]>
Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/mm/nommu.c | |
4 | * | |
5 | * Replacement code for mm functions to support CPU's that don't | |
6 | * have any form of memory management unit (thus no virtual memory). | |
7 | * | |
8 | * See Documentation/nommu-mmap.txt | |
9 | * | |
8feae131 | 10 | * Copyright (c) 2004-2008 David Howells <[email protected]> |
1da177e4 LT |
11 | * Copyright (c) 2000-2003 David McCullough <[email protected]> |
12 | * Copyright (c) 2000-2001 D Jeff Dionne <[email protected]> | |
13 | * Copyright (c) 2002 Greg Ungerer <[email protected]> | |
29c185e5 | 14 | * Copyright (c) 2007-2010 Paul Mundt <[email protected]> |
1da177e4 LT |
15 | */ |
16 | ||
b1de0d13 MH |
17 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
18 | ||
b95f1b31 | 19 | #include <linux/export.h> |
1da177e4 | 20 | #include <linux/mm.h> |
6e84f315 | 21 | #include <linux/sched/mm.h> |
615d6e87 | 22 | #include <linux/vmacache.h> |
1da177e4 LT |
23 | #include <linux/mman.h> |
24 | #include <linux/swap.h> | |
25 | #include <linux/file.h> | |
26 | #include <linux/highmem.h> | |
27 | #include <linux/pagemap.h> | |
28 | #include <linux/slab.h> | |
29 | #include <linux/vmalloc.h> | |
1da177e4 LT |
30 | #include <linux/blkdev.h> |
31 | #include <linux/backing-dev.h> | |
3b32123d | 32 | #include <linux/compiler.h> |
1da177e4 LT |
33 | #include <linux/mount.h> |
34 | #include <linux/personality.h> | |
35 | #include <linux/security.h> | |
36 | #include <linux/syscalls.h> | |
120a795d | 37 | #include <linux/audit.h> |
b1de0d13 | 38 | #include <linux/printk.h> |
1da177e4 | 39 | |
7c0f6ba6 | 40 | #include <linux/uaccess.h> |
1da177e4 LT |
41 | #include <asm/tlb.h> |
42 | #include <asm/tlbflush.h> | |
eb8cdec4 | 43 | #include <asm/mmu_context.h> |
8feae131 DH |
44 | #include "internal.h" |
45 | ||
1da177e4 | 46 | void *high_memory; |
944b6874 | 47 | EXPORT_SYMBOL(high_memory); |
1da177e4 LT |
48 | struct page *mem_map; |
49 | unsigned long max_mapnr; | |
5b8bf307 | 50 | EXPORT_SYMBOL(max_mapnr); |
4266c97a | 51 | unsigned long highest_memmap_pfn; |
fc4d5c29 | 52 | int sysctl_nr_trim_pages = CONFIG_NOMMU_INITIAL_TRIM_EXCESS; |
1da177e4 LT |
53 | int heap_stack_gap = 0; |
54 | ||
33e5d769 | 55 | atomic_long_t mmap_pages_allocated; |
8feae131 | 56 | |
1da177e4 | 57 | EXPORT_SYMBOL(mem_map); |
1da177e4 | 58 | |
8feae131 DH |
59 | /* list of mapped, potentially shareable regions */ |
60 | static struct kmem_cache *vm_region_jar; | |
61 | struct rb_root nommu_region_tree = RB_ROOT; | |
62 | DECLARE_RWSEM(nommu_region_sem); | |
1da177e4 | 63 | |
f0f37e2f | 64 | const struct vm_operations_struct generic_file_vm_ops = { |
1da177e4 LT |
65 | }; |
66 | ||
1da177e4 LT |
67 | /* |
68 | * Return the total memory allocated for this pointer, not | |
69 | * just what the caller asked for. | |
70 | * | |
71 | * Doesn't have to be accurate, i.e. may have races. | |
72 | */ | |
73 | unsigned int kobjsize(const void *objp) | |
74 | { | |
75 | struct page *page; | |
76 | ||
4016a139 MH |
77 | /* |
78 | * If the object we have should not have ksize performed on it, | |
79 | * return size of 0 | |
80 | */ | |
5a1603be | 81 | if (!objp || !virt_addr_valid(objp)) |
6cfd53fc PM |
82 | return 0; |
83 | ||
84 | page = virt_to_head_page(objp); | |
6cfd53fc PM |
85 | |
86 | /* | |
87 | * If the allocator sets PageSlab, we know the pointer came from | |
88 | * kmalloc(). | |
89 | */ | |
1da177e4 LT |
90 | if (PageSlab(page)) |
91 | return ksize(objp); | |
92 | ||
ab2e83ea PM |
93 | /* |
94 | * If it's not a compound page, see if we have a matching VMA | |
95 | * region. This test is intentionally done in reverse order, | |
96 | * so if there's no VMA, we still fall through and hand back | |
97 | * PAGE_SIZE for 0-order pages. | |
98 | */ | |
99 | if (!PageCompound(page)) { | |
100 | struct vm_area_struct *vma; | |
101 | ||
102 | vma = find_vma(current->mm, (unsigned long)objp); | |
103 | if (vma) | |
104 | return vma->vm_end - vma->vm_start; | |
105 | } | |
106 | ||
6cfd53fc PM |
107 | /* |
108 | * The ksize() function is only guaranteed to work for pointers | |
5a1603be | 109 | * returned by kmalloc(). So handle arbitrary pointers here. |
6cfd53fc | 110 | */ |
a50b854e | 111 | return page_size(page); |
1da177e4 LT |
112 | } |
113 | ||
dfc2f91a PM |
114 | /** |
115 | * follow_pfn - look up PFN at a user virtual address | |
116 | * @vma: memory mapping | |
117 | * @address: user virtual address | |
118 | * @pfn: location to store found PFN | |
119 | * | |
120 | * Only IO mappings and raw PFN mappings are allowed. | |
121 | * | |
122 | * Returns zero and the pfn at @pfn on success, -ve otherwise. | |
123 | */ | |
124 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, | |
125 | unsigned long *pfn) | |
126 | { | |
127 | if (!(vma->vm_flags & (VM_IO | VM_PFNMAP))) | |
128 | return -EINVAL; | |
129 | ||
130 | *pfn = address >> PAGE_SHIFT; | |
131 | return 0; | |
132 | } | |
133 | EXPORT_SYMBOL(follow_pfn); | |
134 | ||
f1c4069e | 135 | LIST_HEAD(vmap_area_list); |
1da177e4 | 136 | |
b3bdda02 | 137 | void vfree(const void *addr) |
1da177e4 LT |
138 | { |
139 | kfree(addr); | |
140 | } | |
b5073173 | 141 | EXPORT_SYMBOL(vfree); |
1da177e4 | 142 | |
dd0fc66f | 143 | void *__vmalloc(unsigned long size, gfp_t gfp_mask, pgprot_t prot) |
1da177e4 LT |
144 | { |
145 | /* | |
8518609d RD |
146 | * You can't specify __GFP_HIGHMEM with kmalloc() since kmalloc() |
147 | * returns only a logical address. | |
1da177e4 | 148 | */ |
84097518 | 149 | return kmalloc(size, (gfp_mask | __GFP_COMP) & ~__GFP_HIGHMEM); |
1da177e4 | 150 | } |
b5073173 | 151 | EXPORT_SYMBOL(__vmalloc); |
1da177e4 | 152 | |
a7c3e901 MH |
153 | void *__vmalloc_node_flags(unsigned long size, int node, gfp_t flags) |
154 | { | |
155 | return __vmalloc(size, flags, PAGE_KERNEL); | |
156 | } | |
157 | ||
f905bc44 PM |
158 | void *vmalloc_user(unsigned long size) |
159 | { | |
160 | void *ret; | |
161 | ||
19809c2d | 162 | ret = __vmalloc(size, GFP_KERNEL | __GFP_ZERO, PAGE_KERNEL); |
f905bc44 PM |
163 | if (ret) { |
164 | struct vm_area_struct *vma; | |
165 | ||
166 | down_write(¤t->mm->mmap_sem); | |
167 | vma = find_vma(current->mm, (unsigned long)ret); | |
168 | if (vma) | |
169 | vma->vm_flags |= VM_USERMAP; | |
170 | up_write(¤t->mm->mmap_sem); | |
171 | } | |
172 | ||
173 | return ret; | |
174 | } | |
175 | EXPORT_SYMBOL(vmalloc_user); | |
176 | ||
b3bdda02 | 177 | struct page *vmalloc_to_page(const void *addr) |
1da177e4 LT |
178 | { |
179 | return virt_to_page(addr); | |
180 | } | |
b5073173 | 181 | EXPORT_SYMBOL(vmalloc_to_page); |
1da177e4 | 182 | |
b3bdda02 | 183 | unsigned long vmalloc_to_pfn(const void *addr) |
1da177e4 LT |
184 | { |
185 | return page_to_pfn(virt_to_page(addr)); | |
186 | } | |
b5073173 | 187 | EXPORT_SYMBOL(vmalloc_to_pfn); |
1da177e4 LT |
188 | |
189 | long vread(char *buf, char *addr, unsigned long count) | |
190 | { | |
9bde916b CG |
191 | /* Don't allow overflow */ |
192 | if ((unsigned long) buf + count < count) | |
193 | count = -(unsigned long) buf; | |
194 | ||
1da177e4 LT |
195 | memcpy(buf, addr, count); |
196 | return count; | |
197 | } | |
198 | ||
199 | long vwrite(char *buf, char *addr, unsigned long count) | |
200 | { | |
201 | /* Don't allow overflow */ | |
202 | if ((unsigned long) addr + count < count) | |
203 | count = -(unsigned long) addr; | |
204 | ||
205 | memcpy(addr, buf, count); | |
ac714904 | 206 | return count; |
1da177e4 LT |
207 | } |
208 | ||
209 | /* | |
e1c05067 | 210 | * vmalloc - allocate virtually contiguous memory |
1da177e4 LT |
211 | * |
212 | * @size: allocation size | |
213 | * | |
214 | * Allocate enough pages to cover @size from the page level | |
e1c05067 | 215 | * allocator and map them into contiguous kernel virtual space. |
1da177e4 | 216 | * |
c1c8897f | 217 | * For tight control over page level allocator and protection flags |
1da177e4 LT |
218 | * use __vmalloc() instead. |
219 | */ | |
220 | void *vmalloc(unsigned long size) | |
221 | { | |
222 | return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM, PAGE_KERNEL); | |
223 | } | |
f6138882 AM |
224 | EXPORT_SYMBOL(vmalloc); |
225 | ||
e1ca7788 | 226 | /* |
e1c05067 | 227 | * vzalloc - allocate virtually contiguous memory with zero fill |
e1ca7788 DY |
228 | * |
229 | * @size: allocation size | |
230 | * | |
231 | * Allocate enough pages to cover @size from the page level | |
e1c05067 | 232 | * allocator and map them into contiguous kernel virtual space. |
e1ca7788 DY |
233 | * The memory allocated is set to zero. |
234 | * | |
235 | * For tight control over page level allocator and protection flags | |
236 | * use __vmalloc() instead. | |
237 | */ | |
238 | void *vzalloc(unsigned long size) | |
239 | { | |
240 | return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM | __GFP_ZERO, | |
241 | PAGE_KERNEL); | |
242 | } | |
243 | EXPORT_SYMBOL(vzalloc); | |
244 | ||
245 | /** | |
246 | * vmalloc_node - allocate memory on a specific node | |
247 | * @size: allocation size | |
248 | * @node: numa node | |
249 | * | |
250 | * Allocate enough pages to cover @size from the page level | |
251 | * allocator and map them into contiguous kernel virtual space. | |
252 | * | |
253 | * For tight control over page level allocator and protection flags | |
254 | * use __vmalloc() instead. | |
255 | */ | |
f6138882 AM |
256 | void *vmalloc_node(unsigned long size, int node) |
257 | { | |
258 | return vmalloc(size); | |
259 | } | |
9a14f653 | 260 | EXPORT_SYMBOL(vmalloc_node); |
e1ca7788 DY |
261 | |
262 | /** | |
263 | * vzalloc_node - allocate memory on a specific node with zero fill | |
264 | * @size: allocation size | |
265 | * @node: numa node | |
266 | * | |
267 | * Allocate enough pages to cover @size from the page level | |
268 | * allocator and map them into contiguous kernel virtual space. | |
269 | * The memory allocated is set to zero. | |
270 | * | |
271 | * For tight control over page level allocator and protection flags | |
272 | * use __vmalloc() instead. | |
273 | */ | |
274 | void *vzalloc_node(unsigned long size, int node) | |
275 | { | |
276 | return vzalloc(size); | |
277 | } | |
278 | EXPORT_SYMBOL(vzalloc_node); | |
1da177e4 | 279 | |
1af446ed PM |
280 | /** |
281 | * vmalloc_exec - allocate virtually contiguous, executable memory | |
282 | * @size: allocation size | |
283 | * | |
284 | * Kernel-internal function to allocate enough pages to cover @size | |
285 | * the page level allocator and map them into contiguous and | |
286 | * executable kernel virtual space. | |
287 | * | |
288 | * For tight control over page level allocator and protection flags | |
289 | * use __vmalloc() instead. | |
290 | */ | |
291 | ||
292 | void *vmalloc_exec(unsigned long size) | |
293 | { | |
294 | return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM, PAGE_KERNEL_EXEC); | |
295 | } | |
296 | ||
b5073173 PM |
297 | /** |
298 | * vmalloc_32 - allocate virtually contiguous memory (32bit addressable) | |
1da177e4 LT |
299 | * @size: allocation size |
300 | * | |
301 | * Allocate enough 32bit PA addressable pages to cover @size from the | |
e1c05067 | 302 | * page level allocator and map them into contiguous kernel virtual space. |
1da177e4 LT |
303 | */ |
304 | void *vmalloc_32(unsigned long size) | |
305 | { | |
306 | return __vmalloc(size, GFP_KERNEL, PAGE_KERNEL); | |
307 | } | |
b5073173 PM |
308 | EXPORT_SYMBOL(vmalloc_32); |
309 | ||
310 | /** | |
311 | * vmalloc_32_user - allocate zeroed virtually contiguous 32bit memory | |
312 | * @size: allocation size | |
313 | * | |
314 | * The resulting memory area is 32bit addressable and zeroed so it can be | |
315 | * mapped to userspace without leaking data. | |
f905bc44 PM |
316 | * |
317 | * VM_USERMAP is set on the corresponding VMA so that subsequent calls to | |
318 | * remap_vmalloc_range() are permissible. | |
b5073173 PM |
319 | */ |
320 | void *vmalloc_32_user(unsigned long size) | |
321 | { | |
f905bc44 PM |
322 | /* |
323 | * We'll have to sort out the ZONE_DMA bits for 64-bit, | |
324 | * but for now this can simply use vmalloc_user() directly. | |
325 | */ | |
326 | return vmalloc_user(size); | |
b5073173 PM |
327 | } |
328 | EXPORT_SYMBOL(vmalloc_32_user); | |
1da177e4 LT |
329 | |
330 | void *vmap(struct page **pages, unsigned int count, unsigned long flags, pgprot_t prot) | |
331 | { | |
332 | BUG(); | |
333 | return NULL; | |
334 | } | |
b5073173 | 335 | EXPORT_SYMBOL(vmap); |
1da177e4 | 336 | |
b3bdda02 | 337 | void vunmap(const void *addr) |
1da177e4 LT |
338 | { |
339 | BUG(); | |
340 | } | |
b5073173 | 341 | EXPORT_SYMBOL(vunmap); |
1da177e4 | 342 | |
eb6434d9 PM |
343 | void *vm_map_ram(struct page **pages, unsigned int count, int node, pgprot_t prot) |
344 | { | |
345 | BUG(); | |
346 | return NULL; | |
347 | } | |
348 | EXPORT_SYMBOL(vm_map_ram); | |
349 | ||
350 | void vm_unmap_ram(const void *mem, unsigned int count) | |
351 | { | |
352 | BUG(); | |
353 | } | |
354 | EXPORT_SYMBOL(vm_unmap_ram); | |
355 | ||
356 | void vm_unmap_aliases(void) | |
357 | { | |
358 | } | |
359 | EXPORT_SYMBOL_GPL(vm_unmap_aliases); | |
360 | ||
1eeb66a1 CH |
361 | /* |
362 | * Implement a stub for vmalloc_sync_all() if the architecture chose not to | |
363 | * have one. | |
364 | */ | |
3b32123d | 365 | void __weak vmalloc_sync_all(void) |
1eeb66a1 CH |
366 | { |
367 | } | |
368 | ||
cd12909c | 369 | struct vm_struct *alloc_vm_area(size_t size, pte_t **ptes) |
29c185e5 PM |
370 | { |
371 | BUG(); | |
372 | return NULL; | |
373 | } | |
374 | EXPORT_SYMBOL_GPL(alloc_vm_area); | |
375 | ||
376 | void free_vm_area(struct vm_struct *area) | |
377 | { | |
378 | BUG(); | |
379 | } | |
380 | EXPORT_SYMBOL_GPL(free_vm_area); | |
381 | ||
b5073173 PM |
382 | int vm_insert_page(struct vm_area_struct *vma, unsigned long addr, |
383 | struct page *page) | |
384 | { | |
385 | return -EINVAL; | |
386 | } | |
387 | EXPORT_SYMBOL(vm_insert_page); | |
388 | ||
a667d745 SJ |
389 | int vm_map_pages(struct vm_area_struct *vma, struct page **pages, |
390 | unsigned long num) | |
391 | { | |
392 | return -EINVAL; | |
393 | } | |
394 | EXPORT_SYMBOL(vm_map_pages); | |
395 | ||
396 | int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages, | |
397 | unsigned long num) | |
398 | { | |
399 | return -EINVAL; | |
400 | } | |
401 | EXPORT_SYMBOL(vm_map_pages_zero); | |
402 | ||
1da177e4 LT |
403 | /* |
404 | * sys_brk() for the most part doesn't need the global kernel | |
405 | * lock, except when an application is doing something nasty | |
406 | * like trying to un-brk an area that has already been mapped | |
407 | * to a regular file. in this case, the unmapping will need | |
408 | * to invoke file system routines that need the global lock. | |
409 | */ | |
6a6160a7 | 410 | SYSCALL_DEFINE1(brk, unsigned long, brk) |
1da177e4 LT |
411 | { |
412 | struct mm_struct *mm = current->mm; | |
413 | ||
414 | if (brk < mm->start_brk || brk > mm->context.end_brk) | |
415 | return mm->brk; | |
416 | ||
417 | if (mm->brk == brk) | |
418 | return mm->brk; | |
419 | ||
420 | /* | |
421 | * Always allow shrinking brk | |
422 | */ | |
423 | if (brk <= mm->brk) { | |
424 | mm->brk = brk; | |
425 | return brk; | |
426 | } | |
427 | ||
428 | /* | |
429 | * Ok, looks good - let it rip. | |
430 | */ | |
cfe79c00 | 431 | flush_icache_range(mm->brk, brk); |
1da177e4 LT |
432 | return mm->brk = brk; |
433 | } | |
434 | ||
8feae131 | 435 | /* |
3edf41d8 | 436 | * initialise the percpu counter for VM and region record slabs |
8feae131 DH |
437 | */ |
438 | void __init mmap_init(void) | |
1da177e4 | 439 | { |
00a62ce9 KM |
440 | int ret; |
441 | ||
908c7f19 | 442 | ret = percpu_counter_init(&vm_committed_as, 0, GFP_KERNEL); |
00a62ce9 | 443 | VM_BUG_ON(ret); |
5d097056 | 444 | vm_region_jar = KMEM_CACHE(vm_region, SLAB_PANIC|SLAB_ACCOUNT); |
1da177e4 | 445 | } |
1da177e4 | 446 | |
3034097a | 447 | /* |
8feae131 DH |
448 | * validate the region tree |
449 | * - the caller must hold the region lock | |
3034097a | 450 | */ |
8feae131 DH |
451 | #ifdef CONFIG_DEBUG_NOMMU_REGIONS |
452 | static noinline void validate_nommu_regions(void) | |
3034097a | 453 | { |
8feae131 DH |
454 | struct vm_region *region, *last; |
455 | struct rb_node *p, *lastp; | |
3034097a | 456 | |
8feae131 DH |
457 | lastp = rb_first(&nommu_region_tree); |
458 | if (!lastp) | |
459 | return; | |
460 | ||
461 | last = rb_entry(lastp, struct vm_region, vm_rb); | |
c9427bc0 GT |
462 | BUG_ON(last->vm_end <= last->vm_start); |
463 | BUG_ON(last->vm_top < last->vm_end); | |
8feae131 DH |
464 | |
465 | while ((p = rb_next(lastp))) { | |
466 | region = rb_entry(p, struct vm_region, vm_rb); | |
467 | last = rb_entry(lastp, struct vm_region, vm_rb); | |
468 | ||
c9427bc0 GT |
469 | BUG_ON(region->vm_end <= region->vm_start); |
470 | BUG_ON(region->vm_top < region->vm_end); | |
471 | BUG_ON(region->vm_start < last->vm_top); | |
3034097a | 472 | |
8feae131 DH |
473 | lastp = p; |
474 | } | |
3034097a | 475 | } |
8feae131 | 476 | #else |
33e5d769 DH |
477 | static void validate_nommu_regions(void) |
478 | { | |
479 | } | |
8feae131 | 480 | #endif |
3034097a DH |
481 | |
482 | /* | |
8feae131 | 483 | * add a region into the global tree |
3034097a | 484 | */ |
8feae131 | 485 | static void add_nommu_region(struct vm_region *region) |
3034097a | 486 | { |
8feae131 DH |
487 | struct vm_region *pregion; |
488 | struct rb_node **p, *parent; | |
3034097a | 489 | |
8feae131 DH |
490 | validate_nommu_regions(); |
491 | ||
8feae131 DH |
492 | parent = NULL; |
493 | p = &nommu_region_tree.rb_node; | |
494 | while (*p) { | |
495 | parent = *p; | |
496 | pregion = rb_entry(parent, struct vm_region, vm_rb); | |
497 | if (region->vm_start < pregion->vm_start) | |
498 | p = &(*p)->rb_left; | |
499 | else if (region->vm_start > pregion->vm_start) | |
500 | p = &(*p)->rb_right; | |
501 | else if (pregion == region) | |
502 | return; | |
503 | else | |
504 | BUG(); | |
3034097a DH |
505 | } |
506 | ||
8feae131 DH |
507 | rb_link_node(®ion->vm_rb, parent, p); |
508 | rb_insert_color(®ion->vm_rb, &nommu_region_tree); | |
3034097a | 509 | |
8feae131 | 510 | validate_nommu_regions(); |
3034097a | 511 | } |
3034097a | 512 | |
930e652a | 513 | /* |
8feae131 | 514 | * delete a region from the global tree |
930e652a | 515 | */ |
8feae131 | 516 | static void delete_nommu_region(struct vm_region *region) |
930e652a | 517 | { |
8feae131 | 518 | BUG_ON(!nommu_region_tree.rb_node); |
930e652a | 519 | |
8feae131 DH |
520 | validate_nommu_regions(); |
521 | rb_erase(®ion->vm_rb, &nommu_region_tree); | |
522 | validate_nommu_regions(); | |
57c8f63e GU |
523 | } |
524 | ||
6fa5f80b | 525 | /* |
8feae131 | 526 | * free a contiguous series of pages |
6fa5f80b | 527 | */ |
8feae131 | 528 | static void free_page_series(unsigned long from, unsigned long to) |
6fa5f80b | 529 | { |
8feae131 DH |
530 | for (; from < to; from += PAGE_SIZE) { |
531 | struct page *page = virt_to_page(from); | |
532 | ||
33e5d769 | 533 | atomic_long_dec(&mmap_pages_allocated); |
8feae131 | 534 | put_page(page); |
6fa5f80b | 535 | } |
6fa5f80b DH |
536 | } |
537 | ||
3034097a | 538 | /* |
8feae131 | 539 | * release a reference to a region |
33e5d769 | 540 | * - the caller must hold the region semaphore for writing, which this releases |
dd8632a1 | 541 | * - the region may not have been added to the tree yet, in which case vm_top |
8feae131 | 542 | * will equal vm_start |
3034097a | 543 | */ |
8feae131 DH |
544 | static void __put_nommu_region(struct vm_region *region) |
545 | __releases(nommu_region_sem) | |
1da177e4 | 546 | { |
8feae131 | 547 | BUG_ON(!nommu_region_tree.rb_node); |
1da177e4 | 548 | |
1e2ae599 | 549 | if (--region->vm_usage == 0) { |
dd8632a1 | 550 | if (region->vm_top > region->vm_start) |
8feae131 DH |
551 | delete_nommu_region(region); |
552 | up_write(&nommu_region_sem); | |
553 | ||
554 | if (region->vm_file) | |
555 | fput(region->vm_file); | |
556 | ||
557 | /* IO memory and memory shared directly out of the pagecache | |
558 | * from ramfs/tmpfs mustn't be released here */ | |
22cc877b | 559 | if (region->vm_flags & VM_MAPPED_COPY) |
dd8632a1 | 560 | free_page_series(region->vm_start, region->vm_top); |
8feae131 DH |
561 | kmem_cache_free(vm_region_jar, region); |
562 | } else { | |
563 | up_write(&nommu_region_sem); | |
1da177e4 | 564 | } |
8feae131 | 565 | } |
1da177e4 | 566 | |
8feae131 DH |
567 | /* |
568 | * release a reference to a region | |
569 | */ | |
570 | static void put_nommu_region(struct vm_region *region) | |
571 | { | |
572 | down_write(&nommu_region_sem); | |
573 | __put_nommu_region(region); | |
1da177e4 LT |
574 | } |
575 | ||
3034097a | 576 | /* |
8feae131 DH |
577 | * add a VMA into a process's mm_struct in the appropriate place in the list |
578 | * and tree and add to the address space's page tree also if not an anonymous | |
579 | * page | |
580 | * - should be called with mm->mmap_sem held writelocked | |
3034097a | 581 | */ |
8feae131 | 582 | static void add_vma_to_mm(struct mm_struct *mm, struct vm_area_struct *vma) |
1da177e4 | 583 | { |
6038def0 | 584 | struct vm_area_struct *pvma, *prev; |
1da177e4 | 585 | struct address_space *mapping; |
6038def0 | 586 | struct rb_node **p, *parent, *rb_prev; |
8feae131 | 587 | |
8feae131 DH |
588 | BUG_ON(!vma->vm_region); |
589 | ||
590 | mm->map_count++; | |
591 | vma->vm_mm = mm; | |
1da177e4 LT |
592 | |
593 | /* add the VMA to the mapping */ | |
594 | if (vma->vm_file) { | |
595 | mapping = vma->vm_file->f_mapping; | |
596 | ||
83cde9e8 | 597 | i_mmap_lock_write(mapping); |
1da177e4 | 598 | flush_dcache_mmap_lock(mapping); |
6b2dbba8 | 599 | vma_interval_tree_insert(vma, &mapping->i_mmap); |
1da177e4 | 600 | flush_dcache_mmap_unlock(mapping); |
83cde9e8 | 601 | i_mmap_unlock_write(mapping); |
1da177e4 LT |
602 | } |
603 | ||
8feae131 | 604 | /* add the VMA to the tree */ |
6038def0 | 605 | parent = rb_prev = NULL; |
8feae131 | 606 | p = &mm->mm_rb.rb_node; |
1da177e4 LT |
607 | while (*p) { |
608 | parent = *p; | |
609 | pvma = rb_entry(parent, struct vm_area_struct, vm_rb); | |
610 | ||
8feae131 DH |
611 | /* sort by: start addr, end addr, VMA struct addr in that order |
612 | * (the latter is necessary as we may get identical VMAs) */ | |
613 | if (vma->vm_start < pvma->vm_start) | |
1da177e4 | 614 | p = &(*p)->rb_left; |
6038def0 NK |
615 | else if (vma->vm_start > pvma->vm_start) { |
616 | rb_prev = parent; | |
1da177e4 | 617 | p = &(*p)->rb_right; |
6038def0 | 618 | } else if (vma->vm_end < pvma->vm_end) |
8feae131 | 619 | p = &(*p)->rb_left; |
6038def0 NK |
620 | else if (vma->vm_end > pvma->vm_end) { |
621 | rb_prev = parent; | |
8feae131 | 622 | p = &(*p)->rb_right; |
6038def0 | 623 | } else if (vma < pvma) |
8feae131 | 624 | p = &(*p)->rb_left; |
6038def0 NK |
625 | else if (vma > pvma) { |
626 | rb_prev = parent; | |
8feae131 | 627 | p = &(*p)->rb_right; |
6038def0 | 628 | } else |
8feae131 | 629 | BUG(); |
1da177e4 LT |
630 | } |
631 | ||
632 | rb_link_node(&vma->vm_rb, parent, p); | |
8feae131 DH |
633 | rb_insert_color(&vma->vm_rb, &mm->mm_rb); |
634 | ||
635 | /* add VMA to the VMA list also */ | |
6038def0 NK |
636 | prev = NULL; |
637 | if (rb_prev) | |
638 | prev = rb_entry(rb_prev, struct vm_area_struct, vm_rb); | |
8feae131 | 639 | |
6038def0 | 640 | __vma_link_list(mm, vma, prev, parent); |
1da177e4 LT |
641 | } |
642 | ||
3034097a | 643 | /* |
8feae131 | 644 | * delete a VMA from its owning mm_struct and address space |
3034097a | 645 | */ |
8feae131 | 646 | static void delete_vma_from_mm(struct vm_area_struct *vma) |
1da177e4 | 647 | { |
615d6e87 | 648 | int i; |
1da177e4 | 649 | struct address_space *mapping; |
8feae131 | 650 | struct mm_struct *mm = vma->vm_mm; |
615d6e87 | 651 | struct task_struct *curr = current; |
8feae131 | 652 | |
8feae131 | 653 | mm->map_count--; |
615d6e87 DB |
654 | for (i = 0; i < VMACACHE_SIZE; i++) { |
655 | /* if the vma is cached, invalidate the entire cache */ | |
314ff785 | 656 | if (curr->vmacache.vmas[i] == vma) { |
e020d5bd | 657 | vmacache_invalidate(mm); |
615d6e87 DB |
658 | break; |
659 | } | |
660 | } | |
1da177e4 LT |
661 | |
662 | /* remove the VMA from the mapping */ | |
663 | if (vma->vm_file) { | |
664 | mapping = vma->vm_file->f_mapping; | |
665 | ||
83cde9e8 | 666 | i_mmap_lock_write(mapping); |
1da177e4 | 667 | flush_dcache_mmap_lock(mapping); |
6b2dbba8 | 668 | vma_interval_tree_remove(vma, &mapping->i_mmap); |
1da177e4 | 669 | flush_dcache_mmap_unlock(mapping); |
83cde9e8 | 670 | i_mmap_unlock_write(mapping); |
1da177e4 LT |
671 | } |
672 | ||
8feae131 DH |
673 | /* remove from the MM's tree and list */ |
674 | rb_erase(&vma->vm_rb, &mm->mm_rb); | |
b951bf2c NK |
675 | |
676 | if (vma->vm_prev) | |
677 | vma->vm_prev->vm_next = vma->vm_next; | |
678 | else | |
679 | mm->mmap = vma->vm_next; | |
680 | ||
681 | if (vma->vm_next) | |
682 | vma->vm_next->vm_prev = vma->vm_prev; | |
8feae131 DH |
683 | } |
684 | ||
685 | /* | |
686 | * destroy a VMA record | |
687 | */ | |
688 | static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma) | |
689 | { | |
8feae131 DH |
690 | if (vma->vm_ops && vma->vm_ops->close) |
691 | vma->vm_ops->close(vma); | |
e9714acf | 692 | if (vma->vm_file) |
8feae131 | 693 | fput(vma->vm_file); |
8feae131 | 694 | put_nommu_region(vma->vm_region); |
3928d4f5 | 695 | vm_area_free(vma); |
8feae131 DH |
696 | } |
697 | ||
698 | /* | |
699 | * look up the first VMA in which addr resides, NULL if none | |
700 | * - should be called with mm->mmap_sem at least held readlocked | |
701 | */ | |
702 | struct vm_area_struct *find_vma(struct mm_struct *mm, unsigned long addr) | |
703 | { | |
704 | struct vm_area_struct *vma; | |
8feae131 DH |
705 | |
706 | /* check the cache first */ | |
615d6e87 DB |
707 | vma = vmacache_find(mm, addr); |
708 | if (likely(vma)) | |
8feae131 DH |
709 | return vma; |
710 | ||
e922c4c5 | 711 | /* trawl the list (there may be multiple mappings in which addr |
8feae131 | 712 | * resides) */ |
e922c4c5 | 713 | for (vma = mm->mmap; vma; vma = vma->vm_next) { |
8feae131 DH |
714 | if (vma->vm_start > addr) |
715 | return NULL; | |
716 | if (vma->vm_end > addr) { | |
615d6e87 | 717 | vmacache_update(addr, vma); |
8feae131 DH |
718 | return vma; |
719 | } | |
720 | } | |
721 | ||
722 | return NULL; | |
723 | } | |
724 | EXPORT_SYMBOL(find_vma); | |
725 | ||
726 | /* | |
727 | * find a VMA | |
728 | * - we don't extend stack VMAs under NOMMU conditions | |
729 | */ | |
730 | struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr) | |
731 | { | |
7561e8ca | 732 | return find_vma(mm, addr); |
8feae131 DH |
733 | } |
734 | ||
735 | /* | |
736 | * expand a stack to a given address | |
737 | * - not supported under NOMMU conditions | |
738 | */ | |
739 | int expand_stack(struct vm_area_struct *vma, unsigned long address) | |
740 | { | |
741 | return -ENOMEM; | |
742 | } | |
743 | ||
744 | /* | |
745 | * look up the first VMA exactly that exactly matches addr | |
746 | * - should be called with mm->mmap_sem at least held readlocked | |
747 | */ | |
748 | static struct vm_area_struct *find_vma_exact(struct mm_struct *mm, | |
749 | unsigned long addr, | |
750 | unsigned long len) | |
751 | { | |
752 | struct vm_area_struct *vma; | |
8feae131 DH |
753 | unsigned long end = addr + len; |
754 | ||
755 | /* check the cache first */ | |
615d6e87 DB |
756 | vma = vmacache_find_exact(mm, addr, end); |
757 | if (vma) | |
8feae131 DH |
758 | return vma; |
759 | ||
e922c4c5 | 760 | /* trawl the list (there may be multiple mappings in which addr |
8feae131 | 761 | * resides) */ |
e922c4c5 | 762 | for (vma = mm->mmap; vma; vma = vma->vm_next) { |
8feae131 DH |
763 | if (vma->vm_start < addr) |
764 | continue; | |
765 | if (vma->vm_start > addr) | |
766 | return NULL; | |
767 | if (vma->vm_end == end) { | |
615d6e87 | 768 | vmacache_update(addr, vma); |
8feae131 DH |
769 | return vma; |
770 | } | |
771 | } | |
772 | ||
773 | return NULL; | |
1da177e4 LT |
774 | } |
775 | ||
776 | /* | |
777 | * determine whether a mapping should be permitted and, if so, what sort of | |
778 | * mapping we're capable of supporting | |
779 | */ | |
780 | static int validate_mmap_request(struct file *file, | |
781 | unsigned long addr, | |
782 | unsigned long len, | |
783 | unsigned long prot, | |
784 | unsigned long flags, | |
785 | unsigned long pgoff, | |
786 | unsigned long *_capabilities) | |
787 | { | |
8feae131 | 788 | unsigned long capabilities, rlen; |
1da177e4 LT |
789 | int ret; |
790 | ||
791 | /* do the simple checks first */ | |
22cc877b | 792 | if (flags & MAP_FIXED) |
1da177e4 | 793 | return -EINVAL; |
1da177e4 LT |
794 | |
795 | if ((flags & MAP_TYPE) != MAP_PRIVATE && | |
796 | (flags & MAP_TYPE) != MAP_SHARED) | |
797 | return -EINVAL; | |
798 | ||
f81cff0d | 799 | if (!len) |
1da177e4 LT |
800 | return -EINVAL; |
801 | ||
f81cff0d | 802 | /* Careful about overflows.. */ |
8feae131 DH |
803 | rlen = PAGE_ALIGN(len); |
804 | if (!rlen || rlen > TASK_SIZE) | |
f81cff0d MF |
805 | return -ENOMEM; |
806 | ||
1da177e4 | 807 | /* offset overflow? */ |
8feae131 | 808 | if ((pgoff + (rlen >> PAGE_SHIFT)) < pgoff) |
f81cff0d | 809 | return -EOVERFLOW; |
1da177e4 LT |
810 | |
811 | if (file) { | |
1da177e4 | 812 | /* files must support mmap */ |
72c2d531 | 813 | if (!file->f_op->mmap) |
1da177e4 LT |
814 | return -ENODEV; |
815 | ||
816 | /* work out if what we've got could possibly be shared | |
817 | * - we support chardevs that provide their own "memory" | |
818 | * - we support files/blockdevs that are memory backed | |
819 | */ | |
b4caecd4 CH |
820 | if (file->f_op->mmap_capabilities) { |
821 | capabilities = file->f_op->mmap_capabilities(file); | |
822 | } else { | |
1da177e4 LT |
823 | /* no explicit capabilities set, so assume some |
824 | * defaults */ | |
496ad9aa | 825 | switch (file_inode(file)->i_mode & S_IFMT) { |
1da177e4 LT |
826 | case S_IFREG: |
827 | case S_IFBLK: | |
b4caecd4 | 828 | capabilities = NOMMU_MAP_COPY; |
1da177e4 LT |
829 | break; |
830 | ||
831 | case S_IFCHR: | |
832 | capabilities = | |
b4caecd4 CH |
833 | NOMMU_MAP_DIRECT | |
834 | NOMMU_MAP_READ | | |
835 | NOMMU_MAP_WRITE; | |
1da177e4 LT |
836 | break; |
837 | ||
838 | default: | |
839 | return -EINVAL; | |
840 | } | |
841 | } | |
842 | ||
843 | /* eliminate any capabilities that we can't support on this | |
844 | * device */ | |
845 | if (!file->f_op->get_unmapped_area) | |
b4caecd4 | 846 | capabilities &= ~NOMMU_MAP_DIRECT; |
6e242a1c | 847 | if (!(file->f_mode & FMODE_CAN_READ)) |
b4caecd4 | 848 | capabilities &= ~NOMMU_MAP_COPY; |
1da177e4 | 849 | |
28d7a6ae GY |
850 | /* The file shall have been opened with read permission. */ |
851 | if (!(file->f_mode & FMODE_READ)) | |
852 | return -EACCES; | |
853 | ||
1da177e4 LT |
854 | if (flags & MAP_SHARED) { |
855 | /* do checks for writing, appending and locking */ | |
856 | if ((prot & PROT_WRITE) && | |
857 | !(file->f_mode & FMODE_WRITE)) | |
858 | return -EACCES; | |
859 | ||
496ad9aa | 860 | if (IS_APPEND(file_inode(file)) && |
1da177e4 LT |
861 | (file->f_mode & FMODE_WRITE)) |
862 | return -EACCES; | |
863 | ||
d7a06983 | 864 | if (locks_verify_locked(file)) |
1da177e4 LT |
865 | return -EAGAIN; |
866 | ||
b4caecd4 | 867 | if (!(capabilities & NOMMU_MAP_DIRECT)) |
1da177e4 LT |
868 | return -ENODEV; |
869 | ||
1da177e4 | 870 | /* we mustn't privatise shared mappings */ |
b4caecd4 | 871 | capabilities &= ~NOMMU_MAP_COPY; |
ac714904 | 872 | } else { |
1da177e4 LT |
873 | /* we're going to read the file into private memory we |
874 | * allocate */ | |
b4caecd4 | 875 | if (!(capabilities & NOMMU_MAP_COPY)) |
1da177e4 LT |
876 | return -ENODEV; |
877 | ||
878 | /* we don't permit a private writable mapping to be | |
879 | * shared with the backing device */ | |
880 | if (prot & PROT_WRITE) | |
b4caecd4 | 881 | capabilities &= ~NOMMU_MAP_DIRECT; |
1da177e4 LT |
882 | } |
883 | ||
b4caecd4 CH |
884 | if (capabilities & NOMMU_MAP_DIRECT) { |
885 | if (((prot & PROT_READ) && !(capabilities & NOMMU_MAP_READ)) || | |
886 | ((prot & PROT_WRITE) && !(capabilities & NOMMU_MAP_WRITE)) || | |
887 | ((prot & PROT_EXEC) && !(capabilities & NOMMU_MAP_EXEC)) | |
3c7b2045 | 888 | ) { |
b4caecd4 | 889 | capabilities &= ~NOMMU_MAP_DIRECT; |
3c7b2045 | 890 | if (flags & MAP_SHARED) { |
22cc877b | 891 | pr_warn("MAP_SHARED not completely supported on !MMU\n"); |
3c7b2045 BS |
892 | return -EINVAL; |
893 | } | |
894 | } | |
895 | } | |
896 | ||
1da177e4 LT |
897 | /* handle executable mappings and implied executable |
898 | * mappings */ | |
90f8572b | 899 | if (path_noexec(&file->f_path)) { |
1da177e4 LT |
900 | if (prot & PROT_EXEC) |
901 | return -EPERM; | |
ac714904 | 902 | } else if ((prot & PROT_READ) && !(prot & PROT_EXEC)) { |
1da177e4 LT |
903 | /* handle implication of PROT_EXEC by PROT_READ */ |
904 | if (current->personality & READ_IMPLIES_EXEC) { | |
b4caecd4 | 905 | if (capabilities & NOMMU_MAP_EXEC) |
1da177e4 LT |
906 | prot |= PROT_EXEC; |
907 | } | |
ac714904 | 908 | } else if ((prot & PROT_READ) && |
1da177e4 | 909 | (prot & PROT_EXEC) && |
b4caecd4 | 910 | !(capabilities & NOMMU_MAP_EXEC) |
1da177e4 LT |
911 | ) { |
912 | /* backing file is not executable, try to copy */ | |
b4caecd4 | 913 | capabilities &= ~NOMMU_MAP_DIRECT; |
1da177e4 | 914 | } |
ac714904 | 915 | } else { |
1da177e4 LT |
916 | /* anonymous mappings are always memory backed and can be |
917 | * privately mapped | |
918 | */ | |
b4caecd4 | 919 | capabilities = NOMMU_MAP_COPY; |
1da177e4 LT |
920 | |
921 | /* handle PROT_EXEC implication by PROT_READ */ | |
922 | if ((prot & PROT_READ) && | |
923 | (current->personality & READ_IMPLIES_EXEC)) | |
924 | prot |= PROT_EXEC; | |
925 | } | |
926 | ||
927 | /* allow the security API to have its say */ | |
e5467859 | 928 | ret = security_mmap_addr(addr); |
1da177e4 LT |
929 | if (ret < 0) |
930 | return ret; | |
931 | ||
932 | /* looks okay */ | |
933 | *_capabilities = capabilities; | |
934 | return 0; | |
935 | } | |
936 | ||
937 | /* | |
938 | * we've determined that we can make the mapping, now translate what we | |
939 | * now know into VMA flags | |
940 | */ | |
941 | static unsigned long determine_vm_flags(struct file *file, | |
942 | unsigned long prot, | |
943 | unsigned long flags, | |
944 | unsigned long capabilities) | |
945 | { | |
946 | unsigned long vm_flags; | |
947 | ||
e6bfb709 | 948 | vm_flags = calc_vm_prot_bits(prot, 0) | calc_vm_flag_bits(flags); |
1da177e4 LT |
949 | /* vm_flags |= mm->def_flags; */ |
950 | ||
b4caecd4 | 951 | if (!(capabilities & NOMMU_MAP_DIRECT)) { |
1da177e4 | 952 | /* attempt to share read-only copies of mapped file chunks */ |
3c7b2045 | 953 | vm_flags |= VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC; |
1da177e4 LT |
954 | if (file && !(prot & PROT_WRITE)) |
955 | vm_flags |= VM_MAYSHARE; | |
3c7b2045 | 956 | } else { |
1da177e4 LT |
957 | /* overlay a shareable mapping on the backing device or inode |
958 | * if possible - used for chardevs, ramfs/tmpfs/shmfs and | |
959 | * romfs/cramfs */ | |
b4caecd4 | 960 | vm_flags |= VM_MAYSHARE | (capabilities & NOMMU_VMFLAGS); |
1da177e4 | 961 | if (flags & MAP_SHARED) |
3c7b2045 | 962 | vm_flags |= VM_SHARED; |
1da177e4 LT |
963 | } |
964 | ||
965 | /* refuse to let anyone share private mappings with this process if | |
966 | * it's being traced - otherwise breakpoints set in it may interfere | |
967 | * with another untraced process | |
968 | */ | |
a288eecc | 969 | if ((flags & MAP_PRIVATE) && current->ptrace) |
1da177e4 LT |
970 | vm_flags &= ~VM_MAYSHARE; |
971 | ||
972 | return vm_flags; | |
973 | } | |
974 | ||
975 | /* | |
8feae131 DH |
976 | * set up a shared mapping on a file (the driver or filesystem provides and |
977 | * pins the storage) | |
1da177e4 | 978 | */ |
8feae131 | 979 | static int do_mmap_shared_file(struct vm_area_struct *vma) |
1da177e4 LT |
980 | { |
981 | int ret; | |
982 | ||
f74ac015 | 983 | ret = call_mmap(vma->vm_file, vma); |
dd8632a1 PM |
984 | if (ret == 0) { |
985 | vma->vm_region->vm_top = vma->vm_region->vm_end; | |
645d83c5 | 986 | return 0; |
dd8632a1 | 987 | } |
1da177e4 LT |
988 | if (ret != -ENOSYS) |
989 | return ret; | |
990 | ||
3fa30460 DH |
991 | /* getting -ENOSYS indicates that direct mmap isn't possible (as |
992 | * opposed to tried but failed) so we can only give a suitable error as | |
993 | * it's not possible to make a private copy if MAP_SHARED was given */ | |
1da177e4 LT |
994 | return -ENODEV; |
995 | } | |
996 | ||
997 | /* | |
998 | * set up a private mapping or an anonymous shared mapping | |
999 | */ | |
8feae131 DH |
1000 | static int do_mmap_private(struct vm_area_struct *vma, |
1001 | struct vm_region *region, | |
645d83c5 DH |
1002 | unsigned long len, |
1003 | unsigned long capabilities) | |
1da177e4 | 1004 | { |
dbc8358c | 1005 | unsigned long total, point; |
1da177e4 | 1006 | void *base; |
8feae131 | 1007 | int ret, order; |
1da177e4 LT |
1008 | |
1009 | /* invoke the file's mapping function so that it can keep track of | |
1010 | * shared mappings on devices or memory | |
1011 | * - VM_MAYSHARE will be set if it may attempt to share | |
1012 | */ | |
b4caecd4 | 1013 | if (capabilities & NOMMU_MAP_DIRECT) { |
f74ac015 | 1014 | ret = call_mmap(vma->vm_file, vma); |
dd8632a1 | 1015 | if (ret == 0) { |
1da177e4 | 1016 | /* shouldn't return success if we're not sharing */ |
dd8632a1 PM |
1017 | BUG_ON(!(vma->vm_flags & VM_MAYSHARE)); |
1018 | vma->vm_region->vm_top = vma->vm_region->vm_end; | |
645d83c5 | 1019 | return 0; |
1da177e4 | 1020 | } |
dd8632a1 PM |
1021 | if (ret != -ENOSYS) |
1022 | return ret; | |
1da177e4 LT |
1023 | |
1024 | /* getting an ENOSYS error indicates that direct mmap isn't | |
1025 | * possible (as opposed to tried but failed) so we'll try to | |
1026 | * make a private copy of the data and map that instead */ | |
1027 | } | |
1028 | ||
8feae131 | 1029 | |
1da177e4 LT |
1030 | /* allocate some memory to hold the mapping |
1031 | * - note that this may not return a page-aligned address if the object | |
1032 | * we're allocating is smaller than a page | |
1033 | */ | |
f67d9b15 | 1034 | order = get_order(len); |
8feae131 | 1035 | total = 1 << order; |
f67d9b15 | 1036 | point = len >> PAGE_SHIFT; |
dd8632a1 | 1037 | |
dbc8358c | 1038 | /* we don't want to allocate a power-of-2 sized page set */ |
22cc877b | 1039 | if (sysctl_nr_trim_pages && total - point >= sysctl_nr_trim_pages) |
dbc8358c | 1040 | total = point; |
8feae131 | 1041 | |
da616534 | 1042 | base = alloc_pages_exact(total << PAGE_SHIFT, GFP_KERNEL); |
dbc8358c JK |
1043 | if (!base) |
1044 | goto enomem; | |
1045 | ||
1046 | atomic_long_add(total, &mmap_pages_allocated); | |
1da177e4 | 1047 | |
8feae131 DH |
1048 | region->vm_flags = vma->vm_flags |= VM_MAPPED_COPY; |
1049 | region->vm_start = (unsigned long) base; | |
f67d9b15 | 1050 | region->vm_end = region->vm_start + len; |
dd8632a1 | 1051 | region->vm_top = region->vm_start + (total << PAGE_SHIFT); |
8feae131 DH |
1052 | |
1053 | vma->vm_start = region->vm_start; | |
1054 | vma->vm_end = region->vm_start + len; | |
1da177e4 LT |
1055 | |
1056 | if (vma->vm_file) { | |
1057 | /* read the contents of a file into the copy */ | |
1da177e4 LT |
1058 | loff_t fpos; |
1059 | ||
1060 | fpos = vma->vm_pgoff; | |
1061 | fpos <<= PAGE_SHIFT; | |
1062 | ||
b4bf802a | 1063 | ret = kernel_read(vma->vm_file, base, len, &fpos); |
1da177e4 LT |
1064 | if (ret < 0) |
1065 | goto error_free; | |
1066 | ||
1067 | /* clear the last little bit */ | |
f67d9b15 BL |
1068 | if (ret < len) |
1069 | memset(base + ret, 0, len - ret); | |
1da177e4 | 1070 | |
bfd40eaf KS |
1071 | } else { |
1072 | vma_set_anonymous(vma); | |
1da177e4 LT |
1073 | } |
1074 | ||
1075 | return 0; | |
1076 | ||
1077 | error_free: | |
7223bb4a | 1078 | free_page_series(region->vm_start, region->vm_top); |
8feae131 DH |
1079 | region->vm_start = vma->vm_start = 0; |
1080 | region->vm_end = vma->vm_end = 0; | |
dd8632a1 | 1081 | region->vm_top = 0; |
1da177e4 LT |
1082 | return ret; |
1083 | ||
1084 | enomem: | |
b1de0d13 | 1085 | pr_err("Allocation of length %lu from process %d (%s) failed\n", |
05ae6fa3 | 1086 | len, current->pid, current->comm); |
9af744d7 | 1087 | show_free_areas(0, NULL); |
1da177e4 LT |
1088 | return -ENOMEM; |
1089 | } | |
1090 | ||
1091 | /* | |
1092 | * handle mapping creation for uClinux | |
1093 | */ | |
1fcfd8db ON |
1094 | unsigned long do_mmap(struct file *file, |
1095 | unsigned long addr, | |
1096 | unsigned long len, | |
1097 | unsigned long prot, | |
1098 | unsigned long flags, | |
1099 | vm_flags_t vm_flags, | |
1100 | unsigned long pgoff, | |
897ab3e0 MR |
1101 | unsigned long *populate, |
1102 | struct list_head *uf) | |
1da177e4 | 1103 | { |
8feae131 DH |
1104 | struct vm_area_struct *vma; |
1105 | struct vm_region *region; | |
1da177e4 | 1106 | struct rb_node *rb; |
1fcfd8db | 1107 | unsigned long capabilities, result; |
1da177e4 LT |
1108 | int ret; |
1109 | ||
41badc15 | 1110 | *populate = 0; |
bebeb3d6 | 1111 | |
1da177e4 LT |
1112 | /* decide whether we should attempt the mapping, and if so what sort of |
1113 | * mapping */ | |
1114 | ret = validate_mmap_request(file, addr, len, prot, flags, pgoff, | |
1115 | &capabilities); | |
22cc877b | 1116 | if (ret < 0) |
1da177e4 LT |
1117 | return ret; |
1118 | ||
06aab5a3 DH |
1119 | /* we ignore the address hint */ |
1120 | addr = 0; | |
f67d9b15 | 1121 | len = PAGE_ALIGN(len); |
06aab5a3 | 1122 | |
1da177e4 LT |
1123 | /* we've determined that we can make the mapping, now translate what we |
1124 | * now know into VMA flags */ | |
1fcfd8db | 1125 | vm_flags |= determine_vm_flags(file, prot, flags, capabilities); |
1da177e4 | 1126 | |
8feae131 DH |
1127 | /* we're going to need to record the mapping */ |
1128 | region = kmem_cache_zalloc(vm_region_jar, GFP_KERNEL); | |
1129 | if (!region) | |
1130 | goto error_getting_region; | |
1131 | ||
490fc053 | 1132 | vma = vm_area_alloc(current->mm); |
8feae131 DH |
1133 | if (!vma) |
1134 | goto error_getting_vma; | |
1da177e4 | 1135 | |
1e2ae599 | 1136 | region->vm_usage = 1; |
8feae131 DH |
1137 | region->vm_flags = vm_flags; |
1138 | region->vm_pgoff = pgoff; | |
1139 | ||
8feae131 DH |
1140 | vma->vm_flags = vm_flags; |
1141 | vma->vm_pgoff = pgoff; | |
1da177e4 | 1142 | |
8feae131 | 1143 | if (file) { |
cb0942b8 AV |
1144 | region->vm_file = get_file(file); |
1145 | vma->vm_file = get_file(file); | |
8feae131 DH |
1146 | } |
1147 | ||
1148 | down_write(&nommu_region_sem); | |
1149 | ||
1150 | /* if we want to share, we need to check for regions created by other | |
1da177e4 | 1151 | * mmap() calls that overlap with our proposed mapping |
8feae131 | 1152 | * - we can only share with a superset match on most regular files |
1da177e4 LT |
1153 | * - shared mappings on character devices and memory backed files are |
1154 | * permitted to overlap inexactly as far as we are concerned for in | |
1155 | * these cases, sharing is handled in the driver or filesystem rather | |
1156 | * than here | |
1157 | */ | |
1158 | if (vm_flags & VM_MAYSHARE) { | |
8feae131 DH |
1159 | struct vm_region *pregion; |
1160 | unsigned long pglen, rpglen, pgend, rpgend, start; | |
1da177e4 | 1161 | |
8feae131 DH |
1162 | pglen = (len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
1163 | pgend = pgoff + pglen; | |
165b2392 | 1164 | |
8feae131 DH |
1165 | for (rb = rb_first(&nommu_region_tree); rb; rb = rb_next(rb)) { |
1166 | pregion = rb_entry(rb, struct vm_region, vm_rb); | |
1da177e4 | 1167 | |
8feae131 | 1168 | if (!(pregion->vm_flags & VM_MAYSHARE)) |
1da177e4 LT |
1169 | continue; |
1170 | ||
1171 | /* search for overlapping mappings on the same file */ | |
496ad9aa AV |
1172 | if (file_inode(pregion->vm_file) != |
1173 | file_inode(file)) | |
1da177e4 LT |
1174 | continue; |
1175 | ||
8feae131 | 1176 | if (pregion->vm_pgoff >= pgend) |
1da177e4 LT |
1177 | continue; |
1178 | ||
8feae131 DH |
1179 | rpglen = pregion->vm_end - pregion->vm_start; |
1180 | rpglen = (rpglen + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1181 | rpgend = pregion->vm_pgoff + rpglen; | |
1182 | if (pgoff >= rpgend) | |
1da177e4 LT |
1183 | continue; |
1184 | ||
8feae131 DH |
1185 | /* handle inexactly overlapping matches between |
1186 | * mappings */ | |
1187 | if ((pregion->vm_pgoff != pgoff || rpglen != pglen) && | |
1188 | !(pgoff >= pregion->vm_pgoff && pgend <= rpgend)) { | |
1189 | /* new mapping is not a subset of the region */ | |
b4caecd4 | 1190 | if (!(capabilities & NOMMU_MAP_DIRECT)) |
1da177e4 LT |
1191 | goto sharing_violation; |
1192 | continue; | |
1193 | } | |
1194 | ||
8feae131 | 1195 | /* we've found a region we can share */ |
1e2ae599 | 1196 | pregion->vm_usage++; |
8feae131 DH |
1197 | vma->vm_region = pregion; |
1198 | start = pregion->vm_start; | |
1199 | start += (pgoff - pregion->vm_pgoff) << PAGE_SHIFT; | |
1200 | vma->vm_start = start; | |
1201 | vma->vm_end = start + len; | |
1202 | ||
22cc877b | 1203 | if (pregion->vm_flags & VM_MAPPED_COPY) |
8feae131 | 1204 | vma->vm_flags |= VM_MAPPED_COPY; |
22cc877b | 1205 | else { |
8feae131 DH |
1206 | ret = do_mmap_shared_file(vma); |
1207 | if (ret < 0) { | |
1208 | vma->vm_region = NULL; | |
1209 | vma->vm_start = 0; | |
1210 | vma->vm_end = 0; | |
1e2ae599 | 1211 | pregion->vm_usage--; |
8feae131 DH |
1212 | pregion = NULL; |
1213 | goto error_just_free; | |
1214 | } | |
1215 | } | |
1216 | fput(region->vm_file); | |
1217 | kmem_cache_free(vm_region_jar, region); | |
1218 | region = pregion; | |
1219 | result = start; | |
1220 | goto share; | |
1da177e4 LT |
1221 | } |
1222 | ||
1da177e4 LT |
1223 | /* obtain the address at which to make a shared mapping |
1224 | * - this is the hook for quasi-memory character devices to | |
1225 | * tell us the location of a shared mapping | |
1226 | */ | |
b4caecd4 | 1227 | if (capabilities & NOMMU_MAP_DIRECT) { |
1da177e4 LT |
1228 | addr = file->f_op->get_unmapped_area(file, addr, len, |
1229 | pgoff, flags); | |
bb005a59 | 1230 | if (IS_ERR_VALUE(addr)) { |
1da177e4 | 1231 | ret = addr; |
bb005a59 | 1232 | if (ret != -ENOSYS) |
8feae131 | 1233 | goto error_just_free; |
1da177e4 LT |
1234 | |
1235 | /* the driver refused to tell us where to site | |
1236 | * the mapping so we'll have to attempt to copy | |
1237 | * it */ | |
bb005a59 | 1238 | ret = -ENODEV; |
b4caecd4 | 1239 | if (!(capabilities & NOMMU_MAP_COPY)) |
8feae131 | 1240 | goto error_just_free; |
1da177e4 | 1241 | |
b4caecd4 | 1242 | capabilities &= ~NOMMU_MAP_DIRECT; |
8feae131 DH |
1243 | } else { |
1244 | vma->vm_start = region->vm_start = addr; | |
1245 | vma->vm_end = region->vm_end = addr + len; | |
1da177e4 LT |
1246 | } |
1247 | } | |
1248 | } | |
1249 | ||
8feae131 | 1250 | vma->vm_region = region; |
1da177e4 | 1251 | |
645d83c5 | 1252 | /* set up the mapping |
b4caecd4 | 1253 | * - the region is filled in if NOMMU_MAP_DIRECT is still set |
645d83c5 | 1254 | */ |
1da177e4 | 1255 | if (file && vma->vm_flags & VM_SHARED) |
8feae131 | 1256 | ret = do_mmap_shared_file(vma); |
1da177e4 | 1257 | else |
645d83c5 | 1258 | ret = do_mmap_private(vma, region, len, capabilities); |
1da177e4 | 1259 | if (ret < 0) |
645d83c5 DH |
1260 | goto error_just_free; |
1261 | add_nommu_region(region); | |
8feae131 | 1262 | |
ea637639 | 1263 | /* clear anonymous mappings that don't ask for uninitialized data */ |
0bf5f949 CH |
1264 | if (!vma->vm_file && |
1265 | (!IS_ENABLED(CONFIG_MMAP_ALLOW_UNINITIALIZED) || | |
1266 | !(flags & MAP_UNINITIALIZED))) | |
ea637639 JZ |
1267 | memset((void *)region->vm_start, 0, |
1268 | region->vm_end - region->vm_start); | |
1269 | ||
1da177e4 | 1270 | /* okay... we have a mapping; now we have to register it */ |
8feae131 | 1271 | result = vma->vm_start; |
1da177e4 | 1272 | |
1da177e4 LT |
1273 | current->mm->total_vm += len >> PAGE_SHIFT; |
1274 | ||
8feae131 DH |
1275 | share: |
1276 | add_vma_to_mm(current->mm, vma); | |
1da177e4 | 1277 | |
cfe79c00 MF |
1278 | /* we flush the region from the icache only when the first executable |
1279 | * mapping of it is made */ | |
1280 | if (vma->vm_flags & VM_EXEC && !region->vm_icache_flushed) { | |
1281 | flush_icache_range(region->vm_start, region->vm_end); | |
1282 | region->vm_icache_flushed = true; | |
1283 | } | |
1da177e4 | 1284 | |
cfe79c00 | 1285 | up_write(&nommu_region_sem); |
1da177e4 | 1286 | |
8feae131 | 1287 | return result; |
1da177e4 | 1288 | |
8feae131 DH |
1289 | error_just_free: |
1290 | up_write(&nommu_region_sem); | |
1291 | error: | |
89a86402 DH |
1292 | if (region->vm_file) |
1293 | fput(region->vm_file); | |
8feae131 | 1294 | kmem_cache_free(vm_region_jar, region); |
89a86402 DH |
1295 | if (vma->vm_file) |
1296 | fput(vma->vm_file); | |
3928d4f5 | 1297 | vm_area_free(vma); |
8feae131 DH |
1298 | return ret; |
1299 | ||
1300 | sharing_violation: | |
1301 | up_write(&nommu_region_sem); | |
22cc877b | 1302 | pr_warn("Attempt to share mismatched mappings\n"); |
8feae131 DH |
1303 | ret = -EINVAL; |
1304 | goto error; | |
1da177e4 | 1305 | |
8feae131 DH |
1306 | error_getting_vma: |
1307 | kmem_cache_free(vm_region_jar, region); | |
22cc877b LR |
1308 | pr_warn("Allocation of vma for %lu byte allocation from process %d failed\n", |
1309 | len, current->pid); | |
9af744d7 | 1310 | show_free_areas(0, NULL); |
1da177e4 LT |
1311 | return -ENOMEM; |
1312 | ||
8feae131 | 1313 | error_getting_region: |
22cc877b LR |
1314 | pr_warn("Allocation of vm region for %lu byte allocation from process %d failed\n", |
1315 | len, current->pid); | |
9af744d7 | 1316 | show_free_areas(0, NULL); |
1da177e4 LT |
1317 | return -ENOMEM; |
1318 | } | |
6be5ceb0 | 1319 | |
a90f590a DB |
1320 | unsigned long ksys_mmap_pgoff(unsigned long addr, unsigned long len, |
1321 | unsigned long prot, unsigned long flags, | |
1322 | unsigned long fd, unsigned long pgoff) | |
66f0dc48 HD |
1323 | { |
1324 | struct file *file = NULL; | |
1325 | unsigned long retval = -EBADF; | |
1326 | ||
120a795d | 1327 | audit_mmap_fd(fd, flags); |
66f0dc48 HD |
1328 | if (!(flags & MAP_ANONYMOUS)) { |
1329 | file = fget(fd); | |
1330 | if (!file) | |
1331 | goto out; | |
1332 | } | |
1333 | ||
1334 | flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE); | |
1335 | ||
ad1ed293 | 1336 | retval = vm_mmap_pgoff(file, addr, len, prot, flags, pgoff); |
66f0dc48 HD |
1337 | |
1338 | if (file) | |
1339 | fput(file); | |
1340 | out: | |
1341 | return retval; | |
1342 | } | |
1343 | ||
a90f590a DB |
1344 | SYSCALL_DEFINE6(mmap_pgoff, unsigned long, addr, unsigned long, len, |
1345 | unsigned long, prot, unsigned long, flags, | |
1346 | unsigned long, fd, unsigned long, pgoff) | |
1347 | { | |
1348 | return ksys_mmap_pgoff(addr, len, prot, flags, fd, pgoff); | |
1349 | } | |
1350 | ||
a4679373 CH |
1351 | #ifdef __ARCH_WANT_SYS_OLD_MMAP |
1352 | struct mmap_arg_struct { | |
1353 | unsigned long addr; | |
1354 | unsigned long len; | |
1355 | unsigned long prot; | |
1356 | unsigned long flags; | |
1357 | unsigned long fd; | |
1358 | unsigned long offset; | |
1359 | }; | |
1360 | ||
1361 | SYSCALL_DEFINE1(old_mmap, struct mmap_arg_struct __user *, arg) | |
1362 | { | |
1363 | struct mmap_arg_struct a; | |
1364 | ||
1365 | if (copy_from_user(&a, arg, sizeof(a))) | |
1366 | return -EFAULT; | |
1824cb75 | 1367 | if (offset_in_page(a.offset)) |
a4679373 CH |
1368 | return -EINVAL; |
1369 | ||
a90f590a DB |
1370 | return ksys_mmap_pgoff(a.addr, a.len, a.prot, a.flags, a.fd, |
1371 | a.offset >> PAGE_SHIFT); | |
a4679373 CH |
1372 | } |
1373 | #endif /* __ARCH_WANT_SYS_OLD_MMAP */ | |
1374 | ||
1da177e4 | 1375 | /* |
8feae131 DH |
1376 | * split a vma into two pieces at address 'addr', a new vma is allocated either |
1377 | * for the first part or the tail. | |
1da177e4 | 1378 | */ |
8feae131 DH |
1379 | int split_vma(struct mm_struct *mm, struct vm_area_struct *vma, |
1380 | unsigned long addr, int new_below) | |
1da177e4 | 1381 | { |
8feae131 DH |
1382 | struct vm_area_struct *new; |
1383 | struct vm_region *region; | |
1384 | unsigned long npages; | |
1da177e4 | 1385 | |
779c1023 DH |
1386 | /* we're only permitted to split anonymous regions (these should have |
1387 | * only a single usage on the region) */ | |
1388 | if (vma->vm_file) | |
8feae131 | 1389 | return -ENOMEM; |
1da177e4 | 1390 | |
8feae131 DH |
1391 | if (mm->map_count >= sysctl_max_map_count) |
1392 | return -ENOMEM; | |
1da177e4 | 1393 | |
8feae131 DH |
1394 | region = kmem_cache_alloc(vm_region_jar, GFP_KERNEL); |
1395 | if (!region) | |
1396 | return -ENOMEM; | |
1da177e4 | 1397 | |
3928d4f5 | 1398 | new = vm_area_dup(vma); |
8feae131 DH |
1399 | if (!new) { |
1400 | kmem_cache_free(vm_region_jar, region); | |
1401 | return -ENOMEM; | |
1402 | } | |
1403 | ||
1404 | /* most fields are the same, copy all, and then fixup */ | |
8feae131 DH |
1405 | *region = *vma->vm_region; |
1406 | new->vm_region = region; | |
1407 | ||
1408 | npages = (addr - vma->vm_start) >> PAGE_SHIFT; | |
1409 | ||
1410 | if (new_below) { | |
dd8632a1 | 1411 | region->vm_top = region->vm_end = new->vm_end = addr; |
8feae131 DH |
1412 | } else { |
1413 | region->vm_start = new->vm_start = addr; | |
1414 | region->vm_pgoff = new->vm_pgoff += npages; | |
1da177e4 | 1415 | } |
8feae131 DH |
1416 | |
1417 | if (new->vm_ops && new->vm_ops->open) | |
1418 | new->vm_ops->open(new); | |
1419 | ||
1420 | delete_vma_from_mm(vma); | |
1421 | down_write(&nommu_region_sem); | |
1422 | delete_nommu_region(vma->vm_region); | |
1423 | if (new_below) { | |
1424 | vma->vm_region->vm_start = vma->vm_start = addr; | |
1425 | vma->vm_region->vm_pgoff = vma->vm_pgoff += npages; | |
1426 | } else { | |
1427 | vma->vm_region->vm_end = vma->vm_end = addr; | |
dd8632a1 | 1428 | vma->vm_region->vm_top = addr; |
8feae131 DH |
1429 | } |
1430 | add_nommu_region(vma->vm_region); | |
1431 | add_nommu_region(new->vm_region); | |
1432 | up_write(&nommu_region_sem); | |
1433 | add_vma_to_mm(mm, vma); | |
1434 | add_vma_to_mm(mm, new); | |
1435 | return 0; | |
1da177e4 LT |
1436 | } |
1437 | ||
3034097a | 1438 | /* |
8feae131 DH |
1439 | * shrink a VMA by removing the specified chunk from either the beginning or |
1440 | * the end | |
3034097a | 1441 | */ |
8feae131 DH |
1442 | static int shrink_vma(struct mm_struct *mm, |
1443 | struct vm_area_struct *vma, | |
1444 | unsigned long from, unsigned long to) | |
1da177e4 | 1445 | { |
8feae131 | 1446 | struct vm_region *region; |
1da177e4 | 1447 | |
8feae131 DH |
1448 | /* adjust the VMA's pointers, which may reposition it in the MM's tree |
1449 | * and list */ | |
1450 | delete_vma_from_mm(vma); | |
1451 | if (from > vma->vm_start) | |
1452 | vma->vm_end = from; | |
1453 | else | |
1454 | vma->vm_start = to; | |
1455 | add_vma_to_mm(mm, vma); | |
1da177e4 | 1456 | |
8feae131 DH |
1457 | /* cut the backing region down to size */ |
1458 | region = vma->vm_region; | |
1e2ae599 | 1459 | BUG_ON(region->vm_usage != 1); |
8feae131 DH |
1460 | |
1461 | down_write(&nommu_region_sem); | |
1462 | delete_nommu_region(region); | |
dd8632a1 PM |
1463 | if (from > region->vm_start) { |
1464 | to = region->vm_top; | |
1465 | region->vm_top = region->vm_end = from; | |
1466 | } else { | |
8feae131 | 1467 | region->vm_start = to; |
dd8632a1 | 1468 | } |
8feae131 DH |
1469 | add_nommu_region(region); |
1470 | up_write(&nommu_region_sem); | |
1471 | ||
1472 | free_page_series(from, to); | |
1473 | return 0; | |
1474 | } | |
1da177e4 | 1475 | |
8feae131 DH |
1476 | /* |
1477 | * release a mapping | |
1478 | * - under NOMMU conditions the chunk to be unmapped must be backed by a single | |
1479 | * VMA, though it need not cover the whole VMA | |
1480 | */ | |
897ab3e0 | 1481 | int do_munmap(struct mm_struct *mm, unsigned long start, size_t len, struct list_head *uf) |
8feae131 DH |
1482 | { |
1483 | struct vm_area_struct *vma; | |
f67d9b15 | 1484 | unsigned long end; |
8feae131 | 1485 | int ret; |
1da177e4 | 1486 | |
f67d9b15 | 1487 | len = PAGE_ALIGN(len); |
8feae131 DH |
1488 | if (len == 0) |
1489 | return -EINVAL; | |
365e9c87 | 1490 | |
f67d9b15 BL |
1491 | end = start + len; |
1492 | ||
8feae131 DH |
1493 | /* find the first potentially overlapping VMA */ |
1494 | vma = find_vma(mm, start); | |
1495 | if (!vma) { | |
ac714904 | 1496 | static int limit; |
33e5d769 | 1497 | if (limit < 5) { |
22cc877b LR |
1498 | pr_warn("munmap of memory not mmapped by process %d (%s): 0x%lx-0x%lx\n", |
1499 | current->pid, current->comm, | |
1500 | start, start + len - 1); | |
33e5d769 DH |
1501 | limit++; |
1502 | } | |
8feae131 DH |
1503 | return -EINVAL; |
1504 | } | |
1da177e4 | 1505 | |
8feae131 DH |
1506 | /* we're allowed to split an anonymous VMA but not a file-backed one */ |
1507 | if (vma->vm_file) { | |
1508 | do { | |
22cc877b | 1509 | if (start > vma->vm_start) |
8feae131 | 1510 | return -EINVAL; |
8feae131 DH |
1511 | if (end == vma->vm_end) |
1512 | goto erase_whole_vma; | |
d75a310c NK |
1513 | vma = vma->vm_next; |
1514 | } while (vma); | |
8feae131 DH |
1515 | return -EINVAL; |
1516 | } else { | |
1517 | /* the chunk must be a subset of the VMA found */ | |
1518 | if (start == vma->vm_start && end == vma->vm_end) | |
1519 | goto erase_whole_vma; | |
22cc877b | 1520 | if (start < vma->vm_start || end > vma->vm_end) |
8feae131 | 1521 | return -EINVAL; |
1824cb75 | 1522 | if (offset_in_page(start)) |
8feae131 | 1523 | return -EINVAL; |
1824cb75 | 1524 | if (end != vma->vm_end && offset_in_page(end)) |
8feae131 | 1525 | return -EINVAL; |
8feae131 DH |
1526 | if (start != vma->vm_start && end != vma->vm_end) { |
1527 | ret = split_vma(mm, vma, start, 1); | |
22cc877b | 1528 | if (ret < 0) |
8feae131 | 1529 | return ret; |
8feae131 DH |
1530 | } |
1531 | return shrink_vma(mm, vma, start, end); | |
1532 | } | |
1da177e4 | 1533 | |
8feae131 DH |
1534 | erase_whole_vma: |
1535 | delete_vma_from_mm(vma); | |
1536 | delete_vma(mm, vma); | |
1da177e4 LT |
1537 | return 0; |
1538 | } | |
b5073173 | 1539 | EXPORT_SYMBOL(do_munmap); |
1da177e4 | 1540 | |
bfce281c | 1541 | int vm_munmap(unsigned long addr, size_t len) |
3034097a | 1542 | { |
bfce281c | 1543 | struct mm_struct *mm = current->mm; |
3034097a | 1544 | int ret; |
3034097a DH |
1545 | |
1546 | down_write(&mm->mmap_sem); | |
897ab3e0 | 1547 | ret = do_munmap(mm, addr, len, NULL); |
3034097a DH |
1548 | up_write(&mm->mmap_sem); |
1549 | return ret; | |
1550 | } | |
a46ef99d LT |
1551 | EXPORT_SYMBOL(vm_munmap); |
1552 | ||
1553 | SYSCALL_DEFINE2(munmap, unsigned long, addr, size_t, len) | |
1554 | { | |
bfce281c | 1555 | return vm_munmap(addr, len); |
a46ef99d | 1556 | } |
3034097a DH |
1557 | |
1558 | /* | |
8feae131 | 1559 | * release all the mappings made in a process's VM space |
3034097a | 1560 | */ |
8feae131 | 1561 | void exit_mmap(struct mm_struct *mm) |
1da177e4 | 1562 | { |
8feae131 | 1563 | struct vm_area_struct *vma; |
1da177e4 | 1564 | |
8feae131 DH |
1565 | if (!mm) |
1566 | return; | |
1da177e4 | 1567 | |
8feae131 | 1568 | mm->total_vm = 0; |
1da177e4 | 1569 | |
8feae131 DH |
1570 | while ((vma = mm->mmap)) { |
1571 | mm->mmap = vma->vm_next; | |
1572 | delete_vma_from_mm(vma); | |
1573 | delete_vma(mm, vma); | |
04c34961 | 1574 | cond_resched(); |
1da177e4 LT |
1575 | } |
1576 | } | |
1577 | ||
5d22fc25 | 1578 | int vm_brk(unsigned long addr, unsigned long len) |
1da177e4 LT |
1579 | { |
1580 | return -ENOMEM; | |
1581 | } | |
1582 | ||
1583 | /* | |
6fa5f80b DH |
1584 | * expand (or shrink) an existing mapping, potentially moving it at the same |
1585 | * time (controlled by the MREMAP_MAYMOVE flag and available VM space) | |
1da177e4 | 1586 | * |
6fa5f80b | 1587 | * under NOMMU conditions, we only permit changing a mapping's size, and only |
8feae131 DH |
1588 | * as long as it stays within the region allocated by do_mmap_private() and the |
1589 | * block is not shareable | |
1da177e4 | 1590 | * |
6fa5f80b | 1591 | * MREMAP_FIXED is not supported under NOMMU conditions |
1da177e4 | 1592 | */ |
4b377bab | 1593 | static unsigned long do_mremap(unsigned long addr, |
1da177e4 LT |
1594 | unsigned long old_len, unsigned long new_len, |
1595 | unsigned long flags, unsigned long new_addr) | |
1596 | { | |
6fa5f80b | 1597 | struct vm_area_struct *vma; |
1da177e4 LT |
1598 | |
1599 | /* insanity checks first */ | |
f67d9b15 BL |
1600 | old_len = PAGE_ALIGN(old_len); |
1601 | new_len = PAGE_ALIGN(new_len); | |
8feae131 | 1602 | if (old_len == 0 || new_len == 0) |
1da177e4 LT |
1603 | return (unsigned long) -EINVAL; |
1604 | ||
1824cb75 | 1605 | if (offset_in_page(addr)) |
8feae131 DH |
1606 | return -EINVAL; |
1607 | ||
1da177e4 LT |
1608 | if (flags & MREMAP_FIXED && new_addr != addr) |
1609 | return (unsigned long) -EINVAL; | |
1610 | ||
8feae131 | 1611 | vma = find_vma_exact(current->mm, addr, old_len); |
6fa5f80b DH |
1612 | if (!vma) |
1613 | return (unsigned long) -EINVAL; | |
1da177e4 | 1614 | |
6fa5f80b | 1615 | if (vma->vm_end != vma->vm_start + old_len) |
1da177e4 LT |
1616 | return (unsigned long) -EFAULT; |
1617 | ||
6fa5f80b | 1618 | if (vma->vm_flags & VM_MAYSHARE) |
1da177e4 LT |
1619 | return (unsigned long) -EPERM; |
1620 | ||
8feae131 | 1621 | if (new_len > vma->vm_region->vm_end - vma->vm_region->vm_start) |
1da177e4 LT |
1622 | return (unsigned long) -ENOMEM; |
1623 | ||
1624 | /* all checks complete - do it */ | |
6fa5f80b | 1625 | vma->vm_end = vma->vm_start + new_len; |
6fa5f80b DH |
1626 | return vma->vm_start; |
1627 | } | |
1628 | ||
6a6160a7 HC |
1629 | SYSCALL_DEFINE5(mremap, unsigned long, addr, unsigned long, old_len, |
1630 | unsigned long, new_len, unsigned long, flags, | |
1631 | unsigned long, new_addr) | |
6fa5f80b DH |
1632 | { |
1633 | unsigned long ret; | |
1634 | ||
1635 | down_write(¤t->mm->mmap_sem); | |
1636 | ret = do_mremap(addr, old_len, new_len, flags, new_addr); | |
1637 | up_write(¤t->mm->mmap_sem); | |
1638 | return ret; | |
1da177e4 LT |
1639 | } |
1640 | ||
df06b37f KB |
1641 | struct page *follow_page(struct vm_area_struct *vma, unsigned long address, |
1642 | unsigned int foll_flags) | |
1da177e4 LT |
1643 | { |
1644 | return NULL; | |
1645 | } | |
1646 | ||
8f3b1327 BL |
1647 | int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr, |
1648 | unsigned long pfn, unsigned long size, pgprot_t prot) | |
1da177e4 | 1649 | { |
8f3b1327 BL |
1650 | if (addr != (pfn << PAGE_SHIFT)) |
1651 | return -EINVAL; | |
1652 | ||
314e51b9 | 1653 | vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP; |
66aa2b4b | 1654 | return 0; |
1da177e4 | 1655 | } |
22c4af40 | 1656 | EXPORT_SYMBOL(remap_pfn_range); |
1da177e4 | 1657 | |
3c0b9de6 LT |
1658 | int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len) |
1659 | { | |
1660 | unsigned long pfn = start >> PAGE_SHIFT; | |
1661 | unsigned long vm_len = vma->vm_end - vma->vm_start; | |
1662 | ||
1663 | pfn += vma->vm_pgoff; | |
1664 | return io_remap_pfn_range(vma, vma->vm_start, pfn, vm_len, vma->vm_page_prot); | |
1665 | } | |
1666 | EXPORT_SYMBOL(vm_iomap_memory); | |
1667 | ||
f905bc44 PM |
1668 | int remap_vmalloc_range(struct vm_area_struct *vma, void *addr, |
1669 | unsigned long pgoff) | |
1670 | { | |
1671 | unsigned int size = vma->vm_end - vma->vm_start; | |
1672 | ||
1673 | if (!(vma->vm_flags & VM_USERMAP)) | |
1674 | return -EINVAL; | |
1675 | ||
1676 | vma->vm_start = (unsigned long)(addr + (pgoff << PAGE_SHIFT)); | |
1677 | vma->vm_end = vma->vm_start + size; | |
1678 | ||
1679 | return 0; | |
1680 | } | |
1681 | EXPORT_SYMBOL(remap_vmalloc_range); | |
1682 | ||
1da177e4 LT |
1683 | unsigned long arch_get_unmapped_area(struct file *file, unsigned long addr, |
1684 | unsigned long len, unsigned long pgoff, unsigned long flags) | |
1685 | { | |
1686 | return -ENOMEM; | |
1687 | } | |
1688 | ||
2bcd6454 | 1689 | vm_fault_t filemap_fault(struct vm_fault *vmf) |
b0e15190 DH |
1690 | { |
1691 | BUG(); | |
d0217ac0 | 1692 | return 0; |
b0e15190 | 1693 | } |
b5073173 | 1694 | EXPORT_SYMBOL(filemap_fault); |
0ec76a11 | 1695 | |
82b0f8c3 | 1696 | void filemap_map_pages(struct vm_fault *vmf, |
bae473a4 | 1697 | pgoff_t start_pgoff, pgoff_t end_pgoff) |
f1820361 KS |
1698 | { |
1699 | BUG(); | |
1700 | } | |
1701 | EXPORT_SYMBOL(filemap_map_pages); | |
1702 | ||
84d77d3f | 1703 | int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm, |
442486ec | 1704 | unsigned long addr, void *buf, int len, unsigned int gup_flags) |
0ec76a11 | 1705 | { |
0ec76a11 | 1706 | struct vm_area_struct *vma; |
442486ec | 1707 | int write = gup_flags & FOLL_WRITE; |
0ec76a11 | 1708 | |
1e426fe2 KK |
1709 | if (down_read_killable(&mm->mmap_sem)) |
1710 | return 0; | |
0ec76a11 DH |
1711 | |
1712 | /* the access must start within one of the target process's mappings */ | |
0159b141 DH |
1713 | vma = find_vma(mm, addr); |
1714 | if (vma) { | |
0ec76a11 DH |
1715 | /* don't overrun this mapping */ |
1716 | if (addr + len >= vma->vm_end) | |
1717 | len = vma->vm_end - addr; | |
1718 | ||
1719 | /* only read or write mappings where it is permitted */ | |
d00c7b99 | 1720 | if (write && vma->vm_flags & VM_MAYWRITE) |
7959722b JZ |
1721 | copy_to_user_page(vma, NULL, addr, |
1722 | (void *) addr, buf, len); | |
d00c7b99 | 1723 | else if (!write && vma->vm_flags & VM_MAYREAD) |
7959722b JZ |
1724 | copy_from_user_page(vma, NULL, addr, |
1725 | buf, (void *) addr, len); | |
0ec76a11 DH |
1726 | else |
1727 | len = 0; | |
1728 | } else { | |
1729 | len = 0; | |
1730 | } | |
1731 | ||
1732 | up_read(&mm->mmap_sem); | |
f55f199b MF |
1733 | |
1734 | return len; | |
1735 | } | |
1736 | ||
1737 | /** | |
b7701a5f | 1738 | * access_remote_vm - access another process' address space |
f55f199b MF |
1739 | * @mm: the mm_struct of the target address space |
1740 | * @addr: start address to access | |
1741 | * @buf: source or destination buffer | |
1742 | * @len: number of bytes to transfer | |
6347e8d5 | 1743 | * @gup_flags: flags modifying lookup behaviour |
f55f199b MF |
1744 | * |
1745 | * The caller must hold a reference on @mm. | |
1746 | */ | |
1747 | int access_remote_vm(struct mm_struct *mm, unsigned long addr, | |
6347e8d5 | 1748 | void *buf, int len, unsigned int gup_flags) |
f55f199b | 1749 | { |
6347e8d5 | 1750 | return __access_remote_vm(NULL, mm, addr, buf, len, gup_flags); |
f55f199b MF |
1751 | } |
1752 | ||
1753 | /* | |
1754 | * Access another process' address space. | |
1755 | * - source/target buffer must be kernel space | |
1756 | */ | |
f307ab6d LS |
1757 | int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, |
1758 | unsigned int gup_flags) | |
f55f199b MF |
1759 | { |
1760 | struct mm_struct *mm; | |
1761 | ||
1762 | if (addr + len < addr) | |
1763 | return 0; | |
1764 | ||
1765 | mm = get_task_mm(tsk); | |
1766 | if (!mm) | |
1767 | return 0; | |
1768 | ||
f307ab6d | 1769 | len = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags); |
f55f199b | 1770 | |
0ec76a11 DH |
1771 | mmput(mm); |
1772 | return len; | |
1773 | } | |
fcd35857 | 1774 | EXPORT_SYMBOL_GPL(access_process_vm); |
7e660872 DH |
1775 | |
1776 | /** | |
1777 | * nommu_shrink_inode_mappings - Shrink the shared mappings on an inode | |
1778 | * @inode: The inode to check | |
1779 | * @size: The current filesize of the inode | |
1780 | * @newsize: The proposed filesize of the inode | |
1781 | * | |
1782 | * Check the shared mappings on an inode on behalf of a shrinking truncate to | |
1783 | * make sure that that any outstanding VMAs aren't broken and then shrink the | |
1784 | * vm_regions that extend that beyond so that do_mmap_pgoff() doesn't | |
1785 | * automatically grant mappings that are too large. | |
1786 | */ | |
1787 | int nommu_shrink_inode_mappings(struct inode *inode, size_t size, | |
1788 | size_t newsize) | |
1789 | { | |
1790 | struct vm_area_struct *vma; | |
7e660872 DH |
1791 | struct vm_region *region; |
1792 | pgoff_t low, high; | |
1793 | size_t r_size, r_top; | |
1794 | ||
1795 | low = newsize >> PAGE_SHIFT; | |
1796 | high = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1797 | ||
1798 | down_write(&nommu_region_sem); | |
1acf2e04 | 1799 | i_mmap_lock_read(inode->i_mapping); |
7e660872 DH |
1800 | |
1801 | /* search for VMAs that fall within the dead zone */ | |
6b2dbba8 | 1802 | vma_interval_tree_foreach(vma, &inode->i_mapping->i_mmap, low, high) { |
7e660872 DH |
1803 | /* found one - only interested if it's shared out of the page |
1804 | * cache */ | |
1805 | if (vma->vm_flags & VM_SHARED) { | |
1acf2e04 | 1806 | i_mmap_unlock_read(inode->i_mapping); |
7e660872 DH |
1807 | up_write(&nommu_region_sem); |
1808 | return -ETXTBSY; /* not quite true, but near enough */ | |
1809 | } | |
1810 | } | |
1811 | ||
1812 | /* reduce any regions that overlap the dead zone - if in existence, | |
1813 | * these will be pointed to by VMAs that don't overlap the dead zone | |
1814 | * | |
1815 | * we don't check for any regions that start beyond the EOF as there | |
1816 | * shouldn't be any | |
1817 | */ | |
1acf2e04 | 1818 | vma_interval_tree_foreach(vma, &inode->i_mapping->i_mmap, 0, ULONG_MAX) { |
7e660872 DH |
1819 | if (!(vma->vm_flags & VM_SHARED)) |
1820 | continue; | |
1821 | ||
1822 | region = vma->vm_region; | |
1823 | r_size = region->vm_top - region->vm_start; | |
1824 | r_top = (region->vm_pgoff << PAGE_SHIFT) + r_size; | |
1825 | ||
1826 | if (r_top > newsize) { | |
1827 | region->vm_top -= r_top - newsize; | |
1828 | if (region->vm_end > region->vm_top) | |
1829 | region->vm_end = region->vm_top; | |
1830 | } | |
1831 | } | |
1832 | ||
1acf2e04 | 1833 | i_mmap_unlock_read(inode->i_mapping); |
7e660872 DH |
1834 | up_write(&nommu_region_sem); |
1835 | return 0; | |
1836 | } | |
c9b1d098 AS |
1837 | |
1838 | /* | |
1839 | * Initialise sysctl_user_reserve_kbytes. | |
1840 | * | |
1841 | * This is intended to prevent a user from starting a single memory hogging | |
1842 | * process, such that they cannot recover (kill the hog) in OVERCOMMIT_NEVER | |
1843 | * mode. | |
1844 | * | |
1845 | * The default value is min(3% of free memory, 128MB) | |
1846 | * 128MB is enough to recover with sshd/login, bash, and top/kill. | |
1847 | */ | |
1848 | static int __meminit init_user_reserve(void) | |
1849 | { | |
1850 | unsigned long free_kbytes; | |
1851 | ||
c41f012a | 1852 | free_kbytes = global_zone_page_state(NR_FREE_PAGES) << (PAGE_SHIFT - 10); |
c9b1d098 AS |
1853 | |
1854 | sysctl_user_reserve_kbytes = min(free_kbytes / 32, 1UL << 17); | |
1855 | return 0; | |
1856 | } | |
a4bc6fc7 | 1857 | subsys_initcall(init_user_reserve); |
4eeab4f5 AS |
1858 | |
1859 | /* | |
1860 | * Initialise sysctl_admin_reserve_kbytes. | |
1861 | * | |
1862 | * The purpose of sysctl_admin_reserve_kbytes is to allow the sys admin | |
1863 | * to log in and kill a memory hogging process. | |
1864 | * | |
1865 | * Systems with more than 256MB will reserve 8MB, enough to recover | |
1866 | * with sshd, bash, and top in OVERCOMMIT_GUESS. Smaller systems will | |
1867 | * only reserve 3% of free pages by default. | |
1868 | */ | |
1869 | static int __meminit init_admin_reserve(void) | |
1870 | { | |
1871 | unsigned long free_kbytes; | |
1872 | ||
c41f012a | 1873 | free_kbytes = global_zone_page_state(NR_FREE_PAGES) << (PAGE_SHIFT - 10); |
4eeab4f5 AS |
1874 | |
1875 | sysctl_admin_reserve_kbytes = min(free_kbytes / 32, 1UL << 13); | |
1876 | return 0; | |
1877 | } | |
a4bc6fc7 | 1878 | subsys_initcall(init_admin_reserve); |