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: <20220502110741.951055904@infradead.org> <20220502111216.350926848@infradead.org>
In-Reply-To: <20220502111216.350926848@infradead.org>
From:   Marco Elver <elver@google.com>
Date:   Mon, 2 May 2022 13:23:49 +0200
Message-ID: <CANpmjNP6k5F0vHg6xiu9yhmC7PotGPxVW8M2_pvSKY25eNCnDQ@mail.gmail.com>
Subject: Re: [PATCH 3/3] jump_label,noinstr: Avoid instrumentation for
 JUMP_LABEL=n builds
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     x86@kernel.org, jpoimboe@redhat.com, linux-kernel@vger.kernel.org,
        jbaron@akamai.com, rostedt@goodmis.org, ardb@kernel.org,
        kernel test robot <lkp@intel.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, 2 May 2022 at 13:12, Peter Zijlstra <peterz@infradead.org> wrote:
>
> When building x86_64 with JUMP_LABEL=n it's possible for
> instrumentation to sneak into noinstr:
>
> vmlinux.o: warning: objtool: exit_to_user_mode+0x14: call to static_key_count.constprop.0() leaves .noinstr.text section
> vmlinux.o: warning: objtool: syscall_exit_to_user_mode+0x2d: call to static_key_count.constprop.0() leaves .noinstr.text section
> vmlinux.o: warning: objtool: irqentry_exit_to_user_mode+0x1b: call to static_key_count.constprop.0() leaves .noinstr.text section
>
> Reported-by: kernel test robot <lkp@intel.com>
> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
> ---
>  include/linux/jump_label.h |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> --- a/include/linux/jump_label.h
> +++ b/include/linux/jump_label.h
> @@ -256,9 +256,9 @@ extern void static_key_disable_cpuslocke
>  #include <linux/atomic.h>
>  #include <linux/bug.h>
>
> -static inline int static_key_count(struct static_key *key)
> +static __always_inline int static_key_count(struct static_key *key)
>  {
> -       return atomic_read(&key->enabled);
> +       return READ_ONCE_NOCHECK(&key->enabled.count);
>  }

Did arch_atomic_read() not work? Or should this also be unchecked in
instrumented functions?