seccheck

package
v0.0.0-...-c27c9a0 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Dec 14, 2024 License: Apache-2.0, MIT Imports: 13 Imported by: 22

README

Introduction

This package provides a remote interface to observe behavior of the application running inside the sandbox. It was built with runtime monitoring in mind, e.g. threat detection, but it can be used for other purposes as well. It allows a process running outside the sandbox to receive a stream of trace data asynchronously. This process can watch actions performed by the application, generate alerts when something unexpected occurs, log these actions, etc.

First, let's go over a few concepts before we get into the details.

Concepts

  • Points: these are discrete places (or points) in the code where instrumentation was added. Each point has a unique name and schema. They can be individually enabled/disabled. For example, container/start is a point that is fired when a new container starts.
  • Point fields: each point may contain fields that carry point data. For example, container/start has a id field with the container ID that is getting started.
  • Optional fields: each point may also have optional fields. By default these fields are not collected and they can be manually set to be collected when the point is configured. These fields are normally more expensive to collect and/or large, e.g. resolve path to FD, or data for read/write.
  • Context fields: these are fields generally available to most events, but are disabled by default. Like optional fields, they can be set to be collected when the point is configured. Context field data comes from context where the point is being fired, for example PID, UID/GID, container ID are fields available to most trace points.
  • Sink: sinks are trace point consumers. Each sink is identified by a name and may handle trace points differently. Later we'll describe in more detailed what sinks are available in the system and how to use them.
  • Session: trace session is a set of points that are enabled with their corresponding configuration. A trace session also has a list of sinks that will receive the trace points. A session is identified by a unique name. Once a session is deleted, all points belonging to the session are disabled and the sinks destroyed.

If you're interested in exploring further, there are more details in the design doc.

Points

Every trance point in the system is identified by a unique name. The naming convention is to scope the point with a main component followed by its name to avoid conflicts. Here are a few examples:

  • sentry/signal_delivered
  • container/start
  • syscall/openat/enter

Note: the syscall trace point contains an extra level to separate the enter/exit points.

Most of the trace points are in the syscall component. They come in 2 flavors: raw, schematized. Raw syscalls include all syscalls in the system and contain the 6 arguments for the given syscall. Schematized trace points exist for many syscalls, but not all. They provide fields that are specific to the syscalls and fetch more information than is available from the raw syscall arguments. For example, here is the schema for the open syscall:

message Open {
  gvisor.common.ContextData context_data = 1;
  Exit exit = 2;
  uint64 sysno = 3;
  int64 fd = 4;
  string fd_path = 5;
  string pathname = 6;
  uint32 flags = 7;
  uint32 mode = 8;
}

As you can see, some fields are in both raw and schematized points, like fd which is also arg1 in the raw syscall, but here it has a name and correct type. It also has fields like pathname that are not available in the raw syscall event. In addition, fd_path is an optional field that can take the fd and translate it into a full path for convenience. In some cases, the same schema can be shared by many syscalls. In this example, message Open is used for open(2), openat(2) and creat(2) syscalls. The sysno field can be used to distinguish between them. The schema for all syscall trace points can be found here.

Other components that exist today are:

  • sentry: trace points fired from within gVisor's kernel (schema).
  • container: container related events (schema).

The following command lists all trace points available in the system:

$ runsc trace metadata
POINTS (973)
Name: container/start, optional fields: [env], context fields: [time|thread_id|task_start_time|group_id|thread_group_start_time|container_id|credentials|cwd|process_name]
Name: sentry/clone, optional fields: [], context fields: [time|thread_id|task_start_time|group_id|thread_group_start_time|container_id|credentials|cwd|process_name]
Name: syscall/accept/enter, optional fields: [fd_path], context fields: [time|thread_id|task_start_time|group_id|thread_group_start_time|container_id|credentials|cwd|process_name]
...

Note: the output format for trace metadata may change without notice.

The list above also includes what optional and context fields are available for each trace point. Optional fields schema is part of the trace point proto, like fd_path we saw above. Context fields are set in context_data field of all points and is defined in gvisor.common.ContextData.

Sinks

Sinks receive enabled trace points and do something useful with them. They are identified by a unique name. The same runsc trace metadata command used above also lists all sinks:

$ runsc trace metadata
...
SINKS (2)
Name: remote
Name: null

Note: the output format for trace metadata may change without notice.

