Re: [PATCHv5 1/4] arm64: io: Use asm-generic high level MMIO accessors

[Sai Prakash Ranjan <quic_saipraka@quicinc.com>]

On 12/6/2021 2:20 PM, Arnd Bergmann wrote:
> On Mon, Dec 6, 2021 at 9:28 AM Sai Prakash Ranjan
> <quic_saipraka@quicinc.com> wrote:
>> Remove custom arm64 MMIO accessors read{b,w,l,q} and their relaxed
>> versions in support to use asm-generic ones. Also define arm64
>> barrier macros to override the asm-generic defined barriers.
>>
>> Suggested-by: Arnd Bergmann <arnd@arndb.de>
>> Signed-off-by: Sai Prakash Ranjan <quic_saipraka@quicinc.com>
> This looks correct, but I would change one detail:
>
>> +#define __io_ar(v)             __io_par(v)
>> +#define __io_bw()              __iowmb()
>> +#define __io_br(v)
>> +#define __io_aw(v)
> The default __io_par() is defined in terms of __io_ar(), so it would
> be more logical
> to remove the custom __io_par() and just define __io_ar() here.

Makes sense, will do this.

> I think it would be even better to flip these around and make the low-level
> definitions __io_ar() and __io_bw(), and then defining the arm64 specific
> macros based on those:
>
> /* arm64-specific, don't use in portable drivers */
> #define __iormb(v)     __io_ar(v)
> #define __iowmb()      __io_bw()
> #define __iomb()        dma_mb()
>
>

So __iormb on arm64 has some dummy control dependency stuff as well based on
("arm64: io: Ensure calls to delay routines are ordered against prior 
readX()") and then we would
need to change __iormb definition to __io_ar which doesn't seem like 
__iormb definition to be exact
right?

Thanks,
Sai