As we by now require working tun/tap support in CI runner to allow running ACTS,
there is no need for using slirp networking.

And who knows, maybe the randomly failing ssh ptest could be related to the QEMU
slirp implementation.

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

The proper way to include SSH server is to use IMAGE_FEATURES.

Signed-off-by: Esben Haabendal <esben@geanix.com>

This reverts commit 312c51f4.

We are seeing multiple different race conditions causing random failures in
ACTS, many of them (maybe all) goes away when we stub out the mostly stub out
JSView::CreateComponent()

As before, this should be dropped again later on when the problems have been
solved.  And is required in order to get graphics support to work.

Signed-off-by: Esben Haabendal <esben@geanix.com>

Previously, the libinput quirks files were not being installed by the
package. This commit installs them to the
appropriate location.

Signed-off-by: Kareem Zarka <kareem.zarka@huawei.com>

Most .service files had the license file to go with it, but a few where
missing. Adding the remaining ones here.

Signed-off-by: Stefan Schmidt <stefan.schmidt@huawei.com>

We removed the samgr_standard.cfg after conversion to systemd, but
forgot thios extra file.

Signed-off-by: Stefan Schmidt <stefan.schmidt@huawei.com>

Alle these should hopefully not be needed anymore, with the workaround for the
race condition in ace engine.

Signed-off-by: Esben Haabendal <esben@geanix.com>

When starting up abilities, there is a race condition between OnSurfaceChanged
and RunPage, which both are executed as PostTask on the same task executor.
This executor works in FIFO mode, and if (when) the OnSurfaceChanged comes
first, it will fail with a nullptr check on runningPage_.  This is because the
runningPage_ variable is set by the RunPage PostTask.

This patch is obviously just a workaround, and not a proper fix.  Hopefully, 500
ms will be enough in all relevant situations, but at least in theory, the
problem can still appear.

Signed-off-by: Esben Haabendal <esben@geanix.com>

Graphics seems to be basically working now, both on QEMU and rpi4 (plus/minus a
few quirks).  This patch is now causing more problems than it is solving, so
let's revert it for now.

We need to take care that slow rendering is handled properly.  We might have
more test cases failing due to this after dropping this patch.

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

This simplifies initialization, by having param_service starting up with all the
default parameter values instead of setting them manually in different places.

Signed-off-by: Esben Haabendal <esben@geanix.com>

This service doesn't seem to be needed for ACTS compliance, and consumes a lot
of CPU time when activated during some ACTS tests.

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

These are used by ACTS test cases, seemingly as a workaround for problems with
hilogd stalling/hanging.

Signed-off-by: Esben Haabendal <esben@geanix.com>

The services relying on the /dev/dri/card0 should not be started before the
device have been created.

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

With the removal of all services from the openharmony-preinit script, this is
now a proper oneshot service, and can/must be run before starting the actual
services.

The goal is still to get rid if this script completely.

Anything it does that is actually needed should be moved to the appropriate OH
service unit files.

Signed-off-by: Esben Haabendal <esben@geanix.com>

This avoids param service log information to end up spamming the console.

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

In OHOS the init system sets RLIMIT_NICE to 40 for all OpenHarmony services.
Without that, we were getting a lot of kernel spam from the binder driver.

Signed-off-by: Esben Haabendal <esben@geanix.com>

Signed-off-by: Esben Haabendal <esben@geanix.com>

With this in place, we are ready to start converting OpenHarmony system
abilities (started with sa_main) to systemd services using Type=notify.

Signed-off-by: Esben Haabendal <esben@geanix.com>

Convert hilogd into a systemd service

- Patch hilogd to use sd_notify(3) to send ready notification to systemd, and
  thus use as a Type=notify service.
- Add log and logd users and groups, with similar uid and gid as in OHOS.
- Place Unix sockets in /run/openharmony/hilog instead of /dev/unix/socket,
  enabling secure creaton and managing of these.
- Use systemd.socket units for the two Unix sockets, to allow creating with
  desired uid/gid and permissions, and allowing to use hilogd with systemd
  socket activation

Signed-off-by: Esben Haabendal <esben@geanix.com>