Merge with Linux 2.4.2.

5 files changed, 1265 insertions, 0 deletions

diff --git a/arch/cris/mm/Makefile b/arch/cris/mm/Makefile
new file mode 100644
index 000000000..d1d21a7b7
--- /dev/null
+++ b/arch/cris/mm/Makefile
@@ -0,0 +1,13 @@
+#
+# Makefile for the linux cris-specific parts of the memory manager.
+#
+# Note! Dependencies are done automagically by 'make dep', which also
+# removes any old dependencies. DON'T put your own dependencies here
+# unless it's something special (ie not a .c file).
+#
+# Note 2! The CFLAGS definition is now in the main makefile...
+
+O_TARGET := mm.o
+obj-y	 := init.o fault.o tlb.o extable.o
+
+include $(TOPDIR)/Rules.make
diff --git a/arch/cris/mm/extable.c b/arch/cris/mm/extable.c
new file mode 100644
index 000000000..a4cf00f14
--- /dev/null
+++ b/arch/cris/mm/extable.c
@@ -0,0 +1,55 @@
+/*
+ * linux/arch/cris/mm/extable.c
+ */
+
+#include <linux/config.h>
+#include <linux/module.h>
+#include <asm/uaccess.h>
+
+extern const struct exception_table_entry _start___ex_table[];
+extern const struct exception_table_entry _stop___ex_table[];
+
+static inline unsigned long
+search_one_table(const struct exception_table_entry *first,
+		 const struct exception_table_entry *last,
+		 unsigned long value)
+{
+        while (first <= last) {
+		const struct exception_table_entry *mid;
+		long diff;
+
+		mid = (last - first) / 2 + first;
+		diff = mid->insn - value;
+                if (diff == 0)
+                        return mid->fixup;
+                else if (diff < 0)
+                        first = mid+1;
+                else
+                        last = mid-1;
+        }
+        return 0;
+}
+
+unsigned long
+search_exception_table(unsigned long addr)
+{
+	unsigned long ret;
+
+#ifndef CONFIG_MODULES
+	/* There is only the kernel to search.  */
+	ret = search_one_table(_start___ex_table, _stop___ex_table-1, addr);
+	if (ret) return ret;
+#else
+	/* The kernel is the last "module" -- no need to treat it special.  */
+	struct module *mp;
+	for (mp = module_list; mp != NULL; mp = mp->next) {
+		if (mp->ex_table_start == NULL)
+			continue;
+		ret = search_one_table(mp->ex_table_start,
+				       mp->ex_table_end - 1, addr);
+		if (ret) return ret;
+	}
+#endif
+
+	return 0;
+}
diff --git a/arch/cris/mm/fault.c b/arch/cris/mm/fault.c
new file mode 100644
index 000000000..a4bb237b4
--- /dev/null
+++ b/arch/cris/mm/fault.c
@@ -0,0 +1,390 @@
+/*
+ *  linux/arch/cris/mm/fault.c
+ *
+ *  Copyright (C) 2000  Axis Communications AB
+ *
+ *  Authors:  Bjorn Wesen 
+ * 
+ *  $Log: fault.c,v $
+ *  Revision 1.8  2000/11/22 14:45:31  bjornw
+ *  * 2.4.0-test10 removed the set_pgdir instantaneous kernel global mapping
+ *    into all processes. Instead we fill in the missing PTE entries on demand.
+ *
+ *  Revision 1.7  2000/11/21 16:39:09  bjornw
+ *  fixup switches frametype
+ *
+ *  Revision 1.6  2000/11/17 16:54:08  bjornw
+ *  More detailed siginfo reporting
+ *
+ *
+ */
+
+#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/interrupt.h>
+
+#include <asm/system.h>
+#include <asm/segment.h>
+#include <asm/pgtable.h>
+#include <asm/uaccess.h>
+#include <asm/svinto.h>
+
+extern void die_if_kernel(const char *,struct pt_regs *,long);
+
+asmlinkage void do_invalid_op (struct pt_regs *, unsigned long);
+asmlinkage void do_page_fault(unsigned long address, struct pt_regs *regs,
+			      int error_code);
+
+/* debug of low-level TLB reload */
+#define D(x)
+/* debug of higher-level faults */
+#define DPG(x)
+
+/* fast TLB-fill fault handler */
+
+void
+handle_mmu_bus_fault(struct pt_regs *regs)
+{
+	int cause, select;
+	int index;
+	int page_id;
+	int miss, we, acc, inv;  
+	struct mm_struct *mm = current->active_mm;
+	pmd_t *pmd;
+	pte_t pte;
+	int errcode = 0;
+	unsigned long address;
+
+	cause = *R_MMU_CAUSE;
+	select = *R_TLB_SELECT;
+
+	address = cause & PAGE_MASK; /* get faulting address */
+	
+	page_id = IO_EXTRACT(R_MMU_CAUSE,  page_id,   cause);
+	miss    = IO_EXTRACT(R_MMU_CAUSE,  miss_excp, cause);
+	we      = IO_EXTRACT(R_MMU_CAUSE,  we_excp,   cause);
+	acc     = IO_EXTRACT(R_MMU_CAUSE,  acc_excp,  cause);
+	inv     = IO_EXTRACT(R_MMU_CAUSE,  inv_excp,  cause);  
+	index  =  IO_EXTRACT(R_TLB_SELECT, index,     select);
+	
+	D(printk("bus_fault from IRP 0x%x: addr 0x%x, miss %d, inv %d, we %d, acc %d, "
+		 "idx %d pid %d\n",
+		 regs->irp, address, miss, inv, we, acc, index, page_id));
+
+	/* for a miss, we need to reload the TLB entry */
+
+	if(miss) {
+
+		/* see if the pte exists at all */
+		
+		pmd = (pmd_t *)pgd_offset(mm, address);
+		if(pmd_none(*pmd))
+			goto dofault;
+		if(pmd_bad(*pmd)) {
+			printk("bad pgdir entry 0x%x at 0x%x\n", *pmd, pmd);
+			pmd_clear(pmd);
+			return;
+		}
+		pte = *pte_offset(pmd, address);
+		if(!pte_present(pte))
+			goto dofault;
+		
+		D(printk(" found pte %x pg %x ", pte_val(pte), pte_page(pte)));
+		D(
+		  {
+			  if(pte_val(pte) & _PAGE_SILENT_WRITE)
+				  printk("Silent-W ");
+			  if(pte_val(pte) & _PAGE_KERNEL)
+				  printk("Kernel ");
+			  if(pte_val(pte) & _PAGE_SILENT_READ)
+				  printk("Silent-R ");
+			  if(pte_val(pte) & _PAGE_GLOBAL)
+				  printk("Global ");
+			  if(pte_val(pte) & _PAGE_PRESENT)
+				  printk("Present ");
+			  if(pte_val(pte) & _PAGE_ACCESSED)
+				  printk("Accessed ");
+			  if(pte_val(pte) & _PAGE_MODIFIED)
+				  printk("Modified ");
+			  if(pte_val(pte) & _PAGE_READ)
+				  printk("Readable ");
+			  if(pte_val(pte) & _PAGE_WRITE)
+				  printk("Writeable ");
+			  printk("\n");
+		  });
+
+		/* load up the chosen TLB entry
+		 * this assumes the pte format is the same as the TLB_LO layout.
+		 *
+		 * the write to R_TLB_LO also writes the vpn and page_id fields from
+		 * R_MMU_CAUSE, which we in this case obviously want to keep
+		 */
+		
+		*R_TLB_LO = pte_val(pte);
+
+		return;
+	} 
+
+	errcode = 0x01 | (we << 1);
+
+ dofault:
+	/* leave it to the MM system fault handler below */
+	D(printk("do_page_fault %p errcode %d\n", address, errcode));
+	do_page_fault(address, regs, errcode);
+}
+
+/*
+ * This routine handles page faults.  It determines the address,
+ * and the problem, and then passes it off to one of the appropriate
+ * routines.
+ *
+ * Notice that the address we're given is aligned to the page the fault
+ * occured in, since we only get the PFN in R_MMU_CAUSE not the complete
+ * address.
+ *
+ * error_code:
+ *	bit 0 == 0 means no page found, 1 means protection fault
+ *	bit 1 == 0 means read, 1 means write
+ *
+ * If this routine detects a bad access, it returns 1, otherwise it
+ * returns 0.
+ */
+
+asmlinkage void
+do_page_fault(unsigned long address, struct pt_regs *regs,
+	      int error_code)
+{
+	struct task_struct *tsk;
+	struct mm_struct *mm;
+	struct vm_area_struct * vma;
+	int writeaccess;
+	int fault;
+	unsigned long fixup;
+	siginfo_t info;
+
+	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.
+	 *
+	 * TODO: On CRIS, we have a PTE Global bit which should be set in
+	 * all the PTE's related to vmalloc in all processes - that means if
+	 * we switch process and a vmalloc PTE is still in the TLB, it won't
+	 * need to be reloaded. It's an optimization.
+	 *
+	 * Linux/CRIS's kernel is not page-mapped, so the comparision below
+	 * should really be >= VMALLOC_START, however, kernel fixup errors
+	 * will be handled more quickly by going through vmalloc_fault and then
+	 * into bad_area_nosemaphore than falling through the find_vma user-mode
+	 * tests.
+         */
+
+        if (address >= TASK_SIZE)
+                goto vmalloc_fault;
+
+	mm = tsk->mm;
+	writeaccess = error_code & 2;
+	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(&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 < rdusp())
+			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 (writeaccess) {
+		if (!(vma->vm_flags & VM_WRITE))
+			goto bad_area;
+	} else {
+		if (!(vma->vm_flags & (VM_READ | VM_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.
+	 */
+
+	switch (handle_mm_fault(mm, vma, address, writeaccess)) {
+        case 1:
+                tsk->min_flt++;
+                break;
+        case 2:
+                tsk->maj_flt++;
+                break;
+        case 0:
+                goto do_sigbus;
+        default:
+                goto out_of_memory;
+	}
+
+	up(&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(&mm->mmap_sem);
+
+ bad_area_nosemaphore:
+	DPG(show_registers(regs));
+
+	/* 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)
+	 */
+
+        if ((fixup = search_exception_table(regs->irp)) != 0) {
+                regs->irp = fixup;
+		regs->frametype = CRIS_FRAME_FIXUP;
+		D(printk("doing fixup to 0x%x\n", 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 %08lx\n",address);
+
+	die_if_kernel("Oops", regs, error_code);
+
+	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:
+        up(&mm->mmap_sem);
+	printk("VM: killing process %s\n", tsk->comm);
+	if(user_mode(regs))
+		do_exit(SIGKILL);
+	goto no_context;
+
+ do_sigbus:
+	up(&mm->mmap_sem);
+
+	/*
+         * Send a sigbus, regardless of whether we were in kernel
+         * or user mode.
+         */
+	info.si_code = 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.
+                 */
+                int offset = pgd_index(address);
+                pgd_t *pgd, *pgd_k;
+                pmd_t *pmd, *pmd_k;
+
+                pgd = tsk->active_mm->pgd + offset;
+                pgd_k = init_mm.pgd + offset;
+
+                if (!pgd_present(*pgd)) {
+                        if (!pgd_present(*pgd_k))
+                                goto bad_area_nosemaphore;
+                        set_pgd(pgd, *pgd_k);
+                        return;
+                }
+
+                pmd = pmd_offset(pgd, address);
+                pmd_k = pmd_offset(pgd_k, address);
+
+                if (pmd_present(*pmd) || !pmd_present(*pmd_k))
+                        goto bad_area_nosemaphore;
+                set_pmd(pmd, *pmd_k);
+                return;
+        }
+
+}
diff --git a/arch/cris/mm/init.c b/arch/cris/mm/init.c
new file mode 100644
index 000000000..3d0ceeffb
--- /dev/null
+++ b/arch/cris/mm/init.c
@@ -0,0 +1,506 @@
+/*
+ *  linux/arch/cris/mm/init.c
+ *
+ *  Copyright (C) 1995  Linus Torvalds
+ *  Copyright (C) 2000  Axis Communications AB
+ *
+ *  Authors:  Bjorn Wesen (bjornw@axis.com)
+ *
+ *  $Log: init.c,v $
+ *  Revision 1.15  2001/01/10 21:12:10  bjornw
+ *  loops_per_sec -> loops_per_jiffy
+ *
+ *  Revision 1.14  2000/11/22 16:23:20  bjornw
+ *  Initialize totalhigh counters to 0 to make /proc/meminfo look nice.
+ *
+ *  Revision 1.13  2000/11/21 16:37:51  bjornw
+ *  Temporarily disable initmem freeing
+ *
+ *  Revision 1.12  2000/11/21 13:55:07  bjornw
+ *  Use CONFIG_CRIS_LOW_MAP for the low VM map instead of explicit CPU type
+ *
+ *  Revision 1.11  2000/10/06 12:38:22  bjornw
+ *  Cast empty_bad_page correctly (should really be of * type from the start..
+ *
+ *  Revision 1.10  2000/10/04 16:53:57  bjornw
+ *  Fix memory-map due to LX features
+ *
+ *  Revision 1.9  2000/09/13 15:47:49  bjornw
+ *  Wrong count in reserved-pages loop
+ *
+ *  Revision 1.8  2000/09/13 14:35:10  bjornw
+ *  2.4.0-test8 added a new arg to free_area_init_node
+ *
+ *  Revision 1.7  2000/08/17 15:35:55  bjornw
+ *  2.4.0-test6 removed MAP_NR and inserted virt_to_page
+ *
+ *
+ */
+
+#include <linux/config.h>
+#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 <asm/system.h>
+#include <asm/segment.h>
+#include <asm/pgalloc.h>
+#include <asm/pgtable.h>
+#include <asm/dma.h>
+#include <asm/svinto.h>
+
+static unsigned long totalram_pages;
+
+struct pgtable_cache_struct quicklists;  /* see asm/pgalloc.h */
+
+const char bad_pmd_string[] = "Bad pmd in pte_alloc: %08lx\n";
+
+extern void die_if_kernel(char *,struct pt_regs *,long);
+extern void show_net_buffers(void);
+extern void tlb_init(void);
+
+/*
+ * empty_bad_page is the page that is used for page faults when linux
+ * is out-of-memory. Older versions of linux just did a
+ * do_exit(), but using this instead means there is less risk
+ * for a process dying in kernel mode, possibly leaving a inode
+ * unused etc..
+ *
+ * the main point is that when a page table error occurs, we want to get
+ * out of the kernel safely before killing the process, so we need something
+ * to feed the MMU with when the fault occurs even if we don't have any
+ * real PTE's or page tables.
+ *
+ * empty_bad_page_table is the accompanying page-table: it is initialized
+ * to point to empty_bad_page writable-shared entries.
+ *
+ * empty_zero_page is a special page that is used for zero-initialized
+ * data and COW.
+ */
+
+unsigned long empty_bad_page_table;
+unsigned long empty_bad_page;
+unsigned long empty_zero_page;
+
+pte_t * __bad_pagetable(void)
+{
+	/* somehow it is enough to just clear it and not fill it with
+	 * bad page PTE's...
+	 */
+	memset((void *)empty_bad_page_table, 0, PAGE_SIZE);
+
+	return (pte_t *) empty_bad_page_table;
+}
+
+pte_t __bad_page(void)
+{
+
+	/* clear the empty_bad_page page. this should perhaps be
+	 * a more simple inlined loop like it is on the other 
+	 * architectures.
+	 */
+
+	memset((void *)empty_bad_page, 0, PAGE_SIZE);
+
+	return pte_mkdirty(__mk_pte((void *)empty_bad_page, PAGE_SHARED));
+}
+
+static pte_t * get_bad_pte_table(void)
+{
+	pte_t *empty_bad_pte_table = (pte_t *)empty_bad_page_table;
+	pte_t v;
+	int i;
+
+	v = __bad_page();
+
+	for (i = 0; i < PAGE_SIZE/sizeof(pte_t); i++)
+		empty_bad_pte_table[i] = v;
+
+	return empty_bad_pte_table;
+}
+
+void __handle_bad_pmd(pmd_t *pmd)
+{
+	pmd_ERROR(*pmd);
+	pmd_set(pmd, get_bad_pte_table());
+}
+
+void __handle_bad_pmd_kernel(pmd_t *pmd)
+{
+	pmd_ERROR(*pmd);
+	pmd_set_kernel(pmd, get_bad_pte_table());
+}
+
+pte_t *get_pte_kernel_slow(pmd_t *pmd, unsigned long offset)
+{
+        pte_t *pte;
+
+        pte = (pte_t *) __get_free_page(GFP_KERNEL);
+        if (pmd_none(*pmd)) {
+                if (pte) {
+                        clear_page(pte);
+			pmd_set_kernel(pmd, pte);
+                        return pte + offset;
+                }
+		pmd_set_kernel(pmd, get_bad_pte_table());
+                return NULL;
+        }
+        free_page((unsigned long)pte);
+        if (pmd_bad(*pmd)) {
+                __handle_bad_pmd_kernel(pmd);
+                return NULL;
+        }
+        return (pte_t *) pmd_page(*pmd) + offset;
+}
+
+pte_t *get_pte_slow(pmd_t *pmd, unsigned long offset)
+{
+	pte_t *pte;
+
+        pte = (pte_t *) __get_free_page(GFP_KERNEL);
+        if (pmd_none(*pmd)) {
+                if (pte) {
+                        clear_page(pte);
+			pmd_set(pmd, pte);
+                        return pte + offset;
+                }
+		pmd_set(pmd, get_bad_pte_table());
+                return NULL;
+        }
+        free_page((unsigned long)pte);
+        if (pmd_bad(*pmd)) {
+                __handle_bad_pmd(pmd);
+                return NULL;
+        }
+        return (pte_t *) pmd_page(*pmd) + offset;
+}
+
+#ifndef CONFIG_NO_PGT_CACHE
+struct pgtable_cache_struct quicklists;
+
+/* trim the page-table cache if necessary */
+
+int do_check_pgt_cache(int low, int high)
+{
+        int freed = 0;
+
+        if(pgtable_cache_size > high) {
+                do {
+                        if(pgd_quicklist) {
+                                free_pgd_slow(get_pgd_fast());
+                                freed++;
+                        }
+                        if(pmd_quicklist) {
+                                free_pmd_slow(get_pmd_fast());
+                                freed++;
+                        }
+                        if(pte_quicklist) {
+                                free_pte_slow(get_pte_fast());
+                                 freed++;
+                        }
+                } while(pgtable_cache_size > low);
+        }
+        return freed;
+}
+#else
+int do_check_pgt_cache(int low, int high)
+{
+        return 0;
+}
+#endif
+
+void show_mem(void)
+{
+	int i,free = 0,total = 0,cached = 0, reserved = 0, nonshared = 0;
+	int shared = 0;
+
+	printk("\nMem-info:\n");
+	show_free_areas();
+	printk("Free swap:       %6dkB\n",nr_swap_pages<<(PAGE_SHIFT-10));
+	i = max_mapnr;
+	while (i-- > 0) {
+		total++;
+		if (PageReserved(mem_map+i))
+			reserved++;
+		else if (PageSwapCache(mem_map+i))
+			cached++;
+		else if (!page_count(mem_map+i))
+			free++;
+		else if (page_count(mem_map+i) == 1)
+			nonshared++;
+		else
+			shared += page_count(mem_map+i) - 1;
+	}
+	printk("%d pages of RAM\n",total);
+	printk("%d free pages\n",free);
+	printk("%d reserved pages\n",reserved);
+	printk("%d pages nonshared\n",nonshared);
+	printk("%d pages shared\n",shared);
+	printk("%d pages swap cached\n",cached);
+	printk("%ld pages in page table cache\n",pgtable_cache_size);
+	show_buffers();
+}
+
+/*
+ * The kernel is already mapped with a kernel segment at kseg_c so 
+ * we don't need to map it with a page table. However head.S also
+ * temporarily mapped it at kseg_4 so we should set up the ksegs again,
+ * clear the TLB and do some other paging setup stuff.
+ */
+
+void __init 
+paging_init(void)
+{
+	int i;
+	unsigned long zones_size[MAX_NR_ZONES];
+
+	printk("Setting up paging and the MMU.\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);
+
+	/* initialise the TLB (tlb.c) */
+
+	tlb_init();
+
+	/* see README.mm for details on the KSEG setup */
+
+#ifdef CONFIG_CRIS_LOW_MAP
+
+	/* Etrax-100 LX version 1 has a bug so that we cannot map anything
+	 * across the 0x80000000 boundary, so we need to shrink the user-virtual
+	 * area to 0x50000000 instead of 0xb0000000 and map things slightly
+	 * different. The unused areas are marked as paged so that we can catch
+	 * freak kernel accesses there.
+	 *
+	 * The Juliette chip is mapped at 0xa so we pass that segment straight
+	 * through. We cannot vremap it because the vmalloc area is below 0x8
+	 * and Juliette needs an uncached area above 0x8.
+	 */
+
+	*R_MMU_KSEG = ( IO_STATE(R_MMU_KSEG, seg_f, page ) | 
+			IO_STATE(R_MMU_KSEG, seg_e, seg  ) |  /* uncached flash */
+			IO_STATE(R_MMU_KSEG, seg_d, page ) | 
+			IO_STATE(R_MMU_KSEG, seg_c, page ) | 
+			IO_STATE(R_MMU_KSEG, seg_b, seg  ) |  /* kernel reg area */
+			IO_STATE(R_MMU_KSEG, seg_a, seg  ) |  /* Juliette etc. */
+			IO_STATE(R_MMU_KSEG, seg_9, page ) |
+			IO_STATE(R_MMU_KSEG, seg_8, page ) |
+			IO_STATE(R_MMU_KSEG, seg_7, page ) |  /* kernel vmalloc area */
+			IO_STATE(R_MMU_KSEG, seg_6, seg  ) |  /* kernel DRAM area */
+			IO_STATE(R_MMU_KSEG, seg_5, seg  ) |  /* cached flash */
+			IO_STATE(R_MMU_KSEG, seg_4, page ) |  /* user area */
+			IO_STATE(R_MMU_KSEG, seg_3, page ) |  /* user area */
+			IO_STATE(R_MMU_KSEG, seg_2, page ) |  /* user area */
+			IO_STATE(R_MMU_KSEG, seg_1, page ) |  /* user area */
+			IO_STATE(R_MMU_KSEG, seg_0, page ) ); /* user area */
+
+	*R_MMU_KBASE_HI = ( IO_FIELD(R_MMU_KBASE_HI, base_f, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_e, 0x8 ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_d, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_c, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_b, 0xb ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_a, 0xa ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_9, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_8, 0x0 ) );
+	
+	*R_MMU_KBASE_LO = ( IO_FIELD(R_MMU_KBASE_LO, base_7, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_6, 0x4 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_5, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_4, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_3, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_2, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_1, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_0, 0x0 ) );
+#else
+	/* This code is for the hopefully corrected Etrax-100 LX version 2... */
+
+	*R_MMU_KSEG = ( IO_STATE(R_MMU_KSEG, seg_f, seg  ) | /* cached flash */
+			IO_STATE(R_MMU_KSEG, seg_e, seg  ) | /* uncached flash */
+			IO_STATE(R_MMU_KSEG, seg_d, page ) | /* vmalloc area */
+			IO_STATE(R_MMU_KSEG, seg_c, seg  ) | /* kernel area */
+			IO_STATE(R_MMU_KSEG, seg_b, seg  ) | /* kernel reg area */
+			IO_STATE(R_MMU_KSEG, seg_a, page ) | /* user area */
+			IO_STATE(R_MMU_KSEG, seg_9, page ) |
+			IO_STATE(R_MMU_KSEG, seg_8, page ) |
+			IO_STATE(R_MMU_KSEG, seg_7, page ) |
+			IO_STATE(R_MMU_KSEG, seg_6, page ) |
+			IO_STATE(R_MMU_KSEG, seg_5, page ) |
+			IO_STATE(R_MMU_KSEG, seg_4, page ) |
+			IO_STATE(R_MMU_KSEG, seg_3, page ) |
+			IO_STATE(R_MMU_KSEG, seg_2, page ) |
+			IO_STATE(R_MMU_KSEG, seg_1, page ) |
+			IO_STATE(R_MMU_KSEG, seg_0, page ) );
+
+	*R_MMU_KBASE_HI = ( IO_FIELD(R_MMU_KBASE_HI, base_f, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_e, 0x8 ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_d, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_c, 0x4 ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_b, 0xb ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_a, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_9, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_HI, base_8, 0x0 ) );
+	
+	*R_MMU_KBASE_LO = ( IO_FIELD(R_MMU_KBASE_LO, base_7, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_6, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_5, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_4, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_3, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_2, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_1, 0x0 ) |
+			    IO_FIELD(R_MMU_KBASE_LO, base_0, 0x0 ) );
+#endif	
+
+	*R_MMU_CONTEXT = ( IO_FIELD(R_MMU_CONTEXT, page_id, 0 ) );
+	
+	/* The MMU has been enabled ever since head.S but just to make
+	 * it totally obvious we do it here as well.
+	 */
+
+	*R_MMU_CTRL = ( IO_STATE(R_MMU_CTRL, inv_excp, enable ) |
+			IO_STATE(R_MMU_CTRL, acc_excp, enable ) |
+			IO_STATE(R_MMU_CTRL, we_excp,  enable ) );
+	
+	*R_MMU_ENABLE = IO_STATE(R_MMU_ENABLE, mmu_enable, enable);
+
+	/*
+	 * initialize the bad page table and bad page to point
+	 * to a couple of allocated pages
+	 */
+
+	empty_bad_page_table = (unsigned long)alloc_bootmem_pages(PAGE_SIZE);
+	empty_bad_page = (unsigned long)alloc_bootmem_pages(PAGE_SIZE);
+	empty_zero_page = (unsigned long)alloc_bootmem_pages(PAGE_SIZE);
+	memset((void *)empty_zero_page, 0, PAGE_SIZE);
+
+	/* All pages are DMA'able in Etrax, so put all in the DMA'able zone */
+
+	zones_size[0] = ((unsigned long)high_memory - PAGE_OFFSET) >> PAGE_SHIFT;
+
+	for (i = 1; i < MAX_NR_ZONES; i++)
+		zones_size[i] = 0;
+
+	/* Use free_area_init_node instead of free_area_init, because the former
+	 * is designed for systems where the DRAM starts at an address substantially
+	 * higher than 0, like us (we start at PAGE_OFFSET). This saves space in the
+	 * mem_map page array.
+	 */
+
+	free_area_init_node(0, 0, 0, zones_size, PAGE_OFFSET, 0);
+}
+
+extern unsigned long loops_per_jiffy; /* init/main.c */
+unsigned long loops_per_usec;
+
+extern char _stext, _edata, _etext;
+extern char __init_begin, __init_end;
+
+void __init
+mem_init(void)
+{
+	int codesize, reservedpages, datasize, initsize;
+	unsigned long tmp;
+
+	if(!mem_map)
+		BUG();
+
+	/* max/min_low_pfn was set by setup.c
+	 * now we just copy it to some other necessary places...
+	 *
+	 * high_memory was also set in setup.c
+	 */
+
+	max_mapnr = num_physpages = max_low_pfn - min_low_pfn;
+ 
+	/* this will put all memory onto the freelists */
+        totalram_pages = free_all_bootmem();
+
+	reservedpages = 0;
+	for (tmp = 0; tmp < max_mapnr; tmp++) {
+		/*
+                 * Only count reserved RAM pages
+                 */
+		if (PageReserved(mem_map + tmp))
+			reservedpages++;
+	}
+
+	codesize =  (unsigned long) &_etext - (unsigned long) &_stext;
+        datasize =  (unsigned long) &_edata - (unsigned long) &_etext;
+        initsize =  (unsigned long) &__init_end - (unsigned long) &__init_begin;
+	
+        printk("Memory: %luk/%luk available (%dk kernel code, %dk reserved, %dk data, "
+	       "%dk init)\n" ,
+	       (unsigned long) nr_free_pages() << (PAGE_SHIFT-10),
+	       max_mapnr << (PAGE_SHIFT-10),
+	       codesize >> 10,
+	       reservedpages << (PAGE_SHIFT-10),
+	       datasize >> 10,
+	       initsize >> 10
+               );
+	
+	/* HACK alert - calculate a loops_per_usec for asm/delay.h here
+	 * since this is called just after calibrate_delay in init/main.c
+	 * but before places which use udelay. cannot be in time.c since
+	 * that is called _before_ calibrate_delay
+	 */
+
+	loops_per_usec = (loops_per_jiffy * HZ) / 1000000;
+
+	return;
+}
+
+/* free the pages occupied by initialization code */
+
+void free_initmem(void)
+{
+#if 0
+	/* currently this is a bad idea since the cramfs image is catted onto
+	 * the vmlinux image, and the end of that image is not page-padded so
+	 * part of the cramfs image will be freed here
+	 */
+        unsigned long addr;
+
+        addr = (unsigned long)(&__init_begin);
+        for (; addr < (unsigned long)(&__init_end); addr += PAGE_SIZE) {
+                ClearPageReserved(virt_to_page(addr));
+                set_page_count(virt_to_page(addr), 1);
+                free_page(addr);
+                totalram_pages++;
+        }
+        printk ("Freeing unused kernel memory: %dk freed\n", 
+		(&__init_end - &__init_begin) >> 10);
+#endif
+}
+
+void si_meminfo(struct sysinfo *val)
+{
+	int i;
+
+	i = max_mapnr;
+	val->totalram = 0;
+	val->sharedram = 0;
+	val->freeram = nr_free_pages();
+	val->bufferram = atomic_read(&buffermem_pages);
+	while (i-- > 0)  {
+		if (PageReserved(mem_map+i))
+			continue;
+		val->totalram++;
+		if (!atomic_read(&mem_map[i].count))
+			continue;
+		val->sharedram += atomic_read(&mem_map[i].count) - 1;
+	}
+	val->mem_unit = PAGE_SIZE;
+	val->totalhigh = 0;
+	val->freehigh = 0;
+}
diff --git a/arch/cris/mm/tlb.c b/arch/cris/mm/tlb.c
new file mode 100644
index 000000000..ed74305ae
--- /dev/null
+++ b/arch/cris/mm/tlb.c
@@ -0,0 +1,301 @@
+/*
+ *  linux/arch/cris/mm/tlb.c
+ *
+ *  Copyright (C) 2000  Axis Communications AB
+ *  
+ *  Authors:   Bjorn Wesen (bjornw@axis.com)
+ *
+ */
+
+#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/pgtable.h>
+#include <asm/svinto.h>
+
+#define D(x)
+
+/* CRIS in Etrax100LX TLB */
+
+#define NUM_TLB_ENTRIES 64
+#define NUM_PAGEID 64
+#define INVALID_PAGEID 63
+#define NO_CONTEXT -1
+
+/* The TLB can host up to 64 different mm contexts at the same time.
