[next] PCI: mediatek-gen3: Add missing null pointer check

Commit Message

Colin King April 29, 2021, 11 a.m. UTC

From: Colin Ian King <[e-mail]>

The call to platform_get_resource_byname can potentially return null, so
add a null pointer check to avoid a null pointer dereference issue.

Addresses-Coverity: ("Dereference null return")
Fixes: 441903d9e8f0 ("PCI: mediatek-gen3: Add MediaTek Gen3 driver for MT8192")
Signed-off-by: Colin Ian King <[e-mail]>
---
 drivers/pci/controller/pcie-mediatek-gen3.c | 2 ++
 1 file changed, 2 insertions(+)

Comments

Jianjun Wang (王建军) April 30, 2021, 7:24 a.m. UTC | #1

Hi Colin,

On Thu, 2021-04-29 at 12:00 +0100, Colin King wrote:
> From: Colin Ian King <[e-mail]>
> 
> The call to platform_get_resource_byname can potentially return null, so
> add a null pointer check to avoid a null pointer dereference issue.
> 
> Addresses-Coverity: ("Dereference null return")
> Fixes: 441903d9e8f0 ("PCI: mediatek-gen3: Add MediaTek Gen3 driver for MT8192")
> Signed-off-by: Colin Ian King <[e-mail]>
> ---
>  drivers/pci/controller/pcie-mediatek-gen3.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/pci/controller/pcie-mediatek-gen3.c b/drivers/pci/controller/pcie-mediatek-gen3.c
> index 20165e4a75b2..3c5b97716d40 100644
> --- a/drivers/pci/controller/pcie-mediatek-gen3.c
> +++ b/drivers/pci/controller/pcie-mediatek-gen3.c
> @@ -721,6 +721,8 @@ static int mtk_pcie_parse_port(struct mtk_pcie_port *port)
>  	int ret;
>  
>  	regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "pcie-mac");
> +	if (!regs)
> +		return -EINVAL;

Thanks for your patch.

devm_ioremap_resource() will check and decode this null pointer
dereference error, so I don't think we need to check here.

Thanks.

>  	port->base = devm_ioremap_resource(dev, regs);
>  	if (IS_ERR(port->base)) {
>  		dev_err(dev, "failed to map register base\n");

Christophe JAILLET April 30, 2021, 7:47 a.m. UTC | #2

Le 29/04/2021 à 13:00, Colin King a écrit :
> From: Colin Ian King <[e-mail]>
> 
> The call to platform_get_resource_byname can potentially return null, so
> add a null pointer check to avoid a null pointer dereference issue.
> 
> Addresses-Coverity: ("Dereference null return")
> Fixes: 441903d9e8f0 ("PCI: mediatek-gen3: Add MediaTek Gen3 driver for MT8192")
> Signed-off-by: Colin Ian King <[e-mail]>
> ---
>   drivers/pci/controller/pcie-mediatek-gen3.c | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/pci/controller/pcie-mediatek-gen3.c b/drivers/pci/controller/pcie-mediatek-gen3.c
> index 20165e4a75b2..3c5b97716d40 100644
> --- a/drivers/pci/controller/pcie-mediatek-gen3.c
> +++ b/drivers/pci/controller/pcie-mediatek-gen3.c
> @@ -721,6 +721,8 @@ static int mtk_pcie_parse_port(struct mtk_pcie_port *port)
>   	int ret;
>   
>   	regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "pcie-mac");
> +	if (!regs)
> +		return -EINVAL;
>   	port->base = devm_ioremap_resource(dev, regs);
>   	if (IS_ERR(port->base)) {
>   		dev_err(dev, "failed to map register base\n");
> 

Nitpick:
    Using 'devm_platform_ioremap_resource_byname' is slightly less 
verbose and should please Coverity.


Also, which git repo are you using? On linux-next ([1)], your proposed 
patch is already part of "PCI: mediatek-gen3: Add MediaTek Gen3 driver 
for MT8192".

[1]: 
https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/log/drivers/pci/controller/pcie-mediatek-gen3.c

CJ

Lorenzo Pieralisi April 30, 2021, 8:40 a.m. UTC | #3

On Thu, Apr 29, 2021 at 12:00:40PM +0100, Colin King wrote:
> From: Colin Ian King <[e-mail]>
> 
> The call to platform_get_resource_byname can potentially return null, so
> add a null pointer check to avoid a null pointer dereference issue.
> 
> Addresses-Coverity: ("Dereference null return")
> Fixes: 441903d9e8f0 ("PCI: mediatek-gen3: Add MediaTek Gen3 driver for MT8192")
> Signed-off-by: Colin Ian King <[e-mail]>
> ---
>  drivers/pci/controller/pcie-mediatek-gen3.c | 2 ++
>  1 file changed, 2 insertions(+)

Squashed into the commit it is fixing, in my pci/mediatek branch.

Thanks,
Lorenzo

> diff --git a/drivers/pci/controller/pcie-mediatek-gen3.c b/drivers/pci/controller/pcie-mediatek-gen3.c
> index 20165e4a75b2..3c5b97716d40 100644
> --- a/drivers/pci/controller/pcie-mediatek-gen3.c
> +++ b/drivers/pci/controller/pcie-mediatek-gen3.c
> @@ -721,6 +721,8 @@ static int mtk_pcie_parse_port(struct mtk_pcie_port *port)
>  	int ret;
>  
>  	regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "pcie-mac");
> +	if (!regs)
> +		return -EINVAL;
>  	port->base = devm_ioremap_resource(dev, regs);
>  	if (IS_ERR(port->base)) {
>  		dev_err(dev, "failed to map register base\n");
> -- 
> 2.30.2
>

Bjorn Helgaas April 30, 2021, 4:34 p.m. UTC | #4

On Fri, Apr 30, 2021 at 09:47:06AM +0200, Christophe JAILLET wrote:
> Le 29/04/2021 à 13:00, Colin King a écrit :
> > From: Colin Ian King <[e-mail]>
> > 
> > The call to platform_get_resource_byname can potentially return null, so
> > add a null pointer check to avoid a null pointer dereference issue.
> > 
> > Addresses-Coverity: ("Dereference null return")
> > Fixes: 441903d9e8f0 ("PCI: mediatek-gen3: Add MediaTek Gen3 driver for MT8192")
> > Signed-off-by: Colin Ian King <[e-mail]>
> > ---
> >   drivers/pci/controller/pcie-mediatek-gen3.c | 2 ++
> >   1 file changed, 2 insertions(+)
> > 
> > diff --git a/drivers/pci/controller/pcie-mediatek-gen3.c b/drivers/pci/controller/pcie-mediatek-gen3.c
> > index 20165e4a75b2..3c5b97716d40 100644
> > --- a/drivers/pci/controller/pcie-mediatek-gen3.c
> > +++ b/drivers/pci/controller/pcie-mediatek-gen3.c
> > @@ -721,6 +721,8 @@ static int mtk_pcie_parse_port(struct mtk_pcie_port *port)
> >   	int ret;
> >   	regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "pcie-mac");
> > +	if (!regs)
> > +		return -EINVAL;
> >   	port->base = devm_ioremap_resource(dev, regs);
> >   	if (IS_ERR(port->base)) {
> >   		dev_err(dev, "failed to map register base\n");
> > 
> 
> Nitpick:
>    Using 'devm_platform_ioremap_resource_byname' is slightly less verbose
> and should please Coverity.

Not a nitpick at all.  Jianjun is correct that devm_ioremap_resource()
does check "regs" for NULL and it fails gracefully before trying to
dereference it, so the extra check shouldn't be needed.  And most
cases in drivers/pci/ look like this, without the extra check:

  res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "app");
  base = devm_ioremap_resource(dev, res);
  if (IS_ERR(base))
    ...

If devm_platform_ioremap_resource_byname() keeps Coverity happy, I
think that's what we should be doing across drivers/pci/.  Coverity
false positives are a hassle.

Seems like something for next cycle since we're in the middle of the
merge window.

> Also, which git repo are you using? On linux-next ([1)], your proposed patch
> is already part of "PCI: mediatek-gen3: Add MediaTek Gen3 driver for
> MT8192".
> 
> [1]: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/log/drivers/pci/controller/pcie-mediatek-gen3.c

I think this is because Lorenzo already squashed Colin's change in.

Bjorn

Dan Carpenter May 3, 2021, 7:03 a.m. UTC | #5

On Fri, Apr 30, 2021 at 11:34:50AM -0500, Bjorn Helgaas wrote:
> On Fri, Apr 30, 2021 at 09:47:06AM +0200, Christophe JAILLET wrote:
> > Le 29/04/2021 à 13:00, Colin King a écrit :
> > > From: Colin Ian King <[e-mail]>
> > > 
> > > The call to platform_get_resource_byname can potentially return null, so
> > > add a null pointer check to avoid a null pointer dereference issue.
> > > 
> > > Addresses-Coverity: ("Dereference null return")
> > > Fixes: 441903d9e8f0 ("PCI: mediatek-gen3: Add MediaTek Gen3 driver for MT8192")
> > > Signed-off-by: Colin Ian King <[e-mail]>
> > > ---
> > >   drivers/pci/controller/pcie-mediatek-gen3.c | 2 ++
> > >   1 file changed, 2 insertions(+)
> > > 
> > > diff --git a/drivers/pci/controller/pcie-mediatek-gen3.c b/drivers/pci/controller/pcie-mediatek-gen3.c
> > > index 20165e4a75b2..3c5b97716d40 100644
> > > --- a/drivers/pci/controller/pcie-mediatek-gen3.c
> > > +++ b/drivers/pci/controller/pcie-mediatek-gen3.c
> > > @@ -721,6 +721,8 @@ static int mtk_pcie_parse_port(struct mtk_pcie_port *port)
> > >   	int ret;
> > >   	regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "pcie-mac");
> > > +	if (!regs)
> > > +		return -EINVAL;
> > >   	port->base = devm_ioremap_resource(dev, regs);
> > >   	if (IS_ERR(port->base)) {
> > >   		dev_err(dev, "failed to map register base\n");
> > > 
> > 
> > Nitpick:
> >    Using 'devm_platform_ioremap_resource_byname' is slightly less verbose
> > and should please Coverity.
> 
> Not a nitpick at all.  Jianjun is correct that devm_ioremap_resource()
> does check "regs" for NULL and it fails gracefully before trying to
> dereference it, so the extra check shouldn't be needed.  And most
> cases in drivers/pci/ look like this, without the extra check:
> 
>   res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "app");
>   base = devm_ioremap_resource(dev, res);
>   if (IS_ERR(base))
>     ...
> 
> If devm_platform_ioremap_resource_byname() keeps Coverity happy, I
> think that's what we should be doing across drivers/pci/.  Coverity
> false positives are a hassle.

Smatch knows that devm_ioremap_resource() will return ERR_PTR(-EINVAL)
when we pass it a NULL.  ;)

regards,
dan carpenter

Patch

diff --git a/drivers/pci/controller/pcie-mediatek-gen3.c b/drivers/pci/controller/pcie-mediatek-gen3.c
index 20165e4a75b2..3c5b97716d40 100644
--- a/drivers/pci/controller/pcie-mediatek-gen3.c
+++ b/drivers/pci/controller/pcie-mediatek-gen3.c
@@ -721,6 +721,8 @@  static int mtk_pcie_parse_port(struct mtk_pcie_port *port)
 	int ret;
 
 	regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "pcie-mac");
+	if (!regs)
+		return -EINVAL;
 	port->base = devm_ioremap_resource(dev, regs);
 	if (IS_ERR(port->base)) {
 		dev_err(dev, "failed to map register base\n");