Recently I wrote:
> Hiho
>
> Hans Lermen <lermen@elserv.ffm.fgan.de> wrote:
>
> > Hi,
> >
> > I guess we need help from Bjorn and Alan.
> >
> > On Mon, 26 Feb 1996, Matthias Sattler wrote:
> >
> > > Hiho
> > >
> > > I've encountered a strange phenomenon. When I use the rootramdisk patches
> > > (bzImage+initrd-1.3.66c) with the 1.3.68 kernel my machine is significantly
> > > less stable. The errors occur in areas that were not changed by the patch?!
> > > Here in one example that I can easily reproduce:
> > > nohup xpilots -noQuit &
> > > xpilot
> > > Now play a bit (a few seconds seem enough) and quit xpilot. As a result I get
> > > the following Oops that I do not get with plain 1.3.68.
>
> > Ne need some more help from you to track this down. In order
> > to isolate the problem, can you please check:
> >
> > 1. What is your system setup _without_ the patch (vanilla 1.3.68) ?
>
> Plain 1.3.68.
>
> >
> > 2. Does the Oops also happen _with_ the patch, but having the same
> > setup as with vanilla 1.3.68 ?
> > If it does, can you please test the same without the initrd part of
the
> > patch, hence testing only the high-load stuff.
> > (bzImage parts starts at top of the patch file until
> > arch/i386/boot/tools/build.c (inclusive) )
>
> I've tested several scenarios:
> Common to all is the patched (bzImage+initrd-1.3.66c) 1.3.68 kernel.
>
> 1.) With kerneld support enabled, kernel loaded high with ramdisk but boot-
> device modules (scsi) loaded by hand (linuxrc was a link to /bin/bash).
> => error
> 2.) With kerneld support enabled, bootdevice compiled in, kernel loaded
> low (zImage), ramdisk feature not used.
> => no error
> 3.) With kerneld support enabled, bootdevice compiled in, kernel loaded
> high (bzImage), ramdisk feature not used.
> => error !!!
> 4.) Without kerneld support enabled, everything compiled in, kernel loaded
> high (bzImage), ramdisk feature not used.
> => no error !!!
>
> The last two seem most informative to me. It seems that kerneld and the
> kernel loaded high don't like each other.
And it gets even stranger. Now that I've compiled a kernel without IPX
support the errors do not show up any more! I will try today if
'Full internal IPX network' matters or not (during my tests it was
configured in).
Matthias
O .---------------. .___________. O
/\/ . `. m_sattle@ ,' / \ +FAX . \/\
__..--- ' /\/ | `._________,' | (___)/ * * \(___) \/ \ ` ---..__
""---__ \/`. | informatik. | / | \ +49 (0)6333 ,'\/ __---""
`.. / | .uni-kl.de | | `...' | -65079 \ ...'
`---------------' `._____.'
--> Don't take life too seriously -- you'll never get out of it alive. <--