Meem, Thirumalai,
I'm looking at CR 6595874, which the two of you were looking at earlier-
some questions:
The root-cause of the problem (as I understand it) is that ilms don't
keep a ref on the ilm_ipif in the way that, for example, ire's do, so
ipif_is_quiescent will return TRUE even though there is an ilm pointing
at the ipif.
My question is: why do we have 3 refs in the ipif: ipif_refcnt,
ipif_ire_cnt, and ipif_saved_ire_cnt? There does not seem to
be an "ipif_ref{hold,rele}_notr" (i.e., no-trace) version- is
there some design rationale behind this? All of ipif_refcnt,
ipif_ire_cnt and ipif_saved_ire_cnt are protected by the
ill_lock, so why manage them as distinct variables in the ipif_t?
--Sowmini
[#] http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6595874
_______________________________________________
networking-discuss mailing list
[email protected]
