Re: [patch 00/14] x86/irq: Plug various vector cleanup races

[Joe Lawrence]

On 01/20/2016 03:26 AM, Borislav Petkov wrote:
> ... tglx says you could also try enabling
> CONFIG_DEBUG_OBJECTS and CONFIG_DEBUG_OBJECTS_TIMERS and collect full
> dmesg from when the corruption happens. That might give us some more
> insights as to what happens.

Hi Boris,

Excellent idea.  With CONFIG_DEBUG_OBJECTS_FREE=y  I've hit this twice:

ODEBUG: free active (active state 0) object type: timer_list hint:
xhci_stop_endpoint_command_watchdog+0x0/0x2b0

So I'll try to hunt that down on the xHCI side and report over to
linux-usb.  Sorry for the unrelated noise on this patch -- my guess is
that timing changed just enough for the xHCI driver to hit this. :)

-- Joe