From 75c8f53cfc24508b0b5699902a2ffad6cfa90918 Mon Sep 17 00:00:00 2001
From: c020627 <1326713348@qq.com>
Date: Tue, 29 Sep 2026 22:41:53 +0800
Subject: [PATCH] fix(voice): cap DashScope TTS audio download size to avoid
 disk-fill DoS
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

DashScopeVoice._download_audio buffered the whole response body with no size
limit, so a hostile or broken TTS audio URL could fill disk (probe P4 —
unbounded media download). Route it through the size-capped
common.media_download.download_to_file (100 MB cap); the import is aliased
because this module already defines its own MAX_FILE_BYTES for ASR input.

The method already degrades to None on any download failure, so an oversized
response is refused the same way. Adds a regression test.

Co-Authored-By: WorkBuddy <noreply@workbuddy.ai>
---
 tests/test_dashscope_voice_download.py | 31 ++++++++++++++++++++++++++
 voice/dashscope/dashscope_voice.py     |  7 ++----
 2 files changed, 33 insertions(+), 5 deletions(-)
 create mode 100644 tests/test_dashscope_voice_download.py

diff --git a/tests/test_dashscope_voice_download.py b/tests/test_dashscope_voice_download.py
new file mode 100644
index 00000000..36765462
--- /dev/null
+++ b/tests/test_dashscope_voice_download.py
@@ -0,0 +1,31 @@
+# encoding:utf-8
+"""Bounded-download guard for ``DashScopeVoice._download_audio`` (probe P4).
+
+Regression guard: before the size-capped downloader was wired in,
+``_download_audio`` buffered ``resp.content`` with no upper bound, so a hostile
+or broken TTS audio URL could fill disk. The downloader now raises
+``MediaTooLargeError`` past the cap, which the method degrades from (returns
+None) like any other download failure.
+"""
+import requests
+from common.media_download import MAX_FILE_BYTES
+from unittest.mock import MagicMock, patch
+
+from voice.dashscope.dashscope_voice import DashScopeVoice
+
+
+def test_download_audio_rejects_oversized_body():
+    url = "http://example.test/audio.mp3"
+    resp = MagicMock()
+    resp.raise_for_status.return_value = None
+    # content is set only so the *old* code path (which wrote resp.content)
+    # can complete without a TypeError; the size-capped downloader aborts on
+    # the Content-Length pre-check before reading the stream.
+    resp.content = b""
+    resp.headers = {"Content-Length": str(MAX_FILE_BYTES + 1)}
+    resp.close.return_value = None
+
+    with patch.object(requests, "get", return_value=resp):
+        # On an oversized response the method must degrade to None rather than
+        # writing an unbounded body to disk.
+        assert DashScopeVoice._download_audio(url) is None
diff --git a/voice/dashscope/dashscope_voice.py b/voice/dashscope/dashscope_voice.py
index e1f0770c..97e25173 100644
--- a/voice/dashscope/dashscope_voice.py
+++ b/voice/dashscope/dashscope_voice.py
@@ -7,11 +7,11 @@ import random
 from typing import Optional
 
 import dashscope
-import requests
 from dashscope import MultiModalConversation
 
 from bridge.reply import Reply, ReplyType
 from common.log import logger
+from common.media_download import MAX_FILE_BYTES as _MEDIA_MAX_BYTES, download_to_file as _media_download_to_file
 from common.tmp_dir import TmpDir
 from config import conf
 from voice import audio_convert
@@ -128,10 +128,7 @@ class DashScopeVoice(Voice):
             if ext not in (".mp3", ".wav", ".m4a", ".aac", ".opus"):
                 ext = ".wav"
             dst = os.path.join(tmp_dir, f"dashscope_tts_{ts}_{random.randint(0, 9999)}{ext}")
-            resp = requests.get(url, timeout=60)
-            resp.raise_for_status()
-            with open(dst, "wb") as f:
-                f.write(resp.content)
+            _media_download_to_file(url, dst, _MEDIA_MAX_BYTES, timeout=60)
             return dst
         except Exception as e:
             logger.error(f"[DashScopeVoice] download audio failed: {e}")