Re: [bug-mes] M2-Planet test23

[Danny Milosavljevic]

Hi Jeremiah,

on a real ARM machine, I get the Linux kernel error message

[ 1375.455579] 1891 (test23-armv7l-b): Uhuuh, elf segment at 00010000 requested 
but the memory is mapped already

when I invoke

 ./test/results/test23-armv7l-binary

without arguments.  That is 100% reproducible.

The following section headers are in the latter file:

There are 7 section headers, starting at offset 0xd0:

Section Headers:
  [Nr] Name              Type            Addr     Off    Size   ES Flg Lk Inf Al
  [ 0]                   NULL            00000000 000000 000000 00      0   0  1
  [ 1] .text             PROGBITS        00010200 000200 007b38 00  AX  0   0  1
  [ 2] .data             PROGBITS        00017d38 007d38 001e04 00  WA  0   0  1
  [ 3] .comment          PROGBITS        00010080 000080 000020 00      0   0  1
  [ 4] .shstrtab         STRTAB          000100a0 0000a0 000030 00      0   0  1
  [ 5] .symtab           SYMTAB          00019b3c 009b3c 001330 10      6   0  1
  [ 6] .strtab           STRTAB          000181ef 0081ef 00194d 00      0   0  1

That doesn't look so bad.

Not so fast:

objdump -x says:

./test/results/test23-armv7l-binary:     file format elf32-littlearm
./test/results/test23-armv7l-binary
architecture: arm, flags 0x00000112:
EXEC_P, HAS_SYMS, D_PAGED
start address 0x00010200

Program Header:
    LOAD off    0x00000000 vaddr 0x00010000 paddr 0x00010000 align 2**16
         filesz 0x0000ae6c memsz 0x0000ae6c flags r-x
    LOAD off    0x00000000 vaddr 0x00010000 paddr 0x00010000 align 2**0
         filesz 0x0000ae6c memsz 0x0000ae6c flags rwx
private flags = 5000200: [Version5 EABI] [soft-float ABI]

I'm not sure it's allowed to have two overlapping entries in Program Header 
(see vaddr, memsz)...

Looking at how the ELF program header gets built 
(./test/common_armv7l/ELF-armv7l-debug.hex2), we see:

:ELF_program_headers
:ELF_program_header__text
01 00 00 00                    # ph_type: PT-LOAD = 1
00 00 00 00                    # ph_offset

&ELF_base                      # ph_vaddr
&ELF_base                      # ph_physaddr

%ELF_end>ELF_base              # ph_filesz
%ELF_end>ELF_base              # ph_memsz

05 00 00 00                    # p_flags
00 00 01 00                    # alignment

# @60
:ELF_program_header__data
01 00 00 00                     # ph_type: PT-LOAD = 1
00 00 00 00                     # ph_offset
&ELF_base                       # ph_vaddr
&ELF_base                       # ph_physaddr
%ELF_end>ELF_base               # ph_filesz


So that's twice the same ph_vaddr, right?  That's bad...

pgpbEenWf70l4.pgp
Description: OpenPGP digital signature