OpenRISC: Memory management

Signed-off-by: Jonas Bonn <jonas@southpole.se>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>

arch/openrisc/mm/fault.c

@@ -0,0 +1,338 @@
+/*
+ * OpenRISC fault.c
+ *
+ * Linux architectural port borrowing liberally from similar works of
+ * others.  All original copyrights apply as per the original source
+ * declaration.
+ *
+ * Modifications for the OpenRISC architecture:
+ * Copyright (C) 2003 Matjaz Breskvar <phoenix@bsemi.com>
+ * Copyright (C) 2010-2011 Jonas Bonn <jonas@southpole.se>
+ *
+ *      This program is free software; you can redistribute it and/or
+ *      modify it under the terms of the GNU General Public License
+ *      as published by the Free Software Foundation; either version
+ *      2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/module.h>
+#include <linux/sched.h>
+
+#include <asm/uaccess.h>
+#include <asm/siginfo.h>
+#include <asm/signal.h>
+
+#define NUM_TLB_ENTRIES 64
+#define TLB_OFFSET(add) (((add) >> PAGE_SHIFT) & (NUM_TLB_ENTRIES-1))
+
+unsigned long pte_misses;	/* updated by do_page_fault() */
+unsigned long pte_errors;	/* updated by do_page_fault() */
+
+/* __PHX__ :: - check the vmalloc_fault in do_page_fault()
+ *            - also look into include/asm-or32/mmu_context.h
+ */
+volatile pgd_t *current_pgd;
+
+extern void die(char *, struct pt_regs *, long);
+
+/*
+ * This routine handles page faults.  It determines the address,
+ * and the problem, and then passes it off to one of the appropriate
+ * routines.
+ *
+ * If this routine detects a bad access, it returns 1, otherwise it
+ * returns 0.
+ */
+
+asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long address,
+			      unsigned long vector, int write_acc)
+{
+	struct task_struct *tsk;
+	struct mm_struct *mm;
+	struct vm_area_struct *vma;
+	siginfo_t info;
+	int fault;
+
+	tsk = current;
+
+	/*
+	 * We fault-in kernel-space virtual memory on-demand. The
+	 * 'reference' page table is init_mm.pgd.
+	 *
+	 * NOTE! We MUST NOT take any locks for this case. We may
+	 * be in an interrupt or a critical region, and should
+	 * only copy the information from the master page table,
+	 * nothing more.
+	 *
+	 * NOTE2: This is done so that, when updating the vmalloc
+	 * mappings we don't have to walk all processes pgdirs and
+	 * add the high mappings all at once. Instead we do it as they
+	 * are used. However vmalloc'ed page entries have the PAGE_GLOBAL
+	 * bit set so sometimes the TLB can use a lingering entry.
+	 *
+	 * This verifies that the fault happens in kernel space
+	 * and that the fault was not a protection error.
+	 */
+
+	if (address >= VMALLOC_START &&
+	    (vector != 0x300 && vector != 0x400) &&
+	    !user_mode(regs))
+		goto vmalloc_fault;
+
+	/* If exceptions were enabled, we can reenable them here */
+	if (user_mode(regs)) {
+		/* Exception was in userspace: reenable interrupts */
+		local_irq_enable();
+	} else {
+		/* If exception was in a syscall, then IRQ's may have
+		 * been enabled or disabled.  If they were enabled,
+		 * reenable them.
+		 */
+		if (regs->sr && (SPR_SR_IEE | SPR_SR_TEE))
+			local_irq_enable();
+	}
+
+	mm = tsk->mm;
+	info.si_code = SEGV_MAPERR;
+
+	/*
+	 * If we're in an interrupt or have no user
+	 * context, we must not take the fault..
+	 */
+
+	if (in_interrupt() || !mm)
+		goto no_context;
+
+	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 (user_mode(regs)) {
+		/*
+		 * accessing the stack below usp is always a bug.
+		 * we get page-aligned addresses so we can only check
+		 * if we're within a page from usp, but that might be
+		 * enough to catch brutal errors at least.
+		 */
+		if (address + PAGE_SIZE < regs->sp)
+			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:
+	info.si_code = SEGV_ACCERR;
+
+	/* first do some preliminary protection checks */
+
+	if (write_acc) {
+		if (!(vma->vm_flags & VM_WRITE))
+			goto bad_area;
+	} else {
+		/* not present */
+		if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
+			goto bad_area;
+	}
+
+	/* are we trying to execute nonexecutable area */
+	if ((vector == 0x400) && !(vma->vm_page_prot.pgprot & _PAGE_EXEC))
+		goto bad_area;
+
+	/*
+	 * If for any reason at all we couldn't handle the fault,
+	 * make sure we exit gracefully rather than endlessly redo
+	 * the fault.
+	 */
+
+	fault = handle_mm_fault(mm, vma, address, write_acc);
+	if (unlikely(fault & VM_FAULT_ERROR)) {
+		if (fault & VM_FAULT_OOM)
+			goto out_of_memory;
+		else if (fault & VM_FAULT_SIGBUS)
+			goto do_sigbus;
+		BUG();
+	}
+	/*RGD modeled on Cris */
+	if (fault & VM_FAULT_MAJOR)
+		tsk->maj_flt++;
+	else
+		tsk->min_flt++;
+
+	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:
+	up_read(&mm->mmap_sem);
+
+bad_area_nosemaphore:
+
+	/* User mode accesses just cause a SIGSEGV */
+
+	if (user_mode(regs)) {
+		info.si_signo = SIGSEGV;
+		info.si_errno = 0;
+		/* info.si_code has been set above */
+		info.si_addr = (void *)address;
+		force_sig_info(SIGSEGV, &info, tsk);
+		return;
+	}
+
+no_context:
+
+	/* Are we prepared to handle this kernel fault?
+	 *
+	 * (The kernel has valid exception-points in the source
+	 *  when it acesses user-memory. When it fails in one
+	 *  of those points, we find it in a table and do a jump
+	 *  to some fixup code that loads an appropriate error
+	 *  code)
+	 */
+
+	{
+		const struct exception_table_entry *entry;
+
+		__asm__ __volatile__("l.nop 42");
+
+		if ((entry = search_exception_tables(regs->pc)) != NULL) {
+			/* Adjust the instruction pointer in the stackframe */
+			regs->pc = entry->fixup;
+			return;
+		}
+	}
+
+	/*
+	 * Oops. The kernel tried to access some bad page. We'll have to
+	 * terminate things with extreme prejudice.
+	 */
+
+	if ((unsigned long)(address) < PAGE_SIZE)
+		printk(KERN_ALERT
+		       "Unable to handle kernel NULL pointer dereference");
+	else
+		printk(KERN_ALERT "Unable to handle kernel access");
+	printk(" at virtual address 0x%08lx\n", address);
+
+	die("Oops", regs, write_acc);
+
+	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:
+	__asm__ __volatile__("l.nop 42");
+	__asm__ __volatile__("l.nop 1");
+
+	up_read(&mm->mmap_sem);
+	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.
+	 */
+	info.si_signo = SIGBUS;
+	info.si_errno = 0;
+	info.si_code = BUS_ADRERR;
+	info.si_addr = (void *)address;
+	force_sig_info(SIGBUS, &info, tsk);
+
+	/* Kernel mode? Handle exceptions or die */
+	if (!user_mode(regs))
+		goto no_context;
+	return;
+
+vmalloc_fault:
+	{
+		/*
+		 * Synchronize this task's top level page-table
+		 * with the 'reference' page table.
+		 *
+		 * Use current_pgd instead of tsk->active_mm->pgd
+		 * since the latter might be unavailable if this
+		 * code is executed in a misfortunately run irq
+		 * (like inside schedule() between switch_mm and
+		 *  switch_to...).
+		 */
+
+		int offset = pgd_index(address);
+		pgd_t *pgd, *pgd_k;
+		pud_t *pud, *pud_k;
+		pmd_t *pmd, *pmd_k;
+		pte_t *pte_k;
+
+/*
+		phx_warn("do_page_fault(): vmalloc_fault will not work, "
+			 "since current_pgd assign a proper value somewhere\n"
+			 "anyhow we don't need this at the moment\n");
+
+		phx_mmu("vmalloc_fault");
+*/
+		pgd = (pgd_t *)current_pgd + offset;
+		pgd_k = init_mm.pgd + offset;
+
+		/* Since we're two-level, we don't need to do both
+		 * set_pgd and set_pmd (they do the same thing). If
+		 * we go three-level at some point, do the right thing
+		 * with pgd_present and set_pgd here.
+		 *
+		 * Also, since the vmalloc area is global, we don't
+		 * need to copy individual PTE's, it is enough to
+		 * copy the pgd pointer into the pte page of the
+		 * root task. If that is there, we'll find our pte if
+		 * it exists.
+		 */
+
+		pud = pud_offset(pgd, address);
+		pud_k = pud_offset(pgd_k, address);
+		if (!pud_present(*pud_k))
+			goto no_context;
+
+		pmd = pmd_offset(pud, address);
+		pmd_k = pmd_offset(pud_k, address);
+
+		if (!pmd_present(*pmd_k))
+			goto bad_area_nosemaphore;
+
+		set_pmd(pmd, *pmd_k);
+
+		/* Make sure the actual PTE exists as well to
+		 * catch kernel vmalloc-area accesses to non-mapped
+		 * addresses. If we don't do this, this will just
+		 * silently loop forever.
+		 */
+
+		pte_k = pte_offset_kernel(pmd_k, address);
+		if (!pte_present(*pte_k))
+			goto no_context;
+
+		return;
+	}
+}

arch/openrisc/mm/init.c

@@ -0,0 +1,283 @@
+/*
+ * OpenRISC idle.c
+ *
+ * Linux architectural port borrowing liberally from similar works of
+ * others.  All original copyrights apply as per the original source
+ * declaration.
+ *
+ * Modifications for the OpenRISC architecture:
+ * Copyright (C) 2003 Matjaz Breskvar <phoenix@bsemi.com>
+ * Copyright (C) 2010-2011 Jonas Bonn <jonas@southpole.se>
+ *
+ *      This program is free software; you can redistribute it and/or
+ *      modify it under the terms of the GNU General Public License
+ *      as published by the Free Software Foundation; either version
+ *      2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/ptrace.h>
+#include <linux/mman.h>
+#include <linux/mm.h>
+#include <linux/swap.h>
+#include <linux/smp.h>
+#include <linux/bootmem.h>
+#include <linux/init.h>
+#include <linux/delay.h>
+#include <linux/blkdev.h>	/* for initrd_* */
+#include <linux/pagemap.h>
+#include <linux/memblock.h>
+
+#include <asm/system.h>
+#include <asm/segment.h>
+#include <asm/pgalloc.h>
+#include <asm/pgtable.h>
+#include <asm/dma.h>
+#include <asm/io.h>
+#include <asm/tlb.h>
+#include <asm/mmu_context.h>
+#include <asm/kmap_types.h>
+#include <asm/fixmap.h>
+#include <asm/tlbflush.h>
+
+int mem_init_done;
+
+DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
+
+static void __init zone_sizes_init(void)
+{
+	unsigned long zones_size[MAX_NR_ZONES];
+
+	/* Clear the zone sizes */
+	memset(zones_size, 0, sizeof(zones_size));
+
+	/*
+	 * We use only ZONE_NORMAL
+	 */
+	zones_size[ZONE_NORMAL] = max_low_pfn;
+
+	free_area_init(zones_size);
+}
+
+extern const char _s_kernel_ro[], _e_kernel_ro[];
+
+/*
+ * Map all physical memory into kernel's address space.
