summaryrefslogtreecommitdiffstatshomepage
diff options
context:
space:
mode:
-rw-r--r--src/crypto/zinc/poly1305/poly1305-arm-glue.h11
-rw-r--r--src/crypto/zinc/poly1305/poly1305-x86_64-glue.h9
2 files changed, 20 insertions, 0 deletions
diff --git a/src/crypto/zinc/poly1305/poly1305-arm-glue.h b/src/crypto/zinc/poly1305/poly1305-arm-glue.h
index c694524..92aa226 100644
--- a/src/crypto/zinc/poly1305/poly1305-arm-glue.h
+++ b/src/crypto/zinc/poly1305/poly1305-arm-glue.h
@@ -50,6 +50,17 @@ struct poly1305_arch_internal {
};
#endif
+/* The NEON code uses base 2^26, while the scalar code uses base 2^64 on 64-bit
+ * and base 2^32 on 32-bit. If we hit the unfortunate situation of using NEON
+ * and then having to go back to scalar -- because the user is silly and has
+ * called the update function from two separate contexts -- then we need to
+ * convert back to the original base before proceeding. The below function is
+ * written for 64-bit integers, and so we have to swap words at the end on
+ * big-endian 32-bit. It is possible to reason that the initial reduction below
+ * is sufficient given the implementation invariants. However, for an avoidance
+ * of doubt and because this is not performance critical, we do the full
+ * reduction anyway.
+ */
static void convert_to_base2_64(void *ctx)
{
struct poly1305_arch_internal *state = ctx;
diff --git a/src/crypto/zinc/poly1305/poly1305-x86_64-glue.h b/src/crypto/zinc/poly1305/poly1305-x86_64-glue.h
index b1248e8..585b579 100644
--- a/src/crypto/zinc/poly1305/poly1305-x86_64-glue.h
+++ b/src/crypto/zinc/poly1305/poly1305-x86_64-glue.h
@@ -67,6 +67,15 @@ struct poly1305_arch_internal {
struct { u32 r2, r1, r4, r3; } rn[9];
};
+/* The AVX code uses base 2^26, while the scalar code uses base 2^64. If we hit
+ * the unfortunate situation of using AVX and then having to go back to scalar
+ * -- because the user is silly and has called the update function from two
+ * separate contexts -- then we need to convert back to the original base before
+ * proceeding. It is possible to reason that the initial reduction below is
+ * sufficient given the implementation invariants. However, for an avoidance of
+ * doubt and because this is not performance critical, we do the full reduction
+ * anyway.
+ */
static void convert_to_base2_64(void *ctx)
{
struct poly1305_arch_internal *state = ctx;