Remote

The remote sink serializes the trace point into protobuf and sends it to a separate process. For threat detection, external monitoring processes can receive connections from remote sinks and be sent a stream of trace points that are occurring in the system. This sink connects to a remote process via Unix domain socket and expects the remote process to be listening for new connections. If you're interested in creating a monitoring process that communicates with the remote sink, this document has more details.

The remote sink has many properties that can be configured when it's created (more on how to configure sinks below):

  • endpoint (mandatory): Unix domain socket address to connect.
  • retries: number of attempts to write the trace point before dropping it in case the remote process is not responding. Note that a high number of retries can significantly delay application execution.
  • backoff: initial backoff time after the first failed attempt. This value doubles with every failed attempt, up to the max.
  • backoff_max: max duration to wait between retries.

Null

The null sink does nothing with the trace points and it's used for testing. Syscall tests enable all trace points, with all optional and context fields to ensure there is no crash with them enabled.

Strace (not implemented)

The strace sink has not been implemented yet. It's meant to replace the strace mechanism that exists in the Sentry to simplify the code and add more trace points to it.

Note: It requires more than one trace session to be supported.

Sessions

Trace sessions scope a set of trace points with their corresponding configuration and a set of sinks that receive the points. Sessions can be created at sandbox initialization time or during runtime. Creating sessions at init time guarantees that no trace points are missed, which is important for threat detection. It is configured using the --pod-init-config flag (more on it below). To manage sessions during runtime, runsc trace create|delete|list is used to manipulate trace sessions. Here are few examples assuming there is a running container with ID=cont123 using Docker:

$ sudo runsc --root /run/docker/runtime-runc/moby trace create --config session.json cont123
$ sudo runsc --root /run/docker/runtime-runc/moby trace list cont123
SESSIONS (1)
"Default"
        Sink: "remote", dropped: 0

$ sudo runsc --root /var/run/docker/runtime-runc/moby trace delete --name Default cont123
$ sudo runsc --root /var/run/docker/runtime-runc/moby trace list cont123
SESSIONS (0)

Note: There is a current limitation that only a single session can exist in the system and it must be called Default. This restriction can be lifted in the future when more than one session is needed.

Config

The event session can be defined using JSON for the runsc trace create command. The session definition has 3 main parts:

  1. name: name of the session being created. Only Default for now.
  2. points: array of points being enabled in the session. Each point has:
    1. name: name of trace point being enabled.
    2. optional_fields: array of optional fields to include with the trace point.
    3. context_fields: array of context fields to include with the trace point.
  3. sinks: array of sinks that will process the trace points.
    1. name: name of the sink.
    2. config: sink specific configuration.
    3. ignore_setup_error: ignores failure to configure the sink. In the remote sink case, for example, it doesn't fail container startup if the remote process cannot be reached.

The session configuration above can also be used with the --pod-init-config flag under the "trace_session" JSON object. There is a full example here

Note: For convenience, the --pod-init-config file can also be used with runsc trace create command. The portions of the Pod init config file that are not related to the session configuration are ignored.

Full Example

Here, we're going to explore a how to use runtime monitoring end to end. Under the examples directory there is an implementation of the monitoring process that accepts connections from remote sinks and prints out all the trace points it receives.

First, let's start the monitoring process and leave it running:

$ bazel run examples/seccheck:server_cc
Socket address /tmp/gvisor_events.sock

The server is now listening on the socket at /tmp/gvisor_events.sock for new gVisor sandboxes to connect. Now let's create a session configuration file with some trace points enabled and the remote sink using the socket address from above:

$ cat <<EOF >session.json
{
  "trace_session": {
    "name": "Default",
    "points": [
      {
        "name": "sentry/clone"
      },
      {
        "name": "syscall/fork/enter",
        "context_fields": [
          "group_id",
          "process_name"
        ]
      },
      {
        "name": "syscall/fork/exit",
        "context_fields": [
          "group_id",
          "process_name"
        ]
      },
      {
        "name": "syscall/execve/enter",
        "context_fields": [
          "group_id",
          "process_name"
        ]
      },
      {
        "name": "syscall/sysno/35/enter",
        "context_fields": [
          "group_id",
          "process_name"
        ]
      },
      {
        "name": "syscall/sysno/35/exit"
      }
    ],
    "sinks": [
      {
        "name": "remote",
        "config": {
          "endpoint": "/tmp/gvisor_events.sock"
        }
      }
    ]
  }
}
EOF

Now, we're ready to start a container and watch it send traces to the monitoring process. The container we're going to create simply loops every 5 seconds and writes something to stdout. While the container is running, we're going to call runsc trace command to create a trace session.

