_save_remote_msgqueue_pa_sn2(struct xpc_channel *ch,
+                               unsigned long msgqueue_pa)
 {
-       XPC_NOTIFY_IRQ_SEND_LOCAL_SN2(ch, XPC_IPI_MSGREQUEST);
+       ch->sn.sn2.remote_msgqueue_pa = msgqueue_pa;
 }
 
 /*
@@ -242,23 +449,24 @@ xpc_IPI_send_local_msgrequest_sn2(struct xpc_channel *ch)
  */
 
 static void
-xpc_mark_partition_engaged_sn2(struct xpc_partition *part)
+xpc_indicate_partition_engaged_sn2(struct xpc_partition *part)
 {
        unsigned long irq_flags;
-       AMO_t *amo = (AMO_t *)__va(part->remote_amos_page_pa +
-                                  (XPC_ENGAGED_PARTITIONS_AMO *
-                                   sizeof(AMO_t)));
+       struct amo *amo = (struct amo *)__va(part->sn.sn2.remote_amos_page_pa +
+                                            (XPC_ENGAGED_PARTITIONS_AMO_SN2 *
+                                            sizeof(struct amo)));
 
        local_irq_save(irq_flags);
 
-       /* set bit corresponding to our partid in remote partition's AMO */
+       /* set bit corresponding to our partid in remote partition's amo */
        FETCHOP_STORE_OP(TO_AMO((u64)&amo->variable), FETCHOP_OR,
-                        (1UL << sn_partition_id));
+                        BIT(sn_partition_id));
+
        /*
         * We must always use the nofault function regardless of whether we
         * are on a Shub 1.1 system or a Shub 1.2 slice 0xc processor. If we
         * didn't, we'd never know that the other partition is down and would
-        * keep sending IPIs and AMOs to it until the heartbeat times out.
+        * keep sending IRQs and amos to it until the heartbeat times out.
         */
        (void)xp_nofault_PIOR((u64 *)GLOBAL_MMR_ADDR(NASID_GET(&amo->
                                                               variable),
@@ -268,229 +476,248 @@ xpc_mark_partition_engaged_sn2(struct xpc_partition *part)
 }
 
 static void
-xpc_mark_partition_disengaged_sn2(struct xpc_partition *part)
+xpc_indicate_partition_disengaged_sn2(struct xpc_partition *part)
 {
+       struct xpc_partition_sn2 *part_sn2 = &part->sn.sn2;
        unsigned long irq_flags;
-       AMO_t *amo = (AMO_t *)__va(part->remote_amos_page_pa +
-                                  (XPC_ENGAGED_PARTITIONS_AMO *
-                                   sizeof(AMO_t)));
+       struct amo *amo = (struct amo *)__va(part_sn2->remote_amos_page_pa +
+                                            (XPC_ENGAGED_PARTITIONS_AMO_SN2 *
+                                            sizeof(struct amo)));
 
        local_irq_save(irq_flags);
 
-       /* clear bit corresponding to our partid in remote partition's AMO */
+       /* clear bit corresponding to our partid in remote partition's amo */
        FETCHOP_STORE_OP(TO_AMO((u64)&amo->variable), FETCHOP_AND,
-                        ~(1UL << sn_partition_id));
+                        ~BIT(sn_partition_id));
+
        /*
         * We must always use the nofault function regardless of whether we
         * are on a Shub 1.1 system or a Shub 1.2 slice 0xc processor. If we
         * didn't, we'd never know that the other partition is down and would
-        * keep sending IPIs and AMOs to it until the heartbeat times out.
+        * keep sending IRQs and amos to it until the heartbeat times out.
         */
        (void)xp_nofault_PIOR((u64 *)GLOBAL_MMR_ADDR(NASID_GET(&amo->
                                                               variable),
                                                     xp_nofault_PIOR_target));
 
        local_irq_restore(irq_flags);
-}
-
-static void
-xpc_request_partition_disengage_sn2(struct xpc_partition *part)
-{
-       unsigned long irq_flags;
-       AMO_t *amo = (AMO_t *)__va(part->remote_amos_page_pa +
-                                  (XPC_DISENGAGE_REQUEST_AMO * sizeof(AMO_t)));
-
-       local_irq_save(irq_flags);
 
-       /* set bit corresponding to our partid in remote partition's AMO */
-       FETCHOP_STORE_OP(TO_AMO((u64)&amo->variable), FETCHOP_OR,
-                        (1UL << sn_partition_id));
        /*
-        * We must always use the nofault function regardless of whether we
-        * are on a Shub 1.1 system or a Shub 1.2 slice 0xc processor. If we
-        * didn't, we'd never know that the other partition is down and would
-        * keep sending IPIs and AMOs to it until the heartbeat times out.
+        * Send activate IRQ to get other side to see that we've cleared our
+        * bit in their engaged partitions amo.
         */
-       (void)xp_nofault_PIOR((u64 *)GLOBAL_MMR_ADDR(NASID_GET(&amo->
-                                                              variable),
-                                                    xp_nofault_PIOR_target));
-
-       local_irq_restore(irq_flags);
+       xpc_send_activate_IRQ_sn2(part_sn2->remote_amos_page_pa,
+                                 cnodeid_to_nasid(0),
+                                 part_sn2->activate_IRQ_nasid,
+                                 part_sn2->activate_IRQ_phys_cpuid);
 }
 
 static void
-xpc_cancel_partition_disengage_request_sn2(struct xpc_partition *part)
+xpc_assume_partition_disengaged_sn2(short partid)
 {
-       unsigned long irq_flags;
-       AMO_t *amo = (AMO_t *)__va(part->remote_amos_page_pa +
-                                  (XPC_DISENGAGE_REQUEST_AMO * sizeof(AMO_t)));
-
-       local_irq_save(irq_flags);
+       struct amo *amo = xpc_vars_sn2->amos_page +
+                         XPC_ENGAGED_PARTITIONS_AMO_SN2;
 
-       /* clear bit corresponding to our partid in remote partition's AMO */
+       /* clear bit(s) based on partid mask in our partition's amo */
        FETCHOP_STORE_OP(TO_AMO((u64)&amo->variable), FETCHOP_AND,
-                        ~(1UL << sn_partition_id));
-       /*
-        * We must always use the nofault function regardless of whether we
-        * are on a Shub 1.1 system or a Shub 1.2 slice 0xc processor. If we
-        * didn't, we'd never know that the other partition is down and would
-        * keep sending IPIs and AMOs to it until the heartbeat times out.
-        */
-       (void)xp_nofault_PIOR((u64 *)GLOBAL_MMR_ADDR(NASID_GET(&amo->
-                                                              variable),
-                                                    xp_nofault_PIOR_target));
-
-       local_irq_restore(irq_flags);
+                        ~BIT(partid));
 }
 
-static u64
-xpc_partition_engaged_sn2(u64 partid_mask)
+static int
+xpc_partition_engaged_sn2(short partid)
 {
-       AMO_t *amo = xpc_vars->amos_page + XPC_ENGAGED_PARTITIONS_AMO;
+       struct amo *amo = xpc_vars_sn2->amos_page +
+                         XPC_ENGAGED_PARTITIONS_AMO_SN2;
 
-       /* return our partition's AMO variable ANDed with partid_mask */
+       /* our partition's amo variable ANDed with partid mask */
        return (FETCHOP_LOAD_OP(TO_AMO((u64)&amo->variable), FETCHOP_LOAD) &
-               partid_mask);
+               BIT(partid)) != 0;
 }
 
-static u64
-xpc_partition_disengage_requested_sn2(u64 partid_mask)
+static int
+xpc_any_partition_engaged_sn2(void)
 {
-       AMO_t *amo = xpc_vars->amos_page + XPC_DISENGAGE_REQUEST_AMO;
+       struct amo *amo = xpc_vars_sn2->amos_page +
+                         XPC_ENGAGED_PARTITIONS_AMO_SN2;
 
-       /* return our partition's AMO variable ANDed with partid_mask */
-       return (FETCHOP_LOAD_OP(TO_AMO((u64)&amo->variable), FETCHOP_LOAD) &
-               partid_mask);
+       /* our partition's amo variable */
+       return FETCHOP_LOAD_OP(TO_AMO((u64)&amo->variable), FETCHOP_LOAD) != 0;
 }
 
-static void
-xpc_clear_partition_engaged_sn2(u64 partid_mask)
+/* original protection values for each node */
+static u64 xpc_prot_vec_sn2[MAX_NUMNODES];
+
+/*
+ * Change protections to allow amo operations on non-Shub 1.1 systems.
+ */
+static enum xp_retval
+xpc_allow_amo_ops_sn2(struct amo *amos_page)
 {
-       AMO_t *amo = xpc_vars->amos_page + XPC_ENGAGED_PARTITIONS_AMO;
+       u64 nasid_array = 0;
+       int ret;
 
-       /* clear bit(s) based on partid_mask in our partition's AMO */
-       FETCHOP_STORE_OP(TO_AMO((u64)&amo->variable), FETCHOP_AND,
-                        ~partid_mask);
+       /*
+        * On SHUB 1.1, we cannot call sn_change_memprotect() since the BIST
+        * collides with memory operations. On those systems we call
+        * xpc_allow_amo_ops_shub_wars_1_1_sn2() instead.
+        */
+       if (!enable_shub_wars_1_1()) {
+               ret = sn_change_memprotect(ia64_tpa((u64)amos_page), PAGE_SIZE,
+                                          SN_MEMPROT_ACCESS_CLASS_1,
+                                          &nasid_array);
+               if (ret != 0)
+                       return xpSalError;
+       }
+       return xpSuccess;
 }
 
+/*
+ * Change protections to allow amo operations on Shub 1.1 systems.
+ */
 static void
-xpc_clear_partition_disengage_request_sn2(u64 partid_mask)
+xpc_allow_amo_ops_shub_wars_1_1_sn2(void)
 {
-       AMO_t *amo = xpc_vars->amos_page + XPC_DISENGAGE_REQUEST_AMO;
+       int node;
+       int nasid;
 
-       /* clear bit(s) based on partid_mask in our partition's AMO */
-       FETCHOP_STORE_OP(TO_AMO((u64)&amo->variable), FETCHOP_AND,
-                        ~partid_mask);
+       if (!enable_shub_wars_1_1())
+               return;
+
+       for_each_online_node(node) {
+               nasid = cnodeid_to_nasid(node);
+               /* save current protection values */
+               xpc_prot_vec_sn2[node] =
+                   (u64)HUB_L((u64 *)GLOBAL_MMR_ADDR(nasid,
+                                                 SH1_MD_DQLP_MMR_DIR_PRIVEC0));
+               /* open up everything */
+               HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid,
+                                            SH1_MD_DQLP_MMR_DIR_PRIVEC0),
+                     -1UL);
+               HUB_S((u64 *)GLOBAL_MMR_ADDR(nasid,
+                                            SH1_MD_DQRP_MMR_DIR_PRIVEC0),
+                     -1UL);
+       }
 }
 
 static enum xp_retval
-xpc_rsvd_page_init_sn2(struct xpc_rsvd_page *rp)
+xpc_get_partition_rsvd_page_pa_sn2(void *buf, u64 *cookie, unsigned long *rp_pa,
+                                  size_t *len)
 {
-       AMO_t *amos_page;
-       u64 nasid_array = 0;
+       s64 status;
+       enum xp_retval ret;
+
+       status = sn_partition_reserved_page_pa((u64)buf, cookie, rp_pa, len);
+       if (status == SALRET_OK)
+               ret = xpSuccess;
+       else if (status == SALRET_MORE_PASSES)
+               ret = xpNeedMoreInfo;
+       else
+               ret = xpSalError;
+
+       return ret;
+}
+
+
+static int
+xpc_setup_rsvd_page_sn_sn2(struct xpc_rsvd_page *rp)
+{
+       struct amo *amos_page;
        int i;
        int ret;
 
-       xpc_vars = XPC_RP_VARS(rp);
+       xpc_vars_sn2 = XPC_RP_VARS(rp);
 
-       rp->sn.vars_pa = __pa(xpc_vars);
+       rp->sn.vars_pa = xp_pa(xpc_vars_sn2);
 
        /* vars_part array follows immediately after vars */
-       xpc_vars_part = (struct xpc_vars_part_sn2 *)((u8 *)XPC_RP_VARS(rp) +
-                                                    XPC_RP_VARS_SIZE);
-
+       xpc_vars_part_sn2 = (struct xpc_vars_part_sn2 *)((u8 *)XPC_RP_VARS(rp) +
+                                                        XPC_RP_VARS_SIZE);
 
        /*
-        * Before clearing xpc_vars, see if a page of AMOs had been previously
-        * allocated. If not we'll need to allocate one and set permissions
-        * so that cross-partition AMOs are allowed.
+        * Before clearing xpc_vars_sn2, see if a page of amos had been
+        * previously allocated. If not we'll need to allocate one and set
+        * permissions so that cross-partition amos are allowed.
         *
-        * The allocated AMO page needs MCA reporting to remain disabled after
+        * The allocated amo page needs MCA reporting to remain disabled after
         * XPC has unloaded.  To make this work, we keep a copy of the pointer
-        * to this page (i.e., amos_page) in the struct xpc_vars structure,
+        * to this page (i.e., amos_page) in the struct xpc_vars_sn2 structure,
         * which is pointed to by the reserved page, and re-use that saved copy
-        * on subsequent loads of XPC. This AMO page is never freed, and its
+        * on subsequent loads of XPC. This amo page is never freed, and its
         * memory protections are never restricted.
         */
-       amos_page = xpc_vars->amos_page;
+       amos_page = xpc_vars_sn2->amos_page;
        if (amos_page == NULL) {
-               amos_page = (AMO_t *)TO_AMO(uncached_alloc_page(0, 1));
+               amos_page = (struct amo *)TO_AMO(uncached_alloc_page(0, 1));
                if (amos_page == NULL) {
-                       dev_err(xpc_part, "can't allocate page of AMOs\n");
-                       return xpNoMemory;
+                       dev_err(xpc_part, "can't allocate page of amos\n");
+                       return -ENOMEM;
                }
 
                /*
-                * Open up AMO-R/W to cpu.  This is done for Shub 1.1 systems
-                * when xpc_allow_IPI_ops() is called via xpc_hb_init().
+                * Open up amo-R/W to cpu.  This is done on Shub 1.1 systems
+                * when xpc_allow_amo_ops_shub_wars_1_1_sn2() is called.
                 */
-               if (!enable_shub_wars_1_1()) {
-                       ret = sn_change_memprotect(ia64_tpa((u64)amos_page),
-                                                  PAGE_SIZE,
-                                                  SN_MEMPROT_ACCESS_CLASS_1,
-                                                  &nasid_array);
-                       if (ret != 0) {
-                               dev_err(xpc_part, "can't change memory "
-                                       "protections\n");
-                               uncached_free_page(__IA64_UNCACHED_OFFSET |
-                                                  TO_PHYS((u64)amos_page), 1);
-                               return xpSalError;
-                       }
+               ret = xpc_allow_amo_ops_sn2(amos_page);
+               if (ret != xpSuccess) {
+                       dev_err(xpc_part, "can't allow amo operations\n");
+                       uncached_free_page(__IA64_UNCACHED_OFFSET |
+                                          TO_PHYS((u64)amos_page), 1);
+                       return -EPERM;
                }
        }
 
-       /* clear xpc_vars */
-       memset(xpc_vars, 0, sizeof(struct xpc_vars_sn2));
+       /* clear xpc_vars_sn2 */
+       memset(xpc_vars_sn2, 0, sizeof(struct xpc_vars_sn2));
 
-       xpc_vars->version = XPC_V_VERSION;
-       xpc_vars->act_nasid = cpuid_to_nasid(0);
-       xpc_vars->act_phys_cpuid = cpu_physical_id(0);
-       xpc_vars->vars_part_pa = __pa(xpc_vars_part);
-       xpc_vars->amos_page_pa = ia64_tpa((u64)amos_page);
-       xpc_vars->amos_page = amos_page;        /* save for next load of XPC */
+       xpc_vars_sn2->version = XPC_V_VERSION;
+       xpc_vars_sn2->activate_IRQ_nasid = cpuid_to_nasid(0);
+       xpc_vars_sn2->activate_IRQ_phys_cpuid = cpu_physical_id(0);
+       xpc_vars_sn2->vars_part_pa = xp_pa(xpc_vars_part_sn2);
+       xpc_vars_sn2->amos_page_pa = ia64_tpa((u64)amos_page);
+       xpc_vars_sn2->amos_page = amos_page;    /* save for next load of XPC */
 
-       /* clear xpc_vars_part */
-       memset((u64 *)xpc_vars_part, 0, sizeof(struct xpc_vars_part_sn2) *
-              xp_max_npartitions);
+       /* clear xpc_vars_part_sn2 */
+       memset((u64 *)xpc_vars_part_sn2, 0, sizeof(struct xpc_vars_part_sn2) *
+              XP_MAX_NPARTITIONS_SN2);
 
-       /* initialize the activate IRQ related AMO variables */
-       for (i = 0; i < xp_nasid_mask_words; i++)
-               (void)xpc_IPI_init_sn2(XPC_ACTIVATE_IRQ_AMOS + i);
+       /* initialize the activate IRQ related amo variables */
+       for (i = 0; i < xpc_nasid_mask_nlongs; i++)
+               (void)xpc_init_IRQ_amo_sn2(XPC_ACTIVATE_IRQ_AMOS_SN2 + i);
 
-       /* initialize the engaged remote partitions related AMO variables */
-       (void)xpc_IPI_init_sn2(XPC_ENGAGED_PARTITIONS_AMO);
-       (void)xpc_IPI_init_sn2(XPC_DISENGAGE_REQUEST_AMO);
+       /* initialize the engaged remote partitions related amo variables */
+       (void)xpc_init_IRQ_amo_sn2(XPC_ENGAGED_PARTITIONS_AMO_SN2);
+       (void)xpc_init_IRQ_amo_sn2(XPC_DEACTIVATE_REQUEST_AMO_SN2);
 
-       return xpSuccess;
+       return 0;
 }
 
 static void
 xpc_increment_heartbeat_sn2(void)
 {
-       xpc_vars->heartbeat++;
+       xpc_vars_sn2->heartbeat++;
 }
 
 static void
 xpc_offline_heartbeat_sn2(void)
 {
        xpc_increment_heartbeat_sn2();
-       xpc_vars->heartbeat_offline = 1;
+       xpc_vars_sn2->heartbeat_offline = 1;
 }
 
 static void
 xpc_online_heartbeat_sn2(void)
 {
        xpc_increment_heartbeat_sn2();
-       xpc_vars->heartbeat_offline = 0;
+       xpc_vars_sn2->heartbeat_offline = 0;
 }
 
 static void
 xpc_heartbeat_init_sn2(void)
 {
-       DBUG_ON(xpc_vars == NULL);
+       DBUG_ON(xpc_vars_sn2 == NULL);
 
-       bitmap_zero(xpc_vars->heartbeating_to_mask, XP_MAX_NPARTITIONS_SN2);
-       xpc_heartbeating_to_mask = &xpc_vars->heartbeating_to_mask[0];
+       bitmap_zero(xpc_vars_sn2->heartbeating_to_mask, XP_MAX_NPARTITIONS_SN2);
+       xpc_heartbeating_to_mask = &xpc_vars_sn2->heartbeating_to_mask[0];
        xpc_online_heartbeat_sn2();
 }
 
@@ -500,61 +727,37 @@ xpc_heartbeat_exit_sn2(void)
        xpc_offline_heartbeat_sn2();
 }
 
-/*
- * At periodic intervals, scan through all active partitions and ensure
- * their heartbeat is still active.  If not, the partition is deactivated.
- */
-static void
-xpc_check_remote_hb_sn2(void)
+static enum xp_retval
+xpc_get_remote_heartbeat_sn2(struct xpc_partition *part)
 {
        struct xpc_vars_sn2 *remote_vars;
-       struct xpc_partition *part;
-       short partid;
        enum xp_retval ret;
 
-       remote_vars = (struct xpc_vars_sn2 *)xpc_remote_copy_buffer;
-
-       for (partid = 0; partid < xp_max_npartitions; partid++) {
-
-               if (xpc_exiting)
-                       break;
-
-               if (partid == sn_partition_id)
-                       continue;
+       remote_vars = (struct xpc_vars_sn2 *)xpc_remote_copy_buffer_sn2;
 
-               part = &xpc_partitions[partid];
-
-               if (part->act_state == XPC_P_INACTIVE ||
-                   part->act_state == XPC_P_DEACTIVATING) {
-                       continue;
-               }
-
-               /* pull the remote_hb cache line */
-               ret = xp_remote_memcpy(remote_vars,
-                                      (void *)part->remote_vars_pa,
-                                      XPC_RP_VARS_SIZE);
-               if (ret != xpSuccess) {
-                       XPC_DEACTIVATE_PARTITION(part, ret);
-                       continue;
-               }
-
-               dev_dbg(xpc_part, "partid = %d, heartbeat = %ld, last_heartbeat"
-                       " = %ld, heartbeat_offline = %ld, HB_mask[0] = 0x%lx\n",
-                       partid, remote_vars->heartbeat, part->last_heartbeat,
-                       remote_vars->heartbeat_offline,
-                       remote_vars->heartbeating_to_mask[0]);
-
-               if (((remote_vars->heartbeat == part->last_heartbeat) &&
-                    (remote_vars->heartbeat_offline == 0)) ||
-                   !xpc_hb_allowed(sn_partition_id,
-                                   &remote_vars->heartbeating_to_mask)) {
-
-                       XPC_DEACTIVATE_PARTITION(part, xpNoHeartbeat);
-                       continue;
-               }
+       /* pull the remote vars structure that contains the heartbeat */
+       ret = xp_remote_memcpy(xp_pa(remote_vars),
+                              part->sn.sn2.remote_vars_pa,
+                              XPC_RP_VARS_SIZE);
+       if (ret != xpSuccess)
+               return ret;
 
+       dev_dbg(xpc_part, "partid=%d, heartbeat=%ld, last_heartbeat=%ld, "
+               "heartbeat_offline=%ld, HB_mask[0]=0x%lx\n", XPC_PARTID(part),
+               remote_vars->heartbeat, part->last_heartbeat,
+               remote_vars->heartbeat_offline,
+               remote_vars->heartbeating_to_mask[0]);
+
+       if ((remote_vars->heartbeat == part->last_heartbeat &&
+           remote_vars->heartbeat_offline == 0) ||
+           !xpc_hb_allowed(sn_partition_id,
+                           &remote_vars->heartbeating_to_mask)) {
+               ret = xpNoHeartbeat;
+       } else {
                part->last_heartbeat = remote_vars->heartbeat;
        }
+
+       return ret;
 }
 
 /*
@@ -564,7 +767,8 @@ xpc_check_remote_hb_sn2(void)
  * assumed to be of size XPC_RP_VARS_SIZE.
  */
 static enum xp_retval
-xpc_get_remote_vars_sn2(u64 remote_vars_pa, struct xpc_vars_sn2 *remote_vars)
+xpc_get_remote_vars_sn2(unsigned long remote_vars_pa,
+                       struct xpc_vars_sn2 *remote_vars)
 {
        enum xp_retval ret;
 
@@ -572,7 +776,7 @@ xpc_get_remote_vars_sn2(u64 remote_vars_pa, struct xpc_vars_sn2 *remote_vars)
                return xpVarsNotSet;
 
        /* pull over the cross partition variables */
-       ret = xp_remote_memcpy(remote_vars, (void *)remote_vars_pa,
+       ret = xp_remote_memcpy(xp_pa(remote_vars), remote_vars_pa,
                               XPC_RP_VARS_SIZE);
        if (ret != xpSuccess)
                return ret;
@@ -586,10 +790,91 @@ xpc_get_remote_vars_sn2(u64 remote_vars_pa, struct xpc_vars_sn2 *remote_vars)
 }
 
 static void
-xpc_initiate_partition_activation_sn2(struct xpc_rsvd_page *remote_rp,
-                                     u64 remote_rp_pa, int nasid)
+xpc_request_partition_activation_sn2(struct xpc_rsvd_page *remote_rp,
+                                    unsigned long remote_rp_pa, int nasid)
+{
+       xpc_send_local_activate_IRQ_sn2(nasid);
+}
+
+static void
+xpc_request_partition_reactivation_sn2(struct xpc_partition *part)
+{
+       xpc_send_local_activate_IRQ_sn2(part->sn.sn2.activate_IRQ_nasid);
+}
+
+static void
+xpc_request_partition_deactivation_sn2(struct xpc_partition *part)
+{
+       struct xpc_partition_sn2 *part_sn2 = &part->sn.sn2;
+       unsigned long irq_flags;
+       struct amo *amo = (struct amo *)__va(part_sn2->remote_amos_page_pa +
+                                            (XPC_DEACTIVATE_REQUEST_AMO_SN2 *
+                                            sizeof(struct amo)));
+
+       local_irq_save(irq_flags);
+
+       /* set bit corresponding to our partid in remote partition's amo */
+       FETCHOP_STORE_OP(TO_AMO((u64)&amo->variable), FETCHOP_OR,
+                        BIT(sn_partition_id));
+
+       /*
+        * We must always use the nofault function regardless of whether we
+        * are on a Shub 1.1 system or a Shub 1.2 slice 0xc processor. If we
+        * didn't, we'd never know that the other partition is down and would
+        * keep sending IRQs and amos to it until the heartbeat times out.
+        */
+       (void)xp_nofault_PIOR((u64 *)GLOBAL_MMR_ADDR(NASID_GET(&amo->
+                                                              variable),
+                                                    xp_nofault_PIOR_target));
+
+       local_irq_restore(irq_flags);
+
+       /*
+        * Send activate IRQ to get other side to see that we've set our
+        * bit in their deactivate request amo.
+        */
+       xpc_send_activate_IRQ_sn2(part_sn2->remote_amos_page_pa,
+                                 cnodeid_to_nasid(0),
+                                 part_sn2->activate_IRQ_nasid,
+                                 part_sn2->activate_IRQ_phys_cpuid);
+}
+
+static void
+xpc_cancel_partition_deactivation_request_sn2(struct xpc_partition *part)
 {
-       xpc_IPI_send_local_activate(nasid);
+       unsigned long irq_flags;
+       struct amo *amo = (struct amo *)__va(part->sn.sn2.remote_amos_page_pa +
+                                            (XPC_DEACTIVATE_REQUEST_AMO_SN2 *
+                                            sizeof(struct amo)));
+
+       local_irq_save(irq_flags);
+
+       /* clear bit corresponding to our partid in remote partition's amo */
+       FETCHOP_STORE_OP(TO_AMO((u64)&amo->variable), FETCHOP_AND,
+                        ~BIT(sn_partition_id));
+
+       /*
+        * We must always use the nofault function regardless of whether we
+        * are on a Shub 1.1 system or a Shub 1.2 slice 0xc processor. If we
+        * didn't, we'd never know that the other partition is down and would
+        * keep sending IRQs and amos to it until the heartbeat times out.
+        */
+       (void)xp_nofault_PIOR((u64 *)GLOBAL_MMR_ADDR(NASID_GET(&amo->
+                                                              variable),
+                                                    xp_nofault_PIOR_target));
+
+       local_irq_restore(irq_flags);
+}
+
+static int
+xpc_partition_deactivation_requested_sn2(short partid)
+{
+       struct amo *amo = xpc_vars_sn2->amos_page +
+                         XPC_DEACTIVATE_REQUEST_AMO_SN2;
+
+       /* our partition's amo variable ANDed with partid mask */
+       return (FETCHOP_LOAD_OP(TO_AMO((u64)&amo->variable), FETCHOP_LOAD) &
+               BIT(partid)) != 0;
 }
 
