Message-ID: <daee329a-b4ae-fae3-8d59-6bcf7377df22@hauke-m.de>
Date:   Fri, 17 Apr 2020 19:05:11 +0200
From:   Hauke Mehrtens <hauke@...ke-m.de>
To:     "Ramuthevar, Vadivel MuruganX" 
        <vadivel.muruganx.ramuthevar@...ux.intel.com>,
        linux-kernel@...r.kernel.org, linux-mtd@...ts.infradead.org,
        devicetree@...r.kernel.org
Cc:     cheol.yong.kim@...el.com, hauke.mehrtens@...el.com,
        andriy.shevchenko@...el.com, anders.roxell@...aro.org,
        vigneshr@...com, arnd@...db.de, richard@....at,
        qi-ming.wu@...el.com, brendanhiggins@...gle.com,
        linux-mips@...r.kernel.org, robh+dt@...nel.org,
        boris.brezillon@...labora.com, miquel.raynal@...tlin.com,
        tglx@...utronix.de, masonccyang@...c.com.tw, piotrs@...ence.com,
        John Crispin <john@...ozen.org>
Subject: Re: [PATCH v2 2/2] mtd: rawnand: Add NAND controller support on Intel
 LGM SoC

On 4/17/20 10:21 AM, Ramuthevar, Vadivel MuruganX wrote:
> From: Ramuthevar Vadivel Murugan <vadivel.muruganx.ramuthevar@...ux.intel.com>

Thanks for adding me in CC, I am using my private mail it took me too
long to configure Outlook.

I compared the register description of the LGM with the VRX200, the EBU
and NAND block are looking very similar. I think the VRX200 also
supports HW ECC and DMA, nobody implimented it in the upstream driver.

I think replacing the old XWAY dirver with this one is a good approach.
Then we can add feature flags to activate the extra features only on the
SoCs which support them.

On older SoCs the EBU (NAND) and the PCI (not express) IP core are
sharing some parts like an endianess switch, this is causeing some more
problems.

> This patch adds the new IP of Nand Flash Controller(NFC) support
> on Intel's Lightning Mountain(LGM) SoC.
> 
> DMA is used for burst data transfer operation, also DMA HW supports
> aligned 32bit memory address and aligned data access by default.
> DMA burst of 8 supported. Data register used to support the read/write
> operation from/to device.
> 
> NAND controller driver implements ->exec_op() to replace legacy hooks,
> these specific call-back method to execute NAND operations.
> 
> Signed-off-by: Ramuthevar Vadivel Murugan <vadivel.muruganx.ramuthevar@...ux.intel.com>

.....

> +config MTD_NAND_INTEL_LGM
> +	tristate "Support for NAND controller on Intel LGM SoC"
> +	depends on X86

Compile test should also work for other archs

> +	help
> +	  Enables support for NAND Flash chips on Intel's LGM SoC.
> +          NAND flash interfaced through the External Bus Unit.
> +
>  comment "Misc"

.....

> +
> +#define LGM_CLC			0x000
> +#define LGM_CLC_RST		0x00000000u
> +
> +#define LGM_NAND_ECC_OFFSET	0x008
It is confusing that this is not a register but a different definition.
Please move it somewheer else.

> +#define LGM_ADDR_SEL(n)		(0x20 + (n) * 4)
> +#define LGM_ADDR_MASK		(5 << 4)
> +#define LGM_ADDR_SEL_REGEN	0x1
> +
> +#define LGM_BUSCON(n)		(0x60 + (n) * 4)
> +#define LGM_BUSCON_CMULT_V4	0x1
> +#define LGM_BUSCON_RECOVC(n)	((n) << 2)
> +#define LGM_BUSCON_HOLDC(n)	((n) << 4)
> +#define LGM_BUSCON_WAITRDC(n)	((n) << 6)
> +#define LGM_BUSCON_WAITWRC(n)	((n) << 8)
> +#define LGM_BUSCON_BCGEN_CS	0x0
> +#define LGM_BUSCON_SETUP_EN	BIT(22)
> +#define LGM_BUSCON_ALEC		0xC000
> +
> +#define NAND_CON		0x0B0
> +#define NAND_CON_NANDM_EN	BIT(0)
> +#define NAND_CON_NANDM_DIS	0x0
> +#define NAND_CON_CSMUX_E_EN	BIT(1)
> +#define NAND_CON_ALE_P_LOW	BIT(2)
> +#define NAND_CON_CLE_P_LOW	BIT(3)
> +#define NAND_CON_CS_P_LOW	BIT(4)
> +#define NAND_CON_SE_P_LOW	BIT(5)
> +#define NAND_CON_WP_P_LOW	BIT(6)
> +#define NAND_CON_PRE_P_LOW	BIT(7)
> +#define NAND_CON_IN_CS_S(n)	((n) << 8)
> +#define NAND_CON_OUT_CS_S(n)	((n) << 10)
> +#define NAND_CON_LAT_EN_CS_P	((0x3D) << 18)
> +
> +#define NAND_WAIT		0x0B4
> +#define NAND_WAIT_RDBY		BIT(0)
> +#define NAND_WAIT_WR_C		BIT(3)

The registers with the LGM_ prefix, NAND_CON and NAND_WAIT are from the
EBU (EBU_N) register block. I would prefer if the have the same prefix.
You should use a different prefix for the register definitions below
this comment, which are from the NAND Controller (HSNAND).

> +#define NAND_CTL1		0x110
> +#define NAND_CTL1_ADDR_3_SHIFT	24
> +
> +#define NAND_CTL2		0x114
> +#define NAND_CTL2_ADDR_5_SHIFT	8
> +#define NAND_CTL2_CYC_N_V5	(0x2 << 16)
> +
> +#define NAND_INT_MSK_CTL	0x124
> +#define NAND_INT_MSK_CTL_WR_C	BIT(4)
> +
> +#define NAND_INT_STA		0x128
> +#define NAND_INT_STA_WR_C	BIT(4)
> +
> +#define NAND_CTL		0x130
> +#define NAND_CTL_MODE_ECC	0x1
> +#define NAND_CTL_GO		BIT(2)
> +#define NAND_CTL_CE_SEL_CS(n)	BIT(3 + (n))
> +#define NAND_CTL_RW_READ	0x0
> +#define NAND_CTL_RW_WRITE	BIT(10)
> +#define NAND_CTL_ECC_OFF_V8TH	BIT(11)
> +#define NAND_CTL_CKFF_EN	0x0
> +#define NAND_CTL_MSG_EN		BIT(17)

Till here I find the same registers you use also in the VRX200
description. I haven't checked all the bits. Danube only has the
registers from the EBU, I haven't checked the SoCs in between.

> +#define NAND_PARA0		0x13c
> +#define NAND_PARA0_PAGE_V8192	0x3
> +#define NAND_PARA0_PIB_V256	(0x3 << 4)
> +#define NAND_PARA0_BYP_EN_NP	0x0
> +#define NAND_PARA0_BYP_DEC_NP	0x0
> +#define NAND_PARA0_TYPE_ONFI	BIT(18)
> +#define NAND_PARA0_ADEP_EN	BIT(21)
> +
> +#define NAND_CMSG_0		0x150
> +#define NAND_CMSG_1		0x154
> +
> +#define NAND_WRITE_CMD		(NAND_CON_CS_P_LOW | NAND_CON_CLE_P_LOW)
> +#define NAND_WRITE_ADDR		(NAND_CON_CS_P_LOW | NAND_CON_ALE_P_LOW)
> +#define NAND_WRITE_DATA		NAND_CON_CS_P_LOW
> +#define NAND_READ_DATA		NAND_CON_CS_P_LOW
> +
> +#define NAND_CHIP_NO_SELECTION	-1
> +#define NAND_CHIP_SELECTION	0x0
> +

....
> +static int lgm_nand_probe(struct platform_device *pdev)
> +{
> +	struct device *dev = &pdev->dev;
> +	struct lgm_nand_host *lgm_host;
> +	struct nand_chip *nand;
> +	phys_addr_t nandaddr_pa;
> +	struct mtd_info *mtd;
> +	struct resource *res;
> +	int ret;
> +	u32 cs;
> +
> +	lgm_host = devm_kzalloc(dev, sizeof(*lgm_host), GFP_KERNEL);
> +	if (!lgm_host)
> +		return -ENOMEM;
> +
> +	lgm_host->dev = dev;
> +	nand_controller_init(&lgm_host->controller);
> +
> +	lgm_host->lgm_va =
> +	devm_platform_ioremap_resource_byname(pdev, "lgmnand");
> +	if (IS_ERR(lgm_host->lgm_va))
> +		return PTR_ERR(lgm_host->lgm_va);
lgm_va is named ebu_n in the register description. Could you rename this
variable to ebu. I think the _va postfix is not needed as it should be
clear that this is a virtual addresss.

> +	lgm_host->hsnand_va =
> +	devm_platform_ioremap_resource_byname(pdev, "hsnand");
> +	if (IS_ERR(lgm_host->hsnand_va))
> +		return PTR_ERR(lgm_host->hsnand_va);
> +
> +	ret = device_property_read_u32(dev, "nand,cs", &cs);
> +	if (ret) {
> +		dev_err(dev, "failed to get chip select: %d\n", ret);
> +		return ret;
> +	}

The cs should be validated, ifsome uses cs == 4 there will be a problem.
LGM supports CS 0 to 3 like the VRX200, danube for example only 0 and 1.

> +	lgm_host->cs = cs;
> +
> +	lgm_host->cs_name = devm_kasprintf(dev, GFP_KERNEL, "nand_cs%d", cs);

cs_name is only used locally in this function you can also use some
memory on the stack to generate the name.

> +	if (IS_ERR(lgm_host->cs_name)) {
> +		ret = PTR_ERR(lgm_host->cs_name);
> +		dev_err(dev, "failed to get chip select name: %d\n", ret);
> +		return ret;
> +	}
> +
> +	res = devm_platform_ioremap_resource_byname(pdev, lgm_host->cs_name);

This will only support one chip select at a time on the SoC. It is not
possible to configure the DTS in a way to talk to two NAND chips on
different chip selects. I think this is an uncommon scenario and I do
not know if other NAND drivers in Linux support this feature.

> +	lgm_host->nandaddr_va = res;
> +	nandaddr_pa = res->start;
> +	if (IS_ERR(lgm_host->nandaddr_va))
> +		return PTR_ERR(lgm_host->nandaddr_va);
> +
> +	lgm_host->clk = devm_clk_get(dev, NULL);
> +	if (IS_ERR(lgm_host->clk)) {
> +		ret = PTR_ERR(lgm_host->clk);
> +		dev_err(dev, "failed to get clock: %d\n", ret);
> +		return ret;
> +	}
> +
> +	ret = clk_prepare_enable(lgm_host->clk);
> +	if (ret) {
> +		dev_err(dev, "failed to enable clock: %d\n", ret);
> +		return ret;
> +	}
> +	lgm_host->clk_rate = clk_get_rate(lgm_host->clk);
> +
> +	ret = lgm_dma_init(dev, lgm_host);
> +	if (ret)
> +		goto disable_clk;
> +
> +	writel(lower_32_bits(nandaddr_pa) | LGM_ADDR_SEL_REGEN | LGM_ADDR_MASK,
> +	       lgm_host->lgm_va + LGM_ADDR_SEL(cs));
> +
> +	writel(LGM_BUSCON_CMULT_V4 | LGM_BUSCON_RECOVC(1) |
> +	       LGM_BUSCON_HOLDC(1) | LGM_BUSCON_WAITRDC(2) |
> +	       LGM_BUSCON_WAITWRC(2) | LGM_BUSCON_BCGEN_CS | LGM_BUSCON_ALEC |
> +	       LGM_BUSCON_SETUP_EN, lgm_host->lgm_va + LGM_BUSCON(cs));
> +
> +	/*
> +	 * NAND physical address selection is based on the chip select
> +	 * and written to ADDR_SEL register to get Memory Region Base address.
> +	 * FPI Bus addresses are compared to this base address in conjunction
> +	 * with the mask control. Driver need to program this field!
> +	 * Address: 0x17400 if chip select is CS_0
> +	 * Address: 0x17C00 if chip select is CS_1
> +	 * Refer the Intel LGM SoC datasheet.

Could you please name the section name in the document, to make it
easier to find it, the number will probably change over time.

> +	 */
> +	writel(0x17400051, lgm_host->lgm_va + LGM_ADDR_SEL(0));
> +	writel(0x17C00051, lgm_host->lgm_va + LGM_ADDR_SEL(cs));

Please do not use magic values here. For example you set here the
LGM_ADDR_SEL_REGEN bit a mask and a base.

> +	nand_set_flash_node(&lgm_host->chip, dev->of_node);
> +	mtd = nand_to_mtd(&lgm_host->chip);
> +	mtd->dev.parent = dev;
> +	lgm_host->dev = dev;
> +
> +	platform_set_drvdata(pdev, lgm_host);
> +	nand_set_controller_data(&lgm_host->chip, lgm_host);
> +
> +	nand = &lgm_host->chip;
> +	nand->controller = &lgm_host->controller;
> +	nand->controller->ops = &lgm_nand_controller_ops;
> +
> +	/* Scan to find existence of the device */
> +	ret = nand_scan(&lgm_host->chip, 1);
> +	if (ret)
> +		goto exit_dma;
> +
> +	ret = mtd_device_register(mtd, NULL, 0);
> +	if (ret)
> +		goto clean_nand;
> +
> +	return 0;
> +
> +clean_nand:
> +	nand_cleanup(&lgm_host->chip);
> +exit_dma:
> +	lgm_dma_exit(lgm_host);
> +disable_clk:
> +	clk_disable_unprepare(lgm_host->clk);
> +
> +	return ret;
> +}

Hauke



Download attachment "signature.asc" of type "application/pgp-signature" (489 bytes)

Powered by blists - more mailing lists