Received-SPF: pass (google.com: domain of linux-kernel+bounces-59484-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Feedback-ID: i56a14606:Fastmail
User-Agent: Cyrus-JMAP/3.11.0-alpha0-144-ge5821d614e-fm-20240125.002-ge5821d61
Precedence: bulk
MIME-Version: 1.0
Message-Id: <6eeba3ce-d66c-45e2-8c34-ad0109ec2ef0@app.fastmail.com>
In-Reply-To: <20240208225608.11987-1-kernel@valentinobst.de>
References: <20240208-alice-mm-v2-2-d821250204a6@google.com>
 <20240208225608.11987-1-kernel@valentinobst.de>
Date: Fri, 09 Feb 2024 15:41:08 +0100
From: "Arnd Bergmann" <arnd@arndb.de>
To: "Valentin Obst" <kernel@valentinobst.de>,
 "Alice Ryhl" <aliceryhl@google.com>
Cc: "Andreas Hindborg" <a.hindborg@samsung.com>,
 "Andrew Morton" <akpm@linux-foundation.org>,
 "Alex Gaynor" <alex.gaynor@gmail.com>,
 =?UTF-8?Q?Arve_Hj=C3=B8nnev=C3=A5g?= <arve@android.com>,
 "Benno Lossin" <benno.lossin@proton.me>,
 =?UTF-8?Q?Bj=C3=B6rn_Roy_Baron?= <bjorn3_gh@protonmail.com>,
 "Boqun Feng" <boqun.feng@gmail.com>,
 "Christian Brauner" <brauner@kernel.org>,
 "Carlos Llamas" <cmllamas@google.com>, "Gary Guo" <gary@garyguo.net>,
 "Greg Kroah-Hartman" <gregkh@linuxfoundation.org>,
 "Joel Fernandes" <joel@joelfernandes.org>,
 "Kees Cook" <keescook@chromium.org>, linux-kernel@vger.kernel.org,
 linux-mm@kvack.org, "Martijn Coenen" <maco@android.com>,
 "Miguel Ojeda" <ojeda@kernel.org>, rust-for-linux@vger.kernel.org,
 "Suren Baghdasaryan" <surenb@google.com>, "Todd Kjos" <tkjos@android.com>,
 "Alexander Viro" <viro@zeniv.linux.org.uk>,
 "Wedson Almeida Filho" <wedsonaf@gmail.com>
Subject: Re: [PATCH v2 2/4] uaccess: always export _copy_[from|to]_user with
 CONFIG_RUST
Content-Type: text/plain

On Thu, Feb 8, 2024, at 23:56, Valentin Obst wrote:
>> -#else
>>  extern __must_check unsigned long
>>  _copy_from_user(void *, const void __user *, unsigned long);
>> -#endif
>
> This function is now unconditionally declared, but only defined if
> `!defined(INLINE_COPY_FROM_USER) || defined(CONFIG_RUST)`, i.e., in the
> common case where it is inlined and Rust is disabled this can lead to
> link-time problems if someone decides to use it.

Yes, that is intentional.

If someone tries to use it when the declaration is not there,
they just get a compile-time error, which is not all that
different from a link-time error in practice.

It's unlikely to make a difference here, but enclosing
declarations in an #ifdef is annoying when you want to
reference it from somewhere that is parsed by the compiler
but not called without the respective options.

The if(IS_ENABLED()) and PTR_IF() constructs in particular
only work when the unused functions are still declared.

       Arnd