Re: [PATCH] mtd: nand: spi: rework detect procedure for different read id op

[Boris Brezillon]

On Thu,  9 Jan 2020 15:54:00 +0800
Chuanhong Guo <gch981213@xxxxxxxxx> wrote:

> @@ -215,15 +204,22 @@ struct spinand_manufacturer_ops {
>   * struct spinand_manufacturer - SPI NAND manufacturer instance
>   * @id: manufacturer ID
>   * @name: manufacturer name
> + * @devid_len: number of bytes in device ID
> + * @spinand_table: array with info for spi nands under current manufacturer
> + * @nchips: number of chips available in spinand_table
>   * @ops: manufacturer operations
>   */
>  struct spinand_manufacturer {
>  	u8 id;
>  	char *name;
> +	u8 devid_len;

IIRC, some manufacturers support more than one scheme which means you
can't really take this decision at the manufacturer level. How about
adding a readid_method field to spinand_info?

enum spinand_readid_method {
	SPINAND_READID_METHOD_OPCODE,
	SPINAND_READID_METHOD_OPCODE_ADDR,
	SPINAND_READID_METHOD_OPCODE_DUMMY,
};

> +	const struct spinand_info *spinand_table;

s/spinand_table/chips/

> +	size_t nchips;
>  	const struct spinand_manufacturer_ops *ops;
>  };