Pass down structs_sym_tab to maps_debug_info, allow vmlinux enums to be used in an indexed format for map declaration

Add support for all eBPF helpers

.gitattributes

@@ -0,0 +1,3 @@
+tests/c-form/vmlinux.h linguist-vendored
+examples/ linguist-vendored
+BCC-Examples/ linguist-vendored

.github/dependabot.yml

@@ -0,0 +1,11 @@
+version: 2
+updates:
+  # Maintain dependencies for GitHub Actions
+  - package-ecosystem: "github-actions"
+    directory: "/"
+    schedule:
+      interval: "weekly"
+    groups:
+      actions:
+        patterns:
+          - "*"

.github/workflows/format.yml

@@ -0,0 +1,19 @@
+# This is a format job. Pre-commit has a first-party GitHub action, so we use
+# that: https://github.com/pre-commit/action
+
+name: Format
+
+on:
+  workflow_dispatch:
+  push:
+
+jobs:
+  pre-commit:
+    name: Format
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v5
+    - uses: actions/setup-python@v6
+      with:
+        python-version: "3.x"
+    - uses: pre-commit/action@v3.0.1

.github/workflows/python-publish.yml

@@ -20,9 +20,9 @@ jobs:
     runs-on: ubuntu-latest
 
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v5
 
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@v6
         with:
           python-version: "3.x"
 
@@ -33,7 +33,7 @@ jobs:
           python -m build
 
       - name: Upload distributions
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@v5
         with:
           name: release-dists
           path: dist/
@@ -59,7 +59,7 @@ jobs:
 
     steps:
       - name: Retrieve release distributions
-        uses: actions/download-artifact@v4
+        uses: actions/download-artifact@v6
         with:
           name: release-dists
           path: dist/

.gitignore

@@ -5,4 +5,8 @@
 .vscode/
 __pycache__/
 *.ll
-*.o
+*.o
+.ipynb_checkpoints/
+vmlinux.py
+~*
+vmlinux.h

.pre-commit-config.yaml

@@ -0,0 +1,59 @@
+# To use:
+#
+#     pre-commit run -a
+#
+# Or:
+#
+#     pre-commit install  # (runs every time you commit in git)
+#
+# To update this file:
+#
+#     pre-commit autoupdate
+#
+# See https://github.com/pre-commit/pre-commit
+
+exclude: 'vmlinux.py'
+
+ci:
+  autoupdate_commit_msg: "chore: update pre-commit hooks"
+  autofix_commit_msg: "style: pre-commit fixes"
+
+repos:
+# Standard hooks
+- repo: https://github.com/pre-commit/pre-commit-hooks
+  rev: v6.0.0
+  hooks:
+  - id: check-added-large-files
+  - id: check-case-conflict
+  - id: check-merge-conflict
+  - id: check-symlinks
+  - id: check-yaml
+    exclude: ^conda\.recipe/meta\.yaml$
+  - id: debug-statements
+  - id: end-of-file-fixer
+  - id: mixed-line-ending
+  - id: requirements-txt-fixer
+  - id: trailing-whitespace
+
+- repo: https://github.com/astral-sh/ruff-pre-commit
+  rev: "v0.13.2"
+  hooks:
+    - id: ruff
+      args: ["--fix", "--show-fixes"]
+    - id: ruff-format
+#      exclude: ^(docs)|^(tests)|^(examples)
+
+# Checking static types
+- repo: https://github.com/pre-commit/mirrors-mypy
+  rev: "v1.18.2"
+  hooks:
+    - id: mypy
+      exclude: ^(tests)|^(examples)
+      additional_dependencies: [types-setuptools]
+
+# Changes tabs to spaces
+- repo: https://github.com/Lucas-C/pre-commit-hooks
+  rev: v1.5.5
+  hooks:
+  - id: remove-tabs
+    exclude: '^(docs)|.*/Makefile$|Makefile$'

BCC-Examples/README.md

@@ -0,0 +1,20 @@
+## BCC examples ported to PythonBPF
+
+This folder contains examples of BCC tutorial examples that have been ported to use **PythonBPF**.
+
+## Requirements
+- install `pythonbpf` and `pylibbpf` using pip.
+- You will also need `matplotlib` for vfsreadlat.py example.
+- You will also need `rich` for vfsreadlat_rich.py example.
+- You will also need `plotly` and `dash` for vfsreadlat_plotly.py example.
+
+## Usage
+- You'll need root privileges to run these examples. If you are using a virtualenv, use the following command to run the scripts:
+  ```bash
+  sudo <path_to_virtualenv>/bin/python3 <script_name>.py
+  ```
+- For vfsreadlat_plotly.py, run the following command to start the Dash server:
+  ```bash
+  sudo <path_to_virtualenv>/bin/python3 vfsreadlat_plotly/bpf_program.py
+  ```
+  Then open your web browser and navigate to the given URL.

BCC-Examples/hello_fields.ipynb

@@ -0,0 +1,83 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "28cf2e27-41e2-461c-a39c-147417141a4e",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pythonbpf import bpf, section, bpfglobal, BPF, trace_fields\n",
+    "from ctypes import c_void_p, c_int64"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "133190e5-5a99-4585-b6e1-91224ed973c2",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "@bpf\n",
+    "@section(\"tracepoint/syscalls/sys_enter_clone\")\n",
+    "def hello_world(ctx: c_void_p) -> c_int64:\n",
+    "    print(\"Hello, World!\")\n",
+    "    return 0\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@bpfglobal\n",
+    "def LICENSE() -> str:\n",
+    "    return \"GPL\"\n",
+    "\n",
+    "\n",
+    "# Compile and load\n",
+    "b = BPF()\n",
+    "b.load()\n",
+    "b.attach_all()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "d3934efb-4043-4545-ae4c-c50ec40a24fd",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# header\n",
+    "print(f\"{'TIME(s)':<18} {'COMM':<16} {'PID':<6} {'MESSAGE'}\")\n",
+    "\n",
+    "# format output\n",
+    "while True:\n",
+    "    try:\n",
+    "        (task, pid, cpu, flags, ts, msg) = trace_fields()\n",
+    "    except ValueError:\n",
+    "        continue\n",
+    "    except KeyboardInterrupt:\n",
+    "        exit()\n",
+    "    print(f\"{ts:<18} {task:<16} {pid:<6} {msg}\")"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.13.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

BCC-Examples/hello_fields.py

@@ -0,0 +1,34 @@
+from pythonbpf import bpf, section, bpfglobal, BPF, trace_fields
+from ctypes import c_void_p, c_int64
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_clone")
+def hello_world(ctx: c_void_p) -> c_int64:
+    print("Hello, World!")
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+# Compile and load
+b = BPF()
+b.load()
+b.attach_all()
+
+# header
+print(f"{'TIME(s)':<18} {'COMM':<16} {'PID':<6} {'MESSAGE'}")
+
+# format output
+while True:
+    try:
+        (task, pid, cpu, flags, ts, msg) = trace_fields()
+    except ValueError:
+        continue
+    except KeyboardInterrupt:
+        exit()
+    print(f"{ts:<18} {task:<16} {pid:<6} {msg}")

BCC-Examples/hello_perf_output.ipynb

@@ -0,0 +1,110 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "79b74928-f4b4-4320-96e3-d973997de2f4",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pythonbpf import bpf, map, struct, section, bpfglobal, BPF\n",
+    "from pythonbpf.helper import ktime, pid, comm\n",
+    "from pythonbpf.maps import PerfEventArray\n",
+    "from ctypes import c_void_p, c_int64"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "5bdb0329-ae2d-45e8-808e-5ed5b1374204",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "@bpf\n",
+    "@struct\n",
+    "class data_t:\n",
+    "    pid: c_int64\n",
+    "    ts: c_int64\n",
+    "    comm: str(16)\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@map\n",
+    "def events() -> PerfEventArray:\n",
+    "    return PerfEventArray(key_size=c_int64, value_size=c_int64)\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@section(\"tracepoint/syscalls/sys_enter_clone\")\n",
+    "def hello(ctx: c_void_p) -> c_int64:\n",
+    "    dataobj = data_t()\n",
+    "    dataobj.pid, dataobj.ts = pid(), ktime()\n",
+    "    comm(dataobj.comm)\n",
+    "    events.output(dataobj)\n",
+    "    return 0\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@bpfglobal\n",
+    "def LICENSE() -> str:\n",
+    "    return \"GPL\"\n",
+    "\n",
+    "\n",
+    "# Compile and load\n",
+    "b = BPF()\n",
+    "b.load()\n",
+    "b.attach_all()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "4bcc7d57-6cc4-48a3-bbd2-42ad6263afdf",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "start = 0\n",
+    "\n",
+    "\n",
+    "def callback(cpu, event):\n",
+    "    global start\n",
+    "    if start == 0:\n",
+    "        start = event.ts\n",
+    "    ts = (event.ts - start) / 1e9\n",
+    "    print(f\"[CPU {cpu}] PID: {event.pid}, TS: {ts}, COMM: {event.comm.decode()}\")\n",
+    "\n",
+    "\n",
+    "perf = b[\"events\"].open_perf_buffer(callback, struct_name=\"data_t\")\n",
+    "print(\"Starting to poll... (Ctrl+C to stop)\")\n",
+    "print(\"Try running: fork() or clone() system calls to trigger events\")\n",
+    "\n",
+    "try:\n",
+    "    while True:\n",
+    "        b[\"events\"].poll(1000)\n",
+    "except KeyboardInterrupt:\n",
+    "    print(\"Stopping...\")"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.13.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

BCC-Examples/hello_perf_output.py

@@ -0,0 +1,61 @@
+from pythonbpf import bpf, map, struct, section, bpfglobal, BPF
+from pythonbpf.helper import ktime, pid, comm
+from pythonbpf.maps import PerfEventArray
+from ctypes import c_void_p, c_int64
+
+
+@bpf
+@struct
+class data_t:
+    pid: c_int64
+    ts: c_int64
+    comm: str(16)  # type: ignore [valid-type]
+
+
+@bpf
+@map
+def events() -> PerfEventArray:
+    return PerfEventArray(key_size=c_int64, value_size=c_int64)
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_clone")
+def hello(ctx: c_void_p) -> c_int64:
+    dataobj = data_t()
+    dataobj.pid, dataobj.ts = pid(), ktime()
+    comm(dataobj.comm)
+    events.output(dataobj)
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+# Compile and load
+b = BPF()
+b.load()
+b.attach_all()
+
+start = 0
+
+
+def callback(cpu, event):
+    global start
+    if start == 0:
+        start = event.ts
+    ts = (event.ts - start) / 1e9
+    print(f"[CPU {cpu}] PID: {event.pid}, TS: {ts}, COMM: {event.comm.decode()}")
+
+
+perf = b["events"].open_perf_buffer(callback, struct_name="data_t")
+print("Starting to poll... (Ctrl+C to stop)")
+print("Try running: fork() or clone() system calls to trigger events")
+
+try:
+    while True:
+        b["events"].poll(1000)
+except KeyboardInterrupt:
+    print("Stopping...")

BCC-Examples/hello_world.ipynb

@@ -0,0 +1,116 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "7d5d3cfb-39ba-4516-9856-b3bed47a0cef",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pythonbpf import bpf, section, bpfglobal, BPF, trace_pipe\n",
+    "from ctypes import c_void_p, c_int64"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "cf1c87aa-e173-4156-8f2d-762225bc6d19",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "@bpf\n",
+    "@section(\"tracepoint/syscalls/sys_enter_clone\")\n",
+    "def hello_world(ctx: c_void_p) -> c_int64:\n",
+    "    print(\"Hello, World!\")\n",
+    "    return 0\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@bpfglobal\n",
+    "def LICENSE() -> str:\n",
+    "    return \"GPL\"\n",
+    "\n",
+    "\n",
+    "b = BPF()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "bd81383d-f75a-4269-8451-3d985d85b124",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "      Cache2 I/O-4716    [003] ...21  8218.000492: bpf_trace_printk: count: 11 with 4716\n",
+      "\n",
+      "      Cache2 I/O-4716    [003] ...21  8218.000499: bpf_trace_printk: Hello, World!\n",
+      "\n",
+      "   WebExtensions-5168    [002] ...21  8219.320392: bpf_trace_printk: count: 13 with 5168\n",
+      "\n",
+      "   WebExtensions-5168    [002] ...21  8219.320399: bpf_trace_printk: Hello, World!\n",
+      "\n",
+      "          python-21155   [001] ...21  8220.933716: bpf_trace_printk: count: 5 with 21155\n",
+      "\n",
+      "          python-21155   [001] ...21  8220.933721: bpf_trace_printk: Hello, World!\n",
+      "\n",
+      "          python-21155   [002] ...21  8221.341290: bpf_trace_printk: count: 6 with 21155\n",
+      "\n",
+      "          python-21155   [002] ...21  8221.341295: bpf_trace_printk: Hello, World!\n",
+      "\n",
+      " Isolated Web Co-5462    [000] ...21  8223.095033: bpf_trace_printk: count: 7 with 5462\n",
+      "\n",
+      " Isolated Web Co-5462    [000] ...21  8223.095043: bpf_trace_printk: Hello, World!\n",
+      "\n",
+      "         firefox-4542    [000] ...21  8227.760067: bpf_trace_printk: count: 8 with 4542\n",
+      "\n",
+      "         firefox-4542    [000] ...21  8227.760080: bpf_trace_printk: Hello, World!\n",
+      "\n",
+      " Isolated Web Co-12404   [003] ...21  8227.917086: bpf_trace_printk: count: 7 with 12404\n",
+      "\n",
+      " Isolated Web Co-12404   [003] ...21  8227.917095: bpf_trace_printk: Hello, World!\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "b.load()\n",
+    "b.attach_all()\n",
+    "trace_pipe()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "01e1f25b-decc-425b-a1aa-a5e701082574",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.13.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

BCC-Examples/hello_world.py

@@ -0,0 +1,23 @@
+from pythonbpf import bpf, section, bpfglobal, BPF, trace_pipe
+from ctypes import c_void_p, c_int64
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_clone")
+def hello_world(ctx: c_void_p) -> c_int64:
+    print("Hello, World!")
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+# Compile and load
+b = BPF()
+b.load()
+b.attach_all()
+
+trace_pipe()

BCC-Examples/sync_count.ipynb

@@ -0,0 +1,107 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "dcab010c-f5e9-446f-9f9f-056cc794ad14",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pythonbpf import bpf, map, section, bpfglobal, BPF, trace_fields\n",
+    "from pythonbpf.helper import ktime\n",
+    "from pythonbpf.maps import HashMap\n",
+    "\n",
+    "from ctypes import c_void_p, c_int64"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "720797e8-9c81-4af6-a385-80f1ec4c0f15",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "@bpf\n",
+    "@map\n",
+    "def last() -> HashMap:\n",
+    "    return HashMap(key=c_int64, value=c_int64, max_entries=2)\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@section(\"tracepoint/syscalls/sys_enter_sync\")\n",
+    "def do_trace(ctx: c_void_p) -> c_int64:\n",
+    "    ts_key, cnt_key = 0, 1\n",
+    "    tsp, cntp = last.lookup(ts_key), last.lookup(cnt_key)\n",
+    "    if not cntp:\n",
+    "        last.update(cnt_key, 0)\n",
+    "        cntp = last.lookup(cnt_key)\n",
+    "    if tsp:\n",
+    "        delta = ktime() - tsp\n",
+    "        if delta < 1000000000:\n",
+    "            time_ms = delta // 1000000\n",
+    "            print(f\"{time_ms} {cntp}\")\n",
+    "        last.delete(ts_key)\n",
+    "    else:\n",
+    "        last.update(ts_key, ktime())\n",
+    "    last.update(cnt_key, cntp + 1)\n",
+    "    return 0\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@bpfglobal\n",
+    "def LICENSE() -> str:\n",
+    "    return \"GPL\"\n",
+    "\n",
+    "\n",
+    "# Compile and load\n",
+    "b = BPF()\n",
+    "b.load()\n",
+    "b.attach_all()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "78a8b82c-7c5f-43c1-9de1-cd982a0f345b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "print(\"Tracing for quick sync's... Ctrl-C to end\")\n",
+    "\n",
+    "# format output\n",
+    "start = 0\n",
+    "while True:\n",
+    "    try:\n",
+    "        task, pid, cpu, flags, ts, msg = trace_fields()\n",
+    "        if start == 0:\n",
+    "            start = ts\n",
+    "        ts -= start\n",
+    "        ms, cnt = msg.split()\n",
+    "        print(f\"At time {ts} s: Multiple syncs detected, last {ms} ms ago. Count {cnt}\")\n",
+    "    except KeyboardInterrupt:\n",
+    "        exit()"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.13.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

BCC-Examples/sync_count.py

@@ -0,0 +1,58 @@
+from pythonbpf import bpf, map, section, bpfglobal, BPF, trace_fields
+from pythonbpf.helper import ktime
+from pythonbpf.maps import HashMap
+
+from ctypes import c_void_p, c_int64
+
+
+@bpf
+@map
+def last() -> HashMap:
+    return HashMap(key=c_int64, value=c_int64, max_entries=2)
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_sync")
+def do_trace(ctx: c_void_p) -> c_int64:
+    ts_key, cnt_key = 0, 1
+    tsp, cntp = last.lookup(ts_key), last.lookup(cnt_key)
+    if not cntp:
+        last.update(cnt_key, 0)
+        cntp = last.lookup(cnt_key)
+    if tsp:
+        delta = ktime() - tsp
+        if delta < 1000000000:
+            time_ms = delta // 1000000
+            print(f"{time_ms} {cntp}")
+        last.delete(ts_key)
+    else:
+        last.update(ts_key, ktime())
+    last.update(cnt_key, cntp + 1)
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+# Compile and load
+b = BPF()
+b.load()
+b.attach_all()
+
+print("Tracing for quick sync's... Ctrl-C to end")
+
+# format output
+start = 0
+while True:
+    try:
+        task, pid, cpu, flags, ts, msg = trace_fields()
+        if start == 0:
+            start = ts
+        ts -= start
+        ms, cnt = msg.split()
+        print(f"At time {ts} s: Multiple syncs detected, last {ms} ms ago. Count {cnt}")
+    except KeyboardInterrupt:
+        exit()

BCC-Examples/sync_perf_output.ipynb

@@ -0,0 +1,134 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "b0d1ab05-0c1f-4578-9c1b-568202b95a5c",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pythonbpf import bpf, map, struct, section, bpfglobal, BPF\n",
+    "from pythonbpf.helper import ktime\n",
+    "from pythonbpf.maps import HashMap, PerfEventArray\n",
+    "from ctypes import c_void_p, c_int64"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "85e50d0a-f9d8-468f-8e03-f5f7128f05d8",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "@bpf\n",
+    "@struct\n",
+    "class data_t:\n",
+    "    ts: c_int64\n",
+    "    ms: c_int64\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@map\n",
+    "def events() -> PerfEventArray:\n",
+    "    return PerfEventArray(key_size=c_int64, value_size=c_int64)\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@map\n",
+    "def last() -> HashMap:\n",
+    "    return HashMap(key=c_int64, value=c_int64, max_entries=1)\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@section(\"tracepoint/syscalls/sys_enter_sync\")\n",
+    "def do_trace(ctx: c_void_p) -> c_int64:\n",
+    "    dat, dat.ts, key = data_t(), ktime(), 0\n",
+    "    tsp = last.lookup(key)\n",
+    "    if tsp:\n",
+    "        delta = ktime() - tsp\n",
+    "        if delta < 1000000000:\n",
+    "            dat.ms = delta // 1000000\n",
+    "            events.output(dat)\n",
+    "        last.delete(key)\n",
+    "    else:\n",
+    "        last.update(key, ktime())\n",
+    "    return 0\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@bpfglobal\n",
+    "def LICENSE() -> str:\n",
+    "    return \"GPL\"\n",
+    "\n",
+    "\n",
+    "# Compile and load\n",
+    "b = BPF()\n",
+    "b.load()\n",
+    "b.attach_all()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "40bb1107-369f-4be7-9f10-37201900c16b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "print(\"Tracing for quick sync's... Ctrl-C to end\")\n",
+    "\n",
+    "# format output\n",
+    "start = 0\n",
+    "\n",
+    "\n",
+    "def callback(cpu, event):\n",
+    "    global start\n",
+    "    if start == 0:\n",
+    "        start = event.ts\n",
+    "    event.ts -= start\n",
+    "    print(\n",
+    "        f\"At time {event.ts / 1e9} s: Multiple sync detected, Last sync: {event.ms} ms ago\"\n",
+    "    )\n",
+    "\n",
+    "\n",
+    "perf = b[\"events\"].open_perf_buffer(callback, struct_name=\"data_t\")\n",
+    "print(\"Starting to poll... (Ctrl+C to stop)\")\n",
+    "print(\"Try running: fork() or clone() system calls to trigger events\")\n",
+    "\n",
+    "try:\n",
+    "    while True:\n",
+    "        b[\"events\"].poll(1000)\n",
+    "except KeyboardInterrupt:\n",
+    "    print(\"Stopping...\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "94a588d9-3a40-437c-a35b-fc40410f3eb7",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.13.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

BCC-Examples/sync_perf_output.py

@@ -0,0 +1,75 @@
+from pythonbpf import bpf, map, struct, section, bpfglobal, BPF
+from pythonbpf.helper import ktime
+from pythonbpf.maps import HashMap, PerfEventArray
+from ctypes import c_void_p, c_int64
+
+
+@bpf
+@struct
+class data_t:
+    ts: c_int64
+    ms: c_int64
+
+
+@bpf
+@map
+def events() -> PerfEventArray:
+    return PerfEventArray(key_size=c_int64, value_size=c_int64)
+
+
+@bpf
+@map
+def last() -> HashMap:
+    return HashMap(key=c_int64, value=c_int64, max_entries=1)
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_sync")
+def do_trace(ctx: c_void_p) -> c_int64:
+    dat, dat.ts, key = data_t(), ktime(), 0
+    tsp = last.lookup(key)
+    if tsp:
+        delta = ktime() - tsp
+        if delta < 1000000000:
+            dat.ms = delta // 1000000
+            events.output(dat)
+        last.delete(key)
+    else:
+        last.update(key, ktime())
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+# Compile and load
+b = BPF()
+b.load()
+b.attach_all()
+
+print("Tracing for quick sync's... Ctrl-C to end")
+
+# format output
+start = 0
+
+
+def callback(cpu, event):
+    global start
+    if start == 0:
+        start = event.ts
+    event.ts -= start
+    print(
+        f"At time {event.ts / 1e9} s: Multiple sync detected, Last sync: {event.ms} ms ago"
+    )
+
+
+perf = b["events"].open_perf_buffer(callback, struct_name="data_t")
+print("Starting to poll... (Ctrl+C to stop)")
+try:
+    while True:
+        b["events"].poll(1000)
+except KeyboardInterrupt:
+    print("Stopping...")

BCC-Examples/sync_timing.ipynb

@@ -0,0 +1,102 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "bfe01ceb-2f27-41b3-b3ba-50ec65cfddda",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pythonbpf import bpf, map, section, bpfglobal, BPF, trace_fields\n",
+    "from pythonbpf.helper import ktime\n",
+    "from pythonbpf.maps import HashMap\n",
+    "\n",
+    "from ctypes import c_void_p, c_int64"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "ddb115f4-20a7-43bc-bb5b-ccbfd6031fc2",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "@bpf\n",
+    "@map\n",
+    "def last() -> HashMap:\n",
+    "    return HashMap(key=c_int64, value=c_int64, max_entries=1)\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@section(\"tracepoint/syscalls/sys_enter_sync\")\n",
+    "def do_trace(ctx: c_void_p) -> c_int64:\n",
+    "    key = 0\n",
+    "    tsp = last.lookup(key)\n",
+    "    if tsp:\n",
+    "        delta = ktime() - tsp\n",
+    "        if delta < 1000000000:\n",
+    "            time_ms = delta // 1000000\n",
+    "            print(f\"{time_ms}\")\n",
+    "        last.delete(key)\n",
+    "    else:\n",
+    "        last.update(key, ktime())\n",
+    "    return 0\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@bpfglobal\n",
+    "def LICENSE() -> str:\n",
+    "    return \"GPL\"\n",
+    "\n",
+    "\n",
+    "# Compile and load\n",
+    "b = BPF()\n",
+    "b.load()\n",
+    "b.attach_all()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "e4f46574-9fd8-46e7-9c7b-27a36d07f200",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "print(\"Tracing for quick sync's... Ctrl-C to end\")\n",
+    "\n",
+    "# format output\n",
+    "start = 0\n",
+    "while True:\n",
+    "    try:\n",
+    "        task, pid, cpu, flags, ts, ms = trace_fields()\n",
+    "        if start == 0:\n",
+    "            start = ts\n",
+    "        ts -= start\n",
+    "        print(f\"At time {ts} s: Multiple syncs detected, last {ms} ms ago\")\n",
+    "    except KeyboardInterrupt:\n",
+    "        exit()"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.13.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

BCC-Examples/sync_timing.py

@@ -0,0 +1,53 @@
+from pythonbpf import bpf, map, section, bpfglobal, BPF, trace_fields
+from pythonbpf.helper import ktime
+from pythonbpf.maps import HashMap
+
+from ctypes import c_void_p, c_int64
+
+
+@bpf
+@map
+def last() -> HashMap:
+    return HashMap(key=c_int64, value=c_int64, max_entries=1)
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_sync")
+def do_trace(ctx: c_void_p) -> c_int64:
+    key = 0
+    tsp = last.lookup(key)
+    if tsp:
+        delta = ktime() - tsp
+        if delta < 1000000000:
+            time_ms = delta // 1000000
+            print(f"{time_ms}")
+        last.delete(key)
+    else:
+        last.update(key, ktime())
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+# Compile and load
+b = BPF()
+b.load()
+b.attach_all()
+
+print("Tracing for quick sync's... Ctrl-C to end")
+
+# format output
+start = 0
+while True:
+    try:
+        task, pid, cpu, flags, ts, ms = trace_fields()
+        if start == 0:
+            start = ts
+        ts -= start
+        print(f"At time {ts} s: Multiple syncs detected, last {ms} ms ago")
+    except KeyboardInterrupt:
+        exit()

BCC-Examples/sys_sync.ipynb

@@ -0,0 +1,73 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "bb49598f-b9cc-4ea8-8391-923cad513711",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pythonbpf import bpf, section, bpfglobal, BPF, trace_pipe\n",
+    "from ctypes import c_void_p, c_int64"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "5da237b0-1c7d-4ec5-8c24-696b1c1d97fa",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "@bpf\n",
+    "@section(\"tracepoint/syscalls/sys_enter_sync\")\n",
+    "def hello_world(ctx: c_void_p) -> c_int64:\n",
+    "    print(\"sys_sync() called\")\n",
+    "    return 0\n",
+    "\n",
+    "\n",
+    "@bpf\n",
+    "@bpfglobal\n",
+    "def LICENSE() -> str:\n",
+    "    return \"GPL\"\n",
+    "\n",
+    "\n",
+    "# Compile and load\n",
+    "b = BPF()\n",
+    "b.load()\n",
+    "b.attach_all()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "e4c218ac-fe47-4fd1-a27b-c07e02f3cd05",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "print(\"Tracing sys_sync()... Ctrl-C to end.\")\n",
+    "trace_pipe()"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.13.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

BCC-Examples/sys_sync.py

@@ -0,0 +1,23 @@
+from pythonbpf import bpf, section, bpfglobal, BPF, trace_pipe
+from ctypes import c_void_p, c_int64
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_sync")
+def hello_world(ctx: c_void_p) -> c_int64:
+    print("sys_sync() called")
+    return c_int64(0)
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+# Compile and load
+b = BPF()
+b.load()
+b.attach_all()
+print("Tracing sys_sync()... Ctrl-C to end.")
+trace_pipe()

BCC-Examples/vfsreadlat.py

@@ -0,0 +1,127 @@
+from pythonbpf import bpf, map, struct, section, bpfglobal, BPF
+from pythonbpf.helper import ktime, pid
+from pythonbpf.maps import HashMap, PerfEventArray
+from ctypes import c_void_p, c_uint64
+import matplotlib.pyplot as plt
+import numpy as np
+
+
+@bpf
+@struct
+class latency_event:
+    pid: c_uint64
+    delta_us: c_uint64  # Latency in microseconds
+
+
+@bpf
+@map
+def start() -> HashMap:
+    return HashMap(key=c_uint64, value=c_uint64, max_entries=10240)
+
+
+@bpf
+@map
+def events() -> PerfEventArray:
+    return PerfEventArray(key_size=c_uint64, value_size=c_uint64)
+
+
+@bpf
+@section("kprobe/vfs_read")
+def do_entry(ctx: c_void_p) -> c_uint64:
+    p, ts = pid(), ktime()
+    start.update(p, ts)
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@section("kretprobe/vfs_read")
+def do_return(ctx: c_void_p) -> c_uint64:
+    p = pid()
+    tsp = start.lookup(p)
+
+    if tsp:
+        delta_ns = ktime() - tsp
+
+        # Only track if latency > 1 microsecond
+        if delta_ns > 1000:
+            evt = latency_event()
+            evt.pid, evt.delta_us = p, delta_ns // 1000
+            events.output(evt)
+
+        start.delete(p)
+
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+# Load BPF
+print("Loading BPF program...")
+b = BPF()
+b.load()
+b.attach_all()
+
+# Collect latencies
+latencies = []
+
+
+def callback(cpu, event):
+    latencies.append(event.delta_us)
+
+
+b["events"].open_perf_buffer(callback, struct_name="latency_event")
+
+print("Tracing vfs_read latency... Hit Ctrl-C to end.")
+
+try:
+    while True:
+        b["events"].poll(1000)
+        if len(latencies) > 0 and len(latencies) % 1000 == 0:
+            print(f"Collected {len(latencies)} samples...")
+
+except KeyboardInterrupt:
+    print(f"Collected {len(latencies)} samples. Generating histogram...")
+
+# Create histogram with matplotlib
+if latencies:
+    # Use log scale for better visualization
+    log_latencies = np.log2(latencies)
+
+    plt.figure(figsize=(12, 6))
+
+    # Plot 1: Linear histogram
+    plt.subplot(1, 2, 1)
+    plt.hist(latencies, bins=50, edgecolor="black", alpha=0.7)
+    plt.xlabel("Latency (microseconds)")
+    plt.ylabel("Count")
+    plt.title("VFS Read Latency Distribution (Linear)")
+    plt.grid(True, alpha=0.3)
+
+    # Plot 2: Log2 histogram (like BCC)
+    plt.subplot(1, 2, 2)
+    plt.hist(log_latencies, bins=50, edgecolor="black", alpha=0.7, color="orange")
+    plt.xlabel("log2(Latency in µs)")
+    plt.ylabel("Count")
+    plt.title("VFS Read Latency Distribution (Log2)")
+    plt.grid(True, alpha=0.3)
+
+    # Add statistics
+    print("Statistics:")
+    print(f"  Count:  {len(latencies)}")
+    print(f"  Min:    {min(latencies)} µs")
+    print(f"  Max:    {max(latencies)} µs")
+    print(f"  Mean:   {np.mean(latencies):.2f} µs")
+    print(f"  Median: {np.median(latencies):.2f} µs")
+    print(f"  P95:    {np.percentile(latencies, 95):.2f} µs")
+    print(f"  P99:    {np.percentile(latencies, 99):.2f} µs")
+
+    plt.tight_layout()
+    plt.savefig("vfs_read_latency.png", dpi=150)
+    print("Histogram saved to vfs_read_latency.png")
+    plt.show()
+else:
+    print("No samples collected!")

BCC-Examples/vfsreadlat_plotly/bpf_program.py

@@ -0,0 +1,101 @@
+"""BPF program for tracing VFS read latency."""
+
+from pythonbpf import bpf, map, struct, section, bpfglobal, BPF
+from pythonbpf.helper import ktime, pid
+from pythonbpf.maps import HashMap, PerfEventArray
+from ctypes import c_void_p, c_uint64
+import argparse
+from data_collector import LatencyCollector
+from dashboard import LatencyDashboard
+
+
+@bpf
+@struct
+class latency_event:
+    pid: c_uint64
+    delta_us: c_uint64
+
+
+@bpf
+@map
+def start() -> HashMap:
+    """Map to store start timestamps by PID."""
+    return HashMap(key=c_uint64, value=c_uint64, max_entries=10240)
+
+
+@bpf
+@map
+def events() -> PerfEventArray:
+    """Perf event array for sending latency events to userspace."""
+    return PerfEventArray(key_size=c_uint64, value_size=c_uint64)
+
+
+@bpf
+@section("kprobe/vfs_read")
+def do_entry(ctx: c_void_p) -> c_uint64:
+    """Record start time when vfs_read is called."""
+    p, ts = pid(), ktime()
+    start.update(p, ts)
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@section("kretprobe/vfs_read")
+def do_return(ctx: c_void_p) -> c_uint64:
+    """Calculate and record latency when vfs_read returns."""
+    p = pid()
+    tsp = start.lookup(p)
+
+    if tsp:
+        delta_ns = ktime() - tsp
+
+        # Only track latencies > 1 microsecond
+        if delta_ns > 1000:
+            evt = latency_event()
+            evt.pid, evt.delta_us = p, delta_ns // 1000
+            events.output(evt)
+
+        start.delete(p)
+
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+def parse_args():
+    """Parse command line arguments."""
+    parser = argparse.ArgumentParser(
+        description="Monitor VFS read latency with live dashboard"
+    )
+    parser.add_argument(
+        "--host", default="0.0.0.0", help="Dashboard host (default: 0.0.0.0)"
+    )
+    parser.add_argument(
+        "--port", type=int, default=8050, help="Dashboard port (default: 8050)"
+    )
+    parser.add_argument(
+        "--buffer", type=int, default=10000, help="Recent data buffer size"
+    )
+    return parser.parse_args()
+
+
+args = parse_args()
+
+# Load BPF program
+print("Loading BPF program...")
+b = BPF()
+b.load()
+b.attach_all()
+print("✅ BPF program loaded and attached")
+
+# Setup data collector
+collector = LatencyCollector(b, buffer_size=args.buffer)
+collector.start()
+
+# Create and run dashboard
+dashboard = LatencyDashboard(collector)
+dashboard.run(host=args.host, port=args.port)

BCC-Examples/vfsreadlat_plotly/dashboard.py

@@ -0,0 +1,282 @@
+"""Plotly Dash dashboard for visualizing latency data."""
+
+import dash
+from dash import dcc, html
+from dash.dependencies import Input, Output
+import plotly.graph_objects as go
+from plotly.subplots import make_subplots
+import numpy as np
+
+
+class LatencyDashboard:
+    """Interactive dashboard for latency visualization."""
+
+    def __init__(self, collector, title: str = "VFS Read Latency Monitor"):
+        self.collector = collector
+        self.app = dash.Dash(__name__)
+        self.app.title = title
+        self._setup_layout()
+        self._setup_callbacks()
+
+    def _setup_layout(self):
+        """Create dashboard layout."""
+        self.app.layout = html.Div(
+            [
+                html.H1(
+                    "🔥 VFS Read Latency Dashboard",
+                    style={
+                        "textAlign": "center",
+                        "color": "#2c3e50",
+                        "marginBottom": 20,
+                    },
+                ),
+                # Stats cards
+                html.Div(
+                    [
+                        self._create_stat_card(
+                            "total-samples", "📊 Total Samples", "#3498db"
+                        ),
+                        self._create_stat_card(
+                            "mean-latency", "⚡ Mean Latency", "#e74c3c"
+                        ),
+                        self._create_stat_card(
+                            "p99-latency", "🔥 P99 Latency", "#f39c12"
+                        ),
+                    ],
+                    style={
+                        "display": "flex",
+                        "justifyContent": "space-around",
+                        "marginBottom": 30,
+                    },
+                ),
+                # Graphs - ✅ Make sure these IDs match the callback outputs
+                dcc.Graph(id="dual-histogram", style={"height": "450px"}),
+                dcc.Graph(id="log2-buckets", style={"height": "350px"}),
+                dcc.Graph(id="timeseries-graph", style={"height": "300px"}),
+                # Auto-update
+                dcc.Interval(id="interval-component", interval=1000, n_intervals=0),
+            ],
+            style={"padding": 20, "fontFamily": "Arial, sans-serif"},
+        )
+
+    def _create_stat_card(self, id_name: str, title: str, color: str):
+        """Create a statistics card."""
+        return html.Div(
+            [
+                html.H3(title, style={"color": color}),
+                html.H2(id=id_name, style={"fontSize": 48, "color": "#2c3e50"}),
+            ],
+            className="stat-box",
+            style={
+                "background": "white",
+                "padding": 20,
+                "borderRadius": 10,
+                "boxShadow": "0 4px 6px rgba(0,0,0,0.1)",
+                "textAlign": "center",
+                "flex": 1,
+                "margin": "0 10px",
+            },
+        )
+
+    def _setup_callbacks(self):
+        """Setup dashboard callbacks."""
+
+        @self.app.callback(
+            [
+                Output("total-samples", "children"),
+                Output("mean-latency", "children"),
+                Output("p99-latency", "children"),
+                Output("dual-histogram", "figure"),  # ✅ Match layout IDs
+                Output("log2-buckets", "figure"),  # ✅ Match layout IDs
+                Output("timeseries-graph", "figure"),  # ✅ Match layout IDs
+            ],
+            [Input("interval-component", "n_intervals")],
+        )
+        def update_dashboard(n):
+            stats = self.collector.get_stats()
+
+            if stats.total == 0:
+                return self._empty_state()
+
+            return (
+                f"{stats.total:,}",
+                f"{stats.mean:.1f} µs",
+                f"{stats.p99:.1f} µs",
+                self._create_dual_histogram(),
+                self._create_log2_buckets(),
+                self._create_timeseries(),
+            )
+
+    def _empty_state(self):
+        """Return empty state for dashboard."""
+        empty_fig = go.Figure()
+        empty_fig.update_layout(
+            title="Waiting for data... Generate some disk I/O!", template="plotly_white"
+        )
+        # ✅ Return 6 values (3 stats + 3 figures)
+        return "0", "0 µs", "0 µs", empty_fig, empty_fig, empty_fig
+
+    def _create_dual_histogram(self) -> go.Figure:
+        """Create side-by-side linear and log2 histograms."""
+        latencies = self.collector.get_all_latencies()
+
+        # Create subplots
+        fig = make_subplots(
+            rows=1,
+            cols=2,
+            subplot_titles=("Linear Scale", "Log2 Scale"),
+            horizontal_spacing=0.12,
+        )
+
+        # Linear histogram
+        fig.add_trace(
+            go.Histogram(
+                x=latencies,
+                nbinsx=50,
+                marker_color="rgb(55, 83, 109)",
+                opacity=0.75,
+                name="Linear",
+            ),
+            row=1,
+            col=1,
+        )
+
+        # Log2 histogram
+        log2_latencies = np.log2(latencies + 1)  # +1 to avoid log2(0)
+        fig.add_trace(
+            go.Histogram(
+                x=log2_latencies,
+                nbinsx=30,
+                marker_color="rgb(243, 156, 18)",
+                opacity=0.75,
+                name="Log2",
+            ),
+            row=1,
+            col=2,
+        )
+
+        # Update axes
+        fig.update_xaxes(title_text="Latency (µs)", row=1, col=1)
+        fig.update_xaxes(title_text="log2(Latency in µs)", row=1, col=2)
+        fig.update_yaxes(title_text="Count", row=1, col=1)
+        fig.update_yaxes(title_text="Count", row=1, col=2)
+
+        fig.update_layout(
+            title_text="📊 Latency Distribution (Linear vs Log2)",
+            template="plotly_white",
+            showlegend=False,
+            height=450,
+        )
+
+        return fig
+
+    def _create_log2_buckets(self) -> go.Figure:
+        """Create bar chart of log2 buckets (like BCC histogram)."""
+        buckets = self.collector.get_histogram_buckets()
+
+        if not buckets:
+            fig = go.Figure()
+            fig.update_layout(
+                title="🔥 Log2 Histogram - Waiting for data...", template="plotly_white"
+            )
+            return fig
+
+        # Sort buckets
+        sorted_buckets = sorted(buckets.keys())
+        counts = [buckets[b] for b in sorted_buckets]
+
+        # Create labels (e.g., "8-16µs", "16-32µs")
+        labels = []
+        hover_text = []
+        for bucket in sorted_buckets:
+            lower = 2**bucket
+            upper = 2 ** (bucket + 1)
+            labels.append(f"{lower}-{upper}")
+
+            # Calculate percentage
+            total = sum(counts)
+            pct = (buckets[bucket] / total) * 100 if total > 0 else 0
+            hover_text.append(
+                f"Range: {lower}-{upper} µs<br>"
+                f"Count: {buckets[bucket]:,}<br>"
+                f"Percentage: {pct:.2f}%"
+            )
+
+        # Create bar chart
+        fig = go.Figure()
+
+        fig.add_trace(
+            go.Bar(
+                x=labels,
+                y=counts,
+                marker=dict(
+                    color=counts,
+                    colorscale="YlOrRd",
+                    showscale=True,
+                    colorbar=dict(title="Count"),
+                ),
+                text=counts,
+                textposition="outside",
+                hovertext=hover_text,
+                hoverinfo="text",
+            )
+        )
+
+        fig.update_layout(
+            title="🔥 Log2 Histogram (BCC-style buckets)",
+            xaxis_title="Latency Range (µs)",
+            yaxis_title="Count",
+            template="plotly_white",
+            height=350,
+            xaxis=dict(tickangle=-45),
+        )
+
+        return fig
+
+    def _create_timeseries(self) -> go.Figure:
+        """Create time series figure."""
+        recent = self.collector.get_recent_latencies()
+
+        if not recent:
+            fig = go.Figure()
+            fig.update_layout(
+                title="⏱️ Real-time Latency - Waiting for data...",
+                template="plotly_white",
+            )
+            return fig
+
+        times = [d["time"] for d in recent]
+        lats = [d["latency"] for d in recent]
+
+        fig = go.Figure()
+        fig.add_trace(
+            go.Scatter(
+                x=times,
+                y=lats,
+                mode="lines",
+                line=dict(color="rgb(231, 76, 60)", width=2),
+                fill="tozeroy",
+                fillcolor="rgba(231, 76, 60, 0.2)",
+            )
+        )
+
+        fig.update_layout(
+            title="⏱️ Real-time Latency (Last 10,000 samples)",
+            xaxis_title="Time (seconds)",
+            yaxis_title="Latency (µs)",
+            template="plotly_white",
+            height=300,
+        )
+
+        return fig
+
+    def run(self, host: str = "0.0.0.0", port: int = 8050, debug: bool = False):
+        """Run the dashboard server."""
+        print(f"\n{'=' * 60}")
+        print(f"🚀 Dashboard running at: http://{host}:{port}")
+        print("   Access from your browser to see live graphs")
+        print(
+            "   Generate disk I/O to see data: dd if=/dev/zero of=/tmp/test bs=1M count=100"
+        )
+        print(f"{'=' * 60}\n")
+        self.app.run(debug=debug, host=host, port=port)

