Re: aarch64: relocation 0x105 is not implemented yet.

[Andrey Borzenkov]

В Вс, 29/12/2013 в 12:44 +0000, Leif Lindholm пишет:
> On 29 December 2013 05:24, Andrey Borzenkov <address@hidden> wrote:
> > В Вс, 29/12/2013 в 01:55 +0000, Leif Lindholm пишет:
> >> It shouldn't, but difficult to say without knowing more details.
> >> Can you run the mkimage with -v and pinpoint the triggering module?
> >
> > This comes from .eh_frame section and fails for kernel.exec already.
> >
> > RELOCATION RECORDS FOR [.eh_frame]:
> > OFFSET           TYPE              VALUE
> > 000000000000001c R_AARCH64_PREL32  .text+0x0000000000009b70
> > 0000000000000044 R_AARCH64_PREL32  .text+0x0000000000009b78
> >
> > Also there are ABS32 relocations in various .debug_* sections but I
> > presume they are handled by --strip-unneeded.
> 
> Ah, yes (Fedora also saw this):
> https://plus.google.com/u/0/106265217227408958782/posts/RYNT8RwFHDB
> 
> While it might be interesting to dig out what causes this toolchain difference
> vs. Linaro, I am reasonably confident this relocation cannot be generated
> in actual C code with -mcmodel=large. So my suggestion would be to ignore
> this relocation. Or should we explicitly strip .eh_frame?
> 

If it is not needed to run the binaries on target - I sure suggest
stripping it off. I was not sure as I expected --strip-unneeded to do it
in this case.

Ignoring does not sound right in case we actually hit in in code.