[PATCH] avr32 architecture

This adds support for the Atmel AVR32 architecture as well as the AT32AP7000
CPU and the AT32STK1000 development board.

AVR32 is a new high-performance 32-bit RISC microprocessor core, designed for
cost-sensitive embedded applications, with particular emphasis on low power
consumption and high code density.  The AVR32 architecture is not binary
compatible with earlier 8-bit AVR architectures.

The AVR32 architecture, including the instruction set, is described by the
AVR32 Architecture Manual, available from

http://www.atmel.com/dyn/resources/prod_documents/doc32000.pdf

The Atmel AT32AP7000 is the first CPU implementing the AVR32 architecture.  It
features a 7-stage pipeline, 16KB instruction and data caches and a full
Memory Management Unit.  It also comes with a large set of integrated
peripherals, many of which are shared with the AT91 ARM-based controllers from
Atmel.

Full data sheet is available from

http://www.atmel.com/dyn/resources/prod_documents/doc32003.pdf

while the CPU core implementation including caches and MMU is documented by
the AVR32 AP Technical Reference, available from

http://www.atmel.com/dyn/resources/prod_documents/doc32001.pdf

Information about the AT32STK1000 development board can be found at

http://www.atmel.com/dyn/products/tools_card.asp?tool_id=3918

including a BSP CD image with an earlier version of this patch, development
tools (binaries and source/patches) and a root filesystem image suitable for
booting from SD card.

Alternatively, there's a preliminary "getting started" guide available at
http://avr32linux.org/twiki/bin/view/Main/GettingStarted which provides links
to the sources and patches you will need in order to set up a cross-compiling
environment for avr32-linux.

This patch, as well as the other patches included with the BSP and the
toolchain patches, is actively supported by Atmel Corporation.

[dmccr@us.ibm.com: Fix more pxx_page macro locations]
[bunk@stusta.de: fix `make defconfig']
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Dave McCracken <dmccr@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

