prom_init.c

/*
 * Procedures for interfacing to Open Firmware.
 *
 * Paul Mackerras	August 1996.
 * Copyright (C) 1996-2005 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.
 */

#undef DEBUG_PROM

#include <stdarg.h>
#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/init.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 <linux/bitops.h>
#include <asm/prom.h>
#include <asm/rtas.h>
#include <asm/page.h>
#include <asm/processor.h>
#include <asm/irq.h>
#include <asm/io.h>
#include <asm/smp.h>
#include <asm/mmu.h>
#include <asm/pgtable.h>
#include <asm/pci.h>
#include <asm/iommu.h>
#include <asm/btext.h>
#include <asm/sections.h>
#include <asm/machdep.h>
#include <asm/opal.h>

#include <linux/linux_logo.h>

/*
 * 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.
 * On ppc32 we compile with -mrelocatable, which means that references
 * to extern and static variables get relocated automatically.
 * ppc64 objects are always relocatable, we just need to relocate the
 * TOC.
 *
 * 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 must be done within prom_init().
 *
 * ADDR is used in calls to call_prom.  The 4th and following
 * arguments to call_prom should be 32-bit values.
 * On ppc64, 64 bit values are truncated to 32 bits (and
 * fortunately don't get interpreted as two arguments).
 */
#define ADDR(x)		(u32)(unsigned long)(x)

#ifdef CONFIG_PPC64
#define OF_WORKAROUNDS	0
#else
#define OF_WORKAROUNDS	of_workarounds
int of_workarounds;
#endif

#define OF_WA_CLAIM	1	/* do phys/virt claim separately, then map */
#define OF_WA_LONGTRAIL	2	/* work around longtrail bugs */

#define PROM_BUG() do {						\
        prom_printf("kernel BUG at %s line 0x%x!\n",		\
		    __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];
};

struct prom_t {
	ihandle root;
	phandle chosen;
	int cpu;
	ihandle stdout;
	ihandle mmumap;
	ihandle memory;
};

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, unsigned long r7, unsigned long r8,
		    unsigned long r9);

#ifdef CONFIG_PPC64
extern int enter_prom(struct prom_args *args, unsigned long entry);
#else
static inline int enter_prom(struct prom_args *args, unsigned long entry)
{
	return ((int (*)(struct prom_args *))entry)(args);
}
#endif

extern void copy_and_flush(unsigned long dest, unsigned long src,
			   unsigned long size, unsigned long offset);

/* prom structure */
static struct prom_t __initdata prom;

static unsigned long prom_entry __initdata;

#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;

#ifdef CONFIG_PPC64
static int __initdata prom_iommu_force_on;
static int __initdata prom_iommu_off;
static unsigned long __initdata prom_tce_alloc_start;
static unsigned long __initdata prom_tce_alloc_end;
#endif

/* Platforms codes are now obsolete in the kernel. Now only used within this
 * file and ultimately gone too. Feel free to change them if you need, they
 * are not shared with anything outside of this file anymore
 */
#define PLATFORM_PSERIES	0x0100
#define PLATFORM_PSERIES_LPAR	0x0101
#define PLATFORM_LPAR		0x0001
#define PLATFORM_POWERMAC	0x0400
#define PLATFORM_GENERIC	0x0500
#define PLATFORM_OPAL		0x0600

static int __initdata of_platform;

static char __initdata prom_cmd_line[COMMAND_LINE_SIZE];

static unsigned long __initdata prom_memory_limit;

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];


/*
 * Error results ... some OF calls will return "-1" on error, some
 * will return 0, some will return either. To simplify, here are
 * macros to use with any ihandle or phandle return value to check if
 * it is valid
 */

#define PROM_ERROR		(-1u)
#define PHANDLE_VALID(p)	((p) != 0 && (p) != PROM_ERROR)
#define IHANDLE_VALID(i)	((i) != 0 && (i) != PROM_ERROR)


