summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2009-02-03 16:52:44 -0800
committerLinus Torvalds <torvalds@linux-foundation.org>2009-02-03 16:52:44 -0800
commitf96c08e8c5935d80bb9fd48c61b5bfa00878519e (patch)
treef0e34c84ce69a9e076c23299c94f0fd583a568e0
parent7420b73dc05ae1faa36b49774210a13d29349c53 (diff)
parent27ad27993313312a4ad0047d0a944c425cd511a5 (diff)
downloadblackbird-op-linux-f96c08e8c5935d80bb9fd48c61b5bfa00878519e.tar.gz
blackbird-op-linux-f96c08e8c5935d80bb9fd48c61b5bfa00878519e.zip
Merge branch 'linux-next' of git://git.infradead.org/ubifs-2.6
* 'linux-next' of git://git.infradead.org/ubifs-2.6: UBIFS: remove fast unmounting UBIFS: return sensible error codes UBIFS: remount ro fixes UBIFS: spelling fix 'date' -> 'data' UBIFS: sync wbufs after syncing inodes and pages UBIFS: fix LPT out-of-space bug (again) UBIFS: fix no_chk_data_crc UBIFS: fix assertions UBIFS: ensure orphan area head is initialized UBIFS: always clean up GC LEB space UBIFS: add re-mount debugging checks UBIFS: fix LEB list freeing UBIFS: simplify locking UBIFS: document dark_wm and dead_wm better UBIFS: do not treat all data as short term UBIFS: constify operations UBIFS: do not commit twice
-rw-r--r--Documentation/filesystems/ubifs.txt7
-rw-r--r--fs/ubifs/budget.c35
-rw-r--r--fs/ubifs/debug.c122
-rw-r--r--fs/ubifs/debug.h36
-rw-r--r--fs/ubifs/dir.c96
-rw-r--r--fs/ubifs/file.c9
-rw-r--r--fs/ubifs/gc.c28
-rw-r--r--fs/ubifs/io.c22
-rw-r--r--fs/ubifs/journal.c2
-rw-r--r--fs/ubifs/lprops.c12
-rw-r--r--fs/ubifs/lpt_commit.c44
-rw-r--r--fs/ubifs/master.c2
-rw-r--r--fs/ubifs/orphan.c38
-rw-r--r--fs/ubifs/super.c195
-rw-r--r--fs/ubifs/tnc.c12
-rw-r--r--fs/ubifs/ubifs.h26
16 files changed, 393 insertions, 293 deletions
diff --git a/Documentation/filesystems/ubifs.txt b/Documentation/filesystems/ubifs.txt
index 84da2a4ba25a..12fedb7834c6 100644
--- a/Documentation/filesystems/ubifs.txt
+++ b/Documentation/filesystems/ubifs.txt
@@ -79,13 +79,6 @@ Mount options
(*) == default.
-norm_unmount (*) commit on unmount; the journal is committed
- when the file-system is unmounted so that the
- next mount does not have to replay the journal
- and it becomes very fast;
-fast_unmount do not commit on unmount; this option makes
- unmount faster, but the next mount slower
- because of the need to replay the journal.
bulk_read read more in one go to take advantage of flash
media that read faster sequentially
no_bulk_read (*) do not bulk-read
diff --git a/fs/ubifs/budget.c b/fs/ubifs/budget.c
index 175f9c590b77..f393620890ee 100644
--- a/fs/ubifs/budget.c
+++ b/fs/ubifs/budget.c
@@ -689,7 +689,7 @@ long long ubifs_reported_space(const struct ubifs_info *c, long long free)
}
/**
- * ubifs_get_free_space - return amount of free space.
+ * ubifs_get_free_space_nolock - return amount of free space.
* @c: UBIFS file-system description object
*
* This function calculates amount of free space to report to user-space.
@@ -704,16 +704,14 @@ long long ubifs_reported_space(const struct ubifs_info *c, long long free)
* traditional file-systems, because they have way less overhead than UBIFS.
* So, to keep users happy, UBIFS tries to take the overhead into account.
*/
-long long ubifs_get_free_space(struct ubifs_info *c)
+long long ubifs_get_free_space_nolock(struct ubifs_info *c)
{
- int min_idx_lebs, rsvd_idx_lebs, lebs;
+ int rsvd_idx_lebs, lebs;
long long available, outstanding, free;
- spin_lock(&c->space_lock);
- min_idx_lebs = c->min_idx_lebs;
- ubifs_assert(min_idx_lebs == ubifs_calc_min_idx_lebs(c));
+ ubifs_assert(c->min_idx_lebs == ubifs_calc_min_idx_lebs(c));
outstanding = c->budg_data_growth + c->budg_dd_growth;
- available = ubifs_calc_available(c, min_idx_lebs);
+ available = ubifs_calc_available(c, c->min_idx_lebs);
/*
* When reporting free space to user-space, UBIFS guarantees that it is
@@ -726,15 +724,14 @@ long long ubifs_get_free_space(struct ubifs_info *c)
* Note, the calculations below are similar to what we have in
* 'do_budget_space()', so refer there for comments.
*/
- if (min_idx_lebs > c->lst.idx_lebs)
- rsvd_idx_lebs = min_idx_lebs - c->lst.idx_lebs;
+ if (c->min_idx_lebs > c->lst.idx_lebs)
+ rsvd_idx_lebs = c->min_idx_lebs - c->lst.idx_lebs;
else
rsvd_idx_lebs = 0;
lebs = c->lst.empty_lebs + c->freeable_cnt + c->idx_gc_cnt -
c->lst.taken_empty_lebs;
lebs -= rsvd_idx_lebs;
available += lebs * (c->dark_wm - c->leb_overhead);
- spin_unlock(&c->space_lock);
if (available > outstanding)
free = ubifs_reported_space(c, available - outstanding);
@@ -742,3 +739,21 @@ long long ubifs_get_free_space(struct ubifs_info *c)
free = 0;
return free;
}
+
+/**
+ * ubifs_get_free_space - return amount of free space.
+ * @c: UBIFS file-system description object
+ *
+ * This function calculates and retuns amount of free space to report to
+ * user-space.
+ */
+long long ubifs_get_free_space(struct ubifs_info *c)
+{
+ long long free;
+
+ spin_lock(&c->space_lock);
+ free = ubifs_get_free_space_nolock(c);
+ spin_unlock(&c->space_lock);
+
+ return free;
+}
diff --git a/fs/ubifs/debug.c b/fs/ubifs/debug.c
index 792c5a16c182..e975bd82f38b 100644
--- a/fs/ubifs/debug.c
+++ b/fs/ubifs/debug.c
@@ -620,9 +620,11 @@ void dbg_dump_budg(struct ubifs_info *c)
c->dark_wm, c->dead_wm, c->max_idx_node_sz);
printk(KERN_DEBUG "\tgc_lnum %d, ihead_lnum %d\n",
c->gc_lnum, c->ihead_lnum);
- for (i = 0; i < c->jhead_cnt; i++)
- printk(KERN_DEBUG "\tjhead %d\t LEB %d\n",
- c->jheads[i].wbuf.jhead, c->jheads[i].wbuf.lnum);
+ /* If we are in R/O mode, journal heads do not exist */
+ if (c->jheads)
+ for (i = 0; i < c->jhead_cnt; i++)
+ printk(KERN_DEBUG "\tjhead %d\t LEB %d\n",
+ c->jheads[i].wbuf.jhead, c->jheads[i].wbuf.lnum);
for (rb = rb_first(&c->buds); rb; rb = rb_next(rb)) {
bud = rb_entry(rb, struct ubifs_bud, rb);
printk(KERN_DEBUG "\tbud LEB %d\n", bud->lnum);
@@ -637,10 +639,7 @@ void dbg_dump_budg(struct ubifs_info *c)
/* Print budgeting predictions */
available = ubifs_calc_available(c, c->min_idx_lebs);
outstanding = c->budg_data_growth + c->budg_dd_growth;
- if (available > outstanding)
- free = ubifs_reported_space(c, available - outstanding);
- else
- free = 0;
+ free = ubifs_get_free_space_nolock(c);
printk(KERN_DEBUG "Budgeting predictions:\n");
printk(KERN_DEBUG "\tavailable: %lld, outstanding %lld, free %lld\n",
available, outstanding, free);
@@ -861,6 +860,65 @@ void dbg_dump_index(struct ubifs_info *c)
}
/**
+ * dbg_save_space_info - save information about flash space.
+ * @c: UBIFS file-system description object
+ *
+ * This function saves information about UBIFS free space, dirty space, etc, in
+ * order to check it later.
+ */
+void dbg_save_space_info(struct ubifs_info *c)
+{
+ struct ubifs_debug_info *d = c->dbg;
+
+ ubifs_get_lp_stats(c, &d->saved_lst);
+
+ spin_lock(&c->space_lock);
+ d->saved_free = ubifs_get_free_space_nolock(c);
+ spin_unlock(&c->space_lock);
+}
+
+/**
+ * dbg_check_space_info - check flash space information.
+ * @c: UBIFS file-system description object
+ *
+ * This function compares current flash space information with the information
+ * which was saved when the 'dbg_save_space_info()' function was called.
+ * Returns zero if the information has not changed, and %-EINVAL it it has
+ * changed.
+ */
+int dbg_check_space_info(struct ubifs_info *c)
+{
+ struct ubifs_debug_info *d = c->dbg;
+ struct ubifs_lp_stats lst;
+ long long avail, free;
+
+ spin_lock(&c->space_lock);
+ avail = ubifs_calc_available(c, c->min_idx_lebs);
+ spin_unlock(&c->space_lock);
+ free = ubifs_get_free_space(c);
+
+ if (free != d->saved_free) {
+ ubifs_err("free space changed from %lld to %lld",
+ d->saved_free, free);
+ goto out;
+ }
+
+ return 0;
+
+out:
+ ubifs_msg("saved lprops statistics dump");
+ dbg_dump_lstats(&d->saved_lst);
+ ubifs_get_lp_stats(c, &lst);
+ ubifs_msg("current lprops statistics dump");
+ dbg_dump_lstats(&d->saved_lst);
+ spin_lock(&c->space_lock);
+ dbg_dump_budg(c);
+ spin_unlock(&c->space_lock);
+ dump_stack();
+ return -EINVAL;
+}
+
+/**
* dbg_check_synced_i_size - check synchronized inode size.
* @inode: inode to check
*
@@ -1349,7 +1407,7 @@ int dbg_check_tnc(struct ubifs_info *c, int extra)
* @c: UBIFS file-system description object
* @leaf_cb: called for each leaf node
* @znode_cb: called for each indexing node
- * @priv: private date which is passed to callbacks
+ * @priv: private data which is passed to callbacks
*
* This function walks the UBIFS index and calls the @leaf_cb for each leaf
* node and @znode_cb for each indexing node. Returns zero in case of success
@@ -2409,7 +2467,7 @@ void ubifs_debugging_exit(struct ubifs_info *c)
* Root directory for UBIFS stuff in debugfs. Contains sub-directories which
* contain the stuff specific to particular file-system mounts.
*/
-static struct dentry *debugfs_rootdir;
+static struct dentry *dfs_rootdir;
/**
* dbg_debugfs_init - initialize debugfs file-system.
@@ -2421,9 +2479,9 @@ static struct dentry *debugfs_rootdir;
*/
int dbg_debugfs_init(void)
{
- debugfs_rootdir = debugfs_create_dir("ubifs", NULL);
- if (IS_ERR(debugfs_rootdir)) {
- int err = PTR_ERR(debugfs_rootdir);
+ dfs_rootdir = debugfs_create_dir("ubifs", NULL);
+ if (IS_ERR(dfs_rootdir)) {
+ int err = PTR_ERR(dfs_rootdir);
ubifs_err("cannot create \"ubifs\" debugfs directory, "
"error %d\n", err);
return err;
@@ -2437,7 +2495,7 @@ int dbg_debugfs_init(void)
*/
void dbg_debugfs_exit(void)
{
- debugfs_remove(debugfs_rootdir);
+ debugfs_remove(dfs_rootdir);
}
static int open_debugfs_file(struct inode *inode, struct file *file)
@@ -2452,13 +2510,13 @@ static ssize_t write_debugfs_file(struct file *file, const char __user *buf,
struct ubifs_info *c = file->private_data;
struct ubifs_debug_info *d = c->dbg;
- if (file->f_path.dentry == d->dump_lprops)
+ if (file->f_path.dentry == d->dfs_dump_lprops)
dbg_dump_lprops(c);
- else if (file->f_path.dentry == d->dump_budg) {
+ else if (file->f_path.dentry == d->dfs_dump_budg) {
spin_lock(&c->space_lock);
dbg_dump_budg(c);
spin_unlock(&c->space_lock);
- } else if (file->f_path.dentry == d->dump_tnc) {
+ } else if (file->f_path.dentry == d->dfs_dump_tnc) {
mutex_lock(&c->tnc_mutex);
dbg_dump_tnc(c);
mutex_unlock(&c->tnc_mutex);
@@ -2469,7 +2527,7 @@ static ssize_t write_debugfs_file(struct file *file, const char __user *buf,
return count;
}
-static const struct file_operations debugfs_fops = {
+static const struct file_operations dfs_fops = {
.open = open_debugfs_file,
.write = write_debugfs_file,
.owner = THIS_MODULE,
@@ -2494,36 +2552,32 @@ int dbg_debugfs_init_fs(struct ubifs_info *c)
struct dentry *dent;
struct ubifs_debug_info *d = c->dbg;
- sprintf(d->debugfs_dir_name, "ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
- d->debugfs_dir = debugfs_create_dir(d->debugfs_dir_name,
- debugfs_rootdir);
- if (IS_ERR(d->debugfs_dir)) {
- err = PTR_ERR(d->debugfs_dir);
+ sprintf(d->dfs_dir_name, "ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
+ d->dfs_dir = debugfs_create_dir(d->dfs_dir_name, dfs_rootdir);
+ if (IS_ERR(d->dfs_dir)) {
+ err = PTR_ERR(d->dfs_dir);
ubifs_err("cannot create \"%s\" debugfs directory, error %d\n",
- d->debugfs_dir_name, err);
+ d->dfs_dir_name, err);
goto out;
}
fname = "dump_lprops";
- dent = debugfs_create_file(fname, S_IWUGO, d->debugfs_dir, c,
- &debugfs_fops);
+ dent = debugfs_create_file(fname, S_IWUGO, d->dfs_dir, c, &dfs_fops);
if (IS_ERR(dent))
goto out_remove;
- d->dump_lprops = dent;
+ d->dfs_dump_lprops = dent;
fname = "dump_budg";
- dent = debugfs_create_file(fname, S_IWUGO, d->debugfs_dir, c,
- &debugfs_fops);
+ dent = debugfs_create_file(fname, S_IWUGO, d->dfs_dir, c, &dfs_fops);
if (IS_ERR(dent))
goto out_remove;
- d->dump_budg = dent;
+ d->dfs_dump_budg = dent;
fname = "dump_tnc";
- dent = debugfs_create_file(fname, S_IWUGO, d->debugfs_dir, c,
- &debugfs_fops);
+ dent = debugfs_create_file(fname, S_IWUGO, d->dfs_dir, c, &dfs_fops);
if (IS_ERR(dent))
goto out_remove;
- d->dump_tnc = dent;
+ d->dfs_dump_tnc = dent;
return 0;
@@ -2531,7 +2585,7 @@ out_remove:
err = PTR_ERR(dent);
ubifs_err("cannot create \"%s\" debugfs directory, error %d\n",
fname, err);
- debugfs_remove_recursive(d->debugfs_dir);
+ debugfs_remove_recursive(d->dfs_dir);
out:
return err;
}
@@ -2542,7 +2596,7 @@ out:
*/
void dbg_debugfs_exit_fs(struct ubifs_info *c)
{
- debugfs_remove_recursive(c->dbg->debugfs_dir);
+ debugfs_remove_recursive(c->dbg->dfs_dir);
}
#endif /* CONFIG_UBIFS_FS_DEBUG */
diff --git a/fs/ubifs/debug.h b/fs/ubifs/debug.h
index 9820d6999f7e..c1cd73b2e06e 100644
--- a/fs/ubifs/debug.h
+++ b/fs/ubifs/debug.h
@@ -41,15 +41,17 @@
* @chk_lpt_wastage: used by LPT tree size checker
* @chk_lpt_lebs: used by LPT tree size checker
* @new_nhead_offs: used by LPT tree size checker
- * @new_ihead_lnum: used by debugging to check ihead_lnum
- * @new_ihead_offs: used by debugging to check ihead_offs
+ * @new_ihead_lnum: used by debugging to check @c->ihead_lnum
+ * @new_ihead_offs: used by debugging to check @c->ihead_offs
*
- * debugfs_dir_name: name of debugfs directory containing this file-system's
- * files
- * debugfs_dir: direntry object of the file-system debugfs directory
- * dump_lprops: "dump lprops" debugfs knob
- * dump_budg: "dump budgeting information" debugfs knob
- * dump_tnc: "dump TNC" debugfs knob
+ * @saved_lst: saved lprops statistics (used by 'dbg_save_space_info()')
+ * @saved_free: saved free space (used by 'dbg_save_space_info()')
+ *
+ * dfs_dir_name: name of debugfs directory containing this file-system's files
+ * dfs_dir: direntry object of the file-system debugfs directory
+ * dfs_dump_lprops: "dump lprops" debugfs knob
+ * dfs_dump_budg: "dump budgeting information" debugfs knob
+ * dfs_dump_tnc: "dump TNC" debugfs knob
*/
struct ubifs_debug_info {
void *buf;
@@ -69,11 +71,14 @@ struct ubifs_debug_info {
int new_ihead_lnum;
int new_ihead_offs;
- char debugfs_dir_name[100];
- struct dentry *debugfs_dir;
- struct dentry *dump_lprops;
- struct dentry *dump_budg;
- struct dentry *dump_tnc;
+ struct ubifs_lp_stats saved_lst;
+ long long saved_free;
+
+ char dfs_dir_name[100];
+ struct dentry *dfs_dir;
+ struct dentry *dfs_dump_lprops;
+ struct dentry *dfs_dump_budg;
+ struct dentry *dfs_dump_tnc;
};
#define ubifs_assert(expr) do { \
@@ -297,7 +302,8 @@ int dbg_walk_index(struct ubifs_info *c, dbg_leaf_callback leaf_cb,
dbg_znode_callback znode_cb, void *priv);
/* Checking functions */
-
+void dbg_save_space_info(struct ubifs_info *c);
+int dbg_check_space_info(struct ubifs_info *c);
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);
@@ -439,6 +445,8 @@ void dbg_debugfs_exit_fs(struct ubifs_info *c);
#define dbg_walk_index(c, leaf_cb, znode_cb, priv) 0
#define dbg_old_index_check_init(c, zroot) 0
+#define dbg_save_space_info(c) ({})
+#define dbg_check_space_info(c) 0
#define dbg_check_old_index(c, zroot) 0
#define dbg_check_cats(c) 0
#define dbg_check_ltab(c) 0
diff --git a/fs/ubifs/dir.c b/fs/ubifs/dir.c
index f448ab1f9c38..f55d523c52bb 100644
--- a/fs/ubifs/dir.c
+++ b/fs/ubifs/dir.c
@@ -482,30 +482,29 @@ static int ubifs_dir_release(struct inode *dir, struct file *file)
}
/**
- * lock_2_inodes - lock two UBIFS inodes.
+ * lock_2_inodes - a wrapper for locking two UBIFS inodes.
* @inode1: first inode
* @inode2: second inode
+ *
+ * We do not implement any tricks to guarantee strict lock ordering, because
+ * VFS has already done it for us on the @i_mutex. So this is just a simple
+ * wrapper function.
*/
static void lock_2_inodes(struct inode *inode1, struct inode *inode2)
{
- if (inode1->i_ino < inode2->i_ino) {
- mutex_lock_nested(&ubifs_inode(inode1)->ui_mutex, WB_MUTEX_2);
- mutex_lock_nested(&ubifs_inode(inode2)->ui_mutex, WB_MUTEX_3);
- } else {
- mutex_lock_nested(&ubifs_inode(inode2)->ui_mutex, WB_MUTEX_2);
- mutex_lock_nested(&ubifs_inode(inode1)->ui_mutex, WB_MUTEX_3);
- }
+ mutex_lock_nested(&ubifs_inode(inode1)->ui_mutex, WB_MUTEX_1);
+ mutex_lock_nested(&ubifs_inode(inode2)->ui_mutex, WB_MUTEX_2);
}
/**
- * unlock_2_inodes - unlock two UBIFS inodes inodes.
+ * unlock_2_inodes - a wrapper for unlocking two UBIFS inodes.
* @inode1: first inode
* @inode2: second inode
*/
static void unlock_2_inodes(struct inode *inode1, struct inode *inode2)
{
- mutex_unlock(&ubifs_inode(inode1)->ui_mutex);
mutex_unlock(&ubifs_inode(inode2)->ui_mutex);
+ mutex_unlock(&ubifs_inode(inode1)->ui_mutex);
}
static int ubifs_link(struct dentry *old_dentry, struct inode *dir,
@@ -527,6 +526,8 @@ static int ubifs_link(struct dentry *old_dentry, struct inode *dir,
dbg_gen("dent '%.*s' to ino %lu (nlink %d) in dir ino %lu",
dentry->d_name.len, dentry->d_name.name, inode->i_ino,
inode->i_nlink, dir->i_ino);
+ ubifs_assert(mutex_is_locked(&dir->i_mutex));
+ ubifs_assert(mutex_is_locked(&inode->i_mutex));
err = dbg_check_synced_i_size(inode);
if (err)
return err;
@@ -580,6 +581,8 @@ static int ubifs_unlink(struct inode *dir, struct dentry *dentry)
dbg_gen("dent '%.*s' from ino %lu (nlink %d) in dir ino %lu",
dentry->d_name.len, dentry->d_name.name, inode->i_ino,
inode->i_nlink, dir->i_ino);
+ ubifs_assert(mutex_is_locked(&dir->i_mutex));
+ ubifs_assert(mutex_is_locked(&inode->i_mutex));
err = dbg_check_synced_i_size(inode);
if (err)
return err;
@@ -667,7 +670,8 @@ static int ubifs_rmdir(struct inode *dir, struct dentry *dentry)
dbg_gen("directory '%.*s', ino %lu in dir ino %lu", dentry->d_name.len,
dentry->d_name.name, inode->i_ino, dir->i_ino);
-
+ ubifs_assert(mutex_is_locked(&dir->i_mutex));
+ ubifs_assert(mutex_is_locked(&inode->i_mutex));
err = check_dir_empty(c, dentry->d_inode);
if (err)
return err;
@@ -922,59 +926,30 @@ out_budg:
}
/**
- * lock_3_inodes - lock three UBIFS inodes for rename.
+ * lock_3_inodes - a wrapper for locking three UBIFS inodes.
* @inode1: first inode
* @inode2: second inode
* @inode3: third inode
*
- * For 'ubifs_rename()', @inode1 may be the same as @inode2 whereas @inode3 may
- * be null.
+ * This function is used for 'ubifs_rename()' and @inode1 may be the same as
+ * @inode2 whereas @inode3 may be %NULL.
+ *
+ * We do not implement any tricks to guarantee strict lock ordering, because
+ * VFS has already done it for us on the @i_mutex. So this is just a simple
+ * wrapper function.
*/
static void lock_3_inodes(struct inode *inode1, struct inode *inode2,
struct inode *inode3)
{
- struct inode *i1, *i2, *i3;
-
- if (!inode3) {
- if (inode1 != inode2) {
- lock_2_inodes(inode1, inode2);
- return;
- }
- mutex_lock_nested(&ubifs_inode(inode1)->ui_mutex, WB_MUTEX_1);
- return;
- }
-
- if (inode1 == inode2) {
- lock_2_inodes(inode1, inode3);
- return;
- }
-
- /* 3 different inodes */
- if (inode1 < inode2) {
- i3 = inode2;
- if (inode1 < inode3) {
- i1 = inode1;
- i2 = inode3;
- } else {
- i1 = inode3;
- i2 = inode1;
- }
- } else {
- i3 = inode1;
- if (inode2 < inode3) {
- i1 = inode2;
- i2 = inode3;
- } else {
- i1 = inode3;
- i2 = inode2;
- }
- }
- mutex_lock_nested(&ubifs_inode(i1)->ui_mutex, WB_MUTEX_1);
- lock_2_inodes(i2, i3);
+ mutex_lock_nested(&ubifs_inode(inode1)->ui_mutex, WB_MUTEX_1);
+ if (inode2 != inode1)
+ mutex_lock_nested(&ubifs_inode(inode2)->ui_mutex, WB_MUTEX_2);
+ if (inode3)
+ mutex_lock_nested(&ubifs_inode(inode3)->ui_mutex, WB_MUTEX_3);
}
/**
- * unlock_3_inodes - unlock three UBIFS inodes for rename.
