From patchwork Fri Oct 12 06:31:52 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Honnappa Nagarahalli X-Patchwork-Id: 148716 Delivered-To: patch@linaro.org Received: by 2002:a2e:8595:0:0:0:0:0 with SMTP id b21-v6csp368042lji; Thu, 11 Oct 2018 23:32:19 -0700 (PDT) X-Google-Smtp-Source: ACcGV63zIqYRR8shf+lU7pYh7viHq3GbqU6romid34sT08uFiHA8z+dpd2egIs/bR9st5tITLbq2 X-Received: by 2002:a50:de47:: with SMTP id a7-v6mr6937393edl.245.1539325939929; Thu, 11 Oct 2018 23:32:19 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1539325939; cv=none; d=google.com; s=arc-20160816; b=b45yZT1sz5fAMEH5IQXPE00nEDitmnYlDpXuXH3neReoFpx6GPhgTyqGP29NCxtgUh HFONJvFupzN18mxTEDphSUDTe1NgqhqVRbBGc7Xy+yoWGAw+KzZzXa93Zk33LUCwbI1G Ie/cTh2CFT/ci6EPFkkfvP5l8D3BON9QYMxx2Sr1BJ87+M3e+eWlrI0+ePWfkjALcW1h FdiO/DZokSgwm3leNjGVNeOK1NsyFwNU5F+0Ui4GzbetQeGj16uux0FzIlQxb2QXYX7j 8ogNtsdJCALQ8b/kluNC1QagQ5tLO25EHt3tO1UrSkIz6Kp2l1rrcIUEFbNejg3quN6l pphQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:list-subscribe:list-help:list-post:list-archive :list-unsubscribe:list-id:precedence:subject:references:in-reply-to :message-id:date:cc:to:from; bh=nQSfvXNPC8/ykQVgy11ixDjPBxf91CKHTsW0WDQrA1g=; b=VB0QwjcE2AGrLc5G/tkGjTQ7fsWZnvmF8Pacydoa45cI5YVxNFqCe6WK9ax9FBYZWP G/uw8EDlFHNxiJ0yTpAihNqcMNMGtLoKQVnZ0JZEOsF6FDrWczjC6Uu3hRrzVj6pSLxh akHpF3T62TrDlueVisXl2pnTnstK2w+rBJ7otWVLTPN86YhJs7DuezdFuxPZpWmVFpW1 ERjbvwJq96C4WLyYu9II1wZyA7AJmR7pYDk7fcxg8yEq8D5Ih+tsupqR4wwp3SnB0PUd hKqFu5QGG+81I/xbtb26fK+LvefUncLkDYKt1QhE2HveJ85B7gWlLvK9doC9qtpBdqp+ iPnA== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Return-Path: Received: from dpdk.org (dpdk.org. [92.243.14.124]) by mx.google.com with ESMTP id r15-v6si561058edi.91.2018.10.11.23.32.19; Thu, 11 Oct 2018 23:32:19 -0700 (PDT) Received-SPF: pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) client-ip=92.243.14.124; Authentication-Results: mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id 5B2746833; Fri, 12 Oct 2018 08:32:12 +0200 (CEST) Received: from foss.arm.com (usa-sjc-mx-foss1.foss.arm.com [217.140.101.70]) by dpdk.org (Postfix) with ESMTP id D0CC95F4A for ; Fri, 12 Oct 2018 08:32:08 +0200 (CEST) Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.72.51.249]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id 1DF3EEBD; Thu, 11 Oct 2018 23:32:08 -0700 (PDT) Received: from 2p2660v4-1.austin.arm.com (2p2660v4-1.austin.arm.com [10.118.12.190]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id AD77B3F5B3; Thu, 11 Oct 2018 23:32:07 -0700 (PDT) From: Honnappa Nagarahalli To: bruce.richardson@intel.com, pablo.de.lara.guarch@intel.com Cc: dev@dpdk.org, yipeng1.wang@intel.com, honnappa.nagarahalli@arm.com, dharmik.thakkar@arm.com, gavin.hu@arm.com, nd@arm.com Date: Fri, 12 Oct 2018 01:31:52 -0500 Message-Id: <1539325918-125438-2-git-send-email-honnappa.nagarahalli@arm.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> References: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> Subject: [dpdk-dev] [PATCH v3 1/7] hash: separate multi-writer from rw-concurrency X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" RW concurrency is required with single writer and multiple reader usecase as well. Hence, multi-writer should not be enabled by default when RW concurrency is enabled. Fixes: f2e3001b53ec ("hash: support read/write concurrency") Cc: yipeng1.wang@intel.com Signed-off-by: Honnappa Nagarahalli Reviewed-by: Gavin Hu --- lib/librte_hash/rte_cuckoo_hash.c | 27 ++++++++++++++++----------- lib/librte_hash/rte_cuckoo_hash.h | 2 ++ test/test/test_hash_readwrite.c | 6 ++++-- 3 files changed, 22 insertions(+), 13 deletions(-) -- 2.7.4 diff --git a/lib/librte_hash/rte_cuckoo_hash.c b/lib/librte_hash/rte_cuckoo_hash.c index 750caf8..7bbe9e9 100644 --- a/lib/librte_hash/rte_cuckoo_hash.c +++ b/lib/librte_hash/rte_cuckoo_hash.c @@ -139,6 +139,7 @@ rte_hash_create(const struct rte_hash_parameters *params) unsigned int hw_trans_mem_support = 0, multi_writer_support = 0; unsigned int ext_table_support = 0; unsigned int readwrite_concur_support = 0; + unsigned int writer_takes_lock = 0; rte_hash_function default_hash_func = (rte_hash_function)rte_jhash; @@ -162,12 +163,14 @@ rte_hash_create(const struct rte_hash_parameters *params) if (params->extra_flag & RTE_HASH_EXTRA_FLAGS_TRANS_MEM_SUPPORT) hw_trans_mem_support = 1; - if (params->extra_flag & RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD) + if (params->extra_flag & RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD) { multi_writer_support = 1; + writer_takes_lock = 1; + } if (params->extra_flag & RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY) { readwrite_concur_support = 1; - multi_writer_support = 1; + writer_takes_lock = 1; } if (params->extra_flag & RTE_HASH_EXTRA_FLAGS_EXT_TABLE) @@ -355,6 +358,7 @@ rte_hash_create(const struct rte_hash_parameters *params) h->multi_writer_support = multi_writer_support; h->readwrite_concur_support = readwrite_concur_support; h->ext_table_support = ext_table_support; + h->writer_takes_lock = writer_takes_lock; #if defined(RTE_ARCH_X86) if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_SSE2)) @@ -363,10 +367,11 @@ rte_hash_create(const struct rte_hash_parameters *params) #endif h->sig_cmp_fn = RTE_HASH_COMPARE_SCALAR; - /* Turn on multi-writer only with explicit flag from user and TM - * support. + /* Writer threads need to take the lock when: + * 1) RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY is enabled OR + * 2) RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD is enabled */ - if (h->multi_writer_support) { + if (h->writer_takes_lock) { h->readwrite_lock = rte_malloc(NULL, sizeof(rte_rwlock_t), RTE_CACHE_LINE_SIZE); if (h->readwrite_lock == NULL) @@ -425,10 +430,10 @@ rte_hash_free(struct rte_hash *h) rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK); - if (h->multi_writer_support) { + if (h->multi_writer_support) rte_free(h->local_free_slots); + if (h->writer_takes_lock) rte_free(h->readwrite_lock); - } rte_ring_free(h->free_slots); rte_ring_free(h->free_ext_bkts); rte_free(h->key_store); @@ -473,9 +478,9 @@ rte_hash_count(const struct rte_hash *h) static inline void __hash_rw_writer_lock(const struct rte_hash *h) { - if (h->multi_writer_support && h->hw_trans_mem_support) + if (h->writer_takes_lock && h->hw_trans_mem_support) rte_rwlock_write_lock_tm(h->readwrite_lock); - else if (h->multi_writer_support) + else if (h->writer_takes_lock) rte_rwlock_write_lock(h->readwrite_lock); } @@ -491,9 +496,9 @@ __hash_rw_reader_lock(const struct rte_hash *h) static inline void __hash_rw_writer_unlock(const struct rte_hash *h) { - if (h->multi_writer_support && h->hw_trans_mem_support) + if (h->writer_takes_lock && h->hw_trans_mem_support) rte_rwlock_write_unlock_tm(h->readwrite_lock); - else if (h->multi_writer_support) + else if (h->writer_takes_lock) rte_rwlock_write_unlock(h->readwrite_lock); } diff --git a/lib/librte_hash/rte_cuckoo_hash.h b/lib/librte_hash/rte_cuckoo_hash.h index 7753cd8..1137651 100644 --- a/lib/librte_hash/rte_cuckoo_hash.h +++ b/lib/librte_hash/rte_cuckoo_hash.h @@ -166,6 +166,8 @@ struct rte_hash { uint8_t readwrite_concur_support; /**< If read-write concurrency support is enabled */ uint8_t ext_table_support; /**< Enable extendable bucket table */ + uint8_t writer_takes_lock; + /**< Indicates if the writer threads need to take lock */ rte_hash_function hash_func; /**< Function used to calculate hash. */ uint32_t hash_func_init_val; /**< Init value used by hash_func. */ rte_hash_cmp_eq_t rte_hash_custom_cmp_eq; diff --git a/test/test/test_hash_readwrite.c b/test/test/test_hash_readwrite.c index 2a4f7b9..a8fadd0 100644 --- a/test/test/test_hash_readwrite.c +++ b/test/test/test_hash_readwrite.c @@ -122,10 +122,12 @@ init_params(int use_htm, int use_jhash) if (use_htm) hash_params.extra_flag = RTE_HASH_EXTRA_FLAGS_TRANS_MEM_SUPPORT | - RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY; + RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY | + RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD; else hash_params.extra_flag = - RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY; + RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY | + RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD; hash_params.name = "tests"; From patchwork Fri Oct 12 06:31:53 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Honnappa Nagarahalli X-Patchwork-Id: 148717 Delivered-To: patch@linaro.org Received: by 2002:a2e:8595:0:0:0:0:0 with SMTP id b21-v6csp368132lji; Thu, 11 Oct 2018 23:32:28 -0700 (PDT) X-Google-Smtp-Source: ACcGV63Xlk2vEPnWcm6uHnHkv0oNKQIa6wRlT7Oxs3qdisPvRsmsGHsujQEl0WBeis+o2OHDlSrM X-Received: by 2002:a50:ba89:: with SMTP id x9-v6mr914718ede.191.1539325948147; Thu, 11 Oct 2018 23:32:28 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1539325948; cv=none; d=google.com; s=arc-20160816; b=D1HbICE9hYvekBwhBfppyGCwT8j9URQilVIbWjGXdXIn1wfpkqco9+leqfm/XH0bN1 KoRxd9Yi8CNHwZtvQy2ekFc44WF2+Tj7TppVl5rP/3ZSv3vP9M6fyxId7FIUHxoh6Ffv 1WwMMYeCQeBiepmxIbn1J3+khNw4MwT+ZqI7IRDFMHKNSpaqzng32Mojcs+x0QFRG03a 5KMRRenmIfXxGSJALsXfbngHVNcDIPj92t5ByiuGb45d9rDR/Wvx5GTJKpacfu1ILQvx lge2BpgPyx5RQoJ+X+W+A5jjys07j5BTtgKNzl33csyAf6ZdfS0VFvR0vK2t6B9uG5Uk /1MA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:list-subscribe:list-help:list-post:list-archive :list-unsubscribe:list-id:precedence:subject:references:in-reply-to :message-id:date:cc:to:from; bh=CTibyfiQL2TMWTZVLgFTxWKuLQtM+EEjwgMbfrP8Li4=; b=pB0qIN9zMm99E+wBH/J05AerzIqe0Nh1F/wEZ2KWkeMjI5pke1sDHfmoagvGGlwmF2 4gpVNcmeei+P7ezszCsDimnLKehH2tVtSKHCVISLJfura8+jioKJBR1Jd9IuNLeoeAu/ UUvdw2y+5Bt0H9ByA6fXuSmvCe54nTQ296iSoalFCMCp45e1ovJ/Xsj9pySjbrScT0oF foMYijix7qAvHY71/06TiN8XXaXdUuj0+mgQo+RHkzbjHEuLh0BncdUlAkc6w8xsdVZC SzYYpRtkZKiQCo/n+HjMbVZsYDijbJ0+epx0lqDE7QEo2PSY7Letn4V8AREgmyCmOEJz YLiQ== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Return-Path: Received: from dpdk.org (dpdk.org. [92.243.14.124]) by mx.google.com with ESMTP id k8-v6si409974ejg.208.2018.10.11.23.32.27; Thu, 11 Oct 2018 23:32:28 -0700 (PDT) Received-SPF: pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) client-ip=92.243.14.124; Authentication-Results: mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id AC9361B110; Fri, 12 Oct 2018 08:32:14 +0200 (CEST) Received: from foss.arm.com (foss.arm.com [217.140.101.70]) by dpdk.org (Postfix) with ESMTP id D1E776833 for ; Fri, 12 Oct 2018 08:32:09 +0200 (CEST) Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.72.51.249]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id 4601F15BF; Thu, 11 Oct 2018 23:32:09 -0700 (PDT) Received: from 2p2660v4-1.austin.arm.com (2p2660v4-1.austin.arm.com [10.118.12.190]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id D401E3F5B3; Thu, 11 Oct 2018 23:32:08 -0700 (PDT) From: Honnappa Nagarahalli To: bruce.richardson@intel.com, pablo.de.lara.guarch@intel.com Cc: dev@dpdk.org, yipeng1.wang@intel.com, honnappa.nagarahalli@arm.com, dharmik.thakkar@arm.com, gavin.hu@arm.com, nd@arm.com Date: Fri, 12 Oct 2018 01:31:53 -0500 Message-Id: <1539325918-125438-3-git-send-email-honnappa.nagarahalli@arm.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> References: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> Subject: [dpdk-dev] [PATCH v3 2/7] hash: support do not recycle on delete X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" rte_hash_lookup_xxx APIs return the index of the element in the key store. Application(reader) can use that index to reference other data structures in its scope. Because of this, the index should not be recycled till the application completes using the index. RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL is introduced to support this. When this flag is enabled rte_hash_del_xxx APIs do not free the key-store index/internal memory associated with the deleted entry. The new API rte_hash_free_key_with_position should be called to free the key-store index/internal memory after calling rte_hash_del_xxx APIs. Suggested-by: Yipeng Wang Signed-off-by: Honnappa Nagarahalli Reviewed-by: Gavin Hu --- lib/librte_hash/rte_cuckoo_hash.c | 50 +++++++++++++- lib/librte_hash/rte_cuckoo_hash.h | 8 +++ lib/librte_hash/rte_hash.h | 40 +++++++++++ test/test/test_hash.c | 140 +++++++++++++++++++++++++++++++++++++- 4 files changed, 234 insertions(+), 4 deletions(-) -- 2.7.4 Reviewed-by: Yipeng Wang diff --git a/lib/librte_hash/rte_cuckoo_hash.c b/lib/librte_hash/rte_cuckoo_hash.c index 7bbe9e9..8de0de3 100644 --- a/lib/librte_hash/rte_cuckoo_hash.c +++ b/lib/librte_hash/rte_cuckoo_hash.c @@ -140,6 +140,7 @@ rte_hash_create(const struct rte_hash_parameters *params) unsigned int ext_table_support = 0; unsigned int readwrite_concur_support = 0; unsigned int writer_takes_lock = 0; + unsigned int recycle_on_del = 1; rte_hash_function default_hash_func = (rte_hash_function)rte_jhash; @@ -176,6 +177,9 @@ rte_hash_create(const struct rte_hash_parameters *params) if (params->extra_flag & RTE_HASH_EXTRA_FLAGS_EXT_TABLE) ext_table_support = 1; + if (params->extra_flag & RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL) + recycle_on_del = 0; + /* Store all keys and leave the first entry as a dummy entry for lookup_bulk */ if (multi_writer_support) /* @@ -359,6 +363,7 @@ rte_hash_create(const struct rte_hash_parameters *params) h->readwrite_concur_support = readwrite_concur_support; h->ext_table_support = ext_table_support; h->writer_takes_lock = writer_takes_lock; + h->recycle_on_del = recycle_on_del; #if defined(RTE_ARCH_X86) if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_SSE2)) @@ -1121,7 +1126,6 @@ remove_entry(const struct rte_hash *h, struct rte_hash_bucket *bkt, unsigned i) unsigned lcore_id, n_slots; struct lcore_cache *cached_free_slots; - bkt->sig_current[i] = NULL_SIGNATURE; if (h->multi_writer_support) { lcore_id = rte_lcore_id(); cached_free_slots = &h->local_free_slots[lcore_id]; @@ -1183,7 +1187,12 @@ search_and_remove(const struct rte_hash *h, const void *key, k = (struct rte_hash_key *) ((char *)keys + bkt->key_idx[i] * h->key_entry_size); if (rte_hash_cmp_eq(key, k->key, h) == 0) { - remove_entry(h, bkt, i); + bkt->sig_current[i] = NULL_SIGNATURE; + /* Do not free the key store element if + * recycle_on_del is disabled. + */ + if (h->recycle_on_del) + remove_entry(h, bkt, i); /* Return index where key is stored, * subtracting the first dummy index @@ -1301,6 +1310,43 @@ rte_hash_get_key_with_position(const struct rte_hash *h, const int32_t position, return 0; } +int __rte_experimental +rte_hash_free_key_with_position(const struct rte_hash *h, + const int32_t position) +{ + RETURN_IF_TRUE(((h == NULL) || (position == EMPTY_SLOT)), -EINVAL); + + unsigned int lcore_id, n_slots; + struct lcore_cache *cached_free_slots; + const int32_t total_entries = h->num_buckets * RTE_HASH_BUCKET_ENTRIES; + + /* Out of bounds */ + if (position >= total_entries) + return -EINVAL; + + if (h->multi_writer_support) { + lcore_id = rte_lcore_id(); + cached_free_slots = &h->local_free_slots[lcore_id]; + /* Cache full, need to free it. */ + if (cached_free_slots->len == LCORE_CACHE_SIZE) { + /* Need to enqueue the free slots in global ring. */ + n_slots = rte_ring_mp_enqueue_burst(h->free_slots, + cached_free_slots->objs, + LCORE_CACHE_SIZE, NULL); + cached_free_slots->len -= n_slots; + } + /* Put index of new free slot in cache. */ + cached_free_slots->objs[cached_free_slots->len] = + (void *)((uintptr_t)position); + cached_free_slots->len++; + } else { + rte_ring_sp_enqueue(h->free_slots, + (void *)((uintptr_t)position)); + } + + return 0; +} + static inline void compare_signatures(uint32_t *prim_hash_matches, uint32_t *sec_hash_matches, const struct rte_hash_bucket *prim_bkt, diff --git a/lib/librte_hash/rte_cuckoo_hash.h b/lib/librte_hash/rte_cuckoo_hash.h index 1137651..5225402 100644 --- a/lib/librte_hash/rte_cuckoo_hash.h +++ b/lib/librte_hash/rte_cuckoo_hash.h @@ -166,6 +166,14 @@ struct rte_hash { uint8_t readwrite_concur_support; /**< If read-write concurrency support is enabled */ uint8_t ext_table_support; /**< Enable extendable bucket table */ + uint8_t recycle_on_del; + /**< If internal memory/key-store entry should be + * freed on calling the rte_hash_del_xxx APIs. + * If this is set, rte_hash_free_key_with_position must be + * called to free the internal memory associated with + * the deleted entry. + * This flag is enabled by default. + */ uint8_t writer_takes_lock; /**< Indicates if the writer threads need to take lock */ rte_hash_function hash_func; /**< Function used to calculate hash. */ diff --git a/lib/librte_hash/rte_hash.h b/lib/librte_hash/rte_hash.h index 6ace64e..3e5d336 100644 --- a/lib/librte_hash/rte_hash.h +++ b/lib/librte_hash/rte_hash.h @@ -14,6 +14,8 @@ #include #include +#include + #ifdef __cplusplus extern "C" { #endif @@ -40,6 +42,11 @@ extern "C" { /** Flag to indicate the extendabe bucket table feature should be used */ #define RTE_HASH_EXTRA_FLAGS_EXT_TABLE 0x08 +/** Flag to disable freeing of internal memory/indices on hash delete. + * Refer to rte_hash_del_xxx APIs for more details. + */ +#define RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL 0x10 + /** * The type of hash value of a key. * It should be a value of at least 32bit with fully random pattern. @@ -236,6 +243,10 @@ rte_hash_add_key_with_hash(const struct rte_hash *h, const void *key, hash_sig_t * and should only be called from one thread by default. * Thread safety can be enabled by setting flag during * table creation. + * If RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL is enabled, + * the hash library's internal memory/index will not be freed by this + * API. rte_hash_free_key_with_position API must be called additionally + * to free the internal memory/index associated with the key. * * @param h * Hash table to remove the key from. @@ -257,6 +268,10 @@ rte_hash_del_key(const struct rte_hash *h, const void *key); * and should only be called from one thread by default. * Thread safety can be enabled by setting flag during * table creation. + * If RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL is enabled, + * the hash library's internal memory/index will not be freed by this + * API. rte_hash_free_key_with_position API must be called additionally + * to free the internal memory/index associated with the key. * * @param h * Hash table to remove the key from. @@ -296,6 +311,31 @@ rte_hash_get_key_with_position(const struct rte_hash *h, const int32_t position, void **key); /** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Free hash library's internal memory/index given the position + * of the key. This operation is not multi-thread safe and should + * only be called from one thread by default. Thread safety + * can be enabled by setting flag during table creation. + * If RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL is enabled, + * the hash library's internal memory/index must be freed using this API + * after the key is deleted using rte_hash_del_key_xxx APIs. + * This API does not validate if the key is already freed. + * + * @param h + * Hash table to free the key from. + * @param position + * Position returned when the key was deleted. + * @return + * - 0 if freed successfully + * - -EINVAL if the parameters are invalid. + */ +int __rte_experimental +rte_hash_free_key_with_position(const struct rte_hash *h, + const int32_t position); + +/** * Find a key-value pair in the hash table. * This operation is multi-thread safe with regarding to other lookup threads. * Read-write concurrency can be enabled by setting flag during diff --git a/test/test/test_hash.c b/test/test/test_hash.c index 815c734..19131e9 100644 --- a/test/test/test_hash.c +++ b/test/test/test_hash.c @@ -260,6 +260,13 @@ static void run_hash_func_tests(void) * - lookup (hit) * - delete * - lookup (miss) + * + * Repeat the test case when 'free on delete' is disabled. + * - add + * - lookup (hit) + * - delete + * - lookup (miss) + * - free */ static int test_add_delete(void) { @@ -295,10 +302,12 @@ static int test_add_delete(void) /* repeat test with precomputed hash functions */ hash_sig_t hash_value; - int pos1, expectedPos1; + int pos1, expectedPos1, delPos1; + ut_params.extra_flag = RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL; handle = rte_hash_create(&ut_params); RETURN_IF_ERROR(handle == NULL, "hash creation failed"); + ut_params.extra_flag = 0; hash_value = rte_hash_hash(handle, &keys[0]); pos1 = rte_hash_add_key_with_hash(handle, &keys[0], hash_value); @@ -315,12 +324,18 @@ static int test_add_delete(void) print_key_info("Del", &keys[0], pos1); RETURN_IF_ERROR(pos1 != expectedPos1, "failed to delete key (pos1=%d)", pos1); + delPos1 = pos1; pos1 = rte_hash_lookup_with_hash(handle, &keys[0], hash_value); print_key_info("Lkp", &keys[0], pos1); RETURN_IF_ERROR(pos1 != -ENOENT, "fail: found key after deleting! (pos1=%d)", pos1); + pos1 = rte_hash_free_key_with_position(handle, delPos1); + print_key_info("Free", &keys[0], delPos1); + RETURN_IF_ERROR(pos1 != 0, + "failed to free key (pos1=%d)", delPos1); + rte_hash_free(handle); return 0; @@ -391,6 +406,84 @@ static int test_add_update_delete(void) } /* + * Sequence of operations for a single key with 'disable free on del' set: + * - delete: miss + * - add + * - lookup: hit + * - add: update + * - lookup: hit (updated data) + * - delete: hit + * - delete: miss + * - lookup: miss + * - free: hit + * - lookup: miss + */ +static int test_add_update_delete_free(void) +{ + struct rte_hash *handle; + int pos0, expectedPos0, delPos0, result; + + ut_params.name = "test2"; + ut_params.extra_flag = RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL; + handle = rte_hash_create(&ut_params); + RETURN_IF_ERROR(handle == NULL, "hash creation failed"); + ut_params.extra_flag = 0; + + pos0 = rte_hash_del_key(handle, &keys[0]); + print_key_info("Del", &keys[0], pos0); + RETURN_IF_ERROR(pos0 != -ENOENT, + "fail: found non-existent key (pos0=%d)", pos0); + + pos0 = rte_hash_add_key(handle, &keys[0]); + print_key_info("Add", &keys[0], pos0); + RETURN_IF_ERROR(pos0 < 0, "failed to add key (pos0=%d)", pos0); + expectedPos0 = pos0; + + pos0 = rte_hash_lookup(handle, &keys[0]); + print_key_info("Lkp", &keys[0], pos0); + RETURN_IF_ERROR(pos0 != expectedPos0, + "failed to find key (pos0=%d)", pos0); + + pos0 = rte_hash_add_key(handle, &keys[0]); + print_key_info("Add", &keys[0], pos0); + RETURN_IF_ERROR(pos0 != expectedPos0, + "failed to re-add key (pos0=%d)", pos0); + + pos0 = rte_hash_lookup(handle, &keys[0]); + print_key_info("Lkp", &keys[0], pos0); + RETURN_IF_ERROR(pos0 != expectedPos0, + "failed to find key (pos0=%d)", pos0); + + delPos0 = rte_hash_del_key(handle, &keys[0]); + print_key_info("Del", &keys[0], delPos0); + RETURN_IF_ERROR(delPos0 != expectedPos0, + "failed to delete key (pos0=%d)", delPos0); + + pos0 = rte_hash_del_key(handle, &keys[0]); + print_key_info("Del", &keys[0], pos0); + RETURN_IF_ERROR(pos0 != -ENOENT, + "fail: deleted already deleted key (pos0=%d)", pos0); + + pos0 = rte_hash_lookup(handle, &keys[0]); + print_key_info("Lkp", &keys[0], pos0); + RETURN_IF_ERROR(pos0 != -ENOENT, + "fail: found key after deleting! (pos0=%d)", pos0); + + result = rte_hash_free_key_with_position(handle, delPos0); + print_key_info("Free", &keys[0], delPos0); + RETURN_IF_ERROR(result != 0, + "failed to free key (pos1=%d)", delPos0); + + pos0 = rte_hash_lookup(handle, &keys[0]); + print_key_info("Lkp", &keys[0], pos0); + RETURN_IF_ERROR(pos0 != -ENOENT, + "fail: found key after deleting! (pos0=%d)", pos0); + + rte_hash_free(handle); + return 0; +} + +/* * Sequence of operations for retrieving a key with its position * * - create table @@ -399,11 +492,20 @@ static int test_add_update_delete(void) * - delete key * - try to get the deleted key: miss * + * Repeat the test case when 'free on delete' is disabled. + * - create table + * - add key + * - get the key with its position: hit + * - delete key + * - try to get the deleted key: hit + * - free key + * - try to get the deleted key: miss + * */ static int test_hash_get_key_with_position(void) { struct rte_hash *handle = NULL; - int pos, expectedPos, result; + int pos, expectedPos, delPos, result; void *key; ut_params.name = "hash_get_key_w_pos"; @@ -427,6 +529,38 @@ static int test_hash_get_key_with_position(void) RETURN_IF_ERROR(result != -ENOENT, "non valid key retrieved"); rte_hash_free(handle); + + ut_params.name = "hash_get_key_w_pos"; + ut_params.extra_flag = RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL; + handle = rte_hash_create(&ut_params); + RETURN_IF_ERROR(handle == NULL, "hash creation failed"); + ut_params.extra_flag = 0; + + pos = rte_hash_add_key(handle, &keys[0]); + print_key_info("Add", &keys[0], pos); + RETURN_IF_ERROR(pos < 0, "failed to add key (pos0=%d)", pos); + expectedPos = pos; + + result = rte_hash_get_key_with_position(handle, pos, &key); + RETURN_IF_ERROR(result != 0, "error retrieving a key"); + + delPos = rte_hash_del_key(handle, &keys[0]); + print_key_info("Del", &keys[0], delPos); + RETURN_IF_ERROR(delPos != expectedPos, + "failed to delete key (pos0=%d)", delPos); + + result = rte_hash_get_key_with_position(handle, delPos, &key); + RETURN_IF_ERROR(result != -ENOENT, "non valid key retrieved"); + + result = rte_hash_free_key_with_position(handle, delPos); + print_key_info("Free", &keys[0], delPos); + RETURN_IF_ERROR(result != 0, + "failed to free key (pos1=%d)", delPos); + + result = rte_hash_get_key_with_position(handle, delPos, &key); + RETURN_IF_ERROR(result != -ENOENT, "non valid key retrieved"); + + rte_hash_free(handle); return 0; } @@ -1609,6 +1743,8 @@ test_hash(void) return -1; if (test_add_update_delete() < 0) return -1; + if (test_add_update_delete_free() < 0) + return -1; if (test_five_keys() < 0) return -1; if (test_full_bucket() < 0) From patchwork Fri Oct 12 06:31:54 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Honnappa Nagarahalli X-Patchwork-Id: 148718 Delivered-To: patch@linaro.org Received: by 2002:a2e:8595:0:0:0:0:0 with SMTP id b21-v6csp368230lji; Thu, 11 Oct 2018 23:32:36 -0700 (PDT) X-Google-Smtp-Source: ACcGV60/mS2Y9x2KaDfu7f+e83aOUUlbC5Jn/fCc0KaGQ1sVa1g5WpvDv8UemKEiA1k0nKyv2dta X-Received: by 2002:a50:a6d5:: with SMTP id f21-v6mr6893809edc.225.1539325956867; Thu, 11 Oct 2018 23:32:36 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1539325956; cv=none; d=google.com; s=arc-20160816; b=je7EWgPHprBzB2XLvEyuvSJSsy4nJyIqzXHKs8pCzXn74kFqqovYULSS4V/t9Scmym 3IbVV0nk8LAFq6gAi9rhsDjCnPhCq/ryHde0xcJZw8Cn1j/hwl4plnlKICYYpAqXN4yu JwEzJZXXb4tuoQ/o40e4W47x2HhI3F5RuoEhMKQcMhkapb0SQIdWnufmcd77Yds4V8AT ztofTJn9ow9BeugEw61pYSdAsVJAcKWFhebixWGstJEFcVU8Ls4a8Q2XlHqi3P3miGNX KbSisjBUD4S/h2cKDEAejnPh9p16UB7IDEtKg1jU6cyazmTTvzQqtnsSrDzQdkedlBkv 66Xw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:list-subscribe:list-help:list-post:list-archive :list-unsubscribe:list-id:precedence:subject:references:in-reply-to :message-id:date:cc:to:from; bh=evikmZaVurXm9knpYgMdLzxCY8RbrJdW+xEjINcXuJw=; b=D4SxrhNSpuf2MwLd3iA9DF9UyJDEySfBc/wyaHyI5qB/nacOKo+wkKWZf28Pn3iC4J TpcYa3+grmYpj9qcNVU58/93HXpn6WN7gAW/lna/UtZ7cHQNnRZXBqdN8AhJl50M8bjN CArCBdJBGd2tfQnN8FRkweQdZoZwiewxo/jdAeKd0paEmSGtHphREizkmy3qqL1dcPea dmgmbnA0j9QtmFErxYZmxR5mgPC3EpklPTJWM5l9hK+rZwLt9EUyxr6ZbYPBy8XCMQqY EOoEuRRm8/BvJv5LLTFCLMvFWGzOm2R+Q97SyWTF1uzrgXQQM+L1tLsgWasicKFkZo4z 1Chw== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Return-Path: Received: from dpdk.org (dpdk.org. [92.243.14.124]) by mx.google.com with ESMTP id h22-v6si503587eda.202.2018.10.11.23.32.36; Thu, 11 Oct 2018 23:32:36 -0700 (PDT) Received-SPF: pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) client-ip=92.243.14.124; Authentication-Results: mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id 1CD8A1B116; Fri, 12 Oct 2018 08:32:17 +0200 (CEST) Received: from foss.arm.com (foss.arm.com [217.140.101.70]) by dpdk.org (Postfix) with ESMTP id 0522C6833 for ; Fri, 12 Oct 2018 08:32:11 +0200 (CEST) Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.72.51.249]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id 7735EEBD; Thu, 11 Oct 2018 23:32:10 -0700 (PDT) Received: from 2p2660v4-1.austin.arm.com (2p2660v4-1.austin.arm.com [10.118.12.190]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 127D63F5B3; Thu, 11 Oct 2018 23:32:10 -0700 (PDT) From: Honnappa Nagarahalli To: bruce.richardson@intel.com, pablo.de.lara.guarch@intel.com Cc: dev@dpdk.org, yipeng1.wang@intel.com, honnappa.nagarahalli@arm.com, dharmik.thakkar@arm.com, gavin.hu@arm.com, nd@arm.com Date: Fri, 12 Oct 2018 01:31:54 -0500 Message-Id: <1539325918-125438-4-git-send-email-honnappa.nagarahalli@arm.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> References: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> Subject: [dpdk-dev] [PATCH v3 3/7] hash: correct key store element alignment X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" Correct the key store array element alignment. This is required to make 'pdata' in 'struct rte_hash_key' align on the correct boundary. Signed-off-by: Honnappa Nagarahalli Reviewed-by: Gavin Hu Reviewed-by: Ola Liljedahl Reviewed-by: Steve Capper --- lib/librte_hash/rte_cuckoo_hash.c | 4 +++- lib/librte_hash/rte_cuckoo_hash.h | 2 +- 2 files changed, 4 insertions(+), 2 deletions(-) -- 2.7.4 Reviewed-by: Yipeng Wang diff --git a/lib/librte_hash/rte_cuckoo_hash.c b/lib/librte_hash/rte_cuckoo_hash.c index 8de0de3..13646d0 100644 --- a/lib/librte_hash/rte_cuckoo_hash.c +++ b/lib/librte_hash/rte_cuckoo_hash.c @@ -279,7 +279,9 @@ rte_hash_create(const struct rte_hash_parameters *params) rte_ring_sp_enqueue(r_ext, (void *)((uintptr_t) i)); } - const uint32_t key_entry_size = sizeof(struct rte_hash_key) + params->key_len; + const uint32_t key_entry_size = + RTE_ALIGN(sizeof(struct rte_hash_key) + params->key_len, + KEY_ALIGNMENT); const uint64_t key_tbl_size = (uint64_t) key_entry_size * num_key_slots; k = rte_zmalloc_socket(NULL, key_tbl_size, diff --git a/lib/librte_hash/rte_cuckoo_hash.h b/lib/librte_hash/rte_cuckoo_hash.h index 5225402..b823d37 100644 --- a/lib/librte_hash/rte_cuckoo_hash.h +++ b/lib/librte_hash/rte_cuckoo_hash.h @@ -123,7 +123,7 @@ struct rte_hash_key { }; /* Variable key size */ char key[0]; -} __attribute__((aligned(KEY_ALIGNMENT))); +}; /* All different signature compare functions */ enum rte_hash_sig_compare_function { From patchwork Fri Oct 12 06:31:55 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Honnappa Nagarahalli X-Patchwork-Id: 148719 Delivered-To: patch@linaro.org Received: by 2002:a2e:8595:0:0:0:0:0 with SMTP id b21-v6csp368329lji; Thu, 11 Oct 2018 23:32:44 -0700 (PDT) X-Google-Smtp-Source: ACcGV63G7sSBuPtOf76Hy+0pCP3NEhLR7rk0R3N118ANaXTZzIOLIjBl1Ot4mNjTJ42hhEcUrki7 X-Received: by 2002:a17:906:4054:: with SMTP id y20-v6mr5667358ejj.30.1539325964301; Thu, 11 Oct 2018 23:32:44 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1539325964; cv=none; d=google.com; s=arc-20160816; b=zUp6c3BqgZ4I9d2BIIry1dztf0L6ZMfslVqDir7hlfngl+COrKfhzRlEBhWQp0Z2/5 1NK1Z4CSN+uZOBrP1gmrY1pf7EREkpGV2gQSRHar5qorMumEvSxDKZUS0kHuC3R6RG4L HgVDgnIN89qAlJ6wIolFpDOOnrPXnlmmIzTEg5A1WlTaoi6aVmalzjtzy4hWT8W5tRdP IgddC8vYg3j5zfx8r+DUnYA7FdI1afw2NkRi9O+p1uSD2NzWncdD/MOzbB2EM1mwdQ6r AQy9DgRGuMdyzCnfi6uh5xbjt9MEcprZUlNWIL4jE9GYNZdrFB4ELQR01VTdSuXv1P6a SaTQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:list-subscribe:list-help:list-post:list-archive :list-unsubscribe:list-id:precedence:subject:references:in-reply-to :message-id:date:cc:to:from; bh=qWhOBAlssDRz487JVtPXhVqwxkb7ne5CMna0DZ6fHww=; b=GNhRmwxT5FidCIv8lPqk2Cttz4aVbZwTJdW87onHo7lDagmUHwCp8Exx14+LdEkeKt BQXXSlAIsxfhS+4REL6MK+NnuTQKsSsIPJCO591MwPIHst9168hrFSkUMpDf8BQD9rYn FVWqN/SYZNq9F6Wqqnwy51ELRd/OEANjvf2m8dZiQHr2fYFWJp5vTLanHqrehZnHjxCg cujnS/QgMsRJfndF2Tt2OssqnWsk7CbM7Jyb3/q0FFxiAQUwW5O7215cm4QyQNW8CNLB 90VSvMQbVIoe+bDf3+MpJuWU/qIGGemhpDib31eldEFq0mEe1ZvZ/aEQ5LdqLE/7D14C 3LBA== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Return-Path: Received: from dpdk.org (dpdk.org. [92.243.14.124]) by mx.google.com with ESMTP id k9-v6si591239eds.404.2018.10.11.23.32.44; Thu, 11 Oct 2018 23:32:44 -0700 (PDT) Received-SPF: pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) client-ip=92.243.14.124; Authentication-Results: mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id F17671B122; Fri, 12 Oct 2018 08:32:18 +0200 (CEST) Received: from foss.arm.com (foss.arm.com [217.140.101.70]) by dpdk.org (Postfix) with ESMTP id 3BAF41B103 for ; Fri, 12 Oct 2018 08:32:12 +0200 (CEST) Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.72.51.249]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id A055C15BF; Thu, 11 Oct 2018 23:32:11 -0700 (PDT) Received: from 2p2660v4-1.austin.arm.com (2p2660v4-1.austin.arm.com [10.118.12.190]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 3B2D13F5B3; Thu, 11 Oct 2018 23:32:11 -0700 (PDT) From: Honnappa Nagarahalli To: bruce.richardson@intel.com, pablo.de.lara.guarch@intel.com Cc: dev@dpdk.org, yipeng1.wang@intel.com, honnappa.nagarahalli@arm.com, dharmik.thakkar@arm.com, gavin.hu@arm.com, nd@arm.com Date: Fri, 12 Oct 2018 01:31:55 -0500 Message-Id: <1539325918-125438-5-git-send-email-honnappa.nagarahalli@arm.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> References: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> Subject: [dpdk-dev] [PATCH v3 4/7] hash: add memory ordering to avoid race conditions X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" Only race condition that can occur is - using the key store element before the key write is completed. Hence, while inserting the element the release memory order is used. Any other race condition is caught by the key comparison. Memory orderings are added only where needed. For ex: reads in the writer's context do not need memory ordering as there is a single writer. key_idx in the bucket entry and pdata in the key store element are used for synchronisation. key_idx is used to release an inserted entry in the bucket to the reader. Use of pdata for synchronisation is required due to updation of an existing entry where-in only the pdata is updated without updating key_idx. Signed-off-by: Honnappa Nagarahalli Reviewed-by: Gavin Hu Reviewed-by: Ola Liljedahl Reviewed-by: Steve Capper Reviewed-by: Yipeng Wang --- lib/librte_hash/rte_cuckoo_hash.c | 116 ++++++++++++++++++++++++++++---------- 1 file changed, 87 insertions(+), 29 deletions(-) -- 2.7.4 diff --git a/lib/librte_hash/rte_cuckoo_hash.c b/lib/librte_hash/rte_cuckoo_hash.c index 13646d0..e55725a 100644 --- a/lib/librte_hash/rte_cuckoo_hash.c +++ b/lib/librte_hash/rte_cuckoo_hash.c @@ -1,5 +1,6 @@ /* SPDX-License-Identifier: BSD-3-Clause * Copyright(c) 2010-2016 Intel Corporation + * Copyright(c) 2018 Arm Limited */ #include @@ -585,7 +586,9 @@ enqueue_slot_back(const struct rte_hash *h, rte_ring_sp_enqueue(h->free_slots, slot_id); } -/* Search a key from bucket and update its data */ +/* Search a key from bucket and update its data. + * Writer holds the lock before calling this. + */ static inline int32_t search_and_update(const struct rte_hash *h, void *data, const void *key, struct rte_hash_bucket *bkt, uint16_t sig) @@ -598,8 +601,13 @@ search_and_update(const struct rte_hash *h, void *data, const void *key, k = (struct rte_hash_key *) ((char *)keys + bkt->key_idx[i] * h->key_entry_size); if (rte_hash_cmp_eq(key, k->key, h) == 0) { - /* Update data */ - k->pdata = data; + /* 'pdata' acts as the synchronization point + * when an existing hash entry is updated. + * Key is not updated in this case. + */ + __atomic_store_n(&k->pdata, + data, + __ATOMIC_RELEASE); /* * Return index where key is stored, * subtracting the first dummy index @@ -655,7 +663,15 @@ rte_hash_cuckoo_insert_mw(const struct rte_hash *h, /* Check if slot is available */ if (likely(prim_bkt->key_idx[i] == EMPTY_SLOT)) { prim_bkt->sig_current[i] = sig; - prim_bkt->key_idx[i] = new_idx; + /* Key can be of arbitrary length, so it is + * not possible to store it atomically. + * Hence the new key element's memory stores + * (key as well as data) should be complete + * before it is referenced. + */ + __atomic_store_n(&prim_bkt->key_idx[i], + new_idx, + __ATOMIC_RELEASE); break; } } @@ -739,8 +755,10 @@ rte_hash_cuckoo_move_insert_mw(const struct rte_hash *h, */ curr_bkt->sig_current[curr_slot] = prev_bkt->sig_current[prev_slot]; - curr_bkt->key_idx[curr_slot] = - prev_bkt->key_idx[prev_slot]; + /* Release the updated bucket entry */ + __atomic_store_n(&curr_bkt->key_idx[curr_slot], + prev_bkt->key_idx[prev_slot], + __ATOMIC_RELEASE); curr_slot = prev_slot; curr_node = prev_node; @@ -748,7 +766,10 @@ rte_hash_cuckoo_move_insert_mw(const struct rte_hash *h, } curr_bkt->sig_current[curr_slot] = sig; - curr_bkt->key_idx[curr_slot] = new_idx; + /* Release the new bucket entry */ + __atomic_store_n(&curr_bkt->key_idx[curr_slot], + new_idx, + __ATOMIC_RELEASE); __hash_rw_writer_unlock(h); @@ -889,8 +910,15 @@ __rte_hash_add_key_with_hash(const struct rte_hash *h, const void *key, new_idx = (uint32_t)((uintptr_t) slot_id); /* Copy key */ rte_memcpy(new_k->key, key, h->key_len); - new_k->pdata = data; - + /* Key can be of arbitrary length, so it is not possible to store + * it atomically. Hence the new key element's memory stores + * (key as well as data) should be complete before it is referenced. + * 'pdata' acts as the synchronization point when an existing hash + * entry is updated. + */ + __atomic_store_n(&new_k->pdata, + data, + __ATOMIC_RELEASE); /* Find an empty slot and insert */ ret = rte_hash_cuckoo_insert_mw(h, prim_bkt, sec_bkt, key, data, @@ -1034,21 +1062,27 @@ search_one_bucket(const struct rte_hash *h, const void *key, uint16_t sig, void **data, const struct rte_hash_bucket *bkt) { int i; + uint32_t key_idx; + void *pdata; struct rte_hash_key *k, *keys = h->key_store; for (i = 0; i < RTE_HASH_BUCKET_ENTRIES; i++) { - if (bkt->sig_current[i] == sig && - bkt->key_idx[i] != EMPTY_SLOT) { + key_idx = __atomic_load_n(&bkt->key_idx[i], + __ATOMIC_ACQUIRE); + if (bkt->sig_current[i] == sig && key_idx != EMPTY_SLOT) { k = (struct rte_hash_key *) ((char *)keys + - bkt->key_idx[i] * h->key_entry_size); + key_idx * h->key_entry_size); + pdata = __atomic_load_n(&k->pdata, + __ATOMIC_ACQUIRE); + if (rte_hash_cmp_eq(key, k->key, h) == 0) { if (data != NULL) - *data = k->pdata; + *data = pdata; /* * Return index where key is stored, * subtracting the first dummy index */ - return bkt->key_idx[i] - 1; + return key_idx - 1; } } } @@ -1173,21 +1207,25 @@ __rte_hash_compact_ll(struct rte_hash_bucket *cur_bkt, int pos) { } } -/* Search one bucket and remove the matched key */ +/* Search one bucket and remove the matched key. + * Writer is expected to hold the lock while calling this + * function. + */ static inline int32_t search_and_remove(const struct rte_hash *h, const void *key, struct rte_hash_bucket *bkt, uint16_t sig, int *pos) { struct rte_hash_key *k, *keys = h->key_store; unsigned int i; - int32_t ret; + uint32_t key_idx; /* Check if key is in bucket */ for (i = 0; i < RTE_HASH_BUCKET_ENTRIES; i++) { - if (bkt->sig_current[i] == sig && - bkt->key_idx[i] != EMPTY_SLOT) { + key_idx = __atomic_load_n(&bkt->key_idx[i], + __ATOMIC_ACQUIRE); + if (bkt->sig_current[i] == sig && key_idx != EMPTY_SLOT) { k = (struct rte_hash_key *) ((char *)keys + - bkt->key_idx[i] * h->key_entry_size); + key_idx * h->key_entry_size); if (rte_hash_cmp_eq(key, k->key, h) == 0) { bkt->sig_current[i] = NULL_SIGNATURE; /* Do not free the key store element if @@ -1196,13 +1234,16 @@ search_and_remove(const struct rte_hash *h, const void *key, if (h->recycle_on_del) remove_entry(h, bkt, i); - /* Return index where key is stored, + __atomic_store_n(&bkt->key_idx[i], + EMPTY_SLOT, + __ATOMIC_RELEASE); + + *pos = i; + /* + * Return index where key is stored, * subtracting the first dummy index */ - ret = bkt->key_idx[i] - 1; - bkt->key_idx[i] = EMPTY_SLOT; - *pos = i; - return ret; + return key_idx - 1; } } } @@ -1402,6 +1443,7 @@ __rte_hash_lookup_bulk(const struct rte_hash *h, const void **keys, uint32_t prim_hitmask[RTE_HASH_LOOKUP_BULK_MAX] = {0}; uint32_t sec_hitmask[RTE_HASH_LOOKUP_BULK_MAX] = {0}; struct rte_hash_bucket *cur_bkt, *next_bkt; + void *pdata[RTE_HASH_LOOKUP_BULK_MAX]; /* Prefetch first keys */ for (i = 0; i < PREFETCH_OFFSET && i < num_keys; i++) @@ -1480,18 +1522,25 @@ __rte_hash_lookup_bulk(const struct rte_hash *h, const void **keys, uint32_t hit_index = __builtin_ctzl(prim_hitmask[i]) >> 1; - uint32_t key_idx = primary_bkt[i]->key_idx[hit_index]; + uint32_t key_idx = + __atomic_load_n( + &primary_bkt[i]->key_idx[hit_index], + __ATOMIC_ACQUIRE); const struct rte_hash_key *key_slot = (const struct rte_hash_key *)( (const char *)h->key_store + key_idx * h->key_entry_size); + + if (key_idx != EMPTY_SLOT) + pdata[i] = __atomic_load_n(&key_slot->pdata, + __ATOMIC_ACQUIRE); /* * If key index is 0, do not compare key, * as it is checking the dummy slot */ if (!!key_idx & !rte_hash_cmp_eq(key_slot->key, keys[i], h)) { if (data != NULL) - data[i] = key_slot->pdata; + data[i] = pdata[i]; hits |= 1ULL << i; positions[i] = key_idx - 1; @@ -1504,11 +1553,19 @@ __rte_hash_lookup_bulk(const struct rte_hash *h, const void **keys, uint32_t hit_index = __builtin_ctzl(sec_hitmask[i]) >> 1; - uint32_t key_idx = secondary_bkt[i]->key_idx[hit_index]; + uint32_t key_idx = + __atomic_load_n( + &secondary_bkt[i]->key_idx[hit_index], + __ATOMIC_ACQUIRE); const struct rte_hash_key *key_slot = (const struct rte_hash_key *)( (const char *)h->key_store + key_idx * h->key_entry_size); + + if (key_idx != EMPTY_SLOT) + pdata[i] = __atomic_load_n(&key_slot->pdata, + __ATOMIC_ACQUIRE); + /* * If key index is 0, do not compare key, * as it is checking the dummy slot @@ -1516,7 +1573,7 @@ __rte_hash_lookup_bulk(const struct rte_hash *h, const void **keys, if (!!key_idx & !rte_hash_cmp_eq(key_slot->key, keys[i], h)) { if (data != NULL) - data[i] = key_slot->pdata; + data[i] = pdata[i]; hits |= 1ULL << i; positions[i] = key_idx - 1; @@ -1612,7 +1669,8 @@ rte_hash_iterate(const struct rte_hash *h, const void **key, void **data, uint32 idx = *next % RTE_HASH_BUCKET_ENTRIES; /* If current position is empty, go to the next one */ - while ((position = h->buckets[bucket_idx].key_idx[idx]) == EMPTY_SLOT) { + while ((position = __atomic_load_n(&h->buckets[bucket_idx].key_idx[idx], + __ATOMIC_ACQUIRE)) == EMPTY_SLOT) { (*next)++; /* End of table */ if (*next == total_entries) From patchwork Fri Oct 12 06:31:56 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Honnappa Nagarahalli X-Patchwork-Id: 148720 Delivered-To: patch@linaro.org Received: by 2002:a2e:8595:0:0:0:0:0 with SMTP id b21-v6csp368402lji; Thu, 11 Oct 2018 23:32:51 -0700 (PDT) X-Google-Smtp-Source: ACcGV60OGwpJ0yqVh3SdKIRSG7V+ueDFg+xEmFayXg67n4VYlQ6tQVInXQnAo6l2WPIRcJTzv6cH X-Received: by 2002:a50:8ba2:: with SMTP id m31-v6mr7383247edm.260.1539325971867; Thu, 11 Oct 2018 23:32:51 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1539325971; cv=none; d=google.com; s=arc-20160816; b=DArUsZbMBXCgAERq1fsVygJYohOmK5y89xNkl0VzV+LiTWOgYHQxx8iSE+r5rx2QuN 5870q+xuttINoLsXQHP0sNDgAKcmuWW6kSbjNQKX/Mn83/r7T2o9EqvpdTX4ZbXAOuL+ 9rvaqGsDYO4Y9JZYDttjiNcTvDAMEFihTtcR2K30kWmB75NZGO7opF8T6IKHnl5JrjJ3 w+kXROKR6fA67FiMw3WbdfhryQT9OpPVIVk80POFMgs9SWNxqMZAIJ3I8KwlyKmsmcKx cZicFX3n50AEaoQsaJn6cElAN34EwpUcTVztq3JXCu8cEfkSkVH9d33P/6IzYqJFlfN+ omng== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:list-subscribe:list-help:list-post:list-archive :list-unsubscribe:list-id:precedence:subject:references:in-reply-to :message-id:date:cc:to:from; bh=DJW4z8XkBoUkezHZG2DH27qO1fYwovo/li9zG/7H6ZM=; b=aOe4ead9htYw066ic2WDaZNXCYxOTvHKlwvh8gsjeHY2xEIsvA0nANaF9+ouzdCrIf R2auHynX3OaIg2JaTzBG7Gqao5q/IeYqw21gE1GKSONtI6/428LDL+zKlY7/gi8ZqJ5x BK3apSduIFEoTU03z3YHrenBobcYX3HR5s+xbYdMsWJ5sRBSJoUwV3Cid+HO+5P4iYui 4aTHyJm414//vROzPQaPk88BxYaQsQOrEa6vcuwz2HvsMY/UEpAj4IJwo/sizDESgpcG gw7Acu5hFaySWnbBNU8HAJkm5vTaHg7x9I04RLqlbKRRvoXlmpzjVD7KLZQRhz20tV0d g5sg== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Return-Path: Received: from dpdk.org (dpdk.org. [92.243.14.124]) by mx.google.com with ESMTP id y94-v6si579311ede.17.2018.10.11.23.32.51; Thu, 11 Oct 2018 23:32:51 -0700 (PDT) Received-SPF: pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) client-ip=92.243.14.124; Authentication-Results: mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id BA57E1B13E; Fri, 12 Oct 2018 08:32:21 +0200 (CEST) Received: from foss.arm.com (foss.arm.com [217.140.101.70]) by dpdk.org (Postfix) with ESMTP id 59D9A1B103 for ; Fri, 12 Oct 2018 08:32:13 +0200 (CEST) Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.72.51.249]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id CA7B4EBD; Thu, 11 Oct 2018 23:32:12 -0700 (PDT) Received: from 2p2660v4-1.austin.arm.com (2p2660v4-1.austin.arm.com [10.118.12.190]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 5CE113F5B3; Thu, 11 Oct 2018 23:32:12 -0700 (PDT) From: Honnappa Nagarahalli To: bruce.richardson@intel.com, pablo.de.lara.guarch@intel.com Cc: dev@dpdk.org, yipeng1.wang@intel.com, honnappa.nagarahalli@arm.com, dharmik.thakkar@arm.com, gavin.hu@arm.com, nd@arm.com Date: Fri, 12 Oct 2018 01:31:56 -0500 Message-Id: <1539325918-125438-6-git-send-email-honnappa.nagarahalli@arm.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> References: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> Subject: [dpdk-dev] [PATCH v3 5/7] hash: fix rw concurrency while moving keys X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" Reader-writer concurrency issue, caused by moving the keys to their alternative locations during key insert, is solved by introducing a global counter(tbl_chng_cnt) indicating a change in table. Signed-off-by: Honnappa Nagarahalli Reviewed-by: Gavin Hu Reviewed-by: Ola Liljedahl Reviewed-by: Steve Capper Reviewed-by: Yipeng Wang --- lib/librte_hash/rte_cuckoo_hash.c | 305 +++++++++++++++++++++++++------------- lib/librte_hash/rte_cuckoo_hash.h | 3 + 2 files changed, 208 insertions(+), 100 deletions(-) -- 2.7.4 diff --git a/lib/librte_hash/rte_cuckoo_hash.c b/lib/librte_hash/rte_cuckoo_hash.c index e55725a..262162c 100644 --- a/lib/librte_hash/rte_cuckoo_hash.c +++ b/lib/librte_hash/rte_cuckoo_hash.c @@ -142,6 +142,7 @@ rte_hash_create(const struct rte_hash_parameters *params) unsigned int readwrite_concur_support = 0; unsigned int writer_takes_lock = 0; unsigned int recycle_on_del = 1; + uint32_t *tbl_chng_cnt = NULL; rte_hash_function default_hash_func = (rte_hash_function)rte_jhash; @@ -293,6 +294,14 @@ rte_hash_create(const struct rte_hash_parameters *params) goto err_unlock; } + tbl_chng_cnt = rte_zmalloc_socket(NULL, sizeof(uint32_t), + RTE_CACHE_LINE_SIZE, params->socket_id); + + if (tbl_chng_cnt == NULL) { + RTE_LOG(ERR, HASH, "memory allocation failed\n"); + goto err_unlock; + } + /* * If x86 architecture is used, select appropriate compare function, * which may use x86 intrinsics, otherwise use memcmp @@ -361,6 +370,8 @@ rte_hash_create(const struct rte_hash_parameters *params) default_hash_func : params->hash_func; h->key_store = k; h->free_slots = r; + h->tbl_chng_cnt = tbl_chng_cnt; + *h->tbl_chng_cnt = 0; h->hw_trans_mem_support = hw_trans_mem_support; h->multi_writer_support = multi_writer_support; h->readwrite_concur_support = readwrite_concur_support; @@ -407,6 +418,7 @@ rte_hash_create(const struct rte_hash_parameters *params) rte_free(buckets); rte_free(buckets_ext); rte_free(k); + rte_free(tbl_chng_cnt); return NULL; } @@ -447,6 +459,7 @@ rte_hash_free(struct rte_hash *h) rte_free(h->key_store); rte_free(h->buckets); rte_free(h->buckets_ext); + rte_free(h->tbl_chng_cnt); rte_free(h); rte_free(te); } @@ -531,6 +544,7 @@ rte_hash_reset(struct rte_hash *h) __hash_rw_writer_lock(h); memset(h->buckets, 0, h->num_buckets * sizeof(struct rte_hash_bucket)); memset(h->key_store, 0, h->key_entry_size * (h->entries + 1)); + *h->tbl_chng_cnt = 0; /* clear the free ring */ while (rte_ring_dequeue(h->free_slots, &ptr) == 0) @@ -744,11 +758,27 @@ rte_hash_cuckoo_move_insert_mw(const struct rte_hash *h, if (unlikely(&h->buckets[prev_alt_bkt_idx] != curr_bkt)) { /* revert it to empty, otherwise duplicated keys */ - curr_bkt->key_idx[curr_slot] = EMPTY_SLOT; + __atomic_store_n(&curr_bkt->key_idx[curr_slot], + EMPTY_SLOT, + __ATOMIC_RELEASE); __hash_rw_writer_unlock(h); return -1; } + /* Inform the previous move. The current move need + * not be informed now as the current bucket entry + * is present in both primary and secondary. + * Since there is one writer, load acquires on + * tbl_chng_cnt are not required. + */ + __atomic_store_n(h->tbl_chng_cnt, + *h->tbl_chng_cnt + 1, + __ATOMIC_RELEASE); + /* The stores to sig_alt and sig_current should not + * move above the store to tbl_chng_cnt. + */ + __atomic_thread_fence(__ATOMIC_RELEASE); + /* Need to swap current/alt sig to allow later * Cuckoo insert to move elements back to its * primary bucket if available @@ -765,6 +795,20 @@ rte_hash_cuckoo_move_insert_mw(const struct rte_hash *h, curr_bkt = curr_node->bkt; } + /* Inform the previous move. The current move need + * not be informed now as the current bucket entry + * is present in both primary and secondary. + * Since there is one writer, load acquires on + * tbl_chng_cnt are not required. + */ + __atomic_store_n(h->tbl_chng_cnt, + *h->tbl_chng_cnt + 1, + __ATOMIC_RELEASE); + /* The stores to sig_alt and sig_current should not + * move above the store to tbl_chng_cnt. + */ + __atomic_thread_fence(__ATOMIC_RELEASE); + curr_bkt->sig_current[curr_slot] = sig; /* Release the new bucket entry */ __atomic_store_n(&curr_bkt->key_idx[curr_slot], @@ -1095,34 +1139,62 @@ __rte_hash_lookup_with_hash(const struct rte_hash *h, const void *key, { uint32_t prim_bucket_idx, sec_bucket_idx; struct rte_hash_bucket *bkt, *cur_bkt; + uint32_t cnt_b, cnt_a; int ret; uint16_t short_sig; short_sig = get_short_sig(sig); prim_bucket_idx = get_prim_bucket_index(h, sig); sec_bucket_idx = get_alt_bucket_index(h, prim_bucket_idx, short_sig); - bkt = &h->buckets[prim_bucket_idx]; __hash_rw_reader_lock(h); - /* Check if key is in primary location */ - ret = search_one_bucket(h, key, short_sig, data, bkt); - if (ret != -1) { - __hash_rw_reader_unlock(h); - return ret; - } - /* Calculate secondary hash */ - bkt = &h->buckets[sec_bucket_idx]; + do { + /* Load the table change counter before the lookup + * starts. Acquire semantics will make sure that + * loads in search_one_bucket are not hoisted. + */ + cnt_b = __atomic_load_n(h->tbl_chng_cnt, + __ATOMIC_ACQUIRE); - /* Check if key is in secondary location */ - FOR_EACH_BUCKET(cur_bkt, bkt) { - ret = search_one_bucket(h, key, short_sig, data, cur_bkt); + /* Check if key is in primary location */ + bkt = &h->buckets[prim_bucket_idx]; + ret = search_one_bucket(h, key, short_sig, data, bkt); if (ret != -1) { __hash_rw_reader_unlock(h); return ret; } - } + /* Calculate secondary hash */ + bkt = &h->buckets[sec_bucket_idx]; + + /* Check if key is in secondary location */ + FOR_EACH_BUCKET(cur_bkt, bkt) { + ret = search_one_bucket(h, key, short_sig, + data, cur_bkt); + if (ret != -1) { + __hash_rw_reader_unlock(h); + return ret; + } + } + + /* The loads of sig_current in search_one_bucket + * should not move below the load from tbl_chng_cnt. + */ + __atomic_thread_fence(__ATOMIC_ACQUIRE); + /* Re-read the table change counter to check if the + * table has changed during search. If yes, re-do + * the search. + * This load should not get hoisted. The load + * acquires on cnt_b, key index in primary bucket + * and key index in secondary bucket will make sure + * that it does not get hoisted. + */ + cnt_a = __atomic_load_n(h->tbl_chng_cnt, + __ATOMIC_ACQUIRE); + } while (cnt_b != cnt_a); + __hash_rw_reader_unlock(h); + return -ENOENT; } @@ -1444,6 +1516,7 @@ __rte_hash_lookup_bulk(const struct rte_hash *h, const void **keys, uint32_t sec_hitmask[RTE_HASH_LOOKUP_BULK_MAX] = {0}; struct rte_hash_bucket *cur_bkt, *next_bkt; void *pdata[RTE_HASH_LOOKUP_BULK_MAX]; + uint32_t cnt_b, cnt_a; /* Prefetch first keys */ for (i = 0; i < PREFETCH_OFFSET && i < num_keys; i++) @@ -1485,106 +1558,138 @@ __rte_hash_lookup_bulk(const struct rte_hash *h, const void **keys, } __hash_rw_reader_lock(h); - /* Compare signatures and prefetch key slot of first hit */ - for (i = 0; i < num_keys; i++) { - compare_signatures(&prim_hitmask[i], &sec_hitmask[i], + do { + /* Load the table change counter before the lookup + * starts. Acquire semantics will make sure that + * loads in compare_signatures are not hoisted. + */ + cnt_b = __atomic_load_n(h->tbl_chng_cnt, + __ATOMIC_ACQUIRE); + + /* Compare signatures and prefetch key slot of first hit */ + for (i = 0; i < num_keys; i++) { + compare_signatures(&prim_hitmask[i], &sec_hitmask[i], primary_bkt[i], secondary_bkt[i], sig[i], h->sig_cmp_fn); - if (prim_hitmask[i]) { - uint32_t first_hit = - __builtin_ctzl(prim_hitmask[i]) >> 1; - uint32_t key_idx = primary_bkt[i]->key_idx[first_hit]; - const struct rte_hash_key *key_slot = - (const struct rte_hash_key *)( - (const char *)h->key_store + - key_idx * h->key_entry_size); - rte_prefetch0(key_slot); - continue; - } + if (prim_hitmask[i]) { + uint32_t first_hit = + __builtin_ctzl(prim_hitmask[i]) + >> 1; + uint32_t key_idx = + primary_bkt[i]->key_idx[first_hit]; + const struct rte_hash_key *key_slot = + (const struct rte_hash_key *)( + (const char *)h->key_store + + key_idx * h->key_entry_size); + rte_prefetch0(key_slot); + continue; + } - if (sec_hitmask[i]) { - uint32_t first_hit = - __builtin_ctzl(sec_hitmask[i]) >> 1; - uint32_t key_idx = secondary_bkt[i]->key_idx[first_hit]; - const struct rte_hash_key *key_slot = - (const struct rte_hash_key *)( - (const char *)h->key_store + - key_idx * h->key_entry_size); - rte_prefetch0(key_slot); + if (sec_hitmask[i]) { + uint32_t first_hit = + __builtin_ctzl(sec_hitmask[i]) + >> 1; + uint32_t key_idx = + secondary_bkt[i]->key_idx[first_hit]; + const struct rte_hash_key *key_slot = + (const struct rte_hash_key *)( + (const char *)h->key_store + + key_idx * h->key_entry_size); + rte_prefetch0(key_slot); + } } - } - /* Compare keys, first hits in primary first */ - for (i = 0; i < num_keys; i++) { - positions[i] = -ENOENT; - while (prim_hitmask[i]) { - uint32_t hit_index = - __builtin_ctzl(prim_hitmask[i]) >> 1; - - uint32_t key_idx = - __atomic_load_n( - &primary_bkt[i]->key_idx[hit_index], - __ATOMIC_ACQUIRE); - const struct rte_hash_key *key_slot = - (const struct rte_hash_key *)( - (const char *)h->key_store + - key_idx * h->key_entry_size); - - if (key_idx != EMPTY_SLOT) - pdata[i] = __atomic_load_n(&key_slot->pdata, - __ATOMIC_ACQUIRE); - /* - * If key index is 0, do not compare key, - * as it is checking the dummy slot - */ - if (!!key_idx & !rte_hash_cmp_eq(key_slot->key, keys[i], h)) { - if (data != NULL) - data[i] = pdata[i]; + /* Compare keys, first hits in primary first */ + for (i = 0; i < num_keys; i++) { + positions[i] = -ENOENT; + while (prim_hitmask[i]) { + uint32_t hit_index = + __builtin_ctzl(prim_hitmask[i]) + >> 1; + uint32_t key_idx = + __atomic_load_n( + &primary_bkt[i]->key_idx[hit_index], + __ATOMIC_ACQUIRE); + const struct rte_hash_key *key_slot = + (const struct rte_hash_key *)( + (const char *)h->key_store + + key_idx * h->key_entry_size); - hits |= 1ULL << i; - positions[i] = key_idx - 1; - goto next_key; + if (key_idx != EMPTY_SLOT) + pdata[i] = __atomic_load_n( + &key_slot->pdata, + __ATOMIC_ACQUIRE); + /* + * If key index is 0, do not compare key, + * as it is checking the dummy slot + */ + if (!!key_idx & + !rte_hash_cmp_eq( + key_slot->key, keys[i], h)) { + if (data != NULL) + data[i] = pdata[i]; + + hits |= 1ULL << i; + positions[i] = key_idx - 1; + goto next_key; + } + prim_hitmask[i] &= ~(3ULL << (hit_index << 1)); } - prim_hitmask[i] &= ~(3ULL << (hit_index << 1)); - } - - while (sec_hitmask[i]) { - uint32_t hit_index = - __builtin_ctzl(sec_hitmask[i]) >> 1; - uint32_t key_idx = - __atomic_load_n( - &secondary_bkt[i]->key_idx[hit_index], - __ATOMIC_ACQUIRE); - const struct rte_hash_key *key_slot = - (const struct rte_hash_key *)( - (const char *)h->key_store + - key_idx * h->key_entry_size); - - if (key_idx != EMPTY_SLOT) - pdata[i] = __atomic_load_n(&key_slot->pdata, - __ATOMIC_ACQUIRE); - - /* - * If key index is 0, do not compare key, - * as it is checking the dummy slot - */ + while (sec_hitmask[i]) { + uint32_t hit_index = + __builtin_ctzl(sec_hitmask[i]) + >> 1; + uint32_t key_idx = + __atomic_load_n( + &secondary_bkt[i]->key_idx[hit_index], + __ATOMIC_ACQUIRE); + const struct rte_hash_key *key_slot = + (const struct rte_hash_key *)( + (const char *)h->key_store + + key_idx * h->key_entry_size); - if (!!key_idx & !rte_hash_cmp_eq(key_slot->key, keys[i], h)) { - if (data != NULL) - data[i] = pdata[i]; + if (key_idx != EMPTY_SLOT) + pdata[i] = __atomic_load_n( + &key_slot->pdata, + __ATOMIC_ACQUIRE); + /* + * If key index is 0, do not compare key, + * as it is checking the dummy slot + */ - hits |= 1ULL << i; - positions[i] = key_idx - 1; - goto next_key; + if (!!key_idx & + !rte_hash_cmp_eq( + key_slot->key, keys[i], h)) { + if (data != NULL) + data[i] = pdata[i]; + + hits |= 1ULL << i; + positions[i] = key_idx - 1; + goto next_key; + } + sec_hitmask[i] &= ~(3ULL << (hit_index << 1)); } - sec_hitmask[i] &= ~(3ULL << (hit_index << 1)); +next_key: + continue; } -next_key: - continue; - } + /* The loads of sig_current in compare_signatures + * should not move below the load from tbl_chng_cnt. + */ + __atomic_thread_fence(__ATOMIC_ACQUIRE); + /* Re-read the table change counter to check if the + * table has changed during search. If yes, re-do + * the search. + * This load should not get hoisted. The load + * acquires on cnt_b, primary key index and secondary + * key index will make sure that it does not get + * hoisted. + */ + cnt_a = __atomic_load_n(h->tbl_chng_cnt, + __ATOMIC_ACQUIRE); + } while (cnt_b != cnt_a); /* all found, do not need to go through ext bkt */ if ((hits == ((1ULL << num_keys) - 1)) || !h->ext_table_support) { diff --git a/lib/librte_hash/rte_cuckoo_hash.h b/lib/librte_hash/rte_cuckoo_hash.h index b823d37..728cd17 100644 --- a/lib/librte_hash/rte_cuckoo_hash.h +++ b/lib/librte_hash/rte_cuckoo_hash.h @@ -1,5 +1,6 @@ /* SPDX-License-Identifier: BSD-3-Clause * Copyright(c) 2016 Intel Corporation + * Copyright(c) 2018 Arm Limited */ /* rte_cuckoo_hash.h @@ -196,6 +197,8 @@ struct rte_hash { rte_rwlock_t *readwrite_lock; /**< Read-write lock thread-safety. */ struct rte_hash_bucket *buckets_ext; /**< Extra buckets array */ struct rte_ring *free_ext_bkts; /**< Ring of indexes of free buckets */ + uint32_t *tbl_chng_cnt; + /**< Indicates if the hash table changed from last read. */ } __rte_cache_aligned; struct queue_node { From patchwork Fri Oct 12 06:31:57 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Honnappa Nagarahalli X-Patchwork-Id: 148721 Delivered-To: patch@linaro.org Received: by 2002:a2e:8595:0:0:0:0:0 with SMTP id b21-v6csp368493lji; Thu, 11 Oct 2018 23:33:00 -0700 (PDT) X-Google-Smtp-Source: ACcGV61j3Cu+3PJVpmVhkz8owcRvpAiJDUFmblp1GdljU6rcOYbNtl2GzidS4oYihIpWM3Cn02vg X-Received: by 2002:a50:adca:: with SMTP id b10-v6mr7184866edd.184.1539325979972; Thu, 11 Oct 2018 23:32:59 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1539325979; cv=none; d=google.com; s=arc-20160816; b=ejIEqsNGlB28jxD+oP73q3nfx0c3QGF1fc+nXBDrXfgPtbTuIq97vxrMe8o8ZLOOuU SGey65E2tWp7w6LRHw3k+v6MTNYH+opOlqrBR/2F6hS4gKKKeRy1MyLe1kIuMdN1f/Sg vtKZHXnV1r1S5pi9I+b7xPCyRjuKazqbQ8A7JlguYfWcvAUoMkfVJWSYp+5y40ks1Lfm pGzV8OotxiZkFEeFavHzf4FLM1r+4Z/Z0cEh+sV4feIhRNxHGa/yboz/J12UL3CzX/Se 5kt6yJZd9lURpHkrTW8HtcqyvXnsUB5FL5gv67b9h3+ynW0689Q17oZoUbOvbKP2JzD2 qKGA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:list-subscribe:list-help:list-post:list-archive :list-unsubscribe:list-id:precedence:subject:references:in-reply-to :message-id:date:cc:to:from; bh=Sl8brzFe7x04jwc/Ki4TuHaZ/1aF1Ncnp+Ww9Go+RNI=; b=Xi6UHE/ngtjLq5cgPE+T1zzsotb1xL6wQ/P3l7puIAuOnrxJxiIQ7ixjcwqJAZMhpt G+vf0PX/cZRFz/zZJziZ4TEu9iuUmKso4Nv+JJvTe1UjlkOLKdxNiCEFrrF0LMPxZofB RBc1QbP+oMV+t3aaX0u9npBKB1N2QOEOP93lYXM+aNJnBSuu1HUH0b2mZuXUuo6L7TVq vlOK1Dw68Y9fwCgtb+CnvuJCItA7n4cRgKPm0VyAiy3KpIyk5HpeTNBSyxKrvZo77UFs DRnsljA3PrbdHS/AFBRKoMMrTWsjynEM8/U72fFhGi/A025azaLgk4g8kZQdOzsyAPvB nGrA== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Return-Path: Received: from dpdk.org (dpdk.org. [92.243.14.124]) by mx.google.com with ESMTP id r17-v6si471211edc.269.2018.10.11.23.32.59; Thu, 11 Oct 2018 23:32:59 -0700 (PDT) Received-SPF: pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) client-ip=92.243.14.124; Authentication-Results: mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id BFD8C1B14F; Fri, 12 Oct 2018 08:32:24 +0200 (CEST) Received: from foss.arm.com (foss.arm.com [217.140.101.70]) by dpdk.org (Postfix) with ESMTP id 6E9F01B10D for ; Fri, 12 Oct 2018 08:32:14 +0200 (CEST) Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.72.51.249]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id DABE315BF; Thu, 11 Oct 2018 23:32:13 -0700 (PDT) Received: from 2p2660v4-1.austin.arm.com (2p2660v4-1.austin.arm.com [10.118.12.190]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 7544D3F5B3; Thu, 11 Oct 2018 23:32:13 -0700 (PDT) From: Honnappa Nagarahalli To: bruce.richardson@intel.com, pablo.de.lara.guarch@intel.com Cc: dev@dpdk.org, yipeng1.wang@intel.com, honnappa.nagarahalli@arm.com, dharmik.thakkar@arm.com, gavin.hu@arm.com, nd@arm.com Date: Fri, 12 Oct 2018 01:31:57 -0500 Message-Id: <1539325918-125438-7-git-send-email-honnappa.nagarahalli@arm.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> References: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> Subject: [dpdk-dev] [PATCH v3 6/7] hash: enable lock-free reader-writer concurrency X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" Add the flag to enable reader-writer concurrency during run time. The rte_hash_del_xxx APIs do not free the keystore element when this flag is enabled. Hence a new API, rte_hash_free_key_with_position, to free the key store element is added. Signed-off-by: Honnappa Nagarahalli Reviewed-by: Gavin Hu Reviewed-by: Ola Liljedahl Reviewed-by: Steve Capper Reviewed-by: Yipeng Wang --- lib/librte_hash/rte_cuckoo_hash.c | 82 ++++++++++++++++++++++++------------ lib/librte_hash/rte_cuckoo_hash.h | 2 + lib/librte_hash/rte_hash.h | 58 +++++++++++++++++++++---- lib/librte_hash/rte_hash_version.map | 7 +++ 4 files changed, 114 insertions(+), 35 deletions(-) -- 2.7.4 diff --git a/lib/librte_hash/rte_cuckoo_hash.c b/lib/librte_hash/rte_cuckoo_hash.c index 262162c..4622ece 100644 --- a/lib/librte_hash/rte_cuckoo_hash.c +++ b/lib/librte_hash/rte_cuckoo_hash.c @@ -143,6 +143,7 @@ rte_hash_create(const struct rte_hash_parameters *params) unsigned int writer_takes_lock = 0; unsigned int recycle_on_del = 1; uint32_t *tbl_chng_cnt = NULL; + unsigned int readwrite_concur_lf_support = 0; rte_hash_function default_hash_func = (rte_hash_function)rte_jhash; @@ -162,6 +163,24 @@ rte_hash_create(const struct rte_hash_parameters *params) return NULL; } + /* Validate correct usage of extra options */ + if ((params->extra_flag & RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY) && + (params->extra_flag & RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF)) { + rte_errno = EINVAL; + RTE_LOG(ERR, HASH, "rte_hash_create choose rw concurrency or " + "rw concurrency lock free\n"); + return NULL; + } + + if ((params->extra_flag & RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF) && + (params->extra_flag & RTE_HASH_EXTRA_FLAGS_EXT_TABLE)) { + rte_errno = EINVAL; + RTE_LOG(ERR, HASH, "rte_hash_create extendable bucket " + "feature not supported with rw concurrency " + "lock free\n"); + return NULL; + } + /* Check extra flags field to check extra options. */ if (params->extra_flag & RTE_HASH_EXTRA_FLAGS_TRANS_MEM_SUPPORT) hw_trans_mem_support = 1; @@ -182,6 +201,12 @@ rte_hash_create(const struct rte_hash_parameters *params) if (params->extra_flag & RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL) recycle_on_del = 0; + if (params->extra_flag & RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF) { + readwrite_concur_lf_support = 1; + /* Disable freeing internal memory/index on delete */ + recycle_on_del = 0; + } + /* Store all keys and leave the first entry as a dummy entry for lookup_bulk */ if (multi_writer_support) /* @@ -378,6 +403,7 @@ rte_hash_create(const struct rte_hash_parameters *params) h->ext_table_support = ext_table_support; h->writer_takes_lock = writer_takes_lock; h->recycle_on_del = recycle_on_del; + h->readwrite_concur_lf_support = readwrite_concur_lf_support; #if defined(RTE_ARCH_X86) if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_SSE2)) @@ -765,19 +791,21 @@ rte_hash_cuckoo_move_insert_mw(const struct rte_hash *h, return -1; } - /* Inform the previous move. The current move need - * not be informed now as the current bucket entry - * is present in both primary and secondary. - * Since there is one writer, load acquires on - * tbl_chng_cnt are not required. - */ - __atomic_store_n(h->tbl_chng_cnt, - *h->tbl_chng_cnt + 1, - __ATOMIC_RELEASE); - /* The stores to sig_alt and sig_current should not - * move above the store to tbl_chng_cnt. - */ - __atomic_thread_fence(__ATOMIC_RELEASE); + if (h->readwrite_concur_lf_support) { + /* Inform the previous move. The current move need + * not be informed now as the current bucket entry + * is present in both primary and secondary. + * Since there is one writer, load acquires on + * tbl_chng_cnt are not required. + */ + __atomic_store_n(h->tbl_chng_cnt, + *h->tbl_chng_cnt + 1, + __ATOMIC_RELEASE); + /* The stores to sig_alt and sig_current should not + * move above the store to tbl_chng_cnt. + */ + __atomic_thread_fence(__ATOMIC_RELEASE); + } /* Need to swap current/alt sig to allow later * Cuckoo insert to move elements back to its @@ -795,19 +823,21 @@ rte_hash_cuckoo_move_insert_mw(const struct rte_hash *h, curr_bkt = curr_node->bkt; } - /* Inform the previous move. The current move need - * not be informed now as the current bucket entry - * is present in both primary and secondary. - * Since there is one writer, load acquires on - * tbl_chng_cnt are not required. - */ - __atomic_store_n(h->tbl_chng_cnt, - *h->tbl_chng_cnt + 1, - __ATOMIC_RELEASE); - /* The stores to sig_alt and sig_current should not - * move above the store to tbl_chng_cnt. - */ - __atomic_thread_fence(__ATOMIC_RELEASE); + if (h->readwrite_concur_lf_support) { + /* Inform the previous move. The current move need + * not be informed now as the current bucket entry + * is present in both primary and secondary. + * Since there is one writer, load acquires on + * tbl_chng_cnt are not required. + */ + __atomic_store_n(h->tbl_chng_cnt, + *h->tbl_chng_cnt + 1, + __ATOMIC_RELEASE); + /* The stores to sig_alt and sig_current should not + * move above the store to tbl_chng_cnt. + */ + __atomic_thread_fence(__ATOMIC_RELEASE); + } curr_bkt->sig_current[curr_slot] = sig; /* Release the new bucket entry */ diff --git a/lib/librte_hash/rte_cuckoo_hash.h b/lib/librte_hash/rte_cuckoo_hash.h index 728cd17..0c23957 100644 --- a/lib/librte_hash/rte_cuckoo_hash.h +++ b/lib/librte_hash/rte_cuckoo_hash.h @@ -175,6 +175,8 @@ struct rte_hash { * the deleted entry. * This flag is enabled by default. */ + uint8_t readwrite_concur_lf_support; + /**< If read-write concurrency lock free support is enabled */ uint8_t writer_takes_lock; /**< Indicates if the writer threads need to take lock */ rte_hash_function hash_func; /**< Function used to calculate hash. */ diff --git a/lib/librte_hash/rte_hash.h b/lib/librte_hash/rte_hash.h index 3e5d336..22def5e 100644 --- a/lib/librte_hash/rte_hash.h +++ b/lib/librte_hash/rte_hash.h @@ -44,9 +44,18 @@ extern "C" { /** Flag to disable freeing of internal memory/indices on hash delete. * Refer to rte_hash_del_xxx APIs for more details. + * This