devil-linux-develop — developer discussions

Re: [Devil-linux-develop] errors in 1.1.1 runtime: smbmount, hotplug

[Diego T. <dt...@co...>]

On Wed, Dec 31, 2003 at 03:51:33PM -0500, Bruce Smith wrote:
> I don't have any SMB shares to mount, but if I type just 
> "smbmount" (no parameters) I get the same error:
> 
> smbmount: relocation error: /usr/lib/samba/charset/CP850.so: undefined symbol: smb_register_charset

shit... that was working a couple of days ago :)

-- 
-- gnupg keyfingerprint -- 48AF 5BF9 8F54 2966 64CC  2327 7CD0 DD91 B09D 5799
-- Use of a keyboard or mouse may be linked to serious injuries or disorders.  
Diego Torres - dtorres at anthalia dot org - Madrid / España

Re: [Devil-linux-develop] errors in 1.1.1 runtime: smbmount, hotplug

[Bruce S. <bw...@ar...>]

> > - smbmount does not work, returns error 
> > "/usr/lib/samba/charset/CP850.so: undefined symbol: 
> > smb_register_chgarset" when trying to mount an smbshare.
> 
> Can't verify this one, can somebody else check on it.

I don't have any SMB shares to mount, but if I type just 
"smbmount" (no parameters) I get the same error:

smbmount: relocation error: /usr/lib/samba/charset/CP850.so: undefined symbol: smb_register_charset

 - BS

Re: [Devil-linux-develop] errors in 1.1.1 runtime: smbmount, hotplug

[Heiko Z. <he...@zu...>]

Tim Tait wrote:

>
> I updated from CVS last night, and updated all source. The build wnet 
> fine, but I have a few problems, maybe they are my fault:

There were a lot of changes lately, so I could have screwed it up, too. ;-)

>
> - smbmount does not work, returns error 
> "/usr/lib/samba/charset/CP850.so: undefined symbol: 
> smb_register_chgarset" when trying to mount an smbshare.

Can't verify this one, can somebody else check on it.

> - hotplug pci.rc does not have executable permissions, I assume it 
> needs it?

That one is fixed in CVS.

Heiko

[Devil-linux-develop] errors in 1.1.1 runtime: smbmount, hotplug

[Tim T. <t....@co...>]

I updated from CVS last night, and updated all source. The build wnet 
fine, but I have a few problems, maybe they are my fault:

- smbmount does not work, returns error 
"/usr/lib/samba/charset/CP850.so: undefined symbol: 
smb_register_chgarset" when trying to mount an smbshare.

- hotplug pci.rc does not have executable permissions, I assume it needs it?

Thanks-

Tim

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Heiko Z. <he...@zu...>]

Bruce Smith wrote:

>>>>>Now when I modprobe cipcb0, I get:
>>>>>
>>>>>Unable to handle kernel NULL pointer dereference at virtual address 00000074
>>>>><followed by a big register dump>
>>>>>Segmentation fault
>>>>>
>>>>>At least there are no longer any undefined symbols!  ;-)
>>>>>
>>>>>          
>>>>>
>>>>Did you update from CVS ?
>>>>        
>>>>
>>>Of course!  :-)
>>>
>>>      
>>>
>>>>Are there any lines left, with CFLAGS ? If yes, remove 'em.
>>>>        
>>>>
>>>Nope.  I simply removed the tmp/cipe* directory, untared it again,
>>>removed the tmp/.done*cipe* files, ran scripts/build.sh build opt=cipe
>>>and then ran:  make clean install iso
>>>
>>>      
>>>
>>I can't try it right now, but you can:
>>add the line "disable_et_dyn" before any configure or make parts in the 
>>cipe script and let it recompile.
>>
>>Let's see what happens then.
>>    
>>
>
>It already had the line before the ./configure, so I added another
>before the make, and another one in the install section.  No difference.
>
>I'll try it with a clean lfssystem & mrproper later.
>  
>
It is necessary only once per script. This command, you find it in 
build/scripts/functions, actually does nothing else then replacing the 
gcc specs file.
Before each script execution, the file is replaced by the default one ( 
enable_et_dyn ).