+ * unlock_3_inodes - a wrapper for unlocking three UBIFS inodes for rename.
* @inode1: first inode
* @inode2: second inode
* @inode3: third inode
@@ -982,11 +957,11 @@ static void lock_3_inodes(struct inode *inode1, struct inode *inode2,
static void unlock_3_inodes(struct inode *inode1, struct inode *inode2,
struct inode *inode3)
{
- mutex_unlock(&ubifs_inode(inode1)->ui_mutex);
- if (inode1 != inode2)
- mutex_unlock(&ubifs_inode(inode2)->ui_mutex);
if (inode3)
mutex_unlock(&ubifs_inode(inode3)->ui_mutex);
+ if (inode1 != inode2)
+ mutex_unlock(&ubifs_inode(inode2)->ui_mutex);
+ mutex_unlock(&ubifs_inode(inode1)->ui_mutex);
}
static int ubifs_rename(struct inode *old_dir, struct dentry *old_dentry,
@@ -1020,6 +995,11 @@ static int ubifs_rename(struct inode *old_dir, struct dentry *old_dentry,
"dir ino %lu", old_dentry->d_name.len, old_dentry->d_name.name,
old_inode->i_ino, old_dir->i_ino, new_dentry->d_name.len,
new_dentry->d_name.name, new_dir->i_ino);
+ ubifs_assert(mutex_is_locked(&old_dir->i_mutex));
+ ubifs_assert(mutex_is_locked(&new_dir->i_mutex));
+ if (unlink)
+ ubifs_assert(mutex_is_locked(&new_inode->i_mutex));
+
if (unlink && is_dir) {
err = check_dir_empty(c, new_inode);
@@ -1199,7 +1179,7 @@ int ubifs_getattr(struct vfsmount *mnt, struct dentry *dentry,
return 0;
}
-struct inode_operations ubifs_dir_inode_operations = {
+const struct inode_operations ubifs_dir_inode_operations = {
.lookup = ubifs_lookup,
.create = ubifs_create,
.link = ubifs_link,
@@ -1219,7 +1199,7 @@ struct inode_operations ubifs_dir_inode_operations = {
#endif
};
-struct file_operations ubifs_dir_operations = {
+const struct file_operations ubifs_dir_operations = {
.llseek = ubifs_dir_llseek,
.release = ubifs_dir_release,
.read = generic_read_dir,
diff --git a/fs/ubifs/file.c b/fs/ubifs/file.c
index bf37374567fa..93b6de51f261 100644
--- a/fs/ubifs/file.c
+++ b/fs/ubifs/file.c
@@ -432,7 +432,6 @@ static int ubifs_write_begin(struct file *file, struct address_space *mapping,
int uninitialized_var(err), appending = !!(pos + len > inode->i_size);
struct page *page;
-
ubifs_assert(ubifs_inode(inode)->ui_size == inode->i_size);
if (unlikely(c->ro_media))
@@ -1541,7 +1540,7 @@ static int ubifs_file_mmap(struct file *file, struct vm_area_struct *vma)
return 0;
}
-struct address_space_operations ubifs_file_address_operations = {
+const struct address_space_operations ubifs_file_address_operations = {
.readpage = ubifs_readpage,
.writepage = ubifs_writepage,
.write_begin = ubifs_write_begin,
@@ -1551,7 +1550,7 @@ struct address_space_operations ubifs_file_address_operations = {
.releasepage = ubifs_releasepage,
};
-struct inode_operations ubifs_file_inode_operations = {
+const struct inode_operations ubifs_file_inode_operations = {
.setattr = ubifs_setattr,
.getattr = ubifs_getattr,
#ifdef CONFIG_UBIFS_FS_XATTR
@@ -1562,14 +1561,14 @@ struct inode_operations ubifs_file_inode_operations = {
#endif
};
-struct inode_operations ubifs_symlink_inode_operations = {
+const struct inode_operations ubifs_symlink_inode_operations = {
.readlink = generic_readlink,
.follow_link = ubifs_follow_link,
.setattr = ubifs_setattr,
.getattr = ubifs_getattr,
};
-struct file_operations ubifs_file_operations = {
+const struct file_operations ubifs_file_operations = {
.llseek = generic_file_llseek,
.read = do_sync_read,
.write = do_sync_write,
diff --git a/fs/ubifs/gc.c b/fs/ubifs/gc.c
index 9832f9abe28e..a711d33b3d3e 100644
--- a/fs/ubifs/gc.c
+++ b/fs/ubifs/gc.c
@@ -31,6 +31,26 @@
* to be reused. Garbage collection will cause the number of dirty index nodes
* to grow, however sufficient space is reserved for the index to ensure the
* commit will never run out of space.
+ *
+ * Notes about dead watermark. At current UBIFS implementation we assume that
+ * LEBs which have less than @c->dead_wm bytes of free + dirty space are full
+ * and not worth garbage-collecting. The dead watermark is one min. I/O unit
+ * size, or min. UBIFS node size, depending on what is greater. Indeed, UBIFS
+ * Garbage Collector has to synchronize the GC head's write buffer before
+ * returning, so this is about wasting one min. I/O unit. However, UBIFS GC can
+ * actually reclaim even very small pieces of dirty space by garbage collecting
+ * enough dirty LEBs, but we do not bother doing this at this implementation.
+ *
+ * Notes about dark watermark. The results of GC work depends on how big are
+ * the UBIFS nodes GC deals with. Large nodes make GC waste more space. Indeed,
+ * if GC move data from LEB A to LEB B and nodes in LEB A are large, GC would
+ * have to waste large pieces of free space at the end of LEB B, because nodes
+ * from LEB A would not fit. And the worst situation is when all nodes are of
+ * maximum size. So dark watermark is the amount of free + dirty space in LEB
+ * which are guaranteed to be reclaimable. If LEB has less space, the GC migh
+ * be unable to reclaim it. So, LEBs with free + dirty greater than dark
+ * watermark are "good" LEBs from GC's point of few. The other LEBs are not so
+ * good, and GC takes extra care when moving them.
*/
#include <linux/pagemap.h>
@@ -381,7 +401,7 @@ int ubifs_garbage_collect_leb(struct ubifs_info *c, struct ubifs_lprops *lp)
/*
* Don't release the LEB until after the next commit, because
- * it may contain date which is needed for recovery. So
+ * it may contain data which is needed for recovery. So
* although we freed this LEB, it will become usable only after
* the commit.
*/
@@ -810,8 +830,9 @@ out:
* ubifs_destroy_idx_gc - destroy idx_gc list.
* @c: UBIFS file-system description object
*
- * This function destroys the idx_gc list. It is called when unmounting or
- * remounting read-only so locks are not needed.
+ * This function destroys the @c->idx_gc list. It is called when unmounting
+ * so locks are not needed. Returns zero in case of success and a negative
+ * error code in case of failure.
*/
void ubifs_destroy_idx_gc(struct ubifs_info *c)
{
@@ -824,7 +845,6 @@ void ubifs_destroy_idx_gc(struct ubifs_info *c)
list_del(&idx_gc->list);
kfree(idx_gc);
}
-
}
/**
diff --git a/fs/ubifs/io.c b/fs/ubifs/io.c
index 01682713af69..e8e632a1dcdf 100644
--- a/fs/ubifs/io.c
+++ b/fs/ubifs/io.c
@@ -29,7 +29,7 @@
* would have been wasted for padding to the nearest minimal I/O unit boundary.
* Instead, data first goes to the write-buffer and is flushed when the
* buffer is full or when it is not used for some time (by timer). This is
- * similarto the mechanism is used by JFFS2.
+ * similar to the mechanism is used by JFFS2.
*
* Write-buffers are defined by 'struct ubifs_wbuf' objects and protected by
* mutexes defined inside these objects. Since sometimes upper-level code
@@ -75,7 +75,7 @@ void ubifs_ro_mode(struct ubifs_info *c, int err)
* @lnum: logical eraseblock number
* @offs: offset within the logical eraseblock
* @quiet: print no messages
- * @chk_crc: indicates whether to always check the CRC
+ * @must_chk_crc: indicates whether to always check the CRC
*
* This function checks node magic number and CRC checksum. This function also
* validates node length to prevent UBIFS from becoming crazy when an attacker
@@ -83,11 +83,17 @@ void ubifs_ro_mode(struct ubifs_info *c, int err)
* node length in the common header could cause UBIFS to read memory outside of
* allocated buffer when checking the CRC checksum.
