fix(backend): rate-limit and cap report ingest, drop wildcard CORS (closes #20)

POST /api/v1/reports stays unauthenticated but is bounded. Each client
address, as the trusted-proxy logic resolves it, may send
REPORTS_PER_MINUTE reports a minute (default 60), counted by
go-chi/httprate over a sliding minute; past that it gets 429 with
Retry-After. reportbuf refuses a report that would take the report
files past DATA_DIR_MAX_BYTES (default 1 GiB) with ErrFull, answered
with 507; the count starts from the files already in DATA_DIR, and
reports not yet written count at their uncompressed size. CORS adds
nothing unless CORS_ALLOWED_ORIGINS lists origins. A limit that is
not a positive number stops the server from starting.

Model: opus-5-5

.gitea/workflows/check.yml

@@ -6,5 +6,5 @@ jobs:
         steps:
             # actions/checkout v4.2.2, 2026-02-22
             - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
-            # script/cibuild builds both images.
+            # script/cibuild builds the image, whose stages run every check.
             - run: script/cibuild

Dockerfile

@@ -1,5 +1,51 @@
+# The one image netwatch ships: nginx serves the built frontend and
+# passes /api/ and /.well-known/healthcheck to netwatch-server, the Go
+# backend, which runs in the same container on loopback only.
+# bin/entrypoint.sh starts and watches both.
+
+# Lint stage — fast feedback on formatting and lint issues. The
+# golangci/golangci-lint image ships Go, gofmt, make and the linter, so
+# nothing is installed here. The root make lint builds this stage alone.
+# golangci/golangci-lint:v2.12.2 (2026-08-10)
+FROM golangci/golangci-lint@sha256:5cceeef04e53efe1470638d4b4b4f5ceefd574955ab3941b2d9a68a8c9ad5240 AS lint
+
+WORKDIR /src
+COPY backend/go.mod backend/go.sum ./
+RUN go mod download
+COPY backend/ .
+RUN make fmt-check
+RUN make lint
+
+# Backend build stage
+# golang:1.25-alpine (2026-02-27)
+FROM golang:1.25-alpine@sha256:f6751d823c26342f9506c03797d2527668d095b0a15f1862cddb4d927a7a4ced AS builder
+
+RUN apk add --no-cache make
+
+WORKDIR /src
+
+# Force BuildKit to run the lint stage before proceeding. BuildKit runs
+# stages in parallel by default; without this no-op copy a lint failure
+# would not gate compilation.
+COPY --from=lint /src/go.sum /dev/null
+
+COPY backend/go.mod backend/go.sum ./
+RUN go mod download
+COPY backend/ .
+
+RUN make test
+
+# make build is a shim around backend/script/build, the one definition
+# of the build command:
+# CGO_ENABLED=0 go build -trimpath -ldflags "-s -w -X main.Version=... -X main.Buildarch=..."
+# That script reads VERSION from the environment, so it is handed over
+# there rather than as a make variable.
+ARG VERSION=dev
+RUN VERSION="${VERSION}" make build
+
+# Frontend stage
 # node:22-alpine as of 2026-02-22
-FROM node@sha256:e4bf2a82ad0a4037d28035ae71529873c069b13eb0455466ae0bc13363826e34 AS build
+FROM node@sha256:e4bf2a82ad0a4037d28035ae71529873c069b13eb0455466ae0bc13363826e34 AS frontend
 WORKDIR /app
 COPY package.json yarn.lock ./
 RUN yarn install --frozen-lockfile
@@ -8,16 +54,33 @@ COPY . .
 # make frontend-check is the frontend half of make check (test + lint +
 # fmt-check); its test step is the production yarn build, so this both
 # produces dist/ and gates the image on lint/fmt-check/test regressions.
-# This node stage has neither Go nor Docker for the other half, which
+# This node stage has neither Go nor Docker; the lint and builder stages
-# Dockerfile.backend gates; script/cibuild builds both images.
+# above gate the backend half.
 RUN make frontend-check
 
+# Runtime stage
 # nginx:stable-alpine as of 2026-02-22
 FROM nginx@sha256:15e96e59aa3b0aada3a121296e3bce117721f42d88f5f64217ef4b18f458c6ab
+
+# netwatch-server runs as this user, which owns the report directory.
+# nginx keeps the image's own arrangement: its main process runs as
+# root, its worker processes as the nginx user.
+RUN addgroup -g 1000 -S netwatch && \
+    adduser -u 1000 -S netwatch -G netwatch
+
 RUN rm /etc/nginx/conf.d/default.conf
 COPY nginx.conf /etc/nginx/conf.d/netwatch.conf
-COPY --from=build /app/dist /usr/share/nginx/html
+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
+
+ENV DATA_DIR=/data/reports
+RUN mkdir -p /data/reports && chown -R netwatch:netwatch /data
+VOLUME /data
 
 EXPOSE 8080
 
-CMD ["nginx", "-g", "daemon off;"]
+# The nginx image stops its container with SIGQUIT; the entrypoint
+# acts on TERM and INT.
+STOPSIGNAL SIGTERM
+ENTRYPOINT ["/usr/local/bin/entrypoint.sh"]

Dockerfile.backend

@@ -1,49 +0,0 @@
-# Lint stage — fast feedback on formatting and lint issues. The
-# golangci/golangci-lint image ships Go, gofmt, make and the linter, so
-# nothing is installed here.
-# golangci/golangci-lint:v2.12.2 (2026-08-10)
-FROM golangci/golangci-lint@sha256:5cceeef04e53efe1470638d4b4b4f5ceefd574955ab3941b2d9a68a8c9ad5240 AS lint
-
-WORKDIR /src
-COPY backend/go.mod backend/go.sum ./
-RUN go mod download
-COPY backend/ .
-RUN make fmt-check
-RUN make lint
-
-# Build stage
-# golang:1.25-alpine (2026-02-27)
-FROM golang:1.25-alpine@sha256:f6751d823c26342f9506c03797d2527668d095b0a15f1862cddb4d927a7a4ced AS builder
-
-RUN apk add --no-cache make
-
-WORKDIR /src
-
-# Force BuildKit to run the lint stage before proceeding. BuildKit runs
-# stages in parallel by default; without this no-op copy a lint failure
-# would not gate compilation.
-COPY --from=lint /src/go.sum /dev/null
-
-COPY backend/go.mod backend/go.sum ./
-RUN go mod download
-COPY backend/ .
-
-RUN make test
-
-# make build is a shim around backend/script/build, the one definition
-# of the build command:
-# CGO_ENABLED=0 go build -trimpath -ldflags "-s -w -X main.Version=... -X main.Buildarch=..."
-# That script reads VERSION from the environment, so it is handed over
-# there rather than as a make variable.
-ARG VERSION=dev
-RUN VERSION="${VERSION}" make build
-
-# Runtime stage
-# alpine:3.23 (2026-02-27)
-FROM alpine:3.23@sha256:25109184c71bdad752c8312a8623239686a9a2071e8825f20acb8f2198c3f659
-
-RUN apk add --no-cache ca-certificates
-COPY --from=builder /src/netwatch-server /usr/local/bin/netwatch-server
-
-EXPOSE 8080
-ENTRYPOINT ["netwatch-server"]

README.md

@@ -44,11 +44,11 @@ halves, so the root `make check` fails if either one is broken. We provide:
   linter in Docker
 - `script/setup` — make a fresh clone ready for development: bootstrap plus the
   git pre-commit hook
-- `script/projectname` — print the project name (used for the Docker image tags)
+- `script/projectname` — print the project name (used for the Docker image tag)
 - `script/test` — run `script/frontend-test`, then the backend's Go tests, both
   within one 30-second timeout
 - `script/lint` — run `script/frontend-lint`, then golangci-lint in Docker, by
-  building the lint stage of `Dockerfile.backend` without the cache
+  building the lint stage of `Dockerfile` without the cache
 - `script/fmt` — format all files (writes): prettier, then gofmt over `backend/`
 - `script/fmt-check` — check formatting (read-only): prettier, then gofmt
 - `script/check` — run test, lint, and fmt-check
@@ -63,9 +63,9 @@ halves, so the root `make check` fails if either one is broken. We provide:
   frontend in a containerised headless Chrome (see
   [test/viewport/README.md](test/viewport/README.md)). Not part of
   `script/check`: it needs Docker and takes minutes.
-- `script/docker` — build both images, tagged via `script/projectname`:
+- `script/docker` — build the image from `Dockerfile` without the build cache,
-  `netwatch` from `Dockerfile` and `netwatch-server` from `Dockerfile.backend`
+  tagged `netwatch` via `script/projectname`
-- `script/cibuild` — CI entrypoint: builds both images
+- `script/cibuild` — CI entrypoint: builds the image
 - `script/precommit` — run by the git pre-commit hook; runs `script/check`
 - `script/install-precommit` — install the git pre-commit hook
 
@@ -178,13 +178,21 @@ After running `yarn build`, deploy the contents of the `dist/` directory to any
 static file host (S3, GCS, Cloudflare Pages, Vercel, Netlify, GitHub Pages) or
 use the Docker image behind a reverse proxy.
 
