On Sat, 3 Nov 2012, Miguel Dardenne wrote: > On 2012-11-03 17:50, Alan Stern wrote: > >> I get a USB disconnect of my external USB HDD about once per day. > >> Enabling USB debug in the kernel shows the first error as 'ehci_hcd > >> detected XactErr'. It happens only when the drive is idle. For > >> example, > >> I had a process writing on the HDD every 3 seconds and the problem > >> did > >> not occur for several weeks. > > > > Correction: These errors do not occur when the drive is idle. You > > can > > see in the log you attached: > > OK thanks for this. I then reformulate my statement: using the drive on > a frequent basis (3s period) prevents the problem from occuring. This suggests a solution: Write a little program that will run in the background and do some trivial disk access every 3 seconds. > > No. The drive remains at full power as long as it is mounted. If > > the > > power level were changed, it would show up in your debugging log. > > The current test is that I have been using it in OHCI-only mode for one > full day (by unloading the ehci kernel module) and result so far is that > I did not encounter the issue. Interestingly here is a quote from the > support team at pcengines.ch : "there are some known issues regarding > the Hi-speed USB modes. The system then reports dropping the device, EMI > troubles, and so on. > There are no other workarounds." That does sound exactly like what you experienced. > An interesting thread on this subject and on this very mailing list is > http://thread.gmane.org/gmane.linux.usb.general/36246 "Geode LX CS5536 > EHCI problems". It is 2 years old but is about disconnects also. Does > anybody know if things have improved since then ? > > > It appears that the drive disconnected itself from the USB bus. > > There's no indication of why this happened. > > I conclude from all these that : > - we face some hardware issue probably related to the AMD CS5536 chip > (or less probably its use on the Alix board) Yes. > - playing with power management won't help Probably not. You can try it, but I would be surprised if it made any difference. > - problem is probably not new I don't know. The errors described in that other thread were different from the error you saw. But maybe they are related somehow. > - problem is most probably not solved at this point > > Do I have to mourn the idea of using the USB 2.0 capabilities of the > board, given that the usage of this particular chip is probably too low > / fixing is probably too hard to trigger enough interest in solving the > issue (and given that I probably cannot fix it all on my own) ? In fact, fixing the problem is probably impossible. But if accessing the disk every 3 seconds will work around the problem then you should be in relatively good shape. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
