Users experiencing random kernel crashes due to new versions of
Marvell's DDR training algorithm can solve the issue by setting DDR
speed to 1333H.
But if kernel crashes, it has to be done in U-Boot, which is impossible
without UART connection.
In order to make it easier for users, use the rescue button mechanism:
when rescue mode 9 is selected (that is when 10 LEDs are ON), U-Boot
will train DDR in 1333H mode and also update EEPROM so that subsequent
boot will use this mode.
User has to use the `eeprom` command in U-Boot or `omnia-eeprom` command
in OS to switch back to 1600K mode.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Use the i2c_eeprom miscellaneous driver for reading Turris Omnia EEPROM
in U-Boot proper. Keep using dm_i2c_read() in SPL build, since adding
the i2c_eeprom driver to SPL build increases the image by 1.5 KiB.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Rename the variable holding the EEPROM udevice from `chip` to `eeprom`.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Support old DDR3 training code on Turris Omnia, selectable via EEPROM
field.
Users experiencing DDR3 initialization failures or random crashes of the
operating system due to incorrect DDR3 configuration can select the old
DDR3 training implementation to fix those issues by setting the EEPROM
field "Use old DDR training":
eeprom update "Use old DDR training" 1
Signed-off-by: Marek Behún <kabel@kernel.org>
Add optional support for using old DDR3 training code from 2017.
The code lives in drivers/ddr/marvell/a38x/old/. To prevent symbol
clashing with new DDR3 training code, a special header which renames all
clashing symbols via macros is included and the symbols are prefixed
with 'old_'.
If old DDR3 training support is selected for a board, then the SPL
initialization code calls a new function
board_use_old_ddr3_training()
to check whether it should use old DDR3 training code. The default
weak implementation returns false, defaulting to new DDR3 training code.
Boards that wish to support this need to select the
ARMADA_38X_SUPPORT_OLD_DDR3_TRAINING
config option and implement the old version of DDR topology provider,
ddr3_get_topology_map().
Signed-off-by: Marek Behún <kabel@kernel.org>
Import DDR training code from commit 1b69ce2fc0 ("arm: mvebu:
ddr3_debug: remove self assignments") into
drivers/ddr/marvell/a38x/old/. The code is not used yet.
Explanation:
Since 2019, on some Turris Omnia boards we have been having problems
with newer versions of Marvell's DDR3 training code for Armada 38x,
which is ported from mv-ddr-marvell [1] to U-Boot into the
drivers/ddr/marvell/a38x/ directory:
- sometimes the DDR3 training fails on some older boards, sometime it
fails on some newer boards
- other times it succeeds, but some boards experience crashes of the
operating system after running for some time.
Using the stock version of Turris Omnia's U-Boot from solved these
issues, but this solution was not satisfactory, since we wanted
features from new U-Boot.
Back in 2020-2022 we have spent several months trying to debug the
issues, working with Marvell, on our own, and also with U-Boot
community, but these issues persist still.
One solution we used back in 2019 was a "hybrid U-Boot": the SPL part
(containing the DDR3 training code) was taken from the stock version,
while the proper part was current U-Boot at the time. This solution also
has its drawbacks, of which the main one is the need to glue binaries
from two separate builds.
Since then there have been some more changes to the DDR3 training code
in upstream mv-ddr-marvell that have been ported to U-Boot. We have
provided our users experimental builds of U-Boot in the TurrisOS so that
they could try upgrading the firmware and let us know if those problems
still exist. And they do.
We do not have the time nor manpower to debug this problem and fix it
properly. Marvell was also no able to provide a solution to this,
probably because they do not have the manpower as well.
I have therefore come up with this "not that pretty" solution: take the
DDR3 training code from an older version of U-Boot that is known to
work, put it into current U-Boot under old/ subdirectory within
drivers/ddr/marvell/a38x/, build into the SPL binary both the old and
new versions and make it possible to select the old version via an env
variable.
[1] https://github.com/MarvellEmbeddedProcessors/mv-ddr-marvell
Signed-off-by: Marek Behún <kabel@kernel.org>
Some Turris Omnia boards experience memory issues, and by
experimentation we found that some of these issues can be solved by
slowing DDR speed.
Add a new field in the extended EEPROM information structure, ddr_speed.
Support several values in this field (for now 1066F, 1333H, and the
default, 1600K) and use it to overwrite the DDR topology parameters
used by the DDR training algorithm.
Signed-off-by: Marek Behún <kabel@kernel.org>
Enable the 'eeprom' command with support for EEPROM layout for Turris
Omnia. Enable the i2c-eeprom driver so that the EEPROM is accessed via
driver model.
Signed-off-by: Marek Behún <kabel@kernel.org>
Implement Turris Omnia EEPROM layout for the 'eeprom' command.
When the 'eeprom' command (with layout support) is enabled, we can now
use the 'eeprom print' and 'eeprom update' commands, for example:
=> eeprom print
Magic constant 34a04103
RAM size in GB 2
Wi-Fi Region
CRC32 checksum cecbc2a1
Signed-off-by: Marek Behún <kabel@kernel.org>
For board revisions where the WAN ethernet PHY reset GPIO is controllable
via MCU we currently insert a phy-reset-gpios property into the ethernet
controller node. The mvneta driver parses this property and uses the
GPIO to reset the PHY.
But this phy-reset-gpios property is not a valid DT binding in upstream
kernel. Instead, a reset-gpios property should be inserted into the
ethernet PHY node. This correct DT binding is supported by the DM ETH PHY
U-Boot driver.
Insert the reset-gpios property into the WAN PHY node instead the
phy-reset-gpios property in WAN ETH node so that Linux will correctly use
the reset GPIO.
Enable the CONFIG_DM_ETH_PHY config option so that U-Boot will also use
the correct DT property.
Note: currently there are 4 ethernet controller drivers parsing the
wrong DT property: dwc_eth_qos, fex_mxc, mvneta and mvpp2. We should
convert all relevant device-trees to use reset-gpios so that we can get
rid of these drivers parsing this property.
Fixes: 1da53ae26a ("arm: mvebu: turris_omnia: Add support for design with SW reset signals")
Signed-off-by: Marek Behún <kabel@kernel.org>
Allow compiling with immutable debug settings:
- DEBUG_LEVEL is always set to DEBUG_LEVEL_ERROR
- register dumps are disabled
This can save around 10 KiB of space in the resulting binary, which is a
lot in U-Boot SPL.
Signed-off-by: Marek Behún <kabel@kernel.org>
The variables is_validate_window_per_if, is_validate_window_per_pup,
sweep_cnt and is_run_leveling_sweep_tests are only used if
DDR_VIEWER_TOOL macro is defined, so define them only in that case.
Make them static since they are only used in ddr3_debug.c.
Signed-off-by: Marek Behún <kabel@kernel.org>
Return from ddr3_tip_print_log() early if we won't print anything
anyway.
This way the compiler can optimize away the VALIDATE_IF_ACTIVE() calls
in the for-loop, so if the SILENT_LIB macro is defined, no code is
generated for the rest of the function, which saves some space.
Signed-off-by: Marek Behún <kabel@kernel.org>
Turris Omnia defconfig is nearing image size limit. Disable ext4 write
support to reserve space for more important stuff.
This reduces the size of the KWB image by ~19 KiB.
If in the future U-Boot supports compressing itself and decompressing
on load, we may enable this again.
Signed-off-by: Marek Behún <kabel@kernel.org>
Extend the 'eeprom' command to allow accessing EEPROMs probed via
driver model, uclass UCLASS_I2C_EEPROM.
When the CONFIG_I2C_EEPROM config option is enabled (and so the
i2c-eeprom driver is built), the 'eeprom' command now accepts driver
model device name as EEPROM specifier for the 'eeprom' command, in
addition to the legacy [[bus] devaddr] specifier.
Moreover if no device specifier is given, then the first
UCLASS_I2C_EEPROM device is used, if found.
Signed-off-by: Marek Behún <kabel@kernel.org>
Don't read/write whole EEPROM if not necessary when printing / updating
EEPROM layout fields. Only read/write layout.data_size bytes.
Signed-off-by: Marek Behún <kabel@kernel.org>
Refactor the eeprom_execute_command() function into separate functions
do_eeprom_rw(), do_eeprom_print() and do_eeprom_update().
Signed-off-by: Marek Behún <kabel@kernel.org>
In preparation for allowing to access eeprom by driver-model device
name, refactor the eeprom device specifier parsing. Instead of filling
two parameters (i2c_bus, i2c_addr), the parsing function now fills one
parameter of type struct eeprom_dev_spec.
Signed-off-by: Marek Behún <kabel@kernel.org>
Add a new Kconfig option EEPROM_LAYOUT_VERSIONS, and hide eeprom
layout versionsing code behind it. Only print the relevant help in
'eeprom' command usage if this option is enabled.
Enable this new option for cm_fx6_defconfig and cm_t43_defconfig.
These are the only boards using EEPROM layout versioning.
Signed-off-by: Marek Behún <kabel@kernel.org>
The bus and devaddr arguments of the eeprom command are optional, and if
only one is given, it is assumed to be devaddr. Change the usage help
from
<bus> <devaddr>
to
[[bus] [devaddr]
Signed-off-by: Marek Behún <kabel@kernel.org>
The byte variable is of type unsigned char, it is never less than zero.
The error case is handled by *endptr, so drop the comparison altogether.
Signed-off-by: Marek Behún <kabel@kernel.org>
The __eeprom_field_update_bin() function is expected to parse a hex
string into bytes (potentially in reverse order), but the
simple_strtoul() function is given 0 as base. This does not work since
the string does not contain '0x' prefix. Add explicit base 16.
Signed-off-by: Marek Behún <kabel@kernel.org>
Split the eeprom layout field finding code from the
eeprom_layout_update_field() function in order to make it usable in
alternative implementations of update method.
Signed-off-by: Marek Behún <kabel@kernel.org>
Assign the default eeprom layout parameter .data_size and methods
.print() and .update() before calling eeprom_layout_assign() in
eeprom_layout_setup().
This allows eeprom_layout_assign() to overwrite these if needed.
Signed-off-by: Marek Behún <kabel@kernel.org>
Now that Turris Omnia has a rng driver provided in the MCU driver,
enable the rng command in defconfig.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Add support for true random number generator provided by the MCU on
Turris Omnia. The MCU firmware supports TRNG if the FEAT_TRNG bit is set
in features. In that case we bind the rng driver.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
In the rng command, print
Abort
instead of
Reading RNG failed
if the error number is -EINTR, which can happen if the user pressed
CTRL-C.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Enable support for the poweroff command via sysreset for Turris Omnia.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Add support for system power off via UCLASS_SYSRESET. Newer versions of
Turris Omnia MCU firmware can power off the board (MCU will disable
almost all voltage regulators and go into low power mode).
Move the MCU driver into drivers/misc and register it under UCLASS_MISC.
The sysreset and gpio device are bound as child devices of the MCU device.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Update firmware features reading to try reading 32 bits of features and
fallback to reading 16 bits.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Use byteorder conversion function instead of manually assembling data
from/to MCU.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Add driver model support for sysreset via mvebu system controller. This is
currently only available for U-Boot proper.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
The mvebu-reset driver is only needed by the mvebu PCIe driver, but
currently it is automatically selected if DM_RESET is enabled. Add the
condition of PCI_MVEBU also being enabled for mvebu-reset to be
selected.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
The system-controller driver for 32-bit Armada is currently registered
as UCLASS_RESET, since it only provides enabling/disabling PCIe ports.
Rework it as UCLASS_SYSCON and bind mvebu-reset as a child device.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Commit 35e29e89a3 ("arm: mvebu: Implement simple mvebu-reset driver
for enabling/disabling PCIe ports") made it so that the mvebu reset
driver for enabling/disabling PCIe ports is build if CONFIG_DM_RESET is
enabled. This is because PCI_MVEBU depends on DM_RESET.
But the driver should not be built for SPL. Indeed the PCI_MVEBU driver
is not supported in SPL now, and so the mvebu-reset driver is not needed.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
If the FEAT_CRYPTO feature bit is present in MCU features, the board
crypto is implemented by MCU and the Atmel SHA chip is not present.
Disable Atmel SHA device-tree node in that case.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
If MCU supports the FEAT_CRYPTO feature, read board ECDSA public key
from MCU and print it.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Implement reading board serial number, first MAC address and board
version from MCU. MCU supports board information if the FEAT_BOARD_INFO
feature bit is set in MCU features.
Prefer getting board information from MCU if supported, fallback to
Atmel SHA chip.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Refactor MCU status word and MCU firmware features reading to make it
simpler to use.
Try reading 32 bits of features, if that fails, read 16 bits. Older MCU
firmware supports only 16-bit wide features, and if more bytes are read,
either 0xff is sent or I2C transaction fails. Handle both cases.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Since commit 8cd4bf7dc9 ("turris: Use checkboard() instead of
show_board_info()") the model is show two times during boot:
Model: Turris Omnia
Model: Turris Omnia
This is because the common function show_board_info() shows it, and
Turris' checkboard() also does.
Remove the second print.
Fixes: 8cd4bf7dc9 ("turris: Use checkboard() instead of show_board_info()")
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Add header containing all MCU command definitions and use it in board
code and in MCU driver.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
U-Boot builds for Turris Omnia are approaching the limit of 0xf0000
bytes, which is the size of the U-Boot partition on Omnia.
Enable LTO to get more size optimized binaries.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Extend Turris Mox board code to support CZ.NIC's RIPE Atlas Probe.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
I messed up application of patch 5a428e7510 ("mmc: fsl_esdhc_spl: Add
support for builds without CONFIG_SYS_MMC_U_BOOT_OFFS"). I took it from
a work-in-progress branch where I changed usage of
CONFIG_SDCARD to CONFIG_SD_BOOT
and refactored
SYS_MPC85XX_NO_RESETVEC
mess.
But these changes aren't in master yet. Fix the wrong usage of these
macros.
Fixes: 5a428e7510 ("mmc: fsl_esdhc_spl: Add support for builds without
CONFIG_SYS_MMC_U_BOOT_OFFS")
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Pali Rohár <pali@kernel.org>
It seems that waiting only 10 ms after releasing LAN switch from reset
is not enough for the strapping pins to latch the requested values.
P6_MODE[0] is latched to 0 instead of 1.
Increasing the delay to 50 ms fixes this issue.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Commit d433c74eec ("Convert CONFIG_SDCARD et al to Kconfig") converted
SYS_EXTRA_OPTIONS=SDCARD or SPIFLASH to config options CONFIG_SDCARD and
CONFIG_SPIFLASH, but left one occurance unchanged.
Fix this.
Fixes: d433c74eec ("Convert CONFIG_SDCARD et al to Kconfig")
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Pali Rohár <pali@kernel.org>
Some new Omnia boards will come with Winbond SPI flash. Add to
defconfig.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
After a discussion with Tom Rini, we've agreed that I am going to take
over custodianship of the MPC85XX platform, since it seems other people
do not have necessary interest or time and getting things done over
there takes too long.
Since I am only working on one MPC85XX board, Turris 1.x, and do not
have time to do thorough reviews of patches for this entire platform
(other than those concerning Turris 1.x board), for other boards I will
only run patches through CI and checkpatch, and then send them via PR
upwards to Tom.
Signed-off-by: Marek Behún <kabel@kernel.org>
Acked-by: Tom Rini <trini@konsulko.com>
There is a bug in Turris Omnia's schematics, whereupon the MPP[26] pin,
which is routed to CN11 pin header, is documented as SPI CS1, but
MPP[26] pin does not support this function. Instead it controls chip
select 2 if in "spi0" mode.
Fix the name of the pin node in pinctrl node and fix the comment in SPI
node.
Signed-off-by: Marek Behún <kabel@kernel.org>
Fix diacritics in some instances of my name and change my e-mail address
to kabel@kernel.org.
Add corresponding .mailmap entries.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Fix MAINTAINERS files for Turris devices, add missing files and add Pali
as maintainer.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Macro MVNETA_GMAC_FORCE_LINK_UP can be dropped from value assignment in
fixed link case, since it's value is written into the register later in
the function for link-down-to-link-up case. The value is written as
MVNETA_GMAC_FORCE_LINK_DOWN | MVNETA_GMAC_FORCE_LINK_PASS, and so the
macro definition can also be dropped.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Guard the code handling the fixed PHY case by
CONFIG_IS_ENABLED(PHY_FIXED).
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Since this member is checked only at two places drop it and inline it's
usage.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Write PHY address just before enabling HW polling of the PHY.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Stop parsing fixed-link in the MAC driver. Instead support only PHY mode
and let the fixed PHY driver handle the fixed-link case.
Enable CONFIG_PHY_FIXED for mvneta boards that need it: Turris Omnia and
ESPRESSObin.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
The CONFIG_* macros are reserved for Kconfig. This was probably done
when this driver was being imported from Linux. Rename the macro.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Drop checking for CONFIG_PHYLIB in mvneta, this is already done in
Kconfig.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Use the bool type instead of int for status_change variable.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Drop one indentation level in the mvneta_adjust_link() function.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
The `ret` variable in mvneta_probe() is unused if DM_GPIO is disabled.
Since the variable is used only once after assigning value, we can
inline the usage and drop the variable.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
This MDIO bus is now handled by a proper mvmdio DM driver. Remove it
from mvneta.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
In order to be able to get rid of the non-DM MDIO bus registered in
mvneta driver, start using the DM registered one in Turris MOX board
code.
This also allows us to drop the hack introduced in MOX' -u-boot.dtsi
file.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Add helper to resolve PHY node from it's ofnode via DM MDIO subsystem.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
In order to be able to get rid of the non-DM MDIO bus registered in
mvneta driver, we need to stop using board_network_enable() and instead
use the DM registered MDIO device to configure switch in
last_stage_init().
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
We don't need to remember PHY address anymore, because since using DM
MDIO for connecting PHY, the address is parsed by mdio-uclass from
the ofnode.
But the driver uses a special value of the address to signal fixed link
usage.
Drop phyaddr add fixed_link in driver private structure. This simplifies
code a little.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Use the modern DM MDIO API for connecting PHY in the mvneta driver.
This requires enabling MVMDIO driver in several config files.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
In mvneta_adjust_link() we need to set MII_SPEED bit only if PHY reports
the speed at 100Mbps.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Drop .of_to_plat() from the mvneta driver and parse the two properties
in .probe().
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Change type of private struct member mdio_base from void * to
phys_addr_t. This allows us to drop 2 casts.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Currently we require PHY interface mode to be known when
finding/creating the PHY - the functions
* phy_connect_phy_id()
* phy_device_create()
* create_phy_by_mask()
* search_for_existing_phy()
* get_phy_device_by_mask()
* phy_find_by_mask()
all require the interface parameter, but the only thing done with it is
that it is assigned to phydev->interface.
This makes it impossible to find a PHY device without overwriting the
set mode.
Since the interface mode is not used during .probe() and should be used
at first in .config(), drop the interface parameter from these
functions. Make the default value of phydev->interface (in
phy_device_create()) to be PHY_INTERFACE_MODE_NA. Move the interface
parameter to phy_connect_dev(), where it should be.
Change all occurrences treewide. In occurrences where we don't call
phy_connect_dev() for some reason (they only configure the PHY without
connecting it to an ethernet controller), set
phydev->interface = value from phy_find_by_mask call.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
The bcmgenet and sun8i_emac drivers call phy_connect(), which finds /
creates the PHY and also connects it to the eth device via
phy_connect_dev(), then set some phydev members (bcmgenet only), and
then call phy_connect_dev() explicitly again.
Drop the second phy_connect_dev(), since it is unnecesary.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Use phydev->is_c45 instead of is_10g_interface(phydev->interface) to
determine whether clause 45 protocol should be used.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
We want to be able to have phydev->interface uninitialized during
->probe(). We should assume that phydev->interface is initialized only
before ->config().
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Move PHY_INTERFACE_MODE_NA to the beginning of the enum definition to
make it have zero value. This makes it possible (although not
encouraged) to test for invalid/nonexistent interface mode with !val
instead of val == PHY_INTERFACE_MODE_NA.
The comment near the definition says "Must be last", because when the
constant was introduced in commit 5f184715ec ("Create PHY Lib for
U-Boot"), it was used as the maximum value when interating over the
constants. But this is no longer true - we use PHY_INTERFACE_MODE_MAX
for that now, and so we can move it.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Rename constant PHY_INTERFACE_MODE_NONE to PHY_INTERFACE_MODE_NA to make
it compatible with Linux' naming.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Rename constant PHY_INTERFACE_MODE_COUNT to PHY_INTERFACE_MODE_MAX to
make it compatible with Linux' naming.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Add helpers ofnode_read_phy_mode() and dev_read_phy_mode() to parse the
"phy-mode" / "phy-connection-type" property. Add corresponding UT test.
Use them treewide.
This allows us to inline the phy_get_interface_by_name() into
ofnode_read_phy_mode(), since the former is not used anymore.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Tested-by: Patrice Chotard <patrice.chotard@foss.st.com>
Use the new dm_mdio_read/write/reset() wrappers treewide, instead of
always getting and dereferencing MDIO operations structure pointer.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Add wrappers dm_mdio_read(), dm_mdio_write() and dm_mdio_reset() for
DM MDIO's .read(), .write() and .reset() operations.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Add helpers ofnode_get_phy_node() and dev_get_phy_node() and use it in
net/mdio-uclass.c function dm_eth_connect_phy_handle(). Also add
corresponding UT test.
This is useful because other part's of U-Boot may want to get PHY ofnode
without connecting a PHY.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Use the ARRAY_SIZE() macro instead of hardcoding sizes of arrays in
macros.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Commit 0934dddc64 ("arm: a37xx: Update DTS files to version from
upstream Linux kernel") ported Linux's device-tree files for Armada 3720
SOCs. This broke USB port on Turris MOX, because in Linux' DTS the bus
voltage supply is described as a `phy-supply` property of connector
node, a mechanism that is not supported in U-Boot yet.
For now, fix this by adding `vbus-supply` to usb3 node.
Fixes: 0934dddc64 ("arm: a37xx: Update DTS files to version from upstream Linux kernel")
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Commit 0934dddc64 ("arm: a37xx: Update DTS files to version from
upstream Linux kernel") ported Linux's device-tree files for Armada 3720
SOCs. This broke network on Turris MOX, because the SOC's MDIO bus in
U-Boot currently isn't probed via DM as it's own device, but is
registered as part of mvneta's driver, which means that pinctrl
definitions are not parsed for the MDIO bus node. Also mvneta driver
does not consider "phy-handle" property, only "phy".
For now, fix this by adding armada-3720-turris-mox-u-boot.dtsi file
returning the MDIO to how it was defined previously.
A better solution (using proper mvmdio DM driver) is being work on, but
will need testing on various boards, and we need the bug fixed now for
the upcoming release.
Fixes: 0934dddc64 ("arm: a37xx: Update DTS files to version from upstream Linux kernel")
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Enable ext4 write support in Turris Omnia's defconfig. Some users find
it useful.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
The state of the current DDR training code for Armada 38x is such that
we cannot be sure it will always train successfully - although after the
last change we were yet unable to find a board that failed DDR training,
from experience in the last 2 years we know that it is possible.
The experience also tells us that in many cases the board fails training
only sometimes, and after a reset the training is successful.
Enable the new option that makes the board reset itself on DDR training
failure immediately. Until now we called hang() in such a case, which
meant that the board was reset by the MCU after 120 seconds.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Reviewed-by: Pali Rohár <pali@kernel.org>
Some boards may occacionally fail DDR training. Currently we hang() in
this case. Add an option that makes the board do an immediate reset in
such a case, so that a new training is tried as soon as possible,
instead of hanging and possibly waiting for watchdog to reset the board.
(If the DDR training fails while booting the image via UART, we will
still hang - it doesn't make sense to reset in such a case, because
after reset the board will try booting from another medium, and the
UART booting utility does not expect that.)
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
In commit 3fc92a215b ("ddr: marvell: a38x: fix SPLIT_OUT_MIX state
decision") I ported a cleaned up and changed version of patch
mv_ddr: a380: fix SPLIT_OUT_MIX state decision
In the port we removed checking for BYTE_HOMOGENEOUS_SPLIT_OUT bit,
because:
- the fix seemed to work without it
- the bit was checked for only at one place out of two, while the second
bit, BYTE_SPLIT_OUT_MIX, was checked for in both cases
- without the removal it didn't work on Allied Telesis' x530 board
We recently had a chance to test on more boards, and it seems that the
change needs to be opposite: instead of removing the check for
BYTE_HOMOGENEOUS_SPLIT_OUT from the first if() statement, the check
needs to be added also to the second one - it needs to be at both
places.
With this change all the Turris Omnia boards I have had available to
test seem to work, I didn't encounter not even one failed DDR training.
As last time, I am noting that I do not understand what this code is
actually doing, I haven't studied the DDR training algorithm and
I suspect that no one will be able to explain it to U-Boot contributors,
so we are left with this blind poking in the code with testing whether
it works on several boards and hoping it doesn't break anything for
anyone :-(.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Tested-by: Chris Packham <judge.packham@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
