blob: 5edccd9c9218cda91a4298d587737e88db29f2c0 [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>
Paul Mundtf905bc42008-02-04 22:29:59 -080013 * Copyright (c) 2007 Paul Mundt <lethal@linux-sh.org>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014 */
15
David Howellsf2b85442007-10-29 13:15:39 +000016#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/mm.h>
18#include <linux/mman.h>
19#include <linux/swap.h>
20#include <linux/file.h>
21#include <linux/highmem.h>
22#include <linux/pagemap.h>
23#include <linux/slab.h>
24#include <linux/vmalloc.h>
Roland McGrathfa8e26c2008-07-25 19:45:50 -070025#include <linux/tracehook.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/blkdev.h>
27#include <linux/backing-dev.h>
28#include <linux/mount.h>
29#include <linux/personality.h>
30#include <linux/security.h>
31#include <linux/syscalls.h>
32
33#include <asm/uaccess.h>
34#include <asm/tlb.h>
35#include <asm/tlbflush.h>
36
37void *high_memory;
38struct page *mem_map;
39unsigned long max_mapnr;
40unsigned long num_physpages;
41unsigned long askedalloc, realalloc;
Alan Cox80119ef2008-05-23 13:04:31 -070042atomic_long_t vm_committed_space = ATOMIC_LONG_INIT(0);
Linus Torvalds1da177e2005-04-16 15:20:36 -070043int sysctl_overcommit_memory = OVERCOMMIT_GUESS; /* heuristic overcommit */
44int sysctl_overcommit_ratio = 50; /* default is 50% */
45int sysctl_max_map_count = DEFAULT_MAX_MAP_COUNT;
46int heap_stack_gap = 0;
47
48EXPORT_SYMBOL(mem_map);
Wu, Bryan6a04de62007-04-11 23:28:47 -070049EXPORT_SYMBOL(num_physpages);
Linus Torvalds1da177e2005-04-16 15:20:36 -070050
51/* list of shareable VMAs */
52struct rb_root nommu_vma_tree = RB_ROOT;
53DECLARE_RWSEM(nommu_vma_sem);
54
55struct vm_operations_struct generic_file_vm_ops = {
56};
57
58/*
59 * Handle all mappings that got truncated by a "truncate()"
60 * system call.
61 *
62 * NOTE! We have to be ready to update the memory sharing
63 * between the file and the memory map for a potential last
64 * incomplete page. Ugly, but necessary.
65 */
66int vmtruncate(struct inode *inode, loff_t offset)
67{
68 struct address_space *mapping = inode->i_mapping;
69 unsigned long limit;
70
71 if (inode->i_size < offset)
72 goto do_expand;
73 i_size_write(inode, offset);
74
75 truncate_inode_pages(mapping, offset);
76 goto out_truncate;
77
78do_expand:
79 limit = current->signal->rlim[RLIMIT_FSIZE].rlim_cur;
80 if (limit != RLIM_INFINITY && offset > limit)
81 goto out_sig;
82 if (offset > inode->i_sb->s_maxbytes)
83 goto out;
84 i_size_write(inode, offset);
85
86out_truncate:
87 if (inode->i_op && inode->i_op->truncate)
88 inode->i_op->truncate(inode);
89 return 0;
90out_sig:
91 send_sig(SIGXFSZ, current, 0);
92out:
93 return -EFBIG;
94}
95
96EXPORT_SYMBOL(vmtruncate);
97
98/*
99 * Return the total memory allocated for this pointer, not
100 * just what the caller asked for.
101 *
102 * Doesn't have to be accurate, i.e. may have races.
103 */
104unsigned int kobjsize(const void *objp)
105{
106 struct page *page;
107
Michael Hennerich4016a132008-04-28 02:13:38 -0700108 /*
109 * If the object we have should not have ksize performed on it,
110 * return size of 0
111 */
Paul Mundt5a1603b2008-06-12 16:29:55 +0900112 if (!objp || !virt_addr_valid(objp))
Paul Mundt6cfd53fc2008-06-05 22:46:08 -0700113 return 0;
114
115 page = virt_to_head_page(objp);
Paul Mundt6cfd53fc2008-06-05 22:46:08 -0700116
117 /*
118 * If the allocator sets PageSlab, we know the pointer came from
119 * kmalloc().
120 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121 if (PageSlab(page))
122 return ksize(objp);
123
Paul Mundt6cfd53fc2008-06-05 22:46:08 -0700124 /*
125 * The ksize() function is only guaranteed to work for pointers
Paul Mundt5a1603b2008-06-12 16:29:55 +0900126 * returned by kmalloc(). So handle arbitrary pointers here.
Paul Mundt6cfd53fc2008-06-05 22:46:08 -0700127 */
Paul Mundt5a1603b2008-06-12 16:29:55 +0900128 return PAGE_SIZE << compound_order(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129}
130
131/*
David Howells7b4d5b82006-09-27 01:50:18 -0700132 * get a list of pages in an address range belonging to the specified process
133 * and indicate the VMA that covers each page
134 * - this is potentially dodgy as we may end incrementing the page count of a
135 * slab page or a secondary page from a compound page
136 * - don't permit access to VMAs that don't support it, such as I/O mappings
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 */
138int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
139 unsigned long start, int len, int write, int force,
140 struct page **pages, struct vm_area_struct **vmas)
141{
Sonic Zhang910e46d2006-09-27 01:50:17 -0700142 struct vm_area_struct *vma;
David Howells7b4d5b82006-09-27 01:50:18 -0700143 unsigned long vm_flags;
144 int i;
145
146 /* calculate required read or write permissions.
147 * - if 'force' is set, we only require the "MAY" flags.
