add boto3 test for STREAMING-UNSIGNED-PAYLOAD-TRAILER

Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/820

.woodpecker/debug.yaml

@@ -1,6 +1,3 @@
-labels:
-  nix: "enabled"
-
 when:
   event:
   - push
@@ -12,32 +9,27 @@ when:
 
 steps:
   - name: check formatting
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     commands:
-      - nix-build -j4 --attr flakePackages.fmt
+      - nix-shell --attr devShell --run "cargo fmt -- --check"
 
   - name: build
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     commands:
       - nix-build -j4 --attr flakePackages.dev
 
   - name: unit + func tests (lmdb)
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     commands:
       - nix-build -j4 --attr flakePackages.tests-lmdb
 
   - name: unit + func tests (sqlite)
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     commands:
       - nix-build -j4 --attr flakePackages.tests-sqlite
 
-  - name: unit + func tests (fjall)
-    image: nixpkgs/nix:nixos-24.05
-    commands:
-      - nix-build -j4 --attr flakePackages.tests-fjall
-
   - name: integration tests
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     commands:
       - nix-build -j4 --attr flakePackages.dev
       - nix-shell --attr ci --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)

.woodpecker/publish.yaml

@@ -1,6 +1,3 @@
-labels:
-  nix: "enabled"
-
 when:
   event:
   - deployment
@@ -11,7 +8,7 @@ depends_on:
 
 steps:
   - name: refresh-index
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     environment:
       AWS_ACCESS_KEY_ID:
         from_secret: garagehq_aws_access_key_id
@@ -22,7 +19,7 @@ steps:
       - nix-shell --attr ci --run "refresh_index"
 
   - name: multiarch-docker
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     environment:
       DOCKER_AUTH:
         from_secret: docker_auth

.woodpecker/release.yaml

@@ -1,6 +1,3 @@
-labels:
-  nix: "enabled"
-
 when:
   event:
   - deployment
@@ -19,17 +16,17 @@ matrix:
 
 steps:
   - name: build
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     commands:
       - nix-build --attr releasePackages.${ARCH} --argstr git_version ${CI_COMMIT_TAG:-$CI_COMMIT_SHA}
 
   - name: check is static binary
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     commands:
       - nix-shell --attr ci --run "./script/not-dynamic.sh result/bin/garage"
 
   - name: integration tests
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     commands:
       - nix-shell --attr ci --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
     when:
@@ -38,8 +35,16 @@ steps:
       - matrix:
           ARCH: i386
 
-  - name: upgrade tests
-    image: nixpkgs/nix:nixos-24.05
+  - name: upgrade tests from v1.0.0
+    image: nixpkgs/nix:nixos-22.05
+    commands:
+      - nix-shell --attr ci --run "./script/test-upgrade.sh v1.0.0 x86_64-unknown-linux-musl" || (cat /tmp/garage.log; false)
+    when:
+      - matrix:
+          ARCH: amd64
+
+  - name: upgrade tests from v0.8.4
+    image: nixpkgs/nix:nixos-22.05
     commands:
       - nix-shell --attr ci --run "./script/test-upgrade.sh v0.8.4 x86_64-unknown-linux-musl" || (cat /tmp/garage.log; false)
     when:
@@ -47,7 +52,7 @@ steps:
           ARCH: amd64
 
   - name: push static binary
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     environment:
       TARGET: "${TARGET}"
       AWS_ACCESS_KEY_ID:
@@ -58,7 +63,7 @@ steps:
       - nix-shell --attr ci --run "to_s3"
 
   - name: docker build and publish
-    image: nixpkgs/nix:nixos-24.05
+    image: nixpkgs/nix:nixos-22.05
     environment:
       DOCKER_PLATFORM: "linux/${ARCH}"
       CONTAINER_NAME: "dxflrs/${ARCH}_garage"

Cargo.toml

@@ -24,18 +24,18 @@ default-members = ["src/garage"]
 
 # Internal Garage crates
 format_table = { version = "0.1.1", path = "src/format-table" }
-garage_api_common = { version = "1.3.1", path = "src/api/common" }
-garage_api_admin = { version = "1.3.1", path = "src/api/admin" }
-garage_api_s3 = { version = "1.3.1", path = "src/api/s3" }
-garage_api_k2v = { version = "1.3.1", path = "src/api/k2v" }
-garage_block = { version = "1.3.1", path = "src/block" }
-garage_db = { version = "1.3.1", path = "src/db", default-features = false }
-garage_model = { version = "1.3.1", path = "src/model", default-features = false }
-garage_net = { version = "1.3.1", path = "src/net" }
-garage_rpc = { version = "1.3.1", path = "src/rpc" }
-garage_table = { version = "1.3.1", path = "src/table" }
-garage_util = { version = "1.3.1", path = "src/util" }
-garage_web = { version = "1.3.1", path = "src/web" }
+garage_api_common = { version = "2.0.0", path = "src/api/common" }
+garage_api_admin = { version = "2.0.0", path = "src/api/admin" }
+garage_api_s3 = { version = "2.0.0", path = "src/api/s3" }
+garage_api_k2v = { version = "2.0.0", path = "src/api/k2v" }
+garage_block = { version = "2.0.0", path = "src/block" }
+garage_db = { version = "2.0.0", path = "src/db", default-features = false }
+garage_model = { version = "2.0.0", path = "src/model", default-features = false }
+garage_net = { version = "2.0.0", path = "src/net" }
+garage_rpc = { version = "2.0.0", path = "src/rpc" }
+garage_table = { version = "2.0.0", path = "src/table" }
+garage_util = { version = "2.0.0", path = "src/util" }
+garage_web = { version = "2.0.0", path = "src/web" }
 k2v-client = { version = "0.0.4", path = "src/k2v-client" }
 
 # External crates from crates.io
@@ -48,15 +48,18 @@ blake2 = "0.10"
 bytes = "1.0"
 bytesize = "1.1"
 cfg-if = "1.0"
-chrono = "0.4"
+chrono = { version = "0.4", features = ["serde"] }
 crc32fast = "1.4"
 crc32c = "0.6"
+crc64fast-nvme = "1.2"
 crypto-common = "0.1"
+err-derive = "0.3"
 gethostname = "0.4"
 git-version = "0.3.4"
 hex = "0.4"
 hexdump = "0.1"
 hmac = "0.12"
+idna = "0.5"
 itertools = "0.12"
 ipnet = "2.9.0"
 lazy_static = "1.4"
@@ -64,8 +67,8 @@ md-5 = "0.10"
 mktemp = "0.5"
 nix = { version = "0.29", default-features = false, features = ["fs"] }
 nom = "7.1"
-parking_lot = "0.12"
 parse_duration = "2.1"
+paste = "1.0"
 pin-project = "1.0.12"
 pnet_datalink = "0.34"
 rand = "0.8"
@@ -83,14 +86,12 @@ pretty_env_logger = "0.5"
 structopt = { version = "0.3", default-features = false }
 syslog-tracing = "0.3"
 tracing = "0.1"
-tracing-journald = "0.3.1"
 tracing-subscriber = { version = "0.3", features = ["env-filter"] }
 
 heed = { version = "0.11", default-features = false, features = ["lmdb"] }
-rusqlite = "0.37"
+rusqlite = "0.31.0"
 r2d2 = "0.8"
-r2d2_sqlite = "0.31"
-fjall = "2.4"
+r2d2_sqlite = "0.24"
 
 async-compression = { version = "0.4", features = ["tokio", "zstd"] }
 zstd = { version = "0.13", default-features = false }
@@ -101,6 +102,7 @@ serde = { version = "1.0", default-features = false, features = ["derive", "rc"]
 serde_bytes = "0.11"
 serde_json = "1.0"
 toml = { version = "0.8", default-features = false, features = ["parse"] }
+utoipa = { version = "5.3.1", features = ["chrono"] }
 
 # newer version requires rust edition 2021
 k8s-openapi = { version = "0.21", features = ["v1_24"] }
@@ -136,7 +138,7 @@ prometheus = "0.13"
 aws-sigv4 = { version = "1.1", default-features = false }
 hyper-rustls = { version = "0.26", default-features = false, features = ["http1", "http2", "ring", "rustls-native-certs"] }
 log = "0.4"
-thiserror = "2.0"
+thiserror = "1.0"
 
 # ---- used only as build / dev dependencies ----
 assert-json-diff = "2.0"
@@ -146,8 +148,12 @@ aws-smithy-runtime = { version = "1.8", default-features = false, features = ["t
 aws-sdk-config = { version = "1.62", default-features = false }
 aws-sdk-s3 = { version = "1.79", default-features = false, features = ["rt-tokio"] }
 
+[profile.dev]
+#lto = "thin"	# disabled for now, adds 2-4 min to each CI build
+lto = "off"
+
 [profile.release]
-lto = "thin"
-codegen-units = 16
-opt-level = 3
-strip = "debuginfo"
+lto = true
+codegen-units = 1
+opt-level = "s"
+strip = true

doc/api/garage-admin-v0.html

@@ -1,7 +1,7 @@
 <!DOCTYPE html>
 <html>
   <head>
-    <title>Garage Adminstration API v0</title>
+    <title>Garage adminstration API v0</title>
     <!-- needed for adaptive design -->
     <meta charset="utf-8"/>
     <meta name="viewport" content="width=device-width, initial-scale=1">

doc/api/garage-admin-v1.html

@@ -1,7 +1,7 @@
 <!DOCTYPE html>
 <html>
   <head>
-    <title>Garage Adminstration API v0</title>
+    <title>Garage adminstration API v1</title>
     <!-- needed for adaptive design -->
     <meta charset="utf-8"/>
     <meta name="viewport" content="width=device-width, initial-scale=1">

doc/api/garage-admin-v2.html

@@ -0,0 +1,24 @@
+<!DOCTYPE html>
+<html>
+  <head>
+    <title>Garage adminstration API v2</title>
+    <!-- needed for adaptive design -->
+    <meta charset="utf-8"/>
+    <meta name="viewport" content="width=device-width, initial-scale=1">
+    <link href="./css/redoc.css" rel="stylesheet">
+
+    <!--
+    Redoc doesn't change outer page styles
+    -->
+    <style>
+      body {
+        margin: 0;
+        padding: 0;
+      }
+    </style>
+  </head>
+  <body>
+    <redoc spec-url='./garage-admin-v2.json'></redoc>
+    <script src="./redoc.standalone.js"> </script>
+  </body>
+</html>

doc/book/connect/apps/index.md

@@ -12,7 +12,7 @@ In this section, we cover the following web applications:
 | [Mastodon](#mastodon)     | ✅       | Natively supported    |
 | [Matrix](#matrix)     | ✅       |  Tested with `synapse-s3-storage-provider`    |
 | [ejabberd](#ejabberd)     | ✅       |  `mod_s3_upload`    |
-| [Pixelfed](#pixelfed)     | ✅       |  Natively supported    |
+| [Pixelfed](#pixelfed)     | ❓       |  Not yet tested    |
 | [Pleroma](#pleroma)     | ❓       |  Not yet tested    |
 | [Lemmy](#lemmy)     | ✅        |  Supported with pict-rs    |
 | [Funkwhale](#funkwhale)     | ❓       | Not yet tested     |
@@ -69,7 +69,7 @@ $CONFIG = array(
         'hostname' => '127.0.0.1', // Can also be a domain name, eg. garage.example.com
         'port' => 3900,            // Put your reverse proxy port or your S3 API port
         'use_ssl' => false,        // Set it to true if you have a TLS enabled reverse proxy
-        'region' => 'garage',      // Garage default region is named "garage", edit according to your cluster config
+        'region' => 'garage',      // Garage has only one region named "garage"
         'use_path_style' => true   // Garage supports only path style, must be set to true
     ],
 ],
@@ -135,7 +135,7 @@ bucket but doesn't also know the secret encryption key.
 *Click on the picture to zoom*
 
 Add a new external storage. Put what you want in "folder name" (eg. "shared"). Select "Amazon S3". Keep "Access Key" for the Authentication field.
-In Configuration, put your bucket name (eg. nextcloud), the host (eg. 127.0.0.1), the port (eg. 3900 or 443), the region ("garage" if you use the default, or the one your configured in your `garage.toml`). Tick the SSL box if you have put an HTTPS proxy in front of garage. You must tick the "Path access" box and you must leave the "Legacy authentication (v2)" box empty. Put your Key ID (eg. GK...) and your Secret Key in the last two input boxes. Finally click on the tick symbol on the right of your screen.
+In Configuration, put your bucket name (eg. nextcloud), the host (eg. 127.0.0.1), the port (eg. 3900 or 443), the region (garage). Tick the SSL box if you have put an HTTPS proxy in front of garage. You must tick the "Path access" box and you must leave the "Legacy authentication (v2)" box empty. Put your Key ID (eg. GK...) and your Secret Key in the last two input boxes. Finally click on the tick symbol on the right of your screen.
 
 Now go to your "Files" app and a new "linked folder" has appeared with the name you chose earlier (eg. "shared").
 
@@ -191,10 +191,10 @@ garage key create peertube-key
 
 Keep the Key ID and the Secret key in a pad, they will be needed later.  
 
-We need two buckets, one for normal videos (named peertube-videos) and one for webtorrent videos (named peertube-playlists).
+We need two buckets, one for normal videos (named peertube-video) and one for webtorrent videos (named peertube-playlist).
 ```bash
 garage bucket create peertube-videos
-garage bucket create peertube-playlists
+garage bucket create peertube-playlist
 ```
 
 Now we allow our key to read and write on these buckets:
@@ -238,7 +238,7 @@ object_storage:
   # Put localhost only if you have a garage instance running on that node
   endpoint: 'http://localhost:3900' # or "garage.example.com" if you have TLS on port 443
 
-  # Garage default region is named "garage", edit according to your config
+  # Garage supports only one region for now, named garage
   region: 'garage'
 
   credentials:
@@ -253,7 +253,7 @@ object_storage:
     proxify_private_files: false
 
   streaming_playlists:
-    bucket_name: 'peertube-playlists'
+    bucket_name: 'peertube-playlist'
 
     # Keep it empty for our example
     prefix: ''
@@ -441,7 +441,7 @@ media_storage_providers:
   store_synchronous: True  # do we want to wait that the file has been written before returning?
   config:
     bucket: matrix       # the name of our bucket, we chose matrix earlier
-    region_name: garage  # "garage" by default, edit according to your cluster config
+    region_name: garage  # only "garage" is supported for the region field
     endpoint_url: http://localhost:3900 # the path to the S3 endpoint
     access_key_id: "GKxxx" # your Key ID
     secret_access_key: "xxxx" # your Secret Key

doc/book/connect/backup.md

@@ -161,49 +161,3 @@ kopia repository validate-provider
 
 You can then run all the standard kopia commands: `kopia snapshot create`, `kopia mount`...
 Everything should work out-of-the-box.
-
-## Plakar
-
-Create your key and bucket on Garage server:
-
-```bash
-garage key create my-plakar-key
-garage bucket create plakar-backups
-garage bucket allow plakar-backups --read --write --key my-plakar-key
-…
-```
-
-On Plakar server, add your Garage as a storage location:
-```bash
-plakar store add garageS3 s3://my-garage.tld/plakar-backups \
-region=garage # Or as you've specified in garage.toml \
-access_key=<Key ID from "garage key info my-plakar-key"> \
-secret_access_key=<Secret key from "garage key info my-plakar-key">
-```
-
-Then create the repository.
-```bash
-plakar at @garageS3 create -plaintext # Unencrypted
-# or
-plakar at @garageS3 create #encrypted
-```
-
-If you encrypt your backups (Plakar default), you will need to define a strong passphrase. Do not forget to save your password safely. It will be needed to decrypt your backups.
-
-
-After the repository has been created, check that everything works as expected (that might give an empty result as no file has been added yet, but no error message):
-```bash
-plakar at @garageS3 check
-```
-
-Now that everything is configure, you can use Garage as your backups storage. For instance sync it with a local backup storage:
-```bash
-$ plakar at ~/backups sync to @garageS3
-```
-
-Or list the S3 storage content:
-```bash
-$ plakar at @garageS3 ls
-```
-
-More information in Plakar documentation: https://www.plakar.io/docs/main/quickstart/

doc/book/cookbook/ansible.md

@@ -8,18 +8,18 @@ have published Ansible roles.  We list them and compare them below.
 
 ## Comparison of Ansible roles
 
-| Feature                            | [ansible-role-garage](#zorun-ansible-role-garage) | [garage-docker-ansible-deploy](#moan0s-garage-docker-ansible-deploy) | [eddster ansible-role-garage](#eddster-ansible-role-garage) |
-|------------------------------------|---------------------------------------------|---------------------------------------------------------------|---------------------------------|
-| **Runtime**                        | Systemd                                     | Docker                                                        | Systemd                         |
-| **Target OS**                      | Any Linux                                   | Any Linux                                                     | Any Linux                       |
-| **Architecture**                   | amd64, arm64, i686                          | amd64, arm64                                                  | arm64, arm, 386, amd64                            |
-| **Additional software**            | None                                        | Traefik                                                       | Ngnix and Keepalived (optional) |
-| **Automatic node connection**      | ❌                                          | ✅                                                            | ✅                              |
-| **Layout management**              | ❌                                          | ✅                                                            | ✅                              |
-| **Manage buckets & keys**          | ❌                                          | ✅ (basic)                                                    | ✅                              |
-| **Allow custom Garage config**     | ✅                                          | ❌                                                            | ❌                              |
-| **Facilitate Garage upgrades**     | ✅                                          | ❌                                                            | ✅                              |
-| **Multiple instances on one host** | ✅                                          | ✅                                                            | ❌                              |
+| Feature                            | [ansible-role-garage](#zorun-ansible-role-garage) | [garage-docker-ansible-deploy](#moan0s-garage-docker-ansible-deploy) |
+|------------------------------------|---------------------------------------------|---------------------------------------------------------------|
+| **Runtime**                        | Systemd                                     | Docker                                                        |
+| **Target OS**                      | Any Linux                                   | Any Linux                                                     |
+| **Architecture**                   | amd64, arm64, i686                          | amd64, arm64                                                  |
+| **Additional software**            | None                                        | Traefik                                                       |
+| **Automatic node connection**      | ❌                                          | ✅                                                            |
+| **Layout management**              | ❌                                          | ✅                                                            |
+| **Manage buckets & keys**          | ❌                                          | ✅ (basic)                                                    |
+| **Allow custom Garage config**     | ✅                                          | ❌                                                            |
+| **Facilitate Garage upgrades**     | ✅                                          | ❌                                                            |
+| **Multiple instances on one host** | ✅                                          | ✅                                                            |
 
 
 ## zorun/ansible-role-garage
@@ -49,15 +49,3 @@ structured DNS names, etc).
 
 As a result, this role makes it easier to start with Garage on Ansible,
 but is less flexible.
-
-## eddster2309/ansible-role-garage
-
-[Source code](https://github.com/eddster2309/ansible-role-garage), [Ansible galaxy](https://galaxy.ansible.com/ui/standalone/roles/eddster2309/garage/)
-
-This role is a opinionated but customisable role using the official Garage
-static binaries and only requires Systemd. As such it should work on any
-Linux based host. It includes all the nesscary configuration to
-automatically setup a clustered Garage deployment. Most Garage
-configuration options are exposed through Ansible variables so while you
-can't provide a custom config you can get very close. It can optionally
-installed a HA nginx deployment with Keepalived.

doc/book/cookbook/binary-packages.md

@@ -15,10 +15,9 @@ Alpine Linux repositories (available since v3.17):
 apk add garage
 ```
 
-The default configuration file is installed to `/etc/garage/garage.toml`. You can run
-Garage using: `rc-service garage start`.
-
-If you don't specify `rpc_secret`, it will be automatically replaced with a random string on the first start.
+The default configuration file is installed to `/etc/garage.toml`. You can run
+Garage using: `rc-service garage start`. If you don't specify `rpc_secret`, it
+will be automatically replaced with a random string on the first start.
 
 Please note that this package is built without Consul discovery, Kubernetes
 discovery, OpenTelemetry exporter, and K2V features (K2V will be enabled once
@@ -27,7 +26,7 @@ it's stable).
 
 ## Arch Linux
 
-Garage is available in the official repositories under [extra](https://archlinux.org/packages/extra/x86_64/garage).
+Garage is available in the [AUR](https://aur.archlinux.org/packages/garage).
 
 ## FreeBSD
 

doc/book/cookbook/kubernetes.md

@@ -11,7 +11,7 @@ Firstly clone the repository:
 
 ```bash
 git clone https://git.deuxfleurs.fr/Deuxfleurs/garage
-cd garage/script/helm
+cd garage/scripts/helm
 ```
 
 Deploy with default options:
@@ -26,13 +26,6 @@ Or deploy with custom values:
 helm install --create-namespace --namespace garage garage ./garage -f values.override.yaml
 ```
 
-If you want to manage the CustomRessourceDefinition used by garage for its `kubernetes_discovery` outside of the helm chart, add `garage.kubernetesSkipCrd: true` to your custom values and use the kustomization before deploying the helm chart:
-
-```bash
-kubectl apply -k ../k8s/crd
-helm install --create-namespace --namespace garage garage ./garage -f values.override.yaml
-```
-
 After deploying, cluster layout must be configured manually as described in [Creating a cluster layout](@/documentation/quick-start/_index.md#creating-a-cluster-layout). Use the following command to access garage CLI:
 
 ```bash

doc/book/cookbook/real-world.md

@@ -96,14 +96,14 @@ to store 2 TB of data in total.
 ## Get a Docker image
 
 Our docker image is currently named `dxflrs/garage` and is stored on the [Docker Hub](https://hub.docker.com/r/dxflrs/garage/tags?page=1&ordering=last_updated).
-We encourage you to use a fixed tag (eg. `v1.3.0`) and not the `latest` tag.
-For this example, we will use the latest published version at the time of the writing which is `v1.3.0` but it's up to you
+We encourage you to use a fixed tag (eg. `v2.0.0`) and not the `latest` tag.
+For this example, we will use the latest published version at the time of the writing which is `v2.0.0` but it's up to you
 to check [the most recent versions on the Docker Hub](https://hub.docker.com/r/dxflrs/garage/tags?page=1&ordering=last_updated).
 
 For example:
 
 ```
-sudo docker pull dxflrs/garage:v1.3.0
+sudo docker pull dxflrs/garage:v2.0.0
 ```
 
 ## Deploying and configuring Garage
@@ -171,7 +171,7 @@ docker run \
   -v /etc/garage.toml:/etc/garage.toml \
   -v /var/lib/garage/meta:/var/lib/garage/meta \
   -v /var/lib/garage/data:/var/lib/garage/data \
-  dxflrs/garage:v1.3.0
+  dxflrs/garage:v2.0.0
 ```
 
 With this command line, Garage should be started automatically at each boot.
@@ -185,7 +185,7 @@ If you want to use `docker-compose`, you may use the following `docker-compose.y
 version: "3"
 services:
   garage:
-    image: dxflrs/garage:v1.3.0
+    image: dxflrs/garage:v2.0.0
     network_mode: "host"
     restart: unless-stopped
     volumes:

doc/book/cookbook/systemd.md

@@ -28,7 +28,6 @@ StateDirectory=garage
 DynamicUser=true
 ProtectHome=true
 NoNewPrivileges=true
-LimitNOFILE=42000
 
 [Install]
 WantedBy=multi-user.target

doc/book/quick-start/_index.md

@@ -129,10 +129,10 @@ docker run \
   -d \
   --name garaged \
   -p 3900:3900 -p 3901:3901 -p 3902:3902 -p 3903:3903 \
-  -v /path/to/garage.toml:/etc/garage.toml \
-  -v /path/to/garage/meta:/var/lib/garage/meta \
-  -v /path/to/garage/data:/var/lib/garage/data \
-  dxflrs/garage:v1.3.0
+  -v /etc/garage.toml:/path/to/garage.toml \
+  -v /var/lib/garage/meta:/path/to/garage/meta \
+  -v /var/lib/garage/data:/path/to/garage/data \
+  dxflrs/garage:v2.0.0
 ```
 
 Under Linux, you can substitute `--network host` for `-p 3900:3900 -p 3901:3901 -p 3902:3902 -p 3903:3903`
@@ -182,12 +182,11 @@ ID                 Hostname  Address         Tag                   Zone  Capacit
 ## Creating a cluster layout
 
 Creating a cluster layout for a Garage deployment means informing Garage
-of the disk space available on each node of the cluster, `-c`,
-as well as the name of the zone (e.g. datacenter), `-z`, each machine is located in.
+of the disk space available on each node of the cluster
+as well as the zone (e.g. datacenter) each machine is located in.
 
-For our test deployment, we are have only one node with zone named `dc1` and a
-capacity of `1G`, though the capacity is ignored for a single node deployment
-and can be changed later when adding new nodes.
+For our test deployment, we are using only one node. The way in which we configure
+it does not matter, you can simply write:
 
 ```bash
 garage layout assign -z dc1 -c 1G <node_id>

doc/book/reference-manual/configuration.md

@@ -24,8 +24,7 @@ db_engine = "lmdb"
 
 block_size = "1M"
 block_ram_buffer_max = "256MiB"
-block_max_concurrent_reads = 16
-block_max_concurrent_writes_per_request =10
+
 lmdb_map_size = "1T"
 
 compression_level = 1
@@ -47,7 +46,6 @@ bootstrap_peers = [
     "212fd62eeaca72c122b45a7f4fa0f55e012aa5e24ac384a72a3016413fa724ff@[fc00:F::1]:3901",
 ]
 
-allow_punycode = false
 
 [consul_discovery]
 api = "catalog"
@@ -82,6 +80,7 @@ add_host_to_metrics = true
 [admin]
 api_bind_addr = "0.0.0.0:3903"
 metrics_token = "BCAdFjoa9G0KJR0WXnHHm7fs1ZAbfpI8iIZ+Z/a2NgI="
+metrics_require_token = true
 admin_token = "UkLeGWEvHnXBqnueR3ISEMWpOnm40jH2tM2HnnL/0F4="
 trace_sink = "http://localhost:4317"
 ```
@@ -94,32 +93,29 @@ The following gives details about each available configuration option.
 
 [Environment variables](#env_variables).
 
-Top-level configuration options, in alphabetical order:
-[`allow_punycode`](#allow_punycode),
+Top-level configuration options:
 [`allow_world_readable_secrets`](#allow_world_readable_secrets),
-[`block_max_concurrent_reads`](`block_max_concurrent_reads),
 [`block_ram_buffer_max`](#block_ram_buffer_max),
-[`block_max_concurrent_writes_per_request`](#block_max_concurrent_writes_per_request),
 [`block_size`](#block_size),
 [`bootstrap_peers`](#bootstrap_peers),
 [`compression_level`](#compression_level),
-[`consistency_mode`](#consistency_mode),
 [`data_dir`](#data_dir),
 [`data_fsync`](#data_fsync),
 [`db_engine`](#db_engine),
 [`disable_scrub`](#disable_scrub),
+[`use_local_tz`](#use_local_tz),
 [`lmdb_map_size`](#lmdb_map_size),
 [`metadata_auto_snapshot_interval`](#metadata_auto_snapshot_interval),
 [`metadata_dir`](#metadata_dir),
 [`metadata_fsync`](#metadata_fsync),
 [`metadata_snapshots_dir`](#metadata_snapshots_dir),
 [`replication_factor`](#replication_factor),
+[`consistency_mode`](#consistency_mode),
 [`rpc_bind_addr`](#rpc_bind_addr),
 [`rpc_bind_outgoing`](#rpc_bind_outgoing),
 [`rpc_public_addr`](#rpc_public_addr),
 [`rpc_public_addr_subnet`](#rpc_public_addr_subnet)
-[`rpc_secret`/`rpc_secret_file`](#rpc_secret),
-[`use_local_tz`](#use_local_tz).
+[`rpc_secret`/`rpc_secret_file`](#rpc_secret).
 
 The `[consul_discovery]` section:
 [`api`](#consul_api),
@@ -150,23 +146,20 @@ The `[s3_web]` section:
 
 The `[admin]` section:
 [`api_bind_addr`](#admin_api_bind_addr),
+[`metrics_require_token`](#admin_metrics_require_token),
 [`metrics_token`/`metrics_token_file`](#admin_metrics_token),
 [`admin_token`/`admin_token_file`](#admin_token),
 [`trace_sink`](#admin_trace_sink),
 
 ### Environment variables {#env_variables}
 
-The following configuration parameters must be specified as environment variables,
-they do not exist in the configuration file:
+The following configuration parameter must be specified as an environment
+variable, it does not exist in the configuration file:
 
 - `GARAGE_LOG_TO_SYSLOG` (since `v0.9.4`): set this to `1` or `true` to make the
   Garage daemon send its logs to `syslog` (using the libc `syslog` function)
   instead of printing to stderr.
 
-- `GARAGE_LOG_TO_JOURNALD` (since `v1.2.0`): set this to `1` or `true` to make the
-  Garage daemon send its logs to `journald` (using the native protocol of `systemd-journald`)
-  instead of printing to stderr.
-
 The following environment variables can be used to override the corresponding
 values in the configuration file:
 
@@ -178,7 +171,7 @@ values in the configuration file:
 
 ### Top-level configuration options
 
-#### `replication_factor` (since `v1.0.0`) {#replication_factor}
+#### `replication_factor` {#replication_factor}
 
 The replication factor can be any positive integer smaller or equal the node count in your cluster.
 The chosen replication factor has a big impact on the cluster's failure tolerancy and performance characteristics.
@@ -226,7 +219,7 @@ is in progress.  In theory, no data should be lost as rebalancing is a
 routine operation for Garage, although we cannot guarantee you that everything
  will go right in such an extreme scenario.
 
-#### `consistency_mode` (since `v1.0.0`) {#consistency_mode}
+#### `consistency_mode` {#consistency_mode}
 
 The consistency mode setting determines the read and write behaviour of your cluster.
 
@@ -336,7 +329,6 @@ Since `v0.8.0`, Garage can use alternative storage backends as follows:
 | --------- | ----------------- | ------------- |
 | [LMDB](https://www.symas.com/lmdb) (since `v0.8.0`, default since `v0.9.0`) | `"lmdb"` | `<metadata_dir>/db.lmdb/` |
 | [Sqlite](https://sqlite.org) (since `v0.8.0`) | `"sqlite"` | `<metadata_dir>/db.sqlite` |
-| [Fjall](https://github.com/fjall-rs/fjall) (**experimental support** since `v1.3.0`) | `"fjall"` | `<metadata_dir>/db.fjall/` |
 | [Sled](https://sled.rs) (old default, removed since `v1.0`) | `"sled"` | `<metadata_dir>/db/` |
 
 Sled was supported until Garage v0.9.x, and was removed in Garage v1.0.
@@ -373,14 +365,6 @@ LMDB works very well, but is known to have the following limitations:
   so it is not the best choice for high-performance storage clusters,
   but it should work fine in many cases.
 
-- Fjall: a storage engine based on LSM trees, which theoretically allow for
-  higher write throughput than other storage engines that are based on B-trees.
-  Using Fjall could potentially improve Garage's performance significantly in
-  write-heavy workloads. **Support for Fjall is experimental at this point**,
-  we have added it to Garage for evaluation purposes only. **Do not use it for
-  production-critical workloads.**
-
-
 It is possible to convert Garage's metadata directory from one format to another
 using the `garage convert-db` command, which should be used as follows:
 
@@ -418,7 +402,6 @@ Here is how this option impacts the different database engines:
 |----------|------------------------------------|-------------------------------|
 | Sqlite   | `PRAGMA synchronous = OFF`         | `PRAGMA synchronous = NORMAL` |
 | LMDB     | `MDB_NOMETASYNC` + `MDB_NOSYNC`    | `MDB_NOMETASYNC`              |
-| Fjall    | default options                    | not supported                 |
 
 Note that the Sqlite database is always ran in `WAL` mode (`PRAGMA journal_mode = WAL`).
 
@@ -525,37 +508,6 @@ node.
 
 The default value is 256MiB.
 
-#### `block_max_concurrent_reads` (since `v1.3.0` / `v2.1.0`) {#block_max_concurrent_reads}
-
-The maximum number of blocks (individual files in the data directory) open
-simultaneously for reading.
-
-Reducing this number does not limit the number of data blocks that can be
-transferred through the network simultaneously. This mechanism was just added
-as a backpressure mechanism for HDD read speed: it helps avoid a situation
-where too many requests are coming in and Garage is reading too many block
-files simultaneously, thus not making timely progress on any of the reads.
-
-When a request to read a data block comes in through the network, the requests
-awaits for one of the `block_max_concurrent_reads` slots to be available
-(internally implemented using a Semaphore object). Once it acquired a read
-slot, it reads the entire block file to RAM and frees the slot as soon as the
-block file is finished reading. Only after the slot is released will the
-block's data start being transferred over the network.  If the request fails to
-acquire a reading slot wihtin 15 seconds, it fails with a timeout error.
-Timeout events can be monitored through the `block_read_semaphore_timeouts`
-metric in Prometheus: a non-zero number of such events indicates an I/O
-bottleneck on HDD read speed.
-
-
-#### `block_max_concurrent_writes_per_request` (since  `v2.1.0`) {#block_max_concurrent_writes_per_request}
-
-This parameter is designed to adapt to the concurrent write performance of 
-different storage media.Maximum number of parallel block writes per put request 
-Higher values improve throughput but increase memory usage.
-
-Default: 3, Recommended: 10-30 for NVMe, 3-10 for HDD
-
 #### `lmdb_map_size` {#lmdb_map_size}
 
 This parameters can be used to set the map size used by LMDB,
@@ -654,7 +606,7 @@ be obtained by running `garage node id` and then included directly in the
 key will be returned by `garage node id` and you will have to add the IP
 yourself.
 
-#### `allow_world_readable_secrets` or `GARAGE_ALLOW_WORLD_READABLE_SECRETS` (env) {#allow_world_readable_secrets}
+### `allow_world_readable_secrets` or `GARAGE_ALLOW_WORLD_READABLE_SECRETS` (env) {#allow_world_readable_secrets}
 
 Garage checks the permissions of your secret files to make sure they're not
 world-readable. In some cases, the check might fail and consider your files as
@@ -666,13 +618,6 @@ permission verification.
 Alternatively, you can set the `GARAGE_ALLOW_WORLD_READABLE_SECRETS`
 environment variable to `true` to bypass the permissions check.
 
-#### `allow_punycode` {#allow_punycode}
-
-Allow creating buckets with names containing punycode. When used for buckets served
-as websites, this allows using almost any unicode character in the domain name.
-
-Default to `false`.
-
 ### The `[consul_discovery]` section
 
 Garage supports discovering other nodes of the cluster using Consul.  For this
@@ -824,10 +769,34 @@ See [administration API reference](@/documentation/reference-manual/admin-api.md
 Alternatively, since `v0.8.5`, a path can be used to create a unix socket. Note that for security reasons,
 the socket will have 0220 mode. Make sure to set user and group permissions accordingly.
 
+#### `admin_token`, `admin_token_file` or `GARAGE_ADMIN_TOKEN`, `GARAGE_ADMIN_TOKEN_FILE` (env) {#admin_token}
+
+The token for accessing all administration functions on the admin endpoint,
+with the exception of the metrics endpoint (see `metrics_token`).
+
+You can use any random string for this value. We recommend generating a random
+token with `openssl rand -base64 32`.
+
+For Garage version earlier than `v2.0`, if this token is not set,
+access to these endpoints is disabled entirely.
+
+Since Garage `v2.0`, additional admin API tokens can be defined dynamically
+in your Garage cluster using administration commands. This new admin token system
+is more flexible since it allows admin tokens to have an expiration date,
+and to have a scope restricted to certain admin API functions. If `admin_token`
+is set, it behaves as an admin token without expiration and with full scope.
+Otherwise, only admin API tokens defined dynamically can be used.
+
+`admin_token` was introduced in Garage `v0.7.2`.
+`admin_token_file` and the `GARAGE_ADMIN_TOKEN` environment variable are supported since Garage `v0.8.2`.
+
+`GARAGE_ADMIN_TOKEN_FILE` is supported since `v0.8.5` / `v0.9.1`.
+
 #### `metrics_token`, `metrics_token_file` or `GARAGE_METRICS_TOKEN`, `GARAGE_METRICS_TOKEN_FILE` (env) {#admin_metrics_token}
 
-The token for accessing the Metrics endpoint. If this token is not set, the
-Metrics endpoint can be accessed without access control.
+The token for accessing the Prometheus metrics endpoint (`/metrics`).
+If this token is not set, and unless `metrics_require_token` is set to `true`,
+the metrics endpoint can be accessed without access control.
 
 You can use any random string for this value. We recommend generating a random token with `openssl rand -base64 32`.
 
@@ -836,17 +805,12 @@ You can use any random string for this value. We recommend generating a random t
 
 `GARAGE_METRICS_TOKEN_FILE` is supported since `v0.8.5` / `v0.9.1`.
 
-#### `admin_token`, `admin_token_file` or `GARAGE_ADMIN_TOKEN`, `GARAGE_ADMIN_TOKEN_FILE` (env) {#admin_token}
+#### `metrics_require_token` (since `v2.0.0`) {#admin_metrics_require_token}
 
-The token for accessing all of the other administration endpoints.  If this
-token is not set, access to these endpoints is disabled entirely.
-
-You can use any random string for this value. We recommend generating a random token with `openssl rand -base64 32`.
-
-`admin_token` was introduced in Garage `v0.7.2`.
-`admin_token_file` and the `GARAGE_ADMIN_TOKEN` environment variable are supported since Garage `v0.8.2`.
-
-`GARAGE_ADMIN_TOKEN_FILE` is supported since `v0.8.5` / `v0.9.1`.
+If this is set to `true`, accessing the metrics endpoint will always require
+an access token. Valid tokens include the `metrics_token` if it is set,
+and admin API token defined dynamicaly in Garage which have
+the `Metrics` endpoint in their scope.
 
 #### `trace_sink` {#admin_trace_sink}
 

doc/book/reference-manual/s3-compatibility.md

@@ -23,17 +23,17 @@ Feel free to open a PR to suggest fixes this table. Minio is missing because the
 - 2022-05-25 - Many Ceph S3 endpoints are not documented but implemented. Following a notification from the Ceph community, we added them.
 
 
+
 ## High-level features
 
 | Feature                      | Garage                           | [Openstack Swift](https://docs.openstack.org/swift/latest/s3_compat.html) | [Ceph Object Gateway](https://docs.ceph.com/en/latest/radosgw/s3/) | [Riak CS](https://docs.riak.com/riak/cs/2.1.1/references/apis/storage/s3/index.html) | [OpenIO](https://docs.openio.io/latest/source/arch-design/s3_compliancy.html) |
 |------------------------------|----------------------------------|-----------------|---------------|---------|-----|
-| [signature v2](https://docs.aws.amazon.com/AmazonS3/latest/API/Appendix-Sigv2.html) (deprecated) | ❌ Missing | ✅ |  ✅ | ✅ | ✅ |
+| [signature v2](https://docs.aws.amazon.com/general/latest/gr/signature-version-2.html) (deprecated) | ❌ Missing | ✅ |  ✅ | ✅ | ✅ |
 | [signature v4](https://docs.aws.amazon.com/AmazonS3/latest/API/sig-v4-authenticating-requests.html) |  ✅ Implemented |  ✅ |  ✅ | ❌ | ✅ |
 | [URL path-style](https://docs.aws.amazon.com/AmazonS3/latest/userguide/VirtualHosting.html#path-style-access) (eg. `host.tld/bucket/key`) |  ✅ Implemented | ✅ |  ✅ | ❓| ✅ |
 | [URL vhost-style](https://docs.aws.amazon.com/AmazonS3/latest/userguide/VirtualHosting.html#virtual-hosted-style-access) URL (eg. `bucket.host.tld/key`) |  ✅ Implemented | ❌| ✅| ✅ | ✅ |
 | [Presigned URLs](https://docs.aws.amazon.com/AmazonS3/latest/userguide/ShareObjectPreSignedURL.html) |  ✅ Implemented | ❌|  ✅ | ✅ |  ✅(❓) |
 | [SSE-C encryption](https://docs.aws.amazon.com/AmazonS3/latest/userguide/ServerSideEncryptionCustomerKeys.html) |  ✅ Implemented | ❓ |  ✅ | ❌ |  ✅ |
-| [Bucket versioning](https://docs.aws.amazon.com/AmazonS3/latest/userguide/Versioning.html) | ❌ Missing | ✅ |  ✅ | ❌ | ✅ |
 
 *Note:* OpenIO does not says if it supports presigned URLs. Because it is part
 of signature v4 and they claim they support it without additional precisions,

doc/drafts/admin-api.md

@@ -13,8 +13,9 @@ We will bump the version numbers prefixed to each API endpoint each time the syn
 or semantics change, meaning that code that relies on these endpoints will break
 when changes are introduced.
 
-The Garage administration API was introduced in version 0.7.2, this document
-does not apply to older versions of Garage.
+The Garage administration API was introduced in version 0.7.2, and was
+changed several times.
+This document applies only to the Garage v2 API (starting with Garage v2.0.0).
 
 
 ## Access control
@@ -52,34 +53,28 @@ Returns an HTTP status 200 if the node is ready to answer user's requests,
 and an HTTP status 503 (Service Unavailable) if there are some partitions
 for which a quorum of nodes is not available.
 A simple textual message is also returned in a body with content-type `text/plain`.
-See `/v1/health` for an API that also returns JSON output.
+See `/v2/GetClusterHealth` for an API that also returns JSON output.
+
+### Other special endpoints
+
+#### CheckDomain `GET /check?domain=<domain>`
+
+Checks whether this Garage cluster serves a website for domain `<domain>`.
+Returns HTTP 200 Ok if yes, or HTTP 4xx if no website is available for this domain.
 
 ### Cluster operations
 
-#### GetClusterStatus `GET /v1/status`
+#### GetClusterStatus `GET /v2/GetClusterStatus`
 
 Returns the cluster's current status in JSON, including:
 
-- ID of the node being queried and its version of the Garage daemon
 - Live nodes
 - Currently configured cluster layout
-- Staged changes to the cluster layout
 
 Example response body:
 
 ```json
 {
-  "node": "b10c110e4e854e5aa3f4637681befac755154b20059ec163254ddbfae86b09df",
-  "garageVersion": "v1.3.0",
-  "garageFeatures": [
-    "k2v",
-    "lmdb",
-    "sqlite",
-    "metrics",
-    "bundled-libs"
-  ],
-  "rustVersion": "1.68.0",
-  "dbEngine": "LMDB (using Heed crate)",
   "layoutVersion": 5,
   "nodes": [
     {
@@ -169,7 +164,7 @@ Example response body:
 }
 ```
 
-#### GetClusterHealth `GET /v1/health`
+#### GetClusterHealth `GET /v2/GetClusterHealth`
 
 Returns the cluster's current health in JSON format, with the following variables:
 
@@ -202,7 +197,7 @@ Example response body:
 }
 ```
 
-#### ConnectClusterNodes `POST /v1/connect`
+#### ConnectClusterNodes `POST /v2/ConnectClusterNodes`
 
 Instructs this Garage node to connect to other Garage nodes at specified addresses.
 
@@ -232,7 +227,7 @@ Example response:
 ]
 ```
 
-#### GetClusterLayout `GET /v1/layout`
+#### GetClusterLayout `GET /v2/GetClusterLayout`
 
 Returns the cluster's current layout in JSON, including:
 
@@ -293,7 +288,7 @@ Example response body:
 }
 ```
 
-#### UpdateClusterLayout `POST /v1/layout`
+#### UpdateClusterLayout `POST /v2/UpdateClusterLayout`
 
 Send modifications to the cluster layout. These modifications will
 be included in the staged role changes, visible in subsequent calls
@@ -330,7 +325,7 @@ This returns the new cluster layout with the proposed staged changes,
 as returned by GetClusterLayout.
 
 
-#### ApplyClusterLayout `POST /v1/layout/apply`
+#### ApplyClusterLayout `POST /v2/ApplyClusterLayout`
 
 Applies to the cluster the layout changes currently registered as
 staged layout changes.
@@ -350,23 +345,11 @@ existing layout in the cluster.
 This returns the message describing all the calculations done to compute the new
 layout, as well as the description of the layout as returned by GetClusterLayout.
 
-#### RevertClusterLayout `POST /v1/layout/revert`
+#### RevertClusterLayout `POST /v2/RevertClusterLayout`
 
 Clears all of the staged layout changes.
 
-Request body format:
-
-```json
-{
-  "version": 13
-}
-```
-
-Reverting the staged changes is done by incrementing the version number
-and clearing the contents of the staged change list.
-Similarly to the CLI, the body must include the incremented
-version number, which MUST be 1 + the value of the currently
-existing layout in the cluster.
+This requests contains an empty body.
 
 This returns the new cluster layout with all changes reverted,
 as returned by GetClusterLayout.
@@ -374,7 +357,7 @@ as returned by GetClusterLayout.
 
 ### Access key operations
 
-#### ListKeys `GET /v1/key`
+#### ListKeys `GET /v2/ListKeys`
 
 Returns all API access keys in the cluster.
 
@@ -393,8 +376,8 @@ Example response:
 ]
 ```
 
-#### GetKeyInfo `GET /v1/key?id=<acces key id>`
-#### GetKeyInfo `GET /v1/key?search=<pattern>`
+#### GetKeyInfo `GET /v2/GetKeyInfo?id=<acces key id>`
+#### GetKeyInfo `GET /v2/GetKeyInfo?search=<pattern>`
 
 Returns information about the requested API access key.
 
@@ -468,7 +451,7 @@ Example response:
 }
 ```
 
-#### CreateKey `POST /v1/key`
+#### CreateKey `POST /v2/CreateKey`
 
 Creates a new API access key.
 
@@ -483,7 +466,7 @@ Request body format:
 This returns the key info, including the created secret key,
 in the same format as the result of GetKeyInfo.
 
-#### ImportKey `POST /v1/key/import`
+#### ImportKey `POST /v2/ImportKey`
 
 Imports an existing API key.
 This will check that the imported key is in the valid format, i.e.
@@ -501,7 +484,7 @@ Request body format:
 
 This returns the key info in the same format as the result of GetKeyInfo.
 
-#### UpdateKey `POST /v1/key?id=<acces key id>`
+#### UpdateKey `POST /v2/UpdateKey?id=<acces key id>`
 
 Updates information about the specified API access key.
 
@@ -523,14 +506,14 @@ The possible flags in `allow` and `deny` are: `createBucket`.
 
 This returns the key info in the same format as the result of GetKeyInfo.
 
-#### DeleteKey `DELETE /v1/key?id=<acces key id>`
+#### DeleteKey `POST /v2/DeleteKey?id=<acces key id>`
 
 Deletes an API access key.
 
 
 ### Bucket operations
 
-#### ListBuckets `GET /v1/bucket`
+#### ListBuckets `GET /v2/ListBuckets`
 
 Returns all storage buckets in the cluster.
 
@@ -572,8 +555,8 @@ Example response:
 ]
 ```
 
-#### GetBucketInfo `GET /v1/bucket?id=<bucket id>`
-#### GetBucketInfo `GET /v1/bucket?globalAlias=<alias>`
+#### GetBucketInfo `GET /v2/GetBucketInfo?id=<bucket id>`
+#### GetBucketInfo `GET /v2/GetBucketInfo?globalAlias=<alias>`
 
 Returns information about the requested storage bucket.
 
@@ -616,7 +599,7 @@ Example response:
 }
 ```
 
-#### CreateBucket `POST /v1/bucket`
+#### CreateBucket `POST /v2/CreateBucket`
 
 Creates a new storage bucket.
 
@@ -656,7 +639,7 @@ or no alias at all.
 Technically, you can also specify both `globalAlias` and `localAlias` and that would create
 two aliases, but I don't see why you would want to do that.
 
-#### UpdateBucket `PUT /v1/bucket?id=<bucket id>`
+#### UpdateBucket `POST /v2/UpdateBucket?id=<bucket id>`
 
 Updates configuration of the given bucket.
 
@@ -688,16 +671,38 @@ In `quotas`: new values of `maxSize` and `maxObjects` must both be specified, or
 to remove the quotas. An absent value will be considered the same as a `null`. It is not possible
 to change only one of the two quotas.
 
-#### DeleteBucket `DELETE /v1/bucket?id=<bucket id>`
+#### DeleteBucket `POST /v2/DeleteBucket?id=<bucket id>`
 
 Deletes a storage bucket. A bucket cannot be deleted if it is not empty.
 
 Warning: this will delete all aliases associated with the bucket!
 
+#### CleanupIncompleteUploads `POST /v2/CleanupIncompleteUploads`
+
+Cleanup all incomplete uploads in a bucket that are older than a specified number
+of seconds.
+
+Request body format:
+
+```json
+{
+    "bucketId": "e6a14cd6a27f48684579ec6b381c078ab11697e6bc8513b72b2f5307e25fff9b",
+    "olderThanSecs": 3600
+}
+```
+
+Response format
+
+```json
+{
+    "uploadsDeleted": 12
+}
+```
+
 
 ### Operations on permissions for keys on buckets
 
-#### BucketAllowKey `POST /v1/bucket/allow`
+#### AllowBucketKey `POST /v2/AllowBucketKey`
 
 Allows a key to do read/write/owner operations on a bucket.
 
@@ -718,7 +723,7 @@ Request body format:
 Flags in `permissions` which have the value `true` will be activated.
 Other flags will remain unchanged.
 
-#### BucketDenyKey `POST /v1/bucket/deny`
+#### DenyBucketKey `POST /v2/DenyBucketKey`
 
 Denies a key from doing read/write/owner operations on a bucket.
 
@@ -742,19 +747,35 @@ Other flags will remain unchanged.
 
 ### Operations on bucket aliases
 
-#### GlobalAliasBucket `PUT /v1/bucket/alias/global?id=<bucket id>&alias=<global alias>`
+#### AddBucketAlias `POST /v2/AddBucketAlias`
 
-Empty body. Creates a global alias for a bucket.
+Creates an alias for a bucket in the namespace of a specific access key.
+To create a global alias, specify the `globalAlias` field.
+To create a local alias, specify the `localAlias` and `accessKeyId` fields.
 
-#### GlobalUnaliasBucket `DELETE /v1/bucket/alias/global?id=<bucket id>&alias=<global alias>`
+Request body format:
 
-Removes a global alias for a bucket.
+```json
+{
+    "bucketId": "e6a14cd6a27f48684579ec6b381c078ab11697e6bc8513b72b2f5307e25fff9b",
+    "globalAlias": "my-bucket"
+}
+```
 
-#### LocalAliasBucket `PUT /v1/bucket/alias/local?id=<bucket id>&accessKeyId=<access key ID>&alias=<local alias>`
+or:
 
-Empty body. Creates a local alias for a bucket in the namespace of a specific access key.
+```json
+{
+    "bucketId": "e6a14cd6a27f48684579ec6b381c078ab11697e6bc8513b72b2f5307e25fff9b",
+    "accessKeyId": "GK31c2f218a2e44f485b94239e",
+    "localAlias": "my-bucket"
+}
+```
 
-#### LocalUnaliasBucket `DELETE /v1/bucket/alias/local?id=<bucket id>&accessKeyId<access key ID>&alias=<local alias>`
+#### RemoveBucketAlias `POST /v2/RemoveBucketAlias`
 
-Removes a local alias for a bucket in the namespace of a specific access key.
+Removes an alias for a bucket in the namespace of a specific access key.
+To remove a global alias, specify the `globalAlias` field.
+To remove a local alias, specify the `localAlias` and `accessKeyId` fields.
 
+Request body format: same as AddBucketAlias.

flake.lock

@@ -50,17 +50,17 @@
     },
     "nixpkgs": {
       "locked": {
-        "lastModified": 1763977559,
-        "narHash": "sha256-g4MKqsIRy5yJwEsI+fYODqLUnAqIY4kZai0nldAP6EM=",
+        "lastModified": 1747825515,
+        "narHash": "sha256-BWpMQymVI73QoKZdcVCxUCCK3GNvr/xa2Dc4DM1o2BE=",
         "owner": "NixOS",
         "repo": "nixpkgs",
-        "rev": "cfe2c7d5b5d3032862254e68c37a6576b633d632",
+        "rev": "cd2812de55cf87df88a9e09bf3be1ce63d50c1a6",
         "type": "github"
       },
       "original": {
         "owner": "NixOS",
         "repo": "nixpkgs",
-        "rev": "cfe2c7d5b5d3032862254e68c37a6576b633d632",
+        "rev": "cd2812de55cf87df88a9e09bf3be1ce63d50c1a6",
         "type": "github"
       }
     },
@@ -80,17 +80,17 @@
         ]
       },
       "locked": {
-        "lastModified": 1763952169,
-        "narHash": "sha256-+PeDBD8P+NKauH+w7eO/QWCIp8Cx4mCfWnh9sJmy9CM=",
+        "lastModified": 1738549608,
+        "narHash": "sha256-GdyT9QEUSx5k/n8kILuNy83vxxdyUfJ8jL5mMpQZWfw=",
         "owner": "oxalica",
         "repo": "rust-overlay",
-        "rev": "ab726555a9a72e6dc80649809147823a813fa95b",
+        "rev": "35c6f8c4352f995ecd53896200769f80a3e8f22d",
         "type": "github"
       },
       "original": {
         "owner": "oxalica",
         "repo": "rust-overlay",
-        "rev": "ab726555a9a72e6dc80649809147823a813fa95b",
+        "rev": "35c6f8c4352f995ecd53896200769f80a3e8f22d",
         "type": "github"
       }
     },

flake.nix

@@ -2,13 +2,13 @@
   description =
     "Garage, an S3-compatible distributed object store for self-hosted deployments";
 
-  # Nixpkgs 25.05 as of 2025-11-24
+  # Nixpkgs 25.05 as of 2025-05-22
   inputs.nixpkgs.url =
-    "github:NixOS/nixpkgs/cfe2c7d5b5d3032862254e68c37a6576b633d632";
+    "github:NixOS/nixpkgs/cd2812de55cf87df88a9e09bf3be1ce63d50c1a6";
 
-  # Rust overlay as of 2025-11-24
+  # Rust overlay as of 2025-02-03
   inputs.rust-overlay.url =
-  "github:oxalica/rust-overlay/ab726555a9a72e6dc80649809147823a813fa95b";
+  "github:oxalica/rust-overlay/35c6f8c4352f995ecd53896200769f80a3e8f22d";
   inputs.rust-overlay.inputs.nixpkgs.follows = "nixpkgs";
 
   inputs.crane.url = "github:ipetkov/crane";
@@ -30,10 +30,6 @@
           inherit system nixpkgs crane rust-overlay extraTestEnv;
           release = false;
         }).garage-test;
-        lints = (compile {
-          inherit system nixpkgs crane rust-overlay;
-          release = false;
-        });
       in
       {
         packages = {
@@ -57,13 +53,6 @@
           tests-sqlite = testWith {
             GARAGE_TEST_INTEGRATION_DB_ENGINE = "sqlite";
           };
-          tests-fjall = testWith {
-            GARAGE_TEST_INTEGRATION_DB_ENGINE = "fjall";
-          };
-
-          # lints (fmt, clippy)
-          fmt = lints.garage-cargo-fmt;
-          clippy = lints.garage-cargo-clippy;
         };
 
         # ---- developpment shell, for making native builds only ----

nix/compile.nix

@@ -48,7 +48,7 @@ let
 
   inherit (pkgs) lib stdenv;
 
-  toolchainFn = (p: p.rust-bin.stable."1.91.0".default.override {
+  toolchainFn = (p: p.rust-bin.stable."1.82.0".default.override {
     targets = lib.optionals (target != null) [ rustTarget ];
     extensions = [
       "rust-src"
@@ -68,13 +68,12 @@ let
   rootFeatures = if features != null then
     features
   else
-    ([ "bundled-libs" "lmdb" "sqlite" "fjall" "k2v" ] ++ (lib.optionals release [
+    ([ "bundled-libs" "lmdb" "sqlite" "k2v" ] ++ (lib.optionals release [
       "consul-discovery"
       "kubernetes-discovery"
       "metrics"
       "telemetry-otlp"
       "syslog"
-      "journald"
     ]));
 
   featuresStr = lib.concatStringsSep "," rootFeatures;
@@ -190,15 +189,4 @@ in rec {
       pkgs.cacert
     ];
   } // extraTestEnv);
-
-  # ---- source code linting ----
-
-  garage-cargo-fmt = craneLib.cargoFmt (commonArgs // {
-    cargoExtraArgs = "";
-  });
-
-  garage-cargo-clippy = craneLib.cargoClippy (commonArgs // {
-    cargoArtifacts = garage-deps;
-    cargoClippyExtraArgs = "--all-targets -- -D warnings";
-  });
 }

script/dev-bucket.sh

@@ -17,13 +17,19 @@ else
 fi
 
 $GARAGE_BIN -c /tmp/config.1.toml bucket create eprouvette
-if [ "$GARAGE_08" = "1" ]; then
+if [ "$GARAGE_OLDVER" = "v08" ]; then
 	KEY_INFO=$($GARAGE_BIN -c /tmp/config.1.toml key new --name opérateur)
-else
+	ACCESS_KEY=`echo $KEY_INFO|grep -Po 'GK[a-f0-9]+'`
+	SECRET_KEY=`echo $KEY_INFO|grep -Po 'Secret key: [a-f0-9]+'|grep -Po '[a-f0-9]+$'`
+elif [ "$GARAGE_OLDVER" = "v1" ]; then
 	KEY_INFO=$($GARAGE_BIN -c /tmp/config.1.toml key create opérateur)
+	ACCESS_KEY=`echo $KEY_INFO|grep -Po 'GK[a-f0-9]+'`
+	SECRET_KEY=`echo $KEY_INFO|grep -Po 'Secret key: [a-f0-9]+'|grep -Po '[a-f0-9]+$'`
+else
+	KEY_INFO=$($GARAGE_BIN -c /tmp/config.1.toml json-api CreateKey '{"name":"opérateur"}')
+	ACCESS_KEY=`echo $KEY_INFO|jq -r .accessKeyId`
+	SECRET_KEY=`echo $KEY_INFO|jq -r .secretAccessKey`
 fi
-ACCESS_KEY=`echo $KEY_INFO|grep -Po 'GK[a-f0-9]+'`
-SECRET_KEY=`echo $KEY_INFO|grep -Po 'Secret key: [a-f0-9]+'|grep -Po '[a-f0-9]+$'`
 $GARAGE_BIN -c /tmp/config.1.toml bucket allow eprouvette --read --write --owner --key $ACCESS_KEY
 echo "$ACCESS_KEY $SECRET_KEY" > /tmp/garage.s3
 

script/dev-configure.sh

@@ -29,7 +29,7 @@ until $GARAGE_BIN -c /tmp/config.1.toml status 2>&1|grep -q HEALTHY ; do
 	sleep 1
 done
 
-if [ "$GARAGE_08" = "1" ]; then
+if [ "$GARAGE_OLDVER" = "v08" ]; then
 	$GARAGE_BIN -c /tmp/config.1.toml status \
 		| grep 'NO ROLE' \
 		| grep -Po '^[0-9a-f]+' \

script/dev-env-aws.sh

@@ -1,7 +1,6 @@
 export AWS_ACCESS_KEY_ID=`cat /tmp/garage.s3 |cut -d' ' -f1`
 export AWS_SECRET_ACCESS_KEY=`cat /tmp/garage.s3 |cut -d' ' -f2`
 export AWS_DEFAULT_REGION='garage'
-export AWS_REQUEST_CHECKSUM_CALCULATION='when_required'
 # FUTUREWORK: set AWS_ENDPOINT_URL instead, once nixpkgs bumps awscli to >=2.13.0.
 function aws { command aws --endpoint-url http://127.0.0.1:3911 $@ ; }
 

script/helm/garage/Chart.yaml

@@ -1,18 +1,24 @@
 apiVersion: v2
 name: garage
 description: S3-compatible object store for small self-hosted geo-distributed deployments
+
+# A chart can be either an 'application' or a 'library' chart.
+#
+# Application charts are a collection of templates that can be packaged into versioned archives
+# to be deployed.
+#
+# Library charts provide useful utilities or functions for the chart developer. They're included as
+# a dependency of application charts to inject those utilities and functions into the rendering
+# pipeline. Library charts do not define any templates and therefore cannot be deployed.
 type: application
-version: 0.7.3
-appVersion: "v1.3.1"
-home: https://garagehq.deuxfleurs.fr/
-icon: https://garagehq.deuxfleurs.fr/images/garage-logo.svg
 
-keywords:
-- geo-distributed
-- read-after-write-consistency
-- s3-compatible
+# This is the chart version. This version number should be incremented each time you make changes
+# to the chart and its templates, including the app version.
+# Versions are expected to follow Semantic Versioning (https://semver.org/)
+version: 0.8.0
 
-sources:
-- https://git.deuxfleurs.fr/Deuxfleurs/garage.git
-
-maintainers: []
+# This is the version number of the application being deployed. This version number should be
+# incremented each time you make changes to the application. Versions are not expected to
+# follow Semantic Versioning. They should reflect the version the application is using.
+# It is recommended to use it with quotes.
+appVersion: "v2.0.0"

script/helm/garage/README.md

@@ -1,15 +1,9 @@
 # garage
 
-![Version: 0.7.3](https://img.shields.io/badge/Version-0.7.3-informational?style=flat-square) ![Type: application](https://img.shields.io/badge/Type-application-informational?style=flat-square) ![AppVersion: v1.3.1](https://img.shields.io/badge/AppVersion-v1.3.1-informational?style=flat-square)
+![Version: 0.6.0](https://img.shields.io/badge/Version-0.6.0-informational?style=flat-square) ![Type: application](https://img.shields.io/badge/Type-application-informational?style=flat-square) ![AppVersion: v1.0.1](https://img.shields.io/badge/AppVersion-v1.0.1-informational?style=flat-square)
 
 S3-compatible object store for small self-hosted geo-distributed deployments
 
-**Homepage:** <https://garagehq.deuxfleurs.fr/>
-
-## Source Code
-
-* <https://git.deuxfleurs.fr/Deuxfleurs/garage.git>
-
 ## Values
 
 | Key | Type | Default | Description |
@@ -29,7 +23,6 @@ S3-compatible object store for small self-hosted geo-distributed deployments
 | garage.existingConfigMap | string | `""` | if not empty string, allow using an existing ConfigMap for the garage.toml, if set, ignores garage.toml |
 | garage.garageTomlString | string | `""` | String Template for the garage configuration if set, ignores above values. Values can be templated, see https://garagehq.deuxfleurs.fr/documentation/reference-manual/configuration/ |
 | garage.kubernetesSkipCrd | bool | `false` | Set to true if you want to use k8s discovery but install the CRDs manually outside of the helm chart, for example if you operate at namespace level without cluster ressources |
-| garage.metadataAutoSnapshotInterval | string | `""` | If this value is set, Garage will automatically take a snapshot of the metadata DB file at a regular interval and save it in the metadata directory. https://garagehq.deuxfleurs.fr/documentation/reference-manual/configuration/#metadata_auto_snapshot_interval |
 | garage.replicationMode | string | `"3"` | Default to 3 replicas, see the replication_mode section at https://garagehq.deuxfleurs.fr/documentation/reference-manual/configuration/#replication-mode |
 | garage.rpcBindAddr | string | `"[::]:3901"` |  |
 | garage.rpcSecret | string | `""` | If not given, a random secret will be generated and stored in a Secret object |
@@ -56,7 +49,6 @@ S3-compatible object store for small self-hosted geo-distributed deployments
 | initImage.pullPolicy | string | `"IfNotPresent"` |  |
 | initImage.repository | string | `"busybox"` |  |
 | initImage.tag | string | `"stable"` |  |
-| livenessProbe | object | `{}` | Specifies a livenessProbe |
 | monitoring.metrics.enabled | bool | `false` | If true, a service for monitoring is created with a prometheus.io/scrape annotation |
 | monitoring.metrics.serviceMonitor.enabled | bool | `false` | If true, a ServiceMonitor CRD is created for a prometheus operator https://github.com/coreos/prometheus-operator |
 | monitoring.metrics.serviceMonitor.interval | string | `"15s"` |  |
@@ -79,7 +71,6 @@ S3-compatible object store for small self-hosted geo-distributed deployments
 | podSecurityContext.runAsGroup | int | `1000` |  |
 | podSecurityContext.runAsNonRoot | bool | `true` |  |
 | podSecurityContext.runAsUser | int | `1000` |  |
-| readinessProbe | object | `{}` | Specifies a readinessProbe |
 | resources | object | `{}` |  |
 | securityContext.capabilities | object | `{"drop":["ALL"]}` | The default security context is heavily restricted, feel free to tune it to your requirements |
 | securityContext.readOnlyRootFilesystem | bool | `true` |  |

script/helm/garage/templates/configmap.yaml

@@ -19,10 +19,6 @@ data:
 
     compression_level = {{ .Values.garage.compressionLevel }}
 
-    {{- if .Values.garage.metadataAutoSnapshotInterval }}
-    metadata_auto_snapshot_interval = {{ .Values.garage.metadataAutoSnapshotInterval | quote }}
-    {{- end }}
-
     rpc_bind_addr = "{{ .Values.garage.rpcBindAddr }}"
     # rpc_secret will be populated by the init container from a k8s secret object
     rpc_secret = "__RPC_SECRET_REPLACE__"

script/helm/garage/templates/service.yaml

@@ -4,10 +4,6 @@ metadata:
   name: {{ include "garage.fullname" . }}
   labels:
     {{- include "garage.labels" . | nindent 4 }}
-  {{- with .Values.service.annotations }}
-  annotations:
-    {{- toYaml . | nindent 4 }}
-  {{- end }}
 spec:
   type: {{ .Values.service.type }}
   ports:

script/helm/garage/templates/workload.yaml

@@ -78,14 +78,15 @@ spec:
           {{- with .Values.extraVolumeMounts }}
             {{- toYaml . | nindent 12 }}
           {{- end }}
-          {{- with .Values.livenessProbe }}
-          livenessProbe:
-            {{- toYaml . | nindent 12 }}
-          {{- end }}
-          {{- with .Values.readinessProbe }}
-          readinessProbe:
-            {{- toYaml . | nindent 12 }}
-          {{- end }}
+          # TODO
+          # livenessProbe:
+          #   httpGet:
+          #     path: /
+          #     port: 3900
+          # readinessProbe:
+          #   httpGet:
+          #     path: /
+          #     port: 3900
           resources:
             {{- toYaml .Values.resources | nindent 12 }}
       volumes:

script/helm/garage/values.yaml

@@ -21,10 +21,6 @@ garage:
   # https://garagehq.deuxfleurs.fr/documentation/reference-manual/configuration/#compression-level
   compressionLevel: "1"
 
-  # -- If this value is set, Garage will automatically take a snapshot of the metadata DB file at a regular interval and save it in the metadata directory.
-  # https://garagehq.deuxfleurs.fr/documentation/reference-manual/configuration/#metadata_auto_snapshot_interval
-  metadataAutoSnapshotInterval: ""
-
   rpcBindAddr: "[::]:3901"
   # -- If not given, a random secret will be generated and stored in a Secret object
   rpcSecret: ""
@@ -124,8 +120,6 @@ service:
   # - NodePort (+ Ingress)
   # - LoadBalancer
   type: ClusterIP
-  # -- Annotations to add to the service
-  annotations: {}
   s3:
     api:
       port: 3900
@@ -197,21 +191,6 @@ resources: {}
   #   cpu: 100m
   #   memory: 512Mi
 
-# -- Specifies a livenessProbe
-livenessProbe: {}
-  #httpGet:
-  #  path: /health
-  #  port: 3903
-  #initialDelaySeconds: 5
-  #periodSeconds: 30
-# -- Specifies a readinessProbe
-readinessProbe: {}
-  #httpGet:
-  #  path: /health
-  #  port: 3903
-  #initialDelaySeconds: 5
-  #periodSeconds: 30
-
 nodeSelector: {}
 
 tolerations: []

script/k8s/crd/garagenodes.deuxfleurs.fr.yaml

@@ -1,43 +0,0 @@
-apiVersion: apiextensions.k8s.io/v1
-kind: CustomResourceDefinition
-metadata:
-  name: garagenodes.deuxfleurs.fr
-spec:
-  conversion:
-    strategy: None
-  group: deuxfleurs.fr
-  names:
-    kind: GarageNode
-    listKind: GarageNodeList
-    plural: garagenodes
-    singular: garagenode
-  scope: Namespaced
-  versions:
-  - name: v1
-    schema:
-      openAPIV3Schema:
-        description: Auto-generated derived type for Node via `CustomResource`
-        properties:
-          spec:
-            properties:
-              address:
-                format: ip
-                type: string
-              hostname:
-                type: string
-              port:
-                format: uint16
-                minimum: 0
-                type: integer
-            required:
-            - address
-            - hostname
-            - port
-            type: object
-        required:
-        - spec
-        title: GarageNode
-        type: object
-    served: true
-    storage: true
-    subresources: {}

script/k8s/crd/kustomization.yaml

@@ -1,5 +0,0 @@
-apiVersion: kustomize.config.k8s.io/v1beta1
-kind: Kustomization
-
-resources:
-- garagenodes.deuxfleurs.fr.yaml

script/test-smoke.sh

@@ -112,6 +112,19 @@ if [ -z "$SKIP_S3CMD" ]; then
   done
 fi
 
+# BOTO3
+if [ -z "$SKIP_BOTO3" ]; then
+  echo "🛠️ Testing with boto3 for STREAMING-UNSIGNED-PAYLOAD-TRAILER"
+  source ${SCRIPT_FOLDER}/dev-env-aws.sh
+  AWS_ENDPOINT_URL=https://localhost:4443 python <<EOF
+import boto3
+client = boto3.client('s3', verify=False)
+client.put_object(Body=b'hello world', Bucket='eprouvette', Key='test.s3.txt')
+client.delete_object(Bucket='eprouvette', Key='test.s3.txt')
+print("OK!")
+EOF
+fi
+
 # Minio Client
 if [ -z "$SKIP_MC" ]; then
   echo "🛠️ Testing with mc (minio client)"

script/test-upgrade.sh

@@ -24,9 +24,17 @@ echo "============= insert data into old version cluster ================="
 export GARAGE_BIN=/tmp/old_garage
 if echo $OLD_VERSION | grep 'v0\.8\.'; then
 	echo "Detected Garage v0.8.x"
-	export GARAGE_08=1
+	export GARAGE_OLDVER=v08
+elif (echo $OLD_VERSION | grep 'v0\.9\.') || (echo $OLD_VERSION | grep 'v1\.'); then
+	echo "Detected Garage v0.9.x / v1.x"
+	export GARAGE_OLDVER=v1
 fi
 
+if echo $OLD_VERSION | grep 'v1\.'; then
+	DO_SSEC_TEST=1
+fi
+SSEC_KEY="u8zCfnEyt5Imo/krN+sxA1DQXxLWtPJavU6T6gOVj1Y="
+
 echo "⏳ Setup cluster using old version"
 $GARAGE_BIN --version
 ${SCRIPT_FOLDER}/dev-clean.sh
@@ -37,7 +45,23 @@ ${SCRIPT_FOLDER}/dev-bucket.sh
 
 echo "🛠️ Inserting data in old cluster"
 source ${SCRIPT_FOLDER}/dev-env-rclone.sh
-rclone copy "${SCRIPT_FOLDER}/../.git/" garage:eprouvette/test_dotgit --stats=1s --stats-log-level=NOTICE --stats-one-line
+rclone copy "${SCRIPT_FOLDER}/../.git/" garage:eprouvette/test_dotgit \
+	--stats=1s --stats-log-level=NOTICE --stats-one-line
+
+if [ "$DO_SSEC_TEST" = "1" ]; then
+	# upload small file (should be single part)
+	rclone copy "${SCRIPT_FOLDER}/test-upgrade.sh" garage:eprouvette/test-ssec \
+		--s3-sse-customer-algorithm AES256 \
+		--s3-sse-customer-key-base64 "$SSEC_KEY" \
+		--stats=1s --stats-log-level=NOTICE --stats-one-line
+	# do a multipart upload
+	dd if=/dev/urandom of=/tmp/randfile-for-upgrade bs=5M count=5
+	rclone copy "/tmp/randfile-for-upgrade" garage:eprouvette/test-ssec \
+		--s3-chunk-size 5M \
+		--s3-sse-customer-algorithm AES256 \
+		--s3-sse-customer-key-base64 "$SSEC_KEY" \
+		--stats=1s --stats-log-level=NOTICE --stats-one-line
+fi
 
 echo "🏁 Stopping old cluster"
 killall -INT old_garage
@@ -47,7 +71,7 @@ killall -9 old_garage || true
 echo "🏁 Removing old garage version"
 rm -rv $GARAGE_BIN
 export -n GARAGE_BIN
-export -n GARAGE_08
+export -n GARAGE_OLDVER
 
 echo "================ read data from new cluster ==================="
 
@@ -60,7 +84,8 @@ ${SCRIPT_FOLDER}/dev-cluster.sh >> /tmp/garage.log 2>&1 &
 sleep 3
 
 echo "🛠️ Retrieving data from old cluster"
-rclone copy garage:eprouvette/test_dotgit /tmp/test_dotgit --stats=1s --stats-log-level=NOTICE --stats-one-line --fast-list
+rclone copy garage:eprouvette/test_dotgit /tmp/test_dotgit \
+	--stats=1s --stats-log-level=NOTICE --stats-one-line --fast-list
 
 if ! diff <(find "${SCRIPT_FOLDER}/../.git" -type f | xargs md5sum | cut -d ' ' -f 1 | sort) <(find /tmp/test_dotgit -type f | xargs md5sum | cut -d ' ' -f 1 | sort); then
 	echo "TEST FAILURE: directories are different"
@@ -68,6 +93,23 @@ if ! diff <(find "${SCRIPT_FOLDER}/../.git" -type f | xargs md5sum | cut -d ' '
 fi
 rm -r /tmp/test_dotgit
 
+if [ "$DO_SSEC_TEST" = "1" ]; then
+	rclone copy garage:eprouvette/test-ssec /tmp/test_ssec_out \
+		--s3-sse-customer-algorithm AES256 \
+		--s3-sse-customer-key-base64 "$SSEC_KEY" \
+		--stats=1s --stats-log-level=NOTICE --stats-one-line
+	if ! diff "/tmp/test_ssec_out/test-upgrade.sh" "${SCRIPT_FOLDER}/test-upgrade.sh"; then
+		echo "SSEC-FAILURE (small file)"
+		exit 1
+	fi
+	if ! diff "/tmp/test_ssec_out/randfile-for-upgrade" "/tmp/randfile-for-upgrade"; then
+		echo "SSEC-FAILURE (big file)"
+		exit 1
+	fi
+	rm -r /tmp/test_ssec_out
+	rm /tmp/randfile-for-upgrade
+fi
+
 echo "🏁 Teardown"
 rm -rf /tmp/garage-{data,meta}-*
 rm -rf /tmp/config.*.toml

shell.nix

@@ -26,6 +26,8 @@ in
       s3cmd
       minio-client
       rclone
+      (python312.withPackages (ps: [ ps.boto3 ]))
+
       socat
       psmisc
       which
@@ -34,8 +36,6 @@ in
       jq
     ];
     shellHook = ''
-      export AWS_REQUEST_CHECKSUM_CALCULATION='when_required'
-
       function to_s3 {
         aws \
             --endpoint-url https://garage.deuxfleurs.fr \

src/api/admin/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "garage_api_admin"
-version = "1.3.1"
+version = "2.0.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -14,7 +14,9 @@ path = "lib.rs"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
+format_table.workspace = true
 garage_model.workspace = true
+garage_block.workspace = true
 garage_table.workspace = true
 garage_util.workspace = true
 garage_rpc.workspace = true
@@ -22,8 +24,11 @@ garage_api_common.workspace = true
 
 argon2.workspace = true
 async-trait.workspace = true
-thiserror.workspace = true
+bytesize.workspace = true
+chrono.workspace = true
+err-derive.workspace = true
 hex.workspace = true
+paste.workspace = true
 tracing.workspace = true
 
 futures.workspace = true
@@ -34,6 +39,7 @@ url.workspace = true
 
 serde.workspace = true
 serde_json.workspace = true
+utoipa.workspace = true
 
 opentelemetry.workspace = true
 opentelemetry-prometheus = { workspace = true, optional = true }
@@ -41,3 +47,4 @@ prometheus = { workspace = true, optional = true }
 
 [features]
 metrics = [ "opentelemetry-prometheus", "prometheus" ]
+k2v = [ "garage_model/k2v" ]

src/api/admin/admin_token.rs

@@ -0,0 +1,235 @@
+use std::sync::Arc;
+
+use chrono::{DateTime, Utc};
+
+use garage_table::*;
+use garage_util::time::now_msec;
+
+use garage_model::admin_token_table::*;
+use garage_model::garage::Garage;
+
+use crate::api::*;
+use crate::error::*;
+use crate::{Admin, RequestHandler};
+
+impl RequestHandler for ListAdminTokensRequest {
+	type Response = ListAdminTokensResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<ListAdminTokensResponse, Error> {
+		let now = now_msec();
+
+		let mut res = garage
+			.admin_token_table
+			.get_range(
+				&EmptyKey,
+				None,
+				Some(KeyFilter::Deleted(DeletedFilter::NotDeleted)),
+				10000,
+				EnumerationOrder::Forward,
+			)
+			.await?
+			.iter()
+			.map(|t| admin_token_info_results(t, now))
+			.collect::<Vec<_>>();
+
+		if garage.config.admin.metrics_token.is_some() {
+			res.insert(
+				0,
+				GetAdminTokenInfoResponse {
+					id: None,
+					created: None,
+					name: "metrics_token (from daemon configuration)".into(),
+					expiration: None,
+					expired: false,
+					scope: vec!["Metrics".into()],
+				},
+			);
+		}
+
+		if garage.config.admin.admin_token.is_some() {
+			res.insert(
+				0,
+				GetAdminTokenInfoResponse {
+					id: None,
+					created: None,
+					name: "admin_token (from daemon configuration)".into(),
+					expiration: None,
+					expired: false,
+					scope: vec!["*".into()],
+				},
+			);
+		}
+
+		Ok(ListAdminTokensResponse(res))
+	}
+}
+
+impl RequestHandler for GetAdminTokenInfoRequest {
+	type Response = GetAdminTokenInfoResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<GetAdminTokenInfoResponse, Error> {
+		let token = match (self.id, self.search) {
+			(Some(id), None) => get_existing_admin_token(garage, &id).await?,
+			(None, Some(search)) => {
+				let candidates = garage
+					.admin_token_table
+					.get_range(
+						&EmptyKey,
+						None,
+						Some(KeyFilter::MatchesAndNotDeleted(search.to_string())),
+						10,
+						EnumerationOrder::Forward,
+					)
+					.await?
+					.into_iter()
+					.collect::<Vec<_>>();
+				if candidates.len() != 1 {
+					return Err(Error::bad_request(format!(
+						"{} matching admin tokens",
+						candidates.len()
+					)));
+				}
+				candidates.into_iter().next().unwrap()
+			}
+			_ => {
+				return Err(Error::bad_request(
+					"Either id or search must be provided (but not both)",
+				));
+			}
+		};
+
+		Ok(admin_token_info_results(&token, now_msec()))
+	}
+}
+
+impl RequestHandler for CreateAdminTokenRequest {
+	type Response = CreateAdminTokenResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<CreateAdminTokenResponse, Error> {
+		let (mut token, secret) = if self.0.name.is_some() {
+			AdminApiToken::new("")
+		} else {
+			AdminApiToken::new(&format!("token_{}", Utc::now().format("%Y%m%d_%H%M")))
+		};
+
+		apply_token_updates(&mut token, self.0)?;
+
+		garage.admin_token_table.insert(&token).await?;
+
+		Ok(CreateAdminTokenResponse {
+			secret_token: secret,
+			info: admin_token_info_results(&token, now_msec()),
+		})
+	}
+}
+
+impl RequestHandler for UpdateAdminTokenRequest {
+	type Response = UpdateAdminTokenResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<UpdateAdminTokenResponse, Error> {
+		let mut token = get_existing_admin_token(&garage, &self.id).await?;
+
+		apply_token_updates(&mut token, self.body)?;
+
+		garage.admin_token_table.insert(&token).await?;
+
+		Ok(UpdateAdminTokenResponse(admin_token_info_results(
+			&token,
+			now_msec(),
+		)))
+	}
+}
+
+impl RequestHandler for DeleteAdminTokenRequest {
+	type Response = DeleteAdminTokenResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<DeleteAdminTokenResponse, Error> {
+		let token = get_existing_admin_token(&garage, &self.id).await?;
+
+		garage
+			.admin_token_table
+			.insert(&AdminApiToken::delete(token.prefix))
+			.await?;
+
+		Ok(DeleteAdminTokenResponse)
+	}
+}
+
+// ---- helpers ----
+
+fn admin_token_info_results(token: &AdminApiToken, now: u64) -> GetAdminTokenInfoResponse {
+	let params = token.params().unwrap();
+
+	GetAdminTokenInfoResponse {
+		id: Some(token.prefix.clone()),
+		created: Some(
+			DateTime::from_timestamp_millis(params.created as i64)
+				.expect("invalid timestamp stored in db"),
+		),
+		name: params.name.get().to_string(),
+		expiration: params.expiration.get().map(|x| {
+			DateTime::from_timestamp_millis(x as i64).expect("invalid timestamp stored in db")
+		}),
+		expired: params.is_expired(now),
+		scope: params.scope.get().0.clone(),
+	}
+}
+
+async fn get_existing_admin_token(garage: &Garage, id: &String) -> Result<AdminApiToken, Error> {
+	garage
+		.admin_token_table
+		.get(&EmptyKey, id)
+		.await?
+		.filter(|k| !k.state.is_deleted())
+		.ok_or_else(|| Error::NoSuchAdminToken(id.to_string()))
+}
+
+fn apply_token_updates(
+	token: &mut AdminApiToken,
+	updates: UpdateAdminTokenRequestBody,
+) -> Result<(), Error> {
+	if updates.never_expires && updates.expiration.is_some() {
+		return Err(Error::bad_request(
+			"cannot specify `expiration` and `never_expires`",
+		));
+	}
+
+	let params = token.params_mut().unwrap();
+
+	if let Some(name) = updates.name {
+		params.name.update(name);
+	}
+	if let Some(expiration) = updates.expiration {
+		params
+			.expiration
+			.update(Some(expiration.timestamp_millis() as u64));
+	}
+	if updates.never_expires {
+		params.expiration.update(None);
+	}
+	if let Some(scope) = updates.scope {
+		params.scope.update(AdminApiTokenScope(scope));
+	}
+
+	Ok(())
+}

src/api/admin/api_server.rs

@@ -1,333 +1,230 @@
-use std::collections::HashMap;
+use std::borrow::Cow;
 use std::sync::Arc;
 
-use argon2::password_hash::PasswordHash;
-
-use http::header::{ACCESS_CONTROL_ALLOW_METHODS, ACCESS_CONTROL_ALLOW_ORIGIN, ALLOW};
-use hyper::{body::Incoming as IncomingBody, Request, Response, StatusCode};
+use http::header::{HeaderValue, ACCESS_CONTROL_ALLOW_ORIGIN, AUTHORIZATION};
+use hyper::{body::Incoming as IncomingBody, Request, Response};
+use serde::{Deserialize, Serialize};
 use tokio::sync::watch;
 
 use opentelemetry::trace::SpanRef;
 
 #[cfg(feature = "metrics")]
 use opentelemetry_prometheus::PrometheusExporter;
-#[cfg(feature = "metrics")]
-use prometheus::{Encoder, TextEncoder};
 
 use garage_model::garage::Garage;
-use garage_rpc::system::ClusterHealthStatus;
+use garage_rpc::{Endpoint as RpcEndpoint, *};
+use garage_table::EmptyKey;
+use garage_util::background::BackgroundRunner;
+use garage_util::data::Uuid;
 use garage_util::error::Error as GarageError;
 use garage_util::socket_address::UnixOrTCPSocketAddress;
+use garage_util::time::now_msec;
 
 use garage_api_common::generic_server::*;
 use garage_api_common::helpers::*;
 
-use crate::bucket::*;
-use crate::cluster::*;
+use crate::api::*;
 use crate::error::*;
-use crate::key::*;
 use crate::router_v0;
-use crate::router_v1::{Authorization, Endpoint};
+use crate::router_v1;
+use crate::Authorization;
+use crate::RequestHandler;
+
+// ---- FOR RPC ----
+
+pub const ADMIN_RPC_PATH: &str = "garage_api/admin/rpc.rs/Rpc";
+
+#[derive(Debug, Serialize, Deserialize)]
+pub enum AdminRpc {
+	Proxy(AdminApiRequest),
+	Internal(LocalAdminApiRequest),
+}
+
+#[derive(Debug, Serialize, Deserialize)]
+pub enum AdminRpcResponse {
+	ProxyApiOkResponse(TaggedAdminApiResponse),
+	InternalApiOkResponse(LocalAdminApiResponse),
+	ApiErrorResponse {
+		http_code: u16,
+		error_code: String,
+		message: String,
+	},
+}
+
+impl Rpc for AdminRpc {
+	type Response = Result<AdminRpcResponse, GarageError>;
+}
+
+impl EndpointHandler<AdminRpc> for AdminApiServer {
+	async fn handle(
+		self: &Arc<Self>,
+		message: &AdminRpc,
+		_from: NodeID,
+	) -> Result<AdminRpcResponse, GarageError> {
+		match message {
+			AdminRpc::Proxy(req) => {
+				info!("Proxied admin API request: {}", req.name());
+				let res = req.clone().handle(&self.garage, &self).await;
+				match res {
+					Ok(res) => Ok(AdminRpcResponse::ProxyApiOkResponse(res.tagged())),
+					Err(e) => Ok(AdminRpcResponse::ApiErrorResponse {
+						http_code: e.http_status_code().as_u16(),
+						error_code: e.code().to_string(),
+						message: e.to_string(),
+					}),
+				}
+			}
+			AdminRpc::Internal(req) => {
+				info!("Internal admin API request: {}", req.name());
+				let res = req.clone().handle(&self.garage, &self).await;
+				match res {
+					Ok(res) => Ok(AdminRpcResponse::InternalApiOkResponse(res)),
+					Err(e) => Ok(AdminRpcResponse::ApiErrorResponse {
+						http_code: e.http_status_code().as_u16(),
+						error_code: e.code().to_string(),
+						message: e.to_string(),
+					}),
+				}
+			}
+		}
+	}
+}
+
+// ---- FOR HTTP ----
 
 pub type ResBody = BoxBody<Error>;
 
 pub struct AdminApiServer {
 	garage: Arc<Garage>,
 	#[cfg(feature = "metrics")]
-	exporter: PrometheusExporter,
+	pub(crate) exporter: PrometheusExporter,
 	metrics_token: Option<String>,
+	metrics_require_token: bool,
 	admin_token: Option<String>,
+	pub(crate) background: Arc<BackgroundRunner>,
+	pub(crate) endpoint: Arc<RpcEndpoint<AdminRpc, Self>>,
+}
+
+pub enum HttpEndpoint {
+	Old(router_v1::Endpoint),
+	New(String),
 }
 
 impl AdminApiServer {
 	pub fn new(
 		garage: Arc<Garage>,
+		background: Arc<BackgroundRunner>,
 		#[cfg(feature = "metrics")] exporter: PrometheusExporter,
-	) -> Self {
+	) -> Arc<Self> {
 		let cfg = &garage.config.admin;
 		let metrics_token = cfg.metrics_token.as_deref().map(hash_bearer_token);
 		let admin_token = cfg.admin_token.as_deref().map(hash_bearer_token);
-		Self {
+		let metrics_require_token = cfg.metrics_require_token;
+
+		let endpoint = garage.system.netapp.endpoint(ADMIN_RPC_PATH.into());
+		let admin = Arc::new(Self {
 			garage,
 			#[cfg(feature = "metrics")]
 			exporter,
 			metrics_token,
+			metrics_require_token,
 			admin_token,
-		}
+			background,
+			endpoint,
+		});
+		admin.endpoint.set_handler(admin.clone());
+		admin
 	}
 
 	pub async fn run(
-		self,
+		self: Arc<Self>,
 		bind_addr: UnixOrTCPSocketAddress,
 		must_exit: watch::Receiver<bool>,
 	) -> Result<(), GarageError> {
 		let region = self.garage.config.s3_api.s3_region.clone();
-		ApiServer::new(region, self)
+		ApiServer::new(region, ArcAdminApiServer(self))
 			.run_server(bind_addr, Some(0o220), must_exit)
 			.await
 	}
 
-	fn handle_options(&self, _req: &Request<IncomingBody>) -> Result<Response<ResBody>, Error> {
-		Ok(Response::builder()
-			.status(StatusCode::NO_CONTENT)
-			.header(ALLOW, "OPTIONS, GET, POST")
-			.header(ACCESS_CONTROL_ALLOW_METHODS, "OPTIONS, GET, POST")
-			.header(ACCESS_CONTROL_ALLOW_ORIGIN, "*")
-			.body(empty_body())?)
-	}
-
-	async fn handle_check_domain(
+	async fn handle_http_api(
 		&self,
 		req: Request<IncomingBody>,
+		endpoint: HttpEndpoint,
 	) -> Result<Response<ResBody>, Error> {
-		let query_params: HashMap<String, String> = req
-			.uri()
-			.query()
-			.map(|v| {
-				url::form_urlencoded::parse(v.as_bytes())
-					.into_owned()
-					.collect()
-			})
-			.unwrap_or_else(HashMap::new);
+		let auth_header = req.headers().get(AUTHORIZATION).cloned();
 
-		let has_domain_key = query_params.contains_key("domain");
-
-		if !has_domain_key {
-			return Err(Error::bad_request("No domain query string found"));
-		}
-
-		let domain = query_params
-			.get("domain")
-			.ok_or_internal_error("Could not parse domain query string")?;
-
-		if self.check_domain(domain).await? {
-			Ok(Response::builder()
-				.status(StatusCode::OK)
-				.body(string_body(format!(
-					"Domain '{domain}' is managed by Garage"
-				)))?)
-		} else {
-			Err(Error::bad_request(format!(
-				"Domain '{domain}' is not managed by Garage"
-			)))
-		}
-	}
-
-	async fn check_domain(&self, domain: &str) -> Result<bool, Error> {
-		// Resolve bucket from domain name, inferring if the website must be activated for the
-		// domain to be valid.
-		let (bucket_name, must_check_website) = if let Some(bname) = self
-			.garage
-			.config
-			.s3_api
-			.root_domain
-			.as_ref()
-			.and_then(|rd| host_to_bucket(domain, rd))
-		{
-			(bname.to_string(), false)
-		} else if let Some(bname) = self
-			.garage
-			.config
-			.s3_web
-			.as_ref()
-			.and_then(|sw| host_to_bucket(domain, sw.root_domain.as_str()))
-		{
-			(bname.to_string(), true)
-		} else {
-			(domain.to_string(), true)
+		let request = match endpoint {
+			HttpEndpoint::Old(endpoint_v1) => AdminApiRequest::from_v1(endpoint_v1, req).await?,
+			HttpEndpoint::New(_) => AdminApiRequest::from_request(req).await?,
 		};
 
-		let bucket_id = match self
-			.garage
-			.bucket_helper()
-			.resolve_global_bucket_name(&bucket_name)
-			.await?
-		{
-			Some(bucket_id) => bucket_id,
-			None => return Ok(false),
-		};
-
-		if !must_check_website {
-			return Ok(true);
-		}
-
-		let bucket = self
-			.garage
-			.bucket_helper()
-			.get_existing_bucket(bucket_id)
-			.await?;
-
-		let bucket_state = bucket.state.as_option().unwrap();
-		let bucket_website_config = bucket_state.website_config.get();
-
-		match bucket_website_config {
-			Some(_v) => Ok(true),
-			None => Ok(false),
-		}
-	}
-
-	fn handle_health(&self) -> Result<Response<ResBody>, Error> {
-		let health = self.garage.system.health();
-
-		let (status, status_str) = match health.status {
-			ClusterHealthStatus::Healthy => (StatusCode::OK, "Garage is fully operational"),
-			ClusterHealthStatus::Degraded => (
-				StatusCode::OK,
-				"Garage is operational but some storage nodes are unavailable",
-			),
-			ClusterHealthStatus::Unavailable => (
-				StatusCode::SERVICE_UNAVAILABLE,
-				"Quorum is not available for some/all partitions, reads and writes will fail",
+		let (global_token_hash, token_required) = match request.authorization_type() {
+			Authorization::None => (None, false),
+			Authorization::MetricsToken => (
+				self.metrics_token.as_deref(),
+				self.metrics_token.is_some() || self.metrics_require_token,
 			),
+			Authorization::AdminToken => (self.admin_token.as_deref(), true),
 		};
-		let status_str = format!(
-			"{}\nConsult the full health check API endpoint at /v1/health for more details\n",
-			status_str
-		);
 
-		Ok(Response::builder()
-			.status(status)
-			.header(http::header::CONTENT_TYPE, "text/plain")
-			.body(string_body(status_str))?)
+		if token_required {
+			verify_authorization(&self.garage, global_token_hash, auth_header, request.name())?;
 		}
 
-	fn handle_metrics(&self) -> Result<Response<ResBody>, Error> {
-		#[cfg(feature = "metrics")]
-		{
-			use opentelemetry::trace::Tracer;
-
-			let mut buffer = vec![];
-			let encoder = TextEncoder::new();
-
-			let tracer = opentelemetry::global::tracer("garage");
-			let metric_families = tracer.in_span("admin/gather_metrics", |_| {
-				self.exporter.registry().gather()
-			});
-
-			encoder
-				.encode(&metric_families, &mut buffer)
-				.ok_or_internal_error("Could not serialize metrics")?;
-
-			Ok(Response::builder()
-				.status(StatusCode::OK)
-				.header(http::header::CONTENT_TYPE, encoder.format_type())
-				.body(bytes_body(buffer.into()))?)
+		match request {
+			AdminApiRequest::Options(req) => req.handle(&self.garage, &self).await,
+			AdminApiRequest::CheckDomain(req) => req.handle(&self.garage, &self).await,
+			AdminApiRequest::Health(req) => req.handle(&self.garage, &self).await,
+			AdminApiRequest::Metrics(req) => req.handle(&self.garage, &self).await,
+			req => {
+				let res = req.handle(&self.garage, &self).await?;
+				let mut res = json_ok_response(&res)?;
+				res.headers_mut()
+					.insert(ACCESS_CONTROL_ALLOW_ORIGIN, HeaderValue::from_static("*"));
+				Ok(res)
+			}
 		}
-		#[cfg(not(feature = "metrics"))]
-		Err(Error::bad_request(
-			"Garage was built without the metrics feature".to_string(),
-		))
 	}
 }
 
-impl ApiHandler for AdminApiServer {
+struct ArcAdminApiServer(Arc<AdminApiServer>);
+
+impl ApiHandler for ArcAdminApiServer {
 	const API_NAME: &'static str = "admin";
 	const API_NAME_DISPLAY: &'static str = "Admin";
 
-	type Endpoint = Endpoint;
+	type Endpoint = HttpEndpoint;
 	type Error = Error;
 
-	fn parse_endpoint(&self, req: &Request<IncomingBody>) -> Result<Endpoint, Error> {
+	fn parse_endpoint(&self, req: &Request<IncomingBody>) -> Result<HttpEndpoint, Error> {
 		if req.uri().path().starts_with("/v0/") {
 			let endpoint_v0 = router_v0::Endpoint::from_request(req)?;
-			Endpoint::from_v0(endpoint_v0)
+			let endpoint_v1 = router_v1::Endpoint::from_v0(endpoint_v0)?;
+			Ok(HttpEndpoint::Old(endpoint_v1))
+		} else if req.uri().path().starts_with("/v1/") {
+			let endpoint_v1 = router_v1::Endpoint::from_request(req)?;
+			Ok(HttpEndpoint::Old(endpoint_v1))
 		} else {
-			Endpoint::from_request(req)
+			Ok(HttpEndpoint::New(req.uri().path().to_string()))
 		}
 	}
 
 	async fn handle(
 		&self,
 		req: Request<IncomingBody>,
-		endpoint: Endpoint,
+		endpoint: HttpEndpoint,
 	) -> Result<Response<ResBody>, Error> {
-		let required_auth_hash =
-			match endpoint.authorization_type() {
-				Authorization::None => None,
-				Authorization::MetricsToken => self.metrics_token.as_deref(),
-				Authorization::AdminToken => match self.admin_token.as_deref() {
-					None => return Err(Error::forbidden(
-						"Admin token isn't configured, admin API access is disabled for security.",
-					)),
-					Some(t) => Some(t),
-				},
-			};
-
-		if let Some(password_hash) = required_auth_hash {
-			match req.headers().get("Authorization") {
-				None => return Err(Error::forbidden("Authorization token must be provided")),
-				Some(authorization) => {
-					verify_bearer_token(&authorization, password_hash)?;
-				}
-			}
-		}
-
-		match endpoint {
-			Endpoint::Options => self.handle_options(&req),
-			Endpoint::CheckDomain => self.handle_check_domain(req).await,
-			Endpoint::Health => self.handle_health(),
-			Endpoint::Metrics => self.handle_metrics(),
-			Endpoint::GetClusterStatus => handle_get_cluster_status(&self.garage).await,
-			Endpoint::GetClusterHealth => handle_get_cluster_health(&self.garage).await,
-			Endpoint::ConnectClusterNodes => handle_connect_cluster_nodes(&self.garage, req).await,
-			// Layout
-			Endpoint::GetClusterLayout => handle_get_cluster_layout(&self.garage).await,
-			Endpoint::UpdateClusterLayout => handle_update_cluster_layout(&self.garage, req).await,
-			Endpoint::ApplyClusterLayout => handle_apply_cluster_layout(&self.garage, req).await,
-			Endpoint::RevertClusterLayout => handle_revert_cluster_layout(&self.garage).await,
-			// Keys
-			Endpoint::ListKeys => handle_list_keys(&self.garage).await,
-			Endpoint::GetKeyInfo {
-				id,
-				search,
-				show_secret_key,
-			} => {
-				let show_secret_key = show_secret_key.map(|x| x == "true").unwrap_or(false);
-				handle_get_key_info(&self.garage, id, search, show_secret_key).await
-			}
-			Endpoint::CreateKey => handle_create_key(&self.garage, req).await,
-			Endpoint::ImportKey => handle_import_key(&self.garage, req).await,
-			Endpoint::UpdateKey { id } => handle_update_key(&self.garage, id, req).await,
-			Endpoint::DeleteKey { id } => handle_delete_key(&self.garage, id).await,
-			// Buckets
-			Endpoint::ListBuckets => handle_list_buckets(&self.garage).await,
-			Endpoint::GetBucketInfo { id, global_alias } => {
-				handle_get_bucket_info(&self.garage, id, global_alias).await
-			}
-			Endpoint::CreateBucket => handle_create_bucket(&self.garage, req).await,
-			Endpoint::DeleteBucket { id } => handle_delete_bucket(&self.garage, id).await,
-			Endpoint::UpdateBucket { id } => handle_update_bucket(&self.garage, id, req).await,
-			// Bucket-key permissions
-			Endpoint::BucketAllowKey => {
-				handle_bucket_change_key_perm(&self.garage, req, true).await
-			}
-			Endpoint::BucketDenyKey => {
-				handle_bucket_change_key_perm(&self.garage, req, false).await
-			}
-			// Bucket aliasing
-			Endpoint::GlobalAliasBucket { id, alias } => {
-				handle_global_alias_bucket(&self.garage, id, alias).await
-			}
-			Endpoint::GlobalUnaliasBucket { id, alias } => {
-				handle_global_unalias_bucket(&self.garage, id, alias).await
-			}
-			Endpoint::LocalAliasBucket {
-				id,
-				access_key_id,
-				alias,
-			} => handle_local_alias_bucket(&self.garage, id, access_key_id, alias).await,
-			Endpoint::LocalUnaliasBucket {
-				id,
-				access_key_id,
-				alias,
-			} => handle_local_unalias_bucket(&self.garage, id, access_key_id, alias).await,
-		}
+		self.0.handle_http_api(req, endpoint).await
 	}
 }
 
-impl ApiEndpoint for Endpoint {
-	fn name(&self) -> &'static str {
-		Endpoint::name(self)
+impl ApiEndpoint for HttpEndpoint {
+	fn name(&self) -> Cow<'static, str> {
+		match self {
+			Self::Old(endpoint_v1) => Cow::Borrowed(endpoint_v1.name()),
+			Self::New(path) => Cow::Owned(path.clone()),
+		}
 	}
 
 	fn add_span_attributes(&self, _span: SpanRef<'_>) {}
@@ -347,20 +244,87 @@ fn hash_bearer_token(token: &str) -> String {
 		.to_string()
 }
 
-fn verify_bearer_token(token: &hyper::http::HeaderValue, password_hash: &str) -> Result<(), Error> {
-	use argon2::{password_hash::PasswordVerifier, Argon2};
+fn verify_authorization(
+	garage: &Garage,
+	global_token_hash: Option<&str>,
+	auth_header: Option<hyper::http::HeaderValue>,
+	endpoint_name: &str,
+) -> Result<(), Error> {
+	use argon2::{password_hash::PasswordHash, password_hash::PasswordVerifier, Argon2};
 
-	let parsed_hash = PasswordHash::new(&password_hash).unwrap();
+	let invalid_msg = "Invalid bearer token";
 
-	token
+	let token = match &auth_header {
+		None => {
+			return Err(Error::forbidden(
+				"Bearer token must be provided in Authorization header",
+			))
+		}
+		Some(authorization) => authorization
 			.to_str()?
 			.strip_prefix("Bearer ")
-		.and_then(|token| {
+			.ok_or_else(|| Error::forbidden("Invalid Authorization header"))?
+			.trim(),
+	};
+
+	let token_hash_string = if let Some((prefix, _)) = token.split_once('.') {
+		garage
+			.admin_token_table
+			.get_local(&EmptyKey, &prefix.to_string())?
+			.and_then(|k| k.state.into_option())
+			.filter(|p| !p.is_expired(now_msec()))
+			.filter(|p| p.has_scope(endpoint_name))
+			.ok_or_else(|| Error::forbidden(invalid_msg))?
+			.token_hash
+	} else {
+		global_token_hash
+			.ok_or_else(|| Error::forbidden(invalid_msg))?
+			.to_string()
+	};
+
+	let token_hash =
+		PasswordHash::new(&token_hash_string).ok_or_internal_error("Could not parse token hash")?;
+
 	Argon2::default()
-				.verify_password(token.trim().as_bytes(), &parsed_hash)
-				.ok()
-		})
-		.ok_or_else(|| Error::forbidden("Invalid authorization token"))?;
+		.verify_password(token.as_bytes(), &token_hash)
+		.map_err(|_| Error::forbidden(invalid_msg))?;
 
 	Ok(())
 }
+
+pub(crate) fn find_matching_nodes(garage: &Garage, spec: &str) -> Result<Vec<Uuid>, Error> {
+	let mut res = vec![];
+	if spec == "*" {
+		res = garage.system.cluster_layout().all_nodes().to_vec();
+		for node in garage.system.get_known_nodes() {
+			if node.is_up && !res.contains(&node.id) {
+				res.push(node.id);
+			}
+		}
+	} else if spec == "self" {
+		res.push(garage.system.id);
+	} else {
+		let layout = garage.system.cluster_layout();
+		let known_nodes = garage.system.get_known_nodes();
+		let all_nodes = layout
+			.all_nodes()
+			.iter()
+			.copied()
+			.chain(known_nodes.iter().filter(|x| x.is_up).map(|x| x.id));
+		for node in all_nodes {
+			if !res.contains(&node) && hex::encode(node).starts_with(spec) {
+				res.push(node);
+			}
+		}
+		if res.is_empty() {
+			return Err(Error::bad_request(format!("No nodes matching {}", spec)));
+		}
+		if res.len() > 1 {
+			return Err(Error::bad_request(format!(
+				"Multiple nodes matching {}: {:?}",
+				spec, res
+			)));
+		}
+	}
+	Ok(res)
+}

src/api/admin/block.rs

@@ -0,0 +1,276 @@
+use std::sync::Arc;
+
+use garage_util::data::*;
+use garage_util::error::Error as GarageError;
+use garage_util::time::now_msec;
+
+use garage_table::EmptyKey;
+
+use garage_model::garage::Garage;
+use garage_model::s3::object_table::*;
+use garage_model::s3::version_table::*;
+
+use garage_api_common::common_error::CommonErrorDerivative;
+
+use crate::api::*;
+use crate::error::*;
+use crate::{Admin, RequestHandler};
+
+impl RequestHandler for LocalListBlockErrorsRequest {
+	type Response = LocalListBlockErrorsResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<LocalListBlockErrorsResponse, Error> {
+		let errors = garage.block_manager.list_resync_errors()?;
+		let now = now_msec();
+		let errors = errors
+			.into_iter()
+			.map(|e| BlockError {
+				block_hash: hex::encode(&e.hash),
+				refcount: e.refcount,
+				error_count: e.error_count,
+				last_try_secs_ago: now.saturating_sub(e.last_try) / 1000,
+				next_try_in_secs: e.next_try.saturating_sub(now) / 1000,
+			})
+			.collect();
+		Ok(LocalListBlockErrorsResponse(errors))
+	}
+}
+
+impl RequestHandler for LocalGetBlockInfoRequest {
+	type Response = LocalGetBlockInfoResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<LocalGetBlockInfoResponse, Error> {
+		let hash = find_block_hash_by_prefix(garage, &self.block_hash)?;
+		let refcount = garage.block_manager.get_block_rc(&hash)?;
+		let block_refs = garage
+			.block_ref_table
+			.get_range(&hash, None, None, 10000, Default::default())
+			.await?;
+		let mut versions = vec![];
+		for br in block_refs {
+			if let Some(v) = garage.version_table.get(&br.version, &EmptyKey).await? {
+				let bl = match &v.backlink {
+					VersionBacklink::MultipartUpload { upload_id } => {
+						if let Some(u) = garage.mpu_table.get(upload_id, &EmptyKey).await? {
+							BlockVersionBacklink::Upload {
+								upload_id: hex::encode(&upload_id),
+								upload_deleted: u.deleted.get(),
+								upload_garbage_collected: false,
+								bucket_id: Some(hex::encode(&u.bucket_id)),
+								key: Some(u.key.to_string()),
+							}
+						} else {
+							BlockVersionBacklink::Upload {
+								upload_id: hex::encode(&upload_id),
+								upload_deleted: true,
+								upload_garbage_collected: true,
+								bucket_id: None,
+								key: None,
+							}
+						}
+					}
+					VersionBacklink::Object { bucket_id, key } => BlockVersionBacklink::Object {
+						bucket_id: hex::encode(&bucket_id),
+						key: key.to_string(),
+					},
+				};
+				versions.push(BlockVersion {
+					version_id: hex::encode(&br.version),
+					ref_deleted: br.deleted.get(),
+					version_deleted: v.deleted.get(),
+					garbage_collected: false,
+					backlink: Some(bl),
+				});
+			} else {
+				versions.push(BlockVersion {
+					version_id: hex::encode(&br.version),
+					ref_deleted: br.deleted.get(),
+					version_deleted: true,
+					garbage_collected: true,
+					backlink: None,
+				});
+			}
+		}
+		Ok(LocalGetBlockInfoResponse {
+			block_hash: hex::encode(&hash),
+			refcount,
+			versions,
+		})
+	}
+}
+
+impl RequestHandler for LocalRetryBlockResyncRequest {
+	type Response = LocalRetryBlockResyncResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<LocalRetryBlockResyncResponse, Error> {
+		match self {
+			Self::All { all: true } => {
+				let blocks = garage.block_manager.list_resync_errors()?;
+				for b in blocks.iter() {
+					garage.block_manager.resync.clear_backoff(&b.hash)?;
+				}
+				Ok(LocalRetryBlockResyncResponse {
+					count: blocks.len() as u64,
+				})
+			}
+			Self::All { all: false } => Err(Error::bad_request("nonsense")),
+			Self::Blocks { block_hashes } => {
+				for hash in block_hashes.iter() {
+					let hash = hex::decode(hash).ok_or_bad_request("invalid hash")?;
+					let hash = Hash::try_from(&hash).ok_or_bad_request("invalid hash")?;
+					garage.block_manager.resync.clear_backoff(&hash)?;
+				}
+				Ok(LocalRetryBlockResyncResponse {
+					count: block_hashes.len() as u64,
+				})
+			}
+		}
+	}
+}
+
+impl RequestHandler for LocalPurgeBlocksRequest {
+	type Response = LocalPurgeBlocksResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<LocalPurgeBlocksResponse, Error> {
+		let mut obj_dels = 0;
+		let mut mpu_dels = 0;
+		let mut ver_dels = 0;
+
+		for hash in self.0.iter() {
+			let hash = hex::decode(hash).ok_or_bad_request("invalid hash")?;
+			let hash = Hash::try_from(&hash).ok_or_bad_request("invalid hash")?;
+			let block_refs = garage
+				.block_ref_table
+				.get_range(&hash, None, None, 10000, Default::default())
+				.await?;
+
+			for br in block_refs {
+				if let Some(version) = garage.version_table.get(&br.version, &EmptyKey).await? {
+					handle_block_purge_version_backlink(
+						garage,
+						&version,
+						&mut obj_dels,
+						&mut mpu_dels,
+					)
+					.await?;
+
+					if !version.deleted.get() {
+						let deleted_version = Version::new(version.uuid, version.backlink, true);
+						garage.version_table.insert(&deleted_version).await?;
+						ver_dels += 1;
+					}
+				}
+			}
+		}
+
+		Ok(LocalPurgeBlocksResponse {
+			blocks_purged: self.0.len() as u64,
+			versions_deleted: ver_dels,
+			objects_deleted: obj_dels,
+			uploads_deleted: mpu_dels,
+		})
+	}
+}
+
+fn find_block_hash_by_prefix(garage: &Arc<Garage>, prefix: &str) -> Result<Hash, Error> {
+	if prefix.len() < 4 {
+		return Err(Error::bad_request(
+			"Please specify at least 4 characters of the block hash",
+		));
+	}
+
+	let prefix_bin = hex::decode(&prefix[..prefix.len() & !1]).ok_or_bad_request("invalid hash")?;
+
+	let iter = garage
+		.block_ref_table
+		.data
+		.store
+		.range(&prefix_bin[..]..)
+		.map_err(GarageError::from)?;
+	let mut found = None;
+	for item in iter {
+		let (k, _v) = item.map_err(GarageError::from)?;
+		let hash = Hash::try_from(&k[..32]).unwrap();
+		if &hash.as_slice()[..prefix_bin.len()] != prefix_bin {
+			break;
+		}
+		if hex::encode(hash.as_slice()).starts_with(prefix) {
+			match &found {
+				Some(x) if *x == hash => (),
+				Some(_) => {
+					return Err(Error::bad_request(format!(
+						"Several blocks match prefix `{}`",
+						prefix
+					)));
+				}
+				None => {
+					found = Some(hash);
+				}
+			}
+		}
+	}
+
+	found.ok_or_else(|| Error::NoSuchBlock(prefix.to_string()))
+}
+
+async fn handle_block_purge_version_backlink(
+	garage: &Arc<Garage>,
+	version: &Version,
+	obj_dels: &mut u64,
+	mpu_dels: &mut u64,
+) -> Result<(), Error> {
+	let (bucket_id, key, ov_id) = match &version.backlink {
+		VersionBacklink::Object { bucket_id, key } => (*bucket_id, key.clone(), version.uuid),
+		VersionBacklink::MultipartUpload { upload_id } => {
+			if let Some(mut mpu) = garage.mpu_table.get(upload_id, &EmptyKey).await? {
+				if !mpu.deleted.get() {
+					mpu.parts.clear();
+					mpu.deleted.set();
+					garage.mpu_table.insert(&mpu).await?;
+					*mpu_dels += 1;
+				}
+				(mpu.bucket_id, mpu.key.clone(), *upload_id)
+			} else {
+				return Ok(());
+			}
+		}
+	};
+
+	if let Some(object) = garage.object_table.get(&bucket_id, &key).await? {
+		let ov = object.versions().iter().rev().find(|v| v.is_complete());
+		if let Some(ov) = ov {
+			if ov.uuid == ov_id {
+				let del_uuid = gen_uuid();
+				let deleted_object = Object::new(
+					bucket_id,
+					key,
+					vec![ObjectVersion {
+						uuid: del_uuid,
+						timestamp: ov.timestamp + 1,
+						state: ObjectVersionState::Complete(ObjectVersionData::DeleteMarker),
+					}],
+				);
+				garage.object_table.insert(&deleted_object).await?;
+				*obj_dels += 1;
+			}
+		}
+	}
+
+	Ok(())
+}

src/api/admin/cluster.rs

@@ -1,23 +1,28 @@
 use std::collections::HashMap;
-use std::net::SocketAddr;
+use std::fmt::Write;
 use std::sync::Arc;
 
-use hyper::{body::Incoming as IncomingBody, Request, Response};
-use serde::{Deserialize, Serialize};
+use format_table::format_table_to_string;
 
-use garage_util::crdt::*;
 use garage_util::data::*;
 
 use garage_rpc::layout;
+use garage_rpc::layout::PARTITION_BITS;
 
 use garage_model::garage::Garage;
 
-use garage_api_common::helpers::{json_ok_response, parse_json_body};
-
-use crate::api_server::ResBody;
+use crate::api::*;
 use crate::error::*;
+use crate::{Admin, RequestHandler};
 
-pub async fn handle_get_cluster_status(garage: &Arc<Garage>) -> Result<Response<ResBody>, Error> {
+impl RequestHandler for GetClusterStatusRequest {
+	type Response = GetClusterStatusResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<GetClusterStatusResponse, Error> {
 		let layout = garage.system.cluster_layout();
 		let mut nodes = garage
 			.system
@@ -28,16 +33,16 @@ pub async fn handle_get_cluster_status(garage: &Arc<Garage>) -> Result<Response<
 					i.id,
 					NodeResp {
 						id: hex::encode(i.id),
+						garage_version: i.status.garage_version,
 						addr: i.addr,
 						hostname: i.status.hostname,
 						is_up: i.is_up,
 						last_seen_secs_ago: i.last_seen_secs_ago,
-					data_partition: i
-						.status
-						.data_disk_avail
-						.map(|(avail, total)| FreeSpaceResp {
+						data_partition: i.status.data_disk_avail.map(|(avail, total)| {
+							FreeSpaceResp {
 								available: avail,
 								total,
+							}
 						}),
 						metadata_partition: i.status.meta_disk_avail.map(|(avail, total)| {
 							FreeSpaceResp {
@@ -53,8 +58,7 @@ pub async fn handle_get_cluster_status(garage: &Arc<Garage>) -> Result<Response<
 
 		for (id, _, role) in layout.current().roles.items().iter() {
 			if let layout::NodeRoleV(Some(r)) = role {
-			let role = NodeRoleResp {
-				id: hex::encode(id),
+				let role = NodeAssignedRole {
 					zone: r.zone.to_string(),
 					capacity: r.capacity,
 					tags: r.tags.clone(),
@@ -103,28 +107,30 @@ pub async fn handle_get_cluster_status(garage: &Arc<Garage>) -> Result<Response<
 		let mut nodes = nodes.into_values().collect::<Vec<_>>();
 		nodes.sort_by(|x, y| x.id.cmp(&y.id));
 
-	let res = GetClusterStatusResponse {
-		node: hex::encode(garage.system.id),
-		garage_version: garage_util::version::garage_version(),
-		garage_features: garage_util::version::garage_features(),
-		rust_version: garage_util::version::rust_version(),
-		db_engine: garage.db.engine(),
+		Ok(GetClusterStatusResponse {
 			layout_version: layout.current().version,
 			nodes,
-	};
-
-	Ok(json_ok_response(&res)?)
+		})
+	}
 }
 
-pub async fn handle_get_cluster_health(garage: &Arc<Garage>) -> Result<Response<ResBody>, Error> {
+impl RequestHandler for GetClusterHealthRequest {
+	type Response = GetClusterHealthResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<GetClusterHealthResponse, Error> {
 		use garage_rpc::system::ClusterHealthStatus;
 		let health = garage.system.health();
-	let health = ClusterHealth {
+		let health = GetClusterHealthResponse {
 			status: match health.status {
 				ClusterHealthStatus::Healthy => "healthy",
 				ClusterHealthStatus::Degraded => "degraded",
 				ClusterHealthStatus::Unavailable => "unavailable",
-		},
+			}
+			.to_string(),
 			known_nodes: health.known_nodes,
 			connected_nodes: health.connected_nodes,
 			storage_nodes: health.storage_nodes,
@@ -133,279 +139,138 @@ pub async fn handle_get_cluster_health(garage: &Arc<Garage>) -> Result<Response<
 			partitions_quorum: health.partitions_quorum,
 			partitions_all_ok: health.partitions_all_ok,
 		};
-	Ok(json_ok_response(&health)?)
+		Ok(health)
+	}
 }
 
-pub async fn handle_connect_cluster_nodes(
-	garage: &Arc<Garage>,
-	req: Request<IncomingBody>,
-) -> Result<Response<ResBody>, Error> {
-	let req = parse_json_body::<Vec<String>, _, Error>(req).await?;
+impl RequestHandler for GetClusterStatisticsRequest {
+	type Response = GetClusterStatisticsResponse;
 
-	let res = futures::future::join_all(req.iter().map(|node| garage.system.connect(node)))
+	// FIXME: return this as a JSON struct instead of text
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<GetClusterStatisticsResponse, Error> {
+		let mut ret = String::new();
+
+		// Gather storage node and free space statistics for current nodes
+		let layout = &garage.system.cluster_layout();
+		let mut node_partition_count = HashMap::<Uuid, u64>::new();
+		for short_id in layout.current().ring_assignment_data.iter() {
+			let id = layout.current().node_id_vec[*short_id as usize];
+			*node_partition_count.entry(id).or_default() += 1;
+		}
+		let node_info = garage
+			.system
+			.get_known_nodes()
+			.into_iter()
+			.map(|n| (n.id, n))
+			.collect::<HashMap<_, _>>();
+
+		let mut table = vec!["  ID\tHostname\tZone\tCapacity\tPart.\tDataAvail\tMetaAvail".into()];
+		for (id, parts) in node_partition_count.iter() {
+			let info = node_info.get(id);
+			let status = info.map(|x| &x.status);
+			let role = layout.current().roles.get(id).and_then(|x| x.0.as_ref());
+			let hostname = status.and_then(|x| x.hostname.as_deref()).unwrap_or("?");
+			let zone = role.map(|x| x.zone.as_str()).unwrap_or("?");
+			let capacity = role
+				.map(|x| x.capacity_string())
+				.unwrap_or_else(|| "?".into());
+			let avail_str = |x| match x {
+				Some((avail, total)) => {
+					let pct = (avail as f64) / (total as f64) * 100.;
+					let avail = bytesize::ByteSize::b(avail);
+					let total = bytesize::ByteSize::b(total);
+					format!("{}/{} ({:.1}%)", avail, total, pct)
+				}
+				None => "?".into(),
+			};
+			let data_avail = avail_str(status.and_then(|x| x.data_disk_avail));
+			let meta_avail = avail_str(status.and_then(|x| x.meta_disk_avail));
+			table.push(format!(
+				"  {:?}\t{}\t{}\t{}\t{}\t{}\t{}",
+				id, hostname, zone, capacity, parts, data_avail, meta_avail
+			));
+		}
+		write!(
+			&mut ret,
+			"Storage nodes:\n{}",
+			format_table_to_string(table)
+		)
+		.unwrap();
+
+		let meta_part_avail = node_partition_count
+			.iter()
+			.filter_map(|(id, parts)| {
+				node_info
+					.get(id)
+					.and_then(|x| x.status.meta_disk_avail)
+					.map(|c| c.0 / *parts)
+			})
+			.collect::<Vec<_>>();
+		let data_part_avail = node_partition_count
+			.iter()
+			.filter_map(|(id, parts)| {
+				node_info
+					.get(id)
+					.and_then(|x| x.status.data_disk_avail)
+					.map(|c| c.0 / *parts)
+			})
+			.collect::<Vec<_>>();
+		if !meta_part_avail.is_empty() && !data_part_avail.is_empty() {
+			let meta_avail =
+				bytesize::ByteSize(meta_part_avail.iter().min().unwrap() * (1 << PARTITION_BITS));
+			let data_avail =
+				bytesize::ByteSize(data_part_avail.iter().min().unwrap() * (1 << PARTITION_BITS));
+			writeln!(
+				&mut ret,
+				"\nEstimated available storage space cluster-wide (might be lower in practice):"
+			)
+			.unwrap();
+			if meta_part_avail.len() < node_partition_count.len()
+				|| data_part_avail.len() < node_partition_count.len()
+			{
+				ret += &format_table_to_string(vec![
+					format!("  data: < {}", data_avail),
+					format!("  metadata: < {}", meta_avail),
+				]);
+				writeln!(&mut ret, "A precise estimate could not be given as information is missing for some storage nodes.").unwrap();
+			} else {
+				ret += &format_table_to_string(vec![
+					format!("  data: {}", data_avail),
+					format!("  metadata: {}", meta_avail),
+				]);
+			}
+		}
+
+		Ok(GetClusterStatisticsResponse { freeform: ret })
+	}
+}
+
+impl RequestHandler for ConnectClusterNodesRequest {
+	type Response = ConnectClusterNodesResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<ConnectClusterNodesResponse, Error> {
+		let res = futures::future::join_all(self.0.iter().map(|node| garage.system.connect(node)))
 			.await
 			.into_iter()
 			.map(|r| match r {
-			Ok(()) => ConnectClusterNodesResponse {
+				Ok(()) => ConnectNodeResponse {
 					success: true,
 					error: None,
 				},
-			Err(e) => ConnectClusterNodesResponse {
+				Err(e) => ConnectNodeResponse {
 					success: false,
 					error: Some(format!("{}", e)),
 				},
 			})
 			.collect::<Vec<_>>();
-
-	Ok(json_ok_response(&res)?)
-}
-
-pub async fn handle_get_cluster_layout(garage: &Arc<Garage>) -> Result<Response<ResBody>, Error> {
-	let res = format_cluster_layout(garage.system.cluster_layout().inner());
-
-	Ok(json_ok_response(&res)?)
-}
-
-fn format_cluster_layout(layout: &layout::LayoutHistory) -> GetClusterLayoutResponse {
-	let roles = layout
-		.current()
-		.roles
-		.items()
-		.iter()
-		.filter_map(|(k, _, v)| v.0.clone().map(|x| (k, x)))
-		.map(|(k, v)| NodeRoleResp {
-			id: hex::encode(k),
-			zone: v.zone.clone(),
-			capacity: v.capacity,
-			tags: v.tags.clone(),
-		})
-		.collect::<Vec<_>>();
-
-	let staged_role_changes = layout
-		.staging
-		.get()
-		.roles
-		.items()
-		.iter()
-		.filter(|(k, _, v)| layout.current().roles.get(k) != Some(v))
-		.map(|(k, _, v)| match &v.0 {
-			None => NodeRoleChange {
-				id: hex::encode(k),
-				action: NodeRoleChangeEnum::Remove { remove: true },
-			},
-			Some(r) => NodeRoleChange {
-				id: hex::encode(k),
-				action: NodeRoleChangeEnum::Update {
-					zone: r.zone.clone(),
-					capacity: r.capacity,
-					tags: r.tags.clone(),
-				},
-			},
-		})
-		.collect::<Vec<_>>();
-
-	GetClusterLayoutResponse {
-		version: layout.current().version,
-		roles,
-		staged_role_changes,
+		Ok(ConnectClusterNodesResponse(res))
 	}
 }
-
-// ----
-
-#[derive(Debug, Clone, Copy, Serialize)]
-#[serde(rename_all = "camelCase")]
-pub struct ClusterHealth {
-	status: &'static str,
-	known_nodes: usize,
-	connected_nodes: usize,
-	storage_nodes: usize,
-	storage_nodes_ok: usize,
-	partitions: usize,
-	partitions_quorum: usize,
-	partitions_all_ok: usize,
-}
-
-#[derive(Serialize)]
-#[serde(rename_all = "camelCase")]
-struct GetClusterStatusResponse {
-	node: String,
-	garage_version: &'static str,
-	garage_features: Option<&'static [&'static str]>,
-	rust_version: &'static str,
-	db_engine: String,
-	layout_version: u64,
-	nodes: Vec<NodeResp>,
-}
-
-#[derive(Serialize)]
-#[serde(rename_all = "camelCase")]
-struct ApplyClusterLayoutResponse {
-	message: Vec<String>,
-	layout: GetClusterLayoutResponse,
-}
-
-#[derive(Serialize)]
-#[serde(rename_all = "camelCase")]
-struct ConnectClusterNodesResponse {
-	success: bool,
-	error: Option<String>,
-}
-
-#[derive(Serialize)]
-#[serde(rename_all = "camelCase")]
-struct GetClusterLayoutResponse {
-	version: u64,
-	roles: Vec<NodeRoleResp>,
-	staged_role_changes: Vec<NodeRoleChange>,
-}
-
-#[derive(Serialize)]
-#[serde(rename_all = "camelCase")]
-struct NodeRoleResp {
-	id: String,
-	zone: String,
-	capacity: Option<u64>,
-	tags: Vec<String>,
-}
-
-#[derive(Serialize, Default)]
-#[serde(rename_all = "camelCase")]
-struct FreeSpaceResp {
-	available: u64,
-	total: u64,
-}
-
-#[derive(Serialize, Default)]
-#[serde(rename_all = "camelCase")]
-struct NodeResp {
-	id: String,
-	role: Option<NodeRoleResp>,
-	addr: Option<SocketAddr>,
-	hostname: Option<String>,
-	is_up: bool,
-	last_seen_secs_ago: Option<u64>,
-	draining: bool,
-	#[serde(skip_serializing_if = "Option::is_none")]
-	data_partition: Option<FreeSpaceResp>,
-	#[serde(skip_serializing_if = "Option::is_none")]
-	metadata_partition: Option<FreeSpaceResp>,
-}
-
-// ---- update functions ----
-
-pub async fn handle_update_cluster_layout(
-	garage: &Arc<Garage>,
-	req: Request<IncomingBody>,
-) -> Result<Response<ResBody>, Error> {
-	let updates = parse_json_body::<UpdateClusterLayoutRequest, _, Error>(req).await?;
-
-	let mut layout = garage.system.cluster_layout().inner().clone();
-
-	let mut roles = layout.current().roles.clone();
-	roles.merge(&layout.staging.get().roles);
-
-	for change in updates {
-		let node = hex::decode(&change.id).ok_or_bad_request("Invalid node identifier")?;
-		let node = Uuid::try_from(&node).ok_or_bad_request("Invalid node identifier")?;
-
-		let new_role = match change.action {
-			NodeRoleChangeEnum::Remove { remove: true } => None,
-			NodeRoleChangeEnum::Update {
-				zone,
-				capacity,
-				tags,
-			} => Some(layout::NodeRole {
-				zone,
-				capacity,
-				tags,
-			}),
-			_ => return Err(Error::bad_request("Invalid layout change")),
-		};
-
-		layout
-			.staging
-			.get_mut()
-			.roles
-			.merge(&roles.update_mutator(node, layout::NodeRoleV(new_role)));
-	}
-
-	garage
-		.system
-		.layout_manager
-		.update_cluster_layout(&layout)
-		.await?;
-
-	let res = format_cluster_layout(&layout);
-	Ok(json_ok_response(&res)?)
-}
-
-pub async fn handle_apply_cluster_layout(
-	garage: &Arc<Garage>,
-	req: Request<IncomingBody>,
-) -> Result<Response<ResBody>, Error> {
-	let param = parse_json_body::<ApplyLayoutRequest, _, Error>(req).await?;
-
-	let layout = garage.system.cluster_layout().inner().clone();
-	let (layout, msg) = layout.apply_staged_changes(Some(param.version))?;
-
-	garage
-		.system
-		.layout_manager
-		.update_cluster_layout(&layout)
-		.await?;
-
-	let res = ApplyClusterLayoutResponse {
-		message: msg,
-		layout: format_cluster_layout(&layout),
-	};
-	Ok(json_ok_response(&res)?)
-}
-
-pub async fn handle_revert_cluster_layout(
-	garage: &Arc<Garage>,
-) -> Result<Response<ResBody>, Error> {
-	let layout = garage.system.cluster_layout().inner().clone();
-	let layout = layout.revert_staged_changes()?;
-	garage
-		.system
-		.layout_manager
-		.update_cluster_layout(&layout)
-		.await?;
-
-	let res = format_cluster_layout(&layout);
-	Ok(json_ok_response(&res)?)
-}
-
-// ----
-
-type UpdateClusterLayoutRequest = Vec<NodeRoleChange>;
-
-#[derive(Deserialize)]
-#[serde(rename_all = "camelCase")]
-struct ApplyLayoutRequest {
-	version: u64,
-}
-
-// ----
-
-#[derive(Serialize, Deserialize)]
-#[serde(rename_all = "camelCase")]
-struct NodeRoleChange {
-	id: String,
-	#[serde(flatten)]
-	action: NodeRoleChangeEnum,
-}
-
-#[derive(Serialize, Deserialize)]
-#[serde(untagged)]
-enum NodeRoleChangeEnum {
-	#[serde(rename_all = "camelCase")]
-	Remove { remove: bool },
-	#[serde(rename_all = "camelCase")]
-	Update {
-		zone: String,
-		capacity: Option<u64>,
-		tags: Vec<String>,
-	},
-}

src/api/admin/error.rs

@@ -1,8 +1,8 @@
 use std::convert::TryFrom;
 
+use err_derive::Error;
 use hyper::header::HeaderValue;
 use hyper::{HeaderMap, StatusCode};
-use thiserror::Error;
 
 pub use garage_model::helper::error::Error as HelperError;
 
@@ -16,17 +16,36 @@ use garage_api_common::helpers::*;
 /// Errors of this crate
 #[derive(Debug, Error)]
 pub enum Error {
-	#[error("{0}")]
+	#[error(display = "{}", _0)]
 	/// Error from common error
-	Common(#[from] CommonError),
+	Common(#[error(source)] CommonError),
 
 	// Category: cannot process
+	/// The admin API token does not exist
+	#[error(display = "Admin token not found: {}", _0)]
+	NoSuchAdminToken(String),
+
 	/// The API access key does not exist
-	#[error("Access key not found: {0}")]
+	#[error(display = "Access key not found: {}", _0)]
 	NoSuchAccessKey(String),
 
+	/// The requested block does not exist
+	#[error(display = "Block not found: {}", _0)]
+	NoSuchBlock(String),
+
+	/// The requested worker does not exist
+	#[error(display = "Worker not found: {}", _0)]
+	NoSuchWorker(u64),
+
+	/// The object requested don't exists
+	#[error(display = "Key not found")]
+	NoSuchKey,
+
 	/// In Import key, the key already exists
-	#[error("Key {0} already exists in data store. Even if it is deleted, we can't let you create a new key with the same ID. Sorry.")]
+	#[error(
+		display = "Key {} already exists in data store. Even if it is deleted, we can't let you create a new key with the same ID. Sorry.",
+		_0
+	)]
 	KeyAlreadyExists(String),
 }
 
@@ -46,11 +65,15 @@ impl From<HelperError> for Error {
 }
 
 impl Error {
-	fn code(&self) -> &'static str {
+	pub fn code(&self) -> &'static str {
 		match self {
 			Error::Common(c) => c.aws_code(),
+			Error::NoSuchAdminToken(_) => "NoSuchAdminToken",
 			Error::NoSuchAccessKey(_) => "NoSuchAccessKey",
+			Error::NoSuchWorker(_) => "NoSuchWorker",
+			Error::NoSuchBlock(_) => "NoSuchBlock",
 			Error::KeyAlreadyExists(_) => "KeyAlreadyExists",
+			Error::NoSuchKey => "NoSuchKey",
 		}
 	}
 }
@@ -60,7 +83,11 @@ impl ApiError for Error {
 	fn http_status_code(&self) -> StatusCode {
 		match self {
 			Error::Common(c) => c.http_status_code(),
-			Error::NoSuchAccessKey(_) => StatusCode::NOT_FOUND,
+			Error::NoSuchAdminToken(_)
+			| Error::NoSuchAccessKey(_)
+			| Error::NoSuchWorker(_)
+			| Error::NoSuchBlock(_)
+			| Error::NoSuchKey => StatusCode::NOT_FOUND,
 			Error::KeyAlreadyExists(_) => StatusCode::CONFLICT,
 		}
 	}

src/api/admin/key.rs

@@ -1,20 +1,24 @@
 use std::collections::HashMap;
 use std::sync::Arc;
 
-use hyper::{body::Incoming as IncomingBody, Request, Response, StatusCode};
-use serde::{Deserialize, Serialize};
+use chrono::DateTime;
 
 use garage_table::*;
+use garage_util::time::now_msec;
 
 use garage_model::garage::Garage;
 use garage_model::key_table::*;
 
-use garage_api_common::helpers::*;
-
-use crate::api_server::ResBody;
+use crate::api::*;
 use crate::error::*;
+use crate::{Admin, RequestHandler};
+
+impl RequestHandler for ListKeysRequest {
+	type Response = ListKeysResponse;
+
+	async fn handle(self, garage: &Arc<Garage>, _admin: &Admin) -> Result<ListKeysResponse, Error> {
+		let now = now_msec();
 
-pub async fn handle_list_keys(garage: &Arc<Garage>) -> Result<Response<ResBody>, Error> {
 		let res = garage
 			.key_table
 			.get_range(
@@ -26,148 +30,161 @@ pub async fn handle_list_keys(garage: &Arc<Garage>) -> Result<Response<ResBody>,
 			)
 			.await?
 			.iter()
-		.map(|k| ListKeyResultItem {
+			.map(|k| {
+				let p = k.params().unwrap();
+
+				ListKeysResponseItem {
 					id: k.key_id.to_string(),
-			name: k.params().unwrap().name.get().clone(),
+					name: p.name.get().clone(),
+					created: p.created.map(|x| {
+						DateTime::from_timestamp_millis(x as i64)
+							.expect("invalid timestamp stored in db")
+					}),
+					expiration: p.expiration.get().map(|x| {
+						DateTime::from_timestamp_millis(x as i64)
+							.expect("invalid timestamp stored in db")
+					}),
+					expired: p.is_expired(now),
+				}
 			})
 			.collect::<Vec<_>>();
 
-	Ok(json_ok_response(&res)?)
+		Ok(ListKeysResponse(res))
+	}
 }
 
-#[derive(Serialize)]
-#[serde(rename_all = "camelCase")]
-struct ListKeyResultItem {
-	id: String,
-	name: String,
-}
+impl RequestHandler for GetKeyInfoRequest {
+	type Response = GetKeyInfoResponse;
 
-pub async fn handle_get_key_info(
+	async fn handle(
+		self,
 		garage: &Arc<Garage>,
-	id: Option<String>,
-	search: Option<String>,
-	show_secret_key: bool,
-) -> Result<Response<ResBody>, Error> {
-	let key = if let Some(id) = id {
-		garage.key_helper().get_existing_key(&id).await?
-	} else if let Some(search) = search {
-		garage
-			.key_helper()
-			.get_existing_matching_key(&search)
+		_admin: &Admin,
+	) -> Result<GetKeyInfoResponse, Error> {
+		let key = match (self.id, self.search) {
+			(Some(id), None) => garage.key_helper().get_existing_key(&id).await?,
+			(None, Some(search)) => {
+				let candidates = garage
+					.key_table
+					.get_range(
+						&EmptyKey,
+						None,
+						Some(KeyFilter::MatchesAndNotDeleted(search.to_string())),
+						10,
+						EnumerationOrder::Forward,
+					)
 					.await?
-	} else {
-		unreachable!();
+					.into_iter()
+					.collect::<Vec<_>>();
+				if candidates.len() != 1 {
+					return Err(Error::bad_request(format!(
+						"{} matching keys",
+						candidates.len()
+					)));
+				}
+				candidates.into_iter().next().unwrap()
+			}
+			_ => {
+				return Err(Error::bad_request(
+					"Either id or search must be provided (but not both)",
+				));
+			}
 		};
 
-	key_info_results(garage, key, show_secret_key).await
+		Ok(key_info_results(garage, key, self.show_secret_key).await?)
+	}
 }
 
-pub async fn handle_create_key(
-	garage: &Arc<Garage>,
-	req: Request<IncomingBody>,
-) -> Result<Response<ResBody>, Error> {
-	let req = parse_json_body::<CreateKeyRequest, _, Error>(req).await?;
+impl RequestHandler for CreateKeyRequest {
+	type Response = CreateKeyResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<CreateKeyResponse, Error> {
+		let mut key = Key::new("Unnamed key");
+
+		apply_key_updates(&mut key, self.0)?;
 
-	let key = Key::new(req.name.as_deref().unwrap_or("Unnamed key"));
 		garage.key_table.insert(&key).await?;
 
-	key_info_results(garage, key, true).await
+		Ok(CreateKeyResponse(
+			key_info_results(garage, key, true).await?,
+		))
+	}
 }
 
-#[derive(Deserialize)]
-#[serde(rename_all = "camelCase")]
-struct CreateKeyRequest {
-	name: Option<String>,
-}
+impl RequestHandler for ImportKeyRequest {
+	type Response = ImportKeyResponse;
 
-pub async fn handle_import_key(
+	async fn handle(
+		self,
 		garage: &Arc<Garage>,
-	req: Request<IncomingBody>,
-) -> Result<Response<ResBody>, Error> {
-	let req = parse_json_body::<ImportKeyRequest, _, Error>(req).await?;
-
-	let prev_key = garage.key_table.get(&EmptyKey, &req.access_key_id).await?;
+		_admin: &Admin,
+	) -> Result<ImportKeyResponse, Error> {
+		let prev_key = garage.key_table.get(&EmptyKey, &self.access_key_id).await?;
 		if prev_key.is_some() {
-		return Err(Error::KeyAlreadyExists(req.access_key_id.to_string()));
+			return Err(Error::KeyAlreadyExists(self.access_key_id.to_string()));
 		}
 
 		let imported_key = Key::import(
-		&req.access_key_id,
-		&req.secret_access_key,
-		req.name.as_deref().unwrap_or("Imported key"),
+			&self.access_key_id,
+			&self.secret_access_key,
+			self.name.as_deref().unwrap_or("Imported key"),
 		)
 		.ok_or_bad_request("Invalid key format")?;
 		garage.key_table.insert(&imported_key).await?;
 
-	key_info_results(garage, imported_key, false).await
+		Ok(ImportKeyResponse(
+			key_info_results(garage, imported_key, false).await?,
+		))
+	}
 }
 
-#[derive(Deserialize)]
-#[serde(rename_all = "camelCase")]
-struct ImportKeyRequest {
-	access_key_id: String,
-	secret_access_key: String,
-	name: Option<String>,
-}
+impl RequestHandler for UpdateKeyRequest {
+	type Response = UpdateKeyResponse;
 
-pub async fn handle_update_key(
+	async fn handle(
+		self,
 		garage: &Arc<Garage>,
-	id: String,
-	req: Request<IncomingBody>,
-) -> Result<Response<ResBody>, Error> {
-	let req = parse_json_body::<UpdateKeyRequest, _, Error>(req).await?;
+		_admin: &Admin,
+	) -> Result<UpdateKeyResponse, Error> {
+		let mut key = garage.key_helper().get_existing_key(&self.id).await?;
 
-	let mut key = garage.key_helper().get_existing_key(&id).await?;
-
-	let key_state = key.state.as_option_mut().unwrap();
-
-	if let Some(new_name) = req.name {
-		key_state.name.update(new_name);
-	}
-	if let Some(allow) = req.allow {
-		if allow.create_bucket {
-			key_state.allow_create_bucket.update(true);
-		}
-	}
-	if let Some(deny) = req.deny {
-		if deny.create_bucket {
-			key_state.allow_create_bucket.update(false);
-		}
-	}
+		apply_key_updates(&mut key, self.body)?;
 
 		garage.key_table.insert(&key).await?;
 
-	key_info_results(garage, key, false).await
+		Ok(UpdateKeyResponse(
+			key_info_results(garage, key, false).await?,
+		))
+	}
 }
 
-#[derive(Deserialize)]
-#[serde(rename_all = "camelCase")]
-struct UpdateKeyRequest {
-	name: Option<String>,
-	allow: Option<KeyPerm>,
-	deny: Option<KeyPerm>,
-}
+impl RequestHandler for DeleteKeyRequest {
+	type Response = DeleteKeyResponse;
 
-pub async fn handle_delete_key(
+	async fn handle(
+		self,
 		garage: &Arc<Garage>,
-	id: String,
-) -> Result<Response<ResBody>, Error> {
+		_admin: &Admin,
+	) -> Result<DeleteKeyResponse, Error> {
 		let helper = garage.locked_helper().await;
 
-	let mut key = helper.key().get_existing_key(&id).await?;
+		let mut key = helper.key().get_existing_key(&self.id).await?;
 
 		helper.delete_key(&mut key).await?;
 
-	Ok(Response::builder()
-		.status(StatusCode::NO_CONTENT)
-		.body(empty_body())?)
+		Ok(DeleteKeyResponse)
+	}
 }
 
 async fn key_info_results(
 	garage: &Arc<Garage>,
 	key: Key,
 	show_secret: bool,
-) -> Result<Response<ResBody>, Error> {
+) -> Result<GetKeyInfoResponse, Error> {
 	let mut relevant_buckets = HashMap::new();
 
 	let key_state = key.state.as_option().unwrap();
@@ -193,8 +210,15 @@ async fn key_info_results(
 		}
 	}
 
-	let res = GetKeyInfoResult {
+	let res = GetKeyInfoResponse {
 		name: key_state.name.get().clone(),
+		created: key_state.created.map(|x| {
+			DateTime::from_timestamp_millis(x as i64).expect("invalid timestamp stored in db")
+		}),
+		expiration: key_state.expiration.get().map(|x| {
+			DateTime::from_timestamp_millis(x as i64).expect("invalid timestamp stored in db")
+		}),
+		expired: key_state.is_expired(now_msec()),
 		access_key_id: key.key_id.clone(),
 		secret_access_key: if show_secret {
 			Some(key_state.secret_key.clone())
@@ -208,7 +232,7 @@ async fn key_info_results(
 			.into_values()
 			.map(|bucket| {
 				let state = bucket.state.as_option().unwrap();
-				KeyInfoBucketResult {
+				KeyInfoBucketResponse {
 					id: hex::encode(bucket.id),
 					global_aliases: state
 						.aliases
@@ -238,43 +262,39 @@ async fn key_info_results(
 			.collect::<Vec<_>>(),
 	};
 
-	Ok(json_ok_response(&res)?)
+	Ok(res)
 }
 
-#[derive(Serialize)]
-#[serde(rename_all = "camelCase")]
-struct GetKeyInfoResult {
-	name: String,
-	access_key_id: String,
-	#[serde(skip_serializing_if = "is_default")]
-	secret_access_key: Option<String>,
-	permissions: KeyPerm,
-	buckets: Vec<KeyInfoBucketResult>,
-}
+fn apply_key_updates(key: &mut Key, updates: UpdateKeyRequestBody) -> Result<(), Error> {
+	if updates.never_expires && updates.expiration.is_some() {
+		return Err(Error::bad_request(
+			"cannot specify `expiration` and `never_expires`",
+		));
+	}
 
-#[derive(Serialize, Deserialize)]
-#[serde(rename_all = "camelCase")]
-struct KeyPerm {
-	#[serde(default)]
-	create_bucket: bool,
-}
+	let key_state = key.state.as_option_mut().unwrap();
 
-#[derive(Serialize)]
-#[serde(rename_all = "camelCase")]
-struct KeyInfoBucketResult {
-	id: String,
-	global_aliases: Vec<String>,
-	local_aliases: Vec<String>,
-	permissions: ApiBucketKeyPerm,
-}
+	if let Some(new_name) = updates.name {
+		key_state.name.update(new_name);
+	}
+	if let Some(expiration) = updates.expiration {
+		key_state
+			.expiration
+			.update(Some(expiration.timestamp_millis() as u64));
+	}
+	if updates.never_expires {
+		key_state.expiration.update(None);
+	}
+	if let Some(allow) = updates.allow {
+		if allow.create_bucket {
+			key_state.allow_create_bucket.update(true);
+		}
+	}
+	if let Some(deny) = updates.deny {
+		if deny.create_bucket {
+			key_state.allow_create_bucket.update(false);
+		}
+	}
 
-#[derive(Serialize, Deserialize, Default)]
-#[serde(rename_all = "camelCase")]
-pub(crate) struct ApiBucketKeyPerm {
-	#[serde(default)]
-	pub(crate) read: bool,
-	#[serde(default)]
-	pub(crate) write: bool,
-	#[serde(default)]
-	pub(crate) owner: bool,
+	Ok(())
 }

src/api/admin/layout.rs

@@ -0,0 +1,406 @@
+use std::sync::Arc;
+
+use garage_util::crdt::*;
+use garage_util::data::*;
+use garage_util::error::Error as GarageError;
+
+use garage_rpc::layout;
+
+use garage_model::garage::Garage;
+
+use crate::api::*;
+use crate::error::*;
+use crate::{Admin, RequestHandler};
+
+impl RequestHandler for GetClusterLayoutRequest {
+	type Response = GetClusterLayoutResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<GetClusterLayoutResponse, Error> {
+		Ok(format_cluster_layout(
+			garage.system.cluster_layout().inner(),
+		))
+	}
+}
+
+fn format_cluster_layout(layout: &layout::LayoutHistory) -> GetClusterLayoutResponse {
+	let current = layout.current();
+
+	let roles = current
+		.roles
+		.items()
+		.iter()
+		.filter_map(|(k, _, v)| v.0.clone().map(|x| (k, x)))
+		.map(|(k, v)| {
+			let stored_partitions = current.get_node_usage(k).ok().map(|x| x as u64);
+			LayoutNodeRole {
+				id: hex::encode(k),
+				zone: v.zone.clone(),
+				capacity: v.capacity,
+				stored_partitions,
+				usable_capacity: stored_partitions.map(|x| x * current.partition_size),
+				tags: v.tags.clone(),
+			}
+		})
+		.collect::<Vec<_>>();
+
+	let staged_role_changes = layout
+		.staging
+		.get()
+		.roles
+		.items()
+		.iter()
+		.filter(|(k, _, v)| current.roles.get(k) != Some(v))
+		.map(|(k, _, v)| match &v.0 {
+			None => NodeRoleChange {
+				id: hex::encode(k),
+				action: NodeRoleChangeEnum::Remove { remove: true },
+			},
+			Some(r) => NodeRoleChange {
+				id: hex::encode(k),
+				action: NodeRoleChangeEnum::Update(NodeAssignedRole {
+					zone: r.zone.clone(),
+					capacity: r.capacity,
+					tags: r.tags.clone(),
+				}),
+			},
+		})
+		.collect::<Vec<_>>();
+
+	let staged_parameters = if *layout.staging.get().parameters.get() != current.parameters {
+		Some((*layout.staging.get().parameters.get()).into())
+	} else {
+		None
+	};
+
+	GetClusterLayoutResponse {
+		version: current.version,
+		roles,
+		partition_size: current.partition_size,
+		parameters: current.parameters.into(),
+		staged_role_changes,
+		staged_parameters,
+	}
+}
+
+impl RequestHandler for GetClusterLayoutHistoryRequest {
+	type Response = GetClusterLayoutHistoryResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<GetClusterLayoutHistoryResponse, Error> {
+		let layout_helper = garage.system.cluster_layout();
+		let layout = layout_helper.inner();
+		let min_stored = layout.min_stored();
+
+		let versions = layout
+			.versions
+			.iter()
+			.rev()
+			.chain(layout.old_versions.iter().rev())
+			.map(|ver| {
+				let status = if ver.version == layout.current().version {
+					ClusterLayoutVersionStatus::Current
+				} else if ver.version >= min_stored {
+					ClusterLayoutVersionStatus::Draining
+				} else {
+					ClusterLayoutVersionStatus::Historical
+				};
+				ClusterLayoutVersion {
+					version: ver.version,
+					status,
+					storage_nodes: ver
+						.roles
+						.items()
+						.iter()
+						.filter(
+							|(_, _, x)| matches!(x, layout::NodeRoleV(Some(c)) if c.capacity.is_some()),
+						)
+						.count() as u64,
+					gateway_nodes: ver
+						.roles
+						.items()
+						.iter()
+						.filter(
+							|(_, _, x)| matches!(x, layout::NodeRoleV(Some(c)) if c.capacity.is_none()),
+						)
+						.count() as u64,
+				}
+			})
+			.collect::<Vec<_>>();
+
+		let all_nodes = layout.get_all_nodes();
+		let min_ack = layout_helper.ack_map_min();
+
+		let update_trackers = if layout.versions.len() > 1 {
+			Some(
+				all_nodes
+					.iter()
+					.map(|node| {
+						(
+							hex::encode(&node),
+							NodeUpdateTrackers {
+								ack: layout.update_trackers.ack_map.get(node, min_stored),
+								sync: layout.update_trackers.sync_map.get(node, min_stored),
+								sync_ack: layout.update_trackers.sync_ack_map.get(node, min_stored),
+							},
+						)
+					})
+					.collect(),
+			)
+		} else {
+			None
+		};
+
+		Ok(GetClusterLayoutHistoryResponse {
+			current_version: layout.current().version,
+			min_ack,
+			versions,
+			update_trackers,
+		})
+	}
+}
+
+// ----
+
+// ---- update functions ----
+
+impl RequestHandler for UpdateClusterLayoutRequest {
+	type Response = UpdateClusterLayoutResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<UpdateClusterLayoutResponse, Error> {
+		let mut layout = garage.system.cluster_layout().inner().clone();
+
+		let mut roles = layout.current().roles.clone();
+		roles.merge(&layout.staging.get().roles);
+
+		for change in self.roles {
+			let node = hex::decode(&change.id).ok_or_bad_request("Invalid node identifier")?;
+			let node = Uuid::try_from(&node).ok_or_bad_request("Invalid node identifier")?;
+
+			let new_role = match change.action {
+				NodeRoleChangeEnum::Remove { remove: true } => None,
+				NodeRoleChangeEnum::Update(NodeAssignedRole {
+					zone,
+					capacity,
+					tags,
+				}) => {
+					if matches!(capacity, Some(cap) if cap < 1024) {
+						return Err(Error::bad_request("Capacity should be at least 1K (1024)"));
+					}
+					Some(layout::NodeRole {
+						zone,
+						capacity,
+						tags,
+					})
+				}
+				_ => return Err(Error::bad_request("Invalid layout change")),
+			};
+
+			layout
+				.staging
+				.get_mut()
+				.roles
+				.merge(&roles.update_mutator(node, layout::NodeRoleV(new_role)));
+		}
+
+		if let Some(param) = self.parameters {
+			if let ZoneRedundancy::AtLeast(r_int) = param.zone_redundancy {
+				if r_int > layout.current().replication_factor {
+					return Err(Error::bad_request(format!(
+						"The zone redundancy must be smaller or equal to the replication factor ({}).",
+						layout.current().replication_factor
+					)));
+				} else if r_int < 1 {
+					return Err(Error::bad_request(
+						"The zone redundancy must be at least 1.",
+					));
+				}
+			}
+			layout.staging.get_mut().parameters.update(param.into());
+		}
+
+		garage
+			.system
+			.layout_manager
+			.update_cluster_layout(&layout)
+			.await?;
+
+		let res = format_cluster_layout(&layout);
+		Ok(UpdateClusterLayoutResponse(res))
+	}
+}
+
+impl RequestHandler for PreviewClusterLayoutChangesRequest {
+	type Response = PreviewClusterLayoutChangesResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<PreviewClusterLayoutChangesResponse, Error> {
+		let layout = garage.system.cluster_layout().inner().clone();
+		let new_ver = layout.current().version + 1;
+		match layout.apply_staged_changes(new_ver) {
+			Err(GarageError::Message(error)) => {
+				Ok(PreviewClusterLayoutChangesResponse::Error { error })
+			}
+			Err(e) => Err(e.into()),
+			Ok((new_layout, msg)) => Ok(PreviewClusterLayoutChangesResponse::Success {
+				message: msg,
+				new_layout: format_cluster_layout(&new_layout),
+			}),
+		}
+	}
+}
+
+impl RequestHandler for ApplyClusterLayoutRequest {
+	type Response = ApplyClusterLayoutResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<ApplyClusterLayoutResponse, Error> {
+		let layout = garage.system.cluster_layout().inner().clone();
+		let (layout, msg) = layout.apply_staged_changes(self.version)?;
+
+		garage
+			.system
+			.layout_manager
+			.update_cluster_layout(&layout)
+			.await?;
+
+		Ok(ApplyClusterLayoutResponse {
+			message: msg,
+			layout: format_cluster_layout(&layout),
+		})
+	}
+}
+
+impl RequestHandler for RevertClusterLayoutRequest {
+	type Response = RevertClusterLayoutResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<RevertClusterLayoutResponse, Error> {
+		let layout = garage.system.cluster_layout().inner().clone();
+		let layout = layout.revert_staged_changes()?;
+		garage
+			.system
+			.layout_manager
+			.update_cluster_layout(&layout)
+			.await?;
+
+		let res = format_cluster_layout(&layout);
+		Ok(RevertClusterLayoutResponse(res))
+	}
+}
+
+impl RequestHandler for ClusterLayoutSkipDeadNodesRequest {
+	type Response = ClusterLayoutSkipDeadNodesResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<ClusterLayoutSkipDeadNodesResponse, Error> {
+		let status = garage.system.get_known_nodes();
+
+		let mut layout = garage.system.cluster_layout().inner().clone();
+		let mut ack_updated = vec![];
+		let mut sync_updated = vec![];
+
+		if layout.versions.len() == 1 {
+			return Err(Error::bad_request(
+				"This command cannot be called when there is only one live cluster layout version",
+			));
+		}
+
+		let min_v = layout.min_stored();
+		if self.version <= min_v || self.version > layout.current().version {
+			return Err(Error::bad_request(format!(
+				"Invalid version, you may use the following version numbers: {}",
+				(min_v + 1..=layout.current().version)
+					.map(|x| x.to_string())
+					.collect::<Vec<_>>()
+					.join(" ")
+			)));
+		}
+
+		let all_nodes = layout.get_all_nodes();
+		for node in all_nodes.iter() {
+			// Update ACK tracker for dead nodes or for all nodes if --allow-missing-data
+			if self.allow_missing_data || !status.iter().any(|x| x.id == *node && x.is_up) {
+				if layout.update_trackers.ack_map.set_max(*node, self.version) {
+					ack_updated.push(hex::encode(node));
+				}
+			}
+
+			// If --allow-missing-data, update SYNC tracker for all nodes.
+			if self.allow_missing_data {
+				if layout.update_trackers.sync_map.set_max(*node, self.version) {
+					sync_updated.push(hex::encode(node));
+				}
+			}
+		}
+
+		garage
+			.system
+			.layout_manager
+			.update_cluster_layout(&layout)
+			.await?;
+
+		Ok(ClusterLayoutSkipDeadNodesResponse {
+			ack_updated,
+			sync_updated,
+		})
+	}
+}
+
+// ----
+
+impl From<layout::ZoneRedundancy> for ZoneRedundancy {
+	fn from(x: layout::ZoneRedundancy) -> Self {
+		match x {
+			layout::ZoneRedundancy::Maximum => ZoneRedundancy::Maximum,
+			layout::ZoneRedundancy::AtLeast(x) => ZoneRedundancy::AtLeast(x),
+		}
+	}
+}
+
+impl Into<layout::ZoneRedundancy> for ZoneRedundancy {
+	fn into(self) -> layout::ZoneRedundancy {
+		match self {
+			ZoneRedundancy::Maximum => layout::ZoneRedundancy::Maximum,
+			ZoneRedundancy::AtLeast(x) => layout::ZoneRedundancy::AtLeast(x),
+		}
+	}
+}
+
+impl From<layout::LayoutParameters> for LayoutParameters {
+	fn from(x: layout::LayoutParameters) -> Self {
+		LayoutParameters {
+			zone_redundancy: x.zone_redundancy.into(),
+		}
+	}
+}
+
+impl Into<layout::LayoutParameters> for LayoutParameters {
+	fn into(self) -> layout::LayoutParameters {
+		layout::LayoutParameters {
+			zone_redundancy: self.zone_redundancy.into(),
+		}
+	}
+}

src/api/admin/lib.rs

@@ -3,9 +3,44 @@ extern crate tracing;
 
 pub mod api_server;
 mod error;
+mod macros;
+
+pub mod api;
+pub mod openapi;
 mod router_v0;
 mod router_v1;
+mod router_v2;
 
+mod admin_token;
 mod bucket;
 mod cluster;
 mod key;
+mod layout;
+mod special;
+
+mod block;
+mod node;
+mod repair;
+mod worker;
+
+use std::sync::Arc;
+
+use garage_model::garage::Garage;
+
+pub use api_server::AdminApiServer as Admin;
+
+pub enum Authorization {
+	None,
+	MetricsToken,
+	AdminToken,
+}
+
+pub trait RequestHandler {
+	type Response;
+
+	fn handle(
+		self,
+		garage: &Arc<Garage>,
+		admin: &Admin,
+	) -> impl std::future::Future<Output = Result<Self::Response, error::Error>> + Send;
+}

src/api/admin/macros.rs

@@ -0,0 +1,206 @@
+macro_rules! admin_endpoints {
+    [
+        $(@special $special_endpoint:ident,)*
+        $($endpoint:ident,)*
+    ] => {
+        paste! {
+            #[derive(Debug, Clone, Serialize, Deserialize)]
+            pub enum AdminApiRequest {
+                $(
+                    $special_endpoint( [<$special_endpoint Request>] ),
+                )*
+                $(
+                    $endpoint( [<$endpoint Request>] ),
+                )*
+            }
+
+            #[derive(Debug, Clone, Serialize)]
+            #[serde(untagged)]
+            pub enum AdminApiResponse {
+                $(
+                    $endpoint( [<$endpoint Response>] ),
+                )*
+            }
+
+            #[derive(Debug, Clone, Serialize, Deserialize)]
+            pub enum TaggedAdminApiResponse {
+                $(
+                    $endpoint( [<$endpoint Response>] ),
+                )*
+            }
+
+            impl AdminApiRequest {
+                pub fn name(&self) -> &'static str {
+                    match self {
+                        $(
+                            Self::$special_endpoint(_) => stringify!($special_endpoint),
+                        )*
+                        $(
+                            Self::$endpoint(_) => stringify!($endpoint),
+                        )*
+                    }
+                }
+            }
+
+            impl AdminApiResponse {
+                pub fn tagged(self) -> TaggedAdminApiResponse {
+                    match self {
+                        $(
+                            Self::$endpoint(res) => TaggedAdminApiResponse::$endpoint(res),
+                        )*
+                    }
+                }
+            }
+
+            $(
+                impl From< [< $endpoint Request >] > for AdminApiRequest {
+                    fn from(req: [< $endpoint Request >]) -> AdminApiRequest {
+                        AdminApiRequest::$endpoint(req)
+                    }
+                }
+
+                impl TryFrom<TaggedAdminApiResponse> for [< $endpoint Response >] {
+                    type Error = TaggedAdminApiResponse;
+                    fn try_from(resp: TaggedAdminApiResponse) -> Result< [< $endpoint Response >], TaggedAdminApiResponse> {
+                        match resp {
+                            TaggedAdminApiResponse::$endpoint(v) => Ok(v),
+                            x => Err(x),
+                        }
+                    }
+                }
+            )*
+
+            impl RequestHandler for AdminApiRequest {
+                type Response = AdminApiResponse;
+
+                async fn handle(self, garage: &Arc<Garage>, admin: &Admin) -> Result<AdminApiResponse, Error> {
+                    Ok(match self {
+                        $(
+                            AdminApiRequest::$special_endpoint(_) => panic!(
+                                concat!(stringify!($special_endpoint), " needs to go through a special handler")
+                            ),
+                        )*
+                        $(
+                            AdminApiRequest::$endpoint(req) => AdminApiResponse::$endpoint(req.handle(garage, admin).await?),
+                        )*
+                    })
+                }
+            }
+        }
+    };
+}
+
+macro_rules! local_admin_endpoints {
+    [
+        $($endpoint:ident,)*
+    ] => {
+        paste! {
+            #[derive(Debug, Clone, Serialize, Deserialize)]
+            pub enum LocalAdminApiRequest {
+                $(
+                    $endpoint( [<Local $endpoint Request>] ),
+                )*
+            }
+
+            #[derive(Debug, Clone, Serialize, Deserialize)]
+            pub enum LocalAdminApiResponse {
+                $(
+                    $endpoint( [<Local $endpoint Response>] ),
+                )*
+            }
+
+            $(
+                pub type [< $endpoint Request >] = MultiRequest< [< Local $endpoint Request >] >;
+
+                pub type [< $endpoint RequestBody >] = [< Local $endpoint Request >];
+
+                pub type [< $endpoint Response >] = MultiResponse< [< Local $endpoint Response >] >;
+
+                impl From< [< Local $endpoint Request >] > for LocalAdminApiRequest {
+                    fn from(req: [< Local $endpoint Request >]) -> LocalAdminApiRequest {
+                        LocalAdminApiRequest::$endpoint(req)
+                    }
+                }
+
+                impl TryFrom<LocalAdminApiResponse> for [< Local $endpoint Response >] {
+                    type Error = LocalAdminApiResponse;
+                    fn try_from(resp: LocalAdminApiResponse) -> Result< [< Local $endpoint Response >], LocalAdminApiResponse> {
+                        match resp {
+                            LocalAdminApiResponse::$endpoint(v) => Ok(v),
+                            x => Err(x),
+                        }
+                    }
+                }
+
+                impl RequestHandler for [< $endpoint Request >] {
+                    type Response = [< $endpoint Response >];
+
+	                async fn handle(self, garage: &Arc<Garage>, admin: &Admin) -> Result<Self::Response, Error> {
+                        let to = find_matching_nodes(garage, self.node.as_str())?;
+
+                        let resps = garage.system.rpc_helper().call_many(&admin.endpoint,
+                            &to,
+                            AdminRpc::Internal(self.body.into()),
+                            RequestStrategy::with_priority(PRIO_NORMAL),
+                        ).await?;
+
+                        let mut ret = [< $endpoint Response >] {
+                            success: HashMap::new(),
+                            error: HashMap::new(),
+                        };
+                        for (node, resp) in resps {
+                            match resp {
+                                Ok(AdminRpcResponse::InternalApiOkResponse(r)) => {
+                                    match [< Local $endpoint Response >]::try_from(r) {
+                                        Ok(r) => {
+                                            ret.success.insert(hex::encode(node), r);
+                                        }
+                                        Err(_) => {
+                                            ret.error.insert(hex::encode(node), "returned invalid value".to_string());
+                                        }
+                                    }
+                                }
+                                Ok(AdminRpcResponse::ApiErrorResponse{error_code, http_code, message}) => {
+                                    ret.error.insert(hex::encode(node), format!("{} ({}): {}", error_code, http_code, message));
+                                }
+                                Ok(_) => {
+                                    ret.error.insert(hex::encode(node), "returned invalid value".to_string());
+                                }
+                                Err(e) => {
+                                    ret.error.insert(hex::encode(node), e.to_string());
+                                }
+                            }
+                        }
+
+                        Ok(ret)
+                    }
+                }
+            )*
+
+            impl LocalAdminApiRequest {
+                pub fn name(&self) -> &'static str {
+                    match self {
+                        $(
+                            Self::$endpoint(_) => stringify!($endpoint),
+                        )*
+                    }
+                }
+            }
+
+            impl RequestHandler for LocalAdminApiRequest {
+                type Response = LocalAdminApiResponse;
+
+                async fn handle(self, garage: &Arc<Garage>, admin: &Admin) -> Result<LocalAdminApiResponse, Error> {
+                    Ok(match self {
+                        $(
+                            LocalAdminApiRequest::$endpoint(req) => LocalAdminApiResponse::$endpoint(req.handle(garage, admin).await?),
+                        )*
+                    })
+                }
+            }
+        }
+    };
+}
+
+pub(crate) use admin_endpoints;
+pub(crate) use local_admin_endpoints;

src/api/admin/node.rs

@@ -0,0 +1,144 @@
+use std::fmt::Write;
+use std::sync::Arc;
+
+use format_table::format_table_to_string;
+
+use garage_util::error::Error as GarageError;
+
+use garage_table::replication::*;
+use garage_table::*;
+
+use garage_model::garage::Garage;
+
+use crate::api::*;
+use crate::error::Error;
+use crate::{Admin, RequestHandler};
+
+impl RequestHandler for LocalGetNodeInfoRequest {
+	type Response = LocalGetNodeInfoResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<LocalGetNodeInfoResponse, Error> {
+		Ok(LocalGetNodeInfoResponse {
+			node_id: hex::encode(garage.system.id),
+			garage_version: garage_util::version::garage_version().to_string(),
+			garage_features: garage_util::version::garage_features()
+				.map(|features| features.iter().map(ToString::to_string).collect()),
+			rust_version: garage_util::version::rust_version().to_string(),
+			db_engine: garage.db.engine(),
+		})
+	}
+}
+
+impl RequestHandler for LocalCreateMetadataSnapshotRequest {
+	type Response = LocalCreateMetadataSnapshotResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<LocalCreateMetadataSnapshotResponse, Error> {
+		garage_model::snapshot::async_snapshot_metadata(garage).await?;
+		Ok(LocalCreateMetadataSnapshotResponse)
+	}
+}
+
+impl RequestHandler for LocalGetNodeStatisticsRequest {
+	type Response = LocalGetNodeStatisticsResponse;
+
+	// FIXME: return this as a JSON struct instead of text
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<LocalGetNodeStatisticsResponse, Error> {
+		let sys_status = garage.system.local_status();
+
+		let mut ret = format_table_to_string(vec![
+			format!("Node ID:\t{:?}", garage.system.id),
+			format!("Hostname:\t{}", sys_status.hostname.unwrap_or_default(),),
+			format!(
+				"Garage version:\t{}",
+				garage_util::version::garage_version(),
+			),
+			format!(
+				"Garage features:\t{}",
+				garage_util::version::garage_features()
+					.map(|list| list.join(", "))
+					.unwrap_or_else(|| "(unknown)".into()),
+			),
+			format!(
+				"Rust compiler version:\t{}",
+				garage_util::version::rust_version(),
+			),
+			format!("Database engine:\t{}", garage.db.engine()),
+		]);
+
+		// Gather table statistics
+		let mut table = vec!["  Table\tItems\tMklItems\tMklTodo\tInsQueue\tGcTodo".into()];
+		table.push(gather_table_stats(&garage.admin_token_table)?);
+		table.push(gather_table_stats(&garage.bucket_table)?);
+		table.push(gather_table_stats(&garage.bucket_alias_table)?);
+		table.push(gather_table_stats(&garage.key_table)?);
+
+		table.push(gather_table_stats(&garage.object_table)?);
+		table.push(gather_table_stats(&garage.object_counter_table.table)?);
+		table.push(gather_table_stats(&garage.mpu_table)?);
+		table.push(gather_table_stats(&garage.mpu_counter_table.table)?);
+		table.push(gather_table_stats(&garage.version_table)?);
+		table.push(gather_table_stats(&garage.block_ref_table)?);
+
+		#[cfg(feature = "k2v")]
+		{
+			table.push(gather_table_stats(&garage.k2v.item_table)?);
+			table.push(gather_table_stats(&garage.k2v.counter_table.table)?);
+		}
+
+		write!(
+			&mut ret,
+			"\nTable stats:\n{}",
+			format_table_to_string(table)
+		)
+		.unwrap();
+
+		// Gather block manager statistics
+		writeln!(&mut ret, "\nBlock manager stats:").unwrap();
+		let rc_len = garage.block_manager.rc_len()?.to_string();
+
+		ret += &format_table_to_string(vec![
+			format!("  number of RC entries:\t{} (~= number of blocks)", rc_len),
+			format!(
+				"  resync queue length:\t{}",
+				garage.block_manager.resync.queue_len()?
+			),
+			format!(
+				"  blocks with resync errors:\t{}",
+				garage.block_manager.resync.errors_len()?
+			),
+		]);
+
+		Ok(LocalGetNodeStatisticsResponse { freeform: ret })
+	}
+}
+
+fn gather_table_stats<F, R>(t: &Arc<Table<F, R>>) -> Result<String, Error>
+where
+	F: TableSchema + 'static,
+	R: TableReplication + 'static,
+{
+	let data_len = t.data.store.len().map_err(GarageError::from)?.to_string();
+	let mkl_len = t.merkle_updater.merkle_tree_len()?.to_string();
+
+	Ok(format!(
+		"  {}\t{}\t{}\t{}\t{}\t{}",
+		F::TABLE_NAME,
+		data_len,
+		mkl_len,
+		t.merkle_updater.todo_len()?,
+		t.data.insert_queue_len()?,
+		t.data.gc_todo_len()?
+	))
+}

src/api/admin/openapi.rs

@@ -0,0 +1,924 @@
+#![allow(dead_code)]
+#![allow(non_snake_case)]
+
+use utoipa::{Modify, OpenApi};
+
+use crate::api::*;
+
+// **********************************************
+//      Special endpoints
+// **********************************************
+
+#[utoipa::path(get,
+    path = "/metrics",
+    tag = "Special endpoints",
+    description = "Prometheus metrics endpoint",
+    security((), ("bearerAuth" = [])),
+	responses(
+            (status = 200, description = "Garage daemon metrics exported in Prometheus format"),
+        ),
+)]
+fn Metrics() -> () {}
+
+#[utoipa::path(get,
+    path = "/health",
+    tag = "Special endpoints",
+    description = "
+Check cluster health. The status code returned by this function indicates
+whether this Garage daemon can answer API requests.
+Garage will return `200 OK` even if some storage nodes are disconnected,
+as long as it is able to have a quorum of nodes for read and write operations.
+    ",
+    security(()),
+	responses(
+            (status = 200, description = "Garage is able to answer requests"),
+            (status = 503, description = "This Garage daemon is not able to handle requests")
+        ),
+)]
+fn Health() -> () {}
+
+#[utoipa::path(get,
+    path = "/check",
+    tag = "Special endpoints",
+    description = "
+Static website domain name check. Checks whether a bucket is configured to serve
+a static website for the requested domain. This is used by reverse proxies such
+as Caddy or Tricot, to avoid requesting TLS certificates for domain names that
+do not correspond to an actual website.
+    ",
+    params(
+        ("domain", description = "The domain name to check for"),
+    ),
+    security(()),
+	responses(
+            (status = 200, description = "The domain name redirects to a static website bucket"),
+            (status = 400, description = "No static website bucket exists for this domain")
+        ),
+)]
+fn CheckDomain() -> () {}
+
+// **********************************************
+//      Cluster operations
+// **********************************************
+
+#[utoipa::path(get,
+    path = "/v2/GetClusterStatus",
+    tag = "Cluster",
+    description = "
+Returns the cluster's current status, including:
+
+- ID of the node being queried and its version of the Garage daemon
+- Live nodes
+- Currently configured cluster layout
+- Staged changes to the cluster layout
+
+*Capacity is given in bytes*
+    ",
+	responses(
+            (status = 200, description = "Cluster status report", body = GetClusterStatusResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetClusterStatus() -> () {}
+
+#[utoipa::path(get,
+    path = "/v2/GetClusterHealth",
+    tag = "Cluster",
+    description = "Returns the global status of the cluster, the number of connected nodes (over the number of known ones), the number of healthy storage nodes (over the declared ones), and the number of healthy partitions (over the total).",
+	responses(
+            (status = 200, description = "Cluster health report", body = GetClusterHealthResponse),
+        ),
+)]
+fn GetClusterHealth() -> () {}
+
+#[utoipa::path(get,
+    path = "/v2/GetClusterStatistics",
+    tag = "Cluster",
+    description = "
+Fetch global cluster statistics.
+
+*Note: do not try to parse the `freeform` field of the response, it is given as a string specifically because its format is not stable.*
+    ",
+	responses(
+            (status = 200, description = "Global cluster statistics", body = GetClusterStatisticsResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetClusterStatistics() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/ConnectClusterNodes",
+    tag = "Cluster",
+    description = "Instructs this Garage node to connect to other Garage nodes at specified `<node_id>@<net_address>`. `node_id` is generated automatically on node start.",
+    request_body=ConnectClusterNodesRequest,
+	responses(
+            (status = 200, description = "The request has been handled correctly but it does not mean that all connection requests succeeded; some might have fail, you need to check the body!", body = ConnectClusterNodesResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn ConnectClusterNodes() -> () {}
+
+// **********************************************
+//      Admin API token operations
+// **********************************************
+
+#[utoipa::path(get,
+    path = "/v2/ListAdminTokens",
+    tag = "Admin API token",
+    description = "Returns all admin API tokens in the cluster.",
+	responses(
+            (status = 200, description = "Returns info about all admin API tokens", body = ListAdminTokensResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn ListAdminTokens() -> () {}
+
+#[utoipa::path(get,
+    path = "/v2/GetAdminTokenInfo",
+    tag = "Admin API token",
+    description = "
+Return information about a specific admin API token.
+You can search by specifying the exact token identifier (`id`) or by specifying a pattern (`search`).
+    ",
+    params(GetAdminTokenInfoRequest),
+	responses(
+            (status = 200, description = "Information about the admin token", body = GetAdminTokenInfoResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetAdminTokenInfo() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/CreateAdminToken",
+    tag = "Admin API token",
+    description = "Creates a new admin API token",
+    request_body = UpdateAdminTokenRequestBody,
+	responses(
+            (status = 200, description = "Admin token has been created", body = CreateAdminTokenResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn CreateAdminToken() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/UpdateAdminToken",
+    tag = "Admin API token",
+    description = "
+Updates information about the specified admin API token.
+    ",
+    request_body = UpdateAdminTokenRequestBody,
+    params(
+        ("id", description = "Admin API token ID"),
+    ),
+	responses(
+            (status = 200, description = "Admin token has been updated", body = UpdateAdminTokenResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn UpdateAdminToken() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/DeleteAdminToken",
+    tag = "Admin API token",
+    description = "Delete an admin API token from the cluster, revoking all its permissions.",
+    params(
+        ("id", description = "Admin API token ID"),
+    ),
+	responses(
+            (status = 200, description = "Admin token has been deleted"),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn DeleteAdminToken() -> () {}
+
+// **********************************************
+//      Layout operations
+// **********************************************
+
+#[utoipa::path(get,
+    path = "/v2/GetClusterLayout",
+    tag = "Cluster layout",
+    description = "
+Returns the cluster's current layout, including:
+
+- Currently configured cluster layout
+- Staged changes to the cluster layout
+
+*Capacity is given in bytes*
+    ",
+	responses(
+            (status = 200, description = "Current cluster layout", body = GetClusterLayoutResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetClusterLayout() -> () {}
+
+#[utoipa::path(get,
+    path = "/v2/GetClusterLayoutHistory",
+    tag = "Cluster layout",
+    description = "
+Returns the history of layouts in the cluster
+    ",
+	responses(
+            (status = 200, description = "Cluster layout history", body = GetClusterLayoutHistoryResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetClusterLayoutHistory() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/UpdateClusterLayout",
+    tag = "Cluster layout",
+    description = "
+Send modifications to the cluster layout. These modifications will be included in the staged role changes, visible in subsequent calls of `GET /GetClusterHealth`. Once the set of staged changes is satisfactory, the user may call `POST /ApplyClusterLayout` to apply the changed changes, or `POST /RevertClusterLayout` to clear all of the staged changes in the layout.
+
+Setting the capacity to `null` will configure the node as a gateway.
+Otherwise, capacity must be now set in bytes (before Garage 0.9 it was arbitrary weights).
+For example to declare 100GB, you must set `capacity: 100000000000`.
+
+Garage uses internally the International System of Units (SI), it assumes that 1kB = 1000 bytes, and displays storage as kB, MB, GB (and not KiB, MiB, GiB that assume 1KiB = 1024 bytes).
+    ",
+    request_body(
+        content=UpdateClusterLayoutRequest,
+        description="
+To add a new node to the layout or to change the configuration of an existing node, simply set the values you want (`zone`, `capacity`, and `tags`).
+To remove a node, simply pass the `remove: true` field.
+This logic is represented in OpenAPI with a 'One Of' object.
+
+Contrary to the CLI that may update only a subset of the fields capacity, zone and tags, when calling this API all of these values must be specified.
+        "
+    ),
+	responses(
+            (status = 200, description = "Proposed changes have been added to the list of pending changes", body = UpdateClusterLayoutResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn UpdateClusterLayout() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/PreviewClusterLayoutChanges",
+    tag = "Cluster layout",
+    description = "
+Computes a new layout taking into account the staged parameters, and returns it with detailed statistics. The new layout is not applied in the cluster.
+
+*Note: do not try to parse the `message` field of the response, it is given as an array of string specifically because its format is not stable.*
+    ",
+	responses(
+            (status = 200, description = "Information about the new layout", body = PreviewClusterLayoutChangesResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn PreviewClusterLayoutChanges() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/ApplyClusterLayout",
+    tag = "Cluster layout",
+    description = "
+Applies to the cluster the layout changes currently registered as staged layout changes.
+
+*Note: do not try to parse the `message` field of the response, it is given as an array of string specifically because its format is not stable.*
+    ",
+    request_body=ApplyClusterLayoutRequest,
+	responses(
+            (status = 200, description = "The updated cluster layout has been applied in the cluster", body = ApplyClusterLayoutResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn ApplyClusterLayout() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/RevertClusterLayout",
+    tag = "Cluster layout",
+    description = "Clear staged layout changes",
+	responses(
+            (status = 200, description = "All pending changes to the cluster layout have been erased", body = RevertClusterLayoutResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn RevertClusterLayout() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/ClusterLayoutSkipDeadNodes",
+    tag = "Cluster layout",
+    description = "Force progress in layout update trackers",
+    request_body = ClusterLayoutSkipDeadNodesRequest,
+	responses(
+            (status = 200, description = "Request has been taken into account", body = ClusterLayoutSkipDeadNodesResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn ClusterLayoutSkipDeadNodes() -> () {}
+
+// **********************************************
+//      Access key operations
+// **********************************************
+
+#[utoipa::path(get,
+    path = "/v2/ListKeys",
+    tag = "Access key",
+    description = "Returns all API access keys in the cluster.",
+	responses(
+            (status = 200, description = "Returns the key identifier (aka `AWS_ACCESS_KEY_ID`) and its associated, human friendly, name if any (otherwise return an empty string)", body = ListKeysResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn ListKeys() -> () {}
+
+#[utoipa::path(get,
+    path = "/v2/GetKeyInfo",
+    tag = "Access key",
+    description = "
+Return information about a specific key like its identifiers, its permissions and buckets on which it has permissions.
+You can search by specifying the exact key identifier (`id`) or by specifying a pattern (`search`).
+
+For confidentiality reasons, the secret key is not returned by default: you must pass the `showSecretKey` query parameter to get it.
+    ",
+    params(GetKeyInfoRequest),
+	responses(
+            (status = 200, description = "Information about the access key", body = GetKeyInfoResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetKeyInfo() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/CreateKey",
+    tag = "Access key",
+    description = "Creates a new API access key.",
+    request_body = CreateKeyRequest,
+	responses(
+            (status = 200, description = "Access key has been created", body = CreateKeyResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn CreateKey() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/ImportKey",
+    tag = "Access key",
+    description = "
+Imports an existing API key. This feature must only be used for migrations and backup restore.
+
+**Do not use it to generate custom key identifiers or you will break your Garage cluster.**
+    ",
+    request_body = ImportKeyRequest,
+	responses(
+            (status = 200, description = "Access key has been imported", body = ImportKeyResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn ImportKey() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/UpdateKey",
+    tag = "Access key",
+    description = "
+Updates information about the specified API access key.
+
+*Note: the secret key is not returned in the response, `null` is sent instead.*
+    ",
+    request_body = UpdateKeyRequestBody,
+    params(
+        ("id", description = "Access key ID"),
+    ),
+	responses(
+            (status = 200, description = "Access key has been updated", body = UpdateKeyResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn UpdateKey() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/DeleteKey",
+    tag = "Access key",
+    description = "Delete a key from the cluster. Its access will be removed from all the buckets. Buckets are not automatically deleted and can be dangling. You should manually delete them before. ",
+    params(
+        ("id", description = "Access key ID"),
+    ),
+	responses(
+            (status = 200, description = "Access key has been deleted"),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn DeleteKey() -> () {}
+
+// **********************************************
+//      Bucket operations
+// **********************************************
+
+#[utoipa::path(get,
+    path = "/v2/ListBuckets",
+    tag = "Bucket",
+    description = "List all the buckets on the cluster with their UUID and their global and local aliases.",
+	responses(
+            (status = 200, description = "Returns the UUID of all the buckets and all their aliases", body = ListBucketsResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn ListBuckets() -> () {}
+
+#[utoipa::path(get,
+    path = "/v2/GetBucketInfo",
+    tag = "Bucket",
+    description = "
+Given a bucket identifier (`id`) or a global alias (`alias`), get its information.
+It includes its aliases, its web configuration, keys that have some permissions
+on it, some statistics (number of objects, size), number of dangling multipart uploads,
+and its quotas (if any).
+    ",
+    params(GetBucketInfoRequest),
+	responses(
+            (status = 200, description = "Returns exhaustive information about the bucket", body = GetBucketInfoResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetBucketInfo() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/CreateBucket",
+    tag = "Bucket",
+    description = "
+Creates a new bucket, either with a global alias, a local one, or no alias at all.
+Technically, you can also specify both `globalAlias` and `localAlias` and that would create two aliases.
+    ",
+    request_body = CreateBucketRequest,
+	responses(
+            (status = 200, description = "Returns exhaustive information about the bucket", body = CreateBucketResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn CreateBucket() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/UpdateBucket",
+    tag = "Bucket",
+    description = "
+All fields (`websiteAccess` and `quotas`) are optional.
+If they are present, the corresponding modifications are applied to the bucket, otherwise nothing is changed.
+
+In `websiteAccess`: if `enabled` is `true`, `indexDocument` must be specified.
+The field `errorDocument` is optional, if no error document is set a generic
+error message is displayed when errors happen. Conversely, if `enabled` is
+`false`, neither `indexDocument` nor `errorDocument` must be specified.
+
+In `quotas`: new values of `maxSize` and `maxObjects` must both be specified, or set to `null`
+to remove the quotas. An absent value will be considered the same as a `null`. It is not possible
+to change only one of the two quotas.
+    ",
+    params(
+        ("id", description = "ID of the bucket to update"),
+    ),
+    request_body = UpdateBucketRequestBody,
+	responses(
+            (status = 200, description = "Bucket has been updated", body = UpdateBucketResponse),
+            (status = 404, description = "Bucket not found"),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn UpdateBucket() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/DeleteBucket",
+    tag = "Bucket",
+    description = "
+Deletes a storage bucket. A bucket cannot be deleted if it is not empty.
+
+**Warning:** this will delete all aliases associated with the bucket!
+    ",
+    params(
+        ("id", description = "ID of the bucket to delete"),
+    ),
+	responses(
+            (status = 200, description = "Bucket has been deleted"),
+            (status = 400, description = "Bucket is not empty"),
+            (status = 404, description = "Bucket not found"),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn DeleteBucket() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/CleanupIncompleteUploads",
+    tag = "Bucket",
+    description = "Removes all incomplete multipart uploads that are older than the specified number of seconds.",
+    request_body = CleanupIncompleteUploadsRequest,
+	responses(
+            (status = 200, description = "The bucket was cleaned up successfully", body = CleanupIncompleteUploadsResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn CleanupIncompleteUploads() -> () {}
+
+#[utoipa::path(get,
+    path = "/v2/InspectObject",
+    tag = "Bucket",
+    description = "
+Returns detailed information about an object in a bucket, including its internal state in Garage.
+
+This API call can be used to list the data blocks referenced by an object,
+as well as to view metadata associated to the object.
+
+This call may return a list of more than one version for the object, for instance in the
+case where there is a currently stored version of the object, and a newer version whose
+upload is in progress and not yet finished.
+    ",
+    params(InspectObjectRequest),
+	responses(
+            (status = 200, description = "Returns exhaustive information about the object", body = InspectObjectResponse),
+            (status = 404, description = "Object not found"),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn InspectObject() -> () {}
+
+// **********************************************
+//      Operations on permissions for keys on buckets
+// **********************************************
+
+#[utoipa::path(post,
+    path = "/v2/AllowBucketKey",
+    tag = "Permission",
+    description = "
+⚠️ **DISCLAIMER**: Garage's developers are aware that this endpoint has an unconventional semantic. Be extra careful when implementing it, its behavior is not obvious.
+
+Allows a key to do read/write/owner operations on a bucket.
+
+Flags in permissions which have the value true will be activated. Other flags will remain unchanged (ie. they will keep their internal value).
+
+For example, if you set read to true, the key will be allowed to read the bucket.
+If you set it to false, the key will keeps its previous read permission.
+If you want to disallow read for the key, check the DenyBucketKey operation.
+    ",
+    request_body = AllowBucketKeyRequest,
+	responses(
+            (status = 200, description = "Returns exhaustive information about the bucket", body = AllowBucketKeyResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn AllowBucketKey() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/DenyBucketKey",
+    tag = "Permission",
+    description = "
+⚠️ **DISCLAIMER**: Garage's developers are aware that this endpoint has an unconventional semantic. Be extra careful when implementing it, its behavior is not obvious.
+
+Denies a key from doing read/write/owner operations on a bucket.
+
+Flags in permissions which have the value true will be deactivated. Other flags will remain unchanged.
+
+For example, if you set read to true, the key will be denied from reading.
+If you set read to false,  the key will keep its previous permissions.
+If you want the key to have the reading permission, check the AllowBucketKey operation.
+    ",
+    request_body = DenyBucketKeyRequest,
+	responses(
+            (status = 200, description = "Returns exhaustive information about the bucket", body = DenyBucketKeyResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn DenyBucketKey() -> () {}
+
+// **********************************************
+//      Operations on bucket aliases
+// **********************************************
+
+#[utoipa::path(post,
+    path = "/v2/AddBucketAlias",
+    tag = "Bucket alias",
+    description = "Add an alias for the target bucket.  This can be either a global or a local alias, depending on which fields are specified.",
+    request_body = AddBucketAliasRequest,
+	responses(
+            (status = 200, description = "Returns exhaustive information about the bucket", body = AddBucketAliasResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn AddBucketAlias() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/RemoveBucketAlias",
+    tag = "Bucket alias",
+    description = "Remove an alias for the target bucket.  This can be either a global or a local alias, depending on which fields are specified.",
+    request_body = RemoveBucketAliasRequest,
+	responses(
+            (status = 200, description = "Returns exhaustive information about the bucket", body = RemoveBucketAliasResponse),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn RemoveBucketAlias() -> () {}
+
+// **********************************************
+//      Node operations
+// **********************************************
+
+#[utoipa::path(get,
+    path = "/v2/GetNodeInfo",
+    tag = "Node",
+    description = "
+Return information about the Garage daemon running on one or several nodes.
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+	responses(
+            (status = 200, description = "Responses from individual cluster nodes", body = MultiResponse<LocalGetNodeInfoResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetNodeInfo() -> () {}
+
+#[utoipa::path(get,
+    path = "/v2/GetNodeStatistics",
+    tag = "Node",
+    description = "
+Fetch statistics for one or several Garage nodes.
+
+*Note: do not try to parse the `freeform` field of the response, it is given as a string specifically because its format is not stable.*
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+	responses(
+            (status = 200, description = "Responses from individual cluster nodes", body = MultiResponse<LocalGetNodeStatisticsResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetNodeStatistics() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/CreateMetadataSnapshot",
+    tag = "Node",
+    description = "
+Instruct one or several nodes to take a snapshot of their metadata databases.
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+	responses(
+            (status = 200, description = "Responses from individual cluster nodes", body = MultiResponse<LocalCreateMetadataSnapshotResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn CreateMetadataSnapshot() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/LaunchRepairOperation",
+    tag = "Node",
+    description = "
+Launch a repair operation on one or several cluster nodes.
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+    request_body = LocalLaunchRepairOperationRequest,
+	responses(
+            (status = 200, description = "Responses from individual cluster nodes", body = MultiResponse<LocalLaunchRepairOperationResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn LaunchRepairOperation() -> () {}
+
+// **********************************************
+//      Worker operations
+// **********************************************
+
+#[utoipa::path(post,
+    path = "/v2/ListWorkers",
+    tag = "Worker",
+    description = "
+List background workers currently running on one or several cluster nodes.
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+    request_body = LocalListWorkersRequest,
+	responses(
+            (status = 200, description = "Responses from individual cluster nodes", body = MultiResponse<LocalListWorkersResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn ListWorkers() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/GetWorkerInfo",
+    tag = "Worker",
+    description = "
+Get information about the specified background worker on one or several cluster nodes.
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+    request_body = LocalGetWorkerInfoRequest,
+	responses(
+            (status = 200, description = "Responses from individual cluster nodes", body = MultiResponse<LocalGetWorkerInfoResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetWorkerInfo() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/GetWorkerVariable",
+    tag = "Worker",
+    description = "
+Fetch values of one or several worker variables, from one or several cluster nodes.
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+    request_body = LocalGetWorkerVariableRequest,
+	responses(
+            (status = 200, description = "Responses from individual cluster nodes", body = MultiResponse<LocalGetWorkerVariableResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetWorkerVariable() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/SetWorkerVariable",
+    tag = "Worker",
+    description = "
+Set the value for a worker variable, on one or several cluster nodes.
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+    request_body = LocalSetWorkerVariableRequest,
+	responses(
+            (status = 200, description = "Responses from individual cluster nodes", body = MultiResponse<LocalSetWorkerVariableResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn SetWorkerVariable() -> () {}
+
+// **********************************************
+//      Block operations
+// **********************************************
+
+#[utoipa::path(get,
+    path = "/v2/ListBlockErrors",
+    tag = "Block",
+    description = "
+List data blocks that are currently in an errored state on one or several Garage nodes.
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+	responses(
+            (status = 200, description = "Responses from individual cluster nodes", body = MultiResponse<LocalListBlockErrorsResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn ListBlockErrors() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/GetBlockInfo",
+    tag = "Block",
+    description = "
+Get detailed information about a data block stored on a Garage node, including all object versions and in-progress multipart uploads that contain a reference to this block.
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+    request_body = LocalGetBlockInfoRequest,
+	responses(
+            (status = 200, description = "Detailed block information", body = MultiResponse<LocalGetBlockInfoResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn GetBlockInfo() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/RetryBlockResync",
+    tag = "Block",
+    description = "
+Instruct Garage node(s) to retry the resynchronization of one or several missing data block(s).
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+    request_body = LocalRetryBlockResyncRequest,
+	responses(
+            (status = 200, description = "Responses from individual cluster nodes", body = MultiResponse<LocalRetryBlockResyncResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn RetryBlockResync() -> () {}
+
+#[utoipa::path(post,
+    path = "/v2/PurgeBlocks",
+    tag = "Block",
+    description = "
+Purge references to one or several missing data blocks.
+
+This will remove all objects and in-progress multipart uploads that contain the specified data block(s). The objects will be permanently deleted from the buckets in which they appear. Use with caution.
+    ",
+    params(
+        ("node", description = "Node ID to query, or `*` for all nodes, or `self` for the node responding to the request"),
+    ),
+    request_body = LocalPurgeBlocksRequest,
+	responses(
+            (status = 200, description = "Responses from individual cluster nodes", body = MultiResponse<LocalPurgeBlocksResponse>),
+            (status = 500, description = "Internal server error")
+        ),
+)]
+fn PurgeBlocks() -> () {}
+
+// **********************************************
+// **********************************************
+// **********************************************
+
+struct SecurityAddon;
+
+impl Modify for SecurityAddon {
+	fn modify(&self, openapi: &mut utoipa::openapi::OpenApi) {
+		use utoipa::openapi::security::*;
+		let components = openapi.components.as_mut().unwrap(); // we can unwrap safely since there already is components registered.
+		components.add_security_scheme(
+			"bearerAuth",
+			SecurityScheme::Http(Http::builder().scheme(HttpAuthScheme::Bearer).build()),
+		)
+	}
+}
+
+#[derive(OpenApi)]
+#[openapi(
+    info(
+        version = "v2.0.0",
+        title = "Garage administration API",
+        description = "Administrate your Garage cluster programatically, including status, layout, keys, buckets, and maintainance tasks.
+
+*Disclaimer: This API may change in future Garage versions. Read the changelog and upgrade your scripts before upgrading. Additionnaly, this specification is early stage and can contain bugs, so be careful and please report any issues on our issue tracker.*",
+        contact(
+            name = "The Garage team",
+            email = "garagehq@deuxfleurs.fr",
+            url = "https://garagehq.deuxfleurs.fr/",
+        ),
+    ),
+    modifiers(&SecurityAddon),
+    security(("bearerAuth" = [])),
+    paths(
+        // Special ops
+        Metrics,
+        Health,
+        CheckDomain,
+        // Cluster operations
+        GetClusterHealth,
+        GetClusterStatus,
+        GetClusterStatistics,
+        ConnectClusterNodes,
+        // Admin token operations
+        ListAdminTokens,
+        GetAdminTokenInfo,
+        CreateAdminToken,
+        UpdateAdminToken,
+        DeleteAdminToken,
+        // Layout operations
+        GetClusterLayout,
+        GetClusterLayoutHistory,
+        UpdateClusterLayout,
+        PreviewClusterLayoutChanges,
+        ApplyClusterLayout,
+        RevertClusterLayout,
+        ClusterLayoutSkipDeadNodes,
+        // Key operations
+        ListKeys,
+        GetKeyInfo,
+        CreateKey,
+        ImportKey,
+        UpdateKey,
+        DeleteKey,
+        // Bucket operations
+        ListBuckets,
+        GetBucketInfo,
+        CreateBucket,
+        UpdateBucket,
+        DeleteBucket,
+        CleanupIncompleteUploads,
+        InspectObject,
+        // Operations on permissions
+        AllowBucketKey,
+        DenyBucketKey,
+        // Operations on aliases
+        AddBucketAlias,
+        RemoveBucketAlias,
+        // Node operations
+        GetNodeInfo,
+        GetNodeStatistics,
+        CreateMetadataSnapshot,
+        LaunchRepairOperation,
+        // Worker operations
+        ListWorkers,
+        GetWorkerInfo,
+        GetWorkerVariable,
+        SetWorkerVariable,
+        // Block operations
+        ListBlockErrors,
+        GetBlockInfo,
+        RetryBlockResync,
+        PurgeBlocks,
+    ),
+    servers(
+        (url = "http://localhost:3903/", description = "A local server")
+    ),
+)]
+pub struct ApiDoc;

src/api/admin/repair.rs

@@ -5,6 +5,14 @@ use std::time::Duration;
 use async_trait::async_trait;
 use tokio::sync::watch;
 
+use garage_util::background::*;
+use garage_util::data::*;
+use garage_util::error::{Error as GarageError, OkOrMessage};
+use garage_util::migrate::Migrate;
+
+use garage_table::replication::*;
+use garage_table::*;
+
 use garage_block::manager::BlockManager;
 use garage_block::repair::ScrubWorkerCommand;
 
@@ -14,25 +22,23 @@ use garage_model::s3::mpu_table::*;
 use garage_model::s3::object_table::*;
 use garage_model::s3::version_table::*;
 
-use garage_table::replication::*;
-use garage_table::*;
-
-use garage_util::background::*;
-use garage_util::data::*;
-use garage_util::error::Error;
-use garage_util::migrate::Migrate;
-
-use crate::*;
+use crate::api::*;
+use crate::error::Error;
+use crate::{Admin, RequestHandler};
 
 const RC_REPAIR_ITER_COUNT: usize = 64;
 
-pub async fn launch_online_repair(
+impl RequestHandler for LocalLaunchRepairOperationRequest {
+	type Response = LocalLaunchRepairOperationResponse;
+
+	async fn handle(
+		self,
 		garage: &Arc<Garage>,
-	bg: &BackgroundRunner,
-	opt: RepairOpt,
-) -> Result<(), Error> {
-	match opt.what {
-		RepairWhat::Tables => {
+		admin: &Admin,
+	) -> Result<LocalLaunchRepairOperationResponse, Error> {
+		let bg = &admin.background;
+		match self.repair_type {
+			RepairType::Tables => {
 				info!("Launching a full sync of tables");
 				garage.bucket_table.syncer.add_full_sync()?;
 				garage.object_table.syncer.add_full_sync()?;
@@ -40,65 +46,52 @@ pub async fn launch_online_repair(
 				garage.block_ref_table.syncer.add_full_sync()?;
 				garage.key_table.syncer.add_full_sync()?;
 			}
-		RepairWhat::Versions => {
+			RepairType::Versions => {
 				info!("Repairing the versions table");
 				bg.spawn_worker(TableRepairWorker::new(garage.clone(), RepairVersions));
 			}
-		RepairWhat::MultipartUploads => {
+			RepairType::MultipartUploads => {
 				info!("Repairing the multipart uploads table");
 				bg.spawn_worker(TableRepairWorker::new(garage.clone(), RepairMpu));
 			}
-		RepairWhat::BlockRefs => {
+			RepairType::BlockRefs => {
 				info!("Repairing the block refs table");
 				bg.spawn_worker(TableRepairWorker::new(garage.clone(), RepairBlockRefs));
 			}
-		RepairWhat::BlockRc => {
+			RepairType::BlockRc => {
 				info!("Repairing the block reference counters");
 				bg.spawn_worker(BlockRcRepair::new(
 					garage.block_manager.clone(),
 					garage.block_ref_table.clone(),
 				));
 			}
-		RepairWhat::Blocks => {
+			RepairType::Blocks => {
 				info!("Repairing the stored blocks");
 				bg.spawn_worker(garage_block::repair::RepairWorker::new(
 					garage.block_manager.clone(),
 				));
 			}
-		RepairWhat::Scrub { cmd } => {
+			RepairType::Scrub(cmd) => {
 				let cmd = match cmd {
-				ScrubCmd::Start => ScrubWorkerCommand::Start,
-				ScrubCmd::Pause => ScrubWorkerCommand::Pause(Duration::from_secs(3600 * 24)),
-				ScrubCmd::Resume => ScrubWorkerCommand::Resume,
-				ScrubCmd::Cancel => ScrubWorkerCommand::Cancel,
-				ScrubCmd::SetTranquility { tranquility } => {
-					garage
-						.block_manager
-						.scrub_persister
-						.set_with(|x| x.tranquility = tranquility)?;
-					return Ok(());
+					ScrubCommand::Start => ScrubWorkerCommand::Start,
+					ScrubCommand::Pause => {
+						ScrubWorkerCommand::Pause(Duration::from_secs(3600 * 24))
 					}
+					ScrubCommand::Resume => ScrubWorkerCommand::Resume,
+					ScrubCommand::Cancel => ScrubWorkerCommand::Cancel,
 				};
 				info!("Sending command to scrub worker: {:?}", cmd);
 				garage.block_manager.send_scrub_command(cmd).await?;
 			}
-		RepairWhat::Rebalance => {
+			RepairType::Rebalance => {
 				info!("Rebalancing the stored blocks among storage locations");
 				bg.spawn_worker(garage_block::repair::RebalanceWorker::new(
 					garage.block_manager.clone(),
 				));
 			}
-		RepairWhat::Aliases => {
-			info!("Repairing bucket aliases (foreground)");
-			garage.locked_helper().await.repair_aliases().await?;
 		}
-		RepairWhat::ClearResyncQueue => {
-			let garage = garage.clone();
-			tokio::task::spawn_blocking(move || garage.block_manager.resync.clear_resync_queue())
-				.await??
+		Ok(LocalLaunchRepairOperationResponse)
 	}
-	}
-	Ok(())
 }
 
 // ----
@@ -112,7 +105,7 @@ trait TableRepair: Send + Sync + 'static {
 		&mut self,
 		garage: &Garage,
 		entry: <<Self as TableRepair>::T as TableSchema>::E,
-	) -> impl Future<Output = Result<bool, Error>> + Send;
+	) -> impl Future<Output = Result<bool, GarageError>> + Send;
 }
 
 struct TableRepairWorker<T: TableRepair> {
@@ -148,7 +141,10 @@ impl<R: TableRepair> Worker for TableRepairWorker<R> {
 		}
 	}
 
-	async fn work(&mut self, _must_exit: &mut watch::Receiver<bool>) -> Result<WorkerState, Error> {
+	async fn work(
+		&mut self,
+		_must_exit: &mut watch::Receiver<bool>,
+	) -> Result<WorkerState, GarageError> {
 		let (item_bytes, next_pos) = match R::table(&self.garage).data.store.get_gt(&self.pos)? {
 			Some((k, v)) => (v, k),
 			None => {
@@ -190,7 +186,7 @@ impl TableRepair for RepairVersions {
 		&garage.version_table
 	}
 
-	async fn process(&mut self, garage: &Garage, version: Version) -> Result<bool, Error> {
+	async fn process(&mut self, garage: &Garage, version: Version) -> Result<bool, GarageError> {
 		if !version.deleted.get() {
 			let ref_exists = match &version.backlink {
 				VersionBacklink::Object { bucket_id, key } => garage
@@ -236,7 +232,11 @@ impl TableRepair for RepairBlockRefs {
 		&garage.block_ref_table
 	}
 
-	async fn process(&mut self, garage: &Garage, mut block_ref: BlockRef) -> Result<bool, Error> {
+	async fn process(
+		&mut self,
+		garage: &Garage,
+		mut block_ref: BlockRef,
+	) -> Result<bool, GarageError> {
 		if !block_ref.deleted.get() {
 			let ref_exists = garage
 				.version_table
@@ -271,7 +271,11 @@ impl TableRepair for RepairMpu {
 		&garage.mpu_table
 	}
 
-	async fn process(&mut self, garage: &Garage, mut mpu: MultipartUpload) -> Result<bool, Error> {
+	async fn process(
+		&mut self,
+		garage: &Garage,
+		mut mpu: MultipartUpload,
+	) -> Result<bool, GarageError> {
 		if !mpu.deleted.get() {
 			let ref_exists = garage
 				.object_table
@@ -338,7 +342,10 @@ impl Worker for BlockRcRepair {
 		}
 	}
 
-	async fn work(&mut self, _must_exit: &mut watch::Receiver<bool>) -> Result<WorkerState, Error> {
+	async fn work(
+		&mut self,
+		_must_exit: &mut watch::Receiver<bool>,
+	) -> Result<WorkerState, GarageError> {
 		for _i in 0..RC_REPAIR_ITER_COUNT {
 			let next1 = self
 				.block_manager

src/api/admin/router_v1.rs

@@ -7,12 +7,6 @@ use garage_api_common::router_macros::*;
 use crate::error::*;
 use crate::router_v0;
 
-pub enum Authorization {
-	None,
-	MetricsToken,
-	AdminToken,
-}
-
 router_match! {@func
 
 /// List of all Admin API endpoints.
@@ -211,15 +205,6 @@ impl Endpoint {
 			))),
 		}
 	}
-	/// Get the kind of authorization which is required to perform the operation.
-	pub fn authorization_type(&self) -> Authorization {
-		match self {
-			Self::Health => Authorization::None,
-			Self::CheckDomain => Authorization::None,
-			Self::Metrics => Authorization::MetricsToken,
-			_ => Authorization::AdminToken,
-		}
-	}
 }
 
 generateQueryParameters! {

src/api/admin/router_v2.rs

@@ -0,0 +1,275 @@
+use std::borrow::Cow;
+
+use hyper::body::Incoming as IncomingBody;
+use hyper::{Method, Request};
+use paste::paste;
+
+use garage_api_common::helpers::*;
+use garage_api_common::router_macros::*;
+
+use crate::api::*;
+use crate::error::*;
+use crate::router_v1;
+use crate::Authorization;
+
+impl AdminApiRequest {
+	/// Determine which S3 endpoint a request is for using the request, and a bucket which was
+	/// possibly extracted from the Host header.
+	/// Returns Self plus bucket name, if endpoint is not Endpoint::ListBuckets
+	pub async fn from_request(req: Request<IncomingBody>) -> Result<Self, Error> {
+		let uri = req.uri().clone();
+		let path = uri.path();
+		let query = uri.query();
+
+		let method = req.method().clone();
+
+		let mut query = QueryParameters::from_query(query.unwrap_or_default())?;
+
+		let res = router_match!(@gen_path_parser_v2 (&method, path, "/v2/", query, req) [
+			@special OPTIONS _ => Options (),
+			@special GET "/check" => CheckDomain (query::domain),
+			@special GET "/health" => Health (),
+			@special GET "/metrics" => Metrics (),
+			// Cluster endpoints
+			GET GetClusterStatus (),
+			GET GetClusterHealth (),
+			POST ConnectClusterNodes (body),
+			// Admin token endpoints
+			GET ListAdminTokens (),
+			GET GetAdminTokenInfo (query_opt::id, query_opt::search),
+			POST CreateAdminToken (body),
+			POST UpdateAdminToken (body_field, query::id),
+			POST DeleteAdminToken (query::id),
+			// Layout endpoints
+			GET GetClusterLayout (),
+			GET GetClusterLayoutHistory (),
+			POST UpdateClusterLayout (body),
+			POST PreviewClusterLayoutChanges (),
+			POST ApplyClusterLayout (body),
+			POST RevertClusterLayout (),
+			POST ClusterLayoutSkipDeadNodes (body),
+			// API key endpoints
+			GET GetKeyInfo (query_opt::id, query_opt::search, parse_default(false)::show_secret_key),
+			POST UpdateKey (body_field, query::id),
+			POST CreateKey (body),
+			POST ImportKey (body),
+			POST DeleteKey (query::id),
+			GET ListKeys (),
+			// Bucket endpoints
+			GET GetBucketInfo (query_opt::id, query_opt::global_alias, query_opt::search),
+			GET ListBuckets (),
+			POST CreateBucket (body),
+			POST DeleteBucket (query::id),
+			POST UpdateBucket (body_field, query::id),
+			POST CleanupIncompleteUploads (body),
+			GET InspectObject (query::bucket_id, query::key),
+			// Bucket-key permissions
+			POST AllowBucketKey (body),
+			POST DenyBucketKey (body),
+			// Bucket aliases
+			POST AddBucketAlias (body),
+			POST RemoveBucketAlias (body),
+			// Node APIs
+			GET GetNodeInfo (default::body, query::node),
+			POST CreateMetadataSnapshot (default::body, query::node),
+			GET GetNodeStatistics (default::body, query::node),
+			GET GetClusterStatistics (),
+			POST LaunchRepairOperation (body_field, query::node),
+			// Worker APIs
+			POST ListWorkers (body_field, query::node),
+			POST GetWorkerInfo (body_field, query::node),
+			POST GetWorkerVariable (body_field, query::node),
+			POST SetWorkerVariable (body_field, query::node),
+			// Block APIs
+			GET ListBlockErrors (default::body, query::node),
+			POST GetBlockInfo (body_field, query::node),
+			POST RetryBlockResync (body_field, query::node),
+			POST PurgeBlocks (body_field, query::node),
+		]);
+
+		if let Some(message) = query.nonempty_message() {
+			debug!("Unused query parameter: {}", message)
+		}
+
+		Ok(res)
+	}
+
+	/// Some endpoints work exactly the same in their v2/ version as they did in their v1/ version.
+	/// For these endpoints, we can convert a v1/ call to its equivalent as if it was made using
+	/// its v2/ URL.
+	pub async fn from_v1(
+		v1_endpoint: router_v1::Endpoint,
+		req: Request<IncomingBody>,
+	) -> Result<Self, Error> {
+		use router_v1::Endpoint;
+
+		match v1_endpoint {
+			// GetClusterStatus semantics changed:
+			// info about local node is no longer returned
+			Endpoint::GetClusterHealth => {
+				Ok(AdminApiRequest::GetClusterHealth(GetClusterHealthRequest))
+			}
+			Endpoint::ConnectClusterNodes => {
+				let req = parse_json_body::<ConnectClusterNodesRequest, _, Error>(req).await?;
+				Ok(AdminApiRequest::ConnectClusterNodes(req))
+			}
+
+			// Layout
+			Endpoint::GetClusterLayout => {
+				Ok(AdminApiRequest::GetClusterLayout(GetClusterLayoutRequest))
+			}
+			// UpdateClusterLayout semantics changed
+			Endpoint::ApplyClusterLayout => {
+				let param = parse_json_body::<ApplyClusterLayoutRequest, _, Error>(req).await?;
+				Ok(AdminApiRequest::ApplyClusterLayout(param))
+			}
+			Endpoint::RevertClusterLayout => Ok(AdminApiRequest::RevertClusterLayout(
+				RevertClusterLayoutRequest,
+			)),
+
+			// Keys
+			Endpoint::ListKeys => Ok(AdminApiRequest::ListKeys(ListKeysRequest)),
+			Endpoint::GetKeyInfo {
+				id,
+				search,
+				show_secret_key,
+			} => {
+				let show_secret_key = show_secret_key.map(|x| x == "true").unwrap_or(false);
+				Ok(AdminApiRequest::GetKeyInfo(GetKeyInfoRequest {
+					id,
+					search,
+					show_secret_key,
+				}))
+			}
+			Endpoint::CreateKey => {
+				let req = parse_json_body::<CreateKeyRequest, _, Error>(req).await?;
+				Ok(AdminApiRequest::CreateKey(req))
+			}
+			Endpoint::ImportKey => {
+				let req = parse_json_body::<ImportKeyRequest, _, Error>(req).await?;
+				Ok(AdminApiRequest::ImportKey(req))
+			}
+			Endpoint::UpdateKey { id } => {
+				let body = parse_json_body::<UpdateKeyRequestBody, _, Error>(req).await?;
+				Ok(AdminApiRequest::UpdateKey(UpdateKeyRequest { id, body }))
+			}
+
+			// DeleteKey semantics changed:
+			// - in v1/ : HTTP DELETE => HTTP 204 No Content
+			// - in v2/ : HTTP POST => HTTP 200 Ok
+			// Endpoint::DeleteKey { id } => Ok(AdminApiRequest::DeleteKey(DeleteKeyRequest { id })),
+
+			// Buckets
+			Endpoint::ListBuckets => Ok(AdminApiRequest::ListBuckets(ListBucketsRequest)),
+			Endpoint::GetBucketInfo { id, global_alias } => {
+				Ok(AdminApiRequest::GetBucketInfo(GetBucketInfoRequest {
+					id,
+					global_alias,
+					search: None,
+				}))
+			}
+			Endpoint::CreateBucket => {
+				let req = parse_json_body::<CreateBucketRequest, _, Error>(req).await?;
+				Ok(AdminApiRequest::CreateBucket(req))
+			}
+
+			// DeleteBucket semantics changed::
+			// - in v1/ : HTTP DELETE => HTTP 204 No Content
+			// - in v2/ : HTTP POST => HTTP 200 Ok
+			// Endpoint::DeleteBucket { id } => {
+			// 	Ok(AdminApiRequest::DeleteBucket(DeleteBucketRequest { id }))
+			// }
+			Endpoint::UpdateBucket { id } => {
+				let body = parse_json_body::<UpdateBucketRequestBody, _, Error>(req).await?;
+				Ok(AdminApiRequest::UpdateBucket(UpdateBucketRequest {
+					id,
+					body,
+				}))
+			}
+
+			// Bucket-key permissions
+			Endpoint::BucketAllowKey => {
+				let req = parse_json_body::<BucketKeyPermChangeRequest, _, Error>(req).await?;
+				Ok(AdminApiRequest::AllowBucketKey(AllowBucketKeyRequest(req)))
+			}
+			Endpoint::BucketDenyKey => {
+				let req = parse_json_body::<BucketKeyPermChangeRequest, _, Error>(req).await?;
+				Ok(AdminApiRequest::DenyBucketKey(DenyBucketKeyRequest(req)))
+			}
+			// Bucket aliasing
+			Endpoint::GlobalAliasBucket { id, alias } => {
+				Ok(AdminApiRequest::AddBucketAlias(AddBucketAliasRequest {
+					bucket_id: id,
+					alias: BucketAliasEnum::Global {
+						global_alias: alias,
+					},
+				}))
+			}
+			Endpoint::GlobalUnaliasBucket { id, alias } => Ok(AdminApiRequest::RemoveBucketAlias(
+				RemoveBucketAliasRequest {
+					bucket_id: id,
+					alias: BucketAliasEnum::Global {
+						global_alias: alias,
+					},
+				},
+			)),
+			Endpoint::LocalAliasBucket {
+				id,
+				access_key_id,
+				alias,
+			} => Ok(AdminApiRequest::AddBucketAlias(AddBucketAliasRequest {
+				bucket_id: id,
+				alias: BucketAliasEnum::Local {
+					local_alias: alias,
+					access_key_id,
+				},
+			})),
+			Endpoint::LocalUnaliasBucket {
+				id,
+				access_key_id,
+				alias,
+			} => Ok(AdminApiRequest::RemoveBucketAlias(
+				RemoveBucketAliasRequest {
+					bucket_id: id,
+					alias: BucketAliasEnum::Local {
+						local_alias: alias,
+						access_key_id,
+					},
+				},
+			)),
+
+			// For endpoints that have different body content syntax, issue
+			// deprecation warning
+			_ => Err(Error::bad_request(format!(
+				"v1/ endpoint is no longer supported: {}",
+				v1_endpoint.name()
+			))),
+		}
+	}
+
+	/// Get the kind of authorization which is required to perform the operation.
+	pub fn authorization_type(&self) -> Authorization {
+		match self {
+			Self::Options(_) | Self::Health(_) | Self::CheckDomain(_) => Authorization::None,
+			Self::Metrics(_) => Authorization::MetricsToken,
+			_ => Authorization::AdminToken,
+		}
+	}
+}
+
+generateQueryParameters! {
+	keywords: [],
+	fields: [
+		"node" => node,
+		"domain" => domain,
+		"format" => format,
+		"id" => id,
+		"search" => search,
+		"globalAlias" => global_alias,
+		"alias" => alias,
+		"accessKeyId" => access_key_id,
+		"showSecretKey" => show_secret_key,
+		"bucketId" => bucket_id,
+		"key" => key
+	]
+}

src/api/admin/special.rs

@@ -0,0 +1,173 @@
+use std::sync::Arc;
+
+use http::header::{
+	ACCESS_CONTROL_ALLOW_HEADERS, ACCESS_CONTROL_ALLOW_METHODS, ACCESS_CONTROL_ALLOW_ORIGIN, ALLOW,
+};
+use hyper::{Response, StatusCode};
+
+#[cfg(feature = "metrics")]
+use prometheus::{Encoder, TextEncoder};
+
+use garage_model::garage::Garage;
+use garage_rpc::system::ClusterHealthStatus;
+
+use garage_api_common::helpers::*;
+
+use crate::api::{CheckDomainRequest, HealthRequest, MetricsRequest, OptionsRequest};
+use crate::api_server::ResBody;
+use crate::error::*;
+use crate::{Admin, RequestHandler};
+
+impl RequestHandler for OptionsRequest {
+	type Response = Response<ResBody>;
+
+	async fn handle(
+		self,
+		_garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<Response<ResBody>, Error> {
+		Ok(Response::builder()
+			.status(StatusCode::OK)
+			.header(ALLOW, "OPTIONS,GET,POST")
+			.header(ACCESS_CONTROL_ALLOW_METHODS, "OPTIONS,GET,POST")
+			.header(ACCESS_CONTROL_ALLOW_HEADERS, "authorization,content-type")
+			.header(ACCESS_CONTROL_ALLOW_ORIGIN, "*")
+			.body(empty_body())?)
+	}
+}
+
+impl RequestHandler for MetricsRequest {
+	type Response = Response<ResBody>;
+
+	async fn handle(
+		self,
+		_garage: &Arc<Garage>,
+		admin: &Admin,
+	) -> Result<Response<ResBody>, Error> {
+		#[cfg(feature = "metrics")]
+		{
+			use opentelemetry::trace::Tracer;
+
+			let mut buffer = vec![];
+			let encoder = TextEncoder::new();
+
+			let tracer = opentelemetry::global::tracer("garage");
+			let metric_families = tracer.in_span("admin/gather_metrics", |_| {
+				admin.exporter.registry().gather()
+			});
+
+			encoder
+				.encode(&metric_families, &mut buffer)
+				.ok_or_internal_error("Could not serialize metrics")?;
+
+			Ok(Response::builder()
+				.status(StatusCode::OK)
+				.header(http::header::CONTENT_TYPE, encoder.format_type())
+				.body(bytes_body(buffer.into()))?)
+		}
+		#[cfg(not(feature = "metrics"))]
+		Err(Error::bad_request(
+			"Garage was built without the metrics feature".to_string(),
+		))
+	}
+}
+
+impl RequestHandler for HealthRequest {
+	type Response = Response<ResBody>;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<Response<ResBody>, Error> {
+		let health = garage.system.health();
+
+		let (status, status_str) = match health.status {
+			ClusterHealthStatus::Healthy => (StatusCode::OK, "Garage is fully operational"),
+			ClusterHealthStatus::Degraded => (
+				StatusCode::OK,
+				"Garage is operational but some storage nodes are unavailable",
+			),
+			ClusterHealthStatus::Unavailable => (
+				StatusCode::SERVICE_UNAVAILABLE,
+				"Quorum is not available for some/all partitions, reads and writes will fail",
+			),
+		};
+		let status_str = format!(
+			"{}\nConsult the full health check API endpoint at /v2/GetClusterHealth for more details\n",
+			status_str
+		);
+
+		Ok(Response::builder()
+			.status(status)
+			.header(http::header::CONTENT_TYPE, "text/plain")
+			.body(string_body(status_str))?)
+	}
+}
+
+impl RequestHandler for CheckDomainRequest {
+	type Response = Response<ResBody>;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<Response<ResBody>, Error> {
+		if check_domain(garage, &self.domain).await? {
+			Ok(Response::builder()
+				.status(StatusCode::OK)
+				.body(string_body(format!(
+					"Domain '{}' is managed by Garage",
+					self.domain
+				)))?)
+		} else {
+			Err(Error::bad_request(format!(
+				"Domain '{}' is not managed by Garage",
+				self.domain
+			)))
+		}
+	}
+}
+
+async fn check_domain(garage: &Arc<Garage>, domain: &str) -> Result<bool, Error> {
+	// Resolve bucket from domain name, inferring if the website must be activated for the
+	// domain to be valid.
+	let (bucket_name, must_check_website) = if let Some(bname) = garage
+		.config
+		.s3_api
+		.root_domain
+		.as_ref()
+		.and_then(|rd| host_to_bucket(domain, rd))
+	{
+		(bname.to_string(), false)
+	} else if let Some(bname) = garage
+		.config
+		.s3_web
+		.as_ref()
+		.and_then(|sw| host_to_bucket(domain, sw.root_domain.as_str()))
+	{
+		(bname.to_string(), true)
+	} else {
+		(domain.to_string(), true)
+	};
+
+	let bucket = match garage
+		.bucket_helper()
+		.resolve_global_bucket_fast(&bucket_name)?
+	{
+		Some(b) => b,
+		None => return Ok(false),
+	};
+
+	if !must_check_website {
+		return Ok(true);
+	}
+
+	let bucket_state = bucket.state.as_option().unwrap();
+	let bucket_website_config = bucket_state.website_config.get();
+
+	match bucket_website_config {
+		Some(_v) => Ok(true),
+		None => Ok(false),
+	}
+}

src/api/admin/worker.rs

@@ -0,0 +1,118 @@
+use std::collections::HashMap;
+use std::sync::Arc;
+
+use garage_util::background::*;
+use garage_util::time::now_msec;
+
+use garage_model::garage::Garage;
+
+use crate::api::*;
+use crate::error::Error;
+use crate::{Admin, RequestHandler};
+
+impl RequestHandler for LocalListWorkersRequest {
+	type Response = LocalListWorkersResponse;
+
+	async fn handle(
+		self,
+		_garage: &Arc<Garage>,
+		admin: &Admin,
+	) -> Result<LocalListWorkersResponse, Error> {
+		let workers = admin.background.get_worker_info();
+		let info = workers
+			.into_iter()
+			.filter(|(_, w)| {
+				(!self.busy_only
+					|| matches!(w.state, WorkerState::Busy | WorkerState::Throttled(_)))
+					&& (!self.error_only || w.errors > 0)
+			})
+			.map(|(id, w)| worker_info_to_api(id as u64, w))
+			.collect::<Vec<_>>();
+		Ok(LocalListWorkersResponse(info))
+	}
+}
+
+impl RequestHandler for LocalGetWorkerInfoRequest {
+	type Response = LocalGetWorkerInfoResponse;
+
+	async fn handle(
+		self,
+		_garage: &Arc<Garage>,
+		admin: &Admin,
+	) -> Result<LocalGetWorkerInfoResponse, Error> {
+		let info = admin
+			.background
+			.get_worker_info()
+			.get(&(self.id as usize))
+			.ok_or(Error::NoSuchWorker(self.id))?
+			.clone();
+		Ok(LocalGetWorkerInfoResponse(worker_info_to_api(
+			self.id, info,
+		)))
+	}
+}
+
+impl RequestHandler for LocalGetWorkerVariableRequest {
+	type Response = LocalGetWorkerVariableResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<LocalGetWorkerVariableResponse, Error> {
+		let mut res = HashMap::new();
+		if let Some(k) = self.variable {
+			res.insert(k.clone(), garage.bg_vars.get(&k)?);
+		} else {
+			let vars = garage.bg_vars.get_all();
+			for (k, v) in vars.iter() {
+				res.insert(k.to_string(), v.to_string());
+			}
+		}
+		Ok(LocalGetWorkerVariableResponse(res))
+	}
+}
+
+impl RequestHandler for LocalSetWorkerVariableRequest {
+	type Response = LocalSetWorkerVariableResponse;
+
+	async fn handle(
+		self,
+		garage: &Arc<Garage>,
+		_admin: &Admin,
+	) -> Result<LocalSetWorkerVariableResponse, Error> {
+		garage.bg_vars.set(&self.variable, &self.value)?;
+
+		Ok(LocalSetWorkerVariableResponse {
+			variable: self.variable,
+			value: self.value,
+		})
+	}
+}
+
+// ---- helper functions ----
+
+fn worker_info_to_api(id: u64, info: WorkerInfo) -> WorkerInfoResp {
+	WorkerInfoResp {
+		id,
+		name: info.name,
+		state: match info.state {
+			WorkerState::Busy => WorkerStateResp::Busy,
+			WorkerState::Throttled(t) => WorkerStateResp::Throttled { duration_secs: t },
+			WorkerState::Idle => WorkerStateResp::Idle,
+			WorkerState::Done => WorkerStateResp::Done,
+		},
+		errors: info.errors as u64,
+		consecutive_errors: info.consecutive_errors as u64,
+		last_error: info.last_error.map(|(message, t)| WorkerLastError {
+			message,
+			secs_ago: now_msec().saturating_sub(t) / 1000,
+		}),
+
+		tranquility: info.status.tranquility,
+		progress: info.status.progress,
+		queue_length: info.status.queue_length,
+		persistent_errors: info.status.persistent_errors,
+		freeform: info.status.freeform,
+	}
+}

src/api/common/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "garage_api_common"
-version = "1.3.1"
+version = "2.0.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -23,11 +23,13 @@ bytes.workspace = true
 chrono.workspace = true
 crc32fast.workspace = true
 crc32c.workspace = true
+crc64fast-nvme.workspace = true
 crypto-common.workspace = true
-thiserror.workspace = true
+err-derive.workspace = true
 hex.workspace = true
 hmac.workspace = true
 md-5.workspace = true
+idna.workspace = true
 tracing.workspace = true
 nom.workspace = true
 pin-project.workspace = true

src/api/common/common_error.rs

@@ -1,7 +1,7 @@
 use std::convert::TryFrom;
 
+use err_derive::Error;
 use hyper::StatusCode;
-use thiserror::Error;
 
 use garage_util::error::Error as GarageError;
 
@@ -12,48 +12,48 @@ use garage_model::helper::error::Error as HelperError;
 pub enum CommonError {
 	// ---- INTERNAL ERRORS ----
 	/// Error related to deeper parts of Garage
-	#[error("Internal error: {0}")]
-	InternalError(#[from] GarageError),
+	#[error(display = "Internal error: {}", _0)]
+	InternalError(#[error(source)] GarageError),
 
 	/// Error related to Hyper
-	#[error("Internal error (Hyper error): {0}")]
-	Hyper(#[from] hyper::Error),
+	#[error(display = "Internal error (Hyper error): {}", _0)]
+	Hyper(#[error(source)] hyper::Error),
 
 	/// Error related to HTTP
-	#[error("Internal error (HTTP error): {0}")]
-	Http(#[from] http::Error),
+	#[error(display = "Internal error (HTTP error): {}", _0)]
+	Http(#[error(source)] http::Error),
 
 	// ---- GENERIC CLIENT ERRORS ----
 	/// Proper authentication was not provided
-	#[error("Forbidden: {0}")]
+	#[error(display = "Forbidden: {}", _0)]
 	Forbidden(String),
 
 	/// Generic bad request response with custom message
-	#[error("Bad request: {0}")]
+	#[error(display = "Bad request: {}", _0)]
 	BadRequest(String),
 
 	/// The client sent a header with invalid value
-	#[error("Invalid header value: {0}")]
-	InvalidHeader(#[from] hyper::header::ToStrError),
+	#[error(display = "Invalid header value: {}", _0)]
+	InvalidHeader(#[error(source)] hyper::header::ToStrError),
 
 	// ---- SPECIFIC ERROR CONDITIONS ----
 	// These have to be error codes referenced in the S3 spec here:
 	// https://docs.aws.amazon.com/AmazonS3/latest/API/ErrorResponses.html#ErrorCodeList
 	/// The bucket requested don't exists
-	#[error("Bucket not found: {0}")]
+	#[error(display = "Bucket not found: {}", _0)]
 	NoSuchBucket(String),
 
 	/// Tried to create a bucket that already exist
-	#[error("Bucket already exists")]
+	#[error(display = "Bucket already exists")]
 	BucketAlreadyExists,
 
 	/// Tried to delete a non-empty bucket
-	#[error("Tried to delete a non-empty bucket")]
+	#[error(display = "Tried to delete a non-empty bucket")]
 	BucketNotEmpty,
 
 	// Category: bad request
 	/// Bucket name is not valid according to AWS S3 specs
-	#[error("Invalid bucket name: {0}")]
+	#[error(display = "Invalid bucket name: {}", _0)]
 	InvalidBucketName(String),
 }
 

src/api/common/cors.rs

@@ -9,9 +9,7 @@ use hyper::{body::Body, body::Incoming as IncomingBody, Request, Response, Statu
 use garage_model::bucket_table::{BucketParams, CorsRule as GarageCorsRule};
 use garage_model::garage::Garage;
 
-use crate::common_error::{
-	helper_error_as_internal, CommonError, OkOrBadRequest, OkOrInternalError,
-};
+use crate::common_error::{CommonError, OkOrBadRequest, OkOrInternalError};
 use crate::helpers::*;
 
 pub fn find_matching_cors_rule<'a, B>(
@@ -76,7 +74,7 @@ pub fn add_cors_headers(
 	Ok(())
 }
 
-pub async fn handle_options_api(
+pub fn handle_options_api(
 	garage: Arc<Garage>,
 	req: &Request<IncomingBody>,
 	bucket_name: Option<String>,
@@ -93,16 +91,8 @@ pub async fn handle_options_api(
 	// OPTIONS calls are not auhtenticated).
 	if let Some(bn) = bucket_name {
 		let helper = garage.bucket_helper();
-		let bucket_id = helper
-			.resolve_global_bucket_name(&bn)
-			.await
-			.map_err(helper_error_as_internal)?;
-		if let Some(id) = bucket_id {
-			let bucket = garage
-				.bucket_helper()
-				.get_existing_bucket(id)
-				.await
-				.map_err(helper_error_as_internal)?;
+		let bucket_opt = helper.resolve_global_bucket_fast(&bn)?;
+		if let Some(bucket) = bucket_opt {
 			let bucket_params = bucket.state.into_option().unwrap();
 			handle_options_for_bucket(req, &bucket_params)
 		} else {

src/api/common/generic_server.rs

@@ -1,3 +1,4 @@
+use std::borrow::Cow;
 use std::convert::Infallible;
 use std::fs::{self, Permissions};
 use std::os::unix::fs::PermissionsExt;
@@ -35,7 +36,7 @@ use garage_util::socket_address::UnixOrTCPSocketAddress;
 use crate::helpers::{BoxBody, ErrorBody};
 
 pub trait ApiEndpoint: Send + Sync + 'static {
-	fn name(&self) -> &'static str;
+	fn name(&self) -> Cow<'static, str>;
 	fn add_span_attributes(&self, span: SpanRef<'_>);
 }
 
@@ -58,12 +59,6 @@ pub trait ApiHandler: Send + Sync + 'static {
 		req: Request<IncomingBody>,
 		endpoint: Self::Endpoint,
 	) -> impl Future<Output = Result<Response<BoxBody<Self::Error>>, Self::Error>> + Send;
-
-	/// Returns the key id used to authenticate this request. The ID returned must be safe to
-	/// log.
-	fn key_id_from_request(&self, _req: &Request<IncomingBody>) -> Option<String> {
-		None
-	}
 }
 
 pub struct ApiServer<A: ApiHandler> {
@@ -148,20 +143,19 @@ impl<A: ApiHandler> ApiServer<A> {
 	) -> Result<Response<BoxBody<A::Error>>, http::Error> {
 		let uri = req.uri().clone();
 
-		let source = if let Ok(forwarded_for_ip_addr) =
+		if let Ok(forwarded_for_ip_addr) =
 			forwarded_headers::handle_forwarded_for_headers(req.headers())
 		{
-			format!("{forwarded_for_ip_addr} (via {addr})")
+			info!(
+				"{} (via {}) {} {}",
+				forwarded_for_ip_addr,
+				addr,
+				req.method(),
+				uri
+			);
 		} else {
-			format!("{addr}")
-		};
-		// we only do this to log the access key, so we can discard any error
-		let key = self
-			.api_handler
-			.key_id_from_request(&req)
-			.map(|k| format!("(key {k}) "))
-			.unwrap_or_default();
-		info!("{source} {key}{} {uri}", req.method());
+			info!("{} {} {}", addr, req.method(), uri);
+		}
 		debug!("{:?}", req);
 
 		let tracer = opentelemetry::global::tracer("garage");
@@ -350,11 +344,7 @@ where
 
 	while !*must_exit.borrow() {
 		let (stream, client_addr) = tokio::select! {
-			acc = listener.accept() => match acc {
-							Ok(r) => r,
-							Err(e) if e.kind() == std::io::ErrorKind::ConnectionAborted => continue,
-							Err(e) => return Err(e.into()),
-						},
+			acc = listener.accept() => acc?,
 			_ = must_exit.changed() => continue,
 		};
 

src/api/common/helpers.rs

@@ -8,6 +8,7 @@ use hyper::{
 	body::{Body, Bytes},
 	Request, Response,
 };
+use idna::domain_to_unicode;
 use serde::{Deserialize, Serialize};
 
 use garage_model::bucket_table::BucketParams;
@@ -96,7 +97,7 @@ pub fn authority_to_host(authority: &str) -> Result<String, Error> {
 			authority
 		))),
 	};
-	authority.map(|h| h.to_ascii_lowercase())
+	authority.map(|h| domain_to_unicode(h).0)
 }
 
 /// Extract the bucket name and the key name from an HTTP path and possibly a bucket provided in

src/api/common/router_macros.rs

@@ -45,6 +45,68 @@ macro_rules! router_match {
             }
         }
     }};
+    (@gen_path_parser_v2 ($method:expr, $reqpath:expr, $pathprefix:literal, $query:expr, $req:expr)
+     [
+     $(@special $spec_meth:ident $spec_path:pat =>  $spec_api:ident $spec_params:tt,)*
+     $($meth:ident $api:ident $params:tt,)*
+     ]) => {{
+        {
+            #[allow(unused_parens)]
+            match ($method, $reqpath) {
+                $(
+                    (&Method::$spec_meth, $spec_path) => AdminApiRequest::$spec_api (
+                        router_match!(@@gen_parse_request $spec_api, $spec_params, $query, $req)
+                    ),
+                )*
+                $(
+                    (&Method::$meth, concat!($pathprefix, stringify!($api)))
+                    => AdminApiRequest::$api (
+                        router_match!(@@gen_parse_request $api, $params, $query, $req)
+                        ),
+                )*
+                (m, p) => {
+                    return Err(Error::bad_request(format!(
+                        "Unknown API endpoint: {} {}",
+                        m, p
+                    )))
+                }
+            }
+        }
+    }};
+    (@@gen_parse_request $api:ident, (), $query: expr, $req:expr) => {{
+        paste!(
+            [< $api Request >]
+        )
+    }};
+    (@@gen_parse_request $api:ident, (body), $query: expr, $req:expr) => {{
+        paste!({
+            parse_json_body::< [<$api Request>], _, Error>($req).await?
+        })
+    }};
+    (@@gen_parse_request $api:ident, (body_field, $($conv:ident $(($conv_arg:expr))? :: $param:ident),*), $query: expr, $req:expr)
+        =>
+    {{
+        paste!({
+            let body = parse_json_body::< [<$api RequestBody>], _, Error>($req).await?;
+            [< $api Request >] {
+                body,
+                $(
+                    $param: router_match!(@@parse_param $query, $conv $(($conv_arg))?, $param),
+                )+
+            }
+        })
+    }};
+    (@@gen_parse_request $api:ident, ($($conv:ident $(($conv_arg:expr))? :: $param:ident),*), $query: expr, $req:expr)
+        =>
+    {{
+        paste!({
+            [< $api Request >] {
+                $(
+                    $param: router_match!(@@parse_param $query, $conv $(($conv_arg))?, $param),
+                )+
+            }
+        })
+    }};
     (@gen_parser ($keyword:expr, $key:ident, $query:expr, $header:expr),
         key: [$($kw_k:ident $(if $required_k:ident)? $(header $header_k:expr)? => $api_k:ident $(($($conv_k:ident :: $param_k:ident),*))?,)*],
         no_key: [$($kw_nk:ident $(if $required_nk:ident)? $(if_header $header_nk:expr)? => $api_nk:ident $(($($conv_nk:ident :: $param_nk:ident),*))?,)*]) => {{
@@ -79,13 +141,19 @@ macro_rules! router_match {
         }
     }};
 
+    (@@parse_param $query:expr, default, $param:ident) => {{
+        Default::default()
+    }};
     (@@parse_param $query:expr, query_opt, $param:ident) => {{
         // extract optional query parameter
         $query.$param.take().map(|param| param.into_owned())
     }};
     (@@parse_param $query:expr, query, $param:ident) => {{
         // extract mendatory query parameter
-        $query.$param.take().ok_or_bad_request("Missing argument for endpoint")?.into_owned()
+        $query.$param.take()
+            .ok_or_bad_request(
+                format!("Missing argument `{}` for endpoint", stringify!($param))
+            )?.into_owned()
     }};
     (@@parse_param $query:expr, opt_parse, $param:ident) => {{
         // extract and parse optional query parameter
@@ -99,10 +167,22 @@ macro_rules! router_match {
     (@@parse_param $query:expr, parse, $param:ident) => {{
         // extract and parse mandatory query parameter
         // both missing and un-parseable parameters are reported as errors
-        $query.$param.take().ok_or_bad_request("Missing argument for endpoint")?
+        $query.$param.take()
+            .ok_or_bad_request(
+                format!("Missing argument `{}` for endpoint", stringify!($param))
+            )?
             .parse()
             .map_err(|_| Error::bad_request("Failed to parse query parameter"))?
     }};
+    (@@parse_param $query:expr, parse_default($default:expr), $param:ident) => {{
+        // extract and parse optional query parameter
+        // using provided value as default if paramter is missing
+        $query.$param.take().map(|x| x
+            .parse()
+            .map_err(|_| Error::bad_request("Failed to parse query parameter")))
+            .transpose()?
+            .unwrap_or($default)
+    }};
     (@func
     $(#[$doc:meta])*
      pub enum Endpoint {
@@ -187,6 +267,7 @@ macro_rules! generateQueryParameters {
                             },
                         )*
                         $(
+                            // FIXME: remove if !v.is_empty() ?
                             $f_param => if !v.is_empty() {
                                 if res.$f_name.replace(v).is_some() {
                                     return Err(Error::bad_request(format!(

src/api/common/signature/checksum.rs

@@ -4,6 +4,7 @@ use std::hash::Hasher;
 use base64::prelude::*;
 use crc32c::Crc32cHasher as Crc32c;
 use crc32fast::Hasher as Crc32;
+use crc64fast_nvme::Digest as Crc64Nvme;
 use md5::{Digest, Md5};
 use sha1::Sha1;
 use sha2::Sha256;
@@ -23,11 +24,14 @@ pub const X_AMZ_CHECKSUM_ALGORITHM: HeaderName =
 pub const X_AMZ_CHECKSUM_MODE: HeaderName = HeaderName::from_static("x-amz-checksum-mode");
 pub const X_AMZ_CHECKSUM_CRC32: HeaderName = HeaderName::from_static("x-amz-checksum-crc32");
 pub const X_AMZ_CHECKSUM_CRC32C: HeaderName = HeaderName::from_static("x-amz-checksum-crc32c");
+pub const X_AMZ_CHECKSUM_CRC64NVME: HeaderName =
+	HeaderName::from_static("x-amz-checksum-crc64nvme");
 pub const X_AMZ_CHECKSUM_SHA1: HeaderName = HeaderName::from_static("x-amz-checksum-sha1");
 pub const X_AMZ_CHECKSUM_SHA256: HeaderName = HeaderName::from_static("x-amz-checksum-sha256");
 
 pub type Crc32Checksum = [u8; 4];
 pub type Crc32cChecksum = [u8; 4];
+pub type Crc64NvmeChecksum = [u8; 8];
 pub type Md5Checksum = [u8; 16];
 pub type Sha1Checksum = [u8; 20];
 pub type Sha256Checksum = [u8; 32];
@@ -45,6 +49,7 @@ pub struct ExpectedChecksums {
 pub struct Checksummer {
 	pub crc32: Option<Crc32>,
 	pub crc32c: Option<Crc32c>,
+	pub crc64nvme: Option<Crc64Nvme>,
 	pub md5: Option<Md5>,
 	pub sha1: Option<Sha1>,
 	pub sha256: Option<Sha256>,
@@ -54,6 +59,7 @@ pub struct Checksummer {
 pub struct Checksums {
 	pub crc32: Option<Crc32Checksum>,
 	pub crc32c: Option<Crc32cChecksum>,
+	pub crc64nvme: Option<Crc64NvmeChecksum>,
 	pub md5: Option<Md5Checksum>,
 	pub sha1: Option<Sha1Checksum>,
 	pub sha256: Option<Sha256Checksum>,
@@ -64,6 +70,7 @@ impl Checksummer {
 		Self {
 			crc32: None,
 			crc32c: None,
+			crc64nvme: None,
 			md5: None,
 			sha1: None,
 			sha256: None,
@@ -96,6 +103,9 @@ impl Checksummer {
 		if matches!(&expected.extra, Some(ChecksumValue::Crc32c(_))) {
 			self.crc32c = Some(Crc32c::default());
 		}
+		if matches!(&expected.extra, Some(ChecksumValue::Crc64Nvme(_))) {
+			self.crc64nvme = Some(Crc64Nvme::default());
+		}
 		if matches!(&expected.extra, Some(ChecksumValue::Sha1(_))) {
 			self.sha1 = Some(Sha1::new());
 		}
@@ -109,6 +119,9 @@ impl Checksummer {
 			Some(ChecksumAlgorithm::Crc32c) => {
 				self.crc32c = Some(Crc32c::default());
 			}
+			Some(ChecksumAlgorithm::Crc64Nvme) => {
+				self.crc64nvme = Some(Crc64Nvme::default());
+			}
 			Some(ChecksumAlgorithm::Sha1) => {
 				self.sha1 = Some(Sha1::new());
 			}
@@ -127,6 +140,9 @@ impl Checksummer {
 		if let Some(crc32c) = &mut self.crc32c {
 			crc32c.write(bytes);
 		}
+		if let Some(crc64nvme) = &mut self.crc64nvme {
+			crc64nvme.write(bytes);
+		}
 		if let Some(md5) = &mut self.md5 {
 			md5.update(bytes);
 		}
@@ -144,6 +160,7 @@ impl Checksummer {
 			crc32c: self
 				.crc32c
 				.map(|x| u32::to_be_bytes(u32::try_from(x.finish()).unwrap())),
+			crc64nvme: self.crc64nvme.map(|x| u64::to_be_bytes(x.sum64())),
 			md5: self.md5.map(|x| x.finalize()[..].try_into().unwrap()),
 			sha1: self.sha1.map(|x| x.finalize()[..].try_into().unwrap()),
 			sha256: self.sha256.map(|x| x.finalize()[..].try_into().unwrap()),
@@ -190,6 +207,9 @@ impl Checksums {
 			None => None,
 			Some(ChecksumAlgorithm::Crc32) => Some(ChecksumValue::Crc32(self.crc32.unwrap())),
 			Some(ChecksumAlgorithm::Crc32c) => Some(ChecksumValue::Crc32c(self.crc32c.unwrap())),
+			Some(ChecksumAlgorithm::Crc64Nvme) => {
+				Some(ChecksumValue::Crc64Nvme(self.crc64nvme.unwrap()))
+			}
 			Some(ChecksumAlgorithm::Sha1) => Some(ChecksumValue::Sha1(self.sha1.unwrap())),
 			Some(ChecksumAlgorithm::Sha256) => Some(ChecksumValue::Sha256(self.sha256.unwrap())),
 		}
@@ -202,6 +222,7 @@ pub fn parse_checksum_algorithm(algo: &str) -> Result<ChecksumAlgorithm, Error>
 	match algo {
 		"CRC32" => Ok(ChecksumAlgorithm::Crc32),
 		"CRC32C" => Ok(ChecksumAlgorithm::Crc32c),
+		"CRC64NVME" => Ok(ChecksumAlgorithm::Crc64Nvme),
 		"SHA1" => Ok(ChecksumAlgorithm::Sha1),
 		"SHA256" => Ok(ChecksumAlgorithm::Sha256),
 		_ => Err(Error::bad_request("invalid checksum algorithm")),
@@ -225,6 +246,7 @@ pub fn request_trailer_checksum_algorithm(
 		None => Ok(None),
 		Some(x) if x == X_AMZ_CHECKSUM_CRC32 => Ok(Some(ChecksumAlgorithm::Crc32)),
 		Some(x) if x == X_AMZ_CHECKSUM_CRC32C => Ok(Some(ChecksumAlgorithm::Crc32c)),
+		Some(x) if x == X_AMZ_CHECKSUM_CRC64NVME => Ok(Some(ChecksumAlgorithm::Crc64Nvme)),
 		Some(x) if x == X_AMZ_CHECKSUM_SHA1 => Ok(Some(ChecksumAlgorithm::Sha1)),
 		Some(x) if x == X_AMZ_CHECKSUM_SHA256 => Ok(Some(ChecksumAlgorithm::Sha256)),
 		_ => Err(Error::bad_request("invalid checksum algorithm")),
@@ -243,6 +265,12 @@ pub fn request_checksum_value(
 	if headers.contains_key(X_AMZ_CHECKSUM_CRC32C) {
 		ret.push(extract_checksum_value(headers, ChecksumAlgorithm::Crc32c)?);
 	}
+	if headers.contains_key(X_AMZ_CHECKSUM_CRC64NVME) {
+		ret.push(extract_checksum_value(
+			headers,
+			ChecksumAlgorithm::Crc64Nvme,
+		)?);
+	}
 	if headers.contains_key(X_AMZ_CHECKSUM_SHA1) {
 		ret.push(extract_checksum_value(headers, ChecksumAlgorithm::Sha1)?);
 	}
@@ -281,6 +309,14 @@ pub fn extract_checksum_value(
 				.ok_or_bad_request("invalid x-amz-checksum-crc32c header")?;
 			Ok(ChecksumValue::Crc32c(crc32c))
 		}
+		ChecksumAlgorithm::Crc64Nvme => {
+			let crc64nvme = headers
+				.get(X_AMZ_CHECKSUM_CRC64NVME)
+				.and_then(|x| BASE64_STANDARD.decode(&x).ok())
+				.and_then(|x| x.try_into().ok())
+				.ok_or_bad_request("invalid x-amz-checksum-crc64nvme header")?;
+			Ok(ChecksumValue::Crc64Nvme(crc64nvme))
+		}
 		ChecksumAlgorithm::Sha1 => {
 			let sha1 = headers
 				.get(X_AMZ_CHECKSUM_SHA1)
@@ -311,6 +347,9 @@ pub fn add_checksum_response_headers(
 		Some(ChecksumValue::Crc32c(crc32c)) => {
 			resp = resp.header(X_AMZ_CHECKSUM_CRC32C, BASE64_STANDARD.encode(&crc32c));
 		}
+		Some(ChecksumValue::Crc64Nvme(crc64nvme)) => {
+			resp = resp.header(X_AMZ_CHECKSUM_CRC64NVME, BASE64_STANDARD.encode(&crc64nvme));
+		}
 		Some(ChecksumValue::Sha1(sha1)) => {
 			resp = resp.header(X_AMZ_CHECKSUM_SHA1, BASE64_STANDARD.encode(&sha1));
 		}

src/api/common/signature/error.rs

@@ -1,4 +1,4 @@
-use thiserror::Error;
+use err_derive::Error;
 
 use crate::common_error::CommonError;
 pub use crate::common_error::{CommonErrorDerivative, OkOrBadRequest, OkOrInternalError};
@@ -6,21 +6,21 @@ pub use crate::common_error::{CommonErrorDerivative, OkOrBadRequest, OkOrInterna
 /// Errors of this crate
 #[derive(Debug, Error)]
 pub enum Error {
-	#[error("{0}")]
+	#[error(display = "{}", _0)]
 	/// Error from common error
 	Common(CommonError),
 
 	/// Authorization Header Malformed
-	#[error("Authorization header malformed, unexpected scope: {0}")]
+	#[error(display = "Authorization header malformed, unexpected scope: {}", _0)]
 	AuthorizationHeaderMalformed(String),
 
 	// Category: bad request
 	/// The request contained an invalid UTF-8 sequence in its path or in other parameters
-	#[error("Invalid UTF-8: {0}")]
-	InvalidUtf8Str(#[from] std::str::Utf8Error),
+	#[error(display = "Invalid UTF-8: {}", _0)]
+	InvalidUtf8Str(#[error(source)] std::str::Utf8Error),
 
 	/// The provided digest (checksum) value was invalid
-	#[error("Invalid digest: {0}")]
+	#[error(display = "Invalid digest: {}", _0)]
 	InvalidDigest(String),
 }
 

src/api/common/signature/mod.rs

@@ -64,12 +64,12 @@ pub struct VerifiedRequest {
 	pub content_sha256_header: ContentSha256Header,
 }
 
-pub async fn verify_request(
+pub fn verify_request(
 	garage: &Garage,
 	mut req: Request<IncomingBody>,
 	service: &'static str,
 ) -> Result<VerifiedRequest, Error> {
-	let checked_signature = payload::check_payload_signature(&garage, &mut req, service).await?;
+	let checked_signature = payload::check_payload_signature(&garage, &mut req, service)?;
 
 	let request = streaming::parse_streaming_body(
 		req,

src/api/common/signature/payload.rs

@@ -9,6 +9,7 @@ use sha2::{Digest, Sha256};
 
 use garage_table::*;
 use garage_util::data::Hash;
+use garage_util::time::now_msec;
 
 use garage_model::garage::Garage;
 use garage_model::key_table::*;
@@ -32,7 +33,7 @@ pub struct CheckedSignature {
 	pub signature_header: Option<String>,
 }
 
-pub async fn check_payload_signature(
+pub fn check_payload_signature(
 	garage: &Garage,
 	request: &mut Request<IncomingBody>,
 	service: &'static str,
@@ -43,9 +44,9 @@ pub async fn check_payload_signature(
 		// We check for presigned-URL-style authentication first, because
 		// the browser or something else could inject an Authorization header
 		// that is totally unrelated to AWS signatures.
-		check_presigned_signature(garage, service, request, query).await
+		check_presigned_signature(garage, service, request, query)
 	} else if request.headers().contains_key(AUTHORIZATION) {
-		check_standard_signature(garage, service, request, query).await
+		check_standard_signature(garage, service, request, query)
 	} else {
 		// Unsigned (anonymous) request
 		let content_sha256 = request
@@ -93,7 +94,7 @@ fn parse_x_amz_content_sha256(header: Option<&str>) -> Result<ContentSha256Heade
 	}
 }
 
-async fn check_standard_signature(
+fn check_standard_signature(
 	garage: &Garage,
 	service: &'static str,
 	request: &Request<IncomingBody>,
@@ -104,7 +105,7 @@ async fn check_standard_signature(
 	// Verify that all necessary request headers are included in signed_headers
 	// The following must be included for all signatures:
 	// - the Host header (mandatory)
-	// - all x-amz-* headers used in the request (except x-amz-content-sha256)
+	// - all x-amz-* headers used in the request
 	// AWS also indicates that the Content-Type header should be signed if
 	// it is used, but Minio client doesn't sign it so we don't check it for compatibility.
 	let signed_headers = split_signed_headers(&authorization)?;
@@ -128,7 +129,7 @@ async fn check_standard_signature(
 	trace!("canonical request:\n{}", canonical_request);
 	trace!("string to sign:\n{}", string_to_sign);
 
-	let key = verify_v4(garage, service, &authorization, string_to_sign.as_bytes()).await?;
+	let key = verify_v4(garage, service, &authorization, string_to_sign.as_bytes())?;
 
 	let content_sha256_header = parse_x_amz_content_sha256(Some(&authorization.content_sha256))?;
 
@@ -139,7 +140,7 @@ async fn check_standard_signature(
 	})
 }
 
-async fn check_presigned_signature(
+fn check_presigned_signature(
 	garage: &Garage,
 	service: &'static str,
 	request: &mut Request<IncomingBody>,
@@ -151,7 +152,7 @@ async fn check_presigned_signature(
 	// Verify that all necessary request headers are included in signed_headers
 	// For AWSv4 pre-signed URLs, the following must be included:
 	// - the Host header (mandatory)
-	// - all x-amz-* headers used in the request (except x-amz-content-sha256)
+	// - all x-amz-* headers used in the request
 	let signed_headers = split_signed_headers(&authorization)?;
 	verify_signed_headers(request.headers(), &signed_headers)?;
 
@@ -178,7 +179,7 @@ async fn check_presigned_signature(
 	trace!("canonical request (presigned url):\n{}", canonical_request);
 	trace!("string to sign (presigned url):\n{}", string_to_sign);
 
-	let key = verify_v4(garage, service, &authorization, string_to_sign.as_bytes()).await?;
+	let key = verify_v4(garage, service, &authorization, string_to_sign.as_bytes())?;
 
 	// In the page on presigned URLs, AWS specifies that if a signed query
 	// parameter and a signed header of the same name have different values,
@@ -268,9 +269,7 @@ fn verify_signed_headers(headers: &HeaderMap, signed_headers: &[HeaderName]) ->
 		return Err(Error::bad_request("Header `Host` should be signed"));
 	}
 	for (name, _) in headers.iter() {
-		// Enforce signature of all x-amz-* headers, except x-amz-content-sh256
-		// because it is included in the canonical request in all cases
-		if name.as_str().starts_with("x-amz-") && name != X_AMZ_CONTENT_SHA256 {
+		if name.as_str().starts_with("x-amz-") {
 			if !signed_headers.contains(name) {
 				return Err(Error::bad_request(format!(
 					"Header `{}` should be signed",
@@ -380,7 +379,7 @@ pub fn parse_date(date: &str) -> Result<DateTime<Utc>, Error> {
 	Ok(Utc.from_utc_datetime(&date))
 }
 
-pub async fn verify_v4(
+pub fn verify_v4(
 	garage: &Garage,
 	service: &str,
 	auth: &Authorization,
@@ -393,12 +392,18 @@ pub async fn verify_v4(
 
 	let key = garage
 		.key_table
-		.get(&EmptyKey, &auth.key_id)
-		.await?
+		.get_local(&EmptyKey, &auth.key_id)?
 		.filter(|k| !k.state.is_deleted())
 		.ok_or_else(|| Error::forbidden(format!("No such key: {}", &auth.key_id)))?;
 	let key_p = key.params().unwrap();
 
+	if key_p.is_expired(now_msec()) {
+		return Err(Error::forbidden(format!(
+			"Access key {} has expired",
+			key.key_id
+		)));
+	}
+
 	let mut hmac = signing_hmac(
 		&auth.date,
 		&key_p.secret_key,
@@ -419,7 +424,7 @@ pub async fn verify_v4(
 // ============ Authorization header, or X-Amz-* query params =========
 
 pub struct Authorization {
-	pub key_id: String,
+	key_id: String,
 	scope: String,
 	signed_headers: String,
 	signature: String,
@@ -428,7 +433,7 @@ pub struct Authorization {
 }
 
 impl Authorization {
-	pub fn parse_header(headers: &HeaderMap) -> Result<Self, Error> {
+	fn parse_header(headers: &HeaderMap) -> Result<Self, Error> {
 		let authorization = headers
 			.get(AUTHORIZATION)
 			.ok_or_bad_request("Missing authorization header")?
@@ -470,7 +475,8 @@ impl Authorization {
 
 		let date = headers
 			.get(X_AMZ_DATE)
-			.ok_or_bad_request("Missing X-Amz-Date field")?
+			.ok_or_bad_request("Missing X-Amz-Date field")
+			.map_err(Error::from)?
 			.to_str()?;
 		let date = parse_date(date)?;
 

src/api/k2v/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "garage_api_k2v"
-version = "1.3.1"
+version = "2.0.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -20,7 +20,7 @@ garage_util = { workspace = true, features = [ "k2v" ] }
 garage_api_common.workspace = true
 
 base64.workspace = true
-thiserror.workspace = true
+err-derive.workspace = true
 tracing.workspace = true
 
 futures.workspace = true

src/api/k2v/api_server.rs

@@ -1,3 +1,4 @@
+use std::borrow::Cow;
 use std::sync::Arc;
 
 use hyper::{body::Incoming as IncomingBody, Method, Request, Response};
@@ -76,25 +77,19 @@ impl ApiHandler for K2VApiServer {
 		// The OPTIONS method is processed early, before we even check for an API key
 		if let Endpoint::Options = endpoint {
 			let options_res = handle_options_api(garage, &req, Some(bucket_name))
-				.await
 				.ok_or_bad_request("Error handling OPTIONS")?;
 			return Ok(options_res.map(|_empty_body: EmptyBody| empty_body()));
 		}
 
-		let verified_request = verify_request(&garage, req, "k2v").await?;
+		let verified_request = verify_request(&garage, req, "k2v")?;
 		let req = verified_request.request;
 		let api_key = verified_request.access_key;
 
-		let bucket_id = garage
-			.bucket_helper()
-			.resolve_bucket(&bucket_name, &api_key)
-			.await
-			.map_err(pass_helper_error)?;
 		let bucket = garage
 			.bucket_helper()
-			.get_existing_bucket(bucket_id)
-			.await
-			.map_err(helper_error_as_internal)?;
+			.resolve_bucket_fast(&bucket_name, &api_key)
+			.map_err(pass_helper_error)?;
+		let bucket_id = bucket.id;
 		let bucket_params = bucket.state.into_option().unwrap();
 
 		let allowed = match endpoint.authorization_type() {
@@ -176,17 +171,11 @@ impl ApiHandler for K2VApiServer {
 
 		Ok(resp_ok)
 	}
-
-	fn key_id_from_request(&self, req: &Request<IncomingBody>) -> Option<String> {
-		garage_api_common::signature::payload::Authorization::parse_header(req.headers())
-			.map(|auth| auth.key_id)
-			.ok()
-	}
 }
 
 impl ApiEndpoint for K2VApiEndpoint {
-	fn name(&self) -> &'static str {
-		self.endpoint.name()
+	fn name(&self) -> Cow<'static, str> {
+		Cow::Borrowed(self.endpoint.name())
 	}
 
 	fn add_span_attributes(&self, span: SpanRef<'_>) {

src/api/k2v/error.rs

@@ -1,9 +1,9 @@
+use err_derive::Error;
 use hyper::header::HeaderValue;
 use hyper::{HeaderMap, StatusCode};
-use thiserror::Error;
 
+pub(crate) use garage_api_common::common_error::pass_helper_error;
 use garage_api_common::common_error::{commonErrorDerivative, CommonError};
-pub(crate) use garage_api_common::common_error::{helper_error_as_internal, pass_helper_error};
 pub use garage_api_common::common_error::{
 	CommonErrorDerivative, OkOrBadRequest, OkOrInternalError,
 };
@@ -14,38 +14,38 @@ use garage_api_common::signature::error::Error as SignatureError;
 /// Errors of this crate
 #[derive(Debug, Error)]
 pub enum Error {
-	#[error("{0}")]
+	#[error(display = "{}", _0)]
 	/// Error from common error
-	Common(#[from] CommonError),
+	Common(#[error(source)] CommonError),
 
 	// Category: cannot process
 	/// Authorization Header Malformed
-	#[error("Authorization header malformed, unexpected scope: {0}")]
+	#[error(display = "Authorization header malformed, unexpected scope: {}", _0)]
 	AuthorizationHeaderMalformed(String),
 
 	/// The provided digest (checksum) value was invalid
-	#[error("Invalid digest: {0}")]
+	#[error(display = "Invalid digest: {}", _0)]
 	InvalidDigest(String),
 
 	/// The object requested don't exists
-	#[error("Key not found")]
+	#[error(display = "Key not found")]
 	NoSuchKey,
 
 	/// Some base64 encoded data was badly encoded
-	#[error("Invalid base64: {0}")]
-	InvalidBase64(#[from] base64::DecodeError),
+	#[error(display = "Invalid base64: {}", _0)]
+	InvalidBase64(#[error(source)] base64::DecodeError),
 
 	/// Invalid causality token
-	#[error("Invalid causality token")]
+	#[error(display = "Invalid causality token")]
 	InvalidCausalityToken,
 
 	/// The client asked for an invalid return format (invalid Accept header)
-	#[error("Not acceptable: {0}")]
+	#[error(display = "Not acceptable: {}", _0)]
 	NotAcceptable(String),
 
 	/// The request contained an invalid UTF-8 sequence in its path or in other parameters
-	#[error("Invalid UTF-8: {0}")]
-	InvalidUtf8Str(#[from] std::str::Utf8Error),
+	#[error(display = "Invalid UTF-8: {}", _0)]
+	InvalidUtf8Str(#[error(source)] std::str::Utf8Error),
 }
 
 commonErrorDerivative!(Error);

src/api/s3/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "garage_api_s3"
-version = "1.3.1"
+version = "2.0.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -29,8 +29,10 @@ bytes.workspace = true
 chrono.workspace = true
 crc32fast.workspace = true
 crc32c.workspace = true
-thiserror.workspace = true
+crc64fast-nvme.workspace = true
+err-derive.workspace = true
 hex.workspace = true
+hmac.workspace = true
 tracing.workspace = true
 md-5.workspace = true
 pin-project.workspace = true

src/api/s3/api_server.rs

@@ -1,3 +1,4 @@
+use std::borrow::Cow;
 use std::sync::Arc;
 
 use hyper::header;
@@ -117,11 +118,11 @@ impl ApiHandler for S3ApiServer {
 			return handle_post_object(garage, req, bucket_name.unwrap()).await;
 		}
 		if let Endpoint::Options = endpoint {
-			let options_res = handle_options_api(garage, &req, bucket_name).await?;
+			let options_res = handle_options_api(garage, &req, bucket_name)?;
 			return Ok(options_res.map(|_empty_body: EmptyBody| empty_body()));
 		}
 
-		let verified_request = verify_request(&garage, req, "s3").await?;
+		let verified_request = verify_request(&garage, req, "s3")?;
 		let req = verified_request.request;
 		let api_key = verified_request.access_key;
 
@@ -139,15 +140,11 @@ impl ApiHandler for S3ApiServer {
 			return handle_create_bucket(&garage, req, &api_key.key_id, bucket_name).await;
 		}
 
-		let bucket_id = garage
-			.bucket_helper()
-			.resolve_bucket(&bucket_name, &api_key)
-			.await
-			.map_err(pass_helper_error)?;
 		let bucket = garage
 			.bucket_helper()
-			.get_existing_bucket(bucket_id)
-			.await?;
+			.resolve_bucket_fast(&bucket_name, &api_key)
+			.map_err(pass_helper_error)?;
+		let bucket_id = bucket.id;
 		let bucket_params = bucket.state.into_option().unwrap();
 
 		let allowed = match endpoint.authorization_type() {
@@ -226,7 +223,6 @@ impl ApiHandler for S3ApiServer {
 			Endpoint::DeleteBucket {} => handle_delete_bucket(ctx).await,
 			Endpoint::GetBucketLocation {} => handle_get_bucket_location(ctx),
 			Endpoint::GetBucketVersioning {} => handle_get_bucket_versioning(),
-			Endpoint::GetBucketAcl {} => handle_get_bucket_acl(ctx),
 			Endpoint::ListObjects {
 				delimiter,
 				encoding_type,
@@ -343,17 +339,11 @@ impl ApiHandler for S3ApiServer {
 
 		Ok(resp_ok)
 	}
-
-	fn key_id_from_request(&self, req: &Request<IncomingBody>) -> Option<String> {
-		garage_api_common::signature::payload::Authorization::parse_header(req.headers())
-			.map(|auth| auth.key_id)
-			.ok()
-	}
 }
 
 impl ApiEndpoint for S3ApiEndpoint {
-	fn name(&self) -> &'static str {
-		self.endpoint.name()
+	fn name(&self) -> Cow<'static, str> {
+		Cow::Borrowed(self.endpoint.name())
 	}
 
 	fn add_span_attributes(&self, span: SpanRef<'_>) {

src/api/s3/bucket.rs

@@ -5,7 +5,7 @@ use hyper::{Request, Response, StatusCode};
 use garage_model::bucket_alias_table::*;
 use garage_model::bucket_table::Bucket;
 use garage_model::garage::Garage;
-use garage_model::key_table::{Key, KeyParams};
+use garage_model::key_table::Key;
 use garage_model::permission::BucketKeyPerm;
 use garage_table::util::*;
 use garage_util::crdt::*;
@@ -44,55 +44,6 @@ pub fn handle_get_bucket_versioning() -> Result<Response<ResBody>, Error> {
 		.body(string_body(xml))?)
 }
 
-pub fn handle_get_bucket_acl(ctx: ReqCtx) -> Result<Response<ResBody>, Error> {
-	let ReqCtx {
-		bucket_id, api_key, ..
-	} = ctx;
-	let key_p = api_key.params().ok_or_internal_error(
-		"Key should not be in deleted state at this point (in handle_get_bucket_acl)",
-	)?;
-
-	let mut grants: Vec<s3_xml::Grant> = vec![];
-	let kp = api_key.bucket_permissions(&bucket_id);
-
-	if kp.allow_owner {
-		grants.push(s3_xml::Grant {
-			grantee: create_grantee(&key_p, &api_key),
-			permission: s3_xml::Value("FULL_CONTROL".to_string()),
-		});
-	} else {
-		if kp.allow_read {
-			grants.push(s3_xml::Grant {
-				grantee: create_grantee(&key_p, &api_key),
-				permission: s3_xml::Value("READ".to_string()),
-			});
-			grants.push(s3_xml::Grant {
-				grantee: create_grantee(&key_p, &api_key),
-				permission: s3_xml::Value("READ_ACP".to_string()),
-			});
-		}
-		if kp.allow_write {
-			grants.push(s3_xml::Grant {
-				grantee: create_grantee(&key_p, &api_key),
-				permission: s3_xml::Value("WRITE".to_string()),
-			});
-		}
-	}
-
-	let access_control_policy = s3_xml::AccessControlPolicy {
-		xmlns: (),
-		owner: None,
-		acl: s3_xml::AccessControlList { entries: grants },
-	};
-
-	let xml = s3_xml::to_xml_with_header(&access_control_policy)?;
-	trace!("xml: {}", xml);
-
-	Ok(Response::builder()
-		.header("Content-Type", "application/xml")
-		.body(string_body(xml))?)
-}
-
 pub async fn handle_list_buckets(
 	garage: &Garage,
 	api_key: &Key,
@@ -192,21 +143,16 @@ pub async fn handle_create_bucket(
 	let api_key = helper.key().get_existing_key(api_key_id).await?;
 	let key_params = api_key.params().unwrap();
 
-	let existing_bucket = if let Some(Some(bucket_id)) = key_params.local_aliases.get(&bucket_name)
-	{
-		Some(*bucket_id)
-	} else {
-		helper
+	let existing_bucket = helper
 		.bucket()
-			.resolve_global_bucket_name(&bucket_name)
-			.await?
-	};
+		.resolve_bucket(&bucket_name, &api_key.key_id)
+		.await?;
 
-	if let Some(bucket_id) = existing_bucket {
+	if let Some(bucket) = existing_bucket {
 		// Check we have write or owner permission on the bucket,
 		// in that case it's fine, return 200 OK, bucket exists;
 		// otherwise return a forbidden error.
-		let kp = api_key.bucket_permissions(&bucket_id);
+		let kp = api_key.bucket_permissions(&bucket.id);
 		if !(kp.allow_write || kp.allow_owner) {
 			return Err(CommonError::BucketAlreadyExists.into());
 		}
@@ -221,7 +167,7 @@ pub async fn handle_create_bucket(
 		}
 
 		// Create the bucket!
-		if !is_valid_bucket_name(&bucket_name, garage.config.allow_punycode) {
+		if !is_valid_bucket_name(&bucket_name) {
 			return Err(Error::bad_request(format!(
 				"{}: {}",
 				bucket_name, INVALID_BUCKET_NAME_MESSAGE
@@ -290,11 +236,11 @@ pub async fn handle_delete_bucket(ctx: ReqCtx) -> Result<Response<ResBody>, Erro
 		// 1. delete bucket alias
 		if is_local_alias {
 			helper
-				.purge_local_bucket_alias(*bucket_id, &api_key.key_id, bucket_name)
+				.unset_local_bucket_alias(*bucket_id, &api_key.key_id, bucket_name)
 				.await?;
 		} else {
 			helper
-				.purge_global_bucket_alias(*bucket_id, bucket_name)
+				.unset_global_bucket_alias(*bucket_id, bucket_name)
 				.await?;
 		}
 
@@ -360,15 +306,6 @@ fn parse_create_bucket_xml(xml_bytes: &[u8]) -> Option<Option<String>> {
 	Some(ret)
 }
 
-fn create_grantee(key_params: &KeyParams, api_key: &Key) -> s3_xml::Grantee {
-	s3_xml::Grantee {
-		xmlns_xsi: (),
-		typ: "CanonicalUser".to_string(),
-		display_name: Some(s3_xml::Value(key_params.name.get().to_string())),
-		id: Some(s3_xml::Value(api_key.key_id.to_string())),
-	}
-}
-
 #[cfg(test)]
 mod tests {
 	use super::*;

src/api/s3/copy.rs

@@ -24,12 +24,11 @@ use garage_api_common::helpers::*;
 use garage_api_common::signature::checksum::*;
 
 use crate::api_server::{ReqBody, ResBody};
-use crate::encryption::EncryptionParams;
+use crate::encryption::{EncryptionParams, OekDerivationInfo};
 use crate::error::*;
-use crate::get::{check_version_not_deleted, full_object_byte_stream, PreconditionHeaders};
+use crate::get::{full_object_byte_stream, PreconditionHeaders};
 use crate::multipart;
 use crate::put::{extract_metadata_headers, save_stream, ChecksumMode, SaveStreamResult};
-use crate::website::X_AMZ_WEBSITE_REDIRECT_LOCATION;
 use crate::xml::{self as s3_xml, xmlns_tag};
 
 pub const X_AMZ_COPY_SOURCE_IF_MATCH: HeaderName =
@@ -66,8 +65,18 @@ pub async fn handle_copy(
 			&ctx.garage,
 			req.headers(),
 			&source_version_meta.encryption,
+			OekDerivationInfo::for_object(&source_object, source_version),
+		)?;
+	let dest_uuid = gen_uuid();
+	let dest_encryption = EncryptionParams::new_from_headers(
+		&ctx.garage,
+		req.headers(),
+		OekDerivationInfo {
+			bucket_id: ctx.bucket_id,
+			version_id: dest_uuid,
+			object_key: dest_key,
+		},
 	)?;
-	let dest_encryption = EncryptionParams::new_from_headers(&ctx.garage, req.headers())?;
 
 	// Extract source checksum info before source_object_meta_inner is consumed
 	let source_checksum = source_object_meta_inner.checksum;
@@ -85,18 +94,7 @@ pub async fn handle_copy(
 			Some(v) if v == hyper::header::HeaderValue::from_static("REPLACE") => {
 				extract_metadata_headers(req.headers())?
 			}
-			_ => {
-				// The x-amz-website-redirect-location header is not copied, instead
-				// it is replaced by the value from the request (or removed if no
-				// value was specified)
-				let is_redirect =
-					|(key, _): &(String, String)| key == X_AMZ_WEBSITE_REDIRECT_LOCATION.as_str();
-				let mut headers: Vec<_> = source_object_meta_inner.headers.clone();
-				headers.retain(|h| !is_redirect(h));
-				let new_headers = extract_metadata_headers(req.headers())?;
-				headers.extend(new_headers.into_iter().filter(is_redirect));
-				headers
-			}
+			_ => source_object_meta_inner.into_owned().headers,
 		},
 		checksum: source_checksum,
 	};
@@ -127,6 +125,7 @@ pub async fn handle_copy(
 		handle_copy_metaonly(
 			ctx,
 			dest_key,
+			dest_uuid,
 			dest_object_meta,
 			dest_encryption,
 			source_version,
@@ -150,6 +149,7 @@ pub async fn handle_copy(
 		handle_copy_reencrypt(
 			ctx,
 			dest_key,
+			dest_uuid,
 			dest_object_meta,
 			dest_encryption,
 			source_version,
@@ -181,6 +181,7 @@ pub async fn handle_copy(
 async fn handle_copy_metaonly(
 	ctx: ReqCtx,
 	dest_key: &str,
+	dest_uuid: Uuid,
 	dest_object_meta: ObjectVersionMetaInner,
 	dest_encryption: EncryptionParams,
 	source_version: &ObjectVersion,
@@ -194,7 +195,6 @@ async fn handle_copy_metaonly(
 	} = ctx;
 
 	// Generate parameters for copied object
-	let new_uuid = gen_uuid();
 	let new_timestamp = now_msec();
 
 	let new_meta = ObjectVersionMeta {
@@ -204,7 +204,7 @@ async fn handle_copy_metaonly(
 	};
 
 	let res = SaveStreamResult {
-		version_uuid: new_uuid,
+		version_uuid: dest_uuid,
 		version_timestamp: new_timestamp,
 		etag: new_meta.etag.clone(),
 	};
@@ -216,7 +216,7 @@ async fn handle_copy_metaonly(
 			// bytes is either plaintext before&after or encrypted with the
 			// same keys, so it's ok to just copy it as is
 			let dest_object_version = ObjectVersion {
-				uuid: new_uuid,
+				uuid: dest_uuid,
 				timestamp: new_timestamp,
 				state: ObjectVersionState::Complete(ObjectVersionData::Inline(
 					new_meta,
@@ -237,13 +237,12 @@ async fn handle_copy_metaonly(
 				.get(&source_version.uuid, &EmptyKey)
 				.await?;
 			let source_version = source_version.ok_or(Error::NoSuchKey)?;
-			check_version_not_deleted(&source_version)?;
 
 			// Write an "uploading" marker in Object table
 			// This holds a reference to the object in the Version table
 			// so that it won't be deleted, e.g. by repair_versions.
 			let tmp_dest_object_version = ObjectVersion {
-				uuid: new_uuid,
+				uuid: dest_uuid,
 				timestamp: new_timestamp,
 				state: ObjectVersionState::Uploading {
 					encryption: new_meta.encryption.clone(),
@@ -263,7 +262,7 @@ async fn handle_copy_metaonly(
 			// marked as deleted (they are marked as deleted only if the Version
 			// doesn't exist or is marked as deleted).
 			let mut dest_version = Version::new(
-				new_uuid,
+				dest_uuid,
 				VersionBacklink::Object {
 					bucket_id: dest_bucket_id,
 					key: dest_key.to_string(),
@@ -282,7 +281,7 @@ async fn handle_copy_metaonly(
 				.iter()
 				.map(|b| BlockRef {
 					block: b.1.hash,
-					version: new_uuid,
+					version: dest_uuid,
 					deleted: false.into(),
 				})
 				.collect::<Vec<_>>();
@@ -298,7 +297,7 @@ async fn handle_copy_metaonly(
 			// with the stuff before, the block's reference counts could be decremented before
 			// they are incremented again for the new version, leading to data being deleted.
 			let dest_object_version = ObjectVersion {
-				uuid: new_uuid,
+				uuid: dest_uuid,
 				timestamp: new_timestamp,
 				state: ObjectVersionState::Complete(ObjectVersionData::FirstBlock(
 					new_meta,
@@ -320,6 +319,7 @@ async fn handle_copy_metaonly(
 async fn handle_copy_reencrypt(
 	ctx: ReqCtx,
 	dest_key: &str,
+	dest_uuid: Uuid,
 	dest_object_meta: ObjectVersionMetaInner,
 	dest_encryption: EncryptionParams,
 	source_version: &ObjectVersion,
@@ -339,6 +339,7 @@ async fn handle_copy_reencrypt(
 
 	save_stream(
 		&ctx,
+		dest_uuid,
 		dest_object_meta,
 		dest_encryption,
 		source_stream.map_err(|e| Error::from(GarageError::from(e))),
@@ -362,7 +363,7 @@ pub async fn handle_upload_part_copy(
 	let dest_upload_id = multipart::decode_upload_id(upload_id)?;
 
 	let dest_key = dest_key.to_string();
-	let (source_object, (_, dest_version, mut dest_mpu)) = futures::try_join!(
+	let (source_object, (dest_object, dest_version, mut dest_mpu)) = futures::try_join!(
 		get_copy_source(&ctx, req),
 		multipart::get_upload(&ctx, &dest_key, &dest_upload_id)
 	)?;
@@ -380,7 +381,10 @@ pub async fn handle_upload_part_copy(
 		&garage,
 		req.headers(),
 		&source_version_meta.encryption,
+		OekDerivationInfo::for_object(&source_object, source_object_version),
 	)?;
+
+	let dest_oek_params = OekDerivationInfo::for_object(&dest_object, &dest_version);
 	let (dest_object_encryption, dest_object_checksum_algorithm) = match dest_version.state {
 		ObjectVersionState::Uploading {
 			encryption,
@@ -389,8 +393,12 @@ pub async fn handle_upload_part_copy(
 		} => (encryption, checksum_algorithm),
 		_ => unreachable!(),
 	};
-	let (dest_encryption, _) =
-		EncryptionParams::check_decrypt(&garage, req.headers(), &dest_object_encryption)?;
+	let (dest_encryption, _) = EncryptionParams::check_decrypt(
+		&garage,
+		req.headers(),
+		&dest_object_encryption,
+		dest_oek_params,
+	)?;
 	let same_encryption = EncryptionParams::is_same(&source_encryption, &dest_encryption);
 
 	// Check source range is valid
@@ -429,7 +437,6 @@ pub async fn handle_upload_part_copy(
 		.get(&source_object_version.uuid, &EmptyKey)
 		.await?
 		.ok_or(Error::NoSuchKey)?;
-	check_version_not_deleted(&source_version)?;
 
 	// We want to reuse blocks from the source version as much as possible.
 	// However, we still need to get the data from these blocks
@@ -561,7 +568,6 @@ pub async fn handle_upload_part_copy(
 
 	let mut current_offset = 0;
 	let mut next_block = defragmenter.next().await?;
-	let mut blocks_to_dup = dest_version.clone();
 
 	// TODO this could be optimized similarly to read_and_put_blocks
 	// low priority because uploadpartcopy is rarely used
@@ -591,7 +597,8 @@ pub async fn handle_upload_part_copy(
 			.unwrap()?;
 		checksummer = checksummer_updated;
 
-		let (version_block_key, version_block) = (
+		dest_version.blocks.clear();
+		dest_version.blocks.put(
 			VersionBlockKey {
 				part_number,
 				offset: current_offset,
@@ -603,23 +610,25 @@ pub async fn handle_upload_part_copy(
 		);
 		current_offset += data_len;
 
-		let next = if let Some(final_data) = data_to_upload {
-			dest_version.blocks.clear();
-			dest_version.blocks.put(version_block_key, version_block);
 		let block_ref = BlockRef {
 			block: final_hash,
 			version: dest_version_id,
 			deleted: false.into(),
 		};
+
 		let (_, _, _, next) = futures::try_join!(
 			// Thing 1: if the block is not exactly a block that existed before,
 			// we need to insert that data as a new block.
-				garage.block_manager.rpc_put_block(
-					final_hash,
-					final_data,
-					dest_encryption.is_encrypted(),
-					None
-				),
+			async {
+				if let Some(final_data) = data_to_upload {
+					garage
+						.block_manager
+						.rpc_put_block(final_hash, final_data, dest_encryption.is_encrypted(), None)
+						.await
+				} else {
+					Ok(())
+				}
+			},
 			// Thing 2: we need to insert the block in the version
 			garage.version_table.insert(&dest_version),
 			// Thing 3: we need to add a block reference
@@ -627,32 +636,11 @@ pub async fn handle_upload_part_copy(
 			// Thing 4: we need to read the next block
 			defragmenter.next(),
 		)?;
-			next
-		} else {
-			blocks_to_dup.blocks.put(version_block_key, version_block);
-			defragmenter.next().await?
-		};
 		next_block = next;
 	}
 
 	assert_eq!(current_offset, source_range.length);
 
-	// Put the duplicated blocks into the version & block_refs tables
-	let block_refs_to_put = blocks_to_dup
-		.blocks
-		.items()
-		.iter()
-		.map(|b| BlockRef {
-			block: b.1.hash,
-			version: dest_version_id,
-			deleted: false.into(),
-		})
-		.collect::<Vec<_>>();
-	futures::try_join!(
-		garage.version_table.insert(&blocks_to_dup),
-		garage.block_ref_table.insert_many(&block_refs_to_put[..]),
-	)?;
-
 	let checksums = checksummer.finalize();
 	let etag = dest_encryption.etag_from_md5(&checksums.md5);
 	let checksum = checksums.extract(dest_object_checksum_algorithm);
@@ -695,16 +683,15 @@ async fn get_copy_source(ctx: &ReqCtx, req: &Request<ReqBody>) -> Result<Object,
 	let copy_source = percent_encoding::percent_decode_str(copy_source).decode_utf8()?;
 
 	let (source_bucket, source_key) = parse_bucket_key(&copy_source, None)?;
-	let source_bucket_id = garage
+	let source_bucket = garage
 		.bucket_helper()
-		.resolve_bucket(&source_bucket.to_string(), api_key)
-		.await
+		.resolve_bucket_fast(&source_bucket.to_string(), api_key)
 		.map_err(pass_helper_error)?;
 
-	if !api_key.allow_read(&source_bucket_id) {
+	if !api_key.allow_read(&source_bucket.id) {
 		return Err(Error::forbidden(format!(
-			"Reading from bucket {} not allowed for this key",
-			source_bucket
+			"Reading from bucket {:?} not allowed for this key",
+			source_bucket.id
 		)));
 	}
 
@@ -712,7 +699,7 @@ async fn get_copy_source(ctx: &ReqCtx, req: &Request<ReqBody>) -> Result<Object,
 
 	let source_object = garage
 		.object_table
-		.get(&source_bucket_id, &source_key.to_string())
+		.get(&source_bucket.id, &source_key.to_string())
 		.await?
 		.ok_or(Error::NoSuchKey)?;
 

src/api/s3/cors.rs

@@ -88,9 +88,7 @@ pub async fn handle_put_cors(
 pub struct CorsConfiguration {
 	#[serde(serialize_with = "xmlns_tag", skip_deserializing)]
 	pub xmlns: (),
-	// "default" is required to be able to parse an empty list of rules,
-	// cf https://docs.rs/quick-xml/latest/quick_xml/de/#sequences-xsall-and-xssequence-xml-schema-types
-	#[serde(rename = "CORSRule", default)]
+	#[serde(rename = "CORSRule")]
 	pub cors_rules: Vec<CorsRule>,
 }
 
@@ -272,26 +270,4 @@ mod tests {
 
 		Ok(())
 	}
-
-	#[test]
-	fn test_deserialize_norules() -> Result<(), Error> {
-		let message = r#"<?xml version="1.0" encoding="UTF-8"?>
-<CORSConfiguration xmlns="http://s3.amazonaws.com/doc/2006-03-01/" />"#;
-		let conf: CorsConfiguration = from_str(message).unwrap();
-		let ref_value = CorsConfiguration {
-			xmlns: (),
-			cors_rules: vec![],
-		};
-		assert_eq! {
-			ref_value,
-			conf
-		};
-
-		let message2 = to_xml_with_header(&ref_value)?;
-
-		let cleanup = |c: &str| c.replace(char::is_whitespace, "");
-		assert_eq!(cleanup(message), cleanup(&message2));
-
-		Ok(())
-	}
 }

src/api/s3/encryption.rs

@@ -11,6 +11,7 @@ use aes_gcm::{
 };
 use base64::prelude::*;
 use bytes::Bytes;
+use sha2::Sha256;
 
 use futures::stream::Stream;
 use futures::task;
@@ -21,12 +22,12 @@ use http::header::{HeaderMap, HeaderName, HeaderValue};
 use garage_net::bytes_buf::BytesBuf;
 use garage_net::stream::{stream_asyncread, ByteStream};
 use garage_rpc::rpc_helper::OrderTag;
-use garage_util::data::Hash;
+use garage_util::data::{Hash, Uuid};
 use garage_util::error::Error as GarageError;
 use garage_util::migrate::Migrate;
 
 use garage_model::garage::Garage;
-use garage_model::s3::object_table::{ObjectVersionEncryption, ObjectVersionMetaInner};
+use garage_model::s3::object_table::*;
 
 use garage_api_common::common_error::*;
 use garage_api_common::signature::checksum::Md5Checksum;
@@ -64,32 +65,45 @@ const STREAM_ENC_CYPER_CHUNK_SIZE: usize = STREAM_ENC_PLAIN_CHUNK_SIZE + 16;
 pub enum EncryptionParams {
 	Plaintext,
 	SseC {
+		/// the value of x-amz-server-side-encryption-customer-key
 		client_key: Key<Aes256Gcm>,
+		/// the value of x-amz-server-side-encryption-customer-key-md5
 		client_key_md5: Md5Output,
+		/// the object encryption key, for uploads created in garage v2+
+		object_key: Option<Key<Aes256Gcm>>,
+		/// the compression level used for compressing data blocks
 		compression_level: Option<i32>,
 	},
 }
 
+#[derive(Clone, Copy)]
+pub struct OekDerivationInfo<'a> {
+	pub bucket_id: Uuid,
+	pub version_id: Uuid,
+	pub object_key: &'a str,
+}
+
 impl EncryptionParams {
 	pub fn is_encrypted(&self) -> bool {
 		!matches!(self, Self::Plaintext)
 	}
 
 	pub fn is_same(a: &Self, b: &Self) -> bool {
-		let relevant_info = |x: &Self| match x {
-			Self::Plaintext => None,
-			Self::SseC {
-				client_key,
-				compression_level,
-				..
-			} => Some((*client_key, compression_level.is_some())),
-		};
-		relevant_info(a) == relevant_info(b)
+		// This function is used in CopyObject and UploadPartCopy to determine
+		// whether the object must be re-encrypted. If this returns true,
+		// data blocks are reused as-is. Since Garage v2, we are using
+		// object-specific encryption keys, so we know that if both source
+		// and destination are encrypted, it can't be with the same key.
+		match (a, b) {
+			(Self::Plaintext, Self::Plaintext) => true,
+			_ => false,
+		}
 	}
 
 	pub fn new_from_headers(
 		garage: &Garage,
 		headers: &HeaderMap,
+		oek_info: OekDerivationInfo<'_>,
 	) -> Result<EncryptionParams, Error> {
 		let key = parse_request_headers(
 			headers,
@@ -101,6 +115,7 @@ impl EncryptionParams {
 			Some((client_key, client_key_md5)) => Ok(EncryptionParams::SseC {
 				client_key,
 				client_key_md5,
+				object_key: Some(oek_info.derive_oek(&client_key)),
 				compression_level: garage.config.compression_level,
 			}),
 			None => Ok(EncryptionParams::Plaintext),
@@ -126,6 +141,7 @@ impl EncryptionParams {
 		garage: &Garage,
 		headers: &HeaderMap,
 		obj_enc: &'a ObjectVersionEncryption,
+		oek_info: OekDerivationInfo<'_>,
 	) -> Result<(Self, Cow<'a, ObjectVersionMetaInner>), Error> {
 		let key = parse_request_headers(
 			headers,
@@ -133,13 +149,14 @@ impl EncryptionParams {
 			&X_AMZ_SERVER_SIDE_ENCRYPTION_CUSTOMER_KEY,
 			&X_AMZ_SERVER_SIDE_ENCRYPTION_CUSTOMER_KEY_MD5,
 		)?;
-		Self::check_decrypt_common(garage, key, obj_enc)
+		Self::check_decrypt_common(garage, key, obj_enc, oek_info)
 	}
 
 	pub fn check_decrypt_for_copy_source<'a>(
 		garage: &Garage,
 		headers: &HeaderMap,
 		obj_enc: &'a ObjectVersionEncryption,
+		oek_info: OekDerivationInfo<'_>,
 	) -> Result<(Self, Cow<'a, ObjectVersionMetaInner>), Error> {
 		let key = parse_request_headers(
 			headers,
@@ -147,22 +164,32 @@ impl EncryptionParams {
 			&X_AMZ_COPY_SOURCE_SERVER_SIDE_ENCRYPTION_CUSTOMER_KEY,
 			&X_AMZ_COPY_SOURCE_SERVER_SIDE_ENCRYPTION_CUSTOMER_KEY_MD5,
 		)?;
-		Self::check_decrypt_common(garage, key, obj_enc)
+		Self::check_decrypt_common(garage, key, obj_enc, oek_info)
 	}
 
 	fn check_decrypt_common<'a>(
 		garage: &Garage,
 		key: Option<(Key<Aes256Gcm>, Md5Output)>,
 		obj_enc: &'a ObjectVersionEncryption,
+		oek_info: OekDerivationInfo<'_>,
 	) -> Result<(Self, Cow<'a, ObjectVersionMetaInner>), Error> {
 		match (key, &obj_enc) {
 			(
 				Some((client_key, client_key_md5)),
-				ObjectVersionEncryption::SseC { inner, compressed },
+				ObjectVersionEncryption::SseC {
+					inner,
+					compressed,
+					use_oek,
+				},
 			) => {
 				let enc = Self::SseC {
 					client_key,
 					client_key_md5,
+					object_key: if *use_oek {
+						Some(oek_info.derive_oek(&client_key))
+					} else {
+						None
+					},
 					compression_level: if *compressed {
 						Some(garage.config.compression_level.unwrap_or(1))
 					} else {
@@ -193,13 +220,16 @@ impl EncryptionParams {
 	) -> Result<ObjectVersionEncryption, Error> {
 		match self {
 			Self::SseC {
-				compression_level, ..
+				compression_level,
+				object_key,
+				..
 			} => {
 				let plaintext = meta.encode().map_err(GarageError::from)?;
 				let ciphertext = self.encrypt_blob(&plaintext)?;
 				Ok(ObjectVersionEncryption::SseC {
 					inner: ciphertext.into_owned(),
 					compressed: compression_level.is_some(),
+					use_oek: object_key.is_some(),
 				})
 			}
 			Self::Plaintext => Ok(ObjectVersionEncryption::Plaintext { inner: meta }),
@@ -228,24 +258,37 @@ impl EncryptionParams {
 	// This is used for encrypting object metadata and inlined data for small objects.
 	// This does not compress anything.
 
-	pub fn encrypt_blob<'a>(&self, blob: &'a [u8]) -> Result<Cow<'a, [u8]>, Error> {
+	fn cipher(&self) -> Option<Aes256Gcm> {
 		match self {
-			Self::SseC { client_key, .. } => {
-				let cipher = Aes256Gcm::new(&client_key);
+			Self::SseC {
+				object_key: Some(oek),
+				..
+			} => Some(Aes256Gcm::new(&oek)),
+			Self::SseC {
+				client_key,
+				object_key: None,
+				..
+			} => Some(Aes256Gcm::new(&client_key)),
+			Self::Plaintext => None,
+		}
+	}
+
+	pub fn encrypt_blob<'a>(&self, blob: &'a [u8]) -> Result<Cow<'a, [u8]>, Error> {
+		match self.cipher() {
+			Some(cipher) => {
 				let nonce = Aes256Gcm::generate_nonce(&mut OsRng);
 				let ciphertext = cipher
 					.encrypt(&nonce, blob)
 					.ok_or_internal_error("Encryption failed")?;
 				Ok(Cow::Owned([nonce.to_vec(), ciphertext].concat()))
 			}
-			Self::Plaintext => Ok(Cow::Borrowed(blob)),
+			None => Ok(Cow::Borrowed(blob)),
 		}
 	}
 
 	pub fn decrypt_blob<'a>(&self, blob: &'a [u8]) -> Result<Cow<'a, [u8]>, Error> {
-		match self {
-			Self::SseC { client_key, .. } => {
-				let cipher = Aes256Gcm::new(&client_key);
+		match self.cipher() {
+			Some(cipher) => {
 				let nonce_size = <Aes256Gcm as AeadCore>::NonceSize::to_usize();
 				let nonce = Nonce::from_slice(
 					blob.get(..nonce_size)
@@ -258,7 +301,7 @@ impl EncryptionParams {
 					)?;
 				Ok(Cow::Owned(plaintext))
 			}
-			Self::Plaintext => Ok(Cow::Borrowed(blob)),
+			None => Ok(Cow::Borrowed(blob)),
 		}
 	}
 
@@ -284,10 +327,12 @@ impl EncryptionParams {
 			Self::Plaintext => stream,
 			Self::SseC {
 				client_key,
+				object_key,
 				compression_level,
 				..
 			} => {
-				let plaintext = DecryptStream::new(stream, *client_key);
+				let key = object_key.as_ref().unwrap_or(client_key);
+				let plaintext = DecryptStream::new(stream, *key);
 				if compression_level.is_some() {
 					let reader = stream_asyncread(Box::pin(plaintext));
 					let reader = BufReader::new(reader);
@@ -307,9 +352,12 @@ impl EncryptionParams {
 			Self::Plaintext => Ok(block),
 			Self::SseC {
 				client_key,
+				object_key,
 				compression_level,
 				..
 			} => {
+				let key = object_key.as_ref().unwrap_or(client_key);
+
 				let block = if let Some(level) = compression_level {
 					Cow::Owned(
 						garage_block::zstd_encode(block.as_ref(), *level)
@@ -325,7 +373,7 @@ impl EncryptionParams {
 				OsRng.fill_bytes(&mut nonce);
 				ret.extend_from_slice(nonce.as_slice());
 
-				let mut cipher = EncryptorLE31::<Aes256Gcm>::new(&client_key, &nonce);
+				let mut cipher = EncryptorLE31::<Aes256Gcm>::new(key, &nonce);
 				let mut iter = block.chunks(STREAM_ENC_PLAIN_CHUNK_SIZE).peekable();
 
 				if iter.peek().is_none() {
@@ -361,6 +409,13 @@ impl EncryptionParams {
 	}
 }
 
+pub fn has_encryption_header(headers: &HeaderMap) -> bool {
+	match headers.get(X_AMZ_SERVER_SIDE_ENCRYPTION_CUSTOMER_ALGORITHM) {
+		Some(h) => h.as_bytes() == CUSTOMER_ALGORITHM_AES256,
+		None => false,
+	}
+}
+
 fn parse_request_headers(
 	headers: &HeaderMap,
 	alg_header: &HeaderName,
@@ -420,6 +475,30 @@ fn parse_request_headers(
 	}
 }
 
+impl<'a> OekDerivationInfo<'a> {
+	pub fn for_object<'b>(object: &'a Object, version: &'b ObjectVersion) -> Self {
+		Self {
+			bucket_id: object.bucket_id,
+			version_id: version.uuid,
+			object_key: &object.key,
+		}
+	}
+
+	fn derive_oek(&self, client_key: &Key<Aes256Gcm>) -> Key<Aes256Gcm> {
+		use hmac::{Hmac, Mac};
+
+		// info = bucket_id + object_name + version_uuid + "garage-object-encryption-key"
+		// oek = hmac_sha256(ssec_key, info)
+		let mut hmac = <Hmac<Sha256> as Mac>::new_from_slice(client_key.as_slice())
+			.expect("create hmac-sha256");
+		hmac.update(b"garage-object-encryption-key");
+		hmac.update(self.bucket_id.as_slice());
+		hmac.update(self.version_id.as_slice());
+		hmac.update(self.object_key.as_bytes());
+		hmac.finalize().into_bytes()
+	}
+}
+
 // ---- encrypt & decrypt streams ----
 
 #[pin_project::pin_project]
@@ -569,6 +648,7 @@ mod tests {
 		let enc = EncryptionParams::SseC {
 			client_key: Aes256Gcm::generate_key(&mut OsRng),
 			client_key_md5: Default::default(), // not needed
+			object_key: Some(Aes256Gcm::generate_key(&mut OsRng)),
 			compression_level,
 		};
 

src/api/s3/error.rs

@@ -1,8 +1,8 @@
 use std::convert::TryInto;
 
+use err_derive::Error;
 use hyper::header::HeaderValue;
 use hyper::{HeaderMap, StatusCode};
-use thiserror::Error;
 
 use garage_model::helper::error::Error as HelperError;
 
@@ -25,67 +25,67 @@ use crate::xml as s3_xml;
 /// Errors of this crate
 #[derive(Debug, Error)]
 pub enum Error {
-	#[error("{0}")]
+	#[error(display = "{}", _0)]
 	/// Error from common error
-	Common(#[from] CommonError),
+	Common(#[error(source)] CommonError),
 
 	// Category: cannot process
 	/// Authorization Header Malformed
-	#[error("Authorization header malformed, unexpected scope: {0}")]
+	#[error(display = "Authorization header malformed, unexpected scope: {}", _0)]
 	AuthorizationHeaderMalformed(String),
 
 	/// The object requested don't exists
-	#[error("Key not found")]
+	#[error(display = "Key not found")]
 	NoSuchKey,
 
 	/// The multipart upload requested don't exists
-	#[error("Upload not found")]
+	#[error(display = "Upload not found")]
 	NoSuchUpload,
 
 	/// Precondition failed (e.g. x-amz-copy-source-if-match)
-	#[error("At least one of the preconditions you specified did not hold")]
+	#[error(display = "At least one of the preconditions you specified did not hold")]
 	PreconditionFailed,
 
 	/// Parts specified in CMU request do not match parts actually uploaded
-	#[error("Parts given to CompleteMultipartUpload do not match uploaded parts")]
+	#[error(display = "Parts given to CompleteMultipartUpload do not match uploaded parts")]
 	InvalidPart,
 
 	/// Parts given to CompleteMultipartUpload were not in ascending order
-	#[error("Parts given to CompleteMultipartUpload were not in ascending order")]
+	#[error(display = "Parts given to CompleteMultipartUpload were not in ascending order")]
 	InvalidPartOrder,
 
 	/// In CompleteMultipartUpload: not enough data
 	/// (here we are more lenient than AWS S3)
-	#[error("Proposed upload is smaller than the minimum allowed object size")]
+	#[error(display = "Proposed upload is smaller than the minimum allowed object size")]
 	EntityTooSmall,
 
 	// Category: bad request
 	/// The request contained an invalid UTF-8 sequence in its path or in other parameters
-	#[error("Invalid UTF-8: {0}")]
-	InvalidUtf8Str(#[from] std::str::Utf8Error),
+	#[error(display = "Invalid UTF-8: {}", _0)]
+	InvalidUtf8Str(#[error(source)] std::str::Utf8Error),
 
 	/// The request used an invalid path
-	#[error("Invalid UTF-8: {0}")]
-	InvalidUtf8String(#[from] std::string::FromUtf8Error),
+	#[error(display = "Invalid UTF-8: {}", _0)]
+	InvalidUtf8String(#[error(source)] std::string::FromUtf8Error),
 
 	/// The client sent invalid XML data
-	#[error("Invalid XML: {0}")]
+	#[error(display = "Invalid XML: {}", _0)]
 	InvalidXml(String),
 
 	/// The client sent a range header with invalid value
-	#[error("Invalid HTTP range: {0:?}")]
-	InvalidRange((http_range::HttpRangeParseError, u64)),
+	#[error(display = "Invalid HTTP range: {:?}", _0)]
+	InvalidRange(#[error(from)] (http_range::HttpRangeParseError, u64)),
 
 	/// The client sent a range header with invalid value
-	#[error("Invalid encryption algorithm: {0:?}, should be AES256")]
+	#[error(display = "Invalid encryption algorithm: {:?}, should be AES256", _0)]
 	InvalidEncryptionAlgorithm(String),
 
 	/// The provided digest (checksum) value was invalid
-	#[error("Invalid digest: {0}")]
+	#[error(display = "Invalid digest: {}", _0)]
 	InvalidDigest(String),
 
 	/// The client sent a request for an action not supported by garage
-	#[error("Unimplemented action: {0}")]
+	#[error(display = "Unimplemented action: {}", _0)]
 	NotImplemented(String),
 }
 
@@ -99,12 +99,6 @@ impl From<HelperError> for Error {
 	}
 }
 
-impl From<(http_range::HttpRangeParseError, u64)> for Error {
-	fn from(err: (http_range::HttpRangeParseError, u64)) -> Error {
-		Error::InvalidRange(err)
-	}
-}
-
 impl From<roxmltree::Error> for Error {
 	fn from(err: roxmltree::Error) -> Self {
 		Self::InvalidXml(format!("{}", err))

src/api/s3/get.rs

@@ -19,19 +19,18 @@ use garage_net::stream::ByteStream;
 use garage_rpc::rpc_helper::OrderTag;
 use garage_table::EmptyKey;
 use garage_util::data::*;
-use garage_util::error::{Error as UtilError, OkOrMessage};
+use garage_util::error::OkOrMessage;
 
 use garage_model::garage::Garage;
 use garage_model::s3::object_table::*;
 use garage_model::s3::version_table::*;
 
-use garage_api_common::common_error::CommonError;
 use garage_api_common::helpers::*;
 use garage_api_common::signature::checksum::{add_checksum_response_headers, X_AMZ_CHECKSUM_MODE};
 
 use crate::api_server::ResBody;
 use crate::copy::*;
-use crate::encryption::EncryptionParams;
+use crate::encryption::{EncryptionParams, OekDerivationInfo};
 use crate::error::*;
 
 const X_AMZ_MP_PARTS_COUNT: HeaderName = HeaderName::from_static("x-amz-mp-parts-count");
@@ -182,8 +181,12 @@ pub async fn handle_head_without_ctx(
 		return Ok(res);
 	}
 
-	let (encryption, headers) =
-		EncryptionParams::check_decrypt(&garage, req.headers(), &version_meta.encryption)?;
+	let (encryption, headers) = EncryptionParams::check_decrypt(
+		&garage,
+		req.headers(),
+		&version_meta.encryption,
+		OekDerivationInfo::for_object(&object, object_version),
+	)?;
 
 	let checksum_mode = checksum_mode(&req);
 
@@ -216,7 +219,6 @@ pub async fn handle_head_without_ctx(
 					.get(&object_version.uuid, &EmptyKey)
 					.await?
 					.ok_or(Error::NoSuchKey)?;
-				check_version_not_deleted(&version)?;
 
 				let (part_offset, part_end) =
 					calculate_part_bounds(&version, pn).ok_or(Error::InvalidPart)?;
@@ -305,8 +307,12 @@ pub async fn handle_get_without_ctx(
 		return Ok(res);
 	}
 
-	let (enc, headers) =
-		EncryptionParams::check_decrypt(&garage, req.headers(), &last_v_meta.encryption)?;
+	let (enc, headers) = EncryptionParams::check_decrypt(
+		&garage,
+		req.headers(),
+		&last_v_meta.encryption,
+		OekDerivationInfo::for_object(&object, last_v),
+	)?;
 
 	let checksum_mode = checksum_mode(&req);
 
@@ -367,21 +373,6 @@ pub async fn handle_get_without_ctx(
 	}
 }
 
-pub(crate) fn check_version_not_deleted(version: &Version) -> Result<(), Error> {
-	if version.deleted.get() {
-		// the version was deleted between when the object_table was consulted
-		// and now, this could mean the object was deleted, or overriden.
-		// Rather than say the key doesn't exist, return a transient error
-		// to signal the client to try again.
-		return Err(CommonError::InternalError(UtilError::Message(
-			"conflict/inconsistency between object and version state, version is deleted"
-				.to_string(),
-		))
-		.into());
-	}
-	Ok(())
-}
-
 async fn handle_get_full(
 	garage: Arc<Garage>,
 	version: &ObjectVersion,
@@ -448,7 +439,6 @@ pub fn full_object_byte_stream(
 						.ok_or_message("channel closed")?;
 
 					let version = version_fut.await.unwrap()?.ok_or(Error::NoSuchKey)?;
-					check_version_not_deleted(&version)?;
 					for (i, (_, vb)) in version.blocks.items().iter().enumerate().skip(1) {
 						let stream_block_i = encryption
 							.get_block(&garage, &vb.hash, Some(order_stream.order(i as u64)))
@@ -464,14 +454,6 @@ pub fn full_object_byte_stream(
 				{
 					Ok(()) => (),
 					Err(e) => {
-						// TODO i think this is a bad idea, we should log
-						// an error and stop there. If the error happens to
-						// be exactly the size of what hasn't been streamed
-						// yet, the client will see the request as a
-						// success
-						// instead truncating the output notify the client
-						// something happened with their download, so that
-						// they can retry it
 						let _ = tx.send(error_stream_item(e)).await;
 					}
 				}
@@ -523,7 +505,7 @@ async fn handle_get_range(
 				.get(&version.uuid, &EmptyKey)
 				.await?
 				.ok_or(Error::NoSuchKey)?;
-			check_version_not_deleted(&version)?;
+
 			let body =
 				body_from_blocks_range(garage, encryption, version.blocks.items(), begin, end);
 			Ok(resp_builder.body(body)?)
@@ -574,8 +556,6 @@ async fn handle_get_part(
 				.await?
 				.ok_or(Error::NoSuchKey)?;
 
-			check_version_not_deleted(&version)?;
-
 			let (begin, end) =
 				calculate_part_bounds(&version, part_number).ok_or(Error::InvalidPart)?;
 
@@ -845,9 +825,7 @@ impl PreconditionHeaders {
 	}
 
 	fn check(&self, v: &ObjectVersion, etag: &str) -> Result<Option<StatusCode>, Error> {
-		// we store date with ms precision, but headers are precise to the second: truncate
-		// the timestamp to handle the same-second edge case
-		let v_date = UNIX_EPOCH + Duration::from_secs(v.timestamp / 1000);
+		let v_date = UNIX_EPOCH + Duration::from_millis(v.timestamp);
 
 		// Implemented from https://datatracker.ietf.org/doc/html/rfc7232#section-6
 

src/api/s3/lifecycle.rs

@@ -27,7 +27,7 @@ pub async fn handle_get_lifecycle(ctx: ReqCtx) -> Result<Response<ResBody>, Erro
 			.body(string_body(xml))?)
 	} else {
 		Ok(Response::builder()
-			.status(StatusCode::NOT_FOUND)
+			.status(StatusCode::NO_CONTENT)
 			.body(empty_body())?)
 	}
 }

src/api/s3/list.rs

@@ -17,7 +17,7 @@ use garage_api_common::encoding::*;
 use garage_api_common::helpers::*;
 
 use crate::api_server::{ReqBody, ResBody};
-use crate::encryption::EncryptionParams;
+use crate::encryption::{EncryptionParams, OekDerivationInfo};
 use crate::error::*;
 use crate::multipart as s3_multipart;
 use crate::xml as s3_xml;
@@ -285,8 +285,16 @@ pub async fn handle_list_parts(
 		ObjectVersionState::Uploading { encryption, .. } => encryption,
 		_ => unreachable!(),
 	};
-	let encryption_res =
-		EncryptionParams::check_decrypt(&ctx.garage, req.headers(), &object_encryption);
+	let encryption_res = EncryptionParams::check_decrypt(
+		&ctx.garage,
+		req.headers(),
+		&object_encryption,
+		OekDerivationInfo {
+			bucket_id: ctx.bucket_id,
+			version_id: upload_id,
+			object_key: &query.key,
+		},
+	);
 
 	let (info, next) = fetch_part_info(query, &mpu)?;
 
@@ -326,6 +334,12 @@ pub async fn handle_list_parts(
 						}
 						_ => None,
 					},
+					checksum_crc64nvme: match &checksum {
+						Some(ChecksumValue::Crc64Nvme(x)) => {
+							Some(s3_xml::Value(BASE64_STANDARD.encode(&x)))
+						}
+						_ => None,
+					},
 					checksum_sha1: match &checksum {
 						Some(ChecksumValue::Sha1(x)) => {
 							Some(s3_xml::Value(BASE64_STANDARD.encode(&x)))

src/api/s3/multipart.rs

@@ -6,6 +6,7 @@ use std::sync::Arc;
 use base64::prelude::*;
 use crc32c::Crc32cHasher as Crc32c;
 use crc32fast::Hasher as Crc32;
+use crc64fast_nvme::Digest as Crc64Nvme;
 use futures::prelude::*;
 use hyper::{Request, Response};
 use md5::{Digest, Md5};
@@ -26,7 +27,7 @@ use garage_api_common::helpers::*;
 use garage_api_common::signature::checksum::*;
 
 use crate::api_server::{ReqBody, ResBody};
-use crate::encryption::EncryptionParams;
+use crate::encryption::{has_encryption_header, EncryptionParams, OekDerivationInfo};
 use crate::error::*;
 use crate::put::*;
 use crate::xml as s3_xml;
@@ -56,7 +57,15 @@ pub async fn handle_create_multipart_upload(
 	};
 
 	// Determine whether object should be encrypted, and if so the key
-	let encryption = EncryptionParams::new_from_headers(&garage, req.headers())?;
+	let encryption = EncryptionParams::new_from_headers(
+		&garage,
+		req.headers(),
+		OekDerivationInfo {
+			bucket_id: *bucket_id,
+			version_id: upload_id,
+			object_key: &key,
+		},
+	)?;
 	let object_encryption = encryption.encrypt_meta(meta)?;
 
 	let checksum_algorithm = request_checksum_algorithm(req.headers())?;
@@ -120,8 +129,7 @@ pub async fn handle_put_part(
 
 	// Before we stream the body, configure the needed checksums.
 	req_body.add_expected_checksums(expected_checksums.clone());
-	// TODO: avoid parsing encryption headers twice...
-	if !EncryptionParams::new_from_headers(&garage, &req_head.headers)?.is_encrypted() {
+	if !has_encryption_header(&req_head.headers) {
 		// For non-encrypted objects, we need to compute the md5sum in all cases
 		// (even if content-md5 is not set), because it is used as an etag of the
 		// part, which is in turn used in the etag computation of the whole object
@@ -134,10 +142,11 @@ pub async fn handle_put_part(
 	let mut chunker = StreamChunker::new(stream, garage.config.block_size);
 
 	// Read first chuck, and at the same time try to get object to see if it exists
-	let ((_, object_version, mut mpu), first_block) =
+	let ((object, object_version, mut mpu), first_block) =
 		futures::try_join!(get_upload(&ctx, &key, &upload_id), chunker.next(),)?;
 
 	// Check encryption params
+	let oek_params = OekDerivationInfo::for_object(&object, &object_version);
 	let (object_encryption, checksum_algorithm) = match object_version.state {
 		ObjectVersionState::Uploading {
 			encryption,
@@ -146,8 +155,12 @@ pub async fn handle_put_part(
 		} => (encryption, checksum_algorithm),
 		_ => unreachable!(),
 	};
-	let (encryption, _) =
-		EncryptionParams::check_decrypt(&garage, &req_head.headers, &object_encryption)?;
+	let (encryption, _) = EncryptionParams::check_decrypt(
+		&garage,
+		&req_head.headers,
+		&object_encryption,
+		oek_params,
+	)?;
 
 	// Check object is valid and part can be accepted
 	let first_block = first_block.ok_or_bad_request("Empty body")?;
@@ -297,6 +310,7 @@ pub async fn handle_complete_multipart_upload(
 		return Err(Error::bad_request("No data was uploaded"));
 	}
 
+	let oek_params = OekDerivationInfo::for_object(&object, &object_version);
 	let (object_encryption, checksum_algorithm) = match object_version.state {
 		ObjectVersionState::Uploading {
 			encryption,
@@ -417,8 +431,12 @@ pub async fn handle_complete_multipart_upload(
 	let object_encryption = match checksum_algorithm {
 		None => object_encryption,
 		Some(_) => {
-			let (encryption, meta) =
-				EncryptionParams::check_decrypt(&garage, &req_head.headers, &object_encryption)?;
+			let (encryption, meta) = EncryptionParams::check_decrypt(
+				&garage,
+				&req_head.headers,
+				&object_encryption,
+				oek_params,
+			)?;
 			let new_meta = ObjectVersionMetaInner {
 				headers: meta.into_owned().headers,
 				checksum: checksum_extra,
@@ -464,6 +482,10 @@ pub async fn handle_complete_multipart_upload(
 			Some(ChecksumValue::Crc32c(x)) => Some(s3_xml::Value(BASE64_STANDARD.encode(&x))),
 			_ => None,
 		},
+		checksum_crc64nvme: match &checksum_extra {
+			Some(ChecksumValue::Crc64Nvme(x)) => Some(s3_xml::Value(BASE64_STANDARD.encode(&x))),
+			_ => None,
+		},
 		checksum_sha1: match &checksum_extra {
 			Some(ChecksumValue::Sha1(x)) => Some(s3_xml::Value(BASE64_STANDARD.encode(&x))),
 			_ => None,
@@ -587,6 +609,15 @@ fn parse_complete_multipart_upload_body(
 						.try_into()
 						.ok()?,
 				))
+			} else if let Some(crc64nvme) = item
+				.children()
+				.find(|e| e.has_tag_name("ChecksumCRC64NVME"))
+			{
+				Some(ChecksumValue::Crc64Nvme(
+					BASE64_STANDARD.decode(crc64nvme.text()?).ok()?[..]
+						.try_into()
+						.ok()?,
+				))
 			} else if let Some(sha1) = item.children().find(|e| e.has_tag_name("ChecksumSHA1")) {
 				Some(ChecksumValue::Sha1(
 					BASE64_STANDARD.decode(sha1.text()?).ok()?[..]
@@ -627,6 +658,7 @@ pub(crate) struct MultipartChecksummer {
 pub(crate) enum MultipartExtraChecksummer {
 	Crc32(Crc32),
 	Crc32c(Crc32c),
+	Crc64Nvme(Crc64Nvme),
 	Sha1(Sha1),
 	Sha256(Sha256),
 }
@@ -643,6 +675,9 @@ impl MultipartChecksummer {
 				Some(ChecksumAlgorithm::Crc32c) => {
 					Some(MultipartExtraChecksummer::Crc32c(Crc32c::default()))
 				}
+				Some(ChecksumAlgorithm::Crc64Nvme) => {
+					Some(MultipartExtraChecksummer::Crc64Nvme(Crc64Nvme::default()))
+				}
 				Some(ChecksumAlgorithm::Sha1) => Some(MultipartExtraChecksummer::Sha1(Sha1::new())),
 				Some(ChecksumAlgorithm::Sha256) => {
 					Some(MultipartExtraChecksummer::Sha256(Sha256::new()))
@@ -672,6 +707,12 @@ impl MultipartChecksummer {
 			) => {
 				crc32c.write(&x);
 			}
+			(
+				Some(MultipartExtraChecksummer::Crc64Nvme(ref mut crc64nvme)),
+				Some(ChecksumValue::Crc64Nvme(x)),
+			) => {
+				crc64nvme.write(&x);
+			}
 			(Some(MultipartExtraChecksummer::Sha1(ref mut sha1)), Some(ChecksumValue::Sha1(x))) => {
 				sha1.update(&x);
 			}
@@ -701,6 +742,9 @@ impl MultipartChecksummer {
 			Some(MultipartExtraChecksummer::Crc32c(crc32c)) => Some(ChecksumValue::Crc32c(
 				u32::to_be_bytes(u32::try_from(crc32c.finish()).unwrap()),
 			)),
+			Some(MultipartExtraChecksummer::Crc64Nvme(crc64nvme)) => Some(
+				ChecksumValue::Crc64Nvme(u64::to_be_bytes(crc64nvme.sum64())),
+			),
 			Some(MultipartExtraChecksummer::Sha1(sha1)) => {
 				Some(ChecksumValue::Sha1(sha1.finalize()[..].try_into().unwrap()))
 			}

src/api/s3/post_object.rs

@@ -15,6 +15,7 @@ use serde::Deserialize;
 
 use garage_model::garage::Garage;
 use garage_model::s3::object_table::*;
+use garage_util::data::gen_uuid;
 
 use garage_api_common::cors::*;
 use garage_api_common::helpers::*;
@@ -22,7 +23,7 @@ use garage_api_common::signature::checksum::*;
 use garage_api_common::signature::payload::{verify_v4, Authorization};
 
 use crate::api_server::ResBody;
-use crate::encryption::EncryptionParams;
+use crate::encryption::{EncryptionParams, OekDerivationInfo};
 use crate::error::*;
 use crate::put::{extract_metadata_headers, save_stream, ChecksumMode};
 use crate::xml as s3_xml;
@@ -103,22 +104,18 @@ pub async fn handle_post_object(
 		key.to_owned()
 	};
 
-	let api_key = verify_v4(&garage, "s3", &authorization, policy.as_bytes()).await?;
+	let api_key = verify_v4(&garage, "s3", &authorization, policy.as_bytes())?;
 
-	let bucket_id = garage
+	let bucket = garage
 		.bucket_helper()
-		.resolve_bucket(&bucket_name, &api_key)
-		.await
+		.resolve_bucket_fast(&bucket_name, &api_key)
 		.map_err(pass_helper_error)?;
+	let bucket_id = bucket.id;
 
 	if !api_key.allow_write(&bucket_id) {
 		return Err(Error::forbidden("Operation is not allowed for this key."));
 	}
 
-	let bucket = garage
-		.bucket_helper()
-		.get_existing_bucket(bucket_id)
-		.await?;
 	let bucket_params = bucket.state.into_option().unwrap();
 	let matching_cors_rule = find_matching_cors_rule(
 		&bucket_params,
@@ -141,26 +138,10 @@ pub async fn handle_post_object(
 
 	let mut conditions = decoded_policy.into_conditions()?;
 
-	// If there are conditions on the bucket name, check these against the actual bucket_name rather
-	// than the one in params, which is allowed to be absent.
-	if let Some(conds) = conditions.params.remove("bucket") {
-		for cond in conds {
-			let ok = match cond {
-				Operation::Equal(s) => s.as_str() == bucket_name,
-				Operation::StartsWith(s) => bucket_name.starts_with(&s),
-			};
-			if !ok {
-				return Err(Error::bad_request(
-					"Key 'bucket' has value not allowed in policy",
-				));
-			}
-		}
-	}
-
 	for (param_key, value) in params.iter() {
 		let param_key = param_key.as_str();
 		match param_key {
-			"policy" | "x-amz-signature" | "bucket" => (), // this is always accepted, as it's required to validate other fields
+			"policy" | "x-amz-signature" => (), // this is always accepted, as it's required to validate other fields
 			"content-type" => {
 				let conds = conditions.params.remove("content-type").ok_or_else(|| {
 					Error::bad_request(format!("Key '{}' is not allowed in policy", param_key))
@@ -247,12 +228,22 @@ pub async fn handle_post_object(
 			.transpose()?,
 	};
 
+	let version_uuid = gen_uuid();
+
 	let meta = ObjectVersionMetaInner {
 		headers,
 		checksum: expected_checksums.extra,
 	};
 
-	let encryption = EncryptionParams::new_from_headers(&garage, &params)?;
+	let encryption = EncryptionParams::new_from_headers(
+		&garage,
+		&params,
+		OekDerivationInfo {
+			bucket_id,
+			version_id: version_uuid,
+			object_key: &key,
+		},
+	)?;
 
 	let stream = file_field.map(|r| r.map_err(Into::into));
 	let ctx = ReqCtx {
@@ -265,6 +256,7 @@ pub async fn handle_post_object(
 
 	let res = save_stream(
 		&ctx,
+		version_uuid,
 		meta,
 		encryption,
 		StreamLimiter::new(stream, conditions.content_length),

src/api/s3/put.rs

@@ -35,10 +35,12 @@ use garage_api_common::signature::body::StreamingChecksumReceiver;
 use garage_api_common::signature::checksum::*;
 
 use crate::api_server::{ReqBody, ResBody};
-use crate::encryption::EncryptionParams;
+use crate::encryption::{EncryptionParams, OekDerivationInfo};
 use crate::error::*;
 use crate::website::X_AMZ_WEBSITE_REDIRECT_LOCATION;
 
+const PUT_BLOCKS_MAX_PARALLEL: usize = 3;
+
 pub(crate) struct SaveStreamResult {
 	pub(crate) version_uuid: Uuid,
 	pub(crate) version_timestamp: u64,
@@ -60,6 +62,10 @@ pub async fn handle_put(
 	req: Request<ReqBody>,
 	key: &String,
 ) -> Result<Response<ResBody>, Error> {
+	// Generate version uuid now, because it is necessary to compute SSE-C
+	// encryption parameters
+	let version_uuid = gen_uuid();
+
 	// Retrieve interesting headers from request
 	let headers = extract_metadata_headers(req.headers())?;
 	debug!("Object headers: {:?}", headers);
@@ -80,7 +86,15 @@ pub async fn handle_put(
 	};
 
 	// Determine whether object should be encrypted, and if so the key
-	let encryption = EncryptionParams::new_from_headers(&ctx.garage, req.headers())?;
+	let encryption = EncryptionParams::new_from_headers(
+		&ctx.garage,
+		req.headers(),
+		OekDerivationInfo {
+			bucket_id: ctx.bucket_id,
+			version_id: version_uuid,
+			object_key: &key,
+		},
+	)?;
 
 	// The request body is a special ReqBody object (see garage_api_common::signature::body)
 	// which supports calculating checksums while streaming the data.
@@ -98,6 +112,7 @@ pub async fn handle_put(
 
 	let res = save_stream(
 		&ctx,
+		version_uuid,
 		meta,
 		encryption,
 		stream,
@@ -119,6 +134,7 @@ pub async fn handle_put(
 
 pub(crate) async fn save_stream<S: Stream<Item = Result<Bytes, Error>> + Unpin>(
 	ctx: &ReqCtx,
+	version_uuid: Uuid,
 	mut meta: ObjectVersionMetaInner,
 	encryption: EncryptionParams,
 	body: S,
@@ -138,7 +154,6 @@ pub(crate) async fn save_stream<S: Stream<Item = Result<Bytes, Error>> + Unpin>(
 	let first_block = first_block_opt.unwrap_or_default();
 
 	// Generate identity of new version
-	let version_uuid = gen_uuid();
 	let version_timestamp = next_timestamp(existing_object.as_ref());
 
 	let mut checksummer = match &checksum_mode {
@@ -491,7 +506,7 @@ pub(crate) async fn read_and_put_blocks<S: Stream<Item = Result<Bytes, Error>> +
 			};
 			let recv_next = async {
 				// If more than a maximum number of writes are in progress, don't add more for now
-				if currently_running >= ctx.garage.config.block_max_concurrent_writes_per_request {
+				if currently_running >= PUT_BLOCKS_MAX_PARALLEL {
 					futures::future::pending().await
 				} else {
 					block_rx3.recv().await

src/api/s3/website.rs

@@ -3,7 +3,7 @@ use quick_xml::de::from_reader;
 use hyper::{header::HeaderName, Request, Response, StatusCode};
 use serde::{Deserialize, Serialize};
 
-use garage_model::bucket_table::*;
+use garage_model::bucket_table::{self, *};
 
 use garage_api_common::helpers::*;
 
@@ -26,7 +26,28 @@ pub async fn handle_get_website(ctx: ReqCtx) -> Result<Response<ResBody>, Error>
 				suffix: Value(website.index_document.to_string()),
 			}),
 			redirect_all_requests_to: None,
-			routing_rules: None,
+			routing_rules: RoutingRules {
+				rules: website
+					.routing_rules
+					.clone()
+					.into_iter()
+					.map(|rule| RoutingRule {
+						condition: rule.condition.map(|cond| Condition {
+							http_error_code: cond.http_error_code.map(|c| IntValue(c as i64)),
+							prefix: cond.prefix.map(Value),
+						}),
+						redirect: Redirect {
+							hostname: rule.redirect.hostname.map(Value),
+							http_redirect_code: Some(IntValue(
+								rule.redirect.http_redirect_code as i64,
+							)),
+							protocol: rule.redirect.protocol.map(Value),
+							replace_full: rule.redirect.replace_key.map(Value),
+							replace_prefix: rule.redirect.replace_key_prefix.map(Value),
+						},
+					})
+					.collect(),
+			},
 		};
 		let xml = to_xml_with_header(&wc)?;
 		Ok(Response::builder()
@@ -97,18 +118,28 @@ pub struct WebsiteConfiguration {
 	pub index_document: Option<Suffix>,
 	#[serde(rename = "RedirectAllRequestsTo")]
 	pub redirect_all_requests_to: Option<Target>,
-	#[serde(rename = "RoutingRules")]
-	pub routing_rules: Option<Vec<RoutingRule>>,
+	#[serde(
+		rename = "RoutingRules",
+		default,
+		skip_serializing_if = "RoutingRules::is_empty"
+	)]
+	pub routing_rules: RoutingRules,
+}
+
+#[derive(Debug, Serialize, Deserialize, PartialEq, Eq, PartialOrd, Ord, Default)]
+pub struct RoutingRules {
+	#[serde(rename = "RoutingRule")]
+	pub rules: Vec<RoutingRule>,
+}
+
+impl RoutingRules {
+	fn is_empty(&self) -> bool {
+		self.rules.is_empty()
+	}
 }
 
 #[derive(Debug, Serialize, Deserialize, PartialEq, Eq, PartialOrd, Ord)]
 pub struct RoutingRule {
-	#[serde(rename = "RoutingRule")]
-	pub inner: RoutingRuleInner,
-}
-
-#[derive(Debug, Serialize, Deserialize, PartialEq, Eq, PartialOrd, Ord)]
-pub struct RoutingRuleInner {
 	#[serde(rename = "Condition")]
 	pub condition: Option<Condition>,
 	#[serde(rename = "Redirect")]
@@ -162,7 +193,7 @@ impl WebsiteConfiguration {
 		if self.redirect_all_requests_to.is_some()
 			&& (self.error_document.is_some()
 				|| self.index_document.is_some()
-				|| self.routing_rules.is_some())
+				|| !self.routing_rules.is_empty())
 		{
 			return Err(Error::bad_request(
 				"Bad XML: can't have RedirectAllRequestsTo and other fields",
@@ -177,10 +208,15 @@ impl WebsiteConfiguration {
 		if let Some(ref rart) = self.redirect_all_requests_to {
 			rart.validate()?;
 		}
-		if let Some(ref rrs) = self.routing_rules {
-			for rr in rrs {
-				rr.inner.validate()?;
+		for rr in &self.routing_rules.rules {
+			rr.validate()?;
 		}
+		if self.routing_rules.rules.len() > 1000 {
+			// we will do linear scans, best to avoid overly long configuration. The
+			// limit was choosen arbitrarily
+			return Err(Error::bad_request(
+				"Bad XML: RoutingRules can't have more than 1000 child elements",
+			));
 		}
 
 		Ok(())
@@ -189,11 +225,7 @@ impl WebsiteConfiguration {
 	pub fn into_garage_website_config(self) -> Result<WebsiteConfig, Error> {
 		if self.redirect_all_requests_to.is_some() {
 			Err(Error::NotImplemented(
-				"S3 website redirects are not currently implemented in Garage.".into(),
-			))
-		} else if self.routing_rules.map(|x| !x.is_empty()).unwrap_or(false) {
-			Err(Error::NotImplemented(
-				"S3 routing rules are not currently implemented in Garage.".into(),
+				"RedirectAllRequestsTo is not currently implemented in Garage, however its effect can be emulated using a single inconditional RoutingRule.".into(),
 			))
 		} else {
 			Ok(WebsiteConfig {
@@ -202,6 +234,36 @@ impl WebsiteConfiguration {
 					.map(|x| x.suffix.0)
 					.unwrap_or_else(|| "index.html".to_string()),
 				error_document: self.error_document.map(|x| x.key.0),
+				redirect_all: None,
+				routing_rules: self
+					.routing_rules
+					.rules
+					.into_iter()
+					.map(|rule| {
+						bucket_table::RoutingRule {
+							condition: rule.condition.map(|condition| {
+								bucket_table::RedirectCondition {
+									http_error_code: condition.http_error_code.map(|c| c.0 as u16),
+									prefix: condition.prefix.map(|p| p.0),
+								}
+							}),
+							redirect: bucket_table::Redirect {
+								hostname: rule.redirect.hostname.map(|h| h.0),
+								protocol: rule.redirect.protocol.map(|p| p.0),
+								// aws default to 301, which i find punitive in case of
+								// missconfiguration (can be permanently cached on the
+								// user agent)
+								http_redirect_code: rule
+									.redirect
+									.http_redirect_code
+									.map(|c| c.0 as u16)
+									.unwrap_or(302),
+								replace_key_prefix: rule.redirect.replace_prefix.map(|k| k.0),
+								replace_key: rule.redirect.replace_full.map(|k| k.0),
+							},
+						}
+					})
+					.collect(),
 			})
 		}
 	}
@@ -242,37 +304,69 @@ impl Target {
 	}
 }
 
-impl RoutingRuleInner {
+impl RoutingRule {
 	pub fn validate(&self) -> Result<(), Error> {
-		let has_prefix = self
-			.condition
-			.as_ref()
-			.and_then(|c| c.prefix.as_ref())
-			.is_some();
-		self.redirect.validate(has_prefix)
+		if let Some(condition) = &self.condition {
+			condition.validate()?;
+		}
+		self.redirect.validate()
+	}
+}
+
+impl Condition {
+	pub fn validate(&self) -> Result<bool, Error> {
+		if let Some(ref error_code) = self.http_error_code {
+			// TODO do other error codes make sense? Aws only allows 4xx and 5xx
+			if error_code.0 != 404 {
+				return Err(Error::bad_request(
+					"Bad XML: HttpErrorCodeReturnedEquals must be 404 or absent",
+				));
+			}
+		}
+		Ok(self.prefix.is_some())
 	}
 }
 
 impl Redirect {
-	pub fn validate(&self, has_prefix: bool) -> Result<(), Error> {
-		if self.replace_prefix.is_some() {
-			if self.replace_full.is_some() {
+	pub fn validate(&self) -> Result<(), Error> {
+		if self.replace_prefix.is_some() && self.replace_full.is_some() {
 			return Err(Error::bad_request(
 				"Bad XML: both ReplaceKeyPrefixWith and ReplaceKeyWith are set",
 			));
 		}
-			if !has_prefix {
-				return Err(Error::bad_request(
-					"Bad XML: ReplaceKeyPrefixWith is set, but  KeyPrefixEquals isn't",
-				));
-			}
-		}
 		if let Some(ref protocol) = self.protocol {
 			if protocol.0 != "http" && protocol.0 != "https" {
 				return Err(Error::bad_request("Bad XML: invalid protocol"));
 			}
 		}
-		// TODO there are probably more invalid cases, but which ones?
+		if let Some(ref http_redirect_code) = self.http_redirect_code {
+			match http_redirect_code.0 {
+				// aws allows all 3xx except 300, but some are non-sensical (not modified,
+				// use proxy...)
+				301 | 302 | 303 | 307 | 308 => {
+					if self.hostname.is_none() && self.protocol.is_some() {
+						return Err(Error::bad_request(
+							"Bad XML: HostName must be set if Protocol is set",
+						));
+					}
+				}
+				// aws doesn't allow these codes, but netlify does, and it seems like a
+				// cool feature (change the page seen without changing the url shown by the
+				// user agent)
+				200 | 404 => {
+					if self.hostname.is_some() || self.protocol.is_some() {
+						// hostname would mean different bucket, protocol doesn't make
+						// sense
+						return Err(Error::bad_request(
+                                        "Bad XML: an HttpRedirectCode of 200 is not acceptable alongside HostName or Protocol",
+                                ));
+					}
+				}
+				_ => {
+					return Err(Error::bad_request("Bad XML: invalid HttpRedirectCode"));
+				}
+			}
+		}
 		Ok(())
 	}
 }
@@ -311,6 +405,15 @@ mod tests {
             <ReplaceKeyWith>fullkey</ReplaceKeyWith>
          </Redirect>
       </RoutingRule>
+      <RoutingRule>
+         <Condition>
+            <KeyPrefixEquals></KeyPrefixEquals>
+         </Condition>
+         <Redirect>
+            <HttpRedirectCode>404</HttpRedirectCode>
+            <ReplaceKeyWith>missing</ReplaceKeyWith>
+         </Redirect>
+      </RoutingRule>
    </RoutingRules>
 </WebsiteConfiguration>"#;
 		let conf: WebsiteConfiguration = from_str(message).unwrap();
@@ -326,8 +429,9 @@ mod tests {
 				hostname: Value("garage.tld".to_owned()),
 				protocol: Some(Value("https".to_owned())),
 			}),
-			routing_rules: Some(vec![RoutingRule {
-				inner: RoutingRuleInner {
+			routing_rules: RoutingRules {
+				rules: vec![
+					RoutingRule {
 						condition: Some(Condition {
 							http_error_code: Some(IntValue(404)),
 							prefix: Some(Value("prefix1".to_owned())),
@@ -340,7 +444,21 @@ mod tests {
 							replace_full: Some(Value("fullkey".to_owned())),
 						},
 					},
-			}]),
+					RoutingRule {
+						condition: Some(Condition {
+							http_error_code: None,
+							prefix: Some(Value("".to_owned())),
+						}),
+						redirect: Redirect {
+							hostname: None,
+							protocol: None,
+							http_redirect_code: Some(IntValue(404)),
+							replace_prefix: None,
+							replace_full: Some(Value("missing".to_owned())),
+						},
+					},
+				],
+			},
 		};
 		assert_eq! {
 			ref_value,

src/api/s3/xml.rs

@@ -13,10 +13,6 @@ pub fn xmlns_tag<S: Serializer>(_v: &(), s: S) -> Result<S::Ok, S::Error> {
 	s.serialize_str("http://s3.amazonaws.com/doc/2006-03-01/")
 }
 
