Krishna Yenduri wrote:
> Folks,
>
> What is the documented/supported behavior for modunload(1M) when it is done
> on a driver module? Does it guarantee that detach(9E) is called for all
> attached device instances before _fini(9e) is called?
>
Close, but not quite. What happens is when _fini is called, it arranges
to call mod_remove().
mod_remove implicitly tries to detach all instances. If it fails to
detach any of them, then it returns an error and _fini() error out as
well. The DDI framework will not unload the driver in this case. (The
error returned is probably EBUSY, IIRC.)
I'm not sure if this amount of detail is documented or not, but huge
amounts of code rely on it, and if Sun ever changed this behavior it
would break pretty much every DR-safe driver in existence.
Just to be clear, the callpath (with some middle bits elided) looks like:
modunload(1M) --> via ioctl --> _fini() --> mod_remove(9F) --> (for
each instance) xxx_detach(9e)
-- Garrett
> The detach(9E) man page implies that this is the case -
>
> If detach() determines a particular instance of the device
> cannot be removed when requested because of some exceptional
> condition, detach() must return DDI_FAILURE, which prevents
> the particular device instance from being detached. This
> also prevents the driver from being unloaded.
>
>
> But, modunload(1M) man page is silent on this.
>
> Thanks,
> -Krishna
>
>
> This message posted from opensolaris.org
> _______________________________________________
> driver-discuss mailing list
> driver-discuss@opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/driver-discuss
>
--
Garrett D'Amore, Principal Software Engineer
Tadpole Computer / Computing Technologies Division,
General Dynamics C4 Systems
http://www.tadpolecomputer.com/
Phone: 951 325-2134 Fax: 951 325-2191
_______________________________________________
driver-discuss mailing list
driver-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/driver-discuss
