Re: Dmesg garbage after kernel recompile
Monique Y. Mudama wrote:
I tried to answer each of your questions one-by-one, so I sliced a bit
your post, sorry about that. :)
What about sharing some of the dmesg output, so we can take a look?
I have to refine a little my previous post.
During the boot the kernel emits only garbage (different barely
printable characters separated by spaces, like paranthesis and so).
These messages are really short, just a few characters on each line.
But after boot if I run dmesg it only displays about one character
(which varies from system to system, mostly a ., but now we also have an
'x' :).
So actually I cannot speak about dmesg output. :-)
There are some easily recognisable strings like:
ACPI initialization, and IRQ listing after
------ start ------
) e 6
4
1 d
g
tsS S1S S5)
)
)
e ]
]
]
] Q 101 121 15)*0d
Qs 3 4 5 6 7 91 111 141 )* d.
Q 101 121 15)*0d
Qs 3 4 5 6 7 91 111 141 )* d.
] e
w w
w w
------ end ------
detecting partitions (8 IDE drives, HDA1,2 HDB1,2 HDC1, HDD1...HDG1)
------ start ------
: 0:p p2
0:p p2
0:3 p1
0:3 p1
0:3 p1
0:3 p1
0:3 p1
0:3 p1
------ end -------
What are the differences between your configuration and the stock
config?
My configs are fully custom, I mean we configured the kernel from
scratch each time. So there are a lot of differences.
What may lead to a bigger diff. is that we apply the FreeSWAN and i2c
modules patch, but we tried even without the patches and no difference.
What version of 2.4.27 are you using?
We are using the stock kernel sources from kernel.org.
Are any of the
machines able to boot from a stock Debian kernel, and if so, what
happens?
They boot happily with the stock Debian kernel, they have the usual
output during the boot, and the dmesg output is how it should be.
Are you getting the kernel source from a debian package or
from kernel.org?
As I stated above straight from kernel.org.
Now that you asked I will try to compile one from a debian package,
maybe I'll succeed.
You mention having several machines; do you have a
sacrificial machine you could install from scratch?
Yes, I do have severals, but in fact I already did with some machines.
We are installing every now and then some servers with Debian, and each
install is done from scratch (to be honest we first do a woody install,
and then dist-upgrade to sarge).
As an aside, I wonder why you're using Sarge? Sarge = testing = not
intended as a user system; it would more accurately be described as
the integration phase of the Debian development cycle. (Hey, we've
been looking for a word better than "testing" for forever -- what do
you guys think about "integration"?) It's not necessarily a middle
ground between the stability of stable and the potential brokenness of
unstable. Things can stay broken in testing a lot longer than they can
in sid.
Why? Because in the last 5-6 years we didn't have any really big
problems with the testing branch.
But mainly because we are using a lot of things that are available in
sarge, and not in woody (or only by 'backports').
Some more symptoms:
the module insert/remove procedure prints also messages like in the
previous example.
I tried right now the remove ETH cable, reinsert ETH cable thingie.
It produced the following output:
remove from the 100Mb NIC: n
insert into the 100Mb NIC: x
remove from the 1Gb NIC: n
insert into the 1Gb NIC: 0 S l c d d
Prints the same string over-and-over it only varies by where the newline
is put between the characters.
Now we even tried to compile a freshly extracted 2.4.27. Only make
menuconfig was run, to produce an acceptable .config file, but no
changes were made in it. It compiled cleanly, and booting it gave the
good ol' kernel messages and dmesg output.
After that we imported our previous config file (menuconfig->Load
configuration), and made a compile. Now this reproduces the garbage.
So the problem may hide in our .config file but what is it?
To the other poster (a.list.address@gmail.com):
We need a custom kernel because we need the following features compiled in:
ReiseFS, devFS, RAID0,1,5, LVM, SMP-PIII, ACPI, cmd680-IDE, freeSWAN,
High-memory support, I2O SCSI support
In modules: iptables, eth NIC drivers, bridge, QoS.
I'll rather not bother the list with the .config file, but if somebody
feels the power to deduce the problem from the options, than ask for it,
and I'll send it in private.
Any help will be very appreciated.
Thank you
Bye
Tylla
Ps: we now made another experiment, downloaded a stock debian kernel,
imported our .config, and bang the same garbage.
Reply to: