powerpc: move iSeries_setup.[ch] and mf.c into platforms/iseries

iSeries_setup.c becomes setup.c
iSeries_setup.h becomes setup.h
mf.c retains its name

Also moved iSeries_[gs]et_rtc_time and iSeries_get_boot_time into
mf.c since they are just small wrappers around mf_ functions.

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>

4 files changed, 2347 insertions, 1 deletions

diff --git a/arch/powerpc/platforms/iseries/Makefile b/arch/powerpc/platforms/iseries/Makefile
index 095471d50d9..f5e11907cab 100644
--- a/arch/powerpc/platforms/iseries/Makefile
+++ b/arch/powerpc/platforms/iseries/Makefile
@@ -1 +1 @@
-obj-$(CONFIG_PPC_ISERIES) += hvcall.o hvlpconfig.o lpardata.o
+obj-y += hvcall.o hvlpconfig.o lpardata.o setup.o mf.o
diff --git a/arch/powerpc/platforms/iseries/mf.c b/arch/powerpc/platforms/iseries/mf.c
new file mode 100644
index 00000000000..82f5abab9af
--- /dev/null
+++ b/arch/powerpc/platforms/iseries/mf.c
@@ -0,0 +1,1316 @@
+/*
+ * Copyright (C) 2001 Troy D. Armstrong  IBM Corporation
+ * Copyright (C) 2004-2005 Stephen Rothwell  IBM Corporation
+ *
+ * This modules exists as an interface between a Linux secondary partition
+ * running on an iSeries and the primary partition's Virtual Service
+ * Processor (VSP) object.  The VSP has final authority over powering on/off
+ * all partitions in the iSeries.  It also provides miscellaneous low-level
+ * machine facility type operations.
+ *
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ */
+
+#include <linux/types.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/completion.h>
+#include <linux/delay.h>
+#include <linux/dma-mapping.h>
+#include <linux/bcd.h>
+
+#include <asm/time.h>
+#include <asm/uaccess.h>
+#include <asm/paca.h>
+#include <asm/iSeries/vio.h>
+#include <asm/iSeries/mf.h>
+#include <asm/iSeries/HvLpConfig.h>
+#include <asm/iSeries/ItLpQueue.h>
+
+#include "setup.h"
+
+extern int piranha_simulator;
+
+/*
+ * This is the structure layout for the Machine Facilites LPAR event
+ * flows.
+ */
+struct vsp_cmd_data {
+	u64 token;
+	u16 cmd;
+	HvLpIndex lp_index;
+	u8 result_code;
+	u32 reserved;
+	union {
+		u64 state;	/* GetStateOut */
+		u64 ipl_type;	/* GetIplTypeOut, Function02SelectIplTypeIn */
+		u64 ipl_mode;	/* GetIplModeOut, Function02SelectIplModeIn */
+		u64 page[4];	/* GetSrcHistoryIn */
+		u64 flag;	/* GetAutoIplWhenPrimaryIplsOut,
+				   SetAutoIplWhenPrimaryIplsIn,
+				   WhiteButtonPowerOffIn,
+				   Function08FastPowerOffIn,
+				   IsSpcnRackPowerIncompleteOut */
+		struct {
+			u64 token;
+			u64 address_type;
+			u64 side;
+			u32 length;
+			u32 offset;
+		} kern;		/* SetKernelImageIn, GetKernelImageIn,
+				   SetKernelCmdLineIn, GetKernelCmdLineIn */
+		u32 length_out;	/* GetKernelImageOut, GetKernelCmdLineOut */
+		u8 reserved[80];
+	} sub_data;
+};
+
+struct vsp_rsp_data {
+	struct completion com;
+	struct vsp_cmd_data *response;
+};
+
+struct alloc_data {
+	u16 size;
+	u16 type;
+	u32 count;
+	u16 reserved1;
+	u8 reserved2;
+	HvLpIndex target_lp;
+};
+
+struct ce_msg_data;
+
+typedef void (*ce_msg_comp_hdlr)(void *token, struct ce_msg_data *vsp_cmd_rsp);
+
+struct ce_msg_comp_data {
+	ce_msg_comp_hdlr handler;
+	void *token;
+};
+
+struct ce_msg_data {
+	u8 ce_msg[12];
+	char reserved[4];
+	struct ce_msg_comp_data *completion;
+};
+
+struct io_mf_lp_event {
+	struct HvLpEvent hp_lp_event;
+	u16 subtype_result_code;
+	u16 reserved1;
+	u32 reserved2;
+	union {
+		struct alloc_data alloc;
+		struct ce_msg_data ce_msg;
+		struct vsp_cmd_data vsp_cmd;
+	} data;
+};
+
+#define subtype_data(a, b, c, d)	\
+		(((a) << 24) + ((b) << 16) + ((c) << 8) + (d))
+
+/*
+ * All outgoing event traffic is kept on a FIFO queue.  The first
+ * pointer points to the one that is outstanding, and all new
+ * requests get stuck on the end.  Also, we keep a certain number of
+ * preallocated pending events so that we can operate very early in
+ * the boot up sequence (before kmalloc is ready).
+ */
+struct pending_event {
+	struct pending_event *next;
+	struct io_mf_lp_event event;
+	MFCompleteHandler hdlr;
+	char dma_data[72];
+	unsigned dma_data_length;
+	unsigned remote_address;
+};
+static spinlock_t pending_event_spinlock;
+static struct pending_event *pending_event_head;
+static struct pending_event *pending_event_tail;
+static struct pending_event *pending_event_avail;
+static struct pending_event pending_event_prealloc[16];
+
+/*
+ * Put a pending event onto the available queue, so it can get reused.
+ * Attention! You must have the pending_event_spinlock before calling!
+ */
+static void free_pending_event(struct pending_event *ev)
+{
+	if (ev != NULL) {
+		ev->next = pending_event_avail;
+		pending_event_avail = ev;
+	}
+}
+
+/*
+ * Enqueue the outbound event onto the stack.  If the queue was
+ * empty to begin with, we must also issue it via the Hypervisor
+ * interface.  There is a section of code below that will touch
+ * the first stack pointer without the protection of the pending_event_spinlock.
+ * This is OK, because we know that nobody else will be modifying
+ * the first pointer when we do this.
+ */
+static int signal_event(struct pending_event *ev)
+{
+	int rc = 0;
+	unsigned long flags;
+	int go = 1;
+	struct pending_event *ev1;
+	HvLpEvent_Rc hv_rc;
+
+	/* enqueue the event */
+	if (ev != NULL) {
+		ev->next = NULL;
+		spin_lock_irqsave(&pending_event_spinlock, flags);
+		if (pending_event_head == NULL)
+			pending_event_head = ev;
+		else {
+			go = 0;
+			pending_event_tail->next = ev;
+		}
+		pending_event_tail = ev;
+		spin_unlock_irqrestore(&pending_event_spinlock, flags);
+	}
+
+	/* send the event */
+	while (go) {
+		go = 0;
+
+		/* any DMA data to send beforehand? */
+		if (pending_event_head->dma_data_length > 0)
+			HvCallEvent_dmaToSp(pending_event_head->dma_data,
+					pending_event_head->remote_address,
+					pending_event_head->dma_data_length,
+					HvLpDma_Direction_LocalToRemote);
+
+		hv_rc = HvCallEvent_signalLpEvent(
+				&pending_event_head->event.hp_lp_event);
+		if (hv_rc != HvLpEvent_Rc_Good) {
+			printk(KERN_ERR "mf.c: HvCallEvent_signalLpEvent() "
+					"failed with %d\n", (int)hv_rc);
+
+			spin_lock_irqsave(&pending_event_spinlock, flags);
+			ev1 = pending_event_head;
+			pending_event_head = pending_event_head->next;
+			if (pending_event_head != NULL)
+				go = 1;
+			spin_unlock_irqrestore(&pending_event_spinlock, flags);
+
+			if (ev1 == ev)
+				rc = -EIO;
+			else if (ev1->hdlr != NULL)
+				(*ev1->hdlr)((void *)ev1->event.hp_lp_event.xCorrelationToken, -EIO);
+
+			spin_lock_irqsave(&pending_event_spinlock, flags);
+			free_pending_event(ev1);
+			spin_unlock_irqrestore(&pending_event_spinlock, flags);
+		}
+	}
+
+	return rc;
+}
+
+/*
+ * Allocate a new pending_event structure, and initialize it.
+ */
+static struct pending_event *new_pending_event(void)
+{
+	struct pending_event *ev = NULL;
+	HvLpIndex primary_lp = HvLpConfig_getPrimaryLpIndex();
+	unsigned long flags;
+	struct HvLpEvent *hev;
+
+	spin_lock_irqsave(&pending_event_spinlock, flags);
+	if (pending_event_avail != NULL) {
+		ev = pending_event_avail;
+		pending_event_avail = pending_event_avail->next;
+	}
+	spin_unlock_irqrestore(&pending_event_spinlock, flags);
+	if (ev == NULL) {
+		ev = kmalloc(sizeof(struct pending_event), GFP_ATOMIC);
+		if (ev == NULL) {
+			printk(KERN_ERR "mf.c: unable to kmalloc %ld bytes\n",
+					sizeof(struct pending_event));
+			return NULL;
+		}
+	}
+	memset(ev, 0, sizeof(struct pending_event));
+	hev = &ev->event.hp_lp_event;
+	hev->xFlags.xValid = 1;
+	hev->xFlags.xAckType = HvLpEvent_AckType_ImmediateAck;
+	hev->xFlags.xAckInd = HvLpEvent_AckInd_DoAck;
+	hev->xFlags.xFunction = HvLpEvent_Function_Int;
+	hev->xType = HvLpEvent_Type_MachineFac;
+	hev->xSourceLp = HvLpConfig_getLpIndex();
+	hev->xTargetLp = primary_lp;
+	hev->xSizeMinus1 = sizeof(ev->event) - 1;
+	hev->xRc = HvLpEvent_Rc_Good;
+	hev->xSourceInstanceId = HvCallEvent_getSourceLpInstanceId(primary_lp,
+			HvLpEvent_Type_MachineFac);
+	hev->xTargetInstanceId = HvCallEvent_getTargetLpInstanceId(primary_lp,
+			HvLpEvent_Type_MachineFac);
+
+	return ev;
+}
+
+static int signal_vsp_instruction(struct vsp_cmd_data *vsp_cmd)
+{
+	struct pending_event *ev = new_pending_event();
+	int rc;
+	struct vsp_rsp_data response;
+
+	if (ev == NULL)
+		return -ENOMEM;
+
+	init_completion(&response.com);
+	response.response = vsp_cmd;
+	ev->event.hp_lp_event.xSubtype = 6;
+	ev->event.hp_lp_event.x.xSubtypeData =
+		subtype_data('M', 'F',  'V',  'I');
+	ev->event.data.vsp_cmd.token = (u64)&response;
+	ev->event.data.vsp_cmd.cmd = vsp_cmd->cmd;
+	ev->event.data.vsp_cmd.lp_index = HvLpConfig_getLpIndex();
+	ev->event.data.vsp_cmd.result_code = 0xFF;
+	ev->event.data.vsp_cmd.reserved = 0;
+	memcpy(&(ev->event.data.vsp_cmd.sub_data),
+			&(vsp_cmd->sub_data), sizeof(vsp_cmd->sub_data));
+	mb();
+
+	rc = signal_event(ev);
+	if (rc == 0)
+		wait_for_completion(&response.com);
+	return rc;
+}
+
+
+/*
+ * Send a 12-byte CE message to the primary partition VSP object
+ */
+static int signal_ce_msg(char *ce_msg, struct ce_msg_comp_data *completion)
+{
+	struct pending_event *ev = new_pending_event();
+
+	if (ev == NULL)
+		return -ENOMEM;
+
+	ev->event.hp_lp_event.xSubtype = 0;
+	ev->event.hp_lp_event.x.xSubtypeData =
+		subtype_data('M',  'F',  'C',  'E');
+	memcpy(ev->event.data.ce_msg.ce_msg, ce_msg, 12);
+	ev->event.data.ce_msg.completion = completion;
+	return signal_event(ev);
+}
+
+/*
+ * Send a 12-byte CE message (with no data) to the primary partition VSP object
+ */
+static int signal_ce_msg_simple(u8 ce_op, struct ce_msg_comp_data *completion)
+{
+	u8 ce_msg[12];
+
+	memset(ce_msg, 0, sizeof(ce_msg));
+	ce_msg[3] = ce_op;
+	return signal_ce_msg(ce_msg, completion);
+}
+
+/*
+ * Send a 12-byte CE message and DMA data to the primary partition VSP object
+ */
+static int dma_and_signal_ce_msg(char *ce_msg,
+		struct ce_msg_comp_data *completion, void *dma_data,
+		unsigned dma_data_length, unsigned remote_address)
+{
+	struct pending_event *ev = new_pending_event();
+
+	if (ev == NULL)
+		return -ENOMEM;
+
+	ev->event.hp_lp_event.xSubtype = 0;
+	ev->event.hp_lp_event.x.xSubtypeData =
+		subtype_data('M', 'F', 'C', 'E');
+	memcpy(ev->event.data.ce_msg.ce_msg, ce_msg, 12);
+	ev->event.data.ce_msg.completion = completion;
+	memcpy(ev->dma_data, dma_data, dma_data_length);
+	ev->dma_data_length = dma_data_length;
+	ev->remote_address = remote_address;
+	return signal_event(ev);
+}
+
+/*
+ * Initiate a nice (hopefully) shutdown of Linux.  We simply are
+ * going to try and send the init process a SIGINT signal.  If
+ * this fails (why?), we'll simply force it off in a not-so-nice
+ * manner.
+ */
+static int shutdown(void)
+{
+	int rc = kill_proc(1, SIGINT, 1);
+
+	if (rc) {
+		printk(KERN_ALERT "mf.c: SIGINT to init failed (%d), "
+				"hard shutdown commencing\n", rc);
+		mf_power_off();
+	} else
+		printk(KERN_INFO "mf.c: init has been successfully notified "
+				"to proceed with shutdown\n");
+	return rc;
+}
+
+/*
+ * The primary partition VSP object is sending us a new
+ * event flow.  Handle it...
+ */
+static void handle_int(struct io_mf_lp_event *event)
+{
+	struct ce_msg_data *ce_msg_data;
+	struct ce_msg_data *pce_msg_data;
+	unsigned long flags;
+	struct pending_event *pev;
+
+	/* ack the interrupt */
+	event->hp_lp_event.xRc = HvLpEvent_Rc_Good;
+	HvCallEvent_ackLpEvent(&event->hp_lp_event);
+
+	/* process interrupt */
+	switch (event->hp_lp_event.xSubtype) {
+	case 0:	/* CE message */
+		ce_msg_data = &event->data.ce_msg;
+		switch (ce_msg_data->ce_msg[3]) {
+		case 0x5B:	/* power control notification */
+			if ((ce_msg_data->ce_msg[5] & 0x20) != 0) {
+				printk(KERN_INFO "mf.c: Commencing partition shutdown\n");
+				if (shutdown() == 0)
+					signal_ce_msg_simple(0xDB, NULL);
+			}
+			break;
+		case 0xC0:	/* get time */
+			spin_lock_irqsave(&pending_event_spinlock, flags);
+			pev = pending_event_head;
+			if (pev != NULL)
+				pending_event_head = pending_event_head->next;
+			spin_unlock_irqrestore(&pending_event_spinlock, flags);
+			if (pev == NULL)
+				break;
+			pce_msg_data = &pev->event.data.ce_msg;
+			if (pce_msg_data->ce_msg[3] != 0x40)
+				break;
+			if (pce_msg_data->completion != NULL) {
+				ce_msg_comp_hdlr handler =
+					pce_msg_data->completion->handler;
+				void *token = pce_msg_data->completion->token;
+
+				if (handler != NULL)
+					(*handler)(token, ce_msg_data);
+			}
+			spin_lock_irqsave(&pending_event_spinlock, flags);
+			free_pending_event(pev);
+			spin_unlock_irqrestore(&pending_event_spinlock, flags);
+			/* send next waiting event */
+			if (pending_event_head != NULL)
+				signal_event(NULL);
+			break;
+		}
+		break;
+	case 1:	/* IT sys shutdown */
+		printk(KERN_INFO "mf.c: Commencing system shutdown\n");
+		shutdown();
+		break;
+	}
+}
+
+/*
+ * The primary partition VSP object is acknowledging the receipt
+ * of a flow we sent to them.  If there are other flows queued
+ * up, we must send another one now...
+ */
+static void handle_ack(struct io_mf_lp_event *event)
+{
+	unsigned long flags;
+	struct pending_event *two = NULL;
+	unsigned long free_it = 0;
+	struct ce_msg_data *ce_msg_data;
+	struct ce_msg_data *pce_msg_data;
+	struct vsp_rsp_data *rsp;
+
+	/* handle current event */
+	if (pending_event_head == NULL) {
+		printk(KERN_ERR "mf.c: stack empty for receiving ack\n");
+		return;
+	}
+
+	switch (event->hp_lp_event.xSubtype) {
+	case 0:     /* CE msg */
+		ce_msg_data = &event->data.ce_msg;
+		if (ce_msg_data->ce_msg[3] != 0x40) {
+			free_it = 1;
+			break;
+		}
+		if (ce_msg_data->ce_msg[2] == 0)
+			break;
+		free_it = 1;
+		pce_msg_data = &pending_event_head->event.data.ce_msg;
+		if (pce_msg_data->completion != NULL) {
+			ce_msg_comp_hdlr handler =
+				pce_msg_data->completion->handler;
+			void *token = pce_msg_data->completion->token;
+
+			if (handler != NULL)
+				(*handler)(token, ce_msg_data);
+		}
+		break;
+	case 4:	/* allocate */
+	case 5:	/* deallocate */
+		if (pending_event_head->hdlr != NULL)
+			(*pending_event_head->hdlr)((void *)event->hp_lp_event.xCorrelationToken, event->data.alloc.count);
+		free_it = 1;
+		break;
+	case 6:
+		free_it = 1;
+		rsp = (struct vsp_rsp_data *)event->data.vsp_cmd.token;
+		if (rsp == NULL) {
+			printk(KERN_ERR "mf.c: no rsp\n");
+			break;
+		}
+		if (rsp->response != NULL)
+			memcpy(rsp->response, &event->data.vsp_cmd,
+					sizeof(event->data.vsp_cmd));
+		complete(&rsp->com);
+		break;
+	}
+
+	/* remove from queue */
+	spin_lock_irqsave(&pending_event_spinlock, flags);
+	if ((pending_event_head != NULL) && (free_it == 1)) {
+		struct pending_event *oldHead = pending_event_head;
+
+		pending_event_head = pending_event_head->next;
+		two = pending_event_head;
+		free_pending_event(oldHead);
+	}
+	spin_unlock_irqrestore(&pending_event_spinlock, flags);
+
+	/* send next waiting event */
+	if (two != NULL)
+		signal_event(NULL);
+}
+
+/*
+ * This is the generic event handler we are registering with
+ * the Hypervisor.  Ensure the flows are for us, and then
+ * parse it enough to know if it is an interrupt or an
+ * acknowledge.
+ */
+static void hv_handler(struct HvLpEvent *event, struct pt_regs *regs)
+{
+	if ((event != NULL) && (event->xType == HvLpEvent_Type_MachineFac)) {
+		switch(event->xFlags.xFunction) {
+		case HvLpEvent_Function_Ack:
+			handle_ack((struct io_mf_lp_event *)event);
+			break;
+		case HvLpEvent_Function_Int:
+			handle_int((struct io_mf_lp_event *)event);
+			break;
+		default:
+			printk(KERN_ERR "mf.c: non ack/int event received\n");
+			break;
+		}
+	} else
+		printk(KERN_ERR "mf.c: alien event received\n");
+}
+
+/*
+ * Global kernel interface to allocate and seed events into the
+ * Hypervisor.
+ */
+void mf_allocate_lp_events(HvLpIndex target_lp, HvLpEvent_Type type,
+		unsigned size, unsigned count, MFCompleteHandler hdlr,
+		void *user_token)
+{
+	struct pending_event *ev = new_pending_event();
+	int rc;
+
+	if (ev == NULL) {
+		rc = -ENOMEM;
+	} else {
+		ev->event.hp_lp_event.xSubtype = 4;
+		ev->event.hp_lp_event.xCorrelationToken = (u64)user_token;
+		ev->event.hp_lp_event.x.xSubtypeData =
+			subtype_data('M', 'F', 'M', 'A');
+		ev->event.data.alloc.target_lp = target_lp;
+		ev->event.data.alloc.type = type;
+		ev->event.data.alloc.size = size;
+		ev->event.data.alloc.count = count;
+		ev->hdlr = hdlr;
+		rc = signal_event(ev);
+	}
+	if ((rc != 0) && (hdlr != NULL))
+		(*hdlr)(user_token, rc);
+}
+EXPORT_SYMBOL(mf_allocate_lp_events);
+
+/*
+ * Global kernel interface to unseed and deallocate events already in
+ * Hypervisor.