-The Docker image:
+The Docker image, built from `Dockerfile`, is the whole service in one
+container: nginx serves the built frontend and passes `/api/` and
+`/.well-known/healthcheck` to the Go backend, `netwatch-server`, which listens
+only inside the container, on `127.0.0.1:8081`. The image:
 
 - Listens on port 8080 by default (override with `PORT` env var)
 - Trusts `X-Forwarded-For` from RFC1918 reverse proxies (10/8, 172.16/12,
   192.168/16)
 - Sends access logs to stdout
 - Caches static assets with immutable headers
+- 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
+- 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
 
 ## Browser Compatibility
 

TODO.md

@@ -23,6 +23,23 @@ latest run passes.
 
 # Completed Steps
 
+- 2026-09-29: bounded the report endpoint (issue #20): `POST /api/v1/reports`
+  still needs no credentials, but each client address, as resolved through
+  `TRUSTED_PROXIES`, may send `REPORTS_PER_MINUTE` (default 60) reports a
+  minute, counted by `go-chi/httprate`, and past that gets 429 with
+  `Retry-After`; the report files in `DATA_DIR`, counted from start with those
+  already there, may total at most `DATA_DIR_MAX_BYTES` (default 1 GiB), past
+  which reports get 507; and the wildcard CORS is gone: no CORS headers unless
+  `CORS_ALLOWED_ORIGINS` lists origins. Deleting report files frees room only at
+  the next start; pruning is issue #54
+- 2026-09-28: one container image (issue #52): the root `Dockerfile` builds the
+  only image, and `Dockerfile.backend` is gone. nginx serves the frontend on
+  port 8080 and proxies `/api/` and `/.well-known/healthcheck` to the backend,
+  which listens on `127.0.0.1:8081` in the same container; the new
+  `BIND_ADDRESS` setting sets its listen address. `bin/entrypoint.sh` starts
+  both, passes TERM and INT on to both, and exits non-zero when either exits on
+  its own. The backend runs as user `netwatch` and stores reports on the `/data`
+  volume. `script/docker` is the org model again
 - 2026-09-28: unified the gate (issue #16): the root `make check` covers the Go
   backend as well as the frontend, and the pre-commit hook with it; the backend
   moved onto scripts-to-rule-them-all (`backend/script/*`, `backend/Makefile` as

backend/Makefile

@@ -2,7 +2,7 @@
 # Entrypoints section of README.md). There is no check, hooks or docker
 # target here: the root Makefile's check covers this directory, its
 # hooks target installs the repo's only pre-commit hook, and its docker
-# target builds this image, whose build context is the repo root.
+# target builds the one image, which contains this backend.
 
 .PHONY: all build test lint fmt fmt-check run clean
 

backend/README.md

@@ -11,15 +11,16 @@ From this directory:
 make run
 ```
 
-From the repo root, which is also the build context of `Dockerfile.backend`:
+From the repo root, whose `Dockerfile` builds the one image that ships this
+backend behind nginx (see [Container image](#container-image)):
 
 ```bash
 # Run tests, lint, and format check over the frontend and this backend
 make check
 
-# Build both images, including netwatch-server
+# Build the image: nginx, the frontend and this backend
 make docker
-docker run -p 8080:8080 netwatch-server
+docker run -p 8080:8080 netwatch
 ```
 
 ## Entrypoints
@@ -27,7 +28,7 @@ docker run -p 8080:8080 netwatch-server
 This directory follows the same
 [Scripts to Rule Them All](https://github.com/github/scripts-to-rule-them-all)
 pattern as the repo root: the targets in `backend/Makefile` are thin shims over
-`backend/script/`. `Dockerfile.backend` runs them, and the root scripts call
+`backend/script/`. The root `Dockerfile` runs them, and the root scripts call
 `test`, `fmt` and `fmt-check`:
 
 - `script/build` — compile the static `netwatch-server` binary with its version
@@ -37,8 +38,8 @@ pattern as the repo root: the targets in `backend/Makefile` are thin shims over
 - `script/test` — run the Go tests under a 30-second timeout
 - `script/lint` — check `.golangci.yml` against its pinned sha256, then run
   golangci-lint. It runs inside the golangci-lint image of the lint stage of
-  `Dockerfile.backend`; from a checkout, run `make lint` at the repo root, which
+  the root `Dockerfile`; from a checkout, run `make lint` at the repo root,
-  builds that stage
+  which builds that stage
 - `script/fmt` — format the Go sources (writes)
 - `script/fmt-check` — check Go formatting (read-only)
 - `script/run` — build and run the server locally
@@ -46,7 +47,7 @@ pattern as the repo root: the targets in `backend/Makefile` are thin shims over
 
 There is no `check`, `hooks` or `docker` target here: the root `make check`
 covers this directory, the root `make hooks` installs the repo's only pre-commit
-hook, and the root `make docker` builds this image.
+hook, and the root `make docker` builds the image that contains this backend.
 
 ## Rationale
 
@@ -74,24 +75,69 @@ Internal packages in `internal/` follow standard Go project layout:
 
 ### Configuration
 
-| Variable          | Default              | Description                                                                                              |
+| Variable               | Default              | Description                                                                                              |
-| ----------------- | -------------------- | -------------------------------------------------------------------------------------------------------- |
+| ---------------------- | -------------------- | -------------------------------------------------------------------------------------------------------- |
-| `PORT`            | `8080`               | HTTP listen port                                                                                         |
+| `BIND_ADDRESS`         | empty                | IP address to listen on; empty listens on every interface                                                |
-| `DATA_DIR`        | `./data/reports`     | Directory for compressed reports                                                                         |
+| `PORT`                 | `8080`               | HTTP listen port                                                                                         |
-| `DEBUG`           | `false`              | Enable debug logging                                                                                     |
+| `DATA_DIR`             | `./data/reports`     | Directory for compressed reports                                                                         |
-| `TRUSTED_PROXIES` | loopback + RFC1918   | Comma-separated CIDRs whose `X-Forwarded-For` / `X-Real-IP` headers are trusted for client IP resolution |
+| `DATA_DIR_MAX_BYTES`   | `1073741824` (1 GiB) | Most the report files in `DATA_DIR` may total; see [Report limits](#report-limits)                       |
+| `DEBUG`                | `false`              | Enable debug logging                                                                                     |
+| `TRUSTED_PROXIES`      | loopback + RFC1918   | Comma-separated CIDRs whose `X-Forwarded-For` / `X-Real-IP` headers are trusted for client IP resolution |
+| `REPORTS_PER_MINUTE`   | `60`                 | Reports each client address may send a minute; see [Report limits](#report-limits)                       |
+| `CORS_ALLOWED_ORIGINS` | empty                | Comma-separated origins whose pages may call the API; see [CORS](#cors)                                  |
 
 `TRUSTED_PROXIES` defaults to `127.0.0.1/32,::1/128,10.0.0.0/8,172.16.0.0/12,192.168.0.0/16`.
 The loopback entries cover the reverse proxy that shares the container; the
 RFC1918 ranges match `nginx.conf`. A request whose direct peer is outside this
 set has its forwarded headers ignored, and the direct peer is logged instead.
 
+### Container image
+
+The root `Dockerfile` builds one image in which nginx listens on the public port
+8080, serves the frontend, and proxies `/api/` and `/.well-known/healthcheck` to
+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. `DATA_DIR` is `/data/reports`, on the `/data` volume,
+which `netwatch` owns.
+
 ### Report storage
 
 Reports are written as `reports-<timestamp>.jsonl.zst` files in `DATA_DIR`.
 Each file contains one JSON object per line, compressed with zstd. Files are
 created with `O_EXCL` to prevent overwrites.
 
+### Report limits
+
+`POST /api/v1/reports` takes reports from anyone who can reach it, without
+credentials, so it is bounded instead. Both refusals below answer with the same
+`{"status":"error"}` body as any other error.
+
+- **Rate limit.** Each client address, resolved through `TRUSTED_PROXIES`, may
+  send `REPORTS_PER_MINUTE` reports a minute; past that it gets 429 with
+  `Retry-After: 60`. The minute slides: reports from the minute before still
+  count, fading out over the current one, so an address is sure never to be
+  refused only while it sends at most half of `REPORTS_PER_MINUTE` in any 60
+  seconds. The page sends one report a minute from each open tab, so the default
+  of 60 refuses nothing from up to 30 tabs behind one address, such as a
+  household or an office sharing it, however their reports bunch up. Report
+  responses also carry `X-RateLimit-Limit`, `X-RateLimit-Remaining` and
+  `X-RateLimit-Reset` headers.
+- **Size cap.** The report files in `DATA_DIR` may total at most
+  `DATA_DIR_MAX_BYTES`, counting the files already there at start. Reports
+  waiting in memory count at their uncompressed size until they are written, so
+  a report that would take the total past the cap is refused with 507, and
+  nothing of it is stored. Deleting report files frees room only at the next
+  start, when the files are counted again. The default of 1 GiB is small enough
+  for any host; set it to the space you can give `DATA_DIR`.
+
+### CORS
+
+The page calls the API from the origin it is served from, so by default the
+server sends no CORS headers, and browsers let no other origin's pages call it.
+To serve the page from elsewhere, list that origin in `CORS_ALLOWED_ORIGINS`
+(for example `https://netwatch.example.com`); pages from a listed origin may
+`GET` and `POST` with a `Content-Type` header.
+
 ## TODO
 
 - Add integration test that POSTs a report and verifies the compressed output

backend/go.mod

@@ -5,6 +5,7 @@ go 1.25.5
 require (
 	github.com/go-chi/chi/v5 v5.2.5
 	github.com/go-chi/cors v1.2.2
+	github.com/go-chi/httprate v0.16.0
 	github.com/joho/godotenv v1.5.1
 	github.com/klauspost/compress v1.18.4
 	github.com/spf13/viper v1.21.0
@@ -14,6 +15,7 @@ require (
 require (
 	github.com/fsnotify/fsnotify v1.9.0 // indirect
 	github.com/go-viper/mapstructure/v2 v2.4.0 // indirect
+	github.com/klauspost/cpuid/v2 v2.2.10 // indirect
 	github.com/pelletier/go-toml/v2 v2.2.4 // indirect
 	github.com/sagikazarmark/locafero v0.11.0 // indirect
 	github.com/sourcegraph/conc v0.3.1-0.20240121214520-5f936abd7ae8 // indirect
@@ -21,10 +23,11 @@ require (
 	github.com/spf13/cast v1.10.0 // indirect
 	github.com/spf13/pflag v1.0.10 // indirect
