No bzImage target for MIPS

No bzImage target for MIPS

[Krishna Kondaka]

Hi,

	There is no "bzImage" target in the mips linux Makefiles. Could
	you tell me why this is missing? If I want to generate a 
	compressed image for linux on MIPS, how do I do it?

Thanks
Krishna

RE: No bzImage target for MIPS

[Marc Karasek]

The support for decompression of a bzImage image is not supported in the
MIPS kernel.  If you do just a "make zImage" I believe this will get you a
"zipped" image file.    

-----Original Message-----
From: Krishna Kondaka
To: linux-mips@oss.sgi.com
Sent: 12/13/01 7:47 PM
Subject: No bzImage target for MIPS

Hi,

	There is no "bzImage" target in the mips linux Makefiles. Could
	you tell me why this is missing? If I want to generate a 
	compressed image for linux on MIPS, how do I do it?

Thanks
Krishna

Re: No bzImage target for MIPS

[Pete Popov]

On Thu, 2001-12-13 at 16:47, Krishna Kondaka wrote:
> Hi,
> 
> 	There is no "bzImage" target in the mips linux Makefiles. Could
> 	you tell me why this is missing? If I want to generate a 
> 	compressed image for linux on MIPS, how do I do it?

Download the linux-mips drop-in tree from sourceforge.net. Take a look
at arch/mips/zboot.  The only zImage support in that directory is for
the Alchemy Pb1000 board.  That's one way to do it and I basically
copied it from the ppc tree. You'll have to do some work to add support
for whatever board you want, but the example should be sufficient, if
you're a developer. If you're a casual user, unfortunately you're out of
luck.

Pete

Re: No bzImage target for MIPS

[Geoffrey Espin]

> > 	There is no "bzImage" target in the mips linux Makefiles. Could
> > 	you tell me why this is missing? If I want to generate a 
> > 	compressed image for linux on MIPS, how do I do it?

I guess Bill Gates was right, he practically created the Open
Source movement by "providing" an excellent BIOS.  </sarcasm>

One of Embedded Linux's unspoken "problems" is lack of truly
universal boot program.  GRUB is as univeral as x86, not!
Ppcboot is one of the best, but notice the 'ppc' in the name?
I gather that PMON is used by many in MIPS... but that is another
secret you won't find revealed, even in Jun Sun's execellent MIPS
Porting Guide. :-)

COming from a certain proprietary RTOS, I believe, using the OS
itself, as the "boot monitor" is best approach.  All reusable
code, fastest bootstrap (i.e. no 2nd stage loading), etc... But,
of course, not optimal for size, though my image is 1.3M which
includes a 580K cramfs ramdisk.

> Download the linux-mips drop-in tree from sourceforge.net. Take a look
> at arch/mips/zboot.  The only zImage support in that directory is for
> the Alchemy Pb1000 board.  That's one way to do it and I basically

I finally did rework this.  I have a single 360 line "<board>/misc.c" file,
a <board>/Boot.make included by arch/linux/boot/Makefile, and <board>/boot.S.
There is no cut & pasting of code elsewhere in the linux tree, there are
no new directories (zboot, compressed, blah-blah).  Its just glue that
sticks in front of vmlinux.bin.gz.  Where <board> is Korva (AKA Markham),
for now.  Support for compressed, or not, plus linking a kernel to run
at low memory or high.  Running at 'high' allows the kernel to load
another kernel into 'low' memory without burning flash.  All selectable
from "make menuconfig".

Keeping up with sf.net is one of my failings... so no patch, sorry.
Attached are the key files, if anyone thinks this is worth pursuing.

Geoff
-- 
Geoffrey Espin espin@idiom.com
-- 

=misc.c=========================================================================
/*
 * For Korva/Markham MIPS by Geoffrey Espin espin@idiom.com
 * Based on numerous predecessors including...
 *
 * Author: Johnnie Peters <jpeters@mvista.com>
 * Editor: Tom Rini <trini@mvista.com>
 * Ported by Pete Popov <ppopov@mvista.com> to
 * Copyright 2000-2001 MontaVista Software Inc.
 *
 * This program is free software; you can redistribute  it and/or modify it
 * under  the terms of  the GNU General  Public License as published by the
 * Free Software Foundation;  either version 2 of the  License, or (at your
 * option) any later version.
 *
 * THIS  SOFTWARE  IS PROVIDED   ``AS  IS'' AND   ANY  EXPRESS OR   IMPLIED
 * WARRANTIES,   INCLUDING, BUT NOT  LIMITED  TO, THE IMPLIED WARRANTIES OF
 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN
 * NO  EVENT  SHALL   THE AUTHOR  BE    LIABLE FOR ANY   DIRECT,  INDIRECT,
 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 * NOT LIMITED   TO, PROCUREMENT OF  SUBSTITUTE GOODS  OR SERVICES; LOSS OF
 * USE, DATA,  OR PROFITS; OR  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
 * ANY THEORY OF LIABILITY, WHETHER IN  CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * You should have received a copy of the  GNU General Public License along
 * with this program; if not, write  to the Free Software Foundation, Inc.,
 * 675 Mass Ave, Cambridge, MA 02139, USA.
 */

#include <linux/config.h>

#include <stdarg.h>
#include <asm/page.h>

#include "../../../fs/jffs2/zlib.c" /**/
#include "../../../lib/ctype.c"
#undef __HAVE_ARCH_MEMCPY
#undef __HAVE_ARCH_MEMSET
#include "../../../lib/string.c"
#define CONFIG_REMOTE_DEBUG
#include "../korva/dbg_io.c"

extern char _end[];		/* end of bss */

int printk(const char *fmt, ...);
void putc(const char c);

void gunzip(void *, int, unsigned char *, int *);

/* NEC KORVA/MARKHAM */
#define	DELAY(n)	{ volatile int x,y; for (x=0;x<100000*(n);x++) y=x; }
#define	LED(n)		*(volatile unsigned char*)0xb000001c = (n)

#define	ZIMAGE_OFFSET	0x00010000	/* 64k per Makefile/VMA/vrboot */
#define	ZIMAGE_SIZE	0x00200000	/* 2M */

#define	AVAIL_RAM_START	(PAGE_ALIGN((int)_end))	/* scratch area after image */
#define	AVAIL_RAM_SIZE	0x00100000	/* 1M */

char *avail_ram;		/* used by zalloc */
char *end_avail;

void
decompress_kernel(void)
{
	char *zimage_start;
	int zimage_size;
	int i;

	for (i = 0; i < 8; i++) {
		LED(1 << (7 - i));
		DELAY(5);
	}

	/* fix for dbg_io.c on Korva/Markham */
	*((volatile uint8 *) (0xb0000000 + 0)) |= 0x8;	/* GMR = 0x8 UCSEL (uart) */
	avail_ram = (char *) AVAIL_RAM_START;	/* used by zalloc */
	end_avail = (char *) (AVAIL_RAM_START + AVAIL_RAM_SIZE);

	zimage_start = (char *) (FLASHADDR + ZIMAGE_OFFSET);
	zimage_size = ZIMAGE_SIZE;

	printk("\nUncompressing from %#x to %#x\n", zimage_start, LOADADDR);

#if	0
	printk("zImage [0-4]:  %#x %#x %#x %#x\n",
	       ((unsigned long *) zimage_start)[0],
	       ((unsigned long *) zimage_start)[1],
	       ((unsigned long *) zimage_start)[2],
	       ((unsigned long *) zimage_start)[3]);
#endif

	gunzip((void *) LOADADDR, 0x00400000, zimage_start, &zimage_size);

#if	0
	printk("entry  [0-4]:  %#x %#x %#x %#x (at %#x size %d)\n",
	       ((unsigned long *) KERNEL_ENTRY)[0],
	       ((unsigned long *) KERNEL_ENTRY)[1],
	       ((unsigned long *) KERNEL_ENTRY)[2],
	       ((unsigned long *) KERNEL_ENTRY)[3], KERNEL_ENTRY, zimage_size);
#endif

	printk("Booting Linux at %#x\n", KERNEL_ENTRY);

	for (i = 0; i < 8; i++) {
		LED(1 << (i));
		DELAY(5);
	}

	/* XXX returning and letting boot.S jump used to work? */
	if (1) {
		__asm__ __volatile__("jr\t%0"::"r"(KERNEL_ENTRY));
	}
}

void
exit(void)
{
	printk("\nexit\n");
	while (1) ;
}

void
putc(const char c)
{
#define serial_putc(ch) putDebugChar(ch)

	serial_putc(c);
	if (c == '\n')
		serial_putc('\r');
}

void *
zalloc(void *x, unsigned items, unsigned size)
{
	void *p = avail_ram;

	size *= items;
	size = (size + 7) & -8;
	avail_ram += size;
	if (avail_ram > end_avail) {
		printk("oops... zalloc(%#x, %d, %#x) out of memory\n",
		       (int) x, items, size);
		p = NULL;
	}
	return p;
}

void
zfree(void *x, void *addr)
{
}

void
gunzip(void *dst, int dstlen, unsigned char *src, int *lenp)
{
#define HEAD_CRC	2
#define EXTRA_FIELD	4
#define ORIG_NAME	8
#define COMMENT		0x10
#define RESERVED	0xe0

#define DEFLATED	8

	z_stream s;
	int r, i, flags;

	/* skip header */
	i = 10;
	flags = src[3];
	if (src[2] != DEFLATED || (flags & RESERVED) != 0) {
		printk("bad gzipped data\n");
		exit();
	}
	if ((flags & EXTRA_FIELD) != 0)
		i = 12 + src[10] + (src[11] << 8);
	if ((flags & ORIG_NAME) != 0)
		while (src[i++] != 0) ;
	if ((flags & COMMENT) != 0)
		while (src[i++] != 0) ;
	if ((flags & HEAD_CRC) != 0)
		i += 2;
	if (i >= *lenp) {
		printk("gunzip: ran out of data in header\n");
		exit();
	}

	s.zalloc = zalloc;
	s.zfree = zfree;
	r = inflateInit2(&s, -MAX_WBITS);
	if (r != Z_OK) {
		printk("inflateInit2 returned %#x\n", r);
		exit();
	}
	s.next_in = src + i;
	s.avail_in = *lenp - i;
	s.next_out = dst;
	s.avail_out = dstlen;
	r = inflate(&s, Z_FINISH);
	if (r != Z_OK && r != Z_STREAM_END) {
		printk("inflate returned %#x\n", r);
		exit();
	}
	*lenp = s.next_out - (unsigned char *) dst;
	inflateEnd(&s);
}

int
_cvt(unsigned long val, char *buf, long radix, char *digits)
{
	char temp[80];
	char *cp = temp;
	int length = 0;
	if (val == 0) {		/* Special case */
		*cp++ = '0';
	} else
		while (val) {
			*cp++ = digits[val % radix];
			val /= radix;
		}
	while (cp != temp) {
		*buf++ = *--cp;
		length++;
	}
	*buf = '\0';
	return (length);
}

#define is_digit(c) ((c >= '0') && (c <= '9'))

void
_vprintk(void (*putc) (const char), const char *fmt0, va_list ap)
{
	char c, sign, *cp = 0;
	int left_prec, right_prec, zero_fill, length = 0, pad, pad_on_right;
	char buf[32];
	long val;

	while ((c = *fmt0++)) {
		if (c == '%') {
			c = *fmt0++;
			left_prec = right_prec = pad_on_right = 0;
			if (c == '#')
				c = *fmt0++;	/* ignore alternate fmt */

			if (c == '-') {
				c = *fmt0++;
				pad_on_right++;
			}
			if (c == '0') {
				zero_fill = 1;
				c = *fmt0++;
			} else {
				zero_fill = 0;
			}
			while (is_digit(c)) {
				left_prec = (left_prec * 10) + (c - '0');
				c = *fmt0++;
			}
			if (c == '.') {
				c = *fmt0++;
				zero_fill++;
				while (is_digit(c)) {
					right_prec =
					    (right_prec * 10) + (c - '0');
					c = *fmt0++;
				}
			} else {
				right_prec = left_prec;
			}

			sign = '\0';
			switch (c) {
			case 'd':
			case 'x':
			case 'X':
				val = va_arg(ap, long);
				switch (c) {
				case 'd':
					if (val < 0) {
						sign = '-';
						val = -val;
					}
					length =
					    _cvt(val, buf, 10, "0123456789");
					break;
				case 'x':
					length =
					    _cvt(val, buf, 16,
						 "0123456789abcdef");
					break;
				case 'X':
					length =
					    _cvt(val, buf, 16,
						 "0123456789ABCDEF");
					break;
				}
				cp = buf;
				break;
			case 's':
				cp = va_arg(ap, char *);
				length = strlen(cp);
				break;
			case 'c':
				c = va_arg(ap, long /*char */ );
				(*putc) (c);
				continue;
			default:
				(*putc) ('?');
			}
			pad = left_prec - length;
			if (sign != '\0') {
				pad--;
			}
			if (zero_fill) {
				c = '0';
				if (sign != '\0') {
					(*putc) (sign);
					sign = '\0';
				}
			} else {
				c = ' ';
			}
			if (!pad_on_right) {
				while (pad-- > 0) {
					(*putc) (c);
				}
			}
			if (sign != '\0') {
				(*putc) (sign);
			}
			while (length-- > 0) {
				(*putc) (c = *cp++);
				if (c == '\n') {
					(*putc) ('\r');
				}
			}
			if (pad_on_right) {
				while (pad-- > 0) {
					(*putc) (c);
				}
			}
		} else {
			(*putc) (c);
			if (c == '\n') {
				(*putc) ('\r');
			}
		}
	}
}

int
printk(const char *fmt, ...)
{
	va_list args;
	va_start(args, fmt);
	_vprintk(putc, fmt, args);
	va_end(args);
	return 0;
}

=boot.S=========================================================================
/*
 * boot.S
 *
 * Modified by Geoffrey Espin <espin@idiom.com> for NEC Electronics
 *
 * Copyright (C) 1999 Bradley D. LaRonde
 * Copyright (C) 2001 Dai Itasaka <Daisaku_Itasaka@el.nec.com>
 *
 * This file is subject to the terms and conditions of the GNU General Public
 * License.  See the file "COPYING" in the main directory of this archive
 * for more details.
 */

#include <linux/config.h>

#include <asm/asm.h>
#include <asm/regdef.h>
#include <asm/addrspace.h>
#include <asm/mipsregs.h>

#define	FLASH_BEGIN	(FLASHADDR+0x00010000) // skip first 64k (this program)
#define	FLASH_END	(FLASHADDR+0x00400000) // 4M; see Boot.make


/*
EC:7(CPU clock / 1), EP:0(1word/1cycle), AD:0(VR4000 compatible)
M16:0(Disable), BE:1(Big endian), CS:1(+10), IC:4(16K), DC:3(8K)
K0:2(Disable)

31  30 28 27 24  23 22 21  19  18  17  16  15 14 13  12 11  9 8  6 5 3 2  0
+---+-----+-----+---+-----+---+---+---+---+---+-----+---+-----+----+---+----+
| 0 | EC  |  EP | AD|  0  |M16| 0 | 1 | 0 | BE| 1 0 | CS| IC  | DC | 0 | K0 |
+---+-----+-----+---+-----+---+---+---+---+---+-----+---+-----+----+---+----+
EC : Tclock rate,  EP : Transfer data pattern setting,
AD : Accelerate data mode,  M16 : MIPS16 ISA mode enable
BE : BigEndian, CS : Cache Size mode Indication(fixed to 1)
IC : Instruction Cache Size,  DC : Data Cache Size,
K0 : kseg0 coherency algorithm

See uPD98501 Preliminary User's Manual, Chapter 2, page 156.
*/

#define KORVA_CONFIG_INIT_VAL_BE	0x7002d8c2   /*BE*/
#define KORVA_CONFIG_INIT_VAL	0x700258c2   /*LE*/


#define DEBUG
#ifdef DEBUG
#define DEBUG_LED      0xb000001c
#define CHECKPOINT(n)  li t4, n; sw t4, DEBUG_LED;
#else
#define CHECKPOINT(n)
#endif

LEAF(__start)
	.set	mips3
	.set	noreorder

	nop  // patched by script if using Kyoto-Micro ROM Emulator
	nop
	nop

	mtc0	zero, CP0_STATUS	// disable interrupts
	mtc0	zero, CP0_CAUSE		// clear interrupts
	li	t0, KORVA_CONFIG_INIT_VAL
	mtc0	t0, CP0_CONFIG		// make sure the cache is disabled
	mtc0	zero, CP0_WATCHLO
	mtc0	zero, CP0_WATCHHI	// disable watch exception

	CHECKPOINT(0xB1)

	/* setup_hardware (was function, now PIC as may be linked at RAM) */

	// check if the CPU is running at 66MHz or 100MHz
	lui	t0, 0xb000
	lw	t1,	4(t0)	// 0xb0000004 = GSR
	li	t0,	2	// CCLKSEL
	and	t0, t0, t1
	beq	zero, t0, clk100
	lui	t0, 0xb000

clk66:
	li	t1,	0x120	// SDMDR
	sw	t1, 0x108(t0)	// *(0xb0000108) = 0x120;
	li	t1,	0x000	// SDPTR
	sw	t1, 0x110(t0)	// *(0xb0000110) = 0x000;

	beq zero, zero, 1f
	nop

clk100:
	li	t1,	0x230	// SDMDR
	sw	t1, 0x108(t0)	// *(0xb0000108) = 0x230;
	li	t1,	0x021	// SDPTR
	sw	t1, 0x110(t0)	// *(0xb0000110) = 0x021;

1:
#if defined(CONFIG_NEC_MARKHAM_32MB)
	li	t1, 0x3c2	// SDTSR 32M
				// CAB=0x02, RAB=0x40 BTM=0x0080 SIZE=0x300(32M)
#else
	li	t1, 0x1a1	// SDTSR  8M
				// CAB=0x01, RAB=0x20 BTM=0x0080 SIZE=0x100(8M)
#endif
	sw	t1, 0x10c(t0)	// *(0xb000010c) = 0x1a1/0x3c2;
	
				// allow writing to flash parts
	li	t1,	0x100	// RMMDR (00 4MB mode) WM=1 unmasked
	sw	t1, 0x100(t0)	// *(0xb0000100) = 0x000;

	li	t1,	0x003	// RMATR   FAT_8
	sw	t1, 0x104(t0)	// *(0xb0000104) = 0x003;

	li	t1,	0x000	// IMR
	sw	t1, 0x00c(t0)	// disable all INTs

	nop
	/* end of setup_hardware */

	CHECKPOINT(0xB2)

	/* invalidate_cache (was function, now PIC as may be linked at RAM) */ 

	mtc0	zero, CP0_TAGLO	// invalidate the instruction cache
	li	s0, 0x80000000
	li	s1, 0x80000000 + (1 << 14) // 16k
1:
	cache	(2 << 2) | 0, (s0)
	add	s0, 0x10
	bne	s0, s1, 1b
	nop

	mtc0	zero, CP0_TAGLO	// invalidate the data cache
	li	s0, 0x80000000
	li	s1, 0x80000000 + (1 << 13) // 8k
1:
	cache	(2 << 2) | 1, (s0)
	add	s0, 0x10
	bne	s0, s1, 1b
	nop

	nop
	/* end of invalidate_cache */

	CHECKPOINT(0xB3)

	mfc0	t0, CP0_CONFIG	// enable kseg0 caching
	ori	t0, 0x1	// cached
	mtc0	t0, CP0_CONFIG

	CHECKPOINT(0xB4)

#ifdef	CONFIG_NEC_KORVA_COMPRESSED
relocate:
	la	sp, .stack
	move	s0, a0
	move	s1, a1
	move	s2, a2
	move	s3, a3

	la	a0, __start

	li	a1, FLASHADDR
	la	a2, _edata
	subu	t1, a2, a0
	srl	t1, t1, 2

	/* copy text section */
	li	t0, 0
1:	lw	v0, 0(a1)
	nop
	sw	v0, 0(a0)
	xor	t0, t0, v0
	addu	a0, 4
	bne	a2, a0, 1b
	addu	a1, 4

	/* clear BSS */
	la	a0, _edata
	la	a2, _end
2:	sw	zero, 0(a0)
	bne	a2, a0, 2b
	addu	a0, 4

	/* now that code is in RAM we can jump okay */

	// decompress kernel to ram
	la      ra, 1f
	la      t4, decompress_kernel	// elsewhere in 'C'
	jr      t4
#else
	/* actually we were linked at ROM so we're okay */

	// copy kernel to ram
	la      ra, 1f
	la      t4, copy_kernel
	jr      t4
#endif
	nop
1:

	CHECKPOINT(0xB5)

	// jump to kernel entry point
	li	t4, KERNEL_ENTRY
	jr	t4
	nop

	nop
END(__start)


LEAF(copy_kernel)
	// copy kernel from ROM to RAM

	// Load s0 = src, s1 = src_end, s2 = dest.
	li	s0, FLASH_BEGIN
	li	s1, FLASH_END

	li	s2, LOADADDR

	sw	zero, (s2)
	sw	zero, 4(s2)
	sw	zero, 8(s2)
	sw	zero, 12(s2)

1:
	// Only update LED once in a while.
	and	t0, s0, 0xffff
	bne	t0, zero, 2f

	// Show copy progress on LED.
	srl	t0, s0, 16
	sh	t0, DEBUG_LED

2:
	// Copy src to dest.
	lw	t0, (s0)
	bne	t0, zero, 3f
	nop
	sw	zero, (s2)
	beq	zero, zero, 4f
	nop
3:
	sw	t0, (s2)
4:
	// Increment src and dest.
	add	s0, 4
	add	s2, 4

	// Loop if still more to go.
	bne	s0, s1, 1b
	nop

	jr	ra
	 nop
END(copy_kernel)

#ifdef	CONFIG_NEC_KORVA_COMPRESSED
	.comm .stack,4096*2,4	// used by relocate to call 'C' program
#endif

=Boot.make======================================================================
# Boot.make - used by ../boot/Makefile

OBJCOPY         = mipsel-linux-objcopy
OBJCOPY_FLAGS   = -R .stab -R .stabstr -R .comment -R .note -R .reginfo -R .mdebug
OBJDUMP         = mipsel-linux-objdump
#OBJDUMP_HFLAGS  = --disassemble-all -b srec -m  mips:4600 -EL

TOPDIR          = ../../..

