CAM breaks USB [was Re: USB causing boot to hang]
Steve Kargl
sgk at troutmask.apl.washington.edu
Mon Dec 9 18:52:51 UTC 2019
On Fri, Dec 06, 2019 at 06:11:48PM -0800, Steve Kargl wrote:
> On Sat, Dec 07, 2019 at 01:16:13AM +0100, Hans Petter Selasky wrote:
> >
> > There is an option you can compile into the kernel which will allow the
> > keyboard to enter the debugger.
> >
> > options ALT_BREAK_TO_DEBUGGER
> >
> > Sounds to me like either a leaked refcount or that one thread is
> > spinning blocking execution of other threads.
> >
>
> I tried setting the sysctls
>
> hw.usb.umass.debug: 1
> hw.usb.xhci.debug: 1
> hw.usb.uhci.debug: 1
> hw.usb.ohci.debug: 1
> hw.usb.ehci.debug: 1
>
> but when I rebooted I forgot to request a verbose boot.
> There was no apparent debugging output to the console.
> I try again on Monday to get additional information.
>
keyboard isn't available prior to the hang.
A kernel built at 355010 hangs. This includes kernels
built with and without INVARIANTS, INVARIANT_SUPPORT,
WITNESS, and WITNESS_SKIPSPIN. This includes adding
"kern.cam.boot_delay=30000" to /boot/loader.conf.
Adding "hw.usb.debug=1", "hw.usb.umass.debug=1",
"hw.usb.ctrl.debug=1", "hw.usb.xhci.debug=1",
"hw.usb.uhci.debug=1", "hw.usb.ohci.debug=1", and
"hw.usb.ehci.debug=1" shows no usb related output on
the console when booting the 355010 kernel; whereas, a
355009 kernel floods the console.
Unfortunately, I'm not a kernel hacker and I have run
out of time trying to report an issue.
--
Steve
More information about the freebsd-current
mailing list