timeout_ext 0.0.0

data/ext/timeout_ext/ccan/timer/_info

@@ -0,0 +1,79 @@
+#include "config.h"
+#include <stdio.h>
+#include <string.h>
+
+/**
+ * timer - efficient implementation of rarely-expiring timers.
+ *
+ * This is a lazy implementation of timers: you can add and delete timers
+ * very quickly, and they are only sorted as their expiry approaches.
+ *
+ * This is a common case for timeouts, which must often be set, but
+ * rarely expire.
+ *
+ * Example:
+ *	// Silly example which outputs strings until timers expire.
+ *	#include <ccan/timer/timer.h>
+ *	#include <ccan/time/time.h>
+ *	#include <stdlib.h>
+ *	#include <stdio.h>
+ *
+ *	struct timed_string {
+ *		struct list_node node;
+ *		struct timer timer;
+ *		const char *string;
+ *	};
+ *
+ *	int main(int argc, char *argv[])
+ *	{
+ *		struct timers timers;
+ *		struct list_head strings;
+ *		struct timer *t;
+ *		struct timed_string *s;
+ *
+ *		(void)argc;
+ *		timers_init(&timers, time_mono());
+ *		list_head_init(&strings);
+ *
+ *		while (argv[1]) {
+ *			s = malloc(sizeof(*s));
+ *			s->string = argv[1];
+ *			timer_addrel(&timers, &s->timer,
+ *				     time_from_msec(atol(argv[2])));
+ *			list_add_tail(&strings, &s->node);
+ *			argv += 2;
+ *		}
+ *
+ *		while (!list_empty(&strings)) {
+ *			struct timemono now = time_mono();
+ *			list_for_each(&strings, s, node)
+ *				printf("%s", s->string);
+ *			while ((t = timers_expire(&timers, now)) != NULL) {
+ *				s = container_of(t, struct timed_string, timer);
+ *				list_del_from(&strings, &s->node);
+ *				free(s);
+ *			}
+ *		}
+ *
+ *		exit(0);
+ *	}
+ *
+ * License: LGPL (v2.1 or any later version)
+ * Author: Rusty Russell <rusty@rustcorp.com.au>
+ */
+int main(int argc, char *argv[])
+{
+	/* Expect exactly one argument */
+	if (argc != 2)
+		return 1;
+
+	if (strcmp(argv[1], "depends") == 0) {
+		printf("ccan/array_size\n");
+		printf("ccan/ilog\n");
+		printf("ccan/list\n");
+		printf("ccan/time\n");
+		return 0;
+	}
+
+	return 1;
+}

data/ext/timeout_ext/ccan/timer/design.txt

@@ -0,0 +1,76 @@
+Cascading timer design.
+
+Inspired by the Linux kernel approach, documented roughly at:
+	https://lwn.net/Articles/152436/
+
+For easy description, we use whole seconds and powers of 10: in the
+implementation, we use powers of 2 (eg. 256 entries) and smaller
+granularities.
+
+We start with a simple data structure:
+
+struct timer_level {
+	struct timer_level *next;
+
+	/* Ten buckets: 	[0] [1] [2] [3] [4] [5] [6] [7] [8] [9] */
+	struct list_head bucket[10];
+};
+
+struct timers {
+	/* We can never have a timer before this, aka "now". */
+	time_t offset;
+
+	struct timer_level *level;
+
+	/* Anything too far in the future. */
+	struct list_head far;
+}
+
+The first level of timers holds anything which will happen in the next
+10 seconds.  The next level holds things which will happen in the next
+100 seconds.  And so on.
+
+When we want to add a new timer into the structure, we need to figure
+out first what level it goes into, and second, which bucket.  Say our
+offset is 500,000,001 (about Tue Nov 5, 1985 in Unix time).  And our
+timer is set to go off in 5 seconds, ie. 500,000,006.
+
+The level is easy: the difference between the timer and the offset is
+5, and that's less than 10, so it's in the first level.  The position,
+however, depends on the absolute time, in this case the last digit 6,
+so it's in bucket 6.
+
+Adding a timer at 500,000,123?  The difference is > 100 and < 1000, so
+it's in the third level.  The bucket is 1.  If there's no third level,
+we just add it to the 'far' list for stuff which is in the far future.
+
+Deleting a timer is as simple as removing it; there is no external
+bookkeeping in this scheme.  This matters, since timers used for
+timeouts are almost always deleted before they expire.
+
+Now, when a second passes, we need to know if there are any timers
+which are due.  We increment the offset to 500,000,002, and look in
+the first level, bucket 2 for any timers, so lookup is simple.
+
+We do this eight more times, and we increment the offset to
+500,000,010.  We've swept around back to bucket 0, though it may not
+be empty if we added more timers as we were going.
+
+But we need to look into the next level since a timer at 500,000,010
+added when the offset was 500,000,000 would have gone up there.  We
+empty bucket 1 (due to the '1' in 500,000,010) into these buckets,
+which will contain timers between 500,000,010 and 500,000,019, which
+all now are less than 10 seconds away, so belong in the bottom level.
+
+Similarly, at 500,000,020 we will empty bucket 1 of the second level
+into the first level.  And at 500,000,100 we will empty bucket 1 of
+the third level into the second level then bucket 0 of the second
+level into the first level.  We do it in this order, since emptying
+bucket 1 on the third level (500,000,100 - 500,000,199) may put more
+entries (500,000,100 - 500,000,109) into bucket 0 on the second level.
+
+When we get to 500,001,000 we should empty the fourth level.  If there
+is no fourth level, that's when we sort through the 'far' list and
+empty any which are less than 500,002,000.  If there are many entries
+in the far list, we should add more levels to reduce the number, or at
+least the frequency we have to check it.

data/ext/timeout_ext/ccan/timer/timer.c

@@ -0,0 +1,524 @@
+/* LGPL (v2.1 or any later version) - see LICENSE file for details */
+#include <ccan/timer/timer.h>
+#include <ccan/array_size/array_size.h>
+#include <ccan/ilog/ilog.h>
+#include <stdlib.h>
+#include <stdio.h>
+
+#define PER_LEVEL (1ULL << TIMER_LEVEL_BITS)
+
+struct timer_level {
+	struct list_head list[PER_LEVEL];
+};
+
+static uint64_t time_to_grains(struct timemono t)
+{
+	return t.ts.tv_sec * ((uint64_t)1000000000 / TIMER_GRANULARITY)
+		+ (t.ts.tv_nsec / TIMER_GRANULARITY);
+}
+
+static struct timemono grains_to_time(uint64_t grains)
+{
+	struct timemono t;
+
+	t.ts.tv_sec = grains / (1000000000 / TIMER_GRANULARITY);
+	t.ts.tv_nsec = (grains % (1000000000 / TIMER_GRANULARITY))
+		* TIMER_GRANULARITY;
+	return t;
+}
+
+void timers_init(struct timers *timers, struct timemono start)
+{
+	unsigned int i;
+
+	list_head_init(&timers->far);
+	timers->base = time_to_grains(start);
+	timers->first = -1ULL;
+	memset(timers->firsts, 0xFF, sizeof(timers->firsts));
+	for (i = 0; i < ARRAY_SIZE(timers->level); i++)
+		timers->level[i] = NULL;
+}
+
+static unsigned int level_of(const struct timers *timers, uint64_t time)
+{
+	uint64_t diff;
+
+	/* Level depends how far away it is. */
+	diff = time - timers->base;
+	return ilog64(diff / 2) / TIMER_LEVEL_BITS;
+}
+
+static void timer_add_raw(struct timers *timers, struct timer *t)
+{
+	struct list_head *l;
+	unsigned int level = level_of(timers, t->time);
+	uint64_t *first;
+
+	if (!timers->level[level]) {
+		l = &timers->far;
+		first = &timers->firsts[ARRAY_SIZE(timers->level)];
+	} else {
+		int off = (t->time >> (level*TIMER_LEVEL_BITS)) & (PER_LEVEL-1);
+		l = &timers->level[level]->list[off];
+		first = &timers->firsts[level];
+	}
+
+	list_add_tail(l, &t->list);
+	if (t->time < *first)
+		*first = t->time;
+}
+
+void timer_init(struct timer *t)
+{
+	list_node_init(&t->list);
+}
+
+static bool list_node_initted(const struct list_node *n)
+{
+	return n->prev == n;
+}
+
+void timer_addrel(struct timers *timers, struct timer *t, struct timerel rel)
+{
+	assert(list_node_initted(&t->list));
+
+	t->time = time_to_grains(timemono_add(time_mono(), rel));
+
+#if TIME_HAVE_MONOTONIC
+	assert(t->time >= timers->base);
+#else
+	/* Added in the past?  Treat it as imminent. */
+	if (t->time < timers->base)
+		t->time = timers->base;
+#endif
