srv0srv.h Support raw disk partitions as data files

srv0start.c	Support raw disk partitions as data files
srv0srv.c	Support raw disk partitions as data files
row0purge.c	< 4 GB rows, doublewrite, hang fixes
row0row.c	< 4 GB rows, doublewrite, hang fixes
row0sel.c	< 4 GB rows, doublewrite, hang fixes
row0uins.c	< 4 GB rows, doublewrite, hang fixes
row0umod.c	< 4 GB rows, doublewrite, hang fixes
row0undo.c	< 4 GB rows, doublewrite, hang fixes
row0upd.c	< 4 GB rows, doublewrite, hang fixes
srv0srv.c	< 4 GB rows, doublewrite, hang fixes
srv0start.c	< 4 GB rows, doublewrite, hang fixes
sync0rw.c	< 4 GB rows, doublewrite, hang fixes
sync0sync.c	< 4 GB rows, doublewrite, hang fixes
trx0purge.c	< 4 GB rows, doublewrite, hang fixes
trx0rec.c	< 4 GB rows, doublewrite, hang fixes
trx0sys.c	< 4 GB rows, doublewrite, hang fixes
btr0btr.c	< 4 GB rows, doublewrite, hang fixes
btr0cur.c	< 4 GB rows, doublewrite, hang fixes
buf0buf.c	< 4 GB rows, doublewrite, hang fixes
buf0flu.c	< 4 GB rows, doublewrite, hang fixes
buf0rea.c	< 4 GB rows, doublewrite, hang fixes
data0data.c	< 4 GB rows, doublewrite, hang fixes
fil0fil.c	< 4 GB rows, doublewrite, hang fixes
fsp0fsp.c	< 4 GB rows, doublewrite, hang fixes
ibuf0ibuf.c	< 4 GB rows, doublewrite, hang fixes
lock0lock.c	< 4 GB rows, doublewrite, hang fixes
log0log.c	< 4 GB rows, doublewrite, hang fixes
log0recv.c	< 4 GB rows, doublewrite, hang fixes
os0file.c	< 4 GB rows, doublewrite, hang fixes
page0cur.c	< 4 GB rows, doublewrite, hang fixes
pars0pars.c	< 4 GB rows, doublewrite, hang fixes
rem0cmp.c	< 4 GB rows, doublewrite, hang fixes
rem0rec.c	< 4 GB rows, doublewrite, hang fixes
row0ins.c	< 4 GB rows, doublewrite, hang fixes
row0mysql.c	< 4 GB rows, doublewrite, hang fixes
univ.i  	< 4 GB rows, doublewrite, hang fixes
data0data.ic	< 4 GB rows, doublewrite, hang fixes
mach0data.ic	< 4 GB rows, doublewrite, hang fixes
rem0rec.ic	< 4 GB rows, doublewrite, hang fixes
row0upd.ic	< 4 GB rows, doublewrite, hang fixes
trx0rec.ic	< 4 GB rows, doublewrite, hang fixes
rem0cmp.h	< 4 GB rows, doublewrite, hang fixes
rem0rec.h	< 4 GB rows, doublewrite, hang fixes
row0ins.h	< 4 GB rows, doublewrite, hang fixes
row0mysql.h	< 4 GB rows, doublewrite, hang fixes
row0row.h	< 4 GB rows, doublewrite, hang fixes
row0upd.h	< 4 GB rows, doublewrite, hang fixes
srv0srv.h	< 4 GB rows, doublewrite, hang fixes
sync0sync.h	< 4 GB rows, doublewrite, hang fixes
trx0rec.h	< 4 GB rows, doublewrite, hang fixes
trx0sys.h	< 4 GB rows, doublewrite, hang fixes
trx0types.h	< 4 GB rows, doublewrite, hang fixes
trx0undo.h	< 4 GB rows, doublewrite, hang fixes
ut0dbg.h	< 4 GB rows, doublewrite, hang fixes
ut0ut.h 	< 4 GB rows, doublewrite, hang fixes
btr0btr.h	< 4 GB rows, doublewrite, hang fixes
btr0cur.h	< 4 GB rows, doublewrite, hang fixes
buf0buf.h	< 4 GB rows, doublewrite, hang fixes
buf0flu.h	< 4 GB rows, doublewrite, hang fixes
data0data.h	< 4 GB rows, doublewrite, hang fixes
dict0mem.h	< 4 GB rows, doublewrite, hang fixes
fil0fil.h	< 4 GB rows, doublewrite, hang fixes
fsp0fsp.h	< 4 GB rows, doublewrite, hang fixes
os0file.h	< 4 GB rows, doublewrite, hang fixes


