From: Maxime Coquelin <maxime.coquelin@redhat.com>
To: Patrick Fu <patrick.fu@intel.com>, dev@dpdk.org, chenbo.xia@intel.com
Cc: zhihong.wang@intel.com, cheng1.jiang@intel.com
Subject: Re: [dpdk-dev] [PATCH v1 2/4] vhost: dynamically alloc async memory
Date: Wed, 23 Sep 2020 11:15:21 +0200 [thread overview]
Message-ID: <1522ca39-d9d7-72ff-8bb6-a5a29d12314f@redhat.com> (raw)
In-Reply-To: <20200911015316.1903181-3-patrick.fu@intel.com>
s/alloc/allocate/
On 9/11/20 3:53 AM, Patrick Fu wrote:
> alloc async internal memory buffer by rte_malloc(), replacing array
Allocate async internal memory buffer with rte_malloc()
> declaration inside vq structure. Dynamic allocation can help to save
> memory footprint when async path is not registered.
>
> Signed-off-by: Patrick Fu <patrick.fu@intel.com>
> ---
> lib/librte_vhost/vhost.c | 49 ++++++++++++++++++++++++----------------
> lib/librte_vhost/vhost.h | 4 ++--
> 2 files changed, 31 insertions(+), 22 deletions(-)
>
> diff --git a/lib/librte_vhost/vhost.c b/lib/librte_vhost/vhost.c
> index eca507836..ba374da67 100644
> --- a/lib/librte_vhost/vhost.c
> +++ b/lib/librte_vhost/vhost.c
> @@ -324,6 +324,24 @@ cleanup_device(struct virtio_net *dev, int destroy)
> }
> }
>
> +static void
> +vhost_free_async_mem(struct vhost_virtqueue *vq)
> +{
> + if (vq->async_pkts_pending)
> + rte_free(vq->async_pkts_pending);
> + if (vq->async_pkts_info)
> + rte_free(vq->async_pkts_info);
> + if (vq->it_pool)
> + rte_free(vq->it_pool);
> + if (vq->vec_pool)
> + rte_free(vq->vec_pool);
> +
> + vq->async_pkts_pending = NULL;
> + vq->async_pkts_info = NULL;
> + vq->it_pool = NULL;
> + vq->vec_pool = NULL;
> +}
> +
> void
> free_vq(struct virtio_net *dev, struct vhost_virtqueue *vq)
> {
> @@ -331,10 +349,7 @@ free_vq(struct virtio_net *dev, struct vhost_virtqueue *vq)
> rte_free(vq->shadow_used_packed);
> else {
> rte_free(vq->shadow_used_split);
> - if (vq->async_pkts_pending)
> - rte_free(vq->async_pkts_pending);
> - if (vq->async_pkts_info)
> - rte_free(vq->async_pkts_info);
> + vhost_free_async_mem(vq);
> }
> rte_free(vq->batch_copy_elems);
> rte_mempool_free(vq->iotlb_pool);
> @@ -1576,13 +1591,15 @@ int rte_vhost_async_channel_register(int vid, uint16_t queue_id,
> vq->async_pkts_info = rte_malloc(NULL,
> vq->size * sizeof(struct async_inflight_info),
> RTE_CACHE_LINE_SIZE);
> - if (!vq->async_pkts_pending || !vq->async_pkts_info) {
> - if (vq->async_pkts_pending)
> - rte_free(vq->async_pkts_pending);
> -
> - if (vq->async_pkts_info)
> - rte_free(vq->async_pkts_info);
> -
> + vq->it_pool = rte_malloc(NULL,
> + VHOST_MAX_ASYNC_IT * sizeof(struct rte_vhost_iov_iter),
> + RTE_CACHE_LINE_SIZE);
> + vq->vec_pool = rte_malloc(NULL,
> + VHOST_MAX_ASYNC_VEC * sizeof(struct iovec),
> + RTE_CACHE_LINE_SIZE);
> + if (!vq->async_pkts_pending || !vq->async_pkts_info ||
> + !vq->it_pool || !vq->vec_pool) {
> + vhost_free_async_mem(vq);
> VHOST_LOG_CONFIG(ERR,
> "async register failed: cannot allocate memory for vq data "
> "(vid %d, qid: %d)\n", vid, queue_id);
> @@ -1630,15 +1647,7 @@ int rte_vhost_async_channel_unregister(int vid, uint16_t queue_id)
> goto out;
> }
>
> - if (vq->async_pkts_pending) {
> - rte_free(vq->async_pkts_pending);
> - vq->async_pkts_pending = NULL;
> - }
> -
> - if (vq->async_pkts_info) {
> - rte_free(vq->async_pkts_info);
> - vq->async_pkts_info = NULL;
> - }
> + vhost_free_async_mem(vq);
>
> vq->async_ops.transfer_data = NULL;
> vq->async_ops.check_completed_copies = NULL;
> diff --git a/lib/librte_vhost/vhost.h b/lib/librte_vhost/vhost.h
> index 28aa77380..0af0ac23d 100644
> --- a/lib/librte_vhost/vhost.h
> +++ b/lib/librte_vhost/vhost.h
> @@ -218,8 +218,8 @@ struct vhost_virtqueue {
> /* operation callbacks for async dma */
> struct rte_vhost_async_channel_ops async_ops;
>
> - struct rte_vhost_iov_iter it_pool[VHOST_MAX_ASYNC_IT];
> - struct iovec vec_pool[VHOST_MAX_ASYNC_VEC];
> + struct rte_vhost_iov_iter *it_pool;
> + struct iovec *vec_pool;
>
> /* async data transfer status */
> uintptr_t **async_pkts_pending;
>
I think you should also take care of reallocating on the same numa node
the ring is (seel numa_realloc().
next prev parent reply other threads:[~2020-09-23 9:15 UTC|newest]
Thread overview: 36+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-09-11 1:53 [dpdk-dev] [PATCH v1 0/4] optimize async data path Patrick Fu
2020-09-11 1:53 ` [dpdk-dev] [PATCH v1 1/4] vhost: simplify async copy completion Patrick Fu
2020-09-23 9:07 ` Maxime Coquelin
2020-09-11 1:53 ` [dpdk-dev] [PATCH v1 2/4] vhost: dynamically alloc async memory Patrick Fu
2020-09-23 9:15 ` Maxime Coquelin [this message]
2020-09-29 5:55 ` Fu, Patrick
2020-09-11 1:53 ` [dpdk-dev] [PATCH v1 3/4] vhost: fix async vec buf overrun Patrick Fu
2020-09-23 9:21 ` Maxime Coquelin
2020-09-29 2:23 ` Fu, Patrick
2020-09-11 1:53 ` [dpdk-dev] [PATCH v1 4/4] vhost: fix async register/unregister deadlock Patrick Fu
2020-09-29 6:29 ` [dpdk-dev] [PATCH v2 0/4] optimize async data path Patrick Fu
2020-09-29 6:29 ` [dpdk-dev] [PATCH v2 1/4] vhost: simplify async copy completion Patrick Fu
2020-09-29 6:29 ` [dpdk-dev] [PATCH v2 2/4] vhost: dynamically allocate async memory Patrick Fu
2020-09-29 6:29 ` [dpdk-dev] [PATCH v2 3/4] vhost: fix async vector buffer overrun Patrick Fu
2020-09-29 6:29 ` [dpdk-dev] [PATCH v2 4/4] vhost: fix async register/unregister deadlock Patrick Fu
2020-09-29 9:29 ` [dpdk-dev] [PATCH v3 0/4] optimize async data path Patrick Fu
2020-09-29 9:29 ` [dpdk-dev] [PATCH v3 1/4] vhost: simplify async copy completion Patrick Fu
2020-10-05 13:46 ` Maxime Coquelin
2020-10-09 11:16 ` Fu, Patrick
2020-09-29 9:29 ` [dpdk-dev] [PATCH v3 2/4] vhost: dynamically allocate async memory Patrick Fu
2020-10-05 13:50 ` Maxime Coquelin
2020-09-29 9:29 ` [dpdk-dev] [PATCH v3 3/4] vhost: fix async vector buffer overrun Patrick Fu
2020-10-05 14:19 ` Maxime Coquelin
2020-09-29 9:29 ` [dpdk-dev] [PATCH v3 4/4] vhost: fix async register/unregister deadlock Patrick Fu
2020-10-05 14:25 ` Maxime Coquelin
2020-10-09 10:54 ` Fu, Patrick
2020-10-13 1:45 ` [dpdk-dev] [PATCH v4 0/4] optimize async data path Patrick Fu
2020-10-13 1:45 ` [dpdk-dev] [PATCH v4 1/4] vhost: simplify async copy completion Patrick Fu
2020-10-14 9:28 ` Maxime Coquelin
2020-10-13 1:45 ` [dpdk-dev] [PATCH v4 2/4] vhost: dynamically allocate async memory Patrick Fu
2020-10-14 9:30 ` Maxime Coquelin
2020-10-13 1:45 ` [dpdk-dev] [PATCH v4 3/4] vhost: fix async vector buffer overrun Patrick Fu
2020-10-14 9:33 ` Maxime Coquelin
2020-10-13 1:45 ` [dpdk-dev] [PATCH v4 4/4] vhost: fix async unregister deadlock Patrick Fu
2020-10-14 9:34 ` Maxime Coquelin
2020-10-15 15:40 ` [dpdk-dev] [PATCH v4 0/4] optimize async data path Maxime Coquelin
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=1522ca39-d9d7-72ff-8bb6-a5a29d12314f@redhat.com \
--to=maxime.coquelin@redhat.com \
--cc=chenbo.xia@intel.com \
--cc=cheng1.jiang@intel.com \
--cc=dev@dpdk.org \
--cc=patrick.fu@intel.com \
--cc=zhihong.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).