blob: 7f3028965064b1af467bfe93028e1277006fabc1 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Initialize MMU support.
3 *
4 * Copyright (C) 1998-2003 Hewlett-Packard Co
5 * David Mosberger-Tang <davidm@hpl.hp.com>
6 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07007#include <linux/kernel.h>
8#include <linux/init.h>
9
10#include <linux/bootmem.h>
11#include <linux/efi.h>
12#include <linux/elf.h>
Tejun Heo98e4ae82011-12-08 10:22:08 -080013#include <linux/memblock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/mm.h>
15#include <linux/mmzone.h>
16#include <linux/module.h>
17#include <linux/personality.h>
18#include <linux/reboot.h>
19#include <linux/slab.h>
20#include <linux/swap.h>
21#include <linux/proc_fs.h>
22#include <linux/bitops.h>
Bob Picco139b8302007-01-30 02:11:09 -080023#include <linux/kexec.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <asm/dma.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <asm/io.h>
27#include <asm/machvec.h>
28#include <asm/numa.h>
29#include <asm/patch.h>
30#include <asm/pgalloc.h>
31#include <asm/sal.h>
32#include <asm/sections.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <asm/tlb.h>
34#include <asm/uaccess.h>
35#include <asm/unistd.h>
36#include <asm/mca.h>
37
Linus Torvalds1da177e2005-04-16 15:20:36 -070038extern void ia64_tlb_init (void);
39
40unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL;
41
42#ifdef CONFIG_VIRTUAL_MEM_MAP
Tejun Heo126b3fc2009-10-02 13:28:55 +090043unsigned long VMALLOC_END = VMALLOC_END_INIT;
44EXPORT_SYMBOL(VMALLOC_END);
Linus Torvalds1da177e2005-04-16 15:20:36 -070045struct page *vmem_map;
46EXPORT_SYMBOL(vmem_map);
47#endif
48
Robin Holtfde740e2005-04-25 13:13:16 -070049struct page *zero_page_memmap_ptr; /* map entry for zero page */
Linus Torvalds1da177e2005-04-16 15:20:36 -070050EXPORT_SYMBOL(zero_page_memmap_ptr);
51
Linus Torvalds1da177e2005-04-16 15:20:36 -070052void
KAMEZAWA Hiroyuki954ffcb2007-10-16 01:25:44 -070053__ia64_sync_icache_dcache (pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -070054{
55 unsigned long addr;
56 struct page *page;
57
Linus Torvalds1da177e2005-04-16 15:20:36 -070058 page = pte_page(pte);
59 addr = (unsigned long) page_address(page);
60
61 if (test_bit(PG_arch_1, &page->flags))
62 return; /* i-cache is already coherent with d-cache */
63
Christoph Lameter273988f2008-04-09 13:05:41 -070064 flush_icache_range(addr, addr + (PAGE_SIZE << compound_order(page)));
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 set_bit(PG_arch_1, &page->flags); /* mark page as clean */
66}
67
Jan Beulichcde14bb2007-02-05 18:46:40 -080068/*
69 * Since DMA is i-cache coherent, any (complete) pages that were written via
70 * DMA can be marked as "clean" so that lazy_mmu_prot_update() doesn't have to
71 * flush them when they get mapped into an executable vm-area.
72 */
73void
74dma_mark_clean(void *addr, size_t size)
75{
76 unsigned long pg_addr, end;
77
78 pg_addr = PAGE_ALIGN((unsigned long) addr);
79 end = (unsigned long) addr + size;
80 while (pg_addr + PAGE_SIZE <= end) {
81 struct page *page = virt_to_page(pg_addr);
82 set_bit(PG_arch_1, &page->flags);
83 pg_addr += PAGE_SIZE;
84 }
85}
86
Linus Torvalds1da177e2005-04-16 15:20:36 -070087inline void
88ia64_set_rbs_bot (void)
89{
Jiri Slaby02b763b2010-01-06 16:24:30 +010090 unsigned long stack_size = rlimit_max(RLIMIT_STACK) & -16;
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
92 if (stack_size > MAX_USER_STACK_SIZE)
93 stack_size = MAX_USER_STACK_SIZE;
KAMEZAWA Hiroyuki83d2cd32007-03-23 12:17:46 +090094 current->thread.rbs_bot = PAGE_ALIGN(current->mm->start_stack - stack_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -070095}
96
97/*
98 * This performs some platform-dependent address space initialization.
