"Narendra Kumar S.S" writes: > On 7/26/07, Liane Praza <lianep at eng.sun.com> wrote: > > > > "Narendra Kumar S.S" writes: > > > Now, I have written a code, which calls smf_enable_instance to enable > > > filesystem/local service. > > > Immediately after returning from this call, the next instruction, tries > > to > > > access a file on one of the local filesystems. > > > So, are there chances that, the filesystem/local is not yet completely > > > enabled and because of that, the local filesystem is not yet mounted and > > > hence this instruction might fail!? > > > > It is completely enabled. However, enabled doesn't mean the service > > is online. It means that we're trying to make it be online, but its > > dependencies may not be satisfied, it may have failed, or it may just > > not be online yet. > > > > As I said: > > > > Currently, an interface which is synchronous to online/disabled is > > only > > > > formally available through svcadm [en|dis]able -s. I believe there's > > an > > > > RFE to make the synchronous semantics through libscf, but don't have > > the > > > > bugid to hand. It'd make a great OpenSolaris RFE, though. :) > > > > In the meantime, you could wait a little bit to see if smf_get_state() > > returns SCF_STATE_STRING_ONLINE before continuing. > > > OK. So, if I have to wait a 'little bit', is there any standard or > predefined value for this wait?
Nope, sorry -- I don't know (and the system doesn't either) a priori how long it could take for a service to start, especially filesystem services that mount an administrator-defined arbitrary number of filesystems. (And that doesn't take into account if its dependencies haven't had the chance to start yet.) That's why it's a bug there isn't a libscf function to do what svcadm enable -s does: to implement a better algorithm than an arbitrary wait. I wish that there was, but we just haven't had the chance yet to do the work. I was trying to give you a potential quick and dirty solution that could work for your situation. You could look at the code in svcadm.c if you want the real algorithm... http://cvs.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/cmd/svc/svcadm/ IIRC, most of the code is in synch.c. liane -- Liane Praza, Solaris Kernel Development liane.praza at sun.com - http://blogs.sun.com/lianep
