linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Cheng Renquan]

Hi Intel Linux Wireless group,

Since 3.2-rcX, it seems no more iwlagn driver? This iwlwifi driver
seems not stable
when I hold laptop moving inside a large office, between multiple APs,
everytime I moved to a new place, the wifi link becomes inactive, and
I need to wait
5 minutes or longer to wait iwlwifi stable, networkmanager get a new
ip address; quite annoying,


[root@archtop ~]# lspci -D -nn -k -d 8086:4238
0000:03:00.0 Network controller [0280]: Intel Corporation Centrino
Ultimate-N 6300 [8086:4238] (rev 35)
	Subsystem: Intel Corporation Centrino Ultimate-N 6300 3x3 AGN [8086:1111]
	Kernel driver in use: iwlwifi
	Kernel modules: iwlwifi

in the large office, it's using WPA2 enterprise encryption with
MSCHAPv2 domain user/pass authentication,


Last iwlagn driver in my Ubuntu-10.10 default kernel (linux-2.6.35.X)
on the same computer
same networkmanager/wpa_supplicant software same office working wonderful,
moving smoothly between multiple APs,

At home, even I set wireless router to no any
encryption/authentication, this new kernel
would becomes wifi link down every couple of hours, while another
laptop and same Ubuntu-10
on this laptop could work smoothly,




Thanks,

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Guy, Wey-Yi]

Hi Renquan,

Thanks you for reporting the issue.It will be greatly appreciated  if
you can enter a bug report in http://bugzilla.intellinuxwireless.org/.
Then we can assign engineer to look into the issue , also keep track of
the progress.

Thanks
Wey

On Tue, 2011-12-20 at 15:20 -0800, Cheng Renquan wHi Intel Linux
Wireless group,
> 
> Since 3.2-rcX, it seems no more iwlagn driver? This iwlwifi driver
> seems not stable
> when I hold laptop moving inside a large office, between multiple APs,
> everytime I moved to a new place, the wifi link becomes inactive, and
> I need to wait
> 5 minutes or longer to wait iwlwifi stable, networkmanager get a new
> ip address; quite annoying,
> 
> 
> [root@archtop ~]# lspci -D -nn -k -d 8086:4238
> 0000:03:00.0 Network controller [0280]: Intel Corporation Centrino
> Ultimate-N 6300 [8086:4238] (rev 35)
> 	Subsystem: Intel Corporation Centrino Ultimate-N 6300 3x3 AGN [8086:1111]
> 	Kernel driver in use: iwlwifi
> 	Kernel modules: iwlwifi
> 
> in the large office, it's using WPA2 enterprise encryption with
> MSCHAPv2 domain user/pass authentication,
> 
> 
> Last iwlagn driver in my Ubuntu-10.10 default kernel (linux-2.6.35.X)
> on the same computer
> same networkmanager/wpa_supplicant software same office working wonderful,
> moving smoothly between multiple APs,
> 
> At home, even I set wireless router to no any
> encryption/authentication, this new kernel
> would becomes wifi link down every couple of hours, while another
> laptop and same Ubuntu-10
> on this laptop could work smoothly,
> 
> 
> 
> 
> Thanks,

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Cheng Renquan]

On Tue, Dec 20, 2011 at 3:46 PM, Guy, Wey-Yi <wey-yi.w.guy@intel.com> wrote:
> Hi Renquan,
>
> Thanks you for reporting the issue.It will be greatly appreciated  if
> you can enter a bug report in http://bugzilla.intellinuxwireless.org/.
> Then we can assign engineer to look into the issue , also keep track of
> the progress.
>
> Thanks
> Wey

have reported http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2338

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Sedat Dilek]

On Wed, Dec 21, 2011 at 3:37 AM, Cheng Renquan <crquan@gmail.com> wrote:
> On Tue, Dec 20, 2011 at 3:46 PM, Guy, Wey-Yi <wey-yi.w.guy@intel.com> wrote:
>> Hi Renquan,
>>
>> Thanks you for reporting the issue.It will be greatly appreciated  if
>> you can enter a bug report in http://bugzilla.intellinuxwireless.org/.
>> Then we can assign engineer to look into the issue , also keep track of
>> the progress.
>>
>> Thanks
>> Wey
>
> have reported http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2338
>

Can you try with linux-3.2-rc6 and the attached patch in [1]?

- Sedat -

[1] http://lkml.org/lkml/2011/12/17/14

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Cheng Renquan]

On Tue, Dec 20, 2011 at 10:22 PM, Sedat Dilek
<sedat.dilek@googlemail.com> wrote:
>> have reported http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2338
>
> Can you try with linux-3.2-rc6 and the attached patch in [1]?
>
> - Sedat -
>
> [1] http://lkml.org/lkml/2011/12/17/14

Is it already in iwlwifi.git? I'm gonna test it when I'm in office,
http://git.kernel.org/?p=linux/kernel/git/iwlwifi/iwlwifi.git

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Sedat Dilek]

On Wed, Dec 21, 2011 at 9:03 AM, Cheng Renquan <crquan@gmail.com> wrote:
> On Tue, Dec 20, 2011 at 10:22 PM, Sedat Dilek
> <sedat.dilek@googlemail.com> wrote:
>>> have reported http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2338
>>
>> Can you try with linux-3.2-rc6 and the attached patch in [1]?
>>
>> - Sedat -
>>
>> [1] http://lkml.org/lkml/2011/12/17/14
>
> Is it already in iwlwifi.git? I'm gonna test it when I'm in office,
> http://git.kernel.org/?p=linux/kernel/git/iwlwifi/iwlwifi.git

/me confused...
Your subject is/was " ...linux 3.2-rcX iwlwifi driver's problem...",
now you are talking about iwlwifi.git.
Not sure what's the base of iwlwifi.git, but I guess it is wireless-testing?
BTW; what do you meant by 3.2-rcX, what is X? -rc1, -rc2...-rc6?
Which of those -rcX are affected?

- Sedat -

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Cheng Renquan]

On Wed, Dec 21, 2011 at 12:53 AM, Sedat Dilek
<sedat.dilek@googlemail.com> wrote:
> /me confused...
> Your subject is/was " ...linux 3.2-rcX iwlwifi driver's problem...",
> now you are talking about iwlwifi.git.
> Not sure what's the base of iwlwifi.git, but I guess it is wireless-testing?
> BTW; what do you meant by 3.2-rcX, what is X? -rc1, -rc2...-rc6?
> Which of those -rcX are affected?

More details are tracked into the bug, the kernel version one between
linus offical 3.2.0-rc5 and 3.2.0-rc6,
http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2338

This kernel has been running on my Lenovo laptop more than 1 week with
no reboot (only suspend & resume),
and wifi roaming problems happened in office everyday,

[root@archtop ~]# modinfo iwlwifi
filename:      /lib/modules/3.2.0-rc5+/kernel/drivers/net/wireless/iwlwifi/iwlwifi.ko

>>> [1] http://lkml.org/lkml/2011/12/17/14
>>
>> Is it already in iwlwifi.git? I'm gonna test it when I'm in office,
>> http://git.kernel.org/?p=linux/kernel/git/iwlwifi/iwlwifi.git

Now I'm asking if the patch mentioned in your above link is already in
iwl git tree,
since it's Intel's offical tree for iwlwifi, I'm gonna clone a latest
tree and test it;

But you're using a googlemail, now I'm not sure if you're working in
that Intel iwl team...

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Sedat Dilek]

On Wed, Dec 21, 2011 at 10:06 AM, Cheng Renquan <crquan@gmail.com> wrote:
> On Wed, Dec 21, 2011 at 12:53 AM, Sedat Dilek
> <sedat.dilek@googlemail.com> wrote:
>> /me confused...
>> Your subject is/was " ...linux 3.2-rcX iwlwifi driver's problem...",
>> now you are talking about iwlwifi.git.
>> Not sure what's the base of iwlwifi.git, but I guess it is wireless-testing?
>> BTW; what do you meant by 3.2-rcX, what is X? -rc1, -rc2...-rc6?
>> Which of those -rcX are affected?
>
> More details are tracked into the bug, the kernel version one between
> linus offical 3.2.0-rc5 and 3.2.0-rc6,
> http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2338
>
> This kernel has been running on my Lenovo laptop more than 1 week with
> no reboot (only suspend & resume),
> and wifi roaming problems happened in office everyday,
>
> [root@archtop ~]# modinfo iwlwifi
> filename:      /lib/modules/3.2.0-rc5+/kernel/drivers/net/wireless/iwlwifi/iwlwifi.ko
>
>>>> [1] http://lkml.org/lkml/2011/12/17/14
>>>
>>> Is it already in iwlwifi.git? I'm gonna test it when I'm in office,
>>> http://git.kernel.org/?p=linux/kernel/git/iwlwifi/iwlwifi.git
>
> Now I'm asking if the patch mentioned in your above link is already in
> iwl git tree,
> since it's Intel's offical tree for iwlwifi, I'm gonna clone a latest
> tree and test it;
>

You read the URL I gave you?

"...unfortunately, this -rc6 is missing some important iwlwifi fixes.
These were included in wireless.git#master (master-2011-12-14).

I have attached a single "wireless" patch to make affected people happy...."

Is John Linville (Linux Maintainer of the wireless sub-tree)
wireless.git not official enough for you?
Just FYI, this is the tree which is regularly pulled into net.git and
from there to Linus-tree (mainline, currently 3.2-rcX).
Though...

- Sedat -

> But you're using a googlemail, now I'm not sure if you're working in
> that Intel iwl team...

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Cheng Renquan]

On Wed, Dec 21, 2011 at 1:20 AM, Sedat Dilek <sedat.dilek@googlemail.com> wrote:
> You read the URL I gave you?
>
> "...unfortunately, this -rc6 is missing some important iwlwifi fixes.
> These were included in wireless.git#master (master-2011-12-14).

Sorry, I didn't notice that sentence, just saw a single patch but I
usually prefer checking out from a git tree,

read these from MAINTAINERS, just want to confirm if iwlwifi.git has more
fresh content, this should be the most upstream, right?

(only here has new fix and fixes are verified correct, then John Linville could
pull from here the iwlwifi.git, right? As I understand, the MAINTAINERS model
should be the way of iwlwifi.git => wireless.git => net.git => linus)

INTEL WIRELESS WIFI LINK (iwlwifi)
M:      Wey-Yi Guy <wey-yi.w.guy@intel.com>
M:      Intel Linux Wireless <ilw@linux.intel.com>
L:      linux-wireless@vger.kernel.org
W:      http://intellinuxwireless.org
T:      git git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi/iwlwifi.git
S:      Supported
F:      drivers/net/wireless/iwlwifi/


By comparing the two, I got that iwlwifi.git really has more, now
prefetched for tomorrow
when in office sometimes Internet is not available, and in case
wireless.git with your patch
still does not work, I could have a try of iwlwifi.git;

http://git.kernel.org/?p=linux/kernel/git/iwlwifi/iwlwifi.git;a=history;f=drivers/net/wireless/iwlwifi
http://git.kernel.org/?p=linux/kernel/git/linville/wireless.git;a=history;f=drivers/net/wireless/iwlwifi

>
> I have attached a single "wireless" patch to make affected people happy...."
>
> Is John Linville (Linux Maintainer of the wireless sub-tree)
> wireless.git not official enough for you?
> Just FYI, this is the tree which is regularly pulled into net.git and
> from there to Linus-tree (mainline, currently 3.2-rcX).
> Though...

Thank you, acknowledged,

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Sedat Dilek]

On Wed, Dec 21, 2011 at 11:36 AM, Cheng Renquan <crquan@gmail.com> wrote:
> On Wed, Dec 21, 2011 at 1:20 AM, Sedat Dilek <sedat.dilek@googlemail.com> wrote:
[...]
> read these from MAINTAINERS, just want to confirm if iwlwifi.git has more
> fresh content, this should be the most upstream, right?
>

Wey-Yi or Johannes can gave you precise answers on iwwifi development.
Of course the iwlwifi GIT (master) tree has latest stuff.
As you might have seen there are different GIT branches for stuff to
go into wireless-testing (experimental sofware), wireless-next
(linux-next.git) and wireless (linux.git).

> (only here has new fix and fixes are verified correct, then John Linville could
> pull from here the iwlwifi.git, right? As I understand, the MAINTAINERS model
> should be the way of iwlwifi.git => wireless.git => net.git => linus)
>

For mainline this workflow seems to be correct, but AFAICS you want an
official statement, so wait for Wey-Yi's answer :-).

> INTEL WIRELESS WIFI LINK (iwlwifi)
> M:      Wey-Yi Guy <wey-yi.w.guy@intel.com>
> M:      Intel Linux Wireless <ilw@linux.intel.com>
> L:      linux-wireless@vger.kernel.org
> W:      http://intellinuxwireless.org
> T:      git git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi/iwlwifi.git
> S:      Supported
> F:      drivers/net/wireless/iwlwifi/
>
>
> By comparing the two, I got that iwlwifi.git really has more, now
> prefetched for tomorrow
> when in office sometimes Internet is not available, and in case
> wireless.git with your patch
> still does not work, I could have a try of iwlwifi.git;
>

If you want latest iwlwifi bits, you can try also the work from there.
But a better strategy would be to test this order:
1. wireless (stable)
2. wireless-next or linux-next (testing)
3. wireless-testing and/or wiwlwifi (experimental)

- Sedat -

> http://git.kernel.org/?p=linux/kernel/git/iwlwifi/iwlwifi.git;a=history;f=drivers/net/wireless/iwlwifi
> http://git.kernel.org/?p=linux/kernel/git/linville/wireless.git;a=history;f=drivers/net/wireless/iwlwifi
>
[...]

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Guy, Wey-Yi]

Hi Renquan,

On Wed, 2011-12-21 at 03:14 -0800, Sedat Dilek wrote:
> On Wed, Dec 21, 2011 at 11:36 AM, Cheng Renquan <crquan@gmail.com> wrote:
> > On Wed, Dec 21, 2011 at 1:20 AM, Sedat Dilek <sedat.dilek@googlemail.com> wrote:
> [...]
> > read these from MAINTAINERS, just want to confirm if iwlwifi.git has more
> > fresh content, this should be the most upstream, right?
> >
> 
> Wey-Yi or Johannes can gave you precise answers on iwwifi development.
> Of course the iwlwifi GIT (master) tree has latest stuff.
> As you might have seen there are different GIT branches for stuff to
> go into wireless-testing (experimental sofware), wireless-next
> (linux-next.git) and wireless (linux.git).
> 
http://git.kernel.org/?p=linux/kernel/git/iwlwifi/iwlwifi.git;a=summary
contain the latest patches(both features and bugs fix) which being
updates on daily base. Like Sadet mention, those are experimental and
not being fully test yet. Once those new patches pass the basic
regression test. the patches will push into two different tree
wireless-next : patch contain new features
http://git.kernel.org/?p=linux/kernel/git/linville/wireless-next.git;a=summary
wireless: bug fix patch
http://git.kernel.org/?p=linux/kernel/git/linville/wireless.git;a=summary
Linville also merge  both wireless-next and wireless  tree  into
wireless-testing for people who want to test the breeding edge wireless
driver changes
http://git.kernel.org/?p=linux/kernel/git/linville/wireless-testing.git;a=summary

of cause, Linville will push the patches to both net-next and net tree
once he believe the patches are ready.

> > (only here has new fix and fixes are verified correct, then John Linville could
> > pull from here the iwlwifi.git, right? As I understand, the MAINTAINERS model
> > should be the way of iwlwifi.git => wireless.git => net.git => linus)
> >
> 
> For mainline this workflow seems to be correct, but AFAICS you want an
> official statement, so wait for Wey-Yi's answer :-).
> 
> > INTEL WIRELESS WIFI LINK (iwlwifi)
> > M:      Wey-Yi Guy <wey-yi.w.guy@intel.com>
> > M:      Intel Linux Wireless <ilw@linux.intel.com>
> > L:      linux-wireless@vger.kernel.org
> > W:      http://intellinuxwireless.org
> > T:      git git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi/iwlwifi.git
> > S:      Supported
> > F:      drivers/net/wireless/iwlwifi/
> >
> >
> > By comparing the two, I got that iwlwifi.git really has more, now
> > prefetched for tomorrow
> > when in office sometimes Internet is not available, and in case
> > wireless.git with your patch
> > still does not work, I could have a try of iwlwifi.git;
> >
> 
> If you want latest iwlwifi bits, you can try also the work from there.
> But a better strategy would be to test this order:
> 1. wireless (stable)
> 2. wireless-next or linux-next (testing)
> 3. wireless-testing and/or wiwlwifi (experimental)
> 
wireless-testing = wireless + wireless-next

Thanks
Wey

> - Sedat -
> 
> > http://git.kernel.org/?p=linux/kernel/git/iwlwifi/iwlwifi.git;a=history;f=drivers/net/wireless/iwlwifi
> > http://git.kernel.org/?p=linux/kernel/git/linville/wireless.git;a=history;f=drivers/net/wireless/iwlwifi
> >
> [...]

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Cheng Renquan]

I cherry-picked these 3 patches on top to v3.2.0-rc6, and now it's working,
Thank you all everyone's effort;

[archuser@archtop linux-git]$ git cherry-pick 123877b80
[b46 55de35a] iwlwifi: do not set the sequence control bit is not needed
 Author: Wey-Yi Guy <wey-yi.w.guy@intel.com>
 1 files changed, 4 insertions(+), 1 deletions(-)
[archuser@archtop linux-git]$ git cherry-pick 81670a49184
[b46 1b0b51d] iwlwifi: tx_sync only on PAN context
 Author: Johannes Berg <johannes.berg@intel.com>
 1 files changed, 6 insertions(+), 0 deletions(-)
[archuser@archtop linux-git]$ git cherry-pick 78feb35b81
[b46 73d79e2] iwlwifi: allow to switch to HT40 if not associated
 Author: Wey-Yi Guy <wey-yi.w.guy@intel.com>
 1 files changed, 2 insertions(+), 2 deletions(-)


When walking to some corner meeting rooms (but still has good
wifi connectivity, verified by another Android phone wifi ;),
the roaming may be from previous AP to none and none to next AP,
the process may need 1 to 2 minutes to be stable again, it's not
ideally swift but acceptable,


Dec 21 14:59:11 archtop NetworkManager[2333]: <info> (wlan0): roamed
from BSSID 24:B6:57:34:CC:5F (blizzard) to 6C:9C:ED:EC:BC:7F
(blizzard)
Dec 21 14:59:41 archtop NetworkManager[2333]: <info> (wlan0): roamed
from BSSID 6C:9C:ED:EC:BC:7F (blizzard) to (none) ((none))
Dec 21 14:59:53 archtop NetworkManager[2333]: <info> (wlan0): roamed
from BSSID (none) ((none)) to 2C:3F:38:F6:A8:AF (blizzard)


Thanks,

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[wwguy]

Hi Renquan,

On Wed, 2011-12-21 at 15:28 -0800, Cheng Renquan wrote:
> I cherry-picked these 3 patches on top to v3.2.0-rc6, and now it's working,
> Thank you all everyone's effort;
> 
> [archuser@archtop linux-git]$ git cherry-pick 123877b80
> [b46 55de35a] iwlwifi: do not set the sequence control bit is not needed
>  Author: Wey-Yi Guy <wey-yi.w.guy@intel.com>
>  1 files changed, 4 insertions(+), 1 deletions(-)
> [archuser@archtop linux-git]$ git cherry-pick 81670a49184
> [b46 1b0b51d] iwlwifi: tx_sync only on PAN context
>  Author: Johannes Berg <johannes.berg@intel.com>
>  1 files changed, 6 insertions(+), 0 deletions(-)
> [archuser@archtop linux-git]$ git cherry-pick 78feb35b81
> [b46 73d79e2] iwlwifi: allow to switch to HT40 if not associated
>  Author: Wey-Yi Guy <wey-yi.w.guy@intel.com>
>  1 files changed, 2 insertions(+), 2 deletions(-)
> 
Thank you for the update, I am glad works for you
> 
> When walking to some corner meeting rooms (but still has good
> wifi connectivity, verified by another Android phone wifi ;),
> the roaming may be from previous AP to none and none to next AP,
> the process may need 1 to 2 minutes to be stable again, it's not
> ideally swift but acceptable,
> 
> 
Yes, I know we have more works to do on roaming, it is on our to-do list

Thanks
Wey
 
> Dec 21 14:59:11 archtop NetworkManager[2333]: <info> (wlan0): roamed
> from BSSID 24:B6:57:34:CC:5F (blizzard) to 6C:9C:ED:EC:BC:7F
> (blizzard)
> Dec 21 14:59:41 archtop NetworkManager[2333]: <info> (wlan0): roamed
> from BSSID 6C:9C:ED:EC:BC:7F (blizzard) to (none) ((none))
> Dec 21 14:59:53 archtop NetworkManager[2333]: <info> (wlan0): roamed
> from BSSID (none) ((none)) to 2C:3F:38:F6:A8:AF (blizzard)
> 
> 
> Thanks,
> --
> To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Cheng Renquan]

I've upgraded to Linux 3.2 and roaming failed randomly again,

after the failure it may recover
sometimes by switch off and on the wifi button on laptop,
sometimes by "modprobe -r iwlwifi && modprobe iwlwifi",
sometimes by "stop networkmanager and start networkmanager",
sometimes recover automatically 10 or 20 minutes later,
sometimes never recover, infinitely looping (can only move to other
place, but same place has good wifi coverage verified by other
colleagues, win/mac users)


have checked it has your recent 78feb35 & 81670a4 commits,
$ git log -n5 --no-merges --oneline v3.2 -- drivers/net/wireless/iwlwifi/
96f1f05 iwlwifi: update SCD BC table for all SCD queues
78feb35 iwlwifi: allow to switch to HT40 if not associated
81670a4 iwlwifi: tx_sync only on PAN context
123877b iwlwifi: do not set the sequence control bit is not needed
9995ffe iwlwifi: change the default behavior of watchdog timer



On Wed, Dec 21, 2011 at 5:40 PM, wwguy <wey-yi.w.guy@intel.com> wrote:
> Yes, I know we have more works to do on roaming, it is on our to-do list
>
> Thanks
> Wey

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Guy, Wey-Yi]

Hi Cheng,

On Tue, 2012-01-10 at 15:17 -0800, Cheng Renquan wrote:
> I've upgraded to Linux 3.2 and roaming failed randomly again,
> 
> after the failure it may recover
> sometimes by switch off and on the wifi button on laptop,
> sometimes by "modprobe -r iwlwifi && modprobe iwlwifi",
> sometimes by "stop networkmanager and start networkmanager",
> sometimes recover automatically 10 or 20 minutes later,
> sometimes never recover, infinitely looping (can only move to other
> place, but same place has good wifi coverage verified by other
> colleagues, win/mac users)
> 
> 
> have checked it has your recent 78feb35 & 81670a4 commits,
> $ git log -n5 --no-merges --oneline v3.2 -- drivers/net/wireless/iwlwifi/
> 96f1f05 iwlwifi: update SCD BC table for all SCD queues
> 78feb35 iwlwifi: allow to switch to HT40 if not associated
> 81670a4 iwlwifi: tx_sync only on PAN context
> 123877b iwlwifi: do not set the sequence control bit is not needed
> 9995ffe iwlwifi: change the default behavior of watchdog timer
> 
Are you saying without those two commits (78feb35 & 81670a4), the
roaming works?

Thanks
Wey 

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Cheng Renquan]

On Tue, Jan 10, 2012 at 2:27 PM, Guy, Wey-Yi <wey-yi.w.guy@intel.com> wrote:
> Hi Cheng,
>
> On Tue, 2012-01-10 at 15:17 -0800, Cheng Renquan wrote:
>> I've upgraded to Linux 3.2 and roaming failed randomly again,
>>
>> after the failure it may recover
>> sometimes by switch off and on the wifi button on laptop,
>> sometimes by "modprobe -r iwlwifi && modprobe iwlwifi",
>> sometimes by "stop networkmanager and start networkmanager",
>> sometimes recover automatically 10 or 20 minutes later,
>> sometimes never recover, infinitely looping (can only move to other
>> place, but same place has good wifi coverage verified by other
>> colleagues, win/mac users)
>>
>>
>> have checked it has your recent 78feb35 & 81670a4 commits,
>> $ git log -n5 --no-merges --oneline v3.2 -- drivers/net/wireless/iwlwifi/
>> 96f1f05 iwlwifi: update SCD BC table for all SCD queues
>> 78feb35 iwlwifi: allow to switch to HT40 if not associated
>> 81670a4 iwlwifi: tx_sync only on PAN context
>> 123877b iwlwifi: do not set the sequence control bit is not needed
>> 9995ffe iwlwifi: change the default behavior of watchdog timer
>>
> Are you saying without those two commits (78feb35 & 81670a4), the
> roaming works?

No, the v3.2-rc6 which doesn't include (78feb35 & 81670a4) also has no
roaming working,

The week before Xmas I tested v3.2-rc7 which include your (78feb35 &
81670a4) really has
roaming working, roamed to new AP stable in 1 minute in most meeting
rooms of the office,

Now this week 3.2 came out, since it has on changes under
"drivers/net/wireless/iwlwifi/",
I think it should be same with 3.2-rc7; but roaming not good again,
and may go off even not moving, staying same place;


[archuser@archtop linux-git]$ git log --no-merges -n5 --oneline
v3.2-rc6 -- drivers/net/wireless/iwlwifi/
9995ffe iwlwifi: change the default behavior of watchdog timer
34a5b4b iwlwifi: do not re-configure HT40 after associated
274b89c iwlagn: fix HW crypto for TX-only keys
43e5885 iwlwifi: avoid a panic when unloading the module with RF Kill
f956c34 iwlwifi: don't perform "echo test" when cmd queue stuck

[archuser@archtop linux-git]$ git log --no-merges -n5 --oneline
v3.2-rc7 -- drivers/net/wireless/iwlwifi/
96f1f05 iwlwifi: update SCD BC table for all SCD queues
78feb35 iwlwifi: allow to switch to HT40 if not associated
81670a4 iwlwifi: tx_sync only on PAN context
123877b iwlwifi: do not set the sequence control bit is not needed
9995ffe iwlwifi: change the default behavior of watchdog timer


The difference with that week before Xmas I think may be now the
office has more people
since everyone has come back from holidays;

In a 1hour meeting on Monday (yesterday) with 10+ people in a meeting room,
every other user (win&mac) are online except me: trying get iwlwifi working,
throughout all 60 minutes my iwlwifi get roaming looping cannot
recover (tried all above methods)
and when the meeting finished, other ones left, my iwlwifi started
working; recovered in 60 minutes; how sad

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Guy, Wey-Yi]

Hi Cheng,

On Tue, 2012-01-10 at 16:01 -0800, Cheng Renquan wrote:
> On Tue, Jan 10, 2012 at 2:27 PM, Guy, Wey-Yi <wey-yi.w.guy@intel.com> wrote:
> > Hi Cheng,
> >
> > On Tue, 2012-01-10 at 15:17 -0800, Cheng Renquan wrote:
> >> I've upgraded to Linux 3.2 and roaming failed randomly again,
> >>
> >> after the failure it may recover
> >> sometimes by switch off and on the wifi button on laptop,
> >> sometimes by "modprobe -r iwlwifi && modprobe iwlwifi",
> >> sometimes by "stop networkmanager and start networkmanager",
> >> sometimes recover automatically 10 or 20 minutes later,
> >> sometimes never recover, infinitely looping (can only move to other
> >> place, but same place has good wifi coverage verified by other
> >> colleagues, win/mac users)
> >>
> >>
> >> have checked it has your recent 78feb35 & 81670a4 commits,
> >> $ git log -n5 --no-merges --oneline v3.2 -- drivers/net/wireless/iwlwifi/
> >> 96f1f05 iwlwifi: update SCD BC table for all SCD queues
> >> 78feb35 iwlwifi: allow to switch to HT40 if not associated
> >> 81670a4 iwlwifi: tx_sync only on PAN context
> >> 123877b iwlwifi: do not set the sequence control bit is not needed
> >> 9995ffe iwlwifi: change the default behavior of watchdog timer
> >>
> > Are you saying without those two commits (78feb35 & 81670a4), the
> > roaming works?
> 
> No, the v3.2-rc6 which doesn't include (78feb35 & 81670a4) also has no
> roaming working,
> 
> The week before Xmas I tested v3.2-rc7 which include your (78feb35 &
> 81670a4) really has
> roaming working, roamed to new AP stable in 1 minute in most meeting
> rooms of the office,
> 
> Now this week 3.2 came out, since it has on changes under
> "drivers/net/wireless/iwlwifi/",
> I think it should be same with 3.2-rc7; but roaming not good again,
> and may go off even not moving, staying same place;
> 
> 
> [archuser@archtop linux-git]$ git log --no-merges -n5 --oneline
> v3.2-rc6 -- drivers/net/wireless/iwlwifi/
> 9995ffe iwlwifi: change the default behavior of watchdog timer
> 34a5b4b iwlwifi: do not re-configure HT40 after associated
> 274b89c iwlagn: fix HW crypto for TX-only keys
> 43e5885 iwlwifi: avoid a panic when unloading the module with RF Kill
> f956c34 iwlwifi: don't perform "echo test" when cmd queue stuck
> 
> [archuser@archtop linux-git]$ git log --no-merges -n5 --oneline
> v3.2-rc7 -- drivers/net/wireless/iwlwifi/
> 96f1f05 iwlwifi: update SCD BC table for all SCD queues
> 78feb35 iwlwifi: allow to switch to HT40 if not associated
> 81670a4 iwlwifi: tx_sync only on PAN context
> 123877b iwlwifi: do not set the sequence control bit is not needed
> 9995ffe iwlwifi: change the default behavior of watchdog timer
> 
> 
> The difference with that week before Xmas I think may be now the
> office has more people
> since everyone has come back from holidays;
> 
> In a 1hour meeting on Monday (yesterday) with 10+ people in a meeting room,
> every other user (win&mac) are online except me: trying get iwlwifi working,
> throughout all 60 minutes my iwlwifi get roaming looping cannot
> recover (tried all above methods)
> and when the meeting finished, other ones left, my iwlwifi started
> working; recovered in 60 minutes; how sad

