*: fix a bug that update statement uses point get and update plan with different tblInfo | tidb-test=e1d0c1e615f749e7139f5be95bc4a2b8cedb7380 (#54183) (#54259)

[ti-chi-bot]

This is an automated cherry-pick of #54259

This is an automated cherry-pick of #54183

What problem does this PR solve?

Issue Number: close #53634

Problem Summary:

Case

Init SQLs：

CREATE TABLE stock ( a int NOT NULL, b char(30) NOT NULL,  c int, d char(64), PRIMARY KEY(a,b)) ;
insert into stock values(1, 'a', 11, 'x'), (2, 'b', 22, 'y');
alter table stock add column cct_1 int default 10;
alter table stock modify cct_1 json;
alter table stock add column adc_1 smallint;

• Step1. using conn1
  prepare statements:

    begin;
    SELECT a, c, d from stock where (a, b) IN ((?, ?),(?, ?)) FOR UPDATE;
    UPDATE stock SET c = ? WHERE a= ? AND b = 'a';
    UPDATE stock SET c = ?, d = 'z' WHERE a= ? AND b = 'b';
    commit;

run a statement:
begin;

• Step2. using conn2
  do DDL:

    alter table stock drop column cct_1;

• Step3. using conn1
  When the DDL is in Write-Only state
  exec statements:

    SELECT a, c, d from stock where (a, b) IN ((?, ?),(?, ?)) FOR UPDATE;
    UPDATE stock SET c = ? WHERE a= ? AND b = 'a';
    UPDATE stock SET c = ?, d = 'z' WHERE a= ? AND b = 'b';
    commit;

• Step4. using conn1
  Check the result
  select * from stock;

• Statement execution order table

conn1	conn2
step1. prepare stmts
step1. begin;
	step2. alter table stock drop column cct_1; (state public -> write-only)
step3. exec stmts	write-only
	step2. alter table stock drop column cct_1; (finish)
step4. select * from stock;

Conclusion

The update statement in step3 uses point get and update plan, but the tblInfo used by the two plans is inconsistent, resulting in incorrect data in real storage.

• After executing step3. select statement, the stock is locked in GetRelatedTableForMDL, so stmt.tbls[i].Meta().Revision != newTbl.Meta().Revision is false. It means schemaNotMatch is false. So we needn't to Preprocess.
  

  tidb/pkg/planner/core/plan_cache.go

  Lines 114 to 119 in d5b89f8

  	newTbl, err := tryLockMDLAndUpdateSchemaIfNecessary(sctx.GetPlanCtx(), stmt.dbName[i], stmt.tbls[i], is)
  	if err != nil {
  	schemaNotMatch = true
  	continue
  	}
  	if stmt.tbls[i].Meta().Revision != newTbl.Meta().Revision {

• Step3. update statement using tblName.TableInfo(get it when preparing statements, cct_1 is public) in newPointGetPlan.
  

  tidb/pkg/planner/core/point_get_plan.go

  Line 1317 in d5b89f8

  tbl := tblName.TableInfo

• Step3. update statement using t gets from is.TableByID(tbl.ID)( cct_1 is write-only)
  

  tidb/pkg/planner/core/point_get_plan.go

  Lines 1945 to 1950 in d5b89f8

  	updatePlan.names = pointPlan.OutputNames()
  	is := ctx.GetInfoSchema().(infoschema.InfoSchema)
  	t, _ := is.TableByID(tbl.ID)
  	updatePlan.tblID2Table = map[int64]table.Table{
  	tbl.ID: t,
  	}

What changed and how does it work?

Add the Revision field comparison of tbl(get from txn infoschema) and newTbl to confirm whether reprocess is required.

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)
• No need to test

  • I checked and no code files have been changed.

Side effects

• Performance regression: Consumes more CPU
• Performance regression: Consumes more Memory
• Breaking backward compatibility

Documentation

• Affects user behaviors
• Contains syntax changes
• Contains variable changes
• Contains experimental features
• Changes MySQL compatibility

Release note

Please refer to Release Notes Language Style Guide to write a quality release note.

Fix an issue where improper use of metadata locks in some scenarios could cause abnormal data to be written when using the plan cache.

[Lloyd-Pottiger]

/cc @zimulala @qw4990 @tangenta

[Lloyd-Pottiger]

/retest

[Lloyd-Pottiger]

/retest

[tangenta]

/retest

[Lloyd-Pottiger]

/retest

[Lloyd-Pottiger]

/retest

[Lloyd-Pottiger]

/retest

[Lloyd-Pottiger]

/retest

[jebter]

/retest

[jebter]

/retest

[jebter]

/test mysql-test

[tiprow]

@jebter: No presubmit jobs available for pingcap/[email protected]

In response to this:

/test mysql-test

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

[Lloyd-Pottiger]

/test all

[tiprow]

@Lloyd-Pottiger: No presubmit jobs available for pingcap/[email protected]

In response to this:

/test all

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

[jebter]

/retest

[codecov]

Codecov Report

Attention: Patch coverage is 53.12500% with 30 lines in your changes missing coverage. Please review.

Please upload report for BASE (release-6.5-20241009-v6.5.10@7972ff1). Learn more about missing BASE report.

Additional details and impacted files

@@                        Coverage Diff                        @@
##             release-6.5-20241009-v6.5.10     #57078   +/-   ##
=================================================================
  Coverage                                ?   73.6185%           
=================================================================
  Files                                   ?       1096           
  Lines                                   ?     352338           
  Branches                                ?          0           
=================================================================
  Hits                                    ?     259386           
  Misses                                  ?      76269           
  Partials                                ?      16683           

[ti-chi-bot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: qw4990, tangenta

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [qw4990,tangenta]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[ti-chi-bot]

[LGTM Timeline notifier]

Timeline:

• 2024-11-02 01:26:56.511939201 +0000 UTC m=+658729.351094747: ☑️ agreed by tangenta.
• 2024-11-03 13:49:33.238536035 +0000 UTC m=+789686.077691577: ☑️ agreed by qw4990.