#include <linux/errno.h>
#include <linux/init.h>
#include <linux/kernel.h>
+#include <linux/workqueue.h>
#include <linux/mm.h>
#include <linux/module.h>
+#include <linux/slab.h>
#include <linux/netdevice.h>
-#include <linux/sched.h>
#include <linux/skbuff.h>
#include <linux/string.h>
#include <linux/types.h>
+#include <net/net_namespace.h>
+#include <linux/sched.h>
#include <net/dst.h>
-/* Locking strategy:
- * 1) Garbage collection state of dead destination cache
- * entries is protected by dst_lock.
- * 2) GC is run only from BH context, and is the only remover
- * of entries.
- * 3) Entries are added to the garbage list from both BH
- * and non-BH context, so local BH disabling is needed.
- * 4) All operations modify state, so a spinlock is used.
+/*
+ * Theory of operations:
+ * 1) We use a list, protected by a spinlock, to add
+ * new entries from both BH and non-BH context.
+ * 2) In order to keep spinlock held for a small delay,
+ * we use a second list where are stored long lived
+ * entries, that are handled by the garbage collect thread
+ * fired by a workqueue.
+ * 3) This list is guarded by a mutex,
+ * so that the gc_task and dst_dev_event() can be synchronized.
*/
-static struct dst_entry *dst_garbage_list;
-#if RT_CACHE_DEBUG >= 2
+#if RT_CACHE_DEBUG >= 2
static atomic_t dst_total = ATOMIC_INIT(0);
#endif
-static DEFINE_SPINLOCK(dst_lock);
-static unsigned long dst_gc_timer_expires;
-static unsigned long dst_gc_timer_inc = DST_GC_MAX;
-static void dst_run_gc(unsigned long);
-static void ___dst_free(struct dst_entry * dst);
+/*
+ * We want to keep lock & list close together
+ * to dirty as few cache lines as possible in __dst_free().
+ * As this is not a very strong hint, we dont force an alignment on SMP.
+ */
+static struct {
+ spinlock_t lock;
+ struct dst_entry *list;
+ unsigned long timer_inc;
+ unsigned long timer_expires;
+} dst_garbage = {
+ .lock = __SPIN_LOCK_UNLOCKED(dst_garbage.lock),
+ .timer_inc = DST_GC_MAX,
+};
+static void dst_gc_task(struct work_struct *work);
+static void ___dst_free(struct dst_entry *dst);
+
+static DECLARE_DELAYED_WORK(dst_gc_work, dst_gc_task);
-static DEFINE_TIMER(dst_gc_timer, dst_run_gc, DST_GC_MIN, 0);
+static DEFINE_MUTEX(dst_gc_mutex);
+/*
+ * long lived entries are maintained in this list, guarded by dst_gc_mutex
+ */
+static struct dst_entry *dst_busy_list;
-static void dst_run_gc(unsigned long dummy)
+static void dst_gc_task(struct work_struct *work)
{
int delayed = 0;
- int work_performed;
- struct dst_entry * dst, **dstp;
-
- if (!spin_trylock(&dst_lock)) {
- mod_timer(&dst_gc_timer, jiffies + HZ/10);
- return;
- }
+ int work_performed = 0;
+ unsigned long expires = ~0L;
+ struct dst_entry *dst, *next, head;
+ struct dst_entry *last = &head;
+#if RT_CACHE_DEBUG >= 2
+ ktime_t time_start = ktime_get();
+ struct timespec elapsed;
+#endif
- del_timer(&dst_gc_timer);
- dstp = &dst_garbage_list;
- work_performed = 0;
- while ((dst = *dstp) != NULL) {
- if (atomic_read(&dst->__refcnt)) {
- dstp = &dst->next;
+ mutex_lock(&dst_gc_mutex);
+ next = dst_busy_list;
+
+loop:
+ while ((dst = next) != NULL) {
+ next = dst->next;
+ prefetch(&next->next);
+ cond_resched();
+ if (likely(atomic_read(&dst->__refcnt))) {
+ last->next = dst;
+ last = dst;
delayed++;
continue;
}
- *dstp = dst->next;
- work_performed = 1;
+ work_performed++;
dst = dst_destroy(dst);
if (dst) {
continue;
___dst_free(dst);
- dst->next = *dstp;
- *dstp = dst;
- dstp = &dst->next;
+ dst->next = next;
+ next = dst;
}
}
- if (!dst_garbage_list) {
- dst_gc_timer_inc = DST_GC_MAX;
- goto out;
+
+ spin_lock_bh(&dst_garbage.lock);
+ next = dst_garbage.list;
+ if (next) {
+ dst_garbage.list = NULL;
+ spin_unlock_bh(&dst_garbage.lock);
+ goto loop;
}
- if (!work_performed) {
- if ((dst_gc_timer_expires += dst_gc_timer_inc) > DST_GC_MAX)
- dst_gc_timer_expires = DST_GC_MAX;
- dst_gc_timer_inc += DST_GC_INC;
- } else {
- dst_gc_timer_inc = DST_GC_INC;
- dst_gc_timer_expires = DST_GC_MIN;
+ last->next = NULL;
+ dst_busy_list = head.next;
+ if (!dst_busy_list)
+ dst_garbage.timer_inc = DST_GC_MAX;
+ else {
+ /*
+ * if we freed less than 1/10 of delayed entries,
+ * we can sleep longer.