Got it, very sorry for the problem you experience, yes, I agree with
you, it might be because the noisy environment. Could you please send me
the dmesg log with the following flags

$sudo modprobe iwlwifi debug=0x1083

Thanks
Wey 

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[Cheng Renquan]

On Tue, Jan 10, 2012 at 3:20 PM, Guy, Wey-Yi <wey-yi.w.guy@intel.com> wrote:
> Hi Cheng,

> Got it, very sorry for the problem you experience, yes, I agree with
> you, it might be because the noisy environment. Could you please send me
> the dmesg log with the following flags
>
> $sudo modprobe iwlwifi debug=0x1083
>
> Thanks
> Wey

Attached iwlwifi-debug-jan10.txt.xz privately to Guy since it may have
some privacy; please do not spread;

tried a few meeting rooms only the last one failed roaming; may be
because many ones have left office right now, don't have so much
noise; will try again another busy time;

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[wwguy]

On Tue, 2012-01-10 at 17:30 -0800, Cheng Renquan wrote:
> On Tue, Jan 10, 2012 at 3:20 PM, Guy, Wey-Yi <wey-yi.w.guy@intel.com> wrote:
> > Hi Cheng,
> 
> > Got it, very sorry for the problem you experience, yes, I agree with
> > you, it might be because the noisy environment. Could you please send me
> > the dmesg log with the following flags
> >
> > $sudo modprobe iwlwifi debug=0x1083
> >
> > Thanks
> > Wey
> 
> Attached iwlwifi-debug-jan10.txt.xz privately to Guy since it may have
> some privacy; please do not spread;
> 
> tried a few meeting rooms only the last one failed roaming; may be
> because many ones have left office right now, don't have so much
> noise; will try again another busy time;

hmm, so you thinking it might be the noisy environment. what  if you
rollback the kernel to 3.2.0-rc7 (the one you have before Christmas) and
try it in the same noisy environment and see if it works. It can help me
understand if it is something break the roaming lately, or the problem
always there.

Thanks
Wey

Re: linux 3.2-rcX iwlwifi driver's problem (wifi link unstable when moving between multiple APs)

[wwguy]

Hi Renquan,

On Tue, 2012-01-10 at 17:30 -0800, Cheng Renquan wrote:
> On Tue, Jan 10, 2012 at 3:20 PM, Guy, Wey-Yi <wey-yi.w.guy@intel.com> wrote:
> > Hi Cheng,
> 
> > Got it, very sorry for the problem you experience, yes, I agree with
> > you, it might be because the noisy environment. Could you please send me
> > the dmesg log with the following flags
> >
> > $sudo modprobe iwlwifi debug=0x1083
> >
> > Thanks
> > Wey
> 
> Attached iwlwifi-debug-jan10.txt.xz privately to Guy since it may have
> some privacy; please do not spread;
> 
> tried a few meeting rooms only the last one failed roaming; may be
> because many ones have left office right now, don't have so much
> noise; will try again another busy time;

>From the log you provide, I can see both "direct probe timeout" and
"authentication timeout".

You mention it is in an noisy environment. The thing I am not sure is
this a device fail to transmit issue or fail to receive "ack" from AP
issue?
 
Is it possible for you to capture the sniffer trace if you have the
tools?


Thanks
Wey