Fix leading serial number zeroes not being included in the database

certificate/certificate.go

@@ -8,6 +8,7 @@ import (
 	"crypto/sha1"
 	"crypto/sha256"
 	"crypto/x509"
+	"encoding/asn1"
 	"encoding/base64"
 	"fmt"
 	"log"
@@ -458,7 +459,21 @@ func CertToStored(cert *x509.Certificate, parentSignature, domain, ip string, TS
 	stored.IPs = make([]string, 0)
 
 	stored.Version = cert.Version
-	stored.Serial = fmt.Sprintf("%X", cert.SerialNumber)
+
+	// Compute the ASN1 representation of the serial number
+	m, err := asn1.Marshal(cert.SerialNumber)
+	if err != nil {
+		stored.Serial = ""
+	}
+	// Unmarshal the ASN1 representation into a RawValue so we can trim
+	// the ASN1 tag and length
+	var rawValue asn1.RawValue
+	_, err = asn1.Unmarshal(m, &rawValue)
+	if err != nil {
+		stored.Serial = ""
+	}
+	stored.Serial = fmt.Sprintf("%X", rawValue.Bytes)
+
 	stored.SignatureAlgorithm = SignatureAlgorithm[cert.SignatureAlgorithm]
 
 	stored.Key, err = getPublicKeyInfo(cert)

tools/fixserialnumber.go

@@ -0,0 +1,188 @@
+package main
+
+import (
+	"crypto/x509"
+	"encoding/asn1"
+	"encoding/base64"
+	"flag"
+	"fmt"
+	"log"
+	"math"
+	"os"
+
+	"github.com/mozilla/tls-observatory/database"
+)
+
+type job struct {
+	id int64
+	// Needs to be *string because apparently serial numbers can be NULL in the db
+	currentSerialNumber *string
+	cert                *x509.Certificate
+}
+
+type result struct {
+	id      int64
+	changed bool
+	err     error
+}
+
+func main() {
+	var workerCount int
+	var batchSize int64
+	var minID int64
+	var maxID int64
+	flag.IntVar(&workerCount, "workers", 4, "Number of workers to use")
+	flag.Int64Var(&batchSize, "batchSize", 1000, "Batch size")
+	flag.Int64Var(&minID, "minID", 0, "Minimum certificate ID to modify")
+	flag.Int64Var(&maxID, "maxID", math.MaxInt64, "Maximum certificate ID to modify")
+	flag.Parse()
+	jobs := make(chan job, batchSize)
+	results := make(chan result, batchSize)
+
+	db, err := database.RegisterConnection(
+		os.Getenv("TLSOBS_POSTGRESDB"),
+		os.Getenv("TLSOBS_POSTGRESUSER"),
+		os.Getenv("TLSOBS_POSTGRESPASS"),
+		os.Getenv("TLSOBS_POSTGRES"),
+		"disable",
+	)
+	if err != nil {
+		log.Fatalf("Error connecting to database: %s", err)
+	}
+	defer db.Close()
+
+	for w := 1; w <= workerCount; w++ {
+		go worker(w, jobs, results, db)
+	}
+	changedCount := 0
+	errorCount := 0
+	total := 0
+	go func() {
+		for {
+			log.Printf("Fetching %d certificates with id > %d", batchSize, minID)
+			nextBatch, err := fetchNextBatchWithRetries(5, db, minID, batchSize)
+			if err != nil {
+				log.Fatalf("Error fetching next batch: %s", err)
+			}
+			if len(nextBatch) == 0 || minID >= maxID {
+				close(jobs)
+				close(results)
+				log.Printf("Done. %d/%d errors. %d/%d changed.", errorCount, total, changedCount, total)
+				return
+			}
+			total += len(nextBatch)
+			for _, j := range nextBatch {
+				jobs <- j
+				minID = j.id
+			}
+		}
+	}()
+	for result := range results {
+		if result.err != nil {
+			errorCount++
+			log.Printf("Received error for cert id %d: %s", result.id, result.err)
+		}
+		if result.changed {
+			changedCount++
+		}
+	}
+}
+
+func fetchNextBatchWithRetries(retries int, db *database.DB, minID int64, batchSize int64) (jobs []job, err error) {
+	for i := 0; i < retries; i++ {
+		jobs, err = fetchNextBatch(db, minID, batchSize)
+		if err == nil {
+			break
+		}
+	}
+	return
+}
+
+func fetchNextBatch(db *database.DB, minID int64, batchSize int64) ([]job, error) {
+	rows, err := db.Query(`SELECT id, serial_number, raw_cert
+							   FROM certificates
+							   WHERE id > $1
+							   ORDER BY id
+							   LIMIT $2`,
+		minID,
+		batchSize,
+	)
+	if err != nil {
+		log.Fatalf("Error querying database: %s", err)
+	}
+	defer rows.Close()
+	var jobs []job
+	for rows.Next() {
+		var j job
+		var b64Crt string
+		if err = rows.Scan(&j.id, &j.currentSerialNumber, &b64Crt); err != nil {
+			return nil, fmt.Errorf("Error scanning row: %s", err)
+		}
+		cert, err := b64RawCertToX509Cert(b64Crt)
+		if err != nil {
+			log.Printf("Error converting database certificate to crypto/x509 certificate: %s", err)
+			continue
+		}
+		j.cert = cert
+		jobs = append(jobs, j)
+	}
+	return jobs, nil
+}
+
+func b64RawCertToX509Cert(b64Crt string) (*x509.Certificate, error) {
+	rawCert, err := base64.StdEncoding.DecodeString(b64Crt)
+	if err != nil {
+		return nil, fmt.Errorf("Error b64 decoding certificate: %s", err)
+	}
+	cert, err := x509.ParseCertificate(rawCert)
+	if err != nil {
+		return nil, fmt.Errorf("Error parsing x509 certificate: %s", err)
+	}
+	return cert, nil
+}
+
+func worker(id int, jobs <-chan job, results chan result, db *database.DB) {
+	for j := range jobs {
+		correctSerialNumberBytes, err := getSerialNumberASN1Value(j.cert)
+		if err != nil {
+			results <- result{id: j.id, err: err}
+			continue
+		}
+		correctSerialNumber := fmt.Sprintf("%X", correctSerialNumberBytes)
+		if correctSerialNumber == *j.currentSerialNumber {
+			// Serial number is already stored correctly in the database
+			results <- result{id: j.id, err: nil}
+			continue
+		}
+		err = updateSerialNumberInDB(db, j.id, correctSerialNumber)
+		if err != nil {
+			results <- result{
+				id:  j.id,
+				err: fmt.Errorf("Error updating serial number in database: %s", err),
+			}
+			continue
+		}
+		results <- result{id: j.id, err: nil, changed: true}
+	}
+}
+
+// getSerialNumberASN1Value returns the ASN1-encoded serial number for the given certificate, excluding the tag and length.
+func getSerialNumberASN1Value(cert *x509.Certificate) ([]byte, error) {
+	m, err := asn1.Marshal(cert.SerialNumber)
+	if err != nil {
+		return nil, fmt.Errorf("Error marshalling certificate's serial number: %s", err)
+	}
+	var rawValue asn1.RawValue
+	_, err = asn1.Unmarshal(m, &rawValue)
+	if err != nil {
+		return nil, fmt.Errorf("Error unmarshaling certificate's serial number from ASN1: %s", err)
+	}
+	return rawValue.Bytes, nil
+}
+
+func updateSerialNumberInDB(db *database.DB, id int64, correctSerialNumber string) error {
+	_, err := db.Exec(`UPDATE certificates
+					SET serial_number = $1
+					WHERE id = $2`, correctSerialNumber, id)
+	return err
+}