Greetings, I've been testing David Brownwell's latest EHCI patch (ehci-0216) along with Tom Collin's usb-storage patch (p1.txt) and have experienced a breakthrough in both performance and stability on Kernel 2.4.21-pre4-ac4 with the VT8235.
Instead of getting around 4Mb/sec, I now get around 20Mb/sec (according to hdparm -tT) and have been able to copy several gigabytes of data to the USB 2.0 HDD without it hanging the system. No longer is there there the situation where there is some data transfer, a delay, then a hang - although the system will hang under particular circumstances. One thing I have noticed is that I can pretty much write as much data to the device as needed without hanging the system, however if I transfer files from the USB 2.0 HDD to another partition (e.g. on an IDE HDD), it's likely to hang - on the VT8235/KT266A test computer I have, this occured on both the onboard VIA USB, and the NEC USB 2.0 card, so there are still issues there. It's unbelievable how close this is to working properly, this is perhaps the most effective update yet. If the same stability could be experienced with reading data as with writing data to the device, the VIA EHCI implementation (and NEC cards attached to a VT8235) would be highly usable. I'm not sure if anyone else had problems with applying Tom Collin's usb-storage patch to 2.4.21-pre4-ac4, but attached is a copy of the patch I used (there were a few rejected lines that needed to be re-applied). Thanks, Jonathan
usbstor-p1-2421pre4ac4.patch
Description: Binary data
