linux-kernel - Re: [PATCH 2/2] perf: arm_pmuv3: Uninvert dependency between {asm,perf}/arm

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: <gsntpljw1apl.fsf@coltonlewis-kvm.c.googlers.com>
Date: Wed, 05 Feb 2025 18:57:10 +0000
From: Colton Lewis <coltonlewis@...gle.com>
To: Oliver Upton <oliver.upton@...ux.dev>
Cc: kvm@...r.kernel.org, linux@...linux.org.uk, catalin.marinas@....com, 
	will@...nel.org, maz@...nel.org, joey.gouly@....com, suzuki.poulose@....com, 
	yuzenghui@...wei.com, mark.rutland@....com, 
	linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org, 
	kvmarm@...ts.linux.dev
Subject: Re: [PATCH 2/2] perf: arm_pmuv3: Uninvert dependency between {asm,perf}/arm_pmuv3.h

Oliver Upton <oliver.upton@...ux.dev> writes:

> On Tue, Feb 04, 2025 at 07:57:08PM +0000, Colton Lewis wrote:
>> perf/arm_pmuv3.h includes asm/arm_pmuv3.h at the bottom of the
>> file. This counterintiutive decision was presumably made so
>> asm/arm_pmuv3.h would be included everywhere perf/arm_pmuv3.h was even
>> though the actual dependency relationship goes the other way because
>> asm/arm_pmuv3.h depends on the PMEVN_SWITCH macro that was presumably
>> put there to avoid duplicating it in the asm files for arm and arm64.

>> Extract the relevant macro to its own file to avoid this unusual
>> structure so it may be included in the asm headers without worrying
>> about ordering issues.

>> Signed-off-by: Colton Lewis <coltonlewis@...gle.com>

> Is the intention of this change to allow asm/arm_pmuv3.h to be directly
> included? If yes, what's the issue with using perf/arm_pmuv3.h?

That isn't the primary intent, but it's a good side effect. Headers that
can't be directly included violate people's expectations.

> We already use definitions from the non-arch header in KVM anyway...

My intention here was just reorganizing a counterintuitive use of
headers. The arch header depends on definitions in the non-arch header
even though the inclusion relationship is the other way around.

But this patch doesn't matter as much as fixing the cyclical dependency.