99 * On IA-64, we want to setup the VM area for the register backing
100 * store (which grows upwards) and install the gateway page which is
101 * used for signal trampolines, etc.
102 */
103void
104ia64_init_addr_space (void)
105{
106 struct vm_area_struct *vma;
107
108 ia64_set_rbs_bot();
109
110 /*
111 * If we're out of memory and kmem_cache_alloc() returns NULL, we simply ignore
112 * the problem. When the process attempts to write to the register backing store
113 * for the first time, it will get a SEGFAULT in this case.
114 */
Robert P. J. Dayc3762222007-02-10 01:45:03 -0800115 vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 if (vma) {
Rik van Riel5beb4932010-03-05 13:42:07 -0800117 INIT_LIST_HEAD(&vma->anon_vma_chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118 vma->vm_mm = current->mm;
119 vma->vm_start = current->thread.rbs_bot & PAGE_MASK;
120 vma->vm_end = vma->vm_start + PAGE_SIZE;
Hugh Dickins46dea3d2005-10-29 18:16:20 -0700121 vma->vm_flags = VM_DATA_DEFAULT_FLAGS|VM_GROWSUP|VM_ACCOUNT;
Coly Li3ed75eb2007-10-18 23:39:15 -0700122 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 down_write(&current->mm->mmap_sem);
124 if (insert_vm_struct(current->mm, vma)) {
125 up_write(&current->mm->mmap_sem);
126 kmem_cache_free(vm_area_cachep, vma);
127 return;
128 }
129 up_write(&current->mm->mmap_sem);
130 }
131
132 /* map NaT-page at address zero to speed up speculative dereferencing of NULL: */
133 if (!(current->personality & MMAP_PAGE_ZERO)) {
Robert P. J. Dayc3762222007-02-10 01:45:03 -0800134 vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 if (vma) {
Rik van Riel5beb4932010-03-05 13:42:07 -0800136 INIT_LIST_HEAD(&vma->anon_vma_chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 vma->vm_mm = current->mm;
138 vma->vm_end = PAGE_SIZE;
139 vma->vm_page_prot = __pgprot(pgprot_val(PAGE_READONLY) | _PAGE_MA_NAT);
Konstantin Khlebnikov314e51b2012-10-08 16:29:02 -0700140 vma->vm_flags = VM_READ | VM_MAYREAD | VM_IO |
141 VM_DONTEXPAND | VM_DONTDUMP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 down_write(&current->mm->mmap_sem);
143 if (insert_vm_struct(current->mm, vma)) {
144 up_write(&current->mm->mmap_sem);
145 kmem_cache_free(vm_area_cachep, vma);
146 return;
147 }
148 up_write(&current->mm->mmap_sem);
149 }
150 }
151}
152
153void
154free_initmem (void)
155{
Jiang Liu11199692013-07-03 15:02:48 -0700156 free_reserved_area(ia64_imva(__init_begin), ia64_imva(__init_end),
Jiang Liudbe67df2013-07-03 15:02:51 -0700157 -1, "unused kernel");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158}
159
Chen, Kenneth Wdae28062006-03-22 16:54:15 -0800160void __init
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161free_initrd_mem (unsigned long start, unsigned long end)
162{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 /*
164 * EFI uses 4KB pages while the kernel can use 4KB or bigger.
165 * Thus EFI and the kernel may have different page sizes. It is
166 * therefore possible to have the initrd share the same page as
167 * the end of the kernel (given current setup).
168 *
169 * To avoid freeing/using the wrong page (kernel sized) we:
170 * - align up the beginning of initrd
171 * - align down the end of initrd
172 *
173 * | |
174 * |=============| a000
175 * | |
176 * | |
177 * | | 9000
178 * |/////////////|
179 * |/////////////|
180 * |=============| 8000
181 * |///INITRD////|
182 * |/////////////|
183 * |/////////////| 7000
184 * | |
185 * |KKKKKKKKKKKKK|
186 * |=============| 6000
187 * |KKKKKKKKKKKKK|
188 * |KKKKKKKKKKKKK|
189 * K=kernel using 8KB pages
190 *
191 * In this example, we must free page 8000 ONLY. So we must align up
192 * initrd_start and keep initrd_end as is.
193 */
194 start = PAGE_ALIGN(start);
195 end = end & PAGE_MASK;
196
197 if (start < end)
198 printk(KERN_INFO "Freeing initrd memory: %ldkB freed\n", (end - start) >> 10);
199
200 for (; start < end; start += PAGE_SIZE) {
201 if (!virt_addr_valid(start))
202 continue;
Jiang Liu66f62592013-04-29 15:06:39 -0700203 free_reserved_page(virt_to_page(start));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 }
205}
206
207/*
208 * This installs a clean page in the kernel's page table.
209 */
Chen, Kenneth Wdae28062006-03-22 16:54:15 -0800210static struct page * __init
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211put_kernel_page (struct page *page, unsigned long address, pgprot_t pgprot)
212{
213 pgd_t *pgd;
214 pud_t *pud;
215 pmd_t *pmd;
216 pte_t *pte;
217
218 if (!PageReserved(page))
219 printk(KERN_ERR "put_kernel_page: page at 0x%p not in reserved memory\n",
220 page_address(page));
221
222 pgd = pgd_offset_k(address); /* note: this is NOT pgd_offset()! */
223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 {
225 pud = pud_alloc(&init_mm, pgd, address);
226 if (!pud)
227 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 pmd = pmd_alloc(&init_mm, pud, address);
229 if (!pmd)
230 goto out;
Hugh Dickins872fec12005-10-29 18:16:21 -0700231 pte = pte_alloc_kernel(pmd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 if (!pte)
233 goto out;
Hugh Dickins872fec12005-10-29 18:16:21 -0700234 if (!pte_none(*pte))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 set_pte(pte, mk_pte(page, pgprot));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 }
Hugh Dickins872fec12005-10-29 18:16:21 -0700238 out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 /* no need for flush_tlb */
240 return page;
241}
242
Chen, Kenneth W914a4ea2006-03-12 09:08:26 -0800243static void __init
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244setup_gate (void)
245{
246 struct page *page;
247
248 /*
David Mosberger-Tangad597bd2005-06-08 10:45:00 -0700249 * Map the gate page twice: once read-only to export the ELF
250 * headers etc. and once execute-only page to enable
251 * privilege-promotion via "epc":
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 */
Luis R. Rodrigueze55645e2015-06-02 11:42:02 -0700253 page = virt_to_page(ia64_imva(__start_gate_section));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 put_kernel_page(page, GATE_ADDR, PAGE_READONLY);
255#ifdef HAVE_BUGGY_SEGREL
Luis R. Rodrigueze55645e2015-06-02 11:42:02 -0700256 page = virt_to_page(ia64_imva(__start_gate_section + PAGE_SIZE));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 put_kernel_page(page, GATE_ADDR + PAGE_SIZE, PAGE_GATE);
258#else
259 put_kernel_page(page, GATE_ADDR + PERCPU_PAGE_SIZE, PAGE_GATE);
David Mosberger-Tangad597bd2005-06-08 10:45:00 -0700260 /* Fill in the holes (if any) with read-only zero pages: */
261 {
262 unsigned long addr;
263
264 for (addr = GATE_ADDR + PAGE_SIZE;
265 addr < GATE_ADDR + PERCPU_PAGE_SIZE;
266 addr += PAGE_SIZE)
267 {
268 put_kernel_page(ZERO_PAGE(0), addr,
269 PAGE_READONLY);
270 put_kernel_page(ZERO_PAGE(0), addr + PERCPU_PAGE_SIZE,
271 PAGE_READONLY);
272 }
273 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274#endif
275 ia64_patch_gate();
276}
277
Andy Lutomirskia6c19df2014-08-08 14:23:40 -0700278static struct vm_area_struct gate_vma;
279
280static int __init gate_vma_init(void)
281{
282 gate_vma.vm_mm = NULL;
283 gate_vma.vm_start = FIXADDR_USER_START;
284 gate_vma.vm_end = FIXADDR_USER_END;
285 gate_vma.vm_flags = VM_READ | VM_MAYREAD | VM_EXEC | VM_MAYEXEC;
286 gate_vma.vm_page_prot = __P101;
287
288 return 0;
289}
290__initcall(gate_vma_init);
291
292struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
293{
294 return &gate_vma;
295}
296
297int in_gate_area_no_mm(unsigned long addr)
298{
299 if ((addr >= FIXADDR_USER_START) && (addr < FIXADDR_USER_END))
300 return 1;
301 return 0;
302}
303
304int in_gate_area(struct mm_struct *mm, unsigned long addr)
305{
306 return in_gate_area_no_mm(addr);
307}
308
Greg Kroah-Hartman5b5e76e2012-12-21 14:05:13 -0800309void ia64_mmu_init(void *my_cpu_data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310{
Chen, Kenneth W00b65982006-10-13 10:08:13 -0700311 unsigned long pta, impl_va_bits;
Greg Kroah-Hartman5b5e76e2012-12-21 14:05:13 -0800312 extern void tlb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
314#ifdef CONFIG_DISABLE_VHPT
315# define VHPT_ENABLE_BIT 0
316#else
317# define VHPT_ENABLE_BIT 1
318#endif
319
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 /*
321 * Check if the virtually mapped linear page table (VMLPT) overlaps with a mapped
322 * address space. The IA-64 architecture guarantees that at least 50 bits of
323 * virtual address space are implemented but if we pick a large enough page size
324 * (e.g., 64KB), the mapped address space is big enough that it will overlap with
325 * VMLPT. I assume that once we run on machines big enough to warrant 64KB pages,
326 * IMPL_VA_MSB will be significantly bigger, so this is unlikely to become a
327 * problem in practice. Alternatively, we could truncate the top of the mapped
328 * address space to not permit mappings that would overlap with the VMLPT.
329 * --davidm 00/12/06
330 */
331# define pte_bits 3
332# define mapped_space_bits (3*(PAGE_SHIFT - pte_bits) + PAGE_SHIFT)
333 /*
334 * The virtual page table has to cover the entire implemented address space within
335 * a region even though not all of this space may be mappable. The reason for
336 * this is that the Access bit and Dirty bit fault handlers perform
337 * non-speculative accesses to the virtual page table, so the address range of the
338 * virtual page table itself needs to be covered by virtual page table.
339 */
340# define vmlpt_bits (impl_va_bits - PAGE_SHIFT + pte_bits)
341# define POW2(n) (1ULL << (n))
342
343 impl_va_bits = ffz(~(local_cpu_data->unimpl_va_mask | (7UL << 61)));
344
345 if (impl_va_bits < 51 || impl_va_bits > 61)
346 panic("CPU has bogus IMPL_VA_MSB value of %lu!\n", impl_va_bits - 1);
Peter Chubb6cf07a82005-08-23 20:07:00 -0700347 /*
348 * mapped_space_bits - PAGE_SHIFT is the total number of ptes we need,
349 * which must fit into "vmlpt_bits - pte_bits" slots. Second half of
350 * the test makes sure that our mapped space doesn't overlap the
351 * unimplemented hole in the middle of the region.
352 */
353 if ((mapped_space_bits - PAGE_SHIFT > vmlpt_bits - pte_bits) ||
354 (mapped_space_bits > impl_va_bits - 1))
355 panic("Cannot build a big enough virtual-linear page table"
356 " to cover mapped address space.\n"
357 " Try using a smaller page size.\n");
358
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359
360 /* place the VMLPT at the end of each page-table mapped region: */
361 pta = POW2(61) - POW2(vmlpt_bits);
362
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 /*
364 * Set the (virtually mapped linear) page table address. Bit
365 * 8 selects between the short and long format, bits 2-7 the
366 * size of the table, and bit 0 whether the VHPT walker is
367 * enabled.
368 */
369 ia64_set_pta(pta | (0 << 8) | (vmlpt_bits << 2) | VHPT_ENABLE_BIT);
370
371 ia64_tlb_init();
372
373#ifdef CONFIG_HUGETLB_PAGE
374 ia64_set_rr(HPAGE_REGION_BASE, HPAGE_SHIFT << 2);
375 ia64_srlz_d();
376#endif
377}
378
379#ifdef CONFIG_VIRTUAL_MEM_MAP
Bob Piccoe44e41d2006-06-28 12:55:43 -0400380int vmemmap_find_next_valid_pfn(int node, int i)
381{
382 unsigned long end_address, hole_next_pfn;
383 unsigned long stop_address;
384 pg_data_t *pgdat = NODE_DATA(node);
385
386 end_address = (unsigned long) &vmem_map[pgdat->node_start_pfn + i];
387 end_address = PAGE_ALIGN(end_address);
Xishi Qiu64080682013-11-12 15:07:17 -0800388 stop_address = (unsigned long) &vmem_map[pgdat_end_pfn(pgdat)];
Bob Piccoe44e41d2006-06-28 12:55:43 -0400389
390 do {
391 pgd_t *pgd;
392 pud_t *pud;
393 pmd_t *pmd;
394 pte_t *pte;
395
396 pgd = pgd_offset_k(end_address);
397 if (pgd_none(*pgd)) {
398 end_address += PGDIR_SIZE;
399 continue;
400 }
401
402 pud = pud_offset(pgd, end_address);
403 if (pud_none(*pud)) {
404 end_address += PUD_SIZE;
405 continue;
406 }
407
408 pmd = pmd_offset(pud, end_address);
409 if (pmd_none(*pmd)) {
410 end_address += PMD_SIZE;
411 continue;
412 }
413
414 pte = pte_offset_kernel(pmd, end_address);
415retry_pte:
416 if (pte_none(*pte)) {
417 end_address += PAGE_SIZE;
418 pte++;
419 if ((end_address < stop_address) &&
420 (end_address != ALIGN(end_address, 1UL << PMD_SHIFT)))
421 goto retry_pte;
422 continue;
423 }
424 /* Found next valid vmem_map page */
425 break;
426 } while (end_address < stop_address);
427
428 end_address = min(end_address, stop_address);
429 end_address = end_address - (unsigned long) vmem_map + sizeof(struct page) - 1;
430 hole_next_pfn = end_address / sizeof(struct page);
431 return hole_next_pfn - pgdat->node_start_pfn;
432}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433
Matthew Wilcoxe088a4a2009-05-22 13:49:49 -0700434int __init create_mem_map_page_table(u64 start, u64 end, void *arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435{
436 unsigned long address, start_page, end_page;
437 struct page *map_start, *map_end;
438 int node;
439 pgd_t *pgd;
440 pud_t *pud;
441 pmd_t *pmd;
442 pte_t *pte;
443
444 map_start = vmem_map + (__pa(start) >> PAGE_SHIFT);
445 map_end = vmem_map + (__pa(end) >> PAGE_SHIFT);
446
447 start_page = (unsigned long) map_start & PAGE_MASK;
448 end_page = PAGE_ALIGN((unsigned long) map_end);
449 node = paddr_to_nid(__pa(start));
450
451 for (address = start_page; address < end_page; address += PAGE_SIZE) {
452 pgd = pgd_offset_k(address);
453 if (pgd_none(*pgd))
454 pgd_populate(&init_mm, pgd, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
455 pud = pud_offset(pgd, address);
456
457 if (pud_none(*pud))
458 pud_populate(&init_mm, pud, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
459 pmd = pmd_offset(pud, address);
460
461 if (pmd_none(*pmd))
462 pmd_populate_kernel(&init_mm, pmd, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
463 pte = pte_offset_kernel(pmd, address);
464
465 if (pte_none(*pte))
466 set_pte(pte, pfn_pte(__pa(alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE)) >> PAGE_SHIFT,
467 PAGE_KERNEL));
468 }
469 return 0;
470}
471
472struct memmap_init_callback_data {
473 struct page *start;
474 struct page *end;
475 int nid;
476 unsigned long zone;
477};
478
Adrian Bunk18b8bef2007-10-29 13:49:47 +0100479static int __meminit
Matthew Wilcoxe088a4a2009-05-22 13:49:49 -0700480virtual_memmap_init(u64 start, u64 end, void *arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481{
482 struct memmap_init_callback_data *args;
483 struct page *map_start, *map_end;
484
485 args = (struct memmap_init_callback_data *) arg;
486 map_start = vmem_map + (__pa(start) >> PAGE_SHIFT);
487 map_end = vmem_map + (__pa(end) >> PAGE_SHIFT);
488
489 if (map_start < args->start)
490 map_start = args->start;
491 if (map_end > args->end)
492 map_end = args->end;
493
494 /*
495 * We have to initialize "out of bounds" struct page elements that fit completely
496 * on the same pages that were allocated for the "in bounds" elements because they
497 * may be referenced later (and found to be "reserved").
