backup: the panel's archive is the command's, private key included

Owner, 2026-10-11: the private key is needed, and a backup from the
panel and one from the console have to be the same.

The commit that made the panel's backup work skipped a file under
certs/ the panel could not open - on the test server the private key,
which the certificate hook leaves to root and Postfix's group. That gave
an archive with a certificate and no key, and two kinds of archive.

- Params.Unprivileged is gone. Nothing is left out on account of who
  makes the archive; the queue is out for everyone (previous commit).
- When the panel cannot read a file under certs/, backup.Check says so
  before the download and the panel refuses with what to do on the host:
  chgrp 998 certs/privkey.pem && chmod 640. backup.UnreadableCert tells
  that case from any other failure.
- 998 is the container's selfpost group, which holds the panel and
  Postfix. Its id was whatever groupadd gave; the Dockerfile now fixes
  it at the value it has always had, because an operator addresses it
  by number from the host.
- Guide: method 1 is the same archive as method 2, with the permission
  the key needs and a note for whatever renews the certificate. Help and
  the box follow the design commit before this one.
- e2e: the restore scenario first makes the key unreadable and expects
  the refusal naming the file and the way out, then takes the backup and
  expects the key in it; nothing is put back by hand before the restart.

  go vet ./internal/backup/ ./internal/web/handlers/: ok
  go test ./internal/backup/ ./cmd/... ./internal/web/...: ok
  make e2e on the test server: ok

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

CHANGELOG.md

@@ -33,11 +33,13 @@ Format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/); version
   and the panel is not that user ("open /data/postfix/queue/active:
   permission denied" in the log). `selfpost-backup` run with `docker exec`
   was not affected - it runs as root. The queue is no longer part of a backup
-  (see Changed), and the panel's archive leaves out a file in `certs/` it may
+  (see Changed), so the two archives are the same again. Everything is read
-  not open (usually the private key, named in the log); the page, Help and
+  through before the download starts, so a failure is a message on the page
-  the guide say so, and the guide says to put the key back before starting a
+  and not a file cut short. One thing the panel may still be unable to read
-  restored instance. Everything is read through before the download starts,
+  is the private key in `certs/`: it then refuses and says what to do -
-  so a failure is a message on the page and not a file cut short.
+  `chgrp 998 certs/privkey.pem && chmod 640 certs/privkey.pem` on the host,
+  998 being the container's `selfpost` group, whose id is now fixed in the
+  image.
 
 ### Security
 

build/Dockerfile

@@ -62,7 +62,10 @@ RUN useradd --system --no-create-home --shell /usr/sbin/nologin panel
 # entrypoint.sh) — lets OpenDKIM read the panel-owned keys and lets Postfix read
 # the sasldb2/sender map, and lets the panel reach the supervisor control socket
 # to signal OpenDKIM/Postfix reloads without any process running as root.
-RUN groupadd --system selfpost \
+# The id is fixed, and it is the one the group has always had: an operator
+# gives the TLS private key on the host to this group by number, so that the
+# panel can put it into a full backup (guide § Taking a full backup).
+RUN groupadd --system --gid 998 selfpost \
  && usermod -aG selfpost panel \
  && usermod -aG selfpost opendkim \
  && usermod -aG selfpost postfix

docs/guide.md

@@ -560,17 +560,25 @@ instance keeps running, what goes into the archive, and what happens on restore.
 
 **1. Panel — while the container is running (no downtime).** *Server › Backup*
 → *Full backup*. *Encrypt with a password* is ticked; untick it for a plain
-archive. The archive of the CLI below, less one thing the panel may not read:
+archive. Same archive as the CLI below. The panel runs as its own user, not as
-it runs as its own user, not as root, and **a file in `certs/` it cannot open**
+root, so **the private key in `certs/` has to be readable by it**: give the
-is left out — usually the private key (the panel's log names it). Put the key
+file to the container's `selfpost` group, which has the fixed id 998 and holds
-back into `certs/` before starting a restored instance. If anything else
+both the panel and Postfix —
-cannot be read, the panel says so and downloads nothing.
+
+```sh
+chgrp 998 certs/privkey.pem && chmod 640 certs/privkey.pem
+```
+
+— and have whatever renews the certificate do the same (a certbot deploy hook,
+for one). A key readable by root alone is enough for Postfix, but then the
+panel refuses to make a backup and names the file: it does not hand over an
+archive without the key.
 