innobase/include/btr0btr.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/btr0cur.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/buf0buf.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/buf0flu.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/data0data.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/dict0mem.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/fil0fil.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/fsp0fsp.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/os0file.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/rem0cmp.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/rem0rec.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/row0ins.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/row0mysql.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/row0row.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/row0upd.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/sync0sync.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/trx0rec.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/trx0sys.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/trx0types.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/trx0undo.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/ut0dbg.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/ut0ut.h:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/data0data.ic:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/mach0data.ic:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/rem0rec.ic:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/row0upd.ic:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/trx0rec.ic:
  < 4 GB rows, doublewrite, hang fixes
innobase/include/univ.i:
  < 4 GB rows, doublewrite, hang fixes
innobase/btr/btr0btr.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/btr/btr0cur.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/buf/buf0buf.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/buf/buf0flu.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/buf/buf0rea.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/data/data0data.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/fil/fil0fil.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/fsp/fsp0fsp.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/ibuf/ibuf0ibuf.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/lock/lock0lock.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/log/log0log.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/log/log0recv.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/os/os0file.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/page/page0cur.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/pars/pars0pars.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/rem/rem0cmp.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/rem/rem0rec.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/row/row0ins.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/row/row0mysql.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/row/row0purge.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/row/row0row.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/row/row0sel.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/row/row0uins.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/row/row0umod.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/row/row0undo.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/row/row0upd.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/sync/sync0rw.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/sync/sync0sync.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/trx/trx0purge.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/trx/trx0rec.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/trx/trx0sys.c:
