platform/kernel/u-boot.git
3 years agosata: Rename SATA_SUPPORT to SATA
Simon Glass [Sun, 8 Aug 2021 18:20:17 +0000 (12:20 -0600)]
sata: Rename SATA_SUPPORT to SATA

Rename this options so that CONFIG_IS_ENABLED can be used with it.

Signed-off-by: Simon Glass <sjg@chromium.org>
3 years agoppc: Rename MPC8XXX_INIT_DDR_SUPPORT to MPC8XXX_INIT_DDR
Simon Glass [Sun, 8 Aug 2021 18:20:16 +0000 (12:20 -0600)]
ppc: Rename MPC8XXX_INIT_DDR_SUPPORT to MPC8XXX_INIT_DDR

Rename these options so that CONFIG_IS_ENABLED can be used with them.

Signed-off-by: Simon Glass <sjg@chromium.org>
3 years agoRename CACHE_SUPPORT to CACHE
Simon Glass [Sun, 8 Aug 2021 18:20:15 +0000 (12:20 -0600)]
Rename CACHE_SUPPORT to CACHE

Rename this option so that CONFIG_IS_ENABLED can be used with it.

Oddly there is already an SPL_CACHE option. Drop it in favour of this one.

Drop the special SPL Makefile rule which is now superfluous.

Signed-off-by: Simon Glass <sjg@chromium.org>
3 years agospi: Rename SPI_SUPPORT to SPI
Simon Glass [Sun, 8 Aug 2021 18:20:14 +0000 (12:20 -0600)]
spi: Rename SPI_SUPPORT to SPI

Rename these options so that CONFIG_IS_ENABLED can be used with them.

Signed-off-by: Simon Glass <sjg@chromium.org>
3 years agoserial: Add a SERIAL config
Simon Glass [Sun, 8 Aug 2021 18:20:13 +0000 (12:20 -0600)]
serial: Add a SERIAL config

At present we have SPL_SERIAL and TPL_SERIAL but not piain SERIAL. This
works because there is a special build rule in Makefile that always
includes the drivers/serial directory.

It is better to have all driver directories included by drivers/Makefile
and there is already a rule in there for this purpose. It just needs a
Kconfig for U-Boot proper, so add one.

It is always enabled, for now, since that mimics the current behaviour.
It should be possible to drop the strange 'SERIAL_PRESENT' option at some
point and use SERIAL instead.

Signed-off-by: Simon Glass <sjg@chromium.org>
3 years agoserial: Rename SERIAL_SUPPORT to SERIAL
Simon Glass [Sun, 8 Aug 2021 18:20:12 +0000 (12:20 -0600)]
serial: Rename SERIAL_SUPPORT to SERIAL

Rename these options so that CONFIG_IS_ENABLED can be used with them.

Signed-off-by: Simon Glass <sjg@chromium.org>
3 years agortc: Rename RTC_SUPPORT to RTC
Simon Glass [Sun, 8 Aug 2021 18:20:11 +0000 (12:20 -0600)]
rtc: Rename RTC_SUPPORT to RTC

Rename these options so that CONFIG_IS_ENABLED can be used with them.

Signed-off-by: Simon Glass <sjg@chromium.org>
3 years agopch: Rename PCH_SUPPORT to PCH
Simon Glass [Sun, 8 Aug 2021 18:20:10 +0000 (12:20 -0600)]
pch: Rename PCH_SUPPORT to PCH

Rename these options so that CONFIG_IS_ENABLED can be used with them.

Signed-off-by: Simon Glass <sjg@chromium.org>
3 years agommc: Rename MMC_SUPPORT to MMC
Simon Glass [Sun, 8 Aug 2021 18:20:09 +0000 (12:20 -0600)]
mmc: Rename MMC_SUPPORT to MMC

Rename these options so that CONFIG_IS_ENABLED can be used with them.

Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Jaehoon Chung <jh80.chung@samsung.com>
[trini: Fixup some incorrect renames]
Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoMerge branch '2021-09-03-xyz-modem-fixes' into next
Tom Rini [Fri, 3 Sep 2021 21:15:44 +0000 (17:15 -0400)]
Merge branch '2021-09-03-xyz-modem-fixes' into next

- Assorted x/y/z modem fixes from Pali

3 years agoxyz-modem: Allow to cancel transfer also by CTRL+C
Pali Rohár [Tue, 3 Aug 2021 14:28:44 +0000 (16:28 +0200)]
xyz-modem: Allow to cancel transfer also by CTRL+C

Currently it is possible to cancel loadx and loady commands by pressing
CTRL+X (CAN character) at least 3 times quickly.

All other U-Boot commands, including loadb and loads can be cancelled by
CTRL+C. So allow it also in xyz-modem code used by loadx and loady
commands. Implement it by handling CTRL+C (ETX character) in the same way
as CTRL+X (CAN character).

Due to how x/y-modem protocol works, it is required to press
CTRL+C or CTRL+X at least 3 times quickly.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Simon Glass <sjg@chromium.org>
3 years agoxyz-modem: Show information about finished transfer
Pali Rohár [Tue, 3 Aug 2021 14:28:43 +0000 (16:28 +0200)]
xyz-modem: Show information about finished transfer

Show "## Start Addr" or "## Binary (...) download aborted" information like
in Kermit "loadb" command.

Signed-off-by: Pali Rohár <pali@kernel.org>
3 years agoxyz-modem: Properly abort/terminate transfer on error
Pali Rohár [Tue, 3 Aug 2021 14:28:42 +0000 (16:28 +0200)]
xyz-modem: Properly abort/terminate transfer on error

Transfer termination tries to instruct sender that transfer was terminated.

Print error message and indicates aborted transfer in return value.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Simon Glass <sjg@chromium.org>
3 years agoxyz-modem: Close stream after processing/sending terminate sequence
Pali Rohár [Tue, 3 Aug 2021 14:28:41 +0000 (16:28 +0200)]
xyz-modem: Close stream after processing/sending terminate sequence

Obviously it is not possible to send terminate sequence over stream after
closing stream.

Signed-off-by: Pali Rohár <pali@kernel.org>
3 years agoxyz-modem: Put xyzModem_stream_close debug diagnostic message into ZM_DEBUG()
Pali Rohár [Tue, 3 Aug 2021 14:28:40 +0000 (16:28 +0200)]
xyz-modem: Put xyzModem_stream_close debug diagnostic message into ZM_DEBUG()

This is how all other debug / diagnostic messages are handled.

Signed-off-by: Pali Rohár <pali@kernel.org>
3 years agoxyz-modem: Fix x-modem "xyzModem_eof error" at the end of file
Pali Rohár [Tue, 3 Aug 2021 14:28:39 +0000 (16:28 +0200)]
xyz-modem: Fix x-modem "xyzModem_eof error" at the end of file

In x-modem protocol EOF is not an error state at the end of file.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Simon Glass <sjg@chromium.org>
3 years agoxyz-modem: Fix crash after cancelling transfer
Pali Rohár [Tue, 3 Aug 2021 14:28:38 +0000 (16:28 +0200)]
xyz-modem: Fix crash after cancelling transfer

Variable xyz.len is set to -1 on error. At the end xyzModem_stream_read()
function calls memcpy() with length from variable xyz.len. If this variable
is set to -1 then value passed to memcpy is casted to unsigned value, which
means to copy whole address space. Which then cause U-Boot crash. E.g. on
arm64 it cause CPU crash: "Synchronous Abort" handler, esr 0x96000006

Fix this issue by checking that value stored in xyz.len is valid prior
trying to use it.

