Simon Glass [Sat, 30 Jul 2022 21:52:34 +0000 (15:52 -0600)]
bootstd: Add vbe bootmeth into sandbox
Update sandbox to include the VBE bootmeth. Update a few existing tests to
take account of this change, specifically that the new bootmeth now
appears when scanning.
Simon Glass [Sat, 30 Jul 2022 21:52:31 +0000 (15:52 -0600)]
event: Add an event for device tree fixups
At present there is a confusing array of functions that handle the
device tree fix-ups needed for booting an OS. We should be able to switch
to using events to clean this up.
As a first step, create a new event type and call it from the standard
place.
Note that this event uses the ofnode interface only, since this can
support live tree which is more efficient when making lots of updates.
Simon Glass [Sat, 30 Jul 2022 21:52:29 +0000 (15:52 -0600)]
bootstd: Drop the system bootdev
This was a work-around for the fact that global bootmeths such as EFI
bootmgr and VBE don't use a particular bootdev, or at least select it
themselves so that we don't need to scan all bootdevs when using that
bootmeth.
Simon Glass [Sat, 30 Jul 2022 21:52:28 +0000 (15:52 -0600)]
bootstd: Always create the EFI bootmgr bootmeth
Now that we can separate this out from the normal bootmeths, update the
code to create it always.
We cannot rely on the device tree to create this, since the EFI project
is quite opposed to having anything in the device tree that helps U-Boot
with its processing.
Simon Glass [Sat, 30 Jul 2022 21:52:27 +0000 (15:52 -0600)]
bootstd: Allow scanning for global bootmeths separately
Typically we want to find and use global bootmeths first, since they have
the best idea of how the system should boot. We then use normal bootmeths
as a fallback.
Add the logic for this, putting global bootmeths at the end of the
ordering. We can then easily scan the global bootmeths first, then drop
them from the list for subsequent bootdev-centric scans.
This changes the ordering of global bootmeths, so update the
bootflow_system() accordingly.
Drop the comment from bootmeth_setup_iter_order() since this is an
exported function and it should be in the header file.
Simon Glass [Sat, 30 Jul 2022 21:52:25 +0000 (15:52 -0600)]
bootstd: Support bootflows with global bootmeths
Add support for handling this concept in bootflows. Update the 'bootflow'
command to allow only the normal bootmeths to be used. This alllows
skipping EFI bootmgr and VBE, for example.
Simon Glass [Sat, 30 Jul 2022 21:52:22 +0000 (15:52 -0600)]
bootstd: Allow EFI bootmgr to support an invalid bootflow
For most testing we don't want this bootmeth to actually do anything. For
the one test where we do, add a test hook to obtain the correct behaviour.
This will allow us to bind the device always, rather than just doing it
for this test.
Simon Glass [Sat, 30 Jul 2022 21:52:21 +0000 (15:52 -0600)]
bootstd: Allow bootmeths to be marked as global
The current way of handling things like EFI bootmgr is a bit odd, since
that bootmeth handles selection of the bootdev itself. VBE needs to work
the same way, so we should support it properly.
Add a flag that indicates that the bootmeth is global, rather than being
invoked on each bootdev. Provide a helper to read a bootflow from the
bootmeth.
Simon Glass [Sat, 30 Jul 2022 21:52:16 +0000 (15:52 -0600)]
bootstd: Drop delays in the tests
Some tests go as far as booting a distribution. In this case a menu is
presented to the user, with a two-second timeout. This adds a total of
12 seconds to the test runs at present.
Avoid this by inserting a response using the console-recording feature.
Simon Glass [Sat, 30 Jul 2022 21:52:13 +0000 (15:52 -0600)]
dm: core: Allow writing to a flat tree with ofnode
In generally it is not permitted to implement an ofnode function only for
flat tree or live tree. Both must be supported. Also the code for
live tree access should be in of_access.c rather than ofnode.c which is
really just for holding the API-conversion code.
Update ofnode_write_prop() accordingly and fix the test so it can work
with flat tree too.
Simon Glass [Sat, 30 Jul 2022 21:52:08 +0000 (15:52 -0600)]
dm: core: Introduce support for multiple trees
At present ofnode only works with a single device tree, for the most part.
This is the control FDT used by U-Boot.
When booting an OS we may obtain a different device tree and want to
modify it. Add some initial support for this into the ofnode API.
Note that we don't permit aliases in this other device tree, since the
of_access implementation maintains a list of aliases collected at
start-up. Also, we don't need aliases to do fixups in the other FDT. So
make sure that flat tree and live tree processing are consistent in this
area.
Simon Glass [Sat, 30 Jul 2022 21:52:07 +0000 (15:52 -0600)]
dm: core: Add a note about how livetree updates work
The unflattening algorithm results in a single block of memory being
allocated for the whole tree. When writing new properties, these are
allocated new memory outside that block. When the block is freed, the
allocated properties remain.
Document how this works and the potential memory leak, as well as
mentioning that updating the livetree is actually supported now.
Simon Glass [Sat, 30 Jul 2022 21:52:06 +0000 (15:52 -0600)]
dm: core: Split out the declaration of ofnode
This is used by a lot of files, but ofnode.h needs to include a lot of
header files. This can create dependency cycles, particularly with
global_data.h which must include various declarations.
Split the core delcarations into a separate file to fix this.
The booting flow is SPL -> OpenSBI -> U-Boot.
The boot hart may change after OpenSBI and may not always be hart0,
so wrap the related branch instruction with M-MODE.
Current DTB setup for XIP is not valid.
There is no chance for CONFIG_SYS_FDT_BASE, the DTB address used
in XIP mode, to be returned. Fix this.
Fixes: 2831fc2c4d19 ("riscv: Remove OF_PRIOR_STAGE from RISC-V boards") Signed-off-by: Rick Chen <rick@andestech.com> Signed-off-by: Leo Yu-Chi Liang <ycliang@andestech.com>
John Keeping [Thu, 28 Jul 2022 10:19:15 +0000 (11:19 +0100)]
boot: allow bootmeth-distro without CONFIG_NET
Remove the dependency on CMD_PXE from BOOTMETH_DISTRO by introducing a
new hidden kconfig symbol to control whether pxe_utils is compiled,
allowing bootstd's distro method to be compiled without needing
networking support enabled.
Signed-off-by: John Keeping <john@metanate.com> Reviewed-by: Tom Rini <trini@konsulko.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Correct build errors when CMD_BOOTM is not enabled: Signed-off-by: Simon Glass <sjg@chromium.org>
lz4: Fix compile warning comparison of distinct pointer types
In file included from include/linux/bitops.h:22,
from include/log.h:15,
from include/linux/printk.h:4,
from include/common.h:20,
from lib/lz4_wrapper.c:6:
lib/lz4_wrapper.c: In function ‘ulz4fn’:
include/linux/kernel.h:184:17: warning: comparison of distinct pointer types lacks a cast
(void) (&_min1 == &_min2); \
^~
lib/lz4_wrapper.c:104:18: note: in expansion of macro ‘min’
size_t size = min((ptrdiff_t)block_size, end - out);
^~~
cmd: inconsistent return type of command_process()
The declarations in the header and in the implementation must match.
Reported-by: Sergei Antonov <saproj@gmail.com> Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Simon Glass [Mon, 1 Aug 2022 13:57:59 +0000 (07:57 -0600)]
test: Add some tests for kconfig.h
The macros in this file are a little confusing and we currently have no
tests to check that they work as expected.
Add some tests which check the macros in C code. Add a few tests which
check that the build errors are generated correctly too, using buildman's
-a option.
Patrice Chotard [Tue, 2 Aug 2022 08:21:35 +0000 (10:21 +0200)]
lmb: Fix LMB_MEMORY_REGIONS flag usage
This patch is fixing a broken boot observed on stm32mp157c-dk2 board.
IS_ENABLED macro should be used to check if a compilation flag is set
to "y" or "m".
LMB_MEMORY_REGIONS is set to a numerical value, IS_ENABLED macro is not
suitable in this case.
Fixes: 75c729cadc20 ("lmb: Fix lmb property's defination under struct lmb") Signed-off-by: Patrice Chotard <patrice.chotard@foss.st.com> Acked-by: Michal Simek <michal.simek@amd.com> Reviewed-by: Tom Rini <trini@konsulko.com>
Simon Glass [Wed, 3 Aug 2022 18:08:29 +0000 (12:08 -0600)]
Makefile: Correct the rule removing old of-platdata files
This makes use of makefile variables that don't exist anymore. Fix it and
also remove the object files in that directory.
Also add FORCE as a dependency as required by the if_changed macro.
