summaryrefslogtreecommitdiffhomepage
path: root/pkg/tcpip/ports/ports.go
blob: edc29ad27815d2e4980ce70068e2299fa953cc87 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
// Copyright 2018 The gVisor Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

// Package ports provides PortManager that manages allocating, reserving and releasing ports.
package ports

import (
	"math"
	"math/rand"
	"sync/atomic"

	"gvisor.dev/gvisor/pkg/sync"
	"gvisor.dev/gvisor/pkg/tcpip"
)

const (
	// FirstEphemeral is the first ephemeral port.
	FirstEphemeral = 16000

	// numEphemeralPorts it the mnumber of available ephemeral ports to
	// Netstack.
	numEphemeralPorts = math.MaxUint16 - FirstEphemeral + 1

	anyIPAddress tcpip.Address = ""
)

type portDescriptor struct {
	network   tcpip.NetworkProtocolNumber
	transport tcpip.TransportProtocolNumber
	port      uint16
}

// Flags represents the type of port reservation.
//
// +stateify savable
type Flags struct {
	// MostRecent represents UDP SO_REUSEADDR.
	MostRecent bool

	// LoadBalanced indicates SO_REUSEPORT.
	//
	// LoadBalanced takes precidence over MostRecent.
	LoadBalanced bool
}

// Bits converts the Flags to their bitset form.
func (f Flags) Bits() BitFlags {
	var rf BitFlags
	if f.MostRecent {
		rf |= MostRecentFlag
	}
	if f.LoadBalanced {
		rf |= LoadBalancedFlag
	}
	return rf
}

// Effective returns the effective behavior of a flag config.
func (f Flags) Effective() Flags {
	e := f
	if e.LoadBalanced && e.MostRecent {
		e.MostRecent = false
	}
	return e
}

// PortManager manages allocating, reserving and releasing ports.
type PortManager struct {
	mu             sync.RWMutex
	allocatedPorts map[portDescriptor]bindAddresses

	// hint is used to pick ports ephemeral ports in a stable order for
	// a given port offset.
	//
	// hint must be accessed using the portHint/incPortHint helpers.
	// TODO(gvisor.dev/issue/940): S/R this field.
	hint uint32
}

// BitFlags is a bitset representation of Flags.
type BitFlags uint32

const (
	// MostRecentFlag represents Flags.MostRecent.
	MostRecentFlag BitFlags = 1 << iota

	// LoadBalancedFlag represents Flags.LoadBalanced.
	LoadBalancedFlag

	// nextFlag is the value that the next added flag will have.
	//
	// It is used to calculate FlagMask below. It is also the number of
	// valid flag states.
	nextFlag

	// FlagMask is a bit mask for BitFlags.
	FlagMask = nextFlag - 1
)

// ToFlags converts the bitset into a Flags struct.
func (f BitFlags) ToFlags() Flags {
	return Flags{
		MostRecent:   f&MostRecentFlag != 0,
		LoadBalanced: f&LoadBalancedFlag != 0,
	}
}

// FlagCounter counts how many references each flag combination has.
type FlagCounter struct {
	// refs stores the count for each possible flag combination, (0 though
	// FlagMask).
	refs [nextFlag]int
}

// AddRef increases the reference count for a specific flag combination.
func (c *FlagCounter) AddRef(flags BitFlags) {
	c.refs[flags]++
}

// DropRef decreases the reference count for a specific flag combination.
func (c *FlagCounter) DropRef(flags BitFlags) {
	c.refs[flags]--
}

// TotalRefs calculates the total number of references for all flag
// combinations.
func (c FlagCounter) TotalRefs() int {
	var total int
	for _, r := range c.refs {
		total += r
	}
	return total
}

// FlagRefs returns the number of references with all specified flags.
func (c FlagCounter) FlagRefs(flags BitFlags) int {
	var total int
	for i, r := range c.refs {
		if BitFlags(i)&flags == flags {
			total += r
		}
	}
	return total
}

// AllRefsHave returns if all references have all specified flags.
func (c FlagCounter) AllRefsHave(flags BitFlags) bool {
	for i, r := range c.refs {
		if BitFlags(i)&flags != flags && r > 0 {
			return false
		}
	}
	return true
}

// IntersectionRefs returns the set of flags shared by all references.
func (c FlagCounter) IntersectionRefs() BitFlags {
	intersection := FlagMask
	for i, r := range c.refs {
		if r > 0 {
			intersection &= BitFlags(i)
		}
	}
	return intersection
}

