linux-kernel - Re: [PATCH] Loongarch: entry: fix syscall_get

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20d3df642aaff2a771e74452b81463709f7c16d1.camel@xry111.site>
Date: Thu, 28 Aug 2025 14:46:10 +0800
From: Xi Ruoyao <xry111@...111.site>
To: Wentao Guan <guanwentao@...ontech.com>, chenhuacai@...nel.org
Cc: kernel@...0n.name, linux-kernel@...r.kernel.org,
 loongarch@...ts.linux.dev, 	zhanjun@...ontech.com
Subject: Re: [PATCH] Loongarch: entry: fix syscall_get_arguments() VS
 no-bultin-memcpy

On Tue, 2025-08-26 at 19:32 +0800, Wentao Guan wrote:
> Loongarch use -fno-builtin-memcpy in Makefile,

I still think we should just remove -fno-builtin-memcpy if GCC >= 14. 
There's some wide-spread misunderstanding like "-fno-builtin-memcpy can
stop the compiler from using memcpy for copying small structs" but the
fact is very much contrary to it.

The compiler just uses memcpy for copying the structs of which the size
is larger than a threshold, no matter -fbuiltin-memcpy or not, but with
-fbuiltin-memcpy the compiler can expand some of them into ld/st
sequences, reducing the actual numbers of memcpy calls.

The reason we didn't use -fbuiltin-memcpy is GCC < 14 expands memcpy for
various sizes in a very stupid way, see https://gcc.gnu.org/PR109465.

-- 
Xi Ruoyao <xry111@...111.site>