DPDK patches and discussions
 help / color / mirror / Atom feed
From: Dharmik Thakkar <Dharmik.Thakkar@arm.com>
To: "Wang, Yipeng1" <yipeng1.wang@intel.com>
Cc: "Gobriel, Sameh" <sameh.gobriel@intel.com>,
	"Richardson, Bruce" <bruce.richardson@intel.com>,
	Ray Kinsella <mdr@ashroe.eu>, Neil Horman <nhorman@tuxdriver.com>,
	"dev@dpdk.org" <dev@dpdk.org>, nd <nd@arm.com>
Subject: Re: [dpdk-dev] [PATCH v5 2/4] lib/hash: integrate RCU QSBR
Date: Wed, 21 Oct 2020 04:52:58 +0000	[thread overview]
Message-ID: <72DB22D2-34AF-41BD-ACAC-7C2DF3121BCC@arm.com> (raw)
In-Reply-To: <BYAPR11MB34948CA2C858032FF6BFBB1EC31C0@BYAPR11MB3494.namprd11.prod.outlook.com>



> On Oct 20, 2020, at 9:42 PM, Wang, Yipeng1 <yipeng1.wang@intel.com> wrote:
> 
>> -----Original Message-----
>> From: Dharmik Thakkar <dharmik.thakkar@arm.com>
>> Sent: Tuesday, October 20, 2020 9:13 AM
>> To: Wang, Yipeng1 <yipeng1.wang@intel.com>; Gobriel, Sameh
>> <sameh.gobriel@intel.com>; Richardson, Bruce
>> <bruce.richardson@intel.com>; Ray Kinsella <mdr@ashroe.eu>; Neil Horman
>> <nhorman@tuxdriver.com>
>> Cc: dev@dpdk.org; nd@arm.com; Dharmik Thakkar
>> <dharmik.thakkar@arm.com>
>> Subject: [PATCH v5 2/4] lib/hash: integrate RCU QSBR
>> 
>> Currently, users have to use external RCU mechanisms to free resources
>> when using lock free hash algorithm.
>> 
>> Integrate RCU QSBR process to make it easier for the applications to use lock
>> free algorithm.
>> Refer to RCU documentation to understand various aspects of integrating RCU
>> library into other libraries.
>> 
>> Suggested-by: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com>
>> Signed-off-by: Dharmik Thakkar <dharmik.thakkar@arm.com>
>> Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
>> Acked-by: Ray Kinsella <mdr@ashroe.eu>
>> ---
>> doc/guides/prog_guide/hash_lib.rst |  11 +-
>> lib/librte_hash/meson.build        |   1 +
>> lib/librte_hash/rte_cuckoo_hash.c  | 302 ++++++++++++++++++++++-------
>> lib/librte_hash/rte_cuckoo_hash.h  |   8 +
>> lib/librte_hash/rte_hash.h         |  77 +++++++-
>> lib/librte_hash/version.map        |   2 +-
>> 6 files changed, 325 insertions(+), 76 deletions(-)
>> 
>> diff --git a/doc/guides/prog_guide/hash_lib.rst
>> b/doc/guides/prog_guide/hash_lib.rst
>> index d06c7de2ead1..63e183ed1f08 100644
>> --- a/doc/guides/prog_guide/hash_lib.rst
>> +++ b/doc/guides/prog_guide/hash_lib.rst
>> @@ -102,6 +102,9 @@ For concurrent writes, and concurrent reads and
>> writes the following flag values
>> *  If the 'do not free on delete' (RTE_HASH_EXTRA_FLAGS_NO_FREE_ON_DEL)
>> flag is set, the position of the entry in the hash table is not freed upon calling
>> delete(). This flag is enabled
>>    by default when the lock free read/write concurrency flag is set. The
>> application should free the position after all the readers have stopped
>> referencing the position.
>>    Where required, the application can make use of RCU mechanisms to
>> determine when the readers have stopped referencing the position.
>> +   RCU QSBR process is integrated within the Hash library for safe freeing of
>> the position. Application has certain responsibilities while using this feature.
>> +   Please refer to resource reclamation framework of :ref:`RCU library
>> <RCU_Library>` for more details.
> 
> [Yipeng]: Maybe also add: rte_hash_rcu_qsbr_add() need to be called to use the embedded RCU mechanism.
> Just to give user a pointer to which API to look.

Copy.

>> +
>> 
>> Extendable Bucket Functionality support
>> ----------------------------------------
>> @@ -109,8 +112,8 @@ An extra flag is used to enable this functionality (flag
>> is not set by default).
>> in the very unlikely case due to excessive hash collisions that a key has failed
>> to be inserted, the hash table bucket is extended with a linked  list to insert
>> these failed keys. This feature is important for the workloads (e.g. telco
>> workloads) that need to insert up to 100% of the  hash table size and can't
>> tolerate any key insertion failure (even if very few).
>> -Please note that with the 'lock free read/write concurrency' flag enabled,
>> users need to call 'rte_hash_free_key_with_position' API in order to free the
>> empty buckets and -deleted keys, to maintain the 100% capacity guarantee.
>> +Please note that with the 'lock free read/write concurrency' flag
>> +enabled, users need to call 'rte_hash_free_key_with_position' API or
>> configure integrated RCU QSBR (or use external RCU mechanisms) in order to
>> free the empty buckets and deleted keys, to maintain the 100% capacity
>> guarantee.
>> 
>> Implementation Details (non Extendable Bucket Case)
>> ---------------------------------------------------
>> @@ -172,7 +175,7 @@ Example of deletion:
>> Similar to lookup, the key is searched in its primary and secondary buckets. If
>> the key is found, the  entry is marked as empty. If the hash table was
>> configured with 'no free on delete' or 'lock free read/write concurrency',  the
>> position of the key is not freed. It is the responsibility of the user to free the
>> position after -readers are not referencing the position anymore.
>> +readers are not referencing the position anymore. User can configure
>> +integrated RCU QSBR or use external RCU mechanisms to safely free the
>> +position on delete
>> 
>> 
>> Implementation Details (with Extendable Bucket) @@ -286,6 +289,8 @@ The
>> flow table operations on the application side are described below:
>> *   Free flow: Free flow key position. If 'no free on delete' or 'lock-free
>> read/write concurrency' flags are set,
>>     wait till the readers are not referencing the position returned during
>> add/delete flow and then free the position.
>>     RCU mechanisms can be used to find out when the readers are not
>> referencing the position anymore.
>> +    RCU QSBR process is integrated within the Hash library for safe freeing of
>> the position. Application has certain responsibilities while using this feature.
>> +    Please refer to resource reclamation framework of :ref:`RCU library
>> <RCU_Library>` for more details.
>> 
>> *   Lookup flow: Lookup for the flow key in the hash.
>>     If the returned position is valid (flow lookup hit), use the returned position
>> to access the flow entry in the flow table.
>> diff --git a/lib/librte_hash/meson.build b/lib/librte_hash/meson.build index
>> 6ab46ae9d768..0977a63fd279 100644
>> --- a/lib/librte_hash/meson.build
>> +++ b/lib/librte_hash/meson.build
>> @@ -10,3 +10,4 @@ headers = files('rte_crc_arm64.h',
>> 
>> sources = files('rte_cuckoo_hash.c', 'rte_fbk_hash.c')  deps += ['ring']
>> +deps += ['rcu']
>> diff --git a/lib/librte_hash/rte_cuckoo_hash.c
>> b/lib/librte_hash/rte_cuckoo_hash.c
>> index aad0c965be5e..b9e4d82a0c14 100644
>> --- a/lib/librte_hash/rte_cuckoo_hash.c
>> +++ b/lib/librte_hash/rte_cuckoo_hash.c
>> @@ -52,6 +52,11 @@ static struct rte_tailq_elem rte_hash_tailq = {  };
>> EAL_REGISTER_TAILQ(rte_hash_tailq)
>> 
>> +struct __rte_hash_rcu_dq_entry {
>> +	uint32_t key_idx;
>> +	uint32_t ext_bkt_idx; /**< Extended bkt index */ };
>> +
>> struct rte_hash *
>> rte_hash_find_existing(const char *name)  { @@ -210,7 +215,10 @@
>> rte_hash_create(const struct rte_hash_parameters *params)
>> 
>> 	if (params->extra_flag &
>> RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF) {
>> 		readwrite_concur_lf_support = 1;
>> -		/* Enable not freeing internal memory/index on delete */
>> +		/* Enable not freeing internal memory/index on delete.
>> +		 * If internal RCU is enabled, freeing of internal memory/index
>> +		 * is done on delete
>> +		 */
>> 		no_free_on_del = 1;
>> 	}
>> 
>> @@ -505,6 +513,10 @@ rte_hash_free(struct rte_hash *h)
>> 
>> 	rte_mcfg_tailq_write_unlock();
>> 
>> +	/* RCU clean up. */
>> +	if (h->dq)
>> +		rte_rcu_qsbr_dq_delete(h->dq);
>> +
>> 	if (h->use_local_cache)
>> 		rte_free(h->local_free_slots);
>> 	if (h->writer_takes_lock)
>> @@ -607,11 +619,21 @@ void
>> rte_hash_reset(struct rte_hash *h)
>> {
>> 	uint32_t tot_ring_cnt, i;
>> +	unsigned int pending;
>> 
>> 	if (h == NULL)
>> 		return;
>> 
>> 	__hash_rw_writer_lock(h);
>> +
>> +	/* RCU QSBR clean up. */
>> +	if (h->dq) {
>> +		/* Reclaim all the resources */
>> +		rte_rcu_qsbr_dq_reclaim(h->dq, ~0, NULL, &pending, NULL);
>> +		if (pending != 0)
>> +			RTE_LOG(ERR, HASH, "RCU reclaim all resources
>> failed\n");
>> +	}
>> +
>> 	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;
>> @@ -952,6 +974,37 @@ rte_hash_cuckoo_make_space_mw(const struct
>> rte_hash *h,
>> 	return -ENOSPC;
>> }
>> 
>> +static inline uint32_t
>> +alloc_slot(const struct rte_hash *h, struct lcore_cache
>> +*cached_free_slots) {
>> +	unsigned int  n_slots;
>> +	uint32_t slot_id;
> 
> [Yipeng]: Blank line after variable declaration.

Copy.

