adjust ctrl memory cache committing condition to be more robust to bad regions. otherwise we were flooding new request to repeatedly recopy from the client, when we effectively already knew it would never be valid.

2026-06-17 01:22:24 -07:00 · 2025-06-02 15:28:26 -07:00
parent 44280f3dd0
commit 594594ec72
1 changed files with 1 additions and 1 deletions
@@ -6745,7 +6745,7 @@ ASYNC_WORK_DEF(ctrl_mem_stream_work)
        {
          if(hs_id_match(range_n->id, key.id))
          {
-            if(!u128_match(u128_zero(), hash))
+            if(pre_read_mem_gen == post_read_mem_gen)
            {
              range_n->mem_gen = post_read_mem_gen;
            }