Help-A-Newbie(tm) RH 4.2: Configure/Build/Install new kernel always fails

Help-A-Newbie(tm) RH 4.2: Configure/Build/Install new kernel always fails

Post by Samuel Lake » Sun, 06 Jul 1997 04:00:00




> Keywords: kernel, symbols, redhat 4.2, linux newbie

> I installed RH 4.2 on my Dell (GXM 5166, fwiw) today.  The install and
> initial configuration seemed to go ok.

> What I am (eventually) aiming for, is to turn this into a
> router/firewall.
> So, opening up the "Kernel HOWTO", I cd'd to /usr/src/linux, and
> proceeded
> to go through "make ()config".

> "So far so good"?  Well, I guess I must be having a really bad life.
> When I rebooted, I got >reams< of cascading messages informing me of a
> variety of "missing symbols" in /lib/modules/2.0.30/...    (Does linux store
> these boot error messages somewhere???  They're not in /var/log/*.  "dmesg"
> also doesn't show any of the error messages.  The boot goes by so quickly,
> you barely have time to notice something's wrong, before the error
> message(s) vanish offscreen!!!)

Hi Tim,

I've been having basically similar problems on my dual PPro system.
I've been trying to configure my machine for SMP under RH4.2 for the
last few days. Everything works fine, except at boot I get a bunch
of "insmod:  Can't find blah" messages, ending with ... "your symbols
in /lib/modules/2.0.30/ don't match." It sounds like we're having the
same problem (mine appear to be related to the PPP-interface, I think)

Since we're both using RH4.2, I wonder if our kernel source is somehow
less than pristine. I'm going to try updating my kernel source tree
tonight.

If you somehow solve your problem, please let me know. I'm having it
too...

Thanks,

another *nix guy still figuring out this Linux thing.

--
*****BEGIN GEEK CODE BLOCK********************************************
|This is me, in Geek (v3.12) GIT/GMU/GH, s:++,!y+**, DI++++, UH/UU++ |

|PS++(+++), PE-,  R+, tv-->---, h-(+)>++, t+, !5--, b+++, D++, d-()  |
|G++, e(+++), r++ X+>+++ Mtg++ I am not your friend. Go away.        |
+====END GEEK CODE BLOCK=============================================+
|"I wanna decide who lives and who dies" - Crow T. Robot, MST3K      |
+--------------------------------------------------------------------+

 
 
 

Help-A-Newbie(tm) RH 4.2: Configure/Build/Install new kernel always fails

Post by Robert Kiesli » Mon, 07 Jul 1997 04:00:00



(Text deleted)

Please take these "hints" with a grain of sand, because I've compiled
a different kernel version (2.0.29) and distribution (Debian).

Quote:

> On a wild guess, I tried moving a bunch of the .o files I found in /lib,
> to separate directories, and that seemed to cut down on a lot of the
> "missing symbol" messages.  (Why??  Where is this documented?  If not,
> why not?  How is a newbie supposed to just "know" these things?)

> But not all "missing symbol" messages are gone -- now I've got a problem
> with iBCS.  A friend told me that's the Binary Compatibility System for
> use with SCO.  (Yes, that's a feature I'd like to take advantage of, but
> please, could somebody kindly tell me how I was supposed to "know"
> that?)
> So now that iBCS is broken (c'mon, people; all I did was compile a some-
> what modified kernel and reboot???) how am I supposed to fix THAT?

You need to ensure three things, it seems to me.  First, that the kernel
supports both ELF and a.out format libraries and binaries.  (Redundant, I
know, but necessary for backwards compatibility.)

Second, make sure that you have the kernel daemon enabled, so that kernel
object modules can be loaded at run-time.

Third, make sure that you have support for the kernel modules enabled in
the first place.  A good number of the kernel options depend on what
you've selected previously.  For example, to enable "IP tunneling" you
need to have enabled "IP Network Support."  Do a "make config", and if you
don't know what an option specifies, type something besides y/n/m, and
config will print a paragraph of infarmation of what that particular
option enables, depends on, and/or requires.

Quote:> Ok.  Next item.  In order to turn a box into a router/firewall, it's
> gotta
> have multiple Ethernet interfaces.  So, I read the Multiple-Ethernet
> HOWTO,
> and tried its (very sort, few) suggestions, but that's also not working.

> The Dell comes with an 3com 3c509 device built-in.  I've also scrounged
> up
> a 3c503 ISA card.  Well, on bootup, Linux has seemed to find the 3c503
> on
> a regular basis (though now that I'm trying to build and boot off a new
> kernel, that also is now failing).  I added the suggested config info to
> /etc/lilo.conf (ie, append = " ... "), and also tried manually typing it
> in at the boot level, but nothing works.

> I've never succeeded in getting both cards up at the same time.
> Actually,
> I had been thinking that maybe my 3c509 was bad because it never seemed
> to
> be detected.  So I yanked out the 3c503 to check.  Only THEN did
> autoprobe
> find the 3c509.  (Please:  somebody tell me why??)

Don't know.  I've got a similar problem with an old QIC02 Tape drive.  Probably
an IRQ conflict because the board is outdated.

Quote:> Related question, is there a way to MANUALLY FORCE an autoprobe, while
> the system's running (pls don't tell me that boot time does that; I
> know)?
> What I'm driving at here is, if autoprobe on bootup is going to refuse
> to work, is there at least some way I can do something like Solaris'
> "ifconfig plumb <interface>"?  (So I can at least work around bootup's,
> autoprobe brokenness, if it's going to refuse to Do What I Want.)

Not sure about Redhat distributions... but look in /etc/rc.d, or similar
directory.  There should be a file called "modules.rc" or something
similar.  You can enable and/or disable various boot-time modprobes by
editing that file.

I would make sure that the hardware is running with an out-of-the-box
kernel before I attempted anything with the software.  Only if you have
some really off-the-wall cards should you need to enter any parameters for
them at boot-time.

HTH

Robert



...No writing gig too small...

 
 
 

Help-A-Newbie(tm) RH 4.2: Configure/Build/Install new kernel always fails

Post by Ian COWEL » Tue, 08 Jul 1997 04:00:00




> > Keywords: kernel, symbols, redhat 4.2, linux newbie

> > I installed RH 4.2 on my Dell (GXM 5166, fwiw) today.  The install and
> > initial configuration seemed to go ok.

> > What I am (eventually) aiming for, is to turn this into a
> > router/firewall.
> > So, opening up the "Kernel HOWTO", I cd'd to /usr/src/linux, and
> > proceeded
> > to go through "make ()config".

> > "So far so good"?  Well, I guess I must be having a really bad life.
> > When I rebooted, I got >reams< of cascading messages informing me of a
> > variety of "missing symbols" in /lib/modules/2.0.30/...    (Does linux store
> > these boot error messages somewhere???  They're not in /var/log/*.  "dmesg"
> > also doesn't show any of the error messages.  The boot goes by so quickly,
> > you barely have time to notice something's wrong, before the error
> > message(s) vanish offscreen!!!)

> Hi Tim,

> I've been having basically similar problems on my dual PPro system.
> I've been trying to configure my machine for SMP under RH4.2 for the
> last few days. Everything works fine, except at boot I get a bunch
> of "insmod:  Can't find blah" messages, ending with ... "your symbols
> in /lib/modules/2.0.30/ don't match." It sounds like we're having the
> same problem (mine appear to be related to the PPP-interface, I think)

> Since we're both using RH4.2, I wonder if our kernel source is somehow
> less than pristine. I'm going to try updating my kernel source tree
> tonight.

> If you somehow solve your problem, please let me know. I'm having it
> too...

> Thanks,

> another *nix guy still figuring out this Linux thing.

