Message-ID: <20140107140502.GE26588@tbergstrom-lnx.Nvidia.com>
Date:	Tue, 7 Jan 2014 16:05:02 +0200
From:	Peter De Schrijver <pdeschrijver@...dia.com>
To:	Stephen Warren <swarren@...dotorg.org>
CC:	"linux-arm-kernel@...ts.infradead.org" 
	<linux-arm-kernel@...ts.infradead.org>,
	"linux-tegra@...r.kernel.org" <linux-tegra@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	Rob Landley <rob@...dley.net>,
	Thierry Reding <thierry.reding@...il.com>,
	Grant Likely <grant.likely@...aro.org>,
	"Rob Herring" <robh+dt@...nel.org>,
	Danny Huang <dahuang@...dia.com>,
	"linux-doc@...r.kernel.org" <linux-doc@...r.kernel.org>,
	"devicetree@...r.kernel.org" <devicetree@...r.kernel.org>
Subject: Re: [PATCH v2 2/6] misc: fuse: Add efuse driver for Tegra

On Mon, Jan 06, 2014 at 09:32:24PM +0100, Stephen Warren wrote:
> On 12/24/2013 06:32 AM, Peter De Schrijver wrote:
> > Implement fuse driver for Tegra20, Tegra30, Tegra114 and Tegra124.
> 
> > diff --git a/drivers/misc/fuse/Kconfig b/drivers/misc/fuse/Kconfig
> 
> > +config FUSE_TEGRA
> > +	tristate "Tegra fuse supprt"
> > +	depends on ARCH_TEGRA && SYSFS
> 
> Since (I think) the Tegra-specific APIs this uses are stubbed if they
> can't build, perhaps this should depend on "|| COMPILE_TEST" too?
> 
> > +	help
> > +	  This drivers provides read-only to the e-fuses in Tegra chips.
> > +	  Parsing of the data is left to userspace.
> > +
> > +	  This driver can also be built as a module. If so, the module
> > +	  will be called tegra_efuse.
> > +endmenu
> 
> I'd expect a blank line before "endmenu" since there's one at the start
> of the menu contents.
> 
> > diff --git a/drivers/misc/fuse/tegra/Makefile b/drivers/misc/fuse/tegra/Makefile
> 
> > +obj-y					+= fuse-tegra.o
> > +obj-y					+= fuse-tegra30.o
> > +obj-$(CONFIG_ARCH_TEGRA_2x_SOC)		+= fuse-tegra20.o
> > +obj-$(CONFIG_ARCH_TEGRA_2x_SOC)		+= tegra20_speedo.o
> > +obj-$(CONFIG_ARCH_TEGRA_3x_SOC)		+= tegra30_speedo.o
> > +obj-$(CONFIG_ARCH_TEGRA_114_SOC)	+= tegra114_speedo.o
> > +obj-$(CONFIG_ARCH_TEGRA_124_SOC)	+= tegra124_speedo.o
> 
> > diff --git a/drivers/misc/fuse/tegra/fuse-tegra20.c b/drivers/misc/fuse/tegra/fuse-tegra20.c
> 
> > +static int fuse_size;
> 
> I don't think that's used.
> 
> > +static u32 tegra20_fuse_readl(const unsigned int offset)
> ...
> > +	ret = tegra_apb_readl_using_dma(fuse_phys + FUSE_BEGIN + offset, &val);
> 
> Shouldn't this use the generic tegra_apb_readl(), so that it works
> irrespective of whether the Tegra20 APB DMA driver is available?
> 

tegra_apb_readl() doesn't work reliably on Tegra20 for reading the fuses.
So if the Tegra20 APB DMA, this driver should also be unavailable.

> > +static const struct of_device_id tegra20_fuse_of_match[] = {
> > +	{ .compatible = "nvidia,tegra20-efuse" },
> > +}
> > +
> > +MODULE_DEVICE_TABLE(of, tegra20_fuse_of_match);
> 
> You'd typically omit that blank line.
> 
> > +static int tegra_fuse_probe(struct platform_device *pdev)
> 
> > +	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > +	fuse_phys = res->start;
> 
> Don't you need to error-check res here?
> 
> > +	fuse_randomness();
> 
> If this is a driver, and particularly if this could be in a module, is
> there any guarantee at all that fuse_randomness() gets called early
> enough to be useful?>
> 

For a module this might be true yes... Should we disallow this making a
module?

> > +	platform_set_drvdata(pdev, NULL);
> 
> That seems pointless; if the drvdata isn't used, there's no need to set
> it to anything in particular at all.
> 
> > +	if (tegra_fuse_sysfs(&pdev->dev, FUSE_SIZE, tegra20_fuse_readl,
> > +			     &sku_info))
> 
> Here (and also for fuse_randomness()), there's no verb in the function
> name. Perhaps use tegra_fuse_create_sysfs() and fuse_add_randomness().
> It wouldn't hurt to be consistent and use a tegra20_ prefix on all the
> function names too.
> 
> > diff --git a/drivers/misc/fuse/tegra/fuse-tegra30.c b/drivers/misc/fuse/tegra/fuse-tegra30.c
> 
> > +u32 tegra30_fuse_readl(const unsigned int offset)
> 
> > +	val = readl_relaxed(fuse_base + FUSE_BEGIN + offset);
> 
> If you aren't going to call tegra_apb_readl() here, I wonder if you
> shouldn't rename tegra_apb_readl() as tegra20_apb_readl() to make it
> obvious that the workaround isn't needed on all chips?
> 
> > +	clk_disable_unprepare(fuse_clk);
> 
> Doesn't the use of readl_**relaxed**() above mean that the
> clk_disable_unprepare() could turn off the clock before the fuse read
> had completed, and hence hang the system?
> 

Our clk_disable() has a barrier which should cover for this.

> > +static int tegra_fuse_probe(struct platform_device *pdev)
> 
> > +	fuse_randomness();
> 
> Here, the same function name is used as in fuse-tegra20.c, which might
> make debugging a bit more annoying.
> 
> > +MODULE_LICENSE("GPLv2");
> 
> s/GPLv2/GPL v2/ Perhaps the same in other files?
> 
> > diff --git a/drivers/misc/fuse/tegra/fuse.h b/drivers/misc/fuse/tegra/fuse.h
> 
> > +#ifdef CONFIG_ARCH_TEGRA_2x_SOC
> > +void tegra20_init_speedo_data(struct tegra_sku_info *sku_info,
> > +			      struct device *dev);
> > +bool tegra20_spare_fuse(int bit);
> > +#else
> > +static inline void tegra20_init_speedo_data(struct tegra_sku_info *sku_info,
> > +					    struct device *dev) {}
> > +static inline bool tegra20_spare_fuse(int bit) {}
> > +#endif
> 
> I suppose it doesn't hurt, but the Tegra20 functions don't need stubs
> since they're only called from files that are only compiled for Tegra20.
> But, I suppose it's fine to be consistent within this file and provide
> stubs anyway.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists