Sebastian Reichel authored 2 weeks ago and Sebastian Reichel committed 1 week ago

The ROCK 5B and ROCK 5B+ are quite similar, but there are also a
few changes, which require a different device tree. Most of the
changes are not easily detectable at runtime, but it is possible
to tell the boards apart by checking the memory type. All ROCK
5B models use LPDDR4 and all ROCK 5B+ models use LPDDR5.

Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>

Sebastian Reichel authored 2 weeks ago and Sebastian Reichel committed 1 week ago

This is a simple copy of the latest RK3588 DT from upstream
Linux including the ROCK 5B+ devicetree, which I'm currently
working on. The goal is a combined U-Boot image, which works
on the ROCK 5B and the ROCK 5B+ and chooses the correct DT
file at runtime.

Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>

Maksim Kiselev authored 5 months ago and Sebastian Reichel committed 1 week ago

PU\PD resistors on CC lines must be configured before running the TCPM
state machine.

Also, when the Type-C port acts as a host (SRC), the VBUS sould be enabled
only after the toggling has been completed. And we have to wait for
the corresponding IRQ to finish the toggling process. But this doesn't
happen because we exit from tcpm_poll_event() if VBUS != OK.

So, Let's check for VBUS state only when the port acts as a host (SRC)
to solve this problem.

Signed-off-by: Maksim Kiselev <bigunclemax@gmail.com>
Link: https://lore.kernel.org/r/20241101103501.21432-4-bigunclemax@gmail.com


Signed-off-by: Sebastian Reichel <sre@ring0.de>

Maksim Kiselev authored 5 months ago and Sebastian Reichel committed 1 week ago

Add support for VBUS supply regulator.

When our type-c port acts as a host(SRC), this regulator
used for control VBUS supply.

Signed-off-by: Maksim Kiselev <bigunclemax@gmail.com>
Link: https://lore.kernel.org/r/20241101103501.21432-3-bigunclemax@gmail.com


Signed-off-by: Sebastian Reichel <sre@ring0.de>

Maksim Kiselev authored 5 months ago and Sebastian Reichel committed 1 week ago

Add newline character in dev_dbg end.

Signed-off-by: Maksim Kiselev <bigunclemax@gmail.com>
Acked-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Link: https://lore.kernel.org/r/20241101103501.21432-2-bigunclemax@gmail.com


Signed-off-by: Sebastian Reichel <sre@ring0.de>

Sebastian Reichel authored 5 months ago and Sebastian Reichel committed 1 week ago

Print the textual representation from the PD 3.2 specification
for control message types instead of the 5 bit message ID.

Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>

Sebastian Reichel authored 5 months ago and Sebastian Reichel committed 1 week ago

On Radxa ROCK 5B I managed to get U-Boot into an endless loop of
printing

fusb302 usb-typec@22: TCPM: data role mismatch, initiating error recovery

messages by changing the data role in Linux and then rebooting the
system. This is happening because the external device (A cheap USB-C hub
powered through a USB-C PD power-supply) kept its state and the error
recovery path for non self-powered devices is not enough to change it.

Avoid this by swapping our own data role. Right now data support is
limited anyways and once proper support is added we can use the data
role swap request to get the desired data direction.

Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>

Sebastian Reichel authored 5 months ago and Sebastian Reichel committed 1 week ago

A USB-PD hard reset involves removing the voltage from VBUS for some
time. So basically it has the same effect as removing the USB-C plug
for a short moment. If the machine is powered from the USB-C port and
does not have a fallback supply (e.g. a battery), this will result in
a full machine reset due to power loss.

Ideally we want to avoid triggering a hard reset on these boards. A
non-working USB-C port is probably better than unplanned reboots. But
boards with a backup supply should do the hard reset to get everything
working again.

In theory it would be enough to check the self_powered property, but
it seems the property might not be configured consistently enough in
system firmwares.

So let's start with just printing an error message when a hard reset is
triggered on systems we expect to be affected. This at least makes
debugging issues on affected systems easier without impacting unaffected
systems too much.

Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>

Sebastian Reichel authored 5 months ago and Sebastian Reichel committed 1 week ago

The current TCPM code implements source capability message handling
according to the USB-PD specification. Unfortunately some USB PD
sources do not properly follow the specification and do not send
source capability messages after a soft reset when they already
negotiated a specific contract before. The current way (and what is
described in the specificiation) to resolve this problem is triggering
a hard reset.

But a hard reset is fatal on batteryless platforms powered via USB-C PD,
since that removes VBUS for some time. Since this is triggered at boot
time, the system may get stuck in a boot loop.

For example I noticed the following behaviour on a Radxa Rock 5B: The
leak voltage from a serial connection is enough to keep the fusb302
register state. Since U-Boot configures the fusb302 into
auto-acknowledge mode, the chip will acknowledge a source capability
message before U-Boot has probed the fusb302/tcpm driver after a
hard reset happened. For a compliant power-supply this is not a problem,
since the driver first of all sends a soft reset. But the non-compliant
power-supply does not start sending new source capability messages and
U-Boot will react with a hard reset resulting in a boot loop.

Note, that I changed the Linux kernel TCPM code in the same way some
time ago. While U-Boot is only affected because of the leak voltage
problem, Linux needs this because U-Boot already handles the
initial communication.

Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>

Sebastian Reichel authored 6 months ago and Sebastian Reichel committed 1 week ago

When the Rock 5B is booted with the current TCPM with its power supplied
by a "Cambrionix PDSync-C4" port it reaches the power-supply ready state.
Once that has happened the hub starts sending GetSinkCap messages, but
U-Boot already stopped processing PD messages. After retrying a bunch of
times the hub instead sends soft resets and with those also not being
replied to a hard reset will cut off the supply voltage.

Since the state machine is already prepared to handle GetSinkCap
messages, try to avoid this by handling incoming messages for another
50ms after reaching the ready state.

Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>

Detlev Casanova authored 11 months ago and Sebastian Reichel committed 1 week ago

Some net driver, like rtl8169, can set/get the MAC address from the
registers and store it in pdata->enetaddr.

When that happens, if there is a mismatch with the environment MAC
address, u-boot will show that the MAC address source is DT. This patch
ensures that the shown source is "driver" instead to avoid confusion.

Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>

Detlev Casanova authored 11 months ago and Sebastian Reichel committed 1 week ago

On some boards, a MAC address is set based on the CPU ID or other
information. This is usually done in the misc_init_r() function.

This becomes a problem for net devices that are probed after the call to
misc_init_r(), for example, when the ethernet is on a PCI port, which
needs to be enumerated.

In this case, misc_init_r() will set the ethaddr variable, then, when
the ethernet device is probed, if it has a ROM address, u-boot will warn
about a MAC address mismatch and use the misc_init_r() address instead
of the one in ROM.

The operating system later will most likely use the ROM MAC address,
which can be confusing.

To avoid that, this commit introduces NET_PREFER_ROM_MAC_ADDR that can
be set for boards that have such an interface.

Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>

Sebastian Reichel authored 1 year ago and Sebastian Reichel committed 1 week ago

We need prebuilt artifacts for Rock-5B board.

Co-developed-by: Christopher Obbard <chris.obbard@collabora.com>
Signed-off-by: Christopher Obbard <chris.obbard@collabora.com>
Co-developed-by: Eugen Hristev <eugen.hristev@collabora.com>
Signed-off-by: Eugen Hristev <eugen.hristev@collabora.com>
Signed-off-by: Sebastian Reichel <sebastian.reichel>

Sebastian Reichel authored 3 months ago and Sebastian Reichel committed 1 week ago

Will be replaced by internal pipeline.

Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>

Eugen Hristev authored 2 years ago and Sebastian Reichel committed 1 week ago

Add .ini file for use with boot merger tool to create DDR blob + SPL binary.

Signed-off-by: Eugen Hristev <eugen.hristev@collabora.com>

Jonas Karlman authored 2 years ago and Sebastian Reichel committed 1 week ago

This needs to be reworked, possible into mkimage parameters that can be
passed from binman. Or let binman add a "keep-embedded" prop singling
mkimage that data should remain embedded.

With this we should also be able to remove the u-boot,spl-fifo-mode prop
from other rk33 arm64 devices.

Signed-off-by: Jonas Karlman <jonas@kwiboo.se>

Ziyuan Xu authored 7 years ago and Sebastian Reichel committed 1 week ago

Per dw_mmc databook, it's recommended to reset the host controller if
some data-related error occurred.
Implement a reset mechanism.

Signed-off-by: Ziyuan Xu <xzy.xu@rock-chips.com>
Co-developed-by: Jason Zhu <jason.zhu@rock-chips.com>
Signed-off-by: Jason Zhu <jason.zhu@rock-chips.com>
[eugen.hristev@collabora.com: modified a bit the variables initialization]
Signed-off-by: Eugen Hristev <eugen.hristev@collabora.com>

Signed-off-by: Tom Rini <trini@konsulko.com>

On Ubuntu 25.04 installing the dependency pyyaml 6.0 fails with

    License classifiers are deprecated

Update PyYAML to the current release.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>

It has been reported that memory corruption can occurred because network
packages where received after EXIT_BOOT_SERVICES. See the thread
starting at [1].

We try to remove all drivers when EXIT_BOOT_SERVICES is called. But

* Some network drivers don't call their own stop method when removed.
* Some network drivers don't have a remove method.
* Some devices have CONFIG_DM_DEVICE_REMOVE=n.

Let's call eth_halt() in EXIT_BOOT_SERVICES explicitly.

[1]
https://lore.kernel.org/all/C101B675-EEE6-44CB-8A44-83F72182FBD6@kohlschutter.com/



Cc: Michael Brown <mcb30@ipxe.org>
Reported-by: Christian Kohlschütter <christian@kohlschutter.com>
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Michael Brown <mcb30@ipxe.org>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Tested-by: Christian Kohlschütter <christian@kohlschutter.com>

A recent change tried to speed up EFI booting by not calling dhcp_run
from eth_bootdev_hunt() every time. PXE so far relied on that dhcp_run
call to get an IP address.

Fix that by adding a dhcp_run call to the pxe bootflow, with autoload
disabled to only get the ip address and nothing else.

Suggested-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Fixes: 1f68057e ("net: eth_bootdev_hunt() should not run DHCP")
Signed-off-by: Heiko Stuebner <heiko@sntech.de>

S/MIME Capabilities (OID: 1.2.840.113549.1.9.15) attributes are
expected to be algorithms but neither data nor MS Inderect Data,
thus the checker for data type is incorrect.

This patch fixes a capsule authentication failure with PKCS#7
message that contains S/MIME capabilities, which formed by the EDK2
GenerateCapsule tool.

S/MIME Capabilities are not common attributes in an EFI capsule,
thus this failure cannot be reproduced with the capsules generated
via mkeficapsule.

Fixes: 7de0d155 ("mbedtls: add PKCS7 parser porting layer")
Reported-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: Raymond Mao <raymond.mao@linaro.org>

As reported by Niklas Sombert, enabling bloblist on these platforms
means that we currently don't pass SMBIOS tables on to later stages. For
the v2025.04 release, disable bloblists as that is sufficient to restore
functionality and we will address the underlying problem for the next
release.

Link: https://lore.kernel.org/u-boot/c1dc7446-d1d0-4ef2-b201-09fc612aa8db@uni-duesseldorf.de


Reported-by: Niklas Sombert <niklas.sombert@uni-duesseldorf.de>
Signed-off-by: Tom Rini <trini@konsulko.com>

- Check interface before comparing it in set_dfu_alt_info() to avoid
  hard crash on capsule update

When set_dfu_alt_info() is called, interface can be NULL when called for
the EFI capsule, so check it before calling it with strcmp().

Fixes: 12ea40d2 ("ARM: meson: add support for Libre Computer aml-s905d3-cc")
Fixes: 9e6e6b03 ("ARM: meson: add support for Libre Computer aml-a311d-cc")
Fixes: 75c87c6c ("board: libre-computer: aml-s805x-cc: Enable capsule updates")
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Link: https://lore.kernel.org/r/20250403-u-boot-fix-set-dfu-alt-info-interface-v1-1-1fdd12463186@linaro.org


Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>

Pull request efi-2025-04-rc6

CI:

* https://source.denx.de/u-boot/custodians/u-boot-efi/-/pipelines/25491

Documentation:

* correct references to overlay-fdt-boot.txt
* build: docker: Fix code highlighting syntax
* build: docker: Use code-block for syntax highlighting
* Fix typo bootfloe to bootflow

UEFI:

* Correct the size of the AcpiReclaimMemory region for ACPI tables
  when using QFW

First set of u-boot-at91 fixes for the 2025.04 cycle:

- A single fix for the reset command on sama5d27 boards.