148 */
149 vm_flags = write ? (VM_WRITE | VM_MAYWRITE) : (VM_READ | VM_MAYREAD);
150 vm_flags &= force ? (VM_MAYREAD | VM_MAYWRITE) : (VM_READ | VM_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
152 for (i = 0; i < len; i++) {
Sonic Zhang910e46d2006-09-27 01:50:17 -0700153 vma = find_vma(mm, start);
David Howells7b4d5b82006-09-27 01:50:18 -0700154 if (!vma)
155 goto finish_or_fault;
156
157 /* protect what we can, including chardevs */
158 if (vma->vm_flags & (VM_IO | VM_PFNMAP) ||
159 !(vm_flags & vma->vm_flags))
160 goto finish_or_fault;
Sonic Zhang910e46d2006-09-27 01:50:17 -0700161
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 if (pages) {
163 pages[i] = virt_to_page(start);
164 if (pages[i])
165 page_cache_get(pages[i]);
166 }
167 if (vmas)
Sonic Zhang910e46d2006-09-27 01:50:17 -0700168 vmas[i] = vma;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 start += PAGE_SIZE;
170 }
David Howells7b4d5b82006-09-27 01:50:18 -0700171
172 return i;
173
174finish_or_fault:
175 return i ? : -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176}
Greg Ungerer66aa2b42005-09-12 11:18:10 +1000177EXPORT_SYMBOL(get_user_pages);
178
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179DEFINE_RWLOCK(vmlist_lock);
180struct vm_struct *vmlist;
181
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800182void vfree(const void *addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183{
184 kfree(addr);
185}
Paul Mundtb5073172007-07-21 04:37:25 -0700186EXPORT_SYMBOL(vfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187
Al Virodd0fc662005-10-07 07:46:04 +0100188void *__vmalloc(unsigned long size, gfp_t gfp_mask, pgprot_t prot)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189{
190 /*
Robert P. J. Day85186092007-10-19 23:11:38 +0200191 * You can't specify __GFP_HIGHMEM with kmalloc() since kmalloc()
192 * returns only a logical address.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 */
Nick Piggin84097512006-03-22 00:08:34 -0800194 return kmalloc(size, (gfp_mask | __GFP_COMP) & ~__GFP_HIGHMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195}
Paul Mundtb5073172007-07-21 04:37:25 -0700196EXPORT_SYMBOL(__vmalloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197
Paul Mundtf905bc42008-02-04 22:29:59 -0800198void *vmalloc_user(unsigned long size)
199{
200 void *ret;
201
202 ret = __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM | __GFP_ZERO,
203 PAGE_KERNEL);
204 if (ret) {
205 struct vm_area_struct *vma;
206
207 down_write(&current->mm->mmap_sem);
208 vma = find_vma(current->mm, (unsigned long)ret);
209 if (vma)
210 vma->vm_flags |= VM_USERMAP;
211 up_write(&current->mm->mmap_sem);
212 }
213
214 return ret;
215}
216EXPORT_SYMBOL(vmalloc_user);
217
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800218struct page *vmalloc_to_page(const void *addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219{
220 return virt_to_page(addr);
221}
Paul Mundtb5073172007-07-21 04:37:25 -0700222EXPORT_SYMBOL(vmalloc_to_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800224unsigned long vmalloc_to_pfn(const void *addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225{
226 return page_to_pfn(virt_to_page(addr));
227}
Paul Mundtb5073172007-07-21 04:37:25 -0700228EXPORT_SYMBOL(vmalloc_to_pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
230long vread(char *buf, char *addr, unsigned long count)
231{
232 memcpy(buf, addr, count);
233 return count;
234}
235
236long vwrite(char *buf, char *addr, unsigned long count)
237{
238 /* Don't allow overflow */
239 if ((unsigned long) addr + count < count)
240 count = -(unsigned long) addr;
241
242 memcpy(addr, buf, count);
243 return(count);
244}
245
246/*
247 * vmalloc - allocate virtually continguos memory
248 *
249 * @size: allocation size
250 *
251 * Allocate enough pages to cover @size from the page level
252 * allocator and map them into continguos kernel virtual space.
253 *
Michael Opdenackerc1c88972006-10-03 23:21:02 +0200254 * For tight control over page level allocator and protection flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 * use __vmalloc() instead.
256 */
257void *vmalloc(unsigned long size)
258{
259 return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM, PAGE_KERNEL);
260}
Andrew Mortonf6138882006-02-28 16:59:18 -0800261EXPORT_SYMBOL(vmalloc);
262
263void *vmalloc_node(unsigned long size, int node)
264{
265 return vmalloc(size);
266}
267EXPORT_SYMBOL(vmalloc_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
Paul Mundtb5073172007-07-21 04:37:25 -0700269/**
270 * vmalloc_32 - allocate virtually contiguous memory (32bit addressable)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 * @size: allocation size
272 *
273 * Allocate enough 32bit PA addressable pages to cover @size from the
274 * page level allocator and map them into continguos kernel virtual space.
275 */
276void *vmalloc_32(unsigned long size)
277{
278 return __vmalloc(size, GFP_KERNEL, PAGE_KERNEL);
279}
Paul Mundtb5073172007-07-21 04:37:25 -0700280EXPORT_SYMBOL(vmalloc_32);
281
282/**
283 * vmalloc_32_user - allocate zeroed virtually contiguous 32bit memory
284 * @size: allocation size
285 *
286 * The resulting memory area is 32bit addressable and zeroed so it can be
287 * mapped to userspace without leaking data.
Paul Mundtf905bc42008-02-04 22:29:59 -0800288 *
289 * VM_USERMAP is set on the corresponding VMA so that subsequent calls to
290 * remap_vmalloc_range() are permissible.
Paul Mundtb5073172007-07-21 04:37:25 -0700291 */
292void *vmalloc_32_user(unsigned long size)
293{
Paul Mundtf905bc42008-02-04 22:29:59 -0800294 /*
295 * We'll have to sort out the ZONE_DMA bits for 64-bit,
296 * but for now this can simply use vmalloc_user() directly.
297 */
298 return vmalloc_user(size);
Paul Mundtb5073172007-07-21 04:37:25 -0700299}
300EXPORT_SYMBOL(vmalloc_32_user);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301
302void *vmap(struct page **pages, unsigned int count, unsigned long flags, pgprot_t prot)
303{
304 BUG();
305 return NULL;
306}
Paul Mundtb5073172007-07-21 04:37:25 -0700307EXPORT_SYMBOL(vmap);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800309void vunmap(const void *addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310{
311 BUG();
312}
Paul Mundtb5073172007-07-21 04:37:25 -0700313EXPORT_SYMBOL(vunmap);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314
315/*
Christoph Hellwig1eeb66a2007-05-08 00:27:03 -0700316 * Implement a stub for vmalloc_sync_all() if the architecture chose not to
317 * have one.
318 */
319void __attribute__((weak)) vmalloc_sync_all(void)
320{
321}
322
Paul Mundtb5073172007-07-21 04:37:25 -0700323int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
324 struct page *page)
325{
326 return -EINVAL;
327}
328EXPORT_SYMBOL(vm_insert_page);
329
Christoph Hellwig1eeb66a2007-05-08 00:27:03 -0700330/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 * sys_brk() for the most part doesn't need the global kernel
332 * lock, except when an application is doing something nasty
333 * like trying to un-brk an area that has already been mapped
334 * to a regular file. in this case, the unmapping will need
335 * to invoke file system routines that need the global lock.
336 */
337asmlinkage unsigned long sys_brk(unsigned long brk)
338{
339 struct mm_struct *mm = current->mm;
340
341 if (brk < mm->start_brk || brk > mm->context.end_brk)
342 return mm->brk;
343
344 if (mm->brk == brk)
345 return mm->brk;
346
347 /*
348 * Always allow shrinking brk
349 */
350 if (brk <= mm->brk) {
351 mm->brk = brk;
352 return brk;
353 }
354
355 /*
356 * Ok, looks good - let it rip.
357 */
358 return mm->brk = brk;
359}
360
361#ifdef DEBUG
362static void show_process_blocks(void)
363{
364 struct vm_list_struct *vml;
365
366 printk("Process blocks %d:", current->pid);
367
368 for (vml = &current->mm->context.vmlist; vml; vml = vml->next) {
369 printk(" %p: %p", vml, vml->vma);
370 if (vml->vma)
371 printk(" (%d @%lx #%d)",
372 kobjsize((void *) vml->vma->vm_start),
373 vml->vma->vm_start,
374 atomic_read(&vml->vma->vm_usage));
375 printk(vml->next ? " ->" : ".\n");
376 }
377}
378#endif /* DEBUG */
379
David Howells30340972006-09-27 01:50:20 -0700380/*
381 * add a VMA into a process's mm_struct in the appropriate place in the list
382 * - should be called with mm->mmap_sem held writelocked
383 */
384static void add_vma_to_mm(struct mm_struct *mm, struct vm_list_struct *vml)
385{
386 struct vm_list_struct **ppv;
387
388 for (ppv = &current->mm->context.vmlist; *ppv; ppv = &(*ppv)->next)
389 if ((*ppv)->vma->vm_start > vml->vma->vm_start)
390 break;
391
392 vml->next = *ppv;
393 *ppv = vml;
394}
395
396/*
397 * look up the first VMA in which addr resides, NULL if none
398 * - should be called with mm->mmap_sem at least held readlocked
399 */
400struct vm_area_struct *find_vma(struct mm_struct *mm, unsigned long addr)
401{
402 struct vm_list_struct *loop, *vml;
403
404 /* search the vm_start ordered list */
405 vml = NULL;
406 for (loop = mm->context.vmlist; loop; loop = loop->next) {
407 if (loop->vma->vm_start > addr)
408 break;
409 vml = loop;
410 }
411
412 if (vml && vml->vma->vm_end > addr)
413 return vml->vma;
414
415 return NULL;
416}
417EXPORT_SYMBOL(find_vma);
418
419/*
David Howells930e6522006-09-27 01:50:22 -0700420 * find a VMA
421 * - we don't extend stack VMAs under NOMMU conditions
422 */
423struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr)
424{
425 return find_vma(mm, addr);
426}
427
Greg Ungerer57c8f632007-07-15 23:38:28 -0700428int expand_stack(struct vm_area_struct *vma, unsigned long address)
429{
430 return -ENOMEM;
431}
432
David Howells930e6522006-09-27 01:50:22 -0700433/*
David Howells6fa5f802006-09-27 01:50:21 -0700434 * look up the first VMA exactly that exactly matches addr
435 * - should be called with mm->mmap_sem at least held readlocked
436 */
437static inline struct vm_area_struct *find_vma_exact(struct mm_struct *mm,
438 unsigned long addr)
439{
440 struct vm_list_struct *vml;
441
442 /* search the vm_start ordered list */
443 for (vml = mm->context.vmlist; vml; vml = vml->next) {
444 if (vml->vma->vm_start == addr)
445 return vml->vma;
446 if (vml->vma->vm_start > addr)
447 break;
448 }
449
450 return NULL;
451}
452
453/*
David Howells30340972006-09-27 01:50:20 -0700454 * find a VMA in the global tree
455 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456static inline struct vm_area_struct *find_nommu_vma(unsigned long start)
457{
458 struct vm_area_struct *vma;
459 struct rb_node *n = nommu_vma_tree.rb_node;
460
461 while (n) {
462 vma = rb_entry(n, struct vm_area_struct, vm_rb);
463
464 if (start < vma->vm_start)
465 n = n->rb_left;
466 else if (start > vma->vm_start)
467 n = n->rb_right;
468 else
469 return vma;
470 }
471
472 return NULL;
473}
474
David Howells30340972006-09-27 01:50:20 -0700475/*
476 * add a VMA in the global tree
477 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478static void add_nommu_vma(struct vm_area_struct *vma)
479{
480 struct vm_area_struct *pvma;
481 struct address_space *mapping;
482 struct rb_node **p = &nommu_vma_tree.rb_node;
483 struct rb_node *parent = NULL;
484
485 /* add the VMA to the mapping */
486 if (vma->vm_file) {
487 mapping = vma->vm_file->f_mapping;
488
489 flush_dcache_mmap_lock(mapping);
490 vma_prio_tree_insert(vma, &mapping->i_mmap);
491 flush_dcache_mmap_unlock(mapping);
492 }
493
494 /* add the VMA to the master list */
495 while (*p) {
496 parent = *p;
497 pvma = rb_entry(parent, struct vm_area_struct, vm_rb);
498
499 if (vma->vm_start < pvma->vm_start) {
500 p = &(*p)->rb_left;
501 }
502 else if (vma->vm_start > pvma->vm_start) {
503 p = &(*p)->rb_right;
504 }
505 else {
506 /* mappings are at the same address - this can only
507 * happen for shared-mem chardevs and shared file
508 * mappings backed by ramfs/tmpfs */
509 BUG_ON(!(pvma->vm_flags & VM_SHARED));
510
511 if (vma < pvma)
512 p = &(*p)->rb_left;
513 else if (vma > pvma)
514 p = &(*p)->rb_right;
515 else
516 BUG();
517 }
518 }
519
520 rb_link_node(&vma->vm_rb, parent, p);
521 rb_insert_color(&vma->vm_rb, &nommu_vma_tree);
522}
523
David Howells30340972006-09-27 01:50:20 -0700524/*
525 * delete a VMA from the global list
526 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527static void delete_nommu_vma(struct vm_area_struct *vma)
528{
529 struct address_space *mapping;
530
531 /* remove the VMA from the mapping */
532 if (vma->vm_file) {
533 mapping = vma->vm_file->f_mapping;
534
535 flush_dcache_mmap_lock(mapping);
536 vma_prio_tree_remove(vma, &mapping->i_mmap);
537 flush_dcache_mmap_unlock(mapping);
538 }
539
540 /* remove from the master list */
541 rb_erase(&vma->vm_rb, &nommu_vma_tree);
542}
543
544/*
545 * determine whether a mapping should be permitted and, if so, what sort of
546 * mapping we're capable of supporting
547 */
548static int validate_mmap_request(struct file *file,
549 unsigned long addr,
550 unsigned long len,
551 unsigned long prot,
552 unsigned long flags,
553 unsigned long pgoff,
554 unsigned long *_capabilities)
555{
556 unsigned long capabilities;
557 unsigned long reqprot = prot;
558 int ret;
559
560 /* do the simple checks first */
561 if (flags & MAP_FIXED || addr) {
562 printk(KERN_DEBUG
563 "%d: Can't do fixed-address/overlay mmap of RAM\n",
564 current->pid);
565 return -EINVAL;
566 }
567
568 if ((flags & MAP_TYPE) != MAP_PRIVATE &&
569 (flags & MAP_TYPE) != MAP_SHARED)
570 return -EINVAL;
571
Mike Frysingerf81cff02006-12-06 12:02:59 +1000572 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 return -EINVAL;
574
Mike Frysingerf81cff02006-12-06 12:02:59 +1000575 /* Careful about overflows.. */
576 len = PAGE_ALIGN(len);
577 if (!len || len > TASK_SIZE)
578 return -ENOMEM;
579
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 /* offset overflow? */
581 if ((pgoff + (len >> PAGE_SHIFT)) < pgoff)
Mike Frysingerf81cff02006-12-06 12:02:59 +1000582 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583
584 if (file) {
585 /* validate file mapping requests */
586 struct address_space *mapping;
587
588 /* files must support mmap */
589 if (!file->f_op || !file->f_op->mmap)
590 return -ENODEV;
591
592 /* work out if what we've got could possibly be shared
593 * - we support chardevs that provide their own "memory"
594 * - we support files/blockdevs that are memory backed
595 */
596 mapping = file->f_mapping;
597 if (!mapping)
Josef Sipeke9536ae2006-12-08 02:37:21 -0800598 mapping = file->f_path.dentry->d_inode->i_mapping;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599
600 capabilities = 0;
601 if (mapping && mapping->backing_dev_info)
602 capabilities = mapping->backing_dev_info->capabilities;
603
604 if (!capabilities) {
605 /* no explicit capabilities set, so assume some
606 * defaults */
Josef Sipeke9536ae2006-12-08 02:37:21 -0800607 switch (file->f_path.dentry->d_inode->i_mode & S_IFMT) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 case S_IFREG:
609 case S_IFBLK:
610 capabilities = BDI_CAP_MAP_COPY;
611 break;
612
613 case S_IFCHR:
614 capabilities =
615 BDI_CAP_MAP_DIRECT |
616 BDI_CAP_READ_MAP |
617 BDI_CAP_WRITE_MAP;
618 break;
619
620 default:
621 return -EINVAL;
622 }
623 }
624
625 /* eliminate any capabilities that we can't support on this
626 * device */
627 if (!file->f_op->get_unmapped_area)
628 capabilities &= ~BDI_CAP_MAP_DIRECT;
629 if (!file->f_op->read)
630 capabilities &= ~BDI_CAP_MAP_COPY;
631
632 if (flags & MAP_SHARED) {
633 /* do checks for writing, appending and locking */
634 if ((prot & PROT_WRITE) &&
635 !(file->f_mode & FMODE_WRITE))
636 return -EACCES;
637
Josef Sipeke9536ae2006-12-08 02:37:21 -0800638 if (IS_APPEND(file->f_path.dentry->d_inode) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 (file->f_mode & FMODE_WRITE))
640 return -EACCES;
641
Josef Sipeke9536ae2006-12-08 02:37:21 -0800642 if (locks_verify_locked(file->f_path.dentry->d_inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 return -EAGAIN;
644
645 if (!(capabilities & BDI_CAP_MAP_DIRECT))
646 return -ENODEV;
647
648 if (((prot & PROT_READ) && !(capabilities & BDI_CAP_READ_MAP)) ||
649 ((prot & PROT_WRITE) && !(capabilities & BDI_CAP_WRITE_MAP)) ||
650 ((prot & PROT_EXEC) && !(capabilities & BDI_CAP_EXEC_MAP))
651 ) {
652 printk("MAP_SHARED not completely supported on !MMU\n");
653 return -EINVAL;
654 }
655
656 /* we mustn't privatise shared mappings */
657 capabilities &= ~BDI_CAP_MAP_COPY;
658 }
659 else {
660 /* we're going to read the file into private memory we
661 * allocate */
662 if (!(capabilities & BDI_CAP_MAP_COPY))
663 return -ENODEV;
664
665 /* we don't permit a private writable mapping to be
666 * shared with the backing device */
667 if (prot & PROT_WRITE)
668 capabilities &= ~BDI_CAP_MAP_DIRECT;
669 }
670
671 /* handle executable mappings and implied executable
672 * mappings */
Josef Sipeke9536ae2006-12-08 02:37:21 -0800673 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 if (prot & PROT_EXEC)
675 return -EPERM;
676 }
677 else if ((prot & PROT_READ) && !(prot & PROT_EXEC)) {
678 /* handle implication of PROT_EXEC by PROT_READ */
679 if (current->personality & READ_IMPLIES_EXEC) {
680 if (capabilities & BDI_CAP_EXEC_MAP)
681 prot |= PROT_EXEC;
682 }
683 }
684 else if ((prot & PROT_READ) &&
685 (prot & PROT_EXEC) &&
686 !(capabilities & BDI_CAP_EXEC_MAP)
687 ) {
688 /* backing file is not executable, try to copy */
689 capabilities &= ~BDI_CAP_MAP_DIRECT;
690 }
691 }
692 else {
693 /* anonymous mappings are always memory backed and can be
694 * privately mapped
695 */
696 capabilities = BDI_CAP_MAP_COPY;
697
698 /* handle PROT_EXEC implication by PROT_READ */
699 if ((prot & PROT_READ) &&
700 (current->personality & READ_IMPLIES_EXEC))
701 prot |= PROT_EXEC;
702 }
703
704 /* allow the security API to have its say */
Eric Parised032182007-06-28 15:55:21 -0400705 ret = security_file_mmap(file, reqprot, prot, flags, addr, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 if (ret < 0)
707 return ret;
708
709 /* looks okay */
710 *_capabilities = capabilities;
711 return 0;
712}
713
714/*
715 * we've determined that we can make the mapping, now translate what we
716 * now know into VMA flags
717 */
718static unsigned long determine_vm_flags(struct file *file,
719 unsigned long prot,
720 unsigned long flags,
721 unsigned long capabilities)
722{
723 unsigned long vm_flags;
724
725 vm_flags = calc_vm_prot_bits(prot) | calc_vm_flag_bits(flags);
726 vm_flags |= VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC;
727 /* vm_flags |= mm->def_flags; */
728
729 if (!(capabilities & BDI_CAP_MAP_DIRECT)) {
730 /* attempt to share read-only copies of mapped file chunks */
731 if (file && !(prot & PROT_WRITE))
732 vm_flags |= VM_MAYSHARE;
733 }
734 else {
735 /* overlay a shareable mapping on the backing device or inode
736 * if possible - used for chardevs, ramfs/tmpfs/shmfs and
737 * romfs/cramfs */
738 if (flags & MAP_SHARED)
739 vm_flags |= VM_MAYSHARE | VM_SHARED;
740 else if ((((vm_flags & capabilities) ^ vm_flags) & BDI_CAP_VMFLAGS) == 0)
741 vm_flags |= VM_MAYSHARE;
742 }
743
744 /* refuse to let anyone share private mappings with this process if
745 * it's being traced - otherwise breakpoints set in it may interfere
746 * with another untraced process
747 */
Roland McGrathfa8e26c2008-07-25 19:45:50 -0700748 if ((flags & MAP_PRIVATE) && tracehook_expect_breakpoints(current))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 vm_flags &= ~VM_MAYSHARE;
750
751 return vm_flags;
752}
753
754/*
755 * set up a shared mapping on a file
756 */
757static int do_mmap_shared_file(struct vm_area_struct *vma, unsigned long len)
758{
759 int ret;
760
761 ret = vma->vm_file->f_op->mmap(vma->vm_file, vma);
762 if (ret != -ENOSYS)
763 return ret;
764
765 /* getting an ENOSYS error indicates that direct mmap isn't
766 * possible (as opposed to tried but failed) so we'll fall
767 * through to making a private copy of the data and mapping
768 * that if we can */
769 return -ENODEV;
770}
771
772/*
773 * set up a private mapping or an anonymous shared mapping
774 */
775static int do_mmap_private(struct vm_area_struct *vma, unsigned long len)
776{
777 void *base;
778 int ret;
779
780 /* invoke the file's mapping function so that it can keep track of