+	if (t->time < timers->first)
+		timers->first = t->time;
+
+	timer_add_raw(timers, t);
+}
+
+void timer_addmono(struct timers *timers, struct timer *t, struct timemono when)
+{
+	assert(list_node_initted(&t->list));
+
+	t->time = time_to_grains(when);
+
+	/* Added in the past?  Treat it as imminent. */
+	if (t->time < timers->base)
+		t->time = timers->base;
+	if (t->time < timers->first)
+		timers->first = t->time;
+
+	timer_add_raw(timers, t);
+}
+
+/* FIXME: inline */
+void timer_del(struct timers *timers UNNEEDED, struct timer *t)
+{
+	list_del_init(&t->list);
+}
+
+static void timers_far_get(struct timers *timers,
+			   struct list_head *list,
+			   uint64_t when)
+{
+	struct timer *i, *next;
+
+	list_for_each_safe(&timers->far, i, next, list) {
+		if (i->time <= when) {
+			list_del_from(&timers->far, &i->list);
+			list_add_tail(list, &i->list);
+		}
+	}
+}
+
+static void add_level(struct timers *timers, unsigned int level)
+{
+	struct timer_level *l;
+	struct timer *t;
+	unsigned int i;
+	struct list_head from_far;
+
+	l = malloc(sizeof(*l));
+	if (!l)
+		return;
+
+	for (i = 0; i < ARRAY_SIZE(l->list); i++)
+		list_head_init(&l->list[i]);
+	timers->level[level] = l;
+
+	list_head_init(&from_far);
+	timers_far_get(timers, &from_far,
+		       timers->base + (1ULL << ((level+1)*TIMER_LEVEL_BITS)) - 1);
+
+	while ((t = list_pop(&from_far, struct timer, list)) != NULL)
+		timer_add_raw(timers, t);
+}
+
+/* We don't need to search past the first at level 0, since the
+ * bucket range is 1; they're all the same. */
+static const struct timer *find_first(const struct list_head *list,
+				      unsigned int level,
+				      const struct timer *prev)
+{
+	struct timer *t;
+
+	list_for_each(list, t, list) {
+		if (!prev || t->time < prev->time)
+			prev = t;
+		if (level == 0)
+			break;
+	}
+	return prev;
+}
+
+/* Update level's first watermark, and return overall first. */
+static const struct timer *first_for_level(struct timers *timers,
+					   size_t level,
+					   const struct timer *level_first,
+					   const struct timer *first)
+{
+	if (level_first) {
+		timers->firsts[level] = level_first->time;
+		if (!first || level_first->time < first->time)
+			first = level_first;
+	} else {
+		timers->firsts[level] = -1ULL;
+	}
+	return first;
+}
+
+static bool level_may_beat(const struct timers *timers, size_t level,
+			   const struct timer *first)
+{
+	return !first || timers->firsts[level] < first->time;
+}
+
+/* FIXME: Suboptimal */
+static const struct timer *brute_force_first(struct timers *timers)
+{
+	unsigned int l, i;
+	const struct timer *found = NULL;
+
+	for (l = 0; l < ARRAY_SIZE(timers->level) && timers->level[l]; l++) {
+		const struct timer *t = NULL;
+
+		/* Do we know they don't have a better one? */
+		if (!level_may_beat(timers, l, found))
+			continue;
+
+		/* Find first timer on this level. */
+		for (i = 0; i < PER_LEVEL; i++)
+			t = find_first(&timers->level[l]->list[i], l, t);
+
+		found = first_for_level(timers, l, t, found);
+	}
+
+	/* Check (and update) far list if there's a chance. */
+	l = ARRAY_SIZE(timers->level);
+	if (level_may_beat(timers, l, found)) {
+		const struct timer *t = find_first(&timers->far, l, NULL);
+		found = first_for_level(timers, l, t, found);
+	}
+
+	return found;
+}
+
+static const struct timer *get_first(struct timers *timers)
+{
+	/* We can have just far timers, for example. */
+	if (timers->level[0]) {
+		/* First search rest of lower buckets; we've already spilled
+		 * so if we find one there we don't need to search further. */
+		unsigned int i, off = timers->base % PER_LEVEL;
+
+		for (i = off; i < PER_LEVEL; i++) {
+			struct list_head *h = &timers->level[0]->list[i];
+			if (!list_empty(h))
+				return find_first(h, 0, NULL);
+		}
+	}
+
+	/* From here on, we're searching non-normalized parts of the
+	 * data structure, which is much subtler.
+	 *
+	 * So we brute force. */
+	return brute_force_first(timers);
+}
+
+static bool update_first(struct timers *timers)
+{
+	const struct timer *found = get_first(timers);
+
+	if (!found) {
+		timers->first = -1ULL;
+		return false;
+ 	}
+
+	timers->first = found->time;
+	return true;
+}
+
+bool timer_earliest(struct timers *timers, struct timemono *first)
+{
+	if (!update_first(timers))
+		return false;
+
+	*first = grains_to_time(timers->first);
+	return true;
+}
+
+/* Assume no timers before 'time', cascade down and update base time. */
+static void timer_fast_forward(struct timers *timers, uint64_t time)
+{
+	unsigned int level, changed;
+	int need_level = -1;
+	struct list_head list;
+	struct timer *i;
+
+	/* How many bits changed between base and time?
+	 * Each time we wrap, we need to empty buckets from above. */
+	if (time == timers->base)
+		return;
+
+	changed = ilog64_nz(time ^ timers->base);
+	level = (changed - 1) / TIMER_LEVEL_BITS;
+
+	/* Buckets always empty downwards, so we could cascade manually,
+	 * but it's rarely very many so we just remove and re-add */
+	list_head_init(&list);
+
+	do {
+		if (!timers->level[level]) {
+			/* We need any which belong on this level. */
+			timers_far_get(timers, &list,
+				       timers->base
+				       + (1ULL << ((level+1)*TIMER_LEVEL_BITS))-1);
+			need_level = level;
+		} else {
+			unsigned src;
+
+			/* Get all timers from this bucket. */
+			src = (time >> (level * TIMER_LEVEL_BITS)) % PER_LEVEL;
+			list_append_list(&list,
+					 &timers->level[level]->list[src]);
+		}
+	} while (level--);
+
+	/* Did we hit the last level?  If so, add. */
+	if (need_level != -1)
+		add_level(timers, need_level);
+
+	/* Fast-forward the time, and re-add everyone. */
+	timers->base = time;
+	while ((i = list_pop(&list, struct timer, list)) != NULL)
+		timer_add_raw(timers, i);
+}
+
+/* Returns an expired timer. */
+struct timer *timers_expire(struct timers *timers, struct timemono expire)
+{
+	uint64_t now = time_to_grains(expire);
+	unsigned int off;
+	struct timer *t;
+
+	assert(now >= timers->base);
+
+	if (!timers->level[0]) {
+		if (list_empty(&timers->far))
+			return NULL;
+		add_level(timers, 0);
+	}
+
+	do {
+		if (timers->first > now) {
+			timer_fast_forward(timers, now);
+			return NULL;
+		}
+
+		timer_fast_forward(timers, timers->first);
+		off = timers->base % PER_LEVEL;
+
+		/* This *may* be NULL, if we deleted the first timer */
+		t = list_pop(&timers->level[0]->list[off], struct timer, list);
+		if (t)
+			list_node_init(&t->list);
+	} while (!t && update_first(timers));
+
+	return t;
+}
+
+static bool timer_list_check(const struct list_head *l,
+			     uint64_t min, uint64_t max, uint64_t first,
+			     const char *abortstr)
+{
+	const struct timer *t;
+
+	if (!list_check(l, abortstr))
+		return false;
+
+	list_for_each(l, t, list) {
+		if (t->time < min || t->time > max) {
+			if (abortstr) {
+				fprintf(stderr,
+					"%s: timer %p %llu not %llu-%llu\n",
+					abortstr, t, (long long)t->time,
+					(long long)min, (long long)max);
+				abort();
+			}
+			return false;
+		}
+		if (t->time < first) {
+			if (abortstr) {
+				fprintf(stderr,
+					"%s: timer %p %llu < minimum %llu\n",
+					abortstr, t, (long long)t->time,
+					(long long)first);
+				abort();
+			}
+			return false;
+		}
+	}
+	return true;
+}
+
+struct timers *timers_check(const struct timers *timers, const char *abortstr)
+{
+	unsigned int l, i, off;
+	uint64_t base;
+
+	l = 0;
+	if (!timers->level[0])
+		goto past_levels;
+
+	/* First level is simple. */
+	off = timers->base % PER_LEVEL;
+	for (i = 0; i < PER_LEVEL; i++) {
+		struct list_head *h;
+
+		h = &timers->level[l]->list[(i+off) % PER_LEVEL];
+		if (!timer_list_check(h, timers->base + i, timers->base + i,
+				      timers->firsts[l], abortstr))
+			return NULL;
+	}
+
+	/* For other levels, "current" bucket has been emptied, and may contain
+	 * entries for the current + level_size bucket. */
+	for (l = 1; l < ARRAY_SIZE(timers->level) && timers->level[l]; l++) {
+		uint64_t per_bucket = 1ULL << (TIMER_LEVEL_BITS * l);
+
+		off = ((timers->base >> (l*TIMER_LEVEL_BITS)) % PER_LEVEL);
+		/* We start at *next* bucket. */
+		base = (timers->base & ~(per_bucket - 1)) + per_bucket;
+
+		for (i = 1; i <= PER_LEVEL; i++) {
+			struct list_head *h;
+
+			h = &timers->level[l]->list[(i+off) % PER_LEVEL];
+			if (!timer_list_check(h, base, base + per_bucket - 1,
+					      timers->firsts[l], abortstr))
+				return NULL;
+			base += per_bucket;
+		}
+	}
+
+past_levels:
+	base = (timers->base & ~((1ULL << (TIMER_LEVEL_BITS * l)) - 1))
+		+ (1ULL << (TIMER_LEVEL_BITS * l)) - 1;
+	if (!timer_list_check(&timers->far, base, -1ULL,
+			      timers->firsts[ARRAY_SIZE(timers->level)],
+			      abortstr))
+		return NULL;
+
+	return (struct timers *)timers;
+}
+
+#ifdef CCAN_TIMER_DEBUG
+static void dump_bucket_stats(FILE *fp, const struct list_head *h)
+{
+	unsigned long long min, max, num;
+	struct timer *t;
+
+	if (list_empty(h)) {
+		printf("\n");
+		return;
+	}
+
+	min = -1ULL;
+	max = 0;
+	num = 0;
+	list_for_each(h, t, list) {
+		if (t->time < min)
+			min = t->time;
+		if (t->time > max)
+			max = t->time;
+		num++;
+	}
+	fprintf(fp, " %llu (%llu-%llu)\n",
+		num, min, max);
+}
+
+void timers_dump(const struct timers *timers, FILE *fp)
+{
+	unsigned int l, i, off;
+	unsigned long long base;
+
+	if (!fp)
+		fp = stderr;
+
+	fprintf(fp, "Base: %llu\n", (unsigned long long)timers->base);
+
+	if (!timers->level[0])
+		goto past_levels;
+
+	fprintf(fp, "Level 0:\n");
+
+	/* First level is simple. */
+	off = timers->base % PER_LEVEL;
+	for (i = 0; i < PER_LEVEL; i++) {
+		const struct list_head *h;
+
+		fprintf(fp, "  Bucket %llu (%lu):",
+			(i+off) % PER_LEVEL, timers->base + i);
+		h = &timers->level[0]->list[(i+off) % PER_LEVEL];
+		dump_bucket_stats(fp, h);
+	}
+
+	/* For other levels, "current" bucket has been emptied, and may contain
+	 * entries for the current + level_size bucket. */
+	for (l = 1; l < ARRAY_SIZE(timers->level) && timers->level[l]; l++) {
+		uint64_t per_bucket = 1ULL << (TIMER_LEVEL_BITS * l);
+
+		off = ((timers->base >> (l*TIMER_LEVEL_BITS)) % PER_LEVEL);
+		/* We start at *next* bucket. */
+		base = (timers->base & ~(per_bucket - 1)) + per_bucket;
+
+		fprintf(fp, "Level %u:\n", l);
+		for (i = 1; i <= PER_LEVEL; i++) {
+			const struct list_head *h;
+
+			fprintf(fp, "  Bucket %llu (%llu - %llu):",
+				(i+off) % PER_LEVEL,
+				base, base + per_bucket - 1);
+
+			h = &timers->level[l]->list[(i+off) % PER_LEVEL];
+			dump_bucket_stats(fp, h);
+			base += per_bucket;
+		}
+	}
+
+past_levels:
+	if (!list_empty(&timers->far)) {
+		fprintf(fp, "Far timers:");
+		dump_bucket_stats(fp, &timers->far);
+	}
+}
+#endif
+
+void timers_cleanup(struct timers *timers)
+{
+	unsigned int l;
+
+	for (l = 0; l < ARRAY_SIZE(timers->level); l++)
+		free(timers->level[l]);
+}