Avoid overwrite in clib_memcpy_le{32,64}

Change-Id: Id4a8b6a31fc3e88af2f075cb97c85d3f9b738d9e
Signed-off-by: Damjan Marion <damarion@cisco.com>

1 files changed, 29 insertions, 27 deletions

diff --git a/src/vppinfra/string.h b/src/vppinfra/string.h
index 4755a9868d6..49848e92f39 100644
--- a/src/vppinfra/string.h
+++ b/src/vppinfra/string.h
@@ -217,63 +217,65 @@ static_always_inline void
 clib_memcpy_le (u8 * dst, u8 * src, u8 len, u8 max_len)
 {
 #if defined (CLIB_HxAVE_VEC256)
-  u8x32 s, d;
+  u8x32 s0, s1, d0, d1;
   u8x32 mask = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17,
     18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31
   };
   u8x32 lv = u8x32_splat (len);
   u8x32 add = u8x32_splat (32);
 
-  s = u8x32_load_unaligned (src);
-  d = u8x32_load_unaligned (dst);
-  d = u8x32_blend (d, s, u8x32_is_greater (lv, mask));
-  u8x32_store_unaligned (d, dst);
+  s0 = u8x32_load_unaligned (src);
+  s1 = u8x32_load_unaligned (src + 32);
+  d0 = u8x32_load_unaligned (dst);
+  d1 = u8x32_load_unaligned (dst + 32);
+
+  d0 = u8x32_blend (d0, s0, u8x32_is_greater (lv, mask));
+  u8x32_store_unaligned (d0, dst);
 
   if (max_len <= 32)
     return;
 
   mask += add;
-  s = u8x32_load_unaligned (src + 32);
-  d = u8x32_load_unaligned (dst + 32);
-  d = u8x32_blend (d, s, u8x32_is_greater (lv, mask));
-  u8x32_store_unaligned (d, dst + 32);
+  d1 = u8x32_blend (d1, s1, u8x32_is_greater (lv, mask));
+  u8x32_store_unaligned (d1, dst + 32);
 
 #elif defined (CLIB_HAVE_VEC128) && !defined (__aarch64__)
-  u8x16 s, d;
+  u8x16 s0, s1, s2, s3, d0, d1, d2, d3;
   u8x16 mask = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 };
   u8x16 lv = u8x16_splat (len);
   u8x16 add = u8x16_splat (16);
 
-  s = u8x16_load_unaligned (src);
-  d = u8x16_load_unaligned (dst);
-  d = u8x16_blend (d, s, u8x16_is_greater (lv, mask));
-  u8x16_store_unaligned (d, dst);
+  s0 = u8x16_load_unaligned (src);
+  s1 = u8x16_load_unaligned (src + 16);
+  s2 = u8x16_load_unaligned (src + 32);
+  s3 = u8x16_load_unaligned (src + 48);
+  d0 = u8x16_load_unaligned (dst);
+  d1 = u8x16_load_unaligned (dst + 16);
+  d2 = u8x16_load_unaligned (dst + 32);
+  d3 = u8x16_load_unaligned (dst + 48);
+
+  d0 = u8x16_blend (d0, s0, u8x16_is_greater (lv, mask));
+  u8x16_store_unaligned (d0, dst);
 
   if (max_len <= 16)
     return;
 
   mask += add;
-  s = u8x16_load_unaligned (src + 16);
-  d = u8x16_load_unaligned (dst + 16);
-  d = u8x16_blend (d, s, u8x16_is_greater (lv, mask));
-  u8x16_store_unaligned (d, dst + 16);
+  d1 = u8x16_blend (d1, s1, u8x16_is_greater (lv, mask));
+  u8x16_store_unaligned (d1, dst + 16);
 
   if (max_len <= 32)
     return;
 
   mask += add;
-  s = u8x16_load_unaligned (src + 32);
-  d = u8x16_load_unaligned (dst + 32);
-  d = u8x16_blend (d, s, u8x16_is_greater (lv, mask));
-  u8x16_store_unaligned (d, dst + 32);
+  d2 = u8x16_blend (d2, s2, u8x16_is_greater (lv, mask));
+  u8x16_store_unaligned (d2, dst + 32);
 
   mask += add;
-  s = u8x16_load_unaligned (src + 48);
-  d = u8x16_load_unaligned (dst + 48);
-  d = u8x16_blend (d, s, u8x16_is_greater (lv, mask));
-  u8x16_store_unaligned (d, dst + 48);
+  d3 = u8x16_blend (d3, s3, u8x16_is_greater (lv, mask));
+  u8x16_store_unaligned (d3, dst + 48);
 #else
-  clib_memcpy_fast (dst, src, len);
+  memmove (dst, src, len);
 #endif
 }