Re: [PATCH 1/2] genirq: add an affinity parameter to irq_create_mapping()

From: Thomas Gleixner <tglx@linutronix.de>
To: Laurent Vivier <lvivier@redhat.com>, linux-kernel@vger.kernel.org
Cc: Paul Mackerras <paulus@samba.org>,
	Benjamin Herrenschmidt <benh@kernel.crashing.org>,
	Michael Ellerman <mpe@ellerman.id.au>,
	linux-block@vger.kernel.org,
	"Michael S . Tsirkin" <mst@redhat.com>,
	Marc Zyngier <maz@kernel.org>,
	linuxppc-dev@lists.ozlabs.org, linux-pci@vger.kernel.org,
	Christoph Hellwig <hch@lst.de>,
	Laurent Vivier <lvivier@redhat.com>
Subject: Re: [PATCH 1/2] genirq: add an affinity parameter to irq_create_mapping()
Date: Tue, 24 Nov 2020 23:19:47 +0100	[thread overview]
Message-ID: <87h7pel7ng.fsf@nanos.tec.linutronix.de> (raw)
In-Reply-To: <20201124200308.1110744-2-lvivier@redhat.com>

On Tue, Nov 24 2020 at 21:03, Laurent Vivier wrote:
> This parameter is needed to pass it to irq_domain_alloc_descs().
>
> This seems to have been missed by
> o06ee6d571f0e ("genirq: Add affinity hint to irq allocation")

No, this has not been missed at all. There was and is no reason to do
this.

> This is needed to implement proper support for multiqueue with
> pseries.

And because pseries needs this _all_ callers need to be changed?

>  123 files changed, 171 insertions(+), 146 deletions(-)

Lots of churn for nothing. 99% of the callers will never need that.

What's wrong with simply adding an interface which takes that parameter,
make the existing one an inline wrapper and and leave the rest alone?

Thanks,

        tglx