Can you add *which* kstats are being used? Individual kstats are
normally not documented, and some of them are more volatile than
others. If this case is going to use kstats, then I think we need to
document which ones are used in case someone decides to change them in
the future.
Thanks.
- Garrett
Jim Walker wrote:
> Garrett D'Amore wrote:
>> Mostly I like this, having reviewed the documentation on the website
>> as well.
>>
>> A few things I'd like to see in this case before I assert +1:
>>
>> 1) a list of the actual interfaces (routines) provided... you'll
>> need to provide man pages for them anyway, so this shouldn't be too
>> onerous ... along with commitment levels for each (probably all
>> Uncommitted)
>>
>> 2) a list of interfaces being *imported*, along with the stability
>> for each, for any non-public interfaces. (I suspect that it uses
>> several such interfaces, although I've not checked the code. Some of
>> the details may be challenging to collect.)
>>
>> 3) for Network interfaces, I'd like to know how it collects the
>> details -- does it use libdlpi? Does it use kstats? libdladm?
>> Will it show up vnics and other crossbow kinds of pseudo-interfaces?
>
> The project team has updated the proposal.txt file
> with the requested information.
>
> http://sac.sfbay/arc/PSARC/2009/583/proposal.txt
> http://arc.opensolaris.org/caselog/PSARC/2009/583/proposal.txt
> (may not be pushed yet)
>
> In addition, output from another statgrab run is provided that
> shows crossbow vnic information. kstats is used.
>
> http://sac.sfbay/arc/PSARC/2009/583/materials/statgrab.vnic.txt
>
> net.vnic0.collisions = 0
> net.vnic0.duplex = unknown
> net.vnic0.ierrors = 0
> net.vnic0.interface_name = vnic0
> net.vnic0.ipackets = 0
> net.vnic0.oerrors = 0
> net.vnic0.opackets = 31
> net.vnic0.rx = 0
> net.vnic0.speed = 0
> net.vnic0.systime = 1256701191
> net.vnic0.tx = 2152
> net.vnic0.up = true
>
> Cheers,
> Jim
