I've done performance testing, and I can generate 50 kilobytes of random data in 6.9 seconds using the TMP2 hardware. This is adequate. The boot-up state of my embedded device is very predictable, and so I don't want any mathematical algorithms for psudeo-randomness present on my device. I am cutting them all out. I don't want to use the TPM2 to feed entropy into the Linux kernel's PRNG. Instead I wish to remove the PRNG from the kernel and replace it with a direct call to the TPM2 chip. Performance is not an issue here.
