Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * mm/fadvise.c
4 *
5 * Copyright (C) 2002, Linus Torvalds
6 *
7 * 11Jan2003 Andrew Morton
8 * Initial version.
9 */
10
11#include <linux/kernel.h>
12#include <linux/file.h>
13#include <linux/fs.h>
14#include <linux/mm.h>
15#include <linux/pagemap.h>
16#include <linux/backing-dev.h>
17#include <linux/pagevec.h>
18#include <linux/fadvise.h>
19#include <linux/writeback.h>
20#include <linux/syscalls.h>
21#include <linux/swap.h>
22
23#include <asm/unistd.h>
24
25#include "internal.h"
26
27/*
28 * POSIX_FADV_WILLNEED could set PG_Referenced, and POSIX_FADV_NOREUSE could
29 * deactivate the pages and clear PG_Referenced.
30 */
31
32int generic_fadvise(struct file *file, loff_t offset, loff_t len, int advice)
33{
34 struct inode *inode;
35 struct address_space *mapping;
36 struct backing_dev_info *bdi;
37 loff_t endbyte; /* inclusive */
38 pgoff_t start_index;
39 pgoff_t end_index;
40 unsigned long nrpages;
41
42 inode = file_inode(file);
43 if (S_ISFIFO(inode->i_mode))
44 return -ESPIPE;
45
46 mapping = file->f_mapping;
47 if (!mapping || len < 0)
48 return -EINVAL;
49
50 bdi = inode_to_bdi(mapping->host);
51
52 if (IS_DAX(inode) || (bdi == &noop_backing_dev_info)) {
53 switch (advice) {
54 case POSIX_FADV_NORMAL:
55 case POSIX_FADV_RANDOM:
56 case POSIX_FADV_SEQUENTIAL:
57 case POSIX_FADV_WILLNEED:
58 case POSIX_FADV_NOREUSE:
59 case POSIX_FADV_DONTNEED:
60 /* no bad return value, but ignore advice */
61 break;
62 default:
63 return -EINVAL;
64 }
65 return 0;
66 }
67
68 /*
69 * Careful about overflows. Len == 0 means "as much as possible". Use
70 * unsigned math because signed overflows are undefined and UBSan
71 * complains.
72 */
73 endbyte = (u64)offset + (u64)len;
74 if (!len || endbyte < len)
75 endbyte = -1;
76 else
77 endbyte--; /* inclusive */
78
79 switch (advice) {
80 case POSIX_FADV_NORMAL:
81 file->f_ra.ra_pages = bdi->ra_pages;
82 spin_lock(&file->f_lock);
83 file->f_mode &= ~FMODE_RANDOM;
84 spin_unlock(&file->f_lock);
85 break;
86 case POSIX_FADV_RANDOM:
87 spin_lock(&file->f_lock);
88 file->f_mode |= FMODE_RANDOM;
89 spin_unlock(&file->f_lock);
90 break;
91 case POSIX_FADV_SEQUENTIAL:
92 file->f_ra.ra_pages = bdi->ra_pages * 2;
93 spin_lock(&file->f_lock);
94 file->f_mode &= ~FMODE_RANDOM;
95 spin_unlock(&file->f_lock);
96 break;
97 case POSIX_FADV_WILLNEED:
98 /* First and last PARTIAL page! */
99 start_index = offset >> PAGE_SHIFT;
100 end_index = endbyte >> PAGE_SHIFT;
101
102 /* Careful about overflow on the "+1" */
103 nrpages = end_index - start_index + 1;
104 if (!nrpages)
105 nrpages = ~0UL;
106
107 force_page_cache_readahead(mapping, file, start_index, nrpages);
108 break;
109 case POSIX_FADV_NOREUSE:
110 break;
111 case POSIX_FADV_DONTNEED:
112 if (!inode_write_congested(mapping->host))
113 __filemap_fdatawrite_range(mapping, offset, endbyte,
114 WB_SYNC_NONE);
115
116 /*
117 * First and last FULL page! Partial pages are deliberately
118 * preserved on the expectation that it is better to preserve
119 * needed memory than to discard unneeded memory.
120 */
121 start_index = (offset+(PAGE_SIZE-1)) >> PAGE_SHIFT;
122 end_index = (endbyte >> PAGE_SHIFT);
123 /*
124 * The page at end_index will be inclusively discarded according
125 * by invalidate_mapping_pages(), so subtracting 1 from
126 * end_index means we will skip the last page. But if endbyte
127 * is page aligned or is at the end of file, we should not skip
128 * that page - discarding the last page is safe enough.
129 */
130 if ((endbyte & ~PAGE_MASK) != ~PAGE_MASK &&
131 endbyte != inode->i_size - 1) {
132 /* First page is tricky as 0 - 1 = -1, but pgoff_t
133 * is unsigned, so the end_index >= start_index
134 * check below would be true and we'll discard the whole
135 * file cache which is not what was asked.
136 */
137 if (end_index == 0)
138 break;
139
140 end_index--;
141 }
142
143 if (end_index >= start_index) {
144 unsigned long count;
145
146 /*
147 * It's common to FADV_DONTNEED right after
148 * the read or write that instantiates the
149 * pages, in which case there will be some
150 * sitting on the local LRU cache. Try to
151 * avoid the expensive remote drain and the
152 * second cache tree walk below by flushing
153 * them out right away.
154 */
155 lru_add_drain();
156
157 count = invalidate_mapping_pages(mapping,
158 start_index, end_index);
159
160 /*
161 * If fewer pages were invalidated than expected then
162 * it is possible that some of the pages were on
163 * a per-cpu pagevec for a remote CPU. Drain all
164 * pagevecs and try again.
165 */
166 if (count < (end_index - start_index + 1)) {
167 lru_add_drain_all();
168 invalidate_mapping_pages(mapping, start_index,
169 end_index);
170 }
171 }
172 break;
173 default:
174 return -EINVAL;
175 }
176 return 0;
177}
178EXPORT_SYMBOL(generic_fadvise);
179
180int vfs_fadvise(struct file *file, loff_t offset, loff_t len, int advice)
181{
182 if (file->f_op->fadvise)
183 return file->f_op->fadvise(file, offset, len, advice);
184
185 return generic_fadvise(file, offset, len, advice);
186}
187EXPORT_SYMBOL(vfs_fadvise);
188
189#ifdef CONFIG_ADVISE_SYSCALLS
190
191int ksys_fadvise64_64(int fd, loff_t offset, loff_t len, int advice)
192{
193 struct fd f = fdget(fd);
194 int ret;
195
196 if (!f.file)
197 return -EBADF;
198
199 ret = vfs_fadvise(f.file, offset, len, advice);
200
201 fdput(f);
202 return ret;
203}
204
205SYSCALL_DEFINE4(fadvise64_64, int, fd, loff_t, offset, loff_t, len, int, advice)
206{
207 return ksys_fadvise64_64(fd, offset, len, advice);
208}
209
210#ifdef __ARCH_WANT_SYS_FADVISE64
211
212SYSCALL_DEFINE4(fadvise64, int, fd, loff_t, offset, size_t, len, int, advice)
213{
214 return ksys_fadvise64_64(fd, offset, len, advice);
215}
216
217#endif
218#endif
1/*
2 * mm/fadvise.c
3 *
4 * Copyright (C) 2002, Linus Torvalds
5 *
6 * 11Jan2003 Andrew Morton
7 * Initial version.
8 */
9
10#include <linux/kernel.h>
11#include <linux/file.h>
12#include <linux/fs.h>
13#include <linux/mm.h>
14#include <linux/pagemap.h>
15#include <linux/backing-dev.h>
16#include <linux/pagevec.h>
17#include <linux/fadvise.h>
18#include <linux/writeback.h>
19#include <linux/syscalls.h>
20
21#include <asm/unistd.h>
22
23/*
24 * POSIX_FADV_WILLNEED could set PG_Referenced, and POSIX_FADV_NOREUSE could
25 * deactivate the pages and clear PG_Referenced.
26 */
27SYSCALL_DEFINE(fadvise64_64)(int fd, loff_t offset, loff_t len, int advice)
28{
29 struct file *file = fget(fd);
30 struct address_space *mapping;
31 struct backing_dev_info *bdi;
32 loff_t endbyte; /* inclusive */
33 pgoff_t start_index;
34 pgoff_t end_index;
35 unsigned long nrpages;
36 int ret = 0;
37
38 if (!file)
39 return -EBADF;
40
41 if (S_ISFIFO(file->f_path.dentry->d_inode->i_mode)) {
42 ret = -ESPIPE;
43 goto out;
44 }
45
46 mapping = file->f_mapping;
47 if (!mapping || len < 0) {
48 ret = -EINVAL;
49 goto out;
50 }
51
52 if (mapping->a_ops->get_xip_mem) {
53 switch (advice) {
54 case POSIX_FADV_NORMAL:
55 case POSIX_FADV_RANDOM:
56 case POSIX_FADV_SEQUENTIAL:
57 case POSIX_FADV_WILLNEED:
58 case POSIX_FADV_NOREUSE:
59 case POSIX_FADV_DONTNEED:
60 /* no bad return value, but ignore advice */
61 break;
62 default:
63 ret = -EINVAL;
64 }
65 goto out;
66 }
67
68 /* Careful about overflows. Len == 0 means "as much as possible" */
69 endbyte = offset + len;
70 if (!len || endbyte < len)
71 endbyte = -1;
72 else
73 endbyte--; /* inclusive */
74
75 bdi = mapping->backing_dev_info;
76
77 switch (advice) {
78 case POSIX_FADV_NORMAL:
79 file->f_ra.ra_pages = bdi->ra_pages;
80 spin_lock(&file->f_lock);
81 file->f_mode &= ~FMODE_RANDOM;
82 spin_unlock(&file->f_lock);
83 break;
84 case POSIX_FADV_RANDOM:
85 spin_lock(&file->f_lock);
86 file->f_mode |= FMODE_RANDOM;
87 spin_unlock(&file->f_lock);
88 break;
89 case POSIX_FADV_SEQUENTIAL:
90 file->f_ra.ra_pages = bdi->ra_pages * 2;
91 spin_lock(&file->f_lock);
92 file->f_mode &= ~FMODE_RANDOM;
93 spin_unlock(&file->f_lock);
94 break;
95 case POSIX_FADV_WILLNEED:
96 if (!mapping->a_ops->readpage) {
97 ret = -EINVAL;
98 break;
99 }
100
101 /* First and last PARTIAL page! */
102 start_index = offset >> PAGE_CACHE_SHIFT;
103 end_index = endbyte >> PAGE_CACHE_SHIFT;
104
105 /* Careful about overflow on the "+1" */
106 nrpages = end_index - start_index + 1;
107 if (!nrpages)
108 nrpages = ~0UL;
109
110 ret = force_page_cache_readahead(mapping, file,
111 start_index,
112 nrpages);
113 if (ret > 0)
114 ret = 0;
115 break;
116 case POSIX_FADV_NOREUSE:
117 break;
118 case POSIX_FADV_DONTNEED:
119 if (!bdi_write_congested(mapping->backing_dev_info))
120 filemap_flush(mapping);
121
122 /* First and last FULL page! */
123 start_index = (offset+(PAGE_CACHE_SIZE-1)) >> PAGE_CACHE_SHIFT;
124 end_index = (endbyte >> PAGE_CACHE_SHIFT);
125
126 if (end_index >= start_index)
127 invalidate_mapping_pages(mapping, start_index,
128 end_index);
129 break;
130 default:
131 ret = -EINVAL;
132 }
133out:
134 fput(file);
135 return ret;
136}
137#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
138asmlinkage long SyS_fadvise64_64(long fd, loff_t offset, loff_t len, long advice)
139{
140 return SYSC_fadvise64_64((int) fd, offset, len, (int) advice);
141}
142SYSCALL_ALIAS(sys_fadvise64_64, SyS_fadvise64_64);
143#endif
144
145#ifdef __ARCH_WANT_SYS_FADVISE64
146
147SYSCALL_DEFINE(fadvise64)(int fd, loff_t offset, size_t len, int advice)
148{
149 return sys_fadvise64_64(fd, offset, len, advice);
150}
151#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
152asmlinkage long SyS_fadvise64(long fd, loff_t offset, long len, long advice)
153{
154 return SYSC_fadvise64((int) fd, offset, (size_t)len, (int)advice);
155}
156SYSCALL_ALIAS(sys_fadvise64, SyS_fadvise64);
157#endif
158
159#endif