Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 29 Dec 2022 05:57:13 +0000
From:   Thomas =?utf-8?Q?Wei=C3=9Fschuh?= <thomas@t-8ch.de>
To:     Rong Tao <rtoax@foxmail.com>
Cc:     jpoimboe@kernel.org, linux-kernel@vger.kernel.org,
        linux@weissschuh.net, peterz@infradead.org,
        Rong Tao <rongtao@cestc.cn>
Subject: Re: [PATCH v2 5/8] objtool: reduce memory usage of struct reloc
Message-ID: <20221229055713.7eqozt57aknqujjl@snowball.t-8ch.de>
References: <20221229022622.khboqwfe23e7u2en@snowball.t-8ch.de>
 <tencent_61428759F197532A46AF38A338882C06D108@qq.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <tencent_61428759F197532A46AF38A338882C06D108@qq.com>
Precedence: bulk

Hi Rong,

On Thu, Dec 29, 2022 at 11:29:07AM +0800, Rong Tao wrote:
> Hi, Thomas:
> 
> In /usr/include/elf.h has:
> 
> #define ELF32_R_TYPE(val)		((val) & 0xff)
> #define ELF64_R_TYPE(i)			((i) & 0xffffffff)
>                                        ^^^^^^^^^^
> 
> So, I still feel that keeping 'unsigned int' is a good option. Can we just
> use __attribute__((packed)) for wasted padding bytes?

As the struct contains addresses, packing it will create a ton of
compiler warnings and will be much more likely to break than reducing
the width of 'type'.

Given that reducing the width of 'type'

* is currently absolutely safe,
* is unlikely to break in the future, as the elf designers seem to be
  actively trying to stay within this range anyways,
* saves 8 bytes from a very heavily allocated struct, having a
  measurable impact on the build process,

I continue to propose this aproach.

Let's let the maintainers decide.

Thomas