| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Fri, 24 Aug 2018 03:07:46 +0000
From: Naoya Horiguchi <n-horiguchi@...jp.nec.com>
To: Mike Kravetz <mike.kravetz@...cle.com>
CC: "linux-mm@...ck.org" <linux-mm@...ck.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"Kirill A . Shutemov" <kirill.shutemov@...ux.intel.com>,
Jérôme Glisse <jglisse@...hat.com>,
Vlastimil Babka <vbabka@...e.cz>,
Davidlohr Bueso <dave@...olabs.net>,
Michal Hocko <mhocko@...nel.org>,
Andrew Morton <akpm@...ux-foundation.org>
Subject: Re: [PATCH v6 2/2] hugetlb: take PMD sharing into account when
flushing tlb/caches
On Thu, Aug 23, 2018 at 01:59:17PM -0700, Mike Kravetz wrote:
> 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.
s/PUD_SUZE/PUD_SIZE/
>
> Signed-off-by: Mike Kravetz <mike.kravetz@...cle.com>
Looks good to me.
Reviewed-by: Naoya Horiguchi <n-horiguchi@...jp.nec.com>
> ---
> mm/hugetlb.c | 53 +++++++++++++++++++++++++++++++++++++++++++---------
> 1 file changed, 44 insertions(+), 9 deletions(-)
>
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index a73c5728e961..082cddf46b4f 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -3333,8 +3333,8 @@ 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));
> @@ -3346,6 +3346,11 @@ 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) {
> @@ -3356,6 +3361,10 @@ 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;
> }
>
> @@ -3438,12 +3447,23 @@ 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);
> }
>
> /*
> @@ -4309,11 +4329,21 @@ 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;
> @@ -4324,6 +4354,7 @@ 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);
> @@ -4359,9 +4390,13 @@ 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.
> @@ -4369,7 +4404,7 @@ 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;
> }
> --
> 2.17.1
>
>
Powered by blists - more mailing lists