> > Tony Nguyen wrote: > >> This is the webrev for > >> > >> 6537863 <http://monaco.sfbay.sun.com/detail.jsp?cr=6537863> > >> set_inst_enabled should always make sure general/enabled exists > >> > >> The fix is to make sure SCF_PG_GENERAL pg exists in both > >> set_inst_enabled functions regardless of whether or not the enabling > >> is temporary.
I looked at the review here: http://cr.grommit.com/~tonyn/6537863/webrev/ * FYI, I was briefly worried about the changes being a bit weird from a auths perspective. That is, consumers who might have sufficient auths to modify the property but not to create the property group would get a SCF_ERROR_PERMISSION_DENIED from smf_[enable|disable]_instance(). I believe I've convinced myself that's at worst OK, and at best entirely appropriate, but wanted to at least raise the point for your consideration. Make sure you've thought through the auths and are similarly convinced you're doing the right thing too. svcadm.c: 582 I'd be more explicit in the comment. "An instance's configuration is incomplete if general/enabled doesn't exist. Create both the property group and property here if they don't exist.", or something. 593 Set explicitly to disabled, not to enable's value. 653 But, if we hit "again" it's because someone else deleted the property out from under us. (ECANCELED from get_bool_prop().) I think you need to leave the previous pg_get_or_add() call here. midlevel.c 545 You need to use SCF_PG_GENERAL_FLAGS, not pgflags here. liane -- Liane Praza, Solaris Kernel Development liane.praza at sun.com - http://blogs.sun.com/lianep
