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
|
// Copyright 2020 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build zos && s390x
package unix
import (
"sync"
)
// This file simulates epoll on z/OS using poll.
// Analogous to epoll_event on Linux.
// TODO(neeilan): Pad is because the Linux kernel expects a 96-bit struct. We never pass this to the kernel; remove?
type EpollEvent struct {
Events uint32
Fd int32
Pad int32
}
const (
EPOLLERR = 0x8
EPOLLHUP = 0x10
EPOLLIN = 0x1
EPOLLMSG = 0x400
EPOLLOUT = 0x4
EPOLLPRI = 0x2
EPOLLRDBAND = 0x80
EPOLLRDNORM = 0x40
EPOLLWRBAND = 0x200
EPOLLWRNORM = 0x100
EPOLL_CTL_ADD = 0x1
EPOLL_CTL_DEL = 0x2
EPOLL_CTL_MOD = 0x3
// The following constants are part of the epoll API, but represent
// currently unsupported functionality on z/OS.
// EPOLL_CLOEXEC = 0x80000
// EPOLLET = 0x80000000
// EPOLLONESHOT = 0x40000000
// EPOLLRDHUP = 0x2000 // Typically used with edge-triggered notis
// EPOLLEXCLUSIVE = 0x10000000 // Exclusive wake-up mode
// EPOLLWAKEUP = 0x20000000 // Relies on Linux's BLOCK_SUSPEND capability
)
// TODO(neeilan): We can eliminate these epToPoll / pToEpoll calls by using identical mask values for POLL/EPOLL
// constants where possible The lower 16 bits of epoll events (uint32) can fit any system poll event (int16).
// epToPollEvt converts epoll event field to poll equivalent.
// In epoll, Events is a 32-bit field, while poll uses 16 bits.
func epToPollEvt(events uint32) int16 {
var ep2p = map[uint32]int16{
EPOLLIN: POLLIN,
EPOLLOUT: POLLOUT,
EPOLLHUP: POLLHUP,
EPOLLPRI: POLLPRI,
EPOLLERR: POLLERR,
}
var pollEvts int16 = 0
for epEvt, pEvt := range ep2p {
if (events & epEvt) != 0 {
pollEvts |= pEvt
}
}
return pollEvts
}
// pToEpollEvt converts 16 bit poll event bitfields to 32-bit epoll event fields.
func pToEpollEvt(revents int16) uint32 {
var p2ep = map[int16]uint32{
POLLIN: EPOLLIN,
POLLOUT: EPOLLOUT,
POLLHUP: EPOLLHUP,
POLLPRI: EPOLLPRI,
POLLERR: EPOLLERR,
}
var epollEvts uint32 = 0
for pEvt, epEvt := range p2ep {
if (revents & pEvt) != 0 {
epollEvts |= epEvt
}
}
return epollEvts
}
// Per-process epoll implementation.
type epollImpl struct {
mu sync.Mutex
epfd2ep map[int]*eventPoll
nextEpfd int
}
// eventPoll holds a set of file descriptors being watched by the process. A process can have multiple epoll instances.
// On Linux, this is an in-kernel data structure accessed through a fd.
type eventPoll struct {
mu sync.Mutex
fds map[int]*EpollEvent
}
// epoll impl for this process.
var impl epollImpl = epollImpl{
epfd2ep: make(map[int]*eventPoll),
nextEpfd: 0,
}
func (e *epollImpl) epollcreate(size int) (epfd int, err error) {
e.mu.Lock()
defer e.mu.Unlock()
epfd = e.nextEpfd
e.nextEpfd++
e.epfd2ep[epfd] = &eventPoll{
fds: make(map[int]*EpollEvent),
}
return epfd, nil
}
func (e *epollImpl) epollcreate1(flag int) (fd int, err error) {
return e.epollcreate(4)
}
func (e *epollImpl) epollctl(epfd int, op int, fd int, event *EpollEvent) (err error) {
e.mu.Lock()
defer e.mu.Unlock()
ep, ok := e.epfd2ep[epfd]
if !ok {
return EBADF
}
switch op {
case EPOLL_CTL_ADD:
// TODO(neeilan): When we make epfds and fds disjoint, detect epoll
// loops here (instances watching each other) and return ELOOP.
if _, ok := ep.fds[fd]; ok {
return EEXIST
}
ep.fds[fd] = event
case EPOLL_CTL_MOD:
if _, ok := ep.fds[fd]; !ok {
return ENOENT
}
ep.fds[fd] = event
case EPOLL_CTL_DEL:
if _, ok := ep.fds[fd]; !ok {
return ENOENT
}
delete(ep.fds, fd)
}
return nil
}
// Must be called while holding ep.mu
func (ep *eventPoll) getFds() []int {
fds := make([]int, len(ep.fds))
for fd := range ep.fds {
fds = append(fds, fd)
}
return fds
}
func (e *epollImpl) epollwait(epfd int, events []EpollEvent, msec int) (n int, err error) {
e.mu.Lock() // in [rare] case of concurrent epollcreate + epollwait
ep, ok := e.epfd2ep[epfd]
if !ok {
e.mu.Unlock()
return 0, EBADF
}
pollfds := make([]PollFd, 4)
for fd, epollevt := range ep.fds {
pollfds = append(pollfds, PollFd{Fd: int32(fd), Events: epToPollEvt(epollevt.Events)})
}
e.mu.Unlock()
n, err = Poll(pollfds, msec)
if err != nil {
return n, err
}
i := 0
for _, pFd := range pollfds {
if pFd.Revents != 0 {
events[i] = EpollEvent{Fd: pFd.Fd, Events: pToEpollEvt(pFd.Revents)}
i++
}
if i == n {
break
}
}
return n, nil
}
func EpollCreate(size int) (fd int, err error) {
return impl.epollcreate(size)
}
func EpollCreate1(flag int) (fd int, err error) {
return impl.epollcreate1(flag)
}
func EpollCtl(epfd int, op int, fd int, event *EpollEvent) (err error) {
return impl.epollctl(epfd, op, fd, event)
}
// Because EpollWait mutates events, the caller is expected to coordinate
// concurrent access if calling with the same epfd from multiple goroutines.
func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
return impl.epollwait(epfd, events, msec)
}
|
