search

daynix / UsbDk

Usb Drivers Development Kit for Windows
Apache License 2.0
553 stars 142 forks source link

Problem with AbortPipe #111

Open AiCabal opened 2 years ago

AiCabal commented 2 years ago

I want to cancel transfer and tried to do it with two way: CancelIoEx (like libusb) and AbortPipe. AbortPipe doesn't work at all while CancelIoEx works. AbortPipe return 0 value and all bytes are transfered. But also for some users doesn't work CancelIoEx it either transmits all bytes and return 1 or doesn't transmit at all and anyway return 1.

ybendito commented 1 year ago

@AiCabal Which kind of transfer you try to cancel by aborting the pipe?

P5-2005 commented 1 year ago

@AiCabal Which kind of transfer you try to cancel by aborting the pipe?

seem like this : https://github.com/mineek/openra1n/blob/4595a5333e4134ade77b43fb2259e880b85801ee/openra1n.c#L785C50-L785C50

ybendito commented 1 year ago

@P5-2005 Disclaimer: If I'm not mistaken: Aborting of the control pipe does not do anything and this does not depend on UsbDk. Control transaction has its own timeout of 5 seconds. You can cancel the request, but the transaction will be finished or stall condition will be automatically cleared. Because this is control pipe. On the rest of pipes the stall need to be cleared manually, for that there is a pipe abort. Details should be checked in the USB spec.

P5-2005 commented 1 year ago

the issue i get on my side, its usbdk doesnt do redirection on usb 2.0 with two controller, i dont have now the debug for such case because i have now only pc with usb 3.0 and one controller host so its fine wonder if there is workaround to choose or precise which controller must forward into usb2.0.txt

P5-2005 commented 1 year ago

@P5-2005 Disclaimer: If I'm not mistaken: Aborting of the control pipe does not do anything and this does not depend on UsbDk. Control transaction has its own timeout of 5 seconds. You can cancel the request, but the transaction will be finished or stall condition will be automatically cleared. Because this is control pipe. On the rest of pipes the stall need to be cleared manually, for that there is a pipe abort. Details should be checked in the USB spec.

also that timeout was add if not using usbdk, but usbdk as backend is better than this timeout implemented with libusb(seem not enough)

P5-2005 commented 1 year ago

i found an old trace debug of pc with two usb host controller and usb2.0 only : usbdkmanyroothub2.0.txt