*
- * This function returns zero in case of success %-EUCLEAN in case of bad CRC
- * or magic.
+ * This function may skip data nodes CRC checking if @c->no_chk_data_crc is
+ * true, which is controlled by corresponding UBIFS mount option. However, if
+ * @must_chk_crc is true, then @c->no_chk_data_crc is ignored and CRC is
+ * checked. Similarly, if @c->always_chk_crc is true, @c->no_chk_data_crc is
+ * ignored and CRC is checked.
+ *
+ * This function returns zero in case of success and %-EUCLEAN in case of bad
+ * CRC or magic.
*/
int ubifs_check_node(const struct ubifs_info *c, const void *buf, int lnum,
- int offs, int quiet, int chk_crc)
+ int offs, int quiet, int must_chk_crc)
{
int err = -EINVAL, type, node_len;
uint32_t crc, node_crc, magic;
@@ -123,9 +129,9 @@ int ubifs_check_node(const struct ubifs_info *c, const void *buf, int lnum,
node_len > c->ranges[type].max_len)
goto out_len;
- if (!chk_crc && type == UBIFS_DATA_NODE && !c->always_chk_crc)
- if (c->no_chk_data_crc)
- return 0;
+ if (!must_chk_crc && type == UBIFS_DATA_NODE && !c->always_chk_crc &&
+ c->no_chk_data_crc)
+ return 0;
crc = crc32(UBIFS_CRC32_INIT, buf + 8, node_len - 8);
node_crc = le32_to_cpu(ch->crc);
diff --git a/fs/ubifs/journal.c b/fs/ubifs/journal.c
index 9b7c54e0cd2a..a11ca0958a23 100644
--- a/fs/ubifs/journal.c
+++ b/fs/ubifs/journal.c
@@ -208,7 +208,7 @@ again:
offs = 0;
out:
- err = ubifs_wbuf_seek_nolock(wbuf, lnum, offs, UBI_SHORTTERM);
+ err = ubifs_wbuf_seek_nolock(wbuf, lnum, offs, wbuf->dtype);
if (err)
goto out_unlock;
diff --git a/fs/ubifs/lprops.c b/fs/ubifs/lprops.c
index dfd2bcece27a..4cdd284dea56 100644
--- a/fs/ubifs/lprops.c
+++ b/fs/ubifs/lprops.c
@@ -635,10 +635,10 @@ const struct ubifs_lprops *ubifs_change_lp(struct ubifs_info *c,
* @c: UBIFS file-system description object
* @st: return statistics
*/
-void ubifs_get_lp_stats(struct ubifs_info *c, struct ubifs_lp_stats *st)
+void ubifs_get_lp_stats(struct ubifs_info *c, struct ubifs_lp_stats *lst)
{
spin_lock(&c->space_lock);
- memcpy(st, &c->lst, sizeof(struct ubifs_lp_stats));
+ memcpy(lst, &c->lst, sizeof(struct ubifs_lp_stats));
spin_unlock(&c->space_lock);
}
@@ -678,6 +678,9 @@ int ubifs_change_one_lp(struct ubifs_info *c, int lnum, int free, int dirty,
out:
ubifs_release_lprops(c);
+ if (err)
+ ubifs_err("cannot change properties of LEB %d, error %d",
+ lnum, err);
return err;
}
@@ -714,6 +717,9 @@ int ubifs_update_one_lp(struct ubifs_info *c, int lnum, int free, int dirty,
out:
ubifs_release_lprops(c);
+ if (err)
+ ubifs_err("cannot update properties of LEB %d, error %d",
+ lnum, err);
return err;
}
@@ -737,6 +743,8 @@ int ubifs_read_one_lp(struct ubifs_info *c, int lnum, struct ubifs_lprops *lp)
lpp = ubifs_lpt_lookup(c, lnum);
if (IS_ERR(lpp)) {
err = PTR_ERR(lpp);
+ ubifs_err("cannot read properties of LEB %d, error %d",
+ lnum, err);
goto out;
}
diff --git a/fs/ubifs/lpt_commit.c b/fs/ubifs/lpt_commit.c
index 96ca95707175..3216a1f277f8 100644
--- a/fs/ubifs/lpt_commit.c
+++ b/fs/ubifs/lpt_commit.c
@@ -556,23 +556,23 @@ no_space:
}
/**
- * next_pnode - find next pnode.
+ * next_pnode_to_dirty - find next pnode to dirty.
* @c: UBIFS file-system description object
* @pnode: pnode
*
- * This function returns the next pnode or %NULL if there are no more pnodes.
+ * This function returns the next pnode to dirty or %NULL if there are no more
+ * pnodes. Note that pnodes that have never been written (lnum == 0) are
+ * skipped.
*/
-static struct ubifs_pnode *next_pnode(struct ubifs_info *c,
- struct ubifs_pnode *pnode)
+static struct ubifs_pnode *next_pnode_to_dirty(struct ubifs_info *c,
+ struct ubifs_pnode *pnode)
{
struct ubifs_nnode *nnode;
int iip;
/* Try to go right */
nnode = pnode->parent;
- iip = pnode->iip + 1;
- if (iip < UBIFS_LPT_FANOUT) {
- /* We assume here that LEB zero is never an LPT LEB */
+ for (iip = pnode->iip + 1; iip < UBIFS_LPT_FANOUT; iip++) {
if (nnode->nbranch[iip].lnum)
return ubifs_get_pnode(c, nnode, iip);
}
@@ -583,8 +583,11 @@ static struct ubifs_pnode *next_pnode(struct ubifs_info *c,
nnode = nnode->parent;
if (!nnode)
return NULL;
- /* We assume here that LEB zero is never an LPT LEB */
- } while (iip >= UBIFS_LPT_FANOUT || !nnode->nbranch[iip].lnum);
+ for (; iip < UBIFS_LPT_FANOUT; iip++) {
+ if (nnode->nbranch[iip].lnum)
+ break;
+ }
+ } while (iip >= UBIFS_LPT_FANOUT);
/* Go right */
nnode = ubifs_get_nnode(c, nnode, iip);
@@ -593,12 +596,29 @@ static struct ubifs_pnode *next_pnode(struct ubifs_info *c,
/* Go down to level 1 */
while (nnode->level > 1) {
- nnode = ubifs_get_nnode(c, nnode, 0);
+ for (iip = 0; iip < UBIFS_LPT_FANOUT; iip++) {
+ if (nnode->nbranch[iip].lnum)
+ break;
+ }
+ if (iip >= UBIFS_LPT_FANOUT) {
+ /*
+ * Should not happen, but we need to keep going
+ * if it does.
+ */
+ iip = 0;
+ }
+ nnode = ubifs_get_nnode(c, nnode, iip);
if (IS_ERR(nnode))
return (void *)nnode;
}
- return ubifs_get_pnode(c, nnode, 0);
+ for (iip = 0; iip < UBIFS_LPT_FANOUT; iip++)
+ if (nnode->nbranch[iip].lnum)
+ break;
+ if (iip >= UBIFS_LPT_FANOUT)
+ /* Should not happen, but we need to keep going if it does */
+ iip = 0;
+ return ubifs_get_pnode(c, nnode, iip);
}
/**
@@ -688,7 +708,7 @@ static int make_tree_dirty(struct ubifs_info *c)
pnode = pnode_lookup(c, 0);
while (pnode) {
do_make_pnode_dirty(c, pnode);
- pnode = next_pnode(c, pnode);
+ pnode = next_pnode_to_dirty(c, pnode);
if (IS_ERR(pnode))
return PTR_ERR(pnode);
}
diff --git a/fs/ubifs/master.c b/fs/ubifs/master.c
index 71d5493bf565..a88f33801b98 100644
--- a/fs/ubifs/master.c
+++ b/fs/ubifs/master.c
@@ -354,7 +354,7 @@ int ubifs_write_master(struct ubifs_info *c)
int err, lnum, offs, len;
if (c->ro_media)
- return -EINVAL;
+ return -EROFS;
lnum = UBIFS_MST_LNUM;
offs = c->mst_offs + c->mst_node_alsz;
diff --git a/fs/ubifs/orphan.c b/fs/ubifs/orphan.c
index 9e6f403f170e..152a7b34a141 100644
--- a/fs/ubifs/orphan.c
+++ b/fs/ubifs/orphan.c
@@ -46,7 +46,7 @@
* Orphans are accumulated in a rb-tree. When an inode's link count drops to
* zero, the inode number is added to the rb-tree. It is removed from the tree
* when the inode is deleted. Any new orphans that are in the orphan tree when
- * the commit is run, are written to the orphan area in 1 or more orph nodes.
+ * the commit is run, are written to the orphan area in 1 or more orphan nodes.
* If the orphan area is full, it is consolidated to make space. There is
* always enough space because validation prevents the user from creating more
* than the maximum number of orphans allowed.
@@ -231,7 +231,7 @@ static int tot_avail_orphs(struct ubifs_info *c)
}
/**
- * do_write_orph_node - write a node
+ * do_write_orph_node - write a node to the orphan head.
* @c: UBIFS file-system description object
* @len: length of node
* @atomic: write atomically
@@ -264,11 +264,11 @@ static int do_write_orph_node(struct ubifs_info *c, int len, int atomic)
}
/**
- * write_orph_node - write an orph node
+ * write_orph_node - write an orphan node.
* @c: UBIFS file-system description object
* @atomic: write atomically
*
- * This function builds an orph node from the cnext list and writes it to the
+ * This function builds an orphan node from the cnext list and writes it to the
* orphan head. On success, %0 is returned, otherwise a negative error code
* is returned.
*/
@@ -326,11 +326,11 @@ static int write_orph_node(struct ubifs_info *c, int atomic)
}
/**
- * write_orph_nodes - write orph nodes until there are no more to commit
+ * write_orph_nodes - write orphan nodes until there are no more to commit.
* @c: UBIFS file-system description object
* @atomic: write atomically
*
- * This function writes orph nodes for all the orphans to commit. On success,
+ * This function writes orphan nodes for all the orphans to commit. On success,
* %0 is returned, otherwise a negative error code is returned.
*/
static int write_orph_nodes(struct ubifs_info *c, int atomic)
@@ -478,14 +478,14 @@ int ubifs_orphan_end_commit(struct ubifs_info *c)
}
/**
- * clear_orphans - erase all LEBs used for orphans.
+ * ubifs_clear_orphans - erase all LEBs used for orphans.
* @c: UBIFS file-system description object
*
* If recovery is not required, then the orphans from the previous session
* are not needed. This function locates the LEBs used to record
* orphans, and un-maps them.
*/
-static int clear_orphans(struct ubifs_info *c)
+int ubifs_clear_orphans(struct ubifs_info *c)
{
int lnum, err;
@@ -547,9 +547,9 @@ static int insert_dead_orphan(struct ubifs_info *c, ino_t inum)
* do_kill_orphans - remove orphan inodes from the index.
* @c: UBIFS file-system description object
* @sleb: scanned LEB
- * @last_cmt_no: cmt_no of last orph node read is passed and returned here
+ * @last_cmt_no: cmt_no of last orphan node read is passed and returned here
* @outofdate: whether the LEB is out of date is returned here
- * @last_flagged: whether the end orph node is encountered
+ * @last_flagged: whether the end orphan node is encountered
*
* This function is a helper to the 'kill_orphans()' function. It goes through
* every orphan node in a LEB and for every inode number recorded, removes
@@ -580,8 +580,8 @@ static int do_kill_orphans(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
/*
* The commit number on the master node may be less, because
* of a failed commit. If there are several failed commits in a
- * row, the commit number written on orph nodes will continue to
- * increase (because the commit number is adjusted here) even
+ * row, the commit number written on orphan nodes will continue
+ * to increase (because the commit number is adjusted here) even
* though the commit number on the master node stays the same
* because the master node has not been re-written.
*/
@@ -589,9 +589,9 @@ static int do_kill_orphans(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
c->cmt_no = cmt_no;
if (cmt_no < *last_cmt_no && *last_flagged) {
/*
- * The last orph node had a higher commit number and was
- * flagged as the last written for that commit number.
- * That makes this orph node, out of date.
+ * The last orphan node had a higher commit number and
+ * was flagged as the last written for that commit
+ * number. That makes this orphan node, out of date.
*/
if (!first) {
ubifs_err("out of order commit number %llu in "
@@ -658,10 +658,10 @@ static int kill_orphans(struct ubifs_info *c)
/*
* Orph nodes always start at c->orph_first and are written to each
* successive LEB in turn. Generally unused LEBs will have been unmapped
- * but may contain out of date orph nodes if the unmap didn't go
- * through. In addition, the last orph node written for each commit is
+ * but may contain out of date orphan nodes if the unmap didn't go
+ * through. In addition, the last orphan node written for each commit is
* marked (top bit of orph->cmt_no is set to 1). It is possible that
- * there are orph nodes from the next commit (i.e. the commit did not
+ * there are orphan nodes from the next commit (i.e. the commit did not
* complete successfully). In that case, no orphans will have been lost
* due to the way that orphans are written, and any orphans added will
* be valid orphans anyway and so can be deleted.
@@ -718,7 +718,7 @@ int ubifs_mount_orphans(struct ubifs_info *c, int unclean, int read_only)
if (unclean)
err = kill_orphans(c);
else if (!read_only)
- err = clear_orphans(c);
+ err = ubifs_clear_orphans(c);
return err;
}
diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c
index 89556ee72518..1182b66a5491 100644
--- a/fs/ubifs/super.c
+++ b/fs/ubifs/super.c
@@ -397,6 +397,7 @@ static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
buf->f_namelen = UBIFS_MAX_NLEN;
buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
+ ubifs_assert(buf->f_bfree <= c->block_cnt);
return 0;
}
@@ -432,33 +433,24 @@ static int ubifs_sync_fs(struct super_block *sb, int wait)
int i, err;
struct ubifs_info *c = sb->s_fs_info;
struct writeback_control wbc = {
- .sync_mode = wait ? WB_SYNC_ALL : WB_SYNC_NONE,
+ .sync_mode = WB_SYNC_ALL,
.range_start = 0,
.range_end = LLONG_MAX,
.nr_to_write = LONG_MAX,
};
/*
- * Note by akpm about WB_SYNC_NONE used above: zero @wait is just an
- * advisory thing to help the file system shove lots of data into the
- * queues. If some gets missed then it'll be picked up on the second
+ * Zero @wait is just an advisory thing to help the file system shove
+ * lots of data into the queues, and there will be the second
* '->sync_fs()' call, with non-zero @wait.
*/
+ if (!wait)
+ return 0;
if (sb->s_flags & MS_RDONLY)
return 0;
/*
- * Synchronize write buffers, because 'ubifs_run_commit()' does not
- * do this if it waits for an already running commit.
- */
- for (i = 0; i < c->jhead_cnt; i++) {
- err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
- if (err)
- return err;
- }
-
- /*
* VFS calls '->sync_fs()' before synchronizing all dirty inodes and
* pages, so synchronize them first, then commit the journal. Strictly
* speaking, it is not necessary to commit the journal here,
@@ -469,6 +461,16 @@ static int ubifs_sync_fs(struct super_block *sb, int wait)
*/
generic_sync_sb_inodes(sb, &wbc);
+ /*
+ * Synchronize write buffers, because 'ubifs_run_commit()' does not
+ * do this if it waits for an already running commit.
+ */
+ for (i = 0; i < c->jhead_cnt; i++) {
+ err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
+ if (err)
+ return err;
+ }
+
err = ubifs_run_commit(c);
if (err)
return err;
@@ -572,15 +574,8 @@ static int init_constants_early(struct ubifs_info *c)
c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
/*
- * Initialize dead and dark LEB space watermarks.
- *
- * Dead space is the space which cannot be used. Its watermark is
- * equivalent to min. I/O unit or minimum node size if it is greater
- * then min. I/O unit.
- *
- * Dark space is the space which might be used, or might not, depending
- * on which node should be written to the LEB. Its watermark is
- * equivalent to maximum UBIFS node size.
+ * Initialize dead and dark LEB space watermarks. See gc.c for comments
+ * about these values.
*/
c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
@@ -741,12 +736,12 @@ static void init_constants_master(struct ubifs_info *c)
* take_gc_lnum - reserve GC LEB.
* @c: UBIFS file-system description object
*
- * This function ensures that the LEB reserved for garbage collection is
- * unmapped and is marked as "taken" in lprops. We also have to set free space
- * to LEB size and dirty space to zero, because lprops may contain out-of-date
- * information if the file-system was un-mounted before it has been committed.
