On Tue, 2006-04-11 at 07:13 -0700, Sangeeta Misra wrote:
> Sebastien,
> 
> Could you provide input to Dan's question regarding the IPv6 code - 
> DM-9, DM-10?

Sure.

> 
> DM-9 General T5 Just a question. Why the hell do we allocate
> an mblk to store an nce? Shouldn't we use
> kmem_alloc or kmem_cache_alloc()?
> 
> RESP: Dont know
> NOTE: Seb provide input

There is no good reason as far as I can tell.  A kmem cache would be
better.  Maybe the original implementer thought that these would get
passed around to external resolvers, but that never actually happens.

> 
> usr/src/uts/common/inet/ip/ip_ire.c
> 
> ------- --------------- ------- 
> -----------------------------------------------
> No. Location Sev. Comment
> ------- --------------- ------- 
> -----------------------------------------------
> 
> DM-10 General T5 I may have answered my own DM-9 question.
> It has to do with passing-to-ARP STREAMS
> crud, doesn't it?
> 
> RESP: Dont think that is the reaosn we don't pass nce_t's to arp (ipv6
> is supposed to do neighbor discovery, not arp). Perhaps some v6 expert 
> can clarify this...
> NOTE: Seb provide input

External IPv6 resolvers don't use nce_t's, they use the crusty arp
STREAMS interfaces (along with the arp data structures).  I really don't
think there's a good reason for allocb() to be used at all, and the
nce_mp should just be blown away.

-Seb


_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to