5.3-RELEASE TODO
Andre Oppermann
andre at freebsd.org
Fri Sep 17 05:33:24 PDT 2004
Scott Long wrote:
> |-------------------+---------------+--------------+---------------------|
> | | | | As part of the |
> | | | | MPSAFE network |
> | | | | stack work, |
> | | | | delivery of routing |
> | | | | socket messages was |
> | | | | moved to queued |
> | | | | dispatch via netisr |
> | | | | rather than direct |
> | | | | dispatch from the |
> | | | | routing code. |
> | Increased and | | | However, the risks |
> | configurable | | Robert | of lost routing |
> | netisr queue max | Needs testing | Watson | messages for |
> | depth for routing | | | routing daemons are |
> | sockets | | | high; respond by |
> | | | | increasing the max |
> | | | | depth beyond a |
> | | | | default interface |
> | | | | max depth of 50 to |
> | | | | 128, and allow it |
> | | | | to be |
> | | | | user-configured. |
> | | | | This change is now |
> | | | | present in CVS HEAD |
> | | | | and RELENG_5. |
> |-------------------+---------------+--------------+---------------------|
You can kill this one. The route(4) man page documents since ages that
under memory shortage conditions route messages can get lost.
--
Andre
More information about the freebsd-current
mailing list