Re: [PATCH v1 1/9] RISC-V: don't parse dt/acpi isa string to get rv32/rv64

From: Conor Dooley <conor@kernel.org>
To: Andrew Jones <ajones@ventanamicro.com>
Cc: devicetree@vger.kernel.org, Albert Ou <aou@eecs.berkeley.edu>,
	linux-kernel@vger.kernel.org,
	Conor Dooley <conor.dooley@microchip.com>,
	Rob Herring <robh+dt@kernel.org>,
	palmer@dabbelt.com,
	Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
	Paul Walmsley <paul.walmsley@sifive.com>,
	linux-riscv@lists.infradead.org,
	Heiko Stuebner <heiko.stuebner@vrull.eu>,
	Evan Green <evan@rivosinc.com>
Subject: Re: [PATCH v1 1/9] RISC-V: don't parse dt/acpi isa string to get rv32/rv64
Date: Mon, 26 Jun 2023 17:16:09 +0100	[thread overview]
Message-ID: <20230626-ragweed-whenever-5b22e180dcd1@spud> (raw)
In-Reply-To: <20230626-4fb963235f3ab08383a6d9ab@orel>

[-- Attachment #1.1: Type: text/plain, Size: 1992 bytes --]

On Mon, Jun 26, 2023 at 06:05:40PM +0200, Andrew Jones wrote:
> On Mon, Jun 26, 2023 at 04:51:29PM +0100, Conor Dooley wrote:
> > On Mon, Jun 26, 2023 at 05:14:15PM +0200, Andrew Jones wrote:
> > > On Mon, Jun 26, 2023 at 12:19:39PM +0100, Conor Dooley wrote:
> > > > From: Heiko Stuebner <heiko.stuebner@vrull.eu>
> > > > @@ -333,8 +335,6 @@ static int c_show(struct seq_file *m, void *v)
> > > >  
> > > >  		of_node_put(node);
> > > >  	} else {
> > > > -		if (!acpi_get_riscv_isa(NULL, cpu_id, &isa))
> > > > -			print_isa(m, isa);
> > > >  
> > > 
> > > Extra blank line here to remove. Actually the whole 'else' can be removed
> > > because the print_mmu() call can be brought up above the
> > > 'if (acpi_disabled)'
> > 
> > Can it be? I intentionally did not make that change - wasn't sure
> > whether re-ordering the fields in there was permissible.
> 
> I agree we shouldn't change the order, but moving print_mmu() up won't,
> afaict.

D'oh, I'm an eejit. Sure, I'll do that for v2. Thanks!

> > One of the few things I know does parsing of /proc/cpuinfo is:
> > https://github.com/google/cpu_features/blob/main/src/impl_riscv_linux.c
> > and that doesn't seem to care about the mmu, but does rely on
> > vendor/uarch ordering.
> > 
> > Makes me wonder, does ACPI break things by leaving out uarch/vendor
> > fields, if there is something that expects them to exist? We should
> > not intentionally break stuff in /proc/cpuinfo, but can't say I feel any
> > sympathy for naively parsing it.
> 
> Yes, it would be nice for ACPI to be consistent. I'm not sure what can be
> done about that.

Print "unknown", until there's a way of passing the info?
Speaking of being an eejit, adding new fields to the file would probably
break some really naive parsers & quite frankly that sort of thing can
keep the pieces IMO. Ditto if adding more extensions breaks someone that
expects _zicbom_zicboz that breaks when _zicbop is slid into the middle.

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

[-- Attachment #2: Type: text/plain, Size: 161 bytes --]

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