From: Sasha Levin <sashal@kernel.org>
To: gregkh@linuxfoundation.org
Cc: Sasha Levin <sashal@kernel.org>,
	andrii@kernel.org,
	ast@kernel.org,
	kpsingh@kernel.org,
	patches@lists.linux.dev,
	stable@vger.kernel.org,
	xukuohai@huawei.com,
	Wentao Guan <guanwentao@uniontech.com>
Subject: Re: [PATCH 6.6 0772/1266] bpf, lsm: Add disabled BPF LSM hook list
Date: Tue, 28 Jul 2026 23:17:55 -0400
Message-ID: <20260728225520.stable-0003@kernel.org>
In-Reply-To: <20260728033810.1870379-1-guanwentao@uniontech.com>
References: <20260728033810.1870379-1-guanwentao@uniontech.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

> backport ("bpf, lsm: Add disabled BPF LSM hook list") for v6.6.y bring the new
> build warning for us, let your know that.
>
> and the lsm_getselfattr from a04a1198088a1378d0389c250cc684f649bcc91e commit
> ("LSM: syscalls for current process attributes"), I suggest that use a not
> upstream commit to just drop 'BTF_ID(func, bpf_lsm_getselfattr)' from 6.6.y.
> Maybe 6.1.y has the problem too.

Thanks for the report, I agree with the analysis.

> Maybe 6.1.y has the problem too.

Yes, 6.1.y is affected as well - the same backport of 21c7063f6d08
("bpf, lsm: Add disabled BPF LSM hook list") is in v6.1.178.

The impact is build noise only: resolve_btfids only prints the WARN and
carries on, the unresolved entry is patched as BTF ID 0, and 0 never
matches a real prog->aux->attach_btf_id, so the enforcement for the other
disabled hooks keeps working. I do not want to revert the pair, since
21c7063f6d08 came in as a dependency of 12091470c6b4 ("bpf: Disable
xfrm_decode_session hook attachment"), which prevents a reachable BUG_ON()
and does belong on these trees.

Could you post a formal patch for this please?

-- 
Thanks,
Sasha