5 files changed, 216 insertions, 31 deletions

diff --git a/fs/jbd2/checkpoint.c b/fs/jbd2/checkpoint.c
index ca0f5eb62b20..30beb11ef928 100644
--- a/fs/jbd2/checkpoint.c
+++ b/fs/jbd2/checkpoint.c
@@ -22,6 +22,7 @@
 #include <linux/jbd2.h>
 #include <linux/errno.h>
 #include <linux/slab.h>
+#include <linux/blkdev.h>
 #include <trace/events/jbd2.h>
 
 /*
@@ -506,6 +507,7 @@ int jbd2_cleanup_journal_tail(journal_t *journal)
 	if (blocknr < journal->j_tail)
 		freed = freed + journal->j_last - journal->j_first;
 
+	trace_jbd2_cleanup_journal_tail(journal, first_tid, blocknr, freed);
 	jbd_debug(1,
 		  "Cleaning journal tail from %d to %d (offset %lu), "
 		  "freeing %lu\n",
@@ -515,6 +517,20 @@ int jbd2_cleanup_journal_tail(journal_t *journal)
 	journal->j_tail_sequence = first_tid;
 	journal->j_tail = blocknr;
 	spin_unlock(&journal->j_state_lock);
+
+	/*
+	 * If there is an external journal, we need to make sure that
+	 * any data blocks that were recently written out --- perhaps
+	 * by jbd2_log_do_checkpoint() --- are flushed out before we
+	 * drop the transactions from the external journal.  It's
+	 * unlikely this will be necessary, especially with a
+	 * appropriately sized journal, but we need this to guarantee
+	 * correctness.  Fortunately jbd2_cleanup_journal_tail()
+	 * doesn't get called all that often.
+	 */
+	if ((journal->j_fs_dev != journal->j_dev) &&
+	    (journal->j_flags & JBD2_BARRIER))
+		blkdev_issue_flush(journal->j_fs_dev, NULL);
 	if (!(journal->j_flags & JBD2_ABORT))
 		jbd2_journal_update_superblock(journal, 1);
 	return 0;
diff --git a/fs/jbd2/commit.c b/fs/jbd2/commit.c
index d4cfd6d2779e..671da7fb7ffd 100644
--- a/fs/jbd2/commit.c
+++ b/fs/jbd2/commit.c
@@ -259,6 +259,7 @@ static int journal_submit_data_buffers(journal_t *journal,
 			ret = err;
 		spin_lock(&journal->j_list_lock);
 		J_ASSERT(jinode->i_transaction == commit_transaction);
+		commit_transaction->t_flushed_data_blocks = 1;
 		jinode->i_flags &= ~JI_COMMIT_RUNNING;
 		wake_up_bit(&jinode->i_flags, __JI_COMMIT_RUNNING);
 	}
@@ -286,7 +287,7 @@ static int journal_finish_inode_data_buffers(journal_t *journal,
 		if (err) {
 			/*
 			 * Because AS_EIO is cleared by
-			 * wait_on_page_writeback_range(), set it again so
+			 * filemap_fdatawait_range(), set it again so
 			 * that user process can get -EIO from fsync().
 			 */
 			set_bit(AS_EIO,
@@ -636,6 +637,10 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 		JBUFFER_TRACE(jh, "ph3: write metadata");
 		flags = jbd2_journal_write_metadata_buffer(commit_transaction,
 						      jh, &new_jh, blocknr);
+		if (flags < 0) {
+			jbd2_journal_abort(journal, flags);
+			continue;
+		}
 		set_bit(BH_JWrite, &jh2bh(new_jh)->b_state);
 		wbuf[bufs++] = jh2bh(new_jh);
 
@@ -704,8 +709,17 @@ start_journal_io:
 		}
 	}
 
-	/* Done it all: now write the commit record asynchronously. */
+	/* 
+	 * If the journal is not located on the file system device,
+	 * then we must flush the file system device before we issue
+	 * the commit record
+	 */
+	if (commit_transaction->t_flushed_data_blocks &&
+	    (journal->j_fs_dev != journal->j_dev) &&
+	    (journal->j_flags & JBD2_BARRIER))
+		blkdev_issue_flush(journal->j_fs_dev, NULL);
 
+	/* Done it all: now write the commit record asynchronously. */
 	if (JBD2_HAS_INCOMPAT_FEATURE(journal,
 				      JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT)) {
 		err = journal_submit_commit_record(journal, commit_transaction,
@@ -716,13 +730,6 @@ start_journal_io:
 			blkdev_issue_flush(journal->j_dev, NULL);
 	}
 
