2 * This file is part of UBIFS.
4 * Copyright (C) 2006-2008 Nokia Corporation.
5 * Copyright (C) 2006, 2007 University of Szeged, Hungary
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.
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
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
20 * Authors: Artem Bityutskiy (Битюцкий Артём)
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
32 * similar to the mechanism is used by JFFS2.
34 * Write-buffers are defined by 'struct ubifs_wbuf' objects and protected by
35 * mutexes defined inside these objects. Since sometimes upper-level code
36 * has to lock the write-buffer (e.g. journal space reservation code), many
37 * functions related to write-buffers have "nolock" suffix which means that the
38 * caller has to lock the write-buffer before calling this function.
40 * UBIFS stores nodes at 64 bit-aligned addresses. If the node length is not
41 * aligned, UBIFS starts the next node from the aligned address, and the padded
42 * bytes may contain any rubbish. In other words, UBIFS does not put padding
43 * bytes in those small gaps. Common headers of nodes store real node lengths,
44 * not aligned lengths. Indexing nodes also store real lengths in branches.
46 * UBIFS uses padding when it pads to the next min. I/O unit. In this case it
47 * uses padding nodes or padding bytes, if the padding node does not fit.
49 * All UBIFS nodes are protected by CRC checksums and UBIFS checks all nodes
50 * every time they are read from the flash media.
53 #include <linux/crc32.h>
54 #include <linux/slab.h>
58 * ubifs_ro_mode - switch UBIFS to read read-only mode.
59 * @c: UBIFS file-system description object
60 * @err: error code which is the reason of switching to R/O mode
62 void ubifs_ro_mode(struct ubifs_info *c, int err)
66 c->no_chk_data_crc = 0;
67 c->vfs_sb->s_flags |= MS_RDONLY;
68 ubifs_warn("switched to read-only mode, error %d", err);
74 * ubifs_check_node - check node.
75 * @c: UBIFS file-system description object
77 * @lnum: logical eraseblock number
78 * @offs: offset within the logical eraseblock
79 * @quiet: print no messages
80 * @must_chk_crc: indicates whether to always check the CRC
82 * This function checks node magic number and CRC checksum. This function also
83 * validates node length to prevent UBIFS from becoming crazy when an attacker
84 * feeds it a file-system image with incorrect nodes. For example, too large
85 * node length in the common header could cause UBIFS to read memory outside of
86 * allocated buffer when checking the CRC checksum.
88 * This function may skip data nodes CRC checking if @c->no_chk_data_crc is
89 * true, which is controlled by corresponding UBIFS mount option. However, if
90 * @must_chk_crc is true, then @c->no_chk_data_crc is ignored and CRC is
91 * checked. Similarly, if @c->mounting or @c->remounting_rw is true (we are
92 * mounting or re-mounting to R/W mode), @c->no_chk_data_crc is ignored and CRC
93 * is checked. This is because during mounting or re-mounting from R/O mode to
94 * R/W mode we may read journal nodes (when replying the journal or doing the
95 * recovery) and the journal nodes may potentially be corrupted, so checking is
98 * This function returns zero in case of success and %-EUCLEAN in case of bad
101 int ubifs_check_node(const struct ubifs_info *c, const void *buf, int lnum,
102 int offs, int quiet, int must_chk_crc)
104 int err = -EINVAL, type, node_len;
105 uint32_t crc, node_crc, magic;
106 const struct ubifs_ch *ch = buf;
108 ubifs_assert(lnum >= 0 && lnum < c->leb_cnt && offs >= 0);
109 ubifs_assert(!(offs & 7) && offs < c->leb_size);
111 magic = le32_to_cpu(ch->magic);
112 if (magic != UBIFS_NODE_MAGIC) {
114 ubifs_err("bad magic %#08x, expected %#08x",
115 magic, UBIFS_NODE_MAGIC);
120 type = ch->node_type;
121 if (type < 0 || type >= UBIFS_NODE_TYPES_CNT) {
123 ubifs_err("bad node type %d", type);
127 node_len = le32_to_cpu(ch->len);
128 if (node_len + offs > c->leb_size)
131 if (c->ranges[type].max_len == 0) {
132 if (node_len != c->ranges[type].len)
134 } else if (node_len < c->ranges[type].min_len ||
135 node_len > c->ranges[type].max_len)
138 if (!must_chk_crc && type == UBIFS_DATA_NODE && !c->mounting &&
139 !c->remounting_rw && c->no_chk_data_crc)
142 crc = crc32(UBIFS_CRC32_INIT, buf + 8, node_len - 8);
143 node_crc = le32_to_cpu(ch->crc);
144 if (crc != node_crc) {
146 ubifs_err("bad CRC: calculated %#08x, read %#08x",
156 ubifs_err("bad node length %d", node_len);
159 ubifs_err("bad node at LEB %d:%d", lnum, offs);
160 dbg_dump_node(c, buf);
167 * ubifs_pad - pad flash space.
168 * @c: UBIFS file-system description object
169 * @buf: buffer to put padding to
170 * @pad: how many bytes to pad
172 * The flash media obliges us to write only in chunks of %c->min_io_size and
173 * when we have to write less data we add padding node to the write-buffer and
174 * pad it to the next minimal I/O unit's boundary. Padding nodes help when the
175 * media is being scanned. If the amount of wasted space is not enough to fit a
176 * padding node which takes %UBIFS_PAD_NODE_SZ bytes, we write padding bytes
177 * pattern (%UBIFS_PADDING_BYTE).
179 * Padding nodes are also used to fill gaps when the "commit-in-gaps" method is
182 void ubifs_pad(const struct ubifs_info *c, void *buf, int pad)
186 ubifs_assert(pad >= 0 && !(pad & 7));
188 if (pad >= UBIFS_PAD_NODE_SZ) {
189 struct ubifs_ch *ch = buf;
190 struct ubifs_pad_node *pad_node = buf;
192 ch->magic = cpu_to_le32(UBIFS_NODE_MAGIC);
193 ch->node_type = UBIFS_PAD_NODE;
194 ch->group_type = UBIFS_NO_NODE_GROUP;
195 ch->padding[0] = ch->padding[1] = 0;
197 ch->len = cpu_to_le32(UBIFS_PAD_NODE_SZ);
198 pad -= UBIFS_PAD_NODE_SZ;
199 pad_node->pad_len = cpu_to_le32(pad);
200 crc = crc32(UBIFS_CRC32_INIT, buf + 8, UBIFS_PAD_NODE_SZ - 8);
201 ch->crc = cpu_to_le32(crc);
202 memset(buf + UBIFS_PAD_NODE_SZ, 0, pad);
204 /* Too little space, padding node won't fit */
205 memset(buf, UBIFS_PADDING_BYTE, pad);
209 * next_sqnum - get next sequence number.
210 * @c: UBIFS file-system description object
212 static unsigned long long next_sqnum(struct ubifs_info *c)
214 unsigned long long sqnum;
216 spin_lock(&c->cnt_lock);
217 sqnum = ++c->max_sqnum;
218 spin_unlock(&c->cnt_lock);
220 if (unlikely(sqnum >= SQNUM_WARN_WATERMARK)) {
221 if (sqnum >= SQNUM_WATERMARK) {
222 ubifs_err("sequence number overflow %llu, end of life",
224 ubifs_ro_mode(c, -EINVAL);
226 ubifs_warn("running out of sequence numbers, end of life soon");
233 * ubifs_prepare_node - prepare node to be written to flash.
234 * @c: UBIFS file-system description object
235 * @node: the node to pad
237 * @pad: if the buffer has to be padded
239 * This function prepares node at @node to be written to the media - it
240 * calculates node CRC, fills the common header, and adds proper padding up to
241 * the next minimum I/O unit if @pad is not zero.
243 void ubifs_prepare_node(struct ubifs_info *c, void *node, int len, int pad)
246 struct ubifs_ch *ch = node;
247 unsigned long long sqnum = next_sqnum(c);
249 ubifs_assert(len >= UBIFS_CH_SZ);
251 ch->magic = cpu_to_le32(UBIFS_NODE_MAGIC);
252 ch->len = cpu_to_le32(len);
253 ch->group_type = UBIFS_NO_NODE_GROUP;
254 ch->sqnum = cpu_to_le64(sqnum);
255 ch->padding[0] = ch->padding[1] = 0;
256 crc = crc32(UBIFS_CRC32_INIT, node + 8, len - 8);
257 ch->crc = cpu_to_le32(crc);
261 pad = ALIGN(len, c->min_io_size) - len;
262 ubifs_pad(c, node + len, pad);
267 * ubifs_prep_grp_node - prepare node of a group to be written to flash.
268 * @c: UBIFS file-system description object
269 * @node: the node to pad
271 * @last: indicates the last node of the group
273 * This function prepares node at @node to be written to the media - it
274 * calculates node CRC and fills the common header.
276 void ubifs_prep_grp_node(struct ubifs_info *c, void *node, int len, int last)
279 struct ubifs_ch *ch = node;
280 unsigned long long sqnum = next_sqnum(c);
282 ubifs_assert(len >= UBIFS_CH_SZ);
284 ch->magic = cpu_to_le32(UBIFS_NODE_MAGIC);
285 ch->len = cpu_to_le32(len);
287 ch->group_type = UBIFS_LAST_OF_NODE_GROUP;
289 ch->group_type = UBIFS_IN_NODE_GROUP;
290 ch->sqnum = cpu_to_le64(sqnum);
291 ch->padding[0] = ch->padding[1] = 0;
292 crc = crc32(UBIFS_CRC32_INIT, node + 8, len - 8);
293 ch->crc = cpu_to_le32(crc);
297 * wbuf_timer_callback - write-buffer timer callback function.
298 * @data: timer data (write-buffer descriptor)
300 * This function is called when the write-buffer timer expires.
302 static enum hrtimer_restart wbuf_timer_callback_nolock(struct hrtimer *timer)
304 struct ubifs_wbuf *wbuf = container_of(timer, struct ubifs_wbuf, timer);
306 dbg_io("jhead %s", dbg_jhead(wbuf->jhead));
308 wbuf->c->need_wbuf_sync = 1;
309 ubifs_wake_up_bgt(wbuf->c);
310 return HRTIMER_NORESTART;
314 * new_wbuf_timer - start new write-buffer timer.
315 * @wbuf: write-buffer descriptor
317 static void new_wbuf_timer_nolock(struct ubifs_wbuf *wbuf)
319 ubifs_assert(!hrtimer_active(&wbuf->timer));
323 dbg_io("set timer for jhead %s, %llu-%llu millisecs",
324 dbg_jhead(wbuf->jhead),
325 div_u64(ktime_to_ns(wbuf->softlimit), USEC_PER_SEC),
326 div_u64(ktime_to_ns(wbuf->softlimit) + wbuf->delta,
328 hrtimer_start_range_ns(&wbuf->timer, wbuf->softlimit, wbuf->delta,
333 * cancel_wbuf_timer - cancel write-buffer timer.
334 * @wbuf: write-buffer descriptor
336 static void cancel_wbuf_timer_nolock(struct ubifs_wbuf *wbuf)
341 hrtimer_cancel(&wbuf->timer);
345 * ubifs_wbuf_sync_nolock - synchronize write-buffer.
346 * @wbuf: write-buffer to synchronize
348 * This function synchronizes write-buffer @buf and returns zero in case of
349 * success or a negative error code in case of failure.
351 int ubifs_wbuf_sync_nolock(struct ubifs_wbuf *wbuf)
353 struct ubifs_info *c = wbuf->c;
356 cancel_wbuf_timer_nolock(wbuf);
357 if (!wbuf->used || wbuf->lnum == -1)
358 /* Write-buffer is empty or not seeked */
361 dbg_io("LEB %d:%d, %d bytes, jhead %s",
362 wbuf->lnum, wbuf->offs, wbuf->used, dbg_jhead(wbuf->jhead));
363 ubifs_assert(!(wbuf->avail & 7));
364 ubifs_assert(wbuf->offs + c->min_io_size <= c->leb_size);
365 ubifs_assert(!c->ro_media && !c->ro_mount);
370 ubifs_pad(c, wbuf->buf + wbuf->used, wbuf->avail);
371 err = ubi_leb_write(c->ubi, wbuf->lnum, wbuf->buf, wbuf->offs,
372 c->min_io_size, wbuf->dtype);
374 ubifs_err("cannot write %d bytes to LEB %d:%d",
375 c->min_io_size, wbuf->lnum, wbuf->offs);
382 spin_lock(&wbuf->lock);
383 wbuf->offs += c->min_io_size;
384 wbuf->avail = c->min_io_size;
387 spin_unlock(&wbuf->lock);
389 if (wbuf->sync_callback)
390 err = wbuf->sync_callback(c, wbuf->lnum,
391 c->leb_size - wbuf->offs, dirt);
396 * ubifs_wbuf_seek_nolock - seek write-buffer.
397 * @wbuf: write-buffer
398 * @lnum: logical eraseblock number to seek to
399 * @offs: logical eraseblock offset to seek to
402 * This function targets the write-buffer to logical eraseblock @lnum:@offs.
403 * The write-buffer is synchronized if it is not empty. Returns zero in case of
404 * success and a negative error code in case of failure.
406 int ubifs_wbuf_seek_nolock(struct ubifs_wbuf *wbuf, int lnum, int offs,
409 const struct ubifs_info *c = wbuf->c;
411 dbg_io("LEB %d:%d, jhead %s", lnum, offs, dbg_jhead(wbuf->jhead));
412 ubifs_assert(lnum >= 0 && lnum < c->leb_cnt);
413 ubifs_assert(offs >= 0 && offs <= c->leb_size);
414 ubifs_assert(offs % c->min_io_size == 0 && !(offs & 7));
415 ubifs_assert(lnum != wbuf->lnum);
417 if (wbuf->used > 0) {
418 int err = ubifs_wbuf_sync_nolock(wbuf);
424 spin_lock(&wbuf->lock);
427 wbuf->avail = c->min_io_size;
429 spin_unlock(&wbuf->lock);
436 * ubifs_bg_wbufs_sync - synchronize write-buffers.
437 * @c: UBIFS file-system description object
439 * This function is called by background thread to synchronize write-buffers.
440 * Returns zero in case of success and a negative error code in case of
443 int ubifs_bg_wbufs_sync(struct ubifs_info *c)
447 ubifs_assert(!c->ro_media && !c->ro_mount);
448 if (!c->need_wbuf_sync)
450 c->need_wbuf_sync = 0;
457 dbg_io("synchronize");
458 for (i = 0; i < c->jhead_cnt; i++) {
459 struct ubifs_wbuf *wbuf = &c->jheads[i].wbuf;
464 * If the mutex is locked then wbuf is being changed, so
465 * synchronization is not necessary.
467 if (mutex_is_locked(&wbuf->io_mutex))
470 mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead);
471 if (!wbuf->need_sync) {
472 mutex_unlock(&wbuf->io_mutex);
476 err = ubifs_wbuf_sync_nolock(wbuf);
477 mutex_unlock(&wbuf->io_mutex);
479 ubifs_err("cannot sync write-buffer, error %d", err);
480 ubifs_ro_mode(c, err);
488 /* Cancel all timers to prevent repeated errors */
489 for (i = 0; i < c->jhead_cnt; i++) {
490 struct ubifs_wbuf *wbuf = &c->jheads[i].wbuf;
492 mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead);
493 cancel_wbuf_timer_nolock(wbuf);
494 mutex_unlock(&wbuf->io_mutex);
500 * ubifs_wbuf_write_nolock - write data to flash via write-buffer.
501 * @wbuf: write-buffer
502 * @buf: node to write
505 * This function writes data to flash via write-buffer @wbuf. This means that
506 * the last piece of the node won't reach the flash media immediately if it
507 * does not take whole minimal I/O unit. Instead, the node will sit in RAM
508 * until the write-buffer is synchronized (e.g., by timer).
510 * This function returns zero in case of success and a negative error code in
511 * case of failure. If the node cannot be written because there is no more
512 * space in this logical eraseblock, %-ENOSPC is returned.
514 int ubifs_wbuf_write_nolock(struct ubifs_wbuf *wbuf, void *buf, int len)
516 struct ubifs_info *c = wbuf->c;
517 int err, written, n, aligned_len = ALIGN(len, 8), offs;
519 dbg_io("%d bytes (%s) to jhead %s wbuf at LEB %d:%d", len,
520 dbg_ntype(((struct ubifs_ch *)buf)->node_type),
521 dbg_jhead(wbuf->jhead), wbuf->lnum, wbuf->offs + wbuf->used);
522 ubifs_assert(len > 0 && wbuf->lnum >= 0 && wbuf->lnum < c->leb_cnt);
523 ubifs_assert(wbuf->offs >= 0 && wbuf->offs % c->min_io_size == 0);
524 ubifs_assert(!(wbuf->offs & 7) && wbuf->offs <= c->leb_size);
525 ubifs_assert(wbuf->avail > 0 && wbuf->avail <= c->min_io_size);
526 ubifs_assert(mutex_is_locked(&wbuf->io_mutex));
527 ubifs_assert(!c->ro_media && !c->ro_mount);
529 if (c->leb_size - wbuf->offs - wbuf->used < aligned_len) {
534 cancel_wbuf_timer_nolock(wbuf);
539 if (aligned_len <= wbuf->avail) {
541 * The node is not very large and fits entirely within
544 memcpy(wbuf->buf + wbuf->used, buf, len);
546 if (aligned_len == wbuf->avail) {
547 dbg_io("flush jhead %s wbuf to LEB %d:%d",
548 dbg_jhead(wbuf->jhead), wbuf->lnum, wbuf->offs);
549 err = ubi_leb_write(c->ubi, wbuf->lnum, wbuf->buf,
550 wbuf->offs, c->min_io_size,
555 spin_lock(&wbuf->lock);
556 wbuf->offs += c->min_io_size;
557 wbuf->avail = c->min_io_size;
560 spin_unlock(&wbuf->lock);
562 spin_lock(&wbuf->lock);
563 wbuf->avail -= aligned_len;
564 wbuf->used += aligned_len;
565 spin_unlock(&wbuf->lock);
572 * The node is large enough and does not fit entirely within current
573 * minimal I/O unit. We have to fill and flush write-buffer and switch
574 * to the next min. I/O unit.
576 dbg_io("flush jhead %s wbuf to LEB %d:%d",
577 dbg_jhead(wbuf->jhead), wbuf->lnum, wbuf->offs);
578 memcpy(wbuf->buf + wbuf->used, buf, wbuf->avail);
579 err = ubi_leb_write(c->ubi, wbuf->lnum, wbuf->buf, wbuf->offs,
580 c->min_io_size, wbuf->dtype);
584 offs = wbuf->offs + c->min_io_size;
586 aligned_len -= wbuf->avail;
587 written = wbuf->avail;
590 * The remaining data may take more whole min. I/O units, so write the
591 * remains multiple to min. I/O unit size directly to the flash media.
592 * We align node length to 8-byte boundary because we anyway flash wbuf
593 * if the remaining space is less than 8 bytes.
595 n = aligned_len >> c->min_io_shift;
597 n <<= c->min_io_shift;
598 dbg_io("write %d bytes to LEB %d:%d", n, wbuf->lnum, offs);
599 err = ubi_leb_write(c->ubi, wbuf->lnum, buf + written, offs, n,
609 spin_lock(&wbuf->lock);
612 * And now we have what's left and what does not take whole
613 * min. I/O unit, so write it to the write-buffer and we are
616 memcpy(wbuf->buf, buf + written, len);
619 wbuf->used = aligned_len;
620 wbuf->avail = c->min_io_size - aligned_len;
622 spin_unlock(&wbuf->lock);
625 if (wbuf->sync_callback) {
626 int free = c->leb_size - wbuf->offs - wbuf->used;
628 err = wbuf->sync_callback(c, wbuf->lnum, free, 0);
634 new_wbuf_timer_nolock(wbuf);
639 ubifs_err("cannot write %d bytes to LEB %d:%d, error %d",
640 len, wbuf->lnum, wbuf->offs, err);
641 dbg_dump_node(c, buf);
643 dbg_dump_leb(c, wbuf->lnum);
648 * ubifs_write_node - write node to the media.
649 * @c: UBIFS file-system description object
650 * @buf: the node to write
652 * @lnum: logical eraseblock number
653 * @offs: offset within the logical eraseblock
654 * @dtype: node life-time hint (%UBI_LONGTERM, %UBI_SHORTTERM, %UBI_UNKNOWN)
656 * This function automatically fills node magic number, assigns sequence
657 * number, and calculates node CRC checksum. The length of the @buf buffer has
658 * to be aligned to the minimal I/O unit size. This function automatically
659 * appends padding node and padding bytes if needed. Returns zero in case of
660 * success and a negative error code in case of failure.
662 int ubifs_write_node(struct ubifs_info *c, void *buf, int len, int lnum,
665 int err, buf_len = ALIGN(len, c->min_io_size);
667 dbg_io("LEB %d:%d, %s, length %d (aligned %d)",
668 lnum, offs, dbg_ntype(((struct ubifs_ch *)buf)->node_type), len,
670 ubifs_assert(lnum >= 0 && lnum < c->leb_cnt && offs >= 0);
671 ubifs_assert(offs % c->min_io_size == 0 && offs < c->leb_size);
672 ubifs_assert(!c->ro_media && !c->ro_mount);
677 ubifs_prepare_node(c, buf, len, 1);
678 err = ubi_leb_write(c->ubi, lnum, buf, offs, buf_len, dtype);
680 ubifs_err("cannot write %d bytes to LEB %d:%d, error %d",
681 buf_len, lnum, offs, err);
682 dbg_dump_node(c, buf);
690 * ubifs_read_node_wbuf - read node from the media or write-buffer.
691 * @wbuf: wbuf to check for un-written data
692 * @buf: buffer to read to
695 * @lnum: logical eraseblock number
696 * @offs: offset within the logical eraseblock
698 * This function reads a node of known type and length, checks it and stores
699 * in @buf. If the node partially or fully sits in the write-buffer, this
700 * function takes data from the buffer, otherwise it reads the flash media.
701 * Returns zero in case of success, %-EUCLEAN if CRC mismatched and a negative
702 * error code in case of failure.
704 int ubifs_read_node_wbuf(struct ubifs_wbuf *wbuf, void *buf, int type, int len,
707 const struct ubifs_info *c = wbuf->c;
708 int err, rlen, overlap;
709 struct ubifs_ch *ch = buf;
711 dbg_io("LEB %d:%d, %s, length %d, jhead %s", lnum, offs,
712 dbg_ntype(type), len, dbg_jhead(wbuf->jhead));
713 ubifs_assert(wbuf && lnum >= 0 && lnum < c->leb_cnt && offs >= 0);
714 ubifs_assert(!(offs & 7) && offs < c->leb_size);
715 ubifs_assert(type >= 0 && type < UBIFS_NODE_TYPES_CNT);
717 spin_lock(&wbuf->lock);
718 overlap = (lnum == wbuf->lnum && offs + len > wbuf->offs);
720 /* We may safely unlock the write-buffer and read the data */
721 spin_unlock(&wbuf->lock);
722 return ubifs_read_node(c, buf, type, len, lnum, offs);
725 /* Don't read under wbuf */
726 rlen = wbuf->offs - offs;
730 /* Copy the rest from the write-buffer */
731 memcpy(buf + rlen, wbuf->buf + offs + rlen - wbuf->offs, len - rlen);
732 spin_unlock(&wbuf->lock);
735 /* Read everything that goes before write-buffer */
736 err = ubi_read(c->ubi, lnum, buf, offs, rlen);
737 if (err && err != -EBADMSG) {
738 ubifs_err("failed to read node %d from LEB %d:%d, "
739 "error %d", type, lnum, offs, err);
745 if (type != ch->node_type) {
746 ubifs_err("bad node type (%d but expected %d)",
747 ch->node_type, type);
751 err = ubifs_check_node(c, buf, lnum, offs, 0, 0);
753 ubifs_err("expected node type %d", type);
757 rlen = le32_to_cpu(ch->len);
759 ubifs_err("bad node length %d, expected %d", rlen, len);
766 ubifs_err("bad node at LEB %d:%d", lnum, offs);
767 dbg_dump_node(c, buf);
773 * ubifs_read_node - read node.
774 * @c: UBIFS file-system description object
775 * @buf: buffer to read to
777 * @len: node length (not aligned)
778 * @lnum: logical eraseblock number
779 * @offs: offset within the logical eraseblock
781 * This function reads a node of known type and and length, checks it and
782 * stores in @buf. Returns zero in case of success, %-EUCLEAN if CRC mismatched
783 * and a negative error code in case of failure.
785 int ubifs_read_node(const struct ubifs_info *c, void *buf, int type, int len,
789 struct ubifs_ch *ch = buf;
791 dbg_io("LEB %d:%d, %s, length %d", lnum, offs, dbg_ntype(type), len);
792 ubifs_assert(lnum >= 0 && lnum < c->leb_cnt && offs >= 0);
793 ubifs_assert(len >= UBIFS_CH_SZ && offs + len <= c->leb_size);
794 ubifs_assert(!(offs & 7) && offs < c->leb_size);
795 ubifs_assert(type >= 0 && type < UBIFS_NODE_TYPES_CNT);
797 err = ubi_read(c->ubi, lnum, buf, offs, len);
798 if (err && err != -EBADMSG) {
799 ubifs_err("cannot read node %d from LEB %d:%d, error %d",
800 type, lnum, offs, err);
804 if (type != ch->node_type) {
805 ubifs_err("bad node type (%d but expected %d)",
806 ch->node_type, type);
810 err = ubifs_check_node(c, buf, lnum, offs, 0, 0);
812 ubifs_err("expected node type %d", type);
816 l = le32_to_cpu(ch->len);
818 ubifs_err("bad node length %d, expected %d", l, len);
825 ubifs_err("bad node at LEB %d:%d, LEB mapping status %d", lnum, offs,
826 ubi_is_mapped(c->ubi, lnum));
827 dbg_dump_node(c, buf);
833 * ubifs_wbuf_init - initialize write-buffer.
834 * @c: UBIFS file-system description object
835 * @wbuf: write-buffer to initialize
837 * This function initializes write-buffer. Returns zero in case of success
838 * %-ENOMEM in case of failure.
840 int ubifs_wbuf_init(struct ubifs_info *c, struct ubifs_wbuf *wbuf)
844 wbuf->buf = kmalloc(c->min_io_size, GFP_KERNEL);
848 size = (c->min_io_size / UBIFS_CH_SZ + 1) * sizeof(ino_t);
849 wbuf->inodes = kmalloc(size, GFP_KERNEL);
857 wbuf->lnum = wbuf->offs = -1;
858 wbuf->avail = c->min_io_size;
859 wbuf->dtype = UBI_UNKNOWN;
860 wbuf->sync_callback = NULL;
861 mutex_init(&wbuf->io_mutex);
862 spin_lock_init(&wbuf->lock);
866 hrtimer_init(&wbuf->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
867 wbuf->timer.function = wbuf_timer_callback_nolock;
868 wbuf->softlimit = ktime_set(WBUF_TIMEOUT_SOFTLIMIT, 0);
869 wbuf->delta = WBUF_TIMEOUT_HARDLIMIT - WBUF_TIMEOUT_SOFTLIMIT;
870 wbuf->delta *= 1000000000ULL;
871 ubifs_assert(wbuf->delta <= ULONG_MAX);
876 * ubifs_wbuf_add_ino_nolock - add an inode number into the wbuf inode array.
877 * @wbuf: the write-buffer where to add
878 * @inum: the inode number
880 * This function adds an inode number to the inode array of the write-buffer.
882 void ubifs_wbuf_add_ino_nolock(struct ubifs_wbuf *wbuf, ino_t inum)
885 /* NOR flash or something similar */
888 spin_lock(&wbuf->lock);
890 wbuf->inodes[wbuf->next_ino++] = inum;
891 spin_unlock(&wbuf->lock);
895 * wbuf_has_ino - returns if the wbuf contains data from the inode.
896 * @wbuf: the write-buffer
897 * @inum: the inode number
899 * This function returns with %1 if the write-buffer contains some data from the
900 * given inode otherwise it returns with %0.
902 static int wbuf_has_ino(struct ubifs_wbuf *wbuf, ino_t inum)
906 spin_lock(&wbuf->lock);
907 for (i = 0; i < wbuf->next_ino; i++)
908 if (inum == wbuf->inodes[i]) {
912 spin_unlock(&wbuf->lock);
918 * ubifs_sync_wbufs_by_inode - synchronize write-buffers for an inode.
919 * @c: UBIFS file-system description object
920 * @inode: inode to synchronize
922 * This function synchronizes write-buffers which contain nodes belonging to
923 * @inode. Returns zero in case of success and a negative error code in case of
926 int ubifs_sync_wbufs_by_inode(struct ubifs_info *c, struct inode *inode)
930 for (i = 0; i < c->jhead_cnt; i++) {
931 struct ubifs_wbuf *wbuf = &c->jheads[i].wbuf;
935 * GC head is special, do not look at it. Even if the
936 * head contains something related to this inode, it is
937 * a _copy_ of corresponding on-flash node which sits
942 if (!wbuf_has_ino(wbuf, inode->i_ino))
945 mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead);
946 if (wbuf_has_ino(wbuf, inode->i_ino))
947 err = ubifs_wbuf_sync_nolock(wbuf);
948 mutex_unlock(&wbuf->io_mutex);
951 ubifs_ro_mode(c, err);