[Patch 000/003] pgdat allocation and update for ia64 of memory hotplug.

[Patch 000/003] pgdat allocation and update for ia64 of memory hotplug.

[Yasunori Goto]

Hello.

These are parts of patches for new nodes addition v4.
When new node is added, new pgdat must be allocated and initialized.
But, ia64 has copies of node_data[] on each node. So, kernel has to
allocate not only pgdat but also its copies area. and all of copies
must be updated at hot-add. These are patches for it.

This patch is for 2.6.17-rc3-mm1.

Please apply.

------------------------------------------------------------

Change log from v4 of node hot-add.
  - update for 2.6.17-rc3-mm1.

V4 of post is here.
<description>
http://marc.theaimsgroup.com/?l=linux-mm&m=114258404023573&w=2
<patches>
http://marc.theaimsgroup.com/?l=linux-mm&w=2&r=1&s=memory+hotplug+node+v.4.&q=b

-- 
Yasunori Goto 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[Patch 001/003] pgdat allocation and update for ia64 of memory hotplug.(hold pgdat address at system running)

[Yasunori Goto]

This is preparing patch to make common code for updating of NODE_DATA()
of ia64 between boottime and hotplug.

Current code remembers pgdat address in mem_data which is used at just boot
time. But its information can be used at hotplug time
by moving to global value.
The next patche use this array.


Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>

 arch/ia64/mm/discontig.c |   19 ++++++++-----------
 1 files changed, 8 insertions(+), 11 deletions(-)

Index: pgdat11/arch/ia64/mm/discontig.c
===================================================================
--- pgdat11.orig/arch/ia64/mm/discontig.c	2006-04-20 11:00:04.000000000 +0900
+++ pgdat11/arch/ia64/mm/discontig.c	2006-04-20 11:00:46.000000000 +0900
@@ -33,7 +33,6 @@
  */
 struct early_node_data {
 	struct ia64_node_data *node_data;
-	pg_data_t *pgdat;
 	unsigned long pernode_addr;
 	unsigned long pernode_size;
 	struct bootmem_data bootmem_data;
@@ -46,6 +45,8 @@ struct early_node_data {
 static struct early_node_data mem_data[MAX_NUMNODES] __initdata;
 static nodemask_t memory_less_mask __initdata;
 
+static pg_data_t *pgdat_list[MAX_NUMNODES];
+
 /*
  * To prevent cache aliasing effects, align per-node structures so that they
  * start at addresses that are strided by node number.
@@ -175,13 +176,13 @@ static void __init fill_pernode(int node
 	pernode += PERCPU_PAGE_SIZE * cpus;
 	pernode += node * L1_CACHE_BYTES;
 
-	mem_data[node].pgdat = __va(pernode);
+	pgdat_list[node] = __va(pernode);
 	pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
 
 	mem_data[node].node_data = __va(pernode);
 	pernode += L1_CACHE_ALIGN(sizeof(struct ia64_node_data));
 
-	mem_data[node].pgdat->bdata = bdp;
+	pgdat_list[node]->bdata = bdp;
 	pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
 
 	cpu_data = per_cpu_node_setup(cpu_data, node);
@@ -268,7 +269,7 @@ static int __init find_pernode_space(uns
 static int __init free_node_bootmem(unsigned long start, unsigned long len,
 				    int node)
 {
-	free_bootmem_node(mem_data[node].pgdat, start, len);
+	free_bootmem_node(pgdat_list[node], start, len);
 
 	return 0;
 }
@@ -287,7 +288,7 @@ static void __init reserve_pernode_space
 	int node;
 
 	for_each_online_node(node) {
-		pg_data_t *pdp = mem_data[node].pgdat;
+		pg_data_t *pdp = pgdat_list[node];
 
 		if (node_isset(node, memory_less_mask))
 			continue;
@@ -317,12 +318,8 @@ static void __init reserve_pernode_space
  */
 static void __init initialize_pernode_data(void)
 {
-	pg_data_t *pgdat_list[MAX_NUMNODES];
 	int cpu, node;
 
-	for_each_online_node(node)
-		pgdat_list[node] = mem_data[node].pgdat;
-
 	/* Copy the pg_data_t list to each node and init the node field */
 	for_each_online_node(node) {
 		memcpy(mem_data[node].node_data->pg_data_ptrs, pgdat_list,
@@ -372,7 +369,7 @@ static void __init *memory_less_node_all
 	if (bestnode == -1)
 		bestnode = anynode;
 
-	ptr = __alloc_bootmem_node(mem_data[bestnode].pgdat, pernodesize,
+	ptr = __alloc_bootmem_node(pgdat_list[bestnode], pernodesize,
 		PERCPU_PAGE_SIZE, __pa(MAX_DMA_ADDRESS));
 
 	return ptr;
@@ -476,7 +473,7 @@ void __init find_memory(void)
 		pernodesize = mem_data[node].pernode_size;
 		map = pernode + pernodesize;
 
-		init_bootmem_node(mem_data[node].pgdat,
+		init_bootmem_node(pgdat_list[node],
 				  map>>PAGE_SHIFT,
 				  bdp->node_boot_start>>PAGE_SHIFT,
 				  bdp->node_low_pfn);

-- 
Yasunori Goto 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[Patch 002/003] pgdat allocation and update for ia64 of memory hotplug. (update pgdat address array)

[Yasunori Goto]

As I mentioned previous patches,
ia64 has copies of information of pgdat address array on each node
as per node data.

At v2 of node_add, this function used stop_machine_run() to update them.
(I wished that they were copied safety as much as possible.)
But, in this patch, this arrays are just copied simply, and
set node_online_map bit after completion of pgdat initialization.

So, kernel must touch NODE_DATA() macro after checking 
node_online_map(). (Current code has already done it.)
This is more simple way for just hot-add.....

Note : It will be problem when hot-remove will occur,
       because, even if online_map bit is set, kernel may
       touch NODE_DATA() due to race condition. :-(


Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>

 arch/ia64/mm/discontig.c       |   24 +++++++++++++++++++-----
 include/asm-ia64/nodedata.h    |   12 ++++++++++++
 include/linux/memory_hotplug.h |    4 +---
 3 files changed, 32 insertions(+), 8 deletions(-)

Index: pgdat12/arch/ia64/mm/discontig.c
===================================================================
--- pgdat12.orig/arch/ia64/mm/discontig.c	2006-04-28 10:24:56.000000000 +0900
+++ pgdat12/arch/ia64/mm/discontig.c	2006-04-28 10:31:49.000000000 +0900
@@ -308,6 +308,17 @@ static void __init reserve_pernode_space
 	}
 }
 
+static void __meminit scatter_node_data(void)
+{
+	pg_data_t **dst;
+	int node;
+
+	for_each_online_node(node){
+		dst = LOCAL_DATA_ADDR(pgdat_list[node])->pg_data_ptrs;
+		memcpy(dst, pgdat_list, sizeof(pgdat_list));
+	}
+}
+
 /**
  * initialize_pernode_data - fixup per-cpu & per-node pointers
  *
@@ -320,11 +331,8 @@ static void __init initialize_pernode_da
 {
 	int cpu, node;
 
-	/* Copy the pg_data_t list to each node and init the node field */
-	for_each_online_node(node) {
-		memcpy(mem_data[node].node_data->pg_data_ptrs, pgdat_list,
-		       sizeof(pgdat_list));
-	}
+	scatter_node_data();
+
 #ifdef CONFIG_SMP
 	/* Set the node_data pointer for each per-cpu struct */
 	for (cpu = 0; cpu < NR_CPUS; cpu++) {
@@ -783,3 +791,9 @@ void __init paging_init(void)
 
 	zero_page_memmap_ptr = virt_to_page(ia64_imva(empty_zero_page));
 }
+
+void arch_refresh_nodedata(int update_node, pg_data_t *update_pgdat)
+{
+	pgdat_list[update_node] = update_pgdat;
+	scatter_node_data();
+}
Index: pgdat12/include/asm-ia64/nodedata.h
===================================================================
--- pgdat12.orig/include/asm-ia64/nodedata.h	2006-04-28 10:24:51.000000000 +0900
+++ pgdat12/include/asm-ia64/nodedata.h	2006-04-28 10:27:40.000000000 +0900
@@ -47,6 +47,18 @@ struct ia64_node_data {
  */
 #define NODE_DATA(nid)		(local_node_data->pg_data_ptrs[nid])
 
+/*
+ * LOCAL_DATA_ADDR - This is to calculate the address of other node's
+ *		     "local_node_data" at hot-plug phase. The local_node_data
+ *		     is pointed by per_cpu_page. Kernel usually use it for
+ *		     just executing cpu. However, when new node is hot-added,
+ *		     the addresses of local data for other nodes are necessary
+ *		     to update all of them.
+ */
+#define LOCAL_DATA_ADDR(pgdat)  			\
+	((struct ia64_node_data *)((u64)(pgdat) + 	\
+				   L1_CACHE_ALIGN(sizeof(struct pglist_data))))
+
 #endif /* CONFIG_NUMA */
 
 #endif /* _ASM_IA64_NODEDATA_H */
Index: pgdat12/include/linux/memory_hotplug.h
===================================================================
--- pgdat12.orig/include/linux/memory_hotplug.h	2006-04-28 10:24:51.000000000 +0900
+++ pgdat12/include/linux/memory_hotplug.h	2006-04-28 10:31:49.000000000 +0900
@@ -91,9 +91,7 @@ static inline pg_data_t *arch_alloc_node
 static inline void arch_free_nodedata(pg_data_t *pgdat)
 {
 }
-static inline void arch_refresh_nodedata(int nid, pg_data_t *pgdat)
-{
-}
+extern void arch_refresh_nodedata(int nid, pg_data_t *pgdat);
 
 #else /* CONFIG_HAVE_ARCH_NODEDATA_EXTENSION */
 

-- 
Yasunori Goto 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[Patch 003/003] pgdat allocation and update for ia64 of memory hotplug.(allocate pgdat and per node data)

[Yasunori Goto]

This is a patch to allocate pgdat and per node data area for ia64.
The size for them can be calculated by compute_pernodesize().

Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>

 arch/ia64/mm/discontig.c       |   16 ++++++++++++++--
 include/linux/memory_hotplug.h |    9 ++-------
 2 files changed, 16 insertions(+), 9 deletions(-)

Index: pgdat12/arch/ia64/mm/discontig.c
===================================================================
--- pgdat12.orig/arch/ia64/mm/discontig.c	2006-04-28 10:31:49.000000000 +0900
+++ pgdat12/arch/ia64/mm/discontig.c	2006-04-28 10:32:31.000000000 +0900
@@ -100,7 +100,7 @@ static int __init build_node_maps(unsign
  * acpi_boot_init() (which builds the node_to_cpu_mask array) hasn't been
  * called yet.  Note that node 0 will also count all non-existent cpus.
  */
-static int __init early_nr_cpus_node(int node)
+static int __meminit early_nr_cpus_node(int node)
 {
 	int cpu, n = 0;
 
@@ -115,7 +115,7 @@ static int __init early_nr_cpus_node(int
  * compute_pernodesize - compute size of pernode data
  * @node: the node id.
  */
-static unsigned long __init compute_pernodesize(int node)
+static unsigned long __meminit compute_pernodesize(int node)
 {
 	unsigned long pernodesize = 0, cpus;
 
@@ -792,6 +792,18 @@ void __init paging_init(void)
 	zero_page_memmap_ptr = virt_to_page(ia64_imva(empty_zero_page));
 }
 
+pg_data_t *arch_alloc_nodedata(int nid)
+{
+	unsigned long size = compute_pernodesize(nid);
+
+	return kzalloc(size, GFP_KERNEL);
+}
+
+void arch_free_nodedata(pg_data_t *pgdat)
+{
+	kfree(pgdat);
+}
+
 void arch_refresh_nodedata(int update_node, pg_data_t *update_pgdat)
 {
 	pgdat_list[update_node] = update_pgdat;
Index: pgdat12/include/linux/memory_hotplug.h
===================================================================
--- pgdat12.orig/include/linux/memory_hotplug.h	2006-04-28 10:31:49.000000000 +0900
+++ pgdat12/include/linux/memory_hotplug.h	2006-04-28 10:33:17.000000000 +0900
@@ -84,13 +84,8 @@ static inline int memofy_add_physaddr_to
  * Now, arch_free_nodedata() is just defined for error path of node_hot_add.
  *
  */
-static inline pg_data_t *arch_alloc_nodedata(int nid)
-{
-	return NULL;
-}
-static inline void arch_free_nodedata(pg_data_t *pgdat)
-{
-}
+extern pg_data_t *arch_alloc_nodedata(int nid);
+extern void arch_free_nodedata(pg_data_t *pgdat);
 extern void arch_refresh_nodedata(int nid, pg_data_t *pgdat);
 
 #else /* CONFIG_HAVE_ARCH_NODEDATA_EXTENSION */

-- 
Yasunori Goto 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>