9 files changed, 1718 insertions, 0 deletions

diff --git a/arch/avr32/mm/Makefile b/arch/avr32/mm/Makefile
new file mode 100644
index 00000000000..0066491f90d
--- /dev/null
+++ b/arch/avr32/mm/Makefile
@@ -0,0 +1,6 @@
+#
+# Makefile for the Linux/AVR32 kernel.
+#
+
+obj-y				+= init.o clear_page.o copy_page.o dma-coherent.o
+obj-y				+= ioremap.o cache.o fault.o tlb.o
diff --git a/arch/avr32/mm/cache.c b/arch/avr32/mm/cache.c
new file mode 100644
index 00000000000..450515b245a
--- /dev/null
+++ b/arch/avr32/mm/cache.c
@@ -0,0 +1,150 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/highmem.h>
+#include <linux/unistd.h>
+
+#include <asm/cacheflush.h>
+#include <asm/cachectl.h>
+#include <asm/processor.h>
+#include <asm/uaccess.h>
+
+/*
+ * If you attempt to flush anything more than this, you need superuser
+ * privileges.  The value is completely arbitrary.
+ */
+#define CACHEFLUSH_MAX_LEN	1024
+
+void invalidate_dcache_region(void *start, size_t size)
+{
+	unsigned long v, begin, end, linesz;
+
+	linesz = boot_cpu_data.dcache.linesz;
+
+	//printk("invalidate dcache: %p + %u\n", start, size);
+
+	/* You asked for it, you got it */
+	begin = (unsigned long)start & ~(linesz - 1);
+	end = ((unsigned long)start + size + linesz - 1) & ~(linesz - 1);
+
+	for (v = begin; v < end; v += linesz)
+		invalidate_dcache_line((void *)v);
+}
+
+void clean_dcache_region(void *start, size_t size)
+{
+	unsigned long v, begin, end, linesz;
+
+	linesz = boot_cpu_data.dcache.linesz;
+	begin = (unsigned long)start & ~(linesz - 1);
+	end = ((unsigned long)start + size + linesz - 1) & ~(linesz - 1);
+
+	for (v = begin; v < end; v += linesz)
+		clean_dcache_line((void *)v);
+	flush_write_buffer();
+}
+
+void flush_dcache_region(void *start, size_t size)
+{
+	unsigned long v, begin, end, linesz;
+
+	linesz = boot_cpu_data.dcache.linesz;
+	begin = (unsigned long)start & ~(linesz - 1);
+	end = ((unsigned long)start + size + linesz - 1) & ~(linesz - 1);
+
+	for (v = begin; v < end; v += linesz)
+		flush_dcache_line((void *)v);
+	flush_write_buffer();
+}
+
+void invalidate_icache_region(void *start, size_t size)
+{
+	unsigned long v, begin, end, linesz;
+
+	linesz = boot_cpu_data.icache.linesz;
+	begin = (unsigned long)start & ~(linesz - 1);
+	end = ((unsigned long)start + size + linesz - 1) & ~(linesz - 1);
+
+	for (v = begin; v < end; v += linesz)
+		invalidate_icache_line((void *)v);
+}
+
+static inline void __flush_icache_range(unsigned long start, unsigned long end)
+{
+	unsigned long v, linesz;
+
+	linesz = boot_cpu_data.dcache.linesz;
+	for (v = start; v < end; v += linesz) {
+		clean_dcache_line((void *)v);
+		invalidate_icache_line((void *)v);
+	}
+
+	flush_write_buffer();
+}
+
+/*
+ * This one is called after a module has been loaded.
+ */
+void flush_icache_range(unsigned long start, unsigned long end)
+{
+	unsigned long linesz;
+
+	linesz = boot_cpu_data.dcache.linesz;
+	__flush_icache_range(start & ~(linesz - 1),
+			     (end + linesz - 1) & ~(linesz - 1));
+}
+
+/*
+ * This one is called from do_no_page(), do_swap_page() and install_page().
+ */
+void flush_icache_page(struct vm_area_struct *vma, struct page *page)
+{
+	if (vma->vm_flags & VM_EXEC) {
+		void *v = kmap(page);
+		__flush_icache_range((unsigned long)v, (unsigned long)v + PAGE_SIZE);
+		kunmap(v);
+	}
+}
+
+/*
+ * This one is used by copy_to_user_page()
+ */
+void flush_icache_user_range(struct vm_area_struct *vma, struct page *page,
+			     unsigned long addr, int len)
+{
+	if (vma->vm_flags & VM_EXEC)
+		flush_icache_range(addr, addr + len);
+}
+
+asmlinkage int sys_cacheflush(int operation, void __user *addr, size_t len)
+{
+	int ret;
+
+	if (len > CACHEFLUSH_MAX_LEN) {
+		ret = -EPERM;
+		if (!capable(CAP_SYS_ADMIN))
+			goto out;
+	}
+
+	ret = -EFAULT;
+	if (!access_ok(VERIFY_WRITE, addr, len))
+		goto out;
+
+	switch (operation) {
+	case CACHE_IFLUSH:
+		flush_icache_range((unsigned long)addr,
+				   (unsigned long)addr + len);
+		ret = 0;
+		break;
+	default:
+		ret = -EINVAL;
+	}
+
+out:
+	return ret;
+}
diff --git a/arch/avr32/mm/clear_page.S b/arch/avr32/mm/clear_page.S
new file mode 100644
index 00000000000..5d70dca0069
--- /dev/null
+++ b/arch/avr32/mm/clear_page.S
@@ -0,0 +1,25 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/linkage.h>
+#include <asm/page.h>
+
+/*
+ * clear_page
+ * r12: P1 address (to)
+ */
+	.text
+	.global clear_page
+clear_page:
+	sub	r9, r12, -PAGE_SIZE
+	mov     r10, 0
+	mov	r11, 0
+0:      st.d    r12++, r10
+	cp      r12, r9
+	brne	0b
+	mov     pc, lr
diff --git a/arch/avr32/mm/copy_page.S b/arch/avr32/mm/copy_page.S
new file mode 100644
index 00000000000..c2b3752946b
--- /dev/null
+++ b/arch/avr32/mm/copy_page.S
@@ -0,0 +1,28 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/linkage.h>
+#include <asm/page.h>
+
+/*
+ * copy_page
+ *
+ * r12		to (P1 address)
+ * r11		from (P1 address)
+ * r8-r10	scratch
+ */
+	.text
+	.global copy_page
+copy_page:
+	sub	r10, r11, -(1 << PAGE_SHIFT)
+	/* pref	r11[0] */
+1:	/* pref	r11[8] */
+	ld.d	r8, r11++
+	st.d	r12++, r8
+	cp	r11, r10
+	brlo	1b
+	mov	pc, lr
diff --git a/arch/avr32/mm/dma-coherent.c b/arch/avr32/mm/dma-coherent.c
new file mode 100644
index 00000000000..44ab8a7bdae
--- /dev/null
+++ b/arch/avr32/mm/dma-coherent.c
@@ -0,0 +1,139 @@
+/*
+ *  Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/dma-mapping.h>
+
+#include <asm/addrspace.h>
+#include <asm/cacheflush.h>
+
+void dma_cache_sync(void *vaddr, size_t size, int direction)
+{
+	/*
+	 * No need to sync an uncached area
+	 */
+	if (PXSEG(vaddr) == P2SEG)
+		return;
+
+	switch (direction) {
+	case DMA_FROM_DEVICE:		/* invalidate only */
+		dma_cache_inv(vaddr, size);
+		break;
+	case DMA_TO_DEVICE:		/* writeback only */
+		dma_cache_wback(vaddr, size);
+		break;
+	case DMA_BIDIRECTIONAL:		/* writeback and invalidate */
+		dma_cache_wback_inv(vaddr, size);
+		break;
+	default:
+		BUG();
+	}
+}
+EXPORT_SYMBOL(dma_cache_sync);
+
+static struct page *__dma_alloc(struct device *dev, size_t size,
+				dma_addr_t *handle, gfp_t gfp)
+{
+	struct page *page, *free, *end;
+	int order;
+
+	size = PAGE_ALIGN(size);
+	order = get_order(size);
+
+	page = alloc_pages(gfp, order);
+	if (!page)
+		return NULL;
+	split_page(page, order);
+
+	/*
+	 * When accessing physical memory with valid cache data, we
+	 * get a cache hit even if the virtual memory region is marked
+	 * as uncached.
+	 *
+	 * Since the memory is newly allocated, there is no point in
+	 * doing a writeback. If the previous owner cares, he should
+	 * have flushed the cache before releasing the memory.
+	 */
+	invalidate_dcache_region(phys_to_virt(page_to_phys(page)), size);
+
+	*handle = page_to_bus(page);
+	free = page + (size >> PAGE_SHIFT);
+	end = page + (1 << order);
+
+	/*
+	 * Free any unused pages
+	 */
+	while (free < end) {
+		__free_page(free);
+		free++;
+	}
+
+	return page;
+}
+
+static void __dma_free(struct device *dev, size_t size,
+		       struct page *page, dma_addr_t handle)
+{
+	struct page *end = page + (PAGE_ALIGN(size) >> PAGE_SHIFT);
+
+	while (page < end)
+		__free_page(page++);
+}
+
+void *dma_alloc_coherent(struct device *dev, size_t size,
+			 dma_addr_t *handle, gfp_t gfp)
+{
+	struct page *page;
+	void *ret = NULL;
+
+	page = __dma_alloc(dev, size, handle, gfp);
+	if (page)
+		ret = phys_to_uncached(page_to_phys(page));
+
+	return ret;
+}
+EXPORT_SYMBOL(dma_alloc_coherent);
+
+void dma_free_coherent(struct device *dev, size_t size,
+		       void *cpu_addr, dma_addr_t handle)
+{
+	void *addr = phys_to_cached(uncached_to_phys(cpu_addr));
+	struct page *page;
+
+	pr_debug("dma_free_coherent addr %p (phys %08lx) size %u\n",
+		 cpu_addr, (unsigned long)handle, (unsigned)size);
+	BUG_ON(!virt_addr_valid(addr));
+	page = virt_to_page(addr);
+	__dma_free(dev, size, page, handle);
+}
+EXPORT_SYMBOL(dma_free_coherent);
+
+#if 0
+void *dma_alloc_writecombine(struct device *dev, size_t size,
+			     dma_addr_t *handle, gfp_t gfp)
+{
+	struct page *page;
+
+	page = __dma_alloc(dev, size, handle, gfp);
+
+	/* Now, map the page into P3 with write-combining turned on */
+	return __ioremap(page_to_phys(page), size, _PAGE_BUFFER);
+}
+EXPORT_SYMBOL(dma_alloc_writecombine);
+
+void dma_free_writecombine(struct device *dev, size_t size,
+			   void *cpu_addr, dma_addr_t handle)
+{
+	struct page *page;
+
+	iounmap(cpu_addr);
+
+	page = bus_to_page(handle);
+	__dma_free(dev, size, page, handle);
+}
+EXPORT_SYMBOL(dma_free_writecombine);
+#endif
diff --git a/arch/avr32/mm/fault.c b/arch/avr32/mm/fault.c
new file mode 100644
index 00000000000..678557260a3
--- /dev/null
+++ b/arch/avr32/mm/fault.c
@@ -0,0 +1,315 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * Based on linux/arch/sh/mm/fault.c:
+ *   Copyright (C) 1999  Niibe Yutaka
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/pagemap.h>
+
+#include <asm/kdebug.h>
+#include <asm/mmu_context.h>
+#include <asm/sysreg.h>
+#include <asm/uaccess.h>
+#include <asm/tlb.h>
+
+#ifdef DEBUG
+static void dump_code(unsigned long pc)
+{
+	char *p = (char *)pc;
+	char val;
+	int i;
+
+
+	printk(KERN_DEBUG "Code:");
+	for (i = 0; i < 16; i++) {
+		if (__get_user(val, p + i))
+			break;
+		printk(" %02x", val);
+	}
+	printk("\n");
+}
+#endif
+
+#ifdef CONFIG_KPROBES
+ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);
+
+/* Hook to register for page fault notifications */
+int register_page_fault_notifier(struct notifier_block *nb)
+{
+	return atomic_notifier_chain_register(&notify_page_fault_chain, nb);
+}
+
+int unregister_page_fault_notifier(struct notifier_block *nb)
+{
+	return atomic_notifier_chain_unregister(&notify_page_fault_chain, nb);
+}
+
+static inline int notify_page_fault(enum die_val val, struct pt_regs *regs,
+				    int trap, int sig)
+{
+	struct die_args args = {
+		.regs = regs,
+		.trapnr = trap,
+	};
+	return atomic_notifier_call_chain(&notify_page_fault_chain, val, &args);
+}
+#else
+static inline int notify_page_fault(enum die_val val, struct pt_regs *regs,
+				    int trap, int sig)
+{
+	return NOTIFY_DONE;
+}
+#endif
+
+/*
+ * This routine handles page faults. It determines the address and the
+ * problem, and then passes it off to one of the appropriate routines.
+ *
+ * ecr is the Exception Cause Register. Possible values are:
+ *   5:  Page not found (instruction access)
+ *   6:  Protection fault (instruction access)
+ *   12: Page not found (read access)
+ *   13: Page not found (write access)
+ *   14: Protection fault (read access)
+ *   15: Protection fault (write access)
+ */
+asmlinkage void do_page_fault(unsigned long ecr, struct pt_regs *regs)
+{
+	struct task_struct *tsk;
+	struct mm_struct *mm;
+	struct vm_area_struct *vma;
+	const struct exception_table_entry *fixup;
+	unsigned long address;
+	unsigned long page;
+	int writeaccess = 0;
+
+	if (notify_page_fault(DIE_PAGE_FAULT, regs,
+			      ecr, SIGSEGV) == NOTIFY_STOP)
+		return;
+
+	address = sysreg_read(TLBEAR);
+
+	tsk = current;
+	mm = tsk->mm;
+
+	/*
+	 * If we're in an interrupt or have no user context, we must
+	 * not take the fault...
+	 */
+	if (in_atomic() || !mm || regs->sr & SYSREG_BIT(GM))
+		goto no_context;
+
+	local_irq_enable();
+
+	down_read(&mm->mmap_sem);
+
+	vma = find_vma(mm, address);
+	if (!vma)
+		goto bad_area;
+	if (vma->vm_start <= address)
+		goto good_area;
+	if (!(vma->vm_flags & VM_GROWSDOWN))
+		goto bad_area;
+	if (expand_stack(vma, address))
+		goto bad_area;
+
+	/*
+	 * Ok, we have a good vm_area for this memory access, so we
+	 * can handle it...
+	 */
+good_area:
+	//pr_debug("good area: vm_flags = 0x%lx\n", vma->vm_flags);
+	switch (ecr) {
+	case ECR_PROTECTION_X:
+	case ECR_TLB_MISS_X:
+		if (!(vma->vm_flags & VM_EXEC))
+			goto bad_area;
+		break;
+	case ECR_PROTECTION_R:
+	case ECR_TLB_MISS_R:
+		if (!(vma->vm_flags & (VM_READ | VM_WRITE | VM_EXEC)))
+			goto bad_area;
+		break;
+	case ECR_PROTECTION_W:
+	case ECR_TLB_MISS_W:
+		if (!(vma->vm_flags & VM_WRITE))
+			goto bad_area;
+		writeaccess = 1;
+		break;
+	default:
+		panic("Unhandled case %lu in do_page_fault!", ecr);
+	}
+
+	/*
+	 * If for any reason at all we couldn't handle the fault, make
+	 * sure we exit gracefully rather than endlessly redo the
+	 * fault.
+	 */
+survive:
+	switch (handle_mm_fault(mm, vma, address, writeaccess)) {
+	case VM_FAULT_MINOR:
+		tsk->min_flt++;
+		break;
+	case VM_FAULT_MAJOR:
+		tsk->maj_flt++;
+		break;
+	case VM_FAULT_SIGBUS:
+		goto do_sigbus;
+	case VM_FAULT_OOM:
+		goto out_of_memory;
+	default:
+		BUG();
+	}
+
+	up_read(&mm->mmap_sem);
+	return;
+
+	/*
+	 * Something tried to access memory that isn't in our memory
+	 * map. Fix it, but check if it's kernel or user first...
+	 */
+bad_area:
+	pr_debug("Bad area [%s:%u]: addr %08lx, ecr %lu\n",
+		 tsk->comm, tsk->pid, address, ecr);
+
+	up_read(&mm->mmap_sem);
+
+	if (user_mode(regs)) {
+		/* Hmm...we have to pass address and ecr somehow... */
+		/* tsk->thread.address = address;
+		   tsk->thread.error_code = ecr; */
+#ifdef DEBUG
+		show_regs(regs);
+		dump_code(regs->pc);
+
+		page = sysreg_read(PTBR);
+		printk("ptbr = %08lx", page);
+		if (page) {
+			page = ((unsigned long *)page)[address >> 22];
+			printk(" pgd = %08lx", page);
+			if (page & _PAGE_PRESENT) {
+				page &= PAGE_MASK;
+				address &= 0x003ff000;
+				page = ((unsigned long *)__va(page))[address >> PAGE_SHIFT];
+				printk(" pte = %08lx\n", page);
+			}
+		}
+#endif
+		pr_debug("Sending SIGSEGV to PID %d...\n",
+			tsk->pid);
+		force_sig(SIGSEGV, tsk);
+		return;
+	}
+
+no_context:
+	pr_debug("No context\n");
+
+	/* Are we prepared to handle this kernel fault? */
+	fixup = search_exception_tables(regs->pc);
+	if (fixup) {
+		regs->pc = fixup->fixup;
+		pr_debug("Found fixup at %08lx\n", fixup->fixup);
+		return;
+	}
+
+	/*
+	 * Oops. The kernel tried to access some bad page. We'll have
+	 * to terminate things with extreme prejudice.
+	 */
+	if (address < PAGE_SIZE)
+		printk(KERN_ALERT
+		       "Unable to handle kernel NULL pointer dereference");
+	else
+		printk(KERN_ALERT
+		       "Unable to handle kernel paging request");
+	printk(" at virtual address %08lx\n", address);
+	printk(KERN_ALERT "pc = %08lx\n", regs->pc);
+
+	page = sysreg_read(PTBR);
+	printk(KERN_ALERT "ptbr = %08lx", page);
+	if (page) {
+		page = ((unsigned long *)page)[address >> 22];
+		printk(" pgd = %08lx", page);
+		if (page & _PAGE_PRESENT) {
+			page &= PAGE_MASK;
+			address &= 0x003ff000;
+			page = ((unsigned long *)__va(page))[address >> PAGE_SHIFT];
+			printk(" pte = %08lx\n", page);
+		}
+	}
+	die("\nOops", regs, ecr);
+	do_exit(SIGKILL);
+
+	/*
+	 * We ran out of memory, or some other thing happened to us
+	 * that made us unable to handle the page fault gracefully.
+	 */
+out_of_memory:
+	printk("Out of memory\n");
+	up_read(&mm->mmap_sem);
+	if (current->pid == 1) {
+		yield();
+		down_read(&mm->mmap_sem);
+		goto survive;
+	}
+	printk("VM: Killing process %s\n", tsk->comm);
+	if (user_mode(regs))
+		do_exit(SIGKILL);
+	goto no_context;
+
+do_sigbus:
+	up_read(&mm->mmap_sem);
+
+	/*
+	 * Send a sigbus, regardless of whether we were in kernel or
+	 * user mode.
+	 */
+	/* address, error_code, trap_no, ... */
+#ifdef DEBUG
+	show_regs(regs);
+	dump_code(regs->pc);
+#endif
+	pr_debug("Sending SIGBUS to PID %d...\n", tsk->pid);
+	force_sig(SIGBUS, tsk);
