From: "Xia, Chenbo" <chenbo.xia@intel.com>
To: "Hu, Jiayu" <jiayu.hu@intel.com>, "dev@dpdk.org" <dev@dpdk.org>
Cc: "maxime.coquelin@redhat.com" <maxime.coquelin@redhat.com>
Subject: Re: [dpdk-dev] [PATCH v4 2/3] vhost: rework async configuration struct
Date: Fri, 16 Jul 2021 06:03:16 +0000 [thread overview]
Message-ID: <MN2PR11MB4063FF9B8FB41FC265845A189C119@MN2PR11MB4063.namprd11.prod.outlook.com> (raw)
In-Reply-To: <1626162383-89674-3-git-send-email-jiayu.hu@intel.com>
Hi Jiayu,
> -----Original Message-----
> From: Hu, Jiayu <jiayu.hu@intel.com>
> Sent: Tuesday, July 13, 2021 3:46 PM
> To: dev@dpdk.org
> Cc: maxime.coquelin@redhat.com; Xia, Chenbo <chenbo.xia@intel.com>; Hu, Jiayu
> <jiayu.hu@intel.com>
> Subject: [PATCH v4 2/3] vhost: rework async configuration struct
Struct -> structure
>
> This patch reworks the async configuration structure to improve code
> readability. In addition, add preserved padding fields on the structure
> for future usage.
>
> Signed-off-by: Jiayu Hu <jiayu.hu@intel.com>
> ---
> doc/guides/prog_guide/vhost_lib.rst | 19 +++++++++++--------
> examples/vhost/main.c | 8 ++++----
> lib/vhost/rte_vhost_async.h | 35 +++++++++++++++++------------------
> lib/vhost/vhost.c | 13 +++++--------
> 4 files changed, 37 insertions(+), 38 deletions(-)
>
> diff --git a/doc/guides/prog_guide/vhost_lib.rst
> b/doc/guides/prog_guide/vhost_lib.rst
> index d18fb98..affdc57 100644
> --- a/doc/guides/prog_guide/vhost_lib.rst
> +++ b/doc/guides/prog_guide/vhost_lib.rst
> @@ -218,26 +218,29 @@ The following is an overview of some key Vhost API
> functions:
>
> Enable or disable zero copy feature of the vhost crypto backend.
>
> -* ``rte_vhost_async_channel_register(vid, queue_id, features, ops)``
> +* ``rte_vhost_async_channel_register(vid, queue_id, config, ops)``
>
> Register a vhost queue with async copy device channel after vring
Not related to this patch. But maybe 'Register an async copy device channel
for a vhost queue' is better?
> - is enabled. Following device ``features`` must be specified together
> + is enabled. Following device ``config`` must be specified together
> with the registration:
>
> - * ``async_inorder``
> + * ``features``
>
> - Async copy device can guarantee the ordering of copy completion
> - sequence. Copies are completed in the same order with that at
> - the submission time.
> + This field is used to specify async copy device features.
>
> - Currently, only ``async_inorder`` capable device is supported by vhost.
> + ``RTE_VHOST_ASYNC_INORDER`` represents the async copy device can
> + guarantee the order of copy completion is the same as the order
> + of copy submission.
> +
> + Currently, only ``RTE_VHOST_ASYNC_INORDER`` capable device is
> + supported by vhost.
>
> * ``async_threshold``
>
> The copy length (in bytes) below which CPU copy will be used even if
> applications call async vhost APIs to enqueue/dequeue data.
>
> - Typical value is 512~1024 depending on the async device capability.
> + Typical value is 256~1024 depending on the async device capability.
>
> Applications must provide following ``ops`` callbacks for vhost lib to
> work with the async copy devices:
> diff --git a/examples/vhost/main.c b/examples/vhost/main.c
> index d2179ea..9cd855a 100644
> --- a/examples/vhost/main.c
> +++ b/examples/vhost/main.c
> @@ -1468,7 +1468,7 @@ new_device(int vid)
> vid, vdev->coreid);
>
> if (async_vhost_driver) {
> - struct rte_vhost_async_features f;
> + struct rte_vhost_async_config config = {0};
> struct rte_vhost_async_channel_ops channel_ops;
>
> if (dma_type != NULL && strncmp(dma_type, "ioat", 4) == 0) {
> @@ -1476,11 +1476,11 @@ new_device(int vid)
> channel_ops.check_completed_copies =
> ioat_check_completed_copies_cb;
>
> - f.async_inorder = 1;
> - f.async_threshold = 256;
> + config.features = RTE_VHOST_ASYNC_INORDER;
> + config.async_threshold = 256;
This is ok as for now we are using ioat API. But I guess there should be
a more user-friendly way to know the value. I mean, no users want to do
complicated tests for his platform to know the value. Maybe there could be
some auto-test in specific driver to show the value to user?
It's off-topic for this patch. But thinking about it should be good.
>
> return rte_vhost_async_channel_register(vid, VIRTIO_RXQ,
> - f.intval, &channel_ops);
> + config, &channel_ops);
> }
> }
>
> diff --git a/lib/vhost/rte_vhost_async.h b/lib/vhost/rte_vhost_async.h
> index 6faa31f..c93490d 100644
> --- a/lib/vhost/rte_vhost_async.h
> +++ b/lib/vhost/rte_vhost_async.h
> @@ -93,18 +93,20 @@ struct async_inflight_info {
> };
>
> /**
> - * dma channel feature bit definition
> + * dma channel features
Let's use 'async channel' which is also the name for API.
It's always good to use one term for one thing.
> */
> -struct rte_vhost_async_features {
> - union {
> - uint32_t intval;
> - struct {
> - uint32_t async_inorder:1;
> - uint32_t resvd_0:15;
> - uint32_t async_threshold:12;
> - uint32_t resvd_1:4;
> - };
> - };
> +enum {
> + RTE_VHOST_ASYNC_FEATURE_UNKNOWN = 0U,
> + RTE_VHOST_ASYNC_INORDER = 1U << 0,
> +};
> +
> +/**
> + * dma channel configuration
Ditto
> + */
> +struct rte_vhost_async_config {
> + uint32_t async_threshold;
> + uint32_t features;
> + uint32_t resvd[2];
Generally we use 'rsvd' for the word 'reserved'.
> };
>
> /**
> @@ -114,12 +116,8 @@ struct rte_vhost_async_features {
> * vhost device id async channel to be attached to
> * @param queue_id
> * vhost queue id async channel to be attached to
> - * @param features
> - * DMA channel feature bit
> - * b0 : DMA supports inorder data transfer
> - * b1 - b15: reserved
> - * b16 - b27: Packet length threshold for DMA transfer
> - * b28 - b31: reserved
> + * @param config
> + * DMA channel configuration structure
DMA -> async
> * @param ops
> * DMA operation callbacks
> * @return
> @@ -127,7 +125,8 @@ struct rte_vhost_async_features {
> */
> __rte_experimental
> int rte_vhost_async_channel_register(int vid, uint16_t queue_id,
> - uint32_t features, struct rte_vhost_async_channel_ops *ops);
> + struct rte_vhost_async_config config,
> + struct rte_vhost_async_channel_ops *ops);
>
> /**
> * unregister a dma channel for vhost
> diff --git a/lib/vhost/vhost.c b/lib/vhost/vhost.c
> index 53a470f..a20f05a 100644
> --- a/lib/vhost/vhost.c
> +++ b/lib/vhost/vhost.c
> @@ -1620,18 +1620,15 @@ int rte_vhost_extern_callback_register(int vid,
> }
>
> int rte_vhost_async_channel_register(int vid, uint16_t queue_id,
> - uint32_t features,
> - struct rte_vhost_async_channel_ops *ops)
> + struct rte_vhost_async_config config,
> + struct rte_vhost_async_channel_ops *ops)
> {
> struct vhost_virtqueue *vq;
> struct virtio_net *dev = get_device(vid);
> - struct rte_vhost_async_features f;
>
> if (dev == NULL || ops == NULL)
> return -1;
>
> - f.intval = features;
> -
> if (queue_id >= VHOST_MAX_VRING)
> return -1;
>
> @@ -1640,7 +1637,7 @@ int rte_vhost_async_channel_register(int vid, uint16_t
> queue_id,
> if (unlikely(vq == NULL || !dev->async_copy))
> return -1;
>
> - if (unlikely(!f.async_inorder)) {
> + if (unlikely(!(config.features & RTE_VHOST_ASYNC_INORDER))) {
> VHOST_LOG_CONFIG(ERR,
> "async copy is not supported on non-inorder mode "
> "(vid %d, qid: %d)\n", vid, queue_id);
> @@ -1720,8 +1717,8 @@ int rte_vhost_async_channel_register(int vid, uint16_t
> queue_id,
> vq->async_ops.check_completed_copies = ops->check_completed_copies;
> vq->async_ops.transfer_data = ops->transfer_data;
>
> - vq->async_inorder = f.async_inorder;
> - vq->async_threshold = f.async_threshold;
> + vq->async_inorder = true;
Do we still need this? It's never used.
> + vq->async_threshold = config.async_threshold;
vq->async_threshold is uint16_t and config.async_threshold is uint32_t.
They should be the same.
Thanks,
Chenbo
>
> vq->async_registered = true;
>
> --
> 2.7.4
next prev parent reply other threads:[~2021-07-16 6:03 UTC|newest]
Thread overview: 49+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-28 8:11 [dpdk-dev] [PATCH 0/2] provide thread unsafe async registration functions Jiayu Hu
2021-05-28 8:11 ` [dpdk-dev] [PATCH 1/2] vhost: fix lock on device readiness notification Jiayu Hu
2021-07-02 7:36 ` Maxime Coquelin
2021-07-07 11:18 ` [dpdk-dev] [PATCH v2 0/3] provide thread unsafe async registration functions Jiayu Hu
2021-07-07 11:18 ` [dpdk-dev] [PATCH v2 1/3] vhost: fix lock on device readiness notification Jiayu Hu
2021-07-09 9:43 ` [dpdk-dev] [PATCH v3 0/3] provide thread unsafe async registration functions Jiayu Hu
2021-07-09 9:43 ` [dpdk-dev] [PATCH v3 1/3] vhost: fix lock on device readiness notification Jiayu Hu
2021-07-13 7:46 ` [dpdk-dev] [PATCH v4 0/3] provide thread unsafe async registration functions Jiayu Hu
2021-07-13 7:46 ` [dpdk-dev] [PATCH v4 1/3] vhost: fix lock on device readiness notification Jiayu Hu
2021-07-16 19:51 ` [dpdk-dev] [PATCH v5 0/3] provide thread unsafe async registration functions Jiayu Hu
2021-07-16 19:51 ` [dpdk-dev] [PATCH v5 1/3] vhost: fix lock on device readiness notification Jiayu Hu
2021-07-19 15:00 ` [dpdk-dev] [PATCH v6 0/3] provide thread unsafe async registration functions Jiayu Hu
2021-07-19 15:00 ` [dpdk-dev] [PATCH v6 1/3] vhost: fix lock on device readiness notification Jiayu Hu
2021-07-19 15:00 ` [dpdk-dev] [PATCH v6 2/3] vhost: rework async configuration structure Jiayu Hu
2021-07-20 7:18 ` Maxime Coquelin
2021-07-20 7:45 ` Hu, Jiayu
2021-07-19 15:00 ` [dpdk-dev] [PATCH v6 3/3] vhost: add thread unsafe async registeration functions Jiayu Hu
2021-07-20 7:25 ` Maxime Coquelin
2021-07-21 6:16 ` [dpdk-dev] [PATCH v6 0/3] provide thread unsafe async registration functions Xia, Chenbo
2021-07-16 19:51 ` [dpdk-dev] [PATCH v5 2/3] vhost: rework async configuration structure Jiayu Hu
2021-07-19 2:25 ` Xia, Chenbo
2021-07-19 7:20 ` Maxime Coquelin
2021-07-16 19:51 ` [dpdk-dev] [PATCH v5 3/3] vhost: add thread unsafe async registeration functions Jiayu Hu
2021-07-19 2:27 ` Xia, Chenbo
2021-07-13 7:46 ` [dpdk-dev] [PATCH v4 2/3] vhost: rework async configuration struct Jiayu Hu
2021-07-16 6:03 ` Xia, Chenbo [this message]
2021-07-16 6:18 ` Hu, Jiayu
2021-07-16 6:27 ` Xia, Chenbo
2021-07-16 6:34 ` Hu, Jiayu
2021-07-13 7:46 ` [dpdk-dev] [PATCH v4 3/3] vhost: add thread unsafe async registeration functions Jiayu Hu
2021-07-16 6:54 ` Xia, Chenbo
2021-07-09 9:43 ` [dpdk-dev] [PATCH v3 2/3] vhost: rework async configuration struct Jiayu Hu
2021-07-09 9:43 ` [dpdk-dev] [PATCH v3 3/3] vhost: add thread unsafe async registeration functions Jiayu Hu
2021-07-07 11:18 ` [dpdk-dev] [PATCH v2 2/3] vhost: rework async feature struct Jiayu Hu
2021-07-07 6:39 ` Maxime Coquelin
2021-07-07 11:18 ` [dpdk-dev] [PATCH v2 3/3] vhost: add thread unsafe async registeration functions Jiayu Hu
2021-05-28 8:11 ` [dpdk-dev] [PATCH 2/2] vhost: add thread unsafe async registration functions Jiayu Hu
2021-07-02 7:40 ` Maxime Coquelin
2021-07-05 1:35 ` Hu, Jiayu
2021-07-05 8:58 ` Maxime Coquelin
2021-07-06 8:36 ` Hu, Jiayu
2021-06-04 7:18 ` [dpdk-dev] [PATCH 0/2] provide " Maxime Coquelin
2021-06-04 7:24 ` Maxime Coquelin
2021-06-07 8:07 ` Hu, Jiayu
2021-06-07 13:20 ` Maxime Coquelin
2021-06-08 6:36 ` Hu, Jiayu
2021-06-29 5:36 ` Hu, Jiayu
2021-07-01 15:42 ` Maxime Coquelin
2021-07-02 0:53 ` Hu, Jiayu
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=MN2PR11MB4063FF9B8FB41FC265845A189C119@MN2PR11MB4063.namprd11.prod.outlook.com \
--to=chenbo.xia@intel.com \
--cc=dev@dpdk.org \
--cc=jiayu.hu@intel.com \
--cc=maxime.coquelin@redhat.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).