[Bug ld/28789] New: RISC-V: ld resolves absolute symbols via PC-relative relocations for position-independent targets

[palmer at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28789

            Bug ID: 28789
           Summary: RISC-V: ld resolves absolute symbols via PC-relative
                    relocations for position-independent targets
           Product: binutils
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: ld
          Assignee: unassigned at sourceware dot org
          Reporter: palmer at gcc dot gnu.org
  Target Milestone: ---

Created attachment 13911
  --> https://sourceware.org/bugzilla/attachment.cgi?id=13911&action=edit
First shot at a patch

During a recent discussion on the psABI spec, it came up that we're allowing
absolute symbols to be resolved via PC-relative addressing sequences when
linked into position-independent targets.  This may not produce the expected
answer at runtime, which is not particularly friendly for users.

I've attached a patch that resolves some of these issues, but see the big FIXME
related to absolute symbols defined by linker scripts.  I'm not sure what the
expected behavior is there, but x86 is explicitly treating these as resolvable
via PC-relative relocations and that's not what I'd expect so I'm assuming I'm
missing something.  Our GP initialization sequences are also using absolute
addresses in this way, which is at best worrisome. 

Given how late we are in the 2.38 cycle and how tricky absolute symbols are, I
think it's best if we hold off on doing anything for now.

-- 
You are receiving this mail because:
You are on the CC list for the bug.