From 5559679229bc961848b121ccdeaa8fa5d79bec98 Mon Sep 17 00:00:00 2001
From: coltonottley <coltonottley@gmail.com>
Date: Sun, 26 Jul 2026 05:53:32 -0600
Subject: [PATCH] [Bugfix][KV Offload] Bound unaligned SWA loads by physical
 GPU blocks (#49052)

Signed-off-by: Colton Ottley <colton@ottleyengineering.com>
Co-authored-by: Colton Ottley <colton@ottleyengineering.com>
Co-authored-by: jasl <jasl9187@hotmail.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: Or Ozeri <oro@il.ibm.com>
---
 .../kv_transfer/kv_connector/v1/offloading/scheduler.py          | 1 +
 1 file changed, 1 insertion(+)

diff --git a/vllm/distributed/kv_transfer/kv_connector/v1/offloading/scheduler.py b/vllm/distributed/kv_transfer/kv_connector/v1/offloading/scheduler.py
index b53a365abd..6ccc820f25 100644
--- a/vllm/distributed/kv_transfer/kv_connector/v1/offloading/scheduler.py
+++ b/vllm/distributed/kv_transfer/kv_connector/v1/offloading/scheduler.py
@@ -860,6 +860,7 @@ class OffloadingConnectorScheduler:
                     num_pending_gpu_blocks
                     <= group_config.sliding_window_size_in_chunks
                     * self.config.blocks_per_chunk
+                    + 1
                 )
 
             num_chunks = cdiv(num_cached_tokens, tokens_per_chunk)