+ */
+ if (work_performed <= delayed/10) {
+ dst_garbage.timer_expires += dst_garbage.timer_inc;
+ if (dst_garbage.timer_expires > DST_GC_MAX)
+ dst_garbage.timer_expires = DST_GC_MAX;
+ dst_garbage.timer_inc += DST_GC_INC;
+ } else {
+ dst_garbage.timer_inc = DST_GC_INC;
+ dst_garbage.timer_expires = DST_GC_MIN;
+ }
+ expires = dst_garbage.timer_expires;
+ /*
+ * if the next desired timer is more than 4 seconds in the
+ * future then round the timer to whole seconds
+ */
+ if (expires > 4*HZ)
+ expires = round_jiffies_relative(expires);
+ schedule_delayed_work(&dst_gc_work, expires);
}
+
+ spin_unlock_bh(&dst_garbage.lock);
+ mutex_unlock(&dst_gc_mutex);
#if RT_CACHE_DEBUG >= 2
- printk("dst_total: %d/%d %ld\n",
- atomic_read(&dst_total), delayed, dst_gc_timer_expires);
+ elapsed = ktime_to_timespec(ktime_sub(ktime_get(), time_start));
+ printk(KERN_DEBUG "dst_total: %d delayed: %d work_perf: %d"
+ " expires: %lu elapsed: %lu us\n",
+ atomic_read(&dst_total), delayed, work_performed,
+ expires,
+ elapsed.tv_sec * USEC_PER_SEC +
+ elapsed.tv_nsec / NSEC_PER_USEC);
#endif
- mod_timer(&dst_gc_timer, jiffies + dst_gc_timer_expires);
-
-out:
- spin_unlock(&dst_lock);
}
-static int dst_discard_in(struct sk_buff *skb)
+int dst_discard(struct sk_buff *skb)
{
kfree_skb(skb);
return 0;
}
+EXPORT_SYMBOL(dst_discard);
-static int dst_discard_out(struct sk_buff *skb)
+void *dst_alloc(struct dst_ops *ops)
{
- kfree_skb(skb);
- return 0;
-}
-
-void * dst_alloc(struct dst_ops * ops)
-{
- struct dst_entry * dst;
+ struct dst_entry *dst;
if (ops->gc && atomic_read(&ops->entries) > ops->gc_thresh) {
- if (ops->gc())
+ if (ops->gc(ops))
return NULL;
}
- dst = kmem_cache_alloc(ops->kmem_cachep, GFP_ATOMIC);
+ dst = kmem_cache_zalloc(ops->kmem_cachep, GFP_ATOMIC);
if (!dst)
return NULL;
- memset(dst, 0, ops->entry_size);
atomic_set(&dst->__refcnt, 0);
dst->ops = ops;
dst->lastuse = jiffies;
dst->path = dst;
- dst->input = dst_discard_in;
- dst->output = dst_discard_out;
-#if RT_CACHE_DEBUG >= 2
+ dst->input = dst->output = dst_discard;
+#if RT_CACHE_DEBUG >= 2
atomic_inc(&dst_total);
#endif
atomic_inc(&ops->entries);
return dst;
}
+EXPORT_SYMBOL(dst_alloc);
-static void ___dst_free(struct dst_entry * dst)
+static void ___dst_free(struct dst_entry *dst)
{
/* The first case (dev==NULL) is required, when
protocol module is unloaded.
