fix(models): filter out reserved tag keys in points parser (#16412)

* fix(models): filter out reserved tag keys in points parser

* chore(models): update changelog to reflect reserved tag key change

CHANGELOG.md

@@ -9,7 +9,9 @@
 1. [16522](https://github.com/influxdata/influxdb/pull/16522): Introduce resource logger to tasks, buckets and organizations
 
 ### Bug Fixes
+
 1. [16656](https://github.com/influxdata/influxdb/pull/16656): Check engine closed before collecting index metrics
+1. [16412](https://github.com/influxdata/influxdb/pull/16412): Reject writes which use any of the reserved tag keys
 
 ### UI Improvements
 

models/points.go

@@ -18,16 +18,31 @@ import (
 	"github.com/influxdata/influxdb/pkg/escape"
 )
 
-// Values used to store the field key and measurement name as special internal tags.
 const (
+	// Values used to store the field key and measurement name as special internal tags.
 	FieldKeyTagKey    = "\xff"
 	MeasurementTagKey = "\x00"
+
+	// reserved tag keys which when present cause the point to be discarded
+	// and an error returned
+	reservedFieldTagKey       = "_field"
+	reservedMeasurementTagKey = "_measurement"
+	reservedTimeTagKey        = "time"
 )
 
-// Predefined byte representations of special tag keys.
 var (
+	// Predefined byte representations of special tag keys.
 	FieldKeyTagKeyBytes    = []byte(FieldKeyTagKey)
 	MeasurementTagKeyBytes = []byte(MeasurementTagKey)
+
+	// set of reserved tag keys which cannot be present when a point is being parsed.
+	reservedTagKeys = [][]byte{
+		FieldKeyTagKeyBytes,
+		MeasurementTagKeyBytes,
+		[]byte(reservedFieldTagKey),
+		[]byte(reservedMeasurementTagKey),
+		[]byte(reservedTimeTagKey),
+	}
 )
 
 type escapeSet struct {
@@ -422,6 +437,18 @@ func scanKey(buf []byte, i int) (int, []byte, error) {
 		}
 	}
 
+	// Iterate over tags keys ensure that we do not encounter any
+	// of the reserved tag keys such as _measurement or _field.
+	for j := 0; j < commas; j++ {
+		_, key := scanTo(buf[indices[j]:indices[j+1]-1], 0, '=')
+
+		for _, reserved := range reservedTagKeys {
+			if bytes.Equal(key, reserved) {
+				return i, buf[start:i], fmt.Errorf("cannot use reserved tag key %q", key)
+			}
+		}
+	}
+
 	// Now we know where the key region is within buf, and the location of tags, we
 	// need to determine if duplicate tags exist and if the tags are sorted. This iterates
 	// over the list comparing each tag in the sequence with each other.

models/points_test.go

@@ -2,6 +2,7 @@ package models_test
 
 import (
 	"bytes"
+	"errors"
 	"fmt"
 	"io"
 	"io/ioutil"
@@ -106,24 +107,35 @@ func TestPoint_Tags(t *testing.T) {
 	examples := []struct {
 		Point string
 		Tags  models.Tags
+		Err   error
 	}{
-		{`cpu value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value"})},
-		{"cpu,tag0=v0 value=1", models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": "v0"})},
-		{"cpu,tag0=v0,tag1=v0 value=1", models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": "v0", "tag1": "v0"})},
-		{`cpu,tag0=v\ 0 value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": "v 0"})},
-		{`cpu,tag0=v\ 0\ 1,tag1=v2 value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": "v 0 1", "tag1": "v2"})},
-		{`cpu,tag0=\, value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": ","})},
-		{`cpu,ta\ g0=\, value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "ta g0": ","})},
-		{`cpu,tag0=\,1 value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": ",1"})},
-		{`cpu,tag0=1\"\",t=k value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": `1\"\"`, "t": "k"})},
+		{`cpu value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value"}), nil},
+		{"cpu,tag0=v0 value=1", models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": "v0"}), nil},
+		{"cpu,tag0=v0,tag1=v0 value=1", models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": "v0", "tag1": "v0"}), nil},
+		{`cpu,tag0=v\ 0 value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": "v 0"}), nil},
+		{`cpu,tag0=v\ 0\ 1,tag1=v2 value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": "v 0 1", "tag1": "v2"}), nil},
+		{`cpu,tag0=\, value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": ","}), nil},
+		{`cpu,ta\ g0=\, value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "ta g0": ","}), nil},
+		{`cpu,tag0=\,1 value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": ",1"}), nil},
+		{`cpu,tag0=1\"\",t=k value=1`, models.NewTags(map[string]string{models.MeasurementTagKey: "cpu", models.FieldKeyTagKey: "value", "tag0": `1\"\"`, "t": "k"}), nil},
+		{"cpu,_measurement=v0,tag0=v0 value=1", nil, errors.New(`unable to parse 'cpu,_measurement=v0,tag0=v0 value=1': cannot use reserved tag key "_measurement"`)},
+		// the following are all unsorted tag keys to ensure this works for both cases
+		{"cpu,tag0=v0,_measurement=v0 value=1", nil, errors.New(`unable to parse 'cpu,tag0=v0,_measurement=v0 value=1': cannot use reserved tag key "_measurement"`)},
+		{"cpu,tag0=v0,_field=v0 value=1", nil, errors.New(`unable to parse 'cpu,tag0=v0,_field=v0 value=1': cannot use reserved tag key "_field"`)},
+		{"cpu,tag0=v0,time=v0 value=1", nil, errors.New(`unable to parse 'cpu,tag0=v0,time=v0 value=1': cannot use reserved tag key "time"`)},
 	}
 
 	for _, example := range examples {
 		t.Run(example.Point, func(t *testing.T) {
 			pts, err := models.ParsePointsString(example.Point, "mm")
 			if err != nil {
-				t.Fatal(err)
-			} else if len(pts) != 1 {
+				if !reflect.DeepEqual(example.Err, err) {
+					t.Fatalf("expected %#v, found %#v", example.Err, err)
+				}
+				return
+			}
+
+			if len(pts) != 1 {
 				t.Fatalf("parsed %d points, expected 1", len(pts))
 			}