Re: [PATCH v3 3/3] New I2C: Add support for TPM devices over I2C bus

[Stefan Berger <stefanb@linux.ibm.com>]

On 3/24/23 09:43, Stefan Berger wrote:
> 
> 
> On 3/24/23 04:24, Cédric Le Goater wrote:
>> Hello Ninad,
>>
> 
>>> +
>>> +/*
>>> + * Convert little endian byte stream into local formated
>>> + * unsigned integer
>>> + */
>>> +static inline uint32_t tpm_i2c_le_bytes_to_uint(TPMStateI2C *i2cst)
>>> +{
>>> +    uint32_t data = 0;
>>> +    int      i;
>>> +
>>> +    assert(i2cst->offset <= 5);  /* Including 0th register value */
>>> +
>>> +    if (tpm_i2c_is_little_endian()) {
>>> +        for (i = 1; i < i2cst->offset; i++) {
>>> +            data |= (((uint32_t)i2cst->data[i]) << (8 * (i - 1)));
>>> +        }
>>> +    } else {
>>> +        for (i = 1; i < i2cst->offset; i++) {
>>> +            data <<= 8;
>>> +            data |= i2cst->data[i];
>>> +        }
>>> +    }
>>> +
>>> +    return data;
>>> +}
>>
>> Why is this endianess test  needed ? Something looks wrong.
>>
>> Could you please explain the endianess used by the various devices and
>> the endianness of the MMIO ops ?
>>
>> If you want to test on a BE host, simply install a debian PPC64 VM/LPAR
>> on a POWER9/10 box. I suppose you have access to one in IBM.
> 
> I found it's not that simple anymore since all distros moved to little endian. I have been using Fedora 28 for big endian tests for swtpm but I have never tried it with QEMU...


... Fedora 28 is still good enough to build QEMU.

    Stefan
> 
>     Stefan
>