|
|
|
|
|
|
|
|
|
|
xen-users
Re: [Xen-users] inittab corrupted/serial console broken on Xen-ified ker
On Mon, 17 Dec 2007, Evan Lavelle wrote:
David Stone wrote:
Here's what I do on an F7 system with a built Xen 3.2 and Xenified
Linux kernel from the Xen soure: (2.6.18):
Add the folloing line to /etc/inittab (after the rest of the getty
lines)
s1:12345:respawn:/sbin/agetty -L 115200 ttyS0 vt100
Add "ttyS0" to /etc/securetty
(Reboot)
For Xen/Dom0:
On Xen "command line" (in grub.conf) add: com1=115200,8n1
On Xenified Linux "command line" (in grub.conf) add: console=ttyS0,
115200
It works for me no problem...it looks like you're doing more or less
the same thing?
Yes, looks pretty much the same:
grub.conf:
kernel /xen.gz-2.6.21-2952.fc8 com1=115200,8n1
aren't you missing the console=com1 part of the above line? Without that I
think you're likely to be screwed. You're not telling XEN to grab the
serial port, and the default XENified kernels don't usually have "real"
serial port drivers in them... at least they didn't used to.... what is
confusing is that your dmesg output from
http://lists.xensource.com/archives/html/xen-users/2007-12/msg00486.html
shows ttyS1 ... but NO ttyS0, which pretty much explains why agetty can't
open it and keeps exiting... causing init to whine about it respawning too
fast. (but without serial drivers, I wouldn't have expected to see a
ttyS1, which is confusing the daylights out of me.)
you might try using mgetty instead of agetty, it has better logging
(better everything except size and simplicity:). You can also try a simple
echo testing > /dev/ttyS0
and see if that generates output or a "no such device" error.
module /vmlinuz-2.6.21-2952.fc8xen ro root=/dev/VolGroup00/LogVol01
rhgb quiet console=ttyS0,115200n81
inittab:
co:12345:respawn:/sbin/agetty ttyS0 115200 vt100
<snip>
I've got several pairs of xen machines, where each is connected to the
console of the other. AFAIK that doesn't work with the default xen linux
kernel builds, due to the missing serial drivers. Basically XEN replaces
the serial layer so it can provide a fake serial console that knows about
the hypervisor... at least it used to. It would be really nice if the
situation has improved to the point where the hypervisor and linux can
both dump console output on one serial port, and you can actually use a
different serial port with minicom. (Maybe that's what xvc0 is about!?) In
my experience, you need to custom compile a kernel, and NOT give XEN the
console=com1 parameter to let the linux console run on ttyS0 and be able
to access ttyS1 from minicom ... but AFAIK, that is rambling and not
related to your question.
-Tom
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
|
|
|
|