lwip-devel
[Top][All Lists]
Advanced

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

RE: [lwip-devel] Small request regarding autoip.h


From: Bill Auerbach
Subject: RE: [lwip-devel] Small request regarding autoip.h
Date: Wed, 14 Apr 2010 15:39:26 -0400

> Simon Goldschmidt:
>
>I guess for AutoIP, 250ms are OK: the minimum delay used is 1 second. We
>only have to keep an eye on the jitter: Currently, when we want to delay
>1 second, we might end up with 900ms (which is OK, I guess). However,
>when increasing the timer to 250ms, we could end up with a 750ms delay
>only, so we might have to increase ttw by one.

I don't see this much jitter - or it's not apparent :)

autoip->ttw = (u16_t)(LWIP_AUTOIP_RAND(netif) % (PROBE_WAIT *
AUTOIP_TICKS_PER_SECOND));

Doesn't it mean now ttw is 0 to 3 (it was 0 to 9 with AUTOIP_TMR_INTERVAL
100).  I would think that changes from being 0 to 149 mS longer because of
the larger value.

I tested it with AUTOIP_TMR_INTERVAL set to 250 and the functionality is the
same.  I didn't specifically time things, but the speed of AutoIP providing
an address was the same as determined by the speed of the debug message I
output.  Because AutoIP takes so long, the change to the interval is likely
negligible in the end.  I also had LWIP_DHCP_AUTOIP_COOP_TRIES set to 2 -
otherwise by default it is long enough for a user to think our product is
faulty because it doesn't connect.  I'm of the opinion that the default of 9
for LWIP_DHCP_AUTOIP_COOP_TRIES is much too long.  DHCP is in the order of a
couple of seconds, not a couple of 10's of seconds.

>For IGMP, I'm not totally sure: The protocol includes a "Max Response
>Time" given in 1/10 seconds (which is why the igmp_timer is set to
>100ms, I guess). However, I don't suspect it to be smaller than 250ms,
>so increasing the timer should be OK when the code is adjusted
>accordingly.

I would test this if I used it or if I knew what effect to look for by
changing the timer.  It sounds like with 1/10th of seconds timing that it's
not a good candidate to change.

Did your timer additions do anything to handle the lwIP timer calls?  Would
it be a nice feature to have a single function that monitors elapsed time
(in whatever resolution the user will call this function) and have this
feature make the appropriate calls at the right intervals to lwIP functions?
I do this but the time base is hard coded.

Thanks Simon!
Bill






reply via email to

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