[U-Boot] [PATCH v4 2/3] cfi_flash: convert to driver model

[Thomas Chou <thomas@wytron.com.tw>]

Hi Simon,

On 2015?11?06? 11:15, Simon Glass wrote:
>> +config CFI_FLASH
>> +       bool "Enable Driver Model for CFI Flash driver"
>> +       depends on MTD
>> +       help
>> +         The Common Flash Interface specification was developed by Intel,
>> +         AMD and other flash manufactures that provides a universal method
>
>           flash manufacturers. It provides a universal method
>

OK. Thanks.

>> +#ifdef CONFIG_CFI_FLASH /* for driver model */
>> +static void cfi_flash_init_dm(void)
>> +{
>> +       struct udevice *dev;
>> +
>> +       cfi_flash_num_flash_banks = 0;
>> +       /*
>> +        * The uclass_first_device() will probe the first device and
>> +        * uclass_next_device() will probe the rest if they exist. So
>> +        * that cfi_flash_probe() will get called assigning the base
>> +        * addresses that are available.
>> +        */
>> +       for (uclass_first_device(UCLASS_MTD, &dev);
>> +            dev;
>> +            uclass_next_device(&dev)) {
>> +       }
>
> Why do you need to probe these before they are used? Also I think you
> need error checking.

This is the place that the device is going to be used. In flash_init() 
the cfi flash is probed to find flash type and size.

There may be other MTD devices, so we scan through them. If there is not 
cfi flash device, cfi_flash_num_flash_banks will be zero and the 
flash_init will return size as zero. This is not an error.

>> +#ifdef CONFIG_FLASH_CFI_MTD
>> +       dev->uclass_priv = &cfi_mtd_info[0];
>
> Could you put cfi_mtd_info in a struct and use auto-allocation to allocate it?

I planned this as follow-up patches which will merge cfi-mtd.c and use 
mtd ops as major flash ops. Then the mtd_info will be auto-allocated. 
This is quite similar to Jagan's work on spi-nor.

Best regards,
Thomas