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

2 files changed, 106 insertions, 10 deletions

diff --git a/innobase/rem/rem0cmp.c b/innobase/rem/rem0cmp.c
index 78f4e450269..cdf1f363946 100644
--- a/innobase/rem/rem0cmp.c
+++ b/innobase/rem/rem0cmp.c
@@ -295,14 +295,18 @@ This function is used to compare a data tuple to a physical record.
 Only dtuple->n_fields_cmp first fields are taken into account for
 the the data tuple! If we denote by n = n_fields_cmp, then rec must
 have either m >= n fields, or it must differ from dtuple in some of
-the m fields rec has. */
+the m fields rec has. If rec has an externally stored field we do not
+compare it but return with value 0 if such a comparison should be
+made. */
 
 int
 cmp_dtuple_rec_with_match(
 /*======================*/	
 				/* out: 1, 0, -1, if dtuple is greater, equal, 
 				less than rec, respectively, when only the 
-				common first fields are compared */
+				common first fields are compared, or
+				until the first externally stored field in
+				rec */
 	dtuple_t*	dtuple,	/* in: data tuple */
 	rec_t*		rec,	/* in: physical record which differs from
 				dtuple in some of the common fields, or which
@@ -344,7 +348,8 @@ cmp_dtuple_rec_with_match(
 	ut_ad(cur_field <= dtuple_get_n_fields_cmp(dtuple));
 	ut_ad(cur_field <= rec_get_n_fields(rec));
 
-	/* Match fields in a loop; stop if we run out of fields in dtuple */
+	/* Match fields in a loop; stop if we run out of fields in dtuple
+	or find an externally stored field */
 
 	while (cur_field < dtuple_get_n_fields_cmp(dtuple)) {
 
@@ -357,7 +362,8 @@ cmp_dtuple_rec_with_match(
 
 		/* If we have matched yet 0 bytes, it may be that one or
 		both the fields are SQL null, or the record or dtuple may be
-		the predefined minimum record */
+		the predefined minimum record, or the field is externally
+		stored */
 
 		if (cur_bytes == 0) {
 			if (cur_field == 0) {
@@ -384,6 +390,15 @@ cmp_dtuple_rec_with_match(
 				}
 			}
 
+			if (rec_get_nth_field_extern_bit(rec, cur_field)) {
+				/* We do not compare to an externally
+				stored field */
+
+				ret = 0;
+
+				goto order_resolved;
+			}
+
 		    	if (dtuple_f_len == UNIV_SQL_NULL
 		            || rec_f_len == UNIV_SQL_NULL) {
 
@@ -604,7 +619,8 @@ cmp_dtuple_rec_prefix_equal(
 
 /*****************************************************************
 This function is used to compare two physical records. Only the common
-first fields are compared. */
+first fields are compared, and if an externally stored field is
+encountered, then 0 is returned. */
 
 int
 cmp_rec_rec_with_match(
@@ -688,8 +704,18 @@ cmp_rec_rec_with_match(
 	
 					goto order_resolved;
 				}
-			} 
+			}
+
+			if (rec_get_nth_field_extern_bit(rec1, cur_field)
+			    || rec_get_nth_field_extern_bit(rec2, cur_field)) {
+				/* We do not compare to an externally
+				stored field */
 
+				ret = 0;
+
+				goto order_resolved;
+			}
+			
 		    	if (rec1_f_len == UNIV_SQL_NULL
 		            || rec2_f_len == UNIV_SQL_NULL) {
 
@@ -812,7 +838,8 @@ order_resolved:
 Used in debug checking of cmp_dtuple_... .
 This function is used to compare a data tuple to a physical record. If
 dtuple has n fields then rec must have either m >= n fields, or it must
-differ from dtuple in some of the m fields rec has. */
+differ from dtuple in some of the m fields rec has. If encounters an
+externally stored field, returns 0. */
 static
 int
 cmp_debug_dtuple_rec_with_match(
@@ -882,6 +909,14 @@ cmp_debug_dtuple_rec_with_match(
 		
 		rec_f_data = rec_get_nth_field(rec, cur_field, &rec_f_len);
 
