开云体育

Actually bootmethod android only support android boot image version 4
and with AB image, some old platform wtill use android boot image
version 2 with AB or without AB slot.

This patchset add support of both version 2 and non-AB slot images.
It's fixed in same time a boot issue seen on khadas vim3{l} board with 16GB eMMC

patchset was tested on khadas VIM3 and VIM3L with AOSP main branch and
android-mainline kernel and with TI AM62P with android release provided
by TI.

Signed-off-by: Guillaume La Roque <glaroque@...>
---
Changes in v3:
- Remove gpio recovery stuff not needed anymore.
- Link to v2:

Changes in v2:
- Drop patch 3 (configs: khadas-vim3{l}: fix userdata size) already
applied
- Apply Tested-by and Reviewed-by from v1
- Fix comments
- Revert malloc/free for slot_suffix
- Remove vim3/vim3l stuff in meson64_android.h
- Link to v1:

---
Guillaume La Roque (5):
bootstd: android: add support of bootimage v2
bootstd: android: add non-A/B image support
configs: khadas-vim3l_android{_ab}: move on bootmeth android
configs: khadas-vim3_android{_ab}: move on bootmeth android
bootstd: Add test for Android boot image v2

arch/sandbox/dts/test.dts | 10 +++-
boot/Kconfig | 1 -
boot/bootmeth_android.c | 78 +++++++++++++++++++++----------
configs/am62x_a53_android.config | 1 +
configs/khadas-vim3_android_ab_defconfig | 7 ++-
configs/khadas-vim3_android_defconfig | 7 ++-
configs/khadas-vim3l_android_ab_defconfig | 7 ++-
configs/khadas-vim3l_android_defconfig | 7 ++-
configs/sandbox_defconfig | 1 +
include/configs/khadas-vim3_android.h | 25 ++++++++--
include/configs/khadas-vim3l_android.h | 25 ++++++++--
include/configs/meson64_android.h | 6 ---
test/boot/bootflow.c | 29 ++++++++++--
test/py/tests/test_ut.py | 49 +++++++++++++++++++
14 files changed, 206 insertions(+), 47 deletions(-)
---
base-commit: ce427e40a2422b75374ee3404e3f5c6536e8984a
change-id: 20241015-adnroidv2-a01dca609707

Best regards,
--
Guillaume La Roque <glaroque@...>

Hi,

On Tue, 19 Nov 2024 12:37:37 +0100, Guillaume La Roque wrote:

Actually bootmethod android only support android boot image version 4
and with AB image, some old platform wtill use android boot image
version 2 with AB or without AB slot.

This patchset add support of both version 2 and non-AB slot images.
It's fixed in same time a boot issue seen on khadas vim3{l} board with 16GB eMMC

[...]

Thanks, Applied to (u-boot-dfu-next)

[1/5] bootstd: android: add support of bootimage v2

[2/5] bootstd: android: add non-A/B image support

[3/5] configs: khadas-vim3l_android{_ab}: move on bootmeth android

[4/5] configs: khadas-vim3_android{_ab}: move on bootmeth android

[5/5] bootstd: Add test for Android boot image v2


--
Mattijs

Hi Guillaume,

On mar., nov. 19, 2024 at 15:16, Mattijs Korpershoek <mkorpershoek@...> wrote:

Hi,

On Tue, 19 Nov 2024 12:37:37 +0100, Guillaume La Roque wrote:

Actually bootmethod android only support android boot image version 4
and with AB image, some old platform wtill use android boot image
version 2 with AB or without AB slot.

This patchset add support of both version 2 and non-AB slot images.
It's fixed in same time a boot issue seen on khadas vim3{l} board with 16GB eMMC

[...]

Thanks, Applied to (u-boot-dfu-next)

[1/5] bootstd: android: add support of bootimage v2

[2/5] bootstd: android: add non-A/B image support

[3/5] configs: khadas-vim3l_android{_ab}: move on bootmeth android

[4/5] configs: khadas-vim3_android{_ab}: move on bootmeth android

[5/5] bootstd: Add test for Android boot image v2

CI has found some issues when applying these series:


Could you have a look please?

Thanks!

--
Mattijs

Hi,

Le 19/11/2024 à 18:21, Mattijs Korpershoek a écrit?:

Hi Guillaume,

On mar., nov. 19, 2024 at 15:16, Mattijs Korpershoek <mkorpershoek@...> wrote:

Hi,

On Tue, 19 Nov 2024 12:37:37 +0100, Guillaume La Roque wrote:

Actually bootmethod android only support android boot image version 4
and with AB image, some old platform wtill use android boot image
version 2 with AB or without AB slot.

This patchset add support of both version 2 and non-AB slot images.
It's fixed in same time a boot issue seen on khadas vim3{l} board with 16GB eMMC

[...]

Thanks, Applied to (u-boot-dfu-next)

[1/5] bootstd: android: add support of bootimage v2

[2/5] bootstd: android: add non-A/B image support

[3/5] configs: khadas-vim3l_android{_ab}: move on bootmeth android

[4/5] configs: khadas-vim3_android{_ab}: move on bootmeth android

[5/5] bootstd: Add test for Android boot image v2

CI has found some issues when applying these series:


Could you have a look please?

i will check

Thanks!

--
Mattijs

Hi,

unfortunately i don't find for now how i can fix issue in test.

i don't really understand with when we launch test individually it's working but if we launch test_ut test not working ,
I see on android test for image version 4 to mmc7 in bootflow list detected.

=> ut bootstd bootflow_android_image_v4
Test: bootflow_android_image_v4: bootflow.c
scan_mmc_android_bootdev: mmc7
order 0: mmc2
order 1: mmc1
order 2: mmc7
order 3: <NULL>
Enabled mmc mmc7 bootdev
????? ab_select_slot() ANDROID: Attempting slot a, tries remaining 7
????? ab_select_slot() ANDROID: Attempting slot a, tries remaining 7
Older order: mmc2
Showing all bootflows
Seq? Method?????? State?? Uclass??? Part Name Filename
---? -----------? ------? --------? ---- ------------------------
----------------
? 0? extlinux???? ready?? mmc????????? 1 mmc1.bootdev.part_1 /extlinux/extlinux.conf
? 1? android????? ready?? mmc????????? 0 mmc7.bootdev.whole
? 2? android????? ready?? mmc????????? 0 mmc7.bootdev.whole
---? -----------? ------? --------? ---- ------------------------
----------------
(3 bootflows, 3 valid)

if we just launch bootflow_android_image_v4 we have only a mmc7.bootdev.whole and so test is ok.

if someone have an idea or can give some idea of what i can try or check .

thanks for helping.

Guillaume

Hi Guillaume,

On mer., nov. 20, 2024 at 18:54, Guillaume LA ROQUE <glaroque@...> wrote:

Hi,

unfortunately i don't find for now how i can fix issue in test.

i don't really understand with when we launch test individually it's
working but if we launch test_ut test not working ,
I see on android test for image version 4 to mmc7 in bootflow list detected.

=> ut bootstd bootflow_android_image_v4
Test: bootflow_android_image_v4: bootflow.c
scan_mmc_android_bootdev: mmc7
order 0: mmc2
order 1: mmc1
order 2: mmc7
order 3: <NULL>
Enabled mmc mmc7 bootdev
????? ab_select_slot() ANDROID: Attempting slot a, tries remaining 7
????? ab_select_slot() ANDROID: Attempting slot a, tries remaining 7
Older order: mmc2
Showing all bootflows
Seq? Method?????? State?? Uclass??? Part Name Filename
---? -----------? ------? --------? ---- ------------------------
----------------
? 0? extlinux???? ready?? mmc????????? 1 mmc1.bootdev.part_1
/extlinux/extlinux.conf
? 1? android????? ready?? mmc????????? 0 mmc7.bootdev.whole
? 2? android????? ready?? mmc????????? 0 mmc7.bootdev.whole
---? -----------? ------? --------? ---- ------------------------
----------------
(3 bootflows, 3 valid)

if we just launch bootflow_android_image_v4 we have only a
mmc7.bootdev.whole and so test is ok.

if someone have an idea or can give some idea of what i can try or check .

thanks for helping.

