From owner-freebsd-current@FreeBSD.ORG Thu Jun 7 10:57:11 2007 Return-Path: X-Original-To: freebsd-current@freebsd.org Delivered-To: freebsd-current@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [69.147.83.52]) by hub.freebsd.org (Postfix) with ESMTP id 9F7BB16A46B for ; Thu, 7 Jun 2007 10:57:11 +0000 (UTC) (envelope-from jhs@berklix.org) Received: from thin.berklix.org (thin.berklix.org [194.246.123.68]) by mx1.freebsd.org (Postfix) with ESMTP id 0018813C4B0 for ; Thu, 7 Jun 2007 10:57:10 +0000 (UTC) (envelope-from jhs@berklix.org) Received: from js.berklix.org (p549A6D98.dip.t-dialin.net [84.154.109.152]) (authenticated bits=128) by thin.berklix.org (8.12.11/8.12.11) with ESMTP id l57AIInO090284; Thu, 7 Jun 2007 12:18:19 +0200 (CEST) (envelope-from jhs@berklix.org) Received: from fire.jhs.private (fire.jhs.private [192.168.91.41]) by js.berklix.org (8.13.6/8.13.6) with ESMTP id l57AIDm9046635; Thu, 7 Jun 2007 12:18:13 +0200 (CEST) (envelope-from jhs@berklix.org) Received: from fire.jhs.private (localhost.jhs.private [127.0.0.1]) by fire.jhs.private (8.13.6/8.13.6) with ESMTP id l57AJCBS086357; Thu, 7 Jun 2007 12:19:12 +0200 (CEST) (envelope-from jhs@fire.jhs.private) Message-Id: <200706071019.l57AJCBS086357@fire.jhs.private> to: freebsd-current@freebsd.org In-reply-to: <499c70c0706070210v39f7016hbd80e9780902e992@mail.gmail.com> References: <499c70c0706070210v39f7016hbd80e9780902e992@mail.gmail.com> Comments: In-reply-to "Abdullah Ibn Hamad Al-Marri" message dated "Thu, 07 Jun 2007 12:10:31 +0300." Date: Thu, 07 Jun 2007 12:19:12 +0200 From: "Julian H. Stacey" Cc: Abdullah Ibn Hamad Al-Marri Subject: Re: Why not remove polling(4) from 7.0? X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 07 Jun 2007 10:57:11 -0000 "Abdullah Ibn Hamad Al-Marri" wrote: > Hello Guys, > > Almost all cases polling(4) adds additional latency. > There are some cases that polling(4) helps a little but most cases it wouldn't. > > So why not remove it or switch to adaptive polling as em(4) instead of > resorting to polling? I don't know the different polling strategies, but would caution before potentially making it yet More difficult for old laptops. /boot/loader.conf hw.pcic.irq="0" is needed on 4.11-RELEASE on a laptop here else no pcmcia ethernet. I know this isn't mobile@, but as topic raised here: I have 6 laptops: 1 only, new, can run 6.2-RELEASE! 1 (Digital HiNote Ultra2000) stuck on old 5.1+ -current I recall, 4 stuck on 4.11 (pcmcia ether &/or disc access problems) FreeBSD-5 & 6 wont boot/install for various reasons, mostly pcmcia & disc. Running current would be overly ambitious till I raised some of those 4 boxes. Losing polling sounds possibly scarey ? I append notes in case anyone has additional suggestions/corrections for me to try (OK, I havent tried all this on all laptops, just most on one): (by chance a Dell Latitude XPi P133ST - disc now works - pcmcia not seen.) beastie_disable="yes" boot_verbose="yes" verbose_loading="yes" debug.bootverbose=1 hint.acpi.0.disabled="1" # set to boot with acpi disabled. Above is set by forth/beastie.4th:bootsafekey loader.acpi_disabled_by_user=1 # Above is set by forth/beastie.4th:bootsafekey hint.apm.0.disabled="1" # found on host=fire with kenv. hint.apic.0.disabled="1" # set to boot in safe mode Above set by forth/beastie.4th:bootsafekey hw.ata.ata_dma=0 # ATA disk DMA mode control needed by Dell Latitude XPi P133ST says Ian Freilich , 07.10.2004 Above referenced by forth/beastie.4th:bootsafekey hw.ata.atapi_dma=0 # ATAPI device DMA mode control Above referenced by forth/beastie.4th:bootsafekey hw.ata.wc=0 # ATA disk write caching Above referenced by forth/beastie.4th:bootsafekey hw.eisa_slots=0 # Above referenced by forth/beastie.4th:bootsafekey hint.kbdmux.0.disabled=1 Above is referenced by forth/beastie.4th:bootsafekey hw.pci.enable_io_modes=0 # use if boot hangs after agp hw.pcic.ignore_pci=1 # ignore pci cardbus bridges hw.pcic.intr_path="1" # Needed by host=lapa Normally interrupts for cardbus bridges are routed over the PCI bus (2). However, some laptops will hang when using PCI interrupts due to bugs in this code. Those bugs can be worked around by forcings ISA interrupts (1). Warner wrote: Libretto-50 and Libretto-70 machines have only ISA PCMICA bridges hw.pcic.irq="0" # Needed by host=lapa Override IRQ configured by system for all pcic devices hw.pcic.pd6729_intr_path Determine interrupt path or method for Cirrus Logic PD6729 and similar I/O space based pcmcia bridge. Chips on a PCI expansion card need a value of 2, while chips installed in a laptop need a value of 1 (which is also the default). This is similar to hw.pcic.intr_path, but separate so that it can default to ISA when intr_path defaults to PCI. hw.pcic.ti12xx_enable_pci_clock Some TI-12xx parts need to have the PCI clock enabled. These designs do not provide a clock themselves. Most of the reference boards have the required oscillator parts, so the number of machines that needs this to be set is vanishingly small. machdep.bios.pci=disable # machdep.bios.pnp=disable # machdep.pccard.pcic_irq="0" # PCCARD controller IRQ (0=polled) machdep.pccard.pcic_irq=11 ------ Julian -- Julian Stacey. Munich Computer Consultant, BSD Unix C Linux. http://berklix.com