blob: 3150ad385672bc190fbe6b2b789e2e71de734d1d [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Written by: Patricia Gaughen <gone@us.ibm.com>, IBM Corporation
3 * August 2002: added remote node KVA remap - Martin J. Bligh
4 *
5 * Copyright (C) 2002, IBM Corp.
6 *
7 * All rights reserved.
8 *
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
13 *
14 * This program is distributed in the hope that it will be useful, but
15 * WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
17 * NON INFRINGEMENT. See the GNU General Public License for more
18 * details.
19 *
20 * You should have received a copy of the GNU General Public License
21 * along with this program; if not, write to the Free Software
22 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
23 */
24
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/mm.h>
26#include <linux/bootmem.h>
27#include <linux/mmzone.h>
28#include <linux/highmem.h>
29#include <linux/initrd.h>
30#include <linux/nodemask.h>
Alexey Dobriyan129f6942005-06-23 00:08:33 -070031#include <linux/module.h>
Eric W. Biederman1bc3b912005-06-25 14:58:01 -070032#include <linux/kexec.h>
Dave Hansen22a98352006-03-27 01:16:04 -080033#include <linux/pfn.h>
Linus Torvalds28aa4832007-05-15 18:45:49 -070034#include <linux/swap.h>
Mel Gorman1b000a52008-01-30 13:33:25 +010035#include <linux/acpi.h>
Eric W. Biederman1bc3b912005-06-25 14:58:01 -070036
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#include <asm/e820.h>
38#include <asm/setup.h>
39#include <asm/mmzone.h>
Alexey Starikovskiyce3fe6b2008-03-17 22:08:17 +030040#include <asm/bios_ebda.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
Ravikiran G Thirumalai6c231b72005-09-06 15:17:45 -070042struct pglist_data *node_data[MAX_NUMNODES] __read_mostly;
Alexey Dobriyan129f6942005-06-23 00:08:33 -070043EXPORT_SYMBOL(node_data);
Adrian Bunkfb8c1772007-10-24 18:24:47 +020044static bootmem_data_t node0_bdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
46/*
Adrian Bunkd254c8f2006-06-30 18:29:51 +020047 * numa interface - we expect the numa architecture specific code to have
Linus Torvalds1da177e2005-04-16 15:20:36 -070048 * populated the following initialisation.
49 *
50 * 1) node_online_map - the map of all nodes configured (online) in the system
Andy Whitcroft05b79bd2005-06-23 00:07:57 -070051 * 2) node_start_pfn - the starting page frame number for a node
Linus Torvalds1da177e2005-04-16 15:20:36 -070052 * 3) node_end_pfn - the ending page fram number for a node
53 */
Ravikiran G Thirumalai6c231b72005-09-06 15:17:45 -070054unsigned long node_start_pfn[MAX_NUMNODES] __read_mostly;
55unsigned long node_end_pfn[MAX_NUMNODES] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070056
Andy Whitcroft05b79bd2005-06-23 00:07:57 -070057
58#ifdef CONFIG_DISCONTIGMEM
Linus Torvalds1da177e2005-04-16 15:20:36 -070059/*
Andy Whitcroft05b79bd2005-06-23 00:07:57 -070060 * 4) physnode_map - the mapping between a pfn and owning node
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 * physnode_map keeps track of the physical memory layout of a generic
Yinghai Luba924c82008-05-31 22:51:51 -070062 * numa node on a 64Mb break (each element of the array will
63 * represent 64Mb of memory and will be marked by the node id. so,
Linus Torvalds1da177e2005-04-16 15:20:36 -070064 * if the first gig is on node 0, and the second gig is on node 1
65 * physnode_map will contain:
66 *
Yinghai Luba924c82008-05-31 22:51:51 -070067 * physnode_map[0-15] = 0;
68 * physnode_map[16-31] = 1;
69 * physnode_map[32- ] = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 */
Ravikiran G Thirumalai6c231b72005-09-06 15:17:45 -070071s8 physnode_map[MAX_ELEMENTS] __read_mostly = { [0 ... (MAX_ELEMENTS - 1)] = -1};
Alexey Dobriyan129f6942005-06-23 00:08:33 -070072EXPORT_SYMBOL(physnode_map);
