Wednesday, December 31, 2003 2:59 PM, James Bottomley wrote:
> The author made the arch/i386 compile depend on NR_VECTORS being
> defined.
>
> This symbol, however, was put only into mach-default/irq_vectors.h
>
> The attached patch adds it to voyager; visws and pc9800 however, are
> still broken.
>
> The code that breaks is this (in arch/i386/kernel/i8259.c):
>
> * us. (some of these will be overridden and become
> * 'special' SMP interrupts)
> */
> - for (i = 0; i < NR_IRQS; i++) {
> + for (i = 0; i < (NR_VECTORS - FIRST_EXTERNAL_VECTOR); i++) {
> int vector = FIRST_EXTERNAL_VECTOR + i;
> + if (i >= NR_IRQS)
> + break;
> if (vector != SYSCALL_VECTOR)
> set_intr_gate(vector, interrupt[i]);
>
> as far as I can see, with NR_VECTORS set at 256, FIRST_EXTERNAL_VECTOR
> at 32 and NR_IRQS set at 224 the two forms of the loop are identical.
> The only case it would make a difference would be for NR_IRQ >
> NR_VECTORS + FIRST_EXTERNAL_VECTOR which doesn't seem to make any
> sense. Perhaps just backing this change out of i8259.c would be
> better? NR_VECTORS seems to have no other defined use in the MSI code.
>> It would make a significant difference when CONFIG_PCI_USE_VECTOR is
>> set to "Y" by users to enable MSI support. The setting of
>> CONFIG_PCI_USE_VECTOR to "Y" sets NR_IRQS at 239 (FIRST_SYSTEM_VECTOR)
>> instead of 224.
> ===== include/asm-i386/mach-voyager/irq_vectors.h 1.4 vs edited =====
> --- 1.4/include/asm-i386/mach-voyager/irq_vectors.h Wed Oct 22 11:34:51
> 2003
> +++ edited/include/asm-i386/mach-voyager/irq_vectors.h Wed Dec 31
> 16:30:15 2003
> @@ -55,6 +55,7 @@
> #define VIC_CPU_BOOT_CPI VIC_CPI_LEVEL0
> #define VIC_CPU_BOOT_ERRATA_CPI (VIC_CPI_LEVEL0 + 8)
>
> +#define NR_VECTORS 256
> #define NR_IRQS 224
> #define NR_IRQ_VECTORS NR_IRQS
>
>> Thanks for providing a fix. The fix looks fine to me.
Thanks,
Long