So far psci_cpu_(on|off) only worked for CPU1. Allow to control
CPU0 too. This allows to run the Linux PSCI checker successfully:
[ 2.213447] psci_checker: PSCI checker started using 2 CPUs
[ 2.219107] psci_checker: Starting hotplug tests
[ 2.223859] psci_checker: Trying to turn off and on again all CPUs
[ 2.267191] IRQ21 no longer affine to CPU0
[ 2.293266] Retrying again to check for CPU kill
[ 2.302269] CPU0 killed.
[ 2.311648] psci_checker: Trying to turn off and on again group 0 (CPUs 0-1)
[ 2.354354] IRQ21 no longer affine to CPU0
[ 2.383222] Retrying again to check for CPU kill
[ 2.392148] CPU0 killed.
[ 2.398063] psci_checker: Hotplug tests passed OK
[ 2.402910] psci_checker: Starting suspend tests (10 cycles per state)
[ 2.410019] psci_checker: cpuidle not available on CPU 0, ignoring
[ 2.416452] psci_checker: cpuidle not available on CPU 1, ignoring
[ 2.422757] psci_checker: Could not start suspend tests on any CPU
[ 2.429370] psci_checker: PSCI checker completed
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
PSCI 1.0 require PSCI_VERSION, PSCI_FEATURES, AFFINITY_INFO and
CPU_SUSPEND to be implemented. Commit 0ec3d98f76 ("mx7_common:
use psci 1.0 instead of 0.1") marked the i.MX 7 implementation to
be PSCI 1.0 compliant but failed to implement those functions.
Especially the missing PSCI version callback was noticeable when
booting Linux:
[ 0.000000] psci: probing for conduit method from DT.
[ 0.000000] psci: PSCIv65535.65535 detected in firmware.
[ 0.000000] psci: Using standard PSCI v0.2 function IDs
[ 0.000000] psci: MIGRATE_INFO_TYPE not supported.
[ 0.000000] psci: SMC Calling Convention v1.0
This patch provides a minimal implementation thereof. With this
patch applied Linux detects PSCI 1.0:
[ 0.000000] psci: probing for conduit method from DT.
[ 0.000000] psci: PSCIv1.0 detected in firmware.
[ 0.000000] psci: Using standard PSCI v0.2 function IDs
[ 0.000000] psci: MIGRATE_INFO_TYPE not supported.
[ 0.000000] psci: SMC Calling Convention v1.0
Fixes: 0ec3d98f76 ("mx7_common: use psci 1.0 instead of 0.1")
Suggested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
There is no need for assembly in the platform specific part of
the PSCI implementation.
Note that this does not make it a complete PSCI 1.0 implementation
yet but aids to do so in upcoming patches.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
A proper stack is required to safely use C code in psci_arch_cpu_entry.
Fixes: 486daaa618 ("arm: psci: add a weak function psci_arch_cpu_entry")
Cc: Patrick Delaunay <patrick.delaunay@st.com>
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
Acked-by: Patrick DELAUNAY <Patrick.delaunay@st.com>
Tested-by: Patrick DELAUNAY <Patrick.delaunay@st.com>
Enable the support for all SPI flash types.
Signed-off-by: Uri Mashiach <uri.mashiach@compulab.co.il>
Signed-off-by: Yaniv Levinsky <yaniv.levinsky@compulab.co.il>
Fix an invalid usage of the gpr_init function for the imx6ul
architecture
Signed-off-by: Michael Trimarchi <michael@amarulasolutions.com>
Acked-by: Jagan Teki <jagan@amarulasolutions.com>
This function can be used only for some of the nxp SoC. Make
it explicit in the comment. This adjust a bit commit
3aa4b703b4 ("imx: imx6: Move gpr_init() function to soc.c")
Signed-off-by: Michael Trimarchi <michael@amarulasolutions.com>
Acked-by: Jagan Teki <jagan@amarulasolutions.com>
i.MX7 does not support BMODE due to the erratum e10574 ("Watchdog:
A watchdog timeout or software trigger will not reset the SOC"), so
remove its support.
Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
lcdif_power_down should not be included in spl build to avoid build
failure introduced by commit eb111bb31d
Signed-off-by: Michael Trimarchi <michael@amarulasolutions.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Add the missing gpio phy reset binding to the gpio and
reset time configuration
Signed-off-by: Michael Trimarchi <michael@amarulasolutions.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Mainline and now the SolidRun 4.9 nxp based tree use the new
reorganization of device-tree files that separate out the emmc
into its own dtb. u-boot will now look for -emmc in the device
tree name if one is detected.
Signed-off-by: Jon Nettleton <jon@solid-run.com>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Previously we had just made broad assumptions with which of our
boards had an eMMC or not even though this is a manufacturing time
assembly option. This takes the guessing away and actually checks for
the existence of an eMMC and sets up the has_emmc environment variable.
Signed-off-by: Jon Nettleton <jon@solid-run.com>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
The HB2 boards as well as rev 1.5 soms support eMMC
booting as well as SDHC. Add the infrastructure to support
booting these devices.
Signed-off-by: Jon Nettleton <jon@solid-run.com>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
This code is useful for testing the existance of devices that
do not have card detect capabilities. This breaks out the core
functionality and leaves the actual init logic and error reporting
in mmc_start_init().
Signed-off-by: Jon Nettleton <jon@solid-run.com>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Reviewed-by: Stefano Babic <sbabic@denx.de>
The IN_PROGRESS macro has been removed in commit bd47c13583 (mmc: Fix
splitting device initialization). Remove it from the mmc_start_init()
function description.
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
The GPMI NAND IP seems to be the same as used in i.MX 6Quad. Use
the fsl,imx6q-gpmi-nand compatible string like Linux devices trees
are.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
This macro allows to detect whether the boot ROM initialized USB
already (serial downloader). This is helpful to reliably detect
if the system has been recovered via USB serial downloader.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
Acked-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
If the bootcounter address is in a cached memory,
a flush of dcache must occur after updateing the bootcounter.
Issue found on i.MX6 where bootcounter is put into the internal
(cached) IRAM.
Signed-off-by: Stefano Babic <sbabic@denx.de>
Add Kconfig options SPL_ENV_* and TPL_ENV_* and simplify Makefile.
This allows SPL/TPL image has different environment setting from
full feature U-Boot.
Signed-off-by: York Sun <york.sun@nxp.com>
Currently the fdtfile environment variable is set to
CONFIG_DEFAULT_DEVICE_TREE which is іnternally used as U-Boot devicetree
source. The OS can use a different filename and Kconfig gives us the
ability to select a default devicetree via CONFIG_DEFAULT_FDT_FILE.
This also gives user configuring U-Boot via menuconfig the behaviour
someone would expect.
Signed-off-by: Klaus Goger <klaus.goger@theobroma-systems.com>
Tested-By: Vagrant Cascadian <vagrant@debian.org>
Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Acked-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
On the veyron board the vcc33_sd regulator is used as vmmc-supply for
the SD card. This regulator is powered in the MMC core during power on
but its value is never actually set.
In the veyron platform the reset value for the LDO output is 1.8V while
the standard (min and max) value for this regulator defined in the DTS
is 3.3V. When the MMC core enable the regulator without setting its
value, the output is automatically set to 1.8V instead of 3.3V.
With this patch we preemptively set the value to 3.3V.
Signed-off-by: Carlo Caione <carlo@endlessm.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Acked-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
The GRF_SOC_CON0.grf_force_jtag bit is automatically set at boot and it
is preventing the SDMMC to work correctly. Disable the JTAG function on
the assumption that a working SD has higher priority over JTAG.
Signed-off-by: Carlo Caione <carlo@endlessm.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Acked-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Use regulators_enable_boot_on() to init all the regulators with
regulator-boot-on property.
Signed-off-by: Carlo Caione <carlo@endlessm.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Acked-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
The first dhcp command consistently fails with a timeout when
the lion-rk3368 board is connected to a Zyxel GS1100-24E
Gigabit Ethernet switch:
ethernet@ff290000 Waiting for PHY auto negotiation to complete......... TIMEOUT !
Increasing PHY_ANEG_TIMEOUT from the default 4000 to 8000 makes the
first dhcp command work reliably.
Signed-off-by: Jakob Unterwurzacher <jakob.unterwurzacher@theobroma-systems.com>
Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Acked-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Without the patch SPL (in case of an error) creates an output like:
U-Boot SPL board initMissing DTB
The patch adds the missing line feed. So now we get:
U-Boot SPL board init
Missing DTB
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Acked-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Change the description relating to CONFIG_SPL_ROCKCHIP_BACK_TO_BROM to
clarify that both RK3288 and RK3036 use
CONFIG_SPL_ROCKCHIP_BACK_TO_BROM=y
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Acked-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
%s/rkflashtool/rkdeveloptool/
We are using rkdeveloptool not rkflashtool.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Acked-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
The order distroboot searches for a boot.scr is fixed at compile time.
To make BIOS_DISABLE work as expected and boot from mmc1 instead of
mmc0 if enabled, we need to change the environment at runtime.
Especially as commit: 482cf22333 ("rockchip: rk3399-puma: add boot-on
regulator to override BIOS_DISABLE") enables the eMMC in U-Boot even
if BIOS_DISABLE is active.
Signed-off-by: Klaus Goger <klaus.goger@theobroma-systems.com>
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
This implements the new 'spl_perform_fixups' hook for RK3399-based
boards and injects the /chosen/u-boot,spl-boot-device with an ofpath
corresponding to the boot device used.
The intended usage is for the full U-Boot stage to evaluate this in
scripts and then adapt its boot-order when using distro-boot.
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Tested-by: Klaus Goger <klaus.goger@theobroma-systems.com>
To let the full U-Boot know where it was booted from (i.e. which of
the entries in /chosen/u-boot,spl-boot-order' contained a valid
image), we define (and document) /chosen/u-boot,spl-boot-device as the
property that could/should automatically be injected by SPL.
This commit only contains a documentation change, which documents the
new property and the intended usage.
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Tested-by: Klaus Goger <klaus.goger@theobroma-systems.com>
On some boards, we want to give the board/architecture-specific code a
chance to look at where the next image has been loaded from and
perform fixups before starting the next image. This is of particular
importance, when we probe multiple devices for bootable payloads and
boot the first one found.
This change adds the following:
- we record the boot_device used into the spl_image structure
- we provide an extension-point for boards/architectures that can
perform late fixups depending on a fully populated spl_image
structure (i.e. we'll know the final boot_device and have info
on the image type and operating system to be booted).
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Tested-by: Klaus Goger <klaus.goger@theobroma-systems.com>
The SPL code for smartweb is close to its limit and adding a few extra
instructions to SPL will cause it to overrun its sram allotement (thus
causing build failures). To allow adding the 'spl_perform_fixups'
extension point to SPL, we'll enable SPL_TINY_MEMSET for smartweb.
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Series-cc: trini
All other rockchip boards have rk_board_late_init() hook,
so add it to rk3188 boards also.
Signed-off-by: Alexander Kochetkov <al.kochet@gmail.com>
Acked-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
rk3066 and rk3188 has two I2C controller implementations.
Current I2C driver wan't work with legacy implementation.
Switching between controllers is performed using a bit inside
GFR_SOC_CON1 register. The bit setting is performed by pinctrl
driver. The patch ask pinctrl to do settings.
Signed-off-by: Alexander Kochetkov <al.kochet@gmail.com>
Acked-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
[fix warnings by including the rk3228 variant in the compatible-list]:
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Explicitly add 'python' call for 'acs_tool.pyc', to avoid failed
execution on some OSes.
Signed-off-by: Loic Devulder <ldevulder@suse.de>
Acked-by: Neil Armstrong <narmstrong@baylibre.com>
Synchronize the Linux Device Tree for Amlogic Meson GX boards from Linux 4.17.5
(Linux commit 54fb3c180d05e9dfda892a93413514e99f0cbb19).
This will enable HDMI_5V for USB Amlogic Meson GXL P212 based boards.
Signed-off-by: Loic Devulder <ldevulder@suse.de>
This tests that the importing of an environment with a specified
whitelist works as intended.
If there are variables passed as parameter to the env import command,
those only should be imported in the current environment.