Signed-off-by: Andrew Davis <afd@ti.com>

doc/uImage.FIT/overlay-fdt-boot.txt does not exist anymore.
Reference the correct section of doc/usage/fit/overlay-fdt-boot.rst.

Fixes: 6f6e8bb6 ("doc: Bring in the FIT overlay information")
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>

Use ..code-block:: syntax highlighting instead of :: so all bash
commands use the same syntax highlighting.

Signed-off-by: Leonard Anderweit <l.anderweit@phytec.de>
Reviewed-by: Quentin Schulz <quentin.schulz@cherry.de>

Remove double :: before .. code-block:: bash to correctly highlight the
following commands.

Signed-off-by: Leonard Anderweit <l.anderweit@phytec.de>
Reviewed-by: Quentin Schulz <quentin.schulz@cherry.de>

The end of the ACPI table is set to 'addr' instead of 'end'. The ACPI
code for QEMU relies on those values to mark memory as 'ACPI Reclaim'
and as a result the ACPI RSDP ends up in Boot services Data.

Reported-by: Björn Töpel <bjorn@kernel.org>
Fixes: commit 638cc363 ("acpi: enable writing ACPI tables on QEMU")
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>

Since commit 61040097 ("reset: at91: Add reset driver for basic
assert/deassert operations") the "atmel,sama5d3-rstc" compatible for
the sama5d2 reset controller in sama5d2.dtsi is not handled by
CONFIG_SYSRESET_AT91 anymore, but by CONFIG_RESET_AT91 now.

This resulted in the following error, when trying to reset from the U-Boot
shell on a sama5d27_wlsom1 board:

 => reset
 resetting ...
 System reset not supported on this platform
 ### ERROR ### Please RESET the board ###

Fix it by enabling the CONFIG_RESET_AT91 driver in all sama5d27 defconfigs.

Tested on a sama5d27_wlsom1 board.

Based on the fix in commit e1ee52ca ("configs: at91: sam9x60: Switch
to new reset driver")

Fixes: 61040097 ("reset: at91: Add reset driver for basic assert/deassert operations")
Signed-off-by: Fabio Estevam <festevam@denx.de>

Pull request net-20250314.

kconfig:
* Fix submenu for network commands

net:
* Remove a useless (commented out) line in net-common.h

net-lwip:
* Remove error print on failed tx
* Fix return code of ping_loop() when no ethernet device is found
* Remove superfluous newline in help text for tftp

The Kconfig parser seems to get confused by the current if conditions
following CMD_NET and displays all network command options directly in
the "Command line interface" menu instead of in a "Network commands"
submenu.

To help out Kconfig we can simplify the if conditions, so that the
definition of CMD_NET is followed immediately by an if/endif block that
contains all network command options. We can also remove nested checks
for CMD_NET or (NET || NET_LWIP).

Fixes: 98ad145d ("net: lwip: add DHCP support and dhcp commmand")
Signed-off-by: Paul Barker <paul.barker.ct@bp.renesas.com>
Reviewed-by: Jerome Forissier <jerome.forissier@linaro.org>

When an ethernet driver fails to send a frame we print an error in lwIP.
But depending on how often that error is it might significantly delay
transmissions.

For example downloading a big file with the rpi4 spams the console with
'send error: -101', but removing the print makes the file download with
an average speed of ~8.5MiB/s since the packets are retransmitted.

So let's move it to a 'debug' in lwIP and expect ethernet drivers to handle
the failure if it's severe.

Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Reviewed-by: Jerome Forissier <jerome.forissier@linaro.org>

Commit 1d5d292b ("net: split net into net{,-common,-legacy,-lwip}")
inadvertendly left a commented out declaration for do_wget() in
net-common.h. Remove it.

Signed-off-by: Jerome Forissier <jerome.forissier@linaro.org>
Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>

do_ping() expects ping_loop() to return a negative value on error, so
that it can propagate it to the caller as CMD_RET_FAILURE. This is not
the case when no ethernet device is found, so fix that.

Signed-off-by: Jerome Forissier <jerome.forissier@linaro.org>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>

The help text has a newline at the end which will lead to an empty
line after the tftpboot when printing the help overview. Remove it.

Fixes: 4d4d7838 ("net: lwip: add TFTP support and tftpboot command")
Signed-off-by: Michael Walle <mwalle@kernel.org>
Reviewed-by: Jerome Forissier <jerome.forissier@linaro.org>