BCC-Examples/vfsreadlat_plotly/data_collector.py

@@ -0,0 +1,96 @@
+"""Data collection and management."""
+
+import threading
+import time
+import numpy as np
+from collections import deque
+from dataclasses import dataclass
+from typing import List, Dict
+
+
+@dataclass
+class LatencyStats:
+    """Statistics computed from latency data."""
+
+    total: int = 0
+    mean: float = 0.0
+    median: float = 0.0
+    min: float = 0.0
+    max: float = 0.0
+    p95: float = 0.0
+    p99: float = 0.0
+
+    @classmethod
+    def from_array(cls, data: np.ndarray) -> "LatencyStats":
+        """Compute stats from numpy array."""
+        if len(data) == 0:
+            return cls()
+
+        return cls(
+            total=len(data),
+            mean=float(np.mean(data)),
+            median=float(np.median(data)),
+            min=float(np.min(data)),
+            max=float(np.max(data)),
+            p95=float(np.percentile(data, 95)),
+            p99=float(np.percentile(data, 99)),
+        )
+
+
+class LatencyCollector:
+    """Collects and manages latency data from BPF."""
+
+    def __init__(self, bpf_object, buffer_size: int = 10000):
+        self.bpf = bpf_object
+        self.all_latencies: List[float] = []
+        self.recent_latencies = deque(maxlen=buffer_size)  # type: ignore [var-annotated]
+        self.start_time = time.time()
+        self._lock = threading.Lock()
+        self._poll_thread = None
+
+    def callback(self, cpu: int, event):
+        """Callback for BPF events."""
+        with self._lock:
+            self.all_latencies.append(event.delta_us)
+            self.recent_latencies.append(
+                {"time": time.time() - self.start_time, "latency": event.delta_us}
+            )
+
+    def start(self):
+        """Start collecting data."""
+        self.bpf["events"].open_perf_buffer(self.callback, struct_name="latency_event")
+
+        def poll_loop():
+            while True:
+                self.bpf["events"].poll(100)
+
+        self._poll_thread = threading.Thread(target=poll_loop, daemon=True)
+        self._poll_thread.start()
+        print("✅ Data collection started")
+
+    def get_all_latencies(self) -> np.ndarray:
+        """Get all latencies as numpy array."""
+        with self._lock:
+            return np.array(self.all_latencies) if self.all_latencies else np.array([])
+
+    def get_recent_latencies(self) -> List[Dict]:
+        """Get recent latencies with timestamps."""
+        with self._lock:
+            return list(self.recent_latencies)
+
+    def get_stats(self) -> LatencyStats:
+        """Compute current statistics."""
+        return LatencyStats.from_array(self.get_all_latencies())
+
+    def get_histogram_buckets(self) -> Dict[int, int]:
+        """Get log2 histogram buckets."""
+        latencies = self.get_all_latencies()
+        if len(latencies) == 0:
+            return {}
+
+        log_buckets = np.floor(np.log2(latencies + 1)).astype(int)
+        buckets = {}  # type: ignore [var-annotated]
+        for bucket in log_buckets:
+            buckets[bucket] = buckets.get(bucket, 0) + 1
+
+        return buckets

BCC-Examples/vfsreadlat_rich.py

@@ -0,0 +1,178 @@
+from pythonbpf import bpf, map, struct, section, bpfglobal, BPF
+from pythonbpf.helper import ktime, pid
+from pythonbpf.maps import HashMap, PerfEventArray
+from ctypes import c_void_p, c_uint64
+
+from rich.console import Console
+from rich.live import Live
+from rich.table import Table
+from rich.panel import Panel
+from rich.layout import Layout
+import numpy as np
+import threading
+import time
+from collections import Counter
+
+# ==================== BPF Setup ====================
+
+
+@bpf
+@struct
+class latency_event:
+    pid: c_uint64
+    delta_us: c_uint64
+
+
+@bpf
+@map
+def start() -> HashMap:
+    return HashMap(key=c_uint64, value=c_uint64, max_entries=10240)
+
+
+@bpf
+@map
+def events() -> PerfEventArray:
+    return PerfEventArray(key_size=c_uint64, value_size=c_uint64)
+
+
+@bpf
+@section("kprobe/vfs_read")
+def do_entry(ctx: c_void_p) -> c_uint64:
+    p, ts = pid(), ktime()
+    start.update(p, ts)
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@section("kretprobe/vfs_read")
+def do_return(ctx: c_void_p) -> c_uint64:
+    p = pid()
+    tsp = start.lookup(p)
+
+    if tsp:
+        delta_ns = ktime() - tsp
+
+        if delta_ns > 1000:
+            evt = latency_event()
+            evt.pid, evt.delta_us = p, delta_ns // 1000
+            events.output(evt)
+
+        start.delete(p)
+
+    return 0  # type: ignore [return-value]
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+console = Console()
+console.print("[bold green]Loading BPF program...[/]")
+
+b = BPF()
+b.load()
+b.attach_all()
+
+# ==================== Data Collection ====================
+
+all_latencies = []
+histogram_buckets = Counter()  # type: ignore [var-annotated]
+
+
+def callback(cpu, event):
+    all_latencies.append(event.delta_us)
+    # Create log2 bucket
+    bucket = int(np.floor(np.log2(event.delta_us + 1)))
+    histogram_buckets[bucket] += 1
+
+
+b["events"].open_perf_buffer(callback, struct_name="latency_event")
+
+
+def poll_events():
+    while True:
+        b["events"].poll(100)
+
+
+poll_thread = threading.Thread(target=poll_events, daemon=True)
+poll_thread.start()
+
+# ==================== Live Display ====================
+
+
+def generate_display():
+    layout = Layout()
+    layout.split_column(
+        Layout(name="header", size=3),
+        Layout(name="stats", size=8),
+        Layout(name="histogram", size=20),
+    )
+
+    # Header
+    layout["header"].update(
+        Panel("[bold cyan]🔥 VFS Read Latency Monitor[/]", style="bold white on blue")
+    )
+
+    # Stats
+    if len(all_latencies) > 0:
+        lats = np.array(all_latencies)
+        stats_table = Table(show_header=False, box=None, padding=(0, 2))
+        stats_table.add_column(style="bold cyan")
+        stats_table.add_column(style="bold yellow")
+
+        stats_table.add_row("📊 Total Samples:", f"{len(lats):,}")
+        stats_table.add_row("⚡ Mean Latency:", f"{np.mean(lats):.2f} µs")
+        stats_table.add_row("📉 Min Latency:", f"{np.min(lats):.2f} µs")
+        stats_table.add_row("📈 Max Latency:", f"{np.max(lats):.2f} µs")
+        stats_table.add_row("🎯 P95 Latency:", f"{np.percentile(lats, 95):.2f} µs")
+        stats_table.add_row("🔥 P99 Latency:", f"{np.percentile(lats, 99):.2f} µs")
+
+        layout["stats"].update(
+            Panel(stats_table, title="Statistics", border_style="green")
+        )
+    else:
+        layout["stats"].update(
+            Panel("[yellow]Waiting for data...[/]", border_style="yellow")
+        )
+
+    # Histogram
+    if histogram_buckets:
+        hist_table = Table(title="Latency Distribution", box=None)
+        hist_table.add_column("Range", style="cyan", no_wrap=True)
+        hist_table.add_column("Count", justify="right", style="yellow")
+        hist_table.add_column("Distribution", style="green")
+
+        max_count = max(histogram_buckets.values())
+
+        for bucket in sorted(histogram_buckets.keys()):
+            count = histogram_buckets[bucket]
+            lower = 2**bucket
+            upper = 2 ** (bucket + 1)
+
+            # Create bar
+            bar_width = int((count / max_count) * 40)
+            bar = "█" * bar_width
+
+            hist_table.add_row(
+                f"{lower:5d}-{upper:5d} µs",
+                f"{count:6d}",
+                f"[green]{bar}[/] {count / len(all_latencies) * 100:.1f}%",
+            )
+
+        layout["histogram"].update(Panel(hist_table, border_style="green"))
+
+    return layout
+
+
+try:
+    with Live(generate_display(), refresh_per_second=2, console=console) as live:
+        while True:
+            time.sleep(0.5)
+            live.update(generate_display())
+except KeyboardInterrupt:
+    console.print("\n[bold red]Stopping...[/]")
+
+    if all_latencies:
+        console.print(f"\n[bold green]✅ Collected {len(all_latencies):,} samples[/]")

LICENSE

@@ -200,4 +200,3 @@
    WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    See the License for the specific language governing permissions and
    limitations under the License.
-   

Makefile

@@ -1,14 +1,10 @@
-compile:
-	chmod +x ./tools/compile.py
-	./tools/compile.py ./examples/execve2.py
-
-install: 
+install:
 	pip install -e .
 
 clean:
 	rm -rf build dist *.egg-info
 	rm -rf examples/*.ll examples/*.o
 
-all: install compile
+all: clean install
 
 .PHONY: all clean

README.md

@@ -1,16 +1,209 @@
-# Python-BPF
-This is an LLVM IR generator for eBPF programs in Python. We use `llvmlite` to generate LLVM IR code from pure Python code. This is then compiled to LLVM object files, which can be loaded into the kernel for execution.
+<picture>
+  <source
+    media="(prefers-color-scheme: light)"
+    srcset="https://github.com/user-attachments/assets/f3738131-d7cb-4b5c-8699-c7010295a159"
+    width="450"
+    alt="Light‐mode image">
+  <img
+    src="https://github.com/user-attachments/assets/b175bf39-23cb-475d-a6e1-7b5c99a1ed72"
+    width="450"
+    alt="Dark‐mode image">
+</picture>
+<!-- Badges -->
+<p align="center">
+  <!-- PyPI -->
+  <a href="https://pypi.org/project/pythonbpf/"><img src="https://img.shields.io/pypi/v/pythonbpf?color=blue" alt="PyPI version"></a>
+  <!-- <a href="https://pypi.org/project/pythonbpf/"><img src="https://img.shields.io/pypi/pyversions/pythonbpf" alt="Python versions"></a> -->
+  <!-- <a href="https://pypi.org/project/pythonbpf/"><img src="https://img.shields.io/pypi/dm/pythonbpf" alt="PyPI downloads"></a> -->
+  <!-- <a href="https://pypi.org/project/pythonbpf/"><img src="https://img.shields.io/pypi/status/pythonbpf" alt="PyPI Status"></a> -->
+  <a href="https://pepy.tech/project/pythonbpf"><img src="https://pepy.tech/badge/pythonbpf" alt="Downloads"></a>
+  <!-- Build & CI -->
+  <a href="https://github.com/pythonbpf/python-bpf/actions"><img src="https://github.com/pythonbpf/python-bpf/actions/workflows/python-publish.yml/badge.svg" alt="Build Status"></a>
+  <!-- Meta -->
+  <a href="https://github.com/pythonbpf/python-bpf/blob/main/LICENSE"><img src="https://img.shields.io/github/license/pythonbpf/python-bpf" alt="License"></a>
+</p>
+
+
+Python-BPF is an LLVM IR generator for eBPF programs written in Python. It uses [llvmlite](https://github.com/numba/llvmlite) to generate LLVM IR and then compiles to LLVM object files. These object files can be loaded into the kernel for execution. Python-BPF performs compilation without relying on BCC.
+
+> **Note**: This project is under active development and not ready for production use.
+
+---
+
+## Overview
+
+* Generate eBPF programs directly from Python.
+* Compile to LLVM object files for kernel execution.
+* Built with `llvmlite` for IR generation.
+* Supports maps, helpers, and global definitions for BPF.
+* Companion project: [pylibbpf](https://github.com/pythonbpf/pylibbpf), which provides the bindings required for object loading and execution.
+
+---
+
+## Try It Out!
+Run
+```bash
+curl -s https://raw.githubusercontent.com/pythonbpf/Python-BPF/refs/heads/master/tools/setup.sh | sudo bash
+```
+
+## Installation
+
+Dependencies:
+
+* `clang`
+* Python ≥ 3.8
+
+Install via pip:
+
+```bash
+pip install pythonbpf pylibbpf
+```
+
+---
+
+## Example Usage
+
+```python
+import time
+from pythonbpf import bpf, map, section, bpfglobal, BPF
+from pythonbpf.helper import pid
+from pythonbpf.maps import HashMap
+from pylibbpf import *
+from ctypes import c_void_p, c_int64, c_uint64, c_int32
+import matplotlib.pyplot as plt
+
+
+# This program attaches an eBPF tracepoint to sys_enter_clone,
+# counts per-PID clone syscalls, stores them in a hash map,
+# and then plots the distribution as a histogram using matplotlib.
+# It provides a quick view of process creation activity over 10 seconds.
+
+@bpf
+@map
+def hist() -> HashMap:
+    return HashMap(key=c_int32, value=c_uint64, max_entries=4096)
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_clone")
+def hello(ctx: c_void_p) -> c_int64:
+    process_id = pid()
+    one = 1
+    prev = hist.lookup(process_id)
+    if prev:
+        previous_value = prev + 1
+        print(f"count: {previous_value} with {process_id}")
+        hist.update(process_id, previous_value)
+        return c_int64(0)
+    else:
+        hist.update(process_id, one)
+    return c_int64(0)
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+b = BPF()
+b.load_and_attach()
+hist = BpfMap(b, hist)
+print("Recording")
+time.sleep(10)
+
+counts = list(hist.values())
+
+plt.hist(counts, bins=20)
+plt.xlabel("Clone calls per PID")
+plt.ylabel("Frequency")
+plt.title("Syscall clone counts")
+plt.show()
+```
+---
+
+## Architecture
+
+Python-BPF provides a complete pipeline to write, compile, and load eBPF programs in Python:
+
+1. **Python Source Code**
+
+   * Users write BPF programs in Python using decorators like `@bpf`, `@map`, `@section`, and `@bpfglobal`.
+   * Maps (hash maps), helpers (e.g., `ktime`, `deref`), and tracepoints are defined using Python constructs, preserving a syntax close to standard Python.
+
+2. **AST Generation**
+
+   * The Python `ast` module parses the source code into an Abstract Syntax Tree (AST).
+   * Decorators and type annotations are captured to determine BPF maps, tracepoints, and global variables.
+
+3. **LLVM IR Emission**
+
+   * The AST is transformed into LLVM Intermediate Representation (IR) using `llvmlite`.
+   * IR captures BPF maps, control flow, assignments, and calls to helper functions.
+   * Debug information is emitted for easier inspection.
+
+4. **LLVM Object File Compilation**
+
+   * The LLVM IR (`.ll`) is compiled into a BPF target object file (`.o`) using `llc -march=bpf -O2`.
+   * This produces a kernel-loadable ELF object file containing the BPF bytecode.
+
+5. **libbpf Integration (via pylibbpf)**
+
+   * The compiled object file can be loaded into the kernel using `pylibbpf`.
+   * Maps, tracepoints, and program sections are initialized, and helper functions are resolved.
+   * Programs are attached to kernel hooks (e.g., syscalls) for execution.
+
+6. **Execution in Kernel**
+
+   * The kernel executes the loaded eBPF program.
+   * Hash maps, helpers, and global variables behave as defined in the Python source.
+   * Output can be read via BPF maps, helper functions, or trace printing.
+
+This architecture eliminates the need for embedding C code in Python, allowing full Python tooling support while generating true BPF object files ready for kernel execution.
+
+---
 
 ## Development
-Step 0. Make a virtual environment and activate it using `python3 -m venv .venv && source .venv/bin/activate`.  
-Step 1. Run `make install` to install the required dependencies.  
-Step 2. Run `make` to see the compilation output of the example.  
-Step 3. Run `check.sh` to check if generated object file passes through the verifier inside the examples directory.  
-Step 4. Run `make` in the `examples/c-form` directory to modify the example C BPF program to check the actual LLVM IR generated by clang.  
 
-### Development Notes
-Run ` ./check.sh run execve2.o;` in examples folder
+1. Create a virtual environment and activate it:
+
+   ```bash
+   python3 -m venv .venv
+   source .venv/bin/activate
+   ```
+
+2. Install dependencies:
+
+   ```bash
+   make install
+   ```
+   Then, run any example in `examples`
+3. Verify an object file with the kernel verifier:
+
+   ```bash
+   ./tools/check.sh check execve2.o
+   ```
+
+5. Run an object file using `bpftool`:
+
+   ```bash
+   ./tools/check.sh run execve2.o
+   ```
+
+6. Explore LLVM IR output from clang in `examples/c-form` by running `make`.
+
+---
+
+## Resources
+
+* [Video demonstration](https://youtu.be/eMyLW8iWbks)
+* [Slide deck](https://docs.google.com/presentation/d/1DsWDIVrpJhM4RgOETO9VWqUtEHo3-c7XIWmNpi6sTSo/edit?usp=sharing)
+
+---
 
 ## Authors
-- [@r41k0u](https://github.com/r41k0u)
-- [@varun-r-mallya](https://github.com/varun-r-mallya)
+
+* [@r41k0u](https://github.com/r41k0u)
+* [@varun-r-mallya](https://github.com/varun-r-mallya)
+
+---

examples/binops_demo.py

@@ -0,0 +1,55 @@
+from pythonbpf import bpf, map, section, bpfglobal, compile
+from pythonbpf.helper import ktime
+from pythonbpf.maps import HashMap
+
+from ctypes import c_void_p, c_int64, c_uint64
+
+# Instructions to how to run this program
+# 1. Install PythonBPF: pip install pythonbpf
+# 2. Run the program: python examples/binops_demo.py
+# 3. Run the program with sudo: sudo tools/check.sh run examples/binops_demo.py
+# 4. Start up any program and watch the output
+
+
+@bpf
+@map
+def last() -> HashMap:
+    return HashMap(key=c_uint64, value=c_uint64, max_entries=3)
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_execve")
+def do_trace(ctx: c_void_p) -> c_int64:
+    key = 0
+    tsp = last.lookup(key)
+    if tsp:
+        kt = ktime()
+        delta = kt - tsp
+        if delta < 1000000000:
+            time_ms = delta // 1000000
+            print(f"Execve syscall entered within last second, last {time_ms} ms ago")
+        last.delete(key)
+    else:
+        kt = ktime()
+        last.update(key, kt)
+    return c_int64(0)
+
+
+@bpf
+@section("tracepoint/syscalls/sys_exit_execve")
+def do_exit(ctx: c_void_p) -> c_int64:
+    va = 8
+    nm = 5 ^ va
+    al = 6 & 3
+    ru = nm + al
+    print(f"this is a variable {ru}")
+    return c_int64(0)
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+compile()

examples/blk_request.py

@@ -0,0 +1,28 @@
+from pythonbpf import bpf, map, section, bpfglobal, compile
+from pythonbpf.helper import ktime
+from pythonbpf.maps import HashMap
+
+from ctypes import c_void_p, c_int32, c_uint64
+
+
+@bpf
+@map
+def last() -> HashMap:
+    return HashMap(key=c_uint64, value=c_uint64, max_entries=3)
+
+
+@bpf
+@section("blk_start_request")
+def trace_start(ctx: c_void_p) -> c_int32:
+    ts = ktime()
+    print(f"req started {ts}")
+    return c_int32(0)
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+compile()

examples/c-form/Makefile

@@ -1,19 +0,0 @@
-BPF_CLANG := clang
-CFLAGS := -O2 -emit-llvm -target bpf -c
-
-SRC := $(wildcard *.bpf.c)
-LL := $(SRC:.bpf.c=.bpf.ll)
-OBJ := $(SRC:.bpf.c=.bpf.o)
-
-.PHONY: all clean
-
-all: $(LL) $(OBJ)
-
-%.bpf.o: %.bpf.c
-	$(BPF_CLANG) -O2 -g -target bpf -c $< -o $@
-
-%.bpf.ll: %.bpf.c
-	$(BPF_CLANG) $(CFLAGS) -g -S $< -o $@
-
-clean:
-	rm -f $(LL) $(OBJ)

examples/c-form/ex2.bpf.c

@@ -1,19 +0,0 @@
-#include <linux/bpf.h>
-#include <bpf/bpf_helpers.h>
-#define u64 unsigned long long
-#define u32 unsigned int
-
-struct {
-    __uint(type, BPF_MAP_TYPE_HASH);
-    __uint(max_entries, 1);
-    __type(key, u32);
-    __type(value, u64);
-} last SEC(".maps");
-
-SEC("tracepoint/syscalls/sys_enter_execve")
-int hello(struct pt_regs *ctx) {
-    bpf_printk("Hello, World!\n");
-    return 0;
-}
-
-char LICENSE[] SEC("license") = "GPL";

examples/clone_plot.py

@@ -0,0 +1,57 @@
+import time
+
+from pythonbpf import bpf, map, section, bpfglobal, BPF
+from pythonbpf.helper import pid
+from pythonbpf.maps import HashMap
+from ctypes import c_void_p, c_int64, c_uint64, c_int32
+import matplotlib.pyplot as plt
+
+# This program attaches an eBPF tracepoint to sys_enter_clone,
+# counts per-PID clone syscalls, stores them in a hash map,
+# and then plots the distribution as a histogram using matplotlib.
+# It provides a quick view of process creation activity over 10 seconds.
+# Everything is done with Python only code and with the new pylibbpf library.
+# Run `sudo /path/to/python/binary/ clone_plot.py`
+
+
+@bpf
+@map
+def hist() -> HashMap:
+    return HashMap(key=c_int32, value=c_uint64, max_entries=4096)
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_clone")
+def hello(ctx: c_void_p) -> c_int64:
+    process_id = pid()
+    one = 1
+    prev = hist.lookup(process_id)
+    if prev:
+        previous_value = prev + 1
+        print(f"count: {previous_value} with {process_id}")
+        hist.update(process_id, previous_value)
+        return c_int64(0)
+    else:
+        hist.update(process_id, one)
+    return c_int64(0)
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+b = BPF()
+b.load()
+b.attach_all()
+print("Recording")
+time.sleep(10)
+
+counts = list(b["hist"].values())
+
+plt.hist(counts, bins=20)
+plt.xlabel("Clone calls per PID")
+plt.ylabel("Frequency")
+plt.title("Syscall clone counts")
+plt.show()

examples/execve2.py

@@ -1,33 +0,0 @@
-from pythonbpf.decorators import bpf, map, section, bpfglobal
-from ctypes import c_void_p, c_int64, c_int32, c_uint64
-from pythonbpf.helpers import bpf_ktime_get_ns
-from pythonbpf.maps import HashMap
-
-
-@bpf
-@map
-def last() -> HashMap:
-    return HashMap(key_type=c_uint64, value_type=c_uint64, max_entries=1)
-
-@bpf
-@section("tracepoint/syscalls/sys_enter_execve")
-def hello(ctx: c_void_p) -> c_int32:
-    print("entered")
-    print("multi constant support")
-    return c_int32(0)
-
-
-@bpf
-@section("tracepoint/syscalls/sys_exit_execve")
-def hello_again(ctx: c_void_p) -> c_int64:
-    print("exited")
-    key = 0
-    tsp = last().lookup(key)
-    print(tsp)
-    ts = bpf_ktime_get_ns()
-    return c_int64(0)
-
-@bpf
-@bpfglobal
-def LICENSE() -> str:
-    return "GPL"

examples/execve3.py

@@ -1,40 +0,0 @@
-from pythonbpf.decorators import bpf, map, section, bpfglobal
-from ctypes import c_void_p, c_int64, c_int32, c_uint64
-from pythonbpf.helpers import bpf_ktime_get_ns
-from pythonbpf.maps import HashMap
-
-
-@bpf
-@map
-def last() -> HashMap:
-    return HashMap(key_type=c_uint64, value_type=c_uint64, max_entries=1)
-
-
-@bpf
-@section("tracepoint/syscalls/sys_enter_execve")
-def hello(ctx: c_void_p) -> c_int32:
-    print("entered")
-    print("multi constant support")
-    return c_int32(0)
-
-
-@bpf
-@section("tracepoint/syscalls/sys_exit_execve")
-def hello_again(ctx: c_void_p) -> c_int64:
-    print("exited")
-    key = 0
-    tsp = last().lookup(key)
-    if tsp:
-        delta = (bpf_ktime_get_ns() - tsp.value)
-        if delta < 1000000000:
-            print("execve called within last second")
-        last().delete(key)
-    ts = bpf_ktime_get_ns()
-    last().update(key, ts)
-    return c_int64(0)
-
-
-@bpf
-@bpfglobal
-def LICENSE() -> str:
-    return "GPL"

examples/hello_world.py

@@ -1,15 +1,26 @@
-# This is what it is going to look like
-# pylint: disable-all# type: ignore
-from pythonbpf.decorators import tracepoint, syscalls, bpfglobal, bpf
-from ctypes import c_void_p, c_int32
+from pythonbpf import bpf, section, bpfglobal, BPF, trace_pipe
+from ctypes import c_void_p, c_int64
+
+# Instructions to how to run this program
+# 1. Install PythonBPF: pip install pythonbpf
+# 2. Run the program: sudo python examples/hello_world.py
+# 4. Start up any program and watch the output
+
 
 @bpf
-@tracepoint(syscalls.sys_clone)
-def trace_clone(ctx: c_void_p) -> c_int32:
+@section("tracepoint/syscalls/sys_enter_execve")
+def hello_world(ctx: c_void_p) -> c_int64:
     print("Hello, World!")
-    return c_int32(0)
+    return c_int64(0)
+
 
 @bpf
 @bpfglobal
 def LICENSE() -> str:
     return "GPL"
+
+
+b = BPF()
+b.load()
+b.attach_all()
+trace_pipe()

examples/kprobes.py

@@ -0,0 +1,29 @@
+from pythonbpf import bpf, section, bpfglobal, BPF, trace_pipe
+from ctypes import c_void_p, c_int64
+
+
+@bpf
+@section("kretprobe/do_unlinkat")
+def hello_world(ctx: c_void_p) -> c_int64:
+    print("Hello, World!")
+    return c_int64(0)
+
+
+@bpf
+@section("kprobe/do_unlinkat")
+def hello_world2(ctx: c_void_p) -> c_int64:
+    print("Hello, World!")
+    return c_int64(0)
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+b = BPF()
+b.load()
+b.attach_all()
+print("running")
+trace_pipe()

examples/struct_and_perf.py

@@ -0,0 +1,41 @@
+from pythonbpf import bpf, map, struct, section, bpfglobal, compile
+from pythonbpf.helper import ktime, pid
+from pythonbpf.maps import PerfEventArray
+
+from ctypes import c_void_p, c_int32, c_uint64
+
+
+@bpf
+@struct
+class data_t:
+    pid: c_uint64
+    ts: c_uint64
+    comm: str(16)
+
+
+@bpf
+@map
+def events() -> PerfEventArray:
+    return PerfEventArray(key_size=c_int32, value_size=c_int32)
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_clone")
+def hello(ctx: c_void_p) -> c_int32:
+    dataobj = data_t()
+    strobj = "hellohellohello"
+    dataobj.pid = pid()
+    dataobj.ts = ktime()
+    # dataobj.comm = strobj
+    print(f"clone called at {dataobj.ts} by pid{dataobj.pid}, comm {strobj}")
+    events.output(dataobj)
+    return c_int32(0)
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+compile()

examples/sys_sync.py

@@ -0,0 +1,44 @@
+from pythonbpf import bpf, map, section, bpfglobal, compile
+from pythonbpf.helper import ktime
+from pythonbpf.maps import HashMap
+
+from ctypes import c_void_p, c_int64, c_uint64
+
+# Instructions to how to run this program
+# 1. Install PythonBPF: pip install pythonbpf
+# 2. Run the program: python examples/sys_sync.py
+# 3. Run the program with sudo: sudo tools/check.sh run examples/sys_sync.o
+# 4. Start a Python repl and `import os` and then keep entering `os.sync()` to see reponses.
+
+
+@bpf
+@map
+def last() -> HashMap:
+    return HashMap(key=c_uint64, value=c_uint64, max_entries=3)
+
+
+@bpf
+@section("tracepoint/syscalls/sys_enter_sync")
+def do_trace(ctx: c_void_p) -> c_int64:
+    key = 0
+    tsp = last.lookup(key)
+    if tsp:
+        kt = ktime()
+        delta = kt - tsp
+        if delta < 1000000000:
+            time_ms = delta // 1000000
+            print(f"sync called within last second, last {time_ms} ms ago")
+        last.delete(key)
+    else:
+        kt = ktime()
+        last.update(key, kt)
+    return c_int64(0)
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+compile()

examples/xdp_pass.py

@@ -0,0 +1,45 @@
+from pythonbpf import bpf, map, section, bpfglobal, compile, compile_to_ir
+from pythonbpf.helper import XDP_PASS
+from pythonbpf.maps import HashMap
+from ctypes import c_int64, c_void_p
+
+
+# Instructions to how to run this program
+# 1. Install PythonBPF: pip install pythonbpf
+# 2. Run the program: python examples/xdp_pass.py
+# 3. Run the program with sudo: sudo tools/check.sh run examples/xdp_pass.o
+# 4. Attach object file to any network device with something like ./check.sh xdp examples/xdp_pass.o tailscale0
+# 5. send traffic through the device and observe effects
+
+
+@bpf
+@map
+def count() -> HashMap:
+    return HashMap(key=c_int64, value=c_int64, max_entries=1)
+
+
+@bpf
+@section("xdp")
+def hello_world(ctx: c_void_p) -> c_int64:
+    key = 0
+    one = 1
+    prev = count.lookup(key)
+    if prev:
+        prevval = prev + 1
+        print(f"count: {prevval}")
+        count.update(key, prevval)
+        return XDP_PASS
+    else:
+        count.update(key, one)
+
+    return XDP_PASS
+
+
+@bpf
+@bpfglobal
+def LICENSE() -> str:
+    return "GPL"
+
+
+compile_to_ir("xdp_pass.py", "xdp_pass.ll")
+compile()

pyproject.toml

@@ -4,19 +4,34 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "pythonbpf"
-version = "0.1.0"
+version = "0.1.6"
 description = "Reduced Python frontend for eBPF"
 authors = [
   { name = "r41k0u", email="pragyanshchaturvedi18@gmail.com" },
   { name = "varun-r-mallya", email="varunrmallya@gmail.com" }
 ]
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Intended Audience :: Developers",
+    "Operating System :: POSIX :: Linux",
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.8",
+    "Programming Language :: Python :: 3.9",
+    "Programming Language :: Python :: 3.10",
+    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python",
+    "Topic :: Software Development :: Libraries :: Python Modules",
+    "Topic :: System :: Operating System Kernels :: Linux",
+]
 readme = "README.md"
 license = {text = "Apache-2.0"}
-requires-python = ">=3.8"
+requires-python = ">=3.10"
 
 dependencies = [
   "llvmlite",
-  "astpretty"
+  "astpretty",
+  "pylibbpf"
 ]
 
 [tool.setuptools.packages.find]

pythonbpf/__init__.py

@@ -0,0 +1,16 @@
+from .decorators import bpf, map, section, bpfglobal, struct
+from .codegen import compile_to_ir, compile, BPF
+from .utils import trace_pipe, trace_fields
+
+__all__ = [
+    "bpf",
+    "map",
+    "section",
+    "bpfglobal",
+    "struct",
+    "compile_to_ir",
+    "compile",
+    "BPF",
+    "trace_pipe",
+    "trace_fields",
+]

pythonbpf/allocation_pass.py

@@ -0,0 +1,374 @@
+import ast
+import logging
+import ctypes
+from llvmlite import ir
+from .local_symbol import LocalSymbol
+from pythonbpf.helper import HelperHandlerRegistry
+from pythonbpf.vmlinux_parser.dependency_node import Field
+from .expr import VmlinuxHandlerRegistry
+from pythonbpf.type_deducer import ctypes_to_ir
+
+logger = logging.getLogger(__name__)
+
+
+def create_targets_and_rvals(stmt):
+    """Create lists of targets and right-hand values from an assignment statement."""
+    if isinstance(stmt.targets[0], ast.Tuple):
+        if not isinstance(stmt.value, ast.Tuple):
+            logger.warning("Mismatched multi-target assignment, skipping allocation")
+            return [], []
+        targets, rvals = stmt.targets[0].elts, stmt.value.elts
+        if len(targets) != len(rvals):
+            logger.warning("length of LHS != length of RHS, skipping allocation")
+            return [], []
+        return targets, rvals
+    return stmt.targets, [stmt.value]
+
+
+def handle_assign_allocation(builder, stmt, local_sym_tab, structs_sym_tab):
+    """Handle memory allocation for assignment statements."""
+
+    logger.info(f"Handling assignment for allocation: {ast.dump(stmt)}")
+
+    # NOTE: Support multi-target assignments (e.g.: a, b = 1, 2)
+    targets, rvals = create_targets_and_rvals(stmt)
+
+    for target, rval in zip(targets, rvals):
+        # Skip non-name targets (e.g., struct field assignments)
+        if isinstance(target, ast.Attribute):
+            logger.debug(
+                f"Struct field assignment to {target.attr}, no allocation needed"
+            )
+            continue
+
+        if not isinstance(target, ast.Name):
+            logger.warning(
+                f"Unsupported assignment target type: {type(target).__name__}"
+            )
+            continue
+
+        var_name = target.id
+        # Skip if already allocated
+        if var_name in local_sym_tab:
+            logger.debug(f"Variable {var_name} already allocated, skipping")
+            continue
+
+        # Determine type and allocate based on rval
+        if isinstance(rval, ast.Call):
+            _allocate_for_call(builder, var_name, rval, local_sym_tab, structs_sym_tab)
+        elif isinstance(rval, ast.Constant):
+            _allocate_for_constant(builder, var_name, rval, local_sym_tab)
+        elif isinstance(rval, ast.BinOp):
+            _allocate_for_binop(builder, var_name, local_sym_tab)
+        elif isinstance(rval, ast.Name):
+            # Variable-to-variable assignment (b = a)
+            _allocate_for_name(builder, var_name, rval, local_sym_tab)
+        elif isinstance(rval, ast.Attribute):
+            # Struct field-to-variable assignment (a = dat.fld)
+            _allocate_for_attribute(
+                builder, var_name, rval, local_sym_tab, structs_sym_tab
+            )
+        else:
+            logger.warning(
+                f"Unsupported assignment value type for {var_name}: {type(rval).__name__}"
+            )
+
+
+def _allocate_for_call(builder, var_name, rval, local_sym_tab, structs_sym_tab):
+    """Allocate memory for variable assigned from a call."""
+
+    if isinstance(rval.func, ast.Name):
+        call_type = rval.func.id
+
+        # C type constructors
+        if call_type in ("c_int32", "c_int64", "c_uint32", "c_uint64", "c_void_p"):
+            ir_type = ctypes_to_ir(call_type)
+            var = builder.alloca(ir_type, name=var_name)
+            var.align = ir_type.width // 8
+            local_sym_tab[var_name] = LocalSymbol(var, ir_type)
+            logger.info(f"Pre-allocated {var_name} as {call_type}")
+
+        # Helper functions
+        elif HelperHandlerRegistry.has_handler(call_type):
+            ir_type = ir.IntType(64)  # Assume i64 return type
+            var = builder.alloca(ir_type, name=var_name)
+            var.align = 8
+            local_sym_tab[var_name] = LocalSymbol(var, ir_type)
+            logger.info(f"Pre-allocated {var_name} for helper {call_type}")
+
+        # Deref function
+        elif call_type == "deref":
+            ir_type = ir.IntType(64)  # Assume i64 return type
+            var = builder.alloca(ir_type, name=var_name)
+            var.align = 8
+            local_sym_tab[var_name] = LocalSymbol(var, ir_type)
+            logger.info(f"Pre-allocated {var_name} for deref")
+
+        # Struct constructors
+        elif call_type in structs_sym_tab:
+            struct_info = structs_sym_tab[call_type]
+            var = builder.alloca(struct_info.ir_type, name=var_name)
+            local_sym_tab[var_name] = LocalSymbol(var, struct_info.ir_type, call_type)
+            logger.info(f"Pre-allocated {var_name} for struct {call_type}")
+
+        else:
+            logger.warning(f"Unknown call type for allocation: {call_type}")
+
+    elif isinstance(rval.func, ast.Attribute):
+        # Map method calls - need double allocation for ptr handling
+        _allocate_for_map_method(builder, var_name, local_sym_tab)
+
+    else:
+        logger.warning(f"Unsupported call function type for {var_name}")
+
+
+def _allocate_for_map_method(builder, var_name, local_sym_tab):
+    """Allocate memory for variable assigned from map method (double alloc)."""
+
+    # Main variable (pointer to pointer)
+    ir_type = ir.PointerType(ir.IntType(64))
+    var = builder.alloca(ir_type, name=var_name)
+    local_sym_tab[var_name] = LocalSymbol(var, ir_type)
+
+    # Temporary variable for computed values
+    tmp_ir_type = ir.IntType(64)
+    var_tmp = builder.alloca(tmp_ir_type, name=f"{var_name}_tmp")
+    local_sym_tab[f"{var_name}_tmp"] = LocalSymbol(var_tmp, tmp_ir_type)
+
+    logger.info(f"Pre-allocated {var_name} and {var_name}_tmp for map method")
+
+
+def _allocate_for_constant(builder, var_name, rval, local_sym_tab):
+    """Allocate memory for variable assigned from a constant."""
+
+    if isinstance(rval.value, bool):
+        ir_type = ir.IntType(1)
+        var = builder.alloca(ir_type, name=var_name)
+        var.align = 1
+        local_sym_tab[var_name] = LocalSymbol(var, ir_type)
+        logger.info(f"Pre-allocated {var_name} as bool")
+
+    elif isinstance(rval.value, int):
+        ir_type = ir.IntType(64)
+        var = builder.alloca(ir_type, name=var_name)
+        var.align = 8
+        local_sym_tab[var_name] = LocalSymbol(var, ir_type)
+        logger.info(f"Pre-allocated {var_name} as i64")
+
+    elif isinstance(rval.value, str):
+        ir_type = ir.PointerType(ir.IntType(8))
+        var = builder.alloca(ir_type, name=var_name)
+        var.align = 8
+        local_sym_tab[var_name] = LocalSymbol(var, ir_type)
+        logger.info(f"Pre-allocated {var_name} as string")
+
+    else:
+        logger.warning(
+            f"Unsupported constant type for {var_name}: {type(rval.value).__name__}"
+        )
+
+
+def _allocate_for_binop(builder, var_name, local_sym_tab):
+    """Allocate memory for variable assigned from a binary operation."""
+    ir_type = ir.IntType(64)  # Assume i64 result
+    var = builder.alloca(ir_type, name=var_name)
+    var.align = 8
+    local_sym_tab[var_name] = LocalSymbol(var, ir_type)
+    logger.info(f"Pre-allocated {var_name} for binop result")
+
+
+def _get_type_name(ir_type):
+    """Get a string representation of an IR type."""
+    if isinstance(ir_type, ir.IntType):
+        return f"i{ir_type.width}"
+    elif isinstance(ir_type, ir.PointerType):
+        return "ptr"
+    elif isinstance(ir_type, ir.ArrayType):
+        return f"[{ir_type.count}x{_get_type_name(ir_type.element)}]"
+    else:
+        return str(ir_type).replace(" ", "")
+
+
+def allocate_temp_pool(builder, max_temps, local_sym_tab):
+    """Allocate the temporary scratch space pool for helper arguments."""
+    if not max_temps:
+        logger.info("No temp pool allocation needed")
+        return
+
+    for tmp_type, cnt in max_temps.items():
+        type_name = _get_type_name(tmp_type)
+        logger.info(f"Allocating temp pool of {cnt} variables of type {type_name}")
+        for i in range(cnt):
+            temp_name = f"__helper_temp_{type_name}_{i}"
+            temp_var = builder.alloca(tmp_type, name=temp_name)
+            temp_var.align = _get_alignment(tmp_type)
+            local_sym_tab[temp_name] = LocalSymbol(temp_var, tmp_type)
+            logger.debug(f"Allocated temp variable: {temp_name}")
+
+
+def _allocate_for_name(builder, var_name, rval, local_sym_tab):
+    """Allocate memory for variable-to-variable assignment (b = a)."""
+    source_var = rval.id
+
+    if source_var not in local_sym_tab:
+        logger.error(f"Source variable '{source_var}' not found in symbol table")
+        return
+
+    # Get type and metadata from source variable
+    source_symbol = local_sym_tab[source_var]
+
+    # Allocate with same type and alignment
+    var = _allocate_with_type(builder, var_name, source_symbol.ir_type)
+    local_sym_tab[var_name] = LocalSymbol(
+        var, source_symbol.ir_type, source_symbol.metadata
+    )
+
+    logger.info(
+        f"Pre-allocated {var_name} from {source_var} with type {source_symbol.ir_type}"
+    )
+
+
+def _allocate_for_attribute(builder, var_name, rval, local_sym_tab, structs_sym_tab):
+    """Allocate memory for struct field-to-variable assignment (a = dat.fld)."""
+    if not isinstance(rval.value, ast.Name):
+        logger.warning(f"Complex attribute access not supported for {var_name}")
+        return
+
+    struct_var = rval.value.id
+    field_name = rval.attr
+
+    # Validate struct and field
+    if struct_var not in local_sym_tab:
+        logger.error(f"Struct variable '{struct_var}' not found")
+        return
+
+    struct_type: type = local_sym_tab[struct_var].metadata
+    if not struct_type or struct_type not in structs_sym_tab:
+        if VmlinuxHandlerRegistry.is_vmlinux_struct(struct_type.__name__):
+            # Handle vmlinux struct field access
+            vmlinux_struct_name = struct_type.__name__
+            if not VmlinuxHandlerRegistry.has_field(vmlinux_struct_name, field_name):
+                logger.error(
+                    f"Field '{field_name}' not found in vmlinux struct '{vmlinux_struct_name}'"
+                )
+                return
+
+            field_type: tuple[ir.GlobalVariable, Field] = (
+                VmlinuxHandlerRegistry.get_field_type(vmlinux_struct_name, field_name)
+            )
+            field_ir, field = field_type
+            # TODO: For now, we only support integer type allocations.
+            # This always assumes first argument of function to be the context struct
+            base_ptr = builder.function.args[0]
+            local_sym_tab[
+                struct_var
+            ].var = base_ptr  # This is repurposing of var to store the pointer of the base type
+            local_sym_tab[struct_var].ir_type = field_ir
+
+            # Determine the actual IR type based on the field's type
+            actual_ir_type = None
+
+            # Check if it's a ctypes primitive
+            if field.type.__module__ == ctypes.__name__:
+                try:
+                    field_size_bytes = ctypes.sizeof(field.type)
+                    field_size_bits = field_size_bytes * 8
+
+                    if field_size_bits in [8, 16, 32, 64]:
+                        # Special case: struct_xdp_md i32 fields should allocate as i64
+                        # because load_ctx_field will zero-extend them to i64
+                        if (
+                            vmlinux_struct_name == "struct_xdp_md"
+                            and field_size_bits == 32
+                        ):
+                            actual_ir_type = ir.IntType(64)
+                            logger.info(
+                                f"Allocating {var_name} as i64 for i32 field from struct_xdp_md.{field_name} "
+                                "(will be zero-extended during load)"
+                            )
+                        else:
+                            actual_ir_type = ir.IntType(field_size_bits)
+                    else:
+                        logger.warning(
+                            f"Unusual field size {field_size_bits} bits for {field_name}"
+                        )
+                        actual_ir_type = ir.IntType(64)
+                except Exception as e:
+                    logger.warning(
+                        f"Could not determine size for ctypes field {field_name}: {e}"
+                    )
+                    actual_ir_type = ir.IntType(64)
+
+            # Check if it's a nested vmlinux struct or complex type
+            elif field.type.__module__ == "vmlinux":
+                # For pointers to structs, use pointer type (64-bit)
+                if field.ctype_complex_type is not None and issubclass(
+                    field.ctype_complex_type, ctypes._Pointer
+                ):
+                    actual_ir_type = ir.IntType(64)  # Pointer is always 64-bit
+                # For embedded structs, this is more complex - might need different handling
+                else:
+                    logger.warning(
+                        f"Field {field_name} is a nested vmlinux struct, using i64 for now"
+                    )
+                    actual_ir_type = ir.IntType(64)
+            else:
+                logger.warning(
+                    f"Unknown field type module {field.type.__module__} for {field_name}"
+                )
+                actual_ir_type = ir.IntType(64)
+
+            # Allocate with the actual IR type, not the GlobalVariable
+            var = _allocate_with_type(builder, var_name, actual_ir_type)
+            local_sym_tab[var_name] = LocalSymbol(var, actual_ir_type, field)
+
+            logger.info(
+                f"Pre-allocated {var_name} from vmlinux struct {vmlinux_struct_name}.{field_name}"
+            )
+            return
+        else:
+            logger.error(f"Struct type '{struct_type}' not found")
+        return
+
+    struct_info = structs_sym_tab[struct_type]
+    if field_name not in struct_info.fields:
+        logger.error(f"Field '{field_name}' not found in struct '{struct_type}'")
+        return
+
+    # Get field type
+    field_type = struct_info.field_type(field_name)
+
+    # Special case: char array -> allocate as i8* pointer instead
+    if (
+        isinstance(field_type, ir.ArrayType)
+        and isinstance(field_type.element, ir.IntType)
+        and field_type.element.width == 8
+    ):
+        alloc_type = ir.PointerType(ir.IntType(8))
+        logger.info(f"Allocating {var_name} as i8* (pointer to char array)")
+    else:
+        alloc_type = field_type
+
+    var = _allocate_with_type(builder, var_name, alloc_type)
+    local_sym_tab[var_name] = LocalSymbol(var, alloc_type)
+
+    logger.info(
+        f"Pre-allocated {var_name} from {struct_var}.{field_name} with type {alloc_type}"
+    )
+
+
+def _allocate_with_type(builder, var_name, ir_type):
+    """Allocate variable with appropriate alignment for type."""
+    var = builder.alloca(ir_type, name=var_name)
+    var.align = _get_alignment(ir_type)
+    return var
+
+
+def _get_alignment(ir_type):
+    """Get appropriate alignment for IR type."""
+    if isinstance(ir_type, ir.IntType):
+        return ir_type.width // 8
+    elif isinstance(ir_type, ir.ArrayType) and isinstance(ir_type.element, ir.IntType):
+        return ir_type.element.width // 8
+    else:
+        return 8  # Default: pointer size

pythonbpf/assign_pass.py

@@ -0,0 +1,252 @@
+import ast
+import logging
+from llvmlite import ir
+from pythonbpf.expr import eval_expr
+from pythonbpf.helper import emit_probe_read_kernel_str_call
+from pythonbpf.type_deducer import ctypes_to_ir
+from pythonbpf.vmlinux_parser.dependency_node import Field
+
+logger = logging.getLogger(__name__)
+
+
+def handle_struct_field_assignment(
+    func, module, builder, target, rval, local_sym_tab, map_sym_tab, structs_sym_tab
+):
+    """Handle struct field assignment (obj.field = value)."""
+
+    var_name = target.value.id
+    field_name = target.attr
+
+    if var_name not in local_sym_tab:
+        logger.error(f"Variable '{var_name}' not found in symbol table")
+        return
+
+    struct_type = local_sym_tab[var_name].metadata
+    struct_info = structs_sym_tab[struct_type]
+
+    if field_name not in struct_info.fields:
+        logger.error(f"Field '{field_name}' not found in struct '{struct_type}'")
+        return
+
+    # Get field pointer and evaluate value
+    field_ptr = struct_info.gep(builder, local_sym_tab[var_name].var, field_name)
+    field_type = struct_info.field_type(field_name)
+    val_result = eval_expr(
+        func, module, builder, rval, local_sym_tab, map_sym_tab, structs_sym_tab
+    )
+
+    if val_result is None:
+        logger.error(f"Failed to evaluate value for {var_name}.{field_name}")
+        return
+
+    val, val_type = val_result
+
+    # Special case: i8* string to [N x i8] char array
+    if _is_char_array(field_type) and _is_i8_ptr(val_type):
+        _copy_string_to_char_array(
+            func,
+            module,
+            builder,
+            val,
+            field_ptr,
+            field_type,
+            local_sym_tab,
+            map_sym_tab,
+            structs_sym_tab,
+        )
+        logger.info(f"Copied string to char array {var_name}.{field_name}")
+        return
+
+    # Regular assignment
+    builder.store(val, field_ptr)
+    logger.info(f"Assigned to struct field {var_name}.{field_name}")
+
+
+def _copy_string_to_char_array(
+    func,
+    module,
+    builder,
+    src_ptr,
+    dst_ptr,
+    array_type,
+    local_sym_tab,
+    map_sym_tab,
+    struct_sym_tab,
+):
+    """Copy string (i8*) to char array ([N x i8]) using bpf_probe_read_kernel_str"""
+
+    array_size = array_type.count
+
+    # Get pointer to first element: [N x i8]* -> i8*
+    dst_i8_ptr = builder.gep(
+        dst_ptr,
+        [ir.Constant(ir.IntType(32), 0), ir.Constant(ir.IntType(32), 0)],
+        inbounds=True,
+    )
+
+    # Use the shared emitter function
+    emit_probe_read_kernel_str_call(builder, dst_i8_ptr, array_size, src_ptr)
+
+
+def _is_char_array(ir_type):
+    """Check if type is [N x i8]."""
+    return (
+        isinstance(ir_type, ir.ArrayType)
+        and isinstance(ir_type.element, ir.IntType)
+        and ir_type.element.width == 8
+    )
+
+
+def _is_i8_ptr(ir_type):
+    """Check if type is i8*."""
+    return (
+        isinstance(ir_type, ir.PointerType)
+        and isinstance(ir_type.pointee, ir.IntType)
+        and ir_type.pointee.width == 8
+    )
+
+
+def handle_variable_assignment(
+    func, module, builder, var_name, rval, local_sym_tab, map_sym_tab, structs_sym_tab
+):
+    """Handle single named variable assignment."""
+
+    if var_name not in local_sym_tab:
+        logger.error(f"Variable {var_name} not declared.")
+        return False
+
+    var_ptr = local_sym_tab[var_name].var
+    var_type = local_sym_tab[var_name].ir_type
+
+    # NOTE: Special case for struct initialization
+    if isinstance(rval, ast.Call) and isinstance(rval.func, ast.Name):
+        struct_name = rval.func.id
+        if struct_name in structs_sym_tab and len(rval.args) == 0:
+            struct_info = structs_sym_tab[struct_name]
+            ir_struct = struct_info.ir_type
+
+            builder.store(ir.Constant(ir_struct, None), var_ptr)
+            logger.info(f"Initialized struct {struct_name} for variable {var_name}")
+            return True
+
+    # Special case: struct field char array -> pointer
+    # Handle this before eval_expr to get the pointer, not the value
+    if isinstance(rval, ast.Attribute) and isinstance(rval.value, ast.Name):
+        converted_val = _try_convert_char_array_to_ptr(
+            rval, var_type, builder, local_sym_tab, structs_sym_tab
+        )
+        if converted_val is not None:
+            builder.store(converted_val, var_ptr)
+            logger.info(f"Assigned char array pointer to {var_name}")
+            return True
+
+    val_result = eval_expr(
+        func, module, builder, rval, local_sym_tab, map_sym_tab, structs_sym_tab
+    )
+    if val_result is None:
+        logger.error(f"Failed to evaluate value for {var_name}")
+        return False
+
+    val, val_type = val_result
+    logger.info(f"Evaluated value for {var_name}: {val} of type {val_type}, {var_type}")
+    if val_type != var_type:
+        if isinstance(val_type, Field):
+            logger.info("Handling assignment to struct field")
+            # Special handling for struct_xdp_md i32 fields that are zero-extended to i64
+            # The load_ctx_field already extended them, so val is i64 but val_type.type shows c_uint
+            if (
+                hasattr(val_type, "type")
+                and val_type.type.__name__ == "c_uint"
+                and isinstance(var_type, ir.IntType)
+                and var_type.width == 64
+            ):
+                # This is the struct_xdp_md case - value is already i64
+                builder.store(val, var_ptr)
+                logger.info(
+                    f"Assigned zero-extended struct_xdp_md i32 field to {var_name} (i64)"
+                )
+                return True
+            # TODO: handling only ctype struct fields for now. Handle other stuff too later.
+            elif var_type == ctypes_to_ir(val_type.type.__name__):
+                builder.store(val, var_ptr)
+                logger.info(f"Assigned ctype struct field to {var_name}")
+                return True
+            else:
+                logger.error(
+                    f"Failed to assign ctype struct field to {var_name}: {val_type} != {var_type}"
+                )
+                return False
+        elif isinstance(val_type, ir.IntType) and isinstance(var_type, ir.IntType):
+            # Allow implicit int widening
+            if val_type.width < var_type.width:
+                val = builder.sext(val, var_type)
+                logger.info(f"Implicitly widened int for variable {var_name}")
+            elif val_type.width > var_type.width:
+                val = builder.trunc(val, var_type)
+                logger.info(f"Implicitly truncated int for variable {var_name}")
+        elif isinstance(val_type, ir.IntType) and isinstance(var_type, ir.PointerType):
+            # NOTE: This is assignment to a PTR_TO_MAP_VALUE_OR_NULL
+            logger.info(
+                f"Creating temporary variable for pointer assignment to {var_name}"
+            )
+            var_ptr_tmp = local_sym_tab[f"{var_name}_tmp"].var
+            builder.store(val, var_ptr_tmp)
+            val = var_ptr_tmp
+        else:
+            logger.error(
+                f"Type mismatch for variable {var_name}: {val_type} vs {var_type}"
+            )
+            return False
+
+    builder.store(val, var_ptr)
+    logger.info(f"Assigned value to variable {var_name}")
+    return True
+
+
+def _try_convert_char_array_to_ptr(
+    rval, var_type, builder, local_sym_tab, structs_sym_tab
+):
+    """Try to convert char array field to i8* pointer"""
+    # Only convert if target is i8*
+    if not (
+        isinstance(var_type, ir.PointerType)
+        and isinstance(var_type.pointee, ir.IntType)
+        and var_type.pointee.width == 8
+    ):
+        return None
+
+    struct_var = rval.value.id
+    field_name = rval.attr
+
+    # Validate struct
+    if struct_var not in local_sym_tab:
+        return None
+
+    struct_type = local_sym_tab[struct_var].metadata
+    if not struct_type or struct_type not in structs_sym_tab:
+        return None
+
+    struct_info = structs_sym_tab[struct_type]
+    if field_name not in struct_info.fields:
+        return None
+
+    field_type = struct_info.field_type(field_name)
+
+    # Check if it's a char array
+    if not (
+        isinstance(field_type, ir.ArrayType)
+        and isinstance(field_type.element, ir.IntType)
+        and field_type.element.width == 8
+    ):
+        return None
+
+    # Get pointer to struct field
+    struct_ptr = local_sym_tab[struct_var].var
+    field_ptr = struct_info.gep(builder, struct_ptr, field_name)
+
+    # GEP to first element: [N x i8]* -> i8*
+    return builder.gep(
+        field_ptr,
+        [ir.Constant(ir.IntType(32), 0), ir.Constant(ir.IntType(32), 0)],
+        inbounds=True,
+    )

pythonbpf/bpf_helper_handler.py

@@ -1,64 +0,0 @@
-import ast
-from llvmlite import ir
-
-
-def bpf_ktime_get_ns_emitter(call, module, builder, func):
-    pass
-
-
-def bpf_map_lookup_elem_emitter(map_ptr, key_ptr, module, builder):
-    """
-    Emit LLVM IR for bpf_map_lookup_elem helper function call.
-    """
-    # Cast pointers to void*
-    map_void_ptr = builder.bitcast(map_ptr, ir.PointerType())
-
-    # Define function type for bpf_map_lookup_elem
-    # The function takes two void* arguments and returns void*
-    fn_type = ir.FunctionType(
-        ir.PointerType(),  # Return type: void*
-        [ir.PointerType(), ir.PointerType()],  # Args: (void*, void*)
-        var_arg=False
-    )
-    fn_ptr_type = ir.PointerType(fn_type)
-
-    # Helper ID 1 is bpf_map_lookup_elem
-    fn_addr = ir.Constant(ir.IntType(64), 1)
-    fn_ptr = builder.inttoptr(fn_addr, fn_ptr_type)
-
-    # Call the helper function
-    result = builder.call(fn_ptr, [map_void_ptr, key_ptr], tail=False)
-
-    return result
-
-
-def bpf_printk_emitter(call, module, builder, func):
-    if not hasattr(func, "_fmt_counter"):
-        func._fmt_counter = 0
-
-    for arg in call.args:
-        if isinstance(arg, ast.Constant) and isinstance(arg.value, str):
-            fmt_str = arg.value + "\n" + "\0"
-            fmt_name = f"{func.name}____fmt{func._fmt_counter}"
-            func._fmt_counter += 1
-
-            fmt_gvar = ir.GlobalVariable(
-                module, ir.ArrayType(ir.IntType(8), len(fmt_str)), name=fmt_name)
-            fmt_gvar.global_constant = True
-            fmt_gvar.initializer = ir.Constant(     # type: ignore
-                ir.ArrayType(ir.IntType(8), len(fmt_str)),
-                bytearray(fmt_str.encode("utf8"))
-            )
-            fmt_gvar.linkage = "internal"
-            fmt_gvar.align = 1      # type: ignore
-
-            fmt_ptr = builder.bitcast(fmt_gvar, ir.PointerType())
-
-            fn_type = ir.FunctionType(ir.IntType(
-                64), [ir.PointerType(), ir.IntType(32)], var_arg=True)
-            fn_ptr_type = ir.PointerType(fn_type)
-            fn_addr = ir.Constant(ir.IntType(64), 6)
-            fn_ptr = builder.inttoptr(fn_addr, fn_ptr_type)
-
-            builder.call(fn_ptr, [fmt_ptr, ir.Constant(
-                ir.IntType(32), len(fmt_str))], tail=True)

pythonbpf/codegen.py

@@ -1,17 +1,65 @@
 import ast
 from llvmlite import ir
 from .license_pass import license_processing
-from .functions_pass import func_proc
-from .maps_pass import maps_proc
-from .globals_pass import globals_processing
+from .functions import func_proc
+from .maps import maps_proc
+from .structs import structs_proc
+from .vmlinux_parser import vmlinux_proc
+from pythonbpf.vmlinux_parser.vmlinux_exports_handler import VmlinuxHandler
+from .expr import VmlinuxHandlerRegistry
+from .globals_pass import (
+    globals_list_creation,
+    globals_processing,
+    populate_global_symbol_table,
+)
+from .debuginfo import DW_LANG_C11, DwarfBehaviorEnum, DebugInfoGenerator
 import os
+import subprocess
+import inspect
+from pathlib import Path
+from pylibbpf import BpfObject
+import tempfile
+from logging import Logger
+import logging
+import re
+
+logger: Logger = logging.getLogger(__name__)
+
+VERSION = "v0.1.6"
+
+
+def finalize_module(original_str):
+    """After all IR generation is complete, we monkey patch btf_ama attribute"""
+
+    # Create a string with applied transformation of btf_ama attribute addition to BTF struct field accesses.
+    pattern = r'(@"llvm\.[^"]+:[^"]*" = external global i64, !llvm\.preserve\.access\.index ![0-9]+)'
+    replacement = r'\1 "btf_ama"'
+    return re.sub(pattern, replacement, original_str)
+
+
+def bpf_passthrough_gen(module):
+    i32_ty = ir.IntType(32)
+    ptr_ty = ir.PointerType(ir.IntType(8))
+    fnty = ir.FunctionType(ptr_ty, [i32_ty, ptr_ty])
+
+    # Declare the intrinsic
+    passthrough = ir.Function(module, fnty, "llvm.bpf.passthrough.p0.p0")
+
+    # Set function attributes
+    # TODO: the ones commented are supposed to be there but cannot be added due to llvmlite limitations at the moment
+    # passthrough.attributes.add("nofree")
+    # passthrough.attributes.add("nosync")
+    passthrough.attributes.add("nounwind")
+    # passthrough.attributes.add("memory(none)")
+
+    return passthrough
 
 
 def find_bpf_chunks(tree):
     """Find all functions decorated with @bpf in the AST."""
     bpf_functions = []
     for node in ast.walk(tree):
-        if isinstance(node, ast.FunctionDef):
+        if isinstance(node, ast.FunctionDef) or isinstance(node, ast.ClassDef):
             for decorator in node.decorator_list:
                 if isinstance(decorator, ast.Name) and decorator.id == "bpf":
                     bpf_functions.append(node)
@@ -21,20 +69,34 @@ def find_bpf_chunks(tree):
 
 def processor(source_code, filename, module):
     tree = ast.parse(source_code, filename)
-    print(ast.dump(tree, indent=4))
+    logger.debug(ast.dump(tree, indent=4))
 
     bpf_chunks = find_bpf_chunks(tree)
     for func_node in bpf_chunks:
-        print(f"Found BPF function: {func_node.name}")
+        logger.info(f"Found BPF function/struct: {func_node.name}")
 
-    map_sym_tab = maps_proc(tree, module, bpf_chunks)
-    func_proc(tree, module, bpf_chunks, map_sym_tab)
+    bpf_passthrough_gen(module)
 
+    vmlinux_symtab = vmlinux_proc(tree, module)
+    if vmlinux_symtab:
+        handler = VmlinuxHandler.initialize(vmlinux_symtab)
+        VmlinuxHandlerRegistry.set_handler(handler)
+
+    populate_global_symbol_table(tree, module)
     license_processing(tree, module)
     globals_processing(tree, module)
+    structs_sym_tab = structs_proc(tree, module, bpf_chunks)
+    map_sym_tab = maps_proc(tree, module, bpf_chunks, structs_sym_tab)
+    func_proc(tree, module, bpf_chunks, map_sym_tab, structs_sym_tab)
+
+    globals_list_creation(tree, module)
+    return structs_sym_tab, map_sym_tab
 
 
-def compile_to_ir(filename: str, output: str):
+def compile_to_ir(filename: str, output: str, loglevel=logging.INFO):
+    logging.basicConfig(
+        level=loglevel, format="%(asctime)s [%(levelname)s] %(name)s: %(message)s"
+    )
     with open(filename) as f:
         source = f.read()
 
@@ -42,53 +104,131 @@ def compile_to_ir(filename: str, output: str):
     module.data_layout = "e-m:e-p:64:64-i64:64-i128:128-n32:64-S128"
     module.triple = "bpf"
 
-    if not hasattr(module, '_debug_compile_unit'):
-        module._file_metadata = module.add_debug_info("DIFile", {       # type: ignore
-            "filename": filename,
-            "directory": os.path.dirname(filename)
-        })
-        
-        module._debug_compile_unit = module.add_debug_info("DICompileUnit", {       # type: ignore
-            "language": 29,  # DW_LANG_C11
-            "file": module._file_metadata,      # type: ignore
-            "producer": "PythonBPF DSL Compiler",
-            "isOptimized": True,
-            "runtimeVersion": 0,
-            "emissionKind": 1,
-            "splitDebugInlining": False,
-            "nameTableKind": 0
-        }, is_distinct=True)
+    if not hasattr(module, "_debug_compile_unit"):
+        debug_generator = DebugInfoGenerator(module)
+        debug_generator.generate_file_metadata(filename, os.path.dirname(filename))
+        debug_generator.generate_debug_cu(
+            DW_LANG_C11,
+            f"PythonBPF {VERSION}",
+            True,  # TODO: This is probably not true
+            # TODO: add a global field here that keeps track of all the globals. Works without it, but I think it might
+            # be required for kprobes.
+            True,
+        )
 
-        module.add_named_metadata("llvm.dbg.cu", module._debug_compile_unit)        # type: ignore
+    structs_sym_tab, maps_sym_tab = processor(source, filename, module)
 
-    processor(source, filename, module)
-
-    wchar_size = module.add_metadata([ir.Constant(ir.IntType(32), 1),
-                                      "wchar_size",
-                                      ir.Constant(ir.IntType(32), 4)])
-    frame_pointer = module.add_metadata([ir.Constant(ir.IntType(32), 7),
-                                         "frame-pointer",
-                                         ir.Constant(ir.IntType(32), 2)])
+    wchar_size = module.add_metadata(
+        [
+            DwarfBehaviorEnum.ERROR_IF_MISMATCH,
+            "wchar_size",
+            ir.Constant(ir.IntType(32), 4),
+        ]
+    )
+    frame_pointer = module.add_metadata(
+        [
+            DwarfBehaviorEnum.OVERRIDE_USE_LARGEST,
+            "frame-pointer",
+            ir.Constant(ir.IntType(32), 2),
+        ]
+    )
     # Add Debug Info Version (3 = DWARF v3, which LLVM expects)
-    debug_info_version = module.add_metadata([ir.Constant(ir.IntType(32), 2),
-                                              "Debug Info Version",
-                                              ir.Constant(ir.IntType(32), 3)])
+    debug_info_version = module.add_metadata(
+        [
+            DwarfBehaviorEnum.WARNING_IF_MISMATCH,
+            "Debug Info Version",
+            ir.Constant(ir.IntType(32), 3),
+        ]
+    )
 
-    # Add explicit DWARF version (4 is common, works with LLVM BPF backend)
-    dwarf_version = module.add_metadata([ir.Constant(ir.IntType(32), 2),
-                                         "Dwarf Version",
-                                         ir.Constant(ir.IntType(32), 4)])
+    # Add explicit DWARF version 5
+    dwarf_version = module.add_metadata(
+        [
+            DwarfBehaviorEnum.OVERRIDE_USE_LARGEST,
+            "Dwarf Version",
+            ir.Constant(ir.IntType(32), 5),
+        ]
+    )
 
     module.add_named_metadata("llvm.module.flags", wchar_size)
     module.add_named_metadata("llvm.module.flags", frame_pointer)
     module.add_named_metadata("llvm.module.flags", debug_info_version)
     module.add_named_metadata("llvm.module.flags", dwarf_version)
 
-    module.add_named_metadata("llvm.ident", ["llvmlite PythonBPF v0.0.1"])
+    module.add_named_metadata("llvm.ident", [f"PythonBPF {VERSION}"])
 
+    module_string: str = finalize_module(str(module))
+
+    logger.info(f"IR written to {output}")
     with open(output, "w") as f:
-        f.write(f"source_filename = \"{filename}\"\n")
-        f.write(str(module))
+        f.write(f'source_filename = "{filename}"\n')
+        f.write(module_string)
         f.write("\n")
 
-    return output
+    return output, structs_sym_tab, maps_sym_tab
+
+
+def _run_llc(ll_file, obj_file):
+    """Compile LLVM IR to BPF object file using llc."""
+
+    logger.info(f"Compiling IR to object: {ll_file} -> {obj_file}")
+    result = subprocess.run(
+        [
+            "llc",
+            "-march=bpf",
+            "-filetype=obj",
+            "-O2",
+            str(ll_file),
+            "-o",
+            str(obj_file),
+        ],
+        check=True,
+        capture_output=True,
+        text=True,
+    )
+
+    if result.returncode == 0:
+        logger.info(f"Object file written to {obj_file}")
+        return True
+    else:
+        logger.error(f"llc compilation failed: {result.stderr}")
+        return False
+
+
+def compile(loglevel=logging.WARNING) -> bool:
+    # Look one level up the stack to the caller of this function
+    caller_frame = inspect.stack()[1]
+    caller_file = Path(caller_frame.filename).resolve()
+
+    ll_file = Path("/tmp") / caller_file.with_suffix(".ll").name
+    o_file = caller_file.with_suffix(".o")
+
+    _, structs_sym_tab, maps_sym_tab = compile_to_ir(
+        str(caller_file), str(ll_file), loglevel=loglevel
+    )
+
+    if not _run_llc(ll_file, o_file):
+        logger.error("Compilation to object file failed.")
+        return False
+
+    logger.info(f"Object written to {o_file}")
+    return True
+
+
+def BPF(loglevel=logging.WARNING) -> BpfObject:
+    caller_frame = inspect.stack()[1]
+    src = inspect.getsource(caller_frame.frame)
+    with (
+        tempfile.NamedTemporaryFile(mode="w+", delete=True, suffix=".py") as f,
+        tempfile.NamedTemporaryFile(mode="w+", delete=True, suffix=".ll") as inter,
+        tempfile.NamedTemporaryFile(mode="w+", delete=False, suffix=".o") as obj_file,
+    ):
+        f.write(src)
+        f.flush()
+        source = f.name
+        _, structs_sym_tab, maps_sym_tab = compile_to_ir(
+            source, str(inter.name), loglevel=loglevel
+        )
+        _run_llc(str(inter.name), str(obj_file.name))
+
+        return BpfObject(str(obj_file.name), structs=structs_sym_tab)

pythonbpf/debuginfo/__init__.py

@@ -0,0 +1,5 @@
+from .dwarf_constants import *  # noqa: F403
+from .dtypes import *  # noqa: F403
+from .debug_info_generator import DebugInfoGenerator
+
+__all__ = ["DebugInfoGenerator"]

pythonbpf/debuginfo/debug_info_generator.py

@@ -0,0 +1,266 @@
+"""
+Debug information generation module for Python-BPF
+Provides utilities for generating DWARF/BTF debug information
+"""
+
+from . import dwarf_constants as dc
+from typing import Any, List
+
+
+class DebugInfoGenerator:
+    def __init__(self, module):
+        self.module = module
+        self._type_cache = {}  # Cache for common debug types
+
+    def generate_file_metadata(self, filename, dirname):
+        self.module._file_metadata = self.module.add_debug_info(
+            "DIFile",
+            {  # type: ignore
+                "filename": filename,
+                "directory": dirname,
+            },
+        )
+
+    def generate_debug_cu(
+        self, language, producer: str, is_optimized: bool, is_distinct: bool
+    ):
+        self.module._debug_compile_unit = self.module.add_debug_info(
+            "DICompileUnit",
+            {  # type: ignore
+                "language": language,
+                "file": self.module._file_metadata,  # type: ignore
+                "producer": producer,
+                "isOptimized": is_optimized,
+                "runtimeVersion": 0,
+                "emissionKind": 1,
+                "splitDebugInlining": False,
+                "nameTableKind": 0,
+            },
+            is_distinct=is_distinct,
+        )
+        self.module.add_named_metadata("llvm.dbg.cu", self.module._debug_compile_unit)  # type: ignore
+
+    def get_basic_type(self, name: str, size: int, encoding: int) -> Any:
+        """Get or create a basic type with caching"""
+        key = (name, size, encoding)
+        if key not in self._type_cache:
+            self._type_cache[key] = self.module.add_debug_info(
+                "DIBasicType", {"name": name, "size": size, "encoding": encoding}
+            )
+        return self._type_cache[key]
+
+    def get_int32_type(self) -> Any:
+        """Get debug info for signed 32-bit integer"""
+        return self.get_basic_type("int", 32, dc.DW_ATE_signed)
+
+    def get_uint32_type(self) -> Any:
+        """Get debug info for unsigned 32-bit integer"""
+        return self.get_basic_type("unsigned int", 32, dc.DW_ATE_unsigned)
+
+    def get_uint64_type(self) -> Any:
+        """Get debug info for unsigned 64-bit integer"""
+        return self.get_basic_type("unsigned long long", 64, dc.DW_ATE_unsigned)
+
+    def create_pointer_type(self, base_type: Any, size: int = 64) -> Any:
+        """Create a pointer type to the given base type"""
+        return self.module.add_debug_info(
+            "DIDerivedType",
+            {"tag": dc.DW_TAG_pointer_type, "baseType": base_type, "size": size},
+        )
+
+    def create_array_type(self, base_type: Any, count: int) -> Any:
+        """Create an array type of the given base type with specified count"""
+        subrange = self.module.add_debug_info("DISubrange", {"count": count})
+        return self.module.add_debug_info(
+            "DICompositeType",
+            {
+                "tag": dc.DW_TAG_array_type,
+                "baseType": base_type,
+                "size": self._compute_array_size(base_type, count),
+                "elements": [subrange],
+            },
+        )
+
+    def create_array_type_vmlinux(self, type_info: Any, count: int) -> Any:
+        """Create an array type of the given base type with specified count"""
+        base_type, type_sizing = type_info
+        subrange = self.module.add_debug_info("DISubrange", {"count": count})
+        return self.module.add_debug_info(
+            "DICompositeType",
+            {
+                "tag": dc.DW_TAG_array_type,
+                "baseType": base_type,
+                "size": type_sizing,
+                "elements": [subrange],
+            },
+        )
+
+    @staticmethod
+    def _compute_array_size(base_type: Any, count: int) -> int:
+        # Extract size from base_type if possible
+        # For simplicity, assuming base_type has a size attribute
+        return getattr(base_type, "size", 32) * count
+
+    def create_struct_member(self, name: str, base_type: Any, offset: int) -> Any:
+        """Create a struct member with the given name, type, and offset"""
+        return self.module.add_debug_info(
+            "DIDerivedType",
+            {
+                "tag": dc.DW_TAG_member,
+                "name": name,
+                "file": self.module._file_metadata,
+                "baseType": base_type,
+                "size": getattr(base_type, "size", 64),
+                "offset": offset,
+            },
+        )
+
+    def create_struct_member_vmlinux(
+        self, name: str, base_type_with_size: Any, offset: int
+    ) -> Any:
+        """Create a struct member with the given name, type, and offset"""
+        base_type, type_size = base_type_with_size
+        return self.module.add_debug_info(
+            "DIDerivedType",
+            {
+                "tag": dc.DW_TAG_member,
+                "name": name,
+                "file": self.module._file_metadata,
+                "baseType": base_type,
+                "size": type_size,
+                "offset": offset,
+            },
+        )
+
+    def create_struct_type(
+        self, members: List[Any], size: int, is_distinct: bool
+    ) -> Any:
+        """Create a struct type with the given members and size"""
+        return self.module.add_debug_info(
+            "DICompositeType",
+            {
+                "tag": dc.DW_TAG_structure_type,
+                "file": self.module._file_metadata,
+                "size": size,
+                "elements": members,
+            },
+            is_distinct=is_distinct,
+        )
+
+    def create_struct_type_with_name(
+        self, name: str, members: List[Any], size: int, is_distinct: bool
+    ) -> Any:
+        """Create a struct type with the given members and size"""
+        return self.module.add_debug_info(
+            "DICompositeType",
+            {
+                "name": name,
+                "tag": dc.DW_TAG_structure_type,
+                "file": self.module._file_metadata,
+                "size": size,
+                "elements": members,
+            },
+            is_distinct=is_distinct,
+        )
+
+    def create_global_var_debug_info(
+        self, name: str, var_type: Any, is_local: bool = False
+    ) -> Any:
+        """Create debug info for a global variable"""
+        global_var = self.module.add_debug_info(
+            "DIGlobalVariable",
+            {
+                "name": name,
+                "scope": self.module._debug_compile_unit,
+                "file": self.module._file_metadata,
+                "type": var_type,
+                "isLocal": is_local,
+                "isDefinition": True,
+            },
+            is_distinct=True,
+        )
+
+        return self.module.add_debug_info(
+            "DIGlobalVariableExpression",
+            {"var": global_var, "expr": self.module.add_debug_info("DIExpression", {})},
+        )
+
+    def get_int64_type(self):
+        return self.get_basic_type("long", 64, dc.DW_ATE_signed)
+
+    def create_subroutine_type(self, return_type, param_types):
+        """
+        Create a DISubroutineType given return type and list of parameter types.
+        Equivalent to: !DISubroutineType(types: !{ret, args...})
+        """
+        type_array = [return_type]
+        if isinstance(param_types, (list, tuple)):
+            type_array.extend(param_types)
+        else:
+            type_array.append(param_types)
+        return self.module.add_debug_info("DISubroutineType", {"types": type_array})
+
+    def create_local_variable_debug_info(
+        self, name: str, arg: int, var_type: Any
+    ) -> Any:
+        """
+        Create debug info for a local variable (DILocalVariable) without scope.
+        Example:
+        !DILocalVariable(name: "ctx", arg: 1, file: !3, line: 20, type: !7)
+        """
+        return self.module.add_debug_info(
+            "DILocalVariable",
+            {
+                "name": name,
+                "arg": arg,
+                "file": self.module._file_metadata,
+                "type": var_type,
+            },
+        )
+
+    def add_scope_to_local_variable(self, local_variable_debug_info, scope_value):
+        """
+        Add scope information to an existing local variable debug info object.
+        """
+        # TODO: this is a workaround a flaw in the debug info generation. Fix this if possible in the future.
+        # We should not be touching llvmlite's internals like this.
+        if hasattr(local_variable_debug_info, "operands"):
+            # LLVM metadata operands is a tuple, so we need to rebuild it
+            existing_operands = local_variable_debug_info.operands
+
+            # Convert tuple to list, add scope, convert back to tuple
+            operands_list = list(existing_operands)
+            operands_list.append(("scope", scope_value))
+
+            # Reassign the new tuple
+            local_variable_debug_info.operands = tuple(operands_list)
+
+    def create_subprogram(
+        self, name: str, subroutine_type: Any, retained_nodes: List[Any]
+    ) -> Any:
+        """
+        Create a DISubprogram for a function.
+
+        Args:
+            name: Function name
+            subroutine_type: DISubroutineType for the function signature
+            retained_nodes: List of DILocalVariable nodes for function parameters/variables
+
+        Returns:
+            DISubprogram metadata
+        """
+        return self.module.add_debug_info(
+            "DISubprogram",
+            {
+                "name": name,
+                "scope": self.module._file_metadata,
+                "file": self.module._file_metadata,
+                "type": subroutine_type,
+                # TODO: the following flags do not exist at the moment in our dwarf constants file. We need to add them.
+                # "flags": dc.DW_FLAG_Prototyped | dc.DW_FLAG_AllCallsDescribed,
+                # "spFlags": dc.DW_SPFLAG_Definition | dc.DW_SPFLAG_Optimized,
+                "unit": self.module._debug_compile_unit,
+                "retainedNodes": retained_nodes,
+            },
+            is_distinct=True,
+        )

pythonbpf/debuginfo/dtypes.py

@@ -0,0 +1,7 @@
+import llvmlite.ir as ir
+
+
+class DwarfBehaviorEnum:
+    ERROR_IF_MISMATCH = ir.Constant(ir.IntType(32), 1)
+    WARNING_IF_MISMATCH = ir.Constant(ir.IntType(32), 2)
+    OVERRIDE_USE_LARGEST = ir.Constant(ir.IntType(32), 7)

pythonbpf/debuginfo/dwarf_constants.py

@@ -7,7 +7,7 @@ DW_UT_skeleton = 0x04
 DW_UT_split_compile = 0x05
 DW_UT_split_type = 0x06
 DW_UT_lo_user = 0x80
-DW_UT_hi_user = 0xff
+DW_UT_hi_user = 0xFF
 
 DW_TAG_array_type = 0x01
 DW_TAG_class_type = 0x02
@@ -15,10 +15,10 @@ DW_TAG_entry_point = 0x03
 DW_TAG_enumeration_type = 0x04
 DW_TAG_formal_parameter = 0x05
 DW_TAG_imported_declaration = 0x08
-DW_TAG_label = 0x0a
-DW_TAG_lexical_block = 0x0b
-DW_TAG_member = 0x0d
-DW_TAG_pointer_type = 0x0f
+DW_TAG_label = 0x0A
+DW_TAG_lexical_block = 0x0B
+DW_TAG_member = 0x0D
+DW_TAG_pointer_type = 0x0F
 DW_TAG_reference_type = 0x10
 DW_TAG_compile_unit = 0x11
 DW_TAG_string_type = 0x12
@@ -28,12 +28,12 @@ DW_TAG_typedef = 0x16
 DW_TAG_union_type = 0x17
 DW_TAG_unspecified_parameters = 0x18
 DW_TAG_variant = 0x19
-DW_TAG_common_block = 0x1a
-DW_TAG_common_inclusion = 0x1b
-DW_TAG_inheritance = 0x1c
-DW_TAG_inlined_subroutine = 0x1d
-DW_TAG_module = 0x1e
-DW_TAG_ptr_to_member_type = 0x1f
+DW_TAG_common_block = 0x1A
+DW_TAG_common_inclusion = 0x1B
+DW_TAG_inheritance = 0x1C
+DW_TAG_inlined_subroutine = 0x1D
+DW_TAG_module = 0x1E
+DW_TAG_ptr_to_member_type = 0x1F
 DW_TAG_set_type = 0x20
 DW_TAG_subrange_type = 0x21
 DW_TAG_with_stmt = 0x22
@@ -44,12 +44,12 @@ DW_TAG_const_type = 0x26
 DW_TAG_constant = 0x27
 DW_TAG_enumerator = 0x28
 DW_TAG_file_type = 0x29
-DW_TAG_friend = 0x2a
-DW_TAG_namelist = 0x2b
-DW_TAG_namelist_item = 0x2c
-DW_TAG_packed_type = 0x2d
-DW_TAG_subprogram = 0x2e
-DW_TAG_template_type_parameter = 0x2f
+DW_TAG_friend = 0x2A
+DW_TAG_namelist = 0x2B
+DW_TAG_namelist_item = 0x2C
+DW_TAG_packed_type = 0x2D
+DW_TAG_subprogram = 0x2E
+DW_TAG_template_type_parameter = 0x2F
 DW_TAG_template_value_parameter = 0x30
 DW_TAG_thrown_type = 0x31
 DW_TAG_try_block = 0x32
@@ -60,11 +60,11 @@ DW_TAG_dwarf_procedure = 0x36
 DW_TAG_restrict_type = 0x37
 DW_TAG_interface_type = 0x38
 DW_TAG_namespace = 0x39
-DW_TAG_imported_module = 0x3a
-DW_TAG_unspecified_type = 0x3b
-DW_TAG_partial_unit = 0x3c
-DW_TAG_imported_unit = 0x3d
-DW_TAG_condition = 0x3f
+DW_TAG_imported_module = 0x3A
+DW_TAG_unspecified_type = 0x3B
+DW_TAG_partial_unit = 0x3C
+DW_TAG_imported_unit = 0x3D
+DW_TAG_condition = 0x3F
 DW_TAG_shared_type = 0x40
 DW_TAG_type_unit = 0x41
 DW_TAG_rvalue_reference_type = 0x42
@@ -75,8 +75,8 @@ DW_TAG_dynamic_type = 0x46
 DW_TAG_atomic_type = 0x47
 DW_TAG_call_site = 0x48
 DW_TAG_call_site_parameter = 0x49
-DW_TAG_skeleton_unit = 0x4a
-DW_TAG_immutable_type = 0x4b
+DW_TAG_skeleton_unit = 0x4A
+DW_TAG_immutable_type = 0x4B
 DW_TAG_lo_user = 0x4080
 DW_TAG_MIPS_loop = 0x4081
 DW_TAG_format_label = 0x4101
@@ -88,8 +88,8 @@ DW_TAG_GNU_template_template_param = 0x4106
 DW_TAG_GNU_template_parameter_pack = 0x4107
 DW_TAG_GNU_formal_parameter_pack = 0x4108
 DW_TAG_GNU_call_site = 0x4109
-DW_TAG_GNU_call_site_parameter = 0x410a
-DW_TAG_hi_user = 0xffff
+DW_TAG_GNU_call_site_parameter = 0x410A
+DW_TAG_hi_user = 0xFFFF
 
