(For the onlookers ... try this patch even if it doesn't solve Rene's problem. I know other folk have had unlinking issues, and it's more probable that this patch will improve those behaviors ...)
On Thursday 12 January 2006 1:38 pm, Rene Herman wrote: > Very noisy obviously, but once things settle and the > external drive is just switched off, I do _not_ get any printk's, while > I still experience the IDE throughput drop. I guess this firmly points > to chip? So long as neither of the two schedules (async, periodic) is enabled, the EHCI controller has no need to perform DMA accesses at all. And excess DMA traffic is the only obvious thing that could account for that sort of performance drop ... but previously you had established that the schedules were empty and the schedules were (mostly) off. So it sure acts like the chip is doing something odd, yes. Less obvious things might include PCI arbitration issues, like BIOS kicking in a mode where the bus idles on EHCI rather than for example the CPU or the last bus master. Seriously ... anyone with a PCI analyser and some time could do a Good Deed and help figure out what's going on here. I know there are lots of Linux fans in hardware groups, and I'd hope one of them has a VIA motherboard they could look at with it for a while. I refuse to believe nobody except Windows people have PCI analysers! > On the other hand, I can switch on the external drive without any bad > effects on IDE throughput as long a ehci-hcd is not loaded. Only once > ehci-hcd loads does trouble begin, and only after unloading it does it > end. This would seem to point to ehci-hcd again... It would seem to point to some side effect of the controller being active, which isn't quite the same thing. A controller that's active but hasn't been given any work (as you observed) shouldn't affect the behavior of any other PCI devices (but in your case it seems to). > Some weird initialization value somewhere maybe? Do you have a contact > at VIA who could maybe tell you about an undocumented bit somewhere you > could flip? Would there be any sort of effect like "being faster to the > bus" with the chip doing that async toggle? The chip's fast in benchmarks... VIA has been particularly useless at providing information about their chips. Which would be no problem if they only worked as expected, and the BIOS writers did sane system init. :( - Dave ------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK! http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click _______________________________________________ [email protected] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
