Merge branch 'master' of git://git.denx.de/u-boot-arm

22 files changed, 1085 insertions, 762 deletions

diff --git a/arch/arm/cpu/arm1136/start.S b/arch/arm/cpu/arm1136/start.S
index 5008ac6..29ed065 100644
--- a/arch/arm/cpu/arm1136/start.S
+++ b/arch/arm/cpu/arm1136/start.S
@@ -89,48 +89,35 @@ _end_vect:
 _TEXT_BASE:
 	.word	CONFIG_SYS_TEXT_BASE
 
-#if defined(CONFIG_SYS_ARM_WITHOUT_RELOC)
-.globl _armboot_start
-_armboot_start:
-	.word _start
-#endif
-
 /*
  * These are defined in the board-specific linker script.
+ * Subtracting _start from them lets the linker put their
+ * relative position in the executable instead of leaving
+ * them null.
  */
-.globl _bss_start
-_bss_start:
-	.word __bss_start
-
-.globl _bss_end
-_bss_end:
-	.word _end
-
-#if !defined(CONFIG_SYS_ARM_WITHOUT_RELOC)
-.globl _datarel_start
-_datarel_start:
-	.word __datarel_start
+.globl _bss_start_ofs
+_bss_start_ofs:
+	.word __bss_start - _start
 
-.globl _datarelrolocal_start
-_datarelrolocal_start:
-	.word __datarelrolocal_start
+.globl _bss_end_ofs
+_bss_end_ofs:
+	.word _end - _start
 
-.globl _datarellocal_start
-_datarellocal_start:
-	.word __datarellocal_start
+.globl _datarel_start_ofs
+_datarel_start_ofs:
+	.word __datarel_start - _start
 
-.globl _datarelro_start
-_datarelro_start:
-	.word __datarelro_start
+.globl _datarelrolocal_start_ofs
+_datarelrolocal_start_ofs:
+	.word __datarelrolocal_start - _start
 
-.globl _got_start
-_got_start:
-	.word __got_start
+.globl _datarellocal_start_ofs
+_datarellocal_start_ofs:
+	.word __datarellocal_start - _start
 
-.globl _got_end
-_got_end:
-	.word __got_end
-#endif
+.globl _datarelro_start_ofs
+_datarelro_start_ofs:
+	.word __datarelro_start - _start
 
 #ifdef CONFIG_USE_IRQ
 /* IRQ stack memory (calculated at run-time) */
@@ -225,9 +212,8 @@ stack_setup:
 
 	adr	r0, _start
 	ldr	r2, _TEXT_BASE
-	ldr	r3, _bss_start
-	sub	r2, r3, r2		/* r2 <- size of armboot	    */
-	add	r2, r0, r2		/* r2 <- source end address	    */
+	ldr	r3, _bss_start_ofs
+	add	r2, r0, r3		/* r2 <- source end address	    */
 	cmp	r0, r6
 	beq	clear_bss
 
@@ -239,36 +225,54 @@ copy_loop:
 	blo	copy_loop
 
 #ifndef CONFIG_PRELOADER
-	/* fix got entries */
-	ldr	r1, _TEXT_BASE
-	mov	r0, r7			/* reloc addr */
-	ldr	r2, _got_start		/* addr in Flash */
-	ldr	r3, _got_end		/* addr in Flash */
-	sub	r3, r3, r1
-	add	r3, r3, r0
-	sub	r2, r2, r1
-	add	r2, r2, r0
-
+	/*
+	 * fix .rel.dyn relocations
+	 */
+	ldr	r0, _TEXT_BASE		/* r0 <- Text base */
+	sub	r9, r7, r0		/* r9 <- relocation offset */
+	ldr	r10, _dynsym_start_ofs	/* r10 <- sym table ofs */
+	add	r10, r10, r0		/* r10 <- sym table in FLASH */
+	ldr	r2, _rel_dyn_start_ofs	/* r2 <- rel dyn start ofs */
+	add	r2, r2, r0		/* r2 <- rel dyn start in FLASH */
+	ldr	r3, _rel_dyn_end_ofs	/* r3 <- rel dyn end ofs */
+	add	r3, r3, r0		/* r3 <- rel dyn end in FLASH */
 fixloop:
-	ldr	r4, [r2]
-	sub	r4, r4, r1
-	add	r4, r4, r0
-	str	r4, [r2]
-	add	r2, r2, #4
+	ldr	r0, [r2]	/* r0 <- location to fix up, IN FLASH! */
+	add	r0, r9		/* r0 <- location to fix up in RAM */
+	ldr	r1, [r2, #4]
+	and	r8, r1, #0xff
+	cmp	r8, #23		/* relative fixup? */
+	beq	fixrel
+	cmp	r8, #2		/* absolute fixup? */
+	beq	fixabs
+	/* ignore unknown type of fixup */
+	b	fixnext
+fixabs:
+	/* absolute fix: set location to (offset) symbol value */
+	mov	r1, r1, LSR #4		/* r1 <- symbol index in .dynsym */
+	add	r1, r10, r1		/* r1 <- address of symbol in table */
+	ldr	r1, [r1, #4]		/* r1 <- symbol value */
+	add	r1, r9			/* r1 <- relocated sym addr */
+	b	fixnext
+fixrel:
+	/* relative fix: increase location by offset */
+	ldr	r1, [r0]
+	add	r1, r1, r9
+fixnext:
+	str	r1, [r0]
+	add	r2, r2, #8	/* each rel.dyn entry is 8 bytes */
 	cmp	r2, r3
-	bne	fixloop
+	ble	fixloop
 #endif
 #endif	/* #ifndef CONFIG_SKIP_RELOCATE_UBOOT */
 
 clear_bss:
 #ifndef CONFIG_PRELOADER
-	ldr	r0, _bss_start
-	ldr	r1, _bss_end
+	ldr	r0, _bss_start_ofs
+	ldr	r1, _bss_end_ofs
 	ldr	r3, _TEXT_BASE		/* Text base */
 	mov	r4, r7			/* reloc addr */
-	sub	r0, r0, r3
 	add	r0, r0, r4
-	sub	r1, r1, r3
 	add	r1, r1, r4
 	mov	r2, #0x00000000		/* clear			    */
 
@@ -283,24 +287,34 @@ clbss_l:str	r2, [r0]		/* clear loop...		    */
  * initialization, now running from RAM.
  */
 #ifdef CONFIG_NAND_SPL
-	ldr     pc, _nand_boot
-
-_nand_boot: .word nand_boot
+	ldr     r0, _nand_boot_ofs
+	adr	r1, _start
+	add	pc, r0, r1
+_nand_boot_ofs
+	: .word nand_boot - _start
 #else
 jump_2_ram:
-	ldr	r0, _TEXT_BASE
-	ldr	r2, _board_init_r
-	sub	r2, r2, r0
-	add	r2, r2, r7	/* position from board_init_r in RAM */
+	ldr	r0, _board_init_r_ofs
+	adr	r1, _start
+	add	r0, r0, r1
+	add	lr, r0, r9
 	/* setup parameters for board_init_r */
 	mov	r0, r5		/* gd_t */
 	mov	r1, r7		/* dest_addr */
 	/* jump to it ... */
-	mov	lr, r2
 	mov	pc, lr
 
-_board_init_r: .word board_init_r
+_board_init_r_ofs:
+	.word board_init_r - _start
 #endif
+
+_rel_dyn_start_ofs:
+	.word __rel_dyn_start - _start
+_rel_dyn_end_ofs:
+	.word __rel_dyn_end - _start
+_dynsym_start_ofs:
+	.word __dynsym_start - _start
+
 #else /* #if !defined(CONFIG_SYS_ARM_WITHOUT_RELOC) */
 /*
  * the actual reset code
@@ -375,8 +389,11 @@ stack_setup:
 	bic	sp, sp, #7		/* 8-byte alignment for ABI compliance */
 
 clear_bss:
-	ldr	r0, _bss_start		/* find start of bss segment	    */
-	ldr	r1, _bss_end		/* stop here			    */
+	adr	r2, _start
+	ldr	r0, _bss_start_ofs	/* find start of bss segment        */
+	add	r0, r0, r2
+	ldr	r1, _bss_end_ofs	/* stop here                        */
+	add	r1, r1, r2
 	mov	r2, #0x00000000		/* clear			    */
 
 #ifndef CONFIG_PRELOADER
@@ -386,15 +403,19 @@ clbss_l:str	r2, [r0]		/* clear loop...		    */
 	bne	clbss_l
 #endif
 
-	ldr	pc, _start_armboot
+	ldr	r0, _start_armboot_ofs
+	adr	r1, _start
+	add	r0, r0, r1
+	ldr	pc, r0
 
+_start_armboot_ofs:
 #ifdef CONFIG_NAND_SPL
-_start_armboot: .word nand_boot
+	.word nand_boot - _start
 #else
 #ifdef CONFIG_ONENAND_IPL
-_start_armboot: .word start_oneboot
+	.word start_oneboot - _start
 #else
-_start_armboot: .word start_armboot
+	.word start_armboot - _start
 #endif /* CONFIG_ONENAND_IPL */
 #endif /* CONFIG_NAND_SPL */
 
@@ -487,7 +508,7 @@ cpu_init_crit:
 #if !defined(CONFIG_SYS_ARM_WITHOUT_RELOC)
 	ldr	r2, IRQ_STACK_START_IN		@ set base 2 words into abort stack
 #else
-	ldr	r2, _armboot_start
+	adr	r2, _start
 	sub	r2, r2, #(CONFIG_SYS_MALLOC_LEN)
 	sub	r2, r2, #(CONFIG_SYS_GBL_DATA_SIZE+8)	@ set base 2 words into abort stack
 #endif
@@ -524,8 +545,8 @@ cpu_init_crit:
 #if !defined(CONFIG_SYS_ARM_WITHOUT_RELOC)
 	ldr	r13, IRQ_STACK_START_IN		@ setup our mode stack (enter in banked mode)
 #else
-	ldr	r13, _armboot_start		@ setup our mode stack (enter in banked mode)
-	sub	r13, r13, #(CONFIG_SYS_MALLOC_LEN)	@ move past malloc pool
+	adr	r13, _start			@ setup our mode stack (enter in banked mode)
+	sub	r13, r13, #(CONFIG_STACKSIZE+CONFIG_SYS_MALLOC_LEN)	@ move past malloc pool
 	sub	r13, r13, #(CONFIG_SYS_GBL_DATA_SIZE+8) @ move to reserved a couple spots for abort stack
 #endif
 
diff --git a/arch/arm/cpu/arm1136/u-boot.lds b/arch/arm/cpu/arm1136/u-boot.lds
index 1db4b49..31f43f0 100644
--- a/arch/arm/cpu/arm1136/u-boot.lds
+++ b/arch/arm/cpu/arm1136/u-boot.lds
@@ -59,11 +59,14 @@ SECTIONS
 		*(.data.rel.ro)
 	}
 
-	__got_start = .;
 	. = ALIGN(4);
-	.got : { *(.got) }
+	__rel_dyn_start = .;
+	.rel.dyn : { *(.rel.dyn) }
+	__rel_dyn_end = .;
+
+	__dynsym_start = .;
+	.dynsym : { *(.dynsym) }
 
-	__got_end = .;
 	. = .;
 	__u_boot_cmd_start = .;
 	.u_boot_cmd : { *(.u_boot_cmd) }
@@ -73,4 +76,10 @@ SECTIONS
 	__bss_start = .;
 	.bss (NOLOAD) : { *(.bss) . = ALIGN(4); }
 	_end = .;
+
+	/DISCARD/ : { *(.dynstr*) }
+	/DISCARD/ : { *(.dynamic*) }
+	/DISCARD/ : { *(.plt*) }
+	/DISCARD/ : { *(.interp*) }
+	/DISCARD/ : { *(.gnu*) }
 }
diff --git a/arch/arm/cpu/arm926ejs/at91/clock.c b/arch/arm/cpu/arm926ejs/at91/clock.c
index ecf91f5..7a10a77 100644
--- a/arch/arm/cpu/arm926ejs/at91/clock.c
+++ b/arch/arm/cpu/arm926ejs/at91/clock.c
@@ -11,47 +11,46 @@
  * (at your option) any later version.
  */
 
-#include <config.h>
+#include <common.h>
 #include <asm/arch/hardware.h>
 #include <asm/arch/io.h>
 #include <asm/arch/at91_pmc.h>
 #include <asm/arch/clk.h>
 
-static unsigned long cpu_clk_rate_hz;
-static unsigned long main_clk_rate_hz;
-static unsigned long mck_rate_hz;
-static unsigned long plla_rate_hz;
-static unsigned long pllb_rate_hz;
-static u32 at91_pllb_usb_init;
+#if !defined(CONFIG_AT91FAMILY)
+# error You need to define CONFIG_AT91FAMILY in your board config!
+#endif
+
+DECLARE_GLOBAL_DATA_PTR;
 
 unsigned long get_cpu_clk_rate(void)
 {
-	return cpu_clk_rate_hz;
+	return gd->cpu_clk_rate_hz;
 }
 
 unsigned long get_main_clk_rate(void)
 {
-	return main_clk_rate_hz;
+	return gd->main_clk_rate_hz;
 }
 
 unsigned long get_mck_clk_rate(void)
 {
-	return mck_rate_hz;
+	return gd->mck_rate_hz;
 }
 
 unsigned long get_plla_clk_rate(void)
 {
-	return plla_rate_hz;
+	return gd->plla_rate_hz;
 }
 
 unsigned long get_pllb_clk_rate(void)
 {
-	return pllb_rate_hz;
+	return gd->pllb_rate_hz;
 }
 
 u32 get_pllb_init(void)
 {
-	return at91_pllb_usb_init;
+	return gd->at91_pllb_usb_init;
 }
 
 static unsigned long at91_css_to_rate(unsigned long css)
@@ -60,11 +59,11 @@ static unsigned long at91_css_to_rate(unsigned long css)
 	case AT91_PMC_MCKR_CSS_SLOW:
 		return AT91_SLOW_CLOCK;
 	case AT91_PMC_MCKR_CSS_MAIN:
-		return main_clk_rate_hz;
+		return gd->main_clk_rate_hz;
 	case AT91_PMC_MCKR_CSS_PLLA:
-		return plla_rate_hz;
+		return gd->plla_rate_hz;
 	case AT91_PMC_MCKR_CSS_PLLB:
-		return pllb_rate_hz;
+		return gd->pllb_rate_hz;
 	}
 
 	return 0;
@@ -163,10 +162,10 @@ int at91_clock_init(unsigned long main_clock)
 		main_clock = tmp * (AT91_SLOW_CLOCK / 16);
 	}
 #endif
-	main_clk_rate_hz = main_clock;
+	gd->main_clk_rate_hz = main_clock;
 
