1.0-8756 and linux-kernel 2.6.16 remap_page_range bug resurfaces!

[adambrewster]

I had the same problem on my freshly installed gentoo box. Problem was that I used the wrong install CD (x86 instead of amd64). Somewhat surprising that this was the first thing to fail.

To see if you made the same mistake, type

hostname# grep CHOST /etc/make.conf

amd64 users should see CHOST="x86_64-...". If you see CHOST="i686-..." or similar then you've probably made the same mistake.

!! Note, that you can't just change this line without rebuilding your whole system !!

My fix was to wipe the whole thing and and reinstall from scratch, but there's probably a better way if you've got any valuable data on your disk.

[tek0]

if anyone's interested, I got around that problem: http://www.nvnews.net/vbulletin/show...000#post905000

[blue666ma]

Updates to the problem:

* I'm now running 2.6.16-gentoo-r9 sources (2.6.16.9)
* I've tried installing (from both the nvidia installer and portage) the new 1.0.8762 kernel module; yep, remap_page_range and pci_find_class (surprise, surprise).
* I'm going to include everything in this post: emerge --info and kernel config, but as you'll see, my CHOST is x86-64 in make.conf.
* The only things I can think of are: kernel headers and/or USE flags.
* The installer still hangs immediately after the licence agreement. The part it appears to hang on is the lookup for pre-complied kernel module.
* I've tried Tek0's alias method, and my alias files under /etc/modules.d/* are not executable. But,

encephalon modules.d # update-modules
* Warning: could not generate /etc/modprobe.conf!

* I have 2X BenQ FP202W's @ 1680x1050 a piece, and I want TWINVIEW, OMG, this is the last hurdle! PLEASE HELP!!!

Relavant parts of kernel config:

Has MTRR, nvidiafb (but I modprobe -r it), Athlon64

[d13f00l]

Hm, I'm not sure if the nvidia installer uses this..but..

Isn't there a kernel-headers package, separate from the kernel? Maybe that is out of date?

[blue666ma]

Trying with new 2.6.16 linux-headers (~amd64 unstable branch) and re-emerging EVERYTHING (-uDN). Will post update.

[blue666ma]

GOT IT !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

New Kernel headers solved it. Emerged just fine. Running 1.0.8762 !!

[kiwijulz]

Another possibility: ccache (which I think a lot of folks using Gentoo probably take advantage of.)

For me, and it seems quite a few others, it's causing some 'random' behaviour in the 'which kernel flags to compile the wrapper module with' detection application.

See this gentoo bug page for details: http://bugs.gentoo.org/show_bug.cgi?id=137000

NB- this bit me after upgrading to 2.6.17..., after compiling it successfully, and having it run. I messed up something else and blew away /lib/modules/2.6.17... and rebuilding all modules. Then I saw random compile warnings and dmesg log statements about one, other, or both of remap_page_range pci_find_class - I ended up recompiling several times.

I then re-installed the kernel-headers, etc- nothing worked for me. Finally looked up the ebuild package... and there's the answer right there, simple.

My bad really; everyone using Gentoo should read the warnings about ccache and confcache if they're using them... sometimes they have to be turned off, or flushed out :-). Same goes with 'interesting' compiler options!

Hope that saves someone an hours or two!

KJ

[theWebster]

Thanks everyone but I'm still at it here, though a lot smarter than I was a month ago when I set out to try gentoo again on a couple of machines. The really old slow one ($2 at a yard sale) has KDE 3.5 now and plays DVD's flawlessly, while my newer AthlonXP with nVidia Corporation NVCrush11 [GeForce2 MX Integrated Graphics] (rev b1) is still causing me grief (no desktop.) It was fine and stable bringing a gentoo-sources 2.6.17-gentoo-r6 ACCEPT_KEYWORDS="~x86" kernel and KDE 3.5 all up to date, checked with revdep-rebuild but when I tried to startx it choked.

I found that xorg-x11-7.1 and nvidia old drivers aren't compatible, and my card isn't supported by the new drivers, so have reluctantly stepped back through downgrading to match this mask set:
>media-video/nvidia-kernel-1.0.6629-r5
>media-video/nvidia-glx-1.0.6629-r7
>=x11-base/xorg-x11-7.1
>=x11-base/xorg-server-1.1
>=x11-drivers/xf86-video-vesa-1.2.0
>=x11-drivers/xf86-input-keyboard-1.1.0
>=x11-drivers/xf86-input-mouse-1.1.1
but still no luck. I checked more and now feature -ccache, re-emerged linux-headers, glibc, the kernel and nvidia drivers and STILL GET
WARNING: /lib/modules/2.6.17-gentoo-r6/video/nvidia.ko needs unknown symbol remap_page_range
WARNING: /lib/modules/2.6.17-gentoo-r6/video/nvidia.ko needs unknown symbol pci_find_class

This graphics card is NOT so old, and I see many cudo's on the web for it being a good solution, I WANT SUPPORT please. PLEASE PLEASE.

[zander]

@theWebster: please use the 1.0-8774 NVIDIA Linux graphics driver release, 1.0-6629 is no longer supported.

[theWebster]

Thank you Zander, have done, also removed all masks and updated my world, very nice. I can see shelling in from my Mac network that all is well, kdm, X and kdm_greet are all running and Xorg.0.log hasn't the nasty remarks now. I'm hoping if I find the right numbers to plug into xorg.conf that I'll have a screen displaying something other than a fat grey barrel... ;p

It is looking very promising, to be sure. I might have to put a nice monitor over there or get a bigger kvm switcher. Thanks nVidia also for supporting my hardware and software. I know it's gonna be great -- soon!

[theWebster]

YES! Thank you thank you, got it cookin' now. The magic numbers were

Section "Monitor"
Identifier "Portrait Pivot 1700"
HorizSync 30.0 - 65.0
VertRefresh 50.0 - 90.0
EndSection # thanks Google!

I can happily report that this combination works very well so far:
1.0-8774 NVIDIA Linux graphics driver
xorg-x11-7.1
kde-3.5.4
linux-2.6.17-gentoo-r6
Athlon XP A7N266
nVidia Corporation NVCrush11 [GeForce2 MX Integrated Graphics] (rev b1) [according to lspci -vvvv]