Re: 0bda:b812 USB "3.0" WLAN devices only report a maximum of 480 Mbps

[Fabian Melzow <fabian.melzow@gmail.com>]

Am Sun, 5 Nov 2023 09:56:06 -0500
schrieb Alan Stern <stern@rowland.harvard.edu>:

> On Sun, Nov 05, 2023 at 02:30:51PM +0100, Fabian Melzow wrote:
> > Am Sun, 5 Nov 2023 14:11:02 +0100
> > schrieb Greg KH <gregkh@linuxfoundation.org>:
> >   
> > > On Sun, Nov 05, 2023 at 01:49:09PM +0100, Fabian Melzow wrote:  
> > > > Am Sun, 5 Nov 2023 13:20:38 +0100
> > > > schrieb Greg KH <gregkh@linuxfoundation.org>:
> > > >     
> > > > > On Sun, Nov 05, 2023 at 01:02:15PM +0100, Fabian Melzow
> > > > > wrote:    
> > > > > > These devices should report 5 Gbps for USB 3.0 in the Binary
> > > > > > Object Store Descriptor, but don't do so.      
> > > > > 
> > > > > Why do you think these should report 5 Gbps?  If the device
> > > > > descriptor does not show it, there's nothing that we can do
> > > > > about it in the operating system, sorry.    
> > > > 
> > > > Thanks. I don't look at the USB specification for USB >2.0,
> > > > which was long times ago and then only for a user space program
> > > > and thought that it at least theoretically could be possible to
> > > > set the device speed at some kind of initialization.    
> > > 
> > > There might be some vendor-specific way of doing this, and then
> > > the device resets itself, but other than that, no, descriptors
> > > are a static description of the device's capabilities.  
> > 
> > When this static description is also used by the host controller to
> > set the device speed, then these WLAN devices are limited to 480
> > Mbps. If  
> 
> The speed is not set by the host controller; it is set by the device.
> 
> If the device is capable of running at a higher speed than 480 Mbps
> then it should do so.  Have you checked the actual connection speed,
> as reported in the "speed" file in the device's sysfs directory or in
> the output from "lsusb -t"?

$ lsusb -t (only relevant output)
/:  Bus 001.Port 001: Dev 001, Class=root_hub, Driver=xhci_hcd/14p, 480M
    |__ Port 004: Dev 003, If 0, Class=Vendor Specific Class, Driver=rtw_8822bu, 480M
> 
> How do you know that the device's USB hardware supports more than 480 
> Mbps?  And even if the hardware does, maybe the firmware doesn't.
> 
> (Note: Just because the device's WLAN interface can run at 780 Mbps, 
> it does not follow that the USB interface can run that fast.)

These USB WLAN sticks are sold as "USB 3.0", and also have a blue
USB-3.0-A-plug. I don't really know the speed, but 480 Mbps is reported
by "lsusb -t" or in the descriptor. I did not open these devices, but
have assumed that they only use a single chip to handle everything,
WLAN and USB.

> > > > > What exactly is the hardware problem?  What is not working
> > > > > "properly" for them?    
> > > > 
> > > > All 0bda:b812 USB-WLAN-devices seems to only operate at 480
> > > > Mbps, but have a WLAN-chip which supports 780 Mbps at maximum.
> > > >   
> > > 
> > > Are you sure that those devices really can run at 480 Mbps?
> > > Perhaps the chip can support higher, but the firmware on the
> > > device, and the other stuff in the device does not?  
> > 
> > The used WLAN chipset should support more then 480 Mbps in hardware,
> > but I don't known whether the host controller limits the speed
> > based on the wrong descriptor of the device.  
> 
> The speed is not determined by the device's descriptors; it is 
> negotiated at the hardware level between the device and the upstream
> hub when the USB port is reset, which happens shortly after the
> device is plugged in, as part of initialization.
> 
> Another thing to watch out for: Even a 10-Gbps-capable device will be 
> forced to run at a measly 480 Mbps if it is connected via a USB-2
> cable rather than a USB-3 cable.

The above lsusb-output, is from plugging the stick direct into a
10 Gbps-USB 3.1-A-port, with it's A-plug in the case.