781 * shared mappings on devices or memory
782 * - VM_MAYSHARE will be set if it may attempt to share
783 */
784 if (vma->vm_file) {
785 ret = vma->vm_file->f_op->mmap(vma->vm_file, vma);
786 if (ret != -ENOSYS) {
787 /* shouldn't return success if we're not sharing */
788 BUG_ON(ret == 0 && !(vma->vm_flags & VM_MAYSHARE));
789 return ret; /* success or a real error */
790 }
791
792 /* getting an ENOSYS error indicates that direct mmap isn't
793 * possible (as opposed to tried but failed) so we'll try to
794 * make a private copy of the data and map that instead */
795 }
796
797 /* allocate some memory to hold the mapping
798 * - note that this may not return a page-aligned address if the object
799 * we're allocating is smaller than a page
800 */
Nick Piggin84097512006-03-22 00:08:34 -0800801 base = kmalloc(len, GFP_KERNEL|__GFP_COMP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 if (!base)
803 goto enomem;
804
805 vma->vm_start = (unsigned long) base;
806 vma->vm_end = vma->vm_start + len;
807 vma->vm_flags |= VM_MAPPED_COPY;
808
809#ifdef WARN_ON_SLACK
810 if (len + WARN_ON_SLACK <= kobjsize(result))
811 printk("Allocation of %lu bytes from process %d has %lu bytes of slack\n",
812 len, current->pid, kobjsize(result) - len);
813#endif
814
815 if (vma->vm_file) {
816 /* read the contents of a file into the copy */
817 mm_segment_t old_fs;
818 loff_t fpos;
819
820 fpos = vma->vm_pgoff;
821 fpos <<= PAGE_SHIFT;
822
823 old_fs = get_fs();
824 set_fs(KERNEL_DS);
825 ret = vma->vm_file->f_op->read(vma->vm_file, base, len, &fpos);
826 set_fs(old_fs);
827
828 if (ret < 0)
829 goto error_free;
830
831 /* clear the last little bit */
832 if (ret < len)
833 memset(base + ret, 0, len - ret);
834
835 } else {
836 /* if it's an anonymous mapping, then just clear it */
837 memset(base, 0, len);
838 }
839
840 return 0;
841
842error_free:
843 kfree(base);
844 vma->vm_start = 0;
845 return ret;
846
847enomem:
848 printk("Allocation of length %lu from process %d failed\n",
849 len, current->pid);
850 show_free_areas();
851 return -ENOMEM;
852}
853
854/*
855 * handle mapping creation for uClinux
856 */
857unsigned long do_mmap_pgoff(struct file *file,
858 unsigned long addr,
859 unsigned long len,
860 unsigned long prot,
861 unsigned long flags,
862 unsigned long pgoff)
863{
864 struct vm_list_struct *vml = NULL;
865 struct vm_area_struct *vma = NULL;
866 struct rb_node *rb;
867 unsigned long capabilities, vm_flags;
868 void *result;
869 int ret;
870
Eric Paris7cd94142007-11-26 18:47:40 -0500871 if (!(flags & MAP_FIXED))
872 addr = round_hint_to_min(addr);
873
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 /* decide whether we should attempt the mapping, and if so what sort of
875 * mapping */
876 ret = validate_mmap_request(file, addr, len, prot, flags, pgoff,
877 &capabilities);
878 if (ret < 0)
879 return ret;
880
881 /* we've determined that we can make the mapping, now translate what we
882 * now know into VMA flags */
883 vm_flags = determine_vm_flags(file, prot, flags, capabilities);
884
885 /* we're going to need to record the mapping if it works */
Burman Yan4668edc2006-12-06 20:38:51 -0800886 vml = kzalloc(sizeof(struct vm_list_struct), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 if (!vml)
888 goto error_getting_vml;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889
890 down_write(&nommu_vma_sem);
891
892 /* if we want to share, we need to check for VMAs created by other
893 * mmap() calls that overlap with our proposed mapping
894 * - we can only share with an exact match on most regular files
895 * - shared mappings on character devices and memory backed files are
896 * permitted to overlap inexactly as far as we are concerned for in
897 * these cases, sharing is handled in the driver or filesystem rather
898 * than here
899 */
900 if (vm_flags & VM_MAYSHARE) {
901 unsigned long pglen = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
902 unsigned long vmpglen;
903
David Howells165b2392007-03-22 00:11:24 -0800904 /* suppress VMA sharing for shared regions */
905 if (vm_flags & VM_SHARED &&
906 capabilities & BDI_CAP_MAP_DIRECT)
907 goto dont_share_VMAs;
908
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 for (rb = rb_first(&nommu_vma_tree); rb; rb = rb_next(rb)) {
910 vma = rb_entry(rb, struct vm_area_struct, vm_rb);
911
912 if (!(vma->vm_flags & VM_MAYSHARE))
913 continue;
914
915 /* search for overlapping mappings on the same file */
Josef Sipeke9536ae2006-12-08 02:37:21 -0800916 if (vma->vm_file->f_path.dentry->d_inode != file->f_path.dentry->d_inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917 continue;
918
919 if (vma->vm_pgoff >= pgoff + pglen)
920 continue;
921
922 vmpglen = vma->vm_end - vma->vm_start + PAGE_SIZE - 1;
923 vmpglen >>= PAGE_SHIFT;
924 if (pgoff >= vma->vm_pgoff + vmpglen)
925 continue;
926
927 /* handle inexactly overlapping matches between mappings */
928 if (vma->vm_pgoff != pgoff || vmpglen != pglen) {
929 if (!(capabilities & BDI_CAP_MAP_DIRECT))
930 goto sharing_violation;
931 continue;
932 }
933
934 /* we've found a VMA we can share */
935 atomic_inc(&vma->vm_usage);
936
937 vml->vma = vma;
938 result = (void *) vma->vm_start;
939 goto shared;
940 }
941
David Howells165b2392007-03-22 00:11:24 -0800942 dont_share_VMAs:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 vma = NULL;
944
945 /* obtain the address at which to make a shared mapping
946 * - this is the hook for quasi-memory character devices to
947 * tell us the location of a shared mapping
948 */
949 if (file && file->f_op->get_unmapped_area) {
950 addr = file->f_op->get_unmapped_area(file, addr, len,
951 pgoff, flags);
952 if (IS_ERR((void *) addr)) {
953 ret = addr;
954 if (ret != (unsigned long) -ENOSYS)
955 goto error;
956
957 /* the driver refused to tell us where to site
958 * the mapping so we'll have to attempt to copy
959 * it */
960 ret = (unsigned long) -ENODEV;
961 if (!(capabilities & BDI_CAP_MAP_COPY))
962 goto error;
963
964 capabilities &= ~BDI_CAP_MAP_DIRECT;
965 }
966 }
967 }
968
969 /* we're going to need a VMA struct as well */
Burman Yan4668edc2006-12-06 20:38:51 -0800970 vma = kzalloc(sizeof(struct vm_area_struct), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 if (!vma)
972 goto error_getting_vma;
973
