4 files changed, 142 insertions, 124 deletions

diff --git a/fs/jbd2/checkpoint.c b/fs/jbd2/checkpoint.c
index 9497718fe920..17159cacbd9e 100644
--- a/fs/jbd2/checkpoint.c
+++ b/fs/jbd2/checkpoint.c
@@ -249,16 +249,14 @@ restart:
         return ret;
 }
 
-#define NR_BATCH        64
-
 static void
-__flush_batch(journal_t *journal, struct buffer_head **bhs, int *batch_count)
+__flush_batch(journal_t *journal, int *batch_count)
 {
         int i;
 
-        ll_rw_block(SWRITE, *batch_count, bhs);
+        ll_rw_block(SWRITE, *batch_count, journal->j_chkpt_bhs);
         for (i = 0; i < *batch_count; i++) {
-                struct buffer_head *bh = bhs[i];
+                struct buffer_head *bh = journal->j_chkpt_bhs[i];
                 clear_buffer_jwrite(bh);
                 BUFFER_TRACE(bh, "brelse");
                 __brelse(bh);
@@ -277,8 +275,7 @@ __flush_batch(journal_t *journal, struct buffer_head **bhs, int *batch_count)
  * Called under jbd_lock_bh_state(jh2bh(jh)), and drops it
  */
 static int __process_buffer(journal_t *journal, struct journal_head *jh,
-                        struct buffer_head **bhs, int *batch_count,
-                        transaction_t *transaction)
+                            int *batch_count, transaction_t *transaction)
 {
         struct buffer_head *bh = jh2bh(jh);
         int ret = 0;
@@ -325,14 +322,14 @@ static int __process_buffer(journal_t *journal, struct journal_head *jh,
                 get_bh(bh);
                 J_ASSERT_BH(bh, !buffer_jwrite(bh));
                 set_buffer_jwrite(bh);
-                bhs[*batch_count] = bh;
+                journal->j_chkpt_bhs[*batch_count] = bh;
                 __buffer_relink_io(jh);
                 jbd_unlock_bh_state(bh);
                 transaction->t_chp_stats.cs_written++;
                 (*batch_count)++;
-                if (*batch_count == NR_BATCH) {
+                if (*batch_count == JBD2_NR_BATCH) {
                         spin_unlock(&journal->j_list_lock);
-                        __flush_batch(journal, bhs, batch_count);
+                        __flush_batch(journal, batch_count);
                         ret = 1;
                 }
         }
@@ -388,7 +385,6 @@ restart:
         if (journal->j_checkpoint_transactions == transaction &&
                         transaction->t_tid == this_tid) {
                 int batch_count = 0;
-                struct buffer_head *bhs[NR_BATCH];
                 struct journal_head *jh;
                 int retry = 0, err;
 
@@ -402,7 +398,7 @@ restart:
                                 retry = 1;
                                 break;
                         }
-                        retry = __process_buffer(journal, jh, bhs, &batch_count,
+                        retry = __process_buffer(journal, jh, &batch_count,
                                                  transaction);
                         if (retry < 0 && !result)
                                 result = retry;
@@ -419,7 +415,7 @@ restart:
                                 spin_unlock(&journal->j_list_lock);
                                 retry = 1;
                         }
-                        __flush_batch(journal, bhs, &batch_count);
+                        __flush_batch(journal, &batch_count);
                 }
 
                 if (retry) {
@@ -686,6 +682,7 @@ int __jbd2_journal_remove_checkpoint(struct journal_head *jh)
            safely remove this transaction from the log */
 
         __jbd2_journal_drop_transaction(journal, transaction);
+        kfree(transaction);
 
         /* Just in case anybody was waiting for more transactions to be
            checkpointed... */
@@ -760,5 +757,4 @@ void __jbd2_journal_drop_transaction(journal_t *journal, transaction_t *transact
         J_ASSERT(journal->j_running_transaction != transaction);
 
         jbd_debug(1, "Dropping transaction %d, all done\n", transaction->t_tid);
-        kfree(transaction);
 }
diff --git a/fs/jbd2/commit.c b/fs/jbd2/commit.c
index c8a1bace685a..62804e57a44c 100644
--- a/fs/jbd2/commit.c
+++ b/fs/jbd2/commit.c
@@ -25,6 +25,7 @@
 #include <linux/crc32.h>
 #include <linux/writeback.h>
 #include <linux/backing-dev.h>
+#include <linux/bio.h>
 
 /*
  * Default IO end handler for temporary BJ_IO buffer_heads.
@@ -137,7 +138,7 @@ static int journal_submit_commit_record(journal_t *journal,
                 set_buffer_ordered(bh);
                 barrier_done = 1;
         }
-        ret = submit_bh(WRITE, bh);
+        ret = submit_bh(WRITE_SYNC, bh);
         if (barrier_done)
                 clear_buffer_ordered(bh);
 
@@ -158,7 +159,7 @@ static int journal_submit_commit_record(journal_t *journal,
                 lock_buffer(bh);
                 set_buffer_uptodate(bh);
                 clear_buffer_dirty(bh);
-                ret = submit_bh(WRITE, bh);
+                ret = submit_bh(WRITE_SYNC, bh);
         }
         *cbh = bh;
         return ret;
@@ -168,12 +169,34 @@ static int journal_submit_commit_record(journal_t *journal,
  * This function along with journal_submit_commit_record
  * allows to write the commit record asynchronously.
  */
-static int journal_wait_on_commit_record(struct buffer_head *bh)
+static int journal_wait_on_commit_record(journal_t *journal,
+                                         struct buffer_head *bh)
 {
         int ret = 0;
 
+retry:
         clear_buffer_dirty(bh);
         wait_on_buffer(bh);
+        if (buffer_eopnotsupp(bh) && (journal->j_flags & JBD2_BARRIER)) {
+                printk(KERN_WARNING
+                       "JBD2: wait_on_commit_record: sync failed on %s - "
+                       "disabling barriers\n", journal->j_devname);
+                spin_lock(&journal->j_state_lock);
+                journal->j_flags &= ~JBD2_BARRIER;
+                spin_unlock(&journal->j_state_lock);
+
+                lock_buffer(bh);
+                clear_buffer_dirty(bh);
+                set_buffer_uptodate(bh);
+                bh->b_end_io = journal_end_buffer_io_sync;
+
+                ret = submit_bh(WRITE_SYNC, bh);
+                if (ret) {
+                        unlock_buffer(bh);
+                        return ret;
+                }
+                goto retry;
+        }
 
         if (unlikely(!buffer_uptodate(bh)))
                 ret = -EIO;
@@ -332,13 +355,15 @@ void jbd2_journal_commit_transaction(journal_t *journal)
         int flags;
         int err;
         unsigned long long blocknr;
+        ktime_t start_time;
+        u64 commit_time;
         char *tagp = NULL;
         journal_header_t *header;
         journal_block_tag_t *tag = NULL;
         int space_left = 0;
         int first_tag = 0;
         int tag_flag;
-        int i;
+        int i, to_free = 0;
         int tag_bytes = journal_tag_bytes(journal);
         struct buffer_head *cbh = NULL; /* For transactional checksums */
         __u32 crc32_sum = ~0;
@@ -458,6 +483,7 @@ void jbd2_journal_commit_transaction(journal_t *journal)
         commit_transaction->t_state = T_FLUSH;
         journal->j_committing_transaction = commit_transaction;
         journal->j_running_transaction = NULL;
+        start_time = ktime_get();
         commit_transaction->t_log_start = journal->j_head;
         wake_up(&journal->j_wait_transaction_locked);
         spin_unlock(&journal->j_state_lock);
@@ -803,7 +829,7 @@ wait_for_iobuf:
                         __jbd2_journal_abort_hard(journal);
         }
         if (!err && !is_journal_aborted(journal))
-                err = journal_wait_on_commit_record(cbh);
+                err = journal_wait_on_commit_record(journal, cbh);
 
         if (err)
                 jbd2_journal_abort(journal, err);
@@ -981,14 +1007,23 @@ restart_loop:
         J_ASSERT(commit_transaction == journal->j_committing_transaction);
         journal->j_commit_sequence = commit_transaction->t_tid;
         journal->j_committing_transaction = NULL;
-        spin_unlock(&journal->j_state_lock);
+        commit_time = ktime_to_ns(ktime_sub(ktime_get(), start_time));
 
