Patchwork PCI: mvebu: return NULL instead of ERR_PTR(ret)

login
register
mail settings
Submitter Jingoo Han
Date Nov. 27, 2013, 1:59 a.m.
Message ID <000101ceeb14$480fee80$d82fcb80$%han@samsung.com>
Download mbox | patch
Permalink /patch/294458/
State Not Applicable
Headers show

Comments

Jingoo Han - Nov. 27, 2013, 1:59 a.m.
On Wednesday, November 27, 2013 3:10 AM, Jason Gunthorpe wrote:
> On Tue, Nov 26, 2013 at 02:31:44PM +0900, Jingoo Han wrote:
> > Previously, I sent the patch in order to fix sparse warning as below:
> > How about this?
> >
> > static void __iomem *mvebu_pcie_map_registers(struct platform_device *pdev,
> > 			struct device_node *np, struct mvebu_pcie_port *port)
> > {
> > 	struct resource regs;
> > 	int ret = 0;
> >
> > 	ret = of_address_to_resource(np, 0, &regs);
> > 	if (ret)
> > -		return ERR_PTR(ret);
> > +		return (void __iomem *)ERR_PTR(ret);
> 
> You should probably ask the sparse folks for guidance 'git grep
> iomem.*ERR_PTR' returns nothing, so this isn't an established pattern.
> 
> It seems like sparse should know that ERR_PTR functions can work with
> any pointer no matter the type? IS_ERR_PTR will have the same problem
> with implicitly dropping the iomem tag.

+cc Christopher Li, sparse mailing-list, Joe Perches, Dan Carpenter,
     Axel Lin, Julia Lawall,

Hi All,

I have some questions about handling sparse warning.

Currently, the following sparse warning happens
at Marvell Armada PCIe driver.

drivers/pci/host/pci-mvebu.c:743:31: warning: incorrect type in return expression (different address spaces)
drivers/pci/host/pci-mvebu.c:743:31:    expected void [noderef] <asn:2>*
drivers/pci/host/pci-mvebu.c:743:31:    got void *

mvebu_pcie_map_registers() returns ERR_PTR(ret),
however ERR_PTR() returns (void *), not (void __iomem *).

./drivers/pci/host/pci-mvebu.c
static void __iomem *mvebu_pcie_map_registers(struct platform_device *pdev,
		struct device_node *np, struct mvebu_pcie_port *port)
{
	struct resource regs;
	int ret = 0;

	ret = of_address_to_resource(np, 0, &regs);
	if (ret)
		return ERR_PTR(ret);

	return devm_ioremap_resource(&pdev->dev, &regs);
}
./include/linux/err.h
static inline void * __must_check ERR_PTR(long error)
{
	return (void *) error;
}


Previously, I submitted the following patch, that adds
(void __iomem *) cast.

./drivers/pci/host/pci-mvebu.c
static void __iomem *mvebu_pcie_map_registers(struct platform_device *pdev,
		struct device_node *np, struct mvebu_pcie_port *port)
{
	.....

	ret = of_address_to_resource(np, 0, &regs);
	if (ret)
-		return ERR_PTR(ret);
+		return (void __iomem *)ERR_PTR(ret);

However, other engineers said that "(void __iomem *)ERR_PTR(ret)"
is not a general pattern. I cannot find the proper method to resolve
this sparse warning.

In this case, how can I resolve this sparse warning?

Then, how about the following?


Thank you for reading this. :-)

Best regards,
Jingoo Han

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Joe Perches - Nov. 27, 2013, 2:26 a.m.
On Wed, 2013-11-27 at 10:59 +0900, Jingoo Han wrote:
[]
> ./drivers/pci/host/pci-mvebu.c
> static void __iomem *mvebu_pcie_map_registers(struct platform_device *pdev,
> 		struct device_node *np, struct mvebu_pcie_port *port)
> {
> 	.....
> 
> 	ret = of_address_to_resource(np, 0, &regs);
> 	if (ret)
> -		return ERR_PTR(ret);
> +		return (void __iomem *)ERR_PTR(ret);
> 
> However, other engineers said that "(void __iomem *)ERR_PTR(ret)"
> is not a general pattern. I cannot find the proper method to resolve
> this sparse warning.
> 
> In this case, how can I resolve this sparse warning?

I think there's no problem using the cast.  It's not a
pattern because it's not been required before as function
returns have not previously been declared __iomem.

Or, perhaps the arm|hexagon specific #define below could
be made generic so it could be used.

arch/arm/include/asm/io.h:#define IOMEM(x)	((void __force __iomem *)(x))

	return IOMEM(ERR_PTR(ret));

There aren't any current uses of return IOMEM(foo) either
though so the direct cast is probably more appropriate.


--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Josh Triplett - Nov. 27, 2013, 2:35 a.m.
On Tue, Nov 26, 2013 at 06:26:36PM -0800, Joe Perches wrote:
> On Wed, 2013-11-27 at 10:59 +0900, Jingoo Han wrote:
> []
> > ./drivers/pci/host/pci-mvebu.c
> > static void __iomem *mvebu_pcie_map_registers(struct platform_device *pdev,
> > 		struct device_node *np, struct mvebu_pcie_port *port)
> > {
> > 	.....
> > 
> > 	ret = of_address_to_resource(np, 0, &regs);
> > 	if (ret)
> > -		return ERR_PTR(ret);
> > +		return (void __iomem *)ERR_PTR(ret);
> > 
> > However, other engineers said that "(void __iomem *)ERR_PTR(ret)"
> > is not a general pattern. I cannot find the proper method to resolve
> > this sparse warning.
> > 
> > In this case, how can I resolve this sparse warning?
> 
> I think there's no problem using the cast.  It's not a
> pattern because it's not been required before as function
> returns have not previously been declared __iomem.
> 
> Or, perhaps the arm|hexagon specific #define below could
> be made generic so it could be used.
> 
> arch/arm/include/asm/io.h:#define IOMEM(x)	((void __force __iomem *)(x))
> 
> 	return IOMEM(ERR_PTR(ret));
> 
> There aren't any current uses of return IOMEM(foo) either
> though so the direct cast is probably more appropriate.

I don't think that #define is appropriate for any non-constant value;
almost any instance of that __force cast applied to a variable would be
better written by fixing types more appropriately.  I'd suggest
updating that #define for both architectures to use BUILD_BUG_ON to
assert __builtin_constant_p on its argument.

- Josh Triplett
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Joe Perches - Nov. 27, 2013, 2:48 a.m.
On Tue, 2013-11-26 at 18:35 -0800, Josh Triplett wrote:
> On Tue, Nov 26, 2013 at 06:26:36PM -0800, Joe Perches wrote:
> > On Wed, 2013-11-27 at 10:59 +0900, Jingoo Han wrote:
> > []
> > > ./drivers/pci/host/pci-mvebu.c
> > > static void __iomem *mvebu_pcie_map_registers(struct platform_device *pdev,
> > > 		struct device_node *np, struct mvebu_pcie_port *port)
> > > {
> > > 	.....
> > > 
> > > 	ret = of_address_to_resource(np, 0, &regs);
> > > 	if (ret)
> > > -		return ERR_PTR(ret);
> > > +		return (void __iomem *)ERR_PTR(ret);
> > > 
> > > However, other engineers said that "(void __iomem *)ERR_PTR(ret)"
> > > is not a general pattern. I cannot find the proper method to resolve
> > > this sparse warning.
> > > 
> > > In this case, how can I resolve this sparse warning?
> > 
> > I think there's no problem using the cast.  It's not a
> > pattern because it's not been required before as function
> > returns have not previously been declared __iomem.
> > 
> > Or, perhaps the arm|hexagon specific #define below could
> > be made generic so it could be used.
> > 
> > arch/arm/include/asm/io.h:#define IOMEM(x)	((void __force __iomem *)(x))
> > 
> > 	return IOMEM(ERR_PTR(ret));
> > 
> > There aren't any current uses of return IOMEM(foo) either
> > though so the direct cast is probably more appropriate.
> 
> I don't think that #define is appropriate for any non-constant value;
> almost any instance of that __force cast applied to a variable would be
> better written by fixing types more appropriately.  I'd suggest
> updating that #define for both architectures to use BUILD_BUG_ON to
> assert __builtin_constant_p on its argument.

That's probably a good idea.

There are at least a couple things would likely need to be
fixed up if the #define were changed.

arch/arm/plat-omap/sram.c
drivers/clocksource/bcm_kona_timer.c

I've no issue with either a direct cast or your idea of
a new static inline like ERR_PTR_IO or IOMEM_ERR_PTR or
any other appropriate name.

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

--- a/drivers/pci/host/pci-mvebu.c
+++ b/drivers/pci/host/pci-mvebu.c
@@ -740,7 +740,7 @@  static void __iomem *mvebu_pcie_map_registers(struct platform_device *pdev,

	ret = of_address_to_resource(np, 0, &regs);
	if (ret)
-		return ERR_PTR(ret);
+		return NULL;

	return devm_ioremap_resource(&pdev->dev, &regs);
 }
@@ -939,7 +939,7 @@  static int mvebu_pcie_probe(struct platform_device *pdev)
		continue;

		port->base = mvebu_pcie_map_registers(pdev, child, port);
-		if (IS_ERR(port->base)) {
+		if (IS_ERR_OR_NULL(port->base)) {
			dev_err(&pdev->dev, "PCIe%d.%d: cannot map registers\n",
				port->port, port->lane);
			port->base = NULL;