[PATCH] hugetlbfs: Fix off-by-one error in hugetlb_vmdelete_list()

[PATCH] hugetlbfs: Fix off-by-one error in hugetlb_vmdelete_list()

[Sean Christopherson]

Pass "end - 1" instead of "end" when walking the interval tree in
hugetlb_vmdelete_list() to fix an inclusive vs. exclusive bug.  The two
callers that pass a non-zero "end" treat it as exclusive, whereas the
interval tree iterator expects an inclusive "last".  E.g. punching a hole
in a file that precisely matches the size of a single hugepage, with a
vma starting right on the boundary, will result in unmap_hugepage_range()
being called twice, with the second call having start==end.

The off-by-one error doesn't cause functional problems as
__unmap_hugepage_range() turns into a massive nop due to short-circuiting
its for-loop on "address < end".  But, the mmu_notifier invocations to
invalid_range_{start,end}() are passed a bogus zero-sized range, which
may be unexpected behavior for secondary MMUs.

The bug was exposed by commit ed922739c919 ("KVM: Use interval tree to do
fast hva lookup in memslots"), currently queued in the KVM tree for 5.17,
which added a WARN to detect ranges with start==end.

Reported-by: syzbot+4e697fe80a31aa7efe21@syzkaller.appspotmail.com
Fixes: 1bfad99ab425 ("hugetlbfs: hugetlb_vmtruncate_list() needs to take a range to delete")
Cc: kvm@vger.kernel.org
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
---

Not sure if this should go to stable@.  It's mostly harmless, and likely
nothing more than a minor performance blip when it's not harmless.

 fs/hugetlbfs/inode.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
index 49d2e686be74..a7c6c7498be0 100644
--- a/fs/hugetlbfs/inode.c
+++ b/fs/hugetlbfs/inode.c
@@ -409,10 +409,11 @@ hugetlb_vmdelete_list(struct rb_root_cached *root, pgoff_t start, pgoff_t end)
 	struct vm_area_struct *vma;
 
 	/*
-	 * end == 0 indicates that the entire range after
-	 * start should be unmapped.
+	 * end == 0 indicates that the entire range after start should be
+	 * unmapped.  Note, end is exclusive, whereas the interval tree takes
+	 * an inclusive "last".
 	 */
-	vma_interval_tree_foreach(vma, root, start, end ? end : ULONG_MAX) {
+	vma_interval_tree_foreach(vma, root, start, end ? end - 1 : ULONG_MAX) {
 		unsigned long v_offset;
 		unsigned long v_end;
 
-- 
2.34.1.448.ga2b2bfdf31-goog

Re: [PATCH] hugetlbfs: Fix off-by-one error in hugetlb_vmdelete_list()

[Mike Kravetz]

+Cc Andrew if he wants to take it though his tree.

On 12/28/21 15:42, Sean Christopherson wrote:
> Pass "end - 1" instead of "end" when walking the interval tree in
> hugetlb_vmdelete_list() to fix an inclusive vs. exclusive bug.  The two
> callers that pass a non-zero "end" treat it as exclusive, whereas the
> interval tree iterator expects an inclusive "last".  E.g. punching a hole
> in a file that precisely matches the size of a single hugepage, with a
> vma starting right on the boundary, will result in unmap_hugepage_range()
> being called twice, with the second call having start==end.
> 
> The off-by-one error doesn't cause functional problems as
> __unmap_hugepage_range() turns into a massive nop due to short-circuiting
> its for-loop on "address < end".  But, the mmu_notifier invocations to
> invalid_range_{start,end}() are passed a bogus zero-sized range, which
> may be unexpected behavior for secondary MMUs.
> 
> The bug was exposed by commit ed922739c919 ("KVM: Use interval tree to do
> fast hva lookup in memslots"), currently queued in the KVM tree for 5.17,
> which added a WARN to detect ranges with start==end.
> 
> Reported-by: syzbot+4e697fe80a31aa7efe21@syzkaller.appspotmail.com
> Fixes: 1bfad99ab425 ("hugetlbfs: hugetlb_vmtruncate_list() needs to take a range to delete")
> Cc: kvm@vger.kernel.org
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Signed-off-by: Sean Christopherson <seanjc@google.com>

Thanks Sean!

Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>

> ---
> 
> Not sure if this should go to stable@.  It's mostly harmless, and likely
> nothing more than a minor performance blip when it's not harmless.

I am also unsure about the need to send to stable.  It is possible automation
will pick it up and make that decision for us.
-- 
Mike Kravetz

Re: [PATCH] hugetlbfs: Fix off-by-one error in hugetlb_vmdelete_list()

[Andrew Morton]

On Tue, 28 Dec 2021 19:52:37 -0800 Mike Kravetz <mike.kravetz@oracle.com> wrote:

> +Cc Andrew if he wants to take it though his tree.

Sure.

> > 
> > Not sure if this should go to stable@.  It's mostly harmless, and likely
> > nothing more than a minor performance blip when it's not harmless.
> 
> I am also unsure about the need to send to stable.  It is possible automation
> will pick it up and make that decision for us.

Automation shouldn't do that for mm/ patches because we asked.  But fs/
material might sneak through.  But it does appear that -stable
won't need this.