diff --git a/src/intel/compiler/brw_fs_lower_regioning.cpp b/src/intel/compiler/brw_fs_lower_regioning.cpp
index 89f3d20171a..e0a10e17cde 100644
--- a/src/intel/compiler/brw_fs_lower_regioning.cpp
+++ b/src/intel/compiler/brw_fs_lower_regioning.cpp
@@ -58,9 +58,8 @@ namespace {
          return MAX2(brw_type_size_bytes(inst->dst.type),
                      byte_stride(inst->dst));
 
-      } else if (has_subdword_integer_region_restriction(devinfo, inst) &&
-                 brw_type_size_bytes(inst->src[i].type) < 4 &&
-                 byte_stride(inst->src[i]) >= 4) {
+      } else if (has_subdword_integer_region_restriction(devinfo, inst,
+                                                         &inst->src[i], 1)) {
          /* Use a stride of 32bits if possible, since that will guarantee that
           * the copy emitted to lower this region won't be affected by the
           * sub-dword integer region restrictions.  This may not be possible
@@ -85,9 +84,8 @@ namespace {
       if (has_dst_aligned_region_restriction(devinfo, inst)) {
          return reg_offset(inst->dst) % (reg_unit(devinfo) * REG_SIZE);
 
-      } else if (has_subdword_integer_region_restriction(devinfo, inst) &&
-                 brw_type_size_bytes(inst->src[i].type) < 4 &&
-                 byte_stride(inst->src[i]) >= 4) {
+      } else if (has_subdword_integer_region_restriction(devinfo, inst,
+                                                         &inst->src[i], 1)) {
          const unsigned dst_byte_stride =
             MAX2(byte_stride(inst->dst), brw_type_size_bytes(inst->dst.type));
          const unsigned src_byte_stride = required_src_byte_stride(devinfo, inst, i);
@@ -650,9 +648,16 @@ namespace {
                         subscript(inst->dst, raw_type, j));
          }
 
-         for (unsigned j = 0; j < n; j++)
-            ibld.at(block, inst->next).MOV(subscript(inst->dst, raw_type, j),
-                                           subscript(tmp, raw_type, j));
+         for (unsigned j = 0; j < n; j++) {
+            fs_inst *jnst = ibld.at(block, inst->next).MOV(subscript(inst->dst, raw_type, j),
+                                                           subscript(tmp, raw_type, j));
+            if (has_subdword_integer_region_restriction(v->devinfo, jnst)) {
+               /* The copy isn't guaranteed to comply with all subdword integer
+                * regioning restrictions in some cases.  Lower it recursively.
+                */
+               lower_instruction(v, block, jnst);
+            }
+         }
 
          /* If the destination was an accumulator, after lowering it will be a
           * GRF. Clear writes_accumulator for the instruction.
diff --git a/src/intel/compiler/brw_ir_fs.h b/src/intel/compiler/brw_ir_fs.h
index b0120178503..997927c9ae9 100644
--- a/src/intel/compiler/brw_ir_fs.h
+++ b/src/intel/compiler/brw_ir_fs.h
@@ -445,8 +445,12 @@ has_subdword_integer_region_restriction(const intel_device_info *devinfo,
             brw_type_size_bytes(inst->dst.type)) < 4) {
       for (unsigned i = 0; i < num_srcs; i++) {
          if (brw_type_is_int(srcs[i].type) &&
-             brw_type_size_bytes(srcs[i].type) < 4 &&
-             byte_stride(srcs[i]) >= 4)
+             ((brw_type_size_bytes(srcs[i].type) < 4 &&
+               byte_stride(srcs[i]) >= 4) ||
+              (MAX2(byte_stride(inst->dst),
+                   brw_type_size_bytes(inst->dst.type)) == 1 &&
+               brw_type_size_bytes(srcs[i].type) == 1 &&
+               byte_stride(srcs[i]) >= 2)))
             return true;
       }
    }