slcl, a suckless cloud

slcl is a simple and fast implementation of a web file server, commonly known as "cloud storage" or simply "cloud", written in C99.

Disclaimer

While slcl might not share some of the philosophical views from the suckless project, it still strives towards minimalism, simplicity and efficiency.

Features

• Private access directory with file uploading, with configurable quota.
• Read-only public file sharing.
• Its own, tiny HTTP/1.0 and 1.1-compatible server.
• A simple JSON file as the credentials database.
• No JavaScript.

TLS

In order to maintain simplicity and reduce the risk for security bugs, slcl does not implement TLS support. Instead, this should be provided by a reverse proxy, such as caddy.

Root permissions

slcl is expected to listen to connections from any port number so that root access is not required. So, in order to avoid the risk for security bugs, please do not run slcl as root.

Encryption

Since no client-side JavaScript is used, files are uploaded unencrypted to slcl. If required, encryption should be done before uploading e.g.: using gpg.

Requirements

• A POSIX environment.
• OpenSSL >= 3.0.
• cJSON >= 1.7.15.
• dynstr (provided as a git submodule).
• xxd (for usergen only).
• CMake (optional).

Ubuntu / Debian

sudo apt install libcjson-dev libssl-dev xxd

How to use

Build

Two build environments are provided for slcl - feel free to choose any of them:

• A mostly POSIX-compliant Makefile.
• A CMakeLists.txt.

slcl can be built using the standard build process:

Make

$ make

CMake

$ mkdir build/
$ cmake ..
$ cmake --build .

Setting up

slcl consumes a path to a directory with the following tree structure:

.
├── db.json
├── public
└── user

Where:

• db.json is the credentials database. Details are explained below.
  • Note: slcl creates a database with no users if not found, with file mode bits set to 0600.
• public is a directory containing read-only files that can be accessed without authentication. Internally, they are implemented as simlinks to other files.
  • Note: this directory must be created before running slcl.
• user is a directory containing user directories, which in turn contain anything users put into them.
  • Note: this directory must be created before running slcl.

A more complete example:

.
├── db.json
├── public
│   └── 416d604c03a1cbb2 -> user/alice/file.txt
└── user
    ├── alice
    │   └── file.txt
    └── john
        └── file2.txt

Note: user directories (alice and john on the example above) must be created before running slcl.

Credentials database

slcl reads credentials from the db.json database, with the following schema:

{
    "users": [{
        "name":	"...",
        "password":	"...",
        "salt":	"...",
        "key":	"...",
        "quota": "..."
    }]
}

usergen is an interactive script that consumes a username, a password and, optionally, a user quota in MiB. Then, usergen writes a JSON object that can be appended to the users JSON array in db.json. A salt is randomly generated using openssl and passwords are hashed multiple times beforehand - see usergen and auth.c for further reference. Also, a random key is generated that is later used to sign HTTP cookies.

When users authenticate from a web browser, slcl sends a SHA256HMAC-signed JSON Web Token, using the random key generated by usergen. No session data is kept on the server.

Running

To run slcl, simply run the executable with the path to a directory including the files listed above. By default, slcl will listen to incoming connections on a random TCP port number. To set a specific port number, use the -p command line option. For example:

slcl -p 7822 ~/my-db/

Why this project?

Previously, I had been recommended Nextcloud as an alternative to proprietary services like Dropbox. Unfortunately, despite being a very flexible piece of software, Nextcloud is way too heavy on resources, specially on lower end hardware such as the Raspberry Pi 3:

• It uses around 30% RAM on my Raspberry Pi 3, configured with 973 MiB of RAM, and of course it gets worse with several simultaneous users.
• Simple operations like searching and previewing files cause large amounts of I/O and RAM usage, so much that it locks the whole server up more often than not.
• Nextcloud pages are bloated. Even the login page is over 15 MiB (!).
• Requires clients to run JavaScript, which also has a significant performance penalty on the web browser. Also, some users do not feel comfortable running JavaScript from their web browsers, and thus prefer to disable it.

After years of recurring frustration as a Nextcloud administrator and user, I looked for alternatives that stripped out most of the unneeded bloat from Nextcloud, while providing the required features listed above. However, I could not find any that fit them, so I felt challenged to design a new implementation.

On the other hand, command line-based solutions like rsync might not be as convenient for non-technical people, compared to a web browser, or might not be even available e.g.: phones.

License

slcl, a suckless cloud.
Copyright (C) 2023  Xavier Del Campo Romero

This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU Affero General Public License for more details.

You should have received a copy of the GNU Affero General Public License
along with this program.  If not, see <https://www.gnu.org/licenses/>.

Also, see LICENSE.