1 | /* |
1 | /* |
2 | * libev epoll fd activity backend |
2 | * libev epoll fd activity backend |
3 | * |
3 | * |
4 | * Copyright (c) 2007,2008 Marc Alexander Lehmann <libev@schmorp.de> |
4 | * Copyright (c) 2007,2008,2009,2010 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. |
52 | * |
57 | * |
53 | * lots of "weird code" and complication handling in this file is due |
58 | * 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 |
59 | * 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 |
60 | * epoll_ctl syscalls for common usage patterns and handle the breakage |
56 | * ensuing from receiving events for closed and otherwise long gone |
61 | * ensuing from receiving events for closed and otherwise long gone |
… | |
… | |
68 | /* |
73 | /* |
69 | * we handle EPOLL_CTL_DEL by ignoring it here |
74 | * we handle EPOLL_CTL_DEL by ignoring it here |
70 | * on the assumption that the fd is gone anyways |
75 | * on the assumption that the fd is gone anyways |
71 | * if that is wrong, we have to handle the spurious |
76 | * if that is wrong, we have to handle the spurious |
72 | * event in epoll_poll. |
77 | * event in epoll_poll. |
73 | * the fd is later added, we try to ADD it, and, if that |
78 | * if the fd is added again, we try to ADD it, and, if that |
74 | * fails, we assume it still has the same eventmask. |
79 | * fails, we assume it still has the same eventmask. |
75 | */ |
80 | */ |
76 | if (!nev) |
81 | if (!nev) |
77 | return; |
82 | return; |
78 | |
83 | |
79 | oldmask = anfds [fd].emask; |
84 | oldmask = anfds [fd].emask; |
80 | anfds [fd].emask = nev; |
85 | anfds [fd].emask = nev; |
81 | |
86 | |
82 | /* store the generation counter in the upper 32 bits */ |
87 | /* store the generation counter in the upper 32 bits, the fd in the lower 32 bits */ |
|
|
88 | ev.data.u64 = (uint64_t)(uint32_t)fd |
83 | ev.data.u64 = fd | ((uint64_t)(uint32_t)++anfds [fd].egen << 32); |
89 | | ((uint64_t)(uint32_t)++anfds [fd].egen << 32); |
84 | ev.events = (nev & EV_READ ? EPOLLIN : 0) |
90 | ev.events = (nev & EV_READ ? EPOLLIN : 0) |
85 | | (nev & EV_WRITE ? EPOLLOUT : 0); |
91 | | (nev & EV_WRITE ? EPOLLOUT : 0); |
86 | |
92 | |
87 | if (expect_true (!epoll_ctl (backend_fd, oev ? EPOLL_CTL_MOD : EPOLL_CTL_ADD, fd, &ev))) |
93 | if (expect_true (!epoll_ctl (backend_fd, oev && oldmask != nev ? EPOLL_CTL_MOD : EPOLL_CTL_ADD, fd, &ev))) |
88 | return; |
94 | return; |
89 | |
95 | |
90 | if (expect_true (errno == ENOENT)) |
96 | if (expect_true (errno == ENOENT)) |
91 | { |
97 | { |
92 | /* if ENOENT then the fd went away, so try to do the right thing */ |
98 | /* if ENOENT then the fd went away, so try to do the right thing */ |
… | |
… | |
116 | |
122 | |
117 | static void |
123 | static void |
118 | epoll_poll (EV_P_ ev_tstamp timeout) |
124 | epoll_poll (EV_P_ ev_tstamp timeout) |
119 | { |
125 | { |
120 | int i; |
126 | int i; |
|
|
127 | int eventcnt; |
|
|
128 | |
|
|
129 | /* epoll wait times cannot be larger than (LONG_MAX - 999UL) / HZ msecs, which is below */ |
|
|
130 | /* the default libev max wait time, however. */ |
|
|
131 | EV_RELEASE_CB; |
121 | int eventcnt = epoll_wait (backend_fd, epoll_events, epoll_eventmax, (int)ceil (timeout * 1000.)); |
132 | eventcnt = epoll_wait (backend_fd, epoll_events, epoll_eventmax, (int)ceil (timeout * 1000.)); |
|
|
133 | EV_ACQUIRE_CB; |
122 | |
134 | |
123 | if (expect_false (eventcnt < 0)) |
135 | if (expect_false (eventcnt < 0)) |
124 | { |
136 | { |
125 | if (errno != EINTR) |
137 | if (errno != EINTR) |
126 | ev_syserr ("(libev) epoll_wait"); |
138 | ev_syserr ("(libev) epoll_wait"); |
… | |
… | |
152 | /* we received an event but are not interested in it, try mod or del */ |
164 | /* we received an event but are not interested in it, try mod or del */ |
153 | /* I don't think we ever need MOD, but let's handle it anyways */ |
165 | /* I don't think we ever need MOD, but let's handle it anyways */ |
154 | ev->events = (want & EV_READ ? EPOLLIN : 0) |
166 | ev->events = (want & EV_READ ? EPOLLIN : 0) |
155 | | (want & EV_WRITE ? EPOLLOUT : 0); |
167 | | (want & EV_WRITE ? EPOLLOUT : 0); |
156 | |
168 | |
|
|
169 | /* pre-2.6.9 kernels require a non-null pointer with EPOLL_CTL_DEL, */ |
|
|
170 | /* which is fortunately easy to do for us. */ |
157 | if (epoll_ctl (backend_fd, want ? EPOLL_CTL_MOD : EPOLL_CTL_DEL, fd, ev)) |
171 | if (epoll_ctl (backend_fd, want ? EPOLL_CTL_MOD : EPOLL_CTL_DEL, fd, ev)) |
158 | { |
172 | { |
159 | postfork = 1; /* an error occured, recreate kernel state */ |
173 | postfork = 1; /* an error occured, recreate kernel state */ |
160 | continue; |
174 | continue; |
161 | } |
175 | } |
… | |
… | |
174 | } |
188 | } |
175 | |
189 | |
176 | int inline_size |
190 | int inline_size |
177 | epoll_init (EV_P_ int flags) |
191 | epoll_init (EV_P_ int flags) |
178 | { |
192 | { |
|
|
193 | #ifdef EPOLL_CLOEXEC |
|
|
194 | backend_fd = epoll_create1 (EPOLL_CLOEXEC); |
|
|
195 | |
|
|
196 | if (backend_fd <= 0) |
|
|
197 | #endif |
179 | backend_fd = epoll_create (256); |
198 | backend_fd = epoll_create (256); |
180 | |
199 | |
181 | if (backend_fd < 0) |
200 | if (backend_fd < 0) |
182 | return 0; |
201 | return 0; |
183 | |
202 | |
184 | fcntl (backend_fd, F_SETFD, FD_CLOEXEC); |
203 | fcntl (backend_fd, F_SETFD, FD_CLOEXEC); |