diff -urN linux-2.5/arch/ppc64/boot/Makefile linux-prom-clean/arch/ppc64/boot/Makefile
--- linux-2.5/arch/ppc64/boot/Makefile	2004-08-26 15:46:30.000000000 +1000
+++ linux-prom-clean/arch/ppc64/boot/Makefile	2004-09-17 14:34:55.000000000 +1000
@@ -31,6 +31,7 @@
 BOOTLD		:= $(CROSS32_COMPILE)ld
 BOOTLFLAGS	:= -Ttext 0x00400000 -e _start -T $(obj)/zImage.lds
 BOOTOBJCOPY	:= $(CROSS32_COMPILE)objcopy
+BOOTSTRIP	:= $(CROSS32_COMPILE)strip
 OBJCOPYFLAGS    := contents,alloc,load,readonly,data
 
 src-boot := crt0.S string.S prom.c main.c zlib.c imagesize.c div64.S
@@ -51,36 +52,31 @@
 #-----------------------------------------------------------
 # ELF sections within the zImage bootloader/wrapper
 #-----------------------------------------------------------
-required := vmlinux .config System.map
+required := vmlinux.strip
 initrd   := initrd
 
 obj-sec = $(foreach section, $(1), $(patsubst %,$(obj)/kernel-%.o, $(section)))
 src-sec = $(foreach section, $(1), $(patsubst %,$(obj)/kernel-%.c, $(section)))
 gz-sec  = $(foreach section, $(1), $(patsubst %,$(obj)/kernel-%.gz, $(section)))
 
-hostprogs-y		:= piggy addnote addSystemMap addRamDisk
+hostprogs-y		:= piggy addnote addRamDisk
 targets 		+= zImage zImage.initrd imagesize.c \
 			   $(patsubst $(obj)/%,%, $(call obj-sec, $(required) $(initrd))) \
 			   $(patsubst $(obj)/%,%, $(call src-sec, $(required) $(initrd))) \
 			   $(patsubst $(obj)/%,%, $(call gz-sec, $(required) $(initrd))) \
-			   vmlinux.sm vmlinux.initrd vmlinux.sminitrd
-extra-y			:= sysmap.o initrd.o
-
-quiet_cmd_sysmap = SYSMAP   $@
-      cmd_sysmap = $(obj)/addSystemMap System.map $< $@
-$(obj)/vmlinux.sm: vmlinux $(obj)/addSystemMap System.map FORCE
-	$(call if_changed,sysmap)
+			   vmlinux.initrd
+extra-y			:= initrd.o
 
 quiet_cmd_ramdisk = RAMDISK $@
-      cmd_ramdisk = $(obj)/addRamDisk $(obj)/ramdisk.image.gz System.map $< $@
-$(obj)/vmlinux.initrd: vmlinux $(obj)/addRamDisk $(obj)/ramdisk.image.gz System.map FORCE
-	$(call if_changed,ramdisk)
+      cmd_ramdisk = $(obj)/addRamDisk $(obj)/ramdisk.image.gz $< $@
 
-$(obj)/vmlinux.sminitrd: $(obj)/vmlinux.sm $(obj)/addRamDisk $(obj)/ramdisk.image.gz FORCE
-	$(call if_changed,ramdisk)
+quiet_cmd_stripvm = STRIP $@
+      cmd_stripvm = $(BOOTSTRIP) -s $< -o $@
 
-$(obj)/sysmap.o: System.map $(obj)/piggyback FORCE
-	$(call if_changed,piggy)
+vmlinux.strip: vmlinux FORCE
+	$(call if_changed,stripvm)
+$(obj)/vmlinux.initrd: vmlinux.strip $(obj)/addRamDisk $(obj)/ramdisk.image.gz FORCE
+	$(call if_changed,ramdisk)
 
 addsection = $(BOOTOBJCOPY) $(1) \
 		--add-section=.kernel:$(strip $(patsubst $(obj)/kernel-%.o,%, $(1)))=$(patsubst %.o,%.gz, $(1)) \
@@ -113,9 +109,9 @@
 $(obj)/zImage.initrd: $(call obj-sec, $(required) $(initrd)) $(obj-boot) $(obj)/addnote FORCE
 	$(call if_changed,addnote)
 
-$(obj)/imagesize.c: vmlinux
+$(obj)/imagesize.c: vmlinux.strip
 	@echo Generating $@
-	ls -l vmlinux | \
+	ls -l vmlinux.strip | \
 	awk '{printf "/* generated -- do not edit! */\n" \
 		"unsigned long vmlinux_filesize = %d;\n", $$5}' > $(obj)/imagesize.c
 	$(CROSS_COMPILE)nm -n vmlinux | tail -n 1 | \
@@ -123,6 +119,6 @@
 		>> $(obj)/imagesize.c
 
 install: $(CONFIGURE) $(obj)/$(BOOTIMAGE)
-	sh -x $(src)/install.sh "$(KERNELRELEASE)" "$(obj)/$(BOOTIMAGE)" "$(TOPDIR)/System.map" "$(INSTALL_PATH)"
+	sh -x $(src)/install.sh "$(KERNELRELEASE)" "$(obj)/$(BOOTIMAGE)" "$(INSTALL_PATH)"
 
 clean-files := $(patsubst $(obj)/%,%, $(obj-boot))
diff -urN linux-2.5/arch/ppc64/boot/addSystemMap.c linux-prom-clean/arch/ppc64/boot/addSystemMap.c
--- linux-2.5/arch/ppc64/boot/addSystemMap.c	2004-07-21 01:35:49.000000000 +1000
+++ /dev/null	2004-09-16 11:45:22.000000000 +1000
@@ -1,248 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <byteswap.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <string.h>
-
-void xlate( char * inb, char * trb, unsigned len )
-{
-	unsigned i;
-	for (  i=0; i<len; ++i )
-	{
-		char c = *inb++;
-		char c1 = c >> 4;
-		char c2 = c & 0xf;
-		if ( c1 > 9 )
-			c1 = c1 + 'A' - 10;
-		else
-			c1 = c1 + '0';
-		if ( c2 > 9 )
-			c2 = c2 + 'A' - 10;
-		else
-			c2 = c2 + '0';
-		*trb++ = c1;
-		*trb++ = c2;
-	}
-	*trb = 0;
-}
-
-#define ElfHeaderSize  (64 * 1024)
-#define ElfPages  (ElfHeaderSize / 4096)
-
-void get4k( /*istream *inf*/FILE *file, char *buf )
-{
-	unsigned j;
-	unsigned num = fread(buf, 1, 4096, file);
-	for ( j=num; j<4096; ++j )
-		buf[j] = 0;
-}
-
-void put4k( /*ostream *outf*/FILE *file, char *buf )
-{
-	fwrite(buf, 1, 4096, file);
-}
-
-int main(int argc, char **argv)
-{
-	char inbuf[4096];
-	FILE *sysmap = NULL;
-	char* ptr_end = NULL; 
-	FILE *inputVmlinux = NULL;
-	FILE *outputVmlinux = NULL;
-	long i = 0;
-	unsigned long sysmapFileLen = 0;
-	unsigned long sysmapLen = 0;
-	unsigned long roundR = 0;
-	unsigned long kernelLen = 0;
-	unsigned long actualKernelLen = 0;
-	unsigned long round = 0;
-	unsigned long roundedKernelLen = 0;
-	unsigned long sysmapStartOffs = 0;
-	unsigned long sysmapPages = 0;
-	unsigned long roundedKernelPages = 0;
-	long padPages = 0;
-	if ( argc < 2 )
-	{
-		fprintf(stderr, "Name of System Map file missing.\n");
-		exit(1);
-	}
-
-	if ( argc < 3 )
-	{
-		fprintf(stderr, "Name of vmlinux file missing.\n");
-		exit(1);
-	}
-
-	if ( argc < 4 )
-	{
-		fprintf(stderr, "Name of vmlinux output file missing.\n");
-		exit(1);
-	}
-
-	sysmap = fopen(argv[1], "r");
-	if ( ! sysmap )
-	{
-		fprintf(stderr, "System Map file \"%s\" failed to open.\n", argv[1]);
-		exit(1);
-	}
-	inputVmlinux = fopen(argv[2], "r");
-	if ( ! inputVmlinux )
-	{
-		fprintf(stderr, "vmlinux file \"%s\" failed to open.\n", argv[2]);
-		exit(1);
-	}
-	outputVmlinux = fopen(argv[3], "w");
-	if ( ! outputVmlinux )
-	{
-		fprintf(stderr, "output vmlinux file \"%s\" failed to open.\n", argv[3]);
-		exit(1);
-	}
-
-
-  
-	fseek(inputVmlinux, 0, SEEK_END);
-	kernelLen = ftell(inputVmlinux);
-	fseek(inputVmlinux, 0, SEEK_SET);
-	printf("kernel file size = %ld\n", kernelLen);
-	if ( kernelLen == 0 )
-	{
-		fprintf(stderr, "You must have a linux kernel specified as argv[2]\n");
-		exit(1);
-	}
-
-
-	actualKernelLen = kernelLen - ElfHeaderSize;
-
-	printf("actual kernel length (minus ELF header) = %ld/%lxx \n", actualKernelLen, actualKernelLen);
-
-	round = actualKernelLen % 4096;
-	roundedKernelLen = actualKernelLen;
-	if ( round )
-		roundedKernelLen += (4096 - round);
-
-	printf("Kernel length rounded up to a 4k multiple = %ld/%lxx \n", roundedKernelLen, roundedKernelLen);
-	roundedKernelPages = roundedKernelLen / 4096;
-	printf("Kernel pages to copy = %ld/%lxx\n", roundedKernelPages, roundedKernelPages);
-
-
-
-	/* Sysmap file */
-	fseek(sysmap, 0, SEEK_END);
-	sysmapFileLen = ftell(sysmap);
-	fseek(sysmap, 0, SEEK_SET);
-	printf("%s file size = %ld\n", argv[1], sysmapFileLen);
-
-	sysmapLen = sysmapFileLen;
-
-	roundR = 4096 - (sysmapLen % 4096);
-	if (roundR)
-	{
-		printf("Rounding System Map file up to a multiple of 4096, adding %ld\n", roundR);
-		sysmapLen += roundR;
-	}
-	printf("Rounded System Map size is %ld\n", sysmapLen);
-  
-  /* Process the Sysmap file to determine the true end of the kernel */
-	sysmapPages = sysmapLen / 4096;
-	printf("System map pages to copy = %ld\n", sysmapPages);
-	/* read the whole file line by line, expect that it doesn't fail */
-	while ( fgets(inbuf, 4096, sysmap) )  ;
-	/* search for _end in the last page of the system map */
-	ptr_end = strstr(inbuf, " _end");
-	if (!ptr_end)
-	{
-		fprintf(stderr, "Unable to find _end in the sysmap file \n");
-		fprintf(stderr, "inbuf: \n");
-		fprintf(stderr, "%s \n", inbuf);
-		exit(1);
-	}
-	printf("Found _end in the last page of the sysmap - backing up 10 characters it looks like %s", ptr_end-10);
-	sysmapStartOffs = (unsigned int)strtol(ptr_end-10, NULL, 16);
-	/* calc how many pages we need to insert between the vmlinux and the start of the sysmap */
-	padPages = sysmapStartOffs/4096 - roundedKernelPages;
-
-	/* Check and see if the vmlinux is larger than _end in System.map */
-	if (padPages < 0)
-	{ /* vmlinux is larger than _end - adjust the offset to start the embedded system map */ 
-		sysmapStartOffs = roundedKernelLen;
-		printf("vmlinux is larger than _end indicates it needs to be - sysmapStartOffs = %lx \n", sysmapStartOffs);
-		padPages = 0;
-		printf("will insert %lx pages between the vmlinux and the start of the sysmap \n", padPages);
-	}
-	else
-	{ /* _end is larger than vmlinux - use the sysmapStartOffs we calculated from the system map */
-		printf("vmlinux is smaller than _end indicates is needed - sysmapStartOffs = %lx \n", sysmapStartOffs);
-		printf("will insert %lx pages between the vmlinux and the start of the sysmap \n", padPages);
-	}
-
-
-
-
-	/* Copy 64K ELF header */
-	for (i=0; i<(ElfPages); ++i)
-	{
-		get4k( inputVmlinux, inbuf );
-		put4k( outputVmlinux, inbuf );
-	}
-
-  
-	/* Copy the vmlinux (as full pages). */
-	fseek(inputVmlinux, ElfHeaderSize, SEEK_SET);
-	for ( i=0; i<roundedKernelPages; ++i )
-	{
-		get4k( inputVmlinux, inbuf );
-    
-		/* Set the offsets (of the start and end) of the embedded sysmap so it is set in the vmlinux.sm */
-		if ( i == 0 )
-		{
-			unsigned long * p;
-			printf("Storing embedded_sysmap_start at 0x3c\n");
-			p = (unsigned long *)(inbuf + 0x3c);
-
-#if (BYTE_ORDER == __BIG_ENDIAN)
-			*p = sysmapStartOffs;
-#else
-			*p = bswap_32(sysmapStartOffs);
-#endif
-
-			printf("Storing embedded_sysmap_end at 0x44\n");
-			p = (unsigned long *)(inbuf + 0x44);
-
-#if (BYTE_ORDER == __BIG_ENDIAN)
-			*p = sysmapStartOffs + sysmapFileLen;
-#else
-			*p = bswap_32(sysmapStartOffs + sysmapFileLen);
-#endif
-		}
-    
-		put4k( outputVmlinux, inbuf );
-	}
-  
-  
-	/* Insert any pad pages between the end of the vmlinux and where the system map needs to be. */
-	for (i=0; i<padPages; ++i)
-	{
-		memset(inbuf, 0, 4096);
-		put4k(outputVmlinux, inbuf);
-	}
-
-
-	/* Copy the system map (as full pages). */
-	fseek(sysmap, 0, SEEK_SET);  /* start reading from begining of the system map */
-	for ( i=0; i<sysmapPages; ++i )
-	{
-		get4k( sysmap, inbuf );
-		put4k( outputVmlinux, inbuf );
-	}
-
-
-	fclose(sysmap);
-	fclose(inputVmlinux);
-	fclose(outputVmlinux);
-	/* Set permission to executable */
-	chmod(argv[3], S_IRUSR|S_IWUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH);
-
-	return 0;
-}
-
diff -urN linux-2.5/arch/ppc64/boot/main.c linux-prom-clean/arch/ppc64/boot/main.c
--- linux-2.5/arch/ppc64/boot/main.c	2004-07-21 01:35:49.000000000 +1000
+++ linux-prom-clean/arch/ppc64/boot/main.c	2004-09-17 22:06:28.000000000 +1000
@@ -31,13 +31,9 @@
 void *memmove(void *dest, const void *src, unsigned long n);
 void *memcpy(void *dest, const void *src, unsigned long n);
 
-static struct bi_record *make_bi_recs(unsigned long);
-
-#define RAM_START	0x00000000
-#define RAM_END		(64<<20)
-
 /* Value picked to match that used by yaboot */
 #define PROG_START	0x01400000
+#define RAM_END		(256<<20) // Fixme: use OF */
 
 char *avail_ram;
 char *begin_avail, *end_avail;
@@ -48,8 +44,6 @@
 extern char _start[];
 extern char _vmlinux_start[];
 extern char _vmlinux_end[];
-extern char _sysmap_start[];
-extern char _sysmap_end[];
 extern char _initrd_start[];
 extern char _initrd_end[];
 extern unsigned long vmlinux_filesize;
@@ -62,7 +56,6 @@
 };
 struct addr_range vmlinux = {0, 0, 0};
 struct addr_range vmlinuz = {0, 0, 0};
-struct addr_range sysmap  = {0, 0, 0};
 struct addr_range initrd  = {0, 0, 0};
 
 static char scratch[128<<10];	/* 128kB of scratch space for gunzip */
@@ -70,7 +63,7 @@
 typedef void (*kernel_entry_t)( unsigned long,
                                 unsigned long,
                                 void *,
-				struct bi_record *);
+				void *);
 
 
 int (*prom)(void *);
@@ -80,12 +73,31 @@
 void *stdout;
 void *stderr;
 
+#define DEBUG
+
+static unsigned long claim_base = PROG_START;
+
+static unsigned long try_claim(unsigned long size)
+{
+	unsigned long addr = 0;
+
+	for(; claim_base < RAM_END; claim_base += 0x100000) {
+#ifdef DEBUG
+		printf("    trying: 0x%08lx\n\r", claim_base);
+#endif
+		addr = (unsigned long)claim(claim_base, size, 0);
+		if ((void *)addr != (void *)-1)
+			break;
+	}
+	if (addr == 0)
+		return 0;
+	claim_base = PAGE_ALIGN(claim_base + size);
+	return addr;
+}
 
-void
-start(unsigned long a1, unsigned long a2, void *promptr)
+void start(unsigned long a1, unsigned long a2, void *promptr)
 {
-	unsigned long i, claim_addr, claim_size;
-	struct bi_record *bi_recs;
+	unsigned long i;
 	kernel_entry_t kernel_entry;
 	Elf64_Ehdr *elf64;
 	Elf64_Phdr *elf64ph;
@@ -102,61 +114,59 @@
 
 	printf("zImage starting: loaded at 0x%x\n\r", (unsigned)_start);
 
-#if 0
-	sysmap.size = (unsigned long)(_sysmap_end - _sysmap_start);
-	sysmap.memsize = sysmap.size;
-	if ( sysmap.size > 0 ) {
-		sysmap.addr = (RAM_END - sysmap.size) & ~0xFFF;
-		claim(sysmap.addr, RAM_END - sysmap.addr, 0);
-		printf("initial ramdisk moving 0x%lx <- 0x%lx (%lx bytes)\n\r",
-		       sysmap.addr, (unsigned long)_sysmap_start, sysmap.size);
-		memcpy((void *)sysmap.addr, (void *)_sysmap_start, sysmap.size);
+	/*
+	 * Now we try to claim some memory for the kernel itself
+	 * our "vmlinux_memsize" is the memory footprint in RAM, _HOWEVER_, what
+	 * our Makefile stuffs in is an image containing all sort of junk including
+	 * an ELF header. We need to do some calculations here to find the right
+	 * size... In practice we add 1Mb, that is enough, but we should really
+	 * consider fixing the Makefile to put a _raw_ kernel in there !
+	 */
+	vmlinux_memsize += 0x100000;
+	printf("Allocating 0x%lx bytes for kernel ...\n\r", vmlinux_memsize);
+	vmlinux.addr = try_claim(vmlinux_memsize);
+	if (vmlinux.addr == 0) {
+		printf("Can't allocate memory for kernel image !\n\r");
+		exit();
 	}
-#endif
+	vmlinuz.addr = (unsigned long)_vmlinux_start;
+	vmlinuz.size = (unsigned long)(_vmlinux_end - _vmlinux_start);
+	vmlinux.size = PAGE_ALIGN(vmlinux_filesize);
+	vmlinux.memsize = vmlinux_memsize;
 
+	/*
+	 * Now we try to claim memory for the initrd (and copy it there)
+	 */
 	initrd.size = (unsigned long)(_initrd_end - _initrd_start);
 	initrd.memsize = initrd.size;
 	if ( initrd.size > 0 ) {
-		initrd.addr = (RAM_END - initrd.size) & ~0xFFF;
-		a1 = a2 = 0;
-		claim(initrd.addr, RAM_END - initrd.addr, 0);
+		printf("Allocating 0x%lx bytes for initrd ...\n\r", initrd.size);
+		initrd.addr = try_claim(initrd.size);
+		if (initrd.addr == 0) {
+			printf("Can't allocate memory for initial ramdisk !\n\r");
+			exit();
+		}
+		a1 = initrd.addr;
+		a2 = initrd.size;
 		printf("initial ramdisk moving 0x%lx <- 0x%lx (%lx bytes)\n\r",
 		       initrd.addr, (unsigned long)_initrd_start, initrd.size);
-		memcpy((void *)initrd.addr, (void *)_initrd_start, initrd.size);
-	}
-
-	vmlinuz.addr = (unsigned long)_vmlinux_start;
-	vmlinuz.size = (unsigned long)(_vmlinux_end - _vmlinux_start);
-	vmlinux.addr = (unsigned long)(void *)-1;
-	vmlinux.size = PAGE_ALIGN(vmlinux_filesize);
-	vmlinux.memsize = vmlinux_memsize;
-
-	claim_size = vmlinux.memsize /* PPPBBB: + fudge for bi_recs */;
-	for(claim_addr = PROG_START; 
-	    claim_addr <= PROG_START * 8; 
-	    claim_addr += 0x100000) {
-#ifdef DEBUG
-		printf("    trying: 0x%08lx\n\r", claim_addr);
-#endif
-		vmlinux.addr = (unsigned long)claim(claim_addr, claim_size, 0);
-		if ((void *)vmlinux.addr != (void *)-1) break;
-	}
-	if ((void *)vmlinux.addr == (void *)-1) {
-		printf("claim error, can't allocate kernel memory\n\r");
-		exit();
+		memmove((void *)initrd.addr, (void *)_initrd_start, initrd.size);
+		printf("initrd head: 0x%lx\n", *((u32 *)initrd.addr));
 	}
 
-	/* PPPBBB: should kernel always be gziped? */
+	/* Eventually gunzip the kernel */
 	if (*(unsigned short *)vmlinuz.addr == 0x1f8b) {
+		int len;
 		avail_ram = scratch;
 		begin_avail = avail_high = avail_ram;
 		end_avail = scratch + sizeof(scratch);
 		printf("gunzipping (0x%lx <- 0x%lx:0x%0lx)...",
 		       vmlinux.addr, vmlinuz.addr, vmlinuz.addr+vmlinuz.size);
+		len = vmlinuz.size;
 		gunzip((void *)vmlinux.addr, vmlinux.size,
-			(unsigned char *)vmlinuz.addr, (int *)&vmlinuz.size);
-		printf("done %lu bytes\n\r", vmlinuz.size);
-		printf("%u bytes of heap consumed, max in use %u\n\r",
+			(unsigned char *)vmlinuz.addr, &len);
+		printf("done 0x%lx bytes\n\r", len);
+		printf("0x%x bytes of heap consumed, max in use 0x%\n\r",
 		       (unsigned)(avail_high - begin_avail), heap_max);
 	} else {
 		memmove((void *)vmlinux.addr,(void *)vmlinuz.addr,vmlinuz.size);
@@ -192,7 +202,8 @@
 
 	flush_cache((void *)vmlinux.addr, vmlinux.memsize);
 
-	bi_recs = make_bi_recs(vmlinux.addr + vmlinux.memsize);
+	if (a1)
+		printf("initrd head: 0x%lx\n\r", *((u32 *)initrd.addr));
 
 	kernel_entry = (kernel_entry_t)vmlinux.addr;
 #ifdef DEBUG
@@ -203,65 +214,16 @@
 		"        prom       = 0x%lx,\n\r"
 		"        bi_recs    = 0x%lx,\n\r",
 		(unsigned long)kernel_entry, a1, a2,
-		(unsigned long)prom, (unsigned long)bi_recs);
+		(unsigned long)prom, NULL);
 #endif
 
-	kernel_entry( a1, a2, prom, bi_recs );
+	kernel_entry( a1, a2, prom, NULL );
 
 	printf("Error: Linux kernel returned to zImage bootloader!\n\r");
 
 	exit();
 }
 
-static struct bi_record *
-make_bi_recs(unsigned long addr)
-{
-	struct bi_record *bi_recs;
-	struct bi_record *rec;
-
-	bi_recs = rec = bi_rec_init(addr);
-
-	rec = bi_rec_alloc(rec, 2);
-	rec->tag = BI_FIRST;
-	/* rec->data[0] = ...;	# Written below before return */
-	/* rec->data[1] = ...;	# Written below before return */
-
-	rec = bi_rec_alloc_bytes(rec, strlen("chrpboot")+1);
-	rec->tag = BI_BOOTLOADER_ID;
-	sprintf( (char *)rec->data, "chrpboot");
-
-	rec = bi_rec_alloc(rec, 2);
-	rec->tag = BI_MACHTYPE;
-	rec->data[0] = PLATFORM_PSERIES;
-	rec->data[1] = 1;
-
-	if ( initrd.size > 0 ) {
-		rec = bi_rec_alloc(rec, 2);
-		rec->tag = BI_INITRD;
-		rec->data[0] = initrd.addr;
-		rec->data[1] = initrd.size;
-	}
-
-	if ( sysmap.size > 0 ) {
-		rec = bi_rec_alloc(rec, 2);
-		rec->tag = BI_SYSMAP;
-		rec->data[0] = (unsigned long)sysmap.addr;
-		rec->data[1] = (unsigned long)sysmap.size;
-	}
-
-	rec = bi_rec_alloc(rec, 1);
-	rec->tag = BI_LAST;
-	rec->data[0] = (bi_rec_field)bi_recs;
-
-	/* Save the _end_ address of the bi_rec's in the first bi_rec
-	 * data field for easy access by the kernel.
-	 */
-	bi_recs->data[0] = (bi_rec_field)rec;
-	bi_recs->data[1] = (bi_rec_field)rec + rec->size - (bi_rec_field)bi_recs;
-
-	return bi_recs;
-}
-
 struct memchunk {
 	unsigned int size;
 	unsigned int pad;
diff -urN linux-2.5/arch/ppc64/boot/zImage.lds linux-prom-clean/arch/ppc64/boot/zImage.lds
--- linux-2.5/arch/ppc64/boot/zImage.lds	2004-07-21 01:35:49.000000000 +1000
+++ linux-prom-clean/arch/ppc64/boot/zImage.lds	2004-09-17 14:34:55.000000000 +1000
@@ -61,20 +61,10 @@
 
   . = ALIGN(4096);
   _vmlinux_start =  .;
-  .kernel:vmlinux : { *(.kernel:vmlinux) }
+  .kernel:vmlinux.strip : { *(.kernel:vmlinux.strip) }
   _vmlinux_end =  .;
 
   . = ALIGN(4096);
-  _dotconfig_start =  .;
-  .kernel:.config : { *(.kernel:.config) }
-  _dotconfig_end =  .;
-
-  . = ALIGN(4096);
-  _sysmap_start =  .;
-  .kernel:System.map : { *(.kernel:System.map) }
-  _sysmap_end =  .;
-
-  . = ALIGN(4096);
   _initrd_start =  .;
   .kernel:initrd : { *(.kernel:initrd) }
   _initrd_end =  .;
diff -urN linux-2.5/arch/ppc64/kernel/Makefile linux-prom-clean/arch/ppc64/kernel/Makefile
--- linux-2.5/arch/ppc64/kernel/Makefile	2004-08-26 15:46:30.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/Makefile	2004-09-13 15:15:14.000000000 +1000
@@ -31,7 +31,8 @@
 obj-$(CONFIG_PPC_PSERIES) += pSeries_pci.o pSeries_lpar.o pSeries_hvCall.o \
 			     eeh.o nvram.o pSeries_nvram.o rtasd.o ras.o \
 			     open_pic.o xics.o pSeries_htab.o rtas.o \
-			     chrp_setup.o i8259.o prom.o pSeries_iommu.o
+			     chrp_setup.o i8259.o prom_init.o prom.o \
+			     pSeries_iommu.o
 
 obj-$(CONFIG_PROC_FS)		+= proc_ppc64.o
 obj-$(CONFIG_RTAS_FLASH)	+= rtas_flash.o
diff -urN linux-2.5/arch/ppc64/kernel/asm-offsets.c linux-prom-clean/arch/ppc64/kernel/asm-offsets.c
--- linux-2.5/arch/ppc64/kernel/asm-offsets.c	2004-09-09 17:33:52.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/asm-offsets.c	2004-09-15 21:53:25.000000000 +1000
@@ -33,7 +33,6 @@
 #include <asm/iSeries/ItLpPaca.h>
 #include <asm/iSeries/ItLpQueue.h>
 #include <asm/iSeries/HvLpEvent.h>
-#include <asm/prom.h>
 #include <asm/rtas.h>
 #include <asm/cputable.h>
 
@@ -108,7 +107,6 @@
         DEFINE(LPPACASRR1, offsetof(struct ItLpPaca, xSavedSrr1));
 	DEFINE(LPPACAANYINT, offsetof(struct ItLpPaca, xIntDword.xAnyInt));
 	DEFINE(LPPACADECRINT, offsetof(struct ItLpPaca, xIntDword.xFields.xDecrInt));
-	DEFINE(PROMENTRY, offsetof(struct prom_t, entry));
 
 	/* RTAS */
 	DEFINE(RTASBASE, offsetof(struct rtas_t, base));
diff -urN linux-2.5/arch/ppc64/kernel/btext.c linux-prom-clean/arch/ppc64/kernel/btext.c
--- linux-2.5/arch/ppc64/kernel/btext.c	2004-07-08 06:27:04.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/btext.c	2004-09-13 15:15:14.000000000 +1000
@@ -45,29 +45,31 @@
 int boot_text_mapped;
 int force_printk_to_btext = 0;
 
+extern int global_width, global_height, global_depth, global_pitch;
+extern unsigned global_address;
 boot_infos_t disp_bi;
 
 /* This function will enable the early boot text when doing OF booting. This
  * way, xmon output should work too
  */
-void __init btext_setup_display(int width, int height, int depth, int pitch,
+void btext_setup_display(int width, int height, int depth, int pitch,
 		    unsigned long address)
 {
-	unsigned long offset = reloc_offset();
-	boot_infos_t* bi = PTRRELOC(&disp_bi);
+	//unsigned long offset = reloc_offset();
+	boot_infos_t* bi = &disp_bi;
 
-	RELOC(g_loc_X) = 0;
-	RELOC(g_loc_Y) = 0;
-	RELOC(g_max_loc_X) = width / 8;
-	RELOC(g_max_loc_Y) = height / 16;
-	bi->logicalDisplayBase = (unsigned char *)address;
-	bi->dispDeviceBase = (unsigned char *)address;
-	bi->dispDeviceRowBytes = pitch;
-	bi->dispDeviceDepth = depth;
+	g_loc_X = 0;
+	g_loc_Y = 0;
+	g_max_loc_X = global_width / 8;
+	g_max_loc_Y = global_height / 16;
+	bi->logicalDisplayBase = (unsigned char *)global_address;
+	bi->dispDeviceBase = (unsigned char *)global_address;
+	bi->dispDeviceRowBytes = global_pitch;
+	bi->dispDeviceDepth = global_depth;
 	bi->dispDeviceRect[0] = bi->dispDeviceRect[1] = 0;
-	bi->dispDeviceRect[2] = width;
-	bi->dispDeviceRect[3] = height;
-	RELOC(boot_text_mapped) = 1;
+	bi->dispDeviceRect[2] = global_width;
+	bi->dispDeviceRect[3] = global_height;
+	boot_text_mapped = 1;
 }
 
 /* Here's a small text engine to use during early boot
diff -urN linux-2.5/arch/ppc64/kernel/chrp_setup.c linux-prom-clean/arch/ppc64/kernel/chrp_setup.c
--- linux-2.5/arch/ppc64/kernel/chrp_setup.c	2004-09-15 16:25:15.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/chrp_setup.c	2004-09-15 21:53:25.000000000 +1000
@@ -205,7 +205,7 @@
 
 	hpte_init_pSeries();
 
