From mboxrd@z Thu Jan 1 00:00:00 1970 From: Daniel Borkmann Subject: Re: [PATCH v1] bpf: Set register type according to is_valid_access() Date: Thu, 22 Sep 2016 21:41:36 +0200 Message-ID: <57E433F0.90407@iogearbox.net> References: <20160922183512.13576-1-mic@digikod.net> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 8bit Cc: Alexei Starovoitov , Andy Lutomirski , Kees Cook , Sargun Dhillon , Tejun Heo , netdev@vger.kernel.org To: =?UTF-8?B?TWlja2HDq2wgU2FsYcO8bg==?= , linux-kernel@vger.kernel.org Return-path: In-Reply-To: <20160922183512.13576-1-mic@digikod.net> Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On 09/22/2016 08:35 PM, Mickaël Salaün wrote: > This fix a pointer leak when an unprivileged eBPF program read a pointer > value from the context. Even if is_valid_access() returns a pointer > type, the eBPF verifier replace it with UNKNOWN_VALUE. The register > value containing an address is then allowed to leak. Moreover, this > prevented unprivileged eBPF programs to use functions with (legitimate) > pointer arguments. > > This bug is not an issue for now because the only unprivileged eBPF > program allowed is of type BPF_PROG_TYPE_SOCKET_FILTER and all the types > from its context are UNKNOWN_VALUE. However, this fix is important for > future unprivileged eBPF program types which could use pointers in their > context. > > Signed-off-by: Mickaël Salaün > Fixes: 969bf05eb3ce ("bpf: direct packet access") > Cc: Alexei Starovoitov > Cc: Andy Lutomirski > Cc: Daniel Borkmann > Cc: Kees Cook > Acked-by: Sargun Dhillon > --- > kernel/bpf/verifier.c | 6 ++---- > 1 file changed, 2 insertions(+), 4 deletions(-) > > diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c > index daea765d72e6..0698ccd67715 100644 > --- a/kernel/bpf/verifier.c > +++ b/kernel/bpf/verifier.c > @@ -794,10 +794,8 @@ static int check_mem_access(struct verifier_env *env, u32 regno, int off, > } > err = check_ctx_access(env, off, size, t, ®_type); > if (!err && t == BPF_READ && value_regno >= 0) { > - mark_reg_unknown_value(state->regs, value_regno); > - if (env->allow_ptr_leaks) > - /* note that reg.[id|off|range] == 0 */ > - state->regs[value_regno].type = reg_type; > + /* note that reg.[id|off|range] == 0 */ > + state->regs[value_regno].type = reg_type; True that it's not an issue currently, since reg_type is only set for PTR_TO_PACKET/PTR_TO_PACKET_END in xdp and tc programs that can only be loaded as privileged. So not an issue for BPF_PROG_TYPE_SOCKET_FILTER. One thing I don't quite follow is why you remove the mark_reg_unknown_value() as this also clears imm? I think this could result in an actual verifier bug when it would reuse previous tracked imm value of that dst register? > } > > } else if (reg->type == FRAME_PTR || reg->type == PTR_TO_STACK) { >