Hello.
I rewrote patches to unify mapping from pxm to node id as ver.3.
In previous patches, I moved MAX_PXM_DOMAINS into
include/acpi/acpi_numa.h to unify pxm_to_node mapping.
Its max number was 256.
However, ACPI spec ver.3 defines pxm's extension. So, pxm can be over 256.
256 was not enough for maximum of it, and u8 was not good for pxm's
definition.
In addition, SGI's SN2 already uses its extension in 2.6.16-git14,
and MAX_PXM_DOMAINS was defined by CONFIG_IA64_NR_NODES
in include/asm-ia64/acpi.h.
I defined CONFIG_NR_NODES for common definition.
This patches are for 2.6.16-git14.
I tested them on ia64(Tiger4) with node emulation.
And I confirmed no compile error against ....
- x86-64
- i386 with summit config.
- ia64's SN2 config.
Please apply.
Thanks.
------------------------
Change log from ver.2
- update for 2.6.16-git14.
- definition of pxm was changed from u8 to int. Pxm can be over 256.
- CONFIG_NR_NODES is defined to configure MAX_PXM_DOMAINS.
- redundant call of pxm_bit_set() is removed at acpi_numa_arch_fixup()
of ia64 like followings. :-P
if (pxm_bit_test(i)) {
:
pxm_bit_set(i); <---------------------- !!!
:
}
Change log from ver.1
- Fix old map from HP and SGI's code by Bob Picco-san.
- Remove MAX_PXM_DOMAINS from asm-ia64/acpi.h. It is already defined at
include/acpi/acpi_numa.h.
- Fix return code of setup_node() at arch/x86_64/mm/srat.c
- Fix ACPI_NUMA config for i386 by Andy Witcroft-san.
- Define dummy functions for i386's compile error.
- Remove garbage nid_to_pxm_map from acpi20_parse_srat()
at arch/i386/kernel/srat.c
----------------------------------
Description.
This patch is to unify mapping from pxm to node id.
In current code, i386, x86-64, and ia64 have its mapping by each own code.
But PXM is defined by ACPI and node id is used generically. So,
I think there is no reason to define it on each arch's code.
This mapping should be written at drivers/acpi/numa.c as a common code.
--
Yasunori Goto
This is to use generic pxm_to_node() function instead of
old pxm_to_nid_map for ia64. And change old CONFIG_IA_NR_NODES
to common one.
Signed-off-by: Yasunori Goto <[email protected]>
arch/ia64/Kconfig | 9 ++-------
arch/ia64/hp/common/sba_iommu.c | 2 +-
arch/ia64/kernel/acpi.c | 24 ++++++++----------------
arch/ia64/pci/pci.c | 2 +-
arch/ia64/sn/kernel/setup.c | 4 ++--
include/asm-ia64/acpi.h | 11 -----------
include/asm-ia64/numnodes.h | 6 +++---
7 files changed, 17 insertions(+), 41 deletions(-)
Index: pxm_ver3/arch/ia64/kernel/acpi.c
===================================================================
--- pxm_ver3.orig/arch/ia64/kernel/acpi.c 2006-03-28 14:24:40.091383224 +0900
+++ pxm_ver3/arch/ia64/kernel/acpi.c 2006-03-28 14:25:34.960523177 +0900
@@ -415,9 +415,6 @@ static int __initdata srat_num_cpus; /*
static u32 __devinitdata pxm_flag[PXM_FLAG_LEN];
#define pxm_bit_set(bit) (set_bit(bit,(void *)pxm_flag))
#define pxm_bit_test(bit) (test_bit(bit,(void *)pxm_flag))
-/* maps to convert between proximity domain and logical node ID */
-int __devinitdata pxm_to_nid_map[MAX_PXM_DOMAINS];
-int __initdata nid_to_pxm_map[MAX_NUMNODES];
static struct acpi_table_slit __initdata *slit_table;
static int get_processor_proximity_domain(struct acpi_table_processor_affinity *pa)
@@ -533,22 +530,17 @@ void __init acpi_numa_arch_fixup(void)
* MCD - This can probably be dropped now. No need for pxm ID to node ID
* mapping with sparse node numbering iff MAX_PXM_DOMAINS <= MAX_NUMNODES.
*/
- /* calculate total number of nodes in system from PXM bitmap */
- memset(pxm_to_nid_map, -1, sizeof(pxm_to_nid_map));
- memset(nid_to_pxm_map, -1, sizeof(nid_to_pxm_map));
nodes_clear(node_online_map);
for (i = 0; i < MAX_PXM_DOMAINS; i++) {
if (pxm_bit_test(i)) {
- int nid = num_online_nodes();
- pxm_to_nid_map[i] = nid;
- nid_to_pxm_map[nid] = i;
+ int nid = acpi_map_pxm_to_node(i);
node_set_online(nid);
}
}
/* set logical node id in memory chunk structure */
for (i = 0; i < num_node_memblks; i++)
- node_memblk[i].nid = pxm_to_nid_map[node_memblk[i].nid];
+ node_memblk[i].nid = pxm_to_node(node_memblk[i].nid);
/* assign memory bank numbers for each chunk on each node */
for_each_online_node(i) {
@@ -562,7 +554,7 @@ void __init acpi_numa_arch_fixup(void)
/* set logical node id in cpu structure */
for (i = 0; i < srat_num_cpus; i++)
- node_cpuid[i].nid = pxm_to_nid_map[node_cpuid[i].nid];
+ node_cpuid[i].nid = pxm_to_node(node_cpuid[i].nid);
printk(KERN_INFO "Number of logical nodes in system = %d\n",
num_online_nodes());
@@ -575,11 +567,11 @@ void __init acpi_numa_arch_fixup(void)
for (i = 0; i < slit_table->localities; i++) {
if (!pxm_bit_test(i))
continue;
- node_from = pxm_to_nid_map[i];
+ node_from = pxm_to_node(i);
for (j = 0; j < slit_table->localities; j++) {
if (!pxm_bit_test(j))
continue;
- node_to = pxm_to_nid_map[j];
+ node_to = pxm_to_node(j);
node_distance(node_from, node_to) =
slit_table->entry[i * slit_table->localities + j];
}
@@ -785,9 +777,9 @@ int acpi_map_cpu2node(acpi_handle handle
/*
* Assuming that the container driver would have set the proximity
- * domain and would have initialized pxm_to_nid_map[pxm_id] && pxm_flag
+ * domain and would have initialized pxm_to_node(pxm_id) && pxm_flag
*/
- node_cpuid[cpu].nid = (pxm_id < 0) ? 0 : pxm_to_nid_map[pxm_id];
+ node_cpuid[cpu].nid = (pxm_id < 0) ? 0 : pxm_to_node(pxm_id);
node_cpuid[cpu].phys_id = physid;
#endif
@@ -966,7 +958,7 @@ acpi_map_iosapic(acpi_handle handle, u32
if (pxm < 0)
return AE_OK;
- node = pxm_to_nid_map[pxm];
+ node = pxm_to_node(pxm);
if (node >= MAX_NUMNODES || !node_online(node) ||
cpus_empty(node_to_cpumask(node)))
Index: pxm_ver3/arch/ia64/pci/pci.c
===================================================================
--- pxm_ver3.orig/arch/ia64/pci/pci.c 2006-03-28 14:24:40.091383224 +0900
+++ pxm_ver3/arch/ia64/pci/pci.c 2006-03-28 14:25:34.961499739 +0900
@@ -353,7 +353,7 @@ pci_acpi_scan_root(struct acpi_device *d
pxm = acpi_get_pxm(controller->acpi_handle);
#ifdef CONFIG_NUMA
if (pxm >= 0)
- controller->node = pxm_to_nid_map[pxm];
+ controller->node = pxm_to_node(pxm);
#endif
acpi_walk_resources(device->handle, METHOD_NAME__CRS, count_window,
Index: pxm_ver3/arch/ia64/hp/common/sba_iommu.c
===================================================================
--- pxm_ver3.orig/arch/ia64/hp/common/sba_iommu.c 2006-03-28 14:24:40.092359787 +0900
+++ pxm_ver3/arch/ia64/hp/common/sba_iommu.c 2006-03-28 14:25:34.962476302 +0900
@@ -1958,7 +1958,7 @@ sba_map_ioc_to_node(struct ioc *ioc, acp
if (pxm < 0)
return;
- node = pxm_to_nid_map[pxm];
+ node = pxm_to_node(pxm);
if (node >= MAX_NUMNODES || !node_online(node))
return;
Index: pxm_ver3/arch/ia64/sn/kernel/setup.c
===================================================================
--- pxm_ver3.orig/arch/ia64/sn/kernel/setup.c 2006-03-28 14:24:40.092359787 +0900
+++ pxm_ver3/arch/ia64/sn/kernel/setup.c 2006-03-28 14:25:34.963452864 +0900
@@ -139,7 +139,7 @@ static int __init pxm_to_nasid(int pxm)
int i;
int nid;
- nid = pxm_to_nid_map[pxm];
+ nid = pxm_to_node(pxm);
for (i = 0; i < num_node_memblks; i++) {
if (node_memblk[i].nid == nid) {
return NASID_GET(node_memblk[i].start_paddr);
@@ -704,7 +704,7 @@ void __init build_cnode_tables(void)
* cnode == node for all C & M bricks.
*/
for_each_online_node(node) {
- nasid = pxm_to_nasid(nid_to_pxm_map[node]);
+ nasid = pxm_to_nasid(node_to_pxm(node));
sn_cnodeid_to_nasid[node] = nasid;
physical_node_map[nasid] = node;
}
Index: pxm_ver3/include/asm-ia64/acpi.h
===================================================================
--- pxm_ver3.orig/include/asm-ia64/acpi.h 2006-03-28 14:24:40.092359787 +0900
+++ pxm_ver3/include/asm-ia64/acpi.h 2006-03-28 14:25:34.963452864 +0900
@@ -109,17 +109,6 @@ extern unsigned int get_cpei_target_cpu(
extern void prefill_possible_map(void);
extern int additional_cpus;
-#ifdef CONFIG_ACPI_NUMA
-/* Proximity bitmap length; _PXM is at most 255 (8 bit)*/
-#ifdef CONFIG_IA64_NR_NODES
-#define MAX_PXM_DOMAINS CONFIG_IA64_NR_NODES
-#else
-#define MAX_PXM_DOMAINS (256)
-#endif
-extern int __devinitdata pxm_to_nid_map[MAX_PXM_DOMAINS];
-extern int __initdata nid_to_pxm_map[MAX_NUMNODES];
-#endif
-
extern u16 ia64_acpiid_to_sapicid[];
/*
Index: pxm_ver3/arch/ia64/Kconfig
===================================================================
--- pxm_ver3.orig/arch/ia64/Kconfig 2006-03-28 14:27:36.475170126 +0900
+++ pxm_ver3/arch/ia64/Kconfig 2006-03-28 14:32:05.280830896 +0900
@@ -260,14 +260,9 @@ config NR_CPUS
than 64 will cause the use of a CPU mask array, causing a small
performance hit.
-config IA64_NR_NODES
- int "Maximum number of NODEs (256-1024)" if (IA64_SGI_SN2 || IA64_GENERIC)
- range 256 1024
+config NR_NODES_CHANGABLE
+ def_bool y
depends on IA64_SGI_SN2 || IA64_GENERIC
- default "256"
- help
- This option specifies the maximum number of nodes in your SSI system.
- If in doubt, use the default.
config HOTPLUG_CPU
bool "Support for hot-pluggable CPUs (EXPERIMENTAL)"
Index: pxm_ver3/include/asm-ia64/numnodes.h
===================================================================
--- pxm_ver3.orig/include/asm-ia64/numnodes.h 2006-03-28 14:27:36.475170126 +0900
+++ pxm_ver3/include/asm-ia64/numnodes.h 2006-03-28 14:33:03.073798938 +0900
@@ -8,11 +8,11 @@
/* Max 32 Nodes */
# define NODES_SHIFT 5
#elif defined(CONFIG_IA64_SGI_SN2) || defined(CONFIG_IA64_GENERIC)
-# if CONFIG_IA64_NR_NODES == 256
+# if CONFIG_NR_NODES == 256
# define NODES_SHIFT 8
-# elif CONFIG_IA64_NR_NODES <= 512
+# elif CONFIG_NR_NODES <= 512
# define NODES_SHIFT 9
-# elif CONFIG_IA64_NR_NODES <= 1024
+# elif CONFIG_NR_NODES <= 1024
# define NODES_SHIFT 10
# endif
#endif
--
Yasunori Goto
This is new generic code for pxm_to_node_map and CONFIG_NR_NODES.
Signed-off-by: Yasunori Goto <[email protected]>
drivers/acpi/numa.c | 48 +++++++++++++++++++++++++++++++++++++++++++++++
include/acpi/acpi_numa.h | 23 ++++++++++++++++++++++
include/linux/acpi.h | 1
mm/Kconfig | 12 +++++++++++
4 files changed, 84 insertions(+)
Index: pxm_ver3/drivers/acpi/numa.c
===================================================================
--- pxm_ver3.orig/drivers/acpi/numa.c 2006-03-28 14:10:02.867761158 +0900
+++ pxm_ver3/drivers/acpi/numa.c 2006-03-28 14:13:30.926352359 +0900
@@ -36,12 +36,60 @@
#define _COMPONENT ACPI_NUMA
ACPI_MODULE_NAME("numa")
+static nodemask_t nodes_found_map = NODE_MASK_NONE;
+#define PXM_INVAL -1
+#define NID_INVAL -1
+
+/* maps to convert between proximity domain and logical node ID */
+int __cpuinitdata pxm_to_node_map[MAX_PXM_DOMAINS]
+ = { [0 ... MAX_PXM_DOMAINS - 1] = NID_INVAL };
+int __cpuinitdata node_to_pxm_map[MAX_NUMNODES]
+ = { [0 ... MAX_NUMNODES - 1] = PXM_INVAL };
+
extern int __init acpi_table_parse_madt_family(enum acpi_table_id id,
unsigned long madt_size,
int entry_id,
acpi_madt_entry_handler handler,
unsigned int max_entries);
+int __cpuinit pxm_to_node(int pxm)
+{
+ if (pxm < 0)
+ return NID_INVAL;
+ return pxm_to_node_map[pxm];
+}
+
+int __cpuinit node_to_pxm(int node)
+{
+ if (node < 0)
+ return PXM_INVAL;
+ return node_to_pxm_map[node];
+}
+
+int __cpuinit acpi_map_pxm_to_node(int pxm)
+{
+ int node = pxm_to_node_map[pxm];
+
+ if (node < 0){
+ if (nodes_weight(nodes_found_map) >= MAX_NUMNODES)
+ return NID_INVAL;
+ node = first_unset_node(nodes_found_map);
+ pxm_to_node_map[pxm] = node;
+ node_to_pxm_map[node] = pxm;
+ node_set(node, nodes_found_map);
+ }
+
+ return node;
+}
+
+void __cpuinit acpi_unmap_pxm_to_node(int node)
+{
+ int pxm = node_to_pxm_map[node];
+ pxm_to_node_map[pxm] = NID_INVAL;
+ node_to_pxm_map[node] = PXM_INVAL;
+ node_clear(node, nodes_found_map);
+}
+
void __init acpi_table_print_srat_entry(acpi_table_entry_header * header)
{
Index: pxm_ver3/include/acpi/acpi_numa.h
===================================================================
--- /dev/null 1970-01-01 00:00:00.000000000 +0000
+++ pxm_ver3/include/acpi/acpi_numa.h 2006-03-28 14:13:30.927328921 +0900
@@ -0,0 +1,23 @@
+#ifndef __ACPI_NUMA_H
+#define __ACPI_NUMA_H
+
+#ifdef CONFIG_ACPI_NUMA
+#include <linux/kernel.h>
+
+/* Proximity bitmap length */
+#ifdef CONFIG_NR_NODES_CHANGABLE
+#define MAX_PXM_DOMAINS CONFIG_NR_NODES
+#else
+#define MAX_PXM_DOMAINS (256)
+#endif
+
+extern int __cpuinitdata pxm_to_node_map[MAX_PXM_DOMAINS];
+extern int __cpuinitdata node_to_pxm_map[MAX_NUMNODES];
+
+extern int __cpuinit pxm_to_node(int);
+extern int __cpuinit node_to_pxm(int);
+extern int __cpuinit acpi_map_pxm_to_node(int);
+extern void __cpuinit acpi_unmap_pxm_to_node(int);
+
+#endif /* CONFIG_ACPI_NUMA */
+#endif /* __ACP_NUMA_H */
Index: pxm_ver3/include/linux/acpi.h
===================================================================
--- pxm_ver3.orig/include/linux/acpi.h 2006-03-28 14:10:02.867761158 +0900
+++ pxm_ver3/include/linux/acpi.h 2006-03-28 14:24:38.740797303 +0900
@@ -38,6 +38,7 @@
#include <acpi/acpi.h>
#include <acpi/acpi_bus.h>
#include <acpi/acpi_drivers.h>
+#include <acpi/acpi_numa.h>
#include <asm/acpi.h>
Index: pxm_ver3/mm/Kconfig
===================================================================
--- pxm_ver3.orig/mm/Kconfig 2006-03-28 14:24:38.009352000 +0900
+++ pxm_ver3/mm/Kconfig 2006-03-28 14:24:53.320875250 +0900
@@ -91,6 +91,18 @@ config HAVE_MEMORY_PRESENT
depends on ARCH_HAVE_MEMORY_PRESENT || SPARSEMEM
#
+# NR_NODES is to configure NODES_SHIFT
+#
+config NR_NODES
+ int "Maximum number of NODEs (256-1024)"
+ range 256 1024
+ depends on NEED_MULTIPLE_NODES && NR_NODES_CHANGABLE
+ default "256"
+ help
+ This option specifies the maximum number of nodes in your SSI system.
+ If in doubt, use the default.
+
+#
# SPARSEMEM_EXTREME (which is the default) does some bootmem
# allocations when memory_present() is called. If this can not
# be done on your architecture, select this option. However,
--
Yasunori Goto
This is to remove the code of pxm_to_node from x86-64 code.
Signed-off-by: Yasunori Goto <[email protected]>
arch/x86_64/mm/srat.c | 33 +--------------------------------
include/asm-x86_64/numa.h | 1 -
2 files changed, 1 insertion(+), 33 deletions(-)
Index: pxm_ver3/arch/x86_64/mm/srat.c
===================================================================
--- pxm_ver3.orig/arch/x86_64/mm/srat.c 2006-03-27 12:09:14.000000000 +0900
+++ pxm_ver3/arch/x86_64/mm/srat.c 2006-03-27 14:07:14.000000000 +0900
@@ -22,35 +22,15 @@
static struct acpi_table_slit *acpi_slit;
static nodemask_t nodes_parsed __initdata;
-static nodemask_t nodes_found __initdata;
static struct bootnode nodes[MAX_NUMNODES] __initdata;
-static u8 pxm2node[256] = { [0 ... 255] = 0xff };
/* Too small nodes confuse the VM badly. Usually they result
from BIOS bugs. */
#define NODE_MIN_SIZE (4*1024*1024)
-static int node_to_pxm(int n);
-
-int pxm_to_node(int pxm)
-{
- if ((unsigned)pxm >= 256)
- return -1;
- /* Extend 0xff to (int)-1 */
- return (signed char)pxm2node[pxm];
-}
-
static __init int setup_node(int pxm)
{
- unsigned node = pxm2node[pxm];
- if (node == 0xff) {
- if (nodes_weight(nodes_found) >= MAX_NUMNODES)
- return -1;
- node = first_unset_node(nodes_found);
- node_set(node, nodes_found);
- pxm2node[pxm] = node;
- }
- return pxm2node[pxm];
+ return acpi_map_pxm_to_node(pxm);
}
static __init int conflicting_nodes(unsigned long start, unsigned long end)
@@ -292,17 +272,6 @@ int __init acpi_scan_nodes(unsigned long
return 0;
}
-static int node_to_pxm(int n)
-{
- int i;
- if (pxm2node[n] == n)
- return n;
- for (i = 0; i < 256; i++)
- if (pxm2node[i] == n)
- return i;
- return 0;
-}
-
int __node_distance(int a, int b)
{
int index;
Index: pxm_ver3/include/asm-x86_64/numa.h
===================================================================
--- pxm_ver3.orig/include/asm-x86_64/numa.h 2006-03-27 12:09:17.000000000 +0900
+++ pxm_ver3/include/asm-x86_64/numa.h 2006-03-27 14:05:12.000000000 +0900
@@ -9,7 +9,6 @@ struct bootnode {
};
extern int compute_hash_shift(struct bootnode *nodes, int numnodes);
-extern int pxm_to_node(int nid);
#define ZONE_ALIGN (1UL << (MAX_ORDER+PAGE_SHIFT))
--
Yasunori Goto
This is to remove the code of pxm_to_nid_map from i386 code.
And, some of changing Kconfig and dummy function for compile.
Signed-off-by: Yasunori Goto <[email protected]>
arch/i386/Kconfig | 6 ++++++
arch/i386/kernel/srat.c | 19 ++-----------------
drivers/acpi/Kconfig | 2 +-
include/linux/acpi.h | 8 ++++++++
4 files changed, 17 insertions(+), 18 deletions(-)
Index: pxm_ver3/arch/i386/kernel/srat.c
===================================================================
--- pxm_ver3.orig/arch/i386/kernel/srat.c 2006-01-05 15:43:10.000000000 +0900
+++ pxm_ver3/arch/i386/kernel/srat.c 2006-03-27 14:08:19.000000000 +0900
@@ -39,7 +39,6 @@
#define NODE_ARRAY_OFFSET(x) ((x) % 8) /* 8 bits/char */
#define BMAP_SET(bmap, bit) ((bmap)[NODE_ARRAY_INDEX(bit)] |= 1 << NODE_ARRAY_OFFSET(bit))
#define BMAP_TEST(bmap, bit) ((bmap)[NODE_ARRAY_INDEX(bit)] & (1 << NODE_ARRAY_OFFSET(bit)))
-#define MAX_PXM_DOMAINS 256 /* 1 byte and no promises about values */
/* bitmap length; _PXM is at most 255 */
#define PXM_BITMAP_LEN (MAX_PXM_DOMAINS / 8)
static u8 pxm_bitmap[PXM_BITMAP_LEN]; /* bitmap of proximity domains */
@@ -213,19 +212,11 @@ static __init void node_read_chunk(int n
node_end_pfn[nid] = memory_chunk->end_pfn;
}
-static u8 pxm_to_nid_map[MAX_PXM_DOMAINS];/* _PXM to logical node ID map */
-
-int pxm_to_node(int pxm)
-{
- return pxm_to_nid_map[pxm];
-}
-
/* Parse the ACPI Static Resource Affinity Table */
static int __init acpi20_parse_srat(struct acpi_table_srat *sratp)
{
u8 *start, *end, *p;
int i, j, nid;
- u8 nid_to_pxm_map[MAX_NUMNODES];/* logical node ID to _PXM map */
start = (u8 *)(&(sratp->reserved) + 1); /* skip header */
p = start;
@@ -235,10 +226,6 @@ static int __init acpi20_parse_srat(stru
memset(node_memory_chunk, 0, sizeof(node_memory_chunk));
memset(zholes_size, 0, sizeof(zholes_size));
- /* -1 in these maps means not available */
- memset(pxm_to_nid_map, -1, sizeof(pxm_to_nid_map));
- memset(nid_to_pxm_map, -1, sizeof(nid_to_pxm_map));
-
num_memory_chunks = 0;
while (p < end) {
switch (*p) {
@@ -278,9 +265,7 @@ static int __init acpi20_parse_srat(stru
nodes_clear(node_online_map);
for (i = 0; i < MAX_PXM_DOMAINS; i++) {
if (BMAP_TEST(pxm_bitmap, i)) {
- nid = num_online_nodes();
- pxm_to_nid_map[i] = nid;
- nid_to_pxm_map[nid] = i;
+ int nid = acpi_map_pxm_to_node(i);
node_set_online(nid);
}
}
@@ -288,7 +273,7 @@ static int __init acpi20_parse_srat(stru
/* set cnode id in memory chunk structure */
for (i = 0; i < num_memory_chunks; i++)
- node_memory_chunk[i].nid = pxm_to_nid_map[node_memory_chunk[i].pxm];
+ node_memory_chunk[i].nid = pxm_to_node(node_memory_chunk[i].pxm);
printk("pxm bitmap: ");
for (i = 0; i < sizeof(pxm_bitmap); i++) {
Index: pxm_ver3/arch/i386/Kconfig
===================================================================
--- pxm_ver3.orig/arch/i386/Kconfig 2006-03-27 12:09:13.000000000 +0900
+++ pxm_ver3/arch/i386/Kconfig 2006-03-27 14:08:19.000000000 +0900
@@ -144,6 +144,12 @@ config ACPI_SRAT
bool
default y
depends on NUMA && (X86_SUMMIT || X86_GENERICARCH)
+ select ACPI_NUMA
+
+config HAVE_ARCH_PARSE_SRAT
+ bool
+ default y
+ depends on ACPI_SRAT
config X86_SUMMIT_NUMA
bool
Index: pxm_ver3/drivers/acpi/Kconfig
===================================================================
--- pxm_ver3.orig/drivers/acpi/Kconfig 2006-03-27 12:09:14.000000000 +0900
+++ pxm_ver3/drivers/acpi/Kconfig 2006-03-27 14:08:19.000000000 +0900
@@ -162,7 +162,7 @@ config ACPI_THERMAL
config ACPI_NUMA
bool "NUMA support"
depends on NUMA
- depends on (IA64 || X86_64)
+ depends on (X86_32 || IA64 || X86_64)
default y if IA64_GENERIC || IA64_SGI_SN2
config ACPI_ASUS
Index: pxm_ver3/include/linux/acpi.h
===================================================================
--- pxm_ver3.orig/include/linux/acpi.h 2006-03-27 13:56:46.000000000 +0900
+++ pxm_ver3/include/linux/acpi.h 2006-03-27 14:08:19.000000000 +0900
@@ -409,10 +409,18 @@ void acpi_table_print_madt_entry (acpi_t
void acpi_table_print_srat_entry (acpi_table_entry_header *srat);
/* the following four functions are architecture-dependent */
+#ifdef CONFIG_HAVE_ARCH_PARSE_SRAT
+#define NR_NODE_MEMBLKS MAX_NUMNODES
+#define acpi_numa_slit_init(slit) do {} while (0)
+#define acpi_numa_processor_affinity_init(pa) do {} while (0)
+#define acpi_numa_memory_affinity_init(ma) do {} while (0)
+#define acpi_numa_arch_fixup() do {} while (0)
+#else
void acpi_numa_slit_init (struct acpi_table_slit *slit);
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);
+#endif
#ifdef CONFIG_ACPI_HOTPLUG_CPU
/* Arch dependent functions for cpu hotplug support */
--
Yasunori Goto
Yasunori Goto <[email protected]> wrote:
>
> +/* Proximity bitmap length */
> +#ifdef CONFIG_NR_NODES_CHANGABLE
> +#define MAX_PXM_DOMAINS CONFIG_NR_NODES
> +#else
> +#define MAX_PXM_DOMAINS (256)
> +#endif
I don't think we need CONFIG_NR_NODES_CHANGABLE (it is spelled
"changeable", btw).
If the architecture wants to support changing of CONFIG_NR_NODES then it
can permit CONFIG_NR_NODES to be changed in its Kconfig implementation.
If the architecture doesn't want to permit changing of CONFIG_NR_NODES
then it should simply hardwire CONFIG_NR_NODES to the chosen value in
its Kconfig.
So all architectures which use acpi_numa must implement CONFIG_NR_NODES.
In fact, it would probably make sense to require that all NUMA-supporting
archtectures implement CONFIG_NR_NODES.
Also, we already have NODES_SHIFT defined in include/asm-*/numnodes.h.
What's the relationship between that and CONFIG_NR_NODES? It seems that we
want to derive NODES_SHIFT from CONFIG_NR_NODES.
Was ia64's CONFIG_IA64_NR_NODES the best choice? Should ia64 instead have
made NODES_SHIFT Kconfigurable, and derived its max-nr_nodes from that?
It's all a bit of a pickle.
I guess for now a suitable approach would be to make all numa-using
architectures define CONFIG_NR_NODES, and to leave that rather
unpleasant-looking code in include/asm-ia64/numnodes.h as it is.
> Yasunori Goto <[email protected]> wrote:
> >
> > +/* Proximity bitmap length */
> > +#ifdef CONFIG_NR_NODES_CHANGABLE
> > +#define MAX_PXM_DOMAINS CONFIG_NR_NODES
> > +#else
> > +#define MAX_PXM_DOMAINS (256)
> > +#endif
>
> I don't think we need CONFIG_NR_NODES_CHANGABLE (it is spelled
> "changeable", btw).
>
> If the architecture wants to support changing of CONFIG_NR_NODES then it
> can permit CONFIG_NR_NODES to be changed in its Kconfig implementation.
>
> If the architecture doesn't want to permit changing of CONFIG_NR_NODES
> then it should simply hardwire CONFIG_NR_NODES to the chosen value in
> its Kconfig.
>
> So all architectures which use acpi_numa must implement CONFIG_NR_NODES.
>
> In fact, it would probably make sense to require that all NUMA-supporting
> archtectures implement CONFIG_NR_NODES.
>
> Also, we already have NODES_SHIFT defined in include/asm-*/numnodes.h.
> What's the relationship between that and CONFIG_NR_NODES? It seems that we
> want to derive NODES_SHIFT from CONFIG_NR_NODES.
>
> Was ia64's CONFIG_IA64_NR_NODES the best choice? Should ia64 instead have
> made NODES_SHIFT Kconfigurable, and derived its max-nr_nodes from that?
>
> It's all a bit of a pickle.
>
>
> I guess for now a suitable approach would be to make all numa-using
> architectures define CONFIG_NR_NODES, and to leave that rather
> unpleasant-looking code in include/asm-ia64/numnodes.h as it is.
>
Ahhh.
I understand what you wish at last.
I thought relationship between pxm and nid is just acpi-using
architecture's issue.
But, it becomes for all numa-using architecture's issue.
Ok. I'll change it.
Thanks.
--
Yasunori Goto