[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <alpine.LFD.2.21.2003250240110.2689954@eddie.linux-mips.org>
Date: Wed, 25 Mar 2020 02:46:47 +0000 (GMT)
From: "Maciej W. Rozycki" <macro@...ux-mips.org>
To: Masahiro Yamada <masahiroy@...nel.org>
cc: Thomas Bogendoerfer <tsbogend@...ha.franken.de>,
linux-mips@...r.kernel.org, linux-kernel@...r.kernel.org,
Alexander Lobakin <alobakin@...nk.ru>,
Paul Burton <paulburton@...nel.org>,
clang-built-linux@...glegroups.com
Subject: Re: [PATCH] MIPS: mark some functions as weak to avoid conflict with
Octeon ones
On Wed, 25 Mar 2020, Masahiro Yamada wrote:
> MIPS provides multiple definitions for the following functions:
>
> fw_init_cmdline
> __delay
> __udelay
> __ndelay
> memmove
> __rmemcpy
> memcpy
> __copy_user
>
> The generic ones are defined in lib-y objects, which are overridden by
> the Octeon ones when CONFIG_CAVIUM_OCTEON_SOC is enabled.
>
> The use of EXPORT_SYMBOL in static libraries potentially causes a
> problem for the llvm linker [1]. So, I want to forcibly link lib-y
> objects to vmlinux when CONFIG_MODULES=y.
>
> As a groundwork, we must fix multiple definitions that have been
> hidden by lib-y.
IIUC that causes known dead code to be included in the kernel image.
Wouldn't it be possible to actually omit replaced functions from output by
keying the build of the sources providing generic code with appropriate
CONFIG_* settings (such as CONFIG_GENERIC_DELAY, CONFIG_GENERIC_MEMCPY,
etc. or suchlike)?
Maciej
Powered by blists - more mailing lists