[diffusion] Bound overlong weight lock filenames (#34825)

python/sglang/multimodal_gen/runtime/loader/weight_utils.py

@@ -91,6 +91,12 @@ def get_lock(model_name_or_path: str | Path, cache_dir: str | None = None):
     hash_name = hashlib.sha256(model_name.encode()).hexdigest()
     # add hash to avoid conflict with old users' lock files
     lock_file_name = hash_name + model_name + ".lock"
+    # Linux filesystems commonly cap one filename at 255 bytes. Absolute
+    # snapshot paths can exceed that even though the full path is valid.
+    # The digest is already collision-resistant, so fall back to it alone
+    # while preserving the historical name for ordinary paths.
+    if len(os.fsencode(lock_file_name)) > 255:
+        lock_file_name = hash_name + ".lock"
     # mode 0o666 is required for the filelock to be shared across users
     lock = filelock.FileLock(os.path.join(lock_dir, lock_file_name), mode=0o666)
     return lock

python/sglang/multimodal_gen/test/unit/test_weight_utils.py

@@ -1,11 +1,14 @@
 # SPDX-License-Identifier: Apache-2.0
-"""The patched find_local_ranks() must never touch torch.distributed."""
+"""Unit tests for multimodal weight-loading utilities."""
 
+import os
+import tempfile
 import unittest
 from unittest.mock import patch
 
 from sglang.multimodal_gen.runtime.loader.weight_utils import (
     _disable_runai_streamer_rank_discovery_collective,
+    get_lock,
 )
 
 _DIST_STREAMER_MOD = "runai_model_streamer.distributed_streamer.distributed_streamer"
@@ -83,5 +86,25 @@ class TestDisableRunaiStreamerRankDiscoveryCollective(unittest.TestCase):
         self.assertFalse(hasattr(_StubParams, "find_local_ranks"))
 
 
+class TestDiffusionWeightLock(unittest.TestCase):
+    def test_long_snapshot_path_uses_bounded_lock_filename(self):
+        component_path = os.path.join(
+            "/scratch",
+            "models--" + "very-long-repository-name-" * 8,
+            "snapshots",
+            "a" * 64,
+            "transformer",
+            "config.json",
+        )
+
+        with tempfile.TemporaryDirectory() as lock_dir:
+            lock = get_lock(component_path, lock_dir)
+            lock_filename = os.path.basename(lock.lock_file)
+
+            self.assertLessEqual(len(os.fsencode(lock_filename)), 255)
+            with lock:
+                self.assertTrue(os.path.exists(lock.lock_file))
+
+
 if __name__ == "__main__":
     unittest.main()