Signed-off-by: Pali Rohár <pali@kernel.org>
Acked-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
3 years agoRevert most of the series for adding vexpress_aemv8r support
Tom Rini [Fri, 3 Sep 2021 14:40:28 +0000 (10:40 -0400)]
Revert most of the series for adding vexpress_aemv8r support

Per a request from Andre Przywara and agreed with by Peter Hoyes, the
vexpress aemv8r support wasn't quite ready to be merged, but the
discussion had moved off list.  We should keep the first patch in the
series for now, but revert the rest.  This reverts the following
commits:

e0bd6f31ce41 doc: Add documentation for the Arm vexpress board configs
30e5a449e8c7 arm: Use armv8_switch_to_el1 env to switch to EL1
b53bbca63bf4 vexpress64: Add BASER_FVP vexpress board variant
2f5b7b74903f armv8: Add ARMv8 MPU configuration logic
37a757e227cc armv8: Ensure EL1&0 VMSA is enabled

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoMerge branch '2021-09-02-assorted-platform-and-bugfixes' into next
Tom Rini [Thu, 2 Sep 2021 22:39:28 +0000 (18:39 -0400)]
Merge branch '2021-09-02-assorted-platform-and-bugfixes' into next

- Add position independent execution support for ARMv7
- Snapdragon, synquacer, vexpress64 fixes / improvements
- Prevent NEON register use on ARMv8
- Other assorted fixes

3 years agotools: env: Handle shorter read calls
Thibault Ferrante [Tue, 24 Aug 2021 15:29:50 +0000 (17:29 +0200)]
tools: env: Handle shorter read calls

On some cases, the actual number of bytes read can be shorter
than what was requested. This can be handled gracefully by
taking this difference into account instead of exiting.

Signed-off-by: Thibault Ferrante <thibault.ferrante@gmail.com>
3 years agopsci: fix double declaration
Oleksandr Suvorov [Mon, 23 Aug 2021 21:55:39 +0000 (00:55 +0300)]
psci: fix double declaration

The prototype of psci_features() duplicated. Remove extra declaration.

Fixed: e21e3ffdd1 ("psci: Fix warnings when compiling with W=1")
Reported-by: Michael Scott <mike@foundries.io>
Signed-off-by: Oleksandr Suvorov <oleksandr.suvorov@foundries.io>
3 years agopci: iproc: fix compilation errors and warnings
Bharat Gooty [Mon, 23 Aug 2021 09:27:24 +0000 (14:57 +0530)]
pci: iproc: fix compilation errors and warnings

Signed-off-by: Bharat Gooty <bharat.gooty@broadcom.com>
3 years agomailmap: Update e-mail for Jernej Skrabec
Jernej Skrabec [Sun, 22 Aug 2021 05:23:48 +0000 (07:23 +0200)]
mailmap: Update e-mail for Jernej Skrabec

Old address doesn't exist anymore. Map it to new one.

Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Acked-by: Andre Przywara <andre.przywara@arm.com>
3 years agoARM: Prevent the compiler from using NEON registers
Samuel Holland [Sun, 22 Aug 2021 00:54:02 +0000 (19:54 -0500)]
ARM: Prevent the compiler from using NEON registers

For ARMv8-A, NEON is standard, so the compiler can use it even when no
special target flags are provided. For example, it can use stores from
NEON registers to zero-initialize large structures. GCC 11 decides to
do this inside the DRAM init code for the Allwinner H6.

However, GCC 11 has a bug where it generates misaligned NEON register
stores even with -mstrict-align. Since the MMU is not enabled this early
in SPL, the misaligned store causes an exception and breaks booting.

Work around this issue by restricting the compiler to using GPRs only,
not vector registers. This prevents any future surprises relating to
NEON use as well.

Signed-off-by: Samuel Holland <samuel@sholland.org>
Acked-by: Andre Przywara <andre.przywara@arm.com>
3 years agodoc: Add documentation for the Arm vexpress board configs
Peter Hoyes [Thu, 19 Aug 2021 15:53:14 +0000 (16:53 +0100)]
doc: Add documentation for the Arm vexpress board configs

Create a new documentation section for Arm Ltd boards with a sub-page
for the vexpress board (FVP-A, FVP-R and Juno).

Document how the armv8_switch_to_el1 environment variable can be used
to switch between booting from S-EL2/S-EL1 at runtime on the BASER_FVP.

Signed-off-by: Peter Hoyes <Peter.Hoyes@arm.com>
3 years agoarm: Use armv8_switch_to_el1 env to switch to EL1
Peter Hoyes [Thu, 19 Aug 2021 15:53:13 +0000 (16:53 +0100)]
arm: Use armv8_switch_to_el1 env to switch to EL1

Use the environment variable armv8_switch_to_el1 to determine whether
to switch to EL1 at runtime. This is an alternative to the
CONFIG_ARMV8_SWITCH_TO_EL1 compile-time option.

The environment variable will be ineffective if the ARMV8_MULTIENTRY
config is used.

This is required by the Armv8r64 architecture, which must be able to
boot at S-EL1 for Linux but may need to boot at other ELs for other
systems.

Signed-off-by: Peter Hoyes <Peter.Hoyes@arm.com>
3 years agovexpress64: Add BASER_FVP vexpress board variant
Peter Hoyes [Thu, 19 Aug 2021 15:53:12 +0000 (16:53 +0100)]
vexpress64: Add BASER_FVP vexpress board variant

The BASER_FVP board variant is implemented on top of the BASE_FVP board
config (which, in turn, is based on the Juno Versatile Express board
config). They all share a similar memory map - for BASER_FVP the map is
inverted from the BASE_FVP
(https://developer.arm.com/documentation/100964/1114/Base-Platform/Base---memory/BaseR-Platform-memory-map)

 * Create new TARGET_VEXPRESS64_BASER_FVP target, which uses the same
   board config as BASE_FVP and JUNO
 * Adapt vexpress_aemv8a.h header file to support BASER_FVP (and rename
   to vexpress_aemv8.h)
 * Enable config to switch to EL1 for the BASER_FVP
 * Create vexpress_aemv8r defconfig
 * Provide an MPU memory map for the BASER_FVP

For now, only single core boot is supported.

Signed-off-by: Peter Hoyes <Peter.Hoyes@arm.com>
[trini: Add MAINTAINERS, move BOOTCOMMAND to defconfig]
Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoarmv8: Add ARMv8 MPU configuration logic
Peter Hoyes [Thu, 19 Aug 2021 15:53:11 +0000 (16:53 +0100)]
armv8: Add ARMv8 MPU configuration logic

Armv8r64 is the first Armv8 platform that only has a PMSA at the
current exception level. The architecture supplement for Armv8r64
describes new fields in ID_AA64MMFR0_EL1 which can be used to detect
whether a VMSA or PMSA is present. These fields are RES0 on Armv8a.