 **2. `selfpost-backup` CLI — while the container is running (no downtime).**
 The panel button and this command call the same code path: a consistent SQLite
 snapshot via `VACUUM INTO`, then a gzip tar of `data/` (minus diagnostics) plus
-deploy files. `docker exec` runs the command as root, so this archive has
+deploy files. `docker exec` runs the command as root, so it reads the private
-everything in `certs/`, the private key included.
+key whatever its permissions.
 
 ```sh
 docker exec <container> selfpost-backup > selfpost-backup.tar.gz
@@ -626,7 +634,7 @@ methods avoid it by stopping first.
 | SQLite consistency | `VACUUM INTO` snapshot | Safe when stopped | Safe when stopped |
 | `data/log/` | Excluded | Included (optional exclude) | Included (optional exclude) |
 | Postfix queue | No | Yes | Yes |
-| `certs/` private key | CLI: yes. Panel: only if the panel may read it | Yes | No — deploy files separate |
+| `certs/` private key | Yes (the panel needs it readable by group 998) | Yes | No — deploy files separate |
 | `data/manifest.json` | Written (version stamp) | Absent | Absent |
 | Restore version check | Yes — must match image tag | No | No |
 | Post-restore *Resync* | Yes, on first boot | No | No |

internal/backup/backup.go

@@ -80,12 +80,6 @@ type Params struct {
 	Version    string
 	DeployRoot string
 	OnWarn     func(string)
-	// Unprivileged says the archive is made by the panel, which runs as its
-	// own user and not as root. A file under certs/ it may not open - usually
-	// the private key, root and Postfix's group only - is then left out
-	// instead of failing the whole archive, and OnWarn names it. The
-	// selfpost-backup command, run as root, leaves this unset and takes it.
-	Unprivileged bool
 }
 
 // QueueDir is the Postfix queue, relative to the data directory. No backup
@@ -164,11 +158,11 @@ func Create(w io.Writer, p Params) error {
 		return err
 	}
 
-	if err := addTree(tw, p.DataDir, DataArchivePrefix, excludedFromArchive, p.skipData()); err != nil {
+	if err := addTree(tw, p.DataDir, DataArchivePrefix, excludedFromArchive, skipQueue); err != nil {
 		return err
 	}
 
-	if err := addDeployFiles(tw, p.DeployRoot, p.OnWarn, p.skipCert()); err != nil {
+	if err := addDeployFiles(tw, p.DeployRoot, p.OnWarn); err != nil {
 		return err
 	}
 
@@ -204,43 +198,20 @@ func validateDeployRoot(deployRoot string) error {
 // left out of the archive. A directory left out is not descended into.
 type skipFunc func(path, rel string, d fs.DirEntry) bool
 
-// skipData leaves the Postfix queue out of every archive.
+// skipQueue leaves the Postfix queue out of every archive.
-func (p Params) skipData() skipFunc {
+func skipQueue(_, rel string, _ fs.DirEntry) bool {
-	return func(_, rel string, _ fs.DirEntry) bool {
 	return filepath.ToSlash(rel) == QueueDir
 }
-}
-
-// skipCert leaves a file under certs/ that cannot be opened out of an
-// unprivileged archive, and says so through OnWarn.
-func (p Params) skipCert() skipFunc {
-	if !p.Unprivileged {
-		return nil
-	}
-	return func(path, rel string, d fs.DirEntry) bool {
-		if d.IsDir() {
-			return false
-		}
-		f, err := os.Open(path)
-		if err == nil {
-			f.Close()
-			return false
-		}
-		if !errors.Is(err, fs.ErrPermission) {
-			return false
-		}
-		if p.OnWarn != nil {
-			p.OnWarn(CertsDirName + "/" + filepath.ToSlash(rel) + " is not readable by the panel and is left out of the backup")
-		}
-		return true
-	}
-}
 
 // Check reads through everything Create would put into the archive, without
 // writing one, and returns the first thing that cannot be read. The panel
 // streams its archive to the browser, so once Create has begun a failure can
