On Wednesday, September 26, 2012, Aaron Lu wrote: > On Tue, Sep 25, 2012 at 11:45:34PM +0200, Rafael J. Wysocki wrote: > > On Tuesday, September 25, 2012, Aaron Lu wrote: > > > On 09/25/2012 10:23 PM, Oliver Neukum wrote: > > > > On Tuesday 25 September 2012 22:20:21 Aaron Lu wrote: > > > >> On Tue, Sep 25, 2012 at 01:47:52PM +0200, Rafael J. Wysocki wrote: > > > >>> On Tuesday, September 25, 2012, Aaron Lu wrote: > > > >>>> I'm thinking of enabling this GPE in sr_suspend once we decided that it > > > >>>> is ready to be powered off, so the time frame between sr_suspend and > > > >>>> when the power is actually removed in libata should be taken care of by > > > >>>> the GPE. If GPE fires, the notification function will request a runtime > > > >>>> resume of the device. Does this sound OK? > > > >>> > > > >>> Well, depending on the implementation. sr_suspend() should be rather > > > >>> generic, but the ACPI association (including the GPE thing) is specific to ATA. > > > >> > > > >> Sorry, but don't quite understand this. > > > >> > > > >> We have ACPI bindings for scsi devices, isn't that for us to use ACPI > > > >> when needed in scsi? > > > > > > > > We don't have ACPI bindings for generic SCSI devices. We have such > > > > bindings for SATA drives. You can put such things only in sr if it applies > > > > to all (maybe most) types of drives. > > > > > > OK. Then these scsi bindings for sata drives will be pretty much of > > > no use I think. > > > > > > > > > > >> BTW, if sr_suspend should be generic, that would suggest I shouldn't > > > >> write any ZPODD related code there, right? Any suggestion where these > > > >> code should go then? > > > > > > > > libata. Maybe some generic hooks can be called in sr_suspend(). > > > > > > Thanks for the suggestion. > > > The problem is, I need to know whether the door is closed and if there > > > is a medium inside. I've no way of getting such information in libata. > > > > How does sr get to know it in the libata case? > > By executing a test_unit_ready command. > > Libata does/should not have any routine to do this, it is one of the > transport of SCSI devices and it relies on SCSI driver to manage the > device(both disk and ODD). > > > > > > > PS: Are you sure sr_suspend() handles DVD-RAMs correctly? > > > > > > No. Is there a spec for it? > > > Considering there are many different drives sr handle, is it possible to > > > write a generic sr_suspend? > > > Maybe your suggestion of callback is the way to go. > > > What about this idea, if we find this is a ZPODD capable drive, we > > > enable runtime suspend for it and write a suspend callback according to > > > ZPODD spec. For other drives that does not have a suspend callback, we > > > do not enable runtime suspend. > > > > You can enable runtime PM for all kinds of drives, but make the suspend > > and resume callbacks only do something for ZPODD ones. This may allow their > > parents to use runtime PM (as Alan said earlier in this thread), even if the > > drives themseleves are not really physically suspended. > > Sounds good. > > > > > > Does this sound reasonable? > > > > First, we need to know when the drive is not in use. That information > > we can get from the sr's runtime PM and it looks like we need to notify > > libata about that somehow. I'm not sure what mechanism is the best for > > that at the moment. > > The current mechanism to notify libata is by rumtime suspend. When scsi > device is runtime suspended, its parent device will be suspended. And > ata_port is one of the ancestor devices of scsi device, and we will > remove its power in ata_port's runtime suspend callback. The problem, then, is that the ata_port's runtime suspend callback would have to know whether or not power can be removed from the drive. I'm going to post patches introducing a "no power off" flag for PM QoS, among other things, today or tomorrow. I suppose this flag might be used to tell the ata_port's suspend not to remove power from the attached device. > > Second, when the device is resumed by remote wakeup, we need to notify > > sr about that. A "resume" alone is not sufficient, though, because it may > > be necessary to open the tray. Perhaps in that case we can use the same > > mechanism by which user events are processed by libata and delivered to sr? > > Thanks for the suggestion. > I'm not aware of any user events processed by libata. Do you mean the > events_checking poll? Yes, basically. In the remote wakeup case libata might report the same status as in the "user pressed the eject button" situation happening normally with power on. Thanks, Rafael -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
