vmem.c 8.9 KB
Newer Older
H
Heiko Carstens 已提交
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
/*
 *  arch/s390/mm/vmem.c
 *
 *    Copyright IBM Corp. 2006
 *    Author(s): Heiko Carstens <heiko.carstens@de.ibm.com>
 */

#include <linux/bootmem.h>
#include <linux/pfn.h>
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/list.h>
#include <asm/pgalloc.h>
#include <asm/pgtable.h>
#include <asm/setup.h>
#include <asm/tlbflush.h>

static DEFINE_MUTEX(vmem_mutex);

struct memory_segment {
	struct list_head list;
	unsigned long start;
	unsigned long size;
};

static LIST_HEAD(mem_segs);

28 29
void __meminit memmap_init(unsigned long size, int nid, unsigned long zone,
			   unsigned long start_pfn)
H
Heiko Carstens 已提交
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
{
	struct page *start, *end;
	struct page *map_start, *map_end;
	int i;

	start = pfn_to_page(start_pfn);
	end = start + size;

	for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
		unsigned long cstart, cend;

		cstart = PFN_DOWN(memory_chunk[i].addr);
		cend = cstart + PFN_DOWN(memory_chunk[i].size);

		map_start = mem_map + cstart;
		map_end = mem_map + cend;

		if (map_start < start)
			map_start = start;
		if (map_end > end)
			map_end = end;

		map_start -= ((unsigned long) map_start & (PAGE_SIZE - 1))
			/ sizeof(struct page);
		map_end += ((PFN_ALIGN((unsigned long) map_end)
			     - (unsigned long) map_end)
			    / sizeof(struct page));

		if (map_start < map_end)
			memmap_init_zone((unsigned long)(map_end - map_start),
D
Dave Hansen 已提交
60 61
					 nid, zone, page_to_pfn(map_start),
					 MEMMAP_EARLY);
H
Heiko Carstens 已提交
62 63 64
	}
}

65
static void __ref *vmem_alloc_pages(unsigned int order)
H
Heiko Carstens 已提交
66 67 68 69 70 71
{
	if (slab_is_available())
		return (void *)__get_free_pages(GFP_KERNEL, order);
	return alloc_bootmem_pages((1 << order) * PAGE_SIZE);
}

72 73 74 75 76 77 78 79 80 81 82 83 84
static inline pud_t *vmem_pud_alloc(void)
{
	pud_t *pud = NULL;

#ifdef CONFIG_64BIT
	pud = vmem_alloc_pages(2);
	if (!pud)
		return NULL;
	pud_val(*pud) = _REGION3_ENTRY_EMPTY;
	memcpy(pud + 1, pud, (PTRS_PER_PUD - 1)*sizeof(pud_t));
#endif
	return pud;
}
M
Martin Schwidefsky 已提交
85

H
Heiko Carstens 已提交
86 87
static inline pmd_t *vmem_pmd_alloc(void)
{
88
	pmd_t *pmd = NULL;
H
Heiko Carstens 已提交
89

90 91
#ifdef CONFIG_64BIT
	pmd = vmem_alloc_pages(2);
H
Heiko Carstens 已提交
92 93
	if (!pmd)
		return NULL;
94 95
	clear_table((unsigned long *) pmd, _SEGMENT_ENTRY_EMPTY, PAGE_SIZE*4);
#endif
H
Heiko Carstens 已提交
96 97 98
	return pmd;
}

99
static pte_t __init_refok *vmem_pte_alloc(void)
H
Heiko Carstens 已提交
100
{
101
	pte_t *pte;
H
Heiko Carstens 已提交
102

103 104 105 106
	if (slab_is_available())
		pte = (pte_t *) page_table_alloc(&init_mm);
	else
		pte = alloc_bootmem(PTRS_PER_PTE * sizeof(pte_t));
H
Heiko Carstens 已提交
107 108
	if (!pte)
		return NULL;
109 110
	clear_table((unsigned long *) pte, _PAGE_TYPE_EMPTY,
		    PTRS_PER_PTE * sizeof(pte_t));
H
Heiko Carstens 已提交
111 112 113 114 115 116 117 118 119 120
	return pte;
}

/*
 * Add a physical memory range to the 1:1 mapping.
 */
static int vmem_add_range(unsigned long start, unsigned long size)
{
	unsigned long address;
	pgd_t *pg_dir;
M
Martin Schwidefsky 已提交
121
	pud_t *pu_dir;
H
Heiko Carstens 已提交
122 123 124 125 126 127 128 129
	pmd_t *pm_dir;
	pte_t *pt_dir;
	pte_t  pte;
	int ret = -ENOMEM;

	for (address = start; address < start + size; address += PAGE_SIZE) {
		pg_dir = pgd_offset_k(address);
		if (pgd_none(*pg_dir)) {
M
Martin Schwidefsky 已提交
130 131 132 133 134 135 136 137
			pu_dir = vmem_pud_alloc();
			if (!pu_dir)
				goto out;
			pgd_populate_kernel(&init_mm, pg_dir, pu_dir);
		}

		pu_dir = pud_offset(pg_dir, address);
		if (pud_none(*pu_dir)) {
H
Heiko Carstens 已提交
138 139 140
			pm_dir = vmem_pmd_alloc();
			if (!pm_dir)
				goto out;
M
Martin Schwidefsky 已提交
141
			pud_populate_kernel(&init_mm, pu_dir, pm_dir);
H
Heiko Carstens 已提交
142 143
		}

M
Martin Schwidefsky 已提交
144
		pm_dir = pmd_offset(pu_dir, address);
H
Heiko Carstens 已提交
145 146 147 148 149 150 151 152 153
		if (pmd_none(*pm_dir)) {
			pt_dir = vmem_pte_alloc();
			if (!pt_dir)
				goto out;
			pmd_populate_kernel(&init_mm, pm_dir, pt_dir);
		}

		pt_dir = pte_offset_kernel(pm_dir, address);
		pte = pfn_pte(address >> PAGE_SHIFT, PAGE_KERNEL);
G
Gerald Schaefer 已提交
154
		*pt_dir = pte;
H
Heiko Carstens 已提交
155 156 157 158 159 160 161 162 163 164 165 166 167 168 169
	}
	ret = 0;
out:
	flush_tlb_kernel_range(start, start + size);
	return ret;
}

/*
 * Remove a physical memory range from the 1:1 mapping.
 * Currently only invalidates page table entries.
 */
static void vmem_remove_range(unsigned long start, unsigned long size)
{
	unsigned long address;
	pgd_t *pg_dir;
M
Martin Schwidefsky 已提交
170
	pud_t *pu_dir;
H
Heiko Carstens 已提交
171 172 173 174 175 176 177
	pmd_t *pm_dir;
	pte_t *pt_dir;
	pte_t  pte;

	pte_val(pte) = _PAGE_TYPE_EMPTY;
	for (address = start; address < start + size; address += PAGE_SIZE) {
		pg_dir = pgd_offset_k(address);
M
Martin Schwidefsky 已提交
178 179
		pu_dir = pud_offset(pg_dir, address);
		if (pud_none(*pu_dir))
H
Heiko Carstens 已提交
180
			continue;
M
Martin Schwidefsky 已提交
181
		pm_dir = pmd_offset(pu_dir, address);
H
Heiko Carstens 已提交
182 183 184
		if (pmd_none(*pm_dir))
			continue;
		pt_dir = pte_offset_kernel(pm_dir, address);
G
Gerald Schaefer 已提交
185
		*pt_dir = pte;
H
Heiko Carstens 已提交
186 187 188 189 190 191 192 193 194 195 196 197
	}
	flush_tlb_kernel_range(start, start + size);
}

/*
 * Add a backed mem_map array to the virtual mem_map array.
 */
static int vmem_add_mem_map(unsigned long start, unsigned long size)
{
	unsigned long address, start_addr, end_addr;
	struct page *map_start, *map_end;
	pgd_t *pg_dir;
M
Martin Schwidefsky 已提交
198
	pud_t *pu_dir;
H
Heiko Carstens 已提交
199 200 201 202 203
	pmd_t *pm_dir;
	pte_t *pt_dir;
	pte_t  pte;
	int ret = -ENOMEM;

204 205
	map_start = VMEM_MAP + PFN_DOWN(start);
	map_end	= VMEM_MAP + PFN_DOWN(start + size);
H
Heiko Carstens 已提交
206 207 208 209 210 211 212

	start_addr = (unsigned long) map_start & PAGE_MASK;
	end_addr = PFN_ALIGN((unsigned long) map_end);

	for (address = start_addr; address < end_addr; address += PAGE_SIZE) {
		pg_dir = pgd_offset_k(address);
		if (pgd_none(*pg_dir)) {
M
Martin Schwidefsky 已提交
213 214 215 216 217 218 219 220
			pu_dir = vmem_pud_alloc();
			if (!pu_dir)
				goto out;
			pgd_populate_kernel(&init_mm, pg_dir, pu_dir);
		}

		pu_dir = pud_offset(pg_dir, address);
		if (pud_none(*pu_dir)) {
H
Heiko Carstens 已提交
221 222 223
			pm_dir = vmem_pmd_alloc();
			if (!pm_dir)
				goto out;
M
Martin Schwidefsky 已提交
224
			pud_populate_kernel(&init_mm, pu_dir, pm_dir);
H
Heiko Carstens 已提交
225 226
		}

M
Martin Schwidefsky 已提交
227
		pm_dir = pmd_offset(pu_dir, address);
H
Heiko Carstens 已提交
228 229 230 231 232 233 234 235 236 237 238 239 240 241 242
		if (pmd_none(*pm_dir)) {
			pt_dir = vmem_pte_alloc();
			if (!pt_dir)
				goto out;
			pmd_populate_kernel(&init_mm, pm_dir, pt_dir);
		}

		pt_dir = pte_offset_kernel(pm_dir, address);
		if (pte_none(*pt_dir)) {
			unsigned long new_page;

			new_page =__pa(vmem_alloc_pages(0));
			if (!new_page)
				goto out;
			pte = pfn_pte(new_page >> PAGE_SHIFT, PAGE_KERNEL);
G
Gerald Schaefer 已提交
243
			*pt_dir = pte;
H
Heiko Carstens 已提交
244 245 246 247 248 249 250 251 252 253 254 255
		}
	}
	ret = 0;
out:
	flush_tlb_kernel_range(start_addr, end_addr);
	return ret;
}

static int vmem_add_mem(unsigned long start, unsigned long size)
{
	int ret;

256
	ret = vmem_add_mem_map(start, size);
H
Heiko Carstens 已提交
257 258
	if (ret)
		return ret;
259
	return vmem_add_range(start, size);
H
Heiko Carstens 已提交
260 261 262 263 264 265 266 267 268 269
}

/*
 * Add memory segment to the segment list if it doesn't overlap with
 * an already present segment.
 */
static int insert_memory_segment(struct memory_segment *seg)
{
	struct memory_segment *tmp;

270
	if (seg->start + seg->size >= VMEM_MAX_PHYS ||
H
Heiko Carstens 已提交
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
	    seg->start + seg->size < seg->start)
		return -ERANGE;

	list_for_each_entry(tmp, &mem_segs, list) {
		if (seg->start >= tmp->start + tmp->size)
			continue;
		if (seg->start + seg->size <= tmp->start)
			continue;
		return -ENOSPC;
	}
	list_add(&seg->list, &mem_segs);
	return 0;
}

/*
 * Remove memory segment from the segment list.
 */
static void remove_memory_segment(struct memory_segment *seg)
{
	list_del(&seg->list);
}

static void __remove_shared_memory(struct memory_segment *seg)
{
	remove_memory_segment(seg);
	vmem_remove_range(seg->start, seg->size);
}

int remove_shared_memory(unsigned long start, unsigned long size)
{
	struct memory_segment *seg;
	int ret;

	mutex_lock(&vmem_mutex);

	ret = -ENOENT;
	list_for_each_entry(seg, &mem_segs, list) {
		if (seg->start == start && seg->size == size)
			break;
	}

	if (seg->start != start || seg->size != size)
		goto out;

	ret = 0;
	__remove_shared_memory(seg);
	kfree(seg);
out:
	mutex_unlock(&vmem_mutex);
	return ret;
}

int add_shared_memory(unsigned long start, unsigned long size)
{
	struct memory_segment *seg;
	struct page *page;
	unsigned long pfn, num_pfn, end_pfn;
	int ret;

	mutex_lock(&vmem_mutex);
	ret = -ENOMEM;
	seg = kzalloc(sizeof(*seg), GFP_KERNEL);
	if (!seg)
		goto out;
	seg->start = start;
	seg->size = size;

	ret = insert_memory_segment(seg);
	if (ret)
		goto out_free;

	ret = vmem_add_mem(start, size);
	if (ret)
		goto out_remove;

	pfn = PFN_DOWN(start);
	num_pfn = PFN_DOWN(size);
	end_pfn = pfn + num_pfn;

	page = pfn_to_page(pfn);
	memset(page, 0, num_pfn * sizeof(struct page));

	for (; pfn < end_pfn; pfn++) {
		page = pfn_to_page(pfn);
		init_page_count(page);
		reset_page_mapcount(page);
		SetPageReserved(page);
		INIT_LIST_HEAD(&page->lru);
	}
	goto out;

out_remove:
	__remove_shared_memory(seg);
out_free:
	kfree(seg);
out:
	mutex_unlock(&vmem_mutex);
	return ret;
}

/*
 * map whole physical memory to virtual memory (identity mapping)
373 374
 * we reserve enough space in the vmalloc area for vmemmap to hotplug
 * additional memory segments.
H
Heiko Carstens 已提交
375 376 377 378 379
 */
void __init vmem_map_init(void)
{
	int i;

380 381 382
	INIT_LIST_HEAD(&init_mm.context.crst_list);
	INIT_LIST_HEAD(&init_mm.context.pgtable_list);
	init_mm.context.noexec = 0;
383
	NODE_DATA(0)->node_mem_map = VMEM_MAP;
H
Heiko Carstens 已提交
384 385 386 387 388 389 390 391 392 393 394 395 396 397
	for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++)
		vmem_add_mem(memory_chunk[i].addr, memory_chunk[i].size);
}

/*
 * Convert memory chunk array to a memory segment list so there is a single
 * list that contains both r/w memory and shared memory segments.
 */
static int __init vmem_convert_memory_chunk(void)
{
	struct memory_segment *seg;
	int i;

	mutex_lock(&vmem_mutex);
398
	for (i = 0; i < MEMORY_CHUNKS; i++) {
H
Heiko Carstens 已提交
399 400 401 402 403 404 405 406 407 408 409 410 411 412
		if (!memory_chunk[i].size)
			continue;
		seg = kzalloc(sizeof(*seg), GFP_KERNEL);
		if (!seg)
			panic("Out of memory...\n");
		seg->start = memory_chunk[i].addr;
		seg->size = memory_chunk[i].size;
		insert_memory_segment(seg);
	}
	mutex_unlock(&vmem_mutex);
	return 0;
}

core_initcall(vmem_convert_memory_chunk);