Fix abusing presence of req.req_pool_idx to indicate the presence of req.kv resources (#29430)

python/sglang/srt/disaggregation/prefill.py

@@ -929,7 +929,11 @@ class SchedulerDisaggregationPrefillMixin:
         else:
             logger.warning(error_message)
         req.time_stats.trace_ctx.abort(abort_info={"reason": error_message})
-        if req.req_pool_idx is not None or self.tree_cache.supports_mamba():
+        if (
+            req.req_pool_idx is not None
+            or req.kv is not None
+            or req.mamba_pool_idx is not None
+        ):
             release_kv_cache(req, self.tree_cache)
         maybe_release_metadata_buffer(req, self.req_to_metadata_buffer_idx_allocator)
         req.pending_bootstrap = False

python/sglang/srt/managers/scheduler.py

@@ -2564,8 +2564,7 @@ class Scheduler(
             req.pending_bootstrap = False
         if self.enable_hicache_storage:
             self.tree_cache.release_aborted_request(req.rid)
-        if req.req_pool_idx is not None or self.tree_cache.supports_mamba():
+        release_kv_cache(req, self.tree_cache, is_insert=False)
-            release_kv_cache(req, self.tree_cache, is_insert=False)
 
         self.chunked_req = None
         self._pending_chunked_abort_req = None

python/sglang/srt/managers/scheduler_components/invariant_checker.py

@@ -246,7 +246,7 @@ class SchedulerInvariantChecker:
         swa_uncached = 0
         for batch in batches:
             for req in batch.reqs:
-                if req.req_pool_idx is None:
+                if req.kv is None:
                     continue
 
                 allocated_len = req.kv.kv_allocated_len

python/sglang/srt/mem_cache/common.py

@@ -630,6 +630,8 @@ def alloc_for_decode(batch: ScheduleBatch, token_per_req: int) -> torch.Tensor:
 
 
 def release_kv_cache(req: Req, tree_cache: BasePrefixCache, is_insert: bool = True):
+    # the two resources currently have the same lifecycle, thus simplify logic below
+    assert (req.req_pool_idx is None) == (req.kv is None)
     # MambaRadixCache may alloc mamba state before alloc KV cache
     if req.req_pool_idx is None:
         assert (
@@ -652,7 +654,8 @@ def release_kv_cache(req: Req, tree_cache: BasePrefixCache, is_insert: bool = Tr
 
     # StreamingSession.cache_finished_req handles speculative tail trim
     # internally, then sets req_pool_idx = None.
-    if req.req_pool_idx is None:
+    assert (req.req_pool_idx is None) == (req.kv is None)
+    if req.req_pool_idx is None and req.kv is None:
         return
 
     start_p, end_p = effective_kv_committed_len, req.kv.kv_allocated_len

python/sglang/srt/session/streaming_session.py

@@ -63,7 +63,7 @@ class SessionSlot:
     @property
     def is_holding_kv(self) -> bool:
         """Whether this slot currently holds KV pool resources."""
-        return self.req_pool_idx is not None
+        return self.kv is not None
 
     def save_from_req(self, req: Req, is_first: bool):
         """Save KV state from a finishing request into this slot."""
@@ -212,7 +212,7 @@ class StreamingSession(BasePrefixCache):
         if not _is_streaming(req):
             return None
         slot = self.slots.get(req.session.session_id)
-        if slot is None or slot.req_pool_idx is None:
+        if slot is None or slot.kv is None:
             return None
         if req.to_finish is not None:
             req.session.abort_req()

python/sglang/test/scripted_runtime/req_handle.py

@@ -42,7 +42,7 @@ class ScriptedReqHandle:
     @property
     def kv_pages(self) -> int:
         req = self.req
-        if req is None or req.req_pool_idx is None:
+        if req is None or req.kv is None:
             return 0
         page_size = self.context.scheduler.page_size
         return (req.kv.kv_allocated_len + page_size - 1) // page_size