Merge phase #2 (PAT updates) of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'x86-v28-for-linus-phase2-B' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (27 commits)
  x86, cpa: make the kernel physical mapping initialization a two pass sequence, fix
  x86, pat: cleanups
  x86: fix pagetable init 64-bit breakage
  x86: track memtype for RAM in page struct
  x86, cpa: srlz cpa(), global flush tlb after splitting big page and before doing cpa
  x86, cpa: remove cpa pool code
  x86, cpa: no need to check alias for __set_pages_p/__set_pages_np
  x86, cpa: dont use large pages for kernel identity mapping with DEBUG_PAGEALLOC
  x86, cpa: make the kernel physical mapping initialization a two pass sequence
  x86, cpa: remove USER permission from the very early identity mapping attribute
  x86, cpa: rename PTE attribute macros for kernel direct mapping in early boot
  x86: make sure the CPA test code's use of _PAGE_UNUSED1 is obvious
  linux-next: fix x86 tree build failure
  x86: have set_memory_array_{uc,wb} coalesce memtypes, fix
  agp: enable optimized agp_alloc_pages methods
  x86: have set_memory_array_{uc,wb} coalesce memtypes.
  x86: {reverve,free}_memtype() take a physical address
  x86: fix pageattr-test
  agp: add agp_generic_destroy_pages()
  agp: generic_alloc_pages()
  ...

1 files changed, 82 insertions, 28 deletions

diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index 770536ebf7e..fb30486c82f 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -271,7 +271,8 @@ static __ref void unmap_low_page(void *adr)
 }
 
 static unsigned long __meminit
-phys_pte_init(pte_t *pte_page, unsigned long addr, unsigned long end)
+phys_pte_init(pte_t *pte_page, unsigned long addr, unsigned long end,
+              pgprot_t prot)
 {
         unsigned pages = 0;
         unsigned long last_map_addr = end;
@@ -289,36 +290,43 @@ phys_pte_init(pte_t *pte_page, unsigned long addr, unsigned long end)
                         break;
                 }
 
+                /*
+                 * We will re-use the existing mapping.
+                 * Xen for example has some special requirements, like mapping
+                 * pagetable pages as RO. So assume someone who pre-setup
+                 * these mappings are more intelligent.
+                 */
                 if (pte_val(*pte))
                         continue;
 
                 if (0)
                         printk("   pte=%p addr=%lx pte=%016lx\n",
                                pte, addr, pfn_pte(addr >> PAGE_SHIFT, PAGE_KERNEL).pte);
-                set_pte(pte, pfn_pte(addr >> PAGE_SHIFT, PAGE_KERNEL));
-                last_map_addr = (addr & PAGE_MASK) + PAGE_SIZE;
                 pages++;
+                set_pte(pte, pfn_pte(addr >> PAGE_SHIFT, prot));
+                last_map_addr = (addr & PAGE_MASK) + PAGE_SIZE;
         }
+
         update_page_count(PG_LEVEL_4K, pages);
 
         return last_map_addr;
 }
 
 static unsigned long __meminit
-phys_pte_update(pmd_t *pmd, unsigned long address, unsigned long end)
+phys_pte_update(pmd_t *pmd, unsigned long address, unsigned long end,
+                pgprot_t prot)
 {
         pte_t *pte = (pte_t *)pmd_page_vaddr(*pmd);
 
-        return phys_pte_init(pte, address, end);
+        return phys_pte_init(pte, address, end, prot);
 }
 
 static unsigned long __meminit
 phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end,
-                         unsigned long page_size_mask)
+              unsigned long page_size_mask, pgprot_t prot)
 {
         unsigned long pages = 0;
         unsigned long last_map_addr = end;
-        unsigned long start = address;
 
         int i = pmd_index(address);
 
