/GPTtrace

Generate eBPF programs and tracing with ChatGPT and natural language

Primary LanguagePythonMIT LicenseMIT

GPTtrace 🤖

License: MIT Actions Status DeepSource CodeFactor

Generate eBPF programs and tracing with ChatGPT and natural language

Key Features 💡

Interact and Tracing your Linux with natural language, it can tell how to write eBPF programs in BCC, libbpf styles.

example: tracing with Count page faults by process

result

Generate eBPF programs with natural language

generate

For detail documents and tutorials about how we train ChatGPT to write eBPF programs, please refer to: bpf-developer-tutorial （a libbpf tool tutorial to teach ChatGPT to write eBPF programs)

Note that the GPTtrace tool now is only a demo project to show how it works, the result may not be accuracy, and it is not recommended to use it in production. We are working to make it more stable and complete!

Usage and Setup 🛠

$ ./GPTtrace.py
usage: GPTtrace [-h] [-i | -v | -e TEXT | -g TEXT] [-u UUID] [-t ACCESS_TOKEN]

Use ChatGPT to write eBPF programs (bpftrace, etc.)

optional arguments:
  -h, --help            show this help message and exit
  -i, --info            Let ChatGPT explain what's eBPF
  -v, --verbose         Print the prompt and receive message
  -e TEXT, --execute TEXT
                        Generate commands using your input with ChatGPT, and run it
  -g TEXT, --generate TEXT
                        Generate eBPF programs using your input with ChatGPT
  -u UUID, --uuid UUID  Conversion UUID to use, or passed through environment variable `GPTTRACE_CONV_UUID`
  -t ACCESS_TOKEN, --access-token ACCESS_TOKEN
                        ChatGPT access token, see `https://chat.openai.com/api/auth/session` or passed through
                        `GPTTRACE_ACCESS_TOKEN`

First: login to ChatGPT

  • get the Conversion ID from ChatGPT, and then set it to the environment variable GPTTRACE_CONV_UUID or use the -u option. The Conversion ID is the last part of the URL of the conversation, for example, the Conversion ID of https://chat.openai.com/conv/1a2b3c4d-0000-0000-0000-1k2l3m4n5o6p is 1a2b3c4d-0000-0000-0000-1k2l3m4n5o6p(example, not usable).
  • get the access token from ChatGPT, and then set it to the environment variable GPTTRACE_ACCESS_TOKEN or use the -t option. see https://chat.openai.com/api/auth/session for the access token.

Use prompts to teach ChatGPT to write eBPF programs

$ ./GPTtrace.py --train
----------------------------
Training ChatGPT with `1.md`
----------------------------
....
Trained session: cbd73f64-64b8-4f1d-80d3-c5f4f2fe292e

This will use the material in the prompts directory to teach ChatGPT to write eBPF programs in bpftrace, libbpf, and BCC styles. You can also do that manually by sending the prompts to ChatGPT in the Website.

start your tracing! 🚀

For example:

./GPTtrace.py -e "Count page faults by process"

If the eBPF program cannot be loaded into the kernel, The error message will be used to correct ChatGPT, and the result will be printed to the console.

How it works

  1. GPTtrace pre-trains its eBPF programs using various eBPF development resources, has multiple conversations with ChatGPT to teach it how to write different types of eBPF programs and bpftrace DSLs.
  2. The user inputs their request in natural language, and GPTtrace calls the ChatGPT API to generate an eBPF program. The generated program is then executed via shell or written to a file for compilation and execution.
  3. If there are errors in compilation or loading, the error is sent back to ChatGPT to generate a new eBPF program or command.

Room for improvement

There is still plenty of room for improvement, including:

  1. Once the ChatGPT can search online, it should be much better to let the tool get sample programs from the bcc/bpftrace repository and learn them, or let the tool look at Stack Overflow or something to see how to write eBPF programs, similar to the method used in new Bing search.
  2. Providing more high-quality documentation and tutorials to improve the accuracy of the output and the quality of the code examples.
  3. Making multiple calls to other tools to execute commands and return results. For example, GPTtrace could output a command, have bpftrace query the current kernel version and supported tracepoints, and return the output as part of the conversation.
  4. Incorporating user feedback to improve the quality of the generated code and refine the natural language processing capabilities of the tool.

And also, new LLM models will certainly lead to more realistic and accurate language generation.

Installation 🔧

./install.sh

Examples

  • Files opened by process
  • Syscall count by program
  • Read bytes by process:
  • Read size distribution by process:
  • Show per-second syscall rates:
  • Trace disk size by process
  • Count page faults by process
  • Count LLC cache misses by process name and PID (uses PMCs):
  • Profile user-level stacks at 99 Hertz, for PID 189:
  • Files opened, for processes in the root cgroup-v2

LICENSE

MIT

🔗 Links