is SADB_X_EXT_FLOW_TYPE IPSP_DIRECTION_IN inbound SADB_EXT_ADDRESS_{DST,SRC} reversed?

[Andrew Cagney]

PFKEYv2 has the general convention that SRC=from-address and
DST=to-address.  This means that an incoming packet has
SRC==peer-address and DST==local-address.
However, in IKED I've noticed:
                /* local address */
                bzero(&sa_local, sizeof(sa_local));
                sa_local.sadb_address_exttype = SADB_EXT_ADDRESS_SRC;
                sa_local.sadb_address_len =
                    (sizeof(sa_local) + ROUNDUP(slocal.ss_len)) / 8;
                /* peer address */
                bzero(&sa_peer, sizeof(sa_peer));
                sa_peer.sadb_address_exttype = SADB_EXT_ADDRESS_DST;
                sa_peer.sadb_address_len =
                    (sizeof(sa_peer) + ROUNDUP(speer.ss_len)) / 8;
i.e., a flow from SRC-to-DST ignores of the direction and always has
SRC==local and DST=peer/remote.  This seems wrong.

With libreswan I ended up mimicing this vis:
                case KERNEL_POLICY_DIR_INBOUND:
                        /* XXX: notice how DST gets SRC's value et.al. */
                        put_sadb_address(&req, SADB_EXT_ADDRESS_DST,
kernel_policy->src.host);
                        put_sadb_address(&req, SADB_EXT_ADDRESS_SRC,
kernel_policy->dst.host);
                        break;