ifeq ($(LINUX_SRC),)
LINUX_SRC=$(TOPDIR)
endif
VMLINUX=$(LINUX_SRC)/vmlinux
SYSTEM_MAP=$(LINUX_SRC)/System.map

ifeq ($(USERLAND),)
USERLAND=$(LINUX_SRC)/../userland
endif

FLASHADDR      = 0xbfc00000

# see arch/mips/Makefile LOADADDR
# 0x10000  bottom 64k reserved by/for boot.S (FLASH_BEGIN)
# 0x02000  additional 8k reserved RAM image vector table (see LOADADDR)

ifdef CONFIG_NEC_MARKHAM_BOOTROM
NN=10#in hex is 16M
LOADADDR      += 0x81002000
ZLOADADDR     += 0x81402000# +4M
VMA            = 0x7f00e000 # 0x0-0x81000000+0x10000-0x2000
ZVMA           = 0x7ebfe000 # 0x0-0x81402000
else
NN=00
LOADADDR      += 0x80002000
ZLOADADDR     += 0x80402000# +4M
VMA            = 0x8000e000 # 0x0-0x80000000+0x10000-0x2000
ZVMA           = 0x7fbfe000 # 0x0-0x80402000
endif

KERNEL_ENTRY=$(shell awk '/kernel_entry/{print "0x" $$1}' $(LINUX_SRC)/System.map)

override CFLAGS += -DKERNEL_ENTRY=$(KERNEL_ENTRY) -DNN=$(NN)
override CFLAGS += -DLOADADDR=$(LOADADDR) -DFLASHADDR=$(FLASHADDR) -DZLOADADDR=$(ZLOADADDR)

ifdef CONFIG_NEC_KORVA_COMPRESSED
zbootroms: bin zsrec
else
bootroms: bin srec
endif

$(TOPDIR)/arch/$(ARCH)/ld.script.$(ZLOADADDR)_z: $(TOPDIR)/arch/$(ARCH)/ld.script.in 
	sed -e 's/kernel_entry/__start/' -e 's/@@LOADADDR@@/$(ZLOADADDR)/' <$< >$@

ZLINKFLAGS += -T $(TOPDIR)/arch/$(ARCH)/ld.script.$(ZLOADADDR)_z

misc.c: ../korva/misc.c
	@rm -f $@
	cp $< $@

boot.S: ../korva/boot.S
	@rm -f $@
	cp $< $@

#vrboot$(NN): boot.o 
vrboot$(NN): boot.o misc.o $(TOPDIR)/arch/$(ARCH)/ld.script.$(ZLOADADDR)_z
	@rm -f $@
ifdef CONFIG_NEC_KORVA_COMPRESSED
	$(LD) $(LDFLAGS) -static -EL -N -o $@ $(ZLINKFLAGS) boot.o misc.o
else
	$(LD) $(LDFLAGS) -static -EL -N -o $@ -Ttext $(FLASHADDR) boot.o
endif
	@rm -f $^ ../boot/misc.c ../boot/boot.S 

vmlinux$(NN): $(VMLINUX)
	@rm -f System$(NN).map
	cp -p $(SYSTEM_MAP) System$(NN).map
	cp -p $< $@

# move sections to 0, then add bfc00000 for ROMulator
vrvmlinux$(NN)-0: vmlinux$(NN) vrboot$(NN)
	@rm -f $@
	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-addresses 0 -O binary \
		vrboot$(NN) vrboot$(NN)-0.bin
	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-addresses $(VMA) \
		vmlinux$(NN) vmlinux$(NN)-0
	$(OBJCOPY) $(OBJCOPY_FLAGS) --add-section .vrboot=vrboot$(NN)-0.bin  \
		--set-section-flags .vrboot=,alloc,load,readonly,code \
		vmlinux$(NN)-0 vrvmlinux$(NN)-0

vrvmlinux$(NN)-bfc: vrvmlinux$(NN)-0
	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-address $(FLASHADDR) \
		vrvmlinux$(NN)-0 vrvmlinux$(NN)-bfc

# xxx-bfc.bin is for 'reload' & 'reburn' used on a running system
vrvmlinux$(NN)-bfc.bin: vrvmlinux$(NN)-0
	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-address $(FLASHADDR) -O binary \
		vrvmlinux$(NN)-0 vrvmlinux$(NN)-bfc.bin

bin: vrvmlinux$(NN)-bfc vrvmlinux$(NN)-bfc.bin

vrvmlinux$(NN)-0.srec: vrvmlinux$(NN)-0
	$(OBJCOPY) $(OBJCOPY_FLAGS) -O srec --srec-forceS3 --gap-fill 0 \
		vrvmlinux$(NN)-0 vrvmlinux$(NN)-0.srec
	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-address $(FLASHADDR) -O srec \
		vrvmlinux$(NN)-0 vrvmlinux$(NN)-bfc.srec

srec: vrvmlinux$(NN)-0.srec

# used by 'reburn' script for Korva (8M) which needs small 256K chunks
split:
	@rm -rf split
	@mkdir split
	cd split && split -b 256k ../vrvmlinux00-bfc.bin vrvm

# compressed...

vmlinux$(NN).bin.gz: vmlinux$(NN)
	$(OBJCOPY) -S -O binary $< $<.bin
	gzip -vf $<.bin

zvrvmlinux$(NN)-0: vmlinux$(NN).bin.gz vrboot$(NN)
	@rm -f $@
	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-addresses=$(ZVMA) \
		vrboot$(NN) zvrboot$(NN)-0
	$(OBJCOPY) $(OBJCOPY_FLAGS) --add-section .vmlinux=vmlinux$(NN).bin.gz \
		--change-section-address .vmlinux=0x10000 \
		--set-section-flags .vmlinux=,alloc,load,readonly,code \
		zvrboot$(NN)-0 zvrvmlinux$(NN)-0

zvrvmlinux$(NN)-0.srec: zvrvmlinux$(NN)-0
	$(OBJCOPY) $(OBJCOPY_FLAGS) -O srec --srec-forceS3 \
		zvrvmlinux$(NN)-0 zvrvmlinux$(NN)-0.srec
	$(OBJCOPY) $(OBJCOPY_FLAGS) -O srec --srec-forceS3 \
		--change-address 0xbfc00000 \
		zvrvmlinux$(NN)-0 zvrvmlinux$(NN)-bfc.srec

zsrec: zvrvmlinux$(NN)-0.srec

clean::
	@rm -f System* vrvmlinux* vmlinux* vrboot* zvrvmlinux*
================================================================================

Re: No bzImage target for MIPS

[Keith Owens]

On Thu, 13 Dec 2001 19:28:46 -0800, 
Geoffrey Espin <espin@idiom.com> wrote:
>=misc.c=========================================================================
>#include "../../../fs/jffs2/zlib.c" /**/
>#include "../../../lib/ctype.c"

I am phasing out the practice of ../ in kernel include paths.  It is
much better to do

#include "zlib.c"
#include "ctype.c"

and the Makefile adds -I$(TOPDIR)/fs/jffs2 -I$(TOPDIR)/lib.  Then when
sources are moved from one directory to another, the source does not
change, only the Makefile.  Relative paths are a pain in the neck in
Makefiles, they are even more of a pain in source code.

>TOPDIR          = ../../..

TOPDIR := $(shell cd ../../..; /bin/pwd)

is better, it returns an absolute path instead of a relative one.

Keith Owens, kernel build maintainer

Re: No bzImage target for MIPS

[Geoffrey Espin]

Keith,

On Fri, Dec 14, 2001 at 02:51:21PM +1100, Keith Owens wrote:
> >#include "../../../fs/jffs2/zlib.c" /**/
> >#include "../../../lib/ctype.c"
> I am phasing out the practice of ../ in kernel include paths.  It is
> much better to do
> #include "zlib.c"
> #include "ctype.c"
> and the Makefile adds -I$(TOPDIR)/fs/jffs2 -I$(TOPDIR)/lib.  Then when

I'll check around the tree.  Doesn't matter to me.  Both have crappy
implications.  I hate it when -I's pile up... and you don't know which
directory takes precedence (esp. if there are duplicate names).

> >TOPDIR          = ../../..
> TOPDIR := $(shell cd ../../..; /bin/pwd)
> is better, it returns an absolute path instead of a relative one.

Hmmm, okay.  I probably shouldn't even have this macro.
I had hacked things earlier so I didn't have to run 'make' from linux/.

Thanks for the quick feedback, mate!

[I'll spare someone pointing out that I should have tried to use
linux/lib/vsprintf.c.  Missed it!  What a newbie.]

Geoff
-- 
Geoffrey Espin
espin@idiom.com

Re: No bzImage target for MIPS

[Pete Popov]

On Thu, 2001-12-13 at 19:28, Geoffrey Espin wrote:
> 
> > > 	There is no "bzImage" target in the mips linux Makefiles. Could
> > > 	you tell me why this is missing? If I want to generate a 
> > > 	compressed image for linux on MIPS, how do I do it?
> 
> I guess Bill Gates was right, he practically created the Open
> Source movement by "providing" an excellent BIOS.  </sarcasm>
> 
> One of Embedded Linux's unspoken "problems" is lack of truly
> universal boot program.  GRUB is as univeral as x86, not!
> Ppcboot is one of the best, but notice the 'ppc' in the name?
> I gather that PMON is used by many in MIPS... but that is another
> secret you won't find revealed, even in Jun Sun's execellent MIPS
> Porting Guide. :-)

The "original" pmon has outlived its purpose, although it's still found
on many boards. I've become a fan of MIPS Tech's yamon; it's small, it
works, it has just the features I need in an embedded system, and, did I
say it works?.  Unfortunately the code is released under NDA only. If
MIPS opened the code, I think you would see it on many more boards.
 
> COming from a certain proprietary RTOS, I believe, using the OS
> itself, as the "boot monitor" is best approach.  All reusable
> code, fastest bootstrap (i.e. no 2nd stage loading), etc... But,
> of course, not optimal for size, though my image is 1.3M which
> includes a 580K cramfs ramdisk.

That's fine once you've ported the kernel to your board. What do you do
before that point?

> > Download the linux-mips drop-in tree from sourceforge.net. Take a look
> > at arch/mips/zboot.  The only zImage support in that directory is for
> > the Alchemy Pb1000 board.  That's one way to do it and I basically
> 
> I finally did rework this.  I have a single 360 line "<board>/misc.c" file,
> a <board>/Boot.make included by arch/linux/boot/Makefile, and <board>/boot.S.
> There is no cut & pasting of code elsewhere in the linux tree, there are
> no new directories (zboot, compressed, blah-blah).  Its just glue that
> sticks in front of vmlinux.bin.gz.  Where <board> is Korva (AKA Markham),
> for now.  Support for compressed, or not, plus linking a kernel to run
> at low memory or high.  Running at 'high' allows the kernel to load
> another kernel into 'low' memory without burning flash.  All selectable
> from "make menuconfig".
> 
> Keeping up with sf.net is one of my failings... so no patch, sorry.
> Attached are the key files, if anyone thinks this is worth pursuing.

That's OK, I'm sure someone with interest in the Korva can create the
patch.  But I fail to see how having this code in every <board>
directory is better than having a separate boot directory like the ppc
arch.  The ppc-like zImage support I added is a bit different from the
patch below. It assumes that you do have a boot loader and the 'kernel
loader' is basically separate from the kernel image itself. Adding
initrd support that's not statically compiled in the kernel should be
pretty easy and I like that approach better than statically compiled
initrd images.

Pete

> Geoff
> -- 
> Geoffrey Espin espin@idiom.com
> -- 
> 
> =misc.c=========================================================================
> /*
>  * For Korva/Markham MIPS by Geoffrey Espin espin@idiom.com
>  * Based on numerous predecessors including...
>  *
>  * Author: Johnnie Peters <jpeters@mvista.com>
>  * Editor: Tom Rini <trini@mvista.com>
>  * Ported by Pete Popov <ppopov@mvista.com> to
>  * Copyright 2000-2001 MontaVista Software Inc.
>  *
>  * This program is free software; you can redistribute  it and/or modify it
>  * under  the terms of  the GNU General  Public License as published by the
>  * Free Software Foundation;  either version 2 of the  License, or (at your
>  * option) any later version.
>  *
>  * THIS  SOFTWARE  IS PROVIDED   ``AS  IS'' AND   ANY  EXPRESS OR   IMPLIED
>  * WARRANTIES,   INCLUDING, BUT NOT  LIMITED  TO, THE IMPLIED WARRANTIES OF
>  * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN
>  * NO  EVENT  SHALL   THE AUTHOR  BE    LIABLE FOR ANY   DIRECT,  INDIRECT,
>  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
>  * NOT LIMITED   TO, PROCUREMENT OF  SUBSTITUTE GOODS  OR SERVICES; LOSS OF
>  * USE, DATA,  OR PROFITS; OR  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
>  * ANY THEORY OF LIABILITY, WHETHER IN  CONTRACT, STRICT LIABILITY, OR TORT
>  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
>  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
>  *
>  * You should have received a copy of the  GNU General Public License along
>  * with this program; if not, write  to the Free Software Foundation, Inc.,
>  * 675 Mass Ave, Cambridge, MA 02139, USA.
>  */
> 
> #include <linux/config.h>
> 
> #include <stdarg.h>
> #include <asm/page.h>
> 
> #include "../../../fs/jffs2/zlib.c" /**/
> #include "../../../lib/ctype.c"
> #undef __HAVE_ARCH_MEMCPY
> #undef __HAVE_ARCH_MEMSET
> #include "../../../lib/string.c"
> #define CONFIG_REMOTE_DEBUG
> #include "../korva/dbg_io.c"
> 
> extern char _end[];		/* end of bss */
> 
> int printk(const char *fmt, ...);
> void putc(const char c);
> 
> void gunzip(void *, int, unsigned char *, int *);
> 
> /* NEC KORVA/MARKHAM */
> #define	DELAY(n)	{ volatile int x,y; for (x=0;x<100000*(n);x++) y=x; }
> #define	LED(n)		*(volatile unsigned char*)0xb000001c = (n)
> 
> #define	ZIMAGE_OFFSET	0x00010000	/* 64k per Makefile/VMA/vrboot */
> #define	ZIMAGE_SIZE	0x00200000	/* 2M */
> 
> #define	AVAIL_RAM_START	(PAGE_ALIGN((int)_end))	/* scratch area after image */
> #define	AVAIL_RAM_SIZE	0x00100000	/* 1M */
> 
> char *avail_ram;		/* used by zalloc */
> char *end_avail;
> 
> void
> decompress_kernel(void)
> {
> 	char *zimage_start;
> 	int zimage_size;
> 	int i;
> 
> 	for (i = 0; i < 8; i++) {
> 		LED(1 << (7 - i));
> 		DELAY(5);
> 	}
> 
> 	/* fix for dbg_io.c on Korva/Markham */
> 	*((volatile uint8 *) (0xb0000000 + 0)) |= 0x8;	/* GMR = 0x8 UCSEL (uart) */
> 	avail_ram = (char *) AVAIL_RAM_START;	/* used by zalloc */
> 	end_avail = (char *) (AVAIL_RAM_START + AVAIL_RAM_SIZE);
> 
> 	zimage_start = (char *) (FLASHADDR + ZIMAGE_OFFSET);
> 	zimage_size = ZIMAGE_SIZE;
> 
> 	printk("\nUncompressing from %#x to %#x\n", zimage_start, LOADADDR);
> 
> #if	0
> 	printk("zImage [0-4]:  %#x %#x %#x %#x\n",
> 	       ((unsigned long *) zimage_start)[0],
> 	       ((unsigned long *) zimage_start)[1],
> 	       ((unsigned long *) zimage_start)[2],
> 	       ((unsigned long *) zimage_start)[3]);
> #endif
> 
> 	gunzip((void *) LOADADDR, 0x00400000, zimage_start, &zimage_size);
> 
> #if	0
> 	printk("entry  [0-4]:  %#x %#x %#x %#x (at %#x size %d)\n",
> 	       ((unsigned long *) KERNEL_ENTRY)[0],
> 	       ((unsigned long *) KERNEL_ENTRY)[1],
> 	       ((unsigned long *) KERNEL_ENTRY)[2],
> 	       ((unsigned long *) KERNEL_ENTRY)[3], KERNEL_ENTRY, zimage_size);
> #endif
> 
> 	printk("Booting Linux at %#x\n", KERNEL_ENTRY);
> 
> 	for (i = 0; i < 8; i++) {
> 		LED(1 << (i));
> 		DELAY(5);
> 	}
> 
> 	/* XXX returning and letting boot.S jump used to work? */
> 	if (1) {
> 		__asm__ __volatile__("jr\t%0"::"r"(KERNEL_ENTRY));
> 	}
> }
> 
> void
> exit(void)
> {
> 	printk("\nexit\n");
> 	while (1) ;
> }
> 
> void
> putc(const char c)
> {
> #define serial_putc(ch) putDebugChar(ch)
> 
> 	serial_putc(c);
> 	if (c == '\n')
> 		serial_putc('\r');
> }
> 
> void *
> zalloc(void *x, unsigned items, unsigned size)
> {
> 	void *p = avail_ram;
> 
> 	size *= items;
> 	size = (size + 7) & -8;
> 	avail_ram += size;
> 	if (avail_ram > end_avail) {
> 		printk("oops... zalloc(%#x, %d, %#x) out of memory\n",
> 		       (int) x, items, size);
> 		p = NULL;
> 	}
> 	return p;
> }
> 
> void
> zfree(void *x, void *addr)
> {
> }
> 
> void
> gunzip(void *dst, int dstlen, unsigned char *src, int *lenp)
> {
> #define HEAD_CRC	2
> #define EXTRA_FIELD	4
> #define ORIG_NAME	8
> #define COMMENT		0x10
> #define RESERVED	0xe0
> 
> #define DEFLATED	8
> 
> 	z_stream s;
> 	int r, i, flags;
> 
> 	/* skip header */
> 	i = 10;
> 	flags = src[3];
> 	if (src[2] != DEFLATED || (flags & RESERVED) != 0) {
> 		printk("bad gzipped data\n");
> 		exit();
> 	}
> 	if ((flags & EXTRA_FIELD) != 0)
> 		i = 12 + src[10] + (src[11] << 8);
> 	if ((flags & ORIG_NAME) != 0)
> 		while (src[i++] != 0) ;
> 	if ((flags & COMMENT) != 0)
> 		while (src[i++] != 0) ;
> 	if ((flags & HEAD_CRC) != 0)
> 		i += 2;
> 	if (i >= *lenp) {
> 		printk("gunzip: ran out of data in header\n");
> 		exit();
> 	}
> 
> 	s.zalloc = zalloc;
> 	s.zfree = zfree;
> 	r = inflateInit2(&s, -MAX_WBITS);
> 	if (r != Z_OK) {
> 		printk("inflateInit2 returned %#x\n", r);
> 		exit();
> 	}
> 	s.next_in = src + i;
> 	s.avail_in = *lenp - i;
> 	s.next_out = dst;
> 	s.avail_out = dstlen;
> 	r = inflate(&s, Z_FINISH);
> 	if (r != Z_OK && r != Z_STREAM_END) {
> 		printk("inflate returned %#x\n", r);
> 		exit();
> 	}
> 	*lenp = s.next_out - (unsigned char *) dst;
> 	inflateEnd(&s);
> }
> 
> int
> _cvt(unsigned long val, char *buf, long radix, char *digits)
> {
> 	char temp[80];
> 	char *cp = temp;
> 	int length = 0;
> 	if (val == 0) {		/* Special case */
> 		*cp++ = '0';
> 	} else
> 		while (val) {
> 			*cp++ = digits[val % radix];
> 			val /= radix;
> 		}
> 	while (cp != temp) {
> 		*buf++ = *--cp;
> 		length++;
> 	}
> 	*buf = '\0';
> 	return (length);
> }
> 
> #define is_digit(c) ((c >= '0') && (c <= '9'))
> 
> void
> _vprintk(void (*putc) (const char), const char *fmt0, va_list ap)
> {
> 	char c, sign, *cp = 0;
> 	int left_prec, right_prec, zero_fill, length = 0, pad, pad_on_right;
> 	char buf[32];
> 	long val;
> 
> 	while ((c = *fmt0++)) {
> 		if (c == '%') {
> 			c = *fmt0++;
> 			left_prec = right_prec = pad_on_right = 0;
> 			if (c == '#')
> 				c = *fmt0++;	/* ignore alternate fmt */
> 
> 			if (c == '-') {
> 				c = *fmt0++;
> 				pad_on_right++;
> 			}
> 			if (c == '0') {
> 				zero_fill = 1;
> 				c = *fmt0++;
> 			} else {
> 				zero_fill = 0;
> 			}
> 			while (is_digit(c)) {
> 				left_prec = (left_prec * 10) + (c - '0');
> 				c = *fmt0++;
> 			}
> 			if (c == '.') {
> 				c = *fmt0++;
> 				zero_fill++;
> 				while (is_digit(c)) {
> 					right_prec =
> 					    (right_prec * 10) + (c - '0');
> 					c = *fmt0++;
> 				}
> 			} else {
> 				right_prec = left_prec;
> 			}
> 
> 			sign = '\0';
> 			switch (c) {
> 			case 'd':
> 			case 'x':
> 			case 'X':
> 				val = va_arg(ap, long);
> 				switch (c) {
> 				case 'd':
> 					if (val < 0) {
> 						sign = '-';
> 						val = -val;
> 					}
> 					length =
> 					    _cvt(val, buf, 10, "0123456789");
> 					break;
> 				case 'x':
> 					length =
> 					    _cvt(val, buf, 16,
> 						 "0123456789abcdef");
> 					break;
> 				case 'X':
> 					length =
> 					    _cvt(val, buf, 16,
> 						 "0123456789ABCDEF");
> 					break;
> 				}
> 				cp = buf;
> 				break;
> 			case 's':
> 				cp = va_arg(ap, char *);
> 				length = strlen(cp);
> 				break;
> 			case 'c':
> 				c = va_arg(ap, long /*char */ );
> 				(*putc) (c);
> 				continue;
> 			default:
> 				(*putc) ('?');
> 			}
> 			pad = left_prec - length;
> 			if (sign != '\0') {
> 				pad--;
> 			}
> 			if (zero_fill) {
> 				c = '0';
> 				if (sign != '\0') {
> 					(*putc) (sign);
> 					sign = '\0';
> 				}
> 			} else {
> 				c = ' ';
> 			}
> 			if (!pad_on_right) {
> 				while (pad-- > 0) {
> 					(*putc) (c);
> 				}
> 			}
> 			if (sign != '\0') {
> 				(*putc) (sign);
> 			}
> 			while (length-- > 0) {
> 				(*putc) (c = *cp++);
> 				if (c == '\n') {
> 					(*putc) ('\r');
> 				}
> 			}
> 			if (pad_on_right) {
> 				while (pad-- > 0) {
> 					(*putc) (c);
> 				}
> 			}
> 		} else {
> 			(*putc) (c);
> 			if (c == '\n') {
> 				(*putc) ('\r');
> 			}
> 		}
> 	}
> }
> 
> int
> printk(const char *fmt, ...)
> {
> 	va_list args;
> 	va_start(args, fmt);
> 	_vprintk(putc, fmt, args);
> 	va_end(args);
> 	return 0;
> }
> 
> =boot.S=========================================================================
> /*
>  * boot.S
>  *
>  * Modified by Geoffrey Espin <espin@idiom.com> for NEC Electronics
>  *
>  * Copyright (C) 1999 Bradley D. LaRonde
>  * Copyright (C) 2001 Dai Itasaka <Daisaku_Itasaka@el.nec.com>
>  *
>  * This file is subject to the terms and conditions of the GNU General Public
>  * License.  See the file "COPYING" in the main directory of this archive
>  * for more details.
>  */
> 
> #include <linux/config.h>
> 
> #include <asm/asm.h>
> #include <asm/regdef.h>
> #include <asm/addrspace.h>
> #include <asm/mipsregs.h>
> 
> #define	FLASH_BEGIN	(FLASHADDR+0x00010000) // skip first 64k (this program)
> #define	FLASH_END	(FLASHADDR+0x00400000) // 4M; see Boot.make
> 
> 
> /*
> EC:7(CPU clock / 1), EP:0(1word/1cycle), AD:0(VR4000 compatible)
> M16:0(Disable), BE:1(Big endian), CS:1(+10), IC:4(16K), DC:3(8K)
> K0:2(Disable)
> 
> 31  30 28 27 24  23 22 21  19  18  17  16  15 14 13  12 11  9 8  6 5 3 2  0
