blob: cbbc1377481981ff26e065dbade356ec90ac0728 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/mm/nommu.c
3 *
4 * Replacement code for mm functions to support CPU's that don't
5 * have any form of memory management unit (thus no virtual memory).
6 *
7 * See Documentation/nommu-mmap.txt
8 *
9 * Copyright (c) 2004-2005 David Howells <dhowells@redhat.com>
10 * Copyright (c) 2000-2003 David McCullough <davidm@snapgear.com>
11 * Copyright (c) 2000-2001 D Jeff Dionne <jeff@uClinux.org>
12 * Copyright (c) 2002 Greg Ungerer <gerg@snapgear.com>
13 */
14
15#include <linux/mm.h>
16#include <linux/mman.h>
17#include <linux/swap.h>
18#include <linux/file.h>
19#include <linux/highmem.h>
20#include <linux/pagemap.h>
21#include <linux/slab.h>
22#include <linux/vmalloc.h>
23#include <linux/ptrace.h>
24#include <linux/blkdev.h>
25#include <linux/backing-dev.h>
26#include <linux/mount.h>
27#include <linux/personality.h>
28#include <linux/security.h>
29#include <linux/syscalls.h>
30
31#include <asm/uaccess.h>
32#include <asm/tlb.h>
33#include <asm/tlbflush.h>
34
35void *high_memory;
36struct page *mem_map;
37unsigned long max_mapnr;
38unsigned long num_physpages;
39unsigned long askedalloc, realalloc;
40atomic_t vm_committed_space = ATOMIC_INIT(0);
41int sysctl_overcommit_memory = OVERCOMMIT_GUESS; /* heuristic overcommit */
42int sysctl_overcommit_ratio = 50; /* default is 50% */
43int sysctl_max_map_count = DEFAULT_MAX_MAP_COUNT;
44int heap_stack_gap = 0;
45
46EXPORT_SYMBOL(mem_map);
Linus Torvalds1da177e2005-04-16 15:20:36 -070047EXPORT_SYMBOL(__vm_enough_memory);
48
49/* list of shareable VMAs */
50struct rb_root nommu_vma_tree = RB_ROOT;
51DECLARE_RWSEM(nommu_vma_sem);
52
53struct vm_operations_struct generic_file_vm_ops = {
54};
55
Greg Ungerer66aa2b42005-09-12 11:18:10 +100056EXPORT_SYMBOL(vfree);
57EXPORT_SYMBOL(vmalloc_to_page);
58EXPORT_SYMBOL(vmalloc_32);
Luke Yang7a9166e2006-02-20 18:28:07 -080059EXPORT_SYMBOL(vmap);
60EXPORT_SYMBOL(vunmap);
Greg Ungerer66aa2b42005-09-12 11:18:10 +100061
Linus Torvalds1da177e2005-04-16 15:20:36 -070062/*
63 * Handle all mappings that got truncated by a "truncate()"
64 * system call.
65 *
66 * NOTE! We have to be ready to update the memory sharing
67 * between the file and the memory map for a potential last
68 * incomplete page. Ugly, but necessary.
69 */
70int vmtruncate(struct inode *inode, loff_t offset)
71{
72 struct address_space *mapping = inode->i_mapping;
73 unsigned long limit;
74
75 if (inode->i_size < offset)
76 goto do_expand;
77 i_size_write(inode, offset);
78
79 truncate_inode_pages(mapping, offset);
80 goto out_truncate;
81
82do_expand:
83 limit = current->signal->rlim[RLIMIT_FSIZE].rlim_cur;
84 if (limit != RLIM_INFINITY && offset > limit)
85 goto out_sig;
86 if (offset > inode->i_sb->s_maxbytes)
87 goto out;
88 i_size_write(inode, offset);
89
90out_truncate:
91 if (inode->i_op && inode->i_op->truncate)
92 inode->i_op->truncate(inode);
93 return 0;
94out_sig:
95 send_sig(SIGXFSZ, current, 0);
96out:
97 return -EFBIG;
98}
99
100EXPORT_SYMBOL(vmtruncate);
101
102/*
103 * Return the total memory allocated for this pointer, not
104 * just what the caller asked for.
105 *
106 * Doesn't have to be accurate, i.e. may have races.
107 */
108unsigned int kobjsize(const void *objp)
109{
110 struct page *page;
111
112 if (!objp || !((page = virt_to_page(objp))))
113 return 0;
114
115 if (PageSlab(page))
116 return ksize(objp);
117
118 BUG_ON(page->index < 0);
119 BUG_ON(page->index >= MAX_ORDER);
120
121 return (PAGE_SIZE << page->index);
122}
123
124/*
David Howells7b4d5b82006-09-27 01:50:18 -0700125 * get a list of pages in an address range belonging to the specified process
126 * and indicate the VMA that covers each page
127 * - this is potentially dodgy as we may end incrementing the page count of a
128 * slab page or a secondary page from a compound page
129 * - don't permit access to VMAs that don't support it, such as I/O mappings
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 */
131int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
132 unsigned long start, int len, int write, int force,
133 struct page **pages, struct vm_area_struct **vmas)
134{
Sonic Zhang910e46d2006-09-27 01:50:17 -0700135 struct vm_area_struct *vma;
David Howells7b4d5b82006-09-27 01:50:18 -0700136 unsigned long vm_flags;
137 int i;
138
139 /* calculate required read or write permissions.
140 * - if 'force' is set, we only require the "MAY" flags.
141 */
142 vm_flags = write ? (VM_WRITE | VM_MAYWRITE) : (VM_READ | VM_MAYREAD);
143 vm_flags &= force ? (VM_MAYREAD | VM_MAYWRITE) : (VM_READ | VM_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144
145 for (i = 0; i < len; i++) {
Sonic Zhang910e46d2006-09-27 01:50:17 -0700146 vma = find_vma(mm, start);
David Howells7b4d5b82006-09-27 01:50:18 -0700147 if (!vma)
148 goto finish_or_fault;
149
150 /* protect what we can, including chardevs */
151 if (vma->vm_flags & (VM_IO | VM_PFNMAP) ||
152 !(vm_flags & vma->vm_flags))
153 goto finish_or_fault;
Sonic Zhang910e46d2006-09-27 01:50:17 -0700154
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 if (pages) {
156 pages[i] = virt_to_page(start);
157 if (pages[i])
158 page_cache_get(pages[i]);
159 }
160 if (vmas)
Sonic Zhang910e46d2006-09-27 01:50:17 -0700161 vmas[i] = vma;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 start += PAGE_SIZE;
163 }
David Howells7b4d5b82006-09-27 01:50:18 -0700164
165 return i;
166
167finish_or_fault:
168 return i ? : -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169}
170
Greg Ungerer66aa2b42005-09-12 11:18:10 +1000171EXPORT_SYMBOL(get_user_pages);
172
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173DEFINE_RWLOCK(vmlist_lock);
174struct vm_struct *vmlist;
175
176void vfree(void *addr)
177{
178 kfree(addr);
179}
180
Al Virodd0fc662005-10-07 07:46:04 +0100181void *__vmalloc(unsigned long size, gfp_t gfp_mask, pgprot_t prot)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182{
183 /*
184 * kmalloc doesn't like __GFP_HIGHMEM for some reason
185 */
Nick Piggin84097512006-03-22 00:08:34 -0800186 return kmalloc(size, (gfp_mask | __GFP_COMP) & ~__GFP_HIGHMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187}
188
189struct page * vmalloc_to_page(void *addr)
190{
191 return virt_to_page(addr);
192}
193
194unsigned long vmalloc_to_pfn(void *addr)
195{
196 return page_to_pfn(virt_to_page(addr));
197}
198
199
200long vread(char *buf, char *addr, unsigned long count)
201{
202 memcpy(buf, addr, count);
203 return count;
204}
205
206long vwrite(char *buf, char *addr, unsigned long count)
207{
208 /* Don't allow overflow */
209 if ((unsigned long) addr + count < count)
210 count = -(unsigned long) addr;
211
212 memcpy(addr, buf, count);
213 return(count);
214}
215
216/*
217 * vmalloc - allocate virtually continguos memory
218 *
219 * @size: allocation size
220 *
221 * Allocate enough pages to cover @size from the page level
222 * allocator and map them into continguos kernel virtual space.
223 *
Michael Opdenackerc1c88972006-10-03 23:21:02 +0200224 * For tight control over page level allocator and protection flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 * use __vmalloc() instead.
226 */
227void *vmalloc(unsigned long size)
228{
229 return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM, PAGE_KERNEL);
230}
Andrew Mortonf6138882006-02-28 16:59:18 -0800231EXPORT_SYMBOL(vmalloc);
232
233void *vmalloc_node(unsigned long size, int node)
234{
235 return vmalloc(size);
236}
237EXPORT_SYMBOL(vmalloc_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238
239/*
240 * vmalloc_32 - allocate virtually continguos memory (32bit addressable)
241 *
242 * @size: allocation size
243 *
244 * Allocate enough 32bit PA addressable pages to cover @size from the
245 * page level allocator and map them into continguos kernel virtual space.
246 */
247void *vmalloc_32(unsigned long size)
248{
249 return __vmalloc(size, GFP_KERNEL, PAGE_KERNEL);
250}
251
252void *vmap(struct page **pages, unsigned int count, unsigned long flags, pgprot_t prot)
253{
254 BUG();
255 return NULL;
256}
257
258void vunmap(void *addr)
259{
260 BUG();
261}
262
263/*
264 * sys_brk() for the most part doesn't need the global kernel
265 * lock, except when an application is doing something nasty
266 * like trying to un-brk an area that has already been mapped
267 * to a regular file. in this case, the unmapping will need
268 * to invoke file system routines that need the global lock.
269 */
270asmlinkage unsigned long sys_brk(unsigned long brk)
271{
272 struct mm_struct *mm = current->mm;
273
274 if (brk < mm->start_brk || brk > mm->context.end_brk)
275 return mm->brk;
276
277 if (mm->brk == brk)
278 return mm->brk;
279
280 /*
281 * Always allow shrinking brk
282 */
283 if (brk <= mm->brk) {
284 mm->brk = brk;
285 return brk;
286 }
287
288 /*
289 * Ok, looks good - let it rip.
290 */
291 return mm->brk = brk;
292}
293
294#ifdef DEBUG
295static void show_process_blocks(void)
296{
297 struct vm_list_struct *vml;
298
299 printk("Process blocks %d:", current->pid);
300
301 for (vml = &current->mm->context.vmlist; vml; vml = vml->next) {
302 printk(" %p: %p", vml, vml->vma);
303 if (vml->vma)
304 printk(" (%d @%lx #%d)",
305 kobjsize((void *) vml->vma->vm_start),
306 vml->vma->vm_start,
307 atomic_read(&vml->vma->vm_usage));
308 printk(vml->next ? " ->" : ".\n");
309 }
310}
311#endif /* DEBUG */
312
David Howells30340972006-09-27 01:50:20 -0700313/*
314 * add a VMA into a process's mm_struct in the appropriate place in the list
315 * - should be called with mm->mmap_sem held writelocked
316 */
317static void add_vma_to_mm(struct mm_struct *mm, struct vm_list_struct *vml)
318{
319 struct vm_list_struct **ppv;
320
321 for (ppv = &current->mm->context.vmlist; *ppv; ppv = &(*ppv)->next)
322 if ((*ppv)->vma->vm_start > vml->vma->vm_start)
323 break;
324
325 vml->next = *ppv;
326 *ppv = vml;
327}
328
329/*
330 * look up the first VMA in which addr resides, NULL if none
331 * - should be called with mm->mmap_sem at least held readlocked
332 */
333struct vm_area_struct *find_vma(struct mm_struct *mm, unsigned long addr)
334{
335 struct vm_list_struct *loop, *vml;
336
337 /* search the vm_start ordered list */
338 vml = NULL;
339 for (loop = mm->context.vmlist; loop; loop = loop->next) {
340 if (loop->vma->vm_start > addr)
341 break;
342 vml = loop;
343 }
344
345 if (vml && vml->vma->vm_end > addr)
346 return vml->vma;
347
348 return NULL;
349}
350EXPORT_SYMBOL(find_vma);
351
352/*
David Howells930e6522006-09-27 01:50:22 -0700353 * find a VMA
354 * - we don't extend stack VMAs under NOMMU conditions
355 */
356struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr)
357{
358 return find_vma(mm, addr);
359}
360
361/*
David Howells6fa5f802006-09-27 01:50:21 -0700362 * look up the first VMA exactly that exactly matches addr
363 * - should be called with mm->mmap_sem at least held readlocked
364 */
365static inline struct vm_area_struct *find_vma_exact(struct mm_struct *mm,
366 unsigned long addr)
367{
368 struct vm_list_struct *vml;
369
370 /* search the vm_start ordered list */
371 for (vml = mm->context.vmlist; vml; vml = vml->next) {
372 if (vml->vma->vm_start == addr)
373 return vml->vma;
374 if (vml->vma->vm_start > addr)
375 break;
376 }
377
378 return NULL;
379}
380
381/*
David Howells30340972006-09-27 01:50:20 -0700382 * find a VMA in the global tree
383 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384static inline struct vm_area_struct *find_nommu_vma(unsigned long start)
385{
386 struct vm_area_struct *vma;
387 struct rb_node *n = nommu_vma_tree.rb_node;
388
389 while (n) {
390 vma = rb_entry(n, struct vm_area_struct, vm_rb);
391
392 if (start < vma->vm_start)
393 n = n->rb_left;
394 else if (start > vma->vm_start)
395 n = n->rb_right;
396 else
397 return vma;
398 }
399
400 return NULL;
401}
402
David Howells30340972006-09-27 01:50:20 -0700403/*
404 * add a VMA in the global tree
405 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406static void add_nommu_vma(struct vm_area_struct *vma)
407{
408 struct vm_area_struct *pvma;
409 struct address_space *mapping;
410 struct rb_node **p = &nommu_vma_tree.rb_node;
411 struct rb_node *parent = NULL;
412
413 /* add the VMA to the mapping */
414 if (vma->vm_file) {
415 mapping = vma->vm_file->f_mapping;
416
417 flush_dcache_mmap_lock(mapping);
418 vma_prio_tree_insert(vma, &mapping->i_mmap);
419 flush_dcache_mmap_unlock(mapping);
420 }
421
422 /* add the VMA to the master list */
423 while (*p) {
424 parent = *p;
425 pvma = rb_entry(parent, struct vm_area_struct, vm_rb);
426
427 if (vma->vm_start < pvma->vm_start) {
428 p = &(*p)->rb_left;
429 }
430 else if (vma->vm_start > pvma->vm_start) {
431 p = &(*p)->rb_right;
432 }
433 else {
434 /* mappings are at the same address - this can only
435 * happen for shared-mem chardevs and shared file
436 * mappings backed by ramfs/tmpfs */
437 BUG_ON(!(pvma->vm_flags & VM_SHARED));
438
439 if (vma < pvma)
440 p = &(*p)->rb_left;
441 else if (vma > pvma)
442 p = &(*p)->rb_right;
443 else
444 BUG();
445 }
446 }
447
448 rb_link_node(&vma->vm_rb, parent, p);
449 rb_insert_color(&vma->vm_rb, &nommu_vma_tree);
450}
451
David Howells30340972006-09-27 01:50:20 -0700452/*
453 * delete a VMA from the global list
454 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455static void delete_nommu_vma(struct vm_area_struct *vma)
456{
457 struct address_space *mapping;
458
459 /* remove the VMA from the mapping */
460 if (vma->vm_file) {
461 mapping = vma->vm_file->f_mapping;
462
463 flush_dcache_mmap_lock(mapping);
464 vma_prio_tree_remove(vma, &mapping->i_mmap);
465 flush_dcache_mmap_unlock(mapping);
466 }
467
468 /* remove from the master list */
469 rb_erase(&vma->vm_rb, &nommu_vma_tree);
470}
471
472/*
473 * determine whether a mapping should be permitted and, if so, what sort of
474 * mapping we're capable of supporting
475 */
476static int validate_mmap_request(struct file *file,
477 unsigned long addr,
478 unsigned long len,
479 unsigned long prot,
480 unsigned long flags,
481 unsigned long pgoff,
482 unsigned long *_capabilities)
483{
484 unsigned long capabilities;
485 unsigned long reqprot = prot;
486 int ret;
487
488 /* do the simple checks first */
489 if (flags & MAP_FIXED || addr) {
490 printk(KERN_DEBUG
491 "%d: Can't do fixed-address/overlay mmap of RAM\n",
492 current->pid);
493 return -EINVAL;
494 }
495
496 if ((flags & MAP_TYPE) != MAP_PRIVATE &&
497 (flags & MAP_TYPE) != MAP_SHARED)
498 return -EINVAL;
499
Mike Frysingerf81cff02006-12-06 12:02:59 +1000500 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 return -EINVAL;
502
Mike Frysingerf81cff02006-12-06 12:02:59 +1000503 /* Careful about overflows.. */
504 len = PAGE_ALIGN(len);
505 if (!len || len > TASK_SIZE)
506 return -ENOMEM;
507
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 /* offset overflow? */
509 if ((pgoff + (len >> PAGE_SHIFT)) < pgoff)
Mike Frysingerf81cff02006-12-06 12:02:59 +1000510 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511
512 if (file) {
513 /* validate file mapping requests */
514 struct address_space *mapping;
515
516 /* files must support mmap */
517 if (!file->f_op || !file->f_op->mmap)
518 return -ENODEV;
519
520 /* work out if what we've got could possibly be shared
521 * - we support chardevs that provide their own "memory"
522 * - we support files/blockdevs that are memory backed
523 */
524 mapping = file->f_mapping;
525 if (!mapping)
Josef Sipeke9536ae2006-12-08 02:37:21 -0800526 mapping = file->f_path.dentry->d_inode->i_mapping;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527
528 capabilities = 0;
529 if (mapping && mapping->backing_dev_info)
530 capabilities = mapping->backing_dev_info->capabilities;
531
532 if (!capabilities) {
533 /* no explicit capabilities set, so assume some
534 * defaults */
Josef Sipeke9536ae2006-12-08 02:37:21 -0800535 switch (file->f_path.dentry->d_inode->i_mode & S_IFMT) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 case S_IFREG:
537 case S_IFBLK:
538 capabilities = BDI_CAP_MAP_COPY;
539 break;
540
541 case S_IFCHR:
542 capabilities =
543 BDI_CAP_MAP_DIRECT |
544 BDI_CAP_READ_MAP |
545 BDI_CAP_WRITE_MAP;
546 break;
547
548 default:
549 return -EINVAL;
550 }
551 }
552
553 /* eliminate any capabilities that we can't support on this
554 * device */
555 if (!file->f_op->get_unmapped_area)
556 capabilities &= ~BDI_CAP_MAP_DIRECT;
557 if (!file->f_op->read)
558 capabilities &= ~BDI_CAP_MAP_COPY;
559
560 if (flags & MAP_SHARED) {
561 /* do checks for writing, appending and locking */
562 if ((prot & PROT_WRITE) &&
563 !(file->f_mode & FMODE_WRITE))
564 return -EACCES;
565
Josef Sipeke9536ae2006-12-08 02:37:21 -0800566 if (IS_APPEND(file->f_path.dentry->d_inode) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 (file->f_mode & FMODE_WRITE))
568 return -EACCES;
569
Josef Sipeke9536ae2006-12-08 02:37:21 -0800570 if (locks_verify_locked(file->f_path.dentry->d_inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 return -EAGAIN;
572
573 if (!(capabilities & BDI_CAP_MAP_DIRECT))
574 return -ENODEV;
575
576 if (((prot & PROT_READ) && !(capabilities & BDI_CAP_READ_MAP)) ||
577 ((prot & PROT_WRITE) && !(capabilities & BDI_CAP_WRITE_MAP)) ||
578 ((prot & PROT_EXEC) && !(capabilities & BDI_CAP_EXEC_MAP))
579 ) {
580 printk("MAP_SHARED not completely supported on !MMU\n");
581 return -EINVAL;
582 }
583
584 /* we mustn't privatise shared mappings */
585 capabilities &= ~BDI_CAP_MAP_COPY;
586 }
587 else {
588 /* we're going to read the file into private memory we
589 * allocate */
590 if (!(capabilities & BDI_CAP_MAP_COPY))
591 return -ENODEV;
592
593 /* we don't permit a private writable mapping to be
594 * shared with the backing device */
595 if (prot & PROT_WRITE)
596 capabilities &= ~BDI_CAP_MAP_DIRECT;
597 }
598
599 /* handle executable mappings and implied executable
600 * mappings */
Josef Sipeke9536ae2006-12-08 02:37:21 -0800601 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 if (prot & PROT_EXEC)
603 return -EPERM;
604 }
605 else if ((prot & PROT_READ) && !(prot & PROT_EXEC)) {
606 /* handle implication of PROT_EXEC by PROT_READ */
607 if (current->personality & READ_IMPLIES_EXEC) {
608 if (capabilities & BDI_CAP_EXEC_MAP)
609 prot |= PROT_EXEC;
610 }
611 }
612 else if ((prot & PROT_READ) &&
613 (prot & PROT_EXEC) &&
614 !(capabilities & BDI_CAP_EXEC_MAP)
615 ) {
616 /* backing file is not executable, try to copy */
617 capabilities &= ~BDI_CAP_MAP_DIRECT;
618 }
619 }
620 else {
621 /* anonymous mappings are always memory backed and can be
622 * privately mapped
623 */
624 capabilities = BDI_CAP_MAP_COPY;
625
626 /* handle PROT_EXEC implication by PROT_READ */
627 if ((prot & PROT_READ) &&
628 (current->personality & READ_IMPLIES_EXEC))
629 prot |= PROT_EXEC;
630 }
631
632 /* allow the security API to have its say */
633 ret = security_file_mmap(file, reqprot, prot, flags);
634 if (ret < 0)
635 return ret;
636
637 /* looks okay */
638 *_capabilities = capabilities;
639 return 0;
640}
641
642/*
643 * we've determined that we can make the mapping, now translate what we
644 * now know into VMA flags
645 */
646static unsigned long determine_vm_flags(struct file *file,
647 unsigned long prot,
648 unsigned long flags,
649 unsigned long capabilities)
650{
651 unsigned long vm_flags;
652
653 vm_flags = calc_vm_prot_bits(prot) | calc_vm_flag_bits(flags);
654 vm_flags |= VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC;
655 /* vm_flags |= mm->def_flags; */
656
657 if (!(capabilities & BDI_CAP_MAP_DIRECT)) {
658 /* attempt to share read-only copies of mapped file chunks */
659 if (file && !(prot & PROT_WRITE))
660 vm_flags |= VM_MAYSHARE;
661 }
662 else {
663 /* overlay a shareable mapping on the backing device or inode
664 * if possible - used for chardevs, ramfs/tmpfs/shmfs and
665 * romfs/cramfs */
666 if (flags & MAP_SHARED)
667 vm_flags |= VM_MAYSHARE | VM_SHARED;
668 else if ((((vm_flags & capabilities) ^ vm_flags) & BDI_CAP_VMFLAGS) == 0)
669 vm_flags |= VM_MAYSHARE;
670 }
671
672 /* refuse to let anyone share private mappings with this process if
673 * it's being traced - otherwise breakpoints set in it may interfere
674 * with another untraced process
675 */
676 if ((flags & MAP_PRIVATE) && (current->ptrace & PT_PTRACED))
677 vm_flags &= ~VM_MAYSHARE;
678
679 return vm_flags;
680}
681
682/*
683 * set up a shared mapping on a file
684 */
685static int do_mmap_shared_file(struct vm_area_struct *vma, unsigned long len)
686{
687 int ret;
688
689 ret = vma->vm_file->f_op->mmap(vma->vm_file, vma);
690 if (ret != -ENOSYS)
691 return ret;
692
693 /* getting an ENOSYS error indicates that direct mmap isn't
694 * possible (as opposed to tried but failed) so we'll fall
695 * through to making a private copy of the data and mapping
696 * that if we can */
697 return -ENODEV;
698}
699
700/*
701 * set up a private mapping or an anonymous shared mapping
702 */
703static int do_mmap_private(struct vm_area_struct *vma, unsigned long len)
704{
705 void *base;
706 int ret;
707
708 /* invoke the file's mapping function so that it can keep track of
709 * shared mappings on devices or memory
710 * - VM_MAYSHARE will be set if it may attempt to share
711 */
712 if (vma->vm_file) {
713 ret = vma->vm_file->f_op->mmap(vma->vm_file, vma);
714 if (ret != -ENOSYS) {
715 /* shouldn't return success if we're not sharing */
716 BUG_ON(ret == 0 && !(vma->vm_flags & VM_MAYSHARE));
717 return ret; /* success or a real error */
718 }
719
720 /* getting an ENOSYS error indicates that direct mmap isn't
721 * possible (as opposed to tried but failed) so we'll try to
722 * make a private copy of the data and map that instead */
723 }
724
725 /* allocate some memory to hold the mapping
726 * - note that this may not return a page-aligned address if the object
727 * we're allocating is smaller than a page
728 */
Nick Piggin84097512006-03-22 00:08:34 -0800729 base = kmalloc(len, GFP_KERNEL|__GFP_COMP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 if (!base)
731 goto enomem;
732
733 vma->vm_start = (unsigned long) base;
734 vma->vm_end = vma->vm_start + len;
735 vma->vm_flags |= VM_MAPPED_COPY;
736
737#ifdef WARN_ON_SLACK
738 if (len + WARN_ON_SLACK <= kobjsize(result))
739 printk("Allocation of %lu bytes from process %d has %lu bytes of slack\n",
740 len, current->pid, kobjsize(result) - len);
741#endif
742
743 if (vma->vm_file) {
744 /* read the contents of a file into the copy */
745 mm_segment_t old_fs;
746 loff_t fpos;
747
748 fpos = vma->vm_pgoff;
749 fpos <<= PAGE_SHIFT;
750
751 old_fs = get_fs();
752 set_fs(KERNEL_DS);
753 ret = vma->vm_file->f_op->read(vma->vm_file, base, len, &fpos);
754 set_fs(old_fs);
755
756 if (ret < 0)
757 goto error_free;
758
759 /* clear the last little bit */
760 if (ret < len)
761 memset(base + ret, 0, len - ret);
762
763 } else {
764 /* if it's an anonymous mapping, then just clear it */
765 memset(base, 0, len);
766 }
767
768 return 0;
769
770error_free:
771 kfree(base);
772 vma->vm_start = 0;
773 return ret;
774
775enomem:
776 printk("Allocation of length %lu from process %d failed\n",
777 len, current->pid);
778 show_free_areas();
779 return -ENOMEM;
780}
781
782/*
783 * handle mapping creation for uClinux
784 */
785unsigned long do_mmap_pgoff(struct file *file,
786 unsigned long addr,
787 unsigned long len,
788 unsigned long prot,
789 unsigned long flags,
790 unsigned long pgoff)
791{
792 struct vm_list_struct *vml = NULL;
793 struct vm_area_struct *vma = NULL;
794 struct rb_node *rb;
795 unsigned long capabilities, vm_flags;
796 void *result;
797 int ret;
798
799 /* decide whether we should attempt the mapping, and if so what sort of
800 * mapping */
801 ret = validate_mmap_request(file, addr, len, prot, flags, pgoff,
802 &capabilities);
803 if (ret < 0)
804 return ret;
805
806 /* we've determined that we can make the mapping, now translate what we
807 * now know into VMA flags */
808 vm_flags = determine_vm_flags(file, prot, flags, capabilities);
809
810 /* we're going to need to record the mapping if it works */
Burman Yan4668edc2006-12-06 20:38:51 -0800811 vml = kzalloc(sizeof(struct vm_list_struct), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 if (!vml)
813 goto error_getting_vml;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814
815 down_write(&nommu_vma_sem);
816
817 /* if we want to share, we need to check for VMAs created by other
818 * mmap() calls that overlap with our proposed mapping
819 * - we can only share with an exact match on most regular files
820 * - shared mappings on character devices and memory backed files are
821 * permitted to overlap inexactly as far as we are concerned for in
822 * these cases, sharing is handled in the driver or filesystem rather
823 * than here
824 */
825 if (vm_flags & VM_MAYSHARE) {
826 unsigned long pglen = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
827 unsigned long vmpglen;
828
David Howells165b2392007-03-22 00:11:24 -0800829 /* suppress VMA sharing for shared regions */
830 if (vm_flags & VM_SHARED &&
831 capabilities & BDI_CAP_MAP_DIRECT)
832 goto dont_share_VMAs;
833
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834 for (rb = rb_first(&nommu_vma_tree); rb; rb = rb_next(rb)) {
835 vma = rb_entry(rb, struct vm_area_struct, vm_rb);
836
837 if (!(vma->vm_flags & VM_MAYSHARE))
838 continue;
839
840 /* search for overlapping mappings on the same file */
Josef Sipeke9536ae2006-12-08 02:37:21 -0800841 if (vma->vm_file->f_path.dentry->d_inode != file->f_path.dentry->d_inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 continue;
843
844 if (vma->vm_pgoff >= pgoff + pglen)
845 continue;
846
847 vmpglen = vma->vm_end - vma->vm_start + PAGE_SIZE - 1;
848 vmpglen >>= PAGE_SHIFT;
849 if (pgoff >= vma->vm_pgoff + vmpglen)
850 continue;
851
852 /* handle inexactly overlapping matches between mappings */
853 if (vma->vm_pgoff != pgoff || vmpglen != pglen) {
854 if (!(capabilities & BDI_CAP_MAP_DIRECT))
855 goto sharing_violation;
856 continue;
857 }
858
859 /* we've found a VMA we can share */
860 atomic_inc(&vma->vm_usage);
861
862 vml->vma = vma;
863 result = (void *) vma->vm_start;
864 goto shared;
865 }
866
David Howells165b2392007-03-22 00:11:24 -0800867 dont_share_VMAs:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 vma = NULL;
869
870 /* obtain the address at which to make a shared mapping
871 * - this is the hook for quasi-memory character devices to
872 * tell us the location of a shared mapping
873 */
874 if (file && file->f_op->get_unmapped_area) {
875 addr = file->f_op->get_unmapped_area(file, addr, len,
876 pgoff, flags);
877 if (IS_ERR((void *) addr)) {
878 ret = addr;
879 if (ret != (unsigned long) -ENOSYS)
880 goto error;
881
882 /* the driver refused to tell us where to site
883 * the mapping so we'll have to attempt to copy
884 * it */
885 ret = (unsigned long) -ENODEV;
886 if (!(capabilities & BDI_CAP_MAP_COPY))
887 goto error;
888
889 capabilities &= ~BDI_CAP_MAP_DIRECT;
890 }
891 }
892 }
893
894 /* we're going to need a VMA struct as well */
Burman Yan4668edc2006-12-06 20:38:51 -0800895 vma = kzalloc(sizeof(struct vm_area_struct), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 if (!vma)
897 goto error_getting_vma;
898
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 INIT_LIST_HEAD(&vma->anon_vma_node);
900 atomic_set(&vma->vm_usage, 1);
901 if (file)
902 get_file(file);
903 vma->vm_file = file;
904 vma->vm_flags = vm_flags;
905 vma->vm_start = addr;
906 vma->vm_end = addr + len;
907 vma->vm_pgoff = pgoff;
908
909 vml->vma = vma;
910
911 /* set up the mapping */
912 if (file && vma->vm_flags & VM_SHARED)
913 ret = do_mmap_shared_file(vma, len);
914 else
915 ret = do_mmap_private(vma, len);
916 if (ret < 0)
917 goto error;
918
919 /* okay... we have a mapping; now we have to register it */
920 result = (void *) vma->vm_start;
921
922 if (vma->vm_flags & VM_MAPPED_COPY) {
923 realalloc += kobjsize(result);
924 askedalloc += len;
925 }
926
927 realalloc += kobjsize(vma);
928 askedalloc += sizeof(*vma);
929
930 current->mm->total_vm += len >> PAGE_SHIFT;
931
932 add_nommu_vma(vma);
933
934 shared:
935 realalloc += kobjsize(vml);
936 askedalloc += sizeof(*vml);
937
David Howells30340972006-09-27 01:50:20 -0700938 add_vma_to_mm(current->mm, vml);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939
940 up_write(&nommu_vma_sem);
941
942 if (prot & PROT_EXEC)
943 flush_icache_range((unsigned long) result,
944 (unsigned long) result + len);
945
946#ifdef DEBUG
947 printk("do_mmap:\n");
948 show_process_blocks();
949#endif
950
951 return (unsigned long) result;
952
953 error:
954 up_write(&nommu_vma_sem);
955 kfree(vml);
956 if (vma) {
Gavin Lambert3fcd03e2006-09-30 23:27:01 -0700957 if (vma->vm_file)
958 fput(vma->vm_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 kfree(vma);
960 }
961 return ret;
962
963 sharing_violation:
964 up_write(&nommu_vma_sem);
965 printk("Attempt to share mismatched mappings\n");
966 kfree(vml);
967 return -EINVAL;
968
969 error_getting_vma:
970 up_write(&nommu_vma_sem);
971 kfree(vml);
Greg Ungerer66aa2b42005-09-12 11:18:10 +1000972 printk("Allocation of vma for %lu byte allocation from process %d failed\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 len, current->pid);
974 show_free_areas();
975 return -ENOMEM;
976
977 error_getting_vml:
978 printk("Allocation of vml for %lu byte allocation from process %d failed\n",
979 len, current->pid);
980 show_free_areas();
981 return -ENOMEM;
982}
983
984/*
985 * handle mapping disposal for uClinux
986 */
987static void put_vma(struct vm_area_struct *vma)
988{
989 if (vma) {
990 down_write(&nommu_vma_sem);
991
992 if (atomic_dec_and_test(&vma->vm_usage)) {
993 delete_nommu_vma(vma);
994
995 if (vma->vm_ops && vma->vm_ops->close)
996 vma->vm_ops->close(vma);
997
998 /* IO memory and memory shared directly out of the pagecache from
999 * ramfs/tmpfs mustn't be released here */
1000 if (vma->vm_flags & VM_MAPPED_COPY) {
1001 realalloc -= kobjsize((void *) vma->vm_start);
1002 askedalloc -= vma->vm_end - vma->vm_start;
1003 kfree((void *) vma->vm_start);
1004 }
1005
1006 realalloc -= kobjsize(vma);
1007 askedalloc -= sizeof(*vma);
1008
1009 if (vma->vm_file)
1010 fput(vma->vm_file);
1011 kfree(vma);
1012 }
1013
1014 up_write(&nommu_vma_sem);
1015 }
1016}
1017
David Howells30340972006-09-27 01:50:20 -07001018/*
1019 * release a mapping
1020 * - under NOMMU conditions the parameters must match exactly to the mapping to
1021 * be removed
1022 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023int do_munmap(struct mm_struct *mm, unsigned long addr, size_t len)
1024{
1025 struct vm_list_struct *vml, **parent;
1026 unsigned long end = addr + len;
1027
1028#ifdef DEBUG
1029 printk("do_munmap:\n");
1030#endif
1031
David Howells30340972006-09-27 01:50:20 -07001032 for (parent = &mm->context.vmlist; *parent; parent = &(*parent)->next) {
1033 if ((*parent)->vma->vm_start > addr)
1034 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 if ((*parent)->vma->vm_start == addr &&
Greg Ungerer66aa2b42005-09-12 11:18:10 +10001036 ((len == 0) || ((*parent)->vma->vm_end == end)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 goto found;
David Howells30340972006-09-27 01:50:20 -07001038 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039
1040 printk("munmap of non-mmaped memory by process %d (%s): %p\n",
1041 current->pid, current->comm, (void *) addr);
1042 return -EINVAL;
1043
1044 found:
1045 vml = *parent;
1046
1047 put_vma(vml->vma);
1048
1049 *parent = vml->next;
1050 realalloc -= kobjsize(vml);
1051 askedalloc -= sizeof(*vml);
1052 kfree(vml);
Hugh Dickins365e9c872005-10-29 18:16:18 -07001053
1054 update_hiwater_vm(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 mm->total_vm -= len >> PAGE_SHIFT;
1056
1057#ifdef DEBUG
1058 show_process_blocks();
1059#endif
1060
1061 return 0;
1062}
1063
David Howells30340972006-09-27 01:50:20 -07001064asmlinkage long sys_munmap(unsigned long addr, size_t len)
1065{
1066 int ret;
1067 struct mm_struct *mm = current->mm;
1068
1069 down_write(&mm->mmap_sem);
1070 ret = do_munmap(mm, addr, len);
1071 up_write(&mm->mmap_sem);
1072 return ret;
1073}
1074
1075/*
1076 * Release all mappings
1077 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078void exit_mmap(struct mm_struct * mm)
1079{
1080 struct vm_list_struct *tmp;
1081
1082 if (mm) {
1083#ifdef DEBUG
1084 printk("Exit_mmap:\n");
1085#endif
1086
1087 mm->total_vm = 0;
1088
1089 while ((tmp = mm->context.vmlist)) {
1090 mm->context.vmlist = tmp->next;
1091 put_vma(tmp->vma);
1092
1093 realalloc -= kobjsize(tmp);
1094 askedalloc -= sizeof(*tmp);
1095 kfree(tmp);
1096 }
1097
1098#ifdef DEBUG
1099 show_process_blocks();
1100#endif
1101 }
1102}
1103
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104unsigned long do_brk(unsigned long addr, unsigned long len)
1105{
1106 return -ENOMEM;
1107}
1108
1109/*
David Howells6fa5f802006-09-27 01:50:21 -07001110 * expand (or shrink) an existing mapping, potentially moving it at the same
1111 * time (controlled by the MREMAP_MAYMOVE flag and available VM space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 *
David Howells6fa5f802006-09-27 01:50:21 -07001113 * under NOMMU conditions, we only permit changing a mapping's size, and only
1114 * as long as it stays within the hole allocated by the kmalloc() call in
1115 * do_mmap_pgoff() and the block is not shareable
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 *
David Howells6fa5f802006-09-27 01:50:21 -07001117 * MREMAP_FIXED is not supported under NOMMU conditions
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 */
1119unsigned long do_mremap(unsigned long addr,
1120 unsigned long old_len, unsigned long new_len,
1121 unsigned long flags, unsigned long new_addr)
1122{
David Howells6fa5f802006-09-27 01:50:21 -07001123 struct vm_area_struct *vma;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124
1125 /* insanity checks first */
1126 if (new_len == 0)
1127 return (unsigned long) -EINVAL;
1128
1129 if (flags & MREMAP_FIXED && new_addr != addr)
1130 return (unsigned long) -EINVAL;
1131
David Howells6fa5f802006-09-27 01:50:21 -07001132 vma = find_vma_exact(current->mm, addr);
1133 if (!vma)
1134 return (unsigned long) -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135
David Howells6fa5f802006-09-27 01:50:21 -07001136 if (vma->vm_end != vma->vm_start + old_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 return (unsigned long) -EFAULT;
1138
David Howells6fa5f802006-09-27 01:50:21 -07001139 if (vma->vm_flags & VM_MAYSHARE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 return (unsigned long) -EPERM;
1141
1142 if (new_len > kobjsize((void *) addr))
1143 return (unsigned long) -ENOMEM;
1144
1145 /* all checks complete - do it */
David Howells6fa5f802006-09-27 01:50:21 -07001146 vma->vm_end = vma->vm_start + new_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147
1148 askedalloc -= old_len;
1149 askedalloc += new_len;
1150
David Howells6fa5f802006-09-27 01:50:21 -07001151 return vma->vm_start;
1152}
1153
1154asmlinkage unsigned long sys_mremap(unsigned long addr,
1155 unsigned long old_len, unsigned long new_len,
1156 unsigned long flags, unsigned long new_addr)
1157{
1158 unsigned long ret;
1159
1160 down_write(&current->mm->mmap_sem);
1161 ret = do_mremap(addr, old_len, new_len, flags, new_addr);
1162 up_write(&current->mm->mmap_sem);
1163 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164}
1165
Linus Torvalds6aab3412005-11-28 14:34:23 -08001166struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001167 unsigned int foll_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168{
1169 return NULL;
1170}
1171
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172int remap_pfn_range(struct vm_area_struct *vma, unsigned long from,
1173 unsigned long to, unsigned long size, pgprot_t prot)
1174{
Greg Ungerer66aa2b42005-09-12 11:18:10 +10001175 vma->vm_start = vma->vm_pgoff << PAGE_SHIFT;
1176 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177}
Luke Yang22c4af42006-07-14 00:24:09 -07001178EXPORT_SYMBOL(remap_pfn_range);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179
1180void swap_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
1181{
1182}
1183
1184unsigned long arch_get_unmapped_area(struct file *file, unsigned long addr,
1185 unsigned long len, unsigned long pgoff, unsigned long flags)
1186{
1187 return -ENOMEM;
1188}
1189
Wolfgang Wander1363c3c2005-06-21 17:14:49 -07001190void arch_unmap_area(struct mm_struct *mm, unsigned long addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191{
1192}
1193
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194void unmap_mapping_range(struct address_space *mapping,
1195 loff_t const holebegin, loff_t const holelen,
1196 int even_cows)
1197{
1198}
Luke Yang22c4af42006-07-14 00:24:09 -07001199EXPORT_SYMBOL(unmap_mapping_range);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200
1201/*
David Howellsd56e03c2007-03-22 00:11:23 -08001202 * ask for an unmapped area at which to create a mapping on a file
1203 */
1204unsigned long get_unmapped_area(struct file *file, unsigned long addr,
1205 unsigned long len, unsigned long pgoff,
1206 unsigned long flags)
1207{
1208 unsigned long (*get_area)(struct file *, unsigned long, unsigned long,
1209 unsigned long, unsigned long);
1210
1211 get_area = current->mm->get_unmapped_area;
1212 if (file && file->f_op && file->f_op->get_unmapped_area)
1213 get_area = file->f_op->get_unmapped_area;
1214
1215 if (!get_area)
1216 return -ENOSYS;
1217
1218 return get_area(file, addr, len, pgoff, flags);
1219}
1220
1221EXPORT_SYMBOL(get_unmapped_area);
1222
1223/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 * Check that a process has enough memory to allocate a new virtual
1225 * mapping. 0 means there is enough memory for the allocation to
1226 * succeed and -ENOMEM implies there is not.
1227 *
1228 * We currently support three overcommit policies, which are set via the
1229 * vm.overcommit_memory sysctl. See Documentation/vm/overcommit-accounting
1230 *
1231 * Strict overcommit modes added 2002 Feb 26 by Alan Cox.
1232 * Additional code 2002 Jul 20 by Robert Love.
1233 *
1234 * cap_sys_admin is 1 if the process has admin privileges, 0 otherwise.
1235 *
1236 * Note this is a helper function intended to be used by LSMs which
1237 * wish to use this logic.
1238 */
1239int __vm_enough_memory(long pages, int cap_sys_admin)
1240{
1241 unsigned long free, allowed;
1242
1243 vm_acct_memory(pages);
1244
1245 /*
1246 * Sometimes we want to use more memory than we have
1247 */
1248 if (sysctl_overcommit_memory == OVERCOMMIT_ALWAYS)
1249 return 0;
1250
1251 if (sysctl_overcommit_memory == OVERCOMMIT_GUESS) {
1252 unsigned long n;
1253
Christoph Lameter347ce432006-06-30 01:55:35 -07001254 free = global_page_state(NR_FILE_PAGES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 free += nr_swap_pages;
1256
1257 /*
1258 * Any slabs which are created with the
1259 * SLAB_RECLAIM_ACCOUNT flag claim to have contents
1260 * which are reclaimable, under pressure. The dentry
1261 * cache and most inode caches should fall into this
1262 */
Christoph Lameter972d1a72006-09-25 23:31:51 -07001263 free += global_page_state(NR_SLAB_RECLAIMABLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264
1265 /*
1266 * Leave the last 3% for root
1267 */
1268 if (!cap_sys_admin)
1269 free -= free / 32;
1270
1271 if (free > pages)
1272 return 0;
1273
1274 /*
1275 * nr_free_pages() is very expensive on large systems,
1276 * only call if we're about to fail.
1277 */
1278 n = nr_free_pages();
Hideo AOKId5ddc792006-04-10 22:53:01 -07001279
1280 /*
1281 * Leave reserved pages. The pages are not for anonymous pages.
1282 */
1283 if (n <= totalreserve_pages)
1284 goto error;
1285 else
1286 n -= totalreserve_pages;
1287
1288 /*
1289 * Leave the last 3% for root
1290 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 if (!cap_sys_admin)
1292 n -= n / 32;
1293 free += n;
1294
1295 if (free > pages)
1296 return 0;
Hideo AOKId5ddc792006-04-10 22:53:01 -07001297
1298 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299 }
1300
1301 allowed = totalram_pages * sysctl_overcommit_ratio / 100;
1302 /*
1303 * Leave the last 3% for root
1304 */
1305 if (!cap_sys_admin)
1306 allowed -= allowed / 32;
1307 allowed += total_swap_pages;
1308
1309 /* Don't let a single process grow too big:
1310 leave 3% of the size of this process for other processes */
1311 allowed -= current->mm->total_vm / 32;
1312
Simon Derr2f60f8d2005-08-04 19:52:03 -07001313 /*
1314 * cast `allowed' as a signed long because vm_committed_space
1315 * sometimes has a negative value
1316 */
1317 if (atomic_read(&vm_committed_space) < (long)allowed)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 return 0;
Hideo AOKId5ddc792006-04-10 22:53:01 -07001319error:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 vm_unacct_memory(pages);
1321
1322 return -ENOMEM;
1323}
1324
1325int in_gate_area_no_task(unsigned long addr)
1326{
1327 return 0;
1328}
David Howellsb0e15192006-01-06 00:11:42 -08001329
1330struct page *filemap_nopage(struct vm_area_struct *area,
1331 unsigned long address, int *type)
1332{
1333 BUG();
1334 return NULL;
1335}
David Howells0ec76a12006-09-27 01:50:15 -07001336
1337/*
1338 * Access another process' address space.
1339 * - source/target buffer must be kernel space
1340 */
1341int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write)
1342{
David Howells0ec76a12006-09-27 01:50:15 -07001343 struct vm_area_struct *vma;
1344 struct mm_struct *mm;
1345
1346 if (addr + len < addr)
1347 return 0;
1348
1349 mm = get_task_mm(tsk);
1350 if (!mm)
1351 return 0;
1352
1353 down_read(&mm->mmap_sem);
1354
1355 /* the access must start within one of the target process's mappings */
David Howells0159b142006-09-27 01:50:16 -07001356 vma = find_vma(mm, addr);
1357 if (vma) {
David Howells0ec76a12006-09-27 01:50:15 -07001358 /* don't overrun this mapping */
1359 if (addr + len >= vma->vm_end)
1360 len = vma->vm_end - addr;
1361
1362 /* only read or write mappings where it is permitted */
David Howellsd00c7b92006-09-27 01:50:19 -07001363 if (write && vma->vm_flags & VM_MAYWRITE)
David Howells0ec76a12006-09-27 01:50:15 -07001364 len -= copy_to_user((void *) addr, buf, len);
David Howellsd00c7b92006-09-27 01:50:19 -07001365 else if (!write && vma->vm_flags & VM_MAYREAD)
David Howells0ec76a12006-09-27 01:50:15 -07001366 len -= copy_from_user(buf, (void *) addr, len);
1367 else
1368 len = 0;
1369 } else {
1370 len = 0;
1371 }
1372
1373 up_read(&mm->mmap_sem);
1374 mmput(mm);
1375 return len;
1376}