problem with ldconfig; nvidia loading fixed
chuckr at chuckr.org
Fri Feb 29 17:11:02 UTC 2008
-----BEGIN PGP SIGNED MESSAGE-----
Ben Kaduk wrote:
> Hi Chuck,
> On 2/28/08, Chuck Robey <chuckr at chuckr.org> wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
> [snip RAID/loader interaction]
>> Well, the other thing that's come up, I can't get my nvidia driver to work
>> along with xorg. It used to show it's arrival both in kldstat, AND also
>> because, in /dev, a file named nvidiactl would show up. This has stopped
>> happening. I had been using the old version 100.14.19, and when I checked
>> the nvidia website, it seems a couple of newer versions had appeared, so I
>> downloaded and built version 169.12. Don't ask me why the crazy version
>> numbering, I looked around for something like a changelog, but had no luck.
>> Anyhow, updating to the newer version wasn't too hard, but didn't show any
>> change. Consulting the /var/log/Xorg.0.log file, even though kldstat says
>> that nvidia.ko is loaded, it never shows a /dev/nvidiactl, and the log file
>> says clearly that it never finds the Nvidia kernel module. Using it with
>> Vesa lets me limp, so I do that, but howcome?
> The nvidia kernel module now needs to be loaded at boot-time;
> it won't supercede the default vga support if loaded later.
> cat nvidia_load="YES"" >> /boot/loader.conf
I'd never heard that about it needing absolutely to load at boot time, but
you were absolutely right, I have it working now.
It also made my loading problem (I don't know if I complained on hackers
about that or not). No, in looking back to by send-cache, I don't think I
did, Immediately after I finally got all my raid array proboems fixed, I
found really that there was only 1 remaining: booting my raid.
I have definitely the boot-manager loaded onto ad1 (my 100G emergency eide
disk) and da0 (my raid). I normally would strike f1 to load the eide for
emergencies, and f5, then f1, to load from the raid. Well, now, no matter
what I press, I load from the eide disk. If I intercept the boot at the
beastie menu, and press 6 to get into the loader, I find currdev is
_always_ set to disk1s1a (the eide, ad1). If I want to boot the raid, I
need to manually do a "set currdev=disk2s1a" then unload, then load the
kernel AND manually load all the modules, because loder.conf isn't getting
Once I went thru all that, the nvidia driver worked fine, as you predicted,
any notion why the loading has gone screwey? I can add one thing: When I
initially installed the system, I was just coming back from using Linux for
some years, and I forgot that while Linux is in love with using a small
/boot partition to store the kernels and the bootloader, FreeBSD isn't. I
had stuck the /boot only onto da1s1d, with my root onto da0s1a. I was
loading my root via /etc/fstab, I thought, but in fact, I had an old
version of a just-useable boot onto myu da1s1a, and that was what was
actually booting. When I got rid of the /boot mount line in fstab, and
expected things to work ok, this odd refusal to correctly boot from
Any idea here? Booting is a real drag now.
> -Ben Kaduk
>> OK, I'm not giving my ldconfig problem to you folks, its quite likely its a
>> FreeBSD-current problem, so let those folks handle that one.
>> Boy, I feel like the little boy who cried wolf!
>> -----BEGIN PGP SIGNATURE-----
>> Version: GnuPG v2.0.4 (FreeBSD)
>> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
>> -----END PGP SIGNATURE-----
>> freebsd-hackers at freebsd.org mailing list
>> To unsubscribe, send any mail to "freebsd-hackers-unsubscribe at freebsd.org"
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.4 (FreeBSD)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
-----END PGP SIGNATURE-----
More information about the freebsd-hackers