# Start the container and copy the container ID for future reference.
$ docker run --rm --runtime=runsc -d bash -c "while true; do echo looping; sleep 5; done"
dee0da1eafc6b15abffeed1abc6ca968c6d816252ae334435de6f3871fb05e61

$ CID=dee0da1eafc6b15abffeed1abc6ca968c6d816252ae334435de6f3871fb05e61

# Create new trace session in the container above.
$ sudo runsc --root /var/run/docker/runtime-runc/moby trace create --config session.json ${CID?}
Trace session "Default" created.

In the terminal where you are running the monitoring process, you'll start seeing messages like this:

Connection accepted
E Fork context_data      { thread_group_id: 1 process_name: "bash" } sysno: 57
CloneInfo => created_thread_id:      110 created_thread_group_id: 110 created_thread_start_time_ns: 1660249219204031676
X Fork context_data      { thread_group_id: 1 process_name: "bash" } exit { result: 110 } sysno: 57
E Execve context_data    { thread_group_id: 110 process_name: "bash" } sysno: 59 pathname: "/bin/sleep" argv: "sleep" argv: "5"
E Syscall context_data   { thread_group_id: 110 process_name: "sleep" } sysno: 35 arg1: 139785970818200 arg2: 139785970818200
X Syscall context_data   { thread_group_id: 110 process_name: "sleep" } exit { } sysno: 35 arg1: 139785970818200 arg2: 139785970818200

The first message in the log is a notification that a new sandbox connected to the monitoring process. The E and X in front of the syscall traces denotes whether the trace belongs to an Enter or eXit syscall trace. The first syscall trace shows a call to fork(2) from a process with group_thread_id (or PID) equal to 1 and the process name is bash. In other words, this is the init process of the container, running bash, and calling fork to execute sleep 5. The next trace is from sentry/clone and informs that the forked process has PID=110. Then, X Fork indicates that fork(2) syscall returned to the parent. The child continues and executes execve(2) to call sleep as can be seen from the pathname and argv fields. Note that at this moment, the PID is 110 (child) but the process name is still bash because it hasn't executed sleep yet. After execve(2) is called the process name changes to sleep as expected. Next, it shows the nanosleep(2) raw syscalls, which have sysno=35 (referred to as syscall/sysno/35 in the configuration file), one for enter with the exit trace happening 5 seconds later.

Let's list all trace sessions that are active in the sandbox:

$ sudo runsc --root /var/run/docker/runtime-runc/moby trace list ${CID?}
SESSIONS (1)
"Default"
        Sink: "remote", dropped: 0

It shows the Default session created above, using the remote sink and no trace points have been dropped. Once we're done, the trace session can be deleted with the command below:

$ sudo runsc --root /var/run/docker/runtime-runc/moby trace delete --name
Default ${CID?} Trace session "Default" deleted.

In the monitoring process you should see a message Connection closed to inform that the sandbox has disconnected.

If you want to set up runsc to connect to the monitoring process automatically before the application starts running, you can set the --pod-init-config flag to the configuration file created above. Here's an example:

$ sudo runsc --install --runtime=runsc-trace -- --pod-init-config=$PWD/session.json

Documentation

Overview

Package seccheck defines a structure for dynamically-configured security checks in the sentry.

Index

Constants

View Source 
const DefaultSessionName = "Default"

DefaultSessionName is the name of the only session that can exist in the system for now. When multiple sessions are supported, this can be removed.

View Source 
const (
	// FieldSyscallExecveEnvv is an optional field to collect list of environment
	// variables. Start after FieldSyscallPath because execveat(2) can collect
	// path from FD.
	FieldSyscallExecveEnvv = FieldSyscallPath + 1
)

Fields for execve*(2) syscalls.

Variables

View Source 
var Points = map[string]PointDesc{}

Points is a map with all the trace points registered in the system.

View Source 
var Sinks = map[string]SinkDesc{}

Sinks is a map with all the sinks registered in the system.

Functions

func Create 

func Create(conf *SessionConfig, force bool) error

Create reads the session configuration and applies it to the system.

func Delete 

func Delete(name string) error

Delete deletes an existing session.

func Initialize 

func Initialize()

Initialize initializes the Points available in the system. Must be called prior to using any of them.

func List 

func List(out *[]SessionConfig)

List lists all existing sessions.

func RegisterSink 

func RegisterSink(sink SinkDesc)

RegisterSink registers a new sink to make it discoverable.

func SetupSinks 

func SetupSinks(sinks []SinkConfig) ([]*os.File, error)

SetupSinks runs the setup step of all sinks in the configuration.

Types

type Field 

type Field uint

Field represents the index of a single optional field to be collect for a Point. 

const (
	FieldCtxtContainerID Field = iota
	FieldCtxtCredentials
	FieldCtxtCwd
	FieldCtxtProcessName
	FieldCtxtThreadGroupID
	FieldCtxtThreadGroupStartTime
	FieldCtxtThreadID
	FieldCtxtThreadStartTime
	FieldCtxtTime
)

FieldCtxtX represents a data field that comes from the Context. 

const (
	// FieldContainerStartEnv is an optional field to collect list of environment
	// variables set for the container start process.
	FieldContainerStartEnv Field = iota
)

Fields for container/start point. 

const (
	// FieldSentryExecveBinaryInfo is an optional field to collect information
	// about the binary being executed.
	FieldSentryExecveBinaryInfo Field = iota
)

Fields for sentry/execve point. 

const (
	// FieldSyscallPath is an optional field to collect path from an FD. Given
	// that many syscalls operate on FDs, this const is used across syscalls.
	FieldSyscallPath Field = iota
)

Fields that are common for many syscalls.

type FieldDesc 

type FieldDesc struct {
	// ID is the numeric identifier of the field.
	ID Field
	// Name is the unique field name.
	Name string
}

FieldDesc describes an optional/context field that is available to be configured.

type FieldMask 

type FieldMask struct {
	// contains filtered or unexported fields
}

FieldMask is a bitmask with a single bit representing an optional field to be collected. The meaning of each bit varies per point. The mask is currently limited to 64 fields. If more are needed, FieldMask can be expanded to support additional fields.

func MakeFieldMask 

func MakeFieldMask(fields ...Field) FieldMask

MakeFieldMask creates a FieldMask from a set of Fields.

func (*FieldMask) Add 

func (fm *FieldMask) Add(field Field)

Add adds a Field to the mask.

func (*FieldMask) Contains 

func (fm *FieldMask) Contains(field Field) bool

Contains returns true if the mask contains the Field.

func (*FieldMask) Empty 

func (fm *FieldMask) Empty() bool

Empty returns true if no bits are set.

func (*FieldMask) Remove 

func (fm *FieldMask) Remove(field Field)

Remove removes a Field from the mask.

type FieldSet 

type FieldSet struct {
	// Local indicates which optional fields from the Point that needs to be
	// collected, e.g. resolving path from an FD, or collecting a large field.
	Local FieldMask

	// Context indicates which optional fields from the Context that needs to be
	// collected, e.g. PID, credentials, current time.
	Context FieldMask
}

FieldSet contains all optional fields to be collected by a given Point.

type Point 

type Point uint

A Point represents a checkpoint, a point at which a security check occurs. 

const (
	PointClone Point = iota
	PointContainerStart
	PointExecve
	PointExitNotifyParent
	PointTaskExit
)

PointX represents the checkpoint X.

func GetPointForSyscall 

func GetPointForSyscall(typ SyscallType, sysno uintptr) Point

GetPointForSyscall translates the syscall number to the corresponding Point.

type PointConfig 

type PointConfig struct {
	// Name is the point to be enabled. The point must exist in the system.
	Name string `json:"name,omitempty"`
	// OptionalFields is the list of optional fields to collect from the point.
	OptionalFields []string `json:"optional_fields,omitempty"`
	// ContextFields is the list of context fields to collect.
	ContextFields []string `json:"context_fields,omitempty"`
}

PointConfig describes a point to be enabled in a given session.

type PointDesc 

type PointDesc struct {
	// ID is the point unique identifier.
	ID Point
	// Name is the point unique name. Convention is to use the following format:
	// namespace/name
	// Examples: container/start, sentry/clone, etc.
	Name string
	// OptionalFields is a list of fields that are available in the point, but not
	// collected unless specified when the Point is configured.
	// Examples: fd_path, data for read/write Points, etc.
	OptionalFields []FieldDesc
	// ContextFields is a list of fields that can be collected from the context,
	// but are not collected unless specified when the Point is configured.
	// Examples: container_id, PID, etc.
	ContextFields []FieldDesc
}

PointDesc describes a Point that is available to be configured. Schema for these points are defined in pkg/sentry/seccheck/points/.

