Re: How to compile module to avoid "unresolved symbols" ??
On Sun, Oct 15, 2000 at 06:05:30PM +0100, Steve Haslam wrote:
...
> The main point to check is that you are actually compiling the module
> against the headers used to build your kernel. Particularly, that
> you're using the right <linux/autoconf.h> file, and hence getting the
> right setting of CONFIG_MODVERSIONS...
>
> So, from the device3dfx directory (where you run debian/buildpkg),
> check like this (output is from my machine):
>
> bash$ nm 3dfx.o | grep printk
> U printk_R1b7d4074
> bash$ grep printk /proc/ksyms
> c0115f4c printk_R1b7d4074
>
> Now, your address and version hash will likely be different BUT the
> vital point is that the version hash (_R1b7d4074) must match in the
> 3dfx.o file and /proc/ksyms. If they don't match, this suggests you
> compiled against the wrong kernel headers. Check your KSRC setting,
> and check you did "make clean" and that a compile is really
> happening. (You should see this sort of command:
...
Steve,
Thnaks for the very quick reply. Checking the version hash gave
me a totally BLANK response for the module itself:
nm 3dfx.o | grep printk
U printk
which made me do some thinking... I went through the timestamps for the
kernel packages I had made and discovered that I had created a kernel for my
other box (an old 486... kernel compiles on my Athlon<4 minutes... on the
486=well over an hour) between the module and the last kernel for this box.
I recompiled a fresh kernel, re-did the module, and viola! no complaints!
(I also got a matching version hash.)
I haven't gotten to the testing phase, but this is obviously on the right
track...
Different question came up when I glanced through my Athlon configuration...
do you know whether framebuffer support should be compiled in for the
Voodoo? I don't currently have that marked...
Many thanks again!
Kenward
--
Hi! I'm a .signature virus! Copy me into your ~/.signature, please!
--
Reply to: