Age | Commit message (Collapse) | Author |
|
In order to allow tighter integration with external thermal throttling
code add local temperature and remote critical shutdown temperature
hooks. Rename the existing hooks for clarity.
|
|
Our platform data erroneously used the forth (bus number 3) rather than
the first (bus number 0) controller instance.
|
|
Get rid of obsolete NAND aka Colibri T30 v1.0a prototype handling.
|
|
Add BERR counter functionality to MCP251x driver.
Signed-off-by: Mitja Spes <mitja@lxnav.com>
|
|
As our hardware only allows triggering an interrupt on over-temperature
shutdown we first use it to catch entering throttle limit and only then
set it up to catch an actual over-temperature shutdown.
While throttling we setup a workqueue to catch leaving it again.
We support the same sysfs interface as on NVIDIA Ventana:
root@colibri-t20:~# cat /sys/kernel/debug/thermal/shutdown
115000
root@colibri-t20:~# cat /sys/kernel/debug/thermal/throttle
90000
As follows some output of before and during a throttling run:
root@colibri-t20:~# tegrastats
RAM 159/345MB (lfb 36x4MB) Carveout 11/128MB (lfb 116MB) GART 0/32MB
(lfb 32MB) IRAM 4/255kB(lfb 248kB) cpu [51%,23%]@216 EMC 666 AVP 72 VDE
240 EDP limit -1 Temperatures CPU 48.7 Board 43.2
root@colibri-t20:~# cat /sys/devices/system/cpu/cpu0/cpufreq/throttle
0
root@colibri-t20:~# cat /sys/bus/i2c/devices/4-004c/temp2_os
90000
root@colibri-t20:~# cat /sys/kernel/debug/thermal/shutdown
115000
root@colibri-t20:~# cat /sys/kernel/debug/thermal/throttle
90000
root@colibri-t20:~# tegrastats
RAM 66/345MB (lfb 5x32kB) Carveout 75/128MB (lfb 53MB) GART 8/32MB
(lfb 24MB) IRAM 20/255kB(lfb 232kB) cpu [85%,83%]@912 EMC 666 AVP 216
VDE 240 EDP limit -1 Temperatures CPU 90.1 Board 77.9
root@colibri-t20:~# cat /sys/devices/system/cpu/cpu0/cpufreq/throttle
1
root@colibri-t20:~# cat /sys/bus/i2c/devices/4-004c/temp2_os
115000
|
|
In order to allow integrating with external thermal throttling code add
a probe callback and remote over-temperature shutdown hooks.
|
|
Integrate remote over-temperature shutdown ROS pin handling.
|
|
Clean-up I2C platform data and bring it more in line with Colibri T30.
|
|
In addition to the critical temperature alarm limit this chip features
a second alarm limit called remote over-temperature shutdown ROS with
a separate pin configurable as trigger output. Implement this feature
and add platform data to allow configuring respective pin as output.
|
|
Rather than using hard-coded register indices and offsets define and
use an enumeration thereof.
|
|
Avoid the following issue on both /dev/ttyHS1 as well as /dev/ttyHS3:
[ 4419.708332] tegra_uart tegra_uart.1: Setting clk_src pll_m
[ 4419.714361] Failed to set parent pll_m for uartb (violates clock
limit 600000000)
[ 4427.469124] tegra_uart tegra_uart.3: Setting clk_src pll_m
[ 4427.476177] Failed to set parent pll_m for uartd (violates clock
limit 600000000)
|
|
Integrate spidev support straight from Colibri T20.
|
|
Instead of copying the boot loader frame buffer which currently neither
displays anything worth preserving nor even what it has in the right
resolution make sure both LVDS as well as HDMI frame buffers get
cleared during initialisation to avoid displaying random garbled stuff.
|
|
Add the following tested (e.g. framebuffer console and X) video modes:
- 1280x1024@60
- 1600x1200@60
- 1680x1050@60
- 1920x1080p
- 1920x1200
And the following portrait modes:
- 480x640
- 540x960
- 720x1280
The define TEGRA_FB_VGA in board-colibri_t20.h can be used to switch
between VGA and 800x480.
While at it clean-up some defines and includes.
|
|
Clean-up I2C platform data.
|
|
Validate pin-muxing for latest Colibri T30 v1.1a and v1.1b module
hardware.
|
|
Clean-up kernel configuration and bring it more in line with Colibri T20.
|
|
Add the following tested (e.g. framebuffer console and X) video modes:
- 1280x1024@60
- 1600x1200@60
- 1680x1050@60
- 1920x1080p
- 1920x1200
And the following portrait modes:
- 480x640
- 540x960
- 720x1280
The define TEGRA_FB_VGA in board-colibri_t20.h can be used to switch
between VGA and 800x480.
While at it fix rename of CAMERA_INTERFACE to COLIBRI_T20_VI as well.
|
|
The irq request flag sometimes does not get reset or is asserted
immediately, but iir does not indicated this, if so ISR is entered with
iir set to 0xc1, i.e. no irq pending.
Try enabling the transmit register empty interrupt, iir becomes 0xc2,
irq gets de-asserted and ier is reverted by the regular code flow in
the ISR.
We try this every 4096 spurious irq.
