Re: [PATCH] Use u64 to calculate the timeout value to avoid overflow

[Anton Vorontsov <cbouatmailru@gmail.com>]

On Mon, Nov 26, 2012 at 04:35:59PM +0800, Haijun Zhang wrote:
> As data timeout_ns use u64 to avoid overflow.
> So we use macro div_u64 to perform a division.
[...]
> --- a/drivers/mmc/host/atmel-mci.c
> +++ b/drivers/mmc/host/atmel-mci.c
> @@ -593,17 +593,16 @@ static void atmci_timeout_timer(unsigned long data)
>  	tasklet_schedule(&host->tasklet);
>  }
>  
> -static inline unsigned int atmci_ns_to_clocks(struct atmel_mci *host,
> -					unsigned int ns)
> +static inline unsigned int atmci_ns_to_clocks(struct atmel_mci *host, u64 ns)
>  {
>  	/*
>  	 * It is easier here to use us instead of ns for the timeout,
>  	 * it prevents from overflows during calculation.
>  	 */
> -	unsigned int us = DIV_ROUND_UP(ns, 1000);
> +	u64 us = DIV_ROUND_UP_ULL(ns, 1000);
>  
>  	/* Maximum clock frequency is host->bus_hz/2 */
> -	return us * (DIV_ROUND_UP(host->bus_hz, 2000000));
> +	return (u32)(us * (DIV_ROUND_UP_ULL(host->bus_hz, 2000000)));

Why do you need u32 cast here?..

Thanks,
Anton.