 DW_CHILDREN_no = 0
 DW_CHILDREN_yes = 1
@@ -98,9 +98,9 @@ DW_AT_sibling = 0x01
 DW_AT_location = 0x02
 DW_AT_name = 0x03
 DW_AT_ordering = 0x09
-DW_AT_byte_size = 0x0b
-DW_AT_bit_offset = 0x0c
-DW_AT_bit_size = 0x0d
+DW_AT_byte_size = 0x0B
+DW_AT_bit_offset = 0x0C
+DW_AT_bit_size = 0x0D
 DW_AT_stmt_list = 0x10
 DW_AT_low_pc = 0x11
 DW_AT_high_pc = 0x12
@@ -110,20 +110,20 @@ DW_AT_discr_value = 0x16
 DW_AT_visibility = 0x17
 DW_AT_import = 0x18
 DW_AT_string_length = 0x19
-DW_AT_common_reference = 0x1a
-DW_AT_comp_dir = 0x1b
-DW_AT_const_value = 0x1c
-DW_AT_containing_type = 0x1d
-DW_AT_default_value = 0x1e
+DW_AT_common_reference = 0x1A
+DW_AT_comp_dir = 0x1B
+DW_AT_const_value = 0x1C
+DW_AT_containing_type = 0x1D
+DW_AT_default_value = 0x1E
 DW_AT_inline = 0x20
 DW_AT_is_optional = 0x21
 DW_AT_lower_bound = 0x22
 DW_AT_producer = 0x25
 DW_AT_prototyped = 0x27
-DW_AT_return_addr = 0x2a
-DW_AT_start_scope = 0x2c
-DW_AT_bit_stride = 0x2e
-DW_AT_upper_bound = 0x2f
+DW_AT_return_addr = 0x2A
+DW_AT_start_scope = 0x2C
+DW_AT_bit_stride = 0x2E
+DW_AT_upper_bound = 0x2F
 DW_AT_abstract_origin = 0x31
 DW_AT_accessibility = 0x32
 DW_AT_address_class = 0x33
@@ -133,12 +133,12 @@ DW_AT_calling_convention = 0x36
 DW_AT_count = 0x37
 DW_AT_data_member_location = 0x38
 DW_AT_decl_column = 0x39
-DW_AT_decl_file = 0x3a
-DW_AT_decl_line = 0x3b
-DW_AT_declaration = 0x3c
-DW_AT_discr_list = 0x3d
-DW_AT_encoding = 0x3e
-DW_AT_external = 0x3f
+DW_AT_decl_file = 0x3A
+DW_AT_decl_line = 0x3B
+DW_AT_declaration = 0x3C
+DW_AT_discr_list = 0x3D
+DW_AT_encoding = 0x3E
+DW_AT_external = 0x3F
 DW_AT_frame_base = 0x40
 DW_AT_friend = 0x41
 DW_AT_identifier_case = 0x42
@@ -149,12 +149,12 @@ DW_AT_segment = 0x46
 DW_AT_specification = 0x47
 DW_AT_static_link = 0x48
 DW_AT_type = 0x49
-DW_AT_use_location = 0x4a
-DW_AT_variable_parameter = 0x4b
-DW_AT_virtuality = 0x4c
-DW_AT_vtable_elem_location = 0x4d
-DW_AT_allocated = 0x4e
-DW_AT_associated = 0x4f
+DW_AT_use_location = 0x4A
+DW_AT_variable_parameter = 0x4B
+DW_AT_virtuality = 0x4C
+DW_AT_vtable_elem_location = 0x4D
+DW_AT_allocated = 0x4E
+DW_AT_associated = 0x4F
 DW_AT_data_location = 0x50
 DW_AT_byte_stride = 0x51
 DW_AT_entry_pc = 0x52
@@ -165,12 +165,12 @@ DW_AT_trampoline = 0x56
 DW_AT_call_column = 0x57
 DW_AT_call_file = 0x58
 DW_AT_call_line = 0x59
-DW_AT_description = 0x5a
-DW_AT_binary_scale = 0x5b
-DW_AT_decimal_scale = 0x5c
-DW_AT_small = 0x5d
-DW_AT_decimal_sign = 0x5e
-DW_AT_digit_count = 0x5f
+DW_AT_description = 0x5A
+DW_AT_binary_scale = 0x5B
+DW_AT_decimal_scale = 0x5C
+DW_AT_small = 0x5D
+DW_AT_decimal_sign = 0x5E
+DW_AT_digit_count = 0x5F
 DW_AT_picture_string = 0x60
 DW_AT_mutable = 0x61
 DW_AT_threads_scaled = 0x62
@@ -181,12 +181,12 @@ DW_AT_elemental = 0x66
 DW_AT_pure = 0x67
 DW_AT_recursive = 0x68
 DW_AT_signature = 0x69
-DW_AT_main_subprogram = 0x6a
-DW_AT_data_bit_offset = 0x6b
-DW_AT_const_expr = 0x6c
-DW_AT_enum_class = 0x6d
-DW_AT_linkage_name = 0x6e
-DW_AT_string_length_bit_size = 0x6f
+DW_AT_main_subprogram = 0x6A
+DW_AT_data_bit_offset = 0x6B
+DW_AT_const_expr = 0x6C
+DW_AT_enum_class = 0x6D
+DW_AT_linkage_name = 0x6E
+DW_AT_string_length_bit_size = 0x6F
 DW_AT_string_length_byte_size = 0x70
 DW_AT_rank = 0x71
 DW_AT_str_offsets_base = 0x72
@@ -196,12 +196,12 @@ DW_AT_dwo_name = 0x76
 DW_AT_reference = 0x77
 DW_AT_rvalue_reference = 0x78
 DW_AT_macros = 0x79
-DW_AT_call_all_calls = 0x7a
-DW_AT_call_all_source_calls = 0x7b
-DW_AT_call_all_tail_calls = 0x7c
-DW_AT_call_return_pc = 0x7d
-DW_AT_call_value = 0x7e
-DW_AT_call_origin = 0x7f
+DW_AT_call_all_calls = 0x7A
+DW_AT_call_all_source_calls = 0x7B
+DW_AT_call_all_tail_calls = 0x7C
+DW_AT_call_return_pc = 0x7D
+DW_AT_call_value = 0x7E
+DW_AT_call_origin = 0x7F
 DW_AT_call_parameter = 0x80
 DW_AT_call_pc = 0x81
 DW_AT_call_tail_call = 0x82
@@ -212,9 +212,9 @@ DW_AT_call_data_value = 0x86
 DW_AT_noreturn = 0x87
 DW_AT_alignment = 0x88
 DW_AT_export_symbols = 0x89
-DW_AT_deleted = 0x8a
-DW_AT_defaulted = 0x8b
-DW_AT_loclists_base = 0x8c
+DW_AT_deleted = 0x8A
+DW_AT_defaulted = 0x8B
+DW_AT_loclists_base = 0x8C
 DW_AT_lo_user = 0x2000
 DW_AT_MIPS_fde = 0x2001
 DW_AT_MIPS_loop_begin = 0x2002
@@ -225,12 +225,12 @@ DW_AT_MIPS_software_pipeline_depth = 0x2006
 DW_AT_MIPS_linkage_name = 0x2007
 DW_AT_MIPS_stride = 0x2008
 DW_AT_MIPS_abstract_name = 0x2009
-DW_AT_MIPS_clone_origin = 0x200a
-DW_AT_MIPS_has_inlines = 0x200b
-DW_AT_MIPS_stride_byte = 0x200c
-DW_AT_MIPS_stride_elem = 0x200d
-DW_AT_MIPS_ptr_dopetype = 0x200e
-DW_AT_MIPS_allocatable_dopetype = 0x200f
+DW_AT_MIPS_clone_origin = 0x200A
+DW_AT_MIPS_has_inlines = 0x200B
+DW_AT_MIPS_stride_byte = 0x200C
+DW_AT_MIPS_stride_elem = 0x200D
+DW_AT_MIPS_ptr_dopetype = 0x200E
+DW_AT_MIPS_allocatable_dopetype = 0x200F
 DW_AT_MIPS_assumed_shape_dopetype = 0x2010
 DW_AT_MIPS_assumed_size = 0x2011
 DW_AT_sf_names = 0x2101
@@ -242,12 +242,12 @@ DW_AT_body_end = 0x2106
 DW_AT_GNU_vector = 0x2107
 DW_AT_GNU_guarded_by = 0x2108
 DW_AT_GNU_pt_guarded_by = 0x2109
-DW_AT_GNU_guarded = 0x210a
-DW_AT_GNU_pt_guarded = 0x210b
-DW_AT_GNU_locks_excluded = 0x210c
-DW_AT_GNU_exclusive_locks_required = 0x210d
-DW_AT_GNU_shared_locks_required = 0x210e
-DW_AT_GNU_odr_signature = 0x210f
+DW_AT_GNU_guarded = 0x210A
+DW_AT_GNU_pt_guarded = 0x210B
+DW_AT_GNU_locks_excluded = 0x210C
+DW_AT_GNU_exclusive_locks_required = 0x210D
+DW_AT_GNU_shared_locks_required = 0x210E
+DW_AT_GNU_odr_signature = 0x210F
 DW_AT_GNU_template_name = 0x2110
 DW_AT_GNU_call_site_value = 0x2111
 DW_AT_GNU_call_site_data_value = 0x2112
@@ -260,7 +260,7 @@ DW_AT_GNU_all_source_call_sites = 0x2118
 DW_AT_GNU_locviews = 0x2137
 DW_AT_GNU_entry_view = 0x2138
 DW_AT_GNU_macros = 0x2119
-DW_AT_GNU_deleted = 0x211a
+DW_AT_GNU_deleted = 0x211A
 DW_AT_GNU_dwo_name = 0x2130
 DW_AT_GNU_dwo_id = 0x2131
 DW_AT_GNU_ranges_base = 0x2132
@@ -270,7 +270,7 @@ DW_AT_GNU_pubtypes = 0x2135
 DW_AT_GNU_numerator = 0x2303
 DW_AT_GNU_denominator = 0x2304
 DW_AT_GNU_bias = 0x2305
-DW_AT_hi_user = 0x3fff
+DW_AT_hi_user = 0x3FFF
 
 DW_FORM_addr = 0x01
 DW_FORM_block2 = 0x03
@@ -280,12 +280,12 @@ DW_FORM_data4 = 0x06
 DW_FORM_data8 = 0x07
 DW_FORM_string = 0x08
 DW_FORM_block = 0x09
-DW_FORM_block1 = 0x0a
-DW_FORM_data1 = 0x0b
-DW_FORM_flag = 0x0c
-DW_FORM_sdata = 0x0d
-DW_FORM_strp = 0x0e
-DW_FORM_udata = 0x0f
+DW_FORM_block1 = 0x0A
+DW_FORM_data1 = 0x0B
+DW_FORM_flag = 0x0C
+DW_FORM_sdata = 0x0D
+DW_FORM_strp = 0x0E
+DW_FORM_udata = 0x0F
 DW_FORM_ref_addr = 0x10
 DW_FORM_ref1 = 0x11
 DW_FORM_ref2 = 0x12
@@ -296,12 +296,12 @@ DW_FORM_indirect = 0x16
 DW_FORM_sec_offset = 0x17
 DW_FORM_exprloc = 0x18
 DW_FORM_flag_present = 0x19
-DW_FORM_strx = 0x1a
-DW_FORM_addrx = 0x1b
-DW_FORM_ref_sup4 = 0x1c
-DW_FORM_strp_sup = 0x1d
-DW_FORM_data16 = 0x1e
-DW_FORM_line_strp = 0x1f
+DW_FORM_strx = 0x1A
+DW_FORM_addrx = 0x1B
+DW_FORM_ref_sup4 = 0x1C
+DW_FORM_strp_sup = 0x1D
+DW_FORM_data16 = 0x1E
+DW_FORM_line_strp = 0x1F
 DW_FORM_ref_sig8 = 0x20
 DW_FORM_implicit_const = 0x21
 DW_FORM_loclistx = 0x22
@@ -312,24 +312,24 @@ DW_FORM_strx2 = 0x26
 DW_FORM_strx3 = 0x27
 DW_FORM_strx4 = 0x28
 DW_FORM_addrx1 = 0x29
-DW_FORM_addrx2 = 0x2a
-DW_FORM_addrx3 = 0x2b
-DW_FORM_addrx4 = 0x2c
-DW_FORM_GNU_addr_index = 0x1f01
-DW_FORM_GNU_str_index = 0x1f02
-DW_FORM_GNU_ref_alt = 0x1f20
-DW_FORM_GNU_strp_alt = 0x1f21
+DW_FORM_addrx2 = 0x2A
+DW_FORM_addrx3 = 0x2B
+DW_FORM_addrx4 = 0x2C
+DW_FORM_GNU_addr_index = 0x1F01
+DW_FORM_GNU_str_index = 0x1F02
+DW_FORM_GNU_ref_alt = 0x1F20
+DW_FORM_GNU_strp_alt = 0x1F21
 
 DW_OP_addr = 0x03
 DW_OP_deref = 0x06
 DW_OP_const1u = 0x08
 DW_OP_const1s = 0x09
-DW_OP_const2u = 0x0a
-DW_OP_const2s = 0x0b
-DW_OP_const4u = 0x0c
-DW_OP_const4s = 0x0d
-DW_OP_const8u = 0x0e
-DW_OP_const8s = 0x0f
+DW_OP_const2u = 0x0A
+DW_OP_const2s = 0x0B
+DW_OP_const4u = 0x0C
+DW_OP_const4s = 0x0D
+DW_OP_const8u = 0x0E
+DW_OP_const8s = 0x0F
 DW_OP_constu = 0x10
 DW_OP_consts = 0x11
 DW_OP_dup = 0x12
@@ -340,12 +340,12 @@ DW_OP_swap = 0x16
 DW_OP_rot = 0x17
 DW_OP_xderef = 0x18
 DW_OP_abs = 0x19
-DW_OP_and = 0x1a
-DW_OP_div = 0x1b
-DW_OP_minus = 0x1c
-DW_OP_mod = 0x1d
-DW_OP_mul = 0x1e
-DW_OP_neg = 0x1f
+DW_OP_and = 0x1A
+DW_OP_div = 0x1B
+DW_OP_minus = 0x1C
+DW_OP_mod = 0x1D
+DW_OP_mul = 0x1E
+DW_OP_neg = 0x1F
 DW_OP_not = 0x20
 DW_OP_or = 0x21
 DW_OP_plus = 0x22
@@ -356,12 +356,12 @@ DW_OP_shra = 0x26
 DW_OP_xor = 0x27
 DW_OP_bra = 0x28
 DW_OP_eq = 0x29
-DW_OP_ge = 0x2a
-DW_OP_gt = 0x2b
-DW_OP_le = 0x2c
-DW_OP_lt = 0x2d
-DW_OP_ne = 0x2e
-DW_OP_skip = 0x2f
+DW_OP_ge = 0x2A
+DW_OP_gt = 0x2B
+DW_OP_le = 0x2C
+DW_OP_lt = 0x2D
+DW_OP_ne = 0x2E
+DW_OP_skip = 0x2F
 DW_OP_lit0 = 0x30
 DW_OP_lit1 = 0x31
 DW_OP_lit2 = 0x32
@@ -372,12 +372,12 @@ DW_OP_lit6 = 0x36
 DW_OP_lit7 = 0x37
 DW_OP_lit8 = 0x38
 DW_OP_lit9 = 0x39
-DW_OP_lit10 = 0x3a
-DW_OP_lit11 = 0x3b
-DW_OP_lit12 = 0x3c
-DW_OP_lit13 = 0x3d
-DW_OP_lit14 = 0x3e
-DW_OP_lit15 = 0x3f
+DW_OP_lit10 = 0x3A
+DW_OP_lit11 = 0x3B
+DW_OP_lit12 = 0x3C
+DW_OP_lit13 = 0x3D
+DW_OP_lit14 = 0x3E
+DW_OP_lit15 = 0x3F
 DW_OP_lit16 = 0x40
 DW_OP_lit17 = 0x41
 DW_OP_lit18 = 0x42
@@ -388,12 +388,12 @@ DW_OP_lit22 = 0x46
 DW_OP_lit23 = 0x47
 DW_OP_lit24 = 0x48
 DW_OP_lit25 = 0x49
-DW_OP_lit26 = 0x4a
-DW_OP_lit27 = 0x4b
-DW_OP_lit28 = 0x4c
-DW_OP_lit29 = 0x4d
-DW_OP_lit30 = 0x4e
-DW_OP_lit31 = 0x4f
+DW_OP_lit26 = 0x4A
+DW_OP_lit27 = 0x4B
+DW_OP_lit28 = 0x4C
+DW_OP_lit29 = 0x4D
+DW_OP_lit30 = 0x4E
+DW_OP_lit31 = 0x4F
 DW_OP_reg0 = 0x50
 DW_OP_reg1 = 0x51
 DW_OP_reg2 = 0x52
@@ -404,12 +404,12 @@ DW_OP_reg6 = 0x56
 DW_OP_reg7 = 0x57
 DW_OP_reg8 = 0x58
 DW_OP_reg9 = 0x59
-DW_OP_reg10 = 0x5a
-DW_OP_reg11 = 0x5b
-DW_OP_reg12 = 0x5c
-DW_OP_reg13 = 0x5d
-DW_OP_reg14 = 0x5e
-DW_OP_reg15 = 0x5f
+DW_OP_reg10 = 0x5A
+DW_OP_reg11 = 0x5B
+DW_OP_reg12 = 0x5C
+DW_OP_reg13 = 0x5D
+DW_OP_reg14 = 0x5E
+DW_OP_reg15 = 0x5F
 DW_OP_reg16 = 0x60
 DW_OP_reg17 = 0x61
 DW_OP_reg18 = 0x62
@@ -420,12 +420,12 @@ DW_OP_reg22 = 0x66
 DW_OP_reg23 = 0x67
 DW_OP_reg24 = 0x68
 DW_OP_reg25 = 0x69
-DW_OP_reg26 = 0x6a
-DW_OP_reg27 = 0x6b
-DW_OP_reg28 = 0x6c
-DW_OP_reg29 = 0x6d
-DW_OP_reg30 = 0x6e
-DW_OP_reg31 = 0x6f
+DW_OP_reg26 = 0x6A
+DW_OP_reg27 = 0x6B
+DW_OP_reg28 = 0x6C
+DW_OP_reg29 = 0x6D
+DW_OP_reg30 = 0x6E
+DW_OP_reg31 = 0x6F
 DW_OP_breg0 = 0x70
 DW_OP_breg1 = 0x71
 DW_OP_breg2 = 0x72
@@ -436,12 +436,12 @@ DW_OP_breg6 = 0x76
 DW_OP_breg7 = 0x77
 DW_OP_breg8 = 0x78
 DW_OP_breg9 = 0x79
-DW_OP_breg10 = 0x7a
-DW_OP_breg11 = 0x7b
-DW_OP_breg12 = 0x7c
-DW_OP_breg13 = 0x7d
-DW_OP_breg14 = 0x7e
-DW_OP_breg15 = 0x7f
+DW_OP_breg10 = 0x7A
+DW_OP_breg11 = 0x7B
+DW_OP_breg12 = 0x7C
+DW_OP_breg13 = 0x7D
+DW_OP_breg14 = 0x7E
+DW_OP_breg15 = 0x7F
 DW_OP_breg16 = 0x80
 DW_OP_breg17 = 0x81
 DW_OP_breg18 = 0x82
@@ -452,12 +452,12 @@ DW_OP_breg22 = 0x86
 DW_OP_breg23 = 0x87
 DW_OP_breg24 = 0x88
 DW_OP_breg25 = 0x89
-DW_OP_breg26 = 0x8a
-DW_OP_breg27 = 0x8b
-DW_OP_breg28 = 0x8c
-DW_OP_breg29 = 0x8d
-DW_OP_breg30 = 0x8e
-DW_OP_breg31 = 0x8f
+DW_OP_breg26 = 0x8A
+DW_OP_breg27 = 0x8B
+DW_OP_breg28 = 0x8C
+DW_OP_breg29 = 0x8D
+DW_OP_breg30 = 0x8E
+DW_OP_breg31 = 0x8F
 DW_OP_regx = 0x90
 DW_OP_fbreg = 0x91
 DW_OP_bregx = 0x92
@@ -468,38 +468,38 @@ DW_OP_nop = 0x96
 DW_OP_push_object_address = 0x97
 DW_OP_call2 = 0x98
 DW_OP_call4 = 0x99
-DW_OP_call_ref = 0x9a
-DW_OP_form_tls_address = 0x9b
-DW_OP_call_frame_cfa = 0x9c
-DW_OP_bit_piece = 0x9d
-DW_OP_implicit_value = 0x9e
-DW_OP_stack_value = 0x9f
-DW_OP_implicit_pointer = 0xa0
-DW_OP_addrx = 0xa1
-DW_OP_constx = 0xa2
-DW_OP_entry_value = 0xa3
-DW_OP_const_type = 0xa4
-DW_OP_regval_type = 0xa5
-DW_OP_deref_type = 0xa6
-DW_OP_xderef_type = 0xa7
-DW_OP_convert = 0xa8
-DW_OP_reinterpret = 0xa9
-DW_OP_GNU_push_tls_address = 0xe0
-DW_OP_GNU_uninit = 0xf0
-DW_OP_GNU_encoded_addr = 0xf1
-DW_OP_GNU_implicit_pointer = 0xf2
-DW_OP_GNU_entry_value = 0xf3
-DW_OP_GNU_const_type = 0xf4
-DW_OP_GNU_regval_type = 0xf5
-DW_OP_GNU_deref_type = 0xf6
-DW_OP_GNU_convert = 0xf7
-DW_OP_GNU_reinterpret = 0xf9
-DW_OP_GNU_parameter_ref = 0xfa
-DW_OP_GNU_addr_index = 0xfb
-DW_OP_GNU_const_index = 0xfc
-DW_OP_GNU_variable_value = 0xfd
-DW_OP_lo_user = 0xe0
-DW_OP_hi_user = 0xff
+DW_OP_call_ref = 0x9A
+DW_OP_form_tls_address = 0x9B
+DW_OP_call_frame_cfa = 0x9C
+DW_OP_bit_piece = 0x9D
+DW_OP_implicit_value = 0x9E
+DW_OP_stack_value = 0x9F
+DW_OP_implicit_pointer = 0xA0
+DW_OP_addrx = 0xA1
+DW_OP_constx = 0xA2
+DW_OP_entry_value = 0xA3
+DW_OP_const_type = 0xA4
+DW_OP_regval_type = 0xA5
+DW_OP_deref_type = 0xA6
+DW_OP_xderef_type = 0xA7
+DW_OP_convert = 0xA8
+DW_OP_reinterpret = 0xA9
+DW_OP_GNU_push_tls_address = 0xE0
+DW_OP_GNU_uninit = 0xF0
+DW_OP_GNU_encoded_addr = 0xF1
+DW_OP_GNU_implicit_pointer = 0xF2
+DW_OP_GNU_entry_value = 0xF3
+DW_OP_GNU_const_type = 0xF4
+DW_OP_GNU_regval_type = 0xF5
+DW_OP_GNU_deref_type = 0xF6
+DW_OP_GNU_convert = 0xF7
+DW_OP_GNU_reinterpret = 0xF9
+DW_OP_GNU_parameter_ref = 0xFA
+DW_OP_GNU_addr_index = 0xFB
+DW_OP_GNU_const_index = 0xFC
+DW_OP_GNU_variable_value = 0xFD
+DW_OP_lo_user = 0xE0
+DW_OP_hi_user = 0xFF
 
 DW_ATE_void = 0x0
 DW_ATE_address = 0x1
@@ -511,17 +511,17 @@ DW_ATE_signed_char = 0x6
 DW_ATE_unsigned = 0x7
 DW_ATE_unsigned_char = 0x8
 DW_ATE_imaginary_float = 0x9
-DW_ATE_packed_decimal = 0xa
-DW_ATE_numeric_string = 0xb
-DW_ATE_edited = 0xc
-DW_ATE_signed_fixed = 0xd
-DW_ATE_unsigned_fixed = 0xe
-DW_ATE_decimal_float = 0xf
+DW_ATE_packed_decimal = 0xA
+DW_ATE_numeric_string = 0xB
+DW_ATE_edited = 0xC
+DW_ATE_signed_fixed = 0xD
+DW_ATE_unsigned_fixed = 0xE
+DW_ATE_decimal_float = 0xF
 DW_ATE_UTF = 0x10
 DW_ATE_UCS = 0x11
 DW_ATE_ASCII = 0x12
 DW_ATE_lo_user = 0x80
-DW_ATE_hi_user = 0xff
+DW_ATE_hi_user = 0xFF
 
 DW_DS_unsigned = 1
 DW_DS_leading_overpunch = 2
@@ -533,7 +533,7 @@ DW_END_default = 0
 DW_END_big = 1
 DW_END_little = 2
 DW_END_lo_user = 0x40
-DW_END_hi_user = 0xff
+DW_END_hi_user = 0xFF
 
 DW_ACCESS_public = 1
 DW_ACCESS_protected = 2
@@ -556,12 +556,12 @@ DW_LANG_Cobol85 = 0x0006
 DW_LANG_Fortran77 = 0x0007
 DW_LANG_Fortran90 = 0x0008
 DW_LANG_Pascal83 = 0x0009
-DW_LANG_Modula2 = 0x000a
-DW_LANG_Java = 0x000b
-DW_LANG_C99 = 0x000c
-DW_LANG_Ada95 = 0x000d
-DW_LANG_Fortran95 = 0x000e
-DW_LANG_PLI = 0x000f
+DW_LANG_Modula2 = 0x000A
+DW_LANG_Java = 0x000B
+DW_LANG_C99 = 0x000C
+DW_LANG_Ada95 = 0x000D
+DW_LANG_Fortran95 = 0x000E
+DW_LANG_PLI = 0x000F
 DW_LANG_ObjC = 0x0010
 DW_LANG_ObjC_plus_plus = 0x0011
 DW_LANG_UPC = 0x0012
@@ -572,12 +572,12 @@ DW_LANG_Go = 0x0016
 DW_LANG_Modula3 = 0x0017
 DW_LANG_Haskell = 0x0018
 DW_LANG_C_plus_plus_03 = 0x0019
-DW_LANG_C_plus_plus_11 = 0x001a
-DW_LANG_OCaml = 0x001b
-DW_LANG_Rust = 0x001c
-DW_LANG_C11 = 0x001d
-DW_LANG_Swift = 0x001e
-DW_LANG_Julia = 0x001f
+DW_LANG_C_plus_plus_11 = 0x001A
+DW_LANG_OCaml = 0x001B
+DW_LANG_Rust = 0x001C
+DW_LANG_C11 = 0x001D
+DW_LANG_Swift = 0x001E
+DW_LANG_Julia = 0x001F
 DW_LANG_Dylan = 0x0020
 DW_LANG_C_plus_plus_14 = 0x0021
 DW_LANG_Fortran03 = 0x0022
@@ -586,7 +586,7 @@ DW_LANG_RenderScript = 0x0024
 DW_LANG_BLISS = 0x0025
 DW_LANG_lo_user = 0x8000
 DW_LANG_Mips_Assembler = 0x8001
-DW_LANG_hi_user = 0xffff
+DW_LANG_hi_user = 0xFFFF
 
 DW_ID_case_sensitive = 0
 DW_ID_up_case = 1
@@ -599,7 +599,7 @@ DW_CC_nocall = 0x3
 DW_CC_pass_by_reference = 0x4
 DW_CC_pass_by_value = 0x5
 DW_CC_lo_user = 0x40
-DW_CC_hi_user = 0xff
+DW_CC_hi_user = 0xFF
 
 DW_INL_not_inlined = 0
 DW_INL_inlined = 1
@@ -622,7 +622,7 @@ DW_LNCT_timestamp = 0x3
 DW_LNCT_size = 0x4
 DW_LNCT_MD5 = 0x5
 DW_LNCT_lo_user = 0x2000
-DW_LNCT_hi_user = 0x3fff
+DW_LNCT_hi_user = 0x3FFF
 
 DW_LNS_copy = 1
 DW_LNS_advance_pc = 2
@@ -659,11 +659,11 @@ DW_MACRO_undef_strp = 0x06
 DW_MACRO_import = 0x07
 DW_MACRO_define_sup = 0x08
 DW_MACRO_undef_sup = 0x09
-DW_MACRO_import_sup = 0x0a
-DW_MACRO_define_strx = 0x0b
-DW_MACRO_undef_strx = 0x0c
-DW_MACRO_lo_user = 0xe0
-DW_MACRO_hi_user = 0xff
+DW_MACRO_import_sup = 0x0A
+DW_MACRO_define_strx = 0x0B
+DW_MACRO_undef_strx = 0x0C
+DW_MACRO_lo_user = 0xE0
+DW_MACRO_hi_user = 0xFF
 
 DW_RLE_end_of_list = 0x0
 DW_RLE_base_addressx = 0x1
@@ -691,7 +691,7 @@ DW_LLE_GNU_start_length_entry = 0x3
 
 DW_CFA_advance_loc = 0x40
 DW_CFA_offset = 0x80
-DW_CFA_restore = 0xc0
+DW_CFA_restore = 0xC0
 DW_CFA_extended = 0
 DW_CFA_nop = 0x00
 DW_CFA_set_loc = 0x01
@@ -703,12 +703,12 @@ DW_CFA_restore_extended = 0x06
 DW_CFA_undefined = 0x07
 DW_CFA_same_value = 0x08
 DW_CFA_register = 0x09
-DW_CFA_remember_state = 0x0a
-DW_CFA_restore_state = 0x0b
-DW_CFA_def_cfa = 0x0c
-DW_CFA_def_cfa_register = 0x0d
-DW_CFA_def_cfa_offset = 0x0e
-DW_CFA_def_cfa_expression = 0x0f
+DW_CFA_remember_state = 0x0A
+DW_CFA_restore_state = 0x0B
+DW_CFA_def_cfa = 0x0C
+DW_CFA_def_cfa_register = 0x0D
+DW_CFA_def_cfa_offset = 0x0E
+DW_CFA_def_cfa_expression = 0x0F
 DW_CFA_expression = 0x10
 DW_CFA_offset_extended_sf = 0x11
 DW_CFA_def_cfa_sf = 0x12
@@ -716,26 +716,26 @@ DW_CFA_def_cfa_offset_sf = 0x13
 DW_CFA_val_offset = 0x14
 DW_CFA_val_offset_sf = 0x15
 DW_CFA_val_expression = 0x16
-DW_CFA_low_user = 0x1c
-DW_CFA_MIPS_advance_loc8 = 0x1d
-DW_CFA_GNU_window_save = 0x2d
-DW_CFA_GNU_args_size = 0x2e
-DW_CFA_GNU_negative_offset_extended = 0x2f
-DW_CFA_high_user = 0x3f
+DW_CFA_low_user = 0x1C
+DW_CFA_MIPS_advance_loc8 = 0x1D
+DW_CFA_GNU_window_save = 0x2D
+DW_CFA_GNU_args_size = 0x2E
+DW_CFA_GNU_negative_offset_extended = 0x2F
+DW_CFA_high_user = 0x3F
 
-DW_CIE_ID_32 = 0xffffffff
-DW_CIE_ID_64 = 0xffffffffffffffff
+DW_CIE_ID_32 = 0xFFFFFFFF
+DW_CIE_ID_64 = 0xFFFFFFFFFFFFFFFF
 
 DW_EH_PE_absptr = 0x00
-DW_EH_PE_omit = 0xff
+DW_EH_PE_omit = 0xFF
 DW_EH_PE_uleb128 = 0x01
 DW_EH_PE_udata2 = 0x02
 DW_EH_PE_udata4 = 0x03
 DW_EH_PE_udata8 = 0x04
 DW_EH_PE_sleb128 = 0x09
-DW_EH_PE_sdata2 = 0x0a
-DW_EH_PE_sdata4 = 0x0b
-DW_EH_PE_sdata8 = 0x0c
+DW_EH_PE_sdata2 = 0x0A
+DW_EH_PE_sdata4 = 0x0B
+DW_EH_PE_sdata8 = 0x0C
 DW_EH_PE_signed = 0x08
 DW_EH_PE_pcrel = 0x10
 DW_EH_PE_textrel = 0x20

pythonbpf/decorators.py

@@ -9,17 +9,27 @@ def bpfglobal(func):
     func._is_bpfglobal = True
     return func
 
+
 def map(func):
     """Decorator to mark a function as a BPF map."""
     func._is_map = True
     return func
 
+
+def struct(cls):
+    """Decorator to mark a class as a BPF struct."""
+    cls._is_struct = True
+    return cls
+
+
 def section(name: str):
     def wrapper(fn):
         fn._section = name
         return fn
+
     return wrapper
 
+
 # from types import SimpleNamespace
 
 # syscalls = SimpleNamespace(

pythonbpf/expr/__init__.py

@@ -0,0 +1,16 @@
+from .expr_pass import eval_expr, handle_expr, get_operand_value
+from .type_normalization import convert_to_bool, get_base_type_and_depth
+from .ir_ops import deref_to_depth
+from .call_registry import CallHandlerRegistry
+from .vmlinux_registry import VmlinuxHandlerRegistry
+
+__all__ = [
+    "eval_expr",
+    "handle_expr",
+    "convert_to_bool",
+    "get_base_type_and_depth",
+    "deref_to_depth",
+    "get_operand_value",
+    "CallHandlerRegistry",
+    "VmlinuxHandlerRegistry",
+]

pythonbpf/expr/call_registry.py

@@ -0,0 +1,20 @@
+class CallHandlerRegistry:
+    """Registry for handling different types of calls (helpers, etc.)"""
+
+    _handler = None
+
+    @classmethod
+    def set_handler(cls, handler):
+        """Set the handler for unknown calls"""
+        cls._handler = handler
+
+    @classmethod
+    def handle_call(
+        cls, call, module, builder, func, local_sym_tab, map_sym_tab, structs_sym_tab
+    ):
+        """Handle a call using the registered handler"""
+        if cls._handler is None:
+            return None
+        return cls._handler(
+            call, module, builder, func, local_sym_tab, map_sym_tab, structs_sym_tab
+        )

pythonbpf/expr/expr_pass.py

@@ -0,0 +1,622 @@
+import ast
+from llvmlite import ir
+from logging import Logger
+import logging
+from typing import Dict
+
+from pythonbpf.type_deducer import ctypes_to_ir, is_ctypes
+from .call_registry import CallHandlerRegistry
+from .type_normalization import (
+    convert_to_bool,
+    handle_comparator,
+    get_base_type_and_depth,
+    deref_to_depth,
+)
+from pythonbpf.vmlinux_parser.assignment_info import Field
+from .vmlinux_registry import VmlinuxHandlerRegistry
+
+logger: Logger = logging.getLogger(__name__)
+
+# ============================================================================
+# Leaf Handlers (No Recursive eval_expr calls)
+# ============================================================================
+
+
+def _handle_name_expr(expr: ast.Name, local_sym_tab: Dict, builder: ir.IRBuilder):
+    """Handle ast.Name expressions."""
+    if expr.id in local_sym_tab:
+        var = local_sym_tab[expr.id].var
+        val = builder.load(var)
+        return val, local_sym_tab[expr.id].ir_type
+    else:
+        # Check if it's a vmlinux enum/constant
+        vmlinux_result = VmlinuxHandlerRegistry.handle_name(expr.id)
+        if vmlinux_result is not None:
+            return vmlinux_result
+
+        raise SyntaxError(f"Undefined variable {expr.id}")
+
+
+def _handle_constant_expr(module, builder, expr: ast.Constant):
+    """Handle ast.Constant expressions."""
+    if isinstance(expr.value, int) or isinstance(expr.value, bool):
+        return ir.Constant(ir.IntType(64), int(expr.value)), ir.IntType(64)
+    elif isinstance(expr.value, str):
+        str_name = f".str.{id(expr)}"
+        str_bytes = expr.value.encode("utf-8") + b"\x00"
+        str_type = ir.ArrayType(ir.IntType(8), len(str_bytes))
+        str_constant = ir.Constant(str_type, bytearray(str_bytes))
+
+        # Create global variable
+        global_str = ir.GlobalVariable(module, str_type, name=str_name)
+        global_str.linkage = "internal"
+        global_str.global_constant = True
+        global_str.initializer = str_constant
+
+        str_ptr = builder.bitcast(global_str, ir.PointerType(ir.IntType(8)))
+        return str_ptr, ir.PointerType(ir.IntType(8))
+    else:
+        logger.error(f"Unsupported constant type {ast.dump(expr)}")
+        return None
+
+
+def _handle_attribute_expr(
+    expr: ast.Attribute,
+    local_sym_tab: Dict,
+    structs_sym_tab: Dict,
+    builder: ir.IRBuilder,
+):
+    """Handle ast.Attribute expressions for struct field access."""
+    if isinstance(expr.value, ast.Name):
+        var_name = expr.value.id
+        attr_name = expr.attr
+        if var_name in local_sym_tab:
+            var_ptr, var_type, var_metadata = local_sym_tab[var_name]
+            logger.info(f"Loading attribute {attr_name} from variable {var_name}")
+            logger.info(
+                f"Variable type: {var_type}, Variable ptr: {var_ptr}, Variable Metadata: {var_metadata}"
+            )
+            if (
+                hasattr(var_metadata, "__module__")
+                and var_metadata.__module__ == "vmlinux"
+            ):
+                # Try vmlinux handler when var_metadata is not a string, but has a module attribute.
+                # This has been done to keep everything separate in vmlinux struct handling.
+                vmlinux_result = VmlinuxHandlerRegistry.handle_attribute(
+                    expr, local_sym_tab, None, builder
+                )
+                if vmlinux_result is not None:
+                    return vmlinux_result
+                else:
+                    raise RuntimeError("Vmlinux struct did not process successfully")
+            metadata = structs_sym_tab[var_metadata]
+            if attr_name in metadata.fields:
+                gep = metadata.gep(builder, var_ptr, attr_name)
+                val = builder.load(gep)
+                field_type = metadata.field_type(attr_name)
+                return val, field_type
+    return None
+
+
+def _handle_deref_call(expr: ast.Call, local_sym_tab: Dict, builder: ir.IRBuilder):
+    """Handle deref function calls."""
+    logger.info(f"Handling deref {ast.dump(expr)}")
+    if len(expr.args) != 1:
+        logger.info("deref takes exactly one argument")
+        return None
+
+    arg = expr.args[0]
+    if (
+        isinstance(arg, ast.Call)
+        and isinstance(arg.func, ast.Name)
+        and arg.func.id == "deref"
+    ):
+        logger.info("Multiple deref not supported")
+        return None
+
+    if isinstance(arg, ast.Name):
+        if arg.id in local_sym_tab:
+            arg_ptr = local_sym_tab[arg.id].var
+        else:
+            logger.info(f"Undefined variable {arg.id}")
+            return None
+    else:
+        logger.info("Unsupported argument type for deref")
+        return None
+
+    if arg_ptr is None:
+        logger.info("Failed to evaluate deref argument")
+        return None
+
+    # Load the value from pointer
+    val = builder.load(arg_ptr)
+    return val, local_sym_tab[arg.id].ir_type
+
+
+# ============================================================================
+# Binary Operations
+# ============================================================================
+
+
+def get_operand_value(
+    func, module, operand, builder, local_sym_tab, map_sym_tab, structs_sym_tab=None
+):
+    """Extract the value from an operand, handling variables and constants."""
+    logger.info(f"Getting operand value for: {ast.dump(operand)}")
+    if isinstance(operand, ast.Name):
+        if operand.id in local_sym_tab:
+            var = local_sym_tab[operand.id].var
+            var_type = var.type
+            base_type, depth = get_base_type_and_depth(var_type)
+            logger.info(f"var is {var}, base_type is {base_type}, depth is {depth}")
+            val = deref_to_depth(func, builder, var, depth)
+            return val
+        else:
+            # Check if it's a vmlinux enum/constant
+            vmlinux_result = VmlinuxHandlerRegistry.handle_name(operand.id)
+            if vmlinux_result is not None:
+                val, _ = vmlinux_result
+                return val
+    elif isinstance(operand, ast.Constant):
+        if isinstance(operand.value, int):
+            cst = ir.Constant(ir.IntType(64), int(operand.value))
+            return cst
+        raise TypeError(f"Unsupported constant type: {type(operand.value)}")
+    elif isinstance(operand, ast.BinOp):
+        res = _handle_binary_op_impl(
+            func, module, operand, builder, local_sym_tab, map_sym_tab, structs_sym_tab
+        )
+        return res
+    else:
+        res = eval_expr(
+            func, module, builder, operand, local_sym_tab, map_sym_tab, structs_sym_tab
+        )
+        if res is None:
+            raise ValueError(f"Failed to evaluate call expression: {operand}")
+        val, _ = res
+        logger.info(f"Evaluated expr to {val} of type {val.type}")
+        base_type, depth = get_base_type_and_depth(val.type)
+        if depth > 0:
+            val = deref_to_depth(func, builder, val, depth)
+        return val
+    raise TypeError(f"Unsupported operand type: {type(operand)}")
+
+
+def _handle_binary_op_impl(
+    func, module, rval, builder, local_sym_tab, map_sym_tab, structs_sym_tab=None
+):
+    op = rval.op
+    left = get_operand_value(
+        func, module, rval.left, builder, local_sym_tab, map_sym_tab, structs_sym_tab
+    )
+    right = get_operand_value(
+        func, module, rval.right, builder, local_sym_tab, map_sym_tab, structs_sym_tab
+    )
+    logger.info(f"left is {left}, right is {right}, op is {op}")
+
+    # NOTE: Before doing the operation, if the operands are integers
+    # we always extend them to i64. The assignment to LHS will take
+    # care of truncation if needed.
+    if isinstance(left.type, ir.IntType) and left.type.width < 64:
+        left = builder.sext(left, ir.IntType(64))
+    if isinstance(right.type, ir.IntType) and right.type.width < 64:
+        right = builder.sext(right, ir.IntType(64))
+
+    # Map AST operation nodes to LLVM IR builder methods
+    op_map = {
+        ast.Add: builder.add,
+        ast.Sub: builder.sub,
+        ast.Mult: builder.mul,
+        ast.Div: builder.sdiv,
+        ast.Mod: builder.srem,
+        ast.LShift: builder.shl,
+        ast.RShift: builder.lshr,
+        ast.BitOr: builder.or_,
+        ast.BitXor: builder.xor,
+        ast.BitAnd: builder.and_,
+        ast.FloorDiv: builder.udiv,
+    }
+
+    if type(op) in op_map:
+        result = op_map[type(op)](left, right)
+        return result
+    else:
+        raise SyntaxError("Unsupported binary operation")
+
+
+def _handle_binary_op(
+    func,
+    module,
+    rval,
+    builder,
+    var_name,
+    local_sym_tab,
+    map_sym_tab,
+    structs_sym_tab=None,
+):
+    result = _handle_binary_op_impl(
+        func, module, rval, builder, local_sym_tab, map_sym_tab, structs_sym_tab
+    )
+    if var_name and var_name in local_sym_tab:
+        logger.info(
+            f"Storing result {result} into variable {local_sym_tab[var_name].var}"
+        )
+        builder.store(result, local_sym_tab[var_name].var)
+    return result, result.type
+
+
+# ============================================================================
+# Comparison and Unary Operations
+# ============================================================================
+
+
+def _handle_ctypes_call(
+    func,
+    module,
+    builder,
+    expr,
+    local_sym_tab,
+    map_sym_tab,
+    structs_sym_tab=None,
+):
+    """Handle ctypes type constructor calls."""
+    if len(expr.args) != 1:
+        logger.info("ctypes constructor takes exactly one argument")
+        return None
+
+    arg = expr.args[0]
+    val = eval_expr(
+        func,
+        module,
+        builder,
+        arg,
+        local_sym_tab,
+        map_sym_tab,
+        structs_sym_tab,
+    )
+    if val is None:
+        logger.info("Failed to evaluate argument to ctypes constructor")
+        return None
+    call_type = expr.func.id
+    expected_type = ctypes_to_ir(call_type)
+
+    # Extract the actual IR value and type
+    # val could be (value, ir_type) or (value, Field)
+    value, val_type = val
+
+    # If val_type is a Field object (from vmlinux struct), get the actual IR type of the value
+    if isinstance(val_type, Field):
+        # The value is already the correct IR value (potentially zero-extended)
+        # Get the IR type from the value itself
+        actual_ir_type = value.type
+        logger.info(
+            f"Converting vmlinux field {val_type.name} (IR type: {actual_ir_type}) to {call_type}"
+        )
+    else:
+        actual_ir_type = val_type
+
+    if actual_ir_type != expected_type:
+        # NOTE: We are only considering casting to and from int types for now
+        if isinstance(actual_ir_type, ir.IntType) and isinstance(
+            expected_type, ir.IntType
+        ):
+            if actual_ir_type.width < expected_type.width:
+                value = builder.sext(value, expected_type)
+                logger.info(
+                    f"Sign-extended from i{actual_ir_type.width} to i{expected_type.width}"
+                )
+            elif actual_ir_type.width > expected_type.width:
+                value = builder.trunc(value, expected_type)
+                logger.info(
+                    f"Truncated from i{actual_ir_type.width} to i{expected_type.width}"
+                )
+            else:
+                # Same width, just use as-is (e.g., both i64)
+                pass
+        else:
+            raise ValueError(
+                f"Type mismatch: expected {expected_type}, got {actual_ir_type} (original type: {val_type})"
+            )
+
+    return value, expected_type
+
+
+def _handle_compare(
+    func, module, builder, cond, local_sym_tab, map_sym_tab, structs_sym_tab=None
+):
+    """Handle ast.Compare expressions."""
+
+    if len(cond.ops) != 1 or len(cond.comparators) != 1:
+        logger.error("Only single comparisons are supported")
+        return None
+    lhs = eval_expr(
+        func,
+        module,
+        builder,
+        cond.left,
+        local_sym_tab,
+        map_sym_tab,
+        structs_sym_tab,
+    )
+    rhs = eval_expr(
+        func,
+        module,
+        builder,
+        cond.comparators[0],
+        local_sym_tab,
+        map_sym_tab,
+        structs_sym_tab,
+    )
+
+    if lhs is None or rhs is None:
+        logger.error("Failed to evaluate comparison operands")
+        return None
+
+    lhs, _ = lhs
+    rhs, _ = rhs
+    return handle_comparator(func, builder, cond.ops[0], lhs, rhs)
+
+
+def _handle_unary_op(
+    func,
+    module,
+    builder,
+    expr: ast.UnaryOp,
+    local_sym_tab,
+    map_sym_tab,
+    structs_sym_tab=None,
+):
+    """Handle ast.UnaryOp expressions."""
+    if not isinstance(expr.op, ast.Not) and not isinstance(expr.op, ast.USub):
+        logger.error("Only 'not' and '-' unary operators are supported")
+        return None
+
+    operand = get_operand_value(
+        func, module, expr.operand, builder, local_sym_tab, map_sym_tab, structs_sym_tab
+    )
+    if operand is None:
+        logger.error("Failed to evaluate operand for unary operation")
+        return None
+
+    if isinstance(expr.op, ast.Not):
+        true_const = ir.Constant(ir.IntType(1), 1)
+        result = builder.xor(convert_to_bool(builder, operand), true_const)
+        return result, ir.IntType(1)
+    elif isinstance(expr.op, ast.USub):
+        # Multiply by -1
+        neg_one = ir.Constant(ir.IntType(64), -1)
+        result = builder.mul(operand, neg_one)
+        return result, ir.IntType(64)
+    return None
+
+
+# ============================================================================
+# Boolean Operations
+# ============================================================================
+
+
+def _handle_and_op(func, builder, expr, local_sym_tab, map_sym_tab, structs_sym_tab):
+    """Handle `and` boolean operations."""
+
+    logger.debug(f"Handling 'and' operator with {len(expr.values)} operands")
+
+    merge_block = func.append_basic_block(name="and.merge")
+    false_block = func.append_basic_block(name="and.false")
+
+    incoming_values = []
+
+    for i, value in enumerate(expr.values):
+        is_last = i == len(expr.values) - 1
+
+        # Evaluate current operand
+        operand_result = eval_expr(
+            func, None, builder, value, local_sym_tab, map_sym_tab, structs_sym_tab
+        )
+        if operand_result is None:
+            logger.error(f"Failed to evaluate operand {i} in 'and' expression")
+            return None
+
+        operand_val, operand_type = operand_result
+
+        # Convert to boolean if needed
+        operand_bool = convert_to_bool(builder, operand_val)
+        current_block = builder.block
+
+        if is_last:
+            # Last operand: result is this value
+            builder.branch(merge_block)
+            incoming_values.append((operand_bool, current_block))
+        else:
+            # Not last: check if true, continue or short-circuit
+            next_check = func.append_basic_block(name=f"and.check_{i + 1}")
+            builder.cbranch(operand_bool, next_check, false_block)
+            builder.position_at_end(next_check)
+
+    # False block: short-circuit with false
+    builder.position_at_end(false_block)
+    builder.branch(merge_block)
+    false_value = ir.Constant(ir.IntType(1), 0)
+    incoming_values.append((false_value, false_block))
+
+    # Merge block: phi node
+    builder.position_at_end(merge_block)
+    phi = builder.phi(ir.IntType(1), name="and.result")
+    for val, block in incoming_values:
+        phi.add_incoming(val, block)
+
+    logger.debug(f"Generated 'and' with {len(incoming_values)} incoming values")
+    return phi, ir.IntType(1)
+
+
+def _handle_or_op(func, builder, expr, local_sym_tab, map_sym_tab, structs_sym_tab):
+    """Handle `or` boolean operations."""
+
+    logger.debug(f"Handling 'or' operator with {len(expr.values)} operands")
+
+    merge_block = func.append_basic_block(name="or.merge")
+    true_block = func.append_basic_block(name="or.true")
+
+    incoming_values = []
+
+    for i, value in enumerate(expr.values):
+        is_last = i == len(expr.values) - 1
+
+        # Evaluate current operand
+        operand_result = eval_expr(
+            func, None, builder, value, local_sym_tab, map_sym_tab, structs_sym_tab
+        )
+        if operand_result is None:
+            logger.error(f"Failed to evaluate operand {i} in 'or' expression")
+            return None
+
+        operand_val, operand_type = operand_result
+
+        # Convert to boolean if needed
+        operand_bool = convert_to_bool(builder, operand_val)
+        current_block = builder.block
+
+        if is_last:
+            # Last operand: result is this value
+            builder.branch(merge_block)
+            incoming_values.append((operand_bool, current_block))
+        else:
+            # Not last: check if false, continue or short-circuit
+            next_check = func.append_basic_block(name=f"or.check_{i + 1}")
+            builder.cbranch(operand_bool, true_block, next_check)
+            builder.position_at_end(next_check)
+
+    # True block: short-circuit with true
+    builder.position_at_end(true_block)
+    builder.branch(merge_block)
+    true_value = ir.Constant(ir.IntType(1), 1)
+    incoming_values.append((true_value, true_block))
+
+    # Merge block: phi node
+    builder.position_at_end(merge_block)
+    phi = builder.phi(ir.IntType(1), name="or.result")
+    for val, block in incoming_values:
+        phi.add_incoming(val, block)
+
+    logger.debug(f"Generated 'or' with {len(incoming_values)} incoming values")
+    return phi, ir.IntType(1)
+
+
+def _handle_boolean_op(
+    func,
+    module,
+    builder,
+    expr: ast.BoolOp,
+    local_sym_tab,
+    map_sym_tab,
+    structs_sym_tab=None,
+):
+    """Handle `and` and `or` boolean operations."""
+
+    if isinstance(expr.op, ast.And):
+        return _handle_and_op(
+            func, builder, expr, local_sym_tab, map_sym_tab, structs_sym_tab
+        )
+    elif isinstance(expr.op, ast.Or):
+        return _handle_or_op(
+            func, builder, expr, local_sym_tab, map_sym_tab, structs_sym_tab
+        )
+    else:
+        logger.error(f"Unsupported boolean operator: {type(expr.op).__name__}")
+        return None
+
+
+# ============================================================================
+# Expression Dispatcher
+# ============================================================================
+
+
+def eval_expr(
+    func,
+    module,
+    builder,
+    expr,
+    local_sym_tab,
+    map_sym_tab,
+    structs_sym_tab=None,
+):
+    logger.info(f"Evaluating expression: {ast.dump(expr)}")
+    if isinstance(expr, ast.Name):
+        return _handle_name_expr(expr, local_sym_tab, builder)
+    elif isinstance(expr, ast.Constant):
+        return _handle_constant_expr(module, builder, expr)
+    elif isinstance(expr, ast.Call):
+        if isinstance(expr.func, ast.Name) and expr.func.id == "deref":
+            return _handle_deref_call(expr, local_sym_tab, builder)
+
+        if isinstance(expr.func, ast.Name) and is_ctypes(expr.func.id):
+            return _handle_ctypes_call(
+                func,
+                module,
+                builder,
+                expr,
+                local_sym_tab,
+                map_sym_tab,
+                structs_sym_tab,
+            )
+
+        result = CallHandlerRegistry.handle_call(
+            expr, module, builder, func, local_sym_tab, map_sym_tab, structs_sym_tab
+        )
+        if result is not None:
+            return result
+
+        logger.warning(f"Unknown call: {ast.dump(expr)}")
+        return None
+    elif isinstance(expr, ast.Attribute):
+        return _handle_attribute_expr(expr, local_sym_tab, structs_sym_tab, builder)
+    elif isinstance(expr, ast.BinOp):
+        return _handle_binary_op(
+            func,
+            module,
+            expr,
+            builder,
+            None,
+            local_sym_tab,
+            map_sym_tab,
+            structs_sym_tab,
+        )
+    elif isinstance(expr, ast.Compare):
+        return _handle_compare(
+            func, module, builder, expr, local_sym_tab, map_sym_tab, structs_sym_tab
+        )
+    elif isinstance(expr, ast.UnaryOp):
+        return _handle_unary_op(
+            func, module, builder, expr, local_sym_tab, map_sym_tab, structs_sym_tab
+        )
+    elif isinstance(expr, ast.BoolOp):
+        return _handle_boolean_op(
+            func, module, builder, expr, local_sym_tab, map_sym_tab, structs_sym_tab
+        )
+    logger.info("Unsupported expression evaluation")
+    return None
+
+
+def handle_expr(
+    func,
+    module,
+    builder,
+    expr,
+    local_sym_tab,
+    map_sym_tab,
+    structs_sym_tab,
+):
+    """Handle expression statements in the function body."""
+    logger.info(f"Handling expression: {ast.dump(expr)}")
+    call = expr.value
+    if isinstance(call, ast.Call):
+        eval_expr(
+            func,
+            module,
+            builder,
+            call,
+            local_sym_tab,
+            map_sym_tab,
+            structs_sym_tab,
+        )
+    else:
+        logger.info("Unsupported expression type")

pythonbpf/expr/ir_ops.py

@@ -0,0 +1,50 @@
+import logging
+from llvmlite import ir
+
+logger = logging.getLogger(__name__)
+
+
+def deref_to_depth(func, builder, val, target_depth):
+    """Dereference a pointer to a certain depth."""
+
+    cur_val = val
+    cur_type = val.type
+
+    for depth in range(target_depth):
+        if not isinstance(val.type, ir.PointerType):
+            logger.error("Cannot dereference further, non-pointer type")
+            return None
+
+        # dereference with null check
+        pointee_type = cur_type.pointee
+        null_check_block = builder.block
+        not_null_block = func.append_basic_block(name=f"deref_not_null_{depth}")
+        merge_block = func.append_basic_block(name=f"deref_merge_{depth}")
+
+        null_ptr = ir.Constant(cur_type, None)
+        is_not_null = builder.icmp_signed("!=", cur_val, null_ptr)
+        logger.debug(f"Inserted null check for pointer at depth {depth}")
+
+        builder.cbranch(is_not_null, not_null_block, merge_block)
+
+        builder.position_at_end(not_null_block)
+        dereferenced_val = builder.load(cur_val)
+        logger.debug(f"Dereferenced to depth {depth - 1}, type: {pointee_type}")
+        builder.branch(merge_block)
+
+        builder.position_at_end(merge_block)
+        phi = builder.phi(pointee_type, name=f"deref_result_{depth}")
+
+        zero_value = (
+            ir.Constant(pointee_type, 0)
+            if isinstance(pointee_type, ir.IntType)
+            else ir.Constant(pointee_type, None)
+        )
+        phi.add_incoming(zero_value, null_check_block)
+
+        phi.add_incoming(dereferenced_val, not_null_block)
+
+        # Continue with phi result
+        cur_val = phi
+        cur_type = pointee_type
+    return cur_val

pythonbpf/expr/type_normalization.py

@@ -0,0 +1,83 @@
+import logging
+import ast
+from llvmlite import ir
+from .ir_ops import deref_to_depth
+
+logger = logging.getLogger(__name__)
+
+COMPARISON_OPS = {
+    ast.Eq: "==",
+    ast.NotEq: "!=",
+    ast.Lt: "<",
+    ast.LtE: "<=",
+    ast.Gt: ">",
+    ast.GtE: ">=",
+    ast.Is: "==",
+    ast.IsNot: "!=",
+}
+
+
+def get_base_type_and_depth(ir_type):
+    """Get the base type for pointer types."""
+    cur_type = ir_type
+    depth = 0
+    while isinstance(cur_type, ir.PointerType):
+        depth += 1
+        cur_type = cur_type.pointee
+    return cur_type, depth
+
+
+def _normalize_types(func, builder, lhs, rhs):
+    """Normalize types for comparison."""
+
+    logger.info(f"Normalizing types: {lhs.type} vs {rhs.type}")
+    if isinstance(lhs.type, ir.IntType) and isinstance(rhs.type, ir.IntType):
+        if lhs.type.width < rhs.type.width:
+            lhs = builder.sext(lhs, rhs.type)
+        else:
+            rhs = builder.sext(rhs, lhs.type)
+        return lhs, rhs
+    elif not isinstance(lhs.type, ir.PointerType) and not isinstance(
+        rhs.type, ir.PointerType
+    ):
+        logger.error(f"Type mismatch: {lhs.type} vs {rhs.type}")
+        return None, None
+    else:
+        lhs_base, lhs_depth = get_base_type_and_depth(lhs.type)
+        rhs_base, rhs_depth = get_base_type_and_depth(rhs.type)
+        if lhs_base == rhs_base:
+            if lhs_depth < rhs_depth:
+                rhs = deref_to_depth(func, builder, rhs, rhs_depth - lhs_depth)
+            elif rhs_depth < lhs_depth:
+                lhs = deref_to_depth(func, builder, lhs, lhs_depth - rhs_depth)
+            return _normalize_types(func, builder, lhs, rhs)
+
+
+def convert_to_bool(builder, val):
+    """Convert a value to boolean."""
+    if val.type == ir.IntType(1):
+        return val
+    if isinstance(val.type, ir.PointerType):
+        zero = ir.Constant(val.type, None)
+    else:
+        zero = ir.Constant(val.type, 0)
+    return builder.icmp_signed("!=", val, zero)
+
+
+def handle_comparator(func, builder, op, lhs, rhs):
+    """Handle comparison operations."""
+
+    if lhs.type != rhs.type:
+        lhs, rhs = _normalize_types(func, builder, lhs, rhs)
+
+    if lhs is None or rhs is None:
+        return None
+
+    if type(op) not in COMPARISON_OPS:
+        logger.error(f"Unsupported comparison operator: {type(op)}")
+        return None
+
+    predicate = COMPARISON_OPS[type(op)]
+    result = builder.icmp_signed(predicate, lhs, rhs)
+    logger.debug(f"Comparison result: {result}")
+    return result, ir.IntType(1)

pythonbpf/expr/vmlinux_registry.py

@@ -0,0 +1,75 @@
+import ast
+
+from pythonbpf.vmlinux_parser.vmlinux_exports_handler import VmlinuxHandler
+
+
+class VmlinuxHandlerRegistry:
+    """Registry for vmlinux handler operations"""
+
+    _handler = None
+
+    @classmethod
+    def set_handler(cls, handler: VmlinuxHandler):
+        """Set the vmlinux handler"""
+        cls._handler = handler
+
+    @classmethod
+    def get_handler(cls):
+        """Get the vmlinux handler"""
+        return cls._handler
+
+    @classmethod
+    def handle_name(cls, name):
+        """Try to handle a name as vmlinux enum/constant"""
+        if cls._handler is None:
+            return None
+        return cls._handler.handle_vmlinux_enum(name)
+
+    @classmethod
+    def handle_attribute(cls, expr, local_sym_tab, module, builder):
+        """Try to handle an attribute access as vmlinux struct field"""
+        if cls._handler is None:
+            return None
+
+        if isinstance(expr.value, ast.Name):
+            var_name = expr.value.id
+            field_name = expr.attr
+            return cls._handler.handle_vmlinux_struct_field(
+                var_name, field_name, module, builder, local_sym_tab
+            )
+        return None
+
+    @classmethod
+    def get_struct_debug_info(cls, name):
+        if cls._handler is None:
+            return False
+        return cls._handler.get_struct_debug_info(name)
+
+    @classmethod
+    def is_vmlinux_struct(cls, name):
+        """Check if a name refers to a vmlinux struct"""
+        if cls._handler is None:
+            return False
+        return cls._handler.is_vmlinux_struct(name)
+
+    @classmethod
+    def get_struct_type(cls, name):
+        """Try to handle a struct name as vmlinux struct"""
+        if cls._handler is None:
+            return None
+        return cls._handler.get_vmlinux_struct_type(name)
+
+    @classmethod
+    def has_field(cls, vmlinux_struct_name, field_name):
+        """Check if a vmlinux struct has a specific field"""
+        if cls._handler is None:
+            return False
+        return cls._handler.has_field(vmlinux_struct_name, field_name)
+
+    @classmethod
+    def get_field_type(cls, vmlinux_struct_name, field_name):
+        """Get the type of a field in a vmlinux struct"""
+        if cls._handler is None:
+            return None
+        assert isinstance(cls._handler, VmlinuxHandler)
+        return cls._handler.get_field_type(vmlinux_struct_name, field_name)

pythonbpf/functions/__init__.py

@@ -0,0 +1,3 @@
+from .functions_pass import func_proc
+
+__all__ = ["func_proc"]

pythonbpf/functions/function_debug_info.py

@@ -0,0 +1,82 @@
+import ast
+
+import llvmlite.ir as ir
+import logging
+from pythonbpf.debuginfo import DebugInfoGenerator
+from pythonbpf.expr import VmlinuxHandlerRegistry
+import ctypes
+
+logger = logging.getLogger(__name__)
+
+
+def generate_function_debug_info(
+    func_node: ast.FunctionDef, module: ir.Module, func: ir.Function
+):
+    generator = DebugInfoGenerator(module)
+    leading_argument = func_node.args.args[0]
+    leading_argument_name = leading_argument.arg
+    annotation = leading_argument.annotation
+    if func_node.returns is None:
+        # TODO: should check if this logic is consistent with function return type handling elsewhere
+        return_type = ctypes.c_int64()
+    elif hasattr(func_node.returns, "id"):
+        return_type = func_node.returns.id
+        if return_type == "c_int32":
+            return_type = generator.get_int32_type()
+        elif return_type == "c_int64":
+            return_type = generator.get_int64_type()
+        elif return_type == "c_uint32":
+            return_type = generator.get_uint32_type()
+        elif return_type == "c_uint64":
+            return_type = generator.get_uint64_type()
+        else:
+            logger.warning(
+                "Return type should be int32, int64, uint32 or uint64 only. Falling back to int64"
+            )
+            return_type = generator.get_int64_type()
+    else:
+        return_type = ctypes.c_int64()
+    # context processing
+    if annotation is None:
+        logger.warning("Type of context of function not found.")
+        return
+    if hasattr(annotation, "id"):
+        ctype_name = annotation.id
+        if ctype_name == "c_void_p":
+            return
+        elif ctype_name.startswith("ctypes"):
+            raise SyntaxError(
+                "The first argument should always be a pointer to a struct or a void pointer"
+            )
+        context_debug_info = VmlinuxHandlerRegistry.get_struct_debug_info(annotation.id)
+
+        # Create pointer to context this must be created fresh for each function
+        # to avoid circular reference issues when the same struct is used in multiple functions
+        pointer_to_context_debug_info = generator.create_pointer_type(
+            context_debug_info, 64
+        )
+
+        # Create subroutine type - also fresh for each function
+        subroutine_type = generator.create_subroutine_type(
+            return_type, pointer_to_context_debug_info
+        )
+
+        # Create local variable - fresh for each function with unique name
+        context_local_variable = generator.create_local_variable_debug_info(
+            leading_argument_name, 1, pointer_to_context_debug_info
+        )
+
+        retained_nodes = [context_local_variable]
+        logger.info(f"Generating debug info for function {func_node.name}")
+
+        # Create subprogram with is_distinct=True to ensure each function gets unique debug info
+        subprogram_debug_info = generator.create_subprogram(
+            func_node.name, subroutine_type, retained_nodes
+        )
+        generator.add_scope_to_local_variable(
+            context_local_variable, subprogram_debug_info
+        )
+        func.set_metadata("dbg", subprogram_debug_info)
+
+    else:
+        logger.error(f"Invalid annotation type for argument '{leading_argument_name}'")

pythonbpf/functions/function_metadata.py

@@ -0,0 +1,88 @@
+import ast
+
+
+def get_probe_string(func_node):
+    """Extract the probe string from the decorator of the function node"""
+    # TODO: right now we have the whole string in the section decorator
+    # But later we can implement typed tuples for tracepoints and kprobes
+    # For helper functions, we return "helper"
+
+    for decorator in func_node.decorator_list:
+        if isinstance(decorator, ast.Name) and decorator.id == "bpfglobal":
+            return None
+        if isinstance(decorator, ast.Call) and isinstance(decorator.func, ast.Name):
+            if decorator.func.id == "section" and len(decorator.args) == 1:
+                arg = decorator.args[0]
+                if isinstance(arg, ast.Constant) and isinstance(arg.value, str):
+                    return arg.value
+    return "helper"
+
+
+def is_global_function(func_node):
+    """Check if the function is a global"""
+    for decorator in func_node.decorator_list:
+        if isinstance(decorator, ast.Name) and decorator.id in (
+            "map",
+            "bpfglobal",
+            "struct",
+        ):
+            return True
+    return False
+
+
+def infer_return_type(func_node: ast.FunctionDef):
+    if not isinstance(func_node, (ast.FunctionDef, ast.AsyncFunctionDef)):
+        raise TypeError("Expected ast.FunctionDef")
+    if func_node.returns is not None:
+        try:
+            return ast.unparse(func_node.returns)
+        except Exception:
+            node = func_node.returns
+            if isinstance(node, ast.Name):
+                return node.id
+            if isinstance(node, ast.Attribute):
+                return getattr(node, "attr", type(node).__name__)
+            try:
+                return str(node)
+            except Exception:
+                return type(node).__name__
+    found_type = None
+
+    def _expr_type(e):
+        if e is None:
+            return "None"
+        if isinstance(e, ast.Constant):
+            return type(e.value).__name__
+        if isinstance(e, ast.Name):
+            return e.id
+        if isinstance(e, ast.Call):
+            f = e.func
+            if isinstance(f, ast.Name):
+                return f.id
+            if isinstance(f, ast.Attribute):
+                try:
+                    return ast.unparse(f)
+                except Exception:
+                    return getattr(f, "attr", type(f).__name__)
+            try:
+                return ast.unparse(f)
+            except Exception:
+                return type(f).__name__
+        if isinstance(e, ast.Attribute):
+            try:
+                return ast.unparse(e)
+            except Exception:
+                return getattr(e, "attr", type(e).__name__)
+        try:
+            return ast.unparse(e)
+        except Exception:
+            return type(e).__name__
+
+    for walked_node in ast.walk(func_node):
+        if isinstance(walked_node, ast.Return):
+            t = _expr_type(walked_node.value)
+            if found_type is None:
+                found_type = t
+            elif found_type != t:
+                raise ValueError(f"Conflicting return types: {found_type} vs {t}")
+    return found_type or "None"

pythonbpf/functions/functions_pass.py

@@ -0,0 +1,512 @@
+from llvmlite import ir
+import ast
+import logging
+
+from pythonbpf.helper import (
+    HelperHandlerRegistry,
+    reset_scratch_pool,
+)
+from pythonbpf.type_deducer import ctypes_to_ir
+from pythonbpf.expr import (
+    eval_expr,
+    handle_expr,
+    convert_to_bool,
+    VmlinuxHandlerRegistry,
+)
+from pythonbpf.assign_pass import (
+    handle_variable_assignment,
+    handle_struct_field_assignment,
+)
+from pythonbpf.allocation_pass import (
+    handle_assign_allocation,
+    allocate_temp_pool,
+    create_targets_and_rvals,
+    LocalSymbol,
+)
+from .function_debug_info import generate_function_debug_info
+from .return_utils import handle_none_return, handle_xdp_return, is_xdp_name
+from .function_metadata import get_probe_string, is_global_function, infer_return_type
+
+
+logger = logging.getLogger(__name__)
+
+
+# ============================================================================
+# SECTION 1: Memory Allocation
+# ============================================================================
+
+
+def count_temps_in_call(call_node, local_sym_tab):
+    """Count the number of temporary variables needed for a function call."""
+
+    count = {}
+    is_helper = False
+
+    # NOTE: We exclude print calls for now
+    if isinstance(call_node.func, ast.Name):
+        if (
+            HelperHandlerRegistry.has_handler(call_node.func.id)
+            and call_node.func.id != "print"
+        ):
+            is_helper = True
+            func_name = call_node.func.id
+    elif isinstance(call_node.func, ast.Attribute):
+        if HelperHandlerRegistry.has_handler(call_node.func.attr):
+            is_helper = True
+            func_name = call_node.func.attr
+
+    if not is_helper:
+        return {}  # No temps needed
+
+    for arg_idx in range(len(call_node.args)):
+        # NOTE: Count all non-name arguments
+        # For struct fields, if it is being passed as an argument,
+        # The struct object should already exist in the local_sym_tab
+        arg = call_node.args[arg_idx]
+        if isinstance(arg, ast.Name) or (
+            isinstance(arg, ast.Attribute) and arg.value.id in local_sym_tab
+        ):
+            continue
+        param_type = HelperHandlerRegistry.get_param_type(func_name, arg_idx)
+        if isinstance(param_type, ir.PointerType):
+            pointee_type = param_type.pointee
+            count[pointee_type] = count.get(pointee_type, 0) + 1
+
+    return count
+
+
+def handle_if_allocation(
+    module, builder, stmt, func, ret_type, map_sym_tab, local_sym_tab, structs_sym_tab
+):
+    """Recursively handle allocations in if/else branches."""
+    if stmt.body:
+        allocate_mem(
+            module,
+            builder,
+            stmt.body,
+            func,
+            ret_type,
+            map_sym_tab,
+            local_sym_tab,
+            structs_sym_tab,
+        )
+    if stmt.orelse:
+        allocate_mem(
+            module,
+            builder,
+            stmt.orelse,
+            func,
+            ret_type,
+            map_sym_tab,
+            local_sym_tab,
+            structs_sym_tab,
+        )
+
+
+def allocate_mem(
+    module, builder, body, func, ret_type, map_sym_tab, local_sym_tab, structs_sym_tab
+):
+    max_temps_needed = {}
+
+    def merge_type_counts(count_dict):
+        nonlocal max_temps_needed
+        for typ, cnt in count_dict.items():
+            max_temps_needed[typ] = max(max_temps_needed.get(typ, 0), cnt)
+
+    def update_max_temps_for_stmt(stmt):
+        nonlocal max_temps_needed
+
+        if isinstance(stmt, ast.If):
+            for s in stmt.body:
+                update_max_temps_for_stmt(s)
+            for s in stmt.orelse:
+                update_max_temps_for_stmt(s)
+            return
+
+        stmt_temps = {}
+        for node in ast.walk(stmt):
+            if isinstance(node, ast.Call):
+                call_temps = count_temps_in_call(node, local_sym_tab)
+                for typ, cnt in call_temps.items():
+                    stmt_temps[typ] = stmt_temps.get(typ, 0) + cnt
+        merge_type_counts(stmt_temps)
+
+    for stmt in body:
+        update_max_temps_for_stmt(stmt)
+
+        # Handle allocations
+        if isinstance(stmt, ast.If):
+            handle_if_allocation(
+                module,
+                builder,
+                stmt,
+                func,
+                ret_type,
+                map_sym_tab,
+                local_sym_tab,
+                structs_sym_tab,
+            )
+        elif isinstance(stmt, ast.Assign):
+            handle_assign_allocation(builder, stmt, local_sym_tab, structs_sym_tab)
+
+    allocate_temp_pool(builder, max_temps_needed, local_sym_tab)
+
+    return local_sym_tab
+
+
+# ============================================================================
+# SECTION 2: Statement Handlers
+# ============================================================================
+
+
+def handle_assign(
+    func, module, builder, stmt, map_sym_tab, local_sym_tab, structs_sym_tab
+):
+    """Handle assignment statements in the function body."""
+
+    # NOTE: Support multi-target assignments (e.g.: a, b = 1, 2)
+    targets, rvals = create_targets_and_rvals(stmt)
+
+    for target, rval in zip(targets, rvals):
+        if isinstance(target, ast.Name):
+            # NOTE: Simple variable assignment case: x = 5
+            var_name = target.id
+            result = handle_variable_assignment(
+                func,
+                module,
+                builder,
+                var_name,
+                rval,
+                local_sym_tab,
+                map_sym_tab,
+                structs_sym_tab,
+            )
+            if not result:
+                logger.error(f"Failed to handle assignment to {var_name}")
+            continue
+
+        if isinstance(target, ast.Attribute):
+            # NOTE: Struct field assignment case: pkt.field = value
+            handle_struct_field_assignment(
+                func,
+                module,
+                builder,
+                target,
+                rval,
+                local_sym_tab,
+                map_sym_tab,
+                structs_sym_tab,
+            )
+            continue
+
+        # Unsupported target type
+        logger.error(f"Unsupported assignment target: {ast.dump(target)}")
+
+
+def handle_cond(
+    func, module, builder, cond, local_sym_tab, map_sym_tab, structs_sym_tab=None
+):
+    val = eval_expr(
+        func, module, builder, cond, local_sym_tab, map_sym_tab, structs_sym_tab
+    )[0]
+    return convert_to_bool(builder, val)
+
+
+def handle_if(
+    func, module, builder, stmt, map_sym_tab, local_sym_tab, structs_sym_tab=None
+):
+    """Handle if statements in the function body."""
+    logger.info("Handling if statement")
+    # start = builder.block.parent
+    then_block = func.append_basic_block(name="if.then")
+    merge_block = func.append_basic_block(name="if.end")
+    if stmt.orelse:
+        else_block = func.append_basic_block(name="if.else")
+    else:
+        else_block = None
+
+    cond = handle_cond(
+        func, module, builder, stmt.test, local_sym_tab, map_sym_tab, structs_sym_tab
+    )
+    if else_block:
+        builder.cbranch(cond, then_block, else_block)
+    else:
+        builder.cbranch(cond, then_block, merge_block)
+
+    builder.position_at_end(then_block)
+    for s in stmt.body:
+        process_stmt(
+            func, module, builder, s, local_sym_tab, map_sym_tab, structs_sym_tab, False
+        )
+    if not builder.block.is_terminated:
+        builder.branch(merge_block)
+
+    if else_block:
+        builder.position_at_end(else_block)
+        for s in stmt.orelse:
+            process_stmt(
+                func,
+                module,
+                builder,
+                s,
+                local_sym_tab,
+                map_sym_tab,
+                structs_sym_tab,
+                False,
+            )
+        if not builder.block.is_terminated:
+            builder.branch(merge_block)
+
+    builder.position_at_end(merge_block)
+
+
+def handle_return(builder, stmt, local_sym_tab, ret_type):
+    logger.info(f"Handling return statement: {ast.dump(stmt)}")
+    if stmt.value is None:
+        return handle_none_return(builder)
+    elif isinstance(stmt.value, ast.Name) and is_xdp_name(stmt.value.id):
+        return handle_xdp_return(stmt, builder, ret_type)
+    else:
+        val = eval_expr(
+            func=None,
+            module=None,
+            builder=builder,
+            expr=stmt.value,
+            local_sym_tab=local_sym_tab,
+            map_sym_tab={},
+            structs_sym_tab={},
+        )
+        logger.info(f"Evaluated return expression to {val}")
+        builder.ret(val[0])
+        return True
+
+
+def process_stmt(
+    func,
+    module,
+    builder,
+    stmt,
+    local_sym_tab,
+    map_sym_tab,
+    structs_sym_tab,
+    did_return,
+    ret_type=ir.IntType(64),
+):
+    logger.info(f"Processing statement: {ast.dump(stmt)}")
+    reset_scratch_pool()
+    if isinstance(stmt, ast.Expr):
+        handle_expr(
+            func,
+            module,
+            builder,
+            stmt,
+            local_sym_tab,
+            map_sym_tab,
+            structs_sym_tab,
+        )
+    elif isinstance(stmt, ast.Assign):
+        handle_assign(
+            func, module, builder, stmt, map_sym_tab, local_sym_tab, structs_sym_tab
+        )
+    elif isinstance(stmt, ast.AugAssign):
+        raise SyntaxError("Augmented assignment not supported")
+    elif isinstance(stmt, ast.If):
+        handle_if(
+            func, module, builder, stmt, map_sym_tab, local_sym_tab, structs_sym_tab
+        )
+    elif isinstance(stmt, ast.Return):
+        did_return = handle_return(
+            builder,
+            stmt,
+            local_sym_tab,
+            ret_type,
+        )
+    return did_return
+
+
+# ============================================================================
+# SECTION 3: Function Body Processing
+# ============================================================================
+
+
+def process_func_body(
+    module,
+    builder,
+    func_node,
+    func,
+    ret_type,
+    map_sym_tab,
+    structs_sym_tab,
+):
+    """Process the body of a bpf function"""
+    # TODO: A lot.  We just have print -> bpf_trace_printk for now
+    did_return = False
+
+    local_sym_tab = {}
+
+    # Add the context parameter (first function argument) to the local symbol table
+    if func_node.args.args and len(func_node.args.args) > 0:
+        context_arg = func_node.args.args[0]
+        context_name = context_arg.arg
+
+        if hasattr(context_arg, "annotation") and context_arg.annotation:
+            if isinstance(context_arg.annotation, ast.Name):
+                context_type_name = context_arg.annotation.id
+            elif isinstance(context_arg.annotation, ast.Attribute):
+                context_type_name = context_arg.annotation.attr
+            else:
+                raise TypeError(
+                    f"Unsupported annotation type: {ast.dump(context_arg.annotation)}"
+                )
+            if VmlinuxHandlerRegistry.is_vmlinux_struct(context_type_name):
+                resolved_type = VmlinuxHandlerRegistry.get_struct_type(
+                    context_type_name
+                )
+                context_type = LocalSymbol(None, None, resolved_type)
+                local_sym_tab[context_name] = context_type
+                logger.info(f"Added argument '{context_name}' to local symbol table")
+
+    # pre-allocate dynamic variables
+    local_sym_tab = allocate_mem(
+        module,
+        builder,
+        func_node.body,
+        func,
+        ret_type,
+        map_sym_tab,
+        local_sym_tab,
+        structs_sym_tab,
+    )
+
+    logger.info(f"Local symbol table: {local_sym_tab.keys()}")
+
+    for stmt in func_node.body:
+        did_return = process_stmt(
+            func,
+            module,
+            builder,
+            stmt,
+            local_sym_tab,
+            map_sym_tab,
+            structs_sym_tab,
+            did_return,
+            ret_type,
+        )
+
+    if not did_return:
+        builder.ret(ir.Constant(ir.IntType(64), 0))
+
+
+def process_bpf_chunk(func_node, module, return_type, map_sym_tab, structs_sym_tab):
+    """Process a single BPF chunk (function) and emit corresponding LLVM IR."""
+
+    func_name = func_node.name
+
+    ret_type = return_type
+
+    # TODO: parse parameters
+    param_types = []
+    if func_node.args.args:
+        # Assume first arg to be ctx
+        param_types.append(ir.PointerType())
+
+    func_ty = ir.FunctionType(ret_type, param_types)
+    func = ir.Function(module, func_ty, func_name)
+
+    func.linkage = "dso_local"
+    func.attributes.add("nounwind")
+    func.attributes.add("noinline")
+    # func.attributes.add("optnone")
+
+    if func_node.args.args:
+        # Only look at the first argument for now
+        param = func.args[0]
+        param.add_attribute("nocapture")
+
+    probe_string = get_probe_string(func_node)
+    if probe_string is not None:
+        func.section = probe_string
+
+    block = func.append_basic_block(name="entry")
+    builder = ir.IRBuilder(block)
+
+    process_func_body(
+        module,
+        builder,
+        func_node,
+        func,
+        ret_type,
+        map_sym_tab,
+        structs_sym_tab,
+    )
+    return func
+
+
+# ============================================================================
+# SECTION 4: Top-Level Function Processor
+# ============================================================================
+
+
+def func_proc(tree, module, chunks, map_sym_tab, structs_sym_tab):
+    for func_node in chunks:
+        if is_global_function(func_node):
+            continue
+        func_type = get_probe_string(func_node)
+        logger.info(f"Found probe_string of {func_node.name}: {func_type}")
+
+        func = process_bpf_chunk(
+            func_node,
+            module,
+            ctypes_to_ir(infer_return_type(func_node)),
+            map_sym_tab,
+            structs_sym_tab,
+        )
+
+        logger.info(f"Generating Debug Info for Function {func_node.name}")
+        generate_function_debug_info(func_node, module, func)
+
+
+# TODO: WIP, for string assignment to fixed-size arrays
+def assign_string_to_array(builder, target_array_ptr, source_string_ptr, array_length):
+    """
+    Copy a string (i8*) to a fixed-size array ([N x i8]*)
+    """
+    # Create a loop to copy characters one by one
+    # entry_block = builder.block
+    copy_block = builder.append_basic_block("copy_char")
+    end_block = builder.append_basic_block("copy_end")
+
+    # Create loop counter
+    i = builder.alloca(ir.IntType(32))
+    builder.store(ir.Constant(ir.IntType(32), 0), i)
+
+    # Start the loop
+    builder.branch(copy_block)
+
+    # Copy loop
+    builder.position_at_end(copy_block)
+    idx = builder.load(i)
+    in_bounds = builder.icmp_unsigned(
+        "<", idx, ir.Constant(ir.IntType(32), array_length)
+    )
+    builder.cbranch(in_bounds, copy_block, end_block)
+
+    with builder.if_then(in_bounds):
+        # Load character from source
+        src_ptr = builder.gep(source_string_ptr, [idx])
+        char = builder.load(src_ptr)
+
+        # Store character in target
+        dst_ptr = builder.gep(target_array_ptr, [ir.Constant(ir.IntType(32), 0), idx])
+        builder.store(char, dst_ptr)
+
+        # Increment counter
+        next_idx = builder.add(idx, ir.Constant(ir.IntType(32), 1))
+        builder.store(next_idx, i)
+
+    builder.position_at_end(end_block)
+
+    # Ensure null termination
+    last_idx = ir.Constant(ir.IntType(32), array_length - 1)
+    null_ptr = builder.gep(target_array_ptr, [ir.Constant(ir.IntType(32), 0), last_idx])
+    builder.store(ir.Constant(ir.IntType(8), 0), null_ptr)

pythonbpf/functions/return_utils.py

@@ -0,0 +1,44 @@
+import logging
+import ast
+
+from llvmlite import ir
+
+logger: logging.Logger = logging.getLogger(__name__)
+
+XDP_ACTIONS = {
+    "XDP_ABORTED": 0,
+    "XDP_DROP": 1,
+    "XDP_PASS": 2,
+    "XDP_TX": 3,
+    "XDP_REDIRECT": 4,
+}
+
+
+def handle_none_return(builder) -> bool:
+    """Handle return or return None -> returns 0."""
+    builder.ret(ir.Constant(ir.IntType(64), 0))
+    logger.debug("Generated default return: 0")
+    return True
+
+
+def is_xdp_name(name: str) -> bool:
+    """Check if a name is an XDP action"""
+    return name in XDP_ACTIONS
+
+
+def handle_xdp_return(stmt: ast.Return, builder, ret_type) -> bool:
+    """Handle XDP returns"""
+    if not isinstance(stmt.value, ast.Name):
+        return False
+
+    action_name = stmt.value.id
+
+    if action_name not in XDP_ACTIONS:
+        raise ValueError(
+            f"Unknown XDP action: {action_name}. Available: {XDP_ACTIONS.keys()}"
+        )
+
+    value = XDP_ACTIONS[action_name]
+    builder.ret(ir.Constant(ret_type, value))
+    logger.debug(f"Generated XDP action return: {action_name} = {value}")
+    return True

pythonbpf/functions_pass.py

@@ -1,261 +0,0 @@
-from llvmlite import ir
-import ast
-
-from .bpf_helper_handler import bpf_printk_emitter, bpf_ktime_get_ns_emitter, bpf_map_lookup_elem_emitter
-from .type_deducer import ctypes_to_ir
-
-
-def get_probe_string(func_node):
-    """Extract the probe string from the decorator of the function node."""
-    # TODO: right now we have the whole string in the section decorator
-    # But later we can implement typed tuples for tracepoints and kprobes
-    # For helper functions, we return "helper"
-
-    for decorator in func_node.decorator_list:
-        if isinstance(decorator, ast.Name) and decorator.id == "bpfglobal":
-            return None
-        if isinstance(decorator, ast.Call) and isinstance(decorator.func, ast.Name):
-            if decorator.func.id == "section" and len(decorator.args) == 1:
-                arg = decorator.args[0]
-                if isinstance(arg, ast.Constant) and isinstance(arg.value, str):
-                    return arg.value
-    return "helper"
-
-
-def handle_assign(module, builder, stmt, map_sym_tab, local_sym_tab):
-    """Handle assignment statements in the function body."""
-    if len(stmt.targets) != 1:
-        print("Unsupported multiassignment")
-        return
-
-    num_types = ("c_int32", "c_int64", "c_uint32", "c_uint64")
-
-    target = stmt.targets[0]
-    if not isinstance(target, ast.Name):
-        print("Unsupported assignment target")
-        return
-    var_name = target.id
-    rval = stmt.value
-    if isinstance(rval, ast.Constant):
-        if isinstance(rval.value, int):
-            # Assume c_int64 for now
-            # TODO: make symtab for this
-            var = builder.alloca(ir.IntType(64), name=var_name)
-            var.align = 8
-            builder.store(ir.Constant(ir.IntType(64), rval.value), var)
-            local_sym_tab[var_name] = var
-            print(f"Assigned constant {rval.value} to {var_name}")
-    elif isinstance(rval, ast.Call):
-        if isinstance(rval.func, ast.Name):
-            call_type = rval.func.id
-            print(f"Assignment call type: {call_type}")
-            if call_type in num_types and len(rval.args) == 1 and isinstance(rval.args[0], ast.Constant) and isinstance(rval.args[0].value, int):
-                ir_type = ctypes_to_ir(call_type)
-                var = builder.alloca(ir_type, name=var_name)
-                var.align = ir_type.width // 8
-                builder.store(ir.Constant(ir_type, rval.args[0].value), var)
-                print(f"Assigned {call_type} constant "
-                      f"{rval.args[0].value} to {var_name}")
-                local_sym_tab[var_name] = var
-            else:
-                print(f"Unsupported assignment call type: {call_type}")
-        elif isinstance(rval.func, ast.Attribute):
-            if isinstance(rval.func.attr, str) and rval.func.attr == "lookup":
-                # Get map name and check symtab
-                # maps are called as funcs
-                if isinstance(rval.func.value, ast.Call) and isinstance(rval.func.value.func, ast.Name):
-                    map_name = rval.func.value.func.id
-                    if map_name in map_sym_tab:
-                        map_global = map_sym_tab[map_name]
-                        print(f"Found map {map_name} in symtab for lookup")
-                        if len(rval.args) != 1:
-                            print("Unsupported lookup with != 1 arg")
-                            return
-                        key_arg = rval.args[0]
-                        print(f"Lookup key arg type: {type(key_arg)}")
-                        # TODO: implement a parse_arg ffs as this can be a fucking expr
-                        if isinstance(key_arg, ast.Constant) and isinstance(key_arg.value, int):
-                            key_val = key_arg.value
-                            key_type = ir.IntType(64)
-                            print(f"Key type: {key_type}")
-                            print(f"Key val: {key_val}")
-                            key_var = builder.alloca(key_type)
-                            key_var.align = key_type // 8
-                            builder.store(ir.Constant(
-                                key_type, key_val), key_var)
-                        elif isinstance(key_arg, ast.Name):
-                            # Check in local symtab first
-                            if key_arg.id in local_sym_tab:
-                                key_var = local_sym_tab[key_arg.id]
-                                key_type = key_var.type.pointee
-                            elif key_arg.id in map_sym_tab:
-                                key_var = map_sym_tab[key_arg.id]
-                                key_type = key_var.type.pointee
-                            else:
-                                print("Key variable "
-                                      f"{key_arg.id} not found in symtabs")
-                                return
-                            print(f"Found key variable {key_arg.id} in symtab")
-                            print(f"Key type: {key_type}")
-                        else:
-                            print("Unsupported lookup key arg")
-                            return
-
-                        # TODO: generate call to bpf_map_lookup_elem
-                        result_ptr = bpf_map_lookup_elem_emitter(
-                            map_global, key_var, module, builder)
-
-                    else:
-                        print(f"Map {map_name} not found in symbol table")
-            else:
-                print("Unsupported assignment from method call")
-
-
-def process_func_body(module, builder, func_node, func, ret_type, map_sym_tab):
-    """Process the body of a bpf function"""
-    # TODO: A lot.  We just have print -> bpf_trace_printk for now
-    did_return = False
-
-    local_sym_tab = {}
-
-    for stmt in func_node.body:
-        if isinstance(stmt, ast.Expr) and isinstance(stmt.value, ast.Call):
-            call = stmt.value
-            if isinstance(call.func, ast.Name) and call.func.id == "print":
-                bpf_printk_emitter(call, module, builder, func)
-            if isinstance(call.func, ast.Name) and call.func.id == "bpf_ktime_get_ns":
-                bpf_ktime_get_ns_emitter(call, module, builder, func)
-        elif isinstance(stmt, ast.Assign):
-            handle_assign(module, builder, stmt, map_sym_tab, local_sym_tab)
-        elif isinstance(stmt, ast.Return):
-            if stmt.value is None:
-                builder.ret(ir.Constant(ir.IntType(32), 0))
-                did_return = True
-            elif isinstance(stmt.value, ast.Call) and isinstance(stmt.value.func, ast.Name) and len(stmt.value.args) == 1 and isinstance(stmt.value.args[0], ast.Constant) and isinstance(stmt.value.args[0].value, int):
-                call_type = stmt.value.func.id
-                if ctypes_to_ir(call_type) != ret_type:
-                    raise ValueError("Return type mismatch: expected"
-                                     f"{ctypes_to_ir(call_type)}, got {call_type}")
-                else:
-                    builder.ret(ir.Constant(
-                        ret_type, stmt.value.args[0].value))
-                    did_return = True
-            else:
-                print("Unsupported return value")
-    if not did_return:
-        builder.ret(ir.Constant(ir.IntType(32), 0))
-
-
-def process_bpf_chunk(func_node, module, return_type, map_sym_tab):
-    """Process a single BPF chunk (function) and emit corresponding LLVM IR."""
-
-    func_name = func_node.name
-
-    ret_type = return_type
-
-    # TODO: parse parameters
-    param_types = []
-    if func_node.args.args:
-        # Assume first arg to be ctx
-        param_types.append(ir.PointerType())
-
-    func_ty = ir.FunctionType(ret_type, param_types)
-    func = ir.Function(module, func_ty, func_name)
-
-    func.linkage = "dso_local"
-    func.attributes.add("nounwind")
-    func.attributes.add("noinline")
-    func.attributes.add("optnone")
-
-    if func_node.args.args:
-        # Only look at the first argument for now
-        param = func.args[0]
-        param.add_attribute("nocapture")
-
-    probe_string = get_probe_string(func_node)
-    if probe_string is not None:
-        func.section = probe_string
-
-    block = func.append_basic_block(name="entry")
-    builder = ir.IRBuilder(block)
-
-    process_func_body(module, builder, func_node, func, ret_type, map_sym_tab)
-
-    return func
-
-
-def func_proc(tree, module, chunks, map_sym_tab):
-    for func_node in chunks:
-        is_global = False
-        for decorator in func_node.decorator_list:
-            if isinstance(decorator, ast.Name) and decorator.id == "map":
-                is_global = True
-                break
-            elif isinstance(decorator, ast.Name) and decorator.id == "bpfglobal":
-                is_global = True
-                break
-        if is_global:
-            continue
-        func_type = get_probe_string(func_node)
-        print(f"Found probe_string of {func_node.name}: {func_type}")
-
-        process_bpf_chunk(func_node, module, ctypes_to_ir(
-            infer_return_type(func_node)), map_sym_tab)
-
-
-def infer_return_type(func_node: ast.FunctionDef):
-    if not isinstance(func_node, (ast.FunctionDef, ast.AsyncFunctionDef)):
-        raise TypeError("Expected ast.FunctionDef")
-    if func_node.returns is not None:
-        try:
-            return ast.unparse(func_node.returns)
-        except Exception:
-            node = func_node.returns
-            if isinstance(node, ast.Name):
-                return node.id
-            if isinstance(node, ast.Attribute):
-                return getattr(node, "attr", type(node).__name__)
-            try:
-                return str(node)
-            except Exception:
-                return type(node).__name__
-    found_type = None
-
-    def _expr_type(e):
-        if e is None:
-            return "None"
-        if isinstance(e, ast.Constant):
-            return type(e.value).__name__
-        if isinstance(e, ast.Name):
-            return e.id
-        if isinstance(e, ast.Call):
-            f = e.func
-            if isinstance(f, ast.Name):
-                return f.id
-            if isinstance(f, ast.Attribute):
-                try:
-                    return ast.unparse(f)
-                except Exception:
-                    return getattr(f, "attr", type(f).__name__)
-            try:
-                return ast.unparse(f)
-            except Exception:
-                return type(f).__name__
-        if isinstance(e, ast.Attribute):
-            try:
-                return ast.unparse(e)
-            except Exception:
-                return getattr(e, "attr", type(e).__name__)
-        try:
-            return ast.unparse(e)
-        except Exception:
-            return type(e).__name__
-    for node in ast.walk(func_node):
-        if isinstance(node, ast.Return):
-            t = _expr_type(node.value)
-            if found_type is None:
-                found_type = t
-            elif found_type != t:
-                raise ValueError("Conflicting return types:"
-                                 f"{found_type} vs {t}")
-    return found_type or "None"

pythonbpf/globals_pass.py

@@ -1,8 +1,121 @@
 from llvmlite import ir
 import ast
 
+from logging import Logger
+import logging
+from .type_deducer import ctypes_to_ir
 
