UEFI: arndale does not store Boot Device config

Bug #1131141 reported by Ryan Harkin
20
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Arndale
Confirmed
Undecided
Rony Nandy
Linaro UEFI
Won't Fix
Undecided
Rony Nandy

Bug Description

Arndale boards do not store the Boot Device config, meaning that each time the board boots, it defaults to the config built into the image.

Ideally, the board should store the config in some persistent storage, eg, flash or SD card.

Fathi Boudra (fboudra)
Changed in linaro-uefi:
status: New → Confirmed
Revision history for this message
Rony Nandy (rony-nandy) wrote :

@ Ryan This feature was also not available in Origen.My question is does the VExpress boards support persistent storage of the configs? I had tried this for Origen but in the Bootmanager menu I could not find any options to save the configuration changed.

eg. in u-boot we have "save" command to do this.

So, my question is what is the front end interface(commands etc) in UEFI to save the configs? once we have it I can try to provide the back end support for Arndale.

Revision history for this message
Ryan Harkin (ryanharkin) wrote :

I don't know *how* vexpress does it, but it stores the config in NOR flash.

Whenever a new boot device is created, it is automatically saved without user intervention. I guess it's via a plugin protocol.

Revision history for this message
Shawn Rosti (shawn-rosti) wrote :

It might be worth checking to see if the Arndale board supports UEFI variable services.

Rony Nandy (rony-nandy)
Changed in arndale:
status: New → Confirmed
assignee: nobody → Rony Nandy (rony-nandy)
Changed in linaro-uefi:
assignee: nobody → Rony Nandy (rony-nandy)
Revision history for this message
Fathi Boudra (fboudra) wrote :

LAVA sort of depends on that feature in uefi. Is there any progress?

Revision history for this message
Rony Nandy (rony-nandy) wrote :

@Fathi
   The Arndale UEFI already had auto boot patch which can boot from any available partition with uImage so there is no need to make any manual configurations to make LAVA work.This is a good to have feature and is of lower priority as of now. This requires working on Variable services enablement for which I am going through the spec.

Revision history for this message
Fathi Boudra (fboudra) wrote :

Rony, adding Dave since he's working on the lava side to implement uefi support.
Dave, could you expand on your issue and does auto boot patch resolve it?

Revision history for this message
Fathi Boudra (fboudra) wrote :

As clarified with Rony, 2013.02 release doesn't have the auto boot patches as it was sent after the release.
The two patches should be in the upcoming 2013.03 release.

Revision history for this message
Rony Nandy (rony-nandy) wrote :
Revision history for this message
Ryan Harkin (ryanharkin) wrote :

The current patch in the tree will autoboot a linaro image from the SD card. The patch Rony refers to is a different approach to autobooting.

Both approaches are insufficient for LAVA.

LAVA needs to keep 3 boot configurations: 1) master 2) ubuntu test image 3) android test image

Each configuration boots from a specific partition. We need persistent config to be able to handle these 3 cases in LAVA.

Revision history for this message
Rony Nandy (rony-nandy) wrote :

@Ryan,
   Fathi informed me that UEFI is failing as the partitions are changing,If autoboot happens properly,it should at least boot.

Regarding 3 boot configurations,The workaound can be to have the configurations in dsc file itself while building. Is it feasible?

Revision history for this message
Ryan Harkin (ryanharkin) wrote : Re: [Bug 1131141] Re: UEFI: arndale does not store Boot Device config

Hey Rony,

On 14 March 2013 08:34, Rony Nandy <email address hidden> wrote:
> @Ryan,
> Fathi informed me that UEFI is failing as the partitions are changing,If autoboot happens properly,it should at least boot.

I'm under the impression that your autoboot patch will only boot from
the first FAT partition, whereas LAVA wants to boot a different FAT
partition depending on the kernel to be booted (master or test).

And my UUID zero patch won't fix the problem either.

>
> Regarding 3 boot configurations,The workaound can be to have the
> configurations in dsc file itself while building. Is it feasible?

I don't believe that's possible, there is only 1 configuration for the
default boot entry. The code could be expanded to add multiple
configurations, but that's time consuming work that needs to be
resourced.

>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1131141
>
> Title:
> UEFI: arndale does not store Boot Device config
>
> Status in Arndale Support:
> Confirmed
> Status in UEFI Support in Linaro:
> Confirmed
>
> Bug description:
> Arndale boards do not store the Boot Device config, meaning that each
> time the board boots, it defaults to the config built into the image.
>
> Ideally, the board should store the config in some persistent storage,
> eg, flash or SD card.
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/arndale/+bug/1131141/+subscriptions

Revision history for this message
Dave Pigott (dpigott) wrote :

If we don't have persistence of the UEFI config then we are going to have to create a new device class in the dispatcher. This may not be a bad thing, but it means it's not a 30 minute fix either.

Essentially, we treat UEFI like a sort of uBoot monitor. With uBoot we look for the appropriate uBoot prompt after issuing commands. With UEFI we say the boot prompt is "Start: " - when we see that we issue the UEFI boot config number (2 for Ubuntu, 3 for Adnroid) and that's it. Without persistent storage we will have to issue the set of options and strings to configure a new boot. Only problem is that as it stands, LAVA will time out (just for fun I tried it to prove my point) waiting for "Start: " after we issue the "a" for Boot Manager.

What I will need to do is create a new device type that will allow us to look for the appropriate messages, and then set up the config. It's not rocket science, but it's going to take time to implement.

The question is, what are the chances of getting persistent UEFI config in the near term?

Revision history for this message
Rony Nandy (rony-nandy) wrote :

@Dave I am not very close to the solution till now.

Revision history for this message
Dave Pigott (dpigott) wrote :

OK - so I'll look at extending the dispatcher

Revision history for this message
Leif Lindholm (leif-lindholm) wrote :

Ardnale no longer actively maintained.

Changed in linaro-uefi:
status: Confirmed → Won't Fix
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.