arch/powerpc/mm/pgtable.c

   1 /*
   2  * This file contains common routines for dealing with free of page tables
   3  * Along with common page table handling code
   4  *
   5  *  Derived from arch/powerpc/mm/tlb_64.c:
   6  *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
   7  *
   8  *  Modifications by Paul Mackerras (PowerMac) (paulus@cs.anu.edu.au)
   9  *  and Cort Dougan (PReP) (cort@cs.nmt.edu)
  10  *    Copyright (C) 1996 Paul Mackerras
  11  *
  12  *  Derived from "arch/i386/mm/init.c"
  13  *    Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
  14  *
  15  *  Dave Engebretsen <engebret@us.ibm.com>
  16  *      Rework for PPC64 port.
  17  *
  18  *  This program is free software; you can redistribute it and/or
  19  *  modify it under the terms of the GNU General Public License
  20  *  as published by the Free Software Foundation; either version
  21  *  2 of the License, or (at your option) any later version.
  22  */
  23
  24 #include <linux/kernel.h>
  25 #include <linux/gfp.h>
  26 #include <linux/mm.h>
  27 #include <linux/percpu.h>
  28 #include <linux/hardirq.h>
  29 #include <linux/hugetlb.h>
  30 #include <asm/pgalloc.h>
  31 #include <asm/tlbflush.h>
  32 #include <asm/tlb.h>
  33
  34 static inline int is_exec_fault(void)
  35 {
  36         return current->thread.regs && TRAP(current->thread.regs) == 0x400;
  37 }
  38
  39 /* We only try to do i/d cache coherency on stuff that looks like
  40  * reasonably "normal" PTEs. We currently require a PTE to be present
  41  * and we avoid _PAGE_SPECIAL and cache inhibited pte. We also only do that
  42  * on userspace PTEs
  43  */
  44 static inline int pte_looks_normal(pte_t pte)
  45 {
  46
  47 #if defined(CONFIG_PPC_BOOK3S_64)
  48         if ((pte_val(pte) & (_PAGE_PRESENT | _PAGE_SPECIAL)) == _PAGE_PRESENT) {
  49                 if (pte_ci(pte))
  50                         return 0;
  51                 if (pte_user(pte))
  52                         return 1;
  53         }
  54         return 0;
  55 #else
  56         return (pte_val(pte) &
  57                 (_PAGE_PRESENT | _PAGE_SPECIAL | _PAGE_NO_CACHE | _PAGE_USER)) ==
  58                 (_PAGE_PRESENT | _PAGE_USER);
  59 #endif
  60 }
  61
  62 static struct page *maybe_pte_to_page(pte_t pte)
  63 {
  64         unsigned long pfn = pte_pfn(pte);
  65         struct page *page;
  66
  67         if (unlikely(!pfn_valid(pfn)))
  68                 return NULL;
  69         page = pfn_to_page(pfn);
  70         if (PageReserved(page))
  71                 return NULL;
  72         return page;
  73 }
  74
  75 #if defined(CONFIG_PPC_STD_MMU) || _PAGE_EXEC == 0
  76
  77 /* Server-style MMU handles coherency when hashing if HW exec permission
  78  * is supposed per page (currently 64-bit only). If not, then, we always
  79  * flush the cache for valid PTEs in set_pte. Embedded CPU without HW exec
  80  * support falls into the same category.
  81  */
  82
  83 static pte_t set_pte_filter(pte_t pte)
  84 {
  85         if (radix_enabled())
  86                 return pte;
  87
  88         pte = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
  89         if (pte_looks_normal(pte) && !(cpu_has_feature(CPU_FTR_COHERENT_ICACHE) ||
  90                                        cpu_has_feature(CPU_FTR_NOEXECUTE))) {
  91                 struct page *pg = maybe_pte_to_page(pte);
  92                 if (!pg)
  93                         return pte;
  94                 if (!test_bit(PG_arch_1, &pg->flags)) {
  95                         flush_dcache_icache_page(pg);
  96                         set_bit(PG_arch_1, &pg->flags);
  97                 }
  98         }
  99         return pte;
 100 }
 101
 102 static pte_t set_access_flags_filter(pte_t pte, struct vm_area_struct *vma,
 103                                      int dirty)
 104 {
 105         return pte;
 106 }
 107
 108 #else /* defined(CONFIG_PPC_STD_MMU) || _PAGE_EXEC == 0 */
 109
 110 /* Embedded type MMU with HW exec support. This is a bit more complicated
 111  * as we don't have two bits to spare for _PAGE_EXEC and _PAGE_HWEXEC so
 112  * instead we "filter out" the exec permission for non clean pages.
