lkml.org 
[lkml]   [2005]   [May]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [patch 1/1] Do not enforce unique IO_APIC_ID for Xeon processors in EM64T mode (x86_64)
On Wed, May 11, 2005 at 02:10:03AM -0500, Protasevich, Natalie wrote:
> I tried the following patch:
>
> --- io_apic.c.orig 2005-05-11 03:42:40.000000000 -0400
> +++ io_apic.c 2005-05-11 03:44:28.000000000 -0400
> @@ -2427,12 +2427,13 @@ int __init io_apic_get_unique_id (int io
> physid_mask_t tmp;
> unsigned long flags;
> int i = 0;
> + static int no_ioapic_check;
>
> /* Don't check I/O APIC IDs for some xAPIC systems. They have
> * no meaning without the serial APIC bus.
> */
>
> - if (NO_IOAPIC_CHECK)
> + if (no_ioapic_check)
> return apic_id;
>
> /*
> @@ -2449,6 +2450,13 @@ int __init io_apic_get_unique_id (int io
> reg_00.raw = io_apic_read(ioapic, 0);
> spin_unlock_irqrestore(&ioapic_lock, flags);
>
> + if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) &&
> + (boot_cpu_data.x86 == 15)) {
> + if (reg_00.bits.delivery_type)
> + no_ioapic_check = 1;
> + return apic_id;
> + }
> +
> if (apic_id >= get_physical_broadcast()) {
> printk(KERN_WARNING "IOAPIC[%d]: Invalid apic_id %d,
> trying "
> "%d\n", ioapic, apic_id, reg_00.bits.ID);
>
>
> To my surprise, it didn't work, and when I dumped the I/O APIC registers
> it turned out that on ES7000 the parallel delivery bit wasn't set!
>
> <7>.... register #00: A4000000
> <7>....... : physical APIC id: A4
> <7>....... : Delivery Type: 0
> <7>....... : LTS : 0
> <7>.... register #01: 00170004
> <7>....... : max redirection entries: 0017
> <7>....... : PRQ implemented: 0
> <7>....... : IO APIC version: 0004
>
> Version number was not 0x11 as James mentioned, it was 4, and I couldn't
> find anything documented on the I/O xAPIC versioning.
>
> Looks like the need in the unique id can only be keyed of the local APIC
> id, and probably it is a good idea to keep the NO_IOAPIC_CHECK for
> subarchs that can override the heuristics?

I prefer not to do that. How about a simple

if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL && boot_cpu_data.x86 < 15)
/* do uniqueness check */
else
/* don't do it */
?

Rationale is that P4s and newer and systems not from Intel don't have serial
APIC busses and don't need this uniqueness check.

-Andi
-
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/

\
 
 \ /
  Last update: 2005-05-11 13:26    [from the cache]
©2003-2011 Jasper Spaans. Advertise on this site