Currently, coalescing parameters are grouped for all transmit and receive
virtqueues. This patch series add support to set or get the parameters for
a specified virtqueue.
When the traffic between virtqueues is unbalanced, for example, one virtqueue
is busy and another virtqueue is idle, then it will be very useful to
control coalescing parameters at the virtqueue granularity.
Example command:
$ ethtool -Q eth5 queue_mask 0x1 --coalesce tx-packets 10
Would set max_packets=10 to VQ 1.
$ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10
Would set max_packets=10 to VQ 0.
$ ethtool -Q eth5 queue_mask 0x1 --show-coalesce
Queue: 0
Adaptive RX: off TX: off
stats-block-usecs: 0
sample-interval: 0
pkt-rate-low: 0
pkt-rate-high: 0
rx-usecs: 222
rx-frames: 0
rx-usecs-irq: 0
rx-frames-irq: 256
tx-usecs: 222
tx-frames: 0
tx-usecs-irq: 0
tx-frames-irq: 256
rx-usecs-low: 0
rx-frame-low: 0
tx-usecs-low: 0
tx-frame-low: 0
rx-usecs-high: 0
rx-frame-high: 0
tx-usecs-high: 0
tx-frame-high: 0
In this patch series:
Patch-1: Extract interrupt coalescing settings to a structure.
Patch-2: Extract get/set interrupt coalesce to a function.
Patch-3: Support per queue interrupt coalesce command.
Patch-4: Enable per queue interrupt coalesce feature.
Gavin Li (4):
virtio_net: extract interrupt coalescing settings to a structure
virtio_net: extract get/set interrupt coalesce to a function
virtio_net: support per queue interrupt coalesce command
virtio_net: enable per queue interrupt coalesce feature
drivers/net/virtio_net.c | 169 ++++++++++++++++++++++++++------
include/uapi/linux/virtio_net.h | 14 +++
2 files changed, 154 insertions(+), 29 deletions(-)
--
2.39.1
Extract interrupt coalescing settings to a structure so that it could be
reused in other data structures.
Signed-off-by: Gavin Li <[email protected]>
Reviewed-by: Dragos Tatulea <[email protected]>
Reviewed-by: Jiri Pirko <[email protected]>
---
drivers/net/virtio_net.c | 35 +++++++++++++++++++----------------
1 file changed, 19 insertions(+), 16 deletions(-)
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 0db14f6b87d3..dd5fec073a27 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -126,6 +126,11 @@ static const struct virtnet_stat_desc virtnet_rq_stats_desc[] = {
#define VIRTNET_SQ_STATS_LEN ARRAY_SIZE(virtnet_sq_stats_desc)
#define VIRTNET_RQ_STATS_LEN ARRAY_SIZE(virtnet_rq_stats_desc)
+struct virtnet_interrupt_coalesce {
+ u32 max_packets;
+ u32 max_usecs;
+};
+
/* Internal representation of a send virtqueue */
struct send_queue {
/* Virtqueue associated with this send _queue */
@@ -281,10 +286,8 @@ struct virtnet_info {
u32 speed;
/* Interrupt coalescing settings */
- u32 tx_usecs;
- u32 rx_usecs;
- u32 tx_max_packets;
- u32 rx_max_packets;
+ struct virtnet_interrupt_coalesce intr_coal_tx;
+ struct virtnet_interrupt_coalesce intr_coal_rx;
unsigned long guest_offloads;
unsigned long guest_offloads_capable;
@@ -3056,8 +3059,8 @@ static int virtnet_send_notf_coal_cmds(struct virtnet_info *vi,
return -EINVAL;
/* Save parameters */
- vi->tx_usecs = ec->tx_coalesce_usecs;
- vi->tx_max_packets = ec->tx_max_coalesced_frames;
+ vi->intr_coal_tx.max_usecs = ec->tx_coalesce_usecs;
+ vi->intr_coal_tx.max_packets = ec->tx_max_coalesced_frames;
vi->ctrl->coal_rx.rx_usecs = cpu_to_le32(ec->rx_coalesce_usecs);
vi->ctrl->coal_rx.rx_max_packets = cpu_to_le32(ec->rx_max_coalesced_frames);
