bpf.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Eduard Zingerman <eddyz87@gmail.com>
To: Paul Chaignon <paul.chaignon@gmail.com>, bpf@vger.kernel.org
Cc: Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	 Andrii Nakryiko <andrii@kernel.org>,
	Yonghong Song <yonghong.song@linux.dev>
Subject: Re: [PATCH bpf-next 3/4] selftests/bpf: Test cross-sign 64bits range refinement
Date: Mon, 21 Jul 2025 14:30:07 -0700	[thread overview]
Message-ID: <755dfeb5b02a1d3b5dd8b87a5aeb822628a93996.camel@gmail.com> (raw)
In-Reply-To: <7cf24829f55fac6eee2b43e09e78fc03f443c8e5.1752934170.git.paul.chaignon@gmail.com>

On Sat, 2025-07-19 at 16:23 +0200, Paul Chaignon wrote:

[...]

> The first patch in this series ("bpf: Improve bounds when s64 crosses
> sign boundary") fixes this by refining ranges before we reach the
> condition, such that the verifier can detect the jump is always taken.
> Indeed, at instruction 7, the ranges look as follows:
> 
>     0          umin=0xfffffcf1                 umax=0xff..ff6e  U64_MAX
>     |                [xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx]        |
>     |----------------------------|------------------------------|
>     |xxxxxxxxxx]                                   [xxxxxxxxxxxx|
>     0    smax=0xeffffeee                       smin=-655        -1

I'd move this diagram to the selftest itself.

> 
> The updated __reg64_deduce_bounds can therefore improve the ranges to
> s64=[-655; -146] (and the u64 equivalent). With this new range, it's
> clear that the condition at instruction 8 is always true: R3's umax is
> 0xffffffff and R0's umin is 0xfffffffffffffd71 ((u64)-655). We avoid the
> dead branch and don't end up with an invariant violation.
> 
> Link: https://syzkaller.appspot.com/bug?extid=c711ce17dd78e5d4fdcf [1]
> Signed-off-by: Paul Chaignon <paul.chaignon@gmail.com>
> ---
>  .../selftests/bpf/progs/verifier_bounds.c     | 23 +++++++++++++++++++
>  1 file changed, 23 insertions(+)
> 
> diff --git a/tools/testing/selftests/bpf/progs/verifier_bounds.c b/tools/testing/selftests/bpf/progs/verifier_bounds.c
> index 63b533ca4933..d104d43ff911 100644
> --- a/tools/testing/selftests/bpf/progs/verifier_bounds.c
> +++ b/tools/testing/selftests/bpf/progs/verifier_bounds.c
> @@ -1550,4 +1550,27 @@ l0_%=:	r0 = 0;				\
>  	: __clobber_all);
>  }
>  
> +SEC("socket")
> +__description("bounds deduction sync cross sign boundary")
> +__success __log_level(2) __flag(BPF_F_TEST_REG_INVARIANTS)
> +__retval(0)
> +__naked void test_invariants(void)

Could you please check deduced range with __msg?

> +{
> +	asm volatile("			\
> +	call %[bpf_get_prandom_u32];	\
> +	w3 = w0;			\
> +	w6 = (s8)w0;			\
> +	r0 = (s8)r0;			\
> +	if w6 >= 0xf0000000 goto l0_%=;	\
> +	r0 += r6;			\
> +	r6 += 400;			\
> +	r0 -= r6;			\
> +	if r3 < r0 goto l0_%=;		\
> +l0_%=:	r0 = 0;				\
> +	exit;				\
> +"	:
> +	: __imm(bpf_get_prandom_u32)
> +	: __clobber_all);
> +}

I think two more test cases are needed:
- when intersection is on the other side of the interval;
- when signed and unsigned intervals overlap in two places.


  reply	other threads:[~2025-07-21 21:30 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-07-19 14:20 [PATCH bpf-next 0/4] bpf: Improve 64bits bounds refinement Paul Chaignon
2025-07-19 14:22 ` [PATCH bpf-next 1/4] bpf: Improve bounds when s64 crosses sign boundary Paul Chaignon
2025-07-21 21:29   ` Eduard Zingerman
2025-07-22  7:32   ` Shung-Hsi Yu
2025-07-22 22:09     ` Paul Chaignon
2025-07-23  7:49       ` Shung-Hsi Yu
2025-07-19 14:22 ` [PATCH bpf-next 2/4] selftests/bpf: Update reg_bound range refinement logic Paul Chaignon
2025-07-21 21:29   ` Eduard Zingerman
2025-07-22 21:20     ` Paul Chaignon
2025-07-22 21:26       ` Eduard Zingerman
2025-07-19 14:23 ` [PATCH bpf-next 3/4] selftests/bpf: Test cross-sign 64bits range refinement Paul Chaignon
2025-07-21 21:30   ` Eduard Zingerman [this message]
2025-07-24 14:03     ` Paul Chaignon
2025-07-19 14:24 ` [PATCH bpf-next 4/4] selftests/bpf: Test invariants on JSLT crossing sign Paul Chaignon

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=755dfeb5b02a1d3b5dd8b87a5aeb822628a93996.camel@gmail.com \
    --to=eddyz87@gmail.com \
    --cc=andrii@kernel.org \
    --cc=ast@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=daniel@iogearbox.net \
    --cc=paul.chaignon@gmail.com \
    --cc=yonghong.song@linux.dev \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).