BUG: CE 4.2 WZC does not reconnect to AP after going out of range

This was reported by a customer of ours and we have also reproduced the issue.

Our Device:
Windows CE 4.2 - QFE Updates through 6/2004. (I have tried with and without
the WZC update from the 3 Qtr 2004 QFE package)
ARMV4I Processor Type

Steps to recreate:
1. Using Wireless Zero Config (WZC) Connect to AP
2. Walk out of range so that it disconnects and stay out of range for 15+
seconds
3. Come back in range

At this point the WZC service for some reason does not reconnect to the AP.
It will have no issue reconnecting if you tell it to, however, you should not
have to manually tell it to reconnect.

I found a KB article concerning XP with a similar if not the same issue:
http://support.microsoft.com/default.aspx?scid=kb;en-us;828940

I also replied to another thread concerning this but the originator put it
on the Embedded VC group and this seemed like a better spot. Note: that
means that someone else has also seen this issue on another device.

Also note: I tried this on an IPAQ running PPC2003 yesterday and it also
exhibited a similar problem however it could not even reconnect when told to.

I believe I could get around this problem on our device by writing an
application that would ping the WZC service every 10 seconds or so and try to
reconnect if it was not actively connected to anything. However, this is not
very effecient, would reduce battery life when out of range (I would guess),
and the preferable solution is of course to have WZC fixed.

Also note, that our customer had a log running on their AP which showed that
the unit was set to little boxes (unrecognized characters), which I believe
is the NO SSID setting. Either that or the SSID field is actually getting
corrupted somehow.
--
Matthew Townsend
Itronix Corp.

Paul
Tue Jan 11 13:22:24 CST 2005

You should report this directly to product support services (the phone
number). Although there are a large number of MS insiders who frequent this
group, getting something like this fixed always ends up being triggered by a
product support report. If it's a bug, it doesn't cost you anything and, if
you bought Platform Builder, you should have a couple of free support
requests with that.

Paul T.

"Matthew Townsend" <MatthewTownsend@discussions.microsoft.com> wrote in
message news:D5AFABAD-83FC-46B1-A83B-707B23CE2CB8@microsoft.com...
> This was reported by a customer of ours and we have also reproduced the
> issue.
>
> Our Device:
> Windows CE 4.2 - QFE Updates through 6/2004. (I have tried with and
> without
> the WZC update from the 3 Qtr 2004 QFE package)
> ARMV4I Processor Type
>
> Steps to recreate:
> 1. Using Wireless Zero Config (WZC) Connect to AP
> 2. Walk out of range so that it disconnects and stay out of range for 15+
> seconds
> 3. Come back in range
>
> At this point the WZC service for some reason does not reconnect to the
> AP.
> It will have no issue reconnecting if you tell it to, however, you should
> not
> have to manually tell it to reconnect.
>
> I found a KB article concerning XP with a similar if not the same issue:
> http://support.microsoft.com/default.aspx?scid=kb;en-us;828940
>
> I also replied to another thread concerning this but the originator put it
> on the Embedded VC group and this seemed like a better spot. Note: that
> means that someone else has also seen this issue on another device.
>
> Also note: I tried this on an IPAQ running PPC2003 yesterday and it also
> exhibited a similar problem however it could not even reconnect when told
> to.
>
> I believe I could get around this problem on our device by writing an
> application that would ping the WZC service every 10 seconds or so and try
> to
> reconnect if it was not actively connected to anything. However, this is
> not
> very effecient, would reduce battery life when out of range (I would
> guess),
> and the preferable solution is of course to have WZC fixed.
>
> Also note, that our customer had a log running on their AP which showed
> that
> the unit was set to little boxes (unrecognized characters), which I
> believe
> is the NO SSID setting. Either that or the SSID field is actually getting
> corrupted somehow.
> --
> Matthew Townsend
> Itronix Corp.

Michel
Tue Jan 11 17:06:53 CST 2005

I've seen the same behaviour. Just to confirm you're not the only one...

Regards,


Michel Verhagen
Windows CE.NET Consultant, eMVP
PTS Software
The Netherlands



"Matthew Townsend" <MatthewTownsend@discussions.microsoft.com> wrote in
message news:D5AFABAD-83FC-46B1-A83B-707B23CE2CB8@microsoft.com...
> This was reported by a customer of ours and we have also reproduced the
> issue.
>
> Our Device:
> Windows CE 4.2 - QFE Updates through 6/2004. (I have tried with and
> without
> the WZC update from the 3 Qtr 2004 QFE package)
> ARMV4I Processor Type
>
> Steps to recreate:
> 1. Using Wireless Zero Config (WZC) Connect to AP
> 2. Walk out of range so that it disconnects and stay out of range for 15+
> seconds
> 3. Come back in range
>
> At this point the WZC service for some reason does not reconnect to the
> AP.
> It will have no issue reconnecting if you tell it to, however, you should
> not
> have to manually tell it to reconnect.
>
> I found a KB article concerning XP with a similar if not the same issue:
> http://support.microsoft.com/default.aspx?scid=kb;en-us;828940
>
> I also replied to another thread concerning this but the originator put it
> on the Embedded VC group and this seemed like a better spot. Note: that
> means that someone else has also seen this issue on another device.
>
> Also note: I tried this on an IPAQ running PPC2003 yesterday and it also
> exhibited a similar problem however it could not even reconnect when told
> to.
>
> I believe I could get around this problem on our device by writing an
> application that would ping the WZC service every 10 seconds or so and try
> to
> reconnect if it was not actively connected to anything. However, this is
> not
> very effecient, would reduce battery life when out of range (I would
> guess),
> and the preferable solution is of course to have WZC fixed.
>
> Also note, that our customer had a log running on their AP which showed
> that
> the unit was set to little boxes (unrecognized characters), which I
> believe
> is the NO SSID setting. Either that or the SSID field is actually getting
> corrupted somehow.
> --
> Matthew Townsend
> Itronix Corp.

ctacke/>
Wed Jan 12 20:51:44 CST 2005

Make that three.

-Chris


"Michel Verhagen [eMVP]" <michel.verhagen@nospam.pts.nl> wrote in message
news:OJ3u8ID%23EHA.220@TK2MSFTNGP09.phx.gbl...
> I've seen the same behaviour. Just to confirm you're not the only one...
>
> Regards,
>
>
> Michel Verhagen
> Windows CE.NET Consultant, eMVP
> PTS Software
> The Netherlands
>
>
>
> "Matthew Townsend" <MatthewTownsend@discussions.microsoft.com> wrote in
> message news:D5AFABAD-83FC-46B1-A83B-707B23CE2CB8@microsoft.com...
>> This was reported by a customer of ours and we have also reproduced the
>> issue.
>>
>> Our Device:
>> Windows CE 4.2 - QFE Updates through 6/2004. (I have tried with and
>> without
>> the WZC update from the 3 Qtr 2004 QFE package)
>> ARMV4I Processor Type
>>
>> Steps to recreate:
>> 1. Using Wireless Zero Config (WZC) Connect to AP
>> 2. Walk out of range so that it disconnects and stay out of range for 15+
>> seconds
>> 3. Come back in range
>>
>> At this point the WZC service for some reason does not reconnect to the
>> AP.
>> It will have no issue reconnecting if you tell it to, however, you should
>> not
>> have to manually tell it to reconnect.
>>
>> I found a KB article concerning XP with a similar if not the same issue:
>> http://support.microsoft.com/default.aspx?scid=kb;en-us;828940
>>
>> I also replied to another thread concerning this but the originator put
>> it
>> on the Embedded VC group and this seemed like a better spot. Note: that
>> means that someone else has also seen this issue on another device.
>>
>> Also note: I tried this on an IPAQ running PPC2003 yesterday and it also
>> exhibited a similar problem however it could not even reconnect when told
>> to.
>>
>> I believe I could get around this problem on our device by writing an
>> application that would ping the WZC service every 10 seconds or so and
>> try to
>> reconnect if it was not actively connected to anything. However, this is
>> not
>> very effecient, would reduce battery life when out of range (I would
>> guess),
>> and the preferable solution is of course to have WZC fixed.
>>
>> Also note, that our customer had a log running on their AP which showed
>> that
>> the unit was set to little boxes (unrecognized characters), which I
>> believe
>> is the NO SSID setting. Either that or the SSID field is actually
>> getting
>> corrupted somehow.
>> --
>> Matthew Townsend
>> Itronix Corp.
>
>

stjong
Mon Jan 17 00:56:49 CST 2005

Matthew,
Thank you for reporting this issue, I will check out the QFE you mentioned.
For your CE 4.2 device (not iPaq in which the periodic rescan may be turn
off), what wireless adapter did you use?
Was the driver shipped in CE?
Did the driver indicate a media disconnect when it got out of range?
Did it actually try to associate and failed, like the one reported in the
QFE for XP?
Is the problem reproducible when the SSID is not hidden?
Thanks,
-- Soemin ---
Windows CE CoreOS/Networking
This posting is provided "AS IS" with no warranties, and confers no rights.

MatthewTownsend
Mon Jan 24 19:11:02 CST 2005

We have seen the issue with 2 devices.

Device 1 (Cisco):

Our customer that reported it is using a Cisco Aironet 350 802.11b card.
The drivers are available at www.cisco.com. They are not using the built in
drivers.

I built an OS with the built in version (the one shipped with CE) of the
Cisco WLAN card driver and it also has the issue.
Here is the WZC log for that test (connect, walk out of range, wait 15+
seconds, walk back in range, wait for 1 minute, then manually connect to AP
to prove that it is indeed in range)
[10:46:59 - WLAGS46B1] ZeroConfig started
[09:47:02 - WLAGS46B1] Preferred List Exhausted
[09:47:32 - CISCO1] ZeroConfig started
[09:47:35 - CISCO1] Preferred List Exhausted
[09:47:58 - CISCO1] Associating with 1engAP...
[09:47:58 - CISCO1] Associated with 1engAP - AP [00-0e-d7-a3-e6-30]
[09:49:25 - CISCO1] Associating with 1engAP...
[09:49:28 - CISCO1] Failed to associate with 1engAP
[09:49:28 - CISCO1] Preferred List Exhausted
[09:50:58 - CISCO1] Associating with 1engAP...
[09:50:59 - CISCO1] Associated with 1engAP - AP [00-0e-d7-a3-e6-30]

Device 2(Agere):
I have been testing with our built in WLAN module. It is an Agere 802.11b
chip. Unfortunately I don't even have debug versions of the drivers. As far
as I know they were built explicitly for CE.


Relevant to both Devices:

In all cases when I got out of range, WZC showed that I had disconnected
(the icon has an X through it).

I tried installing the WZC update from the 3rd Qtr. 2004 QFE package and it
still had the issue.

Yes, the problem occurs when the SSID is not hidden.


""Soemin Tjong [MSFT]"" wrote:

> Matthew,
> Thank you for reporting this issue, I will check out the QFE you mentioned.
> For your CE 4.2 device (not iPaq in which the periodic rescan may be turn
> off), what wireless adapter did you use?
> Was the driver shipped in CE?
> Did the driver indicate a media disconnect when it got out of range?
> Did it actually try to associate and failed, like the one reported in the
> QFE for XP?
> Is the problem reproducible when the SSID is not hidden?
> Thanks,
> -- Soemin ---
> Windows CE CoreOS/Networking
> This posting is provided "AS IS" with no warranties, and confers no rights.
>
>

MatthewTownsend
Wed Feb 02 10:11:06 CST 2005

Have you been able to find anything about this? Should I open up a support
request with the platform builder group? Thanks.

