A number of CONFIG symbols have crept in that are never referenced in
code, so drop them here. Further, we have two symbols being enabled
in headers while already enabled correctly in Kconfig, so these lines
can also be removed.
Signed-off-by: Tom Rini <trini@konsulko.com>
The rest of the unmigrated CONFIG symbols in the CONFIG_SYS namespace do
not easily transition to Kconfig. In many cases they likely should come
from the device tree instead. Move these out of CONFIG namespace and in
to CFG namespace.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
The rest of the unmigrated CONFIG symbols in the CONFIG_SYS_SDRAM
namespace do not easily transition to Kconfig. In many cases they likely
should come from the device tree instead. Move these out of CONFIG
namespace and in to CFG namespace.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
The RD-AC5X-32G16HVG6HLG-A0 development board main components and
features include:
* Main 12V/54V power supply
* 270 Gbps throughput packet processor on the main board
* DDR4:
* SR1: 2GB DDR4 2400MT/S(1GB x 2 pcs ) with ECC(1GB x 1 pcs)
* SR2: 4GB DDR4 2400MT/S(2GB x 2 pcs ) with ECC(2GB x 1 pcs)
* PCB co-layout with 4GB device to support 8GB (Dual CS) requirement
* 16GB eMMC (Samsung KLMAG1JETD-B041006)
* 16MB SPI NOR(GD25Q127C)
* 32 x 1000 Base-T interfaces
* 16 x 2500 Base-T interfaces
* SR1: 88E2540*4
* SR2: 88E2580*1+88E2540*2
* Six (6) x 25G Base-R SFP28 interfaces
* One (1) x RJ-45 console connector, interfacing to the on board UART
* One (1) x USB Type-A connector, interfacing to the USB 2.0 port (0)
* One (1) x USB Type-mini B connector, interfacing to the USB 2.0 port (1)
* One (1) x RJ-45 1G Base-T Management port, interfacing to the host
port (shared with PCIe) Connected to 88E1512 Gigabit Ethernet Phy
* One (1) x Oculink port, interfacing to the PCIe port for external CPU
connection
* POE 802.3AT support on Port 1 ~ Port 32, 802.3BT support on Port 33 ~
Port 48 (Microsemi PD69208T4, PD69208M or TI TPS2388,TPS23881
solution)
* POE total power budget 780W
* LED interfaces per network port/POE
* LED interfaces (common) showing system status
* PTP TC mode Supported (Reserved M.2 connector to support BC mode)
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Add support for the Allecat5/Alleycat5X SoC. These are L3 switches with
an integrated CPU (referred to as the CnM block in Marvell's
documentation). These have dual ARMv8.2 CPUs (Cortex-A55). This support
has been ported from Marvell's SDK which is based on a much older
version of U-Boot.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
VHV_Enable GPIO is required to enable during eFuse programming on Armada
SoCs not from 3700 family. Add support for enabling and disabling VHV pin
via GPIO during eFuse programming, when specified.
All details are in Marvell AN-389: ARMADA VHV Power document
(Doc. No. MV-S302545-00 Rev. C, August 2, 2016).
Note that due to HW Errata 3.6 eFuse erroneous burning (Ref #: HWE-3718342)
VHV power must be disabled while core voltage is off to prevent erroneous
eFuse programming.
This is specified in Marvell ARMADA 380/385/388 Functional Errata,
Guidelines, and Restrictions document
(Doc. No. MV-S501377-00 Rev. D, December 1, 2016).
Signed-off-by: Pali Rohár <pali@kernel.org>
This patch implements LD eFuse programming support. Armada 385 contains two
LD eFuse lines, each is 256 bit long with one additional lock bit. LD 0
line is mapped to U-Boot fuse bank 64 and LD 1 line to fuse bank 65. U-Boot
32-bit fuse words 0-8 are mapped to LD eFuse line bits 0-255. U-Boot fuse
word 9 is mapped to LD eFuse line lock bit.
So to program LD 1 General Purpose Data line, use U-Boot fuse command:
=> fuse prog -y 65 0 0x76543210
=> fuse prog -y 65 1 0xfedcba98
=> fuse prog -y 65 2 0x76543210
=> fuse prog -y 65 3 0xfedcba98
=> fuse prog -y 65 4 0x76543210
=> fuse prog -y 65 5 0xfedcba98
=> fuse prog -y 65 6 0x76543210
=> fuse prog -y 65 7 0xfedcba98
=> fuse prog -y 65 8 0x1
Signed-off-by: Pali Rohár <pali@kernel.org>
Since the move to CONFIG_TIMER with support for CONFIG_TIMER_EARLY, this
platform specific init_timer() function is not needed any more. Let's
remove it completely.
