diff options
Diffstat (limited to 'fs/ubifs/replay.c')
-rw-r--r-- | fs/ubifs/replay.c | 468 |
1 files changed, 235 insertions, 233 deletions
diff --git a/fs/ubifs/replay.c b/fs/ubifs/replay.c index eed0fcff8d7..6617280d167 100644 --- a/fs/ubifs/replay.c +++ b/fs/ubifs/replay.c @@ -33,43 +33,32 @@ */ #include "ubifs.h" - -/* - * Replay flags. - * - * REPLAY_DELETION: node was deleted - * REPLAY_REF: node is a reference node - */ -enum { - REPLAY_DELETION = 1, - REPLAY_REF = 2, -}; +#include <linux/list_sort.h> /** - * struct replay_entry - replay tree entry. + * struct replay_entry - replay list entry. * @lnum: logical eraseblock number of the node * @offs: node offset * @len: node length + * @deletion: non-zero if this entry corresponds to a node deletion * @sqnum: node sequence number - * @flags: replay flags - * @rb: links the replay tree + * @list: links the replay list * @key: node key * @nm: directory entry name * @old_size: truncation old size * @new_size: truncation new size - * @free: amount of free space in a bud - * @dirty: amount of dirty space in a bud from padding and deletion nodes * - * UBIFS journal replay must compare node sequence numbers, which means it must - * build a tree of node information to insert into the TNC. + * The replay process first scans all buds and builds the replay list, then + * sorts the replay list in nodes sequence number order, and then inserts all + * the replay entries to the TNC. */ struct replay_entry { int lnum; int offs; int len; + unsigned int deletion:1; unsigned long long sqnum; - int flags; - struct rb_node rb; + struct list_head list; union ubifs_key key; union { struct qstr nm; @@ -77,10 +66,6 @@ struct replay_entry { loff_t old_size; loff_t new_size; }; - struct { - int free; - int dirty; - }; }; }; @@ -88,57 +73,64 @@ struct replay_entry { * struct bud_entry - entry in the list of buds to replay. * @list: next bud in the list * @bud: bud description object - * @free: free bytes in the bud * @sqnum: reference node sequence number + * @free: free bytes in the bud + * @dirty: dirty bytes in the bud */ struct bud_entry { struct list_head list; struct ubifs_bud *bud; - int free; unsigned long long sqnum; + int free; + int dirty; }; /** * set_bud_lprops - set free and dirty space used by a bud. * @c: UBIFS file-system description object - * @r: replay entry of bud + * @b: bud entry which describes the bud + * + * This function makes sure the LEB properties of bud @b are set correctly + * after the replay. Returns zero in case of success and a negative error code + * in case of failure. */ -static int set_bud_lprops(struct ubifs_info *c, struct replay_entry *r) +static int set_bud_lprops(struct ubifs_info *c, struct bud_entry *b) { const struct ubifs_lprops *lp; int err = 0, dirty; ubifs_get_lprops(c); - lp = ubifs_lpt_lookup_dirty(c, r->lnum); + lp = ubifs_lpt_lookup_dirty(c, b->bud->lnum); if (IS_ERR(lp)) { err = PTR_ERR(lp); goto out; } dirty = lp->dirty; - if (r->offs == 0 && (lp->free != c->leb_size || lp->dirty != 0)) { + if (b->bud->start == 0 && (lp->free != c->leb_size || lp->dirty != 0)) { /* * The LEB was added to the journal with a starting offset of * zero which means the LEB must have been empty. The LEB - * property values should be lp->free == c->leb_size and - * lp->dirty == 0, but that is not the case. The reason is that - * the LEB was garbage collected. The garbage collector resets - * the free and dirty space without recording it anywhere except - * lprops, so if there is not a commit then lprops does not have - * that information next time the file system is mounted. + * property values should be @lp->free == @c->leb_size and + * @lp->dirty == 0, but that is not the case. The reason is that + * the LEB had been garbage collected before it became the bud, + * and there was not commit inbetween. The garbage collector + * resets the free and dirty space without recording it + * anywhere except lprops, so if there was no commit then + * lprops does not have that information. * * We do not need to adjust free space because the scan has told * us the exact value which is recorded in the replay entry as - * r->free. + * @b->free. * * However we do need to subtract from the dirty space the * amount of space that the garbage collector reclaimed, which * is the whole LEB minus the amount of space that was free. */ - dbg_mnt("bud LEB %d was GC'd (%d free, %d dirty)", r->lnum, + dbg_mnt("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum, lp->free, lp->dirty); - dbg_gc("bud LEB %d was GC'd (%d free, %d dirty)", r->lnum, + dbg_gc("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum, lp->free, lp->dirty); dirty -= c->leb_size - lp->free; /* @@ -150,21 +142,48 @@ static int set_bud_lprops(struct ubifs_info *c, struct replay_entry *r) */ if (dirty != 0) dbg_msg("LEB %d lp: %d free %d dirty " - "replay: %d free %d dirty", r->lnum, lp->free, - lp->dirty, r->free, r->dirty); + "replay: %d free %d dirty", b->bud->lnum, + lp->free, lp->dirty, b->free, b->dirty); } - lp = ubifs_change_lp(c, lp, r->free, dirty + r->dirty, + lp = ubifs_change_lp(c, lp, b->free, dirty + b->dirty, lp->flags | LPROPS_TAKEN, 0); if (IS_ERR(lp)) { err = PTR_ERR(lp); goto out; } + + /* Make sure the journal head points to the latest bud */ + err = ubifs_wbuf_seek_nolock(&c->jheads[b->bud->jhead].wbuf, + b->bud->lnum, c->leb_size - b->free, + UBI_SHORTTERM); + out: ubifs_release_lprops(c); return err; } /** + * set_buds_lprops - set free and dirty space for all replayed buds. + * @c: UBIFS file-system description object + * + * This function sets LEB properties for all replayed buds. Returns zero in + * case of success and a negative error code in case of failure. + */ +static int set_buds_lprops(struct ubifs_info *c) +{ + struct bud_entry *b; + int err; + + list_for_each_entry(b, &c->replay_buds, list) { + err = set_bud_lprops(c, b); + if (err) + return err; + } + + return 0; +} + +/** * trun_remove_range - apply a replay entry for a truncation to the TNC. * @c: UBIFS file-system description object * @r: replay entry of truncation @@ -200,24 +219,22 @@ static int trun_remove_range(struct ubifs_info *c, struct replay_entry *r) */ static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r) { - int err, deletion = ((r->flags & REPLAY_DELETION) != 0); + int err; - dbg_mnt("LEB %d:%d len %d flgs %d sqnum %llu %s", r->lnum, - r->offs, r->len, r->flags, r->sqnum, DBGKEY(&r->key)); + dbg_mnt("LEB %d:%d len %d deletion %d sqnum %llu %s", r->lnum, + r->offs, r->len, r->deletion, r->sqnum, DBGKEY(&r->key)); /* Set c->replay_sqnum to help deal with dangling branches. */ c->replay_sqnum = r->sqnum; - if (r->flags & REPLAY_REF) - err = set_bud_lprops(c, r); - else if (is_hash_key(c, &r->key)) { - if (deletion) + if (is_hash_key(c, &r->key)) { + if (r->deletion) err = ubifs_tnc_remove_nm(c, &r->key, &r->nm); else err = ubifs_tnc_add_nm(c, &r->key, r->lnum, r->offs, r->len, &r->nm); } else { - if (deletion) + if (r->deletion) switch (key_type(c, &r->key)) { case UBIFS_INO_KEY: { @@ -240,7 +257,7 @@ static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r) return err; if (c->need_recovery) - err = ubifs_recover_size_accum(c, &r->key, deletion, + err = ubifs_recover_size_accum(c, &r->key, r->deletion, r->new_size); } @@ -248,68 +265,77 @@ static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r) } /** - * destroy_replay_tree - destroy the replay. - * @c: UBIFS file-system description object + * replay_entries_cmp - compare 2 replay entries. + * @priv: UBIFS file-system description object + * @a: first replay entry + * @a: second replay entry * - * Destroy the replay tree. + * This is a comparios function for 'list_sort()' which compares 2 replay + * entries @a and @b by comparing their sequence numer. Returns %1 if @a has + * greater sequence number and %-1 otherwise. */ -static void destroy_replay_tree(struct ubifs_info *c) +static int replay_entries_cmp(void *priv, struct list_head *a, + struct list_head *b) { - struct rb_node *this = c->replay_tree.rb_node; - struct replay_entry *r; - - while (this) { - if (this->rb_left) { - this = this->rb_left; - continue; - } else if (this->rb_right) { - this = this->rb_right; - continue; - } - r = rb_entry(this, struct replay_entry, rb); - this = rb_parent(this); - if (this) { - if (this->rb_left == &r->rb) - this->rb_left = NULL; - else - this->rb_right = NULL; - } - if (is_hash_key(c, &r->key)) - kfree(r->nm.name); - kfree(r); - } - c->replay_tree = RB_ROOT; + struct replay_entry *ra, *rb; + + cond_resched(); + if (a == b) + return 0; + + ra = list_entry(a, struct replay_entry, list); + rb = list_entry(b, struct replay_entry, list); + ubifs_assert(ra->sqnum != rb->sqnum); + if (ra->sqnum > rb->sqnum) + return 1; + return -1; } /** - * apply_replay_tree - apply the replay tree to the TNC. + * apply_replay_list - apply the replay list to the TNC. * @c: UBIFS file-system description object * - * Apply the replay tree. - * Returns zero in case of success and a negative error code in case of - * failure. + * Apply all entries in the replay list to the TNC. Returns zero in case of + * success and a negative error code in case of failure. */ -static int apply_replay_tree(struct ubifs_info *c) +static int apply_replay_list(struct ubifs_info *c) { - struct rb_node *this = rb_first(&c->replay_tree); + struct replay_entry *r; + int err; - while (this) { - struct replay_entry *r; - int err; + list_sort(c, &c->replay_list, &replay_entries_cmp); + list_for_each_entry(r, &c->replay_list, list) { cond_resched(); - r = rb_entry(this, struct replay_entry, rb); err = apply_replay_entry(c, r); if (err) return err; - this = rb_next(this); } + return 0; } /** - * insert_node - insert a node to the replay tree. + * destroy_replay_list - destroy the replay. + * @c: UBIFS file-system description object + * + * Destroy the replay list. + */ +static void destroy_replay_list(struct ubifs_info *c) +{ + struct replay_entry *r, *tmp; + + list_for_each_entry_safe(r, tmp, &c->replay_list, list) { + if (is_hash_key(c, &r->key)) + kfree(r->nm.name); + list_del(&r->list); + kfree(r); + } +} + +/** + * insert_node - insert a node to the replay list * @c: UBIFS file-system description object * @lnum: node logical eraseblock number * @offs: node offset @@ -321,39 +347,25 @@ static int apply_replay_tree(struct ubifs_info *c) * @old_size: truncation old size * @new_size: truncation new size * - * This function inserts a scanned non-direntry node to the replay tree. The - * replay tree is an RB-tree containing @struct replay_entry elements which are - * indexed by the sequence number. The replay tree is applied at the very end - * of the replay process. Since the tree is sorted in sequence number order, - * the older modifications are applied first. This function returns zero in - * case of success and a negative error code in case of failure. + * This function inserts a scanned non-direntry node to the replay list. The + * replay list contains @struct replay_entry elements, and we sort this list in + * sequence number order before applying it. The replay list is applied at the + * very end of the replay process. Since the list is sorted in sequence number + * order, the older modifications are applied first. This function returns zero + * in case of success and a negative error code in case of failure. */ static int insert_node(struct ubifs_info *c, int lnum, int offs, int len, union ubifs_key *key, unsigned long long sqnum, int deletion, int *used, loff_t old_size, loff_t new_size) { - struct rb_node **p = &c->replay_tree.rb_node, *parent = NULL; struct replay_entry *r; + dbg_mnt("add LEB %d:%d, key %s", lnum, offs, DBGKEY(key)); + if (key_inum(c, key) >= c->highest_inum) c->highest_inum = key_inum(c, key); - dbg_mnt("add LEB %d:%d, key %s", lnum, offs, DBGKEY(key)); - while (*p) { - parent = *p; - r = rb_entry(parent, struct replay_entry, rb); - if (sqnum < r->sqnum) { - p = &(*p)->rb_left; - continue; - } else if (sqnum > r->sqnum) { - p = &(*p)->rb_right; - continue; - } - ubifs_err("duplicate sqnum in replay"); - return -EINVAL; - } - r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL); if (!r) return -ENOMEM; @@ -363,19 +375,18 @@ static int insert_node(struct ubifs_info *c, int lnum, int offs, int len, r->lnum = lnum; r->offs = offs; r->len = len; + r->deletion = !!deletion; r->sqnum = sqnum; - r->flags = (deletion ? REPLAY_DELETION : 0); + key_copy(c, key, &r->key); r->old_size = old_size; r->new_size = new_size; - key_copy(c, key, &r->key); - rb_link_node(&r->rb, parent, p); - rb_insert_color(&r->rb, &c->replay_tree); + list_add_tail(&r->list, &c->replay_list); return 0; } /** - * insert_dent - insert a directory entry node into the replay tree. + * insert_dent - insert a directory entry node into the replay list. * @c: UBIFS file-system description object * @lnum: node logical eraseblock number * @offs: node offset @@ -387,43 +398,25 @@ static int insert_node(struct ubifs_info *c, int lnum, int offs, int len, * @deletion: non-zero if this is a deletion * @used: number of bytes in use in a LEB * - * This function inserts a scanned directory entry node to the replay tree. - * Returns zero in case of success and a negative error code in case of - * failure. - * - * This function is also used for extended attribute entries because they are - * implemented as directory entry nodes. + * This function inserts a scanned directory entry node or an extended + * attribute entry to the replay list. Returns zero in case of success and a + * negative error code in case of failure. */ static int insert_dent(struct ubifs_info *c, int lnum, int offs, int len, union ubifs_key *key, const char *name, int nlen, unsigned long long sqnum, int deletion, int *used) { - struct rb_node **p = &c->replay_tree.rb_node, *parent = NULL; struct replay_entry *r; char *nbuf; + dbg_mnt("add LEB %d:%d, key %s", lnum, offs, DBGKEY(key)); if (key_inum(c, key) >= c->highest_inum) c->highest_inum = key_inum(c, key); - dbg_mnt("add LEB %d:%d, key %s", lnum, offs, DBGKEY(key)); - while (*p) { - parent = *p; - r = rb_entry(parent, struct replay_entry, rb); - if (sqnum < r->sqnum) { - p = &(*p)->rb_left; - continue; - } - if (sqnum > r->sqnum) { - p = &(*p)->rb_right; - continue; - } - ubifs_err("duplicate sqnum in replay"); - return -EINVAL; - } - r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL); if (!r) return -ENOMEM; + nbuf = kmalloc(nlen + 1, GFP_KERNEL); if (!nbuf) { kfree(r); @@ -435,17 +428,15 @@ static int insert_dent(struct ubifs_info *c, int lnum, int offs, int len, r->lnum = lnum; r->offs = offs; r->len = len; + r->deletion = !!deletion; r->sqnum = sqnum; + key_copy(c, key, &r->key); r->nm.len = nlen; memcpy(nbuf, name, nlen); nbuf[nlen] = '\0'; r->nm.name = nbuf; - r->flags = (deletion ? REPLAY_DELETION : 0); - key_copy(c, key, &r->key); - ubifs_assert(!*p); - rb_link_node(&r->rb, parent, p); - rb_insert_color(&r->rb, &c->replay_tree); + list_add_tail(&r->list, &c->replay_list); return 0; } @@ -482,29 +473,92 @@ int ubifs_validate_entry(struct ubifs_info *c, } /** + * is_last_bud - check if the bud is the last in the journal head. + * @c: UBIFS file-system description object + * @bud: bud description object + * + * This function checks if bud @bud is the last bud in its journal head. This + * information is then used by 'replay_bud()' to decide whether the bud can + * have corruptions or not. Indeed, only last buds can be corrupted by power + * cuts. Returns %1 if this is the last bud, and %0 if not. + */ +static int is_last_bud(struct ubifs_info *c, struct ubifs_bud *bud) +{ + struct ubifs_jhead *jh = &c->jheads[bud->jhead]; + struct ubifs_bud *next; + uint32_t data; + int err; + + if (list_is_last(&bud->list, &jh->buds_list)) + return 1; + + /* + * The following is a quirk to make sure we work correctly with UBIFS + * images used with older UBIFS. + * + * Normally, the last bud will be the last in the journal head's list + * of bud. However, there is one exception if the UBIFS image belongs + * to older UBIFS. This is fairly unlikely: one would need to use old + * UBIFS, then have a power cut exactly at the right point, and then + * try to mount this image with new UBIFS. + * + * The exception is: it is possible to have 2 buds A and B, A goes + * before B, and B is the last, bud B is contains no data, and bud A is + * corrupted at the end. The reason is that in older versions when the + * journal code switched the next bud (from A to B), it first added a + * log reference node for the new bud (B), and only after this it + * synchronized the write-buffer of current bud (A). But later this was + * changed and UBIFS started to always synchronize the write-buffer of + * the bud (A) before writing the log reference for the new bud (B). + * + * But because older UBIFS always synchronized A's write-buffer before + * writing to B, we can recognize this exceptional situation but + * checking the contents of bud B - if it is empty, then A can be + * treated as the last and we can recover it. + * + * TODO: remove this piece of code in a couple of years (today it is + * 16.05.2011). + */ + next = list_entry(bud->list.next, struct ubifs_bud, list); + if (!list_is_last(&next->list, &jh->buds_list)) + return 0; + + err = ubi_read(c->ubi, next->lnum, (char *)&data, + next->start, 4); + if (err) + return 0; + + return data == 0xFFFFFFFF; +} + +/** * replay_bud - replay a bud logical eraseblock. * @c: UBIFS file-system description object - * @lnum: bud logical eraseblock number to replay - * @offs: bud start offset - * @jhead: journal head to which this bud belongs - * @free: amount of free space in the bud is returned here - * @dirty: amount of dirty space from padding and deletion nodes is returned - * here + * @b: bud entry which describes the bud * - * This function returns zero in case of success and a negative error code in - * case of failure. + * This function replays bud @bud, recovers it if needed, and adds all nodes + * from this bud to the replay list. Returns zero in case of success and a + * negative error code in case of failure. */ -static int replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead, - int *free, int *dirty) +static int replay_bud(struct ubifs_info *c, struct bud_entry *b) { - int err = 0, used = 0; + int is_last = is_last_bud(c, b->bud); + int err = 0, used = 0, lnum = b->bud->lnum, offs = b->bud->start; struct ubifs_scan_leb *sleb; struct ubifs_scan_node *snod; - struct ubifs_bud *bud; - dbg_mnt("replay bud LEB %d, head %d", lnum, jhead); - if (c->need_recovery) - sleb = ubifs_recover_leb(c, lnum, offs, c->sbuf, jhead != GCHD); + dbg_mnt("replay bud LEB %d, head %d, offs %d, is_last %d", + lnum, b->bud->jhead, offs, is_last); + + if (c->need_recovery && is_last) + /* + * Recover only last LEBs in the journal heads, because power + * cuts may cause corruptions only in these LEBs, because only + * these LEBs could possibly be written to at the power cut + * time. + */ + sleb = ubifs_recover_leb(c, lnum, offs, c->sbuf, + b->bud->jhead != GCHD); else sleb = ubifs_scan(c, lnum, offs, c->sbuf, 0); if (IS_ERR(sleb)) @@ -620,19 +674,13 @@ static int replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead, goto out; } - bud = ubifs_search_bud(c, lnum); - if (!bud) - BUG(); - + ubifs_assert(ubifs_search_bud(c, lnum)); ubifs_assert(sleb->endpt - offs >= used); ubifs_assert(sleb->endpt % c->min_io_size == 0); - if (sleb->endpt + c->min_io_size <= c->leb_size && !c->ro_mount) - err = ubifs_wbuf_seek_nolock(&c->jheads[jhead].wbuf, lnum, - sleb->endpt, UBI_SHORTTERM); - - *dirty = sleb->endpt - offs - used; - *free = c->leb_size - sleb->endpt; + b->dirty = sleb->endpt - offs - used; + b->free = c->leb_size - sleb->endpt; + dbg_mnt("bud LEB %d replied: dirty %d, free %d", lnum, b->dirty, b->free); out: ubifs_scan_destroy(sleb); @@ -646,55 +694,6 @@ out_dump: } /** - * insert_ref_node - insert a reference node to the replay tree. - * @c: UBIFS file-system description object - * @lnum: node logical eraseblock number - * @offs: node offset - * @sqnum: sequence number - * @free: amount of free space in bud - * @dirty: amount of dirty space from padding and deletion nodes - * - * This function inserts a reference node to the replay tree and returns zero - * in case of success or a negative error code in case of failure. - */ -static int insert_ref_node(struct ubifs_info *c, int lnum, int offs, - unsigned long long sqnum, int free, int dirty) -{ - struct rb_node **p = &c->replay_tree.rb_node, *parent = NULL; - struct replay_entry *r; - - dbg_mnt("add ref LEB %d:%d", lnum, offs); - while (*p) { - parent = *p; - r = rb_entry(parent, struct replay_entry, rb); - if (sqnum < r->sqnum) { - p = &(*p)->rb_left; - continue; - } else if (sqnum > r->sqnum) { - p = &(*p)->rb_right; - continue; - } - ubifs_err("duplicate sqnum in replay tree"); - return -EINVAL; - } - - r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL); - if (!r) - return -ENOMEM; - - r->lnum = lnum; - r->offs = offs; - r->sqnum = sqnum; - r->flags = REPLAY_REF; - r->free = free; - r->dirty = dirty; - - rb_link_node(&r->rb, parent, p); - rb_insert_color(&r->rb, &c->replay_tree); - return 0; -} - -/** * replay_buds - replay all buds. * @c: UBIFS file-system description object * @@ -704,17 +703,16 @@ static int insert_ref_node(struct ubifs_info *c, int lnum, int offs, static int replay_buds(struct ubifs_info *c) { struct bud_entry *b; - int err, uninitialized_var(free), uninitialized_var(dirty); + int err; + unsigned long long prev_sqnum = 0; list_for_each_entry(b, &c->replay_buds, list) { - err = replay_bud(c, b->bud->lnum, b->bud->start, b->bud->jhead, - &free, &dirty); - if (err) - return err; - err = insert_ref_node(c, b->bud->lnum, b->bud->start, b->sqnum, - free, dirty); + err = replay_bud(c, b); if (err) return err; + + ubifs_assert(b->sqnum > prev_sqnum); + prev_sqnum = b->sqnum; } return 0; @@ -1054,25 +1052,29 @@ int ubifs_replay_journal(struct ubifs_info *c) if (err) goto out; - err = apply_replay_tree(c); + err = apply_replay_list(c); + if (err) + goto out; + + err = set_buds_lprops(c); if (err) goto out; /* - * UBIFS budgeting calculations use @c->budg_uncommitted_idx variable - * to roughly estimate index growth. Things like @c->min_idx_lebs + * UBIFS budgeting calculations use @c->bi.uncommitted_idx variable + * to roughly estimate index growth. Things like @c->bi.min_idx_lebs * depend on it. This means we have to initialize it to make sure * budgeting works properly. */ - c->budg_uncommitted_idx = atomic_long_read(&c->dirty_zn_cnt); - c->budg_uncommitted_idx *= c->max_idx_node_sz; + c->bi.uncommitted_idx = atomic_long_read(&c->dirty_zn_cnt); + c->bi.uncommitted_idx *= c->max_idx_node_sz; ubifs_assert(c->bud_bytes <= c->max_bud_bytes || c->need_recovery); dbg_mnt("finished, log head LEB %d:%d, max_sqnum %llu, " "highest_inum %lu", c->lhead_lnum, c->lhead_offs, c->max_sqnum, (unsigned long)c->highest_inum); out: - destroy_replay_tree(c); + destroy_replay_list(c); destroy_bud_list(c); c->replaying = 0; return err; |