""Soemin Tjong [MSFT]"" wrote:

> Matthew,
> Thank you for reporting this issue, I will check out the QFE you mentioned.
> For your CE 4.2 device (not iPaq in which the periodic rescan may be turn
> off), what wireless adapter did you use?
> Was the driver shipped in CE?
> Did the driver indicate a media disconnect when it got out of range?
> Did it actually try to associate and failed, like the one reported in the
> QFE for XP?
> Is the problem reproducible when the SSID is not hidden?
> Thanks,
> -- Soemin ---
> Windows CE CoreOS/Networking
> This posting is provided "AS IS" with no warranties, and confers no rights.
>
>

MatthewTownsend
Mon Mar 07 14:43:04 CST 2005

After opening a support request with Microsoft I discovered that...

If you do not have the "Notify when new networks available" checkbox checked
then WZC will not reconnect to the existing AP connection. Apparently this
is by design but in my opinion the check box text should either explain this
nuance or there should be 2 checkboxes (preferable) 1 for the reconnection
and 1 to turn off those annoying "new network available" messages that our
customers complain about.

As it stands I will just implement my own fix.

"Matthew Townsend" wrote:

> Have you been able to find anything about this? Should I open up a support
> request with the platform builder group? Thanks.
>
> ""Soemin Tjong [MSFT]"" wrote:
>
> > Matthew,
> > Thank you for reporting this issue, I will check out the QFE you mentioned.
> > For your CE 4.2 device (not iPaq in which the periodic rescan may be turn
> > off), what wireless adapter did you use?
> > Was the driver shipped in CE?
> > Did the driver indicate a media disconnect when it got out of range?
> > Did it actually try to associate and failed, like the one reported in the
> > QFE for XP?
> > Is the problem reproducible when the SSID is not hidden?
> > Thanks,
> > -- Soemin ---
> > Windows CE CoreOS/Networking
> > This posting is provided "AS IS" with no warranties, and confers no rights.
> >
> >

stjong
Tue Mar 29 20:39:27 CST 2005

Mark, sorry for not getting back earlier, it has been a busy few months
working on our release.
My understanding is that the check box "Notify me when new wireless
networks are available" is for turning off those annoying messages and not
for preventing it from reconnection.

Going back on this thread, your log seems to confirm that zeroconfig was
trying to reconnect to the AP as you came back in range:

[10:46:59 - WLAGS46B1] ZeroConfig started
[09:47:02 - WLAGS46B1] Preferred List Exhausted
[09:47:32 - CISCO1] ZeroConfig started
[09:47:35 - CISCO1] Preferred List Exhausted
[09:47:58 - CISCO1] Associating with 1engAP...
[09:47:58 - CISCO1] Associated with 1engAP - AP [00-0e-d7-a3-e6-30] <--
This is when you were first connected.
[09:49:25 - CISCO1] Associating with 1engAP...
[09:49:28 - CISCO1] Failed to associate with 1engAP <-- This is when you
walked back in range, but somehow the adapter failed to associate with the
AP.
[09:49:28 - CISCO1] Preferred List Exhausted
[09:50:58 - CISCO1] Associating with 1engAP...
[09:50:59 - CISCO1] Associated with 1engAP - AP [00-0e-d7-a3-e6-30] <--
This is when you did manual connection to the same AP.

I did check with the XP QFE team but couldn't get a definitive answer for
this particular change since it is combined with a number of other fixes
for the XP SP2 release.

I take it with the check box enabled, you couldn't repro the issue with the
check box checked. Is that right? Thanks.

-- Soemin ---
Windows CE CoreOS/Networking
This posting is provided "AS IS" with no warranties, and confers no rights.

MatthewTownsend
Wed Mar 30 10:07:03 CST 2005

Actually, that failure "[09:49:28 - CISCO1] Failed to associate with 1engAP"
was when I was still out of range and it tried to reconnect after loosing the
connection. Upon coming back in range it did not do anything until I
manually told it to do so.

Yes, with the checkbox checked it reconnected just fine.

The Microsoft support people said that it had to be checked in order for it
to reconnect. Is that not true?

Thanks.

