AX210 loses SMB session mid-copy on 802.11ac/802.11ax

[ahershler]

I installed an after market Intel AX210 in my Lenovo Legion Y740 after the OEM Killer network card died.

Ever since, I have experienced a strange phenomenon:

If I start copying files over the WiFi network to a NAS using the SMB protocol, and, while the copy operation is taking place, also browse the Internet (for example using Firefox or Chrome, or opening an internet-hosted PDF with Adobe Reader), the copy operation fails. If I make sure nothing uses the Internet while the copy operation is ongoing, the copy completes just fine. Please note I never lose WiFi connectivity (verified by checking the Event Log). The computer is located 2 meters from the AP without any obstructions.

I can easily reproduce this issue by attempting to copy a large number of largish files while refreshing a page in the browser. It usually takes just 2-3 refreshes for the copy operation to fail.

To verify that this issue is caused by the AX210, I disabled the AX210 and connected my Y740 to the same WiFi network using USB3 tethering on an Android phone, with the phone being connected to the same WiFi router, using 802.11ac. When the Y740 is connected in this way, I can complete a large SMB copy while browsing the Internet without any issues.

If I change the "802.11n/802.11ac/802.11ax Wireless Mode" setting (in the Intel AX210 Properties dialog, Advanced tab) to 802.11n, the issue does not occur.

Technical details:

Lenovo Legion Y740 with Intel i7-9750H CPU and 64 GB RAM

OS: Windows Pro 10.0.19044.2251 (21H2)

AX210 drivers: 22.170.0.3 (latest currently available, but the issue can also be reproduced on earlier versions)

AP: Asus RT-AC68 (access point issues have been ruled out by the above described test using USB tethering a phone to connect to the same AP. Furthermore, other computers connected to the same AP are not experiencing these issues)

To rule out any interference from Bluetooth, Bluetooth was turned off in Windows.

Windows Event log for SMBClient contains Event 30804 (Error) followed by Events 30805 and 30807 (Warning), followed immediately afterwards (at the same second) by Events 30806 and 30808 (Informational). No Netwtw10 events are found around the same time.

WireShark capture shows SMB starting off with a SMB2 SessionSetup Request, copying proceeding normally with ACKs, when, at the time of the 30804 error in the Event Log, the WireShark capture contains an ARP broadcast by the AX210 asking for the address of the NAS, followed immediately by a SMB2 NegotiateProtocol Request. This, to me, indicates the loss of the session and the reconnection immediately afterwards. If it didn't happen in the midst of a copying operation, it would be unnoticeable. The ARP broadcast suggests to me the AX210 goes through a type of reset or restart.

Attempts to resolve:

Followed instructions in https://www.intel.com/content/www/us/en/support/articles/000024678/wireless.html to use Intel Recommended Settings for 802.11ac Connectivity

Followed instructions in https://www.intel.com/content/www/us/en/support/articles/000022173/wireless.html for a Clean Installation of Wireless Drivers

My issue somewhat resembles the issue described in this post by @DDiepenbrock: https://community.intel.com/t5/Wireless/Data-loss-with-Windows-Intel-drivers-for-ax201-and-ax210/m-p/1422254#M43653 Data Loss with Windows Intel drivers for ax201 and ax210, in that this post likewise found that the issue disappeared when forcing the NIC to 802.11n.

I'd greatly appreciate any help in trying to resolve this issue.

[Jean_Intel]

Hello ahershler,

I appreciate the information provided.

You Should know that Intel® does not support wireless integrations. We recommend you speak to the manufacturer of your equipment for support on wireless integrations and replacement parts to ensure compliance with the regulatory requirements of your country and support, as mentioned in my previous post. It is important to check for compatibility with the motherboard manufacturer, as there could be certain cards that will not work with the system.

For more information, read the following article about Regulatory Information Regarding Wireless Hardware Installation or Upgrade.

Please keep in mind that this thread will no longer be monitored by Intel. Thank you for your understanding.

Best regards, 

Jean O.  

Intel Customer Support Technician

[DDiepenbrock]

ahershler, this sounds exactly like issues I've seen with dropped zoom calls when configured for 802.11ac or 802.11ax.  When I configure for 802.11n I don't have a problem.  I was able to demonstrate significant data loss when using iperf3, which I've written this up here:  https://community.intel.com/t5/Wireless/Data-loss-with-Windows-Intel-drivers-for-ax201-and-ax210/m-p/1422254

Unfortunately Intel has thus far provided no meaningful help, and only questioned that perhaps it's environmental -- which is nonsense, I've clearly explained how that is not the case.

I sure hope Intel can actually get some engineers to look into this problem with their driver.  I'm happy to provide more details, but so far I've only been able to communicate with customer support techs who don't really understand networking.

[ahershler]

Thank you for your reply @DDiepenbrock, and also thanks for mentioning my post in your original thread.

As you will have noticed, @Jean_Intel wrote above that "this thread will no longer be monitored by Intel", which, to me, shows how Intel customer support prefers to "handle" these issues.

I have performed tests with iperf3 and was able to replicate your results, that is: I see the same data loss you have documented in your thread.

I wonder whether you would be able to replicate my results (which, I think, would be valuable as it would (further) rule out that this issue could be caused by the computer, the access point, the environment, etc.). Are you able to easily perform a copy of a largish file over SMB2/SMB3 to a file server, while refreshing a web page in Firefox or Chrome? The file should be large enough so that the copy operation will take 15 seconds or longer over your network, providing enough opportunity to refresh a web page twice or three times during the copy. Try this with the NIC configured for 802.11ac or 802.11ax - the file copy should fail. Then try again with the NIC configured for 802.11n - the file copy should succeed.

Please note that the use of SMB2 or SMB3 is required for this test, to force SMB to communicate completely over TCP/IP. The server should be configured to refuse SMB1.

There are two issues I do not (yet) completely understand:
1. According to spec, SMB2/SMB3 is using TCP on port 445. However, the data loss demonstrated using iperf3 only occurs when using UDP; if using TCP, I do not see any data loss, as TCP includes "guaranteed delivery". So even if data loss occurs, TCP resends the lost packets, and a higher level protocol like SMB2/SMB3 should not even be aware that data loss occurred. In other words, the data loss issue demonstrated using iperf3 cannot, by itself, be the cause for the issue I see during SMB2/SMB3 file copy. As I wrote above, the entries in the event log suggest to me that the AX210 is going through some sort of reset. As SMB is a session-based protocol, if this reset results in losing the session, it would cause a failed copy operation.

2. Why do I need to refresh a web page to cause the SMB file copy operation to fail? If the cause of the failure is data loss (as reproduced by the iperf3 test), then the SMB file copy operation should fail even when it is the only network communication taking place. However, if I close all other programs communicating over the network, my SMB file copy operations succeed.

[DDiepenbrock]

@ahershler , yes I'd be happy to give this a try.  I may have time tomorrow or this weekend.

It is difficult to know why a webpage refresh would have an impact without having access to the source code.  

[DDiepenbrock]

I haven't had a chance to test this yet, but plan to.  Intel is now telling me my problems are due to the antenna.  Their support people are even less helpful than Dell's. 

[DDiepenbrock]

Unfortunately I haven't yet had time to setup the appropriate environment.  I can say that I've moved the driver setting to 2.4ghz only and the problems have generally gone away, at the significant expense of performance.