The main reason is to send pmufw cfg overlay from U-Boot to PMUFW to enable
access to DP. Overlay is sent when cls command is called and for that IP
has to be enabled in carrier cards.
And IP needs to be also enabled in SOM dt because with DTB reselection new
DT is not parsed in pre reloc U-Boot instance. It is called from board_f
via embedded_dtb_select(). That's why bind function is not able to allocate
memory and it ends up with error:
"Video device 'display@fd4a0000' cannot allocate frame buffer memory
-ensure the device is set up before relocation"
To avoid this situation DP is placed also to SOM where bind function is
called and frame buffer memory is allocated and just reused after DTB
reselection. Result is the same. There could be a problem in Linux with
different DP configurations but that's need to be solved there because
console should be on from u-boot already.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Link: https://lore.kernel.org/r/c4f31641f917fddb09d976f56875057c658f264c.1645629459.git.michal.simek@xilinx.com
Use ethernet-phy-id compatible string to properly describe phy reset on
kv260 boards. Previous description wasn't correct because reset was done
for mdio bus to operate and it was in this case used for different purpose
which was eth phy reset. With ethernet-phy-id phy reset happens only for
the phy via phy framework.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Link: https://lore.kernel.org/r/73b64f1a2b873b4e26bd2b365364bdf313794ae2.1645629459.git.michal.simek@xilinx.com
With limited low level configuration done via psu-init only IPs connected
on SOM are initialized and configured. All IPs connected to carrier card
are not initialized. There is a need to do proper reset, pin configuration
and also clock setting.
The patch targets the last part which is setting up proper clock for USBs
and SDs.
Also setup proper bus width for SD cards.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: Sai Krishna Potthuri <lakshmi.sai.krishna.potthuri@xilinx.com>
Link: https://lore.kernel.org/r/d9f80b2551bd246c3d7ecb09b516806c8dc83ed9.1645629459.git.michal.simek@xilinx.com
slg7xl45106 is i2c based 8-bit gpo expander, gpo pins are set and get by
writing and reading corresponding gpo bit value into its data register.
Signed-off-by: T Karthik Reddy <t.karthik.reddy@xilinx.com>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Link: https://lore.kernel.org/r/839f475cc75c97ffb3496a4caa93de2faabdbca2.1645629688.git.michal.simek@xilinx.com
In the workaround added with 'commit b6f44082d5 ("mmc: zynq_sdhci: Wait
till sd card detect state is stable")' the timeout variable has post
decrement. Whenever timeout happens, this post decrement is making
timeout=0xffffffff, so timeout error print and return statement are
never reached. Fix it by decrementing it inside the while loop.
Signed-off-by: Ashok Reddy Soma <ashok.reddy.soma@xilinx.com>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Link: https://lore.kernel.org/r/61fc1160ada0dd622cd29e381a74af7bf3d9a200.1645625609.git.michal.simek@xilinx.com
Add support to read MAC addresses from mac address multirecord.
Check if multi record is found, then jump to mac address multirecord by
comparing the record type field. If it matches mac address
multirecord(0xD2), then copy mac addresses.
Copy these read MAC address in xilinx_read_eeprom_fru so that they are
updated to eth*addr in board_late_init_xilinx().
Signed-off-by: Ashok Reddy Soma <ashok.reddy.soma@xilinx.com>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Link: https://lore.kernel.org/r/18f31bc528820934854ea5fd9dc581778fc1e40c.1645624855.git.michal.simek@xilinx.com
In board_late_init_xilinx() eth*addr are updated from the values read from
eeprom. Ideally the MAC addresses are updated sequencially. So if any
MAC address is invalid, it means there are no further valid values.
So optimise this logic by replacing continue with break.
Signed-off-by: Ashok Reddy Soma <ashok.reddy.soma@xilinx.com>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Link: https://lore.kernel.org/r/efef0d07add5d5777396ea111ad75411dc402db3.1645624855.git.michal.simek@xilinx.com
fru_checksum function is simply adding all the bytes and returning the
sum. If the data passed to this function is all zero's then it will
return 0, and the functions calling this api will assume that checksum
is correct. Ideally this is not good. Fix this by returning error if all
the data is 0's.
Signed-off-by: Ashok Reddy Soma <ashok.reddy.soma@xilinx.com>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Link: https://lore.kernel.org/r/ac0366fe55c60a818a3f9ed33d96826c817d5520.1645624855.git.michal.simek@xilinx.com
Since OpenSSL 1.1.0, EVP_MD_CTX_create() is EVP_MD_CTX_new()
EVP_MD_CTX_destroy() is EVP_MD_CTX_free()
EVP_MD_CTX_init() is EVP_MD_CTX_reset()
As there's no need to reset a newly created EVP_MD_CTX, moreover
EVP_DigestSignInit() does the reset, thus call to EVP_MD_CTX_init()
can be dropped.