For each variable passed as parameter, if
- foo is bar in current env and bar2 in exported env, after importing
exported env, foo shall be bar2,
- foo does not exist in current env and foo is bar2 in exported env,
after importing exported env, foo shall be bar2,
- foo is bar in current env and does not exist in exported env (but is
passed as parameter), after importing exported env, foo shall be empty
ONLY if the -d option is passed to env import, otherwise foo shall be
bar,
Any variable not passed as parameter should be left untouched.
Two other tests are made to test that size cannot be '-' if the checksum
protection is enabled.
Signed-off-by: Quentin Schulz <quentin.schulz@bootlin.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
Tested-by: Stephen Warren <swarren@nvidia.com>
While the `env export` can take as parameters variables to be exported,
`env import` does not have such a mechanism of variable selection.
Let's add the ability to add parameters at the end of the command for
variables to be imported.
Every env variable from the env to be imported passed by parameter to
this command will override the value of the variable in the current env.
If a variable exists in the current env but not in the imported env, if
this variable is passed as a parameter to env import, the variable will
be unset ONLY if the -d option is passed to env import, otherwise the
current value of the variable is kept.
If a variable exists in the imported env, the variable in the current
env will be set to the value of the one from the imported env.
All the remaining variables are left untouched.
As the size parameter of env import is positional but optional, let's
add the possibility to use the sentinel '-' for when we don't want to
give the size parameter (when the env is '\0' terminated) but we pass a
list of variables at the end of the command.
env import addr
env import addr -
env import addr size
env import addr - foo1 foo2
env import addr size foo1 foo2
are all valid.
env import -c addr
env import -c addr -
env import -c addr - foo1 foo2
are all invalid because they don't pass the size parameter required for
checking, while the following are valid.
env import addr size
env import addr size foo1 foo2
Nothing's changed for the other parameters or the overall behaviour.
One of its use case could be to load a secure environment from the
signed U-Boot binary and load only a handful of variables from an
other, unsecure, environment without completely losing control of
U-Boot.
Signed-off-by: Quentin Schulz <quentin.schulz@bootlin.com>
Tested-by: Alex Kiernan <alex.kiernan@gmail.com>
Tested-by: Stephen Warren <swarren@nvidia.com>
When vars are passed to the himport_r function with H_NOCLEAR flag,
those vars will be overridden in the current environment and if one of
those vars is not in the imported environment, it'll be deleted in the
current environment whatever the flag passed to himport_r.
The H_NOCLEAR flag is used to clear the whole environment whether vars
are passed to the function or not.
This leads to incoherent behaviour. If one passes vars to himport_r
with the H_NOCLEAR flag, if a var in vars is not in the imported env,
that var will be removed from the current env.
If one passes vars to himport_r without the H_NOCLEAR flag, the whole
environment will be removed and vars will be imported from the
environment in RAM.
It makes more sense to keep the variable that is in the current
environment but not in the imported environment if the H_NOCLEAR flag is
set and remove only that variable if the H_NOCLEAR flag is not set.
Let's clear the whole environment only if H_NOCLEAR and vars are not
passed to himport_r.
Let's remove variables that are in the current environment but not in
the imported env only if the H_NOCLEAR flag is not passed.
Suggested-by: Wolfgang Denk <wd@denx.de>
Signed-off-by: Quentin Schulz <quentin.schulz@bootlin.com>
Some functions have different behaviour when the given address is 0
(assumed to be NULL by the function).
find_ram_base() does not return 0 anymore so it's safe to remove those
offsets.
Suggested-by: Stephen Warren <swarren@wwwdotorg.org>
Signed-off-by: Quentin Schulz <quentin.schulz@bootlin.com>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
Tested-by: Stephen Warren <swarren@nvidia.com>
Some functions test that the given address is not NULL (0) and fail or
have a different behaviour if that's the case (e.g. hexport_r).
Let's make the RAM base address to be not zero by setting it to 2MiB if
that's the case.
2MiB is chosen because it represents the size of an ARM LPAE/v8 section.
Suggested-by: Stephen Warren <swarren@wwwdotorg.org>
Signed-off-by: Quentin Schulz <quentin.schulz@bootlin.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
Tested-by: Stephen Warren <swarren@nvidia.com>
