Re: KASAN: use-after-free Read in filemap_fault

["Kirill A. Shutemov" <kirill@shutemov.name>]

On Fri, Dec 28, 2018 at 01:09:38PM -0800, Andrew Morton wrote:
> On Fri, 28 Dec 2018 12:51:04 -0800 syzbot <syzbot+b437b5a429d680cf2217@syzkaller.appspotmail.com> wrote:
> 
> > Hello,
> > 
> > syzbot found the following crash on:
> 
> uh-oh.  Josef, could you please take a look?
> 
> :	page = find_get_page(mapping, offset);
> : 	if (likely(page) && !(vmf->flags & FAULT_FLAG_TRIED)) {
> : 		/*
> : 		 * We found the page, so try async readahead before
> : 		 * waiting for the lock.
> : 		 */
> : 		fpin = do_async_mmap_readahead(vmf, page);
> : 	} else if (!page) {
> : 		/* No page in the page cache at all */
> : 		fpin = do_sync_mmap_readahead(vmf);
> : 		count_vm_event(PGMAJFAULT);
> : 		count_memcg_event_mm(vmf->vma->vm_mm, PGMAJFAULT);
> 
> vmf->vma has been freed at this point.
> 
> : 		ret = VM_FAULT_MAJOR;
> : retry_find:
> : 		page = pagecache_get_page(mapping, offset,
> : 					  FGP_CREAT|FGP_FOR_MMAP,
> : 					  vmf->gfp_mask);
> : 		if (!page) {
> : 			if (fpin)
> : 				goto out_retry;
> : 			return vmf_error(-ENOMEM);
> : 		}
> : 	}
> 

Here's a fixup for "filemap: drop the mmap_sem for all blocking operations".

do_sync_mmap_readahead() drops mmap_sem now, so by the time of
dereferencing vmf->vma for count_memcg_event_mm() the VMA can be gone.

diff --git a/mm/filemap.c b/mm/filemap.c
index 00a9315f45d4..65c85c47bdb1 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2554,10 +2554,10 @@ vm_fault_t filemap_fault(struct vm_fault *vmf)
 		fpin = do_async_mmap_readahead(vmf, page);
 	} else if (!page) {
 		/* No page in the page cache at all */
-		fpin = do_sync_mmap_readahead(vmf);
 		count_vm_event(PGMAJFAULT);
 		count_memcg_event_mm(vmf->vma->vm_mm, PGMAJFAULT);
 		ret = VM_FAULT_MAJOR;
+		fpin = do_sync_mmap_readahead(vmf);
 retry_find:
 		page = pagecache_get_page(mapping, offset,
 					  FGP_CREAT|FGP_FOR_MMAP,
-- 
 Kirill A. Shutemov