On Thu, 29 May 2014, Huang Rui wrote: > On Thu, May 29, 2014 at 10:50:03AM -0400, Alan Stern wrote: > > On Thu, 29 May 2014, Huang Rui wrote: > > > > > Currently, in usb_sg_wait routine there isn't a timeout on waiting for > > > completion signal. When using TEST 5 -> 8 at usbtest module to verify > > > a new device controller with gadget zero, it would be hang on this > > > phase and this routine waits forever. > > > > Why did it hang? That's the real problem; you should fix that instead > > of adding a timeout. > > > > Right, it might be a HW issue and still under analysis. > > > > So it should add a timeout about > > > 10s to export a TIMEOUT error to return. > > > > What makes you think 10 seconds is a good timeout value? > > > > This value is rough estimate, and can be changed. > > > Sorry, this is not acceptable. Any timeout value you specify has to > > work for _all_ places that call this routine. 10 minutes might be > > okay. > > > > But as a developer's view, when I run the tests, it should be better > to report a error even though it fails. But in some specail error > cases, usbtest TEST 5 -> 8 which used usb_sg_wait might wait forever > and the test is unable to stop. I think it is still not expected, > right? So do you have any idea to resolve this issue to make the > usbtest report the error codes and complete normally? You could disconnect the gadget from the host. Then usb_sg_wait() would complete. :-) Seriously, the right way to handle timeouts with usb_sg_wait() is to have a timer routine call usb_sg_cancel() when the timeout expires. You could add such a timer to tests 5-8 in usbtest. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
