From: Alan H. <aho...@va...> - 2000-09-14 14:48:38
|
Bill, Try again, I've just synced the tdfx-2-branch with the trunk which holds the latest on the kernel modules. Alan. On Thu, Sep 14, 2000 at 10:07:19AM -0400, Bill White wrote: > I can't seem to build the tdfx kernel module on the tdfx-2-1-branch > branch with kernel 2.4.0test8. I get these undefined symbols. > > [crasher]: sudo insmod ./tdfx.o > ./tdfx.o: unresolved symbol copy_from_user_ret > ./tdfx.o: unresolved symbol copy_to_user_ret > ./tdfx.o: unresolved symbol lock_kernel > ./tdfx.o: unresolved symbol unlock_kernel > > This is what I did: > 1.) I have rebuilt and installed the kernel, and the result of > 'uname -r' is "2.4.0-test8". > 2.) I have also updated the tdfx-2-1-branch X tree. > 3.) I enabled AGP and disabled DRM in the kernel. > 4.) A grep of /proc/ksyms for the symbols above shows that there are > no symbols of those names, but there are __generic_copy_to_user > and __generic_copy_from_user symbols defined. There are no > symbols which look anything like lock_kernel. > 5.) I compiled the kernel module with this command. > make -f Makefile.linux \ > TREE=/3dfx/obj/Linux/2.4.0test8/crasher/linux/include \ > tdfx.o > I keep the kernel sources in /3dfx/src/Linux, and make links > with lndir in /3dfx/obj/Linux, just like with the X tree. This > lets me keep multiple configurations of the same source tree. > 6.) The kernel is configured without versioned symbols. The > relevant lines of .config are: > # > # Loadable module support > # > CONFIG_MODULES=y > # CONFIG_MODVERSIONS is not set > CONFIG_KMOD=y > 7.) The symbols copy_from_user_ret and copy_to_user_ret are not > defined in the kernel as far as etags can find. > 8.) All occurrences of lock_kernel that etags can find are inlined > or macros, but they don't seem to be expanded that way in the > kernel code. They seem to be all defined by including the > file smp_lock.h. However, this file doesn't seem to be > included anywhere in the X tree. > > > I think this catches all of the obvious sources of this problem. > > > _______________________________________________ > Dri-devel mailing list > Dri...@li... > http://lists.sourceforge.net/mailman/listinfo/dri-devel |