Merge git://git.kernel.org/pub/scm/linux/kernel/git/bunk/trivial
[safe/jmp/linux-2.6] / arch / arm / mm / mm-armv.c
index c3bd503..b0242c6 100644 (file)
@@ -1,7 +1,7 @@
 /*
  *  linux/arch/arm/mm/mm-armv.c
  *
- *  Copyright (C) 1998-2002 Russell King
+ *  Copyright (C) 1998-2005 Russell King
  *
  * 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
@@ -9,7 +9,6 @@
  *
  *  Page table sludge for ARM v3 and v4 processor architectures.
  */
-#include <linux/config.h>
 #include <linux/module.h>
 #include <linux/mm.h>
 #include <linux/init.h>
@@ -19,7 +18,6 @@
 
 #include <asm/pgalloc.h>
 #include <asm/page.h>
-#include <asm/io.h>
 #include <asm/setup.h>
 #include <asm/tlbflush.h>
 
@@ -180,11 +178,6 @@ pgd_t *get_pgd_slow(struct mm_struct *mm)
 
        if (!vectors_high()) {
                /*
-                * This lock is here just to satisfy pmd_alloc and pte_lock
-                */
-               spin_lock(&mm->page_table_lock);
-
-               /*
                 * On ARM, first page must always be allocated since it
                 * contains the machine vectors.
                 */
@@ -201,23 +194,14 @@ pgd_t *get_pgd_slow(struct mm_struct *mm)
                set_pte(new_pte, *init_pte);
                pte_unmap_nested(init_pte);
                pte_unmap(new_pte);
-
-               spin_unlock(&mm->page_table_lock);
        }
 
        return new_pgd;
 
 no_pte:
-       spin_unlock(&mm->page_table_lock);
        pmd_free(new_pmd);
-       free_pages((unsigned long)new_pgd, 2);
-       return NULL;
-
 no_pmd:
-       spin_unlock(&mm->page_table_lock);
        free_pages((unsigned long)new_pgd, 2);
-       return NULL;
-
 no_pgd:
        return NULL;
 }
@@ -242,7 +226,8 @@ void free_pgd_slow(pgd_t *pgd)
 
        pte = pmd_page(*pmd);
        pmd_clear(pmd);
-       dec_page_state(nr_page_table_pages);
+       dec_zone_page_state(virt_to_page((unsigned long *)pgd), NR_PAGETABLE);
+       pte_lock_deinit(pte);
        pte_free(pte);
        pmd_free(pmd);
 free:
@@ -275,11 +260,9 @@ alloc_init_supersection(unsigned long virt, unsigned long phys, int prot)
        int i;
 
        for (i = 0; i < 16; i += 1) {
-               alloc_init_section(virt, phys & SUPERSECTION_MASK,
-                                  prot | PMD_SECT_SUPER);
+               alloc_init_section(virt, phys, prot | PMD_SECT_SUPER);
 
                virt += (PGDIR_SIZE / 2);
-               phys += (PGDIR_SIZE / 2);
        }
 }
 
@@ -297,30 +280,16 @@ alloc_init_page(unsigned long virt, unsigned long phys, unsigned int prot_l1, pg
        pte_t *ptep;
 
        if (pmd_none(*pmdp)) {
-               unsigned long pmdval;
                ptep = alloc_bootmem_low_pages(2 * PTRS_PER_PTE *
                                               sizeof(pte_t));
 
-               pmdval = __pa(ptep) | prot_l1;
-               pmdp[0] = __pmd(pmdval);
-               pmdp[1] = __pmd(pmdval + 256 * sizeof(pte_t));
-               flush_pmd_entry(pmdp);
+               __pmd_populate(pmdp, __pa(ptep) | prot_l1);
        }
        ptep = pte_offset_kernel(pmdp, virt);
 
        set_pte(ptep, pfn_pte(phys >> PAGE_SHIFT, prot));
 }
 
-/*
- * Clear any PGD mapping.  On a two-level page table system,
- * the clearance is done by the middle-level functions (pmd)
- * rather than the top-level (pgd) functions.
- */
-static inline void clear_mapping(unsigned long virt)
-{
-       pmd_clear(pmd_off_k(virt));
-}
-
 struct mem_types {
        unsigned int    prot_pte;
        unsigned int    prot_l1;
@@ -373,16 +342,23 @@ static struct mem_types mem_types[] __initdata = {
                                PMD_SECT_AP_WRITE | PMD_SECT_BUFFERABLE |
                                PMD_SECT_TEX(1),
                .domain    = DOMAIN_IO,
+       },
+       [MT_NONSHARED_DEVICE] = {
+               .prot_l1   = PMD_TYPE_TABLE,
+               .prot_sect = PMD_TYPE_SECT | PMD_SECT_NONSHARED_DEV |
+                               PMD_SECT_AP_WRITE,
+               .domain    = DOMAIN_IO,
        }
 };
 
 /*
  * Adjust the PMD section entries according to the CPU in use.
  */
-static void __init build_mem_type_table(void)
+void __init build_mem_type_table(void)
 {
        struct cachepolicy *cp;
        unsigned int cr = get_cr();
+       unsigned int user_pgprot, kern_pgprot;
        int cpu_arch = cpu_architecture();
        int i;
 
@@ -399,7 +375,7 @@ static void __init build_mem_type_table(void)
                ecc_mask = 0;
        }
 
-       if (cpu_arch <= CPU_ARCH_ARMv5) {
+       if (cpu_arch <= CPU_ARCH_ARMv5TEJ && !cpu_is_xscale()) {
                for (i = 0; i < ARRAY_SIZE(mem_types); i++) {
                        if (mem_types[i].prot_l1)
                                mem_types[i].prot_l1 |= PMD_BIT4;
@@ -408,6 +384,20 @@ static void __init build_mem_type_table(void)
                }
        }
 
+       cp = &cache_policies[cachepolicy];
+       kern_pgprot = user_pgprot = cp->pte;
+
+       /*
+        * Enable CPU-specific coherency if supported.
+        * (Only available on XSC3 at the moment.)
+        */
+       if (arch_is_coherent()) {
+               if (cpu_is_xsc3()) {
+                       mem_types[MT_MEMORY].prot_sect |= PMD_SECT_S;
+                       mem_types[MT_MEMORY].prot_pte |= L_PTE_COHERENT;
+               }
+       }
+
        /*
         * ARMv6 and above have extended page tables.
         */
@@ -418,6 +408,7 @@ static void __init build_mem_type_table(void)
                 */
                mem_types[MT_MEMORY].prot_sect &= ~PMD_BIT4;
                mem_types[MT_ROM].prot_sect &= ~PMD_BIT4;
+
                /*
                 * Mark cache clean areas and XIP ROM read only
                 * from SVC mode and no access from userspace.
@@ -425,34 +416,59 @@ static void __init build_mem_type_table(void)
                mem_types[MT_ROM].prot_sect |= PMD_SECT_APX|PMD_SECT_AP_WRITE;
                mem_types[MT_MINICLEAN].prot_sect |= PMD_SECT_APX|PMD_SECT_AP_WRITE;
                mem_types[MT_CACHECLEAN].prot_sect |= PMD_SECT_APX|PMD_SECT_AP_WRITE;
+
+               /*
+                * Mark the device area as "shared device"
+                */
+               mem_types[MT_DEVICE].prot_pte |= L_PTE_BUFFERABLE;
+               mem_types[MT_DEVICE].prot_sect |= PMD_SECT_BUFFERED;
+
+               /*
+                * User pages need to be mapped with the ASID
+                * (iow, non-global)
+                */
+               user_pgprot |= L_PTE_ASID;
+
+#ifdef CONFIG_SMP
+               /*
+                * Mark memory with the "shared" attribute for SMP systems
+                */
+               user_pgprot |= L_PTE_SHARED;
+               kern_pgprot |= L_PTE_SHARED;
+               mem_types[MT_MEMORY].prot_sect |= PMD_SECT_S;
+#endif
        }
 
-       cp = &cache_policies[cachepolicy];
+       for (i = 0; i < 16; i++) {
+               unsigned long v = pgprot_val(protection_map[i]);
+               v = (v & ~(L_PTE_BUFFERABLE|L_PTE_CACHEABLE)) | user_pgprot;
+               protection_map[i] = __pgprot(v);
+       }
+
+       mem_types[MT_LOW_VECTORS].prot_pte |= kern_pgprot;
+       mem_types[MT_HIGH_VECTORS].prot_pte |= kern_pgprot;
 
        if (cpu_arch >= CPU_ARCH_ARMv5) {
-               mem_types[MT_LOW_VECTORS].prot_pte |= cp->pte & PTE_CACHEABLE;
-               mem_types[MT_HIGH_VECTORS].prot_pte |= cp->pte & PTE_CACHEABLE;
+#ifndef CONFIG_SMP
+               /*
+                * Only use write-through for non-SMP systems
+                */
+               mem_types[MT_LOW_VECTORS].prot_pte &= ~L_PTE_BUFFERABLE;
+               mem_types[MT_HIGH_VECTORS].prot_pte &= ~L_PTE_BUFFERABLE;
+#endif
        } else {
-               mem_types[MT_LOW_VECTORS].prot_pte |= cp->pte;
-               mem_types[MT_HIGH_VECTORS].prot_pte |= cp->pte;
                mem_types[MT_MINICLEAN].prot_sect &= ~PMD_SECT_TEX(1);
        }
 
+       pgprot_kernel = __pgprot(L_PTE_PRESENT | L_PTE_YOUNG |
+                                L_PTE_DIRTY | L_PTE_WRITE |
+                                L_PTE_EXEC | kern_pgprot);
+
        mem_types[MT_LOW_VECTORS].prot_l1 |= ecc_mask;
        mem_types[MT_HIGH_VECTORS].prot_l1 |= ecc_mask;
        mem_types[MT_MEMORY].prot_sect |= ecc_mask | cp->pmd;
        mem_types[MT_ROM].prot_sect |= cp->pmd;
 
-       for (i = 0; i < 16; i++) {
-               unsigned long v = pgprot_val(protection_map[i]);
-               v &= (~(PTE_BUFFERABLE|PTE_CACHEABLE)) | cp->pte;
-               protection_map[i] = __pgprot(v);
-       }
-
-       pgprot_kernel = __pgprot(L_PTE_PRESENT | L_PTE_YOUNG |
-                                L_PTE_DIRTY | L_PTE_WRITE |
-                                L_PTE_EXEC | cp->pte);
-
        switch (cp->pmd) {
        case PMD_SECT_WT:
                mem_types[MT_CACHECLEAN].prot_sect |= PMD_SECT_WT;
@@ -475,25 +491,25 @@ static void __init build_mem_type_table(void)
  * offsets, and we take full advantage of sections and
  * supersections.
  */
-static void __init create_mapping(struct map_desc *md)
+void __init create_mapping(struct map_desc *md)
 {
        unsigned long virt, length;
        int prot_sect, prot_l1, domain;
        pgprot_t prot_pte;
-       long off;
+       unsigned long off = (u32)__pfn_to_phys(md->pfn);
 
        if (md->virtual != vectors_base() && md->virtual < TASK_SIZE) {
                printk(KERN_WARNING "BUG: not creating mapping for "
-                      "0x%08lx at 0x%08lx in user region\n",
-                      md->physical, md->virtual);
+                      "0x%08llx at 0x%08lx in user region\n",
+                      __pfn_to_phys((u64)md->pfn), md->virtual);
                return;
        }
 
        if ((md->type == MT_DEVICE || md->type == MT_ROM) &&
            md->virtual >= PAGE_OFFSET && md->virtual < VMALLOC_END) {
-               printk(KERN_WARNING "BUG: mapping for 0x%08lx at 0x%08lx "
+               printk(KERN_WARNING "BUG: mapping for 0x%08llx at 0x%08lx "
                       "overlaps vmalloc space\n",
-                      md->physical, md->virtual);
+                      __pfn_to_phys((u64)md->pfn), md->virtual);
        }
 
        domain    = mem_types[md->type].domain;
@@ -501,15 +517,40 @@ static void __init create_mapping(struct map_desc *md)
        prot_l1   = mem_types[md->type].prot_l1 | PMD_DOMAIN(domain);
        prot_sect = mem_types[md->type].prot_sect | PMD_DOMAIN(domain);
 
