Merge branch 'timeouts_v2_squashed'

.gitignore

@@ -180,6 +180,7 @@ uptime-*.json
 /src/test/test-memwipe
 /src/test/test-ntor-cl
 /src/test/test-switch-id
+/src/test/test-timers
 /src/test/test_workqueue
 /src/test/test.exe
 /src/test/test-slow.exe
@@ -188,6 +189,7 @@ uptime-*.json
 /src/test/test-ntor-cl.exe
 /src/test/test-memwipe.exe
 /src/test/test-switch-id.exe
+/src/test/test-timers.exe
 /src/test/test_workqueue.exe
 /src/test/test_zero_length_keys.sh
 /src/test/test_ntor.sh

changes/timeouts

@@ -0,0 +1,7 @@
+  o Minor features (infrastructure):
+    - Tor now includes an improved timer backend, so that we can efficiently
+      support tens or hundreds of thousands of concurrent timers, as will be
+      needed for some of our planned anti-traffic-analysis work. This code
+      is based on William Ahern's "timeout.c" project, which implements
+      a "tickless hierarchical timing wheel". Closes ticket #18365.
+

src/common/include.am

@@ -94,7 +94,9 @@ LIBOR_CRYPTO_A_SOURCES = \
 
 LIBOR_EVENT_A_SOURCES = \
 	src/common/compat_libevent.c \
-	src/common/procmon.c
+	src/common/procmon.c         \
+	src/common/timers.c          \
+	src/ext/timeouts/timeout.c
 
 src_common_libor_a_SOURCES = $(LIBOR_A_SOURCES)
 src_common_libor_crypto_a_SOURCES = $(LIBOR_CRYPTO_A_SOURCES)
@@ -134,6 +136,7 @@ COMMONHEADERS = \
   src/common/procmon.h				\
   src/common/sandbox.h				\
   src/common/testsupport.h			\
+  src/common/timers.h				\
   src/common/torgzip.h				\
   src/common/torint.h				\
   src/common/torlog.h				\

src/common/timers.c

@@ -0,0 +1,297 @@
+/* Copyright (c) 2016, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+/**
+ * \file timers.c
+ * \brief Wrapper around William Ahern's fast hierarchical timer wheel
+ *   implementation, to tie it in with a libevent backend.
+ *
+ * Only use these functions from the main thread.
+ *
+ * The main advantage of tor_timer_t over using libevent's timers is that
+ * they're way more efficient if we need to have thousands or millions of
+ * them.  For more information, see
+ *   http://www.25thandclement.com/~william/projects/timeout.c.html
+ *
+ * Periodic timers are available in the backend, but I've turned them off.
+ * We can turn them back on if needed.
+ */
+
+/* Notes:
+ *
+ * The use of tor_gettimeofday_cached_monotonic() is kind of ugly.  It would
+ * be neat to fix it.
+ *
+ * Having a way to free all timers on shutdown would free people from the
+ * need to track them.  Not sure if that's clever though.
+ *
+ * In an ideal world, Libevent would just switch to use this backend, and we
+ * could throw this file away.  But even if Libevent does switch, we'll be
+ * stuck with legacy libevents for some time.
+ */
+
+#include "orconfig.h"
+
+#include "compat.h"
+#include "compat_libevent.h"
+#include "timers.h"
+#include "torlog.h"
+#include "util.h"
+
+#ifdef HAVE_EVENT2_EVENT_H
+#include <event2/event.h>
+#else
+#include <event.h>
+#endif
+
+struct timeout_cb {
+  timer_cb_fn_t cb;
+  void *arg;
+};
+
+/*
+ * These definitions are for timeouts.c  and timeouts.h.
+ */
+#ifdef __GNUC__
+/* We're not exposing any of the functions outside this file. */
+#define TIMEOUT_PUBLIC __attribute__((__unused__)) static
+#else
+/* We're not exposing any of the functions outside this file. */
+#define TIMEOUT_PUBLIC static
+#endif
+/* We're not using periodic events. */
+#define TIMEOUT_DISABLE_INTERVALS
+/* We always know the global_timeouts object, so we don't need each timeout
+ * to keep a pointer to it. */
+#define TIMEOUT_DISABLE_RELATIVE_ACCESS
+/* We're providing our own struct timeout_cb. */
+#define TIMEOUT_CB_OVERRIDE
+/* We're going to support timers that are pretty far out in advance. Making
+ * this big can be inefficient, but having a significant number of timers
+ * above TIMEOUT_MAX can also be super-inefficent. Choosing 5 here sets
+ * timeout_max to 2^30 ticks, or 29 hours with our value for USEC_PER_TICK */
+#define WHEEL_NUM 5
+#include "src/ext/timeouts/timeout.c"
+
+static struct timeouts *global_timeouts = NULL;
+static struct event *global_timer_event = NULL;
+
+/** We need to choose this value carefully.  Because we're using timer wheels,
+ * it actually costs us to have extra resolution we don't use.  So for now,
+ * I'm going to define our resolution as .1 msec, and hope that's good enough.
+ *
+ * Note that two of the most popular libevent backends (epoll without timerfd,
+ * and windows select), simply can't support sub-millisecond resolution,
+ * do this is optimistic for a lot of users.
+ */
+#define USEC_PER_TICK 100
+
+/** One million microseconds in a second */
+#define USEC_PER_SEC 1000000
+
+/** Check at least once every N seconds. */
+#define MIN_CHECK_SECONDS 3600
+
+/** Check at least once every N ticks. */
+#define MIN_CHECK_TICKS \
+  (((timeout_t)MIN_CHECK_SECONDS) * (1000000 / USEC_PER_TICK))
+
+/**
+ * Convert the timeval in <b>tv</b> to a timeout_t, and return it.
+ *
+ * The output resolution is set by USEC_PER_TICK, and the time corresponding
+ * to 0 is the same as the time corresponding to 0 from
+ * tor_gettimeofday_cached_monotonic().
+ */
+static timeout_t
+tv_to_timeout(const struct timeval *tv)
+{
+  uint64_t usec = tv->tv_usec;
+  usec += ((uint64_t)USEC_PER_SEC) * tv->tv_sec;
+  return usec / USEC_PER_TICK;
+}
+
+/**
+ * Convert the timeout in <b>t</b> to a timeval in <b>tv_out</b>
+ */
+static void
+timeout_to_tv(timeout_t t, struct timeval *tv_out)
+{
+  t *= USEC_PER_TICK;
+  tv_out->tv_usec = (int)(t % USEC_PER_SEC);
+  tv_out->tv_sec = (time_t)(t / USEC_PER_SEC);
+}
+
+/**
+ * Update the timer <b>tv</b> to the current time in <b>tv</b>.
+ */
+static void
+timer_advance_to_cur_time(const struct timeval *tv)
+{
+  timeout_t cur_tick = tv_to_timeout(tv);
+  if (BUG(cur_tick < timeouts_get_curtime(global_timeouts))) {
+    cur_tick = timeouts_get_curtime(global_timeouts); // LCOV_EXCL_LINE
+  }
+  timeouts_update(global_timeouts, cur_tick);
+}
+
+/**
+ * Adjust the time at which the libevent timer should fire based on
+ * the next-expiring time in <b>global_timeouts</b>
+ */
+static void
+libevent_timer_reschedule(void)
+{
+  struct timeval now;
+  tor_gettimeofday_cached_monotonic(&now);
+  timer_advance_to_cur_time(&now);
+
+  timeout_t delay = timeouts_timeout(global_timeouts);
+  struct timeval d;
+  if (delay > MIN_CHECK_TICKS)
+    delay = MIN_CHECK_TICKS;
+  timeout_to_tv(delay, &d);
+  event_add(global_timer_event, &d);
+}
+
+/**
+ * Invoked when the libevent timer has expired: see which tor_timer_t events
+ * have fired, activate their callbacks, and reschedule the libevent timer.
+ */
+static void
+libevent_timer_callback(evutil_socket_t fd, short what, void *arg)
+{
+  (void)fd;
+  (void)what;
+  (void)arg;
+
+  struct timeval now;
+  tor_gettimeofday_cache_clear();
+  tor_gettimeofday_cached_monotonic(&now);
+  timer_advance_to_cur_time(&now);
+
+  tor_timer_t *t;
+  while ((t = timeouts_get(global_timeouts))) {
+    t->callback.cb(t, t->callback.arg, &now);
+  }
+
+  tor_gettimeofday_cache_clear();
+  libevent_timer_reschedule();
+}
+
+/**
+ * Initialize the timers subsystem.  Requires that libevent has already been
+ * initialized.
+ */
+void
+timers_initialize(void)
+{
+  if (BUG(global_timeouts))
+    return; // LCOV_EXCL_LINE
+
+  timeout_error_t err;
+  global_timeouts = timeouts_open(0, &err);
+  if (!global_timeouts) {
+    // LCOV_EXCL_START -- this can only fail on malloc failure.
+    log_err(LD_BUG, "Unable to open timer backend: %s", strerror(err));
+    tor_assert(0);
+    // LCOV_EXCL_STOP
+  }
+
+  struct event *timer_event;
+  timer_event = tor_event_new(tor_libevent_get_base(),
+                              -1, 0, libevent_timer_callback, NULL);
+  tor_assert(timer_event);
+  global_timer_event = timer_event;
+
+  libevent_timer_reschedule();
+}
+
+/**
+ * Release all storage held in the timers subsystem.  Does not fire timers.
+ */
+void
+timers_shutdown(void)
+{
+  if (global_timer_event) {
+    tor_event_free(global_timer_event);
+    global_timer_event = NULL;
+  }
+  if (global_timeouts) {
+    timeouts_close(global_timeouts);
+    global_timeouts = NULL;
+  }
+}
+
+/**
+ * Allocate and return a new timer, with given callback and argument.
+ */
+tor_timer_t *
+timer_new(timer_cb_fn_t cb, void *arg)
+{
+  tor_timer_t *t = tor_malloc(sizeof(tor_timer_t));
+  timeout_init(t, 0);
+  timer_set_cb(t, cb, arg);
+  return t;
+}
+
+/**
+ * Release all storage held by <b>t</b>, and unschedule it if was already
+ * scheduled.
+ */
+void
+timer_free(tor_timer_t *t)
+{
+  if (! t)
+    return;
+
+  timeouts_del(global_timeouts, t);
+  tor_free(t);
+}
+
+/**
+ * Change the callback and argument associated with a timer <b>t</b>.
+ */
+void
+timer_set_cb(tor_timer_t *t, timer_cb_fn_t cb, void *arg)
+{
+  t->callback.cb = cb;
+  t->callback.arg = arg;
+}
+
+/**
+ * Schedule the timer t to fire at the current time plus a delay of <b>tv</b>.
+ * All times are relative to tor_gettimeofday_cached_monotonic.
+ */
+void
+timer_schedule(tor_timer_t *t, const struct timeval *tv)
+{
+  const timeout_t when = tv_to_timeout(tv);
+  struct timeval now;
+  tor_gettimeofday_cached_monotonic(&now);
+  timer_advance_to_cur_time(&now);
+
+  /* Take the old timeout value. */
+  timeout_t to = timeouts_timeout(global_timeouts);
+
+  timeouts_add(global_timeouts, t, when);
+
+  /* Should we update the libevent timer? */
+  if (to <= when) {
+    return; /* we're already going to fire before this timer would trigger. */
+  }
+  libevent_timer_reschedule();
+}
+
+/**
+ * Cancel the timer <b>t</b> if it is currently scheduled.  (It's okay to call
+ * this on an unscheduled timer.
+ */
+void
+timer_disable(tor_timer_t *t)
+{
+  timeouts_del(global_timeouts, t);
+  /* We don't reschedule the libevent timer here, since it's okay if it fires
+   * early. */
+}
+

src/common/timers.h

@@ -0,0 +1,22 @@
+/* Copyright (c) 2016, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+#ifndef TOR_TIMERS_H
+#define TOR_TIMERS_H
+
+#include "orconfig.h"
+#include "testsupport.h"
+
+typedef struct timeout tor_timer_t;
+typedef void (*timer_cb_fn_t)(tor_timer_t *, void *, const struct timeval *);
+tor_timer_t *timer_new(timer_cb_fn_t cb, void *arg);
+void timer_set_cb(tor_timer_t *t, timer_cb_fn_t cb, void *arg);
+void timer_schedule(tor_timer_t *t, const struct timeval *delay);
+void timer_disable(tor_timer_t *t);
+void timer_free(tor_timer_t *t);
+
+void timers_initialize(void);
+void timers_shutdown(void);
+
+#endif
+

src/ext/README

@@ -73,3 +73,7 @@ readpassphrase.[ch]
 
     Portable readpassphrase implementation from OpenSSH portable, version
     6.8p1.
+
+timeouts/
+
+    William Ahern's hierarchical timer-wheel implementation. MIT license.

src/ext/include.am

@@ -12,7 +12,11 @@ EXTHEADERS = \
   src/ext/strlcpy.c	\
   src/ext/tinytest_macros.h \
   src/ext/tor_queue.h	\
-  src/ext/siphash.h
+  src/ext/siphash.h \
+  src/ext/timeouts/timeout.h \
+  src/ext/timeouts/timeout-debug.h \
+  src/ext/timeouts/timeout-bitops.c \
+  src/ext/timeouts/timeout.c
 
 noinst_HEADERS+= $(EXTHEADERS)
 
@@ -148,3 +152,21 @@ noinst_HEADERS += $(LIBKECCAK_TINY_HDRS)
 LIBKECCAK_TINY=src/ext/keccak-tiny/libkeccak-tiny.a
 noinst_LIBRARIES += $(LIBKECCAK_TINY)
 
+EXTRA_DIST += \
+	timeouts/bench/bench-add.lua 		\
+	timeouts/bench/bench-aux.lua		\
+	timeouts/bench/bench.c			\
+	timeouts/bench/bench-del.lua		\
+	timeouts/bench/bench-expire.lua		\
+	timeouts/bench/bench.h			\
+	timeouts/bench/bench-heap.c		\
+	timeouts/bench/bench-llrb.c		\
+	timeouts/bench/bench.plt		\
+	timeouts/bench/bench-wheel.c		\
+	timeouts/bench/Rules.mk			\
+	timeouts/lua/Rules.mk			\
+	timeouts/lua/timeout-lua.c		\
+	timeouts/Makefile			\
+	timeouts/Rules.shrc			\
+	timeouts/test-timeout.c
+

src/ext/timeouts/Makefile

@@ -0,0 +1,68 @@
+# NOTE: GNU Make 3.81 won't export MAKEFLAGS if .POSIX is specified, but
+# Solaris make won't export MAKEFLAGS unless .POSIX is specified.
+$(firstword ignore).POSIX:
+
+.DEFAULT_GOAL = all
+
+.SUFFIXES:
+
+all:
+
+#
+# USER-MODIFIABLE MACROS
+#
+top_srcdir = .
+top_builddir = .
+
+CFLAGS = -O2 -march=native -g -Wall -Wextra -Wno-unused-parameter -Wno-unused-function
+SOFLAGS = $$(auto_soflags)
+LIBS = $$(auto_libs)
+
+ALL_CPPFLAGS = -I$(top_srcdir) -DWHEEL_BIT=$(WHEEL_BIT) -DWHEEL_NUM=$(WHEEL_NUM) $(CPPFLAGS)
+ALL_CFLAGS = $(CFLAGS)
+ALL_SOFLAGS = $(SOFLAGS)
+ALL_LDFLAGS = $(LDFLAGS)
+ALL_LIBS = $(LIBS)
+
+LUA_API = 5.3
+LUA = lua
+LUA51_CPPFLAGS = $(LUA_CPPFLAGS)
+LUA52_CPPFLAGS = $(LUA_CPPFLAGS)
+LUA53_CPPFLAGS = $(LUA_CPPFLAGS)
+
+WHEEL_BIT = 6
+WHEEL_NUM = 4
+
+RM = rm -f
+
+# END MACROS
+
+SHRC = \
+	top_srcdir="$(top_srcdir)"; \
+	top_builddir="$(top_builddir)"; \
+	. "$${top_srcdir}/Rules.shrc"
+
+LUA_APIS = 5.1 5.2 5.3
+
+include $(top_srcdir)/lua/Rules.mk
+include $(top_srcdir)/bench/Rules.mk
+
+all: test-timeout
+
+timeout.o: $(top_srcdir)/timeout.c
+test-timeout.o: $(top_srcdir)/test-timeout.c
+
+timeout.o test-timeout.o:
+	@$(SHRC); echo_cmd $(CC) $(ALL_CFLAGS) -c -o $@ $${top_srcdir}/$(@F:%.o=%.c) $(ALL_CPPFLAGS)
+
+test-timeout: timeout.o test-timeout.o
+	@$(SHRC); echo_cmd $(CC) $(ALL_CPPFLAGS) $(ALL_CFLAGS) -o $@ timeout.o test-timeout.o
+
+.PHONY: clean clean~
+
+clean:
+	$(RM) $(top_builddir)/test-timeout $(top_builddir)/*.o
+	$(RM) -r $(top_builddir)/*.dSYM
+
+clean~:
+	find $(top_builddir) $(top_srcdir) -name "*~" -exec $(RM) -- {} "+"

src/ext/timeouts/Rules.shrc

@@ -0,0 +1,40 @@
+# convert to absolute paths
+top_srcdir="$(cd "${top_srcdir}" && pwd -L)"
+top_builddir="$(cd "${top_builddir}" && pwd -L)"
+
+# Paths for Lua modules (benchmarks and installed modules)
+export LUA_CPATH="${top_builddir}/lua/5.1/?.so;${top_builddir}/bench/?.so;;"
+export LUA_PATH="${top_srcdir}/lua/?.lua;${top_srcdir}/bench/?.lua;;"
+export LUA_CPATH_5_2="${top_builddir}/lua/5.2/?.so;${top_builddir}/bench/?.so;;"
+export LUA_PATH_5_2="${top_srcdir}/lua/?.lua;${top_srcdir}/bench/?.lua;;"
+export LUA_CPATH_5_3="${top_builddir}/lua/5.3/?.so;${top_builddir}/bench/?.so;;"
+export LUA_PATH_5_3="${top_srcdir}/lua/?.lua;${top_srcdir}/bench/?.lua;;"
+
+# preserve stdout so we can print commands to terminal
+exec 9>&1;
+echo_cmd() {
+	printf "%s\n" "$*" >&9;
+	"$@";
+}
+
+auto_soflags() {
+	case "$(uname -s)" in
+	Darwin)
+		printf -- "-bundle -undefined dynamic_lookup"
+		;;
+	*)
+		printf -- "-fPIC -shared"
+		;;
+	esac
+}
+
+auto_libs() {
+	case "$(uname -s)" in
+	Linux)
+		printf -- "-lrt"
+		;;
+	*)
+		;;
+	esac
+}
+

src/ext/timeouts/bench/Rules.mk

@@ -0,0 +1,49 @@
+BENCH_MODS = bench.so $(BENCH_ALGOS:%=bench-%.so)
+BENCH_ALGOS = wheel heap llrb
+BENCH_OPS = add del expire
+
+$(top_builddir)/bench/bench.so: $(top_srcdir)/bench/bench.c
+$(top_builddir)/bench/bench-wheel.so: $(top_srcdir)/bench/bench-wheel.c
+$(top_builddir)/bench/bench-heap.so: $(top_srcdir)/bench/bench-heap.c
+$(top_builddir)/bench/bench-llrb.so: $(top_srcdir)/bench/bench-llrb.c
+
+$(BENCH_MODS:%=$(top_builddir)/bench/%): $(top_srcdir)/timeout.h $(top_srcdir)/timeout.c $(top_srcdir)/bench/bench.h
+	mkdir -p $(@D)
+	@$(SHRC); echo_cmd $(CC) -o $@ $(top_srcdir)/bench/$(@F:%.so=%.c) $(ALL_CPPFLAGS) $(ALL_CFLAGS) $(ALL_SOFLAGS) $(ALL_LDFLAGS) $(ALL_LIBS)
+
+$(BENCH_OPS:%=$(top_builddir)/bench/wheel-%.dat): $(top_builddir)/bench/bench-wheel.so $(top_builddir)/bench/bench.so $(top_srcdir)/bench/bench-aux.lua
+$(BENCH_OPS:%=$(top_builddir)/bench/heap-%.dat): $(top_builddir)/bench/bench-heap.so $(top_builddir)/bench/bench.so $(top_srcdir)/bench/bench-aux.lua
+$(BENCH_OPS:%=$(top_builddir)/bench/llrb-%.dat): $(top_builddir)/bench/bench-llrb.so $(top_builddir)/bench/bench.so $(top_srcdir)/bench/bench-aux.lua
+
+$(BENCH_ALGOS:%=$(top_builddir)/bench/%-add.dat): $(top_srcdir)/bench/bench-add.lua
+	@$(SHRC); echo_cmd cd $(@D) && echo_cmd $(LUA) $${top_srcdir}/bench/bench-add.lua $${top_builddir}/bench/bench-$(@F:%-add.dat=%).so > $(@F).tmp
+	mv $@.tmp $@
+
+$(BENCH_ALGOS:%=$(top_builddir)/bench/%-del.dat): $(top_srcdir)/bench/bench-del.lua
+	@$(SHRC); echo_cmd cd $(@D) && echo_cmd $(LUA) $${top_srcdir}/bench/bench-del.lua $${top_builddir}/bench/bench-$(@F:%-del.dat=%).so > $(@F).tmp
+	mv $@.tmp $@
+
+$(BENCH_ALGOS:%=$(top_builddir)/bench/%-expire.dat): $(top_srcdir)/bench/bench-expire.lua
+	@$(SHRC); echo_cmd cd $(@D) && echo_cmd $(LUA) $${top_srcdir}/bench/bench-expire.lua $${top_builddir}/bench/bench-$(@F:%-expire.dat=%).so > $(@F).tmp
+	mv $@.tmp $@
+
+$(top_builddir)/bench/bench.eps: \
+	$(BENCH_OPS:%=$(top_builddir)/bench/wheel-%.dat) \
+	$(BENCH_OPS:%=$(top_builddir)/bench/heap-%.dat)
+#	$(BENCH_OPS:%=$(top_builddir)/bench/llrb-%.dat)
+
+$(top_builddir)/bench/bench.eps: $(top_srcdir)/bench/bench.plt
+	@$(SHRC); echo_cmd cd $(@D) && echo_cmd gnuplot $${top_srcdir}/bench/bench.plt > $(@F).tmp
+	mv $@.tmp $@
+
+$(top_builddir)/bench/bench.pdf: $(top_builddir)/bench/bench.eps
+	@$(SHRC); echo_cmd ps2pdf $${top_builddir}/bench/bench.eps $@
+
+bench-mods: $(BENCH_MODS:%=$(top_builddir)/bench/%)
+
+bench-all: $(top_builddir)/bench/bench.pdf
+
+bench-clean:
+	$(RM) -r $(top_builddir)/bench/*.so $(top_builddir)/bench/*.dSYM
+	$(RM) $(top_builddir)/bench/*.dat $(top_builddir)/bench/*.tmp
+	$(RM) $(top_builddir)/bench/bench.{eps,pdf}

src/ext/timeouts/bench/bench-add.lua

@@ -0,0 +1,30 @@
+#!/usr/bin/env lua
+
+local bench = require"bench"
+local aux = require"bench-aux"
+
+local lib = ... or aux.optenv("BENCH_L", "bench-wheel.so")
+local limit = tonumber(aux.optenv("BENCH_N", 1000000))
+local step  = tonumber(aux.optenv("BENCH_S", limit / 100))
+local exp_step = tonumber(aux.optenv("BENCH_E", 1.0))
+local verbose = aux.toboolean(os.getenv("BENCH_V", false))
+
+local B = bench.new(lib, count, nil, verbose)
+local fill_count, fill_last = B:fill(limit)
+
+for i=0,limit,step do
+	local exp_elapsed, fill_elapsed, fill_rate
+
+	-- expire all timeouts
+	--exp_elapsed = aux.time(B.expire, B, fill_count, fill_last * exp_step)
+	exp_elapsed = aux.time(B.del, B, 0, fill_count)
+	assert(B:empty())
+
+	-- add i timeouts
+	fill_elapsed, fill_count, fill_last = aux.time(B.fill, B, i)
+	assert(fill_count == i)
+	fill_rate = fill_elapsed > 0 and (fill_count / fill_elapsed) or 0
+
+	local fmt = verbose and "%d\t%f\t(%d/s)\t(exp:%f)" or "%d\t%f"
+	aux.say(fmt, i, fill_elapsed, fill_rate, exp_elapsed)
+end

src/ext/timeouts/bench/bench-aux.lua

@@ -0,0 +1,30 @@
+local bench = require"bench"
+local clock = bench.clock
+
+local aux = {}
+
+local function time_return(begun, ...)
+	local duration = clock() - begun
+	return duration, ...
+end
+
+function aux.time(f, ...)
+	local begun = clock()
+	return time_return(begun, f(...))
+end
+
+function aux.say(...)
+	print(string.format(...))
+end
+
+function aux.toboolean(s)
+	return tostring(s):match("^[1TtYy]") and true or false
+end
+
+function aux.optenv(k, def)
+	local s = os.getenv(k)
+
+	return (s and #s > 0 and s) or def
+end
+
+return aux

src/ext/timeouts/bench/bench-del.lua

@@ -0,0 +1,25 @@
+#!/usr/bin/env lua
+
+local bench = require"bench"
+local aux = require"bench-aux"
+
+local lib = ... or aux.optenv("BENCH_L", "bench-wheel.so")
+local limit = tonumber(aux.optenv("BENCH_N", 1000000))
+local step  = tonumber(aux.optenv("BENCH_S", limit / 100))
+local verbose = aux.toboolean(os.getenv("BENCH_V", false))
+
+local B = bench.new(lib, count)
+
+for i=0,limit,step do
+	-- add i timeouts
+	local fill_elapsed, fill_count = aux.time(B.fill, B, i, 60 * 1000000)
+	assert(i == fill_count)
+
+	--- delete i timeouts
+	local del_elapsed = aux.time(B.del, B, 0, fill_count)
+	assert(B:empty())
+	local del_rate = i > 0 and i / del_elapsed or 0
+
+	local fmt = verbose and "%d\t%f\t(%d/s)\t(fill:%f)" or "%d\t%f"
+	aux.say(fmt, i, del_elapsed, del_rate, fill_elapsed)
+end

src/ext/timeouts/bench/bench-expire.lua

@@ -0,0 +1,29 @@
+#!/usr/bin/env lua
+
+local bench = require"bench"
+local aux = require"bench-aux"
+
+local lib = ... or aux.optenv("BENCH_L", "bench-wheel.so")
+local limit = tonumber(aux.optenv("BENCH_N", 1000000))
+local step  = tonumber(aux.optenv("BENCH_S", limit / 100))
+-- expire 1/1000 * #timeouts per clock update
+local exp_step = tonumber(aux.optenv("BENCH_E", 0.0001)) 
+local verbose = aux.toboolean(os.getenv("BENCH_V", false))
+
+local B = require"bench".new(lib, count)
+
+for i=0,limit,step do
+	-- add i timeouts
+	local fill_elapsed, fill_count, fill_last  = aux.time(B.fill, B, i)
+
+	-- expire timeouts by iteratively updating clock. exp_step is the
+	-- approximate number of timeouts (as a fraction of the total number
+	-- of timeouts) that will expire per update.
+	local exp_elapsed, exp_count = aux.time(B.expire, B, fill_count, math.floor(fill_last * exp_step))
+	assert(exp_count == i)
+	assert(B:empty())
+	local exp_rate = i > 0 and i / exp_elapsed or 0
+
+	local fmt = verbose and "%d\t%f\t(%d/s)\t(fill:%f)" or "%d\t%f"
+	aux.say(fmt, i, exp_elapsed, exp_rate, fill_elapsed)
+end

src/ext/timeouts/bench/bench-heap.c

@@ -0,0 +1,236 @@
+/*
+ * Copyright (c) 2006 Maxim Yegorushkin <maxim.yegorushkin@gmail.com>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. The name of the author may not be used to endorse or promote products
+ *    derived from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#ifndef _MIN_HEAP_H_
+#define _MIN_HEAP_H_
+
+#include <stdlib.h>
+#include <err.h>
+#include "timeout.h"
+#include "bench.h"
+
+#define min_heap_idx interval
+
+typedef timeout_t min_heap_idx_t;
+
+typedef struct min_heap
+{
+    struct timeout** p;
+    unsigned n, a;
+    timeout_t curtime;
+} min_heap_t;
+
+static inline void           min_heap_ctor(min_heap_t* s);
+static inline void           min_heap_dtor(min_heap_t* s);
+static inline void           min_heap_elem_init(struct timeout* e);
+static inline int            min_heap_elem_greater(struct timeout *a, struct timeout *b);
+static inline int            min_heap_empty(min_heap_t* s);
+static inline unsigned       min_heap_size(min_heap_t* s);
+static inline struct timeout*  min_heap_top(min_heap_t* s);
+static inline int            min_heap_reserve(min_heap_t* s, unsigned n);
+static inline int            min_heap_push(min_heap_t* s, struct timeout* e);
+static inline struct timeout*  min_heap_pop(min_heap_t* s);
+static inline int            min_heap_erase(min_heap_t* s, struct timeout* e);
+static inline void           min_heap_shift_up_(min_heap_t* s, unsigned hole_index, struct timeout* e);
+static inline void           min_heap_shift_down_(min_heap_t* s, unsigned hole_index, struct timeout* e);
+
+int min_heap_elem_greater(struct timeout *a, struct timeout *b)
+{
+    return a->expires > b->expires;
+}
+
+void min_heap_ctor(min_heap_t* s) { s->p = 0; s->n = 0; s->a = 0; }
+void min_heap_dtor(min_heap_t* s) { if(s->p) free(s->p); }
+void min_heap_elem_init(struct timeout* e) { e->min_heap_idx = -1; }
+int min_heap_empty(min_heap_t* s) { return 0u == s->n; }
+unsigned min_heap_size(min_heap_t* s) { return s->n; }
+struct timeout* min_heap_top(min_heap_t* s) { return s->n ? *s->p : 0; }
+
+int min_heap_push(min_heap_t* s, struct timeout* e)
+{
+    if(min_heap_reserve(s, s->n + 1))
+        return -1;
+    min_heap_shift_up_(s, s->n++, e);
+    return 0;
+}
+
+struct timeout* min_heap_pop(min_heap_t* s)
+{
+    if(s->n)
+    {
+        struct timeout* e = *s->p;
+        min_heap_shift_down_(s, 0u, s->p[--s->n]);
+        e->min_heap_idx = -1;
+        return e;
+    }
+    return 0;
+}
+
+int min_heap_erase(min_heap_t* s, struct timeout* e)
+{
+    if(((min_heap_idx_t)-1) != e->min_heap_idx)
+    {
+        struct timeout *last = s->p[--s->n];
+        unsigned parent = (e->min_heap_idx - 1) / 2;
+	/* we replace e with the last element in the heap.  We might need to
+	   shift it upward if it is less than its parent, or downward if it is
+	   greater than one or both its children. Since the children are known
+	   to be less than the parent, it can't need to shift both up and
+	   down. */
+        if (e->min_heap_idx > 0 && min_heap_elem_greater(s->p[parent], last))
+             min_heap_shift_up_(s, e->min_heap_idx, last);
+        else
+             min_heap_shift_down_(s, e->min_heap_idx, last);
+        e->min_heap_idx = -1;
+        return 0;
+    }
+    return -1;
+}
+
+int min_heap_reserve(min_heap_t* s, unsigned n)
+{
+    if(s->a < n)
+    {
+        struct timeout** p;
+        unsigned a = s->a ? s->a * 2 : 8;
+        if(a < n)
+            a = n;
+        if(!(p = (struct timeout**)realloc(s->p, a * sizeof *p)))
+            return -1;
+        s->p = p;
+        s->a = a;
+    }
+    return 0;
+}
+
+void min_heap_shift_up_(min_heap_t* s, unsigned hole_index, struct timeout* e)
+{
+    unsigned parent = (hole_index - 1) / 2;
+    while(hole_index && min_heap_elem_greater(s->p[parent], e))
+    {
+        (s->p[hole_index] = s->p[parent])->min_heap_idx = hole_index;
+        hole_index = parent;
+        parent = (hole_index - 1) / 2;
+    }
+    (s->p[hole_index] = e)->min_heap_idx = hole_index;
+}
+
+void min_heap_shift_down_(min_heap_t* s, unsigned hole_index, struct timeout* e)
+{
+    unsigned min_child = 2 * (hole_index + 1);
+    while(min_child <= s->n)
+	{
+        min_child -= min_child == s->n || min_heap_elem_greater(s->p[min_child], s->p[min_child - 1]);
+        if(!(min_heap_elem_greater(e, s->p[min_child])))
+            break;
+        (s->p[hole_index] = s->p[min_child])->min_heap_idx = hole_index;
+        hole_index = min_child;
+        min_child = 2 * (hole_index + 1);
+	}
+    min_heap_shift_up_(s, hole_index,  e);
+}
+
+#endif /* _MIN_HEAP_H_ */
+
+
+static void *init(struct timeout *timeout, size_t count, int verbose) {
+    min_heap_t *H;
+    size_t i;
+
+    H = calloc(1, sizeof *H);
+
+    min_heap_ctor(H);
+    if (0 != min_heap_reserve(H, count))
+        err(1, "realloc");
+
+    for (i = 0; i < count; i++) {
+        min_heap_elem_init(&timeout[i]);
+    }
+
+    return H;
+} /* init() */
+
+
+static void add(void *ctx, struct timeout *to, timeout_t expires) {
+    min_heap_t *H = ctx;
+    min_heap_erase(H, to);
+    to->expires = H->curtime + expires;
+    if (0 != min_heap_push(H, to))
+        err(1, "realloc");
+} /* add() */
+
+
+static void del(void *ctx, struct timeout *to) {
+    min_heap_erase(ctx, to);
+} /* del() */
+
+
+static struct timeout *get(void *ctx) {
+    min_heap_t *H = ctx;
+    struct timeout *to;
+
+    if ((to = min_heap_top(H)) && to->expires <= H->curtime)
+        return min_heap_pop(H);
+
+    return NULL;
+} /* get() */
+
+
+static void update(void *ctx, timeout_t ts) {
+    min_heap_t *H = ctx;
+    H->curtime = ts;
+} /* update() */
+
+
+static void check(void *ctx) {
+    return;
+} /* check() */
+
+
+static int empty(void *ctx) {
+    min_heap_t *H = ctx;
+
+    return (NULL == min_heap_top(H));
+} /* empty() */
+
+
+static void destroy(void *H) {
+    free(H);
+    return;
+} /* destroy() */
+
+
+const struct benchops benchops = {
+    .init    = &init,
+    .add     = &add,
+    .del     = &del,
+    .get     = &get,
+    .update  = &update,
+    .check   = &check,
+    .empty   = &empty,
+    .destroy = &destroy,
+};
+

