ofp-actions: Waste less memory in learn actions.

Make the immediate data member 'src_imm' of a learn spec allocated at
the end of the action for just the right size.  This, together with
some structure packing saves on average of ~128 bytes for each learn
spec in each learn action.  Typical learn actions have about 4 specs
each, so this amounts to saving about 0.5kb for each learn action.

Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>

1 files changed, 28 insertions, 14 deletions

diff --git a/include/openvswitch/ofp-actions.h b/include/openvswitch/ofp-actions.h
index 310ec335b..b30b270e7 100644
--- a/include/openvswitch/ofp-actions.h
+++ b/include/openvswitch/ofp-actions.h
@@ -651,19 +651,6 @@ struct ofpact_resubmit {
     uint8_t table_id;
 };
 
-/* Part of struct ofpact_learn, below. */
-struct ofpact_learn_spec {
-    int n_bits;                 /* Number of bits in source and dest. */
-
-    int src_type;               /* One of NX_LEARN_SRC_*. */
-    struct mf_subfield src;     /* NX_LEARN_SRC_FIELD only. */
-    union mf_subvalue src_imm;  /* NX_LEARN_SRC_IMMEDIATE only. */
-
-    int dst_type;             /* One of NX_LEARN_DST_*. */
-    struct mf_subfield dst;   /* NX_LEARN_DST_MATCH, NX_LEARN_DST_LOAD only. */
-};
-
-
 /* Bits for 'flags' in struct nx_action_learn.
  *
  * If NX_LEARN_F_SEND_FLOW_REM is set, then the learned flows will have their
@@ -708,6 +695,33 @@ enum nx_learn_flags {
 #define NX_LEARN_DST_RESERVED  (3 << 11) /* Not yet defined. */
 #define NX_LEARN_DST_MASK      (3 << 11)
 
+/* Part of struct ofpact_learn, below. */
+struct ofpact_learn_spec {
+    struct mf_subfield src;    /* NX_LEARN_SRC_FIELD only. */
+    struct mf_subfield dst;    /* NX_LEARN_DST_MATCH, NX_LEARN_DST_LOAD only. */
+    uint16_t src_type;         /* One of NX_LEARN_SRC_*. */
+    uint16_t dst_type;         /* One of NX_LEARN_DST_*. */
+    uint8_t n_bits;            /* Number of bits in source and dest. */
+    uint64_t src_imm[];        /* OFPACT_ALIGNTO (uint64_t) aligned. */
+};
+BUILD_ASSERT_DECL(offsetof(struct ofpact_learn_spec, src_imm)
+                  % OFPACT_ALIGNTO == 0);
+BUILD_ASSERT_DECL(offsetof(struct ofpact_learn_spec, src_imm)
+                  == sizeof(struct ofpact_learn_spec));
+
+static inline const struct ofpact_learn_spec *
+ofpact_learn_spec_next(const struct ofpact_learn_spec *spec)
+{
+    if (spec->src_type == NX_LEARN_SRC_IMMEDIATE) {
+        unsigned int n_uint64s
+            = OFPACT_ALIGN(DIV_ROUND_UP(spec->n_bits, 8)) / sizeof (uint64_t);
+        return (const struct ofpact_learn_spec *)
+            ((const uint64_t *)(spec + 1) + n_uint64s);
+    } else {
+        return spec + 1;
+    }
+}
+
 /* OFPACT_LEARN.
  *
  * Used for NXAST_LEARN. */
@@ -718,8 +732,8 @@ struct ofpact_learn {
     uint16_t hard_timeout;      /* Max time before discarding (seconds). */
     uint16_t priority;          /* Priority level of flow entry. */
     uint8_t table_id;           /* Table to insert flow entry. */
-    ovs_be64 cookie;            /* Cookie for new flow. */
     enum nx_learn_flags flags;  /* NX_LEARN_F_*. */
+    ovs_be64 cookie;            /* Cookie for new flow. */
     uint16_t fin_idle_timeout;  /* Idle timeout after FIN, if nonzero. */
     uint16_t fin_hard_timeout;  /* Hard timeout after FIN, if nonzero. */