From: Dragos Tatulea <dtatulea@nvidia.com>
To: "Michael S. Tsirkin" <mst@redhat.com>,
"Jason Wang" <jasowang@redhat.com>,
"Xuan Zhuo" <xuanzhuo@linux.alibaba.com>,
"Eugenio Pérez" <eperezma@redhat.com>,
"Saeed Mahameed" <saeedm@nvidia.com>,
"Leon Romanovsky" <leon@kernel.org>,
"Tariq Toukan" <tariqt@nvidia.com>,
"Si-Wei Liu" <si-wei.liu@oracle.com>
Cc: <virtualization@lists.linux.dev>, <linux-kernel@vger.kernel.org>,
<linux-rdma@vger.kernel.org>, <netdev@vger.kernel.org>,
Dragos Tatulea <dtatulea@nvidia.com>,
Cosmin Ratiu <cratiu@nvidia.com>
Subject: [PATCH vhost 21/23] vdpa/mlx5: Re-create HW VQs under certain conditions
Date: Mon, 17 Jun 2024 18:07:55 +0300 [thread overview]
Message-ID: <20240617-stage-vdpa-vq-precreate-v1-21-8c0483f0ca2a@nvidia.com> (raw)
In-Reply-To: <20240617-stage-vdpa-vq-precreate-v1-0-8c0483f0ca2a@nvidia.com>
There are a few conditions under which the hardware VQs need a full
teardown and setup:
- VQ size changed to something else than default value. Hardware VQ size
modification is not supported.
- User turns off certain device features: mergeable buffers, checksum
virtio 1.0 compliance. In these cases, the TIR and RQT need to be
re-created.
Add a needs_teardown configuration variable and set it when detecting
the above scenarios. On next DRIVER_OK, the resources will be torn down
first.
Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Cosmin Ratiu <cratiu@nvidia.com>
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 15 +++++++++++++++
drivers/vdpa/mlx5/net/mlx5_vnet.h | 1 +
2 files changed, 16 insertions(+)
diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c b/drivers/vdpa/mlx5/net/mlx5_vnet.c
index b2836fd3d1dd..d80d6b47da61 100644
--- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
+++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c
@@ -2390,6 +2390,7 @@ static void mlx5_vdpa_set_vq_num(struct vdpa_device *vdev, u16 idx, u32 num)
}
mvq = &ndev->vqs[idx];
+ ndev->needs_teardown = num != mvq->num_ent;
mvq->num_ent = num;
}
@@ -2800,6 +2801,7 @@ static int mlx5_vdpa_set_driver_features(struct vdpa_device *vdev, u64 features)
struct mlx5_vdpa_dev *mvdev = to_mvdev(vdev);
struct mlx5_vdpa_net *ndev = to_mlx5_vdpa_ndev(mvdev);
u64 old_features = mvdev->actual_features;
+ u64 diff_features;
int err;
print_features(mvdev, features, true);
@@ -2822,6 +2824,14 @@ static int mlx5_vdpa_set_driver_features(struct vdpa_device *vdev, u64 features)
}
}
+ /* When below features diverge from initial device features, VQs need a full teardown. */
+#define NEEDS_TEARDOWN_MASK (BIT_ULL(VIRTIO_NET_F_MRG_RXBUF) | \
+ BIT_ULL(VIRTIO_NET_F_CSUM) | \
+ BIT_ULL(VIRTIO_F_VERSION_1))
+
+ diff_features = mvdev->mlx_features ^ mvdev->actual_features;
+ ndev->needs_teardown = !!(diff_features & NEEDS_TEARDOWN_MASK);
+
update_cvq_info(mvdev);
return err;
}
@@ -3038,6 +3048,7 @@ static void teardown_vq_resources(struct mlx5_vdpa_net *ndev)
destroy_rqt(ndev);
teardown_virtqueues(ndev);
ndev->setup = false;
+ ndev->needs_teardown = false;
}
static int setup_cvq_vring(struct mlx5_vdpa_dev *mvdev)
@@ -3078,6 +3089,10 @@ static void mlx5_vdpa_set_status(struct vdpa_device *vdev, u8 status)
goto err_setup;
}
register_link_notifier(ndev);
+
+ if (ndev->needs_teardown)
+ teardown_vq_resources(ndev);
+
if (ndev->setup) {
err = resume_vqs(ndev);
if (err) {
diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.h b/drivers/vdpa/mlx5/net/mlx5_vnet.h
index 2ada29767cc5..da7318f82d2a 100644
--- a/drivers/vdpa/mlx5/net/mlx5_vnet.h
+++ b/drivers/vdpa/mlx5/net/mlx5_vnet.h
@@ -56,6 +56,7 @@ struct mlx5_vdpa_net {
struct dentry *rx_dent;
struct dentry *rx_table_dent;
bool setup;
+ bool needs_teardown;
u32 cur_num_vqs;
u32 rqt_size;
u16 default_queue_size;
--
2.45.1
next prev parent reply other threads:[~2024-06-17 15:09 UTC|newest]
Thread overview: 53+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-06-17 15:07 [PATCH vhost 00/23] vdpa/mlx5: Pre-create HW VQs to reduce LM downtime Dragos Tatulea
2024-06-17 15:07 ` [PATCH vhost 01/23] vdpa/mlx5: Clarify meaning thorough function rename Dragos Tatulea
2024-06-19 10:37 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 02/23] vdpa/mlx5: Make setup/teardown_vq_resources() symmetrical Dragos Tatulea
2024-06-19 10:38 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 03/23] vdpa/mlx5: Drop redundant code Dragos Tatulea
2024-06-19 10:55 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 04/23] vdpa/mlx5: Drop redundant check in teardown_virtqueues() Dragos Tatulea
2024-06-19 10:56 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 05/23] vdpa/mlx5: Iterate over active VQs during suspend/resume Dragos Tatulea
2024-06-19 11:04 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 06/23] vdpa/mlx5: Remove duplicate suspend code Dragos Tatulea
2024-06-19 11:02 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 07/23] vdpa/mlx5: Initialize and reset device with one queue pair Dragos Tatulea
2024-06-17 15:07 ` [PATCH vhost 08/23] vdpa/mlx5: Clear and reinitialize software VQ data on reset Dragos Tatulea
2024-06-19 11:28 ` Eugenio Perez Martin
2024-06-19 17:03 ` Dragos Tatulea
2024-06-17 15:07 ` [PATCH vhost 09/23] vdpa/mlx5: Add support for modifying the virtio_version VQ field Dragos Tatulea
2024-06-17 15:07 ` [PATCH vhost 10/23] vdpa/mlx5: Add support for modifying the VQ features field Dragos Tatulea
2024-06-17 15:07 ` [PATCH vhost 11/23] vdpa/mlx5: Set an initial size on the VQ Dragos Tatulea
2024-06-19 15:08 ` Eugenio Perez Martin
2024-06-19 17:06 ` Dragos Tatulea
2024-06-17 15:07 ` [PATCH vhost 12/23] vdpa/mlx5: Start off rqt_size with max VQPs Dragos Tatulea
2024-06-19 15:33 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 13/23] vdpa/mlx5: Set mkey modified flags on all VQs Dragos Tatulea
2024-06-19 15:33 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 14/23] vdpa/mlx5: Allow creation of blank VQs Dragos Tatulea
2024-06-17 15:07 ` [PATCH vhost 15/23] vdpa/mlx5: Accept Init -> Ready VQ transition in resume_vq() Dragos Tatulea
2024-06-17 15:07 ` [PATCH vhost 16/23] vdpa/mlx5: Add error code for suspend/resume VQ Dragos Tatulea
2024-06-19 15:41 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 17/23] vdpa/mlx5: Consolidate all VQ modify to Ready to use resume_vq() Dragos Tatulea
2024-06-19 15:43 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 18/23] vdpa/mlx5: Forward error in suspend/resume device Dragos Tatulea
2024-06-23 11:19 ` Zhu Yanjun
2024-06-26 9:28 ` Dragos Tatulea
2024-06-17 15:07 ` [PATCH vhost 19/23] vdpa/mlx5: Use suspend/resume during VQP change Dragos Tatulea
2024-06-19 15:46 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 20/23] vdpa/mlx5: Pre-create hardware VQs at vdpa .dev_add time Dragos Tatulea
2024-06-19 15:54 ` Eugenio Perez Martin
2024-06-26 9:27 ` Dragos Tatulea
2024-07-03 16:01 ` Eugenio Perez Martin
2024-07-08 11:01 ` Dragos Tatulea
2024-07-08 11:11 ` Michael S. Tsirkin
2024-07-08 11:17 ` Dragos Tatulea
2024-07-08 11:25 ` Michael S. Tsirkin
2024-07-08 16:22 ` Zhu Yanjun
2024-07-08 16:43 ` Dragos Tatulea
2024-06-17 15:07 ` Dragos Tatulea [this message]
2024-06-19 16:04 ` [PATCH vhost 21/23] vdpa/mlx5: Re-create HW VQs under certain conditions Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 22/23] vdpa/mlx5: Don't reset VQs more than necessary Dragos Tatulea
2024-06-19 16:14 ` Eugenio Perez Martin
2024-06-17 15:07 ` [PATCH vhost 23/23] vdpa/mlx5: Don't enable non-active VQs in .set_vq_ready() Dragos Tatulea
2024-06-19 16:39 ` Eugenio Perez Martin
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=20240617-stage-vdpa-vq-precreate-v1-21-8c0483f0ca2a@nvidia.com \
--to=dtatulea@nvidia.com \
--cc=cratiu@nvidia.com \
--cc=eperezma@redhat.com \
--cc=jasowang@redhat.com \
--cc=leon@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-rdma@vger.kernel.org \
--cc=mst@redhat.com \
--cc=netdev@vger.kernel.org \
--cc=saeedm@nvidia.com \
--cc=si-wei.liu@oracle.com \
--cc=tariqt@nvidia.com \
--cc=virtualization@lists.linux.dev \
--cc=xuanzhuo@linux.alibaba.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).