src/ext/timeouts/bench/bench-llrb.c

@@ -0,0 +1,425 @@
+/* ==========================================================================
+ * llrb.h - Iterative Left-leaning Red-Black Tree.
+ * --------------------------------------------------------------------------
+ * Copyright (c) 2011, 2013  William Ahern <william@25thandClement.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to permit
+ * persons to whom the Software is furnished to do so, subject to the
+ * following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN
+ * NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * --------------------------------------------------------------------------
+ * CREDITS:
+ * 	o Algorithm courtesy of Robert Sedgewick, "Left-leaning Red-Black
+ * 	  Trees" (September 2008); and Robert Sedgewick and Kevin Wayne,
+ * 	  Algorithms (4th ed. 2011).
+ *
+ * 	  Sedgewick touts the simplicity of the recursive implementation,
+ * 	  but at least for the 2-3 tree variant the iterative approach is
+ * 	  almost line-for-line identical. The magic of C pointers helps;
+ * 	  it'd be uglier with Java.
+ *
+ * 	  A couple of missing NULL checks were added to Sedgewick's deletion
+ * 	  example, and insert was optimized to short-circuit rotations when
+ * 	  walking up the tree.
+ * 
+ * 	o Code implemented in the fashion of Niels Provos' excellent *BSD
+ * 	  sys/tree.h pre-processor library.
+ *
+ * 	  Regarding relative performance, I've refrained from sharing my own
+ * 	  benchmarks. Differences in run-time speed were too correlated to
+ * 	  compiler options and other external factors.
+ *
+ * 	  Provos' delete implementation doesn't need to start at the root of
+ * 	  the tree. However, RB_REMOVE must be passed the actual node to be
+ * 	  removed. LLRB_REMOVE merely requires a key, much like
+ * 	  RB_FIND/LLRB_FIND.
+ * ==========================================================================
+ */
+#ifndef LLRB_H
+#define LLRB_H
+
+#define LLRB_VENDOR "william@25thandClement.com"
+#define LLRB_VERSION 0x20130925
+
+#ifndef LLRB_STATIC
+#ifdef __GNUC__
+#define LLRB_STATIC __attribute__((__unused__)) static
+#else
+#define LLRB_STATIC static
+#endif
+#endif
+
+#define LLRB_HEAD(name, type) \
+struct name { struct type *rbh_root; }
+
+#define LLRB_INITIALIZER(root) { 0 }
+
+#define LLRB_INIT(root) do { (root)->rbh_root = 0; } while (0)
+
+#define LLRB_BLACK 0
+#define LLRB_RED   1
+
+#define LLRB_ENTRY(type) \
+struct { struct type *rbe_left, *rbe_right, *rbe_parent; _Bool rbe_color; }
+
+#define LLRB_LEFT(elm, field) (elm)->field.rbe_left
+#define LLRB_RIGHT(elm, field) (elm)->field.rbe_right
+#define LLRB_PARENT(elm, field) (elm)->field.rbe_parent
+#define LLRB_EDGE(head, elm, field) (((elm) == LLRB_ROOT(head))? &LLRB_ROOT(head) : ((elm) == LLRB_LEFT(LLRB_PARENT((elm), field), field))? &LLRB_LEFT(LLRB_PARENT((elm), field), field) : &LLRB_RIGHT(LLRB_PARENT((elm), field), field))
+#define LLRB_COLOR(elm, field) (elm)->field.rbe_color
+#define LLRB_ROOT(head) (head)->rbh_root
+#define LLRB_EMPTY(head) ((head)->rbh_root == 0)
+#define LLRB_ISRED(elm, field) ((elm) && LLRB_COLOR((elm), field) == LLRB_RED)
+
+#define LLRB_PROTOTYPE(name, type, field, cmp) \
+	LLRB_PROTOTYPE_INTERNAL(name, type, field, cmp,)
+#define LLRB_PROTOTYPE_STATIC(name, type, field, cmp) \
+	LLRB_PROTOTYPE_INTERNAL(name, type, field, cmp, LLRB_STATIC)
+#define LLRB_PROTOTYPE_INTERNAL(name, type, field, cmp, attr) \
+attr struct type *name##_LLRB_INSERT(struct name *, struct type *); \
+attr struct type *name##_LLRB_DELETE(struct name *, struct type *); \
+attr struct type *name##_LLRB_FIND(struct name *, struct type *); \
+attr struct type *name##_LLRB_MIN(struct type *); \
+attr struct type *name##_LLRB_MAX(struct type *); \
+attr struct type *name##_LLRB_NEXT(struct type *);
+
+#define LLRB_GENERATE(name, type, field, cmp) \
+	LLRB_GENERATE_INTERNAL(name, type, field, cmp,)
+#define LLRB_GENERATE_STATIC(name, type, field, cmp) \
+	LLRB_GENERATE_INTERNAL(name, type, field, cmp, LLRB_STATIC)
+#define LLRB_GENERATE_INTERNAL(name, type, field, cmp, attr) \
+static inline void name##_LLRB_ROTL(struct type **pivot) { \
+	struct type *a = *pivot; \
+	struct type *b = LLRB_RIGHT(a, field); \
+	if ((LLRB_RIGHT(a, field) = LLRB_LEFT(b, field))) \
+		LLRB_PARENT(LLRB_RIGHT(a, field), field) = a; \
+	LLRB_LEFT(b, field) = a; \
+	LLRB_COLOR(b, field) = LLRB_COLOR(a, field); \
+	LLRB_COLOR(a, field) = LLRB_RED; \
+	LLRB_PARENT(b, field) = LLRB_PARENT(a, field); \
+	LLRB_PARENT(a, field) = b; \
+	*pivot = b; \
+} \
+static inline void name##_LLRB_ROTR(struct type **pivot) { \
+	struct type *b = *pivot; \
+	struct type *a = LLRB_LEFT(b, field); \
+	if ((LLRB_LEFT(b, field) = LLRB_RIGHT(a, field))) \
+		LLRB_PARENT(LLRB_LEFT(b, field), field) = b; \
+	LLRB_RIGHT(a, field) = b; \
+	LLRB_COLOR(a, field) = LLRB_COLOR(b, field); \
+	LLRB_COLOR(b, field) = LLRB_RED; \
+	LLRB_PARENT(a, field) = LLRB_PARENT(b, field); \
+	LLRB_PARENT(b, field) = a; \
+	*pivot = a; \
+} \
+static inline void name##_LLRB_FLIP(struct type *root) { \
+	LLRB_COLOR(root, field) = !LLRB_COLOR(root, field); \
+	LLRB_COLOR(LLRB_LEFT(root, field), field) = !LLRB_COLOR(LLRB_LEFT(root, field), field); \
+	LLRB_COLOR(LLRB_RIGHT(root, field), field) = !LLRB_COLOR(LLRB_RIGHT(root, field), field); \
+} \
+static inline void name##_LLRB_FIXUP(struct type **root) { \
+	if (LLRB_ISRED(LLRB_RIGHT(*root, field), field) && !LLRB_ISRED(LLRB_LEFT(*root, field), field)) \
+		name##_LLRB_ROTL(root); \
+	if (LLRB_ISRED(LLRB_LEFT(*root, field), field) && LLRB_ISRED(LLRB_LEFT(LLRB_LEFT(*root, field), field), field)) \
+		name##_LLRB_ROTR(root); \
+	if (LLRB_ISRED(LLRB_LEFT(*root, field), field) && LLRB_ISRED(LLRB_RIGHT(*root, field), field)) \
+		name##_LLRB_FLIP(*root); \
+} \
+attr struct type *name##_LLRB_INSERT(struct name *head, struct type *elm) { \
+	struct type **root = &LLRB_ROOT(head); \
+	struct type *parent = 0; \
+	while (*root) { \
+		int comp = (cmp)((elm), (*root)); \
+		parent = *root; \
+		if (comp < 0) \
+			root = &LLRB_LEFT(*root, field); \
+		else if (comp > 0) \
+			root = &LLRB_RIGHT(*root, field); \
+		else \
+			return *root; \
+	} \
+	LLRB_LEFT((elm), field) = 0; \
+	LLRB_RIGHT((elm), field) = 0; \
+	LLRB_COLOR((elm), field) = LLRB_RED; \
+	LLRB_PARENT((elm), field) = parent; \
+	*root = (elm); \
+	while (parent && (LLRB_ISRED(LLRB_LEFT(parent, field), field) || LLRB_ISRED(LLRB_RIGHT(parent, field), field))) { \
+		root = LLRB_EDGE(head, parent, field); \
+		parent = LLRB_PARENT(parent, field); \
+		name##_LLRB_FIXUP(root); \
+	} \
+	LLRB_COLOR(LLRB_ROOT(head), field) = LLRB_BLACK; \
+	return 0; \
+} \
+static inline void name##_LLRB_MOVL(struct type **pivot) { \
+	name##_LLRB_FLIP(*pivot); \
+	if (LLRB_ISRED(LLRB_LEFT(LLRB_RIGHT(*pivot, field), field), field)) { \
+		name##_LLRB_ROTR(&LLRB_RIGHT(*pivot, field)); \
+		name##_LLRB_ROTL(pivot); \
+		name##_LLRB_FLIP(*pivot); \
+	} \
+} \
+static inline void name##_LLRB_MOVR(struct type **pivot) { \
+	name##_LLRB_FLIP(*pivot); \
+	if (LLRB_ISRED(LLRB_LEFT(LLRB_LEFT(*pivot, field), field), field)) { \
+		name##_LLRB_ROTR(pivot); \
+		name##_LLRB_FLIP(*pivot); \
+	} \
+} \
+static inline struct type *name##_DELETEMIN(struct name *head, struct type **root) { \
+	struct type **pivot = root, *deleted, *parent; \
+	while (LLRB_LEFT(*pivot, field)) { \
+		if (!LLRB_ISRED(LLRB_LEFT(*pivot, field), field) && !LLRB_ISRED(LLRB_LEFT(LLRB_LEFT(*pivot, field), field), field)) \
+			name##_LLRB_MOVL(pivot); \
+		pivot = &LLRB_LEFT(*pivot, field); \
+	} \
+	deleted = *pivot; \
+	parent = LLRB_PARENT(*pivot, field); \
+	*pivot = 0; \
+	while (root != pivot) { \
+		pivot = LLRB_EDGE(head, parent, field); \
+		parent = LLRB_PARENT(parent, field); \
+		name##_LLRB_FIXUP(pivot); \
+	} \
+	return deleted; \
+} \
+attr struct type *name##_LLRB_DELETE(struct name *head, struct type *elm) { \
+	struct type **root = &LLRB_ROOT(head), *parent = 0, *deleted = 0; \
+	int comp; \
+	while (*root) { \
+		parent = LLRB_PARENT(*root, field); \
+		comp = (cmp)(elm, *root); \
+		if (comp < 0) { \
+			if (LLRB_LEFT(*root, field) && !LLRB_ISRED(LLRB_LEFT(*root, field), field) && !LLRB_ISRED(LLRB_LEFT(LLRB_LEFT(*root, field), field), field)) \
+				name##_LLRB_MOVL(root); \
+			root = &LLRB_LEFT(*root, field); \
+		} else { \
+			if (LLRB_ISRED(LLRB_LEFT(*root, field), field)) { \
+				name##_LLRB_ROTR(root); \
+				comp = (cmp)(elm, *root); \
+			} \
+			if (!comp && !LLRB_RIGHT(*root, field)) { \
+				deleted = *root; \
+				*root = 0; \
+				break; \
+			} \
+			if (LLRB_RIGHT(*root, field) && !LLRB_ISRED(LLRB_RIGHT(*root, field), field) && !LLRB_ISRED(LLRB_LEFT(LLRB_RIGHT(*root, field), field), field)) { \
+				name##_LLRB_MOVR(root); \
+				comp = (cmp)(elm, *root); \
+			} \
+			if (!comp) { \
+				struct type *orphan = name##_DELETEMIN(head, &LLRB_RIGHT(*root, field)); \
+				LLRB_COLOR(orphan, field) = LLRB_COLOR(*root, field); \
+				LLRB_PARENT(orphan, field) = LLRB_PARENT(*root, field); \
+				if ((LLRB_RIGHT(orphan, field) = LLRB_RIGHT(*root, field))) \
+					LLRB_PARENT(LLRB_RIGHT(orphan, field), field) = orphan; \
+				if ((LLRB_LEFT(orphan, field) = LLRB_LEFT(*root, field))) \
+					LLRB_PARENT(LLRB_LEFT(orphan, field), field) = orphan; \
+				deleted = *root; \
+				*root = orphan; \
+				parent = *root; \
+				break; \
+			} else \
+				root = &LLRB_RIGHT(*root, field); \
+		} \
+	} \
+	while (parent) { \
+		root = LLRB_EDGE(head, parent, field); \
+		parent = LLRB_PARENT(parent, field); \
+		name##_LLRB_FIXUP(root); \
+	} \
+	if (LLRB_ROOT(head)) \
+		LLRB_COLOR(LLRB_ROOT(head), field) = LLRB_BLACK; \
+	return deleted; \
+} \
+attr struct type *name##_LLRB_FIND(struct name *head, struct type *key) { \
+	struct type *elm = LLRB_ROOT(head); \
+	while (elm) { \
+		int comp = (cmp)(key, elm); \
+		if (comp < 0) \
+			elm = LLRB_LEFT(elm, field); \
+		else if (comp > 0) \
+			elm = LLRB_RIGHT(elm, field); \
+		else \
+			return elm; \
+	} \
+	return 0; \
+} \
+attr struct type *name##_LLRB_MIN(struct type *elm) { \
+	while (elm && LLRB_LEFT(elm, field)) \
+		elm = LLRB_LEFT(elm, field); \
+	return elm; \
+} \
+attr struct type *name##_LLRB_MAX(struct type *elm) { \
+	while (elm && LLRB_RIGHT(elm, field)) \
+		elm = LLRB_RIGHT(elm, field); \
+	return elm; \
+} \
+attr struct type *name##_LLRB_NEXT(struct type *elm) { \
+	if (LLRB_RIGHT(elm, field)) { \
+		return name##_LLRB_MIN(LLRB_RIGHT(elm, field)); \
+	} else if (LLRB_PARENT(elm, field)) { \
+		if (elm == LLRB_LEFT(LLRB_PARENT(elm, field), field)) \
+			return LLRB_PARENT(elm, field); \
+		while (LLRB_PARENT(elm, field) && elm == LLRB_RIGHT(LLRB_PARENT(elm, field), field)) \
+			elm = LLRB_PARENT(elm, field); \
+		return LLRB_PARENT(elm, field); \
+	} else return 0; \
+}
+
+#define LLRB_INSERT(name, head, elm) name##_LLRB_INSERT((head), (elm))
+#define LLRB_DELETE(name, head, elm) name##_LLRB_DELETE((head), (elm))
+#define LLRB_REMOVE(name, head, elm) name##_LLRB_DELETE((head), (elm))
+#define LLRB_FIND(name, head, elm) name##_LLRB_FIND((head), (elm))
+#define LLRB_MIN(name, head) name##_LLRB_MIN(LLRB_ROOT((head)))
+#define LLRB_MAX(name, head) name##_LLRB_MAX(LLRB_ROOT((head)))
+#define LLRB_NEXT(name, head, elm) name##_LLRB_NEXT((elm))
+
+#define LLRB_FOREACH(elm, name, head) \
+for ((elm) = LLRB_MIN(name, head); (elm); (elm) = name##_LLRB_NEXT((elm)))
+
+#endif /* LLRB_H */
+
+
+#include <stdlib.h>
+
+#include "timeout.h"
+#include "bench.h"
+
+
+struct rbtimeout {
+	timeout_t expires;
+
+	int pending;
+
+	LLRB_ENTRY(rbtimeout) rbe;
+};
+
+struct rbtimeouts {
+	timeout_t curtime;
+	LLRB_HEAD(tree, rbtimeout) tree;
+};
+
+
+static int timeoutcmp(struct rbtimeout *a, struct rbtimeout *b) {
+	if (a->expires < b->expires) {
+		return -1;
+	} else if (a->expires > b->expires) {
+		return 1;
+	} else if (a < b) {
+		return -1;
+	} else if (a > b) {
+		return 1;
+	} else {
+		return 0;
+	}
+} /* timeoutcmp() */
+
+LLRB_GENERATE_STATIC(tree, rbtimeout, rbe, timeoutcmp)
+
+static void *init(struct timeout *timeout, size_t count, int verbose) {
+	struct rbtimeouts *T;
+	size_t i;
+
+	T = malloc(sizeof *T);
+	T->curtime = 0;
+	LLRB_INIT(&T->tree);
+
+	for (i = 0; i < count; i++) {
+		struct rbtimeout *to = (void *)&timeout[i];
+		to->expires = 0;
+		to->pending = 0;
+	}
+
+	return T;
+} /* init() */
+
+
+static void add(void *ctx, struct timeout *_to, timeout_t expires) {
+	struct rbtimeouts *T = ctx;
+	struct rbtimeout *to = (void *)_to;
+
+	if (to->pending)
+		LLRB_REMOVE(tree, &T->tree, to);
+
+	to->expires = T->curtime + expires;
+	LLRB_INSERT(tree, &T->tree, to);
+	to->pending = 1;
+} /* add() */
+
+
+static void del(void *ctx, struct timeout *_to) {
+	struct rbtimeouts *T = ctx;
+	struct rbtimeout *to = (void *)_to;
+
+	LLRB_REMOVE(tree, &T->tree, to);
+	to->pending = 0;
+	to->expires = 0;
+} /* del() */
+
+
+static struct timeout *get(void *ctx) {
+	struct rbtimeouts *T = ctx;
+	struct rbtimeout *to;
+
+	if ((to = LLRB_MIN(tree, &T->tree)) && to->expires <= T->curtime) {
+		LLRB_REMOVE(tree, &T->tree, to);
+		to->pending = 0;
+		to->expires = 0;
+
+		return (void *)to;
+	}
+
+	return NULL;
+} /* get() */
+
+
+static void update(void *ctx, timeout_t ts) {
+	struct rbtimeouts *T = ctx;
+	T->curtime = ts;
+} /* update() */
+
+
+static void check(void *ctx) {
+	return;
+} /* check() */
+
+
+static int empty(void *ctx) {
+	struct rbtimeouts *T = ctx;
+
+	return LLRB_EMPTY(&T->tree);
+} /* empty() */
+
+
+static void destroy(void *ctx) {
+	free(ctx);
+	return;
+} /* destroy() */
+
+
+const struct benchops benchops = {
+	.init    = &init,
+	.add     = &add,
+	.del     = &del,
+	.get     = &get,
+	.update  = &update,
+	.check   = &check,
+	.empty   = &empty,
+	.destroy = &destroy,
+};
+

src/ext/timeouts/bench/bench-wheel.c

@@ -0,0 +1,81 @@
+#include <stdlib.h>
+
+#define TIMEOUT_PUBLIC static
+
+#include "timeout.h"
+#include "timeout.c"
+#include "bench.h"
+
+
+static void *init(struct timeout *timeout, size_t count, int verbose) {
+	struct timeouts *T;
+	size_t i;
+	int error;
+
+	T = timeouts_open(TIMEOUT_mHZ, &error);
+
+	for (i = 0; i < count; i++) {
+		timeout_init(&timeout[i], 0);
+	}
+
+#if TIMEOUT_DEBUG - 0
+	timeout_debug = verbose;
+#endif
+
+	return T;
+} /* init() */
+
+
+static void add(void *T, struct timeout *to, timeout_t expires) {
+	timeouts_add(T, to, expires);
+} /* add() */
+
+
+static void del(void *T, struct timeout *to) {
+	timeouts_del(T, to);
+} /* del() */
+
+
+static struct timeout *get(void *T) {
+	return timeouts_get(T);
+} /* get() */
+
+
+static void update(void *T, timeout_t ts) {
+	timeouts_update(T, ts);
+} /* update() */
+
+
+static void (check)(void *T) {
+	if (!timeouts_check(T, stderr))
+		_Exit(1);
+} /* check() */
+
+
+static int empty(void *T) {
+	return !(timeouts_pending(T) || timeouts_expired(T));
+} /* empty() */
+
+
+static struct timeout *next(void *T, struct timeouts_it *it) {
+	return timeouts_next(T, it);
+} /* next() */
+
+
+static void destroy(void *T) {
+	timeouts_close(T);
+} /* destroy() */
+
+
+const struct benchops benchops = {
+	.init    = &init,
+	.add     = &add,
+	.del     = &del,
+	.get     = &get,
+	.update  = &update,
+	.check   = &check,
+	.empty   = &empty,
+	.next    = &next,
+	.destroy = &destroy
+};
+