I'll have a look

Guillaume
Le 20/11/2024 à 13:56, Guillaume LA ROQUE a écrit?:

Hi,

Le 19/11/2024 à 18:21, Mattijs Korpershoek a écrit?:

Hi Guillaume,

On mar., nov. 19, 2024 at 15:16, Mattijs Korpershoek
<mkorpershoek@...> wrote:

Hi,

On Tue, 19 Nov 2024 12:37:37 +0100, Guillaume La Roque wrote:

Actually bootmethod android only support android boot image version 4
and with AB image, some old platform wtill use android boot image
version 2 with AB or without AB slot.

This patchset add support of both version 2 and non-AB slot images.
It's fixed in same time a boot issue seen on khadas vim3{l} board
with 16GB eMMC

[...]

Thanks, Applied to
(u-boot-dfu-next)

[1/5] bootstd: android: add support of bootimage v2

[2/5] bootstd: android: add non-A/B image support

[3/5] configs: khadas-vim3l_android{_ab}: move on bootmeth android

[4/5] configs: khadas-vim3_android{_ab}: move on bootmeth android

[5/5] bootstd: Add test for Android boot image v2

CI has found some issues when applying these series:


Could you have a look please?

i will check

Thanks!

--
Mattijs

Hi,

On jeu., nov. 21, 2024 at 10:40, Mattijs Korpershoek <mkorpershoek@...> wrote:

Hi Guillaume,

On mer., nov. 20, 2024 at 18:54, Guillaume LA ROQUE <glaroque@...> wrote:

Hi,

unfortunately i don't find for now how i can fix issue in test.

i don't really understand with when we launch test individually it's
working but if we launch test_ut test not working ,
I see on android test for image version 4 to mmc7 in bootflow list detected.

=> ut bootstd bootflow_android_image_v4
Test: bootflow_android_image_v4: bootflow.c
scan_mmc_android_bootdev: mmc7
order 0: mmc2
order 1: mmc1
order 2: mmc7
order 3: <NULL>
Enabled mmc mmc7 bootdev
????? ab_select_slot() ANDROID: Attempting slot a, tries remaining 7
????? ab_select_slot() ANDROID: Attempting slot a, tries remaining 7
Older order: mmc2
Showing all bootflows
Seq? Method?????? State?? Uclass??? Part Name Filename
---? -----------? ------? --------? ---- ------------------------
----------------
? 0? extlinux???? ready?? mmc????????? 1 mmc1.bootdev.part_1
/extlinux/extlinux.conf
? 1? android????? ready?? mmc????????? 0 mmc7.bootdev.whole
? 2? android????? ready?? mmc????????? 0 mmc7.bootdev.whole
---? -----------? ------? --------? ---- ------------------------
----------------
(3 bootflows, 3 valid)

if we just launch bootflow_android_image_v4 we have only a
mmc7.bootdev.whole and so test is ok.

if someone have an idea or can give some idea of what i can try or check .

thanks for helping.

I'll have a look

The following command fails:
$ ./test/py/test.py --bd sandbox --build -k test_ut

Running both test at the same time fails as well:
$ ./test/py/test.py --bd sandbox --build -k bootflow_android_image_v

However, running individually, no issues are observed:
$ ./test/py/test.py --bd sandbox --build -k bootflow_android_image_v4
$ ./test/py/test.py --bd sandbox --build -k bootflow_android_image_v2

Going back to look at the v1 series [1], the tests were working fine.

If we use the following base (from [1]), it works fine:
commit 9e1cd2f2cb86 ("Merge )

However, the don't when using this base:
commit b7d4c80fce44 ("Merge tag 'efi-next-2024-11-18' of into next")

I've bisected the issue and I've found that the following patch from
Simon causes the issue:
commit fbdac8155c89 ("test: Expand implementation of ut_list_has_dm_tests()")

Reverting the above patch fixes the problem when running with:
$ ./test/py/test.py --bd sandbox --build -k test_ut

However, I don't understand why Simon's patch causes a duplicate mmc7
dev to be mounted.

I will continue the investigation.

[1]

Guillaume
Le 20/11/2024 à 13:56, Guillaume LA ROQUE a écrit?:

Hi,

Le 19/11/2024 à 18:21, Mattijs Korpershoek a écrit?:

Hi Guillaume,

On mar., nov. 19, 2024 at 15:16, Mattijs Korpershoek
<mkorpershoek@...> wrote:

Hi,

On Tue, 19 Nov 2024 12:37:37 +0100, Guillaume La Roque wrote:

Actually bootmethod android only support android boot image version 4
and with AB image, some old platform wtill use android boot image
version 2 with AB or without AB slot.

This patchset add support of both version 2 and non-AB slot images.
It's fixed in same time a boot issue seen on khadas vim3{l} board
with 16GB eMMC

[...]

Thanks, Applied to
(u-boot-dfu-next)

[1/5] bootstd: android: add support of bootimage v2

[2/5] bootstd: android: add non-A/B image support

[3/5] configs: khadas-vim3l_android{_ab}: move on bootmeth android

[4/5] configs: khadas-vim3_android{_ab}: move on bootmeth android

[5/5] bootstd: Add test for Android boot image v2

CI has found some issues when applying these series:


Could you have a look please?

i will check

Thanks!

--
Mattijs

On jeu., nov. 21, 2024 at 12:11, "Mattijs Korpershoek via groups.io" <mkorpershoek@...> wrote:

Hi,

On jeu., nov. 21, 2024 at 10:40, Mattijs Korpershoek <mkorpershoek@...> wrote:

Hi Guillaume,

On mer., nov. 20, 2024 at 18:54, Guillaume LA ROQUE <glaroque@...> wrote:

Hi,

unfortunately i don't find for now how i can fix issue in test.

i don't really understand with when we launch test individually it's
working but if we launch test_ut test not working ,
I see on android test for image version 4 to mmc7 in bootflow list detected.

=> ut bootstd bootflow_android_image_v4
Test: bootflow_android_image_v4: bootflow.c
scan_mmc_android_bootdev: mmc7
order 0: mmc2
order 1: mmc1
order 2: mmc7
order 3: <NULL>
Enabled mmc mmc7 bootdev
????? ab_select_slot() ANDROID: Attempting slot a, tries remaining 7
????? ab_select_slot() ANDROID: Attempting slot a, tries remaining 7
Older order: mmc2
Showing all bootflows
Seq? Method?????? State?? Uclass??? Part Name Filename
---? -----------? ------? --------? ---- ------------------------
----------------
? 0? extlinux???? ready?? mmc????????? 1 mmc1.bootdev.part_1
/extlinux/extlinux.conf
? 1? android????? ready?? mmc????????? 0 mmc7.bootdev.whole
? 2? android????? ready?? mmc????????? 0 mmc7.bootdev.whole
---? -----------? ------? --------? ---- ------------------------
----------------
(3 bootflows, 3 valid)

if we just launch bootflow_android_image_v4 we have only a
mmc7.bootdev.whole and so test is ok.

if someone have an idea or can give some idea of what i can try or check .

thanks for helping.

I'll have a look

The following command fails:
$ ./test/py/test.py --bd sandbox --build -k test_ut

Running both test at the same time fails as well:
$ ./test/py/test.py --bd sandbox --build -k bootflow_android_image_v

However, running individually, no issues are observed:
$ ./test/py/test.py --bd sandbox --build -k bootflow_android_image_v4
$ ./test/py/test.py --bd sandbox --build -k bootflow_android_image_v2

Going back to look at the v1 series [1], the tests were working fine.

If we use the following base (from [1]), it works fine:
commit 9e1cd2f2cb86 ("Merge )

However, the don't when using this base:
commit b7d4c80fce44 ("Merge tag 'efi-next-2024-11-18' of into next")

I've bisected the issue and I've found that the following patch from
Simon causes the issue:
commit fbdac8155c89 ("test: Expand implementation of ut_list_has_dm_tests()")

Reverting the above patch fixes the problem when running with:
$ ./test/py/test.py --bd sandbox --build -k test_ut

However, I don't understand why Simon's patch causes a duplicate mmc7
dev to be mounted.

I will continue the investigation.

Ok, I found the issue. it's not a duplicate mmc7, but we bind the
bootmeth multiple times, which causes issues.

This can also be reproduced on next with:
$ ./test/py/test.py --bd sandbox --build -k test_ut
$ ./test/py/test.py --bd sandbox --build -k "bootflow_android or bootflow_cros"

I will send a patch.

[1]

Guillaume
Le 20/11/2024 à 13:56, Guillaume LA ROQUE a écrit?:

Hi,

Le 19/11/2024 à 18:21, Mattijs Korpershoek a écrit?:

Hi Guillaume,

On mar., nov. 19, 2024 at 15:16, Mattijs Korpershoek
<mkorpershoek@...> wrote:

Hi,

On Tue, 19 Nov 2024 12:37:37 +0100, Guillaume La Roque wrote:

Actually bootmethod android only support android boot image version 4
and with AB image, some old platform wtill use android boot image
version 2 with AB or without AB slot.

This patchset add support of both version 2 and non-AB slot images.
It's fixed in same time a boot issue seen on khadas vim3{l} board
with 16GB eMMC

[...]

Thanks, Applied to
(u-boot-dfu-next)

[1/5] bootstd: android: add support of bootimage v2

[2/5] bootstd: android: add non-A/B image support

[3/5] configs: khadas-vim3l_android{_ab}: move on bootmeth android

[4/5] configs: khadas-vim3_android{_ab}: move on bootmeth android

[5/5] bootstd: Add test for Android boot image v2

CI has found some issues when applying these series:


Could you have a look please?

i will check

Thanks!

--
Mattijs

On jeu., nov. 21, 2024 at 15:38, Mattijs Korpershoek <mkorpershoek@...> wrote:

On jeu., nov. 21, 2024 at 12:11, "Mattijs Korpershoek via groups.io" <mkorpershoek@...> wrote:

Hi,

On jeu., nov. 21, 2024 at 10:40, Mattijs Korpershoek <mkorpershoek@...> wrote:

Hi Guillaume,

On mer., nov. 20, 2024 at 18:54, Guillaume LA ROQUE <glaroque@...> wrote:

Hi,

unfortunately i don't find for now how i can fix issue in test.

i don't really understand with when we launch test individually it's
working but if we launch test_ut test not working ,
I see on android test for image version 4 to mmc7 in bootflow list detected.

=> ut bootstd bootflow_android_image_v4
Test: bootflow_android_image_v4: bootflow.c
scan_mmc_android_bootdev: mmc7
order 0: mmc2
order 1: mmc1
order 2: mmc7
order 3: <NULL>
Enabled mmc mmc7 bootdev
????? ab_select_slot() ANDROID: Attempting slot a, tries remaining 7
????? ab_select_slot() ANDROID: Attempting slot a, tries remaining 7
Older order: mmc2
Showing all bootflows
Seq? Method?????? State?? Uclass??? Part Name Filename
---? -----------? ------? --------? ---- ------------------------
----------------
? 0? extlinux???? ready?? mmc????????? 1 mmc1.bootdev.part_1
/extlinux/extlinux.conf
? 1? android????? ready?? mmc????????? 0 mmc7.bootdev.whole
? 2? android????? ready?? mmc????????? 0 mmc7.bootdev.whole
---? -----------? ------? --------? ---- ------------------------
----------------
(3 bootflows, 3 valid)

if we just launch bootflow_android_image_v4 we have only a
mmc7.bootdev.whole and so test is ok.

if someone have an idea or can give some idea of what i can try or check .

thanks for helping.

I'll have a look

The following command fails:
$ ./test/py/test.py --bd sandbox --build -k test_ut

Running both test at the same time fails as well:
$ ./test/py/test.py --bd sandbox --build -k bootflow_android_image_v

However, running individually, no issues are observed:
$ ./test/py/test.py --bd sandbox --build -k bootflow_android_image_v4
$ ./test/py/test.py --bd sandbox --build -k bootflow_android_image_v2

Going back to look at the v1 series [1], the tests were working fine.

If we use the following base (from [1]), it works fine:
commit 9e1cd2f2cb86 ("Merge )

However, the don't when using this base:
commit b7d4c80fce44 ("Merge tag 'efi-next-2024-11-18' of into next")

I've bisected the issue and I've found that the following patch from
Simon causes the issue:
commit fbdac8155c89 ("test: Expand implementation of ut_list_has_dm_tests()")

Reverting the above patch fixes the problem when running with:
$ ./test/py/test.py --bd sandbox --build -k test_ut

However, I don't understand why Simon's patch causes a duplicate mmc7
dev to be mounted.

I will continue the investigation.

Ok, I found the issue. it's not a duplicate mmc7, but we bind the
bootmeth multiple times, which causes issues.

This can also be reproduced on next with:
$ ./test/py/test.py --bd sandbox --build -k test_ut
$ ./test/py/test.py --bd sandbox --build -k "bootflow_android or bootflow_cros"

I will send a patch.

[1]

Guillaume
Le 20/11/2024 à 13:56, Guillaume LA ROQUE a écrit?:

Hi,

Le 19/11/2024 à 18:21, Mattijs Korpershoek a écrit?:

Hi Guillaume,

On mar., nov. 19, 2024 at 15:16, Mattijs Korpershoek
<mkorpershoek@...> wrote:

Hi,

On Tue, 19 Nov 2024 12:37:37 +0100, Guillaume La Roque wrote:

Actually bootmethod android only support android boot image version 4
and with AB image, some old platform wtill use android boot image
version 2 with AB or without AB slot.

This patchset add support of both version 2 and non-AB slot images.
It's fixed in same time a boot issue seen on khadas vim3{l} board
with 16GB eMMC

[...]

Thanks, Applied to
(u-boot-dfu-next)

[1/5] bootstd: android: add support of bootimage v2

[2/5] bootstd: android: add non-A/B image support

[3/5] configs: khadas-vim3l_android{_ab}: move on bootmeth android

[4/5] configs: khadas-vim3_android{_ab}: move on bootmeth android

[5/5] bootstd: Add test for Android boot image v2

CI has found some issues when applying these series:


Could you have a look please?

i will check

Thanks!

--
Mattijs

Hello,

On mar., nov. 19, 2024 at 15:16, "Mattijs Korpershoek via groups.io" <mkorpershoek@...> wrote:

Hi,

On Tue, 19 Nov 2024 12:37:37 +0100, Guillaume La Roque wrote:

Actually bootmethod android only support android boot image version 4
and with AB image, some old platform wtill use android boot image
version 2 with AB or without AB slot.

This patchset add support of both version 2 and non-AB slot images.
It's fixed in same time a boot issue seen on khadas vim3{l} board with 16GB eMMC

[...]

Thanks, Applied to (u-boot-dfu-next)

[1/5] bootstd: android: add support of bootimage v2

[2/5] bootstd: android: add non-A/B image support

[3/5] configs: khadas-vim3l_android{_ab}: move on bootmeth android

[4/5] configs: khadas-vim3_android{_ab}: move on bootmeth android

[5/5] bootstd: Add test for Android boot image v2

Due to the test regressions discussed in this thread, I'm dropping this
series from the u-boot-dfu tree.

Please submit a v4 targetting next once [1] lands in next.

Thanks,
Mattijs

[1]

--
Mattijs

Hi,

Le 25/11/2024 à 13:06, Mattijs Korpershoek a écrit?:

Hello,

On mar., nov. 19, 2024 at 15:16, "Mattijs Korpershoek via groups.io" <mkorpershoek@...> wrote:

Hi,

On Tue, 19 Nov 2024 12:37:37 +0100, Guillaume La Roque wrote:

Actually bootmethod android only support android boot image version 4
and with AB image, some old platform wtill use android boot image
version 2 with AB or without AB slot.

This patchset add support of both version 2 and non-AB slot images.
It's fixed in same time a boot issue seen on khadas vim3{l} board with 16GB eMMC

[...]

Thanks, Applied to (u-boot-dfu-next)

[1/5] bootstd: android: add support of bootimage v2

[2/5] bootstd: android: add non-A/B image support

[3/5] configs: khadas-vim3l_android{_ab}: move on bootmeth android

[4/5] configs: khadas-vim3_android{_ab}: move on bootmeth android

[5/5] bootstd: Add test for Android boot image v2

Due to the test regressions discussed in this thread, I'm dropping this
series from the u-boot-dfu tree.

Please submit a v4 targetting next once [1] lands in next.

next was update so i will send v4 today, thanks for fix on test.

Guillaume

Thanks,
Mattijs

[1]

--
Mattijs