From: Si-Wei Liu <si-wei.liu@oracle.com>
To: Eli Cohen <elic@nvidia.com>, mst@redhat.com, jasowang@redhat.com
Cc: virtualization@lists.linux-foundation.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v3 1/2] vdpa: Add support for querying vendor statistics
Date: Mon, 2 May 2022 16:47:56 -0700 [thread overview]
Message-ID: <f279bbbb-6af2-95f4-08c3-97ba72deb6cb@oracle.com> (raw)
In-Reply-To: <20220502102201.190357-2-elic@nvidia.com>
On 5/2/2022 3:22 AM, Eli Cohen wrote:
> Allows to read vendor statistics of a vdpa device. The specific
> statistics data are received from the upstream driver in the form of an
> (attribute name, attribute value) pairs.
>
> An example of statistics for mlx5_vdpa device are:
>
> received_desc - number of descriptors received by the virtqueue
> completed_desc - number of descriptors completed by the virtqueue
>
> A descriptor using indirect buffers is still counted as 1. In addition,
> N chained descriptors are counted correctly N times as one would expect.
>
> A new callback was added to vdpa_config_ops which provides the means for
> the vdpa driver to return statistics results.
>
> The interface allows for reading all the supported virtqueues, including
> the control virtqueue if it exists.
>
> Below are some examples taken from mlx5_vdpa which are introduced in the
> following patch:
>
> 1. Read statistics for the virtqueue at index 1
>
> $ vdpa dev vstats show vdpa-a qidx 1
> vdpa-a:
> queue_type tx queue_index 1 received_desc 3844836 completed_desc 3844836
>
> 2. Read statistics for the virtqueue at index 32
> $ vdpa dev vstats show vdpa-a qidx 32
> vdpa-a:
> queue_type control_vq queue_index 32 received_desc 62 completed_desc 62
>
> 3. Read statisitics for the virtqueue at index 0 with json output
> $ vdpa -j dev vstats show vdpa-a qidx 0
> {"vstats":{"vdpa-a":{
> "queue_type":"rx","queue_index":0,"name":"received_desc","value":417776,\
> "name":"completed_desc","value":417548}}}
>
> 4. Read statistics for the virtqueue at index 0 with preety json output
> $ vdpa -jp dev vstats show vdpa-a qidx 0
> {
> "vstats": {
> "vdpa-a": {
>
> "queue_type": "rx",
> "queue_index": 0,
> "name": "received_desc",
> "value": 417776,
> "name": "completed_desc",
> "value": 417548
> }
> }
> }
>
> Signed-off-by: Eli Cohen <elic@nvidia.com>
> ---
> drivers/vdpa/vdpa.c | 129 ++++++++++++++++++++++++++++++++++++++
> include/linux/vdpa.h | 5 ++
> include/uapi/linux/vdpa.h | 6 ++
> 3 files changed, 140 insertions(+)
>
> diff --git a/drivers/vdpa/vdpa.c b/drivers/vdpa/vdpa.c
> index 2b75c00b1005..933466f61ca8 100644
> --- a/drivers/vdpa/vdpa.c
> +++ b/drivers/vdpa/vdpa.c
> @@ -909,6 +909,74 @@ vdpa_dev_config_fill(struct vdpa_device *vdev, struct sk_buff *msg, u32 portid,
> return err;
> }
>
> +static int vdpa_fill_stats_rec(struct vdpa_device *vdev, struct sk_buff *msg,
> + struct genl_info *info, u32 index)
> +{
> + int err;
> +
> + err = vdev->config->get_vendor_vq_stats(vdev, index, msg, info->extack);
> + if (err)
> + return err;
> +
> + if (nla_put_u32(msg, VDPA_ATTR_DEV_QUEUE_INDEX, index))
> + return -EMSGSIZE;
> +
> + return 0;
> +}
> +
> +static int vendor_stats_fill(struct vdpa_device *vdev, struct sk_buff *msg,
> + struct genl_info *info, u32 index)
> +{
> + int err;
> +
> + if (!vdev->config->get_vendor_vq_stats)
> + return -EOPNOTSUPP;
> +
> + err = vdpa_fill_stats_rec(vdev, msg, info, index);
> + if (err)
> + return err;
> +
> + return 0;
> +}
> +
> +static int vdpa_dev_vendor_stats_fill(struct vdpa_device *vdev,
> + struct sk_buff *msg,
> + struct genl_info *info, u32 index)
> +{
> + u32 device_id;
> + void *hdr;
> + int err;
> + u32 portid = info->snd_portid;
> + u32 seq = info->snd_seq;
> + u32 flags = 0;
> +
> + hdr = genlmsg_put(msg, portid, seq, &vdpa_nl_family, flags,
> + VDPA_CMD_DEV_VSTATS_GET);
> + if (!hdr)
> + return -EMSGSIZE;
> +
> + if (nla_put_string(msg, VDPA_ATTR_DEV_NAME, dev_name(&vdev->dev))) {
> + err = -EMSGSIZE;
> + goto undo_msg;
> + }
> +
> + device_id = vdev->config->get_device_id(vdev);
> + if (nla_put_u32(msg, VDPA_ATTR_DEV_ID, device_id)) {
> + err = -EMSGSIZE;
> + goto undo_msg;
> + }
> +
You seem to miss VDPA_ATTR_DEV_NEGOTIATED_FEATURES from this function,
otherwise I can't image how you can ensure the atomicity to infer
queue_type for control vq.
And should we make sure VIRTIO_CONFIG_S_FEATURES_OK is set before call
into vendor_stats_fill()?
> + err = vendor_stats_fill(vdev, msg, info, index);
> +
> + genlmsg_end(msg, hdr);
> +
> + return err;
> +
> +undo_msg:
> + genlmsg_cancel(msg, hdr);
> + return err;
> +}
> +
> static int vdpa_nl_cmd_dev_config_get_doit(struct sk_buff *skb, struct genl_info *info)
> {
> struct vdpa_device *vdev;
> @@ -990,6 +1058,60 @@ vdpa_nl_cmd_dev_config_get_dumpit(struct sk_buff *msg, struct netlink_callback *
> return msg->len;
> }
>
> +static int vdpa_nl_cmd_dev_stats_get_doit(struct sk_buff *skb,
> + struct genl_info *info)
> +{
> + struct vdpa_device *vdev;
> + struct sk_buff *msg;
> + const char *devname;
> + struct device *dev;
> + u32 index;
> + int err;
> +
> + if (!info->attrs[VDPA_ATTR_DEV_NAME])
> + return -EINVAL;
> +
> + if (!info->attrs[VDPA_ATTR_DEV_QUEUE_INDEX])
> + return -EINVAL;
> +
> + devname = nla_data(info->attrs[VDPA_ATTR_DEV_NAME]);
> + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
> + if (!msg)
> + return -ENOMEM;
> +
> + index = nla_get_u32(info->attrs[VDPA_ATTR_DEV_QUEUE_INDEX]);
> + mutex_lock(&vdpa_dev_mutex);
Given that stats_get() is a read-only operation that might happen quite
often from time to time, I wonder if it is now a good timing to convert
vdpa_dev_mutex to a semaphore?
> + dev = bus_find_device(&vdpa_bus, NULL, devname, vdpa_name_match);
> + if (!dev) {
> + NL_SET_ERR_MSG_MOD(info->extack, "device not found");
> + err = -ENODEV;
> + goto dev_err;
Missing nlmsg_free().
> + }
> + vdev = container_of(dev, struct vdpa_device, dev);
> + if (!vdev->mdev) {
> + NL_SET_ERR_MSG_MOD(info->extack, "unmanaged vdpa device");
> + err = -EINVAL;
> + goto mdev_err;
Missing nlmsg_free().
Otherwise looks fine.
Acked-by: Si-Wei Liu <si-wei.liu@oracle.com>
-Siwei
> + }
> + err = vdpa_dev_vendor_stats_fill(vdev, msg, info, index);
> + if (!err)
> + err = genlmsg_reply(msg, info);
> +
> + put_device(dev);
> + mutex_unlock(&vdpa_dev_mutex);
> +
> + if (err)
> + nlmsg_free(msg);
> +
> + return err;
> +
> +mdev_err:
> + put_device(dev);
> +dev_err:
> + mutex_unlock(&vdpa_dev_mutex);
> + return err;
> +}
> +
> static const struct nla_policy vdpa_nl_policy[VDPA_ATTR_MAX + 1] = {
> [VDPA_ATTR_MGMTDEV_BUS_NAME] = { .type = NLA_NUL_STRING },
> [VDPA_ATTR_MGMTDEV_DEV_NAME] = { .type = NLA_STRING },
> @@ -997,6 +1119,7 @@ static const struct nla_policy vdpa_nl_policy[VDPA_ATTR_MAX + 1] = {
> [VDPA_ATTR_DEV_NET_CFG_MACADDR] = NLA_POLICY_ETH_ADDR,
> /* virtio spec 1.1 section 5.1.4.1 for valid MTU range */
> [VDPA_ATTR_DEV_NET_CFG_MTU] = NLA_POLICY_MIN(NLA_U16, 68),
> + [VDPA_ATTR_DEV_QUEUE_INDEX] = NLA_POLICY_RANGE(NLA_U32, 0, 65535),
> };
>
> static const struct genl_ops vdpa_nl_ops[] = {
> @@ -1030,6 +1153,12 @@ static const struct genl_ops vdpa_nl_ops[] = {
> .doit = vdpa_nl_cmd_dev_config_get_doit,
> .dumpit = vdpa_nl_cmd_dev_config_get_dumpit,
> },
> + {
> + .cmd = VDPA_CMD_DEV_VSTATS_GET,
> + .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
> + .doit = vdpa_nl_cmd_dev_stats_get_doit,
> + .flags = GENL_ADMIN_PERM,
> + },
> };
>
> static struct genl_family vdpa_nl_family __ro_after_init = {
> diff --git a/include/linux/vdpa.h b/include/linux/vdpa.h
> index 8943a209202e..48ed1fc00830 100644
> --- a/include/linux/vdpa.h
> +++ b/include/linux/vdpa.h
> @@ -276,6 +276,9 @@ struct vdpa_config_ops {
> const struct vdpa_vq_state *state);
> int (*get_vq_state)(struct vdpa_device *vdev, u16 idx,
> struct vdpa_vq_state *state);
> + int (*get_vendor_vq_stats)(struct vdpa_device *vdev, u16 idx,
> + struct sk_buff *msg,
> + struct netlink_ext_ack *extack);
> struct vdpa_notification_area
> (*get_vq_notification)(struct vdpa_device *vdev, u16 idx);
> /* vq irq is not expected to be changed once DRIVER_OK is set */
> @@ -473,4 +476,6 @@ struct vdpa_mgmt_dev {
> int vdpa_mgmtdev_register(struct vdpa_mgmt_dev *mdev);
> void vdpa_mgmtdev_unregister(struct vdpa_mgmt_dev *mdev);
>
> +#define VDPA_INVAL_QUEUE_INDEX 0xffff
> +
> #endif /* _LINUX_VDPA_H */
> diff --git a/include/uapi/linux/vdpa.h b/include/uapi/linux/vdpa.h
> index 1061d8d2d09d..25c55cab3d7c 100644
> --- a/include/uapi/linux/vdpa.h
> +++ b/include/uapi/linux/vdpa.h
> @@ -18,6 +18,7 @@ enum vdpa_command {
> VDPA_CMD_DEV_DEL,
> VDPA_CMD_DEV_GET, /* can dump */
> VDPA_CMD_DEV_CONFIG_GET, /* can dump */
> + VDPA_CMD_DEV_VSTATS_GET,
> };
>
> enum vdpa_attr {
> @@ -46,6 +47,11 @@ enum vdpa_attr {
> VDPA_ATTR_DEV_NEGOTIATED_FEATURES, /* u64 */
> VDPA_ATTR_DEV_MGMTDEV_MAX_VQS, /* u32 */
> VDPA_ATTR_DEV_SUPPORTED_FEATURES, /* u64 */
> +
> + VDPA_ATTR_DEV_QUEUE_INDEX, /* u32 */
> + VDPA_ATTR_DEV_VENDOR_ATTR_NAME, /* string */
> + VDPA_ATTR_DEV_VENDOR_ATTR_VALUE, /* u64 */
> +
> /* new attributes must be added above here */
> VDPA_ATTR_MAX,
> };
next prev parent reply other threads:[~2022-05-02 23:50 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-05-02 10:21 [PATCH v3 0/2] Show statistics for a vdpa device Eli Cohen
2022-05-02 10:22 ` [PATCH v3 1/2] vdpa: Add support for querying vendor statistics Eli Cohen
2022-05-02 23:47 ` Si-Wei Liu [this message]
2022-05-03 5:13 ` Eli Cohen
2022-05-04 4:43 ` Si-Wei Liu
2022-05-04 5:44 ` Eli Cohen
2022-05-05 8:10 ` Jason Wang
2022-05-06 0:35 ` Si-Wei Liu
2022-05-08 7:44 ` Eli Cohen
2022-05-10 17:58 ` Si-Wei Liu
2022-05-11 10:07 ` Eli Cohen
2022-05-02 10:22 ` [PATCH v3 2/2] vdpa/mlx5: Add support for reading descriptor statistics Eli Cohen
2022-05-03 0:36 ` Si-Wei Liu
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=f279bbbb-6af2-95f4-08c3-97ba72deb6cb@oracle.com \
--to=si-wei.liu@oracle.com \
--cc=elic@nvidia.com \
--cc=jasowang@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mst@redhat.com \
--cc=virtualization@lists.linux-foundation.org \
/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).