AMD supports secure booting on all devices using latest authentication methods to prevent unauthorized or modified code from being run on AMD devices. AMD supports various encryption techniques to make sure only authorized programs access the images. For hardware security features by device, see the following sections.
Secure and Non-Secure Modes in Zynq 7000 SoC Devices
For security reasons, CPU 0 is always the first device out of reset among all master modules within the PS. CPU 1 is held in an WFE state. While the BootROM is running, the JTAG is always disabled, regardless of the reset type, to ensure security. After the BootROM runs, JTAG is enabled if the boot mode is non-secure.
The BootROM code is also responsible for loading the FSBL/User code. When the BootROM releases control to stage 1, your software assumes full control of the entire system. The only way to execute the BootROM again is by generating one of the system resets. The FSBL/User code size (encrypted/unencrypted) is limited to 192 KB. This limit does not apply with the non-secure execute-in-place option.
The PS boot source is selected using the BOOT_MODE
strapping pins (indicated by a weak pull-up or pull-down
resistor), that are sampled once during power-on reset (POR). The sampled values are
stored in the slcr.BOOT_MODE
register.
The BootROM supports encrypted/authenticated, and unencrypted images referred to as secure boot and non-secure boot, respectively. The BootROM supports execution of the stage 1 image directly from NOR or Quad-SPI when using the execute-in-place (xip_mode) option, but only for non-secure boot images. Execute-in-place is possible only for NOR and Quad-SPI boot modes.
- In secure boot, the CPU, running the BootROM code decrypts and authenticates the user PS image on the boot device, stores it in the OCM, and then branches to it.
- In non-secure boot, the CPU, running the BootROM code disables all secure boot features including the AES unit within the PL before branching to the user image in the OCM memory or the flash device (if execute-in-place (XIP) is used).
Any subsequent boot stages for either the PS or the PL are the responsibility of you, the developer, and are under your control. The BootROM code is not accessible to you. Following a stage 1 secure boot, you can proceed with either secure or non-secure subsequent boot stages. Following a non-secure first stage boot, only non-secure subsequent boot stages are possible.
Zynq UltraScale+ MPSoC Device Security
In an AMD Zynq™ UltraScale+™ MPSoC device, the secure boot is accomplished by using the hardware root of trust boot mechanism, which also provides a way to encrypt all of the boot or configuration files. This architecture provides the required confidentiality, integrity, and authentication to host the most secure of applications.
See the section "Security" in the Zynq UltraScale+ Device Technical Reference Manual (UG1085) for more information.
Versal Adaptive SoC Security
On AMD Versal™ adaptive SoCs, secure boot ensures the confidentiality, integrity, and authentication of the firmware and software loaded onto the device. The root of trust starts with the PMC ROM, that authenticates and/or decrypts the PLM software. Now that the PLM software is trusted, the PLM handles loading the rest of the firmware and software in a secure manner. Additionally, if secure boot is not desired then software can at least be validated with a simple checksum.
See Versal Adaptive SoC Technical Reference Manual (AM011) for more information. Also, see the Versal Adaptive SoC Security Manual (UG1508). This manual requires an active NDA to be downloaded from the Design Security Lounge.