> 
>> +	if (h->use_local_cache) {
>> +		/* Try to get a free slot from the local cache */
>> +		if (cached_free_slots->len == 0) {
>> +			/* Need to get another burst of free slots from global
>> ring */
>> +			n_slots = rte_ring_mc_dequeue_burst_elem(h-
>>> free_slots,
>> +					cached_free_slots->objs,
>> +					sizeof(uint32_t),
>> +					LCORE_CACHE_SIZE, NULL);
>> +			if (n_slots == 0)
>> +				return EMPTY_SLOT;
>> +
>> +			cached_free_slots->len += n_slots;
>> +		}
>> +
>> +		/* Get a free slot from the local cache */
>> +		cached_free_slots->len--;
>> +		slot_id = cached_free_slots->objs[cached_free_slots->len];
>> +	} else {
>> +		if (rte_ring_sc_dequeue_elem(h->free_slots, &slot_id,
>> +						sizeof(uint32_t)) != 0)
>> +			return EMPTY_SLOT;
>> +	}
>> +
>> +	return slot_id;
>> +}
>> +
>> static inline int32_t
>> __rte_hash_add_key_with_hash(const struct rte_hash *h, const void *key,
>> 						hash_sig_t sig, void *data)
>> @@ -963,7 +1016,6 @@ __rte_hash_add_key_with_hash(const struct
>> rte_hash *h, const void *key,
>> 	uint32_t ext_bkt_id = 0;
>> 	uint32_t slot_id;
>> 	int ret;
>> -	unsigned n_slots;
>> 	unsigned lcore_id;
>> 	unsigned int i;
>> 	struct lcore_cache *cached_free_slots = NULL; @@ -1001,28
>> +1053,20 @@ __rte_hash_add_key_with_hash(const struct rte_hash *h,
>> const void *key,
>> 	if (h->use_local_cache) {
>> 		lcore_id = rte_lcore_id();
>> 		cached_free_slots = &h->local_free_slots[lcore_id];
>> -		/* Try to get a free slot from the local cache */
>> -		if (cached_free_slots->len == 0) {
>> -			/* Need to get another burst of free slots from global
>> ring */
>> -			n_slots = rte_ring_mc_dequeue_burst_elem(h-
>>> free_slots,
>> -					cached_free_slots->objs,
>> -					sizeof(uint32_t),
>> -					LCORE_CACHE_SIZE, NULL);
>> -			if (n_slots == 0) {
>> -				return -ENOSPC;
>> -			}
>> -
>> -			cached_free_slots->len += n_slots;
>> +	}
>> +	slot_id = alloc_slot(h, cached_free_slots);
>> +	if (slot_id == EMPTY_SLOT) {
>> +		if (h->dq) {
>> +			__hash_rw_writer_lock(h);
>> +			ret = rte_rcu_qsbr_dq_reclaim(h->dq,
>> +					h->hash_rcu_cfg->max_reclaim_size,
>> +					NULL, NULL, NULL);
>> +			__hash_rw_writer_unlock(h);
>> +			if (ret == 0)
>> +				slot_id = alloc_slot(h, cached_free_slots);
>> 		}
>> -
>> -		/* Get a free slot from the local cache */
>> -		cached_free_slots->len--;
>> -		slot_id = cached_free_slots->objs[cached_free_slots->len];
>> -	} else {
>> -		if (rte_ring_sc_dequeue_elem(h->free_slots, &slot_id,
>> -						sizeof(uint32_t)) != 0) {
>> +		if (slot_id == EMPTY_SLOT)
>> 			return -ENOSPC;
>> -		}
>> 	}
>> 
>> 	new_k = RTE_PTR_ADD(keys, slot_id * h->key_entry_size); @@ -
>> 1118,8 +1162,19 @@ __rte_hash_add_key_with_hash(const struct rte_hash
>> *h, const void *key,
>> 	if (rte_ring_sc_dequeue_elem(h->free_ext_bkts, &ext_bkt_id,
>> 						sizeof(uint32_t)) != 0 ||
>> 					ext_bkt_id == 0) {
>> -		ret = -ENOSPC;
>> -		goto failure;
>> +		if (h->dq) {
>> +			if (rte_rcu_qsbr_dq_reclaim(h->dq,
>> +					h->hash_rcu_cfg->max_reclaim_size,
>> +					NULL, NULL, NULL) == 0) {
>> +				rte_ring_sc_dequeue_elem(h-
>>> free_ext_bkts,
>> +							 &ext_bkt_id,
>> +							 sizeof(uint32_t));
>> +			}
>> +		}
>> +		if (ext_bkt_id == 0) {
>> +			ret = -ENOSPC;
>> +			goto failure;
>> +		}
>> 	}
>> 
>> 	/* Use the first location of the new bucket */ @@ -1395,12 +1450,12
>> @@ rte_hash_lookup_data(const struct rte_hash *h, const void *key, void
>> **data)
>> 	return __rte_hash_lookup_with_hash(h, key, rte_hash_hash(h, key),
>> data);  }
>> 
>> -static inline void
>> -remove_entry(const struct rte_hash *h, struct rte_hash_bucket *bkt,
>> unsigned i)
>> +static int
>> +free_slot(const struct rte_hash *h, uint32_t slot_id)
>> {
>> 	unsigned lcore_id, n_slots;
>> -	struct lcore_cache *cached_free_slots;
>> -
>> +	struct lcore_cache *cached_free_slots = NULL;
>> +	/* Return key indexes to free slot ring */
>> 	if (h->use_local_cache) {