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Michael Walle <michael@walle.cc>
Cc: Pali Rohár <pali@kernel.org>
Commit 37dc958947 ("global_data.h: Change ram_top type to phys_addr_t")
changed type of ram_top member from ulong to phys_addr_t but did not
changed types in board_get_usable_ram_top() function which returns value
for ram_top.
So change ulong to phys_addr_t type also in board_get_usable_ram_top()
signature and implementations.
Fixes: 37dc958947 ("global_data.h: Change ram_top type to phys_addr_t")
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Simon Glass <sjg@chromium.org>
Now that the new timer support is available for these platforms, let's
select this IF for all these platforms. This way it's not necessary
that each board changes it's config header.
Signed-off-by: Stefan Roese <sr@denx.de>
Tested-by: Tony Dinh <mibodhi@gmail.com>
Commit 5bb2c550b1 ("arm: mvebu: Move internal registers in
arch_very_early_init() function") moved code from file cpu.c to lowlevel.c,
which moves Marvell internal registers from address INTREG_BASE_ADDR_REG to
SOC_REGS_PHY_BASE.
But the steps describing how to do it correctly were documented only in
older U-Boot versions and commit cefd764222 ("arm: mvebu: Fix internal
register config on A38x") probably unintentionally removed important
details about MMU from code comments around.
Commit 5bb2c550b1 ("arm: mvebu: Move internal registers in
arch_very_early_init() function") implemented code movement according to
(now incomplete) comments which resulted in semi-broken code.
The result is that I-cache is currently disabled for all Armada 38x boards
and maybe there are some other (unreported / undetected) issues.
Reimplement it correctly. First flush all caches, then disable MMU and L2
cache and then move Marvell internal registers. There is no need to
explicitly disable I-cache.
After this change lzmadec command with lzma image of 0x7000000 bytes is
doing decompression just 5 seconds. Before this change it was 30 seconds.
To make lowlevel.S code more readable, extend asm/pl310.h header file to be
compatible with assembler and use macros from this file.
Fixes: 5bb2c550b1 ("arm: mvebu: Move internal registers in arch_very_early_init() function")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
For some unknown reason when L2 cache is disabled on Armada 385 then loadb,
loadx and loady commands do not work with higher baudrates than 115200
(they just abort transfer) and lzmadec command with lzma image of size
0x7000000 (maybe even smaller, we tested this one) is doing decompression
for more than 2 minutes. After enabling L2 cache decompression takes only
30s and loadb, loadx and loady are stable and working fine.
git bisect identified problematic commit 3308933d2f ("arm: mvebu: Avoid
reading MVEBU_REG_PCIE_DEVID register too many times"). Before this commit
above issues were not present.
But investigation showed that above issue was possible to reproduce also by
reverting that commit and forcing compiler to do inline optimization of
mvebu_soc_family() function. Which seems that the root of this issue is in
caches and position of instruction of segments. So currently it is unknown
what is or was broken, but code movement, code inlining or other compiler
optimization triggered it.
Commit 3e5ce7ceeb ("arm: mvebu: Enable L2 cache on Armada XP") mentioned
that enabling L2 cache on Armada XP improved performance and that Armada
38x has L2 disabled (which is default state) and if needed it has to be
enabled in separate patch. As enabling L2 cache also improve performance
on Armada 38x, enable it.
Note that Aurora cache in no outer mode is available only on Armada XP,
hence it is not touched for Armada 38x code.
Fixes: 3308933d2f ("arm: mvebu: Avoid reading MVEBU_REG_PCIE_DEVID register too many times")
Reported-by: Marek Behún <kabel@kernel.org>
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Replace magic constant 1 when disabling MMU by macro CR_M from include
header file asm/system.h.
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Commit c86d53fd88 ("arm: mvebu: Don't disable cache at startup on Armada
XP at all") introduced branch for non-AXP code which was guarded by A38X
condition. Fix this issue by checking for AXP platform, not by A38X.
