Fix critical race conditions and unsafe disk report command

services/disk_report.py

@@ -1,11 +1,48 @@
 import os
+import re
 from typing import Any
 
 from services.runner import run_cmd
 
 
 def _top_dirs_cmd(path: str, limit: int) -> list[str]:
-    return ["bash", "-lc", f"du -xhd1 {path} 2>/dev/null | sort -h | tail -n {limit}"]
+    _ = limit
+    return ["du", "-x", "-h", "-d", "1", path]
+
+
+_SIZE_RE = re.compile(r"^\s*([0-9]+(?:\.[0-9]+)?)([KMGTP]?)(i?B?)?$", re.IGNORECASE)
+
+
+def _size_to_bytes(value: str) -> float:
+    m = _SIZE_RE.match(value.strip())
+    if not m:
+        return -1.0
+    num = float(m.group(1))
+    unit = (m.group(2) or "").upper()
+    mul = {
+        "": 1,
+        "K": 1024,
+        "M": 1024**2,
+        "G": 1024**3,
+        "T": 1024**4,
+        "P": 1024**5,
+    }.get(unit, 1)
+    return num * mul
+
+
+def _format_top_dirs(raw: str, limit: int) -> str:
+    rows: list[tuple[float, str]] = []
+    for line in raw.splitlines():
+        line = line.strip()
+        if not line:
+            continue
+        parts = line.split(maxsplit=1)
+        if len(parts) != 2:
+            continue
+        size, name = parts
+        rows.append((_size_to_bytes(size), f"{size}\t{name}"))
+    rows.sort(key=lambda x: x[0])
+    return "\n".join(line for _sz, line in rows[-max(1, limit):])
 
 
 async def build_disk_report(cfg: dict[str, Any], mount: str, usage: int) -> str:
@@ -15,24 +52,27 @@ async def build_disk_report(cfg: dict[str, Any], mount: str, usage: int) -> str:
 
     rc, out = await run_cmd(_top_dirs_cmd(mount, limit), timeout=30)
     if rc == 0 and out.strip():
+        top_out = _format_top_dirs(out, limit)
         lines.append("")
         lines.append("Top directories:")
-        lines.append(out.strip())
+        lines.append(top_out)
 
     docker_dir = cfg.get("disk_report", {}).get("docker_dir", "/var/lib/docker")
     if docker_dir and os.path.exists(docker_dir):
         rc2, out2 = await run_cmd(_top_dirs_cmd(docker_dir, limit), timeout=30)
         if rc2 == 0 and out2.strip():
+            top_out2 = _format_top_dirs(out2, limit)
             lines.append("")
             lines.append(f"Docker dir: {docker_dir}")
-            lines.append(out2.strip())
+            lines.append(top_out2)
 
     logs_dir = cfg.get("disk_report", {}).get("logs_dir", "/var/log")
     if logs_dir and os.path.exists(logs_dir):
         rc3, out3 = await run_cmd(_top_dirs_cmd(logs_dir, limit), timeout=30)
         if rc3 == 0 and out3.strip():
+            top_out3 = _format_top_dirs(out3, limit)
             lines.append("")
             lines.append(f"Logs dir: {logs_dir}")
-            lines.append(out3.strip())
+            lines.append(top_out3)
 
     return "\n".join(lines)

services/runtime_state.py

@@ -1,9 +1,13 @@
 import json
 import os
+import threading
+import tempfile
 from typing import Any, Dict
 
 _PATH = "/var/server-bot/runtime.json"
 _STATE: Dict[str, Any] = {}
+_LOCK = threading.RLock()
+_LOADED = False
 
 
 def configure(path: str | None):
@@ -13,40 +17,57 @@ def configure(path: str | None):
 
 
 def _load_from_disk():
-    global _STATE
+    global _STATE, _LOADED
     if not os.path.exists(_PATH):
         _STATE = {}
+        _LOADED = True
         return
     try:
         with open(_PATH, "r", encoding="utf-8") as f:
             _STATE = json.load(f)
     except Exception:
         _STATE = {}
+    _LOADED = True
 
 
 def _save():
-    os.makedirs(os.path.dirname(_PATH), exist_ok=True)
+    directory = os.path.dirname(_PATH) or "."
+    os.makedirs(directory, exist_ok=True)
     try:
-        with open(_PATH, "w", encoding="utf-8") as f:
-            json.dump(_STATE, f)
+        fd, tmp_path = tempfile.mkstemp(prefix=".runtime.", suffix=".json", dir=directory)
+        try:
+            with os.fdopen(fd, "w", encoding="utf-8") as f:
+                json.dump(_STATE, f, ensure_ascii=False)
+                f.flush()
+                os.fsync(f.fileno())
+            os.replace(tmp_path, _PATH)
+        finally:
+            if os.path.exists(tmp_path):
+                try:
+                    os.unlink(tmp_path)
+                except Exception:
+                    pass
     except Exception:
         pass
 
 
 def get_state() -> Dict[str, Any]:
-    if not _STATE:
-        _load_from_disk()
-    return _STATE
+    with _LOCK:
+        if not _LOADED:
+            _load_from_disk()
+        return _STATE
 
 
 def set_state(key: str, value: Any):
-    if not _STATE:
-        _load_from_disk()
-    _STATE[key] = value
-    _save()
+    with _LOCK:
+        if not _LOADED:
+            _load_from_disk()
+        _STATE[key] = value
+        _save()
 
 
 def get(key: str, default: Any = None) -> Any:
-    if not _STATE:
-        _load_from_disk()
-    return _STATE.get(key, default)
+    with _LOCK:
+        if not _LOADED:
+            _load_from_disk()
+        return _STATE.get(key, default)