add idl4k kernel firmware version 1.13.0.105

kernel/fs/jbd/Kconfig

@@ -0,0 +1,30 @@
+config JBD
+	tristate
+	help
+	  This is a generic journalling layer for block devices.  It is
+	  currently used by the ext3 file system, but it could also be
+	  used to add journal support to other file systems or block
+	  devices such as RAID or LVM.
+
+	  If you are using the ext3 file system, you need to say Y here.
+	  If you are not using ext3 then you will probably want to say N.
+
+	  To compile this device as a module, choose M here: the module will be
+	  called jbd.  If you are compiling ext3 into the kernel, you
+	  cannot compile this code as a module.
+
+config JBD_DEBUG
+	bool "JBD (ext3) debugging support"
+	depends on JBD && DEBUG_FS
+	help
+	  If you are using the ext3 journaled file system (or potentially any
+	  other file system/device using JBD), this option allows you to
+	  enable debugging output while the system is running, in order to
+	  help track down any problems you are having.  By default the
+	  debugging output will be turned off.
+
+	  If you select Y here, then you will be able to turn on debugging
+	  with "echo N > /sys/kernel/debug/jbd/jbd-debug", where N is a
+	  number between 1 and 5, the higher the number, the more debugging
+	  output is generated.  To turn debugging off again, do
+	  "echo 0 > /sys/kernel/debug/jbd/jbd-debug".

kernel/fs/jbd/Makefile

@@ -0,0 +1,7 @@
+#
+# Makefile for the linux journaling routines.
+#
+
+obj-$(CONFIG_JBD) += jbd.o
+
+jbd-objs := transaction.o commit.o recovery.o checkpoint.o revoke.o journal.o

kernel/fs/jbd/checkpoint.c

@@ -0,0 +1,755 @@
+/*
+ * linux/fs/jbd/checkpoint.c
+ *
+ * Written by Stephen C. Tweedie <sct@redhat.com>, 1999
+ *
+ * Copyright 1999 Red Hat Software --- All Rights Reserved
+ *
+ * This file is part of the Linux kernel and is made available under
+ * the terms of the GNU General Public License, version 2, or at your
+ * option, any later version, incorporated herein by reference.
+ *
+ * Checkpoint routines for the generic filesystem journaling code.
+ * Part of the ext2fs journaling system.
+ *
+ * Checkpointing is the process of ensuring that a section of the log is
+ * committed fully to disk, so that that portion of the log can be
+ * reused.
+ */
+
+#include <linux/time.h>
+#include <linux/fs.h>
+#include <linux/jbd.h>
+#include <linux/errno.h>
+#include <linux/slab.h>
+
+/*
+ * Unlink a buffer from a transaction checkpoint list.
+ *
+ * Called with j_list_lock held.
+ */
+static inline void __buffer_unlink_first(struct journal_head *jh)
+{
+	transaction_t *transaction = jh->b_cp_transaction;
+
+	jh->b_cpnext->b_cpprev = jh->b_cpprev;
+	jh->b_cpprev->b_cpnext = jh->b_cpnext;
+	if (transaction->t_checkpoint_list == jh) {
+		transaction->t_checkpoint_list = jh->b_cpnext;
+		if (transaction->t_checkpoint_list == jh)
+			transaction->t_checkpoint_list = NULL;
+	}
+}
+
+/*
+ * Unlink a buffer from a transaction checkpoint(io) list.
+ *
+ * Called with j_list_lock held.
+ */
+static inline void __buffer_unlink(struct journal_head *jh)
+{
+	transaction_t *transaction = jh->b_cp_transaction;
+
+	__buffer_unlink_first(jh);
+	if (transaction->t_checkpoint_io_list == jh) {
+		transaction->t_checkpoint_io_list = jh->b_cpnext;
+		if (transaction->t_checkpoint_io_list == jh)
+			transaction->t_checkpoint_io_list = NULL;
+	}
+}
+
+/*
+ * Move a buffer from the checkpoint list to the checkpoint io list
+ *
+ * Called with j_list_lock held
+ */
+static inline void __buffer_relink_io(struct journal_head *jh)
+{
+	transaction_t *transaction = jh->b_cp_transaction;
+
+	__buffer_unlink_first(jh);
+
+	if (!transaction->t_checkpoint_io_list) {
+		jh->b_cpnext = jh->b_cpprev = jh;
+	} else {
+		jh->b_cpnext = transaction->t_checkpoint_io_list;
+		jh->b_cpprev = transaction->t_checkpoint_io_list->b_cpprev;
+		jh->b_cpprev->b_cpnext = jh;
+		jh->b_cpnext->b_cpprev = jh;
+	}
+	transaction->t_checkpoint_io_list = jh;
+}
+
+/*
+ * Try to release a checkpointed buffer from its transaction.
+ * Returns 1 if we released it and 2 if we also released the
+ * whole transaction.
+ *
+ * Requires j_list_lock
+ * Called under jbd_lock_bh_state(jh2bh(jh)), and drops it
+ */
+static int __try_to_free_cp_buf(struct journal_head *jh)
+{
+	int ret = 0;
+	struct buffer_head *bh = jh2bh(jh);
+
+	if (jh->b_jlist == BJ_None && !buffer_locked(bh) &&
+	    !buffer_dirty(bh) && !buffer_write_io_error(bh)) {
+		JBUFFER_TRACE(jh, "remove from checkpoint list");
+		ret = __journal_remove_checkpoint(jh) + 1;
+		jbd_unlock_bh_state(bh);
+		journal_remove_journal_head(bh);
+		BUFFER_TRACE(bh, "release");
+		__brelse(bh);
+	} else {
+		jbd_unlock_bh_state(bh);
+	}
+	return ret;
+}
+
+/*
+ * __log_wait_for_space: wait until there is space in the journal.
+ *
+ * Called under j-state_lock *only*.  It will be unlocked if we have to wait
+ * for a checkpoint to free up some space in the log.
+ */
+void __log_wait_for_space(journal_t *journal)
+{
+	int nblocks, space_left;
+	assert_spin_locked(&journal->j_state_lock);
+
+	nblocks = jbd_space_needed(journal);
+	while (__log_space_left(journal) < nblocks) {
+		if (journal->j_flags & JFS_ABORT)
+			return;
+		spin_unlock(&journal->j_state_lock);
+		mutex_lock(&journal->j_checkpoint_mutex);
+
+		/*
+		 * Test again, another process may have checkpointed while we
+		 * were waiting for the checkpoint lock. If there are no
+		 * transactions ready to be checkpointed, try to recover
+		 * journal space by calling cleanup_journal_tail(), and if
+		 * that doesn't work, by waiting for the currently committing
+		 * transaction to complete.  If there is absolutely no way
+		 * to make progress, this is either a BUG or corrupted
+		 * filesystem, so abort the journal and leave a stack
+		 * trace for forensic evidence.
+		 */
+		spin_lock(&journal->j_state_lock);
+		spin_lock(&journal->j_list_lock);
+		nblocks = jbd_space_needed(journal);
+		space_left = __log_space_left(journal);
+		if (space_left < nblocks) {
+			int chkpt = journal->j_checkpoint_transactions != NULL;
+			tid_t tid = 0;
+
+			if (journal->j_committing_transaction)
+				tid = journal->j_committing_transaction->t_tid;
+			spin_unlock(&journal->j_list_lock);
+			spin_unlock(&journal->j_state_lock);
+			if (chkpt) {
+				log_do_checkpoint(journal);
+			} else if (cleanup_journal_tail(journal) == 0) {
+				/* We were able to recover space; yay! */
+				;
+			} else if (tid) {
+				log_wait_commit(journal, tid);
+			} else {
+				printk(KERN_ERR "%s: needed %d blocks and "
+				       "only had %d space available\n",
+				       __func__, nblocks, space_left);
+				printk(KERN_ERR "%s: no way to get more "
+				       "journal space\n", __func__);
+				WARN_ON(1);
+				journal_abort(journal, 0);
+			}
+			spin_lock(&journal->j_state_lock);
+		} else {
+			spin_unlock(&journal->j_list_lock);
+		}
+		mutex_unlock(&journal->j_checkpoint_mutex);
+	}
+}
+
+/*
+ * We were unable to perform jbd_trylock_bh_state() inside j_list_lock.
+ * The caller must restart a list walk.  Wait for someone else to run
+ * jbd_unlock_bh_state().
+ */
+static void jbd_sync_bh(journal_t *journal, struct buffer_head *bh)
+	__releases(journal->j_list_lock)
+{
+	get_bh(bh);
+	spin_unlock(&journal->j_list_lock);
+	jbd_lock_bh_state(bh);
+	jbd_unlock_bh_state(bh);
+	put_bh(bh);
+}
+
+/*
+ * Clean up transaction's list of buffers submitted for io.
+ * We wait for any pending IO to complete and remove any clean
+ * buffers. Note that we take the buffers in the opposite ordering
+ * from the one in which they were submitted for IO.
+ *
+ * Return 0 on success, and return <0 if some buffers have failed
+ * to be written out.
+ *
+ * Called with j_list_lock held.
+ */
+static int __wait_cp_io(journal_t *journal, transaction_t *transaction)
+{
+	struct journal_head *jh;
+	struct buffer_head *bh;
+	tid_t this_tid;
+	int released = 0;
+	int ret = 0;
+
+	this_tid = transaction->t_tid;
+restart:
+	/* Did somebody clean up the transaction in the meanwhile? */
+	if (journal->j_checkpoint_transactions != transaction ||
+			transaction->t_tid != this_tid)
+		return ret;
+	while (!released && transaction->t_checkpoint_io_list) {
+		jh = transaction->t_checkpoint_io_list;
+		bh = jh2bh(jh);
+		if (!jbd_trylock_bh_state(bh)) {
+			jbd_sync_bh(journal, bh);
+			spin_lock(&journal->j_list_lock);
+			goto restart;
+		}
+		if (buffer_locked(bh)) {
+			atomic_inc(&bh->b_count);
+			spin_unlock(&journal->j_list_lock);
+			jbd_unlock_bh_state(bh);
+			wait_on_buffer(bh);
+			/* the journal_head may have gone by now */
+			BUFFER_TRACE(bh, "brelse");
+			__brelse(bh);
+			spin_lock(&journal->j_list_lock);
+			goto restart;
+		}
+		if (unlikely(buffer_write_io_error(bh)))
+			ret = -EIO;
+
+		/*
+		 * Now in whatever state the buffer currently is, we know that
+		 * it has been written out and so we can drop it from the list
+		 */
+		released = __journal_remove_checkpoint(jh);
+		jbd_unlock_bh_state(bh);
+		journal_remove_journal_head(bh);
+		__brelse(bh);
+	}
+
+	return ret;
+}
+
+#define NR_BATCH	64
+
+static void
+__flush_batch(journal_t *journal, struct buffer_head **bhs, int *batch_count)
+{
+	int i;
+
+	ll_rw_block(SWRITE, *batch_count, bhs);
+	for (i = 0; i < *batch_count; i++) {
+		struct buffer_head *bh = bhs[i];
+		clear_buffer_jwrite(bh);
+		BUFFER_TRACE(bh, "brelse");
+		__brelse(bh);
+	}
+	*batch_count = 0;
+}
+
+/*
+ * Try to flush one buffer from the checkpoint list to disk.
+ *
+ * Return 1 if something happened which requires us to abort the current
+ * scan of the checkpoint list.  Return <0 if the buffer has failed to
+ * be written out.
+ *
+ * Called with j_list_lock held and drops it if 1 is returned
+ * 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)
+{
+	struct buffer_head *bh = jh2bh(jh);
+	int ret = 0;
+
+	if (buffer_locked(bh)) {
+		atomic_inc(&bh->b_count);
+		spin_unlock(&journal->j_list_lock);
+		jbd_unlock_bh_state(bh);
+		wait_on_buffer(bh);
+		/* the journal_head may have gone by now */
+		BUFFER_TRACE(bh, "brelse");
+		__brelse(bh);
+		ret = 1;
+	} else if (jh->b_transaction != NULL) {
+		transaction_t *t = jh->b_transaction;
+		tid_t tid = t->t_tid;
+
+		spin_unlock(&journal->j_list_lock);
+		jbd_unlock_bh_state(bh);
+		log_start_commit(journal, tid);
+		log_wait_commit(journal, tid);
+		ret = 1;
+	} else if (!buffer_dirty(bh)) {
+		ret = 1;
+		if (unlikely(buffer_write_io_error(bh)))
+			ret = -EIO;
+		J_ASSERT_JH(jh, !buffer_jbddirty(bh));
+		BUFFER_TRACE(bh, "remove from checkpoint");
+		__journal_remove_checkpoint(jh);
+		spin_unlock(&journal->j_list_lock);
+		jbd_unlock_bh_state(bh);
+		journal_remove_journal_head(bh);
+		__brelse(bh);
+	} else {
+		/*
+		 * Important: we are about to write the buffer, and
+		 * possibly block, while still holding the journal lock.
+		 * We cannot afford to let the transaction logic start
+		 * messing around with this buffer before we write it to
+		 * disk, as that would break recoverability.
+		 */
+		BUFFER_TRACE(bh, "queue");
+		get_bh(bh);
+		J_ASSERT_BH(bh, !buffer_jwrite(bh));
+		set_buffer_jwrite(bh);
+		bhs[*batch_count] = bh;
+		__buffer_relink_io(jh);
+		jbd_unlock_bh_state(bh);
+		(*batch_count)++;
+		if (*batch_count == NR_BATCH) {
+			spin_unlock(&journal->j_list_lock);
+			__flush_batch(journal, bhs, batch_count);
+			ret = 1;
+		}
+	}
+	return ret;
+}
+
+/*
+ * Perform an actual checkpoint. We take the first transaction on the
+ * list of transactions to be checkpointed and send all its buffers
+ * to disk. We submit larger chunks of data at once.
+ *
+ * The journal should be locked before calling this function.
+ * Called with j_checkpoint_mutex held.
+ */
+int log_do_checkpoint(journal_t *journal)
+{
+	transaction_t *transaction;
+	tid_t this_tid;
+	int result;
+
+	jbd_debug(1, "Start checkpoint\n");
+
+	/*
+	 * First thing: if there are any transactions in the log which
+	 * don't need checkpointing, just eliminate them from the
+	 * journal straight away.
+	 */
+	result = cleanup_journal_tail(journal);
+	jbd_debug(1, "cleanup_journal_tail returned %d\n", result);
+	if (result <= 0)
+		return result;
+
+	/*
+	 * OK, we need to start writing disk blocks.  Take one transaction
+	 * and write it.
+	 */
+	result = 0;
+	spin_lock(&journal->j_list_lock);
+	if (!journal->j_checkpoint_transactions)
+		goto out;
+	transaction = journal->j_checkpoint_transactions;
+	this_tid = transaction->t_tid;
+restart:
+	/*
+	 * If someone cleaned up this transaction while we slept, we're
+	 * done (maybe it's a new transaction, but it fell at the same
+	 * address).
+	 */
+	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;
+
+		while (!retry && transaction->t_checkpoint_list) {
+			struct buffer_head *bh;
+
+			jh = transaction->t_checkpoint_list;
+			bh = jh2bh(jh);
+			if (!jbd_trylock_bh_state(bh)) {
+				jbd_sync_bh(journal, bh);
+				retry = 1;
+				break;
+			}
+			retry = __process_buffer(journal, jh, bhs,&batch_count);
+			if (retry < 0 && !result)
+				result = retry;
+			if (!retry && (need_resched() ||
+				spin_needbreak(&journal->j_list_lock))) {
+				spin_unlock(&journal->j_list_lock);
+				retry = 1;
+				break;
+			}
+		}
+
+		if (batch_count) {
+			if (!retry) {
+				spin_unlock(&journal->j_list_lock);
+				retry = 1;
+			}
+			__flush_batch(journal, bhs, &batch_count);
+		}
+
+		if (retry) {
+			spin_lock(&journal->j_list_lock);
+			goto restart;
+		}
+		/*
+		 * Now we have cleaned up the first transaction's checkpoint
+		 * list. Let's clean up the second one
+		 */
+		err = __wait_cp_io(journal, transaction);
+		if (!result)
+			result = err;
+	}
+out:
+	spin_unlock(&journal->j_list_lock);
+	if (result < 0)
+		journal_abort(journal, result);
+	else
+		result = cleanup_journal_tail(journal);
+
+	return (result < 0) ? result : 0;
+}
+
+/*
+ * Check the list of checkpoint transactions for the journal to see if
+ * we have already got rid of any since the last update of the log tail
+ * in the journal superblock.  If so, we can instantly roll the
+ * superblock forward to remove those transactions from the log.
+ *
+ * Return <0 on error, 0 on success, 1 if there was nothing to clean up.
+ *
+ * Called with the journal lock held.
+ *
+ * This is the only part of the journaling code which really needs to be
+ * aware of transaction aborts.  Checkpointing involves writing to the
+ * main filesystem area rather than to the journal, so it can proceed
+ * even in abort state, but we must not update the super block if
+ * checkpointing may have failed.  Otherwise, we would lose some metadata
+ * buffers which should be written-back to the filesystem.
+ */
+
+int cleanup_journal_tail(journal_t *journal)
+{
+	transaction_t * transaction;
+	tid_t		first_tid;
+	unsigned int	blocknr, freed;
+
+	if (is_journal_aborted(journal))
+		return 1;
+
+	/* OK, work out the oldest transaction remaining in the log, and
+	 * the log block it starts at.
+	 *
+	 * If the log is now empty, we need to work out which is the
+	 * next transaction ID we will write, and where it will
+	 * start. */
+
+	spin_lock(&journal->j_state_lock);
+	spin_lock(&journal->j_list_lock);
+	transaction = journal->j_checkpoint_transactions;
+	if (transaction) {
+		first_tid = transaction->t_tid;
+		blocknr = transaction->t_log_start;
+	} else if ((transaction = journal->j_committing_transaction) != NULL) {
+		first_tid = transaction->t_tid;
+		blocknr = transaction->t_log_start;
+	} else if ((transaction = journal->j_running_transaction) != NULL) {
+		first_tid = transaction->t_tid;
+		blocknr = journal->j_head;
+	} else {
+		first_tid = journal->j_transaction_sequence;
+		blocknr = journal->j_head;
+	}
+	spin_unlock(&journal->j_list_lock);
+	J_ASSERT(blocknr != 0);
+
+	/* If the oldest pinned transaction is at the tail of the log
+           already then there's not much we can do right now. */
+	if (journal->j_tail_sequence == first_tid) {
+		spin_unlock(&journal->j_state_lock);
+		return 1;
+	}
+
+	/* OK, update the superblock to recover the freed space.
+	 * Physical blocks come first: have we wrapped beyond the end of
+	 * the log?  */
+	freed = blocknr - journal->j_tail;
+	if (blocknr < journal->j_tail)
+		freed = freed + journal->j_last - journal->j_first;
+
+	jbd_debug(1,
+		  "Cleaning journal tail from %d to %d (offset %u), "
+		  "freeing %u\n",
+		  journal->j_tail_sequence, first_tid, blocknr, freed);
+
+	journal->j_free += freed;
+	journal->j_tail_sequence = first_tid;
+	journal->j_tail = blocknr;
+	spin_unlock(&journal->j_state_lock);
+	if (!(journal->j_flags & JFS_ABORT))
+		journal_update_superblock(journal, 1);
+	return 0;
+}
+
+
+/* Checkpoint list management */
+
+/*
+ * journal_clean_one_cp_list
+ *
+ * Find all the written-back checkpoint buffers in the given list and release them.
+ *
+ * Called with the journal locked.
+ * Called with j_list_lock held.
+ * Returns number of bufers reaped (for debug)
+ */
+
+static int journal_clean_one_cp_list(struct journal_head *jh, int *released)
+{
+	struct journal_head *last_jh;
+	struct journal_head *next_jh = jh;
+	int ret, freed = 0;
+
+	*released = 0;
+	if (!jh)
+		return 0;
+
+	last_jh = jh->b_cpprev;
+	do {
+		jh = next_jh;
+		next_jh = jh->b_cpnext;
+		/* Use trylock because of the ranking */
+		if (jbd_trylock_bh_state(jh2bh(jh))) {
+			ret = __try_to_free_cp_buf(jh);
+			if (ret) {
+				freed++;
+				if (ret == 2) {
+					*released = 1;
+					return freed;
+				}
+			}
+		}
+		/*
+		 * This function only frees up some memory
+		 * if possible so we dont have an obligation
+		 * to finish processing. Bail out if preemption
+		 * requested:
+		 */
+		if (need_resched())
+			return freed;
+	} while (jh != last_jh);
+
+	return freed;
+}
+
+/*
+ * journal_clean_checkpoint_list
+ *
+ * Find all the written-back checkpoint buffers in the journal and release them.
+ *
+ * Called with the journal locked.
+ * Called with j_list_lock held.
+ * Returns number of buffers reaped (for debug)
+ */
+
+int __journal_clean_checkpoint_list(journal_t *journal)
+{
+	transaction_t *transaction, *last_transaction, *next_transaction;
+	int ret = 0;
+	int released;
+
+	transaction = journal->j_checkpoint_transactions;
+	if (!transaction)
+		goto out;
+
+	last_transaction = transaction->t_cpprev;
+	next_transaction = transaction;
+	do {
+		transaction = next_transaction;
+		next_transaction = transaction->t_cpnext;
+		ret += journal_clean_one_cp_list(transaction->
+				t_checkpoint_list, &released);
+		/*
+		 * This function only frees up some memory if possible so we
+		 * dont have an obligation to finish processing. Bail out if
+		 * preemption requested:
+		 */
+		if (need_resched())
+			goto out;
+		if (released)
+			continue;
+		/*
+		 * It is essential that we are as careful as in the case of
+		 * t_checkpoint_list with removing the buffer from the list as
+		 * we can possibly see not yet submitted buffers on io_list
+		 */
+		ret += journal_clean_one_cp_list(transaction->
+				t_checkpoint_io_list, &released);
+		if (need_resched())
+			goto out;
+	} while (transaction != last_transaction);
+out:
+	return ret;
+}
+
+/*
+ * journal_remove_checkpoint: called after a buffer has been committed
+ * to disk (either by being write-back flushed to disk, or being
+ * committed to the log).
+ *
+ * We cannot safely clean a transaction out of the log until all of the
+ * buffer updates committed in that transaction have safely been stored
+ * elsewhere on disk.  To achieve this, all of the buffers in a
+ * transaction need to be maintained on the transaction's checkpoint
+ * lists until they have been rewritten, at which point this function is
+ * called to remove the buffer from the existing transaction's
+ * checkpoint lists.
+ *
+ * The function returns 1 if it frees the transaction, 0 otherwise.
+ *
+ * This function is called with the journal locked.
+ * This function is called with j_list_lock held.
+ * This function is called with jbd_lock_bh_state(jh2bh(jh))
+ */
+
+int __journal_remove_checkpoint(struct journal_head *jh)
+{
+	transaction_t *transaction;
+	journal_t *journal;
+	int ret = 0;
+
+	JBUFFER_TRACE(jh, "entry");
+
+	if ((transaction = jh->b_cp_transaction) == NULL) {
+		JBUFFER_TRACE(jh, "not on transaction");
+		goto out;
+	}
+	journal = transaction->t_journal;
+
+	__buffer_unlink(jh);
+	jh->b_cp_transaction = NULL;
+
+	if (transaction->t_checkpoint_list != NULL ||
+	    transaction->t_checkpoint_io_list != NULL)
+		goto out;
+	JBUFFER_TRACE(jh, "transaction has no more buffers");
+
+	/*
+	 * There is one special case to worry about: if we have just pulled the
+	 * buffer off a running or committing transaction's checkpoing list,
+	 * then even if the checkpoint list is empty, the transaction obviously
+	 * cannot be dropped!
+	 *
+	 * The locking here around t_state is a bit sleazy.
+	 * See the comment at the end of journal_commit_transaction().
+	 */
+	if (transaction->t_state != T_FINISHED) {
+		JBUFFER_TRACE(jh, "belongs to running/committing transaction");
+		goto out;
+	}
+
+	/* OK, that was the last buffer for the transaction: we can now
+	   safely remove this transaction from the log */
+
+	__journal_drop_transaction(journal, transaction);
+
+	/* Just in case anybody was waiting for more transactions to be
+           checkpointed... */
+	wake_up(&journal->j_wait_logspace);
+	ret = 1;
+out:
+	JBUFFER_TRACE(jh, "exit");
+	return ret;
+}
+
+/*
+ * journal_insert_checkpoint: put a committed buffer onto a checkpoint
+ * list so that we know when it is safe to clean the transaction out of
+ * the log.
+ *
+ * Called with the journal locked.
+ * Called with j_list_lock held.
+ */
+void __journal_insert_checkpoint(struct journal_head *jh,
+			       transaction_t *transaction)
+{
+	JBUFFER_TRACE(jh, "entry");
+	J_ASSERT_JH(jh, buffer_dirty(jh2bh(jh)) || buffer_jbddirty(jh2bh(jh)));
+	J_ASSERT_JH(jh, jh->b_cp_transaction == NULL);
+
+	jh->b_cp_transaction = transaction;
+
+	if (!transaction->t_checkpoint_list) {
+		jh->b_cpnext = jh->b_cpprev = jh;
+	} else {
+		jh->b_cpnext = transaction->t_checkpoint_list;
+		jh->b_cpprev = transaction->t_checkpoint_list->b_cpprev;
+		jh->b_cpprev->b_cpnext = jh;
+		jh->b_cpnext->b_cpprev = jh;
+	}
+	transaction->t_checkpoint_list = jh;
+}
+
+/*
+ * We've finished with this transaction structure: adios...
+ *
+ * The transaction must have no links except for the checkpoint by this
+ * point.
+ *
+ * Called with the journal locked.
+ * Called with j_list_lock held.
+ */
+
+void __journal_drop_transaction(journal_t *journal, transaction_t *transaction)
+{
+	assert_spin_locked(&journal->j_list_lock);
+	if (transaction->t_cpnext) {
+		transaction->t_cpnext->t_cpprev = transaction->t_cpprev;
+		transaction->t_cpprev->t_cpnext = transaction->t_cpnext;
+		if (journal->j_checkpoint_transactions == transaction)
+			journal->j_checkpoint_transactions =
+				transaction->t_cpnext;
+		if (journal->j_checkpoint_transactions == transaction)
+			journal->j_checkpoint_transactions = NULL;
+	}
+
+	J_ASSERT(transaction->t_state == T_FINISHED);
+	J_ASSERT(transaction->t_buffers == NULL);
+	J_ASSERT(transaction->t_sync_datalist == NULL);
+	J_ASSERT(transaction->t_forget == NULL);
+	J_ASSERT(transaction->t_iobuf_list == NULL);
+	J_ASSERT(transaction->t_shadow_list == NULL);
+	J_ASSERT(transaction->t_log_list == NULL);
+	J_ASSERT(transaction->t_checkpoint_list == NULL);
+	J_ASSERT(transaction->t_checkpoint_io_list == NULL);
+	J_ASSERT(transaction->t_updates == 0);
+	J_ASSERT(journal->j_committing_transaction != transaction);
+	J_ASSERT(journal->j_running_transaction != transaction);
+
+	jbd_debug(1, "Dropping transaction %d, all done\n", transaction->t_tid);
+	kfree(transaction);
+}

kernel/fs/jbd/commit.c

@@ -0,0 +1,977 @@
+/*
+ * linux/fs/jbd/commit.c
+ *
+ * Written by Stephen C. Tweedie <sct@redhat.com>, 1998
+ *
+ * Copyright 1998 Red Hat corp --- All Rights Reserved
+ *
+ * This file is part of the Linux kernel and is made available under
+ * the terms of the GNU General Public License, version 2, or at your
+ * option, any later version, incorporated herein by reference.
+ *
+ * Journal commit routines for the generic filesystem journaling code;
+ * part of the ext2fs journaling system.
+ */
+
+#include <linux/time.h>
+#include <linux/fs.h>
+#include <linux/jbd.h>
+#include <linux/errno.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
+#include <linux/pagemap.h>
+#include <linux/bio.h>
+
+/*
+ * Default IO end handler for temporary BJ_IO buffer_heads.
+ */
+static void journal_end_buffer_io_sync(struct buffer_head *bh, int uptodate)
+{
+	BUFFER_TRACE(bh, "");
+	if (uptodate)
+		set_buffer_uptodate(bh);
+	else
+		clear_buffer_uptodate(bh);
+	unlock_buffer(bh);
+}
+
+/*
+ * When an ext3-ordered file is truncated, it is possible that many pages are
+ * not successfully freed, because they are attached to a committing transaction.
+ * After the transaction commits, these pages are left on the LRU, with no
+ * ->mapping, and with attached buffers.  These pages are trivially reclaimable
+ * by the VM, but their apparent absence upsets the VM accounting, and it makes
+ * the numbers in /proc/meminfo look odd.
+ *
+ * So here, we have a buffer which has just come off the forget list.  Look to
+ * see if we can strip all buffers from the backing page.
+ *
+ * Called under journal->j_list_lock.  The caller provided us with a ref
+ * against the buffer, and we drop that here.
+ */
+static void release_buffer_page(struct buffer_head *bh)
+{
+	struct page *page;
+
+	if (buffer_dirty(bh))
+		goto nope;
+	if (atomic_read(&bh->b_count) != 1)
+		goto nope;
+	page = bh->b_page;
+	if (!page)
+		goto nope;
+	if (page->mapping)
+		goto nope;
+
+	/* OK, it's a truncated page */
+	if (!trylock_page(page))
+		goto nope;
+
+	page_cache_get(page);
+	__brelse(bh);
+	try_to_free_buffers(page);
+	unlock_page(page);
+	page_cache_release(page);
+	return;
+
+nope:
+	__brelse(bh);
+}
+
+/*
+ * Decrement reference counter for data buffer. If it has been marked
+ * 'BH_Freed', release it and the page to which it belongs if possible.
+ */
+static void release_data_buffer(struct buffer_head *bh)
+{
+	if (buffer_freed(bh)) {
+		clear_buffer_freed(bh);
+		release_buffer_page(bh);
+	} else
+		put_bh(bh);
+}
+
+/*
+ * Try to acquire jbd_lock_bh_state() against the buffer, when j_list_lock is
+ * held.  For ranking reasons we must trylock.  If we lose, schedule away and
+ * return 0.  j_list_lock is dropped in this case.
+ */
+static int inverted_lock(journal_t *journal, struct buffer_head *bh)
+{
+	if (!jbd_trylock_bh_state(bh)) {
+		spin_unlock(&journal->j_list_lock);
+		schedule();
+		return 0;
+	}
+	return 1;
+}
+
+/* Done it all: now write the commit record.  We should have
+ * cleaned up our previous buffers by now, so if we are in abort
+ * mode we can now just skip the rest of the journal write
+ * entirely.
+ *
+ * Returns 1 if the journal needs to be aborted or 0 on success
+ */
+static int journal_write_commit_record(journal_t *journal,
+					transaction_t *commit_transaction)
+{
+	struct journal_head *descriptor;
+	struct buffer_head *bh;
+	journal_header_t *header;
+	int ret;
+	int barrier_done = 0;
+
+	if (is_journal_aborted(journal))
+		return 0;
+
+	descriptor = journal_get_descriptor_buffer(journal);
+	if (!descriptor)
+		return 1;
+
+	bh = jh2bh(descriptor);
+
+	header = (journal_header_t *)(bh->b_data);
+	header->h_magic = cpu_to_be32(JFS_MAGIC_NUMBER);
+	header->h_blocktype = cpu_to_be32(JFS_COMMIT_BLOCK);
+	header->h_sequence = cpu_to_be32(commit_transaction->t_tid);
+
+	JBUFFER_TRACE(descriptor, "write commit block");
+	set_buffer_dirty(bh);
+	if (journal->j_flags & JFS_BARRIER) {
+		set_buffer_ordered(bh);
+		barrier_done = 1;
+	}
+	ret = sync_dirty_buffer(bh);
+	if (barrier_done)
+		clear_buffer_ordered(bh);
+	/* is it possible for another commit to fail at roughly
+	 * the same time as this one?  If so, we don't want to
+	 * trust the barrier flag in the super, but instead want
+	 * to remember if we sent a barrier request
+	 */
+	if (ret == -EOPNOTSUPP && barrier_done) {
+		char b[BDEVNAME_SIZE];
+
+		printk(KERN_WARNING
+			"JBD: barrier-based sync failed on %s - "
+			"disabling barriers\n",
+			bdevname(journal->j_dev, b));
+		spin_lock(&journal->j_state_lock);
+		journal->j_flags &= ~JFS_BARRIER;
+		spin_unlock(&journal->j_state_lock);
+
+		/* And try again, without the barrier */
+		set_buffer_uptodate(bh);
+		set_buffer_dirty(bh);
+		ret = sync_dirty_buffer(bh);
+	}
+	put_bh(bh);		/* One for getblk() */
+	journal_put_journal_head(descriptor);
+
+	return (ret == -EIO);
+}
+
+static void journal_do_submit_data(struct buffer_head **wbuf, int bufs,
+				   int write_op)
+{
+	int i;
+
+	for (i = 0; i < bufs; i++) {
+		wbuf[i]->b_end_io = end_buffer_write_sync;
+		/* We use-up our safety reference in submit_bh() */
+		submit_bh(write_op, wbuf[i]);
+	}
+}
+
+/*
+ *  Submit all the data buffers to disk
+ */
+static int journal_submit_data_buffers(journal_t *journal,
+				       transaction_t *commit_transaction,
+				       int write_op)
+{
+	struct journal_head *jh;
+	struct buffer_head *bh;
+	int locked;
+	int bufs = 0;
+	struct buffer_head **wbuf = journal->j_wbuf;
+	int err = 0;
+
+	/*
+	 * Whenever we unlock the journal and sleep, things can get added
+	 * onto ->t_sync_datalist, so we have to keep looping back to
+	 * write_out_data until we *know* that the list is empty.
+	 *
+	 * Cleanup any flushed data buffers from the data list.  Even in
+	 * abort mode, we want to flush this out as soon as possible.
+	 */
+write_out_data:
+	cond_resched();
+	spin_lock(&journal->j_list_lock);
+
+	while (commit_transaction->t_sync_datalist) {
+		jh = commit_transaction->t_sync_datalist;
+		bh = jh2bh(jh);
+		locked = 0;
+
+		/* Get reference just to make sure buffer does not disappear
+		 * when we are forced to drop various locks */
+		get_bh(bh);
+		/* If the buffer is dirty, we need to submit IO and hence
+		 * we need the buffer lock. We try to lock the buffer without
+		 * blocking. If we fail, we need to drop j_list_lock and do
+		 * blocking lock_buffer().
+		 */
+		if (buffer_dirty(bh)) {
+			if (!trylock_buffer(bh)) {
+				BUFFER_TRACE(bh, "needs blocking lock");
+				spin_unlock(&journal->j_list_lock);
+				/* Write out all data to prevent deadlocks */
+				journal_do_submit_data(wbuf, bufs, write_op);
+				bufs = 0;
+				lock_buffer(bh);
+				spin_lock(&journal->j_list_lock);
+			}
+			locked = 1;
+		}
+		/* We have to get bh_state lock. Again out of order, sigh. */
+		if (!inverted_lock(journal, bh)) {
+			jbd_lock_bh_state(bh);
+			spin_lock(&journal->j_list_lock);
+		}
+		/* Someone already cleaned up the buffer? */
+		if (!buffer_jbd(bh) || bh2jh(bh) != jh
+			|| jh->b_transaction != commit_transaction
+			|| jh->b_jlist != BJ_SyncData) {
+			jbd_unlock_bh_state(bh);
+			if (locked)
+				unlock_buffer(bh);
+			BUFFER_TRACE(bh, "already cleaned up");
+			release_data_buffer(bh);
+			continue;
+		}
+		if (locked && test_clear_buffer_dirty(bh)) {
+			BUFFER_TRACE(bh, "needs writeout, adding to array");
+			wbuf[bufs++] = bh;
+			__journal_file_buffer(jh, commit_transaction,
+						BJ_Locked);
+			jbd_unlock_bh_state(bh);
+			if (bufs == journal->j_wbufsize) {
+				spin_unlock(&journal->j_list_lock);
+				journal_do_submit_data(wbuf, bufs, write_op);
+				bufs = 0;
+				goto write_out_data;
+			}
+		} else if (!locked && buffer_locked(bh)) {
+			__journal_file_buffer(jh, commit_transaction,
+						BJ_Locked);
+			jbd_unlock_bh_state(bh);
+			put_bh(bh);
+		} else {
+			BUFFER_TRACE(bh, "writeout complete: unfile");
+			if (unlikely(!buffer_uptodate(bh)))
+				err = -EIO;
+			__journal_unfile_buffer(jh);
+			jbd_unlock_bh_state(bh);
+			if (locked)
+				unlock_buffer(bh);
+			journal_remove_journal_head(bh);
+			/* One for our safety reference, other for
+			 * journal_remove_journal_head() */
+			put_bh(bh);
+			release_data_buffer(bh);
+		}
+
+		if (need_resched() || spin_needbreak(&journal->j_list_lock)) {
+			spin_unlock(&journal->j_list_lock);
+			goto write_out_data;
+		}
+	}
+	spin_unlock(&journal->j_list_lock);
+	journal_do_submit_data(wbuf, bufs, write_op);
+
+	return err;
+}
+
+/*
+ * journal_commit_transaction
+ *
+ * The primary function for committing a transaction to the log.  This
+ * function is called by the journal thread to begin a complete commit.
+ */
+void journal_commit_transaction(journal_t *journal)
+{
+	transaction_t *commit_transaction;
+	struct journal_head *jh, *new_jh, *descriptor;
+	struct buffer_head **wbuf = journal->j_wbuf;
+	int bufs;
+	int flags;
+	int err;
+	unsigned int 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 write_op = WRITE;
+
+	/*
+	 * First job: lock down the current transaction and wait for
+	 * all outstanding updates to complete.
+	 */
+
+#ifdef COMMIT_STATS
+	spin_lock(&journal->j_list_lock);
+	summarise_journal_usage(journal);
+	spin_unlock(&journal->j_list_lock);
+#endif
+
+	/* Do we need to erase the effects of a prior journal_flush? */
+	if (journal->j_flags & JFS_FLUSHED) {
+		jbd_debug(3, "super block updated\n");
+		journal_update_superblock(journal, 1);
+	} else {
+		jbd_debug(3, "superblock not updated\n");
+	}
+
+	J_ASSERT(journal->j_running_transaction != NULL);
+	J_ASSERT(journal->j_committing_transaction == NULL);
+
+	commit_transaction = journal->j_running_transaction;
+	J_ASSERT(commit_transaction->t_state == T_RUNNING);
+
+	jbd_debug(1, "JBD: starting commit of transaction %d\n",
+			commit_transaction->t_tid);
+
+	spin_lock(&journal->j_state_lock);
+	commit_transaction->t_state = T_LOCKED;
+
+	/*
+	 * Use plugged writes here, since we want to submit several before
+	 * we unplug the device. We don't do explicit unplugging in here,
+	 * instead we rely on sync_buffer() doing the unplug for us.
+	 */
+	if (commit_transaction->t_synchronous_commit)
+		write_op = WRITE_SYNC_PLUG;
+	spin_lock(&commit_transaction->t_handle_lock);
+	while (commit_transaction->t_updates) {
+		DEFINE_WAIT(wait);
+
+		prepare_to_wait(&journal->j_wait_updates, &wait,
+					TASK_UNINTERRUPTIBLE);
+		if (commit_transaction->t_updates) {
+			spin_unlock(&commit_transaction->t_handle_lock);
+			spin_unlock(&journal->j_state_lock);
+			schedule();
+			spin_lock(&journal->j_state_lock);
+			spin_lock(&commit_transaction->t_handle_lock);
+		}
+		finish_wait(&journal->j_wait_updates, &wait);
+	}
+	spin_unlock(&commit_transaction->t_handle_lock);
+
+	J_ASSERT (commit_transaction->t_outstanding_credits <=
+			journal->j_max_transaction_buffers);
+
+	/*
+	 * First thing we are allowed to do is to discard any remaining
+	 * BJ_Reserved buffers.  Note, it is _not_ permissible to assume
+	 * that there are no such buffers: if a large filesystem
+	 * operation like a truncate needs to split itself over multiple
+	 * transactions, then it may try to do a journal_restart() while
+	 * there are still BJ_Reserved buffers outstanding.  These must
+	 * be released cleanly from the current transaction.
+	 *
+	 * In this case, the filesystem must still reserve write access
+	 * again before modifying the buffer in the new transaction, but
+	 * we do not require it to remember exactly which old buffers it
+	 * has reserved.  This is consistent with the existing behaviour
+	 * that multiple journal_get_write_access() calls to the same
+	 * buffer are perfectly permissable.
+	 */
+	while (commit_transaction->t_reserved_list) {
+		jh = commit_transaction->t_reserved_list;
+		JBUFFER_TRACE(jh, "reserved, unused: refile");
+		/*
+		 * A journal_get_undo_access()+journal_release_buffer() may
+		 * leave undo-committed data.
+		 */
+		if (jh->b_committed_data) {
+			struct buffer_head *bh = jh2bh(jh);
+
+			jbd_lock_bh_state(bh);
+			jbd_free(jh->b_committed_data, bh->b_size);
+			jh->b_committed_data = NULL;
+			jbd_unlock_bh_state(bh);
+		}
+		journal_refile_buffer(journal, jh);
+	}
+
+	/*
+	 * Now try to drop any written-back buffers from the journal's
+	 * checkpoint lists.  We do this *before* commit because it potentially
+	 * frees some memory
+	 */
+	spin_lock(&journal->j_list_lock);
+	__journal_clean_checkpoint_list(journal);
+	spin_unlock(&journal->j_list_lock);
+
+	jbd_debug (3, "JBD: commit phase 1\n");
+
+	/*
+	 * Switch to a new revoke table.
+	 */
+	journal_switch_revoke_table(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);
+
+	jbd_debug (3, "JBD: commit phase 2\n");
+
+	/*
+	 * Now start flushing things to disk, in the order they appear
+	 * on the transaction lists.  Data blocks go first.
+	 */
+	err = journal_submit_data_buffers(journal, commit_transaction,
+					  write_op);
+
+	/*
+	 * Wait for all previously submitted IO to complete.
+	 */
+	spin_lock(&journal->j_list_lock);
+	while (commit_transaction->t_locked_list) {
+		struct buffer_head *bh;
+
+		jh = commit_transaction->t_locked_list->b_tprev;
+		bh = jh2bh(jh);
+		get_bh(bh);
+		if (buffer_locked(bh)) {
+			spin_unlock(&journal->j_list_lock);
+			wait_on_buffer(bh);
+			spin_lock(&journal->j_list_lock);
+		}
+		if (unlikely(!buffer_uptodate(bh))) {
+			if (!trylock_page(bh->b_page)) {
+				spin_unlock(&journal->j_list_lock);
+				lock_page(bh->b_page);
+				spin_lock(&journal->j_list_lock);
+			}
+			if (bh->b_page->mapping)
+				set_bit(AS_EIO, &bh->b_page->mapping->flags);
+
+			unlock_page(bh->b_page);
+			SetPageError(bh->b_page);
+			err = -EIO;
+		}
+		if (!inverted_lock(journal, bh)) {
+			put_bh(bh);
+			spin_lock(&journal->j_list_lock);
+			continue;
+		}
+		if (buffer_jbd(bh) && bh2jh(bh) == jh &&
+		    jh->b_transaction == commit_transaction &&
+		    jh->b_jlist == BJ_Locked) {
+			__journal_unfile_buffer(jh);
+			jbd_unlock_bh_state(bh);
+			journal_remove_journal_head(bh);
+			put_bh(bh);
+		} else {
+			jbd_unlock_bh_state(bh);
+		}
+		release_data_buffer(bh);
+		cond_resched_lock(&journal->j_list_lock);
+	}
+	spin_unlock(&journal->j_list_lock);
+
+	if (err) {
+		char b[BDEVNAME_SIZE];
+
+		printk(KERN_WARNING
+			"JBD: Detected IO errors while flushing file data "
+			"on %s\n", bdevname(journal->j_fs_dev, b));
+		if (journal->j_flags & JFS_ABORT_ON_SYNCDATA_ERR)
+			journal_abort(journal, err);
+		err = 0;
+	}
+
+	journal_write_revoke_records(journal, commit_transaction, write_op);
+
+	/*
+	 * If we found any dirty or locked buffers, then we should have
+	 * looped back up to the write_out_data label.  If there weren't
+	 * any then journal_clean_data_list should have wiped the list
+	 * clean by now, so check that it is in fact empty.
+	 */
+	J_ASSERT (commit_transaction->t_sync_datalist == NULL);
+
+	jbd_debug (3, "JBD: commit phase 3\n");
+
+	/*
+	 * Way to go: we have now written out all of the data for a
+	 * transaction!  Now comes the tricky part: we need to write out
+	 * metadata.  Loop over the transaction's entire buffer list:
+	 */
+	spin_lock(&journal->j_state_lock);
+	commit_transaction->t_state = T_COMMIT;
+	spin_unlock(&journal->j_state_lock);
+
+	J_ASSERT(commit_transaction->t_nr_buffers <=
+		 commit_transaction->t_outstanding_credits);
+
+	descriptor = NULL;
+	bufs = 0;
+	while (commit_transaction->t_buffers) {
+
+		/* Find the next buffer to be journaled... */
+
+		jh = commit_transaction->t_buffers;
+
+		/* If we're in abort mode, we just un-journal the buffer and
+		   release it. */
+
+		if (is_journal_aborted(journal)) {
+			clear_buffer_jbddirty(jh2bh(jh));
+			JBUFFER_TRACE(jh, "journal is aborting: refile");
+			journal_refile_buffer(journal, jh);
+			/* If that was the last one, we need to clean up
+			 * any descriptor buffers which may have been
+			 * already allocated, even if we are now
+			 * aborting. */
+			if (!commit_transaction->t_buffers)
+				goto start_journal_io;
+			continue;
+		}
+
+		/* Make sure we have a descriptor block in which to
+		   record the metadata buffer. */
+
+		if (!descriptor) {
+			struct buffer_head *bh;
+
+			J_ASSERT (bufs == 0);
+
+			jbd_debug(4, "JBD: get descriptor\n");
+
+			descriptor = journal_get_descriptor_buffer(journal);
+			if (!descriptor) {
+				journal_abort(journal, -EIO);
+				continue;
+			}
+
+			bh = jh2bh(descriptor);
+			jbd_debug(4, "JBD: got buffer %llu (%p)\n",
+				(unsigned long long)bh->b_blocknr, bh->b_data);
+			header = (journal_header_t *)&bh->b_data[0];
+			header->h_magic     = cpu_to_be32(JFS_MAGIC_NUMBER);
+			header->h_blocktype = cpu_to_be32(JFS_DESCRIPTOR_BLOCK);
+			header->h_sequence  = cpu_to_be32(commit_transaction->t_tid);
+
+			tagp = &bh->b_data[sizeof(journal_header_t)];
+			space_left = bh->b_size - sizeof(journal_header_t);
+			first_tag = 1;
+			set_buffer_jwrite(bh);
+			set_buffer_dirty(bh);
+			wbuf[bufs++] = bh;
+
+			/* Record it so that we can wait for IO
+                           completion later */
+			BUFFER_TRACE(bh, "ph3: file as descriptor");
+			journal_file_buffer(descriptor, commit_transaction,
+					BJ_LogCtl);
+		}
+
+		/* Where is the buffer to be written? */
+
+		err = journal_next_log_block(journal, &blocknr);
+		/* If the block mapping failed, just abandon the buffer
+		   and repeat this loop: we'll fall into the
+		   refile-on-abort condition above. */
+		if (err) {
+			journal_abort(journal, err);
+			continue;
+		}
+
+		/*
+		 * start_this_handle() uses t_outstanding_credits to determine
+		 * the free space in the log, but this counter is changed
+		 * by journal_next_log_block() also.
+		 */
+		commit_transaction->t_outstanding_credits--;
+
+		/* Bump b_count to prevent truncate from stumbling over
+                   the shadowed buffer!  @@@ This can go if we ever get
+                   rid of the BJ_IO/BJ_Shadow pairing of buffers. */
+		atomic_inc(&jh2bh(jh)->b_count);
+
+		/* Make a temporary IO buffer with which to write it out
+                   (this will requeue both the metadata buffer and the
+                   temporary IO buffer). new_bh goes on BJ_IO*/
+
+		set_bit(BH_JWrite, &jh2bh(jh)->b_state);
+		/*
+		 * akpm: journal_write_metadata_buffer() sets
+		 * new_bh->b_transaction to commit_transaction.
+		 * We need to clean this up before we release new_bh
+		 * (which is of type BJ_IO)
+		 */
+		JBUFFER_TRACE(jh, "ph3: write metadata");
+		flags = journal_write_metadata_buffer(commit_transaction,
+						      jh, &new_jh, blocknr);
+		set_bit(BH_JWrite, &jh2bh(new_jh)->b_state);
+		wbuf[bufs++] = jh2bh(new_jh);
+
+		/* Record the new block's tag in the current descriptor
+                   buffer */
+
+		tag_flag = 0;
+		if (flags & 1)
+			tag_flag |= JFS_FLAG_ESCAPE;
+		if (!first_tag)
+			tag_flag |= JFS_FLAG_SAME_UUID;
+
+		tag = (journal_block_tag_t *) tagp;
+		tag->t_blocknr = cpu_to_be32(jh2bh(jh)->b_blocknr);
+		tag->t_flags = cpu_to_be32(tag_flag);
+		tagp += sizeof(journal_block_tag_t);
+		space_left -= sizeof(journal_block_tag_t);
+
+		if (first_tag) {
+			memcpy (tagp, journal->j_uuid, 16);
+			tagp += 16;
+			space_left -= 16;
+			first_tag = 0;
+		}
+
+		/* If there's no more to do, or if the descriptor is full,
+		   let the IO rip! */
+
+		if (bufs == journal->j_wbufsize ||
+		    commit_transaction->t_buffers == NULL ||
+		    space_left < sizeof(journal_block_tag_t) + 16) {
+
+			jbd_debug(4, "JBD: Submit %d IOs\n", bufs);
+
+			/* Write an end-of-descriptor marker before
+                           submitting the IOs.  "tag" still points to
+                           the last tag we set up. */
+
+			tag->t_flags |= cpu_to_be32(JFS_FLAG_LAST_TAG);
+
+start_journal_io:
+			for (i = 0; i < bufs; i++) {
+				struct buffer_head *bh = wbuf[i];
+				lock_buffer(bh);
+				clear_buffer_dirty(bh);
+				set_buffer_uptodate(bh);
+				bh->b_end_io = journal_end_buffer_io_sync;
+				submit_bh(write_op, bh);
+			}
+			cond_resched();
+
+			/* Force a new descriptor to be generated next
+                           time round the loop. */
+			descriptor = NULL;
+			bufs = 0;
+		}
+	}
+
+	/* Lo and behold: we have just managed to send a transaction to
+           the log.  Before we can commit it, wait for the IO so far to
+           complete.  Control buffers being written are on the
+           transaction's t_log_list queue, and metadata buffers are on
+           the t_iobuf_list queue.
+
+	   Wait for the buffers in reverse order.  That way we are
+	   less likely to be woken up until all IOs have completed, and
+	   so we incur less scheduling load.
+	*/
+
+	jbd_debug(3, "JBD: commit phase 4\n");
+
+	/*
+	 * akpm: these are BJ_IO, and j_list_lock is not needed.
+	 * See __journal_try_to_free_buffer.
+	 */
+wait_for_iobuf:
+	while (commit_transaction->t_iobuf_list != NULL) {
+		struct buffer_head *bh;
+
+		jh = commit_transaction->t_iobuf_list->b_tprev;
+		bh = jh2bh(jh);
+		if (buffer_locked(bh)) {
+			wait_on_buffer(bh);
+			goto wait_for_iobuf;
+		}
+		if (cond_resched())
+			goto wait_for_iobuf;
+
+		if (unlikely(!buffer_uptodate(bh)))
+			err = -EIO;
+
+		clear_buffer_jwrite(bh);
+
+		JBUFFER_TRACE(jh, "ph4: unfile after journal write");
+		journal_unfile_buffer(journal, jh);
+
+		/*
+		 * ->t_iobuf_list should contain only dummy buffer_heads
+		 * which were created by journal_write_metadata_buffer().
+		 */
+		BUFFER_TRACE(bh, "dumping temporary bh");
+		journal_put_journal_head(jh);
+		__brelse(bh);
+		J_ASSERT_BH(bh, atomic_read(&bh->b_count) == 0);
+		free_buffer_head(bh);
+
+		/* We also have to unlock and free the corresponding
+                   shadowed buffer */
+		jh = commit_transaction->t_shadow_list->b_tprev;
+		bh = jh2bh(jh);
+		clear_bit(BH_JWrite, &bh->b_state);
+		J_ASSERT_BH(bh, buffer_jbddirty(bh));
+
+		/* The metadata is now released for reuse, but we need
+                   to remember it against this transaction so that when
+                   we finally commit, we can do any checkpointing
+                   required. */
+		JBUFFER_TRACE(jh, "file as BJ_Forget");
+		journal_file_buffer(jh, commit_transaction, BJ_Forget);
+		/*
+		 * Wake up any transactions which were waiting for this
+		 * IO to complete. The barrier must be here so that changes
+		 * by journal_file_buffer() take effect before wake_up_bit()
+		 * does the waitqueue check.
+		 */
+		smp_mb();
+		wake_up_bit(&bh->b_state, BH_Unshadow);
+		JBUFFER_TRACE(jh, "brelse shadowed buffer");
+		__brelse(bh);
+	}
+
+	J_ASSERT (commit_transaction->t_shadow_list == NULL);
+
+	jbd_debug(3, "JBD: commit phase 5\n");
+
+	/* Here we wait for the revoke record and descriptor record buffers */
+ wait_for_ctlbuf:
+	while (commit_transaction->t_log_list != NULL) {
+		struct buffer_head *bh;
+
+		jh = commit_transaction->t_log_list->b_tprev;
+		bh = jh2bh(jh);
+		if (buffer_locked(bh)) {
+			wait_on_buffer(bh);
+			goto wait_for_ctlbuf;
+		}
+		if (cond_resched())
+			goto wait_for_ctlbuf;
+
+		if (unlikely(!buffer_uptodate(bh)))
+			err = -EIO;
+
+		BUFFER_TRACE(bh, "ph5: control buffer writeout done: unfile");
+		clear_buffer_jwrite(bh);
+		journal_unfile_buffer(journal, jh);
+		journal_put_journal_head(jh);
+		__brelse(bh);		/* One for getblk */
+		/* AKPM: bforget here */
+	}
+
+	if (err)
+		journal_abort(journal, err);
+
+	jbd_debug(3, "JBD: commit phase 6\n");
+
+	if (journal_write_commit_record(journal, commit_transaction))
+		err = -EIO;
+
+	if (err)
+		journal_abort(journal, err);
+
+	/* End of a transaction!  Finally, we can do checkpoint
+           processing: any buffers committed as a result of this
+           transaction can be removed from any checkpoint list it was on
+           before. */
+
+	jbd_debug(3, "JBD: commit phase 7\n");
+
+	J_ASSERT(commit_transaction->t_sync_datalist == NULL);
+	J_ASSERT(commit_transaction->t_buffers == NULL);
+	J_ASSERT(commit_transaction->t_checkpoint_list == NULL);
+	J_ASSERT(commit_transaction->t_iobuf_list == NULL);
+	J_ASSERT(commit_transaction->t_shadow_list == NULL);
+	J_ASSERT(commit_transaction->t_log_list == NULL);
+
+restart_loop:
+	/*
+	 * As there are other places (journal_unmap_buffer()) adding buffers
+	 * to this list we have to be careful and hold the j_list_lock.
+	 */
+	spin_lock(&journal->j_list_lock);
+	while (commit_transaction->t_forget) {
+		transaction_t *cp_transaction;
+		struct buffer_head *bh;
+
+		jh = commit_transaction->t_forget;
+		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);
+
+		/*
+		 * If there is undo-protected committed data against
+		 * this buffer, then we can remove it now.  If it is a
+		 * buffer needing such protection, the old frozen_data
+		 * field now points to a committed version of the
+		 * buffer, so rotate that field to the new committed
+		 * data.
+		 *
+		 * Otherwise, we can just throw away the frozen data now.
+		 */
+		if (jh->b_committed_data) {
+			jbd_free(jh->b_committed_data, bh->b_size);
+			jh->b_committed_data = NULL;
+			if (jh->b_frozen_data) {
+				jh->b_committed_data = jh->b_frozen_data;
+				jh->b_frozen_data = NULL;
+			}
+		} else if (jh->b_frozen_data) {
+			jbd_free(jh->b_frozen_data, bh->b_size);
+			jh->b_frozen_data = NULL;
+		}
+
+		spin_lock(&journal->j_list_lock);
+		cp_transaction = jh->b_cp_transaction;
+		if (cp_transaction) {
+			JBUFFER_TRACE(jh, "remove from old cp transaction");
+			__journal_remove_checkpoint(jh);
+		}
+
+		/* Only re-checkpoint the buffer_head if it is marked
+		 * dirty.  If the buffer was added to the BJ_Forget list
+		 * by journal_forget, it may no longer be dirty and
+		 * there's no point in keeping a checkpoint record for
+		 * it. */
+
+		/* 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
+		 * use in a different page. */
+		if (buffer_freed(bh)) {
+			clear_buffer_freed(bh);
+			clear_buffer_jbddirty(bh);
+		}
+
+		if (buffer_jbddirty(bh)) {
+			JBUFFER_TRACE(jh, "add to new checkpointing trans");
+			__journal_insert_checkpoint(jh, commit_transaction);
+			if (is_journal_aborted(journal))
+				clear_buffer_jbddirty(bh);
+			JBUFFER_TRACE(jh, "refile for checkpoint writeback");
+			__journal_refile_buffer(jh);
+			jbd_unlock_bh_state(bh);
+		} else {
+			J_ASSERT_BH(bh, !buffer_dirty(bh));
+			/* The buffer on BJ_Forget list and not jbddirty means
+			 * it has been freed by this transaction and hence it
+			 * could not have been reallocated until this
+			 * transaction has committed. *BUT* it could be
+			 * reallocated once we have written all the data to
+			 * disk and before we process the buffer on BJ_Forget
+			 * list. */
+			JBUFFER_TRACE(jh, "refile or unfile freed buffer");
+			__journal_refile_buffer(jh);
+			if (!jh->b_transaction) {
+				jbd_unlock_bh_state(bh);
+				 /* needs a brelse */
+				journal_remove_journal_head(bh);
+				release_buffer_page(bh);
+			} else
+				jbd_unlock_bh_state(bh);
+		}
+		cond_resched_lock(&journal->j_list_lock);
+	}
+	spin_unlock(&journal->j_list_lock);
+	/*
+	 * This is a bit sleazy.  We use j_list_lock to protect transition
+	 * of a transaction into T_FINISHED state and calling
+	 * __journal_drop_transaction(). Otherwise we could race with
+	 * other checkpointing code processing the transaction...
+	 */
+	spin_lock(&journal->j_state_lock);
+	spin_lock(&journal->j_list_lock);
+	/*
+	 * Now recheck if some buffers did not get attached to the transaction
+	 * while the lock was dropped...
+	 */
+	if (commit_transaction->t_forget) {
+		spin_unlock(&journal->j_list_lock);
+		spin_unlock(&journal->j_state_lock);
+		goto restart_loop;
+	}
+
+	/* Done with this transaction! */
+
+	jbd_debug(3, "JBD: commit phase 8\n");
+
+	J_ASSERT(commit_transaction->t_state == T_COMMIT);
+
+	commit_transaction->t_state = T_FINISHED;
+	J_ASSERT(commit_transaction == journal->j_committing_transaction);
+	journal->j_commit_sequence = commit_transaction->t_tid;
+	journal->j_committing_transaction = NULL;
+	commit_time = ktime_to_ns(ktime_sub(ktime_get(), start_time));
+
+	/*
+	 * weight the commit time higher than the average time so we don't
+	 * react too strongly to vast changes in commit time
+	 */
+	if (likely(journal->j_average_commit_time))
+		journal->j_average_commit_time = (commit_time*3 +
+				journal->j_average_commit_time) / 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) {
+		__journal_drop_transaction(journal, commit_transaction);
+	} else {
+		if (journal->j_checkpoint_transactions == NULL) {
+			journal->j_checkpoint_transactions = commit_transaction;
+			commit_transaction->t_cpnext = commit_transaction;
+			commit_transaction->t_cpprev = commit_transaction;
+		} else {
+			commit_transaction->t_cpnext =
+				journal->j_checkpoint_transactions;
+			commit_transaction->t_cpprev =
+				commit_transaction->t_cpnext->t_cpprev;
+			commit_transaction->t_cpnext->t_cpprev =
+				commit_transaction;
+			commit_transaction->t_cpprev->t_cpnext =
+				commit_transaction;
+		}
+	}
+	spin_unlock(&journal->j_list_lock);
+
+	jbd_debug(1, "JBD: commit %d complete, head %d\n",
+		  journal->j_commit_sequence, journal->j_tail_sequence);
+
+	wake_up(&journal->j_wait_done_commit);
+}

kernel/fs/jbd/recovery.c

@@ -0,0 +1,595 @@
+/*
+ * linux/fs/jbd/recovery.c
+ *
+ * Written by Stephen C. Tweedie <sct@redhat.com>, 1999
+ *
+ * Copyright 1999-2000 Red Hat Software --- All Rights Reserved
+ *
+ * This file is part of the Linux kernel and is made available under
+ * the terms of the GNU General Public License, version 2, or at your
+ * option, any later version, incorporated herein by reference.
+ *
+ * Journal recovery routines for the generic filesystem journaling code;
+ * part of the ext2fs journaling system.
+ */
+
+#ifndef __KERNEL__
+#include "jfs_user.h"
+#else
+#include <linux/time.h>
+#include <linux/fs.h>
+#include <linux/jbd.h>
+#include <linux/errno.h>
+#include <linux/slab.h>
+#endif
+
+/*
+ * Maintain information about the progress of the recovery job, so that
+ * the different passes can carry information between them.
+ */
+struct recovery_info
+{
+	tid_t		start_transaction;
+	tid_t		end_transaction;
+
+	int		nr_replays;
+	int		nr_revokes;
+	int		nr_revoke_hits;
+};
+
+enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY};
+static int do_one_pass(journal_t *journal,
+				struct recovery_info *info, enum passtype pass);
+static int scan_revoke_records(journal_t *, struct buffer_head *,
+				tid_t, struct recovery_info *);
+
+#ifdef __KERNEL__
+
+/* Release readahead buffers after use */
+static void journal_brelse_array(struct buffer_head *b[], int n)
+{
+	while (--n >= 0)
+		brelse (b[n]);
+}
+
+
+/*
+ * When reading from the journal, we are going through the block device
+ * layer directly and so there is no readahead being done for us.  We
+ * need to implement any readahead ourselves if we want it to happen at
+ * all.  Recovery is basically one long sequential read, so make sure we
+ * do the IO in reasonably large chunks.
+ *
+ * This is not so critical that we need to be enormously clever about
+ * the readahead size, though.  128K is a purely arbitrary, good-enough
+ * fixed value.
+ */
+
+#define MAXBUF 8
+static int do_readahead(journal_t *journal, unsigned int start)
+{
+	int err;
+	unsigned int max, nbufs, next;
+	unsigned int blocknr;
+	struct buffer_head *bh;
+
+	struct buffer_head * bufs[MAXBUF];
+
+	/* Do up to 128K of readahead */
+	max = start + (128 * 1024 / journal->j_blocksize);
+	if (max > journal->j_maxlen)
+		max = journal->j_maxlen;
+
+	/* Do the readahead itself.  We'll submit MAXBUF buffer_heads at
+	 * a time to the block device IO layer. */
+
+	nbufs = 0;
+
+	for (next = start; next < max; next++) {
+		err = journal_bmap(journal, next, &blocknr);
+
+		if (err) {
+			printk (KERN_ERR "JBD: bad block at offset %u\n",
+				next);
+			goto failed;
+		}
+
+		bh = __getblk(journal->j_dev, blocknr, journal->j_blocksize);
+		if (!bh) {
+			err = -ENOMEM;
+			goto failed;
+		}
+
+		if (!buffer_uptodate(bh) && !buffer_locked(bh)) {
+			bufs[nbufs++] = bh;
+			if (nbufs == MAXBUF) {
+				ll_rw_block(READ, nbufs, bufs);
+				journal_brelse_array(bufs, nbufs);
+				nbufs = 0;
+			}
+		} else
+			brelse(bh);
+	}
+
+	if (nbufs)
+		ll_rw_block(READ, nbufs, bufs);
+	err = 0;
+
+failed:
+	if (nbufs)
+		journal_brelse_array(bufs, nbufs);
+	return err;
+}
+
+#endif /* __KERNEL__ */
+
+
+/*
+ * Read a block from the journal
+ */
+
+static int jread(struct buffer_head **bhp, journal_t *journal,
+		 unsigned int offset)
+{
+	int err;
+	unsigned int blocknr;
+	struct buffer_head *bh;
+
+	*bhp = NULL;
+
+	if (offset >= journal->j_maxlen) {
+		printk(KERN_ERR "JBD: corrupted journal superblock\n");
+		return -EIO;
+	}
+
+	err = journal_bmap(journal, offset, &blocknr);
+
+	if (err) {
+		printk (KERN_ERR "JBD: bad block at offset %u\n",
+			offset);
+		return err;
+	}
+
+	bh = __getblk(journal->j_dev, blocknr, journal->j_blocksize);
+	if (!bh)
+		return -ENOMEM;
+
+	if (!buffer_uptodate(bh)) {
+		/* If this is a brand new buffer, start readahead.
+                   Otherwise, we assume we are already reading it.  */
+		if (!buffer_req(bh))
+			do_readahead(journal, offset);
+		wait_on_buffer(bh);
+	}
+
+	if (!buffer_uptodate(bh)) {
+		printk (KERN_ERR "JBD: Failed to read block at offset %u\n",
+			offset);
+		brelse(bh);
+		return -EIO;
+	}
+
+	*bhp = bh;
+	return 0;
+}
+
+
+/*
+ * Count the number of in-use tags in a journal descriptor block.
+ */
+
+static int count_tags(struct buffer_head *bh, int size)
+{
+	char *			tagp;
+	journal_block_tag_t *	tag;
+	int			nr = 0;
+
+	tagp = &bh->b_data[sizeof(journal_header_t)];
+
+	while ((tagp - bh->b_data + sizeof(journal_block_tag_t)) <= size) {
+		tag = (journal_block_tag_t *) tagp;
+
+		nr++;
+		tagp += sizeof(journal_block_tag_t);
+		if (!(tag->t_flags & cpu_to_be32(JFS_FLAG_SAME_UUID)))
+			tagp += 16;
+
+		if (tag->t_flags & cpu_to_be32(JFS_FLAG_LAST_TAG))
+			break;
+	}
+
+	return nr;
+}
+
+
+/* Make sure we wrap around the log correctly! */
+#define wrap(journal, var)						\
+do {									\
+	if (var >= (journal)->j_last)					\
+		var -= ((journal)->j_last - (journal)->j_first);	\
+} while (0)
+
+/**
+ * journal_recover - recovers a on-disk journal
+ * @journal: the journal to recover
+ *
+ * The primary function for recovering the log contents when mounting a
+ * journaled device.
+ *
+ * Recovery is done in three passes.  In the first pass, we look for the
+ * end of the log.  In the second, we assemble the list of revoke
+ * blocks.  In the third and final pass, we replay any un-revoked blocks
+ * in the log.
+ */
+int journal_recover(journal_t *journal)
+{
+	int			err, err2;
+	journal_superblock_t *	sb;
+
+	struct recovery_info	info;
+
+	memset(&info, 0, sizeof(info));
+	sb = journal->j_superblock;
+
+	/*
+	 * The journal superblock's s_start field (the current log head)
+	 * is always zero if, and only if, the journal was cleanly
+	 * unmounted.
+	 */
+
+	if (!sb->s_start) {
+		jbd_debug(1, "No recovery required, last transaction %d\n",
+			  be32_to_cpu(sb->s_sequence));
+		journal->j_transaction_sequence = be32_to_cpu(sb->s_sequence) + 1;
+		return 0;
+	}
+
+	err = do_one_pass(journal, &info, PASS_SCAN);
+	if (!err)
+		err = do_one_pass(journal, &info, PASS_REVOKE);
+	if (!err)
+		err = do_one_pass(journal, &info, PASS_REPLAY);
+
+	jbd_debug(1, "JBD: recovery, exit status %d, "
+		  "recovered transactions %u to %u\n",
+		  err, info.start_transaction, info.end_transaction);
+	jbd_debug(1, "JBD: Replayed %d and revoked %d/%d blocks\n",
+		  info.nr_replays, info.nr_revoke_hits, info.nr_revokes);
+
+	/* Restart the log at the next transaction ID, thus invalidating
+	 * any existing commit records in the log. */
+	journal->j_transaction_sequence = ++info.end_transaction;
+
+	journal_clear_revoke(journal);
+	err2 = sync_blockdev(journal->j_fs_dev);
+	if (!err)
+		err = err2;
+
+	return err;
+}
+
+/**
+ * journal_skip_recovery - Start journal and wipe exiting records
+ * @journal: journal to startup
+ *
+ * Locate any valid recovery information from the journal and set up the
+ * journal structures in memory to ignore it (presumably because the
+ * caller has evidence that it is out of date).
+ * This function does'nt appear to be exorted..
+ *
+ * We perform one pass over the journal to allow us to tell the user how
+ * much recovery information is being erased, and to let us initialise
+ * the journal transaction sequence numbers to the next unused ID.
+ */
+int journal_skip_recovery(journal_t *journal)
+{
+	int			err;
+	journal_superblock_t *	sb;
+
+	struct recovery_info	info;
+
+	memset (&info, 0, sizeof(info));
+	sb = journal->j_superblock;
+
+	err = do_one_pass(journal, &info, PASS_SCAN);
+
+	if (err) {
+		printk(KERN_ERR "JBD: error %d scanning journal\n", err);
+		++journal->j_transaction_sequence;
+	} else {
+#ifdef CONFIG_JBD_DEBUG
+		int dropped = info.end_transaction - be32_to_cpu(sb->s_sequence);
+#endif
+		jbd_debug(1,
+			  "JBD: ignoring %d transaction%s from the journal.\n",
+			  dropped, (dropped == 1) ? "" : "s");
+		journal->j_transaction_sequence = ++info.end_transaction;
+	}
+
+	journal->j_tail = 0;
+	return err;
+}
+
+static int do_one_pass(journal_t *journal,
+			struct recovery_info *info, enum passtype pass)
+{
+	unsigned int		first_commit_ID, next_commit_ID;
+	unsigned int		next_log_block;
+	int			err, success = 0;
+	journal_superblock_t *	sb;
+	journal_header_t *	tmp;
+	struct buffer_head *	bh;
+	unsigned int		sequence;
+	int			blocktype;
+
+	/* Precompute the maximum metadata descriptors in a descriptor block */
+	int			MAX_BLOCKS_PER_DESC;
+	MAX_BLOCKS_PER_DESC = ((journal->j_blocksize-sizeof(journal_header_t))
+			       / sizeof(journal_block_tag_t));
+
+	/*
+	 * First thing is to establish what we expect to find in the log
+	 * (in terms of transaction IDs), and where (in terms of log
+	 * block offsets): query the superblock.
+	 */
+
+	sb = journal->j_superblock;
+	next_commit_ID = be32_to_cpu(sb->s_sequence);
+	next_log_block = be32_to_cpu(sb->s_start);
+
+	first_commit_ID = next_commit_ID;
+	if (pass == PASS_SCAN)
+		info->start_transaction = first_commit_ID;
+
+	jbd_debug(1, "Starting recovery pass %d\n", pass);
+
+	/*
+	 * Now we walk through the log, transaction by transaction,
+	 * making sure that each transaction has a commit block in the
+	 * expected place.  Each complete transaction gets replayed back
+	 * into the main filesystem.
+	 */
+
+	while (1) {
+		int			flags;
+		char *			tagp;
+		journal_block_tag_t *	tag;
+		struct buffer_head *	obh;
+		struct buffer_head *	nbh;
+
+		cond_resched();
+
+		/* If we already know where to stop the log traversal,
+		 * check right now that we haven't gone past the end of
+		 * the log. */
+
+		if (pass != PASS_SCAN)
+			if (tid_geq(next_commit_ID, info->end_transaction))
+				break;
+
+		jbd_debug(2, "Scanning for sequence ID %u at %u/%u\n",
+			  next_commit_ID, next_log_block, journal->j_last);
+
+		/* Skip over each chunk of the transaction looking
+		 * either the next descriptor block or the final commit
+		 * record. */
+
+		jbd_debug(3, "JBD: checking block %u\n", next_log_block);
+		err = jread(&bh, journal, next_log_block);
+		if (err)
+			goto failed;
+
+		next_log_block++;
+		wrap(journal, next_log_block);
+
+		/* What kind of buffer is it?
+		 *
+		 * If it is a descriptor block, check that it has the
+		 * expected sequence number.  Otherwise, we're all done
+		 * here. */
+
+		tmp = (journal_header_t *)bh->b_data;
+
+		if (tmp->h_magic != cpu_to_be32(JFS_MAGIC_NUMBER)) {
+			brelse(bh);
+			break;
+		}
+
+		blocktype = be32_to_cpu(tmp->h_blocktype);
+		sequence = be32_to_cpu(tmp->h_sequence);
+		jbd_debug(3, "Found magic %d, sequence %d\n",
+			  blocktype, sequence);
+
+		if (sequence != next_commit_ID) {
+			brelse(bh);
+			break;
+		}
+
+		/* OK, we have a valid descriptor block which matches
+		 * all of the sequence number checks.  What are we going
+		 * to do with it?  That depends on the pass... */
+
+		switch(blocktype) {
+		case JFS_DESCRIPTOR_BLOCK:
+			/* If it is a valid descriptor block, replay it
+			 * in pass REPLAY; otherwise, just skip over the
+			 * blocks it describes. */
+			if (pass != PASS_REPLAY) {
+				next_log_block +=
+					count_tags(bh, journal->j_blocksize);
+				wrap(journal, next_log_block);
+				brelse(bh);
+				continue;
+			}
+
+			/* A descriptor block: we can now write all of
+			 * the data blocks.  Yay, useful work is finally
+			 * getting done here! */
+
+			tagp = &bh->b_data[sizeof(journal_header_t)];
+			while ((tagp - bh->b_data +sizeof(journal_block_tag_t))
+			       <= journal->j_blocksize) {
+				unsigned int io_block;
+
+				tag = (journal_block_tag_t *) tagp;
+				flags = be32_to_cpu(tag->t_flags);
+
+				io_block = next_log_block++;
+				wrap(journal, next_log_block);
+				err = jread(&obh, journal, io_block);
+				if (err) {
+					/* Recover what we can, but
+					 * report failure at the end. */
+					success = err;
+					printk (KERN_ERR
+						"JBD: IO error %d recovering "
+						"block %u in log\n",
+						err, io_block);
+				} else {
+					unsigned int blocknr;
+
+					J_ASSERT(obh != NULL);
+					blocknr = be32_to_cpu(tag->t_blocknr);
+
+					/* If the block has been
+					 * revoked, then we're all done
+					 * here. */
+					if (journal_test_revoke
+					    (journal, blocknr,
+					     next_commit_ID)) {
+						brelse(obh);
+						++info->nr_revoke_hits;
+						goto skip_write;
+					}
+
+					/* Find a buffer for the new
+					 * data being restored */
+					nbh = __getblk(journal->j_fs_dev,
+							blocknr,
+							journal->j_blocksize);
+					if (nbh == NULL) {
+						printk(KERN_ERR
+						       "JBD: Out of memory "
+						       "during recovery.\n");
+						err = -ENOMEM;
+						brelse(bh);
+						brelse(obh);
+						goto failed;
+					}
+
+					lock_buffer(nbh);
+					memcpy(nbh->b_data, obh->b_data,
+							journal->j_blocksize);
+					if (flags & JFS_FLAG_ESCAPE) {
+						*((__be32 *)nbh->b_data) =
+						cpu_to_be32(JFS_MAGIC_NUMBER);
+					}
+
+					BUFFER_TRACE(nbh, "marking dirty");
+					set_buffer_uptodate(nbh);
+					mark_buffer_dirty(nbh);
+					BUFFER_TRACE(nbh, "marking uptodate");
+					++info->nr_replays;
+					/* ll_rw_block(WRITE, 1, &nbh); */
+					unlock_buffer(nbh);
+					brelse(obh);
+					brelse(nbh);
+				}
+
+			skip_write:
+				tagp += sizeof(journal_block_tag_t);
+				if (!(flags & JFS_FLAG_SAME_UUID))
+					tagp += 16;
+
+				if (flags & JFS_FLAG_LAST_TAG)
+					break;
+			}
+
+			brelse(bh);
+			continue;
+
+		case JFS_COMMIT_BLOCK:
+			/* Found an expected commit block: not much to
+			 * do other than move on to the next sequence
+			 * number. */
+			brelse(bh);
+			next_commit_ID++;
+			continue;
+
+		case JFS_REVOKE_BLOCK:
+			/* If we aren't in the REVOKE pass, then we can
+			 * just skip over this block. */
+			if (pass != PASS_REVOKE) {
+				brelse(bh);
+				continue;
+			}
+
+			err = scan_revoke_records(journal, bh,
+						  next_commit_ID, info);
+			brelse(bh);
+			if (err)
+				goto failed;
+			continue;
+
+		default:
+			jbd_debug(3, "Unrecognised magic %d, end of scan.\n",
+				  blocktype);
+			brelse(bh);
+			goto done;
+		}
+	}
+
+ done:
+	/*
+	 * We broke out of the log scan loop: either we came to the
+	 * known end of the log or we found an unexpected block in the
+	 * log.  If the latter happened, then we know that the "current"
+	 * transaction marks the end of the valid log.
+	 */
+
+	if (pass == PASS_SCAN)
+		info->end_transaction = next_commit_ID;
+	else {
+		/* It's really bad news if different passes end up at
+		 * different places (but possible due to IO errors). */
+		if (info->end_transaction != next_commit_ID) {
+			printk (KERN_ERR "JBD: recovery pass %d ended at "
+				"transaction %u, expected %u\n",
+				pass, next_commit_ID, info->end_transaction);
+			if (!success)
+				success = -EIO;
+		}
+	}
+
+	return success;
+
+ failed:
+	return err;
+}
+
+
+/* Scan a revoke record, marking all blocks mentioned as revoked. */
+
+static int scan_revoke_records(journal_t *journal, struct buffer_head *bh,
+			       tid_t sequence, struct recovery_info *info)
+{
+	journal_revoke_header_t *header;
+	int offset, max;
+
+	header = (journal_revoke_header_t *) bh->b_data;
+	offset = sizeof(journal_revoke_header_t);
+	max = be32_to_cpu(header->r_count);
+
+	while (offset < max) {
+		unsigned int blocknr;
+		int err;
+
+		blocknr = be32_to_cpu(* ((__be32 *) (bh->b_data+offset)));
+		offset += 4;
+		err = journal_set_revoke(journal, blocknr, sequence);
+		if (err)
+			return err;
+		++info->nr_revokes;
+	}
+	return 0;
+}

kernel/fs/jbd/revoke.c

@@ -0,0 +1,706 @@
+/*
+ * linux/fs/jbd/revoke.c
+ *
+ * Written by Stephen C. Tweedie <sct@redhat.com>, 2000
+ *
+ * Copyright 2000 Red Hat corp --- All Rights Reserved
+ *
+ * This file is part of the Linux kernel and is made available under
+ * the terms of the GNU General Public License, version 2, or at your
+ * option, any later version, incorporated herein by reference.
+ *
+ * Journal revoke routines for the generic filesystem journaling code;
+ * part of the ext2fs journaling system.
+ *
+ * Revoke is the mechanism used to prevent old log records for deleted
+ * metadata from being replayed on top of newer data using the same
+ * blocks.  The revoke mechanism is used in two separate places:
+ *
+ * + Commit: during commit we write the entire list of the current
+ *   transaction's revoked blocks to the journal
+ *
+ * + Recovery: during recovery we record the transaction ID of all
+ *   revoked blocks.  If there are multiple revoke records in the log
+ *   for a single block, only the last one counts, and if there is a log
+ *   entry for a block beyond the last revoke, then that log entry still
+ *   gets replayed.
+ *
+ * We can get interactions between revokes and new log data within a
+ * single transaction:
+ *
+ * Block is revoked and then journaled:
+ *   The desired end result is the journaling of the new block, so we
+ *   cancel the revoke before the transaction commits.
+ *
+ * Block is journaled and then revoked:
+ *   The revoke must take precedence over the write of the block, so we
+ *   need either to cancel the journal entry or to write the revoke
+ *   later in the log than the log block.  In this case, we choose the
+ *   latter: journaling a block cancels any revoke record for that block
+ *   in the current transaction, so any revoke for that block in the
+ *   transaction must have happened after the block was journaled and so
+ *   the revoke must take precedence.
+ *
+ * Block is revoked and then written as data:
+ *   The data write is allowed to succeed, but the revoke is _not_
+ *   cancelled.  We still need to prevent old log records from
+ *   overwriting the new data.  We don't even need to clear the revoke
+ *   bit here.
+ *
+ * Revoke information on buffers is a tri-state value:
+ *
+ * RevokeValid clear:	no cached revoke status, need to look it up
+ * RevokeValid set, Revoked clear:
+ *			buffer has not been revoked, and cancel_revoke
+ *			need do nothing.
+ * RevokeValid set, Revoked set:
+ *			buffer has been revoked.
+ *
+ * Locking rules:
+ * We keep two hash tables of revoke records. One hashtable belongs to the
+ * running transaction (is pointed to by journal->j_revoke), the other one
+ * belongs to the committing transaction. Accesses to the second hash table
+ * happen only from the kjournald and no other thread touches this table.  Also
+ * journal_switch_revoke_table() which switches which hashtable belongs to the
+ * running and which to the committing transaction is called only from
+ * kjournald. Therefore we need no locks when accessing the hashtable belonging
+ * to the committing transaction.
+ *
+ * All users operating on the hash table belonging to the running transaction
+ * have a handle to the transaction. Therefore they are safe from kjournald
+ * switching hash tables under them. For operations on the lists of entries in
+ * the hash table j_revoke_lock is used.
+ *
+ * Finally, also replay code uses the hash tables but at this moment noone else
+ * can touch them (filesystem isn't mounted yet) and hence no locking is
+ * needed.
+ */
+
+#ifndef __KERNEL__
+#include "jfs_user.h"
+#else
+#include <linux/time.h>
+#include <linux/fs.h>
+#include <linux/jbd.h>
+#include <linux/errno.h>
+#include <linux/slab.h>
+#include <linux/list.h>
+#include <linux/init.h>
+#include <linux/bio.h>
+#endif
+#include <linux/log2.h>
+
+static struct kmem_cache *revoke_record_cache;
+static struct kmem_cache *revoke_table_cache;
+
+/* Each revoke record represents one single revoked block.  During
+   journal replay, this involves recording the transaction ID of the
+   last transaction to revoke this block. */
+
+struct jbd_revoke_record_s
+{
+	struct list_head  hash;
+	tid_t		  sequence;	/* Used for recovery only */
+	unsigned int	  blocknr;
+};
+
+
+/* The revoke table is just a simple hash table of revoke records. */
+struct jbd_revoke_table_s
+{
+	/* It is conceivable that we might want a larger hash table
+	 * for recovery.  Must be a power of two. */
+	int		  hash_size;
+	int		  hash_shift;
+	struct list_head *hash_table;
+};
+
+
+#ifdef __KERNEL__
+static void write_one_revoke_record(journal_t *, transaction_t *,
+				    struct journal_head **, int *,
+				    struct jbd_revoke_record_s *, int);
+static void flush_descriptor(journal_t *, struct journal_head *, int, int);
+#endif
+
+/* Utility functions to maintain the revoke table */
+
+/* Borrowed from buffer.c: this is a tried and tested block hash function */
+static inline int hash(journal_t *journal, unsigned int block)
+{
+	struct jbd_revoke_table_s *table = journal->j_revoke;
+	int hash_shift = table->hash_shift;
+
+	return ((block << (hash_shift - 6)) ^
+		(block >> 13) ^
+		(block << (hash_shift - 12))) & (table->hash_size - 1);
+}
+
+static int insert_revoke_hash(journal_t *journal, unsigned int blocknr,
+			      tid_t seq)
+{
+	struct list_head *hash_list;
+	struct jbd_revoke_record_s *record;
+
+repeat:
+	record = kmem_cache_alloc(revoke_record_cache, GFP_NOFS);
+	if (!record)
+		goto oom;
+
+	record->sequence = seq;
+	record->blocknr = blocknr;
+	hash_list = &journal->j_revoke->hash_table[hash(journal, blocknr)];
+	spin_lock(&journal->j_revoke_lock);
+	list_add(&record->hash, hash_list);
+	spin_unlock(&journal->j_revoke_lock);
+	return 0;
+
+oom:
+	if (!journal_oom_retry)
+		return -ENOMEM;
+	jbd_debug(1, "ENOMEM in %s, retrying\n", __func__);
+	yield();
+	goto repeat;
+}
+
+/* Find a revoke record in the journal's hash table. */
+
+static struct jbd_revoke_record_s *find_revoke_record(journal_t *journal,
+						      unsigned int blocknr)
+{
+	struct list_head *hash_list;
+	struct jbd_revoke_record_s *record;
+
+	hash_list = &journal->j_revoke->hash_table[hash(journal, blocknr)];
+
+	spin_lock(&journal->j_revoke_lock);
+	record = (struct jbd_revoke_record_s *) hash_list->next;
+	while (&(record->hash) != hash_list) {
+		if (record->blocknr == blocknr) {
+			spin_unlock(&journal->j_revoke_lock);
+			return record;
+		}
+		record = (struct jbd_revoke_record_s *) record->hash.next;
+	}
+	spin_unlock(&journal->j_revoke_lock);
+	return NULL;
+}
+
+void journal_destroy_revoke_caches(void)
+{
+	if (revoke_record_cache) {
+		kmem_cache_destroy(revoke_record_cache);
+		revoke_record_cache = NULL;
+	}
+	if (revoke_table_cache) {
+		kmem_cache_destroy(revoke_table_cache);
+		revoke_table_cache = NULL;
+	}
+}
+
+int __init journal_init_revoke_caches(void)
+{
+	J_ASSERT(!revoke_record_cache);
+	J_ASSERT(!revoke_table_cache);
+
+	revoke_record_cache = kmem_cache_create("revoke_record",
+					   sizeof(struct jbd_revoke_record_s),
+					   0,
+					   SLAB_HWCACHE_ALIGN|SLAB_TEMPORARY,
+					   NULL);
+	if (!revoke_record_cache)
+		goto record_cache_failure;
+
+	revoke_table_cache = kmem_cache_create("revoke_table",
+					   sizeof(struct jbd_revoke_table_s),
+					   0, SLAB_TEMPORARY, NULL);
+	if (!revoke_table_cache)
+		goto table_cache_failure;
+
+	return 0;
+
+table_cache_failure:
+	journal_destroy_revoke_caches();
+record_cache_failure:
+	return -ENOMEM;
+}
+
+static struct jbd_revoke_table_s *journal_init_revoke_table(int hash_size)
+{
+	int shift = 0;
+	int tmp = hash_size;
+	struct jbd_revoke_table_s *table;
+
+	table = kmem_cache_alloc(revoke_table_cache, GFP_KERNEL);
+	if (!table)
+		goto out;
+
+	while((tmp >>= 1UL) != 0UL)
+		shift++;
+
+	table->hash_size = hash_size;
+	table->hash_shift = shift;
+	table->hash_table =
+		kmalloc(hash_size * sizeof(struct list_head), GFP_KERNEL);
+	if (!table->hash_table) {
+		kmem_cache_free(revoke_table_cache, table);
+		table = NULL;
+		goto out;
+	}
+
+	for (tmp = 0; tmp < hash_size; tmp++)
+		INIT_LIST_HEAD(&table->hash_table[tmp]);
+
+out:
+	return table;
+}
+
+static void journal_destroy_revoke_table(struct jbd_revoke_table_s *table)
+{
+	int i;
+	struct list_head *hash_list;
+
+	for (i = 0; i < table->hash_size; i++) {
+		hash_list = &table->hash_table[i];
+		J_ASSERT(list_empty(hash_list));
+	}
+
+	kfree(table->hash_table);
+	kmem_cache_free(revoke_table_cache, table);
+}
+
+/* Initialise the revoke table for a given journal to a given size. */
+int journal_init_revoke(journal_t *journal, int hash_size)
+{
+	J_ASSERT(journal->j_revoke_table[0] == NULL);
+	J_ASSERT(is_power_of_2(hash_size));
+
+	journal->j_revoke_table[0] = journal_init_revoke_table(hash_size);
+	if (!journal->j_revoke_table[0])
+		goto fail0;
+
+	journal->j_revoke_table[1] = journal_init_revoke_table(hash_size);
+	if (!journal->j_revoke_table[1])
+		goto fail1;
+
+	journal->j_revoke = journal->j_revoke_table[1];
+
+	spin_lock_init(&journal->j_revoke_lock);
+
+	return 0;
+
+fail1:
+	journal_destroy_revoke_table(journal->j_revoke_table[0]);
+fail0:
+	return -ENOMEM;
+}
+
+/* Destroy a journal's revoke table.  The table must already be empty! */
+void journal_destroy_revoke(journal_t *journal)
+{
+	journal->j_revoke = NULL;
+	if (journal->j_revoke_table[0])
+		journal_destroy_revoke_table(journal->j_revoke_table[0]);
+	if (journal->j_revoke_table[1])
+		journal_destroy_revoke_table(journal->j_revoke_table[1]);
+}
+
+
+#ifdef __KERNEL__
+
+/*
+ * journal_revoke: revoke a given buffer_head from the journal.  This
+ * prevents the block from being replayed during recovery if we take a
+ * crash after this current transaction commits.  Any subsequent
+ * metadata writes of the buffer in this transaction cancel the
+ * revoke.
+ *
+ * Note that this call may block --- it is up to the caller to make
+ * sure that there are no further calls to journal_write_metadata
+ * before the revoke is complete.  In ext3, this implies calling the
+ * revoke before clearing the block bitmap when we are deleting
+ * metadata.
+ *
+ * Revoke performs a journal_forget on any buffer_head passed in as a
+ * parameter, but does _not_ forget the buffer_head if the bh was only
+ * found implicitly.
+ *
+ * bh_in may not be a journalled buffer - it may have come off
+ * the hash tables without an attached journal_head.
+ *
+ * If bh_in is non-zero, journal_revoke() will decrement its b_count
+ * by one.
+ */
+
+int journal_revoke(handle_t *handle, unsigned int blocknr,
+		   struct buffer_head *bh_in)
+{
+	struct buffer_head *bh = NULL;
+	journal_t *journal;
+	struct block_device *bdev;
+	int err;
+
+	might_sleep();
+	if (bh_in)
+		BUFFER_TRACE(bh_in, "enter");
+
+	journal = handle->h_transaction->t_journal;
+	if (!journal_set_features(journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)){
+		J_ASSERT (!"Cannot set revoke feature!");
+		return -EINVAL;
+	}
+
+	bdev = journal->j_fs_dev;
+	bh = bh_in;
+
+	if (!bh) {
+		bh = __find_get_block(bdev, blocknr, journal->j_blocksize);
+		if (bh)
+			BUFFER_TRACE(bh, "found on hash");
+	}
+#ifdef JBD_EXPENSIVE_CHECKING
+	else {
+		struct buffer_head *bh2;
+
+		/* If there is a different buffer_head lying around in
+		 * memory anywhere... */
+		bh2 = __find_get_block(bdev, blocknr, journal->j_blocksize);
+		if (bh2) {
+			/* ... and it has RevokeValid status... */
+			if (bh2 != bh && buffer_revokevalid(bh2))
+				/* ...then it better be revoked too,
+				 * since it's illegal to create a revoke
+				 * record against a buffer_head which is
+				 * not marked revoked --- that would
+				 * risk missing a subsequent revoke
+				 * cancel. */
+				J_ASSERT_BH(bh2, buffer_revoked(bh2));
+			put_bh(bh2);
+		}
+	}
+#endif
+
+	/* We really ought not ever to revoke twice in a row without
+           first having the revoke cancelled: it's illegal to free a
+           block twice without allocating it in between! */
+	if (bh) {
+		if (!J_EXPECT_BH(bh, !buffer_revoked(bh),
+				 "inconsistent data on disk")) {
+			if (!bh_in)
+				brelse(bh);
+			return -EIO;
+		}
+		set_buffer_revoked(bh);
+		set_buffer_revokevalid(bh);
+		if (bh_in) {
+			BUFFER_TRACE(bh_in, "call journal_forget");
+			journal_forget(handle, bh_in);
+		} else {
+			BUFFER_TRACE(bh, "call brelse");
+			__brelse(bh);
+		}
+	}
+
+	jbd_debug(2, "insert revoke for block %u, bh_in=%p\n", blocknr, bh_in);
+	err = insert_revoke_hash(journal, blocknr,
+				handle->h_transaction->t_tid);
+	BUFFER_TRACE(bh_in, "exit");
+	return err;
+}
+
+/*
+ * Cancel an outstanding revoke.  For use only internally by the
+ * journaling code (called from journal_get_write_access).
+ *
+ * We trust buffer_revoked() on the buffer if the buffer is already
+ * being journaled: if there is no revoke pending on the buffer, then we
+ * don't do anything here.
+ *
+ * This would break if it were possible for a buffer to be revoked and
+ * discarded, and then reallocated within the same transaction.  In such
+ * a case we would have lost the revoked bit, but when we arrived here
+ * the second time we would still have a pending revoke to cancel.  So,
+ * do not trust the Revoked bit on buffers unless RevokeValid is also
+ * set.
+ */
+int journal_cancel_revoke(handle_t *handle, struct journal_head *jh)
+{
+	struct jbd_revoke_record_s *record;
+	journal_t *journal = handle->h_transaction->t_journal;
+	int need_cancel;
+	int did_revoke = 0;	/* akpm: debug */
+	struct buffer_head *bh = jh2bh(jh);
+
+	jbd_debug(4, "journal_head %p, cancelling revoke\n", jh);
+
+	/* Is the existing Revoke bit valid?  If so, we trust it, and
+	 * only perform the full cancel if the revoke bit is set.  If
+	 * not, we can't trust the revoke bit, and we need to do the
+	 * full search for a revoke record. */
+	if (test_set_buffer_revokevalid(bh)) {
+		need_cancel = test_clear_buffer_revoked(bh);
+	} else {
+		need_cancel = 1;
+		clear_buffer_revoked(bh);
+	}
+
+	if (need_cancel) {
+		record = find_revoke_record(journal, bh->b_blocknr);
+		if (record) {
+			jbd_debug(4, "cancelled existing revoke on "
+				  "blocknr %llu\n", (unsigned long long)bh->b_blocknr);
+			spin_lock(&journal->j_revoke_lock);
+			list_del(&record->hash);
+			spin_unlock(&journal->j_revoke_lock);
+			kmem_cache_free(revoke_record_cache, record);
+			did_revoke = 1;
+		}
+	}
+
+#ifdef JBD_EXPENSIVE_CHECKING
+	/* There better not be one left behind by now! */
+	record = find_revoke_record(journal, bh->b_blocknr);
+	J_ASSERT_JH(jh, record == NULL);
+#endif
+
+	/* Finally, have we just cleared revoke on an unhashed
+	 * buffer_head?  If so, we'd better make sure we clear the
+	 * revoked status on any hashed alias too, otherwise the revoke
+	 * state machine will get very upset later on. */
+	if (need_cancel) {
+		struct buffer_head *bh2;
+		bh2 = __find_get_block(bh->b_bdev, bh->b_blocknr, bh->b_size);
+		if (bh2) {
+			if (bh2 != bh)
+				clear_buffer_revoked(bh2);
+			__brelse(bh2);
+		}
+	}
+	return did_revoke;
+}
+
+/* journal_switch_revoke table select j_revoke for next transaction
+ * we do not want to suspend any processing until all revokes are
+ * written -bzzz
+ */
+void journal_switch_revoke_table(journal_t *journal)
+{
+	int i;
+
+	if (journal->j_revoke == journal->j_revoke_table[0])
+		journal->j_revoke = journal->j_revoke_table[1];
+	else
+		journal->j_revoke = journal->j_revoke_table[0];
+
+	for (i = 0; i < journal->j_revoke->hash_size; i++)
+		INIT_LIST_HEAD(&journal->j_revoke->hash_table[i]);
+}
+
+/*
+ * Write revoke records to the journal for all entries in the current
+ * revoke hash, deleting the entries as we go.
+ */
+void journal_write_revoke_records(journal_t *journal,
+				  transaction_t *transaction, int write_op)
+{
+	struct journal_head *descriptor;
+	struct jbd_revoke_record_s *record;
+	struct jbd_revoke_table_s *revoke;
+	struct list_head *hash_list;
+	int i, offset, count;
+
+	descriptor = NULL;
+	offset = 0;
+	count = 0;
+
+	/* select revoke table for committing transaction */
+	revoke = journal->j_revoke == journal->j_revoke_table[0] ?
+		journal->j_revoke_table[1] : journal->j_revoke_table[0];
+
+	for (i = 0; i < revoke->hash_size; i++) {
+		hash_list = &revoke->hash_table[i];
+
+		while (!list_empty(hash_list)) {
+			record = (struct jbd_revoke_record_s *)
+				hash_list->next;
+			write_one_revoke_record(journal, transaction,
+						&descriptor, &offset,
+						record, write_op);
+			count++;
+			list_del(&record->hash);
+			kmem_cache_free(revoke_record_cache, record);
+		}
+	}
+	if (descriptor)
+		flush_descriptor(journal, descriptor, offset, write_op);
+	jbd_debug(1, "Wrote %d revoke records\n", count);
+}
+
+/*
+ * Write out one revoke record.  We need to create a new descriptor
+ * block if the old one is full or if we have not already created one.
+ */
+
+static void write_one_revoke_record(journal_t *journal,
+				    transaction_t *transaction,
+				    struct journal_head **descriptorp,
+				    int *offsetp,
+				    struct jbd_revoke_record_s *record,
+				    int write_op)
+{
+	struct journal_head *descriptor;
+	int offset;
+	journal_header_t *header;
+
+	/* If we are already aborting, this all becomes a noop.  We
+           still need to go round the loop in
+           journal_write_revoke_records in order to free all of the
+           revoke records: only the IO to the journal is omitted. */
+	if (is_journal_aborted(journal))
+		return;
+
+	descriptor = *descriptorp;
+	offset = *offsetp;
+
+	/* Make sure we have a descriptor with space left for the record */
+	if (descriptor) {
+		if (offset == journal->j_blocksize) {
+			flush_descriptor(journal, descriptor, offset, write_op);
+			descriptor = NULL;
+		}
+	}
+
+	if (!descriptor) {
+		descriptor = journal_get_descriptor_buffer(journal);
+		if (!descriptor)
+			return;
+		header = (journal_header_t *) &jh2bh(descriptor)->b_data[0];
+		header->h_magic     = cpu_to_be32(JFS_MAGIC_NUMBER);
+		header->h_blocktype = cpu_to_be32(JFS_REVOKE_BLOCK);
+		header->h_sequence  = cpu_to_be32(transaction->t_tid);
+
+		/* Record it so that we can wait for IO completion later */
+		JBUFFER_TRACE(descriptor, "file as BJ_LogCtl");
+		journal_file_buffer(descriptor, transaction, BJ_LogCtl);
+
+		offset = sizeof(journal_revoke_header_t);
+		*descriptorp = descriptor;
+	}
+
+	* ((__be32 *)(&jh2bh(descriptor)->b_data[offset])) =
+		cpu_to_be32(record->blocknr);
+	offset += 4;
+	*offsetp = offset;
+}
+
+/*
+ * Flush a revoke descriptor out to the journal.  If we are aborting,
+ * this is a noop; otherwise we are generating a buffer which needs to
+ * be waited for during commit, so it has to go onto the appropriate
+ * journal buffer list.
+ */
+
+static void flush_descriptor(journal_t *journal,
+			     struct journal_head *descriptor,
+			     int offset, int write_op)
+{
+	journal_revoke_header_t *header;
+	struct buffer_head *bh = jh2bh(descriptor);
+
+	if (is_journal_aborted(journal)) {
+		put_bh(bh);
+		return;
+	}
+
+	header = (journal_revoke_header_t *) jh2bh(descriptor)->b_data;
+	header->r_count = cpu_to_be32(offset);
+	set_buffer_jwrite(bh);
+	BUFFER_TRACE(bh, "write");
+	set_buffer_dirty(bh);
+	ll_rw_block((write_op == WRITE) ? SWRITE : SWRITE_SYNC_PLUG, 1, &bh);
+}
+#endif
+
+/*
+ * Revoke support for recovery.
+ *
+ * Recovery needs to be able to:
+ *
+ *  record all revoke records, including the tid of the latest instance
+ *  of each revoke in the journal
+ *
+ *  check whether a given block in a given transaction should be replayed
+ *  (ie. has not been revoked by a revoke record in that or a subsequent
+ *  transaction)
+ *
+ *  empty the revoke table after recovery.
+ */
+
+/*
+ * First, setting revoke records.  We create a new revoke record for
+ * every block ever revoked in the log as we scan it for recovery, and
+ * we update the existing records if we find multiple revokes for a
+ * single block.
+ */
+
+int journal_set_revoke(journal_t *journal,
+		       unsigned int blocknr,
+		       tid_t sequence)
+{
+	struct jbd_revoke_record_s *record;
+
+	record = find_revoke_record(journal, blocknr);
+	if (record) {
+		/* If we have multiple occurrences, only record the
+		 * latest sequence number in the hashed record */
+		if (tid_gt(sequence, record->sequence))
+			record->sequence = sequence;
+		return 0;
+	}
+	return insert_revoke_hash(journal, blocknr, sequence);
+}
+
+/*
+ * Test revoke records.  For a given block referenced in the log, has
+ * that block been revoked?  A revoke record with a given transaction
+ * sequence number revokes all blocks in that transaction and earlier
+ * ones, but later transactions still need replayed.
+ */
+
+int journal_test_revoke(journal_t *journal,
+			unsigned int blocknr,
+			tid_t sequence)
+{
+	struct jbd_revoke_record_s *record;
+
+	record = find_revoke_record(journal, blocknr);
+	if (!record)
+		return 0;
+	if (tid_gt(sequence, record->sequence))
+		return 0;
+	return 1;
+}
+
+/*
+ * Finally, once recovery is over, we need to clear the revoke table so
+ * that it can be reused by the running filesystem.
+ */
+
+void journal_clear_revoke(journal_t *journal)
+{
+	int i;
+	struct list_head *hash_list;
+	struct jbd_revoke_record_s *record;
+	struct jbd_revoke_table_s *revoke;
+
+	revoke = journal->j_revoke;
+
+	for (i = 0; i < revoke->hash_size; i++) {
+		hash_list = &revoke->hash_table[i];
+		while (!list_empty(hash_list)) {
+			record = (struct jbd_revoke_record_s*) hash_list->next;
+			list_del(&record->hash);
+			kmem_cache_free(revoke_record_cache, record);
+		}
+	}
+}