2024-01-10 22:24:01

by Lad, Prabhakar

[permalink] [raw]
Subject: [PATCH] dmaengine: usb-dmac: Avoid format-overflow warning

From: Lad Prabhakar <[email protected]>

gcc points out that the fix-byte buffer might be too small:
drivers/dma/sh/usb-dmac.c: In function 'usb_dmac_probe':
drivers/dma/sh/usb-dmac.c:720:34: warning: '%u' directive writing between 1 and 10 bytes into a region of size 3 [-Wformat-overflow=]
720 | sprintf(pdev_irqname, "ch%u", index);
| ^~
In function 'usb_dmac_chan_probe',
inlined from 'usb_dmac_probe' at drivers/dma/sh/usb-dmac.c:814:9:
drivers/dma/sh/usb-dmac.c:720:31: note: directive argument in the range [0, 4294967294]
720 | sprintf(pdev_irqname, "ch%u", index);
| ^~~~~~
drivers/dma/sh/usb-dmac.c:720:9: note: 'sprintf' output between 4 and 13 bytes into a destination of size 5
720 | sprintf(pdev_irqname, "ch%u", index);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Maximum number of channels for USB-DMAC as per the driver is 1-99 so use
u8 instead of unsigned int/int for DMAC channel indexing and make the
pdev_irqname string long enough to avoid the warning.

While at it use scnprintf() instead of sprintf() to make the code more
robust.

Signed-off-by: Lad Prabhakar <[email protected]>
---
drivers/dma/sh/usb-dmac.c | 10 +++++-----
1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/dma/sh/usb-dmac.c b/drivers/dma/sh/usb-dmac.c
index a9b4302f6050..f7cd0cad056c 100644
--- a/drivers/dma/sh/usb-dmac.c
+++ b/drivers/dma/sh/usb-dmac.c
@@ -706,10 +706,10 @@ static const struct dev_pm_ops usb_dmac_pm = {

static int usb_dmac_chan_probe(struct usb_dmac *dmac,
struct usb_dmac_chan *uchan,
- unsigned int index)
+ u8 index)
{
struct platform_device *pdev = to_platform_device(dmac->dev);
- char pdev_irqname[5];
+ char pdev_irqname[6];
char *irqname;
int ret;

@@ -717,7 +717,7 @@ static int usb_dmac_chan_probe(struct usb_dmac *dmac,
uchan->iomem = dmac->iomem + USB_DMAC_CHAN_OFFSET(index);

/* Request the channel interrupt. */
- sprintf(pdev_irqname, "ch%u", index);
+ scnprintf(pdev_irqname, sizeof(pdev_irqname), "ch%u", index);
uchan->irq = platform_get_irq_byname(pdev, pdev_irqname);
if (uchan->irq < 0)
return -ENODEV;
@@ -768,8 +768,8 @@ static int usb_dmac_probe(struct platform_device *pdev)
const enum dma_slave_buswidth widths = USB_DMAC_SLAVE_BUSWIDTH;
struct dma_device *engine;
struct usb_dmac *dmac;
- unsigned int i;
int ret;
+ u8 i;

dmac = devm_kzalloc(&pdev->dev, sizeof(*dmac), GFP_KERNEL);
if (!dmac)
@@ -869,7 +869,7 @@ static void usb_dmac_chan_remove(struct usb_dmac *dmac,
static void usb_dmac_remove(struct platform_device *pdev)
{
struct usb_dmac *dmac = platform_get_drvdata(pdev);
- int i;
+ u8 i;

for (i = 0; i < dmac->n_channels; ++i)
usb_dmac_chan_remove(dmac, &dmac->channels[i]);
--
2.34.1



2024-01-10 22:41:28

by Kees Cook

[permalink] [raw]
Subject: Re: [PATCH] dmaengine: usb-dmac: Avoid format-overflow warning

On Wed, Jan 10, 2024 at 10:22:10PM +0000, Prabhakar wrote:
> From: Lad Prabhakar <[email protected]>
>
> gcc points out that the fix-byte buffer might be too small:
> drivers/dma/sh/usb-dmac.c: In function 'usb_dmac_probe':
> drivers/dma/sh/usb-dmac.c:720:34: warning: '%u' directive writing between 1 and 10 bytes into a region of size 3 [-Wformat-overflow=]
> 720 | sprintf(pdev_irqname, "ch%u", index);
> | ^~
> In function 'usb_dmac_chan_probe',
> inlined from 'usb_dmac_probe' at drivers/dma/sh/usb-dmac.c:814:9:
> drivers/dma/sh/usb-dmac.c:720:31: note: directive argument in the range [0, 4294967294]
> 720 | sprintf(pdev_irqname, "ch%u", index);
> | ^~~~~~
> drivers/dma/sh/usb-dmac.c:720:9: note: 'sprintf' output between 4 and 13 bytes into a destination of size 5
> 720 | sprintf(pdev_irqname, "ch%u", index);
> | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
> Maximum number of channels for USB-DMAC as per the driver is 1-99 so use
> u8 instead of unsigned int/int for DMAC channel indexing and make the
> pdev_irqname string long enough to avoid the warning.
>
> While at it use scnprintf() instead of sprintf() to make the code more
> robust.
>
> Signed-off-by: Lad Prabhakar <[email protected]>

This looks like good fixes; thanks! I see n_channels is sanity checked
during the probe in usb_dmac_chan_probe(), so this looks good.

(Is there a reason not to also change n_channels to a u8?)

-Kees

> ---
> drivers/dma/sh/usb-dmac.c | 10 +++++-----
> 1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/dma/sh/usb-dmac.c b/drivers/dma/sh/usb-dmac.c
> index a9b4302f6050..f7cd0cad056c 100644
> --- a/drivers/dma/sh/usb-dmac.c
> +++ b/drivers/dma/sh/usb-dmac.c
> @@ -706,10 +706,10 @@ static const struct dev_pm_ops usb_dmac_pm = {
>
> static int usb_dmac_chan_probe(struct usb_dmac *dmac,
> struct usb_dmac_chan *uchan,
> - unsigned int index)
> + u8 index)
> {
> struct platform_device *pdev = to_platform_device(dmac->dev);
> - char pdev_irqname[5];
> + char pdev_irqname[6];
> char *irqname;
> int ret;
>
> @@ -717,7 +717,7 @@ static int usb_dmac_chan_probe(struct usb_dmac *dmac,
> uchan->iomem = dmac->iomem + USB_DMAC_CHAN_OFFSET(index);
>
> /* Request the channel interrupt. */
> - sprintf(pdev_irqname, "ch%u", index);
> + scnprintf(pdev_irqname, sizeof(pdev_irqname), "ch%u", index);
> uchan->irq = platform_get_irq_byname(pdev, pdev_irqname);
> if (uchan->irq < 0)
> return -ENODEV;
> @@ -768,8 +768,8 @@ static int usb_dmac_probe(struct platform_device *pdev)
> const enum dma_slave_buswidth widths = USB_DMAC_SLAVE_BUSWIDTH;
> struct dma_device *engine;
> struct usb_dmac *dmac;
> - unsigned int i;
> int ret;
> + u8 i;
>
> dmac = devm_kzalloc(&pdev->dev, sizeof(*dmac), GFP_KERNEL);
> if (!dmac)
> @@ -869,7 +869,7 @@ static void usb_dmac_chan_remove(struct usb_dmac *dmac,
> static void usb_dmac_remove(struct platform_device *pdev)
> {
> struct usb_dmac *dmac = platform_get_drvdata(pdev);
> - int i;
> + u8 i;
> for (i = 0; i < dmac->n_channels; ++i)
> usb_dmac_chan_remove(dmac, &dmac->channels[i]);
> --
> 2.34.1
>

--
Kees Cook

2024-01-10 22:47:14

by Lad, Prabhakar

[permalink] [raw]
Subject: Re: [PATCH] dmaengine: usb-dmac: Avoid format-overflow warning

Hi Kees,

Thank you for the review.

On Wed, Jan 10, 2024 at 10:41 PM Kees Cook <[email protected]> wrote:
>
> On Wed, Jan 10, 2024 at 10:22:10PM +0000, Prabhakar wrote:
> > From: Lad Prabhakar <[email protected]>
> >
> > gcc points out that the fix-byte buffer might be too small:
> > drivers/dma/sh/usb-dmac.c: In function 'usb_dmac_probe':
> > drivers/dma/sh/usb-dmac.c:720:34: warning: '%u' directive writing between 1 and 10 bytes into a region of size 3 [-Wformat-overflow=]
> > 720 | sprintf(pdev_irqname, "ch%u", index);
> > | ^~
> > In function 'usb_dmac_chan_probe',
> > inlined from 'usb_dmac_probe' at drivers/dma/sh/usb-dmac.c:814:9:
> > drivers/dma/sh/usb-dmac.c:720:31: note: directive argument in the range [0, 4294967294]
> > 720 | sprintf(pdev_irqname, "ch%u", index);
> > | ^~~~~~
> > drivers/dma/sh/usb-dmac.c:720:9: note: 'sprintf' output between 4 and 13 bytes into a destination of size 5
> > 720 | sprintf(pdev_irqname, "ch%u", index);
> > | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >
> > Maximum number of channels for USB-DMAC as per the driver is 1-99 so use
> > u8 instead of unsigned int/int for DMAC channel indexing and make the
> > pdev_irqname string long enough to avoid the warning.
> >
> > While at it use scnprintf() instead of sprintf() to make the code more
> > robust.
> >
> > Signed-off-by: Lad Prabhakar <[email protected]>
>
> This looks like good fixes; thanks! I see n_channels is sanity checked
> during the probe in usb_dmac_chan_probe(), so this looks good.
>
> (Is there a reason not to also change n_channels to a u8?)
>
Good point, I oversighted it by just looking at the loop indices. I
will send a v2 with that change.

Cheers,
Prabhakar

> -Kees
>
> > ---
> > drivers/dma/sh/usb-dmac.c | 10 +++++-----
> > 1 file changed, 5 insertions(+), 5 deletions(-)
> >
> > diff --git a/drivers/dma/sh/usb-dmac.c b/drivers/dma/sh/usb-dmac.c
> > index a9b4302f6050..f7cd0cad056c 100644
> > --- a/drivers/dma/sh/usb-dmac.c
> > +++ b/drivers/dma/sh/usb-dmac.c
> > @@ -706,10 +706,10 @@ static const struct dev_pm_ops usb_dmac_pm = {
> >
> > static int usb_dmac_chan_probe(struct usb_dmac *dmac,
> > struct usb_dmac_chan *uchan,
> > - unsigned int index)
> > + u8 index)
> > {
> > struct platform_device *pdev = to_platform_device(dmac->dev);
> > - char pdev_irqname[5];
> > + char pdev_irqname[6];
> > char *irqname;
> > int ret;
> >
> > @@ -717,7 +717,7 @@ static int usb_dmac_chan_probe(struct usb_dmac *dmac,
> > uchan->iomem = dmac->iomem + USB_DMAC_CHAN_OFFSET(index);
> >
> > /* Request the channel interrupt. */
> > - sprintf(pdev_irqname, "ch%u", index);
> > + scnprintf(pdev_irqname, sizeof(pdev_irqname), "ch%u", index);
> > uchan->irq = platform_get_irq_byname(pdev, pdev_irqname);
> > if (uchan->irq < 0)
> > return -ENODEV;
> > @@ -768,8 +768,8 @@ static int usb_dmac_probe(struct platform_device *pdev)
> > const enum dma_slave_buswidth widths = USB_DMAC_SLAVE_BUSWIDTH;
> > struct dma_device *engine;
> > struct usb_dmac *dmac;
> > - unsigned int i;
> > int ret;
> > + u8 i;
> >
> > dmac = devm_kzalloc(&pdev->dev, sizeof(*dmac), GFP_KERNEL);
> > if (!dmac)
> > @@ -869,7 +869,7 @@ static void usb_dmac_chan_remove(struct usb_dmac *dmac,
> > static void usb_dmac_remove(struct platform_device *pdev)
> > {
> > struct usb_dmac *dmac = platform_get_drvdata(pdev);
> > - int i;
> > + u8 i;
> > for (i = 0; i < dmac->n_channels; ++i)
> > usb_dmac_chan_remove(dmac, &dmac->channels[i]);
> > --
> > 2.34.1
> >
>
> --
> Kees Cook

2024-01-10 23:58:30

by Kees Cook

[permalink] [raw]
Subject: Re: [PATCH] dmaengine: usb-dmac: Avoid format-overflow warning

