firerunner vsock connection fails but controller works

[LedgeDash]

Calling a function through firerunner directly fails with zero output:

$ cat single_req.json | ./target/debug/firerunner --kernel vmlinux --rootfs images/nodejs.ext4 --appfs images/loremjs.ext4

^C
$

No messaging showing vsock connection succeeds. And after a while the VM exits so ctrl-C just kills the firerunner process.

However, controller works fine:

[luzhuo@Jasper] ./target/debug/controller --appfs_dir images --kernel vmlinux --runtimefs_dir images --requests ctest.jsonl
lorempy2: {"body": "Dolorem voluptatem aliquam quaerat ut dolor.", "request": {"function": "lorempy2", "payload": {"request": 42}}}
Warm 0, Active: 2
loremjs: {"request":{"function":"loremjs","payload":{"request":42}},"sentence":"Aliqua magna anim id ullamco officia exercitation officia ea nostrud esse."}
Warm 0, Active: 2
lorempy2: {"body": "Ut dolor numquam modi.", "request": {"function": "lorempy2", "payload": {"request": 42}}}
Warm 0, Active: 2
loremjs: {"request":{"function":"loremjs","payload":{"request":42}},"sentence":"Ad do adipisicing incididunt ullamco labore nulla Lorem occaecat ut."}
Warm 0, Active: 2
lorempy2: {"body": "Aliquam porro sed neque sed sit.", "request": {"function": "lorempy2", "payload": {"request": 42}}}
Warm 0, Active: 2
loremjs: {"request":{"function":"loremjs","payload":{"request":42}},"sentence":"Ut elit et nisi ipsum ex sint qui nulla officia reprehenderit consectetur enim."}
Warm 0, Active: 2
lorempy2: {"body": "Quiquia eius tempora modi tempora.", "request": {"function": "lorempy2", "payload": {"request": 42}}}
Warm 0, Active: 2
loremjs: {"request":{"function":"loremjs","payload":{"request":42}},"sentence":"Ut est duis adipisicing incididunt labore elit id irure commodo magna id nulla ea duis."}
Warm 0, Active: 2
lorempy2: {"body": "Velit labore etincidunt velit etincidunt consectetur eius labore.", "request": {"function": "lorempy2", "payload": {"request": 42}}}
Warm 0, Active: 2
loremjs: {"request":{"function":"loremjs","payload":{"request":42}},"sentence":"Occaecat est fugiat quis sit aliquip exercitation duis dolore excepteur mollit."}
Warm 0, Active: 2
lorempy2: {"body": "Quisquam dolorem dolor aliquam modi non dolorem.", "request": {"function": "lorempy2", "payload": {"request": 42}}}
Warm 0, Active: 2
loremjs: {"request":{"function":"loremjs","payload":{"request":42}},"sentence":"Ex cupidatat deserunt ad amet occaecat cupidatat minim nostrud deserunt amet deserunt cillum ullamco."}
Warm 0, Active: 2
lorempy2: {"body": "Numquam tempora porro magnam dolore sed quaerat.", "request": {"function": "lorempy2", "payload": {"request": 42}}}
Warm 1, Active: 1
loremjs: {"request":{"function":"loremjs","payload":{"request":42}},"sentence":"Ut consectetur nisi minim magna dolore nulla culpa minim enim incididunt cillum ullamco sit."}
Warm 2, Active: 0

This is related to #12 . Need to improve code around vsock handling.

[LedgeDash]

It's not always the case that VM would exit. Sometimes the VM seems to be stuck and can only be terminated with kill since the VM treats SIGINT (ctrl-c) as nop.

[LedgeDash]

It is also possible for the following to happen:

[davidliu@sns59] ./target/debug/firerunner --kernel ../images/vmlinux --rootfs images/nodejs.ext4 --appfs loremjs.ext4
Connection from VsockAddr { port: 1024, cid: 42 }

vsock connection seems to be successful but nothing seems to go through. and the vm just hangs.

Not sure if this is related at all to @tan-yue 's problem with vsock when booting from a snapshot...?? Maybe our currently implementation around vsock is just broken...

[LedgeDash]

Problem is "resolved" after system reboot. So likely there's something wrong with vsock states in the kernel module. function VM can run and output results but now I see:

[luzhuo@Jasper] ./target/debug/firerunner --kernel vmlinux --rootfs images/nodejs.ext4 --appfs images/loremjs.ext4 < single_req.json
Connection from VsockAddr { port: 1024, cid: 42 }
{"request":{"function":"lorempy2","payload":{}},"sentence":"Sunt non sit ullamco adipisicing officia ipsum in ad commodo."}
thread 'main' panicked at 'Failed to kill child: Sys(ESRCH)', src/libcore/result.rs:999:5
note: Run with `RUST_BACKTRACE=1` environment variable to display a backtrace.

This is panic is happening with Amit's images. (previously, I had this panic issue with images that i created (#11),but it went away when testing with Amit's images. )

[alevy]

We are almost definitely getting rid of vsock, so probably not worth debugging