Fixes ae639183906 ("Makefile: Remove old of-platdata files before regenerating") Reported-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com> Signed-off-by: Simon Glass <sjg@chromium.org>
Marek Vasut [Thu, 14 Jul 2022 23:58:24 +0000 (01:58 +0200)]
mmc: Do not send status of send_status is false
Commit d8fefc8a1fd ("mmc: Fix mmc_switch excessive timeout") introduced
a side effect where CMD13 SEND_STATUS is issued in case mmc_wait_dat0()
does not return -ENOSYS and $send_status is not set. This happens on all
hardware which does implement .mmc_wait_dat0 callback, e.g. i.MX8M .
This leads to lengthy timeout before booting OS in case of eMMC in one
of the HS200/HS400 modes, since the card cannot respond to CMD13 while
downgrading from HS200/HS400 to regular HS mode.
Fix this by adding the missing conditional.
Fixes: d8fefc8a1fd ("mmc: Fix mmc_switch excessive timeout") Signed-off-by: Marek Vasut <marex@denx.de> Cc: Jaehoon Chung <jh80.chung@samsung.com> Cc: Kirill Kapranov <kirill.kapranov@compulab.co.il> Cc: Marek Behún <marek.behun@nic.cz> Cc: Pantelis Antoniou <panto@antoniou-consulting.com> Cc: Ye Li <ye.li@nxp.com> Reviewed-by: Jaehoon Chung <jh80.chung@samsung.com>
Pali Rohár [Thu, 4 Aug 2022 11:03:44 +0000 (13:03 +0200)]
misc: atsha204a: Don't check for error when waking up the device
The device ignores any levels or transitions on the SCL pin when the device
is idle, asleep or during waking up.
Linux kernel driver for atsha204a (atmel-sha204a.ko) also ignores return
value from i2c wakeup send command, see:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/crypto/atmel-i2c.c?h=v5.19#n174
And also userspace Turris libatsha204 library ignores return value from
wakeup send command, see:
https://gitlab.nic.cz/turris/libatsha204/-/blob/v29.2/src/libatsha204/layer_ni2c.c#L75-76
U-Boot driver should do same thing.
Fixes waking up ATSHA204 on Turris 1.x boards.
Signed-off-by: Pali Rohár <pali@kernel.org> Reviewed-by: Stefan Roese <sr@denx.de> Tested-by: Paweł Anikiel <pan@semihalf.com> Reviewed-by: Marek Behún <kabel@kernel.org>
Pali Rohár [Thu, 4 Aug 2022 10:41:54 +0000 (12:41 +0200)]
arm64: a37xx: pinctrl: Fix definitions for MPP pins 20-22
All 3 MPP pins (20, 21 and 22) can be configured individually and also can
be configured to GPIO functions. Fix definitions for these MPP pins in
existing pin groups. After this change GPIO function can be enabled just
for one of these 3 pins.
Signed-off-by: Pali Rohár <pali@kernel.org> Reviewed-by: Stefan Roese <sr@denx.de>
Chris Packham [Thu, 4 Aug 2022 09:06:26 +0000 (21:06 +1200)]
ARM: kirkwood: SBx81LIFKW: disable KIRKWOOD_GPIO
DM_GPIO was already enabled so the MVEBU_GPIO was already available.
Having updated the board code to use the DM_GPIO APIs the KIRKWOOD_GPIO
driver became unnecessary. Disable it for SBx81LIFKW.
Signed-off-by: Chris Packham <judge.packham@gmail.com> Reviewed-by: Stefan Roese <sr@denx.de>
Pali Rohár [Wed, 3 Aug 2022 11:00:51 +0000 (13:00 +0200)]
arm: mvebu: Remove redundant u-boot, dm-pre-reloc from all 32-bit Armada SoCs
Replace it by including of mvebu-u-boot.dtsi file. When board does not use
-u-boot.dtsi then mvebu-u-boot.dtsi is included automatically by makefile
scripts/Makefile.lib.
Signed-off-by: Pali Rohár <pali@kernel.org> Reviewed-by: Stefan Roese <sr@denx.de> Tested-by: Stefan Roese <sr@denx.de>
Pali Rohár [Wed, 3 Aug 2022 11:00:50 +0000 (13:00 +0200)]
arm: mvebu: Introduce mvebu-u-boot.dtsi for 32-bit Armada SoCs
Set u-boot,dm-pre-reloc for /soc/, /soc/internal-regs/ and &uart0 nodes as
it is required on every 32-bit Armada SoCs. And set also u-boot,dm-pre-reloc
for &spi0 when going to boot from SPI because otherwise SPL SPI drivers do
not load.