On Wed, Jan 10, 2024 at 10:46:02PM +0000, Lad, Prabhakar wrote:
> Hi Kees,
>
> Thank you for the review.
>
> On Wed, Jan 10, 2024 at 10:41 PM Kees Cook <[email protected]> wrote:
> >
> > On Wed, Jan 10, 2024 at 10:22:10PM +0000, Prabhakar wrote:
> > > From: Lad Prabhakar <[email protected]>
> > >
> > > gcc points out that the fix-byte buffer might be too small:
> > > drivers/dma/sh/usb-dmac.c: In function 'usb_dmac_probe':
> > > drivers/dma/sh/usb-dmac.c:720:34: warning: '%u' directive writing between 1 and 10 bytes into a region of size 3 [-Wformat-overflow=]
> > > 720 | sprintf(pdev_irqname, "ch%u", index);
> > > | ^~
> > > In function 'usb_dmac_chan_probe',
> > > inlined from 'usb_dmac_probe' at drivers/dma/sh/usb-dmac.c:814:9:
> > > drivers/dma/sh/usb-dmac.c:720:31: note: directive argument in the range [0, 4294967294]
> > > 720 | sprintf(pdev_irqname, "ch%u", index);
> > > | ^~~~~~
> > > drivers/dma/sh/usb-dmac.c:720:9: note: 'sprintf' output between 4 and 13 bytes into a destination of size 5
> > > 720 | sprintf(pdev_irqname, "ch%u", index);
> > > | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > >
> > > Maximum number of channels for USB-DMAC as per the driver is 1-99 so use
> > > u8 instead of unsigned int/int for DMAC channel indexing and make the
> > > pdev_irqname string long enough to avoid the warning.
> > >
> > > While at it use scnprintf() instead of sprintf() to make the code more
> > > robust.
> > >
> > > Signed-off-by: Lad Prabhakar <[email protected]>
> >
> > This looks like good fixes; thanks! I see n_channels is sanity checked
> > during the probe in usb_dmac_chan_probe(), so this looks good.
> >
> > (Is there a reason not to also change n_channels to a u8?)
> >
> Good point, I oversighted it by just looking at the loop indices. I
> will send a v2 with that change.

I think you'll need a bounce variable in usb_dmac_chan_probe() since it
looks like it's reading a 32-bit value from DT, but otherwise, it should
be okay.

-Kees

--
Kees Cook

2024-01-11 09:06:38

by Geert Uytterhoeven

[permalink] [raw]
Subject: Re: [PATCH] dmaengine: usb-dmac: Avoid format-overflow warning

Hi Prabhakar,

On Wed, Jan 10, 2024 at 11:23 PM Prabhakar <[email protected]> wrote:
> From: Lad Prabhakar <[email protected]>
>
> gcc points out that the fix-byte buffer might be too small:
> drivers/dma/sh/usb-dmac.c: In function 'usb_dmac_probe':
> drivers/dma/sh/usb-dmac.c:720:34: warning: '%u' directive writing between 1 and 10 bytes into a region of size 3 [-Wformat-overflow=]
> 720 | sprintf(pdev_irqname, "ch%u", index);
> | ^~
> In function 'usb_dmac_chan_probe',
> inlined from 'usb_dmac_probe' at drivers/dma/sh/usb-dmac.c:814:9:
> drivers/dma/sh/usb-dmac.c:720:31: note: directive argument in the range [0, 4294967294]
> 720 | sprintf(pdev_irqname, "ch%u", index);
> | ^~~~~~
> drivers/dma/sh/usb-dmac.c:720:9: note: 'sprintf' output between 4 and 13 bytes into a destination of size 5
> 720 | sprintf(pdev_irqname, "ch%u", index);
> | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
> Maximum number of channels for USB-DMAC as per the driver is 1-99 so use
> u8 instead of unsigned int/int for DMAC channel indexing and make the
> pdev_irqname string long enough to avoid the warning.
>
> While at it use scnprintf() instead of sprintf() to make the code more
> robust.
>
> Signed-off-by: Lad Prabhakar <[email protected]>

Reviewed-by: Geert Uytterhoeven <[email protected]>

One nit below.

> --- a/drivers/dma/sh/usb-dmac.c
> +++ b/drivers/dma/sh/usb-dmac.c

> @@ -768,8 +768,8 @@ static int usb_dmac_probe(struct platform_device *pdev)
> const enum dma_slave_buswidth widths = USB_DMAC_SLAVE_BUSWIDTH;
> struct dma_device *engine;
> struct usb_dmac *dmac;
> - unsigned int i;
> int ret;
> + u8 i;

Personally, I'm not much a fan of making loop counters smaller than
(unsigned) int. If you do go this way, there are more loops over all
channels still using int.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68korg

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

2024-01-11 11:42:32

by Lad, Prabhakar

[permalink] [raw]
Subject: Re: [PATCH] dmaengine: usb-dmac: Avoid format-overflow warning

Hi Geert,

Thank you for the review.

