From: Ondřej Surý Date: Mon, 3 Feb 2025 12:36:27 +0000 (+0100) Subject: Reduce false sharing in dns_qpcache X-Git-Tag: v9.21.5~3^2~3 X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=c602d76c1fb6cfe54da9094ae1d12912d852d359;p=thirdparty%2Fbind9.git Reduce false sharing in dns_qpcache Instead of having many node_lock_count * sizeof() arrays, pack all the members into a qpcache_bucket_t struct that is cacheline aligned and have a single array of those. Additionaly, make both the head and the tail of isc_queue_t padded, not just the head, to prevent false sharing of the lock-free structure with the lock that follows it. --- diff --git a/lib/dns/qpcache.c b/lib/dns/qpcache.c index 560812a6d17..74bdba0223b 100644 --- a/lib/dns/qpcache.c +++ b/lib/dns/qpcache.c @@ -14,6 +14,7 @@ /*! \file */ #include +#include #include #include @@ -21,8 +22,6 @@ #include #include #include -#include -#include #include #include #include @@ -217,6 +216,41 @@ struct qpcnode { isc_queue_node_t deadlink; }; +/*% + * One bucket structure will be created for each loop, and + * nodes in the database will evenly distributed among buckets + * to reduce contention between threads. + */ +typedef struct qpcache_bucket { + /*% + * Temporary storage for stale cache nodes and dynamically + * deleted nodes that await being cleaned up. + */ + isc_queue_t deadnodes; + + /* Per-bucket lock. */ + isc_rwlock_t lock; + + /* + * Linked list used to implement LRU cache cleaning. + */ + dns_slabheaderlist_t lru; + + /* + * The heap is used for TTL based expiry. Note that qpcache->hmctx + * is the memory context to use for heap memory; this differs from + * the main database memory context, which is qpcache->common.mctx. + */ + isc_heap_t *heap; + + /* Padding to prevent false sharing between locks. */ + uint8_t __padding[ISC_OS_CACHELINE_SIZE - + (sizeof(dns_slabheaderlist_t) + sizeof(isc_heap_t *) + + sizeof(isc_rwlock_t)) % + ISC_OS_CACHELINE_SIZE]; + +} qpcache_bucket_t; + typedef struct qpcache qpcache_t; struct qpcache { /* Unlocked. */ @@ -245,10 +279,6 @@ struct qpcache { */ isc_refcount_t references; - /* Locks for individual tree nodes */ - unsigned int node_lock_count; - isc_rwlock_t *node_locks; - dns_stats_t *rrsetstats; isc_stats_t *cachestats; @@ -262,13 +292,6 @@ struct qpcache { */ uint32_t serve_stale_refresh; - /* - * This is an array of linked lists used to implement the LRU cache. - * There will be node_lock_count linked lists here. Nodes in bucket 1 - * will be placed on the linked list lru[1]. - */ - dns_slabheaderlist_t *lru; - /* * Start point % node_lock_count for next LRU cleanup. */ @@ -280,24 +303,14 @@ struct qpcache { */ _Atomic(isc_stdtime_t) last_used; - /*% - * Temporary storage for stale cache nodes and dynamically deleted - * nodes that await being cleaned up. - */ - isc_queue_t *deadnodes; - - /* - * Heaps. These are used for TTL based expiry in a cache, - * or for zone resigning in a zone DB. hmctx is the memory - * context to use for the heap (which differs from the main - * database memory context in the case of a cache). - */ - isc_mem_t *hmctx; - isc_heap_t **heaps; - /* Locked by tree_lock. */ dns_qp_t *tree; dns_qp_t *nsec; + + isc_mem_t *hmctx; /* Memory context for the heaps */ + + size_t buckets_count; + qpcache_bucket_t buckets[]; /* attribute((counted_by(buckets_count))) */ }; #ifdef DNS_DB_NODETRACE @@ -542,9 +555,11 @@ update_header(qpcache_t *qpdb, dns_slabheader_t *header, isc_stdtime_t now) { /* To be checked: can we really assume this? XXXMLG */ INSIST(ISC_LINK_LINKED(header, link)); - ISC_LIST_UNLINK(qpdb->lru[HEADERNODE(header)->locknum], header, link); + ISC_LIST_UNLINK(qpdb->buckets[HEADERNODE(header)->locknum].lru, header, + link); header->last_used = now; - ISC_LIST_PREPEND(qpdb->lru[HEADERNODE(header)->locknum], header, link); + ISC_LIST_PREPEND(qpdb->buckets[HEADERNODE(header)->locknum].lru, header, + link); } /* @@ -773,7 +788,7 @@ qpcnode_release(qpcache_t *qpdb, qpcnode_t *node, isc_rwlocktype_t *nlocktypep, * erefs (but NOT references!), upgrade the node lock, * decrement erefs again, and see if it's still zero. */ - isc_rwlock_t *nlock = &qpdb->node_locks[node->locknum]; + isc_rwlock_t *nlock = &qpdb->buckets[node->locknum].lock; qpcnode_erefs_increment(qpdb, node, *nlocktypep, *tlocktypep DNS__DB_FLARG_PASS); NODE_FORCEUPGRADE(nlock, nlocktypep); @@ -833,8 +848,9 @@ qpcnode_release(qpcache_t *qpdb, qpcnode_t *node, isc_rwlocktype_t *nlocktypep, *tlocktypep DNS__DB_FLARG_PASS); isc_queue_node_init(&node->deadlink); - if (!isc_queue_enqueue_entry(&qpdb->deadnodes[node->locknum], - node, deadlink)) + if (!isc_queue_enqueue_entry( + &qpdb->buckets[node->locknum].deadnodes, node, + deadlink)) { /* Queue was empty, trigger new cleaning */ isc_loop_t *loop = isc_loop_get(qpdb->loopmgr, @@ -1161,7 +1177,8 @@ setup_delegation(qpc_search_t *search, dns_dbnode_t **nodep, } if (rdataset != NULL) { isc_rwlocktype_t nlocktype = isc_rwlocktype_none; - isc_rwlock_t *nlock = &search->qpdb->node_locks[node->locknum]; + isc_rwlock_t *nlock = + &search->qpdb->buckets[node->locknum].lock; NODE_RDLOCK(nlock, &nlocktype); bindrdataset(search->qpdb, node, search->zonecut_header, search->now, nlocktype, tlocktype, @@ -1299,7 +1316,7 @@ check_zonecut(qpcnode_t *node, void *arg DNS__DB_FLARG) { REQUIRE(search->zonecut == NULL); - nlock = &search->qpdb->node_locks[node->locknum]; + nlock = &search->qpdb->buckets[node->locknum].lock; NODE_RDLOCK(nlock, &nlocktype); /* @@ -1372,7 +1389,7 @@ find_deepest_zonecut(qpc_search_t *search, qpcnode_t *node, isc_rwlocktype_t nlocktype = isc_rwlocktype_none; dns_qpchain_node(&search->chain, i, NULL, (void **)&node, NULL); - nlock = &qpdb->node_locks[node->locknum]; + nlock = &qpdb->buckets[node->locknum].lock; NODE_RDLOCK(nlock, &nlocktype); @@ -1522,7 +1539,7 @@ find_coveringnsec(qpc_search_t *search, const dns_name_t *name, } dns_name_copy(&node->name, fname); - nlock = &search->qpdb->node_locks[node->locknum]; + nlock = &search->qpdb->buckets[node->locknum].lock; NODE_RDLOCK(nlock, &nlocktype); for (header = node->data; header != NULL; header = header_next) { header_next = header->next; @@ -1700,7 +1717,7 @@ qpcache_find(dns_db_t *db, const dns_name_t *name, dns_dbversion_t *version, * We now go looking for rdata... */ - nlock = &search.qpdb->node_locks[node->locknum]; + nlock = &search.qpdb->buckets[node->locknum].lock; NODE_RDLOCK(nlock, &nlocktype); /* @@ -2004,7 +2021,7 @@ tree_exit: if (search.need_cleanup) { node = search.zonecut; INSIST(node != NULL); - nlock = &search.qpdb->node_locks[node->locknum]; + nlock = &search.qpdb->buckets[node->locknum].lock; NODE_RDLOCK(nlock, &nlocktype); qpcnode_release(search.qpdb, node, &nlocktype, &tlocktype, @@ -2088,7 +2105,7 @@ qpcache_findzonecut(dns_db_t *db, const dns_name_t *name, unsigned int options, * We now go looking for an NS rdataset at the node. */ - nlock = &search.qpdb->node_locks[node->locknum]; + nlock = &search.qpdb->buckets[node->locknum].lock; NODE_RDLOCK(nlock, &nlocktype); for (header = node->data; header != NULL; header = header_next) { @@ -2210,7 +2227,7 @@ qpcache_findrdataset(dns_db_t *db, dns_dbnode_t *node, dns_dbversion_t *version, now = isc_stdtime_now(); } - nlock = &qpdb->node_locks[qpnode->locknum]; + nlock = &qpdb->buckets[qpnode->locknum].lock; NODE_RDLOCK(nlock, &nlocktype); matchtype = DNS_TYPEPAIR_VALUE(type, covers); @@ -2355,7 +2372,7 @@ expiredata(dns_db_t *db, dns_dbnode_t *node, void *data) { dns_slabheader_t *header = data; isc_rwlocktype_t nlocktype = isc_rwlocktype_none; isc_rwlocktype_t tlocktype = isc_rwlocktype_none; - isc_rwlock_t *nlock = &qpdb->node_locks[qpnode->locknum]; + isc_rwlock_t *nlock = &qpdb->buckets[qpnode->locknum].lock; NODE_WRLOCK(nlock, &nlocktype); expireheader(header, &nlocktype, &tlocktype, @@ -2381,10 +2398,10 @@ expire_lru_headers(qpcache_t *qpdb, unsigned int locknum, dns_slabheader_t *header = NULL; size_t purged = 0; - for (header = ISC_LIST_TAIL(qpdb->lru[locknum]); + for (header = ISC_LIST_TAIL(qpdb->buckets[locknum].lru); header != NULL && header->last_used <= qpdb->last_used && purged <= purgesize; - header = ISC_LIST_TAIL(qpdb->lru[locknum])) + header = ISC_LIST_TAIL(qpdb->buckets[locknum].lru)) { size_t header_size = rdataset_size(header); @@ -2395,7 +2412,7 @@ expire_lru_headers(qpcache_t *qpdb, unsigned int locknum, * referenced any more (so unlinking is safe) since the * TTL will be reset to 0. */ - ISC_LIST_UNLINK(qpdb->lru[locknum], header, link); + ISC_LIST_UNLINK(qpdb->buckets[locknum].lru, header, link); expireheader(header, nlocktypep, tlocktypep, dns_expire_lru DNS__DB_FLARG_PASS); purged += header_size; @@ -2417,7 +2434,7 @@ expire_lru_headers(qpcache_t *qpdb, unsigned int locknum, static void overmem(qpcache_t *qpdb, dns_slabheader_t *newheader, isc_rwlocktype_t *tlocktypep DNS__DB_FLARG) { - uint32_t locknum_start = qpdb->lru_sweep++ % qpdb->node_lock_count; + uint32_t locknum_start = qpdb->lru_sweep++ % qpdb->buckets_count; uint32_t locknum = locknum_start; size_t purgesize, purged = 0; isc_stdtime_t min_last_used = 0; @@ -2437,7 +2454,7 @@ overmem(qpcache_t *qpdb, dns_slabheader_t *newheader, again: do { isc_rwlocktype_t nlocktype = isc_rwlocktype_none; - isc_rwlock_t *nlock = &qpdb->node_locks[locknum]; + isc_rwlock_t *nlock = &qpdb->buckets[locknum].lock; NODE_WRLOCK(nlock, &nlocktype); purged += expire_lru_headers( @@ -2448,14 +2465,15 @@ again: * Work out the oldest remaining last_used values of the list * tails as we walk across the array of lru lists. */ - dns_slabheader_t *header = ISC_LIST_TAIL(qpdb->lru[locknum]); + dns_slabheader_t *header = + ISC_LIST_TAIL(qpdb->buckets[locknum].lru); if (header != NULL && (min_last_used == 0 || header->last_used < min_last_used)) { min_last_used = header->last_used; } NODE_UNLOCK(nlock, &nlocktype); - locknum = (locknum + 1) % qpdb->node_lock_count; + locknum = (locknum + 1) % qpdb->buckets_count; } while (locknum != locknum_start && purged <= purgesize); /* @@ -2527,40 +2545,15 @@ qpcache__destroy(qpcache_t *qpdb) { if (dns_name_dynamic(&qpdb->common.origin)) { dns_name_free(&qpdb->common.origin, qpdb->common.mctx); } - for (i = 0; i < qpdb->node_lock_count; i++) { - NODE_DESTROYLOCK(&qpdb->node_locks[i]); - } + for (i = 0; i < qpdb->buckets_count; i++) { + NODE_DESTROYLOCK(&qpdb->buckets[i].lock); - /* - * Clean up LRU / re-signing order lists. - */ - if (qpdb->lru != NULL) { - for (i = 0; i < qpdb->node_lock_count; i++) { - INSIST(ISC_LIST_EMPTY(qpdb->lru[i])); - } - isc_mem_cput(qpdb->common.mctx, qpdb->lru, - qpdb->node_lock_count, - sizeof(dns_slabheaderlist_t)); - } - /* - * Clean up dead node buckets. - */ - for (i = 0; i < qpdb->node_lock_count; i++) { - INSIST(isc_queue_empty(&qpdb->deadnodes[i])); - isc_queue_destroy(&qpdb->deadnodes[i]); - } - isc_mem_cput(qpdb->common.mctx, qpdb->deadnodes, qpdb->node_lock_count, - sizeof(qpdb->deadnodes[0])); + INSIST(ISC_LIST_EMPTY(qpdb->buckets[i].lru)); - /* - * Clean up heap objects. - */ - if (qpdb->heaps != NULL) { - for (i = 0; i < qpdb->node_lock_count; i++) { - isc_heap_destroy(&qpdb->heaps[i]); - } - isc_mem_cput(qpdb->hmctx, qpdb->heaps, qpdb->node_lock_count, - sizeof(isc_heap_t *)); + INSIST(isc_queue_empty(&qpdb->buckets[i].deadnodes)); + isc_queue_destroy(&qpdb->buckets[i].deadnodes); + + isc_heap_destroy(&qpdb->buckets[i].heap); } if (qpdb->rrsetstats != NULL) { @@ -2570,8 +2563,6 @@ qpcache__destroy(qpcache_t *qpdb) { isc_stats_detach(&qpdb->cachestats); } - isc_mem_cput(qpdb->common.mctx, qpdb->node_locks, qpdb->node_lock_count, - sizeof(qpdb->node_locks[0])); TREE_DESTROYLOCK(&qpdb->tree_lock); isc_refcount_destroy(&qpdb->references); isc_refcount_destroy(&qpdb->common.references); @@ -2581,7 +2572,9 @@ qpcache__destroy(qpcache_t *qpdb) { qpdb->common.impmagic = 0; isc_mem_detach(&qpdb->hmctx); - isc_mem_putanddetach(&qpdb->common.mctx, qpdb, sizeof(*qpdb)); + isc_mem_putanddetach(&qpdb->common.mctx, qpdb, + sizeof(*qpdb) + qpdb->buckets_count * + sizeof(qpdb->buckets[0])); } static void @@ -2603,18 +2596,19 @@ cleanup_deadnodes(void *arg) { uint16_t locknum = isc_tid(); isc_rwlocktype_t tlocktype = isc_rwlocktype_none; isc_rwlocktype_t nlocktype = isc_rwlocktype_none; - isc_rwlock_t *nlock = &qpdb->node_locks[locknum]; + isc_rwlock_t *nlock = &qpdb->buckets[locknum].lock; qpcnode_t *qpnode = NULL, *qpnext = NULL; isc_queue_t deadnodes; - INSIST(locknum < qpdb->node_lock_count); + INSIST(locknum < qpdb->buckets_count); isc_queue_init(&deadnodes); TREE_WRLOCK(&qpdb->tree_lock, &tlocktype); NODE_WRLOCK(nlock, &nlocktype); - RUNTIME_CHECK(isc_queue_splice(&deadnodes, &qpdb->deadnodes[locknum])); + RUNTIME_CHECK(isc_queue_splice(&deadnodes, + &qpdb->buckets[locknum].deadnodes)); isc_queue_for_each_entry_safe(&deadnodes, qpnode, qpnext, deadlink) { qpcnode_release(qpdb, qpnode, &nlocktype, &tlocktype, false); } @@ -2636,7 +2630,7 @@ static void reactivate_node(qpcache_t *qpdb, qpcnode_t *node, isc_rwlocktype_t tlocktype ISC_ATTR_UNUSED DNS__DB_FLARG) { isc_rwlocktype_t nlocktype = isc_rwlocktype_none; - isc_rwlock_t *nlock = &qpdb->node_locks[node->locknum]; + isc_rwlock_t *nlock = &qpdb->buckets[node->locknum].lock; NODE_RDLOCK(nlock, &nlocktype); qpcnode_acquire(qpdb, node, nlocktype, tlocktype DNS__DB_FLARG_PASS); @@ -2649,11 +2643,9 @@ new_qpcnode(qpcache_t *qpdb, const dns_name_t *name) { *newdata = (qpcnode_t){ .name = DNS_NAME_INITEMPTY, .references = ISC_REFCOUNT_INITIALIZER(1), - .locknum = isc_random_uniform(qpdb->node_lock_count), + .locknum = isc_random_uniform(qpdb->buckets_count), }; - INSIST(newdata->locknum < qpdb->node_lock_count); - isc_mem_attach(qpdb->common.mctx, &newdata->mctx); dns_name_dupwithoffsets(name, newdata->mctx, &newdata->name); @@ -2728,7 +2720,7 @@ qpcache_detachnode(dns_db_t *db, dns_dbnode_t **nodep DNS__DB_FLARG) { node = (qpcnode_t *)(*nodep); *nodep = NULL; - nlock = &qpdb->node_locks[node->locknum]; + nlock = &qpdb->buckets[node->locknum].lock; /* * We can't destroy qpcache while holding a nodelock, so @@ -3022,11 +3014,13 @@ find_header: } if (header->last_used != now) { ISC_LIST_UNLINK( - qpdb->lru[HEADERNODE(header)->locknum], + qpdb->buckets[HEADERNODE(header)->locknum] + .lru, header, link); header->last_used = now; ISC_LIST_PREPEND( - qpdb->lru[HEADERNODE(header)->locknum], + qpdb->buckets[HEADERNODE(header)->locknum] + .lru, header, link); } if (header->noqname == NULL && @@ -3084,11 +3078,13 @@ find_header: } if (header->last_used != now) { ISC_LIST_UNLINK( - qpdb->lru[HEADERNODE(header)->locknum], + qpdb->buckets[HEADERNODE(header)->locknum] + .lru, header, link); header->last_used = now; ISC_LIST_PREPEND( - qpdb->lru[HEADERNODE(header)->locknum], + qpdb->buckets[HEADERNODE(header)->locknum] + .lru, header, link); } if (header->noqname == NULL && @@ -3117,15 +3113,14 @@ find_header: idx = HEADERNODE(newheader)->locknum; if (ZEROTTL(newheader)) { newheader->last_used = qpdb->last_used + 1; - ISC_LIST_APPEND(qpdb->lru[idx], newheader, - link); + ISC_LIST_APPEND(qpdb->buckets[idx].lru, + newheader, link); } else { - ISC_LIST_PREPEND(qpdb->lru[idx], newheader, - link); + ISC_LIST_PREPEND(qpdb->buckets[idx].lru, + newheader, link); } - INSIST(qpdb->heaps != NULL); - isc_heap_insert(qpdb->heaps[idx], newheader); - newheader->heap = qpdb->heaps[idx]; + isc_heap_insert(qpdb->buckets[idx].heap, newheader); + newheader->heap = qpdb->buckets[idx].heap; /* * There are no other references to 'header' when @@ -3142,16 +3137,15 @@ find_header: dns_slabheader_destroy(&header); } else { idx = HEADERNODE(newheader)->locknum; - INSIST(qpdb->heaps != NULL); - isc_heap_insert(qpdb->heaps[idx], newheader); - newheader->heap = qpdb->heaps[idx]; + isc_heap_insert(qpdb->buckets[idx].heap, newheader); + newheader->heap = qpdb->buckets[idx].heap; if (ZEROTTL(newheader)) { newheader->last_used = qpdb->last_used + 1; - ISC_LIST_APPEND(qpdb->lru[idx], newheader, - link); + ISC_LIST_APPEND(qpdb->buckets[idx].lru, + newheader, link); } else { - ISC_LIST_PREPEND(qpdb->lru[idx], newheader, - link); + ISC_LIST_PREPEND(qpdb->buckets[idx].lru, + newheader, link); } if (topheader_prev != NULL) { topheader_prev->next = newheader; @@ -3181,12 +3175,14 @@ find_header: } idx = HEADERNODE(newheader)->locknum; - isc_heap_insert(qpdb->heaps[idx], newheader); - newheader->heap = qpdb->heaps[idx]; + isc_heap_insert(qpdb->buckets[idx].heap, newheader); + newheader->heap = qpdb->buckets[idx].heap; if (ZEROTTL(newheader)) { - ISC_LIST_APPEND(qpdb->lru[idx], newheader, link); + ISC_LIST_APPEND(qpdb->buckets[idx].lru, newheader, + link); } else { - ISC_LIST_PREPEND(qpdb->lru[idx], newheader, link); + ISC_LIST_PREPEND(qpdb->buckets[idx].lru, newheader, + link); } if (topheader != NULL) { @@ -3470,7 +3466,7 @@ qpcache_addrdataset(dns_db_t *db, dns_dbnode_t *node, dns_dbversion_t *version, overmem(qpdb, newheader, &tlocktype DNS__DB_FLARG_PASS); } - nlock = &qpdb->node_locks[qpnode->locknum]; + nlock = &qpdb->buckets[qpnode->locknum].lock; NODE_WRLOCK(nlock, &nlocktype); @@ -3557,7 +3553,7 @@ qpcache_deleterdataset(dns_db_t *db, dns_dbnode_t *node, setttl(newheader, 0); atomic_init(&newheader->attributes, DNS_SLABHEADERATTR_NONEXISTENT); - nlock = &qpdb->node_locks[qpnode->locknum]; + nlock = &qpdb->buckets[qpnode->locknum].lock; NODE_WRLOCK(nlock, &nlocktype); result = add(qpdb, qpnode, NULL, newheader, DNS_DBADD_FORCE, false, NULL, 0, nlocktype, @@ -3596,7 +3592,7 @@ locknode(dns_db_t *db, dns_dbnode_t *node, isc_rwlocktype_t type) { qpcache_t *qpdb = (qpcache_t *)db; qpcnode_t *qpnode = (qpcnode_t *)node; - RWLOCK(&qpdb->node_locks[qpnode->locknum], type); + RWLOCK(&qpdb->buckets[qpnode->locknum].lock, type); } static void @@ -3604,7 +3600,7 @@ unlocknode(dns_db_t *db, dns_dbnode_t *node, isc_rwlocktype_t type) { qpcache_t *qpdb = (qpcache_t *)db; qpcnode_t *qpnode = (qpcnode_t *)node; - RWUNLOCK(&qpdb->node_locks[qpnode->locknum], type); + RWUNLOCK(&qpdb->buckets[qpnode->locknum].lock, type); } isc_result_t @@ -3616,12 +3612,15 @@ dns__qpcache_create(isc_mem_t *mctx, const dns_name_t *origin, isc_mem_t *hmctx = mctx; isc_loop_t *loop = isc_loop(); int i; + isc_loopmgr_t *loopmgr = isc_loop_getloopmgr(loop); + size_t nloops = isc_loopmgr_nloops(loopmgr); /* This database implementation only supports cache semantics */ REQUIRE(type == dns_dbtype_cache); REQUIRE(loop != NULL); - qpdb = isc_mem_get(mctx, sizeof(*qpdb)); + qpdb = isc_mem_get(mctx, + sizeof(*qpdb) + nloops * sizeof(qpdb->buckets[0])); *qpdb = (qpcache_t){ .common.methods = &qpdb_cachemethods, .common.origin = DNS_NAME_INITEMPTY, @@ -3630,6 +3629,7 @@ dns__qpcache_create(isc_mem_t *mctx, const dns_name_t *origin, .common.references = 1, .loopmgr = isc_loop_getloopmgr(loop), .references = 1, + .buckets_count = nloops, }; /* @@ -3642,38 +3642,19 @@ dns__qpcache_create(isc_mem_t *mctx, const dns_name_t *origin, isc_rwlock_init(&qpdb->lock); TREE_INITLOCK(&qpdb->tree_lock); - qpdb->node_lock_count = isc_loopmgr_nloops(qpdb->loopmgr); - qpdb->node_locks = isc_mem_cget(mctx, qpdb->node_lock_count, - sizeof(qpdb->node_locks[0])); + qpdb->buckets_count = isc_loopmgr_nloops(qpdb->loopmgr); dns_rdatasetstats_create(mctx, &qpdb->rrsetstats); - qpdb->lru = isc_mem_cget(mctx, qpdb->node_lock_count, - sizeof(dns_slabheaderlist_t)); - for (i = 0; i < (int)qpdb->node_lock_count; i++) { - ISC_LIST_INIT(qpdb->lru[i]); - } + for (i = 0; i < (int)qpdb->buckets_count; i++) { + ISC_LIST_INIT(qpdb->buckets[i].lru); - /* - * Create the heaps. - */ - qpdb->heaps = isc_mem_cget(hmctx, qpdb->node_lock_count, - sizeof(isc_heap_t *)); - for (i = 0; i < (int)qpdb->node_lock_count; i++) { + qpdb->buckets[i].heap = NULL; isc_heap_create(hmctx, ttl_sooner, set_index, 0, - &qpdb->heaps[i]); - } + &qpdb->buckets[i].heap); - /* - * Create deadnode lists. - */ - qpdb->deadnodes = isc_mem_cget(mctx, qpdb->node_lock_count, - sizeof(qpdb->deadnodes[0])); - for (i = 0; i < (int)(qpdb->node_lock_count); i++) { - isc_queue_init(&qpdb->deadnodes[i]); - } + isc_queue_init(&qpdb->buckets[i].deadnodes); - for (i = 0; i < (int)(qpdb->node_lock_count); i++) { - NODE_INITLOCK(&qpdb->node_locks[i]); + NODE_INITLOCK(&qpdb->buckets[i].lock); } /* @@ -3756,7 +3737,7 @@ rdatasetiter_first(dns_rdatasetiter_t *it DNS__DB_FLARG) { qpcnode_t *qpnode = (qpcnode_t *)iterator->common.node; dns_slabheader_t *header = NULL, *top_next = NULL; isc_rwlocktype_t nlocktype = isc_rwlocktype_none; - isc_rwlock_t *nlock = &qpdb->node_locks[qpnode->locknum]; + isc_rwlock_t *nlock = &qpdb->buckets[qpnode->locknum].lock; NODE_RDLOCK(nlock, &nlocktype); @@ -3802,7 +3783,7 @@ rdatasetiter_next(dns_rdatasetiter_t *it DNS__DB_FLARG) { dns_typepair_t type, negtype; dns_rdatatype_t rdtype, covers; isc_rwlocktype_t nlocktype = isc_rwlocktype_none; - isc_rwlock_t *nlock = &qpdb->node_locks[qpnode->locknum]; + isc_rwlock_t *nlock = &qpdb->buckets[qpnode->locknum].lock; bool expiredok = EXPIREDOK(iterator); header = iterator->current; @@ -3890,7 +3871,7 @@ rdatasetiter_current(dns_rdatasetiter_t *it, qpcnode_t *qpnode = (qpcnode_t *)iterator->common.node; dns_slabheader_t *header = NULL; isc_rwlocktype_t nlocktype = isc_rwlocktype_none; - isc_rwlock_t *nlock = &qpdb->node_locks[qpnode->locknum]; + isc_rwlock_t *nlock = &qpdb->buckets[qpnode->locknum].lock; header = iterator->current; REQUIRE(header != NULL); @@ -3934,7 +3915,7 @@ dereference_iter_node(qpc_dbit_t *qpdbiter DNS__DB_FLARG) { REQUIRE(tlocktype != isc_rwlocktype_write); - nlock = &qpdb->node_locks[node->locknum]; + nlock = &qpdb->buckets[node->locknum].lock; NODE_RDLOCK(nlock, &nlocktype); qpcnode_release(qpdb, node, &nlocktype, &qpdbiter->tree_locked, false DNS__DB_FLARG_PASS); @@ -4252,7 +4233,7 @@ deletedata(dns_db_t *db ISC_ATTR_UNUSED, dns_dbnode_t *node ISC_ATTR_UNUSED, if (ISC_LINK_LINKED(header, link)) { int idx = HEADERNODE(header)->locknum; - ISC_LIST_UNLINK(qpdb->lru[idx], header, link); + ISC_LIST_UNLINK(qpdb->buckets[idx].lru, header, link); } if (header->noqname != NULL) { @@ -4270,7 +4251,7 @@ static void expire_ttl_headers(qpcache_t *qpdb, unsigned int locknum, isc_rwlocktype_t *nlocktypep, isc_rwlocktype_t *tlocktypep, isc_stdtime_t now, bool cache_is_overmem DNS__DB_FLARG) { - isc_heap_t *heap = qpdb->heaps[locknum]; + isc_heap_t *heap = qpdb->buckets[locknum].heap; for (size_t i = 0; i < DNS_QPDB_EXPIRE_TTL_COUNT; i++) { dns_slabheader_t *header = isc_heap_element(heap, 1); diff --git a/lib/isc/include/isc/queue.h b/lib/isc/include/isc/queue.h index d34f551262d..a2abe24c15b 100644 --- a/lib/isc/include/isc/queue.h +++ b/lib/isc/include/isc/queue.h @@ -20,9 +20,11 @@ STATIC_ASSERT(sizeof(struct __cds_wfcq_head) <= ISC_OS_CACHELINE_SIZE, typedef struct isc_queue { struct __cds_wfcq_head head; - uint8_t __padding[ISC_OS_CACHELINE_SIZE - - sizeof(struct __cds_wfcq_head)]; + uint8_t __padding_head[ISC_OS_CACHELINE_SIZE - + sizeof(struct __cds_wfcq_head)]; struct cds_wfcq_tail tail; + uint8_t __padding_tail[ISC_OS_CACHELINE_SIZE - + sizeof(struct __cds_wfcq_head)]; } isc_queue_t; typedef struct cds_wfcq_node isc_queue_node_t;