hugetlb: take PMD sharing into account when flushing tlb/caches

When fixing an issue with PMD sharing and migration, it was discovered via
code inspection that other callers of huge_pmd_unshare potentially have an
issue with cache and tlb flushing.

Use the routine adjust_range_if_pmd_sharing_possible() to calculate worst
case ranges for mmu notifiers.  Ensure that this range is flushed if
huge_pmd_unshare succeeds and unmaps a PUD_SUZE area.

Link: http://lkml.kernel.org/r/20180823205917.16297-3-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index b903d746e1320209e15c22e681f606982f73af50..5c390f5a5207b5c0b4d1d55524d2f655c7edcae1 100644 (file)
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -3326,8 +3326,8 @@ void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
        struct page *page;
        struct hstate *h = hstate_vma(vma);
        unsigned long sz = huge_page_size(h);
-       const unsigned long mmun_start = start; /* For mmu_notifiers */
-       const unsigned long mmun_end   = end;   /* For mmu_notifiers */
+       unsigned long mmun_start = start;       /* For mmu_notifiers */
+       unsigned long mmun_end   = end;         /* For mmu_notifiers */
 
        WARN_ON(!is_vm_hugetlb_page(vma));
        BUG_ON(start & ~huge_page_mask(h));
@@ -3339,6 +3339,11 @@ void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
         */
        tlb_remove_check_page_size_change(tlb, sz);
        tlb_start_vma(tlb, vma);
+
+       /*
+        * If sharing possible, alert mmu notifiers of worst case.
+        */
+       adjust_range_if_pmd_sharing_possible(vma, &mmun_start, &mmun_end);
        mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
        address = start;
        for (; address < end; address += sz) {
@@ -3349,6 +3354,10 @@ void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
                ptl = huge_pte_lock(h, mm, ptep);
                if (huge_pmd_unshare(mm, &address, ptep)) {
                        spin_unlock(ptl);
+                       /*
+                        * We just unmapped a page of PMDs by clearing a PUD.
+                        * The caller's TLB flush range should cover this area.
+                        */
                        continue;
                }
 
@@ -3431,12 +3440,23 @@ void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
 {
        struct mm_struct *mm;
        struct mmu_gather tlb;
+       unsigned long tlb_start = start;
+       unsigned long tlb_end = end;
+
+       /*
+        * If shared PMDs were possibly used within this vma range, adjust
+        * start/end for worst case tlb flushing.
+        * Note that we can not be sure if PMDs are shared until we try to
+        * unmap pages.  However, we want to make sure TLB flushing covers
+        * the largest possible range.
+        */
+       adjust_range_if_pmd_sharing_possible(vma, &tlb_start, &tlb_end);
 
        mm = vma->vm_mm;
 
-       tlb_gather_mmu(&tlb, mm, start, end);
+       tlb_gather_mmu(&tlb, mm, tlb_start, tlb_end);
        __unmap_hugepage_range(&tlb, vma, start, end, ref_page);
-       tlb_finish_mmu(&tlb, start, end);
+       tlb_finish_mmu(&tlb, tlb_start, tlb_end);
 }
 
 /*
@@ -4298,11 +4318,21 @@ unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
        pte_t pte;
        struct hstate *h = hstate_vma(vma);
        unsigned long pages = 0;
+       unsigned long f_start = start;
+       unsigned long f_end = end;
+       bool shared_pmd = false;
+
+       /*
+        * In the case of shared PMDs, the area to flush could be beyond
+        * start/end.  Set f_start/f_end to cover the maximum possible
+        * range if PMD sharing is possible.
+        */
+       adjust_range_if_pmd_sharing_possible(vma, &f_start, &f_end);
 
        BUG_ON(address >= end);
-       flush_cache_range(vma, address, end);
+       flush_cache_range(vma, f_start, f_end);
 
-       mmu_notifier_invalidate_range_start(mm, start, end);
+       mmu_notifier_invalidate_range_start(mm, f_start, f_end);
        i_mmap_lock_write(vma->vm_file->f_mapping);
        for (; address < end; address += huge_page_size(h)) {
                spinlock_t *ptl;
@@ -4313,6 +4343,7 @@ unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
                if (huge_pmd_unshare(mm, &address, ptep)) {
                        pages++;
                        spin_unlock(ptl);
+                       shared_pmd = true;
                        continue;
                }
                pte = huge_ptep_get(ptep);
@@ -4348,9 +4379,13 @@ unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
         * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
         * may have cleared our pud entry and done put_page on the page table:
         * once we release i_mmap_rwsem, another task can do the final put_page
-        * and that page table be reused and filled with junk.
+        * and that page table be reused and filled with junk.  If we actually
+        * did unshare a page of pmds, flush the range corresponding to the pud.
         */
-       flush_hugetlb_tlb_range(vma, start, end);
+       if (shared_pmd)
+               flush_hugetlb_tlb_range(vma, f_start, f_end);
+       else
+               flush_hugetlb_tlb_range(vma, start, end);
        /*
         * No need to call mmu_notifier_invalidate_range() we are downgrading
         * page table protection not changing it to point to a new page.
@@ -4358,7 +4393,7 @@ unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
         * See Documentation/vm/mmu_notifier.rst
         */
        i_mmap_unlock_write(vma->vm_file->f_mapping);
-       mmu_notifier_invalidate_range_end(mm, start, end);
+       mmu_notifier_invalidate_range_end(mm, f_start, f_end);
 
        return pages << h->order;
 }