blob: db5fd1795298764d06abc9a6fdbc1657cc3dcd2b [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 *
David Howells8feae132009-01-08 12:04:47 +00009 * Copyright (c) 2004-2008 David Howells <dhowells@redhat.com>
Linus Torvalds1da177e2005-04-16 15:20:36 -070010 * 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 Mundt29c185e2010-12-24 12:08:30 +090013 * Copyright (c) 2007-2010 Paul Mundt <lethal@linux-sh.org>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014 */
15
Mitchel Humpherysb1de0d12014-06-06 14:38:30 -070016#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17
Paul Gortmakerb95f1b312011-10-16 02:01:52 -040018#include <linux/export.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/mm.h>
Davidlohr Bueso615d6e82014-04-07 15:37:25 -070020#include <linux/vmacache.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/mman.h>
22#include <linux/swap.h>
23#include <linux/file.h>
24#include <linux/highmem.h>
25#include <linux/pagemap.h>
26#include <linux/slab.h>
27#include <linux/vmalloc.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include <linux/blkdev.h>
29#include <linux/backing-dev.h>
Gideon Israel Dsouza3b321232014-04-07 15:37:26 -070030#include <linux/compiler.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <linux/mount.h>
32#include <linux/personality.h>
33#include <linux/security.h>
34#include <linux/syscalls.h>
Al Viro120a7952010-10-30 02:54:44 -040035#include <linux/audit.h>
Mitchel Humpherysb1de0d12014-06-06 14:38:30 -070036#include <linux/printk.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
38#include <asm/uaccess.h>
39#include <asm/tlb.h>
40#include <asm/tlbflush.h>
Bernd Schmidteb8cdec2009-09-21 17:03:57 -070041#include <asm/mmu_context.h>
David Howells8feae132009-01-08 12:04:47 +000042#include "internal.h"
43
Linus Torvalds1da177e2005-04-16 15:20:36 -070044void *high_memory;
Arnd Bergmann944b6872015-02-05 12:25:12 -080045EXPORT_SYMBOL(high_memory);
Linus Torvalds1da177e2005-04-16 15:20:36 -070046struct page *mem_map;
47unsigned long max_mapnr;
gchen gchen5b8bf302015-03-12 16:26:05 -070048EXPORT_SYMBOL(max_mapnr);
Hugh Dickins4266c972009-09-23 17:05:53 +010049unsigned long highest_memmap_pfn;
David Howellsfc4d5c22009-05-06 16:03:05 -070050int sysctl_nr_trim_pages = CONFIG_NOMMU_INITIAL_TRIM_EXCESS;
Linus Torvalds1da177e2005-04-16 15:20:36 -070051int heap_stack_gap = 0;
52
David Howells33e5d7692009-04-02 16:56:32 -070053atomic_long_t mmap_pages_allocated;
David Howells8feae132009-01-08 12:04:47 +000054
Linus Torvalds1da177e2005-04-16 15:20:36 -070055EXPORT_SYMBOL(mem_map);
Linus Torvalds1da177e2005-04-16 15:20:36 -070056
David Howells8feae132009-01-08 12:04:47 +000057/* list of mapped, potentially shareable regions */
58static struct kmem_cache *vm_region_jar;
59struct rb_root nommu_region_tree = RB_ROOT;
60DECLARE_RWSEM(nommu_region_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +040062const struct vm_operations_struct generic_file_vm_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070063};
64
65/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070066 * Return the total memory allocated for this pointer, not
67 * just what the caller asked for.
68 *
69 * Doesn't have to be accurate, i.e. may have races.
70 */
71unsigned int kobjsize(const void *objp)
72{
73 struct page *page;
74
Michael Hennerich4016a132008-04-28 02:13:38 -070075 /*
76 * If the object we have should not have ksize performed on it,
77 * return size of 0
78 */
Paul Mundt5a1603b2008-06-12 16:29:55 +090079 if (!objp || !virt_addr_valid(objp))
Paul Mundt6cfd53fc2008-06-05 22:46:08 -070080 return 0;
81
82 page = virt_to_head_page(objp);
Paul Mundt6cfd53fc2008-06-05 22:46:08 -070083
84 /*
85 * If the allocator sets PageSlab, we know the pointer came from
86 * kmalloc().
87 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 if (PageSlab(page))
89 return ksize(objp);
90
Paul Mundt6cfd53fc2008-06-05 22:46:08 -070091 /*
Paul Mundtab2e83e2009-01-08 12:04:48 +000092 * If it's not a compound page, see if we have a matching VMA
93 * region. This test is intentionally done in reverse order,
94 * so if there's no VMA, we still fall through and hand back
95 * PAGE_SIZE for 0-order pages.
96 */
97 if (!PageCompound(page)) {
98 struct vm_area_struct *vma;
99
100 vma = find_vma(current->mm, (unsigned long)objp);
101 if (vma)
102 return vma->vm_end - vma->vm_start;
103 }
104
105 /*
Paul Mundt6cfd53fc2008-06-05 22:46:08 -0700106 * The ksize() function is only guaranteed to work for pointers
Paul Mundt5a1603b2008-06-12 16:29:55 +0900107 * returned by kmalloc(). So handle arbitrary pointers here.
Paul Mundt6cfd53fc2008-06-05 22:46:08 -0700108 */
Paul Mundt5a1603b2008-06-12 16:29:55 +0900109 return PAGE_SIZE << compound_order(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110}
111
Michel Lespinasse28a35712013-02-22 16:35:55 -0800112long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
113 unsigned long start, unsigned long nr_pages,
114 unsigned int foll_flags, struct page **pages,
115 struct vm_area_struct **vmas, int *nonblocking)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116{
Sonic Zhang910e46d2006-09-27 01:50:17 -0700117 struct vm_area_struct *vma;
David Howells7b4d5b82006-09-27 01:50:18 -0700118 unsigned long vm_flags;
119 int i;
120
121 /* calculate required read or write permissions.
Hugh Dickins58fa8792009-09-21 17:03:31 -0700122 * If FOLL_FORCE is set, we only require the "MAY" flags.
David Howells7b4d5b82006-09-27 01:50:18 -0700123 */
Hugh Dickins58fa8792009-09-21 17:03:31 -0700124 vm_flags = (foll_flags & FOLL_WRITE) ?
125 (VM_WRITE | VM_MAYWRITE) : (VM_READ | VM_MAYREAD);
126 vm_flags &= (foll_flags & FOLL_FORCE) ?
127 (VM_MAYREAD | VM_MAYWRITE) : (VM_READ | VM_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128
Peter Zijlstra9d737772009-06-25 11:58:55 +0200129 for (i = 0; i < nr_pages; i++) {
David Howells7561e8c2010-03-25 16:48:38 +0000130 vma = find_vma(mm, start);
David Howells7b4d5b82006-09-27 01:50:18 -0700131 if (!vma)
132 goto finish_or_fault;
133
134 /* protect what we can, including chardevs */
Hugh Dickins1c3aff12009-09-21 17:03:24 -0700135 if ((vma->vm_flags & (VM_IO | VM_PFNMAP)) ||
136 !(vm_flags & vma->vm_flags))
David Howells7b4d5b82006-09-27 01:50:18 -0700137 goto finish_or_fault;
Sonic Zhang910e46d2006-09-27 01:50:17 -0700138
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139 if (pages) {
140 pages[i] = virt_to_page(start);
141 if (pages[i])
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300142 get_page(pages[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 }
144 if (vmas)
Sonic Zhang910e46d2006-09-27 01:50:17 -0700145 vmas[i] = vma;
David Howellse1ee65d2010-03-25 16:48:44 +0000146 start = (start + PAGE_SIZE) & PAGE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 }
David Howells7b4d5b82006-09-27 01:50:18 -0700148
149 return i;
150
151finish_or_fault:
152 return i ? : -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153}
Nick Pigginb291f002008-10-18 20:26:44 -0700154
Nick Pigginb291f002008-10-18 20:26:44 -0700155/*
156 * get a list of pages in an address range belonging to the specified process
157 * and indicate the VMA that covers each page
158 * - this is potentially dodgy as we may end incrementing the page count of a
159 * slab page or a secondary page from a compound page
160 * - don't permit access to VMAs that don't support it, such as I/O mappings
161 */
Ingo Molnarc12d2da2016-04-04 10:24:58 +0200162long get_user_pages(unsigned long start, unsigned long nr_pages,
Lorenzo Stoakes768ae302016-10-13 01:20:16 +0100163 unsigned int gup_flags, struct page **pages,
Michel Lespinasse28a35712013-02-22 16:35:55 -0800164 struct vm_area_struct **vmas)
Nick Pigginb291f002008-10-18 20:26:44 -0700165{
Lorenzo Stoakes768ae302016-10-13 01:20:16 +0100166 return __get_user_pages(current, current->mm, start, nr_pages,
167 gup_flags, pages, vmas, NULL);
Nick Pigginb291f002008-10-18 20:26:44 -0700168}
Ingo Molnarc12d2da2016-04-04 10:24:58 +0200169EXPORT_SYMBOL(get_user_pages);
Greg Ungerer66aa2b42005-09-12 11:18:10 +1000170
Ingo Molnarc12d2da2016-04-04 10:24:58 +0200171long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
Lorenzo Stoakes3b913172016-10-13 01:20:14 +0100172 unsigned int gup_flags, struct page **pages,
Dave Hansencde70142016-02-12 13:01:55 -0800173 int *locked)
Andrea Arcangelif0818f42015-02-11 15:27:17 -0800174{
Lorenzo Stoakes768ae302016-10-13 01:20:16 +0100175 return get_user_pages(start, nr_pages, gup_flags, pages, NULL);
Andrea Arcangelif0818f42015-02-11 15:27:17 -0800176}
Ingo Molnarc12d2da2016-04-04 10:24:58 +0200177EXPORT_SYMBOL(get_user_pages_locked);
Andrea Arcangelif0818f42015-02-11 15:27:17 -0800178
Andrea Arcangeli0fd71a52015-02-11 15:27:20 -0800179long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
180 unsigned long start, unsigned long nr_pages,
Lorenzo Stoakesd4944b02016-10-13 01:20:12 +0100181 struct page **pages, unsigned int gup_flags)
Andrea Arcangelif0818f42015-02-11 15:27:17 -0800182{
183 long ret;
184 down_read(&mm->mmap_sem);
Dave Hansencde70142016-02-12 13:01:55 -0800185 ret = __get_user_pages(tsk, mm, start, nr_pages, gup_flags, pages,
186 NULL, NULL);
Andrea Arcangelif0818f42015-02-11 15:27:17 -0800187 up_read(&mm->mmap_sem);
188 return ret;
189}
Andrea Arcangeli0fd71a52015-02-11 15:27:20 -0800190EXPORT_SYMBOL(__get_user_pages_unlocked);
191
Ingo Molnarc12d2da2016-04-04 10:24:58 +0200192long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
Lorenzo Stoakesc1641542016-10-13 01:20:13 +0100193 struct page **pages, unsigned int gup_flags)
Andrea Arcangeli0fd71a52015-02-11 15:27:20 -0800194{
Dave Hansencde70142016-02-12 13:01:55 -0800195 return __get_user_pages_unlocked(current, current->mm, start, nr_pages,
Lorenzo Stoakesc1641542016-10-13 01:20:13 +0100196 pages, gup_flags);
Andrea Arcangeli0fd71a52015-02-11 15:27:20 -0800197}
Ingo Molnarc12d2da2016-04-04 10:24:58 +0200198EXPORT_SYMBOL(get_user_pages_unlocked);
Andrea Arcangelif0818f42015-02-11 15:27:17 -0800199
Paul Mundtdfc2f912009-06-26 04:31:57 +0900200/**
201 * follow_pfn - look up PFN at a user virtual address
202 * @vma: memory mapping
203 * @address: user virtual address
204 * @pfn: location to store found PFN
205 *
206 * Only IO mappings and raw PFN mappings are allowed.
207 *
208 * Returns zero and the pfn at @pfn on success, -ve otherwise.
209 */
210int follow_pfn(struct vm_area_struct *vma, unsigned long address,
211 unsigned long *pfn)
212{
213 if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
214 return -EINVAL;
215
216 *pfn = address >> PAGE_SHIFT;
217 return 0;
218}
219EXPORT_SYMBOL(follow_pfn);
220
Joonsoo Kimf1c40692013-04-29 15:07:37 -0700221LIST_HEAD(vmap_area_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800223void vfree(const void *addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224{
225 kfree(addr);
226}
Paul Mundtb5073172007-07-21 04:37:25 -0700227EXPORT_SYMBOL(vfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228
Al Virodd0fc662005-10-07 07:46:04 +0100229void *__vmalloc(unsigned long size, gfp_t gfp_mask, pgprot_t prot)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230{
231 /*
Robert P. J. Day85186092007-10-19 23:11:38 +0200232 * You can't specify __GFP_HIGHMEM with kmalloc() since kmalloc()
233 * returns only a logical address.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 */
Nick Piggin84097512006-03-22 00:08:34 -0800235 return kmalloc(size, (gfp_mask | __GFP_COMP) & ~__GFP_HIGHMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236}
Paul Mundtb5073172007-07-21 04:37:25 -0700237EXPORT_SYMBOL(__vmalloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238
Paul Mundtf905bc42008-02-04 22:29:59 -0800239void *vmalloc_user(unsigned long size)
240{
241 void *ret;
242
243 ret = __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM | __GFP_ZERO,
244 PAGE_KERNEL);
245 if (ret) {
246 struct vm_area_struct *vma;
247
248 down_write(&current->mm->mmap_sem);
249 vma = find_vma(current->mm, (unsigned long)ret);
250 if (vma)
251 vma->vm_flags |= VM_USERMAP;
252 up_write(&current->mm->mmap_sem);
253 }
254
255 return ret;
256}
257EXPORT_SYMBOL(vmalloc_user);
258
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800259struct page *vmalloc_to_page(const void *addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260{
261 return virt_to_page(addr);
262}
Paul Mundtb5073172007-07-21 04:37:25 -0700263EXPORT_SYMBOL(vmalloc_to_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800265unsigned long vmalloc_to_pfn(const void *addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266{
267 return page_to_pfn(virt_to_page(addr));
268}
Paul Mundtb5073172007-07-21 04:37:25 -0700269EXPORT_SYMBOL(vmalloc_to_pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
271long vread(char *buf, char *addr, unsigned long count)
272{
Chen Gang9bde9162013-07-03 15:02:36 -0700273 /* Don't allow overflow */
274 if ((unsigned long) buf + count < count)
275 count = -(unsigned long) buf;
276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 memcpy(buf, addr, count);
278 return count;
279}
280
281long vwrite(char *buf, char *addr, unsigned long count)
282{
283 /* Don't allow overflow */
284 if ((unsigned long) addr + count < count)
285 count = -(unsigned long) addr;
286
287 memcpy(addr, buf, count);
Choi Gi-yongac714902014-04-07 15:37:36 -0700288 return count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289}
290
291/*
Masahiro Yamadae1c05062015-07-07 10:14:59 +0900292 * vmalloc - allocate virtually contiguous memory
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 *
294 * @size: allocation size
295 *
296 * Allocate enough pages to cover @size from the page level
Masahiro Yamadae1c05062015-07-07 10:14:59 +0900297 * allocator and map them into contiguous kernel virtual space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 *
Michael Opdenackerc1c88972006-10-03 23:21:02 +0200299 * For tight control over page level allocator and protection flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 * use __vmalloc() instead.
301 */
302void *vmalloc(unsigned long size)
303{
304 return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM, PAGE_KERNEL);
305}
Andrew Mortonf6138882006-02-28 16:59:18 -0800306EXPORT_SYMBOL(vmalloc);
307
Dave Younge1ca7782010-10-26 14:22:06 -0700308/*
Masahiro Yamadae1c05062015-07-07 10:14:59 +0900309 * vzalloc - allocate virtually contiguous memory with zero fill
Dave Younge1ca7782010-10-26 14:22:06 -0700310 *
311 * @size: allocation size
312 *
313 * Allocate enough pages to cover @size from the page level
Masahiro Yamadae1c05062015-07-07 10:14:59 +0900314 * allocator and map them into contiguous kernel virtual space.
Dave Younge1ca7782010-10-26 14:22:06 -0700315 * The memory allocated is set to zero.
316 *
317 * For tight control over page level allocator and protection flags
318 * use __vmalloc() instead.
319 */
320void *vzalloc(unsigned long size)
321{
322 return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM | __GFP_ZERO,
323 PAGE_KERNEL);
324}
325EXPORT_SYMBOL(vzalloc);
326
327/**
328 * vmalloc_node - allocate memory on a specific node
329 * @size: allocation size
330 * @node: numa node
331 *
332 * Allocate enough pages to cover @size from the page level
333 * allocator and map them into contiguous kernel virtual space.
334 *
335 * For tight control over page level allocator and protection flags
336 * use __vmalloc() instead.
337 */
Andrew Mortonf6138882006-02-28 16:59:18 -0800338void *vmalloc_node(unsigned long size, int node)
339{
340 return vmalloc(size);
341}
Paul Mundt9a14f652010-12-24 11:50:34 +0900342EXPORT_SYMBOL(vmalloc_node);
Dave Younge1ca7782010-10-26 14:22:06 -0700343
344/**
345 * vzalloc_node - allocate memory on a specific node with zero fill
346 * @size: allocation size
347 * @node: numa node
348 *
349 * Allocate enough pages to cover @size from the page level
350 * allocator and map them into contiguous kernel virtual space.
351 * The memory allocated is set to zero.
352 *
353 * For tight control over page level allocator and protection flags
354 * use __vmalloc() instead.
355 */
356void *vzalloc_node(unsigned long size, int node)
357{
358 return vzalloc(size);
359}
360EXPORT_SYMBOL(vzalloc_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361
Paul Mundt1af446e2008-08-04 16:01:47 +0900362#ifndef PAGE_KERNEL_EXEC
363# define PAGE_KERNEL_EXEC PAGE_KERNEL
364#endif
365
366/**
367 * vmalloc_exec - allocate virtually contiguous, executable memory
368 * @size: allocation size
369 *
370 * Kernel-internal function to allocate enough pages to cover @size
371 * the page level allocator and map them into contiguous and
372 * executable kernel virtual space.
373 *
374 * For tight control over page level allocator and protection flags
375 * use __vmalloc() instead.
376 */
377
378void *vmalloc_exec(unsigned long size)
379{
380 return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM, PAGE_KERNEL_EXEC);
381}
382
Paul Mundtb5073172007-07-21 04:37:25 -0700383/**
384 * vmalloc_32 - allocate virtually contiguous memory (32bit addressable)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 * @size: allocation size
386 *
387 * Allocate enough 32bit PA addressable pages to cover @size from the
Masahiro Yamadae1c05062015-07-07 10:14:59 +0900388 * page level allocator and map them into contiguous kernel virtual space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 */
390void *vmalloc_32(unsigned long size)
391{
392 return __vmalloc(size, GFP_KERNEL, PAGE_KERNEL);
393}
Paul Mundtb5073172007-07-21 04:37:25 -0700394EXPORT_SYMBOL(vmalloc_32);
395
396/**
397 * vmalloc_32_user - allocate zeroed virtually contiguous 32bit memory
398 * @size: allocation size
399 *
400 * The resulting memory area is 32bit addressable and zeroed so it can be
401 * mapped to userspace without leaking data.
Paul Mundtf905bc42008-02-04 22:29:59 -0800402 *
403 * VM_USERMAP is set on the corresponding VMA so that subsequent calls to
404 * remap_vmalloc_range() are permissible.
Paul Mundtb5073172007-07-21 04:37:25 -0700405 */
406void *vmalloc_32_user(unsigned long size)
407{
Paul Mundtf905bc42008-02-04 22:29:59 -0800408 /*
409 * We'll have to sort out the ZONE_DMA bits for 64-bit,
410 * but for now this can simply use vmalloc_user() directly.
411 */
412 return vmalloc_user(size);
Paul Mundtb5073172007-07-21 04:37:25 -0700413}
414EXPORT_SYMBOL(vmalloc_32_user);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415
416void *vmap(struct page **pages, unsigned int count, unsigned long flags, pgprot_t prot)
417{
418 BUG();
419 return NULL;
420}
Paul Mundtb5073172007-07-21 04:37:25 -0700421EXPORT_SYMBOL(vmap);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800423void vunmap(const void *addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424{
425 BUG();
426}
Paul Mundtb5073172007-07-21 04:37:25 -0700427EXPORT_SYMBOL(vunmap);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428
Paul Mundteb6434d2009-01-21 17:45:47 +0900429void *vm_map_ram(struct page **pages, unsigned int count, int node, pgprot_t prot)
430{
431 BUG();
432 return NULL;
433}
434EXPORT_SYMBOL(vm_map_ram);
435
436void vm_unmap_ram(const void *mem, unsigned int count)
437{
438 BUG();
439}
440EXPORT_SYMBOL(vm_unmap_ram);
441
442void vm_unmap_aliases(void)
443{
444}
445EXPORT_SYMBOL_GPL(vm_unmap_aliases);
446
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447/*
Christoph Hellwig1eeb66a2007-05-08 00:27:03 -0700448 * Implement a stub for vmalloc_sync_all() if the architecture chose not to
449 * have one.
450 */
Gideon Israel Dsouza3b321232014-04-07 15:37:26 -0700451void __weak vmalloc_sync_all(void)
Christoph Hellwig1eeb66a2007-05-08 00:27:03 -0700452{
453}
454
Paul Mundt29c185e2010-12-24 12:08:30 +0900455/**
456 * alloc_vm_area - allocate a range of kernel address space
457 * @size: size of the area
458 *
459 * Returns: NULL on failure, vm_struct on success
460 *
461 * This function reserves a range of kernel address space, and
462 * allocates pagetables to map that range. No actual mappings
463 * are created. If the kernel address space is not shared
464 * between processes, it syncs the pagetable across all
465 * processes.
466 */
David Vrabelcd129092011-09-29 16:53:32 +0100467struct vm_struct *alloc_vm_area(size_t size, pte_t **ptes)
Paul Mundt29c185e2010-12-24 12:08:30 +0900468{
469 BUG();
470 return NULL;
471}
472EXPORT_SYMBOL_GPL(alloc_vm_area);
473
474void free_vm_area(struct vm_struct *area)
475{
476 BUG();
477}
478EXPORT_SYMBOL_GPL(free_vm_area);
479
Paul Mundtb5073172007-07-21 04:37:25 -0700480int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
481 struct page *page)
482{
483 return -EINVAL;
484}
485EXPORT_SYMBOL(vm_insert_page);
486
Christoph Hellwig1eeb66a2007-05-08 00:27:03 -0700487/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 * sys_brk() for the most part doesn't need the global kernel
489 * lock, except when an application is doing something nasty
490 * like trying to un-brk an area that has already been mapped
491 * to a regular file. in this case, the unmapping will need
492 * to invoke file system routines that need the global lock.
493 */
Heiko Carstens6a6160a2009-01-14 14:14:15 +0100494SYSCALL_DEFINE1(brk, unsigned long, brk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495{
496 struct mm_struct *mm = current->mm;
497
498 if (brk < mm->start_brk || brk > mm->context.end_brk)
499 return mm->brk;
500
501 if (mm->brk == brk)
502 return mm->brk;
503
504 /*
505 * Always allow shrinking brk
506 */
507 if (brk <= mm->brk) {
508 mm->brk = brk;
509 return brk;
510 }
511
512 /*
513 * Ok, looks good - let it rip.
