[U-Boot] [PATCH v2 2/3] efi_loader: enumerate disk devices every time

[AKASHI Takahiro]

On Tue, Jan 22, 2019 at 10:04:55PM +0100, Heinrich Schuchardt wrote:
> On 1/22/19 8:39 PM, Simon Glass wrote:
> > Hi Alex,
> > 
> > On Tue, 22 Jan 2019 at 22:08, Alexander Graf <agraf at suse.de> wrote:
> >>
> >>
> >>
> >> On 22.01.19 09:29, AKASHI Takahiro wrote:
> >>> Alex, Simon,
> >>>
> >>> Apologies for my slow response on this matter,
> >>>
> >>> On Fri, Jan 11, 2019 at 08:57:05AM +0100, Alexander Graf wrote:
> >>>>
> >>>>
> >>>> On 11.01.19 05:29, AKASHI Takahiro wrote:
> >>>>> Alex, Heinrich and Simon,
> >>>>>
> >>>>> Thank you for your comments, they are all valuable but also make me
> >>>>> confused as different people have different requirements :)
> >>>>> I'm not sure that all of us share the same *ultimate* goal here.
> >>>>
> >>>> The shared ultimate goal is to "merge" (as Simon put it) dm and efi objects.
> >>>
> >>> I don't still understand what "merge" means very well.
> >>
> >> It basically means that "struct efi_object" moves into "struct udevice".
> >> Every udevice instance of type UCLASS_BLK would expose the block and
> >> device_path protocols.
> >>
> >> This will be a slightly bigger rework, but eventually allows us to
> >> basically get rid of efi_init_obj_list() I think.
> > 
> > I envisaged something like:
> > 
> > - EFI objects have their own UCLASS_EFI uclass
> > - DM uclasses which support EFI would create a child EFI device (e.g.
> > a UCLASS_EFI child of each UCLASS_BLK device)
> > - EFI-uclass devices would thus be bound as needed
> > - Probing an EFI device would causes its parents to be probed
> > - We can use all the existing DM hooks (probe, remove, parent/child
> > data, operations), to implement EFI things
> > 
> > I'm assuming that a small percentage of devices would have EFI
> > children, so that this is more efficient than trying to merge the data
> > structures. It also allows EFI to maintain some separate from the core
> > DM code.
> 
> Dear Simon,
> 
> thanks to your suggestions.
> 
> I am not yet convinced that an UCLASS_EFI child device will be helpful.
> It is not an EFI object.
> 
> A DM uclass is the equivalent to an EFI driver, i.e. a handle with the
> EFI_DRIVER_BINDING_PROTOCOL installed on it [1]. So the natural thing
> for a uclass supporting EFI would be to provide such a handle.
> 
> For the actual devices we also need handles.
> 
> In the EFI world partitions are devices. How does this fit into your
> picture?

This is one of my concerns, too.
The only solution, I can image, in the *ultimate* goal is
that all the partitions of block devices be converted to
*real* block devices at probe() or any other enumerating processes,
say, "mmc rescan" or "usb start."

This, however, requires an extensive change across all the file system types.

-Takahiro Akashi


> [1] https://lists.denx.de/pipermail/u-boot/2019-January/354359.html
> [RFC] Device model for block devices - integration with EFI subsystem
> 
> Best regards
> 
> Heinrich
> 
> > 
> >>
> >>>
> >>>> But we have this annoying interim state where we would lose a few boards
> >>>> because they haven't been converted to DM. That's what keeps us from it.
> >>>>
> >>>> I think what this discussion boils down to is that someone needs to
> >>>> start prototyping the DM/EFI integration. Start off with a simple
> >>>> subsystem, like BLK.
> >>>
> >>> Even in the current implementation,
> >>> * UEFI disk is implemented using UCLASS_BLK devices
> >>>   (We can ignore !CONFIG_BLK case now as we have agreed.)
> >>> * UEFI-specific block device can be seen as UCLASS_BLK/IF_TYPE_EFI
> >>>
> >>> So how essentially different is the *ultimate* goal from the current form
> >>> regarding block devices?
> >>
> >> The ultimate goal is that efi_disk_register() and efi_obj_list disappear.
> >>
> >> Functionality wise we should be 100% identical to today, so all test
> >> cases would still apply the same way as they do now. This is purely
> >> internal rework, nothing visible to UEFI payloads.
> > 
> > Yes.
> > 
> >>
> >>> In order to identify UEFI disks with u-boot devices transparently, we will
> >>> have to have some sort of *hook* (or hotplug in Alex's language?), either
> >>> in create_block_devices or bind/probe?  I don't know, but Simon seems
> >>> to be in denial about this idea.
> >>
> >> Well, if a udevice *is* an efi device, we no longer need hooks. The
> >> object list would simply change.
> >>
> >> We may still need to have event notifications at that stage, but that's
> >> a different story.
> > 
> > Yes, it's something that I think will need to be added to DM. I
> > haven't got to this as I have not run into an important use case yet.
> > Maybe something like:
> > 
> > Controlled by CONFIG_EVENT
> > 
> > - int dev_ev_register(struct udevice *dev, enum event_t type,
> > event_handler_func_t handler, void *userdata)
> > 
> > which calls handler(struct udevice *dev, void *userdata) when an event is fired
> > 
> > - int dev_ev_unregister() to unregister
> > 
> > - int dev_ev_send(struct udevice *dev, enum struct event_info *info)
> > 
> > which sends events to registered listeners.
> > 
> > struct event_info {
> >   enum event_t type;
> >   union {
> >      struct ev_data_probed probed;
> >      struct ev_data_removed removed;
> >      ...
> >  } d;
> > };
> > 
> >>
> >> As transitioning period, we could probably implement 2 efi object roots:
> >> efi_obj_list as well as the udevice based one. Every piece of code that
> >> iterates through devices then just iterates over both. That way we
> >> should be able to slowly move devices from the old object model to the
> >> new one.
> > 
> > Will the uclass I mentioned above you can iterate through UCLASS_EFI
> > and thus you have a list of EFI devices.
> > 
> > [...]
> > 
> > Regards,
> > Simon
> > 
>