Thanks Ken-
This definitely got me further, and was good advice. Looks like it's been an issue for some time w/ Fedora kernel source.
http://www.g-loaded.eu/2005/12/14/the-complete-fedora-kernel-headers/

Unfortunately, there were many more potholes ahead- mostly implicit declaration errors that required additional include lines for kernel headers. But then the showstoppers came up... struct device changes, function argument mismatches, etc.

I wish the ofa_kernel package development wasn't so anchored by enterprise distros. Making it compatible w/ more recent kernels is work that eventually needs to be done anyway- doing it sooner would allow for that much more pre-enterprise testing. (and I wouldn't have to rebuild all the applications that I prefer not to be dark age versions in order to have a distro that is ofed compatible) :(

    Jeremy

On 11/23/2010 8:22 AM, Ken Cain wrote:
Jeremy Enos wrote:
As does gcc-4.3.5- same error.

Is anyone looking into this on the OFED team?

Jeremy,

This is very informal / unofficial, but it may help you get beyond your build problem. It's quite possible that my approach may not be the best, or may have flaws I do not yet appreciate. Please accept it in the spirit of trying to be helpful.

In my experience I have had to install the kernel .src.rpm package, issue an rpmbuild -bp on the kernel sources, and then do a certain amount of kernel sources prep in the installation area in order to successfully compile the ofa_kernel package (the kernel-devel and kernel-headers packages to not seem to be enough).

rpm -i kernel-2.6.xx...src.rpm
rpmbuild -bp --target=$(uname -m) /usr/src/redhat/SPECS/kernel.spec
cd  /usr/src/redhat/BUILD/kernel-2.6.xx/linux-2.6.xx.x86_64
make oldconfig && make prepare && make scripts

Then when invoking the OFED install.pl refer that script to your installed kernel sources (-s argument):
/usr/src/redhat/BUILD/kernel-2.6.xx/linux-2.6.xx.x86_64

I hope this helps. Thanks,

-Ken
_______________________________________________
ewg mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg

Reply via email to