On Sat, 24 Jan 2009 17:38:48 +0000, Nicholas Blachford wrote > Before copying the files to the HD I first copy them onto another > BFS formatted USB drive. Unfortunately this is a bit of a slow > process as it's seemingly the slowest USB drive on the planet. If > it boots OK I copy it to the HD, another 15 minute wait. Yeah, many flash drives are slow to write. It's a question of the cell quality and type (single/multi level). You can usually get better performance with SD cards and then getting those extreme models. I have a stick here as well that is pretty fast for reading and takes ages for writing anything. > After getting ever more annoyed at the slowness of this process I > upgraded to a bigger and much faster USB drive - A Sandisk cruzer > 8GB to be exact. Unfortunately Haiku doesn't like it... What you describe is interesting, but putting it here is not really the right place. Please open a bug report at dev.haiku-os.org. Also please state what revision you're using. I've fixed a pretty severe issue in this context last weekend (r28934 and up). > You can't immediately mount the drive, you have to eject and > reinsert it. That's because the partition info isn't updated. A "rescan" might trigger it, I'm not sure that's implemented though. > It most likely wont mount and if you want to try again you have to > pull out and reinsert. Yup, because if errors occure it'll just report a media removal and you need to replug to get it back. > The odd thing is the exact same drive when formatted to DOS works > just fine. That's indeed interesting. Though the access pattern of BFS and FAT filesystems are of course pretty different. It's easily possible that BFS just does enough actions to trigger it. > The driver appears to be leaving the drive in a strange state - > there is a light on the drive which flashes if it's in use and > pulses slowly if it's not. I've seen the drive left with the light > left on or off. It probably just hangs. As mentioned, the severe bug did make the driver unable to recover the device from an error condition. Some devices apparently worked fine with a half-working reset, but most had issues (which is fine). Please check with a current revision and see if the trouble persists. If it does, please open a bug report mentioning the exact messages you see in the KDL. Probably it's something along the lines of "could not write back block x". Regards Michael