Re: [PATCH bpf 2/2] libbpf: fix possible use after free in xsk_socket__delete

[Andrii Nakryiko <andrii.nakryiko@gmail.com>]

On Wed, Nov 4, 2020 at 12:27 AM Magnus Karlsson
<magnus.karlsson@gmail.com> wrote:
>
> On Tue, Nov 3, 2020 at 8:05 PM Andrii Nakryiko
> <andrii.nakryiko@gmail.com> wrote:
> >
> > On Tue, Nov 3, 2020 at 1:42 AM Magnus Karlsson
> > <magnus.karlsson@gmail.com> wrote:
> > >
> > > From: Magnus Karlsson <magnus.karlsson@intel.com>
> > >
> > > Fix a possible use after free in xsk_socket__delete that will happen
> > > if xsk_put_ctx() frees the ctx. To fix, save the umem reference taken
> > > from the context and just use that instead.
> > >
> > > Fixes: 2f6324a3937f ("libbpf: Support shared umems between queues and devices")
> > > Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
> > > ---
> > >  tools/lib/bpf/xsk.c | 6 ++++--
> > >  1 file changed, 4 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/tools/lib/bpf/xsk.c b/tools/lib/bpf/xsk.c
> > > index 504b7a8..9bc537d 100644
> > > --- a/tools/lib/bpf/xsk.c
> > > +++ b/tools/lib/bpf/xsk.c
> > > @@ -892,6 +892,7 @@ void xsk_socket__delete(struct xsk_socket *xsk)
> > >  {
> > >         size_t desc_sz = sizeof(struct xdp_desc);
> > >         struct xdp_mmap_offsets off;
> > > +       struct xsk_umem *umem;
> > >         struct xsk_ctx *ctx;
> > >         int err;
> > >
> > > @@ -899,6 +900,7 @@ void xsk_socket__delete(struct xsk_socket *xsk)
> > >                 return;
> > >
> > >         ctx = xsk->ctx;
> > > +       umem = ctx->umem;
> > >         if (ctx->prog_fd != -1) {
> > >                 xsk_delete_bpf_maps(xsk);
> > >                 close(ctx->prog_fd);
> > > @@ -918,11 +920,11 @@ void xsk_socket__delete(struct xsk_socket *xsk)
> > >
> > >         xsk_put_ctx(ctx);
> > >
> > > -       ctx->umem->refcount--;
> > > +       umem->refcount--;
> >
> > if you moved ctx->umem->refcount--; to before xdk_put_ctx(ctx), would
> > that also work?
>
> Yes, it would for that statement, but I still need the umem pointer
> for the statement below. And this statement of potentially closing the
> fd needs to be after xsk_put_ctx(). So we might as well keep
> ujmem->refcount-- where it is, if that is ok with you?

Ah, missed the umem->fd below. Then it makes sense, thanks.

Acked-by: Andrii Nakryiko <andrii@kernel.org>

>
> > >         /* Do not close an fd that also has an associated umem connected
> > >          * to it.
> > >          */
> > > -       if (xsk->fd != ctx->umem->fd)
> > > +       if (xsk->fd != umem->fd)
> > >                 close(xsk->fd);
> > >         free(xsk);
> > >  }
> > > --
> > > 2.7.4
> > >