- Per user and per directory permissions: list directories content, upload, overwrite, download, delete, rename, create directories, create symlinks, changing owner/group and mode, changing access and modification times can be enabled or disabled.
- Per user files/folders ownership: you can map all the users to the system account that runs SFTPGo (all platforms are supported) or you can run SFTPGo as root user and map each user or group of users to a different system account (*NIX only).
Binary releases for Linux, macOS and Windows are available, please visit the [releases](https://github.com/drakkan/sftpgo/releases "releases") page.
Sample Dockerfiles for [Debian](https://www.debian.org "Debian") and [Alpine](https://alpinelinux.org "Alpine") are available inside the source tree [docker](./docker "docker") directory.
- [sftpgo](https://aur.archlinux.org/packages/sftpgo/). This package follow stable releases. It requires `git`, `gcc` and `go` to build.
- [sftpgo-bin](https://aur.archlinux.org/packages/sftpgo-bin/). This package follow stable releases downloading the prebuilt linux binary from GitHub. It does not require `git`, `gcc` and `go` to build.
- [sftpgo-git](https://aur.archlinux.org/packages/sftpgo-git/). This package build and install the latest git master. It requires `git`, `gcc` and `go` to build.
Alternately you can install the package to your [$GOPATH](https://github.com/golang/go/wiki/GOPATH "GOPATH") with the [go tool](https://golang.org/cmd/go/ "go command") from shell:
On Linux and macOS a compiler is easy to install or already installed, on Windows you need to download [MinGW-w64](https://sourceforge.net/projects/mingw-w64/files/) and build SFTPGo from its command prompt.
If you don't need SQLite, you can also get/build SFTPGo setting the environment variable `GCO_ENABLED` to 0, this way SQLite support will be disabled but PostgreSQL, MySQL, bbolt and memory data providers will work and you don't need a `C` compiler for building.
For macOS a `launchd` sample [service](./init/com.github.drakkan.sftpgo.plist "launchd plist") can be found inside the source tree. The `launchd` plist assumes that `sftpgo` has `/usr/local/opt/sftpgo` as base directory.
-`--config-dir` string. Location of the config dir. This directory should contain the `sftpgo` configuration file and is used as the base for files with a relative path (eg. the private keys for the SFTP server, the SQLite or bblot database if you use SQLite or bbolt as data provider). The default value is "." or the value of `SFTPGO_CONFIG_DIR` environment variable.
-`--config-file` string. Name of the configuration file. It must be the name of a file stored in config-dir not the absolute path to the configuration file. The specified file name must have no extension we automatically load JSON, YAML, TOML, HCL and Java properties. The default value is "sftpgo" (and therefore `sftpgo.json`, `sftpgo.yaml` and so on are searched) or the value of `SFTPGO_CONFIG_FILE` environment variable.
-`--log-compress` boolean. Determine if the rotated log files should be compressed using gzip. Default `false` or the value of `SFTPGO_LOG_COMPRESS` environment variable (1 or `true`, 0 or `false`). It is unused if `log-file-path` is empty.
-`--log-file-path` string. Location for the log file, default "sftpgo.log" or the value of `SFTPGO_LOG_FILE_PATH` environment variable. Leave empty to write logs to the standard error.
-`--log-max-age` int. Maximum number of days to retain old log files. Default 28 or the value of `SFTPGO_LOG_MAX_AGE` environment variable. It is unused if `log-file-path` is empty.
-`--log-max-backups` int. Maximum number of old log files to retain. Default 5 or the value of `SFTPGO_LOG_MAX_BACKUPS` environment variable. It is unused if `log-file-path` is empty.
-`--log-max-size` int. Maximum size in megabytes of the log file before it gets rotated. Default 10 or the value of `SFTPGO_LOG_MAX_SIZE` environment variable. It is unused if `log-file-path` is empty.
If you don't configure any private host keys, the daemon will use `id_rsa` in the configuration directory. If that file doesn't exist, the daemon will attempt to autogenerate it (if the user that executes SFTPGo has write access to the config-dir). The server supports any private key format supported by [`crypto/ssh`](https://github.com/golang/crypto/blob/master/ssh/keys.go#L32).
