On Thu, 30 Dec 2021 16:48:09 +0100 Greg Kroah-Hartman wrote:

On Thu, Dec 30, 2021 at 11:36:12PM +0800, Yaqin Pan wrote:

quoted

On Thu, 30 Dec 2021 15:12:27 +0100 Greg Kroah-Hartman wrote:

quoted

quoted

This quirk is only for dwc3 host mode.
the dwc3 controller can't emurate some devices successfully.
For example, TF card reader (aaaa:8816):
failed log
usb 1-1: new high-speed USB device number 2 using xhci-hcd
usb 1-1: device descriptor read/all, error -110

quoted

From the usb analyzer, always return NAK in the data phase.

if enable the GUCTL.SPRSCTRLTRANSEN bit. then the log is:
usb 2-1: new high-speed USB device number 3 using xhci-hcd
usb 2-1: New USB device found, idVendor=aaaa,
idProduct=8816, bcdDevice=13.08
usb 2-1: New USB device strings: Mfr=1, Product=2, SerialNumber=3
usb 2-1: Product: MXT USB Device
usb 2-1: Manufacturer: MXTronics
usb 2-1: SerialNumber: 150101v01
usb 2-1: New USB device found, VID=aaaa, PID=8816

Some devices are slow in responding to Control transfers.
Scheduling mulitiple transactions in one microframe/frame
can cause the devices to misbehave. if this qurik is enabled,
the host controller schedules transations for a Control transfer
in defferent microframes/frame.

If this is needed for all devices (i.e. you do not know what device is
going to be plugged in), why not just enable it for all controllers?
Why whould you NOT want this enabled?

Or is this a broken hardware device and only specific host controllers
need this?  If so, how do we know which ones need this set and which do
not?

I think not all dwc3 controllers need this. For cell phone,customers may
use various usb devices, we can enable this quirk to fix some compatibility
issues. For some chip platform of qcom, i encounter this issue, not every
platform i encounter this problem.

If enabled for all controllers, it will reduce the speed of Control transfers. 
So i think it would be better for user to enable it by their own purposes.

But how do hardware vendors know to enable this?  Can we trigger off of
PCI ids?  Do we need a list of quirks to show which host controllers are
broken this way?

Burying something as basic as "reliable device connection" in a DT quirk
seems very sloppy to me.  We want reliable systems, right?

Yes, we want reliable systems. But i don't have a good ideal about this issue.
when we meet this problem, and from the dwc-usb3 controller datasheet,we know
enable one bit in dwc-usb3 controller's register can fixed this issue.

Of course, i can list the host controllers that i used broken this way if needed.

thanks,

Yaqin pan