This is the mail archive of the
ecos-discuss@sources.redhat.com
mailing list for the eCos project.
Re: DHCP crash with Novell Server
On Tue, Jun 19, 2001 at 03:44:51PM -0500, Grant Edwards wrote:
> Here are the fixes I plan on trying:
>
> 1) in alarm_function() change the state variable _before_
> re-enabling the alarm. That will limit the nesting level
> to 3.
>
> With no other changes, this will cause the lease to expire
> (AFAICT).
[...]
> Perhaps the proper behavior is just to wait for the IP lease to
> expire and start from scratch (start with a DISCOVERY rather
> than a REQUEST), but it doesn't appear that this is what the
> eCos routines do when the lease expires. Rather, they shut
> down the interface and give up, right?
FYI, I tested it with just the alarm_function() fix, and it did
get the lease back, bit it didn't look quite kosher. When the
lease first expired it didn't request a new lease -- sometime
later (after another lease expiry time period, I think) it
started a successful negotiation.
It's more well-behaved with the default T1 and T2 values (as
per the patch).
--
Grant Edwards
grante@visi.com