Re: [PATCH 2/2] rockchip: efuse: add efuse driver for rk3288 efuse

From: Heiko Stübner
Date: Mon Dec 01 2014 - 09:07:39 EST


Hi Jianqun,

Am Montag, 1. Dezember 2014, 15:34:41 schrieb Jianqun Xu:
> Add driver for efuse found on rk3288 board based on rk3288 SoC.
> Driver will read fuse information of chip at the boot stage of
> kernel, this information new is for further usage.
>
> Signed-off-by: Jianqun Xu <jay.xu@xxxxxxxxxxxxxx>

General question would be, what is the purpose of this driver?
I don't see any publically usable functions and the only thing happening is
the
dev_info(efuse->dev, "leakage (%d %d %d)\n",...
output that emits some information from the efuse to the kernel log?


In the dt-binding doc you write:
> The 32x32 eFuse can only be accessed by APB bus when IO_SECURITYsel is high.

While the TRM also says this, IO_SECURITY is not mentioned anywhere else in
the document. Is this a pin or a bit somewhere in the GRF - i.e. something
whichs state is readable?


Some more comments inline.

> ---
> arch/arm/mach-rockchip/efuse.c | 165
> +++++++++++++++++++++++++++++++++++++++++ arch/arm/mach-rockchip/efuse.h |
> 15 ++++
> 2 files changed, 180 insertions(+)
> create mode 100644 arch/arm/mach-rockchip/efuse.c
> create mode 100644 arch/arm/mach-rockchip/efuse.h
>
> diff --git a/arch/arm/mach-rockchip/efuse.c b/arch/arm/mach-rockchip/efuse.c
> new file mode 100644
> index 0000000..326d81e
> --- /dev/null
> +++ b/arch/arm/mach-rockchip/efuse.c

a driver like this should probably live in something like
drivers/soc/rockchip.


> @@ -0,0 +1,165 @@
> +/* mach-rockchip/efuse.c
> + *
> + * Copyright (c) 2014 Rockchip Electronics Co. Ltd.
> + * Author: Jianqun Xu <jay.xu@xxxxxxxxxxxxxx>
> + *
> + * Tmis program is free software; you can redistribute it and/or modify it
> + * under the terms of version 2 of the GNU General Public License as
> + * published by the Free Software Foundation.
> + *
> + * Tmis program is distributed in the hope that it will be useful, but

type Tmis -> This


> WITHOUT + * ANY WARRANTY; without even the implied warranty of
> MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> General Public License for + * more details.
> + *
> + * You should have received a copy of the GNU General Public License along
> with + * tmis program; if not, write to the Free Software Foundation, Inc.,
> + * 51 Franklin Street, Fifth Floor, Boston, MA 02110, USA
> + *
> + * Tme full GNU General Public License is included in this distribution in
> the + * file called LICENSE.
> + */
> +
> +#include <linux/kernel.h>
> +#include <linux/module.h>
> +#include <linux/device.h>
> +#include <linux/delay.h>
> +#include <linux/platform_device.h>
> +#include <linux/of_address.h>
> +#include <linux/io.h>
> +
> +#include "efuse.h"
> +
> +#define EFUSE_BUF_SIZE (32)
> +#define EFUSE_BUF_LKG_CPU (23)
> +#define EFUSE_BUF_LKG_GPU (24)
> +#define EFUSE_BUF_LKG_LOG (25)

no braces needed for those numbers


> +
> +struct rk_efuse_info {
> + /* Platform device */
> + struct device *dev;
> +
> + /* Hardware resources */
> + void __iomem *regs;
> +
> + /* buffer to store registers' values */
> + unsigned int buf[EFUSE_BUF_SIZE];
> +};
> +
> +static void efuse_writel(struct rk_efuse_info *efuse,
> + unsigned int value,
> + unsigned int offset)
> +{
> + writel_relaxed(value, efuse->regs + offset);
> +}
> +
> +static unsigned int efuse_readl(struct rk_efuse_info *efuse,
> + unsigned int offset)
> +{
> + return readl_relaxed(efuse->regs + offset);
> +}

why these indirections for readl and writel? They don't seem to provide any
additional benefit over calling writel_relaxed/readl_relaxed directly below.


> +
> +static unsigned int rockchip_efuse_leakage(struct rk_efuse_info *efuse,
> + int channel)
> +{
> + switch (channel) {
> + case EFUSE_BUF_LKG_CPU:
> + case EFUSE_BUF_LKG_GPU:
> + case EFUSE_BUF_LKG_LOG:
> + return efuse->buf[channel];
> + default:
> + dev_err(efuse->dev, "unknown channel\n");
> + return -EINVAL;
> + }
> +
> + return 0;
> +}
> +
> +static void rockchip_efuse_info(struct rk_efuse_info *efuse)
> +{
> + dev_info(efuse->dev, "leakage (%d %d %d)\n",
> + rockchip_efuse_leakage(efuse, EFUSE_BUF_LKG_CPU),
> + rockchip_efuse_leakage(efuse, EFUSE_BUF_LKG_GPU),
> + rockchip_efuse_leakage(efuse, EFUSE_BUF_LKG_LOG));
> +}
> +
> +static int rockchip_efuse_init(struct rk_efuse_info *efuse)
> +{
> + int start = 0;
> + int ret = 0;
> +
> + efuse_writel(efuse, EFUSE_CSB, REG_EFUSE_CTRL);
> + efuse_writel(efuse, EFUSE_LOAD | EFUSE_PGENB, REG_EFUSE_CTRL);
> + udelay(2);
> +
> + for (start = 0; start <= EFUSE_BUF_SIZE; start++) {
> + efuse_writel(efuse, efuse_readl(efuse, REG_EFUSE_CTRL) &
> + (~(EFUSE_A_MASK << EFUSE_A_SHIFT)),
> + REG_EFUSE_CTRL);
> + efuse_writel(efuse, efuse_readl(efuse, REG_EFUSE_CTRL) |
> + ((start & EFUSE_A_MASK) << EFUSE_A_SHIFT),
> + REG_EFUSE_CTRL);
> + udelay(2);
> + efuse_writel(efuse, efuse_readl(efuse, REG_EFUSE_CTRL) |
> + EFUSE_STROBE, REG_EFUSE_CTRL);
> + udelay(2);
> +
> + efuse->buf[start] = efuse_readl(efuse, REG_EFUSE_DOUT);
> +
> + efuse_writel(efuse, efuse_readl(efuse, REG_EFUSE_CTRL) &
> + (~EFUSE_STROBE), REG_EFUSE_CTRL);
> + udelay(2);
> + }
> +
> + udelay(2);
> + efuse_writel(efuse, efuse_readl(efuse, REG_EFUSE_CTRL) |
> + EFUSE_CSB, REG_EFUSE_CTRL);
> + udelay(2);
> +
> + return ret;
> +}
> +
> +static int rockchip_efuse_probe(struct platform_device *pdev)
> +{
> + struct rk_efuse_info *efuse;
> + struct resource *mem;
> + int ret = 0;
> +
> + efuse = devm_kzalloc(&pdev->dev, sizeof(*efuse), GFP_KERNEL);
> + if (!efuse)
> + return -ENOMEM;
> +
> + mem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> + efuse->regs = devm_ioremap_resource(&pdev->dev, mem);
> + if (IS_ERR(efuse->regs))
> + return PTR_ERR(efuse->regs);
> +
> + platform_set_drvdata(pdev, efuse);
> + efuse->dev = &pdev->dev;
> +
> + ret = rockchip_efuse_init(efuse);
> + if (!ret)
> + rockchip_efuse_info(efuse);
> +
> + return ret;
> +}
> +
> +static const struct of_device_id rockchip_efuse_match[] = {
> + { .compatible = "rockchip,rk3288-efuse", },

what about other Rockchip SoCs? At least the rk3188 seems to contain an efuse
[though the TRM I have only directs to a RK3188 eFuse.pdf without describing
the component. So I don't know if it's the same type.


> + {},
> +};
> +
> +static struct platform_driver rockchip_efuse_driver = {
> + .probe = rockchip_efuse_probe,
> + .driver = {
> + .name = "rk3288-efuse",
> + .owner = THIS_MODULE,

.owner gets already set through module_platform_driver


> + .of_match_table = of_match_ptr(rockchip_efuse_match),
> + },
> +};
> +
> +module_platform_driver(rockchip_efuse_driver);
> +
> +MODULE_DESCRIPTION("Rockchip eFuse Driver");
> +MODULE_AUTHOR("Jianqun Xu <jay.xu@xxxxxxxxxxxxxx>");
> +MODULE_LICENSE("GPL v2");
> diff --git a/arch/arm/mach-rockchip/efuse.h b/arch/arm/mach-rockchip/efuse.h
> new file mode 100644
> index 0000000..3fdcf6d
> --- /dev/null
> +++ b/arch/arm/mach-rockchip/efuse.h

why does this need to be a separate header? The stuff below could very well
simply live inside the fuse.c


> @@ -0,0 +1,15 @@
> +#ifndef _ARCH_ROCKCHIP_EFUSE_H_
> +#define _ARCH_ROCKCHIP_EFUSE_H_
> +
> +/* Rockchip eFuse controller register */
> +#define EFUSE_A_SHIFT (6)
> +#define EFUSE_A_MASK (0x3FF)
> +#define EFUSE_PGENB (1 << 3)

please use BIT(3) instead of (1 << 3)
Same for the bits below.


> +#define EFUSE_LOAD (1 << 2)
> +#define EFUSE_STROBE (1 << 1)
> +#define EFUSE_CSB (1 << 0)
> +
> +#define REG_EFUSE_CTRL (0x0000)
> +#define REG_EFUSE_DOUT (0x0004)

no braces necessary for basic numerals


> +
> +#endif /* _ARCH_ROCKCHIP_EFUSE_H_ */

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/