From owner-freebsd-current Thu Dec 26 4:59:35 2002 Delivered-To: freebsd-current@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 9D45D37B401 for ; Thu, 26 Dec 2002 04:59:33 -0800 (PST) Received: from sccrmhc01.attbi.com (sccrmhc01.attbi.com [204.127.202.61]) by mx1.FreeBSD.org (Postfix) with ESMTP id 13DF343EE5 for ; Thu, 26 Dec 2002 04:59:32 -0800 (PST) (envelope-from altea@attbi.com) Received: from dhcppc0.we.client2.attbi.com (c-24-126-197-60.we.client2.attbi.com[24.126.197.60]) by sccrmhc01.attbi.com (sccrmhc01) with SMTP id <20021226125931001003ktqie>; Thu, 26 Dec 2002 12:59:31 +0000 Subject: NVida Ideas may help. From: Joseph Altea To: David Holm Cc: current@freebsd.org In-Reply-To: <200212260057.19240.david@realityrift.com> References: <200212251454.48551.david@realityrift.com> <200212251651.16258.cbiffle@safety.net> <200212260057.19240.david@realityrift.com> Content-Type: text/plain Content-Transfer-Encoding: 7bit X-Mailer: Ximian Evolution 1.0.8 Date: 26 Dec 2002 04:58:20 -0800 Message-Id: <1040907501.4556.24.camel@dhcppc0.we.client2.attbi.com> Mime-Version: 1.0 Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Hi Becuase of the memory access model with debuging on I found that a interrupt stack overflow in the kernel seems to occur from the interaction with interrupts and the memory on the nVida card. I have not nailed it down yet to which interrupts, but without debugging it reduces it an thus the freezing of the box. But in 3D mode the card is doing alot of the work and is notifing the X system and the kernel through interupts. Seems this overwrites the kernel and cause the "Frozen" deseret effect, without the desert. Maybe we need to have a kernel parameter for the agp based cards for increasing the interrupt stack size or the like. I watched it on one of our hardware developement PC's I have with a hardware debug card in it and shows that the cards blows up the kernel from video memory DMA transfer interrupts to/from X. regards, Please check what happens on IRQ 11 on PCI2 (on my Intel based debug machines) Joseph Altea, CEO ImerXion,LLC altea@attbi.com On Wed, 2002-12-25 at 15:57, David Holm wrote: > On Thursday 26 December 2002 00:51, Cliff L. Biffle wrote: > > > > I've managed to reproduce a very similar situation using RC1. > > > > Does your kernel have various debugging features in place? If so...try > > turning them /off/. I kid you not. On the RC1 box here the nVidia drivers > > were very unstable doing 3D until the kernel was recompiled without > > debugging. (Chris, the owner of the machine, wanted to do it for > > 'performance' reasons -- getting the graphics working was an unexpected > > boon.) > > > > I can get you the kernel config, or make it available to the list if > > anyone's interested in this fleeting heisenbug. > > > > -Cliff L. Biffle > > Yes, I've turned off all the debugging stuff. I've also removed Load "dri" > from XF86Config etc. > Please post your kernel conf, maybe I can see where they differ (if they do). > > //David Holm > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-current" in the body of the message To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message