Helge Hafting wrote:
> I'll try it. Have you considered resubmitting the patch,
> hidden behind a CONFIG_BROKEN_APIC? That'll keep the code
> clean for those with better hardware.
We might as well move the kick_IO_APIC_irq call to the
arch/i386/kernel/irq.c:ack_none function then, surrounded by proper
#ifdefs. The ack_none is the function that does the printk("unexpected
IRQ trap at vector %02x\n", irq), which I see everytime the bug
triggers.
And looking at the comment of the end_level_ioapic_irq function in
io_apic.c, is there a possibility to replace the kick_IO_APIC_irq call
entirely with a end_level_ioapic_irq call? I see lots of similarities in
these two functions.
I didn't test this yet, as I'm still running on Raphael's patch, waiting
for the bug to trigger. (Anyone got a reliable way of triggering it?)
Regards,
- Robbert Kouprie