platform: tegra: rtcpu: refactor pm and probing

Firmware provides now the PM features for suspend and resume. Firmware
activates the WDT after boot synchronization and deactivates it when
suspended.  The boot synchronization and the IVC service
synchronization can be redone at any time.

The driver requires firmware version SM4 because the earlier firmwares
can not resume gracefully from the suspended state.

Reorder initialization in order to avoid accessing hardware before it
is fully powered. Note that powering HW with the APE power domain
during the probe() may happen when pm_runtime_get_sync() is called and
both resume and runtime_resume may get invoked in the process. Add
checks to avoid any premature resumes.

Assert resets in the reverse order they are deasserted. Enable clocks
before touching any resources, including getting interrupt resources.

Fix the unbalanced early_resume/late_suspend.

The power domain does not track the last busy from the children of the
rtcpu device.  Mark the rtcpu device last busy always when the ivc
channel devices are used, too.

The IVC channels are now created only after the initial boot sync and
version exchange with the firmware. When the IVC sync is done the
RTCPU and device are both up and ready.

Do not expose boot and halt methods. Instead provide the reboot
method, which will only proceed when the rtcpu device is in the active
runtime state.

Follow suspend procedure from APE TRM. Wait for APE WFI using an
interrupt. Wait for APE L2 cache to idle. Allocate an interrupt for
APE WFI.

Move all register handling to core-specific functions (tegra_ape_cam*(),
tegra_sce_cam*()).

Do not use the ahub clock.

Power off RTCPU and remove the tracer during remove. Tracer can be
disabled or enabled in the device tree. Treat it as an optional feature.

Jira CRTC-910
Bug 1857548
Jira CRTC-966
Bug 1864505
Bug 200269632
Bug 200292354

Change-Id: I75df878932a4307181a243261f99f0a853d395b1
Signed-off-by: Pekka Pessi <ppessi@nvidia.com>
Reviewed-on: http://git-master/r/1298769
Reviewed-by: David Wang (SW-TEGRA) <davidw@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Mika Liljeberg <mliljeberg@nvidia.com>

1 files changed, 2 insertions, 2 deletions

diff --git a/include/linux/tegra-camera-rtcpu.h b/include/linux/tegra-camera-rtcpu.h
index 73eab094b..76b63ade2 100644
--- a/include/linux/tegra-camera-rtcpu.h
+++ b/include/linux/tegra-camera-rtcpu.h
@@ -20,10 +20,10 @@ struct device;
 
 int tegra_camrtc_iovm_setup(struct device *dev, dma_addr_t iova);
 ssize_t tegra_camrtc_print_version(struct device *dev, char *buf, size_t size);
-int tegra_camrtc_boot(struct device *dev);
-int tegra_camrtc_halt(struct device *dev);
+int tegra_camrtc_reboot(struct device *dev);
 int tegra_camrtc_restore(struct device *dev);
 bool tegra_camrtc_is_rtcpu_alive(struct device *dev);
+
 #define TEGRA_CAMRTC_VERSION_LEN 128
 
 int tegra_camrtc_command(struct device *dev, u32 command, long timeout);