Keith Owens on Mon 12/06 09:37 +1000:
> > The only way I could fix it was to turn off symbol versioning for
> > modules. This is the first kernel I've had to do that for. If I
> > recall the symbol versions appended to the names were some kinds of
> > numbers for the working modules, but the ones that failed to load
> > had some acronyms containing strings like "RLversion_here" or
> > something along those lines.
>
> Whenever you see _R__ver_ in a kernel symbol it means that the
> Makefile stuffed up the symbol versioning. The only guaranteed fix is
> to remove everything and recompile from scratch. You have to use make
> mrproper to remove the broken version data but make mrproper deletes
> .config so save that first.
No, that doesn't help. After a distclean/oldconfig one still gets the
error. Many of the symbols actually have versions and allow the modules
to be installed (nothing is unresolved) but I'd say more than half of
them just have the _R__ver_ business with no actual number. Those are
the ones that, of course, can't find those symbols...
I've never had this problem before, it seems to be specific to 2.2.16.
I have to build without symbol versioning; then everything goes ok.
Note that I've grepped through the symbols in the objects and there
really is nothing that defines the symbols they want...it's like the
symbol table wasn't finished being created or something. For example
they may want enable_irq_R0998fd84 or something like that, but the only
symbol in /proc/ksyms is enable_irq_R__ver_ (but can't recall exactly
from memory). Or, that might be reversed...I can't recall and machine
is not with me right now.
-- Scott- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Jun 15 2000 - 21:00:24 EST