NETFILTER: ctnetlink: check for status attribute existence on conntrack creation
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / jffs2 / readinode.c
blobf1695642d0f7f218757fbdcff49ac0e0ed440a35
1 /*
2 * JFFS2 -- Journalling Flash File System, Version 2.
4 * Copyright (C) 2001-2003 Red Hat, Inc.
6 * Created by David Woodhouse <dwmw2@infradead.org>
8 * For licensing information, see the file 'LICENCE' in this directory.
10 * $Id: readinode.c,v 1.143 2005/11/07 11:14:41 gleixner Exp $
14 #include <linux/kernel.h>
15 #include <linux/sched.h>
16 #include <linux/slab.h>
17 #include <linux/fs.h>
18 #include <linux/crc32.h>
19 #include <linux/pagemap.h>
20 #include <linux/mtd/mtd.h>
21 #include <linux/compiler.h>
22 #include "nodelist.h"
25 * Put a new tmp_dnode_info into the temporaty RB-tree, keeping the list in
26 * order of increasing version.
28 static void jffs2_add_tn_to_tree(struct jffs2_tmp_dnode_info *tn, struct rb_root *list)
30 struct rb_node **p = &list->rb_node;
31 struct rb_node * parent = NULL;
32 struct jffs2_tmp_dnode_info *this;
34 while (*p) {
35 parent = *p;
36 this = rb_entry(parent, struct jffs2_tmp_dnode_info, rb);
38 /* There may actually be a collision here, but it doesn't
39 actually matter. As long as the two nodes with the same
40 version are together, it's all fine. */
41 if (tn->version > this->version)
42 p = &(*p)->rb_left;
43 else
44 p = &(*p)->rb_right;
47 rb_link_node(&tn->rb, parent, p);
48 rb_insert_color(&tn->rb, list);
51 static void jffs2_free_tmp_dnode_info_list(struct rb_root *list)
53 struct rb_node *this;
54 struct jffs2_tmp_dnode_info *tn;
56 this = list->rb_node;
58 /* Now at bottom of tree */
59 while (this) {
60 if (this->rb_left)
61 this = this->rb_left;
62 else if (this->rb_right)
63 this = this->rb_right;
64 else {
65 tn = rb_entry(this, struct jffs2_tmp_dnode_info, rb);
66 jffs2_free_full_dnode(tn->fn);
67 jffs2_free_tmp_dnode_info(tn);
69 this = this->rb_parent;
70 if (!this)
71 break;
73 if (this->rb_left == &tn->rb)
74 this->rb_left = NULL;
75 else if (this->rb_right == &tn->rb)
76 this->rb_right = NULL;
77 else BUG();
80 list->rb_node = NULL;
83 static void jffs2_free_full_dirent_list(struct jffs2_full_dirent *fd)
85 struct jffs2_full_dirent *next;
87 while (fd) {
88 next = fd->next;
89 jffs2_free_full_dirent(fd);
90 fd = next;
94 /* Returns first valid node after 'ref'. May return 'ref' */
95 static struct jffs2_raw_node_ref *jffs2_first_valid_node(struct jffs2_raw_node_ref *ref)
97 while (ref && ref->next_in_ino) {
98 if (!ref_obsolete(ref))
99 return ref;
100 dbg_noderef("node at 0x%08x is obsoleted. Ignoring.\n", ref_offset(ref));
101 ref = ref->next_in_ino;
103 return NULL;
107 * Helper function for jffs2_get_inode_nodes().
108 * It is called every time an directory entry node is found.
110 * Returns: 0 on succes;
111 * 1 if the node should be marked obsolete;
112 * negative error code on failure.
114 static inline int read_direntry(struct jffs2_sb_info *c, struct jffs2_raw_node_ref *ref,
115 struct jffs2_raw_dirent *rd, size_t read, struct jffs2_full_dirent **fdp,
116 uint32_t *latest_mctime, uint32_t *mctime_ver)
118 struct jffs2_full_dirent *fd;
120 /* The direntry nodes are checked during the flash scanning */
121 BUG_ON(ref_flags(ref) == REF_UNCHECKED);
122 /* Obsoleted. This cannot happen, surely? dwmw2 20020308 */
123 BUG_ON(ref_obsolete(ref));
125 /* Sanity check */
126 if (unlikely(PAD((rd->nsize + sizeof(*rd))) != PAD(je32_to_cpu(rd->totlen)))) {
127 JFFS2_ERROR("illegal nsize in node at %#08x: nsize %#02x, totlen %#04x\n",
128 ref_offset(ref), rd->nsize, je32_to_cpu(rd->totlen));
129 return 1;
132 fd = jffs2_alloc_full_dirent(rd->nsize + 1);
133 if (unlikely(!fd))
134 return -ENOMEM;
136 fd->raw = ref;
137 fd->version = je32_to_cpu(rd->version);
138 fd->ino = je32_to_cpu(rd->ino);
139 fd->type = rd->type;
141 /* Pick out the mctime of the latest dirent */
142 if(fd->version > *mctime_ver && je32_to_cpu(rd->mctime)) {
143 *mctime_ver = fd->version;
144 *latest_mctime = je32_to_cpu(rd->mctime);
148 * Copy as much of the name as possible from the raw
149 * dirent we've already read from the flash.
151 if (read > sizeof(*rd))
152 memcpy(&fd->name[0], &rd->name[0],
153 min_t(uint32_t, rd->nsize, (read - sizeof(*rd)) ));
155 /* Do we need to copy any more of the name directly from the flash? */
156 if (rd->nsize + sizeof(*rd) > read) {
157 /* FIXME: point() */
158 int err;
159 int already = read - sizeof(*rd);
161 err = jffs2_flash_read(c, (ref_offset(ref)) + read,
162 rd->nsize - already, &read, &fd->name[already]);
163 if (unlikely(read != rd->nsize - already) && likely(!err))
164 return -EIO;
166 if (unlikely(err)) {
167 JFFS2_ERROR("read remainder of name: error %d\n", err);
168 jffs2_free_full_dirent(fd);
169 return -EIO;
173 fd->nhash = full_name_hash(fd->name, rd->nsize);
174 fd->next = NULL;
175 fd->name[rd->nsize] = '\0';
178 * Wheee. We now have a complete jffs2_full_dirent structure, with
179 * the name in it and everything. Link it into the list
181 jffs2_add_fd_to_list(c, fd, fdp);
183 return 0;
187 * Helper function for jffs2_get_inode_nodes().
188 * It is called every time an inode node is found.
190 * Returns: 0 on succes;
191 * 1 if the node should be marked obsolete;
192 * negative error code on failure.
194 static inline int read_dnode(struct jffs2_sb_info *c, struct jffs2_raw_node_ref *ref,
195 struct jffs2_raw_inode *rd, struct rb_root *tnp, int rdlen,
196 uint32_t *latest_mctime, uint32_t *mctime_ver)
198 struct jffs2_tmp_dnode_info *tn;
199 uint32_t len, csize;
200 int ret = 1;
202 /* Obsoleted. This cannot happen, surely? dwmw2 20020308 */
203 BUG_ON(ref_obsolete(ref));
205 tn = jffs2_alloc_tmp_dnode_info();
206 if (!tn) {
207 JFFS2_ERROR("failed to allocate tn (%d bytes).\n", sizeof(*tn));
208 return -ENOMEM;
211 tn->partial_crc = 0;
212 csize = je32_to_cpu(rd->csize);
214 /* If we've never checked the CRCs on this node, check them now */
215 if (ref_flags(ref) == REF_UNCHECKED) {
216 uint32_t crc;
218 crc = crc32(0, rd, sizeof(*rd) - 8);
219 if (unlikely(crc != je32_to_cpu(rd->node_crc))) {
220 JFFS2_NOTICE("header CRC failed on node at %#08x: read %#08x, calculated %#08x\n",
221 ref_offset(ref), je32_to_cpu(rd->node_crc), crc);
222 goto free_out;
225 /* Sanity checks */
226 if (unlikely(je32_to_cpu(rd->offset) > je32_to_cpu(rd->isize)) ||
227 unlikely(PAD(je32_to_cpu(rd->csize) + sizeof(*rd)) != PAD(je32_to_cpu(rd->totlen)))) {
228 JFFS2_WARNING("inode node header CRC is corrupted at %#08x\n", ref_offset(ref));
229 jffs2_dbg_dump_node(c, ref_offset(ref));
230 goto free_out;
233 if (jffs2_is_writebuffered(c) && csize != 0) {
234 /* At this point we are supposed to check the data CRC
235 * of our unchecked node. But thus far, we do not
236 * know whether the node is valid or obsolete. To
237 * figure this out, we need to walk all the nodes of
238 * the inode and build the inode fragtree. We don't
239 * want to spend time checking data of nodes which may
240 * later be found to be obsolete. So we put off the full
241 * data CRC checking until we have read all the inode
242 * nodes and have started building the fragtree.
244 * The fragtree is being built starting with nodes
245 * having the highest version number, so we'll be able
246 * to detect whether a node is valid (i.e., it is not
247 * overlapped by a node with higher version) or not.
248 * And we'll be able to check only those nodes, which
249 * are not obsolete.
251 * Of course, this optimization only makes sense in case
252 * of NAND flashes (or other flashes whith
253 * !jffs2_can_mark_obsolete()), since on NOR flashes
254 * nodes are marked obsolete physically.
256 * Since NAND flashes (or other flashes with
257 * jffs2_is_writebuffered(c)) are anyway read by
258 * fractions of c->wbuf_pagesize, and we have just read
259 * the node header, it is likely that the starting part
260 * of the node data is also read when we read the
261 * header. So we don't mind to check the CRC of the
262 * starting part of the data of the node now, and check
263 * the second part later (in jffs2_check_node_data()).
264 * Of course, we will not need to re-read and re-check
265 * the NAND page which we have just read. This is why we
266 * read the whole NAND page at jffs2_get_inode_nodes(),
267 * while we needed only the node header.
269 unsigned char *buf;
271 /* 'buf' will point to the start of data */
272 buf = (unsigned char *)rd + sizeof(*rd);
273 /* len will be the read data length */
274 len = min_t(uint32_t, rdlen - sizeof(*rd), csize);
275 tn->partial_crc = crc32(0, buf, len);
277 dbg_readinode("Calculates CRC (%#08x) for %d bytes, csize %d\n", tn->partial_crc, len, csize);
279 /* If we actually calculated the whole data CRC
280 * and it is wrong, drop the node. */
281 if (len >= csize && unlikely(tn->partial_crc != je32_to_cpu(rd->data_crc))) {
282 JFFS2_NOTICE("wrong data CRC in data node at 0x%08x: read %#08x, calculated %#08x.\n",
283 ref_offset(ref), tn->partial_crc, je32_to_cpu(rd->data_crc));
284 goto free_out;
287 } else if (csize == 0) {
289 * We checked the header CRC. If the node has no data, adjust
290 * the space accounting now. For other nodes this will be done
291 * later either when the node is marked obsolete or when its
292 * data is checked.
294 struct jffs2_eraseblock *jeb;
296 dbg_readinode("the node has no data.\n");
297 jeb = &c->blocks[ref->flash_offset / c->sector_size];
298 len = ref_totlen(c, jeb, ref);
300 spin_lock(&c->erase_completion_lock);
301 jeb->used_size += len;
302 jeb->unchecked_size -= len;
303 c->used_size += len;
304 c->unchecked_size -= len;
305 ref->flash_offset = ref_offset(ref) | REF_NORMAL;
306 spin_unlock(&c->erase_completion_lock);
310 tn->fn = jffs2_alloc_full_dnode();
311 if (!tn->fn) {
312 JFFS2_ERROR("alloc fn failed\n");
313 ret = -ENOMEM;
314 goto free_out;
317 tn->version = je32_to_cpu(rd->version);
318 tn->fn->ofs = je32_to_cpu(rd->offset);
319 tn->data_crc = je32_to_cpu(rd->data_crc);
320 tn->csize = csize;
321 tn->fn->raw = ref;
323 /* There was a bug where we wrote hole nodes out with
324 csize/dsize swapped. Deal with it */
325 if (rd->compr == JFFS2_COMPR_ZERO && !je32_to_cpu(rd->dsize) && csize)
326 tn->fn->size = csize;
327 else // normal case...
328 tn->fn->size = je32_to_cpu(rd->dsize);
330 dbg_readinode("dnode @%08x: ver %u, offset %#04x, dsize %#04x, csize %#04x\n",
331 ref_offset(ref), je32_to_cpu(rd->version), je32_to_cpu(rd->offset), je32_to_cpu(rd->dsize), csize);
333 jffs2_add_tn_to_tree(tn, tnp);
335 return 0;
337 free_out:
338 jffs2_free_tmp_dnode_info(tn);
339 return ret;
343 * Helper function for jffs2_get_inode_nodes().
344 * It is called every time an unknown node is found.
346 * Returns: 0 on succes;
347 * 1 if the node should be marked obsolete;
348 * negative error code on failure.
350 static inline int read_unknown(struct jffs2_sb_info *c, struct jffs2_raw_node_ref *ref, struct jffs2_unknown_node *un)
352 /* We don't mark unknown nodes as REF_UNCHECKED */
353 BUG_ON(ref_flags(ref) == REF_UNCHECKED);
355 un->nodetype = cpu_to_je16(JFFS2_NODE_ACCURATE | je16_to_cpu(un->nodetype));
357 if (crc32(0, un, sizeof(struct jffs2_unknown_node) - 4) != je32_to_cpu(un->hdr_crc)) {
358 /* Hmmm. This should have been caught at scan time. */
359 JFFS2_NOTICE("node header CRC failed at %#08x. But it must have been OK earlier.\n", ref_offset(ref));
360 jffs2_dbg_dump_node(c, ref_offset(ref));
361 return 1;
362 } else {
363 switch(je16_to_cpu(un->nodetype) & JFFS2_COMPAT_MASK) {
365 case JFFS2_FEATURE_INCOMPAT:
366 JFFS2_ERROR("unknown INCOMPAT nodetype %#04X at %#08x\n",
367 je16_to_cpu(un->nodetype), ref_offset(ref));
368 /* EEP */
369 BUG();
370 break;
372 case JFFS2_FEATURE_ROCOMPAT:
373 JFFS2_ERROR("unknown ROCOMPAT nodetype %#04X at %#08x\n",
374 je16_to_cpu(un->nodetype), ref_offset(ref));
375 BUG_ON(!(c->flags & JFFS2_SB_FLAG_RO));
376 break;
378 case JFFS2_FEATURE_RWCOMPAT_COPY:
379 JFFS2_NOTICE("unknown RWCOMPAT_COPY nodetype %#04X at %#08x\n",
380 je16_to_cpu(un->nodetype), ref_offset(ref));
381 break;
383 case JFFS2_FEATURE_RWCOMPAT_DELETE:
384 JFFS2_NOTICE("unknown RWCOMPAT_DELETE nodetype %#04X at %#08x\n",
385 je16_to_cpu(un->nodetype), ref_offset(ref));
386 return 1;
390 return 0;
394 * Helper function for jffs2_get_inode_nodes().
395 * The function detects whether more data should be read and reads it if yes.
397 * Returns: 0 on succes;
398 * negative error code on failure.
400 static int read_more(struct jffs2_sb_info *c, struct jffs2_raw_node_ref *ref,
401 int right_size, int *rdlen, unsigned char *buf, unsigned char *bufstart)
403 int right_len, err, len;
404 size_t retlen;
405 uint32_t offs;
407 if (jffs2_is_writebuffered(c)) {
408 right_len = c->wbuf_pagesize - (bufstart - buf);
409 if (right_size + (int)(bufstart - buf) > c->wbuf_pagesize)
410 right_len += c->wbuf_pagesize;
411 } else
412 right_len = right_size;
414 if (*rdlen == right_len)
415 return 0;
417 /* We need to read more data */
418 offs = ref_offset(ref) + *rdlen;
419 if (jffs2_is_writebuffered(c)) {
420 bufstart = buf + c->wbuf_pagesize;
421 len = c->wbuf_pagesize;
422 } else {
423 bufstart = buf + *rdlen;
424 len = right_size - *rdlen;
427 dbg_readinode("read more %d bytes\n", len);
429 err = jffs2_flash_read(c, offs, len, &retlen, bufstart);
430 if (err) {
431 JFFS2_ERROR("can not read %d bytes from 0x%08x, "
432 "error code: %d.\n", len, offs, err);
433 return err;
436 if (retlen < len) {
437 JFFS2_ERROR("short read at %#08x: %d instead of %d.\n",
438 offs, retlen, len);
439 return -EIO;
442 *rdlen = right_len;
444 return 0;
447 /* Get tmp_dnode_info and full_dirent for all non-obsolete nodes associated
448 with this ino, returning the former in order of version */
449 static int jffs2_get_inode_nodes(struct jffs2_sb_info *c, struct jffs2_inode_info *f,
450 struct rb_root *tnp, struct jffs2_full_dirent **fdp,
451 uint32_t *highest_version, uint32_t *latest_mctime,
452 uint32_t *mctime_ver)
454 struct jffs2_raw_node_ref *ref, *valid_ref;
455 struct rb_root ret_tn = RB_ROOT;
456 struct jffs2_full_dirent *ret_fd = NULL;
457 unsigned char *buf = NULL;
458 union jffs2_node_union *node;
459 size_t retlen;
460 int len, err;
462 *mctime_ver = 0;
464 dbg_readinode("ino #%u\n", f->inocache->ino);
466 if (jffs2_is_writebuffered(c)) {
468 * If we have the write buffer, we assume the minimal I/O unit
469 * is c->wbuf_pagesize. We implement some optimizations which in
470 * this case and we need a temporary buffer of size =
471 * 2*c->wbuf_pagesize bytes (see comments in read_dnode()).
472 * Basically, we want to read not only the node header, but the
473 * whole wbuf (NAND page in case of NAND) or 2, if the node
474 * header overlaps the border between the 2 wbufs.
476 len = 2*c->wbuf_pagesize;
477 } else {
479 * When there is no write buffer, the size of the temporary
480 * buffer is the size of the larges node header.
482 len = sizeof(union jffs2_node_union);
485 /* FIXME: in case of NOR and available ->point() this
486 * needs to be fixed. */
487 buf = kmalloc(len, GFP_KERNEL);
488 if (!buf)
489 return -ENOMEM;
491 spin_lock(&c->erase_completion_lock);
492 valid_ref = jffs2_first_valid_node(f->inocache->nodes);
493 if (!valid_ref && f->inocache->ino != 1)
494 JFFS2_WARNING("Eep. No valid nodes for ino #%u.\n", f->inocache->ino);
495 while (valid_ref) {
496 unsigned char *bufstart;
498 /* We can hold a pointer to a non-obsolete node without the spinlock,
499 but _obsolete_ nodes may disappear at any time, if the block
500 they're in gets erased. So if we mark 'ref' obsolete while we're
501 not holding the lock, it can go away immediately. For that reason,
502 we find the next valid node first, before processing 'ref'.
504 ref = valid_ref;
505 valid_ref = jffs2_first_valid_node(ref->next_in_ino);
506 spin_unlock(&c->erase_completion_lock);
508 cond_resched();
511 * At this point we don't know the type of the node we're going
512 * to read, so we do not know the size of its header. In order
513 * to minimize the amount of flash IO we assume the node has
514 * size = JFFS2_MIN_NODE_HEADER.
516 if (jffs2_is_writebuffered(c)) {
518 * We treat 'buf' as 2 adjacent wbufs. We want to
519 * adjust bufstart such as it points to the
520 * beginning of the node within this wbuf.
522 bufstart = buf + (ref_offset(ref) % c->wbuf_pagesize);
523 /* We will read either one wbuf or 2 wbufs. */
524 len = c->wbuf_pagesize - (bufstart - buf);
525 if (JFFS2_MIN_NODE_HEADER + (int)(bufstart - buf) > c->wbuf_pagesize) {
526 /* The header spans the border of the first wbuf */
527 len += c->wbuf_pagesize;
529 } else {
530 bufstart = buf;
531 len = JFFS2_MIN_NODE_HEADER;
534 dbg_readinode("read %d bytes at %#08x(%d).\n", len, ref_offset(ref), ref_flags(ref));
536 /* FIXME: point() */
537 err = jffs2_flash_read(c, ref_offset(ref), len,
538 &retlen, bufstart);
539 if (err) {
540 JFFS2_ERROR("can not read %d bytes from 0x%08x, " "error code: %d.\n", len, ref_offset(ref), err);
541 goto free_out;
544 if (retlen < len) {
545 JFFS2_ERROR("short read at %#08x: %d instead of %d.\n", ref_offset(ref), retlen, len);
546 err = -EIO;
547 goto free_out;
550 node = (union jffs2_node_union *)bufstart;
552 switch (je16_to_cpu(node->u.nodetype)) {
554 case JFFS2_NODETYPE_DIRENT:
556 if (JFFS2_MIN_NODE_HEADER < sizeof(struct jffs2_raw_dirent)) {
557 err = read_more(c, ref, sizeof(struct jffs2_raw_dirent), &len, buf, bufstart);
558 if (unlikely(err))
559 goto free_out;
562 err = read_direntry(c, ref, &node->d, retlen, &ret_fd, latest_mctime, mctime_ver);
563 if (err == 1) {
564 jffs2_mark_node_obsolete(c, ref);
565 break;
566 } else if (unlikely(err))
567 goto free_out;
569 if (je32_to_cpu(node->d.version) > *highest_version)
570 *highest_version = je32_to_cpu(node->d.version);
572 break;
574 case JFFS2_NODETYPE_INODE:
576 if (JFFS2_MIN_NODE_HEADER < sizeof(struct jffs2_raw_inode)) {
577 err = read_more(c, ref, sizeof(struct jffs2_raw_inode), &len, buf, bufstart);
578 if (unlikely(err))
579 goto free_out;
582 err = read_dnode(c, ref, &node->i, &ret_tn, len, latest_mctime, mctime_ver);
583 if (err == 1) {
584 jffs2_mark_node_obsolete(c, ref);
585 break;
586 } else if (unlikely(err))
587 goto free_out;
589 if (je32_to_cpu(node->i.version) > *highest_version)
590 *highest_version = je32_to_cpu(node->i.version);
592 break;
594 default:
595 if (JFFS2_MIN_NODE_HEADER < sizeof(struct jffs2_unknown_node)) {
596 err = read_more(c, ref, sizeof(struct jffs2_unknown_node), &len, buf, bufstart);
597 if (unlikely(err))
598 goto free_out;
601 err = read_unknown(c, ref, &node->u);
602 if (err == 1) {
603 jffs2_mark_node_obsolete(c, ref);
604 break;
605 } else if (unlikely(err))
606 goto free_out;
609 spin_lock(&c->erase_completion_lock);
612 spin_unlock(&c->erase_completion_lock);
613 *tnp = ret_tn;
614 *fdp = ret_fd;
615 kfree(buf);
617 dbg_readinode("nodes of inode #%u were read, the highest version is %u, latest_mctime %u, mctime_ver %u.\n",
618 f->inocache->ino, *highest_version, *latest_mctime, *mctime_ver);
619 return 0;
621 free_out:
622 jffs2_free_tmp_dnode_info_list(&ret_tn);
623 jffs2_free_full_dirent_list(ret_fd);
624 kfree(buf);
625 return err;
628 static int jffs2_do_read_inode_internal(struct jffs2_sb_info *c,
629 struct jffs2_inode_info *f,
630 struct jffs2_raw_inode *latest_node)
632 struct jffs2_tmp_dnode_info *tn;
633 struct rb_root tn_list;
634 struct rb_node *rb, *repl_rb;
635 struct jffs2_full_dirent *fd_list;
636 struct jffs2_full_dnode *fn, *first_fn = NULL;
637 uint32_t crc;
638 uint32_t latest_mctime, mctime_ver;
639 size_t retlen;
640 int ret;
642 dbg_readinode("ino #%u nlink is %d\n", f->inocache->ino, f->inocache->nlink);
644 /* Grab all nodes relevant to this ino */
645 ret = jffs2_get_inode_nodes(c, f, &tn_list, &fd_list, &f->highest_version, &latest_mctime, &mctime_ver);
647 if (ret) {
648 JFFS2_ERROR("cannot read nodes for ino %u, returned error is %d\n", f->inocache->ino, ret);
649 if (f->inocache->state == INO_STATE_READING)
650 jffs2_set_inocache_state(c, f->inocache, INO_STATE_CHECKEDABSENT);
651 return ret;
653 f->dents = fd_list;
655 rb = rb_first(&tn_list);
657 while (rb) {
658 cond_resched();
659 tn = rb_entry(rb, struct jffs2_tmp_dnode_info, rb);
660 fn = tn->fn;
661 ret = 1;
662 dbg_readinode("consider node ver %u, phys offset "
663 "%#08x(%d), range %u-%u.\n", tn->version,
664 ref_offset(fn->raw), ref_flags(fn->raw),
665 fn->ofs, fn->ofs + fn->size);
667 if (fn->size) {
668 ret = jffs2_add_older_frag_to_fragtree(c, f, tn);
669 /* TODO: the error code isn't checked, check it */
670 jffs2_dbg_fragtree_paranoia_check_nolock(f);
671 BUG_ON(ret < 0);
672 if (!first_fn && ret == 0)
673 first_fn = fn;
674 } else if (!first_fn) {
675 first_fn = fn;
676 f->metadata = fn;
677 ret = 0; /* Prevent freeing the metadata update node */
678 } else
679 jffs2_mark_node_obsolete(c, fn->raw);
681 BUG_ON(rb->rb_left);
682 if (rb->rb_parent && rb->rb_parent->rb_left == rb) {
683 /* We were then left-hand child of our parent. We need
684 * to move our own right-hand child into our place. */
685 repl_rb = rb->rb_right;
686 if (repl_rb)
687 repl_rb->rb_parent = rb->rb_parent;
688 } else
689 repl_rb = NULL;
691 rb = rb_next(rb);
693 /* Remove the spent tn from the tree; don't bother rebalancing
694 * but put our right-hand child in our own place. */
695 if (tn->rb.rb_parent) {
696 if (tn->rb.rb_parent->rb_left == &tn->rb)
697 tn->rb.rb_parent->rb_left = repl_rb;
698 else if (tn->rb.rb_parent->rb_right == &tn->rb)
699 tn->rb.rb_parent->rb_right = repl_rb;
700 else BUG();
701 } else if (tn->rb.rb_right)
702 tn->rb.rb_right->rb_parent = NULL;
704 jffs2_free_tmp_dnode_info(tn);
705 if (ret) {
706 dbg_readinode("delete dnode %u-%u.\n",
707 fn->ofs, fn->ofs + fn->size);
708 jffs2_free_full_dnode(fn);
711 jffs2_dbg_fragtree_paranoia_check_nolock(f);
713 BUG_ON(first_fn && ref_obsolete(first_fn->raw));
715 fn = first_fn;
716 if (unlikely(!first_fn)) {
717 /* No data nodes for this inode. */
718 if (f->inocache->ino != 1) {
719 JFFS2_WARNING("no data nodes found for ino #%u\n", f->inocache->ino);
720 if (!fd_list) {
721 if (f->inocache->state == INO_STATE_READING)
722 jffs2_set_inocache_state(c, f->inocache, INO_STATE_CHECKEDABSENT);
723 return -EIO;
725 JFFS2_NOTICE("but it has children so we fake some modes for it\n");
727 latest_node->mode = cpu_to_jemode(S_IFDIR|S_IRUGO|S_IWUSR|S_IXUGO);
728 latest_node->version = cpu_to_je32(0);
729 latest_node->atime = latest_node->ctime = latest_node->mtime = cpu_to_je32(0);
730 latest_node->isize = cpu_to_je32(0);
731 latest_node->gid = cpu_to_je16(0);
732 latest_node->uid = cpu_to_je16(0);
733 if (f->inocache->state == INO_STATE_READING)
734 jffs2_set_inocache_state(c, f->inocache, INO_STATE_PRESENT);
735 return 0;
738 ret = jffs2_flash_read(c, ref_offset(fn->raw), sizeof(*latest_node), &retlen, (void *)latest_node);
739 if (ret || retlen != sizeof(*latest_node)) {
740 JFFS2_ERROR("failed to read from flash: error %d, %zd of %zd bytes read\n",
741 ret, retlen, sizeof(*latest_node));
742 /* FIXME: If this fails, there seems to be a memory leak. Find it. */
743 up(&f->sem);
744 jffs2_do_clear_inode(c, f);
745 return ret?ret:-EIO;
748 crc = crc32(0, latest_node, sizeof(*latest_node)-8);
749 if (crc != je32_to_cpu(latest_node->node_crc)) {
750 JFFS2_ERROR("CRC failed for read_inode of inode %u at physical location 0x%x\n",
751 f->inocache->ino, ref_offset(fn->raw));
752 up(&f->sem);
753 jffs2_do_clear_inode(c, f);
754 return -EIO;
757 switch(jemode_to_cpu(latest_node->mode) & S_IFMT) {
758 case S_IFDIR:
759 if (mctime_ver > je32_to_cpu(latest_node->version)) {
760 /* The times in the latest_node are actually older than
761 mctime in the latest dirent. Cheat. */
762 latest_node->ctime = latest_node->mtime = cpu_to_je32(latest_mctime);
764 break;
767 case S_IFREG:
768 /* If it was a regular file, truncate it to the latest node's isize */
769 jffs2_truncate_fragtree(c, &f->fragtree, je32_to_cpu(latest_node->isize));
770 break;
772 case S_IFLNK:
773 /* Hack to work around broken isize in old symlink code.
774 Remove this when dwmw2 comes to his senses and stops
775 symlinks from being an entirely gratuitous special
776 case. */
777 if (!je32_to_cpu(latest_node->isize))
778 latest_node->isize = latest_node->dsize;
780 if (f->inocache->state != INO_STATE_CHECKING) {
781 /* Symlink's inode data is the target path. Read it and
782 * keep in RAM to facilitate quick follow symlink
783 * operation. */
784 f->target = kmalloc(je32_to_cpu(latest_node->csize) + 1, GFP_KERNEL);
785 if (!f->target) {
786 JFFS2_ERROR("can't allocate %d bytes of memory for the symlink target path cache\n", je32_to_cpu(latest_node->csize));
787 up(&f->sem);
788 jffs2_do_clear_inode(c, f);
789 return -ENOMEM;
792 ret = jffs2_flash_read(c, ref_offset(fn->raw) + sizeof(*latest_node),
793 je32_to_cpu(latest_node->csize), &retlen, (char *)f->target);
795 if (ret || retlen != je32_to_cpu(latest_node->csize)) {
796 if (retlen != je32_to_cpu(latest_node->csize))
797 ret = -EIO;
798 kfree(f->target);
799 f->target = NULL;
800 up(&f->sem);
801 jffs2_do_clear_inode(c, f);
802 return -ret;
805 f->target[je32_to_cpu(latest_node->csize)] = '\0';
806 dbg_readinode("symlink's target '%s' cached\n", f->target);
809 /* fall through... */
811 case S_IFBLK:
812 case S_IFCHR:
813 /* Certain inode types should have only one data node, and it's
814 kept as the metadata node */
815 if (f->metadata) {
816 JFFS2_ERROR("Argh. Special inode #%u with mode 0%o had metadata node\n",
817 f->inocache->ino, jemode_to_cpu(latest_node->mode));
818 up(&f->sem);
819 jffs2_do_clear_inode(c, f);
820 return -EIO;
822 if (!frag_first(&f->fragtree)) {
823 JFFS2_ERROR("Argh. Special inode #%u with mode 0%o has no fragments\n",
824 f->inocache->ino, jemode_to_cpu(latest_node->mode));
825 up(&f->sem);
826 jffs2_do_clear_inode(c, f);
827 return -EIO;
829 /* ASSERT: f->fraglist != NULL */
830 if (frag_next(frag_first(&f->fragtree))) {
831 JFFS2_ERROR("Argh. Special inode #%u with mode 0x%x had more than one node\n",
832 f->inocache->ino, jemode_to_cpu(latest_node->mode));
833 /* FIXME: Deal with it - check crc32, check for duplicate node, check times and discard the older one */
834 up(&f->sem);
835 jffs2_do_clear_inode(c, f);
836 return -EIO;
838 /* OK. We're happy */
839 f->metadata = frag_first(&f->fragtree)->node;
840 jffs2_free_node_frag(frag_first(&f->fragtree));
841 f->fragtree = RB_ROOT;
842 break;
844 if (f->inocache->state == INO_STATE_READING)
845 jffs2_set_inocache_state(c, f->inocache, INO_STATE_PRESENT);
847 return 0;
850 /* Scan the list of all nodes present for this ino, build map of versions, etc. */
851 int jffs2_do_read_inode(struct jffs2_sb_info *c, struct jffs2_inode_info *f,
852 uint32_t ino, struct jffs2_raw_inode *latest_node)
854 dbg_readinode("read inode #%u\n", ino);
856 retry_inocache:
857 spin_lock(&c->inocache_lock);
858 f->inocache = jffs2_get_ino_cache(c, ino);
860 if (f->inocache) {
861 /* Check its state. We may need to wait before we can use it */
862 switch(f->inocache->state) {
863 case INO_STATE_UNCHECKED:
864 case INO_STATE_CHECKEDABSENT:
865 f->inocache->state = INO_STATE_READING;
866 break;
868 case INO_STATE_CHECKING:
869 case INO_STATE_GC:
870 /* If it's in either of these states, we need
871 to wait for whoever's got it to finish and
872 put it back. */
873 dbg_readinode("waiting for ino #%u in state %d\n", ino, f->inocache->state);
874 sleep_on_spinunlock(&c->inocache_wq, &c->inocache_lock);
875 goto retry_inocache;
877 case INO_STATE_READING:
878 case INO_STATE_PRESENT:
879 /* Eep. This should never happen. It can
880 happen if Linux calls read_inode() again
881 before clear_inode() has finished though. */
882 JFFS2_ERROR("Eep. Trying to read_inode #%u when it's already in state %d!\n", ino, f->inocache->state);
883 /* Fail. That's probably better than allowing it to succeed */
884 f->inocache = NULL;
885 break;
887 default:
888 BUG();
891 spin_unlock(&c->inocache_lock);
893 if (!f->inocache && ino == 1) {
894 /* Special case - no root inode on medium */
895 f->inocache = jffs2_alloc_inode_cache();
896 if (!f->inocache) {
897 JFFS2_ERROR("cannot allocate inocache for root inode\n");
898 return -ENOMEM;
900 dbg_readinode("creating inocache for root inode\n");
901 memset(f->inocache, 0, sizeof(struct jffs2_inode_cache));
902 f->inocache->ino = f->inocache->nlink = 1;
903 f->inocache->nodes = (struct jffs2_raw_node_ref *)f->inocache;
904 f->inocache->state = INO_STATE_READING;
905 jffs2_add_ino_cache(c, f->inocache);
907 if (!f->inocache) {
908 JFFS2_ERROR("requestied to read an nonexistent ino %u\n", ino);
909 return -ENOENT;
912 return jffs2_do_read_inode_internal(c, f, latest_node);
915 int jffs2_do_crccheck_inode(struct jffs2_sb_info *c, struct jffs2_inode_cache *ic)
917 struct jffs2_raw_inode n;
918 struct jffs2_inode_info *f = kmalloc(sizeof(*f), GFP_KERNEL);
919 int ret;
921 if (!f)
922 return -ENOMEM;
924 memset(f, 0, sizeof(*f));
925 init_MUTEX_LOCKED(&f->sem);
926 f->inocache = ic;
928 ret = jffs2_do_read_inode_internal(c, f, &n);
929 if (!ret) {
930 up(&f->sem);
931 jffs2_do_clear_inode(c, f);
933 kfree (f);
934 return ret;
937 void jffs2_do_clear_inode(struct jffs2_sb_info *c, struct jffs2_inode_info *f)
939 struct jffs2_full_dirent *fd, *fds;
940 int deleted;
942 down(&f->sem);
943 deleted = f->inocache && !f->inocache->nlink;
945 if (f->inocache && f->inocache->state != INO_STATE_CHECKING)
946 jffs2_set_inocache_state(c, f->inocache, INO_STATE_CLEARING);
948 if (f->metadata) {
949 if (deleted)
950 jffs2_mark_node_obsolete(c, f->metadata->raw);
951 jffs2_free_full_dnode(f->metadata);
954 jffs2_kill_fragtree(&f->fragtree, deleted?c:NULL);
956 if (f->target) {
957 kfree(f->target);
958 f->target = NULL;
961 fds = f->dents;
962 while(fds) {
963 fd = fds;
964 fds = fd->next;
965 jffs2_free_full_dirent(fd);
968 if (f->inocache && f->inocache->state != INO_STATE_CHECKING) {
969 jffs2_set_inocache_state(c, f->inocache, INO_STATE_CHECKEDABSENT);
970 if (f->inocache->nodes == (void *)f->inocache)
971 jffs2_del_ino_cache(c, f->inocache);
974 up(&f->sem);