log0log.c

这是linux下运行的mysql软件包,可用于linux 下安装 php + mysql + apach 的网络配置

	
	log_flush_do_unlocks(unlock);

	mutex_exit(&(log_sys->mutex));

	return;
	
do_waits:
	mutex_exit(&(log_sys->mutex));

	if (wait == LOG_WAIT_ONE_GROUP) {
		os_event_wait(log_sys->one_flushed_event);
	} else if (wait == LOG_WAIT_ALL_GROUPS) {
		os_event_wait(log_sys->no_flush_event);
	} else {
		ut_ad(wait == LOG_NO_WAIT);
	}			
}

/********************************************************************
Does a syncronous flush of the log buffer to disk. */

void
log_buffer_flush_to_disk(void)
/*==========================*/
{
	dulint	lsn;

	mutex_enter(&(log_sys->mutex));

	lsn = log_sys->lsn;

	mutex_exit(&(log_sys->mutex));

	log_write_up_to(lsn, LOG_WAIT_ALL_GROUPS, TRUE);
}

/********************************************************************
Tries to establish a big enough margin of free space in the log buffer, such
that a new log entry can be catenated without an immediate need for a flush. */
static
void
log_flush_margin(void)
/*==================*/
{
	ibool	do_flush	= FALSE;
	log_t*	log		= log_sys;
	dulint	lsn;

	mutex_enter(&(log->mutex));

	if (log->buf_free > log->max_buf_free) {
		
		if (log->n_pending_writes > 0) {
			/* A flush is running: hope that it will provide enough
			free space */
		} else {
			do_flush = TRUE;
			lsn = log->lsn;
		}
	}

	mutex_exit(&(log->mutex));

	if (do_flush) {
		log_write_up_to(lsn, LOG_NO_WAIT, FALSE);
	}
}

/********************************************************************
Advances the smallest lsn for which there are unflushed dirty blocks in the
buffer pool. NOTE: this function may only be called if the calling thread owns
no synchronization objects! */

ibool
log_preflush_pool_modified_pages(
/*=============================*/
				/* out: FALSE if there was a flush batch of
				the same type running, which means that we
				could not start this flush batch */
	dulint	new_oldest,	/* in: try to advance oldest_modified_lsn
				at least to this lsn */
	ibool	sync)		/* in: TRUE if synchronous operation is
				desired */
{
	ulint	n_pages;

	if (recv_recovery_on) {
		/* If the recovery is running, we must first apply all
		log records to their respective file pages to get the
		right modify lsn values to these pages: otherwise, there
		might be pages on disk which are not yet recovered to the
		current lsn, and even after calling this function, we could
		not know how up-to-date the disk version of the database is,
		and we could not make a new checkpoint on the basis of the
		info on the buffer pool only. */
	
		recv_apply_hashed_log_recs(TRUE);
	}

	n_pages = buf_flush_batch(BUF_FLUSH_LIST, ULINT_MAX, new_oldest);

	if (sync) {
		buf_flush_wait_batch_end(BUF_FLUSH_LIST);
	}

	if (n_pages == ULINT_UNDEFINED) {

		return(FALSE);
	}

	return(TRUE);
}

/**********************************************************
Completes a checkpoint. */
static
void
log_complete_checkpoint(void)
/*=========================*/
{
#ifdef UNIV_SYNC_DEBUG
	ut_ad(mutex_own(&(log_sys->mutex)));
#endif /* UNIV_SYNC_DEBUG */
	ut_ad(log_sys->n_pending_checkpoint_writes == 0);

	log_sys->next_checkpoint_no
			= ut_dulint_add(log_sys->next_checkpoint_no, 1);

	log_sys->last_checkpoint_lsn = log_sys->next_checkpoint_lsn;

	rw_lock_x_unlock_gen(&(log_sys->checkpoint_lock), LOG_CHECKPOINT);
}

/**********************************************************
Completes an asynchronous checkpoint info write i/o to a log file. */
static
void
log_io_complete_checkpoint(void)
/*============================*/
{
	mutex_enter(&(log_sys->mutex));

	ut_ad(log_sys->n_pending_checkpoint_writes > 0);
	
	log_sys->n_pending_checkpoint_writes--;

	if (log_sys->n_pending_checkpoint_writes == 0) {
		log_complete_checkpoint();
	}	
	
	mutex_exit(&(log_sys->mutex));
}

/***********************************************************************
Writes info to a checkpoint about a log group. */
static
void
log_checkpoint_set_nth_group_info(
/*==============================*/
	byte*	buf,	/* in: buffer for checkpoint info */
	ulint	n,	/* in: nth slot */
	ulint	file_no,/* in: archived file number */
	ulint	offset)	/* in: archived file offset */
{
	ut_ad(n < LOG_MAX_N_GROUPS);

	mach_write_to_4(buf + LOG_CHECKPOINT_GROUP_ARRAY
			+ 8 * n + LOG_CHECKPOINT_ARCHIVED_FILE_NO, file_no);
	mach_write_to_4(buf + LOG_CHECKPOINT_GROUP_ARRAY
			+ 8 * n + LOG_CHECKPOINT_ARCHIVED_OFFSET, offset);
}

/***********************************************************************
Gets info from a checkpoint about a log group. */

void
log_checkpoint_get_nth_group_info(
/*==============================*/
	byte*	buf,	/* in: buffer containing checkpoint info */
	ulint	n,	/* in: nth slot */
	ulint*	file_no,/* out: archived file number */
	ulint*	offset)	/* out: archived file offset */
{
	ut_ad(n < LOG_MAX_N_GROUPS);

	*file_no = mach_read_from_4(buf + LOG_CHECKPOINT_GROUP_ARRAY
				+ 8 * n + LOG_CHECKPOINT_ARCHIVED_FILE_NO);
	*offset = mach_read_from_4(buf + LOG_CHECKPOINT_GROUP_ARRAY
				+ 8 * n + LOG_CHECKPOINT_ARCHIVED_OFFSET);
}

/**********************************************************
Writes the checkpoint info to a log group header. */
static
void
log_group_checkpoint(
/*=================*/
	log_group_t*	group)	/* in: log group */
{
	log_group_t*	group2;
#ifdef UNIV_LOG_ARCHIVE
	dulint	archived_lsn;
	dulint	next_archived_lsn;
#endif /* UNIV_LOG_ARCHIVE */
	ulint	write_offset;
	ulint	fold;
	byte*	buf;
	ulint	i;

#ifdef UNIV_SYNC_DEBUG
	ut_ad(mutex_own(&(log_sys->mutex)));
#endif /* UNIV_SYNC_DEBUG */
	ut_a(LOG_CHECKPOINT_SIZE <= OS_FILE_LOG_BLOCK_SIZE);
	
	buf = group->checkpoint_buf;
	
	mach_write_to_8(buf + LOG_CHECKPOINT_NO, log_sys->next_checkpoint_no);
	mach_write_to_8(buf + LOG_CHECKPOINT_LSN,
						log_sys->next_checkpoint_lsn);

	mach_write_to_4(buf + LOG_CHECKPOINT_OFFSET,
			log_group_calc_lsn_offset(
					log_sys->next_checkpoint_lsn, group));
								
	mach_write_to_4(buf + LOG_CHECKPOINT_LOG_BUF_SIZE, log_sys->buf_size);

#ifdef UNIV_LOG_ARCHIVE
	if (log_sys->archiving_state == LOG_ARCH_OFF) {
		archived_lsn = ut_dulint_max;
	} else {
		archived_lsn = log_sys->archived_lsn;

		if (0 != ut_dulint_cmp(archived_lsn,
					log_sys->next_archived_lsn)) {
			next_archived_lsn = log_sys->next_archived_lsn;
			/* For debugging only */
		}	
	}

	mach_write_to_8(buf + LOG_CHECKPOINT_ARCHIVED_LSN, archived_lsn);
#else /* UNIV_LOG_ARCHIVE */
	mach_write_to_8(buf + LOG_CHECKPOINT_ARCHIVED_LSN, ut_dulint_max);
#endif /* UNIV_LOG_ARCHIVE */

	for (i = 0; i < LOG_MAX_N_GROUPS; i++) {
		log_checkpoint_set_nth_group_info(buf, i, 0, 0);
	}
	
	group2 = UT_LIST_GET_FIRST(log_sys->log_groups);

	while (group2) {
		log_checkpoint_set_nth_group_info(buf, group2->id,
#ifdef UNIV_LOG_ARCHIVE
						group2->archived_file_no,
						group2->archived_offset
#else /* UNIV_LOG_ARCHIVE */
						0, 0
#endif /* UNIV_LOG_ARCHIVE */
						);

		group2 = UT_LIST_GET_NEXT(log_groups, group2);
	}

	fold = ut_fold_binary(buf, LOG_CHECKPOINT_CHECKSUM_1);
	mach_write_to_4(buf + LOG_CHECKPOINT_CHECKSUM_1, fold);

	fold = ut_fold_binary(buf + LOG_CHECKPOINT_LSN,
			LOG_CHECKPOINT_CHECKSUM_2 - LOG_CHECKPOINT_LSN);
	mach_write_to_4(buf + LOG_CHECKPOINT_CHECKSUM_2, fold);

	/* Starting from InnoDB-3.23.50, we also write info on allocated
	size in the tablespace */

	mach_write_to_4(buf + LOG_CHECKPOINT_FSP_FREE_LIMIT,
						log_fsp_current_free_limit);

	mach_write_to_4(buf + LOG_CHECKPOINT_FSP_MAGIC_N,
					LOG_CHECKPOINT_FSP_MAGIC_N_VAL);

	/* We alternate the physical place of the checkpoint info in the first
	log file */
	
	if (ut_dulint_get_low(log_sys->next_checkpoint_no) % 2 == 0) {
		write_offset = LOG_CHECKPOINT_1;
	} else {
		write_offset = LOG_CHECKPOINT_2;
	}
					
	if (log_do_write) {
		if (log_sys->n_pending_checkpoint_writes == 0) {
	
			rw_lock_x_lock_gen(&(log_sys->checkpoint_lock),
							LOG_CHECKPOINT);
		}

		log_sys->n_pending_checkpoint_writes++;

		log_sys->n_log_ios++;
		
		/* We send as the last parameter the group machine address
		added with 1, as we want to distinguish between a normal log
		file write and a checkpoint field write */
		
		fil_io(OS_FILE_WRITE | OS_FILE_LOG, FALSE, group->space_id,
				write_offset / UNIV_PAGE_SIZE,
				write_offset % UNIV_PAGE_SIZE,
				OS_FILE_LOG_BLOCK_SIZE,
				buf, ((byte*)group + 1));

		ut_ad(((ulint)group & 0x1UL) == 0);
	}
}

/**********************************************************
Writes info to a buffer of a log group when log files are created in
backup restoration. */

void
log_reset_first_header_and_checkpoint(
/*==================================*/
	byte*	hdr_buf,/* in: buffer which will be written to the start
			of the first log file */
	dulint	start)	/* in: lsn of the start of the first log file;
			we pretend that there is a checkpoint at
			start + LOG_BLOCK_HDR_SIZE */
{
	ulint	fold;
	byte*	buf;
	dulint	lsn;
	
	mach_write_to_4(hdr_buf + LOG_GROUP_ID, 0);
	mach_write_to_8(hdr_buf + LOG_FILE_START_LSN, start);

	lsn = ut_dulint_add(start, LOG_BLOCK_HDR_SIZE);

	/* Write the label of ibbackup --restore */
	strcpy((char*) hdr_buf + LOG_FILE_WAS_CREATED_BY_HOT_BACKUP,
				"ibbackup ");
	ut_sprintf_timestamp(
			(char*) hdr_buf + (LOG_FILE_WAS_CREATED_BY_HOT_BACKUP
						+ (sizeof "ibbackup ") - 1));
	buf = hdr_buf + LOG_CHECKPOINT_1;
	
	mach_write_to_8(buf + LOG_CHECKPOINT_NO, ut_dulint_zero);
	mach_write_to_8(buf + LOG_CHECKPOINT_LSN, lsn);

	mach_write_to_4(buf + LOG_CHECKPOINT_OFFSET,
				LOG_FILE_HDR_SIZE + LOG_BLOCK_HDR_SIZE);
								
	mach_write_to_4(buf + LOG_CHECKPOINT_LOG_BUF_SIZE, 2 * 1024 * 1024);

	mach_write_to_8(buf + LOG_CHECKPOINT_ARCHIVED_LSN, ut_dulint_max);

	fold = ut_fold_binary(buf, LOG_CHECKPOINT_CHECKSUM_1);
	mach_write_to_4(buf + LOG_CHECKPOINT_CHECKSUM_1, fold);

	fold = ut_fold_binary(buf + LOG_CHECKPOINT_LSN,
			LOG_CHECKPOINT_CHECKSUM_2 - LOG_CHECKPOINT_LSN);
	mach_write_to_4(buf + LOG_CHECKPOINT_CHECKSUM_2, fold);

	/* Starting from InnoDB-3.23.50, we should also write info on
	allocated size in the tablespace, but unfortunately we do not
	know it here */
}

/**********************************************************
Reads a checkpoint info from a log group header to log_sys->checkpoint_buf. */

void
log_group_read_checkpoint_info(
/*===========================*/
	log_group_t*	group,	/* in: log group */
	ulint		field)	/* in: LOG_CHECKPOINT_1 or LOG_CHECKPOINT_2 */
{
#ifdef UNIV_SYNC_DEBUG
	ut_ad(mutex_own(&(log_sys->mutex)));
#endif /* UNIV_SYNC_DEBUG */

	log_sys->n_log_ios++;
	
	fil_io(OS_FILE_READ | OS_FILE_LOG, TRUE, group->space_id,
			field / UNIV_PAGE_SIZE, field % UNIV_PAGE_SIZE,
			OS_FILE_LOG_BLOCK_SIZE, log_sys->checkpoint_buf, NULL);
}

/**********************************************************
Writes checkpoint info to groups. */

void
log_groups_write_checkpoint_info(void)
/*==================================*/
{
	log_group_t*	group;

#ifdef UNIV_SYNC_DEBUG
	ut_ad(mutex_own(&(log_sys->mutex)));
#endif /* UNIV_SYNC_DEBUG */

	group = UT_LIST_GET_FIRST(log_sys->log_groups);

	while (group) {
		log_group_checkpoint(group);

		group = UT_LIST_GET_NEXT(log_groups, group);
	}
}

/**********************************************************
Makes a checkpoint. Note that this function does not flush dirty
blocks from the buffer pool: it only checks what is lsn of the oldest
modification in the pool, and writes information about the lsn in
log files. Use log_make_checkpoint_at to flush also the pool. */

ibool
log_checkpoint(
/*===========*/
				/* out: TRUE if success, FALSE if a checkpoint
				write was already running */
	ibool	sync,		/* in: TRUE if synchronous operation is
				desired */
	ibool	write_always)	/* in: the function normally checks if the
				the new checkpoint would have a greater
				lsn than the previous one: if not, then no
				physical write is done; by setting this
				parameter TRUE, a physical write will always be
				made to log files */
{
	dulint	oldest_lsn;

	if (recv_recovery_is_on()) {
		recv_apply_hashed_log_recs(TRUE);
	}

	if (srv_unix_file_flush_method != SRV_UNIX_NOSYNC) {
	        fil_flush_file_spaces(FIL_TABLESPACE);
	}

	mutex_enter(&(log_sys->mutex));

	oldest_lsn = log_buf_pool_get_oldest_modification();

	mutex_exit(&(log_sys->mutex));

	/* Because log also contains headers and dummy log records,
	if the buffer pool contains no dirty buffers, oldest_lsn
	gets the value log_sys->lsn from the previous function,
	and we must make sure that the log is flushed up to that
	lsn. If there are dirty buffers in the buffer pool, then our
	write-ahead-logging algorithm ensures that the log has been flushed
	up to oldest_lsn. */

	log_write_up_to(oldest_lsn, LOG_WAIT_ALL_GROUPS, TRUE);

	mutex_enter(&(log_sys->mutex));

	if (!write_always && ut_dulint_cmp(
			log_sys->last_checkpoint_lsn, oldest_lsn) >= 0) {

		mutex_exit(&(log_sys->mutex));

		return(TRUE);
	}

	ut_ad(ut_dulint_cmp(log_sys->written_to_all_lsn, oldest_lsn) >= 0);
	
	if (log_sys->n_pending_checkpoint_writes > 0) {
		/* A checkpoint write is running */

		mutex_exit(&(log_sys->mutex));

		if (sync) {
			/* Wait for the checkpoint write to complete */
			rw_lock_s_lock(&(log_sys->checkpoint_lock));
			rw_lock_s_unlock(&(log_sys->checkpoint_lock));
		}

		return(FALSE);
	}

	log_sys->next_checkpoint_lsn = oldest_lsn;

#ifdef UNIV_DEBUG
	if (log_debug_writes) {
		fprintf(stderr, "Making checkpoint no %lu at lsn %lu %lu\n",
			(ulong) ut_dulint_get_low(log_sys->next_checkpoint_no),
			(ulong) ut_dulint_get_high(oldest_lsn),
			(ulong) ut_dulint_get_low(oldest_lsn));
	}
#endif /* UNIV_DEBUG */

	log_groups_write_checkpoint_info();

	mutex_exit(&(log_sys->mutex));

	if (sync) {