From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeff Garzik Subject: Re: [PATCH] [resend] drivers/net: remove network drivers' last few uses of IRQF_SAMPLE_RANDOM Date: Wed, 13 May 2009 15:39:11 -0400 Message-ID: <4A0B21DF.9050702@garzik.org> References: <20090513060850.GZ31071@waste.org> <20090513142535.GB31071@waste.org> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: Chris Peterson , linux-kernel@vger.kernel.org, NetDev To: Matt Mackall Return-path: Received: from srv5.dvmed.net ([207.36.208.214]:57321 "EHLO mail.dvmed.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755452AbZEMTjQ (ORCPT ); Wed, 13 May 2009 15:39:16 -0400 In-Reply-To: <20090513142535.GB31071@waste.org> Sender: netdev-owner@vger.kernel.org List-ID: Matt Mackall wrote: > On Wed, May 13, 2009 at 12:17:29AM -0700, Chris Peterson wrote: >>> The future model will continue to sample network >>> devices on theory that they -might- be less than 100% observable and >>> that can only increase our total (unmeasurable) amount of entropy. >> That sounds reasonable to me. So should all net drivers now specify >> IRQF_SAMPLE_RANDOM? >> >> Or even simpler: could request_irq() assume IRQF_SAMPLE_RANDOM for any >> interrupt that is not (say) IRQF_IRQPOLL or IRQF_PERCPU? > > Maybe. We don't want IRQ latency to suffer. So before we turn on > sampling of -all- sources, we need to make sampling lighter weight and > we need a way to say 'we have enough' so that we're not consuming CPU > when our pools are 'full'. We could turn it on now and rely on the > current trickle logic, but it's nice to have the water main off when > doing significant plumbing. So, until such time, let's be consistent in net driver land and not IRQF_SAMPLE_RANDOM. Jeff