Page MenuHomePhorge
Files F114869

No OneTemporary
Actions

Size
17 KB
Referenced Files
None
Subscribers
None

View Options

diff --git a/README.md b/README.md
index a182566..7d1d402 100644
--- a/README.md
+++ b/README.md
@@ -1,155 +1,155 @@
# GenMagic
**GenMagic** provides supervised and customisable access to [libmagic](http://man7.org/linux/man-pages/man3/libmagic.3.html) using a supervised external process.
With this library, you can start an one-off process to run a single check, or run the process as a daemon if you expect to run many checks.
## Installation
The package can be installed by adding `gen_magic` to your list of dependencies in `mix.exs`:
```elixir
def deps do
[
{:gen_magic, "~> 1.0.0"}
]
end
```
You must also have [libmagic](http://man7.org/linux/man-pages/man3/libmagic.3.html) installed locally with headers, alongside common compilation tools (i.e. build-essential). These can be acquired by apt-get, yum, brew, etc.
Compilation of the underlying C program is automatic and handled by [elixir_make](https://github.com/elixir-lang/elixir_make).
## Usage
Depending on the use case, you may utilise a single (one-off) GenMagic process without reusing it as a daemon, or utilise a connection pool (such as Poolboy) in your application to run multiple persistent GenMagic processes.
To use GenMagic directly, you can use `GenMagic.Helpers.perform_once/1`:
```elixir
-iex(1)> GenMagic.Helpers.perform_once "."
+iex(1)> GenMagic.perform(".", once: true)
{:ok,
%GenMagic.Result{
content: "directory",
encoding: "binary",
mime_type: "inode/directory"
}}
```
To use the GenMagic server as a daemon, you can start it first, keep a reference, then feed messages to it as you require:
```elixir
{:ok, pid} = GenMagic.Server.start_link([])
-{:ok, result} = GenMagic.Server.perform(pid, path)
+{:ok, result} = GenMagic.perform(path, server: pid)
```
See `GenMagic.Server.start_link/1` and `t:GenMagic.Server.option/0` for more information on startup parameters.
See `GenMagic.Result` for details on the result provided.
## Configuration
When using `GenMagic.Server.start_link/1` to start a persistent server, or `GenMagic.Helpers.perform_once/2` to run an ad-hoc request, you can override specific options to suit your use case.
| Name | Default | Description |
| - | - | - |
| `:startup_timeout` | 1000 | Number of milliseconds to wait for client startup |
| `:process_timeout` | 30000 | Number of milliseconds to process each request |
| `:recycle_threshold` | 10 | Number of cycles before the C process is replaced |
| `:database_patterns` | `[:default]` | Databases to load |
See `t:GenMagic.Server.option/0` for details.
### Use Cases
### Ad-Hoc Requests
For ad-hoc requests, you can use the helper method `GenMagic.Helpers.perform_once/2`:
```elixir
-iex(1)> GenMagic.Helpers.perform_once(Path.join(File.cwd!(), "Makefile"))
+iex(1)> GenMagic.perform(Path.join(File.cwd!(), "Makefile"), once: true)
{:ok,
%GenMagic.Result{
content: "makefile script, ASCII text",
encoding: "us-ascii",
mime_type: "text/x-makefile"
}}
```
### Supervised Requests
The Server should be run under a supervisor which provides resiliency.
Here we run it under a supervisor:
```elixir
iex(1)> {:ok, pid} = Supervisor.start_link([{GenMagic.Server, name: :gen_magic}], strategy: :one_for_one)
{:ok, #PID<0.199.0>}
```
Now we can ask it to inspect a file:
```elixir
-iex(2)> GenMagic.Server.perform(:gen_magic, Path.expand("~/.bash_history"))
-{:ok, [mime_type: "text/plain", encoding: "us-ascii", content: "ASCII text"]}
+iex(2)> GenMagic.perform(Path.expand("~/.bash_history"), server: :gen_magic)
+{:ok, %GenMagic.Result{mime_type: "text/plain", encoding: "us-ascii", content: "ASCII text"}}
```
Note that in this case we have opted to use a named process.
### Pool
For concurrency *and* resiliency, you may start the `GenMagic.Pool`. By default, it will start a `GenMagic.Server`
worker per online scheduler:
You can add a pool in your application supervisor by adding it as a child:
```
children =
[
# ...
{GenMagic.Pool, [name: YourApp.GenMagicPool, pool_size: 2]}
]
opts = [strategy: :one_for_one, name: Pleroma.Supervisor]
Supervisor.start_link(children, opts)
```
-And then you can use it with `GenMagic.Pool.perform/2`:
+And then you can use it with `GenMagic.perform/2` with `pool: YourApp.GenMagicPool` option:
```
-iex(1)> GenMagic.Pool.perform(YourApp.GenMagicPool, Path.expand("~/.bash_history"))
-{:ok, [mime_type: "text/plain", encoding: "us-ascii", content: "ASCII text"]}
+iex(1)> GenMagic.perform(Path.expand("~/.bash_history"), pool: YourApp.GenMagicPool)
+{:ok, %GenMagic.Result{mime_type: "text/plain", encoding: "us-ascii", content: "ASCII text"}}
```
### Check Uploaded Files
If you use Phoenix, you can inspect the file from your controller:
```elixir
def upload(conn, %{"upload" => %{path: path}}) do,
- {:ok, result} = GenMagic.Helpers.perform_once(:gen_magic, path)
+ {:ok, result} = GenMagic.perform(path, server: :gen_magic)
text(conn, "Received your file containing #{result.content}")
end
```
Obviously, it will be more ideal if you have wrapped `GenMagic.Server` in a pool such as Poolboy, to avoid constantly starting and stopping the underlying C program.
## Notes
### Soak Test
Run an endless cycle to prove that the program is resilient:
```bash
find /usr/share/ -name *png | xargs mix run test/soak.exs
find . -name *ex | xargs mix run test/soak.exs
```
## Acknowledgements
During design and prototype development of this library, the Author has drawn inspiration from the following individuals, and therefore thanks all contributors for their generosity:
- Mr [James Every](https://github.com/devstopfix)
- Enhanced Elixir Wrapper (based on GenServer)
- Initial Hex packaging (v.0.22)
- Soak Testing
diff --git a/lib/gen_magic.ex b/lib/gen_magic.ex
index f598521..292dfe8 100644
--- a/lib/gen_magic.ex
+++ b/lib/gen_magic.ex
@@ -1,7 +1,40 @@
defmodule GenMagic do
@moduledoc """
Top-level namespace for GenMagic, the libmagic client for Elixir.
See `GenMagic.Server` or the README for usage.
"""
+
+ @doc """
+ Perform on `path`.
+
+ An option of `server: ServerName`, `pool: PoolName` or `once: true` must be passed.
+ """
+ @type option :: name
+ when name: {:pool, atom()} | {:server, GenMagic.Server.t()} | {:once, true}
+
+ @spec perform(GenMagic.Server.target(), [option()]) :: GenMagic.Server.result()
+ def perform(path, opts, timeout \\ 5000) do
+ mod = cond do
+ Keyword.has_key?(opts, :pool) -> {GenMagic.Pool, Keyword.get(opts, :pool)}
+ Keyword.has_key?(opts, :server) -> {GenMagic.Server, Keyword.get(opts, :server)}
+ Keyword.has_key?(opts, :once) -> {GenMagic.Helpers, nil}
+ true -> nil
+ end
+
+ if mod do
+ do_perform(mod, path, timeout)
+ else
+ {:error, :no_method}
+ end
+ end
+
+ defp do_perform({GenMagic.Helpers, _}, path, timeout) do
+ GenMagic.Helpers.perform_once(path, timeout)
+ end
+
+ defp do_perform({mod, name}, path, timeout) do
+ mod.perform(name, path, tiemout)
+ end
+
end
diff --git a/lib/gen_magic/server.ex b/lib/gen_magic/server.ex
index fc686b2..0e34586 100644
--- a/lib/gen_magic/server.ex
+++ b/lib/gen_magic/server.ex
@@ -1,313 +1,315 @@
defmodule GenMagic.Server do
@moduledoc """
Provides access to the underlying libmagic client, which performs file introspection.
The Server needs to be supervised, since it will terminate if it receives any unexpected error.
"""
@behaviour :gen_statem
alias GenMagic.Result
alias GenMagic.Server.Data
alias GenMagic.Server.Status
import Kernel, except: [send: 2]
@typedoc """
Represents the reference to the underlying server, as returned by `:gen_statem`.
"""
@type t :: :gen_statem.server_ref()
@typedoc """
Represents values accepted as startup options, which can be passed to `start_link/1`.
- `:name`: If present, this will be the registered name for the underlying process.
Note that `:gen_statem` requires `{:local, name}`, but given widespread GenServer convention,
atoms are accepted and will be converted to `{:local, name}`.
- `:startup_timeout`: Specifies how long the Server waits for the C program to initialise.
However, if the underlying C program exits, then the process exits immediately.
Can be set to `:infinity`.
- `:process_timeout`: Specifies how long the Server waits for each request to complete.
Can be set to `:infinity`.
Please note that, if you have chosen a custom timeout value, you should also pass it when
using `GenMagic.Server.perform/3`.
- `:recycle_threshold`: Specifies the number of requests processed before the underlying C
program is recycled.
Can be set to `:infinity` if you do not wish for the program to be recycled.
- `:database_patterns`: Specifies what magic databases to load; you can specify a list of either
Path Patterns (see `Path.wildcard/2`) or `:default` to instruct the C program to load the
appropriate databases.
For example, if you have had to add custom magics, then you can set this value to:
[:default, "path/to/my/magic"]
"""
@type option ::
{:name, atom() | :gen_statem.server_name()}
| {:startup_timeout, timeout()}
| {:process_timeout, timeout()}
| {:recycle_threshold, non_neg_integer() | :infinity}
| {:database_patterns, nonempty_list(:default | Path.t())}
+ @type target :: Path.t() | {:bytes, binary()}
+ @type result :: {:ok, Result.t()} | {:error, term() | String.t()}
+
@typedoc """
Current state of the Server:
- `:pending`: This is the initial state; the Server will attempt to start the underlying Port
and the libmagic client, then automatically transition to either Available or Crashed.
- `:available`: This is the default state. In this state the Server is able to accept requests
and they will be replied in the same order.
- `:processing`: This is the state the Server will be in if it is processing requests. In this
state, further requests can still be lodged and they will be processed when the Server is
available again.
For proper concurrency, use a process pool like Poolboy, Sbroker, etc.
- `:recycling`: This is the state the Server will be in, if its underlying C program needs to be
recycled. This state is triggered whenever the cycle count reaches the defined value as per
`:recycle_threshold`.
In this state, the Server is able to accept requests, but they will not be processed until the
underlying C server program has been started again.
"""
@type state :: :starting | :processing | :available | :recycling
@spec child_spec([option()]) :: Supervisor.child_spec()
@spec start_link([option()]) :: :gen_statem.start_ret()
- @spec perform(t(), Path.t() | {:bytes, binary()}, timeout()) ::
- {:ok, Result.t()} | {:error, term() | String.t()}
+ @spec perform(t(), target(), timeout()) :: result()
@spec status(t(), timeout()) :: {:ok, Status.t()} | {:error, term()}
@spec stop(t(), term(), timeout()) :: :ok
@doc """
Returns the default Child Specification for this Server for use in Supervisors.
You can override this with `Supervisor.child_spec/2` as required.
"""
def child_spec(options) do
%{
id: __MODULE__,
start: {__MODULE__, :start_link, [options]},
type: :worker,
restart: :permanent,
shutdown: 500
}
end
@doc """
Starts a new Server.
See `t:option/0` for further details.
"""
def start_link(options) do
{name, options} = Keyword.pop(options, :name)
case name do
nil -> :gen_statem.start_link(__MODULE__, options, [])
name when is_atom(name) -> :gen_statem.start_link({:local, name}, __MODULE__, options, [])
{:global, _} -> :gen_statem.start_link(name, __MODULE__, options, [])
{:via, _, _} -> :gen_statem.start_link(name, __MODULE__, options, [])
{:local, _} -> :gen_statem.start_link(name, __MODULE__, options, [])
end
end
@doc """
Determines the type of the file provided.
"""
def perform(server_ref, path, timeout \\ 5000) do
case :gen_statem.call(server_ref, {:perform, path}, timeout) do
{:ok, %Result{} = result} -> {:ok, result}
{:error, reason} -> {:error, reason}
end
end
@doc """
Returns status of the Server.
"""
def status(server_ref, timeout \\ 5000) do
:gen_statem.call(server_ref, :status, timeout)
end
@doc """
Stops the Server with reason `:normal` and timeout `:infinity`.
"""
def stop(server_ref) do
:gen_statem.stop(server_ref)
end
@doc """
Stops the Server with the specified reason and timeout.
"""
def stop(server_ref, reason, timeout) do
