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

		fprintf(stderr, "Log flushed to group %lu\n", (ulong) group->id);
	}
#endif /* UNIV_DEBUG */
	return(0);
}

/**********************************************************
Checks if a flush is completed and does the completion routine if yes. */
static
ulint
log_sys_check_flush_completion(void)
/*================================*/
			/* out: LOG_UNLOCK_FLUSH_LOCK or 0 */
{
	ulint	move_start;
	ulint	move_end;

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

	if (log_sys->n_pending_writes == 0) {
	
		log_sys->written_to_all_lsn = log_sys->write_lsn;
		log_sys->buf_next_to_write = log_sys->write_end_offset;

		if (log_sys->write_end_offset > log_sys->max_buf_free / 2) {
			/* Move the log buffer content to the start of the
			buffer */

			move_start = ut_calc_align_down(
						log_sys->write_end_offset,
						OS_FILE_LOG_BLOCK_SIZE);
			move_end = ut_calc_align(log_sys->buf_free,
						OS_FILE_LOG_BLOCK_SIZE);

			ut_memmove(log_sys->buf, log_sys->buf + move_start,
						move_end - move_start);
			log_sys->buf_free -= move_start;
		
			log_sys->buf_next_to_write -= move_start;
		}

		return(LOG_UNLOCK_FLUSH_LOCK);
	}

	return(0);
}

/**********************************************************
Completes an i/o to a log file. */

void
log_io_complete(
/*============*/
	log_group_t*	group)	/* in: log group or a dummy pointer */
{
	ulint	unlock;

#ifdef UNIV_LOG_ARCHIVE
	if ((byte*)group == &log_archive_io) {
		/* It was an archive write */

		log_io_complete_archive();

		return;
	}
#endif /* UNIV_LOG_ARCHIVE */

	if ((ulint)group & 0x1UL) {
		/* It was a checkpoint write */
		group = (log_group_t*)((ulint)group - 1);

		if (srv_unix_file_flush_method != SRV_UNIX_O_DSYNC
		   && srv_unix_file_flush_method != SRV_UNIX_NOSYNC) {
		
		        fil_flush(group->space_id);
		}

#ifdef UNIV_DEBUG
		if (log_debug_writes) {
			fprintf(stderr,
				"Checkpoint info written to group %lu\n",
				group->id);
		}
#endif /* UNIV_DEBUG */
		log_io_complete_checkpoint();

		return;
	}

	ut_error;	/* We currently use synchronous writing of the
			logs and cannot end up here! */

	if (srv_unix_file_flush_method != SRV_UNIX_O_DSYNC
	    && srv_unix_file_flush_method != SRV_UNIX_NOSYNC
	    && srv_flush_log_at_trx_commit != 2) {

	        fil_flush(group->space_id);
	}

	mutex_enter(&(log_sys->mutex));

	ut_a(group->n_pending_writes > 0);
	ut_a(log_sys->n_pending_writes > 0);
	
	group->n_pending_writes--;
	log_sys->n_pending_writes--;

	unlock = log_group_check_flush_completion(group);
	unlock = unlock | log_sys_check_flush_completion();
	
	log_flush_do_unlocks(unlock);

	mutex_exit(&(log_sys->mutex));
}

/**********************************************************
Writes a log file header to a log file space. */
static
void
log_group_file_header_flush(
/*========================*/
	log_group_t*	group,		/* in: log group */
	ulint		nth_file,	/* in: header to the nth file in the
					log file space */
	dulint		start_lsn)	/* in: log file data starts at this
					lsn */
{
	byte*	buf;
	ulint	dest_offset;
#ifdef UNIV_SYNC_DEBUG
	ut_ad(mutex_own(&(log_sys->mutex)));
#endif /* UNIV_SYNC_DEBUG */

	ut_a(nth_file < group->n_files);

	buf = *(group->file_header_bufs + nth_file);

	mach_write_to_4(buf + LOG_GROUP_ID, group->id);
	mach_write_to_8(buf + LOG_FILE_START_LSN, start_lsn);

	/* Wipe over possible label of ibbackup --restore */
	memcpy(buf + LOG_FILE_WAS_CREATED_BY_HOT_BACKUP, "    ", 4);

	dest_offset = nth_file * group->file_size;

#ifdef UNIV_DEBUG
	if (log_debug_writes) {
		fprintf(stderr,
			"Writing log file header to group %lu file %lu\n",
			(ulong) group->id, (ulong) nth_file);
	}
#endif /* UNIV_DEBUG */
	if (log_do_write) {
		log_sys->n_log_ios++;	
		
                srv_os_log_pending_writes++;
                
		fil_io(OS_FILE_WRITE | OS_FILE_LOG, TRUE, group->space_id,
				dest_offset / UNIV_PAGE_SIZE,
				dest_offset % UNIV_PAGE_SIZE,
				OS_FILE_LOG_BLOCK_SIZE,
				buf, group);

                srv_os_log_pending_writes--;
	}
}

/**********************************************************
Stores a 4-byte checksum to the trailer checksum field of a log block
before writing it to a log file. This checksum is used in recovery to
check the consistency of a log block. */
static
void
log_block_store_checksum(
/*=====================*/
	byte*	block)	/* in/out: pointer to a log block */
{
	log_block_set_checksum(block, log_block_calc_checksum(block));
}
	
/**********************************************************
Writes a buffer to a log file group. */

void
log_group_write_buf(
/*================*/
	log_group_t*	group,		/* in: log group */
	byte*		buf,		/* in: buffer */
	ulint		len,		/* in: buffer len; must be divisible
					by OS_FILE_LOG_BLOCK_SIZE */
	dulint		start_lsn,	/* in: start lsn of the buffer; must
					be divisible by
					OS_FILE_LOG_BLOCK_SIZE */
	ulint		new_data_offset)/* in: start offset of new data in
					buf: this parameter is used to decide
					if we have to write a new log file
					header */
{
	ulint	write_len;
	ibool	write_header;
	ulint	next_offset;
	ulint	i;
	
#ifdef UNIV_SYNC_DEBUG
	ut_ad(mutex_own(&(log_sys->mutex)));
#endif /* UNIV_SYNC_DEBUG */
	ut_a(len % OS_FILE_LOG_BLOCK_SIZE == 0);
	ut_a(ut_dulint_get_low(start_lsn) % OS_FILE_LOG_BLOCK_SIZE == 0);

	if (new_data_offset == 0) {
		write_header = TRUE;
	} else {
		write_header = FALSE;
	}
loop:
	if (len == 0) {

		return;
	}

	next_offset = log_group_calc_lsn_offset(start_lsn, group);

	if ((next_offset % group->file_size == LOG_FILE_HDR_SIZE)
	   						&& write_header) {
		/* We start to write a new log file instance in the group */

		log_group_file_header_flush(group,
				next_offset / group->file_size, start_lsn);
                srv_os_log_written+= OS_FILE_LOG_BLOCK_SIZE;
                srv_log_writes++;
	}

	if ((next_offset % group->file_size) + len > group->file_size) {

		write_len = group->file_size
					- (next_offset % group->file_size);
	} else {
		write_len = len;
	}

#ifdef UNIV_DEBUG
	if (log_debug_writes) {

		fprintf(stderr,
		"Writing log file segment to group %lu offset %lu len %lu\n"
		"start lsn %lu %lu\n"
		"First block n:o %lu last block n:o %lu\n",
			(ulong) group->id, (ulong) next_offset,
		        (ulong) write_len,
			(ulong) ut_dulint_get_high(start_lsn),
			(ulong) ut_dulint_get_low(start_lsn),
			(ulong) log_block_get_hdr_no(buf),
			(ulong) log_block_get_hdr_no(
				buf + write_len - OS_FILE_LOG_BLOCK_SIZE));
		ut_a(log_block_get_hdr_no(buf)
			== log_block_convert_lsn_to_no(start_lsn));
		
