Merge remote-tracking branch 'origin/stable' into v2

example.config.toml

 # define a global timeout on establishing of network connections. idle
 # means a timeout on pumping data between sockset when nothing is
 # happening.
-#
-# please be noticed that handshakes have no timeouts intentionally. You can
-# find a reasoning here:
-# https://www.ndss-symposium.org/wp-content/uploads/2020/02/23087-paper.pdf
 [network.timeout]
 tcp = "5s"
 http = "10s"
-idle = "1m"
+idle = "5m"
+handshake = "10s"
+
+# this defines a configuration for TCP keep alives. Default values are taken
+# from Golang default behavior.
+[network.keep-alive]
+disabled = false
+# idle means a time period after which we start sending TCP Keep Alive probes
+idle = "15s"
+# interval is a period between 2 consecutive probes
+interval = "15s"
+# if we miss that many probes, a connection will be considered as a dead one.
+count = 9
 
 # mtg has to mimic real websites. It does not mean domain fronting, it also
 # means that traffic characteristics should be similar to real world traffic.

go.mod

 
 require (
 	github.com/OneOfOne/xxhash v1.2.8
-	github.com/alecthomas/kong v1.14.0
+	github.com/alecthomas/kong v1.15.0
 	github.com/alecthomas/units v0.0.0-20240927000941-0f3dac36c52b
 	github.com/armon/go-socks5 v0.0.0-20160902184237-e75332964ef5
 	github.com/babolivier/go-doh-client v0.0.0-20201028162107-a76cff4cb8b6
 
 require (
 	github.com/beevik/ntp v1.5.0
-	github.com/ncruces/go-dns v1.3.2
+	github.com/ncruces/go-dns v1.3.3
 	github.com/pelletier/go-toml/v2 v2.3.0
 	github.com/pires/go-proxyproto v0.11.0
 	github.com/things-go/go-socks5 v0.1.0

go.sum

 github.com/OneOfOne/xxhash v1.2.8/go.mod h1:eZbhyaAYD41SGSSsnmcpxVoRiQ/MPUTjUdIIOT9Um7Q=
 github.com/alecthomas/assert/v2 v2.11.0 h1:2Q9r3ki8+JYXvGsDyBXwH3LcJ+WK5D0gc5E8vS6K3D0=
 github.com/alecthomas/assert/v2 v2.11.0/go.mod h1:Bze95FyfUr7x34QZrjL+XP+0qgp/zg8yS+TtBj1WA3k=
-github.com/alecthomas/kong v1.14.0 h1:gFgEUZWu2ZmZ+UhyZ1bDhuutbKN1nTtJTwh19Wsn21s=
-github.com/alecthomas/kong v1.14.0/go.mod h1:wrlbXem1CWqUV5Vbmss5ISYhsVPkBb1Yo7YKJghju2I=
+github.com/alecthomas/kong v1.15.0 h1:BVJstKbpO73zKpmIu+m/aLRrNmWwxXPIGTNin9VmLVI=
+github.com/alecthomas/kong v1.15.0/go.mod h1:wrlbXem1CWqUV5Vbmss5ISYhsVPkBb1Yo7YKJghju2I=
 github.com/alecthomas/repr v0.5.2 h1:SU73FTI9D1P5UNtvseffFSGmdNci/O6RsqzeXJtP0Qs=
 github.com/alecthomas/repr v0.5.2/go.mod h1:Fr0507jx4eOXV7AlPV6AVZLYrLIuIeSOWtW57eE/O/4=
 github.com/alecthomas/units v0.0.0-20240927000941-0f3dac36c52b h1:mimo19zliBX/vSQ6PWWSL9lK8qwHozUj03+zLoEB8O0=
 github.com/miekg/dns v1.1.51/go.mod h1:2Z9d3CP1LQWihRZUf29mQ19yDThaI4DAYzte2CaQW5c=
 github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 h1:C3w9PqII01/Oq1c1nUAm88MOHcQC9l5mIlSMApZMrHA=
 github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822/go.mod h1:+n7T8mK8HuQTcFwEeznm/DIxMOiR9yIdICNftLE1DvQ=
-github.com/ncruces/go-dns v1.3.2 h1:kBLuUZBgkQ4qF4WDXZRQ4rG0Gk6sLVJQ5tESkWrxUa0=
-github.com/ncruces/go-dns v1.3.2/go.mod h1:tuzixNY8PY/M7yUzcvRbUaeLs3ifIdydpi5H2bfRU+s=
+github.com/ncruces/go-dns v1.3.3 h1:59OV7XoJrTCoUMZjWRVs4GOjtntMTZqiQ5Mn+BT13hk=
+github.com/ncruces/go-dns v1.3.3/go.mod h1:tuzixNY8PY/M7yUzcvRbUaeLs3ifIdydpi5H2bfRU+s=
 github.com/panjf2000/ants/v2 v2.12.0 h1:u9JhESo83i/GkZnhfTNuFMMWcNt7mnV1bGJ6FT4wXH8=
 github.com/panjf2000/ants/v2 v2.12.0/go.mod h1:tSQuaNQ6r6NRhPt+IZVUevvDyFMTs+eS4ztZc52uJTY=
 github.com/patrickmn/go-cache v2.1.0+incompatible h1:HRMgzkcYKYpi3C8ajMPV8OFXaaRUnok+kx1WdO15EQc=