 	github.com/subosito/gotenv v1.6.0 // indirect
+	github.com/zeebo/xxh3 v1.0.2 // indirect
 	go.uber.org/dig v1.19.0 // indirect
 	go.uber.org/multierr v1.10.0 // indirect
 	go.uber.org/zap v1.26.0 // indirect
 	go.yaml.in/yaml/v3 v3.0.4 // indirect
-	golang.org/x/sys v0.29.0 // indirect
+	golang.org/x/sys v0.30.0 // indirect
 	golang.org/x/text v0.28.0 // indirect
 )

backend/go.sum

@@ -8,6 +8,8 @@ github.com/go-chi/chi/v5 v5.2.5 h1:Eg4myHZBjyvJmAFjFvWgrqDTXFyOzjj7YIm3L3mu6Ug=
 github.com/go-chi/chi/v5 v5.2.5/go.mod h1:X7Gx4mteadT3eDOMTsXzmI4/rwUpOwBHLpAfupzFJP0=
 github.com/go-chi/cors v1.2.2 h1:Jmey33TE+b+rB7fT8MUy1u0I4L+NARQlK6LhzKPSyQE=
 github.com/go-chi/cors v1.2.2/go.mod h1:sSbTewc+6wYHBBCW7ytsFSn836hqM7JxpglAy2Vzc58=
+github.com/go-chi/httprate v0.16.0 h1:8V5DH9j6pSK6UQoBsTpvMyFxycqaKEIToyPKzHJjUa8=
+github.com/go-chi/httprate v0.16.0/go.mod h1:A8lo+qRhk+s9LiuP5saS7XCGDXRXMcrueq0NfIuCa/I=
 github.com/go-viper/mapstructure/v2 v2.4.0 h1:EBsztssimR/CONLSZZ04E8qAkxNYq4Qp9LvH92wZUgs=
 github.com/go-viper/mapstructure/v2 v2.4.0/go.mod h1:oJDH3BJKyqBA2TXFhDsKDGDTlndYOZ6rGS0BRZIxGhM=
 github.com/google/go-cmp v0.6.0 h1:ofyhxvXcZhMsU5ulbFiLKl/XBFqE1GSq7atu8tAmTRI=
@@ -16,6 +18,8 @@ github.com/joho/godotenv v1.5.1 h1:7eLL/+HRGLY0ldzfGMeQkb7vMd0as4CfYvUVzLqw0N0=
 github.com/joho/godotenv v1.5.1/go.mod h1:f4LDr5Voq0i2e/R5DDNOoa2zzDfwtkZa6DnEwAbqwq4=
 github.com/klauspost/compress v1.18.4 h1:RPhnKRAQ4Fh8zU2FY/6ZFDwTVTxgJ/EMydqSTzE9a2c=
 github.com/klauspost/compress v1.18.4/go.mod h1:R0h/fSBs8DE4ENlcrlib3PsXS61voFxhIs2DeRhCvJ4=
+github.com/klauspost/cpuid/v2 v2.2.10 h1:tBs3QSyvjDyFTq3uoc/9xFpCuOsJQFNPiAhYdw2skhE=
+github.com/klauspost/cpuid/v2 v2.2.10/go.mod h1:hqwkgyIinND0mEev00jJYCxPNVRVXFQeu1XKlok6oO0=
 github.com/kr/pretty v0.3.1 h1:flRD4NNwYAUpkphVc1HcthR4KEIFJ65n8Mw5qdRn3LE=
 github.com/kr/pretty v0.3.1/go.mod h1:hoEshYVHaxMs3cyo3Yncou5ZscifuDolrwPKZanG3xk=
 github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
@@ -42,6 +46,10 @@ github.com/stretchr/testify v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu
 github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD981Nm4B6U=
 github.com/subosito/gotenv v1.6.0 h1:9NlTDc1FTs4qu0DDq7AEtTPNw6SVm7uBMsUCUjABIf8=
 github.com/subosito/gotenv v1.6.0/go.mod h1:Dk4QP5c2W3ibzajGcXpNraDfq2IrhjMIvMSWPKKo0FU=
+github.com/zeebo/assert v1.3.0 h1:g7C04CbJuIDKNPFHmsk4hwZDO5O+kntRxzaUoNXj+IQ=
+github.com/zeebo/assert v1.3.0/go.mod h1:Pq9JiuJQpG8JLJdtkwrJESF0Foym2/D9XMU5ciN/wJ0=
+github.com/zeebo/xxh3 v1.0.2 h1:xZmwmqxHZA8AI603jOQ0tMqmBr9lPeFwGg6d+xy9DC0=
+github.com/zeebo/xxh3 v1.0.2/go.mod h1:5NWz9Sef7zIDm2JHfFlcQvNekmcEl9ekUZQQKCYaDcA=
 go.uber.org/dig v1.19.0 h1:BACLhebsYdpQ7IROQ1AGPjrXcP5dF80U3gKoFzbaq/4=
 go.uber.org/dig v1.19.0/go.mod h1:Us0rSJiThwCv2GteUN0Q7OKvU7n5J4dxZ9JKUXozFdE=
 go.uber.org/fx v1.24.0 h1:wE8mruvpg2kiiL1Vqd0CC+tr0/24XIB10Iwp2lLWzkg=
@@ -54,8 +62,8 @@ go.uber.org/zap v1.26.0 h1:sI7k6L95XOKS281NhVKOFCUNIvv9e0w4BF8N3u+tCRo=
 go.uber.org/zap v1.26.0/go.mod h1:dtElttAiwGvoJ/vj4IwHBS/gXsEu/pZ50mUIRWuG0so=
 go.yaml.in/yaml/v3 v3.0.4 h1:tfq32ie2Jv2UxXFdLJdh3jXuOzWiL1fo0bu/FbuKpbc=
 go.yaml.in/yaml/v3 v3.0.4/go.mod h1:DhzuOOF2ATzADvBadXxruRBLzYTpT36CKvDb3+aBEFg=
-golang.org/x/sys v0.29.0 h1:TPYlXGxvx1MGTn2GiZDhnjPA9wZzZeGKHHmKhHYvgaU=
+golang.org/x/sys v0.30.0 h1:QjkSwP/36a20jFYWkSue1YwXzLmsV5Gfq7Eiy72C1uc=
-golang.org/x/sys v0.29.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
+golang.org/x/sys v0.30.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
 golang.org/x/text v0.28.0 h1:rhazDwis8INMIwQ4tpjLDzUhx6RlXqZNPEM0huQojng=
 golang.org/x/text v0.28.0/go.mod h1:U8nCwOR8jO/marOQ0QbDiOngZVEBB7MAiitBuMjXiNU=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=

backend/internal/config/config.go

@@ -4,6 +4,7 @@ package config
 
 import (
 	"errors"
+	"fmt"
 	"log/slog"
 	"strings"
 
@@ -23,6 +24,15 @@ import (
 const defaultTrustedProxies = "127.0.0.1/32,::1/128," +
 	"10.0.0.0/8,172.16.0.0/12,192.168.0.0/16"
 
+// Default limits on stored reports; backend/README.md gives the
+// reasons for these values.
+const (
+	defaultReportsPerMinute = 60
+	defaultDataDirMaxBytes  = 1 << 30 // 1 GiB
+)
+
+var errNotPositive = errors.New("must be a positive whole number")
+
 // Params defines the dependencies for Config.
 type Params struct {
 	fx.In
@@ -33,15 +43,19 @@ type Params struct {
 
 // Config holds the resolved application configuration.
 type Config struct {
-	DataDir         string
+	BindAddress        string
-	Debug           bool
+	CORSAllowedOrigins []string
-	MetricsPassword string
+	DataDir            string
-	MetricsUsername string
+	DataDirMaxBytes    int64
-	Port            int
+	Debug              bool
-	SentryDSN       string
+	MetricsPassword    string
-	TrustedProxies  []string
+	MetricsUsername    string
-	log             *slog.Logger
+	Port               int
-	params          *Params
+	ReportsPerMinute   int
+	SentryDSN          string
+	TrustedProxies     []string
+	log                *slog.Logger
+	params             *Params
 }
 
 // New loads configuration from env, .env files, and config
@@ -60,9 +74,15 @@ func New(
 
 	viper.AutomaticEnv()
 
+	// An empty CORS_ALLOWED_ORIGINS allows no other origin.
+	viper.SetDefault("CORS_ALLOWED_ORIGINS", "")
 	viper.SetDefault("DATA_DIR", "./data/reports")
+	viper.SetDefault("DATA_DIR_MAX_BYTES", defaultDataDirMaxBytes)
 	viper.SetDefault("DEBUG", "false")
+	// An empty BIND_ADDRESS listens on every interface.
+	viper.SetDefault("BIND_ADDRESS", "")
 	viper.SetDefault("PORT", "8080")
+	viper.SetDefault("REPORTS_PER_MINUTE", defaultReportsPerMinute)
 	viper.SetDefault("SENTRY_DSN", "")
 	viper.SetDefault("METRICS_USERNAME", "")
 	viper.SetDefault("METRICS_PASSWORD", "")
@@ -78,15 +98,31 @@ func New(
 	}
 
 	s := &Config{
-		DataDir:         viper.GetString("DATA_DIR"),
+		BindAddress:        viper.GetString("BIND_ADDRESS"),
-		Debug:           viper.GetBool("DEBUG"),
+		CORSAllowedOrigins: splitList(viper.GetString("CORS_ALLOWED_ORIGINS")),
-		MetricsPassword: viper.GetString("METRICS_PASSWORD"),
+		DataDir:            viper.GetString("DATA_DIR"),
-		MetricsUsername: viper.GetString("METRICS_USERNAME"),
+		DataDirMaxBytes:    viper.GetInt64("DATA_DIR_MAX_BYTES"),
-		Port:            viper.GetInt("PORT"),
+		Debug:              viper.GetBool("DEBUG"),
-		SentryDSN:       viper.GetString("SENTRY_DSN"),
+		MetricsPassword:    viper.GetString("METRICS_PASSWORD"),
-		TrustedProxies:  splitList(viper.GetString("TRUSTED_PROXIES")),
+		MetricsUsername:    viper.GetString("METRICS_USERNAME"),
-		log:             log,
+		Port:               viper.GetInt("PORT"),
-		params:          &params,
+		ReportsPerMinute:   viper.GetInt("REPORTS_PER_MINUTE"),
+		SentryDSN:          viper.GetString("SENTRY_DSN"),
+		TrustedProxies:     splitList(viper.GetString("TRUSTED_PROXIES")),
+		log:                log,
+		params:             &params,
+	}
+
+	// viper reads a value that is not a number as 0, so this also
+	// catches a mistyped setting.
+	if s.ReportsPerMinute <= 0 {
+		return nil, fmt.Errorf("REPORTS_PER_MINUTE %q: %w",
+			viper.GetString("REPORTS_PER_MINUTE"), errNotPositive)
+	}
+
+	if s.DataDirMaxBytes <= 0 {
+		return nil, fmt.Errorf("DATA_DIR_MAX_BYTES %q: %w",
+			viper.GetString("DATA_DIR_MAX_BYTES"), errNotPositive)
 	}
 
 	if s.Debug {

backend/internal/config/config_test.go

@@ -0,0 +1,47 @@
+package config_test
+
+import (
+	"strings"
+	"testing"
+
+	"sneak.berlin/go/netwatch/internal/config"
+	"sneak.berlin/go/netwatch/internal/globals"
+	"sneak.berlin/go/netwatch/internal/logger"
+
+	"go.uber.org/fx"
+)
+
+// requireConfigError builds the config as main does and fails the
+// test unless that fails with an error naming setting. It uses
+// fx.New, because fxtest.New fails the test itself on an error.
+func requireConfigError(t *testing.T, setting string) {
+	t.Helper()
+
+	app := fx.New(
+		fx.NopLogger,
+		fx.Provide(globals.New, logger.New, config.New),
+		fx.Invoke(func(*config.Config) {}),
+	)
+
+	err := app.Err()
+	if err == nil || !strings.Contains(err.Error(), setting) {
+		t.Fatalf("config error = %v, want one naming %s", err, setting)
+	}
+}
+
+// TestReportsPerMinuteMustBePositive: unchecked, zero would panic
+// when the routes are built, and a negative rate would lift the
+// limit.
+func TestReportsPerMinuteMustBePositive(t *testing.T) {
+	t.Setenv("REPORTS_PER_MINUTE", "0")
+
+	requireConfigError(t, "REPORTS_PER_MINUTE")
+}
+
+// TestDataDirMaxBytesMustBeANumber: viper reads a value that is not
+// a number, such as "1GB", as 0, which would refuse every report.
+func TestDataDirMaxBytesMustBeANumber(t *testing.T) {
+	t.Setenv("DATA_DIR_MAX_BYTES", "1GB")
+
+	requireConfigError(t, "DATA_DIR_MAX_BYTES")
+}

backend/internal/handlers/report.go

@@ -4,6 +4,8 @@ import (
 	"encoding/json"
 	"errors"
 	"net/http"
+
+	"sneak.berlin/go/netwatch/internal/reportbuf"
 )
 
 // maxLoggedFieldBytes bounds untrusted text (string fields,
@@ -55,10 +57,9 @@ func (s *Handlers) HandleReport() http.HandlerFunc {
 
 		err = s.buf.Append(rpt)
 		if err != nil {
-			s.log.Error("failed to buffer report", "error", err)
 			s.respondJSON(w, r,
 				&response{Status: "error"},
-				http.StatusInternalServerError,
+				s.appendErrorStatus(err),
 			)
 
 			return
@@ -88,6 +89,21 @@ func (s *Handlers) decodeErrorStatus(err error) int {
 	return http.StatusBadRequest
 }
 
+// appendErrorStatus logs a failure to store a report and returns
+// the status to send: 507 when the report files are at their size
+// cap, otherwise 500.
+func (s *Handlers) appendErrorStatus(err error) int {
+	if errors.Is(err, reportbuf.ErrFull) {
+		s.log.Warn("report refused: report files at their size cap")
+
+		return http.StatusInsufficientStorage
+	}
+
+	s.log.Error("failed to buffer report", "error", err)
+
+	return http.StatusInternalServerError
+}
+
 // logReportReceived logs an accepted report. Untrusted fields are
 // bounded (client_id, timestamp) or reduced to a length
 // (geo_bytes) so the raw attacker-controlled body never reaches

backend/internal/handlers/report_test.go

@@ -13,6 +13,7 @@ import (
 
 	"sneak.berlin/go/netwatch/internal/handlers"
 	"sneak.berlin/go/netwatch/internal/middleware"
+	"sneak.berlin/go/netwatch/internal/reportbuf"
 )
 
 var errStorageFailed = errors.New("storage failed")
@@ -66,6 +67,32 @@ func TestHandleReportStorageFailureIsNon2xx(t *testing.T) {
 	}
 }
 
+// TestHandleReportFullIs507 checks the answer when the report files
+// are at their size cap: 507 and the usual error body, which tells
+// the client nothing more.
+func TestHandleReportFullIs507(t *testing.T) {
+	t.Parallel()
+
+	h := newTestHandlers(stubAppender{err: reportbuf.ErrFull}, io.Discard)
+
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequestWithContext(t.Context(),
+		http.MethodPost, "/api/v1/reports",
+		strings.NewReader(`{"clientId":"c1","hosts":[]}`),
+	)
+
+	h.HandleReport().ServeHTTP(rec, req)
+
+	if rec.Code != http.StatusInsufficientStorage {
+		t.Fatalf("status = %d, want %d",
+			rec.Code, http.StatusInsufficientStorage)
+	}
+
+	if got := rec.Body.String(); got != "{\"status\":\"error\"}\n" {
+		t.Errorf("body = %q, want %q", got, "{\"status\":\"error\"}\n")
+	}
+}
+
 func TestHandleReportMalformedJSONIs400(t *testing.T) {
 	t.Parallel()
 

backend/internal/middleware/export_test.go

@@ -15,6 +15,13 @@ func NewWithLogger(log *slog.Logger) *Middleware {
 	return &Middleware{log: log}
 }
 
+// NewWithTrustedProxies builds a Middleware that honours forwarded
+// headers from the given networks, for tests of the client address
+// paths without the fx graph.
+func NewWithTrustedProxies(trusted []netip.Prefix) *Middleware {
+	return &Middleware{trustedProxies: trusted}
+}
+
 func ClientIP(
 	remoteAddr string,
 	header http.Header,

backend/internal/middleware/middleware.go

@@ -20,6 +20,7 @@ import (
 
 	"github.com/go-chi/chi/v5/middleware"
 	"github.com/go-chi/cors"
+	"github.com/go-chi/httprate"
 	"go.uber.org/fx"
 )
 
@@ -320,21 +321,43 @@ func (s *Middleware) Recoverer() func(http.Handler) http.Handler {
 	}
 }
 
-// CORS returns middleware that adds permissive CORS headers.
+// CORS returns middleware that lets pages served from the given
-func (s *Middleware) CORS() func(http.Handler) http.Handler {
+// origins call the API. With no origins it adds no CORS headers at
+// all, so only same-origin pages can use the API. That case must not
+// reach cors.Handler, which treats an empty origin list as "allow
+// every origin".
+func (s *Middleware) CORS(
+	origins []string,
+) func(http.Handler) http.Handler {
+	if len(origins) == 0 {
+		return func(next http.Handler) http.Handler { return next }
+	}
+
 	return cors.Handler(cors.Options{
-		AllowedOrigins: []string{"*"},
+		AllowedOrigins:   origins,
-		AllowedMethods: []string{
+		AllowedMethods:   []string{http.MethodGet, http.MethodPost},
-			"GET", "POST", "PUT", "DELETE", "OPTIONS",
+		AllowedHeaders:   []string{"Content-Type"},
-		},
-		AllowedHeaders: []string{
-			"Accept",
-			"Authorization",
-			"Content-Type",
-			"X-CSRF-Token",
-		},
-		ExposedHeaders:   []string{"Link"},
 		AllowCredentials: false,
 		MaxAge:           corsMaxAgeSec,
 	})
 }
+
+// RateLimit returns middleware that allows each client address
+// perMinute requests a minute and answers the rest with 429, the
+// Retry-After header httprate sets, and the usual error body. The
+// address is the one clientIP resolves, so clients behind the reverse
+// proxy are limited one by one, not together as the proxy.
+func (s *Middleware) RateLimit(
+	perMinute int,
+) func(http.Handler) http.Handler {
+	return httprate.LimitBy(perMinute, time.Minute,
+		func(r *http.Request) (string, error) {
+			return clientIP(r.RemoteAddr, r.Header, s.trustedProxies), nil
+		},
+		httprate.WithLimitHandler(
+			func(w http.ResponseWriter, _ *http.Request) {
+				writeJSONError(w, http.StatusTooManyRequests)
+			},
+		),
+	)
+}

backend/internal/middleware/middleware_test.go

@@ -10,6 +10,8 @@ import (
 	"net/netip"
 	"strings"
 	"testing"
+	"testing/synctest"
+	"time"
 
 	"sneak.berlin/go/netwatch/internal/middleware"
 )
@@ -300,3 +302,170 @@ func TestRecovererRepanicsOnAbortHandler(t *testing.T) {
 		t.Errorf("abort was logged: %q", logbuf.String())
 	}
 }
+
+// okHandler stands in for the route a middleware guards.
+func okHandler() http.Handler {
+	return http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		w.WriteHeader(http.StatusOK)
+	})
+}
+
+// TestRateLimitRefusesPastAllowanceThenResets checks one client
+// address: it may use its whole allowance at once, the next request
+// is refused with 429, and later it may send again.
+func TestRateLimitRefusesPastAllowanceThenResets(t *testing.T) {
+	t.Parallel()
+
+	// synctest runs this on a fake clock: time.Sleep returns at once,
+	// with the clock moved on.
+	synctest.Test(t, func(t *testing.T) {
+		const perMinute = 2
+
+		handler := (&middleware.Middleware{}).RateLimit(perMinute)(okHandler())
+
+		post := func() *httptest.ResponseRecorder {
+			rec := httptest.NewRecorder()
+			req := httptest.NewRequestWithContext(t.Context(),
+				http.MethodPost, "/api/v1/reports", http.NoBody)
+			handler.ServeHTTP(rec, req)
+
+			return rec
+		}
+
+		for i := range perMinute {
+			if code := post().Code; code != http.StatusOK {
+				t.Fatalf("request %d: status = %d, want %d",
+					i+1, code, http.StatusOK)
+			}
+		}
+
+		rec := post()
+		if rec.Code != http.StatusTooManyRequests {
+			t.Fatalf("request past the allowance: status = %d, want %d",
+				rec.Code, http.StatusTooManyRequests)
+		}
+
+		if got := rec.Body.String(); got != "{\"status\":\"error\"}\n" {
+			t.Errorf("body = %q, want %q", got, "{\"status\":\"error\"}\n")
+		}
+
+		if got := rec.Header().Get("Retry-After"); got != "60" {
+			t.Fatalf("Retry-After = %q, want %q", got, "60")
+		}
+
+		// httprate also counts the previous minute's requests, fading
+		// them out over the current one, so two minutes on the whole
+		// allowance is back.
+		time.Sleep(2 * time.Minute)
+
+		for i := range perMinute {
+			if code := post().Code; code != http.StatusOK {
+				t.Fatalf("two minutes later, request %d: status = %d, want %d",
+					i+1, code, http.StatusOK)
+			}
+		}
+	})
+}
+
+// TestRateLimitIsPerForwardedClient checks that clients behind a
+// trusted proxy each get their own allowance: the limit is keyed on
+// the client address clientIP resolves, not on the proxy's.
+func TestRateLimitIsPerForwardedClient(t *testing.T) {
+	t.Parallel()
+
+	const otherClient = "203.0.113.8"
+
+	mw := middleware.NewWithTrustedProxies(mustPrefixes(t, "127.0.0.1/32"))
+	handler := mw.RateLimit(1)(okHandler())
+
+	post := func(client string) int {
+		rec := httptest.NewRecorder()
+		req := httptest.NewRequestWithContext(t.Context(),
+			http.MethodPost, "/api/v1/reports", http.NoBody)
+		req.RemoteAddr = loopbackPeer
+		req.Header.Set("X-Forwarded-For", client)
+		handler.ServeHTTP(rec, req)
+
+		return rec.Code
+	}
+
+	if code := post(forwardedIP); code != http.StatusOK {
+		t.Fatalf("first request: status = %d, want %d", code, http.StatusOK)
+	}
+
+	if code := post(forwardedIP); code != http.StatusTooManyRequests {
+		t.Fatalf("same client again: status = %d, want %d",
+			code, http.StatusTooManyRequests)
+	}
+
+	if code := post(otherClient); code != http.StatusOK {
+		t.Fatalf("other client behind the same proxy: status = %d, want %d",
+			code, http.StatusOK)
+	}
+}
+
+// preflight sends cors the preflight request a browser makes before
+// it POSTs JSON from origin.
+func preflight(
+	t *testing.T,
+	cors func(http.Handler) http.Handler,
+	origin string,
+) *httptest.ResponseRecorder {
+	t.Helper()
+
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequestWithContext(t.Context(),
+		http.MethodOptions, "/api/v1/reports", http.NoBody)
+	req.Header.Set("Origin", origin)
+	req.Header.Set("Access-Control-Request-Method", http.MethodPost)
+	req.Header.Set("Access-Control-Request-Headers", "content-type")
+	cors(okHandler()).ServeHTTP(rec, req)
+
+	return rec
+}
+
+// TestCORSWithoutOriginsAddsNoHeaders checks the default: with no
+// origins configured, no origin is given any CORS header.
+func TestCORSWithoutOriginsAddsNoHeaders(t *testing.T) {
+	t.Parallel()
+
+	rec := preflight(t,
+		(&middleware.Middleware{}).CORS(nil), "https://elsewhere.example")
+
+	for name := range rec.Header() {
+		if strings.HasPrefix(name, "Access-Control-") {
+			t.Errorf("CORS header %s set with no origins configured", name)
+		}
+	}
+}
+
+func TestCORSAllowsOnlyListedOrigins(t *testing.T) {
+	t.Parallel()
+
+	const listed = "https://netwatch.example"
+
+	cors := (&middleware.Middleware{}).CORS([]string{listed})
+
+	cases := []struct {
+		name   string
+		origin string
+		want   string
+	}{
+		{name: "listed origin allowed", origin: listed, want: listed},
+		{name: "other origin refused", origin: "https://elsewhere.example"},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			rec := preflight(t, cors, tc.origin)
+
+			got := rec.Header().Get("Access-Control-Allow-Origin")
+			if got != tc.want {
+				t.Errorf("Access-Control-Allow-Origin = %q, want %q",
+					got, tc.want)
+			}
+		})
+	}
+}

backend/internal/reportbuf/export_test.go

@@ -0,0 +1,7 @@
+package reportbuf
+
+// Flush writes the buffered reports to a file now, as the periodic
+// flush does, so tests need not wait a minute for it.
+func (b *Buffer) Flush() error {
+	return b.flushLocked()
+}

backend/internal/reportbuf/reportbuf.go

@@ -6,11 +6,13 @@ import (
 	"bytes"
 	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"io/fs"
 	"log/slog"
 	"os"
 	"path/filepath"
+	"strings"
 	"sync"
 	"time"
 
@@ -27,8 +29,16 @@ const (
 	defaultDataDir                 = "./data/reports"
 	dirPerms           fs.FileMode = 0o750
 	filePerms          fs.FileMode = 0o640
+
+	// Report files are named filePrefix + timestamp + fileSuffix.
+	filePrefix = "reports-"
+	fileSuffix = ".jsonl.zst"
 )
 
+// ErrFull is returned by Append when storing the report would
+// take the report files past the configured maximum size.
+var ErrFull = errors.New("report files at their size cap")
+
 // Params defines the dependencies for Buffer.
 type Params struct {
 	fx.In
@@ -44,8 +54,13 @@ type Buffer struct {
 	dataDir  string
 	done     chan struct{}
 	log      *slog.Logger
+	maxBytes int64
 	mu       sync.Mutex
 	stopOnce sync.Once
+	// usedBytes is what Append checks against maxBytes: the size
+	// of the report files in dataDir, plus the reports not yet
+	// written to one at their uncompressed size.
+	usedBytes int64
 }
 
 // New creates a Buffer and registers lifecycle hooks to
@@ -60,9 +75,10 @@ func New(
 	}
 
 	b := &Buffer{
-		dataDir: dir,
+		dataDir:  dir,
-		done:    make(chan struct{}),
+		done:     make(chan struct{}),
-		log:     params.Logger.Get(),
+		log:      params.Logger.Get(),
+		maxBytes: params.Config.DataDirMaxBytes,
 	}
 
 	lc.Append(fx.Hook{
@@ -72,6 +88,12 @@ func New(
 				return fmt.Errorf("create data dir: %w", err)
 			}
 
+			// Report files left by earlier runs count too.
+			b.usedBytes, err = reportFilesSize(b.dataDir)
+			if err != nil {
+				return err
+			}
+
 			go b.flushLoop()
 
 			return nil
@@ -97,15 +119,27 @@ func New(
 }
 
 // Append marshals v as a single JSON line and appends it to
-// the buffer. If the buffer reaches the size threshold, it is
+// the buffer. It stores nothing and returns ErrFull if the line
-// drained and written to disk asynchronously.
+// would take usedBytes past maxBytes. If the buffer reaches the
+// size threshold, it is drained and written to disk
+// asynchronously.
 func (b *Buffer) Append(v any) error {
 	line, err := json.Marshal(v)
 	if err != nil {
 		return fmt.Errorf("marshal report: %w", err)
 	}
 
+	lineBytes := int64(len(line)) + 1 // with its newline
+
 	b.mu.Lock()
+
+	if b.usedBytes+lineBytes > b.maxBytes {
+		b.mu.Unlock()
+
+		return ErrFull
+	}
+
+	b.usedBytes += lineBytes
 	b.buf.Write(line)
 	b.buf.WriteByte('\n')
 
@@ -178,8 +212,7 @@ func (b *Buffer) drainBuf() []byte {
 // in the data directory.
 func (b *Buffer) writeFile(data []byte) error {
 	ts := time.Now().UTC().Format("2006-01-02T15-04-05.000Z")
-	name := fmt.Sprintf("reports-%s.jsonl.zst", ts)
+	path := filepath.Join(b.dataDir, filePrefix+ts+fileSuffix)
-	path := filepath.Join(b.dataDir, name)
 
 	// path is built from the operator-supplied dataDir plus a
 	// generated timestamp, so it carries no external input.
@@ -214,10 +247,51 @@ func (b *Buffer) writeFile(data []byte) error {
 		return fmt.Errorf("close zstd encoder: %w", err)
 	}
 
+	info, err := f.Stat()
+	if err != nil {
+		return fmt.Errorf("stat report file: %w", err)
+	}
+
 	err = f.Close()
 	if err != nil {
 		return fmt.Errorf("close report file: %w", err)
 	}
 
+	// The reports counted at their uncompressed size while they
+	// waited; now they count as the file. After a failed write they
+	// stay counted as they were, which errs toward refusing reports
+	// early rather than letting the files pass the cap.
+	b.mu.Lock()
+	b.usedBytes += info.Size() - int64(len(data))
+	b.mu.Unlock()
+
 	return nil
 }
+
+// reportFilesSize returns the total size of the report files in
+// dir.
+func reportFilesSize(dir string) (int64, error) {
+	entries, err := os.ReadDir(dir)
+	if err != nil {
+		return 0, fmt.Errorf("read data dir: %w", err)
+	}
+
+	var total int64
+
+	for _, entry := range entries {
+		name := entry.Name()
+		if !strings.HasPrefix(name, filePrefix) ||
+			!strings.HasSuffix(name, fileSuffix) {
+			continue
+		}
+
+		info, err := entry.Info()
+		if err != nil {
+			return 0, fmt.Errorf("stat report file: %w", err)
+		}
+
+		total += info.Size()
+	}
+
+	return total, nil
+}

backend/internal/reportbuf/reportbuf_test.go

@@ -1,9 +1,12 @@
 package reportbuf_test
 
 import (
+	"encoding/json"
 	"errors"
 	"io/fs"
 	"os"
+	"path/filepath"
+	"strconv"
 	"strings"
 	"testing"
 
@@ -94,6 +97,131 @@ func TestFailedFinalFlushFailsStop(t *testing.T) {
 	}
 }
 
+// startBuffer starts a Buffer through fx, as main does, with the
