Different kernel version is booting than the one submitted in the hwpack

[Naresh Kamboju]

Different kernel version is booting than the one submitted in the hwpack.

CI submitted LAVA job #128425 [1] to test image from build 66 [2]. But the LAVA is trying to boot a image which is not generated from Kernel CI.

If it would have been a kernel CI test image the kernel version should look like.
Linux version 3.10.40-linaro-vexpress64 (buildslave@x86-64-07)

But I see on the job is
Linux version 3.14.0-rc6-0.1-Juno-00037-g285a24a (dliviu@e106497-lin)

[1] https://validation.linaro.org/scheduler/job/128425/log_file#L_27_157
[2] https://ci.linaro.org/jenkins/job/linux-linaro-stable-lsk-pre-merge/hwpack=vexpress64,label=build/66/

[Milosz Wasilewski (mwasilew)]

After the short investigation it seems there are 2 entries in UEFI menu that were not there before:
https://validation.linaro.org/scheduler/job/128425/log_file#L_27_125

So LAVA boots from the NOR flash instead of the TESTBOOT partition:

 [1] Firmware Volume (4068 MB)
 [2] Firmware Volume (4068 MB)
 [3] NOR Flash (63 MB)
 [4] boot (99 MB)
 [5] TESTBOOT (99 MB)
 [6] SDCARD (2043 MB)
 [7] VenHw(E7223039-5836-41E1-B542-D7EC736C5E59)
 [8] VenHw(02118005-9DA7-443A-92D5-781F022AEDBB)
 [9] PXE on MAC Address: 00:02:F7:00:57:DF
 [10] TFTP on MAC Address: 00:02:F7:00:57:DF
 Select the Boot Device: 3
 File path of the EFI Application or the kernel: Image
 Has FDT support? [y/n] y
 Add an initrd: [y/n] n
 Arguments to pass to the binary: console=ttyAMA0,115200 earlyprintk=pl011,0x7ff80000 root=/dev/sda5 rootwait verbose debug
 Description for this new Entry: LAVA OE Test Image

[Matthew Hart (matthew-hart)]

Ideally LAVA should use expect to detect which boot option to use, rather than going for a hardcoded '3', leaving this bug assigned to LAVA Dispatcher but bumped it to critical

[Alan Bennett (akbennett)]

It looks like the UEFI Firmware has changed breaking Automation in the middle of the release cycle.

works:
https://validation.linaro.org/scheduler/job/128188/log_file#L_27_89
UEFI firmware (version 0.1 built at 11:07:04 on Mar 14 2014)

Fails:
https://validation.linaro.org/scheduler/job/128425/log_file#L_27_85
UEFI firmware (version 0.2 built at 10:09:57 on May 15 2014)

[Fathi Boudra (fboudra)]

Matt's commetn still stand, LAVA shouldn't hardcode 3. If we select 5, it will boot.

[Alan Bennett (akbennett)]

Moving hard-coded support from a partition number to a partition name is a little better, but not ideal.

I propose we review of 'better' boot automation of the fast model during the next LAVA cycle.
  Review
    UEFI shell options
    UEFI startup.nsh
    note: It would be nice if the both fastmodels and known hardware platforms. worked

If, after this investigation, we have a path forward we should implement (or plan to implement) the new support. If it doesn't look like there is a clean path forward, we should look at a better way identify the boot partition to better support a changing UEFI boot menu.

The blueprint for this work is: https://cards.linaro.org/browse/LAVA-1340

[Fathi Boudra (fboudra)]

the firmware has been reverted. pre-built image build is in progress, we should get the results in ~10min

[Fathi Boudra (fboudra)]

https://validation.linaro.org/scheduler/job/128633 is booting fine.

[Milosz Wasilewski (mwasilew)]

It seems that board died in the middle of the test. It didn't boot to master image:
https://validation.linaro.org/scheduler/job/128633/log_file#L_98_0

[Fathi Boudra (fboudra)]

right, I noticed the failure but it isn't related to the firmware:
sync
Command error SYNC
ErrorMessage: Unable to intercept MCC boot prompt
Lava failed at action lava_test_shell with error:Unable to intercept MCC boot prompt