[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lwip-users] Timeout wraparound bug

From: Fotis Panagiotopoulos
Subject: Re: [lwip-users] Timeout wraparound bug
Date: Sat, 18 Feb 2017 13:55:41 +0200


To begin with, I do call sys_check_timeouts() very often. I also tried v2.0.1, where the problem vanishes. I believe it is fixed by the following addition:

        } else if (timeout->time > msecs) {
          /* If this is the case, 'timeouts_last_time' and 'now' differs too much.
             This can be due to sys_check_timeouts() not being called at the right
             times, but also when stopping in a breakpoint. Anyway, let's assume
             this is not wanted, so add the first timer's time instead of 'diff' */
          timeout->time = msecs + next_timeout->time;
        timeout->next = t->next;

which looks more like a work-around for me, rather an actual fix.

I tried the snippet Sergio suggested and it correctly printed OK. So no compiler issues.

The only thing that remains is to check the overflow of my sys_now() function. I assume that it wraps around at 2^32-1, but I will cross check it.

Fotis Panagiotopoulos

Mechatronics Engineer



Αμφιαράου 25,, Τ.Κ. 104 42, Αθήνα, Ελλάδα
Τηλ : +30 210 5907000 – ext. 28 | Fax: +30 210 5912711
e-mail: address@hidden | web: http://www.amco.gr


Please consider the environment before printing this e-mail.

On Fri, Feb 17, 2017 at 5:30 PM, Sergio R. Caprile <address@hidden> wrote:
Nope, the difference will be preserved by type promotion, unless the wrap around occurs more than once between calls or the counter counted more than half length counts (which also includes the former...).
Are you calling frequently enough sys_check_timeouts() ?
You should (read 'must').

Otherwise, you have a compiler problem, try this:
#include <stdio.h>
#include <stdint.h>

uint32_t now = 1;
uint32_t timeouts_last_time = 0xFFFFFFFF;

uint32_t diff = now - timeouts_last_time;

        if((now - timeouts_last_time) > 2)

$ ./td

lwip-users mailing list

reply via email to

[Prev in Thread] Current Thread [Next in Thread]