Vendor specific device class about usbip-win

[cezuni]

@wrenchonline : (1) Yeah, 255 class had not been treated.. But this case can be easily improved. (2) You mentioned IRP_MN_REMOVE_DEVICE happens. Honestly, I have no idea for that case. More thoughts needed?

[cezuni]

@wrenchonline : (1) Yeah, 255 class had not been treated.. But this case can be easily improved. (2) You mentioned IRP_MN_REMOVE_DEVICE happens. Honestly, I have no idea for that case. More thoughts needed?

After more thought, IRP_MN_REMOVE_DEVICE may be issued by an upper USB class driver. This problem was addressed at #6 (comment) .

[cezuni]

@wrenchonline : In my opinion, IOCTL_INTERNAL_USB_GET_TOPOLOGY_ADDRESS is not related to device removal. As noted before, I suspect that USB command timeout (500msec) makes PNP manager to issue IRP_MN_REMOVE_DEVICE. By the way, do your logs have any cancellation of an issued IRP ? The reason why i ask is that IRP cancellation is preceded before removing a device.

[cezuni]

@wrenchonline : If you comment out following lines, vhci cannot handle a real faulty error.

https://github.com/cezuni/usbip-win/blob/27ec58531963eeca4d679a7ad70e705f00bdb4e4/driver/vhci/vhci_write.c#L239-L242

It seems the return STATUS_UNSUCCESSFUL;Is the upper driver issued IRP_MN_REMOVE_DEVICE key!

I finally have realized why IRP cancellation was not found when your device was removed. In your case, IRP actually failed!! Yeah, usbip-win logging is too bad to miss an important event.
I think that dropping STATUS_UNSUCCESSFUL cannot be a desirable solution. How about trying to find the reason why usbip command fails at linux box. Linux kernel logs may provide a useful information.

[cezuni]

@wrenchonline : Your hub on linux doesn't seem to work correctly. "device descriptor read/64, error -110" is generated by drivers/usb/core/hub.c in linux kernel. -110 error is ETIMEDOUT.
I recommend you to check a physical or virtual hub.

[cezuni]

@wrenchonline : I'm happy to see your smile. Maybe, returning STATUS_SUCCESS is more reasonable. I 'll check it out more.

[cezuni]

@wrenchonline : Is your vendor specific class device working correctly if a recent vhci driver is applied ?