On Thu, Jan 11, 2024 at 9:05 AM Geert Uytterhoeven <[email protected]> wrote:
>
> Hi Prabhakar,
>
> On Wed, Jan 10, 2024 at 11:23 PM Prabhakar <[email protected]> wrote:
> > From: Lad Prabhakar <[email protected]>
> >
> > gcc points out that the fix-byte buffer might be too small:
> > drivers/dma/sh/usb-dmac.c: In function 'usb_dmac_probe':
> > drivers/dma/sh/usb-dmac.c:720:34: warning: '%u' directive writing between 1 and 10 bytes into a region of size 3 [-Wformat-overflow=]
> > 720 | sprintf(pdev_irqname, "ch%u", index);
> > | ^~
> > In function 'usb_dmac_chan_probe',
> > inlined from 'usb_dmac_probe' at drivers/dma/sh/usb-dmac.c:814:9:
> > drivers/dma/sh/usb-dmac.c:720:31: note: directive argument in the range [0, 4294967294]
> > 720 | sprintf(pdev_irqname, "ch%u", index);
> > | ^~~~~~
> > drivers/dma/sh/usb-dmac.c:720:9: note: 'sprintf' output between 4 and 13 bytes into a destination of size 5
> > 720 | sprintf(pdev_irqname, "ch%u", index);
> > | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >
> > Maximum number of channels for USB-DMAC as per the driver is 1-99 so use
> > u8 instead of unsigned int/int for DMAC channel indexing and make the
> > pdev_irqname string long enough to avoid the warning.
> >
> > While at it use scnprintf() instead of sprintf() to make the code more
> > robust.
> >
> > Signed-off-by: Lad Prabhakar <[email protected]>
>
> Reviewed-by: Geert Uytterhoeven <[email protected]>
>
> One nit below.
>
> > --- a/drivers/dma/sh/usb-dmac.c
> > +++ b/drivers/dma/sh/usb-dmac.c
>
> > @@ -768,8 +768,8 @@ static int usb_dmac_probe(struct platform_device *pdev)
> > const enum dma_slave_buswidth widths = USB_DMAC_SLAVE_BUSWIDTH;
> > struct dma_device *engine;
> > struct usb_dmac *dmac;
> > - unsigned int i;
> > int ret;
> > + u8 i;
>
> Personally, I'm not much a fan of making loop counters smaller than
> (unsigned) int. If you do go this way, there are more loops over all
> channels still using int.
>
Agreed. So shall I drop Kees suggestion and leave the patch as is?

Cheers,
Prabhakar

2024-01-19 12:51:24

by Vinod Koul

[permalink] [raw]
Subject: Re: [PATCH] dmaengine: usb-dmac: Avoid format-overflow warning


On Wed, 10 Jan 2024 22:22:10 +0000, Prabhakar wrote:
> gcc points out that the fix-byte buffer might be too small:
> drivers/dma/sh/usb-dmac.c: In function 'usb_dmac_probe':
> drivers/dma/sh/usb-dmac.c:720:34: warning: '%u' directive writing between 1 and 10 bytes into a region of size 3 [-Wformat-overflow=]
> 720 | sprintf(pdev_irqname, "ch%u", index);
> | ^~
> In function 'usb_dmac_chan_probe',
> inlined from 'usb_dmac_probe' at drivers/dma/sh/usb-dmac.c:814:9:
> drivers/dma/sh/usb-dmac.c:720:31: note: directive argument in the range [0, 4294967294]
> 720 | sprintf(pdev_irqname, "ch%u", index);
> | ^~~~~~
> drivers/dma/sh/usb-dmac.c:720:9: note: 'sprintf' output between 4 and 13 bytes into a destination of size 5
> 720 | sprintf(pdev_irqname, "ch%u", index);
> | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
> [...]

Applied, thanks!

[1/1] dmaengine: usb-dmac: Avoid format-overflow warning
commit: 62b68a88795942512936896b9fec1ee7d5fa9922

Best regards,
--
~Vinod