thp: remove assumptions on pgtable_t type
authorGerald Schaefer <gerald.schaefer@de.ibm.com>
Mon, 8 Oct 2012 23:30:07 +0000 (16:30 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 9 Oct 2012 07:22:29 +0000 (16:22 +0900)
The thp page table pre-allocation code currently assumes that pgtable_t is
of type "struct page *".  This may not be true for all architectures, so
this patch removes that assumption by replacing the functions
prepare_pmd_huge_pte() and get_pmd_huge_pte() with two new functions that
can be defined architecture-specific.

It also removes two VM_BUG_ON checks for page_count() and page_mapcount()
operating on a pgtable_t.  Apart from the VM_BUG_ON removal, there will be
no functional change introduced by this patch.

Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/asm-generic/pgtable.h
include/linux/huge_mm.h
mm/huge_memory.c
mm/pgtable-generic.c

index c9a612069c8ebd2726ce0d819156c2ec7bc1297e..044939f21c166c980ce309fc54fc7b167e2dda63 100644 (file)
@@ -162,6 +162,14 @@ extern void pmdp_splitting_flush(struct vm_area_struct *vma,
                                 unsigned long address, pmd_t *pmdp);
 #endif
 
+#ifndef __HAVE_ARCH_PGTABLE_DEPOSIT
+extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable);
+#endif
+
+#ifndef __HAVE_ARCH_PGTABLE_WITHDRAW
+extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm);
+#endif
+
 #ifndef __HAVE_ARCH_PTE_SAME
 static inline int pte_same(pte_t pte_a, pte_t pte_b)
 {
index 4c59b11311870e74ebbeb110f77e2017be97c3bb..6ab47af5a849b395440b4aa3a2ba88e641f13a0c 100644 (file)
@@ -11,7 +11,6 @@ extern int copy_huge_pmd(struct mm_struct *dst_mm, struct mm_struct *src_mm,
 extern int do_huge_pmd_wp_page(struct mm_struct *mm, struct vm_area_struct *vma,
                               unsigned long address, pmd_t *pmd,
                               pmd_t orig_pmd);
-extern pgtable_t get_pmd_huge_pte(struct mm_struct *mm);
 extern struct page *follow_trans_huge_pmd(struct mm_struct *mm,
                                          unsigned long addr,
                                          pmd_t *pmd,
index e19cc426c522e32b19922532d2380b9d9c37c7a9..9ea6d1953765e5a19dd237175eacd0e2e70e5565 100644 (file)
@@ -598,19 +598,6 @@ out:
 }
 __setup("transparent_hugepage=", setup_transparent_hugepage);
 
-static void prepare_pmd_huge_pte(pgtable_t pgtable,
-                                struct mm_struct *mm)
-{
-       assert_spin_locked(&mm->page_table_lock);
-
-       /* FIFO */
-       if (!mm->pmd_huge_pte)
-               INIT_LIST_HEAD(&pgtable->lru);
-       else
-               list_add(&pgtable->lru, &mm->pmd_huge_pte->lru);
-       mm->pmd_huge_pte = pgtable;
-}
-
 static inline pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma)
 {
        if (likely(vma->vm_flags & VM_WRITE))
@@ -652,7 +639,7 @@ static int __do_huge_pmd_anonymous_page(struct mm_struct *mm,
                 */
                page_add_new_anon_rmap(page, vma, haddr);
                set_pmd_at(mm, haddr, pmd, entry);
-               prepare_pmd_huge_pte(pgtable, mm);
+               pgtable_trans_huge_deposit(mm, pgtable);
                add_mm_counter(mm, MM_ANONPAGES, HPAGE_PMD_NR);
                mm->nr_ptes++;
                spin_unlock(&mm->page_table_lock);
@@ -778,7 +765,7 @@ int copy_huge_pmd(struct mm_struct *dst_mm, struct mm_struct *src_mm,
        pmdp_set_wrprotect(src_mm, addr, src_pmd);
        pmd = pmd_mkold(pmd_wrprotect(pmd));
        set_pmd_at(dst_mm, addr, dst_pmd, pmd);
-       prepare_pmd_huge_pte(pgtable, dst_mm);
+       pgtable_trans_huge_deposit(dst_mm, pgtable);
        dst_mm->nr_ptes++;
 
        ret = 0;
@@ -789,25 +776,6 @@ out:
        return ret;
 }
 
-/* no "address" argument so destroys page coloring of some arch */
-pgtable_t get_pmd_huge_pte(struct mm_struct *mm)
-{
-       pgtable_t pgtable;
-
-       assert_spin_locked(&mm->page_table_lock);
-
-       /* FIFO */
-       pgtable = mm->pmd_huge_pte;
-       if (list_empty(&pgtable->lru))
-               mm->pmd_huge_pte = NULL;
-       else {
-               mm->pmd_huge_pte = list_entry(pgtable->lru.next,
-                                             struct page, lru);
-               list_del(&pgtable->lru);
-       }
-       return pgtable;
-}
-
 static int do_huge_pmd_wp_page_fallback(struct mm_struct *mm,
                                        struct vm_area_struct *vma,
                                        unsigned long address,
@@ -863,7 +831,7 @@ static int do_huge_pmd_wp_page_fallback(struct mm_struct *mm,
        pmdp_clear_flush_notify(vma, haddr, pmd);
        /* leave pmd empty until pte is filled */
 
-       pgtable = get_pmd_huge_pte(mm);
+       pgtable = pgtable_trans_huge_withdraw(mm);
        pmd_populate(mm, &_pmd, pgtable);
 
        for (i = 0; i < HPAGE_PMD_NR; i++, haddr += PAGE_SIZE) {
@@ -1028,7 +996,7 @@ int zap_huge_pmd(struct mmu_gather *tlb, struct vm_area_struct *vma,
        if (__pmd_trans_huge_lock(pmd, vma) == 1) {
                struct page *page;
                pgtable_t pgtable;
-               pgtable = get_pmd_huge_pte(tlb->mm);
+               pgtable = pgtable_trans_huge_withdraw(tlb->mm);
                page = pmd_page(*pmd);
                pmd_clear(pmd);
                tlb_remove_pmd_tlb_entry(tlb, pmd, addr);
@@ -1345,11 +1313,11 @@ static int __split_huge_page_map(struct page *page,
        pmd = page_check_address_pmd(page, mm, address,
                                     PAGE_CHECK_ADDRESS_PMD_SPLITTING_FLAG);
        if (pmd) {
-               pgtable = get_pmd_huge_pte(mm);
+               pgtable = pgtable_trans_huge_withdraw(mm);
                pmd_populate(mm, &_pmd, pgtable);
 
-               for (i = 0, haddr = address; i < HPAGE_PMD_NR;
-                    i++, haddr += PAGE_SIZE) {
+               haddr = address;
+               for (i = 0; i < HPAGE_PMD_NR; i++, haddr += PAGE_SIZE) {
                        pte_t *pte, entry;
                        BUG_ON(PageCompound(page+i));
                        entry = mk_pte(page + i, vma->vm_page_prot);
@@ -2017,8 +1985,6 @@ static void collapse_huge_page(struct mm_struct *mm,
        pte_unmap(pte);
        __SetPageUptodate(new_page);
        pgtable = pmd_pgtable(_pmd);
-       VM_BUG_ON(page_count(pgtable) != 1);
-       VM_BUG_ON(page_mapcount(pgtable) != 0);
 
        _pmd = mk_pmd(new_page, vma->vm_page_prot);
        _pmd = maybe_pmd_mkwrite(pmd_mkdirty(_pmd), vma);
@@ -2036,7 +2002,7 @@ static void collapse_huge_page(struct mm_struct *mm,
        page_add_new_anon_rmap(new_page, vma, address);
        set_pmd_at(mm, address, pmd, _pmd);
        update_mmu_cache(vma, address, _pmd);
-       prepare_pmd_huge_pte(pgtable, mm);
+       pgtable_trans_huge_deposit(mm, pgtable);
        spin_unlock(&mm->page_table_lock);
 
        *hpage = NULL;
index 74c0ddaa6fa0df019c590994fa79b136dd1b4abd..29867e083d37277a10327afaa4eccb53df37ca36 100644 (file)
@@ -120,3 +120,42 @@ void pmdp_splitting_flush(struct vm_area_struct *vma, unsigned long address,
 }
 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
 #endif
+
+#ifndef __HAVE_ARCH_PGTABLE_DEPOSIT
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable)
+{
+       assert_spin_locked(&mm->page_table_lock);
+
+       /* FIFO */
+       if (!mm->pmd_huge_pte)
+               INIT_LIST_HEAD(&pgtable->lru);
+       else
+               list_add(&pgtable->lru, &mm->pmd_huge_pte->lru);
+       mm->pmd_huge_pte = pgtable;
+}
+#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
+#endif
+
+#ifndef __HAVE_ARCH_PGTABLE_WITHDRAW
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+/* no "address" argument so destroys page coloring of some arch */
+pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm)
+{
+       pgtable_t pgtable;
+
+       assert_spin_locked(&mm->page_table_lock);
+
+       /* FIFO */
+       pgtable = mm->pmd_huge_pte;
+       if (list_empty(&pgtable->lru))
+               mm->pmd_huge_pte = NULL;
+       else {
+               mm->pmd_huge_pte = list_entry(pgtable->lru.next,
+                                             struct page, lru);
+               list_del(&pgtable->lru);
+       }
+       return pgtable;
+}
+#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
+#endif