WiFi - AX200 F38 - Bandwidth issue after upgrade

Ask Fedora
,
1

Hi,

I have been enjoying a smooth experience for the last 12+ months.
Upgrades between versions went fine with no issues, until recent update when the WiFi started acting weirdly.

Download is 8-10 times slower than before, what is even more weird that it’s only in ONE direction - Download. Upload speeds are the same.
It’s definitely the problem with OS/Laptop, as from the same spot, connected to the same AP on T480s with W10 max out the speeds.

Had similar issue about 1.5/2 years ago when never kernel upgrade caused same issue.

Setup:
Thinkpad T480 with AX200 installed.
U6-pro
Latest kernel (6.3.8), latest FWs installed.
linux-firmware-whence-20230515-150.fc38.noarch
linux-firmware-20230515-150.fc38.noarch

iwl:

[   16.237307] iwlwifi 0000:03:00.0: enabling device (0000 -> 0002)
[   16.355668] iwlwifi 0000:03:00.0: **api flags index 2 larger than supported by driver**
[   16.355693] iwlwifi 0000:03:00.0: TLV_FW_FSEQ_VERSION: FSEQ Version: 89.3.35.37
[   16.356201] iwlwifi 0000:03:00.0: loaded firmware version 74.a5e9588b.0 cc-a0-74.ucode op_mode iwlmvm
[   16.752638] iwlwifi 0000:03:00.0: Detected Intel(R) Wi-Fi 6 AX200 160MHz, REV=0x340
[   16.910202] iwlwifi 0000:03:00.0: Detected RF HR B3, rfid=0x10a100
[   16.981395] iwlwifi 0000:03:00.0: base HW address: 8c:c6:81:d7:df:43
[   17.179051] iwlwifi 0000:03:00.0 wlp3s0: renamed from wlan0

iwconfig

wlp3s0    IEEE 802.11  ESSID:"XXXXXXXX"  
          Mode:Managed  Frequency:5.22 GHz  Access Point: D0:21:F9:55:35:67   
          Bit Rate=1.4413 Gb/s   Tx-Power=20 dBm   
          Retry short limit:7   RTS thr:off   Fragment thr:off
          Power Management:off
          Link Quality=49/70  Signal level=-61 dBm  
          Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:0
          Tx excessive retries:0  **Invalid misc:233**   Missed beacon:0

Have the below defined for ages, tried to disable it, as there were some indications in the journal.
Even disabling does not help - restarted sysctl and the laptop - no change if enabled or disabled.

net.ipv4.tcp_no_metrics_save = 1
net.core.default_qdisc=fq
net.ipv4.tcp_congestion_control=bbr

journal -b
Jun 17 08:23:59 onewithforce systemd-sysctl[300]: Couldn't write 'bbr' to 'net/ipv4/tcp_congestion_control', ignoring: No such file or directory

Problem:

fallocate -l 3G test_image.img

Uploading to a local storage:
[alex@onewithforce Downloads]$ rsync -avh --progress test_image.img root@192.168.1.150:/Data/public/
sending incremental file list
test_image.img
          3.22G 100%   97.93MB/s    0:00:31 (xfr#1, to-chk=0/1)

sent 3.22G bytes  received 35 bytes  102.29M bytes/sec
total size is 3.22G  speedup is 1.00

Downloading same file locally (to nvme drive)
rm test_image.img
[alex@onewithforce Downloads]$ rsync -avh --progress root@192.168.1.150:/Data/public/test_image.img .
receiving incremental file list
test_image.img
          3.22G 100%    8.87MB/s    0:05:46 (xfr#1, to-chk=0/1)

sent 43 bytes  received 3.22G bytes  9.27M bytes/sec

Problem started more than a week ago after system upgrade and restart.
Tried moving back to 6.2, did not help. linux-firmware package downgrade does fail due to conflict of previous package.

Anyone else facing similar issues ?

P.S: One observation that Invalid misc counter does increase. Value before the test is above, after the test counter almost doubled:

wlp3s0    IEEE 802.11  ESSID:"XXXXXXXX"  
          Mode:Managed  Frequency:5.22 GHz  Access Point: D0:21:F9:55:35:67   
          Bit Rate=1.4413 Gb/s   Tx-Power=20 dBm   
          Retry short limit:7   RTS thr:off   Fragment thr:off
          Power Management:off
          Link Quality=53/70  Signal level=-57 dBm  
          Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:0
          Tx excessive retries:0  **Invalid misc:423**   Missed beacon:0
2

Does the problem go away when you boot an older kernel? Does the system boot another OS? Do you have the latest BIOS updates?

Laptops usually configure wifi at startup to support network booting, so there can be conflicts when BIOS updates are out of sync with OS drivers. Problems sometimes disappear with BIOS updates or powering down rather than a reboot.

Issue with wifi are not unusual, so I find it handy to have a USB wifi dongle to use while waiting for an update that resolves the problem.

1 Like
3

I do keep only a few kernels , booted on 6.2 same issue.
It’s a single OS laptop, have a VM in boxes, as there is no VCDS for Linux.

Latest BIOS an firmware:

[alex@onewithforce ~]$ sudo fwupdmgr update
[sudo] password for alex: 
Devices with no available firmware updates: 
 • SSD 970 EVO 500GB
 • Thunderbolt host controller
 • UEFI Device Firmware
 • UEFI Device Firmware
Devices with the latest available firmware version:
 • Embedded Controller
 • Intel Management Engine
 • System Firmware
 • UEFI dbx

I have had a rock solid connection for more than an year.
Issue is even more strange, as it seems like after a sleep/hibernation it’s gone away.

[alex@onewithforce Downloads]$ uname -a;date;uptime
Linux onewithforce 6.3.8-200.fc38.x86_64 #1 SMP PREEMPT_DYNAMIC Thu Jun 15 02:15:40 UTC 2023 x86_64 GNU/Linux
Mon Jun 19 08:05:19 PM EEST 2023
 20:05:19 up 2 days, 11:41,  2 users,  load average: 0.30, 0.73, 0.96
[alex@onewithforce Downloads]$ rsync -avh --progress root@192.168.1.150:/Data/public/test_image.img .
receiving incremental file list
test_image.img
          3.22G 100%   75.80MB/s    0:00:40 (xfr#1, to-chk=0/1)

sent 43 bytes  received 3.22G bytes  77.64M bytes/sec
total size is 3.22G  speedup is 1.00

Nothing was changed on the system, since I posted it.
I barely reboot the laptop, but for the sake of results can do it in the upcoming days to see if the performance would be as initially posted.

4

Others have reported similar experiences. If there was a failure in transition from
BIOS (netboot?) wifi configuration to Fedora’s configuration (different firmware?) that could explain the issue. It may be that recovery from sleep/hibernation transitions from a low-power wifi mode (to support wake-on-lan) and the restart bypasses the BIOS configuration step. There may be BIOS options to disable wake-on-lan and/or netboot. If the issue was a transition from BIOS to Fedora wifi configuration it may go away once updates have made both configurations place nicely. Maybe there is a way to boot directly to sleep/hibernation to get the benefits of wakeup.

5

Netboot/WOL/PXE are from the first things I disable in BIOS.
All firmware updates went through LVFS.
Invalid misc counter stopped after the sleep/hibernation.

image
image925×289 35.7 KB

6

I wonder if the WiFi is initially started in low-power mode and gets full power at wakeup. It is not unknown for vendors to play tricks with certification requirements for power consumption.

7

Actually it does make sense.
Some fresh observations from yesterday. Rebooted intentionally to validate the bios settings:
WOL/PXE/Network boot disabled.
After a reboot the WiFi speed is again slow.
Initial hibernation/sleep does not restore the speed. Either it require more time or more hibernation/sleep cycles.
Slow download after a reboot:

image
image944×509 59.2 KB

Config after reboot, surprise, surprise - wifi power mgmt is on.

wlp3s0    IEEE 802.11  ESSID:"XXXXXX"  
          Mode:Managed  Frequency:5.22 GHz  Access Point: D0:21:F9:55:35:67   
          Bit Rate=1.9215 Gb/s   Tx-Power=20 dBm   
          Retry short limit:7   RTS thr:off   Fragment thr:off
          **Power Management:on**
          Link Quality=55/70  Signal level=-55 dBm  
          Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:0
          Tx excessive retries:0  **Invalid misc:167**   Missed beacon:0

Disabled it in a following way

[connection]
wifi.powersave = 2
systemctl restart NetworkManager

New value:

wlp3s0    IEEE 802.11  ESSID:"XXXXXXX"  
          Mode:Managed  Frequency:5.22 GHz  Access Point: D0:21:F9:55:35:67   
          Bit Rate=-2.13367e+06 kb/s   Tx-Power=20 dBm   
          Retry short limit:7   RTS thr:off   Fragment thr:off
          **Power Management:off**
          Link Quality=55/70  Signal level=-55 dBm  
          Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:0
          Tx excessive retries:0  **Invalid misc:29**   Missed beacon:0

Another message, but I have seen it before as well:

[17113.160729] wlp3s0: Limiting TX power to 20 (23 - 3) dBm as advertised by d0:21:f9:55:35:67

Slow download speed again:

image
image906×482 58 KB

8

After a bit of fighting with this random issue, turned out it had nothing to do with upgrade.
Tried many things rolling back firmware/drivers and etc and made a good move to retest with my work thinkpad (w10) and this time got slower speed as well.

Turned out the main problem have been the link between the PoE switch and AP and it did happen occasionally.
There were some CRC errors reported on the port powering the AP, counter increased when I was doing the tests.
Plugged again the ports, power cycled the switch and all looks ok now. No CRC errors and speed is as it should. Not sure if it was a bad connector or a netgear wanting to be replaced with Mikrotik.

image
image934×198 15.9 KB 

[alex@onewithforce ~]$ rsync -avh --progress root@192.168.1.150:/Data/public/test_image.img .
receiving incremental file list
test_image.img
          3.22G 100%   89.42MB/s    0:00:34 (xfr#1, to-chk=0/1)

sent 43 bytes  received 3.22G bytes  90.76M bytes/sec
total size is 3.22G  speedup is 1.00
9

Sometimes routers & APs seem to flake out (as do some computers) and a simple power cycle or reboot gets things back on track.

If your problem has been solved please mark the post that provided the solution so others may find the answer should they have a similar issue.

10

See: Wikipedia Soft Errors. If the problem is repeatable, check for updates from the vendor as well as vendor forums where such issues may be discussed.