-`max_auth_tries` integer. Maximum number of authentication attempts permitted per connection. If set to a negative number, the number of attempts are unlimited. If set to zero, the number of attempts are limited to 6.
-`upload_mode` integer. 0 means standard, the files are uploaded directly to the requested path. 1 means atomic: files are uploaded to a temporary path and renamed to the requested path when the client ends the upload. Atomic mode avoids problems such as a web server that serves partial files when the files are being uploaded. In atomic mode if there is an upload error the temporary file is deleted and so the requested upload path will not contain a partial file. 2 means atomic with resume support: as atomic but if there is an upload error the temporary file is renamed to the requested path and not deleted, this way a client can reconnect and resume the upload.
-`actions`, struct. It contains the command to execute and/or the HTTP URL to notify and the trigger conditions. See the "Custom Actions" paragraph for more details
-`execute_on`, list of strings. Valid values are `download`, `upload`, `delete`, `rename`, `ssh_cmd`. Leave empty to disable actions.
-`command`, string. Absolute path to the command to execute. Leave empty to disable.
-`http_notification_url`, a valid URL. An HTTP GET request will be executed to this URL. Leave empty to disable.
-`keys`, struct array. It contains the daemon's private keys. If empty or missing the daemon will search or try to generate `id_rsa` in the configuration directory.
-`private_key`, path to the private key file. It can be a path relative to the config dir or an absolute one.
-`enable_scp`, boolean. Default disabled. Set to `true` to enable the experimental SCP support. This setting is deprecated and will be removed in future versions, please add `scp` to the `enabled_ssh_commands` list to enable it
-`kex_algorithms`, list of strings. Available KEX (Key Exchange) algorithms in preference order. Leave empty to use default values. The supported values can be found here: [`crypto/ssh`](https://github.com/golang/crypto/blob/master/ssh/common.go#L46 "Supported kex algos")
-`ciphers`, list of strings. Allowed ciphers. Leave empty to use default values. The supported values can be found here: [`crypto/ssh`](https://github.com/golang/crypto/blob/master/ssh/common.go#L28 "Supported ciphers")
-`macs`, list of strings. available MAC (message authentication code) algorithms in preference order. Leave empty to use default values. The supported values can be found here: [`crypto/ssh`](https://github.com/golang/crypto/blob/master/ssh/common.go#L84 "Supported MACs")
-`login_banner_file`, path to the login banner file. The contents of the specified file, if any, are sent to the remote user before authentication is allowed. It can be a path relative to the config dir or an absolute one. Leave empty to send no login banner
-`setstat_mode`, integer. 0 means "normal mode": requests for changing permissions, owner/group and access/modification times are executed. 1 means "ignore mode": requests for changing permissions, owner/group and access/modification times are silently ignored.
-`enabled_ssh_commands`, list of enabled SSH commands. These SSH commands are enabled by default: `md5sum`, `sha1sum`, `cd`, `pwd`. `*` enables all supported commands. Some commands are implemented directly inside SFTPGo, while for other commands we use system commands that need to be installed and in your system's `PATH`. For system commands we have no direct control on file creation/deletion and so we cannot support remote filesystems, such as S3, and quota check is suboptimal: if quota is enabled, the number of files is checked at the command begin and not while new files are created. The allowed size is calculated as the difference between the max quota and the used one and it is checked against the bytes transferred via SSH. The command is aborted if it uploads more bytes than the remaining allowed size calculated at the command start. Anyway we see the bytes that the remote command send to the local command via SSH, these bytes contain both protocol commands and files and so the size of the files is different from the size trasferred via SSH: for example a command can send compressed files or a protocol command (few bytes) could delete a big file. To mitigate this issue quotas are recalculated at the command end with a full home directory scan, this could be heavy for big directories. If you need system commands and quotas you could consider to disable quota restrictions and periodically update quota usage yourself using the REST API. We support the following SSH commands:
-`scp`, SCP is an experimental feature, we have our own SCP implementation since we can't rely on "scp" system command to proper handle quotas and user's home dir restrictions. The SCP protocol is quite simple but there is no official docs about it, so we need more testing and feedbacks before enabling it by default. We may not handle some borderline cases or have sneaky bugs. Please do accurate tests yourself before enabling SCP and let us known if something does not work as expected for your use cases. SCP between two remote hosts is supported using the `-3` scp option.
