Greg KH <greg@...> writes: > > On Mon, Jan 05, 2009 at 11:21:57AM +0000, Martin Willms wrote: > > > > Just an update on our problem. > > > > 1.) Thanks for your answers. They helped me a lot. > > > > 2.) We now changed the interface between the backplane hardware (EZ USB > > Controller) and Linux PC in a way, that each time no data is available on the > > backplane anymore, the backplane sends zero length packets. This way the > > usb_bulk_msg function never gets into timeout. I tested this a lot and I am > > quiet sure, that our problem is solved now and no data gets lost anymore. > > > > We implemented it this way, because I got a hint from an Windows USB developer, > > that aborting an USB transfer due to an timeout is not a good idea, because the > > host controller operates on DMA accessible memory. He said, that this is the > > standard way USB transfers are cancelled in Windows. It works fine with Linux, > > too. > > > > This solution is OK for us. I realy did not find any problem in the way we > > implemented the interface before. The only difference is, that the usb_bulk_msg > > function does not get into a timeout anymore. As most of you said, this should > > not be a problem, but in my application it is. > > > > Can it be that there still is a problem in the Linux USB implementation when > > aborting transfers due to a timeout of the usb_bulk_msg function? > > Not that I know of, but it sounds like your application was having > problems with it, not the kernel, right? > > thanks, > > greg k-h > -- > To unsubscribe from this list: send the line "unsubscribe linux-usb" in > the body of a message to majordomo@... > More majordomo info at http://vger.kernel.org/majordomo-info.html > > I am not sure. For me it looks, as if the application driver is OK. I asked Mister Stern about my handling of the usb_bulk_msg (...) function and he said it is OK. I do not know where data gets lost. All I know is, that the correct amount of data is send by my measurement hardware (Cypress EZ USB) and that I evaluate all Bytes, that are received by the usb_bulk_msg (...) functions in the application level driver. So I know that the usb_bulk_msg (...) function allready misses some data. With the bugfix of sending empty USB packets to cancel a transmission no data gets lost anymore. On the Linux PC side the only difference I can see is, that now the usb_bulk_msg (...) function does not get into timeout anymore (no return value ETIMEOUT / -110). The application is working correctly now so I will leave it this way. Nevertheless I do not undertstand why it sporadically lost data before, when the usb_bulk_msg (...) function got into timeouts. thanks, Martin -- 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