 	/* report if PLLA is more than mildly overclocked */
-	plla_rate_hz = at91_pll_rate(main_clock, readl(&pmc->pllar));
+	gd->plla_rate_hz = at91_pll_rate(main_clock, readl(&pmc->pllar));
 
 #ifdef CONFIG_USB_ATMEL
 	/*
@@ -175,9 +174,9 @@ int at91_clock_init(unsigned long main_clock)
 	 *
 	 * REVISIT:  assumes MCK doesn't derive from PLLB!
 	 */
-	at91_pllb_usb_init = at91_pll_calc(main_clock, 48000000 * 2) |
+	gd->at91_pllb_usb_init = at91_pll_calc(main_clock, 48000000 * 2) |
 			     AT91_PMC_PLLBR_USBDIV_2;
-	pllb_rate_hz = at91_pll_rate(main_clock, at91_pllb_usb_init);
+	gd->pllb_rate_hz = at91_pll_rate(main_clock, gd->at91_pllb_usb_init);
 #endif
 
 	/*
@@ -187,30 +186,30 @@ int at91_clock_init(unsigned long main_clock)
 	mckr = readl(&pmc->mckr);
 #if defined(CONFIG_AT91SAM9G45) || defined(CONFIG_AT91SAM9M10G45)
 	/* plla divisor by 2 */
-	plla_rate_hz /= (1 << ((mckr & 1 << 12) >> 12));
+	gd->plla_rate_hz /= (1 << ((mckr & 1 << 12) >> 12));
 #endif
-	mck_rate_hz = at91_css_to_rate(mckr & AT91_PMC_MCKR_CSS_MASK);
-	freq = mck_rate_hz;
+	gd->mck_rate_hz = at91_css_to_rate(mckr & AT91_PMC_MCKR_CSS_MASK);
+	freq = gd->mck_rate_hz;
 
 	freq /= (1 << ((mckr & AT91_PMC_MCKR_PRES_MASK) >> 2));	/* prescale */
 #if defined(CONFIG_AT91RM9200)
 	/* mdiv */
-	mck_rate_hz = freq / (1 + ((mckr & AT91_PMC_MCKR_MDIV_MASK) >> 8));
+	gd->mck_rate_hz = freq / (1 + ((mckr & AT91_PMC_MCKR_MDIV_MASK) >> 8));
 #elif defined(CONFIG_AT91SAM9G20)
 	/* mdiv ; (x >> 7) = ((x >> 8) * 2) */
-	mck_rate_hz = (mckr & AT91_PMC_MCKR_MDIV_MASK) ?
+	gd->mck_rate_hz = (mckr & AT91_PMC_MCKR_MDIV_MASK) ?
 		freq / ((mckr & AT91_PMC_MCKR_MDIV_MASK) >> 7) : freq;
 	if (mckr & AT91_PMC_MCKR_MDIV_MASK)
 		freq /= 2;			/* processor clock division */
 #elif defined(CONFIG_AT91SAM9G45) || defined(CONFIG_AT91SAM9M10G45)
-	mck_rate_hz = (mckr & AT91_PMC_MCKR_MDIV_MASK) ==
+	gd->mck_rate_hz = (mckr & AT91_PMC_MCKR_MDIV_MASK) ==
 		(AT91_PMC_MCKR_MDIV_2 | AT91_PMC_MCKR_MDIV_4)
 		? freq / 3
 		: freq / (1 << ((mckr & AT91_PMC_MCKR_MDIV_MASK) >> 8));
 #else
-	mck_rate_hz = freq / (1 << ((mckr & AT91_PMC_MCKR_MDIV_MASK) >> 8));
+	gd->mck_rate_hz = freq / (1 << ((mckr & AT91_PMC_MCKR_MDIV_MASK) >> 8));
 #endif
-	cpu_clk_rate_hz = freq;
+	gd->cpu_clk_rate_hz = freq;
 
 	return 0;
 }
diff --git a/arch/arm/cpu/arm926ejs/at91/timer.c b/arch/arm/cpu/arm926ejs/at91/timer.c
index 8efc34b..82b8d7e 100644
--- a/arch/arm/cpu/arm926ejs/at91/timer.c
+++ b/arch/arm/cpu/arm926ejs/at91/timer.c
@@ -30,55 +30,63 @@
 #include <asm/arch/io.h>
 #include <div64.h>
 
+#if !defined(CONFIG_AT91FAMILY)
+# error You need to define CONFIG_AT91FAMILY in your board config!
+#endif
+
+DECLARE_GLOBAL_DATA_PTR;
+
 /*
  * We're using the AT91CAP9/SAM9 PITC in 32 bit mode, by
  * setting the 20 bit counter period to its maximum (0xfffff).
+ * (See the relevant data sheets to understand that this really works)
+ *
+ * We do also mimic the typical powerpc way of incrementing
+ * two 32 bit registers called tbl and tbu.
+ *
+ * Those registers increment at 1/16 the main clock rate.
  */
-#define TIMER_LOAD_VAL	0xfffff
 
-static ulong timestamp;
-static ulong lastinc;
-static ulong timer_freq;
+#define TIMER_LOAD_VAL	0xfffff
 
 static inline unsigned long long tick_to_time(unsigned long long tick)
 {
 	tick *= CONFIG_SYS_HZ;
-	do_div(tick, timer_freq);
+	do_div(tick, gd->timer_rate_hz);
 
 	return tick;
 }
 
 static inline unsigned long long usec_to_tick(unsigned long long usec)
 {
-	usec *= timer_freq;
+	usec *= gd->timer_rate_hz;
 	do_div(usec, 1000000);
 
 	return usec;
 }
 
-/* nothing really to do with interrupts, just starts up a counter. */
+/*
+ * Use the PITC in full 32 bit incrementing mode
+ */
 int timer_init(void)
 {
 	at91_pmc_t *pmc = (at91_pmc_t *) AT91_PMC_BASE;
 	at91_pit_t *pit = (at91_pit_t *) AT91_PIT_BASE;
-	/*
-	 * Enable PITC Clock
-	 * The clock is already enabled for system controller in boot
-	 */
+
+	/* Enable PITC Clock */
 	writel(1 << AT91_ID_SYS, &pmc->pcer);
 
 	/* Enable PITC */
 	writel(TIMER_LOAD_VAL | AT91_PIT_MR_EN , &pit->mr);
 
-	reset_timer_masked();
-
-	timer_freq = get_mck_clk_rate() >> 4;
+	gd->timer_rate_hz = gd->mck_rate_hz / 16;
+	gd->tbu = gd->tbl = 0;
 
 	return 0;
 }
 
 /*
- * timer without interrupts
+ * Get the current 64 bit timer tick count
  */
 unsigned long long get_ticks(void)
 {
@@ -86,28 +94,11 @@ unsigned long long get_ticks(void)
 
 	ulong now = readl(&pit->piir);
 
-	if (now >= lastinc)	/* normal mode (non roll) */
-		/* move stamp forward with absolut diff ticks */
-		timestamp += (now - lastinc);
-	else			/* we have rollover of incrementer */
-		timestamp += (0xFFFFFFFF - lastinc) + now;
-	lastinc = now;
-	return timestamp;
-}
-
-void reset_timer_masked(void)
-{
-	/* reset time */
-	at91_pit_t *pit = (at91_pit_t *) AT91_PIT_BASE;
-
-	/* capture current incrementer value time */
-	lastinc = readl(&pit->piir);
-	timestamp = 0; /* start "advancing" time stamp from 0 */
-}
-
-ulong get_timer_masked(void)
-{
-	return tick_to_time(get_ticks());
+	/* increment tbu if tbl has rolled over */
+	if (now < gd->tbl)
+		gd->tbu++;
+	gd->tbl = now;
+	return (((unsigned long long)gd->tbu) << 32) | gd->tbl;
 }
 
 void __udelay(unsigned long usec)
@@ -119,24 +110,32 @@ void __udelay(unsigned long usec)
 	tmp = get_ticks() + tmo;	/* get current timestamp */
 
 	while (get_ticks() < tmp)	/* loop till event */
-		 /*NOP*/;
+		;
 }
 
+/*
+ * reset_timer() and get_timer(base) are a pair of functions that are used by
+ * some timeout/sleep mechanisms in u-boot.
+ *
+ * reset_timer() marks the current time as epoch and
+ * get_timer(base) works relative to that epoch.
+ *
+ * The time is used in CONFIG_SYS_HZ units!
+ */
 void reset_timer(void)
 {
-	reset_timer_masked();
+	gd->timer_reset_value = get_ticks();
 }
 
 ulong get_timer(ulong base)
 {
-	return get_timer_masked () - base;
+	return tick_to_time(get_ticks() - gd->timer_reset_value) - base;
 }
 
 /*
- * This function is derived from PowerPC code (timebase clock frequency).
- * On ARM it returns the number of timer ticks per second.
+ * Return the number of timer ticks per second.
  */
 ulong get_tbclk(void)
 {
-	return timer_freq;
+	return gd->timer_rate_hz;
 }
diff --git a/arch/arm/cpu/arm926ejs/start.S b/arch/arm/cpu/arm926ejs/start.S
index 8cbe3e7..863de3b 100644
--- a/arch/arm/cpu/arm926ejs/start.S
+++ b/arch/arm/cpu/arm926ejs/start.S
@@ -10,6 +10,7 @@
  *  Copyright (c) 2002	Gary Jennejohn <garyj@denx.de>
  *  Copyright (c) 2003	Richard Woodruff <r-woodruff2@ti.com>
  *  Copyright (c) 2003	Kshitij <kshitij@ti.com>
+ *  Copyright (c) 2010	Albert Aribaud <albert.aribaud@free.fr>
  *
  * See file CREDITS for list of people who contributed to this
  * project.
@@ -118,22 +119,19 @@ _fiq:
 _TEXT_BASE:
 	.word	CONFIG_SYS_TEXT_BASE
 
-#if defined(CONFIG_SYS_ARM_WITHOUT_RELOC)
-.globl _armboot_start
-_armboot_start:
-	.word _start
-#endif
-
 /*
  * These are defined in the board-specific linker script.
+ * Subtracting _start from them lets the linker put their
+ * relative position in the executable instead of leaving
+ * them null.
  */
-.globl _bss_start
-_bss_start:
-	.word __bss_start
+.globl _bss_start_ofs
+_bss_start_ofs:
+	.word __bss_start - _start
 
-.globl _bss_end
-_bss_end:
-	.word _end
+.globl _bss_end_ofs
+_bss_end_ofs:
+	.word _end - _start
 
 #ifdef CONFIG_USE_IRQ
 /* IRQ stack memory (calculated at run-time) */
@@ -153,30 +151,6 @@ FIQ_STACK_START:
 IRQ_STACK_START_IN:
 	.word	0x0badc0de
 
-.globl _datarel_start
-_datarel_start:
-	.word __datarel_start
-
-.globl _datarelrolocal_start
-_datarelrolocal_start:
-	.word __datarelrolocal_start
-
-.globl _datarellocal_start
-_datarellocal_start:
-	.word __datarellocal_start
-
-.globl _datarelro_start
-_datarelro_start:
-	.word __datarelro_start
-
-.globl _got_start
-_got_start:
-	.word __got_start
-
-.globl _got_end
-_got_end:
-	.word __got_end
-
 /*
  * the actual reset code
  */
@@ -226,9 +200,8 @@ stack_setup:
 
 	adr	r0, _start
 	ldr	r2, _TEXT_BASE
-	ldr	r3, _bss_start
-	sub	r2, r3, r2		/* r2 <- size of armboot	    */
-	add	r2, r0, r2		/* r2 <- source end address	    */
+	ldr	r3, _bss_start_ofs
+	add	r2, r0, r3		/* r2 <- source end address	    */
 	cmp	r0, r6
 	beq	clear_bss
 
@@ -240,36 +213,54 @@ copy_loop:
 	blo	copy_loop
 
 #ifndef CONFIG_PRELOADER
-	/* fix got entries */
-	ldr	r1, _TEXT_BASE		/* Text base */
-	mov	r0, r7			/* reloc addr */
-	ldr	r2, _got_start		/* addr in Flash */
-	ldr	r3, _got_end		/* addr in Flash */
-	sub	r3, r3, r1
-	add	r3, r3, r0
-	sub	r2, r2, r1
-	add	r2, r2, r0
-
+	/*
+	 * fix .rel.dyn relocations
+	 */
+	ldr	r0, _TEXT_BASE		/* r0 <- Text base */
+	sub	r9, r7, r0		/* r9 <- relocation offset */
+	ldr	r10, _dynsym_start_ofs	/* r10 <- sym table ofs */
+	add	r10, r10, r0		/* r10 <- sym table in FLASH */
+	ldr	r2, _rel_dyn_start_ofs	/* r2 <- rel dyn start ofs */
+	add	r2, r2, r0		/* r2 <- rel dyn start in FLASH */
+	ldr	r3, _rel_dyn_end_ofs	/* r3 <- rel dyn end ofs */
+	add	r3, r3, r0		/* r3 <- rel dyn end in FLASH */
 fixloop:
-	ldr	r4, [r2]
-	sub	r4, r4, r1
-	add	r4, r4, r0
-	str	r4, [r2]
-	add	r2, r2, #4
+	ldr	r0, [r2]	/* r0 <- location to fix up, IN FLASH! */
+	add	r0, r9		/* r0 <- location to fix up in RAM */
+	ldr	r1, [r2, #4]
+	and	r8, r1, #0xff
+	cmp	r8, #23		/* relative fixup? */
+	beq	fixrel
+	cmp	r8, #2		/* absolute fixup? */
+	beq	fixabs
+	/* ignore unknown type of fixup */
+	b	fixnext
+fixabs:
+	/* absolute fix: set location to (offset) symbol value */
+	mov	r1, r1, LSR #4		/* r1 <- symbol index in .dynsym */
+	add	r1, r10, r1		/* r1 <- address of symbol in table */
+	ldr	r1, [r1, #4]		/* r1 <- symbol value */
+	add	r1, r9			/* r1 <- relocated sym addr */
+	b	fixnext
+fixrel:
+	/* relative fix: increase location by offset */
+	ldr	r1, [r0]
+	add	r1, r1, r9
+fixnext:
+	str	r1, [r0]
+	add	r2, r2, #8	/* each rel.dyn entry is 8 bytes */
 	cmp	r2, r3
-	bne	fixloop
+	blo	fixloop
 #endif
 #endif	/* #ifndef CONFIG_SKIP_RELOCATE_UBOOT */
 
 clear_bss:
 #ifndef CONFIG_PRELOADER
-	ldr	r0, _bss_start
-	ldr	r1, _bss_end
+	ldr	r0, _bss_start_ofs
+	ldr	r1, _bss_end_ofs
 	ldr	r3, _TEXT_BASE		/* Text base */
 	mov	r4, r7			/* reloc addr */
-	sub	r0, r0, r3
 	add	r0, r0, r4
-	sub	r1, r1, r3
 	add	r1, r1, r4
 	mov	r2, #0x00000000		/* clear			    */
 
@@ -287,24 +278,33 @@ clbss_l:str	r2, [r0]		/* clear loop...		    */
  * initialization, now running from RAM.
  */
 #ifdef CONFIG_NAND_SPL
-	ldr     pc, _nand_boot
+	ldr     r0, _nand_boot_ofs
+	mov	pc, r0
 
-_nand_boot: .word nand_boot
+_nand_boot_ofs:
+	.word nand_boot
 #else
-	ldr	r0, _TEXT_BASE
-	ldr	r2, _board_init_r
-	sub	r2, r2, r0
-	add	r2, r2, r7	/* position from board_init_r in RAM */
+	ldr	r0, _board_init_r_ofs
+	adr	r1, _start
+	add	r0, r0, r1
+	add	lr, r0, r9
 	/* setup parameters for board_init_r */
 	mov	r0, r5		/* gd_t */
 	mov	r1, r7		/* dest_addr */
 	/* jump to it ... */
-	mov	lr, r2
 	mov	pc, lr
 
-_board_init_r: .word board_init_r
+_board_init_r_ofs:
+	.word board_init_r - _start
 #endif
 
+_rel_dyn_start_ofs:
+	.word __rel_dyn_start - _start
+_rel_dyn_end_ofs:
+	.word __rel_dyn_end - _start
+_dynsym_start_ofs:
+	.word __dynsym_start - _start
+
 #else /* #if !defined(CONFIG_SYS_ARM_WITHOUT_RELOC) */
 /*
  * the actual reset code
@@ -333,10 +333,8 @@ relocate:				/* relocate U-Boot to RAM	    */
 	ldr	r1, _TEXT_BASE		/* test if we run from flash or RAM */
 	cmp     r0, r1                  /* don't reloc during debug         */
 	beq     stack_setup
-	ldr	r2, _armboot_start
-	ldr	r3, _bss_start
-	sub	r2, r3, r2		/* r2 <- size of armboot            */
-	add	r2, r0, r2		/* r2 <- source end address         */
+	ldr	r3, _bss_start_ofs	/* r3 <- _bss_start - _start	    */
+	add	r2, r0, r3		/* r2 <- source end address         */
 
 copy_loop:
 	ldmia	r0!, {r3-r10}		/* copy from source address [r0]    */
@@ -360,8 +358,11 @@ stack_setup:
 	bic	sp, sp, #7		/* 8-byte alignment for ABI compliance */
 
 clear_bss:
-	ldr	r0, _bss_start		/* find start of bss segment        */
-	ldr	r1, _bss_end		/* stop here                        */
+	adr	r2, _start
+	ldr	r0, _bss_start_ofs	/* find start of bss segment        */
+	add	r0, r0, r2
+	ldr	r1, _bss_end_ofs	/* stop here                        */
+	add	r1, r1, r2
 	mov	r2, #0x00000000		/* clear                            */
 
 #ifndef CONFIG_PRELOADER
@@ -374,13 +375,16 @@ clbss_l:str	r2, [r0]		/* clear loop...                    */
 	bl red_LED_on
 #endif /* CONFIG_PRELOADER */
 
-	ldr	pc, _start_armboot
+	ldr	r0, _start_armboot_ofs
+	adr	r1, _start
+	add	r0, r0, r1
+	ldr	pc, r0
 
-_start_armboot:
+_start_armboot_ofs:
 #ifdef CONFIG_NAND_SPL
-	.word nand_boot
+	.word nand_boot - _start
 #else
-	.word start_armboot
+	.word start_armboot - _start
 #endif /* CONFIG_NAND_SPL */
 #endif /* #if !defined(CONFIG_SYS_ARM_WITHOUT_RELOC) */
 
@@ -469,7 +473,7 @@ cpu_init_crit:
 	sub	sp, sp, #S_FRAME_SIZE
 	stmia	sp, {r0 - r12}	@ Save user registers (now in svc mode) r0-r12
 #if defined(CONFIG_SYS_ARM_WITHOUT_RELOC)
-	ldr	r2, _armboot_start
+	adr	r2, _start
 	sub	r2, r2, #(CONFIG_STACKSIZE+CONFIG_SYS_MALLOC_LEN)
 	sub	r2, r2, #(CONFIG_SYS_GBL_DATA_SIZE+8)  @ set base 2 words into abort stack
 #else
@@ -507,7 +511,7 @@ cpu_init_crit:
 
 	.macro get_bad_stack
 #if defined(CONFIG_SYS_ARM_WITHOUT_RELOC)
-	ldr	r13, _armboot_start		@ setup our mode stack
+	adr	r13, _start		@ setup our mode stack
 	sub	r13, r13, #(CONFIG_STACKSIZE+CONFIG_SYS_MALLOC_LEN)
 	sub	r13, r13, #(CONFIG_SYS_GBL_DATA_SIZE+8) @ reserved a couple spots in abort stack
 #else
diff --git a/arch/arm/cpu/arm926ejs/u-boot.lds b/arch/arm/cpu/arm926ejs/u-boot.lds
index 02eb8ca..72f45f8 100644
--- a/arch/arm/cpu/arm926ejs/u-boot.lds
+++ b/arch/arm/cpu/arm926ejs/u-boot.lds
@@ -41,21 +41,19 @@ SECTIONS
 	. = ALIGN(4);
 	.data : {
 		*(.data)
-	__datarel_start = .;
-		*(.data.rel)
-	__datarelrolocal_start = .;
-		*(.data.rel.ro.local)
-	__datarellocal_start = .;
-		*(.data.rel.local)
-	__datarelro_start = .;
-		*(.data.rel.ro)
 	}
 
-	__got_start = .;
 	. = ALIGN(4);
-	.got : { *(.got) }
 
-	__got_end = .;
+	__rel_dyn_start = .;
+	.rel.dyn : { *(.rel.dyn) }
+	__rel_dyn_end = .;
+
+	__dynsym_start = .;
+	.dynsym : { *(.dynsym) }
+
+	. = ALIGN(4);
+
 	. = .;
 	__u_boot_cmd_start = .;
 	.u_boot_cmd : { *(.u_boot_cmd) }
@@ -65,4 +63,10 @@ SECTIONS
 	__bss_start = .;
 	.bss (NOLOAD) : { *(.bss) . = ALIGN(4); }
 	_end = .;
+
+	/DISCARD/ : { *(.dynstr*) }
+	/DISCARD/ : { *(.dynamic*) }
+	/DISCARD/ : { *(.plt*) }
+	/DISCARD/ : { *(.interp*) }
+	/DISCARD/ : { *(.gnu*) }
 }
diff --git a/arch/arm/cpu/armv7/mx51/Makefile b/arch/arm/cpu/armv7/mx5/Makefile
index 7cfaa2c..7cfaa2c 100644
--- a/arch/arm/cpu/armv7/mx51/Makefile
+++ b/arch/arm/cpu/armv7/mx5/Makefile
diff --git a/arch/arm/cpu/armv7/mx51/clock.c b/arch/arm/cpu/armv7/mx5/clock.c
index a27227d..00f649c 100644
--- a/arch/arm/cpu/armv7/mx51/clock.c
+++ b/arch/arm/cpu/armv7/mx5/clock.c
@@ -71,7 +71,7 @@ u32 get_mcu_main_clk(void)
 
 	reg = (__raw_readl(&mxc_ccm->cacrr) & MXC_CCM_CACRR_ARM_PODF_MASK) >>
 		MXC_CCM_CACRR_ARM_PODF_OFFSET;
-	freq = decode_pll(mxc_plls[PLL1_CLOCK], CONFIG_MX51_HCLK_FREQ);
+	freq = decode_pll(mxc_plls[PLL1_CLOCK], CONFIG_SYS_MX5_HCLK);
 	return freq / (reg + 1);
 }
 
@@ -84,14 +84,14 @@ static u32 get_periph_clk(void)
 
 	reg = __raw_readl(&mxc_ccm->cbcdr);
 	if (!(reg & MXC_CCM_CBCDR_PERIPH_CLK_SEL))
-		return decode_pll(mxc_plls[PLL2_CLOCK], CONFIG_MX51_HCLK_FREQ);
+		return decode_pll(mxc_plls[PLL2_CLOCK], CONFIG_SYS_MX5_HCLK);
 	reg = __raw_readl(&mxc_ccm->cbcmr);
 	switch ((reg & MXC_CCM_CBCMR_PERIPH_CLK_SEL_MASK) >>
 		MXC_CCM_CBCMR_PERIPH_CLK_SEL_OFFSET) {
 	case 0:
-		return decode_pll(mxc_plls[PLL1_CLOCK], CONFIG_MX51_HCLK_FREQ);
+		return decode_pll(mxc_plls[PLL1_CLOCK], CONFIG_SYS_MX5_HCLK);
 	case 1:
-		return decode_pll(mxc_plls[PLL3_CLOCK], CONFIG_MX51_HCLK_FREQ);
+		return decode_pll(mxc_plls[PLL3_CLOCK], CONFIG_SYS_MX5_HCLK);
 	default:
 		return 0;
 	}
@@ -146,15 +146,15 @@ static u32 get_uart_clk(void)
 		MXC_CCM_CSCMR1_UART_CLK_SEL_OFFSET) {
 	case 0x0:
 		freq = decode_pll(mxc_plls[PLL1_CLOCK],
-				    CONFIG_MX51_HCLK_FREQ);
+				    CONFIG_SYS_MX5_HCLK);
 		break;
 	case 0x1:
 		freq = decode_pll(mxc_plls[PLL2_CLOCK],
-				    CONFIG_MX51_HCLK_FREQ);
+				    CONFIG_SYS_MX5_HCLK);
 		break;
 	case 0x2:
 		freq = decode_pll(mxc_plls[PLL3_CLOCK],
-				    CONFIG_MX51_HCLK_FREQ);
+				    CONFIG_SYS_MX5_HCLK);
 		break;
 	default:
 		return 66500000;
@@ -181,7 +181,7 @@ u32 get_lp_apm(void)
 	u32 ccsr = __raw_readl(&mxc_ccm->ccsr);
 
 	if (((ccsr >> 9) & 1) == 0)
-		ret_val = CONFIG_MX51_HCLK_FREQ;
+		ret_val = CONFIG_SYS_MX5_HCLK;
 	else
 		ret_val = ((32768 * 1024));
 
@@ -207,17 +207,17 @@ u32 imx_get_cspiclk(void)
 	switch (clk_sel) {
 	case 0:
 		ret_val = decode_pll(mxc_plls[PLL1_CLOCK],
-					CONFIG_MX51_HCLK_FREQ) /
+					CONFIG_SYS_MX5_HCLK) /
 					((pre_pdf + 1) * (pdf + 1));
 		break;
 	case 1:
 		ret_val = decode_pll(mxc_plls[PLL2_CLOCK],
-					CONFIG_MX51_HCLK_FREQ) /
+					CONFIG_SYS_MX5_HCLK) /
 					((pre_pdf + 1) * (pdf + 1));
 		break;
 	case 2:
 		ret_val = decode_pll(mxc_plls[PLL3_CLOCK],
-					CONFIG_MX51_HCLK_FREQ) /
+					CONFIG_SYS_MX5_HCLK) /
 					((pre_pdf + 1) * (pdf + 1));
 		break;
 	default:
@@ -248,7 +248,7 @@ unsigned int mxc_get_clock(enum mxc_clock clk)
 		return imx_get_cspiclk();
 	case MXC_FEC_CLK:
 		return decode_pll(mxc_plls[PLL1_CLOCK],
-				    CONFIG_MX51_HCLK_FREQ);
+				    CONFIG_SYS_MX5_HCLK);
 	default:
 		break;
 	}
@@ -269,16 +269,16 @@ u32 imx_get_fecclk(void)
 /*
  * Dump some core clockes.
  */
-int do_mx51_showclocks(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
+int do_mx5_showclocks(cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 {
 	u32 freq;
 
-	freq = decode_pll(mxc_plls[PLL1_CLOCK], CONFIG_MX51_HCLK_FREQ);
-	printf("mx51 pll1: %dMHz\n", freq / 1000000);
-	freq = decode_pll(mxc_plls[PLL2_CLOCK], CONFIG_MX51_HCLK_FREQ);
-	printf("mx51 pll2: %dMHz\n", freq / 1000000);
-	freq = decode_pll(mxc_plls[PLL3_CLOCK], CONFIG_MX51_HCLK_FREQ);
-	printf("mx51 pll3: %dMHz\n", freq / 1000000);
+	freq = decode_pll(mxc_plls[PLL1_CLOCK], CONFIG_SYS_MX5_HCLK);
+	printf("pll1: %dMHz\n", freq / 1000000);
+	freq = decode_pll(mxc_plls[PLL2_CLOCK], CONFIG_SYS_MX5_HCLK);
+	printf("pll2: %dMHz\n", freq / 1000000);
+	freq = decode_pll(mxc_plls[PLL3_CLOCK], CONFIG_SYS_MX5_HCLK);
+	printf("pll3: %dMHz\n", freq / 1000000);
 	printf("ipg clock     : %dHz\n", mxc_get_clock(MXC_IPG_CLK));
 	printf("ipg per clock : %dHz\n", mxc_get_clock(MXC_IPG_PERCLK));
 
@@ -288,7 +288,7 @@ int do_mx51_showclocks(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[]
 /***************************************************/
 
 U_BOOT_CMD(
-	clockinfo,	CONFIG_SYS_MAXARGS,	1,	do_mx51_showclocks,
-	"display mx51 clocks\n",
+	clockinfo,	CONFIG_SYS_MAXARGS,	1,	do_mx5_showclocks,
+	"display clocks\n",
 	""
 );
diff --git a/arch/arm/cpu/armv7/mx51/iomux.c b/arch/arm/cpu/armv7/mx5/iomux.c
index 62b2954..e8928d5 100644
--- a/arch/arm/cpu/armv7/mx51/iomux.c
+++ b/arch/arm/cpu/armv7/mx5/iomux.c
@@ -23,7 +23,7 @@
 #include <common.h>
 #include <asm/io.h>
 #include <asm/arch/imx-regs.h>
-#include <asm/arch/mx51_pins.h>
+#include <asm/arch/mx5x_pins.h>
 #include <asm/arch/iomux.h>
 #include <asm/arch/sys_proto.h>
 
diff --git a/arch/arm/cpu/armv7/mx51/lowlevel_init.S b/arch/arm/cpu/armv7/mx5/lowlevel_init.S
index 783c81f..e984870 100644
--- a/arch/arm/cpu/armv7/mx51/lowlevel_init.S
+++ b/arch/arm/cpu/armv7/mx5/lowlevel_init.S
@@ -257,20 +257,6 @@ lowlevel_init:
 	orr r1, r1, #(1 << 23)
 	str r1, [r0, #0x4]
 
