netdev - Re: [PATCH bpf-next 2/8] riscv, bpf: add support for far branching

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <2d5d1f2d-d4ab-2449-37c6-e5b319a778d6@iogearbox.net>
Date:   Mon, 9 Dec 2019 22:27:27 +0100
From:   Daniel Borkmann <daniel@...earbox.net>
To:     Luke Nelson <lukenels@...washington.edu>,
        Björn Töpel <bjorn.topel@...il.com>
Cc:     Alexei Starovoitov <ast@...nel.org>,
        Netdev <netdev@...r.kernel.org>, linux-riscv@...ts.infradead.org,
        bpf <bpf@...r.kernel.org>, Xi Wang <xi.wang@...il.com>
Subject: Re: [PATCH bpf-next 2/8] riscv, bpf: add support for far branching

On 12/9/19 10:08 PM, Luke Nelson wrote:
[...]
> We have been developing a formal verification tool for BPF JIT
> compilers, which we have used in the past to find bugs in the RV64
> and x32 BPF JITs:
> 
> https://unsat.cs.washington.edu/projects/serval/
> 
> Recently I added support for verifying the JIT for branch and jump
> instructions, and thought it a good opportunity to verify these
> patches that add support for far jumps and branching.
> 
> I ported these patches to our tool and ran verification, which
> didn't find any bugs according to our specification of BPF and
> RISC-V.
> 
> The tool and code are publicly available, and you can read a more
> detailed writeup of the results here:
> 
> https://github.com/uw-unsat/bpf-jit-verif/tree/far-jump-review
> 
> Currently the tool works on a manually translated version of the
> JIT from C to Rosette, but we are experimenting with ways of making
> this process more automated.

This is awesome work! Did you also check for other architectures aside
from riscv and x86-32, e.g. x86-64 or arm64?

It would be great if we could add such verification tool under tools/bpf/
which would then take the in-tree JIT-code as-is for its analysis and
potentially even trigger a run out of BPF selftests. Any thoughts whether
such path would be feasible wrt serval?

> Reviewed-by: Luke Nelson <lukenels@...washington.edu>
> Cc: Xi Wang <xi.wang@...il.com>

Thanks,
Daniel