As there's no need to reset an EVP_MD_CTX before it's destroyed,
as it will be reset by EVP_MD_CTX_free(), call to EVP_MD_CTX_reset()
is not needed and can be dropped.
Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Based on looking at top contributors it was seen that top statistics from
top contributors don't include all contributions from different email
addresses. That's why I checked all top contributors are checked it.
git shortlog -n $START..$END -e -s
The patch is adding mapping for Bin Meng, Marek Vasut, Masahiro Yamada,
Michal Simek, Tom Rini, Wolfgang Denk.
And also use mapping for Stefan Roese and Wolfgang Denk to be properly
counted.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Acked-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
If parameter -F is given but FIT support is missing, a NULL pointer might
dereferenced (Coverity CID 350249).
If incorrect parameters are given, provide a message and show usage.
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Recent unrelated fixes (9876ae7db6) revealed that we were missing bits
from 2af181b53e in the IOT2050 dt. Add them, but only for main U-Boot.
SPL loads from QSPI only, thus cannot use DMA.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
We only want to call do_board_detect() if CONFIG_TI_I2C_BOARD_DETECT
is set. Same as done for am64.
This makes it possible to add a custom am65 based board design to
U-Boot that does not use this board detection mechanism.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Currently, any u-boot bootloader for ti armv7 platforms using
DEFAULT_FIT_TI_ARGS to boot with a fitimage (boot_fit = 1)
doesn't boot when built with Yocto Poky (openembedded-core).
## Loading kernel from FIT Image at 90000000 ...
Could not find configuration node
ERROR: can't get kernel image!
Arago forked the kernel-fitimage class [1] and altered the
configuration nodes naming while adding the OPTEE support by
using FITIMAGE_CONF_BY_NAME by default [2].
The "upstream" kernel-fitimage class from openembedded-core still
add the "conf-" prefix for each configuration nodes [3].
The ITS file format (from doc/uImage.FIT/source_file_format.txt)
is not really accurate with the expected naming of these nodes.
But in practice the "conf-" prefix is widely used.
When the FIT image support has been added for ti armv7 platforms
the naming from Arago has been used [3]. Fix this issue by adding
the prefix expected by the ITS file generated by kernel-fitimage
class from openembedded-core.
[1] http://arago-project.org/git/meta-arago.git?p=meta-arago.git;a=commitdiff;h=719ab1b2098bcdc59c249e3529fa82cb1b9130e6
[2] http://arago-project.org/git/meta-arago.git?p=meta-arago.git;a=commitdiff;h=f23f2876a0cda89241d031bb7ba0b4256ed90035
[3] https://git.openembedded.org/openembedded-core/tree/meta/classes/kernel-fitimage.bbclass?h=yocto-3.1.13#n290
[3] 1e93cc8473
Signed-off-by: Romain Naour <romain.naour@smile.fr>
Cc: Tom Rini <trini@konsulko.com>
Reviewed-by: Denys Dmytriyenko <denys@konsulko.com>
Public documents about BootROM of some Marvell SoCs are available in the
public Web Archive. Put this information into source code.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
Testes proved that current kwboot version supports also Avanta SoCs.
It looks like that Avanta SoCs are using same kwbimage format as Armada.
Signed-off-by: Pali Rohár <pali@kernel.org>
Tested-by: Tony Dinh <mibodhi@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
Document -D, -b, -d, -q and -s options.
Add common examples how to use kwboot.
Add information about Armada 38x BootROM bug for debug console mode and how
to workaround it.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
Add all supported Armada SoCs and document -b and -d options in usage.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
Marvell BootROM recognize only '\b' byte as backspace. Use terminfo
for retrieving current backspace sequence and replace any occurrence of
backspace sequence by the '\b' byte.
Reading terminfo database is possible via tigetstr() function from system
library libtinfo.so.*. So link kwboot with -ltinfo.
Normally terminfo functions are in <term.h> system header file. But this
header file conflicts with U-Boot "termios_linux.h" header file. So declare
terminfo functions manually.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
-d option is currently broken. In most cases BootROM does not detect this
message pattern. For sending debug message pattern it is needed to do same
steps as for boot message pattern.
Implement sending debug message pattern via same separate thread like it is
for boot message pattern.