+ * The running context is R_MMU_CONTEXT, and each TLB entry contains a
+ * page_id that has to match to give a hit. In page_id_map, we keep track
+ * of which mm's we have assigned which page_id's, so that we know when
+ * to invalidate TLB entries.
+ *
+ * The last page_id is never running - it is used as an invalid page_id
+ * so we can make TLB entries that will never match.
+ */
+
+struct mm_struct *page_id_map[NUM_PAGEID];
+
+static int map_replace_ptr = 1;  /* which page_id_map entry to replace next */
+
+/* invalidate all TLB entries */
+
+void
+flush_tlb_all()
+{
+	int i;
+
+	/* the vpn of i & 0xf is so we dont write similar TLB entries
+	 * in the same 4-way entry group. details.. 
+	 */
+
+	for(i = 0; i < NUM_TLB_ENTRIES; i++) {
+		*R_TLB_SELECT = ( IO_FIELD(R_TLB_SELECT, index, i) );
+
+		*R_TLB_HI = ( IO_FIELD(R_TLB_HI, page_id, INVALID_PAGEID ) |
+			      IO_FIELD(R_TLB_HI, vpn,     i & 0xf ) );
+		
+		*R_TLB_LO = ( IO_STATE(R_TLB_LO, global,no       ) |
+			      IO_STATE(R_TLB_LO, valid, no       ) |
+			      IO_STATE(R_TLB_LO, kernel,no	 ) |
+			      IO_STATE(R_TLB_LO, we,    no       ) |
+			      IO_FIELD(R_TLB_LO, pfn,   0        ) );
+	}
+	D(printk("tlb: flushed all\n"));
+}
+
+/* invalidate the selected mm context only */
+
+void
+flush_tlb_mm(struct mm_struct *mm)
+{
+	int i;
+	int page_id = mm->context;
+
+	D(printk("tlb: flush mm context %d (%p)\n", page_id, mm));
+
+	if(page_id == NO_CONTEXT)
+		return;
+	
+	/* mark the TLB entries that match the page_id as invalid.
+	 * here we could also check the _PAGE_GLOBAL bit and NOT flush
+	 * global pages. is it worth the extra I/O ? 
+	 */
+
+	for(i = 0; i < NUM_TLB_ENTRIES; i++) {
+		*R_TLB_SELECT = IO_FIELD(R_TLB_SELECT, index, i);
+		if (IO_EXTRACT(R_TLB_HI, page_id, *R_TLB_HI) == page_id) {
+			*R_TLB_HI = ( IO_FIELD(R_TLB_HI, page_id, INVALID_PAGEID ) |
+				      IO_FIELD(R_TLB_HI, vpn,     i & 0xf ) );
+			
+			*R_TLB_LO = ( IO_STATE(R_TLB_LO, global,no  ) |
+				      IO_STATE(R_TLB_LO, valid, no  ) |
+				      IO_STATE(R_TLB_LO, kernel,no  ) |
+				      IO_STATE(R_TLB_LO, we,    no  ) |
+				      IO_FIELD(R_TLB_LO, pfn,   0   ) );
+		}
+	}
+}
+
+/* invalidate a single page */
+
+void
+flush_tlb_page(struct vm_area_struct *vma, 
+	       unsigned long addr)
+{
+	struct mm_struct *mm = vma->vm_mm;
+	int page_id = mm->context;
+	int i;
+
+	D(printk("tlb: flush page %p in context %d (%p)\n", addr, page_id, mm));
+
+	if(page_id == NO_CONTEXT)
+		return;
+
+	addr &= PAGE_MASK; /* perhaps not necessary */
+
+	/* invalidate those TLB entries that match both the mm context
+	 * and the virtual address requested 
+	 */
+
+	for(i = 0; i < NUM_TLB_ENTRIES; i++) {
+		unsigned long tlb_hi;
+		*R_TLB_SELECT = IO_FIELD(R_TLB_SELECT, index, i);
+		tlb_hi = *R_TLB_HI;
+		if (IO_EXTRACT(R_TLB_HI, page_id, tlb_hi) == page_id &&
+		    (tlb_hi & PAGE_MASK) == addr) {
+			*R_TLB_HI = IO_FIELD(R_TLB_HI, page_id, INVALID_PAGEID ) |
+				addr; /* same addr as before works. */
+			
+			*R_TLB_LO = ( IO_STATE(R_TLB_LO, global,no  ) |
+				      IO_STATE(R_TLB_LO, valid, no  ) |
+				      IO_STATE(R_TLB_LO, kernel,no  ) |
+				      IO_STATE(R_TLB_LO, we,    no  ) |
+				      IO_FIELD(R_TLB_LO, pfn,   0   ) );
+		}
+	}
+}
+
+/* invalidate a page range */
+
+void
