USB regression issue with amd64 -current

[Scott Ellis <scotte%intrepid.warped.com@localhost>]

I updated my kernel and userland from NetBSD/amd64-current of ~April 
20th to May 14th, and find that USB (on ehci or uhci it seems) causes 
the kernel to panic.  I haven't witnessed it first hand, nor been able 
to reproduce it on demand, but with my daily backup scripts (which rsync 
to a USB HDD), I have seen 3 times:

May 15 15:19:47 intrepid /netbsd: fatal page fault in supervisor mode
May 15 15:19:47 intrepid /netbsd: trap type 6 code 10 rip 0 cs 8 rflags 
10202 cr2  0 cpl 2 rsp ffff8000534b1cc8
May 15 15:19:47 intrepid /netbsd: panic: trap
May 15 15:19:47 intrepid /netbsd: sd0(umass0:0:0:0): generic HBA error
May 15 15:19:47 intrepid /netbsd: fatal page fault in supervisor mode
May 15 15:19:47 intrepid /netbsd: trap type 6 code 10 rip 0 cs 8 rflags 
10246 cr2  0 cpl 4 rsp ffff8000534aac88
May 15 15:19:47 intrepid /netbsd: Skipping crash dump on recursive panic
May 15 15:19:47 intrepid /netbsd: panic: trap


I thought I saw a commit message fly by on source-changes relating to 
USB transfers (something about crossing word boundaries?), but for the 
life of me can't find it in the mail logs or the CVS repository.

Can somebody clue me in on recent USB commits that I can try and revert 
to try and track down the offender?  Or have another clue on what to try?

Since I can't make the panics happen (on demand) (I tried running a 
backup manually, and it worked fine!), I don't want to let the kernel 
stop in DDB or similar, and sit there for hours when I'm not around. :-)

        ScottE