[U-Boot] [PATCH v4 04/12] SPEAr : smi driver support for SPEAr SoCs
Vipin Kumar
hasherror at gmail.com
Thu Jan 14 11:50:39 CET 2010
Hello Tom,
> +
> +#define ST_M25Pxx_ID 0x00002020
> +
> +static struct flash_dev flash_ids[] = {
> + {0x10, 64 * 1024, 2},
> + {0x11, 128 * 1024, 4},
> + {0x12, 256 * 1024, 4},
> + {0x13, 512 * 1024, 8},
> + {0x14, 1 * 1024 * 1024, 16},
> + {0x15, 2 * 1024 * 1024, 32},
> + {0x16, 4 * 1024 * 1024, 64},
> + {0x17, 8 * 1024 * 1024, 128},
> + {0x18, 16 * 1024 * 1024, 64},
> + {0x00,}
> +};
> +
> Could be accessed like
> {
> .densisty = xxx,
> .size = xxx,
> .sector_count = xxx,
> }
>
This is meant to be a look up table, that's why I think the previous
way is batter.
Actually, I got the inspiration from drivers/mtd/nand/nand_ids.c
Still, I have changed it to look as below
{0x10, 0x10000, 2},
{0x11, 0x20000, 4},
{0x12, 0x40000, 4},
{0x13, 0x80000, 8},
size is changed to hex to stay vertically aligned
Please let me know if this is OK
> + *
> + * Detect the SMI flash by reading the ID. Initializes the flash_info structure
> + * with size, sector count etc.
> + */
> +static ulong flash_get_size(ulong base, int banknum)
> +{
> + flash_info_t *info = &flash_info[banknum];
> + struct flash_dev *dev;
> + unsigned int value;
> + unsigned int density;
> + int i;
> +
> + value = smi_read_id(info, banknum);
> + density = (value >> 16) & 0xff;
> +
> + for (i = 0, dev = &flash_ids[0]; dev->density != 0x0;
> + i++, dev = &flash_ids[i]) {
> + if (dev->density == density) {
> + info->size = dev->size;
> + info->sector_count = dev->sector_count;
> + break;
> + }
> + }
> From the flash_ids's struct, it looks like 'density' field is
> unique and increasing by 1. You may be able to replay this loop
> with somthing like
>
> density -= DENSITY_START
> if (density < 0)
> bail
> else if (density > DESITY_END)
> bail
>
> then use desity as an index into the flash_ids array
> info->size = flash_ids[density].size
>
density is one byte of the read id which is read from flash itself
as of now, we are supporting consecutive values but these values
may be rendom in future
Thats why I used a look up table
> +static int smi_wait_till_ready(int bank, int timeout)
> +{
> + int count;
> + int sr;
> +
> + /* One chip guarantees max 5 msec wait here after page writes,
> + but potentially three seconds (!) after page erase. */
> + for (count = 0; count < timeout; count++) {
> +
> + sr = smi_read_sr(bank);
> + if (sr < 0)
> + break;
> + else if (!(sr & WIP_BIT))
> + return 0;
> +
> + /* Try again after 1m-sec */
> + udelay(1000);
> + }
> + printf("SMI controller is still in wait, timeout=%d\n", timeout);
> + return -EIO;
>
> Always failure ?
>
There is a return 0 in case WIP_BIT gets reset
else if (!(sr & WIP_BIT))
return 0;
All the other comments are accepted and I would send a patchset v5 soon
Regards
Vipin
More information about the U-Boot
mailing list