[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: occult nvidia kernel version skew -- getting desperate



On Mon, Jun 19, 2006 at 08:48:42AM -0600, Paul E Condon wrote:
> I haven't been following this thread, but this approach to a problem
> seems to me to be overly complicated.
> 
> On Mon, Jun 19, 2006 at 09:26:42AM -0400, hendrik@topoi.pooq.com wrote:
> > On Sun, Jun 18, 2006 at 06:18:03PM -0400, hendrik@topoi.pooq.com wrote:
> > > Guessing that my problems probably resulted from a faulty upgrade (and 
> > > there was plenty of that in the xorg6.9->7.0 transition), my next step 
> > > would seem to be to purge *all* nvidia-specific packages, remove their 
> > > .debs from apt-cache, reconstruct the locate database, look for any 
> > > remaining files with a reasonable facsimile of "nvidia" in their names 
> > > and delete them, look for any stray files not part of any package and 
> > > delete any that look suspicious, and then begin to install the stuff 
> > > again from scratch, downloading .debs, compiling kernel modules, etc.
> > 
> > I have now purged every package that contains 'nvidia' as part of its 
> > name, and cleared the apt-cache so that if I ever start installing such 
> > packages I can be sure of getting a fresh download.
> > 
> > But I still find nvidia-related files:
> > /lib/modules/2.6.12-1-amd64-generic/kernel/drivers/video/nvidia
> > /lib/modules/2.6.12-1-amd64-generic/kernel/drivers/video/nvidia/nvidiafb.ko
> > /lib/modules/2.6.15-1-amd64-generic/kernel/drivers/video/nvidia
> > /lib/modules/2.6.15-1-amd64-generic/kernel/drivers/video/nvidia/nvidiafb.ko
> > 
> > These should have vanished when I purged 
> > nvidia-kernel-2.6.12-1-amd64-generic and the corresponding package for 
> > 2.6.15, shouldn't they have?
> > 
> > Can I delete them?  I am desperately trying to clean everything out 
> > before I start a reinstall.
> > 
> > The nvidia-kernel-2.6.15... package has disappeared from aptitude's list 
> > of packages, by the way, but nvidia-kernel-2.6.12-1-amd64-generic is 
> > still in the list, marked 'p  S' as purged but (I presume) from a 
> > suspicious source.
> > 
> > 
> > There;s also a lot of nvidia stuff in /usr/src/linus-headers...
> > 
> > /usr/src/linux-headers-2.6.12-1-amd64-generic/include/config/fb/nvidia
> > /usr/src/linux-headers-2.6.12-1-amd64-generic/include/config/fb/nvidia/i2c.h
> > /usr/src/linux-headers-2.6.12-1-amd64-generic/include/config/fb/nvidia/module.h
> > /usr/src/linux-headers-2.6.12-1/drivers/video/nvidia
> > /usr/src/linux-headers-2.6.12-1/drivers/video/nvidia/Makefile
> > /usr/src/linux-headers-2.6.15-1-amd64-generic/include/config/fb/nvidia
> > /usr/src/linux-headers-2.6.15-1-amd64-generic/include/config/fb/nvidia/i2c.h
> > /usr/src/linux-headers-2.6.15-1-amd64-generic/include/config/fb/nvidia/module.h
> > /usr/src/linux-headers-2.6.15-1/drivers/video/nvidia
> > /usr/src/linux-headers-2.6.15-1/drivers/video/nvidia/Makefile
> > 
> > and a huge file tree under /usr/src/modules/nvidia-kernel
> > 
> > Can I just delete all this /usr/src stuff?
> > 
> > -- hendrik
> > 
> IMHO, you will never be sure that *all* files pertaining to a particular
> package have been removed from a malfunctioning computer, whatever the
> issue or malfunction. 

True.  But I thought it might help to delete all the files that appeared 
to be related to the nvidia packages.  What I was hoping to find out is 
whether any of these files definitely belonged to other stuff *before* 
I risked rendering my machine nonbootable.

> 
> You should consider, for *this* situation, starting an install on a
> new partition or new hard disk.

I was trying to avoid this.

> Preferably, a new clean hard disk. This
> way there cannot be any 'nvidia' or whatever. After you have solved
> the problem that you attribute to nvidia,

The problem was pretty likely to have to do with nvidia: specifically 
that when starting X, the nvidia software complained that its 
components had inconsistent version numbers.  The versions in question 
were the version I had installed in the past and the version I had 
installed at the present.  It seemed reasonable that some files left 
from the old installation had survived to cause trouble in the new.

I have made another post in this thread speculating how this situation 
may have arisen.  I would appreciate anyone who knows more about how the 
nvidia-kernel-source package creates the nvidia-kernel package to 
comment on that;  there may be a problem with the instructions in its 
README file if my theory is correct.

> you can try installing old
> files from backups, if you really need them. Install them a few at
> a time, always checking that you have not broken you new, nvidia free,
> environment.

Since I had localised the problem to the presence of inconsistent 
versions, I thought that wiping all the nvidia stuff would suffice. 
But, as you point out, that seems to be harder to do than expected.

> 
> All this might not be necessary if you felt you were really in charge
> of what is on your computer, but IMHO you display a serious lack of
> confidence on this point. So do something that puts you in control
> again.

Is the question whether I feel in charge of my computer, or whether I 
actually *am* in charge?

> 
> And if your new nvidia free installation displays the same wrong
> behavior as you now experience, perhaps, just perhaps, nvidia is
> not the problem.

It is doubtful that, without nvidia, it will complain about the 
versions of nvidia software.

> 

Still, if nothing else works, I suppose I will have to resort to 
complete reinstallation.  There is a spare partition which I left free
long ago, just in case.

Thanks.

-- hendrik

> 
> -- 
> Paul E Condon           
> pecondon@mesanetworks.net
> 
> 
> -- 
> To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org 
> with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
> 



Reply to: