Re: IRQ handling for ehci and/or uhci causes extremely slow firmware transfer over USB

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, 22 Jul 2009, Michael Conrad wrote:

> Hi, I've filed a bug at http://bugzilla.kernel.org/show_bug.cgi?id=13779
> and was asked to bring it here for discussion.
> 
> The short of it is that I have a board that ends up with ehci and uhci on  
> the same interrupt, and in kernel 2.6.27 something changed that causes  
> these two to conflict in some way that causes a firmware upload to take  
> ridiculously long to complete on a USB 2.0 device.
> 
> Thanks for any/all help.
> 
> 
> Here's the text of the bug, for convenience:
> 
> The 950Q TV tuner from Hauppauge uploads its firmware 64 bytes at a time  
> over
> USB, when you tune a channel.  In linux stock kernel 2.6.26 this works  
> fine,
> but in 2.6.27 and later (tested up to .30) the firmware transfer takes 15
> minutes, rendering the device unusable.  This bug is motherboard  
> dependent, and
> working with Mike Krufky at Hauppauge (a dvb/v4l contributor) we have
> determined that it is a resource conflict between ehci and uhci drivers.
> 
> Our main diagnosis is that (using 2.6.30) by disabling the UHCI driver, the
> firmware upload no longer exhibits the problem.  Our main clue was that on  
> our
> hardware, ehci and one instance of uhci were sharing an interrupt.  They  
> were
> also sharing an interrupt in 2.6.26, but did not have the slow transfer  
> problem
> in that version of the kernel.  The problem showed up in 2.6.27 and later.
> 
> Our platform is an embedded Intel Atom board.  Other systems with 2.6.30  
> and
> uchi enabled do not have this problem, however we have no other systems  
> where
> ehci and uhci share an interrupt.
> 
> We are willing to help debug and test.

Let's concentrate on 2.6.30.  First, build the kernel with 
CONFIG_USB_DEBUG enabled, just for the sake of thoroughness.

Next, acquire a couple of usbmon traces showing what happens during the 
firmware upload both with and without uhci-hcd loaded.  (See 
Documentation/usb/usbmon.txt for instructions.)  Attach those traces to 
the bug report and let us know when they are ready.

Also attach the dmesg log; it might contain some useful debugging 
information.

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

[Index of Archives]     [Linux Media]     [Linux Input]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [Old Linux USB Devel Archive]

  Powered by Linux