internal/cli/doctor.go

 		conf.Network.Timeout.TCP.Get(10*time.Second),
 		conf.Network.Timeout.HTTP.Get(0),
 		conf.Network.Timeout.Idle.Get(0),
+		net.KeepAliveConfig{
+			Enable:   !conf.Network.KeepAlive.Disabled.Get(false),
+			Idle:     conf.Network.KeepAlive.Idle.Get(0),
+			Interval: conf.Network.KeepAlive.Interval.Get(0),
+			Count:    int(conf.Network.KeepAlive.Count.Get(0)),
+		},
 	)
 
 	fmt.Println("Validate native network connectivity")

internal/cli/run_proxy.go

 	"fmt"
 	"net"
 	"os"
-	"time"
 
 	"github.com/9seconds/mtg/v2/antireplay"
 	"github.com/9seconds/mtg/v2/events"
 		conf.Network.Timeout.TCP.Get(0),
 		conf.Network.Timeout.HTTP.Get(0),
 		conf.Network.Timeout.Idle.Get(0),
+		net.KeepAliveConfig{
+			Enable:   !conf.Network.KeepAlive.Disabled.Get(false),
+			Idle:     conf.Network.KeepAlive.Idle.Get(0),
+			Interval: conf.Network.KeepAlive.Interval.Get(0),
+			Count:    int(conf.Network.KeepAlive.Count.Get(0)),
+		},
 	)
 
 	proxyDialers := make([]mtglib.Network, len(conf.Network.Proxies))
 
 		AllowFallbackOnUnknownDC: conf.AllowFallbackOnUnknownDC.Get(false),
 		TolerateTimeSkewness:     conf.TolerateTimeSkewness.Value,
-		IdleTimeout:              conf.Network.Timeout.Idle.Get(time.Minute),
+		IdleTimeout:              conf.Network.Timeout.Idle.Get(mtglib.DefaultIdleTimeout),
+		HandshakeTimeout:         conf.Network.Timeout.Handshake.Get(mtglib.DefaultHandshakeTimeout),
 
 		DoppelGangerURLs:    doppelGangerURLs,
 		DoppelGangerPerRaid: conf.Defense.Doppelganger.Repeats.Get(mtglib.DoppelGangerPerRaid),

internal/config/config.go

 	} `json:"defense"`
 	Network struct {
 		Timeout struct {
-			TCP  TypeDuration `json:"tcp"`
-			HTTP TypeDuration `json:"http"`
-			Idle TypeDuration `json:"idle"`
+			TCP       TypeDuration `json:"tcp"`
+			HTTP      TypeDuration `json:"http"`
+			Idle      TypeDuration `json:"idle"`
+			Handshake TypeDuration `json:"handshake"`
 		} `json:"timeout"`
+		KeepAlive struct {
+			Disabled TypeBool        `json:"disabled"`
+			Idle     TypeDuration    `json:"idle"`
+			Interval TypeDuration    `json:"interval"`
+			Count    TypeConcurrency `json:"count"`
+		} `json:"keepAlive"`
 		DOHIP   TypeIP         `json:"dohIp"`
 		DNS     TypeDNSURI     `json:"dns"`
 		Proxies []TypeProxyURL `json:"proxies"`

internal/config/parse.go

 	} `toml:"defense" json:"defense,omitempty"`
 	Network struct {
 		Timeout struct {
-			TCP  string `toml:"tcp" json:"tcp,omitempty"`
-			HTTP string `toml:"http" json:"http,omitempty"`
-			Idle string `toml:"idle" json:"idle,omitempty"`
+			TCP       string `toml:"tcp" json:"tcp,omitempty"`
+			HTTP      string `toml:"http" json:"http,omitempty"`
+			Idle      string `toml:"idle" json:"idle,omitempty"`
+			Handshake string `toml:"handshake" json:"handshake,omitempty"`
 		} `toml:"timeout" json:"timeout,omitempty"`
+		KeepAlive struct {
+			Disabled bool   `toml:"disabled" json:"disabled,omitempty"`
+			Idle     string `toml:"idle" json:"idle,omitempty"`
+			Interval string `toml:"interval" json:"interval,omitempty"`
+			Count    uint   `toml:"count" json:"count,omitempty"`
+		} `toml:"keep-alive" json:"keepAlive,omitempty"`
 		DOHIP   string   `toml:"doh-ip" json:"dohIp,omitempty"`
 		DNS     string   `toml:"dns" json:"dns,omitempty"`
 		Proxies []string `toml:"proxies" json:"proxies,omitempty"`

mtglib/init.go

 	// DefaultIdleTimeout is a default timeout for closing a connection in case of
 	// idling.
 	//
-	// Deprecated: no longer in use because of changed TCP relay algorithm.
-	DefaultIdleTimeout = time.Minute
+	// Set to 5 minutes to survive typical mobile sleep periods (2-5 min) and
+	// avoid racing with MTProto ping_delay_disconnect (~60s interval).
+	DefaultIdleTimeout = 5 * time.Minute
+
+	// DefaultHandshakeTimeout defines a time period during which the
+	// all handshake ceremonies must be completed.
+	DefaultHandshakeTimeout = 10 * time.Second
 