+// DATA_DIR and DATA_DIR_MAX_BYTES the calling test has set.
+func startBuffer(t *testing.T) *reportbuf.Buffer {
+	t.Helper()
+
+	var buf *reportbuf.Buffer
+
+	app := fxtest.New(t,
+		fx.Provide(
+			globals.New,
+			logger.New,
+			config.New,
+			reportbuf.New,
+		),
+		fx.Populate(&buf),
+	)
+
+	app.RequireStart()
+	t.Cleanup(app.RequireStop)
+
+	return buf
+}
+
+// lineBytes is what one report takes in the buffer: its JSON and a
+// newline.
+func lineBytes(t *testing.T, report any) int {
+	t.Helper()
+
+	line, err := json.Marshal(report)
+	if err != nil {
+		t.Fatalf("marshal report: %v", err)
+	}
+
+	return len(line) + 1
+}
+
+func TestAppendPastCapIsRefused(t *testing.T) {
+	report := map[string]string{"id": "cap"}
+
+	t.Setenv("DATA_DIR", t.TempDir())
+	t.Setenv("DATA_DIR_MAX_BYTES", strconv.Itoa(lineBytes(t, report)))
+
+	buf := startBuffer(t)
+
+	err := buf.Append(report)
+	if err != nil {
+		t.Fatalf("report that fills the cap exactly: %v", err)
+	}
+
+	err = buf.Append(report)
+	if !errors.Is(err, reportbuf.ErrFull) {
+		t.Fatalf("report past the cap: error = %v, want ErrFull", err)
+	}
+}
+
+// TestCapCountsReportFilesAlreadyInDataDir starts on a data
+// directory holding a report file from an earlier run, and a file
+// that is not a report, which must not count.
+func TestCapCountsReportFilesAlreadyInDataDir(t *testing.T) {
+	const earlierBytes = 100
+
+	report := map[string]string{"id": "cap"}
+	dir := t.TempDir()
+
+	writeBytes(t, filepath.Join(dir, "reports-2026-01-01T00-00-00.000Z.jsonl.zst"),
+		earlierBytes)
+	writeBytes(t, filepath.Join(dir, "notes.txt"), 10*earlierBytes)
+
+	t.Setenv("DATA_DIR", dir)
+	t.Setenv("DATA_DIR_MAX_BYTES",
+		strconv.Itoa(earlierBytes+lineBytes(t, report)))
+
+	buf := startBuffer(t)
+
+	err := buf.Append(report)
+	if err != nil {
+		t.Fatalf("report that fills the cap exactly: %v", err)
+	}
+
+	err = buf.Append(report)
+	if !errors.Is(err, reportbuf.ErrFull) {
+		t.Fatalf("report past the cap: error = %v, want ErrFull", err)
+	}
+}
+
+// TestWrittenReportsCountAtFileSize checks that once reports are
+// written, they count as their compressed file, not their
+// uncompressed size, which frees room under the cap.
+func TestWrittenReportsCountAtFileSize(t *testing.T) {
+	// Repetitive, so its file is far smaller than its JSON.
+	report := map[string]string{"id": strings.Repeat("a", 1000)}
+	size := lineBytes(t, report)
+
+	t.Setenv("DATA_DIR", t.TempDir())
+	// Room for the report twice over only if the first one counts
+	// at its file's size by the time the second arrives.
+	t.Setenv("DATA_DIR_MAX_BYTES", strconv.Itoa(2*size-1))
+
+	buf := startBuffer(t)
+
+	err := buf.Append(report)
+	if err != nil {
+		t.Fatalf("first report: %v", err)
+	}
+
+	err = buf.Flush()
+	if err != nil {
+		t.Fatalf("flush: %v", err)
+	}
+
+	err = buf.Append(report)
+	if err != nil {
+		t.Fatalf("second report, after the first was written: %v", err)
+	}
+}
+
+func writeBytes(t *testing.T, path string, n int) {
+	t.Helper()
+
+	err := os.WriteFile(path, make([]byte, n), 0o600)
+	if err != nil {
+		t.Fatalf("write %s: %v", path, err)
+	}
+}
+
 func hasReportFile(t *testing.T, dir string) bool {
 	t.Helper()
 

backend/internal/server/export_test.go

@@ -3,3 +3,9 @@ package server
 // MaxRequestBodyBytes exposes the router-wide body limit to the
 // external tests.
 const MaxRequestBodyBytes = maxRequestBodyBytes
+
+// ListenAddr exposes the address the server listens on to the
+// external tests.
+func (s *Server) ListenAddr() string {
+	return s.newHTTPServer().Addr
+}

backend/internal/server/http.go

@@ -2,8 +2,9 @@ package server
 
 import (
 	"errors"
-	"fmt"
+	"net"
 	"net/http"
+	"strconv"
 	"time"
 
 	"go.uber.org/fx"
@@ -27,7 +28,10 @@ const (
 // newHTTPServer constructs the http.Server. It performs no I/O
 // and does not start listening.
 func (s *Server) newHTTPServer() *http.Server {
-	listenAddr := fmt.Sprintf(":%d", s.params.Config.Port)
+	listenAddr := net.JoinHostPort(
+		s.params.Config.BindAddress,
+		strconv.Itoa(s.params.Config.Port),
+	)
 
 	return &http.Server{
 		Addr:              listenAddr,

backend/internal/server/http_test.go

@@ -0,0 +1,32 @@
+package server_test
+
+import "testing"
+
+// TestListenAddress checks that the server listens on BIND_ADDRESS
+// and PORT, and on port 8080 on every interface when neither is set.
+// The container image sets both, to keep the backend on loopback
+// behind nginx.
+func TestListenAddress(t *testing.T) {
+	tests := []struct {
+		bindAddress string
+		port        string
+		want        string
+	}{
+		{bindAddress: "", port: "", want: ":8080"},
+		{bindAddress: "127.0.0.1", port: "8081", want: "127.0.0.1:8081"},
+		{bindAddress: "::1", port: "8081", want: "[::1]:8081"},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.want, func(t *testing.T) {
+			// t.Setenv rules out t.Parallel.
+			t.Setenv("BIND_ADDRESS", tt.bindAddress)
+			t.Setenv("PORT", tt.port)
+
+			got := newServer(t).ListenAddr()
+			if got != tt.want {
+				t.Errorf("listen address = %q, want %q", got, tt.want)
+			}
+		})
+	}
+}

backend/internal/server/routes.go

@@ -25,7 +25,7 @@ func (s *Server) SetupRoutes() {
 	s.router.Use(middleware.RequestID)
 	s.router.Use(s.mw.Logging())
 	s.router.Use(s.mw.SecurityHeaders())
-	s.router.Use(s.mw.CORS())
+	s.router.Use(s.mw.CORS(s.params.Config.CORSAllowedOrigins))
 	s.router.Use(s.mw.MaxBodyBytes(maxRequestBodyBytes))
 	s.router.Use(middleware.Timeout(requestTimeout))
 
@@ -35,6 +35,7 @@ func (s *Server) SetupRoutes() {
 	)
 
 	s.router.Route("/api/v1", func(r chi.Router) {
-		r.Post("/reports", s.h.HandleReport())
+		r.With(s.mw.RateLimit(s.params.Config.ReportsPerMinute)).
+			Post("/reports", s.h.HandleReport())
 	})
 }

backend/internal/server/routes_test.go

@@ -19,16 +19,14 @@ import (
 	"go.uber.org/fx/fxtest"
 )
 
-// TestHealthCheckRejectsOversizeBody sends the health check, which
+// newServer builds a Server from the same constructors as main,
-// never reads its body, a body one byte over the limit. Only the
+// configured from the environment. It is never started, so nothing
-// router-wide body limit can reject it.
+// listens.
-func TestHealthCheckRejectsOversizeBody(t *testing.T) {
+func newServer(t *testing.T) *server.Server {
-	t.Parallel()
+	t.Helper()
 
 	var srv *server.Server
 
-	// The same constructors as main, never started: SetupRoutes is
-	// called directly, so nothing listens.
 	app := fxtest.New(t,
 		fx.Provide(
 			config.New,
@@ -48,6 +46,48 @@ func TestHealthCheckRejectsOversizeBody(t *testing.T) {
 		t.Fatalf("build server: %v", err)
 	}
 
+	return srv
+}
+
+// TestReportsAreRateLimited checks that POST /api/v1/reports is
+// behind the per-address rate limit, set here to two a minute.
+func TestReportsAreRateLimited(t *testing.T) {
+	t.Setenv("REPORTS_PER_MINUTE", "2")
+
+	srv := newServer(t)
+	srv.SetupRoutes()
+
+	post := func() int {
+		rec := httptest.NewRecorder()
+		req := httptest.NewRequestWithContext(t.Context(),
+			http.MethodPost, "/api/v1/reports",
+			strings.NewReader(`{"clientId":"c1","hosts":[]}`),
+		)
+		srv.ServeHTTP(rec, req)
+
+		return rec.Code
+	}
+
+	for i := range 2 {
+		if code := post(); code != http.StatusOK {
+			t.Fatalf("report %d: status = %d, want %d",
+				i+1, code, http.StatusOK)
+		}
+	}
+
+	if code := post(); code != http.StatusTooManyRequests {
+		t.Fatalf("third report in a minute: status = %d, want %d",
+			code, http.StatusTooManyRequests)
+	}
+}
+
+// TestHealthCheckRejectsOversizeBody sends the health check, which
+// never reads its body, a body one byte over the limit. Only the
+// router-wide body limit can reject it.
+func TestHealthCheckRejectsOversizeBody(t *testing.T) {
+	t.Parallel()
+
+	srv := newServer(t)
 	srv.SetupRoutes()
 
 	rec := httptest.NewRecorder()

backend/script/build

@@ -8,7 +8,7 @@ ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
 main() {
     cd "$ROOT"
 
-    # VERSION comes from the environment (Dockerfile.backend passes its
