On (03/04/09 01:52), Peter Memishian wrote:
> 
> While this will work for ipadm, I don't think this is architecturally
> sound.  For instance, consider the case where an application links against
> a library (e.g., libnwam) that in turn links against libipadm.  Now that
> application needs to have something in /etc/security/exec_attr to satisfy
> an implementation detail of libnwam (the fact that it uses libipadm).

It would need to have file_dac_write iff it was doing IP configuration.

> As Seb mentioned earlier, dlmgmtd provides this facility for libdladm.  It
> also synchronizes all changes (e.g., two applications using libdladm at
> the same time cannot clobber each other), and provides a central point for
> other facilities, such as posting sysevent notifications for new datalinks.

Requiring a daemon-per-library to broker with the library's datastore
is also not a very good design (that would be reinventing svc.configd per 
library). It may be that we are forced to have the file_dac_write constraint 
until all of the *adm applications move away from flat-file repository
to smf.

--Sowmini


Reply via email to