commit.c

linux 内核源代码

		if (buffer_jbd(bh) && jh->b_jlist == BJ_Locked) {
			__jbd2_journal_unfile_buffer(jh);
			jbd_unlock_bh_state(bh);
			jbd2_journal_remove_journal_head(bh);
			put_bh(bh);
		} else {
			jbd_unlock_bh_state(bh);
		}
		put_bh(bh);
		cond_resched_lock(&journal->j_list_lock);
	}
	spin_unlock(&journal->j_list_lock);

	if (err)
		jbd2_journal_abort(journal, err);

	jbd2_journal_write_revoke_records(journal, commit_transaction);

	jbd_debug(3, "JBD: commit phase 2\n");

	/*
	 * 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:
	 */
	commit_transaction->t_state = T_COMMIT;

	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 for background writing. */

		if (is_journal_aborted(journal)) {
			JBUFFER_TRACE(jh, "journal is aborting: refile");
			jbd2_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 = jbd2_journal_get_descriptor_buffer(journal);
			if (!descriptor) {
				jbd2_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(JBD2_MAGIC_NUMBER);
			header->h_blocktype = cpu_to_be32(JBD2_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");
			jbd2_journal_file_buffer(descriptor, commit_transaction,
					BJ_LogCtl);
		}

		/* Where is the buffer to be written? */

		err = jbd2_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) {
			jbd2_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 jbd2_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: jbd2_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 = jbd2_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 |= JBD2_FLAG_ESCAPE;
		if (!first_tag)
			tag_flag |= JBD2_FLAG_SAME_UUID;

		tag = (journal_block_tag_t *) tagp;
		write_tag_block(tag_bytes, tag, jh2bh(jh)->b_blocknr);
		tag->t_flags = cpu_to_be32(tag_flag);
		tagp += tag_bytes;
		space_left -= tag_bytes;

		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 < tag_bytes + 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(JBD2_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 (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");
		jbd2_journal_unfile_buffer(journal, jh);

		/*
		 * ->t_iobuf_list should contain only dummy buffer_heads
		 * which were created by jbd2_journal_write_metadata_buffer().
		 */
		BUFFER_TRACE(bh, "dumping temporary bh");
		jbd2_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");
		jbd2_journal_file_buffer(jh, commit_transaction, BJ_Forget);
		/* Wake up any transactions which were waiting for this
		   IO to complete */
		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);
		jbd2_journal_unfile_buffer(journal, jh);
		jbd2_journal_put_journal_head(jh);
		__brelse(bh);		/* One for getblk */
		/* AKPM: bforget here */
	}

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

	if (journal_write_commit_record(journal, commit_transaction))
		err = -EIO;

	if (err)
		jbd2_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) {
			jbd2_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) {
			jbd2_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");
			__jbd2_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 jbd2_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");
			__jbd2_journal_insert_checkpoint(jh, commit_transaction);
			JBUFFER_TRACE(jh, "refile for checkpoint writeback");
			__jbd2_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");
			__jbd2_journal_refile_buffer(jh);
			if (!jh->b_transaction) {
				jbd_unlock_bh_state(bh);
				 /* needs a brelse */
				jbd2_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 borrow j_list_lock to protect
	 * journal->j_committing_transaction in __jbd2_journal_remove_checkpoint.
	 * Really, __jbd2_journal_remove_checkpoint should be using j_state_lock but
	 * it's a bit hassle to hold that across __jbd2_journal_remove_checkpoint
	 */
	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;
	spin_unlock(&journal->j_state_lock);

	if (commit_transaction->t_checkpoint_list == NULL &&
	    commit_transaction->t_checkpoint_io_list == NULL) {
		__jbd2_journal_drop_transaction(journal, commit_transaction);
	} 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);
}