Re: [PATCH net-next 3/3] tcp: use monotonic timestamps for PAWS

[Eric Dumazet]

On 06/18/2018 08:22 AM, Arnd Bergmann wrote:
> Using get_seconds() for timestamps is deprecated since it can lead
> to overflows on 32-bit systems. While the interface generally doesn't
> overflow until year 2106, the specific implementation of the TCP PAWS
> algorithm breaks in 2038 when the intermediate signed 32-bit timestamps
> overflow.
>
...

>
>  
>  static inline u32 tcp_cookie_time(void)
> @@ -1361,7 +1362,7 @@ static inline bool tcp_paws_check(const struct tcp_options_received *rx_opt,
>  {
>  	if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win)
>  		return true;
> -	if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS))
> +	if (unlikely(ktime_get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS))
>  		return true;
>  	/*
>  	 * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0,
> @@ -1391,7 +1392,7 @@ static inline bool tcp_paws_reject(const struct tcp_options_received *rx_opt,
>  
>  	   However, we can relax time bounds for RST segments to MSL.
>  	 */
> -	if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
> +	if (rst && ktime_get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
>  		return false;
>  	return true;



Please use the time_after32(), since ktime_get_seconds() is time64_t while ts_recent_stamp is int.

Same remark for tcp_twsk_unique()

Lets clean up this stuff, thanks !