commit.c

来自「Linux Kernel 2.6.9 for OMAP1710」· C语言 代码 · 共 842 行 · 第 1/2 页

			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_hard(journal);
			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 == ARRAY_SIZE(wbuf) ||
		    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, 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 (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 */
		wake_up_buffer(bh);
		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 (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 */
	}

	jbd_debug(3, "JBD: commit phase 6\n");

	if (is_journal_aborted(journal))
		goto skip_commit;

	/* 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. */

	descriptor = journal_get_descriptor_buffer(journal);
	if (!descriptor) {
		__journal_abort_hard(journal);
		goto skip_commit;
	}

	/* AKPM: buglet - add `i' to tmp! */
	for (i = 0; i < jh2bh(descriptor)->b_size; i += 512) {
		journal_header_t *tmp =
			(journal_header_t*)jh2bh(descriptor)->b_data;
		tmp->h_magic = cpu_to_be32(JFS_MAGIC_NUMBER);
		tmp->h_blocktype = cpu_to_be32(JFS_COMMIT_BLOCK);
		tmp->h_sequence = cpu_to_be32(commit_transaction->t_tid);
	}

	JBUFFER_TRACE(descriptor, "write commit block");
	{
		struct buffer_head *bh = jh2bh(descriptor);
		int ret;
		int barrier_done = 0;

		set_buffer_dirty(bh);
		if (journal->j_flags & JFS_BARRIER) {
			set_buffer_ordered(bh);
			barrier_done = 1;
		}
		ret = sync_dirty_buffer(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 */
			clear_buffer_ordered(bh);
			set_buffer_uptodate(bh);
			set_buffer_dirty(bh);
			ret = sync_dirty_buffer(bh);
		}
		if (unlikely(ret == -EIO))
			err = -EIO;
		put_bh(bh);		/* One for getblk() */
		journal_put_journal_head(descriptor);
	}

	/* 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. */

skip_commit: /* The journal should be unlocked by now. */

	if (err)
		__journal_abort_hard(journal);

	/*
	 * Call any callbacks that had been registered for handles in this
	 * transaction.  It is up to the callback to free any allocated
	 * memory.
	 *
	 * The spinlocking (t_jcb_lock) here is surely unnecessary...
	 */
	spin_lock(&commit_transaction->t_jcb_lock);
	if (!list_empty(&commit_transaction->t_jcb)) {
		struct list_head *p, *n;
		int error = is_journal_aborted(journal);

		list_for_each_safe(p, n, &commit_transaction->t_jcb) {
			struct journal_callback *jcb;

			jcb = list_entry(p, struct journal_callback, jcb_list);
			list_del(p);
			spin_unlock(&commit_transaction->t_jcb_lock);
			jcb->jcb_func(jcb, error);
			spin_lock(&commit_transaction->t_jcb_lock);
		}
	}
	spin_unlock(&commit_transaction->t_jcb_lock);

	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);

	while (commit_transaction->t_forget) {
		transaction_t *cp_transaction;
		struct buffer_head *bh;

		jh = commit_transaction->t_forget;
		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) {
			kfree(jh->b_committed_data);
			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) {
			kfree(jh->b_frozen_data);
			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);
			JBUFFER_TRACE(jh, "refile for checkpoint writeback");
			__journal_refile_buffer(jh);
			jbd_unlock_bh_state(bh);
		} else {
			J_ASSERT_BH(bh, !buffer_dirty(bh));
			J_ASSERT_JH(jh, jh->b_next_transaction == NULL);
			__journal_unfile_buffer(jh);
			jbd_unlock_bh_state(bh);
			journal_remove_journal_head(bh);  /* needs a brelse */
			release_buffer_page(bh);
		}
		spin_unlock(&journal->j_list_lock);
	}

	/* Done with this transaction! */

	jbd_debug(3, "JBD: commit phase 8\n");

	J_ASSERT(commit_transaction->t_state == T_COMMIT);

	/*
	 * This is a bit sleazy.  We borrow j_list_lock to protect
	 * journal->j_committing_transaction in __journal_remove_checkpoint.
	 * Really, __jornal_remove_checkpoint should be using j_state_lock but
	 * it's a bit hassle to hold that across __journal_remove_checkpoint
	 */
	spin_lock(&journal->j_state_lock);
	spin_lock(&journal->j_list_lock);
	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;
	spin_unlock(&journal->j_state_lock);

	if (commit_transaction->t_checkpoint_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);
}