On Wed, 15 Feb 2017, Gabriel C wrote: > On 14.02.2017 17:45, Thomas Gleixner wrote: > The only place I see your debug patch triggers is : > > .. > > [ 7.029816] calling goldfish_pdev_bus_driver_init+0x0/0x14 @ 1 > [ 7.031707] Retrigger (null) 4 > [ 7.033556] Goldfish IRQ Ok. So the interrupt is pending for unknown reasons when it gets requested by that goldfish driver and the enable logic retriggers it. The retrigger invokes that interrupt handler which then goes into infinite loop mode. Having interrupts marked pending after boot happens occasionally and with proper written drivers and interrupt handlers that's a non issue. Every interrupt routine must be able to handle spurious interrupts. So that's not a some mysterious low level interrupt handling problem, it's just a completely bogus driver. > So data->chip->name being NULL here. Ok. That's the lapic_controller chip missing a chip name. I'll fix that along with a command line option which makes that goldfish nonsense not register that platform device unconditionally. Thanks, tglx