[U-Boot] [PATCH 2/9] cmd: nand: abstract global variable usage for dm conversion

[Scott Wood]

[Please CC me at this address rather than my NXP address]

On Fri, 2016-04-01 at 16:59 +0530, Mugunthan V N wrote:
> @@ -432,12 +435,7 @@ static int do_nand(cmd_tbl_t *cmdtp, int flag, int
> argc, char * const argv[])
>  	 * one before these commands can run, even if a partition specifier
>  	 * for another device is to be used.
>  	 */
> -	if (dev < 0 || dev >= CONFIG_SYS_MAX_NAND_DEVICE ||
> -	    !nand_info[dev].name) {
> -		puts("\nno devices available\n");
> -		return 1;
> -	}
> -	nand = &nand_info[dev];
> +	nand = get_nand_dev_by_index(dev);
>  
>  	if (strcmp(cmd, "bad") == 0) {

You eliminated the error check -- now a NULL deref is likely if a bad dev is
requested.  Even if it's checked elsewhere when setting nand_curr_device, it's
possible that the initial default is bad (no NAND devices present, or device 0
failed).

>  		printf("\nDevice %d bad blocks:\n", dev);
> @@ -496,13 +494,13 @@ static int do_nand(cmd_tbl_t *cmdtp, int flag, int
> argc, char * const argv[])
>  		/* skip first two or three arguments, look for offset and
> size */
>  		if (mtd_arg_off_size(argc - o, argv + o, &dev, &off, &size,
>  				     &maxsize, MTD_DEV_TYPE_NAND,
> -				     nand_info[dev].size) != 0)
> +				     nand->size) != 0)
>  			return 1;
>  
>  		if (set_dev(dev))
>  			return 1;
>  
> -		nand = &nand_info[dev];
> +		nand = get_nand_dev_by_index(dev);

Maybe have set_dev return the dev pointer?

Or have a global for the pointer rather than just the index, saving a bunch of
these calls.

-Scott