Re: [PATCH 09/13] media: qcom: camss: Add CSID Gen3 support for SM8550

[Krzysztof Kozlowski <krzk@kernel.org>]

On 11/07/2024 14:00, Bryan O'Donoghue wrote:
> On 11/07/2024 12:41, Depeng Shao wrote:
>>>> Yes, these are some sequences to initialize the HW.
>>>
>>> Hm? It's like you ignore the problem and just answer with whatever to
>>> shut up the reviewer. Instead of replying with the same, address the
>>> problem. Why ordering is not a problem here?
>>>
>>
>> Sorry, I didn't mean that, was trying to understand the problem, then 
>> just sent out the mail by mistake.
>> Do you mean we should use writel to ensure the strict sequences?
>> Thanks for catching this problem, this problem is also in the the 
>> existing camss driver. I will check all of them in this series, but the 
>> problem in some existing camss drivers, maybe Bryan from Linaro can help 
>> to fix them, since I don't have these devices to verify the modifications.
> 
> _relaxed is used I'm sure because that's what's always been used and 
> what downstream does.
> 
> Is there a good reason for it ? None that I can think of.
> 
> Krzysztof is right, there's no good reason to use relaxed() here at all, 
> you should drop it.
> 

In many cases relaxed will be fine, but in few might lead to tricky to
debug issues thus people introducing msleep() or other workarounds.
Usually init sequences are "sequences" for a reason, but of course here
maybe it does not matter.

Best regards,
Krzysztof