Re: [PATCH v4 2/3] mtd: mtk-nor: mtk serial flash controller driver

[Brian Norris <computersforpeace@gmail.com>]

Hi Bayi,

I'm looking over your v5, and I still have a lot of comments. I'll send
those soon. But I still question one of your responses here:

On Sun, Oct 18, 2015 at 10:20:35PM +0800, bayi.cheng wrote:
> On Fri, 2015-10-16 at 00:39 -0700, Brian Norris wrote:
> > On Tue, Oct 13, 2015 at 05:39:19PM +0800, Bayi Cheng wrote:

> > > diff --git a/drivers/mtd/spi-nor/mtk-quadspi.c b/drivers/mtd/spi-nor/mtk-quadspi.c
> > > new file mode 100644
> > > index 0000000..c6ac366
> > > --- /dev/null
> > > +++ b/drivers/mtd/spi-nor/mtk-quadspi.c
> > > @@ -0,0 +1,486 @@
[...]
> > > +#define MTK_NOR_CMD_REG			0x00
> > > +#define MTK_NOR_CNT_REG			0x04
> > > +#define MTK_NOR_RDSR_REG		0x08
> > > +#define MTK_NOR_RDATA_REG		0x0c
> > > +#define MTK_NOR_RADR0_REG		0x10
> > > +#define MTK_NOR_RADR1_REG		0x14
> > > +#define MTK_NOR_RADR2_REG		0x18
> > > +#define MTK_NOR_WDATA_REG		0x1c
> > > +#define MTK_NOR_PRGDATA0_REG		0x20
> > > +#define MTK_NOR_PRGDATA1_REG		0x24
> > > +#define MTK_NOR_PRGDATA2_REG		0x28
> > > +#define MTK_NOR_PRGDATA3_REG		0x2c
> > > +#define MTK_NOR_PRGDATA4_REG		0x30
> > > +#define MTK_NOR_PRGDATA5_REG		0x34

^^ so you have 6 "TX" registers.

> > > +#define MTK_NOR_SHREG0_REG		0x38
> > > +#define MTK_NOR_SHREG1_REG		0x3c
> > > +#define MTK_NOR_SHREG2_REG		0x40
> > > +#define MTK_NOR_SHREG3_REG		0x44
> > > +#define MTK_NOR_SHREG4_REG		0x48
> > > +#define MTK_NOR_SHREG5_REG		0x4c

^^ and you have at least 6 "RX" registers (looking at the mt8173 manual,
I'm not sure what the SHREG{6..9} are for). So I don't see why you can't
program MTK_NOR_CNT_REG to 48 (6 bytes * 8 bits/byte) to get 6 bytes
TX/RX, or IOW, 1 byte of opcode and 5 bytes of either RX or TX data. So
I'm still not convinced about your claim below.

I was able to rewrite your driver to do the above, and I can read out 5
bytes no problem, but unfortunately, the flash device I have has only 3
bytes of ID, so the last two bytes are zero, which doesn't tell me too
much if this is working right.

> > > +#define MTK_NOR_SHREG6_REG		0x50
> > > +#define MTK_NOR_SHREG7_REG		0x54
> > > +#define MTK_NOR_SHREG8_REG		0x58
> > > +#define MTK_NOR_SHREG9_REG		0x5c
> > > +#define MTK_NOR_CFG1_REG		0x60
> > > +#define MTK_NOR_CFG2_REG		0x64
> > > +#define MTK_NOR_CFG3_REG		0x68
> > > +#define MTK_NOR_STATUS0_REG		0x70
> > > +#define MTK_NOR_STATUS1_REG		0x74
> > > +#define MTK_NOR_STATUS2_REG		0x78
> > > +#define MTK_NOR_STATUS3_REG		0x7c
> > > +#define MTK_NOR_FLHCFG_REG		0x84
> > > +#define MTK_NOR_TIME_REG		0x94
> > > +#define MTK_NOR_PP_DATA_REG		0x98
> > > +#define MTK_NOR_PREBUF_STUS_REG		0x9c
> > > +#define MTK_NOR_DELSEL0_REG		0xa0
> > > +#define MTK_NOR_DELSEL1_REG		0xa4
> > > +#define MTK_NOR_INTRSTUS_REG		0xa8
> > > +#define MTK_NOR_INTREN_REG		0xac
> > > +#define MTK_NOR_CHKSUM_CTL_REG		0xb8
> > > +#define MTK_NOR_CHKSUM_REG		0xbc
> > > +#define MTK_NOR_CMD2_REG		0xc0
> > > +#define MTK_NOR_WRPROT_REG		0xc4
> > > +#define MTK_NOR_RADR3_REG		0xc8
> > > +#define MTK_NOR_DUAL_REG		0xcc
> > > +#define MTK_NOR_DELSEL2_REG		0xd0
> > > +#define MTK_NOR_DELSEL3_REG		0xd4
> > > +#define MTK_NOR_DELSEL4_REG		0xd8

...

> > > +static int mt8173_nor_read_reg(struct spi_nor *nor, u8 opcode, u8 *buf, int len)
> > > +{
> > > +	int ret;
> > > +	struct mt8173_nor *mt8173_nor = nor->priv;
> > > +
> > > +	/* mtk nor controller doesn't supoort SPINOR_OP_RDCR */
> > > +	switch (opcode) {
> > > +	case SPINOR_OP_RDID:
> > > +		/* read JEDEC ID need 4 bytes commands */
> > > +		ret = mt8173_nor_set_cmd(mt8173_nor, 0, 32, SPINOR_OP_RDID);
> > > +		if (ret < 0)
> > > +			return ret;
> > > +
> > > +		/* mtk nor flash controller only support 3 bytes IDs */
> > 
> > Are you absolutely sure of this? That would be highly unfortunate, but
> > I also don't believe it's true.
> > 
> Yes, for this issue I have asked our designer of nor flash controller,
> unfortunately, it is true, and I have tried to read more IDs, but our
> controller just accept 3 IDs from nor flash, and our next generation IC
> may solve this problem.

How exactly did you try? Did you do what I suggested above? Are the
"shift registers" not all actually functional?

> > > +		buf[2] = readb(mt8173_nor->base + MTK_NOR_SHREG0_REG);
> > > +		buf[1] = readb(mt8173_nor->base + MTK_NOR_SHREG1_REG);
> > > +		buf[0] = readb(mt8173_nor->base + MTK_NOR_SHREG2_REG);
> > > +		break;

[...]

Brian