From: Andrew Rybchenko <arybchenko@solarflare.com>
To: Olivier MATZ <olivier.matz@6wind.com>
Cc: <dev@dpdk.org>, "Artem V. Andreev" <Artem.Andreev@oktetlabs.ru>
Subject: Re: [dpdk-dev] [RFC PATCH 3/6] mempool/bucket: implement bucket mempool manager
Date: Wed, 17 Jan 2018 18:06:20 +0300 [thread overview]
Message-ID: <98fe3f77-ae62-6a68-2e31-212033d41383@solarflare.com> (raw)
In-Reply-To: <20171214133821.gn4ak6w77gqlyysn@platinum>
On 12/14/2017 04:38 PM, Olivier MATZ wrote:
> On Fri, Nov 24, 2017 at 04:06:28PM +0000, Andrew Rybchenko wrote:
>> From: "Artem V. Andreev" <Artem.Andreev@oktetlabs.ru>
>>
>> The manager provides a way to allocate physically and virtually
>> contiguous set of objects.
>>
>> Note: due to the way objects are organized in the bucket manager,
>> the get_avail_count may return less objects than were enqueued.
>> That breaks the expectation of mempool and mempool_perf tests.
> To me, this can be problematic. The driver should respect the
> API, or it will trigger hard-to-debug issues in applications. Can't
> this be fixed in some way or another?
As I understand there is no requirements on how fast get_count
works. If so, it is doable and we'll fix it in RFCv2.
> [...]
>
>> --- a/config/common_base
>> +++ b/config/common_base
>> @@ -608,6 +608,8 @@ CONFIG_RTE_LIBRTE_MEMPOOL_DEBUG=n
>> #
>> # Compile Mempool drivers
>> #
>> +CONFIG_RTE_DRIVER_MEMPOOL_BUCKET=y
>> +CONFIG_RTE_DRIVER_MEMPOOL_BUCKET_SIZE_KB=32
>> CONFIG_RTE_DRIVER_MEMPOOL_RING=y
>> CONFIG_RTE_DRIVER_MEMPOOL_STACK=y
>>
> Why 32KB?
> Why not more, or less?
> Can it be a runtime parameter?
> I guess it won't work with too large objects.
We have no good understanding of how driver-specific parameters
should be passed on mempool creation. We've simply kept it for
future since it looks like separate task.
If you have ideas, please, share - we'll be thankful.
> [...]
>
>> +struct bucket_data {
>> + unsigned int header_size;
>> + unsigned int chunk_size;
>> + unsigned int bucket_size;
>> + uintptr_t bucket_page_mask;
>> + struct rte_ring *shared_bucket_ring;
>> + struct bucket_stack *buckets[RTE_MAX_LCORE];
>> + /*
>> + * Multi-producer single-consumer ring to hold objects that are
>> + * returned to the mempool at a different lcore than initially
>> + * dequeued
>> + */
>> + struct rte_ring *adoption_buffer_rings[RTE_MAX_LCORE];
>> + struct rte_ring *shared_orphan_ring;
>> + struct rte_mempool *pool;
>> +
>> +};
> I'm seeing per-core structures. Will it work on non-dataplane cores?
> For instance, if a control thread wants to allocate a mbuf?
May be I don't understand something. Does the control thread has
valid rte_lcore_id()?
> If possible, these fields should be more documented (or just renamed).
> For instance, I suggest chunk_size could be called obj_per_bucket, which
> better described the content of the field.
Thanks, we'll do.
> [...]
>
>> +static int
>> +bucket_enqueue_single(struct bucket_data *data, void *obj)
>> +{
>> + int rc = 0;
>> + uintptr_t addr = (uintptr_t)obj;
>> + struct bucket_header *hdr;
>> + unsigned int lcore_id = rte_lcore_id();
>> +
>> + addr &= data->bucket_page_mask;
>> + hdr = (struct bucket_header *)addr;
>> +
>> + if (likely(hdr->lcore_id == lcore_id)) {
>> + if (hdr->fill_cnt < data->bucket_size - 1) {
>> + hdr->fill_cnt++;
>> + } else {
>> + hdr->fill_cnt = 0;
>> + /* Stack is big enough to put all buckets */
>> + bucket_stack_push(data->buckets[lcore_id], hdr);
>> + }
>> + } else if (hdr->lcore_id != LCORE_ID_ANY) {
>> + struct rte_ring *adopt_ring =
>> + data->adoption_buffer_rings[hdr->lcore_id];
>> +
>> + rc = rte_ring_enqueue(adopt_ring, obj);
>> + /* Ring is big enough to put all objects */
>> + RTE_ASSERT(rc == 0);
>> + } else if (hdr->fill_cnt < data->bucket_size - 1) {
>> + hdr->fill_cnt++;
>> + } else {
>> + hdr->fill_cnt = 0;
>> + rc = rte_ring_enqueue(data->shared_bucket_ring, hdr);
>> + /* Ring is big enough to put all buckets */
>> + RTE_ASSERT(rc == 0);
>> + }
>> +
>> + return rc;
>> +}
> [...]
>
>> +static int
>> +bucket_dequeue_buckets(struct bucket_data *data, void **obj_table,
>> + unsigned int n_buckets)
>> +{
>> + struct bucket_stack *cur_stack = data->buckets[rte_lcore_id()];
>> + unsigned int n_buckets_from_stack = RTE_MIN(n_buckets, cur_stack->top);
>> + void **obj_table_base = obj_table;
>> +
>> + n_buckets -= n_buckets_from_stack;
>> + while (n_buckets_from_stack-- > 0) {
>> + void *obj = bucket_stack_pop_unsafe(cur_stack);
>> +
>> + obj_table = bucket_fill_obj_table(data, &obj, obj_table,
>> + data->bucket_size);
>> + }
>> + while (n_buckets-- > 0) {
>> + struct bucket_header *hdr;
>> +
>> + if (unlikely(rte_ring_dequeue(data->shared_bucket_ring,
>> + (void **)&hdr) != 0)) {
>> + /* Return the already-dequeued buffers
>> + * back to the mempool
>> + */
>> + bucket_enqueue(data->pool, obj_table_base,
>> + obj_table - obj_table_base);
>> + rte_errno = ENOBUFS;
>> + return -rte_errno;
>> + }
>> + hdr->lcore_id = rte_lcore_id();
>> + obj_table = bucket_fill_obj_table(data, (void **)&hdr,
>> + obj_table, data->bucket_size);
>> + }
>> +
>> + return 0;
>> +}
> [...]
>
>> +static int
>> +bucket_dequeue(struct rte_mempool *mp, void **obj_table, unsigned int n)
>> +{
>> + struct bucket_data *data = mp->pool_data;
>> + unsigned int n_buckets = n / data->bucket_size;
>> + unsigned int n_orphans = n - n_buckets * data->bucket_size;
>> + int rc = 0;
>> +
>> + bucket_adopt_orphans(data);
>> +
>> + if (unlikely(n_orphans > 0)) {
>> + rc = bucket_dequeue_orphans(data, obj_table +
>> + (n_buckets * data->bucket_size),
>> + n_orphans);
>> + if (rc != 0)
>> + return rc;
>> + }
>> +
>> + if (likely(n_buckets > 0)) {
>> + rc = bucket_dequeue_buckets(data, obj_table, n_buckets);
>> + if (unlikely(rc != 0) && n_orphans > 0) {
>> + rte_ring_enqueue_bulk(data->shared_orphan_ring,
>> + obj_table + (n_buckets *
>> + data->bucket_size),
>> + n_orphans, NULL);
>> + }
>> + }
>> +
>> + return rc;
>> +}
> If my understanding is correct, at initialization, all full buckets will
> go to the data->shared_bucket_ring ring, with lcore_id == ANY (this is
> done in register_mem).
>
> (note: I feel 'data' is not an ideal name for bucket_data)
Yes, agree. We'll rename it. It is really too generic.
> If the core 0 allocates all the mbufs, and then frees them all, they
> will be stored in the per-core stack, with hdr->lcoreid == 0. Is it
> right?
Right.
> If yes, can core 1 allocate a mbuf after that?
We'll add threshold for per-core stack. If it is exceeded, buckets will be
flushed into shared ring.
>> +static unsigned int
>> +bucket_get_count(const struct rte_mempool *mp)
>> +{
>> + const struct bucket_data *data = mp->pool_data;
>> + const struct bucket_stack *local_bucket_stack =
>> + data->buckets[rte_lcore_id()];
>> +
>> + return data->bucket_size * local_bucket_stack->top +
>> + data->bucket_size * rte_ring_count(data->shared_bucket_ring) +
>> + rte_ring_count(data->shared_orphan_ring);
>> +}
> It looks that get_count only rely on the current core stack usage
> and ignore the other core stacks.
We'll fix it to provide more accurate return value which is required
to pass self-test and make it usable for debugging.
> [...]
>
>> +static int
>> +bucket_register_memory_area(__rte_unused const struct rte_mempool *mp,
>> + char *vaddr, __rte_unused phys_addr_t paddr,
>> + size_t len)
>> +{
>> + /* mp->pool_data may be still uninitialized at this point */
>> + unsigned int chunk_size = mp->header_size + mp->elt_size +
>> + mp->trailer_size;
>> + unsigned int bucket_mem_size =
>> + (BUCKET_MEM_SIZE / chunk_size) * chunk_size;
>> + unsigned int bucket_page_sz = rte_align32pow2(bucket_mem_size);
>> + uintptr_t align;
>> + char *iter;
>> +
>> + align = RTE_PTR_ALIGN_CEIL(vaddr, bucket_page_sz) - vaddr;
>> +
>> + for (iter = vaddr + align; iter < vaddr + len; iter += bucket_page_sz) {
>> + /* librte_mempool uses the header part for its own bookkeeping,
>> + * but the librte_mempool's object header is adjacent to the
>> + * data; it is small enough and the header is guaranteed to be
>> + * at least CACHE_LINE_SIZE (i.e. 64) bytes, so we do have
>> + * plenty of space at the start of the header. So the layout
>> + * looks like this:
>> + * [bucket_header] ... unused ... [rte_mempool_objhdr] [data...]
>> + */
> This is not always true.
> If a use creates a mempool with the NO_CACHE_ALIGN, the header will be
> small, without padding.
Thanks. I think it can be handled when bucket mempool implements own
callback to populate objects.
next prev parent reply other threads:[~2018-01-17 15:06 UTC|newest]
Thread overview: 197+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-11-24 16:06 [dpdk-dev] [RFC PATCH 0/6] mempool: add bucket mempool driver Andrew Rybchenko
2017-11-24 16:06 ` [dpdk-dev] [RFC PATCH 1/6] mempool: implement abstract mempool info API Andrew Rybchenko
2017-12-14 13:36 ` Olivier MATZ
2018-01-17 15:03 ` Andrew Rybchenko
2017-11-24 16:06 ` [dpdk-dev] [RFC PATCH 2/6] mempool: implement clustered object allocation Andrew Rybchenko
2017-12-14 13:37 ` Olivier MATZ
2018-01-17 15:03 ` Andrew Rybchenko
2018-01-17 15:55 ` santosh
2018-01-17 16:37 ` Andrew Rybchenko
2017-11-24 16:06 ` [dpdk-dev] [RFC PATCH 3/6] mempool/bucket: implement bucket mempool manager Andrew Rybchenko
2017-12-14 13:38 ` Olivier MATZ
2018-01-17 15:06 ` Andrew Rybchenko [this message]
2017-11-24 16:06 ` [dpdk-dev] [RFC PATCH 4/6] mempool: add a function to flush default cache Andrew Rybchenko
2017-12-14 13:38 ` Olivier MATZ
2018-01-17 15:07 ` Andrew Rybchenko
2017-11-24 16:06 ` [dpdk-dev] [RFC PATCH 5/6] mempool: support block dequeue operation Andrew Rybchenko
2017-12-14 13:38 ` Olivier MATZ
2017-11-24 16:06 ` [dpdk-dev] [RFC PATCH 6/6] mempool/bucket: implement " Andrew Rybchenko
2017-12-14 13:36 ` [dpdk-dev] [RFC PATCH 0/6] mempool: add bucket mempool driver Olivier MATZ
2018-01-17 15:03 ` Andrew Rybchenko
2018-01-23 13:15 ` [dpdk-dev] [RFC v2 00/17] " Andrew Rybchenko
2018-01-23 13:15 ` [dpdk-dev] [RFC v2 01/17] mempool: fix phys contig check if populate default skipped Andrew Rybchenko
2018-01-31 16:45 ` Olivier Matz
2018-02-01 5:05 ` santosh
2018-02-01 6:54 ` Andrew Rybchenko
2018-02-01 9:09 ` santosh
2018-02-01 9:18 ` Andrew Rybchenko
2018-02-01 9:30 ` santosh
2018-02-01 10:00 ` Andrew Rybchenko
2018-02-01 10:14 ` Olivier Matz
2018-02-01 10:33 ` santosh
2018-02-01 14:02 ` Andrew Rybchenko
2018-02-01 10:17 ` santosh
2018-02-01 14:02 ` [dpdk-dev] [PATCH] " Andrew Rybchenko
2018-02-05 23:53 ` [dpdk-dev] [dpdk-stable] " Thomas Monjalon
2018-01-23 13:15 ` [dpdk-dev] [RFC v2 02/17] mempool: add op to calculate memory size to be allocated Andrew Rybchenko
2018-01-31 16:45 ` Olivier Matz
2018-02-01 7:15 ` Andrew Rybchenko
2018-01-23 13:15 ` [dpdk-dev] [RFC v2 03/17] mempool/octeontx: add callback to calculate memory size Andrew Rybchenko
[not found] ` <BN3PR07MB2513732462EB5FE5E1B05713E3FA0@BN3PR07MB2513.namprd07.prod.outlook.com>
2018-02-01 10:01 ` santosh
2018-02-01 13:40 ` santosh
2018-03-10 15:49 ` Andrew Rybchenko
2018-03-11 6:31 ` santosh
2018-01-23 13:15 ` [dpdk-dev] [RFC v2 04/17] mempool: add op to populate objects using provided memory Andrew Rybchenko
2018-01-31 16:45 ` Olivier Matz
2018-02-01 8:51 ` Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 05/17] mempool/octeontx: implement callback to populate objects Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 06/17] mempool: remove callback to get capabilities Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 07/17] mempool: deprecate xmem functions Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 08/17] mempool/octeontx: prepare to remove register memory area op Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 09/17] mempool/dpaa: convert to use populate driver op Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 10/17] mempool: remove callback to register memory area Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 11/17] mempool: ensure the mempool is initialized before populating Andrew Rybchenko
2018-01-31 16:45 ` Olivier Matz
2018-02-01 8:53 ` Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 12/17] mempool/bucket: implement bucket mempool manager Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 13/17] mempool: support flushing the default cache of the mempool Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 14/17] mempool: implement abstract mempool info API Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 15/17] mempool: support block dequeue operation Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 16/17] mempool/bucket: implement " Andrew Rybchenko
2018-01-23 13:16 ` [dpdk-dev] [RFC v2 17/17] mempool/bucket: do not allow one lcore to grab all buckets Andrew Rybchenko
2018-01-31 16:44 ` [dpdk-dev] [RFC v2 00/17] mempool: add bucket mempool driver Olivier Matz
2018-03-10 15:39 ` [dpdk-dev] [PATCH v1 0/9] mempool: prepare to add bucket driver Andrew Rybchenko
2018-03-10 15:39 ` [dpdk-dev] [PATCH v1 1/9] mempool: add op to calculate memory size to be allocated Andrew Rybchenko
2018-03-11 12:51 ` santosh
2018-03-12 6:53 ` Andrew Rybchenko
2018-03-19 17:03 ` Olivier Matz
2018-03-20 10:29 ` Andrew Rybchenko
2018-03-20 14:41 ` Bruce Richardson
2018-03-10 15:39 ` [dpdk-dev] [PATCH v1 2/9] mempool: add op to populate objects using provided memory Andrew Rybchenko
2018-03-19 17:04 ` Olivier Matz
2018-03-21 7:05 ` Andrew Rybchenko
2018-03-10 15:39 ` [dpdk-dev] [PATCH v1 3/9] mempool: remove callback to get capabilities Andrew Rybchenko
2018-03-14 14:40 ` Burakov, Anatoly
2018-03-14 16:12 ` Andrew Rybchenko
2018-03-14 16:53 ` Burakov, Anatoly
2018-03-14 17:24 ` Andrew Rybchenko
2018-03-15 9:48 ` Burakov, Anatoly
2018-03-15 11:49 ` Andrew Rybchenko
2018-03-15 12:00 ` Burakov, Anatoly
2018-03-15 12:44 ` Andrew Rybchenko
2018-03-19 17:05 ` Olivier Matz
2018-03-19 17:06 ` Olivier Matz
2018-03-10 15:39 ` [dpdk-dev] [PATCH v1 4/9] mempool: deprecate xmem functions Andrew Rybchenko
2018-03-10 15:39 ` [dpdk-dev] [PATCH v1 5/9] mempool/octeontx: prepare to remove register memory area op Andrew Rybchenko
2018-03-10 15:39 ` [dpdk-dev] [PATCH v1 6/9] mempool/dpaa: " Andrew Rybchenko
2018-03-10 15:39 ` [dpdk-dev] [PATCH v1 7/9] mempool: remove callback to register memory area Andrew Rybchenko
2018-03-10 15:39 ` [dpdk-dev] [PATCH v1 8/9] mempool: ensure the mempool is initialized before populating Andrew Rybchenko
2018-03-19 17:06 ` Olivier Matz
2018-03-20 13:32 ` Andrew Rybchenko
2018-03-20 16:57 ` Olivier Matz
2018-03-10 15:39 ` [dpdk-dev] [PATCH v1 9/9] mempool: support flushing the default cache of the mempool Andrew Rybchenko
2018-03-14 15:49 ` [dpdk-dev] [PATCH v1 0/9] mempool: prepare to add bucket driver santosh
2018-03-14 15:57 ` Andrew Rybchenko
2018-03-19 17:03 ` Olivier Matz
2018-03-20 10:09 ` Andrew Rybchenko
2018-03-20 11:04 ` Thomas Monjalon
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 00/11] " Andrew Rybchenko
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 01/11] mempool: fix memhdr leak when no objects are populated Andrew Rybchenko
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 02/11] mempool: rename flag to control IOVA-contiguous objects Andrew Rybchenko
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 03/11] mempool: ensure the mempool is initialized before populating Andrew Rybchenko
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 04/11] mempool: add op to calculate memory size to be allocated Andrew Rybchenko
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 05/11] mempool: add op to populate objects using provided memory Andrew Rybchenko
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 06/11] mempool: remove callback to get capabilities Andrew Rybchenko
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 07/11] mempool: deprecate xmem functions Andrew Rybchenko
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 08/11] mempool/octeontx: prepare to remove register memory area op Andrew Rybchenko
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 09/11] mempool/dpaa: " Andrew Rybchenko
2018-03-26 7:13 ` Andrew Rybchenko
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 10/11] mempool: remove callback to register memory area Andrew Rybchenko
2018-03-25 16:20 ` [dpdk-dev] [PATCH v2 11/11] mempool: support flushing the default cache of the mempool Andrew Rybchenko
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 00/11] mempool: prepare to add bucket driver Andrew Rybchenko
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 01/11] mempool: fix memhdr leak when no objects are populated Andrew Rybchenko
2018-04-06 15:50 ` Olivier Matz
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 02/11] mempool: rename flag to control IOVA-contiguous objects Andrew Rybchenko
2018-04-06 15:50 ` Olivier Matz
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 03/11] mempool: ensure the mempool is initialized before populating Andrew Rybchenko
2018-04-04 15:06 ` santosh
2018-04-06 15:50 ` Olivier Matz
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 04/11] mempool: add op to calculate memory size to be allocated Andrew Rybchenko
2018-04-04 15:08 ` santosh
2018-04-06 15:51 ` Olivier Matz
2018-04-12 15:22 ` Burakov, Anatoly
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 05/11] mempool: add op to populate objects using provided memory Andrew Rybchenko
2018-04-04 15:09 ` santosh
2018-04-06 15:51 ` Olivier Matz
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 06/11] mempool: remove callback to get capabilities Andrew Rybchenko
2018-04-04 15:10 ` santosh
2018-04-06 15:51 ` Olivier Matz
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 07/11] mempool: deprecate xmem functions Andrew Rybchenko
2018-04-06 15:52 ` Olivier Matz
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 08/11] mempool/octeontx: prepare to remove register memory area op Andrew Rybchenko
2018-04-04 15:12 ` santosh
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 09/11] mempool/dpaa: " Andrew Rybchenko
2018-04-05 8:25 ` Hemant Agrawal
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 10/11] mempool: remove callback to register memory area Andrew Rybchenko
2018-04-04 15:13 ` santosh
2018-04-06 15:52 ` Olivier Matz
2018-03-26 16:09 ` [dpdk-dev] [PATCH v3 11/11] mempool: support flushing the default cache of the mempool Andrew Rybchenko
2018-04-06 15:53 ` Olivier Matz
2018-03-26 16:12 ` [dpdk-dev] [PATCH v1 0/6] mempool: add bucket driver Andrew Rybchenko
2018-03-26 16:12 ` [dpdk-dev] [PATCH v1 1/6] mempool/bucket: implement bucket mempool manager Andrew Rybchenko
2018-03-26 16:12 ` [dpdk-dev] [PATCH v1 2/6] mempool: implement abstract mempool info API Andrew Rybchenko
2018-04-19 16:42 ` Olivier Matz
2018-04-25 9:57 ` Andrew Rybchenko
2018-04-25 10:26 ` Olivier Matz
2018-03-26 16:12 ` [dpdk-dev] [PATCH v1 3/6] mempool: support block dequeue operation Andrew Rybchenko
2018-04-19 16:41 ` Olivier Matz
2018-04-25 9:49 ` Andrew Rybchenko
2018-03-26 16:12 ` [dpdk-dev] [PATCH v1 4/6] mempool/bucket: implement " Andrew Rybchenko
2018-03-26 16:12 ` [dpdk-dev] [PATCH v1 5/6] mempool/bucket: do not allow one lcore to grab all buckets Andrew Rybchenko
2018-03-26 16:12 ` [dpdk-dev] [PATCH v1 6/6] doc: advertise bucket mempool driver Andrew Rybchenko
2018-04-19 16:43 ` Olivier Matz
2018-04-19 16:41 ` [dpdk-dev] [PATCH v1 0/6] mempool: add bucket driver Olivier Matz
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 00/11] mempool: prepare to " Andrew Rybchenko
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 01/11] mempool: fix memhdr leak when no objects are populated Andrew Rybchenko
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 02/11] mempool: rename flag to control IOVA-contiguous objects Andrew Rybchenko
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 03/11] mempool: ensure the mempool is initialized before populating Andrew Rybchenko
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 04/11] mempool: add op to calculate memory size to be allocated Andrew Rybchenko
2018-04-16 15:33 ` Olivier Matz
2018-04-16 15:41 ` Andrew Rybchenko
2018-04-17 10:23 ` Burakov, Anatoly
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 05/11] mempool: add op to populate objects using provided memory Andrew Rybchenko
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 06/11] mempool: remove callback to get capabilities Andrew Rybchenko
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 07/11] mempool: deprecate xmem functions Andrew Rybchenko
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 08/11] mempool/octeontx: prepare to remove register memory area op Andrew Rybchenko
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 09/11] mempool/dpaa: " Andrew Rybchenko
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 10/11] mempool: remove callback to register memory area Andrew Rybchenko
2018-04-16 13:24 ` [dpdk-dev] [PATCH v4 11/11] mempool: support flushing the default cache of the mempool Andrew Rybchenko
2018-04-24 0:20 ` [dpdk-dev] [PATCH v4 00/11] mempool: prepare to add bucket driver Thomas Monjalon
2018-04-16 13:33 ` [dpdk-dev] [PATCH v2 0/6] mempool: " Andrew Rybchenko
2018-04-16 13:33 ` [dpdk-dev] [PATCH v2 1/6] mempool/bucket: implement bucket mempool manager Andrew Rybchenko
2018-04-16 13:33 ` [dpdk-dev] [PATCH v2 2/6] mempool: implement abstract mempool info API Andrew Rybchenko
2018-04-25 8:44 ` Olivier Matz
2018-04-16 13:33 ` [dpdk-dev] [PATCH v2 3/6] mempool: support block dequeue operation Andrew Rybchenko
2018-04-25 8:45 ` Olivier Matz
2018-04-16 13:33 ` [dpdk-dev] [PATCH v2 4/6] mempool/bucket: implement " Andrew Rybchenko
2018-04-16 13:33 ` [dpdk-dev] [PATCH v2 5/6] mempool/bucket: do not allow one lcore to grab all buckets Andrew Rybchenko
2018-04-16 13:33 ` [dpdk-dev] [PATCH v2 6/6] doc: advertise bucket mempool driver Andrew Rybchenko
2018-04-24 23:00 ` [dpdk-dev] [PATCH v2 0/6] mempool: add bucket driver Thomas Monjalon
2018-04-25 8:43 ` Olivier Matz
2018-04-25 16:32 ` [dpdk-dev] [PATCH v3 " Andrew Rybchenko
2018-04-25 16:32 ` [dpdk-dev] [PATCH v3 1/6] mempool/bucket: implement bucket mempool manager Andrew Rybchenko
2018-04-25 16:32 ` [dpdk-dev] [PATCH v3 2/6] mempool: implement abstract mempool info API Andrew Rybchenko
2018-04-25 16:32 ` [dpdk-dev] [PATCH v3 3/6] mempool: support block dequeue operation Andrew Rybchenko
2018-04-25 16:32 ` [dpdk-dev] [PATCH v3 4/6] mempool/bucket: implement " Andrew Rybchenko
2018-04-25 16:32 ` [dpdk-dev] [PATCH v3 5/6] mempool/bucket: do not allow one lcore to grab all buckets Andrew Rybchenko
2018-04-25 16:32 ` [dpdk-dev] [PATCH v3 6/6] doc: advertise bucket mempool driver Andrew Rybchenko
2018-04-25 21:56 ` Thomas Monjalon
2018-04-25 22:04 ` Thomas Monjalon
2018-04-26 9:50 ` Andrew Rybchenko
2018-04-26 10:59 ` [dpdk-dev] [PATCH v4 0/5] mempool: add bucket driver Andrew Rybchenko
2018-04-26 10:59 ` [dpdk-dev] [PATCH v4 1/5] mempool/bucket: implement bucket mempool manager Andrew Rybchenko
2018-04-26 10:59 ` [dpdk-dev] [PATCH v4 2/5] mempool: implement abstract mempool info API Andrew Rybchenko
2018-04-26 10:59 ` [dpdk-dev] [PATCH v4 3/5] mempool: support block dequeue operation Andrew Rybchenko
2018-04-26 10:59 ` [dpdk-dev] [PATCH v4 4/5] mempool/bucket: implement " Andrew Rybchenko
2018-04-26 10:59 ` [dpdk-dev] [PATCH v4 5/5] mempool/bucket: do not allow one lcore to grab all buckets Andrew Rybchenko
2018-04-26 21:35 ` [dpdk-dev] [PATCH v4 0/5] mempool: add bucket driver Thomas Monjalon
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=98fe3f77-ae62-6a68-2e31-212033d41383@solarflare.com \
--to=arybchenko@solarflare.com \
--cc=Artem.Andreev@oktetlabs.ru \
--cc=dev@dpdk.org \
--cc=olivier.matz@6wind.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).