[PATCH] ppc64: sparsemem memory model

Provide the architecture specific implementation for SPARSEMEM for PPC64 systems. Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Dave Hansen <haveblue@us.ibm.com> Signed-off-by: Mike Kravetz <kravetz@us.ibm.com> (in part) Signed-off-by: Martin Bligh <mbligh@aracnet.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ppc64: sparsemem memory model
Provide the architecture specific implementation for SPARSEMEM for PPC64 systems. Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Dave Hansen <haveblue@us.ibm.com> Signed-off-by: Mike Kravetz <kravetz@us.ibm.com> (in part) Signed-off-by: Martin Bligh <mbligh@aracnet.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Andy Whitcroft · Linus Torvalds
1 parent 74b30be2e1
Showing 7 changed files with 68 additions and 21 deletions Side-by-side Diff
arch/ppc64/Kconfig
arch/ppc64/kernel/setup.c
arch/ppc64/mm/Makefile
arch/ppc64/mm/init.c
include/asm-ppc64/mmzone.h
include/asm-ppc64/page.h
include/asm-ppc64/sparsemem.h
@@ -198,6 +198,13 @@
 	  This option enables hardware multithreading on RS64 cpus.
 	  pSeries systems p620 and p660 have such a cpu type.
  
+config ARCH_SELECT_MEMORY_MODEL
+	def_bool y
+
+config ARCH_FLATMEM_ENABLE
+       def_bool y
+       depends on !NUMA
+
 config ARCH_DISCONTIGMEM_ENABLE
 	def_bool y
 	depends on SMP && PPC_PSERIES
@@ -209,6 +216,10 @@
 config ARCH_FLATMEM_ENABLE
 	def_bool y
  
+config ARCH_SPARSEMEM_ENABLE
+	def_bool y
+	depends on ARCH_DISCONTIGMEM_ENABLE
+
 source "mm/Kconfig"
  
 config HAVE_ARCH_EARLY_PFN_TO_NID
@@ -229,7 +240,7 @@
  
 config NUMA
 	bool "NUMA support"
-	depends on DISCONTIGMEM
+	default y if DISCONTIGMEM || SPARSEMEM
  
 config SCHED_SMT
 	bool "SMT (Hyperthreading) scheduler support"
@@ -1055,6 +1055,7 @@
  
 	/* set up the bootmem stuff with available memory */
 	do_init_bootmem();
+	sparse_init();
  
 	/* initialize the syscall map in systemcfg */
 	setup_syscall_map();
@@ -6,7 +6,7 @@
  
 obj-y := fault.o init.o imalloc.o hash_utils.o hash_low.o tlb.o \
 	slb_low.o slb.o stab.o mmap.o
-obj-$(CONFIG_DISCONTIGMEM) += numa.o
+obj-$(CONFIG_NEED_MULTIPLE_NODES) += numa.o
 obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
 obj-$(CONFIG_PPC_MULTIPLATFORM) += hash_native.o
@@ -531,7 +531,7 @@
  * Initialize the bootmem system and give it all the memory we
  * have available.
  */
-#ifndef CONFIG_DISCONTIGMEM
+#ifndef CONFIG_NEED_MULTIPLE_NODES
 void __init do_init_bootmem(void)
 {
 	unsigned long i;
  
  
@@ -553,12 +553,20 @@
  
 	max_pfn = max_low_pfn;
  
-	/* add all physical memory to the bootmem map. Also find the first */
+	/* Add all physical memory to the bootmem map, mark each area
+	 * present.
+	 */
 	for (i=0; i < lmb.memory.cnt; i++) {
 		unsigned long physbase, size;
+		unsigned long start_pfn, end_pfn;
  
 		physbase = lmb.memory.region[i].physbase;
 		size = lmb.memory.region[i].size;
+
+		start_pfn = physbase >> PAGE_SHIFT;
+		end_pfn = start_pfn + (size >> PAGE_SHIFT);
+		memory_present(0, start_pfn, end_pfn);
+
 		free_bootmem(physbase, size);
 	}
  
@@ -597,7 +605,7 @@
 	free_area_init_node(0, NODE_DATA(0), zones_size,
 			    __pa(PAGE_OFFSET) >> PAGE_SHIFT, zholes_size);
 }
-#endif /* CONFIG_DISCONTIGMEM */
+#endif /* ! CONFIG_NEED_MULTIPLE_NODES */
  
 static struct kcore_list kcore_vmem;
  
@@ -628,7 +636,7 @@
  
 void __init mem_init(void)
 {
-#ifdef CONFIG_DISCONTIGMEM
+#ifdef CONFIG_NEED_MULTIPLE_NODES
 	int nid;
 #endif
 	pg_data_t *pgdat;
@@ -639,7 +647,7 @@
 	num_physpages = max_low_pfn;	/* RAM is assumed contiguous */
 	high_memory = (void *) __va(max_low_pfn * PAGE_SIZE);
  
-#ifdef CONFIG_DISCONTIGMEM
+#ifdef CONFIG_NEED_MULTIPLE_NODES
         for_each_online_node(nid) {
 		if (NODE_DATA(nid)->node_spanned_pages != 0) {
 			printk("freeing bootmem node %x\n", nid);
@@ -10,9 +10,20 @@
 #include <linux/config.h>
 #include <asm/smp.h>
  
-#ifdef CONFIG_DISCONTIGMEM
+/* generic non-linear memory support:
+ *
+ * 1) we will not split memory into more chunks than will fit into the
+ *    flags field of the struct page
+ */
  
+
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+
 extern struct pglist_data *node_data[];
+/*
+ * Return a pointer to the node data for node n.
+ */
+#define NODE_DATA(nid)		(node_data[nid])
  
 /*
  * Following are specific to this numa platform.
  
  
@@ -47,30 +58,27 @@
 	return nid;
 }
  
-#define pfn_to_nid(pfn)		pa_to_nid((pfn) << PAGE_SHIFT)
-
-/*
- * Return a pointer to the node data for node n.
- */
-#define NODE_DATA(nid)		(node_data[nid])
-
 #define node_localnr(pfn, nid)	((pfn) - NODE_DATA(nid)->node_start_pfn)
  
 /*
  * Following are macros that each numa implmentation must define.
  */
  
-/*
- * Given a kernel address, find the home node of the underlying memory.
- */
-#define kvaddr_to_nid(kaddr)	pa_to_nid(__pa(kaddr))
-
 #define node_start_pfn(nid)	(NODE_DATA(nid)->node_start_pfn)
 #define node_end_pfn(nid)	(NODE_DATA(nid)->node_end_pfn)
  
 #define local_mapnr(kvaddr) \
 	( (__pa(kvaddr) >> PAGE_SHIFT) - node_start_pfn(kvaddr_to_nid(kvaddr)) 
  
+#ifdef CONFIG_DISCONTIGMEM
+
+/*
+ * Given a kernel address, find the home node of the underlying memory.
+ */
+#define kvaddr_to_nid(kaddr)	pa_to_nid(__pa(kaddr))
+
+#define pfn_to_nid(pfn)		pa_to_nid((unsigned long)(pfn) << PAGE_SHIFT)
+
 /* Written this way to avoid evaluating arguments twice */
 #define discontigmem_pfn_to_page(pfn) \
 ({ \
@@ -90,6 +98,8 @@
 #define discontigmem_pfn_valid(pfn)		((pfn) < num_physpages)
  
 #endif /* CONFIG_DISCONTIGMEM */
+
+#endif /* CONFIG_NEED_MULTIPLE_NODES */
  
 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
 #define early_pfn_to_nid(pfn)  pa_to_nid(((unsigned long)pfn) << PAGE_SHIFT)
@@ -217,7 +217,8 @@
 #define page_to_pfn(page)	discontigmem_page_to_pfn(page)
 #define pfn_to_page(pfn)	discontigmem_pfn_to_page(pfn)
 #define pfn_valid(pfn)		discontigmem_pfn_valid(pfn)
-#else
+#endif
+#ifdef CONFIG_FLATMEM
 #define pfn_to_page(pfn)	(mem_map + (pfn))
 #define page_to_pfn(page)	((unsigned long)((page) - mem_map))
 #define pfn_valid(pfn)		((pfn) < max_mapnr)
+#ifndef _ASM_PPC64_SPARSEMEM_H
+#define _ASM_PPC64_SPARSEMEM_H 1
+
+#ifdef CONFIG_SPARSEMEM
+/*
+ * SECTION_SIZE_BITS		2^N: how big each section will be
+ * MAX_PHYSADDR_BITS		2^N: how much physical address space we have
+ * MAX_PHYSMEM_BITS		2^N: how much memory we can have in that space
+ */
+#define SECTION_SIZE_BITS       24
+#define MAX_PHYSADDR_BITS       38
+#define MAX_PHYSMEM_BITS        36
+
+#endif /* CONFIG_SPARSEMEM */
+
+#endif /* _ASM_PPC64_SPARSEMEM_H */
...	...	@@ -198,6 +198,13 @@
198	198	This option enables hardware multithreading on RS64 cpus.
199	199	pSeries systems p620 and p660 have such a cpu type.
200	200
	201	+config ARCH_SELECT_MEMORY_MODEL
	202	+ def_bool y
	203	+
	204	+config ARCH_FLATMEM_ENABLE
	205	+ def_bool y
	206	+ depends on !NUMA
	207	+
201	208	config ARCH_DISCONTIGMEM_ENABLE
202	209	def_bool y
203	210	depends on SMP && PPC_PSERIES
...	...	@@ -209,6 +216,10 @@
209	216	config ARCH_FLATMEM_ENABLE
210	217	def_bool y
211	218
	219	+config ARCH_SPARSEMEM_ENABLE
	220	+ def_bool y
	221	+ depends on ARCH_DISCONTIGMEM_ENABLE
	222	+
212	223	source "mm/Kconfig"
213	224
214	225	config HAVE_ARCH_EARLY_PFN_TO_NID
...	...	@@ -229,7 +240,7 @@
229	240
230	241	config NUMA
231	242	bool "NUMA support"
232		- depends on DISCONTIGMEM
	243	+ default y if DISCONTIGMEM \|\| SPARSEMEM
233	244
234	245	config SCHED_SMT
235	246	bool "SMT (Hyperthreading) scheduler support"
...	...	@@ -1055,6 +1055,7 @@
1055	1055
1056	1056	/* set up the bootmem stuff with available memory */
1057	1057	do_init_bootmem();
	1058	+ sparse_init();
1058	1059
1059	1060	/* initialize the syscall map in systemcfg */
1060	1061	setup_syscall_map();
...	...	@@ -6,7 +6,7 @@
6	6
7	7	obj-y := fault.o init.o imalloc.o hash_utils.o hash_low.o tlb.o \
8	8	slb_low.o slb.o stab.o mmap.o
9		-obj-$(CONFIG_DISCONTIGMEM) += numa.o
	9	+obj-$(CONFIG_NEED_MULTIPLE_NODES) += numa.o
10	10	obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
11	11	obj-$(CONFIG_PPC_MULTIPLATFORM) += hash_native.o
...	...	@@ -531,7 +531,7 @@
531	531	* Initialize the bootmem system and give it all the memory we
532	532	* have available.
533	533	*/
534		-#ifndef CONFIG_DISCONTIGMEM
	534	+#ifndef CONFIG_NEED_MULTIPLE_NODES
535	535	void __init do_init_bootmem(void)
536	536	{
537	537	unsigned long i;
538	538
539	539
...	...	@@ -553,12 +553,20 @@
553	553
554	554	max_pfn = max_low_pfn;
555	555
556		- /* add all physical memory to the bootmem map. Also find the first */
	556	+ /* Add all physical memory to the bootmem map, mark each area
	557	+ * present.
	558	+ */
557	559	for (i=0; i < lmb.memory.cnt; i++) {
558	560	unsigned long physbase, size;
	561	+ unsigned long start_pfn, end_pfn;
559	562
560	563	physbase = lmb.memory.region[i].physbase;
561	564	size = lmb.memory.region[i].size;
	565	+
	566	+ start_pfn = physbase >> PAGE_SHIFT;
	567	+ end_pfn = start_pfn + (size >> PAGE_SHIFT);
	568	+ memory_present(0, start_pfn, end_pfn);
	569	+
562	570	free_bootmem(physbase, size);
563	571	}
564	572
...	...	@@ -597,7 +605,7 @@
597	605	free_area_init_node(0, NODE_DATA(0), zones_size,
598	606	__pa(PAGE_OFFSET) >> PAGE_SHIFT, zholes_size);
599	607	}
600		-#endif /* CONFIG_DISCONTIGMEM */
	608	+#endif /* ! CONFIG_NEED_MULTIPLE_NODES */
601	609
602	610	static struct kcore_list kcore_vmem;
603	611
...	...	@@ -628,7 +636,7 @@
628	636
629	637	void __init mem_init(void)
630	638	{
631		-#ifdef CONFIG_DISCONTIGMEM
	639	+#ifdef CONFIG_NEED_MULTIPLE_NODES
632	640	int nid;
633	641	#endif
634	642	pg_data_t *pgdat;
...	...	@@ -639,7 +647,7 @@
639	647	num_physpages = max_low_pfn; /* RAM is assumed contiguous */
640	648	high_memory = (void ) __va(max_low_pfn PAGE_SIZE);
641	649
642		-#ifdef CONFIG_DISCONTIGMEM
	650	+#ifdef CONFIG_NEED_MULTIPLE_NODES
643	651	for_each_online_node(nid) {
644	652	if (NODE_DATA(nid)->node_spanned_pages != 0) {
645	653	printk("freeing bootmem node %x\n", nid);
...	...	@@ -10,9 +10,20 @@
10	10	#include <linux/config.h>
11	11	#include <asm/smp.h>
12	12
13		-#ifdef CONFIG_DISCONTIGMEM
	13	+/* generic non-linear memory support:
	14	+ *
	15	+ * 1) we will not split memory into more chunks than will fit into the
	16	+ * flags field of the struct page
	17	+ */
14	18
	19	+
	20	+#ifdef CONFIG_NEED_MULTIPLE_NODES
	21	+
15	22	extern struct pglist_data *node_data[];
	23	+/*
	24	+ * Return a pointer to the node data for node n.
	25	+ */
	26	+#define NODE_DATA(nid) (node_data[nid])
16	27
17	28	/*
18	29	* Following are specific to this numa platform.
19	30
20	31
...	...	@@ -47,30 +58,27 @@
47	58	return nid;
48	59	}
49	60
50		-#define pfn_to_nid(pfn) pa_to_nid((pfn) << PAGE_SHIFT)
51		-
52		-/*
53		- * Return a pointer to the node data for node n.
54		- */
55		-#define NODE_DATA(nid) (node_data[nid])
56		-
57	61	#define node_localnr(pfn, nid) ((pfn) - NODE_DATA(nid)->node_start_pfn)
58	62
59	63	/*
60	64	* Following are macros that each numa implmentation must define.
61	65	*/
62	66
63		-/*
64		- * Given a kernel address, find the home node of the underlying memory.
65		- */
66		-#define kvaddr_to_nid(kaddr) pa_to_nid(__pa(kaddr))
67		-
68	67	#define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
69	68	#define node_end_pfn(nid) (NODE_DATA(nid)->node_end_pfn)
70	69
71	70	#define local_mapnr(kvaddr) \
72	71	( (__pa(kvaddr) >> PAGE_SHIFT) - node_start_pfn(kvaddr_to_nid(kvaddr))
73	72
	73	+#ifdef CONFIG_DISCONTIGMEM
	74	+
	75	+/*
	76	+ * Given a kernel address, find the home node of the underlying memory.
	77	+ */
	78	+#define kvaddr_to_nid(kaddr) pa_to_nid(__pa(kaddr))
	79	+
	80	+#define pfn_to_nid(pfn) pa_to_nid((unsigned long)(pfn) << PAGE_SHIFT)
	81	+
74	82	/* Written this way to avoid evaluating arguments twice */
75	83	#define discontigmem_pfn_to_page(pfn) \
76	84	({ \
...	...	@@ -90,6 +98,8 @@
90	98	#define discontigmem_pfn_valid(pfn) ((pfn) < num_physpages)
91	99
92	100	#endif /* CONFIG_DISCONTIGMEM */
	101	+
	102	+#endif /* CONFIG_NEED_MULTIPLE_NODES */
93	103
94	104	#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
95	105	#define early_pfn_to_nid(pfn) pa_to_nid(((unsigned long)pfn) << PAGE_SHIFT)
...	...	@@ -217,7 +217,8 @@
217	217	#define page_to_pfn(page) discontigmem_page_to_pfn(page)
218	218	#define pfn_to_page(pfn) discontigmem_pfn_to_page(pfn)
219	219	#define pfn_valid(pfn) discontigmem_pfn_valid(pfn)
220		-#else
	220	+#endif
	221	+#ifdef CONFIG_FLATMEM
221	222	#define pfn_to_page(pfn) (mem_map + (pfn))
222	223	#define page_to_pfn(page) ((unsigned long)((page) - mem_map))
223	224	#define pfn_valid(pfn) ((pfn) < max_mapnr)
	1	+#ifndef _ASM_PPC64_SPARSEMEM_H
	2	+#define _ASM_PPC64_SPARSEMEM_H 1
	3	+
	4	+#ifdef CONFIG_SPARSEMEM
	5	+/*
	6	+ * SECTION_SIZE_BITS 2^N: how big each section will be
	7	+ * MAX_PHYSADDR_BITS 2^N: how much physical address space we have
	8	+ * MAX_PHYSMEM_BITS 2^N: how much memory we can have in that space
	9	+ */
	10	+#define SECTION_SIZE_BITS 24
	11	+#define MAX_PHYSADDR_BITS 38
	12	+#define MAX_PHYSMEM_BITS 36
	13	+
	14	+#endif /* CONFIG_SPARSEMEM */
	15	+
	16	+#endif /* _ASM_PPC64_SPARSEMEM_H */