MDEV-19522 InnoDB commit fails when FTS_DOC_ID value is greater than …

…4294967295 InnoDB commit fails when consecutive FTS_DOC_ID value is greater than 4294967295. Fix is that InnoDB should remove the delta FTS_DOC_ID value limitations and fts should encode 8 byte value, remove FTS_DOC_ID_MAX_STEP variable. Replaced the fts0vlc.ic file with fts0vlc.h fts_encode_int(): Should be able to encode 10 bytes value fts_get_encoded_len(): Should get the length of the value which has 10 bytes fts_decode_vlc(): Add debug assertion to verify the maximum length allowed is 10. mach_read_uint64_little_endian(): Reads 64 bit stored in little endian format Added a unit test case which check for minimum and maximum value to do the fts encoding
MariaDB · Oct 21, 2021 · 8ce8c26 · 8ce8c26
1 parent 6b4fad9
commit 8ce8c26
Show file tree

Hide file tree

Showing 18 changed files with 286 additions and 235 deletions.
diff --git a/mysql-test/suite/innodb_fts/r/basic.result b/mysql-test/suite/innodb_fts/r/basic.result
@@ -313,9 +313,7 @@ FTS_DOC_ID
 65536
 131071
 drop table t1;
-call mtr.add_suppression("\\[ERROR\\] InnoDB: Doc ID 20030101000000 is too big. Its difference with largest used Doc ID 0 cannot exceed or equal to 65535");
 CREATE TABLE t1 (FTS_DOC_ID BIGINT UNSIGNED AUTO_INCREMENT NOT NULL PRIMARY KEY,
 title VARCHAR(200), FULLTEXT(title)) ENGINE=InnoDB;
 INSERT INTO t1 VALUES (NULL, NULL), (20030101000000, 20030102000000);
-ERROR HY000: Invalid InnoDB FTS Doc ID
 DROP TABLE t1;
diff --git a/mysql-test/suite/innodb_fts/r/innodb_fts_misc_1.result b/mysql-test/suite/innodb_fts/r/innodb_fts_misc_1.result
@@ -972,3 +972,24 @@ SELECT * FROM information_schema.innodb_ft_deleted;
 DOC_ID
 DROP TABLE t1;
 SET GLOBAL innodb_ft_aux_table=DEFAULT;
+#
+# MDEV-19522 InnoDB commit fails when FTS_DOC_ID value
+#                is greater than 4294967295
+#
+CREATE TABLE t1(
+FTS_DOC_ID BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
+f1 TEXT, f2 TEXT, PRIMARY KEY (FTS_DOC_ID),
+FULLTEXT KEY (f1)) ENGINE=InnoDB;
+INSERT INTO t1 VALUES (1,'txt','bbb');
+UPDATE t1 SET FTS_DOC_ID = 4294967298;
+SELECT * FROM t1 WHERE match(f1) against("txt");
+FTS_DOC_ID	f1	f2
+4294967298	txt	bbb
+SET @@session.insert_id = 100000000000;
+INSERT INTO t1(f1, f2) VALUES ('aaa', 'bbb');
+CREATE FULLTEXT INDEX i ON t1 (f2);
+SELECT * FROM t1 WHERE match(f2) against("bbb");
+FTS_DOC_ID	f1	f2
+4294967298	txt	bbb
+100000000000	aaa	bbb
+DROP TABLE t1;
diff --git a/mysql-test/suite/innodb_fts/t/basic.test b/mysql-test/suite/innodb_fts/t/basic.test
@@ -277,9 +277,7 @@ insert into t1(f1, f2) values(3, "This is the third record");
 select FTS_DOC_ID from t1;
 drop table t1;
 