@@ -3069,8 +3072,8 @@ static int virtnet_send_notf_coal_cmds(struct virtnet_info *vi,
return -EINVAL;
/* Save parameters */
- vi->rx_usecs = ec->rx_coalesce_usecs;
- vi->rx_max_packets = ec->rx_max_coalesced_frames;
+ vi->intr_coal_rx.max_usecs = ec->rx_coalesce_usecs;
+ vi->intr_coal_rx.max_packets = ec->rx_max_coalesced_frames;
return 0;
}
@@ -3132,10 +3135,10 @@ static int virtnet_get_coalesce(struct net_device *dev,
struct virtnet_info *vi = netdev_priv(dev);
if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL)) {
- ec->rx_coalesce_usecs = vi->rx_usecs;
- ec->tx_coalesce_usecs = vi->tx_usecs;
- ec->tx_max_coalesced_frames = vi->tx_max_packets;
- ec->rx_max_coalesced_frames = vi->rx_max_packets;
+ ec->rx_coalesce_usecs = vi->intr_coal_rx.max_usecs;
+ ec->tx_coalesce_usecs = vi->intr_coal_tx.max_usecs;
+ ec->tx_max_coalesced_frames = vi->intr_coal_tx.max_packets;
+ ec->rx_max_coalesced_frames = vi->intr_coal_rx.max_packets;
} else {
ec->rx_max_coalesced_frames = 1;
@@ -4119,10 +4122,10 @@ static int virtnet_probe(struct virtio_device *vdev)
}
if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL)) {
- vi->rx_usecs = 0;
- vi->tx_usecs = 0;
- vi->tx_max_packets = 0;
- vi->rx_max_packets = 0;
+ vi->intr_coal_rx.max_usecs = 0;
+ vi->intr_coal_tx.max_usecs = 0;
+ vi->intr_coal_tx.max_packets = 0;
+ vi->intr_coal_rx.max_packets = 0;
}
if (virtio_has_feature(vdev, VIRTIO_NET_F_HASH_REPORT))
--
2.39.1
On Mon, 10 Jul 2023 12:20:01 +0300, Gavin Li <[email protected]> wrote:
As far as I know, Heng Qi does that. I'm not sure, it's the same piece.
cc @Heng Qi
Thanks.
> Currently, coalescing parameters are grouped for all transmit and receive
> virtqueues. This patch series add support to set or get the parameters for
> a specified virtqueue.
>
> When the traffic between virtqueues is unbalanced, for example, one virtqueue
> is busy and another virtqueue is idle, then it will be very useful to
> control coalescing parameters at the virtqueue granularity.
>
> Example command:
> $ ethtool -Q eth5 queue_mask 0x1 --coalesce tx-packets 10
> Would set max_packets=10 to VQ 1.
> $ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10
> Would set max_packets=10 to VQ 0.
> $ ethtool -Q eth5 queue_mask 0x1 --show-coalesce
> Queue: 0
> Adaptive RX: off TX: off
> stats-block-usecs: 0
> sample-interval: 0
> pkt-rate-low: 0
> pkt-rate-high: 0
>
> rx-usecs: 222
> rx-frames: 0
> rx-usecs-irq: 0
> rx-frames-irq: 256
>
> tx-usecs: 222
> tx-frames: 0
> tx-usecs-irq: 0
> tx-frames-irq: 256
>
> rx-usecs-low: 0
> rx-frame-low: 0
> tx-usecs-low: 0
> tx-frame-low: 0
>
> rx-usecs-high: 0
> rx-frame-high: 0
> tx-usecs-high: 0
> tx-frame-high: 0
>
> In this patch series:
> Patch-1: Extract interrupt coalescing settings to a structure.
> Patch-2: Extract get/set interrupt coalesce to a function.
> Patch-3: Support per queue interrupt coalesce command.
> Patch-4: Enable per queue interrupt coalesce feature.
>
> Gavin Li (4):
> virtio_net: extract interrupt coalescing settings to a structure
> virtio_net: extract get/set interrupt coalesce to a function
> virtio_net: support per queue interrupt coalesce command
> virtio_net: enable per queue interrupt coalesce feature
>
> drivers/net/virtio_net.c | 169 ++++++++++++++++++++++++++------
> include/uapi/linux/virtio_net.h | 14 +++
> 2 files changed, 154 insertions(+), 29 deletions(-)
>
> --
> 2.39.1
>
Add interrupt_coalesce config in send_queue and receive_queue to cache user
config.
