[PATCH] userfaultfd_release: always remove uffd flags and clear vm_userfaultfd_ctx

[PATCH] userfaultfd_release: always remove uffd flags and clear vm_userfaultfd_ctx

[Oleg Nesterov]

userfaultfd_release() should clear vm_flags/vm_userfaultfd_ctx even
if mm->core_state != NULL.

Otherwise a page fault can see userfaultfd_missing() == T and use an
already freed userfaultfd_ctx.

Reported-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Fixes: 04f5866e41fb ("coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping")
Cc: stable@vger.kernel.org
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
---
 fs/userfaultfd.c | 25 +++++++++++++------------
 1 file changed, 13 insertions(+), 12 deletions(-)

diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c
index ccbdbd6..fe6d804 100644
--- a/fs/userfaultfd.c
+++ b/fs/userfaultfd.c
@@ -880,6 +880,7 @@ static int userfaultfd_release(struct inode *inode, struct file *file)
 	/* len == 0 means wake all */
 	struct userfaultfd_wake_range range = { .len = 0, };
 	unsigned long new_flags;
+	bool still_valid;
 
 	WRITE_ONCE(ctx->released, true);
 
@@ -895,8 +896,7 @@ static int userfaultfd_release(struct inode *inode, struct file *file)
 	 * taking the mmap_sem for writing.
 	 */
 	down_write(&mm->mmap_sem);
-	if (!mmget_still_valid(mm))
-		goto skip_mm;
+	still_valid = mmget_still_valid(mm);
 	prev = NULL;
 	for (vma = mm->mmap; vma; vma = vma->vm_next) {
 		cond_resched();
@@ -907,19 +907,20 @@ static int userfaultfd_release(struct inode *inode, struct file *file)
 			continue;
 		}
 		new_flags = vma->vm_flags & ~(VM_UFFD_MISSING | VM_UFFD_WP);
-		prev = vma_merge(mm, prev, vma->vm_start, vma->vm_end,
-				 new_flags, vma->anon_vma,
-				 vma->vm_file, vma->vm_pgoff,
-				 vma_policy(vma),
-				 NULL_VM_UFFD_CTX);
-		if (prev)
-			vma = prev;
-		else
-			prev = vma;
+		if (still_valid) {
+			prev = vma_merge(mm, prev, vma->vm_start, vma->vm_end,
+					 new_flags, vma->anon_vma,
+					 vma->vm_file, vma->vm_pgoff,
+					 vma_policy(vma),
+					 NULL_VM_UFFD_CTX);
+			if (prev)
+				vma = prev;
+			else
+				prev = vma;
+		}
 		vma->vm_flags = new_flags;
 		vma->vm_userfaultfd_ctx = NULL_VM_UFFD_CTX;
 	}
-skip_mm:
 	up_write(&mm->mmap_sem);
 	mmput(mm);
 wakeup:
-- 
2.5.0

Re: [PATCH] userfaultfd_release: always remove uffd flags and clear vm_userfaultfd_ctx

[Andrea Arcangeli]

On Tue, Aug 20, 2019 at 06:02:38PM +0200, Oleg Nesterov wrote:
> userfaultfd_release() should clear vm_flags/vm_userfaultfd_ctx even
> if mm->core_state != NULL.
> 
> Otherwise a page fault can see userfaultfd_missing() == T and use an
> already freed userfaultfd_ctx.
> 
> Reported-by: Kefeng Wang <wangkefeng.wang@huawei.com>
> Fixes: 04f5866e41fb ("coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping")
> Cc: stable@vger.kernel.org
> Signed-off-by: Oleg Nesterov <oleg@redhat.com>
> ---
>  fs/userfaultfd.c | 25 +++++++++++++------------
>  1 file changed, 13 insertions(+), 12 deletions(-)

Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>

Thanks,
Andrea

Re: [PATCH] userfaultfd_release: always remove uffd flags and clear vm_userfaultfd_ctx

[Kefeng Wang]

On 2019/8/21 0:02, Oleg Nesterov wrote:
> userfaultfd_release() should clear vm_flags/vm_userfaultfd_ctx even
> if mm->core_state != NULL.
>
> Otherwise a page fault can see userfaultfd_missing() == T and use an
> already freed userfaultfd_ctx.
>
> Reported-by: Kefeng Wang <wangkefeng.wang@huawei.com>
> Fixes: 04f5866e41fb ("coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping")
> Cc: stable@vger.kernel.org
> Signed-off-by: Oleg Nesterov <oleg@redhat.com>
> ---
>  fs/userfaultfd.c | 25 +++++++++++++------------
>  1 file changed, 13 insertions(+), 12 deletions(-)

Tested on lts4.4 and 5.3-rc4, Thanks.