+ *
+ * This is explicitly coded for two-level page tables, so if you need
+ * something else then this needs to change.
+ */
+static void __init map_ram(void)
+{
+	unsigned long v, p, e;
+	pgprot_t prot;
+	pgd_t *pge;
+	pud_t *pue;
+	pmd_t *pme;
+	pte_t *pte;
+	/* These mark extents of read-only kernel pages...
+	 * ...from vmlinux.lds.S
+	 */
+	struct memblock_region *region;
+
+	v = PAGE_OFFSET;
+
+	for_each_memblock(memory, region) {
+		p = (u32) region->base & PAGE_MASK;
+		e = p + (u32) region->size;
+
+		v = (u32) __va(p);
+		pge = pgd_offset_k(v);
+
+		while (p < e) {
+			int j;
+			pue = pud_offset(pge, v);
+			pme = pmd_offset(pue, v);
+
+			if ((u32) pue != (u32) pge || (u32) pme != (u32) pge) {
+				panic("%s: OR1K kernel hardcoded for "
+				      "two-level page tables",
+				     __func__);
+			}
+
+			/* Alloc one page for holding PTE's... */
+			pte = (pte_t *) alloc_bootmem_low_pages(PAGE_SIZE);
+			set_pmd(pme, __pmd(_KERNPG_TABLE + __pa(pte)));
+
+			/* Fill the newly allocated page with PTE'S */
+			for (j = 0; p < e && j < PTRS_PER_PGD;
+			     v += PAGE_SIZE, p += PAGE_SIZE, j++, pte++) {
+				if (v >= (u32) _e_kernel_ro ||
+				    v < (u32) _s_kernel_ro)
+					prot = PAGE_KERNEL;
+				else
+					prot = PAGE_KERNEL_RO;
+
+				set_pte(pte, mk_pte_phys(p, prot));
+			}
+
+			pge++;
+		}
+
+		printk(KERN_INFO "%s: Memory: 0x%x-0x%x\n", __func__,
+		       region->base, region->base + region->size);
+	}
+}
+
+void __init paging_init(void)
+{
+	extern void tlb_init(void);
+
+	unsigned long end;
+	int i;
+
+	printk(KERN_INFO "Setting up paging and PTEs.\n");
+
+	/* clear out the init_mm.pgd that will contain the kernel's mappings */
+
+	for (i = 0; i < PTRS_PER_PGD; i++)
+		swapper_pg_dir[i] = __pgd(0);
+
+	/* make sure the current pgd table points to something sane
+	 * (even if it is most probably not used until the next
+	 *  switch_mm)
+	 */
+	current_pgd = init_mm.pgd;
+
+	end = (unsigned long)__va(max_low_pfn * PAGE_SIZE);
+
+	map_ram();
+
+	zone_sizes_init();
+
+	/* self modifying code ;) */
+	/* Since the old TLB miss handler has been running up until now,
+	 * the kernel pages are still all RW, so we can still modify the
+	 * text directly... after this change and a TLB flush, the kernel
+	 * pages will become RO.
+	 */
+	{
+		extern unsigned long dtlb_miss_handler;
+		extern unsigned long itlb_miss_handler;
+
+		unsigned long *dtlb_vector = __va(0x900);
+		unsigned long *itlb_vector = __va(0xa00);
+
+		printk(KERN_INFO "dtlb_miss_handler %p\n", &dtlb_miss_handler);
+		*dtlb_vector = ((unsigned long)&dtlb_miss_handler -
+				(unsigned long)dtlb_vector) >> 2;
+
+		printk(KERN_INFO "itlb_miss_handler %p\n", &itlb_miss_handler);
+		*itlb_vector = ((unsigned long)&itlb_miss_handler -
+				(unsigned long)itlb_vector) >> 2;
+	}
+
+	/* Invalidate instruction caches after code modification */
+	mtspr(SPR_ICBIR, 0x900);
+	mtspr(SPR_ICBIR, 0xa00);
+
+	/* New TLB miss handlers and kernel page tables are in now place.
+	 * Make sure that page flags get updated for all pages in TLB by
+	 * flushing the TLB and forcing all TLB entries to be recreated
+	 * from their page table flags.
+	 */
+	flush_tlb_all();
+}
+
+/* References to section boundaries */
+
+extern char _stext, _etext, _edata, __bss_start, _end;
+extern char __init_begin, __init_end;
+
+static int __init free_pages_init(void)
+{
+	int reservedpages, pfn;
+
+	/* this will put all low memory onto the freelists */
+	totalram_pages = free_all_bootmem();
+
+	reservedpages = 0;
+	for (pfn = 0; pfn < max_low_pfn; pfn++) {
+		/*
+		 * Only count reserved RAM pages
+		 */
+		if (PageReserved(mem_map + pfn))
+			reservedpages++;
+	}
+
+	return reservedpages;
+}
+
+static void __init set_max_mapnr_init(void)
+{
+	max_mapnr = num_physpages = max_low_pfn;
+}
+
+void __init mem_init(void)
+{
+	int codesize, reservedpages, datasize, initsize;
+
+	if (!mem_map)
+		BUG();
+
+	set_max_mapnr_init();
+
+	high_memory = (void *)__va(max_low_pfn * PAGE_SIZE);
+
+	/* clear the zero-page */
+	memset((void *)empty_zero_page, 0, PAGE_SIZE);
+
+	reservedpages = free_pages_init();
+
+	codesize = (unsigned long)&_etext - (unsigned long)&_stext;
+	datasize = (unsigned long)&_edata - (unsigned long)&_etext;
+	initsize = (unsigned long)&__init_end - (unsigned long)&__init_begin;
+
+	printk(KERN_INFO
+	       "Memory: %luk/%luk available (%dk kernel code, %dk reserved, %dk data, %dk init, %ldk highmem)\n",
+	       (unsigned long)nr_free_pages() << (PAGE_SHIFT - 10),
+	       max_mapnr << (PAGE_SHIFT - 10), codesize >> 10,
+	       reservedpages << (PAGE_SHIFT - 10), datasize >> 10,
+	       initsize >> 10, (unsigned long)(0 << (PAGE_SHIFT - 10))
+	    );
+
+	printk("mem_init_done ...........................................\n");
+	mem_init_done = 1;
+	return;
+}
+
+#ifdef CONFIG_BLK_DEV_INITRD
+void free_initrd_mem(unsigned long start, unsigned long end)
+{
+	printk(KERN_INFO "Freeing initrd memory: %ldk freed\n",
+	       (end - start) >> 10);
+
+	for (; start < end; start += PAGE_SIZE) {
+		ClearPageReserved(virt_to_page(start));
+		init_page_count(virt_to_page(start));
+		free_page(start);
+		totalram_pages++;
+	}
+}
+#endif
+
+void free_initmem(void)
+{
+	unsigned long addr;
+
+	addr = (unsigned long)(&__init_begin);
+	for (; addr < (unsigned long)(&__init_end); addr += PAGE_SIZE) {
+		ClearPageReserved(virt_to_page(addr));
+		init_page_count(virt_to_page(addr));
+		free_page(addr);
+		totalram_pages++;
+	}
+	printk(KERN_INFO "Freeing unused kernel memory: %luk freed\n",
+	       ((unsigned long)&__init_end -
+		(unsigned long)&__init_begin) >> 10);
+}

arch/openrisc/mm/ioremap.c

@@ -0,0 +1,137 @@
+/*
+ * OpenRISC ioremap.c
+ *
+ * Linux architectural port borrowing liberally from similar works of
+ * others.  All original copyrights apply as per the original source
+ * declaration.