@@ -326,6 +334,7 @@ phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end,
                 unsigned long pte_phys;
                 pmd_t *pmd = pmd_page + pmd_index(address);
                 pte_t *pte;
+                pgprot_t new_prot = prot;
 
                 if (address >= end) {
                         if (!after_bootmem) {
@@ -339,27 +348,40 @@ phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end,
                         if (!pmd_large(*pmd)) {
                                 spin_lock(&init_mm.page_table_lock);
                                 last_map_addr = phys_pte_update(pmd, address,
-                                                                end);
+                                                                end, prot);
                                 spin_unlock(&init_mm.page_table_lock);
+                                continue;
                         }
-                        /* Count entries we're using from level2_ident_pgt */
-                        if (start == 0)
-                                pages++;
-                        continue;
+                        /*
+                         * If we are ok with PG_LEVEL_2M mapping, then we will
+                         * use the existing mapping,
+                         *
+                         * Otherwise, we will split the large page mapping but
+                         * use the same existing protection bits except for
+                         * large page, so that we don't violate Intel's TLB
+                         * Application note (317080) which says, while changing
+                         * the page sizes, new and old translations should
+                         * not differ with respect to page frame and
+                         * attributes.
+                         */
+                        if (page_size_mask & (1 << PG_LEVEL_2M))
+                                continue;
+                        new_prot = pte_pgprot(pte_clrhuge(*(pte_t *)pmd));
                 }
 
                 if (page_size_mask & (1<<PG_LEVEL_2M)) {
                         pages++;
                         spin_lock(&init_mm.page_table_lock);
                         set_pte((pte_t *)pmd,
-                                pfn_pte(address >> PAGE_SHIFT, PAGE_KERNEL_LARGE));
+                                pfn_pte(address >> PAGE_SHIFT,
+                                        __pgprot(pgprot_val(prot) | _PAGE_PSE)));
                         spin_unlock(&init_mm.page_table_lock);
                         last_map_addr = (address & PMD_MASK) + PMD_SIZE;
                         continue;
                 }
 
                 pte = alloc_low_page(&pte_phys);
-                last_map_addr = phys_pte_init(pte, address, end);
+                last_map_addr = phys_pte_init(pte, address, end, new_prot);
                 unmap_low_page(pte);
 
                 spin_lock(&init_mm.page_table_lock);
@@ -372,12 +394,12 @@ phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end,
 
 static unsigned long __meminit
 phys_pmd_update(pud_t *pud, unsigned long address, unsigned long end,
-                         unsigned long page_size_mask)
+                unsigned long page_size_mask, pgprot_t prot)
 {
         pmd_t *pmd = pmd_offset(pud, 0);
         unsigned long last_map_addr;
 
-        last_map_addr = phys_pmd_init(pmd, address, end, page_size_mask);
+        last_map_addr = phys_pmd_init(pmd, address, end, page_size_mask, prot);
         __flush_tlb_all();
         return last_map_addr;
 }
@@ -394,6 +416,7 @@ phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end,
                 unsigned long pmd_phys;
                 pud_t *pud = pud_page + pud_index(addr);
                 pmd_t *pmd;
+                pgprot_t prot = PAGE_KERNEL;
 
                 if (addr >= end)
                         break;
@@ -405,10 +428,26 @@ phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end,
                 }
 
                 if (pud_val(*pud)) {
-                        if (!pud_large(*pud))
+                        if (!pud_large(*pud)) {
                                 last_map_addr = phys_pmd_update(pud, addr, end,
-                                                         page_size_mask);
-                        continue;
+                                                         page_size_mask, prot);
+                                continue;
+                        }
+                        /*
+                         * If we are ok with PG_LEVEL_1G mapping, then we will
+                         * use the existing mapping.
+                         *
+                         * Otherwise, we will split the gbpage mapping but use
+                         * the same existing protection  bits except for large
+                         * page, so that we don't violate Intel's TLB
+                         * Application note (317080) which says, while changing
+                         * the page sizes, new and old translations should
+                         * not differ with respect to page frame and
+                         * attributes.
+                         */
+                        if (page_size_mask & (1 << PG_LEVEL_1G))
+                                continue;
+                        prot = pte_pgprot(pte_clrhuge(*(pte_t *)pud));
                 }
 
                 if (page_size_mask & (1<<PG_LEVEL_1G)) {
@@ -422,7 +461,8 @@ phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end,
                 }
 
