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).
7 * See Documentation/nommu-mmap.txt
9 * Copyright (c) 2004-2005 David Howells <dhowells@redhat.com>
10 * Copyright (c) 2000-2003 David McCullough <davidm@snapgear.com>
11 * Copyright (c) 2000-2001 D Jeff Dionne <jeff@uClinux.org>
12 * Copyright (c) 2002 Greg Ungerer <gerg@snapgear.com>
16 #include <linux/mman.h>
17 #include <linux/swap.h>
18 #include <linux/file.h>
19 #include <linux/highmem.h>
20 #include <linux/pagemap.h>
21 #include <linux/slab.h>
22 #include <linux/vmalloc.h>
23 #include <linux/ptrace.h>
24 #include <linux/blkdev.h>
25 #include <linux/backing-dev.h>
26 #include <linux/mount.h>
27 #include <linux/personality.h>
28 #include <linux/security.h>
29 #include <linux/syscalls.h>
31 #include <asm/uaccess.h>
33 #include <asm/tlbflush.h>
37 unsigned long max_mapnr
;
38 unsigned long num_physpages
;
39 unsigned long askedalloc
, realalloc
;
40 atomic_t vm_committed_space
= ATOMIC_INIT(0);
41 int sysctl_overcommit_memory
= OVERCOMMIT_GUESS
; /* heuristic overcommit */
42 int sysctl_overcommit_ratio
= 50; /* default is 50% */
43 int sysctl_max_map_count
= DEFAULT_MAX_MAP_COUNT
;
44 int heap_stack_gap
= 0;
46 EXPORT_SYMBOL(mem_map
);
47 EXPORT_SYMBOL(__vm_enough_memory
);
49 /* list of shareable VMAs */
50 struct rb_root nommu_vma_tree
= RB_ROOT
;
51 DECLARE_RWSEM(nommu_vma_sem
);
53 struct vm_operations_struct generic_file_vm_ops
= {
57 EXPORT_SYMBOL(vmalloc_to_page
);
58 EXPORT_SYMBOL(vmalloc_32
);
60 EXPORT_SYMBOL(vunmap
);
63 * Handle all mappings that got truncated by a "truncate()"
66 * NOTE! We have to be ready to update the memory sharing
67 * between the file and the memory map for a potential last
68 * incomplete page. Ugly, but necessary.
70 int vmtruncate(struct inode
*inode
, loff_t offset
)
72 struct address_space
*mapping
= inode
->i_mapping
;
75 if (inode
->i_size
< offset
)
77 i_size_write(inode
, offset
);
79 truncate_inode_pages(mapping
, offset
);
83 limit
= current
->signal
->rlim
[RLIMIT_FSIZE
].rlim_cur
;
84 if (limit
!= RLIM_INFINITY
&& offset
> limit
)
86 if (offset
> inode
->i_sb
->s_maxbytes
)
88 i_size_write(inode
, offset
);
91 if (inode
->i_op
&& inode
->i_op
->truncate
)
92 inode
->i_op
->truncate(inode
);
95 send_sig(SIGXFSZ
, current
, 0);
100 EXPORT_SYMBOL(vmtruncate
);
103 * Return the total memory allocated for this pointer, not
104 * just what the caller asked for.
106 * Doesn't have to be accurate, i.e. may have races.
108 unsigned int kobjsize(const void *objp
)
112 if (!objp
|| !((page
= virt_to_page(objp
))))
118 BUG_ON(page
->index
< 0);
119 BUG_ON(page
->index
>= MAX_ORDER
);
121 return (PAGE_SIZE
<< page
->index
);
125 * The nommu dodgy version :-)
127 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
128 unsigned long start
, int len
, int write
, int force
,
129 struct page
**pages
, struct vm_area_struct
**vmas
)
132 static struct vm_area_struct dummy_vma
;
134 for (i
= 0; i
< len
; i
++) {
136 pages
[i
] = virt_to_page(start
);
138 page_cache_get(pages
[i
]);
141 vmas
[i
] = &dummy_vma
;
147 EXPORT_SYMBOL(get_user_pages
);
149 DEFINE_RWLOCK(vmlist_lock
);
150 struct vm_struct
*vmlist
;
152 void vfree(void *addr
)
157 void *__vmalloc(unsigned long size
, gfp_t gfp_mask
, pgprot_t prot
)
160 * kmalloc doesn't like __GFP_HIGHMEM for some reason
162 return kmalloc(size
, (gfp_mask
| __GFP_COMP
) & ~__GFP_HIGHMEM
);
165 struct page
* vmalloc_to_page(void *addr
)
167 return virt_to_page(addr
);
170 unsigned long vmalloc_to_pfn(void *addr
)
172 return page_to_pfn(virt_to_page(addr
));
176 long vread(char *buf
, char *addr
, unsigned long count
)
178 memcpy(buf
, addr
, count
);
182 long vwrite(char *buf
, char *addr
, unsigned long count
)
184 /* Don't allow overflow */
185 if ((unsigned long) addr
+ count
< count
)
186 count
= -(unsigned long) addr
;
188 memcpy(addr
, buf
, count
);
193 * vmalloc - allocate virtually continguos memory
195 * @size: allocation size
197 * Allocate enough pages to cover @size from the page level
198 * allocator and map them into continguos kernel virtual space.
200 * For tight cotrol over page level allocator and protection flags
201 * use __vmalloc() instead.
203 void *vmalloc(unsigned long size
)
205 return __vmalloc(size
, GFP_KERNEL
| __GFP_HIGHMEM
, PAGE_KERNEL
);
207 EXPORT_SYMBOL(vmalloc
);
209 void *vmalloc_node(unsigned long size
, int node
)
211 return vmalloc(size
);
213 EXPORT_SYMBOL(vmalloc_node
);
216 * vmalloc_32 - allocate virtually continguos memory (32bit addressable)
218 * @size: allocation size
220 * Allocate enough 32bit PA addressable pages to cover @size from the
221 * page level allocator and map them into continguos kernel virtual space.
223 void *vmalloc_32(unsigned long size
)
225 return __vmalloc(size
, GFP_KERNEL
, PAGE_KERNEL
);
228 void *vmap(struct page
**pages
, unsigned int count
, unsigned long flags
, pgprot_t prot
)
234 void vunmap(void *addr
)
240 * sys_brk() for the most part doesn't need the global kernel
241 * lock, except when an application is doing something nasty
242 * like trying to un-brk an area that has already been mapped
243 * to a regular file. in this case, the unmapping will need
244 * to invoke file system routines that need the global lock.
246 asmlinkage
unsigned long sys_brk(unsigned long brk
)
248 struct mm_struct
*mm
= current
->mm
;
250 if (brk
< mm
->start_brk
|| brk
> mm
->context
.end_brk
)
257 * Always allow shrinking brk
259 if (brk
<= mm
->brk
) {
265 * Ok, looks good - let it rip.
267 return mm
->brk
= brk
;
271 static void show_process_blocks(void)
273 struct vm_list_struct
*vml
;
275 printk("Process blocks %d:", current
->pid
);
277 for (vml
= ¤t
->mm
->context
.vmlist
; vml
; vml
= vml
->next
) {
278 printk(" %p: %p", vml
, vml
->vma
);
280 printk(" (%d @%lx #%d)",
281 kobjsize((void *) vml
->vma
->vm_start
),
283 atomic_read(&vml
->vma
->vm_usage
));
284 printk(vml
->next
? " ->" : ".\n");
289 static inline struct vm_area_struct
*find_nommu_vma(unsigned long start
)
291 struct vm_area_struct
*vma
;
292 struct rb_node
*n
= nommu_vma_tree
.rb_node
;
295 vma
= rb_entry(n
, struct vm_area_struct
, vm_rb
);
297 if (start
< vma
->vm_start
)
299 else if (start
> vma
->vm_start
)
308 static void add_nommu_vma(struct vm_area_struct
*vma
)
310 struct vm_area_struct
*pvma
;
311 struct address_space
*mapping
;
312 struct rb_node
**p
= &nommu_vma_tree
.rb_node
;
313 struct rb_node
*parent
= NULL
;
315 /* add the VMA to the mapping */
317 mapping
= vma
->vm_file
->f_mapping
;
319 flush_dcache_mmap_lock(mapping
);
320 vma_prio_tree_insert(vma
, &mapping
->i_mmap
);
321 flush_dcache_mmap_unlock(mapping
);
324 /* add the VMA to the master list */
327 pvma
= rb_entry(parent
, struct vm_area_struct
, vm_rb
);
329 if (vma
->vm_start
< pvma
->vm_start
) {
332 else if (vma
->vm_start
> pvma
->vm_start
) {
336 /* mappings are at the same address - this can only
337 * happen for shared-mem chardevs and shared file
338 * mappings backed by ramfs/tmpfs */
339 BUG_ON(!(pvma
->vm_flags
& VM_SHARED
));
350 rb_link_node(&vma
->vm_rb
, parent
, p
);
351 rb_insert_color(&vma
->vm_rb
, &nommu_vma_tree
);
354 static void delete_nommu_vma(struct vm_area_struct
*vma
)
356 struct address_space
*mapping
;
358 /* remove the VMA from the mapping */
360 mapping
= vma
->vm_file
->f_mapping
;
362 flush_dcache_mmap_lock(mapping
);
363 vma_prio_tree_remove(vma
, &mapping
->i_mmap
);
364 flush_dcache_mmap_unlock(mapping
);
367 /* remove from the master list */
368 rb_erase(&vma
->vm_rb
, &nommu_vma_tree
);
372 * determine whether a mapping should be permitted and, if so, what sort of
373 * mapping we're capable of supporting
375 static int validate_mmap_request(struct file
*file
,
381 unsigned long *_capabilities
)
383 unsigned long capabilities
;
384 unsigned long reqprot
= prot
;
387 /* do the simple checks first */
388 if (flags
& MAP_FIXED
|| addr
) {
390 "%d: Can't do fixed-address/overlay mmap of RAM\n",
395 if ((flags
& MAP_TYPE
) != MAP_PRIVATE
&&
396 (flags
& MAP_TYPE
) != MAP_SHARED
)
399 if (PAGE_ALIGN(len
) == 0)
405 /* offset overflow? */
406 if ((pgoff
+ (len
>> PAGE_SHIFT
)) < pgoff
)
410 /* validate file mapping requests */
411 struct address_space
*mapping
;
413 /* files must support mmap */
414 if (!file
->f_op
|| !file
->f_op
->mmap
)
417 /* work out if what we've got could possibly be shared
418 * - we support chardevs that provide their own "memory"
419 * - we support files/blockdevs that are memory backed
421 mapping
= file
->f_mapping
;
423 mapping
= file
->f_dentry
->d_inode
->i_mapping
;
426 if (mapping
&& mapping
->backing_dev_info
)
427 capabilities
= mapping
->backing_dev_info
->capabilities
;
430 /* no explicit capabilities set, so assume some
432 switch (file
->f_dentry
->d_inode
->i_mode
& S_IFMT
) {
435 capabilities
= BDI_CAP_MAP_COPY
;
450 /* eliminate any capabilities that we can't support on this
452 if (!file
->f_op
->get_unmapped_area
)
453 capabilities
&= ~BDI_CAP_MAP_DIRECT
;
454 if (!file
->f_op
->read
)
455 capabilities
&= ~BDI_CAP_MAP_COPY
;
457 if (flags
& MAP_SHARED
) {
458 /* do checks for writing, appending and locking */
459 if ((prot
& PROT_WRITE
) &&
460 !(file
->f_mode
& FMODE_WRITE
))
463 if (IS_APPEND(file
->f_dentry
->d_inode
) &&
464 (file
->f_mode
& FMODE_WRITE
))
467 if (locks_verify_locked(file
->f_dentry
->d_inode
))
470 if (!(capabilities
& BDI_CAP_MAP_DIRECT
))
473 if (((prot
& PROT_READ
) && !(capabilities
& BDI_CAP_READ_MAP
)) ||
474 ((prot
& PROT_WRITE
) && !(capabilities
& BDI_CAP_WRITE_MAP
)) ||
475 ((prot
& PROT_EXEC
) && !(capabilities
& BDI_CAP_EXEC_MAP
))
477 printk("MAP_SHARED not completely supported on !MMU\n");
481 /* we mustn't privatise shared mappings */
482 capabilities
&= ~BDI_CAP_MAP_COPY
;
485 /* we're going to read the file into private memory we
487 if (!(capabilities
& BDI_CAP_MAP_COPY
))
490 /* we don't permit a private writable mapping to be
491 * shared with the backing device */
492 if (prot
& PROT_WRITE
)
493 capabilities
&= ~BDI_CAP_MAP_DIRECT
;
496 /* handle executable mappings and implied executable
498 if (file
->f_vfsmnt
->mnt_flags
& MNT_NOEXEC
) {
499 if (prot
& PROT_EXEC
)
502 else if ((prot
& PROT_READ
) && !(prot
& PROT_EXEC
)) {
503 /* handle implication of PROT_EXEC by PROT_READ */
504 if (current
->personality
& READ_IMPLIES_EXEC
) {
505 if (capabilities
& BDI_CAP_EXEC_MAP
)
509 else if ((prot
& PROT_READ
) &&
510 (prot
& PROT_EXEC
) &&
511 !(capabilities
& BDI_CAP_EXEC_MAP
)
513 /* backing file is not executable, try to copy */
514 capabilities
&= ~BDI_CAP_MAP_DIRECT
;
518 /* anonymous mappings are always memory backed and can be
521 capabilities
= BDI_CAP_MAP_COPY
;
523 /* handle PROT_EXEC implication by PROT_READ */
524 if ((prot
& PROT_READ
) &&
525 (current
->personality
& READ_IMPLIES_EXEC
))
529 /* allow the security API to have its say */
530 ret
= security_file_mmap(file
, reqprot
, prot
, flags
);
535 *_capabilities
= capabilities
;
540 * we've determined that we can make the mapping, now translate what we
541 * now know into VMA flags
543 static unsigned long determine_vm_flags(struct file
*file
,
546 unsigned long capabilities
)
548 unsigned long vm_flags
;
550 vm_flags
= calc_vm_prot_bits(prot
) | calc_vm_flag_bits(flags
);
551 vm_flags
|= VM_MAYREAD
| VM_MAYWRITE
| VM_MAYEXEC
;
552 /* vm_flags |= mm->def_flags; */
554 if (!(capabilities
& BDI_CAP_MAP_DIRECT
)) {
555 /* attempt to share read-only copies of mapped file chunks */
556 if (file
&& !(prot
& PROT_WRITE
))
557 vm_flags
|= VM_MAYSHARE
;
560 /* overlay a shareable mapping on the backing device or inode
561 * if possible - used for chardevs, ramfs/tmpfs/shmfs and
563 if (flags
& MAP_SHARED
)
564 vm_flags
|= VM_MAYSHARE
| VM_SHARED
;
565 else if ((((vm_flags
& capabilities
) ^ vm_flags
) & BDI_CAP_VMFLAGS
) == 0)
566 vm_flags
|= VM_MAYSHARE
;
569 /* refuse to let anyone share private mappings with this process if
570 * it's being traced - otherwise breakpoints set in it may interfere
571 * with another untraced process
