[wing/wip/nacho/poll: 2/2] Add wing_poll
- From: Ignacio Casal Quinteiro <icq src gnome org>
- To: commits-list gnome org
- Cc:
- Subject: [wing/wip/nacho/poll: 2/2] Add wing_poll
- Date: Wed, 5 Dec 2018 13:57:58 +0000 (UTC)
commit 1de9e035b0c3e849ff03c5aaddde457792f9b5b4
Author: Ignacio Casal Quinteiro <qignacio amazon com>
Date: Fri Nov 30 17:02:00 2018 +0100
Add wing_poll
This is a new polling method allowing to poll more than 64 handles
based on the glib one.
When we reach the limit of 64 we create a thread and we poll
on that thread for a batch of handles this way we overcome the limit.
tests/meson.build | 5 +-
tests/poll.c | 619 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
wing/meson.build | 2 +
wing/wing.h | 1 +
wing/wingpoll.c | 403 +++++++++++++++++++++++++++++++++++
wing/wingpoll.h | 46 ++++
6 files changed, 1075 insertions(+), 1 deletion(-)
---
diff --git a/tests/meson.build b/tests/meson.build
index fbbfc14..9079a7e 100644
--- a/tests/meson.build
+++ b/tests/meson.build
@@ -1,10 +1,13 @@
unit_tests = [
'named-pipe',
+ 'poll',
]
+winsock2_dep = cc.find_library('ws2_32')
+
foreach unit: unit_tests
exe = executable(unit, unit + '.c',
- dependencies: wing_dep,
+ dependencies: [ wing_dep, winsock2_dep ],
include_directories: wing_inc)
test(unit, exe, args: [ '--tap', '-k' ])
endforeach
diff --git a/tests/poll.c b/tests/poll.c
new file mode 100644
index 0000000..2e372da
--- /dev/null
+++ b/tests/poll.c
@@ -0,0 +1,619 @@
+/* Unit test for W32 version of wing_poll()
+ *
+ * Copyright © 2017 Руслан Ижбулатов <lrn1986 gmail com>
+ *
+ * This is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this library; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+#define _WIN32_WINNT 0x0600
+
+#include <Winsock2.h>
+#include <wing/wing.h>
+#include <glib.h>
+
+#define NUM_POLLEES 999
+#define NUM_POLLFDS 1000
+
+#define ASYNC_CONNECT_OK(r) (r == 0 || (r < 0 && GetLastError () == WSAEWOULDBLOCK))
+
+#define REPEAT 1
+
+static void
+init_networking (void)
+{
+ WSADATA wsadata;
+
+ if (WSAStartup (MAKEWORD (2, 0), &wsadata) != 0)
+ g_error ("Windows Sockets could not be initialized");
+}
+
+static void
+prepare_fds (SOCKET sockets[],
+ GPollFD fds[],
+ int num_pollees)
+{
+ gint i;
+
+ for (i = 0; i < num_pollees; i++)
+ {
+ fds[i].fd = (gintptr) WSACreateEvent ();
+ g_assert (WSAEventSelect (sockets[i], (HANDLE) fds[i].fd, FD_READ | FD_CLOSE) == 0);
+ }
+}
+
+static void
+reset_fds (GPollFD fds[],
+ int num_pollees)
+{
+ gint i;
+
+ for (i = 0; i < num_pollees; i++)
+ {
+ WSAResetEvent ((HANDLE) fds[i].fd);
+ fds[i].events = G_IO_IN | G_IO_OUT | G_IO_ERR;
+ fds[i].revents = 0;
+ }
+}
+
+static void
+reset_fds_msg (GPollFD fds[],
+ int num_pollfds)
+{
+ fds[num_pollfds - 1].fd = G_WIN32_MSG_HANDLE;
+ fds[num_pollfds - 1].events = G_IO_IN;
+ fds[num_pollfds - 1].revents = 0;
+}
+
+static void
+check_fds (SOCKET sockets[],
+ GPollFD fds[],
+ int num_pollees)
+{
+ gint i;
+
+ for (i = 0; i < num_pollees; i++)
+ {
+ if (fds[i].revents != 0)
+ {
+ WSANETWORKEVENTS events;
+ g_assert (WSAEnumNetworkEvents (sockets[i], 0, &events) == 0);
+
+ fds[i].revents = 0;
+ if (events.lNetworkEvents & (FD_READ | FD_ACCEPT))
+ fds[i].revents |= G_IO_IN;
+
+ if (events.lNetworkEvents & FD_WRITE)
+ fds[i].revents |= G_IO_OUT;
+ else
+ {
+ /* We have called WSAEnumNetworkEvents() above but it didn't
+ * set FD_WRITE.
+ */
+ if (events.lNetworkEvents & FD_CONNECT)
+ {
+ if (events.iErrorCode[FD_CONNECT_BIT] == 0)
+ fds[i].revents |= G_IO_OUT;
+ else
+ fds[i].revents |= (G_IO_HUP | G_IO_ERR);
+ }
+ if (fds[i].revents == 0 && (events.lNetworkEvents & (FD_CLOSE)))
+ fds[i].revents |= G_IO_HUP;
+ }
+ }
+ }
+}
+
+static void
+prepare_sockets (SOCKET sockets[],
+ SOCKET opp_sockets[],
+ GPollFD fds[],
+ int num_pollees)
+{
+ gint i;
+ SOCKET server;
+ struct sockaddr_in sa;
+ unsigned long ul = 1;
+ int sa_size;
+ int r;
+
+ server = socket (AF_INET, SOCK_STREAM, IPPROTO_TCP);
+ g_assert (server != INVALID_SOCKET);
+
+ memset(&sa, 0, sizeof sa);
+
+ sa.sin_family = AF_INET;
+ sa.sin_port = 0;
+ sa.sin_addr.s_addr = htonl (INADDR_LOOPBACK);
+ sa_size = sizeof (sa);
+
+ g_assert (bind (server, (const struct sockaddr *) &sa, sa_size) == 0);
+ g_assert (getsockname (server, (struct sockaddr *) &sa, &sa_size) == 0);
+ g_assert (listen (server, 1) == 0);
+
+ for (i = 0; i < num_pollees; i++)
+ {
+ opp_sockets[i] = socket (AF_INET, SOCK_STREAM, IPPROTO_TCP);
+ g_assert (opp_sockets[i] != INVALID_SOCKET);
+ g_assert (ioctlsocket (opp_sockets[i], FIONBIO, &ul) == 0);
+
+ r = connect (opp_sockets[i], (const struct sockaddr *) &sa, sizeof (sa));
+ g_assert (ASYNC_CONNECT_OK (r));
+
+ sockets[i] = accept (server, NULL, NULL);
+ g_assert (sockets[i] != INVALID_SOCKET);
+ g_assert (ioctlsocket (sockets[i], FIONBIO, &ul) == 0);
+ }
+
+ closesocket (server);
+}
+
+static void
+cleanup_sockets (SOCKET sockets[],
+ SOCKET opp_sockets[],
+ int num_pollees)
+{
+ gint i;
+
+ for (i = 0; i < num_pollees; i++)
+ {
+ closesocket (sockets[i]);
+ closesocket (opp_sockets[i]);
+ }
+}
+
+static void
+bucketize (gint64 val,
+ gint buckets[],
+ gint64 bucket_limits[],
+ gint count)
+{
+ gint i;
+
+ if (val > bucket_limits[count - 1])
+ {
+ buckets[count - 1] += 1;
+ return;
+ }
+
+ for (i = count - 1; i > 0; i--)
+ if (val < bucket_limits[i] && val >= bucket_limits[i - 1])
+ {
+ buckets[i] += 1;
+ return;
+ }
+
+ buckets[0] += 1;
+}
+
+static void
+print_buckets (gint buckets[],
+ gint64 bucket_limits[],
+ gint count)
+{
+ gint i;
+
+ for (i = 0; i < count; i++)
+ if (i < count - 1)
+ g_print ("%-4lld-%4lld|", i == 0 ? 0 : bucket_limits[i - 1], bucket_limits[i] - 1);
+ else
+ g_print (" >= %-4lld|", bucket_limits[i - 1]);
+
+ g_print ("\n");
+
+ for (i = 0; i < count; i++)
+ {
+ gint len;
+ gint padding;
+ gint j;
+ if (buckets[i] < 10)
+ len = 1;
+ else if (buckets[i] < 100)
+ len = 2;
+ else if (buckets[i] < 1000)
+ len = 3;
+ else
+ len = 4;
+ padding = 9 - len;
+ for (j = 0; j < padding / 2; j++)
+ g_print (" ");
+ if (buckets[i] != 0)
+ g_print ("%*d", len, buckets[i]);
+ else
+ g_print (" ");
+ for (j = padding / 2; j < padding; j++)
+ g_print (" ");
+ g_print (" ");
+ }
+
+ g_print ("\n\n");
+}
+
+static void
+test_wing_poll (void)
+{
+ SOCKET sockets[NUM_POLLEES];
+ GPollFD fds[NUM_POLLFDS];
+ SOCKET opp_sockets[NUM_POLLEES];
+ gint i;
+ gint activatable;
+ gint64 times[REPEAT][2];
+#define BUCKET_COUNT 25
+ gint64 bucket_limits[BUCKET_COUNT] = {3, 5, 10, 15, 20, 25, 30, 35, 40, 50, 60, 70, 80, 90, 100, 120, 150,
180, 220, 280, 350, 450, 600, 800, 1000};
+ gint buckets[BUCKET_COUNT];
+ gint64 times_avg = 0, times_min = G_MAXINT64, times_max = 0;
+
+ prepare_sockets (sockets, opp_sockets, fds, NUM_POLLEES);
+ prepare_fds (sockets, fds, NUM_POLLEES);
+
+ times_avg = 0;
+ times_min = G_MAXINT64;
+ times_max = 0;
+ memset (buckets, 0, sizeof (gint) * BUCKET_COUNT);
+
+ for (i = 0; i < REPEAT; i++)
+ {
+ gint r;
+ gint64 diff;
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ times[i][0] = g_get_monotonic_time ();
+ r = wing_poll (fds, NUM_POLLFDS, 0);
+ times[i][1] = g_get_monotonic_time ();
+ g_assert (r == 0);
+ diff = times[i][1] - times[i][0];
+ if (times_min > diff)
+ times_min = diff;
+ if (times_max < diff)
+ times_max = diff;
+ times_avg += diff;
+ bucketize (diff, buckets, bucket_limits, BUCKET_COUNT);
+ }
+
+ times_avg /= NUM_POLLEES;
+ g_print ("\nempty poll time:\n%4lldns - %4lldns, average %4lldns\n", times_min, times_max, times_avg);
+ print_buckets (buckets, bucket_limits, BUCKET_COUNT);
+
+ times_avg = 0;
+ times_min = G_MAXINT64;
+ times_max = 0;
+ memset (buckets, 0, sizeof (gint) * BUCKET_COUNT);
+
+ activatable = 0;
+
+ for (i = 0; i < REPEAT; i++)
+ {
+ gint r, s, v, t;
+ gint64 diff;
+ MSG msg;
+ gboolean found_app;
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ s = send (opp_sockets[activatable], (const char *) &t, 1, 0);
+ g_assert (PostMessage (NULL, WM_APP, 1, 2));
+ /* This is to ensure that all sockets catch up, otherwise some might not poll active */
+ g_usleep (G_USEC_PER_SEC / 1000);
+
+ times[i][0] = g_get_monotonic_time ();
+ r = wing_poll (fds, NUM_POLLFDS, 1000);
+ times[i][1] = g_get_monotonic_time ();
+
+ check_fds (sockets, fds, NUM_POLLEES);
+ v = recv (sockets[activatable], (char *) &t, 1, 0);
+ found_app = FALSE;
+ while (!found_app && PeekMessage (&msg, NULL, 0, 0, PM_REMOVE))
+ if (msg.message == WM_APP && msg.wParam == 1 && msg.lParam == 2)
+ found_app = TRUE;
+ g_assert (s == 1);
+ g_assert (r == 2);
+ g_assert (v == 1);
+ g_assert (found_app);
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ r = wing_poll (fds, NUM_POLLFDS, 0);
+ check_fds (sockets, fds, NUM_POLLEES);
+ g_assert (r == 0);
+ diff = times[i][1] - times[i][0];
+ if (times_min > diff)
+ times_min = diff;
+ if (times_max < diff)
+ times_max = diff;
+ times_avg += diff;
+ activatable = (activatable + 1) % NUM_POLLEES;
+ bucketize (diff, buckets, bucket_limits, BUCKET_COUNT);
+ }
+
+ times_avg /= NUM_POLLEES;
+ g_print ("1-socket + msg poll time:\n%4lldns - %4lldns, average %4lldns\n", times_min, times_max,
times_avg);
+ print_buckets (buckets, bucket_limits, BUCKET_COUNT);
+
+ times_avg = 0;
+ times_min = G_MAXINT64;
+ times_max = 0;
+ memset (buckets, 0, sizeof (gint) * BUCKET_COUNT);
+
+ activatable = 0;
+
+ for (i = 0; i < REPEAT; i++)
+ {
+ gint r, s, v, t;
+ gint64 diff;
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ s = send (opp_sockets[activatable], (const char *) &t, 1, 0);
+
+ g_usleep (G_USEC_PER_SEC / 1000);
+
+ times[i][0] = g_get_monotonic_time ();
+ r = wing_poll (fds, NUM_POLLFDS, 1000);
+ times[i][1] = g_get_monotonic_time ();
+
+ check_fds (sockets, fds, NUM_POLLEES);
+ v = recv (sockets[activatable], (char *) &t, 1, 0);
+ g_assert (s == 1);
+ g_assert (r == 1);
+ g_assert (v == 1);
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ r = wing_poll (fds, NUM_POLLFDS, 0);
+ check_fds (sockets, fds, NUM_POLLEES);
+ g_assert (r == 0);
+
+ diff = times[i][1] - times[i][0];
+ if (times_min > diff)
+ times_min = diff;
+ if (times_max < diff)
+ times_max = diff;
+ times_avg += diff;
+ activatable = (activatable + 1) % NUM_POLLEES;
+ bucketize (diff, buckets, bucket_limits, BUCKET_COUNT);
+ }
+
+ times_avg /= NUM_POLLEES;
+ g_print ("1-socket poll time:\n%4lldns - %4lldns, average %4lldns\n", times_min, times_max, times_avg);
+ print_buckets (buckets, bucket_limits, BUCKET_COUNT);
+
+ times_avg = 0;
+ times_min = G_MAXINT64;
+ times_max = 0;
+ memset (buckets, 0, sizeof (gint) * BUCKET_COUNT);
+
+ for (i = 0; i < REPEAT; i++)
+ {
+ gint r, s, v, t;
+ gint64 diff;
+ gint j;
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ s = v = 0;
+
+ for (j = 0; j < NUM_POLLEES / 2; j++)
+ s += send (opp_sockets[j], (const char *) &t, 1, 0) == 1 ? 1 : 0;
+
+ g_usleep (G_USEC_PER_SEC / 1000);
+
+ times[i][0] = g_get_monotonic_time ();
+ r = wing_poll (fds, NUM_POLLFDS, 1000);
+ times[i][1] = g_get_monotonic_time ();
+ check_fds (sockets, fds, NUM_POLLEES);
+ for (j = 0; j < NUM_POLLEES / 2; j++)
+ v += recv (sockets[j], (char *) &t, 1, 0) == 1 ? 1 : 0;
+ g_assert (s == NUM_POLLEES / 2);
+ g_assert (r == NUM_POLLEES / 2);
+ g_assert (v == NUM_POLLEES / 2);
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ r = wing_poll (fds, NUM_POLLFDS, 0);
+ check_fds (sockets, fds, NUM_POLLEES);
+ g_assert (r == 0);
+
+ diff = times[i][1] - times[i][0];
+ if (times_min > diff)
+ times_min = diff;
+ if (times_max < diff)
+ times_max = diff;
+ times_avg += diff;
+ bucketize (diff, buckets, bucket_limits, BUCKET_COUNT);
+ }
+
+ times_avg /= NUM_POLLEES;
+ g_print ("half-socket poll time:\n%4lldns - %4lldns, average %4lldns\n", times_min, times_max, times_avg);
+ print_buckets (buckets, bucket_limits, BUCKET_COUNT);
+
+ times_avg = 0;
+ times_min = G_MAXINT64;
+ times_max = 0;
+ memset (buckets, 0, sizeof (gint) * BUCKET_COUNT);
+
+ for (i = 0; i < REPEAT; i++)
+ {
+ gint r, s, v, t;
+ gint64 diff;
+ gint j;
+ MSG msg;
+ gboolean found_app;
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ s = v = 0;
+
+ for (j = 0; j < NUM_POLLEES / 2; j++)
+ s += send (opp_sockets[j], (const char *) &t, 1, 0) == 1 ? 1 : 0;
+ g_assert (PostMessage (NULL, WM_APP, 1, 2));
+
+ /* This is to ensure that all sockets catch up, otherwise some might not poll active */
+ g_usleep (G_USEC_PER_SEC / 1000);
+
+ times[i][0] = g_get_monotonic_time ();
+ r = wing_poll (fds, NUM_POLLFDS, 1000);
+ times[i][1] = g_get_monotonic_time ();
+ check_fds (sockets, fds, NUM_POLLEES);
+ for (j = 0; j < NUM_POLLEES / 2; j++)
+ v += recv (sockets[j], (char *) &t, 1, 0) == 1 ? 1 : 0;
+ found_app = FALSE;
+ while (!found_app && PeekMessage (&msg, NULL, 0, 0, PM_REMOVE))
+ if (msg.message == WM_APP && msg.wParam == 1 && msg.lParam == 2)
+ found_app = TRUE;
+ g_assert (s == NUM_POLLEES / 2);
+ g_assert (r == NUM_POLLEES / 2 + 1);
+ g_assert (v == NUM_POLLEES / 2);
+ g_assert (found_app);
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ r = wing_poll (fds, NUM_POLLFDS, 0);
+ check_fds (sockets, fds, NUM_POLLEES);
+ g_assert (r == 0);
+
+ diff = times[i][1] - times[i][0];
+ if (times_min > diff)
+ times_min = diff;
+ if (times_max < diff)
+ times_max = diff;
+ times_avg += diff;
+ bucketize (diff, buckets, bucket_limits, BUCKET_COUNT);
+ }
+
+ times_avg /= NUM_POLLEES;
+ g_print ("half-socket + msg poll time:\n%4lldns - %4lldns, average %4lldns\n", times_min, times_max,
times_avg);
+ print_buckets (buckets, bucket_limits, BUCKET_COUNT);
+
+ times_avg = 0;
+ times_min = G_MAXINT64;
+ times_max = 0;
+ memset (buckets, 0, sizeof (gint) * BUCKET_COUNT);
+
+ for (i = 0; i < REPEAT; i++)
+ {
+ gint r, s, v, t;
+ gint64 diff;
+ gint j;
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ s = v = 0;
+
+ for (j = 0; j < NUM_POLLEES; j++)
+ s += send (opp_sockets[j], (const char *) &t, 1, 0) == 1 ? 1 : 0;
+
+ g_usleep (G_USEC_PER_SEC / 1000);
+
+ times[i][0] = g_get_monotonic_time ();
+ r = wing_poll (fds, NUM_POLLFDS, 1000);
+ times[i][1] = g_get_monotonic_time ();
+ check_fds (sockets, fds, NUM_POLLEES);
+ for (j = 0; j < NUM_POLLEES; j++)
+ v += recv (sockets[j], (char *) &t, 1, 0) == 1 ? 1 : 0;
+ g_assert (s == NUM_POLLEES);
+ g_assert (r == NUM_POLLEES);
+ g_assert (v == NUM_POLLEES);
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ r = wing_poll (fds, NUM_POLLFDS, 0);
+ check_fds (sockets, fds, NUM_POLLEES);
+ g_assert (r == 0);
+
+ diff = times[i][1] - times[i][0];
+ if (times_min > diff)
+ times_min = diff;
+ if (times_max < diff)
+ times_max = diff;
+ times_avg += diff;
+ bucketize (diff, buckets, bucket_limits, BUCKET_COUNT);
+ }
+
+ times_avg /= NUM_POLLEES;
+ g_print ("%d-socket poll time: \n%4lldns - %4lldns, average %4lldns\n", NUM_POLLEES, times_min, times_max,
times_avg);
+ print_buckets (buckets, bucket_limits, BUCKET_COUNT);
+
+ activatable = 0;
+ times_avg = 0;
+ times_min = G_MAXINT64;
+ times_max = 0;
+ memset (buckets, 0, sizeof (gint) * BUCKET_COUNT);
+
+ for (i = 0; i < REPEAT; i++)
+ {
+ gint r, s, v, t;
+ gint64 diff;
+ gint j;
+ MSG msg;
+ gboolean found_app;
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ s = v = 0;
+
+ for (j = 0; j < activatable; j++)
+ s += send (opp_sockets[j], (const char *) &t, 1, 0) == 1 ? 1 : 0;
+ g_assert (PostMessage (NULL, WM_APP, 1, 2));
+
+ g_usleep (G_USEC_PER_SEC / 1000);
+
+ times[i][0] = g_get_monotonic_time ();
+ r = wing_poll (fds, NUM_POLLFDS, 1000);
+ times[i][1] = g_get_monotonic_time ();
+ check_fds (sockets, fds, NUM_POLLEES);
+ for (j = 0; j < activatable; j++)
+ v += recv (sockets[j], (char *) &t, 1, 0) == 1 ? 1 : 0;
+ found_app = FALSE;
+ while (!found_app && PeekMessage (&msg, NULL, 0, 0, PM_REMOVE))
+ if (msg.message == WM_APP && msg.wParam == 1 && msg.lParam == 2)
+ found_app = TRUE;
+ g_assert (s == activatable);
+ g_assert (r == activatable + 1);
+ g_assert (v == activatable);
+ g_assert (found_app);
+
+ reset_fds (fds, NUM_POLLEES);
+ reset_fds_msg (fds, NUM_POLLFDS);
+ r = wing_poll (fds, NUM_POLLFDS, 0);
+ check_fds (sockets, fds, NUM_POLLEES);
+ g_assert (r == 0);
+
+ diff = times[i][1] - times[i][0];
+ if (times_min > diff)
+ times_min = diff;
+ if (times_max < diff)
+ times_max = diff;
+ times_avg += diff;
+ bucketize (diff, buckets, bucket_limits, BUCKET_COUNT);
+ activatable = (activatable + 1) % NUM_POLLEES;
+ }
+
+ times_avg /= NUM_POLLEES;
+ g_print ("variable socket number + msg poll time: \n%4lldns - %4lldns, average %4lldns\n", times_min,
times_max, times_avg);
+ print_buckets (buckets, bucket_limits, BUCKET_COUNT);
+
+ cleanup_sockets (sockets, opp_sockets, NUM_POLLEES);
+}
+
+int
+main (int argc,
+ char *argv[])
+{
+ g_test_init (&argc, &argv, NULL);
+ init_networking ();
+
+ g_test_add_func ("/poll/wing-poll", test_wing_poll);
+
+ return g_test_run ();
+}
diff --git a/wing/meson.build b/wing/meson.build
index 596e283..8da554b 100644
--- a/wing/meson.build
+++ b/wing/meson.build
@@ -10,6 +10,7 @@ headers = [
'wingnamedpipeconnection.h',
'wingnamedpipelistener.h',
'wingoutputstream.h',
+ 'wingpoll.h',
'wingservice.h',
'wingservicemanager.h',
'wingsource.h',
@@ -27,6 +28,7 @@ sources = [
'wingnamedpipeconnection.c',
'wingnamedpipelistener.c',
'wingoutputstream.c',
+ 'wingpoll.c',
'wingservice.c',
'wingservice-private.h',
'wingservicemanager.c',
diff --git a/wing/wing.h b/wing/wing.h
index 7113f06..ce3badd 100644
--- a/wing/wing.h
+++ b/wing/wing.h
@@ -29,6 +29,7 @@
#include <wing/wingnamedpipelistener.h>
#include <wing/wingnamedpipeconnection.h>
#include <wing/wingoutputstream.h>
+#include <wing/wingpoll.h>
#include <wing/wingservice.h>
#include <wing/wingservicemanager.h>
#include <wing/wingsource.h>
diff --git a/wing/wingpoll.c b/wing/wingpoll.c
new file mode 100644
index 0000000..0b63aa6
--- /dev/null
+++ b/wing/wingpoll.c
@@ -0,0 +1,403 @@
+/*
+ * Copyright (C) 1995-1997 Peter Mattis, Spencer Kimball and Josh MacDonald
+ *
+ * gpoll.c: poll(2) abstraction
+ * Copyright 1998 Owen Taylor
+ * Copyright 2008 Red Hat, Inc.
+ * Copyright (C) 2018 NICE s.r.l.
+ *
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+/*
+ * Modified by the GLib Team and others 1997-2000. See the AUTHORS
+ * file for a list of people on the GLib Team. See the ChangeLog
+ * files for a list of changes. These files are distributed with
+ * GLib at ftp://ftp.gtk.org/pub/gtk/.
+ */
+
+#include "wingpoll.h"
+
+#ifdef G_OS_WIN32
+#define STRICT
+#include <windows.h>
+#include <process.h>
+#endif /* G_OS_WIN32 */
+
+#ifdef _WIN32
+/* Always enable debugging printout on Windows, as it is more often
+ * needed there...
+ */
+#define G_MAIN_POLL_DEBUG
+#endif
+
+#ifdef G_MAIN_POLL_DEBUG
+static gboolean _wing_main_poll_debug = FALSE;
+#endif
+
+static int
+poll_rest (GPollFD *msg_fd,
+ GPollFD *stop_fd,
+ HANDLE *handles,
+ GPollFD *handle_to_fd[],
+ gint nhandles,
+ gint timeout)
+{
+ DWORD ready;
+ GPollFD *f;
+ int recursed_result;
+
+ if (msg_fd != NULL)
+ {
+ /* Wait for either messages or handles
+ * -> Use MsgWaitForMultipleObjectsEx
+ */
+ if (_wing_main_poll_debug)
+ g_print (" MsgWaitForMultipleObjectsEx(%d, %d)\n", nhandles, timeout);
+
+ ready = MsgWaitForMultipleObjectsEx (nhandles, handles, timeout,
+ QS_ALLINPUT, MWMO_ALERTABLE);
+
+ if (ready == WAIT_FAILED)
+ {
+ gchar *emsg;
+
+ emsg = g_win32_error_message (GetLastError ());
+ g_warning ("MsgWaitForMultipleObjectsEx failed: %s", emsg);
+ g_free (emsg);
+ }
+ }
+ else if (nhandles == 0)
+ {
+ /* No handles to wait for, just the timeout */
+ if (timeout == INFINITE)
+ ready = WAIT_FAILED;
+ else
+ {
+ /* Wait for the current process to die, more efficient than SleepEx(). */
+ WaitForSingleObjectEx (GetCurrentProcess (), timeout, TRUE);
+ ready = WAIT_TIMEOUT;
+ }
+ }
+ else
+ {
+ /* Wait for just handles
+ * -> Use WaitForMultipleObjectsEx
+ */
+ if (_wing_main_poll_debug)
+ g_print (" WaitForMultipleObjectsEx(%d, %d)\n", nhandles, timeout);
+
+ ready = WaitForMultipleObjectsEx (nhandles, handles, FALSE, timeout, TRUE);
+ if (ready == WAIT_FAILED)
+ {
+ gchar *emsg;
+
+ emsg = g_win32_error_message (GetLastError ());
+ g_warning ("WaitForMultipleObjectsEx failed: %s", emsg);
+ g_free (emsg);
+ }
+ }
+
+ if (_wing_main_poll_debug)
+ g_print (" wait returns %ld%s\n",
+ ready,
+ (ready == WAIT_FAILED ? " (WAIT_FAILED)" :
+ (ready == WAIT_TIMEOUT ? " (WAIT_TIMEOUT)" :
+ (msg_fd != NULL && ready == WAIT_OBJECT_0 + nhandles ? " (msg)" : ""))));
+
+ if (ready == WAIT_FAILED)
+ return -1;
+ else if (ready == WAIT_TIMEOUT ||
+ ready == WAIT_IO_COMPLETION)
+ return 0;
+ else if (msg_fd != NULL && ready == WAIT_OBJECT_0 + nhandles)
+ {
+ msg_fd->revents |= G_IO_IN;
+
+ /* If we have a timeout, or no handles to poll, be satisfied
+ * with just noticing we have messages waiting.
+ */
+ if (timeout != 0 || nhandles == 0)
+ return 1;
+
+ /* If no timeout and handles to poll, recurse to poll them,
+ * too.
+ */
+ recursed_result = poll_rest (NULL, stop_fd, handles, handle_to_fd, nhandles, 0);
+ return (recursed_result == -1) ? -1 : 1 + recursed_result;
+ }
+ else if (ready >= WAIT_OBJECT_0 && ready < WAIT_OBJECT_0 + nhandles)
+ {
+ int retval;
+
+ f = handle_to_fd[ready - WAIT_OBJECT_0];
+ f->revents = f->events;
+ if (_wing_main_poll_debug)
+ g_print (" got event %p\n", (HANDLE) f->fd);
+
+ /* Do not count the stop_fd */
+ retval = (f != stop_fd) ? 1 : 0;
+
+ /* If no timeout and polling several handles, recurse to poll
+ * the rest of them.
+ */
+ if (timeout == 0 && nhandles > 1)
+ {
+ /* Poll the handles with index > ready */
+ HANDLE *shorter_handles;
+ GPollFD **shorter_handle_to_fd;
+ gint shorter_nhandles;
+
+ shorter_handles = &handles[ready - WAIT_OBJECT_0 + 1];
+ shorter_handle_to_fd = &handle_to_fd[ready - WAIT_OBJECT_0 + 1];
+ shorter_nhandles = nhandles - (ready - WAIT_OBJECT_0 + 1);
+
+ recursed_result = poll_rest (NULL, stop_fd, shorter_handles, shorter_handle_to_fd,
shorter_nhandles, 0);
+ return (recursed_result == -1) ? -1 : retval + recursed_result;
+ }
+ return retval;
+ }
+
+ return 0;
+}
+
+typedef struct
+{
+ HANDLE handles[MAXIMUM_WAIT_OBJECTS];
+ GPollFD *handle_to_fd[MAXIMUM_WAIT_OBJECTS];
+ GPollFD *msg_fd;
+ GPollFD *stop_fd;
+ gint nhandles;
+ gint timeout;
+} GWin32PollThreadData;
+
+static gint
+poll_single_thread (GWin32PollThreadData *data)
+{
+ int retval;
+
+ /* Polling for several things? */
+ if (data->nhandles > 1 || (data->nhandles > 0 && data->msg_fd != NULL))
+ {
+ /* First check if one or several of them are immediately
+ * available
+ */
+ retval = poll_rest (data->msg_fd, data->stop_fd, data->handles, data->handle_to_fd, data->nhandles, 0);
+
+ /* If not, and we have a significant timeout, poll again with
+ * timeout then. Note that this will return indication for only
+ * one event, or only for messages.
+ */
+ if (retval == 0 && (data->timeout == INFINITE || data->timeout > 0))
+ retval = poll_rest (data->msg_fd, data->stop_fd, data->handles, data->handle_to_fd, data->nhandles,
data->timeout);
+ }
+ else
+ {
+ /* Just polling for one thing, so no need to check first if
+ * available immediately
+ */
+ retval = poll_rest (data->msg_fd, data->stop_fd, data->handles, data->handle_to_fd, data->nhandles,
data->timeout);
+ }
+
+ return retval;
+}
+
+static void
+fill_poll_thread_data (GPollFD *fds,
+ guint nfds,
+ gint timeout,
+ GPollFD *stop_fd,
+ GWin32PollThreadData *data)
+{
+ GPollFD *f;
+
+ data->timeout = timeout;
+
+ if (stop_fd != NULL)
+ {
+ if (_wing_main_poll_debug)
+ g_print (" Stop FD: %p", (HANDLE) stop_fd->fd);
+
+ data->stop_fd = stop_fd;
+ data->handle_to_fd[data->nhandles] = stop_fd;
+ data->handles[data->nhandles++] = (HANDLE) stop_fd->fd;
+ }
+
+ for (f = fds; f < &fds[nfds]; ++f)
+ {
+ if ((data->nhandles == MAXIMUM_WAIT_OBJECTS) ||
+ (data->msg_fd != NULL && (data->nhandles == MAXIMUM_WAIT_OBJECTS - 1)))
+ {
+ g_warning ("Too many handles to wait for!");
+ break;
+ }
+
+ if (f->fd == G_WIN32_MSG_HANDLE && (f->events & G_IO_IN))
+ {
+ if (_wing_main_poll_debug && data->msg_fd == NULL)
+ g_print (" MSG");
+ data->msg_fd = f;
+ }
+ else if (f->fd > 0)
+ {
+ if (_wing_main_poll_debug)
+ g_print (" %p", (HANDLE) f->fd);
+ data->handle_to_fd[data->nhandles] = f;
+ data->handles[data->nhandles++] = (HANDLE) f->fd;
+ }
+
+ f->revents = 0;
+ }
+}
+
+static guint __stdcall
+poll_thread_run (gpointer user_data)
+{
+ GWin32PollThreadData *data = user_data;
+
+ /* Docs say that it is safer to call _endthreadex by our own */
+ _endthreadex (poll_single_thread (data));
+
+ g_assert_not_reached ();
+
+ return 0;
+}
+
+/* One slot for a possible msg object and another for the stop event */
+#define MAXIMUM_WAIT_OBJECTS_PER_THREAD (MAXIMUM_WAIT_OBJECTS - 2)
+
+gint
+wing_poll (GPollFD *fds,
+ guint nfds,
+ gint timeout)
+{
+ guint nthreads, threads_remain;
+ HANDLE thread_handles[MAXIMUM_WAIT_OBJECTS];
+ GWin32PollThreadData *threads_data;
+ GPollFD stop_event = { 0, };
+ GPollFD *f;
+ guint i, fds_idx = 0;
+ DWORD ready;
+ DWORD thread_retval;
+ int retval;
+ GPollFD *msg_fd = NULL;
+
+ if (timeout == -1)
+ timeout = INFINITE;
+
+ /* Simple case without extra threads */
+ if (nfds <= MAXIMUM_WAIT_OBJECTS)
+ {
+ GWin32PollThreadData data = { 0, };
+
+ if (_wing_main_poll_debug)
+ g_print ("wing_poll: waiting for");
+
+ fill_poll_thread_data (fds, nfds, timeout, NULL, &data);
+
+ if (_wing_main_poll_debug)
+ g_print ("\n");
+
+ retval = poll_single_thread (&data);
+ if (retval == -1)
+ for (f = fds; f < &fds[nfds]; ++f)
+ f->revents = 0;
+
+ return retval;
+ }
+
+ if (_wing_main_poll_debug)
+ g_print ("wing_poll: polling with threads\n");
+
+ nthreads = nfds / MAXIMUM_WAIT_OBJECTS_PER_THREAD;
+ threads_remain = nfds % MAXIMUM_WAIT_OBJECTS_PER_THREAD;
+ if (threads_remain > 0)
+ nthreads++;
+
+ if (nthreads > MAXIMUM_WAIT_OBJECTS)
+ {
+ g_warning ("Too many handles to wait for in threads!");
+ nthreads = MAXIMUM_WAIT_OBJECTS;
+ }
+
+#if GLIB_SIZEOF_VOID_P == 8
+ stop_event.fd = (gint64)CreateEventW (NULL, TRUE, FALSE, NULL);
+#else
+ stop_event.fd = (gint)CreateEventW (NULL, TRUE, FALSE, NULL);
+#endif
+ stop_event.events = G_IO_IN;
+
+ threads_data = g_new0 (GWin32PollThreadData, nthreads);
+ for (i = 0; i < nthreads; i++)
+ {
+ guint thread_fds;
+ guint ignore;
+
+ if (i == (nthreads - 1) && threads_remain > 0)
+ thread_fds = threads_remain;
+ else
+ thread_fds = MAXIMUM_WAIT_OBJECTS_PER_THREAD;
+
+ fill_poll_thread_data (fds + fds_idx, thread_fds, timeout, &stop_event, &threads_data[i]);
+ fds_idx += thread_fds;
+
+ /* We must poll for messages from the same thread, so poll it along with the threads */
+ if (threads_data[i].msg_fd != NULL)
+ {
+ msg_fd = threads_data[i].msg_fd;
+ threads_data[i].msg_fd = NULL;
+ }
+
+ thread_handles[i] = (HANDLE) _beginthreadex (NULL, 0, poll_thread_run, &threads_data[i], 0, &ignore);
+ }
+
+ /* Wait for at least one thread to return */
+ if (msg_fd != NULL)
+ ready = MsgWaitForMultipleObjectsEx (nthreads, thread_handles, timeout,
+ QS_ALLINPUT, MWMO_ALERTABLE);
+ else
+ ready = WaitForMultipleObjects (nthreads, thread_handles, timeout > 0, timeout);
+
+ /* Signal the stop in case any of the threads did not stop yet */
+ SetEvent ((HANDLE)stop_event.fd);
+
+ /* Wait for the rest of the threads to finish */
+ WaitForMultipleObjects (nthreads, thread_handles, TRUE, INFINITE);
+
+ /* The return value of all the threads give us all the fds that changed state */
+ retval = 0;
+ if (msg_fd != NULL && ready == WAIT_OBJECT_0 + nthreads)
+ {
+ msg_fd->revents |= G_IO_IN;
+ retval = 1;
+ }
+
+ for (i = 0; i < nthreads; i++)
+ {
+ if (GetExitCodeThread (thread_handles[i], &thread_retval))
+ retval = retval == -1 ? -1 : thread_retval == -1 ? -1 : retval + thread_retval;
+
+ CloseHandle (thread_handles[i]);
+ }
+
+ if (retval == -1)
+ for (f = fds; f < &fds[nfds]; ++f)
+ f->revents = 0;
+
+ g_free (threads_data);
+ CloseHandle ((HANDLE)stop_event.fd);
+
+ return retval;
+}
diff --git a/wing/wingpoll.h b/wing/wingpoll.h
new file mode 100644
index 0000000..111cbb6
--- /dev/null
+++ b/wing/wingpoll.h
@@ -0,0 +1,46 @@
+/*
+ * Copyright (C) 1995-1997 Peter Mattis, Spencer Kimball and Josh MacDonald
+ *
+ * gpoll.c: poll(2) abstraction
+ * Copyright 1998 Owen Taylor
+ * Copyright 2008 Red Hat, Inc.
+ * Copyright (C) 2018 NICE s.r.l.
+ *
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+/*
+ * Modified by the GLib Team and others 1997-2000. See the AUTHORS
+ * file for a list of people on the GLib Team. See the ChangeLog
+ * files for a list of changes. These files are distributed with
+ * GLib at ftp://ftp.gtk.org/pub/gtk/.
+ */
+
+#ifndef WING_POLL_H
+#define WING_POLL_H
+
+#include <glib.h>
+#include <wing/wingversionmacros.h>
+
+G_BEGIN_DECLS
+
+WING_AVAILABLE_IN_ALL
+gint wing_poll (GPollFD *fds,
+ guint nfds,
+ gint timeout);
+
+G_END_DECLS
+
+#endif /* WING_UTILS_H */
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]