type PointReq 

type PointReq struct {
	Pt     Point
	Fields FieldSet
}

PointReq indicates what Point a corresponding Sink runs at, and what information it requires at those Points.

type SessionConfig 

type SessionConfig struct {
	// Name is the unique session name.
	Name string `json:"name,omitempty"`
	// Points is the set of points to enable in this session.
	Points []PointConfig `json:"points,omitempty"`
	// IgnoreMissing skips point and optional/context fields not found. This can
	// be used to apply a single configuration file with newer points/fields with
	// older versions which do not have them yet. Note that it may hide typos in
	// the configuration.
	//
	// This field does NOT apply to sinks.
	IgnoreMissing bool `json:"ignore_missing,omitempty"`
	// Sinks are the sinks that will process the points enabled above.
	Sinks []SinkConfig `json:"sinks,omitempty"`
}

SessionConfig describes a new session configuration. A session consists of a set of points to be enabled and sinks where the points are sent to.

type Sink 

type Sink interface {
	// Name return the sink name.
	Name() string
	// Status returns the sink runtime status.
	Status() SinkStatus
	// Stop requests the sink to stop.
	Stop()

	Clone(ctx context.Context, fields FieldSet, info *pb.CloneInfo) error
	Execve(ctx context.Context, fields FieldSet, info *pb.ExecveInfo) error
	ExitNotifyParent(ctx context.Context, fields FieldSet, info *pb.ExitNotifyParentInfo) error
	TaskExit(context.Context, FieldSet, *pb.TaskExit) error

	ContainerStart(context.Context, FieldSet, *pb.Start) error

	Syscall(context.Context, FieldSet, *pb.ContextData, pb.MessageType, proto.Message) error
	RawSyscall(context.Context, FieldSet, *pb.Syscall) error
}

A Sink performs security checks at checkpoints.

Each Sink method X is called at checkpoint X; if the method may return a non-nil error and does so, it causes the checked operation to fail immediately (without calling subsequent Sinks) and return the error. The info argument contains information relevant to the check. The mask argument indicates what fields in info are valid; the mask should usually be a superset of fields requested by the Sink's corresponding PointReq, but may be missing requested fields in some cases (e.g. if the Sink is registered concurrently with invocations of checkpoints).

type SinkConfig 

type SinkConfig struct {
	// Name is the sink to be created. The sink must exist in the system.
	Name string `json:"name,omitempty"`
	// Config is a opaque json object that is passed to the sink.
	Config map[string]any `json:"config,omitempty"`
	// IgnoreSetupError makes errors during sink setup to be ignored. Otherwise,
	// failures will prevent the container from starting.
	IgnoreSetupError bool `json:"ignore_setup_error,omitempty"`
	// Status is the runtime status for the sink.
	Status SinkStatus `json:"status,omitempty"`
	// FD is the endpoint returned from Setup. It may be nil.
	FD *fd.FD `json:"-"`
}

SinkConfig describes the sink that will process the points in a given session.

type SinkDefaults 

type SinkDefaults struct{}

SinkDefaults may be embedded by implementations of Sink to obtain no-op implementations of Sink methods that may be explicitly overridden.

func (SinkDefaults) Clone 

func (SinkDefaults) Clone(context.Context, FieldSet, *pb.CloneInfo) error

Clone implements Sink.Clone.

func (SinkDefaults) ContainerStart 

func (SinkDefaults) ContainerStart(context.Context, FieldSet, *pb.Start) error

ContainerStart implements Sink.ContainerStart.

func (SinkDefaults) Execve 

func (SinkDefaults) Execve(context.Context, FieldSet, *pb.ExecveInfo) error

Execve implements Sink.Execve.

func (SinkDefaults) ExitNotifyParent 

func (SinkDefaults) ExitNotifyParent(context.Context, FieldSet, *pb.ExitNotifyParentInfo) error

ExitNotifyParent implements Sink.ExitNotifyParent.

func (SinkDefaults) RawSyscall 

func (SinkDefaults) RawSyscall(context.Context, FieldSet, *pb.Syscall) error

RawSyscall implements Sink.RawSyscall.

func (SinkDefaults) Status 

func (SinkDefaults) Status() SinkStatus

Status implements Sink.Status.

func (SinkDefaults) Stop 

func (SinkDefaults) Stop()

Stop implements Sink.Stop.

func (SinkDefaults) Syscall 

func (SinkDefaults) Syscall(context.Context, FieldSet, *pb.ContextData, pb.MessageType, proto.Message) error

Syscall implements Sink.Syscall.

func (SinkDefaults) TaskExit 

func (SinkDefaults) TaskExit(context.Context, FieldSet, *pb.TaskExit) error

TaskExit implements Sink.TaskExit.

type SinkDesc 

type SinkDesc struct {
	// Name is a unique identifier for the sink.
	Name string
	// Setup is called outside the protection of the sandbox. This is done to
	// allow the sink to do whatever is necessary to set it up. If it returns a
	// file, this file is donated to the sandbox and passed to the sink when New
	// is called. config is an opaque json object passed to the sink.
	Setup func(config map[string]any) (*os.File, error)
	// New creates a new sink. config is an opaque json object passed to the sink.
	// endpoing is a file descriptor to the file returned in Setup. It's set to -1
	// if Setup returned nil.
	New func(config map[string]any, endpoint *fd.FD) (Sink, error)
}

SinkDesc describes a sink that is available to be configured.

type SinkStatus 

type SinkStatus struct {
	// DroppedCount is the number of trace points dropped.
	DroppedCount uint64
}

SinkStatus represents stats about each Sink instance.

type State 

type State struct {
	// contains filtered or unexported fields
}

State is the type of global, and is separated out for testing. 

var Global State

Global is the method receiver of all seccheck functions.

func (*State) AddSyscallFlagListener 

func (s *State) AddSyscallFlagListener(listener SyscallFlagListener)

AddSyscallFlagListener adds a listener to the State.

The listener will be notified whenever syscall point enablement changes.

func (*State) AppendSink 

func (s *State) AppendSink(c Sink, reqs []PointReq)

AppendSink registers the given Sink to execute at checkpoints. The Sink will execute after all previously-registered sinks, and only if those Sinks return a nil error.

func (*State) Enabled 

func (s *State) Enabled(p Point) bool

Enabled returns true if any Sink is registered for the given checkpoint.

func (*State) GetFieldSet 

func (s *State) GetFieldSet(p Point) FieldSet

GetFieldSet returns the FieldSet that has been configured for a given Point.

func (*State) SentToSinks 

func (s *State) SentToSinks(fn func(c Sink) error) error

SentToSinks iterates over all sinks and calls fn for each one of them.

func (*State) SyscallEnabled 

func (s *State) SyscallEnabled(typ SyscallType, sysno uintptr) bool

SyscallEnabled checks if the corresponding point for the syscall is enabled.

type SyscallFlagListener 

type SyscallFlagListener interface {
	// UpdateSecCheck is called each time the system call point enablement may have changed.
	// This is called with seccheck.State.mu held, so it is expected to be fast and not re-entrant
	// with seccheck.State functions that attempt to re-lock it.
	UpdateSecCheck(state *State)
}

SyscallFlagListener is an interface that is notified when syscall point enablement changes.

It is used to notify the kernel's syscall table about syscall points, without introducing a direct dependency on it.

type SyscallType 

type SyscallType int

SyscallType is an enum that denotes different types of syscall points. There are 2 types of syscall point: fully-schematized and raw. Schematizes are points that have syscall specific format, e.g. open => {path, flags, mode}. Raw uses a generic schema that contains syscall number and 6 arguments. Each of these type have a corresponding enter and exit points. Exit points include return value and errno information. 

const (
	// SyscallEnter represents schematized/enter syscall.
	SyscallEnter SyscallType = iota
	// SyscallExit represents schematized/exit syscall.
	SyscallExit
	// SyscallRawEnter represents raw/enter syscall.
	SyscallRawEnter
	// SyscallRawExit represents raw/exit syscall.
	SyscallRawExit
)

Source Files

View all Source files

Directories

Path Synopsis
sinks
null
Package null defines a seccheck.Sink that does nothing with the trace points, akin to /dev/null.
 Package null defines a seccheck.Sink that does nothing with the trace points, akin to /dev/null.
remote
Package remote defines a seccheck.Sink that serializes points to a remote process.
 Package remote defines a seccheck.Sink that serializes points to a remote process.
remote/server
Package server provides a common server implementation that can connect with remote.Remote.
 Package server provides a common server implementation that can connect with remote.Remote.
remote/test
Package test provides functionality used to test the remote sink.
 Package test provides functionality used to test the remote sink.
remote/wire
Package wire defines structs used in the wire format for the remote checker.
 Package wire defines structs used in the wire format for the remote checker.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.