Re: [Qemu-devel] [PATCH v1] linux-user: elf: ELF_HWCAP for s390x

[David Hildenbrand <david@redhat.com>]

On 04.06.19 14:54, Laurent Vivier wrote:
> Le 04/06/2019 à 11:36, David Hildenbrand a écrit :
>> Let's add all HWCAPs that we can support under TCG right now, when the
>> respective CPU facilities are enabled.
>>
>> Cc: Riku Voipio <riku.voipio@iki.fi>
>> Cc: Laurent Vivier <laurent@vivier.eu>
>> Cc: Cornelia Huck <cohuck@redhat.com>
>> Cc: Laurent Vivier <laurent@vivier.eu>
>> Cc: Richard Henderson <richard.henderson@linaro.org>
>> Signed-off-by: David Hildenbrand <david@redhat.com>
>> ---
>>
>> This patch is based on "[PATCH v2 00/22] s390x/tcg: Vector Instruction
>> Support Part 4". I guess it is best if I just include it in my next
>> pull request.
> Based-on: <20190603090635.10631-1-david@redhat.com>
>           [PATCH v2 00/22] s390x/tcg: Vector Instruction Support Part 4
> 
>>
>> ---
>>  include/elf.h        |  1 +
>>  linux-user/elfload.c | 39 +++++++++++++++++++++++++++++++++++++++
>>  2 files changed, 40 insertions(+)
>>
>> diff --git a/include/elf.h b/include/elf.h
>> index ea7708a4ea..3501e0c8d0 100644
>> --- a/include/elf.h
>> +++ b/include/elf.h
>> @@ -598,6 +598,7 @@ typedef struct {
>>  #define HWCAP_S390_ETF3EH       256
>>  #define HWCAP_S390_HIGH_GPRS    512
>>  #define HWCAP_S390_TE           1024
>> +#define HWCAP_S390_VXRS         2048
>>  
>>  /* M68K specific definitions. */
>>  /* We use the top 24 bits to encode information about the
>> diff --git a/linux-user/elfload.c b/linux-user/elfload.c
>> index a57b7049dd..59b813066c 100644
>> --- a/linux-user/elfload.c
>> +++ b/linux-user/elfload.c
>> @@ -1308,6 +1308,45 @@ static inline void init_thread(struct target_pt_regs *regs,
>>  #define ELF_DATA	ELFDATA2MSB
>>  #define ELF_ARCH	EM_S390
>>  
>> +#include "elf.h"
>> +
>> +#define ELF_HWCAP get_elf_hwcap()
>> +
>> +static uint32_t get_elf_hwcap(void)
>> +{
>> +    uint32_t hwcap = 0;
>> +
>> +    if (s390_has_feat(S390_FEAT_ESAN3)) {
>> +        hwcap |= HWCAP_S390_ESAN3;
>> +    }
>> +    if (s390_has_feat(S390_FEAT_ZARCH)) {
>> +        hwcap |= HWCAP_S390_ZARCH;
>> +    }
>> +    if (s390_has_feat(S390_FEAT_STFLE)) {
>> +        hwcap |= HWCAP_S390_STFLE;
>> +    }
>> +    if (s390_has_feat(S390_FEAT_MSA)) {
>> +        hwcap |= HWCAP_S390_MSA;
>> +    }
>> +    if (s390_has_feat(S390_FEAT_LONG_DISPLACEMENT)) {
>> +        hwcap |= HWCAP_S390_LDISP;
>> +    }
>> +    if (s390_has_feat(S390_FEAT_EXTENDED_IMMEDIATE)) {
>> +        hwcap |= HWCAP_S390_EIMM;
>> +    }
>> +    if (s390_has_feat(S390_FEAT_EXTENDED_TRANSLATION_3) &&
>> +        s390_has_feat(S390_FEAT_ETF3_ENH)) {
>> +        hwcap |= HWCAP_S390_ETF3EH;
>> +    }
>> +    /* 31-bit processes can use 64-bit registers */
>> +    hwcap |= HWCAP_S390_HIGH_GPRS;
>> +    if (s390_has_feat(S390_FEAT_VECTOR)) {
>> +        hwcap |= HWCAP_S390_VXRS;
>> +    }
>> +
>> +    return hwcap;
>> +}
>> +
>>  static inline void init_thread(struct target_pt_regs *regs, struct image_info *infop)
>>  {
>>      regs->psw.addr = infop->entry;
>>
> 
> Acked-by: Laurent Vivier <laurent@vivier.eu>
> 
> You can also do like the other architectures that use shortcuts like this:
> 
> #define GET_FEATURE(feat, hwcap) \
>     do { if (s390_has_feat(feat)) { hwcaps |= hwcap; } } while (0)
> ...
> GET_FEATURE(S390_FEAT_VECTOR, HWCAP_S390_VXRS);
> ...

Thanks, I'll use this.

Cheers!

-- 

Thanks,

David / dhildenb