 113  */
 114 static pte_t set_pte_filter(pte_t pte)
 115 {
 116         struct page *pg;
 117
 118         /* No exec permission in the first place, move on */
 119         if (!(pte_val(pte) & _PAGE_EXEC) || !pte_looks_normal(pte))
 120                 return pte;
 121
 122         /* If you set _PAGE_EXEC on weird pages you're on your own */
 123         pg = maybe_pte_to_page(pte);
 124         if (unlikely(!pg))
 125                 return pte;
 126
 127         /* If the page clean, we move on */
 128         if (test_bit(PG_arch_1, &pg->flags))
 129                 return pte;
 130
 131         /* If it's an exec fault, we flush the cache and make it clean */
 132         if (is_exec_fault()) {
 133                 flush_dcache_icache_page(pg);
 134                 set_bit(PG_arch_1, &pg->flags);
 135                 return pte;
 136         }
 137
 138         /* Else, we filter out _PAGE_EXEC */
 139         return __pte(pte_val(pte) & ~_PAGE_EXEC);
 140 }
 141
 142 static pte_t set_access_flags_filter(pte_t pte, struct vm_area_struct *vma,
 143                                      int dirty)
 144 {
 145         struct page *pg;
 146
 147         /* So here, we only care about exec faults, as we use them
 148          * to recover lost _PAGE_EXEC and perform I$/D$ coherency
 149          * if necessary. Also if _PAGE_EXEC is already set, same deal,
 150          * we just bail out
 151          */
 152         if (dirty || (pte_val(pte) & _PAGE_EXEC) || !is_exec_fault())
 153                 return pte;
 154
 155 #ifdef CONFIG_DEBUG_VM
 156         /* So this is an exec fault, _PAGE_EXEC is not set. If it was
 157          * an error we would have bailed out earlier in do_page_fault()
 158          * but let's make sure of it
 159          */
 160         if (WARN_ON(!(vma->vm_flags & VM_EXEC)))
 161                 return pte;
 162 #endif /* CONFIG_DEBUG_VM */
 163
 164         /* If you set _PAGE_EXEC on weird pages you're on your own */
 165         pg = maybe_pte_to_page(pte);
 166         if (unlikely(!pg))
 167                 goto bail;
 168
 169         /* If the page is already clean, we move on */
 170         if (test_bit(PG_arch_1, &pg->flags))
 171                 goto bail;
 172
 173         /* Clean the page and set PG_arch_1 */
 174         flush_dcache_icache_page(pg);
 175         set_bit(PG_arch_1, &pg->flags);
 176
 177  bail:
 178         return __pte(pte_val(pte) | _PAGE_EXEC);
 179 }
 180
 181 #endif /* !(defined(CONFIG_PPC_STD_MMU) || _PAGE_EXEC == 0) */
 182
 183 /*
 184  * set_pte stores a linux PTE into the linux page table.
 185  */
 186 void set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep,
 187                 pte_t pte)
 188 {
 189         /*
 190          * When handling numa faults, we already have the pte marked
 191          * _PAGE_PRESENT, but we can be sure that it is not in hpte.
 192          * Hence we can use set_pte_at for them.
 193          */
 194         VM_WARN_ON(pte_present(*ptep) && !pte_protnone(*ptep));
 195
 196         /*
 197          * Add the pte bit when tryint set a pte
 198          */
 199         pte = __pte(pte_val(pte) | _PAGE_PTE);
 200
 201         /* Note: mm->context.id might not yet have been assigned as
 202          * this context might not have been activated yet when this
 203          * is called.
 204          */
 205         pte = set_pte_filter(pte);
 206
 207         /* Perform the setting of the PTE */
 208         __set_pte_at(mm, addr, ptep, pte, 0);
 209 }
 210
 211 /*
 212  * This is called when relaxing access to a PTE. It's also called in the page
 213  * fault path when we don't hit any of the major fault cases, ie, a minor
 214  * update of _PAGE_ACCESSED, _PAGE_DIRTY, etc... The generic code will have
 215  * handled those two for us, we additionally deal with missing execute
 216  * permission here on some processors
 217  */
 218 int ptep_set_access_flags(struct vm_area_struct *vma, unsigned long address,
 219                           pte_t *ptep, pte_t entry, int dirty)
 220 {
 221         int changed;
 222         entry = set_access_flags_filter(entry, vma, dirty);
 223         changed = !pte_same(*(ptep), entry);
 224         if (changed) {
 225                 if (!is_vm_hugetlb_page(vma))
 226                         assert_pte_locked(vma->vm_mm, address);
 227                 __ptep_set_access_flags(ptep, entry);
 228                 flush_tlb_page_nohash(vma, address);
 229         }
 230         return changed;
 231 }
 232
 233 #ifdef CONFIG_DEBUG_VM
 234 void assert_pte_locked(struct mm_struct *mm, unsigned long addr)
 235 {
 236         pgd_t *pgd;
 237         pud_t *pud;
 238         pmd_t *pmd;
 239
 240         if (mm == &init_mm)
 241                 return;
 242         pgd = mm->pgd + pgd_index(addr);
 243         BUG_ON(pgd_none(*pgd));
 244         pud = pud_offset(pgd, addr);
 245         BUG_ON(pud_none(*pud));
 246         pmd = pmd_offset(pud, addr);
 247         /*
 248          * khugepaged to collapse normal pages to hugepage, first set
 249          * pmd to none to force page fault/gup to take mmap_sem. After
 250          * pmd is set to none, we do a pte_clear which does this assertion
 251          * so if we find pmd none, return.
 252          */
 253         if (pmd_none(*pmd))
 254                 return;
 255         BUG_ON(!pmd_present(*pmd));
 256         assert_spin_locked(pte_lockptr(mm, pmd));
 257 }
 258 #endif /* CONFIG_DEBUG_VM */
 259
 260 unsigned long vmalloc_to_phys(void *va)
 261 {
 262         unsigned long pfn = vmalloc_to_pfn(va);
 263
 264         BUG_ON(!pfn);
 265         return __pa(pfn_to_kaddr(pfn)) + offset_in_page(va);
 266 }
 267 EXPORT_SYMBOL_GPL(vmalloc_to_phys);