Re: [PATCH v10] mtd: spi-nor: add hisilicon spi-nor flash controller driver

From: Jiancheng Xue
Date: Wed May 04 2016 - 04:27:37 EST


Hi Cyrille,

On 2016/4/27 19:55, Cyrille Pitchen wrote:
> Hi Jiancheng,
>
> Le 19/04/2016 09:27, Jiancheng Xue a écrit :
>> From: Jiancheng Xue <xuejiancheng@xxxxxxxxxx>
>>
>> Add hisilicon spi-nor flash controller driver
>
> [...]
>> +enum hifmc_iftype {
>> + IF_TYPE_STD,
>> + IF_TYPE_DUAL,
>> + IF_TYPE_DIO,
>> + IF_TYPE_QUAD,
>> + IF_TYPE_QIO,
>> +};
>
> Just for my own information, the hisilicon controller supports:
> - SPI 1-1-1 : IF_TYPE_STD
> - SPI 1-1-2 : IF_TYPE_DUAL
> - SPI 1-2-2 : IF_TYPE_DIO
> - SPI 1-1-4 : IF_TYPE_QUAD
> - SPI 1-4-4 : IF_TYPE_QIO
>
> but not the SPI protocols 2-2-2 or 4-4-4, does it?

You are right.

> If I ask you this question, it's because I wonder how to make the difference
> between SPI controllers supporting both SPI 1-4-4 and 4-4-4 and those
> supporting only SPI 1-4-4 in the case they rely on the m25p80 driver as an
> adaptation layer between the spi-nor framework et the spi framework.
>
> I guess the "spi-tx-bus-width" and "spi-rx-bus-width" DT properties are not
> enough to make the difference between these two kinds of SPI controller.
>
> I understand your driver doesn't use the m25p80 driver as it directly calls
> spi_nor_scan() and implements the .read_reg, .write_reg, .read, .write and
> .erase hooks, but its a general question :)
>
IMO, the "spi-tx-bus-width" and "spi-rx-bus-width" DT properties which
will be stored in spi_device.mode just represent modes of the spi device,
but not the ones of the spi controller. We should add a specific member for
the controller to indicate all modes which it supports.

The definition of spi_nor_scan can be modified to
int spi_nor_scan(struct spi_nor *nor, const char *name, int mode).
The argument "mode" represents all modes which the controller supports.
For example,
#define SPI_1_1_1 (1 << 0)
#define SPI_1_1_4 (1 << 0)
#define SPI_1_4_4 (1 << 2)
#define SPI_4_4_4 (1 << 3)
If the controller supports SPI_1_1_1, SPI_1_1_4 and SPI_1_4_4, the mode
passed to spi_nor_scan will be (SPI_1_1_1 | SPI_1_1_4 | SPI_1_4_4).

The really read_mode we use when transferring should depend on both controller's
modes(i.e. the argument "mode") and device's modes(i.e. struct flash_info.flags,
which can be got by spi_nor_read_id()).

>> +static int get_if_type(enum read_mode flash_read)
>> +{
>> + enum hifmc_iftype if_type;
>> +
>> + switch (flash_read) {
>> + case SPI_NOR_DUAL:
>> + if_type = IF_TYPE_DUAL;
>> + break;
>> + case SPI_NOR_QUAD:
>> + if_type = IF_TYPE_QUAD;
>> + break;
>> + case SPI_NOR_NORMAL:
>> + case SPI_NOR_FAST:
>> + default:
>> + if_type = IF_TYPE_STD;
>> + break;
>> + }
>> +
>> + return if_type;
>> +}
>
> Here I understand your QSPI controller could support SPI 1-4-4 and SPI 1-2-2
> but you limit it to SPI 1-1-4 and SPI 1-1-2 because the spi-nor framework
> doesn't support those protocols yet.
>
Yes. You're right.
It is one of the reasons that the spi-nor framework doesn't support those two protocols.
We also found protocols supported were enough for our product solutions now. So
we didn't implement SPI-1-4-4 and SPI-1-2-2 in this patch.

> [...]
>
>> +static void hisi_spi_nor_cmd_prepare(struct hifmc_host *host, u8 cmd,
>> + u32 *opcfg)
>> +{
>> + u32 reg;
>> +
>> + *opcfg |= FMC_OP_CMD1_EN;
>> + switch (cmd) {
>> + case SPINOR_OP_RDID:
>> + case SPINOR_OP_RDSR:
>> + case SPINOR_OP_RDCR:
>> + *opcfg |= FMC_OP_READ_DATA_EN;
>> + break;
>> + case SPINOR_OP_WREN:
>> + reg = readl(host->regbase + FMC_GLOBAL_CFG);
>> + if (reg & FMC_GLOBAL_CFG_WP_ENABLE) {
>> + reg &= ~FMC_GLOBAL_CFG_WP_ENABLE;
>> + writel(reg, host->regbase + FMC_GLOBAL_CFG);
>> + }
>> + break;
>> + case SPINOR_OP_WRSR:
>> + *opcfg |= FMC_OP_WRITE_DATA_EN;
>> + break;
>> + case SPINOR_OP_BE_4K:
>> + case SPINOR_OP_BE_4K_PMC:
>> + case SPINOR_OP_SE_4B:
>> + case SPINOR_OP_SE:
>> + *opcfg |= FMC_OP_ADDR_EN;
>> + break;
>> + case SPINOR_OP_EN4B:
>> + reg = readl(host->regbase + FMC_CFG);
>> + reg |= SPI_NOR_ADDR_MODE;
>> + writel(reg, host->regbase + FMC_CFG);
>> + break;
>> + case SPINOR_OP_EX4B:
>> + reg = readl(host->regbase + FMC_CFG);
>> + reg &= ~SPI_NOR_ADDR_MODE;
>> + writel(reg, host->regbase + FMC_CFG);
>> + break;
>> + case SPINOR_OP_CHIP_ERASE:
>> + default:
>> + break;
>> + }
>> +}
>
> IMHO, this is exactly what should be avoided in (Q)SPI controller drivers:
> they should not analyse the command op code to guess some settings such as
> the SPI protocol to be used or in your case whether some address or data
> cycles are included inside the command.
>
> Why? Just because your driver won't know what to do when we introduce new
> op codes. Right now, hisi_spi_nor_cmd_prepare() is not aware of op code 0x15
> use by Macronix to read its Configuration Register or Micron op codes to
> read/write their Volatile Configuration Register and Enhanced Volatile
> Configuration Register. So your driver won't support those memories any longer
> when new features using those registers will be added in the spi-nor framework.
>
>
> Since your driver implements struct spi_nor hooks, here is what you could do
> instead:
>
> 1 - hisi_spi_nor_read_reg(..., u8 *buf, int len)
>
> if buf != NULL then you should set FMC_OP_READ_DATA_EN, don't set it otherwise.
>
>
> 2 - hisi_spi_nor_write_reg(..., u8 *buf, int len)
>
> buf should always be != NULL so I guess you can always set FMC_OP_WRITE_DATA_EN
>
> For the special case of SPINOR_OP_WREN (Write Enable), your driver seems
> to clear the FMC_GLOBAL_CFG_WP_ENABLE bit from the FMC_GLOBAL_CFG register, if
> set, but never set it again... So why not clear this bit once for all?
>
> Reading the current source code, the support of the hardware write protect
> feature is a little bit odd, isn't it?
>
> 3 - hisi_spi_nor_read(struct spi_nor *nor, ...)
>
> your driver doesn't seem to call hisi_spi_nor_send_cmd() /
> hisi_spi_nor_cmd_prepare() from _read() so I don't know whether you have to
> set the FMC_OP_READ_DATA_EN bit here.
>
> Also you can check nor->addr_width to know whether you need SPI_NOR_ADDR_MODE.
>
>
> 4 - hisi_spi_nor_write(struct spi_nor *nor, ...)
>
> Almost the same comment as for _read(): just replace FMC_OP_READ_DATA_EN by
> FMC_OP_WRITE_DATA_EN I guess.
>
>
> 5 - hisi_spi_nor_erase(struct spi_nor *nor, ...)
>
> Here again you should check nor->addr_width to know when to set the
> SPI_NOR_ADDR_MODE bit in the FMC_CFG register.
>
> The FMC_OP_ADDR_EN bit should always be set for erase operations.
>
>
>> +static int hisi_spi_nor_send_cmd(struct spi_nor *nor, u8 cmd, int len)
>> +{
>> + struct hifmc_priv *priv = nor->priv;
>> + struct hifmc_host *host = priv->host;
>> + u32 reg, op_cfg = 0;
>> +
>> + hisi_spi_nor_cmd_prepare(host, cmd, &op_cfg);
> [...]
>
> Just get rid of hisi_spi_nor_cmd_prepare(), it would simplify the support of
> your driver a lot. For sure, I don't know the hisilicon spi-nor flash
> controller as much as you do but I'm pretty sure its driver doesn't need to
> analyse the op code to guess some hardware settings. There are other means to
> find out these settings.
>
> Anyway, I hope these few comments will help you.
>
Thank you very much for your detailed comments. They help me a lot. I looked into
the datasheet and this driver. I found it was really a big mistake for
hisi_spi_nor_cmd_prepare to configure registers according to the command op code.

