[U-Boot] [PATCH 2/2] travis: Wire Xilinx Versal Virt platform

Tue Jan 15 08:58:02 UTC 2019

On 15.01.19 09:56, Michal Simek wrote:
> On 15. 01. 19 9:54, Alexander Graf wrote:
>>
>>
>> On 15.01.19 08:07, Michal Simek wrote:
>>> On 07. 01. 19 23:59, Tom Rini wrote:
>>>> On Mon, Jan 07, 2019 at 09:15:17PM +0100, Alexander Graf wrote:
>>>>>
>>>>>
>>>>> On 07.01.19 18:13, Tom Rini wrote:
>>>>>> On Mon, Jan 07, 2019 at 05:57:10PM +0100, Alexander Graf wrote:
>>>>>>> On 01/07/2019 01:46 PM, Michal Simek wrote:
>>>>>>>> On 07. 01. 19 13:28, Tom Rini wrote:
>>>>>>>>> On Mon, Jan 07, 2019 at 01:18:46PM +0100, Michal Simek wrote:
>>>>>>>>>> On 04. 01. 19 4:01, Tom Rini wrote:
>>>>>>>>>>> On Fri, Jan 04, 2019 at 09:29:47AM +0800, Bin Meng wrote:
>>>>>>>>>>>> On Thu, Jan 3, 2019 at 9:39 PM Tom Rini <trini at konsulko.com> wrote:
>>>>>>>>>>>>> On Thu, Jan 03, 2019 at 09:00:41AM +0100, Michal Simek wrote:
>>>>>>>>>>>>>> On 26. 12. 18 15:29, Tom Rini wrote:
>>>>>>>>>>>>>>> On Thu, Dec 20, 2018 at 04:30:19PM +0800, Bin Meng wrote:
>>>>>>>>>>>>>>>> Hi Michal,
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> On Thu, Dec 20, 2018 at 4:17 PM Michal Simek <michal.simek at xilinx.com> wrote:
>>>>>>>>>>>>>>>>> On 20. 12. 18 9:09, Bin Meng wrote:
>>>>>>>>>>>>>>>>>> On Thu, Dec 20, 2018 at 3:55 PM Michal Simek <michal.simek at xilinx.com> wrote:
>>>>>>>>>>>>>>>>>>> Test Xilinx Versal Virt platform running on the v3.1.0 Qemu.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Signed-off-by: Michal Simek <michal.simek at xilinx.com>
>>>>>>>>>>>>>>>>>>> ---
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Patch needs to be applied when this PR is merged.
>>>>>>>>>>>>>>>>>>> https://github.com/swarren/uboot-test-hooks/pull/21
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> ---
>>>>>>>>>>>>>>>>>>>  .travis.yml | 8 ++++++++
>>>>>>>>>>>>>>>>>>>  1 file changed, 8 insertions(+)
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> diff --git a/.travis.yml b/.travis.yml
>>>>>>>>>>>>>>>>>>> index 6e4b4ba0a34a..e14e6987e4cf 100644
>>>>>>>>>>>>>>>>>>> --- a/.travis.yml
>>>>>>>>>>>>>>>>>>> +++ b/.travis.yml
>>>>>>>>>>>>>>>>>>> @@ -454,6 +454,14 @@ matrix:
>>>>>>>>>>>>>>>>>>>            QEMU_TARGET="arm-softmmu"
>>>>>>>>>>>>>>>>>>>            TEST_PY_ID="--id qemu"
>>>>>>>>>>>>>>>>>>>            BUILDMAN="^zynq_zc702$"
>>>>>>>>>>>>>>>>>>> +    - name: "test/py xilinx_versal_virt"
>>>>>>>>>>>>>>>>>>> +      env:
>>>>>>>>>>>>>>>>>>> +        - TEST_PY_BD="xilinx_versal_virt"
>>>>>>>>>>>>>>>>>>> +          TEST_PY_TEST_SPEC="not sleep"
>>>>>>>>>>>>>>>>>>> +          QEMU_TARGET="aarch64-softmmu"
>>>>>>>>>>>>>>>>>>> +          QEMU_VERSION="v3.1.0"
>>>>>>>>>>>>>>>>>>> +          TEST_PY_ID="--id qemu"
>>>>>>>>>>>>>>>>>>> +          BUILDMAN="^xilinx_versal_virt$"
>>>>>>>>>>>>>>>>>> Can we turn on 3.1.0 for all QEMU targets?
>>>>>>>>>>>>>>>>> I expected this question.
>>>>>>>>>>>>>>>>> First of all I wanted to enable an option to be able to wire whatever
>>>>>>>>>>>>>>>>> version. Even sha1 for not released version.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> I think we can try to test v3.1.0 qemu version for all boards but I will
>>>>>>>>>>>>>>>>> let Tom to decide is we can move or not.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Yes, please try to test that. If everything is good, I see no reason
>>>>>>>>>>>>>>>> not using it.
>>>>>>>>>>>>>>> Yes, I would also like to move everyone up to v3.1.0 and also having the
>>>>>>>>>>>>>>> version we use be spelled out will help with future reproducibility.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>> v3.1 is not working on vexpress_ca9x4 and vexpress_ca15_tc2.
>>>>>>>>>>>>>> take a look at https://travis-ci.org/michalsimek/u-boot/builds/470873127
>>>>>>>>>>>>>> I am happy to keep them on v3.0 and move the rest to v3.1 and let
>>>>>>>>>>>>>> vexpress guys to figured out what's wrong with v3.1 and configs around.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> What do you think?
>>>>>>>>>>>>> The error is:
>>>>>>>>>>>>> ---------------------------- Captured stdout setup -----------------------------
>>>>>>>>>>>>> +u-boot-test-reset vexpress_ca15_tc2 qemu
>>>>>>>>>>>>> qemu-system-arm: -tftp: invalid option
>>>>>>>>>>>>>
>>>>>>>>>>>>> And I guess the problem is that by v3.1.0 -tftp has been removed and we
>>>>>>>>>>>>> need to use -netdev user,id=net0,tftp=${UBOOT_TRAVIS_BUILD_DIR} like on
>>>>>>>>>>>>> the other QEMU-based targets.  Can you update uboot-test-hooks (change
>>>>>>>>>>>>> .travis.yml to clone your fork of uboot-test-hooks rather than
>>>>>>>>>>>>> swarren's) to have the modern syntax?  Thanks!
>>>>>>>>>>>> I think the correct approach is to update uboot-test-hooks and send PR
>>>>>>>>>>>> to Stephen, then apply this 3.1.0 patch on the U-Boot side. We should
>>>>>>>>>>>> still have U-Boot's .travis.yml point to swarren's git repo, no?
>>>>>>>>>>> Yes, sorry, to be clear, I was just saying how to debug / confirm the
>>>>>>>>>>> changes to uboot-test-hooks.
>>>>>>>>>> I have created new pull request for fixing that network issue here.
>>>>>>>>>> https://github.com/swarren/uboot-test-hooks/pull/22
>>>>>>>>>>
>>>>>>>>>> But for vexpress_ca15_tc2 there is still an issue with helloworld efi.
>>>>>>>>>> It works on v3.0 but not on v3.1.
>>>>>>>>>>
>>>>>>>>>> => => tftpboot 80000000 lib/efi_loader/helloworld.efi
>>>>>>>>>> smc911x: MAC 52:54:00:12:34:56
>>>>>>>>>> smc911x: detected LAN9118 controller
>>>>>>>>>> smc911x: phy initialized
>>>>>>>>>> smc911x: MAC 52:54:00:12:34:56
>>>>>>>>>> Using smc911x-0 device
>>>>>>>>>> TFTP from server 10.0.2.2; our IP address is 10.0.2.15
>>>>>>>>>> Filename 'lib/efi_loader/helloworld.efi'.
>>>>>>>>>> Load address: 0x80000000
>>>>>>>>>> Loading: #
>>>>>>>>>> 	 2 MiB/s
>>>>>>>>>> done
>>>>>>>>>> Bytes transferred = 2144 (860 hex)
>>>>>>>>>> smc911x: MAC 52:54:00:12:34:56
>>>>>>>>>> => => crc32 80000000 $filesize
>>>>>>>>>> CRC32 for 80000000 ... 8000085f ==> b76d87c0
>>>>>>>>>> => => bootefi 80000000
>>>>>>>>>> Scanning disks on mmc...
>>>>>>>>>> ^[[61;213RCard did not respond to voltage select!
>>>>>>>>>> MMC Device 1 not found
>>>>>>>>>> MMC Device 2 not found
>>>>>>>>>> MMC Device 3 not found
>>>>>>>>>> Found 0 disks
>>>>>>>>>> WARNING: booting without device tree
>>>>>>>>>> ## Starting EFI application at 80000000 ...
>>>>>>>>>> Fs+u-boot-test-reset vexpress_ca15_tc2 qemu
>>>>>>>>>>
>>>>>>>>>> I need to move to something else instead of hacking this target.
>>>>>>>>>>
>>>>>>>>>> My goal is to get Versal to travis to make sure that none is breaking it
>>>>>>>>>> not debugging issues with platform I don't know.
>>>>>>>>>> Can I keep this platform on v3.0 and move the rest to 3.1?
>>>>>>>>> Alex, any ideas?  Can you take a look?  Thanks!
>>>>>>>>>
>>>>>>>> Just in case you want to see some logs from travis.
>>>>>>>> https://travis-ci.org/michalsimek/u-boot/builds/475350468
>>>>>>>
>>>>>>>
>>>>>>> So the offending change is a new capability in QEMU that exposes EL2 (HYP)
>>>>>>> mode on the Cortex-A15 core which this vexpress model uses. There are 2
>>>>>>> steps here IMHO:
>>>>>>>
>>>>>>> 1) Disable EL2 again by hand. This is easily doable by passing "-cpu
>>>>>>> cortex-a15,has_el2=off" to the QEMU command line. That restores the previous
>>>>>>> behavior
>>>>>>
>>>>>> ... which is one more change to the uboot-test-scripts repo, can someone
>>>>>> please?  Thanks.
>>>>>
>>>>> Done: https://github.com/swarren/uboot-test-hooks/pull/23
>>>>
>>>> aaaand this breaks older QEMU.  I think for the moment, can you please
>>>> revert this Stephen?  We'll take the QEMU version bump early on in the
>>>> next release and I'll let you know off-list when I'm grabbing it.
>>>
>>> Alex: I expect this hook patch is not needed when you qemu patch is
>>> applied. Did it go to qemu tree?
>>
>> Peter promised he would apply it (he maintains the arm tree in QEMU),
>> but it's not in QEMU's master yet.
>>
>>> Tom: I am proposing this way for upgrade.
>>>
>>> Apply this one:
>>> https://patchwork.ozlabs.org/patch/1016580/
>>>
>>> Bump up qemu version to v3.1.0 with setting up
>>> a) QEMU_VERSION="v3.0.0" for vexpress
>>> or
>>> b) QEMU_VERSION="qemu with Alex fix sha1" for vexpress
>>
>> v3.0.0 for now, then move everyone over to v3.2.0 (or v3.1.1) when it's
>> out I'd say.
> 
> Then I would like to use that patches I have sent already and enabled
> verson on v3.1.0 just for versal.

Ah, sorry, what I was trying to say was "v3.1.0 for everyone but
vexpress, v3.0.0 for vexpress_a15" and then "v3.2.0 for everyone once
it's out".

If we have to have one exception, we may as well have the broken case be
the exception, which is vexpress_15 :).

Alex