ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/ev_epoll.c
(Generate patch)

Comparing libev/ev_epoll.c (file contents):
Revision 1.47 by root, Sun Jul 19 04:11:27 2009 UTC vs.
Revision 1.55 by root, Wed Jan 5 04:21:20 2011 UTC

1/* 1/*
2 * libev epoll fd activity backend 2 * libev epoll fd activity backend
3 * 3 *
4 * Copyright (c) 2007,2008,2009 Marc Alexander Lehmann <libev@schmorp.de> 4 * Copyright (c) 2007,2008,2009,2010,2011 Marc Alexander Lehmann <libev@schmorp.de>
5 * All rights reserved. 5 * All rights reserved.
6 * 6 *
7 * Redistribution and use in source and binary forms, with or without modifica- 7 * Redistribution and use in source and binary forms, with or without modifica-
8 * tion, are permitted provided that the following conditions are met: 8 * tion, are permitted provided that the following conditions are met:
9 * 9 *
… …
47 * b) the fact that ADD != MOD creates a lot of extra syscalls due to a) 47 * b) the fact that ADD != MOD creates a lot of extra syscalls due to a)
48 * and seems not to have any advantage. 48 * and seems not to have any advantage.
49 * c) the inability to handle fork or file descriptors (think dup) 49 * c) the inability to handle fork or file descriptors (think dup)
50 * limits the applicability over poll, so this is not a generic 50 * limits the applicability over poll, so this is not a generic
51 * poll replacement. 51 * poll replacement.
52 * d) epoll doesn't work the same as select with many file descriptors
53 * (such as files). while not critical, no other advanced interface
54 * seems to share this (rather non-unixy) limitation.
55 * e) epoll claims to be embeddable, but in practise you never get
56 * a ready event for the epoll fd (broken: <=2.6.26, working: >=2.6.32).
57 * f) epoll_ctl returning EPERM means the fd is always ready.
52 * 58 *
53 * lots of "weird code" and complication handling in this file is due 59 * lots of "weird code" and complication handling in this file is due
54 * to these design problems with epoll, as we try very hard to avoid 60 * to these design problems with epoll, as we try very hard to avoid
55 * epoll_ctl syscalls for common usage patterns and handle the breakage 61 * epoll_ctl syscalls for common usage patterns and handle the breakage
56 * ensuing from receiving events for closed and otherwise long gone 62 * ensuing from receiving events for closed and otherwise long gone
57 * file descriptors. 63 * file descriptors.
58 */ 64 */
59 65
60#include <sys/epoll.h> 66#include <sys/epoll.h>
67
68#define EV_EMASK_EPERM 0x80
61 69
62static void 70static void
63epoll_modify (EV_P_ int fd, int oev, int nev) 71epoll_modify (EV_P_ int fd, int oev, int nev)
64{ 72{
65 struct epoll_event ev; 73 struct epoll_event ev;
… …
83 ev.data.u64 = (uint64_t)(uint32_t)fd 91 ev.data.u64 = (uint64_t)(uint32_t)fd
84 | ((uint64_t)(uint32_t)++anfds [fd].egen << 32); 92 | ((uint64_t)(uint32_t)++anfds [fd].egen << 32);
85 ev.events = (nev & EV_READ ? EPOLLIN : 0) 93 ev.events = (nev & EV_READ ? EPOLLIN : 0)
86 | (nev & EV_WRITE ? EPOLLOUT : 0); 94 | (nev & EV_WRITE ? EPOLLOUT : 0);
87 95
88 if (expect_true (!epoll_ctl (backend_fd, oev ? EPOLL_CTL_MOD : EPOLL_CTL_ADD, fd, &ev))) 96 if (expect_true (!epoll_ctl (backend_fd, oev && oldmask != nev ? EPOLL_CTL_MOD : EPOLL_CTL_ADD, fd, &ev)))
89 return; 97 return;
90 98
91 if (expect_true (errno == ENOENT)) 99 if (expect_true (errno == ENOENT))
92 { 100 {
93 /* if ENOENT then the fd went away, so try to do the right thing */ 101 /* if ENOENT then the fd went away, so try to do the right thing */
… …
105 goto dec_egen; 113 goto dec_egen;
106 114
107 if (!epoll_ctl (backend_fd, EPOLL_CTL_MOD, fd, &ev)) 115 if (!epoll_ctl (backend_fd, EPOLL_CTL_MOD, fd, &ev))
108 return; 116 return;
109 } 117 }
118 else if (expect_true (errno == EPERM))
119 {
120 anfds [fd].emask = EV_EMASK_EPERM;
121
122 /* add fd to epoll_eperms, if not already inside */
123 if (!(oldmask & EV_EMASK_EPERM))
124 {
125 array_needsize (int, epoll_eperms, epoll_epermmax, epoll_epermcnt + 1, EMPTY2);
126 epoll_eperms [epoll_epermcnt++] = fd;
127 }
128
129 return;
130 }
110 131
111 fd_kill (EV_A_ fd); 132 fd_kill (EV_A_ fd);