-def emit_globals(module: ir.Module, names: list[str]):
+logger: Logger = logging.getLogger(__name__)
+
+# TODO: this is going to be a huge fuck of a headache in the future.
+global_sym_tab = []
+
+
+def populate_global_symbol_table(tree, module: ir.Module):
+    for node in tree.body:
+        if isinstance(node, ast.FunctionDef):
+            for dec in node.decorator_list:
+                if (
+                    isinstance(dec, ast.Call)
+                    and isinstance(dec.func, ast.Name)
+                    and dec.func.id == "section"
+                    and len(dec.args) == 1
+                    and isinstance(dec.args[0], ast.Constant)
+                    and isinstance(dec.args[0].value, str)
+                ):
+                    global_sym_tab.append(node)
+                elif isinstance(dec, ast.Name) and dec.id == "bpfglobal":
+                    global_sym_tab.append(node)
+
+                elif isinstance(dec, ast.Name) and dec.id == "map":
+                    global_sym_tab.append(node)
+    return False
+
+
+def emit_global(module: ir.Module, node, name):
+    logger.info(f"global identifier {name} processing")
+    # deduce LLVM type from the annotated return
+    if not isinstance(node.returns, ast.Name):
+        raise ValueError(f"Unsupported return annotation {ast.dump(node.returns)}")
+    ty = ctypes_to_ir(node.returns.id)
+
+    # extract the return expression
+    # TODO: turn this return extractor into a generic function I can use everywhere.
+    ret_stmt = node.body[0]
+    if not isinstance(ret_stmt, ast.Return) or ret_stmt.value is None:
+        raise ValueError(f"Global '{name}' has no valid return")
+
+    init_val = ret_stmt.value
+
+    # simple constant like "return 0"
+    if isinstance(init_val, ast.Constant):
+        llvm_init = ir.Constant(ty, init_val.value)
+
+    # variable reference like "return SOME_CONST"
+    elif isinstance(init_val, ast.Name):
+        # need symbol resolution here, stub as 0 for now
+        raise ValueError(f"Name reference {init_val.id} not yet supported")
+
+    # constructor call like "return c_int64(0)" or dataclass(...)
+    elif isinstance(init_val, ast.Call):
+        if len(init_val.args) >= 1 and isinstance(init_val.args[0], ast.Constant):
+            llvm_init = ir.Constant(ty, init_val.args[0].value)
+        else:
+            logger.info("Defaulting to zero as no constant argument found")
+            llvm_init = ir.Constant(ty, 0)
+    else:
+        raise ValueError(f"Unsupported return expr {ast.dump(init_val)}")
+
+    gvar = ir.GlobalVariable(module, ty, name=name)
+    gvar.initializer = llvm_init
+    gvar.align = 8
+    gvar.linkage = "dso_local"
+    gvar.global_constant = False
+    return gvar
+
+
+def globals_processing(tree, module):
+    """Process stuff decorated with @bpf and @bpfglobal except license and return the section name"""
+    globals_sym_tab = []
+
+    for node in tree.body:
+        # Skip non-assignment and non-function nodes
+        if not (isinstance(node, ast.FunctionDef)):
+            continue
+
+        # Get the name based on node type
+        if isinstance(node, ast.FunctionDef):
+            name = node.name
+        else:
+            continue
+
+        # Check for duplicate names
+        if name in globals_sym_tab:
+            raise SyntaxError(f"ERROR: Global name '{name}' previously defined")
+        else:
+            globals_sym_tab.append(name)
+
+        if isinstance(node, ast.FunctionDef) and node.name != "LICENSE":
+            decorators = [
+                dec.id for dec in node.decorator_list if isinstance(dec, ast.Name)
+            ]
+            if "bpf" in decorators and "bpfglobal" in decorators:
+                if (
+                    len(node.body) == 1
+                    and isinstance(node.body[0], ast.Return)
+                    and node.body[0].value is not None
+                    and isinstance(
+                        node.body[0].value, (ast.Constant, ast.Name, ast.Call)
+                    )
+                ):
+                    emit_global(module, node, name)
+                else:
+                    raise SyntaxError(f"ERROR: Invalid syntax for {name} global")
+
+    return None
+
+
+def emit_llvm_compiler_used(module: ir.Module, names: list[str]):
     """
     Emit the @llvm.compiler.used global given a list of function/global names.
     """
@@ -24,7 +137,7 @@ def emit_globals(module: ir.Module, names: list[str]):
     gv.section = "llvm.metadata"
 
 
-def globals_processing(tree, module: ir.Module):
+def globals_list_creation(tree, module: ir.Module):
     collected = ["LICENSE"]
 
     for node in tree.body:
@@ -40,10 +153,11 @@ def globals_processing(tree, module: ir.Module):
                 ):
                     collected.append(node.name)
 
-                elif isinstance(dec, ast.Name) and dec.id == "bpfglobal":
-                    collected.append(node.name)
+                # NOTE: all globals other than
+                # elif isinstance(dec, ast.Name) and dec.id == "bpfglobal":
+                #     collected.append(node.name)
 
                 elif isinstance(dec, ast.Name) and dec.id == "map":
                     collected.append(node.name)
 
-    emit_globals(module, collected)
+    emit_llvm_compiler_used(module, collected)

pythonbpf/helper/__init__.py

@@ -0,0 +1,90 @@
+from .helper_registry import HelperHandlerRegistry
+from .helper_utils import reset_scratch_pool
+from .bpf_helper_handler import handle_helper_call, emit_probe_read_kernel_str_call
+from .helpers import (
+    ktime,
+    pid,
+    deref,
+    comm,
+    probe_read_str,
+    random,
+    probe_read,
+    smp_processor_id,
+    uid,
+    skb_store_bytes,
+    get_stack,
+    XDP_DROP,
+    XDP_PASS,
+)
+
+
+# Register the helper handler with expr module
+def _register_helper_handler():
+    """Register helper call handler with the expression evaluator"""
+    from pythonbpf.expr.expr_pass import CallHandlerRegistry
+
+    def helper_call_handler(
+        call, module, builder, func, local_sym_tab, map_sym_tab, structs_sym_tab
+    ):
+        """Check if call is a helper and handle it"""
+        import ast
+
+        # Check for direct helper calls (e.g., ktime(), print())
+        if isinstance(call.func, ast.Name):
+            if HelperHandlerRegistry.has_handler(call.func.id):
+                return handle_helper_call(
+                    call,
+                    module,
+                    builder,
+                    func,
+                    local_sym_tab,
+                    map_sym_tab,
+                    structs_sym_tab,
+                )
+
+        # Check for method calls (e.g., map.lookup())
+        elif isinstance(call.func, ast.Attribute):
+            method_name = call.func.attr
+
+            # Handle: my_map.lookup(key)
+            if isinstance(call.func.value, ast.Name):
+                obj_name = call.func.value.id
+                if map_sym_tab and obj_name in map_sym_tab:
+                    if HelperHandlerRegistry.has_handler(method_name):
+                        return handle_helper_call(
+                            call,
+                            module,
+                            builder,
+                            func,
+                            local_sym_tab,
+                            map_sym_tab,
+                            structs_sym_tab,
+                        )
+
+        return None
+
+    CallHandlerRegistry.set_handler(helper_call_handler)
+
+
+# Register on module import
+_register_helper_handler()
+
+__all__ = [
+    "HelperHandlerRegistry",
+    "reset_scratch_pool",
+    "handle_helper_call",
+    "emit_probe_read_kernel_str_call",
+    "ktime",
+    "pid",
+    "deref",
+    "comm",
+    "probe_read_str",
+    "random",
+    "probe_read",
+    "smp_processor_id",
+    "uid",
+    "skb_store_bytes",
+    "get_stack",
+    "XDP_DROP",
+    "XDP_PASS",
+]

pythonbpf/helper/helper_registry.py

@@ -0,0 +1,61 @@
+from dataclasses import dataclass
+from llvmlite import ir
+from typing import Callable
+
+
+@dataclass
+class HelperSignature:
+    """Signature of a BPF helper function"""
+
+    arg_types: list[ir.Type]
+    return_type: ir.Type
+    func: Callable
+
+
+class HelperHandlerRegistry:
+    """Registry for BPF helpers"""
+
+    _handlers: dict[str, HelperSignature] = {}
+
+    @classmethod
+    def register(cls, helper_name, param_types=None, return_type=None):
+        """Decorator to register a handler function for a helper"""
+
+        def decorator(func):
+            helper_sig = HelperSignature(
+                arg_types=param_types, return_type=return_type, func=func
+            )
+            cls._handlers[helper_name] = helper_sig
+            return func
+
+        return decorator
+
+    @classmethod
+    def get_handler(cls, helper_name):
+        """Get the handler function for a helper"""
+        handler = cls._handlers.get(helper_name)
+        return handler.func if handler else None
+
+    @classmethod
+    def has_handler(cls, helper_name):
+        """Check if a handler function is registered for a helper"""
+        return helper_name in cls._handlers
+
+    @classmethod
+    def get_signature(cls, helper_name):
+        """Get the signature of a helper function"""
+        return cls._handlers.get(helper_name)
+
+    @classmethod
+    def get_param_type(cls, helper_name, index):
+        """Get the type of a parameter of a helper function by the index"""
+        signature = cls.get_signature(helper_name)
+        if signature and signature.arg_types and 0 <= index < len(signature.arg_types):
+            return signature.arg_types[index]
+        return None
+
+    @classmethod
+    def get_return_type(cls, helper_name):
+        """Get the return type of a helper function"""
+        signature = cls.get_signature(helper_name)
+        return signature.return_type if signature else None

pythonbpf/helper/helper_utils.py

@@ -0,0 +1,372 @@
+import ast
+import logging
+
+from llvmlite import ir
+from pythonbpf.expr import (
+    get_operand_value,
+    eval_expr,
+)
+
+logger = logging.getLogger(__name__)
+
+
+class ScratchPoolManager:
+    """Manage the temporary helper variables in local_sym_tab"""
+
+    def __init__(self):
+        self._counters = {}
+
+    @property
+    def counter(self):
+        return sum(self._counters.values())
+
+    def reset(self):
+        self._counters.clear()
+        logger.debug("Scratch pool counter reset to 0")
+
+    def _get_type_name(self, ir_type):
+        if isinstance(ir_type, ir.PointerType):
+            return "ptr"
+        elif isinstance(ir_type, ir.IntType):
+            return f"i{ir_type.width}"
+        elif isinstance(ir_type, ir.ArrayType):
+            return f"[{ir_type.count}x{self._get_type_name(ir_type.element)}]"
+        else:
+            return str(ir_type).replace(" ", "")
+
+    def get_next_temp(self, local_sym_tab, expected_type=None):
+        # Default to i64 if no expected type provided
+        type_name = self._get_type_name(expected_type) if expected_type else "i64"
+        if type_name not in self._counters:
+            self._counters[type_name] = 0
+
+        counter = self._counters[type_name]
+        temp_name = f"__helper_temp_{type_name}_{counter}"
+        self._counters[type_name] += 1
+
+        if temp_name not in local_sym_tab:
+            raise ValueError(
+                f"Scratch pool exhausted or inadequate: {temp_name}. "
+                f"Type: {type_name} Counter: {counter}"
+            )
+
+        logger.debug(f"Using {temp_name} for type {type_name}")
+        return local_sym_tab[temp_name].var, temp_name
+
+
+_temp_pool_manager = ScratchPoolManager()  # Singleton instance
+
+
+def reset_scratch_pool():
+    """Reset the scratch pool counter"""
+    _temp_pool_manager.reset()
+
+
+# ============================================================================
+# Argument Preparation
+# ============================================================================
+
+
+def get_var_ptr_from_name(var_name, local_sym_tab):
+    """Get a pointer to a variable from the symbol table."""
+    if local_sym_tab and var_name in local_sym_tab:
+        return local_sym_tab[var_name].var
+    raise ValueError(f"Variable '{var_name}' not found in local symbol table")
+
+
+def create_int_constant_ptr(value, builder, local_sym_tab, int_width=64):
+    """Create a pointer to an integer constant."""
+
+    int_type = ir.IntType(int_width)
+    ptr, temp_name = _temp_pool_manager.get_next_temp(local_sym_tab, int_type)
+    logger.info(f"Using temp variable '{temp_name}' for int constant {value}")
+    const_val = ir.Constant(int_type, value)
+    builder.store(const_val, ptr)
+    return ptr
+
+
+def get_or_create_ptr_from_arg(
+    func,
+    module,
+    arg,
+    builder,
+    local_sym_tab,
+    map_sym_tab,
+    struct_sym_tab=None,
+    expected_type=None,
+):
+    """Extract or create pointer from the call arguments."""
+
+    logger.info(f"Getting pointer from arg: {ast.dump(arg)}")
+    sz = None
+    if isinstance(arg, ast.Name):
+        # Stack space is already allocated
+        ptr = get_var_ptr_from_name(arg.id, local_sym_tab)
+    elif isinstance(arg, ast.Constant) and isinstance(arg.value, int):
+        int_width = 64  # Default to i64
+        if expected_type and isinstance(expected_type, ir.IntType):
+            int_width = expected_type.width
+        ptr = create_int_constant_ptr(arg.value, builder, local_sym_tab, int_width)
+    elif isinstance(arg, ast.Attribute):
+        # A struct field
+        struct_name = arg.value.id
+        field_name = arg.attr
+
+        if not local_sym_tab or struct_name not in local_sym_tab:
+            raise ValueError(f"Struct '{struct_name}' not found")
+
+        struct_type = local_sym_tab[struct_name].metadata
+        if not struct_sym_tab or struct_type not in struct_sym_tab:
+            raise ValueError(f"Struct type '{struct_type}' not found")
+
+        struct_info = struct_sym_tab[struct_type]
+        if field_name not in struct_info.fields:
+            raise ValueError(
+                f"Field '{field_name}' not found in struct '{struct_name}'"
+            )
+
+        field_type = struct_info.field_type(field_name)
+        struct_ptr = local_sym_tab[struct_name].var
+
+        # Special handling for char arrays
+        if (
+            isinstance(field_type, ir.ArrayType)
+            and isinstance(field_type.element, ir.IntType)
+            and field_type.element.width == 8
+        ):
+            ptr, sz = get_char_array_ptr_and_size(
+                arg, builder, local_sym_tab, struct_sym_tab
+            )
+            if not ptr:
+                raise ValueError("Failed to get char array pointer from struct field")
+        else:
+            ptr = struct_info.gep(builder, struct_ptr, field_name)
+
+    else:
+        # NOTE: For any integer expression reaching this branch, it is probably a struct field or a binop
+        # Evaluate the expression and store the result in a temp variable
+        val = get_operand_value(
+            func, module, arg, builder, local_sym_tab, map_sym_tab, struct_sym_tab
+        )
+        if val is None:
+            raise ValueError("Failed to evaluate expression for helper arg.")
+
+        ptr, temp_name = _temp_pool_manager.get_next_temp(local_sym_tab, expected_type)
+        logger.info(f"Using temp variable '{temp_name}' for expression result")
+        if (
+            isinstance(val.type, ir.IntType)
+            and expected_type
+            and val.type.width > expected_type.width
+        ):
+            val = builder.trunc(val, expected_type)
+        builder.store(val, ptr)
+
+    # NOTE: For char arrays, also return size
+    if sz:
+        return ptr, sz
+
+    return ptr
+
+
+def get_flags_val(arg, builder, local_sym_tab):
+    """Extract or create flags value from the call arguments."""
+    if not arg:
+        return 0
+
+    if isinstance(arg, ast.Name):
+        if local_sym_tab and arg.id in local_sym_tab:
+            flags_ptr = local_sym_tab[arg.id].var
+            return builder.load(flags_ptr)
+        else:
+            raise ValueError(f"Variable '{arg.id}' not found in local symbol table")
+    elif isinstance(arg, ast.Constant) and isinstance(arg.value, int):
+        return arg.value
+
+    raise NotImplementedError(
+        "Only var names or int consts are supported as map helpers flags."
+    )
+
+
+def get_data_ptr_and_size(data_arg, local_sym_tab, struct_sym_tab):
+    """Extract data pointer and size information for perf event output."""
+    if isinstance(data_arg, ast.Name):
+        data_name = data_arg.id
+        if local_sym_tab and data_name in local_sym_tab:
+            data_ptr = local_sym_tab[data_name].var
+        else:
+            raise ValueError(
+                f"Data variable {data_name} not found in local symbol table."
+            )
+
+        # Check if data_name is a struct
+        data_type = local_sym_tab[data_name].metadata
+        if data_type in struct_sym_tab:
+            struct_info = struct_sym_tab[data_type]
+            size_val = ir.Constant(ir.IntType(64), struct_info.size)
+            return data_ptr, size_val
+        else:
+            raise ValueError(f"Struct {data_type} for {data_name} not in symbol table.")
+    else:
+        raise NotImplementedError(
+            "Only simple object names are supported as data in perf event output."
+        )
+
+
+def get_buffer_ptr_and_size(buf_arg, builder, local_sym_tab, struct_sym_tab):
+    """Extract buffer pointer and size from either a struct field or variable."""
+
+    # Case 1: Struct field (obj.field)
+    if isinstance(buf_arg, ast.Attribute):
+        if not isinstance(buf_arg.value, ast.Name):
+            raise ValueError(
+                "Only simple struct field access supported (e.g., obj.field)"
+            )
+
+        struct_name = buf_arg.value.id
+        field_name = buf_arg.attr
+
+        # Lookup struct
+        if not local_sym_tab or struct_name not in local_sym_tab:
+            raise ValueError(f"Struct '{struct_name}' not found")
+
+        struct_type = local_sym_tab[struct_name].metadata
+        if not struct_sym_tab or struct_type not in struct_sym_tab:
+            raise ValueError(f"Struct type '{struct_type}' not found")
+
+        struct_info = struct_sym_tab[struct_type]
+
+        # Get field pointer and type
+        struct_ptr = local_sym_tab[struct_name].var
+        field_ptr = struct_info.gep(builder, struct_ptr, field_name)
+        field_type = struct_info.field_type(field_name)
+
+        if not isinstance(field_type, ir.ArrayType):
+            raise ValueError(f"Field '{field_name}' must be an array type")
+
+        return field_ptr, field_type.count
+
+    # Case 2: Variable name
+    elif isinstance(buf_arg, ast.Name):
+        var_name = buf_arg.id
+
+        if not local_sym_tab or var_name not in local_sym_tab:
+            raise ValueError(f"Variable '{var_name}' not found")
+
+        var_ptr = local_sym_tab[var_name].var
+        var_type = local_sym_tab[var_name].ir_type
+
+        if not isinstance(var_type, ir.ArrayType):
+            raise ValueError(f"Variable '{var_name}' must be an array type")
+
+        return var_ptr, var_type.count
+
+    else:
+        raise ValueError(
+            "comm expects either a struct field (obj.field) or variable name"
+        )
+
+
+def get_char_array_ptr_and_size(buf_arg, builder, local_sym_tab, struct_sym_tab):
+    """Get pointer to char array and its size."""
+
+    # Struct field: obj.field
+    if isinstance(buf_arg, ast.Attribute) and isinstance(buf_arg.value, ast.Name):
+        var_name = buf_arg.value.id
+        field_name = buf_arg.attr
+
+        if not (local_sym_tab and var_name in local_sym_tab):
+            raise ValueError(f"Variable '{var_name}' not found")
+
+        struct_type = local_sym_tab[var_name].metadata
+        if not (struct_sym_tab and struct_type in struct_sym_tab):
+            raise ValueError(f"Struct type '{struct_type}' not found")
+
+        struct_info = struct_sym_tab[struct_type]
+        if field_name not in struct_info.fields:
+            raise ValueError(f"Field '{field_name}' not found")
+
+        field_type = struct_info.field_type(field_name)
+        if not _is_char_array(field_type):
+            logger.info(
+                "Field is not a char array, falling back to int or ptr detection"
+            )
+            return None, 0
+
+        struct_ptr = local_sym_tab[var_name].var
+        field_ptr = struct_info.gep(builder, struct_ptr, field_name)
+
+        # GEP to first element: [N x i8]* -> i8*
+        buf_ptr = builder.gep(
+            field_ptr,
+            [ir.Constant(ir.IntType(32), 0), ir.Constant(ir.IntType(32), 0)],
+            inbounds=True,
+        )
+        return buf_ptr, field_type.count
+
+    elif isinstance(buf_arg, ast.Name):
+        # NOTE: We shouldn't be doing this as we can't get size info
+        var_name = buf_arg.id
+        if not (local_sym_tab and var_name in local_sym_tab):
+            raise ValueError(f"Variable '{var_name}' not found")
+
+        var_ptr = local_sym_tab[var_name].var
+        var_type = local_sym_tab[var_name].ir_type
+
+        if not isinstance(var_type, ir.PointerType) or not isinstance(
+            var_type.pointee, ir.IntType(8)
+        ):
+            raise ValueError("Expected str ptr variable")
+
+        return var_ptr, 256  # Size unknown for str ptr, using 256 as default
+
+    else:
+        raise ValueError("Expected struct field or variable name")
+
+
+def _is_char_array(ir_type):
+    """Check if IR type is [N x i8]."""
+    return (
+        isinstance(ir_type, ir.ArrayType)
+        and isinstance(ir_type.element, ir.IntType)
+        and ir_type.element.width == 8
+    )
+
+
+def get_ptr_from_arg(
+    arg, func, module, builder, local_sym_tab, map_sym_tab, struct_sym_tab
+):
+    """Evaluate argument and return pointer value"""
+
+    result = eval_expr(
+        func, module, builder, arg, local_sym_tab, map_sym_tab, struct_sym_tab
+    )
+
+    if not result:
+        raise ValueError("Failed to evaluate argument")
+
+    val, val_type = result
+
+    if not isinstance(val_type, ir.PointerType):
+        raise ValueError(f"Expected pointer type, got {val_type}")
+
+    return val, val_type
+
+
+def get_int_value_from_arg(
+    arg, func, module, builder, local_sym_tab, map_sym_tab, struct_sym_tab
+):
+    """Evaluate argument and return integer value"""
+
+    result = eval_expr(
+        func, module, builder, arg, local_sym_tab, map_sym_tab, struct_sym_tab
+    )
+
+    if not result:
+        raise ValueError("Failed to evaluate argument")
+
+    val, val_type = result
+
+    if not isinstance(val_type, ir.IntType):
+        raise ValueError(f"Expected integer type, got {val_type}")
+
+    return val

pythonbpf/helper/helpers.py

@@ -0,0 +1,64 @@
+import ctypes
+
+
+def ktime():
+    """get current ktime"""
+    return ctypes.c_int64(0)
+
+
+def pid():
+    """get current process id"""
+    return ctypes.c_int32(0)
+
+
+def deref(ptr):
+    """dereference a pointer"""
+    result = ctypes.cast(ptr, ctypes.POINTER(ctypes.c_void_p)).contents.value
+    return result if result is not None else 0
+
+
+def comm(buf):
+    """get current process command name"""
+    return ctypes.c_int64(0)
+
+
+def probe_read_str(dst, src):
+    """Safely read a null-terminated string from kernel memory"""
+    return ctypes.c_int64(0)
+
+
+def random():
+    """get a pseudorandom u32 number"""
+    return ctypes.c_int32(0)
+
+
+def probe_read(dst, size, src):
+    """Safely read data from kernel memory"""
+    return ctypes.c_int64(0)
+
+
+def smp_processor_id():
+    """get the current CPU id"""
+    return ctypes.c_int32(0)
+
+
+def uid():
+    """get current user id"""
+    return ctypes.c_int32(0)
+
+
+def skb_store_bytes(offset, from_buf, size, flags=0):
+    """store bytes into a socket buffer"""
+    return ctypes.c_int64(0)
+
+
+def get_stack(buf, flags=0):
+    """get the current stack trace"""
+    return ctypes.c_int64(0)
+
+
+XDP_ABORTED = ctypes.c_int64(0)
+XDP_DROP = ctypes.c_int64(1)
+XDP_PASS = ctypes.c_int64(2)
+XDP_TX = ctypes.c_int64(3)
+XDP_REDIRECT = ctypes.c_int64(4)

pythonbpf/helper/printk_formatter.py

@@ -0,0 +1,272 @@
+import ast
+import logging
+
+from llvmlite import ir
+from pythonbpf.expr import eval_expr, get_base_type_and_depth, deref_to_depth
+from pythonbpf.expr.vmlinux_registry import VmlinuxHandlerRegistry
+from pythonbpf.helper.helper_utils import get_char_array_ptr_and_size
+
+logger = logging.getLogger(__name__)
+
+
+def simple_string_print(string_value, module, builder, func):
+    """Prepare arguments for bpf_printk from a simple string value"""
+    fmt_str = string_value + "\n\0"
+    fmt_ptr = _create_format_string_global(fmt_str, func, module, builder)
+
+    args = [fmt_ptr, ir.Constant(ir.IntType(32), len(fmt_str))]
+    return args
+
+
+def handle_fstring_print(
+    joined_str,
+    module,
+    builder,
+    func,
+    local_sym_tab=None,
+    struct_sym_tab=None,
+):
+    """Handle f-string formatting for bpf_printk emitter."""
+    fmt_parts = []
+    exprs = []
+
+    for value in joined_str.values:
+        logger.debug(f"Processing f-string value: {ast.dump(value)}")
+
+        if isinstance(value, ast.Constant):
+            _process_constant_in_fstring(value, fmt_parts, exprs)
+        elif isinstance(value, ast.FormattedValue):
+            _process_fval(
+                value,
+                fmt_parts,
+                exprs,
+                local_sym_tab,
+                struct_sym_tab,
+            )
+        else:
+            raise NotImplementedError(f"Unsupported f-string value type: {type(value)}")
+
+    fmt_str = "".join(fmt_parts)
+    args = simple_string_print(fmt_str, module, builder, func)
+
+    # NOTE: Process expressions (limited to 3 due to BPF constraints)
+    if len(exprs) > 3:
+        logger.warning("bpf_printk supports up to 3 args, extra args will be ignored.")
+
+    for expr in exprs[:3]:
+        arg_value = _prepare_expr_args(
+            expr,
+            func,
+            module,
+            builder,
+            local_sym_tab,
+            struct_sym_tab,
+        )
+        args.append(arg_value)
+
+    return args
+
+
+# ============================================================================
+# Internal Helpers
+# ============================================================================
+
+
+def _process_constant_in_fstring(cst, fmt_parts, exprs):
+    """Process constant values in f-string."""
+    if isinstance(cst.value, str):
+        fmt_parts.append(cst.value)
+    elif isinstance(cst.value, int):
+        fmt_parts.append("%lld")
+        exprs.append(ir.Constant(ir.IntType(64), cst.value))
+    else:
+        raise NotImplementedError(
+            f"Unsupported constant type in f-string: {type(cst.value)}"
+        )
+
+
+def _process_fval(fval, fmt_parts, exprs, local_sym_tab, struct_sym_tab):
+    """Process formatted values in f-string."""
+    logger.debug(f"Processing formatted value: {ast.dump(fval)}")
+
+    if isinstance(fval.value, ast.Name):
+        _process_name_in_fval(fval.value, fmt_parts, exprs, local_sym_tab)
+    elif isinstance(fval.value, ast.Attribute):
+        _process_attr_in_fval(
+            fval.value,
+            fmt_parts,
+            exprs,
+            local_sym_tab,
+            struct_sym_tab,
+        )
+    else:
+        raise NotImplementedError(
+            f"Unsupported formatted value in f-string: {type(fval.value)}"
+        )
+
+
+def _process_name_in_fval(name_node, fmt_parts, exprs, local_sym_tab):
+    """Process name nodes in formatted values."""
+    if local_sym_tab and name_node.id in local_sym_tab:
+        _, var_type, tmp = local_sym_tab[name_node.id]
+        _populate_fval(var_type, name_node, fmt_parts, exprs)
+    else:
+        # Try to resolve through vmlinux registry if not in local symbol table
+        result = VmlinuxHandlerRegistry.handle_name(name_node.id)
+        if result:
+            val, var_type = result
+            _populate_fval(var_type, name_node, fmt_parts, exprs)
+        else:
+            raise ValueError(
+                f"Variable '{name_node.id}' not found in symbol table or vmlinux"
+            )
+
+
+def _process_attr_in_fval(attr_node, fmt_parts, exprs, local_sym_tab, struct_sym_tab):
+    """Process attribute nodes in formatted values."""
+    if (
+        isinstance(attr_node.value, ast.Name)
+        and local_sym_tab
+        and attr_node.value.id in local_sym_tab
+    ):
+        var_name = attr_node.value.id
+        field_name = attr_node.attr
+
+        var_type = local_sym_tab[var_name].metadata
+        if var_type not in struct_sym_tab:
+            raise ValueError(
+                f"Struct '{var_type}' for '{var_name}' not in symbol table"
+            )
+
+        struct_info = struct_sym_tab[var_type]
+        if field_name not in struct_info.fields:
+            raise ValueError(f"Field '{field_name}' not found in struct '{var_type}'")
+
+        field_type = struct_info.field_type(field_name)
+        _populate_fval(field_type, attr_node, fmt_parts, exprs)
+    else:
+        raise NotImplementedError(
+            "Only simple attribute on local vars is supported in f-strings."
+        )
+
+
+def _populate_fval(ftype, node, fmt_parts, exprs):
+    """Populate format parts and expressions based on field type."""
+    if isinstance(ftype, ir.IntType):
+        # TODO: We print as signed integers only for now
+        if ftype.width == 64:
+            fmt_parts.append("%lld")
+            exprs.append(node)
+        elif ftype.width == 32:
+            fmt_parts.append("%d")
+            exprs.append(node)
+        else:
+            raise NotImplementedError(
+                f"Unsupported integer width in f-string: {ftype.width}"
+            )
+    elif isinstance(ftype, ir.PointerType):
+        target, depth = get_base_type_and_depth(ftype)
+        if isinstance(target, ir.IntType):
+            if target.width == 64:
+                fmt_parts.append("%lld")
+                exprs.append(node)
+            elif target.width == 32:
+                fmt_parts.append("%d")
+                exprs.append(node)
+            elif target.width == 8 and depth == 1:
+                # NOTE: Assume i8* is a string
+                fmt_parts.append("%s")
+                exprs.append(node)
+            else:
+                raise NotImplementedError(
+                    f"Unsupported pointer target type in f-string: {target}"
+                )
+        else:
+            raise NotImplementedError(
+                f"Unsupported pointer target type in f-string: {target}"
+            )
+    elif isinstance(ftype, ir.ArrayType):
+        if isinstance(ftype.element, ir.IntType) and ftype.element.width == 8:
+            # Char array
+            fmt_parts.append("%s")
+            exprs.append(node)
+        else:
+            raise NotImplementedError(
+                f"Unsupported array element type in f-string: {ftype.element}"
+            )
+    else:
+        raise NotImplementedError(f"Unsupported field type in f-string: {ftype}")
+
+
+def _create_format_string_global(fmt_str, func, module, builder):
+    """Create a global variable for the format string."""
+    fmt_name = f"{func.name}____fmt{func._fmt_counter}"
+    func._fmt_counter += 1
+
+    fmt_gvar = ir.GlobalVariable(
+        module, ir.ArrayType(ir.IntType(8), len(fmt_str)), name=fmt_name
+    )
+    fmt_gvar.global_constant = True
+    fmt_gvar.initializer = ir.Constant(
+        ir.ArrayType(ir.IntType(8), len(fmt_str)), bytearray(fmt_str.encode("utf8"))
+    )
+    fmt_gvar.linkage = "internal"
+    fmt_gvar.align = 1
+
+    return builder.bitcast(fmt_gvar, ir.PointerType())
+
+
+def _prepare_expr_args(expr, func, module, builder, local_sym_tab, struct_sym_tab):
+    """Evaluate and prepare an expression to use as an arg for bpf_printk."""
+
+    # Special case: struct field char array needs pointer to first element
+    if isinstance(expr, ast.Attribute):
+        char_array_ptr, _ = get_char_array_ptr_and_size(
+            expr, builder, local_sym_tab, struct_sym_tab
+        )
+        if char_array_ptr:
+            return char_array_ptr
+
+    # Regular expression evaluation
+    val, _ = eval_expr(func, module, builder, expr, local_sym_tab, None, struct_sym_tab)
+
+    if not val:
+        logger.warning("Failed to evaluate expression for bpf_printk, defaulting to 0")
+        return ir.Constant(ir.IntType(64), 0)
+
+    # Convert value to bpf_printk compatible type
+    if isinstance(val.type, ir.PointerType):
+        return _handle_pointer_arg(val, func, builder)
+    elif isinstance(val.type, ir.IntType):
+        return _handle_int_arg(val, builder)
+    else:
+        logger.warning(f"Unsupported type {val.type} in bpf_printk, defaulting to 0")
+        return ir.Constant(ir.IntType(64), 0)
+
+
+def _handle_pointer_arg(val, func, builder):
+    """Convert pointer type for bpf_printk."""
+    target, depth = get_base_type_and_depth(val.type)
+
+    if not isinstance(target, ir.IntType):
+        logger.warning("Only int pointers supported in bpf_printk, defaulting to 0")
+        return ir.Constant(ir.IntType(64), 0)
+
+    # i8* is string - use as-is
+    if target.width == 8 and depth == 1:
+        return val
+
+    # Integer pointers: dereference and sign-extend to i64
+    if target.width >= 32:
+        val = deref_to_depth(func, builder, val, depth)
+        return builder.sext(val, ir.IntType(64))
+
+    logger.warning("Unsupported pointer width in bpf_printk, defaulting to 0")
+    return ir.Constant(ir.IntType(64), 0)
+
+
+def _handle_int_arg(val, builder):
+    """Convert integer type for bpf_printk (sign-extend to i64)."""
+    if val.type.width < 64:
+        return builder.sext(val, ir.IntType(64))
+    return val

pythonbpf/helpers.py

@@ -1,4 +0,0 @@
-import ctypes
-
-def bpf_ktime_get_ns():
-    return ctypes.c_int64(0)

pythonbpf/license_pass.py

@@ -1,5 +1,9 @@
 from llvmlite import ir
 import ast
+from logging import Logger
+import logging
+
+logger: Logger = logging.getLogger(__name__)
 
 
 def emit_license(module: ir.Module, license_str: str):
@@ -11,10 +15,10 @@ def emit_license(module: ir.Module, license_str: str):
 
     gvar.initializer = ir.Constant(ty, elems)  # type: ignore
 
-    gvar.align = 1                      # type: ignore
-    gvar.linkage = "dso_local"          # type: ignore
+    gvar.align = 1  # type: ignore
+    gvar.linkage = "dso_local"  # type: ignore
     gvar.global_constant = False
-    gvar.section = "license"            # type: ignore
+    gvar.section = "license"  # type: ignore
 
     return gvar
 
@@ -26,7 +30,8 @@ def license_processing(tree, module):
         if isinstance(node, ast.FunctionDef) and node.name == "LICENSE":
             # check decorators
             decorators = [
-                dec.id for dec in node.decorator_list if isinstance(dec, ast.Name)]
+                dec.id for dec in node.decorator_list if isinstance(dec, ast.Name)
+            ]
             if "bpf" in decorators and "bpfglobal" in decorators:
                 if count == 0:
                     count += 1
@@ -40,9 +45,9 @@ def license_processing(tree, module):
                         emit_license(module, node.body[0].value.value)
                         return "LICENSE"
                     else:
-                        print("ERROR: LICENSE() must return a string literal")
+                        logger.info("ERROR: LICENSE() must return a string literal")
                         return None
                 else:
-                    print("ERROR: LICENSE already defined")
+                    logger.info("ERROR: LICENSE already defined")
                     return None
     return None

pythonbpf/local_symbol.py

@@ -0,0 +1,15 @@
+import llvmlite.ir as ir
+from dataclasses import dataclass
+from typing import Any
+
+
+@dataclass
+class LocalSymbol:
+    var: ir.AllocaInstr
+    ir_type: ir.Type
+    metadata: Any = None
+
+    def __iter__(self):
+        yield self.var
+        yield self.ir_type
+        yield self.metadata

pythonbpf/maps.py

@@ -1,24 +0,0 @@
-class HashMap:
-    def __init__(self, key_type, value_type, max_entries):
-        self.key_type = key_type
-        self.value_type = value_type
-        self.max_entries = max_entries
-        self.entries = {}
-
-    def lookup(self, key):
-        if key in self.entries:
-            return self.entries[key]
-        else:
-            return None
-            
-    def delete(self, key):
-        if key in self.entries:
-            del self.entries[key]
-        else:
-            raise KeyError(f"Key {key} not found in map")
-            
-    def update(self, key, value):
-        if key in self.entries:
-            self.entries[key] = value
-        else:
-            raise KeyError(f"Key {key} not found in map")

pythonbpf/maps/__init__.py

@@ -0,0 +1,5 @@
+from .maps import HashMap, PerfEventArray, RingBuffer
+from .maps_pass import maps_proc
+from .map_types import BPFMapType
+
+__all__ = ["HashMap", "PerfEventArray", "maps_proc", "RingBuffer", "BPFMapType"]

pythonbpf/maps/map_debug_info.py

@@ -0,0 +1,99 @@
+from pythonbpf.debuginfo import DebugInfoGenerator
+from .map_types import BPFMapType
+
+
+def create_map_debug_info(module, map_global, map_name, map_params, structs_sym_tab):
+    """Generate debug info metadata for BPF maps HASH and PERF_EVENT_ARRAY"""
+    generator = DebugInfoGenerator(module)
+
+    uint_type = generator.get_uint32_type()
+    ulong_type = generator.get_uint64_type()
+    array_type = generator.create_array_type(
+        uint_type, map_params.get("type", BPFMapType.UNSPEC).value
+    )
+    type_ptr = generator.create_pointer_type(array_type, 64)
+    key_ptr = generator.create_pointer_type(
+        array_type if "key_size" in map_params else ulong_type, 64
+    )
+    value_ptr = generator.create_pointer_type(
+        array_type if "value_size" in map_params else ulong_type, 64
+    )
+
+    elements_arr = []
+
+    # Create struct members
+    # scope field does not appear for some reason
+    cnt = 0
+    for elem in map_params:
+        if elem == "max_entries":
+            continue
+        if elem == "type":
+            ptr = type_ptr
+        elif "key" in elem:
+            ptr = key_ptr
+        else:
+            ptr = value_ptr
+        # TODO: the best way to do this is not 64, but get the size each time. this will not work for structs.
+        member = generator.create_struct_member(elem, ptr, cnt * 64)
+        elements_arr.append(member)
+        cnt += 1
+
+    if "max_entries" in map_params:
+        max_entries_array = generator.create_array_type(
+            uint_type, map_params["max_entries"]
+        )
+        max_entries_ptr = generator.create_pointer_type(max_entries_array, 64)
+        max_entries_member = generator.create_struct_member(
+            "max_entries", max_entries_ptr, cnt * 64
+        )
+        elements_arr.append(max_entries_member)
+
+    # Create the struct type
+    struct_type = generator.create_struct_type(
+        elements_arr, 64 * len(elements_arr), is_distinct=True
+    )
+
+    # Create global variable debug info
+    global_var = generator.create_global_var_debug_info(
+        map_name, struct_type, is_local=False
+    )
+
+    # Attach debug info to the global variable
+    map_global.set_metadata("dbg", global_var)
+
+    return global_var
+
+
+# TODO: This should not be exposed outside of the module.
+# Ideally we should expose a single create_map_debug_info function that handles all map types.
+# We can probably use a registry pattern to register different map types and their debug info generators.
+# map_params["type"] will be used to determine which generator to use.
+def create_ringbuf_debug_info(
+    module, map_global, map_name, map_params, structs_sym_tab
+):
+    """Generate debug information metadata for BPF RINGBUF map"""
+    generator = DebugInfoGenerator(module)
+
+    int_type = generator.get_int32_type()
+
+    type_array = generator.create_array_type(
+        int_type, map_params.get("type", BPFMapType.RINGBUF).value
+    )
+    type_ptr = generator.create_pointer_type(type_array, 64)
+    type_member = generator.create_struct_member("type", type_ptr, 0)
+
+    max_entries_array = generator.create_array_type(int_type, map_params["max_entries"])
+    max_entries_ptr = generator.create_pointer_type(max_entries_array, 64)
+    max_entries_member = generator.create_struct_member(
+        "max_entries", max_entries_ptr, 64
+    )
+
+    elements_arr = [type_member, max_entries_member]
+
+    struct_type = generator.create_struct_type(elements_arr, 128, is_distinct=True)
+
+    global_var = generator.create_global_var_debug_info(
+        map_name, struct_type, is_local=False
+    )
+    map_global.set_metadata("dbg", global_var)
+    return global_var

pythonbpf/maps/map_types.py

@@ -0,0 +1,39 @@
+from enum import Enum
+
+
+class BPFMapType(Enum):
+    UNSPEC = 0
+    HASH = 1
+    ARRAY = 2
+    PROG_ARRAY = 3
+    PERF_EVENT_ARRAY = 4
+    PERCPU_HASH = 5
+    PERCPU_ARRAY = 6
+    STACK_TRACE = 7
+    CGROUP_ARRAY = 8
+    LRU_HASH = 9
+    LRU_PERCPU_HASH = 10
+    LPM_TRIE = 11
+    ARRAY_OF_MAPS = 12
+    HASH_OF_MAPS = 13
+    DEVMAP = 14
+    SOCKMAP = 15
+    CPUMAP = 16
+    XSKMAP = 17
+    SOCKHASH = 18
+    CGROUP_STORAGE_DEPRECATED = 19
+    CGROUP_STORAGE = 19
+    REUSEPORT_SOCKARRAY = 20
+    PERCPU_CGROUP_STORAGE_DEPRECATED = 21
+    PERCPU_CGROUP_STORAGE = 21
+    QUEUE = 22
+    STACK = 23
+    SK_STORAGE = 24
+    DEVMAP_HASH = 25
+    STRUCT_OPS = 26
+    RINGBUF = 27
+    INODE_STORAGE = 28
+    TASK_STORAGE = 29
+    BLOOM_FILTER = 30
+    USER_RINGBUF = 31
+    CGRP_STORAGE = 32

pythonbpf/maps/maps.py

@@ -0,0 +1,57 @@
+# This file provides type  and function hints only and does not actually give any functionality.
+class HashMap:
+    def __init__(self, key, value, max_entries):
+        self.key = key
+        self.value = value
+        self.max_entries = max_entries
+        self.entries = {}
+
+    def lookup(self, key):
+        if key in self.entries:
+            return self.entries[key]
+        else:
+            return None
+
+    def delete(self, key):
+        if key in self.entries:
+            del self.entries[key]
+        else:
+            raise KeyError(f"Key {key} not found in map")
+
+    # TODO: define the flags that can be added
+    def update(self, key, value, flags=None):
+        if key in self.entries:
+            self.entries[key] = value
+        else:
+            raise KeyError(f"Key {key} not found in map")
+
+
+class PerfEventArray:
+    def __init__(self, key_size, value_size):
+        self.key_type = key_size
+        self.value_type = value_size
+        self.entries = {}
+
+    def output(self, data):
+        pass  # Placeholder for output method
+
+
+class RingBuffer:
+    def __init__(self, max_entries):
+        self.max_entries = max_entries
+
+    def output(self, data, flags=0):
+        pass
+
+    def reserve(self, size: int):
+        if size > self.max_entries:
+            raise ValueError("size cannot be greater than set maximum entries")
+        return 0
+
+    def submit(self, data, flags=0):
+        pass
+
+    def discard(self, data, flags=0):
+        pass
+
+    # add discard, output and also give names to flags and stuff

pythonbpf/maps/maps_pass.py

@@ -0,0 +1,164 @@
+import ast
+import logging
+from logging import Logger
+from llvmlite import ir
+
+from .maps_utils import MapProcessorRegistry, MapSymbol
+from .map_types import BPFMapType
+from .map_debug_info import create_map_debug_info, create_ringbuf_debug_info
+from pythonbpf.expr.vmlinux_registry import VmlinuxHandlerRegistry
+
+
+logger: Logger = logging.getLogger(__name__)
+
+
+def maps_proc(tree, module, chunks, structs_sym_tab):
+    """Process all functions decorated with @map to find BPF maps"""
+    map_sym_tab = {}
+    for func_node in chunks:
+        if is_map(func_node):
+            logger.info(f"Found BPF map: {func_node.name}")
+            map_sym_tab[func_node.name] = process_bpf_map(
+                func_node, module, structs_sym_tab
+            )
+    return map_sym_tab
+
+
+def is_map(func_node):
+    return any(
+        isinstance(decorator, ast.Name) and decorator.id == "map"
+        for decorator in func_node.decorator_list
+    )
+
+
+def create_bpf_map(module, map_name, map_params):
+    """Create a BPF map in the module with given parameters and debug info"""
+
+    # Create the anonymous struct type for BPF map
+    map_struct_type = ir.LiteralStructType(
+        [ir.PointerType() for _ in range(len(map_params))]
+    )
+
+    # Create the global variable
+    map_global = ir.GlobalVariable(module, map_struct_type, name=map_name)
+    map_global.linkage = "dso_local"
+    map_global.global_constant = False
+    map_global.initializer = ir.Constant(map_struct_type, None)
+    map_global.section = ".maps"
+    map_global.align = 8
+
+    logger.info(f"Created BPF map: {map_name} with params {map_params}")
+    return MapSymbol(type=map_params["type"], sym=map_global)
+
+
+def _parse_map_params(rval, expected_args=None):
+    """Parse map parameters from call arguments and keywords."""
+
+    params = {}
+    handler = VmlinuxHandlerRegistry.get_handler()
+    # Parse positional arguments
+    if expected_args:
+        for i, arg_name in enumerate(expected_args):
+            if i < len(rval.args):
+                arg = rval.args[i]
+                if isinstance(arg, ast.Name):
+                    result = _get_vmlinux_enum(handler, arg.id)
+                    params[arg_name] = result if result is not None else arg.id
+                elif isinstance(arg, ast.Constant):
+                    params[arg_name] = arg.value
+
+    # Parse keyword arguments (override positional)
+    for keyword in rval.keywords:
+        if isinstance(keyword.value, ast.Name):
+            name = keyword.value.id
+            result = _get_vmlinux_enum(handler, name)
+            params[keyword.arg] = result if result is not None else name
+        elif isinstance(keyword.value, ast.Constant):
+            params[keyword.arg] = keyword.value.value
+
+    return params
+
+
+def _get_vmlinux_enum(handler, name):
+    if handler and handler.is_vmlinux_enum(name):
+        return handler.get_vmlinux_enum_value(name)
+
+
+@MapProcessorRegistry.register("RingBuffer")
+def process_ringbuf_map(map_name, rval, module, structs_sym_tab):
+    """Process a BPF_RINGBUF map declaration"""
+    logger.info(f"Processing Ringbuf: {map_name}")
+    map_params = _parse_map_params(rval, expected_args=["max_entries"])
+    map_params["type"] = BPFMapType.RINGBUF
+
+    # NOTE: constraints borrowed from https://docs.ebpf.io/linux/map-type/BPF_MAP_TYPE_RINGBUF/
+    max_entries = map_params.get("max_entries")
+    if (
+        not isinstance(max_entries, int)
+        or max_entries < 4096
+        or (max_entries & (max_entries - 1)) != 0
+    ):
+        raise ValueError(
+            "Ringbuf max_entries must be a power of two greater than or equal to the page size (4096)"
+        )
+
+    logger.info(f"Ringbuf map parameters: {map_params}")
+
+    map_global = create_bpf_map(module, map_name, map_params)
+    create_ringbuf_debug_info(module, map_global.sym, map_name, map_params)
+    return map_global
+
+
+@MapProcessorRegistry.register("HashMap")
+def process_hash_map(map_name, rval, module, structs_sym_tab):
+    """Process a BPF_HASH map declaration"""
+    logger.info(f"Processing HashMap: {map_name}")
+    map_params = _parse_map_params(rval, expected_args=["key", "value", "max_entries"])
+    map_params["type"] = BPFMapType.HASH
+
+    logger.info(f"Map parameters: {map_params}")
+    map_global = create_bpf_map(module, map_name, map_params)
+    # Generate debug info for BTF
+    create_map_debug_info(module, map_global.sym, map_name, map_params)
+    return map_global
+
+
+@MapProcessorRegistry.register("PerfEventArray")
+def process_perf_event_map(map_name, rval, module, structs_sym_tab):
+    """Process a BPF_PERF_EVENT_ARRAY map declaration"""
+    logger.info(f"Processing PerfEventArray: {map_name}")
+    map_params = _parse_map_params(rval, expected_args=["key_size", "value_size"])
+    map_params["type"] = BPFMapType.PERF_EVENT_ARRAY
+
+    logger.info(f"Map parameters: {map_params}")
+    map_global = create_bpf_map(module, map_name, map_params)
+    # Generate debug info for BTF
+    create_map_debug_info(module, map_global.sym, map_name, map_params)
+    return map_global
+
+
+def process_bpf_map(func_node, module, structs_sym_tab):
+    """Process a BPF map (a function decorated with @map)"""
+    map_name = func_node.name
+    logger.info(f"Processing BPF map: {map_name}")
+
+    # For now, assume single return statement
+    return_stmt = None
+    for stmt in func_node.body:
+        if isinstance(stmt, ast.Return):
+            return_stmt = stmt
+            break
+    if return_stmt is None:
+        raise ValueError("BPF map must have a return statement")
+
+    rval = return_stmt.value
+
+    if isinstance(rval, ast.Call) and isinstance(rval.func, ast.Name):
+        handler = MapProcessorRegistry.get_processor(rval.func.id)
+        if handler:
+            return handler(map_name, rval, module, structs_sym_tab)
+        else:
+            logger.warning(f"Unknown map type {rval.func.id}, defaulting to HashMap")
+            return process_hash_map(map_name, rval, module)
+    else:
+        raise ValueError("Function under @map must return a map")

pythonbpf/maps/maps_utils.py

@@ -0,0 +1,34 @@
+from collections.abc import Callable
+from dataclasses import dataclass
+from llvmlite import ir
+from typing import Any
+from .map_types import BPFMapType
+
+
+@dataclass
+class MapSymbol:
+    """Class representing a symbol on the map"""
+
+    type: BPFMapType
+    sym: ir.GlobalVariable
+
+
+class MapProcessorRegistry:
+    """Registry for map processor functions"""
+
+    _processors: dict[str, Callable[..., Any]] = {}
+
+    @classmethod
+    def register(cls, map_type_name):
+        """Decorator to register a processor function for a map type"""
+
+        def decorator(func):
+            cls._processors[map_type_name] = func
+            return func
+
+        return decorator
+
+    @classmethod
+    def get_processor(cls, map_type_name):
+        """Get the processor function for a map type"""
+        return cls._processors.get(map_type_name)

pythonbpf/maps_pass.py

@@ -1,219 +0,0 @@
-import ast
-from llvmlite import ir
-from .type_deducer import ctypes_to_ir
-from . import dwarf_constants as dc
-
-map_sym_tab = {}
-
-
-def maps_proc(tree, module, chunks):
-    for func_node in chunks:
-        # Check if this function is a map
-        is_map = False
-        for decorator in func_node.decorator_list:
-            if isinstance(decorator, ast.Name) and decorator.id == "map":
-                is_map = True
-                break
-        if is_map:
-            print(f"Found BPF map: {func_node.name}")
-            process_bpf_map(func_node, module)
-            continue
-    return map_sym_tab
-
-
-def create_bpf_map(module, map_name, map_params):
-    """Create a BPF map in the module with the given parameters and debug info"""
-
-    # Create the anonymous struct type for BPF map
-    map_struct_type = ir.LiteralStructType([
-        ir.PointerType(),
-        ir.PointerType(),
-        ir.PointerType(),
-        ir.PointerType()
-    ])
-
-    # Create the global variable
-    map_global = ir.GlobalVariable(module, map_struct_type, name=map_name)
-    map_global.linkage = 'dso_local'
-    map_global.global_constant = False
-    map_global.initializer = ir.Constant(map_struct_type, None)     # type: ignore
-    map_global.section = ".maps"
-    map_global.align = 8        # type: ignore
-
-    # Generate debug info for BTF
-    create_map_debug_info(module, map_global, map_name, map_params)
-
-    print(f"Created BPF map: {map_name}")
-    map_sym_tab[map_name] = map_global
-    return map_global
-
-def create_map_debug_info(module, map_global, map_name, map_params):
-    """Generate debug information metadata for BPF map"""
-    file_metadata = module._file_metadata
-    compile_unit = module._debug_compile_unit
-
-    # Create basic type for unsigned int (32-bit)
-    uint_type = module.add_debug_info("DIBasicType", {
-        "name": "unsigned int",
-        "size": 32,
-        "encoding": dc.DW_ATE_unsigned
-    })
-
-    # Create basic type for unsigned long long (64-bit)
-    ulong_type = module.add_debug_info("DIBasicType", {
-        "name": "unsigned long long",
-        "size": 64,
-        "encoding": dc.DW_ATE_unsigned
-    })
-
-    # Create array type for map type field (array of 1 unsigned int)
-    array_subrange = module.add_debug_info("DISubrange", {"count": 1})
-    array_type = module.add_debug_info("DICompositeType", {
-        "tag": dc.DW_TAG_array_type,
-        "baseType": uint_type,
-        "size": 32,
-        "elements": [array_subrange]
-    })
-
-    # Create pointer types
-    type_ptr = module.add_debug_info("DIDerivedType", {
-        "tag": dc.DW_TAG_pointer_type,
-        "baseType": array_type,
-        "size": 64
-    })
-
-    max_entries_ptr = module.add_debug_info("DIDerivedType", {
-        "tag": dc.DW_TAG_pointer_type,
-        "baseType": array_type,
-        "size": 64
-    })
-
-    key_ptr = module.add_debug_info("DIDerivedType", {
-        "tag": dc.DW_TAG_pointer_type,
-        "baseType": uint_type,  # Adjust based on actual key type
-        "size": 64
-    })
-
-    value_ptr = module.add_debug_info("DIDerivedType", {
-        "tag": dc.DW_TAG_pointer_type,
-        "baseType": ulong_type,  # Adjust based on actual value type
-        "size": 64
-    })
-
-    # Create struct members
-    # scope field does not appear for some reason
-    type_member = module.add_debug_info("DIDerivedType", {
-        "tag": dc.DW_TAG_member,
-        "name": "type",
-        "file": file_metadata,
-        "baseType": type_ptr,
-        "size": 64,
-        "offset": 0
-    })
-
-    max_entries_member = module.add_debug_info("DIDerivedType", {
-        "tag": dc.DW_TAG_member,
-        "name": "max_entries",
-        "file": file_metadata,
-        "baseType": max_entries_ptr,
-        "size": 64,
-        "offset": 64
-    })
-
-    key_member = module.add_debug_info("DIDerivedType", {
-        "tag": dc.DW_TAG_member,
-        "name": "key",
-        "file": file_metadata,
-        "baseType": key_ptr,
-        "size": 64,
-        "offset": 128
-    })
-
-    value_member = module.add_debug_info("DIDerivedType", {
-        "tag": dc.DW_TAG_member,
-        "name": "value",
-        "file": file_metadata,
-        "baseType": value_ptr,
-        "size": 64,
-        "offset": 192
-    })
-
-    # Create the struct type
-    struct_type = module.add_debug_info("DICompositeType", {
-        "tag": dc.DW_TAG_structure_type,
-        "file": file_metadata,
-        "size": 256,  # 4 * 64-bit pointers
-        "elements": [type_member, max_entries_member, key_member, value_member]
-    }, is_distinct=True)
-
-    # Create global variable debug info
-    global_var = module.add_debug_info("DIGlobalVariable", {
-        "name": map_name,
-        "scope": compile_unit,
-        "file": file_metadata,
-        "type": struct_type,
-        "isLocal": False,
-        "isDefinition": True
-    }, is_distinct=True)
-
-    # Create global variable expression
-    global_var_expr = module.add_debug_info("DIGlobalVariableExpression", {
-        "var": global_var,
-        "expr": module.add_debug_info("DIExpression", {})
-    })
-
-    # Attach debug info to the global variable
-    map_global.set_metadata("dbg", global_var_expr)
-
-    return global_var_expr
-
-
-def process_hash_map(map_name, rval, module):
-    print(f"Creating HashMap map: {map_name}")
-    map_params: dict[str, object] = {"map_type": "HASH"}
-
-    # Assuming order: key_type, value_type, max_entries
-    if len(rval.args) >= 1 and isinstance(rval.args[0], ast.Name):
-        map_params["key_type"] = rval.args[0].id
-    if len(rval.args) >= 2 and isinstance(rval.args[1], ast.Name):
-        map_params["value_type"] = rval.args[1].id
-    if len(rval.args) >= 3 and isinstance(rval.args[2], ast.Constant):
-        const_val = rval.args[2].value
-        if isinstance(const_val, (int, str)):  # safe check
-            map_params["max_entries"] = const_val
-
-    for keyword in rval.keywords:
-        if keyword.arg == "key_type" and isinstance(keyword.value, ast.Name):
-            map_params["key_type"] = keyword.value.id
-        elif keyword.arg == "value_type" and isinstance(keyword.value, ast.Name):
-            map_params["value_type"] = keyword.value.id
-        elif keyword.arg == "max_entries" and isinstance(keyword.value, ast.Constant):
-            const_val = keyword.value.value
-            if isinstance(const_val, (int, str)):
-                map_params["max_entries"] = const_val
-
-    print(f"Map parameters: {map_params}")
-    return create_bpf_map(module, map_name, map_params)
-
-
-def process_bpf_map(func_node, module):
-    """Process a BPF map (a function decorated with @map)"""
-    map_name = func_node.name
-    print(f"Processing BPF map: {map_name}")
-
-    # For now, assume single return statement
-    return_stmt = None
-    for stmt in func_node.body:
-        if isinstance(stmt, ast.Return):
-            return_stmt = stmt
-            break
-    if return_stmt is None:
-        raise ValueError("BPF map must have a return statement")
-
-    rval = return_stmt.value
-
-    # Handle only HashMap maps
-    if isinstance(rval, ast.Call) and isinstance(rval.func, ast.Name) and rval.func.id == "HashMap":
-        process_hash_map(map_name, rval, module)
-    else:
-        raise ValueError("Function under @map must return a map")

pythonbpf/structs/__init__.py

@@ -0,0 +1,3 @@
+from .structs_pass import structs_proc
+
+__all__ = ["structs_proc"]

pythonbpf/structs/struct_type.py

@@ -0,0 +1,33 @@
+from llvmlite import ir
+
+
+class StructType:
+    def __init__(self, ir_type, fields, size):
+        self.ir_type = ir_type
+        self.fields = fields
+        self.size = size
+
+    def field_idx(self, field_name):
+        return list(self.fields.keys()).index(field_name)
+
+    def field_type(self, field_name):
+        return self.fields[field_name]
+
+    def gep(self, builder, ptr, field_name):
+        idx = self.field_idx(field_name)
+        return builder.gep(
+            ptr,
+            [ir.Constant(ir.IntType(32), 0), ir.Constant(ir.IntType(32), idx)],
+            inbounds=True,
+        )
+
+    def field_size(self, field_name):
+        fld = self.fields[field_name]
+        if isinstance(fld, ir.ArrayType):
+            return fld.count * (fld.element.width // 8)
+        elif isinstance(fld, ir.IntType):
+            return fld.width // 8
+        elif isinstance(fld, ir.PointerType):
+            return 8
+
+        raise TypeError(f"Unsupported field type: {fld}")

pythonbpf/structs/structs_pass.py

@@ -0,0 +1,95 @@
+import ast
+import logging
+from llvmlite import ir
+from pythonbpf.type_deducer import ctypes_to_ir
+from .struct_type import StructType
+
+logger = logging.getLogger(__name__)
+
+# TODO: Shall we allow the following syntax:
+# struct MyStruct:
+#     field1: int
+#     field2: str(32)
+# Where int is mapped to c_uint64?
+# Shall we just int64, int32 and uint32 similarly?
+
+
+def structs_proc(tree, module, chunks):
+    """Process all class definitions to find BPF structs"""
+    structs_sym_tab = {}
+    for cls_node in chunks:
+        if is_bpf_struct(cls_node):
+            logger.info(f"Found BPF struct: {cls_node.name}")
+            struct_info = process_bpf_struct(cls_node, module)
+            structs_sym_tab[cls_node.name] = struct_info
+    return structs_sym_tab
+
+
+def is_bpf_struct(cls_node):
+    return any(
+        isinstance(decorator, ast.Name) and decorator.id == "struct"
+        for decorator in cls_node.decorator_list
+    )
+
+
+def process_bpf_struct(cls_node, module):
+    """Process a single BPF struct definition"""
+
+    fields = parse_struct_fields(cls_node)
+    field_types = list(fields.values())
+    total_size = calc_struct_size(field_types)
+    struct_type = ir.LiteralStructType(field_types)
+    logger.info(f"Created struct {cls_node.name} with fields {fields.keys()}")
+    return StructType(struct_type, fields, total_size)
+
+
+def parse_struct_fields(cls_node):
+    """Parse fields of a struct class node"""
+    fields = {}
+
+    for item in cls_node.body:
+        if isinstance(item, ast.AnnAssign) and isinstance(item.target, ast.Name):
+            fields[item.target.id] = get_type_from_ann(item.annotation)
+        else:
+            logger.error(f"Unsupported struct field: {ast.dump(item)}")
+            raise TypeError(f"Unsupported field in {ast.dump(cls_node)}")
+    return fields
+
+
+def get_type_from_ann(annotation):
+    """Convert an AST annotation node to an LLVM IR type for struct fields"""
+    if isinstance(annotation, ast.Call) and isinstance(annotation.func, ast.Name):
+        if annotation.func.id == "str":
+            # Char array
+            # Assumes constant integer argument
+            length = annotation.args[0].value
+            return ir.ArrayType(ir.IntType(8), length)
+    elif isinstance(annotation, ast.Name):
+        # Int type, written as c_int64, c_uint32, etc.
+        return ctypes_to_ir(annotation.id)
+
+    raise TypeError(f"Unsupported annotation type: {ast.dump(annotation)}")
+
+
+def calc_struct_size(field_types):
+    """Calculate total size of the struct with alignment and padding"""
+    curr_offset = 0
+    for ftype in field_types:
+        if isinstance(ftype, ir.IntType):
+            fsize = ftype.width // 8
+            alignment = fsize
+        elif isinstance(ftype, ir.ArrayType):
+            fsize = ftype.count * (ftype.element.width // 8)
+            alignment = ftype.element.width // 8
+        elif isinstance(ftype, ir.PointerType):
+            # We won't encounter this rn, but for the future
+            fsize = 8
+            alignment = 8
+        else:
+            raise TypeError(f"Unsupported field type: {ftype}")
+
+        padding = (alignment - (curr_offset % alignment)) % alignment
+        curr_offset += padding + fsize
+
+    final_padding = (8 - (curr_offset % 8)) % 8
+    return curr_offset + final_padding

pythonbpf/trace.py

@@ -1,54 +0,0 @@
-class TraceEvent:
-    def __init__(self, timestamp, comm, pid, cpu, flags, message):
-        """Represents a parsed trace pipe event"""
-        self.timestamp = timestamp  # float: timestamp in seconds
-        self.comm = comm            # str: command name
-        self.pid = pid              # int: process ID
-        self.cpu = cpu              # int: CPU number
-        self.flags = flags          # str: trace flags
-        self.message = message      # str: the actual message
-
-    def __iter__(self):
-        """Allow unpacking like the original BCC tuple"""
-        yield self.comm
-        yield self.pid
-        yield self.cpu
-        yield self.flags
-        yield self.timestamp
-        yield self.message
-
-
-class TraceReader:
-    def __init__(self, trace_pipe_path="/sys/kernel/debug/tracing/trace_pipe"):
-        self.trace_pipe_path = trace_pipe_path
-        self.file = None
-
-    def __enter__(self):
-        self.file = open(self.trace_pipe_path, "r")
-        return self
-
-    def __exit__(self, exc_type, exc_val, exc_tb):
-        if self.file:
-            self.file.close()
-
-    def __iter__(self):
-        while True:
-            event = self.trace_fields()
-            if event:
-                yield event
-
-    def trace_fields(self):
-        """Read and parse one line from the trace pipe"""
-        if not self.file:
-            self.file = open(self.trace_pipe_path, "r")
-        line = self.file.readline()
-        if not line:
-            return None
-        # Parse the line into components (simplified)
-        # Real implementation would need more robust parsing
-        parts = self._parse_trace_line(line)
-        return TraceEvent(*parts)
-
-    def _parse_trace_line(self, line):
-        # TODO: Implement
-        pass

pythonbpf/type_deducer.py

@@ -1,24 +1,33 @@
 from llvmlite import ir
 
 # TODO: THIS IS NOT SUPPOSED TO MATCH STRINGS :skull:
+mapping = {
+    "c_int8": ir.IntType(8),
+    "c_uint8": ir.IntType(8),
+    "c_int16": ir.IntType(16),
+    "c_uint16": ir.IntType(16),
+    "c_int32": ir.IntType(32),
+    "c_uint32": ir.IntType(32),
+    "c_int64": ir.IntType(64),
+    "c_uint64": ir.IntType(64),
+    "c_float": ir.FloatType(),
+    "c_double": ir.DoubleType(),
+    "c_void_p": ir.IntType(64),
+    "c_long": ir.IntType(64),
+    "c_ulong": ir.IntType(64),
+    "c_longlong": ir.IntType(64),
+    "c_uint": ir.IntType(32),
+    "c_int": ir.IntType(32),
+    # Not so sure about this one
+    "str": ir.PointerType(ir.IntType(8)),
+}
 
 
 def ctypes_to_ir(ctype: str):
-    mapping = {
-        "c_int8": ir.IntType(8),
-        "c_uint8": ir.IntType(8),
-        "c_int16": ir.IntType(16),
-        "c_uint16": ir.IntType(16),
-        "c_int32": ir.IntType(32),
-        "c_uint32": ir.IntType(32),
-        "c_int64": ir.IntType(64),
-        "c_uint64": ir.IntType(64),
-        "c_float": ir.FloatType(),
-        "c_double": ir.DoubleType(),
-        "c_void_p": ir.IntType(64),
-        # Not so sure about this one
-        "str": ir.PointerType(ir.IntType(8))
-    }
     if ctype in mapping:
         return mapping[ctype]
     raise NotImplementedError(f"No mapping for {ctype}")
+
+
+def is_ctypes(ctype: str) -> bool:
+    return ctype in mapping

pythonbpf/utils.py

@@ -0,0 +1,58 @@
+import subprocess
+
+
+def trace_pipe():
+    """Util to read from the trace pipe."""
+    try:
+        subprocess.run(["cat", "/sys/kernel/tracing/trace_pipe"])
+    except KeyboardInterrupt:
+        print("Tracing stopped.")
+    except (FileNotFoundError, PermissionError) as e:
+        print(f"Error accessing trace_pipe: {e}. Try running as root.")
+
+
+def trace_fields():
+    """Parse one line from trace_pipe into fields."""
+    with open("/sys/kernel/tracing/trace_pipe", "rb", buffering=0) as f:
+        while True:
+            line = f.readline().rstrip()
+
+            if not line:
+                continue
+
+            # Skip lost event lines
+            if line.startswith(b"CPU:"):
+                continue
+
+            # Parse BCC-style: first 16 bytes = task
+            task = line[:16].lstrip().decode("utf-8")
+            line = line[17:]  # Skip past task field and space
+
+            # Find the colon that ends "pid cpu flags timestamp"
+            ts_end = line.find(b":")
+            if ts_end == -1:
+                raise ValueError("Cannot parse trace line")
+
+            # Split "pid [cpu] flags timestamp"
+            try:
+                parts = line[:ts_end].split()
+                if len(parts) < 4:
+                    raise ValueError("Not enough fields")
+
+                pid = int(parts[0])
+                cpu = parts[1][1:-1]  # Remove brackets from [cpu]
+                cpu = int(cpu)
+                flags = parts[2]
+                ts = float(parts[3])
+            except (ValueError, IndexError):
+                raise ValueError("Cannot parse trace line")
+
+            # Get message: skip ": symbol:" part
+            line = line[ts_end + 1 :]  # Skip first ":"
+            sym_end = line.find(b":")
+            if sym_end != -1:
+                msg = line[sym_end + 2 :].decode("utf-8")  # Skip ": " after symbol
+            else:
+                msg = line.lstrip().decode("utf-8")
+
+            return (task, pid, cpu, flags, ts, msg)

pythonbpf/vmlinux_parser/__init__.py

@@ -0,0 +1,3 @@
+from .import_detector import vmlinux_proc
+
+__all__ = ["vmlinux_proc"]

pythonbpf/vmlinux_parser/assignment_info.py

@@ -0,0 +1,36 @@
+from enum import Enum, auto
+from typing import Any, Dict, List, Optional
+from dataclasses import dataclass
+import llvmlite.ir as ir
+
+from pythonbpf.vmlinux_parser.dependency_node import Field
+
+
+class AssignmentType(Enum):
+    CONSTANT = auto()
+    STRUCT = auto()
+    ARRAY = auto()  # probably won't be used
+    FUNCTION_POINTER = auto()
+    POINTER = auto()  # again, probably won't be used
+
+
+@dataclass
+class FunctionSignature:
+    return_type: str
+    param_types: List[str]
+    varargs: bool
+
+
+# Thew name of the assignment will be in the dict that uses this class
+@dataclass
+class AssignmentInfo:
+    value_type: AssignmentType
+    python_type: type
+    value: Optional[Any]
+    pointer_level: Optional[int]
+    signature: Optional[FunctionSignature]  # For function pointers
+    # The key of the dict is the name of the field.
+    #   Value is a tuple that contains the global variable representing that field
+    #   along with all the information about that field as a Field type.
+    members: Optional[Dict[str, tuple[ir.GlobalVariable, Field]]]  # For structs.
+    debug_info: Any

pythonbpf/vmlinux_parser/class_handler.py

@@ -0,0 +1,255 @@
+import logging
+from functools import lru_cache
+import importlib
+
+from .dependency_handler import DependencyHandler
+from .dependency_node import DependencyNode
+import ctypes
+from typing import Optional, Any, Dict
+
+logger = logging.getLogger(__name__)
+
+
+@lru_cache(maxsize=1)
+def get_module_symbols(module_name: str):
+    imported_module = importlib.import_module(module_name)
+    return [name for name in dir(imported_module)], imported_module
+
+
+def process_vmlinux_class(
+    node,
+    llvm_module,
+    handler: DependencyHandler,
+):
+    symbols_in_module, imported_module = get_module_symbols("vmlinux")
+    if node.name in symbols_in_module:
+        vmlinux_type = getattr(imported_module, node.name)
+        process_vmlinux_post_ast(vmlinux_type, llvm_module, handler)
+    else:
+        raise ImportError(f"{node.name} not in vmlinux")
+
+
+def process_vmlinux_post_ast(
+    elem_type_class,
+    llvm_handler,
+    handler: DependencyHandler,
+    processing_stack=None,
+):
+    # Initialize processing stack on first call
+    if processing_stack is None:
+        processing_stack = set()
+    symbols_in_module, imported_module = get_module_symbols("vmlinux")
+
+    current_symbol_name = elem_type_class.__name__
+    logger.info(f"Begin {current_symbol_name} Processing")
+    field_table: Dict[str, list] = {}
+    is_complex_type = False
+    containing_type: Optional[Any] = None
+    ctype_complex_type: Optional[Any] = None
+    type_length: Optional[int] = None
+    module_name = getattr(elem_type_class, "__module__", None)
+
+    # Check if already processed
+    if handler.has_node(current_symbol_name):
+        logger.debug(f"Node {current_symbol_name} already processed and ready")
+        return True
+
+    # XXX:Check its use. It's probably not being used.
+    if current_symbol_name in processing_stack:
+        logger.debug(
+            f"Dependency already in processing stack for {current_symbol_name}, skipping"
+        )
+        return True
+
+    processing_stack.add(current_symbol_name)
+
+    if module_name == "vmlinux":
+        if hasattr(elem_type_class, "_type_"):
+            pass
+        else:
+            new_dep_node = DependencyNode(name=current_symbol_name)
+
+            # elem_type_class is the actual vmlinux struct/class
+            new_dep_node.set_ctype_struct(elem_type_class)
+
+            handler.add_node(new_dep_node)
+            class_obj = getattr(imported_module, current_symbol_name)
+            # Inspect the class fields
+            if hasattr(class_obj, "_fields_"):
+                for field_elem in class_obj._fields_:
+                    field_name: str = ""
+                    field_type: Optional[Any] = None
+                    bitfield_size: Optional[int] = None
+                    if len(field_elem) == 2:
+                        field_name, field_type = field_elem
+                    elif len(field_elem) == 3:
+                        field_name, field_type, bitfield_size = field_elem
+                    field_table[field_name] = [field_type, bitfield_size]
+            elif hasattr(class_obj, "__annotations__"):
+                for field_elem in class_obj.__annotations__.items():
+                    if len(field_elem) == 2:
+                        field_name, field_type = field_elem
+                        bitfield_size = None
+                    elif len(field_elem) == 3:
+                        field_name, field_type, bitfield_size = field_elem
+                    else:
+                        raise ValueError(
+                            "Number of fields in items() of class object unexpected"
+                        )
+                    field_table[field_name] = [field_type, bitfield_size]
+            else:
+                raise TypeError("Could not get required class and definition")
+
+            logger.debug(f"Extracted fields for {current_symbol_name}: {field_table}")
+            for elem in field_table.items():
+                elem_name, elem_temp_list = elem
+                [elem_type, elem_bitfield_size] = elem_temp_list
+                local_module_name = getattr(elem_type, "__module__", None)
+                new_dep_node.add_field(elem_name, elem_type, ready=False)
+
+                if local_module_name == ctypes.__name__:
+                    # TODO: need to process pointer to ctype and also CFUNCTYPES here recursively. Current processing is a single dereference
+                    new_dep_node.set_field_bitfield_size(elem_name, elem_bitfield_size)
+
+                    # Process pointer to ctype
+                    if isinstance(elem_type, type) and issubclass(
+                        elem_type, ctypes._Pointer
+                    ):
+                        # Get the pointed-to type
+                        pointed_type = elem_type._type_
+                        logger.debug(f"Found pointer to type: {pointed_type}")
+                        new_dep_node.set_field_containing_type(elem_name, pointed_type)
+                        new_dep_node.set_field_ctype_complex_type(
+                            elem_name, ctypes._Pointer
+                        )
+                        new_dep_node.set_field_ready(elem_name, is_ready=True)
+
+                    # Process function pointers (CFUNCTYPE)
+                    elif hasattr(elem_type, "_restype_") and hasattr(
+                        elem_type, "_argtypes_"
+                    ):
+                        # This is a CFUNCTYPE or similar
+                        logger.info(
+                            f"Function pointer detected for {elem_name} with return type {elem_type._restype_} and arguments {elem_type._argtypes_}"
+                        )
+                        # Set the field as ready but mark it with special handling
+                        new_dep_node.set_field_ctype_complex_type(
+                            elem_name, ctypes.CFUNCTYPE
+                        )
+                        new_dep_node.set_field_ready(elem_name, is_ready=True)
+                        logger.warning(
+                            "Blindly processing CFUNCTYPE ctypes to ensure compilation. Unsupported"
+                        )
+
+                    else:
+                        # Regular ctype
+                        new_dep_node.set_field_ready(elem_name, is_ready=True)
+                        logger.debug(
+                            f"Field {elem_name} is direct ctypes type: {elem_type}"
+                        )
+                elif local_module_name == "vmlinux":
+                    new_dep_node.set_field_bitfield_size(elem_name, elem_bitfield_size)
+                    logger.debug(
+                        f"Processing vmlinux field: {elem_name}, type: {elem_type}"
+                    )
+                    if hasattr(elem_type, "_type_"):
+                        is_complex_type = True
+                        containing_type = elem_type._type_
+                        if hasattr(elem_type, "_length_") and is_complex_type:
+                            type_length = elem_type._length_
+
+                        if containing_type.__module__ == "vmlinux":
+                            new_dep_node.add_dependent(
+                                elem_type._type_.__name__
+                                if hasattr(elem_type._type_, "__name__")
+                                else str(elem_type._type_)
+                            )
+                        elif containing_type.__module__ == ctypes.__name__:
+                            if isinstance(elem_type, type):
+                                if issubclass(elem_type, ctypes.Array):
+                                    ctype_complex_type = ctypes.Array
+                                elif issubclass(elem_type, ctypes._Pointer):
+                                    ctype_complex_type = ctypes._Pointer
+                                else:
+                                    raise ImportError(
+                                        "Non Array and Pointer type ctype imports not supported in current version"
+                                    )
+                            else:
+                                raise TypeError("Unsupported ctypes subclass")
+                        else:
+                            raise ImportError(
+                                f"Unsupported module of {containing_type}"
+                            )
+                        logger.debug(
+                            f"{containing_type} containing type of parent {elem_name} with {elem_type} and ctype {ctype_complex_type} and length {type_length}"
+                        )
+                        new_dep_node.set_field_containing_type(
+                            elem_name, containing_type
+                        )
+                        new_dep_node.set_field_type_size(elem_name, type_length)
+                        new_dep_node.set_field_ctype_complex_type(
+                            elem_name, ctype_complex_type
+                        )
+                        new_dep_node.set_field_type(elem_name, elem_type)
+                        if containing_type.__module__ == "vmlinux":
+                            containing_type_name = (
+                                containing_type.__name__
+                                if hasattr(containing_type, "__name__")
+                                else str(containing_type)
+                            )
+
+                            # Check for self-reference or already processed
+                            if containing_type_name == current_symbol_name:
+                                # Self-referential pointer
+                                logger.debug(
+                                    f"Self-referential pointer in {current_symbol_name}.{elem_name}"
+                                )
+                                new_dep_node.set_field_ready(elem_name, True)
+                            elif handler.has_node(containing_type_name):
+                                # Already processed
+                                logger.debug(
+                                    f"Reusing already processed {containing_type_name}"
+                                )
+                                new_dep_node.set_field_ready(elem_name, True)
+                            else:
+                                # Process recursively - THIS WAS MISSING
+                                new_dep_node.add_dependent(containing_type_name)
+                                process_vmlinux_post_ast(
+                                    containing_type,
+                                    llvm_handler,
+                                    handler,
+                                    processing_stack,
+                                )
+                                new_dep_node.set_field_ready(elem_name, True)
+                        elif containing_type.__module__ == ctypes.__name__:
+                            logger.debug(f"Processing ctype internal{containing_type}")
+                            new_dep_node.set_field_ready(elem_name, True)
+                        else:
+                            raise TypeError(
+                                "Module not supported in recursive resolution"
+                            )
+                    else:
+                        new_dep_node.add_dependent(
+                            elem_type.__name__
+                            if hasattr(elem_type, "__name__")
+                            else str(elem_type)
+                        )
+                        process_vmlinux_post_ast(
+                            elem_type,
+                            llvm_handler,
+                            handler,
+                            processing_stack,
+                        )
+                        new_dep_node.set_field_ready(elem_name, True)
+                else:
+                    raise ValueError(
+                        f"{elem_name} with type {elem_type} from module {module_name} not supported in recursive resolver"
+                    )
+
+    else:
+        raise ImportError("UNSUPPORTED Module")
+
+    logger.info(
+        f"{current_symbol_name} processed and handler readiness {handler.is_ready}"
+    )
+    return True

pythonbpf/vmlinux_parser/dependency_handler.py

@@ -0,0 +1,173 @@
+from typing import Optional, Dict, List, Iterator
+from .dependency_node import DependencyNode
+
+
+class DependencyHandler:
+    """
+    Manages a collection of DependencyNode objects with no duplicates.
+
+    Ensures that no two nodes with the same name can be added and provides
+    methods to check readiness and retrieve specific nodes.
+
+    Example usage:
+        # Create a handler
+        handler = DependencyHandler()
+
+        # Create some dependency nodes
+        node1 = DependencyNode(name="node1")
+        node1.add_field("field1", str)
+        node1.set_field_value("field1", "value1")
+
+        node2 = DependencyNode(name="node2")
+        node2.add_field("field1", int)
+
+        # Add nodes to the handler
+        handler.add_node(node1)
+        handler.add_node(node2)
+
+        # Check if a specific node exists
+        print(handler.has_node("node1"))  # True
+
+        # Get a reference to a node and modify it
+        node = handler.get_node("node2")
+        node.set_field_value("field1", 42)
+
+        # Check if all nodes are ready
+        print(handler.is_ready)  # False (node2 is ready, but node1 isn't)
+    """
+
+    def __init__(self):
+        # Using a dictionary with node names as keys ensures name uniqueness
+        # and provides efficient lookups
+        self._nodes: Dict[str, DependencyNode] = {}
+
+    def add_node(self, node: DependencyNode) -> bool:
+        """
+        Add a dependency node to the handler.
+
+        Args:
+            node: The DependencyNode to add
+
+        Returns:
+            bool: True if the node was added, False if a node with the same name already exists
+
+        Raises:
+            TypeError: If the provided object is not a DependencyNode
+        """
+        if not isinstance(node, DependencyNode):
+            raise TypeError(f"Expected DependencyNode, got {type(node).__name__}")
+
+        # Check if a node with this name already exists
+        if node.name in self._nodes:
+            return False
+
+        self._nodes[node.name] = node
+        return True
+
+    @property
+    def is_ready(self) -> bool:
+        """
+        Check if all nodes are ready.
+
+        Returns:
+            bool: True if all nodes are ready (or if there are no nodes), False otherwise
+        """
+        if not self._nodes:
+            return True
+
+        return all(node.is_ready for node in self._nodes.values())
+
+    def has_node(self, name: str) -> bool:
+        """
+        Check if a node with the given name exists.
+
+        Args:
+            name: The name to check
+
+        Returns:
+            bool: True if a node with the given name exists, False otherwise
+        """
+        return name in self._nodes
+
+    def get_node(self, name: str) -> Optional[DependencyNode]:
+        """
+        Get a node by name for manipulation.
+
+        Args:
+            name: The name of the node to retrieve
+
+        Returns:
+            Optional[DependencyNode]: The node with the given name, or None if not found
+        """
+        return self._nodes.get(name)
+
+    def remove_node(self, node_or_name) -> bool:
+        """
+        Remove a node by name or reference.
+
+        Args:
+            node_or_name: The node to remove or its name
+
+        Returns:
+            bool: True if the node was removed, False if not found
+        """
+        if isinstance(node_or_name, DependencyNode):
+            name = node_or_name.name
+        else:
+            name = node_or_name
+
+        if name in self._nodes:
+            del self._nodes[name]
+            return True
+        return False
+
+    def get_all_nodes(self) -> List[DependencyNode]:
+        """
+        Get all nodes stored in the handler.
+
+        Returns:
+            List[DependencyNode]: List of all nodes
+        """
+        return list(self._nodes.values())
+
+    def __iter__(self) -> Iterator[DependencyNode]:
+        """
+        Iterate over all nodes.
+
+        Returns:
+            Iterator[DependencyNode]: Iterator over all nodes
+        """
+        return iter(self._nodes.values())
+
+    def __len__(self) -> int:
+        """
+        Get the number of nodes in the handler.
+
+        Returns:
+            int: The number of nodes
+        """
+        return len(self._nodes)
+
+    def __getitem__(self, name: str) -> DependencyNode:
+        """
+        Get a node by name using dictionary-style access.
+
+        Args:
+            name: The name of the node to retrieve
+
+        Returns:
+            DependencyNode: The node with the given name
+
+        Raises:
+            KeyError: If no node with the given name exists
+
+        Example:
+            node = handler["some-dep_node_name"]
+        """
+        if name not in self._nodes:
+            raise KeyError(f"No node with name '{name}' found")
+        return self._nodes[name]
+
+    @property
+    def nodes(self):
+        return self._nodes