// deviceNode is never empty. When it has no elements, it is removed from the
// map that references it.
type deviceNode map[tcpip.NICID]FlagCounter

// isAvailable checks whether binding is possible by device. If not binding to a
// device, check against all FlagCounters. If binding to a specific device, check
// against the unspecified device and the provided device.
//
// If either of the port reuse flags is enabled on any of the nodes, all nodes
// sharing a port must share at least one reuse flag. This matches Linux's
// behavior.
func (d deviceNode) isAvailable(flags Flags, bindToDevice tcpip.NICID) bool {
	flagBits := flags.Bits()
	if bindToDevice == 0 {
		// Trying to binding all devices.
		if flagBits == 0 {
			// Can't bind because the (addr,port) is already bound.
			return false
		}
		intersection := FlagMask
		for _, p := range d {
			i := p.IntersectionRefs()
			intersection &= i
			if intersection&flagBits == 0 {
				// Can't bind because the (addr,port) was
				// previously bound without reuse.
				return false
			}
		}
		return true
	}

	intersection := FlagMask

	if p, ok := d[0]; ok {
		intersection = p.IntersectionRefs()
		if intersection&flagBits == 0 {
			return false
		}
	}

	if p, ok := d[bindToDevice]; ok {
		i := p.IntersectionRefs()
		intersection &= i
		if intersection&flagBits == 0 {
			return false
		}
	}

	return true
}

// bindAddresses is a set of IP addresses.
type bindAddresses map[tcpip.Address]deviceNode

// isAvailable checks whether an IP address is available to bind to. If the
// address is the "any" address, check all other addresses. Otherwise, just
// check against the "any" address and the provided address.
func (b bindAddresses) isAvailable(addr tcpip.Address, flags Flags, bindToDevice tcpip.NICID) bool {
	if addr == anyIPAddress {
		// If binding to the "any" address then check that there are no conflicts
		// with all addresses.
		for _, d := range b {
			if !d.isAvailable(flags, bindToDevice) {
				return false
			}
		}
		return true
	}

	// Check that there is no conflict with the "any" address.
	if d, ok := b[anyIPAddress]; ok {
		if !d.isAvailable(flags, bindToDevice) {
			return false
		}
	}

	// Check that this is no conflict with the provided address.
	if d, ok := b[addr]; ok {
		if !d.isAvailable(flags, bindToDevice) {
			return false
		}
	}

	return true
}

// NewPortManager creates new PortManager.
func NewPortManager() *PortManager {
	return &PortManager{allocatedPorts: make(map[portDescriptor]bindAddresses)}
}

// PickEphemeralPort randomly chooses a starting point and iterates over all
// possible ephemeral ports, allowing the caller to decide whether a given port
// is suitable for its needs, and stopping when a port is found or an error
// occurs.
func (s *PortManager) PickEphemeralPort(testPort func(p uint16) (bool, *tcpip.Error)) (port uint16, err *tcpip.Error) {
	offset := uint32(rand.Int31n(numEphemeralPorts))
	return s.pickEphemeralPort(offset, numEphemeralPorts, testPort)
}

// portHint atomically reads and returns the s.hint value.
func (s *PortManager) portHint() uint32 {
	return atomic.LoadUint32(&s.hint)
}

// incPortHint atomically increments s.hint by 1.
func (s *PortManager) incPortHint() {
	atomic.AddUint32(&s.hint, 1)
}

// PickEphemeralPortStable starts at the specified offset + s.portHint and
// iterates over all ephemeral ports, allowing the caller to decide whether a
// given port is suitable for its needs and stopping when a port is found or an
// error occurs.
func (s *PortManager) PickEphemeralPortStable(offset uint32, testPort func(p uint16) (bool, *tcpip.Error)) (port uint16, err *tcpip.Error) {
	p, err := s.pickEphemeralPort(s.portHint()+offset, numEphemeralPorts, testPort)
	if err == nil {
		s.incPortHint()
	}
	return p, err

}

// pickEphemeralPort starts at the offset specified from the FirstEphemeral port
// and iterates over the number of ports specified by count and allows the
// caller to decide whether a given port is suitable for its needs, and stopping
// when a port is found or an error occurs.
func (s *PortManager) pickEphemeralPort(offset, count uint32, testPort func(p uint16) (bool, *tcpip.Error)) (port uint16, err *tcpip.Error) {
	for i := uint32(0); i < count; i++ {
		port = uint16(FirstEphemeral + (offset+i)%count)
		ok, err := testPort(port)
		if err != nil {
			return 0, err
		}

		if ok {
			return port, nil
		}
	}

	return 0, tcpip.ErrNoPortAvailable
}

// IsPortAvailable tests if the given port is available on all given protocols.
func (s *PortManager) IsPortAvailable(networks []tcpip.NetworkProtocolNumber, transport tcpip.TransportProtocolNumber, addr tcpip.Address, port uint16, flags Flags, bindToDevice tcpip.NICID) bool {
	s.mu.Lock()
	defer s.mu.Unlock()
	return s.isPortAvailableLocked(networks, transport, addr, port, flags, bindToDevice)
}

func (s *PortManager) isPortAvailableLocked(networks []tcpip.NetworkProtocolNumber, transport tcpip.TransportProtocolNumber, addr tcpip.Address, port uint16, flags Flags, bindToDevice tcpip.NICID) bool {
	for _, network := range networks {
		desc := portDescriptor{network, transport, port}
		if addrs, ok := s.allocatedPorts[desc]; ok {
			if !addrs.isAvailable(addr, flags, bindToDevice) {
				return false
			}
		}
	}
	return true
}

// ReservePort marks a port/IP combination as reserved so that it cannot be
// reserved by another endpoint. If port is zero, ReservePort will search for
// an unreserved ephemeral port and reserve it, returning its value in the
// "port" return value.
func (s *PortManager) ReservePort(networks []tcpip.NetworkProtocolNumber, transport tcpip.TransportProtocolNumber, addr tcpip.Address, port uint16, flags Flags, bindToDevice tcpip.NICID) (reservedPort uint16, err *tcpip.Error) {
	s.mu.Lock()
	defer s.mu.Unlock()

	// If a port is specified, just try to reserve it for all network
	// protocols.
	if port != 0 {
		if !s.reserveSpecificPort(networks, transport, addr, port, flags, bindToDevice) {
			return 0, tcpip.ErrPortInUse
		}
		return port, nil
	}

	// A port wasn't specified, so try to find one.
	return s.PickEphemeralPort(func(p uint16) (bool, *tcpip.Error) {
		return s.reserveSpecificPort(networks, transport, addr, p, flags, bindToDevice), nil
	})
}

// reserveSpecificPort tries to reserve the given port on all given protocols.
func (s *PortManager) reserveSpecificPort(networks []tcpip.NetworkProtocolNumber, transport tcpip.TransportProtocolNumber, addr tcpip.Address, port uint16, flags Flags, bindToDevice tcpip.NICID) bool {
	if !s.isPortAvailableLocked(networks, transport, addr, port, flags, bindToDevice) {
		return false
	}
	flagBits := flags.Bits()

	// Reserve port on all network protocols.
	for _, network := range networks {
		desc := portDescriptor{network, transport, port}
		m, ok := s.allocatedPorts[desc]
		if !ok {
			m = make(bindAddresses)
			s.allocatedPorts[desc] = m
		}
		d, ok := m[addr]
		if !ok {
			d = make(deviceNode)
			m[addr] = d
		}
		n := d[bindToDevice]
		n.AddRef(flagBits)
		d[bindToDevice] = n
	}

	return true
}

// ReleasePort releases the reservation on a port/IP combination so that it can
// be reserved by other endpoints.
func (s *PortManager) ReleasePort(networks []tcpip.NetworkProtocolNumber, transport tcpip.TransportProtocolNumber, addr tcpip.Address, port uint16, flags Flags, bindToDevice tcpip.NICID) {
	s.mu.Lock()
	defer s.mu.Unlock()

	flagBits := flags.Bits()

	for _, network := range networks {
		desc := portDescriptor{network, transport, port}
		if m, ok := s.allocatedPorts[desc]; ok {
			d, ok := m[addr]
			if !ok {
				continue
			}
			n, ok := d[bindToDevice]
			if !ok {
				continue
			}
			n.refs[flagBits]--
			d[bindToDevice] = n
			if n.TotalRefs() == 0 {
				delete(d, bindToDevice)
			}
			if len(d) == 0 {
				delete(m, addr)
			}
			if len(m) == 0 {
				delete(s.allocatedPorts, desc)
			}
		}
	}
}