9.6 KiB
Rust support for Event Tracing for Windows (ETW)
Provides the #[trace_logging_provider]
macro, which allows you to define a
Trace Logging Provider
for use with the Event Tracing for Windows (ETW)
framework.
This macro is intended for use only when targeting Windows. When targeting other platforms, this macro will still work, but will generate code that does nothing.
This framework allows applications to log schematized events, rather than textual strings. ETW analysis tools can reliably identify fields within your events, and treat them as strongly-typed data, rather than text strings.
To use tracing with ETW, see tracing-etw.
How to create and use an event provider
In ETW, an event provider is a software object that generates events. Event controllers set up event logging sessions, and event consumers read and interpret event data. This crate focuses on enabling applications to create event providers.
Add crate dependencies
Add these dependencies to your Cargo.toml
file:
[dependencies]
win_etw_macros = "0.1.*"
win_etw_provider = "0.1.*"
win_etw_macros
contains the procedural macro that generates eventing code.
win_etw_provider
contains library code that is called by the code that is generated by
win_etw_macros
.
Define the event provider and its events
Add a trait definition to your source code and annotate it with the #[trace_logging_provider]
macro. The #[trace_logging_provider]
macro consumes the trait definition and produces a struct
definition with the same name and the same method signatures. (The trait is not available for
use as an ordinary trait.):
#[trace_logging_provider]
pub trait MyAppEvents {}
Each event provider must have a unique name and GUID. ETW uses this GUID to identify events that are generated by your provider. Windows contains many event providers, so it is important to be able to select only the events generated by your application. This GUID is also used internally by ETW to identify event metadata (field types), so it is important that your GUID be unique. Otherwise, events from conflicting sources that use the same GUID may be incorrectly interpreted.
Specify your provider name
Unless overridden, #[trace_logging_provider]
uses the name of your trait definition as the name of ETW provider ("MyAppEvents" in the example aboved).
To specify a different name, specify the name with #[trace_logging_provider(name = "MyCompany.MyComponent")]
.
Generate a GUID for your event provider
The #[trace_logging_provider]
macro will generate a .NET EventSource
-compatible name-based GUID if you do not
specify a guid
parameter. The generated GUID is identical to the one generated by the following PowerShell code:
[System.Diagnostics.Tracing.EventSource]::new("MyCompany.MyComponent").Guid
. The GUID is accessible from your Rust code via the associated constant named PROVIDER_GUID
(e.g., MyAppEvents::PROVIDER_GUID
).
If you are not interested in using a name-based GUID, you can generate a GUID using a tool like
uuidgen
(available from Visual Studio command line, or an Ubuntu shell) and specify it with
#[trace_logging_provider(guid = "... your guid here ...")]
.
Add events to your provider
In the trait definition, add method signatures. Each method signature defines an event type. The parameters of each method define the fields of the event type. Only a limited set of field types are supported (enumerated below).
use win_etw_macros::trace_logging_provider;
#[trace_logging_provider(name = "MyCompany.MyComponent")]
pub trait MyAppEvents {
fn http_request(client_address: &SockAddr, is_https: bool, status_code: u32, status: &str);
fn database_connection_created(connection_id: u64, server: &str);
fn database_connection_closed(connection_id: u64);
// ...
}
Create an instance of the event provider
At initialization time (in your fn main()
, etc.), create an instance of the event provider:
let my_app_events = MyAppEvents::new();
Your application should only create a single instance of each event provider, per process. That is, you should create a single instance of your event provider and share it across your process. Typically, an instance is stored in static variable, using a lazy / atomic assignment. There are many crates and types which can support this usage pattern.
Call event methods to report events
To report an event, call one of the methods defined on the event provider. The method will call into ETW to report the event, but there is no guarantee that the event is stored or forwarded; events can be dropped if event buffer resources are scarce.
my_app_events.client_connected(None, &"192.168.0.42:6667".parse(), false, 100, "OK");
Note that all generated event methods have an added first parameters,
options: Option<&EventOptions>
. This parameter allows you to override per-event parameters,
such as the event level and event correlation IDs. In most cases, you should pass None
.
Supported field types
Only a limited set of field types are supported.
- Integer primitives up to 64 bits:
i8
,i16
,i32
,i64
,u8
,u16
,u32
,u64
- Floating point primitives:
f32
,f64
- Architecture-dependent sizes:
usize
,isize
. - Boolean:
bool
- Slices of all of the supported primitives:
&[u8]
,&[u16]
, etc. - Windows
FILETIME
. The type must be declared exactly asFILETIME
; type aliases or fully-qualified paths (such aswinapi::shared::minwindef::FILETIME
) will not work. The parameter type in the generated code will bewin_etw_provider::FILETIME
, which is a newtype overu64
. std::time::SystemTime
is supported, but it must be declared exactly asSystemTime
; type aliases or fully-qualified paths (such asstd::time::SystemTime
) will not work.SockAddr
,SockAddrV4
, andSockAddrV6
are supported. They must be declared exactly as shown, not using fully-qualified names or type aliases.
How to capture and view events
There are a variety of tools which can be used to capture and view ETW events.
The simplest tool is the TraceView
tool from the Windows SDK. Typically it is installed
at this path: C:\Program Files (x86)\Windows Kits\10\bin\10.0.<xxxxx>.0\x64\traceview.exe
,
where <xxxxx>
is the release number of the Windows SDK.
Run TraceView
, then select "File", then "Create New Log Session". Select "Manually Entered
GUID or Hashed Name" and enter the GUID that you have assigned to your event provider. Click OK.
The next dialog will prompt you to choose a source of WPP format information; select Auto
and click OK.
At this point, TraceView
should be capturing events (for your assigned GUID) and displaying
them in real time, regardless of which process reported the events.
These tools can also be used to capture ETW events:
- Windows Performance Recorder This tool is intended for capturing system-wide event streams. It is not useful for capturing events for a specific event provider.
- logman is a command-line tool for managing events.
- Tracelog
There are other tools, such as the Windows Performance Recorder, which can capture ETW events.
Ideas for improvement
- Better handling of per-event overrides, rather than using
Option<&EventOptions>
.
References
- Event Tracing for Windows (ETW) Simplified
- TraceLogging for Event Tracing for Windows (ETW)
- Record and View TraceLogging Events
- TraceLoggingOptionGroup
- Provider Traits
- TRACELOGGING_DEFINE_PROVIDER macro
Contributing
This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.
When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.