Re: 48GB NUMA-Q boots, with major IO-APIC hassles

From: Zwane Mwaikambo (zwane@holomorphy.com)
Date: Sat Jan 18 2003 - 21:32:22 EST


On Sat, 18 Jan 2003, William Lee Irwin III wrote:

> +
> +int vector_to_irq[MAX_NUMNODES][FIRST_SYSTEM_VECTOR - FIRST_DEVICE_VECTOR + 1];
> +int apic_pin_to_irq[MAX_IO_APICS][24];
> +
> +/*
> + * timer vectors must always go to 0
> + * vectors < FIRST_DEVICE_VECTOR are 1:1
> + * everything else goes through the table
> + */

Careful.

NR Log Phy Mask Trig IRR Pol Stat Dest Deli Vect:
 00 001 01 0 0 0 0 0 1 1 31
 01 001 01 0 0 0 0 0 1 1 39
 02 000 00 1 0 0 0 0 0 0 00
 03 001 01 0 0 0 0 0 1 1 41
 04 001 01 0 0 0 0 0 1 1 49
 05 001 01 0 0 0 0 0 1 1 51
 06 001 01 0 0 0 0 0 1 1 59
 07 001 01 0 0 0 0 0 1 1 61

> +int irq_of_vector(int vector)
> +{
> + int irq;
> + if (vector < FIRST_DEVICE_VECTOR)
> + irq = vector;
> + else
> + irq = vector_to_irq[numa_node_id()][vector-FIRST_DEVICE_VECTOR];
> + return irq;
> +}

Same as above.

> +static int __init assign_irq_vector(int irq)
> +{
> + static int current_vector = FIRST_DEVICE_VECTOR+1;
> + if (!irq)
> + return FIRST_DEVICE_VECTOR;
> + else if (!irq_vector[irq]) {
> + irq_vector[irq] = current_vector;
> + current_vector = next_irq_vector(current_vector);
> + }
> + return irq_vector[irq];
> +}

You'll drop irqs when you have collisions with devices
attached to other busses/ioapics

> +#define init_vector_to_irq() do {} while (0)
> +#define set_irq_of_vector(a,v,i) do {} while (0)
> +#define set_irq_of_pin(a,v,i) do {} while (0)
> +
> +int irq_of_vector(int vector) { return vector; }

This would need major warnings not to use on non NUMAQ since that
information is relevant on Walmart SMP.

> +#ifdef CONFIG_X86_NUMAQ
> + set_intr_gate(vector, interrupt[vector]);
> +#else
> set_intr_gate(vector, interrupt[irq]);
> +#endif

hmm..

> + if (irq < 0) {
> + printk("bad vector %ld, irq %d\n", regs.orig_eax & 0xff, irq);
> + dump_stack();
> + return 1;
> + }
> +

Oh my =)

        Zwane

-- 
function.linuxpower.ca

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



This archive was generated by hypermail 2b29 : Thu Jan 23 2003 - 22:00:20 EST