Re: [tip:irq/urgent] BUILD REGRESSION 9f7488f24c7571d349d938061e0ede7a39b65d6b

From: Alexandre Ghiti <alex@ghiti.fr>
To: Thomas Gleixner <tglx@linutronix.de>,
	kernel test robot <lkp@intel.com>, x86-ml <x86@kernel.org>,
	Marco Bonelli <marco@mebeim.net>
Cc: linux-kernel@vger.kernel.org, linux-riscv@lists.infradead.org
Subject: Re: [tip:irq/urgent] BUILD REGRESSION 9f7488f24c7571d349d938061e0ede7a39b65d6b
Date: Thu, 7 Aug 2025 13:48:25 +0200	[thread overview]
Message-ID: <d5e49344-e0c2-4095-bd1f-d2d23a8e6534@ghiti.fr> (raw)
In-Reply-To: <87h5yjcxqh.ffs@tglx>

On 8/7/25 11:00, Thomas Gleixner wrote:
> On Thu, Aug 07 2025 at 15:10, kernel test robot wrote:
>> tree/branch: https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git irq/urgent
>> branch HEAD: 9f7488f24c7571d349d938061e0ede7a39b65d6b  irqchip/mvebu-gicp: Use resource_size() for ioremap()
>>
>> Error/Warning (recently discovered and may have been fixed):
>>
>>      https://lore.kernel.org/oe-kbuild-all/202508070607.QKVa8DiU-lkp@intel.com
>>
>>      ERROR: modpost: __ex_table+0x1584 references non-executable section '.rodata.__func__.103779'
>>
>> Error/Warning ids grouped by kconfigs:
>>
>> recent_errors
>> `-- riscv-randconfig-002-20250807
>>      `-- ERROR:__ex_table-references-non-executable-section-.rodata.__func__.
> This has nothing to do with the above commit. That's just a new one
> of the same type as the ones which existed before:
>
> WARNING: modpost: vmlinux: section mismatch in reference: xp_fill_cb+0x46 (section: .text.xp_fill_cb) -> .L62 (section: .init.text)
> ...
> WARNING: modpost: vmlinux: section mismatch in reference: 0x19e4 (section: __ex_table) -> .LASF400 (section: .debug_str)
>
> There are gazillions of them and that's a purely riscv specific related
> problem.

Yes that's a riscv issue that was already reported by Marco [1] who also 
found the underlying issue [2]. The problem is that we emit too many 
relocations which overflow the number of possible relocations in 32-bit. 
After discussion with some riscv gcc folks, it appears to be caused by 
relaxations which, according to the ABI, require the emission of local 
symbol. We don't have a workaround for now and we'll discuss this issue 
with the ABI authors.

[1] 
https://lore.kernel.org/linux-riscv/960240908.630790.1748641210849@privateemail.com/
[2] 
https://lore.kernel.org/all/1714337938.319508.1750244108368@privateemail.com/

Sorry for the noise,

Alex

>
> Thanks,
>
>          tglx
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv