2
0
mirror of https://gitlab.isc.org/isc-projects/bind9 synced 2025-08-22 18:19:42 +00:00
bind/lib/dns/badcache.c
Ondřej Surý 4dacdde28f
Refactor dns_badcache to use cds_lfht lock-free hashtable
The dns_badcache unit had (yet another) own locked hashtable
implementation.  Replace the hashtable used by dns_badcache with
lock-free cds_lfht implementation from liburcu.
2023-07-31 15:51:15 +02:00

395 lines
9.7 KiB
C

/*
* Copyright (C) Internet Systems Consortium, Inc. ("ISC")
*
* SPDX-License-Identifier: MPL-2.0
*
* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, you can obtain one at https://mozilla.org/MPL/2.0/.
*
* See the COPYRIGHT file distributed with this work for additional
* information regarding copyright ownership.
*/
/*! \file */
#include <inttypes.h>
#include <stdbool.h>
#include <isc/atomic.h>
#include <isc/buffer.h>
#include <isc/hash.h>
#include <isc/log.h>
#include <isc/mem.h>
#include <isc/mutex.h>
#include <isc/refcount.h>
#include <isc/rwlock.h>
#include <isc/spinlock.h>
#include <isc/stdtime.h>
#include <isc/string.h>
#include <isc/thread.h>
#include <isc/time.h>
#include <isc/urcu.h>
#include <isc/util.h>
#include <dns/badcache.h>
#include <dns/fixedname.h>
#include <dns/name.h>
#include <dns/rdatatype.h>
#include <dns/types.h>
typedef struct dns_bcentry dns_bcentry_t;
struct dns_badcache {
unsigned int magic;
isc_mem_t *mctx;
struct cds_lfht *ht;
atomic_bool purge_in_progress;
};
#define BADCACHE_MAGIC ISC_MAGIC('B', 'd', 'C', 'a')
#define VALID_BADCACHE(m) ISC_MAGIC_VALID(m, BADCACHE_MAGIC)
#define BADCACHE_INIT_SIZE (1 << 10) /* Must be power of 2 */
#define BADCACHE_MIN_SIZE (1 << 8) /* Must be power of 2 */
struct dns_bcentry {
isc_mem_t *mctx;
dns_rdatatype_t type;
_Atomic(isc_stdtime_t) expire;
atomic_uint_fast32_t flags;
dns_fixedname_t fname;
dns_name_t *name;
struct cds_lfht_node ht_node;
struct rcu_head rcu_head;
};
static void
bcentry_print(dns_bcentry_t *bad, isc_stdtime_t now, FILE *fp);
static void
bcentry_destroy(struct rcu_head *rcu_head);
dns_badcache_t *
dns_badcache_new(isc_mem_t *mctx) {
REQUIRE(mctx != NULL);
dns_badcache_t *bc = isc_mem_get(mctx, sizeof(*bc));
*bc = (dns_badcache_t){
.magic = BADCACHE_MAGIC,
};
bc->ht = cds_lfht_new(BADCACHE_INIT_SIZE, BADCACHE_MIN_SIZE, 0,
CDS_LFHT_AUTO_RESIZE | CDS_LFHT_ACCOUNTING, NULL);
INSIST(bc->ht != NULL);
isc_mem_attach(mctx, &bc->mctx);
return (bc);
}
void
dns_badcache_destroy(dns_badcache_t **bcp) {
REQUIRE(bcp != NULL && *bcp != NULL);
REQUIRE(VALID_BADCACHE(*bcp));
dns_badcache_t *bc = *bcp;
*bcp = NULL;
bc->magic = 0;
dns_bcentry_t *bad = NULL;
struct cds_lfht_iter iter;
cds_lfht_for_each_entry(bc->ht, &iter, bad, ht_node) {
INSIST(!cds_lfht_del(bc->ht, &bad->ht_node));
bcentry_destroy(&bad->rcu_head);
}
RUNTIME_CHECK(!cds_lfht_destroy(bc->ht, NULL));
isc_mem_putanddetach(&bc->mctx, bc, sizeof(dns_badcache_t));
}
static int
bcentry_match(struct cds_lfht_node *ht_node, const void *key) {
const dns_name_t *name = key;
dns_bcentry_t *bad = caa_container_of(ht_node, dns_bcentry_t, ht_node);
return (dns_name_equal(bad->name, name));
}
static dns_bcentry_t *
bcentry_new(dns_badcache_t *bc, const dns_name_t *name,
const dns_rdatatype_t type, const uint32_t flags,
const isc_stdtime_t expire) {
dns_bcentry_t *bad = isc_mem_get(bc->mctx, sizeof(*bad));
*bad = (dns_bcentry_t){
.type = type,
.flags = flags,
.expire = expire,
};
isc_mem_attach(bc->mctx, &bad->mctx);
bad->name = dns_fixedname_initname(&bad->fname);
dns_name_copy(name, bad->name);
return (bad);
}
static void
bcentry_destroy(struct rcu_head *rcu_head) {
dns_bcentry_t *bad = caa_container_of(rcu_head, dns_bcentry_t,
rcu_head);
isc_mem_putanddetach(&bad->mctx, bad, sizeof(*bad));
}
static void
bcentry_evict(struct cds_lfht *ht, dns_bcentry_t *bad) {
/*
* The hashtable isn't locked in a traditional sense, so multiple
* threads can lookup and evict the same record at the same time.
*
* This is amplified by the bcentry_purge_next() that walks a few more
* records in the hashtable and evicts them if they are expired.
*
* We need to destroy the bcentry only once - from the thread that has
* deleted the entry from the hashtable, all other calls to this
* function were redundant.
*/
if (!cds_lfht_del(ht, &bad->ht_node)) {
call_rcu(&bad->rcu_head, bcentry_destroy);
}
}
static bool
bcentry_alive(struct cds_lfht *ht, dns_bcentry_t *bad, isc_stdtime_t now) {
if (cds_lfht_is_node_deleted(&bad->ht_node)) {
return (false);
} else if (atomic_load_relaxed(&bad->expire) < now) {
bcentry_evict(ht, bad);
return (false);
}
return (true);
}
#define cds_lfht_for_each_entry_next(ht, iter, pos, member) \
for (cds_lfht_next(ht, iter), \
pos = cds_lfht_entry(cds_lfht_iter_get_node(iter), \
__typeof__(*(pos)), member); \
pos != NULL; /**/ \
cds_lfht_next(ht, iter), \
pos = cds_lfht_entry(cds_lfht_iter_get_node(iter), \
__typeof__(*(pos)), member))
static void
bcentry_purge_next(struct cds_lfht *ht, struct cds_lfht_iter *iter,
isc_stdtime_t now) {
/* Lazy-purge the table */
size_t count = 10;
dns_bcentry_t *bad;
cds_lfht_for_each_entry_next(ht, iter, bad, ht_node) {
if (!bcentry_alive(ht, bad, now)) {
break;
}
if (--count == 0) {
break;
}
}
}
void
dns_badcache_add(dns_badcache_t *bc, const dns_name_t *name,
dns_rdatatype_t type, bool update, uint32_t flags,
isc_stdtime_t expire) {
REQUIRE(VALID_BADCACHE(bc));
REQUIRE(name != NULL);
isc_stdtime_t now = isc_stdtime_now();
if (expire < now) {
expire = now;
}
rcu_read_lock();
struct cds_lfht *ht = rcu_dereference(bc->ht);
INSIST(ht != NULL);
dns_bcentry_t *bad = NULL;
uint32_t hashval = dns_name_hash(name);
struct cds_lfht_iter iter;
dns_bcentry_t *found = NULL;
cds_lfht_for_each_entry_duplicate(ht, hashval, bcentry_match, name,
&iter, bad, ht_node) {
if (bcentry_alive(ht, bad, now) && bad->type == type) {
found = bad;
/*
* We could bail-out on first match, but:
* 1. there could be duplicate .type entries
* 2. we want to check expire for all entries
*/
}
}
if (found == NULL) {
/*
* In theory, this could result in multiple entries for the same
* type, but we don't care much, as they are going to be roughly
* the same, and the last will always trump and the former
* entries will expire (see above).
*/
bad = bcentry_new(bc, name, type, flags, expire);
cds_lfht_add(ht, hashval, &bad->ht_node);
} else if (update) {
atomic_store_relaxed(&found->expire, expire);
atomic_store_relaxed(&found->flags, flags);
}
rcu_read_unlock();
}
isc_result_t
dns_badcache_find(dns_badcache_t *bc, const dns_name_t *name,
dns_rdatatype_t type, uint32_t *flagp, isc_stdtime_t now) {
REQUIRE(VALID_BADCACHE(bc));
REQUIRE(name != NULL);
isc_result_t result = ISC_R_NOTFOUND;
rcu_read_lock();
struct cds_lfht *ht = rcu_dereference(bc->ht);
INSIST(ht != NULL);
dns_bcentry_t *bad = NULL;
uint32_t hashval = dns_name_hash(name);
struct cds_lfht_iter iter;
dns_bcentry_t *found = NULL;
cds_lfht_for_each_entry_duplicate(ht, hashval, bcentry_match, name,
&iter, bad, ht_node) {
if (bad->type == type && bcentry_alive(ht, bad, now)) {
found = bad;
}
}
if (found) {
result = ISC_R_SUCCESS;
if (flagp != NULL) {
*flagp = atomic_load_relaxed(&found->flags);
}
bcentry_purge_next(ht, &iter, now);
}
rcu_read_unlock();
return (result);
}
void
dns_badcache_flush(dns_badcache_t *bc) {
REQUIRE(VALID_BADCACHE(bc));
struct cds_lfht *ht =
cds_lfht_new(BADCACHE_INIT_SIZE, BADCACHE_MIN_SIZE, 0,
CDS_LFHT_AUTO_RESIZE | CDS_LFHT_ACCOUNTING, NULL);
INSIST(ht != NULL);
/* First swap the hashtables */
rcu_read_lock();
ht = rcu_xchg_pointer(&bc->ht, ht);
rcu_read_unlock();
/* Make sure nobody is using the old hash table */
synchronize_rcu();
/* Flush the old hash table */
dns_bcentry_t *bad = NULL;
struct cds_lfht_iter iter;
cds_lfht_for_each_entry(ht, &iter, bad, ht_node) {
INSIST(!cds_lfht_del(ht, &bad->ht_node));
bcentry_destroy(&bad->rcu_head);
}
RUNTIME_CHECK(!cds_lfht_destroy(ht, NULL));
}
void
dns_badcache_flushname(dns_badcache_t *bc, const dns_name_t *name) {
REQUIRE(VALID_BADCACHE(bc));
REQUIRE(name != NULL);
rcu_read_lock();
struct cds_lfht *ht = rcu_dereference(bc->ht);
INSIST(ht != NULL);
dns_bcentry_t *bad = NULL;
uint32_t hashval = dns_name_hash(name);
struct cds_lfht_iter iter;
cds_lfht_for_each_entry_duplicate(ht, hashval, bcentry_match, name,
&iter, bad, ht_node) {
bcentry_evict(ht, bad);
}
rcu_read_unlock();
}
void
dns_badcache_flushtree(dns_badcache_t *bc, const dns_name_t *name) {
dns_bcentry_t *bad;
isc_stdtime_t now = isc_stdtime_now();
REQUIRE(VALID_BADCACHE(bc));
REQUIRE(name != NULL);
rcu_read_lock();
struct cds_lfht *ht = rcu_dereference(bc->ht);
INSIST(ht != NULL);
struct cds_lfht_iter iter;
cds_lfht_for_each_entry(ht, &iter, bad, ht_node) {
if (dns_name_issubdomain(bad->name, name)) {
bcentry_evict(ht, bad);
} else if (!bcentry_alive(ht, bad, now)) {
/* Flush all the expired entries */
}
}
rcu_read_unlock();
}
static void
bcentry_print(dns_bcentry_t *bad, isc_stdtime_t now, FILE *fp) {
char namebuf[DNS_NAME_FORMATSIZE];
char typebuf[DNS_RDATATYPE_FORMATSIZE];
dns_name_format(bad->name, namebuf, sizeof(namebuf));
dns_rdatatype_format(bad->type, typebuf, sizeof(typebuf));
fprintf(fp, "; %s/%s [ttl %" PRIu32 "]\n", namebuf, typebuf,
atomic_load_relaxed(&bad->expire) - now);
}
void
dns_badcache_print(dns_badcache_t *bc, const char *cachename, FILE *fp) {
dns_bcentry_t *bad;
isc_stdtime_t now = isc_stdtime_now();
REQUIRE(VALID_BADCACHE(bc));
REQUIRE(fp != NULL);
fprintf(fp, ";\n; %s\n;\n", cachename);
rcu_read_lock();
struct cds_lfht *ht = rcu_dereference(bc->ht);
INSIST(ht != NULL);
struct cds_lfht_iter iter;
cds_lfht_for_each_entry(ht, &iter, bad, ht_node) {
if (bcentry_alive(ht, bad, now)) {
bcentry_print(bad, now, fp);
}
}
rcu_read_unlock();
}