+		if (rec_get_nth_field_extern_bit(rec, cur_field)) {
+			/* We do not compare to an externally stored field */
+
+			ret = 0;
+
+			goto order_resolved;
+		}
+
 		ret = cmp_data_data(cur_type, dtuple_f_data, dtuple_f_len,
 							rec_f_data, rec_f_len);
 		if (ret != 0) {
diff --git a/innobase/rem/rem0rec.c b/innobase/rem/rem0rec.c
index 9ddfe7a4b9a..88009f2f5c9 100644
--- a/innobase/rem/rem0rec.c
+++ b/innobase/rem/rem0rec.c
@@ -1,7 +1,7 @@
 /************************************************************************
 Record manager
 
-(c) 1994-1996 Innobase Oy
+(c) 1994-2001 Innobase Oy
 
 Created 5/30/1994 Heikki Tuuri
 *************************************************************************/
@@ -12,6 +12,9 @@ Created 5/30/1994 Heikki Tuuri
 #include "rem0rec.ic"
 #endif
 
+#include "mtr0mtr.h"
+#include "mtr0log.h"
+
 /*			PHYSICAL RECORD
 			===============
 
@@ -21,7 +24,10 @@ found in index pages of the database, has the following format
 represented on a higher text line):
 
 | offset of the end of the last field of data, the most significant
-  bit is set to 1 if and only if the field is SQL-null | 
+  bit is set to 1 if and only if the field is SQL-null,
+  if the offset is 2-byte, then the second most significant
+  bit is set to 1 if the field is stored on another page:
+  mostly this will occur in the case of big BLOB fields |
 ... 
 | offset of the end of the first field of data + the SQL-null bit |
 | 4 bits used to delete mark a record, and mark a predefined
@@ -122,7 +128,8 @@ rec_get_nth_field(
 			return(rec + os);
 		}
 
-		next_os = next_os & ~REC_2BYTE_SQL_NULL_MASK;
+		next_os = next_os & ~(REC_2BYTE_SQL_NULL_MASK
+						| REC_2BYTE_EXTERN_MASK);
 	}
 	
 	*len = next_os - os;
@@ -170,6 +177,60 @@ rec_set_nth_field_null_bit(
 	rec_2_set_field_end_info(rec, i, info);
 }
 
+/***************************************************************
+Sets the value of the ith field extern storage bit. */
+
+void
+rec_set_nth_field_extern_bit(
+/*=========================*/
+	rec_t*	rec,	/* in: record */
+	ulint	i,	/* in: ith field */
+	ibool	val,	/* in: value to set */
+	mtr_t*	mtr)	/* in: mtr holding an X-latch to the page where
+			rec is, or NULL; in the NULL case we do not
+			write to log about the change */
+{
+	ulint	info;
+
+	ut_a(!rec_get_1byte_offs_flag(rec));
+	ut_a(i < rec_get_n_fields(rec));
+	
+	info = rec_2_get_field_end_info(rec, i);
+
+	if (val) {
+		info = info | REC_2BYTE_EXTERN_MASK;
+	} else {
+		info = info & ~REC_2BYTE_EXTERN_MASK;
+	}
+
+	if (mtr) {
+		mlog_write_ulint(rec - REC_N_EXTRA_BYTES - 2 * (i + 1), info,
+							MLOG_2BYTES, mtr);
+	} else {
+		rec_2_set_field_end_info(rec, i, info);
+	}
+}
+
+/***************************************************************
+Sets TRUE the extern storage bits of fields mentioned in an array. */
+
+void
+rec_set_field_extern_bits(
+/*======================*/
+	rec_t*	rec,		/* in: record */
+	ulint*	vec,		/* in: array of field numbers */
+	ulint	n_fields,	/* in: number of fields numbers */
+	mtr_t*	mtr)		/* in: mtr holding an X-latch to the page
+				where rec is, or NULL; in the NULL case we
+				do not write to log about the change */
+{
+	ulint	i;
+	
+	for (i = 0; i < n_fields; i++) {
+		rec_set_nth_field_extern_bit(rec, vec[i], TRUE, mtr);
+	}
+}
+
 /*************************************************************** 
 Sets a record field to SQL null. The physical size of the field is not
 changed. */