/* This is the one and *ONLY* place where we actually call open
 * firmware.
 */

static int __init call_prom(const char *service, int nargs, int nret, ...)
{
	int i;
	struct prom_args args;
	va_list list;

	args.service = ADDR(service);
	args.nargs = nargs;
	args.nret = nret;

	va_start(list, nret);
	for (i = 0; i < nargs; i++)
		args.args[i] = va_arg(list, prom_arg_t);
	va_end(list);

	for (i = 0; i < nret; i++)
		args.args[nargs+i] = 0;

	if (enter_prom(&args, prom_entry) < 0)
		return PROM_ERROR;

	return (nret > 0) ? args.args[nargs] : 0;
}

static int __init call_prom_ret(const char *service, int nargs, int nret,
				prom_arg_t *rets, ...)
{
	int i;
	struct prom_args args;
	va_list list;

	args.service = ADDR(service);
	args.nargs = nargs;
	args.nret = nret;

	va_start(list, rets);
	for (i = 0; i < nargs; i++)
		args.args[i] = va_arg(list, prom_arg_t);
	va_end(list);

	for (i = 0; i < nret; i++)
		args.args[nargs+i] = 0;

	if (enter_prom(&args, prom_entry) < 0)
		return PROM_ERROR;

	if (rets != NULL)
		for (i = 1; i < nret; ++i)
			rets[i-1] = args.args[nargs+i];

	return (nret > 0) ? args.args[nargs] : 0;
}


static void __init prom_print(const char *msg)
{
	const char *p, *q;

	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)
{
	int i, nibbles = sizeof(val)*2;
	char buf[sizeof(val)*2+1];

	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);
}

/* max number of decimal digits in an unsigned long */
#define UL_DIGITS 21
static void __init prom_print_dec(unsigned long val)
{
	int i, size;
	char buf[UL_DIGITS+1];

	for (i = UL_DIGITS-1; i >= 0;  i--) {
		buf[i] = (val % 10) + '0';
		val = val/10;
		if (val == 0)
			break;
	}
	/* shift stuff down */
	size = UL_DIGITS - i;
	call_prom("write", 3, 1, prom.stdout, buf+i, size);
}

static void __init prom_printf(const char *format, ...)
{
	const char *p, *q, *s;
	va_list args;
	unsigned long v;
	long vs;

	va_start(args, format);
	for (p = 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;
		case 'd':
			++q;
			vs = va_arg(args, int);
			if (vs < 0) {
				prom_print("-");
				vs = -vs;
			}
			prom_print_dec(vs);
			break;
		case 'l':
			++q;
			if (*q == 0)
				break;
			else if (*q == 'x') {
				++q;
				v = va_arg(args, unsigned long);
				prom_print_hex(v);
			} else if (*q == 'u') { /* '%lu' */
				++q;
				v = va_arg(args, unsigned long);
				prom_print_dec(v);
			} else if (*q == 'd') { /* %ld */
				++q;
				vs = va_arg(args, long);
				if (vs < 0) {
					prom_print("-");
					vs = -vs;
				}
				prom_print_dec(vs);
			}
			break;
		}
	}
}


static unsigned int __init prom_claim(unsigned long virt, unsigned long size,
				unsigned long align)
{

	if (align == 0 && (OF_WORKAROUNDS & OF_WA_CLAIM)) {
		/*
		 * Old OF requires we claim physical and virtual separately
		 * and then map explicitly (assuming virtual mode)
		 */
		int ret;
		prom_arg_t result;

		ret = call_prom_ret("call-method", 5, 2, &result,
				    ADDR("claim"), prom.memory,
				    align, size, virt);
		if (ret != 0 || result == -1)
			return -1;
		ret = call_prom_ret("call-method", 5, 2, &result,
				    ADDR("claim"), prom.mmumap,
				    align, size, virt);
		if (ret != 0) {
			call_prom("call-method", 4, 1, ADDR("release"),
				  prom.memory, size, virt);
			return -1;
		}
		/* the 0x12 is M (coherence) + PP == read/write */
		call_prom("call-method", 6, 1,
			  ADDR("map"), prom.mmumap, 0x12, size, virt, virt);
		return virt;
	}
	return call_prom("claim", 3, 1, (prom_arg_t)virt, (prom_arg_t)size,
			 (prom_arg_t)align);
}

static void __init __attribute__((noreturn)) prom_panic(const char *reason)
{
	prom_print(reason);
	/* Do not call exit because it clears the screen on pmac
	 * it also causes some sort of double-fault on early pmacs */
	if (of_platform == PLATFORM_POWERMAC)
		asm("trap\n");

	/* ToDo: should put up an SRC here on pSeries */
	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 inline prom_getprop(phandle node, const char *pname,
			       void *value, size_t valuelen)
{
	return call_prom("getprop", 4, 1, node, ADDR(pname),
			 (u32)(unsigned long) value, (u32) valuelen);
}

static int inline prom_getproplen(phandle node, const char *pname)
{
	return call_prom("getproplen", 2, 1, node, ADDR(pname));
}

static void add_string(char **str, const char *q)
{
	char *p = *str;

	while (*q)
		*p++ = *q++;
	*p++ = ' ';
	*str = p;
}

static char *tohex(unsigned int x)
{
	static char digits[] = "0123456789abcdef";
	static char result[9];
	int i;

	result[8] = 0;
	i = 8;
	do {
		--i;
		result[i] = digits[x & 0xf];
		x >>= 4;
	} while (x != 0 && i > 0);
	return &result[i];
}

static int __init prom_setprop(phandle node, const char *nodename,
			       const char *pname, void *value, size_t valuelen)
{
	char cmd[256], *p;

	if (!(OF_WORKAROUNDS & OF_WA_LONGTRAIL))
		return call_prom("setprop", 4, 1, node, ADDR(pname),
				 (u32)(unsigned long) value, (u32) valuelen);

	/* gah... setprop doesn't work on longtrail, have to use interpret */
	p = cmd;
	add_string(&p, "dev");
	add_string(&p, nodename);
	add_string(&p, tohex((u32)(unsigned long) value));
	add_string(&p, tohex(valuelen));
	add_string(&p, tohex(ADDR(pname)));
	add_string(&p, tohex(strlen(pname)));
	add_string(&p, "property");
	*p = 0;
	return call_prom("interpret", 1, 1, (u32)(unsigned long) cmd);
}

/* We can't use the standard versions because of relocation headaches. */
#define isxdigit(c)	(('0' <= (c) && (c) <= '9') \
			 || ('a' <= (c) && (c) <= 'f') \
			 || ('A' <= (c) && (c) <= 'F'))

#define isdigit(c)	('0' <= (c) && (c) <= '9')
#define islower(c)	('a' <= (c) && (c) <= 'z')
#define toupper(c)	(islower(c) ? ((c) - 'a' + 'A') : (c))

unsigned long prom_strtoul(const char *cp, const char **endp)
{
	unsigned long result = 0, base = 10, value;

	if (*cp == '0') {
		base = 8;
		cp++;
		if (toupper(*cp) == 'X') {
			cp++;
			base = 16;
		}
	}

	while (isxdigit(*cp) &&
	       (value = isdigit(*cp) ? *cp - '0' : toupper(*cp) - 'A' + 10) < base) {
		result = result * base + value;
		cp++;
	}

	if (endp)
		*endp = cp;

	return result;
}

unsigned long prom_memparse(const char *ptr, const char **retptr)
{
	unsigned long ret = prom_strtoul(ptr, retptr);
	int shift = 0;

	/*
	 * We can't use a switch here because GCC *may* generate a
	 * jump table which won't work, because we're not running at
	 * the address we're linked at.
	 */
	if ('G' == **retptr || 'g' == **retptr)
		shift = 30;

	if ('M' == **retptr || 'm' == **retptr)
		shift = 20;

	if ('K' == **retptr || 'k' == **retptr)
		shift = 10;

	if (shift) {
		ret <<= shift;
		(*retptr)++;
	}

	return ret;
}

/*
 * Early parsing of the command line passed to the kernel, used for
 * "mem=x" and the options that affect the iommu
 */
static void __init early_cmdline_parse(void)
{
	const char *opt;

	char *p;
	int l = 0;

	prom_cmd_line[0] = 0;
	p = 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 || p[0] == '\0') /* dbl check */
		strlcpy(prom_cmd_line,
			CONFIG_CMDLINE, sizeof(prom_cmd_line));
#endif /* CONFIG_CMDLINE */
	prom_printf("command line: %s\n", prom_cmd_line);