-pub fn xmlns_xsi_tag<S: Serializer>(_v: &(), s: S) -> Result<S::Ok, S::Error> {
-	s.serialize_str("http://www.w3.org/2001/XMLSchema-instance")
-}
-
 #[derive(Debug, Serialize, Deserialize, PartialEq, Eq, PartialOrd, Ord)]
 pub struct Value(#[serde(rename = "$value")] pub String);
 
@@ -139,6 +135,8 @@ pub struct CompleteMultipartUploadResult {
 	pub checksum_crc32: Option<Value>,
 	#[serde(rename = "ChecksumCRC32C")]
 	pub checksum_crc32c: Option<Value>,
+	#[serde(rename = "ChecksumCR64NVME")]
+	pub checksum_crc64nvme: Option<Value>,
 	#[serde(rename = "ChecksumSHA1")]
 	pub checksum_sha1: Option<Value>,
 	#[serde(rename = "ChecksumSHA256")]
@@ -213,6 +211,8 @@ pub struct PartItem {
 	pub checksum_crc32: Option<Value>,
 	#[serde(rename = "ChecksumCRC32C")]
 	pub checksum_crc32c: Option<Value>,
+	#[serde(rename = "ChecksumCRC64NVME")]
+	pub checksum_crc64nvme: Option<Value>,
 	#[serde(rename = "ChecksumSHA1")]
 	pub checksum_sha1: Option<Value>,
 	#[serde(rename = "ChecksumSHA256")]
@@ -323,42 +323,6 @@ pub struct PostObject {
 	pub etag: Value,
 }
 
-#[derive(Debug, Serialize, PartialEq, Eq)]
-pub struct Grantee {
-	#[serde(rename = "xmlns:xsi", serialize_with = "xmlns_xsi_tag")]
-	pub xmlns_xsi: (),
-	#[serde(rename = "xsi:type")]
-	pub typ: String,
-	#[serde(rename = "DisplayName")]
-	pub display_name: Option<Value>,
-	#[serde(rename = "ID")]
-	pub id: Option<Value>,
-}
-
-#[derive(Debug, Serialize, PartialEq, Eq)]
-pub struct Grant {
-	#[serde(rename = "Grantee")]
-	pub grantee: Grantee,
-	#[serde(rename = "Permission")]
-	pub permission: Value,
-}
-
-#[derive(Debug, Serialize, PartialEq, Eq)]
-pub struct AccessControlList {
-	#[serde(rename = "Grant")]
-	pub entries: Vec<Grant>,
-}
-
-#[derive(Debug, Serialize, PartialEq, Eq)]
-pub struct AccessControlPolicy {
-	#[serde(serialize_with = "xmlns_tag")]
-	pub xmlns: (),
-	#[serde(rename = "Owner")]
-	pub owner: Option<Owner>,
-	#[serde(rename = "AccessControlList")]
-	pub acl: AccessControlList,
-}
-
 #[cfg(test)]
 mod tests {
 	use super::*;
@@ -467,43 +431,6 @@ mod tests {
 		Ok(())
 	}
 
-	#[test]
-	fn get_bucket_acl_result() -> Result<(), ApiError> {
-		let grant = Grant {
-			grantee: Grantee {
-				xmlns_xsi: (),
-				typ: "CanonicalUser".to_string(),
-				display_name: Some(Value("owner_name".to_string())),
-				id: Some(Value("qsdfjklm".to_string())),
-			},
-			permission: Value("FULL_CONTROL".to_string()),
-		};
-
-		let get_bucket_acl = AccessControlPolicy {
-			xmlns: (),
-			owner: None,
-			acl: AccessControlList {
-				entries: vec![grant],
-			},
-		};
-		assert_eq!(
-			to_xml_with_header(&get_bucket_acl)?,
-			"<?xml version=\"1.0\" encoding=\"UTF-8\"?>\
-<AccessControlPolicy xmlns=\"http://s3.amazonaws.com/doc/2006-03-01/\">\
-	<AccessControlList>\
-		<Grant>\
-			<Grantee xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\" xsi:type=\"CanonicalUser\">\
-				<DisplayName>owner_name</DisplayName>\
-				<ID>qsdfjklm</ID>\
-			</Grantee>\
-			<Permission>FULL_CONTROL</Permission>\
-		</Grant>\
-	</AccessControlList>\
-</AccessControlPolicy>"
-		);
-		Ok(())
-	}
-
 	#[test]
 	fn delete_result() -> Result<(), ApiError> {
 		let delete_result = DeleteResult {
@@ -595,6 +522,7 @@ mod tests {
 			etag: Value("\"3858f62230ac3c915f300c664312c11f-9\"".to_string()),
 			checksum_crc32: None,
 			checksum_crc32c: None,
+			checksum_crc64nvme: None,
 			checksum_sha1: Some(Value("ZJAnHyG8PeKz9tI8UTcHrJos39A=".into())),
 			checksum_sha256: None,
 		};
@@ -880,6 +808,7 @@ mod tests {
 					size: IntValue(10485760),
 					checksum_crc32: None,
 					checksum_crc32c: None,
+					checksum_crc64nvme: None,
 					checksum_sha256: Some(Value(
 						"5RQ3A5uk0w7ojNjvegohch4JRBBGN/cLhsNrPzfv/hA=".into(),
 					)),
@@ -893,6 +822,7 @@ mod tests {
 					checksum_sha256: None,
 					checksum_crc32c: None,
 					checksum_crc32: Some(Value("ZJAnHyG8=".into())),
+					checksum_crc64nvme: None,
 					checksum_sha1: None,
 				},
 			],

src/block/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "garage_block"
-version = "1.3.1"
+version = "2.0.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -18,7 +18,6 @@ garage_db.workspace = true
 garage_net.workspace = true
 garage_rpc.workspace = true
 garage_util.workspace = true
-garage_table.workspace = true
 
 opentelemetry.workspace = true
 

src/block/manager.rs

@@ -33,8 +33,6 @@ use garage_rpc::rpc_helper::OrderTag;
 use garage_rpc::system::System;
 use garage_rpc::*;
 
-use garage_table::replication::{TableReplication, TableShardedReplication};
-
 use crate::block::*;
 use crate::layout::*;
 use crate::metrics::*;
@@ -50,8 +48,6 @@ pub const INLINE_THRESHOLD: usize = 3072;
 // to delete the block locally.
 pub(crate) const BLOCK_GC_DELAY: Duration = Duration::from_secs(600);
 
-const BLOCK_READ_SEMAPHORE_TIMEOUT: Duration = Duration::from_secs(15);
-
 /// RPC messages used to share blocks of data between nodes
 #[derive(Debug, Serialize, Deserialize)]
 pub enum BlockRpc {
@@ -76,8 +72,8 @@ impl Rpc for BlockRpc {
 
 /// The block manager, handling block exchange between nodes, and block storage on local node
 pub struct BlockManager {
-	/// Replication strategy, allowing to find on which node blocks should be located
-	pub replication: TableShardedReplication,
+	/// Quorum of nodes for write operations
+	pub write_quorum: usize,
 
 	/// Data layout
 	pub(crate) data_layout: ArcSwap<DataLayout>,
@@ -89,7 +85,6 @@ pub struct BlockManager {
 	disable_scrub: bool,
 
 	mutation_lock: Vec<Mutex<BlockManagerLocked>>,
-	read_semaphore: Semaphore,
 
 	pub rc: BlockRc,
 	pub resync: BlockResyncManager,
@@ -125,7 +120,7 @@ impl BlockManager {
 	pub fn new(
 		db: &db::Db,
 		config: &Config,
-		replication: TableShardedReplication,
+		write_quorum: usize,
 		system: Arc<System>,
 	) -> Result<Arc<Self>, Error> {
 		// Load or compute layout, i.e. assignment of data blocks to the different data directories
@@ -169,7 +164,7 @@ impl BlockManager {
 		let scrub_persister = PersisterShared::new(&system.metadata_dir, "scrub_info");
 
 		let block_manager = Arc::new(Self {
-			replication,
+			write_quorum,
 			data_layout: ArcSwap::new(Arc::new(data_layout)),
 			data_layout_persister,
 			data_fsync: config.data_fsync,
@@ -179,8 +174,6 @@ impl BlockManager {
 				.iter()
 				.map(|_| Mutex::new(BlockManagerLocked()))
 				.collect::<Vec<_>>(),
-
-			read_semaphore: Semaphore::new(config.block_max_concurrent_reads),
 			rc,
 			resync,
 			system,
@@ -343,6 +336,19 @@ impl BlockManager {
 		Err(err)
 	}
 
+	/// Returns the set of nodes that should store a copy of a given block.
+	/// These are the nodes assigned to the block's hash in the current
+	/// layout version only: since blocks are immutable, we don't need to
+	/// do complex logic when several layour versions are active at once,
+	/// just move them directly to the new nodes.
+	pub(crate) fn storage_nodes_of(&self, hash: &Hash) -> Vec<Uuid> {
+		self.system
+			.cluster_layout()
+			.current()
+			.nodes_of(hash)
+			.collect()
+	}
+
 	// ---- Public interface ----
 
 	/// Ask nodes that might have a block for it, return it as a stream
@@ -375,7 +381,7 @@ impl BlockManager {
 		prevent_compression: bool,
 		order_tag: Option<OrderTag>,
 	) -> Result<(), Error> {
-		let who = self.system.cluster_layout().current_storage_nodes_of(&hash);
+		let who = self.storage_nodes_of(&hash);
 
 		let compression_level = self.compression_level.filter(|_| !prevent_compression);
 		let (header, bytes) = DataBlock::from_buffer(data, compression_level)
@@ -405,7 +411,7 @@ impl BlockManager {
 				put_block_rpc,
 				RequestStrategy::with_priority(PRIO_NORMAL | PRIO_SECONDARY)
 					.with_drop_on_completion(permit)
-					.with_quorum(self.replication.write_quorum()),
+					.with_quorum(self.write_quorum),
 			)
 			.await?;
 
@@ -413,8 +419,8 @@ impl BlockManager {
 	}
 
 	/// Get number of items in the refcount table
-	pub fn rc_approximate_len(&self) -> Result<usize, Error> {
-		Ok(self.rc.rc_table.approximate_len()?)
+	pub fn rc_len(&self) -> Result<usize, Error> {
+		Ok(self.rc.rc_table.len()?)
 	}
 
 	/// Send command to start/stop/manager scrub worker
@@ -432,7 +438,7 @@ impl BlockManager {
 
 	/// List all resync errors
 	pub fn list_resync_errors(&self) -> Result<Vec<BlockResyncErrorInfo>, Error> {
-		let mut blocks = Vec::with_capacity(self.resync.errors.approximate_len()?);
+		let mut blocks = Vec::with_capacity(self.resync.errors.len()?);
 		for ent in self.resync.errors.iter()? {
 			let (hash, cnt) = ent?;
 			let cnt = ErrorCounter::decode(&cnt);
@@ -562,6 +568,9 @@ impl BlockManager {
 			match self.find_block(hash).await {
 				Some(p) => self.read_block_from(hash, &p).await,
 				None => {
+					// Not found but maybe we should have had it ??
+					self.resync
+						.put_to_resync(hash, 2 * self.system.rpc_helper().rpc_timeout())?;
 					return Err(Error::Message(format!(
 						"block {:?} not found on node",
 						hash
@@ -583,15 +592,6 @@ impl BlockManager {
 	) -> Result<DataBlock, Error> {
 		let (header, path) = block_path.as_parts_ref();
 
-		let permit = tokio::select! {
-			sem = self.read_semaphore.acquire() => sem.ok_or_message("acquire read semaphore")?,
-			_ = tokio::time::sleep(BLOCK_READ_SEMAPHORE_TIMEOUT) => {
-				self.metrics.block_read_semaphore_timeouts.add(1);
-				debug!("read block {:?}: read_semaphore acquire timeout", hash);
-				return Err(Error::Message("read block: read_semaphore acquire timeout".into()));
-			}
-		};
-
 		let mut f = fs::File::open(&path).await?;
 		let mut data = vec![];
 		f.read_to_end(&mut data).await?;
@@ -616,8 +616,6 @@ impl BlockManager {
 			return Err(Error::CorruptData(*hash));
 		}
 
-		drop(permit);
-
 		Ok(data)
 	}
 
@@ -783,7 +781,6 @@ impl BlockManagerLocked {
 
 		let mut f = fs::File::create(&path_tmp).await?;
 		f.write_all(data).await?;
-		f.flush().await?;
 		mgr.metrics.bytes_written.add(data.len() as u64);
 
 		if mgr.data_fsync {

src/block/metrics.rs

@@ -22,7 +22,6 @@ pub struct BlockManagerMetrics {
 
 	pub(crate) bytes_read: BoundCounter<u64>,
 	pub(crate) block_read_duration: BoundValueRecorder<f64>,
-	pub(crate) block_read_semaphore_timeouts: BoundCounter<u64>,
 	pub(crate) bytes_written: BoundCounter<u64>,
 	pub(crate) block_write_duration: BoundValueRecorder<f64>,
 	pub(crate) delete_counter: BoundCounter<u64>,
@@ -51,7 +50,7 @@ impl BlockManagerMetrics {
 				.init(),
 			_rc_size: meter
 				.u64_value_observer("block.rc_size", move |observer| {
-					if let Ok(value) = rc_tree.approximate_len() {
+					if let Ok(value) = rc_tree.len() {
 						observer.observe(value as u64, &[])
 					}
 				})
@@ -59,7 +58,7 @@ impl BlockManagerMetrics {
 				.init(),
 			_resync_queue_len: meter
 				.u64_value_observer("block.resync_queue_length", move |observer| {
-					if let Ok(value) = resync_queue.approximate_len() {
+					if let Ok(value) = resync_queue.len() {
 						observer.observe(value as u64, &[]);
 					}
 				})
@@ -69,7 +68,7 @@ impl BlockManagerMetrics {
 				.init(),
 			_resync_errored_blocks: meter
 				.u64_value_observer("block.resync_errored_blocks", move |observer| {
-					if let Ok(value) = resync_errors.approximate_len() {
+					if let Ok(value) = resync_errors.len() {
 						observer.observe(value as u64, &[]);
 					}
 				})
@@ -120,11 +119,6 @@ impl BlockManagerMetrics {
 				.with_description("Duration of block read operations")
 				.init()
 				.bind(&[]),
-			block_read_semaphore_timeouts: meter
-				.u64_counter("block.read_semaphore_timeouts")
-				.with_description("Number of block reads that failed due to semaphore acquire timeout")
-				.init()
-				.bind(&[]),
 			bytes_written: meter
 				.u64_counter("block.bytes_written")
 				.with_description("Number of bytes written to disk")

src/block/resync.rs

@@ -27,8 +27,6 @@ use garage_util::tranquilizer::Tranquilizer;
 use garage_rpc::system::System;
 use garage_rpc::*;
 
-use garage_table::replication::TableReplication;
-
 use crate::manager::*;
 
 // The delay between the time where a resync operation fails
@@ -106,13 +104,13 @@ impl BlockResyncManager {
 	}
 
 	/// Get length of resync queue
-	pub fn queue_approximate_len(&self) -> Result<usize, Error> {
-		Ok(self.queue.approximate_len()?)
+	pub fn queue_len(&self) -> Result<usize, Error> {
+		Ok(self.queue.len()?)
 	}
 
 	/// Get number of blocks that have an error
-	pub fn errors_approximate_len(&self) -> Result<usize, Error> {
-		Ok(self.errors.approximate_len()?)
+	pub fn errors_len(&self) -> Result<usize, Error> {
+		Ok(self.errors.len()?)
 	}
 
 	/// Clear the error counter for a block and put it in queue immediately
@@ -133,14 +131,6 @@ impl BlockResyncManager {
 		)))
 	}
 
-	/// Clear the entire resync queue and list of errored blocks
-	/// Corresponds to `garage repair clear-resync-queue`
-	pub fn clear_resync_queue(&self) -> Result<(), Error> {
-		self.queue.clear()?;
-		self.errors.clear()?;
-		Ok(())
-	}
-
 	pub fn register_bg_vars(&self, vars: &mut vars::BgVars) {
 		let notify = self.notify.clone();
 		vars.register_rw(
@@ -385,11 +375,8 @@ impl BlockResyncManager {
 			info!("Resync block {:?}: offloading and deleting", hash);
 			let existing_path = existing_path.unwrap();
 
-			let mut who = manager
-				.system
-				.cluster_layout()
-				.current_storage_nodes_of(hash);
-			if who.len() < manager.replication.write_quorum() {
+			let mut who = manager.storage_nodes_of(hash);
+			if who.len() < manager.write_quorum {
 				return Err(Error::Message("Not trying to offload block because we don't have a quorum of nodes to write to".to_string()));
 			}
 			who.retain(|id| *id != manager.system.id);
@@ -471,10 +458,7 @@ impl BlockResyncManager {
 
 			// First, check whether we are still supposed to store that
 			// block in the latest cluster layout version.
-			let storage_nodes = manager
-				.system
-				.cluster_layout()
-				.current_storage_nodes_of(&hash);
+			let storage_nodes = manager.storage_nodes_of(&hash);
 
 			if !storage_nodes.contains(&manager.system.id) {
 				info!(
@@ -556,11 +540,9 @@ impl Worker for ResyncWorker {
 		}
 
 		WorkerStatus {
-			queue_length: Some(self.manager.resync.queue_approximate_len().unwrap_or(0) as u64),
+			queue_length: Some(self.manager.resync.queue_len().unwrap_or(0) as u64),
 			tranquility: Some(tranquility),
-			persistent_errors: Some(
-				self.manager.resync.errors_approximate_len().unwrap_or(0) as u64
-			),
+			persistent_errors: Some(self.manager.resync.errors_len().unwrap_or(0) as u64),
 			..Default::default()
 		}
 	}

src/db/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "garage_db"
-version = "1.3.1"
+version = "2.0.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -12,18 +12,14 @@ readme = "../../README.md"
 path = "lib.rs"
 
 [dependencies]
-thiserror.workspace = true
+err-derive.workspace = true
 tracing.workspace = true
 
 heed = { workspace = true, optional = true }
-
 rusqlite = { workspace = true, optional = true, features = ["backup"] }
 r2d2 = { workspace = true, optional = true }
 r2d2_sqlite = { workspace = true, optional = true }
 
-fjall = { workspace = true, optional = true }
-parking_lot = { workspace = true, optional = true }
-
 [dev-dependencies]
 mktemp.workspace = true
 
@@ -31,5 +27,4 @@ mktemp.workspace = true
 default = [ "lmdb", "sqlite" ]
 bundled-libs = [ "rusqlite?/bundled" ]
 lmdb = [ "heed" ]
-fjall = [ "dep:fjall", "dep:parking_lot" ]
 sqlite = [ "rusqlite", "r2d2", "r2d2_sqlite" ]

src/db/fjall_adapter.rs

@@ -1,453 +0,0 @@
-use core::ops::Bound;
-
-use std::path::PathBuf;
-use std::sync::Arc;
-
-use parking_lot::{MappedRwLockReadGuard, RwLock, RwLockReadGuard};
-
-use fjall::{
-	PartitionCreateOptions, PersistMode, TransactionalKeyspace, TransactionalPartitionHandle,
-	WriteTransaction,
-};
-
-use crate::{
-	open::{Engine, OpenOpt},
-	Db, Error, IDb, ITx, ITxFn, OnCommit, Result, TxError, TxFnResult, TxOpError, TxOpResult,
-	TxResult, TxValueIter, Value, ValueIter,
-};
-
-pub use fjall;
-
-// --
-
-pub(crate) fn open_db(path: &PathBuf, opt: &OpenOpt) -> Result<Db> {
-	info!("Opening Fjall database at: {}", path.display());
-	if opt.fsync {
-		return Err(Error(
-			"metadata_fsync is not supported with the Fjall database engine".into(),
-		));
-	}
-	let mut config = fjall::Config::new(path);
-	if let Some(block_cache_size) = opt.fjall_block_cache_size {
-		config = config.cache_size(block_cache_size as u64);
-	}
-	let keyspace = config.open_transactional()?;
-	Ok(FjallDb::init(keyspace))
-}
-
-// -- err
-
-impl From<fjall::Error> for Error {
-	fn from(e: fjall::Error) -> Error {
-		Error(format!("fjall: {}", e).into())
-	}
-}
-
-impl From<fjall::LsmError> for Error {
-	fn from(e: fjall::LsmError) -> Error {
-		Error(format!("fjall lsm_tree: {}", e).into())
-	}
-}
-
-impl From<fjall::Error> for TxOpError {
-	fn from(e: fjall::Error) -> TxOpError {
-		TxOpError(e.into())
-	}
-}
-
-// -- db
-
-pub struct FjallDb {
-	keyspace: TransactionalKeyspace,
-	trees: RwLock<Vec<(String, TransactionalPartitionHandle)>>,
-}
-
-type ByteRefRangeBound<'r> = (Bound<&'r [u8]>, Bound<&'r [u8]>);
-
-impl FjallDb {
-	pub fn init(keyspace: TransactionalKeyspace) -> Db {
-		let s = Self {
-			keyspace,
-			trees: RwLock::new(Vec::new()),
-		};
-		Db(Arc::new(s))
-	}
-
-	fn get_tree(
-		&self,
-		i: usize,
-	) -> Result<MappedRwLockReadGuard<'_, TransactionalPartitionHandle>> {
-		RwLockReadGuard::try_map(self.trees.read(), |trees: &Vec<_>| {
-			trees.get(i).map(|tup| &tup.1)
-		})
-		.map_err(|_| Error("invalid tree id".into()))
-	}
-}
-
-impl IDb for FjallDb {
-	fn engine(&self) -> String {
-		"Fjall (EXPERIMENTAL!)".into()
-	}
-
-	fn open_tree(&self, name: &str) -> Result<usize> {
-		let mut trees = self.trees.write();
-		let safe_name = encode_name(name)?;
-		if let Some(i) = trees.iter().position(|(name, _)| *name == safe_name) {
-			Ok(i)
-		} else {
-			let tree = self
-				.keyspace
-				.open_partition(&safe_name, PartitionCreateOptions::default())?;
-			let i = trees.len();
-			trees.push((safe_name, tree));
-			Ok(i)
-		}
-	}
-
-	fn list_trees(&self) -> Result<Vec<String>> {
-		Ok(self
-			.keyspace
-			.list_partitions()
-			.iter()
-			.map(|n| decode_name(&n))
-			.collect::<Result<Vec<_>>>()?)
-	}
-
-	fn snapshot(&self, base_path: &PathBuf) -> Result<()> {
-		std::fs::create_dir_all(base_path)?;
-		let path = Engine::Fjall.db_path(base_path);
-
-		let source_state = self.keyspace.read_tx();
-		let copy_keyspace = fjall::Config::new(path).open()?;
-
-		for partition_name in self.keyspace.list_partitions() {
-			let source_partition = self
-				.keyspace
-				.open_partition(&partition_name, PartitionCreateOptions::default())?;
-			let copy_partition =
-				copy_keyspace.open_partition(&partition_name, PartitionCreateOptions::default())?;
-
-			for entry in source_state.iter(&source_partition) {
-				let (key, value) = entry?;
-				copy_partition.insert(key, value)?;
-			}
-		}
-
-		copy_keyspace.persist(PersistMode::SyncAll)?;
-		Ok(())
-	}
-
-	// ----
-
-	fn get(&self, tree_idx: usize, key: &[u8]) -> Result<Option<Value>> {
-		let tree = self.get_tree(tree_idx)?;
-		let tx = self.keyspace.read_tx();
-		let val = tx.get(&tree, key)?;
-		match val {
-			None => Ok(None),
-			Some(v) => Ok(Some(v.to_vec())),
-		}
-	}
-
-	fn approximate_len(&self, tree_idx: usize) -> Result<usize> {
-		let tree = self.get_tree(tree_idx)?;
-		Ok(tree.approximate_len())
-	}
-	fn is_empty(&self, tree_idx: usize) -> Result<bool> {
-		let tree = self.get_tree(tree_idx)?;
-		let tx = self.keyspace.read_tx();
-		Ok(tx.is_empty(&tree)?)
-	}
-
-	fn insert(&self, tree_idx: usize, key: &[u8], value: &[u8]) -> Result<()> {
-		let tree = self.get_tree(tree_idx)?;
-		let mut tx = self.keyspace.write_tx();
-		tx.insert(&tree, key, value);
-		tx.commit()?;
-		Ok(())
-	}
-
-	fn remove(&self, tree_idx: usize, key: &[u8]) -> Result<()> {
-		let tree = self.get_tree(tree_idx)?;
-		let mut tx = self.keyspace.write_tx();
-		tx.remove(&tree, key);
-		tx.commit()?;
-		Ok(())
-	}
-
-	fn clear(&self, tree_idx: usize) -> Result<()> {
-		let mut trees = self.trees.write();
-
-		if tree_idx >= trees.len() {
-			return Err(Error("invalid tree id".into()));
-		}
-		let (name, tree) = trees.remove(tree_idx);
-
-		self.keyspace.delete_partition(tree)?;
-		let tree = self
-			.keyspace
-			.open_partition(&name, PartitionCreateOptions::default())?;
-		trees.insert(tree_idx, (name, tree));
-
-		Ok(())
-	}
-
-	fn iter(&self, tree_idx: usize) -> Result<ValueIter<'_>> {
-		let tree = self.get_tree(tree_idx)?;
-		let tx = self.keyspace.read_tx();
-		Ok(Box::new(tx.iter(&tree).map(iterator_remap)))
-	}
-
-	fn iter_rev(&self, tree_idx: usize) -> Result<ValueIter<'_>> {
-		let tree = self.get_tree(tree_idx)?;
-		let tx = self.keyspace.read_tx();
-		Ok(Box::new(tx.iter(&tree).rev().map(iterator_remap)))
-	}
-
-	fn range<'r>(
-		&self,
-		tree_idx: usize,
-		low: Bound<&'r [u8]>,
-		high: Bound<&'r [u8]>,
-	) -> Result<ValueIter<'_>> {
-		let tree = self.get_tree(tree_idx)?;
-		let tx = self.keyspace.read_tx();
-		Ok(Box::new(
-			tx.range::<&'r [u8], ByteRefRangeBound>(&tree, (low, high))
-				.map(iterator_remap),
-		))
-	}
-	fn range_rev<'r>(
-		&self,
-		tree_idx: usize,
-		low: Bound<&'r [u8]>,
-		high: Bound<&'r [u8]>,
-	) -> Result<ValueIter<'_>> {
-		let tree = self.get_tree(tree_idx)?;
-		let tx = self.keyspace.read_tx();
-		Ok(Box::new(
-			tx.range::<&'r [u8], ByteRefRangeBound>(&tree, (low, high))
-				.rev()
-				.map(iterator_remap),
-		))
-	}
-
-	// ----
-
-	fn transaction(&self, f: &dyn ITxFn) -> TxResult<OnCommit, ()> {
-		let trees = self.trees.read();
-		let mut tx = FjallTx {
-			trees: &trees[..],
-			tx: self.keyspace.write_tx(),
-		};
-
-		let res = f.try_on(&mut tx);
-		match res {
-			TxFnResult::Ok(on_commit) => {
-				tx.tx.commit().map_err(Error::from).map_err(TxError::Db)?;
-				Ok(on_commit)
-			}
-			TxFnResult::Abort => {
-				tx.tx.rollback();
-				Err(TxError::Abort(()))
-			}
-			TxFnResult::DbErr => {
-				tx.tx.rollback();
-				Err(TxError::Db(Error(
-					"(this message will be discarded)".into(),
-				)))
-			}
-		}
-	}
-}
-
-// ----
-
-struct FjallTx<'a> {
-	trees: &'a [(String, TransactionalPartitionHandle)],
-	tx: WriteTransaction<'a>,
-}
-
-impl<'a> FjallTx<'a> {
-	fn get_tree(&self, i: usize) -> TxOpResult<&TransactionalPartitionHandle> {
-		self.trees.get(i).map(|tup| &tup.1).ok_or_else(|| {
-			TxOpError(Error(
-				"invalid tree id (it might have been openned after the transaction started)".into(),
-			))
-		})
-	}
-}
-
-impl<'a> ITx for FjallTx<'a> {
-	fn get(&self, tree_idx: usize, key: &[u8]) -> TxOpResult<Option<Value>> {
-		let tree = self.get_tree(tree_idx)?;
-		match self.tx.get(tree, key)? {
-			Some(v) => Ok(Some(v.to_vec())),
-			None => Ok(None),
-		}
-	}
-	fn len(&self, tree_idx: usize) -> TxOpResult<usize> {
-		let tree = self.get_tree(tree_idx)?;
-		Ok(self.tx.len(tree)? as usize)
-	}
-
-	fn insert(&mut self, tree_idx: usize, key: &[u8], value: &[u8]) -> TxOpResult<()> {
-		let tree = self.get_tree(tree_idx)?.clone();
-		self.tx.insert(&tree, key, value);
-		Ok(())
-	}
-	fn remove(&mut self, tree_idx: usize, key: &[u8]) -> TxOpResult<()> {
-		let tree = self.get_tree(tree_idx)?.clone();
-		self.tx.remove(&tree, key);
-		Ok(())
-	}
-	fn clear(&mut self, _tree_idx: usize) -> TxOpResult<()> {
-		unimplemented!("LSM tree clearing in cross-partition transaction is not supported")
-	}
-
-	fn iter(&self, tree_idx: usize) -> TxOpResult<TxValueIter<'_>> {
-		let tree = self.get_tree(tree_idx)?.clone();
-		Ok(Box::new(self.tx.iter(&tree).map(iterator_remap_tx)))
-	}
-	fn iter_rev(&self, tree_idx: usize) -> TxOpResult<TxValueIter<'_>> {
-		let tree = self.get_tree(tree_idx)?.clone();
-		Ok(Box::new(self.tx.iter(&tree).rev().map(iterator_remap_tx)))
-	}
-
-	fn range<'r>(
-		&self,
-		tree_idx: usize,
-		low: Bound<&'r [u8]>,
-		high: Bound<&'r [u8]>,
-	) -> TxOpResult<TxValueIter<'_>> {
-		let tree = self.get_tree(tree_idx)?;
-		let low = clone_bound(low);
-		let high = clone_bound(high);
-		Ok(Box::new(
-			self.tx
-				.range::<Vec<u8>, ByteVecRangeBounds>(&tree, (low, high))
-				.map(iterator_remap_tx),
-		))
-	}
-	fn range_rev<'r>(
-		&self,
-		tree_idx: usize,
-		low: Bound<&'r [u8]>,
-		high: Bound<&'r [u8]>,
-	) -> TxOpResult<TxValueIter<'_>> {
-		let tree = self.get_tree(tree_idx)?;
-		let low = clone_bound(low);
-		let high = clone_bound(high);
-		Ok(Box::new(
-			self.tx
-				.range::<Vec<u8>, ByteVecRangeBounds>(&tree, (low, high))
-				.rev()
-				.map(iterator_remap_tx),
-		))
-	}
-}
-
-// -- maps fjall's (k, v) to ours
-
-fn iterator_remap(r: fjall::Result<(fjall::Slice, fjall::Slice)>) -> Result<(Value, Value)> {
-	r.map(|(k, v)| (k.to_vec(), v.to_vec()))
-		.map_err(|e| e.into())
-}
-
-fn iterator_remap_tx(r: fjall::Result<(fjall::Slice, fjall::Slice)>) -> TxOpResult<(Value, Value)> {
-	r.map(|(k, v)| (k.to_vec(), v.to_vec()))
-		.map_err(|e| e.into())
-}
-
-// -- utils to deal with Garage's tightness on Bound lifetimes
-
-type ByteVecBound = Bound<Vec<u8>>;
-type ByteVecRangeBounds = (ByteVecBound, ByteVecBound);
-
-fn clone_bound(bound: Bound<&[u8]>) -> ByteVecBound {
-	let value = match bound {
-		Bound::Excluded(v) | Bound::Included(v) => v.to_vec(),
-		Bound::Unbounded => vec![],
-	};
-
-	match bound {
-		Bound::Included(_) => Bound::Included(value),
-		Bound::Excluded(_) => Bound::Excluded(value),
-		Bound::Unbounded => Bound::Unbounded,
-	}
-}
-
-// -- utils to encode table names --
-
-fn encode_name(s: &str) -> Result<String> {
-	let base = 'A' as u32;
-
-	let mut ret = String::with_capacity(s.len() + 10);
-	for c in s.chars() {
-		if c.is_alphanumeric() || c == '_' || c == '-' || c == '#' {
-			ret.push(c);
-		} else if c <= u8::MAX as char {
-			ret.push('$');
-			let c_hi = c as u32 / 16;
-			let c_lo = c as u32 % 16;
-			ret.push(char::from_u32(base + c_hi).unwrap());
-			ret.push(char::from_u32(base + c_lo).unwrap());
-		} else {
-			return Err(Error(
-				format!("table name {} could not be safely encoded", s).into(),
-			));
-		}
-	}
-	Ok(ret)
-}
-
-fn decode_name(s: &str) -> Result<String> {
-	use std::convert::TryFrom;
-
-	let errfn = || Error(format!("encoded table name {} is invalid", s).into());
-	let c_map = |c: char| {
-		let c = c as u32;
-		let base = 'A' as u32;
-		if (base..base + 16).contains(&c) {
-			Some(c - base)
-		} else {
-			None
-		}
-	};
-
-	let mut ret = String::with_capacity(s.len());
-	let mut it = s.chars();
-	while let Some(c) = it.next() {
-		if c == '$' {
-			let c_hi = it.next().and_then(c_map).ok_or_else(errfn)?;
-			let c_lo = it.next().and_then(c_map).ok_or_else(errfn)?;
-			let c_dec = char::try_from(c_hi * 16 + c_lo).map_err(|_| errfn())?;
-			ret.push(c_dec);
-		} else {
-			ret.push(c);
-		}
-	}
-	Ok(ret)
-}
-
-#[cfg(test)]
-mod tests {
-	use super::*;
-
-	#[test]
-	fn test_encdec_name() {
-		for name in [
-			"testname",
-			"test_name",
-			"test name",
-			"test$name",
-			"test:name@help.me$get/this**right",
-		] {
-			let encname = encode_name(name).unwrap();
-			assert!(!encname.contains(' '));
-			assert!(!encname.contains('.'));
-			assert!(!encname.contains('*'));
-			assert_eq!(*name, decode_name(&encname).unwrap());
-		}
-	}
-}

src/db/lib.rs

@@ -1,8 +1,6 @@
 #[macro_use]
 extern crate tracing;
 
-#[cfg(feature = "fjall")]
-pub mod fjall_adapter;
 #[cfg(feature = "lmdb")]
 pub mod lmdb_adapter;
 #[cfg(feature = "sqlite")]
@@ -20,7 +18,7 @@ use std::cell::Cell;
 use std::path::PathBuf;
 use std::sync::Arc;
 
-use thiserror::Error;
+use err_derive::Error;
 
 pub use open::*;
 
@@ -44,7 +42,7 @@ pub type TxValueIter<'a> = Box<dyn std::iter::Iterator<Item = TxOpResult<(Value,
 // ----
 
 #[derive(Debug, Error)]
-#[error("{0}")]
+#[error(display = "{}", _0)]
 pub struct Error(pub Cow<'static, str>);
 
 impl From<std::io::Error> for Error {
@@ -56,7 +54,7 @@ impl From<std::io::Error> for Error {
 pub type Result<T> = std::result::Result<T, Error>;
 
 #[derive(Debug, Error)]
-#[error("{0}")]
+#[error(display = "{}", _0)]
 pub struct TxOpError(pub(crate) Error);
 pub type TxOpResult<T> = std::result::Result<T, TxOpError>;
 
@@ -106,44 +104,32 @@ impl Db {
 			result: Cell::new(None),
 		};
 		let tx_res = self.0.transaction(&f);
-		let fn_res = f.result.into_inner();
+		let ret = f
+			.result
+			.into_inner()
+			.expect("Transaction did not store result");
 
-		match (tx_res, fn_res) {
-			(Ok(on_commit), Some(Ok(value))) => {
-				// Transaction succeeded
-				// TxFn stored the value to return to the user in fn_res
-				// tx_res contains the on_commit list of callbacks, run them now
+		match tx_res {
+			Ok(on_commit) => match ret {
+				Ok(value) => {
 					on_commit.into_iter().for_each(|f| f());
 					Ok(value)
 				}
-			(Err(TxError::Abort(())), Some(Err(TxError::Abort(e)))) => {
-				// Transaction was aborted by user code
-				// The abort error value is stored in fn_res
-				Err(TxError::Abort(e))
-			}
-			(Err(TxError::Db(_tx_e)), Some(Err(TxError::Db(fn_e)))) => {
-				// Transaction encountered a DB error in user code
-				// The error value encountered is the one in fn_res,
-				// tx_res contains only a dummy error message
-				Err(TxError::Db(fn_e))
-			}
-			(Err(TxError::Db(tx_e)), None) => {
-				// Transaction encounterred a DB error when initializing the transaction,
-				// before user code was called
-				Err(TxError::Db(tx_e))
-			}
-			(Err(TxError::Db(tx_e)), Some(Ok(_))) => {
-				// Transaction encounterred a DB error when commiting the transaction,
-				// after user code was called
-				Err(TxError::Db(tx_e))
-			}
-			(tx_res, fn_res) => {
-				panic!(
-					"unexpected error case: tx_res={:?}, fn_res={:?}",
-					tx_res.map(|_| "..."),
-					fn_res.map(|x| x.map(|_| "...").map_err(|_| "..."))
-				);
-			}
+				_ => unreachable!(),
+			},
+			Err(TxError::Abort(())) => match ret {
+				Err(TxError::Abort(e)) => Err(TxError::Abort(e)),
+				_ => unreachable!(),
+			},
+			Err(TxError::Db(e2)) => match ret {
+				// Ok was stored -> the error occurred when finalizing
+				// transaction
+				Ok(_) => Err(TxError::Db(e2)),
+				// An error was already stored: that's the one we want to
+				// return
+				Err(TxError::Db(e)) => Err(TxError::Db(e)),
+				_ => unreachable!(),
+			},
 		}
 	}
 
@@ -166,7 +152,7 @@ impl Db {
 		let tree_names = other.list_trees()?;
 		for name in tree_names {
 			let tree = self.open_tree(&name)?;
-			if !tree.is_empty()? {
+			if tree.len()? > 0 {
 				return Err(Error(format!("tree {} already contains data", name).into()));
 			}
 
@@ -208,12 +194,8 @@ impl Tree {
 		self.0.get(self.1, key.as_ref())
 	}
 	#[inline]
-	pub fn approximate_len(&self) -> Result<usize> {
-		self.0.approximate_len(self.1)
-	}
-	#[inline]
-	pub fn is_empty(&self) -> Result<bool> {
-		self.0.is_empty(self.1)
+	pub fn len(&self) -> Result<usize> {
+		self.0.len(self.1)
 	}
 
 	#[inline]
@@ -351,8 +333,7 @@ pub(crate) trait IDb: Send + Sync {
 	fn snapshot(&self, path: &PathBuf) -> Result<()>;
 
 	fn get(&self, tree: usize, key: &[u8]) -> Result<Option<Value>>;
-	fn approximate_len(&self, tree: usize) -> Result<usize>;
-	fn is_empty(&self, tree: usize) -> Result<bool>;
+	fn len(&self, tree: usize) -> Result<usize>;
 
 	fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result<()>;
 	fn remove(&self, tree: usize, key: &[u8]) -> Result<()>;

src/db/lmdb_adapter.rs

@@ -1,8 +1,8 @@
 use core::ops::Bound;
+use core::ptr::NonNull;
 
 use std::collections::HashMap;
 use std::convert::TryInto;
-use std::marker::PhantomPinned;
 use std::path::PathBuf;
 use std::pin::Pin;
 use std::sync::{Arc, RwLock};
@@ -11,55 +11,12 @@ use heed::types::ByteSlice;
 use heed::{BytesDecode, Env, RoTxn, RwTxn, UntypedDatabase as Database};
 
 use crate::{
-	open::{Engine, OpenOpt},
 	Db, Error, IDb, ITx, ITxFn, OnCommit, Result, TxError, TxFnResult, TxOpError, TxOpResult,
 	TxResult, TxValueIter, Value, ValueIter,
 };
 
 pub use heed;
 
-// ---- top-level open function
-
-pub(crate) fn open_db(path: &PathBuf, opt: &OpenOpt) -> Result<Db> {
-	info!("Opening LMDB database at: {}", path.display());
-	if let Err(e) = std::fs::create_dir_all(&path) {
-		return Err(Error(
-			format!("Unable to create LMDB data directory: {}", e).into(),
-		));
-	}
-
-	let map_size = match opt.lmdb_map_size {
-		None => recommended_map_size(),
-		Some(v) => v - (v % 4096),
-	};
-
-	let mut env_builder = heed::EnvOpenOptions::new();
-	env_builder.max_dbs(100);
-	env_builder.map_size(map_size);
-	env_builder.max_readers(2048);
-	unsafe {
-		env_builder.flag(heed::flags::Flags::MdbNoRdAhead);
-		env_builder.flag(heed::flags::Flags::MdbNoMetaSync);
-		if !opt.fsync {
-			env_builder.flag(heed::flags::Flags::MdbNoSync);
-		}
-	}
-	match env_builder.open(&path) {
-		Err(heed::Error::Io(e)) if e.kind() == std::io::ErrorKind::OutOfMemory => {
-			return Err(Error(
-				"OutOfMemory error while trying to open LMDB database. This can happen \
-                if your operating system is not allowing you to use sufficient virtual \
-                memory address space. Please check that no limit is set (ulimit -v). \
-                You may also try to set a smaller `lmdb_map_size` configuration parameter. \
-                On 32-bit machines, you should probably switch to another database engine."
-					.into(),
-			))
-		}
-		Err(e) => Err(Error(format!("Cannot open LMDB database: {}", e).into())),
-		Ok(db) => Ok(LmdbDb::init(db)),
-	}
-}
-
 // -- err
 
 impl From<heed::Error> for Error {
@@ -147,11 +104,12 @@ impl IDb for LmdbDb {
 		Ok(ret2)
 	}
 
-	fn snapshot(&self, base_path: &PathBuf) -> Result<()> {
-		std::fs::create_dir_all(base_path)?;
-		let path = Engine::Lmdb.db_path(base_path);
+	fn snapshot(&self, to: &PathBuf) -> Result<()> {
+		std::fs::create_dir_all(to)?;
+		let mut path = to.clone();
+		path.push("data.mdb");
 		self.db
-			.copy_to_path(path, heed::CompactionOption::Enabled)?;
+			.copy_to_path(path, heed::CompactionOption::Disabled)?;
 		Ok(())
 	}
 
@@ -168,16 +126,11 @@ impl IDb for LmdbDb {
 		}
 	}
 
-	fn approximate_len(&self, tree: usize) -> Result<usize> {
+	fn len(&self, tree: usize) -> Result<usize> {
 		let tree = self.get_tree(tree)?;
 		let tx = self.db.read_txn()?;
 		Ok(tree.len(&tx)?.try_into().unwrap())
 	}
-	fn is_empty(&self, tree: usize) -> Result<bool> {
-		let tree = self.get_tree(tree)?;
-		let tx = self.db.read_txn()?;
-		Ok(tree.is_empty(&tx)?)
-	}
 
 	fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result<()> {
 		let tree = self.get_tree(tree)?;
@@ -206,15 +159,13 @@ impl IDb for LmdbDb {
 	fn iter(&self, tree: usize) -> Result<ValueIter<'_>> {
 		let tree = self.get_tree(tree)?;
 		let tx = self.db.read_txn()?;
-		// Safety: the cloture does not store its argument anywhere,
-		unsafe { TxAndIterator::make(tx, |tx| Ok(tree.iter(tx)?)) }
+		TxAndIterator::make(tx, |tx| Ok(tree.iter(tx)?))
 	}
 
 	fn iter_rev(&self, tree: usize) -> Result<ValueIter<'_>> {
 		let tree = self.get_tree(tree)?;
 		let tx = self.db.read_txn()?;
-		// Safety: the cloture does not store its argument anywhere,
-		unsafe { TxAndIterator::make(tx, |tx| Ok(tree.rev_iter(tx)?)) }
+		TxAndIterator::make(tx, |tx| Ok(tree.rev_iter(tx)?))
 	}
 
 	fn range<'r>(
@@ -225,8 +176,7 @@ impl IDb for LmdbDb {
 	) -> Result<ValueIter<'_>> {
 		let tree = self.get_tree(tree)?;
 		let tx = self.db.read_txn()?;
-		// Safety: the cloture does not store its argument anywhere,
-		unsafe { TxAndIterator::make(tx, |tx| Ok(tree.range(tx, &(low, high))?)) }
+		TxAndIterator::make(tx, |tx| Ok(tree.range(tx, &(low, high))?))
 	}
 	fn range_rev<'r>(
 		&self,
@@ -236,8 +186,7 @@ impl IDb for LmdbDb {
 	) -> Result<ValueIter<'_>> {
 		let tree = self.get_tree(tree)?;
 		let tx = self.db.read_txn()?;
-		// Safety: the cloture does not store its argument anywhere,
-		unsafe { TxAndIterator::make(tx, |tx| Ok(tree.rev_range(tx, &(low, high))?)) }
+		TxAndIterator::make(tx, |tx| Ok(tree.rev_range(tx, &(low, high))?))
 	}
 
 	// ----
@@ -367,41 +316,28 @@ where
 {
 	tx: RoTxn<'a>,
 	iter: Option<I>,
-	_pin: PhantomPinned,
 }
 
 impl<'a, I> TxAndIterator<'a, I>
 where
 	I: Iterator<Item = IteratorItem<'a>> + 'a,
 {
-	fn iter(self: Pin<&mut Self>) -> &mut Option<I> {
-		// Safety: iter is not structural
-		unsafe { &mut self.get_unchecked_mut().iter }
-	}
-
-	/// Safety: iterfun must not store its argument anywhere but in its result.
-	unsafe fn make<F>(tx: RoTxn<'a>, iterfun: F) -> Result<ValueIter<'a>>
+	fn make<F>(tx: RoTxn<'a>, iterfun: F) -> Result<ValueIter<'a>>
 	where
 		F: FnOnce(&'a RoTxn<'a>) -> Result<I>,
 	{
-		let res = TxAndIterator {
-			tx,
-			iter: None,
-			_pin: PhantomPinned,
-		};
+		let res = TxAndIterator { tx, iter: None };
 		let mut boxed = Box::pin(res);
 
-		let tx_lifetime_overextended: &'a RoTxn<'a> = {
-			let tx = &boxed.tx;
-			// Safety: Artificially extending the lifetime because
-			// this reference will only be stored and accessed from the
-			// returned ValueIter which guarantees that it is destroyed
-			// before the tx it is pointing  to.
-			unsafe { &*&raw const *tx }
-		};
-		let iter = iterfun(&tx_lifetime_overextended)?;
+		// This unsafe allows us to bypass lifetime checks
+		let tx = unsafe { NonNull::from(&boxed.tx).as_ref() };
+		let iter = iterfun(tx)?;
 
-		*boxed.as_mut().iter() = Some(iter);
+		let mut_ref = Pin::as_mut(&mut boxed);
+		// This unsafe allows us to write in a field of the pinned struct
+		unsafe {
+			Pin::get_unchecked_mut(mut_ref).iter = Some(iter);
+		}
 
 		Ok(Box::new(TxAndIteratorPin(boxed)))
 	}
@@ -412,10 +348,8 @@ where
 	I: Iterator<Item = IteratorItem<'a>> + 'a,
 {
 	fn drop(&mut self) {
-		// Safety: `new_unchecked` is okay because we know this value is never
-		// used again after being dropped.
-		let this = unsafe { Pin::new_unchecked(self) };
-		drop(this.iter().take());
+		// ensure the iterator is dropped before the RoTxn it references
+		drop(self.iter.take());
 	}
 }
 
@@ -431,12 +365,13 @@ where
 
 	fn next(&mut self) -> Option<Self::Item> {
 		let mut_ref = Pin::as_mut(&mut self.0);
-		let next = mut_ref.iter().as_mut()?.next()?;
-		let res = match next {
-			Err(e) => Err(e.into()),
-			Ok((k, v)) => Ok((k.to_vec(), v.to_vec())),
-		};
-		Some(res)
+		// This unsafe allows us to mutably access the iterator field
+		let next = unsafe { Pin::get_unchecked_mut(mut_ref).iter.as_mut()?.next() };
+		match next {
+			None => None,
+			Some(Err(e)) => Some(Err(e.into())),
+			Some(Ok((k, v))) => Some(Ok((k.to_vec(), v.to_vec()))),
+		}
 	}
 }
 

src/db/open.rs

@@ -11,7 +11,6 @@ use crate::{Db, Error, Result};
 pub enum Engine {
 	Lmdb,
 	Sqlite,
-	Fjall,
 }
 
 impl Engine {
@@ -20,26 +19,8 @@ impl Engine {
 		match self {
 			Self::Lmdb => "lmdb",
 			Self::Sqlite => "sqlite",
-			Self::Fjall => "fjall",
 		}
 	}
-
-	/// Return engine-specific DB path from base path
-	pub fn db_path(&self, base_path: &PathBuf) -> PathBuf {
-		let mut ret = base_path.clone();
-		match self {
-			Self::Lmdb => {
-				ret.push("db.lmdb");
-			}
-			Self::Sqlite => {
-				ret.push("db.sqlite");
-			}
-			Self::Fjall => {
-				ret.push("db.fjall");
-			}
-		}
-		ret
-	}
 }
 
 impl std::fmt::Display for Engine {
@@ -55,11 +36,10 @@ impl std::str::FromStr for Engine {
 		match text {
 			"lmdb" | "heed" => Ok(Self::Lmdb),
 			"sqlite" | "sqlite3" | "rusqlite" => Ok(Self::Sqlite),
-            "fjall" => Ok(Self::Fjall),
 			"sled" => Err(Error("Sled is no longer supported as a database engine. Converting your old metadata db can be done using an older Garage binary (e.g. v0.9.4).".into())),
 			kind => Err(Error(
 				format!(
-					"Invalid DB engine: {} (options are: lmdb, sqlite, fjall)",
+					"Invalid DB engine: {} (options are: lmdb, sqlite)",
 					kind
 				)
 				.into(),
@@ -71,7 +51,6 @@ impl std::str::FromStr for Engine {
 pub struct OpenOpt {
 	pub fsync: bool,
 	pub lmdb_map_size: Option<usize>,
-	pub fjall_block_cache_size: Option<usize>,
 }
 
 impl Default for OpenOpt {
@@ -79,7 +58,6 @@ impl Default for OpenOpt {
 		Self {
 			fsync: false,
 			lmdb_map_size: None,
-			fjall_block_cache_size: None,
 		}
 	}
 }
@@ -88,15 +66,53 @@ pub fn open_db(path: &PathBuf, engine: Engine, opt: &OpenOpt) -> Result<Db> {
 	match engine {
 		// ---- Sqlite DB ----
 		#[cfg(feature = "sqlite")]
-		Engine::Sqlite => crate::sqlite_adapter::open_db(path, opt),
+		Engine::Sqlite => {
+			info!("Opening Sqlite database at: {}", path.display());
+			let manager = r2d2_sqlite::SqliteConnectionManager::file(path);
+			Ok(crate::sqlite_adapter::SqliteDb::new(manager, opt.fsync)?)
+		}
 
 		// ---- LMDB DB ----
 		#[cfg(feature = "lmdb")]
-		Engine::Lmdb => crate::lmdb_adapter::open_db(path, opt),
+		Engine::Lmdb => {
+			info!("Opening LMDB database at: {}", path.display());
+			if let Err(e) = std::fs::create_dir_all(&path) {
+				return Err(Error(
+					format!("Unable to create LMDB data directory: {}", e).into(),
+				));
+			}
 
-		// ---- Fjall DB ----
-		#[cfg(feature = "fjall")]
-		Engine::Fjall => crate::fjall_adapter::open_db(path, opt),
+			let map_size = match opt.lmdb_map_size {
+				None => crate::lmdb_adapter::recommended_map_size(),
+				Some(v) => v - (v % 4096),
+			};
+
+			let mut env_builder = heed::EnvOpenOptions::new();
+			env_builder.max_dbs(100);
+			env_builder.map_size(map_size);
+			env_builder.max_readers(2048);
+			unsafe {
+				env_builder.flag(crate::lmdb_adapter::heed::flags::Flags::MdbNoRdAhead);
+				env_builder.flag(crate::lmdb_adapter::heed::flags::Flags::MdbNoMetaSync);
+				if !opt.fsync {
+					env_builder.flag(heed::flags::Flags::MdbNoSync);
+				}
+			}
+			match env_builder.open(&path) {
+				Err(heed::Error::Io(e)) if e.kind() == std::io::ErrorKind::OutOfMemory => {
+					return Err(Error(
+						"OutOfMemory error while trying to open LMDB database. This can happen \
+                        if your operating system is not allowing you to use sufficient virtual \
+                        memory address space. Please check that no limit is set (ulimit -v). \
+                        You may also try to set a smaller `lmdb_map_size` configuration parameter. \
+                        On 32-bit machines, you should probably switch to another database engine."
+							.into(),
+					))
+				}
+				Err(e) => Err(Error(format!("Cannot open LMDB database: {}", e).into())),
+				Ok(db) => Ok(crate::lmdb_adapter::LmdbDb::init(db)),
+			}
+		}
 
