Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S264182AbUD2MRa (ORCPT ); Thu, 29 Apr 2004 08:17:30 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264355AbUD2MR3 (ORCPT ); Thu, 29 Apr 2004 08:17:29 -0400 Received: from legolas.restena.lu ([158.64.1.34]:21697 "EHLO smtp.restena.lu") by vger.kernel.org with ESMTP id S264182AbUD2MQW (ORCPT ); Thu, 29 Apr 2004 08:16:22 -0400 Subject: Re: IO-APIC on nforce2 [PATCH] + [PATCH] for nmi_debug=1 + [PATCH] for idle=C1halt, 2.6.5 From: Craig Bradney To: ross@datscreative.com.au Cc: Jesse Allen , Len Brown , a.verweij@student.tudelft.nl, "Prakash K. Cheemplavam" , christian.kroener@tu-harburg.de, linux-kernel@vger.kernel.org, "Maciej W. Rozycki" , Jamie Lokier , Daniel Drake , Ian Kumlien , Allen Martin In-Reply-To: <200404292144.37479.ross@datscreative.com.au> References: <200404282133.34887.ross@datscreative.com.au> <20040428205938.GA1995@tesore.local> <200404292144.37479.ross@datscreative.com.au> Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="=-ZDpf+2Z8bzrhgeH+04IG" Message-Id: <1083240974.8367.14.camel@amilo.bradney.info> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Thu, 29 Apr 2004 14:16:14 +0200 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 6024 Lines: 161 --=-ZDpf+2Z8bzrhgeH+04IG Content-Type: text/plain Content-Transfer-Encoding: quoted-printable On Thu, 2004-04-29 at 13:44, Ross Dickson wrote: > On Thursday 29 April 2004 06:59, Jesse Allen wrote: > > On Wed, Apr 28, 2004 at 09:33:34PM +1000, Ross Dickson wrote: > > > >=20 > > > > It may be this board never hangs no matter what, > > > > or perhaps C1 disconnect was simply disabled in that BIOS > > > > b/c there was no option for it in Advanced Chipset Features > > > > like there is for the most recent BIOS. > > >=20 > > > Maybe other MOBO manufacturers skimp on filter caps and regulator dam= ping > > > ability and a resonance occurs in the on-board supply rails? Do Shutt= le make > > > any claims to using an improved on board regulator? Or Shuttle may ha= ve=20 > > > always programmed more time in C1 cycle handshakes if such is=20 > > > configurable?=20 > >=20 > > Do you really think so? I think there may be a resonance occuring, eve= n with=20 > > this new BIOS. I plugged in new headphones into my nforce2 onboard sou= nd, and=20 > > get a high pitched noise. Now here is where it gets weird: This noise= does=20 > > not occur on boot until sometime after the IDE driver is loaded. I als= o=20 > > believe it varies under a high load. If you disable C1 disconnect, it'= s gone. =20 > > Also I've heard a high pitched noise at certain times coming right from= the=20 > > copmuter (very faint, but I do have very good hearing, I can even hear = a hush=20 > > sounding from my router. my brother was quite astonished when I pointe= d that=20 > > out) I try to distinguish whats doing it. It could be the hard drive.= But=20 > > when I found the other sound in the head phones, I found that the sound= varies=20 > > almost in unison with the sound coming from the computer. Maybe the ID= E or=20 > > hard drive is related, but it is too much related to C1 disconnect. >=20 > I think I might break out my oscilloscope this weekend and have a look at= how=20 > clean the supply rails are around the cpu and northbridge and southbridge= .=20 > Who knows I might get lucky and see some unexpected ripple or spikes. >=20 > >=20 > > Whether it is really possible that my board can really generate this so= und, I=20 > > don't know. Though, I have once determined that resonance was occuring= in an=20 > > old system, causing unstable CPU operation. It wasn't that I heard a s= ound=20 > > coming from it =3D). But what I thought was the case was causing it, a= nd pulled=20 > > it out of the case. I ran it on the table and found it to be stable. = That=20 > > was the only thing wrong. I've also studied resonance before a bit. I= know=20 > > resonance can break systems. But to think that my board is doing emmit= ting=20 > > noise like that is pretty bizarre. >=20 > Not as bizarre as you may think. I have heard coils and even capacitors "= sing" > in years past whilst servicing electronics. >=20 > >=20 > > It may be true that this Shuttle board may have resonance problems. So= that=20 > > would indicate that they did something much like you describe by changi= ng the=20 > > C1 handshake time? Isn't that much like what your patch does? >=20 > I had not really thought about it from that perspective. Whilst my patch = cannot=20 > alter the handshake times it does prevent consecutive C1 cycles from occu= rring > too close together. Too close together I think being less than about 800n= s. I=20 > guess I could look at that with a cro too - use an appropriate pin as the= trigger > source and see if supply rails have load dump voltage rises when going in= to > disconnect. Maybe rail voltage rings for about 700ns and might be out of=20 > tolerence inside Athlon during that time. Would be very interesting if a=20 > few hundred picofarad of low esr decoupling cap placed on a supply rail n= ear a > chip makes a difference? A pinout of the nforce2 chipset would help a gre= at deal > here but I do not have one. Can anyone oblige me? >=20 > >=20 > >=20 > > >=20 > > > > hang issue is completely explained and solved. > > >=20 > > > I have had good (100%) success in reproducing the fault with the Alba= tron=20 > > > KM18G pro MOBO. I needed m-atx form factor and distributor was local = to me. > > > Makes very nice - cheap and stable system but only with the lockup wo= rkaround. > > >=20 > > > I also recollect that Windows had lockups with nforce2 for a while de= pending=20 > > > whether you ran the Nvidia or Microsoft driver. > > > http://lkml.org/lkml/2003/12/13/5 > > > Anybody got the inside running on that one and what was different bet= ween the=20 > > > two drivers? > > >=20 > >=20 > > Yeah, unfortunately, I didn't save a link to the message board that I f= ound=20 > > that on. But the issue is pretty common. I'm sure more info can be fo= und on i > > the windows side. >=20 > No tech info but this link shows user had Lockups with Nvidia's ide drive= r but > OK with MS one. > http://club.cdfreaks.com/showthread/t-91381.html >=20 > - This has become a rather interesting problem to watch from afar. The Athlon here seems to have no issues with the NForce driver under Windows (I dont burn a lot of DVDs on it tho). Whenever its in Linux, its mainly a testing machine these days. It will be interesting to see if theres a real hardware problem and then if it can be worked around in software (cant image a single product recall happening). --=-ZDpf+2Z8bzrhgeH+04IG Content-Type: application/pgp-signature; name=signature.asc Content-Description: This is a digitally signed message part -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (GNU/Linux) iD8DBQBAkPIOi+pIEYrr7mQRApmZAJ9G5/sHOLiYF3y1P08dvXnFVxCgdgCcDjvI XMDLFVdz+z7A+iW3uoVHPko= =CXCq -----END PGP SIGNATURE----- --=-ZDpf+2Z8bzrhgeH+04IG-- - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/