  < 4 GB rows, doublewrite, hang fixes
innobase/srv/srv0srv.c:
  Support raw disk partitions as data files
innobase/srv/srv0start.c:
  Support raw disk partitions as data files
innobase/include/srv0srv.h:
  Support raw disk partitions as data files

1 files changed, 318 insertions, 1 deletions

diff --git a/innobase/trx/trx0sys.c b/innobase/trx/trx0sys.c
index 99ec5b50237..b056975d28a 100644
--- a/innobase/trx/trx0sys.c
+++ b/innobase/trx/trx0sys.c
@@ -19,9 +19,326 @@ Created 3/26/1996 Heikki Tuuri
 #include "trx0undo.h"
 #include "srv0srv.h"
 #include "trx0purge.h"
+#include "log0log.h"
 
 /* The transaction system */
-trx_sys_t*	trx_sys 	= NULL;
+trx_sys_t*		trx_sys 	= NULL;
+trx_doublewrite_t*	trx_doublewrite = NULL;
+
+/********************************************************************
+Creates or initialializes the doublewrite buffer at a database start. */
+static
+void
+trx_doublewrite_init(
+/*=================*/
+	byte*	doublewrite)	/* in: pointer to the doublewrite buf
+				header on trx sys page */
+{
+	trx_doublewrite = mem_alloc(sizeof(trx_doublewrite_t));
+
+	mutex_create(&(trx_doublewrite->mutex));
+	mutex_set_level(&(trx_doublewrite->mutex), SYNC_DOUBLEWRITE);
+
+	trx_doublewrite->first_free = 0;
+
+	trx_doublewrite->block1 = mach_read_from_4(
+						doublewrite
+						+ TRX_SYS_DOUBLEWRITE_BLOCK1);
+	trx_doublewrite->block2 = mach_read_from_4(
+						doublewrite
+						+ TRX_SYS_DOUBLEWRITE_BLOCK2);
+	trx_doublewrite->write_buf_unaligned = 
+				ut_malloc(
+				(1 + 2 * TRX_SYS_DOUBLEWRITE_BLOCK_SIZE)
+				* UNIV_PAGE_SIZE);
+						
+	trx_doublewrite->write_buf = ut_align(
+					trx_doublewrite->write_buf_unaligned,
+					UNIV_PAGE_SIZE);
+	trx_doublewrite->buf_block_arr = mem_alloc(
+					2 * TRX_SYS_DOUBLEWRITE_BLOCK_SIZE
+					* sizeof(void*));
+}
+
+/********************************************************************
+Creates the doublewrite buffer at a database start. The header of the
+doublewrite buffer is placed on the trx system header page. */
+
+void
+trx_sys_create_doublewrite_buf(void)
+/*================================*/
+{
+	page_t*	page;
+	page_t*	page2;
+	page_t*	new_page;
+	byte*	doublewrite;
+	byte*	fseg_header;
+	ulint	page_no;
+	ulint	prev_page_no;
+	ulint	i;
+	mtr_t	mtr;
+
+	if (trx_doublewrite) {
+		/* Already inited */
+
+		return;
+	}
+
+start_again:	
+	mtr_start(&mtr);
+
+	page = buf_page_get(TRX_SYS_SPACE, TRX_SYS_PAGE_NO, RW_X_LATCH, &mtr);
+	buf_page_dbg_add_level(page, SYNC_NO_ORDER_CHECK);
+
+	doublewrite = page + TRX_SYS_DOUBLEWRITE;
+	
+	if (mach_read_from_4(doublewrite + TRX_SYS_DOUBLEWRITE_MAGIC)
+					== TRX_SYS_DOUBLEWRITE_MAGIC_N) {
+
+		/* The doublewrite buffer has already been created:
+		just read in some numbers */
+
+		trx_doublewrite_init(doublewrite);
+		
+		mtr_commit(&mtr);
+	} else {
+		fprintf(stderr,
+		"InnoDB: Doublewrite buffer not found: creating new\n");
+
+		if (buf_pool_get_curr_size() <
+					(2 * TRX_SYS_DOUBLEWRITE_BLOCK_SIZE
+						+ FSP_EXTENT_SIZE / 2 + 100)
+					* UNIV_PAGE_SIZE) {
+			fprintf(stderr,
+			"InnoDB: Cannot create doublewrite buffer: you must\n"
+			"InnoDB: increase your buffer pool size.\n"
+			"InnoDB: Cannot continue operation.\n");
+
+			exit(1);
+		}
+	
+		page2 = fseg_create(TRX_SYS_SPACE, TRX_SYS_PAGE_NO,
+			TRX_SYS_DOUBLEWRITE + TRX_SYS_DOUBLEWRITE_FSEG, &mtr);
+
+		/* fseg_create acquires a second latch on the page,
+		therefore we must declare it: */
+
+		buf_page_dbg_add_level(page2, SYNC_NO_ORDER_CHECK);
+
+		if (page2 == NULL) {
+			fprintf(stderr,
+			"InnoDB: Cannot create doublewrite buffer: you must\n"
+			"InnoDB: increase your tablespace size.\n"
+			"InnoDB: Cannot continue operation.\n");
+
+			/* We exit without committing the mtr to prevent
+			its modifications to the database getting to disk */
+			
+			exit(1);
+		}
+
+		fseg_header = page + TRX_SYS_DOUBLEWRITE
+						+ TRX_SYS_DOUBLEWRITE_FSEG;
+		prev_page_no = 0;
+
+		for (i = 0; i < 2 * TRX_SYS_DOUBLEWRITE_BLOCK_SIZE
+						+ FSP_EXTENT_SIZE / 2; i++) {
+			page_no = fseg_alloc_free_page(fseg_header,
+			 				prev_page_no + 1,
+							FSP_UP, &mtr);
+			if (page_no == FIL_NULL) {
+				fprintf(stderr,
+			"InnoDB: Cannot create doublewrite buffer: you must\n"
+			"InnoDB: increase your tablespace size.\n"
+			"InnoDB: Cannot continue operation.\n");
+
+				exit(1);
+			}
+
+			/* We read the allocated pages to the buffer pool;
+			when they are written to disk in a flush, the space
+			id and page number fields are also written to the
+			pages. When we at database startup read pages
+			from the doublewrite buffer, we know that if the
+			space id and page number in them are the same as
+			the page position in the tablespace, then the page
+			has not been written to in doublewrite. */
+			
+			new_page = buf_page_get(TRX_SYS_SPACE, page_no,
+							RW_X_LATCH, &mtr);
+			buf_page_dbg_add_level(new_page, SYNC_NO_ORDER_CHECK);
+
+			/* Make a dummy change to the page to ensure it will
+			be written to disk in a flush */
+
+			mlog_write_ulint(new_page + FIL_PAGE_DATA,
+					TRX_SYS_DOUBLEWRITE_MAGIC_N,
+					MLOG_4BYTES, &mtr);
+
+			if (i == FSP_EXTENT_SIZE / 2) {
+				mlog_write_ulint(doublewrite
+						+ TRX_SYS_DOUBLEWRITE_BLOCK1,
+						page_no, MLOG_4BYTES, &mtr);
+				mlog_write_ulint(doublewrite
+						+ TRX_SYS_DOUBLEWRITE_REPEAT
+						+ TRX_SYS_DOUBLEWRITE_BLOCK1,
+						page_no, MLOG_4BYTES, &mtr);
+			} else if (i == FSP_EXTENT_SIZE / 2
+					+ TRX_SYS_DOUBLEWRITE_BLOCK_SIZE) {
+				mlog_write_ulint(doublewrite
+						+ TRX_SYS_DOUBLEWRITE_BLOCK2,
+						page_no, MLOG_4BYTES, &mtr);
+				mlog_write_ulint(doublewrite
+						+ TRX_SYS_DOUBLEWRITE_REPEAT
+						+ TRX_SYS_DOUBLEWRITE_BLOCK2,
+						page_no, MLOG_4BYTES, &mtr);
+			} else if (i > FSP_EXTENT_SIZE / 2) {
+				ut_a(page_no == prev_page_no + 1);
+			}
+
+			prev_page_no = page_no;
+		}
+
+		mlog_write_ulint(doublewrite + TRX_SYS_DOUBLEWRITE_MAGIC,
+				TRX_SYS_DOUBLEWRITE_MAGIC_N, MLOG_4BYTES, &mtr);
+		mlog_write_ulint(doublewrite + TRX_SYS_DOUBLEWRITE_MAGIC
+						+ TRX_SYS_DOUBLEWRITE_REPEAT,
+				TRX_SYS_DOUBLEWRITE_MAGIC_N, MLOG_4BYTES, &mtr);
+		mtr_commit(&mtr);
+		
+		/* Flush the modified pages to disk and make a checkpoint */
+		log_make_checkpoint_at(ut_dulint_max, TRUE);
+
+		fprintf(stderr, "InnoDB: Doublewrite buffer created\n");
+
+		goto start_again;
+	}
+}
+
+/********************************************************************
+At a database startup uses a possible doublewrite buffer to restore
+half-written pages in the data files. */
+
+void
+trx_sys_doublewrite_restore_corrupt_pages(void)
+/*===========================================*/
