]>
Commit | Line | Data |
---|---|---|
1e51764a AB |
1 | /* |
2 | * This file is part of UBIFS. | |
3 | * | |
4 | * Copyright (C) 2006-2008 Nokia Corporation. | |
5 | * Copyright (C) 2006, 2007 University of Szeged, Hungary | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify it | |
8 | * under the terms of the GNU General Public License version 2 as published by | |
9 | * the Free Software Foundation. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, but WITHOUT | |
12 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
13 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
14 | * more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License along with | |
17 | * this program; if not, write to the Free Software Foundation, Inc., 51 | |
18 | * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
19 | * | |
20 | * Authors: Artem Bityutskiy (Битюцкий Артём) | |
21 | * Adrian Hunter | |
22 | * Zoltan Sogor | |
23 | */ | |
24 | ||
25 | /* | |
26 | * This file implements UBIFS I/O subsystem which provides various I/O-related | |
27 | * helper functions (reading/writing/checking/validating nodes) and implements | |
28 | * write-buffering support. Write buffers help to save space which otherwise | |
29 | * would have been wasted for padding to the nearest minimal I/O unit boundary. | |
30 | * Instead, data first goes to the write-buffer and is flushed when the | |
31 | * buffer is full or when it is not used for some time (by timer). This is | |
6f7ab6d4 | 32 | * similar to the mechanism is used by JFFS2. |
1e51764a | 33 | * |
6c7f74f7 AB |
34 | * UBIFS distinguishes between minimum write size (@c->min_io_size) and maximum |
35 | * write size (@c->max_write_size). The latter is the maximum amount of bytes | |
36 | * the underlying flash is able to program at a time, and writing in | |
37 | * @c->max_write_size units should presumably be faster. Obviously, | |
38 | * @c->min_io_size <= @c->max_write_size. Write-buffers are of | |
39 | * @c->max_write_size bytes in size for maximum performance. However, when a | |
40 | * write-buffer is flushed, only the portion of it (aligned to @c->min_io_size | |
41 | * boundary) which contains data is written, not the whole write-buffer, | |
42 | * because this is more space-efficient. | |
43 | * | |
44 | * This optimization adds few complications to the code. Indeed, on the one | |
45 | * hand, we want to write in optimal @c->max_write_size bytes chunks, which | |
46 | * also means aligning writes at the @c->max_write_size bytes offsets. On the | |
47 | * other hand, we do not want to waste space when synchronizing the write | |
48 | * buffer, so during synchronization we writes in smaller chunks. And this makes | |
49 | * the next write offset to be not aligned to @c->max_write_size bytes. So the | |
50 | * have to make sure that the write-buffer offset (@wbuf->offs) becomes aligned | |
51 | * to @c->max_write_size bytes again. We do this by temporarily shrinking | |
52 | * write-buffer size (@wbuf->size). | |
53 | * | |
1e51764a AB |
54 | * Write-buffers are defined by 'struct ubifs_wbuf' objects and protected by |
55 | * mutexes defined inside these objects. Since sometimes upper-level code | |
56 | * has to lock the write-buffer (e.g. journal space reservation code), many | |
57 | * functions related to write-buffers have "nolock" suffix which means that the | |
58 | * caller has to lock the write-buffer before calling this function. | |
59 | * | |
60 | * UBIFS stores nodes at 64 bit-aligned addresses. If the node length is not | |
61 | * aligned, UBIFS starts the next node from the aligned address, and the padded | |
62 | * bytes may contain any rubbish. In other words, UBIFS does not put padding | |
63 | * bytes in those small gaps. Common headers of nodes store real node lengths, | |
64 | * not aligned lengths. Indexing nodes also store real lengths in branches. | |
65 | * | |
66 | * UBIFS uses padding when it pads to the next min. I/O unit. In this case it | |
67 | * uses padding nodes or padding bytes, if the padding node does not fit. | |
68 | * | |
6c7f74f7 AB |
69 | * All UBIFS nodes are protected by CRC checksums and UBIFS checks CRC when |
70 | * they are read from the flash media. | |
1e51764a AB |
71 | */ |
72 | ||
73 | #include <linux/crc32.h> | |
5a0e3ad6 | 74 | #include <linux/slab.h> |
1e51764a AB |
75 | #include "ubifs.h" |
76 | ||
ff46d7b3 AH |
77 | /** |
78 | * ubifs_ro_mode - switch UBIFS to read read-only mode. | |
79 | * @c: UBIFS file-system description object | |
80 | * @err: error code which is the reason of switching to R/O mode | |
81 | */ | |
82 | void ubifs_ro_mode(struct ubifs_info *c, int err) | |
83 | { | |
2680d722 AB |
84 | if (!c->ro_error) { |
85 | c->ro_error = 1; | |
ccb3eba7 | 86 | c->no_chk_data_crc = 0; |
2fde99cb | 87 | c->vfs_sb->s_flags |= MS_RDONLY; |
235c362b | 88 | ubifs_warn(c, "switched to read-only mode, error %d", err); |
d033c98b | 89 | dump_stack(); |
ff46d7b3 AH |
90 | } |
91 | } | |
92 | ||
83cef708 AB |
93 | /* |
94 | * Below are simple wrappers over UBI I/O functions which include some | |
95 | * additional checks and UBIFS debugging stuff. See corresponding UBI function | |
96 | * for more information. | |
97 | */ | |
98 | ||
99 | int ubifs_leb_read(const struct ubifs_info *c, int lnum, void *buf, int offs, | |
100 | int len, int even_ebadmsg) | |
101 | { | |
102 | int err; | |
103 | ||
104 | err = ubi_read(c->ubi, lnum, buf, offs, len); | |
105 | /* | |
106 | * In case of %-EBADMSG print the error message only if the | |
107 | * @even_ebadmsg is true. | |
108 | */ | |
109 | if (err && (err != -EBADMSG || even_ebadmsg)) { | |
235c362b | 110 | ubifs_err(c, "reading %d bytes from LEB %d:%d failed, error %d", |
83cef708 | 111 | len, lnum, offs, err); |
7c46d0ae | 112 | dump_stack(); |
83cef708 AB |
113 | } |
114 | return err; | |
115 | } | |
116 | ||
117 | int ubifs_leb_write(struct ubifs_info *c, int lnum, const void *buf, int offs, | |
b36a261e | 118 | int len) |
83cef708 AB |
119 | { |
120 | int err; | |
121 | ||
122 | ubifs_assert(!c->ro_media && !c->ro_mount); | |
123 | if (c->ro_error) | |
124 | return -EROFS; | |
125 | if (!dbg_is_tst_rcvry(c)) | |
b36a261e | 126 | err = ubi_leb_write(c->ubi, lnum, buf, offs, len); |
83cef708 | 127 | else |
b36a261e | 128 | err = dbg_leb_write(c, lnum, buf, offs, len); |
83cef708 | 129 | if (err) { |
235c362b | 130 | ubifs_err(c, "writing %d bytes to LEB %d:%d failed, error %d", |
83cef708 AB |
131 | len, lnum, offs, err); |
132 | ubifs_ro_mode(c, err); | |
7c46d0ae | 133 | dump_stack(); |
83cef708 AB |
134 | } |
135 | return err; | |
136 | } | |
137 | ||
b36a261e | 138 | int ubifs_leb_change(struct ubifs_info *c, int lnum, const void *buf, int len) |
83cef708 AB |
139 | { |
140 | int err; | |
141 | ||
142 | ubifs_assert(!c->ro_media && !c->ro_mount); | |
143 | if (c->ro_error) | |
144 | return -EROFS; | |
145 | if (!dbg_is_tst_rcvry(c)) | |
b36a261e | 146 | err = ubi_leb_change(c->ubi, lnum, buf, len); |
83cef708 | 147 | else |
b36a261e | 148 | err = dbg_leb_change(c, lnum, buf, len); |
83cef708 | 149 | if (err) { |
235c362b | 150 | ubifs_err(c, "changing %d bytes in LEB %d failed, error %d", |
83cef708 AB |
151 | len, lnum, err); |
152 | ubifs_ro_mode(c, err); | |
7c46d0ae | 153 | dump_stack(); |
83cef708 AB |
154 | } |
155 | return err; | |
156 | } | |
157 | ||
158 | int ubifs_leb_unmap(struct ubifs_info *c, int lnum) | |
159 | { | |
160 | int err; | |
161 | ||
162 | ubifs_assert(!c->ro_media && !c->ro_mount); | |
163 | if (c->ro_error) | |
164 | return -EROFS; | |
165 | if (!dbg_is_tst_rcvry(c)) | |
166 | err = ubi_leb_unmap(c->ubi, lnum); | |
167 | else | |
f57cb188 | 168 | err = dbg_leb_unmap(c, lnum); |
83cef708 | 169 | if (err) { |
235c362b | 170 | ubifs_err(c, "unmap LEB %d failed, error %d", lnum, err); |
83cef708 | 171 | ubifs_ro_mode(c, err); |
7c46d0ae | 172 | dump_stack(); |
83cef708 AB |
173 | } |
174 | return err; | |
175 | } | |
176 | ||
b36a261e | 177 | int ubifs_leb_map(struct ubifs_info *c, int lnum) |
83cef708 AB |
178 | { |
179 | int err; | |
180 | ||
181 | ubifs_assert(!c->ro_media && !c->ro_mount); | |
182 | if (c->ro_error) | |
183 | return -EROFS; | |
184 | if (!dbg_is_tst_rcvry(c)) | |
b36a261e | 185 | err = ubi_leb_map(c->ubi, lnum); |
83cef708 | 186 | else |
b36a261e | 187 | err = dbg_leb_map(c, lnum); |
83cef708 | 188 | if (err) { |
235c362b | 189 | ubifs_err(c, "mapping LEB %d failed, error %d", lnum, err); |
83cef708 | 190 | ubifs_ro_mode(c, err); |
7c46d0ae | 191 | dump_stack(); |
83cef708 AB |
192 | } |
193 | return err; | |
194 | } | |
195 | ||
196 | int ubifs_is_mapped(const struct ubifs_info *c, int lnum) | |
197 | { | |
198 | int err; | |
199 | ||
200 | err = ubi_is_mapped(c->ubi, lnum); | |
201 | if (err < 0) { | |
235c362b | 202 | ubifs_err(c, "ubi_is_mapped failed for LEB %d, error %d", |
83cef708 | 203 | lnum, err); |
7c46d0ae | 204 | dump_stack(); |
83cef708 AB |
205 | } |
206 | return err; | |
207 | } | |
208 | ||
1e51764a AB |
209 | /** |
210 | * ubifs_check_node - check node. | |
211 | * @c: UBIFS file-system description object | |
212 | * @buf: node to check | |
213 | * @lnum: logical eraseblock number | |
214 | * @offs: offset within the logical eraseblock | |
215 | * @quiet: print no messages | |
6f7ab6d4 | 216 | * @must_chk_crc: indicates whether to always check the CRC |
1e51764a AB |
217 | * |
218 | * This function checks node magic number and CRC checksum. This function also | |
219 | * validates node length to prevent UBIFS from becoming crazy when an attacker | |
220 | * feeds it a file-system image with incorrect nodes. For example, too large | |
221 | * node length in the common header could cause UBIFS to read memory outside of | |
222 | * allocated buffer when checking the CRC checksum. | |
223 | * | |
6f7ab6d4 AB |
224 | * This function may skip data nodes CRC checking if @c->no_chk_data_crc is |
225 | * true, which is controlled by corresponding UBIFS mount option. However, if | |
226 | * @must_chk_crc is true, then @c->no_chk_data_crc is ignored and CRC is | |
18d1d7fb AB |
227 | * checked. Similarly, if @c->mounting or @c->remounting_rw is true (we are |
228 | * mounting or re-mounting to R/W mode), @c->no_chk_data_crc is ignored and CRC | |
229 | * is checked. This is because during mounting or re-mounting from R/O mode to | |
230 | * R/W mode we may read journal nodes (when replying the journal or doing the | |
231 | * recovery) and the journal nodes may potentially be corrupted, so checking is | |
232 | * required. | |
6f7ab6d4 AB |
233 | * |
234 | * This function returns zero in case of success and %-EUCLEAN in case of bad | |
235 | * CRC or magic. | |
1e51764a AB |
236 | */ |
237 | int ubifs_check_node(const struct ubifs_info *c, const void *buf, int lnum, | |
6f7ab6d4 | 238 | int offs, int quiet, int must_chk_crc) |
1e51764a AB |
239 | { |
240 | int err = -EINVAL, type, node_len; | |
241 | uint32_t crc, node_crc, magic; | |
242 | const struct ubifs_ch *ch = buf; | |
243 | ||
244 | ubifs_assert(lnum >= 0 && lnum < c->leb_cnt && offs >= 0); | |
245 | ubifs_assert(!(offs & 7) && offs < c->leb_size); | |
246 | ||
247 | magic = le32_to_cpu(ch->magic); | |
248 | if (magic != UBIFS_NODE_MAGIC) { | |
249 | if (!quiet) | |
235c362b | 250 | ubifs_err(c, "bad magic %#08x, expected %#08x", |
1e51764a AB |
251 | magic, UBIFS_NODE_MAGIC); |
252 | err = -EUCLEAN; | |
253 | goto out; | |
254 | } | |
255 | ||
256 | type = ch->node_type; | |
257 | if (type < 0 || type >= UBIFS_NODE_TYPES_CNT) { | |
258 | if (!quiet) | |
235c362b | 259 | ubifs_err(c, "bad node type %d", type); |
1e51764a AB |
260 | goto out; |
261 | } | |
262 | ||
263 | node_len = le32_to_cpu(ch->len); | |
264 | if (node_len + offs > c->leb_size) | |
265 | goto out_len; | |
266 | ||
267 | if (c->ranges[type].max_len == 0) { | |
268 | if (node_len != c->ranges[type].len) | |
269 | goto out_len; | |
270 | } else if (node_len < c->ranges[type].min_len || | |
271 | node_len > c->ranges[type].max_len) | |
272 | goto out_len; | |
273 | ||
18d1d7fb AB |
274 | if (!must_chk_crc && type == UBIFS_DATA_NODE && !c->mounting && |
275 | !c->remounting_rw && c->no_chk_data_crc) | |
6f7ab6d4 | 276 | return 0; |
2953e73f | 277 | |
1e51764a AB |
278 | crc = crc32(UBIFS_CRC32_INIT, buf + 8, node_len - 8); |
279 | node_crc = le32_to_cpu(ch->crc); | |
280 | if (crc != node_crc) { | |
281 | if (!quiet) | |
235c362b | 282 | ubifs_err(c, "bad CRC: calculated %#08x, read %#08x", |
1e51764a AB |
283 | crc, node_crc); |
284 | err = -EUCLEAN; | |
285 | goto out; | |
286 | } | |
287 | ||
288 | return 0; | |
289 | ||
290 | out_len: | |
291 | if (!quiet) | |
235c362b | 292 | ubifs_err(c, "bad node length %d", node_len); |
1e51764a AB |
293 | out: |
294 | if (!quiet) { | |
235c362b | 295 | ubifs_err(c, "bad node at LEB %d:%d", lnum, offs); |
edf6be24 | 296 | ubifs_dump_node(c, buf); |
7c46d0ae | 297 | dump_stack(); |
1e51764a AB |
298 | } |
299 | return err; | |
300 | } | |
301 | ||
302 | /** | |
303 | * ubifs_pad - pad flash space. | |
304 | * @c: UBIFS file-system description object | |
305 | * @buf: buffer to put padding to | |
306 | * @pad: how many bytes to pad | |
307 | * | |
308 | * The flash media obliges us to write only in chunks of %c->min_io_size and | |
309 | * when we have to write less data we add padding node to the write-buffer and | |
310 | * pad it to the next minimal I/O unit's boundary. Padding nodes help when the | |
311 | * media is being scanned. If the amount of wasted space is not enough to fit a | |
312 | * padding node which takes %UBIFS_PAD_NODE_SZ bytes, we write padding bytes | |
313 | * pattern (%UBIFS_PADDING_BYTE). | |
314 | * | |
315 | * Padding nodes are also used to fill gaps when the "commit-in-gaps" method is | |
316 | * used. | |
317 | */ | |
318 | void ubifs_pad(const struct ubifs_info *c, void *buf, int pad) | |
319 | { | |
320 | uint32_t crc; | |
321 | ||
322 | ubifs_assert(pad >= 0 && !(pad & 7)); | |
323 | ||
324 | if (pad >= UBIFS_PAD_NODE_SZ) { | |
325 | struct ubifs_ch *ch = buf; | |
326 | struct ubifs_pad_node *pad_node = buf; | |
327 | ||
328 | ch->magic = cpu_to_le32(UBIFS_NODE_MAGIC); | |
329 | ch->node_type = UBIFS_PAD_NODE; | |
330 | ch->group_type = UBIFS_NO_NODE_GROUP; | |
331 | ch->padding[0] = ch->padding[1] = 0; | |
332 | ch->sqnum = 0; | |
333 | ch->len = cpu_to_le32(UBIFS_PAD_NODE_SZ); | |
334 | pad -= UBIFS_PAD_NODE_SZ; | |
335 | pad_node->pad_len = cpu_to_le32(pad); | |
336 | crc = crc32(UBIFS_CRC32_INIT, buf + 8, UBIFS_PAD_NODE_SZ - 8); | |
337 | ch->crc = cpu_to_le32(crc); | |
338 | memset(buf + UBIFS_PAD_NODE_SZ, 0, pad); | |
339 | } else if (pad > 0) | |
340 | /* Too little space, padding node won't fit */ | |
341 | memset(buf, UBIFS_PADDING_BYTE, pad); | |
342 | } | |
343 | ||
344 | /** | |
345 | * next_sqnum - get next sequence number. | |
346 | * @c: UBIFS file-system description object | |
347 | */ | |
348 | static unsigned long long next_sqnum(struct ubifs_info *c) | |
349 | { | |
350 | unsigned long long sqnum; | |
351 | ||
352 | spin_lock(&c->cnt_lock); | |
353 | sqnum = ++c->max_sqnum; | |
354 | spin_unlock(&c->cnt_lock); | |
355 | ||
356 | if (unlikely(sqnum >= SQNUM_WARN_WATERMARK)) { | |
357 | if (sqnum >= SQNUM_WATERMARK) { | |
235c362b | 358 | ubifs_err(c, "sequence number overflow %llu, end of life", |
1e51764a AB |
359 | sqnum); |
360 | ubifs_ro_mode(c, -EINVAL); | |
361 | } | |
235c362b | 362 | ubifs_warn(c, "running out of sequence numbers, end of life soon"); |
1e51764a AB |
363 | } |
364 | ||
365 | return sqnum; | |
366 | } | |
367 | ||
368 | /** | |
369 | * ubifs_prepare_node - prepare node to be written to flash. | |
370 | * @c: UBIFS file-system description object | |
371 | * @node: the node to pad | |
372 | * @len: node length | |
373 | * @pad: if the buffer has to be padded | |
374 | * | |
375 | * This function prepares node at @node to be written to the media - it | |
376 | * calculates node CRC, fills the common header, and adds proper padding up to | |
377 | * the next minimum I/O unit if @pad is not zero. | |
378 | */ | |
379 | void ubifs_prepare_node(struct ubifs_info *c, void *node, int len, int pad) | |
380 | { | |
381 | uint32_t crc; | |
382 | struct ubifs_ch *ch = node; | |
383 | unsigned long long sqnum = next_sqnum(c); | |
384 | ||
385 | ubifs_assert(len >= UBIFS_CH_SZ); | |
386 | ||
387 | ch->magic = cpu_to_le32(UBIFS_NODE_MAGIC); | |
388 | ch->len = cpu_to_le32(len); | |
389 | ch->group_type = UBIFS_NO_NODE_GROUP; | |
390 | ch->sqnum = cpu_to_le64(sqnum); | |
391 | ch->padding[0] = ch->padding[1] = 0; | |
392 | crc = crc32(UBIFS_CRC32_INIT, node + 8, len - 8); | |
393 | ch->crc = cpu_to_le32(crc); | |
394 | ||
395 | if (pad) { | |
396 | len = ALIGN(len, 8); | |
397 | pad = ALIGN(len, c->min_io_size) - len; | |
398 | ubifs_pad(c, node + len, pad); | |
399 | } | |
400 | } | |
401 | ||
402 | /** | |
403 | * ubifs_prep_grp_node - prepare node of a group to be written to flash. | |
404 | * @c: UBIFS file-system description object | |
405 | * @node: the node to pad | |
406 | * @len: node length | |
407 | * @last: indicates the last node of the group | |
408 | * | |
409 | * This function prepares node at @node to be written to the media - it | |
410 | * calculates node CRC and fills the common header. | |
411 | */ | |
412 | void ubifs_prep_grp_node(struct ubifs_info *c, void *node, int len, int last) | |
413 | { | |
414 | uint32_t crc; | |
415 | struct ubifs_ch *ch = node; | |
416 | unsigned long long sqnum = next_sqnum(c); | |
417 | ||
418 | ubifs_assert(len >= UBIFS_CH_SZ); | |
419 | ||
420 | ch->magic = cpu_to_le32(UBIFS_NODE_MAGIC); | |
421 | ch->len = cpu_to_le32(len); | |
422 | if (last) | |
423 | ch->group_type = UBIFS_LAST_OF_NODE_GROUP; | |
424 | else | |
425 | ch->group_type = UBIFS_IN_NODE_GROUP; | |
426 | ch->sqnum = cpu_to_le64(sqnum); | |
427 | ch->padding[0] = ch->padding[1] = 0; | |
428 | crc = crc32(UBIFS_CRC32_INIT, node + 8, len - 8); | |
429 | ch->crc = cpu_to_le32(crc); | |
430 | } | |
431 | ||
432 | /** | |
433 | * wbuf_timer_callback - write-buffer timer callback function. | |
39274a1e | 434 | * @timer: timer data (write-buffer descriptor) |
1e51764a AB |
435 | * |
436 | * This function is called when the write-buffer timer expires. | |
437 | */ | |
f2c5dbd7 | 438 | static enum hrtimer_restart wbuf_timer_callback_nolock(struct hrtimer *timer) |
1e51764a | 439 | { |
f2c5dbd7 | 440 | struct ubifs_wbuf *wbuf = container_of(timer, struct ubifs_wbuf, timer); |
1e51764a | 441 | |
77a7ae58 | 442 | dbg_io("jhead %s", dbg_jhead(wbuf->jhead)); |
1e51764a AB |
443 | wbuf->need_sync = 1; |
444 | wbuf->c->need_wbuf_sync = 1; | |
445 | ubifs_wake_up_bgt(wbuf->c); | |
f2c5dbd7 | 446 | return HRTIMER_NORESTART; |
1e51764a AB |
447 | } |
448 | ||
449 | /** | |
450 | * new_wbuf_timer - start new write-buffer timer. | |
451 | * @wbuf: write-buffer descriptor | |
452 | */ | |
453 | static void new_wbuf_timer_nolock(struct ubifs_wbuf *wbuf) | |
454 | { | |
1b7fc2c0 RM |
455 | ktime_t softlimit = ms_to_ktime(dirty_writeback_interval * 10); |
456 | unsigned long long delta = dirty_writeback_interval; | |
854826c9 | 457 | |
1b7fc2c0 RM |
458 | /* centi to milli, milli to nano, then 10% */ |
459 | delta *= 10ULL * NSEC_PER_MSEC / 10ULL; | |
854826c9 | 460 | |
f2c5dbd7 | 461 | ubifs_assert(!hrtimer_active(&wbuf->timer)); |
854826c9 | 462 | ubifs_assert(delta <= ULONG_MAX); |
1e51764a | 463 | |
0b335b9d | 464 | if (wbuf->no_timer) |
1e51764a | 465 | return; |
77a7ae58 AB |
466 | dbg_io("set timer for jhead %s, %llu-%llu millisecs", |
467 | dbg_jhead(wbuf->jhead), | |
854826c9 RM |
468 | div_u64(ktime_to_ns(softlimit), USEC_PER_SEC), |
469 | div_u64(ktime_to_ns(softlimit) + delta, USEC_PER_SEC)); | |
470 | hrtimer_start_range_ns(&wbuf->timer, softlimit, delta, | |
f2c5dbd7 | 471 | HRTIMER_MODE_REL); |
1e51764a AB |
472 | } |
473 | ||
474 | /** | |
475 | * cancel_wbuf_timer - cancel write-buffer timer. | |
476 | * @wbuf: write-buffer descriptor | |
477 | */ | |
478 | static void cancel_wbuf_timer_nolock(struct ubifs_wbuf *wbuf) | |
479 | { | |
0b335b9d AB |
480 | if (wbuf->no_timer) |
481 | return; | |
1e51764a | 482 | wbuf->need_sync = 0; |
f2c5dbd7 | 483 | hrtimer_cancel(&wbuf->timer); |
1e51764a AB |
484 | } |
485 | ||
486 | /** | |
487 | * ubifs_wbuf_sync_nolock - synchronize write-buffer. | |
488 | * @wbuf: write-buffer to synchronize | |
489 | * | |
490 | * This function synchronizes write-buffer @buf and returns zero in case of | |
491 | * success or a negative error code in case of failure. | |
6c7f74f7 AB |
492 | * |
493 | * Note, although write-buffers are of @c->max_write_size, this function does | |
494 | * not necessarily writes all @c->max_write_size bytes to the flash. Instead, | |
495 | * if the write-buffer is only partially filled with data, only the used part | |
496 | * of the write-buffer (aligned on @c->min_io_size boundary) is synchronized. | |
497 | * This way we waste less space. | |
1e51764a AB |
498 | */ |
499 | int ubifs_wbuf_sync_nolock(struct ubifs_wbuf *wbuf) | |
500 | { | |
501 | struct ubifs_info *c = wbuf->c; | |
6c7f74f7 | 502 | int err, dirt, sync_len; |
1e51764a AB |
503 | |
504 | cancel_wbuf_timer_nolock(wbuf); | |
505 | if (!wbuf->used || wbuf->lnum == -1) | |
506 | /* Write-buffer is empty or not seeked */ | |
507 | return 0; | |
508 | ||
77a7ae58 AB |
509 | dbg_io("LEB %d:%d, %d bytes, jhead %s", |
510 | wbuf->lnum, wbuf->offs, wbuf->used, dbg_jhead(wbuf->jhead)); | |
1e51764a | 511 | ubifs_assert(!(wbuf->avail & 7)); |
3c89f396 AB |
512 | ubifs_assert(wbuf->offs + wbuf->size <= c->leb_size); |
513 | ubifs_assert(wbuf->size >= c->min_io_size); | |
514 | ubifs_assert(wbuf->size <= c->max_write_size); | |
515 | ubifs_assert(wbuf->size % c->min_io_size == 0); | |
2ef13294 | 516 | ubifs_assert(!c->ro_media && !c->ro_mount); |
6c7f74f7 | 517 | if (c->leb_size - wbuf->offs >= c->max_write_size) |
c4361570 | 518 | ubifs_assert(!((wbuf->offs + wbuf->size) % c->max_write_size)); |
1e51764a | 519 | |
2680d722 | 520 | if (c->ro_error) |
1e51764a AB |
521 | return -EROFS; |
522 | ||
6c7f74f7 AB |
523 | /* |
524 | * Do not write whole write buffer but write only the minimum necessary | |
525 | * amount of min. I/O units. | |
526 | */ | |
527 | sync_len = ALIGN(wbuf->used, c->min_io_size); | |
528 | dirt = sync_len - wbuf->used; | |
529 | if (dirt) | |
530 | ubifs_pad(c, wbuf->buf + wbuf->used, dirt); | |
b36a261e | 531 | err = ubifs_leb_write(c, wbuf->lnum, wbuf->buf, wbuf->offs, sync_len); |
987226a5 | 532 | if (err) |
1e51764a | 533 | return err; |
1e51764a | 534 | |
1e51764a | 535 | spin_lock(&wbuf->lock); |
6c7f74f7 AB |
536 | wbuf->offs += sync_len; |
537 | /* | |
538 | * Now @wbuf->offs is not necessarily aligned to @c->max_write_size. | |
539 | * But our goal is to optimize writes and make sure we write in | |
540 | * @c->max_write_size chunks and to @c->max_write_size-aligned offset. | |
541 | * Thus, if @wbuf->offs is not aligned to @c->max_write_size now, make | |
542 | * sure that @wbuf->offs + @wbuf->size is aligned to | |
543 | * @c->max_write_size. This way we make sure that after next | |
544 | * write-buffer flush we are again at the optimal offset (aligned to | |
545 | * @c->max_write_size). | |
546 | */ | |
547 | if (c->leb_size - wbuf->offs < c->max_write_size) | |
548 | wbuf->size = c->leb_size - wbuf->offs; | |
549 | else if (wbuf->offs & (c->max_write_size - 1)) | |
550 | wbuf->size = ALIGN(wbuf->offs, c->max_write_size) - wbuf->offs; | |
551 | else | |
552 | wbuf->size = c->max_write_size; | |
553 | wbuf->avail = wbuf->size; | |
1e51764a AB |
554 | wbuf->used = 0; |
555 | wbuf->next_ino = 0; | |
556 | spin_unlock(&wbuf->lock); | |
557 | ||
558 | if (wbuf->sync_callback) | |
559 | err = wbuf->sync_callback(c, wbuf->lnum, | |
560 | c->leb_size - wbuf->offs, dirt); | |
561 | return err; | |
562 | } | |
563 | ||
564 | /** | |
565 | * ubifs_wbuf_seek_nolock - seek write-buffer. | |
566 | * @wbuf: write-buffer | |
567 | * @lnum: logical eraseblock number to seek to | |
568 | * @offs: logical eraseblock offset to seek to | |
1e51764a | 569 | * |
cb54ef8b | 570 | * This function targets the write-buffer to logical eraseblock @lnum:@offs. |
cb14a184 AB |
571 | * The write-buffer has to be empty. Returns zero in case of success and a |
572 | * negative error code in case of failure. | |
1e51764a | 573 | */ |
b36a261e | 574 | int ubifs_wbuf_seek_nolock(struct ubifs_wbuf *wbuf, int lnum, int offs) |
1e51764a AB |
575 | { |
576 | const struct ubifs_info *c = wbuf->c; | |
577 | ||
77a7ae58 | 578 | dbg_io("LEB %d:%d, jhead %s", lnum, offs, dbg_jhead(wbuf->jhead)); |
1e51764a AB |
579 | ubifs_assert(lnum >= 0 && lnum < c->leb_cnt); |
580 | ubifs_assert(offs >= 0 && offs <= c->leb_size); | |
581 | ubifs_assert(offs % c->min_io_size == 0 && !(offs & 7)); | |
582 | ubifs_assert(lnum != wbuf->lnum); | |
cb14a184 | 583 | ubifs_assert(wbuf->used == 0); |
1e51764a AB |
584 | |
585 | spin_lock(&wbuf->lock); | |
586 | wbuf->lnum = lnum; | |
587 | wbuf->offs = offs; | |
6c7f74f7 AB |
588 | if (c->leb_size - wbuf->offs < c->max_write_size) |
589 | wbuf->size = c->leb_size - wbuf->offs; | |
590 | else if (wbuf->offs & (c->max_write_size - 1)) | |
591 | wbuf->size = ALIGN(wbuf->offs, c->max_write_size) - wbuf->offs; | |
592 | else | |
593 | wbuf->size = c->max_write_size; | |
594 | wbuf->avail = wbuf->size; | |
1e51764a AB |
595 | wbuf->used = 0; |
596 | spin_unlock(&wbuf->lock); | |
1e51764a AB |
597 | |
598 | return 0; | |
599 | } | |
600 | ||
601 | /** | |
602 | * ubifs_bg_wbufs_sync - synchronize write-buffers. | |
603 | * @c: UBIFS file-system description object | |
604 | * | |
605 | * This function is called by background thread to synchronize write-buffers. | |
606 | * Returns zero in case of success and a negative error code in case of | |
607 | * failure. | |
608 | */ | |
609 | int ubifs_bg_wbufs_sync(struct ubifs_info *c) | |
610 | { | |
611 | int err, i; | |
612 | ||
2ef13294 | 613 | ubifs_assert(!c->ro_media && !c->ro_mount); |
1e51764a AB |
614 | if (!c->need_wbuf_sync) |
615 | return 0; | |
616 | c->need_wbuf_sync = 0; | |
617 | ||
2680d722 | 618 | if (c->ro_error) { |
1e51764a AB |
619 | err = -EROFS; |
620 | goto out_timers; | |
621 | } | |
622 | ||
623 | dbg_io("synchronize"); | |
624 | for (i = 0; i < c->jhead_cnt; i++) { | |
625 | struct ubifs_wbuf *wbuf = &c->jheads[i].wbuf; | |
626 | ||
627 | cond_resched(); | |
628 | ||
629 | /* | |
630 | * If the mutex is locked then wbuf is being changed, so | |
631 | * synchronization is not necessary. | |
632 | */ | |
633 | if (mutex_is_locked(&wbuf->io_mutex)) | |
634 | continue; | |
635 | ||
636 | mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead); | |
637 | if (!wbuf->need_sync) { | |
638 | mutex_unlock(&wbuf->io_mutex); | |
639 | continue; | |
640 | } | |
641 | ||
642 | err = ubifs_wbuf_sync_nolock(wbuf); | |
643 | mutex_unlock(&wbuf->io_mutex); | |
644 | if (err) { | |
235c362b | 645 | ubifs_err(c, "cannot sync write-buffer, error %d", err); |
1e51764a AB |
646 | ubifs_ro_mode(c, err); |
647 | goto out_timers; | |
648 | } | |
649 | } | |
650 | ||
651 | return 0; | |
652 | ||
653 | out_timers: | |
654 | /* Cancel all timers to prevent repeated errors */ | |
655 | for (i = 0; i < c->jhead_cnt; i++) { | |
656 | struct ubifs_wbuf *wbuf = &c->jheads[i].wbuf; | |
657 | ||
658 | mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead); | |
659 | cancel_wbuf_timer_nolock(wbuf); | |
660 | mutex_unlock(&wbuf->io_mutex); | |
661 | } | |
662 | return err; | |
663 | } | |
664 | ||
665 | /** | |
666 | * ubifs_wbuf_write_nolock - write data to flash via write-buffer. | |
667 | * @wbuf: write-buffer | |
668 | * @buf: node to write | |
669 | * @len: node length | |
670 | * | |
671 | * This function writes data to flash via write-buffer @wbuf. This means that | |
672 | * the last piece of the node won't reach the flash media immediately if it | |
6c7f74f7 AB |
673 | * does not take whole max. write unit (@c->max_write_size). Instead, the node |
674 | * will sit in RAM until the write-buffer is synchronized (e.g., by timer, or | |
675 | * because more data are appended to the write-buffer). | |
1e51764a AB |
676 | * |
677 | * This function returns zero in case of success and a negative error code in | |
678 | * case of failure. If the node cannot be written because there is no more | |
679 | * space in this logical eraseblock, %-ENOSPC is returned. | |
680 | */ | |
681 | int ubifs_wbuf_write_nolock(struct ubifs_wbuf *wbuf, void *buf, int len) | |
682 | { | |
683 | struct ubifs_info *c = wbuf->c; | |
12f33891 | 684 | int err, written, n, aligned_len = ALIGN(len, 8); |
1e51764a | 685 | |
77a7ae58 AB |
686 | dbg_io("%d bytes (%s) to jhead %s wbuf at LEB %d:%d", len, |
687 | dbg_ntype(((struct ubifs_ch *)buf)->node_type), | |
688 | dbg_jhead(wbuf->jhead), wbuf->lnum, wbuf->offs + wbuf->used); | |
1e51764a AB |
689 | ubifs_assert(len > 0 && wbuf->lnum >= 0 && wbuf->lnum < c->leb_cnt); |
690 | ubifs_assert(wbuf->offs >= 0 && wbuf->offs % c->min_io_size == 0); | |
691 | ubifs_assert(!(wbuf->offs & 7) && wbuf->offs <= c->leb_size); | |
3c89f396 AB |
692 | ubifs_assert(wbuf->avail > 0 && wbuf->avail <= wbuf->size); |
693 | ubifs_assert(wbuf->size >= c->min_io_size); | |
694 | ubifs_assert(wbuf->size <= c->max_write_size); | |
695 | ubifs_assert(wbuf->size % c->min_io_size == 0); | |
1e51764a | 696 | ubifs_assert(mutex_is_locked(&wbuf->io_mutex)); |
2ef13294 | 697 | ubifs_assert(!c->ro_media && !c->ro_mount); |
4f1ab9b0 | 698 | ubifs_assert(!c->space_fixup); |
6c7f74f7 | 699 | if (c->leb_size - wbuf->offs >= c->max_write_size) |
c4361570 | 700 | ubifs_assert(!((wbuf->offs + wbuf->size) % c->max_write_size)); |
1e51764a AB |
701 | |
702 | if (c->leb_size - wbuf->offs - wbuf->used < aligned_len) { | |
703 | err = -ENOSPC; | |
704 | goto out; | |
705 | } | |
706 | ||
707 | cancel_wbuf_timer_nolock(wbuf); | |
708 | ||
2680d722 | 709 | if (c->ro_error) |
1e51764a AB |
710 | return -EROFS; |
711 | ||
712 | if (aligned_len <= wbuf->avail) { | |
713 | /* | |
714 | * The node is not very large and fits entirely within | |
715 | * write-buffer. | |
716 | */ | |
717 | memcpy(wbuf->buf + wbuf->used, buf, len); | |
718 | ||
719 | if (aligned_len == wbuf->avail) { | |
77a7ae58 AB |
720 | dbg_io("flush jhead %s wbuf to LEB %d:%d", |
721 | dbg_jhead(wbuf->jhead), wbuf->lnum, wbuf->offs); | |
987226a5 | 722 | err = ubifs_leb_write(c, wbuf->lnum, wbuf->buf, |
b36a261e | 723 | wbuf->offs, wbuf->size); |
1e51764a AB |
724 | if (err) |
725 | goto out; | |
726 | ||
727 | spin_lock(&wbuf->lock); | |
6c7f74f7 AB |
728 | wbuf->offs += wbuf->size; |
729 | if (c->leb_size - wbuf->offs >= c->max_write_size) | |
730 | wbuf->size = c->max_write_size; | |
731 | else | |
732 | wbuf->size = c->leb_size - wbuf->offs; | |
733 | wbuf->avail = wbuf->size; | |
1e51764a AB |
734 | wbuf->used = 0; |
735 | wbuf->next_ino = 0; | |
736 | spin_unlock(&wbuf->lock); | |
737 | } else { | |
738 | spin_lock(&wbuf->lock); | |
739 | wbuf->avail -= aligned_len; | |
740 | wbuf->used += aligned_len; | |
741 | spin_unlock(&wbuf->lock); | |
742 | } | |
743 | ||
744 | goto exit; | |
745 | } | |
746 | ||
6c7f74f7 | 747 | written = 0; |
1e51764a | 748 | |
6c7f74f7 AB |
749 | if (wbuf->used) { |
750 | /* | |
751 | * The node is large enough and does not fit entirely within | |
752 | * current available space. We have to fill and flush | |
753 | * write-buffer and switch to the next max. write unit. | |
754 | */ | |
755 | dbg_io("flush jhead %s wbuf to LEB %d:%d", | |
756 | dbg_jhead(wbuf->jhead), wbuf->lnum, wbuf->offs); | |
757 | memcpy(wbuf->buf + wbuf->used, buf, wbuf->avail); | |
987226a5 | 758 | err = ubifs_leb_write(c, wbuf->lnum, wbuf->buf, wbuf->offs, |
b36a261e | 759 | wbuf->size); |
6c7f74f7 AB |
760 | if (err) |
761 | goto out; | |
762 | ||
12f33891 | 763 | wbuf->offs += wbuf->size; |
6c7f74f7 AB |
764 | len -= wbuf->avail; |
765 | aligned_len -= wbuf->avail; | |
766 | written += wbuf->avail; | |
767 | } else if (wbuf->offs & (c->max_write_size - 1)) { | |
768 | /* | |
769 | * The write-buffer offset is not aligned to | |
770 | * @c->max_write_size and @wbuf->size is less than | |
771 | * @c->max_write_size. Write @wbuf->size bytes to make sure the | |
772 | * following writes are done in optimal @c->max_write_size | |
773 | * chunks. | |
774 | */ | |
775 | dbg_io("write %d bytes to LEB %d:%d", | |
776 | wbuf->size, wbuf->lnum, wbuf->offs); | |
987226a5 | 777 | err = ubifs_leb_write(c, wbuf->lnum, buf, wbuf->offs, |
b36a261e | 778 | wbuf->size); |
6c7f74f7 AB |
779 | if (err) |
780 | goto out; | |
781 | ||
12f33891 | 782 | wbuf->offs += wbuf->size; |
6c7f74f7 AB |
783 | len -= wbuf->size; |
784 | aligned_len -= wbuf->size; | |
785 | written += wbuf->size; | |
786 | } | |
1e51764a AB |
787 | |
788 | /* | |
6c7f74f7 AB |
789 | * The remaining data may take more whole max. write units, so write the |
790 | * remains multiple to max. write unit size directly to the flash media. | |
1e51764a AB |
791 | * We align node length to 8-byte boundary because we anyway flash wbuf |
792 | * if the remaining space is less than 8 bytes. | |
793 | */ | |
6c7f74f7 | 794 | n = aligned_len >> c->max_write_shift; |
1e51764a | 795 | if (n) { |
6c7f74f7 | 796 | n <<= c->max_write_shift; |
12f33891 AB |
797 | dbg_io("write %d bytes to LEB %d:%d", n, wbuf->lnum, |
798 | wbuf->offs); | |
987226a5 | 799 | err = ubifs_leb_write(c, wbuf->lnum, buf + written, |
b36a261e | 800 | wbuf->offs, n); |
1e51764a AB |
801 | if (err) |
802 | goto out; | |
12f33891 | 803 | wbuf->offs += n; |
1e51764a AB |
804 | aligned_len -= n; |
805 | len -= n; | |
806 | written += n; | |
807 | } | |
808 | ||
809 | spin_lock(&wbuf->lock); | |
810 | if (aligned_len) | |
811 | /* | |
812 | * And now we have what's left and what does not take whole | |
6c7f74f7 | 813 | * max. write unit, so write it to the write-buffer and we are |
1e51764a AB |
814 | * done. |
815 | */ | |
816 | memcpy(wbuf->buf, buf + written, len); | |
817 | ||
6c7f74f7 AB |
818 | if (c->leb_size - wbuf->offs >= c->max_write_size) |
819 | wbuf->size = c->max_write_size; | |
820 | else | |
821 | wbuf->size = c->leb_size - wbuf->offs; | |
822 | wbuf->avail = wbuf->size - aligned_len; | |
1e51764a | 823 | wbuf->used = aligned_len; |
1e51764a AB |
824 | wbuf->next_ino = 0; |
825 | spin_unlock(&wbuf->lock); | |
826 | ||
827 | exit: | |
828 | if (wbuf->sync_callback) { | |
829 | int free = c->leb_size - wbuf->offs - wbuf->used; | |
830 | ||
831 | err = wbuf->sync_callback(c, wbuf->lnum, free, 0); | |
832 | if (err) | |
833 | goto out; | |
834 | } | |
835 | ||
836 | if (wbuf->used) | |
837 | new_wbuf_timer_nolock(wbuf); | |
838 | ||
839 | return 0; | |
840 | ||
841 | out: | |
235c362b | 842 | ubifs_err(c, "cannot write %d bytes to LEB %d:%d, error %d", |
1e51764a | 843 | len, wbuf->lnum, wbuf->offs, err); |
edf6be24 | 844 | ubifs_dump_node(c, buf); |
7c46d0ae | 845 | dump_stack(); |
edf6be24 | 846 | ubifs_dump_leb(c, wbuf->lnum); |
1e51764a AB |
847 | return err; |
848 | } | |
849 | ||
850 | /** | |
851 | * ubifs_write_node - write node to the media. | |
852 | * @c: UBIFS file-system description object | |
853 | * @buf: the node to write | |
854 | * @len: node length | |
855 | * @lnum: logical eraseblock number | |
856 | * @offs: offset within the logical eraseblock | |
1e51764a AB |
857 | * |
858 | * This function automatically fills node magic number, assigns sequence | |
859 | * number, and calculates node CRC checksum. The length of the @buf buffer has | |
860 | * to be aligned to the minimal I/O unit size. This function automatically | |
861 | * appends padding node and padding bytes if needed. Returns zero in case of | |
862 | * success and a negative error code in case of failure. | |
863 | */ | |
864 | int ubifs_write_node(struct ubifs_info *c, void *buf, int len, int lnum, | |
b36a261e | 865 | int offs) |
1e51764a AB |
866 | { |
867 | int err, buf_len = ALIGN(len, c->min_io_size); | |
868 | ||
869 | dbg_io("LEB %d:%d, %s, length %d (aligned %d)", | |
870 | lnum, offs, dbg_ntype(((struct ubifs_ch *)buf)->node_type), len, | |
871 | buf_len); | |
872 | ubifs_assert(lnum >= 0 && lnum < c->leb_cnt && offs >= 0); | |
873 | ubifs_assert(offs % c->min_io_size == 0 && offs < c->leb_size); | |
2ef13294 | 874 | ubifs_assert(!c->ro_media && !c->ro_mount); |
4f1ab9b0 | 875 | ubifs_assert(!c->space_fixup); |
1e51764a | 876 | |
2680d722 | 877 | if (c->ro_error) |
1e51764a AB |
878 | return -EROFS; |
879 | ||
880 | ubifs_prepare_node(c, buf, len, 1); | |
b36a261e | 881 | err = ubifs_leb_write(c, lnum, buf, offs, buf_len); |
987226a5 | 882 | if (err) |
edf6be24 | 883 | ubifs_dump_node(c, buf); |
1e51764a AB |
884 | |
885 | return err; | |
886 | } | |
887 | ||
888 | /** | |
889 | * ubifs_read_node_wbuf - read node from the media or write-buffer. | |
890 | * @wbuf: wbuf to check for un-written data | |
891 | * @buf: buffer to read to | |
892 | * @type: node type | |
893 | * @len: node length | |
894 | * @lnum: logical eraseblock number | |
895 | * @offs: offset within the logical eraseblock | |
896 | * | |
897 | * This function reads a node of known type and length, checks it and stores | |
898 | * in @buf. If the node partially or fully sits in the write-buffer, this | |
899 | * function takes data from the buffer, otherwise it reads the flash media. | |
900 | * Returns zero in case of success, %-EUCLEAN if CRC mismatched and a negative | |
901 | * error code in case of failure. | |
902 | */ | |
903 | int ubifs_read_node_wbuf(struct ubifs_wbuf *wbuf, void *buf, int type, int len, | |
904 | int lnum, int offs) | |
905 | { | |
906 | const struct ubifs_info *c = wbuf->c; | |
907 | int err, rlen, overlap; | |
908 | struct ubifs_ch *ch = buf; | |
909 | ||
77a7ae58 AB |
910 | dbg_io("LEB %d:%d, %s, length %d, jhead %s", lnum, offs, |
911 | dbg_ntype(type), len, dbg_jhead(wbuf->jhead)); | |
1e51764a AB |
912 | ubifs_assert(wbuf && lnum >= 0 && lnum < c->leb_cnt && offs >= 0); |
913 | ubifs_assert(!(offs & 7) && offs < c->leb_size); | |
914 | ubifs_assert(type >= 0 && type < UBIFS_NODE_TYPES_CNT); | |
915 | ||
916 | spin_lock(&wbuf->lock); | |
917 | overlap = (lnum == wbuf->lnum && offs + len > wbuf->offs); | |
918 | if (!overlap) { | |
919 | /* We may safely unlock the write-buffer and read the data */ | |
920 | spin_unlock(&wbuf->lock); | |
921 | return ubifs_read_node(c, buf, type, len, lnum, offs); | |
922 | } | |
923 | ||
924 | /* Don't read under wbuf */ | |
925 | rlen = wbuf->offs - offs; | |
926 | if (rlen < 0) | |
927 | rlen = 0; | |
928 | ||
929 | /* Copy the rest from the write-buffer */ | |
930 | memcpy(buf + rlen, wbuf->buf + offs + rlen - wbuf->offs, len - rlen); | |
931 | spin_unlock(&wbuf->lock); | |
932 | ||
933 | if (rlen > 0) { | |
934 | /* Read everything that goes before write-buffer */ | |
d304820a AB |
935 | err = ubifs_leb_read(c, lnum, buf, offs, rlen, 0); |
936 | if (err && err != -EBADMSG) | |
1e51764a | 937 | return err; |
1e51764a AB |
938 | } |
939 | ||
940 | if (type != ch->node_type) { | |
235c362b | 941 | ubifs_err(c, "bad node type (%d but expected %d)", |
1e51764a AB |
942 | ch->node_type, type); |
943 | goto out; | |
944 | } | |
945 | ||
2953e73f | 946 | err = ubifs_check_node(c, buf, lnum, offs, 0, 0); |
1e51764a | 947 | if (err) { |
235c362b | 948 | ubifs_err(c, "expected node type %d", type); |
1e51764a AB |
949 | return err; |
950 | } | |
951 | ||
952 | rlen = le32_to_cpu(ch->len); | |
953 | if (rlen != len) { | |
235c362b | 954 | ubifs_err(c, "bad node length %d, expected %d", rlen, len); |
1e51764a AB |
955 | goto out; |
956 | } | |
957 | ||
958 | return 0; | |
959 | ||
960 | out: | |
235c362b | 961 | ubifs_err(c, "bad node at LEB %d:%d", lnum, offs); |
edf6be24 | 962 | ubifs_dump_node(c, buf); |
7c46d0ae | 963 | dump_stack(); |
1e51764a AB |
964 | return -EINVAL; |
965 | } | |
966 | ||
967 | /** | |
968 | * ubifs_read_node - read node. | |
969 | * @c: UBIFS file-system description object | |
970 | * @buf: buffer to read to | |
971 | * @type: node type | |
972 | * @len: node length (not aligned) | |
973 | * @lnum: logical eraseblock number | |
974 | * @offs: offset within the logical eraseblock | |
975 | * | |
976 | * This function reads a node of known type and and length, checks it and | |
977 | * stores in @buf. Returns zero in case of success, %-EUCLEAN if CRC mismatched | |
978 | * and a negative error code in case of failure. | |
979 | */ | |
980 | int ubifs_read_node(const struct ubifs_info *c, void *buf, int type, int len, | |
981 | int lnum, int offs) | |
982 | { | |
983 | int err, l; | |
984 | struct ubifs_ch *ch = buf; | |
985 | ||
986 | dbg_io("LEB %d:%d, %s, length %d", lnum, offs, dbg_ntype(type), len); | |
987 | ubifs_assert(lnum >= 0 && lnum < c->leb_cnt && offs >= 0); | |
988 | ubifs_assert(len >= UBIFS_CH_SZ && offs + len <= c->leb_size); | |
989 | ubifs_assert(!(offs & 7) && offs < c->leb_size); | |
990 | ubifs_assert(type >= 0 && type < UBIFS_NODE_TYPES_CNT); | |
991 | ||
d304820a AB |
992 | err = ubifs_leb_read(c, lnum, buf, offs, len, 0); |
993 | if (err && err != -EBADMSG) | |
1e51764a | 994 | return err; |
1e51764a AB |
995 | |
996 | if (type != ch->node_type) { | |
90bea5a3 DG |
997 | ubifs_errc(c, "bad node type (%d but expected %d)", |
998 | ch->node_type, type); | |
1e51764a AB |
999 | goto out; |
1000 | } | |
1001 | ||
2953e73f | 1002 | err = ubifs_check_node(c, buf, lnum, offs, 0, 0); |
1e51764a | 1003 | if (err) { |
90bea5a3 | 1004 | ubifs_errc(c, "expected node type %d", type); |
1e51764a AB |
1005 | return err; |
1006 | } | |
1007 | ||
1008 | l = le32_to_cpu(ch->len); | |
1009 | if (l != len) { | |
90bea5a3 | 1010 | ubifs_errc(c, "bad node length %d, expected %d", l, len); |
1e51764a AB |
1011 | goto out; |
1012 | } | |
1013 | ||
1014 | return 0; | |
1015 | ||
1016 | out: | |
90bea5a3 DG |
1017 | ubifs_errc(c, "bad node at LEB %d:%d, LEB mapping status %d", lnum, |
1018 | offs, ubi_is_mapped(c->ubi, lnum)); | |
1019 | if (!c->probing) { | |
1020 | ubifs_dump_node(c, buf); | |
1021 | dump_stack(); | |
1022 | } | |
1e51764a AB |
1023 | return -EINVAL; |
1024 | } | |
1025 | ||
1026 | /** | |
1027 | * ubifs_wbuf_init - initialize write-buffer. | |
1028 | * @c: UBIFS file-system description object | |
1029 | * @wbuf: write-buffer to initialize | |
1030 | * | |
cb54ef8b | 1031 | * This function initializes write-buffer. Returns zero in case of success |
1e51764a AB |
1032 | * %-ENOMEM in case of failure. |
1033 | */ | |
1034 | int ubifs_wbuf_init(struct ubifs_info *c, struct ubifs_wbuf *wbuf) | |
1035 | { | |
1036 | size_t size; | |
1037 | ||
6c7f74f7 | 1038 | wbuf->buf = kmalloc(c->max_write_size, GFP_KERNEL); |
1e51764a AB |
1039 | if (!wbuf->buf) |
1040 | return -ENOMEM; | |
1041 | ||
6c7f74f7 | 1042 | size = (c->max_write_size / UBIFS_CH_SZ + 1) * sizeof(ino_t); |
1e51764a AB |
1043 | wbuf->inodes = kmalloc(size, GFP_KERNEL); |
1044 | if (!wbuf->inodes) { | |
1045 | kfree(wbuf->buf); | |
1046 | wbuf->buf = NULL; | |
1047 | return -ENOMEM; | |
1048 | } | |
1049 | ||
1050 | wbuf->used = 0; | |
1051 | wbuf->lnum = wbuf->offs = -1; | |
6c7f74f7 AB |
1052 | /* |
1053 | * If the LEB starts at the max. write size aligned address, then | |
1054 | * write-buffer size has to be set to @c->max_write_size. Otherwise, | |
1055 | * set it to something smaller so that it ends at the closest max. | |
1056 | * write size boundary. | |
1057 | */ | |
1058 | size = c->max_write_size - (c->leb_start % c->max_write_size); | |
1059 | wbuf->avail = wbuf->size = size; | |
1e51764a AB |
1060 | wbuf->sync_callback = NULL; |
1061 | mutex_init(&wbuf->io_mutex); | |
1062 | spin_lock_init(&wbuf->lock); | |
1e51764a | 1063 | wbuf->c = c; |
1e51764a AB |
1064 | wbuf->next_ino = 0; |
1065 | ||
f2c5dbd7 AB |
1066 | hrtimer_init(&wbuf->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1067 | wbuf->timer.function = wbuf_timer_callback_nolock; | |
1e51764a AB |
1068 | return 0; |
1069 | } | |
1070 | ||
1071 | /** | |
1072 | * ubifs_wbuf_add_ino_nolock - add an inode number into the wbuf inode array. | |
cb54ef8b | 1073 | * @wbuf: the write-buffer where to add |
1e51764a AB |
1074 | * @inum: the inode number |
1075 | * | |
1076 | * This function adds an inode number to the inode array of the write-buffer. | |
1077 | */ | |
1078 | void ubifs_wbuf_add_ino_nolock(struct ubifs_wbuf *wbuf, ino_t inum) | |
1079 | { | |
1080 | if (!wbuf->buf) | |
1081 | /* NOR flash or something similar */ | |
1082 | return; | |
1083 | ||
1084 | spin_lock(&wbuf->lock); | |
1085 | if (wbuf->used) | |
1086 | wbuf->inodes[wbuf->next_ino++] = inum; | |
1087 | spin_unlock(&wbuf->lock); | |
1088 | } | |
1089 | ||
1090 | /** | |
1091 | * wbuf_has_ino - returns if the wbuf contains data from the inode. | |
1092 | * @wbuf: the write-buffer | |
1093 | * @inum: the inode number | |
1094 | * | |
1095 | * This function returns with %1 if the write-buffer contains some data from the | |
1096 | * given inode otherwise it returns with %0. | |
1097 | */ | |
1098 | static int wbuf_has_ino(struct ubifs_wbuf *wbuf, ino_t inum) | |
1099 | { | |
1100 | int i, ret = 0; | |
1101 | ||
1102 | spin_lock(&wbuf->lock); | |
1103 | for (i = 0; i < wbuf->next_ino; i++) | |
1104 | if (inum == wbuf->inodes[i]) { | |
1105 | ret = 1; | |
1106 | break; | |
1107 | } | |
1108 | spin_unlock(&wbuf->lock); | |
1109 | ||
1110 | return ret; | |
1111 | } | |
1112 | ||
1113 | /** | |
1114 | * ubifs_sync_wbufs_by_inode - synchronize write-buffers for an inode. | |
1115 | * @c: UBIFS file-system description object | |
1116 | * @inode: inode to synchronize | |
1117 | * | |
1118 | * This function synchronizes write-buffers which contain nodes belonging to | |
1119 | * @inode. Returns zero in case of success and a negative error code in case of | |
1120 | * failure. | |
1121 | */ | |
1122 | int ubifs_sync_wbufs_by_inode(struct ubifs_info *c, struct inode *inode) | |
1123 | { | |
1124 | int i, err = 0; | |
1125 | ||
1126 | for (i = 0; i < c->jhead_cnt; i++) { | |
1127 | struct ubifs_wbuf *wbuf = &c->jheads[i].wbuf; | |
1128 | ||
1129 | if (i == GCHD) | |
1130 | /* | |
1131 | * GC head is special, do not look at it. Even if the | |
1132 | * head contains something related to this inode, it is | |
1133 | * a _copy_ of corresponding on-flash node which sits | |
1134 | * somewhere else. | |
1135 | */ | |
1136 | continue; | |
1137 | ||
1138 | if (!wbuf_has_ino(wbuf, inode->i_ino)) | |
1139 | continue; | |
1140 | ||
1141 | mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead); | |
1142 | if (wbuf_has_ino(wbuf, inode->i_ino)) | |
1143 | err = ubifs_wbuf_sync_nolock(wbuf); | |
1144 | mutex_unlock(&wbuf->io_mutex); | |
1145 | ||
1146 | if (err) { | |
1147 | ubifs_ro_mode(c, err); | |
1148 | return err; | |
1149 | } | |
1150 | } | |
1151 | return 0; | |
1152 | } |