On Tuesday 13 of September 2005 22:53, matthieu castet wrote: > > Any comments on this ??? > > > I was waiting usbatm maitainer comment (did you send it on usb atm ML ?). > But he doen't reply for other patches... I know that ... and because patch affect only eagle modems I send it only here, after review I planed send it upstream.
> Could you explain a bit more what it does ? > It is for the firmware bug where atm cell are incomplete ? > > Does it solve some errors in bulk mode for high speed rate ? Yes, chunk of incoming data from bulk urb is not multiple of atm cell length (stride in usbatm name convention) and one cell is coming partially (remain of cell commes in next urb). Patch simply merge broken atm cell in one place. In final effect errors from /proc/net/atm/ueagle* disappear. > > BTW, I think this line should have form > > int timeout = (data <= 0x20ffff) ? 1 : 0; > > but of course check is needed. > Why do you think that ? I have connection speed less then 1Mbits/s and have better results with timeout==1 > > How anyone know if ADI930 firmware have similar bug? > > If yes that would be good reason to merge patch to usbatm module, > > because ADI930 use only bulk mode. > > > I don't know, but it don't support UEA_SEND_TIMEOUT... Incomplete atm cell bug may be independent from timeout command (?). __ Staszek Gruszka