Send per virtqueue interrupt moderation config to underline device in order
to have more efficient interrupt moderation and cpu utilization of guest
VM.
Signed-off-by: Gavin Li <[email protected]>
Reviewed-by: Dragos Tatulea <[email protected]>
Reviewed-by: Jiri Pirko <[email protected]>
---
drivers/net/virtio_net.c | 117 ++++++++++++++++++++++++++++----
include/uapi/linux/virtio_net.h | 14 ++++
2 files changed, 119 insertions(+), 12 deletions(-)
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 802ed21453f5..333a38e1941f 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -144,6 +144,8 @@ struct send_queue {
struct virtnet_sq_stats stats;
+ struct virtnet_interrupt_coalesce intr_coal;
+
struct napi_struct napi;
/* Record whether sq is in reset state. */
@@ -161,6 +163,8 @@ struct receive_queue {
struct virtnet_rq_stats stats;
+ struct virtnet_interrupt_coalesce intr_coal;
+
/* Chain pages by the private ptr. */
struct page *pages;
@@ -3078,6 +3082,56 @@ static int virtnet_send_notf_coal_cmds(struct virtnet_info *vi,
return 0;
}
+static int virtnet_send_ctrl_coal_vq_cmd(struct virtnet_info *vi,
+ u16 vqn, u32 max_usecs, u32 max_packets)
+{
+ struct virtio_net_ctrl_coal_vq coal_vq = {};
+ struct scatterlist sgs;
+
+ coal_vq.vqn = cpu_to_le16(vqn);
+ coal_vq.coal.max_usecs = cpu_to_le32(max_usecs);
+ coal_vq.coal.max_packets = cpu_to_le32(max_packets);
+ sg_init_one(&sgs, &coal_vq, sizeof(coal_vq));
+
+ if (!virtnet_send_command(vi, VIRTIO_NET_CTRL_NOTF_COAL,
+ VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET,
+ &sgs))
+ return -EINVAL;
+
+ return 0;
+}
+
+static int virtnet_send_notf_coal_vq_cmds(struct virtnet_info *vi,
+ struct ethtool_coalesce *ec,
+ u16 queue)
+{
+ int err;
+
+ if (ec->rx_coalesce_usecs || ec->rx_max_coalesced_frames) {
+ err = virtnet_send_ctrl_coal_vq_cmd(vi, rxq2vq(queue),
+ ec->rx_coalesce_usecs,
+ ec->rx_max_coalesced_frames);
+ if (err)
+ return err;
+ /* Save parameters */
+ vi->rq[queue].intr_coal.max_usecs = ec->rx_coalesce_usecs;
+ vi->rq[queue].intr_coal.max_packets = ec->rx_max_coalesced_frames;
+ }
+
+ if (ec->tx_coalesce_usecs || ec->tx_max_coalesced_frames) {
+ err = virtnet_send_ctrl_coal_vq_cmd(vi, txq2vq(queue),
+ ec->tx_coalesce_usecs,
+ ec->tx_max_coalesced_frames);
+ if (err)
+ return err;
+ /* Save parameters */
+ vi->sq[queue].intr_coal.max_usecs = ec->tx_coalesce_usecs;
+ vi->sq[queue].intr_coal.max_packets = ec->tx_max_coalesced_frames;
+ }
+
+ return 0;
+}
+
static int virtnet_coal_params_supported(struct ethtool_coalesce *ec)
{
/* usecs coalescing is supported only if VIRTIO_NET_F_NOTF_COAL
@@ -3094,23 +3148,36 @@ static int virtnet_coal_params_supported(struct ethtool_coalesce *ec)
}
static int virtnet_set_coalesce_one(struct net_device *dev,
- struct ethtool_coalesce *ec)
+ struct ethtool_coalesce *ec,
+ bool per_queue,
+ u32 queue)
{
struct virtnet_info *vi = netdev_priv(dev);
- int ret, i, napi_weight;
