[Bug 231790] objcopy: corrupts relocation entries in big-endian mips64 output when adjusting symbol indexes

bugzilla-noreply at freebsd.org bugzilla-noreply at freebsd.org
Mon Oct 1 18:40:18 UTC 2018


https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=231790

--- Comment #6 from Ed Maste <emaste at freebsd.org> ---
As far as I can tell readelf is broken before and after r338478 on BE mips64.

It looks like it is indeed a false-positive invocation of the le conversion - a
breakpoint on _libelf_mips64el_r_info_tom when running "objcopy --strip-debug
geom_mirror.ko.works out" stops on:

* thread #1, name = 'objcopy', stop reason = breakpoint 2.1
  * frame #0: 0x000000528b50f5c4
libelf.so.2`_libelf_mips64el_r_info_tom(r_info=8589934621) at
gelf_mips64el.c:74
    frame #1: 0x000000528b50e9f2 libelf.so.2`gelf_getrela(ed=<unavailable>,
ndx=<unavailable>, dst=0x00007fffffffe400) at gelf_rela.c:97
    frame #2: 0x00000000002124b7 objcopy`copy_content [inlined]
update_reloc(ecp=<unavailable>) at sections.c:832
    frame #3: 0x00000000002123c1 objcopy`copy_content(ecp=<unavailable>) at
sections.c:604
    frame #4: 0x000000000020db67 objcopy`create_elf(ecp=0x0000000800877000) at
main.c:389
    frame #5: 0x000000000020f0a3 objcopy`create_file(ecp=0x0000000800877000,
src="geom_mirror.ko.works", dst="geom_mirror.ko.works") at main.c:705
    frame #6: 0x000000000020ec81 objcopy`main [inlined]
elfcopy_main(ecp=<unavailable>, argc=<unavailable>, argv=<unavailable>) at
main.c:1028
    frame #7: 0x000000000020ec79 objcopy`main(argc=4, argv=0x00007fffffffe750)
at main.c:1597
    frame #8: 0x0000000000209095 objcopy`_start(ap=<unavailable>,
cleanup=<unavailable>) at crt1.c:74

-- 
You are receiving this mail because:
You are the assignee for the bug.


More information about the freebsd-mips mailing list