Another intel iwlwifi issue in 16.04

[ericwoud]

Here the same problem. Restarting network manager does not work. Nothing seems to work to reset the adapter, except a reboot. See part of dmesg log when the first error comes in:

[ 410.000601] iwlwifi 0000:02:00.0: Queue 16 stuck for 10000 ms.
[ 410.000674] iwlwifi 0000:02:00.0: Current SW read_ptr 145 write_ptr 147
[ 410.033748] ------------[ cut here ]------------
[ 410.033806] WARNING: CPU: 3 PID: 0 at /build/linux-fEAasX/linux-4.9.0/drivers/net/wireless/intel/iwlwifi/pcie/trans.c:1864 iwl_trans_pcie_grab_nic_access+0xeb/0xf0 [iwlwifi]
[ 410.033813] Timeout waiting for hardware access (CSR_GP_CNTRL 0xffffffff)

To reset the adapter, so i could use wifi again, wrote the following script:

#find /sys -name wlan0 -path *devices/pci*
# result: /sys/devices/pci0000:00/0000:00:1c.1/0000:02:00.0/net/wlan0
echo -n 1 > /sys/devices/pci0000:00/0000:00:1c.1/0000:02:00.0/remove
echo -n 1 > /sys/devices/pci0000:00/0000:00:1c.1/rescan
service network-manager restart

Use the find command in the first line to find the correct adress in your system. Exclude the part with /net/wlan0

However, after being stuck and applied once, the problem does not reappear until after the next reboot. So as a 'fix' let it start at boottime with systemd:

sudo gedit /etc/systemd/system/startup.service

[Unit]
Description=Startup Service

[Service]
Type=simple
ExecStart=/bin/su -c "/bin/echo -n 1 > /sys/devices/pci0000:00/0000:00:1c.1/0000:02:00.0/remove ; \
/bin/echo -n 1 > /sys/devices/pci0000:00/0000:00:1c.1/rescan ; \
/usr/sbin/service network-manager restart"

[Install]
WantedBy=multi-user.target

And install with:
sudo systemctl preset startup.service

Somehow this fixes the problem of my 3160 intel wireless adapter to hang after 10 minutes use or so... This works on my Lenovo Yoga 300 11-IBY (80M0) with intel 3160 wifi adapter.

[DiCon]

Thank you so much, ericwoud. The same problem was bothering me for a while now on an Intel 7260 card and I just stumbled on your workaround. And indeed, it does not just get the network running again without rebooting, it actually fixes the problem until the next reboot, which strike me as rather odd...

So, for reference, that it seems to be the same problem, here is a typical dmesg excerpt from my system (which is a Dell XPS 12 9q33):

[  779.792010] iwlwifi 0000:06:00.0: Queue 16 stuck for 10000 ms.
[  779.792034] iwlwifi 0000:06:00.0: Current SW read_ptr 96 write_ptr 162
[  779.809645] ------------[ cut here ]------------
[  779.809670] WARNING: CPU: 2 PID: 0 at /home/kernel/COD/linux/drivers/net/wireless/intel/iwlwifi/pcie/trans.c:1864 iwl_trans_pcie_grab_nic_access+0xeb/0xf0 [iwlwifi]
[  779.809673] Timeout waiting for hardware access (CSR_GP_CNTRL 0xffffffff)

Interestingly, for me the problem seems to be related to the load of a network. At home this bug occurs about once a week, but in my office at the uni it happens randomly about 10 times a day. In an empty lecture hall, while preparing the lecture, it happens about once per hour, but when the lecture hall is filled with 300 students, I can hardly check my mail before the connection drops and I am more or less constantly rebooting to get a connection. (Well, not anymore, thanks to you!)

Did you report this anywhere else or do you have any idea if this is Ubuntu-specifiy or an iwl problem? I would like to submit a bug report and contribute to a "real" solution... So far, I reproduced this bug with Ubuntu 16.10 from a USB stick, my regular 16.10 installation and with 16.10 and the most recent mainline kernel.

[chili555]

I have two Intel wireless devices, including a 7260 that runs perfectly. I'd like to offer some suggestions.