+flush_tlb_range(struct mm_struct *mm, 
+		unsigned long start,
+		unsigned long end)
+{
+	int page_id = mm->context;
+	int i;
+
+	D(printk("tlb: flush range %p<->%p in context %d (%p)\n",
+		 start, end, page_id, mm));
+
+	if(page_id == NO_CONTEXT)
+		return;
+
+	start &= PAGE_MASK;  /* probably not necessary */
+	end &= PAGE_MASK;    /* dito */
+
+	/* invalidate those TLB entries that match both the mm context
+	 * and the virtual address range
+	 */
+
+	for(i = 0; i < NUM_TLB_ENTRIES; i++) {
+		unsigned long tlb_hi, vpn;
+		*R_TLB_SELECT = IO_FIELD(R_TLB_SELECT, index, i);
+		tlb_hi = *R_TLB_HI;
+		vpn = tlb_hi & PAGE_MASK;
+		if (IO_EXTRACT(R_TLB_HI, page_id, tlb_hi) == page_id &&
+		    vpn >= start && vpn < end) {
+			*R_TLB_HI = ( IO_FIELD(R_TLB_HI, page_id, INVALID_PAGEID ) |
+				      IO_FIELD(R_TLB_HI, vpn,     i & 0xf ) );
+			
+			*R_TLB_LO = ( IO_STATE(R_TLB_LO, global,no  ) |
+				      IO_STATE(R_TLB_LO, valid, no  ) |
+				      IO_STATE(R_TLB_LO, kernel,no  ) |
+				      IO_STATE(R_TLB_LO, we,    no  ) |
+				      IO_FIELD(R_TLB_LO, pfn,   0   ) );
+		}
+	}
+}
+
+/*
+ * 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;
+}
+
+/* the following functions are similar to those used in the PPC port */
+
+static inline void
+alloc_context(struct mm_struct *mm)
+{
+	struct mm_struct *old_mm;
+
+	D(printk("tlb: alloc context %d (%p)\n", map_replace_ptr, mm));
+
+	/* did we replace an mm ? */
+
+	old_mm = page_id_map[map_replace_ptr];
+
+	if(old_mm) {
+		/* throw out any TLB entries belonging to the mm we replace
+		 * in the map
+		 */
+		flush_tlb_mm(old_mm);
+
+		old_mm->context = NO_CONTEXT;
+	}
+
+	/* insert it into the page_id_map */
+
+	mm->context = map_replace_ptr;
+	page_id_map[map_replace_ptr] = mm;
+
+	map_replace_ptr++;
+
+	if(map_replace_ptr == INVALID_PAGEID)
+		map_replace_ptr = 0;         /* wrap around */
+	
+}
+
+/* 
+ * if needed, get a new MMU context for the mm. otherwise nothing is done.
+ */
+
+void
+get_mmu_context(struct mm_struct *mm)
+{
+	if(mm->context == NO_CONTEXT)
+		alloc_context(mm);
+}
+
+/* called in schedule() just before actually doing the switch_to */
+
+void 
+switch_mm(struct mm_struct *prev, struct mm_struct *next,
+	  struct task_struct *tsk, int cpu)
+{
+	/* make sure we have a context */
+
+	get_mmu_context(next);
+
+	/* switch context in the MMU */
+	
+	D(printk("switching mmu_context to %d (%p)\n", next->context, next));
+
+	*R_MMU_CONTEXT = IO_FIELD(R_MMU_CONTEXT, page_id, next->context);
+}
+
+
+/* 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.
+ *
+ * the only thing we really need to do here is mark the used PID slot
+ * as empty.
+ */
+
+void
+destroy_context(struct mm_struct *mm)
+{
+	if(mm->context != NO_CONTEXT) {
+		D(printk("destroy_context %d (%p)\n", mm->context, mm));
+		flush_tlb_mm(mm);  /* TODO this might be redundant ? */
+		page_id_map[mm->context] = NULL;
+		/* mm->context = NO_CONTEXT; redundant.. mm will be freed */
+	}
+}
+
+/* called once during VM initialization, from init.c */
+
+void __init
+tlb_init(void)
+{
+	int i;
+
+	/* clear the page_id map */
+
+	for(i = 0; i < 64; i++)
+		page_id_map[i] = NULL;
+	
+	/* invalidate the entire TLB */
+
+	flush_tlb_all();
+
+	/* the init_mm has context 0 from the boot */
+
+	page_id_map[0] = &init_mm;
+}