> +---+-----+-----+---+-----+---+---+---+---+---+-----+---+-----+----+---+----+
> | 0 | EC  |  EP | AD|  0  |M16| 0 | 1 | 0 | BE| 1 0 | CS| IC  | DC | 0 | K0 |
> +---+-----+-----+---+-----+---+---+---+---+---+-----+---+-----+----+---+----+
> EC : Tclock rate,  EP : Transfer data pattern setting,
> AD : Accelerate data mode,  M16 : MIPS16 ISA mode enable
> BE : BigEndian, CS : Cache Size mode Indication(fixed to 1)
> IC : Instruction Cache Size,  DC : Data Cache Size,
> K0 : kseg0 coherency algorithm
> 
> See uPD98501 Preliminary User's Manual, Chapter 2, page 156.
> */
> 
> #define KORVA_CONFIG_INIT_VAL_BE	0x7002d8c2   /*BE*/
> #define KORVA_CONFIG_INIT_VAL	0x700258c2   /*LE*/
> 
> 
> #define DEBUG
> #ifdef DEBUG
> #define DEBUG_LED      0xb000001c
> #define CHECKPOINT(n)  li t4, n; sw t4, DEBUG_LED;
> #else
> #define CHECKPOINT(n)
> #endif
> 
> LEAF(__start)
> 	.set	mips3
> 	.set	noreorder
> 
> 	nop  // patched by script if using Kyoto-Micro ROM Emulator
> 	nop
> 	nop
> 
> 	mtc0	zero, CP0_STATUS	// disable interrupts
> 	mtc0	zero, CP0_CAUSE		// clear interrupts
> 	li	t0, KORVA_CONFIG_INIT_VAL
> 	mtc0	t0, CP0_CONFIG		// make sure the cache is disabled
> 	mtc0	zero, CP0_WATCHLO
> 	mtc0	zero, CP0_WATCHHI	// disable watch exception
> 
> 	CHECKPOINT(0xB1)
> 
> 	/* setup_hardware (was function, now PIC as may be linked at RAM) */
> 
> 	// check if the CPU is running at 66MHz or 100MHz
> 	lui	t0, 0xb000
> 	lw	t1,	4(t0)	// 0xb0000004 = GSR
> 	li	t0,	2	// CCLKSEL
> 	and	t0, t0, t1
> 	beq	zero, t0, clk100
> 	lui	t0, 0xb000
> 
> clk66:
> 	li	t1,	0x120	// SDMDR
> 	sw	t1, 0x108(t0)	// *(0xb0000108) = 0x120;
> 	li	t1,	0x000	// SDPTR
> 	sw	t1, 0x110(t0)	// *(0xb0000110) = 0x000;
> 
> 	beq zero, zero, 1f
> 	nop
> 
> clk100:
> 	li	t1,	0x230	// SDMDR
> 	sw	t1, 0x108(t0)	// *(0xb0000108) = 0x230;
> 	li	t1,	0x021	// SDPTR
> 	sw	t1, 0x110(t0)	// *(0xb0000110) = 0x021;
> 
> 1:
> #if defined(CONFIG_NEC_MARKHAM_32MB)
> 	li	t1, 0x3c2	// SDTSR 32M
> 				// CAB=0x02, RAB=0x40 BTM=0x0080 SIZE=0x300(32M)
> #else
> 	li	t1, 0x1a1	// SDTSR  8M
> 				// CAB=0x01, RAB=0x20 BTM=0x0080 SIZE=0x100(8M)
> #endif
> 	sw	t1, 0x10c(t0)	// *(0xb000010c) = 0x1a1/0x3c2;
> 	
> 				// allow writing to flash parts
> 	li	t1,	0x100	// RMMDR (00 4MB mode) WM=1 unmasked
> 	sw	t1, 0x100(t0)	// *(0xb0000100) = 0x000;
> 
> 	li	t1,	0x003	// RMATR   FAT_8
> 	sw	t1, 0x104(t0)	// *(0xb0000104) = 0x003;
> 
> 	li	t1,	0x000	// IMR
> 	sw	t1, 0x00c(t0)	// disable all INTs
> 
> 	nop
> 	/* end of setup_hardware */
> 
> 	CHECKPOINT(0xB2)
> 
> 	/* invalidate_cache (was function, now PIC as may be linked at RAM) */ 
> 
> 	mtc0	zero, CP0_TAGLO	// invalidate the instruction cache
> 	li	s0, 0x80000000
> 	li	s1, 0x80000000 + (1 << 14) // 16k
> 1:
> 	cache	(2 << 2) | 0, (s0)
> 	add	s0, 0x10
> 	bne	s0, s1, 1b
> 	nop
> 
> 	mtc0	zero, CP0_TAGLO	// invalidate the data cache
> 	li	s0, 0x80000000
> 	li	s1, 0x80000000 + (1 << 13) // 8k
> 1:
> 	cache	(2 << 2) | 1, (s0)
> 	add	s0, 0x10
> 	bne	s0, s1, 1b
> 	nop
> 
> 	nop
> 	/* end of invalidate_cache */
> 
> 	CHECKPOINT(0xB3)
> 
> 	mfc0	t0, CP0_CONFIG	// enable kseg0 caching
> 	ori	t0, 0x1	// cached
> 	mtc0	t0, CP0_CONFIG
> 
> 	CHECKPOINT(0xB4)
> 
> #ifdef	CONFIG_NEC_KORVA_COMPRESSED
> relocate:
> 	la	sp, .stack
> 	move	s0, a0
> 	move	s1, a1
> 	move	s2, a2
> 	move	s3, a3
> 
> 	la	a0, __start
> 
> 	li	a1, FLASHADDR
> 	la	a2, _edata
> 	subu	t1, a2, a0
> 	srl	t1, t1, 2
> 
> 	/* copy text section */
> 	li	t0, 0
> 1:	lw	v0, 0(a1)
> 	nop
> 	sw	v0, 0(a0)
> 	xor	t0, t0, v0
> 	addu	a0, 4
> 	bne	a2, a0, 1b
> 	addu	a1, 4
> 
> 	/* clear BSS */
> 	la	a0, _edata
> 	la	a2, _end
> 2:	sw	zero, 0(a0)
> 	bne	a2, a0, 2b
> 	addu	a0, 4
> 
> 	/* now that code is in RAM we can jump okay */
> 
> 	// decompress kernel to ram
> 	la      ra, 1f
> 	la      t4, decompress_kernel	// elsewhere in 'C'
> 	jr      t4
> #else
> 	/* actually we were linked at ROM so we're okay */
> 
> 	// copy kernel to ram
> 	la      ra, 1f
> 	la      t4, copy_kernel
> 	jr      t4
> #endif
> 	nop
> 1:
> 
> 	CHECKPOINT(0xB5)
> 
> 	// jump to kernel entry point
> 	li	t4, KERNEL_ENTRY
> 	jr	t4
> 	nop
> 
> 	nop
> END(__start)
> 
> 
> LEAF(copy_kernel)
> 	// copy kernel from ROM to RAM
> 
> 	// Load s0 = src, s1 = src_end, s2 = dest.
> 	li	s0, FLASH_BEGIN
> 	li	s1, FLASH_END
> 
> 	li	s2, LOADADDR
> 
> 	sw	zero, (s2)
> 	sw	zero, 4(s2)
> 	sw	zero, 8(s2)
> 	sw	zero, 12(s2)
> 
> 1:
> 	// Only update LED once in a while.
> 	and	t0, s0, 0xffff
> 	bne	t0, zero, 2f
> 
> 	// Show copy progress on LED.
> 	srl	t0, s0, 16
> 	sh	t0, DEBUG_LED
> 
> 2:
> 	// Copy src to dest.
> 	lw	t0, (s0)
> 	bne	t0, zero, 3f
> 	nop
> 	sw	zero, (s2)
> 	beq	zero, zero, 4f
> 	nop
> 3:
> 	sw	t0, (s2)
> 4:
> 	// Increment src and dest.
> 	add	s0, 4
> 	add	s2, 4
> 
> 	// Loop if still more to go.
> 	bne	s0, s1, 1b
> 	nop
> 
> 	jr	ra
> 	 nop
> END(copy_kernel)
> 
> #ifdef	CONFIG_NEC_KORVA_COMPRESSED
> 	.comm .stack,4096*2,4	// used by relocate to call 'C' program
> #endif
> 
> =Boot.make======================================================================
> # Boot.make - used by ../boot/Makefile
> 
> OBJCOPY         = mipsel-linux-objcopy
> OBJCOPY_FLAGS   = -R .stab -R .stabstr -R .comment -R .note -R .reginfo -R .mdebug
> OBJDUMP         = mipsel-linux-objdump
> #OBJDUMP_HFLAGS  = --disassemble-all -b srec -m  mips:4600 -EL
> 
> TOPDIR          = ../../..
> 
> ifeq ($(LINUX_SRC),)
> LINUX_SRC=$(TOPDIR)
> endif
> VMLINUX=$(LINUX_SRC)/vmlinux
> SYSTEM_MAP=$(LINUX_SRC)/System.map
> 
> ifeq ($(USERLAND),)
> USERLAND=$(LINUX_SRC)/../userland
> endif
> 
> FLASHADDR      = 0xbfc00000
> 
> # see arch/mips/Makefile LOADADDR
> # 0x10000  bottom 64k reserved by/for boot.S (FLASH_BEGIN)
> # 0x02000  additional 8k reserved RAM image vector table (see LOADADDR)
> 
> ifdef CONFIG_NEC_MARKHAM_BOOTROM
> NN=10#in hex is 16M
> LOADADDR      += 0x81002000
> ZLOADADDR     += 0x81402000# +4M
> VMA            = 0x7f00e000 # 0x0-0x81000000+0x10000-0x2000
> ZVMA           = 0x7ebfe000 # 0x0-0x81402000
> else
> NN=00
> LOADADDR      += 0x80002000
> ZLOADADDR     += 0x80402000# +4M
> VMA            = 0x8000e000 # 0x0-0x80000000+0x10000-0x2000
> ZVMA           = 0x7fbfe000 # 0x0-0x80402000
> endif
> 
> KERNEL_ENTRY=$(shell awk '/kernel_entry/{print "0x" $$1}' $(LINUX_SRC)/System.map)
> 
> override CFLAGS += -DKERNEL_ENTRY=$(KERNEL_ENTRY) -DNN=$(NN)
> override CFLAGS += -DLOADADDR=$(LOADADDR) -DFLASHADDR=$(FLASHADDR) -DZLOADADDR=$(ZLOADADDR)
> 
> ifdef CONFIG_NEC_KORVA_COMPRESSED
> zbootroms: bin zsrec
> else
> bootroms: bin srec
> endif
> 
> $(TOPDIR)/arch/$(ARCH)/ld.script.$(ZLOADADDR)_z: $(TOPDIR)/arch/$(ARCH)/ld.script.in 
> 	sed -e 's/kernel_entry/__start/' -e 's/@@LOADADDR@@/$(ZLOADADDR)/' <$< >$@
> 
> ZLINKFLAGS += -T $(TOPDIR)/arch/$(ARCH)/ld.script.$(ZLOADADDR)_z
> 
> misc.c: ../korva/misc.c
> 	@rm -f $@
> 	cp $< $@
> 
> boot.S: ../korva/boot.S
> 	@rm -f $@
> 	cp $< $@
> 
> #vrboot$(NN): boot.o 
> vrboot$(NN): boot.o misc.o $(TOPDIR)/arch/$(ARCH)/ld.script.$(ZLOADADDR)_z
> 	@rm -f $@
> ifdef CONFIG_NEC_KORVA_COMPRESSED
> 	$(LD) $(LDFLAGS) -static -EL -N -o $@ $(ZLINKFLAGS) boot.o misc.o
> else
> 	$(LD) $(LDFLAGS) -static -EL -N -o $@ -Ttext $(FLASHADDR) boot.o
> endif
> 	@rm -f $^ ../boot/misc.c ../boot/boot.S 
> 
> vmlinux$(NN): $(VMLINUX)
> 	@rm -f System$(NN).map
> 	cp -p $(SYSTEM_MAP) System$(NN).map
> 	cp -p $< $@
> 
> # move sections to 0, then add bfc00000 for ROMulator
> vrvmlinux$(NN)-0: vmlinux$(NN) vrboot$(NN)
> 	@rm -f $@
> 	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-addresses 0 -O binary \
> 		vrboot$(NN) vrboot$(NN)-0.bin
> 	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-addresses $(VMA) \
> 		vmlinux$(NN) vmlinux$(NN)-0
> 	$(OBJCOPY) $(OBJCOPY_FLAGS) --add-section .vrboot=vrboot$(NN)-0.bin  \
> 		--set-section-flags .vrboot=,alloc,load,readonly,code \
> 		vmlinux$(NN)-0 vrvmlinux$(NN)-0
> 
> vrvmlinux$(NN)-bfc: vrvmlinux$(NN)-0
> 	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-address $(FLASHADDR) \
> 		vrvmlinux$(NN)-0 vrvmlinux$(NN)-bfc
> 
> # xxx-bfc.bin is for 'reload' & 'reburn' used on a running system
> vrvmlinux$(NN)-bfc.bin: vrvmlinux$(NN)-0
> 	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-address $(FLASHADDR) -O binary \
> 		vrvmlinux$(NN)-0 vrvmlinux$(NN)-bfc.bin
> 
> bin: vrvmlinux$(NN)-bfc vrvmlinux$(NN)-bfc.bin
> 
> vrvmlinux$(NN)-0.srec: vrvmlinux$(NN)-0
> 	$(OBJCOPY) $(OBJCOPY_FLAGS) -O srec --srec-forceS3 --gap-fill 0 \
> 		vrvmlinux$(NN)-0 vrvmlinux$(NN)-0.srec
> 	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-address $(FLASHADDR) -O srec \
> 		vrvmlinux$(NN)-0 vrvmlinux$(NN)-bfc.srec
> 
> srec: vrvmlinux$(NN)-0.srec
> 
> # used by 'reburn' script for Korva (8M) which needs small 256K chunks
> split:
> 	@rm -rf split
> 	@mkdir split
> 	cd split && split -b 256k ../vrvmlinux00-bfc.bin vrvm
> 
> # compressed...
> 
> vmlinux$(NN).bin.gz: vmlinux$(NN)
> 	$(OBJCOPY) -S -O binary $< $<.bin
> 	gzip -vf $<.bin
> 
> zvrvmlinux$(NN)-0: vmlinux$(NN).bin.gz vrboot$(NN)
> 	@rm -f $@
> 	$(OBJCOPY) $(OBJCOPY_FLAGS) --change-addresses=$(ZVMA) \
> 		vrboot$(NN) zvrboot$(NN)-0
> 	$(OBJCOPY) $(OBJCOPY_FLAGS) --add-section .vmlinux=vmlinux$(NN).bin.gz \
> 		--change-section-address .vmlinux=0x10000 \
> 		--set-section-flags .vmlinux=,alloc,load,readonly,code \
> 		zvrboot$(NN)-0 zvrvmlinux$(NN)-0
> 
> zvrvmlinux$(NN)-0.srec: zvrvmlinux$(NN)-0
> 	$(OBJCOPY) $(OBJCOPY_FLAGS) -O srec --srec-forceS3 \
> 		zvrvmlinux$(NN)-0 zvrvmlinux$(NN)-0.srec
> 	$(OBJCOPY) $(OBJCOPY_FLAGS) -O srec --srec-forceS3 \
> 		--change-address 0xbfc00000 \
> 		zvrvmlinux$(NN)-0 zvrvmlinux$(NN)-bfc.srec
> 
> zsrec: zvrvmlinux$(NN)-0.srec
> 
> clean::
> 	@rm -f System* vrvmlinux* vmlinux* vrboot* zvrvmlinux*
> ================================================================================

Re: No bzImage target for MIPS

[Geoffrey Espin]

> The "original" pmon has outlived its purpose, although it's still found
> on many boards. I've become a fan of MIPS Tech's yamon; it's small, it
> works, it has just the features I need in an embedded system, and, did I
> say it works?.  Unfortunately the code is released under NDA only. If
> MIPS opened the code, I think you would see it on many more boards.

I rest my case.  Linux is GPL, last I checked.  "LinuxMon", I'll
dub it, solves the problem... for all architectures. :-)  Attached
is the trivial 'reload' script for loading linux on top of linux.
It's just a wrapper for 'dd'.  Oh yeah, your reset.c:xxx_restart()
needs a quick check to see if a new kernel got loaded, and jump
to it, instead of PROM.

> > itself, as the "boot monitor" is best approach.  All reusable
> > code, fastest bootstrap (i.e. no 2nd stage loading), etc... But,
> > of course, not optimal for size, though my image is 1.3M which
> > includes a 580K cramfs ramdisk.
> That's fine once you've ported the kernel to your board. What do you do
> before that point?

With a ROM Emulator, bringing up Linux is no harder than bringing
up a boot monitor or any embedded RTOS, no?  A day or three, if
you are lucky and brilliant.  A somewhat longer if you're not
lucky.  :-)  Getting serial I/O is always the first nut to crack...
with Monta Vista's dbg_io.c, this is greatly simplified.  Of
course, I'm not counting all the extra drivers (Ethernet, ...)
and assuming a well supported architecture.

> patch.  But I fail to see how having this code in every <board>
> directory is better than having a separate boot directory like the ppc
> arch.  The ppc-like zImage support I added is a bit different from the

I quite agree that initrd stuff is not right, or rather builtin
ramdisk.o is not good for embedded.  Having it attached to the
end of vmlinux, possibly padded/aligned to a nice flash sector
boundary should be the general rule.  Having <board>/Boot.make
so one can customize the "make boot" target rule to do the right
thing on a board-by-board basis seems to be the way to go.  E.g.
in arch/mips/boot/Makefile, simply:

    #
    # This file is subject to the terms and conditions of the GNU General Public
    # License.  See the file "COPYING" in the main directory of this archive
    # for more details.
    #
    # Copyright (C) 1995, 1998, 2001 by Ralf Baechle
    #

    .S.s:
	    $(CPP) $(CFLAGS) $< -o $*.s
    .S.o:
	    $(CC) $(CFLAGS) -c $< -o $*.o
    ...
    ifdef CONFIG_NEC_KORVA
    include ../korva/Boot.make
    endif

    all: vmlinux.ecoff addinitrd
    ...

BTW, does any actually build 'vmlinux.ecoff'?

And sure, cluttering up yet another Makefile with CONFIG_XXX_BOARD
is not desirable... rather some magic like:

    include $?{CONFIG_BOARD:common}/Boot.make

would be nice.

Geoff
-- 
Geoffrey Espin espin@idiom.com
--
=reload.sh======================================================================
#!/bin/sh

IMAGE=vrvmlinux00-bfc.bin  # 0M
SKIP=64	                   # 64K for vrboot
SEEK=8                     # 8k (80002000) skip excep vecs to kernel offset

if [ `sed -n -e 's/8\([0-9]\).*bzero/\1/p' /proc/ksyms` -eq 0 ] ; then
	echo "$0: kernel not running at high memory (e.g. 16M), use reboot"
	exit 1
fi

if [ $# -eq 0 ] ; then
	echo "usage (nfs): $0 <user>"
	echo "usage (ftp): $0 <user> <password>"
	exit 1
fi

USER=$1
shift

VM="/home/${USER}/linux/arch/mips/boot/${IMAGE}"

if [ $# -eq 1 ] ; then
	PW=$1
	shift
	VMI="ftp://${USER}:${PW}@10.1.1.1${VM}"
	VM=/tmp/${IMAGE}
	echo wget -c ${VMI} -O ${VM}
	wget -c ${VMI} -O ${VM}
	if [ $? -ne 0 ] ; then
		echo "wget failed"
		exit 1
	fi
else
	# using NFS
fi

echo dd if=${VM} of=/dev/mem bs=1k skip=${SKIP} seek=${SEEK} conv=notrunc 
dd if=${VM} of=/dev/mem bs=1k skip=${SKIP} seek=${SEEK} conv=notrunc 

if [ $? -ne 0 ] ; then
	echo "dd failed"
	exit 1
fi

reboot

--

Re: No bzImage target for MIPS

[Pete Popov]

On Fri, 2001-12-14 at 10:52, Geoffrey Espin wrote:
> 
> > The "original" pmon has outlived its purpose, although it's still found
> > on many boards. I've become a fan of MIPS Tech's yamon; it's small, it
> > works, it has just the features I need in an embedded system, and, did I
> > say it works?.  Unfortunately the code is released under NDA only. If
> > MIPS opened the code, I think you would see it on many more boards.
> 
> I rest my case.  Linux is GPL, last I checked.  "LinuxMon", I'll
> dub it, solves the problem... for all architectures. :-)  Attached
> is the trivial 'reload' script for loading linux on top of linux.
> It's just a wrapper for 'dd'.  Oh yeah, your reset.c:xxx_restart()
> needs a quick check to see if a new kernel got loaded, and jump
> to it, instead of PROM.
> 
> > > itself, as the "boot monitor" is best approach.  All reusable
> > > code, fastest bootstrap (i.e. no 2nd stage loading), etc... But,
> > > of course, not optimal for size, though my image is 1.3M which
> > > includes a 580K cramfs ramdisk.
> > That's fine once you've ported the kernel to your board. What do you do
> > before that point?
> 
> With a ROM Emulator, bringing up Linux is no harder than bringing
> up a boot monitor or any embedded RTOS, no?  A day or three, if
> you are lucky and brilliant.  A somewhat longer if you're not
> lucky.  :-)  Getting serial I/O is always the first nut to crack...
> with Monta Vista's dbg_io.c, this is greatly simplified.  Of
> course, I'm not counting all the extra drivers (Ethernet, ...)
> and assuming a well supported architecture.

True, if you can use a rom emulator on the given board, which you can't
do with all boards.  But then you need to connect the emulator to the
board every time you want to do something ... I don't know, give me a
working boot loader with enet support and I'll take that instead.
 
> > patch.  But I fail to see how having this code in every <board>
> > directory is better than having a separate boot directory like the ppc
> > arch.  The ppc-like zImage support I added is a bit different from the
> 
> I quite agree that initrd stuff is not right, or rather builtin
> ramdisk.o is not good for embedded.  Having it attached to the
> end of vmlinux, possibly padded/aligned to a nice flash sector
> boundary should be the general rule.  Having <board>/Boot.make
> so one can customize the "make boot" target rule to do the right
> thing on a board-by-board basis seems to be the way to go.  

Exactly. That's why the <board> directory in arch/mips/zboot :-) BTW, I
added the new zboot directory in sourceforge because arch/mips/boot was
in the way and I didn't want to mess with that one. 

Ultimately I don't care if the zImage support is in each <board>
directory or in a new directory like arch/mips/zboot, as long as the
support for zImages and, eventually, initrd, is there. 

BTW, if you don't package up your work in a patch that Ralf would take,
the bits get lost over time. I don't know if you care, but that's how
linux evolves so I hope you submit your patches upstream rather than
allowing them to get lost in mailing lists.

Pete

> E.g.
> in arch/mips/boot/Makefile, simply:
> 
>     #
>     # This file is subject to the terms and conditions of the GNU General Public
>     # License.  See the file "COPYING" in the main directory of this archive
>     # for more details.
>     #
>     # Copyright (C) 1995, 1998, 2001 by Ralf Baechle
>     #
> 
>     .S.s:
> 	    $(CPP) $(CFLAGS) $< -o $*.s
>     .S.o:
> 	    $(CC) $(CFLAGS) -c $< -o $*.o
>     ...
>     ifdef CONFIG_NEC_KORVA
>     include ../korva/Boot.make
>     endif
> 
>     all: vmlinux.ecoff addinitrd
>     ...
> 
> BTW, does any actually build 'vmlinux.ecoff'?
> 
> And sure, cluttering up yet another Makefile with CONFIG_XXX_BOARD
> is not desirable... rather some magic like:
> 
>     include $?{CONFIG_BOARD:common}/Boot.make
> 
> would be nice.
> 
> Geoff
> -- 
> Geoffrey Espin espin@idiom.com
> --
> =reload.sh======================================================================
> #!/bin/sh
> 
> IMAGE=vrvmlinux00-bfc.bin  # 0M
> SKIP=64	                   # 64K for vrboot
> SEEK=8                     # 8k (80002000) skip excep vecs to kernel offset
> 
> if [ `sed -n -e 's/8\([0-9]\).*bzero/\1/p' /proc/ksyms` -eq 0 ] ; then
> 	echo "$0: kernel not running at high memory (e.g. 16M), use reboot"
> 	exit 1
> fi
> 
> if [ $# -eq 0 ] ; then
> 	echo "usage (nfs): $0 <user>"
> 	echo "usage (ftp): $0 <user> <password>"
> 	exit 1
> fi
> 
> USER=$1
> shift
> 
> VM="/home/${USER}/linux/arch/mips/boot/${IMAGE}"
> 
> if [ $# -eq 1 ] ; then
> 	PW=$1
> 	shift
> 	VMI="ftp://${USER}:${PW}@10.1.1.1${VM}"
> 	VM=/tmp/${IMAGE}
> 	echo wget -c ${VMI} -O ${VM}
> 	wget -c ${VMI} -O ${VM}
> 	if [ $? -ne 0 ] ; then
> 		echo "wget failed"
> 		exit 1
> 	fi
> else
> 	# using NFS
> fi
> 
> echo dd if=${VM} of=/dev/mem bs=1k skip=${SKIP} seek=${SEEK} conv=notrunc 
> dd if=${VM} of=/dev/mem bs=1k skip=${SKIP} seek=${SEEK} conv=notrunc 
> 
> if [ $? -ne 0 ] ; then
> 	echo "dd failed"
> 	exit 1
> fi
> 
> reboot
> 
> --

RE: No bzImage target for MIPS

[Marc Karasek]

 What about redboot?  It is Open Source and has support for a variety of
processors.  The structure is such that adding support for a new
platform/processor is very easy.  It has support for booting a linux kernel
and as an added bonus support for ecos is pretty much done once you get it
ported. 



-----Original Message-----
From: Pete Popov
To: Geoffrey Espin
Cc: linux-mips
Sent: 12/14/01 2:11 PM
Subject: Re: No bzImage target for MIPS

On Fri, 2001-12-14 at 10:52, Geoffrey Espin wrote:
> 
> > The "original" pmon has outlived its purpose, although it's still
found
> > on many boards. I've become a fan of MIPS Tech's yamon; it's small,
it
> > works, it has just the features I need in an embedded system, and,
did I
> > say it works?.  Unfortunately the code is released under NDA only.
If
> > MIPS opened the code, I think you would see it on many more boards.
> 
> I rest my case.  Linux is GPL, last I checked.  "LinuxMon", I'll
> dub it, solves the problem... for all architectures. :-)  Attached
> is the trivial 'reload' script for loading linux on top of linux.
> It's just a wrapper for 'dd'.  Oh yeah, your reset.c:xxx_restart()
> needs a quick check to see if a new kernel got loaded, and jump
> to it, instead of PROM.
> 
> > > itself, as the "boot monitor" is best approach.  All reusable
> > > code, fastest bootstrap (i.e. no 2nd stage loading), etc... But,
> > > of course, not optimal for size, though my image is 1.3M which
> > > includes a 580K cramfs ramdisk.
> > That's fine once you've ported the kernel to your board. What do you
do
> > before that point?
> 
> With a ROM Emulator, bringing up Linux is no harder than bringing
> up a boot monitor or any embedded RTOS, no?  A day or three, if
> you are lucky and brilliant.  A somewhat longer if you're not
> lucky.  :-)  Getting serial I/O is always the first nut to crack...
> with Monta Vista's dbg_io.c, this is greatly simplified.  Of
> course, I'm not counting all the extra drivers (Ethernet, ...)
> and assuming a well supported architecture.

True, if you can use a rom emulator on the given board, which you can't
do with all boards.  But then you need to connect the emulator to the
board every time you want to do something ... I don't know, give me a
working boot loader with enet support and I'll take that instead.
 
> > patch.  But I fail to see how having this code in every <board>
> > directory is better than having a separate boot directory like the
ppc
> > arch.  The ppc-like zImage support I added is a bit different from
the
> 
> I quite agree that initrd stuff is not right, or rather builtin
> ramdisk.o is not good for embedded.  Having it attached to the
> end of vmlinux, possibly padded/aligned to a nice flash sector
> boundary should be the general rule.  Having <board>/Boot.make
> so one can customize the "make boot" target rule to do the right
> thing on a board-by-board basis seems to be the way to go.  

Exactly. That's why the <board> directory in arch/mips/zboot :-) BTW, I
added the new zboot directory in sourceforge because arch/mips/boot was
in the way and I didn't want to mess with that one. 

Ultimately I don't care if the zImage support is in each <board>
directory or in a new directory like arch/mips/zboot, as long as the
support for zImages and, eventually, initrd, is there. 

BTW, if you don't package up your work in a patch that Ralf would take,
the bits get lost over time. I don't know if you care, but that's how
linux evolves so I hope you submit your patches upstream rather than
allowing them to get lost in mailing lists.

Pete

> E.g.
> in arch/mips/boot/Makefile, simply:
> 
>     #
>     # This file is subject to the terms and conditions of the GNU
General Public
>     # License.  See the file "COPYING" in the main directory of this
archive
>     # for more details.
>     #
>     # Copyright (C) 1995, 1998, 2001 by Ralf Baechle
>     #
> 
>     .S.s:
> 	    $(CPP) $(CFLAGS) $< -o $*.s
>     .S.o:
> 	    $(CC) $(CFLAGS) -c $< -o $*.o
>     ...
>     ifdef CONFIG_NEC_KORVA
>     include ../korva/Boot.make
>     endif
> 
>     all: vmlinux.ecoff addinitrd
>     ...
> 
> BTW, does any actually build 'vmlinux.ecoff'?
> 
> And sure, cluttering up yet another Makefile with CONFIG_XXX_BOARD
> is not desirable... rather some magic like:
> 
>     include $?{CONFIG_BOARD:common}/Boot.make
> 
> would be nice.
> 
> Geoff
> -- 
> Geoffrey Espin espin@idiom.com
> --
>
=reload.sh==============================================================
========
> #!/bin/sh
> 
> IMAGE=vrvmlinux00-bfc.bin  # 0M
> SKIP=64	                   # 64K for vrboot
> SEEK=8                     # 8k (80002000) skip excep vecs to kernel
offset
> 
> if [ `sed -n -e 's/8\([0-9]\).*bzero/\1/p' /proc/ksyms` -eq 0 ] ; then
> 	echo "$0: kernel not running at high memory (e.g. 16M), use
reboot"
> 	exit 1
> fi
> 
> if [ $# -eq 0 ] ; then
> 	echo "usage (nfs): $0 <user>"
> 	echo "usage (ftp): $0 <user> <password>"
> 	exit 1
> fi
> 
> USER=$1
> shift
> 
> VM="/home/${USER}/linux/arch/mips/boot/${IMAGE}"
> 
> if [ $# -eq 1 ] ; then
> 	PW=$1
> 	shift
> 	VMI="ftp://${USER}:${PW}@10.1.1.1${VM}"
> 	VM=/tmp/${IMAGE}
> 	echo wget -c ${VMI} -O ${VM}
> 	wget -c ${VMI} -O ${VM}
> 	if [ $? -ne 0 ] ; then
> 		echo "wget failed"
> 		exit 1
> 	fi
> else
> 	# using NFS
> fi
> 
> echo dd if=${VM} of=/dev/mem bs=1k skip=${SKIP} seek=${SEEK}
conv=notrunc 
> dd if=${VM} of=/dev/mem bs=1k skip=${SKIP} seek=${SEEK} conv=notrunc 
> 
> if [ $? -ne 0 ] ; then
> 	echo "dd failed"
> 	exit 1
> fi
> 
> reboot
> 
> --

RE: No bzImage target for MIPS

[Pete Popov]

On Fri, 2001-12-14 at 14:59, Marc Karasek wrote:
>  What about redboot?  It is Open Source and has support for a variety of
> processors.  The structure is such that adding support for a new
> platform/processor is very easy.  It has support for booting a linux kernel
> and as an added bonus support for ecos is pretty much done once you get it
> ported. 

If you're doing your own board, redboot may be fine. But typically
vendors of referrence boards want you to use whatever boot code they
already have running because they are not willing to pay for new boot
code. 

Pete

> 
> 
> 
> -----Original Message-----
> From: Pete Popov
> To: Geoffrey Espin
> Cc: linux-mips
> Sent: 12/14/01 2:11 PM
> Subject: Re: No bzImage target for MIPS
> 
> On Fri, 2001-12-14 at 10:52, Geoffrey Espin wrote:
> > 
> > > The "original" pmon has outlived its purpose, although it's still
> found
> > > on many boards. I've become a fan of MIPS Tech's yamon; it's small,
> it
> > > works, it has just the features I need in an embedded system, and,
> did I
> > > say it works?.  Unfortunately the code is released under NDA only.
> If
> > > MIPS opened the code, I think you would see it on many more boards.
> > 
> > I rest my case.  Linux is GPL, last I checked.  "LinuxMon", I'll
> > dub it, solves the problem... for all architectures. :-)  Attached
> > is the trivial 'reload' script for loading linux on top of linux.
> > It's just a wrapper for 'dd'.  Oh yeah, your reset.c:xxx_restart()
> > needs a quick check to see if a new kernel got loaded, and jump
> > to it, instead of PROM.
> > 
> > > > itself, as the "boot monitor" is best approach.  All reusable
> > > > code, fastest bootstrap (i.e. no 2nd stage loading), etc... But,
> > > > of course, not optimal for size, though my image is 1.3M which
> > > > includes a 580K cramfs ramdisk.
> > > That's fine once you've ported the kernel to your board. What do you
> do
> > > before that point?
> > 
> > With a ROM Emulator, bringing up Linux is no harder than bringing
> > up a boot monitor or any embedded RTOS, no?  A day or three, if
> > you are lucky and brilliant.  A somewhat longer if you're not
> > lucky.  :-)  Getting serial I/O is always the first nut to crack...
> > with Monta Vista's dbg_io.c, this is greatly simplified.  Of
> > course, I'm not counting all the extra drivers (Ethernet, ...)
> > and assuming a well supported architecture.
> 
> True, if you can use a rom emulator on the given board, which you can't
> do with all boards.  But then you need to connect the emulator to the
> board every time you want to do something ... I don't know, give me a
> working boot loader with enet support and I'll take that instead.
>  
> > > patch.  But I fail to see how having this code in every <board>
> > > directory is better than having a separate boot directory like the
> ppc
> > > arch.  The ppc-like zImage support I added is a bit different from
> the
> > 
> > I quite agree that initrd stuff is not right, or rather builtin
> > ramdisk.o is not good for embedded.  Having it attached to the
> > end of vmlinux, possibly padded/aligned to a nice flash sector
> > boundary should be the general rule.  Having <board>/Boot.make
> > so one can customize the "make boot" target rule to do the right
> > thing on a board-by-board basis seems to be the way to go.  
> 
> Exactly. That's why the <board> directory in arch/mips/zboot :-) BTW, I
> added the new zboot directory in sourceforge because arch/mips/boot was
> in the way and I didn't want to mess with that one. 
> 
> Ultimately I don't care if the zImage support is in each <board>
> directory or in a new directory like arch/mips/zboot, as long as the
> support for zImages and, eventually, initrd, is there. 
> 
> BTW, if you don't package up your work in a patch that Ralf would take,
> the bits get lost over time. I don't know if you care, but that's how
> linux evolves so I hope you submit your patches upstream rather than
> allowing them to get lost in mailing lists.
> 
> Pete
> 
> > E.g.
> > in arch/mips/boot/Makefile, simply:
> > 
> >     #
> >     # This file is subject to the terms and conditions of the GNU
> General Public
> >     # License.  See the file "COPYING" in the main directory of this
> archive
> >     # for more details.
> >     #
> >     # Copyright (C) 1995, 1998, 2001 by Ralf Baechle
> >     #
> > 
> >     .S.s:
> > 	    $(CPP) $(CFLAGS) $< -o $*.s
> >     .S.o:
> > 	    $(CC) $(CFLAGS) -c $< -o $*.o
> >     ...
> >     ifdef CONFIG_NEC_KORVA
> >     include ../korva/Boot.make
> >     endif
> > 
> >     all: vmlinux.ecoff addinitrd
> >     ...
> > 
> > BTW, does any actually build 'vmlinux.ecoff'?
> > 
> > And sure, cluttering up yet another Makefile with CONFIG_XXX_BOARD
> > is not desirable... rather some magic like:
> > 
> >     include $?{CONFIG_BOARD:common}/Boot.make
> > 
> > would be nice.
> > 
> > Geoff
> > -- 
> > Geoffrey Espin espin@idiom.com
> > --
> >
> =reload.sh==============================================================
> ========
> > #!/bin/sh
> > 
> > IMAGE=vrvmlinux00-bfc.bin  # 0M
> > SKIP=64	                   # 64K for vrboot
> > SEEK=8                     # 8k (80002000) skip excep vecs to kernel
> offset
> > 
> > if [ `sed -n -e 's/8\([0-9]\).*bzero/\1/p' /proc/ksyms` -eq 0 ] ; then
> > 	echo "$0: kernel not running at high memory (e.g. 16M), use
> reboot"
> > 	exit 1
> > fi
> > 
> > if [ $# -eq 0 ] ; then
> > 	echo "usage (nfs): $0 <user>"
> > 	echo "usage (ftp): $0 <user> <password>"
> > 	exit 1
> > fi
> > 
> > USER=$1
> > shift
> > 
> > VM="/home/${USER}/linux/arch/mips/boot/${IMAGE}"
> > 
> > if [ $# -eq 1 ] ; then
> > 	PW=$1
> > 	shift
> > 	VMI="ftp://${USER}:${PW}@10.1.1.1${VM}"
> > 	VM=/tmp/${IMAGE}
> > 	echo wget -c ${VMI} -O ${VM}
> > 	wget -c ${VMI} -O ${VM}
> > 	if [ $? -ne 0 ] ; then
> > 		echo "wget failed"
> > 		exit 1
> > 	fi
> > else
> > 	# using NFS
> > fi
> > 
> > echo dd if=${VM} of=/dev/mem bs=1k skip=${SKIP} seek=${SEEK}
> conv=notrunc 
> > dd if=${VM} of=/dev/mem bs=1k skip=${SKIP} seek=${SEEK} conv=notrunc 
> > 
> > if [ $? -ne 0 ] ; then
> > 	echo "dd failed"
> > 	exit 1
> > fi
> > 
> > reboot
> > 
> > --
> 

Re: No bzImage target for MIPS

[Karsten Merker]

On Fri, Dec 14, 2001 at 10:52:57AM -0800, Geoffrey Espin wrote:

>     all: vmlinux.ecoff addinitrd
>     ...
> 
> BTW, does any actually build 'vmlinux.ecoff'?

Yes, it is (among other uses) the default target for DECstations,
the DECstation firmware cannot tftp-boot anything else.

Greetings,
Karsten
-- 
#include <standard_disclaimer>
Nach Paragraph 28 Abs. 3 Bundesdatenschutzgesetz widerspreche ich der Nutzung
oder Uebermittlung meiner Daten fuer Werbezwecke oder fuer die Markt- oder
Meinungsforschung.

Re: No bzImage target for MIPS

[Keith Owens]

On Sat, 15 Dec 2001 09:31:01 +0100, 
Karsten Merker <karsten@excalibur.cologne.de> wrote:
>On Fri, Dec 14, 2001 at 10:52:57AM -0800, Geoffrey Espin wrote:
>> BTW, does any actually build 'vmlinux.ecoff'?
>
>Yes, it is (among other uses) the default target for DECstations,
>the DECstation firmware cannot tftp-boot anything else.

I am standardizing the list of supported formats for linux boot, as
part of the kbuild 2.5 rewrite, I will add vmlinux.ecoff to the master
list.

AFAICT ecoff is only used on mips but, since ecoff is not an arch
specific object format, it makes sense to make it a generic kbuild
target, like elf, srec and bin.  To that end, I looked at moving
elf2ecoff and addinitrd to an arch independent directory so everybody
could use those tools, alas both contain mips specific code.  Any idea
how much work is required to make elf2ecoff and addinitrd into generic
utilities?  Is it worth the effort or should they stay as mips only?

Re: No bzImage target for MIPS

[Florian Lohoff]

[-- Attachment #1: Type: text/plain, Size: 352 bytes --]

On Fri, Dec 14, 2001 at 10:52:57AM -0800, Geoffrey Espin wrote:
> 
> BTW, does any actually build 'vmlinux.ecoff'?
> 

Yep - Some Indy/Indigo2 proms only boot ecoff images via tftp.

Flo
-- 
Florian Lohoff                  flo@rfc822.org             +49-5201-669912
Nine nineth on september the 9th              Welcome to the new billenium

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: No bzImage target for MIPS

[Florian Lohoff]

[-- Attachment #1: Type: text/plain, Size: 947 bytes --]

On Sat, Dec 15, 2001 at 08:14:59PM +1100, Keith Owens wrote:

> target, like elf, srec and bin.  To that end, I looked at moving
> elf2ecoff and addinitrd to an arch independent directory so everybody
> could use those tools, alas both contain mips specific code.  Any idea
> how much work is required to make elf2ecoff and addinitrd into generic
> utilities?  Is it worth the effort or should they stay as mips only?

Last i had a look elf2ecoff and addinitrd were not really nice
recovering from toolchain problems like pages of null between sections
which was a common fault a couple months back. This lead to addinitrd
not beeing able to attach a ramdisk that the kernel would be able to
find it. 

BTW: Shouldn't objcopy be able to convert an ELF into an ECOFF ?

Flo
-- 
Florian Lohoff                  flo@rfc822.org             +49-5201-669912
Nine nineth on september the 9th              Welcome to the new billenium

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: No bzImage target for MIPS

[Maciej W. Rozycki]

On Sat, 15 Dec 2001, Keith Owens wrote:

> AFAICT ecoff is only used on mips but, since ecoff is not an arch
> specific object format, it makes sense to make it a generic kbuild
> target, like elf, srec and bin.  To that end, I looked at moving
> elf2ecoff and addinitrd to an arch independent directory so everybody
> could use those tools, alas both contain mips specific code.  Any idea
> how much work is required to make elf2ecoff and addinitrd into generic
> utilities?  Is it worth the effort or should they stay as mips only?

 Elf2ecoff is probably going not to be used one day anymore, i.e. once
binutils are fixed.  Right now there is a problem in marking ELF
executables impure (non-paged).  Ultimately either objcopy or even ld
directly is going to be used for ECOFF binary creation. 

-- 
+  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
+--------------------------------------------------------------+
+        e-mail: macro@ds2.pg.gda.pl, PGP key available        +

Re: No bzImage target for MIPS

[Ralf Baechle]

On Sat, Dec 15, 2001 at 01:46:04PM +0100, Florian Lohoff wrote:

> > target, like elf, srec and bin.  To that end, I looked at moving
> > elf2ecoff and addinitrd to an arch independent directory so everybody
> > could use those tools, alas both contain mips specific code.  Any idea
> > how much work is required to make elf2ecoff and addinitrd into generic
> > utilities?  Is it worth the effort or should they stay as mips only?
> 
> Last i had a look elf2ecoff and addinitrd were not really nice
> recovering from toolchain problems like pages of null between sections
> which was a common fault a couple months back. This lead to addinitrd
> not beeing able to attach a ramdisk that the kernel would be able to
> find it. 
> 
> BTW: Shouldn't objcopy be able to convert an ELF into an ECOFF ?

In theory yes, in practice that was found to be so problematic that we
just gave up and use those NetBSD tools.

  Ralf