src/ext/timeouts/bench/bench.c

@@ -0,0 +1,293 @@
+#include <stdlib.h>
+#include <string.h>
+#include <time.h>
+#include <errno.h>
+#include <unistd.h>
+#include <dlfcn.h>
+
+#if __APPLE__
+#include <mach/mach_time.h>
+#endif
+
+#include <lua.h>
+#include <lualib.h>
+#include <lauxlib.h>
+
+#include "timeout.h"
+#include "bench.h"
+
+#if LUA_VERSION_NUM < 502
+static int lua_absindex(lua_State *L, int idx) {
+	return (idx > 0 || idx <= LUA_REGISTRYINDEX)? idx : lua_gettop(L) + idx + 1;
+} /* lua_absindex() */
+
+static void luaL_setfuncs(lua_State *L, const luaL_Reg *l, int nup) {
+	int i, t = lua_absindex(L, -1 - nup);
+
+	for (; l->name; l++) {
+		for (i = 0; i < nup; i++)
+			lua_pushvalue(L, -nup);
+		lua_pushcclosure(L, l->func, nup);
+		lua_setfield(L, t, l->name);
+	}
+
+	lua_pop(L, nup);
+} /* luaL_setfuncs() */
+
+#define luaL_newlibtable(L, l) \
+	lua_createtable(L, 0, (sizeof (l) / sizeof *(l)) - 1)
+
+#define luaL_newlib(L, l) \
+	(luaL_newlibtable((L), (l)), luaL_setfuncs((L), (l), 0))
+#endif
+
+#ifndef MAX
+#define MAX(a, b) (((a) > (b))? (a) : (b))
+#endif
+
+
+struct bench {
+	const char *path;
+	void *solib;
+	size_t count;
+	timeout_t timeout_max;
+	int verbose;
+
+	void *state;
+	struct timeout *timeout;
+	struct benchops ops;
+	timeout_t curtime;
+}; /* struct bench */
+
+
+#if __APPLE__
+static mach_timebase_info_data_t timebase;
+#endif
+
+
+static int long long monotime(void) {
+#if __APPLE__
+	unsigned long long abt;
+
+	abt = mach_absolute_time();
+	abt = abt * timebase.numer / timebase.denom;
+
+	return abt / 1000LL;
+#else
+	struct timespec ts;
+
+	clock_gettime(CLOCK_MONOTONIC, &ts);
+
+	return (ts.tv_sec * 1000000L) + (ts.tv_nsec / 1000L);
+#endif
+} /* monotime() */
+
+
+static int bench_clock(lua_State *L) {
+	lua_pushnumber(L, (double)monotime() / 1000000L);
+
+	return 1;
+} /* bench_clock() */
+
+
+static int bench_new(lua_State *L) {
+	const char *path = luaL_checkstring(L, 1);
+	size_t count = luaL_optinteger(L, 2, 1000000);
+	timeout_t timeout_max = luaL_optinteger(L, 3, 300 * 1000000L);
+	int verbose = (lua_isnone(L, 4))? 0 : lua_toboolean(L, 4);
+	struct bench *B;
+	struct benchops *ops;
+
+	B = lua_newuserdata(L, sizeof *B);
+	memset(B, 0, sizeof *B);
+
+	luaL_getmetatable(L, "BENCH*");
+	lua_setmetatable(L, -2);
+
+	B->count = count;
+	B->timeout_max = timeout_max;
+	B->verbose = verbose;
+
+	if (!(B->timeout = calloc(count, sizeof *B->timeout)))
+		return luaL_error(L, "%s", strerror(errno));
+
+	if (!(B->solib = dlopen(path, RTLD_NOW|RTLD_LOCAL)))
+		return luaL_error(L, "%s: %s", path, dlerror());
+
+	if (!(ops = dlsym(B->solib, "benchops")))
+		return luaL_error(L, "%s: %s", path, dlerror());
+
+	B->ops = *ops;
+	B->state = B->ops.init(B->timeout, B->count, B->verbose);
+
+	return 1;
+} /* bench_new() */
+
+
+static int bench_add(lua_State *L) {
+	struct bench *B = lua_touserdata(L, 1);
+	unsigned i;
+	timeout_t t;
+
+	i = (lua_isnoneornil(L, 2))? random() % B->count : (unsigned)luaL_checkinteger(L, 2);
+	t = (lua_isnoneornil(L, 3))? random() % B->timeout_max : (unsigned)luaL_checkinteger(L, 3);
+
+	B->ops.add(B->state, &B->timeout[i], t);
+
+	return 0;
+} /* bench_add() */
+
+
+static int bench_del(lua_State *L) {
+	struct bench *B = lua_touserdata(L, 1);
+	size_t i = luaL_optinteger(L, 2, random() % B->count);
+	size_t j = luaL_optinteger(L, 3, i);
+
+	while (i <= j && i < B->count) {
+		B->ops.del(B->state, &B->timeout[i]);
+		++i;
+	}
+
+	return 0;
+} /* bench_del() */
+
+
+static int bench_fill(lua_State *L) {
+	struct bench *B = lua_touserdata(L, 1);
+	size_t count = luaL_optinteger(L, 2, B->count);
+	long timeout_inc = luaL_optinteger(L, 3, -1), timeout_max = 0, timeout;
+	size_t i;
+
+	if (timeout_inc < 0) {
+		for (i = 0; i < count; i++) {
+			timeout = random() % B->timeout_max;
+			B->ops.add(B->state, &B->timeout[i], timeout);
+			timeout_max = MAX(timeout, timeout_max);
+		}
+	} else {
+		for (i = 0; i < count; i++) {
+			timeout = timeout_inc + i;
+			B->ops.add(B->state, &B->timeout[i], timeout_inc + i);
+			timeout_max = MAX(timeout, timeout_max);
+		}
+	}
+
+	lua_pushinteger(L, (lua_Integer)count);
+	lua_pushinteger(L, (lua_Integer)timeout_max);
+
+	return 2;
+} /* bench_fill() */
+
+
+static int bench_expire(lua_State *L) {
+	struct bench *B = lua_touserdata(L, 1);
+	unsigned count = luaL_optinteger(L, 2, B->count);
+	unsigned step = luaL_optinteger(L, 3, 300000);
+	size_t i = 0;
+
+	while (i < count && !B->ops.empty(B->state)) {
+		B->curtime += step;
+		B->ops.update(B->state, B->curtime);
+
+		while (B->ops.get(B->state))
+			i++;
+	}
+
+	lua_pushinteger(L, (lua_Integer)i);
+
+	return 1;
+} /* bench_expire() */
+
+
+static int bench_empty(lua_State *L) {
+	struct bench *B = lua_touserdata(L, 1);
+
+	lua_pushboolean(L, B->ops.empty(B->state));
+
+	return 1;
+} /* bench_empty() */
+
+
+static int bench__next(lua_State *L) {
+	struct bench *B = lua_touserdata(L, lua_upvalueindex(1));
+	struct timeouts_it *it = lua_touserdata(L, lua_upvalueindex(2));
+	struct timeout *to;
+
+	if (!B->ops.next || !(to = B->ops.next(B->state, it)))
+		return 0;
+
+	lua_pushinteger(L, luaL_optinteger(L, 2, 0) + 1);
+
+	lua_newtable(L);
+	lua_pushinteger(L, to->expires);
+	lua_setfield(L, -2, "expires");
+
+	return 2;
+} /* bench__next() */
+
+static int bench__pairs(lua_State *L) {
+	struct timeouts_it *it;
+
+	lua_settop(L, 1);
+
+	it = lua_newuserdata(L, sizeof *it);
+	TIMEOUTS_IT_INIT(it, TIMEOUTS_ALL);
+
+	lua_pushcclosure(L, &bench__next, 2);
+	lua_pushvalue(L, 1);
+	lua_pushinteger(L, 0);
+
+	return 3;
+} /* bench__pairs() */
+
+
+static int bench__gc(lua_State *L) {
+	struct bench *B = lua_touserdata(L, 1);
+
+	if (B->state) {
+		B->ops.destroy(B->state);
+		B->state = NULL;
+	}
+
+	return 0;
+} /* bench__gc() */
+
+
+static const luaL_Reg bench_methods[] = {
+	{ "add",    &bench_add },
+	{ "del",    &bench_del },
+	{ "fill",   &bench_fill },
+	{ "expire", &bench_expire },
+	{ "empty",  &bench_empty },
+	{ "close",  &bench__gc },
+	{ NULL,     NULL }
+};
+
+static const luaL_Reg bench_metatable[] = {
+	{ "__pairs", &bench__pairs },
+	{ "__gc",    &bench__gc },
+	{ NULL,      NULL }
+};
+
+static const luaL_Reg bench_globals[] = {
+	{ "new",   &bench_new },
+	{ "clock", &bench_clock },
+	{ NULL,  NULL }
+};
+
+int luaopen_bench(lua_State *L) {
+#if __APPLE__
+	mach_timebase_info(&timebase);
+#endif
+
+	if (luaL_newmetatable(L, "BENCH*")) {
+		luaL_setfuncs(L, bench_metatable, 0);
+		luaL_newlib(L, bench_methods);
+		lua_setfield(L, -2, "__index");
+	}
+
+	luaL_newlib(L, bench_globals);
+
+	return 1;
+} /* luaopen_bench() */
+

src/ext/timeouts/bench/bench.h

@@ -0,0 +1,11 @@
+struct benchops {
+	void *(*init)(struct timeout *, size_t, int);
+	void (*add)(void *, struct timeout *, timeout_t);
+	void (*del)(void *, struct timeout *);
+	struct timeout *(*get)(void *);
+	void (*update)(void *, timeout_t);
+	void (*check)(void *);
+	int (*empty)(void *);
+	struct timeout *(*next)(void *, struct timeouts_it *);
+	void (*destroy)(void *);
+}; /* struct benchops() */

src/ext/timeouts/bench/bench.plt

@@ -0,0 +1,19 @@
+set terminal postscript color
+
+set key top left
+set xlabel "Number of timeouts"
+set ylabel "Time\n(microseconds)"
+#set logscale x
+
+set title "Time spent installing timeouts" font ",20"
+plot 'heap-add.dat' using 1:($2*1000000) title "min-heap" with lines ls 1 lw 3 lc "red", \
+     'wheel-add.dat' using 1:($2*1000000) title "hierarchical wheel" with lines ls 1 lw 3 lc "forest-green"
+
+set title "Time spent deleting timeouts" font ",20"
+plot 'heap-del.dat' using 1:($2*1000000) title "min-heap" with lines ls 1 lw 3 lc "red", \
+     'wheel-del.dat' using 1:($2*1000000) title "hierarchical wheel" with lines ls 1 lw 3 lc "forest-green"
+
+set title "Time spent expiring timeouts\n(by iteratively updating clock ~1000 times)" font ",20"
+plot 'heap-expire.dat' using 1:($2*1000000) title "min-heap" with lines ls 1 lw 3 lc "red", \
+     'wheel-expire.dat' using 1:($2*1000000) title "hierarchical wheel" with lines ls 1 lw 3 lc "forest-green"
+

src/ext/timeouts/lua/Rules.mk

@@ -0,0 +1,20 @@
+$(LUA_APIS:%=$(top_builddir)/lua/%/timeout.so): $(top_srcdir)/lua/timeout-lua.c $(top_srcdir)/timeout.h $(top_srcdir)/timeout.c
+	mkdir -p $(@D)
+	@$(SHRC); echo_cmd $(CC) -o $@ $(top_srcdir)/lua/timeout-lua.c -I$(top_srcdir) -DWHEEL_BIT=$(WHEEL_BIT) -DWHEEL_NUM=$(WHEEL_NUM) $(LUA53_CPPFLAGS) $(ALL_CPPFLAGS) $(ALL_CFLAGS) $(ALL_SOFLAGS) $(ALL_LDFLAGS) $(ALL_LIBS)
+
+$(top_builddir)/lua/5.1/timeouts.so: $(top_builddir)/lua/5.1/timeout.so
+$(top_builddir)/lua/5.2/timeouts.so: $(top_builddir)/lua/5.2/timeout.so
+$(top_builddir)/lua/5.3/timeouts.so: $(top_builddir)/lua/5.3/timeout.so
+
+$(LUA_APIS:%=$(top_builddir)/lua/%/timeouts.so):
+	cd $(@D) && ln -fs timeout.so timeouts.so
+
+lua-5.1: $(top_builddir)/lua/5.1/timeout.so $(top_builddir)/lua/5.1/timeouts.so
+lua-5.2: $(top_builddir)/lua/5.2/timeout.so $(top_builddir)/lua/5.2/timeouts.so
+lua-5.3: $(top_builddir)/lua/5.3/timeout.so $(top_builddir)/lua/5.3/timeouts.so
+
+lua-clean:
+	$(RM) -r $(top_builddir)/lua/5.?
+
+clean: lua-clean
+

