[PATCH] ddr: socfpga: Fix integer overflow

[PATCH] ddr: socfpga: Fix integer overflow

[Jit Loon Lim]

From: Tien Fong Chee <tien.fong.chee@intel.com>

dramaddrw data structure are defined as u32. Compiler assume 32-bit
arithmetic on expression and the shifting operation to be done on a
32-bit value. So, the expression would overflow if the shifting
is more than 32-bit.

Signed-off-by: Tien Fong Chee <tien.fong.chee@intel.com>
---
 drivers/ddr/altera/sdram_soc64.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/ddr/altera/sdram_soc64.c b/drivers/ddr/altera/sdram_soc64.c
index 9b1710c135..81ad02118b 100644
--- a/drivers/ddr/altera/sdram_soc64.c
+++ b/drivers/ddr/altera/sdram_soc64.c
@@ -246,7 +246,7 @@ phys_size_t sdram_calculate_size(struct altera_sdram_plat *plat)
 			 DRAMADDRW_CFG_ROW_ADDR_WIDTH(dramaddrw) +
 			 DRAMADDRW_CFG_COL_ADDR_WIDTH(dramaddrw));
 
-	size *= (2 << (hmc_ecc_readl(plat, DDRIOCTRL) &
+	size *= ((phys_size_t)2 << (hmc_ecc_readl(plat, DDRIOCTRL) &
 			DDR_HMC_DDRIOCTRL_IOSIZE_MSK));
 
 	return size;
-- 
2.26.2

Re: [PATCH] ddr: socfpga: Fix integer overflow

[Marek Vasut]

On 8/30/22 08:42, Jit Loon Lim wrote:
> From: Tien Fong Chee <tien.fong.chee@intel.com>
> 
> dramaddrw data structure are defined as u32. Compiler assume 32-bit
> arithmetic on expression and the shifting operation to be done on a
> 32-bit value. So, the expression would overflow if the shifting
> is more than 32-bit.
> 
> Signed-off-by: Tien Fong Chee <tien.fong.chee@intel.com>
> ---
>   drivers/ddr/altera/sdram_soc64.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/ddr/altera/sdram_soc64.c b/drivers/ddr/altera/sdram_soc64.c
> index 9b1710c135..81ad02118b 100644
> --- a/drivers/ddr/altera/sdram_soc64.c
> +++ b/drivers/ddr/altera/sdram_soc64.c
> @@ -246,7 +246,7 @@ phys_size_t sdram_calculate_size(struct altera_sdram_plat *plat)
>   			 DRAMADDRW_CFG_ROW_ADDR_WIDTH(dramaddrw) +
>   			 DRAMADDRW_CFG_COL_ADDR_WIDTH(dramaddrw));
>   
> -	size *= (2 << (hmc_ecc_readl(plat, DDRIOCTRL) &
> +	size *= ((phys_size_t)2 << (hmc_ecc_readl(plat, DDRIOCTRL) &

What about using 2UL or 2ULL here, would that work too ?