Motiviation: We incorrectly did ignore NS servers during redirect which had no ADDITIONAL record. This could at worse have the affect that we failed the query completely as none of the NS servers had a ADDITIONAL record. Beside this using a DnsCache to cache authoritative nameservers does not work in practise as we we need different features and semantics when cache these servers (for example we also want to cache unresolved nameservers and resolve these on the fly when needed). Modifications: - Correctly take NS records into account that have no matching ADDITIONAL record - Correctly handle multiple ADDITIONAL records for the same NS record - Introduce AuthoritativeDnsServerCache as a replacement of the DnsCache when caching authoritative nameservers + adding default implementation - Add an adapter layer to reduce API breakage as much as possible - Replace DnsNameResolver.uncachedRedirectDnsServerStream(...) with newRedirectDnsServerStream(...) - Add unit tests Result: Our DnsResolver now correctly handle redirects in all cases.
293 lines
11 KiB
Java
293 lines
11 KiB
Java
/*
|
|
* Copyright 2018 The Netty Project
|
|
*
|
|
* The Netty Project licenses this file to you under the Apache License,
|
|
* version 2.0 (the "License"); you may not use this file except in compliance
|
|
* with the License. You may obtain a copy of the License at:
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
|
|
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
|
|
* License for the specific language governing permissions and limitations
|
|
* under the License.
|
|
*/
|
|
package io.netty.resolver.dns;
|
|
|
|
import io.netty.channel.EventLoop;
|
|
import io.netty.util.internal.PlatformDependent;
|
|
|
|
import java.util.ArrayList;
|
|
import java.util.Collections;
|
|
import java.util.Iterator;
|
|
import java.util.List;
|
|
import java.util.Map;
|
|
import java.util.Map.Entry;
|
|
import java.util.concurrent.ConcurrentMap;
|
|
import java.util.concurrent.Delayed;
|
|
import java.util.concurrent.ScheduledFuture;
|
|
import java.util.concurrent.TimeUnit;
|
|
import java.util.concurrent.atomic.AtomicReference;
|
|
import java.util.concurrent.atomic.AtomicReferenceFieldUpdater;
|
|
|
|
import static java.util.Collections.singletonList;
|
|
|
|
/**
|
|
* Abstract cache that automatically removes entries for a hostname once the TTL for an entry is reached.
|
|
*
|
|
* @param <E>
|
|
*/
|
|
abstract class Cache<E> {
|
|
private static final AtomicReferenceFieldUpdater<Cache.Entries, ScheduledFuture> FUTURE_UPDATER =
|
|
AtomicReferenceFieldUpdater.newUpdater(Cache.Entries.class, ScheduledFuture.class, "expirationFuture");
|
|
|
|
private static final ScheduledFuture<?> CANCELLED = new ScheduledFuture<Object>() {
|
|
|
|
@Override
|
|
public boolean cancel(boolean mayInterruptIfRunning) {
|
|
return false;
|
|
}
|
|
|
|
@Override
|
|
public long getDelay(TimeUnit unit) {
|
|
// We ignore unit and always return the minimum value to ensure the TTL of the cancelled marker is
|
|
// the smallest.
|
|
return Long.MIN_VALUE;
|
|
}
|
|
|
|
@Override
|
|
public int compareTo(Delayed o) {
|
|
throw new UnsupportedOperationException();
|
|
}
|
|
|
|
@Override
|
|
public boolean isCancelled() {
|
|
return true;
|
|
}
|
|
|
|
@Override
|
|
public boolean isDone() {
|
|
return true;
|
|
}
|
|
|
|
@Override
|
|
public Object get() {
|
|
throw new UnsupportedOperationException();
|
|
}
|
|
|
|
@Override
|
|
public Object get(long timeout, TimeUnit unit) {
|
|
throw new UnsupportedOperationException();
|
|
}
|
|
};
|
|
|
|
// Two years are supported by all our EventLoop implementations and so safe to use as maximum.
|
|
// See also: https://github.com/netty/netty/commit/b47fb817991b42ec8808c7d26538f3f2464e1fa6
|
|
static final int MAX_SUPPORTED_TTL_SECS = (int) TimeUnit.DAYS.toSeconds(365 * 2);
|
|
|
|
private final ConcurrentMap<String, Entries> resolveCache = PlatformDependent.newConcurrentHashMap();
|
|
|
|
/**
|
|
* Remove everything from the cache.
|
|
*/
|
|
final void clear() {
|
|
while (!resolveCache.isEmpty()) {
|
|
for (Iterator<Entry<String, Entries>> i = resolveCache.entrySet().iterator(); i.hasNext();) {
|
|
Map.Entry<String, Entries> e = i.next();
|
|
i.remove();
|
|
|
|
e.getValue().clearAndCancel();
|
|
}
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Clear all entries (if anything exists) for the given hostname and return {@code true} if anything was removed.
|
|
*/
|
|
final boolean clear(String hostname) {
|
|
Entries entries = resolveCache.remove(hostname);
|
|
return entries != null && entries.clearAndCancel();
|
|
}
|
|
|
|
/**
|
|
* Returns all caches entries for the given hostname.
|
|
*/
|
|
final List<? extends E> get(String hostname) {
|
|
Entries entries = resolveCache.get(hostname);
|
|
return entries == null ? null : entries.get();
|
|
}
|
|
|
|
/**
|
|
* Cache a value for the given hostname that will automatically expire once the TTL is reached.
|
|
*/
|
|
final void cache(String hostname, E value, int ttl, EventLoop loop) {
|
|
Entries entries = resolveCache.get(hostname);
|
|
if (entries == null) {
|
|
entries = new Entries(hostname);
|
|
Entries oldEntries = resolveCache.putIfAbsent(hostname, entries);
|
|
if (oldEntries != null) {
|
|
entries = oldEntries;
|
|
}
|
|
}
|
|
entries.add(value, ttl, loop);
|
|
}
|
|
|
|
/**
|
|
* Return the number of hostames for which we have cached something.
|
|
*/
|
|
final int size() {
|
|
return resolveCache.size();
|
|
}
|
|
|
|
/**
|
|
* Returns {@code true} if this entry should replace all other entries that are already cached for the hostname.
|
|
*/
|
|
protected abstract boolean shouldReplaceAll(E entry);
|
|
|
|
/**
|
|
* Sort the {@link List} for a {@code hostname} before caching these.
|
|
*/
|
|
protected void sortEntries(
|
|
@SuppressWarnings("unused") String hostname, @SuppressWarnings("unused") List<E> entries) {
|
|
// NOOP.
|
|
}
|
|
|
|
/**
|
|
* Returns {@code true} if both entries are equal.
|
|
*/
|
|
protected abstract boolean equals(E entry, E otherEntry);
|
|
|
|
// Directly extend AtomicReference for intrinsics and also to keep memory overhead low.
|
|
private final class Entries extends AtomicReference<List<E>> implements Runnable {
|
|
|
|
private final String hostname;
|
|
// Needs to be package-private to be able to access it via the AtomicReferenceFieldUpdater
|
|
volatile ScheduledFuture<?> expirationFuture;
|
|
|
|
Entries(String hostname) {
|
|
super(Collections.<E>emptyList());
|
|
this.hostname = hostname;
|
|
}
|
|
|
|
void add(E e, int ttl, EventLoop loop) {
|
|
if (!shouldReplaceAll(e)) {
|
|
for (;;) {
|
|
List<E> entries = get();
|
|
if (!entries.isEmpty()) {
|
|
final E firstEntry = entries.get(0);
|
|
if (shouldReplaceAll(firstEntry)) {
|
|
assert entries.size() == 1;
|
|
|
|
if (compareAndSet(entries, singletonList(e))) {
|
|
scheduleCacheExpirationIfNeeded(ttl, loop);
|
|
return;
|
|
} else {
|
|
// Need to try again as CAS failed
|
|
continue;
|
|
}
|
|
}
|
|
|
|
// Create a new List for COW semantics
|
|
List<E> newEntries = new ArrayList<E>(entries.size() + 1);
|
|
int i = 0;
|
|
E replacedEntry = null;
|
|
do {
|
|
E entry = entries.get(i);
|
|
// Only add old entry if the address is not the same as the one we try to add as well.
|
|
// In this case we will skip it and just add the new entry as this may have
|
|
// more up-to-date data and cancel the old after we were able to update the cache.
|
|
if (!Cache.this.equals(e, entry)) {
|
|
newEntries.add(entry);
|
|
} else {
|
|
replacedEntry = entry;
|
|
newEntries.add(e);
|
|
|
|
++i;
|
|
for (; i < entries.size(); ++i) {
|
|
newEntries.add(entries.get(i));
|
|
}
|
|
break;
|
|
}
|
|
} while (++i < entries.size());
|
|
if (replacedEntry == null) {
|
|
newEntries.add(e);
|
|
}
|
|
sortEntries(hostname, newEntries);
|
|
|
|
if (compareAndSet(entries, Collections.unmodifiableList(newEntries))) {
|
|
scheduleCacheExpirationIfNeeded(ttl, loop);
|
|
return;
|
|
}
|
|
} else if (compareAndSet(entries, singletonList(e))) {
|
|
scheduleCacheExpirationIfNeeded(ttl, loop);
|
|
return;
|
|
}
|
|
}
|
|
} else {
|
|
set(singletonList(e));
|
|
scheduleCacheExpirationIfNeeded(ttl, loop);
|
|
}
|
|
}
|
|
|
|
private void scheduleCacheExpirationIfNeeded(int ttl, EventLoop loop) {
|
|
for (;;) {
|
|
// We currently don't calculate a new TTL when we need to retry the CAS as we don't expect this to
|
|
// be invoked very concurrently and also we use SECONDS anyway. If this ever becomes a problem
|
|
// we can reconsider.
|
|
ScheduledFuture<?> oldFuture = FUTURE_UPDATER.get(this);
|
|
if (oldFuture == null || oldFuture.getDelay(TimeUnit.SECONDS) > ttl) {
|
|
ScheduledFuture<?> newFuture = loop.schedule(this, ttl, TimeUnit.SECONDS);
|
|
// It is possible that
|
|
// 1. task will fire in between this line, or
|
|
// 2. multiple timers may be set if there is concurrency
|
|
// (1) Shouldn't be a problem because we will fail the CAS and then the next loop will see CANCELLED
|
|
// so the ttl will not be less, and we will bail out of the loop.
|
|
// (2) This is a trade-off to avoid concurrency resulting in contention on a synchronized block.
|
|
if (FUTURE_UPDATER.compareAndSet(this, oldFuture, newFuture)) {
|
|
if (oldFuture != null) {
|
|
oldFuture.cancel(true);
|
|
}
|
|
break;
|
|
} else {
|
|
// There was something else scheduled in the meantime... Cancel and try again.
|
|
newFuture.cancel(true);
|
|
}
|
|
} else {
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
boolean clearAndCancel() {
|
|
List<E> entries = getAndSet(Collections.<E>emptyList());
|
|
if (entries.isEmpty()) {
|
|
return false;
|
|
}
|
|
|
|
ScheduledFuture<?> expirationFuture = FUTURE_UPDATER.getAndSet(this, CANCELLED);
|
|
if (expirationFuture != null) {
|
|
expirationFuture.cancel(false);
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
@Override
|
|
public void run() {
|
|
// We always remove all entries for a hostname once one entry expire. This is not the
|
|
// most efficient to do but this way we can guarantee that if a DnsResolver
|
|
// be configured to prefer one ip family over the other we will not return unexpected
|
|
// results to the enduser if one of the A or AAAA records has different TTL settings.
|
|
//
|
|
// As a TTL is just a hint of the maximum time a cache is allowed to cache stuff it's
|
|
// completely fine to remove the entry even if the TTL is not reached yet.
|
|
//
|
|
// See https://github.com/netty/netty/issues/7329
|
|
resolveCache.remove(hostname, this);
|
|
|
|
clearAndCancel();
|
|
}
|
|
}
|
|
}
|