- * This function returns zero in case of success and a negative error code in
- * case of failure.
+ * This function ensures that the LEB reserved for garbage collection is marked
+ * as "taken" in lprops. We also have to set free space to LEB size and dirty
+ * space to zero, because lprops may contain out-of-date information if the
+ * file-system was un-mounted before it has been committed. This function
+ * returns zero in case of success and a negative error code in case of
+ * failure.
*/
static int take_gc_lnum(struct ubifs_info *c)
{
@@ -757,10 +752,6 @@ static int take_gc_lnum(struct ubifs_info *c)
return -EINVAL;
}
- err = ubifs_leb_unmap(c, c->gc_lnum);
- if (err)
- return err;
-
/* And we have to tell lprops that this LEB is taken */
err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
LPROPS_TAKEN, 0, 0);
@@ -966,13 +957,16 @@ static int ubifs_parse_options(struct ubifs_info *c, char *options,
token = match_token(p, tokens, args);
switch (token) {
+ /*
+ * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
+ * We accepte them in order to be backware-compatible. But this
+ * should be removed at some point.
+ */
case Opt_fast_unmount:
c->mount_opts.unmount_mode = 2;
- c->fast_unmount = 1;
break;
case Opt_norm_unmount:
c->mount_opts.unmount_mode = 1;
- c->fast_unmount = 0;
break;
case Opt_bulk_read:
c->mount_opts.bulk_read = 2;
@@ -1094,12 +1088,7 @@ static int check_free_space(struct ubifs_info *c)
ubifs_err("insufficient free space to mount in read/write mode");
dbg_dump_budg(c);
dbg_dump_lprops(c);
- /*
- * We return %-EINVAL instead of %-ENOSPC because it seems to
- * be the closest error code mentioned in the mount function
- * documentation.
- */
- return -EINVAL;
+ return -ENOSPC;
}
return 0;
}
@@ -1286,10 +1275,19 @@ static int mount_ubifs(struct ubifs_info *c)
if (err)
goto out_orphans;
err = ubifs_rcvry_gc_commit(c);
- } else
+ } else {
err = take_gc_lnum(c);
- if (err)
- goto out_orphans;
+ if (err)
+ goto out_orphans;
+
+ /*
+ * GC LEB may contain garbage if there was an unclean
+ * reboot, and it should be un-mapped.
+ */
+ err = ubifs_leb_unmap(c, c->gc_lnum);
+ if (err)
+ return err;
+ }
err = dbg_check_lprops(c);
if (err)
@@ -1298,6 +1296,16 @@ static int mount_ubifs(struct ubifs_info *c)
err = ubifs_recover_size(c);
if (err)
goto out_orphans;
+ } else {
+ /*
+ * Even if we mount read-only, we have to set space in GC LEB
+ * to proper value because this affects UBIFS free space
+ * reporting. We do not want to have a situation when
+ * re-mounting from R/O to R/W changes amount of free space.
+ */
+ err = take_gc_lnum(c);
+ if (err)
+ goto out_orphans;
}
spin_lock(&ubifs_infos_lock);
@@ -1310,14 +1318,17 @@ static int mount_ubifs(struct ubifs_info *c)
else {
c->need_recovery = 0;
ubifs_msg("recovery completed");
+ /* GC LEB has to be empty and taken at this point */
+ ubifs_assert(c->lst.taken_empty_lebs == 1);
}
- }
+ } else
+ ubifs_assert(c->lst.taken_empty_lebs == 1);
- err = dbg_debugfs_init_fs(c);
+ err = dbg_check_filesystem(c);
if (err)
goto out_infos;
- err = dbg_check_filesystem(c);
+ err = dbg_debugfs_init_fs(c);
if (err)
goto out_infos;
@@ -1351,7 +1362,6 @@ static int mount_ubifs(struct ubifs_info *c)
c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
- dbg_msg("fast unmount: %d", c->fast_unmount);
dbg_msg("big_lpt %d", c->big_lpt);
dbg_msg("log LEBs: %d (%d - %d)",
c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
@@ -1475,10 +1485,8 @@ static int ubifs_remount_rw(struct ubifs_info *c)
{
int err, lnum;
- if (c->ro_media)
- return -EINVAL;
-
mutex_lock(&c->umount_mutex);
+ dbg_save_space_info(c);
c->remounting_rw = 1;
c->always_chk_crc = 1;
@@ -1514,6 +1522,12 @@ static int ubifs_remount_rw(struct ubifs_info *c)
err = ubifs_recover_inl_heads(c, c->sbuf);
if (err)
goto out;
+ } else {
+ /* A readonly mount is not allowed to have orphans */
+ ubifs_assert(c->tot_orphans == 0);
+ err = ubifs_clear_orphans(c);
+ if (err)
+ goto out;
}
if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
@@ -1569,7 +1583,7 @@ static int ubifs_remount_rw(struct ubifs_info *c)
if (c->need_recovery)
err = ubifs_rcvry_gc_commit(c);
else
- err = take_gc_lnum(c);
+ err = ubifs_leb_unmap(c, c->gc_lnum);
if (err)
goto out;
@@ -1582,8 +1596,9 @@ static int ubifs_remount_rw(struct ubifs_info *c)
c->vfs_sb->s_flags &= ~MS_RDONLY;
c->remounting_rw = 0;
c->always_chk_crc = 0;
+ err = dbg_check_space_info(c);
mutex_unlock(&c->umount_mutex);
- return 0;
+ return err;
out:
vfree(c->orph_buf);
@@ -1603,43 +1618,18 @@ out:
}
/**
- * commit_on_unmount - commit the journal when un-mounting.
- * @c: UBIFS file-system description object
- *
- * This function is called during un-mounting and re-mounting, and it commits
- * the journal unless the "fast unmount" mode is enabled.
- */
-static void commit_on_unmount(struct ubifs_info *c)
-{
- struct super_block *sb = c->vfs_sb;
- long long bud_bytes;
-
- /*
- * This function is called before the background thread is stopped, so
- * we may race with ongoing commit, which means we have to take
- * @c->bud_lock to access @c->bud_bytes.
- */
- spin_lock(&c->buds_lock);
- bud_bytes = c->bud_bytes;
- spin_unlock(&c->buds_lock);
-
- if (!c->fast_unmount && !(sb->s_flags & MS_RDONLY) && bud_bytes)
- ubifs_run_commit(c);
-}
-
-/**
* ubifs_remount_ro - re-mount in read-only mode.
* @c: UBIFS file-system description object
*
- * We rely on VFS to have stopped writing. Possibly the background thread could
- * be running a commit, however kthread_stop will wait in that case.
+ * We assume VFS has stopped writing. Possibly the background thread could be
+ * running a commit, however kthread_stop will wait in that case.
*/
static void ubifs_remount_ro(struct ubifs_info *c)
{
int i, err;
ubifs_assert(!c->need_recovery);
- commit_on_unmount(c);
+ ubifs_assert(!(c->vfs_sb->s_flags & MS_RDONLY));
mutex_lock(&c->umount_mutex);
if (c->bgt) {
@@ -1647,27 +1637,29 @@ static void ubifs_remount_ro(struct ubifs_info *c)
c->bgt = NULL;
}
+ dbg_save_space_info(c);
+
for (i = 0; i < c->jhead_cnt; i++) {
ubifs_wbuf_sync(&c->jheads[i].wbuf);
del_timer_sync(&c->jheads[i].wbuf.timer);
}
- if (!c->ro_media) {
- c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
- c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
- c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
- err = ubifs_write_master(c);
- if (err)
- ubifs_ro_mode(c, err);
- }
+ c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
+ c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
+ c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
+ err = ubifs_write_master(c);
+ if (err)
+ ubifs_ro_mode(c, err);
- ubifs_destroy_idx_gc(c);
free_wbufs(c);
vfree(c->orph_buf);
c->orph_buf = NULL;
vfree(c->ileb_buf);
c->ileb_buf = NULL;
ubifs_lpt_free(c, 1);
+ err = dbg_check_space_info(c);
+ if (err)
+ ubifs_ro_mode(c, err);
mutex_unlock(&c->umount_mutex);
}
@@ -1760,11 +1752,20 @@ static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
}
if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
+ if (c->ro_media) {
+ ubifs_msg("cannot re-mount due to prior errors");
+ return -EROFS;
+ }
err = ubifs_remount_rw(c);
if (err)
return err;
- } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
+ } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY)) {
+ if (c->ro_media) {
+ ubifs_msg("cannot re-mount due to prior errors");
+ return -EROFS;
+ }
ubifs_remount_ro(c);
+ }
if (c->bulk_read == 1)
bu_init(c);
@@ -1774,10 +1775,11 @@ static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
c->bu.buf = NULL;
}
+ ubifs_assert(c->lst.taken_empty_lebs == 1);
return 0;
}
-struct super_operations ubifs_super_operations = {
+const struct super_operations ubifs_super_operations = {
.alloc_inode = ubifs_alloc_inode,
.destroy_inode = ubifs_destroy_inode,
.put_super = ubifs_put_super,
@@ -2044,15 +2046,6 @@ out_close:
static void ubifs_kill_sb(struct super_block *sb)
{
- struct ubifs_info *c = sb->s_fs_info;
-
- /*
- * We do 'commit_on_unmount()' here instead of 'ubifs_put_super()'
- * in order to be outside BKL.
