Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored Aug 30, 2022 and Marek Behun committed Aug 30, 2022

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

This makes objdump reporting .ro section of wtmi.elf with READONLY flag.

Reported-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored Aug 23, 2022 and Marek Behun committed Aug 26, 2022

Topaz in indirect mode aliases data register 0x01 to all remaining MDIO
registers (0x02 - 0x1F). Check that values in registers 0x02 and 0x03 are
same to improve Topaz detection on some ESPRESSObin boards.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored Aug 23, 2022 and Marek Behun committed Aug 26, 2022

It seems that some devices have insufficient MDIO pull-up resistor,
which makes mdio_read() return 0x7fff instead of 0xffff when reading
from an even MDIO address when no device is responding on that address.

Check for non-existence of a MDIO device by reading from an odd MDIO
address to force MDIO to stay at high enough level to workaround this
insufficient pull-up issue.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored Aug 22, 2022 and Marek Behun committed Aug 26, 2022

Armada 3720 Devel Board has only one 88E1512 PHY, either at MDIO address
0x00 (in SGMII mode), or at address 0x01 (in RGMII mode).

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

It is now done when mox-imager's GPP code prints version number.

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored May 30, 2022 and Marek Behun committed May 30, 2022

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

This fixes issue whereupon the reset_handler symbol, defined in
startup_asm.S, was saved into the ISR vector without the LSB bit set,
thus not declaring the function as Thumb, which caused reload_helper to
failed to load payload after decompressing.

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

This is already done in wtmi, do it also for compressed and
reload_helper.

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored May 26, 2022 and Marek Behun committed May 26, 2022

This makes it possible to compile wtmi without libgcc when using
DEBUG_UART=1.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored Apr 15, 2022 and Marek Behun committed Apr 25, 2022

This decrease size of the final wtmi_app.bin binary with DEBUG_UART=1
option from 33200 bytes to just 31664 bytes.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored Mar 02, 2022 and Marek Behun committed Apr 07, 2022

Implement OTP write commands with the same API as in Marvell's fuse.bin
firmware.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored Oct 14, 2021 and Marek Behun committed Feb 28, 2022

Implement OTP read commands with the same API as in Marvell's fuse.bin
firmware.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored Oct 14, 2021 and Marek Behun committed Feb 28, 2022

Enable commands MBOX_CMD_OTP_READ and MBOX_CMD_OTP_WRITE. The write
command is for now enabled only in non-secure boot mode.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Fix function efuse_write_row_with_ecc_lock() for the case when ECC row
is masked:
- do not add bits from _efuse_read_row(), since if the row is masked,
  the bits are invalid; only program the new bits
- if the ECC row is masked, do not check for whether the ECC field for
  requested row is already programmed, since the check can't work
  because reading the masked ECC row returns invalid value
  0x0bad0acc0bad0acc

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Don't allow writing masked rows in efuse_write_row_with_ecc_lock().

To compute ECC code the row needs to be readable, so this function can't
work with masked rows.

Signed-off-by: Marek Behún <marek.behun@nic.cz>

In efuse_raw_write() we need first to read the OTP row, otherwise
writing won't work. We currently add the already written bits to the
value about to be written with bitwise OR. But if the row is masked,
reading will not return the real value of the row, but special value
0x0bad0acc0bad0acc. We can't add these bits to the value about to be
written.

Do not add bits returned from _efuse_read_row() to the values about to
be written.

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Let compiler decide whether this function should be inlined.

Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored Oct 14, 2021 and Marek Behun committed Feb 27, 2022

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>

Pali Rohár authored Oct 14, 2021 and Marek Behun committed Feb 27, 2022

Remove duplicate checks for row number in cmd_otp_read() and
cmd_otp_write().

Functions efuse_read_row_no_ecc() and efuse_write_row_no_ecc()
already check for valid row number.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <marek.behun@nic.cz>