 // only cut the download short; this is run first, while a refusal can still be
 // shown as a page.
+//
+// The panel is not root, and nothing is left out on that account: its archive
+// is the one selfpost-backup makes, or it is not made. What it most likely
+// cannot read is the private key under certs/ - see UnreadableCert.
 func Check(p Params) error {
 	if p.DataDir == "" || p.DBPath == "" {
 		return fmt.Errorf("backup: DataDir and DBPath are required")
@@ -248,15 +219,28 @@ func Check(p Params) error {
 	if err := ValidateDeployRoot(p.DeployRoot); err != nil {
 		return err
 	}
-	quiet := p
+	if err := addTree(nil, p.DataDir, DataArchivePrefix, excludedFromArchive, skipQueue); err != nil {
-	quiet.OnWarn = nil
-	if err := addTree(nil, p.DataDir, DataArchivePrefix, excludedFromArchive, quiet.skipData()); err != nil {
 		return err
 	}
-	return addDeployFiles(nil, p.DeployRoot, nil, quiet.skipCert())
+	return addDeployFiles(nil, p.DeployRoot, nil)
 }
 
-func addDeployFiles(tw *tar.Writer, deployRoot string, onWarn func(string), skip skipFunc) error {
+// UnreadableCert reports whether err, from Check or Create, is a file under
+// the deploy root's certs/ that the caller is not allowed to read, and names
+// it relative to the deploy root ("certs/privkey.pem").
+func UnreadableCert(err error, deployRoot string) (name string, ok bool) {
+	var pathErr *fs.PathError
+	if !errors.As(err, &pathErr) || !errors.Is(err, fs.ErrPermission) {
+		return "", false
+	}
+	rel, relErr := filepath.Rel(filepath.Join(deployRoot, CertsDirName), pathErr.Path)
+	if relErr != nil || rel == ".." || strings.HasPrefix(filepath.ToSlash(rel), "../") {
+		return "", false
+	}
+	return CertsDirName + "/" + filepath.ToSlash(rel), true
+}
+
+func addDeployFiles(tw *tar.Writer, deployRoot string, onWarn func(string)) error {
 	for _, name := range []string{ComposeFileName, EnvFileName} {
 		src := filepath.Join(deployRoot, name)
 		info, err := os.Stat(src)
@@ -281,7 +265,7 @@ func addDeployFiles(tw *tar.Writer, deployRoot string, onWarn func(string), skip
 		}
 		return fmt.Errorf("backup: stat %s: %w", CertsDirName, err)
 	}
-	return addTree(tw, certsDir, CertsDirName+"/", nil, skip)
+	return addTree(tw, certsDir, CertsDirName+"/", nil, nil)
 }
 
 // addTree walks root and adds every regular file (and directory, to preserve

internal/backup/backup_test.go

@@ -5,7 +5,9 @@ import (
 	"bytes"
 	"compress/gzip"
 	"encoding/json"
+	"fmt"
 	"io"
+	"io/fs"
 	"os"
 	"path/filepath"
 	"strings"
@@ -236,11 +238,11 @@ func TestCheckRestoreWrongFormatRejected(t *testing.T) {
 	}
 }
 
-// The panel's archive: like any other it has no Postfix queue, the rest is
+// No archive holds the Postfix queue; the rest is there, the private key
-// there, and Check agrees that it can be made.
+// included, and Check agrees that it can be made.
-func TestCreateUnprivilegedLeavesTheQueueOut(t *testing.T) {
+func TestCreateLeavesTheQueueOut(t *testing.T) {
 	dataDir, dbPath, deployRoot := seedDataDir(t)
-	p := Params{DataDir: dataDir, DBPath: dbPath, Version: "1.2.3", DeployRoot: deployRoot, Unprivileged: true}
+	p := Params{DataDir: dataDir, DBPath: dbPath, Version: "1.2.3", DeployRoot: deployRoot}
 
 	if err := Check(p); err != nil {
 		t.Fatalf("Check: %v", err)
@@ -262,6 +264,7 @@ func TestCreateUnprivilegedLeavesTheQueueOut(t *testing.T) {
 		DataArchivePrefix + "postfix/sender_login_maps",
 		ComposeFileName, EnvFileName,
 		CertsDirName + "/fullchain.pem",
+		CertsDirName + "/privkey.pem",
 	} {
 		if _, ok := files[name]; !ok {
 			t.Errorf("archive missing %s", name)
@@ -269,6 +272,28 @@ func TestCreateUnprivilegedLeavesTheQueueOut(t *testing.T) {
 	}