Fixes: c86d53fd88 ("arm: mvebu: Don't disable cache at startup on Armada XP at all")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Commit 3308933d2f ("arm: mvebu: Avoid reading MVEBU_REG_PCIE_DEVID
register too many times") broke support for caches on all Armada SoCs.
Before that commit there was code:
if (mvebu_soc_family() != MVEBU_SOC_A375) {
dcache_enable();
}
And after that commit there is code:
if (IS_ENABLED(CONFIG_ARMADA_375)) {
dcache_enable();
}
Comment above this code says that d-cache should be disabled on Armada 375.
But new code inverted logic and broke Armada 375 and slowed down all other
Armada SoCs (including A38x).
Fix this issue by changing logic to:
if (!IS_ENABLED(CONFIG_ARMADA_375)) {
dcache_enable();
}
Which matches behavior prior that commit.
Fixes: 3308933d2f ("arm: mvebu: Avoid reading MVEBU_REG_PCIE_DEVID register too many times")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
This converts CONFIG_SYS_L2_PL310 to Kconfig.
For omap2 and mvebu the 'select SYS_L2_PL310' locations were
determined using ./tools/moveconfig -i CONFIG_SYS_L2_PL310.
For mx6 I manually chose ARCH_MX6 as 'select' location. The
correctness has been verified using
$ ./tools/moveconfig.py -f ARCH_MX6 ~SYS_L2_PL310 ~SYS_L2CACHE_OFF
0 matches
That means whenever an ARCH_MX6 board had SYS_L2_PL310 disabled, this
was correctly reflected in SYS_L2CACHE_OFF. Thus it's safe to insert
the 'select' statement under ARCH_MX6.
Signed-off-by: Philip Oberfichtner <pro@denx.de>
If we switch to CONFIG_TIMER, we don't need the legacy timer macros and
functions anymore. Add the proper guards to exclude them from compiling.
Cc: Pali Rohár <pali@kernel.org>
Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
mbus driver is initialized from arch_cpu_init() callback which is called
before relocation. This driver stores lot of functions and structure
pointers into global variables, so it is data position dependent.
Therefore after relocations all pointers are invalid and driver does not
work anymore as all pointers referes to the old memory, which overlaps with
CONFIG_SYS_LOAD_ADDR and ${loadaddr}.
For example U-Boot fuse command crashes if loadaddr memory is cleared or
rewritten by some image loaded by U-Boot load command.
mw.w ${loadaddr} 0x0 10000
fuse read 0 1 2
Fix this issue by removing of all mbus global variables in which are stored
pointers to structures or functions which changes during relocation. And
replace it by direct function calls (not via pointers). With this change
fuse command finally works.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Reviewed-by: Stefan Roese <sr@denx.de>
SPI0 CS0 Flash is mapped to address range 0xD4000000 - 0xD7FFFFFF by BootROM.
Proper U-Boot removes this direct mapping. So it is available only in SPL.
This applies for all 32-bit Armada BootROMs. SPL mvebu code is used only on
32-bit Armada SoCs. So move env_sf_get_env_addr() function from Turris
Omnia board to common SPL mvebu code and add proper checks for SPI0 CS0.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
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>
Change detection of platform/cpu from runtime to compile time via config
define. This completely eliminates compiling code which is not going to run
on selected platform. Code which parses and prints device / revision id
still reads device id from MVEBU_REG_PCIE_DEVID register, but only once.
Signed-off-by: Pali Rohár <pali@kernel.org>
Acked-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Fix diacritics in some instances of my name and change my e-mail address
to kabel@kernel.org.
Add corresponding .mailmap entries.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
board_get_usable_ram_top() conflated the RAM size with the top address
of RAM. On systems where RAM starts at address 0 these numbers are the
same so it went unnoticed. Update board_get_usable_ram_top() to take
CONFIG_SYS_SDRAM_BASE into account when determining the top address.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Update the way KWB_CFG_SEC_BOOT_DEV is determined to use
CONFIG_MVEBU_SPL_BOOT_DEVICE_{SPI,MMC} instead of
CONFIG_SPL_BOOT_DEVICE.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
- Make all users of CUSTOM_SYS_INIT_SP_ADDR reference SYS_INIT_SP_ADDR
- Introduce HAS_CUSTOM_SYS_INIT_SP_ADDR to allow for setting the stack
pointer directly, otherwise we use the common calculation.
