Re: [Qemu-devel] [PATCH] linux-user, s390x: ignore OS ABI value in ELF header

[Laurent Vivier]

Le 12/11/2017 à 07:56, Thomas Huth a écrit :
> Am Fri, 10 Nov 2017 20:49:35 +0100
> schrieb Laurent Vivier <address@hidden>:
> 
>> I have this error:
>> bash: /sbin/ldconfig: cannot execute binary file: Exec format error
>>
>> because /sbin/ldconfig is:
>> ELF 64-bit MSB executable, IBM S/390, version 1 (GNU/Linux),
>> statically linked, for GNU/Linux 3.2.0,
>> BuildID[sha1]=90b64604014aafac9c1a0623b1cf447281d1a382, stripped
>>
>> OS ABI is GNU/linux
>>
>> "/bin/ls" works well:
>>
>> ELF 64-bit MSB shared object, IBM S/390, version 1 (SYSV),
>> dynamically linked, interpreter /lib/ld64.so.1, for GNU/Linux 3.2.0,
>> BuildID[sha1]=be9b19143d4657678846f6e5277383071fc1059a, stripped
>>
>> OS ABI is SYSV
>>
>> To be able to execute ldconfig, this patch modifies s390x binfmt mask
>> to ignore the OS ABI value (EI_OSABI, byte 7).
>>
>> Signed-off-by: Laurent Vivier <address@hidden>
>> ---
>>  scripts/qemu-binfmt-conf.sh | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/scripts/qemu-binfmt-conf.sh b/scripts/qemu-binfmt-conf.sh
>> index 8afc3eb5bb..e2e1b7544d 100755
>> --- a/scripts/qemu-binfmt-conf.sh
>> +++ b/scripts/qemu-binfmt-conf.sh
>> @@ -85,7 +85,7 @@
>> sh4eb_mask='\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff
>> sh4eb_family=sh4 
>>  
>> s390x_magic='\x7fELF\x02\x02\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x16'
>> -s390x_mask='\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff'
>> +s390x_mask='\xff\xff\xff\xff\xff\xff\xff\x00\xff\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff'
> 
> If I've got that right, the OSABI field should either be 0 for
> "No extensions or unspecified" (which is then printed  by "file"
> as SYSV) or 3 for "GNU/Linux".
> Thus wouldn't it be better to use a mask of 0xfc here instead, so
> that we refuse at least everything with a value > 3 here?
No, because with this mask you allow HPUX and NetBSD (See
/usr/include/elf.h), I don't think it's really better. The simplest way
to manage this is to ignore the field. A better way should be to have
two binfmt entries, one for SYSV and one for GNU/Linux, but as our goal
is to run linux binaries on linux system, is it worth it?

Most of the other archs use already 0.

> 
> Also I wonder whether i386 has the same problem, too?
> 

Yes, I think all targets can have the same problem (I have with m68k,
but with the gcc-7/cc1 binary).

Thanks,
Laurent