linux-kernel - Re: [PATCH V3 13/14] LoongArch: Adjust default config files for 32BIT/64BIT

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] [day] [month] [year] [list]

Message-ID: <CAAhV-H5D=4ZhDLQOupyO0OOkuq2AWOGFBLpR8yAxaJcXb=QmRA@mail.gmail.com>
Date: Wed, 26 Nov 2025 16:02:54 +0800
From: Huacai Chen <chenhuacai@...nel.org>
To: Arnd Bergmann <arnd@...db.de>
Cc: Thomas Weißschuh <thomas@...ch.de>, 
	Huacai Chen <chenhuacai@...ngson.cn>, loongarch@...ts.linux.dev, 
	Linux-Arch <linux-arch@...r.kernel.org>, Xuefeng Li <lixuefeng@...ngson.cn>, 
	guoren <guoren@...nel.org>, WANG Xuerui <kernel@...0n.name>, 
	Jiaxun Yang <jiaxun.yang@...goat.com>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH V3 13/14] LoongArch: Adjust default config files for 32BIT/64BIT

On Wed, Nov 26, 2025 at 1:55 AM Arnd Bergmann <arnd@...db.de> wrote:
>
> On Sat, Nov 22, 2025, at 15:13, Huacai Chen wrote:
> > On Sat, Nov 22, 2025 at 7:57 PM Arnd Bergmann <arnd@...db.de> wrote:
> >> On Sat, Nov 22, 2025, at 10:45, Thomas Weißschuh wrote:
> >> > On 2025-11-22 12:36:33+0800, Huacai Chen wrote:
> >> >> diff --git a/arch/loongarch/Makefile b/arch/loongarch/Makefile
> >> >> index 96ca1a688984..cf9373786969 100644
> >> >> --- a/arch/loongarch/Makefile
> >> >> +++ b/arch/loongarch/Makefile
> >> >> @@ -5,7 +5,12 @@
> >> >>
> >> >>  boot        := arch/loongarch/boot
> >> >>
> >> >> -KBUILD_DEFCONFIG := loongson3_defconfig
> >> >> +ifdef CONFIG_32BIT
> >> >
> >> > Testing for CONFIG options here doesn't make sense, as the config is not yet
> >> > created.
> >>
> >> > Either test for $(ARCH) or uname or just use one unconditionally.
> >>
> >> I don't really like the $(ARCH) hacks, nobody is going to build kernels
> >> natively on loongarch32, and for the rest it's fine to set the option.
> > OK, I will use 'uname -m' for checking. Though native builds on
> > loongarch32 will hardly happen, we can give a small chance to use
> > loongson32_defconfig, in all other cases, let's use
> > loongson64_defconfig.
>
> What I meant is not to guess the CONFIG_64BIT option from any
> part of the build environment other than the actual .config settings.
>
> $(ARCH) is initialized from 'uname -m' in scripts/subarch.include,
> and I think this should stay unchanged.
>
> >> > Also as mentioned before, snippets can reduce the duplication.
> >> >
> >> >> +KBUILD_DEFCONFIG := loongson32_defconfig
> >> >> +else
> >> >> +KBUILD_DEFCONFIG := loongson64_defconfig
> >> >> +endif
> >> >> +
> >>
> >> This is also not the change I had suggested in my review. I think this
> >> should be a fragment along the lines of arch/mips/configs/generic/32r2.config
> >> and arch/powerpc/configs/book3s_32.config.
> >
> > Sorry for that. I know that the default config file is usually
> > generated by 'make savedefconfig', and the main purpose is
> > significantly reducing file size. I did that, but manually add some
> > lines such as CONFIG_LOONGARCH, CONFIG_32BIT/64BIT, CONFIG_ACPI,
> > CONFIG_EFI, etc. My original goal of doing this is to let users easily
> > know those fundamental options without reading Kconfig files (of
> > course we should not increase the defconfig too much).
> >
> > Sorry again for not explaining about this in the previous version.
>
> I don't mind having the extra options in the generic defconfig
> file, that is unusual but I see your logic there.
>
> My point here is that you should absolutely not duplicate the
> contents but only put the differences into the new file, such
> as
>
> +++ arch/loongarch/configs/loongson32.config
> CONFIG_32BIT=y
> # CONFIG_64BIT is not set
> CONFIG_32BIT_STANDARD=y
> CONFIG_MACH_LOONGSON32=y
> # CONFIG_SMP is not set
> # CONFIG_LOONGSON3_CPUFREQ is not set
> # CONFIG_IPMI_HANDLER is not set
> # CONFIG_TCG_TPM is not set
> # CONFIG_I2C_LS2X is not set
> # CONFIG_PINCTRL_LOONGSON2 is not set
> CONFIG_GPIO_LOONGSON1=y
> # CONFIG_GPIO_LOONGSON is not set
> CONFIG_WATCHDOG=y
> CONFIG_LOONGSON1_WDT=m
> # CONFIG_LOONGSON2_THERMAL=m is not set
>  # CONFIG_IOMMU_SUPPORT is not set
> CONFIG_LOONGSON1_APB_DMA=y
> # CONFIG_LOONGSON2_APB_DMA is not set
> # CONFIG_PWM_LOONGSON is not set
OK,  now I completely understand your point, thank you.

>
> This way, running 'make defconfig; make loongson32.config'
> should produce the same output as your current
> 'make loongson32_defconfig', but it is much easier to keep
> the two in sync because any future changes only
> have to be done in the shared file.
This solution solve the synchronization issue, but break a widely used workflow:
cp xxxx_defconfig .config
make menuconfig  # or something similar
make

If there is a solution similar to dts/dtsi (different dts include the
same dtsi as a common basis), that will be better.

But for now, I prefer to use two independent config files. Fortunately
there are only two so synchronization is not a big problem...

Huacai

>
>      Arnd
>