intel usb 3.0 host controller for skylake cpu

I also meet this problem using Braswell Platform,

We will be reporting the problem to Intel. The problem seems more frequent when the host controller is under high load. There is also some evidence that the problem might also be linked to load. It wouldn't be too surprising if the problem also effected other devices but we haven't done extensive testing on 3rd party devices. In summary there seems to be a compatibility bug between some Intel system on a chip (SoC) USB3 host controllers and at least the Cypress FX3 device chip. This unexpected behaviour confuses the USB devices, it could be argued that that our plug should eventually recover from this bad behaviour, but the problem is really in the underlying silicon chip (a Cypress FX3 in this case). When the host controller does start transmitting again the interval count skips forward (e.g to a value of 902 instead of 858. In our understanding each bus interval boundary is 125us, but the Intel host controller failed to send anything for 5.512ms.ĥms is an eternity when you are expecting a transmission every 125us. The ITP packet should be send from the host each bus interval boundary to all links that aren't in a low power state. The ITP contains a timestamp value based on the Bus Interval Counter and it's value varies from 0 to 0x3FFF before it rolls over, back to zero. The Intel USB3 host controller is failing to send Isochronous Timestamp Packets (ITP) on a regular basis. National Instruments did a protocol trace for us on a similar Intel board and it looks like it is a combination of two bugs.

Intel usb 3.0 host controller for skylake cpu