Re: [PATCH 1/3] scsi: ufshcd: switch to a version macro

[Caleb Connolly <caleb@connolly.tech>]

Hi Christoph,

On 08/03/2021 8:00 am, Christoph Hellwig wrote:
> This looks like a really nice improvement!
>
> A bunch of comments below:
>
>> @@ -696,10 +685,21 @@ static inline u32 ufshcd_get_intr_mask(struct ufs_hba *hba)
>>    */
>>   static inline u32 ufshcd_get_ufs_version(struct ufs_hba *hba)
>>   {
>> +	u32 ufshci_ver;
> missing eempty line after the declaration.
>
>>   	if (hba->quirks & UFSHCD_QUIRK_BROKEN_UFS_HCI_VERSION)
>> +		ufshci_ver = ufshcd_vops_get_ufs_hci_version(hba);
>> +	else
>> +		ufshci_ver = ufshcd_readl(hba, REG_UFS_VERSION);
>>
>> +	/*
>> +	 * UFSHCI v1.x uses a different version scheme, in order
>> +	 * to allow the use of comparisons with the UFSHCI_VER
>> +	 * macro, we convert it to the same scheme as ufs 2.0+.
>> +	 */
>> +	if (ufshci_ver & 0x00010000)
>> +		ufshci_ver = UFSHCI_VER(1, ufshci_ver & 0x00000100);
>> +
>> +	return ufshci_ver;
> I'd use early returns here to clean this up a bit:
>
>   	if (hba->quirks & UFSHCD_QUIRK_BROKEN_UFS_HCI_VERSION)
> 		ufshci_ver = ufshcd_vops_get_ufs_hci_version(hba);
>
> 	...
> 	ufshci_ver = ufshcd_readl(hba, REG_UFS_VERSION);
> 	if (ufshci_ver & 0x00010000)
> 		return UFSHCI_VER(1, ufshci_ver & 0x00000100);
> 	return ufshci_ver;
>
>> +#define UFSHCI_VER(major, minor) \
>> +	((major << 8) + (minor << 4))
> This needs braces around major and minor.  Or better just convert it
> to an inline function (and use a lower case name).

Other (similar) implementations, like NVME_VS() use a macro here, is an 
inline function just personal preference?

I'm perfectly happy either way, so I'll switch to your suggestion.

Thanks for the review.

Regards,

Caleb