+
+	/* Kernel mode? Handle exceptions or die */
+	if (!user_mode(regs))
+		goto no_context;
+}
+
+asmlinkage void do_bus_error(unsigned long addr, int write_access,
+			     struct pt_regs *regs)
+{
+	printk(KERN_ALERT
+	       "Bus error at physical address 0x%08lx (%s access)\n",
+	       addr, write_access ? "write" : "read");
+	printk(KERN_INFO "DTLB dump:\n");
+	dump_dtlb();
+	die("Bus Error", regs, write_access);
+	do_exit(SIGKILL);
+}
+
+/*
+ * This functionality is currently not possible to implement because
+ * we're using segmentation to ensure a fixed mapping of the kernel
+ * virtual address space.
+ *
+ * It would be possible to implement this, but it would require us to
+ * disable segmentation at startup and load the kernel mappings into
+ * the TLB like any other pages. There will be lots of trickery to
+ * avoid recursive invocation of the TLB miss handler, though...
+ */
+#ifdef CONFIG_DEBUG_PAGEALLOC
+void kernel_map_pages(struct page *page, int numpages, int enable)
+{
+
+}
+EXPORT_SYMBOL(kernel_map_pages);
+#endif
diff --git a/arch/avr32/mm/init.c b/arch/avr32/mm/init.c
new file mode 100644
index 00000000000..3e6c4103980
--- /dev/null
+++ b/arch/avr32/mm/init.c
@@ -0,0 +1,480 @@
+/*
+ * Copyright (C) 2004-2006 Atmel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/swap.h>
+#include <linux/init.h>
+#include <linux/initrd.h>
+#include <linux/mmzone.h>
+#include <linux/bootmem.h>
+#include <linux/pagemap.h>
+#include <linux/pfn.h>
+#include <linux/nodemask.h>
+
+#include <asm/page.h>
+#include <asm/mmu_context.h>
+#include <asm/tlb.h>
+#include <asm/io.h>
+#include <asm/dma.h>
+#include <asm/setup.h>
+#include <asm/sections.h>
+
+DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
+
+pgd_t swapper_pg_dir[PTRS_PER_PGD];
+
+struct page *empty_zero_page;
+
+/*
+ * Cache of MMU context last used.
+ */
+unsigned long mmu_context_cache = NO_CONTEXT;
+
+#define START_PFN	(NODE_DATA(0)->bdata->node_boot_start >> PAGE_SHIFT)
+#define MAX_LOW_PFN	(NODE_DATA(0)->bdata->node_low_pfn)
+
+void show_mem(void)
+{
+	int total = 0, reserved = 0, cached = 0;
+	int slab = 0, free = 0, shared = 0;
+	pg_data_t *pgdat;
+
+	printk("Mem-info:\n");
+	show_free_areas();
+
+	for_each_online_pgdat(pgdat) {
+		struct page *page, *end;
+
+		page = pgdat->node_mem_map;
+		end = page + pgdat->node_spanned_pages;
+
+		do {
+			total++;
+			if (PageReserved(page))
+				reserved++;
+			else if (PageSwapCache(page))
+				cached++;
+			else if (PageSlab(page))
+				slab++;
+			else if (!page_count(page))
+				free++;
+			else
+				shared += page_count(page) - 1;
+			page++;
+		} while (page < end);
+	}
+
+	printk ("%d pages of RAM\n", total);
+	printk ("%d free pages\n", free);
+	printk ("%d reserved pages\n", reserved);
+	printk ("%d slab pages\n", slab);
+	printk ("%d pages shared\n", shared);
+	printk ("%d pages swap cached\n", cached);
+}
+
+static void __init print_memory_map(const char *what,
+				    struct tag_mem_range *mem)
+{
+	printk ("%s:\n", what);
+	for (; mem; mem = mem->next) {
+		printk ("  %08lx - %08lx\n",
+			(unsigned long)mem->addr,
+			(unsigned long)(mem->addr + mem->size));
+	}
+}
+
+#define MAX_LOWMEM	HIGHMEM_START
+#define MAX_LOWMEM_PFN	PFN_DOWN(MAX_LOWMEM)
+
+/*
+ * Sort a list of memory regions in-place by ascending address.
+ *
+ * We're using bubble sort because we only have singly linked lists
+ * with few elements.
+ */
+static void __init sort_mem_list(struct tag_mem_range **pmem)
+{
+	int done;
+	struct tag_mem_range **a, **b;
+
+	if (!*pmem)
+		return;
+
+	do {
+		done = 1;
+		a = pmem, b = &(*pmem)->next;
+		while (*b) {
+			if ((*a)->addr > (*b)->addr) {
+				struct tag_mem_range *tmp;
+				tmp = (*b)->next;
+				(*b)->next = *a;
+				*a = *b;
+				*b = tmp;
+				done = 0;
+			}
+			a = &(*a)->next;
+			b = &(*a)->next;
+		}
