Dmitrii Kuvaiskii

@anjalirai-intel Would it be possible to run with `loader.log_level = "all"` and show the log? I have a feeling that Python workload on RHEL creates several SGX enclaves, which is...

Ok, now I see that this log is just a concatenation of two `gramine-sgx` invocations. So each Gramine-SGX run creates a single enclave, as expected. > Just wanted to say,...

A simple way to test my hypothesis above is to add a sleep of e.g. 3 minutes in-between Gramine invocations. In particular, add `sleep 180` in these places: - https://github.com/gramineproject/gramine/blob/5fb315426a3bcc306969d5ace203513265844c38/CI-Examples/python/run-tests.sh#L17...

That's bad. Do you see anything interesting in `dmesg`? Any errors that Linux prints?

> `[1125918.789338] Out of memory: Killed process 2810877 (loader) total-vm:133895024kB` This line is not good. Your machine seems to have ~128GB of RAM, and somehow this RAM gets filled by...

@anjalirai-intel What is the situation with this issue? Is this the one we debugged and had a theory that it's the SGX driver in the Linux kernel doing something wrong...

Ok, let's keep this issue open, though I don't know how to debug it further.

> Currently in Gramine we return those weird strings via `readlink()` In particular, this code does it: https://github.com/gramineproject/gramine/blob/b7ffa83a4203bdc2b560cbfef140c0e8ab6b7a58/libos/src/fs/proc/thread.c#L273-L293 However, there is no inode information, just a hard-coded `[?]`. That's not...

> At the same time, this will probably a big change, and I'd say it's not worth it unless we have a real app (that cannot circumvent this limitation). I...

This is a bug in Busybox, it is incompatible with the newer Linux kernel versions (and as a result with everything that works closely with newer Linux versions, like Glibc):...