Re: [PATCH v10 2/2] riscv: Add runtime constant support

From: Klara Modin <klarasmodin@gmail.com>
To: Charlie Jenkins <charlie@rivosinc.com>
Cc: Paul Walmsley <paul.walmsley@sifive.com>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	Ard Biesheuvel <ardb@kernel.org>,
	Ben Dooks <ben.dooks@codethink.co.uk>,
	Pasha Bouzarjomehri <pasha@rivosinc.com>,
	Emil Renner Berthing <emil.renner.berthing@canonical.com>,
	Alexandre Ghiti <alexghiti@rivosinc.com>,
	Steven Rostedt <rostedt@goodmis.org>,
	Masami Hiramatsu <mhiramat@kernel.org>,
	Mark Rutland <mark.rutland@arm.com>,
	Albert Ou <aou@eecs.berkeley.edu>,
	Peter Zijlstra <peterz@infradead.org>,
	Josh Poimboeuf <jpoimboe@kernel.org>,
	Jason Baron <jbaron@akamai.com>,
	Andrew Jones <ajones@ventanamicro.com>,
	linux-riscv@lists.infradead.org, linux-kernel@vger.kernel.org,
	linux-trace-kernel@vger.kernel.org
Subject: Re: [PATCH v10 2/2] riscv: Add runtime constant support
Date: Fri, 28 Mar 2025 21:22:15 +0100	[thread overview]
Message-ID: <dfc8ee31-2f81-4895-a11d-413641daf8c0@gmail.com> (raw)
In-Reply-To: <Z-b9xGX7SaPgQq15@ghost>

On 3/28/25 20:51, Charlie Jenkins wrote:
> On Fri, Mar 28, 2025 at 04:42:42PM +0100, Klara Modin wrote:
>> Hi,
>>
>> On 3/19/25 19:35, Charlie Jenkins wrote:
>>> Implement the runtime constant infrastructure for riscv. Use this
>>> infrastructure to generate constants to be used by the d_hash()
>>> function.
>>>
>>> This is the riscv variant of commit 94a2bc0f611c ("arm64: add 'runtime
>>> constant' support") and commit e3c92e81711d ("runtime constants: add
>>> x86 architecture support").
>>
>> This patch causes the following build failure for me:
>>
>> fs/dcache.c: Assembler messages:
>> fs/dcache.c:157: Error: attempt to move .org backwards
>> fs/dcache.c:157: Error: attempt to move .org backwards
>> fs/dcache.c:157: Error: attempt to move .org backwards
>> fs/dcache.c:157: Error: attempt to move .org backwards
>> fs/dcache.c:157: Error: attempt to move .org backwards
>> make[3]: *** [scripts/Makefile.build:203: fs/dcache.o] Error 1
> 
> Thank you for the report, this seems like a binutils issue potentially.
> I will look into it. Here is a minimally reproducible example:
> 
> 886 :
> .option push
> .option norvc
> 	nop
> 	nop
> .option pop
> 887 :
> 888 :
> .option push
> .option norvc
> .option arch,+zba
> 	nop
> 	nop
> .option pop
> 889 :
> .org	. - (887b - 886b) + (889b - 888b)
> .org	. - (889b - 888b) + (887b - 886b)
> 
> Removing the ".option arch,+zba" fixes the issue but that shouldn't
> matter...

I tried again with GCC 14.2 and 12.4 (with the same binutils version) 
after Alex's answer and couldn't see the issue with these. I got the 
same result with your example. If I invoke `as` directly it doesn't 
happen either.

The issue might be with GCC 15 then?

Regards,
Klara Modin

> 
> - Charlie
> 
>>
>> The value of CONFIG_RISCV_ISA_ZBKB doesn't seem to have an impact. Reverting
>> the patch on top of next-20250328 resolved the issue for me. I attached the
>> generated fs/dcache.s.
>>
>> Please let me know if there's anything else you need.
>>
>> Regards,
>> Klara Modin
>>

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