GitLab

The ARM architectural timer stops when crust disables OSC24M. This
property causes the kernel to ignore the architectural timer when
calculating sleep time and fall back to reading the RTC.
Signed-off-by: Samuel Holland <samuel@sholland.org>

CONFIG_HID_MULTITOUCH is required for the touchpad on the PineTab
keyboard dock. CONFIG_HIDRAW is not strictly required, but helpful for
both PinePhone and PineTb.

This adjusts sync values according to the datasheet

Newer PineTab may switch to K101-IM2BYL02.
Signed-off-by: Icenowy Zheng <icenowy@aosc.io>

Feixin K101-IM2BYL02 is a new panel by Feixin designed as a replacement
to their K101-IM2BA02 panel. This panel utilizes the Ilitek ILI9881C
controller.

Add this panel's initialzation sequence and timing to ILI9881C driver.
Signed-off-by: Icenowy Zheng <icenowy@aosc.io>

Feixin K101-IM2BYL02 is a drop-in replacement of K101-IM2BA02 panel
(which is already supported by panel-feixin-k101-im2ba02 driver) with
the same pinout. It utilizes an Ilitek ILI9881C controller chip, so its
compatible string should be added to ilitek,ili9881c file.

Add the compatible string for it.
Signed-off-by: Icenowy Zheng <icenowy@aosc.io>
Acked-by: Rob Herring <robh@kernel.org>

There're more panels with ILI9881C controller than the Bananapi one
supported by this driver.

Extract the mode and init sequence part, to prepare the driver for
adding new panels.
Signed-off-by: Icenowy Zheng <icenowy@aosc.io>

Suggested-by: default avatarOndrej Jirman <megous@megous.com>
Signed-off-by: Bhushan Shah's avatarBhushan Shah <bshah@kde.org>

Some outputs from the RTL8723CS are connected to the PL port, which
is at 1.8V. This causes backlight consistency issues.
Signed-off-by: Ondrej Jirman <megous@megous.com>

The runtime power management features in the 8723cs firmware on the
PinePhone don't seem to work correctly. While we expect the chipset to
leave a low power state when it's time to send or receive data, instead
we get a delay of 10-20 seconds before any packets are sent.

sunxi does not support setting ULPI buscontrol, it's used to set PHY to
use external vBus. This stubs out MUSB_ULPI_BUSCONTROL to be able to
handle save/restore events on suspend/resume events.

SECURITY_APPARMOR_BOOTPARAM_VALUE and CONFIG_DEFAULT_SECURITY_APPARMOR
were previously used to set AppArmor as the default LSM. This changed in
0102fb83, now the LSM ordering via CONFIG_LSM sets which major LSM
will be loaded first.

af_socket mediation did not make it into 4.17 so add remaining out
of tree patch

Forward-port of
https://gitlab.com/apparmor/apparmor/-/blob/master/kernel-patches/v4.17/0002-apparmor-af_unix-mediation.patch

Forward-port of
https://gitlab.com/apparmor/apparmor/-/blob/master/kernel-patches/v4.17/0001-apparmor-patch-to-provide-compatibility-with-v2.x-ne.patch

This moves vibrator from "input" to "led", this way we can use LED
triggers. This is what android does and works great.

See:
https://github.com/torvalds/linux/commit/9f6b68774f29692f3e5b87e8eae29da61c3b1171

Squashed commit by @NotKit to get rid of contradicting changes and
make rebasing easier later on.

Signed-off-by: Samuel Holland <samuel@sholland.org>

Enable runtime pm by default so GPU suspend when idle
for 200ms. This value can be changed by
autosuspend_delay_ms in device's power sysfs dir.

On Allwinner H3 lima_device_resume takes ~40us and
lima_device_suspend takes ~20us.
Signed-off-by: Qiang Yu <yuq825@gmail.com>

Add driver pm system and runtime hardware resume/suspend ops.
Note this won't enable runtime pm of the device yet.

v2:
Do clock and power gating when suspend/resume.
Signed-off-by: Qiang Yu <yuq825@gmail.com>

For being used by both device init/fini and suspend/resume.
Signed-off-by: Qiang Yu <yuq825@gmail.com>

For called when PM do resume/suspend.
Signed-off-by: Qiang Yu <yuq825@gmail.com>

Prepare resume/suspend PM.

v2:
Fix lima_pmu_wait_cmd timeout when mali400 case.
Signed-off-by: Qiang Yu <yuq825@gmail.com>

Used for device resume/suspend in the following commits.
Signed-off-by: Qiang Yu <yuq825@gmail.com>

We need to flush TLB anyway before every task start, and the
page directory will be set to empty vm after suspend/resume,
so always set it to the task vm even no ctx switch happens.
Signed-off-by: Qiang Yu <yuq825@gmail.com>

No need to handle this check before calling lima_vm_put.
Signed-off-by: Qiang Yu <yuq825@gmail.com>