		for (i = 0; i < write_len / OS_FILE_LOG_BLOCK_SIZE; i++) {

			ut_a(log_block_get_hdr_no(buf) + i
				== log_block_get_hdr_no(buf
					+ i * OS_FILE_LOG_BLOCK_SIZE));
		}
	}
#endif /* UNIV_DEBUG */
	/* Calculate the checksums for each log block and write them to
	the trailer fields of the log blocks */

	for (i = 0; i < write_len / OS_FILE_LOG_BLOCK_SIZE; i++) {
		log_block_store_checksum(buf + i * OS_FILE_LOG_BLOCK_SIZE);
	}

	if (log_do_write) {
		log_sys->n_log_ios++;	

                srv_os_log_pending_writes++;                
                
		fil_io(OS_FILE_WRITE | OS_FILE_LOG, TRUE, group->space_id,
			next_offset / UNIV_PAGE_SIZE,
			next_offset % UNIV_PAGE_SIZE, write_len, buf, group);

                srv_os_log_pending_writes--;

                srv_os_log_written+= write_len; 
                srv_log_writes++;
	}

	if (write_len < len) {
		start_lsn = ut_dulint_add(start_lsn, write_len);
		len -= write_len;
		buf += write_len;

		write_header = TRUE;

		goto loop;
	}
}

/**********************************************************
This function is called, e.g., when a transaction wants to commit. It checks
that the log has been written to the log file up to the last log entry written
by the transaction. If there is a flush running, it waits and checks if the
flush flushed enough. If not, starts a new flush. */

void
log_write_up_to(
/*============*/
	dulint	lsn,	/* in: log sequence number up to which the log should
			be written, ut_dulint_max if not specified */
	ulint	wait,	/* in: LOG_NO_WAIT, LOG_WAIT_ONE_GROUP,
			or LOG_WAIT_ALL_GROUPS */
	ibool	flush_to_disk)
			/* in: TRUE if we want the written log also to be
			flushed to disk */
{
	log_group_t*	group;
	ulint		start_offset;
	ulint		end_offset;
	ulint		area_start;
	ulint		area_end;
	ulint		loop_count;
	ulint		unlock;

	if (recv_no_ibuf_operations) {
		/* Recovery is running and no operations on the log files are
		allowed yet (the variable name .._no_ibuf_.. is misleading) */

		return;
	}

	loop_count = 0;
loop:
	loop_count++;

	ut_ad(loop_count < 5);

	if (loop_count > 2) {
/*		fprintf(stderr, "Log loop count %lu\n", loop_count); */
	}
	
	mutex_enter(&(log_sys->mutex));

	if (flush_to_disk
	    && ut_dulint_cmp(log_sys->flushed_to_disk_lsn, lsn) >= 0) {
	    
		mutex_exit(&(log_sys->mutex));

		return;
	}

	if (!flush_to_disk
	    && (ut_dulint_cmp(log_sys->written_to_all_lsn, lsn) >= 0
	    	|| (ut_dulint_cmp(log_sys->written_to_some_lsn, lsn) >= 0
	        			&& wait != LOG_WAIT_ALL_GROUPS))) {

		mutex_exit(&(log_sys->mutex));

		return;
	}
	
	if (log_sys->n_pending_writes > 0) {
		/* A write (+ possibly flush to disk) is running */

		if (flush_to_disk
		    && ut_dulint_cmp(log_sys->current_flush_lsn, lsn) >= 0) {
			/* The write + flush will write enough: wait for it to
			complete  */
	    
			goto do_waits;
		}

		if (!flush_to_disk
		    && ut_dulint_cmp(log_sys->write_lsn, lsn) >= 0) {
			/* The write will write enough: wait for it to
			complete  */

			goto do_waits;
		}
		
		mutex_exit(&(log_sys->mutex));

		/* Wait for the write to complete and try to start a new
		write */

		os_event_wait(log_sys->no_flush_event);

		goto loop;
	}

	if (!flush_to_disk
			&& log_sys->buf_free == log_sys->buf_next_to_write) {
		/* Nothing to write and no flush to disk requested */

		mutex_exit(&(log_sys->mutex));

		return;
	}

#ifdef UNIV_DEBUG
	if (log_debug_writes) {
		fprintf(stderr,
			"Writing log from %lu %lu up to lsn %lu %lu\n",
			(ulong) ut_dulint_get_high(log_sys->written_to_all_lsn),
			(ulong) ut_dulint_get_low(log_sys->written_to_all_lsn),
			(ulong) ut_dulint_get_high(log_sys->lsn),
			(ulong)	ut_dulint_get_low(log_sys->lsn));
	}
#endif /* UNIV_DEBUG */
	log_sys->n_pending_writes++;

	group = UT_LIST_GET_FIRST(log_sys->log_groups);
	group->n_pending_writes++; 	/* We assume here that we have only
					one log group! */

	os_event_reset(log_sys->no_flush_event);
	os_event_reset(log_sys->one_flushed_event);

	start_offset = log_sys->buf_next_to_write;
	end_offset = log_sys->buf_free;

	area_start = ut_calc_align_down(start_offset, OS_FILE_LOG_BLOCK_SIZE);
	area_end = ut_calc_align(end_offset, OS_FILE_LOG_BLOCK_SIZE);

	ut_ad(area_end - area_start > 0);

	log_sys->write_lsn = log_sys->lsn;

	if (flush_to_disk) {
		log_sys->current_flush_lsn = log_sys->lsn;
	}

	log_sys->one_flushed = FALSE;
	
	log_block_set_flush_bit(log_sys->buf + area_start, TRUE);
	log_block_set_checkpoint_no(
			log_sys->buf + area_end - OS_FILE_LOG_BLOCK_SIZE,
			log_sys->next_checkpoint_no);
	
	/* Copy the last, incompletely written, log block a log block length
	up, so that when the flush operation writes from the log buffer, the
	segment to write will not be changed by writers to the log */
	
	ut_memcpy(log_sys->buf + area_end,
			log_sys->buf + area_end - OS_FILE_LOG_BLOCK_SIZE,
			OS_FILE_LOG_BLOCK_SIZE);

	log_sys->buf_free += OS_FILE_LOG_BLOCK_SIZE;
	log_sys->write_end_offset = log_sys->buf_free;

	group = UT_LIST_GET_FIRST(log_sys->log_groups);

	/* Do the write to the log files */

	while (group) {
		log_group_write_buf(group,
			log_sys->buf + area_start,
			area_end - area_start,
			ut_dulint_align_down(log_sys->written_to_all_lsn,
						OS_FILE_LOG_BLOCK_SIZE),
			start_offset - area_start);

		log_group_set_fields(group, log_sys->write_lsn);
						
		group = UT_LIST_GET_NEXT(log_groups, group);
	}

	mutex_exit(&(log_sys->mutex));

	if (srv_unix_file_flush_method == SRV_UNIX_O_DSYNC) {
		/* O_DSYNC means the OS did not buffer the log file at all:
		so we have also flushed to disk what we have written */

		log_sys->flushed_to_disk_lsn = log_sys->write_lsn;

	} else if (flush_to_disk) {

		group = UT_LIST_GET_FIRST(log_sys->log_groups);

	        fil_flush(group->space_id);
		log_sys->flushed_to_disk_lsn = log_sys->write_lsn;
	}

	mutex_enter(&(log_sys->mutex));

	group = UT_LIST_GET_FIRST(log_sys->log_groups);

	ut_a(group->n_pending_writes == 1);
	ut_a(log_sys->n_pending_writes == 1);
	
	group->n_pending_writes--;
	log_sys->n_pending_writes--;

	unlock = log_group_check_flush_completion(group);
	unlock = unlock | log_sys_check_flush_completion();