- On some platforms that were using the standard calculation but did not
set CONFIG_SYS_INIT_RAM_SIZE / CONFIG_SYS_INIT_RAM_ADDR, set them.
- On a small number of platforms that were not subtracting
GENERATED_GBL_DATA_SIZE do so now via the standard calculation.
- CONFIG_SYS_INIT_SP_OFFSET is now widely unused, so remove it from most
board config header files.
Signed-off-by: Tom Rini <trini@konsulko.com>
The definition of CONFIG_SPL_BOOTROM_SAVE is always a fixed
CONFIG_SPL_STACK + 4, while CONFIG_SPL_STACK is not constant. This
change will make it clear where the location is still, once
CONFIG_SPL_STACK moves to Kconfig.
Cc: Stefan Roese <sr@denx.de>
Signed-off-by: Tom Rini <trini@konsulko.com>
Moving of internal registers from INTREG_BASE_ADDR_REG to SOC_REGS_PHY_BASE
needs to be done very early, prior calling any function which may touch
internal registers, like debug_uart_init().
So do it earlier in arch_very_early_init() instead of arch_cpu_init().
Movement is done in proper U-Boot, not in SPL. SPL may return to bootrom
and bootrom requires internal registers at (old) expected location.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Nothing selects ARMADA_64BIT. Instead the 64-bit SoCs just select ARM64
directly. Remove the unused config item.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Implement write support for Security OTP values via mailbox API commands
MBOX_CMD_OTP_WRITE_32B and MBOX_CMD_OTP_WRITE.
Write support for North and South Bridge OTPs are not implemented as these
OTPs are already burned in factory with some data.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
U-Boot SPL is on 32-bit mvebu executed by the BootROM. And BootROM expects
that U-Boot SPL returns execution back to the BootROM. Vectors during
execution of U-Boot SPL should not be changed as BootROM does not expect it
and uses its own vectors. So do not overwrite vectors in SPL build.
Signed-off-by: Pali Rohár <pali@kernel.org>
Armada 385 contains 64 lines of HD eFuse and 2 lines of LD eFuse. HD eFuse
is used for secure boot and each line is 64 bits long + 1 lock bit. LD
eFuse lines are 256 bits long + 1 lock bit. LD 0 line is reserved for
Marvell Internal Use and LD 1 line is for General Purpose Data. U-Boot
already contains HD eFuse reading and programming support.
This patch implements LD eFuse reading support. LD 0 line is mapped to
U-Boot fuse bank 64 and LD 1 line to fuse bank 65.
LD 0 Marvell Internal Use line seems that was burned in factory with some
data and can be read by U-Boot fuse command:
=> fuse read 64 0 9
LD 1 General Purpose Data line is by default empty and can be read by
U-Boot fuse command:
=> fuse read 65 0 9
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
It is not possible for the A53 core (on which U-Boot is running) to read it
directly. For this purpose Marvell defined mbox API for sending OTP
commands between CM3 and A53 cores.
Implement these Marvell fuse reading mbox commands via U-Boot fuse API.
Banks 0-43 are used for accessing Security OTP (44 rows with 67 bits via 44
banks and words 0-2).
Note that of the 67 bits, the 3 upper bits are: 1 lock bit and 2
auxiliary bits (meant for testing during the manufacture of the SOC, as
I understand it).
Also note that the lock bit and the auxiliary bits are not readable
via Marvell commands.
With CZ.NIC's commands the lock bit is readable.
Write support is not implemented yet.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Allow to specify input parameters, define all available mbox commands
supported by CZ.NIC's secure firmware and also Marvell's fuse.bin firmware
and fix parsing response from Marvell OTP commands.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Generic A3720 mbox code is currently in Turris Mox specific board file
board/CZ.NIC/turris_mox/mox_sp.c. Move it to board independent arch file
arch/arm/mach-mvebu/armada3700/mbox.c.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Implement reading NB and SB fuses of Armada 37xx SOC via U-Boot fuse API.
Banks 0-43 are reserved for accessing Security OTP (not implemented yet).
Bank 44 is used for accessing North Bridge OTP (69 bits via words 0-2).
Bank 45 is used for accessing South Bridge OTP (97 bits via words 0-3).
Write support is not implemented yet because it looks like that both North
and South Bridge OTPs are already burned in factory with some data. The
meaning of some bits of North Bridge is documented in WTMI source code.