 }
 
+// A file under certs/ that may not be read is told apart from any other
+// failure, so the panel can say what to do about it.
+func TestUnreadableCert(t *testing.T) {
+	root := filepath.Join("srv", "deploy")
+	denied := func(path string) error {
+		return fmt.Errorf("backup: open %s: %w", path, &fs.PathError{Op: "open", Path: path, Err: fs.ErrPermission})
+	}
+	if name, ok := UnreadableCert(denied(filepath.Join(root, "certs", "privkey.pem")), root); !ok || name != "certs/privkey.pem" {
+		t.Errorf("certs/privkey.pem: got %q, %v", name, ok)
+	}
+	if name, ok := UnreadableCert(denied(filepath.Join(root, ".env")), root); ok {
+		t.Errorf(".env was taken for a certificate file: %q", name)
+	}
+	if name, ok := UnreadableCert(denied(filepath.Join("data", "postfix", "x")), root); ok {
+		t.Errorf("a data file was taken for a certificate file: %q", name)
+	}
+	missing := &fs.PathError{Op: "open", Path: filepath.Join(root, "certs", "privkey.pem"), Err: fs.ErrNotExist}
+	if _, ok := UnreadableCert(missing, root); ok {
+		t.Error("a missing file was taken for one that may not be read")
+	}
+}
+
 // Check finds what Create would stumble over, before anything is written.
 func TestCheckReportsWhatCannotBeRead(t *testing.T) {
 	dataDir, dbPath, deployRoot := seedDataDir(t)

internal/web/handlers/handlers_backup.go

@@ -70,14 +70,11 @@ func (h *Handlers) HandleBackup(w http.ResponseWriter, r *http.Request) {
 		h.renderBackupPageWith(w, r, http.StatusBadRequest, "", pwErr)
 		return
 	}
-	// The panel is not root: its archive leaves out the Postfix queue and a
-	// certificate file it may not open (backup.Params.Unprivileged).
 	params := backup.Params{
 		DataDir:    h.cfg.DataDir,
 		DBPath:     h.cfg.DBPath,
 		Version:    h.cfg.Version,
 		DeployRoot: h.cfg.DeployRoot,
-		Unprivileged: true,
 		OnWarn: func(msg string) {
 			logf("panel: full backup: %s", msg)
 		},
@@ -90,11 +87,11 @@ func (h *Handlers) HandleBackup(w http.ResponseWriter, r *http.Request) {
 	// Everything is read through once before the first byte goes out. Past the
 	// headers a failure can only cut the download short - which is what the
 	// button did from 1.3.0 to 2.0.0, with a 200 and ten bytes, because the
-	// queue was not the panel's to read.
+	// queue was not the panel's to read. Nothing is skipped to make the archive
+	// possible: it is the one selfpost-backup makes, or it is refused.
 	if err := backup.Check(params); err != nil {
 		logf("panel: full backup: %v", err)
-		h.renderBackupPageWith(w, r, http.StatusInternalServerError, "",
+		h.renderBackupPageWith(w, r, http.StatusInternalServerError, "", unreadableBackupErr(err, h.cfg.DeployRoot))
-			"Could not read everything a backup holds. Nothing was downloaded; the reason is in the panel's log.")
 		return
 	}
 
@@ -333,6 +330,18 @@ func decryptErrorMessage(err error) string {
 	}
 }
 
+// unreadableBackupErr phrases what backup.Check found. The usual case is the
+// private key: the panel runs as its own user, and a key readable by root
+// alone is not its to read. The way out is on the host, so the message says
+// what to do there (guide § Taking a full backup).
+func unreadableBackupErr(err error, deployRoot string) string {
+	if name, ok := backup.UnreadableCert(err, deployRoot); ok {
+		return "The panel is not allowed to read " + name + ", and a backup without it would not bring the server back. " +
+			"On the host, give the file to group 998 and make it group-readable (chgrp 998, chmod 640), then try again. Nothing was downloaded."
+	}
+	return "Could not read everything a backup holds. Nothing was downloaded; the reason is in the panel's log."
+}
+
 // deployBackupErr phrases a pre-flight backup failure for the operator.
 func deployBackupErr(err error) string {
 	if strings.Contains(err.Error(), "DeployRoot") || strings.Contains(err.Error(), "deploy root") {

internal/web/view/help.go

@@ -193,7 +193,8 @@ var serverTopics = []HelpTopic{
 			". The Postfix queue is not in it. The backup is taken while the server is running, and Postfix keeps moving messages through the queue, "+
 				"so a copy could hold a message twice or miss it; mail still waiting for delivery is therefore not carried over, and a restored server "+
 				"starts with an empty queue. To keep the queue, stop the container and archive the project directory instead. The private key in ", Code("certs/"),
-			" is usually missing too, because the panel is not allowed to read it; ", Code("selfpost-backup"), " run in the container takes the key. It holds secrets, so leave Encrypt with a password ticked: the file is sealed as ", Code(".spbk"),
+			" has to be readable by the panel: give it to group 998 on the host (", Code("chgrp 998"), ", ", Code("chmod 640"),
+			"). Until then the panel refuses to make a backup and names the file, rather than hand over an archive without the key. It holds secrets, so leave Encrypt with a password ticked: the file is sealed as ", Code(".spbk"),
 			", the password is not stored, and without it nothing can be recovered. To restore, unpack it into an empty project directory (decrypt a ",
 			Code(".spbk"), " first with ", Code("selfpost-backup -decrypt"),
 			") and start the same SelfPost version that made it; the panel refuses data from another version. The reverse-proxy configuration and ",

internal/web/view/templates/backup.html

@@ -13,7 +13,7 @@
   <div class="column is-6">
 {{template "box_open" .Full}}
 {{template "body_open"}}
-<p class="mb-3"><code>data/</code> (database, DKIM keys, credentials), <code>docker-compose.yml</code>, <code>.env</code> and <code>certs/</code> in one archive. The Postfix queue is left out: it cannot be copied reliably while the server is running. So is a certificate file the panel may not read. Restore into an empty project directory and start the <strong>same SelfPost version</strong>. The reverse-proxy vhost is not included.</p>
+<p class="mb-3"><code>data/</code> (database, DKIM keys, credentials), <code>docker-compose.yml</code>, <code>.env</code> and <code>certs/</code> in one archive. The Postfix queue is left out: it cannot be copied reliably while the server is running. Restore into an empty project directory and start the <strong>same SelfPost version</strong>. The reverse-proxy vhost is not included.</p>
 <form class="sp-form" method="post" action="{{.FullAction}}">
   <div class="field"><label class="checkbox"><input type="checkbox" name="encrypt" value="1"{{if .Encrypt}} checked{{end}} data-encrypt-toggle> Encrypt with a password (<code>.spbk</code>)</label></div>
   <div data-encrypt-fields>

test/e2e/restore_test.go

@@ -17,12 +17,30 @@ import (
 // testFullBackupRestore takes a full backup from the panel, throws the
 // installation away and brings it back the way the guide tells an operator to
 // (§ Full backup and restore): unpack the archive into an empty project
-// directory, put the private key back, start the same image. What has to come back is what an operator
+// directory, start the same image. What has to come back is what an operator
 // cannot recreate - the administrator, the domain with its DKIM key, the
 // application with its password - and the proof is the same one the first
 // send gave: a message signed with the restored key verifies against the
 // record that was published before the backup.
