At 12:30 PM -0800 2/23/02, Matthew Dharm wrote:
Well, I finally got the latest version running (see my other message to
this list... where did that go, anyway?). The problem was that the code to
deal with CONFIG_PROC_FS is causing a crash... and it looks like it might
be a toolchain bug.
I'm building with CONFIG_PROC_FS, and it works fine for me. It seems
that your problem may be with ld or insmod; it seems that your .data
section is getting placed at zero. Does the address for objects in
.data seem reasonable when you get a map from insmod?
Here is my gcc command line to build e1000_main.o for a RM7000.
/opt/local/mips-ayr-linux-gnu/bin//gcc -Wall -DLINUX -D__KERNEL__
-DMODULE -DEXPORT_SYMTAB -O2 -pipe -I../../../linux-2.4.2/include -I.
-DMODVERSIONS -include
../../../linux-2.4.2/include/linux/modversions.h -mno-abicalls
-fno-pic -mcpu=r8000 -mips2 -Wa,--trap -mlong-calls
-fomit-frame-pointer -fno-strict-aliasing -G 0 -c e1000_main.c
What toolchain are you using?
gcc --version reports 2.97. We got our tools from MonteVista Hard Hat
Linux... I'm not sure about the actual toolchain version, but I can
ask around here if you need.
I'm running the 3.5.19 driver and the only modifications that I made
are due to our hardware (we use the i82543 but in our own board, so
the usage of the software defined I/O pins is different than intel),
and we don't use an intel format serial eeprom for mac address and
setup.
I also had 3.0.<whatever> working without problems.
Our hardware is RM7000, big endian.
I hope this helps... good luck with your bringup.
Kevin
--
|