First, let's disable power saving in Network Manager; from the terminal:

sudo sed -i 's/wifi.powersave = 3/wifi.powersave = 2/' /etc/NetworkManager/conf.d/default-wifi-powersave-on.conf

Next, be certain that you have the latest firmware:

wget http://mirrors.kernel.org/ubuntu/pool/main/l/linux-firmware/linux-firmware_1.162_all.deb
sudo dpkg -i linux-firmware*.deb

Reboot.

Next, check the settings in the router. WPA2-AES is preferred; not any WPA and WPA2 mixed mode and certainly not TKIP. Second, if your router is capable of N speeds, you may have better connectivity with a channel width of 20 MHz in the 2.4 GHz band instead of automatic 20/40 MHz, although it is likely to affect N speeds. I also have better luck with a fixed channel, either 1, 6 or 11, rather than automatic channel selection. Also, be certain the router is not set to use N speeds only; auto B, G and N is preferred. After making these changes, reboot the router.

If you are in a situation where changing the router is not feasible, a landlord, university, etc., you may get a connection even with TKIP with a driver parameter:

sudo -i
echo "options iwlwifi swcrypto=1"  >>  /etc/modprobe.d/iwlwifi.conf
exit

Then set your regulatory domain explicitly. Check yours:

sudo iw reg get

If you get 00, that is a one-size-maybe-fits-all setting. Find yours here: http://en.wikipedia.org/wiki/ISO_3166-1_alpha-2 Then set it temporarily:

sudo iw reg set IS

Of course, substitute your country code if not Iceland. Set it permanently:

gksudo gedit /etc/default/crda

Use nano or kate or leafpad if you don't have the text editor gedit.

Change the last line to read:

REGDOMAIN=IS

Proofread carefully, save and close the text editor.

Next, I'd set IPv6 to Ignore in Network Manager: http://docs.fedoraproject.org/en-US/...pv6-ignore.png This example is for ethernet, but you want wireless.

After making these changes and after a reboot, check the logs again:

dmesg | grep iwl

Any more of those ugly stuck queue messages?

[DiCon]

First of all, I have to take back that the script fixes the problem until the next reboot. Seems like I was just lucky, because today, the script allows me to reconnect to the network without rebooting (still better than without), but it will break down again after a while. So maybe it's not the same bug after all...

@chili555:
Thanks for all the suggestions. I already have the latest firmware (iwlwifi-7260-17.ucode, also check that it's used with modinfo iwlwifi) and the correct regulatory domain (de). Also, I experience the problems on different networks with different severity. Not sure about the settings at home, but here at the uni it is a mixed mode (which I cannot change) network using CCMP (also certainly no TKIP at home). So, I just tested the power saving mode, but it did not help...

Also, the Wifi worked great for almost three years before these problems occurred a few months ago - not exactly sure when, but I remember that the problem was fairly new and that I was hoping that it would be resolved when updating to 16.10, which I did on the 5th of November according to my logs. So, maybe I should grab an older Ubuntu version, run it from USB and upgrade until it breaks again...

[chili555]

I already have the latest firmware (iwlwifi-7260-17.ucode, also check that it's used with modinfo iwlwifi

modinfo shows the firmware that the driver version will use if available. The only way to verify what is actually loaded is:

dmesg | grep iwl

My system loads -17:

[    4.511808] iwlwifi 0000:03:00.0: loaded firmware version 17.352738.0 op_mode iwlmvm

Incidentally, the driver will sometimes load a newer version than listed in modinfo; for example -18 or -19. Generally, however, experimental not-ready-for-prime-time firmware files are not helpful.

Did you try swcrypto=1?

[DiCon]

I have to admit, that I didn't try it at first as I thought you found it helpful in TKIP networks. Now I tried and still have disconnects... It took a while longer (at least I think so - it was disconnected when I returned to the office after a meeting), but that's probably due to the randomness of the problem.

About the firmware version: It's definitely 17. I tried to downgrade it to 16 since the release of 17 into the Ubuntu repository is close to my estimate of when the trouble started. But my iwlwifi does not accept anything below 17 and also 17 is the most recent available - so I will have to test an older kernel... (Of course I now also double-checked with dmesg and it's identical to your version.)

If I understand it correctly, there will be no new versions after 17 unless there are backports - but since you mention experimental firmware: Is there a source for such firmware that I am not aware of?

[chili555]

Is there a source for such firmware that I am not aware of?

My resource is here: https://github.com/OpenELEC/iwlwifi-...aster/firmware As you can see, -17 is the newest they have.

I have to admit, that I didn't try it at first as I thought you found it helpful in TKIP networks. Now I tried and still have disconnects..

At this point, I'd try anything.

If you'd rather remove it, simply edit the file and remove the last line that you added above.

[ericwoud]

Then I think we are talking about different bugs. My solution applies to a Lenovo Yoga 300 11-IBY (80M0). I can confirm the solution still works, after a month still no stuck wifi adapter (intel 3160) at all.

[rob.johnson]

I hope I'm not too late, but I'm running into a very similar issue on 16.04.3 with an Intel 8265 card. I too get the 'Warning: CPU' messages when allowing iwlwifi on boot. chili555 responded to my question on AskUbuntu to suggest adding iwlwifi to the modprobe blacklist, and that does indeed let the machine boot up correctly. I notice that with iwlwifi enabled, i have absolutely no network connections once the GUI loads, even the Intel onboard ethernet on my Dell Precision M3520. Attempting to sudo causes the terminal to hang.

So back on topic, with iwlwifi blacklisted at boot, the GUI loads. I've then manually modprobe'd iwlwifi once the machine is fully up and wifi then works properly. At a guess, a dependency isn't being loaded, or is being loaded in the wrong order. Out of paranoia, I then decided to unload the module before putting my laptop into sleep mode to improve my chances of getting it back:

root@M3520:~# rmmod iwlwifi
rmmod: ERROR: Module iwlwifi is in use by: iwlmvm
root@M3520:~# rmmod iwlmvm

At this point, the terminal hangs again. If I switch to another tab and attempt to sudo, that hangs as well. So based on this, I think the culprit may be this iwlmvm module instead. For now, manually loading iwlwifi when I need wireless is a viable workaround. Anyone else found a solution?

[chili555]

I think the correct method is:

sudo modprobe -r iwlwifi

And then it should take out all the dependencies with it.

You might also be certain that the default /etc/modprobe.d/iwlwifi.conf file is in place.

# /etc/modprobe.d/iwlwifi.conf
# iwlwifi will dyamically load either iwldvm or iwlmvm depending on the
# microcode file installed on the system.  When removing iwlwifi, first
# remove the iwl?vm module and then iwlwifi.
remove iwlwifi \
(/sbin/lsmod | grep -o -e ^iwlmvm -e ^iwldvm -e ^iwlwifi | xargs /sbin/rmmod) \
&& /sbin/modprobe -r mac80211

FYI, the module dependencies are found with modinfo:

$ modinfo iwlwifi
filename:       /lib/modules/4.13.0-25-generic/kernel/drivers/net/wireless/intel/iwlwifi/iwlwifi.ko
license:        GPL
author:         Copyright(c) 2003- 2015 Intel Corporation <linuxwifi@intel.com>
description:    Intel(R) Wireless WiFi driver for Linux
<snip>
depends:        cfg80211

As well as:

$ modinfo iwlmvm
filename:       /lib/modules/4.13.0-25-generic/kernel/drivers/net/wireless/intel/iwlwifi/mvm/iwlmvm.ko
license:        GPL
author:         Copyright(c) 2003- 2015 Intel Corporation <linuxwifi@intel.com>
description:    The new Intel(R) wireless AGN driver for Linux
srcversion:     EBD7611E6805E2BDE8304CC
depends:        iwlwifi,mac80211,cfg80211
<snip>

You can get the module to load even though it's blacklisted after a short wait by adding to /etc/rc.local so that it reads:

sleep 3
modprobe iwlwifi

exit  0

You might have to play with the sleep 3 (seconds) a bit. It might need 4 or it might work perfectly well with 2 or 1.