Add logic to read these fields and, for the protection of the memory
used by U-Boot, initialize the MPU instead of the MMU during init, then
clear the MPU regions before transition to the next stage.

Provide a default (blank) MPU memory map, which can be overridden by
board configurations.

Signed-off-by: Peter Hoyes <Peter.Hoyes@arm.com>
3 years agoarmv8: Ensure EL1&0 VMSA is enabled
Peter Hoyes [Thu, 19 Aug 2021 15:53:10 +0000 (16:53 +0100)]
armv8: Ensure EL1&0 VMSA is enabled

On Armv8-R, the EL1&0 memory system architecture is configurable as a
VMSA or PMSA, and resets to an "architecturally unknown" value.

Add code to armv8_switch_to_el1_m which detects whether the MSA at
EL1&0 is configurable using the id_aa64mmfr0_el1 register MSA fields.
If it is we must ensure the VMSA is enabled so that a rich OS can boot.

The MSA and MSA_FRAC fields are described in the Armv8-R architecture
profile supplement (section G1.3.7):
https://developer.arm.com/documentation/ddi0600/latest/

Signed-off-by: Peter Hoyes <Peter.Hoyes@arm.com>
3 years agoarmv8: Disable pointer authentication traps for EL1
Peter Hoyes [Thu, 19 Aug 2021 15:53:09 +0000 (16:53 +0100)]
armv8: Disable pointer authentication traps for EL1

The use of ARMv8.3 pointer authentication (PAuth) is governed by fields
in HCR_EL2, which trigger a 'trap to EL2' if not enabled. The reset
value of these fields is 'architecturally unknown' so we must ensure
that the fields are enabled (to disable the traps) if we are entering
the kernel at EL1.

The APK field disables PAuth instruction traps and the API field
disables PAuth register traps

Add code to disable the traps in armv8_switch_to_el1_m. Prior to doing
so, it checks fields in the ID_AA64ISAR1_EL1 register to ensure pointer
authentication is supported by the hardware.

The runtime checks require a second temporary register, so add this to
the EL1 transition macro signature and update 2 call sites.

Signed-off-by: Peter Hoyes <Peter.Hoyes@arm.com>
3 years agoRevert "configs: synquacer: Make U-Boot binary position independent"
Masami Hiramatsu [Thu, 19 Aug 2021 13:15:25 +0000 (22:15 +0900)]
Revert "configs: synquacer: Make U-Boot binary position independent"

This reverts commit f7e16bb0c5362c9b01d7e6e96bf6c77fd6b3d89e, since
the U-Boot doesn't boot if it is booted directly from SPI-NOR with
CONFIG_POSITION_INDEPENDENT=y. Unless fixing this issue, it is better
to revert this change.

Signed-off-by: Masami Hiramatsu <masami.hiramatsu@linaro.org>
3 years agolib: add crc16.o if CONFIG_MMC_SPI_CRC_ON for SPL build
Andreas Schwab [Tue, 17 Aug 2021 10:51:34 +0000 (12:51 +0200)]
lib: add crc16.o if CONFIG_MMC_SPI_CRC_ON for SPL build

CONFIG_MMC_SPI_CRC_ON needs the crc16 functions, but it was not included
in an SPL build.  For non-SPL builds, crc16.o is already added
unconditionally.  This also removes CONFIG_SPL_YMODEM_SUPPORT from the
sifive board configs, which is only relevant for some ARM boards and was
only set for its side effect of adding crc16.o.

3 years agocmd/part: correct alignment of partition list
Heinrich Schuchardt [Thu, 5 Aug 2021 22:51:46 +0000 (00:51 +0200)]
cmd/part: correct alignment of partition list

When running 'part list' for an ISO partition the numbers are not under the
labels.

Correct the alignment of the ISO partition list. With the patch the
output looks like:

    Part   Start     Sect x Size Type
      1     3720     5024    512 U-Boot

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
3 years agoarm: mach-snapdrgon: misc: Simplify msm_generate_mac_addr()
Stephan Gerhold [Tue, 3 Aug 2021 10:12:38 +0000 (12:12 +0200)]
arm: mach-snapdrgon: misc: Simplify msm_generate_mac_addr()

The logic in msm_generate_mac_addr() was originally taken from the LK
bootloader where the serial number is a string and must be parsed first.
However, in U-Boot msm_board_serial() returns an u32 and
msm_generate_mac_addr() has quite complicated code that will first
print it as a hex string and then immediately parse it again.

What this function actually does at the end is to put the serial number
encoded as big endian (the order used for the hex string) into the u8 *mac.
Use put_unaligned_be32() to do that with bit shifts instead of going
through the string format.

This should be slightly more efficient and cleaner but does not result
in any functional difference.

Cc: Ramon Fried <rfried.dev@gmail.com>
Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
3 years agoarm: mach-snapdragon: misc: Initialize eMMC if necessary
Stephan Gerhold [Tue, 3 Aug 2021 10:12:37 +0000 (12:12 +0200)]
arm: mach-snapdragon: misc: Initialize eMMC if necessary

At the moment U-Boot produces an empty MAC address (02:00:00:00:00:00)
if the eMMC is not used by anything in U-Boot (e.g. with
CONFIG_ENV_IS_NOWHERE=y instead of having the environment on eMMC).
This happens because then there is nothing that actually initializes
the eMMC and reads the "cid" that is later accessed.

To fix this, call mmc_init() to ensure the eMMC is initialized.
There is no functional difference if the eMMC is already initialized
since then mmc_init() will just return without doing anything.

Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
3 years agoarmv7: Add Position Independent Execution support
Chia-Wei Wang [Tue, 3 Aug 2021 02:50:10 +0000 (10:50 +0800)]
armv7: Add Position Independent Execution support

A U-Boot image could be loaded and executed at a different
location than it was linked at.

For example, Aspeed takes a stable release version of U-Boot image
as the golden one for recovery purposes. When the primary storage
such as flash is corrupted, the golden image would be loaded to any
SRAM/DRAM address on demands through ethernet/UART/etc and run for
rescue.

To deal with this condition, the PIE is needed as there is only one
signed, golden image, which could be however executed at different
places.

This patch adds the PIE support for ARMv7 platform.

Signed-off-by: Chia-Wei Wang <chiawei_wang@aspeedtech.com>
3 years agoarm: Fix option dependency with Kconfig language
Chia-Wei Wang [Tue, 3 Aug 2021 02:50:09 +0000 (10:50 +0800)]
arm: Fix option dependency with Kconfig language

Use Kconfig 'depends on' instead of #if macro to
express the option depdencies.

Signed-off-by: Chia-Wei Wang <chiawei_wang@aspeedtech.com>
3 years agoMerge branch '2021-09-02-assorted-fixes' into next
Tom Rini [Thu, 2 Sep 2021 13:25:43 +0000 (09:25 -0400)]
Merge branch '2021-09-02-assorted-fixes' into next

- Drop old OpenSSL support
- Add DM_HASH support, use it.
- Assorted "stemmy" platform updates
- Various bugfixes

3 years agoata: ahci-pci: Add new option CONFIG_SPL_AHCI_PCI
Pali Rohár [Sun, 15 Aug 2021 14:27:37 +0000 (16:27 +0200)]
ata: ahci-pci: Add new option CONFIG_SPL_AHCI_PCI

This new option allows to disable ahci-pci driver in SPL. Disabling it is
needed when SPL_PCI is not enabled as ahci-pci depends on PCI.

This change fixes following compile error when CONFIG_SPL_SATA_SUPPORT is
enabled and SPL_PCI is disabled.

    LD      spl/u-boot-spl
  arm-linux-gnueabihf-ld.bfd: drivers/ata/ahci.o: in function `ahci_probe_scsi_pci':
  drivers/ata/ahci.c:1205: undefined reference to `dm_pci_map_bar'
  arm-linux-gnueabihf-ld.bfd: drivers/ata/ahci.c:1215: undefined reference to `dm_pci_read_config16'
  arm-linux-gnueabihf-ld.bfd: drivers/ata/ahci.c:1216: undefined reference to `dm_pci_read_config16'
  arm-linux-gnueabihf-ld.bfd: drivers/ata/ahci.c:1220: undefined reference to `dm_pci_map_bar'
  make[1]: *** [scripts/Makefile.spl:512: spl/u-boot-spl] Error 1
  make: *** [Makefile:1977: spl/u-boot-spl] Error 2

    LD      spl/u-boot-spl
  arm-linux-gnueabihf-ld.bfd: drivers/ata/ahci-pci.o: in function `ahci_pci_probe':
  drivers/ata/ahci-pci.c:21: undefined reference to `ahci_probe_scsi_pci'
  make[1]: *** [scripts/Makefile.spl:512: spl/u-boot-spl] Error 1
  make: *** [Makefile:1977: spl/u-boot-spl] Error 2

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
3 years agoata: ahci-pci: Fix dependency on DM_PCI
Pali Rohár [Sun, 15 Aug 2021 14:27:36 +0000 (16:27 +0200)]
ata: ahci-pci: Fix dependency on DM_PCI

File drivers/ata/ahci-pci.c calls function ahci_probe_scsi_pci() which is
compiled only when DM_PCI is enabled. So add missing dependency into
Kconfig.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
3 years agortc: ds1307: Fix incorrect clock reset for DS13xx
Callum Sinclair [Tue, 10 Aug 2021 02:51:15 +0000 (14:51 +1200)]
rtc: ds1307: Fix incorrect clock reset for DS13xx

The ds1307 driver also supports the DS1339 and DS1340.
However, in ds1307_rtc_reset the register writes assume that the chip
is a DS1307. This is evident in the writing of bits SQWE, RS1, RS0 to
the control register. While this applies correctly to the DS1307, on a
DS1340 the control register doesn't contain those bits (instead, the
register is used for clock calibration). By writing these bits the
clock calibration will be changed and the chip can become
non-functional after a reset call.

Signed-off-by: Callum Sinclair <callum.sinclair@alliedtelesis.co.nz>
3 years agoarm: u8500: Prefer building in thumb mode by default
Stephan Gerhold [Sat, 7 Aug 2021 13:07:24 +0000 (15:07 +0200)]
arm: u8500: Prefer building in thumb mode by default

Enabling CONFIG_SYS_THUMB_BUILD produces a significantly smaller
U-Boot binary (250 KiB vs 320 KiB) that still seems to be fully
functional. Make use of that by default but keep it as "imply" so it
can be disabled for testing in case this causes trouble for someone.

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
3 years agoboard: stemmy: Update documentation
Stephan Gerhold [Sat, 7 Aug 2021 13:07:23 +0000 (15:07 +0200)]
board: stemmy: Update documentation

Over the time, the "stemmy" U-Boot board was tested on several other
Samsung smartphones based on ST-Ericsson NovaThor Ux500. Convert the
documentation to reStructuredText at doc/board/ste/stemmy.rst and
make the device list complete. Also note that the board now boots
into USB Fastboot instead of just ending up at the U-Boot prompt.

The device table is mostly taken from the postmarketOS wiki article
(https://wiki.postmarketos.org/wiki/ST-Ericsson_NovaThor_U8500).
All the newly added devices were tested by Linus Walleij.

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
3 years agoboard: stemmy: Add basic Fastboot support
Stephan Gerhold [Sat, 7 Aug 2021 13:07:22 +0000 (15:07 +0200)]
board: stemmy: Add basic Fastboot support

Make use of the new drivers for ARM U8500 introduced in the U-Boot
2021.10 merge window by adding basic support for USB Fastboot with
the "stemmy" board. As a first step this will always boot directly
into USB Fastboot for now with the console displayed on the screen
to make that obvious.

Samsung uses quite strange GPT partition labels on these boards,
so also add a bunch of fastboot_partition_alias_* to make this more
easy to use.

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
3 years agoarm: dts: u8500: u-boot: Add fixed clock for eMMC
Stephan Gerhold [Sat, 7 Aug 2021 13:07:21 +0000 (15:07 +0200)]
arm: dts: u8500: u-boot: Add fixed clock for eMMC

So far there is no need for a clock driver in U-Boot because the
previous boot stage leaves all the necessary clocks on. However,
some drivers in U-Boot (e.g. arm_pl180_mmci) depend on having a clock
driver to obtain the clock frequency.

Setting up the clock drivers properly is a bit tricky on U8500,
so for now add a simple fixed-clock for the eMMC that allows obtaining
the clock frequency. This should be replaced eventually if some board
actually requires enabling some of the clocks.

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
3 years agoarm: dts: u8500: Update from Linux ux500-dts-for-v5.15
Stephan Gerhold [Sat, 7 Aug 2021 13:07:20 +0000 (15:07 +0200)]
arm: dts: u8500: Update from Linux ux500-dts-for-v5.15

Update ste-dbx5x0.dtsi, ste-ab8500.dtsi and ste-ab8505.dtsi with
the changes made in upstream Linux. They are taken from
https://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-nomadik.git/
branch "ux500-dts-for-v5.15" (pending merge for Linux 5.15).

The only relevant change for U-Boot here is
"ARM: dts: ux500: ab8500: Link USB PHY to USB controller node" [1]
which links the USB PHY to the USB controller. This is necessary on
U-Boot because the PHY driver is implemented as part of the generic
PHY subsystem that makes use of these bindings.

[1]: https://lore.kernel.org/linux-arm-kernel/20210709182234.47232-1-stephan@gerhold.net/

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
3 years agoarm: u8500: Imply options for new drivers
Stephan Gerhold [Sat, 7 Aug 2021 13:07:19 +0000 (15:07 +0200)]
arm: u8500: Imply options for new drivers

Imply the options for new drivers added for ARM U8500 during the
U-Boot 2021.10 merge window. Adding these as "imply" in the Kconfig
avoids having to add them to all the board defconfigs but still allows
disabling them if wanted.

Also select DM_USB_GADGET if DM_USB is selected because otherwise
the Ux500 MUSB glue driver does not show up in the configuration.

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
3 years agoMAINTAINERS: Add new drivers for ARM U8500
Stephan Gerhold [Sat, 7 Aug 2021 13:07:18 +0000 (15:07 +0200)]
MAINTAINERS: Add new drivers for ARM U8500

Update MAINTAINERS with various drivers for ARM U8500 that were
added during the U-Boot 2021.10 merge window.