-#ifdef ENABLE_IMPRECISE_ABORT
-	mrs r1, spsr		/* save old spsr */
-	mrs r0, cpsr		/* read out the cpsr */
-	bic r0, r0, #0x100	/* clear the A bit */
-	msr spsr, r0		/* update spsr */
-	add lr, pc, #0x8	/* update lr */
-	movs pc, lr		/* update cpsr */
-	nop
-	nop
-	nop
-	nop
-	msr spsr, r1		/* restore old spsr */
-#endif
-
 	init_l2cc
 
 	init_aips
diff --git a/arch/arm/cpu/armv7/mx51/soc.c b/arch/arm/cpu/armv7/mx5/soc.c
index f22ebe9..7c7a565 100644
--- a/arch/arm/cpu/armv7/mx51/soc.c
+++ b/arch/arm/cpu/armv7/mx5/soc.c
@@ -33,28 +33,33 @@
 #include <fsl_esdhc.h>
 #endif
 
+#if defined(CONFIG_MX51)
+#define CPU_TYPE 0x51000
+#else
+#error "CPU_TYPE not defined"
+#endif
+
 u32 get_cpu_rev(void)
 {
-	int reg;
-	int system_rev;
+	int system_rev = CPU_TYPE;
+	int reg = __raw_readl(ROM_SI_REV);
 
-	reg = __raw_readl(ROM_SI_REV);
 	switch (reg) {
 	case 0x02:
-		system_rev = 0x51000 | CHIP_REV_1_1;
+		system_rev |= CHIP_REV_1_1;
 		break;
 	case 0x10:
 		if ((__raw_readl(GPIO1_BASE_ADDR + 0x0) & (0x1 << 22)) == 0)
-			system_rev = 0x51000 | CHIP_REV_2_5;
+			system_rev |= CHIP_REV_2_5;
 		else
-			system_rev = 0x51000 | CHIP_REV_2_0;
+			system_rev |= CHIP_REV_2_0;
 		break;
 	case 0x20:
-		system_rev = 0x51000 | CHIP_REV_3_0;
+		system_rev |= CHIP_REV_3_0;
 		break;
 	return system_rev;
 	default:
-		system_rev = 0x51000 | CHIP_REV_1_0;
+		system_rev |= CHIP_REV_1_0;
 		break;
 	}
 	return system_rev;
@@ -67,9 +72,10 @@ int print_cpuinfo(void)
 	u32 cpurev;
 
 	cpurev = get_cpu_rev();
-	printf("CPU:   Freescale i.MX51 family rev%d.%d at %d MHz\n",
-		(cpurev & 0xF0) >> 4,
-		(cpurev & 0x0F) >> 4,
+	printf("CPU:   Freescale i.MX%x family rev%d.%d at %d MHz\n",
+		(cpurev & 0xFF000) >> 12,
+		(cpurev & 0x000F0) >> 4,
+		(cpurev & 0x0000F) >> 0,
 		mxc_get_clock(MXC_ARM_CLK) / 1000000);
 	return 0;
 }
diff --git a/arch/arm/cpu/armv7/mx51/speed.c b/arch/arm/cpu/armv7/mx5/speed.c
index a444def..a444def 100644
--- a/arch/arm/cpu/armv7/mx51/speed.c
+++ b/arch/arm/cpu/armv7/mx5/speed.c
diff --git a/arch/arm/cpu/armv7/mx51/timer.c b/arch/arm/cpu/armv7/mx5/timer.c
index 110edbf..3044fcf 100644
--- a/arch/arm/cpu/armv7/mx51/timer.c
+++ b/arch/arm/cpu/armv7/mx5/timer.c
@@ -75,18 +75,18 @@ void reset_timer(void)
 void reset_timer_masked(void)
 {
 	ulong val = __raw_readl(&cur_gpt->counter);
-	lastinc = val / (CONFIG_MX51_CLK32 / CONFIG_SYS_HZ);
+	lastinc = val / (CONFIG_SYS_MX5_CLK32 / CONFIG_SYS_HZ);
 	timestamp = 0;
 }
 
 ulong get_timer_masked(void)
 {
 	ulong val = __raw_readl(&cur_gpt->counter);
-	val /= (CONFIG_MX51_CLK32 / CONFIG_SYS_HZ);
+	val /= (CONFIG_SYS_MX5_CLK32 / CONFIG_SYS_HZ);
 	if (val >= lastinc)
 		timestamp += (val - lastinc);
 	else
-		timestamp += ((0xFFFFFFFF / (CONFIG_MX51_CLK32 / CONFIG_SYS_HZ))
+		timestamp += ((0xFFFFFFFF / (CONFIG_SYS_MX5_CLK32 / CONFIG_SYS_HZ))
 				- lastinc) + val;
 	lastinc = val;
 	return timestamp;
@@ -106,7 +106,7 @@ void set_timer(ulong t)
 void __udelay(unsigned long usec)
 {
 	unsigned long now, start, tmo;
-	tmo = usec * (CONFIG_MX51_CLK32 / 1000) / 1000;
+	tmo = usec * (CONFIG_SYS_MX5_CLK32 / 1000) / 1000;
 
 	if (!tmo)
 		tmo = 1;
diff --git a/arch/arm/cpu/armv7/mx51/u-boot.lds b/arch/arm/cpu/armv7/mx5/u-boot.lds
index 55d6599..55d6599 100644
--- a/arch/arm/cpu/armv7/mx51/u-boot.lds
+++ b/arch/arm/cpu/armv7/mx5/u-boot.lds
diff --git a/arch/arm/cpu/armv7/start.S b/arch/arm/cpu/armv7/start.S
index 26f335a..64c86e9 100644
--- a/arch/arm/cpu/armv7/start.S
+++ b/arch/arm/cpu/armv7/start.S
@@ -78,13 +78,13 @@ _armboot_start:
 /*
  * These are defined in the board-specific linker script.
  */
-.globl _bss_start
-_bss_start:
-	.word __bss_start
+.globl _bss_start_ofs
+_bss_start_ofs:
+	.word __bss_start - _start
 
-.globl _bss_end
-_bss_end:
-	.word _end
+.globl _bss_end_ofs
+_bss_end_ofs:
+	.word _end - _start
 
 #ifdef CONFIG_USE_IRQ
 /* IRQ stack memory (calculated at run-time) */
@@ -104,29 +104,29 @@ FIQ_STACK_START:
 IRQ_STACK_START_IN:
 	.word	0x0badc0de
 
-.globl _datarel_start
-_datarel_start:
-	.word __datarel_start
+.globl _datarel_start_ofs
+_datarel_start_ofs:
+	.word __datarel_start - _start
 
-.globl _datarelrolocal_start
-_datarelrolocal_start:
-	.word __datarelrolocal_start
+.globl _datarelrolocal_start_ofs
+_datarelrolocal_start_ofs:
+	.word __datarelrolocal_start - _start
 
-.globl _datarellocal_start
-_datarellocal_start:
-	.word __datarellocal_start
+.globl _datarellocal_start_ofs
+_datarellocal_start_ofs:
+	.word __datarellocal_start - _start
 
-.globl _datarelro_start
-_datarelro_start:
-	.word __datarelro_start
+.globl _datarelro_start_ofs
+_datarelro_start_ofs:
+	.word __datarelro_start - _start
 
-.globl _got_start
-_got_start:
-	.word __got_start
+.globl _got_start_ofs
+_got_start_ofs:
+	.word __got_start - _start
 
-.globl _got_end
-_got_end:
-	.word __got_end
+.globl _got_end_Ofs
+_got_end_ofs:
+	.word __got_end - _start
 
 /*
  * the actual reset code
@@ -198,9 +198,8 @@ stack_setup:
 #ifndef CONFIG_SKIP_RELOCATE_UBOOT
 	adr	r0, _start
 	ldr	r2, _TEXT_BASE
-	ldr	r3, _bss_start
-	sub	r2, r3, r2		/* r2 <- size of armboot	    */
-	add	r2, r0, r2		/* r2 <- source end address	    */
+	ldr	r3, _bss_start_ofs
+	add	r2, r0, r3		/* r2 <- source end address	    */
 	cmp	r0, r6
 #ifndef CONFIG_PRELOADER
 	beq	jump_2_ram
@@ -213,33 +212,51 @@ copy_loop:
 	blo	copy_loop
 
 #ifndef CONFIG_PRELOADER
-	/* fix got entries */
-	ldr	r1, _TEXT_BASE
-	mov	r0, r7			/* reloc addr */
-	ldr	r2, _got_start		/* addr in Flash */
-	ldr	r3, _got_end		/* addr in Flash */
-	sub	r3, r3, r1
-	add	r3, r3, r0
-	sub	r2, r2, r1
-	add	r2, r2, r0
-
+	/*
+	 * fix .rel.dyn relocations
+	 */
+	ldr	r0, _TEXT_BASE		/* r0 <- Text base */
+	sub	r9, r7, r0		/* r9 <- relocation offset */
+	ldr	r10, _dynsym_start_ofs	/* r10 <- sym table ofs */
+	add	r10, r10, r0		/* r10 <- sym table in FLASH */
+	ldr	r2, _rel_dyn_start_ofs	/* r2 <- rel dyn start ofs */
+	add	r2, r2, r0		/* r2 <- rel dyn start in FLASH */
+	ldr	r3, _rel_dyn_end_ofs	/* r3 <- rel dyn end ofs */
+	add	r3, r3, r0		/* r3 <- rel dyn end in FLASH */
 fixloop:
-	ldr	r4, [r2]
-	sub	r4, r4, r1
-	add	r4, r4, r0
-	str	r4, [r2]
-	add	r2, r2, #4
+	ldr	r0, [r2]	/* r0 <- location to fix up, IN FLASH! */
+	add	r0, r9		/* r0 <- location to fix up in RAM */
+	ldr	r1, [r2, #4]
+	and	r8, r1, #0xff
+	cmp	r8, #23		/* relative fixup? */
+	beq	fixrel
+	cmp	r8, #2		/* absolute fixup? */
+	beq	fixabs
+	/* ignore unknown type of fixup */
+	b	fixnext
+fixabs:
+	/* absolute fix: set location to (offset) symbol value */
+	mov	r1, r1, LSR #4		/* r1 <- symbol index in .dynsym */
+	add	r1, r10, r1		/* r1 <- address of symbol in table */
+	ldr	r1, [r1, #4]		/* r1 <- symbol value */
+	add	r1, r9			/* r1 <- relocated sym addr */
+	b	fixnext
+fixrel:
+	/* relative fix: increase location by offset */
+	ldr	r1, [r0]
+	add	r1, r1, r9
+fixnext:
+	str	r1, [r0]
+	add	r2, r2, #8	/* each rel.dyn entry is 8 bytes */
 	cmp	r2, r3
-	bne	fixloop
+	blo	fixloop
 
 clear_bss:
-	ldr	r0, _bss_start
-	ldr	r1, _bss_end
+	ldr	r0, _bss_start_ofs
+	ldr	r1, _bss_end_ofs
 	ldr	r3, _TEXT_BASE		/* Text base */
 	mov	r4, r7			/* reloc addr */
-	sub	r0, r0, r3
 	add	r0, r0, r4
-	sub	r1, r1, r3
 	add	r1, r1, r4
 	mov	r2, #0x00000000		/* clear			    */
 
@@ -255,18 +272,26 @@ clbss_l:str	r2, [r0]		/* clear loop...		    */
  * initialization, now running from RAM.
  */
 jump_2_ram:
-	ldr	r0, _TEXT_BASE
-	ldr	r2, _board_init_r
-	sub	r2, r2, r0
-	add	r2, r2, r7	/* position from board_init_r in RAM */
+	ldr	r0, _board_init_r_ofs
+	adr	r1, _start
+	add	r0, r0, r1
+	add	lr, r0, r9
 	/* setup parameters for board_init_r */
 	mov	r0, r5		/* gd_t */
 	mov	r1, r7		/* dest_addr */
 	/* jump to it ... */
-	mov	lr, r2
 	mov	pc, lr
 
-_board_init_r: .word board_init_r
+_board_init_r_ofs:
+	.word board_init_r - _start
+
+_rel_dyn_start_ofs:
+	.word __rel_dyn_start - _start
+_rel_dyn_end_ofs:
+	.word __rel_dyn_end - _start
+_dynsym_start_ofs:
+	.word __dynsym_start - _start
+
 #else /* #if !defined(CONFIG_SYS_ARM_WITHOUT_RELOC) */
 /*
  * the actual reset code
diff --git a/arch/arm/cpu/armv7/u-boot.lds b/arch/arm/cpu/armv7/u-boot.lds
index d4fd3fc..88a0fec 100644
--- a/arch/arm/cpu/armv7/u-boot.lds
+++ b/arch/arm/cpu/armv7/u-boot.lds
@@ -53,6 +53,13 @@ SECTIONS
 	__datarelro_start = .;
 		*(.data.rel.ro)
 	}
+	. = ALIGN(4);
+	__rel_dyn_start = .;
+	.rel.dyn : { *(.rel.dyn) }
+	__rel_dyn_end = .;
+
+	__dynsym_start = .;
+	.dynsym : { *(.dynsym) }
 
 	__got_start = .;
 	. = ALIGN(4);
diff --git a/arch/arm/cpu/pxa/cpu.c b/arch/arm/cpu/pxa/cpu.c
index 800d120..3ea3458 100644
--- a/arch/arm/cpu/pxa/cpu.c
+++ b/arch/arm/cpu/pxa/cpu.c
@@ -30,10 +30,11 @@
  * CPU specific code
  */
 
-#include <common.h>
+#include <asm/io.h>
+#include <asm/system.h>
 #include <command.h>
+#include <common.h>
 #include <asm/arch/pxa-regs.h>
-#include <asm/system.h>
 
 static void cache_flush(void);
 
