On Mon, 2004-01-26 at 18:12, Andre Costa wrote: > On Sun, 25 Jan 2004 14:02:24 +0100 > Emiliano Brunetti <emiliano_brunetti@xxxxxx> wrote: > > [...] > > Well, this is weird indeed. > > > > Here it goes: [...] > > - LABEL is fully supported with RAM disk as you suggested (hitn: put > > it somewhere on a FAQ! :) ) > > Agreed. It took me a while to figure it out... =/ I have reports that it doesn't work with vanilla kernels. > > - at home, no way. Still fails to mkinitrd, and can't build a initrd > > image manually. System complains that it can't find aic7xxx module. > > However, modulues seem to be there in /lib/modules, and it is > > definitely a major problem as this is a pure SCSI box. > > Now this is really weird. I would go for a 'clean' rebuild as suggested > above to see if probl persists. I can't see no reason for a module > effectively present at /lib/modules/[...] to be loaded -- does 'depmod > -a' fail? > > Also, since you didn't seem very sure about the module being there, make > sure you're looking at the right place. The root modules dir for a > kernel is /lib/modules/[uname -r] (when the kernel is running -- > otherwise you will get the version number for the wrong kernel!). Indeed it was also a headless configuration problem. My custom kernel is performing really bad, system is almost unusable. Just FYI: - PII400 - 384 Mb SDRAM - pure SCSI system, though a little old, with two UW SCSI II disk - a very old G200 Being a pure SCSI system i never really experienced bottleneck in disk writing. And the kernel that comes out of FC1 was definitely better in performance, though being big and 'general purpose'. I will have to go for a much better kernel configuration. Any hint is absolutely welcome of course. > > I am trying one last time, compiling aic7xxx support as a module. Then > > i'll try *again* to erase and reinstall kernel source from rpm. After > > this, i'd be kinda lost. > > You could try to compile it builtin (i.e. not as a module). Still, there > is no reason for the module not to load cleanly, something's really > wrong. > > Let us know if you make any improvements. This is what happened: i can't compile it builtin, as it fails in generating a initrd, but it does compile as a module. If i compile as a module, initrd seems to be good but performance is definitly disgutsing. I need a little time to track it down. Really weird. E agora vou trabalhar. ;) Muito obrigado E.