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 INIT_LIST_HEAD(&vma->anon_vma_node);
975 atomic_set(&vma->vm_usage, 1);
Matt Helsley925d1c42008-04-29 01:01:36 -0700976 if (file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 get_file(file);
Matt Helsley925d1c42008-04-29 01:01:36 -0700978 if (vm_flags & VM_EXECUTABLE) {
979 added_exe_file_vma(current->mm);
980 vma->vm_mm = current->mm;
981 }
982 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 vma->vm_file = file;
984 vma->vm_flags = vm_flags;
985 vma->vm_start = addr;
986 vma->vm_end = addr + len;
987 vma->vm_pgoff = pgoff;
988
989 vml->vma = vma;
990
991 /* set up the mapping */
992 if (file && vma->vm_flags & VM_SHARED)
993 ret = do_mmap_shared_file(vma, len);
994 else
995 ret = do_mmap_private(vma, len);
996 if (ret < 0)
997 goto error;
998
999 /* okay... we have a mapping; now we have to register it */
1000 result = (void *) vma->vm_start;
1001
1002 if (vma->vm_flags & VM_MAPPED_COPY) {
1003 realalloc += kobjsize(result);
1004 askedalloc += len;
1005 }
1006
1007 realalloc += kobjsize(vma);
1008 askedalloc += sizeof(*vma);
1009
1010 current->mm->total_vm += len >> PAGE_SHIFT;
1011
1012 add_nommu_vma(vma);
1013
1014 shared:
1015 realalloc += kobjsize(vml);
1016 askedalloc += sizeof(*vml);
1017
David Howells30340972006-09-27 01:50:20 -07001018 add_vma_to_mm(current->mm, vml);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
1020 up_write(&nommu_vma_sem);
1021
1022 if (prot & PROT_EXEC)
1023 flush_icache_range((unsigned long) result,
1024 (unsigned long) result + len);
1025
1026#ifdef DEBUG
1027 printk("do_mmap:\n");
1028 show_process_blocks();
1029#endif
1030
1031 return (unsigned long) result;
1032
1033 error:
1034 up_write(&nommu_vma_sem);
1035 kfree(vml);
1036 if (vma) {
Matt Helsley925d1c42008-04-29 01:01:36 -07001037 if (vma->vm_file) {
Gavin Lambert3fcd03e2006-09-30 23:27:01 -07001038 fput(vma->vm_file);
Matt Helsley925d1c42008-04-29 01:01:36 -07001039 if (vma->vm_flags & VM_EXECUTABLE)
1040 removed_exe_file_vma(vma->vm_mm);
1041 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 kfree(vma);
1043 }
1044 return ret;
1045
1046 sharing_violation:
1047 up_write(&nommu_vma_sem);
1048 printk("Attempt to share mismatched mappings\n");
1049 kfree(vml);
1050 return -EINVAL;
1051
1052 error_getting_vma:
1053 up_write(&nommu_vma_sem);
1054 kfree(vml);
Greg Ungerer66aa2b42005-09-12 11:18:10 +10001055 printk("Allocation of vma for %lu byte allocation from process %d failed\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 len, current->pid);
1057 show_free_areas();
1058 return -ENOMEM;
1059
1060 error_getting_vml:
1061 printk("Allocation of vml for %lu byte allocation from process %d failed\n",
1062 len, current->pid);
1063 show_free_areas();
1064 return -ENOMEM;
1065}
Paul Mundtb5073172007-07-21 04:37:25 -07001066EXPORT_SYMBOL(do_mmap_pgoff);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067
1068/*
1069 * handle mapping disposal for uClinux
1070 */
Matt Helsley925d1c42008-04-29 01:01:36 -07001071static void put_vma(struct mm_struct *mm, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072{
1073 if (vma) {
1074 down_write(&nommu_vma_sem);
1075
1076 if (atomic_dec_and_test(&vma->vm_usage)) {
1077 delete_nommu_vma(vma);
1078
1079 if (vma->vm_ops && vma->vm_ops->close)
1080 vma->vm_ops->close(vma);
1081
1082 /* IO memory and memory shared directly out of the pagecache from
1083 * ramfs/tmpfs mustn't be released here */
1084 if (vma->vm_flags & VM_MAPPED_COPY) {
1085 realalloc -= kobjsize((void *) vma->vm_start);
1086 askedalloc -= vma->vm_end - vma->vm_start;
1087 kfree((void *) vma->vm_start);
1088 }
1089
1090 realalloc -= kobjsize(vma);
1091 askedalloc -= sizeof(*vma);
1092
Matt Helsley925d1c42008-04-29 01:01:36 -07001093 if (vma->vm_file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 fput(vma->vm_file);
Matt Helsley925d1c42008-04-29 01:01:36 -07001095 if (vma->vm_flags & VM_EXECUTABLE)
1096 removed_exe_file_vma(mm);
1097 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 kfree(vma);
1099 }
1100
1101 up_write(&nommu_vma_sem);
1102 }
1103}
1104
David Howells30340972006-09-27 01:50:20 -07001105/*
1106 * release a mapping
1107 * - under NOMMU conditions the parameters must match exactly to the mapping to
1108 * be removed
1109 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110int do_munmap(struct mm_struct *mm, unsigned long addr, size_t len)
1111{
1112 struct vm_list_struct *vml, **parent;
1113 unsigned long end = addr + len;
1114
1115#ifdef DEBUG
1116 printk("do_munmap:\n");
1117#endif
1118
David Howells30340972006-09-27 01:50:20 -07001119 for (parent = &mm->context.vmlist; *parent; parent = &(*parent)->next) {
1120 if ((*parent)->vma->vm_start > addr)
1121 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 if ((*parent)->vma->vm_start == addr &&
Greg Ungerer66aa2b42005-09-12 11:18:10 +10001123 ((len == 0) || ((*parent)->vma->vm_end == end)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 goto found;
David Howells30340972006-09-27 01:50:20 -07001125 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126
1127 printk("munmap of non-mmaped memory by process %d (%s): %p\n",
1128 current->pid, current->comm, (void *) addr);
1129 return -EINVAL;
1130
1131 found:
1132 vml = *parent;
1133
Matt Helsley925d1c42008-04-29 01:01:36 -07001134 put_vma(mm, vml->vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135
1136 *parent = vml->next;
1137 realalloc -= kobjsize(vml);
1138 askedalloc -= sizeof(*vml);
1139 kfree(vml);
Hugh Dickins365e9c872005-10-29 18:16:18 -07001140
1141 update_hiwater_vm(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 mm->total_vm -= len >> PAGE_SHIFT;
1143
1144#ifdef DEBUG
1145 show_process_blocks();
1146#endif
1147
1148 return 0;
1149}
Paul Mundtb5073172007-07-21 04:37:25 -07001150EXPORT_SYMBOL(do_munmap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151
David Howells30340972006-09-27 01:50:20 -07001152asmlinkage long sys_munmap(unsigned long addr, size_t len)
1153{
1154 int ret;
1155 struct mm_struct *mm = current->mm;
1156
1157 down_write(&mm->mmap_sem);
1158 ret = do_munmap(mm, addr, len);
1159 up_write(&mm->mmap_sem);
1160 return ret;
1161}
1162
1163/*
1164 * Release all mappings
1165 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166void exit_mmap(struct mm_struct * mm)
1167{
1168 struct vm_list_struct *tmp;