-        if (journal->j_commit_callback)
-                journal->j_commit_callback(journal, commit_transaction);
+        /*
+         * weight the commit time higher than the average time so we don't
+         * react too strongly to vast changes in the commit time
+         */
+        if (likely(journal->j_average_commit_time))
+                journal->j_average_commit_time = (commit_time +
+                                journal->j_average_commit_time*3) / 4;
+        else
+                journal->j_average_commit_time = commit_time;
+        spin_unlock(&journal->j_state_lock);
 
         if (commit_transaction->t_checkpoint_list == NULL &&
             commit_transaction->t_checkpoint_io_list == NULL) {
                 __jbd2_journal_drop_transaction(journal, commit_transaction);
+                to_free = 1;
         } else {
                 if (journal->j_checkpoint_transactions == NULL) {
                         journal->j_checkpoint_transactions = commit_transaction;
@@ -1007,11 +1042,16 @@ restart_loop:
         }
         spin_unlock(&journal->j_list_lock);
 
+        if (journal->j_commit_callback)
+                journal->j_commit_callback(journal, commit_transaction);
+
         trace_mark(jbd2_end_commit, "dev %s transaction %d head %d",
-                   journal->j_devname, journal->j_commit_sequence,
+                   journal->j_devname, commit_transaction->t_tid,
                    journal->j_tail_sequence);
         jbd_debug(1, "JBD: commit %d complete, head %d\n",
                   journal->j_commit_sequence, journal->j_tail_sequence);
+        if (to_free)
+                kfree(commit_transaction);
 
         wake_up(&journal->j_wait_done_commit);
 }
diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index f6bff9d6f8df..56675306ed81 100644
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -40,6 +40,7 @@
 
 #include <asm/uaccess.h>
 #include <asm/page.h>
+#include <asm/div64.h>
 
 EXPORT_SYMBOL(jbd2_journal_start);
 EXPORT_SYMBOL(jbd2_journal_restart);
@@ -66,7 +67,6 @@ EXPORT_SYMBOL(jbd2_journal_update_format);
 EXPORT_SYMBOL(jbd2_journal_check_used_features);
 EXPORT_SYMBOL(jbd2_journal_check_available_features);
 EXPORT_SYMBOL(jbd2_journal_set_features);
-EXPORT_SYMBOL(jbd2_journal_create);
 EXPORT_SYMBOL(jbd2_journal_load);
 EXPORT_SYMBOL(jbd2_journal_destroy);
 EXPORT_SYMBOL(jbd2_journal_abort);
@@ -132,8 +132,9 @@ static int kjournald2(void *arg)
         journal->j_task = current;
         wake_up(&journal->j_wait_done_commit);
 
-        printk(KERN_INFO "kjournald2 starting.  Commit interval %ld seconds\n",
-                        journal->j_commit_interval / HZ);
+        printk(KERN_INFO "kjournald2 starting: pid %d, dev %s, "
+               "commit interval %ld seconds\n", current->pid,
+               journal->j_devname, journal->j_commit_interval / HZ);
 
         /*
          * And now, wait forever for commit wakeup events.
@@ -650,6 +651,8 @@ struct journal_head *jbd2_journal_get_descriptor_buffer(journal_t *journal)
                 return NULL;
 
         bh = __getblk(journal->j_dev, blocknr, journal->j_blocksize);
+        if (!bh)
+                return NULL;
         lock_buffer(bh);
         memset(bh->b_data, 0, journal->j_blocksize);
         set_buffer_uptodate(bh);
@@ -843,6 +846,8 @@ static int jbd2_seq_info_show(struct seq_file *seq, void *v)
             jiffies_to_msecs(s->stats->u.run.rs_flushing / s->stats->ts_tid));
         seq_printf(seq, "  %ums logging transaction\n",
             jiffies_to_msecs(s->stats->u.run.rs_logging / s->stats->ts_tid));
+        seq_printf(seq, "  %luus average transaction commit time\n",
+                   do_div(s->journal->j_average_commit_time, 1000));
         seq_printf(seq, "  %lu handles per transaction\n",
             s->stats->u.run.rs_handle_count / s->stats->ts_tid);
         seq_printf(seq, "  %lu blocks per transaction\n",
@@ -980,6 +985,8 @@ static journal_t * journal_init_common (void)
         spin_lock_init(&journal->j_state_lock);
 
         journal->j_commit_interval = (HZ * JBD2_DEFAULT_MAX_COMMIT_AGE);
+        journal->j_min_batch_time = 0;
+        journal->j_max_batch_time = 15000; /* 15ms */
 
         /* The journal is marked for error until we succeed with recovery! */
         journal->j_flags = JBD2_ABORT;
@@ -1035,15 +1042,14 @@ journal_t * jbd2_journal_init_dev(struct block_device *bdev,
 
         /* journal descriptor can store up to n blocks -bzzz */
         journal->j_blocksize = blocksize;
+        jbd2_stats_proc_init(journal);
         n = journal->j_blocksize / sizeof(journal_block_tag_t);
         journal->j_wbufsize = n;
         journal->j_wbuf = kmalloc(n * sizeof(struct buffer_head*), GFP_KERNEL);
         if (!journal->j_wbuf) {
                 printk(KERN_ERR "%s: Cant allocate bhs for commit thread\n",
                         __func__);
-                kfree(journal);
-                journal = NULL;
-                goto out;
+                goto out_err;
         }
         journal->j_dev = bdev;
         journal->j_fs_dev = fs_dev;
@@ -1053,14 +1059,22 @@ journal_t * jbd2_journal_init_dev(struct block_device *bdev,
         p = journal->j_devname;
         while ((p = strchr(p, '/')))
                 *p = '!';
-        jbd2_stats_proc_init(journal);
 
         bh = __getblk(journal->j_dev, start, journal->j_blocksize);
-        J_ASSERT(bh != NULL);
+        if (!bh) {
+                printk(KERN_ERR
+                       "%s: Cannot get buffer for journal superblock\n",
+                       __func__);
+                goto out_err;
+        }
         journal->j_sb_buffer = bh;
         journal->j_superblock = (journal_superblock_t *)bh->b_data;
-out:
+
         return journal;
+out_err:
+        jbd2_stats_proc_exit(journal);
+        kfree(journal);
+        return NULL;
 }
 
 /**
@@ -1108,9 +1122,7 @@ journal_t * jbd2_journal_init_inode (struct inode *inode)
         if (!journal->j_wbuf) {
                 printk(KERN_ERR "%s: Cant allocate bhs for commit thread\n",
                         __func__);
-                jbd2_stats_proc_exit(journal);
-                kfree(journal);
-                return NULL;
+                goto out_err;
         }
 
         err = jbd2_journal_bmap(journal, 0, &blocknr);
@@ -1118,17 +1130,24 @@ journal_t * jbd2_journal_init_inode (struct inode *inode)
         if (err) {
                 printk(KERN_ERR "%s: Cannnot locate journal superblock\n",
                        __func__);
-                jbd2_stats_proc_exit(journal);
-                kfree(journal);
-                return NULL;
+                goto out_err;
         }
 
         bh = __getblk(journal->j_dev, blocknr, journal->j_blocksize);
-        J_ASSERT(bh != NULL);
+        if (!bh) {
+                printk(KERN_ERR
+                       "%s: Cannot get buffer for journal superblock\n",
+                       __func__);
+                goto out_err;
+        }
         journal->j_sb_buffer = bh;
         journal->j_superblock = (journal_superblock_t *)bh->b_data;
 
         return journal;
+out_err:
+        jbd2_stats_proc_exit(journal);
+        kfree(journal);
+        return NULL;
 }
 
 /*
@@ -1177,77 +1196,6 @@ static int journal_reset(journal_t *journal)
 }
 
 /**
- * int jbd2_journal_create() - Initialise the new journal file
- * @journal: Journal to create. This structure must have been initialised
- *
- * Given a journal_t structure which tells us which disk blocks we can
- * use, create a new journal superblock and initialise all of the
- * journal fields from scratch.
- **/
-int jbd2_journal_create(journal_t *journal)
-{
-        unsigned long long blocknr;
-        struct buffer_head *bh;
-        journal_superblock_t *sb;
-        int i, err;
-
-        if (journal->j_maxlen < JBD2_MIN_JOURNAL_BLOCKS) {
-                printk (KERN_ERR "Journal length (%d blocks) too short.\n",
-                        journal->j_maxlen);
-                journal_fail_superblock(journal);
-                return -EINVAL;
-        }
-
-        if (journal->j_inode == NULL) {
-                /*
-                 * We don't know what block to start at!
-                 */
-                printk(KERN_EMERG
-                       "%s: creation of journal on external device!\n",
-                       __func__);
-                BUG();
-        }
-
-        /* Zero out the entire journal on disk.  We cannot afford to
-           have any blocks on disk beginning with JBD2_MAGIC_NUMBER. */
-        jbd_debug(1, "JBD: Zeroing out journal blocks...\n");
-        for (i = 0; i < journal->j_maxlen; i++) {
-                err = jbd2_journal_bmap(journal, i, &blocknr);
-                if (err)
-                        return err;
-                bh = __getblk(journal->j_dev, blocknr, journal->j_blocksize);
-                lock_buffer(bh);
-                memset (bh->b_data, 0, journal->j_blocksize);
-                BUFFER_TRACE(bh, "marking dirty");
-                mark_buffer_dirty(bh);
-                BUFFER_TRACE(bh, "marking uptodate");
-                set_buffer_uptodate(bh);
-                unlock_buffer(bh);
-                __brelse(bh);
-        }
-
-        sync_blockdev(journal->j_dev);
-        jbd_debug(1, "JBD: journal cleared.\n");
-
-        /* OK, fill in the initial static fields in the new superblock */
-        sb = journal->j_superblock;
-
-        sb->s_header.h_magic     = cpu_to_be32(JBD2_MAGIC_NUMBER);
-        sb->s_header.h_blocktype = cpu_to_be32(JBD2_SUPERBLOCK_V2);
-
-        sb->s_blocksize = cpu_to_be32(journal->j_blocksize);
-        sb->s_maxlen    = cpu_to_be32(journal->j_maxlen);
-        sb->s_first     = cpu_to_be32(1);
-
-        journal->j_transaction_sequence = 1;
-
-        journal->j_flags &= ~JBD2_ABORT;
-        journal->j_format_version = 2;
-
-        return journal_reset(journal);
-}
-
-/**
  * void jbd2_journal_update_superblock() - Update journal sb on disk.
  * @journal: The journal to update.
  * @wait: Set to '0' if you don't want to wait for IO completion.
@@ -1491,7 +1439,9 @@ int jbd2_journal_destroy(journal_t *journal)
         spin_lock(&journal->j_list_lock);
         while (journal->j_checkpoint_transactions != NULL) {
                 spin_unlock(&journal->j_list_lock);
+                mutex_lock(&journal->j_checkpoint_mutex);
                 jbd2_log_do_checkpoint(journal);
+                mutex_unlock(&journal->j_checkpoint_mutex);
                 spin_lock(&journal->j_list_lock);
         }
 
diff --git a/fs/jbd2/transaction.c b/fs/jbd2/transaction.c
index 4f925a4f3d05..46b4e347ed7d 100644
--- a/fs/jbd2/transaction.c
+++ b/fs/jbd2/transaction.c
@@ -25,6 +25,7 @@
 #include <linux/timer.h>
 #include <linux/mm.h>
 #include <linux/highmem.h>
+#include <linux/hrtimer.h>
 
 static void __jbd2_journal_temp_unlink_buffer(struct journal_head *jh);
 
@@ -48,6 +49,7 @@ jbd2_get_transaction(journal_t *journal, transaction_t *transaction)
 {
         transaction->t_journal = journal;
         transaction->t_state = T_RUNNING;
+        transaction->t_start_time = ktime_get();
         transaction->t_tid = journal->j_transaction_sequence++;
         transaction->t_expires = jiffies + journal->j_commit_interval;
         spin_lock_init(&transaction->t_handle_lock);
@@ -1240,7 +1242,7 @@ int jbd2_journal_stop(handle_t *handle)
 {
         transaction_t *transaction = handle->h_transaction;
         journal_t *journal = transaction->t_journal;
-        int old_handle_count, err;
+        int err;
         pid_t pid;
 
         J_ASSERT(journal_current_handle() == handle);
@@ -1263,24 +1265,54 @@ int jbd2_journal_stop(handle_t *handle)
         /*
          * Implement synchronous transaction batching.  If the handle
          * was synchronous, don't force a commit immediately.  Let's
-         * yield and let another thread piggyback onto this transaction.
-         * Keep doing that while new threads continue to arrive.
-         * It doesn't cost much - we're about to run a commit and sleep
-         * on IO anyway.  Speeds up many-threaded, many-dir operations
-         * by 30x or more...
+         * yield and let another thread piggyback onto this
+         * transaction.  Keep doing that while new threads continue to
+         * arrive.  It doesn't cost much - we're about to run a commit
+         * and sleep on IO anyway.  Speeds up many-threaded, many-dir
+         * operations by 30x or more...
+         *
+         * We try and optimize the sleep time against what the
+         * underlying disk can do, instead of having a static sleep
+         * time.  This is useful for the case where our storage is so
+         * fast that it is more optimal to go ahead and force a flush
+         * and wait for the transaction to be committed than it is to
+         * wait for an arbitrary amount of time for new writers to
+         * join the transaction.  We achieve this by measuring how
+         * long it takes to commit a transaction, and compare it with
+         * how long this transaction has been running, and if run time
+         * < commit time then we sleep for the delta and commit.  This
+         * greatly helps super fast disks that would see slowdowns as
+         * more threads started doing fsyncs.
          *
-         * But don't do this if this process was the most recent one to
-         * perform a synchronous write.  We do this to detect the case where a
-         * single process is doing a stream of sync writes.  No point in waiting
-         * for joiners in that case.
+         * But don't do this if this process was the most recent one
+         * to perform a synchronous write.  We do this to detect the
+         * case where a single process is doing a stream of sync
+         * writes.  No point in waiting for joiners in that case.
          */
         pid = current->pid;
         if (handle->h_sync && journal->j_last_sync_writer != pid) {
+                u64 commit_time, trans_time;
+
                 journal->j_last_sync_writer = pid;
-                do {
-                        old_handle_count = transaction->t_handle_count;
-                        schedule_timeout_uninterruptible(1);
-                } while (old_handle_count != transaction->t_handle_count);
+
+                spin_lock(&journal->j_state_lock);
+                commit_time = journal->j_average_commit_time;
+                spin_unlock(&journal->j_state_lock);
+
+                trans_time = ktime_to_ns(ktime_sub(ktime_get(),
+                                                   transaction->t_start_time));
+
+                commit_time = max_t(u64, commit_time,
+                                    1000*journal->j_min_batch_time);
+                commit_time = min_t(u64, commit_time,
+                                    1000*journal->j_max_batch_time);
+
+                if (trans_time < commit_time) {
+                        ktime_t expires = ktime_add_ns(ktime_get(),
+                                                       commit_time);
+                        set_current_state(TASK_UNINTERRUPTIBLE);
+                        schedule_hrtimeout(&expires, HRTIMER_MODE_ABS);
+                }
         }
 
         current->journal_info = NULL;