> --

Did you recompile the modules? See:
/usr/src/linux/Documentation/modules.txt
I am using Redhat 4.1 which uses kernel 2.0.27
So that I could go back to the old set of modules if I wanted,
I set up this structure in /lib/modules/
ls -l /lib/modules/
total 2
lrwxrwxrwx   1 root     root            8 Apr  4 06:26 2.0.27 ->
2.0.27.1/
drwxr-xr-x  10 root     root         1024 Apr  3 13:43 2.0.27.0/
drwxr-xr-x   6 root     root         1024 Apr  6 01:49 2.0.27.1/

--
Ian Cowell

----------------------

 
 
 

Help-A-Newbie(tm) RH 4.2: Configure/Build/Install new kernel always fails

Post by James Youngma » Tue, 08 Jul 1997 04:00:00



> Hi Tim,

> I've been having basically similar problems on my dual PPro system.
> I've been trying to configure my machine for SMP under RH4.2 for the
> last few days. Everything works fine, except at boot I get a bunch
> of "insmod:  Can't find blah" messages, ending with ... "your symbols
> in /lib/modules/2.0.30/ don't match." It sounds like we're having the
> same problem (mine appear to be related to the PPP-interface, I think)

> Since we're both using RH4.2, I wonder if our kernel source is somehow
> less than pristine. I'm going to try updating my kernel source tree
> tonight.

No.   I don't think you have a problem with your kernel sources.

Move the old modules out of the way before make modules_install.
Things should than be much quieter.

 
 
 

Help-A-Newbie(tm) RH 4.2: Configure/Build/Install new kernel always fails

Post by James Youngma » Tue, 08 Jul 1997 04:00:00


[Quoted material has been reformatted to be less than 80 columns wide.]

Quote:> After that, make dep, make clean, make zImage, make modules.  I
> installed vmlinuz in /boot/vml_tph, added appropriate entries
> to /etc/lilo.conf, and then did a make modules_install.

You probably wanted to move the OLD modules out of the way first.

Quote:

> "So far so good"?  Well, I guess I must be having a really bad
> life.  When I rebooted, I got >reams< of cascading messages
> informing me of a variety of "missing symbols" in
> /lib/modules/2.0.30/...  (Does linux store these boot error
> messages somewhere???  They're not in /var/log/*.  "dmesg" also
> doesn't show any of the error messages.  The boot goes by so
> quickly, you barely have time to notice something's wrong,
> before the error message(s) vanish offscreen!!!)

These mesages are likely to have been caused by the older modules
still being there too.  To see those error messages again jut
type "depmod -a".

Quote:> But not all "missing symbol" messages are gone -- now I've got
> a problem with iBCS.  A friend told me that's the Binary
> Compatibility System for use with SCO.  (Yes, that's a feature
> I'd like to take advantage of, but please, could somebody
> kindly tell me how I was supposed to "know" that?)  So now that
> iBCS is broken (c'mon, people; all I did was compile a some-
> what modified kernel and reboot???) how am I supposed to fix
> THAT?

I don't use it myself.  I suspect that you need a newer version
of the iBCS module, but -- more likely -- you just need to
recompile this module as you did the other modules.

Quote:> Ok.  Next item.  In order to turn a box into a router/firewall,
> it's gotta have multiple Ethernet interfaces.  So, I read the
> Multiple-Ethernet HOWTO, and tried its (very sort, few)
> suggestions, but that's also not working.

My firewall uses two identical 3com "Vortex" cards.  Both driven
by the 3c59x.o module.

Quote:> The Dell comes with an 3com 3c509 device built-in.  I've also
> scrounged up a 3c503 ISA card.  Well, on bootup, Linux has
> seemed to find the 3c503 on a regular basis (though now that
> I'm trying to build and boot off a new kernel, that also is now
> failing).  I added the suggested config info to /etc/lilo.conf
> (ie, append = " ... "), and also tried manually typing it in at
> the boot level, but nothing works.

You need aliases in /etc/conf.modules, perhaps something like...

alias eth0=3c509
alias eth1=3c503

(This probably means the Multiple-Ethernet HOWTO needs updating).

Quote:> I've never succeeded in getting both cards up at the same time.
> Actually, I had been thinking that maybe my 3c509 was bad
> because it never seemed to be detected.  So I yanked out the
> 3c503 to check.  Only THEN did autoprobe find the 3c509.
> (Please: somebody tell me why??)

Ethernet cards live in funny places, sometimes.  The list of
places an Ethernet driver can look for a card intersects with the
places that other stuff lives.   If you probe many locations you
increase the risk that you touch the wrong address and poke
hardware that isn't what you think it is, and lock the machine.
this is why the Ethernet probing stops by default once it has
found the first Ethernet card.

Quote:> Related question, is there a way to MANUALLY FORCE an
> autoprobe, while the system's running (pls don't tell me that
> boot time does that; I know)?  What I'm driving at here is, if
> autoprobe on bootup is going to refuse to work, is there at
> least some way I can do something like Solaris' "ifconfig plumb
> <interface>"?  (So I can at least work around bootup's,
> autoprobe brokenness, if it's going to refuse to Do What I
> Want.)

modprobe -t net

(but note, from the manpage:-

 
 
 

Help-A-Newbie(tm) RH 4.2: Configure/Build/Install new kernel always fails

Post by almon » Tue, 15 Jul 1997 04:00:00


Quote:> No.   I don't think you have a problem with your kernel sources.

> Move the old modules out of the way before make modules_install.
> Things should than be much quieter.

Yes. But it still hav problem. I create a test kernel a few days
ago and encounter the following problem.

1. do make menuconfig to change items in network
2. do make dep ; make clean ; make zImage
3. copy zImage to /boot
4. remove /lib/modules/2.0.27
5. do make modules and make modules_install
6. do lilo and add the new zImage as an option

Upon reboot, when loading the scsi modules, I got the following:

loading aic7xxxx.o
scsi_command_site:wrong version or undefined
scsi_register_module:wrong version or undefined
.
.
VFS:cannot open root device 08:02
kernel panic:VFS:Unable to mount root fs on 08:02

Can anybody tell me what I am doing wrong ? There are many people
have problem after kernel compilation. I don't need something PnP,
but as least has enough message tell me what I am wrong.

Thanks in advance.

 
 
 

Help-A-Newbie(tm) RH 4.2: Configure/Build/Install new kernel always fails

Post by Frank Sweetse » Tue, 15 Jul 1997 04:00:00



> > No.   I don't think you have a problem with your kernel sources.

> > Move the old modules out of the way before make modules_install.
> > Things should than be much quieter.

> Yes. But it still hav problem. I create a test kernel a few days
> ago and encounter the following problem.

> 1. do make menuconfig to change items in network
> 2. do make dep ; make clean ; make zImage
> 3. copy zImage to /boot
> 4. remove /lib/modules/2.0.27
> 5. do make modules and make modules_install
> 6. do lilo and add the new zImage as an option

> Upon reboot, when loading the scsi modules, I got the following:

> loading aic7xxxx.o
> scsi_command_site:wrong version or undefined
> scsi_register_module:wrong version or undefined
> .
> .
> VFS:cannot open root device 08:02
> kernel panic:VFS:Unable to mount root fs on 08:02

> Can anybody tell me what I am doing wrong ? There are many people
> have problem after kernel compilation. I don't need something PnP,
> but as least has enough message tell me what I am wrong.

So, you're running your root off of a scsi device, with the scsi driver as
a module, correct?  If so, you have to rerun the initrd script.  The
initial ramdisk image that's being loaded still contains the old aic
module, and the new one won't be used until your run initrd.

--
Windows: I can play Doom!              |RedHat Linux 2.0.31pre-2 i486
Linux: I can be a file server, be a Web|Because reboots are for upgrades!
server, run the accounting package with|http://www.wpi.edu/~rasmusin/pgp.html
twelve terminals AND play Doom!        |for pgp key.        frank sweetser

 
 
 

Help-A-Newbie(tm) RH 4.2: Configure/Build/Install new kernel always fails

Post by Ed Grim » Tue, 15 Jul 1997 04:00:00




>> Keywords: kernel, symbols, redhat 4.2, linux newbie
>> "So far so good"?  Well, I guess I must be having a really bad life.
>> When I rebooted, I got >reams< of cascading messages informing me of
>> a variety of "missing symbols" in /lib/modules/2.0.30/...    (Does
>> linux store these boot error messages somewhere???  They're not in
>> /var/log/*.  "dmesg" also doesn't show any of the error messages.
>> The boot goes by so quickly, you barely have time to notice
>> something's wrong, before the error message(s) vanish offscreen!!!)

> I've been having basically similar problems on my dual PPro system.
> I've been trying to configure my machine for SMP under RH4.2 for the
> last few days. Everything works fine, except at boot I get a bunch
> of "insmod:  Can't find blah" messages, ending with ... "your symbols
> in /lib/modules/2.0.30/ don't match." It sounds like we're having the
> same problem (mine appear to be related to the PPP-interface, I think)

> Since we're both using RH4.2, I wonder if our kernel source is somehow
> less than pristine. I'm going to try updating my kernel source tree
> tonight.

> If you somehow solve your problem, please let me know. I'm having it
> too...

Sam, I expected better from you...  Admittedly, it's damned esoteric.  I
haven't even found out how to reproduce it at will...  I can prevent it,
though, which is what I'm going to tell everybody here, and I know two
of the required factors.

The magic file is called System.map, and it's left sitting lonely in
your top-level kernel source directory.  It needs a home, preferably in
/ or /boot.  My system has gotten to be silly with these things, since I
have a bunch of boot configurations which had problems, running
different kernels & distributions.  (The distribution doesn't matter -
it's only the kernel version & config that matter.  Which modules are
added doesn't matter, but which features are compiled into the kernel
definitely do.)  My system's /etc/rc.d/rc.S, /etc/init.d/boot, and
/etc/rc/boot (Slackware, Debian (been a while, I'm not certain of the
name) and Frank & Stein) actually do some checks to determine which
System.map to use, and make a symlink.  (It also makes a symlink for the
modules & another for the linux source.)

However, for normal use, that's completely unnecessary.  When you
upgrade your kernel, back up the old system map (most distribs put it in
/boot), and move the new one over.  Install your new modules; probably
is a good idea to move your old modules to another directory - this lets
you go back easily, and it makes certain that you won't have any old
modules lying around.  Reboot, and you should have no problems.

The error messages you get are sent to stderr, I believe, which isn't
logged by any of the current distributions I've seen.  However, you can
run the command which caused them, it's located somewhere inside
/etc/rc.d/rc.modules, /etc/init.d/modules, or someplace similar (Been
too long since I've done Red Hat for me to give the location there.),
the command is to rebuild module dependancies, and it's commented as
such.  grep -n depend /etc/*.d/{rc.,}modules, and it should tell you
aproximately where it is.  It may very well be 'depmod -a', but my
memory is fuzzy.

Btw, if you're lazy, you can "fix" the problem by commenting out the
depmod (or whatever) line from your module startup file, but in this
case, your module dependancies won't work.  (Of course, they're already
not working, so you're no worse off.)  This is the other factor I know
about.

--
Ed Grimm

 
 
 

Help-A-Newbie(tm) RH 4.2: Configure/Build/Install new kernel always fails

Post by Alexander Hvosto » Tue, 22 Jul 1997 04:00:00


Hi...

Good thing I read this post!  I KNEW there was something I wasn't doing
with my modules... I abandoned them after something seemed not to work.

Alex



> [Quoted material has been reformatted to be less than 80 columns wide.]

> > After that, make dep, make clean, make zImage, make modules.  I
> > installed vmlinuz in /boot/vml_tph, added appropriate entries
> > to /etc/lilo.conf, and then did a make modules_install.

> You probably wanted to move the OLD modules out of the way first.

> > "So far so good"?  Well, I guess I must be having a really bad
> > life.  When I rebooted, I got >reams< of cascading messages
> > informing me of a variety of "missing symbols" in
> > /lib/modules/2.0.30/...  (Does linux store these boot error
> > messages somewhere???  They're not in /var/log/*.  "dmesg" also
> > doesn't show any of the error messages.  The boot goes by so
> > quickly, you barely have time to notice something's wrong,
> > before the error message(s) vanish offscreen!!!)

> These mesages are likely to have been caused by the older modules
> still being there too.  To see those error messages again jut
> type "depmod -a".

> > But not all "missing symbol" messages are gone -- now I've got
> > a problem with iBCS.  A friend told me that's the Binary
> > Compatibility System for use with SCO.  (Yes, that's a feature
> > I'd like to take advantage of, but please, could somebody
> > kindly tell me how I was supposed to "know" that?)  So now that
> > iBCS is broken (c'mon, people; all I did was compile a some-
> > what modified kernel and reboot???) how am I supposed to fix
> > THAT?

