From 2e244d08369b510923fe8290d40d51b0e88bfebe Mon Sep 17 00:00:00 2001 From: Sebastian Siewior Date: Thu, 17 Jul 2008 14:16:09 +0200 Subject: Documentation: fix typo in ubifs.txt Signed-off-by: Sebastian Siewior Signed-off-by: Artem Bityutskiy --- Documentation/filesystems/ubifs.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Documentation/filesystems/ubifs.txt b/Documentation/filesystems/ubifs.txt index 540e9e7f59c5..6a0d70a22f05 100644 --- a/Documentation/filesystems/ubifs.txt +++ b/Documentation/filesystems/ubifs.txt @@ -57,7 +57,7 @@ Similarly to JFFS2, UBIFS supports on-the-flight compression which makes it possible to fit quite a lot of data to the flash. Similarly to JFFS2, UBIFS is tolerant of unclean reboots and power-cuts. -It does not need stuff like ckfs.ext2. UBIFS automatically replays its +It does not need stuff like fsck.ext2. UBIFS automatically replays its journal and recovers from crashes, ensuring that the on-flash data structures are consistent. -- cgit v1.2.3 From ce769caa50a3fc835b4fc1a6e1463ada127a2e8a Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Fri, 18 Jul 2008 12:54:21 +0300 Subject: UBIFS: print volume name as well We encouredge people to mount using volume name, not device numbers. So print the name of the mounted UBI volume, not just IDs. Signed-off-by: Artem Bityutskiy --- fs/ubifs/super.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c index ca1e2d4e03cc..43af934a7558 100644 --- a/fs/ubifs/super.c +++ b/fs/ubifs/super.c @@ -1122,8 +1122,8 @@ static int mount_ubifs(struct ubifs_info *c) if (err) goto out_infos; - ubifs_msg("mounted UBI device %d, volume %d", c->vi.ubi_num, - c->vi.vol_id); + ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"", + c->vi.ubi_num, c->vi.vol_id, c->vi.name); if (mounted_read_only) ubifs_msg("mounted read-only"); x = (long long)c->main_lebs * c->leb_size; -- cgit v1.2.3 From 182854b46f9feb6f1b03abe747bb2beeebf2adb0 Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Fri, 18 Jul 2008 18:54:29 +0300 Subject: UBIFS: fix budgeting calculations The 'ubifs_release_dirty_inode_budget()' was buggy and incorrectly freed the budget, which led to not freeing all dirty data budget. This patch fixes that. Also, this patch fixes ubifs_mkdir() which passed 1 in dirty_ino_d, which makes no sense. Well, it is harmless though. Also, add few more useful assertions. And improve few debugging messages. Signed-off-by: Artem Bityutskiy --- fs/ubifs/budget.c | 5 +++-- fs/ubifs/dir.c | 3 +-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/fs/ubifs/budget.c b/fs/ubifs/budget.c index d81fb9ed2b8e..12a1717db87c 100644 --- a/fs/ubifs/budget.c +++ b/fs/ubifs/budget.c @@ -686,9 +686,10 @@ void ubifs_convert_page_budget(struct ubifs_info *c) void ubifs_release_dirty_inode_budget(struct ubifs_info *c, struct ubifs_inode *ui) { - struct ubifs_budget_req req = {.dd_growth = c->inode_budget, - .dirtied_ino_d = ui->data_len}; + struct ubifs_budget_req req; + memset(&req, 0, sizeof(struct ubifs_budget_req)); + req.dd_growth = c->inode_budget + ui->data_len; ubifs_release_budget(c, &req); } diff --git a/fs/ubifs/dir.c b/fs/ubifs/dir.c index e90374be7d3b..a79e850fee6d 100644 --- a/fs/ubifs/dir.c +++ b/fs/ubifs/dir.c @@ -727,8 +727,7 @@ static int ubifs_mkdir(struct inode *dir, struct dentry *dentry, int mode) struct ubifs_inode *dir_ui = ubifs_inode(dir); struct ubifs_info *c = dir->i_sb->s_fs_info; int err, sz_change = CALC_DENT_SIZE(dentry->d_name.len); - struct ubifs_budget_req req = { .new_ino = 1, .new_dent = 1, - .dirtied_ino_d = 1 }; + struct ubifs_budget_req req = { .new_ino = 1, .new_dent = 1 }; /* * Budget request settings: new inode, new direntry and changing parent -- cgit v1.2.3 From 7d32c2bb143fa1ca3b0c420feb08a832d65395be Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Fri, 18 Jul 2008 18:54:29 +0300 Subject: UBIFS: improve debugging 1. Print inode mode in some of debugging messages 2. Add few more useful assertions Signed-off-by: Artem Bityutskiy --- fs/ubifs/file.c | 3 ++- fs/ubifs/super.c | 6 ++++-- 2 files changed, 6 insertions(+), 3 deletions(-) diff --git a/fs/ubifs/file.c b/fs/ubifs/file.c index 8565e586e533..01598f28020b 100644 --- a/fs/ubifs/file.c +++ b/fs/ubifs/file.c @@ -941,7 +941,8 @@ int ubifs_setattr(struct dentry *dentry, struct iattr *attr) struct inode *inode = dentry->d_inode; struct ubifs_info *c = inode->i_sb->s_fs_info; - dbg_gen("ino %lu, ia_valid %#x", inode->i_ino, attr->ia_valid); + dbg_gen("ino %lu, mode %#x, ia_valid %#x", + inode->i_ino, inode->i_mode, attr->ia_valid); err = inode_change_ok(inode, attr); if (err) return err; diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c index 43af934a7558..06e3b22a0c1b 100644 --- a/fs/ubifs/super.c +++ b/fs/ubifs/super.c @@ -299,7 +299,7 @@ static int ubifs_write_inode(struct inode *inode, int wait) return 0; } - dbg_gen("inode %lu", inode->i_ino); + dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode); err = ubifs_jnl_write_inode(c, inode, 0); if (err) ubifs_err("can't write inode %lu, error %d", inode->i_ino, err); @@ -323,9 +323,10 @@ static void ubifs_delete_inode(struct inode *inode) */ goto out; - dbg_gen("inode %lu", inode->i_ino); + dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode); ubifs_assert(!atomic_read(&inode->i_count)); ubifs_assert(inode->i_nlink == 0); + ubifs_assert(!ubifs_inode(inode)->dirty); truncate_inode_pages(&inode->i_data, 0); if (is_bad_inode(inode)) @@ -1469,6 +1470,7 @@ static void ubifs_put_super(struct super_block *sb) */ ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0); ubifs_assert(c->budg_idx_growth == 0); + ubifs_assert(c->budg_dd_growth == 0); ubifs_assert(c->budg_data_growth == 0); /* -- cgit v1.2.3 From 1e0f358e29cc91c8eb09e10cbf1f6bb58a62c795 Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Mon, 21 Jul 2008 10:59:53 +0300 Subject: UBIFS: free budget in delete_inode as well Although the inode is marked as clean when it is being deleted, it might stay and be used as orphan, and be marked as dirty. So we have to free the budget when we delete it. Signed-off-by: Artem Bityutskiy --- fs/ubifs/super.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c index 06e3b22a0c1b..884beed1dcb8 100644 --- a/fs/ubifs/super.c +++ b/fs/ubifs/super.c @@ -314,8 +314,9 @@ static void ubifs_delete_inode(struct inode *inode) { int err; struct ubifs_info *c = inode->i_sb->s_fs_info; + struct ubifs_inode *ui = ubifs_inode(inode); - if (ubifs_inode(inode)->xattr) + if (ui->xattr) /* * Extended attribute inode deletions are fully handled in * 'ubifs_removexattr()'. These inodes are special and have @@ -326,13 +327,12 @@ static void ubifs_delete_inode(struct inode *inode) dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode); ubifs_assert(!atomic_read(&inode->i_count)); ubifs_assert(inode->i_nlink == 0); - ubifs_assert(!ubifs_inode(inode)->dirty); truncate_inode_pages(&inode->i_data, 0); if (is_bad_inode(inode)) goto out; - ubifs_inode(inode)->ui_size = inode->i_size = 0; + ui->ui_size = inode->i_size = 0; err = ubifs_jnl_write_inode(c, inode, 1); if (err) /* @@ -341,6 +341,8 @@ static void ubifs_delete_inode(struct inode *inode) */ ubifs_err("can't write inode %lu, error %d", inode->i_ino, err); out: + if (ui->dirty) + ubifs_release_dirty_inode_budget(c, ui); clear_inode(inode); } -- cgit v1.2.3 From 16dfd804b44ef7156d1c201f100bd0d9dc6b7c4b Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Fri, 18 Jul 2008 16:47:41 +0300 Subject: UBIFS: fix error return in failure mode UBIFS recovery testing debug facility simulates media failures. When simulating an IO error, the error code returned must be -EIO but it was not always if the user switched off the debug recovery testing option at the same time. Signed-off-by: Adrian Hunter --- fs/ubifs/debug.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/fs/ubifs/debug.c b/fs/ubifs/debug.c index 4e3aaeba4eca..0adfb29b8503 100644 --- a/fs/ubifs/debug.c +++ b/fs/ubifs/debug.c @@ -2208,16 +2208,17 @@ int dbg_leb_read(struct ubi_volume_desc *desc, int lnum, char *buf, int offset, int dbg_leb_write(struct ubi_volume_desc *desc, int lnum, const void *buf, int offset, int len, int dtype) { - int err; + int err, failing; if (in_failure_mode(desc)) return -EIO; - if (do_fail(desc, lnum, 1)) + failing = do_fail(desc, lnum, 1); + if (failing) cut_data(buf, len); err = ubi_leb_write(desc, lnum, buf, offset, len, dtype); if (err) return err; - if (in_failure_mode(desc)) + if (failing) return -EIO; return 0; } -- cgit v1.2.3 From 2fb42b11f61cbcef7dfc225c1d26c4511436583d Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Fri, 18 Jul 2008 17:56:37 +0300 Subject: UBIFS: ensure UBIFS switches to read-only on error UBI transparently handles write errors by automatically copying and remapping the affected eraseblock. If UBI is unable to do that, for example its pool of eraseblocks reserved for bad block handling is empty, then the error is propagated to UBIFS. UBIFS must protect the media from falling into an inconsistent state by immediately switching to read-only mode. In the case of log updates, this was not being done. Signed-off-by: Adrian Hunter --- fs/ubifs/log.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/fs/ubifs/log.c b/fs/ubifs/log.c index 36857b9ed59e..e14829e50693 100644 --- a/fs/ubifs/log.c +++ b/fs/ubifs/log.c @@ -317,6 +317,8 @@ int ubifs_add_bud_to_log(struct ubifs_info *c, int jhead, int lnum, int offs) return 0; out_unlock: + if (err != -EAGAIN) + ubifs_ro_mode(c, err); mutex_unlock(&c->log_mutex); kfree(ref); kfree(bud); -- cgit v1.2.3 From ff46d7b3e0870a70331b069372c36fbc43018c2d Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Mon, 21 Jul 2008 15:39:05 +0300 Subject: UBIFS: make ubifs_ro_mode() not inline We use ubifs_ro_mode() quite a lot, and not in fast-path, so there is no reason to blow the code up by having it inlined. Also, we usually want R/O mode change to be seen to other CPUs as soon as possible, so when we make this a function call, we will automatically have a memory barrier. Signed-off-by: Adrian Hunter Signed-off-by: Artem Bityutskiy --- fs/ubifs/io.c | 14 ++++++++++++++ fs/ubifs/misc.h | 14 -------------- fs/ubifs/ubifs.h | 1 + 3 files changed, 15 insertions(+), 14 deletions(-) diff --git a/fs/ubifs/io.c b/fs/ubifs/io.c index 3374f91b6709..054363f2b207 100644 --- a/fs/ubifs/io.c +++ b/fs/ubifs/io.c @@ -53,6 +53,20 @@ #include #include "ubifs.h" +/** + * ubifs_ro_mode - switch UBIFS to read read-only mode. + * @c: UBIFS file-system description object + * @err: error code which is the reason of switching to R/O mode + */ +void ubifs_ro_mode(struct ubifs_info *c, int err) +{ + if (!c->ro_media) { + c->ro_media = 1; + ubifs_warn("switched to read-only mode, error %d", err); + dbg_dump_stack(); + } +} + /** * ubifs_check_node - check node. * @c: UBIFS file-system description object diff --git a/fs/ubifs/misc.h b/fs/ubifs/misc.h index 4beccfc256d2..cd83ffc8101c 100644 --- a/fs/ubifs/misc.h +++ b/fs/ubifs/misc.h @@ -79,20 +79,6 @@ static inline struct ubifs_inode *ubifs_inode(const struct inode *inode) return container_of(inode, struct ubifs_inode, vfs_inode); } -/** - * ubifs_ro_mode - switch UBIFS to read read-only mode. - * @c: UBIFS file-system description object - * @err: error code which is the reason of switching to R/O mode - */ -static inline void ubifs_ro_mode(struct ubifs_info *c, int err) -{ - if (!c->ro_media) { - c->ro_media = 1; - ubifs_warn("switched to read-only mode, error %d", err); - dbg_dump_stack(); - } -} - /** * ubifs_compr_present - check if compressor was compiled in. * @compr_type: compressor type to check diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h index e4f89f271827..c488d43b6359 100644 --- a/fs/ubifs/ubifs.h +++ b/fs/ubifs/ubifs.h @@ -1346,6 +1346,7 @@ extern struct backing_dev_info ubifs_backing_dev_info; extern struct ubifs_compressor *ubifs_compressors[UBIFS_COMPR_TYPES_CNT]; /* io.c */ +void ubifs_ro_mode(struct ubifs_info *c, int err); int ubifs_wbuf_write_nolock(struct ubifs_wbuf *wbuf, void *buf, int len); int ubifs_wbuf_seek_nolock(struct ubifs_wbuf *wbuf, int lnum, int offs, int dtype); -- cgit v1.2.3 From fbfa6c884aae2aff479eb8c996c564b1a34eae30 Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Tue, 22 Jul 2008 11:52:52 +0300 Subject: UBIFS: do not write orphans back Orphan inodes are deleted inodes which will disappear after FS re-mount. There is not need to write orphan inodes back, because they are not needed on the flash media. So optimize orphans a little by not writing them back. Just mark them as clean, free the budget, and report success to VFS. Signed-off-by: Artem Bityutskiy --- fs/ubifs/super.c | 18 +++++++++++++----- 1 file changed, 13 insertions(+), 5 deletions(-) diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c index 884beed1dcb8..13e90b0dd95d 100644 --- a/fs/ubifs/super.c +++ b/fs/ubifs/super.c @@ -278,7 +278,7 @@ static void ubifs_destroy_inode(struct inode *inode) */ static int ubifs_write_inode(struct inode *inode, int wait) { - int err; + int err = 0; struct ubifs_info *c = inode->i_sb->s_fs_info; struct ubifs_inode *ui = ubifs_inode(inode); @@ -299,10 +299,18 @@ static int ubifs_write_inode(struct inode *inode, int wait) return 0; } - dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode); - err = ubifs_jnl_write_inode(c, inode, 0); - if (err) - ubifs_err("can't write inode %lu, error %d", inode->i_ino, err); + /* + * As an optimization, do not write orphan inodes to the media just + * because this is not needed. + */ + dbg_gen("inode %lu, mode %#x, nlink %u", + inode->i_ino, (int)inode->i_mode, inode->i_nlink); + if (inode->i_nlink) { + err = ubifs_jnl_write_inode(c, inode, 0); + if (err) + ubifs_err("can't write inode %lu, error %d", + inode->i_ino, err); + } ui->dirty = 0; mutex_unlock(&ui->ui_mutex); -- cgit v1.2.3 From 1f28681ad34a0c7e51dc5070c84b53f7bd34f44c Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Tue, 22 Jul 2008 12:06:13 +0300 Subject: UBIFS: remove unneeded function parameter Simplify 'ubifs_jnl_write_inode()' by removing the 'deletion' parameter which is not really needed because we may test inode->i_nlink and check whether this is a deletion or not. Signed-off-by: Artem Bityutskiy --- fs/ubifs/journal.c | 19 +++++++------------ fs/ubifs/super.c | 4 ++-- fs/ubifs/ubifs.h | 3 +-- 3 files changed, 10 insertions(+), 16 deletions(-) diff --git a/fs/ubifs/journal.c b/fs/ubifs/journal.c index 283155abe5f5..666ad82ec51a 100644 --- a/fs/ubifs/journal.c +++ b/fs/ubifs/journal.c @@ -750,30 +750,25 @@ out_free: * ubifs_jnl_write_inode - flush inode to the journal. * @c: UBIFS file-system description object * @inode: inode to flush - * @deletion: inode has been deleted * * This function writes inode @inode to the journal. If the inode is * synchronous, it also synchronizes the write-buffer. Returns zero in case of * success and a negative error code in case of failure. */ -int ubifs_jnl_write_inode(struct ubifs_info *c, const struct inode *inode, - int deletion) +int ubifs_jnl_write_inode(struct ubifs_info *c, const struct inode *inode) { - int err, len, lnum, offs, sync = 0; + int err, lnum, offs; struct ubifs_ino_node *ino; struct ubifs_inode *ui = ubifs_inode(inode); + int sync = 0, len = UBIFS_INO_NODE_SZ, last_reference = !inode->i_nlink; - dbg_jnl("ino %lu%s", inode->i_ino, - deletion ? " (last reference)" : ""); - if (deletion) - ubifs_assert(inode->i_nlink == 0); + dbg_jnl("ino %lu, nlink %u", inode->i_ino, inode->i_nlink); - len = UBIFS_INO_NODE_SZ; /* * If the inode is being deleted, do not write the attached data. No * need to synchronize the write-buffer either. */ - if (!deletion) { + if (!last_reference) { len += ui->data_len; sync = IS_SYNC(inode); } @@ -786,7 +781,7 @@ int ubifs_jnl_write_inode(struct ubifs_info *c, const struct inode *inode, if (err) goto out_free; - pack_inode(c, ino, inode, 1, deletion); + pack_inode(c, ino, inode, 1, last_reference); err = write_head(c, BASEHD, ino, len, &lnum, &offs, sync); if (err) goto out_release; @@ -795,7 +790,7 @@ int ubifs_jnl_write_inode(struct ubifs_info *c, const struct inode *inode, inode->i_ino); release_head(c, BASEHD); - if (deletion) { + if (last_reference) { err = ubifs_tnc_remove_ino(c, inode->i_ino); if (err) goto out_ro; diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c index 13e90b0dd95d..cf1fb6cffa09 100644 --- a/fs/ubifs/super.c +++ b/fs/ubifs/super.c @@ -306,7 +306,7 @@ static int ubifs_write_inode(struct inode *inode, int wait) dbg_gen("inode %lu, mode %#x, nlink %u", inode->i_ino, (int)inode->i_mode, inode->i_nlink); if (inode->i_nlink) { - err = ubifs_jnl_write_inode(c, inode, 0); + err = ubifs_jnl_write_inode(c, inode); if (err) ubifs_err("can't write inode %lu, error %d", inode->i_ino, err); @@ -341,7 +341,7 @@ static void ubifs_delete_inode(struct inode *inode) goto out; ui->ui_size = inode->i_size = 0; - err = ubifs_jnl_write_inode(c, inode, 1); + err = ubifs_jnl_write_inode(c, inode); if (err) /* * Worst case we have a lost orphan inode wasting space, so a diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h index c488d43b6359..6ddd1de2ea64 100644 --- a/fs/ubifs/ubifs.h +++ b/fs/ubifs/ubifs.h @@ -1400,8 +1400,7 @@ int ubifs_jnl_update(struct ubifs_info *c, const struct inode *dir, int deletion, int xent); int ubifs_jnl_write_data(struct ubifs_info *c, const struct inode *inode, const union ubifs_key *key, const void *buf, int len); -int ubifs_jnl_write_inode(struct ubifs_info *c, const struct inode *inode, - int last_reference); +int ubifs_jnl_write_inode(struct ubifs_info *c, const struct inode *inode); int ubifs_jnl_rename(struct ubifs_info *c, const struct inode *old_dir, const struct dentry *old_dentry, const struct inode *new_dir, -- cgit v1.2.3 From fd6c6b51e3677937090314b20b00f2194900d81b Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Tue, 22 Jul 2008 12:19:09 +0300 Subject: UBIFS: remove another unneeded function parameter The 'last_reference' parameter of 'pack_inode()' is not really needed because 'inode->i_nlink' may be tested instead. Zap it. Signed-off-by: Artem Bityutskiy --- fs/ubifs/journal.c | 30 ++++++++++++++---------------- 1 file changed, 14 insertions(+), 16 deletions(-) diff --git a/fs/ubifs/journal.c b/fs/ubifs/journal.c index 666ad82ec51a..3bc3fc947099 100644 --- a/fs/ubifs/journal.c +++ b/fs/ubifs/journal.c @@ -447,13 +447,11 @@ static int get_dent_type(int mode) * @ino: buffer in which to pack inode node * @inode: inode to pack * @last: indicates the last node of the group - * @last_reference: non-zero if this is a deletion inode */ static void pack_inode(struct ubifs_info *c, struct ubifs_ino_node *ino, - const struct inode *inode, int last, - int last_reference) + const struct inode *inode, int last) { - int data_len = 0; + int data_len = 0, last_reference = !inode->i_nlink; struct ubifs_inode *ui = ubifs_inode(inode); ino->ch.node_type = UBIFS_INO_NODE; @@ -596,9 +594,9 @@ int ubifs_jnl_update(struct ubifs_info *c, const struct inode *dir, ubifs_prep_grp_node(c, dent, dlen, 0); ino = (void *)dent + aligned_dlen; - pack_inode(c, ino, inode, 0, last_reference); + pack_inode(c, ino, inode, 0); ino = (void *)ino + aligned_ilen; - pack_inode(c, ino, dir, 1, 0); + pack_inode(c, ino, dir, 1); if (last_reference) { err = ubifs_add_orphan(c, inode->i_ino); @@ -781,7 +779,7 @@ int ubifs_jnl_write_inode(struct ubifs_info *c, const struct inode *inode) if (err) goto out_free; - pack_inode(c, ino, inode, 1, last_reference); + pack_inode(c, ino, inode, 1); err = write_head(c, BASEHD, ino, len, &lnum, &offs, sync); if (err) goto out_release; @@ -912,16 +910,16 @@ int ubifs_jnl_rename(struct ubifs_info *c, const struct inode *old_dir, p = (void *)dent2 + aligned_dlen2; if (new_inode) { - pack_inode(c, p, new_inode, 0, last_reference); + pack_inode(c, p, new_inode, 0); p += ALIGN(ilen, 8); } if (!move) - pack_inode(c, p, old_dir, 1, 0); + pack_inode(c, p, old_dir, 1); else { - pack_inode(c, p, old_dir, 0, 0); + pack_inode(c, p, old_dir, 0); p += ALIGN(plen, 8); - pack_inode(c, p, new_dir, 1, 0); + pack_inode(c, p, new_dir, 1); } if (last_reference) { @@ -1126,7 +1124,7 @@ int ubifs_jnl_truncate(struct ubifs_info *c, const struct inode *inode, if (err) goto out_free; - pack_inode(c, ino, inode, 0, 0); + pack_inode(c, ino, inode, 0); ubifs_prep_grp_node(c, trun, UBIFS_TRUN_NODE_SZ, dlen ? 0 : 1); if (dlen) ubifs_prep_grp_node(c, dn, dlen, 1); @@ -1246,9 +1244,9 @@ int ubifs_jnl_delete_xattr(struct ubifs_info *c, const struct inode *host, ubifs_prep_grp_node(c, xent, xlen, 0); ino = (void *)xent + aligned_xlen; - pack_inode(c, ino, inode, 0, 1); + pack_inode(c, ino, inode, 0); ino = (void *)ino + UBIFS_INO_NODE_SZ; - pack_inode(c, ino, host, 1, 0); + pack_inode(c, ino, host, 1); err = write_head(c, BASEHD, xent, len, &lnum, &xent_offs, sync); if (!sync && !err) @@ -1339,8 +1337,8 @@ int ubifs_jnl_change_xattr(struct ubifs_info *c, const struct inode *inode, if (err) goto out_free; - pack_inode(c, ino, host, 0, 0); - pack_inode(c, (void *)ino + aligned_len1, inode, 1, 0); + pack_inode(c, ino, host, 0); + pack_inode(c, (void *)ino + aligned_len1, inode, 1); err = write_head(c, BASEHD, ino, aligned_len, &lnum, &offs, 0); if (!sync && !err) { -- cgit v1.2.3 From 014eb04b03202dc75c1c749df4246d98045f5e69 Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Mon, 21 Jul 2008 17:14:29 +0300 Subject: UBIFS: increment commit number earlier Increment the commit number at the beginnig of the commit, instead of doing this after the commit. This is needed for further optimizations. Signed-off-by: Artem Bityutskiy --- fs/ubifs/commit.c | 3 ++- fs/ubifs/log.c | 2 +- fs/ubifs/orphan.c | 4 ++-- fs/ubifs/ubifs.h | 3 ++- 4 files changed, 7 insertions(+), 5 deletions(-) diff --git a/fs/ubifs/commit.c b/fs/ubifs/commit.c index 3b516316c9b3..0a6aa2cc78f0 100644 --- a/fs/ubifs/commit.c +++ b/fs/ubifs/commit.c @@ -74,6 +74,7 @@ static int do_commit(struct ubifs_info *c) goto out_up; } + c->cmt_no += 1; err = ubifs_gc_start_commit(c); if (err) goto out_up; @@ -115,7 +116,7 @@ static int do_commit(struct ubifs_info *c) goto out; mutex_lock(&c->mst_mutex); - c->mst_node->cmt_no = cpu_to_le64(++c->cmt_no); + c->mst_node->cmt_no = cpu_to_le64(c->cmt_no); c->mst_node->log_lnum = cpu_to_le32(new_ltail_lnum); c->mst_node->root_lnum = cpu_to_le32(zroot.lnum); c->mst_node->root_offs = cpu_to_le32(zroot.offs); diff --git a/fs/ubifs/log.c b/fs/ubifs/log.c index e14829e50693..3e0aa7367556 100644 --- a/fs/ubifs/log.c +++ b/fs/ubifs/log.c @@ -412,7 +412,7 @@ int ubifs_log_start_commit(struct ubifs_info *c, int *ltail_lnum) return -ENOMEM; cs->ch.node_type = UBIFS_CS_NODE; - cs->cmt_no = cpu_to_le64(c->cmt_no + 1); + cs->cmt_no = cpu_to_le64(c->cmt_no); ubifs_prepare_node(c, cs, UBIFS_CS_NODE_SZ, 0); /* diff --git a/fs/ubifs/orphan.c b/fs/ubifs/orphan.c index 3afeb9242c6a..02d3462f4d3e 100644 --- a/fs/ubifs/orphan.c +++ b/fs/ubifs/orphan.c @@ -310,10 +310,10 @@ static int write_orph_node(struct ubifs_info *c, int atomic) c->cmt_orphans -= cnt; spin_unlock(&c->orphan_lock); if (c->cmt_orphans) - orph->cmt_no = cpu_to_le64(c->cmt_no + 1); + orph->cmt_no = cpu_to_le64(c->cmt_no); else /* Mark the last node of the commit */ - orph->cmt_no = cpu_to_le64((c->cmt_no + 1) | (1ULL << 63)); + orph->cmt_no = cpu_to_le64((c->cmt_no) | (1ULL << 63)); ubifs_assert(c->ohead_offs + len <= c->leb_size); ubifs_assert(c->ohead_lnum >= c->orph_first); ubifs_assert(c->ohead_lnum <= c->orph_last); diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h index 6ddd1de2ea64..21502b6040f0 100644 --- a/fs/ubifs/ubifs.h +++ b/fs/ubifs/ubifs.h @@ -865,7 +865,8 @@ struct ubifs_mount_opts { * @highest_inum: highest used inode number * @vfs_gen: VFS inode generation counter * @max_sqnum: current global sequence number - * @cmt_no: commit number (last successfully completed commit) + * @cmt_no: commit number of the last successfully completed commit, protected + * by @commit_sem * @cnt_lock: protects @highest_inum, @vfs_gen, and @max_sqnum counters * @fmt_version: UBIFS on-flash format version * @uuid: UUID from super block -- cgit v1.2.3 From de94eb558b542873d3f6f9ede1b8575fb5662248 Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Tue, 22 Jul 2008 13:06:20 +0300 Subject: UBIFS: optimize deletions Every time anything is deleted, UBIFS writes the deletion inode node twice - once in 'ubifs_jnl_update()' and the second time in 'ubifs_jnl_write_inode()'. However, the second write is not needed if no commit happened after 'ubifs_jnl_update()'. This patch checks that condition and avoids writing the deletion inode for the second time. Signed-off-by: Artem Bityutskiy --- fs/ubifs/journal.c | 60 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ fs/ubifs/super.c | 6 ++++-- fs/ubifs/ubifs.h | 12 ++++++++--- 3 files changed, 73 insertions(+), 5 deletions(-) diff --git a/fs/ubifs/journal.c b/fs/ubifs/journal.c index 3bc3fc947099..0bcee7d221e8 100644 --- a/fs/ubifs/journal.c +++ b/fs/ubifs/journal.c @@ -604,6 +604,7 @@ int ubifs_jnl_update(struct ubifs_info *c, const struct inode *dir, release_head(c, BASEHD); goto out_finish; } + ui->del_cmtno = c->cmt_no; } err = write_head(c, BASEHD, dent, len, &lnum, &dent_offs, sync); @@ -820,6 +821,64 @@ out_free: return err; } +/** + * ubifs_jnl_write_inode - delete an inode. + * @c: UBIFS file-system description object + * @inode: inode to delete + * + * This function deletes inode @inode which includes removing it from orphans, + * deleting it from TNC and, in some cases, writing a deletion inode to the + * journal. + * + * When regular file inodes are unlinked or a directory inode is removed, the + * 'ubifs_jnl_update()' function write corresponding deletion inode and + * direntry to the media, and adds the inode to orphans. After this, when the + * last reference to this inode has been dropped, this function is called. In + * general, it has to write one more deletion inode to the media, because if + * a commit happened between 'ubifs_jnl_update()' and + * 'ubifs_jnl_delete_inode()', the deletion inode is not in the journal + * anymore, and in fact it might be not on the flash anymore, becouse it might + * have been garbage-collected already. And for optimization reasond UBIFS does + * not read the orphan area if it has been unmounted cleanly, so it would have + * no indication in the journal that there is a deleted inode which has to be + * removed from TNC. + * + * However, if there was no commit between 'ubifs_jnl_update()' and + * 'ubifs_jnl_delete_inode()', then there is no need to write the deletion + * inode to the media for the second time. And this is quite typical case. + * + * This function returns zero in case of success and a negative error code in + * case of failure. + */ +int ubifs_jnl_delete_inode(struct ubifs_info *c, const struct inode *inode) +{ + int err; + struct ubifs_inode *ui = ubifs_inode(inode); + + ubifs_assert(inode->i_nlink == 0); + + if (ui->del_cmtno != c->cmt_no) + /* A commit happened for sure */ + return ubifs_jnl_write_inode(c, inode); + + down_read(&c->commit_sem); + /* + * Check commit number again, because the first test has been done + * without @c->commit_sem, so a commit might have happened. + */ + if (ui->del_cmtno != c->cmt_no) { + up_read(&c->commit_sem); + return ubifs_jnl_write_inode(c, inode); + } + + ubifs_delete_orphan(c, inode->i_ino); + err = ubifs_tnc_remove_ino(c, inode->i_ino); + if (err) + ubifs_ro_mode(c, err); + up_read(&c->commit_sem); + return err; +} + /** * ubifs_jnl_rename - rename a directory entry. * @c: UBIFS file-system description object @@ -928,6 +987,7 @@ int ubifs_jnl_rename(struct ubifs_info *c, const struct inode *old_dir, release_head(c, BASEHD); goto out_finish; } + new_ui->del_cmtno = c->cmt_no; } err = write_head(c, BASEHD, dent, len, &lnum, &offs, sync); diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c index cf1fb6cffa09..6cc4175f23c1 100644 --- a/fs/ubifs/super.c +++ b/fs/ubifs/super.c @@ -341,13 +341,15 @@ static void ubifs_delete_inode(struct inode *inode) goto out; ui->ui_size = inode->i_size = 0; - err = ubifs_jnl_write_inode(c, inode); + err = ubifs_jnl_delete_inode(c, inode); if (err) /* * Worst case we have a lost orphan inode wasting space, so a * simple error message is ok here. */ - ubifs_err("can't write inode %lu, error %d", inode->i_ino, err); + ubifs_err("can't delete inode %lu, error %d", + inode->i_ino, err); + out: if (ui->dirty) ubifs_release_dirty_inode_budget(c, ui); diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h index 21502b6040f0..dfb4b93614ff 100644 --- a/fs/ubifs/ubifs.h +++ b/fs/ubifs/ubifs.h @@ -322,6 +322,8 @@ struct ubifs_gced_idx_leb { * struct ubifs_inode - UBIFS in-memory inode description. * @vfs_inode: VFS inode description object * @creat_sqnum: sequence number at time of creation + * @del_cmtno: commit number corresponding to the time the inode was deleted, + * protected by @c->commit_sem; * @xattr_size: summarized size of all extended attributes in bytes * @xattr_cnt: count of extended attributes this inode has * @xattr_names: sum of lengths of all extended attribute names belonging to @@ -372,7 +374,10 @@ struct ubifs_gced_idx_leb { */ struct ubifs_inode { struct inode vfs_inode; - unsigned long long creat_sqnum; + union { + unsigned long long creat_sqnum; + unsigned long long del_cmtno; + }; unsigned int xattr_size; unsigned int xattr_cnt; unsigned int xattr_names; @@ -779,7 +784,7 @@ struct ubifs_compressor { /** * struct ubifs_budget_req - budget requirements of an operation. * - * @fast: non-zero if the budgeting should try to aquire budget quickly and + * @fast: non-zero if the budgeting should try to acquire budget quickly and * should not try to call write-back * @recalculate: non-zero if @idx_growth, @data_growth, and @dd_growth fields * have to be re-calculated @@ -860,7 +865,7 @@ struct ubifs_mount_opts { * struct ubifs_info - UBIFS file-system description data structure * (per-superblock). * @vfs_sb: VFS @struct super_block object - * @bdi: backing device info object to make VFS happy and disable readahead + * @bdi: backing device info object to make VFS happy and disable read-ahead * * @highest_inum: highest used inode number * @vfs_gen: VFS inode generation counter @@ -1402,6 +1407,7 @@ int ubifs_jnl_update(struct ubifs_info *c, const struct inode *dir, int ubifs_jnl_write_data(struct ubifs_info *c, const struct inode *inode, const union ubifs_key *key, const void *buf, int len); int ubifs_jnl_write_inode(struct ubifs_info *c, const struct inode *inode); +int ubifs_jnl_delete_inode(struct ubifs_info *c, const struct inode *inode); int ubifs_jnl_rename(struct ubifs_info *c, const struct inode *old_dir, const struct dentry *old_dentry, const struct inode *new_dir, -- cgit v1.2.3 From bc813355c704e5916a86dd4b96fd226bfa3fc6ca Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Wed, 23 Jul 2008 15:23:11 +0300 Subject: UBIFS: do not union creat_sqnum and del_cmtno The values in these two fields need to be preserved independently and so a union cannot be used. Signed-off-by: Adrian Hunter --- fs/ubifs/ubifs.h | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h index dfb4b93614ff..d342c6907244 100644 --- a/fs/ubifs/ubifs.h +++ b/fs/ubifs/ubifs.h @@ -374,10 +374,8 @@ struct ubifs_gced_idx_leb { */ struct ubifs_inode { struct inode vfs_inode; - union { - unsigned long long creat_sqnum; - unsigned long long del_cmtno; - }; + unsigned long long creat_sqnum; + unsigned long long del_cmtno; unsigned int xattr_size; unsigned int xattr_cnt; unsigned int xattr_names; -- cgit v1.2.3 From 7d62ff2c396470bb62a3853f14d3962eac1da974 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Wed, 23 Jul 2008 15:48:39 +0300 Subject: UBIFS: fix typos in comments Signed-off-by: Adrian Hunter --- fs/ubifs/journal.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/fs/ubifs/journal.c b/fs/ubifs/journal.c index 0bcee7d221e8..25de6fde383f 100644 --- a/fs/ubifs/journal.c +++ b/fs/ubifs/journal.c @@ -822,7 +822,7 @@ out_free: } /** - * ubifs_jnl_write_inode - delete an inode. + * ubifs_jnl_delete_inode - delete an inode. * @c: UBIFS file-system description object * @inode: inode to delete * @@ -831,21 +831,21 @@ out_free: * journal. * * When regular file inodes are unlinked or a directory inode is removed, the - * 'ubifs_jnl_update()' function write corresponding deletion inode and + * 'ubifs_jnl_update()' function writes a corresponding deletion inode and * direntry to the media, and adds the inode to orphans. After this, when the * last reference to this inode has been dropped, this function is called. In * general, it has to write one more deletion inode to the media, because if * a commit happened between 'ubifs_jnl_update()' and * 'ubifs_jnl_delete_inode()', the deletion inode is not in the journal - * anymore, and in fact it might be not on the flash anymore, becouse it might - * have been garbage-collected already. And for optimization reasond UBIFS does + * anymore, and in fact it might not be on the flash anymore, because it might + * have been garbage-collected already. And for optimization reasons UBIFS does * not read the orphan area if it has been unmounted cleanly, so it would have * no indication in the journal that there is a deleted inode which has to be * removed from TNC. * * However, if there was no commit between 'ubifs_jnl_update()' and * 'ubifs_jnl_delete_inode()', then there is no need to write the deletion - * inode to the media for the second time. And this is quite typical case. + * inode to the media for the second time. And this is quite a typical case. * * This function returns zero in case of success and a negative error code in * case of failure. -- cgit v1.2.3 From f769108424a19c7758546d1d7d19f098b1a33759 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Wed, 23 Jul 2008 16:55:55 +0300 Subject: UBIFS: correct orphan deletion order The debug function that checks orphans, does so using the TNC mutex. That means it will not see a correct picture if the inode is removed from the orphan tree before it is removed from TNC. Signed-off-by: Adrian Hunter --- fs/ubifs/journal.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/ubifs/journal.c b/fs/ubifs/journal.c index 25de6fde383f..acdae00aaa54 100644 --- a/fs/ubifs/journal.c +++ b/fs/ubifs/journal.c @@ -871,10 +871,11 @@ int ubifs_jnl_delete_inode(struct ubifs_info *c, const struct inode *inode) return ubifs_jnl_write_inode(c, inode); } - ubifs_delete_orphan(c, inode->i_ino); err = ubifs_tnc_remove_ino(c, inode->i_ino); if (err) ubifs_ro_mode(c, err); + else + ubifs_delete_orphan(c, inode->i_ino); up_read(&c->commit_sem); return err; } -- cgit v1.2.3 From 547000da6412c45456ff2ff44a171d01027bd727 Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Thu, 24 Jul 2008 14:42:05 +0300 Subject: UBIFS: improve budgeting checks Budgeting is a crucial UBIFS subsystem - add more assertions to improve requests checking. This is not compiled in when UBIFS debugging is disabled. Signed-off-by: Artem Bityutskiy --- fs/ubifs/budget.c | 12 ++++++++++++ fs/ubifs/ubifs.h | 8 +++++++- 2 files changed, 19 insertions(+), 1 deletion(-) diff --git a/fs/ubifs/budget.c b/fs/ubifs/budget.c index 12a1717db87c..f5afce5f37bd 100644 --- a/fs/ubifs/budget.c +++ b/fs/ubifs/budget.c @@ -543,6 +543,12 @@ int ubifs_budget_space(struct ubifs_info *c, struct ubifs_budget_req *req) int err, idx_growth, data_growth, dd_growth; struct retries_info ri; + ubifs_assert(req->new_page <= 1); + ubifs_assert(req->dirtied_page <= 1); + ubifs_assert(req->new_dent <= 1); + ubifs_assert(req->mod_dent <= 1); + ubifs_assert(req->new_ino <= 1); + ubifs_assert(req->new_ino_d <= UBIFS_MAX_INO_DATA); ubifs_assert(req->dirtied_ino <= 4); ubifs_assert(req->dirtied_ino_d <= UBIFS_MAX_INO_DATA * 4); @@ -618,6 +624,12 @@ again: */ void ubifs_release_budget(struct ubifs_info *c, struct ubifs_budget_req *req) { + ubifs_assert(req->new_page <= 1); + ubifs_assert(req->dirtied_page <= 1); + ubifs_assert(req->new_dent <= 1); + ubifs_assert(req->mod_dent <= 1); + ubifs_assert(req->new_ino <= 1); + ubifs_assert(req->new_ino_d <= UBIFS_MAX_INO_DATA); ubifs_assert(req->dirtied_ino <= 4); ubifs_assert(req->dirtied_ino_d <= UBIFS_MAX_INO_DATA * 4); if (!req->recalculate) { diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h index d342c6907244..565dca2ec0bd 100644 --- a/fs/ubifs/ubifs.h +++ b/fs/ubifs/ubifs.h @@ -812,17 +812,23 @@ struct ubifs_compressor { struct ubifs_budget_req { unsigned int fast:1; unsigned int recalculate:1; +#ifndef UBIFS_DEBUG unsigned int new_page:1; unsigned int dirtied_page:1; unsigned int new_dent:1; unsigned int mod_dent:1; unsigned int new_ino:1; unsigned int new_ino_d:13; -#ifndef UBIFS_DEBUG unsigned int dirtied_ino:4; unsigned int dirtied_ino_d:15; #else /* Not bit-fields to check for overflows */ + unsigned int new_page; + unsigned int dirtied_page; + unsigned int new_dent; + unsigned int mod_dent; + unsigned int new_ino; + unsigned int new_ino_d; unsigned int dirtied_ino; unsigned int dirtied_ino_d; #endif -- cgit v1.2.3 From dab4b4d2f915a65022343012a795f4ae4ae7e83c Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Thu, 24 Jul 2008 14:52:45 +0300 Subject: UBIFS: align inode data to eight UBIFS aligns node lengths to 8, so budgeting has to do the same. Well, direntry, inode, and page budgets are already aligned, but not inode data budget (e.g., data in special devices or symlinks). Do this for inode data as well. Also, add corresponding debugging checks. Signed-off-by: Artem Bityutskiy --- fs/ubifs/budget.c | 10 +++++++++- fs/ubifs/dir.c | 10 ++++++---- fs/ubifs/file.c | 4 ++-- fs/ubifs/ubifs.h | 4 ++++ fs/ubifs/xattr.c | 4 ++-- 5 files changed, 23 insertions(+), 9 deletions(-) diff --git a/fs/ubifs/budget.c b/fs/ubifs/budget.c index f5afce5f37bd..a3978ba4215e 100644 --- a/fs/ubifs/budget.c +++ b/fs/ubifs/budget.c @@ -551,6 +551,8 @@ int ubifs_budget_space(struct ubifs_info *c, struct ubifs_budget_req *req) ubifs_assert(req->new_ino_d <= UBIFS_MAX_INO_DATA); ubifs_assert(req->dirtied_ino <= 4); ubifs_assert(req->dirtied_ino_d <= UBIFS_MAX_INO_DATA * 4); + ubifs_assert(!(req->new_ino_d & 7)); + ubifs_assert(!(req->dirtied_ino_d & 7)); data_growth = calc_data_growth(c, req); dd_growth = calc_dd_growth(c, req); @@ -632,6 +634,8 @@ void ubifs_release_budget(struct ubifs_info *c, struct ubifs_budget_req *req) ubifs_assert(req->new_ino_d <= UBIFS_MAX_INO_DATA); ubifs_assert(req->dirtied_ino <= 4); ubifs_assert(req->dirtied_ino_d <= UBIFS_MAX_INO_DATA * 4); + ubifs_assert(!(req->new_ino_d & 7)); + ubifs_assert(!(req->dirtied_ino_d & 7)); if (!req->recalculate) { ubifs_assert(req->idx_growth >= 0); ubifs_assert(req->data_growth >= 0); @@ -659,7 +663,11 @@ void ubifs_release_budget(struct ubifs_info *c, struct ubifs_budget_req *req) ubifs_assert(c->budg_idx_growth >= 0); ubifs_assert(c->budg_data_growth >= 0); + ubifs_assert(c->budg_dd_growth >= 0); ubifs_assert(c->min_idx_lebs < c->main_lebs); + ubifs_assert(!(c->budg_idx_growth & 7)); + ubifs_assert(!(c->budg_data_growth & 7)); + ubifs_assert(!(c->budg_dd_growth & 7)); spin_unlock(&c->space_lock); } @@ -701,7 +709,7 @@ void ubifs_release_dirty_inode_budget(struct ubifs_info *c, struct ubifs_budget_req req; memset(&req, 0, sizeof(struct ubifs_budget_req)); - req.dd_growth = c->inode_budget + ui->data_len; + req.dd_growth = c->inode_budget + ALIGN(ui->data_len, 8); ubifs_release_budget(c, &req); } diff --git a/fs/ubifs/dir.c b/fs/ubifs/dir.c index a79e850fee6d..eba3a8a7c333 100644 --- a/fs/ubifs/dir.c +++ b/fs/ubifs/dir.c @@ -525,7 +525,7 @@ static int ubifs_link(struct dentry *old_dentry, struct inode *dir, struct ubifs_inode *dir_ui = ubifs_inode(dir); int err, sz_change = CALC_DENT_SIZE(dentry->d_name.len); struct ubifs_budget_req req = { .new_dent = 1, .dirtied_ino = 2, - .dirtied_ino_d = ui->data_len }; + .dirtied_ino_d = ALIGN(ui->data_len, 8) }; /* * Budget request settings: new direntry, changing the target inode, @@ -788,7 +788,8 @@ static int ubifs_mknod(struct inode *dir, struct dentry *dentry, int sz_change = CALC_DENT_SIZE(dentry->d_name.len); int err, devlen = 0; struct ubifs_budget_req req = { .new_ino = 1, .new_dent = 1, - .new_ino_d = devlen, .dirtied_ino = 1 }; + .new_ino_d = ALIGN(devlen, 8), + .dirtied_ino = 1 }; /* * Budget request settings: new inode, new direntry and changing parent @@ -862,7 +863,8 @@ static int ubifs_symlink(struct inode *dir, struct dentry *dentry, int err, len = strlen(symname); int sz_change = CALC_DENT_SIZE(dentry->d_name.len); struct ubifs_budget_req req = { .new_ino = 1, .new_dent = 1, - .new_ino_d = len, .dirtied_ino = 1 }; + .new_ino_d = ALIGN(len, 8), + .dirtied_ino = 1 }; /* * Budget request settings: new inode, new direntry and changing parent @@ -1011,7 +1013,7 @@ static int ubifs_rename(struct inode *old_dir, struct dentry *old_dentry, struct ubifs_budget_req req = { .new_dent = 1, .mod_dent = 1, .dirtied_ino = 3 }; struct ubifs_budget_req ino_req = { .dirtied_ino = 1, - .dirtied_ino_d = old_inode_ui->data_len }; + .dirtied_ino_d = ALIGN(old_inode_ui->data_len, 8) }; struct timespec time; /* diff --git a/fs/ubifs/file.c b/fs/ubifs/file.c index 01598f28020b..9fecab2f30bc 100644 --- a/fs/ubifs/file.c +++ b/fs/ubifs/file.c @@ -890,7 +890,7 @@ static int do_setattr(struct ubifs_info *c, struct inode *inode, loff_t new_size = attr->ia_size; struct ubifs_inode *ui = ubifs_inode(inode); struct ubifs_budget_req req = { .dirtied_ino = 1, - .dirtied_ino_d = ui->data_len }; + .dirtied_ino_d = ALIGN(ui->data_len, 8) }; err = ubifs_budget_space(c, &req); if (err) @@ -1052,7 +1052,7 @@ static int update_mctime(struct ubifs_info *c, struct inode *inode) if (mctime_update_needed(inode, &now)) { int err, release; struct ubifs_budget_req req = { .dirtied_ino = 1, - .dirtied_ino_d = ui->data_len }; + .dirtied_ino_d = ALIGN(ui->data_len, 8) }; err = ubifs_budget_space(c, &req); if (err) diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h index 565dca2ec0bd..73ca8a009798 100644 --- a/fs/ubifs/ubifs.h +++ b/fs/ubifs/ubifs.h @@ -808,6 +808,10 @@ struct ubifs_compressor { * An inode may contain 4KiB of data at max., thus the widths of @new_ino_d * is 13 bits, and @dirtied_ino_d - 15, because up to 4 inodes may be made * dirty by the re-name operation. + * + * Note, UBIFS aligns node lengths to 8-bytes boundary, so the requester has to + * make sure the amount of inode data which contribute to @new_ino_d and + * @dirtied_ino_d fields are aligned. */ struct ubifs_budget_req { unsigned int fast:1; diff --git a/fs/ubifs/xattr.c b/fs/ubifs/xattr.c index 1388a078e1a9..39e831d074ce 100644 --- a/fs/ubifs/xattr.c +++ b/fs/ubifs/xattr.c @@ -103,8 +103,8 @@ static int create_xattr(struct ubifs_info *c, struct inode *host, struct inode *inode; struct ubifs_inode *ui, *host_ui = ubifs_inode(host); struct ubifs_budget_req req = { .new_ino = 1, .new_dent = 1, - .new_ino_d = size, .dirtied_ino = 1, - .dirtied_ino_d = host_ui->data_len}; + .new_ino_d = size, .dirtied_ino = 1, + .dirtied_ino_d = ALIGN(host_ui->data_len, 8)}; if (host_ui->xattr_cnt >= MAX_XATTRS_PER_INODE) return -ENOSPC; -- cgit v1.2.3 From 1de9415906bccab51fb74c6adf575948610f0909 Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Fri, 25 Jul 2008 12:58:38 +0300 Subject: UBIFS: print pid in dump function Useful when something fails and there are many processes racing. Signed-off-by: Artem Bityutskiy --- fs/ubifs/debug.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/fs/ubifs/debug.c b/fs/ubifs/debug.c index 0adfb29b8503..b9cb77473758 100644 --- a/fs/ubifs/debug.c +++ b/fs/ubifs/debug.c @@ -568,8 +568,8 @@ void dbg_dump_budget_req(const struct ubifs_budget_req *req) void dbg_dump_lstats(const struct ubifs_lp_stats *lst) { spin_lock(&dbg_lock); - printk(KERN_DEBUG "Lprops statistics: empty_lebs %d, idx_lebs %d\n", - lst->empty_lebs, lst->idx_lebs); + printk(KERN_DEBUG "(pid %d) Lprops statistics: empty_lebs %d, " + "idx_lebs %d\n", current->pid, lst->empty_lebs, lst->idx_lebs); printk(KERN_DEBUG "\ttaken_empty_lebs %d, total_free %lld, " "total_dirty %lld\n", lst->taken_empty_lebs, lst->total_free, lst->total_dirty); @@ -587,8 +587,8 @@ void dbg_dump_budg(struct ubifs_info *c) struct ubifs_gced_idx_leb *idx_gc; spin_lock(&dbg_lock); - printk(KERN_DEBUG "Budgeting info: budg_data_growth %lld, " - "budg_dd_growth %lld, budg_idx_growth %lld\n", + printk(KERN_DEBUG "(pid %d) Budgeting info: budg_data_growth %lld, " + "budg_dd_growth %lld, budg_idx_growth %lld\n", current->pid, c->budg_data_growth, c->budg_dd_growth, c->budg_idx_growth); printk(KERN_DEBUG "\tdata budget sum %lld, total budget sum %lld, " "freeable_cnt %d\n", c->budg_data_growth + c->budg_dd_growth, @@ -634,7 +634,7 @@ void dbg_dump_lprops(struct ubifs_info *c) struct ubifs_lprops lp; struct ubifs_lp_stats lst; - printk(KERN_DEBUG "Dumping LEB properties\n"); + printk(KERN_DEBUG "(pid %d) Dumping LEB properties\n", current->pid); ubifs_get_lp_stats(c, &lst); dbg_dump_lstats(&lst); @@ -655,7 +655,7 @@ void dbg_dump_leb(const struct ubifs_info *c, int lnum) if (dbg_failure_mode) return; - printk(KERN_DEBUG "Dumping LEB %d\n", lnum); + printk(KERN_DEBUG "(pid %d) Dumping LEB %d\n", current->pid, lnum); sleb = ubifs_scan(c, lnum, 0, c->dbg_buf); if (IS_ERR(sleb)) { @@ -720,8 +720,8 @@ void dbg_dump_heap(struct ubifs_info *c, struct ubifs_lpt_heap *heap, int cat) { int i; - printk(KERN_DEBUG "Dumping heap cat %d (%d elements)\n", - cat, heap->cnt); + printk(KERN_DEBUG "(pid %d) Dumping heap cat %d (%d elements)\n", + current->pid, cat, heap->cnt); for (i = 0; i < heap->cnt; i++) { struct ubifs_lprops *lprops = heap->arr[i]; @@ -736,7 +736,7 @@ void dbg_dump_pnode(struct ubifs_info *c, struct ubifs_pnode *pnode, { int i; - printk(KERN_DEBUG "Dumping pnode:\n"); + printk(KERN_DEBUG "(pid %d) Dumping pnode:\n", current->pid); printk(KERN_DEBUG "\taddress %zx parent %zx cnext %zx\n", (size_t)pnode, (size_t)parent, (size_t)pnode->cnext); printk(KERN_DEBUG "\tflags %lu iip %d level %d num %d\n", @@ -755,7 +755,7 @@ void dbg_dump_tnc(struct ubifs_info *c) int level; printk(KERN_DEBUG "\n"); - printk(KERN_DEBUG "Dumping the TNC tree\n"); + printk(KERN_DEBUG "(pid %d) Dumping the TNC tree\n", current->pid); znode = ubifs_tnc_levelorder_next(c->zroot.znode, NULL); level = znode->level; printk(KERN_DEBUG "== Level %d ==\n", level); -- cgit v1.2.3 From b364b41aeb0289be402be83eebca92eb90bfcb8b Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Fri, 25 Jul 2008 14:38:51 +0300 Subject: UBIFS: reserve more space for index At the moment UBIFS reserves twice old index size space for the index. But this is not enough in some cases, because if the indexing node are very fragmented and there are many small gaps, while the dirty index has big znodes - in-the-gaps method would fail. Thus, reserve trise as more, in which case we are guaranteed that we can commit in any case. Signed-off-by: Artem Bityutskiy --- fs/ubifs/budget.c | 8 ++++---- fs/ubifs/find.c | 9 +++++++-- fs/ubifs/misc.h | 2 +- fs/ubifs/ubifs-media.h | 4 ++-- 4 files changed, 14 insertions(+), 9 deletions(-) diff --git a/fs/ubifs/budget.c b/fs/ubifs/budget.c index a3978ba4215e..323d83a4d099 100644 --- a/fs/ubifs/budget.c +++ b/fs/ubifs/budget.c @@ -263,8 +263,8 @@ int ubifs_calc_min_idx_lebs(struct ubifs_info *c) idx_size = c->old_idx_sz + c->budg_idx_growth + c->budg_uncommitted_idx; - /* And make sure we have twice the index size of space reserved */ - idx_size <<= 1; + /* And make sure we have trice the index size of space reserved */ + idx_size = idx_size + (idx_size << 1); /* * We do not maintain 'old_idx_size' as 'old_idx_lebs'/'old_idx_bytes' @@ -388,11 +388,11 @@ static int can_use_rp(struct ubifs_info *c) * This function makes sure UBIFS has enough free eraseblocks for index growth * and data. * - * When budgeting index space, UBIFS reserves twice as more LEBs as the index + * When budgeting index space, UBIFS reserves trice as more LEBs as the index * would take if it was consolidated and written to the flash. This guarantees * that the "in-the-gaps" commit method always succeeds and UBIFS will always * be able to commit dirty index. So this function basically adds amount of - * budgeted index space to the size of the current index, multiplies this by 2, + * budgeted index space to the size of the current index, multiplies this by 3, * and makes sure this does not exceed the amount of free eraseblocks. * * Notes about @c->min_idx_lebs and @c->lst.idx_lebs variables: diff --git a/fs/ubifs/find.c b/fs/ubifs/find.c index 10394c548367..c70c7679c1bf 100644 --- a/fs/ubifs/find.c +++ b/fs/ubifs/find.c @@ -290,9 +290,14 @@ int ubifs_find_dirty_leb(struct ubifs_info *c, struct ubifs_lprops *ret_lp, idx_lp = idx_heap->arr[0]; sum = idx_lp->free + idx_lp->dirty; /* - * Since we reserve twice as more space for the index than it + * Since we reserve trice as more space for the index than it * actually takes, it does not make sense to pick indexing LEBs - * with less than half LEB of dirty space. + * with less than, say, half LEB of dirty space. May be half is + * not the optimal boundary - this should be tested and + * checked. This boundary should determine how much we use + * in-the-gaps to consolidate the index comparing to how much + * we use garbage collector to consolidate it. The "half" + * criteria just feels to be fine. */ if (sum < min_space || sum < c->half_leb_size) idx_lp = NULL; diff --git a/fs/ubifs/misc.h b/fs/ubifs/misc.h index cd83ffc8101c..87dabf9fe742 100644 --- a/fs/ubifs/misc.h +++ b/fs/ubifs/misc.h @@ -308,7 +308,7 @@ static inline long long ubifs_reported_space(const struct ubifs_info *c, { int divisor, factor; - divisor = UBIFS_MAX_DATA_NODE_SZ + (c->max_idx_node_sz << 1); + divisor = UBIFS_MAX_DATA_NODE_SZ + (c->max_idx_node_sz * 3); factor = UBIFS_MAX_DATA_NODE_SZ - UBIFS_DATA_NODE_SZ; do_div(free, divisor); diff --git a/fs/ubifs/ubifs-media.h b/fs/ubifs/ubifs-media.h index 0cc7da9bed47..bd2121f3426e 100644 --- a/fs/ubifs/ubifs-media.h +++ b/fs/ubifs/ubifs-media.h @@ -228,10 +228,10 @@ enum { /* Minimum number of orphan area logical eraseblocks */ #define UBIFS_MIN_ORPH_LEBS 1 /* - * Minimum number of main area logical eraseblocks (buds, 2 for the index, 1 + * Minimum number of main area logical eraseblocks (buds, 3 for the index, 1 * for GC, 1 for deletions, and at least 1 for committed data). */ -#define UBIFS_MIN_MAIN_LEBS (UBIFS_MIN_BUD_LEBS + 5) +#define UBIFS_MIN_MAIN_LEBS (UBIFS_MIN_BUD_LEBS + 6) /* Minimum number of logical eraseblocks */ #define UBIFS_MIN_LEB_CNT (UBIFS_SB_LEBS + UBIFS_MST_LEBS + \ -- cgit v1.2.3 From 0010f18afc5f8ba25e1d20e3165894c32a65af02 Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Fri, 25 Jul 2008 16:39:44 +0300 Subject: UBIFS: minor tweaks in commit No functional changes, just lessen the amount of indentations. Signed-off-by: Artem Bityutskiy --- fs/ubifs/tnc_commit.c | 37 ++++++++++++++++++------------------- 1 file changed, 18 insertions(+), 19 deletions(-) diff --git a/fs/ubifs/tnc_commit.c b/fs/ubifs/tnc_commit.c index 8117e65ba2e9..8ac76b1c2d55 100644 --- a/fs/ubifs/tnc_commit.c +++ b/fs/ubifs/tnc_commit.c @@ -372,26 +372,25 @@ static int layout_in_gaps(struct ubifs_info *c, int cnt) written = layout_leb_in_gaps(c, p); if (written < 0) { err = written; - if (err == -ENOSPC) { - if (!dbg_force_in_the_gaps_enabled) { - /* - * Do not print scary warnings if the - * debugging option which forces - * in-the-gaps is enabled. - */ - ubifs_err("out of space"); - spin_lock(&c->space_lock); - dbg_dump_budg(c); - spin_unlock(&c->space_lock); - dbg_dump_lprops(c); - } - /* Try to commit anyway */ - err = 0; - break; + if (err != -ENOSPC) { + kfree(c->gap_lebs); + c->gap_lebs = NULL; + return err; } - kfree(c->gap_lebs); - c->gap_lebs = NULL; - return err; + if (!dbg_force_in_the_gaps_enabled) { + /* + * Do not print scary warnings if the debugging + * option which forces in-the-gaps is enabled. + */ + ubifs_err("out of space"); + spin_lock(&c->space_lock); + dbg_dump_budg(c); + spin_unlock(&c->space_lock); + dbg_dump_lprops(c); + } + /* Try to commit anyway */ + err = 0; + break; } p++; cnt -= written; -- cgit v1.2.3 From 22bc7fa8c5da09805edc6a6199ce81373b2c207d Mon Sep 17 00:00:00 2001 From: Zoltan Sogor Date: Mon, 28 Jul 2008 16:28:49 +0200 Subject: UBIFS: support splice_write Signed-off-by: Zoltan Sogor Signed-off-by: Artem Bityutskiy --- fs/ubifs/file.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/ubifs/file.c b/fs/ubifs/file.c index 9fecab2f30bc..4071d1cae29f 100644 --- a/fs/ubifs/file.c +++ b/fs/ubifs/file.c @@ -1271,6 +1271,7 @@ struct file_operations ubifs_file_operations = { .fsync = ubifs_fsync, .unlocked_ioctl = ubifs_ioctl, .splice_read = generic_file_splice_read, + .splice_write = generic_file_splice_write, #ifdef CONFIG_COMPAT .compat_ioctl = ubifs_compat_ioctl, #endif -- cgit v1.2.3 From 3a13252c6f3a029ac992a36910e945f361482797 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Wed, 30 Jul 2008 12:18:02 +0300 Subject: UBIFS: correct spelling of "thrice". Signed-off-by: Adrian Hunter --- fs/ubifs/budget.c | 4 ++-- fs/ubifs/find.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/fs/ubifs/budget.c b/fs/ubifs/budget.c index 323d83a4d099..154098157473 100644 --- a/fs/ubifs/budget.c +++ b/fs/ubifs/budget.c @@ -263,7 +263,7 @@ int ubifs_calc_min_idx_lebs(struct ubifs_info *c) idx_size = c->old_idx_sz + c->budg_idx_growth + c->budg_uncommitted_idx; - /* And make sure we have trice the index size of space reserved */ + /* And make sure we have thrice the index size of space reserved */ idx_size = idx_size + (idx_size << 1); /* @@ -388,7 +388,7 @@ static int can_use_rp(struct ubifs_info *c) * This function makes sure UBIFS has enough free eraseblocks for index growth * and data. * - * When budgeting index space, UBIFS reserves trice as more LEBs as the index + * When budgeting index space, UBIFS reserves thrice as many LEBs as the index * would take if it was consolidated and written to the flash. This guarantees * that the "in-the-gaps" commit method always succeeds and UBIFS will always * be able to commit dirty index. So this function basically adds amount of diff --git a/fs/ubifs/find.c b/fs/ubifs/find.c index c70c7679c1bf..adee7b5ddeab 100644 --- a/fs/ubifs/find.c +++ b/fs/ubifs/find.c @@ -290,7 +290,7 @@ int ubifs_find_dirty_leb(struct ubifs_info *c, struct ubifs_lprops *ret_lp, idx_lp = idx_heap->arr[0]; sum = idx_lp->free + idx_lp->dirty; /* - * Since we reserve trice as more space for the index than it + * Since we reserve thrice as much space for the index than it * actually takes, it does not make sense to pick indexing LEBs * with less than, say, half LEB of dirty space. May be half is * not the optimal boundary - this should be tested and -- cgit v1.2.3 From 81ffa38e1558f54db190e2d11e7260ab09c4acf2 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Fri, 1 Aug 2008 15:35:08 +0300 Subject: UBIFS: always set i_generation to 0 UBIFS does not presently re-use inode numbers, so leaving i_generation zero is most appropriate for now. Signed-off-by: Adrian Hunter Signed-off-by: Artem Bityutskiy --- fs/ubifs/dir.c | 1 - fs/ubifs/super.c | 2 -- fs/ubifs/ubifs.h | 4 +--- 3 files changed, 1 insertion(+), 6 deletions(-) diff --git a/fs/ubifs/dir.c b/fs/ubifs/dir.c index eba3a8a7c333..0d1ab8967a4c 100644 --- a/fs/ubifs/dir.c +++ b/fs/ubifs/dir.c @@ -165,7 +165,6 @@ struct inode *ubifs_new_inode(struct ubifs_info *c, const struct inode *dir, } inode->i_ino = ++c->highest_inum; - inode->i_generation = ++c->vfs_gen; /* * The creation sequence number remains with this inode for its * lifetime. All nodes for this inode have a greater sequence number, diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c index 6cc4175f23c1..2c268a476413 100644 --- a/fs/ubifs/super.c +++ b/fs/ubifs/super.c @@ -30,7 +30,6 @@ #include #include #include -#include #include #include #include @@ -1671,7 +1670,6 @@ static int ubifs_fill_super(struct super_block *sb, void *data, int silent) INIT_LIST_HEAD(&c->orph_new); c->highest_inum = UBIFS_FIRST_INO; - get_random_bytes(&c->vfs_gen, sizeof(int)); c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM; ubi_get_volume_info(ubi, &c->vi); diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h index 73ca8a009798..f2dd749d7989 100644 --- a/fs/ubifs/ubifs.h +++ b/fs/ubifs/ubifs.h @@ -876,11 +876,10 @@ struct ubifs_mount_opts { * @bdi: backing device info object to make VFS happy and disable read-ahead * * @highest_inum: highest used inode number - * @vfs_gen: VFS inode generation counter * @max_sqnum: current global sequence number * @cmt_no: commit number of the last successfully completed commit, protected * by @commit_sem - * @cnt_lock: protects @highest_inum, @vfs_gen, and @max_sqnum counters + * @cnt_lock: protects @highest_inum and @max_sqnum counters * @fmt_version: UBIFS on-flash format version * @uuid: UUID from super block * @@ -1117,7 +1116,6 @@ struct ubifs_info { struct backing_dev_info bdi; ino_t highest_inum; - unsigned int vfs_gen; unsigned long long max_sqnum; unsigned long long cmt_no; spinlock_t cnt_lock; -- cgit v1.2.3 From 840dc6b891d521f18bf081bd5a32e4a1f8110abc Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Fri, 1 Aug 2008 18:13:37 +0300 Subject: UBIFS: improve arguments checking in debugging messages Use "if (0) printk()" construct in debugging print macros to make the debugging messages be checked even if debugging is off. This patch also removes some unneeded spaces and blank lines. Signed-off-by: Artem Bityutskiy --- fs/ubifs/debug.h | 143 +++++++++++++++++++++++++++---------------------------- 1 file changed, 69 insertions(+), 74 deletions(-) diff --git a/fs/ubifs/debug.h b/fs/ubifs/debug.h index 3c4f1e93c9e0..50315fc57185 100644 --- a/fs/ubifs/debug.h +++ b/fs/ubifs/debug.h @@ -27,7 +27,7 @@ #define UBIFS_DBG(op) op -#define ubifs_assert(expr) do { \ +#define ubifs_assert(expr) do { \ if (unlikely(!(expr))) { \ printk(KERN_CRIT "UBIFS assert failed in %s at %u (pid %d)\n", \ __func__, __LINE__, current->pid); \ @@ -73,50 +73,50 @@ const char *dbg_key_str1(const struct ubifs_info *c, const union ubifs_key *key); /* - * DBGKEY macros require dbg_lock to be held, which it is in the dbg message + * DBGKEY macros require @dbg_lock to be held, which it is in the dbg message * macros. */ #define DBGKEY(key) dbg_key_str0(c, (key)) #define DBGKEY1(key) dbg_key_str1(c, (key)) /* General messages */ -#define dbg_gen(fmt, ...) dbg_do_msg(UBIFS_MSG_GEN, fmt, ##__VA_ARGS__) +#define dbg_gen(fmt, ...) dbg_do_msg(UBIFS_MSG_GEN, fmt, ##__VA_ARGS__) /* Additional journal messages */ -#define dbg_jnl(fmt, ...) dbg_do_msg(UBIFS_MSG_JNL, fmt, ##__VA_ARGS__) +#define dbg_jnl(fmt, ...) dbg_do_msg(UBIFS_MSG_JNL, fmt, ##__VA_ARGS__) /* Additional TNC messages */ -#define dbg_tnc(fmt, ...) dbg_do_msg(UBIFS_MSG_TNC, fmt, ##__VA_ARGS__) +#define dbg_tnc(fmt, ...) dbg_do_msg(UBIFS_MSG_TNC, fmt, ##__VA_ARGS__) /* Additional lprops messages */ -#define dbg_lp(fmt, ...) dbg_do_msg(UBIFS_MSG_LP, fmt, ##__VA_ARGS__) +#define dbg_lp(fmt, ...) dbg_do_msg(UBIFS_MSG_LP, fmt, ##__VA_ARGS__) /* Additional LEB find messages */ -#define dbg_find(fmt, ...) dbg_do_msg(UBIFS_MSG_FIND, fmt, ##__VA_ARGS__) +#define dbg_find(fmt, ...) dbg_do_msg(UBIFS_MSG_FIND, fmt, ##__VA_ARGS__) /* Additional mount messages */ -#define dbg_mnt(fmt, ...) dbg_do_msg(UBIFS_MSG_MNT, fmt, ##__VA_ARGS__) +#define dbg_mnt(fmt, ...) dbg_do_msg(UBIFS_MSG_MNT, fmt, ##__VA_ARGS__) /* Additional I/O messages */ -#define dbg_io(fmt, ...) dbg_do_msg(UBIFS_MSG_IO, fmt, ##__VA_ARGS__) +#define dbg_io(fmt, ...) dbg_do_msg(UBIFS_MSG_IO, fmt, ##__VA_ARGS__) /* Additional commit messages */ -#define dbg_cmt(fmt, ...) dbg_do_msg(UBIFS_MSG_CMT, fmt, ##__VA_ARGS__) +#define dbg_cmt(fmt, ...) dbg_do_msg(UBIFS_MSG_CMT, fmt, ##__VA_ARGS__) /* Additional budgeting messages */ -#define dbg_budg(fmt, ...) dbg_do_msg(UBIFS_MSG_BUDG, fmt, ##__VA_ARGS__) +#define dbg_budg(fmt, ...) dbg_do_msg(UBIFS_MSG_BUDG, fmt, ##__VA_ARGS__) /* Additional log messages */ -#define dbg_log(fmt, ...) dbg_do_msg(UBIFS_MSG_LOG, fmt, ##__VA_ARGS__) +#define dbg_log(fmt, ...) dbg_do_msg(UBIFS_MSG_LOG, fmt, ##__VA_ARGS__) /* Additional gc messages */ -#define dbg_gc(fmt, ...) dbg_do_msg(UBIFS_MSG_GC, fmt, ##__VA_ARGS__) +#define dbg_gc(fmt, ...) dbg_do_msg(UBIFS_MSG_GC, fmt, ##__VA_ARGS__) /* Additional scan messages */ -#define dbg_scan(fmt, ...) dbg_do_msg(UBIFS_MSG_SCAN, fmt, ##__VA_ARGS__) +#define dbg_scan(fmt, ...) dbg_do_msg(UBIFS_MSG_SCAN, fmt, ##__VA_ARGS__) /* Additional recovery messages */ -#define dbg_rcvry(fmt, ...) dbg_do_msg(UBIFS_MSG_RCVRY, fmt, ##__VA_ARGS__) +#define dbg_rcvry(fmt, ...) dbg_do_msg(UBIFS_MSG_RCVRY, fmt, ##__VA_ARGS__) /* * Debugging message type flags (must match msg_type_names in debug.c). @@ -239,34 +239,23 @@ typedef int (*dbg_leaf_callback)(struct ubifs_info *c, struct ubifs_zbranch *zbr, void *priv); typedef int (*dbg_znode_callback)(struct ubifs_info *c, struct ubifs_znode *znode, void *priv); - int dbg_walk_index(struct ubifs_info *c, dbg_leaf_callback leaf_cb, dbg_znode_callback znode_cb, void *priv); /* Checking functions */ int dbg_check_lprops(struct ubifs_info *c); - int dbg_old_index_check_init(struct ubifs_info *c, struct ubifs_zbranch *zroot); int dbg_check_old_index(struct ubifs_info *c, struct ubifs_zbranch *zroot); - int dbg_check_cats(struct ubifs_info *c); - int dbg_check_ltab(struct ubifs_info *c); - int dbg_check_synced_i_size(struct inode *inode); - int dbg_check_dir_size(struct ubifs_info *c, const struct inode *dir); - int dbg_check_tnc(struct ubifs_info *c, int extra); - int dbg_check_idx_size(struct ubifs_info *c, long long idx_size); - int dbg_check_filesystem(struct ubifs_info *c); - void dbg_check_heap(struct ubifs_info *c, struct ubifs_lpt_heap *heap, int cat, int add_pos); - int dbg_check_lprops(struct ubifs_info *c); int dbg_check_lpt_nodes(struct ubifs_info *c, struct ubifs_cnode *cnode, int row, int col); @@ -329,71 +318,77 @@ static inline int dbg_change(struct ubi_volume_desc *desc, int lnum, #else /* !CONFIG_UBIFS_FS_DEBUG */ #define UBIFS_DBG(op) -#define ubifs_assert(expr) ({}) -#define ubifs_assert_cmt_locked(c) + +/* Use "if (0)" to make compiler check arguments even if debugging is off */ +#define ubifs_assert(expr) do { \ + if (0 && (expr)) \ + printk(KERN_CRIT "UBIFS assert failed in %s at %u (pid %d)\n", \ + __func__, __LINE__, current->pid); \ +} while (0) + +#define dbg_err(fmt, ...) do { \ + if (0) \ + ubifs_err(fmt, ##__VA_ARGS__); \ +} while (0) + +#define dbg_msg(fmt, ...) do { \ + if (0) \ + printk(KERN_DEBUG "UBIFS DBG (pid %d): %s: " fmt "\n", \ + current->pid, __func__, ##__VA_ARGS__); \ +} while (0) + #define dbg_dump_stack() -#define dbg_err(fmt, ...) ({}) -#define dbg_msg(fmt, ...) ({}) -#define dbg_key(c, key, fmt, ...) ({}) - -#define dbg_gen(fmt, ...) ({}) -#define dbg_jnl(fmt, ...) ({}) -#define dbg_tnc(fmt, ...) ({}) -#define dbg_lp(fmt, ...) ({}) -#define dbg_find(fmt, ...) ({}) -#define dbg_mnt(fmt, ...) ({}) -#define dbg_io(fmt, ...) ({}) -#define dbg_cmt(fmt, ...) ({}) -#define dbg_budg(fmt, ...) ({}) -#define dbg_log(fmt, ...) ({}) -#define dbg_gc(fmt, ...) ({}) -#define dbg_scan(fmt, ...) ({}) -#define dbg_rcvry(fmt, ...) ({}) - -#define dbg_ntype(type) "" -#define dbg_cstate(cmt_state) "" -#define dbg_get_key_dump(c, key) ({}) -#define dbg_dump_inode(c, inode) ({}) -#define dbg_dump_node(c, node) ({}) -#define dbg_dump_budget_req(req) ({}) -#define dbg_dump_lstats(lst) ({}) -#define dbg_dump_budg(c) ({}) -#define dbg_dump_lprop(c, lp) ({}) -#define dbg_dump_lprops(c) ({}) -#define dbg_dump_leb(c, lnum) ({}) -#define dbg_dump_znode(c, znode) ({}) -#define dbg_dump_heap(c, heap, cat) ({}) -#define dbg_dump_pnode(c, pnode, parent, iip) ({}) -#define dbg_dump_tnc(c) ({}) -#define dbg_dump_index(c) ({}) +#define ubifs_assert_cmt_locked(c) -#define dbg_walk_index(c, leaf_cb, znode_cb, priv) 0 +#define dbg_gen(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_jnl(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_tnc(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_lp(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_find(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_mnt(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_io(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_cmt(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_budg(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_log(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_gc(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_scan(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) +#define dbg_rcvry(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__) + +#define DBGKEY(key) ((char *)(key)) +#define DBGKEY1(key) ((char *)(key)) + +#define dbg_ntype(type) "" +#define dbg_cstate(cmt_state) "" +#define dbg_get_key_dump(c, key) ({}) +#define dbg_dump_inode(c, inode) ({}) +#define dbg_dump_node(c, node) ({}) +#define dbg_dump_budget_req(req) ({}) +#define dbg_dump_lstats(lst) ({}) +#define dbg_dump_budg(c) ({}) +#define dbg_dump_lprop(c, lp) ({}) +#define dbg_dump_lprops(c) ({}) +#define dbg_dump_leb(c, lnum) ({}) +#define dbg_dump_znode(c, znode) ({}) +#define dbg_dump_heap(c, heap, cat) ({}) +#define dbg_dump_pnode(c, pnode, parent, iip) ({}) +#define dbg_dump_tnc(c) ({}) +#define dbg_dump_index(c) ({}) +#define dbg_walk_index(c, leaf_cb, znode_cb, priv) 0 #define dbg_old_index_check_init(c, zroot) 0 #define dbg_check_old_index(c, zroot) 0 - #define dbg_check_cats(c) 0 - #define dbg_check_ltab(c) 0 - #define dbg_check_synced_i_size(inode) 0 - #define dbg_check_dir_size(c, dir) 0 - #define dbg_check_tnc(c, x) 0 - #define dbg_check_idx_size(c, idx_size) 0 - #define dbg_check_filesystem(c) 0 - #define dbg_check_heap(c, heap, cat, add_pos) ({}) - #define dbg_check_lprops(c) 0 #define dbg_check_lpt_nodes(c, cnode, row, col) 0 - #define dbg_force_in_the_gaps_enabled 0 #define dbg_force_in_the_gaps() 0 - #define dbg_failure_mode 0 #define dbg_failure_mode_registration(c) ({}) #define dbg_failure_mode_deregistration(c) ({}) -- cgit v1.2.3 From 5acd6ff8ac09eb71f3aef2ccccefab658be8aff4 Mon Sep 17 00:00:00 2001 From: Zoltan Sogor Date: Tue, 12 Aug 2008 13:54:54 +0300 Subject: UBIFS: fix budgeting request alignment in xattr code Data length has to be aligned in the budgeting request. Code in xattr.c did not do this. Signed-off-by: Zoltan Sogor Signed-off-by: Artem Bityutskiy --- fs/ubifs/xattr.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/fs/ubifs/xattr.c b/fs/ubifs/xattr.c index 39e831d074ce..6f493dea561e 100644 --- a/fs/ubifs/xattr.c +++ b/fs/ubifs/xattr.c @@ -103,8 +103,8 @@ static int create_xattr(struct ubifs_info *c, struct inode *host, struct inode *inode; struct ubifs_inode *ui, *host_ui = ubifs_inode(host); struct ubifs_budget_req req = { .new_ino = 1, .new_dent = 1, - .new_ino_d = size, .dirtied_ino = 1, - .dirtied_ino_d = ALIGN(host_ui->data_len, 8)}; + .new_ino_d = ALIGN(size, 8), .dirtied_ino = 1, + .dirtied_ino_d = ALIGN(host_ui->data_len, 8) }; if (host_ui->xattr_cnt >= MAX_XATTRS_PER_INODE) return -ENOSPC; @@ -200,7 +200,7 @@ static int change_xattr(struct ubifs_info *c, struct inode *host, struct ubifs_inode *host_ui = ubifs_inode(host); struct ubifs_inode *ui = ubifs_inode(inode); struct ubifs_budget_req req = { .dirtied_ino = 2, - .dirtied_ino_d = size + host_ui->data_len }; + .dirtied_ino_d = ALIGN(size, 8) + ALIGN(host_ui->data_len, 8) }; ubifs_assert(ui->data_len == inode->i_size); err = ubifs_budget_space(c, &req); @@ -497,8 +497,8 @@ static int remove_xattr(struct ubifs_info *c, struct inode *host, int err; struct ubifs_inode *host_ui = ubifs_inode(host); struct ubifs_inode *ui = ubifs_inode(inode); - struct ubifs_budget_req req = { .dirtied_ino = 1, .mod_dent = 1, - .dirtied_ino_d = host_ui->data_len }; + struct ubifs_budget_req req = { .dirtied_ino = 2, .mod_dent = 1, + .dirtied_ino_d = ALIGN(host_ui->data_len, 8) }; ubifs_assert(ui->data_len == inode->i_size); -- cgit v1.2.3 From 0a883a05c54b326bcf99c0902af28dae0386be0a Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Wed, 13 Aug 2008 14:13:26 +0300 Subject: UBIFS: few commentary fixes Signed-off-by: Artem Bityutskiy --- fs/ubifs/super.c | 8 ++++---- fs/ubifs/ubifs.h | 2 -- 2 files changed, 4 insertions(+), 6 deletions(-) diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c index 2c268a476413..f71e6b8822c4 100644 --- a/fs/ubifs/super.c +++ b/fs/ubifs/super.c @@ -148,7 +148,7 @@ struct inode *ubifs_iget(struct super_block *sb, unsigned long inum) if (err) goto out_invalid; - /* Disable readahead */ + /* Disable read-ahead */ inode->i_mapping->backing_dev_info = &c->bdi; switch (inode->i_mode & S_IFMT) { @@ -344,7 +344,7 @@ static void ubifs_delete_inode(struct inode *inode) if (err) /* * Worst case we have a lost orphan inode wasting space, so a - * simple error message is ok here. + * simple error message is OK here. */ ubifs_err("can't delete inode %lu, error %d", inode->i_ino, err); @@ -1683,10 +1683,10 @@ static int ubifs_fill_super(struct super_block *sb, void *data, int silent) } /* - * UBIFS provids 'backing_dev_info' in order to disable readahead. For + * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For * UBIFS, I/O is not deferred, it is done immediately in readpage, * which means the user would have to wait not just for their own I/O - * but the readahead I/O as well i.e. completely pointless. + * but the read-ahead I/O as well i.e. completely pointless. * * Read-ahead will be disabled because @c->bdi.ra_pages is 0. */ diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h index f2dd749d7989..d7f706f7a302 100644 --- a/fs/ubifs/ubifs.h +++ b/fs/ubifs/ubifs.h @@ -20,8 +20,6 @@ * Adrian Hunter */ -/* Implementation version 0.7 */ - #ifndef __UBIFS_H__ #define __UBIFS_H__ -- cgit v1.2.3 From 720b499c806200d06f4f22c668d46db784117089 Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Wed, 13 Aug 2008 16:16:31 +0300 Subject: UBIFS: remove unneeded check Commit d70b67c8bc72ee23b55381bd6a884f4796692f77 fixed VFS and it never calls FS lookup function in deleted directories now. We may remove corresponding UBIFS check. Signed-off-by: Artem Bityutskiy --- fs/ubifs/dir.c | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-) diff --git a/fs/ubifs/dir.c b/fs/ubifs/dir.c index 0d1ab8967a4c..5c96f1fb7016 100644 --- a/fs/ubifs/dir.c +++ b/fs/ubifs/dir.c @@ -219,15 +219,7 @@ static struct dentry *ubifs_lookup(struct inode *dir, struct dentry *dentry, err = ubifs_tnc_lookup_nm(c, &key, dent, &dentry->d_name); if (err) { - /* - * Do not hash the direntry if parent 'i_nlink' is zero, because - * this has side-effects - '->delete_inode()' call will not be - * called for the parent orphan inode, because 'd_count' of its - * direntry will stay 1 (it'll be negative direntry I guess) - * and prevent 'iput_final()' until the dentry is destroyed due - * to unmount or memory pressure. - */ - if (err == -ENOENT && dir->i_nlink != 0) { + if (err == -ENOENT) { dbg_gen("not found"); goto done; } -- cgit v1.2.3 From c78c7e35a4709b55d3126624662c8f6d7e3d1a5e Mon Sep 17 00:00:00 2001 From: Artem Bityutskiy Date: Tue, 12 Aug 2008 16:30:12 +0300 Subject: UBIFS: xattr bugfixes Xattr code has not been tested for a while and there were serveral bugs. One of them is using wrong inode in 'ubifs_jnl_change_xattr()'. The other is a deadlock in 'ubifs_setxattr()': the i_mutex is locked in 'cap_inode_need_killpriv()' path, so deadlock happens when 'ubifs_setxattr()' tries to lock it again. Thanks to Zoltan Sogor for finding these bugs. Signed-off-by: Artem Bityutskiy --- fs/ubifs/journal.c | 2 +- fs/ubifs/xattr.c | 44 +++++++++++++++++--------------------------- 2 files changed, 18 insertions(+), 28 deletions(-) diff --git a/fs/ubifs/journal.c b/fs/ubifs/journal.c index acdae00aaa54..22993f867d19 100644 --- a/fs/ubifs/journal.c +++ b/fs/ubifs/journal.c @@ -1374,7 +1374,7 @@ int ubifs_jnl_change_xattr(struct ubifs_info *c, const struct inode *inode, const struct inode *host) { int err, len1, len2, aligned_len, aligned_len1, lnum, offs; - struct ubifs_inode *host_ui = ubifs_inode(inode); + struct ubifs_inode *host_ui = ubifs_inode(host); struct ubifs_ino_node *ino; union ubifs_key key; int sync = IS_DIRSYNC(host); diff --git a/fs/ubifs/xattr.c b/fs/ubifs/xattr.c index 6f493dea561e..649bec78b645 100644 --- a/fs/ubifs/xattr.c +++ b/fs/ubifs/xattr.c @@ -61,7 +61,7 @@ /* * Limit the number of extended attributes per inode so that the total size - * (xattr_size) is guaranteeded to fit in an 'unsigned int'. + * (@xattr_size) is guaranteeded to fit in an 'unsigned int'. */ #define MAX_XATTRS_PER_INODE 65535 @@ -110,7 +110,7 @@ static int create_xattr(struct ubifs_info *c, struct inode *host, return -ENOSPC; /* * Linux limits the maximum size of the extended attribute names list - * to %XATTR_LIST_MAX. This means we should not allow creating more* + * to %XATTR_LIST_MAX. This means we should not allow creating more * extended attributes if the name list becomes larger. This limitation * is artificial for UBIFS, though. */ @@ -128,7 +128,6 @@ static int create_xattr(struct ubifs_info *c, struct inode *host, goto out_budg; } - mutex_lock(&host_ui->ui_mutex); /* Re-define all operations to be "nothing" */ inode->i_mapping->a_ops = &none_address_operations; inode->i_op = &none_inode_operations; @@ -141,23 +140,19 @@ static int create_xattr(struct ubifs_info *c, struct inode *host, ui->data = kmalloc(size, GFP_NOFS); if (!ui->data) { err = -ENOMEM; - goto out_unlock; + goto out_free; } - memcpy(ui->data, value, size); + inode->i_size = ui->ui_size = size; + ui->data_len = size; + + mutex_lock(&host_ui->ui_mutex); host->i_ctime = ubifs_current_time(host); host_ui->xattr_cnt += 1; host_ui->xattr_size += CALC_DENT_SIZE(nm->len); host_ui->xattr_size += CALC_XATTR_BYTES(size); host_ui->xattr_names += nm->len; - /* - * We do not use i_size_write() because nobody can race with us as we - * are holding host @host->i_mutex - every xattr operation for this - * inode is serialized by it. - */ - inode->i_size = ui->ui_size = size; - ui->data_len = size; err = ubifs_jnl_update(c, host, nm, inode, 0, 1); if (err) goto out_cancel; @@ -172,8 +167,8 @@ out_cancel: host_ui->xattr_cnt -= 1; host_ui->xattr_size -= CALC_DENT_SIZE(nm->len); host_ui->xattr_size -= CALC_XATTR_BYTES(size); -out_unlock: mutex_unlock(&host_ui->ui_mutex); +out_free: make_bad_inode(inode); iput(inode); out_budg: @@ -207,22 +202,21 @@ static int change_xattr(struct ubifs_info *c, struct inode *host, if (err) return err; - mutex_lock(&host_ui->ui_mutex); - host->i_ctime = ubifs_current_time(host); - host_ui->xattr_size -= CALC_XATTR_BYTES(ui->data_len); - host_ui->xattr_size += CALC_XATTR_BYTES(size); - kfree(ui->data); ui->data = kmalloc(size, GFP_NOFS); if (!ui->data) { err = -ENOMEM; - goto out_unlock; + goto out_free; } - memcpy(ui->data, value, size); inode->i_size = ui->ui_size = size; ui->data_len = size; + mutex_lock(&host_ui->ui_mutex); + host->i_ctime = ubifs_current_time(host); + host_ui->xattr_size -= CALC_XATTR_BYTES(ui->data_len); + host_ui->xattr_size += CALC_XATTR_BYTES(size); + /* * It is important to write the host inode after the xattr inode * because if the host inode gets synchronized (via 'fsync()'), then @@ -240,9 +234,9 @@ static int change_xattr(struct ubifs_info *c, struct inode *host, out_cancel: host_ui->xattr_size -= CALC_XATTR_BYTES(size); host_ui->xattr_size += CALC_XATTR_BYTES(ui->data_len); - make_bad_inode(inode); -out_unlock: mutex_unlock(&host_ui->ui_mutex); + make_bad_inode(inode); +out_free: ubifs_release_budget(c, &req); return err; } @@ -312,6 +306,7 @@ int ubifs_setxattr(struct dentry *dentry, const char *name, dbg_gen("xattr '%s', host ino %lu ('%.*s'), size %zd", name, host->i_ino, dentry->d_name.len, dentry->d_name.name, size); + ubifs_assert(mutex_is_locked(&host->i_mutex)); if (size > UBIFS_MAX_INO_DATA) return -ERANGE; @@ -384,7 +379,6 @@ ssize_t ubifs_getxattr(struct dentry *dentry, const char *name, void *buf, if (!xent) return -ENOMEM; - mutex_lock(&host->i_mutex); xent_key_init(c, &key, host->i_ino, &nm); err = ubifs_tnc_lookup_nm(c, &key, xent, &nm); if (err) { @@ -419,7 +413,6 @@ ssize_t ubifs_getxattr(struct dentry *dentry, const char *name, void *buf, out_iput: iput(inode); out_unlock: - mutex_unlock(&host->i_mutex); kfree(xent); return err; } @@ -449,8 +442,6 @@ ssize_t ubifs_listxattr(struct dentry *dentry, char *buffer, size_t size) return -ERANGE; lowest_xent_key(c, &key, host->i_ino); - - mutex_lock(&host->i_mutex); while (1) { int type; @@ -479,7 +470,6 @@ ssize_t ubifs_listxattr(struct dentry *dentry, char *buffer, size_t size) pxent = xent; key_read(c, &xent->key, &key); } - mutex_unlock(&host->i_mutex); kfree(pxent); if (err != -ENOENT) { -- cgit v1.2.3