From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail.gmx.net (pop.gmx.net [213.165.64.20]) by ozlabs.org (Postfix) with SMTP id B68102BF0E for ; Mon, 3 Jan 2005 08:33:38 +1100 (EST) From: Sebastian Heutling To: Sven Luther In-Reply-To: <20050102093908.GA25739@pegasos> References: <200412261435.13562.marc.dietrich@ap.physik.uni-giessen.de> <41CF77A9.9040405@gmx.net> <20041231095557.GA16480@pegasos> <1104505249.6794.5.camel@weizen.left.earth> <20041231222808.GB8078@pegasos> <1104623790.10369.14.camel@weizen.left.earth> <20050102093908.GA25739@pegasos> Content-Type: text/plain; charset=ISO-8859-15 Date: Sun, 02 Jan 2005 22:33:37 +0100 Message-Id: <1104701617.7898.12.camel@weizen.left.earth> Mime-Version: 1.0 Cc: linuxppc-dev@ozlabs.org, Marc Dietrich Subject: Re: [FYI] linux 2.6 still not working with PReP (ppc32) Reply-To: sheutlin@gmx.de List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Sun, 2005-01-02 at 10:39 +0100, Sven Luther wrote: > On Sun, Jan 02, 2005 at 12:56:29AM +0100, Sebastian Heutling wrote: > > On Fri, 2004-12-31 at 23:28 +0100, Sven Luther wrote: > > > On Fri, Dec 31, 2004 at 04:00:49PM +0100, Sebastian Heutling wrote: > > > > On Fri, 2004-12-31 at 10:55 +0100, Sven Luther wrote: > > > > > On Mon, Dec 27, 2004 at 03:47:05AM +0100, Christian wrote: > > > > > > Marc Dietrich schrieb: > > > > > > >Hi all, > > > > > > > > > > > > > >I read this thread about booting a PowerStack with newer kernels. > > > > > > >Motivated by this, I tried to compile a 2.6.10 kernel on my machine (with > > > > > > >sym53c825a onboard). Sadly, the scsi controller always timeouts during > > > > > > >initialisation. I > > > > > > > > > > I also have some strange issues with network support using the standard debian > > > > > 2.6.8 kernel. ifconfig upping the decchip ethernet on the powerstack freezes > > > > > the kernel, and i hear reports of problems on IBM PReP hardware too with not > > > > > working ethernet, altough there is no freeze. > > > > > > > > > > Any idea on this ? > > > > > > > > Does it contain the IRQ patch? Where does it boot from SCSI or net? What > > > > PCI-Devices are installed? > > > > > > Probably not, just the standard debian kernel, i donĂ't remember any irq > > > patches, but there where a bunch of patches from leighbb. it is done > > > > The patch is the one that is included in 2.6.10 which is mentioned above > > and this thread already has somewhere in the beginning. I attached it > > below. > > I haven't had the time yet to test the 2.6.10 kernel. Maybe there is > > another problem now ... > > The one with three lines of modified irq lines from tom rini ? I found it in > google, but i also saw another bunch of prep changes which got in 2.6.9 i Hmmm ... I didn't realise any other prep changes in 2.6.9 > think. Since debian is still using 2.6.8 and a bunch of backports, these may > be relevant, i think. You mean debian/testing? > > Also, i think for some reason i was unsubscribed while the list moved to > ozlab, and lost maybe the start of the thread, thanks for the attachement. That explains a lot ;-) > > > > netbooting, and the only pci device installed is the cirruss logic graphic > > > card. > > > > If you use the standard debian kernel - ok, it doesn't work because of > > wrong IRQs on the PowerstackII (Utah). > > Which i will now fix now that i am aware of the problem. There seem to be also > a problem on IBM prep hardware, may it be related ? Looking at the IRQ tables in prep_pci.c it very possible. There are a lot structures which have a lot of low numbered slots initialised to 0. Might be that its the same problem: PCI slotnumbering has changed resulting in wrong used entries in this tables. People should look at this particular place if they have problems like lost interrupts, kernel output saying that a device got IRQ 0, timeouts or something like that. > > > BTW: The onboard Winbond IDE chip works as well although no connector is > > provided. You only need a little bit of soldering and adjusting the IRQ > > of Slot 1 from 0 to 4. :-) > > Ah, wondered about that, and saw the slots are not soldered. I wonder why go > to all the trouble including a IDE chip, but not solder the connectors. Seems > to me like peanut savings or something. I am not sure but I think there are two versions of this machine: One with only IDE connectors and one with only SCSI connector. But to be honest: I don't know as well why they didn't solder all connectors. Sebastian Heutling