Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
MIME-Version: 1.0
References: <20220421072212.608884-1-song@kernel.org> <CAHk-=wi3eu8mdKmXOCSPeTxABVbstbDg1q5Fkak+A9kVwF+fVw@mail.gmail.com>
 <CAADnVQKyDwXUMCfmdabbVE0vSGxdpqmWAwHRBqbPLW=LdCnHBQ@mail.gmail.com> <CAHk-=whFeBezdSrPy31iYv-UZNnNavymrhqrwCptE4uW8aeaHw@mail.gmail.com>
In-Reply-To: <CAHk-=whFeBezdSrPy31iYv-UZNnNavymrhqrwCptE4uW8aeaHw@mail.gmail.com>
From:   Song Liu <song@kernel.org>
Date:   Thu, 21 Apr 2022 12:40:56 -0700
Message-ID: <CAPhsuW7M6exGD3C1cPBGjhU0Y5efxtJ3=0BWNnbuH87TgQMzdg@mail.gmail.com>
Subject: Re: [PATCH bpf] bpf: invalidate unused part of bpf_prog_pack
To:     Linus Torvalds <torvalds@linux-foundation.org>
Cc:     Alexei Starovoitov <alexei.starovoitov@gmail.com>,
        bpf <bpf@vger.kernel.org>, Linux-MM <linux-mm@kvack.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Alexei Starovoitov <ast@kernel.org>,
        Daniel Borkmann <daniel@iogearbox.net>,
        Kernel Team <kernel-team@fb.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        "Edgecombe, Rick P" <rick.p.edgecombe@intel.com>,
        Christoph Hellwig <hch@infradead.org>,
        Andrii Nakryiko <andrii@kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi Linus,

On Thu, Apr 21, 2022 at 11:59 AM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> On Thu, Apr 21, 2022 at 11:24 AM Alexei Starovoitov
> <alexei.starovoitov@gmail.com> wrote:
> >
> > Let's not complicate the logic by dragging jit_fill_hole
> > further into generic allocation.
>
> I agree that just zeroing the page is probably perfectly fine in
> practice on x86, but I'm also not really seeing the "complication" of
> just doing things right.
>
> > The existing bpf_prog_pack code still does memset(0xcc)
> > a random range of bytes before and after jit-ed bpf code.
>
> That is actually wishful thinking, and not based on reality.
>
> From what I can tell, the end of the jit'ed bpf code is actually the
> exception table entries, so we have that data being marked executable.
>
> Honestly, what is wrong with this trivial patch?

This version would fill the memory with illegal instruction when we
allocate the bpf_prog_pack.

The extra logic I had in the original patch was to erase the memory
when a BPF program is freed. In this case, the memory will be
returned to the bpf_prog_pack, and stays as RO+X. Actually, I
am not quite sure whether we need this logic. If not, we only need
the much simpler version.

Thanks,
Song