Frank Morane wrote:
I still can't get a console BUT I can access via ssh the booted domU; then I saw nash using 95% of the cpu and I killed it.
The problem is that this is far from what I want since I cannot access the domU via a console, when I try I get these:
printk: 185 messages suppressed. 4gb seg fixup, process ntpd (pid 1230), cs:ip 73:008ac17a
[...]
and yes I have the "hwcap 0 nosegneg" in /etc/ld.so.conf. There is critical information missing to get older generation (Centos 4) domU booting on late FC5, FC6 and FC7.
nosegneg only works if the kernel supports it, and as far as I can find out, this capability was only added relatively recently (earlier this year).
http://www.uwsg.iu.edu/hypermail/linux/kernel/0704.2/2872.html
In addition, this flags affects how (or rather, which) glibc is dynamically linked in to the binary. /sbin/nash is statically linked, but /usr/sbin/ntpd is dynamically linked (at least on FC6, and I expect on FC5 too).
So if you're getting the above messages concerning ntpd, then it sounds a lot like you're still linking to a version of glibc which is using negative offsets with segment register overrides, which is something that Xen has to emulate, slowly.
As to how to fix this: In Debian the fix is well-known. Rename the directory /lib/tls (on 64 bit Fedora: /lib64/tls) as /lib/tls.disabled, and reboot the domU.
Rich.