-	if (ppc64_iommu_off)
+	if (of_chosen && get_property(of_chosen, "linux,iommu-off", NULL))
 		pci_dma_init_direct();
 	else
 		tce_init_pSeries();
@@ -223,9 +223,7 @@
 	ppc_md.udbg_getc_poll = udbg_getc_poll;
 }
 
-void __init
-chrp_init(unsigned long r3, unsigned long r4, unsigned long r5,
-	  unsigned long r6, unsigned long r7)
+void __init chrp_init(void)
 {
 	ppc_md.setup_arch     = chrp_setup_arch;
 	ppc_md.get_cpuinfo    = chrp_get_cpuinfo;
diff -urN linux-2.5/arch/ppc64/kernel/entry.S linux-prom-clean/arch/ppc64/kernel/entry.S
--- linux-2.5/arch/ppc64/kernel/entry.S	2004-09-06 16:43:57.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/entry.S	2004-09-15 17:05:42.000000000 +1000
@@ -777,11 +777,8 @@
 	std	r11,_MSR(r1)
 
 	/* Get the PROM entrypoint */
-        bl      .reloc_offset
-	LOADADDR(r12,prom)
-	sub	r12,r12,r3
-	ld	r12,PROMENTRY(r12)
-	mtlr	r12
+	ld	r0,GPR4(r1)
+	mtlr	r0
 
 	/* Switch MSR to 32 bits mode
 	 */
diff -urN linux-2.5/arch/ppc64/kernel/head.S linux-prom-clean/arch/ppc64/kernel/head.S
--- linux-2.5/arch/ppc64/kernel/head.S	2004-09-15 16:25:15.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/head.S	2004-09-16 13:49:50.000000000 +1000
@@ -84,8 +84,9 @@
 	.globl  _stext
 _stext:
 #ifdef CONFIG_PPC_PSERIES
-_STATIC(__start)
+_GLOBAL(__start)
 	/* NOP this out unconditionally */
+	li	r28,0x5a00
 BEGIN_FTR_SECTION
 	b .__start_initialization_pSeries
 END_FTR_SECTION(0, 1)
@@ -1223,7 +1224,7 @@
 #endif
 	b 	1b			/* Loop until told to go	 */
 #ifdef CONFIG_PPC_ISERIES
-_GLOBAL(__start_initialization_iSeries)
+_STATIC(__start_initialization_iSeries)
 	/* Clear out the BSS */
 	LOADADDR(r11,__bss_stop)
 
@@ -1274,7 +1275,7 @@
 
 #ifdef CONFIG_PPC_PSERIES
 
-_STATIC(mmu_off)
+_STATIC(__mmu_off)
 	mfmsr	r3
 	andi.	r0,r3,MSR_IR|MSR_DR
 	beqlr
@@ -1284,13 +1285,83 @@
 	sync
 	rfid
 	b	.	/* prevent speculative execution */
+
+/*
+ * Here is our main kernel entry point. We support currently 2 kind of entries
+ * depending on the value of r5.
+ *
+ *   r5 != NULL -> OF entry, we go to prom_init, "legacy" parameter content
+ *                 in r3...r7
+ *   
+ *   r5 == NULL -> kexec style entry. r3 is a physical pointer to the
+ *                 DT block, r4 is a physical pointer to the kernel itself
+ *		   also TODO: pass klimit in there (r6)
+ */
 _GLOBAL(__start_initialization_pSeries)
-	mr	r31,r3			/* save parameters */
+	/*
+	 * Are we booted from a PROM Of-type client-interface ?
+	 */
+	cmpldi	cr0,r5,0
+	bne	.__boot_from_prom		/* yes -> prom */
+
+	li	r28,0x5a01
+	
+	/*
+	 * We are booted from a kexec-style call
+	 */
+
+	/* Save parameters */
+	mr	r31,r3
+	mr	r30,r4
+	mr	r29,r6
+
+	/*
+	 * Make sure we are running in 64 bits mode
+	 */
+	bl	.enable_64b_mode
+
+	li	r28,0x5a02
+
+	/* put a relocation offset into r3 */
+	bl	.reloc_offset
+
+	LOADADDR(r2,__toc_start)
+	addi	r2,r2,0x4000
+	addi	r2,r2,0x4000
+
+	/* Relocate the TOC from a virt addr to a real addr */
+	sub	r2,r2,r3
+
+	/* Setup some critical 970 SPRs before switching MMU off */
+	bl	.__970_cpu_preinit
+
+	li	r28,0x5a03
+
+	li	r24,0			/* cpu # */
+
+	/* Switch off MMU if not already */
+	LOADADDR(r4, .__after_prom_start - KERNELBASE)
+	add	r4,r4,r30
+	bl	.__mmu_off
+	b	.__after_prom_start
+
+_STATIC(__boot_from_prom)
+	/*
+	 * We are booted from a PROM
+	 */
+
+	/* Save parameters */
+	mr	r31,r3
 	mr	r30,r4
 	mr	r29,r5
 	mr	r28,r6
 	mr	r27,r7
 
+	/*
+	 * Make sure we are running in 64 bits mode
+	 * XXX kill that when no longer necessary, that is when prom_init finally
+	 * becomes a 32 bits separate binary :)
+	 */
 	bl	.enable_64b_mode
 
 	/* put a relocation offset into r3 */
@@ -1303,7 +1374,7 @@
 	/* Relocate the TOC from a virt addr to a real addr */
 	sub	r2,r2,r3
 
-	/* Save parameters */
+	/* Restore parameters */
 	mr	r3,r31
 	mr	r4,r30
 	mr	r5,r29
@@ -1312,17 +1383,8 @@
 
 	/* Do all of the interaction with OF client interface */
 	bl	.prom_init
-	mr	r23,r3			/* Save phys address we are running at */
-
-	/* Setup some critical 970 SPRs before switching MMU off */
-	bl	.__970_cpu_preinit
-
-	li	r24,0			/* cpu # */
-
-	/* Switch off MMU if not already */
-	LOADADDR(r4, .__after_prom_start - KERNELBASE)
-	add	r4,r4,r23
-	bl	.mmu_off	
+	/* We never return */
+	b	.
 
 /*
  * At this point, r3 contains the physical address we are running at,
@@ -1348,7 +1410,7 @@
 
 	li	r3,0			/* target addr */
 
-	// XXX FIXME: Use phys returned by OF (r23)
+	// XXX FIXME: Use phys returned by OF (r30)
 	sub	r4,r27,r26 		/* source addr			 */
 					/* current address of _start	 */
 					/*   i.e. where we are running	 */
@@ -1368,7 +1430,7 @@
 	mtctr	r0			/* that we just made/relocated	 */
 	bctr
 
-4:	LOADADDR(r5,klimit)
+4:	LOADADDR(r5,klimit) // XXX FIXME: change klimit based on parameter r6 */
 	sub	r5,r5,r26
 	ld	r5,0(r5)		/* get the value of klimit */
 	sub	r5,r5,r27
@@ -1875,8 +1937,7 @@
 	mr	r5,r26
 	bl	.identify_cpu
 
-	/* Get the pointer to the segment table which is used by		*/
-	/* stab_initialize						 */
+	/* Setup a valid physical PACA pointer in SPRG3 for early_init */
 	LOADADDR(r27, boot_cpuid)
 	sub	r27,r27,r26
 	lwz	r27,0(r27)
@@ -1909,12 +1970,18 @@
 98:					/* !(rpa hypervisor) || !(star) */
 	mtasr	r4			/* set the stab location	*/
 99:
-	mfspr	r6,SPRG3
-	ld	r3,PACASTABREAL(r6)	/* restore r3 for stab_initialize */
 
-	/* Initialize an initial memory mapping and turn on relocation.	*/
-	bl	.stab_initialize
-	bl	.htab_initialize
+#ifdef CONFIG_BOOTX_TEXT
+/*	bl	.btext_setup_display */
+#endif
+	/* Do very early kernel initializations, including initial hash table setup
+	 * and turn on relocation.	*/
+
+	/* Restore parameters passed from prom_init/kexec */
+	mr	r3,r31
+	mr	r4,r30
+	mr	r5,r29
+ 	bl	.early_init
 
 	li	r3,SYSTEMCFG_PHYS_ADDR	/* r3 = ptr to systemcfg */
 	lwz	r3,PLATFORM(r3)		/* r3 = platform flags */
@@ -1980,13 +2047,7 @@
 	ld	r2,PACATOC(r13)
 	std	r1,PACAKSAVE(r13)
 
-	/* Restore the parms passed in from the bootloader. */
-	mr	r3,r31
-	mr	r4,r30
-	mr	r5,r29
-	mr	r6,r28
-	mr	r7,r27
-
+	/* Call some more initializations with a complete mmu environment */
 	bl	.setup_system
 
 	/* Load up the kernel context */
diff -urN linux-2.5/arch/ppc64/kernel/lmb.c linux-prom-clean/arch/ppc64/kernel/lmb.c
--- linux-2.5/arch/ppc64/kernel/lmb.c	2004-09-06 16:43:57.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/lmb.c	2004-09-17 20:58:41.000000000 +1000
@@ -22,6 +22,43 @@
 
 struct lmb lmb;
 
+#undef DEBUG
+
+void lmb_dump_all(void)
+{
+#ifdef DEBUG
+	unsigned long i;
+	struct lmb *_lmb  = &lmb;
+
+	udbg_printf("lmb_dump_all:\n");
+	udbg_printf("    memory.cnt		  = 0x%lx\n",
+		    _lmb->memory.cnt);
+	udbg_printf("    memory.size		  = 0x%lx\n",
+		    _lmb->memory.size);
+	for (i=0; i < _lmb->memory.cnt ;i++) {
+		udbg_printf("    memory.region[0x%x].base       = 0x%lx\n",
+			    i, _lmb->memory.region[i].base);
+		udbg_printf("		      .physbase = 0x%lx\n",
+			    _lmb->memory.region[i].physbase);
+		udbg_printf("		      .size     = 0x%lx\n",
+			    _lmb->memory.region[i].size);
+	}
+
+	udbg_printf("\n    reserved.cnt	  = 0x%lx\n",
+		    _lmb->reserved.cnt);
+	udbg_printf("    reserved.size	  = 0x%lx\n",
+		    _lmb->reserved.size);
+	for (i=0; i < _lmb->reserved.cnt ;i++) {
+		udbg_printf("    reserved.region[0x%x].base       = 0x%lx\n",
+			    i, _lmb->reserved.region[i].base);
+		udbg_printf("		      .physbase = 0x%lx\n",
+			    _lmb->reserved.region[i].physbase);
+		udbg_printf("		      .size     = 0x%lx\n",
+			    _lmb->reserved.region[i].size);
+	}
+#endif /* DEBUG */
+}
+
 static unsigned long __init
 lmb_addrs_overlap(unsigned long base1, unsigned long size1,
                   unsigned long base2, unsigned long size2)
@@ -71,8 +108,7 @@
 void __init
 lmb_init(void)
 {
-	unsigned long offset = reloc_offset();
-	struct lmb *_lmb = PTRRELOC(&lmb);
+	struct lmb *_lmb = &lmb;
 
 	/* Create a dummy zero size LMB which will get coalesced away later.
 	 * This simplifies the lmb_add() code below...
@@ -94,8 +130,7 @@
 	unsigned long i;
 	unsigned long mem_size = 0;
 	unsigned long size_mask = 0;
-	unsigned long offset = reloc_offset();
-	struct lmb *_lmb = PTRRELOC(&lmb);
+	struct lmb *_lmb = &lmb;
 #ifdef CONFIG_MSCHUNKS
 	unsigned long physbase = 0;
 #endif
@@ -178,8 +213,7 @@
 long __init
 lmb_add(unsigned long base, unsigned long size)
 {
-	unsigned long offset = reloc_offset();
-	struct lmb *_lmb = PTRRELOC(&lmb);
+	struct lmb *_lmb = &lmb;
 	struct lmb_region *_rgn = &(_lmb->memory);
 
 	/* On pSeries LPAR systems, the first LMB is our RMO region. */
@@ -193,8 +227,7 @@
 long __init
 lmb_reserve(unsigned long base, unsigned long size)
 {
-	unsigned long offset = reloc_offset();
-	struct lmb *_lmb = PTRRELOC(&lmb);
+	struct lmb *_lmb = &lmb;
 	struct lmb_region *_rgn = &(_lmb->reserved);
 
 	return lmb_add_region(_rgn, base, size);
@@ -227,8 +260,7 @@
 {
 	long i, j;
 	unsigned long base = 0;
-	unsigned long offset = reloc_offset();
-	struct lmb *_lmb = PTRRELOC(&lmb);
+	struct lmb *_lmb = &lmb;
 	struct lmb_region *_mem = &(_lmb->memory);
 	struct lmb_region *_rsv = &(_lmb->reserved);
 
@@ -263,8 +295,7 @@
 unsigned long __init
 lmb_phys_mem_size(void)
 {
-	unsigned long offset = reloc_offset();
-	struct lmb *_lmb = PTRRELOC(&lmb);
+	struct lmb *_lmb = &lmb;
 #ifdef CONFIG_MSCHUNKS
 	return _lmb->memory.size;
 #else
@@ -282,8 +313,7 @@
 unsigned long __init
 lmb_end_of_DRAM(void)
 {
-	unsigned long offset = reloc_offset();
-	struct lmb *_lmb = PTRRELOC(&lmb);
+	struct lmb *_lmb = &lmb;
 	struct lmb_region *_mem = &(_lmb->memory);
 	int idx = _mem->cnt - 1;
 
@@ -300,8 +330,7 @@
 lmb_abs_to_phys(unsigned long aa)
 {
 	unsigned long i, pa = aa;
-	unsigned long offset = reloc_offset();
-	struct lmb *_lmb = PTRRELOC(&lmb);
+	struct lmb *_lmb = &lmb;
 	struct lmb_region *_mem = &(_lmb->memory);
 
 	for (i=0; i < _mem->cnt; i++) {
diff -urN linux-2.5/arch/ppc64/kernel/misc.S linux-prom-clean/arch/ppc64/kernel/misc.S
--- linux-2.5/arch/ppc64/kernel/misc.S	2004-09-15 16:25:15.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/misc.S	2004-09-15 21:53:26.000000000 +1000
@@ -454,17 +454,6 @@
 	sync
 	blr	
 
-_GLOBAL(_get_PVR)
-	mfspr	r3,PVR
-	blr
-
-_GLOBAL(_get_PIR)
-	mfspr	r3,PIR
-	blr
-
-_GLOBAL(_get_HID0)
-	mfspr	r3,HID0
-	blr
 
 _GLOBAL(cvt_fd)
 	lfd	0,0(r5)		/* load up fpscr value */
diff -urN linux-2.5/arch/ppc64/kernel/open_pic.c linux-prom-clean/arch/ppc64/kernel/open_pic.c
--- linux-2.5/arch/ppc64/kernel/open_pic.c	2004-07-29 08:46:55.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/open_pic.c	2004-09-17 17:28:11.000000000 +1000
@@ -362,7 +362,6 @@
 	find_ISUs();
 
 	/* Initialize timer interrupts */
-	ppc64_boot_msg(0x21, "OpenPic Timer");
 	for (i = 0; i < OPENPIC_NUM_TIMERS; i++) {
 		/* Disabled, Priority 0 */
 		openpic_inittimer(i, 0, openpic_vec_timer+i);
@@ -372,7 +371,6 @@
 
 #ifdef CONFIG_SMP
 	/* Initialize IPI interrupts */
-	ppc64_boot_msg(0x22, "OpenPic IPI");
 	openpic_test_broken_IPI();
 	for (i = 0; i < OPENPIC_NUM_IPI; i++) {
 		/* Disabled, Priority 10..13 */
@@ -384,8 +382,6 @@
 #endif
 
 	/* Initialize external interrupts */
-	ppc64_boot_msg(0x23, "OpenPic Ext");
-
 	openpic_set_priority(0xf);
 
 	/* SIOint (8259 cascade) is special */
@@ -420,7 +416,6 @@
 		irq_desc[i].handler = &open_pic;
 
 	/* Initialize the spurious interrupt */
-	ppc64_boot_msg(0x24, "OpenPic Spurious");
 	openpic_set_spurious(openpic_vec_spurious);
 
 	openpic_set_priority(0);
diff -urN linux-2.5/arch/ppc64/kernel/open_pic_u3.c linux-prom-clean/arch/ppc64/kernel/open_pic_u3.c
--- linux-2.5/arch/ppc64/kernel/open_pic_u3.c	2004-07-08 06:27:06.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/open_pic_u3.c	2004-09-17 17:28:22.000000000 +1000
@@ -344,5 +344,5 @@
 	openpic2_set_priority(0);
 	openpic2_disable_8259_pass_through();
 
-	ppc64_boot_msg(0x25, "OpenPic2 Done");
+	ppc64_boot_msg(0x25, "OpenPic U3 Done");
 }
diff -urN linux-2.5/arch/ppc64/kernel/pSeries_htab.c linux-prom-clean/arch/ppc64/kernel/pSeries_htab.c
--- linux-2.5/arch/ppc64/kernel/pSeries_htab.c	2004-09-17 22:20:11.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/pSeries_htab.c	2004-09-15 17:30:31.000000000 +1000
@@ -220,10 +220,12 @@
 	if ((cur_cpu_spec->cpu_features & CPU_FTR_TLBIEL) && !large && local) {
 		tlbiel(va);
 	} else {
-		if (!(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE))
+		int lock_tlbie = !(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE);
+
+		if (lock_tlbie)
 			spin_lock(&pSeries_tlbie_lock);
 		tlbie(va, large);
-		if (!(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE))
+		if (lock_tlbie)
 			spin_unlock(&pSeries_tlbie_lock);
 	}
 
@@ -243,6 +245,7 @@
 	unsigned long vsid, va, vpn, flags;
 	long slot;
 	HPTE *hptep;
+	int lock_tlbie = !(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE);
 
 	vsid = get_kernel_vsid(ea);
 	va = (vsid << 28) | (ea & 0x0fffffff);
@@ -256,10 +259,10 @@
 	set_pp_bit(newpp, hptep);
 
 	/* Ensure it is out of the tlb too */
-	if (!(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE))
+	if (lock_tlbie)
 		spin_lock_irqsave(&pSeries_tlbie_lock, flags);
 	tlbie(va, 0);
-	if (!(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE))
+	if (lock_tlbie)
 		spin_unlock_irqrestore(&pSeries_tlbie_lock, flags);
 }
 
@@ -270,6 +273,7 @@
 	Hpte_dword0 dw0;
 	unsigned long avpn = va >> 23;
 	unsigned long flags;
+	int lock_tlbie = !(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE);
 
 	if (large)
 		avpn &= ~0x1UL;
@@ -291,10 +295,10 @@
 	if ((cur_cpu_spec->cpu_features & CPU_FTR_TLBIEL) && !large && local) {
 		tlbiel(va);
 	} else {
-		if (!(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE))
+		if (lock_tlbie)
 			spin_lock(&pSeries_tlbie_lock);
 		tlbie(va, large);
-		if (!(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE))
+		if (lock_tlbie)
 			spin_unlock(&pSeries_tlbie_lock);
 	}
 	local_irq_restore(flags);
@@ -364,8 +368,9 @@
 
 		asm volatile("ptesync":::"memory");
 	} else {
-		/* XXX double check that it is safe to take this late */
-		if (!(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE))
+		int lock_tlbie = !(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE);
+
+		if (lock_tlbie)
 			spin_lock(&pSeries_tlbie_lock);
 
 		asm volatile("ptesync":::"memory");
@@ -375,7 +380,7 @@
 
 		asm volatile("eieio; tlbsync; ptesync":::"memory");
 
-		if (!(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE))
+		if (lock_tlbie)
 			spin_unlock(&pSeries_tlbie_lock);
 	}
 
diff -urN linux-2.5/arch/ppc64/kernel/pSeries_iommu.c linux-prom-clean/arch/ppc64/kernel/pSeries_iommu.c
--- linux-2.5/arch/ppc64/kernel/pSeries_iommu.c	2004-08-26 15:46:30.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/pSeries_iommu.c	2004-09-16 13:46:10.000000000 +1000
@@ -166,24 +166,25 @@
 				 struct device_node *dn,
 				 struct iommu_table *tbl) 
 {
-	phandle node;
-	unsigned long i;
-	struct of_tce_table *oft;
-
-	node = ((struct device_node *)(phb->arch_data))->node;
-
-	oft = NULL;
-
-	for (i=0; of_tce_table[i].node; i++)
-		if(of_tce_table[i].node == node) {
-			oft = &of_tce_table[i];
-			break;
-		}
-
-	if (!oft)
-		panic("PCI_DMA: iommu_table_setparms: Can't find phb named '%s' in of_tce_table\n", dn->full_name);
-
-	memset((void *)oft->base, 0, oft->size);
+	struct device_node *node;
+	unsigned long *basep;
+	unsigned int *sizep;
+
+	node = (struct device_node *)phb->arch_data;
+
+	if (get_property(node, "linux,has-tce-table", NULL) == NULL) {
+		printk(KERN_ERR "PCI_DMA: iommu_table_setparms: %s has no tce table !\n",
+		      dn->full_name);
+		return;
+	}
+	basep = (unsigned long *)get_property(node, "linux,tce-base", NULL);
+	sizep = (unsigned int *)get_property(node, "linux,tce-size", NULL);
+	if (basep == NULL || sizep == NULL) {
+		printk(KERN_ERR "PCI_DMA: iommu_table_setparms: %s has missing tce"
+		       " entries !\n", dn->full_name);
+		return;
+	}
+	memset((void *)(*basep), 0, *sizep);
 
 	tbl->it_busno = phb->bus->number;
 	
@@ -207,7 +208,7 @@
 	if (phb->dma_window_base_cur > (1 << 19))
 		panic("PCI_DMA: Unexpected number of IOAs under this PHB.\n"); 
 	
-	tbl->it_base = oft->base;
+	tbl->it_base = *basep;
 	tbl->it_index = 0;
 	tbl->it_entrysize = sizeof(union tce_entry);
 	tbl->it_blocksize = 16;
diff -urN linux-2.5/arch/ppc64/kernel/pSeries_lpar.c linux-prom-clean/arch/ppc64/kernel/pSeries_lpar.c
--- linux-2.5/arch/ppc64/kernel/pSeries_lpar.c	2004-09-17 22:20:11.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/pSeries_lpar.c	2004-09-17 22:18:42.000000000 +1000
@@ -352,24 +352,37 @@
 	}
 }
 
+/* call this from early_init() for a working debug console on
+ * vterm capable LPAR machines
+ */
+#if 0
+void udbg_init_debug_lpar(void)
+{
+	vtermno = 0;
+	ppc_md.udbg_putc = udbg_putcLP;
+	ppc_md.udbg_getc = udbg_getcLP;
+	ppc_md.udbg_getc_poll = udbg_getc_pollLP;
+}
+#endif
+
 /* returns 0 if couldn't find or use /chosen/stdout as console */
 static int find_udbg_vterm(void)
 {
 	struct device_node *stdout_node;
+	phandle *stdout_ph;
 	u32 *termno;
 	char *name;
 	int found = 0;
 
 	/* find the boot console from /chosen/stdout */
-	if (!of_stdout_device) {
-		printk(KERN_WARNING "couldn't get path from /chosen/stdout!\n");
-		return found;
-	}
-	stdout_node = of_find_node_by_path(of_stdout_device);
-	if (!stdout_node) {
-		printk(KERN_WARNING "couldn't find node from /chosen/stdout\n");
-		return found;
-	}
+	if (!of_chosen)
+		return 0;
+	stdout_ph = (phandle *)get_property(of_chosen, "linux,stdout-package", NULL);
+	if (stdout_ph == NULL)
+		return 0;
+	stdout_node = of_find_node_by_phandle(*stdout_ph);
+	if (!stdout_node)
+		return 0;
 
 	/* now we have the stdout node; figure out what type of device it is. */
 	name = (char *)get_property(stdout_node, "name", NULL);
@@ -391,15 +404,17 @@
 		} else {
 			/* XXX implement udbg_putcLP_vtty for hvterm-protocol1 case */
 			printk(KERN_WARNING "%s doesn't speak hvterm1; "
-					"can't print udbg messages\n", of_stdout_device);
+					"can't print udbg messages\n",
+			       stdout_node->full_name);
 		}
 	} else if (strncmp(name, "serial", 6)) {
 		/* XXX fix ISA serial console */
 		printk(KERN_WARNING "serial stdout on LPAR ('%s')! "
-				"can't print udbg messages\n", of_stdout_device);
+				"can't print udbg messages\n",
+		       stdout_node->full_name);
 	} else {
 		printk(KERN_WARNING "don't know how to print to stdout '%s'\n",
-				of_stdout_device);
+		       stdout_node->full_name);
 	}
 
 out:
@@ -532,6 +547,19 @@
 	return -1;
 }
 
+void pSeries_lpar_hptab_clear(void)
+{
+	unsigned long size_bytes = 1UL << naca->pftSize;
+	unsigned long hpte_count = size_bytes >> 4;
+	unsigned long lpar_rc;
+	unsigned long dummy1, dummy2;
+	int i;
+
+	/* TODO: Use bulk call */
+	for (i = 0; i < hpte_count; i++)
+		plpar_pte_remove(0, i, 0, &dummy1, &dummy2);
+}
+
 /*
  * NOTE: for updatepp ops we are fortunate that the linux "newpp" bits and
  * the low 3 bits of flags happen to line up.  So no transform is needed.
@@ -660,14 +688,15 @@
 	int i;
 	unsigned long flags;
 	struct ppc64_tlb_batch *batch = &__get_cpu_var(ppc64_tlb_batch);
+	int lock_tlbie = !(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE);
 
-	if (!(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE))
+	if (lock_tlbie)
 		spin_lock_irqsave(&pSeries_lpar_tlbie_lock, flags);
 
 	for (i = 0; i < number; i++)
 		flush_hash_page(context, batch->addr[i], batch->pte[i], local);
 
-	if (!(cur_cpu_spec->cpu_features & CPU_FTR_LOCKLESS_TLBIE))
+	if (lock_tlbie)
 		spin_unlock_irqrestore(&pSeries_lpar_tlbie_lock, flags);
 }
 
diff -urN linux-2.5/arch/ppc64/kernel/pSeries_pci.c linux-prom-clean/arch/ppc64/kernel/pSeries_pci.c
--- linux-2.5/arch/ppc64/kernel/pSeries_pci.c	2004-09-15 16:25:15.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/pSeries_pci.c	2004-09-15 21:53:26.000000000 +1000
@@ -765,7 +765,8 @@
 	phbs_fixup_io();
 	chrp_request_regions();
 	pci_fix_bus_sysdata();
-	if (!ppc64_iommu_off)
+
+	if (!of_chosen || !get_property(of_chosen, "linux,iommu-off", NULL))
 		iommu_setup_pSeries();
 }
 
diff -urN linux-2.5/arch/ppc64/kernel/pmac_feature.c linux-prom-clean/arch/ppc64/kernel/pmac_feature.c
--- linux-2.5/arch/ppc64/kernel/pmac_feature.c	2004-09-06 16:43:57.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/pmac_feature.c	2004-09-17 14:34:55.000000000 +1000
@@ -44,9 +44,9 @@
 #undef DEBUG_FEATURE
 
 #ifdef DEBUG_FEATURE
-#define DBG(fmt,...) printk(KERN_DEBUG fmt)
+#define DBG(fmt...) printk(KERN_DEBUG fmt)
 #else
-#define DBG(fmt,...)
+#define DBG(fmt...)
 #endif
 
 /*
diff -urN linux-2.5/arch/ppc64/kernel/pmac_iommu.c linux-prom-clean/arch/ppc64/kernel/pmac_iommu.c
--- linux-2.5/arch/ppc64/kernel/pmac_iommu.c	2004-07-08 06:27:06.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/pmac_iommu.c	2004-09-17 14:34:55.000000000 +1000
@@ -37,13 +37,14 @@
 #include <linux/vmalloc.h>
 #include <asm/io.h>
 #include <asm/prom.h>
-#include <asm/rtas.h>
 #include <asm/ppcdebug.h>
 #include <asm/iommu.h>
 #include <asm/pci-bridge.h>
 #include <asm/machdep.h>
 #include <asm/abs_addr.h>
 #include <asm/cacheflush.h>
+#include <asm/lmb.h>
+
 #include "pci.h"
 
 
@@ -76,8 +77,8 @@
 #define DARTMAP_RPNMASK 0x00ffffff
 
 /* Physical base address and size of the DART table */
-unsigned long dart_tablebase;
-unsigned long dart_tablesize;
+unsigned long dart_tablebase; /* exported to htab_initialize */
+static unsigned long dart_tablesize;
 
 /* Virtual base address of the DART table */
 static u32 *dart_vbase;
@@ -300,6 +301,22 @@
 	}
 }
 
+void __init pmac_iommu_alloc(void)
+{
+	/* Only reserve DART space if machine has more than 2GB of RAM
+	 * or if requested with iommu=on on cmdline.
+	 */
+	if (lmb_end_of_DRAM() <= 0x80000000ull &&
+	    get_property(of_chosen, "linux,iommu-force-on", NULL) == NULL)
+		return;
 
+	/* 512 pages (2MB) is max DART tablesize. */
+	dart_tablesize = 1UL << 21;
+	/* 16MB (1 << 24) alignment. We allocate a full 16Mb chuck since we
+	 * will blow up an entire large page anyway in the kernel mapping
+	 */
+	dart_tablebase = (unsigned long)
+		abs_to_virt(lmb_alloc_base(1UL<<24, 1UL<<24, 0x80000000L));
 
-
+	printk(KERN_INFO "U3-DART allocated at: %lx\n", dart_tablebase);
+}
diff -urN linux-2.5/arch/ppc64/kernel/pmac_setup.c linux-prom-clean/arch/ppc64/kernel/pmac_setup.c
--- linux-2.5/arch/ppc64/kernel/pmac_setup.c	2004-07-08 06:27:06.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/pmac_setup.c	2004-09-17 14:34:55.000000000 +1000
@@ -77,14 +77,10 @@
 #define DEFAULT_ROOT_DEVICE Root_SDA1	/* sda1 - slightly silly choice */
 
 extern  int powersave_nap;
-int sccdbg;
+int sccdbg = 1;
 
 extern void udbg_init_scc(struct device_node *np);
 
-#ifdef CONFIG_BOOTX_TEXT
-void pmac_progress(char *s, unsigned short hex);
-#endif
-
 void __pmac pmac_show_cpuinfo(struct seq_file *m)
 {
 	struct device_node *np;
@@ -183,67 +179,13 @@
 #endif
 }
 
-extern char *bootpath;
-extern char *bootdevice;
-void *boot_host;
-int boot_target;
-int boot_part;
-extern dev_t boot_dev;
-
 #ifdef CONFIG_SCSI
-void __init note_scsi_host(struct device_node *node, void *host)
+void note_scsi_host(struct device_node *node, void *host)
 {
-	int l;
-	char *p;
-
-	l = strlen(node->full_name);
-	if (bootpath != NULL && bootdevice != NULL
-	    && strncmp(node->full_name, bootdevice, l) == 0
-	    && (bootdevice[l] == '/' || bootdevice[l] == 0)) {
-		boot_host = host;
-		/*
-		 * There's a bug in OF 1.0.5.  (Why am I not surprised.)
-		 * If you pass a path like scsi/sd@1:0 to canon, it returns
-		 * something like /bandit@F2000000/gc@10/53c94@10000/sd@0,0
-		 * That is, the scsi target number doesn't get preserved.
-		 * So we pick the target number out of bootpath and use that.
-		 */
-		p = strstr(bootpath, "/sd@");
-		if (p != NULL) {
-			p += 4;
-			boot_target = simple_strtoul(p, NULL, 10);
-			p = strchr(p, ':');
-			if (p != NULL)
-				boot_part = simple_strtoul(p + 1, NULL, 10);
-		}
-	}
+	/* Obsolete */
 }
 #endif
 
-#if defined(CONFIG_BLK_DEV_IDE) && defined(CONFIG_BLK_DEV_IDE_PMAC)
-static dev_t __init find_ide_boot(void)
-{
-	char *p;
-	int n;
-	dev_t __init pmac_find_ide_boot(char *bootdevice, int n);
-
-	if (bootdevice == NULL)
-		return 0;
-	p = strrchr(bootdevice, '/');
-	if (p == NULL)
-		return 0;
-	n = p - bootdevice;
-
-	return pmac_find_ide_boot(bootdevice, n);
-}
-#endif /* CONFIG_BLK_DEV_IDE && CONFIG_BLK_DEV_IDE_PMAC */
-
-void __init find_boot_device(void)
-{
-#if defined(CONFIG_BLK_DEV_IDE) && defined(CONFIG_BLK_DEV_IDE_PMAC)
-	boot_dev = find_ide_boot();
-#endif
-}
 
 static int initializing = 1;
 
@@ -258,7 +200,7 @@
 /* can't be __init - can be called whenever a disk is first accessed */
 void __pmac note_bootable_part(dev_t dev, int part, int goodness)
 {
-	static int found_boot = 0;
+	extern dev_t boot_dev;
 	char *p;
 
 	if (!initializing)
@@ -270,10 +212,6 @@
 	if (p != NULL && (p == saved_command_line || p[-1] == ' '))
 		return;
 
-	if (!found_boot) {
-		find_boot_device();
-		found_boot = 1;
-	}
 	if (!boot_dev || dev == boot_dev) {
 		ROOT_DEV = dev + part;
 		boot_dev = 0;
@@ -315,9 +253,6 @@
 
 /* 
  * Early initialization.
- * Relocation is on but do not reference unbolted pages
- * Also, device-tree hasn't been "finished", so don't muck with
- * it too much
  */
 void __init pmac_init_early(void)
 {
@@ -417,17 +352,30 @@
 
 core_initcall(pmac_irq_cascade_init);
 
-void __init pmac_init(unsigned long r3, unsigned long r4, unsigned long r5,
-		      unsigned long r6, unsigned long r7)
+static void __init pmac_progress(char *s, unsigned short hex)
+{
+	if (sccdbg) {
+		udbg_puts(s);
+		udbg_putc('\n');
+	}
+#ifdef CONFIG_BOOTX_TEXT
+	else if (boot_text_mapped) {
+		btext_drawstring(s);
+		btext_drawchar('\n');
+	}
+#endif /* CONFIG_BOOTX_TEXT */
+}
+
+void __init pmac_init(void)
 {
 	/* Probe motherboard chipset */
 	pmac_feature_init();
 
 	/* Init SCC */
-	if (strstr(cmd_line, "sccdbg")) {
+       	if (strstr(cmd_line, "sccdbg")) {
 		sccdbg = 1;
 		udbg_init_scc(NULL);
-	}
+       	}
 
 	/* Fill up the machine description */
 	ppc_md.setup_arch     = pmac_setup_arch;
@@ -450,28 +398,12 @@
 	ppc_md.feature_call   = pmac_do_feature_call;
 
 
-#ifdef CONFIG_BOOTX_TEXT
 	ppc_md.progress       = pmac_progress;
-#endif /* CONFIG_BOOTX_TEXT */
 
 	if (ppc_md.progress) ppc_md.progress("pmac_init(): exit", 0);
 
 }
 
-#ifdef CONFIG_BOOTX_TEXT
-void __init pmac_progress(char *s, unsigned short hex)
-{
-	if (sccdbg) {
-		udbg_puts(s);
-		udbg_putc('\n');
-	}
-	else if (boot_text_mapped) {
-		btext_drawstring(s);
-		btext_drawchar('\n');
-	}
-}
-#endif /* CONFIG_BOOTX_TEXT */
-
 static int __init pmac_declare_of_platform_devices(void)
 {
 	struct device_node *np;
diff -urN linux-2.5/arch/ppc64/kernel/ppc_ksyms.c linux-prom-clean/arch/ppc64/kernel/ppc_ksyms.c
--- linux-2.5/arch/ppc64/kernel/ppc_ksyms.c	2004-09-09 17:33:52.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/ppc_ksyms.c	2004-09-15 21:53:26.000000000 +1000
@@ -119,7 +119,6 @@
 EXPORT_SYMBOL(kernel_thread);
 
 EXPORT_SYMBOL(flush_instruction_cache);
-EXPORT_SYMBOL(_get_PVR);
 EXPORT_SYMBOL(giveup_fpu);
 #ifdef CONFIG_ALTIVEC
 EXPORT_SYMBOL(giveup_altivec);
diff -urN linux-2.5/arch/ppc64/kernel/prom.c linux-prom-clean/arch/ppc64/kernel/prom.c
--- linux-2.5/arch/ppc64/kernel/prom.c	2004-09-15 16:25:15.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/prom.c	2004-09-17 22:10:35.000000000 +1000
@@ -15,7 +15,7 @@
  *      2 of the License, or (at your option) any later version.
  */
 
-#undef DEBUG_PROM
+#undef DEBUG
 
 #include <stdarg.h>
 #include <linux/config.h>
@@ -54,1696 +54,53 @@
 #include <asm/machdep.h>
 #include "open_pic.h"
 
