(Please visit the site to view this file)
(Please visit the site to view this file)
Hello,
We are testing the SecureBoot feature with our SPI1 flash. The chip select pin is SPI1_SCS0 as required by SPRAB41E.
A total of four files have been generated with the following commands:
- SecureHexAIS_OMAP-L138 -ini "none_busWidth.ini" -otype binary -o "aNone.bin" "ARM_testBoot.out"
- SecureHexAIS_OMAP-L138 -ini "SPIMASTER_AddrWidth.ini" -otype binary -o "aSPI.bin" "ARM_testBoot.out"
- SecureHexAIS_OMAP-L138 -ini "none_busWidth.ini" -otype binary -o "dNone.bin" "DSP_testBoot.out"
- SecureHexAIS_OMAP-L138 -ini "SPIMASTER_AddrWidth.ini" -otype binary -o "dSPI.bin" "DSP_testBoot.out"
So there are actually one DSP .out file and one ARM .out file with exactly the same code: toggling a multitude of GPIO pins with while(delay), even without using timer. The pins being toggled are
GPIO {0_1,0_2,0_5,0_15,1_9,1_11,1_13,5_6,5_12,6_2,6_3,6_14,7_0,7_2,7_3,7_4,7_5,7_6,7_7,7_9,7_10,7_11,7_12,7_13,7_14,7_15,8_2,8_6,8_8,8_9,8_10,8_11,8_13}
All ARM content is allocated and linked into 128KB shared RAM (start at 0x80000000), and all DSP content is linked to L2 (start at 0x11800000) so their running doesn’t require DDR setting. At 300MHz the ARM toggles GPIO pins at about 800Hz, and DSP toggles them at about 6000Hz.
After we burn the flash and attach them to sockets on to ZWTA3E (secure version) boards, none booted successfully. Oscilloscope detects a brief period of SPI_CLK pulses (about a dozen μs) upon power on the CLK pin of the flash, but then stops permanently. Then we remove them from secure boards and examine the flash content, we found the image was not altered at all, which means the structure of the Key Data was not encrypted by KEK in the eFuse.
The “SecureHexAIS_OMAP-L138” was not altered/recompiled, and is the same as downloaded from TI.
Regarding Secure Kernel API, none were included in the code of the GPIO toggling applications.
We also verified the four files all begin with 41504954h 58535920h BE40C0DEh. In addition, the secure board itself is perfect fine with BOOTMODE pins wired for SPI1 flash booting, and have successfully booted many images generated by non-secure AIS_GEN on non-secure L138 processors.
We are very frustrated as we have been testing with SecureBoot for days, and have re-made board with removable Flash sockets with quite some investment, but the experiments doesn’t go through.
The content of the two .ini files are here attached, and we are also sending .out and .bin files privately, soliciting TI expert’s help.
Jeff