:gen_statem.stop(server_ref, reason, timeout)
end
@impl :gen_statem
def init(options) do
import GenMagic.Config
data = %Data{
port_name: get_port_name(),
port_options: get_port_options(options),
startup_timeout: get_startup_timeout(options),
process_timeout: get_process_timeout(options),
recycle_threshold: get_recycle_threshold(options)
}
{:ok, :starting, data}
end
@impl :gen_statem
def callback_mode do
[:state_functions, :state_enter]
end
@doc false
def starting(:enter, _, %{request: nil, port: nil} = data) do
port = Port.open(data.port_name, data.port_options)
{:keep_state, %{data | port: port}, data.startup_timeout}
end
@doc false
def starting({:call, from}, :status, data) do
handle_status_call(from, :starting, data)
end
@doc false
def starting({:call, _from}, {:perform, _path}, _data) do
{:keep_state_and_data, :postpone}
end
@doc false
def starting(:info, {port, {:data, ready}}, %{port: port} = data) do
case :erlang.binary_to_term(ready) do
:ready -> {:next_state, :available, data}
end
end
def starting(:info, {port, {:exit_status, code}}, %{port: port} = data) do
error =
case code do
1 -> :no_database
2 -> :no_argument
3 -> :missing_database
4 -> :ei_alloc_failed
5 -> :ei_bad_term
code -> {:unexpected_error, code}
end
{:stop, {:error, error}, data}
end
@doc false
def available(:enter, _old_state, %{request: nil}) do
:keep_state_and_data
end
@doc false
def available({:call, from}, {:perform, path}, data) do
data = %{data | cycles: data.cycles + 1, request: {path, from, :erlang.now()}}
arg =
case path do
path when is_binary(path) -> {:file, path}
{:bytes, bytes} -> {:bytes, bytes}
end
send(data.port, arg)
{:next_state, :processing, data}
end
@doc false
def available({:call, from}, :status, data) do
handle_status_call(from, :available, data)
end
@doc false
def processing(:enter, _old_state, %{request: {_path, _from, _time}} = data) do
{:keep_state_and_data, data.process_timeout}
end
@doc false
def processing({:call, _from}, {:perform, _path}, _data) do
{:keep_state_and_data, :postpone}
end
@doc false
def processing({:call, from}, :status, data) do
handle_status_call(from, :processing, data)
end
@doc false
def processing(:info, {port, {:data, response}}, %{port: port, request: {_, from, _}} = data) do
response = {:reply, from, handle_response(response)}
next_state = (data.cycles >= data.recycle_threshold && :recycling) || :available
{:next_state, next_state, %{data | request: nil}, [response, :hibernate]}
end
@doc false
def recycling(:enter, _, %{request: nil, port: port} = data) when is_port(port) do
send(data.port, {:stop, :recycle})
{:keep_state_and_data, data.startup_timeout}
end
@doc false
def recycling({:call, _from}, {:perform, _path}, _data) do
{:keep_state_and_data, :postpone}
end
@doc false
def recycling({:call, from}, :status, data) do
handle_status_call(from, :recycling, data)
end
@doc false
def recycling(:info, {port, {:exit_status, 0}}, %{port: port} = data) do
{:next_state, :starting, %{data | port: nil, cycles: 0}}
end
@doc false
@impl :gen_statem
def terminate(_, _, %{port: port}) do
Kernel.send(port, {self(), :close})
end
def terminate(_, _, _) do
:ok
end
defp send(port, command) do
Kernel.send(port, {self(), {:command, :erlang.term_to_binary(command)}})
end
@errnos %{
2 => :enoent,
13 => :eaccess,
20 => :enotdir,
12 => :enomem,
24 => :emfile,
36 => :enametoolong
}
@errno Map.keys(@errnos)
defp handle_response(data) do
case :erlang.binary_to_term(data) do
{:ok, {mime_type, encoding, content}} -> {:ok, Result.build(mime_type, encoding, content)}
{:error, {errno, _}} when errno in @errno -> {:error, @errnos[errno]}
{:error, {errno, string}} -> {:error, "#{errno}: #{string}"}
{:error, _} = error -> error
end
end
defp handle_status_call(from, state, data) do
response = {:ok, %__MODULE__.Status{state: state, cycles: data.cycles}}
{:keep_state_and_data, {:reply, from, response}}
end
end

File Metadata

Mime Type
text/x-diff
Expires
Tue, Nov 26, 6:24 PM (1 d, 11 h)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
40207
Default Alt Text
(17 KB)

Event Timeline