Olivier Blin <[EMAIL PROTECTED]> writes:
>> If I could get alsa-0.9.8 and nVidia drivers, I'd keep running this
>> baby. Very responsive compared 2.4.22-21 on my 1.3G 9.2 cooker
>> machine!
>
> I'll try to build the nvidia drivers later today. If I succeed, i'll
> write a small howto :)
Added to the Nvidia 2.6 patch, I had to add one more patch to be able
to build on a different running kernel, and somme specs changes IIRC.
#
# Spec file for NVIDIA kernel module.
#
# $Header: //sw/OEM/nvr40_linux_4490/drivers/resman/NVIDIA_kernel.spec#1 $
#
%define RELEASE 4496
%define CUSTOM_RELEASE 4496
%define rel %{?CUSTOM_RELEASE} %{!?CUSTOM_RELEASE:%RELEASE}
%define ver 1.0
Summary: NVIDIA kernel module for NVIDIA Architecture support.
Name: NVIDIA_kernel
Version: %ver
Release: %rel
Copyright: NVIDIA Corp. 2000
Group: User Interface/X Hardware Support
Source:
ftp://ftp1.detonator.nvidia.com/pub/drivers/english/XFree86_40/NVIDIA_kernel-%{ver}-%{RELEASE}.tar.gz
Patch: NVIDIA_kernel-1.0-4496-2.6.diff
Patch1: NVIDIA_kernel-1.0-4496-kerver.patch
URL: http://www.nvidia.com
Vendor: NVIDIA Corp.
BuildRoot: %{_tmppath}/%{name}-%{PACKAGE_VERSION}
Packager: Buildmeister
Requires: kernel >= 2.2.0
Requires: rpm >= 3.0.4
%description
NVIDIA Architecture support for systems with updated or custom kernels.
%prep
%setup -q -n ${RPM_PACKAGE_NAME}-${RPM_PACKAGE_VERSION}-${RPM_PACKAGE_RELEASE}
%patch -p1 -z .26
%patch1 -p1 -z .kerver
%build
# don't exit immediately when something bad happens.
set +e
ln -s Makefile.kbuild Makefile
# building an rpm from an srpm does not check the requires field.
# make sure that rpm is good enough for us.
RPMVERSION=`rpm -q --queryformat "%{VERSION}" rpm`
RPMMAJOR=`echo $RPMVERSION | awk -F \. '{print $1;}'`
RPMMINOR=`echo $RPMVERSION | awk -F \. '{print $2;}'`
RPMPATCH=`echo $RPMVERSION | awk -F \. '{print $3;}'`
if [ $RPMMAJOR -le 3 -a $RPMMINOR -le 0 -a $RPMPATCH -lt 4 ]; then
echo -e "\nNeed rpm version 3.0.4 or greater\n";
exit 1;
fi
#make nvidia.o
# This will install the module itself onto the user's system, but not
# load it into the running system. This will allow our auto-builds to
# execute properly
%install
# create a file list for nvidia.o, to make sure the path is correct
# it really should be under $RPM_BUILD_ROOT, but the %files section
# does not honor it. Bummer.
rm -f /tmp/files.lst
# allow a build script to override the targetted kernel
# otherwise default to current running kernel
if [ -z "$TARGET_KERNEL" ]; then
export TARGET_KERNEL=`uname -r`
fi
if [ -d $RPM_ROOT/lib/modules/$TARGET_KERNEL/kernel ]; then
INSTALLPATH=/lib/modules/$TARGET_KERNEL/kernel/drivers/video
else
INSTALLPATH=/lib/modules/$TARGET_KERNEL/video
fi
mkdir -p $RPM_BUILD_ROOT/$INSTALLPATH
IGNORE_CC_MISMATCH=1 make install
install -m 0444 nvidia.ko $RPM_BUILD_ROOT/$INSTALLPATH
echo "%attr(444 root root) " $INSTALLPATH/nvidia.ko > /tmp/files.lst
unset RPM_BUILD_ROOT # trick RPM 4.0 into not stripping us
%clean
[ -n "$RPM_BUILD_ROOT" -a "$RPM_BUILD_ROOT" != / ] && rm -rf $RPM_BUILD_ROOT
# Old versions of the tarball Makefile would install NVdriver in the
# wrong directory. Remove the old NVdriver, if that leaves it parent
# directory empty remove it as well. Note that we always install the
# nvidia.o module in the correct directory, so we don't need to worry
# about it here.
%pre
WRONGDIR=/lib/modules/`uname -r`/kernel/video
if [ -d $WRONGDIR ]; then
rm -f $WRONGDIR/NVdriver
rmdir --ignore-fail-on-non-empty $WRONGDIR
fi
# In %post we basically do "run-time" installation
# Note that our files are installed onto the system, and the packages
# are cleaned up and gone. There are no Makefiles or fancy environment
# variables to play with. I basically repeated the needed parts of the
# install.sh script here
%post
# unload it quietly if already loaded..
/sbin/rmmod NVdriver >& /dev/null
/sbin/rmmod nvidia >& /dev/null
rmmod_failed=0
/sbin/lsmod | grep nvidia >& /dev/null && rmmod_failed=1
/sbin/lsmod | grep NVdriver >& /dev/null && rmmod_failed=1
if [ $rmmod_failed -eq 1 ]; then
echo
echo "Unable to remove existing NVIDIA kernel module."
echo "Please be sure you have exited X before attempting"
echo "to install the NVIDIA_kernel rpm."
echo
exit 1
fi
# delete the old NVdriver
# XXX is there a better way to do this?
if [ -z "$TARGET_KERNEL" ]; then
export TARGET_KERNEL=`uname -r`
fi
if [ -d $RPM_ROOT/lib/modules/$TARGET_KERNEL/kernel ]; then
INSTALLPATH=/lib/modules/$TARGET_KERNEL/kernel/drivers/video
else
INSTALLPATH=/lib/modules/$TARGET_KERNEL/video
fi
rm -f $INSTALLPATH/NVdriver
function error {
echo "$(basename $0): fatal error, $1 failed"
exit 1
}
modconf=""
tmp=`mktemp -q /tmp/nvidia.XXXXXX` || error "mktemp"
paths[0]=/etc/conf.modules
paths[1]=/etc/modules.conf
paths[2]=/etc/modutils/aliases
for path in [EMAIL PROTECTED]; do
test -f $path && modconf=$path
done
if [ ! -z $modconf ]; then
# Initialize the permissions on $tmp so that we don't change the
# permissions of $modconf when we mv it into place
cp -p $modconf $tmp
sed '/^alias.*\(NVdriver\|nvidia\)/d' < $modconf > $tmp
if [ -c /dev/.devfsd ]; then
echo "alias /dev/nvidia* nvidia" >> $tmp
else
echo "alias char-major-195 nvidia" >> $tmp
fi
mv -f $tmp $modconf
test -x /sbin/update-modules && /sbin/update-modules
else
if [ -f /etc/rc.d/rc.modules ]; then
# Initialize the permissions on $tmp so that we don't change the
# permissions of /etc/rc.d/rc.modules when we mv it into place
cp -p /etc/rc.d/rc.modules $tmp
sed '/.*\(NVdriver\|nvidia\).*/d' < /etc/rc.d/rc.modules > $tmp
echo "/sbin/modprobe nvidia" >> $tmp
mv -f $tmp /etc/rc.d/rc.modules
fi
fi
if [ ! -c /dev/.devfsd ]; then
for i in 0 1 2 3; do
node="/dev/nvidia$i"
rm -f $node
mknod $node c 195 $i || error "mknod \"$node\""
chmod 0666 $node || error "chmod \"$node\""
done
node="/dev/nvidiactl"
rm -f $node
mknod $node c 195 255 || error "mknod \"$node\""
chmod 0666 $node || error "chmod \"$node\""
fi
# Add it to the module dependency list.
/sbin/depmod -a && /sbin/modprobe nvidia
# Check to see if the module is actually loaded
if ! /sbin/lsmod | grep nvidia >& /dev/null ; then
echo
echo "Failed to install nvidia.ko"
echo
exit 1
fi
# It worked
echo
echo nvidia.ko installed successfully
echo
exit 0
%preun
if [ $1 = 0 ] ; then
#
# unload it quietly if already loaded..
#
# Only unload if unused, otherwise this scriptlet fails and
# RPM won't continue with uninstallation. So we also check
# if the usage count is zero.
#
if [ x`/sbin/lsmod | grep nvidia | tr -s " " | cut -f 3 -d " "` == "x0" ]; then
/sbin/rmmod nvidia >& /dev/null
fi
fi
%files -f /tmp/files.lst
%changelog
* Sun Jan 05 2003 Joseph Kain <[EMAIL PROTECTED]>
- Correctly preserve permissions on /etc/rc.d/rc.modules.
- Pass "-f" option to mv.
- Fix typos.
* Fri Jan 03 2003 Joseph Kain <[EMAIL PROTECTED]>
- Preserve permissions on modules conf file.
* Thu Nov 07 2002 Joseph Kain <[EMAIL PROTECTED]>
- Suppress lsmod output
* Fri Oct 11 2002 Andy Ritger <[EMAIL PROTECTED]>
- Check that the module is loaded before reporting success.
* Fri Sep 06 2002 Gareth Hughes <[EMAIL PROTECTED]>
- Change NVdriver to nvidia.o everywhere.
* Mon Nov 26 2001 Joseph Kain <[EMAIL PROTECTED]>
- Use %{_tmppath} instead of hardcoding /var/tmp
* Tue Aug 24 2001 Joseph Kain <[EMAIL PROTECTED]>
- in preun check usage count of NVdriver before trying to unload.
* Thu Aug 31 2000 Terence Ripperda <[EMAIL PROTECTED]>
- allow build scripts to specify a kernel version via TARGET_KERNEL variable
otherwise fallback to `uname -r`
- when searching for agpgart module, search in RPM_BUILD_ROOT, so we don't
pick up an errant symbol on a build machine.
* Tue Aug 15 2000 Mark Cooke <[EMAIL PROTECTED]>
- Change permissions on NVdriver to 444 to avoid symbol-stripping issues
with versions of RPM after 3.0.4
- Add a test to make sure the NVdriver registers correctly.
* Wed Aug 02 2000 Terence Ripperda <[EMAIL PROTECTED]>
- regress use of /proc/devices to get major number
* Tue May 02 2000 Terence Ripperda <[EMAIL PROTECTED]>
- regress use of "make install" in post, no Makefile present
* Fri Apr 27 2000 Nick Triantos <[EMAIL PROTECTED]>
- Use "NVdriver" and "install" targets on make
- Remove BuildReq item from header, was causing build error
* Thu Apr 27 2000 Gregory McLean <[EMAIL PROTECTED]>
- Cleaned out some redundant parts of the spec file.
- Set the buildroot.
- Made sure the preun gets called properly.
- Added a %clean section to keep building machine tidy.
- Set up logic for epoc builds if needed.
--- NVIDIA_kernel-1.0-4496/Makefile.kbuild.kerver 2003-09-07 08:26:42.923982911 +0200
+++ NVIDIA_kernel-1.0-4496/Makefile.kbuild 2003-09-07 08:27:18.675144332 +0200
@@ -71,11 +71,11 @@
ifdef KERNDIR
KERNEL_SOURCES := $(KERNDIR)
KERNEL_HEADERS := -I$(KERNEL_SOURCES)/include
- MODULE_ROOT := /lib/modules/$(shell uname -r)/kernel/drivers # XXX
+ MODULE_ROOT := /lib/modules/$(TARGET_KERNEL)/kernel/drivers # XXX
else
- KERNEL_SOURCES := /lib/modules/$(shell uname -r)/build
+ KERNEL_SOURCES := /lib/modules/$(TARGET_KERNEL)/build
KERNEL_HEADERS := -I$(KERNEL_SOURCES)/include
- MODULE_ROOT := /lib/modules/$(shell uname -r)/kernel/drivers
+ MODULE_ROOT := /lib/modules/$(TARGET_KERNEL)/kernel/drivers
endif
#
--
Warly