-`md5sum`, `sha1sum`, `sha256sum`, `sha384sum`, `sha512sum`. Useful to check message digests for uploaded files. These commands are implemented inside SFTPGo so they work even if the matching system commands are not available, for example on Windows.
-`cd`, `pwd`. Some SFTP clients does not support the SFTP SSH_FXP_REALPATH packet type and so they use `cd` and `pwd` SSH commands to get the initial directory. Currently `cd` do nothing and `pwd` always returns the `/` path.
-`git-receive-pack`, `git-upload-pack`, `git-upload-archive`. These commands enable support for Git repositories over SSH, they need to be installed and in your system's `PATH`.
-`rsync`. The `rsync` command need to be installed and in your system's `PATH`. We cannot avoid that rsync create symlinks so if the user has the permission to create symlinks we add the option `--safe-links` to the received rsync command if it is not already set. This should prevent to create symlinks that point outside the home dir. If the user cannot create symlinks we add the option `--munge-links`, if it is not already set. This should make symlinks unusable (but manually recoverable)
-`keyboard_interactive_auth_program`, string. Absolute path to an external program to use for keyboard interactive authentication. See the "Keyboard Interactive Authentication" paragraph for more details.
-`name`, string. Database name. For driver `sqlite` this can be the database name relative to the config dir or the absolute path to the SQLite database. For driver `memory` this is the (optional) path relative to the config dir or the absolute path to the users dump to load.
-`sslmode`, integer. Used for drivers `mysql` and `postgresql`. 0 disable SSL/TLS connections, 1 require ssl, 2 set ssl mode to `verify-ca` for driver `postgresql` and `skip-verify` for driver `mysql`, 3 set ssl mode to `verify-full` for driver `postgresql` and `preferred` for driver `mysql`
-`connectionstring`, string. Provide a custom database connection string. If not empty this connection string will be used instead of build one using the previous parameters. Leave empty for drivers `bolt` and `memory`
- 2, quota is updated each time a user upload or delete a file but only for users with quota restrictions. With this configuration the "quota scan" REST API can still be used to periodically update space usage for users without quota restrictions
-`users_base_dir`, string. Users' default base directory. If no home dir is defined while adding a new user, and this value is a valid absolute path, then the user home dir will be automatically defined as the path obtained joining the base dir and the username
-`actions`, struct. It contains the command to execute and/or the HTTP URL to notify and the trigger conditions. See the "Custom Actions" paragraph for more details
-`execute_on`, list of strings. Valid values are `add`, `update`, `delete`. `update` action will not be fired for internal updates such as the last login or the user quota fields.
-`external_auth_program`, string. Absolute path to an external program to use for users authentication. See the "External Authentication" paragraph for more details.
-`external_auth_scope`, integer. 0 means all supported authetication scopes (passwords, public keys and keyboard interactive). 1 means passwords only. 2 means public keys only. 4 means key keyboard interactive only. The flags can be combined, for example 6 means public keys and keyboard interactive
-`credentials_path`, string. It defines the directory for storing user provided credential files such as Google Cloud Storage credentials. This can be an absolute path or a path relative to the config dir
-`backups_path`, string. Path to the backup directory. This can be an absolute path or a path relative to the config dir. We don't allow backups in arbitrary paths for security reasons
-`auth_user_file`, string. Path to a file used to store usernames and password for basic authentication. This can be an absolute path or a path relative to the config dir. We support HTTP basic authentication and the file format must conform to the one generated using the Apache tool. The supported password formats are bcrypt (`$2y$` prefix) and md5 crypt (`$apr1$` prefix). If empty HTTP authentication is disabled.
-`certificate_file`, string. Certificate for HTTPS. This can be an absolute path or a path relative to the config dir.
-`certificate_key_file`, string. Private key matching the above certificate. This can be an absolute path or a path relative to the config dir. If both the certificate and the private key are provided the the server will expect HTTPS connections.