+	} while (!done);
+}
+
+/*
+ * Find a free memory region large enough for storing the
+ * bootmem bitmap.
+ */
+static unsigned long __init
+find_bootmap_pfn(const struct tag_mem_range *mem)
+{
+	unsigned long bootmap_pages, bootmap_len;
+	unsigned long node_pages = PFN_UP(mem->size);
+	unsigned long bootmap_addr = mem->addr;
+	struct tag_mem_range *reserved = mem_reserved;
+	struct tag_mem_range *ramdisk = mem_ramdisk;
+	unsigned long kern_start = virt_to_phys(_stext);
+	unsigned long kern_end = virt_to_phys(_end);
+
+	bootmap_pages = bootmem_bootmap_pages(node_pages);
+	bootmap_len = bootmap_pages << PAGE_SHIFT;
+
+	/*
+	 * Find a large enough region without reserved pages for
+	 * storing the bootmem bitmap. We can take advantage of the
+	 * fact that all lists have been sorted.
+	 *
+	 * We have to check explicitly reserved regions as well as the
+	 * kernel image and any RAMDISK images...
+	 *
+	 * Oh, and we have to make sure we don't overwrite the taglist
+	 * since we're going to use it until the bootmem allocator is
+	 * fully up and running.
+	 */
+	while (1) {
+		if ((bootmap_addr < kern_end) &&
+		    ((bootmap_addr + bootmap_len) > kern_start))
+			bootmap_addr = kern_end;
+
+		while (reserved &&
+		       (bootmap_addr >= (reserved->addr + reserved->size)))
+			reserved = reserved->next;
+
+		if (reserved &&
+		    ((bootmap_addr + bootmap_len) >= reserved->addr)) {
+			bootmap_addr = reserved->addr + reserved->size;
+			continue;
+		}
+
+		while (ramdisk &&
+		       (bootmap_addr >= (ramdisk->addr + ramdisk->size)))
+			ramdisk = ramdisk->next;
+
+		if (!ramdisk ||
+		    ((bootmap_addr + bootmap_len) < ramdisk->addr))
+			break;
+
+		bootmap_addr = ramdisk->addr + ramdisk->size;
+	}
+
+	if ((PFN_UP(bootmap_addr) + bootmap_len) >= (mem->addr + mem->size))
+		return ~0UL;
+
+	return PFN_UP(bootmap_addr);
+}
+
+void __init setup_bootmem(void)
+{
+	unsigned bootmap_size;
+	unsigned long first_pfn, bootmap_pfn, pages;
+	unsigned long max_pfn, max_low_pfn;
+	unsigned long kern_start = virt_to_phys(_stext);
+	unsigned long kern_end = virt_to_phys(_end);
+	unsigned node = 0;
+	struct tag_mem_range *bank, *res;
+
+	sort_mem_list(&mem_phys);
+	sort_mem_list(&mem_reserved);
+
+	print_memory_map("Physical memory", mem_phys);
+	print_memory_map("Reserved memory", mem_reserved);
+
+	nodes_clear(node_online_map);
+
+	if (mem_ramdisk) {
+#ifdef CONFIG_BLK_DEV_INITRD
+		initrd_start = __va(mem_ramdisk->addr);
+		initrd_end = initrd_start + mem_ramdisk->size;
+
+		print_memory_map("RAMDISK images", mem_ramdisk);
+		if (mem_ramdisk->next)
+			printk(KERN_WARNING
+			       "Warning: Only the first RAMDISK image "
+			       "will be used\n");
+		sort_mem_list(&mem_ramdisk);
+#else
+		printk(KERN_WARNING "RAM disk image present, but "
+		       "no initrd support in kernel!\n");
+#endif
+	}
+
+	if (mem_phys->next)
+		printk(KERN_WARNING "Only using first memory bank\n");
+
+	for (bank = mem_phys; bank; bank = NULL) {
+		first_pfn = PFN_UP(bank->addr);
+		max_low_pfn = max_pfn = PFN_DOWN(bank->addr + bank->size);
+		bootmap_pfn = find_bootmap_pfn(bank);
+		if (bootmap_pfn > max_pfn)
+			panic("No space for bootmem bitmap!\n");
+
+		if (max_low_pfn > MAX_LOWMEM_PFN) {
+			max_low_pfn = MAX_LOWMEM_PFN;
+#ifndef CONFIG_HIGHMEM
+			/*
+			 * Lowmem is memory that can be addressed
+			 * directly through P1/P2
+			 */
+			printk(KERN_WARNING
+			       "Node %u: Only %ld MiB of memory will be used.\n",
+			       node, MAX_LOWMEM >> 20);
+			printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n");
+#else
+#error HIGHMEM is not supported by AVR32 yet
+#endif
+		}
+
+		/* Initialize the boot-time allocator with low memory only. */
+		bootmap_size = init_bootmem_node(NODE_DATA(node), bootmap_pfn,
+						 first_pfn, max_low_pfn);
+
+		printk("Node %u: bdata = %p, bdata->node_bootmem_map = %p\n",
+		       node, NODE_DATA(node)->bdata,
+		       NODE_DATA(node)->bdata->node_bootmem_map);
+
+		/*
+		 * Register fully available RAM pages with the bootmem
+		 * allocator.
+		 */
+		pages = max_low_pfn - first_pfn;
+		free_bootmem_node (NODE_DATA(node), PFN_PHYS(first_pfn),
+				   PFN_PHYS(pages));
+
+		/*
+		 * Reserve space for the kernel image (if present in
+		 * this node)...
+		 */
+		if ((kern_start >= PFN_PHYS(first_pfn)) &&
+		    (kern_start < PFN_PHYS(max_pfn))) {
+			printk("Node %u: Kernel image %08lx - %08lx\n",
+			       node, kern_start, kern_end);
+			reserve_bootmem_node(NODE_DATA(node), kern_start,
+					     kern_end - kern_start);
+		}
+
+		/* ...the bootmem bitmap... */
+		reserve_bootmem_node(NODE_DATA(node),
+				     PFN_PHYS(bootmap_pfn),
+				     bootmap_size);
+
+		/* ...any RAMDISK images... */
+		for (res = mem_ramdisk; res; res = res->next) {
+			if (res->addr > PFN_PHYS(max_pfn))
+				break;
+
+			if (res->addr >= PFN_PHYS(first_pfn)) {
+				printk("Node %u: RAMDISK %08lx - %08lx\n",
+				       node,
+				       (unsigned long)res->addr,
+				       (unsigned long)(res->addr + res->size));
+				reserve_bootmem_node(NODE_DATA(node),
+						     res->addr, res->size);
+			}
+		}
+
+		/* ...and any other reserved regions. */
+		for (res = mem_reserved; res; res = res->next) {
+			if (res->addr > PFN_PHYS(max_pfn))
+				break;
+
+			if (res->addr >= PFN_PHYS(first_pfn)) {
+				printk("Node %u: Reserved %08lx - %08lx\n",
+				       node,
+				       (unsigned long)res->addr,
+				       (unsigned long)(res->addr + res->size));
+				reserve_bootmem_node(NODE_DATA(node),
+						     res->addr, res->size);
+			}
+		}
+
+		node_set_online(node);
+	}
+}
+
+/*
+ * paging_init() sets up the page tables
+ *
+ * This routine also unmaps the page at virtual kernel address 0, so
+ * that we can trap those pesky NULL-reference errors in the kernel.
+ */
+void __init paging_init(void)
+{
+	extern unsigned long _evba;
+	void *zero_page;
+	int nid;
+
+	/*
+	 * Make sure we can handle exceptions before enabling
+	 * paging. Not that we should ever _get_ any exceptions this
+	 * early, but you never know...
+	 */
+	printk("Exception vectors start at %p\n", &_evba);
+	sysreg_write(EVBA, (unsigned long)&_evba);
+
+	/*
+	 * Since we are ready to handle exceptions now, we should let
+	 * the CPU generate them...
+	 */
+	__asm__ __volatile__ ("csrf %0" : : "i"(SR_EM_BIT));
+
+	/*
+	 * Allocate the zero page. The allocator will panic if it
+	 * can't satisfy the request, so no need to check.
+	 */
+	zero_page = alloc_bootmem_low_pages_node(NODE_DATA(0),
+						 PAGE_SIZE);
+
+	{
+		pgd_t *pg_dir;
+		int i;
+
+		pg_dir = swapper_pg_dir;
+		sysreg_write(PTBR, (unsigned long)pg_dir);
+
+		for (i = 0; i < PTRS_PER_PGD; i++)
+			pgd_val(pg_dir[i]) = 0;
+
+		enable_mmu();
+		printk ("CPU: Paging enabled\n");
+	}
+
+	for_each_online_node(nid) {
+		pg_data_t *pgdat = NODE_DATA(nid);
+		unsigned long zones_size[MAX_NR_ZONES];
+		unsigned long low, start_pfn;
+
+		start_pfn = pgdat->bdata->node_boot_start;
+		start_pfn >>= PAGE_SHIFT;
+		low = pgdat->bdata->node_low_pfn;
+
+		memset(zones_size, 0, sizeof(zones_size));
+		zones_size[ZONE_NORMAL] = low - start_pfn;
+
+		printk("Node %u: start_pfn = 0x%lx, low = 0x%lx\n",
+		       nid, start_pfn, low);
+
+		free_area_init_node(nid, pgdat, zones_size, start_pfn, NULL);
+
+		printk("Node %u: mem_map starts at %p\n",
+		       pgdat->node_id, pgdat->node_mem_map);
+	}
+
+	mem_map = NODE_DATA(0)->node_mem_map;
+
+	memset(zero_page, 0, PAGE_SIZE);
+	empty_zero_page = virt_to_page(zero_page);
+	flush_dcache_page(empty_zero_page);
+}
+
+void __init mem_init(void)
+{
+	int codesize, reservedpages, datasize, initsize;
+	int nid, i;
+
+	reservedpages = 0;
+	high_memory = NULL;
+