#ifdef CONFIG_PPC64
	opt = strstr(prom_cmd_line, "iommu=");
	if (opt) {
		prom_printf("iommu opt is: %s\n", opt);
		opt += 6;
		while (*opt && *opt == ' ')
			opt++;
		if (!strncmp(opt, "off", 3))
			prom_iommu_off = 1;
		else if (!strncmp(opt, "force", 5))
			prom_iommu_force_on = 1;
	}
#endif
	opt = strstr(prom_cmd_line, "mem=");
	if (opt) {
		opt += 4;
		prom_memory_limit = prom_memparse(opt, (const char **)&opt);
#ifdef CONFIG_PPC64
		/* Align to 16 MB == size of ppc64 large page */
		prom_memory_limit = ALIGN(prom_memory_limit, 0x1000000);
#endif
	}
}

#if defined(CONFIG_PPC_PSERIES) || defined(CONFIG_PPC_POWERNV)
/*
 * The architecture vector has an array of PVR mask/value pairs,
 * followed by # option vectors - 1, followed by the option vectors.
 *
 * See prom.h for the definition of the bits specified in the
 * architecture vector.
 *
 * Because the description vector contains a mix of byte and word
 * values, we declare it as an unsigned char array, and use this
 * macro to put word values in.
 */
#define W(x)	((x) >> 24) & 0xff, ((x) >> 16) & 0xff, \
		((x) >> 8) & 0xff, (x) & 0xff

unsigned char ibm_architecture_vec[] = {
	W(0xfffe0000), W(0x003a0000),	/* POWER5/POWER5+ */
	W(0xffff0000), W(0x003e0000),	/* POWER6 */
	W(0xffff0000), W(0x003f0000),	/* POWER7 */
	W(0xffff0000), W(0x004b0000),	/* POWER8 */
	W(0xffffffff), W(0x0f000004),	/* all 2.07-compliant */
	W(0xffffffff), W(0x0f000003),	/* all 2.06-compliant */
	W(0xffffffff), W(0x0f000002),	/* all 2.05-compliant */
	W(0xfffffffe), W(0x0f000001),	/* all 2.04-compliant and earlier */
	6 - 1,				/* 6 option vectors */

	/* option vector 1: processor architectures supported */
	3 - 2,				/* length */
	0,				/* don't ignore, don't halt */
	OV1_PPC_2_00 | OV1_PPC_2_01 | OV1_PPC_2_02 | OV1_PPC_2_03 |
	OV1_PPC_2_04 | OV1_PPC_2_05 | OV1_PPC_2_06 | OV1_PPC_2_07,

	/* option vector 2: Open Firmware options supported */
	34 - 2,				/* length */
	OV2_REAL_MODE,
	0, 0,
	W(0xffffffff),			/* real_base */
	W(0xffffffff),			/* real_size */
	W(0xffffffff),			/* virt_base */
	W(0xffffffff),			/* virt_size */
	W(0xffffffff),			/* load_base */
	W(256),				/* 256MB min RMA */
	W(0xffffffff),			/* full client load */
	0,				/* min RMA percentage of total RAM */
	48,				/* max log_2(hash table size) */

	/* option vector 3: processor options supported */
	3 - 2,				/* length */
	0,				/* don't ignore, don't halt */
	OV3_FP | OV3_VMX | OV3_DFP,

	/* option vector 4: IBM PAPR implementation */
	3 - 2,				/* length */
	0,				/* don't halt */
	OV4_MIN_ENT_CAP,		/* minimum VP entitled capacity */

	/* option vector 5: PAPR/OF options */
	19 - 2,				/* length */
	0,				/* don't ignore, don't halt */
	OV5_FEAT(OV5_LPAR) | OV5_FEAT(OV5_SPLPAR) | OV5_FEAT(OV5_LARGE_PAGES) |
	OV5_FEAT(OV5_DRCONF_MEMORY) | OV5_FEAT(OV5_DONATE_DEDICATE_CPU) |
#ifdef CONFIG_PCI_MSI
	/* PCIe/MSI support.  Without MSI full PCIe is not supported */
	OV5_FEAT(OV5_MSI),
#else
	0,
#endif
	0,
#ifdef CONFIG_PPC_SMLPAR
	OV5_FEAT(OV5_CMO) | OV5_FEAT(OV5_XCMO),
#else
	0,
#endif
	OV5_FEAT(OV5_TYPE1_AFFINITY) | OV5_FEAT(OV5_PRRN),
	0,
	0,
	0,
	/* WARNING: The offset of the "number of cores" field below
	 * must match by the macro below. Update the definition if
	 * the structure layout changes.
	 */
#define IBM_ARCH_VEC_NRCORES_OFFSET	117
	W(NR_CPUS),			/* number of cores supported */
	0,
	0,
	0,
	0,
	OV5_FEAT(OV5_PFO_HW_RNG) | OV5_FEAT(OV5_PFO_HW_ENCR) |
	OV5_FEAT(OV5_PFO_HW_842),
	OV5_FEAT(OV5_SUB_PROCESSORS),
	/* option vector 6: IBM PAPR hints */
	4 - 2,				/* length */
	0,
	0,
	OV6_LINUX,

};

/* Old method - ELF header with PT_NOTE sections */
static struct fake_elf {
	Elf32_Ehdr	elfhdr;
	Elf32_Phdr	phdr[2];
	struct chrpnote {
		u32	namesz;
		u32	descsz;
		u32	type;
		char	name[8];	/* "PowerPC" */
		struct chrpdesc {
			u32	real_mode;
			u32	real_base;
			u32	real_size;
			u32	virt_base;
			u32	virt_size;
			u32	load_base;
		} chrpdesc;
	} chrpnote;
	struct rpanote {
		u32	namesz;
		u32	descsz;
		u32	type;
		char	name[24];	/* "IBM,RPA-Client-Config" */
		struct rpadesc {
			u32	lpar_affinity;
			u32	min_rmo_size;
			u32	min_rmo_percent;
			u32	max_pft_size;
			u32	splpar;
			u32	min_load;
			u32	new_mem_def;
			u32	ignore_me;
		} rpadesc;
	} rpanote;
} fake_elf = {
	.elfhdr = {
		.e_ident = { 0x7f, 'E', 'L', 'F',
			     ELFCLASS32, ELFDATA2MSB, EV_CURRENT },
		.e_type = ET_EXEC,	/* yeah right */
		.e_machine = EM_PPC,
		.e_version = EV_CURRENT,
		.e_phoff = offsetof(struct fake_elf, phdr),
		.e_phentsize = sizeof(Elf32_Phdr),
		.e_phnum = 2
	},
	.phdr = {
		[0] = {
			.p_type = PT_NOTE,
			.p_offset = offsetof(struct fake_elf, chrpnote),
			.p_filesz = sizeof(struct chrpnote)
		}, [1] = {
			.p_type = PT_NOTE,
			.p_offset = offsetof(struct fake_elf, rpanote),
			.p_filesz = sizeof(struct rpanote)
		}
	},
	.chrpnote = {
		.namesz = sizeof("PowerPC"),
		.descsz = sizeof(struct chrpdesc),
		.type = 0x1275,
		.name = "PowerPC",
		.chrpdesc = {
			.real_mode = ~0U,	/* ~0 means "don't care" */
			.real_base = ~0U,
			.real_size = ~0U,
			.virt_base = ~0U,
			.virt_size = ~0U,
			.load_base = ~0U
		},
	},
	.rpanote = {
		.namesz = sizeof("IBM,RPA-Client-Config"),
		.descsz = sizeof(struct rpadesc),
		.type = 0x12759999,
		.name = "IBM,RPA-Client-Config",
		.rpadesc = {
			.lpar_affinity = 0,
			.min_rmo_size = 64,	/* in megabytes */
			.min_rmo_percent = 0,
			.max_pft_size = 48,	/* 2^48 bytes max PFT size */
			.splpar = 1,
			.min_load = ~0U,
			.new_mem_def = 0
		}
	}
};

static int __init prom_count_smt_threads(void)
{
	phandle node;
	char type[64];
	unsigned int plen;

	/* Pick up th first CPU node we can find */
	for (node = 0; prom_next_node(&node); ) {
		type[0] = 0;
		prom_getprop(node, "device_type", type, sizeof(type));

		if (strcmp(type, "cpu"))
			continue;
		/*
		 * There is an entry for each smt thread, each entry being
		 * 4 bytes long.  All cpus should have the same number of
		 * smt threads, so return after finding the first.
		 */
		plen = prom_getproplen(node, "ibm,ppc-interrupt-server#s");
		if (plen == PROM_ERROR)
			break;
		plen >>= 2;
		prom_debug("Found %lu smt threads per core\n", (unsigned long)plen);

