[CentOS] Several kernel problems

Deim Ágoston

ago at lsc.hu
Thu Aug 4 22:52:05 UTC 2005


hi all,

I have to configure an Intel desktop motherboard based "server" with 4
disks, using linux software raid (2 raid1 arrays). The type of motherboard
is Intel Desktop Board D925XEBC2 "BlackCanyon". The gigabit ethernet
controller is Marvell Yukon based (Syskonnekt) and stock kernel has
problems using it. It doesn't complain, lspci "sees" it (it's not a big
thing at all) but nothing under /proc/net/sk98lin/ (there should be an
eth0 at least). I upgraded the stcok kernel to the newer one, 2.6.9-11 EL
kernel but it doesn't work too. I tried to compile it as a module and to
compile the driver into the kernel.
After I downloaded the tar.gz and newest driver from marvell.com and
compiled it against the official EL kernel source it worked... for 2
seconds. After that I got a kernel panic. I downloaded the latest vanilla
kernel, version 2.6.12.3, compiled it with the patched driver. Right now
the kernel sees the ethernet card but there are three problems:
- this kernel is not the overpatched, stress tested kernel source (lots of
applications and 3rd party drivers require RH kernels) but I can live with
it
- the boot is terribly sloooooooow. at startup when udev starts the boot
process hangs for a while. That should be the version mismatch between
kernel version (2.6.12.3) and udev (latest is 0.60, CentOS/RHEL includes
0.39). As far as I've seen the changelogs and lkml archives an udev
upgrade should help. But I don't want to replace everything.
- it seems that the system constantly uses the disks! The HDD activity LED
full time active. If I boot the stock kernel it doesn't do so this was
after the upgrade. anybody seen such thing before? (Maybe that's udev
related too but I don't think so). It seems that there's no constant
writes but it's strange. I used dstat to get this result maybe there's a
better option.

Any comments appriciated.

bye,
Ago




More information about the CentOS mailing list