Entrypoint checks DATA_DIR in full before acting on it as root (closes #80)

`bin/entrypoint.sh` now checks `DATA_DIR` before it creates anything or
changes an owner or mode: it must be `/data` or a path below it with no
`.`, `..` or empty part, and no part of it that exists, `/data`
included, may be a symbolic link. Anything else stops the start with one
message naming `DATA_DIR`. Only then does it create `DATA_DIR`, give
`/data` and everything in it to `netwatch` (`chown -R -h`, so a link in
it is not followed) and set mode 750 on `/data` and `DATA_DIR`. The
README section "Running under upaas" says which values are accepted.

Model: opus-5-5

Dockerfile

@@ -77,8 +77,9 @@ COPY --from=frontend /app/dist /usr/share/nginx/html
 COPY --from=builder /src/netwatch-server /usr/local/bin/netwatch-server
 COPY bin/entrypoint.sh /usr/local/bin/entrypoint.sh
 
+# bin/entrypoint.sh creates DATA_DIR at start and gives it and /data to
+# the netwatch user, whatever is mounted there.
 ENV DATA_DIR=/data/reports
-RUN mkdir -p /data/reports && chown -R netwatch:netwatch /data
 VOLUME /data
 
 # The default public port; PORT changes it.

README.md

@@ -192,8 +192,9 @@ only inside the container, on `127.0.0.1:8081`. The image:
 - Sends the security headers `REPO_POLICIES.md` requires on every response, as
   `security-headers.conf` sets them, in place of the backend's own
 - Stores reports in `DATA_DIR`, `/data/reports` by default, on the `/data`
-  volume. The backend runs as user `netwatch` (uid 1000), so a directory
-  bind-mounted at `/data` must be writable by uid 1000
+  volume. Before the backend starts, the image creates `DATA_DIR` and gives it
+  and `/data` to user `netwatch` (uid 1000), which the backend runs as, so a
+  host directory bind-mounted at `/data` ends up owned by uid 1000
 - Writes buffered reports to disk on `docker stop`, and exits non-zero if nginx
   or the backend exits on its own, so the platform restarts it
 
@@ -203,16 +204,6 @@ What the [upaas](https://git.eeqj.de/sneak/upaas) app for netwatch needs:
 
 - **Port:** container port `8080`.
 - **Volume:** container path `/data`; the reports are kept in `/data/reports`.
-- **First run:** upaas bind-mounts the host directory it is given and does not
-  create it, and the backend, which runs as uid 1000, does not start unless it
-  can write there. Create the directory, owned by uid 1000, before the first
-  deploy:
-
-    ```bash
-    mkdir -p /path/to/data
-    chown 1000:1000 /path/to/data
-    ```
-
 - **Environment variables:** none is required. An empty one counts as unset, and
   one set to a value netwatch cannot use stops the container at start, with the
   reason in its log.
@@ -225,8 +216,10 @@ What the [upaas](https://git.eeqj.de/sneak/upaas) app for netwatch needs:
     - `CORS_ALLOWED_ORIGINS`, default empty: other origins whose pages may call
       the API
     - `DEBUG`, default `false`: debug logging
-    - `DATA_DIR`, default `/data/reports`: leave unset; reports kept outside
-      `/data` do not survive a redeploy
+    - `DATA_DIR`, default `/data/reports`: the directory the reports are kept
+      in: `/data` or a path below it, with no `.` or `..` part and no extra `/`.
+      The container also stops if a part of the path that exists, `/data`
+      included, is a symbolic link
     - `TRUSTED_PROXIES`, default empty: set it to the address the reverse proxy
       in front of the container connects from, as an IP address or CIDR; several
       are separated by commas. nginx takes the client address from

TODO.md

@@ -23,6 +23,23 @@ latest run passes.
 