The meaning of bits in South Bridge is unknown.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
This symbol is only possibly useful on some mvebu platforms, so move the
symbol there and select it for the only current user.
Cc: Mario Six <mario.six@gdsys.cc>
Cc: Stefan Roese <sr@denx.de>
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Stefan Roese <sr@denx.de>
This register is referenced in one location and does not seem
configurable, so remove it from CONFIG namespace.
Cc: Stefan Roese <sr@denx.de>
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Stefan Roese <sr@denx.de>
A big part is the DM pinctrl driver, which allows us to get rid of quite
some custom pinmux code and make the whole port much more robust. Many
thanks to Samuel for that nice contribution! There are some more or less
cosmetic warnings about missing clocks right now, I will send the trivial
fixes for that later.
Another big chunk is the mkimage upgrade, which adds RISC-V and TOC0
(secure images) support. Both features are unused at the moment, but I
have an always-secure board that will use that once the DT lands in the
kernel.
On top of those big things we have some smaller fixes, improving the
I2C DM support, fixing some H6/H616 early clock setup and improving the
eMMC boot partition support.
The gitlab CI completed successfully, including the build test for all
161 sunxi boards. I also boot tested on a A64, A20, H3, H6, and F1C100
board. USB, SD card, eMMC, and Ethernet all work there (where applicable).
Platforms can overwrite the weak definition of spl_mmc_boot_mode() to
determine where to load U-Boot proper from.
For most of them this is a trivial decision based on Kconfig variables,
but it might be desirable the probe the actual device to answer this
question.
Pass the pointer to the mmc struct to that function, so implementations
can make use of that.
Compile-tested for all users changed.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Stefano Babic <sbabic@denx.de>
Reviewed-by: Ley Foon Tan <ley.foon.tan@inte.com> (for SoCFPGA)
Acked-by: Lokesh Vutla <lokeshvutla@ti.com> (for OMAP and K3)
Reviewed-by: Simon Glass <sjg@chromium.org>
The value CONFIG_DB_784MP_GP is only used in the DDR code to refer to
CONFIG_TARGET_DB_MV784MP_GP so just use that second value directly.
Cc: Stefan Roese <sr@denx.de>
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Stefan Roese <sr@denx.de>
The a3700_fdt_fix_pcie_regions() function still computes nonsense.
It computes the fixup offset from the PCI address taken from the first
row of the "ranges" array, which means that:
- PCI address must equal CPU address (otherwise the computed fix offset
will be wrong),
- the first row must contain the lowest address.
This is the case for the default device-tree, which is why we didn't
notice it.
It also adds the fixup offset to all PCI and CPU addresses, which is
wrong.
Instead:
1) The fixup offset must be computed from the CPU address, not PCI
address.
2) The fixup offset must be computed from the row containing the lowest
CPU address, which is not necessarily contained in the first row.
3) The PCI address - the address to which the PCIe controller remaps the
address space as seen from the point of view of the PCIe device -
must be fixed by the fix offset in the same way as the CPU address
only in the special case when the CPU adn PCI addresses are the same.
Same addresses means that remapping is disabled, and thus if we
change the CPU address, we need also to change the PCI address so
that the remapping is still disabled afterwards.
Consider an example:
The ranges entries contain:
PCI address CPU address
70000000 EA000000
E9000000 E9000000
EB000000 EB000000
By default CPU PCIe window is at: E8000000 - F0000000
Consider the case when TF-A moves it to: F2000000 - FA000000
Until now the function would take the PCI address of the first entry:
70000000, and the new base, F2000000, to compute the fix offset:
F2000000 - 70000000 = 82000000, and then add 8200000 to all addresses,
resulting in
PCI address CPU address
F2000000 6C000000
6B000000 6B000000
6D000000 6D000000
which is complete nonsense - none of the CPU addresses is in the
requested window.
Now it will take the lowest CPU address, which is in second row,
E9000000, and compute the fix offset F2000000 - E9000000 = 09000000,
and then add it to all CPU addresses and those PCI addresses which
equal to their corresponding CPU addresses, resulting in
PCI address CPU address
70000000 F3000000
F2000000 F2000000
F4000000 F4000000
where all of the CPU addresses are in the needed window.
Fixes: 4a82fca8e3 ("arm: a37xx: pci: Fix a3700_fdt_fix_pcie_regions() function")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