 	// DefaultTolerateTimeSkewness is a default timeout for time skewness on a
 	// faketls timeout verification.

mtglib/internal/tls/fake/client_side.go

 	"crypto/sha256"
 	"crypto/subtle"
 	"encoding/binary"
+	"errors"
 	"fmt"
 	"io"
 	"net"
 	// record_type(1) + version(2) + size(2) + handshake_type(1) + uint24_length(3) + client_version(2)
 	RandomOffset = 1 + 2 + 2 + 1 + 3 + 2
 
+	// https://datatracker.ietf.org/doc/html/rfc8701#name-grease-values
+	// https://medium.com/asecuritysite-when-bob-met-alice/in-cybersecurity-what-is-grease-9f8850558dea
+	GreaseMask      = 0x0f0f
+	GreaseValueType = 0x0a0a
+
 	sniDNSNamesListType = 0
 )
 
 var (
 	emptyRandom = [RandomLen]byte{}
 	extTypeSNI  = [2]byte{}
+
+	ErrCannotFindCipher = errors.New("cannot find a cipher")
 )
 
 type ClientHello struct {
 	hostname string,
 	tolerateTimeSkewness time.Duration,
 ) (*ClientHello, error) {
-	if err := conn.SetReadDeadline(time.Now().Add(ClientHelloReadTimeout)); err != nil {
-		return nil, fmt.Errorf("cannot set read deadline: %w", err)
-	}
-	defer conn.SetReadDeadline(resetDeadline) //nolint: errcheck
-
 	// This is how FakeTLS is organized:
 	//  1. We create sha256 HMAC with a given secret
 	//  2. We dump there a whole TLS frame except of the fact that random
 
 	cipherSuiteLen := int64(binary.BigEndian.Uint16(header[:]))
 
-	// we do not care about picking up any cipher. we pick the first one,
-	// so it is always should be present.
-	if _, err := io.ReadFull(r, header[:]); err != nil {
-		return nil, fmt.Errorf("cannot read first cipher suite: %w", err)
-	}
+	// Pick the first non-GREASE cipher suite from the list.
+	// Real TLS servers never select GREASE values (RFC 8701, pattern 0x?a?a),
+	// so echoing them back is a trivial DPI fingerprint.
+	// cipherSuiteLen is in bytes; each cipher suite is 2 bytes.
+	for range cipherSuiteLen / 2 {
+		if _, err := io.ReadFull(r, header[:]); err != nil {
+			return nil, fmt.Errorf("cannot read cipher suite: %w", err)
+		}
+
+		if hello.CipherSuite != 0 {
+			// do not forget we have to scan until the end
+			continue
+		}
 
-	hello.CipherSuite = binary.BigEndian.Uint16(header[:])
+		if cs := binary.BigEndian.Uint16(header[:]); cs&GreaseMask != GreaseValueType {
+			hello.CipherSuite = cs
+		}
+	}
 
-	if _, err := io.CopyN(io.Discard, r, cipherSuiteLen-2); err != nil {
-		return nil, fmt.Errorf("cannot skip remaining cipher suites: %w", err)
+	if hello.CipherSuite == 0 {
+		return nil, ErrCannotFindCipher
 	}
 
 	if _, err := io.ReadFull(r, header[:1]); err != nil {

mtglib/internal/tls/fake/client_side_snapshot_test.go

 	"github.com/9seconds/mtg/v2/mtglib"
 	"github.com/9seconds/mtg/v2/mtglib/internal/tls/fake"
 	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/mock"
 	"github.com/stretchr/testify/require"
 	"github.com/stretchr/testify/suite"
 )
 		readBuf: readBuf,
 	}
 
-	connMock.
-		On("SetReadDeadline", mock.AnythingOfType("time.Time")).
-		Twice().
-		Return(nil)
-
 	return connMock
 }
 

mtglib/internal/tls/fake/client_side_test.go

 
 import (
 	"bytes"
+	cryptotls "crypto/tls"
 	"encoding/binary"
 	"encoding/json"
-	"errors"
 	"io"
 	"os"
 	"testing"
 	"github.com/9seconds/mtg/v2/mtglib"
 	"github.com/9seconds/mtg/v2/mtglib/internal/tls"
 	"github.com/9seconds/mtg/v2/mtglib/internal/tls/fake"
-	"github.com/stretchr/testify/mock"
 	"github.com/stretchr/testify/require"
 	"github.com/stretchr/testify/suite"
 )
 	suite.connMock = &parseClientHelloConnMock{
 		readBuf: suite.readBuf,
 	}
-
-	suite.connMock.
-		On("SetReadDeadline", mock.AnythingOfType("time.Time")).
-		Twice().
-		Return(nil)
 }
 
 func (suite *ParseClientHelloTestSuite) TearDownTest() {
 }
 
 func (suite *ParseClientHello_TLSHeaderTestSuite) TestEmpty() {
-	suite.connMock.ExpectedCalls = []*mock.Call{}
-	suite.connMock.
-		On("SetReadDeadline", mock.AnythingOfType("time.Time")).
-		Once().
-		Return(errors.New("fail"))
-
 	_, err := fake.ReadClientHello(suite.connMock, suite.secret.Key[:], suite.secret.Host, TolerateTime)
-	suite.ErrorContains(err, "fail")
+	suite.ErrorContains(err, "cannot read client hello")
 }
 
 func (suite *ParseClientHello_TLSHeaderTestSuite) TestNothing() {
-	suite.connMock.ExpectedCalls = []*mock.Call{}
-	suite.connMock.
-		On("SetReadDeadline", mock.AnythingOfType("time.Time")).
-		Twice().
-		Return(nil)
-
 	_, err := fake.ReadClientHello(suite.connMock, suite.secret.Key[:], suite.secret.Host, TolerateTime)
 	suite.ErrorIs(err, io.EOF)
 }
 }
 
 func (suite *ParseClientHelloHandshakeBodyTestSuite) TestCannotReadFirstCipherSuite() {
-	body := make([]byte, 2+fake.RandomLen+1+2)
+	body := make([]byte, 2+fake.RandomLen+1+2+1) // cipherSuiteLen=2 but only 1 byte available
+	binary.BigEndian.PutUint16(body[2+fake.RandomLen+1:], 2)
 
 	suite.writeBody(body)
 
 	_, err := fake.ReadClientHello(suite.connMock, suite.secret.Key[:], suite.secret.Host, TolerateTime)
-	suite.ErrorContains(err, "cannot read first cipher suite")
+	suite.ErrorContains(err, "cannot read cipher suite")
 }
 
 func (suite *ParseClientHelloHandshakeBodyTestSuite) TestCannotSkipRemainingCipherSuites() {
 	suite.writeBody(body)
 
 	_, err := fake.ReadClientHello(suite.connMock, suite.secret.Key[:], suite.secret.Host, TolerateTime)
-	suite.ErrorContains(err, "cannot skip remaining cipher suites")
+	suite.ErrorContains(err, "cannot read cipher suite")
+}
+
+func (suite *ParseClientHelloHandshakeBodyTestSuite) TestCannotFindCipher() {
+	// All cipher suites are GREASE values — must return ErrCannotFindCipher.
+	body := make([]byte, 2+fake.RandomLen+1+2+4+1)
+	binary.BigEndian.PutUint16(body[2+fake.RandomLen+1:], 4)
+	binary.BigEndian.PutUint16(body[2+fake.RandomLen+1+2:], 0x0a0a)
+	binary.BigEndian.PutUint16(body[2+fake.RandomLen+1+2+2:], 0x1a1a)
+	body[2+fake.RandomLen+1+2+4] = 1
+
+	suite.writeBody(body)
+
+	_, err := fake.ReadClientHello(suite.connMock, suite.secret.Key[:], suite.secret.Host, TolerateTime)
+	suite.ErrorIs(err, fake.ErrCannotFindCipher)
 }
 
 func (suite *ParseClientHelloHandshakeBodyTestSuite) TestCannotReadCompressionMethodsLength() {
 	body := make([]byte, 2+fake.RandomLen+1+2+2)
 	binary.BigEndian.PutUint16(body[2+fake.RandomLen+1:], 2)
+	binary.BigEndian.PutUint16(body[2+fake.RandomLen+1+2:], cryptotls.TLS_AES_128_GCM_SHA256)
 
 	suite.writeBody(body)
 
 func (suite *ParseClientHelloHandshakeBodyTestSuite) TestCannotSkipCompressionMethods() {
 	body := make([]byte, 2+fake.RandomLen+1+2+2+1)
 	binary.BigEndian.PutUint16(body[2+fake.RandomLen+1:], 2)
+	binary.BigEndian.PutUint16(body[2+fake.RandomLen+1+2:], cryptotls.TLS_AES_128_GCM_SHA256)
 	body[2+fake.RandomLen+1+2+2] = 1
 
 	suite.writeBody(body)
 	// cipherSuite(2) + compressionLen(1) + compression(1) = 41
 	body := make([]byte, 41)
 	binary.BigEndian.PutUint16(body[35:], 2)
+	binary.BigEndian.PutUint16(body[37:], cryptotls.TLS_AES_128_GCM_SHA256)
 	body[39] = 1
 
 	suite.readBuf.Write(body)
 		readBuf: readBuf,
 	}
 
-	connMock.
-		On("SetReadDeadline", mock.AnythingOfType("time.Time")).
-		Twice().
-		Return(nil)
-
 	return connMock
 }
 

mtglib/internal/tls/fake/init.go

 
 import (
 	"errors"
-	"time"
 )
 
-const (
-	ClientHelloReadTimeout = 5 * time.Second
-)
-
-var (
-	resetDeadline time.Time
-
-	ErrBadDigest = errors.New("incorrect client random")
-)
+var ErrBadDigest = errors.New("incorrect client random")

mtglib/internal/tls/fake/server_side_test.go

 	recordType, length, err := tls.ReadRecord(suite.buf, &rec)
 	suite.NoError(err)
 	suite.Equal(byte(tls.TypeApplicationData), recordType)
-	suite.Greater(length, int64(2500))
+	suite.GreaterOrEqual(length, int64(2500))
 
 	suite.Empty(suite.buf.Bytes())
 }

mtglib/internal/tls/fake/testdata/client-hello-ok-grease-first.json

+{
+  "time": 1617181365,
+  "random": "w4TaDfYg/aUKdx1oi68vxMKvHJczRNvtRRppLETzeNE=",
+  "sessionId": "St2BZ2uHMFn3B2trD1jfdtpjoJOOg6JBeLhFcyCMCq4=",
+  "host": "storage.googleapis.com",
+  "cipherSuite": 4867,
+  "full": "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"
+}

mtglib/proxy.go

 	allowFallbackOnUnknownDC    bool
 	tolerateTimeSkewness        time.Duration
 	idleTimeout                 time.Duration
+	handshakeTimeout            time.Duration
 	domainFrontingPort          int
 	domainFrontingIP            string
 	domainFrontingProxyProtocol bool
 	ctx := newStreamContext(p.ctx, p.logger, conn)
 	defer ctx.Close()
 
+	if err := ctx.clientConn.SetDeadline(time.Now().Add(p.handshakeTimeout)); err != nil {
+		ctx.logger.WarningError("cannot set handshake timeout", err)
+		return
+	}
+
 	stop := context.AfterFunc(ctx, func() {
 		ctx.Close()
 	})
 		return
 	}
 
+	if err := ctx.clientConn.SetDeadline(time.Time{}); err != nil {
+		ctx.logger.WarningError("cannot set deadline", err)
+		return
+	}
+
 	if err := p.doTelegramCall(ctx); err != nil {
 		ctx.logger.WarningError("cannot dial to telegram", err)
 		return
 		domainFrontingIP:         opts.DomainFrontingIP,
 		tolerateTimeSkewness:     opts.getTolerateTimeSkewness(),
 		idleTimeout:              opts.getIdleTimeout(),
+		handshakeTimeout:         opts.getHandshakeTimeout(),
 		allowFallbackOnUnknownDC: opts.AllowFallbackOnUnknownDC,
 		telegram:                 tg,
 		doppelGanger: doppel.NewGanger(

mtglib/proxy_opts.go

 	// This is an optional setting.
 	IdleTimeout time.Duration
 
+	// HandshakeTimeout is a timeout during which all handshake ceremonies must
+	// be completed, otherwise this process will be aborted
+	//
+	// This is an optional setting.
+	HandshakeTimeout time.Duration
+
 	// TolerateTimeSkewness is a time boundary that defines a time range where
 	// faketls timestamp is acceptable.
 	//
 	return p.PreferIP
 }
 
+func (p ProxyOpts) getHandshakeTimeout() time.Duration {
+	if p.HandshakeTimeout == 0 {
+		return DefaultHandshakeTimeout
+	}
+
+	return p.HandshakeTimeout
+}
+
 func (p ProxyOpts) getIdleTimeout() time.Duration {
 	if p.IdleTimeout == 0 {
-		return time.Minute
+		return DefaultIdleTimeout
 	}
 
 	return p.IdleTimeout

network/init.go

 
 	// DefaultTCPKeepAlivePeriod defines a time period between 2 consequitive
 	// probes.
+	//
+	// Deprecated: use DefaultKeepAliveIdle and DefaultKeepAliveInterval instead.
 	DefaultTCPKeepAlivePeriod = 10 * time.Second
 
+	// DefaultKeepAliveIdle is the time a connection must be idle before
+	// the first keepalive probe is sent.
+	DefaultKeepAliveIdle = 30 * time.Second
+
+	// DefaultKeepAliveInterval is the time between consecutive keepalive
+	// probes.
+	DefaultKeepAliveInterval = 10 * time.Second
+
+	// DefaultKeepAliveCount is the number of unacknowledged probes before
+	// the connection is considered dead.
+	DefaultKeepAliveCount = 3
+
 	// ProxyDialerOpenThreshold is used for load balancing SOCKS5 dialer only.
 	//
 	// This dialer uses circuit breaker with of 3 stages: OPEN, HALF_OPEN and

network/sockopts.go

 }
 
 func setCommonSocketOptions(conn *net.TCPConn) error {
-	if err := conn.SetKeepAlivePeriod(DefaultTCPKeepAlivePeriod); err != nil {
-		return fmt.Errorf("cannot set time period of TCP keepalive probes: %w", err)
+	if err := conn.SetKeepAliveConfig(net.KeepAliveConfig{
+		Enable:   true,
+		Idle:     DefaultKeepAliveIdle,
+		Interval: DefaultKeepAliveInterval,
+		Count:    DefaultKeepAliveCount,
+	}); err != nil {
+		return fmt.Errorf("cannot configure TCP keepalive: %w", err)
 	}
 
 	if err := conn.SetLinger(tcpLingerTimeout); err != nil {

network/sockopts_test.go

+//go:build linux || darwin
+// +build linux darwin
+
+package network_test
+
+import (
+	"net"
+	"runtime"
+	"syscall"
+	"testing"
+	"time"
+
+	"github.com/9seconds/mtg/v2/network"
+	"github.com/stretchr/testify/require"
+	"golang.org/x/sys/unix"
+)
+
+func tcpKeepIdleOption() int {
+	if runtime.GOOS == "darwin" {
+		return 0x10 // TCP_KEEPALIVE on macOS
+	}
+
+	return 0x4 // TCP_KEEPIDLE on Linux
+}
+
+func TestSetClientSocketOptionsKeepAlive(t *testing.T) {
+	t.Parallel()
+
+	listener, err := net.Listen("tcp", "127.0.0.1:0")
+	require.NoError(t, err)
+	defer func() {
+		err := listener.Close()
+		require.NoError(t, err)
+	}()
+
+	type dialResult struct {
+		conn net.Conn
+		err  error
+	}
+
+	dialDone := make(chan dialResult, 1)
+
+	go func() {
+		c, err := net.Dial("tcp", listener.Addr().String())
+		dialDone <- dialResult{conn: c, err: err}
+	}()
+
+	tcpListener, ok := listener.(*net.TCPListener)
+	require.True(t, ok, "listener must be a *net.TCPListener")
+
+	require.NoError(t, tcpListener.SetDeadline(time.Now().Add(5*time.Second)))
+
+	accepted, err := listener.Accept()
+	require.NoError(t, err)
+	defer func() {
+		err := accepted.Close()
+		require.NoError(t, err)
+	}()
+
+	dr := <-dialDone
+	require.NoError(t, dr.err)
+	defer func() {
+		err := dr.conn.Close()
+		require.NoError(t, err)
+	}()
+
+	err = network.SetClientSocketOptions(accepted, 0)
+	require.NoError(t, err)
+
+	tcpConn := accepted.(*net.TCPConn)
+
+	rawConn, err := tcpConn.SyscallConn()
+	require.NoError(t, err)
+
+	err = rawConn.Control(func(fd uintptr) {
+		val, err := unix.GetsockoptInt(int(fd), syscall.SOL_SOCKET, syscall.SO_KEEPALIVE)
+		require.NoError(t, err)
+		require.NotEqual(t, 0, val, "SO_KEEPALIVE should be enabled")
+
+		idle, err := unix.GetsockoptInt(int(fd), syscall.IPPROTO_TCP, tcpKeepIdleOption())
+		require.NoError(t, err)
+		require.Equal(t, int(network.DefaultKeepAliveIdle.Seconds()), idle)
+
+		interval, err := unix.GetsockoptInt(int(fd), syscall.IPPROTO_TCP, unix.TCP_KEEPINTVL)
+		require.NoError(t, err)
+		require.Equal(t, int(network.DefaultKeepAliveInterval.Seconds()), interval)
+
+		count, err := unix.GetsockoptInt(int(fd), syscall.IPPROTO_TCP, unix.TCP_KEEPCNT)
+		require.NoError(t, err)
+		require.Equal(t, network.DefaultKeepAliveCount, count)
+	})
+	require.NoError(t, err)
+}

network/v2/base_http_test.go

 }
 
 func (suite *BaseHTTPTestSuite) SetupTest() {
-	suite.client = network.New(nil, "mtg/1", 0, 0, 0).MakeHTTPClient(nil)
+	suite.client = network.New(nil, "mtg/1", 0, 0, 0, network.DefaultKeepAliveConfig).MakeHTTPClient(nil)
 }
 
 func (suite *BaseHTTPTestSuite) TestGet() {

network/v2/base_network_test.go

 func (suite *BaseNetworkTestSuite) SetupSuite() {
 	suite.EchoServerTestSuite.SetupSuite()
 
-	suite.net = network.New(nil, "agent", 0, 0, 0)
+	suite.net = network.New(nil, "agent", 0, 0, 0, network.DefaultKeepAliveConfig)
 }
 
 func (suite *BaseNetworkTestSuite) TestDialUnknownNetwork() {

network/v2/init.go

 
 import (
 	"errors"
+	"net"
 	"time"
 )
 
 
 	// DefaultTCPKeepAlivePeriod defines a time period between 2 consecuitive
 	// probes.
+	//
+	// Deprecated: use DefaultKeepAliveConfig
 	DefaultTCPKeepAlivePeriod = 10 * time.Second
 
+	// DefaultKeepAliveIdle is the time a connection must be idle before
+	// the first keepalive probe is sent.
+	//
+	// Deprecated: use DefaultKeepAliveConfig
+	DefaultKeepAliveIdle = 30 * time.Second
+
+	// DefaultKeepAliveInterval is the time between consecutive keepalive
+	// probes.
+	//
+	// Deprecated: use DefaultKeepAliveConfig
+	DefaultKeepAliveInterval = 10 * time.Second
+
+	// DefaultKeepAliveCount is the number of unacknowledged probes before
+	// the connection is considered dead.
+	//
+	// Deprecated: use DefaultKeepAliveConfig
+	DefaultKeepAliveCount = 3
+
 	// User Agent to use in HTTP client.
 	UserAgent = "curl/8.5.0"
 
 	// tcpLingerTimeout defines a number of seconds to wait for sending
 	// unacknowledged data.
 	tcpLingerTimeout = 1
+
+	// tcpNotSentLowat limits the amount of unsent data queued in the
+	// kernel write buffer per socket. When the unsent data drops below
+	// this threshold, the socket becomes writable again. This reduces
+	// per-connection memory usage and bufferbloat by applying
+	// back-pressure to the relay loop instead of piling up data in
+	// kernel buffers.
+	tcpNotSentLowat = 128 * 1024
 )
 