-	/*
-	 * This is the right place to wait for data buffers both for ASYNC
-	 * and !ASYNC commit. If commit is ASYNC, we need to wait only after
-	 * the commit block went to disk (which happens above). If commit is
-	 * SYNC, we need to wait for data buffers before we start writing
-	 * commit block, which happens below in such setting.
-	 */
 	err = journal_finish_inode_data_buffers(journal, commit_transaction);
 	if (err) {
 		printk(KERN_WARNING
@@ -876,8 +883,7 @@ restart_loop:
 		spin_unlock(&journal->j_list_lock);
 		bh = jh2bh(jh);
 		jbd_lock_bh_state(bh);
-		J_ASSERT_JH(jh,	jh->b_transaction == commit_transaction ||
-			jh->b_transaction == journal->j_running_transaction);
+		J_ASSERT_JH(jh,	jh->b_transaction == commit_transaction);
 
 		/*
 		 * If there is undo-protected committed data against
@@ -923,12 +929,12 @@ restart_loop:
 		/* A buffer which has been freed while still being
 		 * journaled by a previous transaction may end up still
 		 * being dirty here, but we want to avoid writing back
-		 * that buffer in the future now that the last use has
-		 * been committed.  That's not only a performance gain,
-		 * it also stops aliasing problems if the buffer is left
-		 * behind for writeback and gets reallocated for another
+		 * that buffer in the future after the "add to orphan"
+		 * operation been committed,  That's not only a performance
+		 * gain, it also stops aliasing problems if the buffer is
+		 * left behind for writeback and gets reallocated for another
 		 * use in a different page. */
-		if (buffer_freed(bh)) {
+		if (buffer_freed(bh) && !jh->b_next_transaction) {
 			clear_buffer_freed(bh);
 			clear_buffer_jbddirty(bh);
 		}
diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index fed85388ee86..c03d4dce4d76 100644
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -39,6 +39,8 @@
 #include <linux/seq_file.h>
 #include <linux/math64.h>
 #include <linux/hash.h>
+#include <linux/log2.h>
+#include <linux/vmalloc.h>
 
 #define CREATE_TRACE_POINTS
 #include <trace/events/jbd2.h>
@@ -78,6 +80,7 @@ EXPORT_SYMBOL(jbd2_journal_errno);
 EXPORT_SYMBOL(jbd2_journal_ack_err);
 EXPORT_SYMBOL(jbd2_journal_clear_err);
 EXPORT_SYMBOL(jbd2_log_wait_commit);
+EXPORT_SYMBOL(jbd2_log_start_commit);
 EXPORT_SYMBOL(jbd2_journal_start_commit);
 EXPORT_SYMBOL(jbd2_journal_force_commit_nested);
 EXPORT_SYMBOL(jbd2_journal_wipe);
@@ -92,6 +95,7 @@ EXPORT_SYMBOL(jbd2_journal_begin_ordered_truncate);
 
 static int journal_convert_superblock_v1(journal_t *, journal_superblock_t *);
 static void __journal_abort_soft (journal_t *journal, int errno);
+static int jbd2_journal_create_slab(size_t slab_size);
 
 /*
  * Helper function used to manage commit timeouts
@@ -358,6 +362,10 @@ repeat:
 
 		jbd_unlock_bh_state(bh_in);
 		tmp = jbd2_alloc(bh_in->b_size, GFP_NOFS);
+		if (!tmp) {
+			jbd2_journal_put_journal_head(new_jh);
+			return -ENOMEM;
+		}
 		jbd_lock_bh_state(bh_in);
 		if (jh_in->b_frozen_data) {
 			jbd2_free(tmp, bh_in->b_size);
@@ -809,7 +817,7 @@ static journal_t * journal_init_common (void)
 	journal_t *journal;
 	int err;
 
-	journal = kzalloc(sizeof(*journal), GFP_KERNEL|__GFP_NOFAIL);
+	journal = kzalloc(sizeof(*journal), GFP_KERNEL);
 	if (!journal)
 		goto fail;
 
@@ -1243,11 +1251,25 @@ int jbd2_journal_load(journal_t *journal)
 		}
 	}
 
+	/*
+	 * Create a slab for this blocksize
+	 */
+	err = jbd2_journal_create_slab(be32_to_cpu(sb->s_blocksize));
+	if (err)
+		return err;
+
 	/* Let the recovery code check whether it needs to recover any
 	 * data from the journal. */
 	if (jbd2_journal_recover(journal))
 		goto recovery_error;
 
+	if (journal->j_failed_commit) {
+		printk(KERN_ERR "JBD2: journal transaction %u on %s "
+		       "is corrupt.\n", journal->j_failed_commit,
+		       journal->j_devname);
+		return -EIO;
+	}
+
 	/* OK, we've finished with the dynamic journal bits:
 	 * reinitialise the dynamic contents of the superblock in memory
 	 * and reset them on disk. */
@@ -1795,6 +1817,127 @@ size_t journal_tag_bytes(journal_t *journal)
 }
 
 /*
+ * JBD memory management
+ *
+ * These functions are used to allocate block-sized chunks of memory
+ * used for making copies of buffer_head data.  Very often it will be
+ * page-sized chunks of data, but sometimes it will be in
+ * sub-page-size chunks.  (For example, 16k pages on Power systems
+ * with a 4k block file system.)  For blocks smaller than a page, we
+ * use a SLAB allocator.  There are slab caches for each block size,
+ * which are allocated at mount time, if necessary, and we only free
+ * (all of) the slab caches when/if the jbd2 module is unloaded.  For
+ * this reason we don't need to a mutex to protect access to
+ * jbd2_slab[] allocating or releasing memory; only in
+ * jbd2_journal_create_slab().
+ */
+#define JBD2_MAX_SLABS 8
+static struct kmem_cache *jbd2_slab[JBD2_MAX_SLABS];
+static DECLARE_MUTEX(jbd2_slab_create_sem);
+
+static const char *jbd2_slab_names[JBD2_MAX_SLABS] = {
+	"jbd2_1k", "jbd2_2k", "jbd2_4k", "jbd2_8k",
+	"jbd2_16k", "jbd2_32k", "jbd2_64k", "jbd2_128k"
+};
+
+
+static void jbd2_journal_destroy_slabs(void)
+{
+	int i;
+
+	for (i = 0; i < JBD2_MAX_SLABS; i++) {
+		if (jbd2_slab[i])
+			kmem_cache_destroy(jbd2_slab[i]);
+		jbd2_slab[i] = NULL;
+	}
+}
+
+static int jbd2_journal_create_slab(size_t size)
+{
+	int i = order_base_2(size) - 10;
+	size_t slab_size;
+
+	if (size == PAGE_SIZE)
+		return 0;
+
+	if (i >= JBD2_MAX_SLABS)
+		return -EINVAL;
+
+	if (unlikely(i < 0))
+		i = 0;
+	down(&jbd2_slab_create_sem);
+	if (jbd2_slab[i]) {
+		up(&jbd2_slab_create_sem);
+		return 0;	/* Already created */
+	}
+
+	slab_size = 1 << (i+10);
+	jbd2_slab[i] = kmem_cache_create(jbd2_slab_names[i], slab_size,
+					 slab_size, 0, NULL);
+	up(&jbd2_slab_create_sem);
+	if (!jbd2_slab[i]) {
+		printk(KERN_EMERG "JBD2: no memory for jbd2_slab cache\n");
+		return -ENOMEM;
+	}
+	return 0;
+}
+
+static struct kmem_cache *get_slab(size_t size)
+{
+	int i = order_base_2(size) - 10;
+
+	BUG_ON(i >= JBD2_MAX_SLABS);
+	if (unlikely(i < 0))
+		i = 0;
+	BUG_ON(jbd2_slab[i] == 0);
+	return jbd2_slab[i];
+}
+
+void *jbd2_alloc(size_t size, gfp_t flags)
+{
+	void *ptr;
+
+	BUG_ON(size & (size-1)); /* Must be a power of 2 */
+
+	flags |= __GFP_REPEAT;
+	if (size == PAGE_SIZE)
+		ptr = (void *)__get_free_pages(flags, 0);
+	else if (size > PAGE_SIZE) {
+		int order = get_order(size);
+
+		if (order < 3)
+			ptr = (void *)__get_free_pages(flags, order);
+		else
+			ptr = vmalloc(size);
+	} else
+		ptr = kmem_cache_alloc(get_slab(size), flags);
+
+	/* Check alignment; SLUB has gotten this wrong in the past,
+	 * and this can lead to user data corruption! */
+	BUG_ON(((unsigned long) ptr) & (size-1));
+
+	return ptr;
+}
+
+void jbd2_free(void *ptr, size_t size)
+{
+	if (size == PAGE_SIZE) {
+		free_pages((unsigned long)ptr, 0);
+		return;
+	}
+	if (size > PAGE_SIZE) {
+		int order = get_order(size);
+
+		if (order < 3)
+			free_pages((unsigned long)ptr, order);
+		else
+			vfree(ptr);
+		return;
+	}
+	kmem_cache_free(get_slab(size), ptr);
+};
+
+/*
  * Journal_head storage management
  */
 static struct kmem_cache *jbd2_journal_head_cache;
@@ -2103,7 +2246,8 @@ static void __init jbd2_create_debugfs_entry(void)
 {
 	jbd2_debugfs_dir = debugfs_create_dir("jbd2", NULL);
 	if (jbd2_debugfs_dir)
-		jbd2_debug = debugfs_create_u8(JBD2_DEBUG_NAME, S_IRUGO,
+		jbd2_debug = debugfs_create_u8(JBD2_DEBUG_NAME,
+					       S_IRUGO | S_IWUSR,
 					       jbd2_debugfs_dir,
 					       &jbd2_journal_enable_debug);
 }
@@ -2191,6 +2335,7 @@ static void jbd2_journal_destroy_caches(void)
 	jbd2_journal_destroy_revoke_caches();
 	jbd2_journal_destroy_jbd2_journal_head_cache();
 	jbd2_journal_destroy_handle_cache();
+	jbd2_journal_destroy_slabs();
 }
 
 static int __init journal_init(void)
diff --git a/fs/jbd2/recovery.c b/fs/jbd2/recovery.c
index 73063285b13f..049281b7cb89 100644
--- a/fs/jbd2/recovery.c
+++ b/fs/jbd2/recovery.c
@@ -20,7 +20,6 @@
 #include <linux/fs.h>
 #include <linux/jbd2.h>
 #include <linux/errno.h>
-#include <linux/slab.h>
 #include <linux/crc32.h>
 #endif
 
diff --git a/fs/jbd2/transaction.c b/fs/jbd2/transaction.c
index a0512700542f..bfc70f57900f 100644
--- a/fs/jbd2/transaction.c
+++ b/fs/jbd2/transaction.c
@@ -1727,6 +1727,21 @@ static int journal_unmap_buffer(journal_t *journal, struct buffer_head *bh)
 	if (!jh)
 		goto zap_buffer_no_jh;
 
+	/*
+	 * We cannot remove the buffer from checkpoint lists until the
+	 * transaction adding inode to orphan list (let's call it T)
+	 * is committed.  Otherwise if the transaction changing the
+	 * buffer would be cleaned from the journal before T is
+	 * committed, a crash will cause that the correct contents of
+	 * the buffer will be lost.  On the other hand we have to
+	 * clear the buffer dirty bit at latest at the moment when the
+	 * transaction marking the buffer as freed in the filesystem
+	 * structures is committed because from that moment on the
+	 * buffer can be reallocated and used by a different page.
+	 * Since the block hasn't been freed yet but the inode has
+	 * already been added to orphan list, it is safe for us to add
+	 * the buffer to BJ_Forget list of the newest transaction.
+	 */
 	transaction = jh->b_transaction;
 	if (transaction == NULL) {
 		/* First case: not on any transaction.  If it
@@ -1783,16 +1798,15 @@ static int journal_unmap_buffer(journal_t *journal, struct buffer_head *bh)
 	} else if (transaction == journal->j_committing_transaction) {
 		JBUFFER_TRACE(jh, "on committing transaction");
 		/*
-		 * If it is committing, we simply cannot touch it.  We
-		 * can remove it's next_transaction pointer from the
-		 * running transaction if that is set, but nothing
-		 * else. */
+		 * The buffer is committing, we simply cannot touch
+		 * it. So we just set j_next_transaction to the
+		 * running transaction (if there is one) and mark
+		 * buffer as freed so that commit code knows it should
+		 * clear dirty bits when it is done with the buffer.
+		 */
 		set_buffer_freed(bh);
-		if (jh->b_next_transaction) {
-			J_ASSERT(jh->b_next_transaction ==
-					journal->j_running_transaction);
-			jh->b_next_transaction = NULL;
-		}
+		if (journal->j_running_transaction && buffer_jbddirty(bh))
+			jh->b_next_transaction = journal->j_running_transaction;
 		jbd2_journal_put_journal_head(jh);
 		spin_unlock(&journal->j_list_lock);
 		jbd_unlock_bh_state(bh);
@@ -1969,7 +1983,7 @@ void jbd2_journal_file_buffer(struct journal_head *jh,
  */
 void __jbd2_journal_refile_buffer(struct journal_head *jh)
 {
-	int was_dirty;
+	int was_dirty, jlist;
 	struct buffer_head *bh = jh2bh(jh);
 
 	J_ASSERT_JH(jh, jbd_is_locked_bh_state(bh));
@@ -1991,8 +2005,13 @@ void __jbd2_journal_refile_buffer(struct journal_head *jh)
 	__jbd2_journal_temp_unlink_buffer(jh);
 	jh->b_transaction = jh->b_next_transaction;
 	jh->b_next_transaction = NULL;
-	__jbd2_journal_file_buffer(jh, jh->b_transaction,
-				jh->b_modified ? BJ_Metadata : BJ_Reserved);
+	if (buffer_freed(bh))
+		jlist = BJ_Forget;
+	else if (jh->b_modified)
+		jlist = BJ_Metadata;
+	else
+		jlist = BJ_Reserved;
+	__jbd2_journal_file_buffer(jh, jh->b_transaction, jlist);
 	J_ASSERT_JH(jh, jh->b_transaction->t_state == T_RUNNING);
 
 	if (was_dirty)