 func testFullBackupRestore(t *testing.T, sc *scenario) {
+	// A private key the panel may not read - the state a server is in until
+	// its operator gives the key to group 998. The panel does not make an
+	// archive without it: it refuses, names the file and says what to do.
+	keyPath := filepath.Join(h.stageDir, "certs", "privkey.pem")
+	if err := os.Chmod(keyPath, 0o600); err != nil {
+		t.Fatal(err)
+	}
+	resp, body, err := sc.panel.postForm("/server/backup", url.Values{})
+	if err != nil {
+		t.Fatalf("full backup with an unreadable key: %v", err)
+	}
+	if resp.StatusCode != 500 || !strings.Contains(body, "not allowed to read certs/privkey.pem") || !strings.Contains(body, "chgrp 998") {
+		t.Fatalf("full backup with an unreadable key: status %d, want a refusal that names the file and the way out:\n%.800s", resp.StatusCode, body)
+	}
+	if err := os.Chmod(keyPath, 0o644); err != nil {
+		t.Fatal(err)
+	}
+
 	resp, archive, err := sc.panel.postForm("/server/backup", url.Values{})
 	if err != nil {
 		t.Fatalf("full backup: %v", err)
@@ -41,14 +59,6 @@ func testFullBackupRestore(t *testing.T, sc *scenario) {
 	// The installation goes: containers down, then everything the archive is
 	// supposed to bring back. The fake zone and the sink's directory are the
 	// outside world and stay.
-	// The private key is the operator's to keep: the panel may not read it (it
-	// is the host user's, mode 0600, here as on a real server) and its archive
-	// comes without it.
-	keyPath := filepath.Join(h.stageDir, "certs", "privkey.pem")
-	key, err := os.ReadFile(keyPath)
-	if err != nil {
-		t.Fatalf("read the stand's private key: %v", err)
-	}
 	h.reclaimData()
 	h.down()
 	for _, name := range []string{"data", "certs", "docker-compose.yml", ".env"} {
@@ -60,21 +70,16 @@ func testFullBackupRestore(t *testing.T, sc *scenario) {
 	if err != nil {
 		t.Fatalf("unpack the backup into the empty project directory: %v", err)
 	}
-	for _, want := range []string{"docker-compose.yml", ".env", "data/manifest.json", "data/selfpost.db", "certs/fullchain.pem"} {
+	for _, want := range []string{"docker-compose.yml", ".env", "data/manifest.json", "data/selfpost.db", "certs/fullchain.pem", "certs/privkey.pem"} {
 		if !names[want] {
 			t.Fatalf("the backup has no %s (it has %d entries)", want, len(names))
 		}
 	}
-	// The panel's archive is the one without the Postfix queue (guide § Taking
+	// No backup holds the Postfix queue (guide § Full backup and restore): the
-	// a full backup): the restored instance has to build an empty one.
+	// restored instance has to build an empty one.
 	for name := range names {
 		if name == "data/postfix/queue" || strings.HasPrefix(name, "data/postfix/queue/") {
-			t.Fatalf("the panel's backup holds %s", name)
+			t.Fatalf("the backup holds %s", name)
-		}
-	}
-	if !names["certs/privkey.pem"] {
-		if err := os.WriteFile(keyPath, key, 0o600); err != nil {
-			t.Fatalf("put the private key back: %v", err)
 		}
 	}
 

test/e2e/stage.go

@@ -113,5 +113,13 @@ func writeSelfSignedCert(certPath, keyPath string) error {
 		return err
 	}
 	keyPEM := pem.EncodeToMemory(&pem.Block{Type: "RSA PRIVATE KEY", Bytes: x509.MarshalPKCS1PrivateKey(key)})
-	return os.WriteFile(keyPath, keyPEM, 0o600)
+	// Readable by everyone, which a real key must never be. On a server the
+	// operator gives the key to the container's selfpost group (998) so the
+	// panel can put it into a full backup; the runner of this suite is not
+	// always root and cannot hand a file to a group it is not in. The key is
+	// thrown away with the stage directory.
+	if err := os.WriteFile(keyPath, keyPEM, 0o644); err != nil {
+		return err
+	}
+	return os.Chmod(keyPath, 0o644)
 }