		/* Sanity check */
		if (plen < 1 || plen > 64) {
			prom_printf("Threads per core %lu out of bounds, assuming 1\n",
				    (unsigned long)plen);
			return 1;
		}
		return plen;
	}
	prom_debug("No threads found, assuming 1 per core\n");

	return 1;

}


static void __init prom_send_capabilities(void)
{
	ihandle elfloader, root;
	prom_arg_t ret;
	u32 *cores;

	root = call_prom("open", 1, 1, ADDR("/"));
	if (root != 0) {
		/* We need to tell the FW about the number of cores we support.
		 *
		 * To do that, we count the number of threads on the first core
		 * (we assume this is the same for all cores) and use it to
		 * divide NR_CPUS.
		 */
		cores = (u32 *)&ibm_architecture_vec[IBM_ARCH_VEC_NRCORES_OFFSET];
		if (*cores != NR_CPUS) {
			prom_printf("WARNING ! "
				    "ibm_architecture_vec structure inconsistent: %lu!\n",
				    *cores);
		} else {
			*cores = DIV_ROUND_UP(NR_CPUS, prom_count_smt_threads());
			prom_printf("Max number of cores passed to firmware: %lu (NR_CPUS = %lu)\n",
				    *cores, NR_CPUS);
		}

		/* try calling the ibm,client-architecture-support method */
		prom_printf("Calling ibm,client-architecture-support...");
		if (call_prom_ret("call-method", 3, 2, &ret,
				  ADDR("ibm,client-architecture-support"),
				  root,
				  ADDR(ibm_architecture_vec)) == 0) {
			/* the call exists... */
			if (ret)
				prom_printf("\nWARNING: ibm,client-architecture"
					    "-support call FAILED!\n");
			call_prom("close", 1, 0, root);
			prom_printf(" done\n");
			return;
		}
		call_prom("close", 1, 0, root);
		prom_printf(" not implemented\n");
	}

	/* no ibm,client-architecture-support call, try the old way */
	elfloader = call_prom("open", 1, 1, ADDR("/packages/elf-loader"));
	if (elfloader == 0) {
		prom_printf("couldn't open /packages/elf-loader\n");
		return;
	}
	call_prom("call-method", 3, 1, ADDR("process-elf-header"),
			elfloader, ADDR(&fake_elf));
	call_prom("close", 1, 0, elfloader);
}
#endif

/*
 * Memory allocation strategy... our layout is normally:
 *
 *  at 14Mb or more we have vmlinux, then a gap and initrd.  In some
 *  rare cases, initrd might end up being 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 this 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.
 *
 *  Note that calls to reserve_mem have to be done explicitly, 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 base = alloc_bottom;
	unsigned long addr = 0;

	if (align)
		base = _ALIGN_UP(base, align);
	prom_debug("alloc_up(%x, %x)\n", size, align);
	if (ram_top == 0)
		prom_panic("alloc_up() called with mem not initialized\n");

	if (align)
		base = _ALIGN_UP(alloc_bottom, align);
	else
		base = alloc_bottom;

	for(; (base + size) <= 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 (addr != PROM_ERROR && addr != 0)
			break;
		addr = 0;
		if (align == 0)
			break;
	}
	if (addr == 0)
		return 0;
	alloc_bottom = addr + size;

	prom_debug(" -> %x\n", addr);
	prom_debug("  alloc_bottom : %x\n", alloc_bottom);
	prom_debug("  alloc_top    : %x\n", alloc_top);
	prom_debug("  alloc_top_hi : %x\n", alloc_top_high);
	prom_debug("  rmo_top      : %x\n", rmo_top);
	prom_debug("  ram_top      : %x\n", ram_top);

	return addr;
}

/*
 * Allocates memory downward, either from top of RMO, or if highmem
 * is set, from the top of RAM.  Note that this one doesn't handle
 * failures.  It does claim memory if highmem is not set.
 */
static unsigned long __init alloc_down(unsigned long size, unsigned long align,
				       int highmem)
{
	unsigned long base, addr = 0;

	prom_debug("alloc_down(%x, %x, %s)\n", size, align,
		   highmem ? "(high)" : "(low)");
	if (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(alloc_top_high - size, align);
		if (addr <= alloc_bottom)
			return 0;
		/* Will we bump into the RMO ? If yes, check out that we