reiserfs: don't lock journal_init()
authorFrederic Weisbecker <fweisbec@gmail.com>
Tue, 10 Jan 2012 23:11:09 +0000 (15:11 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 11 Jan 2012 00:30:53 +0000 (16:30 -0800)
journal_init() doesn't need the lock since no operation on the filesystem
is involved there.  journal_read() and get_list_bitmap() have yet to be
reviewed carefully though before removing the lock there.  Just keep the
it around these two calls for safety.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Jeff Mahoney <jeffm@suse.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/reiserfs/journal.c
fs/reiserfs/super.c

index cce8e87..c3cf54f 100644 (file)
@@ -2678,16 +2678,10 @@ int journal_init(struct super_block *sb, const char *j_dev_name,
        char b[BDEVNAME_SIZE];
        int ret;
 
-       /*
-        * Unlock here to avoid various RECLAIM-FS-ON <-> IN-RECLAIM-FS
-        * dependency inversion warnings.
-        */
-       reiserfs_write_unlock(sb);
        journal = SB_JOURNAL(sb) = vzalloc(sizeof(struct reiserfs_journal));
        if (!journal) {
                reiserfs_warning(sb, "journal-1256",
                                 "unable to get memory for journal structure");
-               reiserfs_write_lock(sb);
                return 1;
        }
        INIT_LIST_HEAD(&journal->j_bitmap_nodes);
@@ -2695,10 +2689,8 @@ int journal_init(struct super_block *sb, const char *j_dev_name,
        INIT_LIST_HEAD(&journal->j_working_list);
        INIT_LIST_HEAD(&journal->j_journal_list);
        journal->j_persistent_trans = 0;
-       ret = reiserfs_allocate_list_bitmaps(sb, journal->j_list_bitmap,
-                                          reiserfs_bmap_count(sb));
-       reiserfs_write_lock(sb);
-       if (ret)
+       if (reiserfs_allocate_list_bitmaps(sb, journal->j_list_bitmap,
+                                          reiserfs_bmap_count(sb)))
                goto free_and_return;
 
        allocate_bitmap_nodes(sb);
@@ -2727,27 +2719,11 @@ int journal_init(struct super_block *sb, const char *j_dev_name,
                goto free_and_return;
        }
 
-       /*
-        * We need to unlock here to avoid creating the following
-        * dependency:
-        * reiserfs_lock -> sysfs_mutex
-        * Because the reiserfs mmap path creates the following dependency:
-        * mm->mmap -> reiserfs_lock, hence we have
-        * mm->mmap -> reiserfs_lock ->sysfs_mutex
-        * This would ends up in a circular dependency with sysfs readdir path
-        * which does sysfs_mutex -> mm->mmap_sem
-        * This is fine because the reiserfs lock is useless in mount path,
-        * at least until we call journal_begin. We keep it for paranoid
-        * reasons.
-        */
-       reiserfs_write_unlock(sb);
        if (journal_init_dev(sb, journal, j_dev_name) != 0) {
-               reiserfs_write_lock(sb);
                reiserfs_warning(sb, "sh-462",
                                 "unable to initialize jornal device");
                goto free_and_return;
        }
-       reiserfs_write_lock(sb);
 
        rs = SB_DISK_SUPER_BLOCK(sb);
 
@@ -2829,9 +2805,7 @@ int journal_init(struct super_block *sb, const char *j_dev_name,
        journal->j_mount_id = 10;
        journal->j_state = 0;
        atomic_set(&(journal->j_jlock), 0);
-       reiserfs_write_unlock(sb);
        journal->j_cnode_free_list = allocate_cnodes(num_cnodes);
-       reiserfs_write_lock(sb);
        journal->j_cnode_free_orig = journal->j_cnode_free_list;
        journal->j_cnode_free = journal->j_cnode_free_list ? num_cnodes : 0;
        journal->j_cnode_used = 0;
@@ -2848,24 +2822,37 @@ int journal_init(struct super_block *sb, const char *j_dev_name,
 
        init_journal_hash(sb);
        jl = journal->j_current_jl;
+
+       /*
+        * get_list_bitmap() may call flush_commit_list() which
+        * requires the lock. Calling flush_commit_list() shouldn't happen
+        * this early but I like to be paranoid.
+        */
+       reiserfs_write_lock(sb);
        jl->j_list_bitmap = get_list_bitmap(sb, jl);
+       reiserfs_write_unlock(sb);
        if (!jl->j_list_bitmap) {
                reiserfs_warning(sb, "journal-2005",
                                 "get_list_bitmap failed for journal list 0");
                goto free_and_return;
        }
-       if (journal_read(sb) < 0) {
+
+       /*
+        * Journal_read needs to be inspected in order to push down
+        * the lock further inside (or even remove it).
+        */
+       reiserfs_write_lock(sb);
+       ret = journal_read(sb);
+       reiserfs_write_unlock(sb);
+       if (ret < 0) {
                reiserfs_warning(sb, "reiserfs-2006",
                                 "Replay Failure, unable to mount");
                goto free_and_return;
        }
 
        reiserfs_mounted_fs_count++;
-       if (reiserfs_mounted_fs_count <= 1) {
-               reiserfs_write_unlock(sb);
+       if (reiserfs_mounted_fs_count <= 1)
                commit_wq = alloc_workqueue("reiserfs", WQ_MEM_RECLAIM, 0);
-               reiserfs_write_lock(sb);
-       }
 
        INIT_DELAYED_WORK(&journal->j_work, flush_async_commits);
        journal->j_work_sb = sb;
index 620dd5d..61b6038 100644 (file)
@@ -1826,6 +1826,17 @@ static int reiserfs_fill_super(struct super_block *s, void *data, int silent)
                printk("reiserfs: using flush barriers\n");
        }
 
+       // set_device_ro(s->s_dev, 1) ;
+       if (journal_init(s, jdev_name, old_format, commit_max_age)) {
+               SWARN(silent, s, "sh-2022",
+                     "unable to initialize journal space");
+               goto error_unlocked;
+       } else {
+               jinit_done = 1; /* once this is set, journal_release must be called
+                                ** if we error out of the mount
+                                */
+       }
+
        /*
         * This path assumed to be called with the BKL in the old times.
         * Now we have inherited the big reiserfs lock from it and many
@@ -1836,16 +1847,6 @@ static int reiserfs_fill_super(struct super_block *s, void *data, int silent)
         */
        reiserfs_write_lock(s);
 
-       // set_device_ro(s->s_dev, 1) ;
-       if (journal_init(s, jdev_name, old_format, commit_max_age)) {
-               SWARN(silent, s, "sh-2022",
-                     "unable to initialize journal space");
-               goto error;
-       } else {
-               jinit_done = 1; /* once this is set, journal_release must be called
-                                ** if we error out of the mount
-                                */
-       }
        if (reread_meta_blocks(s)) {
                SWARN(silent, s, "jmacd-9",
                      "unable to reread meta blocks after journal init");