@@ -71,17 +72,249 @@ void set_GPIO_mode(int gpio_mode)
 {
 	int gpio = gpio_mode & GPIO_MD_MASK_NR;
 	int fn = (gpio_mode & GPIO_MD_MASK_FN) >> 8;
-	int gafr;
+	int val;
+
+	/* This below changes direction setting of GPIO "gpio" */
+	val = readl(GPDR(gpio));
 
 	if (gpio_mode & GPIO_MD_MASK_DIR)
-	{
-		GPDR(gpio) |= GPIO_bit(gpio);
-	}
+		val |= GPIO_bit(gpio);
 	else
-	{
-		GPDR(gpio) &= ~GPIO_bit(gpio);
-	}
-	gafr = GAFR(gpio) & ~(0x3 << (((gpio) & 0xf)*2));
-	GAFR(gpio) = gafr |  (fn  << (((gpio) & 0xf)*2));
+		val &= ~GPIO_bit(gpio);
+
+	writel(val, GPDR(gpio));
+
+	/* This below updates only AF of GPIO "gpio" */
+	val = readl(GAFR(gpio));
+	val &= ~(0x3 << (((gpio) & 0xf) * 2));
+	val |= fn << (((gpio) & 0xf) * 2);
+	writel(val, GAFR(gpio));
 }
 #endif /* CONFIG_CPU_MONAHANS */
+
+void pxa_wait_ticks(int ticks)
+{
+	writel(0, OSCR);
+	while (readl(OSCR) < ticks)
+		asm volatile("":::"memory");
+}
+
+inline void writelrb(uint32_t val, uint32_t addr)
+{
+	writel(val, addr);
+	asm volatile("":::"memory");
+	readl(addr);
+	asm volatile("":::"memory");
+}
+
+void pxa_dram_init(void)
+{
+	uint32_t tmp;
+	int i;
+	/*
+	 * 1) Initialize Asynchronous static memory controller
+	 */
+
+	writelrb(CONFIG_SYS_MSC0_VAL, MSC0);
+	writelrb(CONFIG_SYS_MSC1_VAL, MSC1);
+	writelrb(CONFIG_SYS_MSC2_VAL, MSC2);
+	/*
+	 * 2) Initialize Card Interface
+	 */
+
+	/* MECR: Memory Expansion Card Register */
+	writelrb(CONFIG_SYS_MECR_VAL, MECR);
+	/* MCMEM0: Card Interface slot 0 timing */
+	writelrb(CONFIG_SYS_MCMEM0_VAL, MCMEM0);
+	/* MCMEM1: Card Interface slot 1 timing */
+	writelrb(CONFIG_SYS_MCMEM1_VAL, MCMEM1);
+	/* MCATT0: Card Interface Attribute Space Timing, slot 0 */
+	writelrb(CONFIG_SYS_MCATT0_VAL, MCATT0);
+	/* MCATT1: Card Interface Attribute Space Timing, slot 1 */
+	writelrb(CONFIG_SYS_MCATT1_VAL, MCATT1);
+	/* MCIO0: Card Interface I/O Space Timing, slot 0 */
+	writelrb(CONFIG_SYS_MCIO0_VAL, MCIO0);
+	/* MCIO1: Card Interface I/O Space Timing, slot 1 */
+	writelrb(CONFIG_SYS_MCIO1_VAL, MCIO1);
+
+	/*
+	 * 3) Configure Fly-By DMA register
+	 */
+
+	writelrb(CONFIG_SYS_FLYCNFG_VAL, FLYCNFG);
+
+	/*
+	 * 4) Initialize Timing for Sync Memory (SDCLK0)
+	 */
+
+	/*
+	 * Before accessing MDREFR we need a valid DRI field, so we set
+	 * this to power on defaults + DRI field.
+	 */
+
+	/* Read current MDREFR config and zero out DRI */
+	tmp = readl(MDREFR) & ~0xfff;
+	/* Add user-specified DRI */
+	tmp |= CONFIG_SYS_MDREFR_VAL & 0xfff;
+	/* Configure important bits */
+	tmp |= MDREFR_K0RUN | MDREFR_SLFRSH;
+	tmp &= ~(MDREFR_APD | MDREFR_E1PIN);
+
+	/* Write MDREFR back */
+	writelrb(tmp, MDREFR);
+
+	/*
+	 * 5) Initialize Synchronous Static Memory (Flash/Peripherals)
+	 */
+
+	/* Initialize SXCNFG register. Assert the enable bits.
+	 *
+	 * Write SXMRS to cause an MRS command to all enabled banks of
+	 * synchronous static memory. Note that SXLCR need not be written
+	 * at this time.
+	 */
+	writelrb(CONFIG_SYS_SXCNFG_VAL, SXCNFG);
+
+	/*
+	 * 6) Initialize SDRAM
+	 */
+
+	writelrb(CONFIG_SYS_MDREFR_VAL & ~MDREFR_SLFRSH, MDREFR);
+	writelrb(CONFIG_SYS_MDREFR_VAL | MDREFR_E1PIN, MDREFR);
+
+	/*
+	 * 7) Write MDCNFG with MDCNFG:DEx deasserted (set to 0), to configure
+	 *    but not enable each SDRAM partition pair.
+	 */
+
+	writelrb(CONFIG_SYS_MDCNFG_VAL &
+		~(MDCNFG_DE0 | MDCNFG_DE1 | MDCNFG_DE2 | MDCNFG_DE3), MDCNFG);
+	/* Wait for the clock to the SDRAMs to stabilize, 100..200 usec. */
+	pxa_wait_ticks(0x300);
+
+	/*
+	 * 8) Trigger a number (usually 8) refresh cycles by attempting
+	 *    non-burst read or write accesses to disabled SDRAM, as commonly
+	 *    specified in the power up sequence documented in SDRAM data
+	 *    sheets. The address(es) used for this purpose must not be
+	 *    cacheable.
+	 */
+	for (i = 9; i >= 0; i--) {
+		writel(i, 0xa0000000);
+		asm volatile("":::"memory");
+	}
+	/*
+	 * 9) Write MDCNFG with enable bits asserted (MDCNFG:DEx set to 1).
+	 */
+
+	tmp = CONFIG_SYS_MDCNFG_VAL &
+		(MDCNFG_DE0 | MDCNFG_DE1 | MDCNFG_DE2 | MDCNFG_DE3);
+	tmp |= readl(MDCNFG);
+	writelrb(tmp, MDCNFG);
+
+	/*
+	 * 10) Write MDMRS.
+	 */
+
+	writelrb(CONFIG_SYS_MDMRS_VAL, MDMRS);
+
+	/*
+	 * 11) Enable APD
+	 */
+
+	if (CONFIG_SYS_MDREFR_VAL & MDREFR_APD) {
+		tmp = readl(MDREFR);
+		tmp |= MDREFR_APD;
+		writelrb(tmp, MDREFR);
+	}
+}
+
+void pxa_gpio_setup(void)
+{
+	writel(CONFIG_SYS_GPSR0_VAL, GPSR0);
+	writel(CONFIG_SYS_GPSR1_VAL, GPSR1);
+	writel(CONFIG_SYS_GPSR2_VAL, GPSR2);
+#if defined(CONFIG_PXA27X) || defined(CONFIG_CPU_MONAHANS)
+	writel(CONFIG_SYS_GPSR3_VAL, GPSR3);
+#endif
+
+	writel(CONFIG_SYS_GPCR0_VAL, GPCR0);
+	writel(CONFIG_SYS_GPCR1_VAL, GPCR1);
+	writel(CONFIG_SYS_GPCR2_VAL, GPCR2);
+#if defined(CONFIG_PXA27X) || defined(CONFIG_CPU_MONAHANS)
+	writel(CONFIG_SYS_GPCR3_VAL, GPCR3);
+#endif
+
+	writel(CONFIG_SYS_GPDR0_VAL, GPDR0);
+	writel(CONFIG_SYS_GPDR1_VAL, GPDR1);
+	writel(CONFIG_SYS_GPDR2_VAL, GPDR2);
+#if defined(CONFIG_PXA27X) || defined(CONFIG_CPU_MONAHANS)
+	writel(CONFIG_SYS_GPDR3_VAL, GPDR3);
+#endif
+
+	writel(CONFIG_SYS_GAFR0_L_VAL, GAFR0_L);
+	writel(CONFIG_SYS_GAFR0_U_VAL, GAFR0_U);
+	writel(CONFIG_SYS_GAFR1_L_VAL, GAFR1_L);
+	writel(CONFIG_SYS_GAFR1_U_VAL, GAFR1_U);
+	writel(CONFIG_SYS_GAFR2_L_VAL, GAFR2_L);
+	writel(CONFIG_SYS_GAFR2_U_VAL, GAFR2_U);
+#if defined(CONFIG_PXA27X) || defined(CONFIG_CPU_MONAHANS)
+	writel(CONFIG_SYS_GAFR3_L_VAL, GAFR3_L);
+	writel(CONFIG_SYS_GAFR3_U_VAL, GAFR3_U);
+#endif
+
+	writel(CONFIG_SYS_PSSR_VAL, PSSR);
+}
+
+void pxa_interrupt_setup(void)
+{
+	writel(0, ICLR);
+	writel(0, ICMR);
+#if defined(CONFIG_PXA27X) || defined(CONFIG_CPU_MONAHANS)
+	writel(0, ICLR2);
+	writel(0, ICMR2);
+#endif
+}
+
+void pxa_clock_setup(void)
+{
+#ifndef CONFIG_CPU_MONAHANS
+	writel(CONFIG_SYS_CKEN, CKEN);
+	writel(CONFIG_SYS_CCCR, CCCR);
+	asm volatile("mcr	p14, 0, %0, c6, c0, 0"::"r"(2));
+#else
+/* Set CKENA/CKENB/ACCR for MH */
+#endif
+
+	/* enable the 32Khz oscillator for RTC and PowerManager */
+	writel(OSCC_OON, OSCC);
+	while(!(readl(OSCC) & OSCC_OOK))
+		asm volatile("":::"memory");
+}
+
+void pxa_wakeup(void)
+{
+	uint32_t rcsr;
+
+	rcsr = readl(RCSR);
+	writel(rcsr & (RCSR_GPR | RCSR_SMR | RCSR_WDR | RCSR_HWR), RCSR);
+
+	/* Wakeup */
+	if (rcsr & RCSR_SMR) {
+		writel(PSSR_PH, PSSR);
+		pxa_dram_init();
+		icache_disable();
+		dcache_disable();
+		asm volatile("mov	pc, %0"::"r"(readl(PSSR)));
+	}
+}
+
+int arch_cpu_init(void)
+{
+	pxa_gpio_setup();
+//	pxa_wait_ticks(0x8000);
+	pxa_wakeup();
+	pxa_interrupt_setup();
+	pxa_clock_setup();
+	return 0;
+}
diff --git a/arch/arm/cpu/pxa/i2c.c b/arch/arm/cpu/pxa/i2c.c
index 6b72ba1..7aa49ae 100644
--- a/arch/arm/cpu/pxa/i2c.c
+++ b/arch/arm/cpu/pxa/i2c.c
@@ -33,6 +33,7 @@
 /* FIXME: this file is PXA255 specific! What about other XScales? */
 
 #include <common.h>
+#include <asm/io.h>
 
 #ifdef CONFIG_HARD_I2C
 
@@ -93,19 +94,21 @@ struct i2c_msg {
 
 static void i2c_reset( void )
 {
-	ICR &= ~ICR_IUE;		/* disable unit */
-	ICR |= ICR_UR;			/* reset the unit */
+	writel(readl(ICR) & ~ICR_IUE, ICR);	/* disable unit */
+	writel(readl(ICR) | ICR_UR, ICR);	/* reset the unit */
 	udelay(100);
-	ICR &= ~ICR_IUE;		/* disable unit */
+	writel(readl(ICR) & ~ICR_IUE, ICR);	/* disable unit */
 #ifdef CONFIG_CPU_MONAHANS
-	CKENB |= (CKENB_4_I2C); /*  | CKENB_1_PWM1 | CKENB_0_PWM0); */
+	/* | CKENB_1_PWM1 | CKENB_0_PWM0); */
+	writel(readl(CKENB) | (CKENB_4_I2C), CKENB);
 #else /* CONFIG_CPU_MONAHANS */
-	CKEN |= CKEN14_I2C;		/* set the global I2C clock on */
+	/* set the global I2C clock on */
+	writel(readl(CKEN) | CKEN14_I2C, CKEN);
 #endif
-	ISAR = I2C_PXA_SLAVE_ADDR;	/* set our slave address */
-	ICR = I2C_ICR_INIT;		/* set control register values */
-	ISR = I2C_ISR_INIT;		/* set clear interrupt bits */
-	ICR |= ICR_IUE;			/* enable unit */
+	writel(I2C_PXA_SLAVE_ADDR, ISAR);	/* set our slave address */
+	writel(I2C_ICR_INIT, ICR);		/* set control reg values */
+	writel(I2C_ISR_INIT, ISR);		/* set clear interrupt bits */
+	writel(readl(ICR) | ICR_IUE, ICR);	/* enable unit */
 	udelay(100);
 }
 
@@ -159,22 +162,26 @@ int i2c_transfer(struct i2c_msg *msg)
 			goto transfer_error_bus_busy;
 
 		/* start transmission */
-		ICR &= ~ICR_START;
-		ICR &= ~ICR_STOP;
-		IDBR = msg->data;
-		if (msg->condition == I2C_COND_START)     ICR |=  ICR_START;
-		if (msg->condition == I2C_COND_STOP)      ICR |=  ICR_STOP;
-		if (msg->acknack   == I2C_ACKNAK_SENDNAK) ICR |=  ICR_ACKNAK;
-		if (msg->acknack   == I2C_ACKNAK_SENDACK) ICR &= ~ICR_ACKNAK;
-		ICR &= ~ICR_ALDIE;
-		ICR |= ICR_TB;
+		writel(readl(ICR) & ~ICR_START, ICR);
+		writel(readl(ICR) & ~ICR_STOP, ICR);
+		writel(msg->data, IDBR);
+		if (msg->condition == I2C_COND_START)
+			writel(readl(ICR) | ICR_START, ICR);
+		if (msg->condition == I2C_COND_STOP)
+			writel(readl(ICR) | ICR_STOP, ICR);
+		if (msg->acknack == I2C_ACKNAK_SENDNAK)
+			writel(readl(ICR) | ICR_ACKNAK, ICR);
+		if (msg->acknack == I2C_ACKNAK_SENDACK)
+			writel(readl(ICR) & ~ICR_ACKNAK, ICR);
+		writel(readl(ICR) & ~ICR_ALDIE, ICR);
+		writel(readl(ICR) | ICR_TB, ICR);
 
 		/* transmit register empty? */
 		if (!i2c_isr_set_cleared(ISR_ITE,0))
 			goto transfer_error_transmit_timeout;
 
 		/* clear 'transmit empty' state */
-		ISR |= ISR_ITE;
+		writel(readl(ISR) | ISR_ITE, ISR);
 
 		/* wait for ACK from slave */
 		if (msg->acknack == I2C_ACKNAK_WAITACK)
@@ -189,23 +196,27 @@ int i2c_transfer(struct i2c_msg *msg)
 			goto transfer_error_bus_busy;
 
 		/* start receive */
-		ICR &= ~ICR_START;
-		ICR &= ~ICR_STOP;
-		if (msg->condition == I2C_COND_START)	  ICR |= ICR_START;
-		if (msg->condition == I2C_COND_STOP)	  ICR |= ICR_STOP;
-		if (msg->acknack   == I2C_ACKNAK_SENDNAK) ICR |=  ICR_ACKNAK;
-		if (msg->acknack   == I2C_ACKNAK_SENDACK) ICR &= ~ICR_ACKNAK;
-		ICR &= ~ICR_ALDIE;
-		ICR |= ICR_TB;
+		writel(readl(ICR) & ~ICR_START, ICR);
+		writel(readl(ICR) & ~ICR_STOP, ICR);
+		if (msg->condition == I2C_COND_START)
+			writel(readl(ICR) | ICR_START, ICR);
+		if (msg->condition == I2C_COND_STOP)
+			writel(readl(ICR) | ICR_STOP, ICR);
+		if (msg->acknack == I2C_ACKNAK_SENDNAK)
+			writel(readl(ICR) | ICR_ACKNAK, ICR);
+		if (msg->acknack == I2C_ACKNAK_SENDACK)
+			writel(readl(ICR) & ~ICR_ACKNAK, ICR);
+		writel(readl(ICR) & ~ICR_ALDIE, ICR);
+		writel(readl(ICR) | ICR_TB, ICR);
 
 		/* receive register full? */
 		if (!i2c_isr_set_cleared(ISR_IRF,0))
 			goto transfer_error_receive_timeout;
 
-		msg->data = IDBR;
+		msg->data = readl(IDBR);
 
 		/* clear 'receive empty' state */
-		ISR |= ISR_IRF;
+		writel(readl(ISR) | ISR_IRF, ISR);
 
 		break;
 
diff --git a/arch/arm/cpu/pxa/pxafb.c b/arch/arm/cpu/pxa/pxafb.c
index 0ee6a75..987fa06 100644
--- a/arch/arm/cpu/pxa/pxafb.c
+++ b/arch/arm/cpu/pxa/pxafb.c
@@ -35,6 +35,7 @@
 #include <stdio_dev.h>
 #include <lcd.h>
 #include <asm/arch/pxa-regs.h>
+#include <asm/io.h>
 
 /* #define DEBUG */
 
@@ -56,26 +57,26 @@
 
 /* 640x480x16 @ 61 Hz */
 vidinfo_t panel_info = {
-	vl_col:		640,
-	vl_row:		480,
-	vl_width:	640,
-	vl_height:	480,
-	vl_clkp:	CONFIG_SYS_HIGH,
-	vl_oep:		CONFIG_SYS_HIGH,
-	vl_hsp:		CONFIG_SYS_HIGH,
-	vl_vsp:		CONFIG_SYS_HIGH,
-	vl_dp:		CONFIG_SYS_HIGH,
-	vl_bpix:	LCD_BPP,
-	vl_lbw:		0,
-	vl_splt:	0,
-	vl_clor:	0,
-	vl_tft:		1,
-	vl_hpw:		40,
-	vl_blw:		56,
-	vl_elw:		56,
-	vl_vpw:		20,
-	vl_bfw:		8,
-	vl_efw:		8,
+	.vl_col		= 640,
+	.vl_row		= 480,
+	.vl_width	= 640,
+	.vl_height	= 480,
+	.vl_clkp	= CONFIG_SYS_HIGH,
+	.vl_oep		= CONFIG_SYS_HIGH,
+	.vl_hsp		= CONFIG_SYS_HIGH,
+	.vl_vsp		= CONFIG_SYS_HIGH,
+	.vl_dp		= CONFIG_SYS_HIGH,
+	.vl_bpix	= LCD_BPP,
+	.vl_lbw		= 0,
+	.vl_splt	= 0,
+	.vl_clor	= 0,
+	.vl_tft		= 1,
+	.vl_hpw		= 40,
+	.vl_blw		= 56,
+	.vl_elw		= 56,
+	.vl_vpw		= 20,
+	.vl_bfw		= 8,
+	.vl_efw		= 8,
 };
 #endif /* CONFIG_PXA_VIDEO */
 
@@ -90,26 +91,26 @@ vidinfo_t panel_info = {
 # define REG_LCCR3	0x0340FF08
 
 vidinfo_t panel_info = {
-	vl_col:		640,
-	vl_row:		480,
-	vl_width:	157,
-	vl_height:	118,
-	vl_clkp:	CONFIG_SYS_HIGH,
-	vl_oep:		CONFIG_SYS_HIGH,
-	vl_hsp:		CONFIG_SYS_HIGH,
-	vl_vsp:		CONFIG_SYS_HIGH,
-	vl_dp:		CONFIG_SYS_HIGH,
-	vl_bpix:	LCD_BPP,
-	vl_lbw:		0,
-	vl_splt:	1,
-	vl_clor:	1,
-	vl_tft:		0,
-	vl_hpw:		1,
-	vl_blw:		3,
-	vl_elw:		3,
-	vl_vpw:		1,
-	vl_bfw:		0,
-	vl_efw:		0,
+	.vl_col		= 640,
+	.vl_row		= 480,
+	.vl_width	= 157,
+	.vl_height	= 118,
+	.vl_clkp	= CONFIG_SYS_HIGH,
+	.vl_oep		= CONFIG_SYS_HIGH,
+	.vl_hsp		= CONFIG_SYS_HIGH,
+	.vl_vsp		= CONFIG_SYS_HIGH,
+	.vl_dp		= CONFIG_SYS_HIGH,
+	.vl_bpix	= LCD_BPP,
+	.vl_lbw		= 0,
+	.vl_splt	= 1,
+	.vl_clor	= 1,
+	.vl_tft		= 0,
+	.vl_hpw		= 1,
+	.vl_blw		= 3,
+	.vl_elw		= 3,
+	.vl_vpw		= 1,
+	.vl_bfw		= 0,
+	.vl_efw		= 0,
 };
 #endif /* CONFIG_SHARP_LM8V31 */
 /*----------------------------------------------------------------------*/
@@ -123,26 +124,26 @@ vidinfo_t panel_info = {
 # define REG_LCCR3	0x0340FF08
 
 vidinfo_t panel_info = {
-	vl_col:		640,
-	vl_row:		480,
-	vl_width:	157,
-	vl_height:	118,
-	vl_clkp:	CONFIG_SYS_HIGH,
-	vl_oep:		CONFIG_SYS_HIGH,
-	vl_hsp:		CONFIG_SYS_HIGH,
-	vl_vsp:		CONFIG_SYS_HIGH,
-	vl_dp:		CONFIG_SYS_HIGH,
-	vl_bpix:	LCD_BPP,
-	vl_lbw:		0,
-	vl_splt:	1,
-	vl_clor:	1,
-	vl_tft:		1,
-	vl_hpw:		32,
-	vl_blw:		144,
-	vl_elw:		32,
-	vl_vpw:		2,
-	vl_bfw:		13,
-	vl_efw:		30,
+	.vl_col		= 640,
+	.vl_row		= 480,
+	.vl_width	= 157,
+	.vl_height	= 118,
+	.vl_clkp	= CONFIG_SYS_HIGH,
+	.vl_oep		= CONFIG_SYS_HIGH,
+	.vl_hsp		= CONFIG_SYS_HIGH,
+	.vl_vsp		= CONFIG_SYS_HIGH,
+	.vl_dp		= CONFIG_SYS_HIGH,
+	.vl_bpix	= LCD_BPP,
+	.vl_lbw		= 0,
+	.vl_splt	= 1,
+	.vl_clor	= 1,
+	.vl_tft		= 1,
+	.vl_hpw		= 32,
+	.vl_blw		= 144,
+	.vl_elw		= 32,
+	.vl_vpw		= 2,
+	.vl_bfw		= 13,
+	.vl_efw		= 30,
 };
 #endif /* CONFIG_VOIPAC_LCD */
 
@@ -156,26 +157,26 @@ vidinfo_t panel_info = {
 #define REG_LCCR3	0x0340FF20
 
 vidinfo_t panel_info = {
-	vl_col:		320,
-	vl_row:		240,
-	vl_width:	167,
-	vl_height:	109,
-	vl_clkp:	CONFIG_SYS_HIGH,
-	vl_oep:		CONFIG_SYS_HIGH,
-	vl_hsp:		CONFIG_SYS_HIGH,
-	vl_vsp:		CONFIG_SYS_HIGH,
-	vl_dp:		CONFIG_SYS_HIGH,
-	vl_bpix:	LCD_BPP,
-	vl_lbw:		1,
-	vl_splt:	0,
-	vl_clor:	1,
-	vl_tft:		0,
-	vl_hpw:		1,
-	vl_blw:		1,
-	vl_elw:		1,
-	vl_vpw:		7,
-	vl_bfw:		0,
-	vl_efw:		0,
+	.vl_col		= 320,
+	.vl_row		= 240,
+	.vl_width	= 167,
+	.vl_height	= 109,
+	.vl_clkp	= CONFIG_SYS_HIGH,
+	.vl_oep		= CONFIG_SYS_HIGH,
+	.vl_hsp		= CONFIG_SYS_HIGH,
+	.vl_vsp		= CONFIG_SYS_HIGH,
+	.vl_dp		= CONFIG_SYS_HIGH,
+	.vl_bpix	= LCD_BPP,
+	.vl_lbw		= 1,
+	.vl_splt	= 0,
+	.vl_clor	= 1,
+	.vl_tft		= 0,
+	.vl_hpw		= 1,
+	.vl_blw		= 1,
+	.vl_elw		= 1,
+	.vl_vpw		= 7,
+	.vl_bfw		= 0,
+	.vl_efw		= 0,
 };
 #endif /* CONFIG_HITACHI_SX14 */
 
@@ -190,31 +191,132 @@ vidinfo_t panel_info = {
 # define REG_LCCR3	0x03b00009
 
 vidinfo_t panel_info = {
-	vl_col:		240,
-	vl_row:		320,
-	vl_width:	240,
-	vl_height:	320,
-	vl_clkp:	CONFIG_SYS_HIGH,
-	vl_oep:		CONFIG_SYS_LOW,
-	vl_hsp:		CONFIG_SYS_LOW,
-	vl_vsp:		CONFIG_SYS_LOW,
-	vl_dp:		CONFIG_SYS_HIGH,
-	vl_bpix:	LCD_BPP,
-	vl_lbw:		0,
-	vl_splt:	1,
-	vl_clor:	1,
-	vl_tft:		1,
-	vl_hpw:		4,
-	vl_blw:		4,
-	vl_elw:		8,
-	vl_vpw:		4,
-	vl_bfw:		4,
-	vl_efw:		8,
+	.vl_col		= 240,
+	.vl_row		= 320,
+	.vl_width	= 240,
+	.vl_height	= 320,
+	.vl_clkp	= CONFIG_SYS_HIGH,
+	.vl_oep		= CONFIG_SYS_LOW,
+	.vl_hsp		= CONFIG_SYS_LOW,
+	.vl_vsp		= CONFIG_SYS_LOW,
+	.vl_dp		= CONFIG_SYS_HIGH,
+	.vl_bpix	= LCD_BPP,
+	.vl_lbw		= 0,
+	.vl_splt	= 1,
+	.vl_clor	= 1,
+	.vl_tft		= 1,
+	.vl_hpw		= 4,
+	.vl_blw		= 4,
+	.vl_elw		= 8,
+	.vl_vpw		= 4,
+	.vl_bfw		= 4,
+	.vl_efw		= 8,
 };
 #endif /* CONFIG_LMS283GF05 */
 
 /*----------------------------------------------------------------------*/
 
+#ifdef CONFIG_ACX517AKN
+
+# define LCD_BPP	LCD_COLOR8
+
+/* you have to set lccr0 and lccr3 (including pcd) */
+# define REG_LCCR0	0x003008f9
+# define REG_LCCR3	0x03700006
+
+vidinfo_t panel_info = {
+	.vl_col		= 320,
+	.vl_row		= 320,
+	.vl_width	= 320,
+	.vl_height	= 320,
+	.vl_clkp	= CONFIG_SYS_HIGH,
+	.vl_oep		= CONFIG_SYS_LOW,
+	.vl_hsp		= CONFIG_SYS_LOW,
+	.vl_vsp		= CONFIG_SYS_LOW,
+	.vl_dp		= CONFIG_SYS_HIGH,
+	.vl_bpix	= LCD_BPP,
+	.vl_lbw		= 0,
+	.vl_splt	= 1,
+	.vl_clor	= 1,
+	.vl_tft		= 1,
+	.vl_hpw		= 0x04,
+	.vl_blw		= 0x1c,
+	.vl_elw		= 0x08,
+	.vl_vpw		= 0x01,
+	.vl_bfw		= 0x07,
+	.vl_efw		= 0x08,
+};
+#endif /* CONFIG_ACX517AKN */
+
+/*----------------------------------------------------------------------*/
+
+#ifdef CONFIG_LQ038J7DH53
+
+# define LCD_BPP	LCD_COLOR8
+
+/* you have to set lccr0 and lccr3 (including pcd) */
+# define REG_LCCR0	0x003008f9
+# define REG_LCCR3	0x03700004
+
+vidinfo_t panel_info = {
+	.vl_col		= 320,
+	.vl_row		= 480,
+	.vl_width	= 320,
+	.vl_height	= 480,
+	.vl_clkp	= CONFIG_SYS_HIGH,
+	.vl_oep		= CONFIG_SYS_LOW,
+	.vl_hsp		= CONFIG_SYS_LOW,
+	.vl_vsp		= CONFIG_SYS_LOW,
+	.vl_dp		= CONFIG_SYS_HIGH,
+	.vl_bpix	= LCD_BPP,
+	.vl_lbw		= 0,
+	.vl_splt	= 1,
+	.vl_clor	= 1,
+	.vl_tft		= 1,
+	.vl_hpw		= 0x04,
+	.vl_blw		= 0x20,
+	.vl_elw		= 0x01,
+	.vl_vpw		= 0x01,
+	.vl_bfw		= 0x04,
+	.vl_efw		= 0x01,
+};
+#endif /* CONFIG_ACX517AKN */
+
+/*----------------------------------------------------------------------*/
+
+#ifdef CONFIG_LITTLETON_LCD
+# define LCD_BPP	LCD_COLOR8
+
+/* you have to set lccr0 and lccr3 (including pcd) */
+# define REG_LCCR0	0x003008f8
+# define REG_LCCR3	0x0300FF04
+
+vidinfo_t panel_info = {
+	.vl_col		= 480,
+	.vl_row		= 640,
+	.vl_width	= 480,
+	.vl_height	= 640,
+	.vl_clkp	= CONFIG_SYS_HIGH,
+	.vl_oep		= CONFIG_SYS_HIGH,
+	.vl_hsp		= CONFIG_SYS_HIGH,
+	.vl_vsp		= CONFIG_SYS_HIGH,
+	.vl_dp		= CONFIG_SYS_HIGH,
+	.vl_bpix	= LCD_BPP,
+	.vl_lbw		= 0,
+	.vl_splt	= 0,
+	.vl_clor	= 0,
+	.vl_tft		= 1,
+	.vl_hpw		= 9,
+	.vl_blw		= 8,
+	.vl_elw		= 24,
+	.vl_vpw		= 2,
+	.vl_bfw		= 2,
+	.vl_efw		= 4,
+};
+#endif /* CONFIG_LITTLETON_LCD */
+
+/*----------------------------------------------------------------------*/
+
 #if LCD_BPP == LCD_COLOR8
 void lcd_setcolreg (ushort regno, ushort red, ushort green, ushort blue);
 #endif
@@ -377,12 +479,14 @@ static void pxafb_setup_gpio (vidinfo_t *vid)
 	{
 		debug("Setting GPIO for 4 bit data\n");
 		/* bits 58-61 */
-		GPDR1 |= (0xf << 26);
-		GAFR1_U = (GAFR1_U & ~(0xff << 20)) | (0xaa << 20);
+		writel(readl(GPDR1) | (0xf << 26), GPDR1);
+		writel((readl(GAFR1_U) & ~(0xff << 20)) | (0xaa << 20),
+			GAFR1_U);
 
 		/* bits 74-77 */
-		GPDR2 |= (0xf << 10);
-		GAFR2_L = (GAFR2_L & ~(0xff << 20)) | (0xaa << 20);
+		writel(readl(GPDR2) | (0xf << 10), GPDR2);
+		writel((readl(GAFR2_L) & ~(0xff << 20)) | (0xaa << 20),
+			GAFR2_L);
 	}
 
 	/* 8 bit interface */
@@ -391,15 +495,17 @@ static void pxafb_setup_gpio (vidinfo_t *vid)
 	{
 		debug("Setting GPIO for 8 bit data\n");
 		/* bits 58-65 */
-		GPDR1 |= (0x3f << 26);
-		GPDR2 |= (0x3);
+		writel(readl(GPDR1) | (0x3f << 26), GPDR1);
+		writel(readl(GPDR2) | (0x3), GPDR2);
 
-		GAFR1_U = (GAFR1_U & ~(0xfff << 20)) | (0xaaa << 20);
-		GAFR2_L = (GAFR2_L & ~0xf) | (0xa);
+		writel((readl(GAFR1_U) & ~(0xfff << 20)) | (0xaaa << 20),
+			GAFR1_U);
+		writel((readl(GAFR2_L) & ~0xf) | (0xa), GAFR2_L);
 
 		/* bits 74-77 */
-		GPDR2 |= (0xf << 10);
-		GAFR2_L = (GAFR2_L & ~(0xff << 20)) | (0xaa << 20);
+		writel(readl(GPDR2) | (0xf << 10), GPDR2);
+		writel((readl(GAFR2_L) & ~(0xff << 20)) | (0xaa << 20),
+			GAFR2_L);
 	}
 
 	/* 16 bit interface */
@@ -407,11 +513,12 @@ static void pxafb_setup_gpio (vidinfo_t *vid)
 	{
 		debug("Setting GPIO for 16 bit data\n");
 		/* bits 58-77 */
-		GPDR1 |= (0x3f << 26);
-		GPDR2 |= 0x00003fff;
+		writel(readl(GPDR1) | (0x3f << 26), GPDR1);
+		writel(readl(GPDR2) | 0x00003fff, GPDR2);
 
-		GAFR1_U = (GAFR1_U & ~(0xfff << 20)) | (0xaaa << 20);
-		GAFR2_L = (GAFR2_L & 0xf0000000) | 0x0aaaaaaa;
+		writel((readl(GAFR1_U) & ~(0xfff << 20)) | (0xaaa << 20),
+			GAFR1_U);
+		writel((readl(GAFR2_L) & 0xf0000000) | 0x0aaaaaaa, GAFR2_L);
 	}
 	else
 	{
@@ -425,26 +532,26 @@ static void pxafb_enable_controller (vidinfo_t *vid)
 	debug("Enabling LCD controller\n");
 
 	/* Sequence from 11.7.10 */
-	LCCR3  = vid->pxa.reg_lccr3;
-	LCCR2  = vid->pxa.reg_lccr2;
-	LCCR1  = vid->pxa.reg_lccr1;
-	LCCR0  = vid->pxa.reg_lccr0 & ~LCCR0_ENB;
-	FDADR0 = vid->pxa.fdadr0;
-	FDADR1 = vid->pxa.fdadr1;
-	LCCR0 |= LCCR0_ENB;
+	writel(vid->pxa.reg_lccr3, LCCR3);
+	writel(vid->pxa.reg_lccr2, LCCR2);
+	writel(vid->pxa.reg_lccr1, LCCR1);
+	writel(vid->pxa.reg_lccr0 & ~LCCR0_ENB, LCCR0);
+	writel(vid->pxa.fdadr0, FDADR0);
+	writel(vid->pxa.fdadr1, FDADR1);
+	writel(readl(LCCR0) | LCCR0_ENB, LCCR0);
 
 #ifdef	CONFIG_CPU_MONAHANS
-	CKENA |= CKENA_1_LCD;
+	writel(readl(CKENA) | CKENA_1_LCD, CKENA);
 #else
-	CKEN |= CKEN16_LCD;
+	writel(readl(CKEN) | CKEN16_LCD, CKEN);
 #endif
 
-	debug("FDADR0 = 0x%08x\n", (unsigned int)FDADR0);
-	debug("FDADR1 = 0x%08x\n", (unsigned int)FDADR1);
-	debug("LCCR0 = 0x%08x\n", (unsigned int)LCCR0);
-	debug("LCCR1 = 0x%08x\n", (unsigned int)LCCR1);
-	debug("LCCR2 = 0x%08x\n", (unsigned int)LCCR2);
-	debug("LCCR3 = 0x%08x\n", (unsigned int)LCCR3);
+	debug("FDADR0 = 0x%08x\n", readl(FDADR0));
+	debug("FDADR1 = 0x%08x\n", readl(FDADR1));
+	debug("LCCR0 = 0x%08x\n", readl(LCCR0));
+	debug("LCCR1 = 0x%08x\n", readl(LCCR1));
+	debug("LCCR2 = 0x%08x\n", readl(LCCR2));
+	debug("LCCR3 = 0x%08x\n", readl(LCCR3));
 }
 
 static int pxafb_init (vidinfo_t *vid)
diff --git a/arch/arm/cpu/pxa/start.S b/arch/arm/cpu/pxa/start.S
index 9c5023b..684e44e 100644
--- a/arch/arm/cpu/pxa/start.S
+++ b/arch/arm/cpu/pxa/start.S
@@ -31,6 +31,14 @@
 #include <config.h>
 #include <version.h>
 #include <asm/arch/pxa-regs.h>
+#include <asm/arch/macro.h>
+
+/* takes care the CP15 update has taken place */
+.macro CPWAIT reg
+mrc  p15,0,\reg,c2,c0,0
+mov  \reg,\reg
+sub  pc,pc,#4
+.endm
 
 .globl _start
 _start: b	reset
@@ -86,11 +94,9 @@ _fiq:			.word fiq
 _TEXT_BASE:
 	.word	CONFIG_SYS_TEXT_BASE
 
-#if defined(CONFIG_SYS_ARM_WITHOUT_RELOC)
 .globl _armboot_start
 _armboot_start:
 	.word _start
-#endif
 
 /*
  * These are defined in the board-specific linker script.
@@ -115,7 +121,7 @@ FIQ_STACK_START:
 	.word 0x0badc0de
 #endif /* CONFIG_USE_IRQ */
 
-#if !defined(CONFIG_SYS_ARM_WITHOUT_RELOC)
+#ifndef CONFIG_PRELOADER
 /* IRQ stack memory (calculated at run-time) + 8 bytes */
 .globl IRQ_STACK_START_IN
 IRQ_STACK_START_IN:
@@ -159,12 +165,84 @@ reset:
 	msr	cpsr,r0
 
 	/*
-	 * we do sys-critical inits only at reboot,
-	 * not when booting from ram!
+	 * Enable MMU to use DCache as DRAM
 	 */
-#ifndef CONFIG_SKIP_LOWLEVEL_INIT
-	bl	cpu_init_crit
-#endif
+	/* Domain access -- enable for all CPs */
+	ldr	r0, =0x0000ffff
+	mcr	p15, 0, r0, c3, c0, 0
+
+	/* Point TTBR to MMU table */
+	ldr	r0, =mmu_table
+	adr	r2, _start
+	orr	r0, r2
+	mcr	p15, 0, r0, c2, c0, 0
+
+/* !!! Hereby, check if the code is running from SRAM !!! */
+/* If the code is running from SRAM, alias SRAM to 0x0 to simulate NOR. The code
+ * is linked to 0x0 too, so this makes things easier. */
+	cmp	r2, #0x5c000000
+
+	ldreq	r1, [r0]
+	orreq	r1, r2
+	streq	r1, [r0]
+
+	/* Kick in MMU, ICache, DCache, BTB */
+	mrc	p15, 0, r0, c1, c0, 0
+	bic	r0, #0x1b00
+	bic	r0, #0x0087
+	orr	r0, #0x1800
+	orr	r0, #0x0005
+	mcr	p15, 0, r0, c1, c0, 0
+	CPWAIT	r0
+
+	/* Unlock Icache, Dcache */
+	mcr	p15, 0, r0, c9, c1, 1
+	mcr	p15, 0, r0, c9, c2, 1
+
+	/* Flush Icache, Dcache, BTB */
+	mcr	p15, 0, r0, c7, c7, 0
+
+	/* Unlock I-TLB, D-TLB */
+	mcr	p15, 0, r0, c10, c4, 1
+	mcr	p15, 0, r0, c10, c8, 1
+
+	/* Flush TLB */
+	mcr	p15, 0, r0, c8, c7, 0
+	/* Allocate 4096 bytes of Dcache as RAM */
+
+	/* Drain pending loads and stores */
+	mcr	p15, 0, r0, c7, c10, 4
+
+	mov	r4, #0x00
+	mov	r5, #0x00
+	mov	r2, #0x01
+	mcr	p15, 0, r0, c9, c2, 0
+	CPWAIT	r0
+
+	/* 128 lines reserved (128 x 32bytes = 4096 bytes total) */
+	mov	r0, #128
+	mov	r1, #0xa0000000
+alloc:
+	mcr	p15, 0, r1, c7, c2, 5
+	/* Drain pending loads and stores */
+	mcr	p15, 0, r0, c7, c10, 4
+	strd	r4, [r1], #8
+	strd	r4, [r1], #8
+	strd	r4, [r1], #8
+	strd	r4, [r1], #8
+	subs	r0, #0x01
+	bne	alloc
+	/* Drain pending loads and stores */
+	mcr	p15, 0, r0, c7, c10, 4
+	mov	r2, #0x00
+	mcr	p15, 0, r2, c9, c2, 0
+	CPWAIT	r0
+
+	/* Jump to 0x0 ( + offset) if running from SRAM */
+	adr	r0, zerojmp
+	bic	r0, #0x5c000000
+	mov	pc, r0
+zerojmp:
 
 /* Set stackpointer in internal RAM to call board_init_f */
 call_board_init_f:
@@ -201,11 +279,13 @@ stack_setup:
 	beq	clear_bss
 
 #ifndef CONFIG_SKIP_RELOCATE_UBOOT
+	stmfd sp!, {r0-r12}
 copy_loop:
-	ldmia	r0!, {r9-r10}		/* copy from source address [r0]    */
-	stmia	r6!, {r9-r10}		/* copy to   target address [r1]    */
+	ldmia	r0!, {r3-r5, r7-r11}	/* copy from source address [r0]    */
+	stmia	r6!, {r3-r5, r7-r11}	/* copy to   target address [r1]    */
 	cmp	r0, r2			/* until source end address [r2]    */
 	blo	copy_loop
+	ldmfd sp!, {r0-r12}
 
 #ifndef CONFIG_PRELOADER
 	/* fix got entries */
@@ -274,218 +354,28 @@ _board_init_r: .word board_init_r
 
 /****************************************************************************/
 /*									    */
-/* the actual reset code						    */
+/* the actual reset code for OneNAND IPL				    */
 /*									    */
 /****************************************************************************/
 
+#ifndef	CONFIG_PXA27X
+#error OneNAND IPL is not supported on PXA25x and 26x due to lack of SRAM
+#endif
+
 reset:
-	mrs	r0,cpsr			/* set the CPU to SVC32 mode	    */
-	bic	r0,r0,#0x1f		/* (superviser mode, M=10011)	    */
+	/* Set CPU to SVC32 mode */
+	mrs	r0,cpsr
+	bic	r0,r0,#0x1f
 	orr	r0,r0,#0x13
 	msr	cpsr,r0
 
-	/*
-	 * we do sys-critical inits only at reboot,
-	 * not when booting from RAM!
-	 */
-#ifndef CONFIG_SKIP_LOWLEVEL_INIT
-	bl	cpu_init_crit		/* we do sys-critical inits	    */
-#endif /* !CONFIG_SKIP_LOWLEVEL_INIT */
-
-#ifndef CONFIG_SKIP_RELOCATE_UBOOT
-relocate:				/* relocate U-Boot to RAM	    */
-	adr	r0, _start		/* r0 <- current position of code   */
-	ldr	r1, _TEXT_BASE		/* test if we run from flash or RAM */
-#ifndef	CONFIG_PRELOADER
-	cmp	r0, r1			/* don't reloc during debug	    */
-	beq	stack_setup
-#endif
-
-	ldr	r2, _armboot_start
-	ldr	r3, _bss_start
-	sub	r2, r3, r2		/* r2 <- size of armboot	    */
-	add	r2, r0, r2		/* r2 <- source end address	    */
-
-copy_loop:
-	ldmia	r0!, {r3-r10}		/* copy from source address [r0]    */
-	stmia	r1!, {r3-r10}		/* copy to   target address [r1]    */
-	cmp	r0, r2			/* until source end address [r2]    */
-	blo	copy_loop
-#endif /* !CONFIG_SKIP_RELOCATE_UBOOT */
+	/* Point stack at the end of SRAM and leave 32 words for abort-stack */
+	ldr	sp, =0x5c03ff80
 
