#define KBL_REVID_A0           0x0
 #define KBL_REVID_B0           0x1
+#define KBL_REVID_C0           0x2
+#define KBL_REVID_D0           0x3
+#define KBL_REVID_E0           0x4
 
 #define IS_KBL_REVID(p, since, until) \
        (IS_KABYLAKE(p) && IS_REVID(p, since, until))
 
                                                uint32_t *const batch,
                                                uint32_t index)
 {
+       struct drm_i915_private *dev_priv = engine->dev->dev_private;
        uint32_t l3sqc4_flush = (0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES);
 
        /*
-        * WaDisableLSQCROPERFforOCL:skl
+        * WaDisableLSQCROPERFforOCL:skl,kbl
         * This WA is implemented in skl_init_clock_gating() but since
         * this batch updates GEN8_L3SQCREG4 with default value we need to
         * set this bit here to retain the WA during flush.
         */
-       if (IS_SKL_REVID(engine->dev, 0, SKL_REVID_E0))
+       if (IS_SKL_REVID(dev_priv, 0, SKL_REVID_E0) ||
+           IS_KBL_REVID(dev_priv, 0, KBL_REVID_E0))
                l3sqc4_flush |= GEN8_LQSC_RO_PERF_DIS;
 
        wa_ctx_emit(batch, index, (MI_STORE_REGISTER_MEM_GEN8 |
 
                WA_SET_BIT_MASKED(HDC_CHICKEN0,
                                  HDC_FENCE_DEST_SLM_DISABLE);
 
+       /* GEN8_L3SQCREG4 has a dependency with WA batch so any new changes
+        * involving this register should also be added to WA batch as required.
+        */
+       if (IS_KBL_REVID(dev_priv, 0, KBL_REVID_E0))
+               /* WaDisableLSQCROPERFforOCL:kbl */
+               I915_WRITE(GEN8_L3SQCREG4, I915_READ(GEN8_L3SQCREG4) |
+                          GEN8_LQSC_RO_PERF_DIS);
+
+       /* WaDisableLSQCROPERFforOCL:kbl */
+       ret = wa_ring_whitelist_reg(engine, GEN8_L3SQCREG4);
+       if (ret)
+               return ret;
+
        return 0;
 }