src/ext/timeouts/lua/timeout-lua.c

@@ -0,0 +1,396 @@
+#include <assert.h>
+#include <string.h>
+
+#include <lua.h>
+#include <lualib.h>
+#include <lauxlib.h>
+
+#if LUA_VERSION_NUM != 503
+#error only Lua 5.3 supported
+#endif
+
+#define TIMEOUT_PUBLIC static
+#include "timeout.h"
+#include "timeout.c"
+
+#define TIMEOUT_METANAME "struct timeout"
+#define TIMEOUTS_METANAME "struct timeouts*"
+
+static struct timeout *
+to_checkudata(lua_State *L, int index)
+{
+	return luaL_checkudata(L, index, TIMEOUT_METANAME);
+}
+
+static struct timeouts *
+tos_checkudata(lua_State *L, int index)
+{
+	return *(struct timeouts **)luaL_checkudata(L, index, TIMEOUTS_METANAME);
+}
+
+static void
+tos_bind(lua_State *L, int tos_index, int to_index)
+{
+	lua_getuservalue(L, tos_index);
+	lua_pushlightuserdata(L, to_checkudata(L, to_index));
+	lua_pushvalue(L, to_index);
+	lua_rawset(L, -3);
+	lua_pop(L, 1);
+}
+
+static void
+tos_unbind(lua_State *L, int tos_index, int to_index)
+{
+	lua_getuservalue(L, tos_index);
+	lua_pushlightuserdata(L, to_checkudata(L, to_index));
+	lua_pushnil(L);
+	lua_rawset(L, -3);
+	lua_pop(L, 1);
+}
+
+static int
+to__index(lua_State *L)
+{
+	struct timeout *to = to_checkudata(L, 1);
+
+	if (lua_type(L, 2 == LUA_TSTRING)) {
+		const char *key = lua_tostring(L, 2);
+
+		if (!strcmp(key, "flags")) {
+			lua_pushinteger(L, to->flags);
+
+			return 1;
+		} else if (!strcmp(key, "expires")) {
+			lua_pushinteger(L, to->expires);
+
+			return 1;
+		}
+	}
+
+	if (LUA_TNIL != lua_getuservalue(L, 1)) {
+		lua_pushvalue(L, 2);
+		if (LUA_TNIL != lua_rawget(L, -2))
+			return 1;
+	}
+
+	lua_pushvalue(L, 2);
+	if (LUA_TNIL != lua_rawget(L, lua_upvalueindex(1)))
+		return 1;
+
+	return 0;
+}
+
+static int
+to__newindex(lua_State *L)
+{
+	if (LUA_TNIL == lua_getuservalue(L, 1)) {
+		lua_newtable(L);
+		lua_pushvalue(L, -1);
+		lua_setuservalue(L, 1);
+	}
+
+	lua_pushvalue(L, 2);
+	lua_pushvalue(L, 3);
+	lua_rawset(L, -3);
+
+	return 0;
+}
+
+static int
+to__gc(lua_State *L)
+{
+	struct timeout *to = to_checkudata(L, 1);
+
+	/*
+	 * NB: On script exit it's possible for a timeout to still be
+	 * associated with a timeouts object, particularly when the timeouts
+	 * object was created first.
+	 */
+	timeout_del(to);
+
+	return 0;
+}
+
+static int
+to_new(lua_State *L)
+{
+	int flags = luaL_optinteger(L, 1, 0);
+	struct timeout *to;
+
+	to = lua_newuserdata(L, sizeof *to);
+	timeout_init(to, flags);
+	luaL_setmetatable(L, TIMEOUT_METANAME);
+
+	return 1;
+}
+
+static const luaL_Reg to_methods[] = {
+	{ NULL,  NULL },
+};
+
+static const luaL_Reg to_metatable[] = {
+	{ "__index",    &to__index },
+	{ "__newindex", &to__newindex },
+	{ "__gc",       &to__gc },
+	{ NULL,         NULL },
+};
+
+static const luaL_Reg to_globals[] = {
+	{ "new", &to_new },
+	{ NULL,  NULL },
+};
+
+static void
+to_newmetatable(lua_State *L)
+{
+	if (luaL_newmetatable(L, TIMEOUT_METANAME)) {
+		/*
+		 * fill metamethod table, capturing the methods table as an
+		 * upvalue for use by __index metamethod
+		 */
+		luaL_newlib(L, to_methods);
+		luaL_setfuncs(L, to_metatable, 1);
+	}
+}
+
+int
+luaopen_timeout(lua_State *L)
+{
+	to_newmetatable(L);
+
+	luaL_newlib(L, to_globals);
+	lua_pushinteger(L, TIMEOUT_INT);
+	lua_setfield(L, -2, "INT");
+	lua_pushinteger(L, TIMEOUT_ABS);
+	lua_setfield(L, -2, "ABS");
+
+	return 1;
+}
+
+static int
+tos_update(lua_State *L)
+{
+	struct timeouts *T = tos_checkudata(L, 1);
+	lua_Number n = luaL_checknumber(L, 2);
+
+	timeouts_update(T, timeouts_f2i(T, n));
+
+	lua_pushvalue(L, 1);
+
+	return 1;
+}
+
+static int
+tos_step(lua_State *L)
+{
+	struct timeouts *T = tos_checkudata(L, 1);
+	lua_Number n = luaL_checknumber(L, 2);
+
+	timeouts_step(T, timeouts_f2i(T, n));
+
+	lua_pushvalue(L, 1);
+
+	return 1;
+}
+
+static int
+tos_timeout(lua_State *L)
+{
+	struct timeouts *T = tos_checkudata(L, 1);
+
+	lua_pushnumber(L, timeouts_i2f(T, timeouts_timeout(T)));
+
+	return 1;
+}
+
+static int
+tos_add(lua_State *L)
+{
+	struct timeouts *T = tos_checkudata(L, 1);
+	struct timeout *to = to_checkudata(L, 2);
+	lua_Number timeout = luaL_checknumber(L, 3);
+
+	tos_bind(L, 1, 2);
+	timeouts_addf(T, to, timeout);
+
+	return lua_pushvalue(L, 1), 1;
+}
+
+static int
+tos_del(lua_State *L)
+{
+	struct timeouts *T = tos_checkudata(L, 1);
+	struct timeout *to = to_checkudata(L, 2);
+
+	timeouts_del(T, to);
+	tos_unbind(L, 1, 2);
+
+	return lua_pushvalue(L, 1), 1;
+}
+
+static int
+tos_get(lua_State *L)
+{
+	struct timeouts *T = tos_checkudata(L, 1);
+	struct timeout *to;
+
+	if (!(to = timeouts_get(T)))
+		return 0;
+
+	lua_getuservalue(L, 1);
+	lua_rawgetp(L, -1, to);
+
+	if (!timeout_pending(to))
+		tos_unbind(L, 1, lua_absindex(L, -1));
+
+	return 1;
+}
+
+static int
+tos_pending(lua_State *L)
+{
+	struct timeouts *T = tos_checkudata(L, 1);
+
+	lua_pushboolean(L, timeouts_pending(T));
+
+	return 1;
+}
+
+static int
+tos_expired(lua_State *L)
+{
+	struct timeouts *T = tos_checkudata(L, 1);
+
+	lua_pushboolean(L, timeouts_expired(T));
+
+	return 1;
+}
+
+static int
+tos_check(lua_State *L)
+{
+	struct timeouts *T = tos_checkudata(L, 1);
+
+	lua_pushboolean(L, timeouts_check(T, NULL));
+
+	return 1;
+}
+
+static int
+tos__next(lua_State *L)
+{
+	struct timeouts *T = tos_checkudata(L, lua_upvalueindex(1));
+	struct timeouts_it *it = lua_touserdata(L, lua_upvalueindex(2));
+	struct timeout *to;
+
+	if (!(to = timeouts_next(T, it)))
+		return 0;
+
+	lua_getuservalue(L, lua_upvalueindex(1));
+	lua_rawgetp(L, -1, to);
+
+	return 1;
+}
+
+static int
+tos_timeouts(lua_State *L)
+{
+	int flags = luaL_checkinteger(L, 2);
+	struct timeouts_it *it;
+
+	tos_checkudata(L, 1);
+	lua_pushvalue(L, 1);
+	it = lua_newuserdata(L, sizeof *it);
+	TIMEOUTS_IT_INIT(it, flags);
+	lua_pushcclosure(L, &tos__next, 2);
+
+	return 1;
+}
+
+static int
+tos__gc(lua_State *L)
+{
+	struct timeouts **tos = luaL_checkudata(L, 1, TIMEOUTS_METANAME);
+	struct timeout *to;
+
+	TIMEOUTS_FOREACH(to, *tos, TIMEOUTS_ALL) {
+		timeouts_del(*tos, to);
+	}
+
+	timeouts_close(*tos);
+	*tos = NULL;
+
+	return 0;
+}
+
+static int
+tos_new(lua_State *L)
+{
+	timeout_t hz = luaL_optinteger(L, 1, 0);
+	struct timeouts **T;
+	int error;
+	
+	T = lua_newuserdata(L, sizeof *T);
+	luaL_setmetatable(L, TIMEOUTS_METANAME);
+
+	lua_newtable(L);
+	lua_setuservalue(L, -2);
+
+	if (!(*T = timeouts_open(hz, &error)))
+		return luaL_error(L, "%s", strerror(error));
+
+	return 1;
+}
+
+static const luaL_Reg tos_methods[] = {
+	{ "update",   &tos_update },
+	{ "step",     &tos_step },
+	{ "timeout",  &tos_timeout },
+	{ "add",      &tos_add },
+	{ "del",      &tos_del },
+	{ "get",      &tos_get },
+	{ "pending",  &tos_pending },
+	{ "expired",  &tos_expired },
+	{ "check",    &tos_check },
+	{ "timeouts", &tos_timeouts },
+	{ NULL,       NULL },
+};
+
+static const luaL_Reg tos_metatable[] = {
+	{ "__gc", &tos__gc },
+	{ NULL,   NULL },
+};
+
+static const luaL_Reg tos_globals[] = {
+	{ "new", &tos_new },
+	{ NULL,  NULL },
+};
+
+static void
+tos_newmetatable(lua_State *L)
+{
+	if (luaL_newmetatable(L, TIMEOUTS_METANAME)) {
+		luaL_setfuncs(L, tos_metatable, 0);
+		luaL_newlib(L, tos_methods);
+		lua_setfield(L, -2, "__index");
+	}
+}
+
+int
+luaopen_timeouts(lua_State *L)
+{
+	to_newmetatable(L);
+	tos_newmetatable(L);
+
+	luaL_newlib(L, tos_globals);
+	lua_pushinteger(L, TIMEOUTS_PENDING);
+	lua_setfield(L, -2, "PENDING");
+	lua_pushinteger(L, TIMEOUTS_EXPIRED);
+	lua_setfield(L, -2, "EXPIRED");
+	lua_pushinteger(L, TIMEOUTS_ALL);
+	lua_setfield(L, -2, "ALL");
+	lua_pushinteger(L, TIMEOUTS_CLEAR);
+	lua_setfield(L, -2, "CLEAR");
+
+	return 1;
+}

src/ext/timeouts/test-timeout.c

@@ -0,0 +1,530 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <assert.h>
+#include <limits.h>
+
+#include "timeout.h"
+
+#define THE_END_OF_TIME ((timeout_t)-1)
+
+static int check_misc(void) {
+	if (TIMEOUT_VERSION != timeout_version())
+		return 1;
+	if (TIMEOUT_V_REL != timeout_v_rel())
+		return 1;
+	if (TIMEOUT_V_API != timeout_v_api())
+		return 1;
+	if (TIMEOUT_V_ABI != timeout_v_abi())
+		return 1;
+	if (strcmp(timeout_vendor(), TIMEOUT_VENDOR))
+		return 1;
+	return 0;
+}
+
+static int check_open_close(timeout_t hz_set, timeout_t hz_expect) {
+	int err=0;
+	struct timeouts *tos = timeouts_open(hz_set, &err);
+	if (!tos)
+		return 1;
+	if (err)
+		return 1;
+	if (hz_expect != timeouts_hz(tos))
+		return 1;
+	timeouts_close(tos);
+	return 0;
+}
+
+/* Not very random */
+static timeout_t random_to(timeout_t min, timeout_t max)
+{
+	if (max <= min)
+		return min;
+	/* Not actually all that random, but should exercise the code. */
+	timeout_t rand64 = random() * (timeout_t)INT_MAX + random();
+	return min + (rand64 % (max-min));
+}
+
+/* configuration for check_randomized */
+struct rand_cfg {
+	/* When creating timeouts, smallest possible delay */
+	timeout_t min_timeout;
+	/* When creating timeouts, largest possible delay */
+	timeout_t max_timeout;
+	/* First time to start the clock at. */
+	timeout_t start_at;
+	/* Do not advance the clock past this time. */
+	timeout_t end_at;
+	/* Number of timeouts to create and monitor. */
+	int n_timeouts;
+	/* Advance the clock by no more than this each step. */
+	timeout_t max_step;
+	/* Use relative timers and stepping */
+	int relative;
+	/* Every time the clock ticks, try removing this many timeouts at
+	 * random. */
+	int try_removing;
+	/* When we're done, advance the clock to the end of time. */
+	int finalize;
+};
+
+static int check_randomized(const struct rand_cfg *cfg)
+{
+#define FAIL() do {					\
+		printf("Failure on line %d\n", __LINE__);	\
+		goto done;					\
+	} while (0)
+
+	int i, err;
+	int rv = 1;
+	struct timeout *t = calloc(cfg->n_timeouts, sizeof(struct timeout));
+	timeout_t *timeouts = calloc(cfg->n_timeouts, sizeof(timeout_t));
+	uint8_t *fired = calloc(cfg->n_timeouts, sizeof(uint8_t));
+        uint8_t *found = calloc(cfg->n_timeouts, sizeof(uint8_t));
+	uint8_t *deleted = calloc(cfg->n_timeouts, sizeof(uint8_t));
+	struct timeouts *tos = timeouts_open(0, &err);
+	timeout_t now = cfg->start_at;
+	int n_added_pending = 0, cnt_added_pending = 0;
+	int n_added_expired = 0, cnt_added_expired = 0;
+        struct timeouts_it it_p, it_e, it_all;
+        int p_done = 0, e_done = 0, all_done = 0;
+	struct timeout *to = NULL;
+	const int rel = cfg->relative;
+
+	if (!t || !timeouts || !tos || !fired || !found || !deleted)
+		FAIL();
+	timeouts_update(tos, cfg->start_at);
+
+	for (i = 0; i < cfg->n_timeouts; ++i) {
+		if (&t[i] != timeout_init(&t[i], rel ? 0 : TIMEOUT_ABS))
+			FAIL();
+		if (timeout_pending(&t[i]))
+			FAIL();
+		if (timeout_expired(&t[i]))
+			FAIL();
+
+		timeouts[i] = random_to(cfg->min_timeout, cfg->max_timeout);
+
+		timeouts_add(tos, &t[i], timeouts[i] - (rel ? now : 0));
+		if (timeouts[i] <= cfg->start_at) {
+			if (timeout_pending(&t[i]))
+				FAIL();
+			if (! timeout_expired(&t[i]))
+				FAIL();
+			++n_added_expired;
+		} else {
+			if (! timeout_pending(&t[i]))
+				FAIL();
+			if (timeout_expired(&t[i]))
+				FAIL();
+			++n_added_pending;
+		}
+	}
+
+	if (!!n_added_pending != timeouts_pending(tos))
+		FAIL();
+	if (!!n_added_expired != timeouts_expired(tos))
+		FAIL();
+
+        /* Test foreach, interleaving a few iterators. */
+        TIMEOUTS_IT_INIT(&it_p, TIMEOUTS_PENDING);
+        TIMEOUTS_IT_INIT(&it_e, TIMEOUTS_EXPIRED);
+        TIMEOUTS_IT_INIT(&it_all, TIMEOUTS_ALL);
+        while (! (p_done && e_done && all_done)) {
+		if (!p_done) {
+			to = timeouts_next(tos, &it_p);
+			if (to) {
+				i = to - &t[0];
+				++found[i];
+				++cnt_added_pending;
+			} else {
+				p_done = 1;
+			}
+		}
+		if (!e_done) {
+			to = timeouts_next(tos, &it_e);
+			if (to) {
+				i = to - &t[0];
+				++found[i];
+				++cnt_added_expired;
+			} else {
+				e_done = 1;
+			}
+		}
+		if (!all_done) {
+			to = timeouts_next(tos, &it_all);
+			if (to) {
+				i = to - &t[0];
+				++found[i];
+			} else {
+				all_done = 1;
+			}
+		}
+        }
+
+	for (i = 0; i < cfg->n_timeouts; ++i) {
+		if (found[i] != 2)
+			FAIL();
+	}
+	if (cnt_added_expired != n_added_expired)
+		FAIL();
+	if (cnt_added_pending != n_added_pending)
+		FAIL();
+
+	while (NULL != (to = timeouts_get(tos))) {
+		i = to - &t[0];
+		assert(&t[i] == to);
+		if (timeouts[i] > cfg->start_at)
+			FAIL(); /* shouldn't have happened yet */
+
+		--n_added_expired; /* drop expired timeouts. */
+		++fired[i];
+	}
+
+	if (n_added_expired != 0)
+		FAIL();
+
+	while (now < cfg->end_at) {
+		int n_fired_this_time = 0;
+		timeout_t first_at = timeouts_timeout(tos) + now;
+
+		timeout_t oldtime = now;
+		timeout_t step = random_to(1, cfg->max_step);
+		int another;
+		now += step;
+		if (rel)
+			timeouts_step(tos, step);
+		else
+			timeouts_update(tos, now);
+
+		for (i = 0; i < cfg->try_removing; ++i) {
+			int idx = random() % cfg->n_timeouts;
+			if (! fired[idx]) {
+				timeout_del(&t[idx]);
+				++deleted[idx];
+			}
+		}
+
+		another = (timeouts_timeout(tos) == 0);
+
+		while (NULL != (to = timeouts_get(tos))) {
+			if (! another)
+				FAIL(); /* Thought we saw the last one! */
+			i = to - &t[0];
+			assert(&t[i] == to);
+			if (timeouts[i] > now)
+				FAIL(); /* shouldn't have happened yet */
+			if (timeouts[i] <= oldtime)
+				FAIL(); /* should have happened already */
+			if (timeouts[i] < first_at)
+				FAIL(); /* first_at should've been earlier */
+			fired[i]++;
+			n_fired_this_time++;
+			another = (timeouts_timeout(tos) == 0);
+		}
+		if (n_fired_this_time && first_at > now)
+			FAIL(); /* first_at should've been earlier */
+		if (another)
+			FAIL(); /* Huh? We think there are more? */
+		if (!timeouts_check(tos, stderr))
+			FAIL();
+	}
+
+	for (i = 0; i < cfg->n_timeouts; ++i) {
+		if (fired[i] > 1)
+			FAIL(); /* Nothing fired twice. */
+		if (timeouts[i] <= now) {
+			if (!(fired[i] || deleted[i]))
+				FAIL();
+		} else {
+			if (fired[i])
+				FAIL();
+		}
+		if (fired[i] && deleted[i])
+			FAIL();
+		if (cfg->finalize > 1) {
+			if (!fired[i])
+				timeout_del(&t[i]);
+		}
+	}
+
+	/* Now nothing more should fire between now and the end of time. */
+	if (cfg->finalize) {
+		timeouts_update(tos, THE_END_OF_TIME);
+		if (cfg->finalize > 1) {
+			if (timeouts_get(tos))
+				FAIL();
+			TIMEOUTS_FOREACH(to, tos, TIMEOUTS_ALL)
+				FAIL();
+		}
+	}
+	rv = 0;
+
+ done:
+	if (tos) timeouts_close(tos);
+	if (t) free(t);
+	if (timeouts) free(timeouts);
+	if (fired) free(fired);
+	if (found) free(found);
+	if (deleted) free(deleted);
+	return rv;
+}
+
+struct intervals_cfg {
+	const timeout_t *timeouts;
+	int n_timeouts;
+	timeout_t start_at;
+	timeout_t end_at;
+	timeout_t skip;
+};
+
+int
+check_intervals(struct intervals_cfg *cfg)
+{
+	int i, err;
+	int rv = 1;
+	struct timeout *to;
+	struct timeout *t = calloc(cfg->n_timeouts, sizeof(struct timeout));
+	unsigned *fired = calloc(cfg->n_timeouts, sizeof(unsigned));
+	struct timeouts *tos = timeouts_open(0, &err);
+
+	timeout_t now = cfg->start_at;
+	if (!t || !tos || !fired)
+		FAIL();
+
+	timeouts_update(tos, now);
+
+	for (i = 0; i < cfg->n_timeouts; ++i) {
+		if (&t[i] != timeout_init(&t[i], TIMEOUT_INT))
+			FAIL();
+		if (timeout_pending(&t[i]))
+			FAIL();
+		if (timeout_expired(&t[i]))
+			FAIL();
+
+		timeouts_add(tos, &t[i], cfg->timeouts[i]);
+		if (! timeout_pending(&t[i]))
+			FAIL();
+		if (timeout_expired(&t[i]))
+			FAIL();
+	}
+
+	while (now < cfg->end_at) {
+		timeout_t delay = timeouts_timeout(tos);
+		if (cfg->skip && delay < cfg->skip)
+			delay = cfg->skip;
+		timeouts_step(tos, delay);
+		now += delay;
+
+		while (NULL != (to = timeouts_get(tos))) {
+			i = to - &t[0];
+			assert(&t[i] == to);
+			fired[i]++;
+			if (0 != (to->expires - cfg->start_at) % cfg->timeouts[i])
+				FAIL();
+			if (to->expires <= now)
+				FAIL();
+			if (to->expires > now + cfg->timeouts[i])
+				FAIL();
+		}
+		if (!timeouts_check(tos, stderr))
+			FAIL();
+	}
+
+	timeout_t duration = now - cfg->start_at;
+	for (i = 0; i < cfg->n_timeouts; ++i) {
+		if (cfg->skip) {
+			if (fired[i] > duration / cfg->timeouts[i])
+				FAIL();
+		} else {
+			if (fired[i] != duration / cfg->timeouts[i])
+				FAIL();
+		}
+		if (!timeout_pending(&t[i]))
+			FAIL();
+	}
+
+	rv = 0;
+ done:
+	if (t) free(t);
+	if (fired) free(fired);
+	if (tos) free(tos);
+	return rv;
+}
+
+int
+main(int argc, char **argv)
+{
+	int j;
+	int n_failed = 0;
+#define DO(fn) do {                             \
+		printf("."); fflush(stdout);	\
+		if (fn) {			\
+			++n_failed;		\
+			printf("%s failed\n", #fn);	\
+		}					\
+        } while (0)
+
+#define DO_N(n, fn) do {			\
+		for (j = 0; j < (n); ++j) {	\
+			DO(fn);			\
+		}				\
+	} while (0)
+
+        DO(check_misc());
+	DO(check_open_close(1000, 1000));
+	DO(check_open_close(0, TIMEOUT_mHZ));
+
+	struct rand_cfg cfg1 = {
+		.min_timeout = 1,
+		.max_timeout = 100,
+		.start_at = 5,
+		.end_at = 1000,
+		.n_timeouts = 1000,
+		.max_step = 10,
+		.relative = 0,
+		.try_removing = 0,
+		.finalize = 2,
+		};
+	DO_N(300,check_randomized(&cfg1));
+
+	struct rand_cfg cfg2 = {
+		.min_timeout = 20,
+		.max_timeout = 1000,
+		.start_at = 10,
+		.end_at = 100,
+		.n_timeouts = 1000,
+		.max_step = 5,
+		.relative = 1,
+		.try_removing = 0,
+		.finalize = 2,
+		};
+	DO_N(300,check_randomized(&cfg2));
+
+	struct rand_cfg cfg2b = {
+		.min_timeout = 20,
+		.max_timeout = 1000,
+		.start_at = 10,
+		.end_at = 100,
+		.n_timeouts = 1000,
+		.max_step = 5,
+		.relative = 1,
+		.try_removing = 0,
+		.finalize = 1,
+		};
+	DO_N(300,check_randomized(&cfg2b));
+
+	struct rand_cfg cfg2c = {
+		.min_timeout = 20,
+		.max_timeout = 1000,
+		.start_at = 10,
+		.end_at = 100,
+		.n_timeouts = 1000,
+		.max_step = 5,
+		.relative = 1,
+		.try_removing = 0,
+		.finalize = 0,
+		};
+	DO_N(300,check_randomized(&cfg2c));
+
+	struct rand_cfg cfg3 = {
+		.min_timeout = 2000,
+		.max_timeout = ((uint64_t)1) << 50,
+		.start_at = 100,
+		.end_at = ((uint64_t)1) << 49,
+		.n_timeouts = 1000,
+		.max_step = 1<<31,
+		.relative = 0,
+		.try_removing = 0,
+		.finalize = 2,
+		};
+	DO_N(10,check_randomized(&cfg3));
+
+	struct rand_cfg cfg3b = {
+		.min_timeout = ((uint64_t)1) << 50,
+		.max_timeout = ((uint64_t)1) << 52,
+		.start_at = 100,
+		.end_at = ((uint64_t)1) << 53,
+		.n_timeouts = 1000,
+		.max_step = ((uint64_t)1)<<48,
+		.relative = 0,
+		.try_removing = 0,
+		.finalize = 2,
+		};
+	DO_N(10,check_randomized(&cfg3b));
+
+	struct rand_cfg cfg4 = {
+		.min_timeout = 2000,
+		.max_timeout = ((uint64_t)1) << 30,
+		.start_at = 100,
+		.end_at = ((uint64_t)1) << 26,
+		.n_timeouts = 10000,
+		.max_step = 1<<16,
+		.relative = 0,
+		.try_removing = 3,
+		.finalize = 2,
+		};
+	DO_N(10,check_randomized(&cfg4));
+
+	const timeout_t primes[] = {
+		2,3,5,7,11,13,17,19,23,29,31,37,41,43,47,53,
+		59,61,67,71,73,79,83,89,97
+	};
+	const timeout_t factors_of_1337[] = {
+		1, 7, 191, 1337
+	};
+	const timeout_t multiples_of_five[] = {
+		5, 10, 15, 20, 25, 30, 35, 40, 45, 50
+	};
+
+	struct intervals_cfg icfg1 = {
+		.timeouts = primes,
+		.n_timeouts = sizeof(primes)/sizeof(timeout_t),
+		.start_at = 50,
+		.end_at = 5322,
+		.skip = 0,
+	};
+	DO(check_intervals(&icfg1));
+
+	struct intervals_cfg icfg2 = {
+		.timeouts = factors_of_1337,
+		.n_timeouts = sizeof(factors_of_1337)/sizeof(timeout_t),
+		.start_at = 50,
+		.end_at = 50000,
+		.skip = 0,
+	};
+	DO(check_intervals(&icfg2));
+
+	struct intervals_cfg icfg3 = {
+		.timeouts = multiples_of_five,
+		.n_timeouts = sizeof(multiples_of_five)/sizeof(timeout_t),
+		.start_at = 49,
+		.end_at = 5333,
+		.skip = 0,
+	};
+	DO(check_intervals(&icfg3));
+
+	struct intervals_cfg icfg4 = {
+		.timeouts = primes,
+		.n_timeouts = sizeof(primes)/sizeof(timeout_t),
+		.start_at = 50,
+		.end_at = 5322,
+		.skip = 16,
+	};
+	DO(check_intervals(&icfg4));
+
+        if (n_failed) {
+          puts("\nFAIL");
+        } else {
+          puts("\nOK");
+        }
+	return !!n_failed;
+}
+
+/* TODO:
+
+ * Solve PR#3.
+
+ * Investigate whether any untaken branches are possible.
+
+ */

src/ext/timeouts/timeout-bitops.c

@@ -0,0 +1,254 @@
+#include <stdint.h>
+#include <limits.h>
+#ifdef _MSC_VER
+#include <intrin.h>     /* _BitScanForward, _BitScanReverse */
+#endif
+
+/* First define ctz and clz functions; these are compiler-dependent if
+ * you want them to be fast. */
+#if defined(__GNUC__) && !defined(TIMEOUT_DISABLE_GNUC_BITOPS)
+
+#ifndef LONG_BIT
+#define LONG_BIT (SIZEOF_LONG*CHAR_BIT)
+#endif
+
+/* On GCC and clang and some others, we can use __builtin functions. They
+ * are not defined for n==0, but timeout.s never calls them with n==0. */
+
+#define ctz64(n) __builtin_ctzll(n)
+#define clz64(n) __builtin_clzll(n)
+#if LONG_BIT == 32
+#define ctz32(n) __builtin_ctzl(n)
+#define clz32(n) __builtin_clzl(n)
+#else
+#define ctz32(n) __builtin_ctz(n)
+#define clz32(n) __builtin_clz(n)
+#endif
+
+#elif defined(_MSC_VER) && !defined(TIMEOUT_DISABLE_MSVC_BITOPS)
+
+/* On MSVC, we have these handy functions. We can ignore their return
+ * values, since we will never supply val == 0. */
+
+static __inline int ctz32(unsigned long val)
+{
+	DWORD zeros = 0;
+	_BitScanForward(&zeros, val);
+	return zeros;
+}
+static __inline int clz32(unsigned long val)
+{
+	DWORD zeros = 0;
+	_BitScanReverse(&zeros, val);
+	return zeros;
+}
+#ifdef _WIN64
+/* According to the documentation, these only exist on Win64. */
+static __inline int ctz64(uint64_t val)
+{
+	DWORD zeros = 0;
+	_BitScanForward64(&zeros, val);
+	return zeros;
+}
+static __inline int clz64(uint64_t val)
+{
+	DWORD zeros = 0;
+	_BitScanReverse64(&zeros, val);
+	return zeros;
+}
+#else
+static __inline int ctz64(uint64_t val)
+{
+	uint32_t lo = (uint32_t) val;
+	uint32_t hi = (uint32_t) (val >> 32);
+	return lo ? ctz32(lo) : 32 + ctz32(hi);
+}
+static __inline int clz64(uint64_t val)
+{
+	uint32_t lo = (uint32_t) val;
+	uint32_t hi = (uint32_t) (val >> 32);
+	return hi ? clz32(hi) : 32 + clz32(lo);
+}
+#endif
+
+/* End of MSVC case. */
+
+#else
+
+/* TODO: There are more clever ways to do this in the generic case. */
+
+
+#define process_(one, cz_bits, bits)					\
+	if (x < ( one << (cz_bits - bits))) { rv += bits; x <<= bits; }
+
+#define process64(bits) process_((UINT64_C(1)), 64, (bits))
+static inline int clz64(uint64_t x)
+{
+	int rv = 0;
+
+	process64(32);
+	process64(16);
+	process64(8);
+	process64(4);
+	process64(2);
+	process64(1);
+	return rv;
+}
+#define process32(bits) process_((UINT32_C(1)), 32, (bits))
+static inline int clz32(uint32_t x)
+{
+	int rv = 0;
+
+	process32(16);
+	process32(8);
+	process32(4);
+	process32(2);
+	process32(1);
+	return rv;
+}
+
+#undef process_
+#undef process32
+#undef process64
+#define process_(one, bits)						\
+	if ((x & ((one << (bits))-1)) == 0) { rv += bits; x >>= bits; }
+
+#define process64(bits) process_((UINT64_C(1)), bits)
+static inline int ctz64(uint64_t x)
+{
+	int rv = 0;
+
+	process64(32);
+	process64(16);
+	process64(8);
+	process64(4);
+	process64(2);
+	process64(1);
+	return rv;
+}
+
+#define process32(bits) process_((UINT32_C(1)), bits)
+static inline int ctz32(uint32_t x)
+{
+	int rv = 0;
+
+	process32(16);
+	process32(8);
+	process32(4);
+	process32(2);
+	process32(1);
+	return rv;
+}
+
+#undef process32
+#undef process64
+#undef process_
+
+/* End of generic case */
+
+#endif /* End of defining ctz */
+
+#ifdef TEST_BITOPS
+#include <stdio.h>
+#include <stdlib.h>
+
+static uint64_t testcases[] = {
+	13371337 * 10,
+	100,
+	385789752,
+	82574,
+	(((uint64_t)1)<<63) + (((uint64_t)1)<<31) + 10101
+};
+
+static int
+naive_clz(int bits, uint64_t v)
+{
+	int r = 0;
+	uint64_t bit = ((uint64_t)1) << (bits-1);
+	while (bit && 0 == (v & bit)) {
+		r++;
+		bit >>= 1;
+	}
+	/* printf("clz(%d,%lx) -> %d\n", bits, v, r); */
+	return r;
+}
+
+static int
+naive_ctz(int bits, uint64_t v)
+{
+	int r = 0;
+	uint64_t bit = 1;
+	while (bit && 0 == (v & bit)) {
+		r++;
+		bit <<= 1;
+		if (r == bits)
+			break;
+	}
+	/* printf("ctz(%d,%lx) -> %d\n", bits, v, r); */
+	return r;
+}
+
+static int
+check(uint64_t vv)
+{
+	uint32_t v32 = (uint32_t) vv;
+
+	if (vv == 0)
+		return 1; /* c[tl]z64(0) is undefined. */
+
+	if (ctz64(vv) != naive_ctz(64, vv)) {
+		printf("mismatch with ctz64: %d\n", ctz64(vv));
+				exit(1);
+		return 0;
+	}
+	if (clz64(vv) != naive_clz(64, vv)) {
+		printf("mismatch with clz64: %d\n", clz64(vv));
+				exit(1);
+		return 0;
+	}
+
+	if (v32 == 0)
+		return 1; /* c[lt]z(0) is undefined. */
+
+	if (ctz32(v32) != naive_ctz(32, v32)) {
+		printf("mismatch with ctz32: %d\n", ctz32(v32));
+		exit(1);
+		return 0;
+	}
+	if (clz32(v32) != naive_clz(32, v32)) {
+		printf("mismatch with clz32: %d\n", clz32(v32));
+				exit(1);
+		return 0;
+	}
+	return 1;
+}
+
+int
+main(int c, char **v)
+{
+	unsigned int i;
+	const unsigned int n = sizeof(testcases)/sizeof(testcases[0]);
+	int result = 0;
+
+	for (i = 0; i <= 63; ++i) {
+		uint64_t x = 1 << i;
+		if (!check(x))
+			result = 1;
+		--x;
+		if (!check(x))
+			result = 1;
+	}
+
+	for (i = 0; i < n; ++i) {
+		if (! check(testcases[i]))
+			result = 1;
+	}
+	if (result) {
+		puts("FAIL");
+	} else {
+		puts("OK");
+	}
+	return result;
+}
+#endif
+

src/ext/timeouts/timeout-debug.h

@@ -0,0 +1,77 @@
+/*
+ * D E B U G  R O U T I N E S
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+#if TIMEOUT_DEBUG - 0
+#include <stdlib.h>
+#include <stdio.h>
+
+#undef TIMEOUT_DEBUG
+#define TIMEOUT_DEBUG 1
+#define DEBUG_LEVEL timeout_debug
+
+static int timeout_debug;
+
+#define SAYit_(lvl, fmt, ...) do { \
+	if (DEBUG_LEVEL >= (lvl)) \
+		fprintf(stderr, fmt "%s", __FILE__, __LINE__, __func__, __VA_ARGS__); \
+} while (0)
+
+#define SAYit(lvl, ...) SAYit_((lvl), "%s:%d:%s: " __VA_ARGS__, "\n")
+
+#define PANIC(...) do { \
+	SAYit(0, __VA_ARGS__); \
+	_Exit(EXIT_FAILURE); \
+} while (0)
+#else
+#undef TIMEOUT_DEBUG
+#define TIMEOUT_DEBUG 0
+#define DEBUG_LEVEL 0
+
+#define SAYit(...) (void)0
+#endif
+
+#define SAY(...) SAYit(1, __VA_ARGS__)
+#define HAI SAY("HAI")
+
+
+static inline char *fmt_(char *buf, uint64_t ts, int wheel_bit, int wheel_num) {
+	char *p = buf;
+	int wheel, n, i;
+
+	for (wheel = wheel_num - 2; wheel >= 0; wheel--) {
+		n = ((1 << wheel_bit) - 1) & (ts >> (wheel * WHEEL_BIT));
+
+		for (i = wheel_bit - 1; i >= 0; i--) {
+			*p++ = '0' + !!(n & (1 << i));
+		}
+
+		if (wheel != 0)
+			*p++ = ':';
+	}
+
+	*p = 0;
+
+	return buf;
+} /* fmt_() */
+
+#define fmt(ts) fmt_(((char[((1 << WHEEL_BIT) * WHEEL_NUM) + WHEEL_NUM + 1]){ 0 }), (ts), WHEEL_BIT, WHEEL_NUM)
+
+
+static inline char *bin64_(char *buf, uint64_t n, int wheel_bit) {
+	char *p = buf;
+	int i;
+
+	for (i = 0; i < (1 << wheel_bit); i++) {
+		*p++ = "01"[0x1 & (n >> (((1 << wheel_bit) - 1) - i))];
+	}
+
+	*p = 0;
+
+	return buf;
+} /* bin64_() */
+
+#define bin64(ts) bin64_(((char[((1 << WHEEL_BIT) * WHEEL_NUM) + 1]){ 0 }), (ts), WHEEL_BIT)
+
+

src/ext/timeouts/timeout.c

@@ -0,0 +1,754 @@
+/* ==========================================================================
+ * timeout.c - Tickless hierarchical timing wheel.
+ * --------------------------------------------------------------------------
+ * Copyright (c) 2013, 2014  William Ahern
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to permit
+ * persons to whom the Software is furnished to do so, subject to the
+ * following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN
+ * NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * ==========================================================================
+ */
+#ifdef HAVE_CONFIG_H
+#include "orconfig.h"
+#endif
+#include <limits.h>    /* CHAR_BIT */
+
+#include <stddef.h>    /* NULL */
+#include <stdlib.h>    /* malloc(3) free(3) */
+#include <stdio.h>     /* FILE fprintf(3) */
+
+#include <inttypes.h>  /* UINT64_C uint64_t */
+
+#include <string.h>    /* memset(3) */
+
+#include <errno.h>     /* errno */
+
+#include <sys/queue.h> /* TAILQ(3) */
+
+#include "timeout.h"
+
+#ifndef TIMEOUT_DEBUG
+#define TIMEOUT_DEBUG 0
+#endif
+
+#if TIMEOUT_DEBUG - 0
+#include "timeout-debug.h"
+#endif
+
+#ifdef TIMEOUT_DISABLE_RELATIVE_ACCESS
+#define TO_SET_TIMEOUTS(to, T) ((void)0)
+#else
+#define TO_SET_TIMEOUTS(to, T) ((to)->timeouts = (T))
+#endif
+
+/*
+ * A N C I L L A R Y  R O U T I N E S
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+#define abstime_t timeout_t /* for documentation purposes */
+#define reltime_t timeout_t /* "" */
+
+#if !defined countof
+#define countof(a) (sizeof (a) / sizeof *(a))
+#endif
+
+#if !defined endof
+#define endof(a) (&(a)[countof(a)])
+#endif
+
+#if !defined MIN
+#define MIN(a, b) (((a) < (b))? (a) : (b))
+#endif
+
+#if !defined MAX
+#define MAX(a, b) (((a) > (b))? (a) : (b))
+#endif
+
+#if !defined TAILQ_CONCAT
+#define TAILQ_CONCAT(head1, head2, field) do {                          \
+	if (!TAILQ_EMPTY(head2)) {                                      \
+		*(head1)->tqh_last = (head2)->tqh_first;                \
+		(head2)->tqh_first->field.tqe_prev = (head1)->tqh_last; \
+		(head1)->tqh_last = (head2)->tqh_last;                  \
+		TAILQ_INIT((head2));                                    \
+	}                                                               \
+} while (0)
+#endif
+
+#if !defined TAILQ_FOREACH_SAFE
+#define TAILQ_FOREACH_SAFE(var, head, field, tvar)                      \
+	for ((var) = TAILQ_FIRST(head);                                 \
+	    (var) && ((tvar) = TAILQ_NEXT(var, field), 1);              \
+	    (var) = (tvar))
+#endif
+
+
+/*
+ * B I T  M A N I P U L A T I O N  R O U T I N E S
+ *
+ * The macros and routines below implement wheel parameterization. The
+ * inputs are:
+ *
+ *   WHEEL_BIT - The number of value bits mapped in each wheel. The
+ *               lowest-order WHEEL_BIT bits index the lowest-order (highest
+ *               resolution) wheel, the next group of WHEEL_BIT bits the
+ *               higher wheel, etc.
+ *
+ *   WHEEL_NUM - The number of wheels. WHEEL_BIT * WHEEL_NUM = the number of
+ *               value bits used by all the wheels. For the default of 6 and
+ *               4, only the low 24 bits are processed. Any timeout value
+ *               larger than this will cycle through again.
+ *
+ * The implementation uses bit fields to remember which slot in each wheel
+ * is populated, and to generate masks of expiring slots according to the
+ * current update interval (i.e. the "tickless" aspect). The slots to
+ * process in a wheel are (populated-set & interval-mask).
+ *
+ * WHEEL_BIT cannot be larger than 6 bits because 2^6 -> 64 is the largest
+ * number of slots which can be tracked in a uint64_t integer bit field.
+ * WHEEL_BIT cannot be smaller than 3 bits because of our rotr and rotl
+ * routines, which only operate on all the value bits in an integer, and
+ * there's no integer smaller than uint8_t.
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+#if !defined WHEEL_BIT
+#define WHEEL_BIT 6
+#endif
+
+#if !defined WHEEL_NUM
+#define WHEEL_NUM 4
+#endif
+
+#define WHEEL_LEN (1U << WHEEL_BIT)
+#define WHEEL_MAX (WHEEL_LEN - 1)
+#define WHEEL_MASK (WHEEL_LEN - 1)
+#define TIMEOUT_MAX ((TIMEOUT_C(1) << (WHEEL_BIT * WHEEL_NUM)) - 1)
+
+#include "timeout-bitops.c"
+
+#if WHEEL_BIT == 6
+#define ctz(n) ctz64(n)
+#define clz(n) clz64(n)
+#define fls(n) ((int)(64 - clz64(n)))
+#else
+#define ctz(n) ctz32(n)
+#define clz(n) clz32(n)
+#define fls(n) ((int)(32 - clz32(n)))
+#endif
+
+#if WHEEL_BIT == 6
+#define WHEEL_C(n) UINT64_C(n)
+#define WHEEL_PRIu PRIu64
+#define WHEEL_PRIx PRIx64
+
+typedef uint64_t wheel_t;
+
+#elif WHEEL_BIT == 5
+
+#define WHEEL_C(n) UINT32_C(n)
+#define WHEEL_PRIu PRIu32
+#define WHEEL_PRIx PRIx32
+
+typedef uint32_t wheel_t;
+
+#elif WHEEL_BIT == 4
+
+#define WHEEL_C(n) UINT16_C(n)
+#define WHEEL_PRIu PRIu16
+#define WHEEL_PRIx PRIx16
+
+typedef uint16_t wheel_t;
+
+#elif WHEEL_BIT == 3
+
+#define WHEEL_C(n) UINT8_C(n)
+#define WHEEL_PRIu PRIu8
+#define WHEEL_PRIx PRIx8
+
+typedef uint8_t wheel_t;
+
+#else
+#error invalid WHEEL_BIT value
+#endif
+
+
+static inline wheel_t rotl(const wheel_t v, int c) {
+	if (!(c &= (sizeof v * CHAR_BIT - 1)))
+		return v;
+
+	return (v << c) | (v >> (sizeof v * CHAR_BIT - c));
+} /* rotl() */
+
+
+static inline wheel_t rotr(const wheel_t v, int c) {
+	if (!(c &= (sizeof v * CHAR_BIT - 1)))
+		return v;
+
+	return (v >> c) | (v << (sizeof v * CHAR_BIT - c));
+} /* rotr() */
+
+
+/*
+ * T I M E R  R O U T I N E S
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+TAILQ_HEAD(timeout_list, timeout);
+
+struct timeouts {
+	struct timeout_list wheel[WHEEL_NUM][WHEEL_LEN], expired;
+
+	wheel_t pending[WHEEL_NUM];
+
+	timeout_t curtime;
+	timeout_t hertz;
+}; /* struct timeouts */
+
+
+static struct timeouts *timeouts_init(struct timeouts *T, timeout_t hz) {
+	unsigned i, j;
+
+	for (i = 0; i < countof(T->wheel); i++) {
+		for (j = 0; j < countof(T->wheel[i]); j++) {
+			TAILQ_INIT(&T->wheel[i][j]);
+		}
+	}
+
+	TAILQ_INIT(&T->expired);
+
+	for (i = 0; i < countof(T->pending); i++) {
+		T->pending[i] = 0;
+	}
+
+	T->curtime = 0;
+	T->hertz = (hz)? hz : TIMEOUT_mHZ;
+
+	return T;
+} /* timeouts_init() */
+
+
+TIMEOUT_PUBLIC struct timeouts *timeouts_open(timeout_t hz, int *error) {
+	struct timeouts *T;
+
+	if ((T = malloc(sizeof *T)))
+		return timeouts_init(T, hz);
+
+	*error = errno;
+
+	return NULL;
+} /* timeouts_open() */
+
+
+static void timeouts_reset(struct timeouts *T) {
+	struct timeout_list reset;
+	struct timeout *to;
+	unsigned i, j;
+
+	TAILQ_INIT(&reset);
+
+	for (i = 0; i < countof(T->wheel); i++) {
+		for (j = 0; j < countof(T->wheel[i]); j++) {
+			TAILQ_CONCAT(&reset, &T->wheel[i][j], tqe);
+		}
+	}
+
+	TAILQ_CONCAT(&reset, &T->expired, tqe);
+
+	TAILQ_FOREACH(to, &reset, tqe) {
+		to->pending = NULL;
+		TO_SET_TIMEOUTS(to, NULL);
+	}
+} /* timeouts_reset() */
+
+
+TIMEOUT_PUBLIC void timeouts_close(struct timeouts *T) {
+	/*
+	 * NOTE: Delete installed timeouts so timeout_pending() and
+	 * timeout_expired() worked as expected.
+	 */
+	timeouts_reset(T);
+
+	free(T);
+} /* timeouts_close() */
+
+
+TIMEOUT_PUBLIC timeout_t timeouts_hz(struct timeouts *T) {
+	return T->hertz;
+} /* timeouts_hz() */
+
+
+TIMEOUT_PUBLIC void timeouts_del(struct timeouts *T, struct timeout *to) {
+	if (to->pending) {
+		TAILQ_REMOVE(to->pending, to, tqe);
+
+		if (to->pending != &T->expired && TAILQ_EMPTY(to->pending)) {
+			ptrdiff_t index = to->pending - &T->wheel[0][0];
+			int wheel = (int) (index / WHEEL_LEN);
+			int slot = index % WHEEL_LEN;
+
+			T->pending[wheel] &= ~(WHEEL_C(1) << slot);
+		}
+
+		to->pending = NULL;
+		TO_SET_TIMEOUTS(to, NULL);
+	}
+} /* timeouts_del() */
+
+
+static inline reltime_t timeout_rem(struct timeouts *T, struct timeout *to) {
+	return to->expires - T->curtime;
+} /* timeout_rem() */
+
+
+static inline int timeout_wheel(timeout_t timeout) {
+	/* must be called with timeout != 0, so fls input is nonzero */
+	return (fls(MIN(timeout, TIMEOUT_MAX)) - 1) / WHEEL_BIT;
+} /* timeout_wheel() */
+
+
+static inline int timeout_slot(int wheel, timeout_t expires) {
+	return WHEEL_MASK & ((expires >> (wheel * WHEEL_BIT)) - !!wheel);
+} /* timeout_slot() */
+
+
+static void timeouts_sched(struct timeouts *T, struct timeout *to, timeout_t expires) {
+	timeout_t rem;
+	int wheel, slot;
+
+	timeouts_del(T, to);
+
+	to->expires = expires;
+
+	TO_SET_TIMEOUTS(to, T);
+
+	if (expires > T->curtime) {
+		rem = timeout_rem(T, to);
+
+		/* rem is nonzero since:
+		 *   rem == timeout_rem(T,to),
+		 *       == to->expires - T->curtime
+		 *   and above we have expires > T->curtime.
+		 */
+		wheel = timeout_wheel(rem);
+		slot = timeout_slot(wheel, to->expires);
+
+		to->pending = &T->wheel[wheel][slot];
+		TAILQ_INSERT_TAIL(to->pending, to, tqe);
+
+		T->pending[wheel] |= WHEEL_C(1) << slot;
+	} else {
+		to->pending = &T->expired;
+		TAILQ_INSERT_TAIL(to->pending, to, tqe);
+	}
+} /* timeouts_sched() */
+
+
+#ifndef TIMEOUT_DISABLE_INTERVALS
+static void timeouts_readd(struct timeouts *T, struct timeout *to) {
+	to->expires += to->interval;
+
+	if (to->expires <= T->curtime) {
+		/* If we've missed the next firing of this timeout, reschedule
+		 * it to occur at the next multiple of its interval after
+		 * the last time that it fired.
+		 */
+		timeout_t n = T->curtime - to->expires;
+		timeout_t r = n % to->interval;
+		to->expires = T->curtime + (to->interval - r);
+	}
+
+	timeouts_sched(T, to, to->expires);
+} /* timeouts_readd() */
+#endif
+
+
+TIMEOUT_PUBLIC void timeouts_add(struct timeouts *T, struct timeout *to, timeout_t timeout) {
+#ifndef TIMEOUT_DISABLE_INTERVALS
+	if (to->flags & TIMEOUT_INT)
+		to->interval = MAX(1, timeout);
+#endif
+
+	if (to->flags & TIMEOUT_ABS)
+		timeouts_sched(T, to, timeout);
+	else
+		timeouts_sched(T, to, T->curtime + timeout);
+} /* timeouts_add() */
+
+
+TIMEOUT_PUBLIC void timeouts_update(struct timeouts *T, abstime_t curtime) {
+	timeout_t elapsed = curtime - T->curtime;
+	struct timeout_list todo;
+	int wheel;
+
+	TAILQ_INIT(&todo);
+
+	/*
+	 * There's no avoiding looping over every wheel. It's best to keep
+	 * WHEEL_NUM smallish.
+	 */
+	for (wheel = 0; wheel < WHEEL_NUM; wheel++) {
+		wheel_t pending;
+
+		/*
+		 * Calculate the slots expiring in this wheel
+		 *
+		 * If the elapsed time is greater than the maximum period of
+		 * the wheel, mark every position as expiring.
+		 *
+		 * Otherwise, to determine the expired slots fill in all the
+		 * bits between the last slot processed and the current
+		 * slot, inclusive of the last slot. We'll bitwise-AND this
+		 * with our pending set below.
+		 *
+		 * If a wheel rolls over, force a tick of the next higher
+		 * wheel.
+		 */
+		if ((elapsed >> (wheel * WHEEL_BIT)) > WHEEL_MAX) {
+			pending = (wheel_t)~WHEEL_C(0);
+		} else {
+			wheel_t _elapsed = WHEEL_MASK & (elapsed >> (wheel * WHEEL_BIT));
+			int oslot, nslot;
+
+			/*
+			 * TODO: It's likely that at least one of the
+			 * following three bit fill operations is redundant
+			 * or can be replaced with a simpler operation.
+			 */
+			oslot = WHEEL_MASK & (T->curtime >> (wheel * WHEEL_BIT));
+			pending = rotl(((UINT64_C(1) << _elapsed) - 1), oslot);
+
+			nslot = WHEEL_MASK & (curtime >> (wheel * WHEEL_BIT));
+			pending |= rotr(rotl(((WHEEL_C(1) << _elapsed) - 1), nslot), (int)_elapsed);
+			pending |= WHEEL_C(1) << nslot;
+		}
+
+		while (pending & T->pending[wheel]) {
+			/* ctz input cannot be zero: loop condition. */
+			int slot = ctz(pending & T->pending[wheel]);
+			TAILQ_CONCAT(&todo, &T->wheel[wheel][slot], tqe);
+			T->pending[wheel] &= ~(UINT64_C(1) << slot);
+		}
+
+		if (!(0x1 & pending))
+			break; /* break if we didn't wrap around end of wheel */
+
+		/* if we're continuing, the next wheel must tick at least once */
+		elapsed = MAX(elapsed, (WHEEL_LEN << (wheel * WHEEL_BIT)));
+	}
+
+	T->curtime = curtime;
+
+	while (!TAILQ_EMPTY(&todo)) {
+		struct timeout *to = TAILQ_FIRST(&todo);
+
+		TAILQ_REMOVE(&todo, to, tqe);
+		to->pending = NULL;
+
+		timeouts_sched(T, to, to->expires);
+	}
+
+	return;
+} /* timeouts_update() */
+
+TIMEOUT_PUBLIC timeout_t timeouts_get_curtime(struct timeouts *T) {
+	return T->curtime;
+} /* timeouts_get_curtime() */
+
+TIMEOUT_PUBLIC void timeouts_step(struct timeouts *T, reltime_t elapsed) {
+	timeouts_update(T, T->curtime + elapsed);
+} /* timeouts_step() */
+
+
+TIMEOUT_PUBLIC bool timeouts_pending(struct timeouts *T) {
+	wheel_t pending = 0;
+	int wheel;
+
+	for (wheel = 0; wheel < WHEEL_NUM; wheel++) {
+		pending |= T->pending[wheel];
+	}
+
+	return !!pending;
+} /* timeouts_pending() */
+
+
+TIMEOUT_PUBLIC bool timeouts_expired(struct timeouts *T) {
+	return !TAILQ_EMPTY(&T->expired);
+} /* timeouts_expired() */
+
+
+/*
+ * Calculate the interval before needing to process any timeouts pending on
+ * any wheel.
+ *
+ * (This is separated from the public API routine so we can evaluate our
+ * wheel invariant assertions irrespective of the expired queue.)
+ *
+ * This might return a timeout value sooner than any installed timeout if
+ * only higher-order wheels have timeouts pending. We can only know when to
+ * process a wheel, not precisely when a timeout is scheduled. Our timeout
+ * accuracy could be off by 2^(N*M)-1 units where N is the wheel number and
+ * M is WHEEL_BIT. Only timeouts which have fallen through to wheel 0 can be
+ * known exactly.
+ *
+ * We should never return a timeout larger than the lowest actual timeout.
+ */
+static timeout_t timeouts_int(struct timeouts *T) {
+	timeout_t timeout = ~TIMEOUT_C(0), _timeout;
+	timeout_t relmask;
+	int wheel, slot;
+
+	relmask = 0;
+
+	for (wheel = 0; wheel < WHEEL_NUM; wheel++) {
+		if (T->pending[wheel]) {
+			slot = WHEEL_MASK & (T->curtime >> (wheel * WHEEL_BIT));
+
+			/* ctz input cannot be zero: T->pending[wheel] is
+			 * nonzero, so rotr() is nonzero. */
+			_timeout = (ctz(rotr(T->pending[wheel], slot)) + !!wheel) << (wheel * WHEEL_BIT);
+			/* +1 to higher order wheels as those timeouts are one rotation in the future (otherwise they'd be on a lower wheel or expired) */
+
+			_timeout -= relmask & T->curtime;
+			/* reduce by how much lower wheels have progressed */
+
+			timeout = MIN(_timeout, timeout);
+		}
+
+		relmask <<= WHEEL_BIT; 
+		relmask |= WHEEL_MASK;
+	}
+
+	return timeout;
+} /* timeouts_int() */
+
+
+/*
+ * Calculate the interval our caller can wait before needing to process
+ * events.
+ */
+TIMEOUT_PUBLIC timeout_t timeouts_timeout(struct timeouts *T) {
+	if (!TAILQ_EMPTY(&T->expired))
+		return 0;
+
+	return timeouts_int(T);
+} /* timeouts_timeout() */
+
+
+TIMEOUT_PUBLIC struct timeout *timeouts_get(struct timeouts *T) {
+	if (!TAILQ_EMPTY(&T->expired)) {
+		struct timeout *to = TAILQ_FIRST(&T->expired);
+
+		TAILQ_REMOVE(&T->expired, to, tqe);
+		to->pending = NULL;
+		TO_SET_TIMEOUTS(to, NULL);
+
+#ifndef TIMEOUT_DISABLE_INTERVALS
+		if ((to->flags & TIMEOUT_INT) && to->interval > 0)
+			timeouts_readd(T, to);
+#endif
+
+		return to;
+	} else {
+		return 0;
+	}
+} /* timeouts_get() */
+
+
+/*
+ * Use dumb looping to locate the earliest timeout pending on the wheel so
+ * our invariant assertions can check the result of our optimized code.
+ */
+static struct timeout *timeouts_min(struct timeouts *T) {
+	struct timeout *to, *min = NULL;
+	unsigned i, j;
+
+	for (i = 0; i < countof(T->wheel); i++) {
+		for (j = 0; j < countof(T->wheel[i]); j++) {
+			TAILQ_FOREACH(to, &T->wheel[i][j], tqe) {
+				if (!min || to->expires < min->expires)
+					min = to;
+			}
+		}
+	}
+
+	return min;
+} /* timeouts_min() */
+
+
+/*
+ * Check some basic algorithm invariants. If these invariants fail then
+ * something is definitely broken.
+ */
+#define report(...) do { \
+	if ((fp)) \
+		fprintf(fp, __VA_ARGS__); \
+} while (0)
+
+#define check(expr, ...) do { \
+	if (!(expr)) { \
+		report(__VA_ARGS__); \
+		return 0; \
+	} \
+} while (0)
+
+TIMEOUT_PUBLIC bool timeouts_check(struct timeouts *T, FILE *fp) {
+	timeout_t timeout;
+	struct timeout *to;
+
+	if ((to = timeouts_min(T))) {
+		check(to->expires > T->curtime, "missed timeout (expires:%" TIMEOUT_PRIu " <= curtime:%" TIMEOUT_PRIu ")\n", to->expires, T->curtime);
+
+		timeout = timeouts_int(T);
+		check(timeout <= to->expires - T->curtime, "wrong soft timeout (soft:%" TIMEOUT_PRIu " > hard:%" TIMEOUT_PRIu ") (expires:%" TIMEOUT_PRIu "; curtime:%" TIMEOUT_PRIu ")\n", timeout, (to->expires - T->curtime), to->expires, T->curtime);
+
+		timeout = timeouts_timeout(T);
+		check(timeout <= to->expires - T->curtime, "wrong soft timeout (soft:%" TIMEOUT_PRIu " > hard:%" TIMEOUT_PRIu ") (expires:%" TIMEOUT_PRIu "; curtime:%" TIMEOUT_PRIu ")\n", timeout, (to->expires - T->curtime), to->expires, T->curtime);
+	} else {
+		timeout = timeouts_timeout(T);
+
+		if (!TAILQ_EMPTY(&T->expired))
+			check(timeout == 0, "wrong soft timeout (soft:%" TIMEOUT_PRIu " != hard:%" TIMEOUT_PRIu ")\n", timeout, TIMEOUT_C(0));
+		else
+			check(timeout == ~TIMEOUT_C(0), "wrong soft timeout (soft:%" TIMEOUT_PRIu " != hard:%" TIMEOUT_PRIu ")\n", timeout, ~TIMEOUT_C(0));
+	}
+
+	return 1;
+} /* timeouts_check() */
+
+
+#define ENTER                                                           \
+	do {                                                            \
+	static const int pc0 = __LINE__;                                \
+	switch (pc0 + it->pc) {                                         \
+	case __LINE__: (void)0
+
+#define SAVE_AND_DO(do_statement)                                       \
+	do {                                                            \
+		it->pc = __LINE__ - pc0;                                \
+		do_statement;                                           \
+		case __LINE__: (void)0;                                 \
+	} while (0)
+
+#define YIELD(rv)                                                       \
+	SAVE_AND_DO(return (rv))
+
+#define LEAVE                                                           \
+	SAVE_AND_DO(break);                                             \
+	}                                                               \
+	} while (0)
+
+TIMEOUT_PUBLIC struct timeout *timeouts_next(struct timeouts *T, struct timeouts_it *it) {
+	struct timeout *to;
+
+	ENTER;
+
+	if (it->flags & TIMEOUTS_EXPIRED) {
+		if (it->flags & TIMEOUTS_CLEAR) {
+			while ((to = timeouts_get(T))) {
+				YIELD(to);
+			}
+		} else {
+			TAILQ_FOREACH_SAFE(to, &T->expired, tqe, it->to) {
+				YIELD(to);
+			}
+		}
+	}
+
+	if (it->flags & TIMEOUTS_PENDING) {
+		for (it->i = 0; it->i < countof(T->wheel); it->i++) {
+			for (it->j = 0; it->j < countof(T->wheel[it->i]); it->j++) {
+				TAILQ_FOREACH_SAFE(to, &T->wheel[it->i][it->j], tqe, it->to) {
+					YIELD(to);
+				}
+			}
+		}
+	}
+
+	LEAVE;
+
+	return NULL;
+} /* timeouts_next */
+
+#undef LEAVE
+#undef YIELD
+#undef SAVE_AND_DO
+#undef ENTER
+
+
+/*
+ * T I M E O U T  R O U T I N E S
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+TIMEOUT_PUBLIC struct timeout *timeout_init(struct timeout *to, int flags) {
+	memset(to, 0, sizeof *to);
+
+	to->flags = flags;
+
+	return to;
+} /* timeout_init() */
+
+
+#ifndef TIMEOUT_DISABLE_RELATIVE_ACCESS
+TIMEOUT_PUBLIC bool timeout_pending(struct timeout *to) {
+	return to->pending && to->pending != &to->timeouts->expired;
+} /* timeout_pending() */
+
+
+TIMEOUT_PUBLIC bool timeout_expired(struct timeout *to) {
+	return to->pending && to->pending == &to->timeouts->expired;
+} /* timeout_expired() */
+
+
+TIMEOUT_PUBLIC void timeout_del(struct timeout *to) {
+	timeouts_del(to->timeouts, to);
+} /* timeout_del() */
+#endif
+
+
+/*
+ * V E R S I O N  I N T E R F A C E S
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+TIMEOUT_PUBLIC int timeout_version(void) {
+	return TIMEOUT_VERSION;
+} /* timeout_version() */
+
+
+TIMEOUT_PUBLIC const char *timeout_vendor(void) {
+	return TIMEOUT_VENDOR;
+} /* timeout_version() */
+
+
+TIMEOUT_PUBLIC int timeout_v_rel(void) {
+	return TIMEOUT_V_REL;
+} /* timeout_version() */
+
+
+TIMEOUT_PUBLIC int timeout_v_abi(void) {
+	return TIMEOUT_V_ABI;
+} /* timeout_version() */
+
+
+TIMEOUT_PUBLIC int timeout_v_api(void) {
+	return TIMEOUT_V_API;
+} /* timeout_version() */
+

src/ext/timeouts/timeout.h

@@ -0,0 +1,256 @@
+/* ==========================================================================
+ * timeout.h - Tickless hierarchical timing wheel.
+ * --------------------------------------------------------------------------
+ * Copyright (c) 2013, 2014  William Ahern
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to permit
+ * persons to whom the Software is furnished to do so, subject to the
+ * following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN
+ * NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * ==========================================================================
+ */
+#ifndef TIMEOUT_H
+#define TIMEOUT_H
+
+#include <stdbool.h>    /* bool */
+#include <stdio.h>      /* FILE */
+
+#include <inttypes.h>   /* PRIu64 PRIx64 PRIX64 uint64_t */
+
+#include <sys/queue.h>  /* TAILQ(3) */
+
+
+/*
+ * V E R S I O N  I N T E R F A C E S
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+#if !defined TIMEOUT_PUBLIC
+#define TIMEOUT_PUBLIC
+#endif
+
+#define TIMEOUT_VERSION TIMEOUT_V_REL
+#define TIMEOUT_VENDOR  "william@25thandClement.com"
+
+#define TIMEOUT_V_REL 0x20160226
+#define TIMEOUT_V_ABI 0x20160224
+#define TIMEOUT_V_API 0x20160226
+
+TIMEOUT_PUBLIC int timeout_version(void);
+
+TIMEOUT_PUBLIC const char *timeout_vendor(void);
+
+TIMEOUT_PUBLIC int timeout_v_rel(void);
+
+TIMEOUT_PUBLIC int timeout_v_abi(void);
+
+TIMEOUT_PUBLIC int timeout_v_api(void);
+
+
+/*
+ * I N T E G E R  T Y P E  I N T E R F A C E S
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+#define TIMEOUT_C(n) UINT64_C(n)
+#define TIMEOUT_PRIu PRIu64
+#define TIMEOUT_PRIx PRIx64
+#define TIMEOUT_PRIX PRIX64
+
+#define TIMEOUT_mHZ TIMEOUT_C(1000)
+#define TIMEOUT_uHZ TIMEOUT_C(1000000)
+#define TIMEOUT_nHZ TIMEOUT_C(1000000000)
+
+typedef uint64_t timeout_t;
+
+#define timeout_error_t int /* for documentation purposes */
+
+
+/*
+ * C A L L B A C K  I N T E R F A C E
+ *
+ * Callback function parameters unspecified to make embedding into existing
+ * applications easier.
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+#ifndef TIMEOUT_CB_OVERRIDE
+struct timeout_cb {
+	void (*fn)(void);
+	void *arg;
+}; /* struct timeout_cb */
+#endif
+
+/*
+ * T I M E O U T  I N T E R F A C E S
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+#ifndef TIMEOUT_DISABLE_INTERVALS
+#define TIMEOUT_INT 0x01 /* interval (repeating) timeout */
+#endif
+#define TIMEOUT_ABS 0x02 /* treat timeout values as absolute */
+
+#define TIMEOUT_INITIALIZER(flags) { (flags) }
+
+#define timeout_setcb(to, fn, arg) do { \
+	(to)->callback.fn = (fn);       \
+	(to)->callback.arg = (arg);     \
+} while (0)
+
+struct timeout {
+	int flags;
+
+	timeout_t expires;
+	/* absolute expiration time */
+
+	struct timeout_list *pending;
+	/* timeout list if pending on wheel or expiry queue */
+
+	TAILQ_ENTRY(timeout) tqe;
+	/* entry member for struct timeout_list lists */
+
+#ifndef TIMEOUT_DISABLE_CALLBACKS
+	struct timeout_cb callback;
+	/* optional callback information */
+#endif
+
+#ifndef TIMEOUT_DISABLE_INTERVALS
+	timeout_t interval;
+	/* timeout interval if periodic */
+#endif
+
+#ifndef TIMEOUT_DISABLE_RELATIVE_ACCESS
+	struct timeouts *timeouts;
+	/* timeouts collection if member of */
+#endif
+}; /* struct timeout */
+
+
+TIMEOUT_PUBLIC struct timeout *timeout_init(struct timeout *, int);
+/* initialize timeout structure (same as TIMEOUT_INITIALIZER) */
+
+#ifndef TIMEOUT_DISABLE_RELATIVE_ACCESS
+TIMEOUT_PUBLIC bool timeout_pending(struct timeout *);
+/* true if on timing wheel, false otherwise */
+ 
+TIMEOUT_PUBLIC bool timeout_expired(struct timeout *);
+/* true if on expired queue, false otherwise */
+
+TIMEOUT_PUBLIC void timeout_del(struct timeout *);
+/* remove timeout from any timing wheel (okay if not member of any) */
+#endif
+
+/*
+ * T I M I N G  W H E E L  I N T E R F A C E S
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+struct timeouts;
+
+TIMEOUT_PUBLIC struct timeouts *timeouts_open(timeout_t, timeout_error_t *);
+/* open a new timing wheel, setting optional HZ (for float conversions) */
+
+TIMEOUT_PUBLIC void timeouts_close(struct timeouts *);
+/* destroy timing wheel */
+
+TIMEOUT_PUBLIC timeout_t timeouts_hz(struct timeouts *);
+/* return HZ setting (for float conversions) */
+
+TIMEOUT_PUBLIC void timeouts_update(struct timeouts *, timeout_t);
+/* update timing wheel with current absolute time */
+
+TIMEOUT_PUBLIC void timeouts_step(struct timeouts *, timeout_t);
+/* step timing wheel by relative time */
+
+TIMEOUT_PUBLIC timeout_t timeouts_get_curtime(struct timeouts *);
+/* Return the current tick. */
+
+TIMEOUT_PUBLIC timeout_t timeouts_timeout(struct timeouts *);
+/* return interval to next required update */
+
+TIMEOUT_PUBLIC void timeouts_add(struct timeouts *, struct timeout *, timeout_t);
+/* add timeout to timing wheel */
+
+TIMEOUT_PUBLIC void timeouts_del(struct timeouts *, struct timeout *);
+/* remove timeout from any timing wheel or expired queue (okay if on neither) */
+
+TIMEOUT_PUBLIC struct timeout *timeouts_get(struct timeouts *);
+/* return any expired timeout (caller should loop until NULL-return) */
+
+TIMEOUT_PUBLIC bool timeouts_pending(struct timeouts *);
+/* return true if any timeouts pending on timing wheel */
+
+TIMEOUT_PUBLIC bool timeouts_expired(struct timeouts *);
+/* return true if any timeouts on expired queue */
+
+TIMEOUT_PUBLIC bool timeouts_check(struct timeouts *, FILE *);
+/* return true if invariants hold. describes failures to optional file handle. */
+
+#define TIMEOUTS_PENDING 0x10
+#define TIMEOUTS_EXPIRED 0x20
+#define TIMEOUTS_ALL     (TIMEOUTS_PENDING|TIMEOUTS_EXPIRED)
+#define TIMEOUTS_CLEAR   0x40
+
+#define TIMEOUTS_IT_INITIALIZER(flags) { (flags), 0, 0, 0, 0 }
+
+#define TIMEOUTS_IT_INIT(cur, _flags) do {                              \
+	(cur)->flags = (_flags);                                        \
+	(cur)->pc = 0;                                                  \
+} while (0)
+
+struct timeouts_it {
+	int flags;
+	unsigned pc, i, j;
+	struct timeout *to;
+}; /* struct timeouts_it */
+
+TIMEOUT_PUBLIC struct timeout *timeouts_next(struct timeouts *, struct timeouts_it *);
+/* return next timeout in pending wheel or expired queue. caller can delete
+ * the returned timeout, but should not otherwise manipulate the timing
+ * wheel. in particular, caller SHOULD NOT delete any other timeout as that
+ * could invalidate cursor state and trigger a use-after-free.
+ */
+
+#define TIMEOUTS_FOREACH(var, T, flags)                                 \
+	struct timeouts_it _it = TIMEOUTS_IT_INITIALIZER((flags));      \
+	while (((var) = timeouts_next((T), &_it)))
+
+
+/*
+ * B O N U S  W H E E L  I N T E R F A C E S
+ *
+ * I usually use floating point timeouts in all my code, but it's cleaner to
+ * separate it to keep the core algorithmic code simple.
+ *
+ * Using macros instead of static inline routines where <math.h> routines
+ * might be used to keep -lm linking optional.
+ *
+ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+#include <math.h> /* ceil(3) */
+
+#define timeouts_f2i(T, f) \
+	((timeout_t)ceil((f) * timeouts_hz((T)))) /* prefer late expiration over early */
+
+#define timeouts_i2f(T, i) \
+	((double)(i) / timeouts_hz((T)))
+
+#define timeouts_addf(T, to, timeout) \
+	timeouts_add((T), (to), timeouts_f2i((T), (timeout)))
+
+#endif /* TIMEOUT_H */

src/test/include.am

@@ -17,6 +17,7 @@ endif
 
 TESTS += src/test/test src/test/test-slow src/test/test-memwipe \
 	src/test/test_workqueue src/test/test_keygen.sh \
+	src/test/test-timers \
 	$(TESTSCRIPTS)
 
 # These flavors are run using automake's test-driver and test-network.sh
@@ -40,7 +41,8 @@ noinst_PROGRAMS+= \
 	src/test/test-memwipe \
 	src/test/test-child \
 	src/test/test_workqueue \
-	src/test/test-switch-id
+	src/test/test-switch-id \
+	src/test/test-timers
 endif
 
 src_test_AM_CPPFLAGS = -DSHARE_DATADIR="\"$(datadir)\"" \
@@ -128,6 +130,8 @@ src_test_test_slow_SOURCES = \
 src_test_test_memwipe_SOURCES = \
 	src/test/test-memwipe.c
 
+src_test_test_timers_SOURCES = \
+	src/test/test-timers.c
 
 src_test_test_CFLAGS = $(AM_CFLAGS) $(TEST_CFLAGS)
 
@@ -191,6 +195,16 @@ src_test_test_workqueue_LDADD = src/or/libtor-testing.a \
 	@TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ @TOR_LIBEVENT_LIBS@ \
 	@TOR_OPENSSL_LIBS@ @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@
 
+src_test_test_timers_CPPFLAGS = $(src_test_test_CPPFLAGS)
+src_test_test_timers_CFLAGS = $(src_test_test_CFLAGS)
+src_test_test_timers_LDADD = \
+	src/common/libor-event-testing.a \
+	src/common/libor-crypto-testing.a $(LIBKECCAK_TINY) $(LIBDONNA) \
+	src/common/libor-testing.a \
+	@TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ @TOR_LIBEVENT_LIBS@ \
+	@TOR_OPENSSL_LIBS@ @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@
+src_test_test_timers_LDFLAGS = $(src_test_test_LDFLAGS)
+
 noinst_HEADERS+= \
 	src/test/fakechans.h \
 	src/test/log_test_helpers.h \

src/test/test-timers.c

@@ -0,0 +1,133 @@
+/* Copyright 2016, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+#include "orconfig.h"
+
+#include <math.h>
+#include <stdio.h>
+#include <string.h>
+
+#ifdef HAVE_EVENT2_EVENT_H
+#include <event2/event.h>
+#else
+#include <event.h>
+#endif
+
+#include "compat.h"
+#include "compat_libevent.h"
+#include "crypto.h"
+#include "timers.h"
+#include "util.h"
+
+#define N_TIMERS 1000
+#define MAX_DURATION 30
+#define N_DISABLE 5
+
+static struct timeval fire_at[N_TIMERS] = {{0,0}};
+static int is_disabled[N_TIMERS] = {0};
+static int fired[N_TIMERS] = {0};
+static struct timeval difference[N_TIMERS] = {{0,0}};
+static tor_timer_t *timers[N_TIMERS] = {NULL};
+
+static int n_active_timers = 0;
+static int n_fired = 0;
+
+static void
+timer_cb(tor_timer_t *t, void *arg, const struct timeval *now)
+{
+  tor_timer_t **t_ptr = arg;
+  tor_assert(*t_ptr == t);
+  int idx = (int) (t_ptr - timers);
+  ++fired[idx];
+  timersub(now, &fire_at[idx], &difference[idx]);
+  ++n_fired;
+  // printf("%d / %d\n",n_fired, N_TIMERS);
+  if (n_fired == n_active_timers) {
+    event_base_loopbreak(tor_libevent_get_base());
+  }
+}
+
+int
+main(int argc, char **argv)
+{
+  (void)argc;
+  (void)argv;
+  tor_libevent_cfg cfg;
+  memset(&cfg, 0, sizeof(cfg));
+  tor_libevent_initialize(&cfg);
+  timers_initialize();
+
+  int i;
+  int ret;
+  struct timeval now;
+  tor_gettimeofday(&now);
+  for (i = 0; i < N_TIMERS; ++i) {
+    struct timeval delay;
+    delay.tv_sec = crypto_rand_int_range(0,MAX_DURATION);
+    delay.tv_usec = crypto_rand_int_range(0,1000000);
+    timeradd(&now, &delay, &fire_at[i]);
+    timers[i] = timer_new(timer_cb, &timers[i]);
+    timer_schedule(timers[i], &delay);
+    ++n_active_timers;
+  }
+
+  /* Disable some; we'll make sure they don't trigger. */
+  for (i = 0; i < N_DISABLE; ++i) {
+    int idx = crypto_rand_int_range(0, N_TIMERS);
+    if (is_disabled[idx])
+      continue;
+    is_disabled[idx] = 1;
+    timer_disable(timers[idx]);
+    --n_active_timers;
+  }
+
+  event_base_loop(tor_libevent_get_base(), 0);
+
+  int64_t total_difference = 0;
+  uint64_t total_square_difference = 0;
+  tor_assert(n_fired == n_active_timers);
+  for (i = 0; i < N_TIMERS; ++i) {
+    if (is_disabled[i]) {
+      tor_assert(fired[i] == 0);
+      continue;
+    }
+    tor_assert(fired[i] == 1);
+    int64_t diff = difference[i].tv_usec + difference[i].tv_sec * 1000000;
+    total_difference += diff;
+    total_square_difference += diff*diff;
+  }
+  const int64_t mean_diff = total_difference / n_active_timers;
+  printf("mean difference: "U64_FORMAT" usec\n",
+         U64_PRINTF_ARG(mean_diff));
+
+  const double mean_sq = ((double)total_square_difference)/ n_active_timers;
+  const double sq_mean = mean_diff * mean_diff;
+  const double stddev = sqrt(mean_sq - sq_mean);
+  printf("standard deviation: %lf usec\n", stddev);
+
+#define MAX_DIFF_USEC (500*1000)
+#define MAX_STDDEV_USEC (500*1000)
+#define ODD_DIFF_USEC (2000)
+#define ODD_STDDEV_USEC (2000)
+
+  if (mean_diff < 0 || mean_diff > MAX_DIFF_USEC || stddev > MAX_STDDEV_USEC) {
+    printf("Either your system is under ridiculous load, or the "
+           "timer backend is broken.\n");
+    ret = 1;
+  } else if (mean_diff > ODD_DIFF_USEC || stddev > ODD_STDDEV_USEC) {
+    printf("Either your system is a bit slow or the "
+           "timer backend is odd.\n");
+    ret = 0;
+  } else {
+    printf("Looks good enough.\n");
+    ret = 0;
+  }
+
+  timer_free(NULL);
+
+  for (i = 0; i < N_TIMERS; ++i) {
+    timer_free(timers[i]);
+  }
+  timers_shutdown();
+  return ret;
+}