Checking if BootROM entered into UART debug mode is different than
detecting UART boot mode. When in boot mode, BootROM sends xmodem NAK
bytes. When in debug mode, BootROM activates console echo and reply back
every written byte (extept \r\n which is interpreted as executing command
and \b which is interpreting as removing the last sent byte).
So in kwboot, check that BootROM send back at least 4 debug message
patterns as a echo reply for debug message patterns which kwboot is sending
in the loop.
Then there is another observation, if host writes too many bytes (as
command) then BootROM command line buffer may overflow after trying to
execute such long command. To workaround this overflow, it is enough to
remove bytes from the input line buffer by sending 3 \b bytes for every
sent character. So do it.
With this change, it is possbile to enter into the UART debug mode with
kwboot -d option.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
After BootROM successfully detects boot message pattern on UART it waits
until host stop sending data on UART. For example Armada 385 BootROM
requires that host does not send anything on UART at least 24 ms. If host
is still sending something then BootROM waits (possibly infinitely).
BootROM successfully detects boot message pattern if it receives it in
small period of time after power on.
So to ensure that host put BootROM into UART boot mode, host must send
continuous stream of boot message pattern with a small gap (for A385 at
least 24 ms) after series of pattern. But this gap cannot be too often or
too long to ensure that it does not cover whole BootROM time window when it
is detecting for boot message pattern.
Therefore it is needed to do following steps in cycle without any delay:
1. send series of boot message pattern over UART
2. wait until kernel transmit all data
3. sleep small period of time
At the same time, host needs to monitor input queue, data received on the
UART and checking if it contains NAK byte by which BootROM informs that
xmodem transfer is ready.
But it is not possible to wait until kernel transmit all data on UART and
at the same time in the one process to also wait for input data. This is
limitation of POSIX tty API and also by linux kernel that it does not
provide asynchronous function for waiting until all data are transmitted.
There is only synchronous variant tcdrain().
So to correctly implement this handshake on systems with linux kernel, it
is needed to use tcdrain() in separate thread.
Implement sending of boot message pattern in one thread and reading of
reply in the main thread. Use pthread library for threads.
This change makes UART booting on Armada 385 more reliable. It is possible
to start kwboot and power on board after minute and kwboot correctly put
board into UART boot mode.
Old implementation without separate thread has an issue that it read just
one byte from UART input queue and then it send 128 message pattern to the
output queue. If some noise was on UART then kwboot was not able to read
BootROM response as its input queue was just overflowed and kwboot was
sending more data than receiving.
This change basically fixed above issue too.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
Function kwboot_debugmsg() is always called with kwboot_msg_debug as msg
and function kwboot_bootmsg() with kwboot_msg_debug as msg. Function
kwboot_bootmsg() is never called with NULL msg.
Simplify, cleanup and remove dead code.
No functional change.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
Variable msg_req_delay is set but never used. So completely remove it.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
Failure of kwboot_tty_send() and tcflush() functions is fatal, it does not
make sense to continue. So return error back to the caller like in other
places where are called these functions.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
Commit 369e532691 ("ddr: marvell: a38x: allow board specific ODT
configuration") added the odt_config member to struct
mv_ddr_topology_map ahead of the clk_enable and ck_delay members. This
means that any boards that configured either of clk_enable or ck_delay
needed to have their board topology updated. This affects the x530 and
clearfog boards. Other A38x boards don't touch any of the trailing
members of mv_ddr_topology_map so don't need updating.
Fixes: 369e532691 ("ddr: marvell: a38x: allow board specific ODT configuration")
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Acked-by: Baruch Siach <baruch@tkos.co.il>
Reviewed-by: Stefan Roese <sr@denx.de>
Replace null pointer by pointer to device registers when calling
armada38x_rtc_write.
Signed-off-by: Francois Berder <fberder@outlook.fr>
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>
Enable ext4 write support in Turris Omnia's defconfig. Some users find
it useful.
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Use "MVPCIE_" prefix instead of generic "PCIE_" prefix for pci_mvebu.c
specific macros. Define offset macros for Root Port registers and use
standard register macros from pci.h when accessing Root Port registers.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Macro SELECT() is unused and struct mvebu_pcie field lane_mask is unused
too. Remove them.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Stefan Roese <sr@denx.de>
Sometimes kwboot after quitting terminal prints error message:
terminal: Bad address
This is caused by trying to call write() syscall with count of (size_t)-1
bytes.
When quit sequence is split into more read() calls then number of input
bytes (nin) at the end of cycle can underflow and be negative. Fix it.
Fixes: de7514046e ("tools: kwboot: Fix detection of quit esc sequence")
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
