xref: /openbmc/linux/mm/mincore.c (revision b627b4ed)
1 /*
2  *	linux/mm/mincore.c
3  *
4  * Copyright (C) 1994-2006  Linus Torvalds
5  */
6 
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>
15 #include <linux/swap.h>
16 #include <linux/swapops.h>
17 
18 #include <asm/uaccess.h>
19 #include <asm/pgtable.h>
20 
21 /*
22  * Later we can get more picky about what "in core" means precisely.
23  * For now, simply check to see if the page is in the page cache,
24  * and is up to date; i.e. that no page-in operation would be required
25  * at this time if an application were to map and access this page.
26  */
27 static unsigned char mincore_page(struct address_space *mapping, pgoff_t pgoff)
28 {
29 	unsigned char present = 0;
30 	struct page *page;
31 
32 	/*
33 	 * When tmpfs swaps out a page from a file, any process mapping that
34 	 * file will not get a swp_entry_t in its pte, but rather it is like
35 	 * any other file mapping (ie. marked !present and faulted in with
36 	 * tmpfs's .fault). So swapped out tmpfs mappings are tested here.
37 	 *
38 	 * However when tmpfs moves the page from pagecache and into swapcache,
39 	 * it is still in core, but the find_get_page below won't find it.
40 	 * No big deal, but make a note of it.
41 	 */
42 	page = find_get_page(mapping, pgoff);
43 	if (page) {
44 		present = PageUptodate(page);
45 		page_cache_release(page);
46 	}
47 
48 	return present;
49 }
50 
51 /*
52  * Do a chunk of "sys_mincore()". We've already checked
53  * all the arguments, we hold the mmap semaphore: we should
54  * just return the amount of info we're asked for.
55  */
56 static long do_mincore(unsigned long addr, unsigned char *vec, unsigned long pages)
57 {
58 	pgd_t *pgd;
59 	pud_t *pud;
60 	pmd_t *pmd;
61 	pte_t *ptep;
62 	spinlock_t *ptl;
63 	unsigned long nr;
64 	int i;
65 	pgoff_t pgoff;
66 	struct vm_area_struct *vma = find_vma(current->mm, addr);
67 
68 	/*
69 	 * find_vma() didn't find anything above us, or we're
70 	 * in an unmapped hole in the address space: ENOMEM.
71 	 */
72 	if (!vma || addr < vma->vm_start)
73 		return -ENOMEM;
74 
75 	/*
76 	 * Calculate how many pages there are left in the last level of the
77 	 * PTE array for our address.
78 	 */
79 	nr = PTRS_PER_PTE - ((addr >> PAGE_SHIFT) & (PTRS_PER_PTE-1));
80 
81 	/*
82 	 * Don't overrun this vma
83 	 */
84 	nr = min(nr, (vma->vm_end - addr) >> PAGE_SHIFT);
85 
86 	/*
87 	 * Don't return more than the caller asked for
88 	 */
89 	nr = min(nr, pages);
90 
91 	pgd = pgd_offset(vma->vm_mm, addr);
92 	if (pgd_none_or_clear_bad(pgd))
93 		goto none_mapped;
94 	pud = pud_offset(pgd, addr);
95 	if (pud_none_or_clear_bad(pud))
96 		goto none_mapped;
97 	pmd = pmd_offset(pud, addr);
98 	if (pmd_none_or_clear_bad(pmd))
99 		goto none_mapped;
100 
101 	ptep = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
102 	for (i = 0; i < nr; i++, ptep++, addr += PAGE_SIZE) {
103 		unsigned char present;
104 		pte_t pte = *ptep;
105 
106 		if (pte_present(pte)) {
107 			present = 1;
108 
109 		} else if (pte_none(pte)) {
110 			if (vma->vm_file) {
111 				pgoff = linear_page_index(vma, addr);
112 				present = mincore_page(vma->vm_file->f_mapping,
113 							pgoff);
114 			} else
115 				present = 0;
116 
117 		} else if (pte_file(pte)) {
118 			pgoff = pte_to_pgoff(pte);
119 			present = mincore_page(vma->vm_file->f_mapping, pgoff);
120 
121 		} else { /* pte is a swap entry */
122 			swp_entry_t entry = pte_to_swp_entry(pte);
123 			if (is_migration_entry(entry)) {
124 				/* migration entries are always uptodate */
125 				present = 1;
126 			} else {
127 #ifdef CONFIG_SWAP
128 				pgoff = entry.val;
129 				present = mincore_page(&swapper_space, pgoff);
130 #else
131 				WARN_ON(1);
132 				present = 1;
133 #endif
134 			}
135 		}
136 
137 		vec[i] = present;
138 	}
139 	pte_unmap_unlock(ptep-1, ptl);
140 
141 	return nr;
142 
143 none_mapped:
144 	if (vma->vm_file) {
145 		pgoff = linear_page_index(vma, addr);
146 		for (i = 0; i < nr; i++, pgoff++)
147 			vec[i] = mincore_page(vma->vm_file->f_mapping, pgoff);
148 	} else {
149 		for (i = 0; i < nr; i++)
150 			vec[i] = 0;
151 	}
152 
153 	return nr;
154 }
155 
156 /*
157  * The mincore(2) system call.
158  *
159  * mincore() returns the memory residency status of the pages in the
160  * current process's address space specified by [addr, addr + len).
161  * The status is returned in a vector of bytes.  The least significant
162  * bit of each byte is 1 if the referenced page is in memory, otherwise
163  * it is zero.
164  *
165  * Because the status of a page can change after mincore() checks it
166  * but before it returns to the application, the returned vector may
167  * contain stale information.  Only locked pages are guaranteed to
168  * remain in memory.
169  *
170  * return values:
171  *  zero    - success
172  *  -EFAULT - vec points to an illegal address
173  *  -EINVAL - addr is not a multiple of PAGE_CACHE_SIZE
174  *  -ENOMEM - Addresses in the range [addr, addr + len] are
175  *		invalid for the address space of this process, or
176  *		specify one or more pages which are not currently
177  *		mapped
178  *  -EAGAIN - A kernel resource was temporarily unavailable.
179  */
180 SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len,
181 		unsigned char __user *, vec)
182 {
183 	long retval;
184 	unsigned long pages;
185 	unsigned char *tmp;
186 
187 	/* Check the start address: needs to be page-aligned.. */
188  	if (start & ~PAGE_CACHE_MASK)
189 		return -EINVAL;
190 
191 	/* ..and we need to be passed a valid user-space range */
192 	if (!access_ok(VERIFY_READ, (void __user *) start, len))
193 		return -ENOMEM;
194 
195 	/* This also avoids any overflows on PAGE_CACHE_ALIGN */
196 	pages = len >> PAGE_SHIFT;
197 	pages += (len & ~PAGE_MASK) != 0;
198 
199 	if (!access_ok(VERIFY_WRITE, vec, pages))
200 		return -EFAULT;
201 
202 	tmp = (void *) __get_free_page(GFP_USER);
203 	if (!tmp)
204 		return -EAGAIN;
205 
206 	retval = 0;
207 	while (pages) {
208 		/*
209 		 * Do at most PAGE_SIZE entries per iteration, due to
210 		 * the temporary buffer size.
211 		 */
212 		down_read(&current->mm->mmap_sem);
213 		retval = do_mincore(start, tmp, min(pages, PAGE_SIZE));
214 		up_read(&current->mm->mmap_sem);
215 
216 		if (retval <= 0)
217 			break;
218 		if (copy_to_user(vec, tmp, retval)) {
219 			retval = -EFAULT;
220 			break;
221 		}
222 		pages -= retval;
223 		vec += retval;
224 		start += retval << PAGE_SHIFT;
225 		retval = 0;
226 	}
227 	free_page((unsigned long) tmp);
228 	return retval;
229 }
230