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
3 years agoloadb: Properly indicate aborted kermit transfer
Pali Rohár [Fri, 6 Aug 2021 16:07:39 +0000 (18:07 +0200)]
loadb: Properly indicate aborted kermit transfer

When k_recv() returns zero it indicates that kermit transfer was aborted.
Function do_load_serial_bin() (caller of load_serial_bin()) interprets
value ~0 as aborted transfer, so properly propagates information about
aborted transfer from k_recv() to do_load_serial_bin().

Signed-off-by: Pali Rohár <pali@kernel.org>
3 years agoscripts/checkpatch.pl: Resync with v5.13
Tom Rini [Tue, 3 Aug 2021 12:31:56 +0000 (08:31 -0400)]
scripts/checkpatch.pl: Resync with v5.13

This resyncs us with the version found in v5.13 of the Linux kernel with
the following exceptions:
- Keep our u-boot specific tests / code area.
- Change the location of checkpatch.rst (which we now import)
- Drop the "use strscpy" test as we don't have that, but do have strlcpy
  and want that used now.
- Keep debug/printf in the list for $logFunctions

And note that we now also include the spdxcheck.py tool that
checkpatch.pl supports calling out to, and include upstream's
checkpatch.rst in our develop section of the documentation.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agomkimage: clarify error message for empty input files
Thomas Hebb [Sun, 1 Aug 2021 22:23:13 +0000 (15:23 -0700)]
mkimage: clarify error message for empty input files

Currently, an empty imput file causes `mmap()` to fail, and you get an
error like "mkimage: Can't read file.img: Invalid argument", which is
extremely unintuitive and hard to diagnose if you don't know what to
look for. Add an explicit check for an empty file and provide a clear
error message instead.

We already bounds check the image size when listing and re-signing
existing images, so we only need this check here, when opening data
files going into a image.

Signed-off-by: Thomas Hebb <tommyhebb@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
3 years agolib: -Wformat-truncation in rsa_engine_get_priv_key
Heinrich Schuchardt [Fri, 30 Jul 2021 15:05:07 +0000 (17:05 +0200)]
lib: -Wformat-truncation in rsa_engine_get_priv_key

With glibc 2.33 (Ubuntu package glibc6 2.33-0ubuntu9) building
sifive_unmatched_defconfig results in:

In file included from /usr/include/stdio.h:866,
                  from ././include/compiler.h:26,
                  from <command-line>:
In function ‘snprintf’,
     inlined from ‘rsa_engine_get_priv_key’ at ./tools/../^:273:4:
/usr/include/riscv64-linux-gnu/bits/stdio2.h:71:10: warning:
‘%s’ directive argument is null [-Wformat-truncation=]
    71 |   return __builtin___snprintf_chk (__s, __n,
__USE_FORTIFY_LEVEL - 1,
       |
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
    72 |        __glibc_objsize (__s), __fmt,
       |        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
    73 |        __va_arg_pack ());
       |        ~~~~~~~~~~~~~~~~~

Avoid passing a NULL string.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
3 years agofit: Use DM hash driver if supported
Chia-Wei Wang [Fri, 30 Jul 2021 01:08:05 +0000 (09:08 +0800)]
fit: Use DM hash driver if supported

Calculate hash using DM driver if supported.
For backward compatibility, the call to legacy
hash functions is reserved.

Signed-off-by: Chia-Wei Wang <chiawei_wang@aspeedtech.com>
3 years agocrypto: hash: Add software hash DM driver
Chia-Wei Wang [Fri, 30 Jul 2021 01:08:04 +0000 (09:08 +0800)]
crypto: hash: Add software hash DM driver

Add purely software-implmented drivers to support multiple
hash operations including CRC, MD5, and SHA family.

This driver is based on the new hash uclass.

Signed-off-by: Chia-Wei Wang <chiawei_wang@aspeedtech.com>
3 years agodm: hash: Add new UCLASS_HASH support
Chia-Wei Wang [Fri, 30 Jul 2021 01:08:03 +0000 (09:08 +0800)]
dm: hash: Add new UCLASS_HASH support

Add UCLASS_HASH for hash driver development. Thus the
hash drivers (SW or HW-accelerated) can be developed
in the DM-based fashion.

Signed-off-by: Chia-Wei Wang <chiawei_wang@aspeedtech.com>
3 years agolib/md5: Export progressive APIs
Chia-Wei Wang [Fri, 30 Jul 2021 01:08:02 +0000 (09:08 +0800)]
lib/md5: Export progressive APIs

Export the MD5 hash init/update/finish progressive APIs
for better flexibility.

Signed-off-by: Chia-Wei Wang <chiawei_wang@aspeedtech.com>
3 years agolib/rsa: Remove support for OpenSSL < 1.1.0 and libressl < 2.7.0
Alexandru Gagniuc [Thu, 29 Jul 2021 18:31:21 +0000 (13:31 -0500)]
lib/rsa: Remove support for OpenSSL < 1.1.0 and libressl < 2.7.0

Older OpenSSL and libressl versions have a slightly different API.
This require #ifdefs to support. However, we still can't support it
because the ECDSA path does not compile with these older versions.
These #ifdefs are truly a vestigial appendage.

Alternatively, the ECDSA path could be updated for older libraries,
but this requires significant extra code, and #ifdefs. Those libraries
are over three years old, and there concerns whether it makes sense to
build modern software for real world use against such old libraries.

Thusly, remove #ifdefs and code for old OpenSSL and LibreSSL support.

Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
3 years agoMerge branch 'next' of https://gitlab.denx.de/u-boot/custodians/u-boot-marvell into...
Tom Rini [Tue, 31 Aug 2021 23:11:15 +0000 (19:11 -0400)]
Merge branch 'next' of https://gitlab.denx.de/u-boot/custodians/u-boot-marvell into next

- Handling all DM watchdogs in watchdog_reset() (Rasmus)

3 years agoMerge branch '2021-08-31-kconfig-migrations-part2' into next
Tom Rini [Tue, 31 Aug 2021 21:52:28 +0000 (17:52 -0400)]
Merge branch '2021-08-31-kconfig-migrations-part2' into next

- Further CONFIG to Kconfig migrations
  - Some DDR related symbols
  - CONFIG_SYS_LOAD_ADDR moved, loadaddr always set in environment now.
  - Finish MX7D, convert IMX_CONFIG
  - Some RAMBOOT related options
  - L1 cache size converted and named consistently for all arches.  A
    further follow-up to rename things for even better clarity is welcome.
  - CONFIG_SKIP_LOWLEVEL_INIT, CONFIG_SYS_MALLOC_LEN

3 years agoKconfig: Remove all default n/no options
Michal Simek [Fri, 27 Aug 2021 06:48:10 +0000 (08:48 +0200)]
Kconfig: Remove all default n/no options

default n/no doesn't need to be specified. It is default option anyway.

Signed-off-by: Michal Simek <michal.simek@xilinx.com>
[trini: Rework FSP_USE_UPD portion]
Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoConvert CONFIG_SYS_MALLOC_LEN to Kconfig
Tom Rini [Sun, 29 Aug 2021 01:34:49 +0000 (21:34 -0400)]
Convert CONFIG_SYS_MALLOC_LEN to Kconfig

This converts the following to Kconfig:
   CONFIG_SYS_MALLOC_LEN

Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Sean Anderson <seanga2@gmail.com>
Reviewed-by: Rick Chen <rick@andestech.com>
3 years agoConvert CONFIG_SKIP_LOWLEVEL_INIT et al to Kconfig
Tom Rini [Sat, 28 Aug 2021 01:18:30 +0000 (21:18 -0400)]
Convert CONFIG_SKIP_LOWLEVEL_INIT et al to Kconfig

This converts the following to Kconfig:
   CONFIG_SKIP_LOWLEVEL_INIT
   CONFIG_SKIP_LOWLEVEL_INIT_ONLY

In order to do this, we need to introduce SPL and TPL variants of these
options so that we can clearly disable these options only in SPL in some
cases, and both instances in other cases.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoFinish converting CONFIG_SYS_CACHELINE_SIZE to Kconfig
Tom Rini [Thu, 26 Aug 2021 15:47:59 +0000 (11:47 -0400)]
Finish converting CONFIG_SYS_CACHELINE_SIZE to Kconfig

We move the SYS_CACHE_SHIFT_N options from arch/arm/Kconfig to
arch/Kconfig, and introduce SYS_CACHE_SHIFT_4 to provide a size of 16.
Introduce select statements for other architectures based on current
usage.  For MIPS, we take the existing arch-specific symbol and migrate
to the generic symbol.  This lets us remove a little bit of otherwise
unused code.

Cc: Alexey Brodkin <alexey.brodkin@synopsys.com>
Cc: Anup Patel <anup.patel@wdc.com>
Cc: Atish Patra <atish.patra@wdc.com>
Cc: Bin Meng <bmeng.cn@gmail.com>
Cc: Leo <ycliang@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Sean Anderson <seanga2@gmail.com>
Cc: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Sean Anderson <seanga2@gmail.com>
Reviewed-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Reviewed-by: Rick Chen <rick@andestech.com>
3 years agols1046ardb_qspi*: Enable RAMBOOT_PBL
Tom Rini [Wed, 25 Aug 2021 03:11:50 +0000 (23:11 -0400)]
ls1046ardb_qspi*: Enable RAMBOOT_PBL

Based on include/configs/ls1046ardb.h it seems that CONFIG_RAMBOOT_PBL
should have been enabled, but was not.  Enable and migrate the values to
Kconfig.

Cc: Mingkai Hu <mingkai.hu@nxp.com>
Cc: Rajesh Bhagat <rajesh.bhagat@nxp.com>
Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoConvert CONFIG_RAMBOOT_PBL et al to Kconfig
Tom Rini [Wed, 25 Aug 2021 03:11:49 +0000 (23:11 -0400)]
Convert CONFIG_RAMBOOT_PBL et al to Kconfig

This converts the following to Kconfig:
   CONFIG_RAMBOOT_PBL
   CONFIG_SYS_FSL_PBL_PBI
   CONFIG_SYS_FSL_PBL_RCW

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoConvert CONFIG_QSPI_BOOT to Kconfig
Tom Rini [Wed, 25 Aug 2021 01:19:12 +0000 (21:19 -0400)]
Convert CONFIG_QSPI_BOOT to Kconfig

This converts the following to Kconfig:
   CONFIG_QSPI_BOOT

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoConvert CONFIG_SYS_FSL_DDR4 to Kconfig
Tom Rini [Wed, 25 Aug 2021 00:47:06 +0000 (20:47 -0400)]
Convert CONFIG_SYS_FSL_DDR4 to Kconfig

This converts the following to Kconfig:
   CONFIG_SYS_FSL_DDR4

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoconfigs: Remove unused IMX_NAND symbol
Tom Rini [Wed, 25 Aug 2021 00:41:04 +0000 (20:41 -0400)]
configs: Remove unused IMX_NAND symbol

The symbol CONFIG_IMX_NAND is not referenced in the code, remove it.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agonitrogen6x: Populate FDTFILE at build-time for all platforms
Tom Rini [Wed, 25 Aug 2021 00:41:03 +0000 (20:41 -0400)]
nitrogen6x: Populate FDTFILE at build-time for all platforms

Rather than using CONFIG_SABRELITE to set FDTFILE for only that
platform, switch to always setting this based on
CONFIG_DEFAULT_DEVICE_TREE as this should always match the kernel
device tree name anyhow.

Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Troy Kisky <troy.kisky@boundarydevices.com>
3 years agonitrogen6x: Migrate DDR_MB to Kconfig
Tom Rini [Wed, 25 Aug 2021 00:41:02 +0000 (20:41 -0400)]
nitrogen6x: Migrate DDR_MB to Kconfig

Move the CONFIG_DDR_MB symbol to Kconfig.  A later clean-up would be to
make dynamic memory size detection work based on how this is done on
other i.MX6 platforms.

Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Troy Kisky <troy.kisky@boundarydevices.com>
3 years agoConvert CONFIG_SPL to Kconfig
Tom Rini [Wed, 25 Aug 2021 00:41:01 +0000 (20:41 -0400)]
Convert CONFIG_SPL to Kconfig

This converts the following to Kconfig:
   CONFIG_SPL

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoimx: Finish migration of IMX_CONFIG to Kconfig
Tom Rini [Wed, 25 Aug 2021 00:41:00 +0000 (20:41 -0400)]
imx: Finish migration of IMX_CONFIG to Kconfig

- Provide a default Kconfig value of the default script
- Largely continue to define this via the board Kconfig file
- For the boards that select a script based on defconfig rather than
  TARGET, keep this within the defconfig.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoimx: Introduce CONFIG_MACH_IMX
Tom Rini [Wed, 25 Aug 2021 00:40:59 +0000 (20:40 -0400)]
imx: Introduce CONFIG_MACH_IMX

Currently, there is no over-arching symbol for access to
arch/arm/mach-imx nor the CONFIG symbols that are common over all of
these related platforms.  This new CONFIG symbol will allow us to start
down this path.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoConvert CONFIG_MX7D to Kconfig
Tom Rini [Mon, 23 Aug 2021 20:35:25 +0000 (16:35 -0400)]
Convert CONFIG_MX7D to Kconfig

This converts the following to Kconfig:
   CONFIG_MX7D

Cc: Oleksandr Suvorov <oleksandr.suvorov@toradex.com>
Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoConvert CONFIG_SYS_LOAD_ADDR to Kconfig
Tom Rini [Mon, 23 Aug 2021 14:25:31 +0000 (10:25 -0400)]
Convert CONFIG_SYS_LOAD_ADDR to Kconfig

Now that we have consistent usage, migrate this symbol to Kconfig.

Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Rick Chen <rick@andestech.com>
3 years agoglobal: Convert CONFIG_LOADADDR to CONFIG_SYS_LOADADDR
Tom Rini [Mon, 23 Aug 2021 14:25:30 +0000 (10:25 -0400)]
global: Convert CONFIG_LOADADDR to CONFIG_SYS_LOADADDR

- In most of the codebase, we reference CONFIG_SYS_LOAD_ADDR and not
  CONFIG_LOADADDR.
- Generally, CONFIG_SYS_LOADADDR is set to CONFIG_LOADADDR and then as
  noted, we use CONFIG_SYS_LOADADDR.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoqfw: Switch to CONFIG_SYS_LOAD_ADDR from CONFIG_LOADADDR
Tom Rini [Mon, 23 Aug 2021 14:25:29 +0000 (10:25 -0400)]
qfw: Switch to CONFIG_SYS_LOAD_ADDR from CONFIG_LOADADDR

All platforms define CONFIG_SYS_LOAD_ADDR, but only some define
CONFIG_LOADADDR.  Very very rarely are these not the same address, and
qemu-ppce500 is one such case.  However, based on reading the history of
the code, this mismatched value was simply a copy-paste from other
PowerPC platforms where it is this unused currently.  Switch the code to
use CONFIG_SYS_LOAD_ADDR and update the documentation.

Cc: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
3 years agoConvert CONFIG_SPL_MPC8XXX_INIT_DDR_SUPPORT to Kconfig
Tom Rini [Sat, 21 Aug 2021 17:50:19 +0000 (13:50 -0400)]
Convert CONFIG_SPL_MPC8XXX_INIT_DDR_SUPPORT to Kconfig

This converts the following to Kconfig:
   CONFIG_SPL_MPC8XXX_INIT_DDR_SUPPORT

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agonxp: Migrate a number of DDR related symbols to Kconfig
Tom Rini [Sat, 21 Aug 2021 17:50:18 +0000 (13:50 -0400)]
nxp: Migrate a number of DDR related symbols to Kconfig

- Guard most of the options in drivers/ddr/fsl/Kconfig with
  SYS_FSL_DDR || SYS_FSL_MMDC.
- Migrate FSL_DMA, DDR_ECC, DDR_ECC_CMD, and ECC_INIT_VIA_DDRCONTROLLER
  to Kconfig.
- Clean up the logic for including the DDR_ECC_CMD code.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agonxp: Migrate CONFIG_DDR_CLK_FREQ to Kconfig
Tom Rini [Sat, 21 Aug 2021 17:50:17 +0000 (13:50 -0400)]
nxp: Migrate CONFIG_DDR_CLK_FREQ to Kconfig

As this symbol can either be a fixed value or the function
get_board_ddr_clk, migration is tricky.  Introduce a choice of DYNAMIC
or STATIC_DDR_CLK_FREQ.  If DYNAMIC, we continue to use the board
defined get_board_ddr_clk function.  If STATIC, set CONFIG_DDR_CLK_FREQ
to that value and now include/clock_legacy.h contains the function
prototype or defines get_board_ddr_clk() to that static value.  Update
callers to test for DYNAMIC or STATIC.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoddr: Migrate DDR_SPD to Kconfig
Tom Rini [Sat, 21 Aug 2021 17:50:16 +0000 (13:50 -0400)]
ddr: Migrate DDR_SPD to Kconfig

Move the symbol that controls building some JEDEC SPD support functions
to Kconfig.  This is required on the TI keystone 2 platforms and very
frequently (but not always) used on large number of Freescale/NXP
platforms, so use imply there.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agomvebe: Migrate CONFIG_DDR_LOG_LEVEL to Kconfig
Tom Rini [Sat, 21 Aug 2021 17:50:15 +0000 (13:50 -0400)]
mvebe: Migrate CONFIG_DDR_LOG_LEVEL to Kconfig

Move this specific option to Kconfig.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agomvebu: ddr: Rename CONFIG_DDR_FIXED_SIZE to CONFIG_SYS_SDRAM_SIZE
Tom Rini [Sat, 21 Aug 2021 17:50:14 +0000 (13:50 -0400)]
mvebu: ddr: Rename CONFIG_DDR_FIXED_SIZE to CONFIG_SYS_SDRAM_SIZE

We have a number of CONFIG symbols to express the fixed size of system
memory.  For now, rename CONFIG_DDR_FIXED_SIZE to CONFIG_SYS_SDRAM_SIZE
and adjust usage to match that CONFIG_SYS_SDRAM_SIZE expects the entire
size rather than MiB.

Cc: Marek Behún <marek.behun@nic.cz>
Cc: Stefan Roese <sr@denx.de>
Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Marek Behún <marek.behun@nic.cz>
3 years agomvebu: Migrate CONFIG_DDR_32BIT/64BIT to Kconfig
Tom Rini [Sat, 21 Aug 2021 17:50:13 +0000 (13:50 -0400)]
mvebu: Migrate CONFIG_DDR_32BIT/64BIT to Kconfig

Move CONFIG_DDR_32BIT/64BIT to Kconfig as a choice for Armada XP
platforms.  Make 64bit the default as this mirrors the current code.

Signed-off-by: Tom Rini <trini@konsulko.com>
Acked-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
3 years agosandbox: add test of wdt-uclass' watchdog_reset()
Rasmus Villemoes [Thu, 19 Aug 2021 09:57:06 +0000 (11:57 +0200)]
sandbox: add test of wdt-uclass' watchdog_reset()

Check that the watchdog_reset() implementation in wdt-uclass behaves
as expected:

- resets all activated watchdog devices
- leaves unactivated/stopped devices alone
- that the rate-limiting works, with a per-device threshold

Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agosandbox: add test of wdt_gpio driver
Rasmus Villemoes [Thu, 19 Aug 2021 09:57:05 +0000 (11:57 +0200)]
sandbox: add test of wdt_gpio driver

It seems that no other test has claimed gpio_a:7 yet, so use that.

The only small wrinkle is modifying the existing wdt test to use
uclass_get_device_by_driver() since we now have two UCLASS_WDT
instances in play, so it's a little more robust to fetch the device by
driver and not merely uclass+index.

Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agowatchdog: add gpio watchdog driver
Rasmus Villemoes [Thu, 19 Aug 2021 09:57:04 +0000 (11:57 +0200)]
watchdog: add gpio watchdog driver

A rather common kind of external watchdog circuit is one that is kept
alive by toggling a gpio. Add a driver for handling such a watchdog.

The corresponding linux driver apparently has support for some
watchdog circuits which can be disabled by tri-stating the gpio, but I
have never actually encountered such a chip in the wild; the whole
point of adding an external watchdog is usually that it is not in any
way under software control. For forward-compatibility, and to make DT
describe the hardware, the current driver only supports devices that
have the always-running property. I went a little back and forth on
whether I should fail ->probe or only ->start, and ended up deciding
->start was the right place.

The compatible string is probably a little odd as it has nothing to do
with linux per se - however, I chose that to make .dts snippets
reusable between device trees used with U-Boot and linux, and this is
the (only) compatible string that linux' corresponding driver and DT
binding accepts. I have asked whether one should/could add "wdt-gpio"
to that binding, but the answer was no:

  https://lore.kernel.org/lkml/CAL_JsqKEGaFpiFV_oAtE+S_bnHkg4qry+bhx2EDs=NSbVf_giA@mail.gmail.com/

If someone feels strongly about this, I can certainly remove the
"linux," part from the string - it probably wouldn't the only place where
one can't reuse a DT snippet as-is between linux and U-Boot.

Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agowatchdog: wdt-uclass.c: handle all DM watchdogs in watchdog_reset()
Rasmus Villemoes [Thu, 19 Aug 2021 09:57:03 +0000 (11:57 +0200)]
watchdog: wdt-uclass.c: handle all DM watchdogs in watchdog_reset()

A board can have and make use of more than one watchdog device, say
one built into the SOC and an external gpio-petted one. Having
wdt-uclass only handle the first is both a little arbitrary and
unexpected.

So change initr_watchdog() so we visit (probe) all DM watchdog
devices, and call the init_watchdog_dev helper for each.

Similarly let watchdog_reset() loop over the whole uclass - each
having their own ratelimiting metadata, and a separate "is this device
running" flag.

This gets rid of the watchdog_dev member of struct global_data.  We
do, however, still need the GD_FLG_WDT_READY set in
initr_watchdog(). This is because watchdog_reset() can get called
before DM is ready, and I don't think we can call uclass_get() that
early.

The current code just returns 0 if "getting" the first device fails -
that can of course happen because there are no devices, but it could
also happen if its ->probe call failed. In keeping with that, continue
with the handling of the remaining devices even if one fails to
probe. This is also why we cannot use uclass_probe_all().

If desired, it's possible to later add a per-device "u-boot,autostart"
boolean property, so that one can do CONFIG_WATCHDOG_AUTOSTART
per-device.

Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agoboard: x530: switch to wdt_stop_all()
Rasmus Villemoes [Thu, 19 Aug 2021 09:57:02 +0000 (11:57 +0200)]
board: x530: switch to wdt_stop_all()

Since the gd->watchdog_dev member is going away, switch to using the
new wdt_stop_all() helper.

While here, clean up the preprocessor conditional: The ->watchdog_dev
member is actually guarded by CONFIG_WDT [disabling that in
x530_defconfig while keeping CONFIG_WATCHDOG breaks the build], and in
the new world order so is the existence of the wdt_stop_all()
function.

Actually, existence of wdt_stop_all() depends on CONFIG_${SPL_}WDT, so
really spell the condition using CONFIG_IS_ENABLED, and make it a C
rather than cpp if.

Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agowatchdog: wdt-uclass.c: add wdt_stop_all() helper
Rasmus Villemoes [Thu, 19 Aug 2021 09:57:01 +0000 (11:57 +0200)]
watchdog: wdt-uclass.c: add wdt_stop_all() helper

Since the watchdog_dev member of struct global_data is going away in
favor of the wdt-uclass handling all watchdog devices, prepare for
that by adding a helper to call wdt_stop() on all known devices.

If an error is encountered, still do wdt_stop() on remaining devices,
but remember and return the first error seen.

Initially, this will only be used in one single
place (board/alliedtelesis/x530/x530.c).

Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agosandbox: disable CONFIG_WATCHDOG_AUTOSTART
Rasmus Villemoes [Thu, 19 Aug 2021 09:57:00 +0000 (11:57 +0200)]
sandbox: disable CONFIG_WATCHDOG_AUTOSTART

For the unit tests, it is more convenient if the tests are in charge
of when the watchdog devices are started and stopped, so prevent
wdt-uclass from doing it automatically.

Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agowatchdog: wdt-uclass.c: keep track of each device's running state
Rasmus Villemoes [Thu, 19 Aug 2021 09:56:59 +0000 (11:56 +0200)]
watchdog: wdt-uclass.c: keep track of each device's running state

As a step towards handling all DM watchdogs in watchdog_reset(), use a
per-device flag to keep track of whether the device has been started
instead of a bit in gd->flags.

We will still need that bit to know whether we are past
initr_watchdog() and hence have populated gd->watchdog_dev -
incidentally, that is how it was used prior to commit 9c44ff1c5f3c.

Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agowatchdog: wdt-uclass.c: refactor initr_watchdog()
Rasmus Villemoes [Thu, 19 Aug 2021 09:56:58 +0000 (11:56 +0200)]
watchdog: wdt-uclass.c: refactor initr_watchdog()

In preparation for handling all DM watchdogs in watchdog_reset(), pull
out the code which handles starting (or not) the gd->watchdog_dev
device.

Include the device name in various printfs.

Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agowatchdog: wdt-uclass.c: neaten UCLASS_DRIVER definition
Rasmus Villemoes [Thu, 19 Aug 2021 09:56:57 +0000 (11:56 +0200)]
watchdog: wdt-uclass.c: neaten UCLASS_DRIVER definition

The addition of .pre_probe and .per_device_auto made this look
bad. Fix it.

Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agowatchdog: wdt-uclass.c: introduce struct wdt_priv
Rasmus Villemoes [Thu, 19 Aug 2021 09:56:56 +0000 (11:56 +0200)]
watchdog: wdt-uclass.c: introduce struct wdt_priv

As preparation for having the wdt-uclass provided watchdog_reset()
function handle all DM watchdog devices, and not just the first such,
introduce a uclass-owned struct to hold the reset_period and
next_reset, so these become per-device instead of being static
variables.

No functional change intended.

Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agowatchdog: wdt-uclass.c: use wdt_start() in wdt_expire_now()
Rasmus Villemoes [Thu, 19 Aug 2021 09:56:55 +0000 (11:56 +0200)]
watchdog: wdt-uclass.c: use wdt_start() in wdt_expire_now()

wdt_start() does the "no ->start? return -ENOSYS" check, don't
open-code that in wdt_expire_now().

Also, wdt_start() maintains some global (and later some per-device)
state, which would get out of sync with this direct method call - not
that it matters much here since the board is supposed to reset very
soon.

Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3 years agoglobal: Remove unused or unnecessary CONFIG symbols related to DDR
Tom Rini [Sat, 21 Aug 2021 17:50:12 +0000 (13:50 -0400)]
global: Remove unused or unnecessary CONFIG symbols related to DDR

These symbols are now either unused or were only used within the config
file to determine other logic, which could be done in a way that doesn't
further pollute the CONFIG namespace.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoMerge branch '2021-08-30-kconfig-migrations-part1' into next
Tom Rini [Mon, 30 Aug 2021 20:44:54 +0000 (16:44 -0400)]
Merge branch '2021-08-30-kconfig-migrations-part1' into next

- Begin merging some Kconfig migration, and CONFIG namespace cleanup
  series in.  This gives us:
  - A number of I2C symbols migrated over
  - DWC2, i8042, altera_spi and a few other areas updated to use CFG not
    CONFIG for the concept of "configuration space" defines.
  - Rename CONFIG_EXTRA_ENV_BOARD_SETTINGS to EXTRA_ENV_BOARD_SETTINGS
  - Some dead code removal.
  - Rename a number of CONFIG symbols that were only referenced within
    the config header to not use CONFIG as a prefix.

3 years agoKconfig: Use spaces not tabs in Kconfig entires
Tom Rini [Thu, 19 Aug 2021 20:23:58 +0000 (16:23 -0400)]
Kconfig: Use spaces not tabs in Kconfig entires

While the Kconfig language seems to accept either form of whitespace, we
use a space throughout the project, except in these spots.

Signed-off-by: Tom Rini <trini@konsulko.com>
3 years agoastro_mcf5373l: Rework ASTRO_ID logic
Tom Rini [Thu, 19 Aug 2021 19:16:15 +0000 (15:16 -0400)]
astro_mcf5373l: Rework ASTRO_ID logic

Rather than using CONFIG namespace for logic internal to
include/configs/astro_mcf5373l.h to select ASTRO_ID (and populate the
default environment), strip CONFIG from the various options used and
set.

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