Avoid remap-cache aliasing across tensor views

The CP shared-KV remap caches may receive tensor views that share the same storage but represent different logical page rows.  Keying only by storage pointer and shape can reuse a stale remap for another view, corrupting cache-hit materialization.  Use the actual tensor data pointer plus stride, storage offset, and version so different views and mutations are not collapsed into one cache entry.

Constraint: CP shared-KV bs>1 cache-hit paths reuse small tensor views over shared backing tensors.

Rejected: Clear the remap cache on every call | would avoid aliasing but add avoidable hot-path churn and hide the identity bug.

Confidence: high

Scope-risk: narrow

Directive: Do not reduce the remap cache key back to storage pointer only; same-storage views are semantically distinct here.

Tested: g0034 cjy-glm5-new PYTHONPATH=python python -m pytest -q test/registered/unit/mem_cache/test_cp_shared_kv_runtime.py::TestCpSharedKVRuntimeHelpers::test_token_slot_remap_cache_distinguishes_same_storage_views test/registered/unit/mem_cache/test_cp_shared_kv_runtime.py::TestCpSharedKVRuntimeHelpers::test_paged_slot_remap_cache_distinguishes_same_storage_views test/registered/unit/speculative/test_eagle_worker_v2_cp_hidden.py

Not-tested: local pytest for CP runtime import is blocked by missing starlette in the local base environment.
(cherry picked from commit 7360ef13565dbf21428e3b121112135a2955f913)

python/sglang/srt/layers/attention/nsa/cp_shared_kv_runtime.py

@@ -241,12 +241,21 @@ class SharedPagedBufferSlotRemap:
     )
 
 
-def _tensor_identity_key(tensor: torch.Tensor) -> tuple[int, tuple[int, ...], str, str]:
+def _tensor_identity_key(
+    tensor: torch.Tensor,
+) -> tuple[int, tuple[int, ...], tuple[int, ...], int, str, str, int | None]:
+    try:
+        version = int(tensor._version)
+    except Exception:
+        version = None
     return (
-        int(tensor.untyped_storage().data_ptr()),
+        int(tensor.data_ptr()),
         tuple(int(dim) for dim in tensor.shape),
+        tuple(int(stride) for stride in tensor.stride()),
+        int(tensor.storage_offset()),
         str(tensor.dtype),
         str(tensor.device),
+        version,
     )
 
 

test/registered/unit/mem_cache/test_cp_shared_kv_runtime.py

@@ -4641,6 +4641,41 @@ class TestCpSharedKVRuntimeHelpers(unittest.TestCase):
             )
         )
 
+    def test_token_slot_remap_cache_distinguishes_same_storage_views(self):
+        from sglang.srt.layers.attention.nsa import cp_shared_kv_runtime as runtime
+        from sglang.srt.mem_cache.cp_shared_kv_layout import CpSharedKVLayout
+
+        layout = CpSharedKVLayout(page_size=4, cp_size=1, cp_rank=0)
+        forward_batch = SimpleNamespace(batch_size=1, forward_mode="extend")
+        kv_cache = torch.zeros((64, 1, 1), dtype=torch.float32)
+        backing_pages = torch.tensor(
+            [
+                [1, 2, 3],
+                [9, 10, 11],
+            ],
+            dtype=torch.int64,
+        )
+        logical_pages_a = backing_pages[:1]
+        logical_pages_b = backing_pages[1:]
+
+        remap_a = runtime.get_or_build_shared_token_kv_slot_remap(
+            forward_batch,
+            kv_cache=kv_cache,
+            remap_logical_pages=logical_pages_a,
+            layout=layout,
+            page_size=4,
+        )
+        remap_b = runtime.get_or_build_shared_token_kv_slot_remap(
+            forward_batch,
+            kv_cache=kv_cache,
+            remap_logical_pages=logical_pages_b,
+            layout=layout,
+            page_size=4,
+        )
+
+        self.assertIsNot(remap_a, remap_b)
+        self.assertEqual(remap_b.slot_logical_pages.tolist(), [9, 10, 11])
+
     def test_token_slot_remap_incomplete_cache_state_logs(self):
         from sglang.srt.layers.attention.nsa import cp_shared_kv_runtime as runtime
         from sglang.srt.mem_cache.cp_shared_kv_layout import CpSharedKVLayout
@@ -4715,6 +4750,39 @@ class TestCpSharedKVRuntimeHelpers(unittest.TestCase):
         self.assertTrue(torch.equal(dense_page_buffer[2], page_buffer[2]))
         self.assertTrue(torch.equal(dense_page_buffer[3], page_buffer[5]))
 
+    def test_paged_slot_remap_cache_distinguishes_same_storage_views(self):
+        from sglang.srt.layers.attention.nsa import cp_shared_kv_runtime as runtime
+        from sglang.srt.mem_cache.cp_shared_kv_layout import CpSharedKVLayout
+
+        layout = CpSharedKVLayout(page_size=4, cp_size=1, cp_rank=0)
+        forward_batch = SimpleNamespace(batch_size=1, forward_mode="extend")
+        page_buffer = torch.zeros((16, 3), dtype=torch.uint8)
+        backing_pages = torch.tensor(
+            [
+                [1, 2, 3],
+                [9, 10, 11],
+            ],
+            dtype=torch.int64,
+        )
+        logical_pages_a = backing_pages[:1]
+        logical_pages_b = backing_pages[1:]
+
+        remap_a = runtime.get_or_build_shared_paged_buffer_slot_remap(
+            forward_batch,
+            page_buffer=page_buffer,
+            logical_pages=logical_pages_a,
+            layout=layout,
+        )
+        remap_b = runtime.get_or_build_shared_paged_buffer_slot_remap(
+            forward_batch,
+            page_buffer=page_buffer,
+            logical_pages=logical_pages_b,
+            layout=layout,
+        )
+
+        self.assertIsNot(remap_a, remap_b)
+        self.assertEqual(remap_b.slot_logical_pages.tolist(), [9, 10, 11])
+
     def test_paged_slot_remap_cache_miss_logs_after_warm_cache(self):
         from sglang.srt.layers.attention.nsa import cp_shared_kv_runtime as runtime
         from sglang.srt.mem_cache.cp_shared_kv_layout import CpSharedKVLayout