Content-Type: multipart/mixed; boundary="===============5435079094861835384==" MIME-Version: 1.0 From: Denis Kenzior Subject: Re: [PATCH v2 1/4] netdev: allow SAE reassociation Date: Thu, 05 Aug 2021 12:23:55 -0500 Message-ID: <4313fd3e-026a-041c-908a-6bffe5aaff9b@gmail.com> In-Reply-To: <20210805165914.293926-1-prestwoj@gmail.com> List-Id: To: iwd@lists.01.org --===============5435079094861835384== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Hi James, On 8/5/21 11:59 AM, James Prestwood wrote: > This adds support in netdev_reassociate for SAE. An SAE auth > protocol is created rather than the connect command and from > here there is virtually no difference to a normal connection > attempt. The only differences are 'in_roam' is set as to Did you mean 'in_reassoc' here? > ignore the inevitable disconnect event and PREV_BSSID is > appended to CMD_ASSOCIATE only for this new reassociation case. > --- > src/netdev.c | 32 ++++++++++++++++++++++++-------- > 1 file changed, 24 insertions(+), 8 deletions(-) > = Please rebase on top of git HEAD. > @@ -2716,6 +2719,11 @@ static void netdev_sae_tx_associate(void *user_dat= a) > = > l_genl_msg_append_attrv(msg, NL80211_ATTR_IE, iov, n_used); > = > + /* If doing a non-FT Reassociation */ > + if (netdev->in_reassoc && !IE_AKM_IS_FT(netdev->handshake->akm_suite)) > + l_genl_msg_append_attr(msg, NL80211_ATTR_PREV_BSSID, 6, > + netdev->prev_bssid); > + Actually I added auth_proto::prev_bssid just recently with exactly this in = mind. > if (!l_genl_family_send(nl80211, msg, netdev_assoc_cb, netdev, NULL)) { > l_genl_msg_unref(msg); > netdev_connect_failed(netdev, NETDEV_RESULT_ASSOCIATION_FAILED, > @@ -3584,7 +3592,7 @@ int netdev_reassociate(struct netdev *netdev, struc= t scan_bss *target_bss, > netdev_event_func_t event_filter, > netdev_connect_cb_t cb, void *user_data) > { > - struct l_genl_msg *cmd_connect; > + struct l_genl_msg *cmd_connect =3D NULL; > struct netdev_handshake_state; > struct handshake_state *old_hs; > struct eapol_sm *sm =3D NULL, *old_sm; > @@ -3594,16 +3602,21 @@ int netdev_reassociate(struct netdev *netdev, str= uct scan_bss *target_bss, > if (netdev_handshake_state_setup_connection_type(hs) < 0) > return -ENOTSUP; > = > - /* TODO: SoftMac SAE/FILS Re-Associations are not suppored yet */ > - if (L_WARN_ON(IE_AKM_IS_SAE(hs->akm_suite) || > - IE_AKM_IS_FILS(hs->akm_suite))) > + /* TODO: SoftMac FILS Re-Associations are not suppored yet */ > + if (L_WARN_ON(IE_AKM_IS_FILS(hs->akm_suite))) > return -ENOTSUP; > = > - cmd_connect =3D netdev_build_cmd_connect(netdev, target_bss, hs, > + if (IE_AKM_IS_SAE(hs->akm_suite)) > + netdev->ap =3D sae_sm_new(hs, netdev_sae_tx_authenticate, > + netdev_sae_tx_associate, > + netdev); > + else { > + cmd_connect =3D netdev_build_cmd_connect(netdev, target_bss, hs, > orig_bss->addr, NULL, 0); This should be taking SAE offload into account, similar to netdev_connect p= ath... > = > - if (is_rsn) > - sm =3D eapol_sm_new(hs); > + if (is_rsn) > + sm =3D eapol_sm_new(hs); > + } > = > old_sm =3D netdev->sm; > old_hs =3D netdev->handshake; Regards, -Denis --===============5435079094861835384==--