514 */
Mike Frysingercfe79c02010-01-06 17:23:23 +0000515 flush_icache_range(mm->brk, brk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 return mm->brk = brk;
517}
518
David Howells8feae132009-01-08 12:04:47 +0000519/*
520 * initialise the VMA and region record slabs
521 */
522void __init mmap_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523{
KOSAKI Motohiro00a62ce2009-04-30 15:08:51 -0700524 int ret;
525
Tejun Heo908c7f12014-09-08 09:51:29 +0900526 ret = percpu_counter_init(&vm_committed_as, 0, GFP_KERNEL);
KOSAKI Motohiro00a62ce2009-04-30 15:08:51 -0700527 VM_BUG_ON(ret);
Vladimir Davydov5d097052016-01-14 15:18:21 -0800528 vm_region_jar = KMEM_CACHE(vm_region, SLAB_PANIC|SLAB_ACCOUNT);
David Howells8feae132009-01-08 12:04:47 +0000529}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530
David Howells8feae132009-01-08 12:04:47 +0000531/*
532 * validate the region tree
533 * - the caller must hold the region lock
534 */
535#ifdef CONFIG_DEBUG_NOMMU_REGIONS
536static noinline void validate_nommu_regions(void)
537{
538 struct vm_region *region, *last;
539 struct rb_node *p, *lastp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
David Howells8feae132009-01-08 12:04:47 +0000541 lastp = rb_first(&nommu_region_tree);
542 if (!lastp)
543 return;
544
545 last = rb_entry(lastp, struct vm_region, vm_rb);
Geliang Tangc9427bc2015-11-05 18:48:38 -0800546 BUG_ON(last->vm_end <= last->vm_start);
547 BUG_ON(last->vm_top < last->vm_end);
David Howells8feae132009-01-08 12:04:47 +0000548
549 while ((p = rb_next(lastp))) {
550 region = rb_entry(p, struct vm_region, vm_rb);
551 last = rb_entry(lastp, struct vm_region, vm_rb);
552
Geliang Tangc9427bc2015-11-05 18:48:38 -0800553 BUG_ON(region->vm_end <= region->vm_start);
554 BUG_ON(region->vm_top < region->vm_end);
555 BUG_ON(region->vm_start < last->vm_top);
David Howells8feae132009-01-08 12:04:47 +0000556
557 lastp = p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 }
559}
David Howells8feae132009-01-08 12:04:47 +0000560#else
David Howells33e5d7692009-04-02 16:56:32 -0700561static void validate_nommu_regions(void)
562{
563}
David Howells8feae132009-01-08 12:04:47 +0000564#endif
565
566/*
567 * add a region into the global tree
568 */
569static void add_nommu_region(struct vm_region *region)
570{
571 struct vm_region *pregion;
572 struct rb_node **p, *parent;
573
574 validate_nommu_regions();
575
David Howells8feae132009-01-08 12:04:47 +0000576 parent = NULL;
577 p = &nommu_region_tree.rb_node;
578 while (*p) {
579 parent = *p;
580 pregion = rb_entry(parent, struct vm_region, vm_rb);
581 if (region->vm_start < pregion->vm_start)
582 p = &(*p)->rb_left;
583 else if (region->vm_start > pregion->vm_start)
584 p = &(*p)->rb_right;
585 else if (pregion == region)
586 return;
587 else
588 BUG();
589 }
590
591 rb_link_node(&region->vm_rb, parent, p);
592 rb_insert_color(&region->vm_rb, &nommu_region_tree);
593
594 validate_nommu_regions();
595}
596
597/*
598 * delete a region from the global tree
599 */
600static void delete_nommu_region(struct vm_region *region)
601{
602 BUG_ON(!nommu_region_tree.rb_node);
603
604 validate_nommu_regions();
605 rb_erase(&region->vm_rb, &nommu_region_tree);
606 validate_nommu_regions();
607}
608
609/*
610 * free a contiguous series of pages
611 */
612static void free_page_series(unsigned long from, unsigned long to)
613{
614 for (; from < to; from += PAGE_SIZE) {
615 struct page *page = virt_to_page(from);
616
David Howells33e5d7692009-04-02 16:56:32 -0700617 atomic_long_dec(&mmap_pages_allocated);
David Howells8feae132009-01-08 12:04:47 +0000618 put_page(page);
619 }
620}
621
622/*
623 * release a reference to a region
David Howells33e5d7692009-04-02 16:56:32 -0700624 * - the caller must hold the region semaphore for writing, which this releases
Paul Mundtdd8632a2009-01-08 12:04:47 +0000625 * - the region may not have been added to the tree yet, in which case vm_top
David Howells8feae132009-01-08 12:04:47 +0000626 * will equal vm_start
627 */
628static void __put_nommu_region(struct vm_region *region)
629 __releases(nommu_region_sem)
630{
David Howells8feae132009-01-08 12:04:47 +0000631 BUG_ON(!nommu_region_tree.rb_node);
632
David Howells1e2ae592010-01-15 17:01:33 -0800633 if (--region->vm_usage == 0) {
Paul Mundtdd8632a2009-01-08 12:04:47 +0000634 if (region->vm_top > region->vm_start)
David Howells8feae132009-01-08 12:04:47 +0000635 delete_nommu_region(region);
636 up_write(&nommu_region_sem);
637
638 if (region->vm_file)
639 fput(region->vm_file);
640
641 /* IO memory and memory shared directly out of the pagecache
642 * from ramfs/tmpfs mustn't be released here */
Leon Romanovsky22cc8772015-06-24 16:57:47 -0700643 if (region->vm_flags & VM_MAPPED_COPY)
Paul Mundtdd8632a2009-01-08 12:04:47 +0000644 free_page_series(region->vm_start, region->vm_top);
David Howells8feae132009-01-08 12:04:47 +0000645 kmem_cache_free(vm_region_jar, region);
646 } else {
647 up_write(&nommu_region_sem);
648 }
649}
650
651/*
652 * release a reference to a region
653 */
654static void put_nommu_region(struct vm_region *region)
655{
656 down_write(&nommu_region_sem);
657 __put_nommu_region(region);
658}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659
David Howells30340972006-09-27 01:50:20 -0700660/*
Bernd Schmidteb8cdec2009-09-21 17:03:57 -0700661 * update protection on a vma
662 */
663static void protect_vma(struct vm_area_struct *vma, unsigned long flags)
664{
665#ifdef CONFIG_MPU
666 struct mm_struct *mm = vma->vm_mm;
667 long start = vma->vm_start & PAGE_MASK;
668 while (start < vma->vm_end) {
669 protect_page(mm, start, flags);
670 start += PAGE_SIZE;
671 }
672 update_protections(mm);
673#endif
674}
675
676/*
David Howells30340972006-09-27 01:50:20 -0700677 * add a VMA into a process's mm_struct in the appropriate place in the list
David Howells8feae132009-01-08 12:04:47 +0000678 * and tree and add to the address space's page tree also if not an anonymous
679 * page
David Howells30340972006-09-27 01:50:20 -0700680 * - should be called with mm->mmap_sem held writelocked
681 */
David Howells8feae132009-01-08 12:04:47 +0000682static void add_vma_to_mm(struct mm_struct *mm, struct vm_area_struct *vma)
David Howells30340972006-09-27 01:50:20 -0700683{
Namhyung Kim6038def2011-05-24 17:11:22 -0700684 struct vm_area_struct *pvma, *prev;
David Howells8feae132009-01-08 12:04:47 +0000685 struct address_space *mapping;
Namhyung Kim6038def2011-05-24 17:11:22 -0700686 struct rb_node **p, *parent, *rb_prev;
David Howells30340972006-09-27 01:50:20 -0700687
David Howells8feae132009-01-08 12:04:47 +0000688 BUG_ON(!vma->vm_region);
689
690 mm->map_count++;
691 vma->vm_mm = mm;
692
Bernd Schmidteb8cdec2009-09-21 17:03:57 -0700693 protect_vma(vma, vma->vm_flags);
694
David Howells8feae132009-01-08 12:04:47 +0000695 /* add the VMA to the mapping */
696 if (vma->vm_file) {
697 mapping = vma->vm_file->f_mapping;
698
Davidlohr Bueso83cde9e2014-12-12 16:54:21 -0800699 i_mmap_lock_write(mapping);
David Howells8feae132009-01-08 12:04:47 +0000700 flush_dcache_mmap_lock(mapping);
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -0700701 vma_interval_tree_insert(vma, &mapping->i_mmap);
David Howells8feae132009-01-08 12:04:47 +0000702 flush_dcache_mmap_unlock(mapping);
Davidlohr Bueso83cde9e2014-12-12 16:54:21 -0800703 i_mmap_unlock_write(mapping);
David Howells8feae132009-01-08 12:04:47 +0000704 }
705
706 /* add the VMA to the tree */
Namhyung Kim6038def2011-05-24 17:11:22 -0700707 parent = rb_prev = NULL;
David Howells8feae132009-01-08 12:04:47 +0000708 p = &mm->mm_rb.rb_node;
709 while (*p) {
710 parent = *p;
711 pvma = rb_entry(parent, struct vm_area_struct, vm_rb);
712
713 /* sort by: start addr, end addr, VMA struct addr in that order
714 * (the latter is necessary as we may get identical VMAs) */
715 if (vma->vm_start < pvma->vm_start)
716 p = &(*p)->rb_left;
Namhyung Kim6038def2011-05-24 17:11:22 -0700717 else if (vma->vm_start > pvma->vm_start) {
718 rb_prev = parent;
David Howells8feae132009-01-08 12:04:47 +0000719 p = &(*p)->rb_right;
Namhyung Kim6038def2011-05-24 17:11:22 -0700720 } else if (vma->vm_end < pvma->vm_end)
David Howells8feae132009-01-08 12:04:47 +0000721 p = &(*p)->rb_left;
Namhyung Kim6038def2011-05-24 17:11:22 -0700722 else if (vma->vm_end > pvma->vm_end) {
723 rb_prev = parent;
David Howells8feae132009-01-08 12:04:47 +0000724 p = &(*p)->rb_right;
Namhyung Kim6038def2011-05-24 17:11:22 -0700725 } else if (vma < pvma)
David Howells8feae132009-01-08 12:04:47 +0000726 p = &(*p)->rb_left;
Namhyung Kim6038def2011-05-24 17:11:22 -0700727 else if (vma > pvma) {
728 rb_prev = parent;
David Howells8feae132009-01-08 12:04:47 +0000729 p = &(*p)->rb_right;
Namhyung Kim6038def2011-05-24 17:11:22 -0700730 } else
David Howells8feae132009-01-08 12:04:47 +0000731 BUG();
732 }
733
734 rb_link_node(&vma->vm_rb, parent, p);
735 rb_insert_color(&vma->vm_rb, &mm->mm_rb);
736
737 /* add VMA to the VMA list also */
Namhyung Kim6038def2011-05-24 17:11:22 -0700738 prev = NULL;
739 if (rb_prev)
740 prev = rb_entry(rb_prev, struct vm_area_struct, vm_rb);
David Howells30340972006-09-27 01:50:20 -0700741
Namhyung Kim6038def2011-05-24 17:11:22 -0700742 __vma_link_list(mm, vma, prev, parent);
David Howells8feae132009-01-08 12:04:47 +0000743}
744
745/*
746 * delete a VMA from its owning mm_struct and address space
747 */
748static void delete_vma_from_mm(struct vm_area_struct *vma)
749{
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700750 int i;
David Howells8feae132009-01-08 12:04:47 +0000751 struct address_space *mapping;
752 struct mm_struct *mm = vma->vm_mm;
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700753 struct task_struct *curr = current;
David Howells8feae132009-01-08 12:04:47 +0000754
Bernd Schmidteb8cdec2009-09-21 17:03:57 -0700755 protect_vma(vma, 0);
756
David Howells8feae132009-01-08 12:04:47 +0000757 mm->map_count--;
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700758 for (i = 0; i < VMACACHE_SIZE; i++) {
759 /* if the vma is cached, invalidate the entire cache */
760 if (curr->vmacache[i] == vma) {
Steven Miaoe020d5b2014-06-23 13:22:02 -0700761 vmacache_invalidate(mm);
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700762 break;
763 }
764 }
David Howells8feae132009-01-08 12:04:47 +0000765
766 /* remove the VMA from the mapping */
767 if (vma->vm_file) {
768 mapping = vma->vm_file->f_mapping;
769
Davidlohr Bueso83cde9e2014-12-12 16:54:21 -0800770 i_mmap_lock_write(mapping);
David Howells8feae132009-01-08 12:04:47 +0000771 flush_dcache_mmap_lock(mapping);
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -0700772 vma_interval_tree_remove(vma, &mapping->i_mmap);
David Howells8feae132009-01-08 12:04:47 +0000773 flush_dcache_mmap_unlock(mapping);
Davidlohr Bueso83cde9e2014-12-12 16:54:21 -0800774 i_mmap_unlock_write(mapping);
David Howells8feae132009-01-08 12:04:47 +0000775 }
776
777 /* remove from the MM's tree and list */
778 rb_erase(&vma->vm_rb, &mm->mm_rb);
Namhyung Kimb951bf22011-05-24 17:11:23 -0700779
780 if (vma->vm_prev)
781 vma->vm_prev->vm_next = vma->vm_next;
782 else
783 mm->mmap = vma->vm_next;
784
785 if (vma->vm_next)
786 vma->vm_next->vm_prev = vma->vm_prev;
David Howells8feae132009-01-08 12:04:47 +0000787}
788
789/*
790 * destroy a VMA record
791 */
792static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma)
793{
David Howells8feae132009-01-08 12:04:47 +0000794 if (vma->vm_ops && vma->vm_ops->close)
795 vma->vm_ops->close(vma);
Konstantin Khlebnikove9714ac2012-10-08 16:28:54 -0700796 if (vma->vm_file)
David Howells8feae132009-01-08 12:04:47 +0000797 fput(vma->vm_file);
David Howells8feae132009-01-08 12:04:47 +0000798 put_nommu_region(vma->vm_region);
799 kmem_cache_free(vm_area_cachep, vma);
David Howells30340972006-09-27 01:50:20 -0700800}
801
802/*
803 * look up the first VMA in which addr resides, NULL if none
804 * - should be called with mm->mmap_sem at least held readlocked
805 */
806struct vm_area_struct *find_vma(struct mm_struct *mm, unsigned long addr)
807{
David Howells8feae132009-01-08 12:04:47 +0000808 struct vm_area_struct *vma;
David Howells30340972006-09-27 01:50:20 -0700809
David Howells8feae132009-01-08 12:04:47 +0000810 /* check the cache first */
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700811 vma = vmacache_find(mm, addr);
812 if (likely(vma))
David Howells8feae132009-01-08 12:04:47 +0000813 return vma;
814
Namhyung Kime922c4c2011-05-24 17:11:24 -0700815 /* trawl the list (there may be multiple mappings in which addr
David Howells8feae132009-01-08 12:04:47 +0000816 * resides) */
Namhyung Kime922c4c2011-05-24 17:11:24 -0700817 for (vma = mm->mmap; vma; vma = vma->vm_next) {
David Howells8feae132009-01-08 12:04:47 +0000818 if (vma->vm_start > addr)
819 return NULL;
820 if (vma->vm_end > addr) {
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700821 vmacache_update(addr, vma);
David Howells8feae132009-01-08 12:04:47 +0000822 return vma;
823 }
David Howells30340972006-09-27 01:50:20 -0700824 }
825
David Howells30340972006-09-27 01:50:20 -0700826 return NULL;
827}
828EXPORT_SYMBOL(find_vma);
829
830/*
David Howells930e6522006-09-27 01:50:22 -0700831 * find a VMA
832 * - we don't extend stack VMAs under NOMMU conditions
833 */
834struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr)
835{
David Howells7561e8c2010-03-25 16:48:38 +0000836 return find_vma(mm, addr);
David Howells930e6522006-09-27 01:50:22 -0700837}
838
David Howells8feae132009-01-08 12:04:47 +0000839/*
840 * expand a stack to a given address
841 * - not supported under NOMMU conditions
842 */
Greg Ungerer57c8f632007-07-15 23:38:28 -0700843int expand_stack(struct vm_area_struct *vma, unsigned long address)
844{
845 return -ENOMEM;
846}
847
David Howells930e6522006-09-27 01:50:22 -0700848/*
David Howells6fa5f802006-09-27 01:50:21 -0700849 * look up the first VMA exactly that exactly matches addr
850 * - should be called with mm->mmap_sem at least held readlocked
851 */
David Howells8feae132009-01-08 12:04:47 +0000852static struct vm_area_struct *find_vma_exact(struct mm_struct *mm,
853 unsigned long addr,
854 unsigned long len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855{
856 struct vm_area_struct *vma;
David Howells8feae132009-01-08 12:04:47 +0000857 unsigned long end = addr + len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858
David Howells8feae132009-01-08 12:04:47 +0000859 /* check the cache first */
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700860 vma = vmacache_find_exact(mm, addr, end);
861 if (vma)
David Howells8feae132009-01-08 12:04:47 +0000862 return vma;
863
Namhyung Kime922c4c2011-05-24 17:11:24 -0700864 /* trawl the list (there may be multiple mappings in which addr
David Howells8feae132009-01-08 12:04:47 +0000865 * resides) */
Namhyung Kime922c4c2011-05-24 17:11:24 -0700866 for (vma = mm->mmap; vma; vma = vma->vm_next) {
David Howells8feae132009-01-08 12:04:47 +0000867 if (vma->vm_start < addr)
868 continue;
869 if (vma->vm_start > addr)
870 return NULL;
871 if (vma->vm_end == end) {
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700872 vmacache_update(addr, vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 return vma;
David Howells8feae132009-01-08 12:04:47 +0000874 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 }
876
877 return NULL;
878}
879
David Howells30340972006-09-27 01:50:20 -0700880/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 * determine whether a mapping should be permitted and, if so, what sort of
882 * mapping we're capable of supporting
883 */
884static int validate_mmap_request(struct file *file,
885 unsigned long addr,
886 unsigned long len,
887 unsigned long prot,
888 unsigned long flags,
889 unsigned long pgoff,
890 unsigned long *_capabilities)
891{
David Howells8feae132009-01-08 12:04:47 +0000892 unsigned long capabilities, rlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 int ret;
894
895 /* do the simple checks first */
Leon Romanovsky22cc8772015-06-24 16:57:47 -0700896 if (flags & MAP_FIXED)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898
899 if ((flags & MAP_TYPE) != MAP_PRIVATE &&
900 (flags & MAP_TYPE) != MAP_SHARED)
901 return -EINVAL;
902
Mike Frysingerf81cff02006-12-06 12:02:59 +1000903 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 return -EINVAL;
905
Mike Frysingerf81cff02006-12-06 12:02:59 +1000906 /* Careful about overflows.. */
David Howells8feae132009-01-08 12:04:47 +0000907 rlen = PAGE_ALIGN(len);
908 if (!rlen || rlen > TASK_SIZE)
Mike Frysingerf81cff02006-12-06 12:02:59 +1000909 return -ENOMEM;
910
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 /* offset overflow? */
David Howells8feae132009-01-08 12:04:47 +0000912 if ((pgoff + (rlen >> PAGE_SHIFT)) < pgoff)
Mike Frysingerf81cff02006-12-06 12:02:59 +1000913 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914
915 if (file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 /* files must support mmap */
Al Viro72c2d532013-09-22 16:27:52 -0400917 if (!file->f_op->mmap)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 return -ENODEV;
919
920 /* work out if what we've got could possibly be shared
921 * - we support chardevs that provide their own "memory"
922 * - we support files/blockdevs that are memory backed
923 */
Christoph Hellwigb4caecd2015-01-14 10:42:32 +0100924 if (file->f_op->mmap_capabilities) {
925 capabilities = file->f_op->mmap_capabilities(file);
926 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 /* no explicit capabilities set, so assume some
928 * defaults */
Al Viro496ad9a2013-01-23 17:07:38 -0500929 switch (file_inode(file)->i_mode & S_IFMT) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 case S_IFREG:
931 case S_IFBLK:
Christoph Hellwigb4caecd2015-01-14 10:42:32 +0100932 capabilities = NOMMU_MAP_COPY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 break;
934
935 case S_IFCHR:
936 capabilities =
Christoph Hellwigb4caecd2015-01-14 10:42:32 +0100937 NOMMU_MAP_DIRECT |
938 NOMMU_MAP_READ |
939 NOMMU_MAP_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 break;
941
942 default:
943 return -EINVAL;
944 }
945 }
946
947 /* eliminate any capabilities that we can't support on this
948 * device */
949 if (!file->f_op->get_unmapped_area)
Christoph Hellwigb4caecd2015-01-14 10:42:32 +0100950 capabilities &= ~NOMMU_MAP_DIRECT;
Al Viro6e242a12015-03-31 12:35:13 -0400951 if (!(file->f_mode & FMODE_CAN_READ))
Christoph Hellwigb4caecd2015-01-14 10:42:32 +0100952 capabilities &= ~NOMMU_MAP_COPY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953
Graff Yang28d7a6a2009-08-18 14:11:17 -0700954 /* The file shall have been opened with read permission. */
955 if (!(file->f_mode & FMODE_READ))
956 return -EACCES;
957
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 if (flags & MAP_SHARED) {
959 /* do checks for writing, appending and locking */
960 if ((prot & PROT_WRITE) &&
961 !(file->f_mode & FMODE_WRITE))
962 return -EACCES;
963
Al Viro496ad9a2013-01-23 17:07:38 -0500964 if (IS_APPEND(file_inode(file)) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 (file->f_mode & FMODE_WRITE))
966 return -EACCES;
967
Jeff Laytond7a06982014-03-10 09:54:15 -0400968 if (locks_verify_locked(file))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 return -EAGAIN;
970
Christoph Hellwigb4caecd2015-01-14 10:42:32 +0100971 if (!(capabilities & NOMMU_MAP_DIRECT))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 return -ENODEV;
973
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 /* we mustn't privatise shared mappings */
Christoph Hellwigb4caecd2015-01-14 10:42:32 +0100975 capabilities &= ~NOMMU_MAP_COPY;
Choi Gi-yongac714902014-04-07 15:37:36 -0700976 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 /* we're going to read the file into private memory we
978 * allocate */
Christoph Hellwigb4caecd2015-01-14 10:42:32 +0100979 if (!(capabilities & NOMMU_MAP_COPY))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 return -ENODEV;
981
982 /* we don't permit a private writable mapping to be
983 * shared with the backing device */
984 if (prot & PROT_WRITE)
Christoph Hellwigb4caecd2015-01-14 10:42:32 +0100985 capabilities &= ~NOMMU_MAP_DIRECT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 }
987
Christoph Hellwigb4caecd2015-01-14 10:42:32 +0100988 if (capabilities & NOMMU_MAP_DIRECT) {
989 if (((prot & PROT_READ) && !(capabilities & NOMMU_MAP_READ)) ||
990 ((prot & PROT_WRITE) && !(capabilities & NOMMU_MAP_WRITE)) ||
991 ((prot & PROT_EXEC) && !(capabilities & NOMMU_MAP_EXEC))
Bernd Schmidt3c7b2042010-05-25 23:43:00 -0700992 ) {
Christoph Hellwigb4caecd2015-01-14 10:42:32 +0100993 capabilities &= ~NOMMU_MAP_DIRECT;
Bernd Schmidt3c7b2042010-05-25 23:43:00 -0700994 if (flags & MAP_SHARED) {
Leon Romanovsky22cc8772015-06-24 16:57:47 -0700995 pr_warn("MAP_SHARED not completely supported on !MMU\n");
Bernd Schmidt3c7b2042010-05-25 23:43:00 -0700996 return -EINVAL;
997 }
998 }
999 }
1000
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 /* handle executable mappings and implied executable
1002 * mappings */
Eric W. Biederman90f85722015-06-29 14:42:03 -05001003 if (path_noexec(&file->f_path)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 if (prot & PROT_EXEC)
1005 return -EPERM;
Choi Gi-yongac714902014-04-07 15:37:36 -07001006 } else if ((prot & PROT_READ) && !(prot & PROT_EXEC)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 /* handle implication of PROT_EXEC by PROT_READ */
1008 if (current->personality & READ_IMPLIES_EXEC) {
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001009 if (capabilities & NOMMU_MAP_EXEC)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 prot |= PROT_EXEC;
1011 }
Choi Gi-yongac714902014-04-07 15:37:36 -07001012 } else if ((prot & PROT_READ) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 (prot & PROT_EXEC) &&
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001014 !(capabilities & NOMMU_MAP_EXEC)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 ) {
1016 /* backing file is not executable, try to copy */
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001017 capabilities &= ~NOMMU_MAP_DIRECT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 }
Choi Gi-yongac714902014-04-07 15:37:36 -07001019 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 /* anonymous mappings are always memory backed and can be
1021 * privately mapped
1022 */
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001023 capabilities = NOMMU_MAP_COPY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024
1025 /* handle PROT_EXEC implication by PROT_READ */
1026 if ((prot & PROT_READ) &&
1027 (current->personality & READ_IMPLIES_EXEC))
1028 prot |= PROT_EXEC;
1029 }
1030
1031 /* allow the security API to have its say */
Al Viroe5467852012-05-30 13:30:51 -04001032 ret = security_mmap_addr(addr);
1033 if (ret < 0)
1034 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035
1036 /* looks okay */
1037 *_capabilities = capabilities;
1038 return 0;
1039}
1040
1041/*
1042 * we've determined that we can make the mapping, now translate what we
1043 * now know into VMA flags
1044 */
1045static unsigned long determine_vm_flags(struct file *file,
1046 unsigned long prot,
1047 unsigned long flags,
1048 unsigned long capabilities)
1049{
1050 unsigned long vm_flags;
1051
Dave Hansene6bfb702016-02-12 13:02:31 -08001052 vm_flags = calc_vm_prot_bits(prot, 0) | calc_vm_flag_bits(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 /* vm_flags |= mm->def_flags; */
1054
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001055 if (!(capabilities & NOMMU_MAP_DIRECT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 /* attempt to share read-only copies of mapped file chunks */
Bernd Schmidt3c7b2042010-05-25 23:43:00 -07001057 vm_flags |= VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 if (file && !(prot & PROT_WRITE))
1059 vm_flags |= VM_MAYSHARE;
Bernd Schmidt3c7b2042010-05-25 23:43:00 -07001060 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 /* overlay a shareable mapping on the backing device or inode
1062 * if possible - used for chardevs, ramfs/tmpfs/shmfs and
1063 * romfs/cramfs */
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001064 vm_flags |= VM_MAYSHARE | (capabilities & NOMMU_VMFLAGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 if (flags & MAP_SHARED)
Bernd Schmidt3c7b2042010-05-25 23:43:00 -07001066 vm_flags |= VM_SHARED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 }
1068
1069 /* refuse to let anyone share private mappings with this process if
1070 * it's being traced - otherwise breakpoints set in it may interfere
1071 * with another untraced process
1072 */
Tejun Heoa288eec2011-06-17 16:50:37 +02001073 if ((flags & MAP_PRIVATE) && current->ptrace)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 vm_flags &= ~VM_MAYSHARE;
1075
1076 return vm_flags;
1077}
1078
1079/*
David Howells8feae132009-01-08 12:04:47 +00001080 * set up a shared mapping on a file (the driver or filesystem provides and
1081 * pins the storage)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 */
David Howells8feae132009-01-08 12:04:47 +00001083static int do_mmap_shared_file(struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084{
1085 int ret;
1086
1087 ret = vma->vm_file->f_op->mmap(vma->vm_file, vma);
Paul Mundtdd8632a2009-01-08 12:04:47 +00001088 if (ret == 0) {
1089 vma->vm_region->vm_top = vma->vm_region->vm_end;
David Howells645d83c2009-09-24 15:13:10 +01001090 return 0;
Paul Mundtdd8632a2009-01-08 12:04:47 +00001091 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 if (ret != -ENOSYS)
1093 return ret;
1094
David Howells3fa30462010-03-23 13:35:21 -07001095 /* getting -ENOSYS indicates that direct mmap isn't possible (as
1096 * opposed to tried but failed) so we can only give a suitable error as
1097 * it's not possible to make a private copy if MAP_SHARED was given */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 return -ENODEV;
1099}
1100
1101/*
1102 * set up a private mapping or an anonymous shared mapping
1103 */
David Howells8feae132009-01-08 12:04:47 +00001104static int do_mmap_private(struct vm_area_struct *vma,
1105 struct vm_region *region,
David Howells645d83c2009-09-24 15:13:10 +01001106 unsigned long len,
1107 unsigned long capabilities)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108{
Joonsoo Kimdbc83582014-12-12 16:55:55 -08001109 unsigned long total, point;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 void *base;
David Howells8feae132009-01-08 12:04:47 +00001111 int ret, order;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
1113 /* invoke the file's mapping function so that it can keep track of
1114 * shared mappings on devices or memory
1115 * - VM_MAYSHARE will be set if it may attempt to share
1116 */
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001117 if (capabilities & NOMMU_MAP_DIRECT) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 ret = vma->vm_file->f_op->mmap(vma->vm_file, vma);
Paul Mundtdd8632a2009-01-08 12:04:47 +00001119 if (ret == 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 /* shouldn't return success if we're not sharing */
Paul Mundtdd8632a2009-01-08 12:04:47 +00001121 BUG_ON(!(vma->vm_flags & VM_MAYSHARE));
1122 vma->vm_region->vm_top = vma->vm_region->vm_end;
David Howells645d83c2009-09-24 15:13:10 +01001123 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 }
Paul Mundtdd8632a2009-01-08 12:04:47 +00001125 if (ret != -ENOSYS)
1126 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127
1128 /* getting an ENOSYS error indicates that direct mmap isn't
1129 * possible (as opposed to tried but failed) so we'll try to
1130 * make a private copy of the data and map that instead */
1131 }
1132
David Howells8feae132009-01-08 12:04:47 +00001133
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 /* allocate some memory to hold the mapping
1135 * - note that this may not return a page-aligned address if the object
1136 * we're allocating is smaller than a page
1137 */
Bob Liuf67d9b12011-05-24 17:12:56 -07001138 order = get_order(len);
David Howells8feae132009-01-08 12:04:47 +00001139 total = 1 << order;
Bob Liuf67d9b12011-05-24 17:12:56 -07001140 point = len >> PAGE_SHIFT;
Paul Mundtdd8632a2009-01-08 12:04:47 +00001141
Joonsoo Kimdbc83582014-12-12 16:55:55 -08001142 /* we don't want to allocate a power-of-2 sized page set */
Leon Romanovsky22cc8772015-06-24 16:57:47 -07001143 if (sysctl_nr_trim_pages && total - point >= sysctl_nr_trim_pages)
Joonsoo Kimdbc83582014-12-12 16:55:55 -08001144 total = point;
David Howells8feae132009-01-08 12:04:47 +00001145
Joonsoo Kimda616532015-02-27 15:51:43 -08001146 base = alloc_pages_exact(total << PAGE_SHIFT, GFP_KERNEL);
Joonsoo Kimdbc83582014-12-12 16:55:55 -08001147 if (!base)
1148 goto enomem;
David Howells8feae132009-01-08 12:04:47 +00001149
Joonsoo Kimdbc83582014-12-12 16:55:55 -08001150 atomic_long_add(total, &mmap_pages_allocated);
1151
David Howells8feae132009-01-08 12:04:47 +00001152 region->vm_flags = vma->vm_flags |= VM_MAPPED_COPY;
1153 region->vm_start = (unsigned long) base;
Bob Liuf67d9b12011-05-24 17:12:56 -07001154 region->vm_end = region->vm_start + len;
Paul Mundtdd8632a2009-01-08 12:04:47 +00001155 region->vm_top = region->vm_start + (total << PAGE_SHIFT);
David Howells8feae132009-01-08 12:04:47 +00001156
1157 vma->vm_start = region->vm_start;
1158 vma->vm_end = region->vm_start + len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159
1160 if (vma->vm_file) {
1161 /* read the contents of a file into the copy */
1162 mm_segment_t old_fs;
1163 loff_t fpos;
1164
1165 fpos = vma->vm_pgoff;
1166 fpos <<= PAGE_SHIFT;
1167
1168 old_fs = get_fs();
1169 set_fs(KERNEL_DS);
Al Viro6e242a12015-03-31 12:35:13 -04001170 ret = __vfs_read(vma->vm_file, base, len, &fpos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 set_fs(old_fs);
1172
1173 if (ret < 0)
1174 goto error_free;
1175
1176 /* clear the last little bit */
Bob Liuf67d9b12011-05-24 17:12:56 -07001177 if (ret < len)
1178 memset(base + ret, 0, len - ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 }
1181
1182 return 0;
1183
1184error_free:
Namhyung Kim7223bb42011-05-24 17:11:26 -07001185 free_page_series(region->vm_start, region->vm_top);
David Howells8feae132009-01-08 12:04:47 +00001186 region->vm_start = vma->vm_start = 0;
1187 region->vm_end = vma->vm_end = 0;
Paul Mundtdd8632a2009-01-08 12:04:47 +00001188 region->vm_top = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 return ret;
1190
1191enomem:
Mitchel Humpherysb1de0d12014-06-06 14:38:30 -07001192 pr_err("Allocation of length %lu from process %d (%s) failed\n",
Greg Ungerer05ae6fa2009-01-13 17:30:22 +10001193 len, current->pid, current->comm);
David Rientjes7bf02ea2011-05-24 17:11:16 -07001194 show_free_areas(0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195 return -ENOMEM;
1196}
1197
1198/*
1199 * handle mapping creation for uClinux
1200 */
Oleg Nesterov1fcfd8d2015-09-09 15:39:29 -07001201unsigned long do_mmap(struct file *file,
1202 unsigned long addr,
1203 unsigned long len,
1204 unsigned long prot,
1205 unsigned long flags,
1206 vm_flags_t vm_flags,
1207 unsigned long pgoff,
1208 unsigned long *populate)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209{
David Howells8feae132009-01-08 12:04:47 +00001210 struct vm_area_struct *vma;
1211 struct vm_region *region;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 struct rb_node *rb;
Oleg Nesterov1fcfd8d2015-09-09 15:39:29 -07001213 unsigned long capabilities, result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 int ret;
1215
Michel Lespinasse41badc12013-02-22 16:32:47 -08001216 *populate = 0;
Michel Lespinassebebeb3d2013-02-22 16:32:37 -08001217
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 /* decide whether we should attempt the mapping, and if so what sort of
1219 * mapping */
1220 ret = validate_mmap_request(file, addr, len, prot, flags, pgoff,
1221 &capabilities);
Leon Romanovsky22cc8772015-06-24 16:57:47 -07001222 if (ret < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223 return ret;
1224
David Howells06aab5a2009-09-24 12:33:48 +01001225 /* we ignore the address hint */
1226 addr = 0;
Bob Liuf67d9b12011-05-24 17:12:56 -07001227 len = PAGE_ALIGN(len);
David Howells06aab5a2009-09-24 12:33:48 +01001228
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 /* we've determined that we can make the mapping, now translate what we
1230 * now know into VMA flags */
Oleg Nesterov1fcfd8d2015-09-09 15:39:29 -07001231 vm_flags |= determine_vm_flags(file, prot, flags, capabilities);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232
David Howells8feae132009-01-08 12:04:47 +00001233 /* we're going to need to record the mapping */
1234 region = kmem_cache_zalloc(vm_region_jar, GFP_KERNEL);
1235 if (!region)
1236 goto error_getting_region;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237
David Howells8feae132009-01-08 12:04:47 +00001238 vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
1239 if (!vma)
1240 goto error_getting_vma;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241
David Howells1e2ae592010-01-15 17:01:33 -08001242 region->vm_usage = 1;
David Howells8feae132009-01-08 12:04:47 +00001243 region->vm_flags = vm_flags;
1244 region->vm_pgoff = pgoff;
1245
Rik van Riel5beb4932010-03-05 13:42:07 -08001246 INIT_LIST_HEAD(&vma->anon_vma_chain);
David Howells8feae132009-01-08 12:04:47 +00001247 vma->vm_flags = vm_flags;
1248 vma->vm_pgoff = pgoff;
1249
1250 if (file) {
Al Virocb0942b2012-08-27 14:48:26 -04001251 region->vm_file = get_file(file);
1252 vma->vm_file = get_file(file);
David Howells8feae132009-01-08 12:04:47 +00001253 }
1254
1255 down_write(&nommu_region_sem);
1256
1257 /* if we want to share, we need to check for regions created by other
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 * mmap() calls that overlap with our proposed mapping
David Howells8feae132009-01-08 12:04:47 +00001259 * - we can only share with a superset match on most regular files
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 * - shared mappings on character devices and memory backed files are
1261 * permitted to overlap inexactly as far as we are concerned for in
1262 * these cases, sharing is handled in the driver or filesystem rather
1263 * than here
1264 */
1265 if (vm_flags & VM_MAYSHARE) {
David Howells8feae132009-01-08 12:04:47 +00001266 struct vm_region *pregion;
1267 unsigned long pglen, rpglen, pgend, rpgend, start;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268
David Howells8feae132009-01-08 12:04:47 +00001269 pglen = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1270 pgend = pgoff + pglen;
David Howells165b2392007-03-22 00:11:24 -08001271
David Howells8feae132009-01-08 12:04:47 +00001272 for (rb = rb_first(&nommu_region_tree); rb; rb = rb_next(rb)) {
1273 pregion = rb_entry(rb, struct vm_region, vm_rb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274
David Howells8feae132009-01-08 12:04:47 +00001275 if (!(pregion->vm_flags & VM_MAYSHARE))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 continue;
1277
1278 /* search for overlapping mappings on the same file */
Al Viro496ad9a2013-01-23 17:07:38 -05001279 if (file_inode(pregion->vm_file) !=
1280 file_inode(file))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 continue;
1282
David Howells8feae132009-01-08 12:04:47 +00001283 if (pregion->vm_pgoff >= pgend)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 continue;
1285
David Howells8feae132009-01-08 12:04:47 +00001286 rpglen = pregion->vm_end - pregion->vm_start;
1287 rpglen = (rpglen + PAGE_SIZE - 1) >> PAGE_SHIFT;
1288 rpgend = pregion->vm_pgoff + rpglen;
1289 if (pgoff >= rpgend)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290 continue;
1291
David Howells8feae132009-01-08 12:04:47 +00001292 /* handle inexactly overlapping matches between
1293 * mappings */
1294 if ((pregion->vm_pgoff != pgoff || rpglen != pglen) &&
1295 !(pgoff >= pregion->vm_pgoff && pgend <= rpgend)) {
1296 /* new mapping is not a subset of the region */
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001297 if (!(capabilities & NOMMU_MAP_DIRECT))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 goto sharing_violation;
1299 continue;
1300 }
1301
David Howells8feae132009-01-08 12:04:47 +00001302 /* we've found a region we can share */
David Howells1e2ae592010-01-15 17:01:33 -08001303 pregion->vm_usage++;
David Howells8feae132009-01-08 12:04:47 +00001304 vma->vm_region = pregion;
1305 start = pregion->vm_start;
1306 start += (pgoff - pregion->vm_pgoff) << PAGE_SHIFT;
1307 vma->vm_start = start;
1308 vma->vm_end = start + len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309
Leon Romanovsky22cc8772015-06-24 16:57:47 -07001310 if (pregion->vm_flags & VM_MAPPED_COPY)
David Howells8feae132009-01-08 12:04:47 +00001311 vma->vm_flags |= VM_MAPPED_COPY;
Leon Romanovsky22cc8772015-06-24 16:57:47 -07001312 else {
David Howells8feae132009-01-08 12:04:47 +00001313 ret = do_mmap_shared_file(vma);
1314 if (ret < 0) {
1315 vma->vm_region = NULL;
1316 vma->vm_start = 0;
1317 vma->vm_end = 0;
David Howells1e2ae592010-01-15 17:01:33 -08001318 pregion->vm_usage--;
David Howells8feae132009-01-08 12:04:47 +00001319 pregion = NULL;
1320 goto error_just_free;
1321 }
1322 }
1323 fput(region->vm_file);
1324 kmem_cache_free(vm_region_jar, region);
1325 region = pregion;
1326 result = start;
1327 goto share;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 }
1329
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 /* obtain the address at which to make a shared mapping
1331 * - this is the hook for quasi-memory character devices to
1332 * tell us the location of a shared mapping
1333 */
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001334 if (capabilities & NOMMU_MAP_DIRECT) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 addr = file->f_op->get_unmapped_area(file, addr, len,
1336 pgoff, flags);
Namhyung Kimbb005a52011-05-24 17:11:27 -07001337 if (IS_ERR_VALUE(addr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 ret = addr;
Namhyung Kimbb005a52011-05-24 17:11:27 -07001339 if (ret != -ENOSYS)
David Howells8feae132009-01-08 12:04:47 +00001340 goto error_just_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341
1342 /* the driver refused to tell us where to site
1343 * the mapping so we'll have to attempt to copy
1344 * it */
Namhyung Kimbb005a52011-05-24 17:11:27 -07001345 ret = -ENODEV;
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001346 if (!(capabilities & NOMMU_MAP_COPY))
David Howells8feae132009-01-08 12:04:47 +00001347 goto error_just_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001349 capabilities &= ~NOMMU_MAP_DIRECT;
David Howells8feae132009-01-08 12:04:47 +00001350 } else {
1351 vma->vm_start = region->vm_start = addr;
1352 vma->vm_end = region->vm_end = addr + len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 }
1354 }
1355 }
1356
David Howells8feae132009-01-08 12:04:47 +00001357 vma->vm_region = region;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358
David Howells645d83c2009-09-24 15:13:10 +01001359 /* set up the mapping
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001360 * - the region is filled in if NOMMU_MAP_DIRECT is still set
David Howells645d83c2009-09-24 15:13:10 +01001361 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 if (file && vma->vm_flags & VM_SHARED)
David Howells8feae132009-01-08 12:04:47 +00001363 ret = do_mmap_shared_file(vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 else
David Howells645d83c2009-09-24 15:13:10 +01001365 ret = do_mmap_private(vma, region, len, capabilities);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366 if (ret < 0)
David Howells645d83c2009-09-24 15:13:10 +01001367 goto error_just_free;
1368 add_nommu_region(region);
David Howells8feae132009-01-08 12:04:47 +00001369
Jie Zhangea637632009-12-14 18:00:02 -08001370 /* clear anonymous mappings that don't ask for uninitialized data */
1371 if (!vma->vm_file && !(flags & MAP_UNINITIALIZED))
1372 memset((void *)region->vm_start, 0,
1373 region->vm_end - region->vm_start);
1374
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 /* okay... we have a mapping; now we have to register it */
David Howells8feae132009-01-08 12:04:47 +00001376 result = vma->vm_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 current->mm->total_vm += len >> PAGE_SHIFT;
1379
David Howells8feae132009-01-08 12:04:47 +00001380share:
1381 add_vma_to_mm(current->mm, vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382
Mike Frysingercfe79c02010-01-06 17:23:23 +00001383 /* we flush the region from the icache only when the first executable
1384 * mapping of it is made */
1385 if (vma->vm_flags & VM_EXEC && !region->vm_icache_flushed) {
1386 flush_icache_range(region->vm_start, region->vm_end);
1387 region->vm_icache_flushed = true;
1388 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389
Mike Frysingercfe79c02010-01-06 17:23:23 +00001390 up_write(&nommu_region_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391
David Howells8feae132009-01-08 12:04:47 +00001392 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393
David Howells8feae132009-01-08 12:04:47 +00001394error_just_free:
1395 up_write(&nommu_region_sem);
1396error:
David Howells89a86402009-10-30 13:13:26 +00001397 if (region->vm_file)
1398 fput(region->vm_file);
David Howells8feae132009-01-08 12:04:47 +00001399 kmem_cache_free(vm_region_jar, region);
David Howells89a86402009-10-30 13:13:26 +00001400 if (vma->vm_file)
1401 fput(vma->vm_file);
David Howells8feae132009-01-08 12:04:47 +00001402 kmem_cache_free(vm_area_cachep, vma);
David Howells8feae132009-01-08 12:04:47 +00001403 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404
David Howells8feae132009-01-08 12:04:47 +00001405sharing_violation:
1406 up_write(&nommu_region_sem);
Leon Romanovsky22cc8772015-06-24 16:57:47 -07001407 pr_warn("Attempt to share mismatched mappings\n");
David Howells8feae132009-01-08 12:04:47 +00001408 ret = -EINVAL;
1409 goto error;
1410
1411error_getting_vma:
1412 kmem_cache_free(vm_region_jar, region);
Leon Romanovsky22cc8772015-06-24 16:57:47 -07001413 pr_warn("Allocation of vma for %lu byte allocation from process %d failed\n",
1414 len, current->pid);
David Rientjes7bf02ea2011-05-24 17:11:16 -07001415 show_free_areas(0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416 return -ENOMEM;
1417
David Howells8feae132009-01-08 12:04:47 +00001418error_getting_region:
Leon Romanovsky22cc8772015-06-24 16:57:47 -07001419 pr_warn("Allocation of vm region for %lu byte allocation from process %d failed\n",
1420 len, current->pid);
David Rientjes7bf02ea2011-05-24 17:11:16 -07001421 show_free_areas(0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 return -ENOMEM;
1423}
Linus Torvalds6be5ceb2012-04-20 17:13:58 -07001424
Hugh Dickins66f0dc42009-12-30 20:17:34 +00001425SYSCALL_DEFINE6(mmap_pgoff, unsigned long, addr, unsigned long, len,
1426 unsigned long, prot, unsigned long, flags,
1427 unsigned long, fd, unsigned long, pgoff)
1428{
1429 struct file *file = NULL;
1430 unsigned long retval = -EBADF;
1431
Al Viro120a7952010-10-30 02:54:44 -04001432 audit_mmap_fd(fd, flags);
Hugh Dickins66f0dc42009-12-30 20:17:34 +00001433 if (!(flags & MAP_ANONYMOUS)) {
1434 file = fget(fd);
1435 if (!file)
1436 goto out;
1437 }
1438
1439 flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);
1440
Greg Ungererad1ed292012-06-04 14:29:59 +10001441 retval = vm_mmap_pgoff(file, addr, len, prot, flags, pgoff);
Hugh Dickins66f0dc42009-12-30 20:17:34 +00001442
1443 if (file)
1444 fput(file);
1445out:
1446 return retval;
1447}
1448
Christoph Hellwiga4679372010-03-10 15:21:15 -08001449#ifdef __ARCH_WANT_SYS_OLD_MMAP
1450struct mmap_arg_struct {
1451 unsigned long addr;
1452 unsigned long len;
1453 unsigned long prot;
1454 unsigned long flags;
1455 unsigned long fd;
1456 unsigned long offset;
1457};
1458
1459SYSCALL_DEFINE1(old_mmap, struct mmap_arg_struct __user *, arg)
1460{
1461 struct mmap_arg_struct a;
1462
1463 if (copy_from_user(&a, arg, sizeof(a)))
1464 return -EFAULT;
Alexander Kuleshov1824cb72015-11-05 18:46:35 -08001465 if (offset_in_page(a.offset))
Christoph Hellwiga4679372010-03-10 15:21:15 -08001466 return -EINVAL;
1467
1468 return sys_mmap_pgoff(a.addr, a.len, a.prot, a.flags, a.fd,
1469 a.offset >> PAGE_SHIFT);
1470}
1471#endif /* __ARCH_WANT_SYS_OLD_MMAP */
1472
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473/*
David Howells8feae132009-01-08 12:04:47 +00001474 * split a vma into two pieces at address 'addr', a new vma is allocated either
1475 * for the first part or the tail.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 */
David Howells8feae132009-01-08 12:04:47 +00001477int split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
1478 unsigned long addr, int new_below)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479{
David Howells8feae132009-01-08 12:04:47 +00001480 struct vm_area_struct *new;
1481 struct vm_region *region;
1482 unsigned long npages;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483
David Howells779c1022010-01-15 17:01:34 -08001484 /* we're only permitted to split anonymous regions (these should have
1485 * only a single usage on the region) */
