Michael Hunter wrote: > On Thu, 05 Nov 2009 15:36:52 -0500 > "Anurag S. Maskey" <Anurag.Maskey at Sun.COM> wrote: > > >> Michael Hunter wrote: >> >>>> http://zhadum.east/export/ws/am223141/checkout-area/nwam1-fixes/webrev/ >>>> >>>> 12479 nwamd state machine can miss events >>>> http://defect.opensolaris.org/bz/show_bug.cgi?id=12479 >>>> >>>> As mentioned in bug comments, fix involves generating the state change >>>> event with LINK_UP events, but checking the state of the NCU to make >>>> sure that it is not ONLINE when the event is actually processed. >>>> >>>> >>> That seems like the right method. >>> >>> How did you test this? >>> >>> >> One of my test machines had a few "link is already up, ignoring" debug >> messages from earlier. The new binary generates the "NCU is already >> online, not going to up state" messages on a couple of restarts of >> nwam. Renee is also testing the binary on the machine that exhibited >> this bug. >> > > I am less worried about testing that the bug fixes what it was designed > for that for testing for unintended side effects. Have you don't any > testing which would stimulate this and related code in various ways? > My test machine every once in a while gets LINK_UP message when the link is already up. I restarted nwam so that this situation happens before and after the fix. The fix worked as expected. My machine never reached the state like Renee's where the link was thought to be down when in fact it was up. Renee confirms that the fix worked with her machine.
Anurag