+ *
+ * Modifications for the OpenRISC architecture:
+ * Copyright (C) 2003 Matjaz Breskvar <phoenix@bsemi.com>
+ * Copyright (C) 2010-2011 Jonas Bonn <jonas@southpole.se>
+ *
+ *      This program is free software; you can redistribute it and/or
+ *      modify it under the terms of the GNU General Public License
+ *      as published by the Free Software Foundation; either version
+ *      2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/vmalloc.h>
+#include <linux/io.h>
+#include <asm/pgalloc.h>
+#include <asm/kmap_types.h>
+#include <asm/fixmap.h>
+#include <asm/bug.h>
+#include <asm/pgtable.h>
+#include <linux/sched.h>
+#include <asm/tlbflush.h>
+
+extern int mem_init_done;
+
+static unsigned int fixmaps_used __initdata;
+
+/*
+ * Remap an arbitrary physical address space into the kernel virtual
+ * address space. Needed when the kernel wants to access high addresses
+ * directly.
+ *
+ * NOTE! We need to allow non-page-aligned mappings too: we will obviously
+ * have to convert them into an offset in a page-aligned mapping, but the
+ * caller shouldn't need to know that small detail.
+ */
+void __iomem *__init_refok
+__ioremap(phys_addr_t addr, unsigned long size, pgprot_t prot)
+{
+	phys_addr_t p;
+	unsigned long v;
+	unsigned long offset, last_addr;
+	struct vm_struct *area = NULL;
+
+	/* Don't allow wraparound or zero size */
+	last_addr = addr + size - 1;
+	if (!size || last_addr < addr)
+		return NULL;
+
+	/*
+	 * Mappings have to be page-aligned
+	 */
+	offset = addr & ~PAGE_MASK;
+	p = addr & PAGE_MASK;
+	size = PAGE_ALIGN(last_addr + 1) - p;
+
+	if (likely(mem_init_done)) {
+		area = get_vm_area(size, VM_IOREMAP);
+		if (!area)
+			return NULL;
+		v = (unsigned long)area->addr;
+	} else {
+		if ((fixmaps_used + (size >> PAGE_SHIFT)) > FIX_N_IOREMAPS)
+			return NULL;
+		v = fix_to_virt(FIX_IOREMAP_BEGIN + fixmaps_used);
+		fixmaps_used += (size >> PAGE_SHIFT);
+	}
+
+	if (ioremap_page_range(v, v + size, p, prot)) {
+		if (likely(mem_init_done))
+			vfree(area->addr);
+		else
+			fixmaps_used -= (size >> PAGE_SHIFT);
+		return NULL;
+	}
+
+	return (void __iomem *)(offset + (char *)v);
+}
+
+void iounmap(void *addr)
+{
+	/* If the page is from the fixmap pool then we just clear out
+	 * the fixmap mapping.
+	 */
+	if (unlikely((unsigned long)addr > FIXADDR_START)) {
+		/* This is a bit broken... we don't really know
+		 * how big the area is so it's difficult to know
+		 * how many fixed pages to invalidate...
+		 * just flush tlb and hope for the best...
+		 * consider this a FIXME
+		 *
+		 * Really we should be clearing out one or more page
+		 * table entries for these virtual addresses so that
+		 * future references cause a page fault... for now, we
+		 * rely on two things:
+		 *   i)  this code never gets called on known boards
+		 *   ii) invalid accesses to the freed areas aren't made
+		 */
+		flush_tlb_all();
+		return;
+	}
+
+	return vfree((void *)(PAGE_MASK & (unsigned long)addr));
+}
+
+/**
+ * OK, this one's a bit tricky... ioremap can get called before memory is
+ * initialized (early serial console does this) and will want to alloc a page
+ * for its mapping.  No userspace pages will ever get allocated before memory
+ * is initialized so this applies only to kernel pages.  In the event that
+ * this is called before memory is initialized we allocate the page using
+ * the memblock infrastructure.
