[U-Boot] [PATCH v4 05/12] SPEAr : nand driver support for SPEAr SoCs

[Tom <Tom.Rix@windriver.com>]

Scott Wood wrote:
> On Wed, Jan 13, 2010 at 07:13:17AM -0600, Tom wrote:
>> SPEAr SoCs contain an FSMC controller which can be used to interface
>> with a range of memories eg. NAND, SRAM, NOR.
>> Currently, this driver supports interfacing FSMC with NAND memories
>>
>> Signed-off-by: Vipin <vipin.kumar@st.com>
>> ---
>>   drivers/mtd/nand/Makefile             |    1 +
>>   drivers/mtd/nand/spr_nand.c           |  123 +++++++++++++++++++++++++++++++++
>>   include/asm-arm/arch-spear/spr_nand.h |   57 +++++++++++++++
>>   3 files changed, 181 insertions(+), 0 deletions(-)
>>   create mode 100755 drivers/mtd/nand/spr_nand.c
>>   create mode 100644 include/asm-arm/arch-spear/spr_nand.h
>>
>> <snip>
>> +}
>> +
>> +static int spear_read_hwecc(struct mtd_info *mtd,
>> +			    const u_char *data, u_char ecc[3])
>> +{
>> +	u_int ecc_tmp;
>> +
>> +	/* read the h/w ECC */
>> +	ecc_tmp = readl(&fsmc_regs_p->genmemctrl_ecc);
>> +
>> +	ecc[0] = (u_char) (ecc_tmp & 0xFF);
>> +	ecc[1] = (u_char) ((ecc_tmp & 0xFF00) >> 8);
>> +	ecc[2] = (u_char) ((ecc_tmp & 0xFF0000) >> 16);
>> +
>> +	return 0;
>>
>> Always returning 0.
>> return can be switched to void
> 
> No it can't, because it needs to match the signature of ecc.calculate.

You are correct!
And I was not..

> 
>> +int spear_nand_init(struct nand_chip *nand)
>> +{
>> +	writel(FSMC_DEVWID_8 | FSMC_DEVTYPE_NAND | FSMC_ENABLE | FSMC_WAITON,
>> +	       &fsmc_regs_p->genmemctrl_pc);
>> +	writel(readl(&fsmc_regs_p->genmemctrl_pc) | FSMC_TCLR_1 | FSMC_TAR_1,
>> +	       &fsmc_regs_p->genmemctrl_pc);
>> +	writel(FSMC_THIZ_1 | FSMC_THOLD_4 | FSMC_TWAIT_6 | FSMC_TSET_0,
>> +	       &fsmc_regs_p->genmemctrl_comm);
>> +	writel(FSMC_THIZ_1 | FSMC_THOLD_4 | FSMC_TWAIT_6 | FSMC_TSET_0,
>> +	       &fsmc_regs_p->genmemctrl_attrib);
>> +
>> +	nand->options = 0;
>> +	nand->ecc.mode = NAND_ECC_HW;
>> +	nand->ecc.layout = &spear_nand_ecclayout;
>> +	nand->ecc.size = 512;
>> +	nand->ecc.bytes = 3;
>> +	nand->ecc.calculate = spear_read_hwecc;
>> +	nand->ecc.hwctl = spear_enable_hwecc;
>> +	nand->ecc.correct = nand_correct_data;
>> +	nand->cmd_ctrl = spear_nand_hwcontrol;
>> +	return 0;
>> Same here for returning void.
> 
> IMHO it's better to preserve the ability to return an error in the future,
> especially in init functions.

Yes. That is perfectly valid.

Tom

> 
> -Scott