On Sun, Nov 29, 2009 at 9:36 AM, Matthew Dharm <mdharm-usb@xxxxxxxxxxxxxxxxxx> wrote: > On Sun, Nov 29, 2009 at 09:20:13AM -0800, Jason Byrne wrote: >> On Sun, Nov 29, 2009 at 8:28 AM, Matthew Dharm >> <mdharm-usb@xxxxxxxxxxxxxxxxxx> wrote: >> > On Sun, Nov 29, 2009 at 10:45:16AM -0500, Alan Stern wrote: >> >> On Sun, 29 Nov 2009, Jason Byrne wrote: >> >> >> >> > Hi, >> >> > >> >> > I've tried several stable kernel versions with an isd200 usb2 cradle, >> >> > and it seems that the last working kernel version with this device may >> >> > be 2.6.26.8. >> >> > >> >> > I'm not sure if this was related to the separation of the isd200 >> >> > driver into a separate module, or if it's an ehci_hcd issue, or what >> >> > the cause is... >> >> >> >> You should collect a pair of usbmon traces, one from a working kernel >> >> and one from a recent non-working kernel. See >> >> Documentation/usb/usbmon.txt for instructions. >> > >> > It looks like the working version doesn't use high-speed, only full speed. >> > Try rmmod'ing ehci_hcd on the newer kernel and then attaching the device. >> >> It does work if I rmmod ehci_hcd and attach the device, but using usb >> 1.1 for this device would be a huge step backward. Historically, there >> was a usb 1.1 version of this particular cradle that may also use an >> isd driver, but I've used this particular cradle for years with usb >> 2.0/high speed. > > It may be a huge step backwards, but it is also a significant datapoint for > debugging. > > Specifically, the usb-storage driver does not have any codepaths which > differentiate between full-speed and high-speed devices. This makes it > likely that this is an EHCI-related issue. Attached is a usbmon dump of the bus with the attached device, with kernel 2.6.31.6, and ehci_hcd module loaded... I've also noticed that the kernel will attach this device successfully ~ 1/5 times, so it doesn't fail 100% of the time with ehci_hcd active.
Attachment:
1.mon.out
Description: Binary data