+ */
+
+pte_t __init_refok *pte_alloc_one_kernel(struct mm_struct *mm,
+					 unsigned long address)
+{
+	pte_t *pte;
+
+	if (likely(mem_init_done)) {
+		pte = (pte_t *) __get_free_page(GFP_KERNEL | __GFP_REPEAT);
+	} else {
+		pte = (pte_t *) alloc_bootmem_low_pages(PAGE_SIZE);
+#if 0
+		/* FIXME: use memblock... */
+		pte = (pte_t *) __va(memblock_alloc(PAGE_SIZE, PAGE_SIZE));
+#endif
+	}
+
+	if (pte)
+		clear_page(pte);
+	return pte;
+}

arch/openrisc/mm/tlb.c

@@ -0,0 +1,193 @@
+/*
+ * OpenRISC tlb.c
+ *
+ * Linux architectural port borrowing liberally from similar works of
+ * others.  All original copyrights apply as per the original source
+ * declaration.
+ *
+ * Modifications for the OpenRISC architecture:
+ * Copyright (C) 2003 Matjaz Breskvar <phoenix@bsemi.com>
+ * Copyright (C) 2010-2011 Julius Baxter <julius.baxter@orsoc.se>
+ * Copyright (C) 2010-2011 Jonas Bonn <jonas@southpole.se>
+ *
+ *      This program is free software; you can redistribute it and/or
+ *      modify it under the terms of the GNU General Public License
+ *      as published by the Free Software Foundation; either version
+ *      2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/ptrace.h>
+#include <linux/mman.h>
+#include <linux/mm.h>
+#include <linux/init.h>
+
+#include <asm/system.h>
+#include <asm/segment.h>
+#include <asm/tlbflush.h>
+#include <asm/pgtable.h>
+#include <asm/mmu_context.h>
+#include <asm/spr_defs.h>
+
+#define NO_CONTEXT -1
+
+#define NUM_DTLB_SETS (1 << ((mfspr(SPR_IMMUCFGR) & SPR_IMMUCFGR_NTS) >> \
+			    SPR_DMMUCFGR_NTS_OFF))
+#define NUM_ITLB_SETS (1 << ((mfspr(SPR_IMMUCFGR) & SPR_IMMUCFGR_NTS) >> \
+			    SPR_IMMUCFGR_NTS_OFF))
+#define DTLB_OFFSET(addr) (((addr) >> PAGE_SHIFT) & (NUM_DTLB_SETS-1))
+#define ITLB_OFFSET(addr) (((addr) >> PAGE_SHIFT) & (NUM_ITLB_SETS-1))
+/*
+ * Invalidate all TLB entries.
+ *
+ * This comes down to setting the 'valid' bit for all xTLBMR registers to 0.
+ * Easiest way to accomplish this is to just zero out the xTLBMR register
+ * completely.
+ *
+ */
+
+void flush_tlb_all(void)
+{
+	int i;
+	unsigned long num_tlb_sets;
+
+	/* Determine number of sets for IMMU. */
+	/* FIXME: Assumption is I & D nsets equal. */
+	num_tlb_sets = NUM_ITLB_SETS;
+
+	for (i = 0; i < num_tlb_sets; i++) {
+		mtspr_off(SPR_DTLBMR_BASE(0), i, 0);
+		mtspr_off(SPR_ITLBMR_BASE(0), i, 0);
+	}
+}
+
+#define have_dtlbeir (mfspr(SPR_DMMUCFGR) & SPR_DMMUCFGR_TEIRI)
+#define have_itlbeir (mfspr(SPR_IMMUCFGR) & SPR_IMMUCFGR_TEIRI)
+
+/*
+ * Invalidate a single page.  This is what the xTLBEIR register is for.