> I don't use it myself.  I suspect that you need a newer version
> of the iBCS module, but -- more likely -- you just need to
> recompile this module as you did the other modules.

> > Ok.  Next item.  In order to turn a box into a router/firewall,
> > it's gotta have multiple Ethernet interfaces.  So, I read the
> > Multiple-Ethernet HOWTO, and tried its (very sort, few)
> > suggestions, but that's also not working.

> My firewall uses two identical 3com "Vortex" cards.  Both driven
> by the 3c59x.o module.

> > The Dell comes with an 3com 3c509 device built-in.  I've also
> > scrounged up a 3c503 ISA card.  Well, on bootup, Linux has
> > seemed to find the 3c503 on a regular basis (though now that
> > I'm trying to build and boot off a new kernel, that also is now
> > failing).  I added the suggested config info to /etc/lilo.conf
> > (ie, append = " ... "), and also tried manually typing it in at
> > the boot level, but nothing works.

> You need aliases in /etc/conf.modules, perhaps something like...

> alias eth0=3c509
> alias eth1=3c503

> (This probably means the Multiple-Ethernet HOWTO needs updating).

> > I've never succeeded in getting both cards up at the same time.
> > Actually, I had been thinking that maybe my 3c509 was bad
> > because it never seemed to be detected.  So I yanked out the
> > 3c503 to check.  Only THEN did autoprobe find the 3c509.
> > (Please: somebody tell me why??)

> Ethernet cards live in funny places, sometimes.  The list of
> places an Ethernet driver can look for a card intersects with the
> places that other stuff lives.   If you probe many locations you
> increase the risk that you touch the wrong address and poke
> hardware that isn't what you think it is, and lock the machine.
> this is why the Ethernet probing stops by default once it has
> found the first Ethernet card.

> > Related question, is there a way to MANUALLY FORCE an
> > autoprobe, while the system's running (pls don't tell me that
> > boot time does that; I know)?  What I'm driving at here is, if
> > autoprobe on bootup is going to refuse to work, is there at
> > least some way I can do something like Solaris' "ifconfig plumb
> > <interface>"?  (So I can at least work around bootup's,
> > autoprobe brokenness, if it's going to refuse to Do What I
> > Want.)

> modprobe -t net

> (but note, from the manpage:-

 
 
 

1. RedHat 4.2 upgrade--kernel build fails

Hello.

System is:  

SuperMicro P5MMS MB, AMD K5 PR120 CPU, Adaptec 2940U SCSI, 2.1 GB Micropolis
HD, 64 MB EDO RAM, decent video, yadda, yadda, yadda...

Fresh install of RedHat 4.1 from CD-ROM.  Recent upgrade to RedHat 4.2
from hard drive (ftp'd down from redhat mirror).

Problem is:

% make config ; make dep ; make clean

all works just fine.

% make zImage

yields this result:


gcc -D__KERNEL__ -I/usr/local/src/linux/include -Wall -Wstrict-prototypes \
-O2 -fomit-frame-pointer -D__SMP__ -pipe -fno-strength-reduce -m486 \
-malign-loops=2 -malign-jumps=2 -malign-functions=2 -DCPU=586 -D__SMP__  \
-c -o init/main.o init/main.c

[at which point about a minute later, I hit ctrl-c and...]

make: *** Deleting file `init/main.o'
make: *** [init/main.o] Interrupt


So gcc dies trying to build the very first kernel file.  Any other file
in the kernel build ends up the same way (I checked).  gcc does work, as
a little hello world program still works just fine, but...

Any help will be appreciated.

Regards,
Ross Bagley

2. Additional question on network cards

3. Newbie configuring RHat 4.2 : help for www, mail server setups

4. How to use ioctl on CDROM Device /dev/cd0

5. RH 4.2 LILO boot problems. Newbie needs help.

6. No sound Redhat 8.0 + Soundblaster 16 PCI

7. RedHat 4.2 Install help for a newbie. Please?

8. wuftp statistics programs

9. Need help new install RedHat 4.2 Boots forever

10. RH 4.1 & 4.2 sendmail local delivery fails

11. Help: NIS failed after installed a new kernel

12. RH 8.0 kernel modules build fails

13. RH 4.2 installation failed....