* [PATCH] CONFIG_SMP=n fixes redux
@ 2004-09-01 0:44 Jesse Barnes
2004-09-01 18:14 ` Jesse Barnes
0 siblings, 1 reply; 2+ messages in thread
From: Jesse Barnes @ 2004-09-01 0:44 UTC (permalink / raw)
To: linux-ia64
[-- Attachment #1: Type: text/plain, Size: 1259 bytes --]
Today both Christoph and Alex mentioned that they thought this patch would be
nice to have in the tree. It allows a generic kernel to build and run when
CONFIG_SMP=n. I'd still like feedback about the per-cpu stuff though, in
particular, this
+#ifdef CONFIG_SMP
/* Set the node_data pointer for each per-cpu struct */
for (cpu = 0; cpu < NR_CPUS; cpu++) {
node = node_cpuid[cpu].nid;
per_cpu(cpu_info, cpu).node_data = mem_data[node].node_data;
}
+#else
+ {
+ struct cpuinfo_ia64 *cpu0_cpu_info;
+ cpu = 0;
+ node = node_cpuid[cpu].nid;
+ cpu0_cpu_info = (struct cpuinfo_ia64 *)(__phys_per_cpu_start +
+ ((char *)&per_cpu__cpu_info - __per_cpu_start));
+ cpu0_cpu_info->node_data = mem_data[node].node_data;
+ }
+#endif /* CONFIG_SMP */
is really ugly. Maybe I should take this opportunity to convert the early
memory allocation code to using the bootmem allocator instead of backdooring
it via find_pernode_space()? I think that would mean we could share more
code between the contig and discontigmem cases, but we'd also have to be
careful to avoid cacheline aliasing.
Thanks,
Jesse
[-- Attachment #2: ia64-generic-no-smp-3.patch --]
[-- Type: text/plain, Size: 7248 bytes --]
===== arch/ia64/kernel/Makefile 1.34 vs edited =====
--- 1.34/arch/ia64/kernel/Makefile 2004-08-16 07:51:35 -07:00
+++ edited/arch/ia64/kernel/Makefile 2004-08-31 17:33:30 -07:00
@@ -15,6 +15,7 @@
obj-$(CONFIG_IOSAPIC) += iosapic.o
obj-$(CONFIG_MODULES) += module.o
obj-$(CONFIG_SMP) += smp.o smpboot.o
+obj-$(CONFIG_NUMA) += numa.o
obj-$(CONFIG_PERFMON) += perfmon_default_smpl.o
obj-$(CONFIG_IA64_CYCLONE) += cyclone.o
===== arch/ia64/kernel/acpi.c 1.74 vs edited =====
--- 1.74/arch/ia64/kernel/acpi.c 2004-08-05 21:40:29 -07:00
+++ edited/arch/ia64/kernel/acpi.c 2004-08-31 17:33:31 -07:00
@@ -624,8 +624,10 @@
if (smp_boot_data.cpu_phys_id[cpu] != hard_smp_processor_id())
node_cpuid[i++].phys_id = smp_boot_data.cpu_phys_id[cpu];
}
- build_cpu_to_node_map();
# endif
+#endif
+#ifdef CONFIG_ACPI_NUMA
+ build_cpu_to_node_map();
#endif
/* Make boot-up look pretty */
printk(KERN_INFO "%d CPUs available, %d CPUs total\n", available_cpus, total_cpus);
===== arch/ia64/kernel/setup.c 1.79 vs edited =====
--- 1.79/arch/ia64/kernel/setup.c 2004-08-04 11:08:05 -07:00
+++ edited/arch/ia64/kernel/setup.c 2004-08-31 17:33:32 -07:00
@@ -317,11 +317,9 @@
machvec_init(acpi_get_sysname());
#endif
-#ifdef CONFIG_SMP
/* If we register an early console, allow CPU 0 to printk */
if (!early_console_setup())
cpu_set(smp_processor_id(), cpu_online_map);
-#endif
#ifdef CONFIG_ACPI_BOOT
/* Initialize the ACPI boot-time table parser */
===== arch/ia64/kernel/smpboot.c 1.60 vs edited =====
--- 1.60/arch/ia64/kernel/smpboot.c 2004-08-24 02:08:43 -07:00
+++ edited/arch/ia64/kernel/smpboot.c 2004-08-31 17:33:33 -07:00
@@ -478,47 +478,6 @@
}
}
-#ifdef CONFIG_NUMA
-
-/* on which node is each logical CPU (one cacheline even for 64 CPUs) */
-u8 cpu_to_node_map[NR_CPUS] __cacheline_aligned;
-EXPORT_SYMBOL(cpu_to_node_map);
-/* which logical CPUs are on which nodes */
-cpumask_t node_to_cpu_mask[MAX_NUMNODES] __cacheline_aligned;
-
-/*
- * Build cpu to node mapping and initialize the per node cpu masks.
- */
-void __init
-build_cpu_to_node_map (void)
-{
- int cpu, i, node;
-
- for(node=0; node<MAX_NUMNODES; node++)
- cpus_clear(node_to_cpu_mask[node]);
- for(cpu = 0; cpu < NR_CPUS; ++cpu) {
- /*
- * All Itanium NUMA platforms I know use ACPI, so maybe we
- * can drop this ifdef completely. [EF]
- */
-#ifdef CONFIG_ACPI_NUMA
- node = -1;
- for (i = 0; i < NR_CPUS; ++i)
- if (cpu_physical_id(cpu) == node_cpuid[i].phys_id) {
- node = node_cpuid[i].nid;
- break;
- }
-#else
-# error Fixme: Dunno how to build CPU-to-node map.
-#endif
- cpu_to_node_map[cpu] = (node >= 0) ? node : 0;
- if (node >= 0)
- cpu_set(cpu, node_to_cpu_mask[node]);
- }
-}
-
-#endif /* CONFIG_NUMA */
-
/*
* Cycle through the APs sending Wakeup IPIs to boot each.
*/
===== arch/ia64/mm/discontig.c 1.18 vs edited =====
--- 1.18/arch/ia64/mm/discontig.c 2004-08-26 23:30:29 -07:00
+++ edited/arch/ia64/mm/discontig.c 2004-08-31 17:33:33 -07:00
@@ -225,6 +225,33 @@
}
/**
+ * per_cpu_node_setup - setup per-cpu areas on each node
+ * @cpu_data: per-cpu area on this node
+ * @node: node to setup
+ *
+ * Copy the static per-cpu data into the region we just set aside and then
+ * setup __per_cpu_offset for each CPU on this node. Return a pointer to
+ * the end of the area.
+ */
+static void *per_cpu_node_setup(void *cpu_data, int node)
+{
+#ifdef CONFIG_SMP
+ int cpu;
+
+ for (cpu = 0; cpu < NR_CPUS; cpu++) {
+ if (node == node_cpuid[cpu].nid) {
+ memcpy(__va(cpu_data), __phys_per_cpu_start,
+ __per_cpu_end - __per_cpu_start);
+ __per_cpu_offset[cpu] = (char*)__va(cpu_data) -
+ __per_cpu_start;
+ cpu_data += PERCPU_PAGE_SIZE;
+ }
+ }
+#endif
+ return cpu_data;
+}
+
+/**
* find_pernode_space - allocate memory for memory map and per-node structures
* @start: physical start of range
* @len: length of range
@@ -255,7 +282,7 @@
static int __init find_pernode_space(unsigned long start, unsigned long len,
int node)
{
- unsigned long epfn, cpu, cpus;
+ unsigned long epfn, cpus;
unsigned long pernodesize = 0, pernode, pages, mapsize;
void *cpu_data;
struct bootmem_data *bdp = &mem_data[node].bootmem_data;
@@ -305,20 +332,7 @@
mem_data[node].pgdat->bdata = bdp;
pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
- /*
- * Copy the static per-cpu data into the region we
- * just set aside and then setup __per_cpu_offset
- * for each CPU on this node.
- */
- for (cpu = 0; cpu < NR_CPUS; cpu++) {
- if (node == node_cpuid[cpu].nid) {
- memcpy(__va(cpu_data), __phys_per_cpu_start,
- __per_cpu_end - __per_cpu_start);
- __per_cpu_offset[cpu] = (char*)__va(cpu_data) -
- __per_cpu_start;
- cpu_data += PERCPU_PAGE_SIZE;
- }
- }
+ cpu_data = per_cpu_node_setup(cpu_data, node);
}
return 0;
@@ -384,8 +398,8 @@
*/
static void __init initialize_pernode_data(void)
{
- int cpu, node;
pg_data_t *pgdat_list[NR_NODES];
+ int cpu, node;
for (node = 0; node < numnodes; node++)
pgdat_list[node] = mem_data[node].pgdat;
@@ -395,12 +409,22 @@
memcpy(mem_data[node].node_data->pg_data_ptrs, pgdat_list,
sizeof(pgdat_list));
}
-
+#ifdef CONFIG_SMP
/* Set the node_data pointer for each per-cpu struct */
for (cpu = 0; cpu < NR_CPUS; cpu++) {
node = node_cpuid[cpu].nid;
per_cpu(cpu_info, cpu).node_data = mem_data[node].node_data;
}
+#else
+ {
+ struct cpuinfo_ia64 *cpu0_cpu_info;
+ cpu = 0;
+ node = node_cpuid[cpu].nid;
+ cpu0_cpu_info = (struct cpuinfo_ia64 *)(__phys_per_cpu_start +
+ ((char *)&per_cpu__cpu_info - __per_cpu_start));
+ cpu0_cpu_info->node_data = mem_data[node].node_data;
+ }
+#endif /* CONFIG_SMP */
}
/**
@@ -464,6 +488,7 @@
find_initrd();
}
+#ifdef CONFIG_SMP
/**
* per_cpu_init - setup per-cpu variables
*
@@ -474,15 +499,15 @@
{
int cpu;
- if (smp_processor_id() == 0) {
- for (cpu = 0; cpu < NR_CPUS; cpu++) {
- per_cpu(local_per_cpu_offset, cpu) =
- __per_cpu_offset[cpu];
- }
- }
+ if (smp_processor_id() != 0)
+ return __per_cpu_start + __per_cpu_offset[smp_processor_id()];
+
+ for (cpu = 0; cpu < NR_CPUS; cpu++)
+ per_cpu(local_per_cpu_offset, cpu) = __per_cpu_offset[cpu];
return __per_cpu_start + __per_cpu_offset[smp_processor_id()];
}
+#endif /* CONFIG_SMP */
/**
* show_mem - give short summary of memory stats
===== include/asm-ia64/smp.h 1.18 vs edited =====
--- 1.18/include/asm-ia64/smp.h 2004-06-21 13:44:12 -07:00
+++ edited/include/asm-ia64/smp.h 2004-08-31 17:33:34 -07:00
@@ -126,6 +126,7 @@
#else
#define cpu_logical_id(cpuid) 0
+#define cpu_physical_id(i) ((ia64_getreg(_IA64_REG_CR_LID) >> 16) & 0xffff)
#endif /* CONFIG_SMP */
#endif /* _ASM_IA64_SMP_H */
===== include/asm-ia64/sn/sn_cpuid.h 1.10 vs edited =====
--- 1.10/include/asm-ia64/sn/sn_cpuid.h 2004-08-18 11:00:37 -07:00
+++ edited/include/asm-ia64/sn/sn_cpuid.h 2004-08-31 17:33:35 -07:00
@@ -83,10 +83,6 @@
*
*/
-#ifndef CONFIG_SMP
-#define cpu_physical_id(cpuid) ((ia64_getreg(_IA64_REG_CR_LID) >> 16) & 0xffff)
-#endif
-
/*
* macros for some of these exist in sn/addrs.h & sn/arch.h, etc. However,
* trying #include these files here causes circular dependencies.
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH] CONFIG_SMP=n fixes redux
2004-09-01 0:44 [PATCH] CONFIG_SMP=n fixes redux Jesse Barnes
@ 2004-09-01 18:14 ` Jesse Barnes
0 siblings, 0 replies; 2+ messages in thread
From: Jesse Barnes @ 2004-09-01 18:14 UTC (permalink / raw)
To: linux-ia64
[-- Attachment #1: Type: text/plain, Size: 1775 bytes --]
On Tuesday, August 31, 2004 5:44 pm, Jesse Barnes wrote:
> Today both Christoph and Alex mentioned that they thought this patch would
> be nice to have in the tree. It allows a generic kernel to build and run
> when CONFIG_SMP=n. I'd still like feedback about the per-cpu stuff though,
> in particular, this
>
> +#ifdef CONFIG_SMP
> /* Set the node_data pointer for each per-cpu struct */
> for (cpu = 0; cpu < NR_CPUS; cpu++) {
> node = node_cpuid[cpu].nid;
> per_cpu(cpu_info, cpu).node_data =
> mem_data[node].node_data; }
> +#else
> + {
> + struct cpuinfo_ia64 *cpu0_cpu_info;
> + cpu = 0;
> + node = node_cpuid[cpu].nid;
> + cpu0_cpu_info = (struct cpuinfo_ia64
> *)(__phys_per_cpu_start + + ((char
> *)&per_cpu__cpu_info - __per_cpu_start)); +
> cpu0_cpu_info->node_data = mem_data[node].node_data; + }
> +#endif /* CONFIG_SMP */
>
> is really ugly. Maybe I should take this opportunity to convert the early
> memory allocation code to using the bootmem allocator instead of
> backdooring it via find_pernode_space()? I think that would mean we could
> share more code between the contig and discontigmem cases, but we'd also
> have to be careful to avoid cacheline aliasing.
The patch I posted here was incomplete (missing numa.c file). This new one
fixes the CONFIG_NUMA=n case too, but has led me to conclude that the whole
thing is a big mess. Our CONFIG_* variables aren't named very well and
there's a lot of confusion between actual NUMA support routines (e.g. smart
page allocation, distance tables, etc.) and simple memory layout information
necessary for discontig support on ia64 boxes.
Jesse
[-- Attachment #2: ia64-generic-config-fixes.patch --]
[-- Type: text/plain, Size: 19646 bytes --]
# This is a BitKeeper generated diff -Nru style patch.
#
# ChangeSet
# 2004/09/01 11:11:28-07:00 jbarnes@tomahawk.engr.sgi.com
# config fixes
#
# arch/ia64/kernel/numa.c
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +57 -0
#
# include/linux/acpi.h
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +4 -0
# config fixes
#
# include/asm-ia64/sn/sn_cpuid.h
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +0 -4
# config fixes
#
# include/asm-ia64/smp.h
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +1 -0
# config fixes
#
# include/asm-ia64/processor.h
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +1 -3
# config fixes
#
# include/asm-ia64/numa.h
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +1 -8
# config fixes
#
# include/asm-ia64/acpi.h
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +1 -0
# config fixes
#
# drivers/acpi/numa.c
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +0 -2
# config fixes
#
# drivers/acpi/Kconfig
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +0 -1
# config fixes
#
# arch/ia64/sn/kernel/setup.c
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +1 -0
# config fixes
#
# arch/ia64/mm/numa.c
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +0 -7
# config fixes
#
# arch/ia64/mm/discontig.c
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +62 -31
# config fixes
#
# arch/ia64/kernel/smpboot.c
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +0 -41
# config fixes
#
# arch/ia64/kernel/setup.c
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +0 -2
# config fixes
#
# arch/ia64/kernel/numa.c
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +0 -0
# BitKeeper file /home/jbarnes/working/linux-2.5-numa/arch/ia64/kernel/numa.c
#
# arch/ia64/kernel/acpi.c
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +21 -3
# config fixes
#
# arch/ia64/kernel/Makefile
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +1 -0
# config fixes
#
# arch/ia64/Kconfig
# 2004/09/01 11:11:18-07:00 jbarnes@tomahawk.engr.sgi.com +0 -2
# config fixes
#
diff -Nru a/arch/ia64/Kconfig b/arch/ia64/Kconfig
--- a/arch/ia64/Kconfig 2004-09-01 11:12:45 -07:00
+++ b/arch/ia64/Kconfig 2004-09-01 11:12:45 -07:00
@@ -44,8 +44,6 @@
config IA64_GENERIC
bool "generic"
- select NUMA
- select ACPI_NUMA
select VIRTUAL_MEM_MAP
select DISCONTIGMEM
help
diff -Nru a/arch/ia64/kernel/Makefile b/arch/ia64/kernel/Makefile
--- a/arch/ia64/kernel/Makefile 2004-09-01 11:12:45 -07:00
+++ b/arch/ia64/kernel/Makefile 2004-09-01 11:12:45 -07:00
@@ -15,6 +15,7 @@
obj-$(CONFIG_IOSAPIC) += iosapic.o
obj-$(CONFIG_MODULES) += module.o
obj-$(CONFIG_SMP) += smp.o smpboot.o
+obj-$(CONFIG_NUMA) += numa.o
obj-$(CONFIG_PERFMON) += perfmon_default_smpl.o
obj-$(CONFIG_IA64_CYCLONE) += cyclone.o
diff -Nru a/arch/ia64/kernel/acpi.c b/arch/ia64/kernel/acpi.c
--- a/arch/ia64/kernel/acpi.c 2004-09-01 11:12:45 -07:00
+++ b/arch/ia64/kernel/acpi.c 2004-09-01 11:12:45 -07:00
@@ -359,6 +359,17 @@
/* maps to convert between proximity domain and logical node ID */
int __initdata pxm_to_nid_map[MAX_PXM_DOMAINS];
int __initdata nid_to_pxm_map[MAX_NUMNODES];
+
+/*
+ * The following structures are usually initialized by ACPI or
+ * similar mechanisms and describe the NUMA characteristics of the machine.
+ */
+int num_node_memblks;
+struct node_memblk_s node_memblk[NR_NODE_MEMBLKS];
+struct node_cpuid_s node_cpuid[NR_CPUS];
+
+#ifdef CONFIG_NUMA
+
static struct acpi_table_slit __initdata *slit_table;
/*
@@ -380,6 +391,7 @@
}
slit_table = slit;
}
+#endif
void __init
acpi_numa_processor_affinity_init (struct acpi_table_processor_affinity *pa)
@@ -434,7 +446,10 @@
void __init
acpi_numa_arch_fixup (void)
{
- int i, j, node_from, node_to;
+ int i, j;
+#ifdef CONFIG_NUMA
+ int node_from, node_to;
+#endif
/* If there's no SRAT, fix the phys_id */
if (srat_num_cpus == 0) {
@@ -475,7 +490,7 @@
printk(KERN_INFO "Number of logical nodes in system = %d\n", numnodes);
printk(KERN_INFO "Number of memory chunks in system = %d\n", num_node_memblks);
-
+#ifdef CONFIG_NUMA
if (!slit_table) return;
memset(numa_slit, -1, sizeof(numa_slit));
for (i=0; i<slit_table->localities; i++) {
@@ -490,6 +505,7 @@
slit_table->entry[i*slit_table->localities + j];
}
}
+#endif
#ifdef SLIT_DEBUG
printk("ACPI 2.0 SLIT locality table:\n");
@@ -624,8 +640,10 @@
if (smp_boot_data.cpu_phys_id[cpu] != hard_smp_processor_id())
node_cpuid[i++].phys_id = smp_boot_data.cpu_phys_id[cpu];
}
- build_cpu_to_node_map();
# endif
+#endif
+#ifdef CONFIG_NUMA
+ build_cpu_to_node_map();
#endif
/* Make boot-up look pretty */
printk(KERN_INFO "%d CPUs available, %d CPUs total\n", available_cpus, total_cpus);
diff -Nru a/arch/ia64/kernel/numa.c b/arch/ia64/kernel/numa.c
--- /dev/null Wed Dec 31 16:00:00 196900
+++ b/arch/ia64/kernel/numa.c 2004-09-01 11:12:45 -07:00
@@ -0,0 +1,57 @@
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ *
+ * ia64 kernel NUMA specific stuff
+ *
+ * Copyright (C) 2002 Erich Focht <efocht@ess.nec.de>
+ * Copyright (C) 2004 Silicon Graphics, Inc.
+ * Jesse Barnes <jbarnes@sgi.com>
+ */
+#include <linux/config.h>
+#include <linux/topology.h>
+#include <linux/module.h>
+#include <asm/processor.h>
+#include <asm/smp.h>
+
+u8 cpu_to_node_map[NR_CPUS] __cacheline_aligned;
+EXPORT_SYMBOL(cpu_to_node_map);
+
+cpumask_t node_to_cpu_mask[MAX_NUMNODES] __cacheline_aligned;
+
+/**
+ * build_cpu_to_node_map - setup cpu to node and node to cpumask arrays
+ *
+ * Build cpu to node mapping and initialize the per node cpu masks using
+ * info from the node_cpuid array handed to us by ACPI.
+ */
+void __init build_cpu_to_node_map(void)
+{
+ int cpu, i, node;
+
+ for(node=0; node < MAX_NUMNODES; node++)
+ cpus_clear(node_to_cpu_mask[node]);
+
+ for(cpu = 0; cpu < NR_CPUS; ++cpu) {
+ node = -1;
+ for (i = 0; i < NR_CPUS; ++i)
+ if (cpu_physical_id(cpu) == node_cpuid[i].phys_id) {
+ node = node_cpuid[i].nid;
+ break;
+ }
+ cpu_to_node_map[cpu] = (node >= 0) ? node : 0;
+ if (node >= 0)
+ cpu_set(cpu, node_to_cpu_mask[node]);
+ }
+}
diff -Nru a/arch/ia64/kernel/setup.c b/arch/ia64/kernel/setup.c
--- a/arch/ia64/kernel/setup.c 2004-09-01 11:12:45 -07:00
+++ b/arch/ia64/kernel/setup.c 2004-09-01 11:12:45 -07:00
@@ -317,11 +317,9 @@
machvec_init(acpi_get_sysname());
#endif
-#ifdef CONFIG_SMP
/* If we register an early console, allow CPU 0 to printk */
if (!early_console_setup())
cpu_set(smp_processor_id(), cpu_online_map);
-#endif
#ifdef CONFIG_ACPI_BOOT
/* Initialize the ACPI boot-time table parser */
diff -Nru a/arch/ia64/kernel/smpboot.c b/arch/ia64/kernel/smpboot.c
--- a/arch/ia64/kernel/smpboot.c 2004-09-01 11:12:45 -07:00
+++ b/arch/ia64/kernel/smpboot.c 2004-09-01 11:12:45 -07:00
@@ -478,47 +478,6 @@
}
}
-#ifdef CONFIG_NUMA
-
-/* on which node is each logical CPU (one cacheline even for 64 CPUs) */
-u8 cpu_to_node_map[NR_CPUS] __cacheline_aligned;
-EXPORT_SYMBOL(cpu_to_node_map);
-/* which logical CPUs are on which nodes */
-cpumask_t node_to_cpu_mask[MAX_NUMNODES] __cacheline_aligned;
-
-/*
- * Build cpu to node mapping and initialize the per node cpu masks.
- */
-void __init
-build_cpu_to_node_map (void)
-{
- int cpu, i, node;
-
- for(node=0; node<MAX_NUMNODES; node++)
- cpus_clear(node_to_cpu_mask[node]);
- for(cpu = 0; cpu < NR_CPUS; ++cpu) {
- /*
- * All Itanium NUMA platforms I know use ACPI, so maybe we
- * can drop this ifdef completely. [EF]
- */
-#ifdef CONFIG_ACPI_NUMA
- node = -1;
- for (i = 0; i < NR_CPUS; ++i)
- if (cpu_physical_id(cpu) == node_cpuid[i].phys_id) {
- node = node_cpuid[i].nid;
- break;
- }
-#else
-# error Fixme: Dunno how to build CPU-to-node map.