- */
- if (sb->s_root)
- commit_on_unmount(c);
- /* The un-mount routine is actually done in put_super() */
generic_shutdown_super(sb);
}
diff --git a/fs/ubifs/tnc.c b/fs/ubifs/tnc.c
index f7e36f545527..fa28a84c6a1b 100644
--- a/fs/ubifs/tnc.c
+++ b/fs/ubifs/tnc.c
@@ -443,6 +443,11 @@ static int tnc_read_node_nm(struct ubifs_info *c, struct ubifs_zbranch *zbr,
* This function performs that same function as ubifs_read_node except that
* it does not require that there is actually a node present and instead
* the return code indicates if a node was read.
+ *
+ * Note, this function does not check CRC of data nodes if @c->no_chk_data_crc
+ * is true (it is controlled by corresponding mount option). However, if
+ * @c->always_chk_crc is true, @c->no_chk_data_crc is ignored and CRC is always
+ * checked.
*/
static int try_read_node(const struct ubifs_info *c, void *buf, int type,
int len, int lnum, int offs)
@@ -470,9 +475,8 @@ static int try_read_node(const struct ubifs_info *c, void *buf, int type,
if (node_len != len)
return 0;
- if (type == UBIFS_DATA_NODE && !c->always_chk_crc)
- if (c->no_chk_data_crc)
- return 0;
+ if (type == UBIFS_DATA_NODE && !c->always_chk_crc && c->no_chk_data_crc)
+ return 1;
crc = crc32(UBIFS_CRC32_INIT, buf + 8, node_len - 8);
node_crc = le32_to_cpu(ch->crc);
@@ -1506,7 +1510,7 @@ out:
*
* Note, if the bulk-read buffer length (@bu->buf_len) is known, this function
* makes sure bulk-read nodes fit the buffer. Otherwise, this function prepares
- * maxumum possible amount of nodes for bulk-read.
+ * maximum possible amount of nodes for bulk-read.
*/
int ubifs_tnc_get_bu_keys(struct ubifs_info *c, struct bu_info *bu)
{
diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h
index fc2a4cc66d03..039a68bee29a 100644
--- a/fs/ubifs/ubifs.h
+++ b/fs/ubifs/ubifs.h
@@ -426,9 +426,9 @@ struct ubifs_unclean_leb {
* LEB properties flags.
*
* LPROPS_UNCAT: not categorized
- * LPROPS_DIRTY: dirty > 0, not index
+ * LPROPS_DIRTY: dirty > free, dirty >= @c->dead_wm, not index
* LPROPS_DIRTY_IDX: dirty + free > @c->min_idx_node_sze and index
- * LPROPS_FREE: free > 0, not empty, not index
+ * LPROPS_FREE: free > 0, dirty < @c->dead_wm, not empty, not index
* LPROPS_HEAP_CNT: number of heaps used for storing categorized LEBs
* LPROPS_EMPTY: LEB is empty, not taken
* LPROPS_FREEABLE: free + dirty == leb_size, not index, not taken
@@ -961,7 +961,6 @@ struct ubifs_debug_info;
* @cs_lock: commit state lock
* @cmt_wq: wait queue to sleep on if the log is full and a commit is running
*
- * @fast_unmount: do not run journal commit before un-mounting
* @big_lpt: flag that LPT is too big to write whole during commit
* @no_chk_data_crc: do not check CRCs when reading data nodes (except during
* recovery)
@@ -1202,7 +1201,6 @@ struct ubifs_info {
spinlock_t cs_lock;
wait_queue_head_t cmt_wq;
- unsigned int fast_unmount:1;
unsigned int big_lpt:1;
unsigned int no_chk_data_crc:1;
unsigned int bulk_read:1;
@@ -1405,13 +1403,13 @@ extern struct list_head ubifs_infos;
extern spinlock_t ubifs_infos_lock;
extern atomic_long_t ubifs_clean_zn_cnt;
extern struct kmem_cache *ubifs_inode_slab;
-extern struct super_operations ubifs_super_operations;
-extern struct address_space_operations ubifs_file_address_operations;
-extern struct file_operations ubifs_file_operations;
-extern struct inode_operations ubifs_file_inode_operations;
-extern struct file_operations ubifs_dir_operations;
-extern struct inode_operations ubifs_dir_inode_operations;
-extern struct inode_operations ubifs_symlink_inode_operations;
+extern const struct super_operations ubifs_super_operations;
+extern const struct address_space_operations ubifs_file_address_operations;
+extern const struct file_operations ubifs_file_operations;
+extern const struct inode_operations ubifs_file_inode_operations;
+extern const struct file_operations ubifs_dir_operations;
+extern const struct inode_operations ubifs_dir_inode_operations;
+extern const struct inode_operations ubifs_symlink_inode_operations;
extern struct backing_dev_info ubifs_backing_dev_info;
extern struct ubifs_compressor *ubifs_compressors[UBIFS_COMPR_TYPES_CNT];
@@ -1428,7 +1426,7 @@ int ubifs_read_node_wbuf(struct ubifs_wbuf *wbuf, void *buf, int type, int len,
int ubifs_write_node(struct ubifs_info *c, void *node, int len, int lnum,
int offs, int dtype);
int ubifs_check_node(const struct ubifs_info *c, const void *buf, int lnum,
- int offs, int quiet, int chk_crc);
+ int offs, int quiet, int must_chk_crc);
void ubifs_prepare_node(struct ubifs_info *c, void *buf, int len, int pad);
void ubifs_prep_grp_node(struct ubifs_info *c, void *node, int len, int last);
int ubifs_io_init(struct ubifs_info *c);
@@ -1495,6 +1493,7 @@ void ubifs_release_ino_dirty(struct ubifs_info *c, struct inode *inode,
void ubifs_cancel_ino_op(struct ubifs_info *c, struct inode *inode,
struct ubifs_budget_req *req);
long long ubifs_get_free_space(struct ubifs_info *c);
+long long ubifs_get_free_space_nolock(struct ubifs_info *c);
int ubifs_calc_min_idx_lebs(struct ubifs_info *c);
void ubifs_convert_page_budget(struct ubifs_info *c);
long long ubifs_reported_space(const struct ubifs_info *c, long long free);
@@ -1603,6 +1602,7 @@ void ubifs_delete_orphan(struct ubifs_info *c, ino_t inum);
int ubifs_orphan_start_commit(struct ubifs_info *c);
int ubifs_orphan_end_commit(struct ubifs_info *c);
int ubifs_mount_orphans(struct ubifs_info *c, int unclean, int read_only);
+int ubifs_clear_orphans(struct ubifs_info *c);
/* lpt.c */
int ubifs_calc_lpt_geom(struct ubifs_info *c);
@@ -1646,7 +1646,7 @@ const struct ubifs_lprops *ubifs_change_lp(struct ubifs_info *c,
const struct ubifs_lprops *lp,
int free, int dirty, int flags,
int idx_gc_cnt);
-void ubifs_get_lp_stats(struct ubifs_info *c, struct ubifs_lp_stats *stats);
+void ubifs_get_lp_stats(struct ubifs_info *c, struct ubifs_lp_stats *lst);
void ubifs_add_to_cat(struct ubifs_info *c, struct ubifs_lprops *lprops,
int cat);
void ubifs_replace_cat(struct ubifs_info *c, struct ubifs_lprops *old_lprops,
OpenPOWER on IntegriCloud