[brussels-dev] code review: several bug fixes

[sowmini.varadhan at sun.com]

On (09/27/07 15:17), Artem Kachitchkine wrote:
> 
>
>>   http://cr.opensolaris.org/~sowmini/arcfixes/
>
> mac.c:
>
>  940         cp = (char *)tmp->b_wptr;
>  941         *cp++ = '\0';
>  942         tmp->b_wptr = (uchar_t *)cp;
>
> Or just:
>
> 	*tmp->b_wptr++ = '\0';
>
> So, for my education, what's the background on non nul-terminated strings 
> in so much of networking code (and the need for precautions like strn*)? I 
> understand when a property value is untyped, i.e. just an array of bytes. 
> But stuff like names (link names, property names, etc) are strings by 
> definition, it should be easy to guarantee their nul-terminatedness. Is 
> this historical?

some of these strings gets passed down on ioctls, so it may 
not be properly null-terminated, hence the precaution. 

but in this particular case, the history is even more murky: the
ndd code was originally written by mentat, and if you wander into
nd_getset, you will find that they have a strange hidden convention
of double-null termination (which I was getting wrong). 
The only known user level program that parses this is /sbin/ndd itself.

See, for example
  http://www.opensolaris.org/jive/thread.jspa?messageID=127515&#127515

--Sowmini