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
|
|
As the new Colibri Evaluation carrier board v3.1a uses SODIMM pin 45 as
IrDA transceiver SD with a 100 K pull-down we revert to using the power
key as high active.
|
|
GPIO keys configured as active-low either require pull-ups on the
carrier board or via pin-muxing. We enable the later for now.
|
|
Use GPIO inputs for function tri-stating rather than pin group
tri-stating to avoid influencing neighbouring pins.
While at it get rid of most pull-up/downs and a lot of commented out
stuff.
|
|
Rename CAMERA_INTERFACE to COLIBRI_T20_VI which stands for video input
to be more in-line with Colibri T30.
|
|
Get rid of optional CAM/GEN3_I2C on SODIMM pin 127/133.
|
|
Verify standard GPIO usage for latest v1.2a modules.
|
|
The Iris carrier board by default uses SODIMM pin 73 as DAC power save.
Note: This conflicts with the CAN interrupt on the new EvalBoard v3.1a.
|
|
VESA VGA mode defines hsync/vsync to be of polarity low rather than
high.
|
|
The HDMI power rail configuration used to be done within the LVDS aka
TFTLCD panel enable function. Now moving it to the platform power
initialisation where it really belongs. This avoids any
duplicate/spurious regulator configuration.
|
|
As the deep-sleep LP0 suspend mode is currently anyway not supported
default to the LP1 suspend mode for now.
|
|
The full PMIC power off decreases power consumption but doing a
subsequent reset won't force a re-boot any more.
|
|
Clean-up the following unused regulator supplies in the platform power
stuff:
p_cam_avdd, avdd_lvds, vddio_sys (panjit_touch), vmmc (sdhci-tegra.3)
and avdd_2v85.
While at it add some more comments concerning specific consumers/rails.
|
|
Instead of copying the boot loader frame buffer which currently neither
displays anything worth preserving nor even what it has in the right
resolution make sure both LVDS as well as HDMI frame buffers get
cleared during initialisation to avoid displaying random garbled stuff.
|
|
The boot logo and framebuffer console where displayed with a wrong
stride caused by a colour depth mismatch. Fix this by initially
configuring the display controllers to 16 bpp.
|
|
There are some classes of devices that are controlled entirely from
user space. Userspace consumer driver provides ability to control power
supplies for such devices. Enable this configuration option to be more
in line with the T30.
|
|
If the console is locked while the system is rebooted, the messages in
the temporary log buffer would not have propagated to all the console
drivers. This option forces the console lock to be released if it
failed to be acquired, which will cause all the pending messages to be
flushed. Enable this configuration option to be more in line with the
T30.
|
|
Get rid of the 2G/2G user/kernel memory split configuration in order to
use the default 3G/1G user/kernel memory split. This is more in line
with the T30 and avoids issues trying to run Android.
|
|
Revise Ethernet MAC address assignment: should now use MAC from EEPROM
if set and further handles up to two instances of custom user MACs (2nd
one with a 0x100000 offset). This way customer does not have to put
EEPROM on a secondary Ethernet on the carrier board and still gets a
valid official MAC address from us (e.g. analogous to how we did it on
our Protea carrier board).
|
|
Fix spurious whitespace in Tegra's UDC driver log message.
|
|
Fix two issues concerning SDMMC4B pinmux (e.g. using a regular SD card
as the boot device instead of the on-module eMMC). First according to
the TRM it is illegal to mux a certain function to two sets of balls
(e.g. the regular SDMMC4 as well as the SDMMC4B muxing). Second make
absolutely sure all unused balls are not only tri-stating their outputs
but also disable their inputs to prevent any illegal internal states.
|
|
Make sure NC balls are not only tri-stating their outputs but also do
not drive their inputs but rather have them pulled down in order to
prevent any unstable internal values.
While at it fix alphabetical order of the pinmuxes.
|
|
This fixes the following build issue:
arch/arm/mach-tegra/board-colibri_t30.c:276:41: error:
'colibri_t30_emmc_platform_data' defined but not used
[-Werror=unused-variable]
cc1: all warnings being treated as errors
|
|
Verify standard GPIO usage for latest v1.1a samples.
While at it get rid of some commented out obsolete stuff.
|
|
Integrate proper Ethernet initialisation analogous to how it is done
for the Colibri T20.
|
|
Add a commented out define for v1.0a modules with optional NAND
variants which are auto detected.
|
|
As we don't want to exclusively reserve the detection GPIO temporarily
enabling it is still required.
|
|
Hack to force WIRELESS_EXT required to build Redpine Signals LiteFi driver.
|
|
Enable kernel configuration for USB audio.
|
|
As Colibris are usually self powered default drawn VBUS to 2 mA rather
than demanding 500 mA. This should fix it running from not self powered
USB hubs as well.
|
|
Activate Android rather than mass storage gadget driver to support
multiple USB functions. The functions can be configured via a board
file and may be enabled and disabled dynamically. E.g. the following
enables regular Ethernet aka RNDIS functionality once booted:
echo rndis > /sys/class/android_usb/android0/functions
echo 1 > /sys/class/android_usb/android0/enable
|
|
Activate Colibri rather than Tegra USB OTG driver as used on
Colibri T20 as well.
|