 /*
@@ -597,56 +882,60 @@ xpc_initiate_partition_activation_sn2(struct xpc_rsvd_page *remote_rp,
  */
 static void
 xpc_update_partition_info_sn2(struct xpc_partition *part, u8 remote_rp_version,
-                             struct timespec *remote_rp_stamp,
-                             u64 remote_rp_pa, u64 remote_vars_pa,
+                             unsigned long *remote_rp_ts_jiffies,
+                             unsigned long remote_rp_pa,
+                             unsigned long remote_vars_pa,
                              struct xpc_vars_sn2 *remote_vars)
 {
+       struct xpc_partition_sn2 *part_sn2 = &part->sn.sn2;
+
        part->remote_rp_version = remote_rp_version;
        dev_dbg(xpc_part, "  remote_rp_version = 0x%016x\n",
                part->remote_rp_version);
 
-       part->remote_rp_stamp = *remote_rp_stamp;
-       dev_dbg(xpc_part, "  remote_rp_stamp (tv_sec = 0x%lx tv_nsec = 0x%lx\n",
-               part->remote_rp_stamp.tv_sec, part->remote_rp_stamp.tv_nsec);
+       part->remote_rp_ts_jiffies = *remote_rp_ts_jiffies;
+       dev_dbg(xpc_part, "  remote_rp_ts_jiffies = 0x%016lx\n",
+               part->remote_rp_ts_jiffies);
 
        part->remote_rp_pa = remote_rp_pa;
        dev_dbg(xpc_part, "  remote_rp_pa = 0x%016lx\n", part->remote_rp_pa);
 
-       part->remote_vars_pa = remote_vars_pa;
+       part_sn2->remote_vars_pa = remote_vars_pa;
        dev_dbg(xpc_part, "  remote_vars_pa = 0x%016lx\n",
-               part->remote_vars_pa);
+               part_sn2->remote_vars_pa);
 
        part->last_heartbeat = remote_vars->heartbeat;
        dev_dbg(xpc_part, "  last_heartbeat = 0x%016lx\n",
                part->last_heartbeat);
 
-       part->remote_vars_part_pa = remote_vars->vars_part_pa;
+       part_sn2->remote_vars_part_pa = remote_vars->vars_part_pa;
        dev_dbg(xpc_part, "  remote_vars_part_pa = 0x%016lx\n",
-               part->remote_vars_part_pa);
+               part_sn2->remote_vars_part_pa);
 
-       part->remote_act_nasid = remote_vars->act_nasid;
-       dev_dbg(xpc_part, "  remote_act_nasid = 0x%x\n",
-               part->remote_act_nasid);
+       part_sn2->activate_IRQ_nasid = remote_vars->activate_IRQ_nasid;
+       dev_dbg(xpc_part, "  activate_IRQ_nasid = 0x%x\n",
+               part_sn2->activate_IRQ_nasid);
 
-       part->remote_act_phys_cpuid = remote_vars->act_phys_cpuid;
-       dev_dbg(xpc_part, "  remote_act_phys_cpuid = 0x%x\n",
-               part->remote_act_phys_cpuid);
+       part_sn2->activate_IRQ_phys_cpuid =
+           remote_vars->activate_IRQ_phys_cpuid;
+       dev_dbg(xpc_part, "  activate_IRQ_phys_cpuid = 0x%x\n",
+               part_sn2->activate_IRQ_phys_cpuid);
 
-       part->remote_amos_page_pa = remote_vars->amos_page_pa;
+       part_sn2->remote_amos_page_pa = remote_vars->amos_page_pa;
        dev_dbg(xpc_part, "  remote_amos_page_pa = 0x%lx\n",
-               part->remote_amos_page_pa);
+               part_sn2->remote_amos_page_pa);
 
-       part->remote_vars_version = remote_vars->version;
+       part_sn2->remote_vars_version = remote_vars->version;
        dev_dbg(xpc_part, "  remote_vars_version = 0x%x\n",
-               part->remote_vars_version);
+               part_sn2->remote_vars_version);
 }
 
 /*
- * Prior code has determined the nasid which generated an IPI.  Inspect
- * that nasid to determine if its partition needs to be activated or
- * deactivated.
+ * Prior code has determined the nasid which generated a activate IRQ.
+ * Inspect that nasid to determine if its partition needs to be activated
+ * or deactivated.
  *
- * A partition is consider "awaiting activation" if our partition
+ * A partition is considered "awaiting activation" if our partition
  * flags indicate it is not active and it has a heartbeat.  A
  * partition is considered "awaiting deactivation" if our partition
  * flags indicate it is active but it has no heartbeat or it is not
@@ -656,23 +945,23 @@ xpc_update_partition_info_sn2(struct xpc_partition *part, u8 remote_rp_version,
  * initialized reserved page.
  */
 static void
-xpc_identify_act_IRQ_req_sn2(int nasid)
+xpc_identify_activate_IRQ_req_sn2(int nasid)
 {
        struct xpc_rsvd_page *remote_rp;
        struct xpc_vars_sn2 *remote_vars;
-       u64 remote_rp_pa;
-       u64 remote_vars_pa;
+       unsigned long remote_rp_pa;
+       unsigned long remote_vars_pa;
        int remote_rp_version;
        int reactivate = 0;
-       int stamp_diff;
-       struct timespec remote_rp_stamp = { 0, 0 }; /*>>> ZERO_STAMP */
+       unsigned long remote_rp_ts_jiffies = 0;
        short partid;
        struct xpc_partition *part;
+       struct xpc_partition_sn2 *part_sn2;
        enum xp_retval ret;
 
        /* pull over the reserved page structure */
 
-       remote_rp = (struct xpc_rsvd_page *)xpc_remote_copy_buffer;
+       remote_rp = (struct xpc_rsvd_page *)xpc_remote_copy_buffer_sn2;
 
        ret = xpc_get_remote_rp(nasid, NULL, remote_rp, &remote_rp_pa);
        if (ret != xpSuccess) {
@@ -683,19 +972,18 @@ xpc_identify_act_IRQ_req_sn2(int nasid)
 
        remote_vars_pa = remote_rp->sn.vars_pa;
        remote_rp_version = remote_rp->version;
-       if (XPC_SUPPORTS_RP_STAMP(remote_rp_version))
-               remote_rp_stamp = remote_rp->stamp;
+       remote_rp_ts_jiffies = remote_rp->ts_jiffies;
 
        partid = remote_rp->SAL_partid;
        part = &xpc_partitions[partid];
+       part_sn2 = &part->sn.sn2;
 
        /* pull over the cross partition variables */
 
-       remote_vars = (struct xpc_vars_sn2 *)xpc_remote_copy_buffer;
+       remote_vars = (struct xpc_vars_sn2 *)xpc_remote_copy_buffer_sn2;
 
        ret = xpc_get_remote_vars_sn2(remote_vars_pa, remote_vars);
        if (ret != xpSuccess) {
-
                dev_warn(xpc_part, "unable to get XPC variables from nasid %d, "
                         "which sent interrupt, reason=%d\n", nasid, ret);
 
@@ -703,31 +991,26 @@ xpc_identify_act_IRQ_req_sn2(int nasid)
                return;
        }
 
-       part->act_IRQ_rcvd++;
+       part->activate_IRQ_rcvd++;
 
        dev_dbg(xpc_part, "partid for nasid %d is %d; IRQs = %d; HB = "
-               "%ld:0x%lx\n", (int)nasid, (int)partid, part->act_IRQ_rcvd,
+               "%ld:0x%lx\n", (int)nasid, (int)partid, part->activate_IRQ_rcvd,
                remote_vars->heartbeat, remote_vars->heartbeating_to_mask[0]);
 
        if (xpc_partition_disengaged(part) &&
-           part->act_state == XPC_P_INACTIVE) {
+           part->act_state == XPC_P_AS_INACTIVE) {
 
                xpc_update_partition_info_sn2(part, remote_rp_version,
-                                             &remote_rp_stamp, remote_rp_pa,
-                                             remote_vars_pa, remote_vars);
-
-               if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version)) {
-                       if (xpc_partition_disengage_requested(1UL << partid)) {
-                               /*
-                                * Other side is waiting on us to disengage,
-                                * even though we already have.
-                                */
-                               return;
-                       }
+                                             &remote_rp_ts_jiffies,
+                                             remote_rp_pa, remote_vars_pa,
+                                             remote_vars);
 
-               } else {
-                       /* other side doesn't support disengage requests */
-                       xpc_clear_partition_disengage_request(1UL << partid);
+               if (xpc_partition_deactivation_requested_sn2(partid)) {
+                       /*
+                        * Other side is waiting on us to deactivate even though
+                        * we already have.
+                        */
+                       return;
                }
 
                xpc_activate_partition(part);
@@ -735,130 +1018,68 @@ xpc_identify_act_IRQ_req_sn2(int nasid)
        }
 
        DBUG_ON(part->remote_rp_version == 0);
-       DBUG_ON(part->remote_vars_version == 0);
-
-       if (!XPC_SUPPORTS_RP_STAMP(part->remote_rp_version)) {
-               DBUG_ON(XPC_SUPPORTS_DISENGAGE_REQUEST(part->
-                                                      remote_vars_version));
-
-               if (!XPC_SUPPORTS_RP_STAMP(remote_rp_version)) {
-                       DBUG_ON(XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->
-                                                              version));
-                       /* see if the other side rebooted */
-                       if (part->remote_amos_page_pa ==
-                           remote_vars->amos_page_pa &&
-                           xpc_hb_allowed(sn_partition_id,
-                                         &remote_vars->heartbeating_to_mask)) {
-                               /* doesn't look that way, so ignore the IPI */
-                               return;
-                       }
-               }
+       DBUG_ON(part_sn2->remote_vars_version == 0);
 
-               /*
-                * Other side rebooted and previous XPC didn't support the
-                * disengage request, so we don't need to do anything special.
-                */
-
-               xpc_update_partition_info_sn2(part, remote_rp_version,
-                                             &remote_rp_stamp, remote_rp_pa,
-                                             remote_vars_pa, remote_vars);
-               part->reactivate_nasid = nasid;
-               XPC_DEACTIVATE_PARTITION(part, xpReactivating);
-               return;
-       }
+       if (remote_rp_ts_jiffies != part->remote_rp_ts_jiffies) {
 
-       DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version));
+               /* the other side rebooted */
 
-       if (!XPC_SUPPORTS_RP_STAMP(remote_rp_version)) {
-               DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->version));
-
-               /*
-                * Other side rebooted and previous XPC did support the
-                * disengage request, but the new one doesn't.
-                */
-
-               xpc_clear_partition_engaged(1UL << partid);
-               xpc_clear_partition_disengage_request(1UL << partid);
+               DBUG_ON(xpc_partition_engaged_sn2(partid));
+               DBUG_ON(xpc_partition_deactivation_requested_sn2(partid));
 
                xpc_update_partition_info_sn2(part, remote_rp_version,
-                                             &remote_rp_stamp, remote_rp_pa,
-                                             remote_vars_pa, remote_vars);
+                                             &remote_rp_ts_jiffies,
+                                             remote_rp_pa, remote_vars_pa,
+                                             remote_vars);
                reactivate = 1;
-
-       } else {
-               DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->version));
-
-               stamp_diff = xpc_compare_stamps(&part->remote_rp_stamp,
-                                               &remote_rp_stamp);
-               if (stamp_diff != 0) {
-                       DBUG_ON(stamp_diff >= 0);
-
-                       /*
-                        * Other side rebooted and the previous XPC did support
-                        * the disengage request, as does the new one.
-                        */
-
-                       DBUG_ON(xpc_partition_engaged(1UL << partid));
-                       DBUG_ON(xpc_partition_disengage_requested(1UL <<
-                                                                 partid));
-
-                       xpc_update_partition_info_sn2(part, remote_rp_version,
-                                                     &remote_rp_stamp,
-                                                     remote_rp_pa,
-                                                     remote_vars_pa,
-                                                     remote_vars);
-                       reactivate = 1;
-               }
        }
 
-       if (part->disengage_request_timeout > 0 &&
-           !xpc_partition_disengaged(part)) {
+       if (part->disengage_timeout > 0 && !xpc_partition_disengaged(part)) {
                /* still waiting on other side to disengage from us */
                return;
        }
 
-       if (reactivate) {
-               part->reactivate_nasid = nasid;
+       if (reactivate)
                XPC_DEACTIVATE_PARTITION(part, xpReactivating);
-
-       } else if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version) &&
-                  xpc_partition_disengage_requested(1UL << partid)) {
+       else if (xpc_partition_deactivation_requested_sn2(partid))
                XPC_DEACTIVATE_PARTITION(part, xpOtherGoingDown);
-       }
 }
 
 /*
- * Loop through the activation AMO variables and process any bits
+ * Loop through the activation amo variables and process any bits
  * which are set.  Each bit indicates a nasid sending a partition
  * activation or deactivation request.
  *
  * Return #of IRQs detected.
  */
 int
-xpc_identify_act_IRQ_sender_sn2(void)
+xpc_identify_activate_IRQ_sender_sn2(void)
 {
-       int word, bit;
-       u64 nasid_mask;
+       int l;
+       int b;
+       unsigned long nasid_mask_long;
        u64 nasid;              /* remote nasid */
        int n_IRQs_detected = 0;
-       AMO_t *act_amos;
+       struct amo *act_amos;
 
-       act_amos = xpc_vars->amos_page + XPC_ACTIVATE_IRQ_AMOS;
+       act_amos = xpc_vars_sn2->amos_page + XPC_ACTIVATE_IRQ_AMOS_SN2;
 
-       /* scan through act AMO variable looking for non-zero entries */
-       for (word = 0; word < xp_nasid_mask_words; word++) {
+       /* scan through activate amo variables looking for non-zero entries */
+       for (l = 0; l < xpc_nasid_mask_nlongs; l++) {
 
                if (xpc_exiting)
                        break;
 
-               nasid_mask = xpc_IPI_receive_sn2(&act_amos[word]);
-               if (nasid_mask == 0) {
-                       /* no IRQs from nasids in this variable */
+               nasid_mask_long = xpc_receive_IRQ_amo_sn2(&act_amos[l]);
+
+               b = find_first_bit(&nasid_mask_long, BITS_PER_LONG);
+               if (b >= BITS_PER_LONG) {
+                       /* no IRQs from nasids in this amo variable */
                        continue;
                }
 
