[v7,21/21] scripts: add loongarch64 binfmt config

Signed-off-by: Song Gao <gaosong@loongson.cn>
Signed-off-by: Xiaojuan Yang <yangxiaojuan@loongson.cn>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 scripts/qemu-binfmt-conf.sh | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

Hi Song,

On 10/18/21 20:47, Song Gao wrote:
> Signed-off-by: Song Gao <gaosong@loongson.cn>
> Signed-off-by: Xiaojuan Yang <yangxiaojuan@loongson.cn>
> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>   scripts/qemu-binfmt-conf.sh | 6 +++++-
>   1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/scripts/qemu-binfmt-conf.sh b/scripts/qemu-binfmt-conf.sh
> index 7de996d..5575bdd 100755
> --- a/scripts/qemu-binfmt-conf.sh
> +++ b/scripts/qemu-binfmt-conf.sh
> @@ -4,7 +4,7 @@
>   qemu_target_list="i386 i486 alpha arm armeb sparc sparc32plus sparc64 \
>   ppc ppc64 ppc64le m68k mips mipsel mipsn32 mipsn32el mips64 mips64el \
>   sh4 sh4eb s390x aarch64 aarch64_be hppa riscv32 riscv64 xtensa xtensaeb \
> -microblaze microblazeel or1k x86_64 hexagon"
> +microblaze microblazeel or1k x86_64 hexagon loongarch64"
>   
>   i386_magic='\x7fELF\x01\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x03\x00'
>   i386_mask='\xff\xff\xff\xff\xff\xfe\xfe\x00\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff\xff'
> @@ -140,6 +140,10 @@ hexagon_magic='\x7fELF\x01\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x
>   hexagon_mask='\xff\xff\xff\xff\xff\xff\xff\x00\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff\xff'
>   hexagon_family=hexagon
>   
> +loongarch64_magic='\x7fELF\x02\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x02\x01'
> +loongarch64_mask='\xff\xff\xff\xff\xff\xff\xff\x00\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff\xff'

Here the EI_OSABI (7th, 0-based offset) byte is ignored which is okay 
(we want at least ELFOSABI_SYSV=0 and ELFOSABI_GNU=3 but not others, 
mask of this byte could be "\xfc" to exclude values > 3).

However, the EI_ABIVERSION (8th) byte is fixed to be zero; according to 
the draft LoongArch ELF psABI spec [1] there might be future revision(s) 
to necessitate a bump of ABI version, but it's highly unlikely to in 
turn require modifications to QEMU code, with almost all logic happening 
in the dynamic loader. I suggest unmasking this whole byte (setting mask 
for this byte to "\x00").

[1]: 
https://github.com/loongson/LoongArch-Documentation/blob/50e62f196afd4fae7f25dc98854167f97528046b/docs/LoongArch-ELF-ABI-EN.adoc

> +loongarch64_family=loongarch
> +
>   qemu_get_family() {
>       cpu=${HOST_ARCH:-$(uname -m)}
>       case "$cpu" in

Hi, Xuerui

On 10/18/2021 11:49 PM, WANG Xuerui wrote:
> Hi Song,
> 
> On 10/18/21 20:47, Song Gao wrote:
>> Signed-off-by: Song Gao <gaosong@loongson.cn>
>> Signed-off-by: Xiaojuan Yang <yangxiaojuan@loongson.cn>
>> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
>> ---
>>   scripts/qemu-binfmt-conf.sh | 6 +++++-
>>   1 file changed, 5 insertions(+), 1 deletion(-)
>>
>> diff --git a/scripts/qemu-binfmt-conf.sh b/scripts/qemu-binfmt-conf.sh
>> index 7de996d..5575bdd 100755
>> --- a/scripts/qemu-binfmt-conf.sh
>> +++ b/scripts/qemu-binfmt-conf.sh
>> @@ -4,7 +4,7 @@
>>   qemu_target_list="i386 i486 alpha arm armeb sparc sparc32plus sparc64 \
>>   ppc ppc64 ppc64le m68k mips mipsel mipsn32 mipsn32el mips64 mips64el \
>>   sh4 sh4eb s390x aarch64 aarch64_be hppa riscv32 riscv64 xtensa xtensaeb \
>> -microblaze microblazeel or1k x86_64 hexagon"
>> +microblaze microblazeel or1k x86_64 hexagon loongarch64"
>>     i386_magic='\x7fELF\x01\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x03\x00'
>>   i386_mask='\xff\xff\xff\xff\xff\xfe\xfe\x00\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff\xff'
>> @@ -140,6 +140,10 @@ hexagon_magic='\x7fELF\x01\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x
>>   hexagon_mask='\xff\xff\xff\xff\xff\xff\xff\x00\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff\xff'
>>   hexagon_family=hexagon
>>   +loongarch64_magic='\x7fELF\x02\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x02\x01'
>> +loongarch64_mask='\xff\xff\xff\xff\xff\xff\xff\x00\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff\xff'
> 
> Here the EI_OSABI (7th, 0-based offset) byte is ignored which is okay (we want at least ELFOSABI_SYSV=0 and ELFOSABI_GNU=3 but not others, mask of this byte could be "\xfc" to exclude values > 3).
> 
The EI_OSABI can not be modified. If we support FreeBSD and OpenBSD in the future.
> However, the EI_ABIVERSION (8th) byte is fixed to be zero; according to the draft LoongArch ELF psABI spec [1] there might be future revision(s) to necessitate a bump of ABI version, but it's highly unlikely to in turn require modifications to QEMU code, with almost all logic happening in the dynamic loader. I suggest unmasking this whole byte (setting mask for this byte to "\x00").
>That's a good suggestion， thank you very much.


Thanks
Song Gao

On 2021/10/21 14:24, Song Gao wrote:
> Hi, Xuerui
>
> On 10/18/2021 11:49 PM, WANG Xuerui wrote:
>> Hi Song,
>>
>> On 10/18/21 20:47, Song Gao wrote:
>>> Signed-off-by: Song Gao <gaosong@loongson.cn>
>>> Signed-off-by: Xiaojuan Yang <yangxiaojuan@loongson.cn>
>>> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
>>> ---
>>>   scripts/qemu-binfmt-conf.sh | 6 +++++-
>>>   1 file changed, 5 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/scripts/qemu-binfmt-conf.sh b/scripts/qemu-binfmt-conf.sh
>>> index 7de996d..5575bdd 100755
>>> --- a/scripts/qemu-binfmt-conf.sh
>>> +++ b/scripts/qemu-binfmt-conf.sh
>>> @@ -4,7 +4,7 @@
>>>   qemu_target_list="i386 i486 alpha arm armeb sparc sparc32plus sparc64 \
>>>   ppc ppc64 ppc64le m68k mips mipsel mipsn32 mipsn32el mips64 mips64el \
>>>   sh4 sh4eb s390x aarch64 aarch64_be hppa riscv32 riscv64 xtensa xtensaeb \
>>> -microblaze microblazeel or1k x86_64 hexagon"
>>> +microblaze microblazeel or1k x86_64 hexagon loongarch64"
>>>     i386_magic='\x7fELF\x01\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x03\x00'
>>>   i386_mask='\xff\xff\xff\xff\xff\xfe\xfe\x00\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff\xff'
>>> @@ -140,6 +140,10 @@ hexagon_magic='\x7fELF\x01\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x
>>>   hexagon_mask='\xff\xff\xff\xff\xff\xff\xff\x00\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff\xff'
>>>   hexagon_family=hexagon
>>>   +loongarch64_magic='\x7fELF\x02\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x02\x01'
>>> +loongarch64_mask='\xff\xff\xff\xff\xff\xff\xff\x00\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff\xff'
>> Here the EI_OSABI (7th, 0-based offset) byte is ignored which is okay (we want at least ELFOSABI_SYSV=0 and ELFOSABI_GNU=3 but not others, mask of this byte could be "\xfc" to exclude values > 3).
>>
> The EI_OSABI can not be modified. If we support FreeBSD and OpenBSD in the future.

Then the respective changes would belong to a different place than here,
i.e. "binfmt" which is IIRC Linux-only? We can't support running BSD
binaries on top of Linux with user-only emulation AFAIK.

But again, leaving this as-is should pose minimal risk. Even if we only
accept 0~3 at binfmt_misc matching time, we would still let through 1
and 2, so that's not enough and wouldn't make any difference. So while I
don't think you're right wrt the reason you stated, I'd still agree with
not touching this EI_OSABI mask byte.