[PATCH -v1] MIPS: add support for gzip/bzip2/lzma compressed kernel images

[PATCH -v1] MIPS: add support for gzip/bzip2/lzma compressed kernel images

[Wu Zhangjin]

This patch will help to generate smaller kernel images for linux-MIPS,

Here is the effect when using lzma:

$ ls -sh vmlinux
7.1M vmlinux
$ ls -sh arch/mips/boot/compressed/vmlinuz
1.5M arch/mips/boot/compressed/vmlinuz

Have tested the 32bit kernel on Qemu/Malta and 64bit kernel on FuLoong
Mini PC. both of them work well.

and this revision incorporates the feedback from Alexander Clouter
<alex@digriz.org.uk>, he helped to test it on AR7[1] based Linksys
WAG54Gv2 and gave good suggestion on board-independence.

NOTE: this should work for the other MIPS-based machines, but I have
used the command bc in the Makefile to calculate the load address of the
compressed kernel. I'm not sure this is suitable.  perhaps I need to
rewrite this part in C program or somebody help to simplify the current
implementation.

Signed-off-by: Wu Zhangjin <wuzhangjin@gmail.com>
---
 arch/mips/Kconfig                      |    3 +
 arch/mips/Makefile                     |   11 ++-
 arch/mips/boot/compressed/Makefile     |   72 +++++++++++++++
 arch/mips/boot/compressed/dbg.c        |   64 ++++++++++++++
 arch/mips/boot/compressed/dbg.h        |   19 ++++
 arch/mips/boot/compressed/decompress.c |  149 +++++++++++++++++++++++++++++++
 arch/mips/boot/compressed/dummy.c      |    4 +
 arch/mips/boot/compressed/head.S       |   86 ++++++++++++++++++
 arch/mips/boot/compressed/ld.script    |  150 ++++++++++++++++++++++++++++++++
 9 files changed, 556 insertions(+), 2 deletions(-)
 create mode 100644 arch/mips/boot/compressed/Makefile
 create mode 100644 arch/mips/boot/compressed/dbg.c
 create mode 100644 arch/mips/boot/compressed/dbg.h
 create mode 100644 arch/mips/boot/compressed/decompress.c
 create mode 100644 arch/mips/boot/compressed/dummy.c
 create mode 100644 arch/mips/boot/compressed/head.S
 create mode 100644 arch/mips/boot/compressed/ld.script

diff --git a/arch/mips/Kconfig b/arch/mips/Kconfig
index 3ca0fe1..fae7029 100644
--- a/arch/mips/Kconfig
+++ b/arch/mips/Kconfig
@@ -4,6 +4,9 @@ config MIPS
 	select HAVE_IDE
 	select HAVE_OPROFILE
 	select HAVE_ARCH_KGDB
+	select HAVE_KERNEL_GZIP
+	select HAVE_KERNEL_BZIP2
+	select HAVE_KERNEL_LZMA
 	# Horrible source of confusion.  Die, die, die ...
 	select EMBEDDED
 	select RTC_LIB
diff --git a/arch/mips/Makefile b/arch/mips/Makefile
index 861da51..300b996 100644
--- a/arch/mips/Makefile
+++ b/arch/mips/Makefile
@@ -709,7 +709,10 @@ vmlinux.64: vmlinux
 
 makeboot =$(Q)$(MAKE) $(build)=arch/mips/boot VMLINUX=$(vmlinux-32) $(1)
 
-all:	$(all-y)
+all:	$(all-y) zImage
+
+zImage: vmlinux
+	$(Q)$(MAKE) $(build)=arch/mips/boot/compressed $@ LOADADDR=$(load-y)
 
 vmlinux.bin: $(vmlinux-32)
 	+@$(call makeboot,$@)
@@ -735,11 +738,13 @@ endif
 
 install:
 	$(Q)install -D -m 755 vmlinux $(INSTALL_PATH)/vmlinux-$(KERNELRELEASE)
+	$(Q)install -D -m 755 vmlinuz $(INSTALL_PATH)/vmlinuz-$(KERNELRELEASE)
 	$(Q)install -D -m 644 .config $(INSTALL_PATH)/config-$(KERNELRELEASE)
 	$(Q)install -D -m 644 System.map $(INSTALL_PATH)/System.map-$(KERNELRELEASE)
 
 archclean:
 	@$(MAKE) $(clean)=arch/mips/boot
+	@$(MAKE) $(clean)=arch/mips/boot/compressed
 	@$(MAKE) $(clean)=arch/mips/lasat
 
 define archhelp
@@ -747,10 +752,12 @@ define archhelp
 	echo '  vmlinux.ecoff        - ECOFF boot image'
 	echo '  vmlinux.bin          - Raw binary boot image'
 	echo '  vmlinux.srec         - SREC boot image'
+	echo '  vmlinuz              - Compressed boot image'
 	echo
 	echo '  These will be default as apropriate for a configured platform.'
 endef
 
