[Bug ld/12763] New: [PATCH] Incorrect error when .tbss starts outside PT_LOAD segment

[marcus at mc dot pp.se]

http://sourceware.org/bugzilla/show_bug.cgi?id=12763

           Summary: [PATCH] Incorrect error when .tbss starts outside
                    PT_LOAD segment
           Product: binutils
           Version: 2.20
            Status: NEW
          Severity: normal
          Priority: P2
         Component: ld
        AssignedTo: address@hidden
        ReportedBy: address@hidden


It can happen that .tbss has a VMA which is after the last VMA of the PT_LOAD
segment it is in, for example due to alignment requirements.  This will cause
ld to report an error "`.tbss' can't be allocated in segment".  But the check
is actually bogus, because the VMA of .tbss is never used.  As long as .tbss
firs in the PT_TLS segment, all is well.

There is already an exception in the macro ELF_SECTION_SIZE, which says that
the size of .tbss should not be considered when the segment is PT_LOAD.  Well,
neither should the VMA.  My suggested fix is the following, which leverages the
already implemented special handling for .tbss:

diff --git a/include/elf/internal.h b/include/elf/internal.h
index 9ea175c..0ce9385 100644
--- a/include/elf/internal.h
+++ b/include/elf/internal.h
@@ -316,6 +316,7 @@ struct elf_segment_map
    /* SHF_ALLOC sections must have VMAs within the segment.  Be                
       careful about segments right at the end of memory.  */           \
    && ((sec_hdr->sh_flags & SHF_ALLOC) == 0                            \
+       || ELF_SECTION_SIZE(sec_hdr,segment) == 0                        \
        || (sec_hdr->sh_addr >= segment->p_vaddr                                
           && (sec_hdr->sh_addr - segment->p_vaddr                      \
               + ELF_SECTION_SIZE(sec_hdr, segment) <= segment->p_memsz))))

The rationale here is that if the section has zero size (either becase it's
actually empty, or because we are ignoring the size because it's .tbbs in
PT_LOAD), then the VMA does not matter, because the address space is empty
anyway.

The bug exists in both 2.20.1 and 2.21.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.