+ int queue_count = per_queue ? 1 : vi->max_queue_pairs;
+ int queue_number = per_queue ? queue : 0;
bool update_napi = false;
+ int ret, i, napi_weight;
+
+ if (queue >= vi->max_queue_pairs)
+ return -EINVAL;
/* Can't change NAPI weight if the link is up */
napi_weight = ec->tx_max_coalesced_frames ? NAPI_POLL_WEIGHT : 0;
- if (napi_weight ^ vi->sq[0].napi.weight) {
- if (dev->flags & IFF_UP)
- return -EBUSY;
- else
+ for (i = queue_number; i < queue_count; i++) {
+ if (napi_weight ^ vi->sq[i].napi.weight) {
+ if (dev->flags & IFF_UP)
+ return -EBUSY;
+
update_napi = true;
+ queue_number = i;
+ break;
+ }
}
- if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL))
+ if (!per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL))
ret = virtnet_send_notf_coal_cmds(vi, ec);
+ else if (per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL))
+ ret = virtnet_send_notf_coal_vq_cmds(vi, ec, queue);
else
ret = virtnet_coal_params_supported(ec);
@@ -3118,7 +3185,7 @@ static int virtnet_set_coalesce_one(struct net_device *dev,
return ret;
if (update_napi) {
- for (i = 0; i < vi->max_queue_pairs; i++)
+ for (i = queue_number; i < queue_count; i++)
vi->sq[i].napi.weight = napi_weight;
}
@@ -3130,19 +3197,29 @@ static int virtnet_set_coalesce(struct net_device *dev,
struct kernel_ethtool_coalesce *kernel_coal,
struct netlink_ext_ack *extack)
{
- return virtnet_set_coalesce_one(dev, ec);
+ return virtnet_set_coalesce_one(dev, ec, false, 0);
}
static int virtnet_get_coalesce_one(struct net_device *dev,
- struct ethtool_coalesce *ec)
+ struct ethtool_coalesce *ec,
+ bool per_queue,
+ u32 queue)
{
struct virtnet_info *vi = netdev_priv(dev);
- if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL)) {
+ if (queue >= vi->max_queue_pairs)
+ return -EINVAL;
+
+ if (!per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL)) {
ec->rx_coalesce_usecs = vi->intr_coal_rx.max_usecs;
ec->tx_coalesce_usecs = vi->intr_coal_tx.max_usecs;
ec->tx_max_coalesced_frames = vi->intr_coal_tx.max_packets;
ec->rx_max_coalesced_frames = vi->intr_coal_rx.max_packets;
+ } else if (per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL)) {
+ ec->rx_coalesce_usecs = vi->rq[queue].intr_coal.max_usecs;
+ ec->tx_coalesce_usecs = vi->sq[queue].intr_coal.max_usecs;
+ ec->tx_max_coalesced_frames = vi->sq[queue].intr_coal.max_packets;
+ ec->rx_max_coalesced_frames = vi->rq[queue].intr_coal.max_packets;
} else {
ec->rx_max_coalesced_frames = 1;
@@ -3158,7 +3235,21 @@ static int virtnet_get_coalesce(struct net_device *dev,
struct kernel_ethtool_coalesce *kernel_coal,
struct netlink_ext_ack *extack)
{
- return virtnet_get_coalesce_one(dev, ec);
+ return virtnet_get_coalesce_one(dev, ec, false, 0);
+}
+
+static int virtnet_set_per_queue_coalesce(struct net_device *dev,
+ u32 queue,
+ struct ethtool_coalesce *ec)
+{
+ return virtnet_set_coalesce_one(dev, ec, true, queue);
+}
+
+static int virtnet_get_per_queue_coalesce(struct net_device *dev,
+ u32 queue,
+ struct ethtool_coalesce *ec)
+{
+ return virtnet_get_coalesce_one(dev, ec, true, queue);
}
static void virtnet_init_settings(struct net_device *dev)
@@ -3291,6 +3382,8 @@ static const struct ethtool_ops virtnet_ethtool_ops = {
.set_link_ksettings = virtnet_set_link_ksettings,
.set_coalesce = virtnet_set_coalesce,
.get_coalesce = virtnet_get_coalesce,
+ .set_per_queue_coalesce = virtnet_set_per_queue_coalesce,
+ .get_per_queue_coalesce = virtnet_get_per_queue_coalesce,
.get_rxfh_key_size = virtnet_get_rxfh_key_size,
.get_rxfh_indir_size = virtnet_get_rxfh_indir_size,
.get_rxfh = virtnet_get_rxfh,
diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h
index 12c1c9699935..cc65ef0f3c3e 100644
--- a/include/uapi/linux/virtio_net.h
+++ b/include/uapi/linux/virtio_net.h
@@ -56,6 +56,7 @@
#define VIRTIO_NET_F_MQ 22 /* Device supports Receive Flow
* Steering */
#define VIRTIO_NET_F_CTRL_MAC_ADDR 23 /* Set MAC address */
+#define VIRTIO_NET_F_VQ_NOTF_COAL 52 /* Device supports virtqueue notification coalescing */
#define VIRTIO_NET_F_NOTF_COAL 53 /* Device supports notifications coalescing */
#define VIRTIO_NET_F_GUEST_USO4 54 /* Guest can handle USOv4 in. */
#define VIRTIO_NET_F_GUEST_USO6 55 /* Guest can handle USOv6 in. */
@@ -391,5 +392,18 @@ struct virtio_net_ctrl_coal_rx {
};
#define VIRTIO_NET_CTRL_NOTF_COAL_RX_SET 1
+#define VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET 2
+#define VIRTIO_NET_CTRL_NOTF_COAL_VQ_GET 3
+
+struct virtio_net_ctrl_coal {
+ __le32 max_packets;
+ __le32 max_usecs;
+};
+
+struct virtio_net_ctrl_coal_vq {
+ __le16 vqn;
+ __le16 reserved;
+ struct virtio_net_ctrl_coal coal;
+};
#endif /* _UAPI_LINUX_VIRTIO_NET_H */
--
2.39.1
Enable per queue interrupt coalesce feature bit in driver and validate its
dependency with control queue.
Signed-off-by: Gavin Li <[email protected]>
Reviewed-by: Dragos Tatulea <[email protected]>
Reviewed-by: Jiri Pirko <[email protected]>
---
drivers/net/virtio_net.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 333a38e1941f..81bac3597a2b 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -4060,6 +4060,8 @@ static bool virtnet_validate_features(struct virtio_device *vdev)
VIRTNET_FAIL_ON(vdev, VIRTIO_NET_F_HASH_REPORT,
"VIRTIO_NET_F_CTRL_VQ") ||
VIRTNET_FAIL_ON(vdev, VIRTIO_NET_F_NOTF_COAL,
+ "VIRTIO_NET_F_CTRL_VQ") ||
+ VIRTNET_FAIL_ON(vdev, VIRTIO_NET_F_VQ_NOTF_COAL,
"VIRTIO_NET_F_CTRL_VQ"))) {
return false;
}
@@ -4484,6 +4486,7 @@ static struct virtio_device_id id_table[] = {
VIRTIO_NET_F_MTU, VIRTIO_NET_F_CTRL_GUEST_OFFLOADS, \
VIRTIO_NET_F_SPEED_DUPLEX, VIRTIO_NET_F_STANDBY, \
VIRTIO_NET_F_RSS, VIRTIO_NET_F_HASH_REPORT, VIRTIO_NET_F_NOTF_COAL, \
+ VIRTIO_NET_F_VQ_NOTF_COAL, \
VIRTIO_NET_F_GUEST_HDRLEN
static unsigned int features[] = {
--
2.39.1
Extract get/set interrupt coalesce to a function to be reused by global
and per queue config.