 		// Pattern is unreachable when all supported DB engines are compiled into binary. The allow
 		// attribute is added so that we won't have to change this match in case stop building

src/db/sqlite_adapter.rs

@@ -11,23 +11,12 @@ use r2d2_sqlite::SqliteConnectionManager;
 use rusqlite::{params, Rows, Statement, Transaction};
 
 use crate::{
-	open::{Engine, OpenOpt},
 	Db, Error, IDb, ITx, ITxFn, OnCommit, Result, TxError, TxFnResult, TxOpError, TxOpResult,
 	TxResult, TxValueIter, Value, ValueIter,
 };
 
 pub use rusqlite;
 
-// ---- top-level open function
-
-pub(crate) fn open_db(path: &PathBuf, opt: &OpenOpt) -> Result<Db> {
-	info!("Opening Sqlite database at: {}", path.display());
-	let manager = r2d2_sqlite::SqliteConnectionManager::file(path);
-	Ok(SqliteDb::new(manager, opt.fsync)?)
-}
-
-// ----
-
 type Connection = r2d2::PooledConnection<SqliteConnectionManager>;
 
 // --- err
@@ -150,18 +139,17 @@ impl IDb for SqliteDb {
 		Ok(trees)
 	}
 
-	fn snapshot(&self, base_path: &PathBuf) -> Result<()> {
-		std::fs::create_dir_all(base_path)?;
-		let path = Engine::Sqlite
-			.db_path(&base_path)
-			.into_os_string()
-			.into_string()
-			.map_err(|_| Error("invalid sqlite path string".into()))?;
-
-		info!("Start sqlite VACUUM INTO `{}`", path);
-		self.db.get()?.execute("VACUUM INTO ?1", params![path])?;
-		info!("Finished sqlite VACUUM INTO `{}`", path);
-
+	fn snapshot(&self, to: &PathBuf) -> Result<()> {
+		fn progress(p: rusqlite::backup::Progress) {
+			let percent = (p.pagecount - p.remaining) * 100 / p.pagecount;
+			info!("Sqlite snapshot progress: {}%", percent);
+		}
+		std::fs::create_dir_all(to)?;
+		let mut path = to.clone();
+		path.push("db.sqlite");
+		self.db
+			.get()?
+			.backup(rusqlite::DatabaseName::Main, path, Some(progress))?;
 		Ok(())
 	}
 
@@ -172,7 +160,7 @@ impl IDb for SqliteDb {
 		self.internal_get(&self.db.get()?, &tree, key)
 	}
 
-	fn approximate_len(&self, tree: usize) -> Result<usize> {
+	fn len(&self, tree: usize) -> Result<usize> {
 		let tree = self.get_tree(tree)?;
 		let db = self.db.get()?;
 
@@ -184,10 +172,6 @@ impl IDb for SqliteDb {
 		}
 	}
 
-	fn is_empty(&self, tree: usize) -> Result<bool> {
-		Ok(self.approximate_len(tree)? == 0)
-	}
-
 	fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result<()> {
 		let tree = self.get_tree(tree)?;
 		let db = self.db.get()?;

src/db/test.rs

@@ -1,7 +1,7 @@
 use crate::*;
 
 fn test_suite(db: Db) {
-	let tree = db.open_tree("tree:this_is_a_tree").unwrap();
+	let tree = db.open_tree("tree").unwrap();
 
 	let ka: &[u8] = &b"test"[..];
 	let kb: &[u8] = &b"zwello"[..];
@@ -14,7 +14,7 @@ fn test_suite(db: Db) {
 
 	assert!(tree.insert(ka, va).is_ok());
 	assert_eq!(tree.get(ka).unwrap().unwrap(), va);
-	assert_eq!(tree.iter().unwrap().count(), 1);
+	assert_eq!(tree.len().unwrap(), 1);
 
 	// ---- test transaction logic ----
 
@@ -148,15 +148,3 @@ fn test_sqlite_db() {
 	let db = SqliteDb::new(manager, false).unwrap();
 	test_suite(db);
 }
-
-#[test]
-#[cfg(feature = "fjall")]
-fn test_fjall_db() {
-	use crate::fjall_adapter::{fjall, FjallDb};
-
-	let path = mktemp::Temp::new_dir().unwrap();
-	let config = fjall::Config::new(path).temporary(true);
-	let keyspace = config.open_transactional().unwrap();
-	let db = FjallDb::init(keyspace);
-	test_suite(db);
-}

src/garage/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "garage"
-version = "1.3.1"
+version = "2.0.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -26,6 +26,7 @@ garage_db.workspace = true
 garage_api_admin.workspace = true
 garage_api_s3.workspace = true
 garage_api_k2v = { workspace = true, optional = true }
+garage_api_common.workspace = true
 garage_block.workspace = true
 garage_model.workspace = true
 garage_net.workspace = true
@@ -37,6 +38,7 @@ garage_web.workspace = true
 backtrace.workspace = true
 bytes.workspace = true
 bytesize.workspace = true
+chrono.workspace = true
 timeago.workspace = true
 parse_duration.workspace = true
 hex.workspace = true
@@ -47,8 +49,8 @@ sha1.workspace = true
 sodiumoxide.workspace = true
 structopt.workspace = true
 git-version.workspace = true
-
-serde.workspace = true
+utoipa.workspace = true
+serde_json.workspace = true
 
 futures.workspace = true
 tokio.workspace = true
@@ -57,7 +59,6 @@ opentelemetry.workspace = true
 opentelemetry-prometheus = { workspace = true, optional = true }
 opentelemetry-otlp = { workspace = true, optional = true }
 syslog-tracing = { workspace = true, optional = true }
-tracing-journald = { workspace = true, optional = true }
 
 [dev-dependencies]
 garage_api_common.workspace = true
@@ -86,12 +87,11 @@ k2v-client.workspace = true
 [features]
 default = [ "bundled-libs", "metrics", "lmdb", "sqlite", "k2v" ]
 
-k2v = [ "garage_util/k2v", "garage_api_k2v" ]
+k2v = [ "garage_util/k2v", "garage_api_k2v", "garage_api_admin/k2v" ]
 
 # Database engines
 lmdb = [ "garage_model/lmdb" ]
 sqlite = [ "garage_model/sqlite" ]
-fjall = [ "garage_model/fjall" ]
 
 # Automatic registration and discovery via Consul API
 consul-discovery = [ "garage_rpc/consul-discovery" ]
@@ -103,8 +103,6 @@ metrics = [ "garage_api_admin/metrics", "opentelemetry-prometheus" ]
 telemetry-otlp = [ "opentelemetry-otlp" ]
 # Logging to syslog
 syslog = [ "syslog-tracing" ]
-# Logging to journald
-journald = [ "tracing-journald" ]
 
 # NOTE: bundled-libs and system-libs should be treat as mutually exclusive;
 # exactly one of them should be enabled.

src/garage/admin/block.rs

@@ -1,243 +0,0 @@
-use garage_util::data::*;
-
-use garage_table::*;
-
-use garage_model::helper::error::{Error, OkOrBadRequest};
-use garage_model::s3::object_table::*;
-use garage_model::s3::version_table::*;
-
-use crate::cli::*;
-
-use super::*;
-
-impl AdminRpcHandler {
-	pub(super) async fn handle_block_cmd(&self, cmd: &BlockOperation) -> Result<AdminRpc, Error> {
-		match cmd {
-			BlockOperation::ListErrors => Ok(AdminRpc::BlockErrorList(
-				self.garage.block_manager.list_resync_errors()?,
-			)),
-			BlockOperation::Info { hash } => self.handle_block_info(hash).await,
-			BlockOperation::RetryNow { all, blocks } => {
-				self.handle_block_retry_now(*all, blocks).await
-			}
-			BlockOperation::Purge { yes, blocks } => self.handle_block_purge(*yes, blocks).await,
-		}
-	}
-
-	async fn handle_block_info(&self, hash: &String) -> Result<AdminRpc, Error> {
-		let hash = self.find_block_hash_by_prefix(hash)?;
-		let refcount = self.garage.block_manager.get_block_rc(&hash)?;
-		let block_refs = self
-			.garage
-			.block_ref_table
-			.get_range(&hash, None, None, 10000, Default::default())
-			.await?;
-		let mut versions = vec![];
-		let mut uploads = vec![];
-		for br in block_refs {
-			if let Some(v) = self
-				.garage
-				.version_table
-				.get(&br.version, &EmptyKey)
-				.await?
-			{
-				if let VersionBacklink::MultipartUpload { upload_id } = &v.backlink {
-					if let Some(u) = self.garage.mpu_table.get(upload_id, &EmptyKey).await? {
-						uploads.push(u);
-					}
-				}
-				versions.push(Ok(v));
-			} else {
-				versions.push(Err(br.version));
-			}
-		}
-		Ok(AdminRpc::BlockInfo {
-			hash,
-			refcount,
-			versions,
-			uploads,
-		})
-	}
-
-	async fn handle_block_retry_now(
-		&self,
-		all: bool,
-		blocks: &[String],
-	) -> Result<AdminRpc, Error> {
-		if all {
-			if !blocks.is_empty() {
-				return Err(Error::BadRequest(
-					"--all was specified, cannot also specify blocks".into(),
-				));
-			}
-			let blocks = self.garage.block_manager.list_resync_errors()?;
-			for b in blocks.iter() {
-				self.garage.block_manager.resync.clear_backoff(&b.hash)?;
-			}
-			Ok(AdminRpc::Ok(format!(
-				"{} blocks returned in queue for a retry now (check logs to see results)",
-				blocks.len()
-			)))
-		} else {
-			for hash in blocks {
-				let hash = hex::decode(hash).ok_or_bad_request("invalid hash")?;
-				let hash = Hash::try_from(&hash).ok_or_bad_request("invalid hash")?;
-				self.garage.block_manager.resync.clear_backoff(&hash)?;
-			}
-			Ok(AdminRpc::Ok(format!(
-				"{} blocks returned in queue for a retry now (check logs to see results)",
-				blocks.len()
-			)))
-		}
-	}
-
-	async fn handle_block_purge(&self, yes: bool, blocks: &[String]) -> Result<AdminRpc, Error> {
-		if !yes {
-			return Err(Error::BadRequest(
-				"Pass the --yes flag to confirm block purge operation.".into(),
-			));
-		}
-
-		let mut obj_dels = 0;
-		let mut mpu_dels = 0;
-		let mut ver_dels = 0;
-		let mut br_dels = 0;
-
-		for hash in blocks {
-			let hash = hex::decode(hash).ok_or_bad_request("invalid hash")?;
-			let hash = Hash::try_from(&hash).ok_or_bad_request("invalid hash")?;
-			let block_refs = self
-				.garage
-				.block_ref_table
-				.get_range(&hash, None, None, 10000, Default::default())
-				.await?;
-
-			for br in block_refs {
-				if let Some(version) = self
-					.garage
-					.version_table
-					.get(&br.version, &EmptyKey)
-					.await?
-				{
-					self.handle_block_purge_version_backlink(
-						&version,
-						&mut obj_dels,
-						&mut mpu_dels,
-					)
-					.await?;
-
-					if !version.deleted.get() {
-						let deleted_version = Version::new(version.uuid, version.backlink, true);
-						self.garage.version_table.insert(&deleted_version).await?;
-						ver_dels += 1;
-					}
-				}
-				if !br.deleted.get() {
-					let mut br = br;
-					br.deleted.set();
-					self.garage.block_ref_table.insert(&br).await?;
-					br_dels += 1;
-				}
-			}
-		}
-
-		Ok(AdminRpc::Ok(format!(
-			"Purged {} blocks: marked {} block refs, {} versions, {} objects and {} multipart uploads as deleted",
-			blocks.len(),
-			br_dels,
-			ver_dels,
-			obj_dels,
-			mpu_dels,
-		)))
-	}
-
-	async fn handle_block_purge_version_backlink(
-		&self,
-		version: &Version,
-		obj_dels: &mut usize,
-		mpu_dels: &mut usize,
-	) -> Result<(), Error> {
-		let (bucket_id, key, ov_id) = match &version.backlink {
-			VersionBacklink::Object { bucket_id, key } => (*bucket_id, key.clone(), version.uuid),
-			VersionBacklink::MultipartUpload { upload_id } => {
-				if let Some(mut mpu) = self.garage.mpu_table.get(upload_id, &EmptyKey).await? {
-					if !mpu.deleted.get() {
-						mpu.parts.clear();
-						mpu.deleted.set();
-						self.garage.mpu_table.insert(&mpu).await?;
-						*mpu_dels += 1;
-					}
-					(mpu.bucket_id, mpu.key.clone(), *upload_id)
-				} else {
-					return Ok(());
-				}
-			}
-		};
-
-		if let Some(object) = self.garage.object_table.get(&bucket_id, &key).await? {
-			let ov = object.versions().iter().rev().find(|v| v.is_complete());
-			if let Some(ov) = ov {
-				if ov.uuid == ov_id {
-					let del_uuid = gen_uuid();
-					let deleted_object = Object::new(
-						bucket_id,
-						key,
-						vec![ObjectVersion {
-							uuid: del_uuid,
-							timestamp: ov.timestamp + 1,
-							state: ObjectVersionState::Complete(ObjectVersionData::DeleteMarker),
-						}],
-					);
-					self.garage.object_table.insert(&deleted_object).await?;
-					*obj_dels += 1;
-				}
-			}
-		}
-
-		Ok(())
-	}
-
-	// ---- helper function ----
-	fn find_block_hash_by_prefix(&self, prefix: &str) -> Result<Hash, Error> {
-		if prefix.len() < 4 {
-			return Err(Error::BadRequest(
-				"Please specify at least 4 characters of the block hash".into(),
-			));
-		}
-
-		let prefix_bin =
-			hex::decode(&prefix[..prefix.len() & !1]).ok_or_bad_request("invalid hash")?;
-
-		let iter = self
-			.garage
-			.block_ref_table
-			.data
-			.store
-			.range(&prefix_bin[..]..)
-			.map_err(GarageError::from)?;
-		let mut found = None;
-		for item in iter {
-			let (k, _v) = item.map_err(GarageError::from)?;
-			let hash = Hash::try_from(&k[..32]).unwrap();
-			if &hash.as_slice()[..prefix_bin.len()] != prefix_bin {
-				break;
-			}
-			if hex::encode(hash.as_slice()).starts_with(prefix) {
-				match &found {
-					Some(x) if *x == hash => (),
-					Some(_) => {
-						return Err(Error::BadRequest(format!(
-							"Several blocks match prefix `{}`",
-							prefix
-						)));
-					}
-					None => {
-						found = Some(hash);
-					}
-				}
-			}
-		}
-
-		found.ok_or_else(|| Error::BadRequest("No matching block found".into()))
-	}
-}

src/garage/admin/bucket.rs

@@ -1,500 +0,0 @@
-use std::collections::HashMap;
-use std::fmt::Write;
-
-use garage_util::crdt::*;
-use garage_util::time::*;
-
-use garage_table::*;
-
-use garage_model::bucket_alias_table::*;
-use garage_model::bucket_table::*;
-use garage_model::helper::error::{Error, OkOrBadRequest};
-use garage_model::permission::*;
-
-use crate::cli::*;
-
-use super::*;
-
-impl AdminRpcHandler {
-	pub(super) async fn handle_bucket_cmd(&self, cmd: &BucketOperation) -> Result<AdminRpc, Error> {
-		match cmd {
-			BucketOperation::List => self.handle_list_buckets().await,
-			BucketOperation::Info(query) => self.handle_bucket_info(query).await,
-			BucketOperation::Create(query) => self.handle_create_bucket(&query.name).await,
-			BucketOperation::Delete(query) => self.handle_delete_bucket(query).await,
-			BucketOperation::Alias(query) => self.handle_alias_bucket(query).await,
-			BucketOperation::Unalias(query) => self.handle_unalias_bucket(query).await,
-			BucketOperation::Allow(query) => self.handle_bucket_allow(query).await,
-			BucketOperation::Deny(query) => self.handle_bucket_deny(query).await,
-			BucketOperation::Website(query) => self.handle_bucket_website(query).await,
-			BucketOperation::SetQuotas(query) => self.handle_bucket_set_quotas(query).await,
-			BucketOperation::CleanupIncompleteUploads(query) => {
-				self.handle_bucket_cleanup_incomplete_uploads(query).await
-			}
-		}
-	}
-
-	async fn handle_list_buckets(&self) -> Result<AdminRpc, Error> {
-		let buckets = self
-			.garage
-			.bucket_table
-			.get_range(
-				&EmptyKey,
-				None,
-				Some(DeletedFilter::NotDeleted),
-				10000,
-				EnumerationOrder::Forward,
-			)
-			.await?;
-
-		Ok(AdminRpc::BucketList(buckets))
-	}
-
-	async fn handle_bucket_info(&self, query: &BucketOpt) -> Result<AdminRpc, Error> {
-		let bucket_id = self
-			.garage
-			.bucket_helper()
-			.admin_get_existing_matching_bucket(&query.name)
-			.await?;
-
-		let bucket = self
-			.garage
-			.bucket_helper()
-			.get_existing_bucket(bucket_id)
-			.await?;
-
-		let counters = self
-			.garage
-			.object_counter_table
-			.table
-			.get(&bucket_id, &EmptyKey)
-			.await?
-			.map(|x| x.filtered_values(&self.garage.system.cluster_layout()))
-			.unwrap_or_default();
-
-		let mpu_counters = self
-			.garage
-			.mpu_counter_table
-			.table
-			.get(&bucket_id, &EmptyKey)
-			.await?
-			.map(|x| x.filtered_values(&self.garage.system.cluster_layout()))
-			.unwrap_or_default();
-
-		let mut relevant_keys = HashMap::new();
-		for (k, _) in bucket
-			.state
-			.as_option()
-			.unwrap()
-			.authorized_keys
-			.items()
-			.iter()
-		{
-			if let Some(key) = self
-				.garage
-				.key_table
-				.get(&EmptyKey, k)
-				.await?
-				.filter(|k| !k.is_deleted())
-			{
-				relevant_keys.insert(k.clone(), key);
-			}
-		}
-		for ((k, _), _, _) in bucket
-			.state
-			.as_option()
-			.unwrap()
-			.local_aliases
-			.items()
-			.iter()
-		{
-			if relevant_keys.contains_key(k) {
-				continue;
-			}
-			if let Some(key) = self.garage.key_table.get(&EmptyKey, k).await? {
-				relevant_keys.insert(k.clone(), key);
-			}
-		}
-
-		Ok(AdminRpc::BucketInfo {
-			bucket,
-			relevant_keys,
-			counters,
-			mpu_counters,
-		})
-	}
-
-	#[allow(clippy::ptr_arg)]
-	async fn handle_create_bucket(&self, name: &String) -> Result<AdminRpc, Error> {
-		if !is_valid_bucket_name(name, self.garage.config.allow_punycode) {
-			return Err(Error::BadRequest(format!(
-				"{}: {}",
-				name, INVALID_BUCKET_NAME_MESSAGE
-			)));
-		}
-
-		let helper = self.garage.locked_helper().await;
-
-		if let Some(alias) = self.garage.bucket_alias_table.get(&EmptyKey, name).await? {
-			if alias.state.get().is_some() {
-				return Err(Error::BadRequest(format!("Bucket {} already exists", name)));
-			}
-		}
-
-		// ---- done checking, now commit ----
-
-		let bucket = Bucket::new();
-		self.garage.bucket_table.insert(&bucket).await?;
-
-		helper.set_global_bucket_alias(bucket.id, name).await?;
-
-		Ok(AdminRpc::Ok(format!("Bucket {} was created.", name)))
-	}
-
-	async fn handle_delete_bucket(&self, query: &DeleteBucketOpt) -> Result<AdminRpc, Error> {
-		let helper = self.garage.locked_helper().await;
-
-		let bucket_id = helper
-			.bucket()
-			.admin_get_existing_matching_bucket(&query.name)
-			.await?;
-
-		// Get the alias, but keep in minde here the bucket name
-		// given in parameter can also be directly the bucket's ID.
-		// In that case bucket_alias will be None, and
-		// we can still delete the bucket if it has zero aliases
-		// (a condition which we try to prevent but that could still happen somehow).
-		// We just won't try to delete an alias entry because there isn't one.
-		let bucket_alias = self
-			.garage
-			.bucket_alias_table
-			.get(&EmptyKey, &query.name)
-			.await?;
-
-		// Check bucket doesn't have other aliases
-		let mut bucket = helper.bucket().get_existing_bucket(bucket_id).await?;
-		let bucket_state = bucket.state.as_option().unwrap();
-		if bucket_state
-			.aliases
-			.items()
-			.iter()
-			.filter(|(_, _, active)| *active)
-			.any(|(name, _, _)| name != &query.name)
-		{
-			return Err(Error::BadRequest(format!("Bucket {} still has other global aliases. Use `bucket unalias` to delete them one by one.", query.name)));
-		}
-		if bucket_state
-			.local_aliases
-			.items()
-			.iter()
-			.any(|(_, _, active)| *active)
-		{
-			return Err(Error::BadRequest(format!("Bucket {} still has other local aliases. Use `bucket unalias` to delete them one by one.", query.name)));
-		}
-
-		// Check bucket is empty
-		if !helper.bucket().is_bucket_empty(bucket_id).await? {
-			return Err(Error::BadRequest(format!(
-				"Bucket {} is not empty",
-				query.name
-			)));
-		}
-
-		if !query.yes {
-			return Err(Error::BadRequest(
-				"Add --yes flag to really perform this operation".to_string(),
-			));
-		}
-
-		// --- done checking, now commit ---
-		// 1. delete authorization from keys that had access
-		for (key_id, _) in bucket.authorized_keys() {
-			helper
-				.set_bucket_key_permissions(bucket.id, key_id, BucketKeyPerm::NO_PERMISSIONS)
-				.await?;
-		}
-
-		// 2. delete bucket alias
-		if bucket_alias.is_some() {
-			helper
-				.purge_global_bucket_alias(bucket_id, &query.name)
-				.await?;
-		}
-
-		// 3. delete bucket
-		bucket.state = Deletable::delete();
-		self.garage.bucket_table.insert(&bucket).await?;
-
-		Ok(AdminRpc::Ok(format!("Bucket {} was deleted.", query.name)))
-	}
-
-	async fn handle_alias_bucket(&self, query: &AliasBucketOpt) -> Result<AdminRpc, Error> {
-		let helper = self.garage.locked_helper().await;
-
-		let bucket_id = helper
-			.bucket()
-			.admin_get_existing_matching_bucket(&query.existing_bucket)
-			.await?;
-
-		if let Some(key_pattern) = &query.local {
-			let key = helper.key().get_existing_matching_key(key_pattern).await?;
-
-			helper
-				.set_local_bucket_alias(bucket_id, &key.key_id, &query.new_name)
-				.await?;
-			Ok(AdminRpc::Ok(format!(
-				"Alias {} now points to bucket {:?} in namespace of key {}",
-				query.new_name, bucket_id, key.key_id
-			)))
-		} else {
-			helper
-				.set_global_bucket_alias(bucket_id, &query.new_name)
-				.await?;
-			Ok(AdminRpc::Ok(format!(
-				"Alias {} now points to bucket {:?}",
-				query.new_name, bucket_id
-			)))
-		}
-	}
-
-	async fn handle_unalias_bucket(&self, query: &UnaliasBucketOpt) -> Result<AdminRpc, Error> {
-		let helper = self.garage.locked_helper().await;
-
-		if let Some(key_pattern) = &query.local {
-			let key = helper.key().get_existing_matching_key(key_pattern).await?;
-
-			let bucket_id = key
-				.state
-				.as_option()
-				.unwrap()
-				.local_aliases
-				.get(&query.name)
-				.cloned()
-				.flatten()
-				.ok_or_bad_request("Bucket not found")?;
-
-			helper
-				.unset_local_bucket_alias(bucket_id, &key.key_id, &query.name)
-				.await?;
-
-			Ok(AdminRpc::Ok(format!(
-				"Alias {} no longer points to bucket {:?} in namespace of key {}",
-				&query.name, bucket_id, key.key_id
-			)))
-		} else {
-			let bucket_id = helper
-				.bucket()
-				.resolve_global_bucket_name(&query.name)
-				.await?
-				.ok_or_bad_request("Bucket not found")?;
-
-			helper
-				.unset_global_bucket_alias(bucket_id, &query.name)
-				.await?;
-
-			Ok(AdminRpc::Ok(format!(
-				"Alias {} no longer points to bucket {:?}",
-				&query.name, bucket_id
-			)))
-		}
-	}
-
-	async fn handle_bucket_allow(&self, query: &PermBucketOpt) -> Result<AdminRpc, Error> {
-		let helper = self.garage.locked_helper().await;
-
-		let bucket_id = helper
-			.bucket()
-			.admin_get_existing_matching_bucket(&query.bucket)
-			.await?;
-		let key = helper
-			.key()
-			.get_existing_matching_key(&query.key_pattern)
-			.await?;
-
-		let allow_read = query.read || key.allow_read(&bucket_id);
-		let allow_write = query.write || key.allow_write(&bucket_id);
-		let allow_owner = query.owner || key.allow_owner(&bucket_id);
-
-		helper
-			.set_bucket_key_permissions(
-				bucket_id,
-				&key.key_id,
-				BucketKeyPerm {
-					timestamp: now_msec(),
-					allow_read,
-					allow_write,
-					allow_owner,
-				},
-			)
-			.await?;
-
-		Ok(AdminRpc::Ok(format!(
-			"New permissions for {} on {}: read {}, write {}, owner {}.",
-			&key.key_id, &query.bucket, allow_read, allow_write, allow_owner
-		)))
-	}
-
-	async fn handle_bucket_deny(&self, query: &PermBucketOpt) -> Result<AdminRpc, Error> {
-		let helper = self.garage.locked_helper().await;
-
-		let bucket_id = helper
-			.bucket()
-			.admin_get_existing_matching_bucket(&query.bucket)
-			.await?;
-		let key = helper
-			.key()
-			.get_existing_matching_key(&query.key_pattern)
-			.await?;
-
-		let allow_read = !query.read && key.allow_read(&bucket_id);
-		let allow_write = !query.write && key.allow_write(&bucket_id);
-		let allow_owner = !query.owner && key.allow_owner(&bucket_id);
-
-		helper
-			.set_bucket_key_permissions(
-				bucket_id,
-				&key.key_id,
-				BucketKeyPerm {
-					timestamp: now_msec(),
-					allow_read,
-					allow_write,
-					allow_owner,
-				},
-			)
-			.await?;
-
-		Ok(AdminRpc::Ok(format!(
-			"New permissions for {} on {}: read {}, write {}, owner {}.",
-			&key.key_id, &query.bucket, allow_read, allow_write, allow_owner
-		)))
-	}
-
-	async fn handle_bucket_website(&self, query: &WebsiteOpt) -> Result<AdminRpc, Error> {
-		let bucket_id = self
-			.garage
-			.bucket_helper()
-			.admin_get_existing_matching_bucket(&query.bucket)
-			.await?;
-
-		let mut bucket = self
-			.garage
-			.bucket_helper()
-			.get_existing_bucket(bucket_id)
-			.await?;
-		let bucket_state = bucket.state.as_option_mut().unwrap();
-
-		if !(query.allow ^ query.deny) {
-			return Err(Error::BadRequest(
-				"You must specify exactly one flag, either --allow or --deny".to_string(),
-			));
-		}
-
-		let website = if query.allow {
-			Some(WebsiteConfig {
-				index_document: query.index_document.clone(),
-				error_document: query.error_document.clone(),
-			})
-		} else {
-			None
-		};
-
-		bucket_state.website_config.update(website);
-		self.garage.bucket_table.insert(&bucket).await?;
-
-		let msg = if query.allow {
-			format!("Website access allowed for {}", &query.bucket)
-		} else {
-			format!("Website access denied for {}", &query.bucket)
-		};
-
-		Ok(AdminRpc::Ok(msg))
-	}
-
-	async fn handle_bucket_set_quotas(&self, query: &SetQuotasOpt) -> Result<AdminRpc, Error> {
-		let bucket_id = self
-			.garage
-			.bucket_helper()
-			.admin_get_existing_matching_bucket(&query.bucket)
-			.await?;
-
-		let mut bucket = self
-			.garage
-			.bucket_helper()
-			.get_existing_bucket(bucket_id)
-			.await?;
-		let bucket_state = bucket.state.as_option_mut().unwrap();
-
-		if query.max_size.is_none() && query.max_objects.is_none() {
-			return Err(Error::BadRequest(
-				"You must specify either --max-size or --max-objects (or both) for this command to do something.".to_string(),
-			));
-		}
-
-		let mut quotas = bucket_state.quotas.get().clone();
-
-		match query.max_size.as_ref().map(String::as_ref) {
-			Some("none") => quotas.max_size = None,
-			Some(v) => {
-				let bs = v
-					.parse::<bytesize::ByteSize>()
-					.ok_or_bad_request(format!("Invalid size specified: {}", v))?;
-				quotas.max_size = Some(bs.as_u64());
-			}
-			_ => (),
-		}
-
-		match query.max_objects.as_ref().map(String::as_ref) {
-			Some("none") => quotas.max_objects = None,
-			Some(v) => {
-				let mo = v
-					.parse::<u64>()
-					.ok_or_bad_request(format!("Invalid number specified: {}", v))?;
-				quotas.max_objects = Some(mo);
-			}
-			_ => (),
-		}
-
-		bucket_state.quotas.update(quotas);
-		self.garage.bucket_table.insert(&bucket).await?;
-
-		Ok(AdminRpc::Ok(format!(
-			"Quotas updated for {}",
-			&query.bucket
-		)))
-	}
-
-	async fn handle_bucket_cleanup_incomplete_uploads(
-		&self,
-		query: &CleanupIncompleteUploadsOpt,
-	) -> Result<AdminRpc, Error> {
-		let mut bucket_ids = vec![];
-		for b in query.buckets.iter() {
-			bucket_ids.push(
-				self.garage
-					.bucket_helper()
-					.admin_get_existing_matching_bucket(b)
-					.await?,
-			);
-		}
-
-		let duration = parse_duration::parse::parse(&query.older_than)
-			.ok_or_bad_request("Invalid duration passed for --older-than parameter")?;
-
-		let mut ret = String::new();
-		for bucket in bucket_ids {
-			let count = self
-				.garage
-				.bucket_helper()
-				.cleanup_incomplete_uploads(&bucket, duration)
-				.await?;
-			writeln!(
-				&mut ret,
-				"Bucket {:?}: {} incomplete uploads aborted",
-				bucket, count
-			)
-			.unwrap();
-		}
-
-		Ok(AdminRpc::Ok(ret))
-	}
-}

src/garage/admin/key.rs

@@ -1,161 +0,0 @@
-use std::collections::HashMap;
-
-use garage_table::*;
-
-use garage_model::helper::error::*;
-use garage_model::key_table::*;
-
-use crate::cli::*;
-
-use super::*;
-
-impl AdminRpcHandler {
-	pub(super) async fn handle_key_cmd(&self, cmd: &KeyOperation) -> Result<AdminRpc, Error> {
-		match cmd {
-			KeyOperation::List => self.handle_list_keys().await,
-			KeyOperation::Info(query) => self.handle_key_info(query).await,
-			KeyOperation::Create(query) => self.handle_create_key(query).await,
-			KeyOperation::Rename(query) => self.handle_rename_key(query).await,
-			KeyOperation::Delete(query) => self.handle_delete_key(query).await,
-			KeyOperation::Allow(query) => self.handle_allow_key(query).await,
-			KeyOperation::Deny(query) => self.handle_deny_key(query).await,
-			KeyOperation::Import(query) => self.handle_import_key(query).await,
-		}
-	}
-
-	async fn handle_list_keys(&self) -> Result<AdminRpc, Error> {
-		let key_ids = self
-			.garage
-			.key_table
-			.get_range(
-				&EmptyKey,
-				None,
-				Some(KeyFilter::Deleted(DeletedFilter::NotDeleted)),
-				10000,
-				EnumerationOrder::Forward,
-			)
-			.await?
-			.iter()
-			.map(|k| (k.key_id.to_string(), k.params().unwrap().name.get().clone()))
-			.collect::<Vec<_>>();
-		Ok(AdminRpc::KeyList(key_ids))
-	}
-
-	async fn handle_key_info(&self, query: &KeyInfoOpt) -> Result<AdminRpc, Error> {
-		let mut key = self
-			.garage
-			.key_helper()
-			.get_existing_matching_key(&query.key_pattern)
-			.await?;
-
-		if !query.show_secret {
-			key.state.as_option_mut().unwrap().secret_key = "(redacted)".into();
-		}
-
-		self.key_info_result(key).await
-	}
-
-	async fn handle_create_key(&self, query: &KeyNewOpt) -> Result<AdminRpc, Error> {
-		let key = Key::new(&query.name);
-		self.garage.key_table.insert(&key).await?;
-		self.key_info_result(key).await
-	}
-
-	async fn handle_rename_key(&self, query: &KeyRenameOpt) -> Result<AdminRpc, Error> {
-		let mut key = self
-			.garage
-			.key_helper()
-			.get_existing_matching_key(&query.key_pattern)
-			.await?;
-		key.params_mut()
-			.unwrap()
-			.name
-			.update(query.new_name.clone());
-		self.garage.key_table.insert(&key).await?;
-		self.key_info_result(key).await
-	}
-
-	async fn handle_delete_key(&self, query: &KeyDeleteOpt) -> Result<AdminRpc, Error> {
-		let helper = self.garage.locked_helper().await;
-
-		let mut key = helper
-			.key()
-			.get_existing_matching_key(&query.key_pattern)
-			.await?;
-
-		if !query.yes {
-			return Err(Error::BadRequest(
-				"Add --yes flag to really perform this operation".to_string(),
-			));
-		}
-
-		helper.delete_key(&mut key).await?;
-
-		Ok(AdminRpc::Ok(format!(
-			"Key {} was deleted successfully.",
-			key.key_id
-		)))
-	}
-
-	async fn handle_allow_key(&self, query: &KeyPermOpt) -> Result<AdminRpc, Error> {
-		let mut key = self
-			.garage
-			.key_helper()
-			.get_existing_matching_key(&query.key_pattern)
-			.await?;
-		if query.create_bucket {
-			key.params_mut().unwrap().allow_create_bucket.update(true);
-		}
-		self.garage.key_table.insert(&key).await?;
-		self.key_info_result(key).await
-	}
-
-	async fn handle_deny_key(&self, query: &KeyPermOpt) -> Result<AdminRpc, Error> {
-		let mut key = self
-			.garage
-			.key_helper()
-			.get_existing_matching_key(&query.key_pattern)
-			.await?;
-		if query.create_bucket {
-			key.params_mut().unwrap().allow_create_bucket.update(false);
-		}
-		self.garage.key_table.insert(&key).await?;
-		self.key_info_result(key).await
-	}
-
-	async fn handle_import_key(&self, query: &KeyImportOpt) -> Result<AdminRpc, Error> {
-		if !query.yes {
-			return Err(Error::BadRequest("This command is intended to re-import keys that were previously generated by Garage. If you want to create a new key, use `garage key new` instead. Add the --yes flag if you really want to re-import a key.".to_string()));
-		}
-
-		let prev_key = self.garage.key_table.get(&EmptyKey, &query.key_id).await?;
-		if prev_key.is_some() {
-			return Err(Error::BadRequest(format!("Key {} already exists in data store. Even if it is deleted, we can't let you create a new key with the same ID. Sorry.", query.key_id)));
-		}
-
-		let imported_key = Key::import(&query.key_id, &query.secret_key, &query.name)
-			.ok_or_bad_request("Invalid key format")?;
-		self.garage.key_table.insert(&imported_key).await?;
-
-		self.key_info_result(imported_key).await
-	}
-
-	async fn key_info_result(&self, key: Key) -> Result<AdminRpc, Error> {
-		let mut relevant_buckets = HashMap::new();
-
-		for (id, _) in key
-			.state
-			.as_option()
-			.unwrap()
-			.authorized_buckets
-			.items()
-			.iter()
-		{
-			if let Some(b) = self.garage.bucket_table.get(&EmptyKey, id).await? {
-				relevant_buckets.insert(*id, b);
-			}
-		}
-
-		Ok(AdminRpc::KeyInfo(key, relevant_buckets))
-	}
-}

src/garage/admin/mod.rs

@@ -1,545 +0,0 @@
-mod block;
-mod bucket;
-mod key;
-
-use std::collections::HashMap;
-use std::fmt::Write;
-use std::future::Future;
-use std::sync::Arc;
-
-use futures::future::FutureExt;
-
-use serde::{Deserialize, Serialize};
-
-use format_table::format_table_to_string;
-
-use garage_util::background::BackgroundRunner;
-use garage_util::data::*;
-use garage_util::error::Error as GarageError;
-
-use garage_table::replication::*;
-use garage_table::*;
-
-use garage_rpc::layout::PARTITION_BITS;
-use garage_rpc::*;
-
-use garage_block::manager::BlockResyncErrorInfo;
-
-use garage_model::bucket_table::*;
-use garage_model::garage::Garage;
-use garage_model::helper::error::{Error, OkOrBadRequest};
-use garage_model::key_table::*;
-use garage_model::s3::mpu_table::MultipartUpload;
-use garage_model::s3::version_table::Version;
-
-use crate::cli::*;
-use crate::repair::online::launch_online_repair;
-
-pub const ADMIN_RPC_PATH: &str = "garage/admin_rpc.rs/Rpc";
-
-#[derive(Debug, Serialize, Deserialize)]
-#[allow(clippy::large_enum_variant)]
-pub enum AdminRpc {
-	BucketOperation(BucketOperation),
-	KeyOperation(KeyOperation),
-	LaunchRepair(RepairOpt),
-	Stats(StatsOpt),
-	Worker(WorkerOperation),
-	BlockOperation(BlockOperation),
-	MetaOperation(MetaOperation),
-
-	// Replies
-	Ok(String),
-	BucketList(Vec<Bucket>),
-	BucketInfo {
-		bucket: Bucket,
-		relevant_keys: HashMap<String, Key>,
-		counters: HashMap<String, i64>,
-		mpu_counters: HashMap<String, i64>,
-	},
-	KeyList(Vec<(String, String)>),
-	KeyInfo(Key, HashMap<Uuid, Bucket>),
-	WorkerList(
-		HashMap<usize, garage_util::background::WorkerInfo>,
-		WorkerListOpt,
-	),
-	WorkerVars(Vec<(Uuid, String, String)>),
-	WorkerInfo(usize, garage_util::background::WorkerInfo),
-	BlockErrorList(Vec<BlockResyncErrorInfo>),
-	BlockInfo {
-		hash: Hash,
-		refcount: u64,
-		versions: Vec<Result<Version, Uuid>>,
-		uploads: Vec<MultipartUpload>,
-	},
-}
-
-impl Rpc for AdminRpc {
-	type Response = Result<AdminRpc, Error>;
-}
-
-pub struct AdminRpcHandler {
-	garage: Arc<Garage>,
-	background: Arc<BackgroundRunner>,
-	endpoint: Arc<Endpoint<AdminRpc, Self>>,
-}
-
-impl AdminRpcHandler {
-	pub fn new(garage: Arc<Garage>, background: Arc<BackgroundRunner>) -> Arc<Self> {
-		let endpoint = garage.system.netapp.endpoint(ADMIN_RPC_PATH.into());
-		let admin = Arc::new(Self {
-			garage,
-			background,
-			endpoint,
-		});
-		admin.endpoint.set_handler(admin.clone());
-		admin
-	}
-
-	// ================ REPAIR COMMANDS ====================
-
-	async fn handle_launch_repair(self: &Arc<Self>, opt: RepairOpt) -> Result<AdminRpc, Error> {
-		if !opt.yes {
-			return Err(Error::BadRequest(
-				"Please provide the --yes flag to initiate repair operations.".to_string(),
-			));
-		}
-		if opt.all_nodes {
-			let mut opt_to_send = opt.clone();
-			opt_to_send.all_nodes = false;
-
-			let mut failures = vec![];
-			let all_nodes = self.garage.system.cluster_layout().all_nodes().to_vec();
-			for node in all_nodes.iter() {
-				let node = (*node).into();
-				let resp = self
-					.endpoint
-					.call(
-						&node,
-						AdminRpc::LaunchRepair(opt_to_send.clone()),
-						PRIO_NORMAL,
-					)
-					.await;
-				if !matches!(resp, Ok(Ok(_))) {
-					failures.push(node);
-				}
-			}
-			if failures.is_empty() {
-				Ok(AdminRpc::Ok("Repair launched on all nodes".to_string()))
-			} else {
-				Err(Error::BadRequest(format!(
-					"Could not launch repair on nodes: {:?} (launched successfully on other nodes)",
-					failures
-				)))
-			}
-		} else {
-			launch_online_repair(&self.garage, &self.background, opt).await?;
-			Ok(AdminRpc::Ok(format!(
-				"Repair launched on {:?}",
-				self.garage.system.id
-			)))
-		}
-	}
-
-	// ================ STATS COMMANDS ====================
-
-	async fn handle_stats(&self, opt: StatsOpt) -> Result<AdminRpc, Error> {
-		if opt.all_nodes {
-			let mut ret = String::new();
-			let mut all_nodes = self.garage.system.cluster_layout().all_nodes().to_vec();
-			for node in self.garage.system.get_known_nodes().iter() {
-				if node.is_up && !all_nodes.contains(&node.id) {
-					all_nodes.push(node.id);
-				}
-			}
-
-			for node in all_nodes.iter() {
-				let mut opt = opt.clone();
-				opt.all_nodes = false;
-				opt.skip_global = true;
-
-				writeln!(&mut ret, "\n======================").unwrap();
-				writeln!(&mut ret, "Stats for node {:?}:", node).unwrap();
-
-				let node_id = (*node).into();
-				match self
-					.endpoint
-					.call(&node_id, AdminRpc::Stats(opt), PRIO_NORMAL)
-					.await
-				{
-					Ok(Ok(AdminRpc::Ok(s))) => writeln!(&mut ret, "{}", s).unwrap(),
-					Ok(Ok(x)) => writeln!(&mut ret, "Bad answer: {:?}", x).unwrap(),
-					Ok(Err(e)) => writeln!(&mut ret, "Remote error: {}", e).unwrap(),
-					Err(e) => writeln!(&mut ret, "Network error: {}", e).unwrap(),
-				}
-			}
-
-			writeln!(&mut ret, "\n======================").unwrap();
-			write!(
-				&mut ret,
-				"Cluster statistics:\n\n{}",
-				self.gather_cluster_stats()
-			)
-			.unwrap();
-
-			Ok(AdminRpc::Ok(ret))
-		} else {
-			Ok(AdminRpc::Ok(self.gather_stats_local(opt)?))
-		}
-	}
-
-	fn gather_stats_local(&self, opt: StatsOpt) -> Result<String, Error> {
-		let mut ret = String::new();
-		writeln!(
-			&mut ret,
-			"\nGarage version: {} [features: {}]\nRust compiler version: {}",
-			garage_util::version::garage_version(),
-			garage_util::version::garage_features()
-				.map(|list| list.join(", "))
-				.unwrap_or_else(|| "(unknown)".into()),
-			garage_util::version::rust_version(),
-		)
-		.unwrap();
-
-		writeln!(&mut ret, "\nDatabase engine: {}", self.garage.db.engine()).unwrap();
-
-		// Gather table statistics
-		let mut table = vec!["  Table\tItems\tMklItems\tMklTodo\tGcTodo".into()];
-		table.push(self.gather_table_stats(&self.garage.bucket_table)?);
-		table.push(self.gather_table_stats(&self.garage.key_table)?);
-		table.push(self.gather_table_stats(&self.garage.object_table)?);
-		table.push(self.gather_table_stats(&self.garage.version_table)?);
-		table.push(self.gather_table_stats(&self.garage.block_ref_table)?);
-		write!(
-			&mut ret,
-			"\nTable stats:\n{}",
-			format_table_to_string(table)
-		)
-		.unwrap();
-
-		// Gather block manager statistics
-		writeln!(&mut ret, "\nBlock manager stats:").unwrap();
-		let rc_len = self.garage.block_manager.rc_approximate_len()?.to_string();
-
-		writeln!(
-			&mut ret,
-			"  number of RC entries (~= number of blocks): {}",
-			rc_len
-		)
-		.unwrap();
-		writeln!(
-			&mut ret,
-			"  resync queue length: {}",
-			self.garage.block_manager.resync.queue_approximate_len()?
-		)
-		.unwrap();
-		writeln!(
-			&mut ret,
-			"  blocks with resync errors: {}",
-			self.garage.block_manager.resync.errors_approximate_len()?
-		)
-		.unwrap();
-
-		if !opt.skip_global {
-			write!(&mut ret, "\n{}", self.gather_cluster_stats()).unwrap();
-		}
-
-		Ok(ret)
-	}
-
-	fn gather_cluster_stats(&self) -> String {
-		let mut ret = String::new();
-
-		// Gather storage node and free space statistics for current nodes
-		let layout = &self.garage.system.cluster_layout();
-		let mut node_partition_count = HashMap::<Uuid, u64>::new();
-		for short_id in layout.current().ring_assignment_data.iter() {
-			let id = layout.current().node_id_vec[*short_id as usize];
-			*node_partition_count.entry(id).or_default() += 1;
-		}
-		let node_info = self
-			.garage
-			.system
-			.get_known_nodes()
-			.into_iter()
-			.map(|n| (n.id, n))
-			.collect::<HashMap<_, _>>();
-
-		let mut table = vec!["  ID\tHostname\tZone\tCapacity\tPart.\tDataAvail\tMetaAvail".into()];
-		for (id, parts) in node_partition_count.iter() {
-			let info = node_info.get(id);
-			let status = info.map(|x| &x.status);
-			let role = layout.current().roles.get(id).and_then(|x| x.0.as_ref());
-			let hostname = status.and_then(|x| x.hostname.as_deref()).unwrap_or("?");
-			let zone = role.map(|x| x.zone.as_str()).unwrap_or("?");
-			let capacity = role
-				.map(|x| x.capacity_string())
-				.unwrap_or_else(|| "?".into());
-			let avail_str = |x| match x {
-				Some((avail, total)) => {
-					let pct = (avail as f64) / (total as f64) * 100.;
-					let avail = bytesize::ByteSize::b(avail);
-					let total = bytesize::ByteSize::b(total);
-					format!("{}/{} ({:.1}%)", avail, total, pct)
-				}
-				None => "?".into(),
-			};
-			let data_avail = avail_str(status.and_then(|x| x.data_disk_avail));
-			let meta_avail = avail_str(status.and_then(|x| x.meta_disk_avail));
-			table.push(format!(
-				"  {:?}\t{}\t{}\t{}\t{}\t{}\t{}",
-				id, hostname, zone, capacity, parts, data_avail, meta_avail
-			));
-		}
-		write!(
-			&mut ret,
-			"Storage nodes:\n{}",
-			format_table_to_string(table)
-		)
-		.unwrap();
-
-		let meta_part_avail = node_partition_count
-			.iter()
-			.filter_map(|(id, parts)| {
-				node_info
-					.get(id)
-					.and_then(|x| x.status.meta_disk_avail)
-					.map(|c| c.0 / *parts)
-			})
-			.collect::<Vec<_>>();
-		let data_part_avail = node_partition_count
-			.iter()
-			.filter_map(|(id, parts)| {
-				node_info
-					.get(id)
-					.and_then(|x| x.status.data_disk_avail)
-					.map(|c| c.0 / *parts)
-			})
-			.collect::<Vec<_>>();
-		if !meta_part_avail.is_empty() && !data_part_avail.is_empty() {
-			let meta_avail =
-				bytesize::ByteSize(meta_part_avail.iter().min().unwrap() * (1 << PARTITION_BITS));
-			let data_avail =
-				bytesize::ByteSize(data_part_avail.iter().min().unwrap() * (1 << PARTITION_BITS));
-			writeln!(
-				&mut ret,
-				"\nEstimated available storage space cluster-wide (might be lower in practice):"
-			)
-			.unwrap();
-			if meta_part_avail.len() < node_partition_count.len()
-				|| data_part_avail.len() < node_partition_count.len()
-			{
-				writeln!(&mut ret, "  data: < {}", data_avail).unwrap();
-				writeln!(&mut ret, "  metadata: < {}", meta_avail).unwrap();
-				writeln!(&mut ret, "A precise estimate could not be given as information is missing for some storage nodes.").unwrap();
-			} else {
-				writeln!(&mut ret, "  data: {}", data_avail).unwrap();
-				writeln!(&mut ret, "  metadata: {}", meta_avail).unwrap();
-			}
-		}
-
-		ret
-	}
-
-	fn gather_table_stats<F, R>(&self, t: &Arc<Table<F, R>>) -> Result<String, Error>
-	where
-		F: TableSchema + 'static,
-		R: TableReplication + 'static,
-	{
-		let data_len = t
-			.data
-			.store
-			.approximate_len()
-			.map_err(GarageError::from)?
-			.to_string();
-		let mkl_len = t.merkle_updater.merkle_tree_approximate_len()?.to_string();
-
-		Ok(format!(
-			"  {}\t{}\t{}\t{}\t{}",
-			F::TABLE_NAME,
-			data_len,
-			mkl_len,
-			t.merkle_updater.todo_approximate_len()?,
-			t.data.gc_todo_approximate_len()?
-		))
-	}
-
-	// ================ WORKER COMMANDS ====================
-
-	async fn handle_worker_cmd(&self, cmd: &WorkerOperation) -> Result<AdminRpc, Error> {
-		match cmd {
-			WorkerOperation::List { opt } => {
-				let workers = self.background.get_worker_info();
-				Ok(AdminRpc::WorkerList(workers, *opt))
-			}
-			WorkerOperation::Info { tid } => {
-				let info = self
-					.background
-					.get_worker_info()
-					.get(tid)
-					.ok_or_bad_request(format!("No worker with TID {}", tid))?
-					.clone();
-				Ok(AdminRpc::WorkerInfo(*tid, info))
-			}
-			WorkerOperation::Get {
-				all_nodes,
-				variable,
-			} => self.handle_get_var(*all_nodes, variable).await,
-			WorkerOperation::Set {
-				all_nodes,
-				variable,
-				value,
-			} => self.handle_set_var(*all_nodes, variable, value).await,
-		}
-	}
-
-	async fn handle_get_var(
-		&self,
-		all_nodes: bool,
-		variable: &Option<String>,
-	) -> Result<AdminRpc, Error> {
-		if all_nodes {
-			let mut ret = vec![];
-			let all_nodes = self.garage.system.cluster_layout().all_nodes().to_vec();
-			for node in all_nodes.iter() {
-				let node = (*node).into();
-				match self
-					.endpoint
-					.call(
-						&node,
-						AdminRpc::Worker(WorkerOperation::Get {
-							all_nodes: false,
-							variable: variable.clone(),
-						}),
-						PRIO_NORMAL,
-					)
-					.await??
-				{
-					AdminRpc::WorkerVars(v) => ret.extend(v),
-					m => return Err(GarageError::unexpected_rpc_message(m).into()),
-				}
-			}
-			Ok(AdminRpc::WorkerVars(ret))
-		} else {
-			#[allow(clippy::collapsible_else_if)]
-			if let Some(v) = variable {
-				Ok(AdminRpc::WorkerVars(vec![(
-					self.garage.system.id,
-					v.clone(),
-					self.garage.bg_vars.get(v)?,
-				)]))
-			} else {
-				let mut vars = self.garage.bg_vars.get_all();
-				vars.sort();
-				Ok(AdminRpc::WorkerVars(
-					vars.into_iter()
-						.map(|(k, v)| (self.garage.system.id, k.to_string(), v))
-						.collect(),
-				))
-			}
-		}
-	}
-
-	async fn handle_set_var(
-		&self,
-		all_nodes: bool,
-		variable: &str,
-		value: &str,
-	) -> Result<AdminRpc, Error> {
-		if all_nodes {
-			let mut ret = vec![];
-			let all_nodes = self.garage.system.cluster_layout().all_nodes().to_vec();
-			for node in all_nodes.iter() {
-				let node = (*node).into();
-				match self
-					.endpoint
-					.call(
-						&node,
-						AdminRpc::Worker(WorkerOperation::Set {
-							all_nodes: false,
-							variable: variable.to_string(),
-							value: value.to_string(),
-						}),
-						PRIO_NORMAL,
-					)
-					.await??
-				{
-					AdminRpc::WorkerVars(v) => ret.extend(v),
-					m => return Err(GarageError::unexpected_rpc_message(m).into()),
-				}
-			}
-			Ok(AdminRpc::WorkerVars(ret))
-		} else {
-			self.garage.bg_vars.set(variable, value)?;
-			Ok(AdminRpc::WorkerVars(vec![(
-				self.garage.system.id,
-				variable.to_string(),
-				value.to_string(),
-			)]))
-		}
-	}
-
-	// ================ META DB COMMANDS ====================
-
-	async fn handle_meta_cmd(self: &Arc<Self>, mo: &MetaOperation) -> Result<AdminRpc, Error> {
-		match mo {
-			MetaOperation::Snapshot { all: true } => {
-				let to = self.garage.system.cluster_layout().all_nodes().to_vec();
-
-				let resps = futures::future::join_all(to.iter().map(|to| async move {
-					let to = (*to).into();
-					self.endpoint
-						.call(
-							&to,
-							AdminRpc::MetaOperation(MetaOperation::Snapshot { all: false }),
-							PRIO_NORMAL,
-						)
-						.await?
-				}))
-				.await;
-
-				let mut ret = vec![];
-				for (to, resp) in to.iter().zip(resps.iter()) {
-					let res_str = match resp {
-						Ok(_) => "ok".to_string(),
-						Err(e) => format!("error: {}", e),
-					};
-					ret.push(format!("{:?}\t{}", to, res_str));
-				}
-
-				if resps.iter().any(Result::is_err) {
-					Err(GarageError::Message(format_table_to_string(ret)).into())
-				} else {
-					Ok(AdminRpc::Ok(format_table_to_string(ret)))
-				}
-			}
-			MetaOperation::Snapshot { all: false } => {
-				garage_model::snapshot::async_snapshot_metadata(&self.garage).await?;
-				Ok(AdminRpc::Ok("Snapshot has been saved.".into()))
-			}
-		}
-	}
-}
-
-impl EndpointHandler<AdminRpc> for AdminRpcHandler {
-	fn handle(
-		self: &Arc<Self>,
-		message: &AdminRpc,
-		_from: NodeID,
-	) -> impl Future<Output = Result<AdminRpc, Error>> + Send {
-		let self2 = self.clone();
-		async move {
-			match message {
-				AdminRpc::BucketOperation(bo) => self2.handle_bucket_cmd(bo).await,
-				AdminRpc::KeyOperation(ko) => self2.handle_key_cmd(ko).await,
-				AdminRpc::LaunchRepair(opt) => self2.handle_launch_repair(opt.clone()).await,
-				AdminRpc::Stats(opt) => self2.handle_stats(opt.clone()).await,
-				AdminRpc::Worker(wo) => self2.handle_worker_cmd(wo).await,
-				AdminRpc::BlockOperation(bo) => self2.handle_block_cmd(bo).await,
-				AdminRpc::MetaOperation(mo) => self2.handle_meta_cmd(mo).await,
-				m => Err(GarageError::unexpected_rpc_message(m).into()),
-			}
-		}
-		.boxed()
-	}
-}