-	/* Set up the stack						    */
-stack_setup:
-	ldr	r0, _TEXT_BASE		/* upper 128 KiB: relocated uboot   */
-#ifdef CONFIG_PRELOADER
-	sub	sp, r0, #128		/* leave 32 words for abort-stack   */
-#else
-	sub	r0, r0, #CONFIG_SYS_MALLOC_LEN /* malloc area		    */
-	sub	r0, r0, #CONFIG_SYS_GBL_DATA_SIZE /* bdinfo		    */
-#ifdef CONFIG_USE_IRQ
-	sub	r0, r0, #(CONFIG_STACKSIZE_IRQ+CONFIG_STACKSIZE_FIQ)
-#endif /* CONFIG_USE_IRQ */
-	sub	sp, r0, #12		/* leave 3 words for abort-stack    */
-	bic	sp, sp, #7		/* 8-byte alignment for ABI compliance */
-#endif
+	/* Start OneNAND IPL */
+	ldr	pc, =start_oneboot
 
-clear_bss:
-	ldr	r0, _bss_start		/* find start of bss segment	    */
-	ldr	r1, _bss_end		/* stop here			    */
-	mov	r2, #0x00000000		/* clear			    */
-
-#ifndef CONFIG_PRELOADER
-clbss_l:str	r2, [r0]		/* clear loop...		    */
-	add	r0, r0, #4
-	cmp	r0, r1
-	blo	clbss_l
-#endif
-
-	ldr	pc, _start_armboot
-
-#ifdef CONFIG_ONENAND_IPL
-_start_armboot: .word start_oneboot
-#else
-_start_armboot: .word start_armboot
-#endif
-#endif /* #if !defined(CONFIG_SYS_ARM_WITHOUT_RELOC) */
-
-/****************************************************************************/
-/*									    */
-/* CPU_init_critical registers						    */
-/*									    */
-/* - setup important registers						    */
-/* - setup memory timing						    */
-/*									    */
-/****************************************************************************/
-/* mk@tbd: Fix this! */
-#undef RCSR
-#undef ICMR
-#undef OSMR3
-#undef OSCR
-#undef OWER
-#undef OIER
-#undef CCCR
-
-/* Interrupt-Controller base address					    */
-IC_BASE:	   .word	   0x40d00000
-#define ICMR	0x04
-
-/* Reset-Controller */
-RST_BASE:	.word	0x40f00030
-#define RCSR	0x00
-
-/* Operating System Timer */
-OSTIMER_BASE:	.word	0x40a00000
-#define OSMR3	0x0C
-#define OSCR	0x10
-#define OWER	0x18
-#define OIER	0x1C
-
-/* Clock Manager Registers						    */
-#ifdef CONFIG_CPU_MONAHANS
-# ifndef CONFIG_SYS_MONAHANS_RUN_MODE_OSC_RATIO
-#  error "You have to define CONFIG_SYS_MONAHANS_RUN_MODE_OSC_RATIO!!"
-# endif /* !CONFIG_SYS_MONAHANS_RUN_MODE_OSC_RATIO */
-# ifndef CONFIG_SYS_MONAHANS_TURBO_RUN_MODE_RATIO
-#  define CONFIG_SYS_MONAHANS_TURBO_RUN_MODE_RATIO 0x1
-# endif /* !CONFIG_SYS_MONAHANS_TURBO_RUN_MODE_RATIO */
-#else /* !CONFIG_CPU_MONAHANS */
-#ifdef CONFIG_SYS_CPUSPEED
-CC_BASE:	.word	0x41300000
-#define CCCR	0x00
-cpuspeed:	.word	CONFIG_SYS_CPUSPEED
-#else /* !CONFIG_SYS_CPUSPEED */
-#error "You have to define CONFIG_SYS_CPUSPEED!!"
-#endif /* CONFIG_SYS_CPUSPEED */
-#endif /* CONFIG_CPU_MONAHANS */
-
-	/* takes care the CP15 update has taken place */
-	.macro CPWAIT reg
-	mrc  p15,0,\reg,c2,c0,0
-	mov  \reg,\reg
-	sub  pc,pc,#4
-	.endm
-
-cpu_init_crit:
-
-	/* mask all IRQs						    */
-#ifndef CONFIG_CPU_MONAHANS
-	ldr	r0, IC_BASE
-	mov	r1, #0x00
-	str	r1, [r0, #ICMR]
-#else /* CONFIG_CPU_MONAHANS */
-	/* Step 1 - Enable CP6 permission */
-	mrc	p15, 0, r1, c15, c1, 0	@ read CPAR
-	orr	r1, r1, #0x40
-		mcr	p15, 0, r1, c15, c1, 0
-	CPWAIT	r1
-
-	/* Step 2 - Mask ICMR & ICMR2 */
-	mov	r1, #0
-	mcr	p6, 0, r1, c1, c0, 0	@ ICMR
-	mcr	p6, 0, r1, c7, c0, 0	@ ICMR2
-
-	/* turn off all clocks but the ones we will definitly require */
-	ldr	r1, =CKENA
-	ldr	r2, =(CKENA_22_FFUART | CKENA_10_SRAM | CKENA_9_SMC | CKENA_8_DMC)
-	str	r2, [r1]
-	ldr	r1, =CKENB
-	ldr	r2, =(CKENB_6_IRQ)
-	str	r2, [r1]
-#endif /* !CONFIG_CPU_MONAHANS */
-
-	/* set clock speed */
-#ifdef CONFIG_CPU_MONAHANS
-	ldr	r0, =ACCR
-	ldr	r1, =(((CONFIG_SYS_MONAHANS_TURBO_RUN_MODE_RATIO<<8) & ACCR_XN_MASK) | (CONFIG_SYS_MONAHANS_RUN_MODE_OSC_RATIO & ACCR_XL_MASK))
-	str	r1, [r0]
-#else /* !CONFIG_CPU_MONAHANS */
-#ifdef CONFIG_SYS_CPUSPEED
-	ldr	r0, CC_BASE
-	ldr	r1, cpuspeed
-	str	r1, [r0, #CCCR]
-	mov	r0, #2
-	mcr	p14, 0, r0, c6, c0, 0
-
-setspeed_done:
-
-#endif /* CONFIG_SYS_CPUSPEED */
-#endif /* CONFIG_CPU_MONAHANS */
-
-	/*
-	 * before relocating, we have to setup RAM timing
-	 * because memory timing is board-dependend, you will
-	 * find a lowlevel_init.S in your board directory.
-	 */
-	mov	ip,	lr
-	bl	lowlevel_init
-	mov	lr,	ip
-
-	/* Memory interfaces are working. Disable MMU and enable I-cache.   */
-	/* mk: hmm, this is not in the monahans docs, leave it now but
-	 *     check here if it doesn't work :-) */
-
-	ldr	r0, =0x2001		/* enable access to all coproc.	    */
-	mcr	p15, 0, r0, c15, c1, 0
-	CPWAIT r0
-
-	mcr	p15, 0, r0, c7, c10, 4	/* drain the write & fill buffers   */
-	CPWAIT r0
-
-	mcr	p15, 0, r0, c7, c7, 0	/* flush Icache, Dcache and BTB	    */
-	CPWAIT r0
-
-	mcr	p15, 0, r0, c8, c7, 0	/* flush instuction and data TLBs   */
-	CPWAIT r0
-
-	/* Enable the Icache						    */
-/*
-	mrc	p15, 0, r0, c1, c0, 0
-	orr	r0, r0, #0x1800
-	mcr	p15, 0, r0, c1, c0, 0
-	CPWAIT
-*/
-	mov	pc, lr
+#endif /* #if !defined(CONFIG_ONENAND_IPL) */
 
 #ifndef CONFIG_PRELOADER
 /****************************************************************************/
@@ -676,6 +566,12 @@ fiq:
 /* perform a watchdog timeout for a soft reset.				    */
 /*									    */
 /****************************************************************************/
+/* Operating System Timer */
+OSTIMER_BASE:	.word	0x40a00000
+#define OSMR3	0x0C
+#define OSCR	0x10
+#define OWER	0x18
+#define OIER	0x1C
 
 	.align	5
 .globl reset_cpu
@@ -703,3 +599,25 @@ reset_cpu:
 reset_endless:
 
 	b	reset_endless
+
+#ifndef CONFIG_PRELOADER
+.section .mmudata, "a"
+	.align	14
+	.globl	mmu_table
+mmu_table:
+	/* 0x00000000 - 0xa0000000 : 1:1, uncached mapping */
+	.set	__base, 0
+	.rept	0xa00
+	.word	(__base << 20) | 0xc12
+	.set	__base, __base + 1
+	.endr
+
+	/* 0xa0000000 - 0xa0100000 : 1:1, cached mapping */
+	.word	(0xa00 << 20) | 0x1c1e
+
+	.set	__base, 0xa01
+	.rept	0x1000 - 0xa01
+	.word	(__base << 20) | 0xc12
+	.set	__base, __base + 1
+	.endr
+#endif
diff --git a/arch/arm/cpu/pxa/timer.c b/arch/arm/cpu/pxa/timer.c
index 8d0f826..ec950c7 100644
--- a/arch/arm/cpu/pxa/timer.c
+++ b/arch/arm/cpu/pxa/timer.c
@@ -26,8 +26,9 @@
  * MA 02111-1307 USA
  */
 
-#include <common.h>
 #include <asm/arch/pxa-regs.h>
+#include <asm/io.h>
+#include <common.h>
 #include <div64.h>
 
 #ifdef CONFIG_USE_IRQ
@@ -86,7 +87,7 @@ void __udelay (unsigned long usec)
 
 void reset_timer_masked (void)
 {
-	OSCR = 0;
+	writel(0, OSCR);
 }
 
 ulong get_timer_masked (void)
@@ -113,7 +114,7 @@ void udelay_masked (unsigned long usec)
  */
 unsigned long long get_ticks(void)
 {
-	return OSCR;
+	return readl(OSCR);
 }
 
 /*
diff --git a/arch/arm/cpu/pxa/usb.c b/arch/arm/cpu/pxa/usb.c
index bd718a6..0311d5e 100644
--- a/arch/arm/cpu/pxa/usb.c
+++ b/arch/arm/cpu/pxa/usb.c
@@ -27,86 +27,79 @@
 # if defined(CONFIG_CPU_MONAHANS) || defined(CONFIG_PXA27X)
 
 #include <asm/arch/pxa-regs.h>
+#include <asm/io.h>
 #include <usb.h>
 
 int usb_cpu_init(void)
 {
 #if defined(CONFIG_CPU_MONAHANS)
 	/* Enable USB host clock. */
-	CKENA |= (CKENA_2_USBHOST |  CKENA_20_UDC);
+	writel(readl(CKENA) | CKENA_2_USBHOST | CKENA_20_UDC, CKENA);
 	udelay(100);
 #endif
 #if defined(CONFIG_PXA27X)
 	/* Enable USB host clock. */
-	CKEN |= CKEN10_USBHOST;
+	writel(readl(CKEN) | CKEN10_USBHOST, CKEN);
 #endif
 
 #if defined(CONFIG_CPU_MONAHANS)
 	/* Configure Port 2 for Host (USB Client Registers) */
-	UP2OCR = 0x3000c;
+	writel(0x3000c, UP2OCR);
 #endif
 
-	UHCHR |= UHCHR_FHR;
+	writel(readl(UHCHR) | UHCHR_FHR, UHCHR);
 	wait_ms(11);
-	UHCHR &= ~UHCHR_FHR;
+	writel(readl(UHCHR) & ~UHCHR_FHR, UHCHR);
 
-	UHCHR |= UHCHR_FSBIR;
-	while (UHCHR & UHCHR_FSBIR)
+	writel(readl(UHCHR) | UHCHR_FSBIR, UHCHR);
+	while (readl(UHCHR) & UHCHR_FSBIR)
 		udelay(1);
 
 #if defined(CONFIG_CPU_MONAHANS)
-	UHCHR &= ~UHCHR_SSEP0;
+	writel(readl(UHCHR) & ~UHCHR_SSEP0, UHCHR);
 #endif
 #if defined(CONFIG_PXA27X)
-	UHCHR &= ~UHCHR_SSEP2;
+	writel(readl(UHCHR) & ~UHCHR_SSEP2, UHCHR);
 #endif
-	UHCHR &= ~UHCHR_SSEP1;
-	UHCHR &= ~UHCHR_SSE;
+	writel(readl(UHCHR) & ~(UHCHR_SSEP1 | UHCHR_SSE), UHCHR);
 
 	return 0;
 }
 
 int usb_cpu_stop(void)
 {
-	UHCHR |= UHCHR_FHR;
+	writel(readl(UHCHR) | UHCHR_FHR, UHCHR);
 	udelay(11);
-	UHCHR &= ~UHCHR_FHR;
+	writel(readl(UHCHR) & ~UHCHR_FHR, UHCHR);
 
-	UHCCOMS |= 1;
+	writel(readl(UHCCOMS) | UHCHR_FHR, UHCCOMS);
 	udelay(10);
 
 #if defined(CONFIG_CPU_MONAHANS)
-	UHCHR |= UHCHR_SSEP0;
+	writel(readl(UHCHR) | UHCHR_SSEP0, UHCHR);
 #endif
 #if defined(CONFIG_PXA27X)
-	UHCHR |= UHCHR_SSEP2;
+	writel(readl(UHCHR) | UHCHR_SSEP2, UHCHR);
 #endif
-	UHCHR |= UHCHR_SSEP1;
-	UHCHR |= UHCHR_SSE;
-
-	return 0;
-}
-
-int usb_cpu_init_fail(void)
-{
-	UHCHR |= UHCHR_FHR;
-	udelay(11);
-	UHCHR &= ~UHCHR_FHR;
-
-	UHCCOMS |= 1;
-	udelay(10);
+	writel(readl(UHCHR) | UHCHR_SSEP1 | UHCHR_SSE, UHCHR);
 
 #if defined(CONFIG_CPU_MONAHANS)
-	UHCHR |= UHCHR_SSEP0;
+	/* Disable USB host clock. */
+	writel(readl(CKENA) & ~(CKENA_2_USBHOST | CKENA_20_UDC), CKENA);
+	udelay(100);
 #endif
 #if defined(CONFIG_PXA27X)
-	UHCHR |= UHCHR_SSEP2;
+	/* Disable USB host clock. */
+	writel(readl(CKEN) & ~CKEN10_USBHOST, CKEN);
 #endif
-	UHCHR |= UHCHR_SSEP1;
-	UHCHR |= UHCHR_SSE;
 
 	return 0;
 }
 
+int usb_cpu_init_fail(void)
+{
+	return usb_cpu_stop();
+}
+
 # endif /* defined(CONFIG_CPU_MONAHANS) || defined(CONFIG_PXA27X) */
 #endif /* defined(CONFIG_USB_OHCI) && defined(CONFIG_SYS_USB_OHCI_CPU_INIT) */