usb/156596: [ehci] Extremely high interrupt rate on ehci/uhci IRQ16 80% cpu utilization on CPU0

[Hans Petter Selasky]

The following reply was made to PR usb/156596; it has been noted by GNATS.

From: Hans Petter Selasky <hselasky at c2i.net>
To: Colin Percival <cperciva at freebsd.org>
Cc: "bug-followup at FreeBSD.org" <bug-followup at freebsd.org>
Subject: Re: usb/156596: [ehci] Extremely high interrupt rate on ehci/uhci IRQ16 80% cpu utilization on CPU0
Date: Sat, 9 Jul 2011 11:33:40 +0200

 On Saturday 09 July 2011 02:26:37 Colin Percival wrote:
 > Hi again,
 > 
 > The attached patch seems to fix the problem while not breaking anything on
 > my laptop.  It makes handles the case of EHCI_REMOVE_QH(sqh, last) with
 > sqh == last and sqh->prev == NULL by setting last = NULL -- the idea being
 > that if sqh is the only entry in the queue, we still ought to delete it
 > even though it doesn't have a predecessor.
 > 
 > I have no idea if this is correct, since I don't know the rest of the code
 > in this file.  Maybe we should also delete the first entry in a queue even
 > when it isn't also the last entry; maybe the problem lies somewhere else
 > entirely.  I'm hoping that someone who knows this code will be able to
 > figure out the right fix easily now that I have an apparently-working
 > hack.
 
 Hi,
 
 The last element should always be there. If it is NULL I would like to see a 
 backtrace from that. KASSERT(last != NULL) in both add and remove. If you 
 clear that the USB queue will stop working I believe. Anyway, I think your 
 clue might give some hints what is wrong. It appears that the EHCI is spinning 
 on the IDLE queue somehow producing a lot of IRQ's.
 
 --HPS