+ *
+ * There's no point in checking the vma for PAGE_EXEC to determine whether it's
+ * the data or instruction TLB that should be flushed... that would take more
+ * than the few instructions that the following compiles down to!
+ *
+ * The case where we don't have the xTLBEIR register really only works for
+ * MMU's with a single way and is hard-coded that way.
+ */
+
+#define flush_dtlb_page_eir(addr) mtspr(SPR_DTLBEIR, addr)
+#define flush_dtlb_page_no_eir(addr) \
+	mtspr_off(SPR_DTLBMR_BASE(0), DTLB_OFFSET(addr), 0);
+
+#define flush_itlb_page_eir(addr) mtspr(SPR_ITLBEIR, addr)
+#define flush_itlb_page_no_eir(addr) \
+	mtspr_off(SPR_ITLBMR_BASE(0), ITLB_OFFSET(addr), 0);
+
+void flush_tlb_page(struct vm_area_struct *vma, unsigned long addr)
+{
+	if (have_dtlbeir)
+		flush_dtlb_page_eir(addr);
+	else
+		flush_dtlb_page_no_eir(addr);
+
+	if (have_itlbeir)
+		flush_itlb_page_eir(addr);
+	else
+		flush_itlb_page_no_eir(addr);
+}
+
+void flush_tlb_range(struct vm_area_struct *vma,
+		     unsigned long start, unsigned long end)
+{
+	int addr;
+	bool dtlbeir;
+	bool itlbeir;
+
+	dtlbeir = have_dtlbeir;
+	itlbeir = have_itlbeir;
+
+	for (addr = start; addr < end; addr += PAGE_SIZE) {
+		if (dtlbeir)
+			flush_dtlb_page_eir(addr);
+		else
+			flush_dtlb_page_no_eir(addr);
+
+		if (itlbeir)
+			flush_itlb_page_eir(addr);
+		else
+			flush_itlb_page_no_eir(addr);
+	}
+}
+
+/*
+ * Invalidate the selected mm context only.
+ *
+ * FIXME: Due to some bug here, we're flushing everything for now.
+ * This should be changed to loop over over mm and call flush_tlb_range.
+ */
+
+void flush_tlb_mm(struct mm_struct *mm)
+{
+
+	/* Was seeing bugs with the mm struct passed to us. Scrapped most of
+	   this function. */
+	/* Several architctures do this */
+	flush_tlb_all();
+}
+
+/* called in schedule() just before actually doing the switch_to */
+
+void switch_mm(struct mm_struct *prev, struct mm_struct *next,
+	       struct task_struct *next_tsk)
+{
+	/* remember the pgd for the fault handlers
+	 * this is similar to the pgd register in some other CPU's.
+	 * we need our own copy of it because current and active_mm
+	 * might be invalid at points where we still need to derefer
+	 * the pgd.
+	 */
+	current_pgd = next->pgd;
+
+	/* We don't have context support implemented, so flush all
+	 * entries belonging to previous map
+	 */
+
+	if (prev != next)
+		flush_tlb_mm(prev);
+
+}
+
+/*
+ * Initialize the context related info for a new mm_struct
+ * instance.
+ */
+
+int init_new_context(struct task_struct *tsk, struct mm_struct *mm)
+{
+	mm->context = NO_CONTEXT;
+	return 0;
+}
+
+/* called by __exit_mm to destroy the used MMU context if any before
+ * destroying the mm itself. this is only called when the last user of the mm
+ * drops it.
+ */
+
+void destroy_context(struct mm_struct *mm)
+{
+	flush_tlb_mm(mm);
+
+}
+
+/* called once during VM initialization, from init.c */
+
+void __init tlb_init(void)
+{
+	/* Do nothing... */
+	/* invalidate the entire TLB */
+	/* flush_tlb_all(); */
+}