Re: [PATCH 7/9] nvme: apple: Add Apple A11 support

[Sven Peter <sven@kernel.org>]

On 11.08.25 15:50, Nick Chan wrote:
> Add support for ANS2 NVMe on Apple A11 SoC.
> 
> This version of ANS2 is less quirky than the one in M1, and does not have
> NVMMU or Linear SQ. However, it still requires a non-standard 128-byte
> SQE.
> 
> Signed-off-by: Nick Chan <towinchenmi@gmail.com>
> ---
>   drivers/nvme/host/apple.c | 228 +++++++++++++++++++++++++++++++---------------

[...]

>   }
>   
>   static void apple_nvme_rtkit_crashed(void *cookie, const void *crashlog, size_t crashlog_size)
> @@ -284,21 +294,8 @@ static void apple_nvme_submit_cmd(struct apple_nvme_queue *q,
>   				  struct nvme_command *cmd)
>   {

Please just create a separate submit function here.
There's just not much code that's shared between the two variants.

[...]

>   }
>   
> @@ -587,10 +618,17 @@ static inline void apple_nvme_handle_cqe(struct apple_nvme_queue *q,
>   {
>   	struct apple_nvme *anv = queue_to_apple_nvme(q);
>   	struct nvme_completion *cqe = &q->cqes[idx];
> -	__u16 command_id = READ_ONCE(cqe->command_id);
>   	struct request *req;
>   
> -	apple_nvmmu_inval(q, command_id);
> +	if (!anv->hw->has_lsq_nvmmu)
> +		cqe->command_id--;
> +
> +	__u16 command_id = READ_ONCE(cqe->command_id);
> +
> +	if (anv->hw->has_lsq_nvmmu)
> +		apple_nvmmu_inval(q, command_id);
> +	else
> +		command_id++;

This entire block here looks weird. First you decrease the command_id
directly inside the shared memory structure, then you read it with
READ_ONCE to a local variable only to increase it again. Why?



Sven