498 */
499 map_start -= ((unsigned long) map_start & (PAGE_SIZE - 1)) / sizeof(struct page);
500 map_end += ((PAGE_ALIGN((unsigned long) map_end) - (unsigned long) map_end)
501 / sizeof(struct page));
502
503 if (map_start < map_end)
504 memmap_init_zone((unsigned long)(map_end - map_start),
Dave Hansena2f3aa022007-01-10 23:15:30 -0800505 args->nid, args->zone, page_to_pfn(map_start),
506 MEMMAP_EARLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 return 0;
508}
509
Adrian Bunk18b8bef2007-10-29 13:49:47 +0100510void __meminit
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511memmap_init (unsigned long size, int nid, unsigned long zone,
512 unsigned long start_pfn)
513{
514 if (!vmem_map)
Dave Hansena2f3aa022007-01-10 23:15:30 -0800515 memmap_init_zone(size, nid, zone, start_pfn, MEMMAP_EARLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 else {
517 struct page *start;
518 struct memmap_init_callback_data args;
519
520 start = pfn_to_page(start_pfn);
521 args.start = start;
522 args.end = start + size;
523 args.nid = nid;
524 args.zone = zone;
525
526 efi_memmap_walk(virtual_memmap_init, &args);
527 }
528}
529
530int
531ia64_pfn_valid (unsigned long pfn)
532{
533 char byte;
534 struct page *pg = pfn_to_page(pfn);
535
536 return (__get_user(byte, (char __user *) pg) == 0)
537 && ((((u64)pg & PAGE_MASK) == (((u64)(pg + 1) - 1) & PAGE_MASK))
538 || (__get_user(byte, (char __user *) (pg + 1) - 1) == 0));
539}
540EXPORT_SYMBOL(ia64_pfn_valid);
541
Matthew Wilcoxe088a4a2009-05-22 13:49:49 -0700542int __init find_largest_hole(u64 start, u64 end, void *arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543{
544 u64 *max_gap = arg;
545
546 static u64 last_end = PAGE_OFFSET;
547
548 /* NOTE: this algorithm assumes efi memmap table is ordered */
549
550 if (*max_gap < (start - last_end))
551 *max_gap = start - last_end;
552 last_end = end;
553 return 0;
554}
Mel Gorman05e0caa2006-09-27 01:49:54 -0700555
Bob Picco139b8302007-01-30 02:11:09 -0800556#endif /* CONFIG_VIRTUAL_MEM_MAP */
557
Matthew Wilcoxe088a4a2009-05-22 13:49:49 -0700558int __init register_active_ranges(u64 start, u64 len, int nid)
Mel Gorman05e0caa2006-09-27 01:49:54 -0700559{
Zoltan Menyhart98075d22008-04-11 15:21:35 -0700560 u64 end = start + len;
Bob Picco139b8302007-01-30 02:11:09 -0800561
Bob Picco139b8302007-01-30 02:11:09 -0800562#ifdef CONFIG_KEXEC
563 if (start > crashk_res.start && start < crashk_res.end)
564 start = crashk_res.end;
565 if (end > crashk_res.start && end < crashk_res.end)
566 end = crashk_res.start;
567#endif
568
569 if (start < end)
Tejun Heo98e4ae82011-12-08 10:22:08 -0800570 memblock_add_node(__pa(start), end - start, nid);
Mel Gorman05e0caa2006-09-27 01:49:54 -0700571 return 0;
572}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573
Zou Nan haia3f5c332007-03-20 13:41:57 -0700574int
Matthew Wilcoxe088a4a2009-05-22 13:49:49 -0700575find_max_min_low_pfn (u64 start, u64 end, void *arg)
Zou Nan haia3f5c332007-03-20 13:41:57 -0700576{
577 unsigned long pfn_start, pfn_end;
578#ifdef CONFIG_FLATMEM
579 pfn_start = (PAGE_ALIGN(__pa(start))) >> PAGE_SHIFT;
580 pfn_end = (PAGE_ALIGN(__pa(end - 1))) >> PAGE_SHIFT;
581#else
582 pfn_start = GRANULEROUNDDOWN(__pa(start)) >> PAGE_SHIFT;
583 pfn_end = GRANULEROUNDUP(__pa(end - 1)) >> PAGE_SHIFT;
584#endif
585 min_low_pfn = min(min_low_pfn, pfn_start);
586 max_low_pfn = max(max_low_pfn, pfn_end);
587 return 0;
588}
589
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590/*
591 * Boot command-line option "nolwsys" can be used to disable the use of any light-weight
592 * system call handler. When this option is in effect, all fsyscalls will end up bubbling
593 * down into the kernel and calling the normal (heavy-weight) syscall handler. This is
594 * useful for performance testing, but conceivably could also come in handy for debugging
595 * purposes.
596 */
597
Chen, Kenneth W03906ea2006-03-12 09:10:59 -0800598static int nolwsys __initdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599
600static int __init
601nolwsys_setup (char *s)
602{
603 nolwsys = 1;
604 return 1;
605}
606
607__setup("nolwsys", nolwsys_setup);
608
Chen, Kenneth Wdae28062006-03-22 16:54:15 -0800609void __init
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610mem_init (void)
611{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 int i;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613
Robin Holtfde740e2005-04-25 13:13:16 -0700614 BUG_ON(PTRS_PER_PGD * sizeof(pgd_t) != PAGE_SIZE);
615 BUG_ON(PTRS_PER_PMD * sizeof(pmd_t) != PAGE_SIZE);
616 BUG_ON(PTRS_PER_PTE * sizeof(pte_t) != PAGE_SIZE);
617
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618#ifdef CONFIG_PCI
619 /*
620 * This needs to be called _after_ the command line has been parsed but _before_
621 * any drivers that may need the PCI DMA interface are initialized or bootmem has
622 * been freed.
623 */
624 platform_dma_init();
625#endif
626
Bob Picco2d4b1fa2005-10-04 15:13:57 -0400627#ifdef CONFIG_FLATMEM
Stoyan Gaydarov80a03e22009-03-10 00:10:30 -0500628 BUG_ON(!mem_map);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629#endif
630
Jiang Liub57b63a2013-07-03 15:04:25 -0700631 set_max_mapnr(max_low_pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 high_memory = __va(max_low_pfn * PAGE_SIZE);
Jiang Liub57b63a2013-07-03 15:04:25 -0700633 free_all_bootmem();
Jiang Liude4bcdd2013-07-03 15:03:58 -0700634 mem_init_print_info(NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635
636 /*
637 * For fsyscall entrpoints with no light-weight handler, use the ordinary
638 * (heavy-weight) handler, but mark it by setting bit 0, so the fsyscall entry
639 * code can tell them apart.
640 */
641 for (i = 0; i < NR_syscalls; ++i) {
Luis R. Rodrigueze55645e2015-06-02 11:42:02 -0700642 extern unsigned long fsyscall_table[NR_syscalls];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 extern unsigned long sys_call_table[NR_syscalls];
644
645 if (!fsyscall_table[i] || nolwsys)
646 fsyscall_table[i] = sys_call_table[i] | 1;
647 }
648 setup_gate();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649}
Yasunori Goto1681b8e2006-01-07 11:50:38 +0900650
651#ifdef CONFIG_MEMORY_HOTPLUG
Yasunori Gotobc02af92006-06-27 02:53:30 -0700652int arch_add_memory(int nid, u64 start, u64 size)
Yasunori Goto1681b8e2006-01-07 11:50:38 +0900653{
654 pg_data_t *pgdat;
655 struct zone *zone;
656 unsigned long start_pfn = start >> PAGE_SHIFT;
657 unsigned long nr_pages = size >> PAGE_SHIFT;
658 int ret;
659
Yasunori Gotobc02af92006-06-27 02:53:30 -0700660 pgdat = NODE_DATA(nid);
Yasunori Goto1681b8e2006-01-07 11:50:38 +0900661
Wang Naned562ae2014-08-06 16:07:43 -0700662 zone = pgdat->node_zones +
663 zone_for_memory(nid, start, size, ZONE_NORMAL);
Gary Hadec04fc582009-01-06 14:39:14 -0800664 ret = __add_pages(nid, zone, start_pfn, nr_pages);
Yasunori Goto1681b8e2006-01-07 11:50:38 +0900665
666 if (ret)
667 printk("%s: Problem encountered in __add_pages() as ret=%d\n",
Harvey Harrisond4ed8082008-03-04 15:15:00 -0800668 __func__, ret);
Yasunori Goto1681b8e2006-01-07 11:50:38 +0900669
670 return ret;
671}
Wen Congyang24d335c2013-02-22 16:32:58 -0800672
673#ifdef CONFIG_MEMORY_HOTREMOVE
674int arch_remove_memory(u64 start, u64 size)
675{
676 unsigned long start_pfn = start >> PAGE_SHIFT;
677 unsigned long nr_pages = size >> PAGE_SHIFT;
678 struct zone *zone;
679 int ret;
680
681 zone = page_zone(pfn_to_page(start_pfn));
682 ret = __remove_pages(zone, start_pfn, nr_pages);
683 if (ret)
684 pr_warn("%s: Problem encountered in __remove_pages() as"
685 " ret=%d\n", __func__, ret);
686
687 return ret;
688}
689#endif
Yasunori Goto1681b8e2006-01-07 11:50:38 +0900690#endif
Huang, Xiaolan839052d2008-05-15 10:18:41 +0800691
Mel Gormanaec6a882014-01-21 15:49:13 -0800692/**
693 * show_mem - give short summary of memory stats
694 *
695 * Shows a simple page count of reserved and used pages in the system.
696 * For discontig machines, it does this on a per-pgdat basis.
697 */
698void show_mem(unsigned int filter)
699{
700 int total_reserved = 0;
701 unsigned long total_present = 0;
702 pg_data_t *pgdat;
703
704 printk(KERN_INFO "Mem-info:\n");
705 show_free_areas(filter);
706 printk(KERN_INFO "Node memory in pages:\n");
707 for_each_online_pgdat(pgdat) {
708 unsigned long present;
709 unsigned long flags;
710 int reserved = 0;
711 int nid = pgdat->node_id;
712 int zoneid;
713
714 if (skip_free_areas_node(filter, nid))
715 continue;
716 pgdat_resize_lock(pgdat, &flags);
717
718 for (zoneid = 0; zoneid < MAX_NR_ZONES; zoneid++) {
719 struct zone *zone = &pgdat->node_zones[zoneid];
720 if (!populated_zone(zone))
721 continue;
722
723 reserved += zone->present_pages - zone->managed_pages;
724 }
725 present = pgdat->node_present_pages;
726
727 pgdat_resize_unlock(pgdat, &flags);
728 total_present += present;
729 total_reserved += reserved;
730 printk(KERN_INFO "Node %4d: RAM: %11ld, rsvd: %8d, ",
731 nid, present, reserved);
732 }
733 printk(KERN_INFO "%ld pages of RAM\n", total_present);
734 printk(KERN_INFO "%d reserved pages\n", total_reserved);
735 printk(KERN_INFO "Total of %ld pages in page table cache\n",
736 quicklist_total_size());
737 printk(KERN_INFO "%ld free buffer pages\n", nr_free_buffer_pages());
738}