*/
- if (dst->dev == NULL || !(dst->dev->flags&IFF_UP)) {
- dst->input = dst_discard_in;
- dst->output = dst_discard_out;
- }
+ if (dst->dev == NULL || !(dst->dev->flags&IFF_UP))
+ dst->input = dst->output = dst_discard;
dst->obsolete = 2;
}
+EXPORT_SYMBOL(__dst_free);
-void __dst_free(struct dst_entry * dst)
+void __dst_free(struct dst_entry *dst)
{
- spin_lock_bh(&dst_lock);
+ spin_lock_bh(&dst_garbage.lock);
___dst_free(dst);
- dst->next = dst_garbage_list;
- dst_garbage_list = dst;
- if (dst_gc_timer_inc > DST_GC_INC) {
- dst_gc_timer_inc = DST_GC_INC;
- dst_gc_timer_expires = DST_GC_MIN;
- mod_timer(&dst_gc_timer, jiffies + dst_gc_timer_expires);
+ dst->next = dst_garbage.list;
+ dst_garbage.list = dst;
+ if (dst_garbage.timer_inc > DST_GC_INC) {
+ dst_garbage.timer_inc = DST_GC_INC;
+ dst_garbage.timer_expires = DST_GC_MIN;
+ cancel_delayed_work(&dst_gc_work);
+ schedule_delayed_work(&dst_gc_work, dst_garbage.timer_expires);
}
- spin_unlock_bh(&dst_lock);
+ spin_unlock_bh(&dst_garbage.lock);
}
struct dst_entry *dst_destroy(struct dst_entry * dst)
dst->ops->destroy(dst);
if (dst->dev)
dev_put(dst->dev);
-#if RT_CACHE_DEBUG >= 2
+#if RT_CACHE_DEBUG >= 2
atomic_dec(&dst_total);
#endif
kmem_cache_free(dst->ops->kmem_cachep, dst);
}
return NULL;
}
+EXPORT_SYMBOL(dst_destroy);
+
+void dst_release(struct dst_entry *dst)
+{
+ if (dst) {
+ int newrefcnt;
+
+ smp_mb__before_atomic_dec();
+ newrefcnt = atomic_dec_return(&dst->__refcnt);
+ WARN_ON(newrefcnt < 0);
+ }
+}
+EXPORT_SYMBOL(dst_release);
/* Dirty hack. We did it in 2.2 (in __dst_free),
* we have _very_ good reasons not to repeat
*
* Commented and originally written by Alexey.
*/
-static inline void dst_ifdown(struct dst_entry *dst, struct net_device *dev,
- int unregister)
+static void dst_ifdown(struct dst_entry *dst, struct net_device *dev,
+ int unregister)
{
if (dst->ops->ifdown)
dst->ops->ifdown(dst, dev, unregister);
return;
if (!unregister) {
- dst->input = dst_discard_in;
- dst->output = dst_discard_out;
+ dst->input = dst->output = dst_discard;
} else {
- dst->dev = &loopback_dev;
- dev_hold(&loopback_dev);
+ dst->dev = dev_net(dst->dev)->loopback_dev;
+ dev_hold(dst->dev);
dev_put(dev);
if (dst->neighbour && dst->neighbour->dev == dev) {
- dst->neighbour->dev = &loopback_dev;
+ dst->neighbour->dev = dst->dev;
+ dev_hold(dst->dev);
dev_put(dev);
- dev_hold(&loopback_dev);
}
}
}
-static int dst_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
+static int dst_dev_event(struct notifier_block *this, unsigned long event,
+ void *ptr)
{
struct net_device *dev = ptr;
- struct dst_entry *dst;
+ struct dst_entry *dst, *last = NULL;
switch (event) {
case NETDEV_UNREGISTER:
case NETDEV_DOWN:
- spin_lock_bh(&dst_lock);
- for (dst = dst_garbage_list; dst; dst = dst->next) {
+ mutex_lock(&dst_gc_mutex);
+ for (dst = dst_busy_list; dst; dst = dst->next) {
+ last = dst;
dst_ifdown(dst, dev, event != NETDEV_DOWN);
}
- spin_unlock_bh(&dst_lock);
+
+ spin_lock_bh(&dst_garbage.lock);
+ dst = dst_garbage.list;
+ dst_garbage.list = NULL;
+ spin_unlock_bh(&dst_garbage.lock);
+
+ if (last)
+ last->next = dst;
+ else
+ dst_busy_list = dst;
+ for (; dst; dst = dst->next)
+ dst_ifdown(dst, dev, event != NETDEV_DOWN);
+ mutex_unlock(&dst_gc_mutex);
break;
}
return NOTIFY_DONE;
{
register_netdevice_notifier(&dst_dev_notifier);
}
-
-EXPORT_SYMBOL(__dst_free);
-EXPORT_SYMBOL(dst_alloc);
-EXPORT_SYMBOL(dst_destroy);