> You know, I didn't realize it at first, but that's rather surprising. > The CSW patch should not have made any difference at all to the failure > mode you were seeing. All it did was tell the computer to expect a larger > CSW packet _from_ the drive; it did not affect the data being sent _to_ > the drive. And the failures occurred not during the CSW phase but during > the data phase, when the computer was writing to the drive, not reading > the CSW. > > You might try reverting that patch to see if the failure rate goes back > up. And if it does, is the failure mode different? > I will probably give it a try next days when I have time, but this sounds like it will require a fair amount of tries to get reasonable statistics...
Actually, the usbsnoopy logs can depend on the driver versions. I have usbstor.sys 5.00.2195.4854 and disk.sys 5.00.2195.5419. > I don't recall if you said whether your system has a VIA USB controller. > If it does, that might explain the larger CSW request size, quite > independently of the drive vendor. > Nope, I have Intel 82801 DB host controller. > Come to think of it, there's one thing that might be a little useful. If > you have a high-speed hub and a second USB device, you can try plugging > the hub into your computer and then both the Genesys drive and the other > device into the hub. Start writing files as before. When the drive > fails, it will be interesting to see if the other USB device still > works. If it does, that would be convincing evidence that there's no > problem with Linux's EHCI driver (not that we have any reason now to > believe there is such a problem). > I will be able to run such a test once I borrow necessary hardware from people around. But that seems to be technically doable. Another comment. I've tested the device with my laptop at home, which has USB 1.1 and kernel 2.4.19. I've fixed the "max_sectors" parameter in the scsiglue.c. I don't know if I really had to do it, but I did not have much time for playing around. Anyways, the thing seems to work, but of course it's very-very slow. So I've copied several hundreds megabytes from and to the drive and everything went OK. Of course it's not the final proof, but it seem that the problem is only (or mostly) relevant to the USB 2.0 mode of operation. Well, I've also managed to copy several hundreds of megabytes with USB 2.0 after the last fixes, that's why I cannot tell for sure that it all works 100% reliable with USB 1.1... Regards, Max ------------------------------------------------------- This SF.Net email is sponsored by: IBM Linux Tutorials Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