""Soemin Tjong [MSFT]"" wrote:

> Mark, sorry for not getting back earlier, it has been a busy few months
> working on our release.
> My understanding is that the check box "Notify me when new wireless
> networks are available" is for turning off those annoying messages and not
> for preventing it from reconnection.
>
> Going back on this thread, your log seems to confirm that zeroconfig was
> trying to reconnect to the AP as you came back in range:
>
> [10:46:59 - WLAGS46B1] ZeroConfig started
> [09:47:02 - WLAGS46B1] Preferred List Exhausted
> [09:47:32 - CISCO1] ZeroConfig started
> [09:47:35 - CISCO1] Preferred List Exhausted
> [09:47:58 - CISCO1] Associating with 1engAP...
> [09:47:58 - CISCO1] Associated with 1engAP - AP [00-0e-d7-a3-e6-30] <--
> This is when you were first connected.
> [09:49:25 - CISCO1] Associating with 1engAP...
> [09:49:28 - CISCO1] Failed to associate with 1engAP <-- This is when you
> walked back in range, but somehow the adapter failed to associate with the
> AP.
> [09:49:28 - CISCO1] Preferred List Exhausted
> [09:50:58 - CISCO1] Associating with 1engAP...
> [09:50:59 - CISCO1] Associated with 1engAP - AP [00-0e-d7-a3-e6-30] <--
> This is when you did manual connection to the same AP.
>
> I did check with the XP QFE team but couldn't get a definitive answer for
> this particular change since it is combined with a number of other fixes
> for the XP SP2 release.
>
> I take it with the check box enabled, you couldn't repro the issue with the
> check box checked. Is that right? Thanks.
>
> -- Soemin ---
> Windows CE CoreOS/Networking
> This posting is provided "AS IS" with no warranties, and confers no rights.
>
>
>

MatthewTownsend
Wed Mar 30 12:37:04 CST 2005

I just fixed this in \WINCE420\PUBLIC\COMMON\OAK\DRIVERS\NETUI

There is a timeout value that is set to 1 minute if the checkbox is checked
but set to infinity if not. I simplay changed it so that it is always set to
1 minute. Now it reconnects when coming back into range (well, within a
minute of).

""Soemin Tjong [MSFT]"" wrote:

> Mark, sorry for not getting back earlier, it has been a busy few months
> working on our release.
> My understanding is that the check box "Notify me when new wireless
> networks are available" is for turning off those annoying messages and not
> for preventing it from reconnection.
>
> Going back on this thread, your log seems to confirm that zeroconfig was
> trying to reconnect to the AP as you came back in range:
>
> [10:46:59 - WLAGS46B1] ZeroConfig started
> [09:47:02 - WLAGS46B1] Preferred List Exhausted
> [09:47:32 - CISCO1] ZeroConfig started
> [09:47:35 - CISCO1] Preferred List Exhausted
> [09:47:58 - CISCO1] Associating with 1engAP...
> [09:47:58 - CISCO1] Associated with 1engAP - AP [00-0e-d7-a3-e6-30] <--
> This is when you were first connected.
> [09:49:25 - CISCO1] Associating with 1engAP...
> [09:49:28 - CISCO1] Failed to associate with 1engAP <-- This is when you
> walked back in range, but somehow the adapter failed to associate with the
> AP.
> [09:49:28 - CISCO1] Preferred List Exhausted
> [09:50:58 - CISCO1] Associating with 1engAP...
> [09:50:59 - CISCO1] Associated with 1engAP - AP [00-0e-d7-a3-e6-30] <--
> This is when you did manual connection to the same AP.
>
> I did check with the XP QFE team but couldn't get a definitive answer for
> this particular change since it is combined with a number of other fixes
> for the XP SP2 release.
>
> I take it with the check box enabled, you couldn't repro the issue with the
> check box checked. Is that right? Thanks.
>
> -- Soemin ---
> Windows CE CoreOS/Networking
> This posting is provided "AS IS" with no warranties, and confers no rights.
>
>
>

stjong
Mon Apr 04 01:18:11 CDT 2005

Okay, thanks for confirming.
Zeroconfig internal periodic scan should not have anything to do with the
notification check box.
Sound more like a bug to me.
We'll check it out.
Thanks,
-- Soemin ---
Windows CE CoreOS/Networking
This posting is provided "AS IS" with no warranties, and confers no rights.