+ */
+void mf_deallocate_lp_events(HvLpIndex target_lp, HvLpEvent_Type type,
+		unsigned count, MFCompleteHandler hdlr, void *user_token)
+{
+	struct pending_event *ev = new_pending_event();
+	int rc;
+
+	if (ev == NULL)
+		rc = -ENOMEM;
+	else {
+		ev->event.hp_lp_event.xSubtype = 5;
+		ev->event.hp_lp_event.xCorrelationToken = (u64)user_token;
+		ev->event.hp_lp_event.x.xSubtypeData =
+			subtype_data('M', 'F', 'M', 'D');
+		ev->event.data.alloc.target_lp = target_lp;
+		ev->event.data.alloc.type = type;
+		ev->event.data.alloc.count = count;
+		ev->hdlr = hdlr;
+		rc = signal_event(ev);
+	}
+	if ((rc != 0) && (hdlr != NULL))
+		(*hdlr)(user_token, rc);
+}
+EXPORT_SYMBOL(mf_deallocate_lp_events);
+
+/*
+ * Global kernel interface to tell the VSP object in the primary
+ * partition to power this partition off.
+ */
+void mf_power_off(void)
+{
+	printk(KERN_INFO "mf.c: Down it goes...\n");
+	signal_ce_msg_simple(0x4d, NULL);
+	for (;;)
+		;
+}
+
+/*
+ * Global kernel interface to tell the VSP object in the primary
+ * partition to reboot this partition.
+ */
+void mf_reboot(void)
+{
+	printk(KERN_INFO "mf.c: Preparing to bounce...\n");
+	signal_ce_msg_simple(0x4e, NULL);
+	for (;;)
+		;
+}
+
+/*
+ * Display a single word SRC onto the VSP control panel.
+ */
+void mf_display_src(u32 word)
+{
+	u8 ce[12];
+
+	memset(ce, 0, sizeof(ce));
+	ce[3] = 0x4a;
+	ce[7] = 0x01;
+	ce[8] = word >> 24;
+	ce[9] = word >> 16;
+	ce[10] = word >> 8;
+	ce[11] = word;
+	signal_ce_msg(ce, NULL);
+}
+
+/*
+ * Display a single word SRC of the form "PROGXXXX" on the VSP control panel.
+ */
+void mf_display_progress(u16 value)
+{
+	u8 ce[12];
+	u8 src[72];
+
+	memcpy(ce, "\x00\x00\x04\x4A\x00\x00\x00\x48\x00\x00\x00\x00", 12);
+	memcpy(src, "\x01\x00\x00\x01\x00\x00\x00\x00\x00\x00\x00\x00"
+		"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00"
+		"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00"
+		"\x00\x00\x00\x00PROGxxxx                        ",
+		72);
+	src[6] = value >> 8;
+	src[7] = value & 255;
+	src[44] = "0123456789ABCDEF"[(value >> 12) & 15];
+	src[45] = "0123456789ABCDEF"[(value >> 8) & 15];
+	src[46] = "0123456789ABCDEF"[(value >> 4) & 15];
+	src[47] = "0123456789ABCDEF"[value & 15];
+	dma_and_signal_ce_msg(ce, NULL, src, sizeof(src), 9 * 64 * 1024);
+}
+
+/*
+ * Clear the VSP control panel.  Used to "erase" an SRC that was
+ * previously displayed.
+ */
+void mf_clear_src(void)
+{
+	signal_ce_msg_simple(0x4b, NULL);
+}
+
+/*
+ * Initialization code here.
+ */
+void mf_init(void)
+{
+	int i;
+
+	/* initialize */
+	spin_lock_init(&pending_event_spinlock);
+	for (i = 0;
+	     i < sizeof(pending_event_prealloc) / sizeof(*pending_event_prealloc);
+	     ++i)
+		free_pending_event(&pending_event_prealloc[i]);
+	HvLpEvent_registerHandler(HvLpEvent_Type_MachineFac, &hv_handler);
+
+	/* virtual continue ack */
+	signal_ce_msg_simple(0x57, NULL);
+
+	/* initialization complete */
+	printk(KERN_NOTICE "mf.c: iSeries Linux LPAR Machine Facilities "
+			"initialized\n");
+}
+
+struct rtc_time_data {
+	struct completion com;
+	struct ce_msg_data ce_msg;
+	int rc;
+};
+
+static void get_rtc_time_complete(void *token, struct ce_msg_data *ce_msg)
+{
+	struct rtc_time_data *rtc = token;
+
+	memcpy(&rtc->ce_msg, ce_msg, sizeof(rtc->ce_msg));
+	rtc->rc = 0;
+	complete(&rtc->com);
+}
+
+static int rtc_set_tm(int rc, u8 *ce_msg, struct rtc_time *tm)
+{
+	tm->tm_wday = 0;
+	tm->tm_yday = 0;
+	tm->tm_isdst = 0;
+	if (rc) {
+		tm->tm_sec = 0;
+		tm->tm_min = 0;
+		tm->tm_hour = 0;
+		tm->tm_mday = 15;
+		tm->tm_mon = 5;
+		tm->tm_year = 52;
+		return rc;
+	}
+
+	if ((ce_msg[2] == 0xa9) ||
+	    (ce_msg[2] == 0xaf)) {
+		/* TOD clock is not set */
+		tm->tm_sec = 1;
+		tm->tm_min = 1;
+		tm->tm_hour = 1;
+		tm->tm_mday = 10;
+		tm->tm_mon = 8;
+		tm->tm_year = 71;
+		mf_set_rtc(tm);
+	}
+	{
+		u8 year = ce_msg[5];
+		u8 sec = ce_msg[6];
+		u8 min = ce_msg[7];
+		u8 hour = ce_msg[8];
+		u8 day = ce_msg[10];
+		u8 mon = ce_msg[11];
+
+		BCD_TO_BIN(sec);
+		BCD_TO_BIN(min);
+		BCD_TO_BIN(hour);
+		BCD_TO_BIN(day);
+		BCD_TO_BIN(mon);
+		BCD_TO_BIN(year);
+
+		if (year <= 69)
+			year += 100;
+
+		tm->tm_sec = sec;
+		tm->tm_min = min;
+		tm->tm_hour = hour;
+		tm->tm_mday = day;
+		tm->tm_mon = mon;
+		tm->tm_year = year;
+	}
+
+	return 0;
+}
+
+int mf_get_rtc(struct rtc_time *tm)
+{
+	struct ce_msg_comp_data ce_complete;
+	struct rtc_time_data rtc_data;
+	int rc;
+
+	memset(&ce_complete, 0, sizeof(ce_complete));
+	memset(&rtc_data, 0, sizeof(rtc_data));
+	init_completion(&rtc_data.com);
+	ce_complete.handler = &get_rtc_time_complete;
+	ce_complete.token = &rtc_data;
+	rc = signal_ce_msg_simple(0x40, &ce_complete);
+	if (rc)
+		return rc;
+	wait_for_completion(&rtc_data.com);
+	return rtc_set_tm(rtc_data.rc, rtc_data.ce_msg.ce_msg, tm);
+}
+
+struct boot_rtc_time_data {
+	int busy;
+	struct ce_msg_data ce_msg;
+	int rc;
+};
+
+static void get_boot_rtc_time_complete(void *token, struct ce_msg_data *ce_msg)
+{
+	struct boot_rtc_time_data *rtc = token;
+
+	memcpy(&rtc->ce_msg, ce_msg, sizeof(rtc->ce_msg));
+	rtc->rc = 0;
+	rtc->busy = 0;
+}
+
+int mf_get_boot_rtc(struct rtc_time *tm)
+{
+	struct ce_msg_comp_data ce_complete;
+	struct boot_rtc_time_data rtc_data;
+	int rc;
+
+	memset(&ce_complete, 0, sizeof(ce_complete));
+	memset(&rtc_data, 0, sizeof(rtc_data));
+	rtc_data.busy = 1;
+	ce_complete.handler = &get_boot_rtc_time_complete;
+	ce_complete.token = &rtc_data;
+	rc = signal_ce_msg_simple(0x40, &ce_complete);
+	if (rc)
+		return rc;
+	/* We need to poll here as we are not yet taking interrupts */
+	while (rtc_data.busy) {
+		if (hvlpevent_is_pending())
+			process_hvlpevents(NULL);
+	}
+	return rtc_set_tm(rtc_data.rc, rtc_data.ce_msg.ce_msg, tm);
+}
+
+int mf_set_rtc(struct rtc_time *tm)
+{
+	char ce_time[12];
+	u8 day, mon, hour, min, sec, y1, y2;
+	unsigned year;
+
+	year = 1900 + tm->tm_year;
+	y1 = year / 100;
+	y2 = year % 100;
+
+	sec = tm->tm_sec;
+	min = tm->tm_min;
+	hour = tm->tm_hour;
+	day = tm->tm_mday;
+	mon = tm->tm_mon + 1;
+
+	BIN_TO_BCD(sec);
+	BIN_TO_BCD(min);
+	BIN_TO_BCD(hour);
+	BIN_TO_BCD(mon);
+	BIN_TO_BCD(day);
+	BIN_TO_BCD(y1);
+	BIN_TO_BCD(y2);
+
+	memset(ce_time, 0, sizeof(ce_time));
+	ce_time[3] = 0x41;
+	ce_time[4] = y1;
+	ce_time[5] = y2;
+	ce_time[6] = sec;
+	ce_time[7] = min;
+	ce_time[8] = hour;
+	ce_time[10] = day;
+	ce_time[11] = mon;
+
+	return signal_ce_msg(ce_time, NULL);
+}
+
+#ifdef CONFIG_PROC_FS
+
+static int proc_mf_dump_cmdline(char *page, char **start, off_t off,
+		int count, int *eof, void *data)
+{
+	int len;
+	char *p;
+	struct vsp_cmd_data vsp_cmd;
+	int rc;
+	dma_addr_t dma_addr;
+
+	/* The HV appears to return no more than 256 bytes of command line */
+	if (off >= 256)
+		return 0;
+	if ((off + count) > 256)
+		count = 256 - off;
+
+	dma_addr = dma_map_single(iSeries_vio_dev, page, off + count,
+			DMA_FROM_DEVICE);
+	if (dma_mapping_error(dma_addr))
+		return -ENOMEM;
+	memset(page, 0, off + count);
+	memset(&vsp_cmd, 0, sizeof(vsp_cmd));
+	vsp_cmd.cmd = 33;
+	vsp_cmd.sub_data.kern.token = dma_addr;
+	vsp_cmd.sub_data.kern.address_type = HvLpDma_AddressType_TceIndex;
+	vsp_cmd.sub_data.kern.side = (u64)data;
+	vsp_cmd.sub_data.kern.length = off + count;
+	mb();
+	rc = signal_vsp_instruction(&vsp_cmd);
+	dma_unmap_single(iSeries_vio_dev, dma_addr, off + count,
+			DMA_FROM_DEVICE);
+	if (rc)
+		return rc;
+	if (vsp_cmd.result_code != 0)
+		return -ENOMEM;
+	p = page;
+	len = 0;
+	while (len < (off + count)) {
+		if ((*p == '\0') || (*p == '\n')) {
+			if (*p == '\0')
+				*p = '\n';
+			p++;
+			len++;
+			*eof = 1;
+			break;
+		}
+		p++;
+		len++;
+	}
+
+	if (len < off) {
+		*eof = 1;
+		len = 0;
+	}
+	return len;
+}
+
+#if 0
+static int mf_getVmlinuxChunk(char *buffer, int *size, int offset, u64 side)
+{
+	struct vsp_cmd_data vsp_cmd;
+	int rc;
+	int len = *size;
+	dma_addr_t dma_addr;
+
+	dma_addr = dma_map_single(iSeries_vio_dev, buffer, len,
+			DMA_FROM_DEVICE);
+	memset(buffer, 0, len);
+	memset(&vsp_cmd, 0, sizeof(vsp_cmd));
+	vsp_cmd.cmd = 32;
+	vsp_cmd.sub_data.kern.token = dma_addr;
+	vsp_cmd.sub_data.kern.address_type = HvLpDma_AddressType_TceIndex;
+	vsp_cmd.sub_data.kern.side = side;
+	vsp_cmd.sub_data.kern.offset = offset;
+	vsp_cmd.sub_data.kern.length = len;
+	mb();
+	rc = signal_vsp_instruction(&vsp_cmd);
+	if (rc == 0) {
+		if (vsp_cmd.result_code == 0)
+			*size = vsp_cmd.sub_data.length_out;
+		else
+			rc = -ENOMEM;
+	}
+
+	dma_unmap_single(iSeries_vio_dev, dma_addr, len, DMA_FROM_DEVICE);
+
+	return rc;
+}
+
+static int proc_mf_dump_vmlinux(char *page, char **start, off_t off,
+		int count, int *eof, void *data)
+{
+	int sizeToGet = count;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EACCES;
+
+	if (mf_getVmlinuxChunk(page, &sizeToGet, off, (u64)data) == 0) {
+		if (sizeToGet != 0) {
+			*start = page + off;
+			return sizeToGet;
+		}
+		*eof = 1;
+		return 0;
+	}
+	*eof = 1;
+	return 0;
+}
+#endif
+
+static int proc_mf_dump_side(char *page, char **start, off_t off,
+		int count, int *eof, void *data)
+{
+	int len;
+	char mf_current_side = ' ';
+	struct vsp_cmd_data vsp_cmd;
+
+	memset(&vsp_cmd, 0, sizeof(vsp_cmd));
+	vsp_cmd.cmd = 2;
+	vsp_cmd.sub_data.ipl_type = 0;
+	mb();
+
+	if (signal_vsp_instruction(&vsp_cmd) == 0) {
+		if (vsp_cmd.result_code == 0) {
+			switch (vsp_cmd.sub_data.ipl_type) {
+			case 0:	mf_current_side = 'A';
+				break;
+			case 1:	mf_current_side = 'B';
+				break;
+			case 2:	mf_current_side = 'C';
+				break;
+			default:	mf_current_side = 'D';
+				break;
+			}
+		}
+	}
+
+	len = sprintf(page, "%c\n", mf_current_side);
+
+	if (len <= (off + count))
+		*eof = 1;
+	*start = page + off;
+	len -= off;
+	if (len > count)
+		len = count;
+	if (len < 0)
+		len = 0;
+	return len;
+}
+
+static int proc_mf_change_side(struct file *file, const char __user *buffer,
+		unsigned long count, void *data)
+{
+	char side;
+	u64 newSide;
+	struct vsp_cmd_data vsp_cmd;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EACCES;
+
+	if (count == 0)
+		return 0;
+
+	if (get_user(side, buffer))
+		return -EFAULT;
+
+	switch (side) {
+	case 'A':	newSide = 0;
+			break;
+	case 'B':	newSide = 1;
+			break;
+	case 'C':	newSide = 2;
+			break;
+	case 'D':	newSide = 3;
+			break;
+	default:
+		printk(KERN_ERR "mf_proc.c: proc_mf_change_side: invalid side\n");
+		return -EINVAL;
+	}
+
+	memset(&vsp_cmd, 0, sizeof(vsp_cmd));
+	vsp_cmd.sub_data.ipl_type = newSide;
+	vsp_cmd.cmd = 10;
+
+	(void)signal_vsp_instruction(&vsp_cmd);
+
+	return count;
+}
+
+#if 0
+static void mf_getSrcHistory(char *buffer, int size)
+{
+	struct IplTypeReturnStuff return_stuff;
+	struct pending_event *ev = new_pending_event();
+	int rc = 0;
+	char *pages[4];
+
+	pages[0] = kmalloc(4096, GFP_ATOMIC);
+	pages[1] = kmalloc(4096, GFP_ATOMIC);
+	pages[2] = kmalloc(4096, GFP_ATOMIC);
+	pages[3] = kmalloc(4096, GFP_ATOMIC);
+	if ((ev == NULL) || (pages[0] == NULL) || (pages[1] == NULL)
+			 || (pages[2] == NULL) || (pages[3] == NULL))
+		return -ENOMEM;
+
+	return_stuff.xType = 0;
+	return_stuff.xRc = 0;
+	return_stuff.xDone = 0;
+	ev->event.hp_lp_event.xSubtype = 6;
+	ev->event.hp_lp_event.x.xSubtypeData =
+		subtype_data('M', 'F', 'V', 'I');
+	ev->event.data.vsp_cmd.xEvent = &return_stuff;
+	ev->event.data.vsp_cmd.cmd = 4;
+	ev->event.data.vsp_cmd.lp_index = HvLpConfig_getLpIndex();
+	ev->event.data.vsp_cmd.result_code = 0xFF;
+	ev->event.data.vsp_cmd.reserved = 0;
+	ev->event.data.vsp_cmd.sub_data.page[0] = ISERIES_HV_ADDR(pages[0]);
+	ev->event.data.vsp_cmd.sub_data.page[1] = ISERIES_HV_ADDR(pages[1]);
+	ev->event.data.vsp_cmd.sub_data.page[2] = ISERIES_HV_ADDR(pages[2]);
+	ev->event.data.vsp_cmd.sub_data.page[3] = ISERIES_HV_ADDR(pages[3]);
+	mb();
+	if (signal_event(ev) != 0)
+		return;
+
+ 	while (return_stuff.xDone != 1)
+ 		udelay(10);
+ 	if (return_stuff.xRc == 0)
+ 		memcpy(buffer, pages[0], size);
+	kfree(pages[0]);
+	kfree(pages[1]);
+	kfree(pages[2]);
+	kfree(pages[3]);
+}
+#endif
+
+static int proc_mf_dump_src(char *page, char **start, off_t off,
+		int count, int *eof, void *data)
+{
+#if 0
+	int len;
+
+	mf_getSrcHistory(page, count);
+	len = count;
+	len -= off;
+	if (len < count) {
+		*eof = 1;
+		if (len <= 0)
+			return 0;
+	} else
+		len = count;
+	*start = page + off;
+	return len;
+#else
+	return 0;
+#endif
+}
+
+static int proc_mf_change_src(struct file *file, const char __user *buffer,
+		unsigned long count, void *data)
+{
+	char stkbuf[10];
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EACCES;
+
+	if ((count < 4) && (count != 1)) {
+		printk(KERN_ERR "mf_proc: invalid src\n");
+		return -EINVAL;
+	}
+
+	if (count > (sizeof(stkbuf) - 1))
+		count = sizeof(stkbuf) - 1;
+	if (copy_from_user(stkbuf, buffer, count))
+		return -EFAULT;
+
+	if ((count == 1) && (*stkbuf == '\0'))
+		mf_clear_src();
+	else
+		mf_display_src(*(u32 *)stkbuf);
+
+	return count;
+}
+
+static int proc_mf_change_cmdline(struct file *file, const char __user *buffer,
+		unsigned long count, void *data)
+{
+	struct vsp_cmd_data vsp_cmd;
+	dma_addr_t dma_addr;
+	char *page;
+	int ret = -EACCES;
+
+	if (!capable(CAP_SYS_ADMIN))
+		goto out;
+
+	dma_addr = 0;
+	page = dma_alloc_coherent(iSeries_vio_dev, count, &dma_addr,
+			GFP_ATOMIC);
+	ret = -ENOMEM;
+	if (page == NULL)
+		goto out;
+
+	ret = -EFAULT;
+	if (copy_from_user(page, buffer, count))
+		goto out_free;
+
+	memset(&vsp_cmd, 0, sizeof(vsp_cmd));
+	vsp_cmd.cmd = 31;
+	vsp_cmd.sub_data.kern.token = dma_addr;
+	vsp_cmd.sub_data.kern.address_type = HvLpDma_AddressType_TceIndex;
+	vsp_cmd.sub_data.kern.side = (u64)data;
+	vsp_cmd.sub_data.kern.length = count;
+	mb();
+	(void)signal_vsp_instruction(&vsp_cmd);
+	ret = count;
+
+out_free:
+	dma_free_coherent(iSeries_vio_dev, count, page, dma_addr);
+out:
+	return ret;
+}
+
+static ssize_t proc_mf_change_vmlinux(struct file *file,
+				      const char __user *buf,
+				      size_t count, loff_t *ppos)
+{
+	struct proc_dir_entry *dp = PDE(file->f_dentry->d_inode);
+	ssize_t rc;
