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
|
// 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 sharedmem
import (
"math"
"golang.org/x/sys/unix"
"gvisor.dev/gvisor/pkg/buffer"
"gvisor.dev/gvisor/pkg/eventfd"
"gvisor.dev/gvisor/pkg/tcpip/link/sharedmem/queue"
)
const (
nilID = math.MaxUint64
)
// tx holds all state associated with a tx queue.
type tx struct {
data []byte
q queue.Tx
ids idManager
bufs bufferManager
eventFD eventfd.Eventfd
sharedData []byte
sharedDataFD int
}
// init initializes all state needed by the tx queue based on the information
// provided.
//
// The caller always retains ownership of all file descriptors passed in. The
// queue implementation will duplicate any that it may need in the future.
func (t *tx) init(bufferSize uint32, c *QueueConfig) error {
// Map in all buffers.
txPipe, err := getBuffer(c.TxPipeFD)
if err != nil {
return err
}
rxPipe, err := getBuffer(c.RxPipeFD)
if err != nil {
unix.Munmap(txPipe)
return err
}
data, err := getBuffer(c.DataFD)
if err != nil {
unix.Munmap(txPipe)
unix.Munmap(rxPipe)
return err
}
sharedData, err := getBuffer(c.SharedDataFD)
if err != nil {
unix.Munmap(txPipe)
unix.Munmap(rxPipe)
unix.Munmap(data)
}
// Initialize state based on buffers.
t.q.Init(txPipe, rxPipe, sharedDataPointer(sharedData))
t.ids.init()
t.bufs.init(0, len(data), int(bufferSize))
t.data = data
t.eventFD = c.EventFD
t.sharedDataFD = c.SharedDataFD
t.sharedData = sharedData
return nil
}
// cleanup releases all resources allocated during init(). It must only be
// called if init() has previously succeeded.
func (t *tx) cleanup() {
a, b := t.q.Bytes()
unix.Munmap(a)
unix.Munmap(b)
unix.Munmap(t.data)
}
// transmit sends a packet made of bufs. Returns a boolean that specifies
// whether the packet was successfully transmitted.
func (t *tx) transmit(transmitBuf buffer.Buffer) bool {
// Pull completions from the tx queue and add their buffers back to the
// pool so that we can reuse them.
for {
id, ok := t.q.CompletedPacket()
if !ok {
break
}
if buf := t.ids.remove(id); buf != nil {
t.bufs.free(buf)
}
}
bSize := t.bufs.entrySize
total := uint32(transmitBuf.Size())
bufCount := (total + bSize - 1) / bSize
// Allocate enough buffers to hold all the data.
var buf *queue.TxBuffer
for i := bufCount; i != 0; i-- {
b := t.bufs.alloc()
if b == nil {
// Failed to get all buffers. Return to the pool
// whatever we had managed to get.
if buf != nil {
t.bufs.free(buf)
}
return false
}
b.Next = buf
buf = b
}
// Copy data into allocated buffers.
nBuf := buf
var dBuf []byte
transmitBuf.Apply(func(v *buffer.View) {
for v.Size() > 0 {
if len(dBuf) == 0 {
dBuf = t.data[nBuf.Offset:][:nBuf.Size]
nBuf = nBuf.Next
}
n := copy(dBuf, v.AsSlice())
v.TrimFront(n)
dBuf = dBuf[n:]
}
})
// Get an id for this packet and send it out.
id := t.ids.add(buf)
if !t.q.Enqueue(id, total, bufCount, buf) {
t.ids.remove(id)
t.bufs.free(buf)
return false
}
return true
}
// notify writes to the tx.eventFD to indicate to the peer that there is data to
// be read.
func (t *tx) notify() {
if t.q.NotificationsEnabled() {
t.eventFD.Notify()
}
}
// idDescriptor is used by idManager to either point to a tx buffer (in case
// the ID is assigned) or to the next free element (if the id is not assigned).
type idDescriptor struct {
buf *queue.TxBuffer
nextFree uint64
}
// idManager is a manager of tx buffer identifiers. It assigns unique IDs to
// tx buffers that are added to it; the IDs can only be reused after they have
// been removed.
//
// The ID assignments are stored so that the tx buffers can be retrieved from
// the IDs previously assigned to them.
type idManager struct {
// ids is a slice containing all tx buffers. The ID is the index into
// this slice.
ids []idDescriptor
// freeList a list of free IDs.
freeList uint64
}
// init initializes the id manager.
func (m *idManager) init() {
m.freeList = nilID
}
// add assigns an ID to the given tx buffer.
func (m *idManager) add(b *queue.TxBuffer) uint64 {
if i := m.freeList; i != nilID {
// There is an id available in the free list, just use it.
m.ids[i].buf = b
m.freeList = m.ids[i].nextFree
return i
}
// We need to expand the id descriptor.
m.ids = append(m.ids, idDescriptor{buf: b})
return uint64(len(m.ids) - 1)
}
// remove retrieves the tx buffer associated with the given ID, and removes the
// ID from the assigned table so that it can be reused in the future.
func (m *idManager) remove(i uint64) *queue.TxBuffer {
if i >= uint64(len(m.ids)) {
return nil
}
desc := &m.ids[i]
b := desc.buf
if b == nil {
// The provided id is not currently assigned.
return nil
}
desc.buf = nil
desc.nextFree = m.freeList
m.freeList = i
return b
}
// bufferManager manages a buffer region broken up into smaller, equally sized
// buffers. Smaller buffers can be allocated and freed.
type bufferManager struct {
freeList *queue.TxBuffer
curOffset uint64
limit uint64
entrySize uint32
}
// init initializes the buffer manager.
func (b *bufferManager) init(initialOffset, size, entrySize int) {
b.freeList = nil
b.curOffset = uint64(initialOffset)
b.limit = uint64(initialOffset + size/entrySize*entrySize)
b.entrySize = uint32(entrySize)
}
// alloc allocates a buffer from the manager, if one is available.
func (b *bufferManager) alloc() *queue.TxBuffer {
if b.freeList != nil {
// There is a descriptor ready for reuse in the free list.
d := b.freeList
b.freeList = d.Next
d.Next = nil
return d
}
if b.curOffset < b.limit {
// There is room available in the never-used range, so create
// a new descriptor for it.
d := &queue.TxBuffer{
Offset: b.curOffset,
Size: b.entrySize,
}
b.curOffset += uint64(b.entrySize)
return d
}
return nil
}
// free returns all buffers in the list to the buffer manager so that they can
// be reused.
func (b *bufferManager) free(d *queue.TxBuffer) {
// Find the last buffer in the list.
last := d
for last.Next != nil {
last = last.Next
}
// Push list onto free list.
last.Next = b.freeList
b.freeList = d
}
|