                 pmd = alloc_low_page(&pmd_phys);
-                last_map_addr = phys_pmd_init(pmd, addr, end, page_size_mask);
+                last_map_addr = phys_pmd_init(pmd, addr, end, page_size_mask,
+                                              prot);
                 unmap_low_page(pmd);
 
                 spin_lock(&init_mm.page_table_lock);
@@ -430,6 +470,7 @@ phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end,
                 spin_unlock(&init_mm.page_table_lock);
         }
         __flush_tlb_all();
+
         update_page_count(PG_LEVEL_1G, pages);
 
         return last_map_addr;
@@ -446,13 +487,14 @@ phys_pud_update(pgd_t *pgd, unsigned long addr, unsigned long end,
         return phys_pud_init(pud, addr, end, page_size_mask);
 }
 
-static void __init find_early_table_space(unsigned long end)
+static void __init find_early_table_space(unsigned long end, int use_pse,
+                                          int use_gbpages)
 {
         unsigned long puds, pmds, ptes, tables, start;
 
         puds = (end + PUD_SIZE - 1) >> PUD_SHIFT;
         tables = roundup(puds * sizeof(pud_t), PAGE_SIZE);
-        if (direct_gbpages) {
+        if (use_gbpages) {
                 unsigned long extra;
                 extra = end - ((end>>PUD_SHIFT) << PUD_SHIFT);
                 pmds = (extra + PMD_SIZE - 1) >> PMD_SHIFT;
@@ -460,7 +502,7 @@ static void __init find_early_table_space(unsigned long end)
                 pmds = (end + PMD_SIZE - 1) >> PMD_SHIFT;
         tables += roundup(pmds * sizeof(pmd_t), PAGE_SIZE);
 
-        if (cpu_has_pse) {
+        if (use_pse) {
                 unsigned long extra;
                 extra = end - ((end>>PMD_SHIFT) << PMD_SHIFT);
                 ptes = (extra + PAGE_SIZE - 1) >> PAGE_SHIFT;
@@ -528,6 +570,7 @@ static unsigned long __init kernel_physical_mapping_init(unsigned long start,
                 pgd_populate(&init_mm, pgd, __va(pud_phys));
                 spin_unlock(&init_mm.page_table_lock);
         }
+        __flush_tlb_all();
 
         return last_map_addr;
 }
@@ -571,6 +614,7 @@ unsigned long __init_refok init_memory_mapping(unsigned long start,
 
         struct map_range mr[NR_RANGE_MR];
         int nr_range, i;
+        int use_pse, use_gbpages;
 
         printk(KERN_INFO "init_memory_mapping\n");
 
@@ -584,9 +628,21 @@ unsigned long __init_refok init_memory_mapping(unsigned long start,
         if (!after_bootmem)
                 init_gbpages();
 
-        if (direct_gbpages)
+#ifdef CONFIG_DEBUG_PAGEALLOC
+        /*
+         * For CONFIG_DEBUG_PAGEALLOC, identity mapping will use small pages.
+         * This will simplify cpa(), which otherwise needs to support splitting
+         * large pages into small in interrupt context, etc.
+         */
+        use_pse = use_gbpages = 0;
+#else
+        use_pse = cpu_has_pse;
+        use_gbpages = direct_gbpages;
+#endif
+
+        if (use_gbpages)
                 page_size_mask |= 1 << PG_LEVEL_1G;
-        if (cpu_has_pse)
+        if (use_pse)
                 page_size_mask |= 1 << PG_LEVEL_2M;
 
         memset(mr, 0, sizeof(mr));
@@ -647,7 +703,7 @@ unsigned long __init_refok init_memory_mapping(unsigned long start,
                          (mr[i].page_size_mask & (1<<PG_LEVEL_2M))?"2M":"4k"));
 
         if (!after_bootmem)
-                find_early_table_space(end);
+                find_early_table_space(end, use_pse, use_gbpages);
 
         for (i = 0; i < nr_range; i++)
                 last_map_addr = kernel_physical_mapping_init(
@@ -806,8 +862,6 @@ void __init mem_init(void)
                 reservedpages << (PAGE_SHIFT-10),
                 datasize >> 10,
                 initsize >> 10);
-
-        cpa_init();
 }
 
 void free_init_pages(char *what, unsigned long begin, unsigned long end)