 # Completed Steps
 
+- 2026-09-29: `bin/entrypoint.sh` checks `DATA_DIR` in full before it acts on it
+  as root (issue #80): `DATA_DIR` must be `/data` or a path below it with no
+  `.`, `..` or empty part, and no part of it that exists, `/data` included, may
+  be a symbolic link; anything else stops the start with a message naming
+  `DATA_DIR`. Only then is `DATA_DIR` created and `/data` given to `netwatch`,
+  so a refused start no longer creates directories outside `/data`, and
+  `DATA_DIR=/etc` no longer gives `/etc` to `netwatch`. The `README.md` section
+  "Running under upaas" says which values are accepted
+- 2026-09-29: the container sets up its own data directory (issue #75):
+  `bin/entrypoint.sh`, still as root, creates `DATA_DIR` if missing and gives it
+  and `/data` to the `netwatch` user with mode 750 before starting the backend
+  as that user, so an empty host directory owned by root, or one holding files
+  from another uid, works with no step on the host. It stops the start instead
+  when a symbolic link is on the path to `DATA_DIR`, since root would change
+  whatever the link points to. The `README.md` first-run step that created and
+  chowned the host directory is gone, and the image no longer sets that
+  ownership at build time
 - 2026-09-29: CI can no longer pass on checks that did not run (issue #37):
   `script/cibuild` is now the org model, byte for byte. It runs
   `script/bootstrap` and `script/check`, then builds the image with `--no-cache`

backend/README.md

@@ -104,7 +104,8 @@ this server. The image's entrypoint, `bin/entrypoint.sh`, starts the server as
 user `netwatch` (uid 1000) with `BIND_ADDRESS=127.0.0.1` and `PORT=8081`, so
 only nginx reaches it, and with `TRUSTED_PROXIES=127.0.0.1/32`, so it takes the
 client address nginx passes on and no other. `DATA_DIR` is `/data/reports`, on
-the `/data` volume, which `netwatch` owns. nginx replaces the security headers
+the `/data` volume; the entrypoint creates it and gives it and `/data` to
+`netwatch` before starting the server. nginx replaces the security headers
 this server sets with those in the root `security-headers.conf`, so those are
 what clients of the image see.
 

bin/entrypoint.sh

@@ -61,6 +61,47 @@ for proxy in $(printf '%s' "$TRUSTED_PROXIES" | tr ',' ' '); do
     echo "set_real_ip_from $cidr;"
 done > /etc/nginx/trusted-proxies.conf
 
+# netwatch-server keeps its report files in DATA_DIR, on the /data
+# volume, which may be a host directory owned by root or by another
+# uid. /data and everything in it are given to the netwatch user here,
+# and /data and DATA_DIR get the mode the server gives a directory it
+# creates, so the host directory needs no preparing.
+#
+# This runs as root, so nothing is created or changed until DATA_DIR is
+# known to be /data or a path below it, with no '.', '..' or empty
+# part, and no part of it that exists, /data included, is a symbolic
+# link: the netwatch user can put one in /data, and root would follow
+# it anywhere in the container. Nothing else runs in the container yet,
+# so no link can appear after the check.
+export DATA_DIR="${DATA_DIR:-/data/reports}"
+data_dir_ok() {
+    # With a / added at the end, a last part of '.' or '..', and a / at
+    # the end, match these patterns too.
+    case "$DATA_DIR/" in
+        */./* | */../* | *//*) return 1 ;;
+        /data/*) ;;
+        *) return 1 ;;
+    esac
+    # Each part from DATA_DIR up to /data. [ -L ] is false for a part
+    # that does not exist.
+    dir="$DATA_DIR"
+    while [ "$dir" != /data ]; do
+        [ -L "$dir" ] && return 1
+        dir="${dir%/*}"
+    done
+    [ ! -L /data ]
+}
+if ! data_dir_ok; then
+    echo "entrypoint: DATA_DIR must be /data or a path below it, with no" \
+        "'.', '..', extra '/' or symbolic link on it, not '$DATA_DIR'" >&2
+    exit 1
+fi
+mkdir -p "$DATA_DIR" || exit 1
+# -h: a symbolic link in /data is itself given to netwatch, not what it
+# points to.
+chown -R -h netwatch:netwatch /data || exit 1
+chmod 750 /data "$DATA_DIR" || exit 1
+
 # A stop signal is only noted here; the loop below acts on it.
 stop_requested=""
 trap 'stop_requested=yes' TERM INT