https://github.com/hermit-os/kernel/pull/1248 has mare it mode likely that we end up reading uninitialized memory for FUSE. I suspect this is coming from subtle codegen changes (aka we are doing something wrong with uninit memory or references).
$ cargo xtask ci qemu --arch x86_64 --profile release --package rftrace-example --virtiofsd
...
Saving traces to disk...!
0xc1b000, Events { ptr: 0xc1b000, len: 2000, cap: 2000 }
Parsing TID 1...!
Writing to disk: 8 events, 128 bytes (/root/tracedir/1.dat)
Parsed all events!
Creating fake uftrace data dir at /root/tracedir..
Creating ./info
feats = TASK_SESSION | SYM_REL_ADDR
info = CMDLINE | TASKINFO
cmdline = 'fakeuftrace'
tid = [1]
Creating ./task.txt
[0][PANIC] panicked at src/fs/fuse.rs:479:47:
called `Result::unwrap()` on an `Err` value: TryFromIntError(())
This happens while converting a u64 into an i64, so this is most definitely not a real value.
https://github.com/hermit-os/kernel/pull/1248 has mare it mode likely that we end up reading uninitialized memory for FUSE. I suspect this is coming from subtle codegen changes (aka we are doing something wrong with uninit memory or references).
This happens while converting a
u64
into ani64
, so this is most definitely not a real value.https://github.com/hermit-os/kernel/blob/a723af0ae2de31d04474e290b2e33d253a3cabfe/src/fs/fuse.rs#L479
I'll disable the test in CI for now.