Signed-off-by: Gavin Li <[email protected]>
Reviewed-by: Dragos Tatulea <[email protected]>
Reviewed-by: Jiri Pirko <[email protected]>
---
drivers/net/virtio_net.c | 22 +++++++++++++++++-----
1 file changed, 17 insertions(+), 5 deletions(-)
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index dd5fec073a27..802ed21453f5 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -3093,10 +3093,8 @@ static int virtnet_coal_params_supported(struct ethtool_coalesce *ec)
return 0;
}
-static int virtnet_set_coalesce(struct net_device *dev,
- struct ethtool_coalesce *ec,
- struct kernel_ethtool_coalesce *kernel_coal,
- struct netlink_ext_ack *extack)
+static int virtnet_set_coalesce_one(struct net_device *dev,
+ struct ethtool_coalesce *ec)
{
struct virtnet_info *vi = netdev_priv(dev);
int ret, i, napi_weight;
@@ -3127,10 +3125,16 @@ static int virtnet_set_coalesce(struct net_device *dev,
return ret;
}
-static int virtnet_get_coalesce(struct net_device *dev,
+static int virtnet_set_coalesce(struct net_device *dev,
struct ethtool_coalesce *ec,
struct kernel_ethtool_coalesce *kernel_coal,
struct netlink_ext_ack *extack)
+{
+ return virtnet_set_coalesce_one(dev, ec);
+}
+
+static int virtnet_get_coalesce_one(struct net_device *dev,
+ struct ethtool_coalesce *ec)
{
struct virtnet_info *vi = netdev_priv(dev);
@@ -3149,6 +3153,14 @@ static int virtnet_get_coalesce(struct net_device *dev,
return 0;
}
+static int virtnet_get_coalesce(struct net_device *dev,
+ struct ethtool_coalesce *ec,
+ struct kernel_ethtool_coalesce *kernel_coal,
+ struct netlink_ext_ack *extack)
+{
+ return virtnet_get_coalesce_one(dev, ec);
+}
+
static void virtnet_init_settings(struct net_device *dev)
{
struct virtnet_info *vi = netdev_priv(dev);
--
2.39.1
On 7/10/2023 5:25 PM, Xuan Zhuo wrote:
> External email: Use caution opening links or attachments
>
>
> On Mon, 10 Jul 2023 12:20:01 +0300, Gavin Li <[email protected]> wrote:
>
> As far as I know, Heng Qi does that. I'm not sure, it's the same piece.
>
> cc @Heng Qi
>
> Thanks.
Do you mean the one below?
https://lists.oasis-open.org/archives/virtio-dev/202303/msg00415.html
The code is to implement what it described and I don't have better
words. So, I copied the text from Heng Qi's.
>
>
>> Currently, coalescing parameters are grouped for all transmit and receive
>> virtqueues. This patch series add support to set or get the parameters for
>> a specified virtqueue.
>>
>> When the traffic between virtqueues is unbalanced, for example, one virtqueue
>> is busy and another virtqueue is idle, then it will be very useful to
>> control coalescing parameters at the virtqueue granularity.
>>
>> Example command:
>> $ ethtool -Q eth5 queue_mask 0x1 --coalesce tx-packets 10
>> Would set max_packets=10 to VQ 1.
>> $ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10
>> Would set max_packets=10 to VQ 0.
>> $ ethtool -Q eth5 queue_mask 0x1 --show-coalesce
>> Queue: 0
>> Adaptive RX: off TX: off
>> stats-block-usecs: 0
>> sample-interval: 0
>> pkt-rate-low: 0
>> pkt-rate-high: 0
>>
>> rx-usecs: 222
>> rx-frames: 0
>> rx-usecs-irq: 0
>> rx-frames-irq: 256
>>
>> tx-usecs: 222
>> tx-frames: 0
>> tx-usecs-irq: 0
>> tx-frames-irq: 256
>>
>> rx-usecs-low: 0
>> rx-frame-low: 0
>> tx-usecs-low: 0
>> tx-frame-low: 0
>>
>> rx-usecs-high: 0
>> rx-frame-high: 0
>> tx-usecs-high: 0
>> tx-frame-high: 0
>>
>> In this patch series:
>> Patch-1: Extract interrupt coalescing settings to a structure.
>> Patch-2: Extract get/set interrupt coalesce to a function.
>> Patch-3: Support per queue interrupt coalesce command.
>> Patch-4: Enable per queue interrupt coalesce feature.
>>
>> Gavin Li (4):
>> virtio_net: extract interrupt coalescing settings to a structure
>> virtio_net: extract get/set interrupt coalesce to a function
>> virtio_net: support per queue interrupt coalesce command
>> virtio_net: enable per queue interrupt coalesce feature
>>
>> drivers/net/virtio_net.c | 169 ++++++++++++++++++++++++++------
>> include/uapi/linux/virtio_net.h | 14 +++
>> 2 files changed, 154 insertions(+), 29 deletions(-)
>>
>> --
>> 2.39.1
>>
On Tue, 11 Jul 2023 10:34:33 +0800, Gavin Li <[email protected]> wrote:
>
> On 7/10/2023 5:25 PM, Xuan Zhuo wrote:
> > External email: Use caution opening links or attachments
> >
> >
> > On Mon, 10 Jul 2023 12:20:01 +0300, Gavin Li <[email protected]> wrote:
> >
> > As far as I know, Heng Qi does that. I'm not sure, it's the same piece.
> >
> > cc @Heng Qi
> >
> > Thanks.
> Do you mean the one below?
