Re: [PATCH] target/riscv/cpu.c: check priv_ver before auto-enable zca/zcd/zcf

From: Conor Dooley <conor@kernel.org>
To: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Cc: qemu-devel@nongnu.org, qemu-riscv@nongnu.org,
	alistair.francis@wdc.com, bmeng@tinylab.org,
	liweiwei@iscas.ac.cn, zhiwei_liu@linux.alibaba.com,
	palmer@rivosinc.com
Subject: Re: [PATCH] target/riscv/cpu.c: check priv_ver before auto-enable zca/zcd/zcf
Date: Mon, 17 Jul 2023 18:28:32 +0100	[thread overview]
Message-ID: <20230717-reattach-ungloved-ab5e4a067a4b@spud> (raw)
In-Reply-To: <20230717154141.60898-1-dbarboza@ventanamicro.com>

[-- Attachment #1: Type: text/plain, Size: 1961 bytes --]

On Mon, Jul 17, 2023 at 12:41:41PM -0300, Daniel Henrique Barboza wrote:
> Commit bd30559568 made changes in how we're checking and disabling
> extensions based on env->priv_ver. One of the changes was to move the
> extension disablement code to the end of realize(), being able to
> disable extensions after we've auto-enabled some of them.
> 
> An unfortunate side effect of this change started to happen with CPUs
> that has an older priv version, like sifive-u54. Starting on commit
> 2288a5ce43e5 we're auto-enabling zca, zcd and zcf if RVC is enabled,
> but these extensions are priv version 1.12.0. When running a cpu that
> has an older priv ver (like sifive-u54) the user is spammed with
> warnings like these:
> 
> qemu-system-riscv64: warning: disabling zca extension for hart 0x0000000000000000 because privilege spec version does not match
> qemu-system-riscv64: warning: disabling zcd extension for hart 0x0000000000000000 because privilege spec version does not match
> 
> The warnings are part of the code that disables the extension, but in this
> case we're throwing user warnings for stuff that we enabled on our own,
> without user intervention. Users are left wondering what they did wrong.
> 
> A quick 8.1 fix for this nuisance is to check the CPU priv spec before
> auto-enabling zca/zcd/zcf. A more appropriate fix will include a more
> robust framework that will account for both priv_ver and user choice
> when auto-enabling/disabling extensions, but for 8.1 we'll make it do
> with this simple check.
> 
> It's also worth noticing that this is the only case where we're
> auto-enabling extensions based on a criteria (in this case RVC) that
> doesn't match the priv spec of the extensions we're enabling. There's no
> need for more 8.1 band-aids.
> 
> Cc: Conor Dooley <conor@kernel.org>

Does the job, thanks for doing this.
Tested-by: Conor Dooley <conor.dooley@microchip.com>

Cheers,
Conor.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]