-#ifdef CONFIG_LOGO_LINUX_CLUT224
-#include <linux/linux_logo.h>
-extern const struct linux_logo logo_linux_clut224;
-#endif
-
-/*
- * Properties whose value is longer than this get excluded from our
- * copy of the device tree. This value does need to be big enough to
- * ensure that we don't lose things like the interrupt-map property
- * on a PCI-PCI bridge.
- */
-#define MAX_PROPERTY_LENGTH	(1UL * 1024 * 1024)
-
-/*
- * prom_init() is called very early on, before the kernel text
- * and data have been mapped to KERNELBASE.  At this point the code
- * is running at whatever address it has been loaded at, so
- * references to extern and static variables must be relocated
- * explicitly.  The procedure reloc_offset() returns the address
- * we're currently running at minus the address we were linked at.
- * (Note that strings count as static variables.)
- *
- * Because OF may have mapped I/O devices into the area starting at
- * KERNELBASE, particularly on CHRP machines, we can't safely call
- * OF once the kernel has been mapped to KERNELBASE.  Therefore all
- * OF calls should be done within prom_init(), and prom_init()
- * and all routines called within it must be careful to relocate
- * references as necessary.
- *
- * Note that the bss is cleared *after* prom_init runs, so we have
- * to make sure that any static or extern variables it accesses
- * are put in the data segment.
- */
-
-
-#define PROM_BUG() do {						\
-        prom_printf("kernel BUG at %s line 0x%x!\n",		\
-		    RELOC(__FILE__), __LINE__);			\
-        __asm__ __volatile__(".long " BUG_ILLEGAL_INSTR);	\
-} while (0)
-
-#ifdef DEBUG_PROM
-#define prom_debug(x...)	prom_printf(x)
-#else
-#define prom_debug(x...)
-#endif
-
-
-struct pci_reg_property {
-	struct pci_address addr;
-	u32 size_hi;
-	u32 size_lo;
-};
-
-
-struct isa_reg_property {
-	u32 space;
-	u32 address;
-	u32 size;
-};
-
-struct pci_intr_map {
-	struct pci_address addr;
-	u32 dunno;
-	phandle int_ctrler;
-	u32 intr;
-};
-
-
-typedef unsigned long interpret_func(struct device_node *, unsigned long,
-				     int, int, int);
-
-#ifndef FB_MAX			/* avoid pulling in all of the fb stuff */
-#define FB_MAX	8
-#endif
-
-/* prom structure */
-struct prom_t prom;
-
-char *prom_display_paths[FB_MAX] __initdata = { NULL, };
-phandle prom_display_nodes[FB_MAX] __initdata;
-unsigned int prom_num_displays = 0;
-char *of_stdout_device = NULL;
-
-static int iommu_force_on;
-int ppc64_iommu_off;
-
-extern struct rtas_t rtas;
-extern unsigned long klimit;
-extern struct lmb lmb;
-
-#define MAX_PHB (32 * 6)  /* 32 drawers * 6 PHBs/drawer */
-struct of_tce_table of_tce_table[MAX_PHB + 1];
-
-char *bootpath = NULL;
-char *bootdevice = NULL;
-
-int boot_cpuid = 0;
-#define MAX_CPU_THREADS 2
-
-struct device_node *allnodes = NULL;
-/* use when traversing tree through the allnext, child, sibling,
- * or parent members of struct device_node.
- */
-static rwlock_t devtree_lock = RW_LOCK_UNLOCKED;
-
-extern unsigned long reloc_offset(void);
-
-extern void enter_prom(struct prom_args *args);
-extern void copy_and_flush(unsigned long dest, unsigned long src,
-			   unsigned long size, unsigned long offset);
-
-unsigned long dev_tree_size;
-unsigned long _get_PIR(void);
-
-#ifdef CONFIG_HMT
-struct {
-	unsigned int pir;
-	unsigned int threadid;
-} hmt_thread_data[NR_CPUS];
-#endif /* CONFIG_HMT */
-
-char testString[] = "LINUX\n"; 
-
-/*
- * This are used in calls to call_prom.  The 4th and following
- * arguments to call_prom should be 32-bit values.  64 bit values
- * are truncated to 32 bits (and fortunately don't get interpreted
- * as two arguments).
- */
-#define ADDR(x)		(u32) ((unsigned long)(x) - offset)
-
-/* This is the one and *ONLY* place where we actually call open
- * firmware from, since we need to make sure we're running in 32b
- * mode when we do.  We switch back to 64b mode upon return.
- */
-
-#define PROM_ERROR	(-1)
-
-static int __init call_prom(const char *service, int nargs, int nret, ...)
-{
-	int i;
-	unsigned long offset = reloc_offset();
-	struct prom_t *_prom = PTRRELOC(&prom);
-	va_list list;
-
-	_prom->args.service = ADDR(service);
-	_prom->args.nargs = nargs;
-	_prom->args.nret = nret;
-	_prom->args.rets = (prom_arg_t *)&(_prom->args.args[nargs]);
-
-	va_start(list, nret);
-	for (i=0; i < nargs; i++)
-		_prom->args.args[i] = va_arg(list, prom_arg_t);
-	va_end(list);
-
-	for (i=0; i < nret ;i++)
-		_prom->args.rets[i] = 0;
-
-	enter_prom(&_prom->args);
-
-	return (nret > 0)? _prom->args.rets[0]: 0;
-}
-
-
-static void __init prom_print(const char *msg)
-{
-	const char *p, *q;
-	unsigned long offset = reloc_offset();
-	struct prom_t *_prom = PTRRELOC(&prom);
-
-	if (_prom->stdout == 0)
-		return;
-
-	for (p = msg; *p != 0; p = q) {
-		for (q = p; *q != 0 && *q != '\n'; ++q)
-			;
-		if (q > p)
-			call_prom("write", 3, 1, _prom->stdout, p, q - p);
-		if (*q == 0)
-			break;
-		++q;
-		call_prom("write", 3, 1, _prom->stdout, ADDR("\r\n"), 2);
-	}
-}
-
-
-static void __init prom_print_hex(unsigned long val)
-{
-	unsigned long offset = reloc_offset();
-	int i, nibbles = sizeof(val)*2;
-	char buf[sizeof(val)*2+1];
-	struct prom_t *_prom = PTRRELOC(&prom);
-
-	for (i = nibbles-1;  i >= 0;  i--) {
-		buf[i] = (val & 0xf) + '0';
-		if (buf[i] > '9')
-			buf[i] += ('a'-'0'-10);
-		val >>= 4;
-	}
-	buf[nibbles] = '\0';
-	call_prom("write", 3, 1, _prom->stdout, buf, nibbles);
-}
-
-
-static void __init prom_printf(const char *format, ...)
-{
-	unsigned long offset = reloc_offset();
-	const char *p, *q, *s;
-	va_list args;
-	unsigned long v;
-	struct prom_t *_prom = PTRRELOC(&prom);
-
-	va_start(args, format);
-	for (p = PTRRELOC(format); *p != 0; p = q) {
-		for (q = p; *q != 0 && *q != '\n' && *q != '%'; ++q)
-			;
-		if (q > p)
-			call_prom("write", 3, 1, _prom->stdout, p, q - p);
-		if (*q == 0)
-			break;
-		if (*q == '\n') {
-			++q;
-			call_prom("write", 3, 1, _prom->stdout,
-				  ADDR("\r\n"), 2);
-			continue;
-		}
-		++q;
-		if (*q == 0)
-			break;
-		switch (*q) {
-		case 's':
-			++q;
-			s = va_arg(args, const char *);
-			prom_print(s);
-			break;
-		case 'x':
-			++q;
-			v = va_arg(args, unsigned long);
-			prom_print_hex(v);
-			break;
-		}
-	}
-}
-
-
-static void __init __attribute__((noreturn)) prom_panic(const char *reason)
-{
-	unsigned long offset = reloc_offset();
-
-	prom_print(PTRRELOC(reason));
-	/* ToDo: should put up an SRC here */
-	call_prom("exit", 0, 0);
-
-	for (;;)			/* should never get here */
-		;
-}
-
-
-static int __init prom_next_node(phandle *nodep)
-{
-	phandle node;
-
-	if ((node = *nodep) != 0
-	    && (*nodep = call_prom("child", 1, 1, node)) != 0)
-		return 1;
-	if ((*nodep = call_prom("peer", 1, 1, node)) != 0)
-		return 1;
-	for (;;) {
-		if ((node = call_prom("parent", 1, 1, node)) == 0)
-			return 0;
-		if ((*nodep = call_prom("peer", 1, 1, node)) != 0)
-			return 1;
-	}
-}
-
-static int __init prom_getprop(phandle node, const char *pname,
-			       void *value, size_t valuelen)
-{
-	unsigned long offset = reloc_offset();
-
-	return call_prom("getprop", 4, 1, node, ADDR(pname),
-			 (u32)(unsigned long) value, (u32) valuelen);
-}
-
-static void __init prom_initialize_naca(void)
-{
-	phandle node;
-	char type[64];
-	unsigned long num_cpus = 0;
-	unsigned long offset = reloc_offset();
-	struct prom_t *_prom = PTRRELOC(&prom);
-	struct naca_struct *_naca = RELOC(naca);
-	struct systemcfg *_systemcfg = RELOC(systemcfg);
-
-	/* NOTE: _naca->debug_switch is already initialized. */
-	prom_debug("prom_initialize_naca: start...\n");
-
-	_naca->pftSize = 0;	/* ilog2 of htab size.  computed below. */
-
-	for (node = 0; prom_next_node(&node); ) {
-		type[0] = 0;
-		prom_getprop(node, "device_type", type, sizeof(type));
-
-		if (!strcmp(type, RELOC("cpu"))) {
-			num_cpus += 1;
-
-			/* We're assuming *all* of the CPUs have the same
-			 * d-cache and i-cache sizes... -Peter
-			 */
-			if ( num_cpus == 1 ) {
-				u32 size, lsize;
-				const char *dc, *ic;
-
-				if (_systemcfg->platform == PLATFORM_POWERMAC){
-					dc = "d-cache-block-size";
-					ic = "i-cache-block-size";
-				} else {
-					dc = "d-cache-line-size";
-					ic = "i-cache-line-size";
-				}
-
-				prom_getprop(node, "d-cache-size",
-					     &size, sizeof(size));
-
-				prom_getprop(node, dc, &lsize, sizeof(lsize));
-
-				_systemcfg->dCacheL1Size = size;
-				_systemcfg->dCacheL1LineSize = lsize;
-				_naca->dCacheL1LogLineSize = __ilog2(lsize);
-				_naca->dCacheL1LinesPerPage = PAGE_SIZE/lsize;
-
-				prom_getprop(node, "i-cache-size",
-					     &size, sizeof(size));
-
-				prom_getprop(node, ic, &lsize, sizeof(lsize));
-
-				_systemcfg->iCacheL1Size = size;
-				_systemcfg->iCacheL1LineSize = lsize;
-				_naca->iCacheL1LogLineSize = __ilog2(lsize);
-				_naca->iCacheL1LinesPerPage = PAGE_SIZE/lsize;
-
-				if (_systemcfg->platform == PLATFORM_PSERIES_LPAR) {
-					u32 pft_size[2];
-					prom_getprop(node, "ibm,pft-size",
-						&pft_size, sizeof(pft_size));
-				/* pft_size[0] is the NUMA CEC cookie */
-					_naca->pftSize = pft_size[1];
-				}
-			}
-		} else if (!strcmp(type, RELOC("serial"))) {
-			phandle isa, pci;
-			struct isa_reg_property reg;
-			union pci_range ranges;
-
-			if (_systemcfg->platform == PLATFORM_POWERMAC)
-				continue;
-			type[0] = 0;
-			prom_getprop(node, "ibm,aix-loc", type, sizeof(type));
-
-			if (strcmp(type, RELOC("S1")))
-				continue;
-
-			prom_getprop(node, "reg", &reg, sizeof(reg));
-
-			isa = call_prom("parent", 1, 1, node);
-			if (!isa)
-				PROM_BUG();
-			pci = call_prom("parent", 1, 1, isa);
-			if (!pci)
-				PROM_BUG();
-
-			prom_getprop(pci, "ranges", &ranges, sizeof(ranges));
-
-			if ( _prom->encode_phys_size == 32 )
-				_naca->serialPortAddr = ranges.pci32.phys+reg.address;
-			else {
-				_naca->serialPortAddr = 
-					((((unsigned long)ranges.pci64.phys_hi) << 32) |
-					 (ranges.pci64.phys_lo)) + reg.address;
-			}
-		}
-	}
-
-	if (_systemcfg->platform == PLATFORM_POWERMAC)
-		_naca->interrupt_controller = IC_OPEN_PIC;
-	else {
-		_naca->interrupt_controller = IC_INVALID;
-		for (node = 0; prom_next_node(&node); ) {
-			type[0] = 0;
-			prom_getprop(node, "name", type, sizeof(type));
-			if (strcmp(type, RELOC("interrupt-controller")))
-				continue;
-			prom_getprop(node, "compatible", type, sizeof(type));
-			if (strstr(type, RELOC("open-pic")))
-				_naca->interrupt_controller = IC_OPEN_PIC;
-			else if (strstr(type, RELOC("ppc-xicp")))
-				_naca->interrupt_controller = IC_PPC_XIC;
-			else
-				prom_printf("prom: failed to recognize"
-					    " interrupt-controller\n");
-			break;
-		}
-	}
-
-	if (_naca->interrupt_controller == IC_INVALID) {
-		prom_printf("prom: failed to find interrupt-controller\n");
-		PROM_BUG();
-	}
-
-	/* We gotta have at least 1 cpu... */
-	if ( (_systemcfg->processorCount = num_cpus) < 1 )
-		PROM_BUG();
-
-	_systemcfg->physicalMemorySize = lmb_phys_mem_size();
-
-	if (_systemcfg->platform == PLATFORM_PSERIES ||
-	    _systemcfg->platform == PLATFORM_POWERMAC) {
-		unsigned long rnd_mem_size, pteg_count;
-
-		/* round mem_size up to next power of 2 */
-		rnd_mem_size = 1UL << __ilog2(_systemcfg->physicalMemorySize);
-		if (rnd_mem_size < _systemcfg->physicalMemorySize)
-			rnd_mem_size <<= 1;
-
-		/* # pages / 2 */
-		pteg_count = (rnd_mem_size >> (12 + 1));
-
-		_naca->pftSize = __ilog2(pteg_count << 7);
-	}
-
-	if (_naca->pftSize == 0) {
-		prom_printf("prom: failed to compute pftSize!\n");
-		PROM_BUG();
-	}
-
-	/* Add an eye catcher and the systemcfg layout version number */
-	strcpy(_systemcfg->eye_catcher, RELOC("SYSTEMCFG:PPC64"));
-	_systemcfg->version.major = SYSTEMCFG_MAJOR;
-	_systemcfg->version.minor = SYSTEMCFG_MINOR;
-	_systemcfg->processor = _get_PVR();
-
-	prom_debug("systemcfg->processorCount       = 0x%x\n",
-		   _systemcfg->processorCount);
-	prom_debug("systemcfg->physicalMemorySize   = 0x%x\n",
-		   _systemcfg->physicalMemorySize);
-	prom_debug("naca->pftSize                   = 0x%x\n",
-		   _naca->pftSize);
-	prom_debug("systemcfg->dCacheL1LineSize     = 0x%x\n",
-		   _systemcfg->dCacheL1LineSize);
-	prom_debug("systemcfg->iCacheL1LineSize     = 0x%x\n",
-		   _systemcfg->iCacheL1LineSize);
-	prom_debug("naca->serialPortAddr            = 0x%x\n",
-		   _naca->serialPortAddr);
-	prom_debug("naca->interrupt_controller      = 0x%x\n",
-		   _naca->interrupt_controller);
-	prom_debug("systemcfg->platform             = 0x%x\n",
-		   _systemcfg->platform);
-	prom_debug("prom_initialize_naca: end...\n");
-}
-
-
-static void __init early_cmdline_parse(void)
-{
-	unsigned long offset = reloc_offset();
-	char *opt;
-#ifndef CONFIG_PMAC_DART
-	struct systemcfg *_systemcfg = RELOC(systemcfg);
-#endif
-
-	opt = strstr(RELOC(cmd_line), RELOC("iommu="));
-	if (opt) {
-		prom_printf("opt is:%s\n", opt);
-		opt += 6;
-		while (*opt && *opt == ' ')
-			opt++;
-		if (!strncmp(opt, RELOC("off"), 3))
-			RELOC(ppc64_iommu_off) = 1;
-		else if (!strncmp(opt, RELOC("force"), 5))
-			RELOC(iommu_force_on) = 1;
-	}
-
-#ifndef CONFIG_PMAC_DART
-	if (_systemcfg->platform == PLATFORM_POWERMAC) {
-		RELOC(ppc64_iommu_off) = 1;
-		prom_printf("DART disabled on PowerMac !\n");
-	}
-#endif
-}
-
-#ifdef DEBUG_PROM
-void prom_dump_lmb(void)
-{
-	unsigned long i;
-	unsigned long offset = reloc_offset();
-	struct lmb *_lmb  = PTRRELOC(&lmb);
-
-	prom_printf("\nprom_dump_lmb:\n");
-	prom_printf("    memory.cnt		  = 0x%x\n",
-		    _lmb->memory.cnt);
-	prom_printf("    memory.size		 = 0x%x\n",
-		    _lmb->memory.size);
-	for (i=0; i < _lmb->memory.cnt ;i++) {
-		prom_printf("    memory.region[0x%x].base       = 0x%x\n",
-			    i, _lmb->memory.region[i].base);
-		prom_printf("		      .physbase = 0x%x\n",
-			    _lmb->memory.region[i].physbase);
-		prom_printf("		      .size     = 0x%x\n",
-			    _lmb->memory.region[i].size);
-	}
-
-	prom_printf("\n    reserved.cnt		  = 0x%x\n",
-		    _lmb->reserved.cnt);
-	prom_printf("    reserved.size		 = 0x%x\n",
-		    _lmb->reserved.size);
-	for (i=0; i < _lmb->reserved.cnt ;i++) {
-		prom_printf("    reserved.region[0x%x\n].base       = 0x%x\n",
-			    i, _lmb->reserved.region[i].base);
-		prom_printf("		      .physbase = 0x%x\n",
-			    _lmb->reserved.region[i].physbase);
-		prom_printf("		      .size     = 0x%x\n",
-			    _lmb->reserved.region[i].size);
-	}
-}
-#endif /* DEBUG_PROM */
-
-static void __init prom_initialize_lmb(void)
-{
-	phandle node;
-	char type[64];
-	unsigned long i, offset = reloc_offset();
-	struct prom_t *_prom = PTRRELOC(&prom);
-	struct systemcfg *_systemcfg = RELOC(systemcfg);
-	union lmb_reg_property reg;
-	unsigned long lmb_base, lmb_size;
-	unsigned long num_regs, bytes_per_reg = (_prom->encode_phys_size*2)/8;
-
-	lmb_init();
-
-	/* XXX Quick HACK. Proper fix is to drop those structures and properly use
-	 * #address-cells. PowerMac has #size-cell set to 1 and #address-cells to 2
-	 */
-	if (_systemcfg->platform == PLATFORM_POWERMAC)
-		bytes_per_reg = 12;
-
-	for (node = 0; prom_next_node(&node); ) {
-		type[0] = 0;
-		prom_getprop(node, "device_type", type, sizeof(type));
-
-		if (strcmp(type, RELOC("memory")))
-			continue;
-
-		num_regs = prom_getprop(node, "reg", &reg, sizeof(reg))
-			/ bytes_per_reg;
-
-		for (i=0; i < num_regs ;i++) {
-			if (_systemcfg->platform == PLATFORM_POWERMAC) {
-				lmb_base = ((unsigned long)reg.addrPM[i].address_hi) << 32;
-				lmb_base |= (unsigned long)reg.addrPM[i].address_lo;
-				lmb_size = reg.addrPM[i].size;
-			} else if (_prom->encode_phys_size == 32) {
-				lmb_base = reg.addr32[i].address;
-				lmb_size = reg.addr32[i].size;
-			} else {
-				lmb_base = reg.addr64[i].address;
-				lmb_size = reg.addr64[i].size;
-			}
-
-			/* We limit memory to 2GB if the IOMMU is off */
-			if (RELOC(ppc64_iommu_off)) {
-				if (lmb_base >= 0x80000000UL)
-					continue;
-
-				if ((lmb_base + lmb_size) > 0x80000000UL)
-					lmb_size = 0x80000000UL - lmb_base;
-			}
-
-			if (lmb_add(lmb_base, lmb_size) < 0)
-				prom_printf("Too many LMB's, discarding this one...\n");
-		}
-
-	}
-
-	lmb_analyze();
-#ifdef DEBUG_PROM
-	prom_dump_lmb();
-#endif /* DEBUG_PROM */
-}
-
-static void __init
-prom_instantiate_rtas(void)
-{
-	unsigned long offset = reloc_offset();
-	struct prom_t *_prom = PTRRELOC(&prom);
-	struct rtas_t *_rtas = PTRRELOC(&rtas);
-	struct systemcfg *_systemcfg = RELOC(systemcfg);
-	ihandle prom_rtas;
-	u32 getprop_rval;
-	char hypertas_funcs[4];
-
-	prom_debug("prom_instantiate_rtas: start...\n");
-
-	prom_rtas = call_prom("finddevice", 1, 1, ADDR("/rtas"));
-	if (prom_rtas != (ihandle) -1) {
-		unsigned long x;
-		x = prom_getprop(prom_rtas, "ibm,hypertas-functions",
-				 hypertas_funcs, sizeof(hypertas_funcs));
-
-		if (x != PROM_ERROR) {
-			prom_printf("Hypertas detected, assuming LPAR !\n");
-			_systemcfg->platform = PLATFORM_PSERIES_LPAR;
-		}
-
-		prom_getprop(prom_rtas, "rtas-size",
-			     &getprop_rval, sizeof(getprop_rval));
-		_rtas->size = getprop_rval;
-		prom_printf("instantiating rtas");
-		if (_rtas->size != 0) {
-			unsigned long rtas_region = RTAS_INSTANTIATE_MAX;
-
-			/* Grab some space within the first RTAS_INSTANTIATE_MAX bytes
-			 * of physical memory (or within the RMO region) because RTAS
-			 * runs in 32-bit mode and relocate off.
-			 */
-			if ( _systemcfg->platform == PLATFORM_PSERIES_LPAR ) {
-				struct lmb *_lmb  = PTRRELOC(&lmb);
-				rtas_region = min(_lmb->rmo_size, RTAS_INSTANTIATE_MAX);
-			}
-
-			_rtas->base = lmb_alloc_base(_rtas->size, PAGE_SIZE, rtas_region);
-
-			prom_printf(" at 0x%x", _rtas->base);
-
-			prom_rtas = call_prom("open", 1, 1, ADDR("/rtas"));
-			prom_printf("...");
-
-			if (call_prom("call-method", 3, 2,
-				      ADDR("instantiate-rtas"),
-				      prom_rtas,
-				      _rtas->base) != PROM_ERROR) {
-				_rtas->entry = (long)_prom->args.rets[1];
-			}
-			RELOC(rtas_rmo_buf)
-				= lmb_alloc_base(RTAS_RMOBUF_MAX, PAGE_SIZE,
-							rtas_region);
-		}
-
-		if (_rtas->entry <= 0) {
-			prom_printf(" failed\n");
-		} else {
-			prom_printf(" done\n");
-		}
-
-		prom_debug("rtas->base		= 0x%x\n", _rtas->base);
-		prom_debug("rtas->entry	       = 0x%x\n", _rtas->entry);
-		prom_debug("rtas->size		= 0x%x\n", _rtas->size);
-	}
-	prom_debug("prom_instantiate_rtas: end...\n");
-}
-
-
-#ifdef CONFIG_PMAC_DART
-static void __init prom_initialize_dart_table(void)
-{
-	unsigned long offset = reloc_offset();
-	extern unsigned long dart_tablebase;
-	extern unsigned long dart_tablesize;
-
-	/* Only reserve DART space if machine has more than 2GB of RAM
-	 * or if requested with iommu=on on cmdline.
-	 */
-	if (lmb_end_of_DRAM() <= 0x80000000ull && !RELOC(iommu_force_on))
-		return;
-
-	/* 512 pages (2MB) is max DART tablesize. */
-	RELOC(dart_tablesize) = 1UL << 21;
-	/* 16MB (1 << 24) alignment. We allocate a full 16Mb chuck since we
-	 * will blow up an entire large page anyway in the kernel mapping
-	 */
-	RELOC(dart_tablebase) = (unsigned long)
-		abs_to_virt(lmb_alloc_base(1UL<<24, 1UL<<24, 0x80000000L));
-
-	prom_printf("Dart at: %x\n", RELOC(dart_tablebase));
-}
-#endif /* CONFIG_PMAC_DART */
-
-static void __init prom_initialize_tce_table(void)
-{
-	phandle node;
-	ihandle phb_node;
-	unsigned long offset = reloc_offset();
-	char compatible[64], path[64], type[64], model[64];
-	unsigned long i, table = 0;
-	unsigned long base, vbase, align;
-	unsigned int minalign, minsize;
-	struct of_tce_table *prom_tce_table = RELOC(of_tce_table);
-	unsigned long tce_entry, *tce_entryp;
-
-	if (RELOC(ppc64_iommu_off))
-		return;
-
-	prom_debug("starting prom_initialize_tce_table\n");
-
-	/* Search all nodes looking for PHBs. */
-	for (node = 0; prom_next_node(&node); ) {
-		if (table == MAX_PHB) {
-			prom_printf("WARNING: PCI host bridge ignored, "
-				    "need to increase MAX_PHB\n");
-			continue;
-		}
-
-		compatible[0] = 0;
-		type[0] = 0;
-		model[0] = 0;
-		prom_getprop(node, "compatible",
-			     compatible, sizeof(compatible));
-		prom_getprop(node, "device_type", type, sizeof(type));
-		prom_getprop(node, "model", model, sizeof(model));
-
-		/* Keep the old logic in tack to avoid regression. */
-		if (compatible[0] != 0) {
-			if ((strstr(compatible, RELOC("python")) == NULL) &&
-			    (strstr(compatible, RELOC("Speedwagon")) == NULL) &&
-			    (strstr(compatible, RELOC("Winnipeg")) == NULL))
-				continue;
-		} else if (model[0] != 0) {
-			if ((strstr(model, RELOC("ython")) == NULL) &&
-			    (strstr(model, RELOC("peedwagon")) == NULL) &&
-			    (strstr(model, RELOC("innipeg")) == NULL))
-				continue;
-		}
-
-		if ((type[0] == 0) || (strstr(type, RELOC("pci")) == NULL)) {
-			continue;
-		}
-
-		if (prom_getprop(node, "tce-table-minalign", &minalign,
-				 sizeof(minalign)) == PROM_ERROR) {
-			minalign = 0;
-		}
-
-		if (prom_getprop(node, "tce-table-minsize", &minsize,
-				 sizeof(minsize)) == PROM_ERROR) {
-			minsize = 4UL << 20;
-		}
-
-		/*
-		 * Even though we read what OF wants, we just set the table
-		 * size to 4 MB.  This is enough to map 2GB of PCI DMA space.
-		 * By doing this, we avoid the pitfalls of trying to DMA to
-		 * MMIO space and the DMA alias hole.
-		 *
-		 * On POWER4, firmware sets the TCE region by assuming
-		 * each TCE table is 8MB. Using this memory for anything
-		 * else will impact performance, so we always allocate 8MB.
-		 * Anton
-		 */
-		if (__is_processor(PV_POWER4) || __is_processor(PV_POWER4p))
-			minsize = 8UL << 20;
-		else
-			minsize = 4UL << 20;
-
-		/* Align to the greater of the align or size */
-		align = max(minalign, minsize);
-
-		/* Carve out storage for the TCE table. */
-		base = lmb_alloc(minsize, align);
-
-		if ( !base ) {
-			prom_panic("ERROR, cannot find space for TCE table.\n");
-		}
-
-		vbase = (unsigned long)abs_to_virt(base);
-
-		/* Save away the TCE table attributes for later use. */
-		prom_tce_table[table].node = node;
-		prom_tce_table[table].base = vbase;
-		prom_tce_table[table].size = minsize;
-
-		prom_debug("TCE table: 0x%x\n", table);
-		prom_debug("\tnode = 0x%x\n", node);
-		prom_debug("\tbase = 0x%x\n", vbase);
-		prom_debug("\tsize = 0x%x\n", minsize);
-
-		/* Initialize the table to have a one-to-one mapping
-		 * over the allocated size.
-		 */
-		tce_entryp = (unsigned long *)base;
-		for (i = 0; i < (minsize >> 3) ;tce_entryp++, i++) {
-			tce_entry = (i << PAGE_SHIFT);
-			tce_entry |= 0x3;
-			*tce_entryp = tce_entry;
-		}
-
-		/* It seems OF doesn't null-terminate the path :-( */
-		memset(path, 0, sizeof(path));
-		/* Call OF to setup the TCE hardware */
-		if (call_prom("package-to-path", 3, 1, node,
-			      path, sizeof(path)-1) == PROM_ERROR) {
-			prom_printf("package-to-path failed\n");
-		} else {
-			prom_printf("opening PHB %s", path);
-		}
-
-		phb_node = call_prom("open", 1, 1, path);
-		if ( (long)phb_node <= 0) {
-			prom_printf("... failed\n");
-		} else {
-			prom_printf("... done\n");
-		}
-		call_prom("call-method", 6, 0, ADDR("set-64-bit-addressing"),
-			  phb_node, -1, minsize,
-			  (u32) base, (u32) (base >> 32));
-		call_prom("close", 1, 0, phb_node);
-
-		table++;
-	}
-
-	/* Flag the first invalid entry */
-	prom_tce_table[table].node = 0;
-	prom_debug("ending prom_initialize_tce_table\n");
-}
-
-/*
- * With CHRP SMP we need to use the OF to start the other
- * processors so we can't wait until smp_boot_cpus (the OF is
- * trashed by then) so we have to put the processors into
- * a holding pattern controlled by the kernel (not OF) before
- * we destroy the OF.
- *
- * This uses a chunk of low memory, puts some holding pattern
- * code there and sends the other processors off to there until
- * smp_boot_cpus tells them to do something.  The holding pattern
- * checks that address until its cpu # is there, when it is that
- * cpu jumps to __secondary_start().  smp_boot_cpus() takes care
- * of setting those values.
- *
- * We also use physical address 0x4 here to tell when a cpu
- * is in its holding pattern code.
- *
- * Fixup comment... DRENG / PPPBBB - Peter
- *
- * -- Cort
- */
-static void __init prom_hold_cpus(unsigned long mem)
-{
-	unsigned long i;
-	unsigned int reg;
-	phandle node;
-	unsigned long offset = reloc_offset();
-	char type[64], *path;
-	int cpuid = 0;
-	unsigned int interrupt_server[MAX_CPU_THREADS];
-	unsigned int cpu_threads, hw_cpu_num;
-	int propsize;
-	extern void __secondary_hold(void);
-	extern unsigned long __secondary_hold_spinloop;
-	extern unsigned long __secondary_hold_acknowledge;
-	unsigned long *spinloop
-		= (void *)virt_to_abs(&__secondary_hold_spinloop);
-	unsigned long *acknowledge
-		= (void *)virt_to_abs(&__secondary_hold_acknowledge);
-	unsigned long secondary_hold
-		= virt_to_abs(*PTRRELOC((unsigned long *)__secondary_hold));
-	struct prom_t *_prom = PTRRELOC(&prom);
-
-	prom_debug("prom_hold_cpus: start...\n");
-	prom_debug("    1) spinloop       = 0x%x\n", (unsigned long)spinloop);
-	prom_debug("    1) *spinloop      = 0x%x\n", *spinloop);
-	prom_debug("    1) acknowledge    = 0x%x\n",
-		   (unsigned long)acknowledge);
-	prom_debug("    1) *acknowledge   = 0x%x\n", *acknowledge);
-	prom_debug("    1) secondary_hold = 0x%x\n", secondary_hold);
-
-	/* Set the common spinloop variable, so all of the secondary cpus
-	 * will block when they are awakened from their OF spinloop.
-	 * This must occur for both SMP and non SMP kernels, since OF will
-	 * be trashed when we move the kernel.
-	 */
-	*spinloop = 0;
-
-#ifdef CONFIG_HMT
-	for (i=0; i < NR_CPUS; i++) {
-		RELOC(hmt_thread_data)[i].pir = 0xdeadbeef;
-	}
-#endif
-	/* look for cpus */
-	for (node = 0; prom_next_node(&node); ) {
-		type[0] = 0;
-		prom_getprop(node, "device_type", type, sizeof(type));
-		if (strcmp(type, RELOC("cpu")) != 0)
-			continue;
-
-		/* Skip non-configured cpus. */
-		prom_getprop(node, "status", type, sizeof(type));
-		if (strcmp(type, RELOC("okay")) != 0)
-			continue;
-
-		reg = -1;
-		prom_getprop(node, "reg", &reg, sizeof(reg));
-
-		path = (char *) mem;
-		memset(path, 0, 256);
-		if (call_prom("package-to-path", 3, 1,
-			      node, path, 255) == PROM_ERROR)
-			continue;
-
-		prom_debug("\ncpuid        = 0x%x\n", cpuid);
-		prom_debug("cpu hw idx   = 0x%x\n", reg);
-
-		/* Init the acknowledge var which will be reset by
-		 * the secondary cpu when it awakens from its OF
-		 * spinloop.
-		 */
-		*acknowledge = (unsigned long)-1;
-
-		propsize = prom_getprop(node, "ibm,ppc-interrupt-server#s",
-					&interrupt_server,
-					sizeof(interrupt_server));
-		if (propsize < 0) {
-			/* no property.  old hardware has no SMT */
-			cpu_threads = 1;
-			interrupt_server[0] = reg; /* fake it with phys id */
-		} else {
-			/* We have a threaded processor */
-			cpu_threads = propsize / sizeof(u32);
-			if (cpu_threads > MAX_CPU_THREADS) {
-				prom_printf("SMT: too many threads!\n"
-					    "SMT: found %x, max is %x\n",
-					    cpu_threads, MAX_CPU_THREADS);
-				cpu_threads = 1; /* ToDo: panic? */
-			}
-		}
-
-		hw_cpu_num = interrupt_server[0];
-		if (hw_cpu_num != _prom->cpu) {
-			/* Primary Thread of non-boot cpu */
-			prom_printf("%x : starting cpu %s... ", cpuid, path);
-			call_prom("start-cpu", 3, 0, node,
-				  secondary_hold, cpuid);
-
-			for ( i = 0 ; (i < 100000000) && 
-			      (*acknowledge == ((unsigned long)-1)); i++ ) ;
-
-			if (*acknowledge == cpuid) {
-				prom_printf(" done\n");
-				/* We have to get every CPU out of OF,
-				 * even if we never start it. */
-				if (cpuid >= NR_CPUS)
-					goto next;
-			} else {
-				prom_printf(" failed: %x\n", *acknowledge);
-			}
-		}
-#ifdef CONFIG_SMP
-		else
-			prom_printf("%x : boot cpu     %s\n", cpuid, path);
-#endif
-next:
-#ifdef CONFIG_SMP
-		/* Init paca for secondary threads.   They start later. */
-		for (i=1; i < cpu_threads; i++) {
-			cpuid++;
-			if (cpuid >= NR_CPUS)
-				continue;
-		}
-#endif
-		cpuid++;
-	}
-#ifdef CONFIG_HMT
-	/* Only enable HMT on processors that provide support. */
-	if (__is_processor(PV_PULSAR) || 
-	    __is_processor(PV_ICESTAR) ||
-	    __is_processor(PV_SSTAR)) {
-		prom_printf("    starting secondary threads\n");
-
-		for (i = 0; i < NR_CPUS; i += 2) {
-			if (!cpu_online(i))
-				continue;
-
-			if (i == 0) {
-				unsigned long pir = _get_PIR();
-				if (__is_processor(PV_PULSAR)) {
-					RELOC(hmt_thread_data)[i].pir = 
-						pir & 0x1f;
-				} else {
-					RELOC(hmt_thread_data)[i].pir = 
-						pir & 0x3ff;
-				}
-			}
-		}
-	} else {
-		prom_printf("Processor is not HMT capable\n");
-	}
-#endif
-
-	if (cpuid > NR_CPUS)
-		prom_printf("WARNING: maximum CPUs (" __stringify(NR_CPUS)
-			    ") exceeded: ignoring extras\n");
-
-	prom_debug("prom_hold_cpus: end...\n");
-}
-
-#ifdef CONFIG_BOOTX_TEXT
-
-/* This function will enable the early boot text when doing OF booting. This
- * way, xmon output should work too
- */
-static void __init setup_disp_fake_bi(ihandle dp)
-{
-	int width = 640, height = 480, depth = 8, pitch;
-	unsigned address;
-	struct pci_reg_property addrs[8];
-	int i, naddrs;
-	char name[64];
-	unsigned long offset = reloc_offset();
-
-	memset(name, 0, sizeof(name));
-	prom_getprop(dp, "name", name, sizeof(name));
-	name[sizeof(name)-1] = 0;
-	prom_printf("Initializing fake screen: %s\n", name);
-
-	prom_getprop(dp, "width", &width, sizeof(width));
-	prom_getprop(dp, "height", &height, sizeof(height));
-	prom_getprop(dp, "depth", &depth, sizeof(depth));
-	pitch = width * ((depth + 7) / 8);
-	prom_getprop(dp, "linebytes", &pitch, sizeof(pitch));
-	if (pitch == 1)
-		pitch = 0x1000;		/* for strange IBM display */
-	address = 0;
-
-	prom_printf("width %x height %x depth %x linebytes %x\n",
-		    width, height, depth, depth);
-
-	prom_getprop(dp, "address", &address, sizeof(address));
-	if (address == 0) {
-		/* look for an assigned address with a size of >= 1MB */
-		naddrs = prom_getprop(dp, "assigned-addresses",
-				      addrs, sizeof(addrs));
-		naddrs /= sizeof(struct pci_reg_property);
-		for (i = 0; i < naddrs; ++i) {
-			if (addrs[i].size_lo >= (1 << 20)) {
-				address = addrs[i].addr.a_lo;
-				/* use the BE aperture if possible */
-				if (addrs[i].size_lo >= (16 << 20))
-					address += (8 << 20);
-				break;
-			}
-		}
-		if (address == 0) {
-			prom_printf("Failed to get address of frame buffer\n");
-			return;
-		}
-	}
-	btext_setup_display(width, height, depth, pitch, address);
-	prom_printf("Addr of fb: %x\n", address);
-	RELOC(boot_text_mapped) = 0;
-}
-#endif /* CONFIG_BOOTX_TEXT */
-
-static void __init prom_init_client_services(unsigned long pp)
-{
-	unsigned long offset = reloc_offset();
-	struct prom_t *_prom = PTRRELOC(&prom);
-
-	/* Get a handle to the prom entry point before anything else */
-	_prom->entry = pp;
-
-	/* Init default value for phys size */
-	_prom->encode_phys_size = 32;
-
-	/* get a handle for the stdout device */
-	_prom->chosen = call_prom("finddevice", 1, 1, ADDR("/chosen"));
-	if ((long)_prom->chosen <= 0)
-		prom_panic("cannot find chosen"); /* msg won't be printed :( */
-
-	/* get device tree root */
-	_prom->root = call_prom("finddevice", 1, 1, ADDR("/"));
-	if ((long)_prom->root <= 0)
-		prom_panic("cannot find device tree root"); /* msg won't be printed :( */
-}
-
-static void __init prom_init_stdout(void)
-{
-	unsigned long offset = reloc_offset();
-	struct prom_t *_prom = PTRRELOC(&prom);
-	u32 val;
-
-	if (prom_getprop(_prom->chosen, "stdout", &val, sizeof(val)) <= 0)
-		prom_panic("cannot find stdout");
-
-	_prom->stdout = val;
-}
-
-static int __init prom_find_machine_type(void)
-{
-	unsigned long offset = reloc_offset();
-	struct prom_t *_prom = PTRRELOC(&prom);
-	char compat[256];
-	int len, i = 0;
-
-	len = prom_getprop(_prom->root, "compatible",
-			   compat, sizeof(compat)-1);
-	if (len > 0) {
-		compat[len] = 0;
-		while (i < len) {
-			char *p = &compat[i];
-			int sl = strlen(p);
-			if (sl == 0)
-				break;
-			if (strstr(p, RELOC("Power Macintosh")) ||
-			    strstr(p, RELOC("MacRISC4")))
-				return PLATFORM_POWERMAC;
-			i += sl + 1;
-		}
-	}
-	/* Default to pSeries */
-	return PLATFORM_PSERIES;
-}
-
-static int __init prom_set_color(ihandle ih, int i, int r, int g, int b)
-{
-	unsigned long offset = reloc_offset();
-
-	return call_prom("call-method", 6, 1, ADDR("color!"), ih, i, b, g, r);
-}
-
-/*
- * If we have a display that we don't know how to drive,
- * we will want to try to execute OF's open method for it
- * later.  However, OF will probably fall over if we do that
- * we've taken over the MMU.
- * So we check whether we will need to open the display,
- * and if so, open it now.
- */
-static unsigned long __init check_display(unsigned long mem)
-{
-	phandle node;
-	ihandle ih;
-	int i, j;
-	unsigned long offset = reloc_offset();
-	struct prom_t *_prom = PTRRELOC(&prom);
-	char type[16], *path;
-	static unsigned char default_colors[] = {
-		0x00, 0x00, 0x00,
-		0x00, 0x00, 0xaa,
-		0x00, 0xaa, 0x00,
-		0x00, 0xaa, 0xaa,
-		0xaa, 0x00, 0x00,
-		0xaa, 0x00, 0xaa,
-		0xaa, 0xaa, 0x00,
-		0xaa, 0xaa, 0xaa,
-		0x55, 0x55, 0x55,
-		0x55, 0x55, 0xff,
-		0x55, 0xff, 0x55,
-		0x55, 0xff, 0xff,
-		0xff, 0x55, 0x55,
-		0xff, 0x55, 0xff,
-		0xff, 0xff, 0x55,
-		0xff, 0xff, 0xff
-	};
-	const unsigned char *clut;
-
-	_prom->disp_node = 0;
-
-	prom_printf("Looking for displays\n");
-	if (RELOC(of_stdout_device) != 0)
-		prom_printf("OF stdout is    : %s\n",
-			    PTRRELOC(RELOC(of_stdout_device)));
-	for (node = 0; prom_next_node(&node); ) {
-		type[0] = 0;
-		prom_getprop(node, "device_type", type, sizeof(type));
-		if (strcmp(type, RELOC("display")) != 0)
-			continue;
-		/* It seems OF doesn't null-terminate the path :-( */
-		path = (char *) mem;
-		memset(path, 0, 256);
-
-		/*
-		 * leave some room at the end of the path for appending extra
-		 * arguments
-		 */
-		if (call_prom("package-to-path", 3, 1, node, path, 250) < 0)
-			continue;
-		prom_printf("found display   : %s\n", path);
-		
-		/*
-		 * If this display is the device that OF is using for stdout,
-		 * move it to the front of the list.
-		 */
-		mem += strlen(path) + 1;
-		i = RELOC(prom_num_displays);
-		RELOC(prom_num_displays) = i + 1;
-		if (RELOC(of_stdout_device) != 0 && i > 0
-		    && strcmp(PTRRELOC(RELOC(of_stdout_device)), path) == 0) {
-			for (; i > 0; --i) {
-				RELOC(prom_display_paths[i])
-					= RELOC(prom_display_paths[i-1]);
-				RELOC(prom_display_nodes[i])
-					= RELOC(prom_display_nodes[i-1]);
-			}
-			_prom->disp_node = node;
-		}
-		RELOC(prom_display_paths[i]) = PTRUNRELOC(path);
-		RELOC(prom_display_nodes[i]) = node;
-		if (_prom->disp_node == 0)
-			_prom->disp_node = node;
-		if (RELOC(prom_num_displays) >= FB_MAX)
-			break;
-	}
-	prom_printf("Opening displays...\n");
-	for (j = RELOC(prom_num_displays) - 1; j >= 0; j--) {
-		path = PTRRELOC(RELOC(prom_display_paths[j]));
-		prom_printf("opening display : %s", path);
-		ih = call_prom("open", 1, 1, path);
-		if (ih == (ihandle)0 || ih == (ihandle)-1) {
-			prom_printf("... failed\n");
-			continue;
-		}
-
-		prom_printf("... done\n");
-
-		/* Setup a useable color table when the appropriate
-		 * method is available. Should update this to set-colors */
-		clut = RELOC(default_colors);
-		for (i = 0; i < 32; i++, clut += 3)
-			if (prom_set_color(ih, i, clut[0], clut[1],
-					   clut[2]) != 0)
-				break;
-
-#ifdef CONFIG_LOGO_LINUX_CLUT224
-		clut = PTRRELOC(RELOC(logo_linux_clut224.clut));
-		for (i = 0; i < RELOC(logo_linux_clut224.clutsize); i++, clut += 3)
-			if (prom_set_color(ih, i + 32, clut[0], clut[1],
-					   clut[2]) != 0)
-				break;
-#endif /* CONFIG_LOGO_LINUX_CLUT224 */
-	}
-
-	return DOUBLEWORD_ALIGN(mem);
-}
-
-/* Return (relocated) pointer to this much memory: moves initrd if reqd. */
-static void __init *__make_room(unsigned long *mem_start, unsigned long *mem_end,
-				unsigned long needed, unsigned long align)
-{
-	void *ret;
-
-	*mem_start = ALIGN(*mem_start, align);
-	if (*mem_start + needed > *mem_end) {
-#ifdef CONFIG_BLK_DEV_INITRD
-		unsigned long offset = reloc_offset();
-		/* FIXME: Apple OF doesn't map unclaimed mem.  If this
-		 * ever happened on G5, we'd need to fix. */
-		unsigned long initrd_len;
-
-		if (*mem_end != RELOC(initrd_start))
-			prom_panic("No memory for copy_device_tree");
-
-		prom_printf("Huge device_tree: moving initrd\n");
-		/* Move by 4M. */
-		initrd_len = RELOC(initrd_end) - RELOC(initrd_start);
-		*mem_end = RELOC(initrd_start) + 4 * 1024 * 1024;
-		memmove((void *)*mem_end, (void *)RELOC(initrd_start),
-			initrd_len);
-		RELOC(initrd_start) = *mem_end;
-		RELOC(initrd_end) = RELOC(initrd_start) + initrd_len;
-#else
-		prom_panic("No memory for copy_device_tree");
-#endif
-	}
-
-	ret = (void *)*mem_start;
-	*mem_start += needed;
-
-	return ret;
-}
-
-#define make_room(startp, endp, type) \
-	__make_room(startp, endp, sizeof(type), __alignof__(type))
-
-static void __init
-inspect_node(phandle node, struct device_node *dad,
-	     unsigned long *mem_start, unsigned long *mem_end,
-	     struct device_node ***allnextpp)
-{
-	int l;
-	phandle child;
-	struct device_node *np;
-	struct property *pp, **prev_propp;
-	char *prev_name, *namep;
-	unsigned char *valp;
-	unsigned long offset = reloc_offset();
-	phandle ibm_phandle;
-
-	np = make_room(mem_start, mem_end, struct device_node);
-	memset(np, 0, sizeof(*np));
-
-	np->node = node;
-	**allnextpp = PTRUNRELOC(np);
-	*allnextpp = &np->allnext;
-	if (dad != 0) {
-		np->parent = PTRUNRELOC(dad);
-		/* we temporarily use the `next' field as `last_child'. */
-		if (dad->next == 0)
-			dad->child = PTRUNRELOC(np);
-		else
-			dad->next->sibling = PTRUNRELOC(np);
-		dad->next = np;
-	}
-
-	/* get and store all properties */
-	prev_propp = &np->properties;
-	prev_name = RELOC("");
-	for (;;) {
-		/* 32 is max len of name including nul. */
-		namep = make_room(mem_start, mem_end, char[32]);
-		if (call_prom("nextprop", 3, 1, node, prev_name, namep) <= 0) {
-			/* No more nodes: unwind alloc */
-			*mem_start = (unsigned long)namep;
-			break;
-		}
-		/* Trim off some if we can */
-		*mem_start = DOUBLEWORD_ALIGN((unsigned long)namep
-					     + strlen(namep) + 1);
-		pp = make_room(mem_start, mem_end, struct property);
-		pp->name = PTRUNRELOC(namep);
-		prev_name = namep;
-
-		pp->length = call_prom("getproplen", 2, 1, node, namep);
-		if (pp->length < 0)
-			continue;
-		if (pp->length > MAX_PROPERTY_LENGTH) {
-			char path[128];
-
-			prom_printf("WARNING: ignoring large property ");
-			/* It seems OF doesn't null-terminate the path :-( */
-			memset(path, 0, sizeof(path));
-			if (call_prom("package-to-path", 3, 1, node,
-				      path, sizeof(path)-1) > 0)
-				prom_printf("[%s] ", path);
-			prom_printf("%s length 0x%x\n", namep, pp->length);
-			continue;
-		}
-		valp = __make_room(mem_start, mem_end, pp->length, 1);
-		pp->value = PTRUNRELOC(valp);
-		call_prom("getprop", 4, 1, node, namep, valp, pp->length);
-		*prev_propp = PTRUNRELOC(pp);
-		prev_propp = &pp->next;
-	}
-
-	/* Add a "linux,phandle" property. */
-	namep = make_room(mem_start, mem_end, char[16]);
-	strcpy(namep, RELOC("linux,phandle"));
-	pp = make_room(mem_start, mem_end, struct property);
-	pp->name = PTRUNRELOC(namep);
-	pp->length = sizeof(phandle);
-	valp = make_room(mem_start, mem_end, phandle);
-	pp->value = PTRUNRELOC(valp);
-	*(phandle *)valp = node;
-	*prev_propp = PTRUNRELOC(pp);
-	pp->next = NULL;
-
-	/* Set np->linux_phandle to the value of the ibm,phandle property
-	   if it exists, otherwise to the phandle for this node. */
-	np->linux_phandle = node;
-	if (prom_getprop(node, "ibm,phandle",
-			 &ibm_phandle, sizeof(ibm_phandle)) > 0)
-		np->linux_phandle = ibm_phandle;
-
-	/* get the node's full name */
-	namep = (char *)*mem_start;
-	l = call_prom("package-to-path", 3, 1, node,
-		      namep, *mem_end - *mem_start);
-	if (l >= 0) {
-		/* Didn't fit?  Get more room. */
-		if (l+1 > *mem_end - *mem_start) {
-			namep = __make_room(mem_start, mem_end, l+1, 1);
-			call_prom("package-to-path", 3, 1, node, namep, l);
-		}
-		np->full_name = PTRUNRELOC(namep);
-		namep[l] = '\0';
-		*mem_start = DOUBLEWORD_ALIGN(*mem_start + l + 1);
-	}
-
-	/* do all our children */
-	child = call_prom("child", 1, 1, node);
-	while (child != (phandle)0) {
-		inspect_node(child, np, mem_start, mem_end,
-					 allnextpp);
-		child = call_prom("peer", 1, 1, child);
-	}
-}
-
-/*
- * Make a copy of the device tree from the PROM.
- */
-static unsigned long __init
-copy_device_tree(unsigned long mem_start)
-{
-	phandle root;
-	struct device_node **allnextp;
-	unsigned long offset = reloc_offset();
-	unsigned long mem_end;
-
-	/* We pass mem_end-mem_start to OF: keep it well under 32-bit */
-	mem_end = mem_start + 1024*1024*1024;
-#ifdef CONFIG_BLK_DEV_INITRD
-	if (RELOC(initrd_start) && RELOC(initrd_start) > mem_start)
-		mem_end = RELOC(initrd_start);
-#endif /* CONFIG_BLK_DEV_INITRD */
-
-	root = call_prom("peer", 1, 1, (phandle)0);
-	if (root == (phandle)0) {
-		prom_panic("couldn't get device tree root\n");
-	}
-	allnextp = &RELOC(allnodes);
-	inspect_node(root, NULL, &mem_start, &mem_end, &allnextp);
-	*allnextp = NULL;
-	return mem_start;
-}
-
-/* Verify bi_recs are good */
-static struct bi_record * __init prom_bi_rec_verify(struct bi_record *bi_recs)
-{
-	struct bi_record *first, *last;
-
-	prom_debug("birec_verify: r6=0x%x\n", (unsigned long)bi_recs);
-	if (bi_recs != NULL)
-		prom_debug("  tag=0x%x\n", bi_recs->tag);
+#ifdef DEBUG
+#define DBG(fmt...) udbg_printf(fmt)
+#else
+#define DBG(fmt...)
+#endif
 
-	if ( bi_recs == NULL || bi_recs->tag != BI_FIRST )
-		return NULL;
+struct pci_reg_property {
+	struct pci_address addr;
+	u32 size_hi;
+	u32 size_lo;
+};
 
-	last = (struct bi_record *)(long)bi_recs->data[0];
+struct isa_reg_property {
+	u32 space;
+	u32 address;
+	u32 size;
+};
 
-	prom_debug("  last=0x%x\n", (unsigned long)last);
-	if (last != NULL)
-		prom_debug("  last_tag=0x%x\n", last->tag);
 
-	if ( last == NULL || last->tag != BI_LAST )
-		return NULL;
+typedef unsigned long interpret_func(struct device_node *, unsigned long,
+				     int, int, int);
 
-	first = (struct bi_record *)(long)last->data[0];
-	prom_debug("  first=0x%x\n", (unsigned long)first);
+extern struct rtas_t rtas;
+extern struct lmb lmb;
+extern unsigned long klimit;
 
-	if ( first == NULL || first != bi_recs )
-		return NULL;
+static int __initdata dt_root_addr_cells;
+static int __initdata dt_root_size_cells;
+static int __initdata iommu_is_off;
+typedef u32 cell_t;
 
-	return bi_recs;
-}
+static struct boot_param_header *initial_boot_params __initdata;
 
-static void __init prom_bi_rec_reserve(void)
-{
-	unsigned long offset = reloc_offset();
-	struct prom_t *_prom = PTRRELOC(&prom);
-	struct bi_record *rec;
-
-	if ( _prom->bi_recs != NULL) {
-
-		for ( rec=_prom->bi_recs;
-		      rec->tag != BI_LAST;
-		      rec=bi_rec_next(rec) ) {
-			prom_debug("bi: 0x%x\n", rec->tag);
-			switch (rec->tag) {
-#ifdef CONFIG_BLK_DEV_INITRD
-			case BI_INITRD:
-				RELOC(initrd_start) = (unsigned long)(rec->data[0]);
-				RELOC(initrd_end) = RELOC(initrd_start) + rec->data[1];
-				break;
-#endif /* CONFIG_BLK_DEV_INITRD */
-			}
-		}
-		/* The next use of this field will be after relocation
-	 	 * is enabled, so convert this physical address into a
-	 	 * virtual address.
-	 	 */
-		_prom->bi_recs = PTRUNRELOC(_prom->bi_recs);
-	}
-}
+static struct device_node *allnodes = NULL;
 
-/*
- * We enter here early on, when the Open Firmware prom is still
- * handling exceptions and the MMU hash table for us.
+/* use when traversing tree through the allnext, child, sibling,
+ * or parent members of struct device_node.
  */
+static rwlock_t devtree_lock = RW_LOCK_UNLOCKED;
 
-unsigned long __init
-prom_init(unsigned long r3, unsigned long r4, unsigned long pp,
-	  unsigned long r6, unsigned long r7)
-{
-	unsigned long mem;
-	ihandle prom_cpu;
-	phandle cpu_pkg;
-	unsigned long offset = reloc_offset();
-	long l;
-	char *p, *d;
-	unsigned long phys;
-	u32 getprop_rval;
-	struct systemcfg *_systemcfg;
-	struct paca_struct *lpaca = PTRRELOC(&paca[0]);
-	struct prom_t *_prom = PTRRELOC(&prom);
-
-	/* First zero the BSS -- use memset, some arches don't have
-	 * caches on yet */
-	memset(PTRRELOC(&__bss_start), 0, __bss_stop - __bss_start);
-
-	/* Setup systemcfg and NACA pointers now */
-	RELOC(systemcfg) = _systemcfg = (struct systemcfg *)(SYSTEMCFG_VIRT_ADDR - offset);
-	RELOC(naca) = (struct naca_struct *)(NACA_VIRT_ADDR - offset);
-
-	/* Init interface to Open Firmware and pickup bi-recs */
-	prom_init_client_services(pp);
-
-	/* Init prom stdout device */
-	prom_init_stdout();
-
-	prom_debug("klimit=0x%x\n", RELOC(klimit));
-	prom_debug("offset=0x%x\n", offset);
-	prom_debug("->mem=0x%x\n", RELOC(klimit) - offset);
-
-	/* check out if we have bi_recs */
-	_prom->bi_recs = prom_bi_rec_verify((struct bi_record *)r6);
-	if ( _prom->bi_recs != NULL ) {
-		RELOC(klimit) = PTRUNRELOC((unsigned long)_prom->bi_recs +
-					   _prom->bi_recs->data[1]);
-		prom_debug("bi_recs=0x%x\n", (unsigned long)_prom->bi_recs);
-		prom_debug("new mem=0x%x\n", RELOC(klimit) - offset);
-	}
-
-	/* If we don't have birec's or didn't find them, check for an initrd
-	 * using the "yaboot" way
-	 */
-#ifdef CONFIG_BLK_DEV_INITRD
-	if ( _prom->bi_recs == NULL && r3 && r4 && r4 != 0xdeadbeef) {
-		RELOC(initrd_start) = (r3 >= KERNELBASE) ? __pa(r3) : r3;
-		RELOC(initrd_end) = RELOC(initrd_start) + r4;
-		RELOC(initrd_below_start_ok) = 1;
-	}
-#endif /* CONFIG_BLK_DEV_INITRD */
-
-	/* Default machine type. */
-	_systemcfg->platform = prom_find_machine_type();
-
-	/* On pSeries, copy the CPU hold code */
-	if (_systemcfg->platform == PLATFORM_PSERIES)
-		copy_and_flush(0, KERNELBASE - offset, 0x100, 0);
-
-	/* Start storing things at klimit */
-	mem = RELOC(klimit) - offset;
-
-	/* Get the full OF pathname of the stdout device */
-	p = (char *) mem;
-	memset(p, 0, 256);
-	call_prom("instance-to-path", 3, 1, _prom->stdout, p, 255);
-	RELOC(of_stdout_device) = PTRUNRELOC(p);
-	mem += strlen(p) + 1;
-
-	getprop_rval = 1;
-	prom_getprop(_prom->root, "#size-cells",
-		     &getprop_rval, sizeof(getprop_rval));
-	_prom->encode_phys_size = (getprop_rval == 1) ? 32 : 64;
-
-	/* Determine which cpu is actually running right _now_ */
-	if (prom_getprop(_prom->chosen, "cpu",
-			 &prom_cpu, sizeof(prom_cpu)) <= 0)
-		prom_panic("cannot find boot cpu");
-
-	cpu_pkg = call_prom("instance-to-package", 1, 1, prom_cpu);
-	prom_getprop(cpu_pkg, "reg", &getprop_rval, sizeof(getprop_rval));
-	_prom->cpu = getprop_rval;
-	lpaca[0].hw_cpu_id = _prom->cpu;
-
-	RELOC(boot_cpuid) = 0;
-
-	prom_debug("Booting CPU hw index = 0x%x\n", _prom->cpu);
-
-	/* Get the boot device and translate it to a full OF pathname. */
-	p = (char *) mem;
-	l = prom_getprop(_prom->chosen, "bootpath", p, 1<<20);
-	if (l > 0) {
-		p[l] = 0;	/* should already be null-terminated */
-		RELOC(bootpath) = PTRUNRELOC(p);
-		mem += l + 1;
-		d = (char *) mem;
-		*d = 0;
-		call_prom("canon", 3, 1, p, d, 1<<20);
-		RELOC(bootdevice) = PTRUNRELOC(d);
-		mem = DOUBLEWORD_ALIGN(mem + strlen(d) + 1);
-	}
-
-	RELOC(cmd_line[0]) = 0;
-#ifdef CONFIG_CMDLINE
-	strlcpy(RELOC(cmd_line), CONFIG_CMDLINE, sizeof(cmd_line));
-#endif /* CONFIG_CMDLINE */
-	if ((long)_prom->chosen > 0) {
-		prom_getprop(_prom->chosen, "bootargs", p, sizeof(cmd_line));
-		if (p != NULL && p[0] != 0)
-			strlcpy(RELOC(cmd_line), p, sizeof(cmd_line));
-	}
-
-	early_cmdline_parse();
-
-	prom_initialize_lmb();
-
-	prom_bi_rec_reserve();
-
-	mem = check_display(mem);
-
-	if (_systemcfg->platform != PLATFORM_POWERMAC)
-		prom_instantiate_rtas();
-
-	/* Initialize some system info into the Naca early... */
-	prom_initialize_naca();
-
-	/* If we are on an SMP machine, then we *MUST* do the
-	 * following, regardless of whether we have an SMP
-	 * kernel or not.
-	 */
-	prom_hold_cpus(mem);
-
-	prom_debug("after basic inits, mem=0x%x\n", mem);
-#ifdef CONFIG_BLK_DEV_INITRD
-	prom_debug("initrd_start=0x%x\n", RELOC(initrd_start));
-	prom_debug("initrd_end=0x%x\n", RELOC(initrd_end));
-#endif /* CONFIG_BLK_DEV_INITRD */
-	prom_debug("copying OF device tree...\n");
-
-	mem = copy_device_tree(mem);
-
-	RELOC(klimit) = mem + offset;
-
-	prom_debug("new klimit is\n");
-	prom_debug("klimit=0x%x\n", RELOC(klimit));
-	prom_debug(" ->mem=0x%x\n", mem);
-
-	lmb_reserve(0, __pa(RELOC(klimit)));
-
-#ifdef CONFIG_BLK_DEV_INITRD
-	if (RELOC(initrd_start)) {
-		unsigned long initrd_len;
-		initrd_len = RELOC(initrd_end) - RELOC(initrd_start);
-
-		/* Move initrd if it's where we're going to copy kernel. */
-		if (RELOC(initrd_start) < __pa(RELOC(klimit))) {
-			memmove((void *)mem, (void *)RELOC(initrd_start),
-				initrd_len);
-			RELOC(initrd_start) = mem;
-			RELOC(initrd_end) = mem + initrd_len;
-		}
-
-		lmb_reserve(RELOC(initrd_start), initrd_len);
-	}
-#endif /* CONFIG_BLK_DEV_INITRD */
-
-	if (_systemcfg->platform == PLATFORM_PSERIES)
-		prom_initialize_tce_table();
-
-#ifdef CONFIG_PMAC_DART
-	if (_systemcfg->platform == PLATFORM_POWERMAC)
-		prom_initialize_dart_table();
-#endif
-
-#ifdef CONFIG_BOOTX_TEXT
-	if (_prom->disp_node) {
-		prom_printf("Setting up bi display...\n");
-		setup_disp_fake_bi(_prom->disp_node);
-	}
-#endif /* CONFIG_BOOTX_TEXT */
-
-	prom_printf("Calling quiesce ...\n");
-	call_prom("quiesce", 0, 0);
-	phys = KERNELBASE - offset;
-
-#ifdef CONFIG_BLK_DEV_INITRD
-	/* If we had an initrd, we convert its address to virtual */
-	if (RELOC(initrd_start)) {
-		RELOC(initrd_start) = (unsigned long)__va(RELOC(initrd_start));
-		RELOC(initrd_end) = (unsigned long)__va(RELOC(initrd_end));
-	}
-#endif /* CONFIG_BLK_DEV_INITRD */
-
-	prom_printf("returning from prom_init\n");
-	return phys;
-}
+/* export that to outside world */
+struct device_node *of_chosen;
 
 /*
  * Find the device_node with a given phandle.
  */
-static struct device_node * __devinit
-find_phandle(phandle ph)
+static struct device_node * find_phandle(phandle ph)
 {
 	struct device_node *np;
 
@@ -1756,8 +113,7 @@
 /*
  * Find the interrupt parent of a node.
  */
-static struct device_node * __devinit
-intr_parent(struct device_node *p)
+static struct device_node * __devinit intr_parent(struct device_node *p)
 {
 	phandle *parp;
 
@@ -1798,14 +154,14 @@
  * Map an interrupt from a device up to the platform interrupt
  * descriptor.
  */
-static int __devinit
-map_interrupt(unsigned int **irq, struct device_node **ictrler,
-	      struct device_node *np, unsigned int *ints, int nintrc)
+static int __devinit map_interrupt(unsigned int **irq, struct device_node **ictrler,
+				   struct device_node *np, unsigned int *ints,
+				   int nintrc)
 {
 	struct device_node *p, *ipar;
 	unsigned int *imap, *imask, *ip;
 	int i, imaplen, match;
-	int newintrc, newaddrc;
+	int newintrc = 0, newaddrc = 0;
 	unsigned int *reg;
 	int naddrc;
 
@@ -1894,9 +250,9 @@
 	return nintrc;
 }
 
-static unsigned long __init
-finish_node_interrupts(struct device_node *np, unsigned long mem_start,
-		       int measure_only)
+static unsigned long __init finish_node_interrupts(struct device_node *np,
+						   unsigned long mem_start,
+						   int measure_only)
 {
 	unsigned int *ints;
 	int intlen, intrcells, intrcount;
@@ -1957,9 +313,10 @@
 	return mem_start;
 }
 
-static unsigned long __init
-interpret_pci_props(struct device_node *np, unsigned long mem_start,
-		    int naddrc, int nsizec, int measure_only)
+static unsigned long __init interpret_pci_props(struct device_node *np,
+						unsigned long mem_start,
+						int naddrc, int nsizec,
+						int measure_only)
 {
 	struct address_range *adr;
 	struct pci_reg_property *pci_addrs;
@@ -1985,9 +342,10 @@
 	return mem_start;
 }
 
-static unsigned long __init
-interpret_dbdma_props(struct device_node *np, unsigned long mem_start,
-		      int naddrc, int nsizec, int measure_only)
+static unsigned long __init interpret_dbdma_props(struct device_node *np,
+						  unsigned long mem_start,
+						  int naddrc, int nsizec,
+						  int measure_only)
 {
 	struct reg_property32 *rp;
 	struct address_range *adr;
@@ -2025,9 +383,10 @@
 	return mem_start;
 }
 
-static unsigned long __init
-interpret_macio_props(struct device_node *np, unsigned long mem_start,
-		      int naddrc, int nsizec, int measure_only)
+static unsigned long __init interpret_macio_props(struct device_node *np,
+						  unsigned long mem_start,
+						  int naddrc, int nsizec,
+						  int measure_only)
 {
 	struct reg_property32 *rp;
 	struct address_range *adr;
@@ -2065,9 +424,10 @@
 	return mem_start;
 }
 
-static unsigned long __init
-interpret_isa_props(struct device_node *np, unsigned long mem_start,
-		    int naddrc, int nsizec, int measure_only)
+static unsigned long __init interpret_isa_props(struct device_node *np,
+						unsigned long mem_start,
+						int naddrc, int nsizec,
+						int measure_only)
 {
 	struct isa_reg_property *rp;
 	struct address_range *adr;
@@ -2093,9 +453,10 @@
 	return mem_start;
 }
 
-static unsigned long __init
-interpret_root_props(struct device_node *np, unsigned long mem_start,
-		     int naddrc, int nsizec, int measure_only)
+static unsigned long __init interpret_root_props(struct device_node *np,
+						 unsigned long mem_start,
+						 int naddrc, int nsizec,
+						 int measure_only)
 {
 	struct address_range *adr;
 	int i, l;
@@ -2123,21 +484,15 @@
 	return mem_start;
 }
 
-static unsigned long __init
-finish_node(struct device_node *np, unsigned long mem_start,
-	    interpret_func *ifunc, int naddrc, int nsizec, int measure_only)
+static unsigned long __init finish_node(struct device_node *np,
+					unsigned long mem_start,
+					interpret_func *ifunc,
+					int naddrc, int nsizec,
+					int measure_only)
 {
 	struct device_node *child;
 	int *ip;
 
-	np->name = get_property(np, "name", NULL);
-	np->type = get_property(np, "device_type", NULL);
-
-	if (!np->name)
-		np->name = "<NULL>";
-	if (!np->type)
-		np->type = "<NULL>";
-
 	/* get the device addresses and interrupts */
 	if (ifunc != NULL)
 		mem_start = ifunc(np, mem_start, naddrc, nsizec, measure_only);
@@ -2188,24 +543,468 @@
 /**
  * finish_device_tree is called once things are running normally
  * (i.e. with text and data mapped to the address they were linked at).
- * It traverses the device tree and fills in the name, type,
- * {n_}addrs and {n_}intrs fields of each node.
+ * It traverses the device tree and fills in some of the additional,
+ * fields in each node like {n_}addrs and {n_}intrs, the virt interrupt
+ * mapping is also initialized at this point.
  */
-void __init
-finish_device_tree(void)
+void __init finish_device_tree(void)
 {
-	unsigned long mem = klimit;
+	unsigned long mem, size;
 
+	/* Initialize virtual IRQ map */
 	virt_irq_init();
 
-	dev_tree_size = finish_node(allnodes, 0, NULL, 0, 0, 1);
-	mem = (long)abs_to_virt(lmb_alloc(dev_tree_size,
-					  __alignof__(struct device_node)));
-	if (finish_node(allnodes, mem, NULL, 0, 0, 0) != mem + dev_tree_size)
+	/* Finish device-tree (pre-parsing some properties etc...) */
+	size = finish_node(allnodes, 0, NULL, 0, 0, 1);
+	mem = (unsigned long)abs_to_virt(lmb_alloc(size, 128));
+	if (finish_node(allnodes, mem, NULL, 0, 0, 0) != mem + size)
 		BUG();
-	rtas.dev = of_find_node_by_name(NULL, "rtas");
 }
 
+#ifdef DEBUG
+#define printk udbg_printf
+#endif
+
+static inline char *find_flat_dt_string(u32 offset)
+{
+	return ((char *)initial_boot_params) + initial_boot_params->off_dt_strings
+		+ offset;
+}
+
+/**
+ * This function is used to scan the flattened device-tree, it is
+ * used to extract the memory informations at boot before we can
+ * unflatten the tree
+ */
+static int __init scan_flat_dt(int (*it)(unsigned long node,
+					 const char *full_path, void *data),
+			       void *data)
+{
+	unsigned long p = ((unsigned long)initial_boot_params) +
+		initial_boot_params->off_dt_struct;
+	int rc = 0;
+
+	do {
+		u32 tag = *((u32 *)p);
+		char *pathp;
+		
+		p += 4;
+		if (tag == OF_DT_END_NODE)
+			continue;
+		if (tag == OF_DT_END)
+			break;
+		if (tag == OF_DT_PROP) {
+			u32 sz = *((u32 *)p);
+			p += 8;
+			p = _ALIGN(p, sz >= 8 ? 8 : 4);
+			p += sz;
+			p = _ALIGN(p, 4);
+			continue;
+		}
+		if (tag != OF_DT_BEGIN_NODE) {
+			printk(KERN_WARNING "Invalid tag %x scanning flattened"
+			       " device tree !\n", tag);
+			return -EINVAL;
+		}
+		pathp = (char *)p;
+		p = _ALIGN(p + strlen(pathp) + 1, 4);
+		rc = it(p, pathp, data);
+		if (rc != 0)
+			break;		
+	} while(1);
+
+	return rc;
+}
+
+/**
+ * This  function can be used within scan_flattened_dt callback to get
+ * access to properties
+ */
+static void* __init get_flat_dt_prop(unsigned long node, const char *name,
+				     unsigned long *size)
+{
+	unsigned long p = node;
+
+	do {
+		u32 tag = *((u32 *)p);
+		u32 sz, noff;
+		const char *nstr;
+
+		p += 4;
+		if (tag != OF_DT_PROP)
+			return NULL;
+
+		sz = *((u32 *)p);
+		noff = *((u32 *)(p + 4));
+		p += 8;
+		p = _ALIGN(p, sz >= 8 ? 8 : 4);
+
+		nstr = find_flat_dt_string(noff);
+		if (nstr == NULL) {
+			printk(KERN_WARNING "Can't find property index name !\n");
+			return NULL;
+		}
+		if (strcmp(name, nstr) == 0) {
+			if (size)
+				*size = sz;
+			return (void *)p;
+		}
+		p += sz;
+		p = _ALIGN(p, 4);
+	} while(1);
+}
+
+static void *__init unflatten_dt_alloc(unsigned long *mem, unsigned long size,
+					       unsigned long align)
+{
+	void *res;
+
+	*mem = _ALIGN(*mem, align);
+	res = (void *)*mem;
+	*mem += size;
+
+	return res;
+}
+
+static unsigned long __init unflatten_dt_node(unsigned long mem,
+					      unsigned long *p,
+					      struct device_node *dad,
+					      struct device_node ***allnextpp)
+{
+	struct device_node *np;
+	struct property *pp, **prev_pp = NULL;
+	char *pathp;
+	u32 tag;
+	unsigned int l;
+
+	tag = *((u32 *)(*p));
+	if (tag != OF_DT_BEGIN_NODE) {
+		printk("Weird tag at start of node: %x\n", tag);
+		return mem;
+	}
+	*p += 4;
+	pathp = (char *)*p;
+	l = strlen(pathp) + 1;
+	*p = _ALIGN(*p + l, 4);
+
+	np = unflatten_dt_alloc(&mem, sizeof(struct device_node) + l,
+				__alignof__(struct device_node));
+	if (allnextpp) {
+		memset(np, 0, sizeof(*np));
+		np->full_name = ((char*)np) + sizeof(struct device_node);
+		memcpy(np->full_name, pathp, l);
+		prev_pp = &np->properties;
+		**allnextpp = np;
+		*allnextpp = &np->allnext;
+		if (dad != NULL) {
+			np->parent = dad;
+			/* we temporarily use the `next' field as `last_child'. */
+			if (dad->next == 0)
+				dad->child = np;
+			else
+				dad->next->sibling = np;
+			dad->next = np;
+		}
+	}
+	while(1) {
+		u32 sz, noff;
+		char *pname;
+
+		tag = *((u32 *)(*p));
+		if (tag != OF_DT_PROP)
+			break;
+		*p += 4;
+		sz = *((u32 *)(*p));
+		noff = *((u32 *)((*p) + 4));
+		*p = _ALIGN((*p) + 8, sz >= 8 ? 8 : 4);
+
+		pname = find_flat_dt_string(noff);
+		if (pname == NULL) {
+			printk("Can't find property name in list !\n");
+			break;
+		}
+		l = strlen(pname) + 1;
+		pp = unflatten_dt_alloc(&mem, sizeof(struct property),
+					__alignof__(struct property));
+		if (allnextpp) {
+			if (strcmp(pname, "linux,phandle") == 0) {
+				np->node = *((u32 *)*p);
+				if (np->linux_phandle == 0)
+					np->linux_phandle = np->node;
+			}
+			if (strcmp(pname, "ibm,phandle") == 0)
+				np->linux_phandle = *((u32 *)*p);
+			pp->name = pname;
+			pp->length = sz;
+			pp->value = (void *)*p;
+			*prev_pp = pp;
+			prev_pp = &pp->next;
+		}
+		*p = _ALIGN((*p) + sz, 4);
+	}
+	if (allnextpp) {
+		*prev_pp = NULL;
+		np->name = get_property(np, "name", NULL);
+		np->type = get_property(np, "device_type", NULL);
+
+		if (!np->name)
+			np->name = "<NULL>";
+		if (!np->type)
+			np->type = "<NULL>";
+	}
+	while (tag == OF_DT_BEGIN_NODE) {
+		mem = unflatten_dt_node(mem, p, np, allnextpp);
+		tag = *((u32 *)(*p));
+	}
+	if (tag != OF_DT_END_NODE) {
+		printk("Weird tag at start of node: %x\n", tag);
+		return mem;
+	}
+	*p += 4;
+	return mem;
+}
+
+
+/**
+ * unflattens the device-tree passed by the firmware, creating the
+ * tree of struct device_node. It also fills the "name" and "type"
+ * pointers of the nodes so the normal device-tree walking functions
+ * can be used (this used to be done by finish_device_tree)
+ */
+void __init unflatten_device_tree(void)
+{
+	unsigned long start, mem, size;
+	struct device_node **allnextp = &allnodes;
+	char *p;
+	int l = 0;
+
+	/* First pass, scan for size */
+	start = ((unsigned long)initial_boot_params) +
+		initial_boot_params->off_dt_struct;
+	size = unflatten_dt_node(0, &start, NULL, NULL);
+
+	/* Allocate memory for the expanded device tree */
+	mem = (unsigned long)abs_to_virt(lmb_alloc(size,
+						   __alignof__(struct device_node)));
+
+	/* Second pass, do actual unflattening */
+	start = ((unsigned long)initial_boot_params) +
+		initial_boot_params->off_dt_struct;
+	unflatten_dt_node(mem, &start, NULL, &allnextp);
+	if (*((u32 *)start) != OF_DT_END)
+		printk(KERN_WARNING "Weird tag at end of tree: %x\n", *((u32 *)start));
+	*allnextp = NULL;
+
+	/* Get pointer to OF "/chosen" node for use everywhere */
+	of_chosen = of_find_node_by_path("/chosen");
+
+	/* Retreive command line */
+	if (of_chosen != NULL) {
+		p = (char *)get_property(of_chosen, "bootargs", &l);
+		if (p != NULL && l > 0)
+			strlcpy(cmd_line, p, min(l, COMMAND_LINE_SIZE));
+	}
+#ifdef CONFIG_CMDLINE
+	if (l == 0) /* dbl check */
+		strlcpy(cmd_line, CONFIG_CMDLINE, COMMAND_LINE_SIZE);
+#endif /* CONFIG_CMDLINE */
+}
+
+
+static int __init early_init_dt_scan_cpus(unsigned long node,
+					  const char *full_path, void *data)
+{
+	char *type = get_flat_dt_prop(node, "device_type", NULL);
+
+	/* We are scanning "cpu" nodes only */
+	if (type == NULL || strcmp(type, "cpu") != 0)
+		return 0;
+
+	/* On LPAR, look for the first ibm,pft-size property for the  hash table size
+	 */
+	if (systemcfg->platform == PLATFORM_PSERIES_LPAR && naca->pftSize == 0) {
+		u32 *pft_size;
+		pft_size = (u32 *)get_flat_dt_prop(node, "ibm,pft-size", NULL);
+		if (pft_size != NULL) {
+			/* pft_size[0] is the NUMA CEC cookie */
+			naca->pftSize = pft_size[1];
+		}
+	}
+
+	/* Check if it's the boot-cpu, set it's hw index in paca now */
+	if (get_flat_dt_prop(node, "linux,boot-cpu", NULL) != NULL) {
+		u32 *prop = get_flat_dt_prop(node, "reg", NULL);
+		paca[0].hw_cpu_id = *prop;
+	}
+
+	return 0;
+}
+
+static int __init early_init_dt_scan_chosen(unsigned long node,
+					    const char *full_path, void *data)
+{
+	u32 *prop;
+
+	if (strcmp(full_path, "/chosen") != 0)
+		return 0;
+
+	/* get platform type */
+	prop = (u32 *)get_flat_dt_prop(node, "linux,platform", NULL);
+	if (prop == NULL)
+		return 0;
+	systemcfg->platform = *prop;
+
+	/* check if iommu is forced off */
+	if (get_flat_dt_prop(node, "linux,iommu-off", NULL) != NULL)
+		iommu_is_off = 1;
+
+	/* break now */
+	return 1;
+}
+
+static int __init early_init_dt_scan_root(unsigned long node,
+					  const char *full_path, void *data)
+{
+	u32 *prop;
+
+	if (strcmp(full_path, "/") != 0)
+		return 0;
+
+	prop = (u32 *)get_flat_dt_prop(node, "#size-cells", NULL);
+	dt_root_size_cells = (prop == NULL) ? 1 : *prop;
+		
+	prop = (u32 *)get_flat_dt_prop(node, "#address-cells", NULL);
+	dt_root_addr_cells = (prop == NULL) ? 2 : *prop;
+	
+	/* break now */
+	return 1;
+}
+
+static unsigned long __init dt_mem_next_cell(int s, cell_t **cellp)
+{
+	cell_t *p = *cellp;
+	unsigned long r = 0;
+
+	/* Ignore more than 2 cells */
+	while (s > 2) {
+		p++;
+		s--;
+	}
+	while (s) {
+		r <<= 32;
+		r |= *(p++);
+		s--;
+	}
+
+	*cellp = p;
+	return r;
+}
+
+
+static int __init early_init_dt_scan_memory(unsigned long node,
+					    const char *full_path, void *data)
+{
+	char *type = get_flat_dt_prop(node, "device_type", NULL);
+	cell_t *reg, *endp;
+	unsigned long l;
+
+	/* We are scanning "memory" nodes only */
+	if (type == NULL || strcmp(type, "memory") != 0)
+		return 0;
+
+	reg = (cell_t *)get_flat_dt_prop(node, "reg", &l);
+	if (reg == NULL)
+		return 0;
+
+	endp = reg + (l / sizeof(cell_t));
+
+	DBG("memory scan node %s ...\n", full_path);
+	while ((endp - reg) >= (dt_root_addr_cells + dt_root_size_cells)) {
+		unsigned long base, size;
+
+		base = dt_mem_next_cell(dt_root_addr_cells, &reg);
+		size = dt_mem_next_cell(dt_root_size_cells, &reg);
+
+		if (size == 0)
+			continue;
+		DBG(" - %lx ,  %lx\n", base, size);
+		if (iommu_is_off) {
+			if (base >= 0x80000000ul)
+				continue;
+			if ((base + size) > 0x80000000ul)
+				size = 0x80000000ul - base;
+		}
+		lmb_add(base, size);
+	}
+	return 0;
+}
+
+static void __init early_reserve_mem(void)
+{
+	u64 base, size;
+	u64 *reserve_map = (u64 *)(((unsigned long)initial_boot_params) +
+				   initial_boot_params->off_mem_rsvmap);
+	while (1) {
+		base = *(reserve_map++);
+		size = *(reserve_map++);
+		if (size == 0)
+			break;
+		lmb_reserve(base, size);
+	}
+
+	DBG("emory reserved, lmbs :\n");
+	lmb_dump_all();
+}
+
+void __init early_init_devtree(void *params)
+{
+	/* Setup flat device-tree pointer */
+	initial_boot_params = params;
+
+	/* By default, hash size is not set */
+	naca->pftSize = 0;
+
+	/* Retreive various informations from the /chosen node of the
+	 * device-tree, including the platform type, initrd location and
+	 * size, TCE reserve, and more ...
+	 */
+	scan_flat_dt(early_init_dt_scan_chosen, NULL);
+
+	/* Scan memory nodes and rebuild LMBs */
+	lmb_init();
+	scan_flat_dt(early_init_dt_scan_root, NULL);
+	scan_flat_dt(early_init_dt_scan_memory, NULL);
+	lmb_analyze();
+	systemcfg->physicalMemorySize = lmb_phys_mem_size();
+
+	DBG("Physical mem : %lx\n", systemcfg->physicalMemorySize);
+
+	/* Reserve LMB regions used by kernel, initrd, dt, etc... */
+	early_reserve_mem();
+
+	/* Retreive hash table size from flattened tree */
+	scan_flat_dt(early_init_dt_scan_cpus, NULL);
+
+	/* If hash size wasn't obtained above, we calculate it now based on
+	 * the total RAM size
+	 */
+	if (naca->pftSize == 0) {
+		unsigned long rnd_mem_size, pteg_count;
+
+		/* round mem_size up to next power of 2 */
+		rnd_mem_size = 1UL << __ilog2(systemcfg->physicalMemorySize);
+		if (rnd_mem_size < systemcfg->physicalMemorySize)
+			rnd_mem_size <<= 1;
+
+		/* # pages / 2 */
+		pteg_count = (rnd_mem_size >> (12 + 1));
+
+		naca->pftSize = __ilog2(pteg_count << 7);
+	}
+}
+
+#undef printk
+
 int
 prom_n_addr_cells(struct device_node* np)
 {
@@ -2518,6 +1317,28 @@
 EXPORT_SYMBOL(of_find_node_by_path);
 
