search

morrownr / 8814au

Linux Driver for USB WiFi Adapters that are based on the RTL8814AU Chipset
Other
531 stars 102 forks source link

(info)[stability] TRx configuration differs depending on USB2/3 #18

Open abroenni opened 3 years ago

abroenni commented 3 years ago

Hi so I've done some more digging into the code base and I have merged some commits into my DEV Branch. The code is ugly.

I first cherrypicked some features I saw on the aircrack-ng version of this driver as I thought more features is better. But while I was going through the debug logs, I noticed that the driver presents the physical device differently to the system depending if it is plugged in into an usb2 port or an usb3 port (This requires properly telling the driver to make use of USB3 with the module param).

This could only be very specific to my device. But I noticed that when in USB3 mode the driver configures the device as a 3T3R with 3 streams. When plugged into a USB2 mode it configures the device as a 2T4R device. This can be confirmed by loocking at the the system journal when the driver is loading, or by doing a cat hal_info under /proc/net/rtl8814au/{devicename}/. I removed the decision block where it checks for USB2/USB3 in the code and now the driver configures the device as a 3T3R regardless of USB2 or 3.

This then subsequently also activated 3 spatial streams in HT and VHT mode, which can be confirmed with iw phyX info This now allows to specify ht_capab=[..][RX-STBC3][..] and vht_capab=[..][RX-STBC-3][..]. I have uploaded some hostapd configs in my branch as well.

I build the driver on raspbian 32bit os and configured it for AC mode with 80Mhz bandwidth. hostpad would come up, but the moment I would use an AC device that can use 80Mhz, the driver driver would be killed by the system and reloaded. the journal would say that there is an over-current on the USB port. But it would not make the entire system hang. My setup is a rpi 3b+ using the built in wifi in station mode to connect to my main wifi on 5Ghz.

I remembered from an earlier project, that the raspbian os has some stability issues when used in 5ghz mode and I had found that arch arm 64 bit is much better. So I setup arch arm 64bit for this as well and installed the driver (no problems at all). Once the same hostapd conf for AC was set up, I ran the test. And this time it would not crash the kernel module and hostapd stayed up.

I even went as far as setting up a second usb wifi device with an older ralink chip(driver rt2800usb) on the same rpi3b+. Now it is hosting 2 wifis: one for 2.4G (ralink) and one for 5G(8814a). Its been up for 24hours so far and clients can connect and disconnect no problem. But like someone else said about this driver, the crashes happen randomly, so I am only testing and sharing it here with no PR.

I might snap some iperf3 stats for the different configurations.

cheers

PS: don't activate Beamforming for AP mode in hostapd nor module config. It really cripples the throughput from what I have found.

morrownr commented 3 years ago

Hi abroenni,

Quote: "The code is ugly."

Reply: That is a serious understatement. I've used meld a lot to compare code between the various drivers I have up here and I don't know how the Realtek dev keeps this stuff sorted out. This 8814au driver shows signs of a lack of love more so than the other drivers. In fact, this driver is getting hard to maintain. I spend most of my time in the code working on updates related to kernel changes and this driver is a real challenge to keep going. We need a cleaned up, modernized version.

I have measured the current on my adapter based on a 8814au chipset. When I put a good load on with iperf3, it will pull nearly 900 mA. The RasPi 3b, 3b+ and 4b allocate only 1200 mA for the entire usb subsystem and I don't think it handles spikes very well. We could use a powered hub except for the backfeeding issue with RasPis. I have seriously tried to use my 8814au adapter with a RasPi4b but have not been able to find a way to have stable operation in AP mode.

The most stable hostapd settings I have found are:

ht_capab=[HT40+][HT40-][SHORT-GI-20][SHORT-GI-40][MAX-AMSDU-7935]
vht_capab=[MAX-MPDU-11454][SHORT-GI-80][HTC-VHT]
# Note: [TX-STBC-2BY1] causes problems

I was checking out your dev branch. Let me know if you want me to test it and what you want it tested with.

If you can figure out how to stabilize the AP mode so that it is usable, you would be a hero to a lot of folks with this adapter.

morrownr commented 3 years ago

@abroenni

Have you looked at this anymore?

abroenni commented 3 years ago

Since the AP fix was applied, the 3R3T:3 streams change doesn't cause any problems for me (the change is in my dev branch). I got a bit confused about the differences and similarities between spatial streams involving antenna count for TX-RX and STBC stream counts.

I literally just figured out my mistake with the 3 streams. It wasn't in the code, but in the hostapd config. Currently, using my dev channel, I get the follwoing iw and I can see 

Band 1:
        Capabilities: 0x1bf3
            RX LDPC
            HT20/HT40
            Static SM Power Save
            RX Greenfield
            RX HT20 SGI
            RX HT40 SGI
            TX STBC
            RX STBC 3-streams
            Max AMSDU length: 7935 bytes
            DSSS/CCK HT40
...
    Band 2:
        Capabilities: 0x1bf3
            RX LDPC
            HT20/HT40
            Static SM Power Save
            RX Greenfield
            RX HT20 SGI
            RX HT40 SGI
            TX STBC
            RX STBC 3-streams
            Max AMSDU length: 7935 bytes
            DSSS/CCK HT40
        Maximum RX AMPDU length 65535 bytes (exponent: 0x003)
        Minimum RX AMPDU time spacing: 16 usec (0x07)
        HT Max RX data rate: 450 Mbps
        HT TX/RX MCS rate indexes supported: 0-23
        VHT Capabilities (0x03c051b2):
            Max MPDU length: 11454
            Supported Channel Width: neither 160 nor 80+80
            RX LDPC
            short GI (80 MHz)
            TX STBC
            SU Beamformee
            +HTC-VHT
        VHT RX MCS set:
            1 streams: MCS 0-9
            2 streams: MCS 0-9
            3 streams: MCS 0-9
...

Setting up an for a 2Ghz AP using this device, and then using another computer to run iw wlanX scan | less and looking for the access point I see the following :

   HT capabilities:
                Capabilities: 0x1bbd
                        RX LDPC
                        HT20
                        SM Power Save disabled
                        RX Greenfield
                        RX HT20 SGI
                        TX STBC
                        RX STBC 3-streams
                        Max AMSDU length: 7935 bytes
                        DSSS/CCK HT40
                Maximum RX AMPDU length 65535 bytes (exponent: 0x003)
                Minimum RX AMPDU time spacing: 2 usec (0x04)
                HT Max RX data rate: 450 Mbps
                HT TX/RX MCS rate indexes supported: 0-23

I have to test where I can see the 3 RX stbc streams in wireshark or so and check if they make a difference. That is not to say, that having all these capabilities on in hostpad conf is neccessarily a good idead, but more that the function can be turned on if so desired.

"[TX-STBC-2BY1] causes problems" : Agreed. The connection speed drops down to around 2Mbit. Should really not be used.