Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:5 as permitted sender) client-ip=2620:137:e000::3:5;
MIME-Version: 1.0
References: <20231214062434.3565630-1-menglong8.dong@gmail.com>
 <20231214062434.3565630-2-menglong8.dong@gmail.com> <CAADnVQ+kKxj2hg33CzH_iXdH5fs8wjwpkPP-Jjh41weqf9BEwA@mail.gmail.com>
 <CADxym3Z1=6R8x7==wMfMA6N+Ha9VtOdpiw5LP7z4frW3=sYjCA@mail.gmail.com>
In-Reply-To: <CADxym3Z1=6R8x7==wMfMA6N+Ha9VtOdpiw5LP7z4frW3=sYjCA@mail.gmail.com>
From:   Alexei Starovoitov <alexei.starovoitov@gmail.com>
Date:   Thu, 14 Dec 2023 08:10:08 -0800
Message-ID: <CAADnVQJddDYv7K6UxqYA3PJk4OsdAphWoRzd5_gkxhU6rUBdKw@mail.gmail.com>
Subject: Re: [PATCH bpf-next v3 1/2] bpf: make the verifier tracks the "not
 equal" for regs
To:     Menglong Dong <menglong8.dong@gmail.com>
Cc:     Andrii Nakryiko <andrii@kernel.org>, Eddy Z <eddyz87@gmail.com>,
        Yonghong Song <yonghong.song@linux.dev>,
        Alexei Starovoitov <ast@kernel.org>,
        Daniel Borkmann <daniel@iogearbox.net>,
        John Fastabend <john.fastabend@gmail.com>,
        Martin KaFai Lau <martin.lau@linux.dev>,
        Song Liu <song@kernel.org>, KP Singh <kpsingh@kernel.org>,
        Stanislav Fomichev <sdf@google.com>,
        Hao Luo <haoluo@google.com>, Jiri Olsa <jolsa@kernel.org>,
        bpf <bpf@vger.kernel.org>, LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Dec 14, 2023 at 6:07=E2=80=AFAM Menglong Dong <menglong8.dong@gmail=
.com> wrote:
>
> Hello,
>
> On Thu, Dec 14, 2023 at 9:49=E2=80=AFPM Alexei Starovoitov
> <alexei.starovoitov@gmail.com> wrote:
> >
> > On Wed, Dec 13, 2023 at 10:28=E2=80=AFPM Menglong Dong <menglong8.dong@=
gmail.com> wrote:
> > >
> > > We can derive some new information for BPF_JNE in regs_refine_cond_op=
().
> > > Take following code for example:
> > >
> > >   /* The type of "a" is u16 */
> > >   if (a > 0 && a < 100) {
> > >     /* the range of the register for a is [0, 99], not [1, 99],
> > >      * and will cause the following error:
> > >      *
> > >      *   invalid zero-sized read
> > >      *
> > >      * as a can be 0.
> > >      */
> > >     bpf_skb_store_bytes(skb, xx, xx, a, 0);
> > >   }
> >
> > Please craft a selftest from above with inline asm
> > (C might not work as compiler might optimize it)
>
> Okay! Should I add this selftests to reg_bounds as a subtest,
> or add a "verifier_reg_edge.c" for verifier testing?

reg_bounds is for automated.
I think it will fit fine in the existing progs/verifier_bounds.c

>
> > Also we call:
> >         /* fallthrough (FALSE) branch */
> >         regs_refine_cond_op(false_reg1, false_reg2,
> > rev_opcode(opcode), is_jmp32);
> >         /* jump (TRUE) branch */
> >         regs_refine_cond_op(true_reg1, true_reg2, opcode, is_jmp32);
> >
> > so despite BPF_JNE is not handled explicitly it still should have
> > caught above due to rev_opcode() ?
>
> Ennn.....I'm a little confused. In this case, the TRUE path is
> handled properly, as the opcode is BPF_JEQ; and the FALSE
> is not handled properly, as the opcode is rev_opcode(BPF_JEQ),
> which is BPF_JNE. And the bpf_skb_store_bytes() will be called
> in the FALSE path. The origin state of false_reg* should be the same
> as true_reg*.

Ahh. I see.
It wasn't clear how 'a > 0 && a < 100' got to be JNE after optimizations.