is enabled by default when RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF + * is enabled. */ #define RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL 0x10 +/** Flag to support lock free reader writer concurrency. Writer can be + * single writer/multi writer. + * Currently, extended bucket table feature is not supported with + * this feature. + */ +#define RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF 0x20 + /** * The type of hash value of a key. * It should be a value of at least 32bit with fully random pattern. @@ -132,7 +141,11 @@ void rte_hash_free(struct rte_hash *h); /** - * Reset all hash structure, by zeroing all entries + * Reset all hash structure, by zeroing all entries. + * When RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled, + * it is application's responsibility to make sure that + * none of the readers are referencing the hash table. + * * @param h * Hash table to reset */ @@ -156,6 +169,10 @@ rte_hash_count(const struct rte_hash *h); * and should only be called from one thread by default. * Thread safety can be enabled by setting flag during * table creation. + * The writer needs to be aware if this API is called to update + * an existing entry. The application should free any memory + * allocated for the existing 'data' only after all the readers + * have stopped referrencing it. * * @param h * Hash table to add the key to. @@ -178,6 +195,10 @@ rte_hash_add_key_data(const struct rte_hash *h, const void *key, void *data); * and should only be called from one thread by default. * Thread safety can be enabled by setting flag during * table creation. + * The writer needs to be aware if this API is called to update + * an existing entry. The application should free any memory + * allocated for the existing 'data' only after all the readers + * have stopped referencing it. * * @param h * Hash table to add the key to. @@ -243,10 +264,15 @@ rte_hash_add_key_with_hash(const struct rte_hash *h, const void *key, hash_sig_t * and should only be called from one thread by default. * Thread safety can be enabled by setting flag during * table creation. - * If RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL is enabled, - * the hash library's internal memory/index will not be freed by this + * If RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL or + * RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled, + * the hash library's internal memory will not be freed by this * API. rte_hash_free_key_with_position API must be called additionally - * to free the internal memory/index associated with the key. + * to free any internal memory associated with the key. + * If RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled, + * rte_hash_free_key_with_position API should be called after all + * the readers have stopped referencing the entry corresponding to + * this key. RCU mechanisms can be used to determine such a state. * * @param h * Hash table to remove the key from. @@ -258,6 +284,8 @@ rte_hash_add_key_with_hash(const struct rte_hash *h, const void *key, hash_sig_t * - A positive value that can be used by the caller as an offset into an * array of user data. This value is unique for this key, and is the same * value that was returned when the key was added. + * When lock free concurrency is enabled, this value should be used + * while calling the rte_hash_free_key_with_position API. */ int32_t rte_hash_del_key(const struct rte_hash *h, const void *key); @@ -268,10 +296,15 @@ rte_hash_del_key(const struct rte_hash *h, const void *key); * and should only be called from one thread by default. * Thread safety can be enabled by setting flag during * table creation. - * If RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL is enabled, - * the hash library's internal memory/index will not be freed by this + * If RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL or + * RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled, + * the hash library's internal memory will not be freed by this * API. rte_hash_free_key_with_position API must be called additionally - * to free the internal memory/index associated with the key. + * to free any internal memory associated with the key. + * If RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled, + * rte_hash_free_key_with_position API should be called after all + * the readers have stopped referencing the entry corresponding to + * this key. RCU mechanisms can be used to determine such a state. * * @param h * Hash table to remove the key from. @@ -285,6 +318,8 @@ rte_hash_del_key(const struct rte_hash *h, const void *key); * - A positive value that can be used by the caller as an offset into an * array of user data. This value is unique for this key, and is the same * value that was returned when the key was added. + * When lock free concurrency is enabled, this value should be used + * while calling the rte_hash_free_key_with_position API. */ int32_t rte_hash_del_key_with_hash(const struct rte_hash *h, const void *key, hash_sig_t sig); @@ -318,10 +353,15 @@ rte_hash_get_key_with_position(const struct rte_hash *h, const int32_t position, * of the key. This operation is not multi-thread safe and should * only be called from one thread by default. Thread safety * can be enabled by setting flag during table creation. - * If RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL is enabled, - * the hash library's internal memory/index must be freed using this API + * If RTE_HASH_EXTRA_FLAGS_RECYCLE_ON_DEL or + * RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled, + * the hash library's internal memory must be freed using this API * after the key is deleted using rte_hash_del_key_xxx APIs. * This API does not validate if the key is already freed. + * If RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled, + * this API should be called only after all the readers have stopped + * referencing the entry corresponding to this key. RCU mechanisms can + * be used to determine such a state. * * @param h * Hash table to free the key from. diff --git a/lib/librte_hash/rte_hash_version.map b/lib/librte_hash/rte_hash_version.map index e216ac8..734ae28 100644 --- a/lib/librte_hash/rte_hash_version.map +++ b/lib/librte_hash/rte_hash_version.map @@ -53,3 +53,10 @@ DPDK_18.08 { rte_hash_count; } DPDK_16.07; + +EXPERIMENTAL { + global: + + rte_hash_free_key_with_position; + +}; From patchwork Fri Oct 12 06:31:58 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Honnappa Nagarahalli X-Patchwork-Id: 148722 Delivered-To: patch@linaro.org Received: by 2002:a2e:8595:0:0:0:0:0 with SMTP id b21-v6csp368572lji; Thu, 11 Oct 2018 23:33:07 -0700 (PDT) X-Google-Smtp-Source: ACcGV63KmtxcgYxIJN8Y7whWBjSEmTErSkhhUU3HCjttbAV0GsJATAs7cUMWayLmPAdq4noHoPAy X-Received: by 2002:a17:906:1d5b:: with SMTP id o27-v6mr5571024ejh.61.1539325987550; Thu, 11 Oct 2018 23:33:07 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1539325987; cv=none; d=google.com; s=arc-20160816; b=CpxnUQmwhaDKag8k0GvqldKOyrsczQ7Yv6jEte809n9gyBduNi4zF8rBTOg49Ote52 4bhP3vbuThF8eHZf9QNwsx15NBne6wS0cYrauI0EtZLyHDBa8mC5EIzRuUWOtkb5s8Mk TVpjDbZEZVHoVoGy5HN0zoz9QzMuSM2Bs28IrbzD6Y8FSsF+nrHUi4hsuHV6RM7d7XpU E6VcmELjijqPg2KWb+zPDcEfEG5uNIM/rkekxDaKABLmWWWRpcDgxn/llzSuhRaAuXrh tL/0mcws2Co+J1uvXFoArpUkEENsYnB2oic+OmaXzpIJNtzCLTjADvTP77L3FBVBQp8I nonw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:list-subscribe:list-help:list-post:list-archive :list-unsubscribe:list-id:precedence:subject:references:in-reply-to :message-id:date:cc:to:from; bh=FmXy4TZ6MSrzxzF6JD9ntjPzCkulSL7RVI+mzZbB95A=; b=DQEmfNFCHE6Xb8h5ybBLNxQKoNdFlT8ustt6BgI78rQmi1H/93E7wsZd/g4TIR6y5o YEunPASDPPj/2CdBhhf0SBTMJmZUGI+zGBz2s/Hv1101voEvoZY9dRLkCSY3hEk3Hpk0 cVazINTNvP7ii67EjJC6Gcm8qyy5DWKlE12lohZsmtq9JyFNZYVmvucl4rELDqif/04A 01aqNOU+ytxA/EATmTnamNSL6CKo3X/r9oKQxrYz6Jj2QSePDMTWP9hM/4dpYy+O2dxf BgakE5CPXIouMF5eWhXamSz6Hdmb61y3v6J8hBocZagtHC8HvRHyj88LhMTYe4G4U/zo pFyg== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Return-Path: Received: from dpdk.org (dpdk.org. [92.243.14.124]) by mx.google.com with ESMTP id ce2-v6si389375ejb.214.2018.10.11.23.33.07; Thu, 11 Oct 2018 23:33:07 -0700 (PDT) Received-SPF: pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) client-ip=92.243.14.124; Authentication-Results: mx.google.com; spf=pass (google.com: domain of dev-bounces@dpdk.org designates 92.243.14.124 as permitted sender) smtp.mailfrom=dev-bounces@dpdk.org Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id E11AB1B157; Fri, 12 Oct 2018 08:32:26 +0200 (CEST) Received: from foss.arm.com (foss.arm.com [217.140.101.70]) by dpdk.org (Postfix) with ESMTP id 8D0A21B100 for ; Fri, 12 Oct 2018 08:32:15 +0200 (CEST) Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.72.51.249]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id 09762EBD; Thu, 11 Oct 2018 23:32:15 -0700 (PDT) Received: from 2p2660v4-1.austin.arm.com (2p2660v4-1.austin.arm.com [10.118.12.190]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 865FA3F5B3; Thu, 11 Oct 2018 23:32:14 -0700 (PDT) From: Honnappa Nagarahalli To: bruce.richardson@intel.com, pablo.de.lara.guarch@intel.com Cc: dev@dpdk.org, yipeng1.wang@intel.com, honnappa.nagarahalli@arm.com, dharmik.thakkar@arm.com, gavin.hu@arm.com, nd@arm.com Date: Fri, 12 Oct 2018 01:31:58 -0500 Message-Id: <1539325918-125438-8-git-send-email-honnappa.nagarahalli@arm.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> References: <1539325918-125438-1-git-send-email-honnappa.nagarahalli@arm.com> Subject: [dpdk-dev] [PATCH v3 7/7] test/hash: read-write lock-free concurrency test X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" From: Dharmik Thakkar Unit tests to check for hash lookup perf with lock-free enabled and with lock-free disabled. Unit tests performed with readers running in parallel with writers. Tests include: - hash lookup on existing keys with: - hash add causing NO key-shifts of existing keys in the table - hash lookup on existing keys likely to be on shift-path with: - hash add causing key-shifts of existing keys in the table - hash lookup on existing keys NOT likely to be on shift-path with: - hash add causing key-shifts of existing keys in the table - hash lookup on non-existing keys with: - hash add causing NO key-shifts of existing keys in the table - hash add causing key-shifts of existing keys in the table - hash lookup on keys likely to be on shift-path with: - multiple writers causing key-shifts of existing keys in the table Signed-off-by: Dharmik Thakkar Reviewed-by: Honnappa Nagarahalli Reviewed-by: Gavin Hu --- test/test/Makefile | 1 + test/test/meson.build | 1 + test/test/test_hash_readwrite_lf.c | 1084 ++++++++++++++++++++++++++++++++++++ 3 files changed, 1086 insertions(+) create mode 100644 test/test/test_hash_readwrite_lf.c -- 2.7.4 diff --git a/test/test/Makefile b/test/test/Makefile index 5d8b1dc..0e0e6c4 100644 --- a/test/test/Makefile +++ b/test/test/Makefile @@ -116,6 +116,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_HASH) += test_hash_functions.c SRCS-$(CONFIG_RTE_LIBRTE_HASH) += test_hash_scaling.c SRCS-$(CONFIG_RTE_LIBRTE_HASH) += test_hash_multiwriter.c SRCS-$(CONFIG_RTE_LIBRTE_HASH) += test_hash_readwrite.c +SRCS-$(CONFIG_RTE_LIBRTE_HASH) += test_hash_readwrite_lf.c SRCS-$(CONFIG_RTE_LIBRTE_LPM) += test_lpm.c SRCS-$(CONFIG_RTE_LIBRTE_LPM) += test_lpm_perf.c diff --git a/test/test/meson.build b/test/test/meson.build index d696f5e..bc3350f 100644 --- a/test/test/meson.build +++ b/test/test/meson.build @@ -46,6 +46,7 @@ test_sources = files('commands.c', 'test_hash_multiwriter.c', 'test_hash_readwrite.c', 'test_hash_perf.c', + 'test_hash_readwrite_lf.c', 'test_hash_scaling.c', 'test_interrupts.c', 'test_kni.c', diff --git a/test/test/test_hash_readwrite_lf.c b/test/test/test_hash_readwrite_lf.c new file mode 100644 index 0000000..841e989 --- /dev/null +++ b/test/test/test_hash_readwrite_lf.c @@ -0,0 +1,1084 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2018 Arm Limited + */ + +#include +#include + +#include +#include +#include +#include +#include +#include +#include +#include + +#include "test.h" + +#ifndef RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF +#define RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF 0 +#endif + +#define RUN_WITH_HTM_DISABLED 0 + +#if (RUN_WITH_HTM_DISABLED) + +#define TOTAL_ENTRY (5*1024) +#define TOTAL_INSERT (5*1024) + +#else + +#define TOTAL_ENTRY (16*1024*1024) +#define TOTAL_INSERT (16*1024*1024) + +#endif + +#define READ_FAIL 0 +#define READ_PASS_NO_KEY_SHIFTS 1 +#define READ_PASS_SHIFT_PATH 2 +#define READ_PASS_NON_SHIFT_PATH 3 + +#define NUM_TEST 3 +unsigned int rwc_core_cnt[NUM_TEST] = {1, 2, 4}; + +struct rwc_perf { + uint32_t w_no_ks_r_pass[NUM_TEST]; + uint32_t w_no_ks_r_fail[NUM_TEST]; + uint32_t w_ks_r_pass_nsp[NUM_TEST]; + uint32_t w_ks_r_pass_sp[NUM_TEST]; + uint32_t w_ks_r_fail[NUM_TEST]; + uint32_t multi_rw[NUM_TEST - 1][NUM_TEST]; +}; + +static struct rwc_perf rwc_lf_results, rwc_non_lf_results; + +struct { + uint32_t *keys; + uint32_t *keys_no_ks; + uint32_t *keys_ks; + uint32_t *keys_absent; + uint32_t *keys_shift_path; + uint32_t *keys_non_shift_path; + uint32_t count_keys_no_ks; + uint32_t count_keys_ks; + uint32_t count_keys_absent; + uint32_t count_keys_shift_path; + uint32_t count_keys_non_shift_path; + uint32_t single_insert; + struct rte_hash *h; +} tbl_rwc_test_param; + +static rte_atomic64_t gread_cycles; +static rte_atomic64_t greads; + +static volatile uint8_t writer_done; +static volatile uint8_t multi_writer_done[4]; +uint8_t num_test; +uint8_t htm; + +uint16_t enabled_core_ids[RTE_MAX_LCORE]; + +uint8_t *scanned_bkts; + +static inline int +get_enabled_cores_list(void) +{ + uint32_t i = 0; + uint16_t core_id; + uint32_t max_cores = rte_lcore_count(); + for (core_id = 0; core_id < RTE_MAX_LCORE && i < max_cores; core_id++) { + if (rte_lcore_is_enabled(core_id)) { + enabled_core_ids[i] = core_id; + i++; + } + } + + if (i != rte_lcore_count()) { + printf("Number of enabled cores in list is different from " + "number given by rte_lcore_count()\n"); + return -1; + } + return 0; +} + +static inline int +check_bucket(uint32_t bkt_idx, uint32_t key) +{ + uint32_t iter; + uint32_t prev_iter; + uint32_t diff; + uint32_t count = 0; + const void *next_key; + void *next_data; + + /* Temporary bucket to hold the keys */ + uint32_t keys_in_bkt[8]; + + iter = bkt_idx * 8; + prev_iter = iter; + while (rte_hash_iterate(tbl_rwc_test_param.h, + &next_key, &next_data, &iter) >= 0) { + + /* Check for duplicate entries */ + if (*(const uint32_t *)next_key == key) + return 1; + + /* Identify if there is any free entry in the bucket */ + diff = iter - prev_iter; + if (diff > 1) + break; + + prev_iter = iter; + keys_in_bkt[count] = *(const uint32_t *)next_key; + count++; + + /* All entries in the bucket are occupied */ + if (count == 8) { + + /* + * Check if bucket was not scanned before, to avoid + * duplicate keys. + */ + if (scanned_bkts[bkt_idx] == 0) { + /* + * Since this bucket (pointed to by bkt_idx) is + * full, it is likely that key(s) in this + * bucket will be on the shift path, when + * collision occurs. Thus, add it to + * keys_shift_path. + */ + memcpy(tbl_rwc_test_param.keys_shift_path + + tbl_rwc_test_param.count_keys_shift_path + , keys_in_bkt, 32); + tbl_rwc_test_param.count_keys_shift_path += 8; + scanned_bkts[bkt_idx] = 1; + } + return -1; + } + } + return 0; +} + +static int +generate_keys(void) +{ + uint32_t *keys = NULL; + uint32_t *keys_no_ks = NULL; + uint32_t *keys_ks = NULL; + uint32_t *keys_absent = NULL; + uint32_t *keys_non_shift_path = NULL; + uint32_t *found = NULL; + uint32_t count_keys_no_ks = 0; + uint32_t count_keys_ks = 0; + uint32_t i; + + /* + * keys will consist of a) keys whose addition to the hash table + * will result in shifting of the existing keys to their alternate + * locations b) keys whose addition to the hash table will not result + * in shifting of the existing keys. + */ + keys = rte_malloc(NULL, sizeof(uint32_t) * TOTAL_INSERT, 0); + if (keys == NULL) { + printf("RTE_MALLOC failed\n"); + goto err; + } + + /* + * keys_no_ks (no key-shifts): Subset of 'keys' - consists of keys that + * will NOT result in shifting of the existing keys to their alternate + * locations. Roughly around 900K keys. + */ + keys_no_ks = rte_malloc(NULL, sizeof(uint32_t) * TOTAL_INSERT, 0); + if (keys_no_ks == NULL) { + printf("RTE_MALLOC failed\n"); + goto err; + } + + /* + * keys_ks (key-shifts): Subset of 'keys' - consists of keys that will + * result in shifting of the existing keys to their alternate locations. + * Roughly around 146K keys. There might be repeating keys. More code is + * required to filter out these keys which will complicate the test case + */ + keys_ks = rte_malloc(NULL, sizeof(uint32_t) * TOTAL_INSERT, 0); + if (keys_ks == NULL) { + printf("RTE_MALLOC failed\n"); + goto err; + } + + /* Used to identify keys not inserted in the hash table */ + found = rte_zmalloc(NULL, sizeof(uint32_t) * TOTAL_INSERT, 0); + if (found == NULL) { + printf("RTE_MALLOC failed\n"); + goto err; + } + + /* + * This consist of keys not inserted to the hash table. + * Used to test perf of lookup on keys that do not exist in the table. + */ + keys_absent = rte_malloc(NULL, sizeof(uint32_t) * TOTAL_INSERT, 0); + if (keys_absent == NULL) { + printf("RTE_MALLOC failed\n"); + goto err; + } + + /* + * This consist of keys which are likely to be on the shift + * path (i.e. being moved to alternate location), when collision occurs + * on addition of a key to an already full primary bucket. + * Used to test perf of lookup on keys that are on the shift path. + */ + tbl_rwc_test_param.keys_shift_path = rte_malloc(NULL, sizeof(uint32_t) * + TOTAL_INSERT, 0); + if (tbl_rwc_test_param.keys_shift_path == NULL) { + printf("RTE_MALLOC failed\n"); + goto err; + } + + /* + * This consist of keys which are never on the shift + * path (i.e. being moved to alternate location), when collision occurs + * on addition of a key to an already full primary bucket. + * Used to test perf of lookup on keys that are not on the shift path. + */ + keys_non_shift_path = rte_malloc(NULL, sizeof(uint32_t) * TOTAL_INSERT, + 0); + if (keys_non_shift_path == NULL) { + printf("RTE_MALLOC failed\n"); + goto err; + } + + /* + * Used to mark bkts in which at least one key was shifted to its + * alternate location + */ + scanned_bkts = rte_malloc(NULL, sizeof(uint8_t) * TOTAL_INSERT / 8, 0); + if (scanned_bkts == NULL) { + printf("RTE_MALLOC failed\n"); + goto err; + } + + tbl_rwc_test_param.keys = keys; + tbl_rwc_test_param.keys_no_ks = keys_no_ks; + tbl_rwc_test_param.keys_ks = keys_ks; + tbl_rwc_test_param.keys_absent = keys_absent; + tbl_rwc_test_param.keys_non_shift_path = keys_non_shift_path; + + hash_sig_t sig; + uint32_t prim_bucket_idx; + int ret; + uint32_t num_buckets; + uint32_t bucket_bitmask; + num_buckets = TOTAL_ENTRY/8; + bucket_bitmask = num_buckets - 1; + + /* Generate keys by adding previous two keys, neglect overflow */ + keys[0] = 0; + keys[1] = 1; + for (i = 2; i < TOTAL_INSERT; i++) + keys[i] = keys[i-1] + keys[i-2]; + + /* Segregate keys into keys_no_ks and keys_ks */ + for (i = 0; i < TOTAL_INSERT; i++) { + /* Check if primary bucket has space.*/ + sig = rte_hash_hash(tbl_rwc_test_param.h, + tbl_rwc_test_param.keys+i); + prim_bucket_idx = sig & bucket_bitmask; + ret = check_bucket(prim_bucket_idx, keys[i]); + if (ret < 0) { + /* + * Primary bucket is full, this key will result in + * shifting of the keys to their alternate locations. + */ + keys_ks[count_keys_ks] = keys[i]; + count_keys_ks++; + } else if (ret == 0) { + /* + * Primary bucket has space, this key will not result in + * shifting of the keys. Hence, add key to the table. + */ + ret = rte_hash_add_key_data(tbl_rwc_test_param.h, + keys+i, + (void *)((uintptr_t)i)); + if (ret < 0) { + printf("writer failed %"PRIu32"\n", i); + break; + } + keys_no_ks[count_keys_no_ks] = keys[i]; + count_keys_no_ks++; + } + } + + for (i = 0; i < count_keys_no_ks; i++) { + /* Identify keys in keys_no_ks with value less than 1M */ + if (keys_no_ks[i] < TOTAL_INSERT) + found[keys_no_ks[i]]++; + } + + for (i = 0; i < count_keys_ks; i++) { + /* Identify keys in keys_ks with value less than 1M */ + if (keys_ks[i] < TOTAL_INSERT) + found[keys_ks[i]]++; + } + + uint32_t count_keys_absent = 0; + for (i = 0; i < TOTAL_INSERT; i++) { + /* Identify missing keys between 0 and 1M */ + if (found[i] == 0) + keys_absent[count_keys_absent++] = i; + } + + /* Find keys that will not be on the shift path */ + uint32_t iter; + const void *next_key; + void *next_data; + uint32_t count = 0; + for (i = 0; i < TOTAL_INSERT / 8; i++) { + /* Check bucket for no keys shifted to alternate locations */ + if (scanned_bkts[i] == 0) { + iter = i * 8; + while (rte_hash_iterate(tbl_rwc_test_param.h, + &next_key, &next_data, &iter) >= 0) { + + /* Check if key belongs to the current bucket */ + if (i >= (iter-1)/8) + keys_non_shift_path[count++] + = *(const uint32_t *)next_key; + else + break; + } + } + } + + tbl_rwc_test_param.count_keys_no_ks = count_keys_no_ks; + tbl_rwc_test_param.count_keys_ks = count_keys_ks; + tbl_rwc_test_param.count_keys_absent = count_keys_absent; + tbl_rwc_test_param.count_keys_non_shift_path = count; + + printf("\nCount of keys NOT causing shifting of existing keys to " + "alternate location: %d\n", tbl_rwc_test_param.count_keys_no_ks); + printf("\nCount of keys causing shifting of existing keys to alternate " + "locations: %d\n\n", tbl_rwc_test_param.count_keys_ks); + printf("Count of absent keys that will never be added to the hash " + "table: %d\n\n", tbl_rwc_test_param.count_keys_absent); + printf("Count of keys likely to be on the shift path: %d\n\n", + tbl_rwc_test_param.count_keys_shift_path); + printf("Count of keys not likely to be on the shift path: %d\n\n", + tbl_rwc_test_param.count_keys_non_shift_path); + + rte_free(found); + rte_hash_free(tbl_rwc_test_param.h); + return 0; + +err: + rte_free(keys); + rte_free(keys_no_ks); + rte_free(keys_ks); + rte_free(keys_absent); + rte_free(found); + rte_free(tbl_rwc_test_param.keys_shift_path); + rte_free(scanned_bkts); + return -1; +} + +static int +init_params(int rwc_lf, int use_jhash) +{ + struct rte_hash *handle; + + struct rte_hash_parameters hash_params = { + .entries = TOTAL_ENTRY, + .key_len = sizeof(uint32_t), + .hash_func_init_val = 0, + .socket_id = rte_socket_id(), + }; + + if (use_jhash) + hash_params.hash_func = rte_jhash; + else + hash_params.hash_func = rte_hash_crc; + + if (rwc_lf) + hash_params.extra_flag = + RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF | + RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD; + else if (htm) + hash_params.extra_flag = + RTE_HASH_EXTRA_FLAGS_TRANS_MEM_SUPPORT | + RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY | + RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD; + else + hash_params.extra_flag = + RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY | + RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD; + + hash_params.name = "tests"; + + handle = rte_hash_create(&hash_params); + if (handle == NULL) { + printf("hash creation failed"); + return -1; + } + + tbl_rwc_test_param.h = handle; + return 0; +} + +static int +test_rwc_reader(__attribute__((unused)) void *arg) +{ + uint32_t i; + int ret; + uint64_t begin, cycles; + uint32_t loop_cnt = 0; + uint8_t read_type = (uint8_t)((uintptr_t)arg); + uint32_t read_cnt; + uint32_t *keys; + + if (read_type == READ_FAIL) { + keys = tbl_rwc_test_param.keys_absent; + read_cnt = tbl_rwc_test_param.count_keys_absent; + } else if (read_type == READ_PASS_NO_KEY_SHIFTS) { + keys = tbl_rwc_test_param.keys_no_ks; + read_cnt = tbl_rwc_test_param.count_keys_no_ks; + } else if (read_type == READ_PASS_SHIFT_PATH) { + keys = tbl_rwc_test_param.keys_shift_path; + read_cnt = tbl_rwc_test_param.count_keys_shift_path; + } else { + keys = tbl_rwc_test_param.keys_non_shift_path; + read_cnt = tbl_rwc_test_param.count_keys_non_shift_path; + } + + begin = rte_rdtsc_precise(); + do { + for (i = 0; i < read_cnt; i++) { + ret = rte_hash_lookup(tbl_rwc_test_param.h, keys + i); + if ((read_type == READ_FAIL && ret != -ENOENT) + || (read_type != READ_FAIL && ret == -ENOENT)) { + printf("lookup failed! %"PRIu32"\n", keys[i]); + return -1; + } + } + loop_cnt++; + } while (!writer_done); + + cycles = rte_rdtsc_precise() - begin; + rte_atomic64_add(&gread_cycles, cycles); + rte_atomic64_add(&greads, i*loop_cnt); + return 0; +} + +static int +write_keys(uint8_t key_shift) +{ + uint32_t i; + int ret; + uint32_t key_cnt; + uint32_t *keys; + if (key_shift) { + key_cnt = tbl_rwc_test_param.count_keys_ks; + keys = tbl_rwc_test_param.keys_ks; + } else { + key_cnt = tbl_rwc_test_param.count_keys_no_ks; + keys = tbl_rwc_test_param.keys_no_ks; + } + for (i = 0; i < key_cnt; i++) { + ret = rte_hash_add_key(tbl_rwc_test_param.h, keys + i); + if (!key_shift && ret < 0) { + printf("writer failed %"PRIu32"\n", i); + return -1; + } + } + return 0; +} + +static int +test_rwc_multi_writer(__attribute__((unused)) void *arg) +{ + uint32_t i, offset; + uint32_t pos_core = (uint32_t)((uintptr_t)arg); + offset = pos_core * tbl_rwc_test_param.single_insert; + for (i = offset; i < offset + tbl_rwc_test_param.single_insert; i++) + rte_hash_add_key(tbl_rwc_test_param.h, + tbl_rwc_test_param.keys_ks + i); + multi_writer_done[pos_core] = 1; + return 0; +} + +/* + * Test lookup perf: + * Reader(s) lookup keys present in the table. + */ +static int +test_hash_add_no_ks_lookup_pass(struct rwc_perf *rwc_perf_results, int rwc_lf) +{ + unsigned int n; + uint64_t i; + int use_jhash = 0; + uint8_t key_shift = 0; + uint8_t read_type = READ_PASS_NO_KEY_SHIFTS; + + rte_atomic64_init(&greads); + rte_atomic64_init(&gread_cycles); + + if (init_params(rwc_lf, use_jhash) != 0) + goto err; + printf("\nTest: Hash add - no key-shifts, read - pass\n"); + for (n = 0; n < num_test; n++) { + unsigned int tot_lcore = rte_lcore_count(); + if (tot_lcore < rwc_core_cnt[n] + 1) + goto finish; + + printf("\nNumber of readers: %u\n", rwc_core_cnt[n]); + + rte_atomic64_clear(&greads); + rte_atomic64_clear(&gread_cycles); + + rte_hash_reset(tbl_rwc_test_param.h); + writer_done = 0; + if (write_keys(key_shift) < 0) + goto err; + writer_done = 1; + for (i = 1; i <= rwc_core_cnt[n]; i++) + rte_eal_remote_launch(test_rwc_reader, + (void *)(uintptr_t)read_type, + enabled_core_ids[i]); + rte_eal_mp_wait_lcore(); + + for (i = 1; i <= rwc_core_cnt[n]; i++) + if (lcore_config[i].ret < 0) + goto err; + + unsigned long long cycles_per_lookup = + rte_atomic64_read(&gread_cycles) / + rte_atomic64_read(&greads); + rwc_perf_results->w_no_ks_r_pass[n] = cycles_per_lookup; + printf("Cycles per lookup: %llu\n", cycles_per_lookup); + } + +finish: + rte_hash_free(tbl_rwc_test_param.h); + return 0; + +err: + rte_hash_free(tbl_rwc_test_param.h); + return -1; +} + +/* + * Test lookup perf: + * Reader(s) lookup keys absent in the table while + * 'Main' thread adds with no key-shifts. + */ +static int +test_hash_add_no_ks_lookup_fail(struct rwc_perf *rwc_perf_results, int rwc_lf) +{ + unsigned int n; + uint64_t i; + int use_jhash = 0; + uint8_t key_shift = 0; + uint8_t read_type = READ_FAIL; + int ret; + + rte_atomic64_init(&greads); + rte_atomic64_init(&gread_cycles); + + if (init_params(rwc_lf, use_jhash) != 0) + goto err; + printf("\nTest: Hash add - no key-shifts, Hash lookup - fail\n"); + for (n = 0; n < num_test; n++) { + unsigned int tot_lcore = rte_lcore_count(); + if (tot_lcore < rwc_core_cnt[n] + 1) + goto finish; + + printf("\nNumber of readers: %u\n", rwc_core_cnt[n]); + + rte_atomic64_clear(&greads); + rte_atomic64_clear(&gread_cycles); + + rte_hash_reset(tbl_rwc_test_param.h); + writer_done = 0; + + for (i = 1; i <= rwc_core_cnt[n]; i++) + rte_eal_remote_launch(test_rwc_reader, + (void *)(uintptr_t)read_type, + enabled_core_ids[i]); + ret = write_keys(key_shift); + writer_done = 1; + rte_eal_mp_wait_lcore(); + + if (ret < 0) + goto err; + for (i = 1; i <= rwc_core_cnt[n]; i++) + if (lcore_config[i].ret < 0) + goto err; + + unsigned long long cycles_per_lookup = + rte_atomic64_read(&gread_cycles) / + rte_atomic64_read(&greads); + rwc_perf_results->w_no_ks_r_fail[n] = cycles_per_lookup; + printf("Cycles per lookup: %llu\n", cycles_per_lookup); + } + +finish: + rte_hash_free(tbl_rwc_test_param.h); + return 0; + +err: + rte_hash_free(tbl_rwc_test_param.h); + return -1; +} + +/* + * Test lookup perf: + * Reader(s) lookup keys present in the table and not likely to be on the + * shift path while 'Main' thread adds keys causing key-shifts. + */ +static int +test_hash_add_ks_lookup_pass_non_sp(struct rwc_perf *rwc_perf_results, + int rwc_lf) +{ + unsigned int n; + uint64_t i; + int use_jhash = 0; + int ret; + uint8_t key_shift; + uint8_t read_type = READ_PASS_NON_SHIFT_PATH; + + rte_atomic64_init(&greads); + rte_atomic64_init(&gread_cycles); + + if (init_params(rwc_lf, use_jhash) != 0) + goto err; + printf("\nTest: Hash add - key shift, Hash lookup - pass" + " (non-shift-path)\n"); + for (n = 0; n < num_test; n++) { + unsigned int tot_lcore = rte_lcore_count(); + if (tot_lcore < rwc_core_cnt[n] + 1) + goto finish; + + printf("\nNumber of readers: %u\n", rwc_core_cnt[n]); + + rte_atomic64_clear(&greads); + rte_atomic64_clear(&gread_cycles); + + rte_hash_reset(tbl_rwc_test_param.h); + writer_done = 0; + key_shift = 0; + if (write_keys(key_shift) < 0) + goto err; + for (i = 1; i <= rwc_core_cnt[n]; i++) + rte_eal_remote_launch(test_rwc_reader, + (void *)(uintptr_t)read_type, + enabled_core_ids[i]); + key_shift = 1; + ret = write_keys(key_shift); + writer_done = 1; + rte_eal_mp_wait_lcore(); + + if (ret < 0) + goto err; + for (i = 1; i <= rwc_core_cnt[n]; i++) + if (lcore_config[i].ret < 0) + goto err; + + unsigned long long cycles_per_lookup = + rte_atomic64_read(&gread_cycles) / + rte_atomic64_read(&greads); + rwc_perf_results->w_ks_r_pass_nsp[n] = cycles_per_lookup; + printf("Cycles per lookup: %llu\n", cycles_per_lookup); + } + +finish: + rte_hash_free(tbl_rwc_test_param.h); + return 0; + +err: + rte_hash_free(tbl_rwc_test_param.h); + return -1; +} + +/* + * Test lookup perf: + * Reader(s) lookup keys present in the table and likely on the shift-path while + * 'Main' thread adds keys causing key-shifts. + */ +static int +test_hash_add_ks_lookup_pass_sp(struct rwc_perf *rwc_perf_results, int rwc_lf) +{ + unsigned int n; + uint64_t i; + int use_jhash = 0; + int ret; + uint8_t key_shift; + uint8_t read_type = READ_PASS_SHIFT_PATH; + + rte_atomic64_init(&greads); + rte_atomic64_init(&gread_cycles); + + if (init_params(rwc_lf, use_jhash) != 0) + goto err; + printf("\nTest: Hash add - key shift, Hash lookup - pass (shift-path)" + "\n"); + + for (n = 0; n < num_test; n++) { + unsigned int tot_lcore = rte_lcore_count(); + if (tot_lcore < rwc_core_cnt[n]) + goto finish; + + printf("\nNumber of readers: %u\n", rwc_core_cnt[n]); + rte_atomic64_clear(&greads); + rte_atomic64_clear(&gread_cycles); + + rte_hash_reset(tbl_rwc_test_param.h); + writer_done = 0; + key_shift = 0; + if (write_keys(key_shift) < 0) + goto err; + for (i = 1; i <= rwc_core_cnt[n]; i++) + rte_eal_remote_launch(test_rwc_reader, + (void *)(uintptr_t)read_type, + enabled_core_ids[i]); + key_shift = 1; + ret = write_keys(key_shift); + writer_done = 1; + rte_eal_mp_wait_lcore(); + + if (ret < 0) + goto err; + for (i = 1; i <= rwc_core_cnt[n]; i++) + if (lcore_config[i].ret < 0) + goto err; + + unsigned long long cycles_per_lookup = + rte_atomic64_read(&gread_cycles) / + rte_atomic64_read(&greads); + rwc_perf_results->w_ks_r_pass_sp[n] = cycles_per_lookup; + printf("Cycles per lookup: %llu\n", cycles_per_lookup); + } + +finish: + rte_hash_free(tbl_rwc_test_param.h); + return 0; + +err: + rte_hash_free(tbl_rwc_test_param.h); + return -1; +} + +/* + * Test lookup perf: + * Reader(s) lookup keys absent in the table while + * 'Main' thread adds keys causing key-shifts. + */ +static int +test_hash_add_ks_lookup_fail(struct rwc_perf *rwc_perf_results, int rwc_lf) +{ + unsigned int n; + uint64_t i; + int use_jhash = 0; + int ret; + uint8_t key_shift; + uint8_t read_type = READ_FAIL; + + rte_atomic64_init(&greads); + rte_atomic64_init(&gread_cycles); + + if (init_params(rwc_lf, use_jhash) != 0) + goto err; + printf("\nTest: Hash add - key shift, Hash lookup - fail\n"); + for (n = 0; n < num_test; n++) { + unsigned int tot_lcore = rte_lcore_count(); + if (tot_lcore < rwc_core_cnt[n] + 1) + goto finish; + + printf("\nNumber of readers: %u\n", rwc_core_cnt[n]); + + rte_atomic64_clear(&greads); + rte_atomic64_clear(&gread_cycles); + + rte_hash_reset(tbl_rwc_test_param.h); + writer_done = 0; + key_shift = 0; + if (write_keys(key_shift) < 0) + goto err; + for (i = 1; i <= rwc_core_cnt[n]; i++) + rte_eal_remote_launch(test_rwc_reader, + (void *)(uintptr_t)read_type, + enabled_core_ids[i]); + key_shift = 1; + ret = write_keys(key_shift); + writer_done = 1; + rte_eal_mp_wait_lcore(); + + if (ret < 0) + goto err; + for (i = 1; i <= rwc_core_cnt[n]; i++) + if (lcore_config[i].ret < 0) + goto err; + + unsigned long long cycles_per_lookup = + rte_atomic64_read(&gread_cycles) / + rte_atomic64_read(&greads); + rwc_perf_results->w_ks_r_fail[n] = cycles_per_lookup; + printf("Cycles per lookup: %llu\n", cycles_per_lookup); + } + +finish: + rte_hash_free(tbl_rwc_test_param.h); + return 0; + +err: + rte_hash_free(tbl_rwc_test_param.h); + return -1; +} + +/* + * Test lookup perf: + * Reader(s) lookup keys present in the table and likely on the shift-path while + * Writers add keys causing key-shifts. + */ +static int +test_hash_multi_add_lookup(struct rwc_perf *rwc_perf_results, int rwc_lf) +{ + unsigned int n, m; + uint64_t i; + int use_jhash = 0; + uint8_t key_shift; + uint8_t read_type = READ_PASS_SHIFT_PATH; + + rte_atomic64_init(&greads); + rte_atomic64_init(&gread_cycles); + + if (init_params(rwc_lf, use_jhash) != 0) + goto err; + printf("\nTest: Multi-add-lookup\n"); + uint8_t pos_core; + for (m = 1; m < num_test; m++) { + /* Calculate keys added by each writer */ + tbl_rwc_test_param.single_insert = + tbl_rwc_test_param.count_keys_ks / rwc_core_cnt[m]; + + for (n = 0; n < num_test; n++) { + unsigned int tot_lcore = rte_lcore_count(); + if (tot_lcore < rwc_core_cnt[n] + rwc_core_cnt[m] + 1) + goto finish; + + printf("\nNumber of writers: %u", rwc_core_cnt[m]); + printf("\nNumber of readers: %u\n", rwc_core_cnt[n]); + + rte_atomic64_clear(&greads); + rte_atomic64_clear(&gread_cycles); + + rte_hash_reset(tbl_rwc_test_param.h); + writer_done = 0; + for (i = 0; i < 4; i++) + multi_writer_done[i] = 0; + key_shift = 0; + if (write_keys(key_shift) < 0) + goto err; + + /* Launch reader(s) */ + for (i = 1; i <= rwc_core_cnt[n]; i++) + rte_eal_remote_launch(test_rwc_reader, + (void *)(uintptr_t)read_type, + enabled_core_ids[i]); + key_shift = 1; + pos_core = 0; + + /* Launch writers */ + for (; i <= rwc_core_cnt[m] + rwc_core_cnt[n]; i++) { + rte_eal_remote_launch(test_rwc_multi_writer, + (void *)(uintptr_t)pos_core, + enabled_core_ids[i]); + pos_core++; + } + + /* Wait for writers to complete */ + for (i = 0; i < rwc_core_cnt[m]; i++) + while + (multi_writer_done[i] == 0); + writer_done = 1; + + rte_eal_mp_wait_lcore(); + + for (i = 1; i <= rwc_core_cnt[n]; i++) + if (lcore_config[i].ret < 0) + goto err; + + unsigned long long cycles_per_lookup = + rte_atomic64_read(&gread_cycles) / + rte_atomic64_read(&greads); + rwc_perf_results->multi_rw[m][n] = cycles_per_lookup; + printf("Cycles per lookup: %llu\n", cycles_per_lookup); + } + } + +finish: + rte_hash_free(tbl_rwc_test_param.h); + return 0; + +err: + rte_hash_free(tbl_rwc_test_param.h); + return -1; +} + +static int +test_hash_readwrite_lf_main(void) +{ + /* + * Variables used to choose different tests. + * rwc_lf indicates if read-write concurrency lock-free support is + * enabled. + * htm indicates if Hardware transactional memory support is enabled. + */ + int rwc_lf = 0; + int use_jhash = 0; + num_test = NUM_TEST; + if (rte_lcore_count() == 1) { + printf("More than one lcore is required " + "to do read write lock-free concurrency test\n"); + return -1; + } + + setlocale(LC_NUMERIC, ""); + + if (rte_tm_supported()) + htm = 1; + else + htm = 0; + + if (init_params(rwc_lf, use_jhash) != 0) + return -1; + if (generate_keys() != 0) + return -1; + if (get_enabled_cores_list() != 0) + return -1; + + if (RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF) { + rwc_lf = 1; + printf("Test lookup with read-write concurrency lock free support" + " enabled\n"); + if (test_hash_add_no_ks_lookup_pass(&rwc_lf_results, rwc_lf) + < 0) + return -1; + if (test_hash_add_no_ks_lookup_fail(&rwc_lf_results, rwc_lf) + < 0) + return -1; + if (test_hash_add_ks_lookup_pass_non_sp(&rwc_lf_results, rwc_lf) + < 0) + return -1; + if (test_hash_add_ks_lookup_pass_sp(&rwc_lf_results, rwc_lf) + < 0) + return -1; + if (test_hash_add_ks_lookup_fail(&rwc_lf_results, rwc_lf) < 0) + return -1; + if (test_hash_multi_add_lookup(&rwc_lf_results, rwc_lf) < 0) + return -1; + } + printf("\nTest lookup with read-write concurrency lock free support" + " disabled\n"); + rwc_lf = 0; + if (!htm) { + printf("With HTM Disabled\n"); + if (!RUN_WITH_HTM_DISABLED) { + printf("Enable RUN_WITH_HTM_DISABLED to test with" + " lock-free disabled"); + goto results; + } + } else + printf("With HTM Enabled\n"); + if (test_hash_add_no_ks_lookup_pass(&rwc_non_lf_results, rwc_lf) < 0) + return -1; + if (test_hash_add_no_ks_lookup_fail(&rwc_non_lf_results, rwc_lf) < 0) + return -1; + if (test_hash_add_ks_lookup_pass_non_sp(&rwc_non_lf_results, rwc_lf) + < 0) + return -1; + if (test_hash_add_ks_lookup_pass_sp(&rwc_non_lf_results, rwc_lf) < 0) + return -1; + if (test_hash_add_ks_lookup_fail(&rwc_non_lf_results, rwc_lf) < 0) + return -1; + if (test_hash_multi_add_lookup(&rwc_non_lf_results, rwc_lf) < 0) + return -1; +results: + printf("\n\t\t\t\t\t\t********** Results summary **********\n\n"); + printf("_______\t\t_______\t\t_________\t___\t\t_________\t\t\t\t\t\t" + "_________________\n"); + int i, j; + printf("Writers\t\tReaders\t\tLock-free\tHTM\t\tTest-case\t\t\t\t\t\t" + "Cycles per lookup\n"); + printf("_______\t\t_______\t\t_________\t___\t\t_________\t\t\t\t\t\t" + "_________________\n"); + for (i = 0; i < NUM_TEST; i++) { + printf("%u\t\t%u\t\t", 1, rwc_core_cnt[i]); + printf("Enabled\t\t"); + printf("N/A\t\t"); + printf("Hash add - no key-shifts, lookup - pass\t\t\t\t%u\n\t\t" + "\t\t\t\t\t\t", rwc_lf_results.w_no_ks_r_pass[i]); + printf("Hash add - no key-shifts, lookup - fail\t\t\t\t%u\n\t\t" + "\t\t\t\t\t\t", rwc_lf_results.w_no_ks_r_fail[i]); + printf("Hash add - key-shifts, lookup - pass (non-shift-path)\t" + "\t%u\n\t\t\t\t\t\t\t\t", + rwc_lf_results.w_ks_r_pass_nsp[i]); + printf("Hash add - key-shifts, lookup - pass (shift-path)\t\t%u" + "\n\t\t\t\t\t\t\t\t", rwc_lf_results.w_ks_r_pass_sp[i]); + printf("Hash add - key-shifts, Hash lookup fail\t\t\t\t%u\n\n" + "\t\t\t\t", rwc_lf_results.w_ks_r_fail[i]); + + printf("Disabled\t"); + if (htm) + printf("Enabled\t\t"); + else + printf("Disabled\t"); + printf("Hash add - no key-shifts, lookup - pass\t\t\t\t%u\n\t\t" + "\t\t\t\t\t\t", rwc_non_lf_results.w_no_ks_r_pass[i]); + printf("Hash add - no key-shifts, lookup - fail\t\t\t\t%u\n\t\t" + "\t\t\t\t\t\t", rwc_non_lf_results.w_no_ks_r_fail[i]); + printf("Hash add - key-shifts, lookup - pass (non-shift-path)\t" + "\t%u\n\t\t\t\t\t\t\t\t", + rwc_non_lf_results.w_ks_r_pass_nsp[i]); + printf("Hash add - key-shifts, lookup - pass (shift-path)\t\t%u" + "\n\t\t\t\t\t\t\t\t", + rwc_non_lf_results.w_ks_r_pass_sp[i]); + printf("Hash add - key-shifts, Hash lookup fail\t\t\t\t%u\n", + rwc_non_lf_results.w_ks_r_fail[i]); + + printf("_______\t\t_______\t\t_________\t___\t\t_________\t\t\t\t" + "\t\t_________________\n"); + } + + for (i = 1; i < NUM_TEST; i++) { + for (j = 0; j < NUM_TEST; j++) { + printf("%u", rwc_core_cnt[i]); + printf("\t\t%u\t\t", rwc_core_cnt[j]); + printf("Enabled\t\t"); + printf("N/A\t\t"); + printf("Multi-add-lookup\t\t\t\t\t\t%u\n\n\t\t\t\t", + rwc_lf_results.multi_rw[i][j]); + printf("Disabled\t"); + if (htm) + printf("Enabled\t\t"); + else + printf("Disabled\t"); + printf("Multi-add-lookup\t\t\t\t\t\t%u\n", + rwc_non_lf_results.multi_rw[i][j]); + + printf("_______\t\t_______\t\t_________\t___\t\t" + "_________\t\t\t\t\t\t_________________\n"); + } + } + + rte_free(tbl_rwc_test_param.keys); + rte_free(tbl_rwc_test_param.keys_no_ks); + rte_free(tbl_rwc_test_param.keys_ks); + rte_free(tbl_rwc_test_param.keys_absent); + rte_free(tbl_rwc_test_param.keys_shift_path); + rte_free(scanned_bkts); + return 0; +} + +REGISTER_TEST_COMMAND(hash_readwrite_lf_autotest, test_hash_readwrite_lf_main);