kernel-smp-2.6.9-34.0.2.EL

[Dag Wieers]

On Fri, 4 Aug 2006, pierre_stephane.baton at alcatelaleniaspace.com wrote:

> On Fri, 4 Aug 2006, Dag Wieers wrote:
> > On Fri, 4 Aug 2006, pierre_stephane.baton at alcatelaleniaspace.com wrote:
> >
> > > Nothing in /var/log/message, I checked there 1st. I can't experiment 
> > > it again, those WS are in prod, and the soft has to ru for hours to 
> > > produce results, so I can't crash it..
> > > 
> > > HeY I just asked my colleague, as he has put a cosole server, this 
> > > is the output! :
> > > 
> > > ^M<3/8/2006 11:19:06 MEST>Kernel BUG at locks:1799
> > > ^M<3/8/2006 11:19:06 MEST>invalid operand: 0000 [1] SMP
> > > ^M<3/8/2006 11:19:06 MEST>CPU 0
> > > ^M<3/8/2006 11:19:06 MEST>Modules linked in: mvfs(U) vnode(U) nfsd 
> > > exportfs md5 ipv6 parport_pc lp parport autofs4 i2c_dev i2c_core nfs 
> > > lockd nfs_acl sunrpc ds yenta_socket pcmcia_core dm_mirror dm_mod 
> > > uhci_hcd ehci_hcd e1000 floppy ext3 jbd raid1 ata_piix libata 
> > > sd_mod scsi_mod
> > > ^M<3/8/2006 11:19:07 MEST>Pid: 13436, comm: StartHostSoftFr 
> > > Tainted: PF 2.6.9-34.0.2.ELsmp
> > 
> > Ok, this kernel is tainted. It has a non GPL-licensed module loaded 
> > and one of the modules has been force-loaded (which is an indication 
> > it may not have been build for your kernel).
> > 
> > Both mvfs and vnode modules are not part of the kernel and the kernel 
> > panic is obviously related to mvfs. I hope you have a 3rd party vendor 
> > to complain to.
> > 
> > Loading kernel modules that are not build against the kernel you are 
> > running is a high risk of messing up your system. You may have been 
> > lucky that you did not loose any data.
>
> I agree you about number of issues on this kernel : too much 8-D
> 
> For mvfs and vnode, those are used by IBM - Rational Clearcase, I'll look 
> by there if they have more info as we have support.

If I can give you one more advice. It might not help, but it is the right 
thing to do anyway. Open a ticket with IBM ClearCase support and demand 
that they fix this the proper way.

They'll probably tell you the kernel modules were not supported by the 
kernel you run and that this is not supported by IBM.

There are 2 issues with this.

 1. The modules should never have been loaded the way they did. If support
    take this lightly they do not understand the problem. It can damage 
    your data ! Let me repeat that: IT CAN DAMAGE YOUR DATA.

    That in itself should be a cause to prevent the situation by 
    IBM by whatever means.

 2. A binary kernel module is never the solutÃion. Vendors often donot 
    understand that, or argue that there is no other way to support Linux. 
    But that's because of their own restrictions, not Linux.

Maybe if enough people make the same complaint they'll get it.

In fact this whole situation will lead to proprietary software being 
inferior to equivalent Open Source offerings and it is important that 
vendors understand this. My experience is that proprietary kernel modules 
(and often proprietary software) on Linux have hidden pitfalls and are 
very hard to support.

Imagine you have one or two other applications or hardware that require 
their own binary kernel modules. There is no way to be sure they work 
together or that they alle support the same kernel and you are basicly 
stuck.

So it is very important to try and steer away from binary kernel modules 
at all cost. Unless you like complexity and headaches :)

Anyway, I found this on the net. Read the enlightening comment left behind 
by another user :)

	The abrupt un-exporting of symbols
	http://lwn.net/Articles/118835/

PS I do understand that you probably cannot steer away from IBM ClearCase 
in your environment. But you get the point.

Kind regards,
--   dag wieers,  dag at wieers.com,  http://dag.wieers.com/   --
[all I want is a warm bed and a kind word and unlimited power]