Re: [Qemu-devel] [PATCH] virtiofsd: fix req use-after-free in lo_flush()

[Dr. David Alan Gilbert]

* Stefan Hajnoczi (address@hidden) wrote:
> Isn't it funny that a commit trying to fix use-after-free introduced its
> own use-after-free?
> 
> fuse_reply_err() frees the request so we cannot call lo_data(req)
> afterwards.
> 
> Fixes: a5081fc6fc3603671383616127b3a5e169cf64ed
>        ("virtiofsd: introduce inode refcount to prevent use-after-free")
> Signed-off-by: Stefan Hajnoczi <address@hidden>

Squashed into the top.

Dave

> ---
>  contrib/virtiofsd/passthrough_ll.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/contrib/virtiofsd/passthrough_ll.c 
> b/contrib/virtiofsd/passthrough_ll.c
> index ee8627446a..a81c01d0d1 100644
> --- a/contrib/virtiofsd/passthrough_ll.c
> +++ b/contrib/virtiofsd/passthrough_ll.c
> @@ -1969,9 +1969,8 @@ static void lo_flush(fuse_req_t req, fuse_ino_t ino, 
> struct fuse_file_info *fi)
>       pthread_mutex_unlock(&inode->plock_mutex);
>  
>       res = close(dup(lo_fi_fd(req, fi)));
> -     fuse_reply_err(req, res == -1 ? errno : 0);
> -
>       lo_inode_put(lo_data(req), &inode);
> +     fuse_reply_err(req, res == -1 ? errno : 0);
>  }
>  
>  static void lo_fsync(fuse_req_t req, fuse_ino_t ino, int datasync,
> -- 
> 2.21.0
> 
--
Dr. David Alan Gilbert / address@hidden / Manchester, UK