>> 		lcore_id = rte_lcore_id();
>> 		cached_free_slots = &h->local_free_slots[lcore_id]; @@ -
>> 1411,18 +1466,127 @@ remove_entry(const struct rte_hash *h, struct
>> rte_hash_bucket *bkt, unsigned i)
>> 						cached_free_slots->objs,
>> 						sizeof(uint32_t),
>> 						LCORE_CACHE_SIZE, NULL);
>> -			ERR_IF_TRUE((n_slots == 0),
>> -				"%s: could not enqueue free slots in global
>> ring\n",
>> -				__func__);
>> +			RETURN_IF_TRUE((n_slots == 0), -EFAULT);
>> 			cached_free_slots->len -= n_slots;
>> 		}
>> -		/* Put index of new free slot in cache. */
>> -		cached_free_slots->objs[cached_free_slots->len] =
>> -							bkt->key_idx[i];
>> -		cached_free_slots->len++;
>> +	}
>> +
>> +	enqueue_slot_back(h, cached_free_slots, slot_id);
>> +	return 0;
>> +}
>> +
>> +static void
>> +__hash_rcu_qsbr_free_resource(void *p, void *e, unsigned int n) {
>> +	void *key_data = NULL;
>> +	int ret;
>> +	struct rte_hash_key *keys, *k;
>> +	struct rte_hash *h = (struct rte_hash *)p;
>> +	struct __rte_hash_rcu_dq_entry rcu_dq_entry =
>> +			*((struct __rte_hash_rcu_dq_entry *)e);
>> +
>> +	RTE_SET_USED(n);
>> +	keys = h->key_store;
>> +
>> +	k = (struct rte_hash_key *) ((char *)keys +
>> +				rcu_dq_entry.key_idx * h->key_entry_size);
>> +	key_data = k->pdata;
>> +	if (h->hash_rcu_cfg->free_key_data_func)
>> +		h->hash_rcu_cfg->free_key_data_func(h->hash_rcu_cfg-
>>> key_data_ptr,
>> +						    key_data);
>> +
>> +	if (h->ext_table_support && rcu_dq_entry.ext_bkt_idx !=
>> EMPTY_SLOT)
>> +		/* Recycle empty ext bkt to free list. */
>> +		rte_ring_sp_enqueue_elem(h->free_ext_bkts,
>> +			&rcu_dq_entry.ext_bkt_idx, sizeof(uint32_t));
>> +
>> +	/* Return key indexes to free slot ring */
>> +	ret = free_slot(h, rcu_dq_entry.key_idx);
>> +	if (ret < 0) {
>> +		RTE_LOG(ERR, HASH,
>> +			"%s: could not enqueue free slots in global ring\n",
>> +				__func__);
>> +	}
>> +}
>> +
>> +int
>> +rte_hash_rcu_qsbr_add(struct rte_hash *h,
>> +				struct rte_hash_rcu_config *cfg)
>> +{
>> +	struct rte_rcu_qsbr_dq_parameters params = {0};
>> +	char rcu_dq_name[RTE_RCU_QSBR_DQ_NAMESIZE];
>> +	struct rte_hash_rcu_config *hash_rcu_cfg = NULL;
>> +
>> +	const uint32_t total_entries = h->use_local_cache ?
>> +		h->entries + (RTE_MAX_LCORE - 1) * (LCORE_CACHE_SIZE - 1)
>> + 1
>> +							: h->entries + 1;
>> +
>> +	if ((h == NULL) || cfg == NULL || cfg->v == NULL) {
>> +		rte_errno = EINVAL;
>> +		return 1;
>> +	}
>> +
>> +	if (h->hash_rcu_cfg) {
>> +		rte_errno = EEXIST;
>> +		return 1;
>> +	}
>> +
>> +	hash_rcu_cfg = rte_zmalloc(NULL, sizeof(struct rte_hash_rcu_config),
>> 0);
>> +	if (hash_rcu_cfg == NULL) {
>> +		RTE_LOG(ERR, HASH, "memory allocation failed\n");
>> +		return 1;
>> +	}
>> +
>> +	if (cfg->mode == RTE_HASH_QSBR_MODE_SYNC) {
>> +		/* No other things to do. */
>> +	} else if (cfg->mode == RTE_HASH_QSBR_MODE_DQ) {
>> +		/* Init QSBR defer queue. */
>> +		snprintf(rcu_dq_name, sizeof(rcu_dq_name),
>> +					"HASH_RCU_%s", h->name);
>> +		params.name = rcu_dq_name;
>> +		params.size = cfg->dq_size;
>> +		if (params.size == 0)
>> +			params.size = total_entries;
>> +		params.trigger_reclaim_limit = cfg->trigger_reclaim_limit;
>> +		if (params.max_reclaim_size == 0)
>> +			params.max_reclaim_size =
>> RTE_HASH_RCU_DQ_RECLAIM_MAX;
>> +		params.esize = sizeof(struct __rte_hash_rcu_dq_entry);
>> +		params.free_fn = __hash_rcu_qsbr_free_resource;
>> +		params.p = h;
>> +		params.v = cfg->v;
>> +		h->dq = rte_rcu_qsbr_dq_create(&params);
>> +		if (h->dq == NULL) {
>> +			rte_free(hash_rcu_cfg);
>> +			RTE_LOG(ERR, HASH, "HASH defer queue creation
>> failed\n");
>> +			return 1;
>> +		}
>> 	} else {
>> -		rte_ring_sp_enqueue_elem(h->free_slots,
>> -				&bkt->key_idx[i], sizeof(uint32_t));
>> +		rte_free(hash_rcu_cfg);
>> +		rte_errno = EINVAL;
>> +		return 1;
>> +	}
>> +
>> +	hash_rcu_cfg->v = cfg->v;
>> +	hash_rcu_cfg->mode = cfg->mode;
>> +	hash_rcu_cfg->dq_size = params.size;
>> +	hash_rcu_cfg->trigger_reclaim_limit = params.trigger_reclaim_limit;
>> +	hash_rcu_cfg->max_reclaim_size = params.max_reclaim_size;
>> +	hash_rcu_cfg->free_key_data_func = cfg->free_key_data_func;
>> +	hash_rcu_cfg->key_data_ptr = cfg->key_data_ptr;
>> +
>> +	h->hash_rcu_cfg = hash_rcu_cfg;
>> +
>> +	return 0;
>> +}
>> +
>> +static inline void
>> +remove_entry(const struct rte_hash *h, struct rte_hash_bucket *bkt,
>> +unsigned i) {
>> +	int ret = free_slot(h, bkt->key_idx[i]);
>> +	if (ret < 0) {
>> +		RTE_LOG(ERR, HASH,
>> +			"%s: could not enqueue free slots in global ring\n",
>> +				__func__);
>> 	}
>> }
>> 
>> @@ -1521,6 +1685,8 @@ __rte_hash_del_key_with_hash(const struct
>> rte_hash *h, const void *key,
>> 	int pos;
>> 	int32_t ret, i;
>> 	uint16_t short_sig;
>> +	uint32_t index = EMPTY_SLOT;
>> +	struct __rte_hash_rcu_dq_entry rcu_dq_entry;
>> 
>> 	short_sig = get_short_sig(sig);
>> 	prim_bucket_idx = get_prim_bucket_index(h, sig); @@ -1555,10
>> +1721,9 @@ __rte_hash_del_key_with_hash(const struct rte_hash *h, const
>> void *key,
>> 
>> /* Search last bucket to see if empty to be recycled */
>> return_bkt:
>> -	if (!last_bkt) {
>> -		__hash_rw_writer_unlock(h);
>> -		return ret;
>> -	}
>> +	if (!last_bkt)
>> +		goto return_key;
>> +
>> 	while (last_bkt->next) {
>> 		prev_bkt = last_bkt;
>> 		last_bkt = last_bkt->next;
>> @@ -1571,11 +1736,11 @@ __rte_hash_del_key_with_hash(const struct
>> rte_hash *h, const void *key,
>> 	/* found empty bucket and recycle */
>> 	if (i == RTE_HASH_BUCKET_ENTRIES) {
>> 		prev_bkt->next = NULL;
>> -		uint32_t index = last_bkt - h->buckets_ext + 1;
>> +		index = last_bkt - h->buckets_ext + 1;
>> 		/* Recycle the empty bkt if
>> 		 * no_free_on_del is disabled.
>> 		 */
>> -		if (h->no_free_on_del)
>> +		if (h->no_free_on_del) {
>> 			/* Store index of an empty ext bkt to be recycled
>> 			 * on calling rte_hash_del_xxx APIs.
>> 			 * When lock free read-write concurrency is enabled,
>> @@ -1583,12 +1748,34 @@ __rte_hash_del_key_with_hash(const struct
>> rte_hash *h, const void *key,
>> 			 * immediately (as readers might be using it still).
>> 			 * Hence freeing of the ext bkt is piggy-backed to
>> 			 * freeing of the key index.
>> +			 * If using external RCU, store this index in an array.
>> 			 */
>> -			h->ext_bkt_to_free[ret] = index;
>> -		else
>> +			if (h->hash_rcu_cfg == NULL)
>> +				h->ext_bkt_to_free[ret] = index;
> 
> 
> [Yipeng]: If using embedded qsbr (not NULL), how did you recycle the ext bkt?

In DQ mode, ext_bkt_idx is enqueued to the defer queue along with the key_idx and recycled as
part of rte_rcu_qsbr_dq_reclaim(). In SYNC mode, it is recycled after rte_rcu_qsbr_synchronize() succeeds.

> 
>> +		} else
>> 			rte_ring_sp_enqueue_elem(h->free_ext_bkts,
>> &index,
>> 							sizeof(uint32_t));
>> 	}
>> +
>> +return_key:
>> +	/* Using internal RCU QSBR */
>> +	if (h->hash_rcu_cfg) {
>> +		/* Key index where key is stored, adding the first dummy
>> index */
>> +		rcu_dq_entry.key_idx = ret + 1;
>> +		rcu_dq_entry.ext_bkt_idx = index;
>> +		if (h->dq == NULL) {
>> +			/* Wait for quiescent state change if using
>> +			 * RTE_HASH_QSBR_MODE_SYNC
>> +			 */
>> +			rte_rcu_qsbr_synchronize(h->hash_rcu_cfg->v,
>> +						 RTE_QSBR_THRID_INVALID);
>> +			__hash_rcu_qsbr_free_resource((void
>> *)((uintptr_t)h),
>> +						      &rcu_dq_entry, 1);
>> +		} else if (h->dq)
>> +			/* Push into QSBR FIFO if using
>> RTE_HASH_QSBR_MODE_DQ */
>> +			if (rte_rcu_qsbr_dq_enqueue(h->dq,
>> &rcu_dq_entry) != 0)
>> +				RTE_LOG(ERR, HASH, "Failed to push QSBR
>> FIFO\n");
>> +	}
>> 	__hash_rw_writer_unlock(h);
>> 	return ret;
>> }
>> @@ -1637,8 +1824,6 @@ rte_hash_free_key_with_position(const struct
>> rte_hash *h,
>> 
>> 	RETURN_IF_TRUE(((h == NULL) || (key_idx == EMPTY_SLOT)), -EINVAL);
>> 
>> -	unsigned int lcore_id, n_slots;
>> -	struct lcore_cache *cached_free_slots;
>> 	const uint32_t total_entries = h->use_local_cache ?
>> 		h->entries + (RTE_MAX_LCORE - 1) * (LCORE_CACHE_SIZE - 1)
>> + 1
>> 							: h->entries + 1;
>> @@ -1656,28 +1841,9 @@ rte_hash_free_key_with_position(const struct
>> rte_hash *h,
>> 		}
>> 	}
>> 
>> -	if (h->use_local_cache) {
>> -		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_elem(h-
>>> free_slots,
>> -						cached_free_slots->objs,
>> -						sizeof(uint32_t),
>> -						LCORE_CACHE_SIZE, NULL);
>> -			RETURN_IF_TRUE((n_slots == 0), -EFAULT);
>> -			cached_free_slots->len -= n_slots;
>> -		}
>> -		/* Put index of new free slot in cache. */
>> -		cached_free_slots->objs[cached_free_slots->len] = key_idx;
>> -		cached_free_slots->len++;
>> -	} else {
>> -		rte_ring_sp_enqueue_elem(h->free_slots, &key_idx,
>> -						sizeof(uint32_t));
>> -	}
>> +	/* Enqueue slot to cache/ring of free slots. */
>> +	return free_slot(h, key_idx);
>> 
>> -	return 0;
>> }
>> 
>> static inline void
>> diff --git a/lib/librte_hash/rte_cuckoo_hash.h
>> b/lib/librte_hash/rte_cuckoo_hash.h
>> index 345de6bf9cfd..85be49d3bbe7 100644
>> --- a/lib/librte_hash/rte_cuckoo_hash.h
>> +++ b/lib/librte_hash/rte_cuckoo_hash.h
>> @@ -168,6 +168,11 @@ struct rte_hash {
>> 	struct lcore_cache *local_free_slots;
>> 	/**< Local cache per lcore, storing some indexes of the free slots */
>> 
>> +	/* RCU config */
>> +	struct rte_hash_rcu_config *hash_rcu_cfg;
>> +	/**< HASH RCU QSBR configuration structure */
>> +	struct rte_rcu_qsbr_dq *dq;	/**< RCU QSBR defer queue. */
>> +
>> 	/* Fields used in lookup */
>> 
>> 	uint32_t key_len __rte_cache_aligned;
>> @@ -230,4 +235,7 @@ struct queue_node {
>> 	int prev_slot;               /* Parent(slot) in search path */
>> };
>> 
>> +/** @internal Default RCU defer queue entries to reclaim in one go. */
>> +#define RTE_HASH_RCU_DQ_RECLAIM_MAX	16
>> +
>> #endif
>> diff --git a/lib/librte_hash/rte_hash.h b/lib/librte_hash/rte_hash.h index
>> bff40251bc98..3d28f177f14a 100644
>> --- a/lib/librte_hash/rte_hash.h
>> +++ b/lib/librte_hash/rte_hash.h
>> @@ -15,6 +15,7 @@
>> #include <stddef.h>
>> 
>> #include <rte_compat.h>
>> +#include <rte_rcu_qsbr.h>
>> 
>> #ifdef __cplusplus
>> extern "C" {
>> @@ -45,7 +46,8 @@ extern "C" {
>> /** Flag to disable freeing of key index 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.
>> + * is enabled. However, if internal RCU is enabled, freeing of internal
>> + * memory/index is done on delete
>>  */
>> #define RTE_HASH_EXTRA_FLAGS_NO_FREE_ON_DEL 0x10
>> 
>> @@ -67,6 +69,13 @@ typedef uint32_t (*rte_hash_function)(const void
>> *key, uint32_t key_len,
>> /** Type of function used to compare the hash key. */  typedef int
>> (*rte_hash_cmp_eq_t)(const void *key1, const void *key2, size_t key_len);
>> 
>> +/**
>> + * Type of function used to free data stored in the key.
>> + * Required when using internal RCU to allow application to free
>> +key-data once
>> + * the key is returned to the the ring of free key-slots.
>> + */
>> +typedef void (*rte_hash_free_key_data)(void *p, void *key_data);
>> +
>> /**
>>  * Parameters used when creating the hash table.
>>  */
>> @@ -81,6 +90,39 @@ struct rte_hash_parameters {
>> 	uint8_t extra_flag;		/**< Indicate if additional parameters
>> are present. */
>> };
>> 
>> +/** RCU reclamation modes */
>> +enum rte_hash_qsbr_mode {
>> +	/** Create defer queue for reclaim. */
>> +	RTE_HASH_QSBR_MODE_DQ = 0,
>> +	/** Use blocking mode reclaim. No defer queue created. */
>> +	RTE_HASH_QSBR_MODE_SYNC
>> +};
>> +
>> +/** HASH RCU QSBR configuration structure. */ struct
>> +rte_hash_rcu_config {
>> +	struct rte_rcu_qsbr *v;		/**< RCU QSBR variable. */
>> +	enum rte_hash_qsbr_mode mode;
>> +	/**< Mode of RCU QSBR. RTE_HASH_QSBR_MODE_xxx
>> +	 * '0' for default: create defer queue for reclaim.
>> +	 */
>> +	uint32_t dq_size;
>> +	/**< RCU defer queue size.
>> +	 * default: total hash table entries.
>> +	 */
>> +	uint32_t trigger_reclaim_limit;	/**< Threshold to trigger auto reclaim.
>> */
>> +	uint32_t max_reclaim_size;
>> +	/**< Max entries to reclaim in one go.
>> +	 * default: RTE_HASH_RCU_DQ_RECLAIM_MAX.
>> +	 */
>> +	void *key_data_ptr;
>> +	/**< Pointer passed to the free function. Typically, this is the
>> +	 * pointer to the data structure to which the resource to free
>> +	 * (key-data) belongs. This can be NULL.
>> +	 */
>> +	rte_hash_free_key_data free_key_data_func;
>> +	/**< Function to call to free the resource (key-data). */ };
>> +
>> /** @internal A hash table structure. */  struct rte_hash;
>> 
>> @@ -287,7 +329,8 @@ rte_hash_add_key_with_hash(const struct rte_hash
>> *h, const void *key, hash_sig_t
>>  * Thread safety can be enabled by setting flag during
>>  * table creation.
>>  * If RTE_HASH_EXTRA_FLAGS_NO_FREE_ON_DEL or
>> - * RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled,
>> + * RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled and
>> + * internal RCU is NOT enabled,
>>  * the key index returned by rte_hash_add_key_xxx APIs will not be
>>  * freed by this API. rte_hash_free_key_with_position API must be called
>>  * additionally to free the index associated with the key.
>> @@ -316,7 +359,8 @@ rte_hash_del_key(const struct rte_hash *h, const
>> void *key);
>>  * Thread safety can be enabled by setting flag during
>>  * table creation.
>>  * If RTE_HASH_EXTRA_FLAGS_NO_FREE_ON_DEL or
>> - * RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled,
>> + * RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled and
>> + * internal RCU is NOT enabled,
>>  * the key index returned by rte_hash_add_key_xxx APIs will not be
>>  * freed by this API. rte_hash_free_key_with_position API must be called
>>  * additionally to free the index associated with the key.
>> @@ -370,7 +414,8 @@ rte_hash_get_key_with_position(const struct
>> rte_hash *h, const int32_t position,
>>  * only be called from one thread by default. Thread safety
>>  * can be enabled by setting flag during table creation.
>>  * If RTE_HASH_EXTRA_FLAGS_NO_FREE_ON_DEL or
>> - * RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled,
>> + * RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF is enabled and
>> + * internal RCU is NOT enabled,
>>  * the key index returned by rte_hash_del_key_xxx APIs must be freed
>>  * using this API. This API should be called after all the readers
>>  * have stopped referencing the entry corresponding to this key.
>> @@ -625,6 +670,30 @@ rte_hash_lookup_bulk(const struct rte_hash *h,
>> const void **keys,
>>  */
>> int32_t
>> rte_hash_iterate(const struct rte_hash *h, const void **key, void **data,
>> uint32_t *next);
>> +
>> +/**
>> + * @warning
>> + * @b EXPERIMENTAL: this API may change without prior notice
>> + *
>> + * Associate RCU QSBR variable with an Hash object.
> [Yipeng]: a Hash object

Copy.

> 
>> + * This API should be called to enable the integrated RCU QSBR support
>> +and
>> + * should be called immediately after creating the Hash object.
>> + *
>> + * @param h
>> + *   the hash object to add RCU QSBR
>> + * @param cfg
>> + *   RCU QSBR configuration
>> + * @return
>> + *   On success - 0
>> + *   On error - 1 with error code set in rte_errno.
>> + *   Possible rte_errno codes are:
>> + *   - EINVAL - invalid pointer
>> + *   - EEXIST - already added QSBR
>> + *   - ENOMEM - memory allocation failure
>> + */
>> +__rte_experimental
>> +int rte_hash_rcu_qsbr_add(struct rte_hash *h,
>> +				struct rte_hash_rcu_config *cfg);
>> #ifdef __cplusplus
>> }
>> #endif
>> diff --git a/lib/librte_hash/version.map b/lib/librte_hash/version.map index
>> c0db81014ff9..c6d73080f478 100644
>> --- a/lib/librte_hash/version.map
>> +++ b/lib/librte_hash/version.map
>> @@ -36,5 +36,5 @@ EXPERIMENTAL {
>> 	rte_hash_lookup_with_hash_bulk;
>> 	rte_hash_lookup_with_hash_bulk_data;
>> 	rte_hash_max_key_id;
>> -
>> +	rte_hash_rcu_qsbr_add;
>> };
>> --
>> 2.17.1
> [Yipeng]:
> 
> Hi, Dharmik, thanks for the work! It generally looks good.
> Just some minor issues to address and one question for the ext table inlined.

Thank you for the comments, Yipeng! I will update the patches.

> 
> Thanks!


  reply	other threads:[~2020-10-21  4:53 UTC|newest]

Thread overview: 49+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-09-01  6:58 [dpdk-dev] [RFC 0/3] RCU integration with hash library Dharmik Thakkar
2019-09-01  6:58 ` [dpdk-dev] [RFC 1/3] net/ixgbe: avoid multpile definitions of 'bool' Dharmik Thakkar
2019-09-05 15:26   ` Stephen Hemminger
2019-09-05 20:10     ` Dharmik Thakkar
2019-09-05 20:23       ` Stephen Hemminger
2019-09-01  6:58 ` [dpdk-dev] [RFC 2/3] lib/hash: integrate RCU QSBR Dharmik Thakkar
2019-09-01  6:58 ` [dpdk-dev] [RFC 3/3] test/hash: add tests for integrated " Dharmik Thakkar
2019-09-05 21:31 ` [dpdk-dev] [RFC 0/3] RCU integration with hash library Wang, Yipeng1
2020-08-19  4:05 ` [dpdk-dev] [RFC v2] lib/hash: integrate RCU QSBR Dharmik Thakkar
2020-08-25 19:59   ` Dharmik Thakkar
2020-08-31 20:47   ` Wang, Yipeng1
2020-09-01 22:01     ` Dharmik Thakkar
2020-10-16 13:53       ` David Marchand
2020-10-16 15:04         ` Dharmik Thakkar
2020-10-16 17:38   ` [dpdk-dev] [PATCH 0/3] hash: " Dharmik Thakkar
2020-10-16 17:38     ` [dpdk-dev] [PATCH v3 1/3] lib/hash: " Dharmik Thakkar
2020-10-19  9:43       ` Kinsella, Ray
2020-10-16 17:38     ` [dpdk-dev] [PATCH v3 2/3] test/hash: replace rte atomic with C11 atomic APIs Dharmik Thakkar
2020-10-16 17:38     ` [dpdk-dev] [PATCH v3 3/3] test/hash: add tests for integrated RCU QSBR Dharmik Thakkar
2020-10-19 14:48     ` [dpdk-dev] [PATCH 0/3] hash: integrate " David Marchand
2020-10-19 16:35     ` [dpdk-dev] [PATCH v4 " Dharmik Thakkar
2020-10-19 16:35       ` [dpdk-dev] [PATCH v4 1/3] lib/hash: " Dharmik Thakkar
2020-10-19 16:35       ` [dpdk-dev] [PATCH v4 2/3] test/hash: replace rte atomic with C11 atomic APIs Dharmik Thakkar
2020-10-19 16:35       ` [dpdk-dev] [PATCH v4 3/3] test/hash: add tests for integrated RCU QSBR Dharmik Thakkar
2020-10-19 21:05       ` [dpdk-dev] [PATCH v4 0/3] hash: integrate " David Marchand
2020-10-20  4:08         ` Dharmik Thakkar
2020-10-20 16:12       ` [dpdk-dev] [PATCH v5 0/4] " Dharmik Thakkar
2020-10-20 16:12         ` [dpdk-dev] [PATCH v5 1/4] rcu: build on Windows Dharmik Thakkar
2020-10-20 16:12         ` [dpdk-dev] [PATCH v5 2/4] lib/hash: integrate RCU QSBR Dharmik Thakkar
2020-10-21  2:42           ` Wang, Yipeng1
2020-10-21  4:52             ` Dharmik Thakkar [this message]
2020-10-20 16:13         ` [dpdk-dev] [PATCH v5 3/4] test/hash: replace rte atomic with C11 atomic APIs Dharmik Thakkar
2020-10-21  2:52           ` Wang, Yipeng1
2020-10-20 16:13         ` [dpdk-dev] [PATCH v5 4/4] test/hash: add tests for integrated RCU QSBR Dharmik Thakkar
2020-10-21  3:54           ` Wang, Yipeng1
2020-10-21  4:55             ` Dharmik Thakkar
2020-10-21 22:34             ` Honnappa Nagarahalli
2020-10-21 22:50         ` [dpdk-dev] [PATCH v6 0/4] hash: integrate " Dharmik Thakkar
2020-10-21 22:50           ` [dpdk-dev] [PATCH v6 1/4] rcu: build on Windows Dharmik Thakkar
2020-10-22  9:05             ` David Marchand
2020-10-22 18:35             ` Kadam, Pallavi
2020-10-21 22:50           ` [dpdk-dev] [PATCH v6 2/4] lib/hash: integrate RCU QSBR Dharmik Thakkar
2020-10-23 21:46             ` Dharmik Thakkar
2020-10-23 22:08             ` Wang, Yipeng1
2020-10-21 22:50           ` [dpdk-dev] [PATCH v6 3/4] test/hash: replace rte atomic with C11 atomic APIs Dharmik Thakkar
2020-10-21 22:50           ` [dpdk-dev] [PATCH v6 4/4] test/hash: add tests for integrated RCU QSBR Dharmik Thakkar
2020-10-23 22:11             ` Wang, Yipeng1
2020-10-24  9:09           ` [dpdk-dev] [PATCH v6 0/4] hash: integrate " David Marchand
2020-10-26 13:56             ` Dharmik Thakkar

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=72DB22D2-34AF-41BD-ACAC-7C2DF3121BCC@arm.com \
    --to=dharmik.thakkar@arm.com \
    --cc=bruce.richardson@intel.com \
    --cc=dev@dpdk.org \
    --cc=mdr@ashroe.eu \
    --cc=nd@arm.com \
    --cc=nhorman@tuxdriver.com \
    --cc=sameh.gobriel@intel.com \
    --cc=yipeng1.wang@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).