Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
f64e02fe TT |
2 | /* |
3 | * linux/fs/ext4/readpage.c | |
4 | * | |
5 | * Copyright (C) 2002, Linus Torvalds. | |
6 | * Copyright (C) 2015, Google, Inc. | |
7 | * | |
8 | * This was originally taken from fs/mpage.c | |
9 | * | |
10 | * The intent is the ext4_mpage_readpages() function here is intended | |
11 | * to replace mpage_readpages() in the general case, not just for | |
12 | * encrypted files. It has some limitations (see below), where it | |
13 | * will fall back to read_block_full_page(), but these limitations | |
14 | * should only be hit when page_size != block_size. | |
15 | * | |
16 | * This will allow us to attach a callback function to support ext4 | |
17 | * encryption. | |
18 | * | |
19 | * If anything unusual happens, such as: | |
20 | * | |
21 | * - encountering a page which has buffers | |
22 | * - encountering a page which has a non-hole after a hole | |
23 | * - encountering a page with non-contiguous blocks | |
24 | * | |
25 | * then this code just gives up and calls the buffer_head-based read function. | |
26 | * It does handle a page which has holes at the end - that is a common case: | |
ea1754a0 | 27 | * the end-of-file on blocksize < PAGE_SIZE setups. |
f64e02fe TT |
28 | * |
29 | */ | |
30 | ||
31 | #include <linux/kernel.h> | |
32 | #include <linux/export.h> | |
33 | #include <linux/mm.h> | |
34 | #include <linux/kdev_t.h> | |
35 | #include <linux/gfp.h> | |
36 | #include <linux/bio.h> | |
37 | #include <linux/fs.h> | |
38 | #include <linux/buffer_head.h> | |
39 | #include <linux/blkdev.h> | |
40 | #include <linux/highmem.h> | |
41 | #include <linux/prefetch.h> | |
42 | #include <linux/mpage.h> | |
43 | #include <linux/writeback.h> | |
44 | #include <linux/backing-dev.h> | |
45 | #include <linux/pagevec.h> | |
46 | #include <linux/cleancache.h> | |
47 | ||
48 | #include "ext4.h" | |
49 | ||
c9c7429c MH |
50 | static inline bool ext4_bio_encrypted(struct bio *bio) |
51 | { | |
643fa961 | 52 | #ifdef CONFIG_FS_ENCRYPTION |
c9c7429c MH |
53 | return unlikely(bio->bi_private != NULL); |
54 | #else | |
55 | return false; | |
56 | #endif | |
57 | } | |
58 | ||
f64e02fe TT |
59 | /* |
60 | * I/O completion handler for multipage BIOs. | |
61 | * | |
62 | * The mpage code never puts partial pages into a BIO (except for end-of-file). | |
63 | * If a page does not map to a contiguous run of blocks then it simply falls | |
64 | * back to block_read_full_page(). | |
65 | * | |
66 | * Why is this? If a page's completion depends on a number of different BIOs | |
67 | * which can complete in any order (or at the same time) then determining the | |
68 | * status of that page is hard. See end_buffer_async_read() for the details. | |
69 | * There is no point in duplicating all that complexity. | |
70 | */ | |
4246a0b6 | 71 | static void mpage_end_io(struct bio *bio) |
f64e02fe TT |
72 | { |
73 | struct bio_vec *bv; | |
6dc4f100 | 74 | struct bvec_iter_all iter_all; |
f64e02fe | 75 | |
c9c7429c | 76 | if (ext4_bio_encrypted(bio)) { |
4e4cbee9 | 77 | if (bio->bi_status) { |
a7550b30 | 78 | fscrypt_release_ctx(bio->bi_private); |
c9c7429c | 79 | } else { |
0cb8dae4 | 80 | fscrypt_enqueue_decrypt_bio(bio->bi_private, bio); |
c9c7429c MH |
81 | return; |
82 | } | |
83 | } | |
2b070cfe | 84 | bio_for_each_segment_all(bv, bio, iter_all) { |
f64e02fe TT |
85 | struct page *page = bv->bv_page; |
86 | ||
4e4cbee9 | 87 | if (!bio->bi_status) { |
f64e02fe TT |
88 | SetPageUptodate(page); |
89 | } else { | |
90 | ClearPageUptodate(page); | |
91 | SetPageError(page); | |
92 | } | |
93 | unlock_page(page); | |
94 | } | |
95 | ||
96 | bio_put(bio); | |
97 | } | |
98 | ||
99 | int ext4_mpage_readpages(struct address_space *mapping, | |
100 | struct list_head *pages, struct page *page, | |
ac22b46a | 101 | unsigned nr_pages, bool is_readahead) |
f64e02fe TT |
102 | { |
103 | struct bio *bio = NULL; | |
f64e02fe TT |
104 | sector_t last_block_in_bio = 0; |
105 | ||
106 | struct inode *inode = mapping->host; | |
107 | const unsigned blkbits = inode->i_blkbits; | |
09cbfeaf | 108 | const unsigned blocks_per_page = PAGE_SIZE >> blkbits; |
f64e02fe TT |
109 | const unsigned blocksize = 1 << blkbits; |
110 | sector_t block_in_file; | |
111 | sector_t last_block; | |
112 | sector_t last_block_in_file; | |
113 | sector_t blocks[MAX_BUF_PER_PAGE]; | |
114 | unsigned page_block; | |
115 | struct block_device *bdev = inode->i_sb->s_bdev; | |
116 | int length; | |
117 | unsigned relative_block = 0; | |
118 | struct ext4_map_blocks map; | |
119 | ||
120 | map.m_pblk = 0; | |
121 | map.m_lblk = 0; | |
122 | map.m_len = 0; | |
123 | map.m_flags = 0; | |
124 | ||
de9e9181 | 125 | for (; nr_pages; nr_pages--) { |
f64e02fe TT |
126 | int fully_mapped = 1; |
127 | unsigned first_hole = blocks_per_page; | |
128 | ||
f64e02fe | 129 | if (pages) { |
f86196ea | 130 | page = lru_to_page(pages); |
d454a273 LX |
131 | |
132 | prefetchw(&page->flags); | |
f64e02fe | 133 | list_del(&page->lru); |
063d99b4 | 134 | if (add_to_page_cache_lru(page, mapping, page->index, |
8a5c743e | 135 | readahead_gfp_mask(mapping))) |
f64e02fe TT |
136 | goto next_page; |
137 | } | |
138 | ||
139 | if (page_has_buffers(page)) | |
140 | goto confused; | |
141 | ||
09cbfeaf | 142 | block_in_file = (sector_t)page->index << (PAGE_SHIFT - blkbits); |
f64e02fe TT |
143 | last_block = block_in_file + nr_pages * blocks_per_page; |
144 | last_block_in_file = (i_size_read(inode) + blocksize - 1) >> blkbits; | |
145 | if (last_block > last_block_in_file) | |
146 | last_block = last_block_in_file; | |
147 | page_block = 0; | |
148 | ||
149 | /* | |
150 | * Map blocks using the previous result first. | |
151 | */ | |
152 | if ((map.m_flags & EXT4_MAP_MAPPED) && | |
153 | block_in_file > map.m_lblk && | |
154 | block_in_file < (map.m_lblk + map.m_len)) { | |
155 | unsigned map_offset = block_in_file - map.m_lblk; | |
156 | unsigned last = map.m_len - map_offset; | |
157 | ||
158 | for (relative_block = 0; ; relative_block++) { | |
159 | if (relative_block == last) { | |
160 | /* needed? */ | |
161 | map.m_flags &= ~EXT4_MAP_MAPPED; | |
162 | break; | |
163 | } | |
164 | if (page_block == blocks_per_page) | |
165 | break; | |
166 | blocks[page_block] = map.m_pblk + map_offset + | |
167 | relative_block; | |
168 | page_block++; | |
169 | block_in_file++; | |
170 | } | |
171 | } | |
172 | ||
173 | /* | |
174 | * Then do more ext4_map_blocks() calls until we are | |
175 | * done with this page. | |
176 | */ | |
177 | while (page_block < blocks_per_page) { | |
178 | if (block_in_file < last_block) { | |
179 | map.m_lblk = block_in_file; | |
180 | map.m_len = last_block - block_in_file; | |
181 | ||
182 | if (ext4_map_blocks(NULL, inode, &map, 0) < 0) { | |
183 | set_error_page: | |
184 | SetPageError(page); | |
185 | zero_user_segment(page, 0, | |
09cbfeaf | 186 | PAGE_SIZE); |
f64e02fe TT |
187 | unlock_page(page); |
188 | goto next_page; | |
189 | } | |
190 | } | |
191 | if ((map.m_flags & EXT4_MAP_MAPPED) == 0) { | |
192 | fully_mapped = 0; | |
193 | if (first_hole == blocks_per_page) | |
194 | first_hole = page_block; | |
195 | page_block++; | |
196 | block_in_file++; | |
197 | continue; | |
198 | } | |
199 | if (first_hole != blocks_per_page) | |
200 | goto confused; /* hole -> non-hole */ | |
201 | ||
202 | /* Contiguous blocks? */ | |
203 | if (page_block && blocks[page_block-1] != map.m_pblk-1) | |
204 | goto confused; | |
205 | for (relative_block = 0; ; relative_block++) { | |
206 | if (relative_block == map.m_len) { | |
207 | /* needed? */ | |
208 | map.m_flags &= ~EXT4_MAP_MAPPED; | |
209 | break; | |
210 | } else if (page_block == blocks_per_page) | |
211 | break; | |
212 | blocks[page_block] = map.m_pblk+relative_block; | |
213 | page_block++; | |
214 | block_in_file++; | |
215 | } | |
216 | } | |
217 | if (first_hole != blocks_per_page) { | |
218 | zero_user_segment(page, first_hole << blkbits, | |
09cbfeaf | 219 | PAGE_SIZE); |
f64e02fe TT |
220 | if (first_hole == 0) { |
221 | SetPageUptodate(page); | |
222 | unlock_page(page); | |
223 | goto next_page; | |
224 | } | |
225 | } else if (fully_mapped) { | |
226 | SetPageMappedToDisk(page); | |
227 | } | |
228 | if (fully_mapped && blocks_per_page == 1 && | |
229 | !PageUptodate(page) && cleancache_get_page(page) == 0) { | |
230 | SetPageUptodate(page); | |
231 | goto confused; | |
232 | } | |
233 | ||
234 | /* | |
235 | * This page will go to BIO. Do we need to send this | |
236 | * BIO off first? | |
237 | */ | |
238 | if (bio && (last_block_in_bio != blocks[0] - 1)) { | |
239 | submit_and_realloc: | |
4e49ea4a | 240 | submit_bio(bio); |
f64e02fe TT |
241 | bio = NULL; |
242 | } | |
243 | if (bio == NULL) { | |
a7550b30 | 244 | struct fscrypt_ctx *ctx = NULL; |
c9c7429c | 245 | |
592ddec7 | 246 | if (IS_ENCRYPTED(inode) && S_ISREG(inode->i_mode)) { |
cd0265fc | 247 | ctx = fscrypt_get_ctx(GFP_NOFS); |
c9c7429c MH |
248 | if (IS_ERR(ctx)) |
249 | goto set_error_page; | |
250 | } | |
f64e02fe | 251 | bio = bio_alloc(GFP_KERNEL, |
b54ffb73 | 252 | min_t(int, nr_pages, BIO_MAX_PAGES)); |
c9c7429c MH |
253 | if (!bio) { |
254 | if (ctx) | |
a7550b30 | 255 | fscrypt_release_ctx(ctx); |
f64e02fe | 256 | goto set_error_page; |
c9c7429c | 257 | } |
74d46992 | 258 | bio_set_dev(bio, bdev); |
f64e02fe TT |
259 | bio->bi_iter.bi_sector = blocks[0] << (blkbits - 9); |
260 | bio->bi_end_io = mpage_end_io; | |
c9c7429c | 261 | bio->bi_private = ctx; |
ac22b46a JA |
262 | bio_set_op_attrs(bio, REQ_OP_READ, |
263 | is_readahead ? REQ_RAHEAD : 0); | |
f64e02fe TT |
264 | } |
265 | ||
266 | length = first_hole << blkbits; | |
267 | if (bio_add_page(bio, page, length, 0) < length) | |
268 | goto submit_and_realloc; | |
269 | ||
270 | if (((map.m_flags & EXT4_MAP_BOUNDARY) && | |
271 | (relative_block == map.m_len)) || | |
272 | (first_hole != blocks_per_page)) { | |
4e49ea4a | 273 | submit_bio(bio); |
f64e02fe TT |
274 | bio = NULL; |
275 | } else | |
276 | last_block_in_bio = blocks[blocks_per_page - 1]; | |
277 | goto next_page; | |
278 | confused: | |
279 | if (bio) { | |
4e49ea4a | 280 | submit_bio(bio); |
f64e02fe TT |
281 | bio = NULL; |
282 | } | |
283 | if (!PageUptodate(page)) | |
284 | block_read_full_page(page, ext4_get_block); | |
285 | else | |
286 | unlock_page(page); | |
287 | next_page: | |
288 | if (pages) | |
09cbfeaf | 289 | put_page(page); |
f64e02fe TT |
290 | } |
291 | BUG_ON(pages && !list_empty(pages)); | |
292 | if (bio) | |
4e49ea4a | 293 | submit_bio(bio); |
f64e02fe TT |
294 | return 0; |
295 | } |