Query regarding APIC timer interval configuration about sgx-step

[ipaljak]

Hi Jo,

you were right, turns out the step interval had to be quite a bit higher on my machine (~40). Thank you very much for the help.

Cheers,
Ivan

[jovanbulck]

Hi Ivan,

From the output, this looks like you're "zero-stepping": your APC timer interval is likely too low so that the interrupt is arriving too early (during ERESUME itself) and the enclave gets interrupted before it can make any progress. That explains why you see the same RIP value all the time: the enclave is stuck at the first page fault and cannot make progress. In the app/bench program, single-stepping starts from the first page fault, which is likely not yet the NOP instruction slide but another function residing on the same page (you could verify this with objdump and the RIP you see being reported here). This is why the parse_nop script reports zero on the whole line.

Moving forward: your CPU seems to have a base frequency of 2.60GHz and the above output was generated with SGX_STEP_TIMER_INTERVAL=20 which is likely too low looking at the table in the README. Try going for something around 23-25? If you keep getting the same RIP, try going even higher depending on the microcode version you're using. Hope this was helpful, let me know if it works for you ^^