Re: Testing CAM wrapper for ata(4) controller drivers

On Sun, Dec 13, 2009 at 02:20:47AM +0200, Alexander Motin wrote:
> Juergen Lock wrote:
> > In article <4B20EDD2.5000102_at_FreeBSD.org> you write:
> >> Andriy Gapon wrote:
> >>> on 10/12/2009 01:07 Alexander Best said the following:
> >>>> ah. i see. thanks for the hint. sorry mav for blaming ATA_CAM. ;) would be
> >>>> nice if this would be fixed at some point. i believe the problem also applies
> >>>> to tape drives, usb memory card readers, etc.? so generally speaking: any
> >>>> devices which allow new media insertion, but don't disconnect/re-attach
> >>>> from/to CAM.
> >>> Yes, we need to get some notification that media is changed and then trigger geom
> >>> action.  Right now there is no notification from hardware in most cases and there
> >>> is no support for handling that in drivers, AFAIK.  Maybe ahci driver starts to
> >>> add support for that.  So either something needs to poll media for changes or a
> >>> user has to trigger some action explicitly.  No magic.
> >> Both ahci and siis drivers already have SATA Asynchronous Notifications
> >> support, that was especially made to do that. Now AN used to receive
> >> messages from PMP about fan-out ports physical events and working fine.
> >>
> >> What is needed: SATA ATAPI device with AN support (haven't checked if
> >> there are ones on the market), enable these messages, improve cd driver
> >> to make some useful activity (have no idea how) on such events.
> > 
> > Doesn't scsi have `unit attention' to notify media changes?  Or does
> > that not work as expected with atapi?
> 
> Haven't checked. Will try it tomorrow. But any way it require device
> polling. SATA AN same time is device-initiated real-time event.

Ah ok.

 Cheers,
	Juergen