Re: [Bugfix] x86, irq: Fix bug in setting IOAPIC pin attributes

From: Thomas Gleixner
Date: Tue Aug 26 2014 - 17:21:21 EST


On Tue, 26 Aug 2014, Mika Westerberg wrote:

> On Tue, Aug 26, 2014 at 04:45:01PM +0800, Jiang Liu wrote:
> > On startup, IOAPIC driver preallocates IRQ descriptors and programs
> > IOAPIC pins with default level and polarity attributes for all legacy
> > IRQs. Later legacy IRQ users may fail to set IOAPIC pin attributes
> > if the requested attributes conflicts with the default IOAPIC pin
> > attributes. So change mp_irqdomain_map() to allow the first legacy IRQ
> > user to reprogram IOAPIC pin with different attributes.
> >
> > Signed-off-by: Jiang Liu <jiang.liu@xxxxxxxxxxxxxxx>
>
> I tried this and now we get further in setting up the IOAPIC but it
> still doesn't work :-(
>
> I noticed that mp_set_gsi_attr() never programs the IOAPIC in case
> info->set is 0 so I added following on top of your patch. With the below
> patch, the system seems to work as expected.
>
> diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
> index 66c2700c9f83..777762a0ad09 100644
> --- a/arch/x86/kernel/apic/io_apic.c
> +++ b/arch/x86/kernel/apic/io_apic.c
> @@ -3942,10 +3942,16 @@ int mp_set_gsi_attr(u32 gsi, int trigger, int polarity, int node)
>
> mutex_lock(&ioapic_mutex);
> if (!info->set) {
> + struct io_apic_irq_attr attr;
> +
> info->trigger = trigger;
> info->polarity = polarity;
> info->node = node;
> info->set = 1;
> +
> + /* Now program the IOAPIC with the new attributes */
> + set_io_apic_irq_attr(&attr, ioapic, pin, trigger, polarity);
> + ret = io_apic_setup_irq_pin(gsi, node, &attr);
> } else if (info->trigger != trigger || info->polarity != polarity) {
> ret = -EBUSY;
> }

Can someone please combine this into a single patch with a proper
changelog?

Thanks,

tglx

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/