Thanks Roamer, As far as I can see, the only difference is the context of mac_link_update() call. In case of polling, xge calls mac_link_update() from timer context. Why is that could be a problem?
Regards. On Mon, 2007-11-05 at 16:05 -0800, Yunsong (Roamer) Lu wrote: > Link status of aggregation is detected through the link up/down > interrupt, so disabling link interrupt will cause the standby problem. > Simply, to report link status appropriately is the minimum requirement > to the driver except changing MAC address. > > Looks you're talking about "polling" on xge, but why did you need to > disable the link interrupt? From the polling mode implemented by > Crossbow, link interrupt is still necessary and only traffic interrupt, > in fact receive interrupt for now, could be disabled/enabled as required > by GLD. > > Am I missing anything from your problem? > > Thanks, > > Roamer > > Dmitry Yusupov wrote: > > Folks, > > > > what are the requirements for GLDv3 LLD for simple case aggregation - > > two links back to back? MAC address set/change functionality seems not > > enough... > > > > Noticed, that link notification is somewhat related. For xge case, if we > > do polling we are always in up/standby case like this: > > > > -bash-3.00# dladm show-aggr > > key: 1 (0x0001) policy: L4 address: 0:c:fc:0:29:46 (auto) > > device address speed duplex > > link state > > xge0 0:c:fc:0:29:46 10000 Mbps full unknown > > standby > > xge1 0:c:fc:0:37:10 10000 Mbps full unknown > > standby > > > > If we enable link interrupt, data link moves to the up/attached like > > this: > > > > -bash-3.00# dladm show-aggr > > key: 1 (0x0001) policy: L4 address: 0:c:fc:0:29:46 (auto) > > device address speed duplex > > link state > > xge0 0:c:fc:0:29:46 10000 Mbps full up > > attached > > xge1 0:c:fc:0:37:10 10000 Mbps full up > > attached > > > > Any hints? > > Thanks. > > _______________________________________________ > > networking-discuss mailing list > > [email protected] > _______________________________________________ networking-discuss mailing list [email protected]