-               dev_dbg(xpc_part, "AMO[%d] gave back 0x%lx\n", word,
-                       nasid_mask);
+               dev_dbg(xpc_part, "amo[%d] gave back 0x%lx\n", l,
+                       nasid_mask_long);
 
                /*
                 * If this nasid has been added to the machine since
@@ -866,176 +1087,132 @@ xpc_identify_act_IRQ_sender_sn2(void)
                 * remote nasid in our reserved pages machine mask.
                 * This is used in the event of module reload.
                 */
-               xpc_mach_nasids[word] |= nasid_mask;
+               xpc_mach_nasids[l] |= nasid_mask_long;
 
                /* locate the nasid(s) which sent interrupts */
 
-               for (bit = 0; bit < (8 * sizeof(u64)); bit++) {
-                       if (nasid_mask & (1UL << bit)) {
-                               n_IRQs_detected++;
-                               nasid = XPC_NASID_FROM_W_B(word, bit);
-                               dev_dbg(xpc_part, "interrupt from nasid %ld\n",
-                                       nasid);
-                               xpc_identify_act_IRQ_req_sn2(nasid);
-                       }
-               }
+               do {
+                       n_IRQs_detected++;
+                       nasid = (l * BITS_PER_LONG + b) * 2;
+                       dev_dbg(xpc_part, "interrupt from nasid %ld\n", nasid);
+                       xpc_identify_activate_IRQ_req_sn2(nasid);
+
+                       b = find_next_bit(&nasid_mask_long, BITS_PER_LONG,
+                                         b + 1);
+               } while (b < BITS_PER_LONG);
        }
        return n_IRQs_detected;
 }
 
 static void
-xpc_process_act_IRQ_rcvd_sn2(int n_IRQs_expected)
+xpc_process_activate_IRQ_rcvd_sn2(void)
 {
+       unsigned long irq_flags;
+       int n_IRQs_expected;
        int n_IRQs_detected;
 
-       n_IRQs_detected = xpc_identify_act_IRQ_sender_sn2();
+       DBUG_ON(xpc_activate_IRQ_rcvd == 0);
+
+       spin_lock_irqsave(&xpc_activate_IRQ_rcvd_lock, irq_flags);
+       n_IRQs_expected = xpc_activate_IRQ_rcvd;
+       xpc_activate_IRQ_rcvd = 0;
+       spin_unlock_irqrestore(&xpc_activate_IRQ_rcvd_lock, irq_flags);
+
+       n_IRQs_detected = xpc_identify_activate_IRQ_sender_sn2();
        if (n_IRQs_detected < n_IRQs_expected) {
-               /* retry once to help avoid missing AMO */
-               (void)xpc_identify_act_IRQ_sender_sn2();
+               /* retry once to help avoid missing amo */
+               (void)xpc_identify_activate_IRQ_sender_sn2();
        }
 }
 
 /*
- * Setup the infrastructure necessary to support XPartition Communication
- * between the specified remote partition and the local one.
+ * Setup the channel structures that are sn2 specific.
  */
 static enum xp_retval
-xpc_setup_infrastructure_sn2(struct xpc_partition *part)
+xpc_setup_ch_structures_sn_sn2(struct xpc_partition *part)
 {
+       struct xpc_partition_sn2 *part_sn2 = &part->sn.sn2;
+       struct xpc_channel_sn2 *ch_sn2;
        enum xp_retval retval;
        int ret;
        int cpuid;
        int ch_number;
-       struct xpc_channel *ch;
        struct timer_list *timer;
        short partid = XPC_PARTID(part);
 
-       /*
-        * Allocate all of the channel structures as a contiguous chunk of
-        * memory.
-        */
-       DBUG_ON(part->channels != NULL);
-       part->channels = kzalloc(sizeof(struct xpc_channel) * XPC_MAX_NCHANNELS,
-                                GFP_KERNEL);
-       if (part->channels == NULL) {
-               dev_err(xpc_chan, "can't get memory for channels\n");
-               return xpNoMemory;
-       }
-
        /* allocate all the required GET/PUT values */
 
-       part->local_GPs = xpc_kzalloc_cacheline_aligned(XPC_GP_SIZE,
-                                                       GFP_KERNEL,
-                                                       &part->local_GPs_base);
-       if (part->local_GPs == NULL) {
+       part_sn2->local_GPs =
+           xpc_kzalloc_cacheline_aligned(XPC_GP_SIZE, GFP_KERNEL,
+                                         &part_sn2->local_GPs_base);
+       if (part_sn2->local_GPs == NULL) {
                dev_err(xpc_chan, "can't get memory for local get/put "
                        "values\n");
-               retval = xpNoMemory;
-               goto out_1;
+               return xpNoMemory;
        }
 
-       part->remote_GPs = xpc_kzalloc_cacheline_aligned(XPC_GP_SIZE,
-                                                        GFP_KERNEL,
-                                                        &part->
-                                                        remote_GPs_base);
-       if (part->remote_GPs == NULL) {
+       part_sn2->remote_GPs =
+           xpc_kzalloc_cacheline_aligned(XPC_GP_SIZE, GFP_KERNEL,
+                                         &part_sn2->remote_GPs_base);
+       if (part_sn2->remote_GPs == NULL) {
                dev_err(xpc_chan, "can't get memory for remote get/put "
                        "values\n");
                retval = xpNoMemory;
-               goto out_2;
+               goto out_1;
        }
 
-       part->remote_GPs_pa = 0;
+       part_sn2->remote_GPs_pa = 0;
 
        /* allocate all the required open and close args */
 
-       part->local_openclose_args =
-           xpc_kzalloc_cacheline_aligned(XPC_OPENCLOSE_ARGS_SIZE, GFP_KERNEL,
-                                         &part->local_openclose_args_base);
-       if (part->local_openclose_args == NULL) {
+       part_sn2->local_openclose_args =
+           xpc_kzalloc_cacheline_aligned(XPC_OPENCLOSE_ARGS_SIZE,
+                                         GFP_KERNEL, &part_sn2->
+                                         local_openclose_args_base);
+       if (part_sn2->local_openclose_args == NULL) {
                dev_err(xpc_chan, "can't get memory for local connect args\n");
                retval = xpNoMemory;
-               goto out_3;
-       }
-
-       part->remote_openclose_args =
-           xpc_kzalloc_cacheline_aligned(XPC_OPENCLOSE_ARGS_SIZE, GFP_KERNEL,
-                                         &part->remote_openclose_args_base);
-       if (part->remote_openclose_args == NULL) {
-               dev_err(xpc_chan, "can't get memory for remote connect args\n");
-               retval = xpNoMemory;
-               goto out_4;
+               goto out_2;
        }
 
-       part->remote_openclose_args_pa = 0;
+       part_sn2->remote_openclose_args_pa = 0;
 
-       part->local_IPI_amo_va = xpc_IPI_init_sn2(partid);
-       part->local_IPI_amo = 0;
-       spin_lock_init(&part->IPI_lock);
+       part_sn2->local_chctl_amo_va = xpc_init_IRQ_amo_sn2(partid);
 
-       part->remote_IPI_nasid = 0;
-       part->remote_IPI_phys_cpuid = 0;
-       part->remote_IPI_amo_va = NULL;
+       part_sn2->notify_IRQ_nasid = 0;
+       part_sn2->notify_IRQ_phys_cpuid = 0;
+       part_sn2->remote_chctl_amo_va = NULL;
 
-       atomic_set(&part->channel_mgr_requests, 1);
-       init_waitqueue_head(&part->channel_mgr_wq);
-
-       sprintf(part->IPI_owner, "xpc%02d", partid);
-       ret = request_irq(SGI_XPC_NOTIFY, xpc_notify_IRQ_handler, IRQF_SHARED,
-                         part->IPI_owner, (void *)(u64)partid);
+       sprintf(part_sn2->notify_IRQ_owner, "xpc%02d", partid);
+       ret = request_irq(SGI_XPC_NOTIFY, xpc_handle_notify_IRQ_sn2,
+                         IRQF_SHARED, part_sn2->notify_IRQ_owner,
+                         (void *)(u64)partid);
        if (ret != 0) {
                dev_err(xpc_chan, "can't register NOTIFY IRQ handler, "
                        "errno=%d\n", -ret);
                retval = xpLackOfResources;
-               goto out_5;
+               goto out_3;
        }
 
-       /* Setup a timer to check for dropped IPIs */
-       timer = &part->dropped_IPI_timer;
+       /* Setup a timer to check for dropped notify IRQs */
+       timer = &part_sn2->dropped_notify_IRQ_timer;
        init_timer(timer);
-       timer->function = (void (*)(unsigned long))xpc_dropped_IPI_check;
+       timer->function =
+           (void (*)(unsigned long))xpc_check_for_dropped_notify_IRQ_sn2;
        timer->data = (unsigned long)part;
-       timer->expires = jiffies + XPC_P_DROPPED_IPI_WAIT_INTERVAL;
+       timer->expires = jiffies + XPC_DROPPED_NOTIFY_IRQ_WAIT_INTERVAL;
        add_timer(timer);
 
-       part->nchannels = XPC_MAX_NCHANNELS;
-
-       atomic_set(&part->nchannels_active, 0);
-       atomic_set(&part->nchannels_engaged, 0);
-
        for (ch_number = 0; ch_number < part->nchannels; ch_number++) {
-               ch = &part->channels[ch_number];
+               ch_sn2 = &part->channels[ch_number].sn.sn2;
 
-               ch->partid = partid;
-               ch->number = ch_number;
-               ch->flags = XPC_C_DISCONNECTED;
+               ch_sn2->local_GP = &part_sn2->local_GPs[ch_number];
+               ch_sn2->local_openclose_args =
+                   &part_sn2->local_openclose_args[ch_number];
 
-               ch->local_GP = &part->local_GPs[ch_number];
-               ch->local_openclose_args =
-                   &part->local_openclose_args[ch_number];
-
-               atomic_set(&ch->kthreads_assigned, 0);
-               atomic_set(&ch->kthreads_idle, 0);
-               atomic_set(&ch->kthreads_active, 0);
-
-               atomic_set(&ch->references, 0);
-               atomic_set(&ch->n_to_notify, 0);
-
-               spin_lock_init(&ch->lock);
-               mutex_init(&ch->msg_to_pull_mutex);
-               init_completion(&ch->wdisconnect_wait);
-
-               atomic_set(&ch->n_on_msg_allocate_wq, 0);
-               init_waitqueue_head(&ch->msg_allocate_wq);
-               init_waitqueue_head(&ch->idle_wq);
+               mutex_init(&ch_sn2->msg_to_pull_mutex);
        }
 
-       /*
-        * With the setting of the partition setup_state to XPC_P_SETUP, we're
-        * declaring that this partition is ready to go.
-        */
-       part->setup_state = XPC_P_SETUP;
-
        /*
         * Setup the per partition specific variables required by the
         * remote partition to establish channel connections with us.
@@ -1043,113 +1220,84 @@ xpc_setup_infrastructure_sn2(struct xpc_partition *part)
         * The setting of the magic # indicates that these per partition
         * specific variables are ready to be used.
         */
-       xpc_vars_part[partid].GPs_pa = __pa(part->local_GPs);
-       xpc_vars_part[partid].openclose_args_pa =
-           __pa(part->local_openclose_args);
-       xpc_vars_part[partid].IPI_amo_pa = __pa(part->local_IPI_amo_va);
+       xpc_vars_part_sn2[partid].GPs_pa = xp_pa(part_sn2->local_GPs);
+       xpc_vars_part_sn2[partid].openclose_args_pa =
+           xp_pa(part_sn2->local_openclose_args);
+       xpc_vars_part_sn2[partid].chctl_amo_pa =
+           xp_pa(part_sn2->local_chctl_amo_va);
        cpuid = raw_smp_processor_id(); /* any CPU in this partition will do */
-       xpc_vars_part[partid].IPI_nasid = cpuid_to_nasid(cpuid);
-       xpc_vars_part[partid].IPI_phys_cpuid = cpu_physical_id(cpuid);
-       xpc_vars_part[partid].nchannels = part->nchannels;
-       xpc_vars_part[partid].magic = XPC_VP_MAGIC1;
+       xpc_vars_part_sn2[partid].notify_IRQ_nasid = cpuid_to_nasid(cpuid);
+       xpc_vars_part_sn2[partid].notify_IRQ_phys_cpuid =
+           cpu_physical_id(cpuid);
+       xpc_vars_part_sn2[partid].nchannels = part->nchannels;
+       xpc_vars_part_sn2[partid].magic = XPC_VP_MAGIC1_SN2;
 
        return xpSuccess;
 
-       /* setup of infrastructure failed */
-out_5:
-       kfree(part->remote_openclose_args_base);
-       part->remote_openclose_args = NULL;
-out_4:
-       kfree(part->local_openclose_args_base);
-       part->local_openclose_args = NULL;
+       /* setup of ch structures failed */
 out_3:
-       kfree(part->remote_GPs_base);
-       part->remote_GPs = NULL;
+       kfree(part_sn2->local_openclose_args_base);
+       part_sn2->local_openclose_args = NULL;
 out_2:
-       kfree(part->local_GPs_base);
-       part->local_GPs = NULL;
+       kfree(part_sn2->remote_GPs_base);
+       part_sn2->remote_GPs = NULL;
 out_1:
-       kfree(part->channels);
-       part->channels = NULL;
+       kfree(part_sn2->local_GPs_base);
+       part_sn2->local_GPs = NULL;
        return retval;
 }
 
 /*
- * Teardown the infrastructure necessary to support XPartition Communication
- * between the specified remote partition and the local one.
+ * Teardown the channel structures that are sn2 specific.
  */
 static void
-xpc_teardown_infrastructure_sn2(struct xpc_partition *part)
+xpc_teardown_ch_structures_sn_sn2(struct xpc_partition *part)
 {
+       struct xpc_partition_sn2 *part_sn2 = &part->sn.sn2;
        short partid = XPC_PARTID(part);
 
        /*
-        * We start off by making this partition inaccessible to local
-        * processes by marking it as no longer setup. Then we make it
-        * inaccessible to remote processes by clearing the XPC per partition
-        * specific variable's magic # (which indicates that these variables
-        * are no longer valid) and by ignoring all XPC notify IPIs sent to
-        * this partition.
-        */
-
-       DBUG_ON(atomic_read(&part->nchannels_engaged) != 0);
-       DBUG_ON(atomic_read(&part->nchannels_active) != 0);
-       DBUG_ON(part->setup_state != XPC_P_SETUP);
-       part->setup_state = XPC_P_WTEARDOWN;
-
-       xpc_vars_part[partid].magic = 0;
-
-       free_irq(SGI_XPC_NOTIFY, (void *)(u64)partid);
-
-       /*
-        * Before proceeding with the teardown we have to wait until all
-        * existing references cease.
+        * Indicate that the variables specific to the remote partition are no
+        * longer available for its use.
         */
-       wait_event(part->teardown_wq, (atomic_read(&part->references) == 0));
-
-       /* now we can begin tearing down the infrastructure */
-
-       part->setup_state = XPC_P_TORNDOWN;
+       xpc_vars_part_sn2[partid].magic = 0;
 
        /* in case we've still got outstanding timers registered... */
-       del_timer_sync(&part->dropped_IPI_timer);
+       del_timer_sync(&part_sn2->dropped_notify_IRQ_timer);
+       free_irq(SGI_XPC_NOTIFY, (void *)(u64)partid);
 
-       kfree(part->remote_openclose_args_base);
-       part->remote_openclose_args = NULL;
-       kfree(part->local_openclose_args_base);
-       part->local_openclose_args = NULL;
-       kfree(part->remote_GPs_base);
-       part->remote_GPs = NULL;
-       kfree(part->local_GPs_base);
-       part->local_GPs = NULL;
-       kfree(part->channels);
-       part->channels = NULL;
-       part->local_IPI_amo_va = NULL;
+       kfree(part_sn2->local_openclose_args_base);
+       part_sn2->local_openclose_args = NULL;
+       kfree(part_sn2->remote_GPs_base);
+       part_sn2->remote_GPs = NULL;
+       kfree(part_sn2->local_GPs_base);
+       part_sn2->local_GPs = NULL;
+       part_sn2->local_chctl_amo_va = NULL;
 }
 
 /*
  * Create a wrapper that hides the underlying mechanism for pulling a cacheline
  * (or multiple cachelines) from a remote partition.
  *
- * src must be a cacheline aligned physical address on the remote partition.
+ * src_pa must be a cacheline aligned physical address on the remote partition.
  * dst must be a cacheline aligned virtual address on this partition.
  * cnt must be cacheline sized
  */
-/* >>> Replace this function by call to xp_remote_memcpy() or bte_copy()? */
+/* ??? Replace this function by call to xp_remote_memcpy() or bte_copy()? */
 static enum xp_retval
 xpc_pull_remote_cachelines_sn2(struct xpc_partition *part, void *dst,
-                              const void *src, size_t cnt)
+                              const unsigned long src_pa, size_t cnt)
 {
        enum xp_retval ret;
 
-       DBUG_ON((u64)src != L1_CACHE_ALIGN((u64)src));
-       DBUG_ON((u64)dst != L1_CACHE_ALIGN((u64)dst));
+       DBUG_ON(src_pa != L1_CACHE_ALIGN(src_pa));
+       DBUG_ON((unsigned long)dst != L1_CACHE_ALIGN((unsigned long)dst));
        DBUG_ON(cnt != L1_CACHE_ALIGN(cnt));
 
-       if (part->act_state == XPC_P_DEACTIVATING)
+       if (part->act_state == XPC_P_AS_DEACTIVATING)
                return part->reason;
 
-       ret = xp_remote_memcpy(dst, src, cnt);
+       ret = xp_remote_memcpy(xp_pa(dst), src_pa, cnt);
        if (ret != xpSuccess) {
                dev_dbg(xpc_chan, "xp_remote_memcpy() from partition %d failed,"
                        " ret=%d\n", XPC_PARTID(part), ret);
@@ -1164,21 +1312,23 @@ xpc_pull_remote_cachelines_sn2(struct xpc_partition *part, void *dst,
 static enum xp_retval
 xpc_pull_remote_vars_part_sn2(struct xpc_partition *part)
 {
+       struct xpc_partition_sn2 *part_sn2 = &part->sn.sn2;
        u8 buffer[L1_CACHE_BYTES * 2];
        struct xpc_vars_part_sn2 *pulled_entry_cacheline =
            (struct xpc_vars_part_sn2 *)L1_CACHE_ALIGN((u64)buffer);
        struct xpc_vars_part_sn2 *pulled_entry;
-       u64 remote_entry_cacheline_pa, remote_entry_pa;
+       unsigned long remote_entry_cacheline_pa;
+       unsigned long remote_entry_pa;
        short partid = XPC_PARTID(part);
        enum xp_retval ret;
 
        /* pull the cacheline that contains the variables we're interested in */
 
-       DBUG_ON(part->remote_vars_part_pa !=
-               L1_CACHE_ALIGN(part->remote_vars_part_pa));
+       DBUG_ON(part_sn2->remote_vars_part_pa !=
+               L1_CACHE_ALIGN(part_sn2->remote_vars_part_pa));
        DBUG_ON(sizeof(struct xpc_vars_part_sn2) != L1_CACHE_BYTES / 2);
 
-       remote_entry_pa = part->remote_vars_part_pa +
+       remote_entry_pa = part_sn2->remote_vars_part_pa +
            sn_partition_id * sizeof(struct xpc_vars_part_sn2);
 
        remote_entry_cacheline_pa = (remote_entry_pa & ~(L1_CACHE_BYTES - 1));
@@ -1188,7 +1338,7 @@ xpc_pull_remote_vars_part_sn2(struct xpc_partition *part)
                                                    (L1_CACHE_BYTES - 1)));
 
        ret = xpc_pull_remote_cachelines_sn2(part, pulled_entry_cacheline,
-                                            (void *)remote_entry_cacheline_pa,
+                                            remote_entry_cacheline_pa,
                                             L1_CACHE_BYTES);
        if (ret != xpSuccess) {
                dev_dbg(xpc_chan, "failed to pull XPC vars_part from "
@@ -1198,8 +1348,8 @@ xpc_pull_remote_vars_part_sn2(struct xpc_partition *part)
 
        /* see if they've been set up yet */
 
-       if (pulled_entry->magic != XPC_VP_MAGIC1 &&
-           pulled_entry->magic != XPC_VP_MAGIC2) {
+       if (pulled_entry->magic != XPC_VP_MAGIC1_SN2 &&
+           pulled_entry->magic != XPC_VP_MAGIC2_SN2) {
 
                if (pulled_entry->magic != 0) {
                        dev_dbg(xpc_chan, "partition %d's XPC vars_part for "
@@ -1212,13 +1362,13 @@ xpc_pull_remote_vars_part_sn2(struct xpc_partition *part)
                return xpRetry;
        }
 
-       if (xpc_vars_part[partid].magic == XPC_VP_MAGIC1) {
+       if (xpc_vars_part_sn2[partid].magic == XPC_VP_MAGIC1_SN2) {
 
                /* validate the variables */
 
                if (pulled_entry->GPs_pa == 0 ||
                    pulled_entry->openclose_args_pa == 0 ||
-                   pulled_entry->IPI_amo_pa == 0) {
+                   pulled_entry->chctl_amo_pa == 0) {
 
                        dev_err(xpc_chan, "partition %d's XPC vars_part for "
                                "partition %d are not valid\n", partid,
@@ -1228,23 +1378,24 @@ xpc_pull_remote_vars_part_sn2(struct xpc_partition *part)
 
                /* the variables we imported look to be valid */
 
-               part->remote_GPs_pa = pulled_entry->GPs_pa;
-               part->remote_openclose_args_pa =
+               part_sn2->remote_GPs_pa = pulled_entry->GPs_pa;
+               part_sn2->remote_openclose_args_pa =
                    pulled_entry->openclose_args_pa;
-               part->remote_IPI_amo_va =
-                   (AMO_t *)__va(pulled_entry->IPI_amo_pa);
-               part->remote_IPI_nasid = pulled_entry->IPI_nasid;
-               part->remote_IPI_phys_cpuid = pulled_entry->IPI_phys_cpuid;
+               part_sn2->remote_chctl_amo_va =
+                   (struct amo *)__va(pulled_entry->chctl_amo_pa);
+               part_sn2->notify_IRQ_nasid = pulled_entry->notify_IRQ_nasid;
+               part_sn2->notify_IRQ_phys_cpuid =
+                   pulled_entry->notify_IRQ_phys_cpuid;
 
                if (part->nchannels > pulled_entry->nchannels)
                        part->nchannels = pulled_entry->nchannels;
 
                /* let the other side know that we've pulled their variables */
 
-               xpc_vars_part[partid].magic = XPC_VP_MAGIC2;
+               xpc_vars_part_sn2[partid].magic = XPC_VP_MAGIC2_SN2;
        }
 
-       if (pulled_entry->magic == XPC_VP_MAGIC1)
+       if (pulled_entry->magic == XPC_VP_MAGIC1_SN2)
                return xpRetry;
 
        return xpSuccess;
@@ -1258,10 +1409,11 @@ xpc_pull_remote_vars_part_sn2(struct xpc_partition *part)
 static enum xp_retval
 xpc_make_first_contact_sn2(struct xpc_partition *part)
 {
+       struct xpc_partition_sn2 *part_sn2 = &part->sn.sn2;
        enum xp_retval ret;
 
        /*
-        * Register the remote partition's AMOs with SAL so it can handle
+        * 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
@@ -1270,7 +1422,7 @@ xpc_make_first_contact_sn2(struct xpc_partition *part)
         * 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,
+       if (sn_register_xp_addr_region(part_sn2->remote_amos_page_pa,
                                       PAGE_SIZE, 1) < 0) {
                dev_warn(xpc_part, "xpc_activating(%d) failed to register "
                         "xp_addr region\n", XPC_PARTID(part));
@@ -1280,7 +1432,14 @@ xpc_make_first_contact_sn2(struct xpc_partition *part)
                return ret;
        }
 
-       xpc_IPI_send_activated(part);
+       /*
+        * Send activate IRQ to get other side to activate if they've not
+        * already begun to do so.
+        */
+       xpc_send_activate_IRQ_sn2(part_sn2->remote_amos_page_pa,
+                                 cnodeid_to_nasid(0),
+                                 part_sn2->activate_IRQ_nasid,
+                                 part_sn2->activate_IRQ_phys_cpuid);
 
        while ((ret = xpc_pull_remote_vars_part_sn2(part)) != xpSuccess) {
                if (ret != xpRetry) {
@@ -1294,7 +1453,7 @@ xpc_make_first_contact_sn2(struct xpc_partition *part)
                /* wait a 1/4 of a second or so */
                (void)msleep_interruptible(250);
 
-               if (part->act_state == XPC_P_DEACTIVATING)
+               if (part->act_state == XPC_P_AS_DEACTIVATING)
                        return part->reason;
        }
 
@@ -1302,30 +1461,31 @@ xpc_make_first_contact_sn2(struct xpc_partition *part)
 }
 
 /*
- * Get the IPI flags and pull the openclose args and/or remote GPs as needed.
+ * Get the chctl flags and pull the openclose args and/or remote GPs as needed.
  */
 static u64
-xpc_get_IPI_flags_sn2(struct xpc_partition *part)
+xpc_get_chctl_all_flags_sn2(struct xpc_partition *part)
 {
+       struct xpc_partition_sn2 *part_sn2 = &part->sn.sn2;
        unsigned long irq_flags;
-       u64 IPI_amo;
+       union xpc_channel_ctl_flags chctl;
        enum xp_retval ret;
 
        /*
-        * See if there are any IPI flags to be handled.
+        * See if there are any chctl flags to be handled.
         */
 
-       spin_lock_irqsave(&part->IPI_lock, irq_flags);
-       IPI_amo = part->local_IPI_amo;
-       if (IPI_amo != 0)
-               part->local_IPI_amo = 0;
+       spin_lock_irqsave(&part->chctl_lock, irq_flags);
+       chctl = part->chctl;
+       if (chctl.all_flags != 0)
+               part->chctl.all_flags = 0;
 
-       spin_unlock_irqrestore(&part->IPI_lock, irq_flags);
+       spin_unlock_irqrestore(&part->chctl_lock, irq_flags);
 
-       if (XPC_ANY_OPENCLOSE_IPI_FLAGS_SET(IPI_amo)) {
-               ret = xpc_pull_remote_cachelines_sn2(part,
-                                                   part->remote_openclose_args,
-                                                    (void *)part->
+       if (xpc_any_openclose_chctl_flags_set(&chctl)) {
+               ret = xpc_pull_remote_cachelines_sn2(part, part->
+                                                    remote_openclose_args,
+                                                    part_sn2->
                                                     remote_openclose_args_pa,
                                                     XPC_OPENCLOSE_ARGS_SIZE);
                if (ret != xpSuccess) {
@@ -1335,14 +1495,14 @@ xpc_get_IPI_flags_sn2(struct xpc_partition *part)
                                "partition %d, ret=%d\n", XPC_PARTID(part),
                                ret);
 
-                       /* don't bother processing IPIs anymore */
-                       IPI_amo = 0;
+                       /* don't bother processing chctl flags anymore */
+                       chctl.all_flags = 0;
                }
        }
 
-       if (XPC_ANY_MSG_IPI_FLAGS_SET(IPI_amo)) {
-               ret = xpc_pull_remote_cachelines_sn2(part, part->remote_GPs,
-                                                   (void *)part->remote_GPs_pa,
+       if (xpc_any_msg_chctl_flags_set(&chctl)) {
+               ret = xpc_pull_remote_cachelines_sn2(part, part_sn2->remote_GPs,
+                                                    part_sn2->remote_GPs_pa,
                                                     XPC_GP_SIZE);
                if (ret != xpSuccess) {
                        XPC_DEACTIVATE_PARTITION(part, ret);
@@ -1350,89 +1510,437 @@ xpc_get_IPI_flags_sn2(struct xpc_partition *part)
                        dev_dbg(xpc_chan, "failed to pull GPs from partition "
                                "%d, ret=%d\n", XPC_PARTID(part), ret);
 
-                       /* don't bother processing IPIs anymore */
-                       IPI_amo = 0;
+                       /* don't bother processing chctl flags anymore */
+                       chctl.all_flags = 0;
+               }
+       }
+
+       return chctl.all_flags;
+}
+
+/*
+ * Allocate the local message queue and the notify queue.
+ */
+static enum xp_retval
+xpc_allocate_local_msgqueue_sn2(struct xpc_channel *ch)
+{
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       unsigned long irq_flags;
+       int nentries;
+       size_t nbytes;
+
+       for (nentries = ch->local_nentries; nentries > 0; nentries--) {
+
+               nbytes = nentries * ch->entry_size;
+               ch_sn2->local_msgqueue =
+                   xpc_kzalloc_cacheline_aligned(nbytes, GFP_KERNEL,
+                                                 &ch_sn2->local_msgqueue_base);
+               if (ch_sn2->local_msgqueue == NULL)
+                       continue;
+
+               nbytes = nentries * sizeof(struct xpc_notify_sn2);
+               ch_sn2->notify_queue = kzalloc(nbytes, GFP_KERNEL);
+               if (ch_sn2->notify_queue == NULL) {
+                       kfree(ch_sn2->local_msgqueue_base);
+                       ch_sn2->local_msgqueue = NULL;
+                       continue;
+               }
+
+               spin_lock_irqsave(&ch->lock, irq_flags);
+               if (nentries < ch->local_nentries) {
+                       dev_dbg(xpc_chan, "nentries=%d local_nentries=%d, "
+                               "partid=%d, channel=%d\n", nentries,
+                               ch->local_nentries, ch->partid, ch->number);
+
+                       ch->local_nentries = nentries;
+               }
+               spin_unlock_irqrestore(&ch->lock, irq_flags);
+               return xpSuccess;
+       }
+
+       dev_dbg(xpc_chan, "can't get memory for local message queue and notify "
+               "queue, partid=%d, channel=%d\n", ch->partid, ch->number);
+       return xpNoMemory;
+}
+
+/*
+ * Allocate the cached remote message queue.
+ */
+static enum xp_retval
+xpc_allocate_remote_msgqueue_sn2(struct xpc_channel *ch)
+{
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       unsigned long irq_flags;
+       int nentries;
+       size_t nbytes;
+
+       DBUG_ON(ch->remote_nentries <= 0);
+
+       for (nentries = ch->remote_nentries; nentries > 0; nentries--) {
+
+               nbytes = nentries * ch->entry_size;
+               ch_sn2->remote_msgqueue =
+                   xpc_kzalloc_cacheline_aligned(nbytes, GFP_KERNEL, &ch_sn2->
+                                                 remote_msgqueue_base);
+               if (ch_sn2->remote_msgqueue == NULL)
+                       continue;
+
+               spin_lock_irqsave(&ch->lock, irq_flags);
+               if (nentries < ch->remote_nentries) {
+                       dev_dbg(xpc_chan, "nentries=%d remote_nentries=%d, "
+                               "partid=%d, channel=%d\n", nentries,
+                               ch->remote_nentries, ch->partid, ch->number);
+
+                       ch->remote_nentries = nentries;
+               }
+               spin_unlock_irqrestore(&ch->lock, irq_flags);
+               return xpSuccess;
+       }
+
+       dev_dbg(xpc_chan, "can't get memory for cached remote message queue, "
+               "partid=%d, channel=%d\n", ch->partid, ch->number);
+       return xpNoMemory;
+}
+
+/*
+ * Allocate message queues and other stuff associated with a channel.
+ *
+ * Note: Assumes all of the channel sizes are filled in.
+ */
+static enum xp_retval
+xpc_setup_msg_structures_sn2(struct xpc_channel *ch)
+{
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       enum xp_retval ret;
+
+       DBUG_ON(ch->flags & XPC_C_SETUP);
+
+       ret = xpc_allocate_local_msgqueue_sn2(ch);
+       if (ret == xpSuccess) {
+
+               ret = xpc_allocate_remote_msgqueue_sn2(ch);
+               if (ret != xpSuccess) {
+                       kfree(ch_sn2->local_msgqueue_base);
+                       ch_sn2->local_msgqueue = NULL;
+                       kfree(ch_sn2->notify_queue);
+                       ch_sn2->notify_queue = NULL;
+               }
+       }
+       return ret;
+}
+
+/*
+ * Free up message queues and other stuff that were allocated for the specified
+ * channel.
+ */
+static void
+xpc_teardown_msg_structures_sn2(struct xpc_channel *ch)
+{
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+
+       DBUG_ON(!spin_is_locked(&ch->lock));
+
+       ch_sn2->remote_msgqueue_pa = 0;
+
+       ch_sn2->local_GP->get = 0;
+       ch_sn2->local_GP->put = 0;
+       ch_sn2->remote_GP.get = 0;
+       ch_sn2->remote_GP.put = 0;
+       ch_sn2->w_local_GP.get = 0;
+       ch_sn2->w_local_GP.put = 0;
+       ch_sn2->w_remote_GP.get = 0;
+       ch_sn2->w_remote_GP.put = 0;
+       ch_sn2->next_msg_to_pull = 0;
+
+       if (ch->flags & XPC_C_SETUP) {
+               dev_dbg(xpc_chan, "ch->flags=0x%x, partid=%d, channel=%d\n",
+                       ch->flags, ch->partid, ch->number);
+
+               kfree(ch_sn2->local_msgqueue_base);
+               ch_sn2->local_msgqueue = NULL;
+               kfree(ch_sn2->remote_msgqueue_base);
+               ch_sn2->remote_msgqueue = NULL;
+               kfree(ch_sn2->notify_queue);
+               ch_sn2->notify_queue = NULL;
+       }
+}
+
+/*
+ * Notify those who wanted to be notified upon delivery of their message.
+ */
+static void
+xpc_notify_senders_sn2(struct xpc_channel *ch, enum xp_retval reason, s64 put)
+{
+       struct xpc_notify_sn2 *notify;
+       u8 notify_type;
+       s64 get = ch->sn.sn2.w_remote_GP.get - 1;
+
+       while (++get < put && atomic_read(&ch->n_to_notify) > 0) {
+
+               notify = &ch->sn.sn2.notify_queue[get % ch->local_nentries];
+
+               /*
+                * See if the notify entry indicates it was associated with
+                * a message who's sender wants to be notified. It is possible
+                * that it is, but someone else is doing or has done the
+                * notification.
+                */
+               notify_type = notify->type;
+               if (notify_type == 0 ||
+                   cmpxchg(&notify->type, notify_type, 0) != notify_type) {
+                       continue;
+               }
+
+               DBUG_ON(notify_type != XPC_N_CALL);
+
+               atomic_dec(&ch->n_to_notify);
+
+               if (notify->func != NULL) {
+                       dev_dbg(xpc_chan, "notify->func() called, notify=0x%p "
+                               "msg_number=%ld partid=%d channel=%d\n",
+                               (void *)notify, get, ch->partid, ch->number);
+
+                       notify->func(reason, ch->partid, ch->number,
+                                    notify->key);
+
+                       dev_dbg(xpc_chan, "notify->func() returned, notify=0x%p"
+                               " msg_number=%ld partid=%d channel=%d\n",
+                               (void *)notify, get, ch->partid, ch->number);
+               }
+       }
+}
+
+static void
+xpc_notify_senders_of_disconnect_sn2(struct xpc_channel *ch)
+{
+       xpc_notify_senders_sn2(ch, ch->reason, ch->sn.sn2.w_local_GP.put);
+}
+
+/*
+ * Clear some of the msg flags in the local message queue.
+ */
+static inline void
+xpc_clear_local_msgqueue_flags_sn2(struct xpc_channel *ch)
+{
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       struct xpc_msg_sn2 *msg;
+       s64 get;
+
+       get = ch_sn2->w_remote_GP.get;
+       do {
+               msg = (struct xpc_msg_sn2 *)((u64)ch_sn2->local_msgqueue +
+                                            (get % ch->local_nentries) *
+                                            ch->entry_size);
+               msg->flags = 0;
+       } while (++get < ch_sn2->remote_GP.get);
+}
+
+/*
+ * Clear some of the msg flags in the remote message queue.
+ */
+static inline void
+xpc_clear_remote_msgqueue_flags_sn2(struct xpc_channel *ch)
+{
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       struct xpc_msg_sn2 *msg;
+       s64 put;
+
+       put = ch_sn2->w_remote_GP.put;
+       do {
+               msg = (struct xpc_msg_sn2 *)((u64)ch_sn2->remote_msgqueue +
+                                            (put % ch->remote_nentries) *
+                                            ch->entry_size);
+               msg->flags = 0;
+       } while (++put < ch_sn2->remote_GP.put);
+}
+
+static int
+xpc_n_of_deliverable_payloads_sn2(struct xpc_channel *ch)
+{
+       return ch->sn.sn2.w_remote_GP.put - ch->sn.sn2.w_local_GP.get;
+}
+
+static void
+xpc_process_msg_chctl_flags_sn2(struct xpc_partition *part, int ch_number)
+{
+       struct xpc_channel *ch = &part->channels[ch_number];
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       int npayloads_sent;
+
+       ch_sn2->remote_GP = part->sn.sn2.remote_GPs[ch_number];
+
+       /* See what, if anything, has changed for each connected channel */
+
+       xpc_msgqueue_ref(ch);
+
+       if (ch_sn2->w_remote_GP.get == ch_sn2->remote_GP.get &&
+           ch_sn2->w_remote_GP.put == ch_sn2->remote_GP.put) {
+               /* nothing changed since GPs were last pulled */
+               xpc_msgqueue_deref(ch);
+               return;
+       }
+
+       if (!(ch->flags & XPC_C_CONNECTED)) {
+               xpc_msgqueue_deref(ch);
+               return;
+       }
+
+       /*
+        * First check to see if messages recently sent by us have been
+        * received by the other side. (The remote GET value will have
+        * changed since we last looked at it.)
+        */
+
+       if (ch_sn2->w_remote_GP.get != ch_sn2->remote_GP.get) {
+
+               /*
+                * We need to notify any senders that want to be notified
+                * that their sent messages have been received by their
+                * intended recipients. We need to do this before updating
+                * w_remote_GP.get so that we don't allocate the same message
+                * queue entries prematurely (see xpc_allocate_msg()).
+                */
+               if (atomic_read(&ch->n_to_notify) > 0) {
+                       /*
+                        * Notify senders that messages sent have been
+                        * received and delivered by the other side.
+                        */
+                       xpc_notify_senders_sn2(ch, xpMsgDelivered,
+                                              ch_sn2->remote_GP.get);
+               }
+
+               /*
+                * Clear msg->flags in previously sent messages, so that
+                * they're ready for xpc_allocate_msg().
+                */
+               xpc_clear_local_msgqueue_flags_sn2(ch);
+
+               ch_sn2->w_remote_GP.get = ch_sn2->remote_GP.get;
+
+               dev_dbg(xpc_chan, "w_remote_GP.get changed to %ld, partid=%d, "
+                       "channel=%d\n", ch_sn2->w_remote_GP.get, ch->partid,
+                       ch->number);
+
+               /*
+                * If anyone was waiting for message queue entries to become
+                * available, wake them up.
+                */
+               if (atomic_read(&ch->n_on_msg_allocate_wq) > 0)
+                       wake_up(&ch->msg_allocate_wq);
+       }
+
+       /*
+        * Now check for newly sent messages by the other side. (The remote
+        * PUT value will have changed since we last looked at it.)
+        */
+
+       if (ch_sn2->w_remote_GP.put != ch_sn2->remote_GP.put) {
+               /*
+                * Clear msg->flags in previously received messages, so that
+                * they're ready for xpc_get_deliverable_payload_sn2().
+                */
+               xpc_clear_remote_msgqueue_flags_sn2(ch);
+
+               ch_sn2->w_remote_GP.put = ch_sn2->remote_GP.put;
+
+               dev_dbg(xpc_chan, "w_remote_GP.put changed to %ld, partid=%d, "
+                       "channel=%d\n", ch_sn2->w_remote_GP.put, ch->partid,
+                       ch->number);
+
+               npayloads_sent = xpc_n_of_deliverable_payloads_sn2(ch);
+               if (npayloads_sent > 0) {
+                       dev_dbg(xpc_chan, "msgs waiting to be copied and "
+                               "delivered=%d, partid=%d, channel=%d\n",
+                               npayloads_sent, ch->partid, ch->number);
+
+                       if (ch->flags & XPC_C_CONNECTEDCALLOUT_MADE)
+                               xpc_activate_kthreads(ch, npayloads_sent);
                }
        }
 
-       return IPI_amo;
+       xpc_msgqueue_deref(ch);
 }
 
-static struct xpc_msg *
+static struct xpc_msg_sn2 *
 xpc_pull_remote_msg_sn2(struct xpc_channel *ch, s64 get)
 {
        struct xpc_partition *part = &xpc_partitions[ch->partid];
-       struct xpc_msg *remote_msg, *msg;
-       u32 msg_index, nmsgs;
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       unsigned long remote_msg_pa;
+       struct xpc_msg_sn2 *msg;
+       u32 msg_index;
+       u32 nmsgs;
        u64 msg_offset;
        enum xp_retval ret;
 
-       if (mutex_lock_interruptible(&ch->msg_to_pull_mutex) != 0) {
+       if (mutex_lock_interruptible(&ch_sn2->msg_to_pull_mutex) != 0) {
                /* we were interrupted by a signal */
                return NULL;
        }
 
-       while (get >= ch->next_msg_to_pull) {
+       while (get >= ch_sn2->next_msg_to_pull) {
 
                /* pull as many messages as are ready and able to be pulled */
 
-               msg_index = ch->next_msg_to_pull % ch->remote_nentries;
+               msg_index = ch_sn2->next_msg_to_pull % ch->remote_nentries;
 
-               DBUG_ON(ch->next_msg_to_pull >= ch->w_remote_GP.put);
-               nmsgs = ch->w_remote_GP.put - ch->next_msg_to_pull;
+               DBUG_ON(ch_sn2->next_msg_to_pull >= ch_sn2->w_remote_GP.put);
+               nmsgs = ch_sn2->w_remote_GP.put - ch_sn2->next_msg_to_pull;
                if (msg_index + nmsgs > ch->remote_nentries) {
                        /* ignore the ones that wrap the msg queue for now */
                        nmsgs = ch->remote_nentries - msg_index;
                }
 
-               msg_offset = msg_index * ch->msg_size;
-               msg = (struct xpc_msg *)((u64)ch->remote_msgqueue + msg_offset);
-               remote_msg = (struct xpc_msg *)(ch->remote_msgqueue_pa +
-                                               msg_offset);
+               msg_offset = msg_index * ch->entry_size;
+               msg = (struct xpc_msg_sn2 *)((u64)ch_sn2->remote_msgqueue +
+                   msg_offset);
+               remote_msg_pa = ch_sn2->remote_msgqueue_pa + msg_offset;
 
-               ret = xpc_pull_remote_cachelines_sn2(part, msg, remote_msg,
-                                                    nmsgs * ch->msg_size);
+               ret = xpc_pull_remote_cachelines_sn2(part, msg, remote_msg_pa,
+                                                    nmsgs * ch->entry_size);
                if (ret != xpSuccess) {
 
                        dev_dbg(xpc_chan, "failed to pull %d msgs starting with"
                                " msg %ld from partition %d, channel=%d, "
-                               "ret=%d\n", nmsgs, ch->next_msg_to_pull,
+                               "ret=%d\n", nmsgs, ch_sn2->next_msg_to_pull,
                                ch->partid, ch->number, ret);
 
                        XPC_DEACTIVATE_PARTITION(part, ret);
 
-                       mutex_unlock(&ch->msg_to_pull_mutex);
+                       mutex_unlock(&ch_sn2->msg_to_pull_mutex);
                        return NULL;
                }
 
-               ch->next_msg_to_pull += nmsgs;
+               ch_sn2->next_msg_to_pull += nmsgs;
        }
 
-       mutex_unlock(&ch->msg_to_pull_mutex);
+       mutex_unlock(&ch_sn2->msg_to_pull_mutex);
 
        /* return the message we were looking for */
-       msg_offset = (get % ch->remote_nentries) * ch->msg_size;
-       msg = (struct xpc_msg *)((u64)ch->remote_msgqueue + msg_offset);
+       msg_offset = (get % ch->remote_nentries) * ch->entry_size;
+       msg = (struct xpc_msg_sn2 *)((u64)ch_sn2->remote_msgqueue + msg_offset);
 
        return msg;
 }
 
 /*
- * Get a message to be delivered.
+ * Get the next deliverable message's payload.
  */
-static struct xpc_msg *
-xpc_get_deliverable_msg_sn2(struct xpc_channel *ch)
+static void *
+xpc_get_deliverable_payload_sn2(struct xpc_channel *ch)
 {
-       struct xpc_msg *msg = NULL;
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       struct xpc_msg_sn2 *msg;
+       void *payload = NULL;
        s64 get;
 
        do {
                if (ch->flags & XPC_C_DISCONNECTING)
                        break;
 
-               get = ch->w_local_GP.get;
+               get = ch_sn2->w_local_GP.get;
                rmb();  /* guarantee that .get loads before .put */
-               if (get == ch->w_remote_GP.put)
+               if (get == ch_sn2->w_remote_GP.put)
                        break;
 
                /* There are messages waiting to be pulled and delivered.
@@ -1442,7 +1950,7 @@ xpc_get_deliverable_msg_sn2(struct xpc_channel *ch)
                 * to try again for the next one.
                 */
 
-               if (cmpxchg(&ch->w_local_GP.get, get, get + 1) == get) {
+               if (cmpxchg(&ch_sn2->w_local_GP.get, get, get + 1) == get) {
                        /* we got the entry referenced by get */
 
                        dev_dbg(xpc_chan, "w_local_GP.get changed to %ld, "
@@ -1454,40 +1962,43 @@ xpc_get_deliverable_msg_sn2(struct xpc_channel *ch)
                        msg = xpc_pull_remote_msg_sn2(ch, get);
 
                        DBUG_ON(msg != NULL && msg->number != get);
-                       DBUG_ON(msg != NULL && (msg->flags & XPC_M_DONE));
-                       DBUG_ON(msg != NULL && !(msg->flags & XPC_M_READY));
+                       DBUG_ON(msg != NULL && (msg->flags & XPC_M_SN2_DONE));
+                       DBUG_ON(msg != NULL && !(msg->flags & XPC_M_SN2_READY));
 
+                       payload = &msg->payload;
                        break;
                }
 
        } while (1);
 
-       return msg;
+       return payload;
 }
 
 /*
  * Now we actually send the messages that are ready to be sent by advancing
- * the local message queue's Put value and then send an IPI to the recipient
- * partition.
+ * the local message queue's Put value and then send a chctl msgrequest to the
+ * recipient partition.
  */
 static void
 xpc_send_msgs_sn2(struct xpc_channel *ch, s64 initial_put)
 {
-       struct xpc_msg *msg;
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       struct xpc_msg_sn2 *msg;
        s64 put = initial_put + 1;
-       int send_IPI = 0;
+       int send_msgrequest = 0;
 
        while (1) {
 
                while (1) {
-                       if (put == ch->w_local_GP.put)
+                       if (put == ch_sn2->w_local_GP.put)
                                break;
 
-                       msg = (struct xpc_msg *)((u64)ch->local_msgqueue +
-                                                (put % ch->local_nentries) *
-                                                ch->msg_size);
+                       msg = (struct xpc_msg_sn2 *)((u64)ch_sn2->
+                                                    local_msgqueue + (put %
+                                                    ch->local_nentries) *
+                                                    ch->entry_size);
 
-                       if (!(msg->flags & XPC_M_READY))
+                       if (!(msg->flags & XPC_M_SN2_READY))
                                break;
 
                        put++;
@@ -1498,10 +2009,10 @@ xpc_send_msgs_sn2(struct xpc_channel *ch, s64 initial_put)
                        break;
                }
 
-               if (cmpxchg_rel(&ch->local_GP->put, initial_put, put) !=
+               if (cmpxchg_rel(&ch_sn2->local_GP->put, initial_put, put) !=
                    initial_put) {
                        /* someone else beat us to it */
-                       DBUG_ON(ch->local_GP->put < initial_put);
+                       DBUG_ON(ch_sn2->local_GP->put < initial_put);
                        break;
                }
 
@@ -1510,18 +2021,18 @@ xpc_send_msgs_sn2(struct xpc_channel *ch, s64 initial_put)
                dev_dbg(xpc_chan, "local_GP->put changed to %ld, partid=%d, "
                        "channel=%d\n", put, ch->partid, ch->number);
 
-               send_IPI = 1;
+               send_msgrequest = 1;
 
                /*
                 * We need to ensure that the message referenced by
-                * local_GP->put is not XPC_M_READY or that local_GP->put
+                * local_GP->put is not XPC_M_SN2_READY or that local_GP->put
                 * equals w_local_GP.put, so we'll go have a look.
                 */
                initial_put = put;
        }
 
-       if (send_IPI)
-               xpc_IPI_send_msgrequest_sn2(ch);
+       if (send_msgrequest)
+               xpc_send_chctl_msgrequest_sn2(ch);
 }
 
 /*
@@ -1530,24 +2041,13 @@ xpc_send_msgs_sn2(struct xpc_channel *ch, s64 initial_put)
  */
 static enum xp_retval
 xpc_allocate_msg_sn2(struct xpc_channel *ch, u32 flags,
-                    struct xpc_msg **address_of_msg)
+                    struct xpc_msg_sn2 **address_of_msg)
 {
-       struct xpc_msg *msg;
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       struct xpc_msg_sn2 *msg;
        enum xp_retval ret;
        s64 put;
 
-       /* this reference will be dropped in xpc_send_msg_sn2() */
-       xpc_msgqueue_ref(ch);
-
-       if (ch->flags & XPC_C_DISCONNECTING) {
-               xpc_msgqueue_deref(ch);
-               return ch->reason;
-       }
-       if (!(ch->flags & XPC_C_CONNECTED)) {
-               xpc_msgqueue_deref(ch);
-               return xpNotConnected;
-       }
-
        /*
         * Get the next available message entry from the local message queue.
         * If none are available, we'll make sure that we grab the latest
@@ -1557,9 +2057,9 @@ xpc_allocate_msg_sn2(struct xpc_channel *ch, u32 flags,
 
        while (1) {
 
-               put = ch->w_local_GP.put;
+               put = ch_sn2->w_local_GP.put;
                rmb();  /* guarantee that .put loads before .get */
-               if (put - ch->w_remote_GP.get < ch->local_nentries) {
+               if (put - ch_sn2->w_remote_GP.get < ch->local_nentries) {
 
                        /* There are available message entries. We need to try
                         * to secure one for ourselves. We'll do this by trying
@@ -1567,7 +2067,8 @@ xpc_allocate_msg_sn2(struct xpc_channel *ch, u32 flags,
                         * doesn't beat us to it. If they do, we'll have to
                         * try again.
                         */
-                       if (cmpxchg(&ch->w_local_GP.put, put, put + 1) == put) {
+                       if (cmpxchg(&ch_sn2->w_local_GP.put, put, put + 1) ==
+                           put) {
                                /* we got the entry referenced by put */
                                break;
                        }
@@ -1578,29 +2079,26 @@ xpc_allocate_msg_sn2(struct xpc_channel *ch, u32 flags,
                 * There aren't any available msg entries at this time.
                 *
                 * In waiting for a message entry to become available,
-                * we set a timeout in case the other side is not
-                * sending completion IPIs. This lets us fake an IPI
-                * that will cause the IPI handler to fetch the latest
-                * GP values as if an IPI was sent by the other side.
+                * we set a timeout in case the other side is not sending
+                * completion interrupts. This lets us fake a notify IRQ
+                * that will cause the notify IRQ handler to fetch the latest
+                * GP values as if an interrupt was sent by the other side.
                 */
                if (ret == xpTimeout)
-                       xpc_IPI_send_local_msgrequest_sn2(ch);
+                       xpc_send_chctl_local_msgrequest_sn2(ch);
 
-               if (flags & XPC_NOWAIT) {
-                       xpc_msgqueue_deref(ch);
+               if (flags & XPC_NOWAIT)
                        return xpNoWait;
-               }
 
                ret = xpc_allocate_msg_wait(ch);
-               if (ret != xpInterrupted && ret != xpTimeout) {
-                       xpc_msgqueue_deref(ch);
+               if (ret != xpInterrupted && ret != xpTimeout)
                        return ret;
-               }
        }
 
        /* get the message's address and initialize it */
-       msg = (struct xpc_msg *)((u64)ch->local_msgqueue +
-                                (put % ch->local_nentries) * ch->msg_size);
+       msg = (struct xpc_msg_sn2 *)((u64)ch_sn2->local_msgqueue +
+                                    (put % ch->local_nentries) *
+                                    ch->entry_size);
 
        DBUG_ON(msg->flags != 0);
        msg->number = put;
@@ -1610,49 +2108,63 @@ xpc_allocate_msg_sn2(struct xpc_channel *ch, u32 flags,
                (void *)msg, msg->number, ch->partid, ch->number);
 
        *address_of_msg = msg;
-
        return xpSuccess;
 }
 
 /*
  * Common code that does the actual sending of the message by advancing the
- * local message queue's Put value and sends an IPI to the partition the
- * message is being sent to.
+ * local message queue's Put value and sends a chctl msgrequest to the
+ * partition the message is being sent to.
  */
 static enum xp_retval
-xpc_send_msg_sn2(struct xpc_channel *ch, struct xpc_msg *msg, u8 notify_type,
-                xpc_notify_func func, void *key)
+xpc_send_payload_sn2(struct xpc_channel *ch, u32 flags, void *payload,
+                    u16 payload_size, u8 notify_type, xpc_notify_func func,
+                    void *key)
 {
        enum xp_retval ret = xpSuccess;
-       struct xpc_notify *notify = notify;
-       s64 put, msg_number = msg->number;
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       struct xpc_msg_sn2 *msg = msg;
+       struct xpc_notify_sn2 *notify = notify;
+       s64 msg_number;
+       s64 put;
 
        DBUG_ON(notify_type == XPC_N_CALL && func == NULL);
-       DBUG_ON((((u64)msg - (u64)ch->local_msgqueue) / ch->msg_size) !=
-               msg_number % ch->local_nentries);
-       DBUG_ON(msg->flags & XPC_M_READY);
+
+       if (XPC_MSG_SIZE(payload_size) > ch->entry_size)
+               return xpPayloadTooBig;
+
+       xpc_msgqueue_ref(ch);
 
        if (ch->flags & XPC_C_DISCONNECTING) {
-               /* drop the reference grabbed in xpc_allocate_msg_sn2() */
-               xpc_msgqueue_deref(ch);
-               return ch->reason;
+               ret = ch->reason;
+               goto out_1;
        }
+       if (!(ch->flags & XPC_C_CONNECTED)) {
+               ret = xpNotConnected;
+               goto out_1;
+       }
+
+       ret = xpc_allocate_msg_sn2(ch, flags, &msg);
+       if (ret != xpSuccess)
+               goto out_1;
+
+       msg_number = msg->number;
 
        if (notify_type != 0) {
                /*
                 * Tell the remote side to send an ACK interrupt when the
                 * message has been delivered.
                 */
-               msg->flags |= XPC_M_INTERRUPT;
+               msg->flags |= XPC_M_SN2_INTERRUPT;
 
                atomic_inc(&ch->n_to_notify);
 
-               notify = &ch->notify_queue[msg_number % ch->local_nentries];
+               notify = &ch_sn2->notify_queue[msg_number % ch->local_nentries];
                notify->func = func;
                notify->key = key;
                notify->type = notify_type;
 
-               /* >>> is a mb() needed here? */
+               /* ??? Is a mb() needed here? */
 
                if (ch->flags & XPC_C_DISCONNECTING) {
                        /*
@@ -1667,28 +2179,27 @@ xpc_send_msg_sn2(struct xpc_channel *ch, struct xpc_msg *msg, u8 notify_type,
                                atomic_dec(&ch->n_to_notify);
                                ret = ch->reason;
                        }
-
-                       /* drop reference grabbed in xpc_allocate_msg_sn2() */
-                       xpc_msgqueue_deref(ch);
-                       return ret;
+                       goto out_1;
                }
        }
 
-       msg->flags |= XPC_M_READY;
+       memcpy(&msg->payload, payload, payload_size);
+
+       msg->flags |= XPC_M_SN2_READY;
 
        /*
         * The preceding store of msg->flags must occur before the following
-        * load of ch->local_GP->put.
+        * load of local_GP->put.
         */
        mb();
 
        /* see if the message is next in line to be sent, if so send it */
 
-       put = ch->local_GP->put;
+       put = ch_sn2->local_GP->put;
        if (put == msg_number)
                xpc_send_msgs_sn2(ch, put);
 
-       /* drop the reference grabbed in xpc_allocate_msg_sn2() */
+out_1:
        xpc_msgqueue_deref(ch);
        return ret;
 }
@@ -1696,26 +2207,31 @@ xpc_send_msg_sn2(struct xpc_channel *ch, struct xpc_msg *msg, u8 notify_type,
 /*
  * Now we actually acknowledge the messages that have been delivered and ack'd
  * by advancing the cached remote message queue's Get value and if requested
- * send an IPI to the message sender's partition.
+ * send a chctl msgrequest to the message sender's partition.
+ *
+ * If a message has XPC_M_SN2_INTERRUPT set, send an interrupt to the partition
+ * that sent the message.
  */
 static void
 xpc_acknowledge_msgs_sn2(struct xpc_channel *ch, s64 initial_get, u8 msg_flags)
 {
-       struct xpc_msg *msg;
+       struct xpc_channel_sn2 *ch_sn2 = &ch->sn.sn2;
+       struct xpc_msg_sn2 *msg;
        s64 get = initial_get + 1;
-       int send_IPI = 0;
+       int send_msgrequest = 0;
 
        while (1) {
 
                while (1) {
-                       if (get == ch->w_local_GP.get)
+                       if (get == ch_sn2->w_local_GP.get)
                                break;
 
-                       msg = (struct xpc_msg *)((u64)ch->remote_msgqueue +
-                                                (get % ch->remote_nentries) *
-                                                ch->msg_size);
+                       msg = (struct xpc_msg_sn2 *)((u64)ch_sn2->
+                                                    remote_msgqueue + (get %
+                                                    ch->remote_nentries) *
+                                                    ch->entry_size);
 
-                       if (!(msg->flags & XPC_M_DONE))
+                       if (!(msg->flags & XPC_M_SN2_DONE))
                                break;
 
                        msg_flags |= msg->flags;
@@ -1727,10 +2243,10 @@ xpc_acknowledge_msgs_sn2(struct xpc_channel *ch, s64 initial_get, u8 msg_flags)
                        break;
                }
 
-               if (cmpxchg_rel(&ch->local_GP->get, initial_get, get) !=
+               if (cmpxchg_rel(&ch_sn2->local_GP->get, initial_get, get) !=
                    initial_get) {
                        /* someone else beat us to it */
-                       DBUG_ON(ch->local_GP->get <= initial_get);
+                       DBUG_ON(ch_sn2->local_GP->get <= initial_get);
                        break;
                }
 
@@ -1739,38 +2255,42 @@ xpc_acknowledge_msgs_sn2(struct xpc_channel *ch, s64 initial_get, u8 msg_flags)
                dev_dbg(xpc_chan, "local_GP->get changed to %ld, partid=%d, "
                        "channel=%d\n", get, ch->partid, ch->number);
 
-               send_IPI = (msg_flags & XPC_M_INTERRUPT);
+               send_msgrequest = (msg_flags & XPC_M_SN2_INTERRUPT);
 
                /*
                 * We need to ensure that the message referenced by
-                * local_GP->get is not XPC_M_DONE or that local_GP->get
+                * local_GP->get is not XPC_M_SN2_DONE or that local_GP->get
                 * equals w_local_GP.get, so we'll go have a look.
                 */
                initial_get = get;
        }
 
-       if (send_IPI)
-               xpc_IPI_send_msgrequest_sn2(ch);
+       if (send_msgrequest)
+               xpc_send_chctl_msgrequest_sn2(ch);
 }
 
 static void
-xpc_received_msg_sn2(struct xpc_channel *ch, struct xpc_msg *msg)
+xpc_received_payload_sn2(struct xpc_channel *ch, void *payload)
 {
+       struct xpc_msg_sn2 *msg;
+       s64 msg_number;
        s64 get;
-       s64 msg_number = msg->number;
+
+       msg = container_of(payload, struct xpc_msg_sn2, payload);
+       msg_number = msg->number;
 
        dev_dbg(xpc_chan, "msg=0x%p, msg_number=%ld, partid=%d, channel=%d\n",
                (void *)msg, msg_number, ch->partid, ch->number);
 
-       DBUG_ON((((u64)msg - (u64)ch->remote_msgqueue) / ch->msg_size) !=
+       DBUG_ON((((u64)msg - (u64)ch->remote_msgqueue) / ch->entry_size) !=
                msg_number % ch->remote_nentries);
-       DBUG_ON(msg->flags & XPC_M_DONE);
+       DBUG_ON(msg->flags & XPC_M_SN2_DONE);
 
-       msg->flags |= XPC_M_DONE;
+       msg->flags |= XPC_M_SN2_DONE;
 
        /*
         * The preceding store of msg->flags must occur before the following
-        * load of ch->local_GP->get.
+        * load of local_GP->get.
         */
        mb();
 
@@ -1778,60 +2298,107 @@ xpc_received_msg_sn2(struct xpc_channel *ch, struct xpc_msg *msg)
         * See if this message is next in line to be acknowledged as having
         * been delivered.
         */
-       get = ch->local_GP->get;
+       get = ch->sn.sn2.local_GP->get;
        if (get == msg_number)
                xpc_acknowledge_msgs_sn2(ch, get, msg->flags);
 }
 
-void
+int
 xpc_init_sn2(void)
 {
-       xpc_rsvd_page_init = xpc_rsvd_page_init_sn2;
+       int ret;
+       size_t buf_size;
+
+       xpc_setup_partitions_sn = xpc_setup_partitions_sn_sn2;
+       xpc_get_partition_rsvd_page_pa = xpc_get_partition_rsvd_page_pa_sn2;
+       xpc_setup_rsvd_page_sn = xpc_setup_rsvd_page_sn_sn2;
        xpc_increment_heartbeat = xpc_increment_heartbeat_sn2;
        xpc_offline_heartbeat = xpc_offline_heartbeat_sn2;
        xpc_online_heartbeat = xpc_online_heartbeat_sn2;
        xpc_heartbeat_init = xpc_heartbeat_init_sn2;
        xpc_heartbeat_exit = xpc_heartbeat_exit_sn2;
-       xpc_check_remote_hb = xpc_check_remote_hb_sn2;
-
-       xpc_initiate_partition_activation =
-           xpc_initiate_partition_activation_sn2;
-       xpc_process_act_IRQ_rcvd = xpc_process_act_IRQ_rcvd_sn2;
-       xpc_setup_infrastructure = xpc_setup_infrastructure_sn2;
-       xpc_teardown_infrastructure = xpc_teardown_infrastructure_sn2;
+       xpc_get_remote_heartbeat = xpc_get_remote_heartbeat_sn2;
+
+       xpc_request_partition_activation = xpc_request_partition_activation_sn2;
+       xpc_request_partition_reactivation =
+           xpc_request_partition_reactivation_sn2;
+       xpc_request_partition_deactivation =
+           xpc_request_partition_deactivation_sn2;
+       xpc_cancel_partition_deactivation_request =
+           xpc_cancel_partition_deactivation_request_sn2;
+
+       xpc_process_activate_IRQ_rcvd = xpc_process_activate_IRQ_rcvd_sn2;
+       xpc_setup_ch_structures_sn = xpc_setup_ch_structures_sn_sn2;
+       xpc_teardown_ch_structures_sn = xpc_teardown_ch_structures_sn_sn2;
        xpc_make_first_contact = xpc_make_first_contact_sn2;
-       xpc_get_IPI_flags = xpc_get_IPI_flags_sn2;
-       xpc_get_deliverable_msg = xpc_get_deliverable_msg_sn2;
-
-       xpc_mark_partition_engaged = xpc_mark_partition_engaged_sn2;
-       xpc_mark_partition_disengaged = xpc_mark_partition_disengaged_sn2;
-       xpc_request_partition_disengage = xpc_request_partition_disengage_sn2;
-       xpc_cancel_partition_disengage_request =
-           xpc_cancel_partition_disengage_request_sn2;
+
+       xpc_get_chctl_all_flags = xpc_get_chctl_all_flags_sn2;
+       xpc_send_chctl_closerequest = xpc_send_chctl_closerequest_sn2;
+       xpc_send_chctl_closereply = xpc_send_chctl_closereply_sn2;
+       xpc_send_chctl_openrequest = xpc_send_chctl_openrequest_sn2;
+       xpc_send_chctl_openreply = xpc_send_chctl_openreply_sn2;
+
+       xpc_save_remote_msgqueue_pa = xpc_save_remote_msgqueue_pa_sn2;
+
+       xpc_setup_msg_structures = xpc_setup_msg_structures_sn2;
+       xpc_teardown_msg_structures = xpc_teardown_msg_structures_sn2;
+
+       xpc_notify_senders_of_disconnect = xpc_notify_senders_of_disconnect_sn2;
+       xpc_process_msg_chctl_flags = xpc_process_msg_chctl_flags_sn2;
+       xpc_n_of_deliverable_payloads = xpc_n_of_deliverable_payloads_sn2;
+       xpc_get_deliverable_payload = xpc_get_deliverable_payload_sn2;
+
+       xpc_indicate_partition_engaged = xpc_indicate_partition_engaged_sn2;
+       xpc_indicate_partition_disengaged =
+           xpc_indicate_partition_disengaged_sn2;
        xpc_partition_engaged = xpc_partition_engaged_sn2;
-       xpc_partition_disengage_requested =
-           xpc_partition_disengage_requested_sn2;
-       xpc_clear_partition_engaged = xpc_clear_partition_engaged_sn2;
-       xpc_clear_partition_disengage_request =
-           xpc_clear_partition_disengage_request_sn2;
+       xpc_any_partition_engaged = xpc_any_partition_engaged_sn2;
+       xpc_assume_partition_disengaged = xpc_assume_partition_disengaged_sn2;
 
-       xpc_IPI_send_local_activate = xpc_IPI_send_local_activate_sn2;
-       xpc_IPI_send_activated = xpc_IPI_send_activated_sn2;
-       xpc_IPI_send_local_reactivate = xpc_IPI_send_local_reactivate_sn2;
-       xpc_IPI_send_disengage = xpc_IPI_send_disengage_sn2;
+       xpc_send_payload = xpc_send_payload_sn2;
+       xpc_received_payload = xpc_received_payload_sn2;
 
-       xpc_IPI_send_closerequest = xpc_IPI_send_closerequest_sn2;
-       xpc_IPI_send_closereply = xpc_IPI_send_closereply_sn2;
-       xpc_IPI_send_openrequest = xpc_IPI_send_openrequest_sn2;
-       xpc_IPI_send_openreply = xpc_IPI_send_openreply_sn2;
+       if (offsetof(struct xpc_msg_sn2, payload) > XPC_MSG_HDR_MAX_SIZE) {
+               dev_err(xpc_part, "header portion of struct xpc_msg_sn2 is "
+                       "larger than %d\n", XPC_MSG_HDR_MAX_SIZE);
+               return -E2BIG;
+       }
 
-       xpc_allocate_msg = xpc_allocate_msg_sn2;
+       buf_size = max(XPC_RP_VARS_SIZE,
+                      XPC_RP_HEADER_SIZE + XP_NASID_MASK_BYTES_SN2);
+       xpc_remote_copy_buffer_sn2 = xpc_kmalloc_cacheline_aligned(buf_size,
+                                                                  GFP_KERNEL,
+                                             &xpc_remote_copy_buffer_base_sn2);
+       if (xpc_remote_copy_buffer_sn2 == NULL) {
+               dev_err(xpc_part, "can't get memory for remote copy buffer\n");
+               return -ENOMEM;
+       }
 
-       xpc_send_msg = xpc_send_msg_sn2;
-       xpc_received_msg = xpc_received_msg_sn2;
+       /* open up protections for IPI and [potentially] amo operations */
+       xpc_allow_IPI_ops_sn2();
+       xpc_allow_amo_ops_shub_wars_1_1_sn2();
+
+       /*
+        * 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_handle_activate_IRQ_sn2, 0,
+                         "xpc hb", NULL);
+       if (ret != 0) {
+               dev_err(xpc_part, "can't register ACTIVATE IRQ handler, "
+                       "errno=%d\n", -ret);
+               xpc_disallow_IPI_ops_sn2();
+               kfree(xpc_remote_copy_buffer_base_sn2);
+       }
+       return ret;
 }
 
 void
 xpc_exit_sn2(void)
 {
+       free_irq(SGI_XPC_ACTIVATE, NULL);
+       xpc_disallow_IPI_ops_sn2();
+       kfree(xpc_remote_copy_buffer_base_sn2);
 }