Re: [PATCH v3 00/11] RTW88: Add support for USB variants

[Sascha Hauer <s.hauer@pengutronix.de>]

On Thu, Nov 24, 2022 at 09:21:58AM +0100, Sascha Hauer wrote:
> On Thu, Nov 24, 2022 at 06:48:23AM +0000, Ping-Ke Shih wrote:
> > 
> > > -----Original Message-----
> > > From: Sascha Hauer <s.hauer@pengutronix.de>
> > > Sent: Tuesday, November 22, 2022 10:55 PM
> > > To: Bernie Huang <phhuang@realtek.com>
> > > Cc: linux-wireless@vger.kernel.org; Ping-Ke Shih <pkshih@realtek.com>; Hans Ulli Kroll
> > > <linux@ulli-kroll.de>; Martin Blumenstingl <martin.blumenstingl@googlemail.com>; netdev@vger.kernel.org;
> > > Kalle Valo <kvalo@kernel.org>; Yan-Hsuan Chuang <tony0620emma@gmail.com>; linux-kernel@vger.kernel.org;
> > > Viktor Petrenko <g0000ga@gmail.com>; Neo Jou <neojou@gmail.com>; Bernie Huang <phhuang@realtek.com>;
> > > kernel@pengutronix.de; Johannes Berg <johannes@sipsolutions.net>; Alexander Hochbaum <alex@appudo.com>;
> > > Da Xue <da@libre.computer>
> > > Subject: Re: [PATCH v3 00/11] RTW88: Add support for USB variants
> > > 
> > > On Tue, Nov 22, 2022 at 03:52:15PM +0100, Sascha Hauer wrote:
> > > > This is the third round of adding support for the USB variants to the
> > > > RTW88 driver. There are a few changes to the last version which make it
> > > > worth looking at this version.
> > > >
> > > > First of all RTL8723du and RTL8821cu are tested working now. The issue
> > > > here was that the txdesc checksum calculation was wrong. I found the
> > > > correct calculation in various downstream drivers found on github.
> > > >
> > > > The second big issue was that TX packet aggregation was wrong. When
> > > > aggregating packets each packet start has to be aligned to eight bytes.
> > > > The necessary alignment was added to the total URB length before
> > > > checking if there is another packet to aggregate, so the URB length
> > > > included that padding after the last packet, which is wrong.  Fixing
> > > > this makes the driver work much more reliably.
> > > >
> > > > I added all people to Cc: who showed interest in this driver and I want
> > > > to welcome you for testing and reviewing.
> > > 
> > > There still is a problem with the RTL8822cu chipset I have here.  When
> > > using NetworkManager I immediately lose the connection to the AP after
> > > it has been connected:
> > > 
> > > [  376.213846] wlan0: authenticate with 76:83:c2:ce:81:b1
> > > [  380.085463] wlan0: send auth to 76:83:c2:ce:81:b1 (try 1/3)
> > > [  380.091446] wlan0: authenticated
> > > [  380.108864] wlan0: associate with 76:83:c2:ce:81:b1 (try 1/3)
> > > [  380.136448] wlan0: RX AssocResp from 76:83:c2:ce:81:b1 (capab=0x1411 status=0 aid=2)
> > > [  380.202955] wlan0: associated
> > > [  380.268140] IPv6: ADDRCONF(NETDEV_CHANGE): wlan0: link becomes ready
> > > [  380.275328] wlan0: Connection to AP 76:83:c2:ce:81:b1 lost
> > > 
> > > That doesn't happen when using plain wpa_supplicant. This seems to go
> > > down to cd96e22bc1da ("rtw88: add beacon filter support"). After being
> > > connected I get a BCN_FILTER_CONNECTION_LOSS beacon. Plain
> > > wpa_supplicant seems to go another code patch and doesn't activate
> > > connection quality monitoring.
> > > 
> > > The connection to the AP works fluently also with NetworkManager though
> > > when I just ignore the BCN_FILTER_CONNECTION_LOSS beacon.
> > > 
> > > Any idea what may be wrong here?
> > > 
> > 
> > Please reference to below patch to see if it can work to you.
> > 
> > https://lore.kernel.org/linux-wireless/20221124064442.28042-1-pkshih@realtek.com/T/#u
> 
> Great! That solves this issue \o/

I am also happy to say that this patch is a real performance boost.
Here are some numbers:

RTL8723DU, no patch
===================
tx [  5]   0.00-10.00  sec  1.08 MBytes   903 Kbits/sec    1 sender
rx [  5]   0.00-10.26  sec  10.3 MBytes  8.41 Mbits/sec    0 sender

RTL8723DU, with patch
=====================
tx [  5]   0.00-10.00  sec  34.7 MBytes  29.1 Mbits/sec    9 sender
rx [  5]   0.00-10.05  sec  34.4 MBytes  28.7 Mbits/sec    5 sender

RTL8821CU, no patch
===================
tx [  5]   0.00-10.00  sec  6.01 MBytes  5.04 Mbits/sec    0 sender
rx [  5]   0.00-10.44  sec  53.4 MBytes  42.9 Mbits/sec    0 sender

RTL8821CU, with patch
===================
tx [  5]   0.00-10.00  sec  57.9 MBytes  48.6 Mbits/sec    0 sender
rx [  5]   0.00-10.15  sec  66.3 MBytes  54.8 Mbits/sec  419 sender

RTL8822CU, no patch
===================
tx [  5]   0.00-10.00  sec  11.8 MBytes  9.91 Mbits/sec    1 sender
rx [  5]   0.00-10.18  sec   105 MBytes  86.3 Mbits/sec  805 sender

RTL8822CU, with patch
===================
tx [  5]   0.00-10.01  sec  80.0 MBytes  67.1 Mbits/sec    0 sender
rx [  5]   0.00-10.12  sec   117 MBytes  97.2 Mbits/sec  658 sender

Tested with:

tx: iperf3 -c 192.168.0.57 -i0
rx: iperf3 -c 192.168.0.57 -i0 -R

Especially the RTL8723DU now starts being useful.

Sascha

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |