[PATCH] md: fix various bugs with aligned reads in RAID5
[wrt350n-kernel.git] / mm / mincore.c
blob8aca6f7167bbb25f858075c90c64d766fcc1e4c3
1 /*
2 * linux/mm/mincore.c
4 * Copyright (C) 1994-2006 Linus Torvalds
5 */
7 /*
8 * The mincore() system call.
9 */
10 #include <linux/slab.h>
11 #include <linux/pagemap.h>
12 #include <linux/mm.h>
13 #include <linux/mman.h>
14 #include <linux/syscalls.h>
16 #include <asm/uaccess.h>
17 #include <asm/pgtable.h>
20 * Later we can get more picky about what "in core" means precisely.
21 * For now, simply check to see if the page is in the page cache,
22 * and is up to date; i.e. that no page-in operation would be required
23 * at this time if an application were to map and access this page.
25 static unsigned char mincore_page(struct vm_area_struct * vma,
26 unsigned long pgoff)
28 unsigned char present = 0;
29 struct address_space * as = vma->vm_file->f_mapping;
30 struct page * page;
32 page = find_get_page(as, pgoff);
33 if (page) {
34 present = PageUptodate(page);
35 page_cache_release(page);
38 return present;
42 * Do a chunk of "sys_mincore()". We've already checked
43 * all the arguments, we hold the mmap semaphore: we should
44 * just return the amount of info we're asked for.
46 static long do_mincore(unsigned long addr, unsigned char *vec, unsigned long pages)
48 unsigned long i, nr, pgoff;
49 struct vm_area_struct *vma = find_vma(current->mm, addr);
52 * find_vma() didn't find anything above us, or we're
53 * in an unmapped hole in the address space: ENOMEM.
55 if (!vma || addr < vma->vm_start)
56 return -ENOMEM;
59 * Ok, got it. But check whether it's a segment we support
60 * mincore() on. Right now, we don't do any anonymous mappings.
62 * FIXME: This is just stupid. And returning ENOMEM is
63 * stupid too. We should just look at the page tables. But
64 * this is what we've traditionally done, so we'll just
65 * continue doing it.
67 if (!vma->vm_file)
68 return -ENOMEM;
71 * Calculate how many pages there are left in the vma, and
72 * what the pgoff is for our address.
74 nr = (vma->vm_end - addr) >> PAGE_SHIFT;
75 if (nr > pages)
76 nr = pages;
78 pgoff = (addr - vma->vm_start) >> PAGE_SHIFT;
79 pgoff += vma->vm_pgoff;
81 /* And then we just fill the sucker in.. */
82 for (i = 0 ; i < nr; i++, pgoff++)
83 vec[i] = mincore_page(vma, pgoff);
85 return nr;
89 * The mincore(2) system call.
91 * mincore() returns the memory residency status of the pages in the
92 * current process's address space specified by [addr, addr + len).
93 * The status is returned in a vector of bytes. The least significant
94 * bit of each byte is 1 if the referenced page is in memory, otherwise
95 * it is zero.
97 * Because the status of a page can change after mincore() checks it
98 * but before it returns to the application, the returned vector may
99 * contain stale information. Only locked pages are guaranteed to
100 * remain in memory.
102 * return values:
103 * zero - success
104 * -EFAULT - vec points to an illegal address
105 * -EINVAL - addr is not a multiple of PAGE_CACHE_SIZE
106 * -ENOMEM - Addresses in the range [addr, addr + len] are
107 * invalid for the address space of this process, or
108 * specify one or more pages which are not currently
109 * mapped
110 * -EAGAIN - A kernel resource was temporarily unavailable.
112 asmlinkage long sys_mincore(unsigned long start, size_t len,
113 unsigned char __user * vec)
115 long retval;
116 unsigned long pages;
117 unsigned char *tmp;
119 /* Check the start address: needs to be page-aligned.. */
120 if (start & ~PAGE_CACHE_MASK)
121 return -EINVAL;
123 /* ..and we need to be passed a valid user-space range */
124 if (!access_ok(VERIFY_READ, (void __user *) start, len))
125 return -ENOMEM;
127 /* This also avoids any overflows on PAGE_CACHE_ALIGN */
128 pages = len >> PAGE_SHIFT;
129 pages += (len & ~PAGE_MASK) != 0;
131 if (!access_ok(VERIFY_WRITE, vec, pages))
132 return -EFAULT;
134 tmp = (void *) __get_free_page(GFP_USER);
135 if (!tmp)
136 return -EAGAIN;
138 retval = 0;
139 while (pages) {
141 * Do at most PAGE_SIZE entries per iteration, due to
142 * the temporary buffer size.
144 down_read(&current->mm->mmap_sem);
145 retval = do_mincore(start, tmp, min(pages, PAGE_SIZE));
146 up_read(&current->mm->mmap_sem);
148 if (retval <= 0)
149 break;
150 if (copy_to_user(vec, tmp, retval)) {
151 retval = -EFAULT;
152 break;
154 pages -= retval;
155 vec += retval;
156 start += retval << PAGE_SHIFT;
157 retval = 0;
159 free_page((unsigned long) tmp);
160 return retval;