112 133
113dec_egen: 134dec_egen:
114 /* we didn't successfully call epoll_ctl, so decrement the generation counter again */ 135 /* we didn't successfully call epoll_ctl, so decrement the generation counter again */
… …
119epoll_poll (EV_P_ ev_tstamp timeout) 140epoll_poll (EV_P_ ev_tstamp timeout)
120{ 141{
121 int i; 142 int i;
122 int eventcnt; 143 int eventcnt;
123 144
145 /* epoll wait times cannot be larger than (LONG_MAX - 999UL) / HZ msecs, which is below */
146 /* the default libev max wait time, however. */
124 EV_RELEASE_CB; 147 EV_RELEASE_CB;
125 eventcnt = epoll_wait (backend_fd, epoll_events, epoll_eventmax, (int)ceil (timeout * 1000.)); 148 eventcnt = epoll_wait (backend_fd, epoll_events, epoll_eventmax, (int)ceil (timeout * 1000.));
126 EV_ACQUIRE_CB; 149 EV_ACQUIRE_CB;
127 150
128 if (expect_false (eventcnt < 0)) 151 if (expect_false (eventcnt < 0))
… …
141 int want = anfds [fd].events; 164 int want = anfds [fd].events;
142 int got = (ev->events & (EPOLLOUT | EPOLLERR | EPOLLHUP) ? EV_WRITE : 0) 165 int got = (ev->events & (EPOLLOUT | EPOLLERR | EPOLLHUP) ? EV_WRITE : 0)
143 | (ev->events & (EPOLLIN | EPOLLERR | EPOLLHUP) ? EV_READ : 0); 166 | (ev->events & (EPOLLIN | EPOLLERR | EPOLLHUP) ? EV_READ : 0);
144 167
145 /* check for spurious notification */ 168 /* check for spurious notification */
169 /* we assume that fd is always in range, as we never shrink the anfds array */
146 if (expect_false ((uint32_t)anfds [fd].egen != (uint32_t)(ev->data.u64 >> 32))) 170 if (expect_false ((uint32_t)anfds [fd].egen != (uint32_t)(ev->data.u64 >> 32)))
147 { 171 {
148 /* recreate kernel state */ 172 /* recreate kernel state */
149 postfork = 1; 173 postfork = 1;
150 continue; 174 continue;
… …
157 /* we received an event but are not interested in it, try mod or del */ 181 /* we received an event but are not interested in it, try mod or del */
158 /* I don't think we ever need MOD, but let's handle it anyways */ 182 /* I don't think we ever need MOD, but let's handle it anyways */
159 ev->events = (want & EV_READ ? EPOLLIN : 0) 183 ev->events = (want & EV_READ ? EPOLLIN : 0)
160 | (want & EV_WRITE ? EPOLLOUT : 0); 184 | (want & EV_WRITE ? EPOLLOUT : 0);
161 185
186 /* pre-2.6.9 kernels require a non-null pointer with EPOLL_CTL_DEL, */
187 /* which is fortunately easy to do for us. */
162 if (epoll_ctl (backend_fd, want ? EPOLL_CTL_MOD : EPOLL_CTL_DEL, fd, ev)) 188 if (epoll_ctl (backend_fd, want ? EPOLL_CTL_MOD : EPOLL_CTL_DEL, fd, ev))
163 { 189 {
164 postfork = 1; /* an error occured, recreate kernel state */ 190 postfork = 1; /* an error occurred, recreate kernel state */
165 continue; 191 continue;
166 } 192 }
167 } 193 }
168 194
169 fd_event (EV_A_ fd, got); 195 fd_event (EV_A_ fd, got);
… …
174 { 200 {
175 ev_free (epoll_events); 201 ev_free (epoll_events);
176 epoll_eventmax = array_nextsize (sizeof (struct epoll_event), epoll_eventmax, epoll_eventmax + 1); 202 epoll_eventmax = array_nextsize (sizeof (struct epoll_event), epoll_eventmax, epoll_eventmax + 1);
177 epoll_events = (struct epoll_event *)ev_malloc (sizeof (struct epoll_event) * epoll_eventmax); 203 epoll_events = (struct epoll_event *)ev_malloc (sizeof (struct epoll_event) * epoll_eventmax);
178 } 204 }
205
206 /* now add events for all fds where epoll fails, while select works... */
207 for (i = epoll_epermcnt; i--; )
208 {
209 int fd = epoll_eperms [i];
210 unsigned char events = anfds [fd].events & (EV_READ | EV_WRITE);
211
212 if (anfds [fd].emask & EV_EMASK_EPERM && events)
213 fd_event (EV_A_ fd, events);
214 else
215 epoll_eperms [i] = epoll_eperms [--epoll_epermcnt];
216 }
179} 217}
180 218
181int inline_size 219int inline_size
182epoll_init (EV_P_ int flags) 220epoll_init (EV_P_ int flags)
183{ 221{
… …
205 243
206void inline_size 244void inline_size
207epoll_destroy (EV_P) 245epoll_destroy (EV_P)
208{ 246{
209 ev_free (epoll_events); 247 ev_free (epoll_events);
248 array_free (epoll_eperm, EMPTY);
210} 249}
211 250
212void inline_size 251void inline_size
213epoll_fork (EV_P) 252epoll_fork (EV_P)
214{ 253{

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines