The rpi_4_32b_defconfig states that only one DRAM bank is present. This
leads to a wrong configuration of the available DRAM. Fix this by
setting the DRAM bank config accordingly.

Fixes: 193279d7 ("RPI: Add defconfigs for rpi4 (32/64)")

Signed-off-by: Matthias Brugger <mbrugger@suse.com>

Up to now we only update the DRAM banks when we are define
CONFIG_BCM2711. But our one binary approach uses a config that supports
BCM2837 and BCM2711. As a result we only see one gibibyte of RAM on
Raspberry Pi 4, even if it has more RAM.
Fix this by calling dram_init_banksize.

Fixes: 56940906 ("ARM: defconfig: add unified config for RPi3 and RPi4")

Signed-off-by: Matthias Brugger <mbrugger@suse.com>

To update the dram bank information from device-tree we use
fdtdec_decode_ram_size() which expectes the the size-cells and
address-cells to be defined in the memory node. For normal system RAM
these values are defined in the root node. When the values differ from
the default values defined in the spec, we can end up with wrong RAM
bank information.

Switch to the "standard" way to update the RAM bank information to
avoid this.

Fixes: 9de5b89e ("rpi4: enable dram bank initialization")

Signed-off-by: Matthias Brugger <mbrugger@suse.com>

- Increase stack size to avoid a stack overflow during distro boot.
- Add hifive-unleashed-a00.dts for SIFIVE FU540.
- Add OF_SEPARATE support for SIFIVE FU540.
- Add SPL support for Andes AX25 AE350.
- Improve U-Boot SPL / OpenSBI smp boot flow for RISC-V.

At the start, OpenSBI relocates itself to its link address. If the link
address ranges of U-Boot SPL and OpenSBI overlap, the relocation can
lead to code corruption if a hart is still running U-Boot SPL during
relocation. To avoid this problem, the main hart is specified as the
preferred boot hart to perform the relocation. This fixes the code
corruption problems based on the assumption that since the main hart
schedules the secondary harts to enter OpenSBI, it will be the last to
enter OpenSBI. However it was reported that this assumption is not
always correct.

To make sure the assumption always holds true, wait for all secondary
harts to acknowledge the call-function request before entering OpenSBI
on the main hart.

Reported-by: Rick Chen <rick@andestech.com>
Signed-off-by: Lukas Auer <lukas.auer@aisec.fraunhofer.de>
Reviewed-by: Rick Chen <rick@andestech.com>
Tested-by: Rick Chen <rick@andestech.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

Add a wait option to smp_call_function() to wait for the secondary harts
to acknowledge the call-function request. The request is considered to
be acknowledged once each secondary hart has cleared the corresponding
IPI.

As part of the call-function request, the secondary harts invalidate the
instruction cache after clearing the IPI. This adds a delay between
acknowledgment (clear IPI) and fulfillment (call function) of the
request. We want to use the acknowledgment to be able to judge when the
request has been completed. Remove the delay by clearing the IPI after
cache invalidation and just before calling the function from the
request.

Signed-off-by: Lukas Auer <lukas.auer@aisec.fraunhofer.de>
Reviewed-by: Rick Chen <rick@andestech.com>
Tested-by: Rick Chen <rick@andestech.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

Add the function riscv_get_ipi() for reading the pending status of IPIs.
The supported controllers are Andes' Platform Level Interrupt Controller
(PLIC), the Supervisor Binary Interface (SBI), and SiFive's Core Local
Interruptor (CLINT).

Signed-off-by: Lukas Auer <lukas.auer@aisec.fraunhofer.de>
Reviewed-by: Rick Chen <rick@andestech.com>

OpenSBI uses a relocation lottery to determine the hart to relocate
OpenSBI to its link address. In the U-Boot SPL boot flow, the main hart
schedules the secondary harts to enter OpenSBI before doing so itself.
One of the secondary harts will therefore always be the winner of the
relocation lottery. This is problematic if the link address ranges of
OpenSBI and U-Boot SPL overlap. OpenSBI will be relocated and therefore
overwrite U-Boot SPL while some harts may still run it, leading to code
corruption.

Avoid this problem by specifying the main hart as the preferred boot
hart to perform the OpenSBI relocation. The main hart will be the last
hart to enter OpenSBI, relocation can therefore occur safely.

The boot hart field was added to version 2 of the OpenSBI FW_DYNAMIC
info structure. The header file include/opensbi.h is synchronized with
include/sbi/fw_dynamic.h from the OpenSBI project to update the info
structure. The header file is recent as of commit
7a13beb21326 ("firmware: Add preferred boot HART field in struct
fw_dynamic_info").

Reported-by: Rick Chen <rick@andestech.com>
Suggested-by: Anup Patel <Anup.Patel@wdc.com>
Signed-off-by: Lukas Auer <lukas.auer@aisec.fraunhofer.de>
Reviewed-by: Rick Chen <rick@andestech.com>
Tested-by: Rick Chen <rick@andestech.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

Add descriptions about U-Boot SPL feature and how to build and run.

Signed-off-by: Rick Chen <rick@andestech.com>
Cc: KC Lin <kclin@andestech.com>
Cc: Alan Kao <alankao@andestech.com>

Those are required for cfi-flash driver to get correct address information.
Also modify size description correctly.

Signed-off-by: Rick Chen <rick@andestech.com>
Cc: KC Lin <kclin@andestech.com>
Cc: Alan Kao <alankao@andestech.com>

Add CPU2 and CPU3 information in cpus node
to support four cores SMP booting.

Signed-off-by: Rick Chen <rick@andestech.com>
Cc: KC Lin <kclin@andestech.com>
Cc: Alan Kao <alankao@andestech.com>

For RV64, it will use sd instruction to clear t0
register, and the increament will be 8 bytes. So
if the difference between__bss_strat and __bss_end
was not 8 bytes aligned, the clear bss loop will
overflow and acks like system hang.

Signed-off-by: Rick Chen <rick@andestech.com>
Cc: KC Lin <kclin@andestech.com>
Cc: Alan Kao <alankao@andestech.com>

When ax25-ae350 try to enable v5l2 cache
driver in SPL configuration, it need this
option for cache support in SPL.

Signed-off-by: Rick Chen <rick@andestech.com>
Cc: KC Lin <kclin@andestech.com>
Cc: Alan Kao <alankao@andestech.com>

The mcache_ctl csr only can be manipulated in M mode.
Add SPL_RISCV_MMODE for U-Boot SPL to control cache
operation.

Signed-off-by: Rick Chen <rick@andestech.com>
Cc: KC Lin <kclin@andestech.com>
Cc: Alan Kao <alankao@andestech.com>

Fix two wrong settings of andes plic driver as below:

1. Fix wrong pending register base definition.
2. Declaring the en variable in enable_ipi() as unsigned int instead of
   int can help to fix wrong plic enabling setting in RV64.

Signed-off-by: Rick Chen <rick@andestech.com>
Cc: KC Lin <kclin@andestech.com>
Cc: Alan Kao <alankao@andestech.com>

To get memory size from device tree instead of
get_ram_size(). This can avoid memory access fault
in U-Boot proper after PMP configurations in OpenSBI.

Signed-off-by: Rick Chen <rick@andestech.com>
Cc: KC Lin <kclin@andestech.com>
Cc: Alan Kao <alankao@andestech.com>

This patch provides four configurations which can support U-Boot SPL
to boot from RAM or FLASH and then boot FIT image including OpenSBI
FW_DYNAMIC firmware and U-Boot proper images from RAM or MMC boot devices.

With ae350_rv[32|64]_spl_defconfigs:

U-Boot SPL will be loaded by gdb or FSBL and runs in RAM in machine mode
and then load FIT image from RAM device on AE350.

With ae350_rv[32|64]_spl_xip_defconfigs:

U-Boot SPL can be burned into SPI flash and run in flash in machine mode
and then load FIT image from SPI flash or MMC device on AE350.

Signed-off-by: Rick Chen <rick@andestech.com>
Cc: KC Lin <kclin@andestech.com>
Cc: Alan Kao <alankao@andestech.com>

The U-Boot SPL will boot in M mode and load the FIT image which
include OpenSBI and U-Boot proper images. After loading progress,
it will jump to OpenSBI first and then U-Boot proper which will
run in S mode.

Also remove V5L2_CACHE due to U-Boot SPL code size consideration.
Without this concern, it can be enable manually for performance.

Signed-off-by: Rick Chen <rick@andestech.com>
Cc: KC Lin <kclin@andestech.com>
Cc: Alan Kao <alankao@andestech.com>

This would help to make the necessary changes in drivers and device trees
in U-Boot tree itself. This feature would also be helpful to not pass
dtb during opensbi builds.

Signed-off-by: Jagan Teki <jagan@amarulasolutions.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Rick Chen <rick@andestech.com>

Sync the hifive-unleashed-a00 dts from Linux with
below commit details:

commit <2993c9b04e616df0848b655d7202a707a70fc876> ("riscv: dts: HiFive
Unleashed: add default chosen/stdout-path")

Idea is to periodically sync the dts from Linux instead of
tweaking internal changes one after another, so better not
add any intermediate changes in between. This would help to
maintain the dts files easy and meaningful since we are
reusing device tree files from Linux.

Signed-off-by: Jagan Teki <jagan@amarulasolutions.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

This fixes a problem, where booting Linux using distro boot will
sometimes lead to an invalid instruction exception on the main hart. The
secondary harts are not affected and boot Linux successfully. The root
cause of this problem is a stack overflow on the main hart.

With distro boot, the current default stack size of 8KiB on RISC-V is
not sufficient and will cause a stack overflow. The stacks are allocated
sequentially. In the case of a stack overflow the stack of the main hart
can reach into that of another hart and be corrupted.

The stack overflow previously did not cause any problems, because only
stack frames, which are not used anymore since the hart enters Linux,
were corrupted. Starting with GCC 9, the stack usage has decreased. Now,
only the most recent stack frame overflows into the stack of a secondary
hart and is corrupted. The illegal instruction exception is caused by
the secondary hart overwriting the return address in the stack frame of
the main hart with an address that does not include valid code.

Increase the default stack size of each hart to 16KiB to avoid this
problem.

Reported-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Lukas Auer <lukas.auer@aisec.fraunhofer.de>
Tested-by: David Abdurachmanov <david.abdurachmanov@sifive.com>
Tested-by: Aurelien Jarno <aurelien@aurel32.net>
Reviewed-by: Rick Chen <rick@andestech.com>

The commit 1b42ab3e ("ARM: DRA7: Fixup DSPEVE, IVA and GPU clock
frequencies based on OPP") added the core logic to update the kernel
device-tree blob to adjust the DSP, IVA and GPU DPLL clocks based on
a one-time OPP choice selected in U-Boot for most of the DRA7xx/AM57xx
family of SoCs.

The DSPs on DRA76xP/DRA77xP SoCs (DRA76x ACD package SoCs) though
provide a higher performance and can run at a higher clock frequency
of 850 MHz at OPP_HIGH instead of 750 MHz. Fix up the logic to use the
correct clock rates on these SoCs. Note that this higher clock rate is
not applicable to other Jacinto 6 Plus SoCs (DRA75xP/DRA74xP SoCs or
AM574x SoCs) that follow the ABZ package.

Signed-off-by: Suman Anna <s-anna@ti.com>
Reviewed-by: Lokesh Vutla <lokeshvutla@ti.com>

On NixOS, cross compiled kernels have long suffixes that cause them to
exceed the current maximum path length. The PXE/TFTP max path length is
used for extlinux.conf support as well, which is where this problem
usually manifest's itself.

Signed-off-by: Ben Wolsieffer <benwolsieffer@gmail.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>

Enable configs related to DMA and Ethernet so as to support networking at
U-Boot prompt

Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Reviewed-by: Grygorii Strashko <grygorii.strashko@ti.com>

Add DT nodes related to DMA and CPSW to -u-boot.dtsi to get networking
up on J721e EVM.

Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Reviewed-by: Grygorii Strashko <grygorii.strashko@ti.com>

Add new compatible to handle J721e SoC

Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Reviewed-by: Grygorii Strashko <grygorii.strashko@ti.com>

Get flow ID information for RX DMA channel using dma_get_cfg() interface
instead of reading from DT. This is required in order to avoid DT update
whenever there is change in the range of flow ID allocated to the host.

Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Reviewed-by: Grygorii Strashko <grygorii.strashko@ti.com>

Implement dma_get_cfg() interface to pass flow id information for DMA
clients to use. This is needed because on K3 SoCs, CPSW (ethernet) and
UDMA (DMA provider) support "flows" within a given RX DMA channel. This
allows different network packets to be segregated while using same RX
DMA channel. In order for basic ethernet to work, CPSW slave must be
aware of the flow ID allocated for the RX channel by the DMA driver.
This interface allows CPSW to query flow ID from DMA provider and
configure it in CPSW HW.

Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Reviewed-by: Grygorii Strashko <grygorii.strashko@ti.com>

Sometimes, there would be a need to exchange data between DMA provider
and DMA client which are very specific to DMA driver of the SoC/platform
and are not generic enough to be put into struct dma. Therefore, introduce
dma_get_cfg() interface to get DMA provider specific data from client
device. Clients can use unique configuration ID flags to get different
configuration data from DMA driver.

Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Reviewed-by: Grygorii Strashko <grygorii.strashko@ti.com>

Fixes: e7668497 ("net: ftgmac100: convert the RX/TX descriptor arrays")
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Joel Stanley <joel@jms.id.au>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>

This bus is used to access internal SoC PHYs.  These PHYs are configured
by the ENETC driver directly, but it's useful to have command line access
to this MDIO to debug the system especially when using new external PHYs.

Signed-off-by: Alex Marginean <alexandru.marginean@nxp.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>

On some boards MDIO may get stuck if it detects echo on the line.  This is
a know hardware issue, there is a board fix for it.  In case we're running
on a board that doesn't have the fix, we don't want to loop here forever
and freeze U-Boot.

Signed-off-by: Alex Marginean <alexandru.marginean@nxp.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>

This reduces the time needed to establish a link as we don't reset the link
each time the interface is used.  Our Link capabilities do not change at
run-time so there is no need to re-apply PHY configuration each time.

Signed-off-by: Alex Marginean <alexandru.marginean@nxp.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>

The original code enabled link speeds up to 1Gbps, but the interface can
go up to 2.5G, enable that speed to in PHY AN mask.

Signed-off-by: Alex Marginean <alexandru.marginean@nxp.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>

Uses the new dm_eth_phy_connect helper to connect to the PHY to simplify
the code.

Signed-off-by: Alex Marginean <alexandru.marginean@nxp.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>

The function connects an ethernet device to a PHY using DT information.
This API is only available for eth devices with an associated device tree
node.

Signed-off-by: Alex Marginean <alexandru.marginean@nxp.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>

Renamed dm_mdio_phy_connect arguments dev to mdiodev and addr to phyaddr
for a bit more clarity and consistency with the following patches.
Also use NULL instead of 0 on error return path.

Signed-off-by: Alex Marginean <alexandru.marginean@nxp.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>

Commit b618b370 ("net: Convert CONFIG_TFTP_BLOCKSIZE to Kconfig")
accidentally set the default *option* TFTP block size to 512 bytes, even
though the comment in the code says that this is a terrible choice. Most
boards didn't define the symbol before, so they got the default block size
of 1468 bytes before, but now use 512 bytes, which is also the fallback.
This leads to both abysmal performance and a lot of hashes printed
on the screen (one character for every 5K), which is both annoying and
slow over serial links.

Set the default block size in Kconfig back to the value it had before.

This improves TFTP performance from 2.8 MB/s to 6.9 MB/s on a Pine64.

Fixes: b618b370 ("net: Convert CONFIG_TFTP_BLOCKSIZE to Kconfig")
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>

Apply 10G PCS init for USXGMII, XFI interface types.

Signed-off-by: Alex Marginean <alexandru.marginean@nxp.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>