 CLEAN_FILES += vmlinux.32 \
 	       vmlinux.64 \
-	       vmlinux.ecoff
+	       vmlinux.ecoff \
+               vmlinuz
diff --git a/arch/mips/boot/compressed/Makefile b/arch/mips/boot/compressed/Makefile
new file mode 100644
index 0000000..f9ccd97
--- /dev/null
+++ b/arch/mips/boot/compressed/Makefile
@@ -0,0 +1,72 @@
+#
+# This file is subject to the terms and conditions of the GNU General Public
+# License.
+#
+# Adapted for MIPS Pete Popov, Dan Malek
+#
+# Copyright (C) 1994 by Linus Torvalds
+# Adapted for PowerPC by Gary Thomas
+# modified by Cort (cort@cs.nmt.edu)
+#
+# Copyright (C) 2009 Lemote Inc. & Insititute of Computing Technology
+# Author: Wu Zhangjin <wuzj@lemote.com>
+#
+
+# The load address of the compressed kernel, VMLINUZ_LOAD_ADDRESS > VMLINUX_LOAD_ADDRESS + VMLINUX_SIZE
+VMLINUX_SIZE := $(shell wc -c $(objtree)/$(KBUILD_IMAGE) 2>/dev/null | cut -d' ' -f1)
+VMLINUX_SIZE := $(shell echo "obase=16;ibase=10;(($(VMLINUX_SIZE)/65536 + 1) * 65536)" | bc | cut -d'.' -f1)
+VMLINUX_LOAD_ADDRESS = $(shell echo $(LOADADDR) | sed -e "s/0xffffffff//g")
+VMLINUZ_LOAD_ADDRESS := $(shell echo "obase=16; ibase=16; ($(VMLINUX_LOAD_ADDRESS) + $(VMLINUX_SIZE))" | bc)
+VMLINUZ_LOAD_ADDRESS := 0x$(if $(CONFIG_64BIT),ffffffff,)$(VMLINUZ_LOAD_ADDRESS)
+LOADADDR := 0x$(if $(CONFIG_64BIT),ffffffff,)$(VMLINUX_LOAD_ADDRESS)
+
+# set the default size of the mallocing area for decompressing
+BOOT_HEAP_SIZE := 0x400000
+
+KBUILD_CFLAGS := $(LINUXINCLUDE) $(KBUILD_CFLAGS) -D__KERNEL__ \
+	-DBOOT_HEAP_SIZE=$(BOOT_HEAP_SIZE) -D"LOADADDR=$(LOADADDR)ull" \
+
+# uncomment the -DDEBUG to enable serial port debugging for your machine
+# and please ensure there is a suitable serial port address defined in dbg.h
+KBUILD_CFLAGS += #-DDEBUG
+
+KBUILD_AFLAGS := $(LINUXINCLUDE) $(KBUILD_AFLAGS) -D__ASSEMBLY__ \
+	-DKERNEL_ENTRY=0x$(shell $(NM) $(objtree)/$(KBUILD_IMAGE) 2>/dev/null | grep " kernel_entry" | cut -f1 -d \ ) \
+	-DBOOT_HEAP_SIZE=$(BOOT_HEAP_SIZE)
+
+OBJECTS := $(obj)/head.o $(obj)/decompress.o $(obj)/dbg.o
+
+OBJCOPYFLAGS_vmlinux.bin := $(OBJCOPYFLAGS) -O binary -R .comment -S
+$(obj)/vmlinux.bin: $(KBUILD_IMAGE)
+	$(call if_changed,objcopy)
+
+suffix_$(CONFIG_KERNEL_GZIP)  = gz
+suffix_$(CONFIG_KERNEL_BZIP2) = bz2
+suffix_$(CONFIG_KERNEL_LZMA)  = lzma
+tool_$(CONFIG_KERNEL_GZIP)    = gzip
+tool_$(CONFIG_KERNEL_BZIP2)   = bzip2
+tool_$(CONFIG_KERNEL_LZMA)    = lzma
+$(obj)/vmlinux.$(suffix_y): $(obj)/vmlinux.bin
+	$(call if_changed,$(tool_y))
+	$(Q)rm -f $<
+
+$(obj)/piggy.o: $(obj)/vmlinux.$(suffix_y) $(obj)/dummy.o
+	$(Q)$(OBJCOPY) $(OBJCOPYFLAGS) \
+		--add-section=.image=$< \
+		--set-section-flags=.image=contents,alloc,load,readonly,data \
+		$(obj)/dummy.o $@
+	$(Q)rm -f $<
+
+LDFLAGS_vmlinuz := $(LDFLAGS) -Ttext $(VMLINUZ_LOAD_ADDRESS) -T
+$(obj)/vmlinuz: $(src)/ld.script $(OBJECTS) $(obj)/piggy.o
+	$(call if_changed,ld)
+	$(Q)$(OBJCOPY) $(OBJCOPYFLAGS) -R .comment -R .stab -R .stabstr -R .initrd -R .sysmap $@
+	$(Q)rm -f $(obj)/piggy.o
+
+zImage: $(obj)/vmlinuz
+	$(Q)ln -sf $< $(objtree)/vmlinuz
+	$(Q)echo "  SYMLINK $(objtree)/vmlinuz"
+
+clean:
+clean-files += *.o \
+	       vmlinu*
diff --git a/arch/mips/boot/compressed/dbg.c b/arch/mips/boot/compressed/dbg.c
new file mode 100644
index 0000000..04db05f
--- /dev/null
+++ b/arch/mips/boot/compressed/dbg.c
@@ -0,0 +1,64 @@
+/*
+ * MIPS-specific debug support for pre-boot environment
+ */
+
+#ifdef DEBUG
+
+#include <linux/types.h>
+#include <linux/serial_reg.h>
+#include <linux/init.h>
+
+#include "dbg.h"
+
+static inline unsigned int serial_in(int offset)
+{
+	return *((char *)PORT(offset));
+}
+
+static inline void serial_out(int offset, int value)
+{
+	*((char *)PORT(offset)) = value;
+}
+
+int putc(char c)
+{
+	while ((serial_in(UART_LSR) & UART_LSR_THRE) == 0)
+		;
+
+	serial_out(UART_TX, c);
+
+	return 1;
+}
+
+void puts(const char *s)
+{
+	char c;
+	while ((c = *s++) != '\0') {
+		putc(c);
+		if (c == '\n')
+			putc('\r');
+	}
+}
+
+void puthex(unsigned long long val)
+{
+
+	unsigned char buf[10];
+	int i;
+	for (i = 7; i >= 0; i--) {
+		buf[i] = "0123456789ABCDEF"[val & 0x0F];
+		val >>= 4;
+	}
+	buf[8] = '\0';
+	puts(buf);
+}
+
+
+#else
+void puts(const char *s)
+{
+}
+void puthex(unsigned long long val)
+{
+}
+#endif
diff --git a/arch/mips/boot/compressed/dbg.h b/arch/mips/boot/compressed/dbg.h
new file mode 100644
index 0000000..ee5f1fd
--- /dev/null
+++ b/arch/mips/boot/compressed/dbg.h
@@ -0,0 +1,19 @@
+/*
+ * board-specific serial port
+ */
+
+#include <asm/addrspace.h>
+
+#ifdef CONFIG_LEMOTE_FULONG
+#define UART_BASE 0x1fd003f8
+#define PORT(offset) (CKSEG1ADDR(UART_BASE) + (offset))
+#endif
+
+#ifdef CONFIG_AR7
+#include <ar7.h>
+#define PORT(offset) (CKSEG1ADDR(AR7_REGS_UART0) + (4 * offset))
+#endif
+
+#ifndef PORT
+#error please define the serial port address for your own machine
+#endif
diff --git a/arch/mips/boot/compressed/decompress.c b/arch/mips/boot/compressed/decompress.c
new file mode 100644
index 0000000..2283e61
--- /dev/null
+++ b/arch/mips/boot/compressed/decompress.c
@@ -0,0 +1,149 @@
+/*
+ * Misc. bootloader code for many machines.
+ *
+ * Copyright 2001 MontaVista Software Inc.
+ * Author: Matt Porter <mporter@mvista.com> Derived from
+ * arch/ppc/boot/prep/misc.c
+ *
+ * Copyright (C) 2009 Lemote, Inc. & Institute of Computing Technology
+ * Author: Wu Zhangjin <wuzj@lemote.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.
+ */
+
+#include <linux/types.h>
+#include <linux/kernel.h>
+
+#include <asm/addrspace.h>
+
+/* These two variables specify the free mem region
+ * that can be used for temporary malloc area
+ */
+unsigned long free_mem_ptr;
+unsigned long free_mem_end_ptr;
+char *zimage_start;
+
+/* The linker tells us where the image is. */
+extern unsigned char __image_begin, __image_end;
+extern unsigned char __ramdisk_begin, __ramdisk_end;
+unsigned long initrd_size;
+
+/* debug interfaces  */
+extern void puts(const char *s);
+extern void puthex(unsigned long long val);
+
+void error(char *x)
+{
+	puts("\n\n");
+	puts(x);
+	puts("\n\n -- System halted");
+
+	while (1)
+		;	/* Halt */
+}
+
+/* activate the code for pre-boot environment */
+#define STATIC static
+
+#ifdef CONFIG_KERNEL_GZIP
+void *memcpy(void *dest, const void *src, size_t n)
+{
+	int i;
+	const char *s = src;
+	char *d = dest;
+
+	for (i = 0; i < n; i++)
+		d[i] = s[i];
+	return dest;
+}
+#include "../../../../lib/decompress_inflate.c"
+#endif
+
+#ifdef CONFIG_KERNEL_BZIP2
+void *memset(void *s, int c, size_t n)
+{
+	int i;
+	char *ss = s;
+
+	for (i = 0; i < n; i++)
+		ss[i] = c;
+	return s;
+}
+#include "../../../../lib/decompress_bunzip2.c"
+#endif
+
+#ifdef CONFIG_KERNEL_LZMA
+#include "../../../../lib/decompress_unlzma.c"
+#endif
+
+void decompress_kernel(unsigned long load_addr, int num_words,
+		       unsigned long cksum, unsigned long boot_heap_start)
+{
+	extern unsigned long start;
+	int zimage_size;
+
+	initrd_size = (unsigned long)(&__ramdisk_end) -
+	    (unsigned long)(&__ramdisk_begin);
+
+	/*
+	 * Reveal where we were loaded at and where we
+	 * were relocated to.
+	 */
+	puts("loaded at:     ");
+	puthex(load_addr);
+	puts(" ");
+	puthex((unsigned long)(load_addr + (4 * num_words)));
+	puts("\n");
+	if ((unsigned long)load_addr != (unsigned long)&start) {
+		puts("relocated to:  ");
+		puthex((unsigned long)&start);
+		puts(" ");
+		puthex((unsigned long)((unsigned long)&start +
+				       (4 * num_words)));
+		puts("\n");
+	}
+
+	/*
+	 * We link ourself to an arbitrary low address.  When we run, we
+	 * relocate outself to that address.  __image_beign points to
+	 * the part of the image where the zImage is. -- Tom
+	 */
+	zimage_start = (char *)(unsigned long)(&__image_begin);
+	zimage_size = (unsigned long)(&__image_end) -
+	    (unsigned long)(&__image_begin);
+
+	/*
+	 * The zImage and initrd will be between start and _end, so they've
+	 * already been moved once.  We're good to go now. -- Tom
+	 */
+	puts("zimage at:     ");
+	puthex((unsigned long)zimage_start);
+	puts(" ");
+	puthex((unsigned long)(zimage_size + zimage_start));
+	puts("\n");
+
+	if (initrd_size) {
+		puts("initrd at:     ");
+		puthex((unsigned long)(&__ramdisk_begin));
+		puts(" ");
+		puthex((unsigned long)(&__ramdisk_end));
+		puts("\n");
+	}
+
+	/* this area are prepared for mallocing when decompressing */
+	free_mem_ptr = boot_heap_start;
+	free_mem_end_ptr = boot_heap_start + BOOT_HEAP_SIZE;
+
+	/* Display standard Linux/MIPS boot prompt for kernel args */
+	puts("Uncompressing Linux at load address ");
+	puthex(LOADADDR);
+	puts("\n");
+	/* Decompress the kernel with according algorithm */
+	decompress(zimage_start, zimage_size, 0, 0,
+		   (void *)LOADADDR, 0, error);
+	/* FIXME: is there a need to flush cache here? */
+	puts("Now, booting the kernel...\n");
+}
diff --git a/arch/mips/boot/compressed/dummy.c b/arch/mips/boot/compressed/dummy.c
new file mode 100644
index 0000000..31dbf45
--- /dev/null
+++ b/arch/mips/boot/compressed/dummy.c
@@ -0,0 +1,4 @@
+int main(void)
+{
+	return 0;
+}
diff --git a/arch/mips/boot/compressed/head.S b/arch/mips/boot/compressed/head.S
new file mode 100644
index 0000000..f4c29fc
--- /dev/null
+++ b/arch/mips/boot/compressed/head.S
@@ -0,0 +1,86 @@
+/*
+ * 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) 1994, 1995 Waldorf Electronics
+ * Written by Ralf Baechle and Andreas Busse
+ * Copyright (C) 1995 - 1999 Ralf Baechle
+ * Copyright (C) 1996 Paul M. Antoine
+ * Modified for DECStation and hence R3000 support by Paul M. Antoine
+ * Further modifications by David S. Miller and Harald Koerfgen
+ * Copyright (C) 1999 Silicon Graphics, Inc.
+ */
+#include <linux/autoconf.h>
+#include <linux/threads.h>
+
+#include <asm/asm.h>
+#include <asm/cacheops.h>
+#include <asm/mipsregs.h>
+#include <asm/asm-offsets.h>
+#include <asm/cachectl.h>
+#include <asm/regdef.h>
+
+	.set noreorder
+	.cprestore
+	LEAF(start)
+start:
+	bal	locate
+	nop
+locate:
+	subu	s8, ra, 8	/* Where we were loaded */
+	PTR_LA	sp, (.stack + 8192)
+
+	move	s0, a0		/* Save boot rom start args */
+	move	s1, a1
+	move	s2, a2
+	move	s3, a3
+
+	PTR_LA	a0, start	/* Where we were linked to run */
+
+	move	a1, s8
+	PTR_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 */
+	PTR_LA	a0, _edata
+	PTR_LA	a2, _end
