lkml.org 
[lkml]   [2010]   [Jan]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
SubjectRe: [PATCH] Revert 2fbd07a5f so machines with BSPs phsyical apic id != 0 can boot
From
Date
On Mon, 2010-01-11 at 14:55 -0800, Linus Torvalds wrote:
>
> On Mon, 11 Jan 2010, Suresh Siddha wrote:
> >
> > So far, I don't see a need for the revert.
>
> It's very easy. It breaks. You don't have a fix. We'll revert it.

Linus, We are in -rc3 and thought we have few days atleast to sort it
out and post the correct fix to the problem, rather than do a quick
revert (as we know that the current code is not fundamentally broken).

But if you want to revert, I have appended a patch to revert this, which
has the correct subject, description etc atleast. I can work with Ananth
and re-submit this for the next release.

---
From: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Subject: Revert "x86, apic: Use logical flat on intel with <= 8 logical cpus"

Revert commit 2fbd07a5f5d1295fa9b0c0564ec27da7c276a75a, as this commit breaks
an IBM platform with quad-core Xeon cpu's.

According to Suresh, this might be an IBM platform issue, as on other Intel
platforms with <= 8 logical cpu's, logical flat mode works fine irrespective of
physical apic id values (inline with the xapic architecture).

Revert this for now because of an IBM platform breakage.

Another version will be re-submitted after the complete analysis.

Signed-off-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Acked-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: stable@kernel.org
---

diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
index 79e5b92..072aea6 100644
--- a/arch/x86/kernel/apic/apic.c
+++ b/arch/x86/kernel/apic/apic.c
@@ -61,7 +61,7 @@ unsigned int boot_cpu_physical_apicid = -1U;
/*
* The highest APIC ID seen during enumeration.
*
- * This determines the messaging protocol we can use: if all APIC IDs
+ * On AMD, this determines the messaging protocol we can use: if all APIC IDs
* are in the 0 ... 7 range, then we can use logical addressing which
* has some performance advantages (better broadcasting).
*
@@ -1915,24 +1915,14 @@ void __cpuinit generic_processor_info(int apicid, int version)
max_physical_apicid = apicid;

#ifdef CONFIG_X86_32
- /*
- * Would be preferable to switch to bigsmp when CONFIG_HOTPLUG_CPU=y
- * but we need to work other dependencies like SMP_SUSPEND etc
- * before this can be done without some confusion.
- * if (CPU_HOTPLUG_ENABLED || num_processors > 8)
- * - Ashok Raj <ashok.raj@intel.com>
- */
- if (max_physical_apicid >= 8) {
- switch (boot_cpu_data.x86_vendor) {
- case X86_VENDOR_INTEL:
- if (!APIC_XAPIC(version)) {
- def_to_bigsmp = 0;
- break;
- }
- /* If P4 and above fall through */
- case X86_VENDOR_AMD:
+ switch (boot_cpu_data.x86_vendor) {
+ case X86_VENDOR_INTEL:
+ if (num_processors > 8)
+ def_to_bigsmp = 1;
+ break;
+ case X86_VENDOR_AMD:
+ if (max_physical_apicid >= 8)
def_to_bigsmp = 1;
- }
}
#endif

diff --git a/arch/x86/kernel/apic/probe_64.c b/arch/x86/kernel/apic/probe_64.c
index 65edc18..c4cbd30 100644
--- a/arch/x86/kernel/apic/probe_64.c
+++ b/arch/x86/kernel/apic/probe_64.c
@@ -64,16 +64,23 @@ void __init default_setup_apic_routing(void)
apic = &apic_x2apic_phys;
else
apic = &apic_x2apic_cluster;
- printk(KERN_INFO "Setting APIC routing to %s\n", apic->name);
}
#endif

if (apic == &apic_flat) {
- if (max_physical_apicid >= 8)
- apic = &apic_physflat;
- printk(KERN_INFO "Setting APIC routing to %s\n", apic->name);
+ switch (boot_cpu_data.x86_vendor) {
+ case X86_VENDOR_INTEL:
+ if (num_processors > 8)
+ apic = &apic_physflat;
+ break;
+ case X86_VENDOR_AMD:
+ if (max_physical_apicid >= 8)
+ apic = &apic_physflat;
+ }
}

+ printk(KERN_INFO "Setting APIC routing to %s\n", apic->name);
+
if (is_vsmp_box()) {
/* need to update phys_pkg_id */
apic->phys_pkg_id = apicid_phys_pkg_id;



\
 
 \ /
  Last update: 2010-01-12 00:49    [W:0.419 / U:0.120 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site