+	dma_addr_t dma_addr;
+	char *page;
+	struct vsp_cmd_data vsp_cmd;
+
+	rc = -EACCES;
+	if (!capable(CAP_SYS_ADMIN))
+		goto out;
+
+	dma_addr = 0;
+	page = dma_alloc_coherent(iSeries_vio_dev, count, &dma_addr,
+			GFP_ATOMIC);
+	rc = -ENOMEM;
+	if (page == NULL) {
+		printk(KERN_ERR "mf.c: couldn't allocate memory to set vmlinux chunk\n");
+		goto out;
+	}
+	rc = -EFAULT;
+	if (copy_from_user(page, buf, count))
+		goto out_free;
+
+	memset(&vsp_cmd, 0, sizeof(vsp_cmd));
+	vsp_cmd.cmd = 30;
+	vsp_cmd.sub_data.kern.token = dma_addr;
+	vsp_cmd.sub_data.kern.address_type = HvLpDma_AddressType_TceIndex;
+	vsp_cmd.sub_data.kern.side = (u64)dp->data;
+	vsp_cmd.sub_data.kern.offset = *ppos;
+	vsp_cmd.sub_data.kern.length = count;
+	mb();
+	rc = signal_vsp_instruction(&vsp_cmd);
+	if (rc)
+		goto out_free;
+	rc = -ENOMEM;
+	if (vsp_cmd.result_code != 0)
+		goto out_free;
+
+	*ppos += count;
+	rc = count;
+out_free:
+	dma_free_coherent(iSeries_vio_dev, count, page, dma_addr);
+out:
+	return rc;
+}
+
+static struct file_operations proc_vmlinux_operations = {
+	.write		= proc_mf_change_vmlinux,
+};
+
+static int __init mf_proc_init(void)
+{
+	struct proc_dir_entry *mf_proc_root;
+	struct proc_dir_entry *ent;
+	struct proc_dir_entry *mf;
+	char name[2];
+	int i;
+
+	mf_proc_root = proc_mkdir("iSeries/mf", NULL);
+	if (!mf_proc_root)
+		return 1;
+
+	name[1] = '\0';
+	for (i = 0; i < 4; i++) {
+		name[0] = 'A' + i;
+		mf = proc_mkdir(name, mf_proc_root);
+		if (!mf)
+			return 1;
+
+		ent = create_proc_entry("cmdline", S_IFREG|S_IRUSR|S_IWUSR, mf);
+		if (!ent)
+			return 1;
+		ent->nlink = 1;
+		ent->data = (void *)(long)i;
+		ent->read_proc = proc_mf_dump_cmdline;
+		ent->write_proc = proc_mf_change_cmdline;
+
+		if (i == 3)	/* no vmlinux entry for 'D' */
+			continue;
+
+		ent = create_proc_entry("vmlinux", S_IFREG|S_IWUSR, mf);
+		if (!ent)
+			return 1;
+		ent->nlink = 1;
+		ent->data = (void *)(long)i;
+		ent->proc_fops = &proc_vmlinux_operations;
+	}
+
+	ent = create_proc_entry("side", S_IFREG|S_IRUSR|S_IWUSR, mf_proc_root);
+	if (!ent)
+		return 1;
+	ent->nlink = 1;
+	ent->data = (void *)0;
+	ent->read_proc = proc_mf_dump_side;
+	ent->write_proc = proc_mf_change_side;
+
+	ent = create_proc_entry("src", S_IFREG|S_IRUSR|S_IWUSR, mf_proc_root);
+	if (!ent)
+		return 1;
+	ent->nlink = 1;
+	ent->data = (void *)0;
+	ent->read_proc = proc_mf_dump_src;
+	ent->write_proc = proc_mf_change_src;
+
+	return 0;
+}
+
+__initcall(mf_proc_init);
+
+#endif /* CONFIG_PROC_FS */
+
+/*
+ * Get the RTC from the virtual service processor
+ * This requires flowing LpEvents to the primary partition
+ */
+void iSeries_get_rtc_time(struct rtc_time *rtc_tm)
+{
+	if (piranha_simulator)
+		return;
+
+	mf_get_rtc(rtc_tm);
+	rtc_tm->tm_mon--;
+}
+
+/*
+ * Set the RTC in the virtual service processor
+ * This requires flowing LpEvents to the primary partition
+ */
+int iSeries_set_rtc_time(struct rtc_time *tm)
+{
+	mf_set_rtc(tm);
+	return 0;
+}
+
+void iSeries_get_boot_time(struct rtc_time *tm)
+{
+	if (piranha_simulator)
+		return;
+
+	mf_get_boot_rtc(tm);
+	tm->tm_mon  -= 1;
+}
diff --git a/arch/powerpc/platforms/iseries/setup.c b/arch/powerpc/platforms/iseries/setup.c
new file mode 100644
index 00000000000..ad78c8581a5
--- /dev/null
+++ b/arch/powerpc/platforms/iseries/setup.c
@@ -0,0 +1,1006 @@
+/*
+ *    Copyright (c) 2000 Mike Corrigan <mikejc@us.ibm.com>
+ *    Copyright (c) 1999-2000 Grant Erickson <grant@lcse.umn.edu>
+ *
+ *    Description:
+ *      Architecture- / platform-specific boot-time initialization code for
+ *      the IBM iSeries LPAR.  Adapted from original code by Grant Erickson and
+ *      code by Gary Thomas, Cort Dougan <cort@fsmlabs.com>, and Dan Malek
+ *      <dan@net4x.com>.
+ *
+ *      This program is free software; you can redistribute it and/or
+ *      modify it under the terms of the GNU General Public License
+ *      as published by the Free Software Foundation; either version
+ *      2 of the License, or (at your option) any later version.
+ */
+
+#undef DEBUG
+
+#include <linux/config.h>
+#include <linux/init.h>
+#include <linux/threads.h>
+#include <linux/smp.h>
+#include <linux/param.h>
+#include <linux/string.h>
+#include <linux/initrd.h>
+#include <linux/seq_file.h>
+#include <linux/kdev_t.h>
+#include <linux/major.h>
+#include <linux/root_dev.h>
+
+#include <asm/processor.h>
+#include <asm/machdep.h>
+#include <asm/page.h>
+#include <asm/mmu.h>
+#include <asm/pgtable.h>
+#include <asm/mmu_context.h>
+#include <asm/cputable.h>
+#include <asm/sections.h>
+#include <asm/iommu.h>
+#include <asm/firmware.h>
+
+#include <asm/time.h>
+#include <asm/naca.h>
+#include <asm/paca.h>
+#include <asm/cache.h>
+#include <asm/sections.h>
+#include <asm/abs_addr.h>
+#include <asm/iSeries/HvCallHpt.h>
+#include <asm/iSeries/HvLpConfig.h>
+#include <asm/iSeries/HvCallEvent.h>
+#include <asm/iSeries/HvCallSm.h>
+#include <asm/iSeries/HvCallXm.h>
+#include <asm/iSeries/ItLpQueue.h>
+#include <asm/iSeries/IoHriMainStore.h>
+#include <asm/iSeries/mf.h>
+#include <asm/iSeries/HvLpEvent.h>
+#include <asm/iSeries/iSeries_irq.h>
+#include <asm/iSeries/IoHriProcessorVpd.h>
+#include <asm/iSeries/ItVpdAreas.h>
+#include <asm/iSeries/LparMap.h>
+
+#include "setup.h"
+
+extern void hvlog(char *fmt, ...);
+
+#ifdef DEBUG
+#define DBG(fmt...) hvlog(fmt)
+#else
+#define DBG(fmt...)
+#endif
+
+/* Function Prototypes */
+extern void ppcdbg_initialize(void);
+
+static void build_iSeries_Memory_Map(void);
+static int iseries_shared_idle(void);
+static int iseries_dedicated_idle(void);
+#ifdef CONFIG_PCI
+extern void iSeries_pci_final_fixup(void);
+#else
+static void iSeries_pci_final_fixup(void) { }
+#endif
+
+/* Global Variables */
+int piranha_simulator;
+
+extern int rd_size;		/* Defined in drivers/block/rd.c */
+extern unsigned long klimit;
+extern unsigned long embedded_sysmap_start;
+extern unsigned long embedded_sysmap_end;
+
+extern unsigned long iSeries_recal_tb;
+extern unsigned long iSeries_recal_titan;
+
+static int mf_initialized;
+
+struct MemoryBlock {
+	unsigned long absStart;
+	unsigned long absEnd;
+	unsigned long logicalStart;
+	unsigned long logicalEnd;
+};
+
+/*
+ * Process the main store vpd to determine where the holes in memory are
+ * and return the number of physical blocks and fill in the array of
+ * block data.
+ */
+static unsigned long iSeries_process_Condor_mainstore_vpd(
+		struct MemoryBlock *mb_array, unsigned long max_entries)
+{
+	unsigned long holeFirstChunk, holeSizeChunks;
+	unsigned long numMemoryBlocks = 1;
+	struct IoHriMainStoreSegment4 *msVpd =
+		(struct IoHriMainStoreSegment4 *)xMsVpd;
+	unsigned long holeStart = msVpd->nonInterleavedBlocksStartAdr;
+	unsigned long holeEnd = msVpd->nonInterleavedBlocksEndAdr;
+	unsigned long holeSize = holeEnd - holeStart;
+
+	printk("Mainstore_VPD: Condor\n");
+	/*
+	 * Determine if absolute memory has any
+	 * holes so that we can interpret the
+	 * access map we get back from the hypervisor
+	 * correctly.
+	 */
+	mb_array[0].logicalStart = 0;
+	mb_array[0].logicalEnd = 0x100000000;
+	mb_array[0].absStart = 0;
+	mb_array[0].absEnd = 0x100000000;
+
+	if (holeSize) {
+		numMemoryBlocks = 2;
+		holeStart = holeStart & 0x000fffffffffffff;
+		holeStart = addr_to_chunk(holeStart);
+		holeFirstChunk = holeStart;
+		holeSize = addr_to_chunk(holeSize);
+		holeSizeChunks = holeSize;
+		printk( "Main store hole: start chunk = %0lx, size = %0lx chunks\n",
+				holeFirstChunk, holeSizeChunks );
+		mb_array[0].logicalEnd = holeFirstChunk;
+		mb_array[0].absEnd = holeFirstChunk;
+		mb_array[1].logicalStart = holeFirstChunk;
+		mb_array[1].logicalEnd = 0x100000000 - holeSizeChunks;
+		mb_array[1].absStart = holeFirstChunk + holeSizeChunks;
+		mb_array[1].absEnd = 0x100000000;
+	}
+	return numMemoryBlocks;
+}
+
+#define MaxSegmentAreas			32
+#define MaxSegmentAdrRangeBlocks	128
+#define MaxAreaRangeBlocks		4
+
+static unsigned long iSeries_process_Regatta_mainstore_vpd(
+		struct MemoryBlock *mb_array, unsigned long max_entries)
+{
+	struct IoHriMainStoreSegment5 *msVpdP =
+		(struct IoHriMainStoreSegment5 *)xMsVpd;
+	unsigned long numSegmentBlocks = 0;
+	u32 existsBits = msVpdP->msAreaExists;
+	unsigned long area_num;
+
+	printk("Mainstore_VPD: Regatta\n");
+
+	for (area_num = 0; area_num < MaxSegmentAreas; ++area_num ) {
+		unsigned long numAreaBlocks;
+		struct IoHriMainStoreArea4 *currentArea;
+
+		if (existsBits & 0x80000000) {
+			unsigned long block_num;
+
+			currentArea = &msVpdP->msAreaArray[area_num];
+			numAreaBlocks = currentArea->numAdrRangeBlocks;
+			printk("ms_vpd: processing area %2ld  blocks=%ld",
+					area_num, numAreaBlocks);
+			for (block_num = 0; block_num < numAreaBlocks;
+					++block_num ) {
+				/* Process an address range block */
+				struct MemoryBlock tempBlock;
+				unsigned long i;
+
+				tempBlock.absStart =
+					(unsigned long)currentArea->xAdrRangeBlock[block_num].blockStart;
+				tempBlock.absEnd =
+					(unsigned long)currentArea->xAdrRangeBlock[block_num].blockEnd;
+				tempBlock.logicalStart = 0;
+				tempBlock.logicalEnd   = 0;
+				printk("\n          block %ld absStart=%016lx absEnd=%016lx",
+						block_num, tempBlock.absStart,
+						tempBlock.absEnd);
+
+				for (i = 0; i < numSegmentBlocks; ++i) {
+					if (mb_array[i].absStart ==
+							tempBlock.absStart)
+						break;
+				}
+				if (i == numSegmentBlocks) {
+					if (numSegmentBlocks == max_entries)
+						panic("iSeries_process_mainstore_vpd: too many memory blocks");
+					mb_array[numSegmentBlocks] = tempBlock;
+					++numSegmentBlocks;
+				} else
+					printk(" (duplicate)");
+			}
+			printk("\n");
+		}
+		existsBits <<= 1;
+	}
+	/* Now sort the blocks found into ascending sequence */
+	if (numSegmentBlocks > 1) {
+		unsigned long m, n;
+
+		for (m = 0; m < numSegmentBlocks - 1; ++m) {
+			for (n = numSegmentBlocks - 1; m < n; --n) {
+				if (mb_array[n].absStart <
+						mb_array[n-1].absStart) {
+					struct MemoryBlock tempBlock;
+
+					tempBlock = mb_array[n];
+					mb_array[n] = mb_array[n-1];
+					mb_array[n-1] = tempBlock;
+				}
+			}
+		}
+	}
+	/*
+	 * Assign "logical" addresses to each block.  These
+	 * addresses correspond to the hypervisor "bitmap" space.
+	 * Convert all addresses into units of 256K chunks.
+	 */
+	{
+	unsigned long i, nextBitmapAddress;
+
+	printk("ms_vpd: %ld sorted memory blocks\n", numSegmentBlocks);
+	nextBitmapAddress = 0;
+	for (i = 0; i < numSegmentBlocks; ++i) {
+		unsigned long length = mb_array[i].absEnd -
+			mb_array[i].absStart;
+
+		mb_array[i].logicalStart = nextBitmapAddress;
+		mb_array[i].logicalEnd = nextBitmapAddress + length;
+		nextBitmapAddress += length;
+		printk("          Bitmap range: %016lx - %016lx\n"
+				"        Absolute range: %016lx - %016lx\n",
+				mb_array[i].logicalStart,
+				mb_array[i].logicalEnd,
+				mb_array[i].absStart, mb_array[i].absEnd);
+		mb_array[i].absStart = addr_to_chunk(mb_array[i].absStart &
+				0x000fffffffffffff);
+		mb_array[i].absEnd = addr_to_chunk(mb_array[i].absEnd &
+				0x000fffffffffffff);
+		mb_array[i].logicalStart =
+			addr_to_chunk(mb_array[i].logicalStart);
+		mb_array[i].logicalEnd = addr_to_chunk(mb_array[i].logicalEnd);
+	}
+	}
+
+	return numSegmentBlocks;
+}
+
+static unsigned long iSeries_process_mainstore_vpd(struct MemoryBlock *mb_array,
+		unsigned long max_entries)
+{
+	unsigned long i;
+	unsigned long mem_blocks = 0;
+
+	if (cpu_has_feature(CPU_FTR_SLB))
+		mem_blocks = iSeries_process_Regatta_mainstore_vpd(mb_array,
+				max_entries);
+	else
+		mem_blocks = iSeries_process_Condor_mainstore_vpd(mb_array,
+				max_entries);
+
+	printk("Mainstore_VPD: numMemoryBlocks = %ld \n", mem_blocks);
+	for (i = 0; i < mem_blocks; ++i) {
+		printk("Mainstore_VPD: block %3ld logical chunks %016lx - %016lx\n"
+		       "                             abs chunks %016lx - %016lx\n",
+			i, mb_array[i].logicalStart, mb_array[i].logicalEnd,
+			mb_array[i].absStart, mb_array[i].absEnd);
+	}
+	return mem_blocks;
+}
+
+static void __init iSeries_get_cmdline(void)
+{
+	char *p, *q;
+
+	/* copy the command line parameter from the primary VSP  */
+	HvCallEvent_dmaToSp(cmd_line, 2 * 64* 1024, 256,
+			HvLpDma_Direction_RemoteToLocal);
+
+	p = cmd_line;
+	q = cmd_line + 255;
+	while(p < q) {
+		if (!*p || *p == '\n')
+			break;
+		++p;
+	}
+	*p = 0;
+}
+
+static void __init iSeries_init_early(void)
+{
+	extern unsigned long memory_limit;
+
+	DBG(" -> iSeries_init_early()\n");
+
+	ppc64_firmware_features = FW_FEATURE_ISERIES;
+
+	ppcdbg_initialize();
+
+	ppc64_interrupt_controller = IC_ISERIES;
+
+#if defined(CONFIG_BLK_DEV_INITRD)
+	/*
+	 * If the init RAM disk has been configured and there is
+	 * a non-zero starting address for it, set it up
+	 */
+	if (naca.xRamDisk) {
+		initrd_start = (unsigned long)__va(naca.xRamDisk);
+		initrd_end = initrd_start + naca.xRamDiskSize * PAGE_SIZE;
+		initrd_below_start_ok = 1;	// ramdisk in kernel space
+		ROOT_DEV = Root_RAM0;
+		if (((rd_size * 1024) / PAGE_SIZE) < naca.xRamDiskSize)
+			rd_size = (naca.xRamDiskSize * PAGE_SIZE) / 1024;
+	} else
+#endif /* CONFIG_BLK_DEV_INITRD */
+	{
+	    /* ROOT_DEV = MKDEV(VIODASD_MAJOR, 1); */
+	}
+
+	iSeries_recal_tb = get_tb();
+	iSeries_recal_titan = HvCallXm_loadTod();
+
+	/*
+	 * Initialize the hash table management pointers
+	 */
+	hpte_init_iSeries();
+
+	/*
+	 * Initialize the DMA/TCE management
+	 */
+	iommu_init_early_iSeries();
+
+	iSeries_get_cmdline();
+
+	/* Save unparsed command line copy for /proc/cmdline */
+	strlcpy(saved_command_line, cmd_line, COMMAND_LINE_SIZE);
+
+	/* Parse early parameters, in particular mem=x */
+	parse_early_param();
+
+	if (memory_limit) {
+		if (memory_limit < systemcfg->physicalMemorySize)
+			systemcfg->physicalMemorySize = memory_limit;
+		else {
+			printk("Ignoring mem=%lu >= ram_top.\n", memory_limit);
+			memory_limit = 0;
+		}
+	}
+
+	/* Initialize machine-dependency vectors */
+#ifdef CONFIG_SMP
+	smp_init_iSeries();
+#endif
+	if (itLpNaca.xPirEnvironMode == 0)
+		piranha_simulator = 1;
+
+	/* Associate Lp Event Queue 0 with processor 0 */
+	HvCallEvent_setLpEventQueueInterruptProc(0, 0);
+
+	mf_init();
+	mf_initialized = 1;
+	mb();
+
+	/* If we were passed an initrd, set the ROOT_DEV properly if the values
+	 * look sensible. If not, clear initrd reference.
+	 */
+#ifdef CONFIG_BLK_DEV_INITRD
+	if (initrd_start >= KERNELBASE && initrd_end >= KERNELBASE &&
+	    initrd_end > initrd_start)
+		ROOT_DEV = Root_RAM0;
+	else
+		initrd_start = initrd_end = 0;
+#endif /* CONFIG_BLK_DEV_INITRD */
+
+	DBG(" <- iSeries_init_early()\n");
+}
+
+struct mschunks_map mschunks_map = {
+	/* XXX We don't use these, but Piranha might need them. */
+	.chunk_size  = MSCHUNKS_CHUNK_SIZE,
+	.chunk_shift = MSCHUNKS_CHUNK_SHIFT,
+	.chunk_mask  = MSCHUNKS_OFFSET_MASK,
+};
+EXPORT_SYMBOL(mschunks_map);
+
+void mschunks_alloc(unsigned long num_chunks)
+{
+	klimit = _ALIGN(klimit, sizeof(u32));
+	mschunks_map.mapping = (u32 *)klimit;
+	klimit += num_chunks * sizeof(u32);
+	mschunks_map.num_chunks = num_chunks;
+}
+
+/*
+ * The iSeries may have very large memories ( > 128 GB ) and a partition
+ * may get memory in "chunks" that may be anywhere in the 2**52 real
+ * address space.  The chunks are 256K in size.  To map this to the
+ * memory model Linux expects, the AS/400 specific code builds a
+ * translation table to translate what Linux thinks are "physical"
+ * addresses to the actual real addresses.  This allows us to make
+ * it appear to Linux that we have contiguous memory starting at
+ * physical address zero while in fact this could be far from the truth.
+ * To avoid confusion, I'll let the words physical and/or real address
+ * apply to the Linux addresses while I'll use "absolute address" to
+ * refer to the actual hardware real address.
