Pass over a config path that runs through a file

Of the places pixa looks for its config file on its own, one whose path
runs through a file, such as under a HOME of /dev/null, aborted startup
with "not a directory", though no file can be there. It is now passed
over like a file that does not exist, as README.md says.

Model: opus-5-5

README.md

@@ -34,6 +34,65 @@ else has a built-in default. A config file mounted at `/etc/pixa/config.yml`
 is optional: it is read when present, and an environment variable wins over
 the same setting in it.
 
+## Deployment
+
+pixa listens on plain HTTP and runs behind a reverse proxy that terminates TLS.
+[`configs/Caddyfile`](configs/Caddyfile) is an example for Caddy, chosen because
+it is the smallest correct one: Caddy gets the TLS certificate itself and does
+everything in this list without further settings. The reverse proxy must:
+
+- terminate TLS, as the login and generator pages work only over HTTPS (see
+  Routes);
+- pass the `Host`, `Origin` and `Referer` headers on unchanged, as pixa refuses
+  a form from those pages unless `Origin` or `Referer` names the host in `Host`,
+  and builds encrypted URLs from `Host`;
+- set `X-Forwarded-For` to the client's address, with `trusted_proxies` set to
+  the address pixa sees the proxy's requests come from, so the login limit
+  counts each client by its own address (see `trusted_proxies` under
+  Configuration);
+- wait for pixa's answer for at least `downstream_timeout` (default `60s`), the
+  longest pixa takes to fetch, convert and send an image.
+
+It may also refuse `/metrics`, as the example does, so that only a scraper that
+reaches pixa directly can read it; pixa itself asks for the metrics username and
+password there.
+
+pixa does the rest itself: it checks signatures and encrypted URLs, applies the
+allowlist, refuses upstream hosts with private or local addresses, limits login
+attempts, upstream response size and image dimensions, and sends the security
+headers, `Strict-Transport-Security` included, with every response.
+
+The state directory (`state_dir`, `/var/lib/pixa` in the container) holds the
+database and the disk cache:
+
+- It needs a persistent volume: without one, every restart starts with an empty
+  cache. In the container, the startup script gives the directory to the user
+  pixa runs as (uid 65532) and sets its mode to `750`; outside it, that user
+  must be able to write the directory.
+- `cache_max_bytes` limits the source and transformed images together. The
+  database, the metadata files, the `.meta` file beside each transformed image
+  and files still being written come on top, and eviction runs in the
+  background, so the cache can pass the limit for a while: leave room on the
+  volume beyond it.
+- Set `cache_max_bytes` for a lasting deployment. Its default is 75% of the
+  space free when pixa starts, which the cache's own files reduce, so a fuller
+  cache gives a smaller limit after a restart.
+
+A load balancer's health check can request `/.well-known/healthcheck.json`,
+which answers 200 whenever pixa is running, in maintenance mode too (see
+`maintenance_mode`).
+
+On SIGTERM or SIGINT pixa stops accepting connections, gives the requests in
+progress and the images being processed 5 seconds to finish, and exits: with 0,
+or with 1 when images were still being processed after those 5 seconds or
+another part of pixa failed to stop. A request not finished by then is cut off.
+`docker stop` waits 10 seconds before it kills the container.
+
+Outside Docker, pixa needs libvips (the image has 8.15) and libheif to run, as
+it uses libvips through CGO; building it also needs their development files,
+`pkg-config` and a C compiler. `script/bootstrap` installs all of these with
+nix, apt, brew or apk.
+
 ## Running under upaas
 
 What the [upaas](https://git.eeqj.de/sneak/upaas) app for pixa needs:
@@ -354,10 +413,10 @@ pixa finds: `/etc/pixa/config.yml`, `/etc/pixa/config.yaml`,
 `~/.config/pixa/config.yml`, `~/.config/pixa/config.yaml`, then `config.yml`
 and `config.yaml` in the working directory. A named file that does not exist,
 cannot be read or does not parse aborts startup. Of the files pixa looks for on
-its own, one it finds but cannot read or parse aborts startup; one it cannot
+its own, only one that does not exist is passed over, without a message. One
-find, for any reason, is passed over without a message, even when the file is
+that pixa cannot read or parse aborts startup, naming the file. So does one in a
-there in a directory pixa may not enter. With no file, pixa uses the environment
+directory pixa may not enter, whether or not it is there, since pixa cannot
-and the defaults.
+tell. With no file, pixa uses the environment and the defaults.
 