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
74void memory_present(int nid, unsigned long start, unsigned long end)
75{
76 unsigned long pfn;
77
78 printk(KERN_INFO "Node: %d, start_pfn: %ld, end_pfn: %ld\n",
79 nid, start, end);
80 printk(KERN_DEBUG " Setting physnode_map array to node %d for pfns:\n", nid);
81 printk(KERN_DEBUG " ");
82 for (pfn = start; pfn < end; pfn += PAGES_PER_ELEMENT) {
83 physnode_map[pfn / PAGES_PER_ELEMENT] = nid;
Yinghai Luba924c82008-05-31 22:51:51 -070084 printk(KERN_CONT "%ld ", pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -070085 }
Yinghai Luba924c82008-05-31 22:51:51 -070086 printk(KERN_CONT "\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -070087}
88
89unsigned long node_memmap_size_bytes(int nid, unsigned long start_pfn,
90 unsigned long end_pfn)
91{
92 unsigned long nr_pages = end_pfn - start_pfn;
93
94 if (!nr_pages)
95 return 0;
96
97 return (nr_pages + 1) * sizeof(struct page);
98}
Andy Whitcroft05b79bd2005-06-23 00:07:57 -070099#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
101extern unsigned long find_max_low_pfn(void);
Dave Hansen05039b92005-10-29 18:16:57 -0700102extern void add_one_highpage_init(struct page *, int, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103extern unsigned long highend_pfn, highstart_pfn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104
105#define LARGE_PAGE_BYTES (PTRS_PER_PTE * PAGE_SIZE)
106
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107unsigned long node_remap_size[MAX_NUMNODES];
Adrian Bunk59659f12007-10-17 18:04:36 +0200108static void *node_remap_start_vaddr[MAX_NUMNODES];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109void set_pmd_pfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags);
110
keith mannthey91023302006-09-25 23:31:03 -0700111static unsigned long kva_start_pfn;
112static unsigned long kva_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113/*
114 * FLAT - support for basic PC memory model with discontig enabled, essentially
115 * a single node with all available processors in it with a flat
116 * memory map.
117 */
118int __init get_memcfg_numa_flat(void)
119{
120 printk("NUMA - single node, flat memory mode\n");
121
122 /* Run the memory configuration and find the top of memory. */
Yinghai Lu05961522008-05-31 22:52:47 -0700123 find_max_pfn();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124 node_start_pfn[0] = 0;
125 node_end_pfn[0] = max_pfn;
126 memory_present(0, 0, max_pfn);
Yinghai Lub66cd722008-05-31 22:53:47 -0700127 node_remap_size[0] = node_memmap_size_bytes(0, 0, max_pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128
129 /* Indicate there is one node available. */
130 nodes_clear(node_online_map);
131 node_set_online(0);
132 return 1;
133}
134
135/*
136 * Find the highest page frame number we have available for the node
137 */
Ingo Molnarfa5c4632008-04-16 02:29:42 +0200138static void __init propagate_e820_map_node(int nid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139{
140 if (node_end_pfn[nid] > max_pfn)
141 node_end_pfn[nid] = max_pfn;
142 /*
143 * if a user has given mem=XXXX, then we need to make sure
144 * that the node _starts_ before that, too, not just ends
145 */
146 if (node_start_pfn[nid] > max_pfn)
147 node_start_pfn[nid] = max_pfn;
Eric Sesterhenn8d8f3cb2006-10-03 23:34:58 +0200148 BUG_ON(node_start_pfn[nid] > node_end_pfn[nid]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149}
150
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151/*
152 * Allocate memory for the pg_data_t for this node via a crude pre-bootmem
153 * method. For node zero take this from the bottom of memory, for
154 * subsequent nodes place them at node_remap_start_vaddr which contains
155 * node local data in physically node local memory. See setup_memory()
156 * for details.
157 */
158static void __init allocate_pgdat(int nid)
159{
160 if (nid && node_has_online_mem(nid))
161 NODE_DATA(nid) = (pg_data_t *)node_remap_start_vaddr[nid];
162 else {
Yinghai Lu16387292008-05-29 12:57:22 -0700163 unsigned long pgdat_phys;
164 pgdat_phys = find_e820_area(min_low_pfn<<PAGE_SHIFT,
165 max_low_pfn<<PAGE_SHIFT, sizeof(pg_data_t),
166 PAGE_SIZE);
167 NODE_DATA(nid) = (pg_data_t *)(pfn_to_kaddr(pgdat_phys>>PAGE_SHIFT));
168 reserve_early(pgdat_phys, pgdat_phys + sizeof(pg_data_t),
169 "NODE_DATA");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 }
Yinghai Lue8c27ac2008-06-01 13:15:22 -0700171 printk(KERN_DEBUG "allocate_pgdat: node %d NODE_DATA %08lx\n",
172 nid, (unsigned long)NODE_DATA(nid));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173}
174
Mel Gorman1b000a52008-01-30 13:33:25 +0100175#ifdef CONFIG_DISCONTIGMEM
176/*
177 * In the discontig memory model, a portion of the kernel virtual area (KVA)
178 * is reserved and portions of nodes are mapped using it. This is to allow
179 * node-local memory to be allocated for structures that would normally require
180 * ZONE_NORMAL. The memory is allocated with alloc_remap() and callers
181 * should be prepared to allocate from the bootmem allocator instead. This KVA
182 * mechanism is incompatible with SPARSEMEM as it makes assumptions about the
183 * layout of memory that are broken if alloc_remap() succeeds for some of the
184 * map and fails for others
185 */
186static unsigned long node_remap_start_pfn[MAX_NUMNODES];
187static void *node_remap_end_vaddr[MAX_NUMNODES];
188static void *node_remap_alloc_vaddr[MAX_NUMNODES];
189static unsigned long node_remap_offset[MAX_NUMNODES];
190
Dave Hansen6f167ec2005-06-23 00:07:39 -0700191void *alloc_remap(int nid, unsigned long size)
192{
193 void *allocation = node_remap_alloc_vaddr[nid];
194
195 size = ALIGN(size, L1_CACHE_BYTES);
196
197 if (!allocation || (allocation + size) >= node_remap_end_vaddr[nid])
198 return 0;
199
200 node_remap_alloc_vaddr[nid] += size;
201 memset(allocation, 0, size);
202
203 return allocation;
204}
205
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206void __init remap_numa_kva(void)
207{
208 void *vaddr;
209 unsigned long pfn;
210 int node;
211
212 for_each_online_node(node) {
Yinghai Lue8c27ac2008-06-01 13:15:22 -0700213 printk(KERN_DEBUG "remap_numa_kva: node %d\n", node);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 for (pfn=0; pfn < node_remap_size[node]; pfn += PTRS_PER_PTE) {
215 vaddr = node_remap_start_vaddr[node]+(pfn<<PAGE_SHIFT);
Yinghai Lue8c27ac2008-06-01 13:15:22 -0700216 printk(KERN_DEBUG "remap_numa_kva: %08lx to pfn %08lx\n",
217 (unsigned long)vaddr,
218 node_remap_start_pfn[node] + pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 set_pmd_pfn((ulong) vaddr,
220 node_remap_start_pfn[node] + pfn,
221 PAGE_KERNEL_LARGE);
222 }
223 }
224}
225
226static unsigned long calculate_numa_remap_pages(void)
227{
228 int nid;
229 unsigned long size, reserve_pages = 0;
Dave Hansen5b505b92005-06-23 00:07:41 -0700230 unsigned long pfn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231
232 for_each_online_node(nid) {
Mel Gorman4cfee882006-09-27 01:49:51 -0700233 unsigned old_end_pfn = node_end_pfn[nid];
234
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 /*
236 * The acpi/srat node info can show hot-add memroy zones
237 * where memory could be added but not currently present.
