Age | Commit message (Collapse) | Author |
|
remount overlay persistence issues
It was reported that changes made after "adb remount" were not
persisting after a reboot. After reproducing and debugging, I
was at a bit of a loss for why.
After quite a bit of floundering around, I finally realized we
were missing first_stage_mount from the fstab arguments. This
prevents the metadata partition from being available early
enough for the overlay images to be located and applied at
bootup.
Signed-off-by: John Stultz <john.stultz@linaro.org>
Change-Id: I63a8cedacbace33cceee7f83121a4b83dd996cc3
|
|
Over the last week I started seeing some occasional trouble
with the f2fs entry in fstab.hikey960 causing init confusion
where the fsck found bad magic value (since the partition is
ext4), but the partition wasn't released before the ext4
mount began and saw it as in use.
This caused userdata not not be mounted at all, and resulted
in a number of keystore errors and the system not booting
fully to UI.
I don't think anyone is really using the f2fs userdata config
with HiKey960, so lets drop it.
Signed-off-by: John Stultz <john.stultz@linaro.org>
Change-Id: Ibc34b74a252fd94618fb454db4b9c46ec9eef1a5
|
|
The maximum size of all dynamic partition groups should not exceed
the super size - DAP metadata size. Today the configuration of
the device doesn't take the metadata into acount.
Since we recommend 4M to partners, reserve 4M here.
Bug: 182431975
Test: build
Change-Id: I2feb06531bf0bbbff8845fa485b5b4917152bad6
|
|
With dynamic partitions we need /metadata mounted to properly
make 'adb remount' work.
So this makes use of the unused cache partition to mount it as
metadata (along with the needed build env values).
With this patch 'adb remount' works.
Signed-off-by: John Stultz <john.stultz@linaro.org>
Change-Id: I33da79ec15b453dbdc3b8a3180863abeb67d4cb6
|
|
While we boot up properly with the dynamic partitions in only
the ramdisk fstab, Yongqin noted adb remount was failing recently.
On looking into the issue, I realized one issue is we don't have
the dynamic paritions listed in the vendor fstab which causes
some of the remount logic to not find the partitions.
This patch adds the dynamic partitions and unifies the fstab
file for hikey960 so we have the same one for both the ramdisk
and vendor image.
NOTE: adb remount still does not fully work with this change. We
will need a follow on patch to fix that.
Signed-off-by: John Stultz <john.stultz@linaro.org>
Change-Id: Idf9c463162398ef89db31fbf69d205d1a47e2cdd
|
|
super.img was not included in the update image and
was therefore not visible to update tools
Test: the super.img is in the hikey960-img-eng.xxx.zip file
Signed-off-by: Yongqin Liu <yongqin.liu@linaro.org>
Change-Id: I1e025c7ac57f4b6c88333ab47065d018afe2f4ee
|
|
This updates the ptable to use the super label instead of system
to avoid using more fragile retrofit options in the build.
Signed-off-by: John Stultz <john.stultz@linaro.org>
Change-Id: Ic57287d7cfd2d6bfa54f3a848d07dc1b858a3810
|
|
Following again in Amit's footsteps, this implements virtual A/B
in the same fashion as Amit's patch for db845c:
https://android-review.googlesource.com/c/device/linaro/dragonboard/+/1547736
As the bootloader doesn't support the slot_suffix, we fix it to
_a statically in the boot args.
Signed-off-by: John Stultz <john.stultz@linaro.org>
Change-Id: I80079d8eeb2635fece1f0f5f24eeaeff3a1c53db
|
|
This is basically a port of work Amit had done against HiKey
board awhile back:
https://android-review.linaro.org/c/device/linaro/hikey/+/20550
Moving system vendor product and system_ext to the super
partition.
For now, instead of reworking the ptable, we simply flash the
super partition to the system partition. This will need to change
so that we change the ptable as well, but that needs some time
as the ptable flashing can be disruptive for testing.
Feedback would be appreciated.
Signed-off-by: John Stultz <john.stultz@linaro.org>
Change-Id: Ie53fed6392116b16063f497978e369490a1fccd2
|
|
|
|
picking the dummy hcd
With some kernels that neable DUMMY_USB_HCD, we end up with
multiple UDC devices, and the system might pick the dummy one
to use which causes adb to fail.
This corrects this by setting the sys.usb.controller property
to the hikey/hikey960 controller name.
Test: adb connection works with android11-5.4 kernel for hikey/hikey960
Change-Id: Ife99b9b16204cfe767c8b755ec5208ce5047c211
Signed-off-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Yongqin Liu <yongqin.liu@linaro.org>
|
|
Bug: 173737347
Test: treehugger
Change-Id: Id7bcbb13c69514884c9859e4593569b36f3d86ec
|
|
to fix the failure of reported the test
android.graphics.cts.VulkanDeqpLevelTest#testVulkanDeqpLevel
of the CtsGraphicsTestCases module
Test: CtsGraphicsTestCases android.graphics.cts.VulkanDeqpLevelTest#testVulkanDeqpLevel passed
Signed-off-by: Yongqin Liu <yongqin.liu@linaro.org>
Change-Id: I989893e559b7387540f9d2458ce45b3512245db1
|
|
With sdcardfs deprecated, we need to include a few tweaks
to get emulated storage functionality working properly.
Many thanks to Martijn Coenen and Daniel Rosenberg for their
help narrowing down the needed changes.
Signed-off-by: John Stultz <john.stultz@linaro.org>
Change-Id: Idbebf0c14914a148dd9ed0d5f071aaed435665b2
|
|
as the "adb remount" checks the /vendor/etc/fstab.hikey960
file to find if the partition is a known partition,
so we need to add vendor here to make it work with adb remount.
Otherwise will get following error:
$ adb remount vendor
Unknown partition vendor, skipping
No partitions to remount
remount failed
$
Or it's not remounted to be writable after adb remount command.
Test: adb remount, and vendor partition is writable.
Signed-off-by: Yongqin Liu <yongqin.liu@linaro.org>
Change-Id: I4a6e0ddc31cd62bcf9b7c42083656c7dbaed5009
|
|
The vendor codec package doesn't work with 4.14+ kernels, and
is now causing build issues, so its time to stop using it.
Stop including the Make file, and remove the installation nag
message from the build so folks don't continue to install it.
Signed-off-by: John Stultz <john.stultz@linaro.org>
Change-Id: If59fc36871cae6c78917660789ecd3109e05aa59
|
|
As FBE is required by Android 10 and higher.
The change here referenced documents here:
https://source.android.com/security/encryption/file-based
Test: ./android-vts/tools/vts-tradefed run vts-kernel -m VtsKernelEncryptionTest
with android-mainline, android-5.4, android-hikey-linaro-4.19, android-hikey-linaro-4.14 kernels
Signed-off-by: Yongqin Liu <yongqin.liu@linaro.org>
Change-Id: Ic40d3fd55d6d10cf159e079e258e7deec4f0bf81
|
|
We're banning ELF prebuilts in PRODUCT_COPY_FILES. Define prebuilt
modules for vendor prebuilts so we can run check_elf_file.py on them.
Use PRODUCT_PACKAGES to install ELF prebuilts.
Bug: 140560012
Test: Build pass. Built artifacts should not change.
Test: lunch hikey960-userdebug
Change-Id: Ibef1ef8e2d600a59fe444484679f25c018bd1ecc
|
|
This reverts commit 5cec2d8bdcbceeca8ba541baf16e7221a1d80e6c.
Reason for revert: Replacing with HIDL-ized version
Change-Id: Ia9cc1f0903aa272a575095d2fee540b3d863d48c
Merged-In: Ia9cc1f0903aa272a575095d2fee540b3d863d48c
|
|
Add the cuttlefish gatekeeper hal to the hikey/hikey960 build
Change-Id: I7b7e66f84dfcdfc097efa0903205bf4ce3561191
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
boot.img is created"
|
|
created
Since https://android-review.googlesource.com/c/platform/build/+/951796 landed,
boot.img stopped being created for hikey/hikey960.
After some digging there, it seems the hikey/hikey960 boards never
set BOARD_BOOTIMAGE_PARTITION_SIZE which is now required.
Change-Id: I820b93879f68ea6e354dd6b5c7319769bc076ce9
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
This is temporary and required due to conflicts in the names
used by the Mali RS driver binaries and AOSP binaries.
Bug: 131709055
Test: manual
Change-Id: Ia1509cb4d12267f35f1e85e2494628140b2881b8
|
|
Instead of conditionally compiling hwcomposer.drm for hikey and
hikey960, compile two different modules and selectively use them
on hikey and hikey960.
Bug: 122332597
Test: none
Change-Id: Ifcb64798bc34101d9e1f851454c3fe361c478c12
|
|
This allows us to deprecate the dtb fstab dependency. Because 4.19 also
does not support android-verity, disable the feature at the same time.
Bug: 121043321
Test: not yet tested
Change-Id: I9c2a8c140ed64faecee0d66e00033fdffa398a4b
Signed-off-by: Alistair Strachan <astrachan@google.com>
|
|
Bug: 117293412
Change-Id: Ic4c5119bc10827484a49a39b895bd82e678c6cf0
|
|
as required by cts test defined here:
https://android.googlesource.com/platform/cts/+/master/tests/tests/permission/src/android/permission/cts/FileSystemPermissionTest.java#1000
Test: run cts -m CtsPermissionTestCases \
-t android.permission.cts.FileSystemPermissionTest#testDataMountedNoSuidNoDev
Change-Id: I162725a8c737bbc89b2dacaa634484258aa3cd95
Signed-off-by: Yongqin Liu <yongqin.liu@linaro.org>
|
|
To easily adjust partitions, avoid using hard-coded value sdd13.
Test: boot
Change-Id: I97f976c781931363c62550e40124471011e0436b
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
|
|
Since the FIQ_debugger is now deprecated in android-4.19,
and its utility on hikey/hikey960 has always been minimal
given there isn't a real FIQ (just an maskable IRQ), lets
just remove the use of the FIQ0 console.
We still have to keep it for 4.4/4.9/4.14 hikey kernels,
as removing it breaks console input. We will need to
land kernel changes and update userspace.
But for hikey960 removing it is ok.
Change-Id: I85ac12660ab0b5d38d64486ffef4ae0cebbe5e8f
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
If set to true, kernel will use builtin EDID for 1920x1080 display
Bug: 117215508
Test: Manual
Change-Id: Ie76e4b0d75f5ec8a4b6f3b8c71b4328865fd487e
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
|
|
This patch provides hifi changes that are needed to work with
FULL_TREBLE enabled.
As /system/bin does not support write permissions moved dec-out.pcm
file to /data/dec-out.pcm
Copied dhifimesg to /vendor/bin in order to capture hifi dsp logs
Bug: 116226155
Change-Id: Ia90f09760198b3fefe5ef3dfbdf5d7e11dd67a4b
Originally-by: Niranjan Yadla <nyadla@cadence.com>
Signed-off-by: Alistair Strachan <astrachan@google.com>
|
|
This does not work with latest GSI on AOSP master
which has cache as a symlink to /data/cache where
trying to mount cache partition fails.
Change-Id: I0621accc580a5ae8ccf2494a76df457f152d51e3
Signed-off-by: Vishal Bohj <vishal.bhoj@linaro.org>
|
|
Re-add the renderscript binaries to the build and
add sepolicy and manifest changes needed to support
ARM's renderscript library.
Change-Id: I743ffe93a4e41d73122dc0d2a1ad8eb962597649
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
This patch adds prebuilt binary/libraries needed by the mali renderscript
code.
Following the renderscript integration instructions in the release
notes from the binary blob package here:
https://developer.arm.com/products/software/mali-drivers/user-space
Specifically:
https://developer.arm.com/-/media/Files/downloads/mali-drivers/user-space/HiKey%20960/mali-G71_r10p0-01rel1_android-8.1_1.tar.gz?revision=56403097-18f0-4ea2-9ff8-de7044b71e5e
These were built by repo sync'ing Android-8.1, and building:
OVERRIDE_RS_DRIVER=libRSDriverArm.so make libclcore.bc libclcore_neon.bc libbcc libLLVM bcc
Then the resulting binaries were copied from the system dir
to here.
Change-Id: Ife139a20c2a9aa1727de0668ba85309d9b09920f
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
From: Yu Chen <chenyu56@huawei.com>
With the usb rework going on against mainline kernels, Android
sometimes gets confused as to which device to use for usb-gadget
functionality.
Thus this adds:
PRODUCT_DEFAULT_PROPERTY_OVERRIDES += sys.usb.controller=ff100000.dwc3
To make sure the right device is selected.
Without this, if we boot using host mode and then plug in the
gadget connection, the system will not setup the gadget device
and adb will not function.
Additionally, this should not have any side-effects on older
kernels. Yu Chen clarifies: the property of "sys.usb.controller"
should always be "ff100000.dwc3" on hikey960. Previously we did
not use the code from drivers/usb/dwc3/drd.c to handle the
data-role-switch" in our older code, and the dwc3 udc always
exists in "/sys/class/udc" in older kernels.
Change-Id: Ic21da64ab1b7cba452ca0ec9580ba90ca59f5d0c
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
Add a build option TARGET_ANDROID_VERITY which can be set to anything
but the empty string to enable dm-android-verity. This option is
introduced because the UEFI bootloader for hikey and hikey960 has a bug
where it incorrectly rejects boot.img files with no ramdisk (see
https://bugs.96boards.org/show_bug.cgi?id=757).
This change *will* work on a hikey960 with the non-UEFI bootloader
installed.
Test: local
Bug: 72722987
Bug: 111829702
Change-Id: I75d0941866d9196e78e7c2943cdf05d35beb1879
Signed-off-by: Alistair Strachan <astrachan@google.com>
|
|
The partition table blob defines partition sizes which do not match the
size of the images created by the Android build system. This tickles a
bug in the libfec userspace code, which is fixed in system/extras
Ia6cdbc0f35a98fdfa1fb4677951ae5c829cabb0f, but we may as well stop
wasting flash space.
Test: local
Bug: 72722987
Change-Id: Ia971c0c3ae5b3e798b107a8c6d571cb67d8ec0b9
Signed-off-by: Alistair Strachan <astrachan@google.com>
|
|
This adds a version check on the hisi vdec package and provides
a URL to the current expected package.
This did get a bit hacky as we can't check values set in the
vendor pkg makefile if we use inherit-product. So we have to
include it directly. I think we could do it better, but we would
have to re-spin the published vendor binary, so lets hack around
it this time and do better with the next package update.
Change-Id: Ic6832850e9b4d03b84358060810aa17e23e5b8ae
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
When we moved the fstab files over, we probably shouldn't have
copied them to /vendor/etc/init/ and instead used /vendor/etc/.
Thus this patch corrects this.
Change-Id: I2143fb6596d50289701158b58c91dfb440943231
Reported-by: Mark Salyzyn <salyzyn@android.com>
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
It is not compatible with treble as it needs additional dependencies to
be built and installed to /vendor. There is also a conflicting rule to
build android.hardware.renderscript@1.0.so and we are about to
deprecate support for this.
build/make/core/base_rules.mk:294: warning: ignoring old commands for target
`out/target/product/hikey960/obj_arm/lib/android.hardware.renderscript@1.0-impl.so'
Bug: 77611511
Bug: 77909036
Bug: 110479079
Change-Id: I4c0f6643c5f9000c6a2131fd6b1d4170fe8b07f4
Signed-off-by: Alistair Strachan <astrachan@google.com>
|
|
This is basically the current state of the rest of the
FULL_TREBLE enablement bits for both hikey/hikey960
This patch also requires kernels that are rebuilt to include the
DTS fstab changes seen here:
https://git.linaro.org/people/john.stultz/android-dev.git/log/?h=dev/treble
Credit for this patch goes to Vishal Bhoj <vishal.bhoj@linaro.org>
who basically was the original author and magician
who got things working.
I've only refactored and split up his changes to help
with the submissions.
Change-Id: Iab0f5fc93bdfd1b86c98aa1d54974fde33514063
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
Remove arm specific bits since we host libraries
in git repository.
Add hisilicon decoder libraries
Change-Id: Ic70cf9dd96579589dbffda390ac8dc913370e95a
Signed-off-by: Vishal Bhoj <vishal.bhoj@linaro.org>
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
Bug: 77482940
Test: Manual
Change-Id: Ic292f3aa79cc8da15e7472da920a2dcdd72a8533
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
|
|
The Vulkan runtime only looks in /vendor/lib[64]/hw for the Vulkan HAL,
not in /vendor/lib[64]/ where it was being installed.
The current driver supports the features required by
FEATURE_VULKAN_HARDWARE_LEVEL=1, so advertise it as required by CTS
VulkanFeaturesTest.
Test: adb shell cmd gpu vkjson
Test: cts-tradefed run cts -m CtsGraphicsTestCases -t android.graphics.cts.VulkanFeaturesTest
Change-Id: I6fccf8a34da9ba303be698032c49b7211f9d3d89
|
|
Make sure new vendor partition is flashed in flashall scripts
Bug: 77684218
Test: Manual
Change-Id: I8cebbd506f8810ec2912b2fa5572ecc0df05e64d
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
Bug: 75290651
Test: Manual
Change-Id: I21939c7b290775ae492acb02dea485dfebba130e
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
|
|
|
|
as it is reported by cts test:
android.os.cts.EnvironmentTest.testNoAtime junit.framework.AssertionFailedError: Found device mounted at /dev/block/platform/soc/f723d000.dwmmc0/by-name/cache without 'noatime' option, which can cause unnecessary flash churn; please update your fstab.
Test: manually with cts android.os.cts.EnvironmentTest#testNoAtime test
Change-Id: Id9f6c7a5cbb74888805047c7da172c1db10f737d
Signed-off-by: Yongqin Liu <yongqin.liu@linaro.org>
|
|
With new licence terms, we can now redistribute mali blobs
without the vendor package type-through license. Instead
the EULA just has to be included with the blobs.
So switch to using included r9p0 binaries rather then
using the vendor package.
Change-Id: If16cd4fbeb317c146d3c90c93cce6a73a6530dd0
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
Bug: 76019495
Test: Manual
Change-Id: I1399cd89caec21e47a27f7a491b91939c1ccdb40
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
|