mm/autonuma: let architecture override how the write bit should be stashed in a protn...
authorAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Fri, 24 Feb 2017 22:59:16 +0000 (14:59 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 25 Feb 2017 01:46:56 +0000 (17:46 -0800)
Patch series "Numabalancing preserve write fix", v2.

This patch series address an issue w.r.t THP migration and autonuma
preserve write feature.  migrate_misplaced_transhuge_page() cannot deal
with concurrent modification of the page.  It does a page copy without
following the migration pte sequence.  IIUC, this was done to keep the
migration simpler and at the time of implemenation we didn't had THP
page cache which would have required a more elaborate migration scheme.
That means thp autonuma migration expect the protnone with saved write
to be done such that both kernel and user cannot update the page
content.  This patch series enables archs like ppc64 to do that.  We are
good with the hash translation mode with the current code, because we
never create a hardware page table entry for a protnone pte.

This patch (of 2):

Autonuma preserves the write permission across numa fault to avoid
taking a writefault after a numa fault (Commit: b191f9b106ea " mm: numa:
preserve PTE write permissions across a NUMA hinting fault").
Architecture can implement protnone in different ways and some may
choose to implement that by clearing Read/ Write/Exec bit of pte.
Setting the write bit on such pte can result in wrong behaviour.  Fix
this up by allowing arch to override how to save the write bit on a
protnone pte.

[aneesh.kumar@linux.vnet.ibm.com: don't mark pte saved write in case of dirty_accountable]
Link: http://lkml.kernel.org/r/1487942884-16517-1-git-send-email-aneesh.kumar@linux.vnet.ibm.com
[aneesh.kumar@linux.vnet.ibm.com: v3]
Link: http://lkml.kernel.org/r/1487498625-10891-2-git-send-email-aneesh.kumar@linux.vnet.ibm.com
Link: http://lkml.kernel.org/r/1487050314-3892-2-git-send-email-aneesh.kumar@linux.vnet.ibm.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Acked-by: Michael Neuling <mikey@neuling.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Michael Ellerman <michaele@au1.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
mm/huge_memory.c
mm/memory.c
mm/mprotect.c

index a0aba0f9c57bee8217c4bc329b8c1fd7711db774..6265411eb2ede39aa1710b4401ff250a411cb700 100644 (file)
@@ -225,6 +225,22 @@ static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addres
 }
 #endif
 
+#ifndef pte_savedwrite
+#define pte_savedwrite pte_write
+#endif
+
+#ifndef pte_mk_savedwrite
+#define pte_mk_savedwrite pte_mkwrite
+#endif
+
+#ifndef pmd_savedwrite
+#define pmd_savedwrite pmd_write
+#endif
+
+#ifndef pmd_mk_savedwrite
+#define pmd_mk_savedwrite pmd_mkwrite
+#endif
+
 #ifndef __HAVE_ARCH_PMDP_SET_WRPROTECT
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
index efddd02141a8f67c972c436a32721ecf1392b74a..92c2ee2dfcf8072688f4e328181df2b6f489ee44 100644 (file)
@@ -1426,7 +1426,7 @@ int do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t pmd)
        }
 
        /* See similar comment in do_numa_page for explanation */
-       if (!pmd_write(pmd))
+       if (!pmd_savedwrite(pmd))
                flags |= TNF_NO_GROUP;
 
        /*
@@ -1489,7 +1489,7 @@ int do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t pmd)
        goto out;
 clear_pmdnuma:
        BUG_ON(!PageLocked(page));
-       was_writable = pmd_write(pmd);
+       was_writable = pmd_savedwrite(pmd);
        pmd = pmd_modify(pmd, vma->vm_page_prot);
        pmd = pmd_mkyoung(pmd);
        if (was_writable)
@@ -1744,7 +1744,7 @@ int change_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd,
                        entry = pmdp_huge_get_and_clear_notify(mm, addr, pmd);
                        entry = pmd_modify(entry, newprot);
                        if (preserve_write)
-                               entry = pmd_mkwrite(entry);
+                               entry = pmd_mk_savedwrite(entry);
                        ret = HPAGE_PMD_NR;
                        set_pmd_at(mm, addr, pmd, entry);
                        BUG_ON(vma_is_anonymous(vma) && !preserve_write &&
index 0c759ba122e12ec960e125023b90ec10ae771059..14fc0b40f0bb6cf3ee50cfff8e7db865ad442cdd 100644 (file)
@@ -3401,7 +3401,7 @@ static int do_numa_page(struct vm_fault *vmf)
        int target_nid;
        bool migrated = false;
        pte_t pte;
-       bool was_writable = pte_write(vmf->orig_pte);
+       bool was_writable = pte_savedwrite(vmf->orig_pte);
        int flags = 0;
 
        /*
index a45b4dc6a7f5d7577b56193d2dd2115bee2a7d59..848e946b08e58e31bf6482bd091338a43bb66fe1 100644 (file)
@@ -99,7 +99,7 @@ static unsigned long change_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
                        ptent = ptep_modify_prot_start(mm, addr, pte);
                        ptent = pte_modify(ptent, newprot);
                        if (preserve_write)
-                               ptent = pte_mkwrite(ptent);
+                               ptent = pte_mk_savedwrite(ptent);
 
                        /* Avoid taking write faults for known dirty pages */
                        if (dirty_accountable && pte_dirty(ptent) &&