[PATCH] usb: musb: Fix signed integer overflow

[PATCH] usb: musb: Fix signed integer overflow

[Siddarth Gundu]

Shifting 1 << 31 on a 32-bit int causes signed integer overflow, which
leads to undefined behavior.

Cast 1 to u32 before performing the shift, eliminating the
undefined behavior.

Signed-off-by: Siddarth Gundu <siddarthsgml@gmail.com>
---
 drivers/usb/musb/cppi_dma.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/usb/musb/cppi_dma.h b/drivers/usb/musb/cppi_dma.h
index 3606be897cb2..33bf5f83f08d 100644
--- a/drivers/usb/musb/cppi_dma.h
+++ b/drivers/usb/musb/cppi_dma.h
@@ -38,7 +38,7 @@ struct cppi_rx_stateram {
 };
 
 /* hw_options bits in CPPI buffer descriptors */
-#define CPPI_SOP_SET	((u32)(1 << 31))
+#define CPPI_SOP_SET	((u32)1 << 31)
 #define CPPI_EOP_SET	((u32)(1 << 30))
 #define CPPI_OWN_SET	((u32)(1 << 29))	/* owned by cppi */
 #define CPPI_EOQ_MASK	((u32)(1 << 28))
@@ -48,7 +48,7 @@ struct cppi_rx_stateram {
 #define CPPI_RECV_PKTLEN_MASK 0xFFFF
 #define CPPI_BUFFER_LEN_MASK 0xFFFF
 
-#define CPPI_TEAR_READY ((u32)(1 << 31))
+#define CPPI_TEAR_READY ((u32)1 << 31)
 
 /* CPPI data structure definitions */
 
-- 
2.43.0

Re: [PATCH] usb: musb: Fix signed integer overflow

[Greg KH]

On Sun, May 11, 2025 at 02:42:17AM +0530, Siddarth Gundu wrote:
> Shifting 1 << 31 on a 32-bit int causes signed integer overflow, which
> leads to undefined behavior.
> 
> Cast 1 to u32 before performing the shift, eliminating the
> undefined behavior.
> 
> Signed-off-by: Siddarth Gundu <siddarthsgml@gmail.com>
> ---
>  drivers/usb/musb/cppi_dma.h | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/usb/musb/cppi_dma.h b/drivers/usb/musb/cppi_dma.h
> index 3606be897cb2..33bf5f83f08d 100644
> --- a/drivers/usb/musb/cppi_dma.h
> +++ b/drivers/usb/musb/cppi_dma.h
> @@ -38,7 +38,7 @@ struct cppi_rx_stateram {
>  };
>  
>  /* hw_options bits in CPPI buffer descriptors */
> -#define CPPI_SOP_SET	((u32)(1 << 31))
> +#define CPPI_SOP_SET	((u32)1 << 31)

Why not just use the proper BIT() macro instead?

And does this actually overflow and cause problems?  What is the result
of the original code where it is an issue in the driver?  Have you
tested it?  Is the assembly output actually different?

thanks,

greg k-h