573 if ((flags
& MAP_PRIVATE
) && (current
->ptrace
& PT_PTRACED
))
574 vm_flags
&= ~VM_MAYSHARE
;
580 * set up a shared mapping on a file
582 static int do_mmap_shared_file(struct vm_area_struct
*vma
, unsigned long len
)
586 ret
= vma
->vm_file
->f_op
->mmap(vma
->vm_file
, vma
);
590 /* getting an ENOSYS error indicates that direct mmap isn't
591 * possible (as opposed to tried but failed) so we'll fall
592 * through to making a private copy of the data and mapping
598 * set up a private mapping or an anonymous shared mapping
600 static int do_mmap_private(struct vm_area_struct
*vma
, unsigned long len
)
605 /* invoke the file's mapping function so that it can keep track of
606 * shared mappings on devices or memory
607 * - VM_MAYSHARE will be set if it may attempt to share
610 ret
= vma
->vm_file
->f_op
->mmap(vma
->vm_file
, vma
);
611 if (ret
!= -ENOSYS
) {
612 /* shouldn't return success if we're not sharing */
613 BUG_ON(ret
== 0 && !(vma
->vm_flags
& VM_MAYSHARE
));
614 return ret
; /* success or a real error */
617 /* getting an ENOSYS error indicates that direct mmap isn't
618 * possible (as opposed to tried but failed) so we'll try to
619 * make a private copy of the data and map that instead */
622 /* allocate some memory to hold the mapping
623 * - note that this may not return a page-aligned address if the object
624 * we're allocating is smaller than a page
626 base
= kmalloc(len
, GFP_KERNEL
|__GFP_COMP
);
630 vma
->vm_start
= (unsigned long) base
;
631 vma
->vm_end
= vma
->vm_start
+ len
;
632 vma
->vm_flags
|= VM_MAPPED_COPY
;
635 if (len
+ WARN_ON_SLACK
<= kobjsize(result
))
636 printk("Allocation of %lu bytes from process %d has %lu bytes of slack\n",
637 len
, current
->pid
, kobjsize(result
) - len
);
641 /* read the contents of a file into the copy */
645 fpos
= vma
->vm_pgoff
;
650 ret
= vma
->vm_file
->f_op
->read(vma
->vm_file
, base
, len
, &fpos
);
656 /* clear the last little bit */
658 memset(base
+ ret
, 0, len
- ret
);
661 /* if it's an anonymous mapping, then just clear it */
662 memset(base
, 0, len
);
673 printk("Allocation of length %lu from process %d failed\n",
680 * handle mapping creation for uClinux
682 unsigned long do_mmap_pgoff(struct file
*file
,
689 struct vm_list_struct
*vml
= NULL
;
690 struct vm_area_struct
*vma
= NULL
;
692 unsigned long capabilities
, vm_flags
;
696 /* decide whether we should attempt the mapping, and if so what sort of
698 ret
= validate_mmap_request(file
, addr
, len
, prot
, flags
, pgoff
,
703 /* we've determined that we can make the mapping, now translate what we
704 * now know into VMA flags */
705 vm_flags
= determine_vm_flags(file
, prot
, flags
, capabilities
);
707 /* we're going to need to record the mapping if it works */
708 vml
= kmalloc(sizeof(struct vm_list_struct
), GFP_KERNEL
);
710 goto error_getting_vml
;
711 memset(vml
, 0, sizeof(*vml
));
713 down_write(&nommu_vma_sem
);
715 /* if we want to share, we need to check for VMAs created by other
716 * mmap() calls that overlap with our proposed mapping
717 * - we can only share with an exact match on most regular files
718 * - shared mappings on character devices and memory backed files are
719 * permitted to overlap inexactly as far as we are concerned for in
720 * these cases, sharing is handled in the driver or filesystem rather
723 if (vm_flags
& VM_MAYSHARE
) {
724 unsigned long pglen
= (len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
725 unsigned long vmpglen
;
727 for (rb
= rb_first(&nommu_vma_tree
); rb
; rb
= rb_next(rb
)) {
728 vma
= rb_entry(rb
, struct vm_area_struct
, vm_rb
);
730 if (!(vma
->vm_flags
& VM_MAYSHARE
))
733 /* search for overlapping mappings on the same file */
734 if (vma
->vm_file
->f_dentry
->d_inode
!= file
->f_dentry
->d_inode
)
737 if (vma
->vm_pgoff
>= pgoff
+ pglen
)
740 vmpglen
= vma
->vm_end
- vma
->vm_start
+ PAGE_SIZE
- 1;
741 vmpglen
>>= PAGE_SHIFT
;
742 if (pgoff
>= vma
->vm_pgoff
+ vmpglen
)
745 /* handle inexactly overlapping matches between mappings */
746 if (vma
->vm_pgoff
!= pgoff
|| vmpglen
!= pglen
) {
747 if (!(capabilities
& BDI_CAP_MAP_DIRECT
))
748 goto sharing_violation
;
752 /* we've found a VMA we can share */
753 atomic_inc(&vma
->vm_usage
);
756 result
= (void *) vma
->vm_start
;
762 /* obtain the address at which to make a shared mapping
763 * - this is the hook for quasi-memory character devices to
764 * tell us the location of a shared mapping
766 if (file
&& file
->f_op
->get_unmapped_area
) {
767 addr
= file
->f_op
->get_unmapped_area(file
, addr
, len
,
769 if (IS_ERR((void *) addr
)) {
771 if (ret
!= (unsigned long) -ENOSYS
)
774 /* the driver refused to tell us where to site
775 * the mapping so we'll have to attempt to copy
777 ret
= (unsigned long) -ENODEV
;
778 if (!(capabilities
& BDI_CAP_MAP_COPY
))
781 capabilities
&= ~BDI_CAP_MAP_DIRECT
;
786 /* we're going to need a VMA struct as well */
787 vma
= kmalloc(sizeof(struct vm_area_struct
), GFP_KERNEL
);
789 goto error_getting_vma
;
791 memset(vma
, 0, sizeof(*vma
));
792 INIT_LIST_HEAD(&vma
->anon_vma_node
);
793 atomic_set(&vma
->vm_usage
, 1);
797 vma
->vm_flags
= vm_flags
;
798 vma
->vm_start
= addr
;
799 vma
->vm_end
= addr
+ len
;
800 vma
->vm_pgoff
= pgoff
;
804 /* set up the mapping */
805 if (file
&& vma
->vm_flags
& VM_SHARED
)
806 ret
= do_mmap_shared_file(vma
, len
);
808 ret
= do_mmap_private(vma
, len
);
812 /* okay... we have a mapping; now we have to register it */
813 result
= (void *) vma
->vm_start
;
815 if (vma
->vm_flags
& VM_MAPPED_COPY
) {
816 realalloc
+= kobjsize(result
);
820 realalloc
+= kobjsize(vma
);
821 askedalloc
+= sizeof(*vma
);
823 current
->mm
->total_vm
+= len
>> PAGE_SHIFT
;
828 realalloc
+= kobjsize(vml
);
829 askedalloc
+= sizeof(*vml
);
831 vml
->next
= current
->mm
->context
.vmlist
;
832 current
->mm
->context
.vmlist
= vml
;
834 up_write(&nommu_vma_sem
);
836 if (prot
& PROT_EXEC
)
837 flush_icache_range((unsigned long) result
,
838 (unsigned long) result
+ len
);
841 printk("do_mmap:\n");
842 show_process_blocks();
845 return (unsigned long) result
;
848 up_write(&nommu_vma_sem
);
857 up_write(&nommu_vma_sem
);
858 printk("Attempt to share mismatched mappings\n");
863 up_write(&nommu_vma_sem
);
865 printk("Allocation of vma for %lu byte allocation from process %d failed\n",
871 printk("Allocation of vml for %lu byte allocation from process %d failed\n",
878 * handle mapping disposal for uClinux
880 static void put_vma(struct vm_area_struct
*vma
)
883 down_write(&nommu_vma_sem
);
885 if (atomic_dec_and_test(&vma
->vm_usage
)) {
886 delete_nommu_vma(vma
);
888 if (vma
->vm_ops
&& vma
->vm_ops
->close
)
889 vma
->vm_ops
->close(vma
);
891 /* IO memory and memory shared directly out of the pagecache from
892 * ramfs/tmpfs mustn't be released here */
893 if (vma
->vm_flags
& VM_MAPPED_COPY
) {
894 realalloc
-= kobjsize((void *) vma
->vm_start
);
895 askedalloc
-= vma
->vm_end
- vma
->vm_start
;
896 kfree((void *) vma
->vm_start
);
899 realalloc
-= kobjsize(vma
);
900 askedalloc
-= sizeof(*vma
);
907 up_write(&nommu_vma_sem
);
911 int do_munmap(struct mm_struct
*mm
, unsigned long addr
, size_t len
)
913 struct vm_list_struct
*vml
, **parent
;
914 unsigned long end
= addr
+ len
;
917 printk("do_munmap:\n");
920 for (parent
= &mm
->context
.vmlist
; *parent
; parent
= &(*parent
)->next
)
921 if ((*parent
)->vma
->vm_start
== addr
&&
922 ((len
== 0) || ((*parent
)->vma
->vm_end
== end
)))
925 printk("munmap of non-mmaped memory by process %d (%s): %p\n",
926 current
->pid
, current
->comm
, (void *) addr
);
935 realalloc
-= kobjsize(vml
);
936 askedalloc
-= sizeof(*vml
);
939 update_hiwater_vm(mm
);
940 mm
->total_vm
-= len
>> PAGE_SHIFT
;
943 show_process_blocks();
949 /* Release all mmaps. */
950 void exit_mmap(struct mm_struct
* mm
)
952 struct vm_list_struct
*tmp
;
956 printk("Exit_mmap:\n");
961 while ((tmp
= mm
->context
.vmlist
)) {
962 mm
->context
.vmlist
= tmp
->next
;
965 realalloc
-= kobjsize(tmp
);
966 askedalloc
-= sizeof(*tmp
);
971 show_process_blocks();
976 asmlinkage
long sys_munmap(unsigned long addr
, size_t len
)
979 struct mm_struct
*mm
= current
->mm
;
981 down_write(&mm
->mmap_sem
);
982 ret
= do_munmap(mm
, addr
, len
);
983 up_write(&mm
->mmap_sem
);
987 unsigned long do_brk(unsigned long addr
, unsigned long len
)
993 * Expand (or shrink) an existing mapping, potentially moving it at the
994 * same time (controlled by the MREMAP_MAYMOVE flag and available VM space)
996 * MREMAP_FIXED option added 5-Dec-1999 by Benjamin LaHaise
997 * This option implies MREMAP_MAYMOVE.
999 * on uClinux, we only permit changing a mapping's size, and only as long as it stays within the
1000 * hole allocated by the kmalloc() call in do_mmap_pgoff() and the block is not shareable
1002 unsigned long do_mremap(unsigned long addr
,
1003 unsigned long old_len
, unsigned long new_len
,
1004 unsigned long flags
, unsigned long new_addr
)
1006 struct vm_list_struct
*vml
= NULL
;
1008 /* insanity checks first */
1010 return (unsigned long) -EINVAL
;
1012 if (flags
& MREMAP_FIXED
&& new_addr
!= addr
)
1013 return (unsigned long) -EINVAL
;
1015 for (vml
= current
->mm
->context
.vmlist
; vml
; vml
= vml
->next
)
1016 if (vml
->vma
->vm_start
== addr
)
1019 return (unsigned long) -EINVAL
;
1022 if (vml
->vma
->vm_end
!= vml
->vma
->vm_start
+ old_len
)
1023 return (unsigned long) -EFAULT
;
1025 if (vml
->vma
->vm_flags
& VM_MAYSHARE
)
1026 return (unsigned long) -EPERM
;
1028 if (new_len
> kobjsize((void *) addr
))
1029 return (unsigned long) -ENOMEM
;
1031 /* all checks complete - do it */
1032 vml
->vma
->vm_end
= vml
->vma
->vm_start
+ new_len
;
1034 askedalloc
-= old_len
;
1035 askedalloc
+= new_len
;
1037 return vml
->vma
->vm_start
;
1041 * Look up the first VMA which satisfies addr < vm_end, NULL if none
1043 struct vm_area_struct
*find_vma(struct mm_struct
*mm
, unsigned long addr
)
1045 struct vm_list_struct
*vml
;
1047 for (vml
= mm
->context
.vmlist
; vml
; vml
= vml
->next
)
1048 if (addr
>= vml
->vma
->vm_start
&& addr
< vml
->vma
->vm_end
)
1054 EXPORT_SYMBOL(find_vma
);
1056 struct page
*follow_page(struct vm_area_struct
*vma
, unsigned long address
,
1057 unsigned int foll_flags
)
1062 struct vm_area_struct
*find_extend_vma(struct mm_struct
*mm
, unsigned long addr
)
1067 int remap_pfn_range(struct vm_area_struct
*vma
, unsigned long from
,
1068 unsigned long to
, unsigned long size
, pgprot_t prot
)
1070 vma
->vm_start
= vma
->vm_pgoff
<< PAGE_SHIFT
;
1074 void swap_unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
1078 unsigned long arch_get_unmapped_area(struct file
*file
, unsigned long addr
,
1079 unsigned long len
, unsigned long pgoff
, unsigned long flags
)
1084 void arch_unmap_area(struct mm_struct
*mm
, unsigned long addr
)
1088 void unmap_mapping_range(struct address_space
*mapping
,
1089 loff_t
const holebegin
, loff_t
const holelen
,
1095 * Check that a process has enough memory to allocate a new virtual
1096 * mapping. 0 means there is enough memory for the allocation to
1097 * succeed and -ENOMEM implies there is not.
1099 * We currently support three overcommit policies, which are set via the
1100 * vm.overcommit_memory sysctl. See Documentation/vm/overcommit-accounting
1102 * Strict overcommit modes added 2002 Feb 26 by Alan Cox.
1103 * Additional code 2002 Jul 20 by Robert Love.
1105 * cap_sys_admin is 1 if the process has admin privileges, 0 otherwise.
1107 * Note this is a helper function intended to be used by LSMs which
1108 * wish to use this logic.
1110 int __vm_enough_memory(long pages
, int cap_sys_admin
)
1112 unsigned long free
, allowed
;
1114 vm_acct_memory(pages
);
1117 * Sometimes we want to use more memory than we have
1119 if (sysctl_overcommit_memory
== OVERCOMMIT_ALWAYS
)
1122 if (sysctl_overcommit_memory
== OVERCOMMIT_GUESS
) {
1125 free
= get_page_cache_size();
1126 free
+= nr_swap_pages
;
1129 * Any slabs which are created with the
1130 * SLAB_RECLAIM_ACCOUNT flag claim to have contents
1131 * which are reclaimable, under pressure. The dentry
1132 * cache and most inode caches should fall into this
1134 free
+= atomic_read(&slab_reclaim_pages
);
1137 * Leave the last 3% for root
1146 * nr_free_pages() is very expensive on large systems,
1147 * only call if we're about to fail.
1149 n
= nr_free_pages();
1156 vm_unacct_memory(pages
);
1160 allowed
= totalram_pages
* sysctl_overcommit_ratio
/ 100;
1162 * Leave the last 3% for root
1165 allowed
-= allowed
/ 32;
1166 allowed
+= total_swap_pages
;
1168 /* Don't let a single process grow too big:
1169 leave 3% of the size of this process for other processes */
1170 allowed
-= current
->mm
->total_vm
/ 32;
1173 * cast `allowed' as a signed long because vm_committed_space
1174 * sometimes has a negative value
1176 if (atomic_read(&vm_committed_space
) < (long)allowed
)
1179 vm_unacct_memory(pages
);
1184 int in_gate_area_no_task(unsigned long addr
)
1189 struct page
*filemap_nopage(struct vm_area_struct
*area
,
1190 unsigned long address
, int *type
)