I rewrote corresponding functions like below.

static int hisi_spi_nor_op_reg(struct spi_nor *nor, u8 opcode, int len, u8 optype)
{
struct hifmc_priv *priv = nor->priv;
struct hifmc_host *host = priv->host;
u32 reg;

reg = FMC_CMD_CMD1(opcode);
writel(reg, host->regbase + FMC_CMD);

reg = FMC_DATA_NUM_CNT(len);
writel(reg, host->regbase + FMC_DATA_NUM);

reg = OP_CFG_FM_CS(priv->chipselect);
writel(reg, host->regbase + FMC_OP_CFG);

writel(0xff, host->regbase + FMC_INT_CLR);
reg = FMC_OP_CMD1_EN;
if (optype == FMC_OP_READ) {
reg |= FMC_OP_READ_DATA_EN;
} else {
reg |= FMC_OP_WRITE_DATA_EN;
}
reg |= FMC_OP_REG_OP_START;
writel(reg, host->regbase + FMC_OP);

return wait_op_finish(host);
}

static int hisi_spi_nor_read_reg(struct spi_nor *nor, u8 opcode, u8 *buf,
int len)
{
struct hifmc_priv *priv = nor->priv;
struct hifmc_host *host = priv->host;
int ret;

ret = hisi_spi_nor_op_reg(nor, opcode, len, FMC_OP_READ);
if (ret)
return ret;

memcpy_fromio(buf, host->iobase, len);
return 0;
}

static int hisi_spi_nor_write_reg(struct spi_nor *nor, u8 opcode,
u8 *buf, int len)
{
struct hifmc_priv *priv = nor->priv;
struct hifmc_host *host = priv->host;

if (len)
memcpy_toio(host->iobase, buf, len);

return hisi_spi_nor_op_reg(nor, opcode, len, FMC_OP_WRITE);
}

static int hisi_spi_nor_register(struct device_node *np,
struct hifmc_host *host)
{
struct spi_nor *nor;
...
nor->read_reg = hisi_spi_nor_read_reg;
nor->write_reg = hisi_spi_nor_write_reg;
nor->erase = NULL; /*use the default implementation*/
...
spi_nor_scan(nor, NULL, SPI_NOR_QUAD);
...
}

Any new comments will be highly appreciated. Thank you.

Regards,
Jiancheng