If you want to use a private key that use an algorithm different from RSA or more than one private key then replace the empty `keys` array with something like this:
The configuration can be read from JSON, TOML, YAML, HCL, envfile and Java properties config files, if your `config-file` flag is set to `sftpgo` (default value) you need to create a configuration file called `sftpgo.json` or `sftpgo.yaml` and so on inside `config-dir`.
You can also override all the available configuration options using environment variables, sftpgo will check for environment variables with a name matching the key uppercased and prefixed with the `SFTPGO_`. You need to use `__` to traverse a struct.
- To set sftpd `bind_port` you need to define the env var `SFTPGO_SFTPD__BIND_PORT`
- To set the `execute_on` actions you need to define the env var `SFTPGO_SFTPD__ACTIONS__EXECUTE_ON` for example `SFTPGO_SFTPD__ACTIONS__EXECUTE_ON=upload,download`
Please note that to override configuration options with environment variables a configuration file containing the options to override is required. You can, for example, deploy the default configuration file and then override the options you need to customize using environment variables.
SQL based data providers (SQLite, MySQL, PostgreSQL) requires the creation of a database containing the required tables. Memory and bolt data providers does not require an initialization.
SQL scripts to create the required database structure can be found inside the source tree [sql](./sql "sql") directory. The SQL scripts filename is, by convention, the date as `YYYYMMDD` and the suffix `.sql`. You need to apply all the SQL scripts for your database ordered by name, for example `20190828.sql` must be applied before `20191112.sql` and so on.
The `memory` provider can load users from a dump obtained using the `dumpdata` REST API. The path to this dump file can be configured using the dataprovider `name` configuration key. It will be loaded at startup and can be reloaded on demand sending a `SIGHUP` signal on Unix based systems and a `paramchange` request to the running service on Windows. The `memory` provider will not modify the provided file so quota usage and last login will not be persisted.
Custom authentication methods can easily be added. SFTPGo supports external authentication modules, and writing a new backend can be as simple as a few lines of shell script.
To enable external authentication you must set the absolute path of your authentication program using `external_auth_program` key in your configuration file.
The external program can read the following environment variables to get info about the user trying to authenticate:
-`SFTPGO_AUTHD_USERNAME`
-`SFTPGO_AUTHD_PASSWORD`, not empty for password authentication
-`SFTPGO_AUTHD_PUBLIC_KEY`, not empty for public key authentication
Previous global environment variables aren't cleared when the script is called. The content of these variables is _not_ quoted. They may contain special characters. They are under the control of a possibly malicious remote user.
The program must respond on the standard output with a valid SFTPGo user serialized as JSON if the authentication succeed or an user with an empty username if the authentication fails.
If the authentication succeed the user will be automatically added/updated inside the defined data provider. Actions defined for user added/updated will not be executed in this case.
The external program should check authentication only, if there are login restrictions such as user disabled, expired, login allowed only from specific IP addresses it is enough to populate the matching user fields and these conditions will be checked in the same way as for built-in users.
If you have an external authentication program that could be useful for others too, for example LDAP/Active Directory authentication, please let us know and/or send a pull request.
Keyboard interactive authentication is in general case a series of question asked by the server with responses provided by the client.
This authentication method is typically used for multi factor authentication.
There is no restrictions on the number of questions asked on a particular authentication stage; there is also no restrictions on the number of stages involving different sets of questions.
To enable keyboard interactive authentication you must set the absolute path of your authentication program using `keyboard_interactive_auth_program` key in your configuration file.
The external program can read the following environment variables to get info about the user trying to authenticate:
-`SFTPGO_AUTHD_USERNAME`
-`SFTPGO_AUTHD_PASSWORD`, this is the hashed password as stored inside the data provider
Previous global environment variables aren't cleared when the script is called. The content of these variables is _not_ quoted. They may contain special characters.
The program must write the questions on its standard output, in a single line, using the following struct JSON serialized:
-`instruction`, string. A short description to show to the user that is trying to authenticate. Can be empty or omitted
-`questions`, list of questions to be asked to the user
-`echos` list of boolean flags corresponding to the questions (so the lengths of both lists must be the same) and indicating whether user's reply for a particular question should be echoed on the screen while they are typing: true if it should be echoed, or false if it should be hidden.
-`auth_result`, integer. Set this field to 1 to indicate successful authentication, 0 is ignored, any other value means authentication error. If this fields is found and it is different from 0 then SFTPGo does not read any other questions from the external program and finalize the authentication.
Keyboard interactive authentication can be chained to the external authentication.
The authentication must finish within 60 seconds.
Let's see a very basic example. Our sample keyboard interactive authentication program will ask for 2 sets of questions and accept the user if the answer to the last question is `answer3`.
```
#!/bin/sh
echo '{"questions":["Question1: ","Question2: "],"instruction":"This is a sample for keyboard interactive authentication","echos":[true,false]}'
SFTPGo allows to configure custom commands and/or HTTP notifications on file upload, download, delete, rename, on SSH commands and on user add, update and delete.
The `actions` struct inside the "sftpd" configuration section allows to configure actions on file upload, download, delete, rename and on SSH commands.
Actions will not be executed if an error is detected and so a partial file is uploaded or downloaded or an SSH command is not successfully completed. The `upload` condition includes both uploads to new files and overwrite of existing files. The `ssh_cmd` condition will be triggered after a command is successfully executed via SSH. `scp` will trigger the `download` and `upload` conditions and not `ssh_cmd`.
The `command`, if defined, is invoked with the following arguments:
The `http_notification_url`, if defined, will be called invoked as http POST. The action is added to the query string, for example `<http_notification_url>?action=update` and the user is sent serialized as JSON inside the POST body
Specifying a different `key_prefix` you can assign different virtual folders of the same bucket to different users. This is similar to a chroot directory for local filesystem. Each SFTP/SCP user can only access to the assigned virtual folder and to its contents The virtual folder identified by `key_prefix` does not need to be pre-created.
- We don't support renaming non empty directories since we should rename all the contents too and this could take long time: think about directories with thousands of files, for each file we should do an AWS API call.
- For server side encryption you have to configure the mapped bucket to automatically encrypt objects.
- A local home directory is still required to store temporary files.
Each user can be mapped with a Google Cloud Storage bucket or a bucket virtual folder, this way the mapped bucket/virtual folder is exposed over SFTP/SCP. This backend is very similar to the S3 backend and it has the same limitations.
## Other Storage backends
Adding new storage backends it's quite easy:
- implement the [Fs interface](./vfs/vfs.go#L18 "interface for filesystem backends").
- update the user method `GetFilesystem` to return the new backend
- update the web interface and the REST API CLI
- add the flags for the new storage backed to the `portable` mode
Anyway some backends require a pay per use account (or they offer free account for a limited time period only), to be able to add support for such backends or to review pull requests please provide a test account. The test account must be available over the time to be able to maintain the backend and do basic tests before each new release.
-C, --advertise-credentials If the SFTP service is advertised via multicast DNS this flag allows to put username/password inside the advertised TXT record
-S, --advertise-service Advertise SFTP service using multicast DNS (default true)
-d, --directory string Path to the directory to serve. This can be an absolute path or a path relative to the current directory (default ".")
-f, --fs-provider int 0 means local filesystem, 1 Amazon S3 compatible, 2 Google Cloud Storage
--gcs-bucket string
--gcs-credentials-file string Google Cloud Storage JSON credentials file
--gcs-key-prefix string Allows to restrict access to the virtual folder identified by this prefix and its contents
--gcs-storage-class string
-h, --help help for portable
-l, --log-file-path string Leave empty to disable logging
-p, --password string Leave empty to use an auto generated value
-g, --permissions strings User's permissions. "*" means any permission (default [list,download])
In portable mode SFTPGo can advertise the SFTP service and, optionally, the credentials via multicast DNS, so there is a standard way to discover the service and to automatically connect to it.
