[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190513172527.GB16567@lakrids.cambridge.arm.com>
Date: Mon, 13 May 2019 18:25:27 +0100
From: Mark Rutland <mark.rutland@....com>
To: Eric Dumazet <edumazet@...gle.com>
Cc: "David S . Miller" <davem@...emloft.net>,
Alexei Starovoitov <ast@...nel.org>,
Daniel Borkmann <daniel@...earbox.net>,
netdev <netdev@...r.kernel.org>,
Eric Dumazet <eric.dumazet@...il.com>,
bpf <bpf@...r.kernel.org>, syzbot <syzkaller@...glegroups.com>,
Petar Penkov <ppenkov@...gle.com>,
Stanislav Fomichev <sdf@...gle.com>
Subject: Re: [PATCH net] flow_dissector: disable preemption around BPF calls
On Mon, May 13, 2019 at 10:20:19AM -0700, 'Eric Dumazet' via syzkaller wrote:
> On Mon, May 13, 2019 at 10:17 AM Mark Rutland <mark.rutland@....com> wrote:
> >
> > On Mon, May 13, 2019 at 09:38:55AM -0700, 'Eric Dumazet' via syzkaller wrote:
> > > Various things in eBPF really require us to disable preemption
> > > before running an eBPF program.
> >
> > Is that true for all eBPF uses? I note that we don't disable preemption
> > in the lib/test_bpf.c module, for example.
> >
> > If it's a general requirement, perhaps it's worth an assertion within
> > BPF_PROG_RUN()?
>
> The assertion is already there :)
>
> This is how syzbot triggered the report.
Ah! :)
I also see I'm wrong about test_bpf.c, so sorry for the noise on both
counts!
Thanks,
Mark.
Powered by blists - more mailing lists