diff --git a/vllm/v1/core/kv_cache_utils.py b/vllm/v1/core/kv_cache_utils.py index 88e60987ba..aa42f90bb9 100644 --- a/vllm/v1/core/kv_cache_utils.py +++ b/vllm/v1/core/kv_cache_utils.py @@ -407,6 +407,20 @@ class FreeKVCacheBlockQueue: curr_block = curr_block.next_free_block return ret + def iter_blocks_after( + self, + cursor: KVCacheBlock | None, + ) -> Iterator[KVCacheBlock]: + """Iterate free blocks in eviction order after the cursor.""" + if cursor is None: + curr_block = self.fake_free_list_head.next_free_block + else: + curr_block = cursor.next_free_block + + while curr_block is not None and curr_block is not self.fake_free_list_tail: + yield curr_block + curr_block = curr_block.next_free_block + def need_extra_keys(request: Request) -> bool: """Check whether the blocks allocated to this request need extra hash keys. diff --git a/vllm/v1/simple_kv_offload/manager.py b/vllm/v1/simple_kv_offload/manager.py index 44ba902f45..f451f503d9 100644 --- a/vllm/v1/simple_kv_offload/manager.py +++ b/vllm/v1/simple_kv_offload/manager.py @@ -487,24 +487,14 @@ class SimpleCPUOffloadScheduler: if self._cursor is not None and self._cursor.ref_cnt > 0: self._cursor = None - # Determine start node. - if self._cursor is None: - node = free_queue.fake_free_list_head.next_free_block - else: - node = self._cursor.next_free_block - - tail = free_queue.fake_free_list_tail gpu_ids: list[int] = [] block_hashes: list[bytes] = [] - covered = 0 last_visited = self._cursor - while ( - node is not None - and node is not tail - and covered < self._target_free - and len(gpu_ids) < num_cpu_free - ): + for covered, node in enumerate(free_queue.iter_blocks_after(self._cursor)): + if covered >= self._target_free or len(gpu_ids) >= num_cpu_free: + break + last_visited = node bhash = node.block_hash @@ -516,9 +506,6 @@ class SimpleCPUOffloadScheduler: gpu_ids.append(node.block_id) block_hashes.append(bhash) - covered += 1 - node = node.next_free_block - self._cursor = last_visited # Batch-allocate CPU blocks and stamp hashes.