238 */
239 if (node_start_pfn[nid] > max_pfn)
240 continue;
241 if (node_end_pfn[nid] > max_pfn)
242 node_end_pfn[nid] = max_pfn;
243
244 /* ensure the remap includes space for the pgdat. */
245 size = node_remap_size[nid] + sizeof(pg_data_t);
246
247 /* convert size to large (pmd size) pages, rounding up */
248 size = (size + LARGE_PAGE_BYTES - 1) / LARGE_PAGE_BYTES;
249 /* now the roundup is correct, convert to PAGE_SIZE pages */
250 size = size * PTRS_PER_PTE;
Dave Hansen5b505b92005-06-23 00:07:41 -0700251
252 /*
253 * Validate the region we are allocating only contains valid
254 * pages.
255 */
256 for (pfn = node_end_pfn[nid] - size;
257 pfn < node_end_pfn[nid]; pfn++)
258 if (!page_is_ram(pfn))
259 break;
260
261 if (pfn != node_end_pfn[nid])
262 size = 0;
263
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 printk("Reserving %ld pages of KVA for lmem_map of node %d\n",
265 size, nid);
266 node_remap_size[nid] = size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 node_remap_offset[nid] = reserve_pages;
Dave Hansen6f167ec2005-06-23 00:07:39 -0700268 reserve_pages += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 printk("Shrinking node %d from %ld pages to %ld pages\n",
270 nid, node_end_pfn[nid], node_end_pfn[nid] - size);
Ravikiran G Thirumalai4b0271e2005-08-07 09:42:50 -0700271
272 if (node_end_pfn[nid] & (PTRS_PER_PTE-1)) {
273 /*
274 * Align node_end_pfn[] and node_remap_start_pfn[] to
275 * pmd boundary. remap_numa_kva will barf otherwise.
276 */
277 printk("Shrinking node %d further by %ld pages for proper alignment\n",
278 nid, node_end_pfn[nid] & (PTRS_PER_PTE-1));
279 size += node_end_pfn[nid] & (PTRS_PER_PTE-1);
280 }
281
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 node_end_pfn[nid] -= size;
283 node_remap_start_pfn[nid] = node_end_pfn[nid];
Mel Gorman4cfee882006-09-27 01:49:51 -0700284 shrink_active_range(nid, old_end_pfn, node_end_pfn[nid]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 }
286 printk("Reserving total of %ld pages for numa KVA remap\n",
287 reserve_pages);
288 return reserve_pages;
289}
290
Mel Gorman1b000a52008-01-30 13:33:25 +0100291static void init_remap_allocator(int nid)
292{
293 node_remap_start_vaddr[nid] = pfn_to_kaddr(
294 kva_start_pfn + node_remap_offset[nid]);
295 node_remap_end_vaddr[nid] = node_remap_start_vaddr[nid] +
296 (node_remap_size[nid] * PAGE_SIZE);
297 node_remap_alloc_vaddr[nid] = node_remap_start_vaddr[nid] +
298 ALIGN(sizeof(pg_data_t), PAGE_SIZE);
299
300 printk ("node %d will remap to vaddr %08lx - %08lx\n", nid,
301 (ulong) node_remap_start_vaddr[nid],
Yinghai Lue8c27ac2008-06-01 13:15:22 -0700302 (ulong) node_remap_end_vaddr[nid]);
Mel Gorman1b000a52008-01-30 13:33:25 +0100303}
304#else
305void *alloc_remap(int nid, unsigned long size)
306{
307 return NULL;
308}
309
310static unsigned long calculate_numa_remap_pages(void)
311{
312 return 0;
313}
314
315static void init_remap_allocator(int nid)
316{
317}
318
319void __init remap_numa_kva(void)
320{
321}
322#endif /* CONFIG_DISCONTIGMEM */
323
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324extern void setup_bootmem_allocator(void);
325unsigned long __init setup_memory(void)
326{
327 int nid;
328 unsigned long system_start_pfn, system_max_low_pfn;
Mel Gormanbac48942008-01-30 13:32:54 +0100329 unsigned long wasted_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330
331 /*
332 * When mapping a NUMA machine we allocate the node_mem_map arrays
333 * from node local memory. They are then mapped directly into KVA
334 * between zone normal and vmalloc space. Calculate the size of
Simon Arlott27b46d72007-10-20 01:13:56 +0200335 * this space and use it to adjust the boundary between ZONE_NORMAL
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 * and ZONE_HIGHMEM.
337 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 get_memcfg_numa();
339
keith mannthey91023302006-09-25 23:31:03 -0700340 kva_pages = calculate_numa_remap_pages();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341
342 /* partially used pages are not usable - thus round upwards */
343 system_start_pfn = min_low_pfn = PFN_UP(init_pg_tables_end);
344
keith mannthey91023302006-09-25 23:31:03 -0700345 kva_start_pfn = find_max_low_pfn() - kva_pages;
346
347#ifdef CONFIG_BLK_DEV_INITRD
348 /* Numa kva area is below the initrd */
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100349 if (initrd_start)
350 kva_start_pfn = PFN_DOWN(initrd_start - PAGE_OFFSET)
H. Peter Anvin30c82642007-10-15 17:13:22 -0700351 - kva_pages;
keith mannthey91023302006-09-25 23:31:03 -0700352#endif
Mel Gormanbac48942008-01-30 13:32:54 +0100353
354 /*
355 * We waste pages past at the end of the KVA for no good reason other
356 * than how it is located. This is bad.
357 */
358 wasted_pages = kva_start_pfn & (PTRS_PER_PTE-1);
359 kva_start_pfn -= wasted_pages;
360 kva_pages += wasted_pages;
keith mannthey91023302006-09-25 23:31:03 -0700361
362 system_max_low_pfn = max_low_pfn = find_max_low_pfn();
363 printk("kva_start_pfn ~ %ld find_max_low_pfn() ~ %ld\n",
364 kva_start_pfn, max_low_pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 printk("max_pfn = %ld\n", max_pfn);
Yinghai Lua5481282008-05-29 12:58:37 -0700366
367 /* avoid clash with initrd */
368 reserve_early(kva_start_pfn<<PAGE_SHIFT,
369 (kva_start_pfn + kva_pages)<<PAGE_SHIFT,
370 "KVA PG");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371#ifdef CONFIG_HIGHMEM
372 highstart_pfn = highend_pfn = max_pfn;
373 if (max_pfn > system_max_low_pfn)
374 highstart_pfn = system_max_low_pfn;
375 printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
376 pages_to_mb(highend_pfn - highstart_pfn));
Jan Beulichba9c2312006-09-26 10:52:31 +0200377 num_physpages = highend_pfn;
378 high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
379#else
380 num_physpages = system_max_low_pfn;
381 high_memory = (void *) __va(system_max_low_pfn * PAGE_SIZE - 1) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382#endif
383 printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
384 pages_to_mb(system_max_low_pfn));
385 printk("min_low_pfn = %ld, max_low_pfn = %ld, highstart_pfn = %ld\n",
386 min_low_pfn, max_low_pfn, highstart_pfn);
387
388 printk("Low memory ends at vaddr %08lx\n",
389 (ulong) pfn_to_kaddr(max_low_pfn));
390 for_each_online_node(nid) {
Mel Gorman1b000a52008-01-30 13:33:25 +0100391 init_remap_allocator(nid);
Dave Hansen6f167ec2005-06-23 00:07:39 -0700392
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 allocate_pgdat(nid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 }
395 printk("High memory starts at vaddr %08lx\n",
396 (ulong) pfn_to_kaddr(highstart_pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 for_each_online_node(nid)
Ingo Molnarfa5c4632008-04-16 02:29:42 +0200398 propagate_e820_map_node(nid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
400 memset(NODE_DATA(0), 0, sizeof(struct pglist_data));
401 NODE_DATA(0)->bdata = &node0_bdata;
402 setup_bootmem_allocator();
403 return max_low_pfn;
404}
405
406void __init zone_sizes_init(void)
407{
408 int nid;
Mel Gorman6391af12006-10-11 01:20:39 -0700409 unsigned long max_zone_pfns[MAX_NR_ZONES];
410 memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
411 max_zone_pfns[ZONE_DMA] =
412 virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
413 max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
Linus Torvalds28aa4832007-05-15 18:45:49 -0700414#ifdef CONFIG_HIGHMEM
Mel Gorman6391af12006-10-11 01:20:39 -0700415 max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
Linus Torvalds28aa4832007-05-15 18:45:49 -0700416#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
Mel Gorman4cfee882006-09-27 01:49:51 -0700418 /* If SRAT has not registered memory, register it now */
419 if (find_max_pfn_with_active_regions() == 0) {
420 for_each_online_node(nid) {
421 if (node_has_online_mem(nid))
422 add_active_range(nid, node_start_pfn[nid],
423 node_end_pfn[nid]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 }
Mel Gorman4cfee882006-09-27 01:49:51 -0700426
427 free_area_init_nodes(max_zone_pfns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 return;
429}
430
431void __init set_highmem_pages_init(int bad_ppro)
432{
433#ifdef CONFIG_HIGHMEM
434 struct zone *zone;
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700435 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436
437 for_each_zone(zone) {
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700438 unsigned long node_pfn, zone_start_pfn, zone_end_pfn;
439
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 if (!is_highmem(zone))
441 continue;
442
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 zone_start_pfn = zone->zone_start_pfn;
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700444 zone_end_pfn = zone_start_pfn + zone->spanned_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700446 printk("Initializing %s for node %d (%08lx:%08lx)\n",
Christoph Lameter89fa3022006-09-25 23:31:55 -0700447 zone->name, zone_to_nid(zone),
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700448 zone_start_pfn, zone_end_pfn);
449
450 for (node_pfn = zone_start_pfn; node_pfn < zone_end_pfn; node_pfn++) {
451 if (!pfn_valid(node_pfn))
452 continue;
453 page = pfn_to_page(node_pfn);
Dave Hansen05039b92005-10-29 18:16:57 -0700454 add_one_highpage_init(page, node_pfn, bad_ppro);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 }
456 }
457 totalram_pages += totalhigh_pages;
458#endif
459}
Yasunori Goto7c7e9422006-12-22 01:11:13 -0800460
461#ifdef CONFIG_MEMORY_HOTPLUG
Adrian Bunkfb8c1772007-10-24 18:24:47 +0200462static int paddr_to_nid(u64 addr)
Yasunori Goto7c7e9422006-12-22 01:11:13 -0800463{
464 int nid;
465 unsigned long pfn = PFN_DOWN(addr);
466
467 for_each_node(nid)
468 if (node_start_pfn[nid] <= pfn &&
469 pfn < node_end_pfn[nid])
470 return nid;
471
472 return -1;
473}
474
475/*
476 * This function is used to ask node id BEFORE memmap and mem_section's
477 * initialization (pfn_to_nid() can't be used yet).
478 * If _PXM is not defined on ACPI's DSDT, node id must be found by this.
479 */
480int memory_add_physaddr_to_nid(u64 addr)
481{
482 int nid = paddr_to_nid(addr);
483 return (nid >= 0) ? nid : 0;
484}
485
486EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
487#endif