aboutsummaryrefslogtreecommitdiff
path: root/arch/alpha/mm/numa.c
blob: a53fda0481ca32c0a33d9f9914f32e3ac94edd98 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
/*
 *  linux/arch/alpha/mm/numa.c
 *
 *  DISCONTIGMEM NUMA alpha support.
 *
 *  Copyright (C) 2001 Andrea Arcangeli <andrea@suse.de> SuSE
 */

#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/bootmem.h>
#include <linux/swap.h>
#include <linux/initrd.h>
#include <linux/pfn.h>
#include <linux/module.h>

#include <asm/hwrpb.h>
#include <asm/pgalloc.h>

pg_data_t node_data[MAX_NUMNODES];
EXPORT_SYMBOL(node_data);

#undef DEBUG_DISCONTIG
#ifdef DEBUG_DISCONTIG
#define DBGDCONT(args...) printk(args)
#else
#define DBGDCONT(args...)
#endif

#define for_each_mem_cluster(memdesc, cluster, i)		\
	for ((cluster) = (memdesc)->cluster, (i) = 0;		\
	     (i) < (memdesc)->numclusters; (i)++, (cluster)++)

static void __init show_mem_layout(void)
{
	struct memclust_struct * cluster;
	struct memdesc_struct * memdesc;
	int i;

	/* Find free clusters, and init and free the bootmem accordingly.  */
	memdesc = (struct memdesc_struct *)
	  (hwrpb->mddt_offset + (unsigned long) hwrpb);

	printk("Raw memory layout:\n");
	for_each_mem_cluster(memdesc, cluster, i) {
		printk(" memcluster %2d, usage %1lx, start %8lu, end %8lu\n",
		       i, cluster->usage, cluster->start_pfn,
		       cluster->start_pfn + cluster->numpages);
	}
}

static void __init
setup_memory_node(int nid, void *kernel_end)
{
	extern unsigned long mem_size_limit;
	struct memclust_struct * cluster;
	struct memdesc_struct * memdesc;
	unsigned long start_kernel_pfn, end_kernel_pfn;
	unsigned long bootmap_size, bootmap_pages, bootmap_start;
	unsigned long start, end;
	unsigned long node_pfn_start, node_pfn_end;
	unsigned long node_min_pfn, node_max_pfn;
	int i;
	unsigned long node_datasz = PFN_UP(sizeof(pg_data_t));
	int show_init = 0;

	/* Find the bounds of current node */
	node_pfn_start = (node_mem_start(nid)) >> PAGE_SHIFT;
	node_pfn_end = node_pfn_start + (node_mem_size(nid) >> PAGE_SHIFT);
	
	/* Find free clusters, and init and free the bootmem accordingly.  */
	memdesc = (struct memdesc_struct *)
	  (hwrpb->mddt_offset + (unsigned long) hwrpb);

	/* find the bounds of this node (node_min_pfn/node_max_pfn) */
	node_min_pfn = ~0UL;
	node_max_pfn = 0UL;
	for_each_mem_cluster(memdesc, cluster, i) {
		/* Bit 0 is console/PALcode reserved.  Bit 1 is
		   non-volatile memory -- we might want to mark
		   this for later.  */
		if (cluster->usage & 3)
			continue;

		start = cluster->start_pfn;
		end = start + cluster->numpages;

		if (start >= node_pfn_end || end <= node_pfn_start)
			continue;

		if (!show_init) {
			show_init = 1;
			printk("Initializing bootmem allocator on Node ID %d\n", nid);
		}
		printk(" memcluster %2d, usage %1lx, start %8lu, end %8lu\n",
		       i, cluster->usage, cluster->start_pfn,
		       cluster->start_pfn + cluster->numpages);

		if (start < node_pfn_start)
			start = node_pfn_start;
		if (end > node_pfn_end)
			end = node_pfn_end;

		if (start < node_min_pfn)
			node_min_pfn = start;
		if (end > node_max_pfn)
			node_max_pfn = end;
	}

	if (mem_size_limit && node_max_pfn > mem_size_limit) {
		static int msg_shown = 0;
		if (!msg_shown) {
			msg_shown = 1;
			printk("setup: forcing memory size to %ldK (from %ldK).\n",
			       mem_size_limit << (PAGE_SHIFT - 10),
			       node_max_pfn    << (PAGE_SHIFT - 10));
		}
		node_max_pfn = mem_size_limit;
	}

	if (node_min_pfn >= node_max_pfn)
		return;

	/* Update global {min,max}_low_pfn from node information. */
	if (node_min_pfn < min_low_pfn)
		min_low_pfn = node_min_pfn;
	if (node_max_pfn > max_low_pfn)
		max_pfn = max_low_pfn = node_max_pfn;

	num_physpages += node_max_pfn - node_min_pfn;

#if 0 /* we'll try this one again in a little while */
	/* Cute trick to make sure our local node data is on local memory */
	node_data[nid] = (pg_data_t *)(__va(node_min_pfn << PAGE_SHIFT));
#endif
	/* Quasi-mark the pg_data_t as in-use */
	node_min_pfn += node_datasz;
	if (node_min_pfn >= node_max_pfn) {
		printk(" not enough mem to reserve NODE_DATA");
		return;
	}
	NODE_DATA(nid)->bdata = &bootmem_node_data[nid];

	printk(" Detected node memory:   start %8lu, end %8lu\n",
	       node_min_pfn, node_max_pfn);

	DBGDCONT(" DISCONTIG: node_data[%d]   is at 0x%p\n", nid, NODE_DATA(nid));
	DBGDCONT(" DISCONTIG: NODE_DATA(%d)->bdata is at 0x%p\n", nid, NODE_DATA(nid)->bdata);

	/* Find the bounds of kernel memory.  */
	start_kernel_pfn = PFN_DOWN(KERNEL_START_PHYS);
	end_kernel_pfn = PFN_UP(virt_to_phys(kernel_end));
	bootmap_start = -1;

	if (!nid && (node_max_pfn < end_kernel_pfn || node_min_pfn > start_kernel_pfn))
		panic("kernel loaded out of ram");

	/* Zone start phys-addr must be 2^(MAX_ORDER-1) aligned.
	   Note that we round this down, not up - node memory
	   has much larger alignment than 8Mb, so it's safe. */
	node_min_pfn &= ~((1UL << (MAX_ORDER-1))-1);

	/* We need to know how many physically contiguous pages
	   we'll need for the bootmap.  */
	bootmap_pages = bootmem_bootmap_pages(node_max_pfn-node_min_pfn);

	/* Now find a good region where to allocate the bootmap.  */
	for_each_mem_cluster(memdesc, cluster, i) {
		if (cluster->usage & 3)
			continue;

		start = cluster->start_pfn;
		end = start + cluster->numpages;

		if (start >= node_max_pfn || end <= node_min_pfn)
			continue;

		if (end > node_max_pfn)
			end = node_max_pfn;
		if (start < node_min_pfn)
			start = node_min_pfn;

		if (start < start_kernel_pfn) {
			if (end > end_kernel_pfn
			    && end - end_kernel_pfn >= bootmap_pages) {
				bootmap_start = end_kernel_pfn;
				break;
			} else if (end > start_kernel_pfn)
				end = start_kernel_pfn;
		} else if (start < end_kernel_pfn)
			start = end_kernel_pfn;
		if (end - start >= bootmap_pages) {
			bootmap_start = start;
			break;
		}
	}

	if (bootmap_start == -1)
		panic("couldn't find a contigous place for the bootmap");

	/* Allocate the bootmap and mark the whole MM as reserved.  */
	bootmap_size = init_bootmem_node(NODE_DATA(nid), bootmap_start,
					 node_min_pfn, node_max_pfn);
	DBGDCONT(" bootmap_start %lu, bootmap_size %lu, bootmap_pages %lu\n",
		 bootmap_start, bootmap_size, bootmap_pages);

	/* Mark the free regions.  */
	for_each_mem_cluster(memdesc, cluster, i) {
		if (cluster->usage & 3)
			continue;

		start = cluster->start_pfn;
		end = cluster->start_pfn + cluster->numpages;

		if (start >= node_max_pfn || end <= node_min_pfn)
			continue;

		if (end > node_max_pfn)
			end = node_max_pfn;
		if (start < node_min_pfn)
			start = node_min_pfn;

		if (start < start_kernel_pfn) {
			if (end > end_kernel_pfn) {
				free_bootmem_node(NODE_DATA(nid), PFN_PHYS(start),
					     (PFN_PHYS(start_kernel_pfn)
					      - PFN_PHYS(start)));
				printk(" freeing pages %ld:%ld\n",
				       start, start_kernel_pfn);
				start = end_kernel_pfn;
			} else if (end > start_kernel_pfn)
				end = start_kernel_pfn;
		} else if (start < end_kernel_pfn)
			start = end_kernel_pfn;
		if (start >= end)
			continue;

		free_bootmem_node(NODE_DATA(nid), PFN_PHYS(start), PFN_PHYS(end) - PFN_PHYS(start));
		printk(" freeing pages %ld:%ld\n", start, end);
	}

	/* Reserve the bootmap memory.  */
	reserve_bootmem_node(NODE_DATA(nid), PFN_PHYS(bootmap_start),
			bootmap_size, BOOTMEM_DEFAULT);
	printk(" reserving pages %ld:%ld\n", bootmap_start, bootmap_start+PFN_UP(bootmap_size));

	node_set_online(nid);
}

void __init
setup_memory(void *kernel_end)
{
	int nid;

	show_mem_layout();

	nodes_clear(node_online_map);

	min_low_pfn = ~0UL;
	max_low_pfn = 0UL;
	for (nid = 0; nid < MAX_NUMNODES; nid++)
		setup_memory_node(nid, kernel_end);

#ifdef CONFIG_BLK_DEV_INITRD
	initrd_start = INITRD_START;
	if (initrd_start) {
		extern void *move_initrd(unsigned long);

		initrd_end = initrd_start+INITRD_SIZE;
		printk("Initial ramdisk at: 0x%p (%lu bytes)\n",
		       (void *) initrd_start, INITRD_SIZE);

		if ((void *)initrd_end > phys_to_virt(PFN_PHYS(max_low_pfn))) {
			if (!move_initrd(PFN_PHYS(max_low_pfn)))
				printk("initrd extends beyond end of memory "
				       "(0x%08lx > 0x%p)\ndisabling initrd\n",
				       initrd_end,
				       phys_to_virt(PFN_PHYS(max_low_pfn)));
		} else {
			nid = kvaddr_to_nid(initrd_start);
			reserve_bootmem_node(NODE_DATA(nid),
					     virt_to_phys((void *)initrd_start),
					     INITRD_SIZE, BOOTMEM_DEFAULT);
		}
	}
#endif /* CONFIG_BLK_DEV_INITRD */
}

void __init paging_init(void)
{
	unsigned int    nid;
	unsigned long   zones_size[MAX_NR_ZONES] = {0, };
	unsigned long	dma_local_pfn;

	/*
	 * The old global MAX_DMA_ADDRESS per-arch API doesn't fit
	 * in the NUMA model, for now we convert it to a pfn and
	 * we interpret this pfn as a local per-node information.
	 * This issue isn't very important since none of these machines
	 * have legacy ISA slots anyways.
	 */
	dma_local_pfn = virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;

	for_each_online_node(nid) {
		bootmem_data_t *bdata = &bootmem_node_data[nid];
		unsigned long start_pfn = bdata->node_boot_start >> PAGE_SHIFT;
		unsigned long end_pfn = bdata->node_low_pfn;

		if (dma_local_pfn >= end_pfn - start_pfn)
			zones_size[ZONE_DMA] = end_pfn - start_pfn;
		else {
			zones_size[ZONE_DMA] = dma_local_pfn;
			zones_size[ZONE_NORMAL] = (end_pfn - start_pfn) - dma_local_pfn;
		}
		free_area_init_node(nid, NODE_DATA(nid), zones_size, start_pfn, NULL);
	}

	/* Initialize the kernel's ZERO_PGE. */
	memset((void *)ZERO_PGE, 0, PAGE_SIZE);
}

void __init mem_init(void)
{
	unsigned long codesize, reservedpages, datasize, initsize, pfn;
	extern int page_is_ram(unsigned long) __init;
	extern char _text, _etext, _data, _edata;
	extern char __init_begin, __init_end;
	unsigned long nid, i;
	high_memory = (void *) __va(max_low_pfn << PAGE_SHIFT);

	reservedpages = 0;
	for_each_online_node(nid) {
		/*
		 * This will free up the bootmem, ie, slot 0 memory
		 */
		totalram_pages += free_all_bootmem_node(NODE_DATA(nid));

		pfn = NODE_DATA(nid)->node_start_pfn;
		for (i = 0; i < node_spanned_pages(nid); i++, pfn++)
			if (page_is_ram(pfn) &&
			    PageReserved(nid_page_nr(nid, i)))
				reservedpages++;
	}

	codesize =  (unsigned long) &_etext - (unsigned long) &_text;
	datasize =  (unsigned long) &_edata - (unsigned long) &_data;
	initsize =  (unsigned long) &__init_end - (unsigned long) &__init_begin;

	printk("Memory: %luk/%luk available (%luk kernel code, %luk reserved, "
	       "%luk data, %luk init)\n",
	       (unsigned long)nr_free_pages() << (PAGE_SHIFT-10),
	       num_physpages << (PAGE_SHIFT-10),
	       codesize >> 10,
	       reservedpages << (PAGE_SHIFT-10),
	       datasize >> 10,
	       initsize >> 10);
#if 0
	mem_stress();
#endif
}

void
show_mem(void)
{
	long i,free = 0,total = 0,reserved = 0;
	long shared = 0, cached = 0;
	int nid;

	printk("\nMem-info:\n");
	show_free_areas();
	printk("Free swap:       %6ldkB\n", nr_swap_pages<<(PAGE_SHIFT-10));
	for_each_online_node(nid) {
		unsigned long flags;
		pgdat_resize_lock(NODE_DATA(nid), &flags);
		i = node_spanned_pages(nid);
		while (i-- > 0) {
			struct page *page = nid_page_nr(nid, i);
			total++;
			if (PageReserved(page))
				reserved++;
			else if (PageSwapCache(page))
				cached++;
			else if (!page_count(page))
				free++;
			else
				shared += page_count(page) - 1;
		}
		pgdat_resize_unlock(NODE_DATA(nid), &flags);
	}
	printk("%ld pages of RAM\n",total);
	printk("%ld free pages\n",free);
	printk("%ld reserved pages\n",reserved);
	printk("%ld pages shared\n",shared);
	printk("%ld pages swap cached\n",cached);
}