+{
+	byte*	buf;
+	byte*	read_buf;
+	byte*	unaligned_read_buf;
+	ulint	block1;
+	ulint	block2;
+	byte*	page;
+	byte*	doublewrite;
+	ulint	space_id;
+	ulint	page_no;
+	ulint	i;
+	
+	/* We do the file i/o past the buffer pool */
+
+	unaligned_read_buf = ut_malloc(2 * UNIV_PAGE_SIZE);
+	read_buf = ut_align(unaligned_read_buf, UNIV_PAGE_SIZE);	
+
+	/* Read the trx sys header to check if we are using the
+	doublewrite buffer */
+
+	fil_io(OS_FILE_READ, TRUE, TRX_SYS_SPACE, TRX_SYS_PAGE_NO, 0,
+					UNIV_PAGE_SIZE, read_buf, NULL);
+
+	doublewrite = read_buf + TRX_SYS_DOUBLEWRITE;
+
+	if (mach_read_from_4(doublewrite + TRX_SYS_DOUBLEWRITE_MAGIC)
+					== TRX_SYS_DOUBLEWRITE_MAGIC_N) {
+		/* The doublewrite buffer has been created */
+		
+		trx_doublewrite_init(doublewrite);
+
+		block1 = trx_doublewrite->block1;
+		block2 = trx_doublewrite->block2;
+
+		buf = trx_doublewrite->write_buf;
+	} else {
+		goto leave_func;
+	}
+
+	/* Read the pages from the doublewrite buffer to memory */
+
+	fil_io(OS_FILE_READ, TRUE, TRX_SYS_SPACE, block1, 0,
+			TRX_SYS_DOUBLEWRITE_BLOCK_SIZE * UNIV_PAGE_SIZE,
+			buf, NULL);
+	fil_io(OS_FILE_READ, TRUE, TRX_SYS_SPACE, block2, 0,
+			TRX_SYS_DOUBLEWRITE_BLOCK_SIZE * UNIV_PAGE_SIZE,
+			buf + TRX_SYS_DOUBLEWRITE_BLOCK_SIZE * UNIV_PAGE_SIZE,
+			NULL);
+	/* Check if any of these pages is half-written in data files, in the
+	intended position */
+
+	page = buf;
+	
+	for (i = 0; i < TRX_SYS_DOUBLEWRITE_BLOCK_SIZE * 2; i++) {
+		
+		space_id = mach_read_from_4(page + FIL_PAGE_SPACE);
+		page_no = mach_read_from_4(page + FIL_PAGE_OFFSET);
+
+		if (!fil_check_adress_in_tablespace(space_id, page_no)) {
+		  	fprintf(stderr,
+	"InnoDB: Warning: an inconsistent page in the doublewrite buffer\n"
+	"InnoDB: space id %lu page number %lu, %lu'th page in dblwr buf.\n",
+				space_id, page_no, i);
+		
+		} else if (space_id == TRX_SYS_SPACE
+		    && (  (page_no >= block1
+			   && page_no
+				< block1 + TRX_SYS_DOUBLEWRITE_BLOCK_SIZE)
+		        || (page_no >= block2
+			   && page_no
+				< block2 + TRX_SYS_DOUBLEWRITE_BLOCK_SIZE))) {
+
+			/* It is an unwritten doublewrite buffer page:
+			do nothing */
+
+		} else {
+			/* Read in the actual page from the data files */
+			
+			fil_io(OS_FILE_READ, TRUE, space_id, page_no, 0,
+					UNIV_PAGE_SIZE, read_buf, NULL);
+			/* Check if the page is corrupt */
+
+			if (buf_page_is_corrupted(read_buf)) {
+
+		  		fprintf(stderr,
+		"InnoDB: Warning: database page corruption or a failed\n"
+		"InnoDB: file read of page %lu.\n", page_no);
+		  		fprintf(stderr,
+		"InnoDB: Trying to recover it from the doublewrite buffer.\n");
+				
+				if (buf_page_is_corrupted(page)) {
+		  			fprintf(stderr,
+		"InnoDB: Also the page in the doublewrite buffer is corrupt.\n"
+		"InnoDB: Cannot continue operation.\n");
+					exit(1);
+				}
+
+				/* Write the good page from the
+				doublewrite buffer to the intended
+				position */
+
+				fil_io(OS_FILE_WRITE, TRUE, space_id,
+					page_no, 0,
+					UNIV_PAGE_SIZE, page, NULL);
+		  		fprintf(stderr,
+		"InnoDB: Recovered the page from the doublewrite buffer.\n");
+			}
+		}
+
+		page += UNIV_PAGE_SIZE;
+	}
+
+	fil_flush_file_spaces(FIL_TABLESPACE);
+	
+leave_func:
+	ut_free(unaligned_read_buf);
+}
 
 /********************************************************************
 Checks that trx is in the trx list. */