Signed-off-by: Pali Rohár <pali@kernel.org> Reviewed-by: Stefan Roese <sr@denx.de> Tested-by: Stefan Roese <sr@denx.de>
Pali Rohár [Fri, 5 Aug 2022 11:37:25 +0000 (13:37 +0200)]
arm: mvebu: dts: Build only arch-compatible dts files
64-bit Armada DTS files are not build correctly during compilation of
32-bit Armada boards and vice versa. So fix makefile build system to
compile only those dts files which are compatible for the current build
(64-bit Armada DTS files only for 64-bit builds and 32-bit Armada DTS files
only for 32-bit builds).
Signed-off-by: Pali Rohár <pali@kernel.org> Reviewed-by: Stefan Roese <sr@denx.de>
Pali Rohár [Tue, 2 Aug 2022 09:55:19 +0000 (11:55 +0200)]
arm: mvebu: spl: Always fallback to BootROM boot method
BootROM boot method should always work so always add it as fallback method
to spl_boot_list. In case U-Boot SPI driver fails it is better to try using
BootROM than hanging as by default only one boot method is specified.
Signed-off-by: Pali Rohár <pali@kernel.org> Tested-by: Tony Dinh <mibodhi@gmail.com> Reviewed-by: Stefan Roese <sr@denx.de>
Pali Rohár [Mon, 1 Aug 2022 10:11:13 +0000 (12:11 +0200)]
gpio: turris_omnia_mcu: Fix usage of CMD_EXT_CONTROL
CMD_GENERAL_CONTROL takes two 8-bit arguments but CMD_EXT_CONTROL takes
two 16-bit arguments. Fix this issue and change CMD_EXT_CONTROL arguments
to 16-bit.
Fixes: c12d8a3b1405 ("gpio: Add Turris Omnia MCU driver") Signed-off-by: Pali Rohár <pali@kernel.org> Reviewed-by: Stefan Roese <sr@denx.de>
Pali Rohár [Mon, 1 Aug 2022 10:02:20 +0000 (12:02 +0200)]
arm: mvebu: turris_omnia: Increase fdt size in fixup_mtd_partitions
Sometimes fixup_mtd_partitions() prints during booting kernel error
"Failed fixing SPI NOR partitions!" because it does not have enough space
for creating all paritions nodes. So increase fdt size.
Fixes: 1f3c494182e8 ("arm: mvebu: turris_omnia: fixup MTD partitions in Linux' DTB") Signed-off-by: Pali Rohár <pali@kernel.org> Reviewed-by: Stefan Roese <sr@denx.de> Reviewed-by: Marek Behún <kabel@kernel.org>
arm: mvebu: turris_omnia: Add support for design with SW reset signals
New Turris Omnia HW board revision requires that software controls
peripheral reset signals, namely PERST# signals on mPCIe slots, ethernet
phy reset and lan switch reset. Those pins are connected to MCU controlled
by MCU i2c API as GPIOs. On new HW board revision those pins stay in reset
after board reset and software has to release these peripherals from reset
manually. MCU announce this requirement by FEAT_PERIPH_MCU bit in
CMD_GET_FEATURES command.
On older HW board revisions when FEAT_PERIPH_MCU is not announced, all
those reset signals are automatically released after board finish reset.
Detect FEAT_PERIPH_MCU bit in board_fix_fdt() and ft_board_setup()
functions and insert into device tree blob pcie "reset-gpios" and eth phy
"phy-reset-gpios" properties with corresponding MCU gpio definitions.
PCIe and eth PHY drivers then automatically release resets during device
initialization. Both U-Boot and Linux kernel drivers support those device
tree reset properties.
Initialization of lan switch on new HW board revision is more complicated.
Switch strapping pins are shared with switch RGMII pins. And strapping pins
must be in specific configuration after releasing switch reset. Due to pin
sharing, it is first required to switch A385 side of switch pins into GPIO
mode, set strapping configuration, release switch from reset and after that
switch A385 pins back to RGMII mode.
Because this complicated setup is not supported by switch DSA drivers and
cannot be expressed easily in device tree, implement it manually in SPL
function spl_board_init(). So in proper U-Boot and OS/kernel would be lan
switch initialized and be in same configuration like it was on old HW board
revisions (where reset sequence did those steps at hardware level).
Signed-off-by: Pali Rohár <pali@kernel.org> Reviewed-by: Stefan Roese <sr@denx.de> Reviewed-by: Marek Behún <kabel@kernel.org>
This allows you to specify the type of rgmii-id that will enable phy
internal delay in ethernet phy-mode.
This adds all RGMII cases to all of get_pinmode() except LD11, because LD11
SoC doesn't support RGMII due to the constraint of the hardware. When RGMII
phy mode is specified in the devicetree for LD11, the driver will abort
with an error.
Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com> Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
net: phy: possible NULL dereference in fixed_phy_create()
We check if phydev is NULL. Only but if it is non-NULL we set one
component of phydev. But even if it is NULL we set another. We should not
dereference NULL in either case.
Fixes: a62b976714fb ("net: phy: don't require PHY interface mode during PHY creation") Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com> Reviewed-by: Marek Behún <kabel@kernel.org>
Ramon Fried [Sun, 5 Jun 2022 00:44:15 +0000 (03:44 +0300)]
net: phy: Remove inline definitions from convinience functions
The convinience functions are not that small and they caused
bloated text segments because of their usage.
There was no need to inline them in the first place, as
they're not part of a fastpath.
Signed-off-by: Ramon Fried <rfried.dev@gmail.com> Reviewed-by: Michael Trimarchi <michael@amarulasolutions.com> Reviewed-by: Marek Behún <kabel@kernel.org> Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Zev Weiss [Tue, 17 May 2022 22:16:39 +0000 (15:16 -0700)]
net: ftgmac100: use bus name in mdio error messages
Previously we'd been using a device name retrieved via
ftgmac100_data->phydev, but the mdio read/write functions may be
called before that member is initialized in ftgmac100_phy_init(),
leading to a NULL pointer dereference while printing the error message
issued if the mdio access fails. We can instead use bus->name, which
is already available at that point.
Signed-off-by: Zev Weiss <zev@bewilderbeest.net> Fixes: 0dd2ef6b54ce ("net: ftgmac100: add MDIO bus and phylib support") Reviewed-by: Cédric Le Goater <clg@kaod.org>
Rasmus Villemoes [Thu, 12 May 2022 07:33:07 +0000 (09:33 +0200)]
net: dwc_eth_qos: remove use of DWC_NET_PHYADDR
Only two boards in the tree set the macro DWC_NET_PHYADDR. Both have
CONFIG_DM_ETH_PHY=y, so should set the phy address in DT if necessary.
The imx8mp_evk does set the correct address in device tree.
The other board seems to be a copy-paste-adapt from an old
version of the imx8mp_evk config header, given the "#ifdef
CONFIG_DWC_ETH_QOS" block that has been removed from imx8mp_evk header
in commit 6735d8b9349. Its device tree doesn't even enable (i.e., set
'status = "okay"') the &eqos node. But the other ethernet device,
&fec, does get enabled, and does have a phy sitting at address 4 (and
it also has a corresponding legacy #define CONFIG_FEC_MXC_PHYADDR
4). So I believe it should be completely safe to remove it from there
as well.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk> Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
[trini: Re-apply to top of tree, update imx93_evk.h] Signed-off-by: Tom Rini <trini@konsulko.com>
Rasmus Villemoes [Wed, 11 May 2022 14:58:41 +0000 (16:58 +0200)]
net: dwc_eth_qos: lift parsing of max-speed DT property to common code
I have an iMX8MP with a ti,dp83867 phy in front of the eqos
interface. The phy is Gbit capable - however, the C and D differential
pairs are not physically routed to the RJ45 connector. So I need to
prevent the phy from advertising 1000Mbps.
The necessary code is almost already there in the form of a
phy_set_supported() call in eqos_start(), but the max-speed DT
property is currently only parsed in
eqos_probe_resources_stm32(). Lift that parsing to eqos_probe().
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk> Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Rasmus Villemoes [Wed, 11 May 2022 14:12:50 +0000 (16:12 +0200)]
net: dwc_eth_qos: fix double resource leak in eqos_remove()
Not only does eqos_remove() fail to free the buffers that have been
allocated by eqos_probe_resources_core(), it repeats those allocations
and thus drops twice as much memory on the floor.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk> Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Marek Vasut [Mon, 25 Apr 2022 18:28:05 +0000 (20:28 +0200)]
net: dm9000: Correctly handle empty FIFO
Assign packet pointer only in case the MAC reports anything in the FIFO.
In case the MAC indicates empty FIFO, return 0 to pass that information
to the network stack.
Signed-off-by: Marek Vasut <marex@denx.de> Cc: Joe Hershberger <joe.hershberger@ni.com> Cc: Ramon Fried <rfried.dev@gmail.com> Reviewed-by: Ramon Fried <rfried.dev@gmail.com>