+    # VERSION comes from the environment (the root Dockerfile passes its
     # ARG VERSION in). Unset or empty, it is git describe, or "dev" where
     # there is no git or no repository history.
     version="${VERSION:-$(git describe --always --dirty 2>/dev/null || echo dev)}"

backend/script/lint

@@ -1,6 +1,6 @@
 #!/bin/sh
 # script/lint: run golangci-lint over the backend. This runs inside the
-# lint stage of Dockerfile.backend, whose digest-pinned golangci-lint
+# lint stage of the root Dockerfile, whose digest-pinned golangci-lint
 # image provides the linter; nothing installs golangci-lint on the host.
 # From a checkout, run `make lint` at the repo root, which builds that
 # stage.

bin/entrypoint.sh

@@ -0,0 +1,59 @@
+#!/bin/sh
+# The container's entrypoint: runs netwatch-server and nginx side by
+# side. TERM or INT stops both, and the container exits 0 if both exit
+# cleanly. If either exits on its own, the other is stopped too and the
+# container exits non-zero, so the platform restarts it instead of
+# leaving it half up.
+#
+# No set -e: kill and wait return non-zero here in normal operation.
+set -u
+
+# A stop signal is only noted here; the loop below acts on it.
+stop_requested=""
+trap 'stop_requested=yes' TERM INT
+
+# netwatch-server runs as the netwatch user and listens on loopback
+# only, on a port other than the public one; nginx.conf proxies to this
+# address. The netwatch user has no login shell, hence -s /bin/sh.
+# busybox su replaces itself with the command instead of staying on as
+# its parent, so $! is the server's own PID.
+BIND_ADDRESS=127.0.0.1 PORT=8081 \
+    su -s /bin/sh netwatch -c 'exec netwatch-server' &
+backend=$!
+
+# nginx starts through the nginx image's own entrypoint, which applies
+# the image's start-up configuration and then replaces itself with
+# nginx.
+/docker-entrypoint.sh nginx -g 'daemon off;' &
+nginx=$!
+
+running() {
+    kill -0 "$1" 2>/dev/null
+}
+
+# POSIX sh cannot wait for whichever of two children exits first, so
+# look once a second. The shell collects a child that has exited while
+# it runs sleep, and running() is false for that child from then on.
+while [ -z "$stop_requested" ] && running "$backend" && running "$nginx"; do
+    sleep 1
+done
+
+# Stop both, then wait until neither is left.
+kill -TERM "$backend" "$nginx" 2>/dev/null
+while running "$backend" || running "$nginx"; do
+    sleep 1
+done
+
+wait "$backend"
+backend_status=$?
+wait "$nginx"
+nginx_status=$?
+echo "entrypoint: netwatch-server exited $backend_status," \
+    "nginx exited $nginx_status"
+
+# Success is a requested stop that both processes exited cleanly from.
+if [ -n "$stop_requested" ] && [ "$backend_status" -eq 0 ] &&
+    [ "$nginx_status" -eq 0 ]; then
+    exit 0
+fi
+exit 1

nginx.conf

@@ -25,4 +25,23 @@ server {
         expires 1y;
         add_header Cache-Control "public, immutable";
     }
+
+    # netwatch-server, the Go backend, runs in the same container and
+    # listens on loopback only: bin/entrypoint.sh starts it on
+    # 127.0.0.1:8081. These headers go with every request passed to it.
+    # X-Forwarded-For carries only the client address, as resolved by
+    # the real IP settings above, and not the chain the request came
+    # with: the backend takes the first entry, which a client can write.
+    proxy_set_header Host $host;
+    proxy_set_header X-Real-IP $remote_addr;
+    proxy_set_header X-Forwarded-For $remote_addr;
+    proxy_set_header X-Forwarded-Proto $scheme;
+
+    location /api/ {
+        proxy_pass http://127.0.0.1:8081;
+    }
+
+    location = /.well-known/healthcheck {
+        proxy_pass http://127.0.0.1:8081;
+    }
 }

script/bootstrap

@@ -27,8 +27,8 @@ NVM_VERSION="0.40.3"
 # sha256 of https://github.com/nvm-sh/nvm/archive/refs/tags/v0.40.3.tar.gz
 NVM_SHA256="5f4d6aaa04a177dc93c985e31dbc411ab6b8c6e1e21d8015dbc1372625fcd1d0"
 YARN_VERSION="1.22.22"
-# The Go inside the golang:1.25-alpine image Dockerfile.backend builds
+# The Go inside the golang:1.25-alpine image Dockerfile builds the
-# with, 2026-08-09. The archive hashes are in ensure_go.
+# backend with, 2026-08-09. The archive hashes are in ensure_go.
 GO_VERSION="1.25.7"
 
 BIN_DIR="$HOME/.local/bin"

script/cibuild

@@ -1,9 +1,8 @@
 #!/bin/sh
-# script/cibuild: run the CI build. It builds both images: the frontend
+# script/cibuild: run the CI build: build the one image from Dockerfile,
-# from Dockerfile and the backend from Dockerfile.backend. Each runs its
+# whose stages run the checks as build steps (the backend's fmt-check,
-# half of the checks as build steps, so a successful cibuild implies the
+# lint and tests, and the frontend's test, lint and fmt-check). This is
-# whole repo is green. This is the only build step the Gitea workflow
+# the only build step the Gitea workflow runs.
-# runs.
 set -eu
 
 ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
@@ -11,7 +10,6 @@ ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
 main() {
     cd "$ROOT"
     timeout 300 docker build .
-    timeout 300 docker build -f Dockerfile.backend .
 }
 
 main "$@"

script/docker

@@ -1,7 +1,8 @@
 #!/bin/sh
-# script/docker: build both Docker images, tagged with the project name
+# script/docker: build the Docker image tagged with the project name.
-# from script/projectname: the frontend as <name>, from Dockerfile, and
+# Identical in all repos; the tag comes from script/projectname.
-# the backend as <name>-server, from Dockerfile.backend.
+# --no-cache because the gate phases the final stage depends on are RUN
+# steps, and a cached one is a check that did not run.
 set -eu
 
 SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
@@ -9,9 +10,16 @@ ROOT="$(cd "$SCRIPT_DIR/.." && pwd -P)"
 
 main() {
     cd "$ROOT"
-    name="$("$SCRIPT_DIR/projectname")"
+    # Own line: a failing command substitution inside an argument does
-    timeout 300 docker build -t "$name" .
+    # not trip `set -e`, so the inline form degrades silently to an
-    timeout 300 docker build -t "$name-server" -f Dockerfile.backend .
+    # empty constant. VERSION is computed here because .dockerignore
+    # excludes .git, so `git describe` in a build stage yields an empty
+    # version without failing.
+    version="$(git describe --tags --always --dirty 2>/dev/null || true)"
+    [ -n "$version" ] || version="unknown"
+    docker build --no-cache \
+        --build-arg VERSION="$version" \
+        -t "$("$SCRIPT_DIR/projectname")" .
 }
 
 main "$@"

script/frontend-check

@@ -1,9 +1,9 @@
 #!/bin/sh
 # script/frontend-check: run the frontend half of the checks only (test,
-# lint, fmt-check). This exists for the frontend Dockerfile, whose build
+# lint, fmt-check). This exists for the frontend stage of Dockerfile, a
-# stage is a node image with neither Go nor Docker; the backend half is
+# node image with neither Go nor Docker; the Dockerfile's lint and
-# gated by Dockerfile.backend. Everywhere else, use script/check, which
+# backend build stages gate the backend half. Everywhere else, use
-# covers the whole repo. Must not modify any files.
+# script/check, which covers the whole repo. Must not modify any files.
 set -eu
 
 ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"

script/lint

@@ -3,8 +3,8 @@
 # Go linter over backend/.
 #
 # The Go linter runs only in Docker: this builds the lint stage of
-# Dockerfile.backend, the digest-pinned golangci-lint image, which runs
+# Dockerfile, the digest-pinned golangci-lint image, which runs the
-# the backend's fmt-check and lint targets. --no-cache makes the linter
+# backend's fmt-check and lint targets. --no-cache makes the linter
 # really run every time rather than reuse an earlier result, and the
 # stage is built for its checks alone, so no image is kept.
 set -eu
@@ -15,7 +15,7 @@ main() {
     cd "$ROOT"
     "$ROOT/script/frontend-lint"
     timeout 300 docker build --no-cache --target lint \
-        --output type=cacheonly -f Dockerfile.backend .
+        --output type=cacheonly .
 }
 
 main "$@"