[Milkymist-devel] new R_LM32_16_GOT relocation

[Sébastien Bourdeauducq]

Hi Jon,

The new binutils for LM32 emits undocumented R_LM32_16_GOT relocations that 
are not implemented in Linux. From what I understood according to the 
lm32_elf_howto_table array, I needed to patch the 16 least significant bits of 
the instruction to be relocated:
http://github.com/lekernel/milkymist-
linux/commit/73de207d007baeb981c2ae70e20c5557834f2a9f

But when running the binary under the new kernel, the machine just freezed. I 
can't tell you much more now, but we should have QEMU support soon which will 
be a lot of help to debug such problems.

Can you tell us what R_LM32_16_GOT is supposed to do? How to handle it if not 
by patching the 16 LSBs?

Thank you,
Sébastien