Here is an example of the advertised service including credentials as seen using `avahi-browse`:
```
= enp0s31f6 IPv4 SFTPGo portable 53705 SFTP File Transfer local
-`password` used for password authentication. For users created using SFTPGo REST API if the password has no known hashing algo prefix it will be stored using argon2id. SFTPGo supports checking passwords stored with bcrypt, pbkdf2, md5crypt and sha512crypt too. For pbkdf2 the supported format is `$<algo>$<iterations>$<salt>$<hashed pwd base64 encoded>`, where algo is `pbkdf2-sha1` or `pbkdf2-sha256` or `pbkdf2-sha512`. For example the `pbkdf2-sha256` of the word `password` using 150000 iterations and `E86a9YMX3zC7` as salt must be stored as `$pbkdf2-sha256$150000$E86a9YMX3zC7$R5J62hsSq+pYw00hLLPKBbcGXmq7fj5+/M0IFoYtZbo=`. For bcrypt the format must be the one supported by golang's [crypto/bcrypt](https://godoc.org/golang.org/x/crypto/bcrypt) package, for example the password `secret` with cost `14` must be stored as `$2a$14$ajq8Q7fbtFRQvXpdCq7Jcuy.Rx1h/L4J60Otx.gyNLbAYctGMJ9tK`. For md5crypt and sha512crypt we support the format used in `/etc/shadow` with the `$1$` and `$6$` prefix, this is useful if you are migrating from Unix system user accounts. We support Apache md5crypt (`$apr1$` prefix) too. Using the REST API you can send a password hashed as bcrypt, pbkdf2, md5crypt or sha512crypt and it will be stored as is.
-`uid`, `gid`. If sftpgo runs as root system user then the created files and directories will be assigned to this system uid/gid. Ignored on windows and if sftpgo runs as non root user: in this case files and directories for all SFTP users will be owned by the system user that runs sftpgo.
-`chmod` changing file or directory permissions is allowed. On Windows, only the 0200 bit (owner writable) of mode is used; it controls whether the file's read-only attribute is set or cleared. The other bits are currently unused. Use mode 0400 for a read-only file and 0600 for a readable+writable file.
-`chown` changing file or directory owner and group is allowed. Changing owner and group is not supported on Windows.
-`allowed_ip`, List of IP/Mask allowed to login. Any IP address not contained in this list cannot login. IP/Mask must be in CIDR notation as defined in RFC 4632 and RFC 4291, for example "192.0.2.0/24" or "2001:db8::/32"
-`denied_ip`, List of IP/Mask not allowed to login. If an IP address is both allowed and denied then login will be denied
- If your accounts are aleady stored inside a supported database, you can create a database view. Since a view is read only, you have to disable user management and quota tracking so SFTPGo will never try to write to the view
- you can import your users inside SFTPGo. Take a look at [sftpgo_api_cli.py](./scripts/README.md "sftpgo_api_cli script"), it can convert and import users from Linux system users and Pure-FTPd/ProFTPD virtual users
SFTPGo exposes REST API to manage, backup and restore users and to get real time reports of the active connections with possibility of forcibly closing a connection.
If quota tracking is enabled in `sftpgo` configuration file, then the used size and number of files are updated each time a file is added/removed. If files are added/removed not using SFTP/SCP or if you change `track_quota` from `2` to `1`, you can rescan the users home dir and update the used quota using the REST API.
REST API can be protected using HTTP basic authentication and exposed via HTTPS, if you need more advanced security features you can setup a reverse proxy using an HTTP Server such as Apache or NGNIX.
You can also generate your own REST client, in your preferred programming language or even bash scripts, using an OpenAPI generator such as [swagger-codegen](https://github.com/swagger-api/swagger-codegen) or [OpenAPI Generator](https://openapi-generator.tech/)
You can easily build your own interface using the exposed REST API, anyway SFTPGo provides also a very basic built-in web interface that allows to manage users and connections.
The web interface can be protected using HTTP basic authentication and exposed via HTTPS, if you need more advanced security features you can setup a reverse proxy as explained for the REST API.
- **"connection failed logs"**, logs for failed attempts to initialize a connection. A connection can fail for an authentication error or other errors such as a client abort or a timeout if the login does not happen in two minutes
The **connection failed logs** can be used for integration in tools such as [Fail2ban](http://www.fail2ban.org/). Example of [jails](./fail2ban/jails) and [filters](./fail2ban/filters) working with `systemd`/`journald` are available in fail2ban directory.