1486 if (vma->vm_file)
David Howells8feae132009-01-08 12:04:47 +00001487 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488
David Howells8feae132009-01-08 12:04:47 +00001489 if (mm->map_count >= sysctl_max_map_count)
1490 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491
David Howells8feae132009-01-08 12:04:47 +00001492 region = kmem_cache_alloc(vm_region_jar, GFP_KERNEL);
1493 if (!region)
1494 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495
David Howells8feae132009-01-08 12:04:47 +00001496 new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
1497 if (!new) {
1498 kmem_cache_free(vm_region_jar, region);
1499 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 }
David Howells8feae132009-01-08 12:04:47 +00001501
1502 /* most fields are the same, copy all, and then fixup */
1503 *new = *vma;
1504 *region = *vma->vm_region;
1505 new->vm_region = region;
1506
1507 npages = (addr - vma->vm_start) >> PAGE_SHIFT;
1508
1509 if (new_below) {
Paul Mundtdd8632a2009-01-08 12:04:47 +00001510 region->vm_top = region->vm_end = new->vm_end = addr;
David Howells8feae132009-01-08 12:04:47 +00001511 } else {
1512 region->vm_start = new->vm_start = addr;
1513 region->vm_pgoff = new->vm_pgoff += npages;
1514 }
1515
1516 if (new->vm_ops && new->vm_ops->open)
1517 new->vm_ops->open(new);
1518
1519 delete_vma_from_mm(vma);
1520 down_write(&nommu_region_sem);
1521 delete_nommu_region(vma->vm_region);
1522 if (new_below) {
1523 vma->vm_region->vm_start = vma->vm_start = addr;
1524 vma->vm_region->vm_pgoff = vma->vm_pgoff += npages;
1525 } else {
1526 vma->vm_region->vm_end = vma->vm_end = addr;
Paul Mundtdd8632a2009-01-08 12:04:47 +00001527 vma->vm_region->vm_top = addr;
David Howells8feae132009-01-08 12:04:47 +00001528 }
1529 add_nommu_region(vma->vm_region);
1530 add_nommu_region(new->vm_region);
1531 up_write(&nommu_region_sem);
1532 add_vma_to_mm(mm, vma);
1533 add_vma_to_mm(mm, new);
1534 return 0;
1535}
1536
1537/*
1538 * shrink a VMA by removing the specified chunk from either the beginning or
1539 * the end
1540 */
1541static int shrink_vma(struct mm_struct *mm,
1542 struct vm_area_struct *vma,
1543 unsigned long from, unsigned long to)
1544{
1545 struct vm_region *region;
1546
David Howells8feae132009-01-08 12:04:47 +00001547 /* adjust the VMA's pointers, which may reposition it in the MM's tree
1548 * and list */
1549 delete_vma_from_mm(vma);
1550 if (from > vma->vm_start)
1551 vma->vm_end = from;
1552 else
1553 vma->vm_start = to;
1554 add_vma_to_mm(mm, vma);
1555
1556 /* cut the backing region down to size */
1557 region = vma->vm_region;
David Howells1e2ae592010-01-15 17:01:33 -08001558 BUG_ON(region->vm_usage != 1);
David Howells8feae132009-01-08 12:04:47 +00001559
1560 down_write(&nommu_region_sem);
1561 delete_nommu_region(region);
Paul Mundtdd8632a2009-01-08 12:04:47 +00001562 if (from > region->vm_start) {
1563 to = region->vm_top;
1564 region->vm_top = region->vm_end = from;
1565 } else {
David Howells8feae132009-01-08 12:04:47 +00001566 region->vm_start = to;
Paul Mundtdd8632a2009-01-08 12:04:47 +00001567 }
David Howells8feae132009-01-08 12:04:47 +00001568 add_nommu_region(region);
1569 up_write(&nommu_region_sem);
1570
1571 free_page_series(from, to);
1572 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573}
1574
David Howells30340972006-09-27 01:50:20 -07001575/*
1576 * release a mapping
David Howells8feae132009-01-08 12:04:47 +00001577 * - under NOMMU conditions the chunk to be unmapped must be backed by a single
1578 * VMA, though it need not cover the whole VMA
David Howells30340972006-09-27 01:50:20 -07001579 */
David Howells8feae132009-01-08 12:04:47 +00001580int do_munmap(struct mm_struct *mm, unsigned long start, size_t len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581{
David Howells8feae132009-01-08 12:04:47 +00001582 struct vm_area_struct *vma;
Bob Liuf67d9b12011-05-24 17:12:56 -07001583 unsigned long end;
David Howells8feae132009-01-08 12:04:47 +00001584 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585
Bob Liuf67d9b12011-05-24 17:12:56 -07001586 len = PAGE_ALIGN(len);
David Howells8feae132009-01-08 12:04:47 +00001587 if (len == 0)
1588 return -EINVAL;
1589
Bob Liuf67d9b12011-05-24 17:12:56 -07001590 end = start + len;
1591
David Howells8feae132009-01-08 12:04:47 +00001592 /* find the first potentially overlapping VMA */
1593 vma = find_vma(mm, start);
1594 if (!vma) {
Choi Gi-yongac714902014-04-07 15:37:36 -07001595 static int limit;
David Howells33e5d7692009-04-02 16:56:32 -07001596 if (limit < 5) {
Leon Romanovsky22cc8772015-06-24 16:57:47 -07001597 pr_warn("munmap of memory not mmapped by process %d (%s): 0x%lx-0x%lx\n",
1598 current->pid, current->comm,
1599 start, start + len - 1);
David Howells33e5d7692009-04-02 16:56:32 -07001600 limit++;
1601 }
David Howells8feae132009-01-08 12:04:47 +00001602 return -EINVAL;
David Howells30340972006-09-27 01:50:20 -07001603 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604
David Howells8feae132009-01-08 12:04:47 +00001605 /* we're allowed to split an anonymous VMA but not a file-backed one */
1606 if (vma->vm_file) {
1607 do {
Leon Romanovsky22cc8772015-06-24 16:57:47 -07001608 if (start > vma->vm_start)
David Howells8feae132009-01-08 12:04:47 +00001609 return -EINVAL;
David Howells8feae132009-01-08 12:04:47 +00001610 if (end == vma->vm_end)
1611 goto erase_whole_vma;
Namhyung Kimd75a3102011-05-24 17:11:25 -07001612 vma = vma->vm_next;
1613 } while (vma);
David Howells8feae132009-01-08 12:04:47 +00001614 return -EINVAL;
1615 } else {
1616 /* the chunk must be a subset of the VMA found */
1617 if (start == vma->vm_start && end == vma->vm_end)
1618 goto erase_whole_vma;
Leon Romanovsky22cc8772015-06-24 16:57:47 -07001619 if (start < vma->vm_start || end > vma->vm_end)
David Howells8feae132009-01-08 12:04:47 +00001620 return -EINVAL;
Alexander Kuleshov1824cb72015-11-05 18:46:35 -08001621 if (offset_in_page(start))
David Howells8feae132009-01-08 12:04:47 +00001622 return -EINVAL;
Alexander Kuleshov1824cb72015-11-05 18:46:35 -08001623 if (end != vma->vm_end && offset_in_page(end))
David Howells8feae132009-01-08 12:04:47 +00001624 return -EINVAL;
David Howells8feae132009-01-08 12:04:47 +00001625 if (start != vma->vm_start && end != vma->vm_end) {
1626 ret = split_vma(mm, vma, start, 1);
Leon Romanovsky22cc8772015-06-24 16:57:47 -07001627 if (ret < 0)
David Howells8feae132009-01-08 12:04:47 +00001628 return ret;
David Howells8feae132009-01-08 12:04:47 +00001629 }
1630 return shrink_vma(mm, vma, start, end);
1631 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632
David Howells8feae132009-01-08 12:04:47 +00001633erase_whole_vma:
1634 delete_vma_from_mm(vma);
1635 delete_vma(mm, vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 return 0;
1637}
Paul Mundtb5073172007-07-21 04:37:25 -07001638EXPORT_SYMBOL(do_munmap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639
Al Virobfce2812012-04-20 21:57:04 -04001640int vm_munmap(unsigned long addr, size_t len)
David Howells30340972006-09-27 01:50:20 -07001641{
Al Virobfce2812012-04-20 21:57:04 -04001642 struct mm_struct *mm = current->mm;
David Howells30340972006-09-27 01:50:20 -07001643 int ret;
David Howells30340972006-09-27 01:50:20 -07001644
1645 down_write(&mm->mmap_sem);
1646 ret = do_munmap(mm, addr, len);
1647 up_write(&mm->mmap_sem);
1648 return ret;
1649}
Linus Torvaldsa46ef992012-04-20 16:20:01 -07001650EXPORT_SYMBOL(vm_munmap);
1651
1652SYSCALL_DEFINE2(munmap, unsigned long, addr, size_t, len)
1653{
Al Virobfce2812012-04-20 21:57:04 -04001654 return vm_munmap(addr, len);
Linus Torvaldsa46ef992012-04-20 16:20:01 -07001655}
David Howells30340972006-09-27 01:50:20 -07001656
1657/*
David Howells8feae132009-01-08 12:04:47 +00001658 * release all the mappings made in a process's VM space
David Howells30340972006-09-27 01:50:20 -07001659 */
David Howells8feae132009-01-08 12:04:47 +00001660void exit_mmap(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661{
David Howells8feae132009-01-08 12:04:47 +00001662 struct vm_area_struct *vma;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663
David Howells8feae132009-01-08 12:04:47 +00001664 if (!mm)
1665 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666
David Howells8feae132009-01-08 12:04:47 +00001667 mm->total_vm = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668
David Howells8feae132009-01-08 12:04:47 +00001669 while ((vma = mm->mmap)) {
1670 mm->mmap = vma->vm_next;
1671 delete_vma_from_mm(vma);
1672 delete_vma(mm, vma);
Steven J. Magnani04c34962010-11-24 12:56:54 -08001673 cond_resched();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 }
1675}
1676
Linus Torvalds5d22fc22016-05-27 15:57:31 -07001677int vm_brk(unsigned long addr, unsigned long len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678{
1679 return -ENOMEM;
1680}
1681
1682/*
David Howells6fa5f802006-09-27 01:50:21 -07001683 * expand (or shrink) an existing mapping, potentially moving it at the same
1684 * time (controlled by the MREMAP_MAYMOVE flag and available VM space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 *
David Howells6fa5f802006-09-27 01:50:21 -07001686 * under NOMMU conditions, we only permit changing a mapping's size, and only
David Howells8feae132009-01-08 12:04:47 +00001687 * as long as it stays within the region allocated by do_mmap_private() and the
1688 * block is not shareable
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 *
David Howells6fa5f802006-09-27 01:50:21 -07001690 * MREMAP_FIXED is not supported under NOMMU conditions
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 */
Al Viro4b377ba2013-03-04 10:47:59 -05001692static unsigned long do_mremap(unsigned long addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 unsigned long old_len, unsigned long new_len,
1694 unsigned long flags, unsigned long new_addr)
1695{
David Howells6fa5f802006-09-27 01:50:21 -07001696 struct vm_area_struct *vma;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697
1698 /* insanity checks first */
Bob Liuf67d9b12011-05-24 17:12:56 -07001699 old_len = PAGE_ALIGN(old_len);
1700 new_len = PAGE_ALIGN(new_len);
David Howells8feae132009-01-08 12:04:47 +00001701 if (old_len == 0 || new_len == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 return (unsigned long) -EINVAL;
1703
Alexander Kuleshov1824cb72015-11-05 18:46:35 -08001704 if (offset_in_page(addr))
David Howells8feae132009-01-08 12:04:47 +00001705 return -EINVAL;
1706
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 if (flags & MREMAP_FIXED && new_addr != addr)
1708 return (unsigned long) -EINVAL;
1709
David Howells8feae132009-01-08 12:04:47 +00001710 vma = find_vma_exact(current->mm, addr, old_len);
David Howells6fa5f802006-09-27 01:50:21 -07001711 if (!vma)
1712 return (unsigned long) -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713
David Howells6fa5f802006-09-27 01:50:21 -07001714 if (vma->vm_end != vma->vm_start + old_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 return (unsigned long) -EFAULT;
1716
David Howells6fa5f802006-09-27 01:50:21 -07001717 if (vma->vm_flags & VM_MAYSHARE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 return (unsigned long) -EPERM;
1719
David Howells8feae132009-01-08 12:04:47 +00001720 if (new_len > vma->vm_region->vm_end - vma->vm_region->vm_start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721 return (unsigned long) -ENOMEM;
1722
1723 /* all checks complete - do it */
David Howells6fa5f802006-09-27 01:50:21 -07001724 vma->vm_end = vma->vm_start + new_len;
David Howells6fa5f802006-09-27 01:50:21 -07001725 return vma->vm_start;
1726}
1727
Heiko Carstens6a6160a2009-01-14 14:14:15 +01001728SYSCALL_DEFINE5(mremap, unsigned long, addr, unsigned long, old_len,
1729 unsigned long, new_len, unsigned long, flags,
1730 unsigned long, new_addr)
David Howells6fa5f802006-09-27 01:50:21 -07001731{
1732 unsigned long ret;
1733
1734 down_write(&current->mm->mmap_sem);
1735 ret = do_mremap(addr, old_len, new_len, flags, new_addr);
1736 up_write(&current->mm->mmap_sem);
1737 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738}
1739
Michel Lespinasse240aade2013-02-22 16:35:56 -08001740struct page *follow_page_mask(struct vm_area_struct *vma,
1741 unsigned long address, unsigned int flags,
1742 unsigned int *page_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743{
Michel Lespinasse240aade2013-02-22 16:35:56 -08001744 *page_mask = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 return NULL;
1746}
1747
Bob Liu8f3b1322011-07-08 15:39:46 -07001748int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
1749 unsigned long pfn, unsigned long size, pgprot_t prot)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750{
Bob Liu8f3b1322011-07-08 15:39:46 -07001751 if (addr != (pfn << PAGE_SHIFT))
1752 return -EINVAL;
1753
Konstantin Khlebnikov314e51b2012-10-08 16:29:02 -07001754 vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
Greg Ungerer66aa2b42005-09-12 11:18:10 +10001755 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756}
Luke Yang22c4af42006-07-14 00:24:09 -07001757EXPORT_SYMBOL(remap_pfn_range);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758
Linus Torvalds3c0b9de2013-04-27 13:25:38 -07001759int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len)
1760{
1761 unsigned long pfn = start >> PAGE_SHIFT;
1762 unsigned long vm_len = vma->vm_end - vma->vm_start;
1763
1764 pfn += vma->vm_pgoff;
1765 return io_remap_pfn_range(vma, vma->vm_start, pfn, vm_len, vma->vm_page_prot);
1766}
1767EXPORT_SYMBOL(vm_iomap_memory);
1768
Paul Mundtf905bc42008-02-04 22:29:59 -08001769int remap_vmalloc_range(struct vm_area_struct *vma, void *addr,
1770 unsigned long pgoff)
1771{
1772 unsigned int size = vma->vm_end - vma->vm_start;
1773
1774 if (!(vma->vm_flags & VM_USERMAP))
1775 return -EINVAL;
1776
1777 vma->vm_start = (unsigned long)(addr + (pgoff << PAGE_SHIFT));
1778 vma->vm_end = vma->vm_start + size;
1779
1780 return 0;
1781}
1782EXPORT_SYMBOL(remap_vmalloc_range);
1783
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784unsigned long arch_get_unmapped_area(struct file *file, unsigned long addr,
1785 unsigned long len, unsigned long pgoff, unsigned long flags)
1786{
1787 return -ENOMEM;
1788}
1789
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790void unmap_mapping_range(struct address_space *mapping,
1791 loff_t const holebegin, loff_t const holelen,
1792 int even_cows)
1793{
1794}
Luke Yang22c4af42006-07-14 00:24:09 -07001795EXPORT_SYMBOL(unmap_mapping_range);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796
Nick Piggind0217ac2007-07-19 01:47:03 -07001797int filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
David Howellsb0e15192006-01-06 00:11:42 -08001798{
1799 BUG();
Nick Piggind0217ac2007-07-19 01:47:03 -07001800 return 0;
David Howellsb0e15192006-01-06 00:11:42 -08001801}
Paul Mundtb5073172007-07-21 04:37:25 -07001802EXPORT_SYMBOL(filemap_fault);
David Howells0ec76a12006-09-27 01:50:15 -07001803
Kirill A. Shutemovbae473a2016-07-26 15:25:20 -07001804void filemap_map_pages(struct fault_env *fe,
1805 pgoff_t start_pgoff, pgoff_t end_pgoff)
Kirill A. Shutemovf1820362014-04-07 15:37:19 -07001806{
1807 BUG();
1808}
1809EXPORT_SYMBOL(filemap_map_pages);
1810
Mike Frysingerf55f1992011-03-29 14:05:12 +01001811static int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
Lorenzo Stoakes442486e2016-10-13 01:20:18 +01001812 unsigned long addr, void *buf, int len, unsigned int gup_flags)
David Howells0ec76a12006-09-27 01:50:15 -07001813{
David Howells0ec76a12006-09-27 01:50:15 -07001814 struct vm_area_struct *vma;
Lorenzo Stoakes442486e2016-10-13 01:20:18 +01001815 int write = gup_flags & FOLL_WRITE;
David Howells0ec76a12006-09-27 01:50:15 -07001816
1817 down_read(&mm->mmap_sem);
1818
1819 /* the access must start within one of the target process's mappings */
David Howells0159b142006-09-27 01:50:16 -07001820 vma = find_vma(mm, addr);
1821 if (vma) {
David Howells0ec76a12006-09-27 01:50:15 -07001822 /* don't overrun this mapping */
1823 if (addr + len >= vma->vm_end)
1824 len = vma->vm_end - addr;
1825
1826 /* only read or write mappings where it is permitted */
David Howellsd00c7b992006-09-27 01:50:19 -07001827 if (write && vma->vm_flags & VM_MAYWRITE)
Jie Zhang79597222010-01-06 17:23:28 +00001828 copy_to_user_page(vma, NULL, addr,
1829 (void *) addr, buf, len);
David Howellsd00c7b992006-09-27 01:50:19 -07001830 else if (!write && vma->vm_flags & VM_MAYREAD)
Jie Zhang79597222010-01-06 17:23:28 +00001831 copy_from_user_page(vma, NULL, addr,
1832 buf, (void *) addr, len);
David Howells0ec76a12006-09-27 01:50:15 -07001833 else
1834 len = 0;
1835 } else {
1836 len = 0;
1837 }
1838
1839 up_read(&mm->mmap_sem);
Mike Frysingerf55f1992011-03-29 14:05:12 +01001840
1841 return len;
1842}
1843
1844/**
1845 * @access_remote_vm - access another process' address space
1846 * @mm: the mm_struct of the target address space
1847 * @addr: start address to access
1848 * @buf: source or destination buffer
1849 * @len: number of bytes to transfer
Lorenzo Stoakes6347e8d2016-10-13 01:20:19 +01001850 * @gup_flags: flags modifying lookup behaviour
Mike Frysingerf55f1992011-03-29 14:05:12 +01001851 *
1852 * The caller must hold a reference on @mm.
1853 */
1854int access_remote_vm(struct mm_struct *mm, unsigned long addr,
Lorenzo Stoakes6347e8d2016-10-13 01:20:19 +01001855 void *buf, int len, unsigned int gup_flags)
Mike Frysingerf55f1992011-03-29 14:05:12 +01001856{
Lorenzo Stoakes6347e8d2016-10-13 01:20:19 +01001857 return __access_remote_vm(NULL, mm, addr, buf, len, gup_flags);
Mike Frysingerf55f1992011-03-29 14:05:12 +01001858}
1859
1860/*
1861 * Access another process' address space.
1862 * - source/target buffer must be kernel space
1863 */
Lorenzo Stoakesf307ab62016-10-13 01:20:20 +01001864int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1865 unsigned int gup_flags)
Mike Frysingerf55f1992011-03-29 14:05:12 +01001866{
1867 struct mm_struct *mm;
1868
1869 if (addr + len < addr)
1870 return 0;
1871
1872 mm = get_task_mm(tsk);
1873 if (!mm)
1874 return 0;
1875
Lorenzo Stoakesf307ab62016-10-13 01:20:20 +01001876 len = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
Mike Frysingerf55f1992011-03-29 14:05:12 +01001877
David Howells0ec76a12006-09-27 01:50:15 -07001878 mmput(mm);
1879 return len;
1880}
David Howells7e660872010-01-15 17:01:39 -08001881
1882/**
1883 * nommu_shrink_inode_mappings - Shrink the shared mappings on an inode
1884 * @inode: The inode to check
1885 * @size: The current filesize of the inode
1886 * @newsize: The proposed filesize of the inode
1887 *
1888 * Check the shared mappings on an inode on behalf of a shrinking truncate to
1889 * make sure that that any outstanding VMAs aren't broken and then shrink the
1890 * vm_regions that extend that beyond so that do_mmap_pgoff() doesn't
1891 * automatically grant mappings that are too large.
1892 */
1893int nommu_shrink_inode_mappings(struct inode *inode, size_t size,
1894 size_t newsize)
1895{
1896 struct vm_area_struct *vma;
David Howells7e660872010-01-15 17:01:39 -08001897 struct vm_region *region;
1898 pgoff_t low, high;
1899 size_t r_size, r_top;
1900
1901 low = newsize >> PAGE_SHIFT;
1902 high = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
1903
1904 down_write(&nommu_region_sem);
Davidlohr Bueso1acf2e02014-12-12 16:54:39 -08001905 i_mmap_lock_read(inode->i_mapping);
David Howells7e660872010-01-15 17:01:39 -08001906
1907 /* search for VMAs that fall within the dead zone */
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07001908 vma_interval_tree_foreach(vma, &inode->i_mapping->i_mmap, low, high) {
David Howells7e660872010-01-15 17:01:39 -08001909 /* found one - only interested if it's shared out of the page
1910 * cache */
1911 if (vma->vm_flags & VM_SHARED) {
Davidlohr Bueso1acf2e02014-12-12 16:54:39 -08001912 i_mmap_unlock_read(inode->i_mapping);
David Howells7e660872010-01-15 17:01:39 -08001913 up_write(&nommu_region_sem);
1914 return -ETXTBSY; /* not quite true, but near enough */
1915 }
1916 }
1917
1918 /* reduce any regions that overlap the dead zone - if in existence,
1919 * these will be pointed to by VMAs that don't overlap the dead zone
1920 *
1921 * we don't check for any regions that start beyond the EOF as there
1922 * shouldn't be any
1923 */
Davidlohr Bueso1acf2e02014-12-12 16:54:39 -08001924 vma_interval_tree_foreach(vma, &inode->i_mapping->i_mmap, 0, ULONG_MAX) {
David Howells7e660872010-01-15 17:01:39 -08001925 if (!(vma->vm_flags & VM_SHARED))
1926 continue;
1927
1928 region = vma->vm_region;
1929 r_size = region->vm_top - region->vm_start;
1930 r_top = (region->vm_pgoff << PAGE_SHIFT) + r_size;
1931
1932 if (r_top > newsize) {
1933 region->vm_top -= r_top - newsize;
1934 if (region->vm_end > region->vm_top)
1935 region->vm_end = region->vm_top;
1936 }
1937 }
1938
Davidlohr Bueso1acf2e02014-12-12 16:54:39 -08001939 i_mmap_unlock_read(inode->i_mapping);
David Howells7e660872010-01-15 17:01:39 -08001940 up_write(&nommu_region_sem);
1941 return 0;
1942}
Andrew Shewmakerc9b1d092013-04-29 15:08:10 -07001943
1944/*
1945 * Initialise sysctl_user_reserve_kbytes.
1946 *
1947 * This is intended to prevent a user from starting a single memory hogging
1948 * process, such that they cannot recover (kill the hog) in OVERCOMMIT_NEVER
1949 * mode.
1950 *
1951 * The default value is min(3% of free memory, 128MB)
1952 * 128MB is enough to recover with sshd/login, bash, and top/kill.
1953 */
1954static int __meminit init_user_reserve(void)
1955{
1956 unsigned long free_kbytes;
1957
1958 free_kbytes = global_page_state(NR_FREE_PAGES) << (PAGE_SHIFT - 10);
1959
1960 sysctl_user_reserve_kbytes = min(free_kbytes / 32, 1UL << 17);
1961 return 0;
1962}
Paul Gortmakera4bc6fc2015-05-01 20:08:20 -04001963subsys_initcall(init_user_reserve);
Andrew Shewmaker4eeab4f2013-04-29 15:08:11 -07001964
1965/*
1966 * Initialise sysctl_admin_reserve_kbytes.
1967 *
1968 * The purpose of sysctl_admin_reserve_kbytes is to allow the sys admin
1969 * to log in and kill a memory hogging process.
1970 *
1971 * Systems with more than 256MB will reserve 8MB, enough to recover
1972 * with sshd, bash, and top in OVERCOMMIT_GUESS. Smaller systems will
1973 * only reserve 3% of free pages by default.
1974 */
1975static int __meminit init_admin_reserve(void)
1976{
1977 unsigned long free_kbytes;
1978
1979 free_kbytes = global_page_state(NR_FREE_PAGES) << (PAGE_SHIFT - 10);
1980
1981 sysctl_admin_reserve_kbytes = min(free_kbytes / 32, 1UL << 13);
1982 return 0;
1983}
Paul Gortmakera4bc6fc2015-05-01 20:08:20 -04001984subsys_initcall(init_admin_reserve);