Heiko

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Bruce S. <bw...@ar...>]

> >>>Now when I modprobe cipcb0, I get:
> >>>
> >>>Unable to handle kernel NULL pointer dereference at virtual address 00000074
> >>><followed by a big register dump>
> >>>Segmentation fault
> >>>
> >>>At least there are no longer any undefined symbols!  ;-)
> >>>
> >>Did you update from CVS ?
> >
> >Of course!  :-)
> >
> >>Are there any lines left, with CFLAGS ? If yes, remove 'em.
> >
> >Nope.  I simply removed the tmp/cipe* directory, untared it again,
> >removed the tmp/.done*cipe* files, ran scripts/build.sh build opt=cipe
> >and then ran:  make clean install iso
> >
> I can't try it right now, but you can:
> add the line "disable_et_dyn" before any configure or make parts in the 
> cipe script and let it recompile.
> 
> Let's see what happens then.

It already had the line before the ./configure, so I added another
before the make, and another one in the install section.  No difference.

I'll try it with a clean lfssystem & mrproper later.

 - BS

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Heiko Z. <he...@zu...>]

Bruce Smith wrote:

>>>OK, I recompiled it on my system with all security turned off.
>>>Now when I modprobe cipcb0, I get:
>>>
>>>Unable to handle kernel NULL pointer dereference at virtual address 00000074
>>><followed by a big register dump>
>>>Segmentation fault
>>>
>>>At least there are no longer any undefined symbols!  ;-)
>>> 
>>>      
>>>
>>Did you update from CVS ?
>>    
>>
>
>Of course!  :-)
>
>  
>
>>Are there any lines left, with CFLAGS ? If yes, remove 'em.
>>    
>>
>
>Nope.  I simply removed the tmp/cipe* directory, untared it again,
>removed the tmp/.done*cipe* files, ran scripts/build.sh build opt=cipe
>and then ran:  make clean install iso
>  
>
I can't try it right now, but you can:
add the line "disable_et_dyn" before any configure or make parts in the 
cipe script and let it recompile.

Let's see what happens then.

Heiko

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Bruce S. <bw...@ar...>]

> >OK, I recompiled it on my system with all security turned off.
> >Now when I modprobe cipcb0, I get:
> >
> >Unable to handle kernel NULL pointer dereference at virtual address 00000074
> ><followed by a big register dump>
> >Segmentation fault
> >
> >At least there are no longer any undefined symbols!  ;-)
> >  
> Did you update from CVS ?

Of course!  :-)

> Are there any lines left, with CFLAGS ? If yes, remove 'em.

Nope.  I simply removed the tmp/cipe* directory, untared it again,
removed the tmp/.done*cipe* files, ran scripts/build.sh build opt=cipe
and then ran:  make clean install iso

 - BS

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Heiko Z. <he...@zu...>]

Bruce Smith wrote:

>>I had to disable the compilation of ET_DYN for cipe.
>>    
>>
>
>OK, I recompiled it on my system with all security turned off.
>Now when I modprobe cipcb0, I get:
>
>Unable to handle kernel NULL pointer dereference at virtual address 00000074
><followed by a big register dump>
>Segmentation fault
>
>At least there are no longer any undefined symbols!  ;-)
>  
>
Did you update from CVS ?

Are there any lines left, with CFLAGS ? If yes, remove 'em.

Heiko

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Bruce S. <bw...@ar...>]

> I had to disable the compilation of ET_DYN for cipe.

OK, I recompiled it on my system with all security turned off.
Now when I modprobe cipcb0, I get:

Unable to handle kernel NULL pointer dereference at virtual address 00000074
<followed by a big register dump>
Segmentation fault

At least there are no longer any undefined symbols!  ;-)

  - BS

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Heiko Z. <he...@zu...>]

Bruce Smith wrote:

>Starting hotplug modules      [ OK ]
>*** can't synthesize input events - /proc/bus/input/devices missing
>  
>
Solved, too.
The updated patch is on the FTP server.