-var ErrCannotDial = errors.New("cannot dial to any address")
+var (
+	ErrCannotDial = errors.New("cannot dial to any address")
+
+	// DefaultKeepAliveConfig defines a default configuration for
+	// keep alive settings. As per official documentation, if keep alive
+	// is enabled, then:
+	//
+	//  Idle = 15 * time.Second
+	//  Interval = 15 * time.Second
+	//  Count = 9
+	DefaultKeepAliveConfig = net.KeepAliveConfig{
+		Enable: true,
+	}
+)

network/v2/network.go

 type network struct {
 	net.Dialer
 
-	httpTimeout time.Duration
-	idleTimeout time.Duration
-	userAgent   string
+	keepAliveConfig net.KeepAliveConfig
+	httpTimeout     time.Duration
+	idleTimeout     time.Duration
+	userAgent       string
 }
 
 func (n *network) Dial(network, address string) (essentials.Conn, error) {
 
 	tcpConn := conn.(*net.TCPConn)
 
-	return tcpConn, setCommonSocketOptions(tcpConn)
+	return tcpConn, setCommonSocketOptions(tcpConn, n.keepAliveConfig)
 }
 
 func (n *network) MakeHTTPClient(
 	tcpTimeout,
 	httpTimeout,
 	idleTimeout time.Duration,
+	keepAliveConfig net.KeepAliveConfig,
 ) mtglib.Network {
 	if dnsResolver == nil {
 		dnsResolver = net.DefaultResolver
 			Resolver:      dnsResolver,
 			FallbackDelay: -1,
 		},
-		userAgent:   userAgent,
-		idleTimeout: idleTimeout,
-		httpTimeout: httpTimeout,
+		userAgent:       userAgent,
+		idleTimeout:     idleTimeout,
+		httpTimeout:     httpTimeout,
+		keepAliveConfig: keepAliveConfig,
 	}
 }

network/v2/sockopts.go

 	"net"
 )
 
-func setCommonSocketOptions(conn *net.TCPConn) error {
-	if err := conn.SetKeepAlivePeriod(DefaultTCPKeepAlivePeriod); err != nil {
-		return fmt.Errorf("cannot set time period of TCP keepalive probes: %w", err)
+func setCommonSocketOptions(conn *net.TCPConn, keepAliveConfig net.KeepAliveConfig) error {
+	if err := conn.SetKeepAliveConfig(keepAliveConfig); err != nil {
+		return fmt.Errorf("cannot configure TCP keepalive: %w", err)
 	}
 
 	if err := conn.SetLinger(tcpLingerTimeout); err != nil {
 		return fmt.Errorf("cannot setup SO_REUSEADDR/PORT: %w", err)
 	}
 
+	setCongestionControl(rawConn)
+	setTCPUserTimeout(rawConn, keepAliveConfig)
+	setNotSentLowat(rawConn)
+
 	return nil
 }

network/v2/sockopts_congestion.go

+//go:build linux
+
+package network
+
+import (
+	"syscall"
+
+	"golang.org/x/sys/unix"
+)
+
+// setCongestionControl sets BBR as the TCP congestion control algorithm.
+// BBR provides better throughput over lossy and high-latency links compared
+// to the default cubic, which is especially beneficial for mobile and
+// home internet clients. This is best-effort: silently ignored if the
+// kernel does not have tcp_bbr available.
+func setCongestionControl(conn syscall.RawConn) {
+	conn.Control(func(fd uintptr) { //nolint: errcheck
+		unix.SetsockoptString(int(fd), unix.IPPROTO_TCP, unix.TCP_CONGESTION, "bbr") //nolint: errcheck
+	})
+}

network/v2/sockopts_congestion_stub.go

+//go:build !linux
+
+package network
+
+import "syscall"
+
+func setCongestionControl(conn syscall.RawConn) {}

network/v2/sockopts_lowat.go

+//go:build linux || darwin
+
+package network
+
+import (
+	"syscall"
+
+	"golang.org/x/sys/unix"
+)
+
+// setNotSentLowat sets TCP_NOTSENT_LOWAT which limits the amount of
+// unsent data queued in the kernel write buffer. Once unsent data drops
+// below this threshold the socket becomes writable again, applying
+// back-pressure to the relay loop instead of piling up data in kernel
+// buffers. This reduces per-connection memory and bufferbloat.
+func setNotSentLowat(conn syscall.RawConn) {
+	conn.Control(func(fd uintptr) { //nolint: errcheck
+		unix.SetsockoptInt(int(fd), unix.IPPROTO_TCP, unix.TCP_NOTSENT_LOWAT, tcpNotSentLowat) //nolint: errcheck
+	})
+}

network/v2/sockopts_lowat_stub.go

+//go:build !linux && !darwin
+
+package network
+
+import "syscall"
+
+func setNotSentLowat(conn syscall.RawConn) {}

network/v2/sockopts_unix.go → network/v2/sockopts_reuseaddr.go

 //go:build !windows
-// +build !windows
 
 package network
 

network/v2/sockopts_windows.go → network/v2/sockopts_reuseaddr_stub.go

 //go:build windows
-// +build windows
 
 package network
 

network/v2/sockopts_test.go

+//go:build linux || darwin
+// +build linux darwin
+
+package network
+
+import (
+	"net"
+	"runtime"
+	"syscall"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+	"golang.org/x/sys/unix"
+)
+
+func tcpKeepIdleOption() int {
+	if runtime.GOOS == "darwin" {
+		return 0x10 // TCP_KEEPALIVE on macOS
+	}
+
+	return 0x4 // TCP_KEEPIDLE on Linux
+}
+
+func TestSetCommonSocketOptionsKeepAlive(t *testing.T) {
+	t.Parallel()
+
+	listener, err := net.Listen("tcp", "127.0.0.1:0")
+	require.NoError(t, err)
+	defer func() {
+		err := listener.Close()
+		require.NoError(t, err)
+	}()
+
+	type dialResult struct {
+		conn net.Conn
+		err  error
+	}
+
+	dialDone := make(chan dialResult, 1)
+
+	go func() {
+		c, err := net.Dial("tcp", listener.Addr().String())
+		dialDone <- dialResult{conn: c, err: err}
+	}()
+
+	tcpListener, ok := listener.(*net.TCPListener)
+	require.True(t, ok, "listener must be a *net.TCPListener")
+
+	require.NoError(t, tcpListener.SetDeadline(time.Now().Add(5*time.Second)))
+
+	accepted, err := listener.Accept()
+	require.NoError(t, err)
+	defer func() {
+		err := accepted.Close()
+		require.NoError(t, err)
+	}()
+
+	dr := <-dialDone
+	require.NoError(t, dr.err)
+	defer func() {
+		err := dr.conn.Close()
+		require.NoError(t, err)
+	}()
+
+	tcpConn := accepted.(*net.TCPConn)
+
+	err = setCommonSocketOptions(tcpConn, DefaultKeepAliveConfig)
+	require.NoError(t, err)
+
+	rawConn, err := tcpConn.SyscallConn()
+	require.NoError(t, err)
+
+	err = rawConn.Control(func(fd uintptr) {
+		val, err := unix.GetsockoptInt(int(fd), syscall.SOL_SOCKET, syscall.SO_KEEPALIVE)
+		require.NoError(t, err)
+		require.NotEqual(t, 0, val, "SO_KEEPALIVE should be enabled")
+
+		idle, err := unix.GetsockoptInt(int(fd), syscall.IPPROTO_TCP, tcpKeepIdleOption())
+		require.NoError(t, err)
+		require.Equal(t, 15, idle, "keepalive idle should match DefaultKeepAliveIdle")
+
+		interval, err := unix.GetsockoptInt(int(fd), syscall.IPPROTO_TCP, unix.TCP_KEEPINTVL)
+		require.NoError(t, err)
+		require.Equal(t, 15, interval, "keepalive interval should match DefaultKeepAliveInterval")
+
+		count, err := unix.GetsockoptInt(int(fd), syscall.IPPROTO_TCP, unix.TCP_KEEPCNT)
+		require.NoError(t, err)
+		require.Equal(t, 9, count, "keepalive count should match DefaultKeepAliveCount")
+	})
+	require.NoError(t, err)
+}

network/v2/sockopts_usertimeout.go

+//go:build linux
+
+package network
+
+import (
+	"net"
+	"syscall"
+	"time"
+
+	"golang.org/x/sys/unix"
+)
+
+// Go runtime defaults for KeepAliveConfig when fields are zero.
+const (
+	goDefaultKeepAliveIdle     = 15 * time.Second
+	goDefaultKeepAliveInterval = 15 * time.Second
+	goDefaultKeepAliveCount    = 9
+)
+
+// setTCPUserTimeout sets TCP_USER_TIMEOUT on a socket. If transmitted
+// data remains unacknowledged for this long, the kernel closes the
+// connection. As recommended by Cloudflare
+// (https://blog.cloudflare.com/when-tcp-sockets-refuse-to-die/),
+// the value is computed as: keepidle + keepintvl * keepcnt. This
+// ensures TCP_USER_TIMEOUT and keepalives agree on when to give up.
+// Best-effort: silently ignored if unsupported.
+func setTCPUserTimeout(conn syscall.RawConn, cfg net.KeepAliveConfig) {
+	idle := cfg.Idle
+	if idle == 0 {
+		idle = goDefaultKeepAliveIdle
+	}
+
+	interval := cfg.Interval
+	if interval == 0 {
+		interval = goDefaultKeepAliveInterval
+	}
+
+	count := cfg.Count
+	if count == 0 {
+		count = goDefaultKeepAliveCount
+	}
+
+	timeout := idle + interval*time.Duration(count)
+
+	conn.Control(func(fd uintptr) { //nolint: errcheck
+		unix.SetsockoptInt(int(fd), unix.IPPROTO_TCP, unix.TCP_USER_TIMEOUT, int(timeout.Milliseconds())) //nolint: errcheck
+	})
+}

network/v2/sockopts_usertimeout_stub.go

+//go:build !linux
+
+package network
+
+import (
+	"net"
+	"syscall"
+)
+
+func setTCPUserTimeout(conn syscall.RawConn, cfg net.KeepAliveConfig) {}

network/v2/socks_proxy_test.go

 	require.NoError(suite.T(), err)
 	suite.authURL = parsed
 
-	suite.baseNetwork = network.New(nil, "mtg", 0, 0, 0)
+	suite.baseNetwork = network.New(nil, "mtg", 0, 0, 0, network.DefaultKeepAliveConfig)
 }
 
 func (suite *SocksProxyTestSuite) TestIncorrectSchema() {