 /**
+ *	of_find_node_by_phandle - Find a node given a phandle
+ *	@handle:	phandle of the node to find
+ *
+ *	Returns a node pointer with refcount incremented, use
+ *	of_node_put() on it when done.
+ */
+struct device_node *of_find_node_by_phandle(phandle handle)
+{
+	struct device_node *np;
+
+	read_lock(&devtree_lock);
+	for (np = allnodes; np != 0; np = np->allnext)
+		if (np->linux_phandle == handle)
+			break;
+	if (np)
+		of_node_get(np);
+	read_unlock(&devtree_lock);
+	return np;
+}
+EXPORT_SYMBOL(of_find_node_by_phandle);
+
+/**
  *	of_find_all_nodes - Get next node in global list
  *	@prev:	Previous node or NULL to start iteration
  *		of_node_put() will be called on it
@@ -3053,3 +1874,13 @@
 	}
 }
 #endif
+
+
+
+
+
+
+
+
+
+
diff -urN linux-2.5/arch/ppc64/kernel/prom_init.c linux-prom-clean/arch/ppc64/kernel/prom_init.c
--- /dev/null	2004-09-16 11:45:22.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/prom_init.c	2004-09-17 22:19:57.000000000 +1000
@@ -0,0 +1,1794 @@
+/*
+ * 
+ *
+ * Procedures for interfacing to Open Firmware.
+ *
+ * Paul Mackerras	August 1996.
+ * Copyright (C) 1996 Paul Mackerras.
+ * 
+ *  Adapted for 64bit PowerPC by Dave Engebretsen and Peter Bergner.
+ *    {engebret|bergner}@us.ibm.com 
+ *
+ *      This program is free software; you can redistribute it and/or
+ *      modify it under the terms of the GNU General Public License
+ *      as published by the Free Software Foundation; either version
+ *      2 of the License, or (at your option) any later version.
+ */
+
+#define DEBUG_PROM
+
+#include <stdarg.h>
+#include <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/init.h>
+#include <linux/version.h>
+#include <linux/threads.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+#include <linux/proc_fs.h>
+#include <linux/stringify.h>
+#include <linux/delay.h>
+#include <linux/initrd.h>
+#include <asm/prom.h>
+#include <asm/rtas.h>
+#include <asm/abs_addr.h>
+#include <asm/page.h>
+#include <asm/processor.h>
+#include <asm/irq.h>
+#include <asm/io.h>
+#include <asm/smp.h>
+#include <asm/system.h>
+#include <asm/mmu.h>
+#include <asm/pgtable.h>
+#include <asm/bitops.h>
+#include <asm/naca.h>
+#include <asm/pci.h>
+#include <asm/iommu.h>
+#include <asm/bootinfo.h>
+#include <asm/ppcdebug.h>
+#include <asm/btext.h>
+#include <asm/sections.h>
+#include <asm/machdep.h>
+#include "open_pic.h"
+
+#ifdef CONFIG_LOGO_LINUX_CLUT224
+#include <linux/linux_logo.h>
+extern const struct linux_logo logo_linux_clut224;
+#endif
+
+/*
+ * Properties whose value is longer than this get excluded from our
+ * copy of the device tree. This value does need to be big enough to
+ * ensure that we don't lose things like the interrupt-map property
+ * on a PCI-PCI bridge.
+ */
+#define MAX_PROPERTY_LENGTH	(1UL * 1024 * 1024)
+
+/*
+ * Eventually bump that one up
+ */
+#define DEVTREE_CHUNK_SIZE	0x100000
+
+/*
+ * This is the size of the local memory reserve map that gets copied
+ * into the boot params passed to the kernel. That size is totally
+ * flexible as the kernel just reads the list until it encounters an
+ * entry with size 0, so it can be changed without breaking binary
+ * compatibility
+ */
+#define MEM_RESERVE_MAP_SIZE	8
+
+/*
+ * prom_init() is called very early on, before the kernel text
+ * and data have been mapped to KERNELBASE.  At this point the code
+ * is running at whatever address it has been loaded at, so
+ * references to extern and static variables must be relocated
+ * explicitly.  The procedure reloc_offset() returns the address
+ * we're currently running at minus the address we were linked at.
+ * (Note that strings count as static variables.)
+ *
+ * Because OF may have mapped I/O devices into the area starting at
+ * KERNELBASE, particularly on CHRP machines, we can't safely call
+ * OF once the kernel has been mapped to KERNELBASE.  Therefore all
+ * OF calls should be done within prom_init(), and prom_init()
+ * and all routines called within it must be careful to relocate
+ * references as necessary.
+ *
+ * Note that the bss is cleared *after* prom_init runs, so we have
+ * to make sure that any static or extern variables it accesses
+ * are put in the data segment.
+ */
+
+
+#define PROM_BUG() do {						\
+        prom_printf("kernel BUG at %s line 0x%x!\n",		\
+		    RELOC(__FILE__), __LINE__);			\
+        __asm__ __volatile__(".long " BUG_ILLEGAL_INSTR);	\
+} while (0)
+
+#ifdef DEBUG_PROM
+#define prom_debug(x...)	prom_printf(x)
+#else
+#define prom_debug(x...)
+#endif
+
+
+typedef u32 prom_arg_t;
+
+struct prom_args {
+        u32 service;
+        u32 nargs;
+        u32 nret;
+        prom_arg_t args[10];
+        prom_arg_t *rets;     /* Pointer to return values in args[16]. */
+};
+
+struct prom_t {
+	unsigned long entry;
+	ihandle root;
+	ihandle chosen;
+	int cpu;
+	ihandle stdout;
+	ihandle disp_node;
+	struct prom_args args;
+	unsigned long version;
+	unsigned long root_size_cells;
+	unsigned long root_addr_cells;
+};
+
+struct pci_reg_property {
+	struct pci_address addr;
+	u32 size_hi;
+	u32 size_lo;
+};
+
+struct mem_map_entry {
+	u64	base;
+	u64	size;
+};
+
+typedef u32 cell_t;
+
+extern void __start(unsigned long r3, unsigned long r4, unsigned long r5,
+		    unsigned long r6);
+
+extern unsigned long reloc_offset(void);
+extern void enter_prom(struct prom_args *args, unsigned long entry);
+extern void copy_and_flush(unsigned long dest, unsigned long src,
+			   unsigned long size, unsigned long offset);
+
+extern unsigned long klimit;
+
+//int global_width = 640, global_height = 480, global_depth = 8, global_pitch;
+//unsigned global_address;
+/* prom structure */
+static struct prom_t __initdata prom;
+
+#define PROM_SCRATCH_SIZE 256
+
+static char __initdata of_stdout_device[256];
+static char __initdata prom_scratch[PROM_SCRATCH_SIZE];
+
+static unsigned long __initdata dt_header_start;
+static unsigned long __initdata dt_struct_start, dt_struct_end;
+static unsigned long __initdata dt_string_start, dt_string_end;
+
+static unsigned long __initdata prom_initrd_start, prom_initrd_end;
+
+static int __initdata iommu_force_on;
+static int __initdata ppc64_iommu_off;
+static int __initdata of_platform;
+
+static char __initdata prom_cmd_line[COMMAND_LINE_SIZE];
+
+static unsigned long __initdata alloc_top;
+static unsigned long __initdata alloc_top_high;
+static unsigned long __initdata alloc_bottom;
+static unsigned long __initdata rmo_top;
+static unsigned long __initdata ram_top;
+
+static struct mem_map_entry __initdata mem_reserve_map[MEM_RESERVE_MAP_SIZE];
+static int __initdata mem_reserve_cnt;
+
+static cell_t __initdata regbuf[1024];
+
+
+#define MAX_CPU_THREADS 2
+
+/* TO GO */
+#ifdef CONFIG_HMT
+struct {
+	unsigned int pir;
+	unsigned int threadid;
+} hmt_thread_data[NR_CPUS];
+#endif /* CONFIG_HMT */
+
+/*
+ * This are used in calls to call_prom.  The 4th and following
+ * arguments to call_prom should be 32-bit values.  64 bit values
+ * are truncated to 32 bits (and fortunately don't get interpreted
+ * as two arguments).
+ */
+#define ADDR(x)		(u32) ((unsigned long)(x) - offset)
+
+/* This is the one and *ONLY* place where we actually call open
+ * firmware from, since we need to make sure we're running in 32b
+ * mode when we do.  We switch back to 64b mode upon return.
+ */
+
+#define PROM_ERROR	(-1)
+
+static int __init call_prom(const char *service, int nargs, int nret, ...)
+{
+	int i;
+	unsigned long offset = reloc_offset();
+	struct prom_t *_prom = PTRRELOC(&prom);
+	va_list list;
+
+	_prom->args.service = ADDR(service);
+	_prom->args.nargs = nargs;
+	_prom->args.nret = nret;
+	_prom->args.rets = (prom_arg_t *)&(_prom->args.args[nargs]);
+
+	va_start(list, nret);
+	for (i=0; i < nargs; i++)
+		_prom->args.args[i] = va_arg(list, prom_arg_t);
+	va_end(list);
+
+	for (i=0; i < nret ;i++)
+		_prom->args.rets[i] = 0;
+
+	enter_prom(&_prom->args, _prom->entry);
+
+	return (nret > 0) ? _prom->args.rets[0] : 0;
+}
+
+
+static unsigned int __init prom_claim(unsigned long virt, unsigned long size,
+				unsigned long align)
+{
+	return (unsigned int)call_prom("claim", 3, 1,
+				       (prom_arg_t)virt, (prom_arg_t)size,
+				       (prom_arg_t)align);
+}
+
+static void __init prom_print(const char *msg)
+{
+	const char *p, *q;
+	unsigned long offset = reloc_offset();
+	struct prom_t *_prom = PTRRELOC(&prom);
+
+	if (_prom->stdout == 0)
+		return;
+
+	for (p = msg; *p != 0; p = q) {
+		for (q = p; *q != 0 && *q != '\n'; ++q)
+			;
+		if (q > p)
+			call_prom("write", 3, 1, _prom->stdout, p, q - p);
+		if (*q == 0)
+			break;
+		++q;
+		call_prom("write", 3, 1, _prom->stdout, ADDR("\r\n"), 2);
+	}
+}
+
+
+static void __init prom_print_hex(unsigned long val)
+{
+	unsigned long offset = reloc_offset();
+	int i, nibbles = sizeof(val)*2;
+	char buf[sizeof(val)*2+1];
+	struct prom_t *_prom = PTRRELOC(&prom);
+
+	for (i = nibbles-1;  i >= 0;  i--) {
+		buf[i] = (val & 0xf) + '0';
+		if (buf[i] > '9')
+			buf[i] += ('a'-'0'-10);
+		val >>= 4;
+	}
+	buf[nibbles] = '\0';
+	call_prom("write", 3, 1, _prom->stdout, buf, nibbles);
+}
+
+
+static void __init prom_printf(const char *format, ...)
+{
+	unsigned long offset = reloc_offset();
+	const char *p, *q, *s;
+	va_list args;
+	unsigned long v;
+	struct prom_t *_prom = PTRRELOC(&prom);
+
+	va_start(args, format);
+	for (p = PTRRELOC(format); *p != 0; p = q) {
+		for (q = p; *q != 0 && *q != '\n' && *q != '%'; ++q)
+			;
+		if (q > p)
+			call_prom("write", 3, 1, _prom->stdout, p, q - p);
+		if (*q == 0)
+			break;
+		if (*q == '\n') {
+			++q;
+			call_prom("write", 3, 1, _prom->stdout,
+				  ADDR("\r\n"), 2);
+			continue;
+		}
+		++q;
+		if (*q == 0)
+			break;
+		switch (*q) {
+		case 's':
+			++q;
+			s = va_arg(args, const char *);
+			prom_print(s);
+			break;
+		case 'x':
+			++q;
+			v = va_arg(args, unsigned long);
+			prom_print_hex(v);
+			break;
+		}
+	}
+}
+
+
+static void __init __attribute__((noreturn)) prom_panic(const char *reason)
+{
+	unsigned long offset = reloc_offset();
+
+	prom_print(PTRRELOC(reason));
+	/* ToDo: should put up an SRC here */
+	call_prom("exit", 0, 0);
+
+	for (;;)			/* should never get here */
+		;
+}
+
+
+static int __init prom_next_node(phandle *nodep)
+{
+	phandle node;
+
+	if ((node = *nodep) != 0
+	    && (*nodep = call_prom("child", 1, 1, node)) != 0)
+		return 1;
+	if ((*nodep = call_prom("peer", 1, 1, node)) != 0)
+		return 1;
+	for (;;) {
+		if ((node = call_prom("parent", 1, 1, node)) == 0)
+			return 0;
+		if ((*nodep = call_prom("peer", 1, 1, node)) != 0)
+			return 1;
+	}
+}
+
+static int __init prom_getprop(phandle node, const char *pname,
+			       void *value, size_t valuelen)
+{
+	unsigned long offset = reloc_offset();
+
+	return call_prom("getprop", 4, 1, node, ADDR(pname),
+			 (u32)(unsigned long) value, (u32) valuelen);
+}
+
+static int __init prom_getproplen(phandle node, const char *pname)
+{
+	unsigned long offset = reloc_offset();
+
+	return call_prom("getproplen", 2, 1, node, ADDR(pname));
+}
+
+static int __init prom_setprop(phandle node, const char *pname,
+			       void *value, size_t valuelen)
+{
+	unsigned long offset = reloc_offset();
+
+	return call_prom("setprop", 4, 1, node, ADDR(pname),
+			 (u32)(unsigned long) value, (u32) valuelen);
+}
+
+
+/*
+ * Early parsing of the command line passed to the kernel, used for
+ * the options that affect the iommu
+ */
+static void __init early_cmdline_parse(void)
+{
+	unsigned long offset = reloc_offset();
+	struct prom_t *_prom = PTRRELOC(&prom);
+	char *opt, *p;
+	int l = 0;
+
+	RELOC(prom_cmd_line[0]) = 0;
+	p = RELOC(prom_cmd_line);
+	if ((long)_prom->chosen > 0)
+		l = prom_getprop(_prom->chosen, "bootargs", p, COMMAND_LINE_SIZE-1);
+#ifdef CONFIG_CMDLINE
+	if (l == 0) /* dbl check */
+		strlcpy(RELOC(prom_cmd_line), CONFIG_CMDLINE, sizeof(prom_cmd_line));
+#endif /* CONFIG_CMDLINE */
+	prom_printf("command line: %s\n", RELOC(prom_cmd_line));
+
+	opt = strstr(RELOC(prom_cmd_line), RELOC("iommu="));
+	if (opt) {
+		prom_printf("iommu opt is: %s\n", opt);
+		opt += 6;
+		while (*opt && *opt == ' ')
+			opt++;
+		if (!strncmp(opt, RELOC("off"), 3))
+			RELOC(ppc64_iommu_off) = 1;
+		else if (!strncmp(opt, RELOC("force"), 5))
+			RELOC(iommu_force_on) = 1;
+	}
+
+#ifndef CONFIG_PMAC_DART
+	if (RELOC(of_platform) == PLATFORM_POWERMAC) {
+		RELOC(ppc64_iommu_off) = 1;
+		prom_printf("DART disabled on PowerMac !\n");
+	}
+#endif
+}
+
+/*
+ * Memory allocation strategy... our layout is normally:
+ *
+ *  at 14Mb or more we vmlinux, then a gap and initrd. In some rare cases, initrd
+ *  might end up beeing before the kernel though. We assume this won't override
+ *  the final kernel at 0, we have no provision to handle that in this version,
+ *  but it should hopefully never happen.
+ *
+ *  alloc_top is set to the top of RMO, eventually shrink down if the TCEs overlap
+ *  alloc_bottom is set to the top of kernel/initrd
+ *
+ *  from there, allocations are done that way : rtas is allocated topmost, and
+ *  the device-tree is allocated from the bottom. We try to grow the device-tree
+ *  allocation as we progress. If we can't, then we fail, we don't currently have
+ *  a facility to restart elsewhere, but that shouldn't be necessary neither
+ *
+ *  Note that calls to reserve_mem have to be done explicitely, memory allocated
+ *  with either alloc_up or alloc_down isn't automatically reserved.
+ */
+
+
+/*
+ * Allocates memory in the RMO upward from the kernel/initrd
+ *
+ * When align is 0, this is a special case, it means to allocate in place
+ * at the current location of alloc_bottom or fail (that is basically
+ * extending the previous allocation). Used for the device-tree flattening
+ */
+static unsigned long __init alloc_up(unsigned long size, unsigned long align)
+{
+	unsigned long offset = reloc_offset();
+	unsigned long base = _ALIGN_UP(RELOC(alloc_bottom), align);
+	unsigned long addr = 0;
+
+	prom_debug("alloc_up(%x, %x)\n", size, align);
+	if (RELOC(ram_top) == 0)
+		prom_panic("alloc_up() called with mem not initialized\n");
+
+	if (align)
+		base = _ALIGN_UP(RELOC(alloc_bottom), align);
+	else
+		base = RELOC(alloc_bottom);
+
+	for(; (base + size) <= RELOC(alloc_top); 
+	    base = _ALIGN_UP(base + 0x100000, align)) {
+		prom_debug("    trying: 0x%x\n\r", base);
+		addr = (unsigned long)prom_claim(base, size, 0);
+		if ((int)addr != PROM_ERROR)
+			break;
+		addr = 0;
+		if (align == 0)
+			break;
+	}
+	if (addr == 0)
+		return 0;
+	RELOC(alloc_bottom) = addr;
+
+	prom_printf(" -> %x\n", addr);
+	prom_printf("  alloc_bottom : %x\n", RELOC(alloc_bottom));
+	prom_printf("  alloc_top    : %x\n", RELOC(alloc_top));
+	prom_printf("  alloc_top_hi : %x\n", RELOC(alloc_top_high));
+	prom_printf("  rmo_top      : %x\n", RELOC(rmo_top));
+	prom_printf("  ram_top      : %x\n", RELOC(ram_top));
+
+	return addr;
+}
+
+/*
+ * Allocates memory downard, either from top of RMO, or if highmem
+ * is set, from the top of RAM. Note that this one doesn't handle
+ * failures. In does claim memory if highmem is not set.
+ */
+static unsigned long __init alloc_down(unsigned long size, unsigned long align,
+				       int highmem)
+{
+	unsigned long offset = reloc_offset();
+	unsigned long base, addr = 0;
+
+	prom_debug("alloc_down(%x, %x, %s)\n", size, align,
+		   highmem ? RELOC("(high)") : RELOC("(low)"));
+	if (RELOC(ram_top) == 0)
+		prom_panic("alloc_down() called with mem not initialized\n");
+
+	if (highmem) {
+		/* Carve out storage for the TCE table. */
+		addr = _ALIGN_DOWN(RELOC(alloc_top_high) - size, align);
+		if (addr <= RELOC(alloc_bottom))
+			return 0;
+		else {
+			/* Will we bump into the RMO ? If yes, check out that we
+			 * didn't overlap existing allocations there, if we did,
+			 * we are dead, we must be the first in town !
+			 */
+			if (addr < RELOC(rmo_top)) {
+				/* Good, we are first */
+				if (RELOC(alloc_top) == RELOC(rmo_top))
+					RELOC(alloc_top) = addr;
+				else
+					return 0;
+			}
+			RELOC(alloc_top_high) = addr;
+		}
+		goto bail;
+	}
+
+	base = _ALIGN_DOWN(RELOC(alloc_top) - size, align);
+	for(; base > RELOC(alloc_bottom); base = _ALIGN_DOWN(base - 0x100000, align))  {
+		prom_debug("    trying: 0x%x\n\r", base);
+		addr = (unsigned long)prom_claim(base, size, 0);
+		if ((int)addr != PROM_ERROR)
+			break;
+		addr = 0;
+	}
+	if (addr == 0)
+		return 0;
+	RELOC(alloc_top) = addr;
+
+ bail:
+	prom_printf(" -> %x\n", addr);
+	prom_printf("  alloc_bottom : %x\n", RELOC(alloc_bottom));
+	prom_printf("  alloc_top    : %x\n", RELOC(alloc_top));
+	prom_printf("  alloc_top_hi : %x\n", RELOC(alloc_top_high));
+	prom_printf("  rmo_top      : %x\n", RELOC(rmo_top));
+	prom_printf("  ram_top      : %x\n", RELOC(ram_top));
+
+	return addr;
+}
+
+/*
+ * Parse a "reg" cell
+ */
+static unsigned long __init prom_next_cell(int s, cell_t **cellp)
+{
+	cell_t *p = *cellp;
+	unsigned long r = 0;
+
+	/* Ignore more than 2 cells */
+	while (s > 2) {
+		p++;
+		s--;
+	}
+	while (s) {
+		r <<= 32;
+		r |= *(p++);
+		s--;
+	}
+
+	*cellp = p;
+	return r;
+}
+
+/*
+ * Very dumb function for adding to the memory reserve list, but
+ * we don't need anything smarter at this point
+ *
+ * XXX Eventually check for collisions. They should NEVER happen
+ * if problems seem to show up, it would be a good start to track
+ * them down.
+ */
+static void reserve_mem(unsigned long base, unsigned long size)
+{
+	unsigned long offset = reloc_offset();
+	unsigned long top = base + size;
+	unsigned long cnt = RELOC(mem_reserve_cnt);
+
+	/* We need to always keep one empty entry so that we
+	 * have our terminator with "size" set to 0 since we are
+	 * dumb and just copy this entire array to the boot params
+	 */
+	base = _ALIGN_DOWN(base, PAGE_SIZE);
+	top = _ALIGN_DOWN(top, PAGE_SIZE);
+	size = top - base;
+
+	if (cnt >= (MEM_RESERVE_MAP_SIZE - 1))
+		prom_panic("Memory reserve map exhausted !\n");
+	RELOC(mem_reserve_map)[cnt].base = base;
+	RELOC(mem_reserve_map)[cnt].size = size;
+	RELOC(mem_reserve_cnt) = cnt + 1;
+}
+
+/*
+ * Initialize memory allocation mecanism, parse "memory" nodes and
+ * obtain that way the top of memory and RMO to setup out local allocator
+ */
+static void __init prom_init_mem(void)
+{
+	phandle node;
+	char *path, type[64];
+	unsigned int plen;
+	cell_t *p, *endp;
+	unsigned long offset = reloc_offset();
+	struct prom_t *_prom = PTRRELOC(&prom);
+
+	/*
+	 * We iterate the memory nodes to find
+	 * 1) top of RMO (first node)
+	 * 2) top of memory
+	 */
+	prom_debug("root_addr_cells: %x\n", (long)_prom->root_addr_cells);
+	prom_debug("root_size_cells: %x\n", (long)_prom->root_size_cells);
+
+	prom_debug("scanning memory:\n");
+	path = RELOC(prom_scratch);
+
+	for (node = 0; prom_next_node(&node); ) {
+		type[0] = 0;
+		prom_getprop(node, "device_type", type, sizeof(type));
+
+		if (strcmp(type, RELOC("memory")))
+			continue;
+	
+		plen = prom_getprop(node, "reg", RELOC(regbuf), sizeof(regbuf));
+		if (plen > sizeof(regbuf)) {
+			prom_printf("memory node too large for buffer !\n");
+			plen = sizeof(regbuf);
+		}
+		p = RELOC(regbuf);
+		endp = p + (plen / sizeof(cell_t));
+
+#ifdef DEBUG_PROM
+		memset(path, 0, PROM_SCRATCH_SIZE);
+		call_prom("package-to-path", 3, 1, node, path, PROM_SCRATCH_SIZE-1);
+		prom_debug("  node %s :\n", path);
+#endif /* DEBUG_PROM */
+
+		while ((endp - p) >= (_prom->root_addr_cells + _prom->root_size_cells)) {
+			unsigned long base, size;
+
+			base = prom_next_cell(_prom->root_addr_cells, &p);
+			size = prom_next_cell(_prom->root_size_cells, &p);
+
+			if (size == 0)
+				continue;
+			prom_debug("    %x %x\n", base, size);
+			if (base == 0)
+				RELOC(rmo_top) = size;
+			if ((base + size) > RELOC(ram_top))
+				RELOC(ram_top) = base + size;
+		}
+	}
+
+	/* Setup our top/bottom alloc points, that is top of RMO or top of
+	 * segment 0 when running non-LPAR
+	 */
+	if ( RELOC(of_platform) == PLATFORM_PSERIES_LPAR )
+		RELOC(alloc_top) = RELOC(rmo_top);
+	else
+		RELOC(alloc_top) = min(0x40000000ul, RELOC(ram_top));
+	RELOC(alloc_bottom) = PAGE_ALIGN(RELOC(klimit) - offset + 0x4000);
+	RELOC(alloc_top_high) = RELOC(ram_top);
+
+	/* Check if we have an initrd after the kernel, if we do move our bottom
+	 * point to after it
+	 */
+	if (RELOC(prom_initrd_start)) {
+		if ((RELOC(prom_initrd_start) + RELOC(prom_initrd_end))
+		    > RELOC(alloc_bottom))
+			RELOC(alloc_bottom) = PAGE_ALIGN(RELOC(prom_initrd_end));
+	}
+
+	prom_printf("memory layout at init:\n");
+	prom_printf("  alloc_bottom : %x\n", RELOC(alloc_bottom));
+	prom_printf("  alloc_top    : %x\n", RELOC(alloc_top));
+	prom_printf("  alloc_top_hi : %x\n", RELOC(alloc_top_high));
+	prom_printf("  rmo_top      : %x\n", RELOC(rmo_top));
+	prom_printf("  ram_top      : %x\n", RELOC(ram_top));
+}
+
+
+/*
+ * Allocate room for and instanciate RTAS
+ */
+static void __init prom_instantiate_rtas(void)
+{
+	unsigned long offset = reloc_offset();
+	struct prom_t *_prom = PTRRELOC(&prom);
+	phandle prom_rtas;
+	u64 base, entry = 0;
+        u32 size;
+
+	prom_debug("prom_instantiate_rtas: start...\n");
+
+	prom_rtas = call_prom("finddevice", 1, 1, ADDR("/rtas"));
+	if (prom_rtas != (phandle) -1) {
+		prom_getprop(prom_rtas, "rtas-size", &size, sizeof(size));
+		if (size != 0) {
+			base = alloc_down(size, PAGE_SIZE, 0);
+			if (base == 0) {
+				prom_printf("RTAS allocation failed !\n");
+				return;
+			}
+			prom_printf("instantiating rtas at 0x%x", base);
+
+			prom_rtas = call_prom("open", 1, 1, ADDR("/rtas"));
+			prom_printf("...");
+
+			if (call_prom("call-method", 3, 2,
+				      ADDR("instantiate-rtas"),
+				      prom_rtas, base) != PROM_ERROR) {
+				entry = (long)_prom->args.rets[1];
+			}
+			if (entry == 0) {
+				prom_printf(" failed\n");
+				return;
+			}
+			prom_printf(" done\n");
+
+	       		reserve_mem(base, size);
+		}
+
+		prom_setprop(_prom->chosen, "linux,rtas-base", &base, sizeof(base));
+		prom_setprop(_prom->chosen, "linux,rtas-entry", &entry, sizeof(entry));
+		prom_setprop(_prom->chosen, "linux,rtas-size", &size, sizeof(size));
+
+		prom_debug("rtas base     = 0x%x\n", base);
+		prom_debug("rtas entry    = 0x%x\n", entry);
+		prom_debug("rtas size     = 0x%x\n", (long)size);
+	}
+	prom_debug("prom_instantiate_rtas: end...\n");
+}
+
+
+/*
+ * Allocate room for and initialize TCE tables
+ */
+static void __init prom_initialize_tce_table(void)
+{
+	phandle node;
+	ihandle phb_node;
+	unsigned long offset = reloc_offset();
+	char compatible[64], type[64], model[64];
+	char *path = RELOC(prom_scratch);
+	u64 i, table = 0;
+	u64 base, vbase, align;
+	u32 minalign, minsize;
+	u64 tce_entry, *tce_entryp;
+	u64 local_alloc_top, local_alloc_bottom;
+
+	if (RELOC(ppc64_iommu_off))
+		return;
+
+	prom_debug("starting prom_initialize_tce_table\n");
+
+	/* Cache current top of allocs so we reserve a single block */
+	local_alloc_top = RELOC(alloc_top_high);
+	local_alloc_bottom = local_alloc_top;
+
+	/* Search all nodes looking for PHBs. */
+	for (node = 0; prom_next_node(&node); ) {
+		compatible[0] = 0;
+		type[0] = 0;
+		model[0] = 0;
+		prom_getprop(node, "compatible",
+			     compatible, sizeof(compatible));
+		prom_getprop(node, "device_type", type, sizeof(type));
+		prom_getprop(node, "model", model, sizeof(model));
+
+		if ((type[0] == 0) || (strstr(type, RELOC("pci")) == NULL))
+			continue;
+
+		/* Keep the old logic in tack to avoid regression. */
+		if (compatible[0] != 0) {
+			if ((strstr(compatible, RELOC("python")) == NULL) &&
+			    (strstr(compatible, RELOC("Speedwagon")) == NULL) &&
+			    (strstr(compatible, RELOC("Winnipeg")) == NULL))
+				continue;
+		} else if (model[0] != 0) {
+			if ((strstr(model, RELOC("ython")) == NULL) &&
+			    (strstr(model, RELOC("peedwagon")) == NULL) &&
+			    (strstr(model, RELOC("innipeg")) == NULL))
+				continue;
+		}
+
+		if (prom_getprop(node, "tce-table-minalign", &minalign,
+				 sizeof(minalign)) == PROM_ERROR)
+			minalign = 0;
+		if (prom_getprop(node, "tce-table-minsize", &minsize,
+				 sizeof(minsize)) == PROM_ERROR)
+			minsize = 4UL << 20;
+
+		/*
+		 * Even though we read what OF wants, we just set the table
+		 * size to 4 MB.  This is enough to map 2GB of PCI DMA space.
+		 * By doing this, we avoid the pitfalls of trying to DMA to
+		 * MMIO space and the DMA alias hole.
+		 *
+		 * On POWER4, firmware sets the TCE region by assuming
+		 * each TCE table is 8MB. Using this memory for anything
+		 * else will impact performance, so we always allocate 8MB.
+		 * Anton
+		 */
+		if (__is_processor(PV_POWER4) || __is_processor(PV_POWER4p))
+			minsize = 8UL << 20;
+		else
+			minsize = 4UL << 20;
+
+		/* Align to the greater of the align or size */
+		align = max(minalign, minsize);
+		base = alloc_down(minsize, align, 1);
+		if (base == 0)
+			prom_panic("ERROR, cannot find space for TCE table.\n");
+		if (base < local_alloc_bottom)
+			local_alloc_bottom = base;
+
+		vbase = (unsigned long)abs_to_virt(base);
+
+		/* Save away the TCE table attributes for later use. */
+		prom_setprop(node, "linux,tce-base", &vbase, sizeof(vbase));
+		prom_setprop(node, "linux,tce-size", &minsize, sizeof(minsize));
+		prom_setprop(node, "linux,has-tce-table", NULL, 0);
+
+		prom_debug("TCE table: 0x%x\n", table);
+		prom_debug("\tnode = 0x%x\n", node);
+		prom_debug("\tbase = 0x%x\n", vbase);
+		prom_debug("\tsize = 0x%x\n", minsize);
+
+		/* Initialize the table to have a one-to-one mapping
+		 * over the allocated size.
+		 */
+		tce_entryp = (unsigned long *)base;
+		for (i = 0; i < (minsize >> 3) ;tce_entryp++, i++) {
+			tce_entry = (i << PAGE_SHIFT);
+			tce_entry |= 0x3;
+			*tce_entryp = tce_entry;
+		}
+
+		/* It seems OF doesn't null-terminate the path :-( */
+		memset(path, 0, sizeof(path));
+		/* Call OF to setup the TCE hardware */
+		if (call_prom("package-to-path", 3, 1, node,
+			      path, PROM_SCRATCH_SIZE-1) == PROM_ERROR) {
+			prom_printf("package-to-path failed\n");
+		} else
+			prom_printf("opening PHB %s", path);
+
+		phb_node = call_prom("open", 1, 1, path);
+		if ( (long)phb_node <= 0)
+			prom_printf("... failed\n");
+		else
+			prom_printf("... done\n");
+
+		call_prom("call-method", 6, 0, ADDR("set-64-bit-addressing"),
+			  phb_node, -1, minsize,
+			  (u32) base, (u32) (base >> 32));
+		call_prom("close", 1, 0, phb_node);
+
+		table++;
+	}
+
+	reserve_mem(local_alloc_bottom, local_alloc_top - local_alloc_bottom);
+
+	/* Flag the first invalid entry */
+	prom_debug("ending prom_initialize_tce_table\n");
+}
+
+/*
+ * With CHRP SMP we need to use the OF to start the other
+ * processors so we can't wait until smp_boot_cpus (the OF is
+ * trashed by then) so we have to put the processors into
+ * a holding pattern controlled by the kernel (not OF) before
+ * we destroy the OF.
+ *
+ * This uses a chunk of low memory, puts some holding pattern
+ * code there and sends the other processors off to there until
+ * smp_boot_cpus tells them to do something.  The holding pattern
+ * checks that address until its cpu # is there, when it is that
+ * cpu jumps to __secondary_start().  smp_boot_cpus() takes care
+ * of setting those values.
+ *
+ * We also use physical address 0x4 here to tell when a cpu
+ * is in its holding pattern code.
+ *
+ * Fixup comment... DRENG / PPPBBB - Peter
+ *
+ * -- Cort
+ */
+static void __init prom_hold_cpus(void)
+{
+	unsigned long i;
+	unsigned int reg;
+	phandle node;
+	unsigned long offset = reloc_offset();
+	char type[64];
+	int cpuid = 0;
+	unsigned int interrupt_server[MAX_CPU_THREADS];
+	unsigned int cpu_threads, hw_cpu_num;
+	int propsize;
+	extern void __secondary_hold(void);
+	extern unsigned long __secondary_hold_spinloop;
+	extern unsigned long __secondary_hold_acknowledge;
+	unsigned long *spinloop
+		= (void *)virt_to_abs(&__secondary_hold_spinloop);
+	unsigned long *acknowledge
+		= (void *)virt_to_abs(&__secondary_hold_acknowledge);
+	unsigned long secondary_hold
+		= virt_to_abs(*PTRRELOC((unsigned long *)__secondary_hold));
+	struct prom_t *_prom = PTRRELOC(&prom);
+
+	prom_debug("prom_hold_cpus: start...\n");
+	prom_debug("    1) spinloop       = 0x%x\n", (unsigned long)spinloop);
+	prom_debug("    1) *spinloop      = 0x%x\n", *spinloop);
+	prom_debug("    1) acknowledge    = 0x%x\n",
+		   (unsigned long)acknowledge);
+	prom_debug("    1) *acknowledge   = 0x%x\n", *acknowledge);
+	prom_debug("    1) secondary_hold = 0x%x\n", secondary_hold);
+
+	/* Set the common spinloop variable, so all of the secondary cpus
+	 * will block when they are awakened from their OF spinloop.
+	 * This must occur for both SMP and non SMP kernels, since OF will
+	 * be trashed when we move the kernel.
+	 */
+	*spinloop = 0;
+
+#ifdef CONFIG_HMT
+	for (i=0; i < NR_CPUS; i++) {
+		RELOC(hmt_thread_data)[i].pir = 0xdeadbeef;
+	}
+#endif
+	/* look for cpus */
+	for (node = 0; prom_next_node(&node); ) {
+		type[0] = 0;
+		prom_getprop(node, "device_type", type, sizeof(type));
+		if (strcmp(type, RELOC("cpu")) != 0)
+			continue;
+
+		/* Skip non-configured cpus. */
+		prom_getprop(node, "status", type, sizeof(type));
+		if (strcmp(type, RELOC("okay")) != 0)
+			continue;
+
+		reg = -1;
+		prom_getprop(node, "reg", &reg, sizeof(reg));
+
+		prom_debug("\ncpuid        = 0x%x\n", cpuid);
+		prom_debug("cpu hw idx   = 0x%x\n", reg);
+
+		/* Init the acknowledge var which will be reset by
+		 * the secondary cpu when it awakens from its OF
+		 * spinloop.
+		 */
+		*acknowledge = (unsigned long)-1;
+
+		propsize = prom_getprop(node, "ibm,ppc-interrupt-server#s",
+					&interrupt_server,
+					sizeof(interrupt_server));
+		if (propsize < 0) {
+			/* no property.  old hardware has no SMT */
+			cpu_threads = 1;
+			interrupt_server[0] = reg; /* fake it with phys id */
+		} else {
+			/* We have a threaded processor */
+			cpu_threads = propsize / sizeof(u32);
+			if (cpu_threads > MAX_CPU_THREADS) {
+				prom_printf("SMT: too many threads!\n"
+					    "SMT: found %x, max is %x\n",
+					    cpu_threads, MAX_CPU_THREADS);
+				cpu_threads = 1; /* ToDo: panic? */
+			}
+		}
+
+		hw_cpu_num = interrupt_server[0];
+		if (hw_cpu_num != _prom->cpu) {
+			/* Primary Thread of non-boot cpu */
+			prom_printf("%x : starting cpu hw idx %x... ", cpuid, reg);
+			call_prom("start-cpu", 3, 0, node,
+				  secondary_hold, cpuid);
+
+			for ( i = 0 ; (i < 100000000) && 
+			      (*acknowledge == ((unsigned long)-1)); i++ ) ;
+
+			if (*acknowledge == cpuid) {
+				prom_printf("done\n");
+				/* We have to get every CPU out of OF,
+				 * even if we never start it. */
+				if (cpuid >= NR_CPUS)
+					goto next;
+			} else {
+				prom_printf("failed: %x\n", *acknowledge);
+			}
+		}
+#ifdef CONFIG_SMP
+		else
+			prom_printf("%x : boot cpu     %x\n", cpuid, reg);
+#endif
+next:
+#ifdef CONFIG_SMP
+		/* Init paca for secondary threads.   They start later. */
+		for (i=1; i < cpu_threads; i++) {
+			cpuid++;
+			if (cpuid >= NR_CPUS)
+				continue;
+		}
+#endif /* CONFIG_SMP */
+		cpuid++;
+	}
+#ifdef CONFIG_HMT
+	/* Only enable HMT on processors that provide support. */
+	if (__is_processor(PV_PULSAR) || 
+	    __is_processor(PV_ICESTAR) ||
+	    __is_processor(PV_SSTAR)) {
+		prom_printf("    starting secondary threads\n");
+
+		for (i = 0; i < NR_CPUS; i += 2) {
+			if (!cpu_online(i))
+				continue;
+
+			if (i == 0) {
+				unsigned long pir = mfspr(SPRN_PIR);
+				if (__is_processor(PV_PULSAR)) {
+					RELOC(hmt_thread_data)[i].pir = 
+						pir & 0x1f;
+				} else {
+					RELOC(hmt_thread_data)[i].pir = 
+						pir & 0x3ff;
+				}
+			}
+		}
+	} else {
+		prom_printf("Processor is not HMT capable\n");
+	}
+#endif
+
+	if (cpuid > NR_CPUS)
+		prom_printf("WARNING: maximum CPUs (" __stringify(NR_CPUS)
+			    ") exceeded: ignoring extras\n");
+
+	prom_debug("prom_hold_cpus: end...\n");
+}
+
+
+#ifdef CONFIG_BOOTX_TEXT__disabled
+
+/* This function will enable the early boot text when doing OF booting. This
+ * way, xmon output should work too
+ */
+static void __init setup_disp_fake_bi(ihandle dp)
+{
+	//int width = 640, height = 480, depth = 8, pitch;
+	//unsigned address;
+	struct pci_reg_property addrs[8];
+	int i, naddrs;
+	char name[64];
+	unsigned long offset = reloc_offset();
+
+	memset(name, 0, sizeof(name));
+	prom_getprop(dp, "name", name, sizeof(name));
+	name[sizeof(name)-1] = 0;
+	prom_printf("Initializing fake screen: %s\n", name);
+
+	prom_getprop(dp, "width", &global_width, sizeof(global_width));
+	prom_getprop(dp, "height", &global_height, sizeof(global_height));
+	prom_getprop(dp, "depth", &global_depth, sizeof(global_depth));
+	global_pitch = global_width * ((global_depth + 7) / 8);
+	prom_getprop(dp, "linebytes", &global_pitch, sizeof(global_pitch));
+	if (global_pitch == 1)
+		global_pitch = 0x1000;		/* for strange IBM display */
+	global_address = 0;
+
+	prom_printf("width %x height %x depth %x linebytes %x\n",
+		    global_width, global_height, global_depth, global_depth);
+
+	prom_getprop(dp, "address", &global_address, sizeof(global_address));
+	if (global_address == 0) {
+		/* look for an assigned address with a size of >= 1MB */
+		naddrs = prom_getprop(dp, "assigned-addresses",
+				      addrs, sizeof(addrs));
+		naddrs /= sizeof(struct pci_reg_property);
+		for (i = 0; i < naddrs; ++i) {
+			if (addrs[i].size_lo >= (1 << 20)) {
+				global_address = addrs[i].addr.a_lo;
+				/* use the BE aperture if possible */
+				if (addrs[i].size_lo >= (16 << 20))
+					global_address += (8 << 20);
+				break;
+			}
+		}
+		if (global_address == 0) {
+			prom_printf("Failed to get address of frame buffer\n");
+			return;
+		}
+	}
+	//btext_setup_display(width, height, depth, pitch, address);
+	//prom_printf("Addr of fb: %x\n", global_address);
+	RELOC(boot_text_mapped) = 0;
+}
+#endif /* CONFIG_BOOTX_TEXT */
+
+static void __init prom_init_client_services(unsigned long pp)
+{
+	unsigned long offset = reloc_offset();
+	struct prom_t *_prom = PTRRELOC(&prom);
+
+	/* Get a handle to the prom entry point before anything else */
+	_prom->entry = pp;
+
+	/* Init default value for phys size */
+	_prom->root_size_cells = 1;
+	_prom->root_addr_cells = 2;
+
+	/* get a handle for the stdout device */
+	_prom->chosen = call_prom("finddevice", 1, 1, ADDR("/chosen"));
+	if ((long)_prom->chosen <= 0)
+		prom_panic("cannot find chosen"); /* msg won't be printed :( */
+
+	/* get device tree root */
+	_prom->root = call_prom("finddevice", 1, 1, ADDR("/"));
+	if ((long)_prom->root <= 0)
+		prom_panic("cannot find device tree root"); /* msg won't be printed :( */
+}
+
+static void __init prom_init_stdout(void)
+{
+	unsigned long offset = reloc_offset();
+	struct prom_t *_prom = PTRRELOC(&prom);
+	char *path = RELOC(of_stdout_device);
+	char type[16];
+	u32 val;
+
+	if (prom_getprop(_prom->chosen, "stdout", &val, sizeof(val)) <= 0)
+		prom_panic("cannot find stdout");
+
+	_prom->stdout = val;
+
+	/* Get the full OF pathname of the stdout device */
+	memset(path, 0, 256);
+	call_prom("instance-to-path", 3, 1, _prom->stdout, path, 255);
+	val = call_prom("instance-to-package", 1, 1, _prom->stdout);
+	prom_setprop(_prom->chosen, "linux,stdout-package", &val, sizeof(val));
+	prom_printf("OF stdout device is: %s\n", RELOC(of_stdout_device));
+
+	/* If it's a display, note it */
+	memset(type, 0, sizeof(type));
+	prom_getprop(val, "device_type", type, sizeof(type));
+	if (strcmp(type, RELOC("display")) == 0) {
+		_prom->disp_node = val;
+		prom_setprop(val, "linux,boot-display", NULL, 0);
+	}
+}
+
+static int __init prom_find_machine_type(void)
+{
+	unsigned long offset = reloc_offset();
+	struct prom_t *_prom = PTRRELOC(&prom);
+	char compat[256];
+	int len, i = 0;
+	phandle rtas;
+
+	len = prom_getprop(_prom->root, "compatible",
+			   compat, sizeof(compat)-1);
+	if (len > 0) {
+		compat[len] = 0;
+		while (i < len) {
+			char *p = &compat[i];
+			int sl = strlen(p);
+			if (sl == 0)
+				break;
+			if (strstr(p, RELOC("Power Macintosh")) ||
+			    strstr(p, RELOC("MacRISC4")))
+				return PLATFORM_POWERMAC;
+			i += sl + 1;
+		}
+	}
+	/* Default to pSeries. We need to know if we are running LPAR */
+	rtas = call_prom("finddevice", 1, 1, ADDR("/rtas"));
+	if (rtas != (phandle) -1) {
+		unsigned long x;
+		x = prom_getproplen(rtas, "ibm,hypertas-functions");
+		if (x != PROM_ERROR) {
+			prom_printf("Hypertas detected, assuming LPAR !\n");
+			return PLATFORM_PSERIES_LPAR;
+		}
+	}
+	return PLATFORM_PSERIES;
+}
+
+static int __init prom_set_color(ihandle ih, int i, int r, int g, int b)
+{
+	unsigned long offset = reloc_offset();
+
+	return call_prom("call-method", 6, 1, ADDR("color!"), ih, i, b, g, r);
+}
+
+/*
+ * If we have a display that we don't know how to drive,
+ * we will want to try to execute OF's open method for it
+ * later.  However, OF will probably fall over if we do that
+ * we've taken over the MMU.
+ * So we check whether we will need to open the display,
+ * and if so, open it now.
+ */
+static unsigned long __init prom_check_displays(void)
+{
+	unsigned long offset = reloc_offset();
+	struct prom_t *_prom = PTRRELOC(&prom);
+	char type[16], *path;
+	phandle node;
+	ihandle ih;
+	int i;
+
+	static unsigned char default_colors[] = {
+		0x00, 0x00, 0x00,
+		0x00, 0x00, 0xaa,
+		0x00, 0xaa, 0x00,
+		0x00, 0xaa, 0xaa,
+		0xaa, 0x00, 0x00,
+		0xaa, 0x00, 0xaa,
+		0xaa, 0xaa, 0x00,
+		0xaa, 0xaa, 0xaa,
+		0x55, 0x55, 0x55,
+		0x55, 0x55, 0xff,
+		0x55, 0xff, 0x55,
+		0x55, 0xff, 0xff,
+		0xff, 0x55, 0x55,
+		0xff, 0x55, 0xff,
+		0xff, 0xff, 0x55,
+		0xff, 0xff, 0xff
+	};
+	const unsigned char *clut;
+
+	prom_printf("Looking for displays\n");
+	for (node = 0; prom_next_node(&node); ) {
+		memset(type, 0, sizeof(type));
+		prom_getprop(node, "device_type", type, sizeof(type));
+		if (strcmp(type, RELOC("display")) != 0)
+			continue;
+
+		/* It seems OF doesn't null-terminate the path :-( */
+		path = RELOC(prom_scratch);
+		memset(path, 0, PROM_SCRATCH_SIZE);
+
+		/*
+		 * leave some room at the end of the path for appending extra
+		 * arguments
+		 */
+		if (call_prom("package-to-path", 3, 1, node, path, PROM_SCRATCH_SIZE-10) < 0)
+			continue;
+		prom_printf("found display   : %s, opening ... ", path);
+		
+		ih = call_prom("open", 1, 1, path);
+		if (ih == (ihandle)0 || ih == (ihandle)-1) {
+			prom_printf("failed\n");
+			continue;
+		}
+
+		/* Success */
+		prom_printf("done\n");
+		prom_setprop(node, "linux,opened", NULL, 0);
+
+		/*
+		 * stdout wasn't a display node, pick the first we can find
+		 * for btext
+		 */
+		if (_prom->disp_node == 0)
+			_prom->disp_node = node;
+
+		/* Setup a useable color table when the appropriate
+		 * method is available. Should update this to set-colors */
+		clut = RELOC(default_colors);
+		for (i = 0; i < 32; i++, clut += 3)
+			if (prom_set_color(ih, i, clut[0], clut[1],
+					   clut[2]) != 0)
+				break;
+
+#ifdef CONFIG_LOGO_LINUX_CLUT224
+		clut = PTRRELOC(RELOC(logo_linux_clut224.clut));
+		for (i = 0; i < RELOC(logo_linux_clut224.clutsize); i++, clut += 3)
+			if (prom_set_color(ih, i + 32, clut[0], clut[1],
+					   clut[2]) != 0)
+				break;
+#endif /* CONFIG_LOGO_LINUX_CLUT224 */
+	}
+}
+
+
+/* Return (relocated) pointer to this much memory: moves initrd if reqd. */
+static void __init *make_room(unsigned long *mem_start, unsigned long *mem_end,
+			      unsigned long needed, unsigned long align)
+{
+	unsigned long offset = reloc_offset();
+	void *ret;
+
+	*mem_start = _ALIGN(*mem_start, align);
+	while ((*mem_start + needed) > *mem_end) {
+		unsigned long room, chunk;
+
+		prom_debug("Chunk exhausted, claiming more at %x...\n",
+			   RELOC(alloc_bottom));
+		room = RELOC(alloc_top) - RELOC(alloc_bottom);
+		if (room > DEVTREE_CHUNK_SIZE)
+			room = DEVTREE_CHUNK_SIZE;
+		if (room < PAGE_SIZE)
+			prom_panic("No memory for flatten_device_tree (no room)");
+		chunk = alloc_up(room, 0);
+		if (chunk == 0)
+			prom_panic("No memory for flatten_device_tree (claim failed)");
+		*mem_end = RELOC(alloc_top);
+	}
+
+	ret = (void *)*mem_start;
+	*mem_start += needed;
+
+	return ret;
+}
+
+#define dt_push_token(token, mem_start, mem_end) \
+	do { *((u32 *)make_room(mem_start, mem_end, 4, 4)) = token; } while(0)
+
+static unsigned long __init dt_find_string(char *str)
+{
+	unsigned long offset = reloc_offset();
+	char *s, *os;
+
+	s = os = (char *)RELOC(dt_string_start);
+	s += 4;
+	while (s <  (char *)RELOC(dt_string_end)) {
+		if (strcmp(s, str) == 0)
+			return s - os;
+		s += strlen(s) + 1;
+	}
+	return 0;
+}
+
+static void __init scan_dt_build_strings(phandle node, unsigned long *mem_start,
+					 unsigned long *mem_end)
+{
+	unsigned long offset = reloc_offset();
+	char *prev_name, *namep, *sstart;
+	unsigned long soff;
+	phandle child;
+
+	sstart =  (char *)RELOC(dt_string_start);
+
+	/* get and store all property names */
+	prev_name = RELOC("");
+	for (;;) {
+		
+		/* 32 is max len of name including nul. */
+		namep = make_room(mem_start, mem_end, 32, 1);
+		if (call_prom("nextprop", 3, 1, node, prev_name, namep) <= 0) {
+			/* No more nodes: unwind alloc */
+			*mem_start = (unsigned long)namep;
+			break;
+		}
+		soff = dt_find_string(namep);
+		if (soff != 0) {
+			*mem_start = (unsigned long)namep;
+			namep = sstart + soff;
+		} else {
+			/* Trim off some if we can */
+			*mem_start = (unsigned long)namep + strlen(namep) + 1;
+			RELOC(dt_string_end) = *mem_start;
+		}
+		prev_name = namep;
+	}
+
+	/* do all our children */
+	child = call_prom("child", 1, 1, node);
+	while (child != (phandle)0) {
+		scan_dt_build_strings(child, mem_start, mem_end);
+		child = call_prom("peer", 1, 1, child);
+	}
+}
+
+static void __init scan_dt_build_struct(phandle node, unsigned long *mem_start,
+					unsigned long *mem_end)
+{
+	int l, align;
+	phandle child;
+	char *namep, *prev_name, *sstart;
+	unsigned long soff;
+	unsigned char *valp;
+	unsigned long offset = reloc_offset();
+	char pname[32];
+	char *path;
+
+	path = RELOC(prom_scratch);
+
+	dt_push_token(OF_DT_BEGIN_NODE, mem_start, mem_end);
+
+	/* get the node's full name */
+	namep = (char *)*mem_start;
+	l = call_prom("package-to-path", 3, 1, node,
+		      namep, *mem_end - *mem_start);
+	if (l >= 0) {
+		/* Didn't fit?  Get more room. */
+		if (l+1 > *mem_end - *mem_start) {
+			namep = make_room(mem_start, mem_end, l+1, 1);
+			call_prom("package-to-path", 3, 1, node, namep, l);
+		}
+		namep[l] = '\0';
+		*mem_start = _ALIGN(((unsigned long) namep) + strlen(namep) + 1, 4);
+	}
+
+	/* get it again for debugging */
+	memset(path, 0, PROM_SCRATCH_SIZE);
+	call_prom("package-to-path", 3, 1, node, path, PROM_SCRATCH_SIZE-1);
+
+	/* get and store all properties */
+	prev_name = RELOC("");
+	sstart = (char *)RELOC(dt_string_start);
+	for (;;) {
+		if (call_prom("nextprop", 3, 1, node, prev_name, pname) <= 0)
+			break;
+
+		/* find string offset */
+		soff = dt_find_string(pname);
+		if (soff == 0) {
+			prom_printf("WARNING: Can't find string index for <%s>, node %s\n",
+				    pname, path);
+			break;
+		}
+		prev_name = sstart + soff;
+
+		/* get length */
+		l = call_prom("getproplen", 2, 1, node, pname);
+
+		/* sanity checks */
+		if (l < 0)
+			continue;
+		if (l > MAX_PROPERTY_LENGTH) {
+			prom_printf("WARNING: ignoring large property ");
+			/* It seems OF doesn't null-terminate the path :-( */
+			prom_printf("[%s] ", path);
+			prom_printf("%s length 0x%x\n", pname, l);
+			continue;
+		}
+
+		/* push property head */
+		dt_push_token(OF_DT_PROP, mem_start, mem_end);
+		dt_push_token(l, mem_start, mem_end);
+		dt_push_token(soff, mem_start, mem_end);
+
+		/* push property content */
+		align = (l >= 8) ? 8 : 4;
+		valp = make_room(mem_start, mem_end, l, align);
+		call_prom("getprop", 4, 1, node, pname, valp, l);
+		*mem_start = _ALIGN(*mem_start, 4);
+	}
+
+	/* Add a "linux,phandle" property. */
+	soff = dt_find_string(RELOC("linux,phandle"));
+	if (soff == 0)
+		prom_printf("WARNING: Can't find string index for <linux-phandle>"
+			    " node %s\n", path);
+	else {
+		dt_push_token(OF_DT_PROP, mem_start, mem_end);
+		dt_push_token(4, mem_start, mem_end);
+		dt_push_token(soff, mem_start, mem_end);
+		valp = make_room(mem_start, mem_end, 4, 4);
+		*(u32 *)valp = node;
+	}
+
+	/* do all our children */
+	child = call_prom("child", 1, 1, node);
+	while (child != (phandle)0) {
+		scan_dt_build_struct(child, mem_start, mem_end);
+		child = call_prom("peer", 1, 1, child);
+	}
+
+	dt_push_token(OF_DT_END_NODE, mem_start, mem_end);
+}
+
+static void __init flatten_device_tree(void)
+{
+	phandle root;
+	unsigned long offset = reloc_offset();
+	unsigned long mem_start, mem_end, room;
+	struct boot_param_header *hdr;
+	char *namep;
+	u64 *rsvmap;
+
+	/*
+	 * Check how much room we have between alloc top & bottom (+/- a
+	 * few pages), crop to 4Mb, as this is our "chuck" size
+	 */
+	room = RELOC(alloc_top) - RELOC(alloc_bottom) - 0x4000;
+	if (room > DEVTREE_CHUNK_SIZE)
+		room = DEVTREE_CHUNK_SIZE;
+	prom_debug("starting device tree allocs at %x\n", RELOC(alloc_bottom));
+
+	/* Now try to claim that */
+	mem_start = (unsigned long)alloc_up(room, PAGE_SIZE);
+	if (mem_start == 0)
+		prom_panic("Can't allocate initial device-tree chunk\n");
+	mem_end = RELOC(alloc_top);
+
+	/* Get root of tree */
+	root = call_prom("peer", 1, 1, (phandle)0);
+	if (root == (phandle)0)
+		prom_panic ("couldn't get device tree root\n");
+
+	/* Build header and make room for mem rsv map */ 
+	mem_start = _ALIGN(mem_start, 4);
+	hdr = make_room(&mem_start, &mem_end, sizeof(struct boot_param_header), 4);
+	RELOC(dt_header_start) = (unsigned long)hdr;
+	rsvmap = make_room(&mem_start, &mem_end, sizeof(mem_reserve_map), 8);
+
+	/* Start of strings */
+	mem_start = PAGE_ALIGN(mem_start);
+	RELOC(dt_string_start) = mem_start;
+	mem_start += 4; /* hole */
+
+	/* Add "linux,phandle" in there, we'll need it */
+	namep = make_room(&mem_start, &mem_end, 16, 1);
+	strcpy(namep, RELOC("linux,phandle"));
+	mem_start = (unsigned long)namep + strlen(namep) + 1;
+	RELOC(dt_string_end) = mem_start;
+
+	/* Build string array */
+	prom_printf("Building dt strings...\n"); 
+	scan_dt_build_strings(root, &mem_start, &mem_end);
+
+	/* Build structure */
+	mem_start = PAGE_ALIGN(mem_start);
+	RELOC(dt_struct_start) = mem_start;
+	prom_printf("Building dt structure...\n"); 
+	scan_dt_build_struct(root, &mem_start, &mem_end);
+	dt_push_token(OF_DT_END, &mem_start, &mem_end);
+	RELOC(dt_struct_end) = PAGE_ALIGN(mem_start);
+
+	/* Finish header */
+	hdr->magic = OF_DT_HEADER;
+	hdr->totalsize = RELOC(dt_struct_end) - RELOC(dt_header_start);
+	hdr->off_dt_struct = RELOC(dt_struct_start) - RELOC(dt_header_start);
+	hdr->off_dt_strings = RELOC(dt_string_start) - RELOC(dt_header_start);
+	hdr->off_mem_rsvmap = ((unsigned long)rsvmap) - RELOC(dt_header_start);
+	hdr->version = OF_DT_VERSION;
+	hdr->last_comp_version = 1;
+
+	/* Reserve the whole thing and copy the reserve map in, we
+	 * also bump mem_reserve_cnt to cause further reservations to
+	 * fail since it's too late.
+	 */
+	reserve_mem(RELOC(dt_header_start), hdr->totalsize);
+	memcpy(rsvmap, RELOC(mem_reserve_map), sizeof(mem_reserve_map));
+
+#ifdef DEBUG_PROM
+	{
+		int i;
+		prom_printf("reserved memory map:\n");
+		for (i = 0; i < RELOC(mem_reserve_cnt); i++)
+			prom_printf("  %x - %x\n", RELOC(mem_reserve_map)[i].base,
+				    RELOC(mem_reserve_map)[i].size);
+	}
+#endif
+	RELOC(mem_reserve_cnt) = MEM_RESERVE_MAP_SIZE;
+
+	prom_printf("Device tree strings 0x%x -> 0x%x\n",
+		    RELOC(dt_string_start), RELOC(dt_string_end)); 
+	prom_printf("Device tree struct  0x%x -> 0x%x\n",
+		    RELOC(dt_struct_start), RELOC(dt_struct_end));
+
+ }
+
+static void __init prom_find_boot_cpu(void)
+{
+	unsigned long offset = reloc_offset();
+       	struct prom_t *_prom = PTRRELOC(&prom);
+	u32 getprop_rval;
+	ihandle prom_cpu;
+	phandle cpu_pkg;
+
+	if (prom_getprop(_prom->chosen, "cpu", &prom_cpu, sizeof(prom_cpu)) <= 0)
+		prom_panic("cannot find boot cpu");
+
+	cpu_pkg = call_prom("instance-to-package", 1, 1, prom_cpu);
+
+	prom_setprop(cpu_pkg, "linux,boot-cpu", NULL, 0);
+	prom_getprop(cpu_pkg, "reg", &getprop_rval, sizeof(getprop_rval));
+	_prom->cpu = getprop_rval;
+
+	prom_debug("Booting CPU hw index = 0x%x\n", _prom->cpu);
+}
+
+static void __init prom_check_initrd(unsigned long r3, unsigned long r4)
+{
+#ifdef CONFIG_BLK_DEV_INITRD
+	unsigned long offset = reloc_offset();
+       	struct prom_t *_prom = PTRRELOC(&prom);
+
+	if ( r3 && r4 && r4 != 0xdeadbeef) {
+		u64 val;
+
+		RELOC(prom_initrd_start) = (r3 >= KERNELBASE) ? __pa(r3) : r3;
+		RELOC(prom_initrd_end) = RELOC(prom_initrd_start) + r4;
+
+		val = (u64)RELOC(prom_initrd_start);
+		prom_setprop(_prom->chosen, "linux,initrd-start", &val, sizeof(val));
+		val = (u64)RELOC(prom_initrd_end);
+		prom_setprop(_prom->chosen, "linux,initrd-end", &val, sizeof(val));
+
+		reserve_mem(RELOC(prom_initrd_start),
+			    RELOC(prom_initrd_end) - RELOC(prom_initrd_start));
+
+		prom_debug("initrd_start=0x%x\n", RELOC(prom_initrd_start));
+		prom_debug("initrd_end=0x%x\n", RELOC(prom_initrd_end));
+	}
+#endif /* CONFIG_BLK_DEV_INITRD */
+}
+
+/*
+ * We enter here early on, when the Open Firmware prom is still
+ * handling exceptions and the MMU hash table for us.
+ */
+
+unsigned long __init prom_init(unsigned long r3, unsigned long r4, unsigned long pp,
+			       unsigned long r6, unsigned long r7)
+{	
+	unsigned long offset = reloc_offset();
+       	struct prom_t *_prom = PTRRELOC(&prom);
+	unsigned long phys;
+	u32 getprop_rval;
+	
+	/*
+	 * First zero the BSS
+	 */
+	memset(PTRRELOC(&__bss_start), 0, __bss_stop - __bss_start);
+
+	/*
+	 * Init interface to Open Firmware, get some node references,
+	 * like /chosen
+	 */
+	prom_init_client_services(pp);
+
+	/*
+	 * Init prom stdout device
+	 */
+	prom_init_stdout();
+	prom_debug("klimit=0x%x\n", RELOC(klimit));
+	prom_debug("offset=0x%x\n", offset);
+
+	/*
+	 * Reserve kernel in reserve map
+	 */
+	reserve_mem(0, __pa(RELOC(klimit)));
+
+	/*
+	 * Check for an initrd
+	 */
+	prom_check_initrd(r3, r4);
+
+	/*
+	 * Get default machine type. At this point, we do not differenciate
+	 * between pSeries SMP and pSeries LPAR
+	 */
+	RELOC(of_platform) = prom_find_machine_type();
+	getprop_rval = RELOC(of_platform);
+	prom_setprop(_prom->chosen, "linux,platform",
+		     &getprop_rval, sizeof(getprop_rval));
+
+	/*
+	 * On pSeries, copy the CPU hold code
+	 */
+       	if (RELOC(of_platform) & PLATFORM_PSERIES)
+       		copy_and_flush(0, KERNELBASE - offset, 0x100, 0);
+
+	/*
+	 * Get memory cells format
+	 */
+	getprop_rval = 1;
+	prom_getprop(_prom->root, "#size-cells",
+		     &getprop_rval, sizeof(getprop_rval));
+	_prom->root_size_cells = getprop_rval;
+	getprop_rval = 2;
+	prom_getprop(_prom->root, "#address-cells",
+		     &getprop_rval, sizeof(getprop_rval));
+	_prom->root_addr_cells = getprop_rval;
+
+	/*
+	 * Do early parsing of command line
+	 */
+	early_cmdline_parse();
+
+	/*
+	 * Initialize memory management within prom_init
+	 */
+	prom_init_mem();
+
+	/*
+	 * Determine which cpu is actually running right _now_
+	 */
+	prom_find_boot_cpu();
+
+	/* 
+	 * Initialize display devices
+	 */
+	prom_check_displays();
+
+	/*
+	 * Initialize IOMMU (TCE tables) on pSeries. Do that before anything else
+	 * that uses the allocator, we need to make sure we get the top of memory
+	 * available for us here...
+	 */
+	if (RELOC(of_platform) == PLATFORM_PSERIES)
+		prom_initialize_tce_table();
+
+	/*
+	 * On non-powermacs, try to instantiate RTAS and puts all CPUs
+	 * in spin-loops. PowerMacs don't have a working RTAS and use
+	 * a different way to spin CPUs
+	 */
+	if (RELOC(of_platform) != PLATFORM_POWERMAC) {
+		prom_instantiate_rtas();
+		prom_hold_cpus();
+	}
+
+	/*
+	 * Fill in some infos for use by the kernel later on
+	 */
+	if (RELOC(ppc64_iommu_off))
+		prom_setprop(_prom->chosen, "linux,iommu-off", NULL, 0);
+	if (RELOC(iommu_force_on))
+		prom_setprop(_prom->chosen, "linux,iommu-force-on", NULL, 0);
+
+
+	// XXX KEXEC TEST HACK
+	if (0) {
+		extern unsigned long klimit;
+		u64 ksize = __pa(RELOC(klimit));
+		u64 kcopy;
+		
+		kcopy = alloc_down(ksize, PAGE_SIZE, 0);
+		if (kcopy != 0) {
+			reserve_mem(kcopy, ksize);
+			prom_printf("duplicating kernel at %x ...\n", kcopy);
+			copy_and_flush(kcopy, KERNELBASE - offset, ksize, 0);
+			prom_setprop(_prom->chosen, "linux,kcopy-base", &kcopy, 8);
+			prom_setprop(_prom->chosen, "linux,kcopy-size", &ksize, 8);
+		}
+	}
+
+
+	/*
+	 * Now finally create the flattened device-tree
+	 */
+       	prom_printf("copying OF device tree ...\n");
+       	flatten_device_tree();
+
+	/*
+	 * Call OF "quiesce" method to shut down pending DMA's from
+	 * devices etc...
+	 */
+	prom_printf("Calling quiesce ...\n");
+	call_prom("quiesce", 0, 0);
+
+	/*
+	 * And finally, call the kernel passing it the flattened device
+	 * tree and NULL as r5, thus triggering the new entry point which
+	 * is common to us and kexec
+	 */
+	phys = KERNELBASE - offset;
+	prom_printf("returning from prom_init\n");
+	prom_debug("->dt_header_start=0x%x\n", RELOC(dt_header_start));
+	prom_debug("->phys=0x%x\n", phys);
+
+	__start(RELOC(dt_header_start), phys, 0, RELOC(klimit));
+
+	return 0;
+}
+
diff -urN linux-2.5/arch/ppc64/kernel/rtas.c linux-prom-clean/arch/ppc64/kernel/rtas.c
--- linux-2.5/arch/ppc64/kernel/rtas.c	2004-09-06 16:43:57.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/rtas.c	2004-09-17 21:52:58.000000000 +1000
@@ -42,6 +42,7 @@
 
 spinlock_t rtas_data_buf_lock = SPIN_LOCK_UNLOCKED;
 char rtas_data_buf[RTAS_DATA_BUF_SIZE]__page_aligned;
+unsigned long rtas_rmo_buf;
 
 void
 call_rtas_display_status(char c)
@@ -425,7 +426,6 @@
 	} while (status == RTAS_BUSY);
 }
 
-unsigned long rtas_rmo_buf = 0;
 
 asmlinkage int ppc_rtas(struct rtas_args __user *uargs)
 {
@@ -536,6 +536,50 @@
 	return rtas_error_log_max;
 }
 
+/*
+ * Call early during boot, before mem init or bootmem, to retreive the RTAS
+ * informations from the device-tree and allocate the RMO buffer for userland
+ * accesses.
+ */
+void __init rtas_initialize(void)
+{
+	/* Get RTAS dev node and fill up our "rtas" structure with infos
+	 * about it.
+	 */
+	rtas.dev = of_find_node_by_name(NULL, "rtas");
+	if (rtas.dev) {
+		u64 *basep, *entryp;
+		u32 *sizep;
+
+		basep = (u64 *)get_property(of_chosen, "linux,rtas-base", NULL);
+		sizep = (u32 *)get_property(of_chosen, "linux,rtas-size", NULL);
+		if (basep != NULL && sizep != NULL) {
+			rtas.base = *basep;
+			rtas.size = *sizep;
+			entryp = (u64 *)get_property(of_chosen, "linux,rtas-entry", NULL);
+			if (entryp == NULL) /* Ugh */
+				rtas.entry = rtas.base;
+			else
+				rtas.entry = *entryp;
+		} else
+			rtas.dev = NULL;
+	}
+	/* If RTAS was found, allocate the RMO buffer for it */
+	if (rtas.dev) {
+		unsigned long rtas_region = RTAS_INSTANTIATE_MAX;
+		if (systemcfg->platform == PLATFORM_PSERIES_LPAR)
+			rtas_region = min(lmb.rmo_size, RTAS_INSTANTIATE_MAX);
+
+		rtas_rmo_buf = lmb_alloc_base(RTAS_RMOBUF_MAX, PAGE_SIZE,
+							rtas_region);
+	}
+
+#ifdef CONFIG_HOTPLUG_CPU
+	rtas_stop_self_args.token = rtas_token("stop-self");
+#endif /* CONFIG_HOTPLUG_CPU */
+}
+
+
 EXPORT_SYMBOL(rtas_firmware_flash_list);
 EXPORT_SYMBOL(rtas_token);
 EXPORT_SYMBOL(rtas_call);
diff -urN linux-2.5/arch/ppc64/kernel/setup.c linux-prom-clean/arch/ppc64/kernel/setup.c
--- linux-2.5/arch/ppc64/kernel/setup.c	2004-09-15 16:25:15.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/setup.c	2004-09-17 22:09:01.000000000 +1000
@@ -48,24 +48,24 @@
 #include <asm/system.h>
 #include <asm/rtas.h>
 
-extern unsigned long klimit;
+struct pci_reg_property {
+	struct pci_address addr;
+	u32 size_hi;
+	u32 size_lo;
+};
+                                                                                
+struct isa_reg_property {
+	u32 space;
+	u32 address;
+	u32 size;
+};
+
 /* extern void *stab; */
 extern HTAB htab_data;
+extern unsigned long klimit;
 
-int have_of = 1;
-
-extern void  chrp_init(unsigned long r3,
-		       unsigned long r4,
-		       unsigned long r5,
-		       unsigned long r6,
-		       unsigned long r7);
-
-extern void  pmac_init(unsigned long r3,
-		       unsigned long r4,
-		       unsigned long r5,
-		       unsigned long r6,
-		       unsigned long r7);
-
+extern void chrp_init(void);
+extern void pmac_init(void);
 extern void fw_feature_init(void);
 extern void iSeries_init_early(void);
 extern void pSeries_init_early(void);
@@ -76,6 +76,11 @@
 extern int  idle_setup(void);
 extern void vpa_init(int cpu);
 extern void iSeries_parse_cmdline(void);
+extern void stab_initialize(unsigned long stab);
+extern void htab_initialize(void);
+extern void early_init_devtree(void *flat_dt);
+extern void unflatten_device_tree(void);
+extern void pmac_iommu_alloc(void);
 
 unsigned long decr_overclock = 1;
 unsigned long decr_overclock_proc0 = 1;
@@ -83,8 +88,7 @@
 unsigned long decr_overclock_proc0_set = 0;
 
 int powersave_nap;
-
-unsigned char aux_device_present;
+int have_of = 1;
 
 #ifdef CONFIG_MAGIC_SYSRQ
 unsigned long SYSRQ_KEY;
@@ -297,7 +301,6 @@
 
 	systemcfg->processorCount = num_present_cpus();
 }
-
 #endif /* !defined(CONFIG_PPC_ISERIES) && defined(CONFIG_SMP) */
 
 #ifdef CONFIG_XMON
@@ -312,12 +315,178 @@
 early_param("xmon", early_xmon);
 #endif
 
+
+#ifdef CONFIG_PPC_PSERIES
+/*
+ * TODO: Kill as much of that as possible and/or move it to setup_system
+ * or setup_arch... or eventually kill setup_system
+ */
+static void __init initialize_naca(void)
+{
+	struct device_node *np;
+	char type[64];
+	char *typep = type;
+	unsigned long num_cpus = 0;
+	struct naca_struct *_naca = naca;
+	struct systemcfg *_systemcfg = systemcfg;
+	unsigned long encode_phys_size = 32;
+	void *prop;
+
+	np = of_find_node_by_path("/");
+	if (!np)
+		return;
+	prop = get_property(np, "#size-cells", NULL);
+	if (prop != NULL)
+		encode_phys_size = (*((u32 *)prop)) << 5;
+	
+	for (np = NULL; (np = of_find_node_by_type(np, "cpu"));) {
+		num_cpus += 1;
+
+		/* We're assuming *all* of the CPUs have the same
+		 * d-cache and i-cache sizes... -Peter
+		 */
+
+		if ( num_cpus == 1 ) {
+			u32 *sizep, *lsizep;
+			const char *dc, *ic;
+
+			/* Then read cache informations */
+			if (_systemcfg->platform == PLATFORM_POWERMAC){
+				dc = "d-cache-block-size";
+				ic = "i-cache-block-size";
+			} else {
+				dc = "d-cache-line-size";
+				ic = "i-cache-line-size";
+			}
+
+			sizep = (u32 *)get_property(np, "d-cache-size", NULL);
+
+			lsizep = (u32 *) get_property(np, dc, NULL);
+
+			_systemcfg->dCacheL1Size = *sizep;
+			_systemcfg->dCacheL1LineSize = *lsizep;
+			_naca->dCacheL1LogLineSize = __ilog2(*lsizep);
+			_naca->dCacheL1LinesPerPage = PAGE_SIZE/(*lsizep);
+
+			sizep = (u32 *)get_property(np, "i-cache-size", NULL);
+
+			lsizep = (u32 *)get_property(np, ic, NULL);
+
+			_systemcfg->iCacheL1Size = *sizep;
+			_systemcfg->iCacheL1LineSize = *lsizep;
+			_naca->iCacheL1LogLineSize = __ilog2(*lsizep);
+			_naca->iCacheL1LinesPerPage = PAGE_SIZE/(*lsizep);
+
+		}
+	}
+
+	for (np = NULL; (np = of_find_node_by_type(np, "serial"));) {
+		struct device_node *isa, *pci;
+		struct isa_reg_property *reg;
+		union pci_range ranges;
+		union pci_range *rangesp = &ranges;
+
+		if (_systemcfg->platform == PLATFORM_POWERMAC)
+			continue;
+		type[0] = 0;
+	 	typep = (char *)get_property(np, "ibm,aix-loc", NULL);
+		if ((typep == NULL) || (typep && strcmp(typep, "S1")))
+			continue;
+
+		reg = (struct isa_reg_property *)get_property(np, "reg", NULL);	
+
+		isa = of_get_parent(np);
+		if (!isa)
+			panic("no isa parent found\n");
+		pci = of_get_parent(isa);
+		if (!pci)
+			panic("no pci parent found\n");
+
+		rangesp = (union pci_range *)get_property(pci, "ranges", NULL);
+
+		if ( encode_phys_size == 32 )
+			_naca->serialPortAddr = ranges.pci32.phys+reg->address;
+		else {
+			_naca->serialPortAddr =
+				((((unsigned long)ranges.pci64.phys_hi) << 32)
+				|
+			(ranges.pci64.phys_lo)) + reg->address;
+		}
+	}
+
+	if (_systemcfg->platform == PLATFORM_POWERMAC)
+		_naca->interrupt_controller = IC_OPEN_PIC;
+	else {
+		__irq_offset_value = NUM_ISA_INTERRUPTS;
+		_naca->interrupt_controller = IC_INVALID;
+		np = NULL;
+		while ((np = of_find_node_by_name(np, "interrupt-controller")))
+		{
+			typep = (char *)get_property(np, "compatible", NULL);
+			if (strstr(typep, "open-pic"))
+				naca->interrupt_controller = IC_OPEN_PIC;
+			else if (strstr(typep, "ppc-xicp"))
+				naca->interrupt_controller = IC_PPC_XIC;
+			else
+				printk("initialize_naca: failed to recognize"
+				" interrupt-controller\n");
+			break;
+		}
+	}
+
+	if (_naca->interrupt_controller == IC_INVALID) {
+		panic("failed to find interrupt-controller\n");
+	}
+	
+	/* We gotta have at least 1 cpu... */
+	if ( (_systemcfg->processorCount = num_cpus) < 1 )
+		 panic("no cpus found");
+
+	/* Add an eye catcher and the systemcfg layout version number */
+	strcpy(_systemcfg->eye_catcher, "SYSTEMCFG:PPC64");
+	_systemcfg->version.major = SYSTEMCFG_MAJOR;
+	_systemcfg->version.minor = SYSTEMCFG_MINOR;
+	_systemcfg->processor = mfspr(SPRN_PVR);
+}
+
+static void __init check_for_initrd(void)
+{
+#ifdef CONFIG_BLK_DEV_INITRD
+	u64 *prop;
+
+	prop = (u64 *)get_property(of_chosen, "linux,initrd-start", NULL);
+	if (prop != NULL) {
+		initrd_start = (unsigned long)__va(*prop);
+		prop = (u64 *)get_property(of_chosen, "linux,initrd-end", NULL);
+		if (prop != NULL) {
+			initrd_end = (unsigned long)__va(*prop);
+			initrd_below_start_ok = 1;
+		} else
+			initrd_start = 0;
+	}
+
+	/* If we were passed an initrd, set the ROOT_DEV properly if the values
+	 * look sensible. If not, clear initrd reference.
+	 */
+	if (initrd_start >= KERNELBASE && initrd_end >= KERNELBASE &&
+	    initrd_end > initrd_start)
+		ROOT_DEV = Root_RAM0;
+	else
+		initrd_start = initrd_end = 0;
+
+	if (initrd_start)
+		printk("Found initrd at 0x%lx:0x%lx\n", initrd_start, initrd_end);
+#endif /* CONFIG_BLK_DEV_INITRD */
+}
+
+
+#endif /* CONFIG_PPC_PSERIES */
+
 /*
  * Do some initial setup of the system.  The parameters are those which 
  * were passed in from the bootloader.
  */
-void setup_system(unsigned long r3, unsigned long r4, unsigned long r5,
-		  unsigned long r6, unsigned long r7)
+void __init setup_system(void)
 {
 #if defined(CONFIG_SMP) && defined(CONFIG_PPC_PSERIES)
 	int ret, i;
@@ -329,6 +498,45 @@
 		systemcfg->platform = PLATFORM_ISERIES_LPAR;
 #endif
 
+#ifdef CONFIG_PPC_PSERIES
+
+	/*
+	 * Unflatten the device-tree passed by prom_init or kexec
+	 */
+	unflatten_device_tree();
+
+	/*
+	 * Fill the naca & systemcfg structures with informations
+	 * retreived from the device-tree. Need to be called before
+	 * finish_device_tree() since the later requires some of the
+	 * informations filled up here to properly parse the interrupt
+	 * tree.
+	 */
+	initialize_naca();
+
+	/*
+	 * "Finish" the device-tree, that is do the actual parsing of
+	 * some of the properties like the interrupt map
+	 */
+	finish_device_tree();
+
+	/*
+	 * Initialize RTAS if available
+	 */
+	rtas_initialize();
+
+	/*
+	 * Check if we have an initrd provided via the device-tree
+	 */
+	check_for_initrd();
+
+#endif /* CONFIG_PPC_PSERIES */
+	
+	/*
+	 * XXX Replace that by a call to a ppc_md call, the correct ppc_md is
+	 * to be found via a probe() call iterating them all done from
+	 * early init. May eventually be merged with setup_arch though ....
+	 */
 	switch (systemcfg->platform) {
 #ifdef CONFIG_PPC_ISERIES
 	case PLATFORM_ISERIES_LPAR:
@@ -353,32 +561,33 @@
 #endif /* CONFIG_PPC_PMAC */
 	}
 
-#ifdef CONFIG_XMON_DEFAULT
-	xmon_init();
-#endif
-	/* If we were passed an initrd, set the ROOT_DEV properly if the values
-	 * look sensible. If not, clear initrd reference.
+	/* Finish initializing the hash table (do the dynamic
+	 * patching for the fast-path hashtable.S code).
+	 * Done here after the above "early" code has setup the
+	 * hash pointer callbacks
 	 */
-#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 */
+	htab_finish_init();
 
+	/*
+	 * If we have BootX text initialized, map it now so it can be used
+	 * for debugging
+	 */
 #ifdef CONFIG_BOOTX_TEXT
 	map_boot_text();
-	if (systemcfg->platform == PLATFORM_POWERMAC) {
-		early_console_initialized = 1;
-		register_console(&udbg_console);
-	}
 #endif /* CONFIG_BOOTX_TEXT */
 
+	/*
+	 * Initialize xmon
+	 */
+#ifdef CONFIG_XMON_DEFAULT
+	xmon_init();
+#endif
+
 #ifdef CONFIG_PPC_PMAC
 	if (systemcfg->platform == PLATFORM_POWERMAC) {
-		finish_device_tree();
-		pmac_init(r3, r4, r5, r6, r7);
+		early_console_initialized = 1;
+		register_console(&udbg_console);
+		pmac_init();
 	}
 #endif /* CONFIG_PPC_PMAC */
 
@@ -386,10 +595,9 @@
 	if (systemcfg->platform & PLATFORM_PSERIES) {
 		early_console_initialized = 1;
 		register_console(&udbg_console);
-		__irq_offset_value = NUM_ISA_INTERRUPTS;
-		finish_device_tree();
-		chrp_init(r3, r4, r5, r6, r7);
+		chrp_init();
 	}
+
 #endif /* CONFIG_PPC_PSERIES */
 
 #ifdef CONFIG_PPC_ISERIES
@@ -429,15 +637,6 @@
 #endif /* CONFIG_PPC_PSERIES */
 #endif /* CONFIG_SMP */
 
-#if defined(CONFIG_HOTPLUG_CPU) &&  !defined(CONFIG_PPC_PMAC)
-	rtas_stop_self_args.token = rtas_token("stop-self");
-#endif /* CONFIG_HOTPLUG_CPU && !CONFIG_PPC_PMAC */
-
-	/* Finish initializing the hash table (do the dynamic
-	 * patching for the fast-path hashtable.S code)
-	 */
-	htab_finish_init();
-
 	printk("Starting Linux PPC64 %s\n", UTS_RELEASE);
 
 	printk("-----------------------------------------------------\n");
@@ -526,7 +725,7 @@
 #ifdef CONFIG_SMP
 	pvr = per_cpu(pvr, cpu_id);
 #else
-	pvr = _get_PVR();
+	pvr = mfpvr(PSRN_PVR);
 #endif
 	maj = (pvr >> 8) & 0xFF;
 	min = pvr & 0xFF;
@@ -599,18 +798,24 @@
 	struct device_node *prom_stdout;
 	char *name;
 	int offset = 0;
+	phandle *stdout_ph;
 
 	/* The user has requested a console so this is already set up. */
 	if (strstr(saved_command_line, "console="))
 		return -EBUSY;
 
-	prom_stdout = find_path_device(of_stdout_device);
+	if (!of_chosen)
+		return -ENODEV;
+	stdout_ph = (phandle *)get_property(of_chosen, "linux,stdout-package", NULL);
+	if (stdout_ph == NULL)
+		return -ENODEV;
+	prom_stdout = of_find_node_by_phandle(*stdout_ph);
 	if (!prom_stdout)
 		return -ENODEV;
 
 	name = (char *)get_property(prom_stdout, "name", NULL);
 	if (!name)
-		return -ENODEV;
+		goto not_found;
 
 	if (strcmp(name, "serial") == 0) {
 		int i;
@@ -631,7 +836,7 @@
 					break;
 				default:
 					/* We dont recognise the serial port */
-					return -ENODEV;
+					goto not_found;
 			}
 		}
 	} else if (strcmp(name, "vty") == 0) {
@@ -649,11 +854,13 @@
  					offset = 1;
  					break;
  				default:
- 					return -ENODEV;
+					goto not_found;
  			}
+			of_node_put(prom_stdout);
  			return add_preferred_console("hvsi", offset, NULL);
  		} else {
  			/* pSeries LPAR virtual console */
+			of_node_put(prom_stdout);
  			return add_preferred_console("hvc", 0, NULL);
  		}
 	} else if (strcmp(name, "ch-a") == 0)
@@ -661,10 +868,14 @@
 	else if (strcmp(name, "ch-b") == 0)
 		offset = 1;
 	else
-		return -ENODEV;
+		goto not_found;
+	of_node_put(prom_stdout);
 
 	return add_preferred_console("ttyS", offset, NULL);
 
+ not_found:
+	of_node_put(prom_stdout);
+	return -ENODEV;
 }
 console_initcall(set_preferred_console);
 #endif
@@ -877,6 +1088,43 @@
 
 }
 
+#ifdef CONFIG_PPC_PSERIES
+
+void __init early_init(unsigned long dt_ptr)
+{
+	struct paca_struct *lpaca = get_paca();
+
+	/*
+	 * Do early initializations using the flattened device
+	 * tree, like retreiving the physical memory map or
+	 * calculating/retreiving the hash table size
+	 */
+	early_init_devtree(__va(dt_ptr));
+
+	/*
+	 * On PowerMac, the DART (iommu) must be allocated now since it
+	 * has an impact on htab_initialize (due to the large page it
+	 * occupies having to be broken up so the DART itself is not
+	 * part of the cacheable linar mapping
+	 */
+#ifdef CONFIG_PMAC_DART
+	pmac_iommu_alloc();
+#endif
+	/*
+	 * Initialize stab / SLB management
+	 */
+	stab_initialize(lpaca->stab_real);
+
+	/*
+	 * Initialize the MMU Hash table and create the linear mapping
+	 * of memory
+	 */
+	htab_initialize();
+}
+
+#endif /* CONFIG_PPC_PSERIES */
+
+
 __setup("spread_lpevents=", set_spread_lpevents );
 __setup("decr_overclock_proc0=", set_decr_overclock_proc0 );
 __setup("decr_overclock=", set_decr_overclock );
diff -urN linux-2.5/arch/ppc64/kernel/smp.c linux-prom-clean/arch/ppc64/kernel/smp.c
--- linux-2.5/arch/ppc64/kernel/smp.c	2004-09-15 16:25:15.000000000 +1000
+++ linux-prom-clean/arch/ppc64/kernel/smp.c	2004-09-15 21:53:26.000000000 +1000
@@ -75,6 +75,7 @@
 			 unsigned long vpa);
 
 int smt_enabled_at_boot = 1;
+int boot_cpuid = 0;
 
 /* Low level assembly function used to backup CPU 0 state */
 extern void __save_cpu_setup(void);
@@ -749,7 +750,7 @@
 
 static void __devinit smp_store_cpu_info(int id)
 {
-	per_cpu(pvr, id) = _get_PVR();
+	per_cpu(pvr, id) = mfspr(SPRN_PVR);
 }
 
 static void __init smp_create_idle(unsigned int cpu)
diff -urN linux-2.5/arch/ppc64/mm/hash_utils.c linux-prom-clean/arch/ppc64/mm/hash_utils.c
--- linux-2.5/arch/ppc64/mm/hash_utils.c	2004-09-15 16:25:15.000000000 +1000
+++ linux-prom-clean/arch/ppc64/mm/hash_utils.c	2004-09-17 20:55:51.000000000 +1000
@@ -147,8 +147,11 @@
 	htab_data.htab_num_ptegs = pteg_count;
 	htab_data.htab_hash_mask = pteg_count - 1;
 
-	if (systemcfg->platform == PLATFORM_PSERIES ||
-	    systemcfg->platform == PLATFORM_POWERMAC) {
+	if (systemcfg->platform & PLATFORM_LPAR) {
+		/* Using a hypervisor which owns the htab */
+		htab_data.htab = NULL;
+		_SDR1 = 0; 
+	} else {
 		/* Find storage for the HPT.  Must be contiguous in
 		 * the absolute address space.
 		 */
@@ -164,10 +167,6 @@
 
 		/* Initialize the HPT with no entries */
 		memset((void *)table, 0, htab_size_bytes);
-	} else {
-		/* Using a hypervisor which owns the htab */
-		htab_data.htab = NULL;
-		_SDR1 = 0; 
 	}
 
 	mode_rw = _PAGE_ACCESSED | _PAGE_COHERENT | PP_RWXX;
@@ -179,7 +178,7 @@
 	if (cur_cpu_spec->cpu_features & CPU_FTR_16M_PAGE)
 		use_largepages = 1;
 
-	/* add all physical memory to the bootmem map */
+	/* create bolted the linear mapping in the hash table */
 	for (i=0; i < lmb.memory.cnt; i++) {
 		unsigned long base, size;
 
diff -urN linux-2.5/arch/ppc64/oprofile/common.c linux-prom-clean/arch/ppc64/oprofile/common.c
--- linux-2.5/arch/ppc64/oprofile/common.c	2004-08-26 15:46:30.000000000 +1000
+++ linux-prom-clean/arch/ppc64/oprofile/common.c	2004-09-15 21:53:26.000000000 +1000
@@ -133,7 +133,7 @@
 {
 	unsigned int pvr;
 
-	pvr = _get_PVR();
+	pvr = mfspr(SPRN_PVR);
 
 	switch (PVR_VER(pvr)) {
 		case PV_630:
diff -urN linux-2.5/include/asm-ppc64/lmb.h linux-prom-clean/include/asm-ppc64/lmb.h
--- linux-2.5/include/asm-ppc64/lmb.h	2004-09-06 16:43:57.000000000 +1000
+++ linux-prom-clean/include/asm-ppc64/lmb.h	2004-09-17 14:35:46.000000000 +1000
@@ -20,12 +20,6 @@
 
 #define MAX_LMB_REGIONS 128
 
-union lmb_reg_property { 
-	struct reg_property32	 addr32[MAX_LMB_REGIONS];
-	struct reg_property64	 addr64[MAX_LMB_REGIONS];
-	struct reg_property_pmac addrPM[MAX_LMB_REGIONS];
-};
-
 #define LMB_ALLOC_ANYWHERE	0
 
 struct lmb_property {
@@ -60,6 +54,8 @@
 extern unsigned long __init lmb_end_of_DRAM(void);
 extern unsigned long __init lmb_abs_to_phys(unsigned long);
 
+extern void lmb_dump_all(void);
+
 extern unsigned long io_hole_start;
 
 #endif /* _PPC64_LMB_H */
diff -urN linux-2.5/include/asm-ppc64/prom.h linux-prom-clean/include/asm-ppc64/prom.h
--- linux-2.5/include/asm-ppc64/prom.h	2004-08-10 10:22:37.000000000 +1000
+++ linux-prom-clean/include/asm-ppc64/prom.h	2004-09-17 17:34:38.000000000 +1000
@@ -24,14 +24,47 @@
 #define LONG_LSW(X) (((unsigned long)X) & 0xffffffff)
 #define LONG_MSW(X) (((unsigned long)X) >> 32)
 
+/* Definitions used by the flattened device tree */
+#define OF_DT_HEADER		0xd00dfeed	/* 4: version, 4: total size */
+#define OF_DT_BEGIN_NODE	0x1		/* Start node: full name */
+#define OF_DT_END_NODE		0x2		/* End node */
+#define OF_DT_PROP		0x3		/* Property: name off, size, content */
+#define OF_DT_END		0x9
+
+#define OF_DT_VERSION		1
+
+/*
+ * This is what gets passed to the kernel by prom_init or kexec
+ *
+ * The dt struct contains the device tree structure, full pathes and
+ * property contents. The dt strings contain a separate block with just
+ * the strings for the property names, and is fully page aligned and
+ * self contained in a page, so that it can be kept around by the kernel,
+ * each property name appears only once in this page (cheap compression)
+ *
+ * the mem_rsvmap contains a map of reserved ranges of physical memory,
+ * passing it here instead of in the device-tree itself greatly simplifies
+ * the job of everybody. It's just a list of u64 pairs (base/size) that
+ * ends when size is 0
+ */
+struct boot_param_header
+{
+	u32	magic;			/* magic word OF_DT_HEADER */
+	u32	totalsize;		/* total size of DT block */
+	u32	off_dt_struct;		/* offset to structure */
+	u32	off_dt_strings;		/* offset to strings */
+	u32	off_mem_rsvmap;		/* offset to memory reserve map */
+	u32	version;		/* format version */
+	u32	last_comp_version;	/* last compatible version */
+};
+
+
+
 typedef u32 phandle;
 typedef u32 ihandle;
 typedef u32 phandle32;
 typedef u32 ihandle32;
 
-extern char *prom_display_paths[];
-extern unsigned int prom_num_displays;
-
 struct address_range {
 	unsigned long space;
 	unsigned long address;
@@ -136,6 +169,7 @@
  */
 struct pci_controller;
 struct iommu_table;
+
 struct device_node {
 	char	*name;
 	char	*type;
@@ -171,6 +205,8 @@
 	unsigned long _flags;
 };
 
+extern struct device_node *of_chosen;
+
 /* flag descriptions */
 #define OF_STALE   0 /* node is slated for deletion */
 #define OF_DYNAMIC 1 /* node and properties were allocated via kmalloc */
@@ -202,34 +238,6 @@
 	dn->addr_link = de;
 }
 
-typedef u32 prom_arg_t;
-
-struct prom_args {
-        u32 service;
-        u32 nargs;
-        u32 nret;
-        prom_arg_t args[10];
-        prom_arg_t *rets;     /* Pointer to return values in args[16]. */
-};
-
-struct prom_t {
-	unsigned long entry;
-	ihandle root;
-	ihandle chosen;
-	int cpu;
-	ihandle stdout;
-	ihandle disp_node;
-	struct prom_args args;
-	unsigned long version;
-	unsigned long encode_phys_size;
-	struct bi_record *bi_recs;
-};
-
-extern struct prom_t prom;
-extern char *of_stdout_device;
-
-extern int boot_cpuid;
-
 /* OBSOLETE: Old stlye node lookup */
 extern struct device_node *find_devices(const char *name);
 extern struct device_node *find_type_devices(const char *type);
@@ -246,6 +254,7 @@
 extern struct device_node *of_find_compatible_node(struct device_node *from,
 	const char *type, const char *compat);
 extern struct device_node *of_find_node_by_path(const char *path);
+extern struct device_node *of_find_node_by_phandle(phandle handle);
 extern struct device_node *of_find_all_nodes(struct device_node *prev);
 extern struct device_node *of_get_parent(const struct device_node *node);
 extern struct device_node *of_get_next_child(const struct device_node *node,
diff -urN linux-2.5/include/asm-ppc64/rtas.h linux-prom-clean/include/asm-ppc64/rtas.h
--- linux-2.5/include/asm-ppc64/rtas.h	2004-09-06 16:43:57.000000000 +1000
+++ linux-prom-clean/include/asm-ppc64/rtas.h	2004-09-17 19:54:46.000000000 +1000
@@ -189,6 +189,7 @@
 extern int rtas_get_power_level(int powerdomain, int *level);
 extern int rtas_set_power_level(int powerdomain, int level, int *setlevel);
 extern int rtas_set_indicator(int indicator, int index, int new_value);
+extern void rtas_initialize(void);
 
 /* Given an RTAS status code of 9900..9905 compute the hinted delay */
 unsigned int rtas_extended_busy_delay_time(int status);
diff -urN linux-2.5/include/asm-ppc64/smp.h linux-prom-clean/include/asm-ppc64/smp.h
--- linux-2.5/include/asm-ppc64/smp.h	2004-09-15 16:25:15.000000000 +1000
+++ linux-prom-clean/include/asm-ppc64/smp.h	2004-09-15 21:53:31.000000000 +1000
@@ -37,6 +37,7 @@
 #define hard_smp_processor_id() (get_paca()->hw_cpu_id)
 
 extern cpumask_t cpu_sibling_map[NR_CPUS];
+extern int boot_cpuid;
 
 /* Since OpenPIC has only 4 IPIs, we use slightly different message numbers.
  *