 | Variable                             | Config key                      | Meaning                                                                      |
 | ------------------------------------ | ------------------------------- | ---------------------------------------------------------------------------- |

TODO.md

@@ -29,6 +29,20 @@ P2: security: referer blacklist
 
 # Completed Steps
 
+- 2026-10-04 a config file pixa cannot read aborts startup (closes #176): of the
+  places pixa looks for its config file on its own, only one where the file does
+  not exist is passed over; any other error, such as a directory on the path
+  that pixa may not enter, aborts startup naming the file, as a file that does
+  not parse already did.
+- 2026-10-04 deployment guide and example Caddy config (closes #89):
+  "Deployment" in `README.md` says what the reverse proxy in front of pixa must
+  do (terminate TLS; pass `Host`, `Origin` and `Referer` on unchanged; set
+  `X-Forwarded-For`, with `trusted_proxies` to match; wait at least
+  `downstream_timeout`; optionally refuse `/metrics`) and what pixa does itself,
+  that the state directory needs a persistent volume and what `cache_max_bytes`
+  counts, the health check for a load balancer, what a stop does and its exit
+  codes, and what running outside Docker needs; `configs/Caddyfile` is the
+  example, checked with `caddy validate`.
 - 2026-10-04 the metrics basic auth, CORS preflight, request logging and
   metrics recording have tests (closes #79): `MetricsAuth` on its own answers
   401 with a challenge without credentials or with a wrong username or password
@@ -485,6 +499,3 @@ P2: security: referer blacklist
     - Prometheus performance metrics
     - integration tests for the image proxy flow
     - load tests to verify the 1k to 5k req/s target
-- P2: documentation
-    - deployment guide
-    - example nginx or caddy reverse proxy config

configs/Caddyfile

@@ -0,0 +1,17 @@
+# Example Caddy config for running pixa behind Caddy; see "Deployment" in
+# README.md. Replace images.example.com with pixa's public host name, and
+# 127.0.0.1:8080 with the address Caddy reaches pixa on.
+#
+# Caddy gets and renews the TLS certificate for the host name, passes the
+# Host, Origin and Referer headers on unchanged, sets X-Forwarded-For to the
+# client's address, and waits for pixa's answer with no time limit of its
+# own, so pixa's downstream_timeout is what ends a slow request.
+
+images.example.com
+
+# pixa asks for metrics.username and metrics.password on /metrics. This
+# line also keeps it off the public address, for a scraper that reaches
+# pixa directly; remove it to read /metrics through Caddy.
+respond /metrics 404
+
+reverse_proxy 127.0.0.1:8080

internal/config/config.go

@@ -4,6 +4,7 @@ package config
 import (
 	"errors"
 	"fmt"
+	"io/fs"
 	"log/slog"
 	"math"
 	"net/netip"
@@ -14,6 +15,7 @@ import (
 	"sort"
 	"strconv"
 	"strings"
+	"syscall"
 	"time"
 
 	"git.eeqj.de/sneak/smartconfig"
@@ -778,19 +780,27 @@ func loadConfigFile(log *slog.Logger, appName string) (*smartconfig.Config, erro
 	for _, path := range configPaths {
 		cleanPath := filepath.Clean(path)
 
+		// Only a config file that does not exist is skipped, including
+		// one whose path runs through a file, such as under a HOME of
+		// /dev/null. One that cannot be read or does not parse is a
+		// fatal startup error.
 		_, statErr := os.Stat(cleanPath)
-		if statErr == nil {
+		if errors.Is(statErr, fs.ErrNotExist) || errors.Is(statErr, syscall.ENOTDIR) {
-			// A config file that exists but does not parse is a fatal
+			continue
-			// startup error, never something to skip over.
-			sc, err := smartconfig.NewFromConfigPath(path)
-			if err != nil {
-				return nil, fmt.Errorf("failed to parse config file %s: %w", path, err)
-			}
-
-			log.Info("loaded config file", "path", path)
-
-			return sc, nil
 		}
+
+		if statErr != nil {
+			return nil, fmt.Errorf("failed to read config file %s: %w", path, statErr)
+		}
+
+		sc, err := smartconfig.NewFromConfigPath(path)
+		if err != nil {
+			return nil, fmt.Errorf("failed to parse config file %s: %w", path, err)
+		}
+
+		log.Info("loaded config file", "path", path)
+
+		return sc, nil
 	}
 
 	return nil, nil //nolint:nilnil // nil config is valid (use defaults)

internal/config/config_validation_internal_test.go

@@ -564,6 +564,116 @@ func TestMalformedConfigFileAbortsStartup(t *testing.T) {
 	t.Logf("got expected error: %v", err)
 }
 
+// TestConfigFileInDirectoryPixaMayNotEnterAbortsStartup checks that a
+// config file pixa cannot read because it may not enter its directory
+// aborts startup instead of being passed over.
+func TestConfigFileInDirectoryPixaMayNotEnterAbortsStartup(t *testing.T) {
+	if os.Geteuid() == 0 {
+		t.Skip("root may enter any directory")
+	}
+
+	home := t.TempDir()
+	configDir := filepath.Join(home, ".config", "pixa-test-nonexistent-app")
+	configPath := filepath.Join(configDir, "config.yml")
+
+	err := os.MkdirAll(configDir, 0o700)
+	if err != nil {
+		t.Fatalf("failed to create config directory: %v", err)
+	}
+
+	err = os.WriteFile(configPath, []byte(signingKeyLine), 0o600)
+	if err != nil {
+		t.Fatalf("failed to write config: %v", err)
+	}
+
+	err = os.Chmod(configDir, 0)
+	if err != nil {
+		t.Fatalf("failed to remove the config directory's permissions: %v", err)
+	}
+
+	// Give the directory back its permissions so t.TempDir can remove it.
+	t.Cleanup(func() {
+		//nolint:gosec // G302: a directory needs its execute bit to be removed
+		_ = os.Chmod(configDir, 0o700)
+	})
+
+	// The ~/.config candidate is the only one that exists: the appname
+	// rules out /etc, and the working directory is empty.
+	t.Setenv("PIXA_CONFIG_PATH", "")
+	t.Setenv("HOME", home)
+	t.Chdir(t.TempDir())
+
+	log := slog.New(slog.DiscardHandler)
+
+	sc, err := loadConfigFile(log, "pixa-test-nonexistent-app")
+	if err == nil {
+		t.Fatalf("config file pixa cannot read must abort startup, got config: %v",
+			sc)
+	}
+
+	t.Logf("got expected error: %v", err)
+
+	if !strings.Contains(err.Error(), configPath) {
+		t.Errorf("error %q does not name the config file %s", err.Error(), configPath)
+	}
+}
+
+// TestConfigFileLinkingToItselfAbortsStartup checks that a config file
+// pixa cannot read for a reason other than not existing aborts startup,
+// as root too: a symbolic link to itself fails with "too many levels of
+// symbolic links".
+func TestConfigFileLinkingToItselfAbortsStartup(t *testing.T) {
+	workDir := t.TempDir()
+
+	err := os.Symlink("config.yml", filepath.Join(workDir, "config.yml"))
+	if err != nil {
+		t.Fatalf("failed to create symbolic link: %v", err)
+	}
+
+	// Only the working directory's config.yml is there: the appname rules
+	// out /etc, and HOME is empty.
+	t.Setenv("PIXA_CONFIG_PATH", "")
+	t.Setenv("HOME", t.TempDir())
+	t.Chdir(workDir)
+
+	log := slog.New(slog.DiscardHandler)
+
+	sc, err := loadConfigFile(log, "pixa-test-nonexistent-app")
+	if err == nil {
+		t.Fatalf("config file pixa cannot read must abort startup, got config: %v",
+			sc)
+	}
+
+	t.Logf("got expected error: %v", err)
+
+	if !strings.Contains(err.Error(), "config.yml") {
+		t.Errorf("error %q does not name the config file config.yml", err.Error())
+	}
+}
+
+// TestConfigPathThroughFileIsPassedOver checks that a config file path
+// that runs through a file, such as one under a HOME of /dev/null, is
+// passed over like one that does not exist, since no file can be there.
+func TestConfigPathThroughFileIsPassedOver(t *testing.T) {
+	// No config file is there: the appname rules out /etc, HOME is
+	// /dev/null, and the working directory is empty.
+	t.Setenv("PIXA_CONFIG_PATH", "")
+	t.Setenv("HOME", os.DevNull)
+	t.Chdir(t.TempDir())
+
+	log := slog.New(slog.DiscardHandler)
+
+	sc, err := loadConfigFile(log, "pixa-test-nonexistent-app")
+	if err != nil {
+		t.Fatalf("a config path through a file must be passed over, got error: %v",
+			err)
+	}
+
+	if sc != nil {
+		t.Errorf("expected no config file, got config: %v", sc)
+	}
+}
+
 func TestEnsureStateDirCreatesDirectory(t *testing.T) {
 	t.Parallel()