pythonbpf/vmlinux_parser/dependency_node.py

@@ -0,0 +1,388 @@
+from dataclasses import dataclass, field
+from typing import Dict, Any, Optional
+import ctypes
+
+
+# TODO: FIX THE FUCKING TYPE NAME CONVENTION.
+@dataclass
+class Field:
+    """Represents a field in a dependency node with its type and readiness state."""
+
+    name: str
+    type: type
+    ctype_complex_type: Optional[Any]
+    containing_type: Optional[Any]
+    type_size: Optional[int]
+    bitfield_size: Optional[int]
+    offset: int
+    value: Any = None
+    ready: bool = False
+
+    def __hash__(self):
+        """
+        Create a hash based on the immutable attributes that define this field's identity.
+        This allows Field objects to be used as dictionary keys.
+        """
+        # Use a tuple of the fields that uniquely identify this field
+        identity = (
+            self.name,
+            id(self.type),  # Use id for non-hashable types
+            id(self.ctype_complex_type) if self.ctype_complex_type else None,
+            id(self.containing_type) if self.containing_type else None,
+            self.type_size,
+            self.bitfield_size,
+            self.offset,
+            self.value if self.value else None,
+        )
+        return hash(identity)
+
+    def __eq__(self, other):
+        """
+        Define equality consistent with the hash function.
+        Two fields are equal if they have they are the same
+        """
+        return self is other
+
+    def set_ready(self, is_ready: bool = True) -> None:
+        """Set the readiness state of this field."""
+        self.ready = is_ready
+
+    def set_value(self, value: Any, mark_ready: bool = False) -> None:
+        """Set the value of this field and optionally mark it as ready."""
+        self.value = value
+        if mark_ready:
+            self.ready = True
+
+    def set_type(self, given_type, mark_ready: bool = False) -> None:
+        """Set value of the type field and mark as ready"""
+        self.type = given_type
+        if mark_ready:
+            self.ready = True
+
+    def set_containing_type(
+        self, containing_type: Optional[Any], mark_ready: bool = False
+    ) -> None:
+        """Set the containing_type of this field and optionally mark it as ready."""
+        self.containing_type = containing_type
+        if mark_ready:
+            self.ready = True
+
+    def set_type_size(self, type_size: Any, mark_ready: bool = False) -> None:
+        """Set the type_size of this field and optionally mark it as ready."""
+        self.type_size = type_size
+        if mark_ready:
+            self.ready = True
+
+    def set_ctype_complex_type(
+        self, ctype_complex_type: Any, mark_ready: bool = False
+    ) -> None:
+        """Set the ctype_complex_type of this field and optionally mark it as ready."""
+        self.ctype_complex_type = ctype_complex_type
+        if mark_ready:
+            self.ready = True
+
+    def set_bitfield_size(self, bitfield_size: Any, mark_ready: bool = False) -> None:
+        """Set the bitfield_size of this field and optionally mark it as ready."""
+        self.bitfield_size = bitfield_size
+        if mark_ready:
+            self.ready = True
+
+    def set_offset(self, offset: int) -> None:
+        """Set the offset of this field"""
+        self.offset = offset
+
+
+@dataclass
+class DependencyNode:
+    """
+    A node with typed fields and readiness tracking.
+
+    Example usage:
+        # Create a dependency node for a Person
+        somestruct = DependencyNode(name="struct_1")
+
+        # Add fields with their types
+        somestruct.add_field("field_1", str)
+        somestruct.add_field("field_2", int)
+        somestruct.add_field("field_3", str)
+
+        # Check if the node is ready (should be False initially)
+        print(f"Is node ready? {somestruct.is_ready}")  # False
+
+        # Set some field values
+        somestruct.set_field_value("field_1", "someproperty")
+        somestruct.set_field_value("field_2", 30)
+
+        # Check if the node is ready (still False because email is not ready)
+        print(f"Is node ready? {somestruct.is_ready}")  # False
+
+        # Set the last field and make the node ready
+        somestruct.set_field_value("field_3", "anotherproperty")
+
+        # Now the node should be ready
+        print(f"Is node ready? {somestruct.is_ready}")  # True
+
+        # You can also mark a field as not ready
+        somestruct.set_field_ready("field_3", False)
+
+        # Now the node is not ready again
+        print(f"Is node ready? {somestruct.is_ready}")  # False
+
+        # Get all field values
+        print(somestruct.get_field_values())  # {'field_1': 'someproperty', 'field_2': 30, 'field_3': 'anotherproperty'}
+
+        # Get only ready fields
+        ready_fields = somestruct.get_ready_fields()
+        print(f"Ready fields: {[field.name for field in ready_fields.values()]}")  # ['field_1', 'field_2']
+    """
+
+    name: str
+    depends_on: Optional[list[str]] = None
+    fields: Dict[str, Field] = field(default_factory=dict)
+    _ready_cache: Optional[bool] = field(default=None, repr=False)
+    current_offset: int = 0
+    ctype_struct: Optional[Any] = field(default=None, repr=False)
+
+    def add_field(
+        self,
+        name: str,
+        field_type: type,
+        initial_value: Any = None,
+        containing_type: Optional[Any] = None,
+        type_size: Optional[int] = None,
+        ctype_complex_type: Optional[int] = None,
+        bitfield_size: Optional[int] = None,
+        ready: bool = False,
+        offset: int = 0,
+    ) -> None:
+        """Add a field to the node with an optional initial value and readiness state."""
+        if self.depends_on is None:
+            self.depends_on = []
+        self.fields[name] = Field(
+            name=name,
+            type=field_type,
+            value=initial_value,
+            ready=ready,
+            containing_type=containing_type,
+            type_size=type_size,
+            ctype_complex_type=ctype_complex_type,
+            bitfield_size=bitfield_size,
+            offset=offset,
+        )
+        # Invalidate readiness cache
+        self._ready_cache = None
+
+    def set_ctype_struct(self, ctype_struct: Any) -> None:
+        """Set the ctypes structure for automatic offset calculation."""
+        self.ctype_struct = ctype_struct
+
+    def __sizeof__(self):
+        # If we have a ctype_struct, use its size
+        if self.ctype_struct is not None:
+            return ctypes.sizeof(self.ctype_struct)
+        return self.current_offset
+
+    def get_field(self, name: str) -> Field:
+        """Get a field by name."""
+        return self.fields[name]
+
+    def set_field_value(self, name: str, value: Any, mark_ready: bool = False) -> None:
+        """Set a field's value and optionally mark it as ready."""
+        if name not in self.fields:
+            raise KeyError(f"Field '{name}' does not exist in node '{self.name}'")
+
+        self.fields[name].set_value(value, mark_ready)
+        # Invalidate readiness cache
+        self._ready_cache = None
+
+    def set_field_type(self, name: str, type: Any, mark_ready: bool = False) -> None:
+        """Set a field's type and optionally mark it as ready."""
+        if name not in self.fields:
+            raise KeyError(f"Field '{name}' does not exist in node '{self.name}'")
+
+        self.fields[name].set_type(type, mark_ready)
+        # Invalidate readiness cache
+        self._ready_cache = None
+
+    def set_field_containing_type(
+        self, name: str, containing_type: Any, mark_ready: bool = False
+    ) -> None:
+        """Set a field's containing_type and optionally mark it as ready."""
+        if name not in self.fields:
+            raise KeyError(f"Field '{name}' does not exist in node '{self.name}'")
+
+        self.fields[name].set_containing_type(containing_type, mark_ready)
+        # Invalidate readiness cache
+        self._ready_cache = None
+
+    def set_field_type_size(
+        self, name: str, type_size: Any, mark_ready: bool = False
+    ) -> None:
+        """Set a field's type_size and optionally mark it as ready."""
+        if name not in self.fields:
+            raise KeyError(f"Field '{name}' does not exist in node '{self.name}'")
+
+        self.fields[name].set_type_size(type_size, mark_ready)
+        # Invalidate readiness cache
+        self._ready_cache = None
+
+    def set_field_ctype_complex_type(
+        self, name: str, ctype_complex_type: Any, mark_ready: bool = False
+    ) -> None:
+        """Set a field's ctype_complex_type and optionally mark it as ready."""
+        if name not in self.fields:
+            raise KeyError(f"Field '{name}' does not exist in node '{self.name}'")
+
+        self.fields[name].set_ctype_complex_type(ctype_complex_type, mark_ready)
+        # Invalidate readiness cache
+        self._ready_cache = None
+
+    def set_field_bitfield_size(
+        self, name: str, bitfield_size: Any, mark_ready: bool = False
+    ) -> None:
+        """Set a field's bitfield_size and optionally mark it as ready."""
+        if name not in self.fields:
+            raise KeyError(f"Field '{name}' does not exist in node '{self.name}'")
+
+        self.fields[name].set_bitfield_size(bitfield_size, mark_ready)
+        # Invalidate readiness cache
+        self._ready_cache = None
+
+    def set_field_ready(
+        self,
+        name: str,
+        is_ready: bool = False,
+        size_of_containing_type: Optional[int] = None,
+    ) -> None:
+        """Mark a field as ready or not ready."""
+        if name not in self.fields:
+            raise KeyError(f"Field '{name}' does not exist in node '{self.name}'")
+
+        self.fields[name].set_ready(is_ready)
+
+        # Use ctypes built-in offset if available
+        if self.ctype_struct is not None:
+            try:
+                self.fields[name].set_offset(getattr(self.ctype_struct, name).offset)
+            except AttributeError:
+                # Fallback to manual calculation if field not found in ctype_struct
+                self.fields[name].set_offset(self.current_offset)
+                self.current_offset += self._calculate_size(
+                    name, size_of_containing_type
+                )
+        else:
+            # Manual offset calculation when no ctype_struct is available
+            self.fields[name].set_offset(self.current_offset)
+            self.current_offset += self._calculate_size(name, size_of_containing_type)
+
+        # Invalidate readiness cache
+        self._ready_cache = None
+
+    def _calculate_size(
+        self, name: str, size_of_containing_type: Optional[int] = None
+    ) -> int:
+        processing_field = self.fields[name]
+        # size_of_field will be in bytes
+        if processing_field.type.__module__ == ctypes.__name__:
+            size_of_field = ctypes.sizeof(processing_field.type)
+            return size_of_field
+        elif processing_field.type.__module__ == "vmlinux":
+            if processing_field.ctype_complex_type is not None:
+                if issubclass(processing_field.ctype_complex_type, ctypes.Array):
+                    if processing_field.containing_type.__module__ == ctypes.__name__:
+                        if (
+                            processing_field.containing_type is not None
+                            and processing_field.type_size is not None
+                        ):
+                            size_of_field = (
+                                ctypes.sizeof(processing_field.containing_type)
+                                * processing_field.type_size
+                            )
+                        else:
+                            raise RuntimeError(
+                                f"{processing_field} has no containing_type or type_size"
+                            )
+                        return size_of_field
+                    elif processing_field.containing_type.__module__ == "vmlinux":
+                        if (
+                            size_of_containing_type is not None
+                            and processing_field.type_size is not None
+                        ):
+                            size_of_field = (
+                                size_of_containing_type * processing_field.type_size
+                            )
+                        else:
+                            raise RuntimeError(
+                                f"{processing_field} has no containing_type or type_size"
+                            )
+                        return size_of_field
+                elif issubclass(processing_field.ctype_complex_type, ctypes._Pointer):
+                    return ctypes.sizeof(ctypes.c_void_p)
+                else:
+                    raise NotImplementedError(
+                        "This subclass of ctype not supported yet"
+                    )
+            elif processing_field.type_size is not None:
+                # Handle vmlinux types with type_size but no ctype_complex_type
+                # This means it's a direct vmlinux struct field (not array/pointer wrapped)
+                # The type_size should already contain the full size of the struct
+                # But if there's a containing_type from vmlinux, we need that size
+                if processing_field.containing_type is not None:
+                    if processing_field.containing_type.__module__ == "vmlinux":
+                        # For vmlinux containing types, we need the pre-calculated size
+                        if size_of_containing_type is not None:
+                            return size_of_containing_type * processing_field.type_size
+                        else:
+                            raise RuntimeError(
+                                f"Field {name}: vmlinux containing_type requires size_of_containing_type"
+                            )
+                    else:
+                        raise ModuleNotFoundError(
+                            f"Containing type module {processing_field.containing_type.__module__} not supported"
+                        )
+                else:
+                    raise RuntimeError("Wrong type found with no containing type")
+            else:
+                # No ctype_complex_type and no type_size, must rely on size_of_containing_type
+                if size_of_containing_type is None:
+                    raise RuntimeError(
+                        f"Size of containing type {size_of_containing_type} is None"
+                    )
+                return size_of_containing_type
+
+        else:
+            raise ModuleNotFoundError("Module is not supported for the operation")
+        raise RuntimeError("control should not reach here")
+
+    @property
+    def is_ready(self) -> bool:
+        """Check if the node is ready (all fields are ready)."""
+        # Use cached value if available
+        if self._ready_cache is not None:
+            return self._ready_cache
+
+        # Calculate readiness only when needed
+        if not self.fields:
+            self._ready_cache = True
+            return True
+
+        self._ready_cache = all(elem.ready for elem in self.fields.values())
+        return self._ready_cache
+
+    def get_field_values(self) -> Dict[str, Any]:
+        """Get a dictionary of field names to their values."""
+        return {name: elem.value for name, elem in self.fields.items()}
+
+    def get_ready_fields(self) -> Dict[str, Field]:
+        """Get all fields that are marked as ready."""
+        return {name: elem for name, elem in self.fields.items() if elem.ready}
+
+    def get_not_ready_fields(self) -> Dict[str, Field]:
+        """Get all fields that are marked as not ready."""
+        return {name: elem for name, elem in self.fields.items() if not elem.ready}
+
+    def add_dependent(self, dep_type):
+        if dep_type in self.depends_on:
+            return
+        else:
+            self.depends_on.append(dep_type)

pythonbpf/vmlinux_parser/import_detector.py

@@ -0,0 +1,163 @@
+import ast
+import logging
+import importlib
+import inspect
+
+from .assignment_info import AssignmentInfo, AssignmentType
+from .dependency_handler import DependencyHandler
+from .ir_gen import IRGenerator
+from .class_handler import process_vmlinux_class
+
+logger = logging.getLogger(__name__)
+
+
+def detect_import_statement(tree: ast.AST) -> list[tuple[str, ast.ImportFrom]]:
+    """
+    Parse AST and detect import statements from vmlinux.
+
+    Returns a list of tuples (module_name, imported_item) for vmlinux imports.
+    Raises SyntaxError for invalid import patterns.
+
+    Args:
+        tree: The AST to parse
+
+    Returns:
+        List of tuples containing (module_name, imported_item) for each vmlinux import
+
+    Raises:
+        SyntaxError: If multiple imports from vmlinux are attempted or import * is used
+    """
+    vmlinux_imports = []
+
+    for node in ast.walk(tree):
+        # Handle "from vmlinux import ..." statements
+        if isinstance(node, ast.ImportFrom):
+            if node.module == "vmlinux":
+                # Check for wildcard import: from vmlinux import *
+                if any(alias.name == "*" for alias in node.names):
+                    raise SyntaxError(
+                        "Wildcard imports from vmlinux are not supported. "
+                        "Please import specific types explicitly."
+                    )
+
+                # Check for multiple imports: from vmlinux import A, B, C
+                if len(node.names) > 1:
+                    imported_names = [alias.name for alias in node.names]
+                    raise SyntaxError(
+                        f"Multiple imports from vmlinux are not supported. "
+                        f"Found: {', '.join(imported_names)}. "
+                        f"Please use separate import statements for each type."
+                    )
+
+                # Check if no specific import is specified (should not happen with valid Python)
+                if len(node.names) == 0:
+                    raise SyntaxError(
+                        "Import from vmlinux must specify at least one type."
+                    )
+
+                # Valid single import
+                for alias in node.names:
+                    import_name = alias.name
+                    # Use alias if provided, otherwise use the original name (commented)
+                    # as_name = alias.asname if alias.asname else alias.name
+                    vmlinux_imports.append(("vmlinux", node))
+                    logger.info(f"Found vmlinux import: {import_name}")
+
+        # Handle "import vmlinux" statements (not typical but should be rejected)
+        elif isinstance(node, ast.Import):
+            for alias in node.names:
+                if alias.name == "vmlinux" or alias.name.startswith("vmlinux."):
+                    raise SyntaxError(
+                        "Direct import of vmlinux module is not supported. "
+                        "Use 'from vmlinux import <type>' instead."
+                    )
+
+    logger.info(f"Total vmlinux imports detected: {len(vmlinux_imports)}")
+    return vmlinux_imports
+
+
+def vmlinux_proc(tree: ast.AST, module):
+    import_statements = detect_import_statement(tree)
+
+    # initialise dependency handler
+    handler = DependencyHandler()
+    # initialise assignment dictionary of name to type
+    assignments: dict[str, AssignmentInfo] = {}
+
+    if not import_statements:
+        logger.info("No vmlinux imports found")
+        return None
+
+    # Import vmlinux module directly
+    try:
+        vmlinux_mod = importlib.import_module("vmlinux")
+    except ImportError:
+        logger.warning("Could not import vmlinux module")
+        return None
+
+    source_file = inspect.getsourcefile(vmlinux_mod)
+    if source_file is None:
+        logger.warning("Cannot find source for vmlinux module")
+        return None
+
+    with open(source_file, "r") as f:
+        mod_ast = ast.parse(f.read(), filename=source_file)
+
+    for import_mod, import_node in import_statements:
+        for alias in import_node.names:
+            imported_name = alias.name
+            found = False
+            for mod_node in mod_ast.body:
+                if (
+                    isinstance(mod_node, ast.ClassDef)
+                    and mod_node.name == imported_name
+                ):
+                    process_vmlinux_class(mod_node, module, handler)
+                    found = True
+                    break
+                if isinstance(mod_node, ast.Assign):
+                    for target in mod_node.targets:
+                        if isinstance(target, ast.Name) and target.id == imported_name:
+                            process_vmlinux_assign(mod_node, module, assignments)
+                            found = True
+                            break
+                if found:
+                    break
+            if not found:
+                logger.info(
+                    f"{imported_name} not found as ClassDef or Assign in vmlinux"
+                )
+
+    IRGenerator(module, handler, assignments)
+    return assignments
+
+
+def process_vmlinux_assign(node, module, assignments: dict[str, AssignmentInfo]):
+    """Process assignments from vmlinux module."""
+    # Only handle single-target assignments
+    if len(node.targets) == 1 and isinstance(node.targets[0], ast.Name):
+        target_name = node.targets[0].id
+
+        # Handle constant value assignments
+        if isinstance(node.value, ast.Constant):
+            # Fixed: using proper TypedDict creation syntax with named arguments
+            assignments[target_name] = AssignmentInfo(
+                value_type=AssignmentType.CONSTANT,
+                python_type=type(node.value.value),
+                value=node.value.value,
+                pointer_level=None,
+                signature=None,
+                members=None,
+                debug_info=None,
+            )
+            logger.info(
+                f"Added assignment: {target_name} = {node.value.value!r} of type {type(node.value.value)}"
+            )
+
+        # Handle other assignment types that we may need to support
+        else:
+            logger.warning(
+                f"Unsupported assignment type for {target_name}: {ast.dump(node.value)}"
+            )
+    else:
+        raise ValueError("Not a simple assignment")

pythonbpf/vmlinux_parser/ir_gen/__init__.py

@@ -0,0 +1,3 @@
+from .ir_generation import IRGenerator
+
+__all__ = ["IRGenerator"]

pythonbpf/vmlinux_parser/ir_gen/debug_info_gen.py

@@ -0,0 +1,161 @@
+from pythonbpf.debuginfo import DebugInfoGenerator, dwarf_constants as dc
+from ..dependency_node import DependencyNode
+import ctypes
+import logging
+from typing import List, Any, Tuple
+
+logger = logging.getLogger(__name__)
+
+
+def debug_info_generation(
+    struct: DependencyNode,
+    llvm_module,
+    generated_debug_info: List[Tuple[DependencyNode, Any]],
+) -> Any:
+    """
+    Generate DWARF debug information for a struct defined in a DependencyNode.
+
+    Args:
+        struct: The dependency node containing struct information
+        llvm_module: The LLVM module to add debug info to
+        generated_debug_info: List of tuples (struct, debug_info) to track generated debug info
+
+    Returns:
+        The generated global variable debug info
+    """
+    # Set up debug info generator
+    generator = DebugInfoGenerator(llvm_module)
+
+    # Check if debug info for this struct has already been generated
+    for existing_struct, debug_info in generated_debug_info:
+        if existing_struct.name == struct.name:
+            return debug_info
+
+    # Process all fields and create members for the struct
+    members = []
+    for field_name, field in struct.fields.items():
+        # Get appropriate debug type for this field
+        field_type = _get_field_debug_type(
+            field_name, field, generator, struct, generated_debug_info
+        )
+        # Create struct member with proper offset
+        member = generator.create_struct_member_vmlinux(
+            field_name, field_type, field.offset * 8
+        )
+        members.append(member)
+
+    if struct.name.startswith("struct_"):
+        struct_name = struct.name.removeprefix("struct_")
+    else:
+        raise ValueError("Unions are not supported in the current version")
+    # Create struct type with all members
+    struct_type = generator.create_struct_type_with_name(
+        struct_name, members, struct.__sizeof__() * 8, is_distinct=True
+    )
+
+    return struct_type
+
+
+def _get_field_debug_type(
+    field_name: str,
+    field,
+    generator: DebugInfoGenerator,
+    parent_struct: DependencyNode,
+    generated_debug_info: List[Tuple[DependencyNode, Any]],
+) -> tuple[Any, int]:
+    """
+    Determine the appropriate debug type for a field based on its Python/ctypes type.
+
+    Args:
+        field_name: Name of the field
+        field: Field object containing type information
+        generator: DebugInfoGenerator instance
+        parent_struct: The parent struct containing this field
+        generated_debug_info: List of already generated debug info
+
+    Returns:
+        The debug info type for this field
+    """
+    # Handle complex types (arrays, pointers)
+    if field.ctype_complex_type is not None:
+        if issubclass(field.ctype_complex_type, ctypes.Array):
+            # Handle array types
+            element_type, base_type_size = _get_basic_debug_type(
+                field.containing_type, generator
+            )
+            return generator.create_array_type_vmlinux(
+                (element_type, base_type_size * field.type_size), field.type_size
+            ), field.type_size * base_type_size
+        elif issubclass(field.ctype_complex_type, ctypes._Pointer):
+            # Handle pointer types
+            pointee_type, _ = _get_basic_debug_type(field.containing_type, generator)
+            return generator.create_pointer_type(pointee_type), 64
+
+    # Handle other vmlinux types (nested structs)
+    if field.type.__module__ == "vmlinux":
+        # If it's a struct from vmlinux, check if we've already generated debug info for it
+        struct_name = field.type.__name__
+
+        # Look for existing debug info in the list
+        for existing_struct, debug_info in generated_debug_info:
+            if existing_struct.name == struct_name:
+                # Use existing debug info
+                return debug_info, existing_struct.__sizeof__()
+
+        # If not found, create a forward declaration
+        # This will be completed when the actual struct is processed
+        logger.warning("Forward declaration in struct created")
+        forward_type = generator.create_struct_type([], 0, is_distinct=True)
+        return forward_type, 0
+
+    # Handle basic C types
+    return _get_basic_debug_type(field.type, generator)
+
+
+def _get_basic_debug_type(ctype, generator: DebugInfoGenerator) -> Any:
+    """
+    Map a ctypes type to a DWARF debug type.
+
+    Args:
+        ctype: A ctypes type or Python type
+        generator: DebugInfoGenerator instance
+
+    Returns:
+        The corresponding debug type
+    """
+    # Map ctypes to debug info types
+    if ctype == ctypes.c_char or ctype == ctypes.c_byte:
+        return generator.get_basic_type("char", 8, dc.DW_ATE_signed_char), 8
+    elif ctype == ctypes.c_ubyte or ctype == ctypes.c_uint8:
+        return generator.get_basic_type("unsigned char", 8, dc.DW_ATE_unsigned_char), 8
+    elif ctype == ctypes.c_short or ctype == ctypes.c_int16:
+        return generator.get_basic_type("short", 16, dc.DW_ATE_signed), 16
+    elif ctype == ctypes.c_ushort or ctype == ctypes.c_uint16:
+        return generator.get_basic_type("unsigned short", 16, dc.DW_ATE_unsigned), 16
+    elif ctype == ctypes.c_int or ctype == ctypes.c_int32:
+        return generator.get_basic_type("int", 32, dc.DW_ATE_signed), 32
+    elif ctype == ctypes.c_uint or ctype == ctypes.c_uint32:
+        return generator.get_basic_type("unsigned int", 32, dc.DW_ATE_unsigned), 32
+    elif ctype == ctypes.c_long:
+        return generator.get_basic_type("long", 64, dc.DW_ATE_signed), 64
+    elif ctype == ctypes.c_ulong:
+        return generator.get_basic_type("unsigned long", 64, dc.DW_ATE_unsigned), 64
+    elif ctype == ctypes.c_longlong or ctype == ctypes.c_int64:
+        return generator.get_basic_type("long long", 64, dc.DW_ATE_signed), 64
+    elif ctype == ctypes.c_ulonglong or ctype == ctypes.c_uint64:
+        return generator.get_basic_type(
+            "unsigned long long", 64, dc.DW_ATE_unsigned
+        ), 64
+    elif ctype == ctypes.c_float:
+        return generator.get_basic_type("float", 32, dc.DW_ATE_float), 32
+    elif ctype == ctypes.c_double:
+        return generator.get_basic_type("double", 64, dc.DW_ATE_float), 64
+    elif ctype == ctypes.c_bool:
+        return generator.get_basic_type("bool", 8, dc.DW_ATE_boolean), 8
+    elif ctype == ctypes.c_char_p:
+        char_type = generator.get_basic_type("char", 8, dc.DW_ATE_signed_char), 8
+        return generator.create_pointer_type(char_type)
+    elif ctype == ctypes.c_void_p:
+        return generator.create_pointer_type(None), 64
+    else:
+        return generator.get_uint64_type(), 64

pythonbpf/vmlinux_parser/ir_gen/ir_generation.py

@@ -0,0 +1,225 @@
+import ctypes
+import logging
+
+from ..assignment_info import AssignmentInfo, AssignmentType
+from ..dependency_handler import DependencyHandler
+from .debug_info_gen import debug_info_generation
+from ..dependency_node import DependencyNode
+import llvmlite.ir as ir
+
+logger = logging.getLogger(__name__)
+
+
+class IRGenerator:
+    # get the assignments dict and add this stuff to it.
+    def __init__(self, llvm_module, handler: DependencyHandler, assignments):
+        self.llvm_module = llvm_module
+        self.handler: DependencyHandler = handler
+        self.generated: list[str] = []
+        self.generated_debug_info: list = []
+        # Use struct_name and field_name as key instead of Field object
+        self.generated_field_names: dict[str, dict[str, ir.GlobalVariable]] = {}
+        self.assignments: dict[str, AssignmentInfo] = assignments
+        if not handler.is_ready:
+            raise ImportError(
+                "Semantic analysis of vmlinux imports failed. Cannot generate IR"
+            )
+        for struct in handler:
+            self.struct_processor(struct)
+
+    def struct_processor(self, struct, processing_stack=None):
+        # Initialize processing stack on first call
+        if processing_stack is None:
+            processing_stack = set()
+
+        # If already generated, skip
+        if struct.name in self.generated:
+            return
+
+        # Detect circular dependency
+        if struct.name in processing_stack:
+            logger.info(
+                f"Circular dependency detected for {struct.name}, skipping recursive processing"
+            )
+            # For circular dependencies, we can either:
+            # 1. Use forward declarations (opaque pointers)
+            # 2. Mark as incomplete and process later
+            # 3. Generate a placeholder type
+            # Here we'll just skip and let it be processed in its own call
+            return
+
+        logger.info(f"IR generating for {struct.name}")
+
+        # Add to processing stack before processing dependencies
+        processing_stack.add(struct.name)
+
+        try:
+            # Process all dependencies first
+            if struct.depends_on is None:
+                pass
+            else:
+                for dependency in struct.depends_on:
+                    if dependency not in self.generated:
+                        # Check if dependency exists in handler
+                        if dependency in self.handler.nodes:
+                            dep_node_from_dependency = self.handler[dependency]
+                            # Pass the processing_stack down to track circular refs
+                            self.struct_processor(
+                                dep_node_from_dependency, processing_stack
+                            )
+                        else:
+                            raise RuntimeError(
+                                f"Warning: Dependency {dependency} not found in handler"
+                            )
+
+            # Generate IR first to populate field names
+            struct_debug_info = self.gen_ir(struct, self.generated_debug_info)
+            self.generated_debug_info.append((struct, struct_debug_info))
+
+            # Fill the assignments dictionary with struct information
+            if struct.name not in self.assignments:
+                # Create a members dictionary for AssignmentInfo
+                members_dict = {}
+                for field_name, field in struct.fields.items():
+                    # Get the generated field name from our dictionary, or use field_name if not found
+                    if (
+                        struct.name in self.generated_field_names
+                        and field_name in self.generated_field_names[struct.name]
+                    ):
+                        field_global_variable = self.generated_field_names[struct.name][
+                            field_name
+                        ]
+                        members_dict[field_name] = (field_global_variable, field)
+                    else:
+                        raise ValueError(
+                            f"llvm global name not found for struct field {field_name}"
+                        )
+                        # members_dict[field_name] = (field_name, field)
+
+                # Add struct to assignments dictionary
+                self.assignments[struct.name] = AssignmentInfo(
+                    value_type=AssignmentType.STRUCT,
+                    python_type=struct.ctype_struct,
+                    value=None,
+                    pointer_level=None,
+                    signature=None,
+                    members=members_dict,
+                    debug_info=struct_debug_info,
+                )
+                logger.info(f"Added struct assignment info for {struct.name}")
+
+            self.generated.append(struct.name)
+
+        finally:
+            # Remove from processing stack after we're done
+            processing_stack.discard(struct.name)
+
+    def gen_ir(self, struct, generated_debug_info):
+        # TODO: we add the btf_ama attribute by monkey patching in the end of compilation, but once llvmlite
+        #  accepts our issue, we will resort to normal accessed attribute based attribute addition
+        # currently we generate all possible field accesses for CO-RE and put into the assignment table
+        debug_info = debug_info_generation(
+            struct, self.llvm_module, generated_debug_info
+        )
+        field_index = 0
+
+        # Make sure the struct has an entry in our field names dictionary
+        if struct.name not in self.generated_field_names:
+            self.generated_field_names[struct.name] = {}
+
+        for field_name, field in struct.fields.items():
+            # does not take arrays and similar types into consideration yet.
+            if field.ctype_complex_type is not None and issubclass(
+                field.ctype_complex_type, ctypes.Array
+            ):
+                array_size = field.type_size
+                containing_type = field.containing_type
+                if containing_type.__module__ == ctypes.__name__:
+                    containing_type_size = ctypes.sizeof(containing_type)
+                    if array_size == 0:
+                        field_co_re_name = self._struct_name_generator(
+                            struct, field, field_index, True, 0, containing_type_size
+                        )
+                        globvar = ir.GlobalVariable(
+                            self.llvm_module, ir.IntType(64), name=field_co_re_name
+                        )
+                        globvar.linkage = "external"
+                        globvar.set_metadata("llvm.preserve.access.index", debug_info)
+                        self.generated_field_names[struct.name][field_name] = globvar
+                        field_index += 1
+                        continue
+                    for i in range(0, array_size):
+                        field_co_re_name = self._struct_name_generator(
+                            struct, field, field_index, True, i, containing_type_size
+                        )
+                        globvar = ir.GlobalVariable(
+                            self.llvm_module, ir.IntType(64), name=field_co_re_name
+                        )
+                        globvar.linkage = "external"
+                        globvar.set_metadata("llvm.preserve.access.index", debug_info)
+                        self.generated_field_names[struct.name][field_name] = globvar
+                    field_index += 1
+            elif field.type_size is not None:
+                array_size = field.type_size
+                containing_type = field.containing_type
+                if containing_type.__module__ == "vmlinux":
+                    containing_type_size = self.handler[
+                        containing_type.__name__
+                    ].current_offset
+                    for i in range(0, array_size):
+                        field_co_re_name = self._struct_name_generator(
+                            struct, field, field_index, True, i, containing_type_size
+                        )
+                        globvar = ir.GlobalVariable(
+                            self.llvm_module, ir.IntType(64), name=field_co_re_name
+                        )
+                        globvar.linkage = "external"
+                        globvar.set_metadata("llvm.preserve.access.index", debug_info)
+                        self.generated_field_names[struct.name][field_name] = globvar
+                    field_index += 1
+            else:
+                field_co_re_name = self._struct_name_generator(
+                    struct, field, field_index
+                )
+                field_index += 1
+                globvar = ir.GlobalVariable(
+                    self.llvm_module, ir.IntType(64), name=field_co_re_name
+                )
+                globvar.linkage = "external"
+                globvar.set_metadata("llvm.preserve.access.index", debug_info)
+                self.generated_field_names[struct.name][field_name] = globvar
+        return debug_info
+
+    def _struct_name_generator(
+        self,
+        struct: DependencyNode,
+        field,
+        field_index: int,
+        is_indexed: bool = False,
+        index: int = 0,
+        containing_type_size: int = 0,
+    ) -> str:
+        # TODO: Does not support Unions as well as recursive pointer and array type naming
+        if is_indexed:
+            name = (
+                "llvm."
+                + struct.name.removeprefix("struct_")
+                + f":0:{field.offset + index * containing_type_size}"
+                + "$"
+                + f"0:{field_index}:{index}"
+            )
+            return name
+        elif struct.name.startswith("struct_"):
+            name = (
+                "llvm."
+                + struct.name.removeprefix("struct_")
+                + f":0:{field.offset}"
+                + "$"
+                + f"0:{field_index}"
+            )
+            return name
+        else:
+            print(self.handler[struct.name])
+            raise TypeError(
+                "Name generation cannot occur due to type name not starting with struct"
+            )

pythonbpf/vmlinux_parser/vmlinux_exports_handler.py

@@ -0,0 +1,255 @@
+import logging
+from typing import Any
+import ctypes
+from llvmlite import ir
+
+from pythonbpf.local_symbol import LocalSymbol
+from pythonbpf.vmlinux_parser.assignment_info import AssignmentType
+
+logger = logging.getLogger(__name__)
+
+
+class VmlinuxHandler:
+    """Handler for vmlinux-related operations"""
+
+    _instance = None
+
+    @classmethod
+    def get_instance(cls):
+        """Get the singleton instance"""
+        if cls._instance is None:
+            logger.warning("VmlinuxHandler used before initialization")
+            return None
+        return cls._instance
+
+    @classmethod
+    def initialize(cls, vmlinux_symtab):
+        """Initialize the handler with vmlinux symbol table"""
+        cls._instance = cls(vmlinux_symtab)
+        return cls._instance
+
+    def __init__(self, vmlinux_symtab):
+        """Initialize with vmlinux symbol table"""
+        self.vmlinux_symtab = vmlinux_symtab
+        logger.info(
+            f"VmlinuxHandler initialized with {len(vmlinux_symtab) if vmlinux_symtab else 0} symbols"
+        )
+
+    def is_vmlinux_enum(self, name):
+        """Check if name is a vmlinux enum constant"""
+        return (
+            name in self.vmlinux_symtab
+            and self.vmlinux_symtab[name].value_type == AssignmentType.CONSTANT
+        )
+
+    def get_struct_debug_info(self, name: str) -> Any:
+        if (
+            name in self.vmlinux_symtab
+            and self.vmlinux_symtab[name].value_type == AssignmentType.STRUCT
+        ):
+            return self.vmlinux_symtab[name].debug_info
+        else:
+            raise ValueError(f"{name} is not a vmlinux struct type")
+
+    def get_vmlinux_struct_type(self, name):
+        """Check if name is a vmlinux struct type"""
+        if (
+            name in self.vmlinux_symtab
+            and self.vmlinux_symtab[name].value_type == AssignmentType.STRUCT
+        ):
+            return self.vmlinux_symtab[name].python_type
+        else:
+            raise ValueError(f"{name} is not a vmlinux struct type")
+
+    def is_vmlinux_struct(self, name):
+        """Check if name is a vmlinux struct"""
+        return (
+            name in self.vmlinux_symtab
+            and self.vmlinux_symtab[name].value_type == AssignmentType.STRUCT
+        )
+
+    def handle_vmlinux_enum(self, name):
+        """Handle vmlinux enum constants by returning LLVM IR constants"""
+        if self.is_vmlinux_enum(name):
+            value = self.vmlinux_symtab[name].value
+            logger.info(f"Resolving vmlinux enum {name} = {value}")
+            return ir.Constant(ir.IntType(64), value), ir.IntType(64)
+        return None
+
+    def get_vmlinux_enum_value(self, name):
+        """Handle vmlinux enum constants by returning LLVM IR constants"""
+        if self.is_vmlinux_enum(name):
+            value = self.vmlinux_symtab[name].value
+            logger.info(f"The value of vmlinux enum {name} = {value}")
+            return value
+        return None
+
+    def handle_vmlinux_struct_field(
+        self, struct_var_name, field_name, module, builder, local_sym_tab
+    ):
+        """Handle access to vmlinux struct fields"""
+        if struct_var_name in local_sym_tab:
+            var_info: LocalSymbol = local_sym_tab[struct_var_name]
+            logger.info(
+                f"Attempting to access field {field_name} of possible vmlinux struct {struct_var_name}"
+            )
+            python_type: type = var_info.metadata
+            struct_name = python_type.__name__
+            globvar_ir, field_data = self.get_field_type(struct_name, field_name)
+            builder.function.args[0].type = ir.PointerType(ir.IntType(8))
+            field_ptr = self.load_ctx_field(
+                builder, builder.function.args[0], globvar_ir, field_data, struct_name
+            )
+            # Return pointer to field and field type
+            return field_ptr, field_data
+        else:
+            raise RuntimeError("Variable accessed not found in symbol table")
+
+    @staticmethod
+    def load_ctx_field(builder, ctx_arg, offset_global, field_data, struct_name=None):
+        """
+        Generate LLVM IR to load a field from BPF context using offset.
+
+        Args:
+            builder: llvmlite IRBuilder instance
+            ctx_arg: The context pointer argument (ptr/i8*)
+            offset_global: Global variable containing the field offset (i64)
+            field_data: contains data about the field
+            struct_name: Name of the struct being accessed (optional)
+        Returns:
+            The loaded value (i64 register or appropriately sized)
+        """
+
+        # Load the offset value
+        offset = builder.load(offset_global)
+
+        # Ensure ctx_arg is treated as i8* (byte pointer)
+        i8_ptr_type = ir.PointerType()
+
+        # Cast ctx_arg to i8* if it isn't already
+        if str(ctx_arg.type) != str(i8_ptr_type):
+            ctx_i8_ptr = builder.bitcast(ctx_arg, i8_ptr_type)
+        else:
+            ctx_i8_ptr = ctx_arg
+
+        # GEP with explicit type - this is the key fix
+        field_ptr = builder.gep(
+            ctx_i8_ptr,
+            [offset],
+            inbounds=False,
+        )
+
+        # Get or declare the BPF passthrough intrinsic
+        module = builder.function.module
+
+        try:
+            passthrough_fn = module.globals.get("llvm.bpf.passthrough.p0.p0")
+            if passthrough_fn is None:
+                raise KeyError
+        except (KeyError, AttributeError):
+            passthrough_type = ir.FunctionType(
+                i8_ptr_type,
+                [ir.IntType(32), i8_ptr_type],
+            )
+            passthrough_fn = ir.Function(
+                module,
+                passthrough_type,
+                name="llvm.bpf.passthrough.p0.p0",
+            )
+
+        # Call passthrough to satisfy BPF verifier
+        verified_ptr = builder.call(
+            passthrough_fn, [ir.Constant(ir.IntType(32), 0), field_ptr], tail=True
+        )
+
+        # Determine the appropriate IR type based on field information
+        int_width = 64  # Default to 64-bit
+        needs_zext = False  # Track if we need zero-extension for xdp_md
+
+        if field_data is not None:
+            # Try to determine the size from field metadata
+            if field_data.type.__module__ == ctypes.__name__:
+                try:
+                    field_size_bytes = ctypes.sizeof(field_data.type)
+                    field_size_bits = field_size_bytes * 8
+
+                    if field_size_bits in [8, 16, 32, 64]:
+                        int_width = field_size_bits
+                        logger.info(f"Determined field size: {int_width} bits")
+
+                        # Special handling for struct_xdp_md i32 fields
+                        # Load as i32 but extend to i64 before storing
+                        if struct_name == "struct_xdp_md" and int_width == 32:
+                            needs_zext = True
+                            logger.info(
+                                "struct_xdp_md i32 field detected, will zero-extend to i64"
+                            )
+                    else:
+                        logger.warning(
+                            f"Unusual field size {field_size_bits} bits, using default 64"
+                        )
+                except Exception as e:
+                    logger.warning(
+                        f"Could not determine field size: {e}, using default 64"
+                    )
+
+            elif field_data.type.__module__ == "vmlinux":
+                # For pointers to structs or complex vmlinux types
+                if field_data.ctype_complex_type is not None and issubclass(
+                    field_data.ctype_complex_type, ctypes._Pointer
+                ):
+                    int_width = 64  # Pointers are always 64-bit
+                    logger.info("Field is a pointer type, using 64 bits")
+                # TODO: Add handling for other complex types (arrays, embedded structs, etc.)
+                else:
+                    logger.warning("Complex vmlinux field type, using default 64 bits")
+
+        # Bitcast to appropriate pointer type based on determined width
+        ptr_type = ir.PointerType(ir.IntType(int_width))
+
+        typed_ptr = builder.bitcast(verified_ptr, ptr_type)
+
+        # Load and return the value
+        value = builder.load(typed_ptr)
+
+        # Zero-extend i32 to i64 for struct_xdp_md fields
+        if needs_zext:
+            value = builder.zext(value, ir.IntType(64))
+            logger.info("Zero-extended i32 value to i64 for struct_xdp_md field")
+
+        return value
+
+    def has_field(self, struct_name, field_name):
+        """Check if a vmlinux struct has a specific field"""
+        if self.is_vmlinux_struct(struct_name):
+            python_type = self.vmlinux_symtab[struct_name].python_type
+            return hasattr(python_type, field_name)
+        return False
+
+    def get_field_type(self, vmlinux_struct_name, field_name):
+        """Get the type of a field in a vmlinux struct"""
+        if self.is_vmlinux_struct(vmlinux_struct_name):
+            python_type = self.vmlinux_symtab[vmlinux_struct_name].python_type
+            if hasattr(python_type, field_name):
+                return self.vmlinux_symtab[vmlinux_struct_name].members[field_name]
+            else:
+                raise ValueError(
+                    f"Field {field_name} not found in vmlinux struct {vmlinux_struct_name}"
+                )
+        else:
+            raise ValueError(f"{vmlinux_struct_name} is not a vmlinux struct")
+
+    def get_field_index(self, vmlinux_struct_name, field_name):
+        """Get the type of a field in a vmlinux struct"""
+        if self.is_vmlinux_struct(vmlinux_struct_name):
+            python_type = self.vmlinux_symtab[vmlinux_struct_name].python_type
+            if hasattr(python_type, field_name):
+                return list(
+                    self.vmlinux_symtab[vmlinux_struct_name].members.keys()
+                ).index(field_name)
+            else:
+                raise ValueError(
+                    f"Field {field_name} not found in vmlinux struct {vmlinux_struct_name}"
+                )
+        else:
+            raise ValueError(f"{vmlinux_struct_name} is not a vmlinux struct")