Merge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx
[sfrench/cifs-2.6.git] / arch / x86 / mm / init_32.c
index 3c76d194fd2ccef4df106a2f47fe9b02a3057264..ee1091a469641f95db2564c558827e972728ee37 100644 (file)
 #include <linux/bootmem.h>
 #include <linux/slab.h>
 #include <linux/proc_fs.h>
-#include <linux/efi.h>
 #include <linux/memory_hotplug.h>
 #include <linux/initrd.h>
 #include <linux/cpumask.h>
 
+#include <asm/asm.h>
 #include <asm/processor.h>
 #include <asm/system.h>
 #include <asm/uaccess.h>
 #include <asm/fixmap.h>
 #include <asm/e820.h>
 #include <asm/apic.h>
+#include <asm/bugs.h>
 #include <asm/tlb.h>
 #include <asm/tlbflush.h>
+#include <asm/pgalloc.h>
 #include <asm/sections.h>
 #include <asm/paravirt.h>
+#include <asm/setup.h>
+#include <asm/cacheflush.h>
 
 unsigned int __VMALLOC_RESERVE = 128 << 20;
 
 DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 unsigned long highstart_pfn, highend_pfn;
 
-static int noinline do_test_wp_bit(void);
+static noinline int do_test_wp_bit(void);
 
 /*
  * Creates a middle page table and puts a pointer to it in the
@@ -61,26 +65,26 @@ static pmd_t * __init one_md_table_init(pgd_t *pgd)
 {
        pud_t *pud;
        pmd_t *pmd_table;
-               
+
 #ifdef CONFIG_X86_PAE
        if (!(pgd_val(*pgd) & _PAGE_PRESENT)) {
                pmd_table = (pmd_t *) alloc_bootmem_low_pages(PAGE_SIZE);
 
-               paravirt_alloc_pd(__pa(pmd_table) >> PAGE_SHIFT);
+               paravirt_alloc_pd(&init_mm, __pa(pmd_table) >> PAGE_SHIFT);
                set_pgd(pgd, __pgd(__pa(pmd_table) | _PAGE_PRESENT));
                pud = pud_offset(pgd, 0);
-               if (pmd_table != pmd_offset(pud, 0))
-                       BUG();
+               BUG_ON(pmd_table != pmd_offset(pud, 0));
        }
 #endif
        pud = pud_offset(pgd, 0);
        pmd_table = pmd_offset(pud, 0);
+
        return pmd_table;
 }
 
 /*
  * Create a page table and place a pointer to it in a middle page
- * directory entry.
+ * directory entry:
  */
 static pte_t * __init one_page_table_init(pmd_t *pmd)
 {
@@ -90,9 +94,10 @@ static pte_t * __init one_page_table_init(pmd_t *pmd)
 #ifdef CONFIG_DEBUG_PAGEALLOC
                page_table = (pte_t *) alloc_bootmem_pages(PAGE_SIZE);
 #endif
-               if (!page_table)
+               if (!page_table) {
                        page_table =
                                (pte_t *)alloc_bootmem_low_pages(PAGE_SIZE);
+               }
 
                paravirt_alloc_pt(&init_mm, __pa(page_table) >> PAGE_SHIFT);
                set_pmd(pmd, __pmd(__pa(page_table) | _PAGE_TABLE));
@@ -103,22 +108,21 @@ static pte_t * __init one_page_table_init(pmd_t *pmd)
 }
 
 /*
- * This function initializes a certain range of kernel virtual memory 
+ * This function initializes a certain range of kernel virtual memory
  * with new bootmem page tables, everywhere page tables are missing in
  * the given range.
- */
-
-/*
- * NOTE: The pagetables are allocated contiguous on the physical space 
- * so we can cache the place of the first one and move around without 
+ *
+ * NOTE: The pagetables are allocated contiguous on the physical space
+ * so we can cache the place of the first one and move around without
  * checking the pgd every time.
  */
-static void __init page_table_range_init (unsigned long start, unsigned long end, pgd_t *pgd_base)
+static void __init
+page_table_range_init(unsigned long start, unsigned long end, pgd_t *pgd_base)
 {
-       pgd_t *pgd;
-       pmd_t *pmd;
        int pgd_idx, pmd_idx;
        unsigned long vaddr;
+       pgd_t *pgd;
+       pmd_t *pmd;
 
        vaddr = start;
        pgd_idx = pgd_index(vaddr);
@@ -128,7 +132,8 @@ static void __init page_table_range_init (unsigned long start, unsigned long end
        for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd++, pgd_idx++) {
                pmd = one_md_table_init(pgd);
                pmd = pmd + pmd_index(vaddr);
-               for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); pmd++, pmd_idx++) {
+               for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end);
+                                                       pmd++, pmd_idx++) {
                        one_page_table_init(pmd);
 
                        vaddr += PMD_SIZE;
@@ -145,17 +150,17 @@ static inline int is_kernel_text(unsigned long addr)
 }
 
 /*
- * This maps the physical memory to kernel virtual address space, a total 
- * of max_low_pfn pages, by creating page tables starting from address 
- * PAGE_OFFSET.
+ * This maps the physical memory to kernel virtual address space, a total
+ * of max_low_pfn pages, by creating page tables starting from address
+ * PAGE_OFFSET:
  */
 static void __init kernel_physical_mapping_init(pgd_t *pgd_base)
 {
+       int pgd_idx, pmd_idx, pte_ofs;
        unsigned long pfn;
        pgd_t *pgd;
        pmd_t *pmd;
        pte_t *pte;
-       int pgd_idx, pmd_idx, pte_ofs;
 
        pgd_idx = pgd_index(PAGE_OFFSET);
        pgd = pgd_base + pgd_idx;
@@ -165,29 +170,43 @@ static void __init kernel_physical_mapping_init(pgd_t *pgd_base)
                pmd = one_md_table_init(pgd);
                if (pfn >= max_low_pfn)
                        continue;
-               for (pmd_idx = 0; pmd_idx < PTRS_PER_PMD && pfn < max_low_pfn; pmd++, pmd_idx++) {
-                       unsigned int address = pfn * PAGE_SIZE + PAGE_OFFSET;
 
-                       /* Map with big pages if possible, otherwise create normal page tables. */
+               for (pmd_idx = 0;
+                    pmd_idx < PTRS_PER_PMD && pfn < max_low_pfn;
+                    pmd++, pmd_idx++) {
+                       unsigned int addr = pfn * PAGE_SIZE + PAGE_OFFSET;
+
+                       /*
+                        * Map with big pages if possible, otherwise
+                        * create normal page tables:
+                        */
                        if (cpu_has_pse) {
-                               unsigned int address2 = (pfn + PTRS_PER_PTE - 1) * PAGE_SIZE + PAGE_OFFSET + PAGE_SIZE-1;
-                               if (is_kernel_text(address) || is_kernel_text(address2))
-                                       set_pmd(pmd, pfn_pmd(pfn, PAGE_KERNEL_LARGE_EXEC));
-                               else
-                                       set_pmd(pmd, pfn_pmd(pfn, PAGE_KERNEL_LARGE));
+                               unsigned int addr2;
+                               pgprot_t prot = PAGE_KERNEL_LARGE;
+
+                               addr2 = (pfn + PTRS_PER_PTE-1) * PAGE_SIZE +
+                                       PAGE_OFFSET + PAGE_SIZE-1;
+
+                               if (is_kernel_text(addr) ||
+                                   is_kernel_text(addr2))
+                                       prot = PAGE_KERNEL_LARGE_EXEC;
+
+                               set_pmd(pmd, pfn_pmd(pfn, prot));
 
                                pfn += PTRS_PER_PTE;
-                       } else {
-                               pte = one_page_table_init(pmd);
-
-                               for (pte_ofs = 0;
-                                    pte_ofs < PTRS_PER_PTE && pfn < max_low_pfn;
-                                    pte++, pfn++, pte_ofs++, address += PAGE_SIZE) {
-                                       if (is_kernel_text(address))
-                                               set_pte(pte, pfn_pte(pfn, PAGE_KERNEL_EXEC));
-                                       else
-                                               set_pte(pte, pfn_pte(pfn, PAGE_KERNEL));
-                               }
+                               continue;
+                       }
+                       pte = one_page_table_init(pmd);
+
+                       for (pte_ofs = 0;
+                            pte_ofs < PTRS_PER_PTE && pfn < max_low_pfn;
+                            pte++, pfn++, pte_ofs++, addr += PAGE_SIZE) {
+                               pgprot_t prot = PAGE_KERNEL;
+
+                               if (is_kernel_text(addr))
+                                       prot = PAGE_KERNEL_EXEC;
+
+                               set_pte(pte, pfn_pte(pfn, prot));
                        }
                }
        }
@@ -200,57 +219,23 @@ static inline int page_kills_ppro(unsigned long pagenr)
        return 0;
 }
 
-int page_is_ram(unsigned long pagenr)
-{
-       int i;
-       unsigned long addr, end;
-
-       if (efi_enabled) {
-               efi_memory_desc_t *md;
-               void *p;
-
-               for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
-                       md = p;
-                       if (!is_available_memory(md))
-                               continue;
-                       addr = (md->phys_addr+PAGE_SIZE-1) >> PAGE_SHIFT;
-                       end = (md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT)) >> PAGE_SHIFT;
-
-                       if ((pagenr >= addr) && (pagenr < end))
-                               return 1;
-               }
-               return 0;
-       }
-
-       for (i = 0; i < e820.nr_map; i++) {
-
-               if (e820.map[i].type != E820_RAM)       /* not usable memory */
-                       continue;
-               /*
-                *      !!!FIXME!!! Some BIOSen report areas as RAM that
-                *      are not. Notably the 640->1Mb area. We need a sanity
-                *      check here.
-                */
-               addr = (e820.map[i].addr+PAGE_SIZE-1) >> PAGE_SHIFT;
-               end = (e820.map[i].addr+e820.map[i].size) >> PAGE_SHIFT;
-               if  ((pagenr >= addr) && (pagenr < end))
-                       return 1;
-       }
-       return 0;
-}
-
 #ifdef CONFIG_HIGHMEM
 pte_t *kmap_pte;
 pgprot_t kmap_prot;
 
-#define kmap_get_fixmap_pte(vaddr)                                     \
-       pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr), vaddr), (vaddr)), (vaddr))
+static inline pte_t *kmap_get_fixmap_pte(unsigned long vaddr)
+{
+       return pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr),
+                       vaddr), vaddr), vaddr);
+}
 
 static void __init kmap_init(void)
 {
        unsigned long kmap_vstart;
 
-       /* cache the first kmap pte */
+       /*
+        * Cache the first kmap pte:
+        */
        kmap_vstart = __fix_to_virt(FIX_KMAP_BEGIN);
        kmap_pte = kmap_get_fixmap_pte(kmap_vstart);
 
@@ -259,11 +244,11 @@ static void __init kmap_init(void)
 
 static void __init permanent_kmaps_init(pgd_t *pgd_base)
 {
+       unsigned long vaddr;
        pgd_t *pgd;
        pud_t *pud;
        pmd_t *pmd;
        pte_t *pte;
-       unsigned long vaddr;
 
        vaddr = PKMAP_BASE;
        page_table_range_init(vaddr, vaddr + PAGE_SIZE*LAST_PKMAP, pgd_base);
@@ -272,7 +257,7 @@ static void __init permanent_kmaps_init(pgd_t *pgd_base)
        pud = pud_offset(pgd, vaddr);
        pmd = pmd_offset(pud, vaddr);
        pte = pte_offset_kernel(pmd, vaddr);
-       pkmap_page_table = pte; 
+       pkmap_page_table = pte;
 }
 
 static void __meminit free_new_highpage(struct page *page)
@@ -291,7 +276,8 @@ void __init add_one_highpage_init(struct page *page, int pfn, int bad_ppro)
                SetPageReserved(page);
 }
 
-static int __meminit add_one_highpage_hotplug(struct page *page, unsigned long pfn)
+static int __meminit
+add_one_highpage_hotplug(struct page *page, unsigned long pfn)
 {
        free_new_highpage(page);
        totalram_pages++;
@@ -299,6 +285,7 @@ static int __meminit add_one_highpage_hotplug(struct page *page, unsigned long p
        max_mapnr = max(pfn, max_mapnr);
 #endif
        num_physpages++;
+
        return 0;
 }
 
@@ -306,7 +293,7 @@ static int __meminit add_one_highpage_hotplug(struct page *page, unsigned long p
  * Not currently handling the NUMA case.
  * Assuming single node and all memory that
  * has been added dynamically that would be
- * onlined here is in HIGHMEM
+ * onlined here is in HIGHMEM.
  */
 void __meminit online_page(struct page *page)
 {
@@ -314,13 +301,11 @@ void __meminit online_page(struct page *page)
        add_one_highpage_hotplug(page, page_to_pfn(page));
 }
 
-
-#ifdef CONFIG_NUMA
-extern void set_highmem_pages_init(int);
-#else
+#ifndef CONFIG_NUMA
 static void __init set_highmem_pages_init(int bad_ppro)
 {
        int pfn;
+
        for (pfn = highstart_pfn; pfn < highend_pfn; pfn++) {
                /*
                 * Holes under sparsemem might not have no mem_map[]:
@@ -330,64 +315,53 @@ static void __init set_highmem_pages_init(int bad_ppro)
        }
        totalram_pages += totalhigh_pages;
 }
-#endif /* CONFIG_FLATMEM */
+#endif /* !CONFIG_NUMA */
 
 #else
-#define kmap_init() do { } while (0)
-#define permanent_kmaps_init(pgd_base) do { } while (0)
-#define set_highmem_pages_init(bad_ppro) do { } while (0)
+# define kmap_init()                           do { } while (0)
+# define permanent_kmaps_init(pgd_base)                do { } while (0)
+# define set_highmem_pages_init(bad_ppro)      do { } while (0)
 #endif /* CONFIG_HIGHMEM */
 