1169
1170 if (mm) {
1171#ifdef DEBUG
1172 printk("Exit_mmap:\n");
1173#endif
1174
1175 mm->total_vm = 0;
1176
1177 while ((tmp = mm->context.vmlist)) {
1178 mm->context.vmlist = tmp->next;
Matt Helsley925d1c42008-04-29 01:01:36 -07001179 put_vma(mm, tmp->vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180
1181 realalloc -= kobjsize(tmp);
1182 askedalloc -= sizeof(*tmp);
1183 kfree(tmp);
1184 }
1185
1186#ifdef DEBUG
1187 show_process_blocks();
1188#endif
1189 }
1190}
1191
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192unsigned long do_brk(unsigned long addr, unsigned long len)
1193{
1194 return -ENOMEM;
1195}
1196
1197/*
David Howells6fa5f802006-09-27 01:50:21 -07001198 * expand (or shrink) an existing mapping, potentially moving it at the same
1199 * time (controlled by the MREMAP_MAYMOVE flag and available VM space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 *
David Howells6fa5f802006-09-27 01:50:21 -07001201 * under NOMMU conditions, we only permit changing a mapping's size, and only
1202 * as long as it stays within the hole allocated by the kmalloc() call in
1203 * do_mmap_pgoff() and the block is not shareable
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 *
David Howells6fa5f802006-09-27 01:50:21 -07001205 * MREMAP_FIXED is not supported under NOMMU conditions
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 */
1207unsigned long do_mremap(unsigned long addr,
1208 unsigned long old_len, unsigned long new_len,
1209 unsigned long flags, unsigned long new_addr)
1210{
David Howells6fa5f802006-09-27 01:50:21 -07001211 struct vm_area_struct *vma;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212
1213 /* insanity checks first */
1214 if (new_len == 0)
1215 return (unsigned long) -EINVAL;
1216
1217 if (flags & MREMAP_FIXED && new_addr != addr)
1218 return (unsigned long) -EINVAL;
1219
David Howells6fa5f802006-09-27 01:50:21 -07001220 vma = find_vma_exact(current->mm, addr);
1221 if (!vma)
1222 return (unsigned long) -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223
David Howells6fa5f802006-09-27 01:50:21 -07001224 if (vma->vm_end != vma->vm_start + old_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225 return (unsigned long) -EFAULT;
1226
David Howells6fa5f802006-09-27 01:50:21 -07001227 if (vma->vm_flags & VM_MAYSHARE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228 return (unsigned long) -EPERM;
1229
1230 if (new_len > kobjsize((void *) addr))
1231 return (unsigned long) -ENOMEM;
1232
1233 /* all checks complete - do it */
David Howells6fa5f802006-09-27 01:50:21 -07001234 vma->vm_end = vma->vm_start + new_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235
1236 askedalloc -= old_len;
1237 askedalloc += new_len;
1238
David Howells6fa5f802006-09-27 01:50:21 -07001239 return vma->vm_start;
1240}
Paul Mundtb5073172007-07-21 04:37:25 -07001241EXPORT_SYMBOL(do_mremap);
David Howells6fa5f802006-09-27 01:50:21 -07001242
1243asmlinkage unsigned long sys_mremap(unsigned long addr,
1244 unsigned long old_len, unsigned long new_len,
1245 unsigned long flags, unsigned long new_addr)
1246{
1247 unsigned long ret;
1248
1249 down_write(&current->mm->mmap_sem);
1250 ret = do_mremap(addr, old_len, new_len, flags, new_addr);
1251 up_write(&current->mm->mmap_sem);
1252 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253}
1254
Linus Torvalds6aab3412005-11-28 14:34:23 -08001255struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001256 unsigned int foll_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257{
1258 return NULL;
1259}
1260
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261int remap_pfn_range(struct vm_area_struct *vma, unsigned long from,
1262 unsigned long to, unsigned long size, pgprot_t prot)
1263{
Greg Ungerer66aa2b42005-09-12 11:18:10 +10001264 vma->vm_start = vma->vm_pgoff << PAGE_SHIFT;
1265 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266}
Luke Yang22c4af42006-07-14 00:24:09 -07001267EXPORT_SYMBOL(remap_pfn_range);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268
Paul Mundtf905bc42008-02-04 22:29:59 -08001269int remap_vmalloc_range(struct vm_area_struct *vma, void *addr,
1270 unsigned long pgoff)
1271{
1272 unsigned int size = vma->vm_end - vma->vm_start;
1273
1274 if (!(vma->vm_flags & VM_USERMAP))
1275 return -EINVAL;
1276
1277 vma->vm_start = (unsigned long)(addr + (pgoff << PAGE_SHIFT));
1278 vma->vm_end = vma->vm_start + size;
1279
1280 return 0;
1281}
1282EXPORT_SYMBOL(remap_vmalloc_range);
1283
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284void swap_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
1285{
1286}
1287
1288unsigned long arch_get_unmapped_area(struct file *file, unsigned long addr,
1289 unsigned long len, unsigned long pgoff, unsigned long flags)
1290{
1291 return -ENOMEM;
1292}
1293
Wolfgang Wander1363c3c2005-06-21 17:14:49 -07001294void arch_unmap_area(struct mm_struct *mm, unsigned long addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295{
1296}
1297
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298void unmap_mapping_range(struct address_space *mapping,
1299 loff_t const holebegin, loff_t const holelen,
1300 int even_cows)
1301{
1302}
Luke Yang22c4af42006-07-14 00:24:09 -07001303EXPORT_SYMBOL(unmap_mapping_range);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304
1305/*
David Howellsd56e03c2007-03-22 00:11:23 -08001306 * ask for an unmapped area at which to create a mapping on a file
1307 */
1308unsigned long get_unmapped_area(struct file *file, unsigned long addr,
1309 unsigned long len, unsigned long pgoff,
1310 unsigned long flags)
1311{
1312 unsigned long (*get_area)(struct file *, unsigned long, unsigned long,
1313 unsigned long, unsigned long);
1314
1315 get_area = current->mm->get_unmapped_area;
1316 if (file && file->f_op && file->f_op->get_unmapped_area)
1317 get_area = file->f_op->get_unmapped_area;
1318
1319 if (!get_area)
1320 return -ENOSYS;
1321
1322 return get_area(file, addr, len, pgoff, flags);
1323}
David Howellsd56e03c2007-03-22 00:11:23 -08001324EXPORT_SYMBOL(get_unmapped_area);
1325
1326/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 * Check that a process has enough memory to allocate a new virtual
1328 * mapping. 0 means there is enough memory for the allocation to
1329 * succeed and -ENOMEM implies there is not.
1330 *
1331 * We currently support three overcommit policies, which are set via the
1332 * vm.overcommit_memory sysctl. See Documentation/vm/overcommit-accounting
1333 *
1334 * Strict overcommit modes added 2002 Feb 26 by Alan Cox.
1335 * Additional code 2002 Jul 20 by Robert Love.
1336 *
1337 * cap_sys_admin is 1 if the process has admin privileges, 0 otherwise.
1338 *
1339 * Note this is a helper function intended to be used by LSMs which
1340 * wish to use this logic.
1341 */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001342int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343{
1344 unsigned long free, allowed;
1345
1346 vm_acct_memory(pages);
1347
1348 /*
1349 * Sometimes we want to use more memory than we have
1350 */
1351 if (sysctl_overcommit_memory == OVERCOMMIT_ALWAYS)
1352 return 0;
1353
1354 if (sysctl_overcommit_memory == OVERCOMMIT_GUESS) {
1355 unsigned long n;
1356
Christoph Lameter347ce432006-06-30 01:55:35 -07001357 free = global_page_state(NR_FILE_PAGES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 free += nr_swap_pages;
1359
1360 /*
1361 * Any slabs which are created with the
1362 * SLAB_RECLAIM_ACCOUNT flag claim to have contents
1363 * which are reclaimable, under pressure. The dentry
1364 * cache and most inode caches should fall into this
1365 */
Christoph Lameter972d1a72006-09-25 23:31:51 -07001366 free += global_page_state(NR_SLAB_RECLAIMABLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367
1368 /*
1369 * Leave the last 3% for root
1370 */
1371 if (!cap_sys_admin)
1372 free -= free / 32;
1373
1374 if (free > pages)
1375 return 0;
1376
1377 /*
1378 * nr_free_pages() is very expensive on large systems,
1379 * only call if we're about to fail.
1380 */
1381 n = nr_free_pages();
Hideo AOKId5ddc792006-04-10 22:53:01 -07001382
1383 /*
1384 * Leave reserved pages. The pages are not for anonymous pages.
1385 */
1386 if (n <= totalreserve_pages)
1387 goto error;
1388 else
1389 n -= totalreserve_pages;
1390
1391 /*
1392 * Leave the last 3% for root
1393 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 if (!cap_sys_admin)
1395 n -= n / 32;
1396 free += n;
1397
1398 if (free > pages)
1399 return 0;
Hideo AOKId5ddc792006-04-10 22:53:01 -07001400
1401 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 }
1403
1404 allowed = totalram_pages * sysctl_overcommit_ratio / 100;
1405 /*
1406 * Leave the last 3% for root
1407 */
1408 if (!cap_sys_admin)
1409 allowed -= allowed / 32;
1410 allowed += total_swap_pages;
1411
1412 /* Don't let a single process grow too big:
1413 leave 3% of the size of this process for other processes */
1414 allowed -= current->mm->total_vm / 32;
1415
Simon Derr2f60f8d2005-08-04 19:52:03 -07001416 /*
1417 * cast `allowed' as a signed long because vm_committed_space
1418 * sometimes has a negative value
1419 */
Alan Cox80119ef2008-05-23 13:04:31 -07001420 if (atomic_long_read(&vm_committed_space) < (long)allowed)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 return 0;
Hideo AOKId5ddc792006-04-10 22:53:01 -07001422error:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423 vm_unacct_memory(pages);
1424
1425 return -ENOMEM;
1426}
1427
1428int in_gate_area_no_task(unsigned long addr)
1429{
1430 return 0;
1431}
David Howellsb0e15192006-01-06 00:11:42 -08001432
Nick Piggind0217ac2007-07-19 01:47:03 -07001433int filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
David Howellsb0e15192006-01-06 00:11:42 -08001434{
1435 BUG();
Nick Piggind0217ac2007-07-19 01:47:03 -07001436 return 0;
David Howellsb0e15192006-01-06 00:11:42 -08001437}
Paul Mundtb5073172007-07-21 04:37:25 -07001438EXPORT_SYMBOL(filemap_fault);
David Howells0ec76a12006-09-27 01:50:15 -07001439
1440/*
1441 * Access another process' address space.
1442 * - source/target buffer must be kernel space
1443 */
1444int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write)
1445{
David Howells0ec76a12006-09-27 01:50:15 -07001446 struct vm_area_struct *vma;
1447 struct mm_struct *mm;
1448
1449 if (addr + len < addr)
1450 return 0;
1451
1452 mm = get_task_mm(tsk);
1453 if (!mm)
1454 return 0;
1455
1456 down_read(&mm->mmap_sem);
1457
1458 /* the access must start within one of the target process's mappings */
David Howells0159b142006-09-27 01:50:16 -07001459 vma = find_vma(mm, addr);
1460 if (vma) {
David Howells0ec76a12006-09-27 01:50:15 -07001461 /* don't overrun this mapping */
1462 if (addr + len >= vma->vm_end)
1463 len = vma->vm_end - addr;
1464
1465 /* only read or write mappings where it is permitted */
David Howellsd00c7b992006-09-27 01:50:19 -07001466 if (write && vma->vm_flags & VM_MAYWRITE)
David Howells0ec76a12006-09-27 01:50:15 -07001467 len -= copy_to_user((void *) addr, buf, len);
David Howellsd00c7b992006-09-27 01:50:19 -07001468 else if (!write && vma->vm_flags & VM_MAYREAD)
David Howells0ec76a12006-09-27 01:50:15 -07001469 len -= copy_from_user(buf, (void *) addr, len);
1470 else
1471 len = 0;
1472 } else {
1473 len = 0;
1474 }
1475
1476 up_read(&mm->mmap_sem);
1477 mmput(mm);
1478 return len;
1479}