From owner-freebsd-net@FreeBSD.ORG Thu Feb 2 19:10:42 2012 Return-Path: Delivered-To: freebsd-net@freebsd.org Received: from mx2.freebsd.org (mx2.freebsd.org [IPv6:2001:4f8:fff6::35]) by hub.freebsd.org (Postfix) with ESMTP id ED8201065670 for ; Thu, 2 Feb 2012 19:10:42 +0000 (UTC) (envelope-from melifaro@FreeBSD.org) Received: from dhcp170-36-red.yandex.net (freefall.freebsd.org [IPv6:2001:4f8:fff6::28]) by mx2.freebsd.org (Postfix) with ESMTP id 76CA614FF5C; Thu, 2 Feb 2012 19:10:41 +0000 (UTC) Message-ID: <4F2ADF5C.1030900@FreeBSD.org> Date: Thu, 02 Feb 2012 23:09:16 +0400 From: "Alexander V. Chernikov" User-Agent: Mozilla/5.0 (X11; FreeBSD amd64; rv:8.0) Gecko/20111117 Thunderbird/8.0 MIME-Version: 1.0 To: Ryan Stone References: <4F29A464.3080302@zonov.org> <4F29E2C8.5000909@FreeBSD.org> <4F2A2EAB.3010700@zonov.org> <1446971288.20120202105912@yandex.ru> <4F2AB0A9.70905@FreeBSD.org> In-Reply-To: Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit Cc: freebsd-net@freebsd.org, =?UTF-8?B?0JrQvtC90YzQutC+0LIg0JXQstCz0LXQvdC40Lk=?= , Andrey Zonov Subject: Re: netisr defered - active only one thread X-BeenThere: freebsd-net@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Networking and TCP/IP with FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 02 Feb 2012 19:10:43 -0000 On 02.02.2012 21:07, Ryan Stone wrote: > 2012/2/2 Alexander V. Chernikov: >> P.S. it is also reasonable to set net.isr.bindthreads to 1 > > I really don't recommend setting this in any release. There is > currently a bug with binding kernel threads that causes unrelated > threads to be unnecessarily bound to CPUs. In the specific case of > net.isr.bindthreads this will cause(among others) the softclock > threads to all be bound to the same CPU, which may well obviate any > performance increase from binding the netisr threads. > As far as I understand, the only effect of setting bindthreads to 1 causes intr_event_bind() to bind soft netisr to appropriate CPU. Can you point me to ML discussion or some other info clarifying this bug? Okay, so one can manually bind those netisrs to CPUs using `procstat -t 0` and cpuset -l X -t ... -- WBR, Alexander