+ *
+ * build_iSeries_Memory_Map gets information from the Hypervisor and
+ * looks at the Main Store VPD to determine the absolute addresses
+ * of the memory that has been assigned to our partition and builds
+ * a table used to translate Linux's physical addresses to these
+ * absolute addresses.  Absolute addresses are needed when
+ * communicating with the hypervisor (e.g. to build HPT entries)
+ */
+
+static void __init build_iSeries_Memory_Map(void)
+{
+	u32 loadAreaFirstChunk, loadAreaLastChunk, loadAreaSize;
+	u32 nextPhysChunk;
+	u32 hptFirstChunk, hptLastChunk, hptSizeChunks, hptSizePages;
+	u32 num_ptegs;
+	u32 totalChunks,moreChunks;
+	u32 currChunk, thisChunk, absChunk;
+	u32 currDword;
+	u32 chunkBit;
+	u64 map;
+	struct MemoryBlock mb[32];
+	unsigned long numMemoryBlocks, curBlock;
+
+	/* Chunk size on iSeries is 256K bytes */
+	totalChunks = (u32)HvLpConfig_getMsChunks();
+	mschunks_alloc(totalChunks);
+
+	/*
+	 * Get absolute address of our load area
+	 * and map it to physical address 0
+	 * This guarantees that the loadarea ends up at physical 0
+	 * otherwise, it might not be returned by PLIC as the first
+	 * chunks
+	 */
+
+	loadAreaFirstChunk = (u32)addr_to_chunk(itLpNaca.xLoadAreaAddr);
+	loadAreaSize =  itLpNaca.xLoadAreaChunks;
+
+	/*
+	 * Only add the pages already mapped here.
+	 * Otherwise we might add the hpt pages
+	 * The rest of the pages of the load area
+	 * aren't in the HPT yet and can still
+	 * be assigned an arbitrary physical address
+	 */
+	if ((loadAreaSize * 64) > HvPagesToMap)
+		loadAreaSize = HvPagesToMap / 64;
+
+	loadAreaLastChunk = loadAreaFirstChunk + loadAreaSize - 1;
+
+	/*
+	 * TODO Do we need to do something if the HPT is in the 64MB load area?
+	 * This would be required if the itLpNaca.xLoadAreaChunks includes
+	 * the HPT size
+	 */
+
+	printk("Mapping load area - physical addr = 0000000000000000\n"
+		"                    absolute addr = %016lx\n",
+		chunk_to_addr(loadAreaFirstChunk));
+	printk("Load area size %dK\n", loadAreaSize * 256);
+
+	for (nextPhysChunk = 0; nextPhysChunk < loadAreaSize; ++nextPhysChunk)
+		mschunks_map.mapping[nextPhysChunk] =
+			loadAreaFirstChunk + nextPhysChunk;
+
+	/*
+	 * Get absolute address of our HPT and remember it so
+	 * we won't map it to any physical address
+	 */
+	hptFirstChunk = (u32)addr_to_chunk(HvCallHpt_getHptAddress());
+	hptSizePages = (u32)HvCallHpt_getHptPages();
+	hptSizeChunks = hptSizePages >> (MSCHUNKS_CHUNK_SHIFT - PAGE_SHIFT);
+	hptLastChunk = hptFirstChunk + hptSizeChunks - 1;
+
+	printk("HPT absolute addr = %016lx, size = %dK\n",
+			chunk_to_addr(hptFirstChunk), hptSizeChunks * 256);
+
+	/* Fill in the hashed page table hash mask */
+	num_ptegs = hptSizePages *
+		(PAGE_SIZE / (sizeof(hpte_t) * HPTES_PER_GROUP));
+	htab_hash_mask = num_ptegs - 1;
+
+	/*
+	 * The actual hashed page table is in the hypervisor,
+	 * we have no direct access
+	 */
+	htab_address = NULL;
+
+	/*
+	 * Determine if absolute memory has any
+	 * holes so that we can interpret the
+	 * access map we get back from the hypervisor
+	 * correctly.
+	 */
+	numMemoryBlocks = iSeries_process_mainstore_vpd(mb, 32);
+
+	/*
+	 * Process the main store access map from the hypervisor
+	 * to build up our physical -> absolute translation table
+	 */
+	curBlock = 0;
+	currChunk = 0;
+	currDword = 0;
+	moreChunks = totalChunks;
+
+	while (moreChunks) {
+		map = HvCallSm_get64BitsOfAccessMap(itLpNaca.xLpIndex,
+				currDword);
+		thisChunk = currChunk;
+		while (map) {
+			chunkBit = map >> 63;
+			map <<= 1;
+			if (chunkBit) {
+				--moreChunks;
+				while (thisChunk >= mb[curBlock].logicalEnd) {
+					++curBlock;
+					if (curBlock >= numMemoryBlocks)
+						panic("out of memory blocks");
+				}
+				if (thisChunk < mb[curBlock].logicalStart)
+					panic("memory block error");
+
+				absChunk = mb[curBlock].absStart +
+					(thisChunk - mb[curBlock].logicalStart);
+				if (((absChunk < hptFirstChunk) ||
+				     (absChunk > hptLastChunk)) &&
+				    ((absChunk < loadAreaFirstChunk) ||
+				     (absChunk > loadAreaLastChunk))) {
+					mschunks_map.mapping[nextPhysChunk] =
+						absChunk;
+					++nextPhysChunk;
+				}
+			}
+			++thisChunk;
+		}
+		++currDword;
+		currChunk += 64;
+	}
+
+	/*
+	 * main store size (in chunks) is
+	 *   totalChunks - hptSizeChunks
+	 * which should be equal to
+	 *   nextPhysChunk
+	 */
+	systemcfg->physicalMemorySize = chunk_to_addr(nextPhysChunk);
+}
+
+/*
+ * Document me.
+ */
+static void __init iSeries_setup_arch(void)
+{
+	unsigned procIx = get_paca()->lppaca.dyn_hv_phys_proc_index;
+
+	if (get_paca()->lppaca.shared_proc) {
+		ppc_md.idle_loop = iseries_shared_idle;
+		printk(KERN_INFO "Using shared processor idle loop\n");
+	} else {
+		ppc_md.idle_loop = iseries_dedicated_idle;
+		printk(KERN_INFO "Using dedicated idle loop\n");
+	}
+
+	/* Setup the Lp Event Queue */
+	setup_hvlpevent_queue();
+
+	printk("Max  logical processors = %d\n",
+			itVpdAreas.xSlicMaxLogicalProcs);
+	printk("Max physical processors = %d\n",
+			itVpdAreas.xSlicMaxPhysicalProcs);
+
+	systemcfg->processor = xIoHriProcessorVpd[procIx].xPVR;
+	printk("Processor version = %x\n", systemcfg->processor);
+}
+
+static void iSeries_get_cpuinfo(struct seq_file *m)
+{
+	seq_printf(m, "machine\t\t: 64-bit iSeries Logical Partition\n");
+}
+
+/*
+ * Document me.
+ * and Implement me.
+ */
+static int iSeries_get_irq(struct pt_regs *regs)
+{
+	/* -2 means ignore this interrupt */
+	return -2;
+}
+
+/*
+ * Document me.
+ */
+static void iSeries_restart(char *cmd)
+{
+	mf_reboot();
+}
+
+/*
+ * Document me.
+ */
+static void iSeries_power_off(void)
+{
+	mf_power_off();
+}
+
+/*
+ * Document me.
+ */
+static void iSeries_halt(void)
+{
+	mf_power_off();
+}
+
+static void __init iSeries_progress(char * st, unsigned short code)
+{
+	printk("Progress: [%04x] - %s\n", (unsigned)code, st);
+	if (!piranha_simulator && mf_initialized) {
+		if (code != 0xffff)
+			mf_display_progress(code);
+		else
+			mf_clear_src();
+	}
+}
+
+static void __init iSeries_fixup_klimit(void)
+{
+	/*
+	 * Change klimit to take into account any ram disk
+	 * that may be included
+	 */
+	if (naca.xRamDisk)
+		klimit = KERNELBASE + (u64)naca.xRamDisk +
+			(naca.xRamDiskSize * PAGE_SIZE);
+	else {
+		/*
+		 * No ram disk was included - check and see if there
+		 * was an embedded system map.  Change klimit to take
+		 * into account any embedded system map
+		 */
+		if (embedded_sysmap_end)
+			klimit = KERNELBASE + ((embedded_sysmap_end + 4095) &
+					0xfffffffffffff000);
+	}
+}
+
+static int __init iSeries_src_init(void)
+{
+        /* clear the progress line */
+        ppc_md.progress(" ", 0xffff);
+        return 0;
+}
+
+late_initcall(iSeries_src_init);
+
+static inline void process_iSeries_events(void)
+{
+	asm volatile ("li 0,0x5555; sc" : : : "r0", "r3");
+}
+
+static void yield_shared_processor(void)
+{
+	unsigned long tb;
+
+	HvCall_setEnabledInterrupts(HvCall_MaskIPI |
+				    HvCall_MaskLpEvent |
+				    HvCall_MaskLpProd |
+				    HvCall_MaskTimeout);
+
+	tb = get_tb();
+	/* Compute future tb value when yield should expire */
+	HvCall_yieldProcessor(HvCall_YieldTimed, tb+tb_ticks_per_jiffy);
+
+	/*
+	 * The decrementer stops during the yield.  Force a fake decrementer
+	 * here and let the timer_interrupt code sort out the actual time.
+	 */
+	get_paca()->lppaca.int_dword.fields.decr_int = 1;
+	process_iSeries_events();
+}
+
+static int iseries_shared_idle(void)
+{
+	while (1) {
+		while (!need_resched() && !hvlpevent_is_pending()) {
+			local_irq_disable();
+			ppc64_runlatch_off();
+
+			/* Recheck with irqs off */
+			if (!need_resched() && !hvlpevent_is_pending())
+				yield_shared_processor();
+
+			HMT_medium();
+			local_irq_enable();
+		}
+
+		ppc64_runlatch_on();
+
+		if (hvlpevent_is_pending())
+			process_iSeries_events();
+
+		schedule();
+	}
+
+	return 0;
+}
+
+static int iseries_dedicated_idle(void)
+{
+	long oldval;
+
+	while (1) {
+		oldval = test_and_clear_thread_flag(TIF_NEED_RESCHED);
+
+		if (!oldval) {
+			set_thread_flag(TIF_POLLING_NRFLAG);
+
+			while (!need_resched()) {
+				ppc64_runlatch_off();
+				HMT_low();
+
+				if (hvlpevent_is_pending()) {
+					HMT_medium();
+					ppc64_runlatch_on();
+					process_iSeries_events();
+				}
+			}
+
+			HMT_medium();
+			clear_thread_flag(TIF_POLLING_NRFLAG);
+		} else {
+			set_need_resched();
+		}
+
+		ppc64_runlatch_on();
+		schedule();
+	}
+
+	return 0;
+}
+
+#ifndef CONFIG_PCI
+void __init iSeries_init_IRQ(void) { }
+#endif
+
+static int __init iseries_probe(int platform)
+{
+	return PLATFORM_ISERIES_LPAR == platform;
+}
+
+struct machdep_calls __initdata iseries_md = {
+	.setup_arch	= iSeries_setup_arch,
+	.get_cpuinfo	= iSeries_get_cpuinfo,
+	.init_IRQ	= iSeries_init_IRQ,
+	.get_irq	= iSeries_get_irq,
+	.init_early	= iSeries_init_early,
+	.pcibios_fixup	= iSeries_pci_final_fixup,
+	.restart	= iSeries_restart,
+	.power_off	= iSeries_power_off,
+	.halt		= iSeries_halt,
+	.get_boot_time	= iSeries_get_boot_time,
+	.set_rtc_time	= iSeries_set_rtc_time,
+	.get_rtc_time	= iSeries_get_rtc_time,
+	.calibrate_decr	= generic_calibrate_decr,
+	.progress	= iSeries_progress,
+	.probe		= iseries_probe,
+	/* XXX Implement enable_pmcs for iSeries */
+};
+
+struct blob {
+	unsigned char data[PAGE_SIZE];
+	unsigned long next;
+};
+
+struct iseries_flat_dt {
+	struct boot_param_header header;
+	u64 reserve_map[2];
+	struct blob dt;
+	struct blob strings;
+};
+
+struct iseries_flat_dt iseries_dt;
+
+void dt_init(struct iseries_flat_dt *dt)
+{
+	dt->header.off_mem_rsvmap =
+		offsetof(struct iseries_flat_dt, reserve_map);
+	dt->header.off_dt_struct = offsetof(struct iseries_flat_dt, dt);
+	dt->header.off_dt_strings = offsetof(struct iseries_flat_dt, strings);
+	dt->header.totalsize = sizeof(struct iseries_flat_dt);
+	dt->header.dt_strings_size = sizeof(struct blob);
+
+	/* There is no notion of hardware cpu id on iSeries */
+	dt->header.boot_cpuid_phys = smp_processor_id();
+
+	dt->dt.next = (unsigned long)&dt->dt.data;
+	dt->strings.next = (unsigned long)&dt->strings.data;
+
+	dt->header.magic = OF_DT_HEADER;
+	dt->header.version = 0x10;
+	dt->header.last_comp_version = 0x10;
+
+	dt->reserve_map[0] = 0;
+	dt->reserve_map[1] = 0;
+}
+
+void dt_check_blob(struct blob *b)
+{
+	if (b->next >= (unsigned long)&b->next) {
+		DBG("Ran out of space in flat device tree blob!\n");
+		BUG();
+	}
+}
+
+void dt_push_u32(struct iseries_flat_dt *dt, u32 value)
+{
+	*((u32*)dt->dt.next) = value;
+	dt->dt.next += sizeof(u32);
+
+	dt_check_blob(&dt->dt);
+}
+
+void dt_push_u64(struct iseries_flat_dt *dt, u64 value)
+{
+	*((u64*)dt->dt.next) = value;
+	dt->dt.next += sizeof(u64);
+
+	dt_check_blob(&dt->dt);
+}
+
+unsigned long dt_push_bytes(struct blob *blob, char *data, int len)
+{
+	unsigned long start = blob->next - (unsigned long)blob->data;
+
+	memcpy((char *)blob->next, data, len);
+	blob->next = _ALIGN(blob->next + len, 4);
+
+	dt_check_blob(blob);
+
+	return start;
+}
+
+void dt_start_node(struct iseries_flat_dt *dt, char *name)
+{
+	dt_push_u32(dt, OF_DT_BEGIN_NODE);
+	dt_push_bytes(&dt->dt, name, strlen(name) + 1);
+}
+
+#define dt_end_node(dt) dt_push_u32(dt, OF_DT_END_NODE)
+
+void dt_prop(struct iseries_flat_dt *dt, char *name, char *data, int len)
+{
+	unsigned long offset;
+
+	dt_push_u32(dt, OF_DT_PROP);
+
+	/* Length of the data */
+	dt_push_u32(dt, len);
+
+	/* Put the property name in the string blob. */
+	offset = dt_push_bytes(&dt->strings, name, strlen(name) + 1);
+
+	/* The offset of the properties name in the string blob. */
+	dt_push_u32(dt, (u32)offset);
+
+	/* The actual data. */
+	dt_push_bytes(&dt->dt, data, len);
+}
+
+void dt_prop_str(struct iseries_flat_dt *dt, char *name, char *data)
+{
+	dt_prop(dt, name, data, strlen(data) + 1); /* + 1 for NULL */
+}
+
+void dt_prop_u32(struct iseries_flat_dt *dt, char *name, u32 data)
+{
+	dt_prop(dt, name, (char *)&data, sizeof(u32));
+}
+
+void dt_prop_u64(struct iseries_flat_dt *dt, char *name, u64 data)
+{
+	dt_prop(dt, name, (char *)&data, sizeof(u64));
+}
+
+void dt_prop_u64_list(struct iseries_flat_dt *dt, char *name, u64 *data, int n)
+{
+	dt_prop(dt, name, (char *)data, sizeof(u64) * n);
+}
+
+void dt_prop_empty(struct iseries_flat_dt *dt, char *name)
+{
+	dt_prop(dt, name, NULL, 0);
+}
+
+void dt_cpus(struct iseries_flat_dt *dt)
+{
+	unsigned char buf[32];
+	unsigned char *p;
+	unsigned int i, index;
+	struct IoHriProcessorVpd *d;
+
+	/* yuck */
+	snprintf(buf, 32, "PowerPC,%s", cur_cpu_spec->cpu_name);
+	p = strchr(buf, ' ');
+	if (!p) p = buf + strlen(buf);
+
+	dt_start_node(dt, "cpus");
+	dt_prop_u32(dt, "#address-cells", 1);
+	dt_prop_u32(dt, "#size-cells", 0);
+
+	for (i = 0; i < NR_CPUS; i++) {
+		if (paca[i].lppaca.dyn_proc_status >= 2)
+			continue;
+
+		snprintf(p, 32 - (p - buf), "@%d", i);
+		dt_start_node(dt, buf);
+
+		dt_prop_str(dt, "device_type", "cpu");
+
+		index = paca[i].lppaca.dyn_hv_phys_proc_index;
+		d = &xIoHriProcessorVpd[index];
+
+		dt_prop_u32(dt, "i-cache-size", d->xInstCacheSize * 1024);
+		dt_prop_u32(dt, "i-cache-line-size", d->xInstCacheOperandSize);
+
+		dt_prop_u32(dt, "d-cache-size", d->xDataL1CacheSizeKB * 1024);
+		dt_prop_u32(dt, "d-cache-line-size", d->xDataCacheOperandSize);
+
+		/* magic conversions to Hz copied from old code */
+		dt_prop_u32(dt, "clock-frequency",
+			((1UL << 34) * 1000000) / d->xProcFreq);
+		dt_prop_u32(dt, "timebase-frequency",
+			((1UL << 32) * 1000000) / d->xTimeBaseFreq);
+
+		dt_prop_u32(dt, "reg", i);
+
+		dt_end_node(dt);
+	}
+
+	dt_end_node(dt);
+}
+
+void build_flat_dt(struct iseries_flat_dt *dt)
+{
+	u64 tmp[2];
+
+	dt_init(dt);
+
+	dt_start_node(dt, "");
+
+	dt_prop_u32(dt, "#address-cells", 2);
+	dt_prop_u32(dt, "#size-cells", 2);
+
+	/* /memory */
+	dt_start_node(dt, "memory@0");
+	dt_prop_str(dt, "name", "memory");
+	dt_prop_str(dt, "device_type", "memory");
+	tmp[0] = 0;
+	tmp[1] = systemcfg->physicalMemorySize;
+	dt_prop_u64_list(dt, "reg", tmp, 2);
+	dt_end_node(dt);
+
+	/* /chosen */
+	dt_start_node(dt, "chosen");
+	dt_prop_u32(dt, "linux,platform", PLATFORM_ISERIES_LPAR);
+	dt_end_node(dt);
+
+	dt_cpus(dt);
+
+	dt_end_node(dt);
+
+	dt_push_u32(dt, OF_DT_END);
+}
+
+void * __init iSeries_early_setup(void)
+{
+	iSeries_fixup_klimit();
+
+	/*
+	 * Initialize the table which translate Linux physical addresses to
+	 * AS/400 absolute addresses
+	 */
+	build_iSeries_Memory_Map();
+
+	build_flat_dt(&iseries_dt);
+
+	return (void *) __pa(&iseries_dt);
+}
diff --git a/arch/powerpc/platforms/iseries/setup.h b/arch/powerpc/platforms/iseries/setup.h
new file mode 100644
index 00000000000..6da89ae991c
--- /dev/null
+++ b/arch/powerpc/platforms/iseries/setup.h
@@ -0,0 +1,24 @@
+/*
+ *    Copyright (c) 2000 Mike Corrigan <mikejc@us.ibm.com>
+ *    Copyright (c) 1999-2000 Grant Erickson <grant@lcse.umn.edu>
+ *
+ *    Description:
+ *      Architecture- / platform-specific boot-time initialization code for
+ *      the IBM AS/400 LPAR. Adapted from original code by Grant Erickson and
+ *      code by Gary Thomas, Cort Dougan <cort@cs.nmt.edu>, and Dan Malek
+ *      <dan@netx4.com>.
+ *
+ *      This program is free software; you can redistribute it and/or
+ *      modify it under the terms of the GNU General Public License
+ *      as published by the Free Software Foundation; either version
+ *      2 of the License, or (at your option) any later version.
+ */
+
+#ifndef	__ISERIES_SETUP_H__
+#define	__ISERIES_SETUP_H__
+
+extern void iSeries_get_boot_time(struct rtc_time *tm);
+extern int iSeries_set_rtc_time(struct rtc_time *tm);
+extern void iSeries_get_rtc_time(struct rtc_time *tm);
+
+#endif /* __ISERIES_SETUP_H__ */