+       /*
+        * Catch 36-bit addresses
+        */
+       if(md->pfn >= 0x100000) {
+               if(domain) {
+                       printk(KERN_ERR "MM: invalid domain in supersection "
+                               "mapping for 0x%08llx at 0x%08lx\n",
+                               __pfn_to_phys((u64)md->pfn), md->virtual);
+                       return;
+               }
+               if((md->virtual | md->length | __pfn_to_phys(md->pfn))
+                       & ~SUPERSECTION_MASK) {
+                       printk(KERN_ERR "MM: cannot create mapping for "
+                               "0x%08llx at 0x%08lx invalid alignment\n",
+                               __pfn_to_phys((u64)md->pfn), md->virtual);
+                       return;
+               }
+
+               /*
+                * Shift bits [35:32] of address into bits [23:20] of PMD
+                * (See ARMv6 spec).
+                */
+               off |= (((md->pfn >> (32 - PAGE_SHIFT)) & 0xF) << 20);
+       }
+
        virt   = md->virtual;
-       off    = md->physical - virt;
+       off   -= virt;
        length = md->length;
 
        if (mem_types[md->type].prot_l1 == 0 &&
            (virt & 0xfffff || (virt + off) & 0xfffff || (virt + length) & 0xfffff)) {
                printk(KERN_WARNING "BUG: map for 0x%08lx at 0x%08lx can not "
                       "be mapped using pages, ignoring.\n",
-                      md->physical, md->virtual);
+                      __pfn_to_phys(md->pfn), md->virtual);
                return;
        }
 
@@ -526,14 +567,24 @@ static void __init create_mapping(struct map_desc *md)
         *      supersections are only allocated for domain 0 regardless
         *      of the actual domain assignments in use.
         */
-       if (cpu_architecture() >= CPU_ARCH_ARMv6 && domain == 0) {
-               /* Align to supersection boundary */
-               while ((virt & ~SUPERSECTION_MASK || (virt + off) &
-                       ~SUPERSECTION_MASK) && length >= (PGDIR_SIZE / 2)) {
-                       alloc_init_section(virt, virt + off, prot_sect);
-
-                       virt   += (PGDIR_SIZE / 2);
-                       length -= (PGDIR_SIZE / 2);
+       if ((cpu_architecture() >= CPU_ARCH_ARMv6 || cpu_is_xsc3())
+               && domain == 0) {
+               /*
+                * Align to supersection boundary if !high pages.
+                * High pages have already been checked for proper
+                * alignment above and they will fail the SUPSERSECTION_MASK
+                * check because of the way the address is encoded into
+                * offset.
+                */
+               if (md->pfn <= 0x100000) {
+                       while ((virt & ~SUPERSECTION_MASK ||
+                               (virt + off) & ~SUPERSECTION_MASK) &&
+                               length >= (PGDIR_SIZE / 2)) {
+                               alloc_init_section(virt, virt + off, prot_sect);
+
+                               virt   += (PGDIR_SIZE / 2);
+                               length -= (PGDIR_SIZE / 2);
+                       }
                }
 
                while (length >= SUPERSECTION_SIZE) {
@@ -569,23 +620,23 @@ static void __init create_mapping(struct map_desc *md)
  */
 void setup_mm_for_reboot(char mode)
 {
-       unsigned long pmdval;
+       unsigned long base_pmdval;
        pgd_t *pgd;
-       pmd_t *pmd;
        int i;
-       int cpu_arch = cpu_architecture();
 
        if (current->mm && current->mm->pgd)
                pgd = current->mm->pgd;
        else
                pgd = init_mm.pgd;
 
-       for (i = 0; i < FIRST_USER_PGD_NR + USER_PTRS_PER_PGD; i++) {
-               pmdval = (i << PGDIR_SHIFT) |
-                        PMD_SECT_AP_WRITE | PMD_SECT_AP_READ |
-                        PMD_TYPE_SECT;
-               if (cpu_arch <= CPU_ARCH_ARMv5)
-                       pmdval |= PMD_BIT4;
+       base_pmdval = PMD_SECT_AP_WRITE | PMD_SECT_AP_READ | PMD_TYPE_SECT;
+       if (cpu_architecture() <= CPU_ARCH_ARMv5TEJ && !cpu_is_xscale())
+               base_pmdval |= PMD_BIT4;
+
+       for (i = 0; i < FIRST_USER_PGD_NR + USER_PTRS_PER_PGD; i++, pgd++) {
+               unsigned long pmdval = (i << PGDIR_SHIFT) | base_pmdval;
+               pmd_t *pmd;
+
                pmd = pmd_off(pgd, i << PGDIR_SHIFT);
                pmd[0] = __pmd(pmdval);
                pmd[1] = __pmd(pmdval + (1 << (PGDIR_SHIFT - 1)));
@@ -593,100 +644,6 @@ void setup_mm_for_reboot(char mode)
        }
 }
 
-extern void _stext, _etext;
-
-/*
- * Setup initial mappings.  We use the page we allocated for zero page to hold
- * the mappings, which will get overwritten by the vectors in traps_init().
- * The mappings must be in virtual address order.
- */
-void __init memtable_init(struct meminfo *mi)
-{
-       struct map_desc *init_maps, *p, *q;
-       unsigned long address = 0;
-       int i;
-
-       build_mem_type_table();
-
-       init_maps = p = alloc_bootmem_low_pages(PAGE_SIZE);
-
-#ifdef CONFIG_XIP_KERNEL
-       p->physical   = CONFIG_XIP_PHYS_ADDR & PMD_MASK;
-       p->virtual    = (unsigned long)&_stext & PMD_MASK;
-       p->length     = ((unsigned long)&_etext - p->virtual + ~PMD_MASK) & PMD_MASK;
-       p->type       = MT_ROM;
-       p ++;
-#endif
-
-       for (i = 0; i < mi->nr_banks; i++) {
-               if (mi->bank[i].size == 0)
-                       continue;
-
-               p->physical   = mi->bank[i].start;
-               p->virtual    = __phys_to_virt(p->physical);
-               p->length     = mi->bank[i].size;
-               p->type       = MT_MEMORY;
-               p ++;
-       }
-
-#ifdef FLUSH_BASE
-       p->physical   = FLUSH_BASE_PHYS;
-       p->virtual    = FLUSH_BASE;
-       p->length     = PGDIR_SIZE;
-       p->type       = MT_CACHECLEAN;
-       p ++;
-#endif
-
-#ifdef FLUSH_BASE_MINICACHE
-       p->physical   = FLUSH_BASE_PHYS + PGDIR_SIZE;
-       p->virtual    = FLUSH_BASE_MINICACHE;
-       p->length     = PGDIR_SIZE;
-       p->type       = MT_MINICLEAN;
-       p ++;
-#endif
-
-       /*
-        * Go through the initial mappings, but clear out any
-        * pgdir entries that are not in the description.
-        */
-       q = init_maps;
-       do {
-               if (address < q->virtual || q == p) {
-                       clear_mapping(address);
-                       address += PGDIR_SIZE;
-               } else {
-                       create_mapping(q);
-
-                       address = q->virtual + q->length;
-                       address = (address + PGDIR_SIZE - 1) & PGDIR_MASK;
-
-                       q ++;
-               }
-       } while (address != 0);
-
-       /*
-        * Create a mapping for the machine vectors at the high-vectors
-        * location (0xffff0000).  If we aren't using high-vectors, also
-        * create a mapping at the low-vectors virtual address.
-        */
-       init_maps->physical   = virt_to_phys(init_maps);
-       init_maps->virtual    = 0xffff0000;
-       init_maps->length     = PAGE_SIZE;
-       init_maps->type       = MT_HIGH_VECTORS;
-       create_mapping(init_maps);
-
-       if (!vectors_high()) {
-               init_maps->virtual = 0;
-               init_maps->type = MT_LOW_VECTORS;
-               create_mapping(init_maps);
-       }
-
-       flush_cache_all();
-       local_flush_tlb_all();
-
-       top_pmd = pmd_off_k(0xffff0000);
-}
-
 /*
  * Create the architecture specific mappings
  */