-call mtr.add_suppression("\\[ERROR\\] InnoDB: Doc ID 20030101000000 is too big. Its difference with largest used Doc ID 0 cannot exceed or equal to 65535");
 CREATE TABLE t1 (FTS_DOC_ID BIGINT UNSIGNED AUTO_INCREMENT NOT NULL PRIMARY KEY,
 		 title VARCHAR(200), FULLTEXT(title)) ENGINE=InnoDB;
---error 182
 INSERT INTO t1 VALUES (NULL, NULL), (20030101000000, 20030102000000);
 DROP TABLE t1;
diff --git a/mysql-test/suite/innodb_fts/t/innodb_fts_misc_1.test b/mysql-test/suite/innodb_fts/t/innodb_fts_misc_1.test
@@ -942,3 +942,21 @@ SET GLOBAL innodb_ft_aux_table='test/t1';
 SELECT * FROM information_schema.innodb_ft_deleted;
 DROP TABLE t1;
 SET GLOBAL innodb_ft_aux_table=DEFAULT;
+
+--echo #
+--echo # MDEV-19522 InnoDB commit fails when FTS_DOC_ID value
+--echo #                is greater than 4294967295
+--echo #
+CREATE TABLE t1(
+        FTS_DOC_ID BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
+        f1 TEXT, f2 TEXT, PRIMARY KEY (FTS_DOC_ID),
+        FULLTEXT KEY (f1)) ENGINE=InnoDB;
+INSERT INTO t1 VALUES (1,'txt','bbb');
+UPDATE t1 SET FTS_DOC_ID = 4294967298;
+SELECT * FROM t1 WHERE match(f1) against("txt");
+SET @@session.insert_id = 100000000000;
+INSERT INTO t1(f1, f2) VALUES ('aaa', 'bbb');
+CREATE FULLTEXT INDEX i ON t1 (f2);
+SELECT * FROM t1 WHERE match(f2) against("bbb");
+# Cleanup
+DROP TABLE t1;
diff --git a/storage/innobase/CMakeLists.txt b/storage/innobase/CMakeLists.txt
@@ -188,3 +188,7 @@ IF(MSVC)
 ENDIF()
 
 ADD_SUBDIRECTORY(${CMAKE_SOURCE_DIR}/extra/mariabackup ${CMAKE_BINARY_DIR}/extra/mariabackup)
+
+IF(WITH_UNIT_TESTS)
+  ADD_SUBDIRECTORY(unittest)
+ENDIF()
diff --git a/storage/innobase/fts/fts0fts.cc b/storage/innobase/fts/fts0fts.cc
@@ -32,7 +32,7 @@ Full Text Search interface
 #include "fts0priv.h"
 #include "fts0types.h"
 #include "fts0types.ic"
-#include "fts0vlc.ic"
+#include "fts0vlc.h"
 #include "fts0plugin.h"
 #include "dict0priv.h"
 #include "dict0stats.h"
@@ -1247,7 +1247,7 @@ fts_cache_node_add_positions(
 	ulint		enc_len;
 	ulint		last_pos;
 	byte*		ptr_start;
-	ulint		doc_id_delta;
+	doc_id_t	doc_id_delta;
 
 #ifdef UNIV_DEBUG
 	if (cache) {
@@ -1258,7 +1258,7 @@ fts_cache_node_add_positions(
 	ut_ad(doc_id >= node->last_doc_id);
 
 	/* Calculate the space required to store the ilist. */
-	doc_id_delta = (ulint)(doc_id - node->last_doc_id);
+	doc_id_delta = doc_id - node->last_doc_id;
 	enc_len = fts_get_encoded_len(doc_id_delta);
 
 	last_pos = 0;
@@ -1307,14 +1307,14 @@ fts_cache_node_add_positions(
 	ptr_start = ptr;
 
 	/* Encode the new fragment. */
-	ptr += fts_encode_int(doc_id_delta, ptr);
+	ptr = fts_encode_int(doc_id_delta, ptr);
 
 	last_pos = 0;
 	for (i = 0; i < ib_vector_size(positions); i++) {
 		ulint	pos = *(static_cast<ulint*>(
 			 ib_vector_get(positions, i)));
 
-		ptr += fts_encode_int(pos - last_pos, ptr);
+		ptr = fts_encode_int(pos - last_pos, ptr);
 		last_pos = pos;
 	}
 

diff --git a/storage/innobase/fts/fts0opt.cc b/storage/innobase/fts/fts0opt.cc
@@ -36,6 +36,7 @@ Completed 2011/7/10 Sunny and Jimmy Yang
 #include "ut0list.h"
 #include "zlib.h"
 #include "fts0opt.h"
+#include "fts0vlc.h"
 
 /** The FTS optimize thread's work queue. */
 ib_wqueue_t* fts_optimize_wq;
@@ -1116,7 +1117,7 @@ fts_optimize_encode_node(
 	ulint		pos_enc_len;
 	doc_id_t	doc_id_delta;
 	dberr_t		error = DB_SUCCESS;
-	byte*		src = enc->src_ilist_ptr;
+	const byte*	src = enc->src_ilist_ptr;
 
 	if (node->first_doc_id == 0) {
 		ut_a(node->last_doc_id == 0);
@@ -1173,7 +1174,7 @@ fts_optimize_encode_node(
 
 	/* Encode the doc id. Cast to ulint, the delta should be small and
 	therefore no loss of precision. */
-	dst += fts_encode_int((ulint) doc_id_delta, dst);
+	dst = fts_encode_int(doc_id_delta, dst);
 
 	/* Copy the encoded pos array. */
 	memcpy(dst, src, pos_enc_len);
@@ -1220,7 +1221,8 @@ fts_optimize_node(
 		doc_id_t	delta;
 		doc_id_t	del_doc_id = FTS_NULL_DOC_ID;
 
-		delta = fts_decode_vlc(&enc->src_ilist_ptr);
+		delta = fts_decode_vlc(
+			(const byte**)&enc->src_ilist_ptr);
 
 test_again:
 		/* Check whether the doc id is in the delete list, if
@@ -1248,7 +1250,7 @@ fts_optimize_node(
 
 			/* Skip the entries for this document. */
 			while (*enc->src_ilist_ptr) {
-				fts_decode_vlc(&enc->src_ilist_ptr);
+				fts_decode_vlc((const byte**)&enc->src_ilist_ptr);
 			}
 
 			/* Skip the end of word position marker. */

diff --git a/storage/innobase/fts/fts0que.cc b/storage/innobase/fts/fts0que.cc
@@ -34,6 +34,7 @@ Completed 2011/7/10 Sunny and Jimmy Yang
 #include "fts0pars.h"
 #include "fts0types.h"
 #include "fts0plugin.h"
+#include "fts0vlc.h"
 
 #include <iomanip>
 #include <vector>
@@ -3224,7 +3225,7 @@ fts_query_filter_doc_ids(
 	ulint			len,		/*!< in: doc id ilist size */
 	ibool			calc_doc_count)	/*!< in: whether to remember doc count */
 {
-	byte*		ptr = static_cast<byte*>(data);
+	const byte*	ptr = static_cast<byte*>(data);
 	doc_id_t	doc_id = 0;
 	ulint		decoded = 0;
 	ib_rbt_t*	doc_freqs = word_freq->doc_freqs;
@@ -3234,8 +3235,8 @@ fts_query_filter_doc_ids(
 		ulint		freq = 0;
 		fts_doc_freq_t*	doc_freq;
 		fts_match_t*	match = NULL;
-		ulint		last_pos = 0;
-		ulint		pos = fts_decode_vlc(&ptr);
+		doc_id_t	last_pos = 0;
+		doc_id_t	pos = fts_decode_vlc(&ptr);
 
 		/* Some sanity checks. */
 		if (doc_id == 0) {

diff --git a/storage/innobase/handler/ha_innodb.cc b/storage/innobase/handler/ha_innodb.cc
@@ -8543,8 +8543,7 @@ calc_row_difference(
 		    && prebuilt->table->fts
 		    && innobase_strcasecmp(
 			field->field_name, FTS_DOC_ID_COL_NAME) == 0) {
-			doc_id = (doc_id_t) mach_read_from_n_little_endian(
-				n_ptr, 8);
+			doc_id = mach_read_uint64_little_endian(n_ptr);
 			if (doc_id == 0) {
 				return(DB_FTS_INVALID_DOCID);
 			}
@@ -8787,16 +8786,6 @@ calc_row_difference(
 					<< innodb_table->name;
 
 				return(DB_FTS_INVALID_DOCID);
-			} else if ((doc_id
-				    - prebuilt->table->fts->cache->next_doc_id)
-				   >= FTS_DOC_ID_MAX_STEP) {
-
-				ib::warn() << "Doc ID " << doc_id << " is too"
-					" big. Its difference with largest"
-					" Doc ID used " << prebuilt->table->fts
-					->cache->next_doc_id - 1
-					<< " cannot exceed or equal to "
-					<< FTS_DOC_ID_MAX_STEP;
 			}
 
 

diff --git a/storage/innobase/handler/i_s.cc b/storage/innobase/handler/i_s.cc
@@ -58,6 +58,7 @@ Modified Dec 29, 2014 Jan Lindström (Added sys_semaphore_waits)
 #include "fil0fil.h"
 #include "fil0crypt.h"
 #include "dict0crea.h"
+#include "fts0vlc.h"
 
 /** The latest successfully looked up innodb_fts_aux_table */
 UNIV_INTERN table_id_t innodb_ft_aux_table_id;
@@ -2775,7 +2776,7 @@ i_s_fts_index_cache_fill_one_index(
 		/* Decrypt the ilist, and display Dod ID and word position */
 		for (ulint i = 0; i < ib_vector_size(word->nodes); i++) {
 			fts_node_t*	node;
-			byte*		ptr;
+			const byte*	ptr;
 			ulint		decoded = 0;
 			doc_id_t	doc_id = 0;
 
@@ -2785,13 +2786,11 @@ i_s_fts_index_cache_fill_one_index(
 			ptr = node->ilist;
 
 			while (decoded < node->ilist_size) {
-				ulint	pos = fts_decode_vlc(&ptr);
 
-				doc_id += pos;
+				doc_id += fts_decode_vlc(&ptr);
 
 				/* Get position info */
 				while (*ptr) {
-					pos = fts_decode_vlc(&ptr);
 
 					OK(field_store_string(
 						   fields[I_S_FTS_WORD],
@@ -2812,7 +2811,7 @@ i_s_fts_index_cache_fill_one_index(
 						   doc_id, true));
 
 					OK(fields[I_S_FTS_ILIST_DOC_POS]->store(
-						   pos, true));
+						   fts_decode_vlc(&ptr), true));
 
 					OK(schema_table_store_record(
 						   thd, table));
@@ -3146,7 +3145,7 @@ i_s_fts_index_table_fill_one_fetch(
 		/* Decrypt the ilist, and display Dod ID and word position */
 		for (ulint i = 0; i < ib_vector_size(word->nodes); i++) {
 			fts_node_t*	node;
-			byte*		ptr;
+			const byte*	ptr;
 			ulint		decoded = 0;
 			doc_id_t	doc_id = 0;
 
@@ -3156,13 +3155,10 @@ i_s_fts_index_table_fill_one_fetch(
 			ptr = node->ilist;
 
 			while (decoded < node->ilist_size) {
-				ulint	pos = fts_decode_vlc(&ptr);
-
-				doc_id += pos;
+				doc_id += fts_decode_vlc(&ptr);
 
 				/* Get position info */
 				while (*ptr) {
-					pos = fts_decode_vlc(&ptr);
 
 					OK(field_store_string(
 						   fields[I_S_FTS_WORD],
@@ -3181,7 +3177,7 @@ i_s_fts_index_table_fill_one_fetch(
 						longlong(doc_id), true));
 
 					OK(fields[I_S_FTS_ILIST_DOC_POS]->store(
-						   pos, true));
+						   fts_decode_vlc(&ptr), true));
 
 					OK(schema_table_store_record(
 						   thd, table));

diff --git a/storage/innobase/include/fts0fts.h b/storage/innobase/include/fts0fts.h
@@ -96,10 +96,6 @@ those defined in mysql file ft_global.h */
 /** Threshold where our optimize thread automatically kicks in */
 #define FTS_OPTIMIZE_THRESHOLD		10000000
 
-/** Threshold to avoid exhausting of doc ids. Consecutive doc id difference
-should not exceed FTS_DOC_ID_MAX_STEP */
-#define FTS_DOC_ID_MAX_STEP		65535
-
 /** Maximum possible Fulltext word length in bytes (assuming mbmaxlen=4) */
 #define FTS_MAX_WORD_LEN		(HA_FT_MAXCHARLEN * 4)
 

diff --git a/storage/innobase/include/fts0types.h b/storage/innobase/include/fts0types.h
@@ -314,16 +314,6 @@ int fts_doc_id_cmp(
 	const void*	p1,			/*!< in: id1 */
 	const void*	p2);			/*!< in: id2 */
 
-/******************************************************************//**
-Decode and return the integer that was encoded using our VLC scheme.*/
-UNIV_INLINE
-ulint
-fts_decode_vlc(
-/*===========*/
-			/*!< out: value decoded */
-	byte**	ptr);	/*!< in: ptr to decode from, this ptr is
-			incremented by the number of bytes decoded */
-
 /******************************************************************//**
 Duplicate a string. */
 UNIV_INLINE
@@ -338,28 +328,6 @@ fts_string_dup(
 	const fts_string_t*	src,		/*!< in: src string */
 	mem_heap_t*		heap);		/*!< in: heap to use */
 
-/******************************************************************//**
-Return length of val if it were encoded using our VLC scheme. */
-UNIV_INLINE
-ulint
-fts_get_encoded_len(
-/*================*/
-						/*!< out: length of value
-						 encoded, in bytes */
-	ulint		val);			/*!< in: value to encode */
-
-/******************************************************************//**
-Encode an integer using our VLC scheme and return the length in bytes. */
-UNIV_INLINE
-ulint
-fts_encode_int(
-/*===========*/
-						/*!< out: length of value
-						encoded, in bytes */
-	ulint		val,			/*!< in: value to encode */
-	byte*		buf);			/*!< in: buffer, must have
-						enough space */
-
 /******************************************************************//**
 Get the selected FTS aux INDEX suffix. */
 UNIV_INLINE
@@ -381,6 +349,5 @@ fts_select_index(
 	ulint			len);
 
 #include "fts0types.ic"
-#include "fts0vlc.ic"
 
 #endif /* INNOBASE_FTS0TYPES_H */