+2:	sw	zero, 0(a0)
+	bne	a2, a0, 2b
+	addu	a0, 4
+
+	move	a0, s8		     /* load address */
+	move	a1, t1               /* length in words */
+	move	a2, t0               /* checksum */
+	PTR_LA	a3, (.heap)          /* heap address */
+
+	PTR_LA	ra, 1f
+	PTR_LA	k0, decompress_kernel
+	jr	k0
+	nop
+1:
+
+	move	a0, s0
+	move	a1, s1
+	move	a2, s2
+	move	a3, s3
+	li	k0, KERNEL_ENTRY
+	jr	k0
+	nop
+3:
+	b 3b
+	END(start)
+
+	.comm .heap, BOOT_HEAP_SIZE, 4
+	.comm .stack,4096*2,4
diff --git a/arch/mips/boot/compressed/ld.script b/arch/mips/boot/compressed/ld.script
new file mode 100644
index 0000000..29e9f4c
--- /dev/null
+++ b/arch/mips/boot/compressed/ld.script
@@ -0,0 +1,150 @@
+OUTPUT_ARCH(mips)
+ENTRY(start)
+SECTIONS
+{
+  /* Read-only sections, merged into text segment: */
+  .init          : { *(.init)		} =0
+  .text      :
+  {
+    _ftext = . ;
+    *(.text)
+    *(.rodata)
+    *(.rodata1)
+    /* .gnu.warning sections are handled specially by elf32.em.  */
+    *(.gnu.warning)
+  } =0
+  .kstrtab : { *(.kstrtab) }
+
+  . = ALIGN(16);		/* Exception table */
+  __start___ex_table = .;
+  __ex_table : { *(__ex_table) }
+  __stop___ex_table = .;
+
+  __start___dbe_table = .;	/* Exception table for data bus errors */
+  __dbe_table : { *(__dbe_table) }
+  __stop___dbe_table = .;
+
+  __start___ksymtab = .;	/* Kernel symbol table */
+  __ksymtab : { *(__ksymtab) }
+  __stop___ksymtab = .;
+
+  _etext = .;
+
+  . = ALIGN(8192);
+  .data.init_task : { *(.data.init_task) }
+
+  /* Startup code */
+  . = ALIGN(4096);
+  __init_begin = .;
+  .text.init : { *(.text.init) }
+  .data.init : { *(.data.init) }
+  . = ALIGN(16);
+  __setup_start = .;
+  .setup.init : { *(.setup.init) }
+  __setup_end = .;
+  __initcall_start = .;
+  .initcall.init : { *(.initcall.init) }
+  __initcall_end = .;
+  . = ALIGN(4096);	/* Align double page for init_task_union */
+  __init_end = .;
+
+  . = ALIGN(4096);
+  .data.page_aligned : { *(.data.idt) }
+
+  . = ALIGN(32);
+  .data.cacheline_aligned : { *(.data.cacheline_aligned) }
+
+  .fini      : { *(.fini)    } =0
+  .reginfo : { *(.reginfo) }
+  /* Adjust the address for the data segment.  We want to adjust up to
+     the same address within the page on the next page up.  It would
+     be more correct to do this:
+       . = .;
+     The current expression does not correctly handle the case of a
+     text segment ending precisely at the end of a page; it causes the
+     data segment to skip a page.  The above expression does not have
+     this problem, but it will currently (2/95) cause BFD to allocate
+     a single segment, combining both text and data, for this case.
+     This will prevent the text segment from being shared among
+     multiple executions of the program; I think that is more
+     important than losing a page of the virtual address space (note
+     that no actual memory is lost; the page which is skipped can not
+     be referenced).  */
+  . = .;
+  .data    :
+  {
+    _fdata = . ;
+    *(.data)
+
+   /* Put the compressed image here, so bss is on the end. */
+   __image_begin = .;
+   *(.image)
+   __image_end = .;
+   /* Align the initial ramdisk image (INITRD) on page boundaries. */
+   . = ALIGN(4096);
+   __ramdisk_begin = .;
+   *(.initrd)
+   __ramdisk_end = .;
+   . = ALIGN(4096);
+
+    CONSTRUCTORS
+  }
+  .data1   : { *(.data1) }
+  _gp = . + 0x8000;
+  .lit8 : { *(.lit8) }
+  .lit4 : { *(.lit4) }
+  .ctors         : { *(.ctors)   }
+  .dtors         : { *(.dtors)   }
+  .got           : { *(.got.plt) *(.got) }
+  .dynamic       : { *(.dynamic) }
+  /* We want the small data sections together, so single-instruction offsets
+     can access them all, and initialized data all before uninitialized, so
+     we can shorten the on-disk segment size.  */
+  .sdata     : { *(.sdata) }
+  . = ALIGN(4);
+  _edata  =  .;
+  PROVIDE (edata = .);
+
+  __bss_start = .;
+  _fbss = .;
+  .sbss      : { *(.sbss) *(.scommon) }
+  .bss       :
+  {
+   *(.dynbss)
+   *(.bss)
+   *(COMMON)
+   .  = ALIGN(4);
+  _end = . ;
+  PROVIDE (end = .);
+  }
+
+  /* Sections to be discarded */
+  /DISCARD/ :
+  {
+        *(.text.exit)
+        *(.data.exit)
+        *(.exitcall.exit)
+  }
+
+  /* This is the MIPS specific mdebug section.  */
+  .mdebug : { *(.mdebug) }
+  /* These are needed for ELF backends which have not yet been
+     converted to the new style linker.  */
+  .stab 0 : { *(.stab) }
+  .stabstr 0 : { *(.stabstr) }
+  /* DWARF debug sections.
+     Symbols in the .debug DWARF section are relative to the beginning of the
+     section so we begin .debug at 0.  It's not clear yet what needs to happen
+     for the others.   */
+  .debug          0 : { *(.debug) }
+  .debug_srcinfo  0 : { *(.debug_srcinfo) }
+  .debug_aranges  0 : { *(.debug_aranges) }
+  .debug_pubnames 0 : { *(.debug_pubnames) }
+  .debug_sfnames  0 : { *(.debug_sfnames) }
+  .line           0 : { *(.line) }
+  /* These must appear regardless of  .  */
+  .gptab.sdata : { *(.gptab.data) *(.gptab.sdata) }
+  .gptab.sbss : { *(.gptab.bss) *(.gptab.sbss) }
+  .comment : { *(.comment) }
+  .note : { *(.note) }
+}
-- 
1.6.2.1

Re: AR7 runtime identification [was:- Re: [PATCH -v1] MIPS: add support for gzip/bzip2/lzma compressed kernel images]

[Florian Fainelli]

Le Monday 10 August 2009 23:42:27 Florian Fainelli, vous avez écrit :
> Hi Alexander,
>
> Le Monday 10 August 2009 12:12:05 Alexander Clouter, vous avez écrit :
> > * Florian Fainelli <florian@openwrt.org> [2009-08-10 12:06:07+0200]:
> > > Hi Alexander,
> > >
> > > Le Monday 10 August 2009 11:03:15 Alexander Clouter, vous avez écrit :
> > > > I notice in arch/mips/ar7/prom.c the while loop hangs on
> > > > "(serial_in(UART_LSR) & UART_LSR_TEMT)" instead, is this because of
> > > > the buggy UART's seen in some revisions of the AR7?  As it stands
> > > > Wu's putc() implementation works fine for me but I'm guessing it is
> > > > because the UART has not much opportunity for over/underruns as not
> > > > much is being sent.
> > >
> > > Even with the silicon bug, this should not be a problem as long as you
> > > keep polling the LSR_TEMT bit in the LSR register. The problem appears
> > > when running the UART with interrupts.
> >
> > Ahhhh, that makes sense.  I'll leave Wu's code alone :)
> >
> > > > The UART on my board is buggy (missing/repeated characters like many
> > > > others seem to get) to, but I remember seeing somewhere (in a chat
> > > > you had years ago with Alan Cox if I remember correctly) the UART is
> > > > not buggy on *all* AR7 based boards?  Is is possible to detect a
> > > > buggy UART at runtime (maybe via a AR7 revision match test)?
> > >
> > > Alan Cox suggested to perform some tests which I did not carry out yet.
> > > By using two different hardware revisions in two different routers I
> > > noticed that the silicon bug is present in TNETD7300GDU revision 4,
> > > while revision 5 does not have it.
> > >
> > > The revision id can certainly help differentiating between the silicon
> > > version, later tonight when I have access to the hardware I will
> > > compare between a WAG54G and a C54APRx. I remember being them different
> > > for theTNETD7300GDU rev 4 and the rev 5.
> >
> > Well, I was offering that I could add an extra datapoint if need be.
> > There must be a way to fix up the UART without adding a 'new' UART type,
> > in a clean way.  I'll have a look into it tonight, but I imagine you
> > have looked at this in far more detail than I could...but hey, it gives
> > me something to do tonight[1]. :)
>
> For your information, the TNETD7300GDU is detected like this:
> TI AR7 (TNETD7300), ID: 0x0005, Revision: 0x02
>
> and the TNETD7300EZDW (ADSL 2+) is detected like this:
> TI AR7 (TNETD7200), ID: 0x002b, Revision: 0x10 which also has the UART bug
> and is wrongly detected as a TNETD7200.
>
> I have left the WAG54G at work and will get my hands back on it tomorow.

The bad news is that my WAG54G v2 which is also a TNEDT7300GDU has this HW bug 
too rendering the runtime detection of the bug more difficult.
-- 
Best regards, Florian Fainelli
Email: florian@openwrt.org
Web: http://openwrt.org
IRC: [florian] on irc.freenode.net
-------------------------------

Re: AR7 runtime identification [was:- Re: [PATCH -v1] MIPS: add support for gzip/bzip2/lzma compressed kernel images]

[Alexander Clouter]

Florian Fainelli <florian@openwrt.org> wrote:
>
>> Le Monday 10 August 2009 12:12:05 Alexander Clouter, vous avez écrit :
>>
>> For your information, the TNETD7300GDU is detected like this:
>> TI AR7 (TNETD7300), ID: 0x0005, Revision: 0x02
>>
>> and the TNETD7300EZDW (ADSL 2+) is detected like this:
>> TI AR7 (TNETD7200), ID: 0x002b, Revision: 0x10 which also has the UART bug
>> and is wrongly detected as a TNETD7200.
>>
>> I have left the WAG54G at work and will get my hands back on it tomorow.
> 
> The bad news is that my WAG54G v2 which is also a TNEDT7300GDU has this HW bug 
> too rendering the runtime detection of the bug more difficult.
>
Well, two options I guess.  Another Kconfig or pass something 
on the command line to the kernel.  I would opt for the latter as the 
bug does not make the machine completely unusable and if you make sure 
the workaround is disabled by default hopefully that will have the 
effect of getting people to contact you to add an extra data point.

Annoyingly I'm guess we are more interested in people who do *not* have 
the bug and we would not hear from them as a result.  Maybe if we 
proactively crippled their serial port.... :)

Cheers

-- 
Alexander Clouter
.sigmonster says: Be braver -- you can't cross a chasm in two small jumps.

Re: AR7 runtime identification [was:- Re: [PATCH -v1] MIPS: add support for gzip/bzip2/lzma compressed kernel images]

[Florian Fainelli]

Le Wednesday 12 August 2009 10:37:04 Alexander Clouter, vous avez écrit :
> Florian Fainelli <florian@openwrt.org> wrote:
> >> Le Monday 10 August 2009 12:12:05 Alexander Clouter, vous avez écrit :
> >>
> >> For your information, the TNETD7300GDU is detected like this:
> >> TI AR7 (TNETD7300), ID: 0x0005, Revision: 0x02
> >>
> >> and the TNETD7300EZDW (ADSL 2+) is detected like this:
> >> TI AR7 (TNETD7200), ID: 0x002b, Revision: 0x10 which also has the UART
> >> bug and is wrongly detected as a TNETD7200.
> >>
> >> I have left the WAG54G at work and will get my hands back on it tomorow.
> >
> > The bad news is that my WAG54G v2 which is also a TNEDT7300GDU has this
> > HW bug too rendering the runtime detection of the bug more difficult.
>
> Well, two options I guess.  Another Kconfig or pass something
> on the command line to the kernel.  I would opt for the latter as the
> bug does not make the machine completely unusable and if you make sure
> the workaround is disabled by default hopefully that will have the
> effect of getting people to contact you to add an extra data point.

Or simply enable the workaround even for sane hardware like it is done here: 
https://dev.openwrt.org/browser/trunk/target/linux/ar7/patches-2.6.30/500-serial_kludge.patch 
as this patch has no side effect on working hardware.

>
> Annoyingly I'm guess we are more interested in people who do *not* have
> the bug and we would not hear from them as a result.  Maybe if we
> proactively crippled their serial port.... :)
>
> Cheers
> --  
Cordialement, Florian Fainelli
------------------------------
-- 
Best regards, Florian Fainelli
Email: florian@openwrt.org
Web: http://openwrt.org
IRC: [florian] on irc.freenode.net
-------------------------------

Re: [PATCH -v1] MIPS: add support for gzip/bzip2/lzma compressed kernel images

[Alexander Clouter]

Hi,

* Wu Zhangjin <wuzhangjin@gmail.com> [2009-08-10 16:49:14+0800]:
>
> This patch will help to generate smaller kernel images for linux-MIPS,
> 
> Here is the effect when using lzma:
> 
> $ ls -sh vmlinux
> 7.1M vmlinux
> $ ls -sh arch/mips/boot/compressed/vmlinuz
> 1.5M arch/mips/boot/compressed/vmlinuz
> 
> Have tested the 32bit kernel on Qemu/Malta and 64bit kernel on FuLoong
> Mini PC. both of them work well.
> 
> and this revision incorporates the feedback from Alexander Clouter
> <alex@digriz.org.uk>, he helped to test it on AR7[1] based Linksys
> WAG54Gv2 and gave good suggestion on board-independence.
> 
> NOTE: this should work for the other MIPS-based machines, but I have
> used the command bc in the Makefile to calculate the load address of the
> compressed kernel. I'm not sure this is suitable.  perhaps I need to
> rewrite this part in C program or somebody help to simplify the current
> implementation.
> 
Finally Fleabay'ed a replacement WAG54Gv2 to replace the one I 
bricked to try this new patch.

Works perfectly....I finally can cooked more or less vanilla kernels on 
my board now :)

> Signed-off-by: Wu Zhangjin <wuzhangjin@gmail.com>
>
Tested-by: Alexander Clouter <alex@digriz.org.uk>

Cheers

-- 
Alexander Clouter
.sigmonster says: My computer can beat up your computer.
                  		-- Karl Lehenbauer

Re: [PATCH -v1] MIPS: add support for gzip/bzip2/lzma compressed kernel images

[Manuel Lauss]

Hi Wu!

On Mon, Aug 10, 2009 at 10:49 AM, Wu Zhangjin <wuzhangjin@gmail.com> wrote:
> This patch will help to generate smaller kernel images for linux-MIPS,

I just gave it quick spin on my Alchemy boards, works quite well!

For the debug code (dbg.[ch]) I suggest you leave the actual uart-banging
to board code (either in the form of cpp macros or inlined functions)
because most chips either don't have a 16550 compatible uart or (in the
alchemy case) have a 16550 with non-standard register layout.

Thanks you for your work!
        Manuel Lauss

Re: [PATCH -v1] MIPS: add support for gzip/bzip2/lzma compressed kernel images

[Manuel Lauss]

Hi Wu!

On Mon, Aug 10, 2009 at 10:49 AM, Wu Zhangjin <wuzhangjin@gmail.com> wrote:
> This patch will help to generate smaller kernel images for linux-MIPS,

I just gave it quick spin on my Alchemy boards, works quite well!

For the debug code (dbg.[ch]) I suggest you leave the actual uart-banging
to board code (either in the form of cpp macros or inlined functions)
because most chips either don't have a 16550 compatible uart or (in the
alchemy case) have a 16550 with non-standard register layout.

Thanks you for your work!
        Manuel Lauss

Re: [PATCH -v1] MIPS: add support for gzip/bzip2/lzma compressed kernel images

[Alexander Clouter]

Hi,

* Wu Zhangjin <wuzhangjin@gmail.com> [2009-08-10 16:49:14+0800]:
>
> This patch will help to generate smaller kernel images for linux-MIPS,
> 
> [snipped]
> 
> NOTE: this should work for the other MIPS-based machines, but I have
> used the command bc in the Makefile to calculate the load address of the
> compressed kernel. I'm not sure this is suitable.  perhaps I need to
> rewrite this part in C program or somebody help to simplify the current
> implementation.
> 
Remembered you wanted a nicer way to work out the load address so this 
is my best attempt so far....I think it's fine to leave it as shell code 
personally; but I'm not a maintainer :)

> + [snipped]
>
> +# The load address of the compressed kernel, VMLINUZ_LOAD_ADDRESS > VMLINUX_LOAD_ADDRESS + VMLINUX_SIZE
> +VMLINUX_SIZE := $(shell wc -c $(objtree)/$(KBUILD_IMAGE) 2>/dev/null | cut -d' ' -f1)
> +VMLINUX_SIZE := $(shell echo "obase=16;ibase=10;(($(VMLINUX_SIZE)/65536 + 1) * 65536)" | bc | cut -d'.' -f1)
> +VMLINUX_LOAD_ADDRESS = $(shell echo $(LOADADDR) | sed -e "s/0xffffffff//g")
> +VMLINUZ_LOAD_ADDRESS := $(shell echo "obase=16; ibase=16; ($(VMLINUX_LOAD_ADDRESS) + $(VMLINUX_SIZE))" | bc)
> +VMLINUZ_LOAD_ADDRESS := 0x$(if $(CONFIG_64BIT),ffffffff,)$(VMLINUZ_LOAD_ADDRESS)
> +LOADADDR := 0x$(if $(CONFIG_64BIT),ffffffff,)$(VMLINUX_LOAD_ADDRESS)
> +
Slightly prettier is to get rid of the 'bc' dependency, a divide and a 
'cut' here and there.

Cheers

--
Alexander Clouter
.sigmonster says: Nice guys finish last.
                                -- Leo Durocher

diff --git a/arch/mips/boot/compressed/Makefile b/arch/mips/boot/compressed/Makefile
index f9ccd97..e531b28 100644
--- a/arch/mips/boot/compressed/Makefile
+++ b/arch/mips/boot/compressed/Makefile
@@ -12,11 +12,11 @@
 # Author: Wu Zhangjin <wuzj@lemote.com>
 #
 
-# The load address of the compressed kernel, VMLINUZ_LOAD_ADDRESS > VMLINUX_LOAD_ADDRESS + VMLINUX_SIZE
-VMLINUX_SIZE := $(shell wc -c $(objtree)/$(KBUILD_IMAGE) 2>/dev/null | cut -d' ' -f1)
-VMLINUX_SIZE := $(shell echo "obase=16;ibase=10;(($(VMLINUX_SIZE)/65536 + 1) * 65536)" | bc | cut -d'.' -f1)
-VMLINUX_LOAD_ADDRESS = $(shell echo $(LOADADDR) | sed -e "s/0xffffffff//g")
-VMLINUZ_LOAD_ADDRESS := $(shell echo "obase=16; ibase=16; ($(VMLINUX_LOAD_ADDRESS) + $(VMLINUX_SIZE))" | bc)
+# compressed kernel load addr: VMLINUZ_LOAD_ADDRESS > VMLINUX_LOAD_ADDRESS + VMLINUX_SIZE
+VMLINUX_SIZE := $(shell wc -c $(objtree)/$(KBUILD_IMAGE) | cut -d' ' -f1)
+VMLINUX_SIZE := $(shell echo $$(($(VMLINUX_SIZE) + (65536 - $(VMLINUX_SIZE) % 65536))))
+VMLINUX_LOAD_ADDRESS = $(shell echo $(LOADADDR) | sed -e 's/0xffffffff//')
+VMLINUZ_LOAD_ADDRESS := $(shell printf %x $$((0x$(VMLINUX_LOAD_ADDRESS) + $(VMLINUX_SIZE))))
 VMLINUZ_LOAD_ADDRESS := 0x$(if $(CONFIG_64BIT),ffffffff,)$(VMLINUZ_LOAD_ADDRESS)
 LOADADDR := 0x$(if $(CONFIG_64BIT),ffffffff,)$(VMLINUX_LOAD_ADDRESS)

Re: [PATCH -v1] MIPS: add support for gzip/bzip2/lzma compressed kernel images

[Ralf Baechle]

On Mon, Aug 10, 2009 at 04:49:14PM +0800, Wu Zhangjin wrote:

> This patch will help to generate smaller kernel images for linux-MIPS,
> 
> Here is the effect when using lzma:
> 
> $ ls -sh vmlinux
> 7.1M vmlinux
> $ ls -sh arch/mips/boot/compressed/vmlinuz
> 1.5M arch/mips/boot/compressed/vmlinuz
> 
> Have tested the 32bit kernel on Qemu/Malta and 64bit kernel on FuLoong
> Mini PC. both of them work well.

  ip27_defconfig:

  CHK     include/linux/compile.h
(standard_in) 1: parse error
(standard_in) 1: parse error
  AS      arch/mips/boot/compressed/head.o
arch/mips/boot/compressed/head.S: Assembler messages:
arch/mips/boot/compressed/head.S:78: Error: Number (0xa800000000021bf0) larger than 32 bits
make[1]: *** [arch/mips/boot/compressed/head.o] Error 1

  rm200_defconfig:

$ make
[...]
  IHEX2FW firmware/whiteheat.fw
  IHEX2FW firmware/keyspan_pda/keyspan_pda.fw
  IHEX2FW firmware/keyspan_pda/xircom_pgs.fw
[ralf@h5 linux-queue]$ ll vmlinuz 
lrwxrwxrwx 1 ralf ralf 33 2009-10-12 03:30 vmlinuz -> arch/mips/boot/compressed/vmlinuz*
[ralf@h5 linux-queue]$ ll arch/mips/boot/compressed/vmlinuz 
-rwxrwxr-x 1 ralf ralf 1675346 2009-10-12 03:30 arch/mips/boot/compressed/vmlinuz*
[ralf@h5 linux-queue]$ make distclean
  CLEAN   arch/mips/boot
(standard_in) 1: parse error
[...]

For rm200 we need an ECOFF zimage.

malta_defconfig:

$ mkdir ../build-malta
$ make O=../build-malta malta_defconfig
[...]
$ cd ../build-malta
$ make
[...]
$ make distclean
make -C /home/ralf/src/linux/linux-queue O=/home/ralf/src/linux/build-malta/. distclean
  CLEAN   arch/mips/boot
(standard_in) 1: parse error
$

By default we generate a binary that is not ELF file for malta, so a
compressed kernel should be in that format also.

$ make help
[...]
Architecture specific targets (mips):
  install              - install kernel into /boot
  vmlinux.ecoff        - ECOFF boot image
  vmlinux.bin          - Raw binary boot image
  vmlinux.srec         - SREC boot image
  vmlinuz              - Compressed boot image
[...]
$ make vmlinuz
make: *** No rule to make target `vmlinuz'.  Stop.
$

> diff --git a/arch/mips/Makefile b/arch/mips/Makefile
> index 861da51..300b996 100644
> --- a/arch/mips/Makefile
> +++ b/arch/mips/Makefile
> @@ -709,7 +709,10 @@ vmlinux.64: vmlinux
>  
>  makeboot =$(Q)$(MAKE) $(build)=arch/mips/boot VMLINUX=$(vmlinux-32) $(1)
>  
> -all:	$(all-y)
> +all:	$(all-y) zImage
> +
> +zImage: vmlinux
> +	$(Q)$(MAKE) $(build)=arch/mips/boot/compressed $@ LOADADDR=$(load-y)

Several systems rely on the firmware for memory usage information.  With
a compressed kernel this information will be inaccurate.  This is a
somewhat hard problem to solve so I suggest to only enable compressed
kernels on systems where they're known to work.

  Ralf

Re: [PATCH -v1] MIPS: add support for gzip/bzip2/lzma compressed kernel images

[Wu Zhangjin]

Hi,

On Mon, 2009-10-12 at 04:34 +0200, Ralf Baechle wrote:
> On Mon, Aug 10, 2009 at 04:49:14PM +0800, Wu Zhangjin wrote:
> 
> > This patch will help to generate smaller kernel images for linux-MIPS,
> > 
> > Here is the effect when using lzma:
> > 
> > $ ls -sh vmlinux
> > 7.1M vmlinux
> > $ ls -sh arch/mips/boot/compressed/vmlinuz
> > 1.5M arch/mips/boot/compressed/vmlinuz
> > 
> > Have tested the 32bit kernel on Qemu/Malta and 64bit kernel on FuLoong
> > Mini PC. both of them work well.
> 
>   ip27_defconfig:
> 
>   CHK     include/linux/compile.h
> (standard_in) 1: parse error
> (standard_in) 1: parse error
>   AS      arch/mips/boot/compressed/head.o
> arch/mips/boot/compressed/head.S: Assembler messages:
> arch/mips/boot/compressed/head.S:78: Error: Number (0xa800000000021bf0) larger than 32 bits
> make[1]: *** [arch/mips/boot/compressed/head.o] Error 1
> 

Seems something wrong in the Makefile, will fix it later, here it is:

[...]
/bin/sh: 0x0xa80000000001c000: value too great for base (error token is
"0x0xa80000000001c000")
  Building modules, stage 2.
  AS      arch/mips/boot/compressed/head.o
  CC      arch/mips/boot/compressed/decompress.o
  GZIP    arch/mips/boot/compressed/vmlinux.gz
  CC      arch/mips/boot/compressed/dbg.o
  MODPOST 137 modules
  CC      arch/mips/boot/compressed/dummy.o
arch/mips/boot/compressed/head.S: Assembler messages:
arch/mips/boot/compressed/head.S:47: Error: Number (0xa800000000021bf0)
larger than 32 bits
[...]

>   rm200_defconfig:
> 
> $ make
> [...]
>   IHEX2FW firmware/whiteheat.fw
>   IHEX2FW firmware/keyspan_pda/keyspan_pda.fw
>   IHEX2FW firmware/keyspan_pda/xircom_pgs.fw
> [ralf@h5 linux-queue]$ ll vmlinuz 
> lrwxrwxrwx 1 ralf ralf 33 2009-10-12 03:30 vmlinuz -> arch/mips/boot/compressed/vmlinuz*
> [ralf@h5 linux-queue]$ ll arch/mips/boot/compressed/vmlinuz 
> -rwxrwxr-x 1 ralf ralf 1675346 2009-10-12 03:30 arch/mips/boot/compressed/vmlinuz*
> [ralf@h5 linux-queue]$ make distclean
>   CLEAN   arch/mips/boot
> (standard_in) 1: parse error
> [...]
> 
> For rm200 we need an ECOFF zimage.
> 

Okay, will fix it later.

> malta_defconfig:
> 
> $ mkdir ../build-malta
> $ make O=../build-malta malta_defconfig
> [...]
> $ cd ../build-malta
> $ make
> [...]
> $ make distclean
> make -C /home/ralf/src/linux/linux-queue O=/home/ralf/src/linux/build-malta/. distclean
>   CLEAN   arch/mips/boot
> (standard_in) 1: parse error

This "parse error" will be fixed by tuning the Makefile via adding
something like " blabla 2>/dev/null".

> $
> 
> By default we generate a binary that is not ELF file for malta, so a
> compressed kernel should be in that format also.
> 
> $ make help
> [...]
> Architecture specific targets (mips):
>   install              - install kernel into /boot
>   vmlinux.ecoff        - ECOFF boot image
>   vmlinux.bin          - Raw binary boot image
>   vmlinux.srec         - SREC boot image
>   vmlinuz              - Compressed boot image
> [...]
> $ make vmlinuz
> make: *** No rule to make target `vmlinuz'.  Stop.
> $
> 

it should be

$ make zImage

will replace "vmlinuz - Compressed boot image" to "zImage - Compressed
boot image", or, fix "make vmlinuz" itself.

> > diff --git a/arch/mips/Makefile b/arch/mips/Makefile
> > index 861da51..300b996 100644
> > --- a/arch/mips/Makefile
> > +++ b/arch/mips/Makefile
> > @@ -709,7 +709,10 @@ vmlinux.64: vmlinux
> >  
> >  makeboot =$(Q)$(MAKE) $(build)=arch/mips/boot VMLINUX=$(vmlinux-32) $(1)
> >  
> > -all:	$(all-y)
> > +all:	$(all-y) zImage
> > +
> > +zImage: vmlinux
> > +	$(Q)$(MAKE) $(build)=arch/mips/boot/compressed $@ LOADADDR=$(load-y)
> 
> Several systems rely on the firmware for memory usage information.  With
> a compressed kernel this information will be inaccurate.  This is a
> somewhat hard problem to solve so I suggest to only enable compressed
> kernels on systems where they're known to work.

Okay, I will add a new kernel option for this, and also, the
DEBUG(dbg.[hc]) part will be selectable with a new kernel option too.

Thanks & Regards,
	Wu Zhangjin