-#endif
- cpu_to_node_map[cpu] = (node >= 0) ? node : 0;
- if (node >= 0)
- cpu_set(cpu, node_to_cpu_mask[node]);
- }
-}
-
-#endif /* CONFIG_NUMA */
-
/*
* Cycle through the APs sending Wakeup IPIs to boot each.
*/
diff -Nru a/arch/ia64/mm/discontig.c b/arch/ia64/mm/discontig.c
--- a/arch/ia64/mm/discontig.c 2004-09-01 11:12:45 -07:00
+++ b/arch/ia64/mm/discontig.c 2004-09-01 11:12:45 -07:00
@@ -9,7 +9,7 @@
/*
* Platform initialization for Discontig Memory
*/
-
+#include <linux/config.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/swap.h>
@@ -40,6 +40,7 @@
static struct early_node_data mem_data[NR_NODES] __initdata;
+#ifdef CONFIG_NUMA
/**
* reassign_cpu_only_nodes - called from find_memory to move CPU-only nodes to a memory node
*
@@ -161,13 +162,16 @@
return;
}
+#else
+static void __init reassign_cpu_only_nodes(void) { }
+#endif /* CONFIG_NUMA */
/*
* To prevent cache aliasing effects, align per-node structures so that they
* start at addresses that are strided by node number.
*/
-#define NODEDATA_ALIGN(addr, node) \
- ((((addr) + 1024*1024-1) & ~(1024*1024-1)) + (node)*PERCPU_PAGE_SIZE)
+#define NODEDATA_ALIGN(addr, node) ((((addr) + 1024*1024-1) & \
+ ~(1024*1024-1)) + (node)*PERCPU_PAGE_SIZE)
/**
* build_node_maps - callback to setup bootmem structs for each node
@@ -213,7 +217,7 @@
* acpi_boot_init() (which builds the node_to_cpu_mask array) hasn't been
* called yet.
*/
-static int early_nr_cpus_node(int node)
+static int __init early_nr_cpus_node(int node)
{
int cpu, n = 0;
@@ -225,6 +229,33 @@
}
/**
+ * per_cpu_node_setup - setup per-cpu areas on each node
+ * @cpu_data: per-cpu area on this node
+ * @node: node to setup
+ *
+ * Copy the static per-cpu data into the region we just set aside and then
+ * setup __per_cpu_offset for each CPU on this node. Return a pointer to
+ * the end of the area.
+ */
+static void __init *per_cpu_node_setup(void *cpu_data, int node)
+{
+#ifdef CONFIG_SMP
+ int cpu;
+
+ for (cpu = 0; cpu < NR_CPUS; cpu++) {
+ if (node == node_cpuid[cpu].nid) {
+ memcpy(__va(cpu_data), __phys_per_cpu_start,
+ __per_cpu_end - __per_cpu_start);
+ __per_cpu_offset[cpu] = (char*)__va(cpu_data) -
+ __per_cpu_start;
+ cpu_data += PERCPU_PAGE_SIZE;
+ }
+ }
+#endif
+ return cpu_data;
+}
+
+/**
* find_pernode_space - allocate memory for memory map and per-node structures
* @start: physical start of range
* @len: length of range
@@ -255,7 +286,7 @@
static int __init find_pernode_space(unsigned long start, unsigned long len,
int node)
{
- unsigned long epfn, cpu, cpus;
+ unsigned long epfn, cpus;
unsigned long pernodesize = 0, pernode, pages, mapsize;
void *cpu_data;
struct bootmem_data *bdp = &mem_data[node].bootmem_data;
@@ -305,20 +336,7 @@
mem_data[node].pgdat->bdata = bdp;
pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
- /*
- * Copy the static per-cpu data into the region we
- * just set aside and then setup __per_cpu_offset
- * for each CPU on this node.
- */
- for (cpu = 0; cpu < NR_CPUS; cpu++) {
- if (node == node_cpuid[cpu].nid) {
- memcpy(__va(cpu_data), __phys_per_cpu_start,
- __per_cpu_end - __per_cpu_start);
- __per_cpu_offset[cpu] = (char*)__va(cpu_data) -
- __per_cpu_start;
- cpu_data += PERCPU_PAGE_SIZE;
- }
- }
+ cpu_data = per_cpu_node_setup(cpu_data, node);
}
return 0;
@@ -384,8 +402,8 @@
*/
static void __init initialize_pernode_data(void)
{
- int cpu, node;
pg_data_t *pgdat_list[NR_NODES];
+ int cpu, node;
for (node = 0; node < numnodes; node++)
pgdat_list[node] = mem_data[node].pgdat;
@@ -395,12 +413,22 @@
memcpy(mem_data[node].node_data->pg_data_ptrs, pgdat_list,
sizeof(pgdat_list));
}
-
+#ifdef CONFIG_SMP
/* Set the node_data pointer for each per-cpu struct */
for (cpu = 0; cpu < NR_CPUS; cpu++) {
node = node_cpuid[cpu].nid;
per_cpu(cpu_info, cpu).node_data = mem_data[node].node_data;
}
+#else
+ {
+ struct cpuinfo_ia64 *cpu0_cpu_info;
+ cpu = 0;
+ node = node_cpuid[cpu].nid;
+ cpu0_cpu_info = (struct cpuinfo_ia64 *)(__phys_per_cpu_start +
+ ((char *)&per_cpu__cpu_info - __per_cpu_start));
+ cpu0_cpu_info->node_data = mem_data[node].node_data;
+ }
+#endif /* CONFIG_SMP */
}
/**
@@ -464,25 +492,26 @@
find_initrd();
}
+#ifdef CONFIG_SMP
/**
* per_cpu_init - setup per-cpu variables
*
* find_pernode_space() does most of this already, we just need to set
* local_per_cpu_offset
*/
-void *per_cpu_init(void)
+void __init *per_cpu_init(void)
{
int cpu;
- if (smp_processor_id() == 0) {
- for (cpu = 0; cpu < NR_CPUS; cpu++) {
- per_cpu(local_per_cpu_offset, cpu) =
- __per_cpu_offset[cpu];
- }
- }
+ if (smp_processor_id() != 0)
+ return __per_cpu_start + __per_cpu_offset[smp_processor_id()];
+
+ for (cpu = 0; cpu < NR_CPUS; cpu++)
+ per_cpu(local_per_cpu_offset, cpu) = __per_cpu_offset[cpu];
return __per_cpu_start + __per_cpu_offset[smp_processor_id()];
}
+#endif /* CONFIG_SMP */
/**
* show_mem - give short summary of memory stats
@@ -533,7 +562,8 @@
* Take this opportunity to round the start address up and the end address
* down to page boundaries.
*/
-void call_pernode_memory(unsigned long start, unsigned long len, void *arg)
+void __init call_pernode_memory(unsigned long start, unsigned long len,
+ void *arg)
{
unsigned long rs, re, end = start + len;
void (*func)(unsigned long, unsigned long, int);
@@ -577,7 +607,8 @@
* for each piece of usable memory and will setup these values for each node.
* Very similar to build_maps().
*/
-static int count_node_pages(unsigned long start, unsigned long len, int node)
+static int __init count_node_pages(unsigned long start, unsigned long len,
+ int node)
{
unsigned long end = start + len;
@@ -602,7 +633,7 @@
* paging_init() sets up the page tables for each node of the system and frees
* the bootmem allocator memory for general use.
*/
-void paging_init(void)
+void __init paging_init(void)
{
unsigned long max_dma;
unsigned long zones_size[MAX_NR_ZONES];
diff -Nru a/arch/ia64/mm/numa.c b/arch/ia64/mm/numa.c
--- a/arch/ia64/mm/numa.c 2004-09-01 11:12:45 -07:00
+++ b/arch/ia64/mm/numa.c 2004-09-01 11:12:45 -07:00
@@ -24,13 +24,6 @@
static struct cpu *sysfs_cpus;
/*
- * The following structures are usually initialized by ACPI or
- * similar mechanisms and describe the NUMA characteristics of the machine.
- */
-int num_node_memblks;
-struct node_memblk_s node_memblk[NR_NODE_MEMBLKS];
-struct node_cpuid_s node_cpuid[NR_CPUS];
-/*
* This is a matrix with "distances" between nodes, they should be
* proportional to the memory access latency ratios.
*/
diff -Nru a/arch/ia64/sn/kernel/setup.c b/arch/ia64/sn/kernel/setup.c
--- a/arch/ia64/sn/kernel/setup.c 2004-09-01 11:12:45 -07:00
+++ b/arch/ia64/sn/kernel/setup.c 2004-09-01 11:12:45 -07:00
@@ -29,6 +29,7 @@
#include <linux/sched.h>
#include <linux/root_dev.h>
+#include <asm/acpi.h>
#include <asm/io.h>
#include <asm/sal.h>
#include <asm/machvec.h>
diff -Nru a/drivers/acpi/Kconfig b/drivers/acpi/Kconfig
--- a/drivers/acpi/Kconfig 2004-09-01 11:12:45 -07:00
+++ b/drivers/acpi/Kconfig 2004-09-01 11:12:45 -07:00
@@ -142,7 +142,6 @@
config ACPI_NUMA
bool "NUMA support"
depends on ACPI_INTERPRETER
- depends on NUMA
depends on IA64
default y if IA64_GENERIC || IA64_SGI_SN2
diff -Nru a/drivers/acpi/numa.c b/drivers/acpi/numa.c
--- a/drivers/acpi/numa.c 2004-09-01 11:12:45 -07:00
+++ b/drivers/acpi/numa.c 2004-09-01 11:12:45 -07:00
@@ -95,9 +95,7 @@
localities = (u32) slit->localities;
printk(KERN_INFO PREFIX "SLIT localities %ux%u\n", localities, localities);
-
acpi_numa_slit_init(slit);
-
return 0;
}
diff -Nru a/include/asm-ia64/acpi.h b/include/asm-ia64/acpi.h
--- a/include/asm-ia64/acpi.h 2004-09-01 11:12:45 -07:00
+++ b/include/asm-ia64/acpi.h 2004-09-01 11:12:45 -07:00
@@ -30,6 +30,7 @@
#ifdef __KERNEL__
+#include <linux/config.h>
#include <linux/init.h>
#include <linux/numa.h>
#include <asm/system.h>
diff -Nru a/include/asm-ia64/numa.h b/include/asm-ia64/numa.h
--- a/include/asm-ia64/numa.h 2004-09-01 11:12:45 -07:00
+++ b/include/asm-ia64/numa.h 2004-09-01 11:12:45 -07:00
@@ -12,9 +12,6 @@
#define _ASM_IA64_NUMA_H
#include <linux/config.h>
-
-#ifdef CONFIG_NUMA
-
#include <linux/cache.h>
#include <linux/cpumask.h>
#include <linux/numa.h>
@@ -58,17 +55,13 @@
* proportional to the memory access latency ratios.
*/
+#ifdef CONFIG_NUMA
extern u8 numa_slit[MAX_NUMNODES * MAX_NUMNODES];
#define node_distance(from,to) (numa_slit[(from) * numnodes + (to)])
-
extern int paddr_to_nid(unsigned long paddr);
-
#define local_nodeid (cpu_to_node_map[smp_processor_id()])
-
#else /* !CONFIG_NUMA */
-
#define paddr_to_nid(addr) 0
-
#endif /* CONFIG_NUMA */
#endif /* _ASM_IA64_NUMA_H */
diff -Nru a/include/asm-ia64/processor.h b/include/asm-ia64/processor.h
--- a/include/asm-ia64/processor.h 2004-09-01 11:12:45 -07:00
+++ b/include/asm-ia64/processor.h 2004-09-01 11:12:45 -07:00
@@ -88,9 +88,7 @@
#include <asm/rse.h>
#include <asm/unwind.h>
#include <asm/atomic.h>
-#ifdef CONFIG_NUMA
#include <asm/nodedata.h>
-#endif
/* like above but expressed as bitfields for more efficient access: */
struct ia64_psr {
@@ -168,7 +166,7 @@
__u8 archrev;
char vendor[16];
-#ifdef CONFIG_NUMA
+#ifdef CONFIG_DISCONTIGMEM
struct ia64_node_data *node_data;
#endif
};
diff -Nru a/include/asm-ia64/smp.h b/include/asm-ia64/smp.h
--- a/include/asm-ia64/smp.h 2004-09-01 11:12:45 -07:00
+++ b/include/asm-ia64/smp.h 2004-09-01 11:12:45 -07:00
@@ -126,6 +126,7 @@
#else
#define cpu_logical_id(cpuid) 0
+#define cpu_physical_id(i) ((ia64_getreg(_IA64_REG_CR_LID) >> 16) & 0xffff)
#endif /* CONFIG_SMP */
#endif /* _ASM_IA64_SMP_H */
diff -Nru a/include/asm-ia64/sn/sn_cpuid.h b/include/asm-ia64/sn/sn_cpuid.h
--- a/include/asm-ia64/sn/sn_cpuid.h 2004-09-01 11:12:45 -07:00
+++ b/include/asm-ia64/sn/sn_cpuid.h 2004-09-01 11:12:45 -07:00
@@ -83,10 +83,6 @@
*
*/
-#ifndef CONFIG_SMP
-#define cpu_physical_id(cpuid) ((ia64_getreg(_IA64_REG_CR_LID) >> 16) & 0xffff)
-#endif
-
/*
* macros for some of these exist in sn/addrs.h & sn/arch.h, etc. However,
* trying #include these files here causes circular dependencies.
diff -Nru a/include/linux/acpi.h b/include/linux/acpi.h
--- a/include/linux/acpi.h 2004-09-01 11:12:45 -07:00
+++ b/include/linux/acpi.h 2004-09-01 11:12:45 -07:00
@@ -391,7 +391,11 @@
void acpi_table_print_srat_entry (acpi_table_entry_header *srat);
/* the following four functions are architecture-dependent */
+#ifdef CONFIG_NUMA
void acpi_numa_slit_init (struct acpi_table_slit *slit);
+#else
+static inline void acpi_numa_slit_init(struct acpi_table_slit *slit) { }
+#endif
void acpi_numa_processor_affinity_init (struct acpi_table_processor_affinity *pa);
void acpi_numa_memory_affinity_init (struct acpi_table_memory_affinity *ma);
void acpi_numa_arch_fixup(void);
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2004-09-01 18:14 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-09-01 0:44 [PATCH] CONFIG_SMP=n fixes redux Jesse Barnes
2004-09-01 18:14 ` Jesse Barnes
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox