Hey all,

I have eventually whittled the problem down to this - the mem_controller
(mem_rc-f0.8-2617-rc3) causes this problem.  I have successfully
compiled the 2.6.17 kernel and booted from it on its own, with the
single resource groups patch (res_group_f0.8_single_2617) and with the
cpu controller patch (cpurc-v0.5-2617).  I also made sure I had updated
all the ancillary programs that the kernel needs (list found in <kernel
source tree>/Documentation/Changes), since I was moving from 2.4 to 2.6.

I would love to attach some sort of log file for you guys to work with,
but I'm not sure what I can do.  /var/log/messages has no record of the
attempted boot, and dmesg only has records for the most recent one.

I've attached the .config files I used to compile the kernel as well,
just in case you want to have a look.  They are based on running make
xconfig on a freshly untar'd and gunzip'd kernel source tree, and
hitting save, and then running make oldconfig.  The patches are applied
after the .config file is first created/updated.

FILE LIST:
.config-rg-stdkernel-pre-patch          .config file following fresh
untar, make xconfig, patch res-groups, make oldconfig
.config-rg-stdkernel-rgwithcpu          same as above, only patched with
cpu_rc as well
.config-rg-stdkernel-rgwithcpuandmem    same as above, only patched with
cpu_rc AND mem_rc (kernel made from this did not work.


I'm happy enough to keep working with the CPU controller on its own, but
any info I can give you guys to debug the mem controller, please ask.

Thanks,

Stephen.




-----Original Message-----
From: Matt Helsley [mailto:[EMAIL PROTECTED] 
Sent: Wednesday, August 23, 2006 6:44 PM
To: Stephen Workman (stworkma)
Cc: CKRM-Tech
Subject: Re: [ckrm-tech] Problems with
Networkafterres_group_f0.8_single_2617.gz

On Tue, 2006-08-22 at 19:38 -0700, Stephen Workman (stworkma) wrote:
> Hey Matt, all,
> 
> I've attached .config files, dmesg and ifconfig output, corresponding 
> to different kernel make options.  Standard refers to the settings 
> already in place on this PC.

OK, I'm confused. It doesn't look like any of the info you posted
corresponds to the system with a working network (none had an eth0).
Being able to compare the two would help.

Incidentally I messed up. Instead of asking for /var/log/dmesg I
should've asked for the output of the dmesg command -- it would be more
complete.

> I'm not sure why, but I've noticed that when the box is booting up, it

> starts setting up eth0, but then complains that the MAC address 
> doesn't match what was expected.  I have changed no initial script

Which distro? If it's debian-based (Debian, Ubuntu, Knoppix..) the
/etc/network/interfaces file could be helpful.

        2.6 kernels when compared to 2.4 often enumerated devices in a
different order. This means if multiple ethernet devices existed they
would have different names depending on the current kernel. One solution
for network devices on both kernels was to rely on its MAC address to
consistently configure the physical device.

>  files, and under a std kernel, it seems to work fine.  There are also

I'd guess your network configuration is looking for a device with a
specific MAC address but no such device exists since the driver failed
to load due to

>  complaints about version mismatching.
>
> Any help would be greatly appreciated.
> 
> Thanks,
> 
> Steve.

The lines ending with:

"disagrees about version of symbol struct_module"

suggests old modules are left from your previous install. I would guess
the network device module failed to load and hence the network wouldn't
come up.

You might try making the module CONFIG_ values consistent.

Or you might try setting LOCALVERSION to indicate which config you're
currently building and testing. For example:

# .config-rg-allmodconfig
CONFIG_LOCALVERSION="allmodconfig"
CONFIG_LOCALVERSION_AUTO=n

# .config-rg-allyesconfig
CONFIG_LOCALVERSION="allyesconfig"
CONFIG_LOCALVERSION_AUTO=n

# .config-rg
CONFIG_LOCALVERSION="standard"
CONFIG_LOCALVERSION_AUTO=n

This should cause the modules directory to be different for each kernel
and prevent the kernels from loading modules that weren't built for
them. (I'd also append these to the resulting /boot files just to keep
track of things)

Or you might try cleaning *everything* between each config switch. For
each build:

make mrproper
cp config-rg .config
make oldconfig
make bzImage
su
rm -rf /lib/modules/2.6.17
make modules_install
etc.

Does any of the above help? If not please send the config, ifconfig -a,
and dmesg output for a working system so I can compare.

Cheers,
        -Matt Helsley

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
ckrm-tech mailing list
https://lists.sourceforge.net/lists/listinfo/ckrm-tech

Reply via email to