| Message ID | 1562019229-15410-1-git-send-email-arend.vanspriel@broadcom.com |
|---|---|
| State | Accepted |
| Headers | show |
| Series | supplicant: use separate flag for 4-way handshake offload | expand |
On 7/2/2019 12:13 AM, Arend van Spriel wrote: > Commit d896874f8689 ("nl80211: Indicate 802.1X 4-way handshake > offload in connect") used the req_key_mgmt_offload flag to > indicate to the driver that it should offload the 802.1X handshake. > However, this field was existing and used for a different offload > API. This causes wpa_supplicant to send a connect request with the typo: should be "a connect request *without*". Should I resend or can this be fixed before applying? Regards, Arend
Am 02.07.19 um 00:13 schrieb Arend van Spriel: > Commit d896874f8689 ("nl80211: Indicate 802.1X 4-way handshake > offload in connect") used the req_key_mgmt_offload flag to > indicate to the driver that it should offload the 802.1X handshake. > However, this field was existing and used for a different offload > API. This causes wpa_supplicant to send a connect request with the > WANT_1X_HS flag and the subsequent set-pmk is rejected causing the > connection to fail. So this patch fixes that by introducing a new > flag req_handshake_offload so the offloads are no longer entangled. > > Reported-by: Stefan Wahren <wahrenst@gmx.net> > Tested-by: Stefan Wahren <wahrenst@gmx.net> > Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com> gentle ping
On Tue, Jul 02, 2019 at 12:13:49AM +0200, Arend van Spriel wrote: > Commit d896874f8689 ("nl80211: Indicate 802.1X 4-way handshake > offload in connect") used the req_key_mgmt_offload flag to > indicate to the driver that it should offload the 802.1X handshake. > However, this field was existing and used for a different offload > API. This causes wpa_supplicant to send a connect request with the > WANT_1X_HS flag and the subsequent set-pmk is rejected causing the > connection to fail. So this patch fixes that by introducing a new > flag req_handshake_offload so the offloads are no longer entangled. Thanks, applied.
On 8/1/2019 6:17 PM, Jouni Malinen wrote: > On Tue, Jul 02, 2019 at 12:13:49AM +0200, Arend van Spriel wrote: >> Commit d896874f8689 ("nl80211: Indicate 802.1X 4-way handshake >> offload in connect") used the req_key_mgmt_offload flag to >> indicate to the driver that it should offload the 802.1X handshake. >> However, this field was existing and used for a different offload >> API. This causes wpa_supplicant to send a connect request with the Don't know if a correction is possible, but reading back my commit message I noticed a typo. It should have been "connect request *without* WANT_1X_HS" here. >> WANT_1X_HS flag and the subsequent set-pmk is rejected causing the >> connection to fail. So this patch fixes that by introducing a new >> flag req_handshake_offload so the offloads are no longer entangled. Regards, Arend
On Fri, Aug 02, 2019 at 10:29:15AM +0200, Arend Van Spriel wrote: > On 8/1/2019 6:17 PM, Jouni Malinen wrote: > > On Tue, Jul 02, 2019 at 12:13:49AM +0200, Arend van Spriel wrote: > > > Commit d896874f8689 ("nl80211: Indicate 802.1X 4-way handshake > > > offload in connect") used the req_key_mgmt_offload flag to > > > indicate to the driver that it should offload the 802.1X handshake. > > > However, this field was existing and used for a different offload > > > API. This causes wpa_supplicant to send a connect request with the > > Don't know if a correction is possible, but reading back my commit message I > noticed a typo. It should have been "connect request *without* WANT_1X_HS" > here. I thought I already fixed that in the commit. Anyway, if it is not correct, the commit message cannot be changed after it has been pushed to the master branch, i.e., the only alternative to changing something would be to revert a commit and make a new one.
On August 3, 2019 3:14:51 PM Jouni Malinen <j@w1.fi> wrote: > On Fri, Aug 02, 2019 at 10:29:15AM +0200, Arend Van Spriel wrote: >> On 8/1/2019 6:17 PM, Jouni Malinen wrote: >>> On Tue, Jul 02, 2019 at 12:13:49AM +0200, Arend van Spriel wrote: >>>> Commit d896874f8689 ("nl80211: Indicate 802.1X 4-way handshake >>>> offload in connect") used the req_key_mgmt_offload flag to >>>> indicate to the driver that it should offload the 802.1X handshake. >>>> However, this field was existing and used for a different offload >>>> API. This causes wpa_supplicant to send a connect request with the >> >> Don't know if a correction is possible, but reading back my commit message I >> noticed a typo. It should have been "connect request *without* WANT_1X_HS" >> here. > > I thought I already fixed that in the commit. Anyway, if it is not > correct, the commit message cannot be changed after it has been pushed > to the master branch, i.e., the only alternative to changing something > would be to revert a commit and make a new one. I checked the repo and you did correct it [1]. Thanks, Arend [1] https://w1.fi/cgit/hostap/commit/?id=cb28bd52e1f
diff --git a/src/drivers/driver.h b/src/drivers/driver.h index 8a5cdb8..2a8459a 100644 --- a/src/drivers/driver.h +++ b/src/drivers/driver.h @@ -1074,6 +1074,14 @@ struct wpa_driver_associate_params { int req_key_mgmt_offload; /** + * req_handshake_offload - Request EAPOL handshake offload + * + * Request EAPOL handshake offload for this connection if the device + * supports it. + */ + int req_handshake_offload; + + /** * Flag for indicating whether this association includes support for * RRM (Radio Resource Measurements) */ diff --git a/src/drivers/driver_nl80211.c b/src/drivers/driver_nl80211.c index 45835a2..c44bd20 100644 --- a/src/drivers/driver_nl80211.c +++ b/src/drivers/driver_nl80211.c @@ -5632,7 +5632,7 @@ static int nl80211_connect_common(struct wpa_driver_nl80211_data *drv, return -1; } - if (params->req_key_mgmt_offload && + if (params->req_handshake_offload && (drv->capa.flags & WPA_DRIVER_FLAGS_4WAY_HANDSHAKE_8021X)) { wpa_printf(MSG_DEBUG, " * WANT_1X_4WAY_HS"); if (nla_put_flag(msg, NL80211_ATTR_WANT_1X_4WAY_HS)) diff --git a/wpa_supplicant/wpa_supplicant.c b/wpa_supplicant/wpa_supplicant.c index 78eef38..911d79d 100644 --- a/wpa_supplicant/wpa_supplicant.c +++ b/wpa_supplicant/wpa_supplicant.c @@ -3229,7 +3229,7 @@ static void wpas_start_assoc_cb(struct wpa_radio_work *work, int deinit) params.key_mgmt_suite == WPA_KEY_MGMT_IEEE8021X_SHA256 || params.key_mgmt_suite == WPA_KEY_MGMT_IEEE8021X_SUITE_B || params.key_mgmt_suite == WPA_KEY_MGMT_IEEE8021X_SUITE_B_192)) - params.req_key_mgmt_offload = 1; + params.req_handshake_offload = 1; if (wpa_s->conf->key_mgmt_offload) { if (params.key_mgmt_suite == WPA_KEY_MGMT_IEEE8021X ||
Commit d896874f8689 ("nl80211: Indicate 802.1X 4-way handshake offload in connect") used the req_key_mgmt_offload flag to indicate to the driver that it should offload the 802.1X handshake. However, this field was existing and used for a different offload API. This causes wpa_supplicant to send a connect request with the WANT_1X_HS flag and the subsequent set-pmk is rejected causing the connection to fail. So this patch fixes that by introducing a new flag req_handshake_offload so the offloads are no longer entangled. Reported-by: Stefan Wahren <wahrenst@gmx.net> Tested-by: Stefan Wahren <wahrenst@gmx.net> Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com> --- Hi Jouni, I do not recall exactly what happened in commit d896874f8689. I suspect I revised the patch using req_key_mgmt_offload without spinning another test. Regards, Arend --- src/drivers/driver.h | 8 ++++++++ src/drivers/driver_nl80211.c | 2 +- wpa_supplicant/wpa_supplicant.c | 2 +- 3 files changed, 10 insertions(+), 2 deletions(-)