Split event.c into per-API files
authorKirill Isakov <bootctl@gmail.com>
Sun, 22 May 2022 07:28:03 +0000 (13:28 +0600)
committerKirill Isakov <bootctl@gmail.com>
Tue, 24 May 2022 03:44:23 +0000 (09:44 +0600)
13 files changed:
src/bsd/darwin/meson.build
src/bsd/event.c [new file with mode: 0644]
src/bsd/meson.build
src/event.c
src/event.h
src/event_select.c [new file with mode: 0644]
src/linux/event.c [new file with mode: 0644]
src/linux/meson.build
src/meson.build
src/signal.c [new file with mode: 0644]
src/solaris/meson.build
src/windows/event.c [new file with mode: 0644]
src/windows/meson.build

index 5672fcd..f7dc99e 100644 (file)
@@ -7,3 +7,6 @@ if dep_tunemu.found() and dep_pcap.found()
   cdata.set('ENABLE_TUNEMU', 1)
 endif
 
+# macOS apparently doesn't support kqueue with TAP devices
+src_tincd += src_event_select
+
diff --git a/src/bsd/event.c b/src/bsd/event.c
new file mode 100644 (file)
index 0000000..a1d5742
--- /dev/null
@@ -0,0 +1,170 @@
+/*
+    event.c -- kqueue support for the BSD family
+    Copyright (C) 2012-2022 Guus Sliepen <guus@tinc-vpn.org>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+*/
+
+#include "../system.h"
+
+#include <sys/event.h>
+
+#include "../event.h"
+#include "../utils.h"
+#include "../net.h"
+
+static bool running = false;
+static int kq = 0;
+
+static inline void event_init(void) {
+       if(!kq) {
+               kq = kqueue();
+
+               if(kq == -1) {
+                       logger(DEBUG_ALWAYS, LOG_EMERG, "Could not initialize kqueue: %s", strerror(errno));
+                       abort();
+               }
+       }
+}
+
+static void event_deinit(void) {
+       if(kq) {
+               close(kq);
+               kq = 0;
+       }
+}
+
+void io_add(io_t *io, io_cb_t cb, void *data, int fd, int flags) {
+       if(io->cb) {
+               return;
+       }
+
+       io->fd = fd;
+       io->cb = cb;
+       io->data = data;
+       io->node.data = io;
+
+       io_set(io, flags);
+}
+
+void io_set(io_t *io, int flags) {
+       event_init();
+
+       if(flags == io->flags) {
+               return;
+       }
+
+       io->flags = flags;
+
+       if(io->fd == -1) {
+               return;
+       }
+
+       const struct kevent change[] = {
+               {
+                       .ident = io->fd,
+                       .filter = EVFILT_READ,
+                       .flags = EV_RECEIPT | (flags & IO_READ ? EV_ADD : EV_DELETE),
+                       .udata = io,
+               },
+               {
+                       .ident = io->fd,
+                       .filter = EVFILT_WRITE,
+                       .flags = EV_RECEIPT | (flags & IO_WRITE ? EV_ADD : EV_DELETE),
+                       .udata = io,
+               },
+       };
+       struct kevent result[2];
+
+       if(kevent(kq, change, 2, result, 2, NULL) < 0) {
+               logger(DEBUG_ALWAYS, LOG_EMERG, "kevent failed: %s", strerror(errno));
+               abort();
+       }
+
+       int rerr = (int)result[0].data;
+       int werr = (int)result[1].data;
+
+       if((rerr && rerr != ENOENT) || (werr && werr != ENOENT)) {
+               logger(DEBUG_ALWAYS, LOG_EMERG, "kevent errors: %s, %s", strerror(rerr), strerror(werr));
+               abort();
+       }
+
+       if(!flags) {
+               io_tree.generation++;
+       }
+}
+
+void io_del(io_t *io) {
+       if(io->cb) {
+               io_set(io, 0);
+               io->cb = NULL;
+       }
+}
+
+bool event_loop(void) {
+       event_init();
+       running = true;
+
+       while(running) {
+               struct timeval diff;
+               struct timeval *tv = timeout_execute(&diff);
+               struct kevent events[MAX_EVENTS_PER_LOOP];
+
+               const struct timespec ts = {
+                       .tv_sec = tv->tv_sec,
+                       .tv_nsec = tv->tv_usec * 1000,
+               };
+
+               int n = kevent(kq, NULL, 0, events, MAX_EVENTS_PER_LOOP, &ts);
+
+               if(n < 0) {
+                       if(sockwouldblock(sockerrno)) {
+                               continue;
+                       } else {
+                               return false;
+                       }
+               }
+
+               if(!n) {
+                       continue;
+               }
+
+               unsigned int curgen = io_tree.generation;
+
+               for(int i = 0; i < n; i++) {
+                       const struct kevent *evt = &events[i];
+                       const io_t *io = evt->udata;
+
+                       if(evt->filter == EVFILT_WRITE) {
+                               io->cb(io->data, IO_WRITE);
+                       } else if(evt->filter == EVFILT_READ) {
+                               io->cb(io->data, IO_READ);
+                       } else {
+                               continue;
+                       }
+
+                       if(curgen != io_tree.generation) {
+                               break;
+                       }
+               }
+       }
+
+       event_deinit();
+       return true;
+}
+
+void event_exit(void) {
+       running = false;
+}
index 91bc550..cc29479 100644 (file)
@@ -6,11 +6,6 @@ check_headers += [
   'net/tun/if_tun.h',
 ]
 
-# macOS apparently doesn't support kqueue with TAP devices
-if os_name != 'darwin'
-  check_headers += 'sys/event.h'
-endif
-
 check_functions += [
   'devname',
   'fdevname',
@@ -18,6 +13,10 @@ check_functions += [
 
 src_tincd += files('device.c')
 
+if os_name != 'darwin'
+  src_tincd += files('event.c')
+endif
+
 if os_name in ['openbsd', 'darwin']
   subdir(os_name)
 endif
index c373eac..56d8fa9 100644 (file)
@@ -1,5 +1,5 @@
 /*
-    event.c -- I/O, timeout and signal event handling
+    event.c -- I/O, timeout, and event handling
     Copyright (C) 2012-2022 Guus Sliepen <guus@tinc-vpn.org>
 
     This program is free software; you can redistribute it and/or modify
 
 #include "system.h"
 
-#ifdef HAVE_WINDOWS
-#  include <assert.h>
-#else
-#  if defined(HAVE_SYS_EPOLL_H)
-#    include <sys/epoll.h>
-#    define HAVE_EPOLL 1
-#  elif defined(HAVE_SYS_EVENT_H)
-#    include <sys/event.h>
-#    define HAVE_KQUEUE 1
-#  else
-#    define HAVE_SELECT 1
-#  endif
-#endif
-
 #include "event.h"
-#include "utils.h"
-#include "net.h"
 
 struct timeval now;
-static bool running;
-
-#if defined(HAVE_EPOLL) || defined(HAVE_KQUEUE)
-static int event_fd = 0;
-#elif defined(HAVE_SELECT)
-static fd_set readfds;
-static fd_set writefds;
-#elif defined(HAVE_WINDOWS)
-static const long READ_EVENTS = FD_READ | FD_ACCEPT | FD_CLOSE;
-static const long WRITE_EVENTS = FD_WRITE | FD_CONNECT;
-static DWORD event_count = 0;
-#endif
-
-static inline void event_init(void) {
-#if defined(HAVE_EPOLL) || defined(HAVE_KQUEUE)
-
-       if(!event_fd) {
-#if defined(HAVE_EPOLL)
-               /* NOTE: 1024 limit is only used on ancient (pre 2.6.27) kernels.
-                  Decent kernels will ignore this value making it unlimited.
-                  epoll_create1 might be better, but these kernels would not be supported
-                  in that case. */
-               event_fd = epoll_create(1024);
-#else
-               event_fd = kqueue();
-#endif
-
-               if(event_fd == -1) {
-                       logger(DEBUG_ALWAYS, LOG_EMERG, "Could not initialize events: %s", strerror(errno));
-                       abort();
-               }
-       }
-
-#endif
-}
-
-static void event_deinit(void) {
-#if defined(HAVE_EPOLL) || defined(HAVE_KQUEUE)
-
-       if(event_fd > 0) {
-               close(event_fd);
-               event_fd = 0;
-       }
-
-#endif
-}
 
 static int io_compare(const io_t *a, const io_t *b) {
 #ifndef HAVE_WINDOWS
@@ -133,178 +71,8 @@ static int timeout_compare(const timeout_t *a, const timeout_t *b) {
        return 0;
 }
 
-static splay_tree_t io_tree = {.compare = (splay_compare_t)io_compare};
-static splay_tree_t timeout_tree = {.compare = (splay_compare_t)timeout_compare};
-
-void io_add(io_t *io, io_cb_t cb, void *data, int fd, int flags) {
-       if(io->cb) {
-               return;
-       }
-
-       io->fd = fd;
-#ifdef HAVE_WINDOWS
-
-       if(io->fd != -1) {
-               io->event = WSACreateEvent();
-
-               if(io->event == WSA_INVALID_EVENT) {
-                       abort();
-               }
-       }
-
-       event_count++;
-#endif
-       io->cb = cb;
-       io->data = data;
-       io->node.data = io;
-
-       io_set(io, flags);
-
-#ifdef HAVE_SELECT
-
-       if(!splay_insert_node(&io_tree, &io->node)) {
-               abort();
-       }
-
-#endif
-}
-
-#ifdef HAVE_WINDOWS
-void io_add_event(io_t *io, io_cb_t cb, void *data, WSAEVENT event) {
-       io->event = event;
-       io_add(io, cb, data, -1, 0);
-}
-#endif
-
-void io_set(io_t *io, int flags) {
-       event_init();
-
-       if(flags == io->flags) {
-               return;
-       }
-
-       io->flags = flags;
-
-       if(io->fd == -1) {
-               return;
-       }
-
-#ifndef HAVE_WINDOWS
-#ifdef HAVE_EPOLL
-       epoll_ctl(event_fd, EPOLL_CTL_DEL, io->fd, NULL);
-
-       struct epoll_event ev = {
-               .events = 0,
-               .data.ptr = io,
-       };
-
-       if(flags & IO_READ) {
-               ev.events |= EPOLLIN;
-       }
-
-       if(flags & IO_WRITE) {
-               ev.events |= EPOLLOUT;
-       } else if(ev.events == 0) {
-               io_tree.generation++;
-               return;
-       }
-
-       if(epoll_ctl(event_fd, EPOLL_CTL_ADD, io->fd, &ev) < 0) {
-               perror("epoll_ctl_add");
-       }
-
-#endif
-
-#ifdef HAVE_KQUEUE
-       const struct kevent change[] = {
-               {
-                       .ident = io->fd,
-                       .filter = EVFILT_READ,
-                       .flags = EV_RECEIPT | (flags & IO_READ ? EV_ADD : EV_DELETE),
-                       .udata = io,
-               },
-               {
-                       .ident = io->fd,
-                       .filter = EVFILT_WRITE,
-                       .flags = EV_RECEIPT | (flags & IO_WRITE ? EV_ADD : EV_DELETE),
-                       .udata = io,
-               },
-       };
-       struct kevent result[2];
-
-       if(kevent(event_fd, change, 2, result, 2, NULL) < 0) {
-               logger(DEBUG_ALWAYS, LOG_EMERG, "kevent failed: %s", strerror(errno));
-               abort();
-       }
-
-       int rerr = (int)result[0].data;
-       int werr = (int)result[1].data;
-
-       if((rerr && rerr != ENOENT) || (werr && werr != ENOENT)) {
-               logger(DEBUG_ALWAYS, LOG_EMERG, "kevent errors: %s, %s", strerror(rerr), strerror(werr));
-               abort();
-       }
-
-       if(!flags) {
-               io_tree.generation++;
-       }
-
-#endif
-
-#ifdef HAVE_SELECT
-
-       if(flags & IO_READ) {
-               FD_SET(io->fd, &readfds);
-       } else {
-               FD_CLR(io->fd, &readfds);
-       }
-
-       if(flags & IO_WRITE) {
-               FD_SET(io->fd, &writefds);
-       } else {
-               FD_CLR(io->fd, &writefds);
-       }
-
-#endif
-
-#else
-       long events = 0;
-
-       if(flags & IO_WRITE) {
-               events |= WRITE_EVENTS;
-       }
-
-       if(flags & IO_READ) {
-               events |= READ_EVENTS;
-       }
-
-       if(WSAEventSelect(io->fd, io->event, events) != 0) {
-               abort();
-       }
-
-#endif
-}
-
-void io_del(io_t *io) {
-       if(!io->cb) {
-               return;
-       }
-
-       io_set(io, 0);
-#ifdef HAVE_WINDOWS
-
-       if(io->fd != -1 && WSACloseEvent(io->event) == FALSE) {
-               abort();
-       }
-
-       event_count--;
-#endif
-
-#if HAVE_SELECT
-       splay_unlink_node(&io_tree, &io->node);
-#endif
-       io->cb = NULL;
-}
+splay_tree_t io_tree = {.compare = (splay_compare_t)io_compare};
+splay_tree_t timeout_tree = {.compare = (splay_compare_t)timeout_compare};
 
 void timeout_add(timeout_t *timeout, timeout_cb_t cb, void *data, const struct timeval *tv) {
        timeout->cb = cb;
@@ -342,79 +110,7 @@ void timeout_del(timeout_t *timeout) {
        };
 }
 
-#ifndef HAVE_WINDOWS
-
-// From Matz's Ruby
-#ifndef NSIG
-# define NSIG (_SIGMAX + 1)      /* For QNX */
-#endif
-
-
-static io_t signalio;
-static int pipefd[2] = {-1, -1};
-static signal_t *signal_handle[NSIG + 1] = {NULL};
-
-static void signal_handler(int signum) {
-       unsigned char num = signum;
-
-       if(write(pipefd[1], &num, 1) != 1) {
-               // Pipe full or broken, nothing we can do about it.
-       }
-}
-
-static void signalio_handler(void *data, int flags) {
-       (void)data;
-       (void)flags;
-       unsigned char signum;
-
-       if(read(pipefd[0], &signum, 1) != 1) {
-               return;
-       }
-
-       signal_t *sig = signal_handle[signum];
-
-       if(sig) {
-               sig->cb(sig->data);
-       }
-}
-
-static void pipe_init(void) {
-       if(!pipe(pipefd)) {
-               io_add(&signalio, signalio_handler, NULL, pipefd[0], IO_READ);
-       }
-}
-
-void signal_add(signal_t *sig, signal_cb_t cb, void *data, int signum) {
-       if(sig->cb) {
-               return;
-       }
-
-       sig->signum = signum;
-       sig->cb = cb;
-       sig->data = data;
-
-       if(pipefd[0] == -1) {
-               pipe_init();
-       }
-
-       signal(signum, signal_handler);
-
-       signal_handle[signum] = sig;
-}
-
-void signal_del(signal_t *sig) {
-       if(!sig->cb) {
-               return;
-       }
-
-       signal(sig->signum, SIG_DFL);
-
-       signal_handle[sig->signum] = NULL;
-       sig->cb = NULL;
-}
-#endif
-
-static struct timeval *timeout_execute(struct timeval *diff) {
+struct timeval *timeout_execute(struct timeval *diff) {
        gettimeofday(&now, NULL);
        struct timeval *tv = NULL;
 
@@ -436,258 +132,3 @@ static struct timeval *timeout_execute(struct timeval *diff) {
 
        return tv;
 }
-
-bool event_loop(void) {
-       event_init();
-       running = true;
-
-#ifndef HAVE_WINDOWS
-
-#ifdef HAVE_SELECT
-       fd_set readable;
-       fd_set writable;
-#endif
-
-       while(running) {
-               struct timeval diff;
-               struct timeval *tv = timeout_execute(&diff);
-
-#ifdef HAVE_SELECT
-               memcpy(&readable, &readfds, sizeof(readable));
-               memcpy(&writable, &writefds, sizeof(writable));
-#endif
-
-#ifdef HAVE_EPOLL
-               struct epoll_event events[MAX_EVENTS_PER_LOOP];
-               long timeout = (tv->tv_sec * 1000) + (tv->tv_usec / 1000);
-
-               if(timeout > INT_MAX) {
-                       timeout = INT_MAX;
-               }
-
-               int n = epoll_wait(event_fd, events, MAX_EVENTS_PER_LOOP, (int)timeout);
-#endif
-
-#ifdef HAVE_KQUEUE
-               struct kevent events[MAX_EVENTS_PER_LOOP];
-
-               const struct timespec ts = {
-                       .tv_sec = tv->tv_sec,
-                       .tv_nsec = tv->tv_usec * 1000,
-               };
-
-               int n = kevent(event_fd, NULL, 0, events, MAX_EVENTS_PER_LOOP, &ts);
-#endif
-
-#ifdef HAVE_SELECT
-               int maxfds =  0;
-
-               if(io_tree.tail) {
-                       io_t *last = io_tree.tail->data;
-                       maxfds = last->fd + 1;
-               }
-
-               int n = select(maxfds, &readable, &writable, NULL, tv);
-#endif
-
-               if(n < 0) {
-                       if(sockwouldblock(sockerrno)) {
-                               continue;
-                       } else {
-                               return false;
-                       }
-               }
-
-               if(!n) {
-                       continue;
-               }
-
-               unsigned int curgen = io_tree.generation;
-
-
-#ifdef HAVE_EPOLL
-
-               for(int i = 0; i < n; i++) {
-                       io_t *io = events[i].data.ptr;
-
-                       if(events[i].events & EPOLLOUT && io->flags & IO_WRITE) {
-                               io->cb(io->data, IO_WRITE);
-                       }
-
-                       if(curgen != io_tree.generation) {
-                               break;
-                       }
-
-                       if(events[i].events & EPOLLIN && io->flags & IO_READ) {
-                               io->cb(io->data, IO_READ);
-                       }
-
-                       if(curgen != io_tree.generation) {
-                               break;
-                       }
-               }
-
-#endif
-
-#ifdef HAVE_KQUEUE
-
-               for(int i = 0; i < n; i++) {
-                       const struct kevent *evt = &events[i];
-                       const io_t *io = evt->udata;
-
-                       if(evt->filter == EVFILT_WRITE) {
-                               io->cb(io->data, IO_WRITE);
-                       } else if(evt->filter == EVFILT_READ) {
-                               io->cb(io->data, IO_READ);
-                       } else {
-                               continue;
-                       }
-
-                       if(curgen != io_tree.generation) {
-                               break;
-                       }
-               }
-
-#endif
-
-#ifdef HAVE_SELECT
-
-               for splay_each(io_t, io, &io_tree) {
-                       if(FD_ISSET(io->fd, &writable)) {
-                               io->cb(io->data, IO_WRITE);
-                       } else if(FD_ISSET(io->fd, &readable)) {
-                               io->cb(io->data, IO_READ);
-                       } else {
-                               continue;
-                       }
-
-                       /*
-                               There are scenarios in which the callback will remove another io_t from the tree
-                               (e.g. closing a double connection). Since splay_each does not support that, we
-                               need to exit the loop if that happens. That's okay, since any remaining events will
-                               get picked up by the next select() call.
-                       */
-                       if(curgen != io_tree.generation) {
-                               break;
-                       }
-               }
-
-#endif
-       }
-
-#else
-       assert(WSA_WAIT_EVENT_0 == 0);
-
-       while(running) {
-               struct timeval diff;
-               struct timeval *tv = timeout_execute(&diff);
-               DWORD timeout_ms = tv ? (DWORD)(tv->tv_sec * 1000 + tv->tv_usec / 1000 + 1) : WSA_INFINITE;
-
-               if(!event_count) {
-                       Sleep(timeout_ms);
-                       continue;
-               }
-
-               /*
-                  For some reason, Microsoft decided to make the FD_WRITE event edge-triggered instead of level-triggered,
-                  which is the opposite of what select() does. In practice, that means that if a FD_WRITE event triggers,
-                  it will never trigger again until a send() returns EWOULDBLOCK. Since the semantics of this event loop
-                  is that write events are level-triggered (i.e. they continue firing until the socket is full), we need
-                  to emulate these semantics by making sure we fire each IO_WRITE that is still writeable.
-
-                  Note that technically FD_CLOSE has the same problem, but it's okay because user code does not rely on
-                  this event being fired again if ignored.
-               */
-               unsigned int curgen = io_tree.generation;
-
-               for splay_each(io_t, io, &io_tree) {
-                       if(io->flags & IO_WRITE && send(io->fd, NULL, 0, 0) == 0) {
-                               io->cb(io->data, IO_WRITE);
-
-                               if(curgen != io_tree.generation) {
-                                       break;
-                               }
-                       }
-               }
-
-               if(event_count > WSA_MAXIMUM_WAIT_EVENTS) {
-                       WSASetLastError(WSA_INVALID_PARAMETER);
-                       return(false);
-               }
-
-               WSAEVENT events[WSA_MAXIMUM_WAIT_EVENTS];
-               io_t *io_map[WSA_MAXIMUM_WAIT_EVENTS];
-               DWORD event_index = 0;
-
-               for splay_each(io_t, io, &io_tree) {
-                       events[event_index] = io->event;
-                       io_map[event_index] = io;
-                       event_index++;
-               }
-
-               /*
-                * If the generation number changes due to event addition
-                * or removal by a callback we restart the loop.
-                */
-               curgen = io_tree.generation;
-
-               for(DWORD event_offset = 0; event_offset < event_count;) {
-                       DWORD result = WSAWaitForMultipleEvents(event_count - event_offset, &events[event_offset], FALSE, timeout_ms, FALSE);
-
-                       if(result == WSA_WAIT_TIMEOUT) {
-                               break;
-                       }
-
-                       if(result >= event_count - event_offset) {
-                               return false;
-                       }
-
-                       /* Look up io in the map by index. */
-                       event_index = result + event_offset;
-                       io_t *io = io_map[event_index];
-
-                       if(io->fd == -1) {
-                               io->cb(io->data, 0);
-
-                               if(curgen != io_tree.generation) {
-                                       break;
-                               }
-                       } else {
-                               WSANETWORKEVENTS network_events;
-
-                               if(WSAEnumNetworkEvents(io->fd, io->event, &network_events) != 0) {
-                                       return(false);
-                               }
-
-                               if(network_events.lNetworkEvents & READ_EVENTS) {
-                                       io->cb(io->data, IO_READ);
-
-                                       if(curgen != io_tree.generation) {
-                                               break;
-                                       }
-                               }
-
-                               /*
-                                   The fd might be available for write too. However, if we already fired the read callback, that
-                                   callback might have deleted the io (e.g. through terminate_connection()), so we can't fire the
-                                   write callback here. Instead, we loop back and let the writable io loop above handle it.
-                                */
-                       }
-
-                       /* Continue checking the rest of the events. */
-                       event_offset = event_index + 1;
-
-                       /* Just poll the next time through. */
-                       timeout_ms = 0;
-               }
-       }
-
-#endif
-
-       event_deinit();
-       return true;
-}
-
-void event_exit(void) {
-       running = false;
-}
index 1443ebc..386c0b5 100644 (file)
@@ -56,6 +56,9 @@ typedef struct signal_t {
 
 extern struct timeval now;
 
+extern splay_tree_t io_tree;
+extern splay_tree_t timeout_tree;
+
 extern void io_add(io_t *io, io_cb_t cb, void *data, int fd, int flags);
 #ifdef HAVE_WINDOWS
 extern void io_add_event(io_t *io, io_cb_t cb, void *data, WSAEVENT event);
@@ -66,6 +69,7 @@ extern void io_set(io_t *io, int flags);
 extern void timeout_add(timeout_t *timeout, timeout_cb_t cb, void *data, const struct timeval *tv);
 extern void timeout_del(timeout_t *timeout);
 extern void timeout_set(timeout_t *timeout, const struct timeval *tv);
+extern struct timeval *timeout_execute(struct timeval *diff);
 
 extern void signal_add(signal_t *sig, signal_cb_t cb, void *data, int signum);
 extern void signal_del(signal_t *sig);
diff --git a/src/event_select.c b/src/event_select.c
new file mode 100644 (file)
index 0000000..cac61e3
--- /dev/null
@@ -0,0 +1,140 @@
+/*
+    event_select.c -- select(2) support
+    Copyright (C) 2012-2022 Guus Sliepen <guus@tinc-vpn.org>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+*/
+
+#include "system.h"
+
+#include "event.h"
+#include "utils.h"
+
+static bool running = false;
+static fd_set readfds;
+static fd_set writefds;
+
+void io_add(io_t *io, io_cb_t cb, void *data, int fd, int flags) {
+       if(io->cb) {
+               return;
+       }
+
+       io->fd = fd;
+       io->cb = cb;
+       io->data = data;
+       io->node.data = io;
+
+       io_set(io, flags);
+
+       if(!splay_insert_node(&io_tree, &io->node)) {
+               abort();
+       }
+}
+
+void io_set(io_t *io, int flags) {
+       if(flags == io->flags) {
+               return;
+       }
+
+       io->flags = flags;
+
+       if(io->fd == -1) {
+               return;
+       }
+
+       if(flags & IO_READ) {
+               FD_SET(io->fd, &readfds);
+       } else {
+               FD_CLR(io->fd, &readfds);
+       }
+
+       if(flags & IO_WRITE) {
+               FD_SET(io->fd, &writefds);
+       } else {
+               FD_CLR(io->fd, &writefds);
+       }
+}
+
+void io_del(io_t *io) {
+       if(io->cb) {
+               io_set(io, 0);
+               splay_unlink_node(&io_tree, &io->node);
+               io->cb = NULL;
+       }
+}
+
+bool event_loop(void) {
+       running = true;
+
+       fd_set readable;
+       fd_set writable;
+
+       while(running) {
+               struct timeval diff;
+               struct timeval *tv = timeout_execute(&diff);
+
+               memcpy(&readable, &readfds, sizeof(readable));
+               memcpy(&writable, &writefds, sizeof(writable));
+
+               int maxfds =  0;
+
+               if(io_tree.tail) {
+                       io_t *last = io_tree.tail->data;
+                       maxfds = last->fd + 1;
+               }
+
+               int n = select(maxfds, &readable, &writable, NULL, tv);
+
+               if(n < 0) {
+                       if(sockwouldblock(sockerrno)) {
+                               continue;
+                       } else {
+                               return false;
+                       }
+               }
+
+               if(!n) {
+                       continue;
+               }
+
+               unsigned int curgen = io_tree.generation;
+
+               for splay_each(io_t, io, &io_tree) {
+                       if(FD_ISSET(io->fd, &writable)) {
+                               io->cb(io->data, IO_WRITE);
+                       } else if(FD_ISSET(io->fd, &readable)) {
+                               io->cb(io->data, IO_READ);
+                       } else {
+                               continue;
+                       }
+
+                       /*
+                           There are scenarios in which the callback will remove another io_t from the tree
+                           (e.g. closing a double connection). Since splay_each does not support that, we
+                           need to exit the loop if that happens. That's okay, since any remaining events will
+                           get picked up by the next select() call.
+                       */
+                       if(curgen != io_tree.generation) {
+                               break;
+                       }
+               }
+       }
+
+       return true;
+}
+
+void event_exit(void) {
+       running = false;
+}
diff --git a/src/linux/event.c b/src/linux/event.c
new file mode 100644 (file)
index 0000000..ebcccfe
--- /dev/null
@@ -0,0 +1,169 @@
+/*
+    event.c -- epoll support for Linux
+    Copyright (C) 2012-2022 Guus Sliepen <guus@tinc-vpn.org>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+*/
+
+#include "../system.h"
+
+#include <sys/epoll.h>
+
+#include "../event.h"
+#include "../utils.h"
+#include "../net.h"
+
+static bool running = false;
+static int epollset = 0;
+
+/* NOTE: 1024 limit is only used on ancient (pre 2.6.27) kernels.
+   Decent kernels will ignore this value making it unlimited.
+   epoll_create1 might be better, but these kernels would not be supported
+   in that case. */
+static inline void event_init(void) {
+       if(!epollset) {
+               epollset = epoll_create(1024);
+
+               if(epollset == -1) {
+                       logger(DEBUG_ALWAYS, LOG_EMERG, "Could not initialize epoll: %s", strerror(errno));
+                       abort();
+               }
+       }
+}
+
+static void event_deinit(void) {
+       if(epollset) {
+               close(epollset);
+               epollset = 0;
+       }
+}
+
+void io_add(io_t *io, io_cb_t cb, void *data, int fd, int flags) {
+       if(io->cb) {
+               return;
+       }
+
+       io->fd = fd;
+       io->cb = cb;
+       io->data = data;
+       io->node.data = io;
+
+       io_set(io, flags);
+}
+
+void io_set(io_t *io, int flags) {
+       event_init();
+
+       if(flags == io->flags) {
+               return;
+       }
+
+       io->flags = flags;
+
+       if(io->fd == -1) {
+               return;
+       }
+
+       epoll_ctl(epollset, EPOLL_CTL_DEL, io->fd, NULL);
+
+       struct epoll_event ev = {
+               .events = 0,
+               .data.ptr = io,
+       };
+
+       if(flags & IO_READ) {
+               ev.events |= EPOLLIN;
+       }
+
+       if(flags & IO_WRITE) {
+               ev.events |= EPOLLOUT;
+       } else if(ev.events == 0) {
+               io_tree.generation++;
+               return;
+       }
+
+       if(epoll_ctl(epollset, EPOLL_CTL_ADD, io->fd, &ev) < 0) {
+               logger(DEBUG_ALWAYS, LOG_EMERG, "epoll_ctl failed: %s", strerror(errno));
+               abort();
+       }
+}
+
+void io_del(io_t *io) {
+       if(io->cb) {
+               io_set(io, 0);
+               io->cb = NULL;
+       }
+}
+
+bool event_loop(void) {
+       event_init();
+       running = true;
+
+       while(running) {
+               struct timeval diff;
+               struct timeval *tv = timeout_execute(&diff);
+
+               struct epoll_event events[MAX_EVENTS_PER_LOOP];
+               long timeout = (tv->tv_sec * 1000) + (tv->tv_usec / 1000);
+
+               if(timeout > INT_MAX) {
+                       timeout = INT_MAX;
+               }
+
+               int n = epoll_wait(epollset, events, MAX_EVENTS_PER_LOOP, (int)timeout);
+
+               if(n < 0) {
+                       if(sockwouldblock(sockerrno)) {
+                               continue;
+                       } else {
+                               return false;
+                       }
+               }
+
+               if(!n) {
+                       continue;
+               }
+
+               unsigned int curgen = io_tree.generation;
+
+               for(int i = 0; i < n; i++) {
+                       io_t *io = events[i].data.ptr;
+
+                       if(events[i].events & EPOLLOUT && io->flags & IO_WRITE) {
+                               io->cb(io->data, IO_WRITE);
+                       }
+
+                       if(curgen != io_tree.generation) {
+                               break;
+                       }
+
+                       if(events[i].events & EPOLLIN && io->flags & IO_READ) {
+                               io->cb(io->data, IO_READ);
+                       }
+
+                       if(curgen != io_tree.generation) {
+                               break;
+                       }
+               }
+
+       }
+
+       event_deinit();
+       return true;
+}
+
+void event_exit(void) {
+       running = false;
+}
index 1b94f95..5c5a273 100644 (file)
@@ -1,12 +1,14 @@
 check_headers += [
   'linux/if_tun.h',
   'netpacket/packet.h',
-  'sys/epoll.h',
 ]
 
 check_functions += 'recvmmsg'
 
-src_tincd += files('device.c')
+src_tincd += files(
+  'device.c',
+  'event.c',
+)
 
 dep_libsystemd = dependency('libsystemd', required: opt_systemd)
 if dep_libsystemd.found()
index d9f7b14..cc64628 100644 (file)
@@ -174,6 +174,12 @@ src_tincd = [
   'subnet.c',
 ]
 
+src_event_select = files('event_select.c')
+
+if os_name != 'windows'
+  src_tincd += 'signal.c'
+endif
+
 cc_flags_tinc = cc_flags
 cc_flags_tincd = cc_flags
 
diff --git a/src/signal.c b/src/signal.c
new file mode 100644 (file)
index 0000000..73f7219
--- /dev/null
@@ -0,0 +1,90 @@
+/*
+    signal.c -- signal handling
+    Copyright (C) 2012-2022 Guus Sliepen <guus@tinc-vpn.org>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+*/
+
+#include "system.h"
+
+#include "event.h"
+
+// From Matz's Ruby
+#ifndef NSIG
+# define NSIG (_SIGMAX + 1)      /* For QNX */
+#endif
+
+static io_t signalio;
+static int pipefd[2] = {-1, -1};
+static signal_t *signal_handle[NSIG + 1] = {NULL};
+
+static void signal_handler(int signum) {
+       unsigned char num = signum;
+
+       if(write(pipefd[1], &num, 1) != 1) {
+               // Pipe full or broken, nothing we can do about it.
+       }
+}
+
+static void signalio_handler(void *data, int flags) {
+       (void)data;
+       (void)flags;
+       unsigned char signum;
+
+       if(read(pipefd[0], &signum, 1) != 1) {
+               return;
+       }
+
+       signal_t *sig = signal_handle[signum];
+
+       if(sig) {
+               sig->cb(sig->data);
+       }
+}
+
+static void pipe_init(void) {
+       if(!pipe(pipefd)) {
+               io_add(&signalio, signalio_handler, NULL, pipefd[0], IO_READ);
+       }
+}
+
+void signal_add(signal_t *sig, signal_cb_t cb, void *data, int signum) {
+       if(sig->cb) {
+               return;
+       }
+
+       sig->signum = signum;
+       sig->cb = cb;
+       sig->data = data;
+
+       if(pipefd[0] == -1) {
+               pipe_init();
+       }
+
+       signal(signum, signal_handler);
+
+       signal_handle[signum] = sig;
+}
+
+void signal_del(signal_t *sig) {
+       if(!sig->cb) {
+               return;
+       }
+
+       signal(sig->signum, SIG_DFL);
+
+       signal_handle[sig->signum] = NULL;
+       sig->cb = NULL;
+}
index 5d071ff..6247af2 100644 (file)
@@ -1,4 +1,7 @@
-src_tincd += files('device.c')
+src_tincd += [
+  files('device.c'),
+  src_event_select,
+]
 
 deps_common += cc.find_library('libsocket')
 
diff --git a/src/windows/event.c b/src/windows/event.c
new file mode 100644 (file)
index 0000000..a4688d0
--- /dev/null
@@ -0,0 +1,223 @@
+/*
+    event.c -- event support for Windows
+    Copyright (C) 2012-2022 Guus Sliepen <guus@tinc-vpn.org>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+*/
+
+#include "../system.h"
+
+#include <assert.h>
+
+#include "../event.h"
+#include "../utils.h"
+#include "../net.h"
+
+static bool running = false;
+static DWORD event_count = 0;
+
+static const long READ_EVENTS = FD_READ | FD_ACCEPT | FD_CLOSE;
+static const long WRITE_EVENTS = FD_WRITE | FD_CONNECT;
+
+void io_add(io_t *io, io_cb_t cb, void *data, int fd, int flags) {
+       if(io->cb) {
+               return;
+       }
+
+       io->fd = fd;
+
+       if(io->fd != -1) {
+               io->event = WSACreateEvent();
+
+               if(io->event == WSA_INVALID_EVENT) {
+                       abort();
+               }
+       }
+
+       event_count++;
+
+       io->cb = cb;
+       io->data = data;
+       io->node.data = io;
+
+       io_set(io, flags);
+
+       if(!splay_insert_node(&io_tree, &io->node)) {
+               abort();
+       }
+}
+
+void io_add_event(io_t *io, io_cb_t cb, void *data, WSAEVENT event) {
+       io->event = event;
+       io_add(io, cb, data, -1, 0);
+}
+
+void io_set(io_t *io, int flags) {
+       if(flags == io->flags) {
+               return;
+       }
+
+       io->flags = flags;
+
+       if(io->fd == -1) {
+               return;
+       }
+
+       long events = 0;
+
+       if(flags & IO_WRITE) {
+               events |= WRITE_EVENTS;
+       }
+
+       if(flags & IO_READ) {
+               events |= READ_EVENTS;
+       }
+
+       if(WSAEventSelect(io->fd, io->event, events) != 0) {
+               abort();
+       }
+}
+
+void io_del(io_t *io) {
+       if(!io->cb) {
+               return;
+       }
+
+       io_set(io, 0);
+
+       if(io->fd != -1 && WSACloseEvent(io->event) == FALSE) {
+               abort();
+       }
+
+       event_count--;
+       splay_unlink_node(&io_tree, &io->node);
+       io->cb = NULL;
+}
+
+bool event_loop(void) {
+       running = true;
+
+       assert(WSA_WAIT_EVENT_0 == 0);
+
+       while(running) {
+               struct timeval diff;
+               struct timeval *tv = timeout_execute(&diff);
+               DWORD timeout_ms = tv ? (DWORD)(tv->tv_sec * 1000 + tv->tv_usec / 1000 + 1) : WSA_INFINITE;
+
+               if(!event_count) {
+                       Sleep(timeout_ms);
+                       continue;
+               }
+
+               /*
+                  For some reason, Microsoft decided to make the FD_WRITE event edge-triggered instead of level-triggered,
+                  which is the opposite of what select() does. In practice, that means that if a FD_WRITE event triggers,
+                  it will never trigger again until a send() returns EWOULDBLOCK. Since the semantics of this event loop
+                  is that write events are level-triggered (i.e. they continue firing until the socket is full), we need
+                  to emulate these semantics by making sure we fire each IO_WRITE that is still writeable.
+
+                  Note that technically FD_CLOSE has the same problem, but it's okay because user code does not rely on
+                  this event being fired again if ignored.
+               */
+               unsigned int curgen = io_tree.generation;
+
+               for splay_each(io_t, io, &io_tree) {
+                       if(io->flags & IO_WRITE && send(io->fd, NULL, 0, 0) == 0) {
+                               io->cb(io->data, IO_WRITE);
+
+                               if(curgen != io_tree.generation) {
+                                       break;
+                               }
+                       }
+               }
+
+               if(event_count > WSA_MAXIMUM_WAIT_EVENTS) {
+                       WSASetLastError(WSA_INVALID_PARAMETER);
+                       return(false);
+               }
+
+               WSAEVENT events[WSA_MAXIMUM_WAIT_EVENTS];
+               io_t *io_map[WSA_MAXIMUM_WAIT_EVENTS];
+               DWORD event_index = 0;
+
+               for splay_each(io_t, io, &io_tree) {
+                       events[event_index] = io->event;
+                       io_map[event_index] = io;
+                       event_index++;
+               }
+
+               /*
+                * If the generation number changes due to event addition
+                * or removal by a callback we restart the loop.
+                */
+               curgen = io_tree.generation;
+
+               for(DWORD event_offset = 0; event_offset < event_count;) {
+                       DWORD result = WSAWaitForMultipleEvents(event_count - event_offset, &events[event_offset], FALSE, timeout_ms, FALSE);
+
+                       if(result == WSA_WAIT_TIMEOUT) {
+                               break;
+                       }
+
+                       if(result >= event_count - event_offset) {
+                               return false;
+                       }
+
+                       /* Look up io in the map by index. */
+                       event_index = result + event_offset;
+                       io_t *io = io_map[event_index];
+
+                       if(io->fd == -1) {
+                               io->cb(io->data, 0);
+
+                               if(curgen != io_tree.generation) {
+                                       break;
+                               }
+                       } else {
+                               WSANETWORKEVENTS network_events;
+
+                               if(WSAEnumNetworkEvents(io->fd, io->event, &network_events) != 0) {
+                                       return(false);
+                               }
+
+                               if(network_events.lNetworkEvents & READ_EVENTS) {
+                                       io->cb(io->data, IO_READ);
+
+                                       if(curgen != io_tree.generation) {
+                                               break;
+                                       }
+                               }
+
+                               /*
+                                   The fd might be available for write too. However, if we already fired the read callback, that
+                                   callback might have deleted the io (e.g. through terminate_connection()), so we can't fire the
+                                   write callback here. Instead, we loop back and let the writable io loop above handle it.
+                                */
+                       }
+
+                       /* Continue checking the rest of the events. */
+                       event_offset = event_index + 1;
+
+                       /* Just poll the next time through. */
+                       timeout_ms = 0;
+               }
+       }
+
+       return true;
+}
+
+void event_exit(void) {
+       running = false;
+}
index b78b901..40f7013 100644 (file)
@@ -15,5 +15,9 @@ foreach libname : win_common_libs
 endforeach
 
 src_lib_common += files('random.c')
-src_tincd += files('device.c')
+
+src_tincd += files(
+  'device.c',
+  'event.c',
+)