2:54 p.m.
When p2p is connecting and other device acts as Group owner then
G_SUPPLICANT_STATE_COMPLETED is received after P2P Group formation is
success. handle_wps_completion() function disconnects the network if
wps_ssid doesn't match with ssid, due to this reason P2P connection is
never successful.
---
plugins/wifi.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/plugins/wifi.c b/plugins/wifi.c
index 74f216d..a874112 100644
--- a/plugins/wifi.c
+++ b/plugins/wifi.c
@@ -2415,6 +2415,9 @@ static void interface_state(GSupplicantInterface *interface)
/* though it should be already stopped: */
stop_autoscan(device);
+ if (wifi->p2p_connecting)
+ break;
+
if (!handle_wps_completion(interface, network, device, wifi))
break;
--
1.9.1
9:24 a.m.
Hi Saurav,
On Mon, Oct 16, 2017 at 3:54 PM, Saurav Babu <saurav.babu(a)samsung.com> wrote:
When p2p is connecting and other device acts as Group owner then
G_SUPPLICANT_STATE_COMPLETED is received after P2P Group formation is
success. handle_wps_completion() function disconnects the network if
wps_ssid doesn't match with ssid, due to this reason P2P connection is
never successful.
Great catch!
plugins/wifi.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/plugins/wifi.c b/plugins/wifi.c
index 74f216d..a874112 100644
--- a/plugins/wifi.c
+++ b/plugins/wifi.c
@@ -2415,6 +2415,9 @@ static void interface_state(GSupplicantInterface *interface)
/* though it should be already stopped: */
stop_autoscan(device);
+ if (wifi->p2p_connecting)
+ break;
+
You are testing this with a wpa_supplicant configuration that does not
create a separate P2P group interface, right? Maybe because the
chipset does not support it or that you manually configured
wpa_supplicant to do so (p2p_no_group_iface=1)?
In a system where such group interface is created, IIRC p2p_connecting
will be set only on the main interface and not on the group interface
which is the one that will go to the state
G_SUPPLICANT_STATE_COMPLETED. If that's true, this patch will not
cover such case. Do you agree?
Do you have the possibility to test this with both configuration?
Thanks,
Jose Blanquicet
6:56 p.m.
Hi Saurav,
On Thu, Oct 19, 2017 at 10:24 AM, Jose Blanquicet <blanquicet(a)gmail.com> wrote:
On Mon, Oct 16, 2017 at 3:54 PM, Saurav Babu
<saurav.babu(a)samsung.com> wrote:
> When p2p is connecting and other device acts as Group owner then
> G_SUPPLICANT_STATE_COMPLETED is received after P2P Group formation is
> success. handle_wps_completion() function disconnects the network if
> wps_ssid doesn't match with ssid, due to this reason P2P connection is
> never successful.
Digging into this issue I reached the conclusion that when P2P is
connecting, we actually do not need to handle interface state changes
G_SUPPLICANT_STATE_*. Instead, the connection success is completely
based on the peer state CONNMAN_PEER_STATE_*.
Therefore, we should ignore all the interface state changes during the
P2P connection, otherwise it could result in an inconsistency. In
fact, I tested your patch and it solves the connection problem but at
the end of the connection the service belonging to the just created
P2P group, goes in "association" state and we do not want this.
Here my proposal:
From f4cfb7707af3c97a3beca1007c0145ba15bf772c Mon Sep 17 00:00:00 2001
From: Jose Blanquicet <jose(a)blanquicet.com>
Date: Sun, 22 Oct 2017 19:39:31 +0200
Subject: [PATCH] wifi: Avoid state machine moves forward with P2P networks
---
plugins/wifi.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/plugins/wifi.c b/plugins/wifi.c
index 74f216d..c2226a3 100644
--- a/plugins/wifi.c
+++ b/plugins/wifi.c
@@ -2822,6 +2822,10 @@ static void
network_associated(GSupplicantNetwork *network)
if (!wifi)
return;
+ /* P2P networks must not be treated as WiFi networks */
+ if (wifi->p2p_connecting || wifi->p2p_device)
+ return;
+
identifier = g_supplicant_network_get_identifier(network);
connman_network = connman_device_get_network(wifi->device, identifier);
--
1.9.1
wifi->p2p_device will handle system's configuration where the P2P
group interface is created and wifi->p2p_connecting where it is not.
Could you please test this patch and give us your feedback?
Thanks,
Jose Blanquicet
6:31 a.m.
Hello Jose,
Sorry for the late reply.
Digging into this issue I reached the conclusion that when P2P is
connecting, we actually do not need to handle interface state changes
G_SUPPLICANT_STATE_*. Instead, the connection success is completely
based on the peer state CONNMAN_PEER_STATE_*.
Therefore, we should ignore all the interface state changes during
the
P2P connection, otherwise it could result in an inconsistency. In
fact, I tested your patch and it solves the connection problem but at
the end of the connection the service belonging to the just created
P2P group, goes in "association" state and we do not want this.
Here my proposal:
From f4cfb7707af3c97a3beca1007c0145ba15bf772c Mon Sep 17 00:00:00
2001
From: Jose Blanquicet
Date: Sun, 22 Oct 2017 19:39:31 +0200
Subject: [PATCH] wifi: Avoid state machine moves forward with P2P networks
---
plugins/wifi.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/plugins/wifi.c b/plugins/wifi.c
index 74f216d..c2226a3 100644
--- a/plugins/wifi.c
+++ b/plugins/wifi.c
@@ -2822,6 +2822,10 @@ static void
network_associated(GSupplicantNetwork *network)
if (!wifi)
return;
+ /* P2P networks must not be treated as WiFi networks */
+ if (wifi->p2p_connecting || wifi->p2p_device)
+ return;
+
identifier = g_supplicant_network_get_identifier(network);
connman_network = connman_device_get_network(wifi->device,
identifier);
--
1.9.1
wifi->p2p_device will handle system's configuration where the
P2P
group interface is created and wifi->p2p_connecting where it is not.
Could you please test this patch and give us your feedback?
This patch worked for me.
Thanks,
Saurav
8:47 p.m.
Hi Jose,
On 12/01/2017 07:31 AM, Saurav Babu wrote:
Hello Jose,
Sorry for the late reply.
> Digging into this issue I reached the conclusion that when P2P is
> connecting, we actually do not need to handle interface state changes
> G_SUPPLICANT_STATE_*. Instead, the connection success is completely
> based on the peer state CONNMAN_PEER_STATE_*.
> Therefore, we should ignore all the interface state changes during the
> P2P connection, otherwise it could result in an inconsistency. In
> fact, I tested your patch and it solves the connection problem but at
> the end of the connection the service belonging to the just created
> P2P group, goes in "association" state and we do not want this.
Could you add your excellent analysis to commit message and resend the
patch with it?
Thanks,
Daniel
1503
days inactive
1558
days old
5 comments
3 participants
participants (3)
-
Daniel Wagner -
Jose Blanquicet -
Saurav Babu