[Olsr-dev] OLSRd 0.6.0 with ETT available

Erik Tromp (spam-protected)
Fri Oct 1 08:45:25 CEST 2010


> 
> Van: Henning Rogge [mailto:(spam-protected)] 
> Verzonden: donderdag 30 september 2010 17:22
> Aan: (spam-protected)
> CC: Erik Tromp; 'Henning Rogge'
> Onderwerp: Re: [Olsr-dev] OLSRd 0.6.0 with ETT available
> 
> 1.) The OLSR RFC says "every OLSR message have to be 4 byte aligned".
> 
> 2.) If you have unaligned fields within your message, you might get
trouble.
> 
> 3.) If you have unaligned LENGTH all messages after you in the same packet
> will have trouble. The more recent OLSR code will just drop all messages
with
> an odd length.
> 
> 4.) ARM_NO_WARN_ALIGN is mostly for byte-arrays... or for cases where we
know
> more than the compiler.
> 
> Henning Rogge


Ok, so no ARM/MIPS cpu crash :-)

B.t.w., are you sure about point 1.) ? I cannot seem to find your quote
("every OLSR message have to be 4 byte aligned") anywhere in RFC 3626 .
In fact I cannot even find the any form of the word "align" in the RFC ...

Point 2.): what trouble? Current RFC-compliant packets already have
unaligned
fields, see e.g. Htime and Willingness in par. 6.1 ("HELLO Message Format").

About point 3.): I removed this check in the OLSR-LC code.

Anyway, never mind: I will update the code. 32-bit alignment seems to
be ... better. But then again, one could reason that 64-bits alignment
would be 'even more better' for the future.

Best Regards,
Erik







More information about the Olsr-dev mailing list