Heiko

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Heiko Z. <he...@zu...>]

Heiko Zuerker wrote:

> Heiko Zuerker wrote:
>
>> Bruce Smith wrote:
>>
>>>> I ran my own compile last night.  mrproper, new lfssystem.  I 
>>>> turned OFF
>>>> _all_ of the security options under:  Build-Configuration -> Security
>>>>
>>>> Starting up final system...
>>>> /bin/chroot: error while loading shared libraries: /lib/libsafe.so.2:
>>>> cannot open shared object file: No such file or directory
>>>> Kernel panic: Attempted to kill init!
>>>>   
>>>
>>>
>>>
>>> OK, I ran a build/install on libsafe manually and rebuilt the ISO.
>>> The system now boots.  If libsafe isn't optional, we should remove it
>>> from menuconfig (or really make it optional).
>>>
>> It is optional.
>> You probably just had the ld.so.preload on the migrated configuraiton.
>>
>>> BTW, CIPE still has unresolved symbols in it's module.  (all security
>>> stuff turned off, except libsafe)
>>>
>> Yes I just mentioned that.
>> I think I found the  problem, test compile is running.
>
>
>
> There seem to be 2 problems, the first one is solved:
> you (Bruce) defined CFLAGS in the build script, in order to turn SSP off.
> Messing around with CFLAGS, you have to be carefull. In the older 
> version we modified KCC to "gcc -fno-stack-protector"

It's OK in CVS now.

I had to disable the compilation of ET_DYN for cipe.

Heiko

Re: [Devil-linux-develop] upgrade-config

[Heiko Z. <he...@zu...>]

Friedrich Lobenstock wrote:

> Heiko Zuerker wrote on 31.12.2003 17:44 MET:
>
>>
>> OK, looks also like "ftp" needs a valid homedirectory, I'll set it to 
>> /var/empty.
>
>
> Hmmm...would /var/tmp/ be a better place for such?
>
The directory needs to be read-only for the user, otherwise vsftpd won't 
start.

Heiko

Re: [Devil-linux-develop] upgrade-config

[Friedrich L. <fl...@fl...>]

Heiko Zuerker wrote on 31.12.2003 17:44 MET:
>
> OK, looks also like "ftp" needs a valid homedirectory, I'll set it to 
> /var/empty.

Hmmm...would /var/tmp/ be a better place for such?

-- 
MfG / Regards
Friedrich Lobenstock
____________________________________________________________________
Friedrich Lobenstock                       Linux Services Lobenstock
URL: http://www.lsl.at/                         Email: fl...@fl...
____________________________________________________________________

Re: [Devil-linux-develop] upgrade-config

[Heiko Z. <he...@zu...>]

Bruce Smith wrote:

>>>Is that where the anonymous ftp files go?
>>>
>>>If so, why don't you specify "anon_root" in /etc/vsftpd.conf instead?
>>>
>>>      
>>>
>>because I was too lazy.....
>>OK I'll set it.
>>    
>>
>
>Works better for upgrades.  :-)
>  
>
OK, looks also like "ftp" needs a valid homedirectory, I'll set it to 
/var/empty.

>>>>As I said, for me it's only the ftp user, but I was thinking that other 
>>>>users could be more "creative".
>>>>        
>>>>
>>>Like allowing "bin" to login?  :-)
>>>
>>>      
>>>
>>Shut up. ;-)
>>    
>>
>
>I can change the program to migrate the info for the stock users, but
>that may cause us other problems.
>
>What happens when we upgrade to a new version of a package, and the 
>home directory for a package's user should be changed to something
>different?  Somone migrates, changes the home directory back to the 
>old value, and now the package doesn't work.
>
>Personally I think we should leave it as-is, but I'll change it if you
>want me to ...  ?
>  
>
OK you're right, let's leave it as it is.

Heiko

Re: [Devil-linux-develop] upgrade-config

[Bruce S. <bw...@ar...>]

> >Is that where the anonymous ftp files go?
> >
> >If so, why don't you specify "anon_root" in /etc/vsftpd.conf instead?
> >
> because I was too lazy.....
> OK I'll set it.

Works better for upgrades.  :-)

> >>As I said, for me it's only the ftp user, but I was thinking that other 
> >>users could be more "creative".
> >
> >Like allowing "bin" to login?  :-)
> >
> Shut up. ;-)

I can change the program to migrate the info for the stock users, but
that may cause us other problems.

What happens when we upgrade to a new version of a package, and the 
home directory for a package's user should be changed to something
different?  Somone migrates, changes the home directory back to the 
old value, and now the package doesn't work.

Personally I think we should leave it as-is, but I'll change it if you
want me to ...  ?

 - BS

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Heiko Z. <he...@zu...>]

Heiko Zuerker wrote:

> Bruce Smith wrote:
>
>>> I ran my own compile last night.  mrproper, new lfssystem.  I turned 
>>> OFF
>>> _all_ of the security options under:  Build-Configuration -> Security
>>>
>>> Starting up final system...
>>> /bin/chroot: error while loading shared libraries: /lib/libsafe.so.2:
>>> cannot open shared object file: No such file or directory
>>> Kernel panic: Attempted to kill init!
>>>   
>>
>>
>> OK, I ran a build/install on libsafe manually and rebuilt the ISO.
>> The system now boots.  If libsafe isn't optional, we should remove it
>> from menuconfig (or really make it optional).
>>
> It is optional.
> You probably just had the ld.so.preload on the migrated configuraiton.
>
>> BTW, CIPE still has unresolved symbols in it's module.  (all security
>> stuff turned off, except libsafe)
>>
> Yes I just mentioned that.
> I think I found the  problem, test compile is running.


There seem to be 2 problems, the first one is solved:
you (Bruce) defined CFLAGS in the build script, in order to turn SSP off.
Messing around with CFLAGS, you have to be carefull. In the older 
version we modified KCC to "gcc -fno-stack-protector"

Heiko

Re: [Devil-linux-develop] upgrade-config

[Heiko Z. <he...@zu...>]

Bruce Smith wrote:

>>No I did actually change the home directory of "ftp", since this is the 
>>user the vsftpd runs under.
>>    
>>
>
>Is that where the anonymous ftp files go?
>
>If so, why don't you specify "anon_root" in /etc/vsftpd.conf instead?
>
because I was too lazy.....
OK I'll set it.

>>As I said, for me it's only the ftp user, but I was thinking that other 
>>users could be more "creative".
>>    
>>
>
>Like allowing "bin" to login?  :-)
>  
>
Shut up. ;-)


Heiko

Re: [Devil-linux-develop] upgrade-config

[Bruce S. <bw...@ar...>]

> No I did actually change the home directory of "ftp", since this is the 
> user the vsftpd runs under.

Is that where the anonymous ftp files go?

If so, why don't you specify "anon_root" in /etc/vsftpd.conf instead?

> As I said, for me it's only the ftp user, but I was thinking that other 
> users could be more "creative".

Like allowing "bin" to login?  :-)

 - BS

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Heiko Z. <he...@zu...>]

Bruce Smith wrote:

>>I ran my own compile last night.  mrproper, new lfssystem.  I turned OFF
>>_all_ of the security options under:  Build-Configuration -> Security
>>
>>Starting up final system...
>>/bin/chroot: error while loading shared libraries: /lib/libsafe.so.2:
>>cannot open shared object file: No such file or directory
>>Kernel panic: Attempted to kill init!
>>    
>>
>
>OK, I ran a build/install on libsafe manually and rebuilt the ISO.
>The system now boots.  If libsafe isn't optional, we should remove it
>from menuconfig (or really make it optional).
>
It is optional.
You probably just had the ld.so.preload on the migrated configuraiton.

>BTW, CIPE still has unresolved symbols in it's module.  (all security
>stuff turned off, except libsafe)
>
Yes I just mentioned that.
I think I found the  problem, test compile is running.

>I'm also getting an error on hotplug (with or w/o the security stuff):
>
>Starting hotplug modules      [ OK ]
>*** can't synthesize input events - /proc/bus/input/devices missing
>  
>
That is OK.
I had to switch to the hotplug CVS version and need to add some more 
stuff to our patch for hotplug.

Heiko

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Bruce S. <bw...@ar...>]

> >Starting up final system...
> >/bin/chroot: error while loading shared libraries: /lib/libsafe.so.2:
> >cannot open shared object file: No such file or directory
> >Kernel panic: Attempted to kill init!
> >  
> >
> This is defined in the file /etc/ld.so.preload .
> Did you migrate this one from an old config?

I didn't get to the migrate part.  It does the chroot before
upgrade-config.  I guess I'll have to nuke the entire config.

> Delete the file and then it works. ;-)

I'll have to take your word for that, since I added libsafe.  :-)

 - BS

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Heiko Z. <he...@zu...>]

Heiko Zuerker wrote:

> Bruce Smith wrote:
>
>>>> There are more problems with all the new security stuff, "sagator" 
>>>> doesn't work properly anymore.
>>>>
>>>> I'm investigating what's going on, it just takes forever to test.....
>>>>     
>>>
>>> The sagator problem is solved.
>>> I start a new compile tonight, so we can see what's going on with cipe.
>>>   
>>
Cipe has unresolved symbols again...
I'll go ahead an re-disable SSP for it, but strangely the missing 
symbols have nothing todo with __guard......


Heiko

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Bruce S. <bw...@ar...>]

> I ran my own compile last night.  mrproper, new lfssystem.  I turned OFF
> _all_ of the security options under:  Build-Configuration -> Security
>
> Starting up final system...
> /bin/chroot: error while loading shared libraries: /lib/libsafe.so.2:
> cannot open shared object file: No such file or directory
> Kernel panic: Attempted to kill init!

OK, I ran a build/install on libsafe manually and rebuilt the ISO.
The system now boots.  If libsafe isn't optional, we should remove it
from menuconfig (or really make it optional).

BTW, CIPE still has unresolved symbols in it's module.  (all security
stuff turned off, except libsafe)

I'm also getting an error on hotplug (with or w/o the security stuff):

Starting hotplug modules      [ OK ]
*** can't synthesize input events - /proc/bus/input/devices missing

Re: [Devil-linux-develop] Kernel crash, was: Unresolved symbols in cipe.

[Heiko Z. <he...@zu...>]

Bruce Smith wrote:

>>>There are more problems with all the new security stuff, "sagator" 
>>>doesn't work properly anymore.
>>>
>>>I'm investigating what's going on, it just takes forever to test.....
>>>      
>>>
>>The sagator problem is solved.
>>I start a new compile tonight, so we can see what's going on with cipe.
>>    
>>
>
>I ran my own compile last night.  mrproper, new lfssystem.  I turned OFF
>_all_ of the security options under:  Build-Configuration -> Security
>
>Now to see if the cipe module (cipcb0.o) loads....  
>Well, I never got that far.  Kernel crash upon boot!
>
>Starting up final system...
>/bin/chroot: error while loading shared libraries: /lib/libsafe.so.2:
>cannot open shared object file: No such file or directory
>Kernel panic: Attempted to kill init!
>  
>
This is defined in the file /etc/ld.so.preload .
Did you migrate this one from an old config?

Delete the file and then it works. ;-)

cu
  Heiko

Re: [Devil-linux-develop] upgrade-config

[Heiko Z. <he...@zu...>]

Bruce Smith wrote:

>>I just ran into a small problem with the upgrade of users/groups.
>>Currently only the passwords of existing users get migrated.
>>
>>This can (in my case did) cause problems, when other things got changed 
>>like group memberships or home directories.
>>    
>>
>
>Can you be more specific on what users you are talking about?
>
>Normal users, added after installation (tom/dick/harry) should have ALL
>of their attributes moved including home directory, shell, etc.
>
>Users in the default passwd file (root,bin,www,mail,fax,...) only have
>their passwords migrated.  This is because only the stock user which has
>rights to actually login is "root".  So, why is this a problem?  Did you
>change root's home directory or shell?  Or are you using "bin" as a
>login user? (shame on you! :)
>  
>
No I did actually change the home directory of "ftp", since this is the 
user the vsftpd runs under.

>>Should we just go ahead and plain-dumb copy the entire lines from the 
>>passwd/shadow/group/gshadow files ?
>>    
>>
>
>No, please give more details of what the hell you're doing, and I'll
>adjust accordingly.  :-)
>  
>
I didn't do anything... :-)))

As I said, for me it's only the ftp user, but I was thinking that other 
users could be more "creative".

cu
  Heiko