[RFC patch 2/4] riscv: Get CPU manufacturer information
Damien Le Moal
Damien.LeMoal at wdc.com
Tue Mar 9 01:59:43 GMT 2021
On 2021/03/09 10:25, Vincent Chen wrote:
>>> #endif /* _ASM_RISCV_PROCESSOR_H */
>>> diff --git a/arch/riscv/include/asm/soc.h b/arch/riscv/include/asm/soc.h
>>> index f494066051a2..03dee6db404c 100644
>>> --- a/arch/riscv/include/asm/soc.h
>>> +++ b/arch/riscv/include/asm/soc.h
>>> @@ -10,6 +10,7 @@
>>> #include <linux/of.h>
>>> #include <linux/linkage.h>
>>> #include <linux/types.h>
>>> +#include <asm/hwcap.h>
>>>
>>> #define SOC_EARLY_INIT_DECLARE(name, compat, fn) \
>>> static const struct of_device_id __soc_early_init__##name \
>>> diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
>>> index ac202f44a670..389162ee19ea 100644
>>> --- a/arch/riscv/kernel/cpufeature.c
>>> +++ b/arch/riscv/kernel/cpufeature.c
>>> @@ -12,6 +12,8 @@
>>> #include <asm/hwcap.h>
>>> #include <asm/smp.h>
>>> #include <asm/switch_to.h>
>>> +#include <asm/sbi.h>
>>> +#include <asm/csr.h>
>>>
>>> unsigned long elf_hwcap __read_mostly;
>>>
>>> @@ -22,6 +24,8 @@ static DECLARE_BITMAP(riscv_isa, RISCV_ISA_EXT_MAX) __read_mostly;
>>> bool has_fpu __read_mostly;
>>> #endif
>>>
>>> +struct cpu_manufacturer_info_t cpu_mfr_info;
>>> +
>>> /**
>>> * riscv_isa_extension_base() - Get base extension word
>>> *
>>> @@ -149,3 +153,16 @@ void riscv_fill_hwcap(void)
>>> has_fpu = true;
>>> #endif
>>> }
>>> +
>>> +void riscv_fill_cpu_manufacturer_info(void)
>>> +{
>>> +#ifndef CONFIG_RISCV_M_MODE
>>> + cpu_mfr_info.vendor_id = sbi_get_vendorid();
>>> + cpu_mfr_info.arch_id = sbi_get_archid();
>>> + cpu_mfr_info.imp_id = sbi_get_impid();
>>> +#else
>>> + cpu_mfr_info.vendor_id = csr_read(CSR_MVENDORID);
>>> + cpu_mfr_info.arch_id = csr_read(CSR_MARCHID);
>>> + cpu_mfr_info.imp_id = csr_read(CSR_MIMPID);
>>> +#endif
>>
>> Why ? reading the registers will work with M-Mode too. It was there before when
>> we temporarily had the builtin DTB lookup based on vendor/arch/imp (see defunct
>> soc_lookup_builtin_dtb() in 5.11).
>>
> Sorry, I cannot fully catch what you mean.
> I agree that reading these registers will work with M-Mode, so I used
> the macro csr_read() to read these three registers when the kernel run
> in the M-mode.
> The definition of csr_read() is here:
> #define csr_read(csr) \
> ({ \
> register unsigned long __v; \
> __asm__ __volatile__ ("csrr %0, " __ASM_STR(csr) \
> : "=r" (__v) : \
> : "memory"); \
> __v; \
> })
> I guess that #if"n"def CONFIG_RISCV_M_MODE may mislead you. #ifndef is
> really unfriendly to read. I will change #ifndef to #ifdef in my next
> version patch.
You are right. I totally misread this :) It was my morning and I needed more
coffee ! Apologies for the noise. This looks fine.
>
>
>>> +}
>>> diff --git a/arch/riscv/kernel/setup.c b/arch/riscv/kernel/setup.c
>>> index e85bacff1b50..03621ce9092c 100644
>>> --- a/arch/riscv/kernel/setup.c
>>> +++ b/arch/riscv/kernel/setup.c
>>> @@ -278,6 +278,8 @@ void __init setup_arch(char **cmdline_p)
>>> #endif
>>>
>>> riscv_fill_hwcap();
>>> +
>>
>> Nit: I do not think the white libe is really necessary here.
>>
> OK, I will remove it.
>>> + riscv_fill_cpu_manufacturer_info();
>>> }
>>>
>>> static int __init topology_init(void)
>>> diff --git a/arch/riscv/kernel/soc.c b/arch/riscv/kernel/soc.c
>>> index a0516172a33c..58f6fd91743a 100644
>>> --- a/arch/riscv/kernel/soc.c
>>> +++ b/arch/riscv/kernel/soc.c
>>> @@ -6,6 +6,7 @@
>>> #include <linux/libfdt.h>
>>> #include <linux/pgtable.h>
>>> #include <asm/soc.h>
>>> +#include <asm/hwcap.h>
>>
>> Why is this necessary ?
>
> I forgot to remove it when doing code cleanup. Thank you for the
> reminder. I will remove it
>
--
Damien Le Moal
Western Digital Research
More information about the linux-riscv
mailing list