1 files changed, 163 insertions, 0 deletions
diff --git a/cmd_snapshot.go b/cmd_snapshot.go
new file mode 100644
index 0000000..5a9c50e
--- /dev/null
+++ b/cmd_snapshot.go
@@ -0,0 +1,163 @@
+package main
+
+import (
+	"bytes"
+	"context"
+	"crypto/sha256"
+	"crypto/x509"
+	"encoding/json"
+	"errors"
+	"fmt"
+	logger "log"
+	"net/http"
+	"os"
+	"time"
+
+	"git.cs.kau.se/rasmoste/ct-sans/internal/merkle"
+	ct "github.com/google/certificate-transparency-go"
+	"github.com/google/certificate-transparency-go/client"
+	"github.com/google/certificate-transparency-go/jsonclient"
+	"gitlab.torproject.org/rgdd/ct/pkg/metadata"
+)
+
+func snapshot(opts options) error {
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	if err := os.MkdirAll(opts.Directory, os.ModePerm); err != nil {
+		return err
+	}
+
+	logger.Printf("INFO: updating metadata file\n")
+	source := metadata.NewHTTPSource(metadata.HTTPSourceOptions{Name: "google"})
+	msg, sig, md, err := source.Load(ctx)
+	if err != nil {
+		return err
+	}
+	if err := os.WriteFile(fmt.Sprintf("%s/%s", opts.Directory, opts.metadataFile), msg, 0644); err != nil {
+		return err
+	}
+	if err := os.WriteFile(fmt.Sprintf("%s/%s", opts.Directory, opts.metadataSignatureFile), sig, 0644); err != nil {
+		return err
+	}
+	timestamp := []byte(fmt.Sprintf("%d", time.Now().Unix()))
+	if err := os.WriteFile(fmt.Sprintf("%s/%s", opts.Directory, opts.metadataTimestampFile), timestamp, 0644); err != nil {
+		return err
+	}
+
+	logger.Printf("INFO: updating signed tree heads\n")
+	for _, log := range logs(md) {
+		id, _ := log.Key.ID()
+		der, _ := x509.MarshalPKIXPublicKey(log.Key)
+		dir := fmt.Sprintf("%s/%x", opts.logDirectory, id)
+		sthFile := fmt.Sprintf("%s/%s", dir, opts.sthFile)
+		if err := os.MkdirAll(dir, os.ModePerm); err != nil {
+			return fmt.Errorf("%s: %v", log.Description, err)
+		}
+
+		// Fetch next STH
+		cli, err := client.New(string(log.URL), &http.Client{}, jsonclient.Options{PublicKeyDER: der, UserAgent: opts.HTTPAgent})
+		if err != nil {
+			return fmt.Errorf("%s: %v", *log.Description, err)
+		}
+		nextSTH, err := cli.GetSTH(ctx)
+		if err != nil {
+			return fmt.Errorf("%s: %v", *log.Description, err)
+		}
+		nextSTHBytes, err := json.Marshal(nextSTH)
+		if err != nil {
+			return fmt.Errorf("%s: %v", *log.Description, err)
+		}
+		//
+		// It's a bit ugly that ct.SignedTreeHead contains fields that
+		// are not populated.  Doesn't cause any prolems here, however.
+		//
+
+		// Bootstrap log if we don't have any STH yet
+		if _, err := os.Stat(sthFile); err != nil {
+			if !errors.Is(err, os.ErrNotExist) {
+				return fmt.Errorf("%s: %v", *log.Description, err)
+			}
+			if err := os.WriteFile(sthFile, nextSTHBytes, 0644); err != nil {
+				return fmt.Errorf("%s: %v", *log.Description, err)
+			}
+
+			logger.Printf("INFO: bootstrapped %s at tree size %d\n", *log.Description, nextSTH.TreeSize)
+			continue
+		}
+
+		// Otherwise: update an existing STH
+		currSTHBytes, err := os.ReadFile(sthFile)
+		if err != nil {
+			return fmt.Errorf("%s: %v", *log.Description, err)
+		}
+		var currSTH ct.SignedTreeHead
+		if err := json.Unmarshal(currSTHBytes, &currSTH); err != nil {
+			return fmt.Errorf("%s: %v", *log.Description, err)
+		}
+		if nextSTH.TreeSize < currSTH.TreeSize {
+			return fmt.Errorf("%s: next tree size is smaller: %s", *log.Description, nextSTHBytes)
+		}
+		if nextSTH.TreeSize == currSTH.TreeSize {
+			if !bytes.Equal(nextSTH.SHA256RootHash[:], currSTH.SHA256RootHash[:]) {
+				return fmt.Errorf("%s: split-view: %s", *log.Description, nextSTHBytes)
+			}
+
+			logger.Printf("INFO: %s is already up-to-date at size %d\n", *log.Description, nextSTH.TreeSize)
+			continue
+		}
+		hashes, err := cli.GetSTHConsistency(ctx, currSTH.TreeSize, nextSTH.TreeSize)
+		if err != nil {
+			return fmt.Errorf("%s: %v", *log.Description, err)
+		}
+		if err := merkle.VerifyConsistency(currSTH.TreeSize,
+			nextSTH.TreeSize,
+			[sha256.Size]byte(currSTH.SHA256RootHash),
+			[sha256.Size]byte(nextSTH.SHA256RootHash),
+			proof(hashes)); err != nil {
+			return fmt.Errorf("%s: inconsistent tree: %v", *log.Description, err)
+		}
+		if err := os.WriteFile(sthFile, nextSTHBytes, 0644); err != nil {
+			return fmt.Errorf("%s: %v", *log.Description, err)
+		}
+		logger.Printf("INFO: updated %s to tree size %d\n", *log.Description, nextSTH.TreeSize)
+	}
+	return nil
+}
+
+// logs select logs that count towards CT-compliance checks.  Logs that don't
+// have a description are skipped after printing a warning.
+func logs(md metadata.Metadata) (logs []metadata.Log) {
+	for _, operators := range md.Operators {
+		for _, log := range operators.Logs {
+			if log.Description == nil {
+				fmt.Fprintf(os.Stderr, "WARNING: skipping log without description")
+				continue
+			}
+			if log.State == nil {
+				continue // skip logs with unknown states
+			}
+			if log.State.Name == metadata.LogStatePending {
+				continue // pending logs do not count towards CT-compliance
+			}
+			if log.State.Name == metadata.LogStateRetired {
+				continue // retired logs are not necessarily reachable
+			}
+			if log.State.Name == metadata.LogStateRejected {
+				continue // rejected logs do not count towards CT-compliance
+			}
+
+			logs = append(logs, log)
+		}
+	}
+	return
+}
+
+// proof formats hashes so that they can be passed to the merkle package
+func proof(hashes [][]byte) (p [][sha256.Size]byte) {
+	for _, hash := range hashes {
+		var h [sha256.Size]byte
+		copy(h[:], hash)
+		p = append(p, h)
+	}
+	return
+}