-unsigned long long __PAGE_KERNEL = _PAGE_KERNEL;
+pteval_t __PAGE_KERNEL = _PAGE_KERNEL;
 EXPORT_SYMBOL(__PAGE_KERNEL);
-unsigned long long __PAGE_KERNEL_EXEC = _PAGE_KERNEL_EXEC;
 
-#ifdef CONFIG_NUMA
-extern void __init remap_numa_kva(void);
-#else
-#define remap_numa_kva() do {} while (0)
-#endif
+pteval_t __PAGE_KERNEL_EXEC = _PAGE_KERNEL_EXEC;
 
 void __init native_pagetable_setup_start(pgd_t *base)
 {
-#ifdef CONFIG_X86_PAE
-       int i;
+       unsigned long pfn, va;
+       pgd_t *pgd;
+       pud_t *pud;
+       pmd_t *pmd;
+       pte_t *pte;
 
        /*
-        * Init entries of the first-level page table to the
-        * zero page, if they haven't already been set up.
-        *
-        * In a normal native boot, we'll be running on a
-        * pagetable rooted in swapper_pg_dir, but not in PAE
-        * mode, so this will end up clobbering the mappings
-        * for the lower 24Mbytes of the address space,
-        * without affecting the kernel address space.
+        * Remove any mappings which extend past the end of physical
+        * memory from the boot time page table:
         */
-       for (i = 0; i < USER_PTRS_PER_PGD; i++)
-               set_pgd(&base[i],
-                       __pgd(__pa(empty_zero_page) | _PAGE_PRESENT));
-
-       /* Make sure kernel address space is empty so that a pagetable
-          will be allocated for it. */
-       memset(&base[USER_PTRS_PER_PGD], 0,
-              KERNEL_PGD_PTRS * sizeof(pgd_t));
-#else
-       paravirt_alloc_pd(__pa(swapper_pg_dir) >> PAGE_SHIFT);
-#endif
+       for (pfn = max_low_pfn + 1; pfn < 1<<(32-PAGE_SHIFT); pfn++) {
+               va = PAGE_OFFSET + (pfn<<PAGE_SHIFT);
+               pgd = base + pgd_index(va);
+               if (!pgd_present(*pgd))
+                       break;
+
+               pud = pud_offset(pgd, va);
+               pmd = pmd_offset(pud, va);
+               if (!pmd_present(*pmd))
+                       break;
+
+               pte = pte_offset_kernel(pmd, va);
+               if (!pte_present(*pte))
+                       break;
+
+               pte_clear(NULL, va, pte);
+       }
+       paravirt_alloc_pd(&init_mm, __pa(base) >> PAGE_SHIFT);
 }
 
 void __init native_pagetable_setup_done(pgd_t *base)
 {
-#ifdef CONFIG_X86_PAE
-       /*
-        * Add low memory identity-mappings - SMP needs it when
-        * starting up on an AP from real-mode. In the non-PAE
-        * case we already have these mappings through head.S.
-        * All user-space mappings are explicitly cleared after
-        * SMP startup.
-        */
-       set_pgd(&base[0], base[USER_PTRS_PER_PGD]);
-#endif
 }
 
 /*
@@ -396,9 +370,8 @@ void __init native_pagetable_setup_done(pgd_t *base)
  * the boot process.
  *
  * If we're booting on native hardware, this will be a pagetable
- * constructed in arch/i386/kernel/head.S, and not running in PAE mode
- * (even if we'll end up running in PAE).  The root of the pagetable
- * will be swapper_pg_dir.
+ * constructed in arch/x86/kernel/head_32.S.  The root of the
+ * pagetable will be swapper_pg_dir.
  *
  * If we're booting paravirtualized under a hypervisor, then there are
  * more options: we may already be running PAE, and the pagetable may
@@ -410,10 +383,10 @@ void __init native_pagetable_setup_done(pgd_t *base)
  * be partially populated, and so it avoids stomping on any existing
  * mappings.
  */
-static void __init pagetable_init (void)
+static void __init pagetable_init(void)
 {
-       unsigned long vaddr, end;
        pgd_t *pgd_base = swapper_pg_dir;
+       unsigned long vaddr, end;
 
        paravirt_pagetable_setup_start(pgd_base);
 
@@ -435,34 +408,36 @@ static void __init pagetable_init (void)
         * Fixed mappings, only the page table structure has to be
         * created - mappings will be set by set_fixmap():
         */
+       early_ioremap_clear();
        vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK;
        end = (FIXADDR_TOP + PMD_SIZE - 1) & PMD_MASK;
        page_table_range_init(vaddr, end, pgd_base);
+       early_ioremap_reset();
 
        permanent_kmaps_init(pgd_base);
 
        paravirt_pagetable_setup_done(pgd_base);
 }
 
-#if defined(CONFIG_HIBERNATION) || defined(CONFIG_ACPI)
+#ifdef CONFIG_ACPI_SLEEP
 /*
- * Swap suspend & friends need this for resume because things like the intel-agp
+ * ACPI suspend needs this for resume, because things like the intel-agp
  * driver might have split up a kernel 4MB mapping.
  */
-char __nosavedata swsusp_pg_dir[PAGE_SIZE]
-       __attribute__ ((aligned (PAGE_SIZE)));
+char swsusp_pg_dir[PAGE_SIZE]
+       __attribute__ ((aligned(PAGE_SIZE)));
 
 static inline void save_pg_dir(void)
 {
        memcpy(swsusp_pg_dir, swapper_pg_dir, PAGE_SIZE);
 }
-#else
+#else /* !CONFIG_ACPI_SLEEP */
 static inline void save_pg_dir(void)
 {
 }
-#endif
+#endif /* !CONFIG_ACPI_SLEEP */
 
-void zap_low_mappings (void)
+void zap_low_mappings(void)
 {
        int i;
 
@@ -474,22 +449,24 @@ void zap_low_mappings (void)
         * Note that "pgd_clear()" doesn't do it for
         * us, because pgd_clear() is a no-op on i386.
         */
-       for (i = 0; i < USER_PTRS_PER_PGD; i++)
+       for (i = 0; i < USER_PTRS_PER_PGD; i++) {
 #ifdef CONFIG_X86_PAE
                set_pgd(swapper_pg_dir+i, __pgd(1 + __pa(empty_zero_page)));
 #else
                set_pgd(swapper_pg_dir+i, __pgd(0));
 #endif
+       }
        flush_tlb_all();
 }
 
-int nx_enabled = 0;
+int nx_enabled;
+
+pteval_t __supported_pte_mask __read_mostly = ~_PAGE_NX;
+EXPORT_SYMBOL_GPL(__supported_pte_mask);
 
 #ifdef CONFIG_X86_PAE
 
-static int disable_nx __initdata = 0;
-u64 __supported_pte_mask __read_mostly = ~_PAGE_NX;
-EXPORT_SYMBOL_GPL(__supported_pte_mask);
+static int disable_nx __initdata;
 
 /*
  * noexec = on|off
@@ -506,11 +483,14 @@ static int __init noexec_setup(char *str)
                        __supported_pte_mask |= _PAGE_NX;
                        disable_nx = 0;
                }
-       } else if (!strcmp(str,"off")) {
-               disable_nx = 1;
-               __supported_pte_mask &= ~_PAGE_NX;
-       } else
-               return -EINVAL;
+       } else {
+               if (!strcmp(str, "off")) {
+                       disable_nx = 1;
+                       __supported_pte_mask &= ~_PAGE_NX;
+               } else {
+                       return -EINVAL;
+               }
+       }
 
        return 0;
 }
@@ -522,6 +502,7 @@ static void __init set_nx(void)
 
        if (cpu_has_pae && (cpuid_eax(0x80000000) > 0x80000001)) {
                cpuid(0x80000001, &v[0], &v[1], &v[2], &v[3]);
+
                if ((v[3] & (1 << 20)) && !disable_nx) {
                        rdmsr(MSR_EFER, l, h);
                        l |= EFER_NX;
@@ -531,35 +512,6 @@ static void __init set_nx(void)
                }
        }
 }
-
-/*
- * Enables/disables executability of a given kernel page and
- * returns the previous setting.
- */
-int __init set_kernel_exec(unsigned long vaddr, int enable)
-{
-       pte_t *pte;
-       int ret = 1;
-
-       if (!nx_enabled)
-               goto out;
-
-       pte = lookup_address(vaddr);
-       BUG_ON(!pte);
-
-       if (!pte_exec_kernel(*pte))
-               ret = 0;
-
-       if (enable)
-               pte->pte_high &= ~(1 << (_PAGE_BIT_NX - 32));
-       else
-               pte->pte_high |= 1 << (_PAGE_BIT_NX - 32);
-       pte_update_defer(&init_mm, vaddr, pte);
-       __flush_tlb_all();
-out:
-       return ret;
-}
-
 #endif
 
 /*
@@ -574,21 +526,12 @@ void __init paging_init(void)
 #ifdef CONFIG_X86_PAE
        set_nx();
        if (nx_enabled)
-               printk("NX (Execute Disable) protection: active\n");
+               printk(KERN_INFO "NX (Execute Disable) protection: active\n");
 #endif
-
        pagetable_init();
 
        load_cr3(swapper_pg_dir);
 
-#ifdef CONFIG_X86_PAE
-       /*
-        * We will bail out later - printk doesn't work right now so
-        * the user would just see a hanging kernel.
-        */
-       if (cpu_has_pae)
-               set_in_cr4(X86_CR4_PAE);
-#endif
        __flush_tlb_all();
 
        kmap_init();
@@ -600,10 +543,10 @@ void __init paging_init(void)
  * used to involve black magic jumps to work around some nasty CPU bugs,
  * but fortunately the switch to using exceptions got rid of all that.
  */
-
 static void __init test_wp_bit(void)
 {
-       printk("Checking if this processor honours the WP bit even in supervisor mode... ");
+       printk(KERN_INFO
+  "Checking if this processor honours the WP bit even in supervisor mode...");
 
        /* Any page-aligned address will do, the test is non-destructive */
        __set_fixmap(FIX_WP_TEST, __pa(&swapper_pg_dir), PAGE_READONLY);
@@ -611,47 +554,46 @@ static void __init test_wp_bit(void)
        clear_fixmap(FIX_WP_TEST);
 
        if (!boot_cpu_data.wp_works_ok) {
-               printk("No.\n");
+               printk(KERN_CONT "No.\n");
 #ifdef CONFIG_X86_WP_WORKS_OK
-               panic("This kernel doesn't support CPU's with broken WP. Recompile it for a 386!");
+               panic(
+  "This kernel doesn't support CPU's with broken WP. Recompile it for a 386!");
 #endif
        } else {
-               printk("Ok.\n");
+               printk(KERN_CONT "Ok.\n");
        }
 }
 
-static struct kcore_list kcore_mem, kcore_vmalloc; 
+static struct kcore_list kcore_mem, kcore_vmalloc;
 
 void __init mem_init(void)
 {
-       extern int ppro_with_ram_bug(void);
        int codesize, reservedpages, datasize, initsize;
-       int tmp;
-       int bad_ppro;
+       int tmp, bad_ppro;
 
 #ifdef CONFIG_FLATMEM
        BUG_ON(!mem_map);
 #endif
-       
        bad_ppro = ppro_with_ram_bug();
 
 #ifdef CONFIG_HIGHMEM
        /* check that fixmap and pkmap do not overlap */
-       if (PKMAP_BASE+LAST_PKMAP*PAGE_SIZE >= FIXADDR_START) {
-               printk(KERN_ERR "fixmap and kmap areas overlap - this will crash\n");
+       if (PKMAP_BASE + LAST_PKMAP*PAGE_SIZE >= FIXADDR_START) {
+               printk(KERN_ERR
+                       "fixmap and kmap areas overlap - this will crash\n");
                printk(KERN_ERR "pkstart: %lxh pkend: %lxh fixstart %lxh\n",
-                               PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE, FIXADDR_START);
+                               PKMAP_BASE, PKMAP_BASE + LAST_PKMAP*PAGE_SIZE,
+                               FIXADDR_START);
                BUG();
        }
 #endif
        /* this will put all low memory onto the freelists */
        totalram_pages += free_all_bootmem();
 
        reservedpages = 0;
        for (tmp = 0; tmp < max_low_pfn; tmp++)
                /*
-                * Only count reserved RAM pages
+                * Only count reserved RAM pages:
                 */
                if (page_is_ram(tmp) && PageReserved(pfn_to_page(tmp)))
                        reservedpages++;
@@ -662,11 +604,12 @@ void __init mem_init(void)
        datasize =  (unsigned long) &_edata - (unsigned long) &_etext;
        initsize =  (unsigned long) &__init_end - (unsigned long) &__init_begin;
 
-       kclist_add(&kcore_mem, __va(0), max_low_pfn << PAGE_SHIFT); 
-       kclist_add(&kcore_vmalloc, (void *)VMALLOC_START, 
+       kclist_add(&kcore_mem, __va(0), max_low_pfn << PAGE_SHIFT);
+       kclist_add(&kcore_vmalloc, (void *)VMALLOC_START,
                   VMALLOC_END-VMALLOC_START);
 
-       printk(KERN_INFO "Memory: %luk/%luk available (%dk kernel code, %dk reserved, %dk data, %dk init, %ldk highmem)\n",
+       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),
                num_physpages << (PAGE_SHIFT-10),
                codesize >> 10,
@@ -677,54 +620,53 @@ void __init mem_init(void)
               );
 
 #if 1 /* double-sanity-check paranoia */
-       printk("virtual kernel memory layout:\n"
-              "    fixmap  : 0x%08lx - 0x%08lx   (%4ld kB)\n"
+       printk(KERN_INFO "virtual kernel memory layout:\n"
+               "    fixmap  : 0x%08lx - 0x%08lx   (%4ld kB)\n"
 #ifdef CONFIG_HIGHMEM
-              "    pkmap   : 0x%08lx - 0x%08lx   (%4ld kB)\n"
+               "    pkmap   : 0x%08lx - 0x%08lx   (%4ld kB)\n"
 #endif
-              "    vmalloc : 0x%08lx - 0x%08lx   (%4ld MB)\n"
-              "    lowmem  : 0x%08lx - 0x%08lx   (%4ld MB)\n"
-              "      .init : 0x%08lx - 0x%08lx   (%4ld kB)\n"
-              "      .data : 0x%08lx - 0x%08lx   (%4ld kB)\n"
-              "      .text : 0x%08lx - 0x%08lx   (%4ld kB)\n",
-              FIXADDR_START, FIXADDR_TOP,
-              (FIXADDR_TOP - FIXADDR_START) >> 10,
+               "    vmalloc : 0x%08lx - 0x%08lx   (%4ld MB)\n"
+               "    lowmem  : 0x%08lx - 0x%08lx   (%4ld MB)\n"
+               "      .init : 0x%08lx - 0x%08lx   (%4ld kB)\n"
+               "      .data : 0x%08lx - 0x%08lx   (%4ld kB)\n"
+               "      .text : 0x%08lx - 0x%08lx   (%4ld kB)\n",
+               FIXADDR_START, FIXADDR_TOP,
+               (FIXADDR_TOP - FIXADDR_START) >> 10,
 
 #ifdef CONFIG_HIGHMEM
-              PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE,
-              (LAST_PKMAP*PAGE_SIZE) >> 10,
+               PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE,
+               (LAST_PKMAP*PAGE_SIZE) >> 10,
 #endif
 
-              VMALLOC_START, VMALLOC_END,
-              (VMALLOC_END - VMALLOC_START) >> 20,
+               VMALLOC_START, VMALLOC_END,
+               (VMALLOC_END - VMALLOC_START) >> 20,
 
-              (unsigned long)__va(0), (unsigned long)high_memory,
-              ((unsigned long)high_memory - (unsigned long)__va(0)) >> 20,
+               (unsigned long)__va(0), (unsigned long)high_memory,
+               ((unsigned long)high_memory - (unsigned long)__va(0)) >> 20,
 
-              (unsigned long)&__init_begin, (unsigned long)&__init_end,
-              ((unsigned long)&__init_end - (unsigned long)&__init_begin) >> 10,
+               (unsigned long)&__init_begin, (unsigned long)&__init_end,
+               ((unsigned long)&__init_end -
+                (unsigned long)&__init_begin) >> 10,
 
-              (unsigned long)&_etext, (unsigned long)&_edata,
-              ((unsigned long)&_edata - (unsigned long)&_etext) >> 10,
+               (unsigned long)&_etext, (unsigned long)&_edata,
+               ((unsigned long)&_edata - (unsigned long)&_etext) >> 10,
 
-              (unsigned long)&_text, (unsigned long)&_etext,
-              ((unsigned long)&_etext - (unsigned long)&_text) >> 10);
+               (unsigned long)&_text, (unsigned long)&_etext,
+               ((unsigned long)&_etext - (unsigned long)&_text) >> 10);
 
 #ifdef CONFIG_HIGHMEM
-       BUG_ON(PKMAP_BASE+LAST_PKMAP*PAGE_SIZE > FIXADDR_START);
-       BUG_ON(VMALLOC_END                     > PKMAP_BASE);
+       BUG_ON(PKMAP_BASE + LAST_PKMAP*PAGE_SIZE        > FIXADDR_START);
+       BUG_ON(VMALLOC_END                              > PKMAP_BASE);
 #endif
-       BUG_ON(VMALLOC_START                   > VMALLOC_END);
-       BUG_ON((unsigned long)high_memory      > VMALLOC_START);
+       BUG_ON(VMALLOC_START                            > VMALLOC_END);
+       BUG_ON((unsigned long)high_memory               > VMALLOC_START);
 #endif /* double-sanity-check paranoia */
 
-#ifdef CONFIG_X86_PAE
-       if (!cpu_has_pae)
-               panic("cannot execute a PAE-enabled kernel on a PAE-less CPU!");
-#endif
        if (boot_cpu_data.wp_works_ok < 0)
                test_wp_bit();
 
+       cpa_init();
+
        /*
         * Subtle. SMP is doing it's boot stuff late (because it has to
         * fork idle threads) - but it also needs low mappings for the
@@ -746,49 +688,35 @@ int arch_add_memory(int nid, u64 start, u64 size)
 
        return __add_pages(zone, start_pfn, nr_pages);
 }
-
 #endif
 
-struct kmem_cache *pmd_cache;
-
-void __init pgtable_cache_init(void)
-{
-       if (PTRS_PER_PMD > 1)
-               pmd_cache = kmem_cache_create("pmd",
-                                             PTRS_PER_PMD*sizeof(pmd_t),
-                                             PTRS_PER_PMD*sizeof(pmd_t),
-                                             SLAB_PANIC,
-                                             pmd_ctor);
-}
-
 /*
  * This function cannot be __init, since exceptions don't work in that
  * section.  Put this after the callers, so that it cannot be inlined.
  */
-static int noinline do_test_wp_bit(void)
+static noinline int do_test_wp_bit(void)
 {
        char tmp_reg;
        int flag;
 
        __asm__ __volatile__(
-               "       movb %0,%1      \n"
-               "1:     movb %1,%0      \n"
-               "       xorl %2,%2      \n"
+               "       movb %0, %1     \n"
+               "1:     movb %1, %0     \n"
+               "       xorl %2, %2     \n"
                "2:                     \n"
-               ".section __ex_table,\"a\"\n"
-               "       .align 4        \n"
-               "       .long 1b,2b     \n"
-               ".previous              \n"
+               _ASM_EXTABLE(1b,2b)
                :"=m" (*(char *)fix_to_virt(FIX_WP_TEST)),
                 "=q" (tmp_reg),
                 "=r" (flag)
                :"2" (1)
                :"memory");
-       
+
        return flag;
 }
 
 #ifdef CONFIG_DEBUG_RODATA
+const int rodata_test_data = 0xC3;
+EXPORT_SYMBOL_GPL(rodata_test_data);
 
 void mark_rodata_ro(void)
 {
@@ -801,32 +729,58 @@ void mark_rodata_ro(void)
        if (num_possible_cpus() <= 1)
 #endif
        {
-               change_page_attr(virt_to_page(start),
-                                size >> PAGE_SHIFT, PAGE_KERNEL_RX);
-               printk("Write protecting the kernel text: %luk\n", size >> 10);
+               set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT);
+               printk(KERN_INFO "Write protecting the kernel text: %luk\n",
+                       size >> 10);
+
+#ifdef CONFIG_CPA_DEBUG
+               printk(KERN_INFO "Testing CPA: Reverting %lx-%lx\n",
+                       start, start+size);
+               set_pages_rw(virt_to_page(start), size>>PAGE_SHIFT);
+
+               printk(KERN_INFO "Testing CPA: write protecting again\n");
+               set_pages_ro(virt_to_page(start), size>>PAGE_SHIFT);
+#endif
        }
 #endif
        start += size;
        size = (unsigned long)__end_rodata - start;
-       change_page_attr(virt_to_page(start),
-                        size >> PAGE_SHIFT, PAGE_KERNEL_RO);
-       printk("Write protecting the kernel read-only data: %luk\n",
-              size >> 10);
+       set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT);
+       printk(KERN_INFO "Write protecting the kernel read-only data: %luk\n",
+               size >> 10);
+       rodata_test();
 
-       /*
-        * change_page_attr() requires a global_flush_tlb() call after it.
-        * We do this after the printk so that if something went wrong in the
-        * change, the printk gets out at least to give a better debug hint
-        * of who is the culprit.
-        */
-       global_flush_tlb();
+#ifdef CONFIG_CPA_DEBUG
+       printk(KERN_INFO "Testing CPA: undo %lx-%lx\n", start, start + size);
+       set_pages_rw(virt_to_page(start), size >> PAGE_SHIFT);
+
+       printk(KERN_INFO "Testing CPA: write protecting again\n");
+       set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT);
+#endif
 }
 #endif
 
 void free_init_pages(char *what, unsigned long begin, unsigned long end)
 {
+#ifdef CONFIG_DEBUG_PAGEALLOC
+       /*
+        * If debugging page accesses then do not free this memory but
+        * mark them not present - any buggy init-section access will
+        * create a kernel page fault:
+        */
+       printk(KERN_INFO "debug: unmapping init memory %08lx..%08lx\n",
+               begin, PAGE_ALIGN(end));
+       set_memory_np(begin, (end - begin) >> PAGE_SHIFT);
+#else
        unsigned long addr;
 
+       /*
+        * We just marked the kernel text read only above, now that
+        * we are going to free part of that, we need to make that
+        * writeable first.
+        */
+       set_memory_rw(begin, (end - begin) >> PAGE_SHIFT);
+
        for (addr = begin; addr < end; addr += PAGE_SIZE) {
                ClearPageReserved(virt_to_page(addr));
                init_page_count(virt_to_page(addr));
@@ -835,6 +789,7 @@ void free_init_pages(char *what, unsigned long begin, unsigned long end)
                totalram_pages++;
        }
        printk(KERN_INFO "Freeing %s: %luk freed\n", what, (end - begin) >> 10);
+#endif
 }
 
 void free_initmem(void)
@@ -850,4 +805,3 @@ void free_initrd_mem(unsigned long start, unsigned long end)
        free_init_pages("initrd memory", start, end);
 }
 #endif
-