YES
>
> https://lists.oasis-open.org/archives/virtio-dev/202303/msg00415.html
>
> The code is to implement what it described and I don't have better
> words. So, I copied the text from Heng Qi's.
Maybe he wrote code for it. I think you should ask him about his plans first.
Thanks.
>
> >
> >
> >> Currently, coalescing parameters are grouped for all transmit and receive
> >> virtqueues. This patch series add support to set or get the parameters for
> >> a specified virtqueue.
> >>
> >> When the traffic between virtqueues is unbalanced, for example, one virtqueue
> >> is busy and another virtqueue is idle, then it will be very useful to
> >> control coalescing parameters at the virtqueue granularity.
> >>
> >> Example command:
> >> $ ethtool -Q eth5 queue_mask 0x1 --coalesce tx-packets 10
> >> Would set max_packets=10 to VQ 1.
> >> $ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10
> >> Would set max_packets=10 to VQ 0.
> >> $ ethtool -Q eth5 queue_mask 0x1 --show-coalesce
> >> Queue: 0
> >> Adaptive RX: off TX: off
> >> stats-block-usecs: 0
> >> sample-interval: 0
> >> pkt-rate-low: 0
> >> pkt-rate-high: 0
> >>
> >> rx-usecs: 222
> >> rx-frames: 0
> >> rx-usecs-irq: 0
> >> rx-frames-irq: 256
> >>
> >> tx-usecs: 222
> >> tx-frames: 0
> >> tx-usecs-irq: 0
> >> tx-frames-irq: 256
> >>
> >> rx-usecs-low: 0
> >> rx-frame-low: 0
> >> tx-usecs-low: 0
> >> tx-frame-low: 0
> >>
> >> rx-usecs-high: 0
> >> rx-frame-high: 0
> >> tx-usecs-high: 0
> >> tx-frame-high: 0
> >>
> >> In this patch series:
> >> Patch-1: Extract interrupt coalescing settings to a structure.
> >> Patch-2: Extract get/set interrupt coalesce to a function.
> >> Patch-3: Support per queue interrupt coalesce command.
> >> Patch-4: Enable per queue interrupt coalesce feature.
> >>
> >> Gavin Li (4):
> >> virtio_net: extract interrupt coalescing settings to a structure
> >> virtio_net: extract get/set interrupt coalesce to a function
> >> virtio_net: support per queue interrupt coalesce command
> >> virtio_net: enable per queue interrupt coalesce feature
> >>
> >> drivers/net/virtio_net.c | 169 ++++++++++++++++++++++++++------
> >> include/uapi/linux/virtio_net.h | 14 +++
> >> 2 files changed, 154 insertions(+), 29 deletions(-)
> >>
> >> --
> >> 2.39.1
> >>
在 2023/7/10 下午5:20, Gavin Li 写道:
> Currently, coalescing parameters are grouped for all transmit and receive
> virtqueues. This patch series add support to set or get the parameters for
> a specified virtqueue.
>
> When the traffic between virtqueues is unbalanced, for example, one virtqueue
> is busy and another virtqueue is idle, then it will be very useful to
> control coalescing parameters at the virtqueue granularity.
We definitely did the same thing, and I'm waiting for our hardware
implementation to be ready to
push the ethtool + netdim implementation.
Since this commit log is completely copied from the implementation of
the virtio spec, consulting
each other's scheduling plan in advance can be friendly to avoid us
doing the same thing and consuming an extra effort.
Thanks.
>
> Example command:
> $ ethtool -Q eth5 queue_mask 0x1 --coalesce tx-packets 10
> Would set max_packets=10 to VQ 1.
> $ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10
> Would set max_packets=10 to VQ 0.
> $ ethtool -Q eth5 queue_mask 0x1 --show-coalesce
> Queue: 0
> Adaptive RX: off TX: off
> stats-block-usecs: 0
> sample-interval: 0
> pkt-rate-low: 0
> pkt-rate-high: 0
>
> rx-usecs: 222
> rx-frames: 0
> rx-usecs-irq: 0
> rx-frames-irq: 256
>
> tx-usecs: 222
> tx-frames: 0
> tx-usecs-irq: 0
> tx-frames-irq: 256
>
> rx-usecs-low: 0
> rx-frame-low: 0
> tx-usecs-low: 0
> tx-frame-low: 0
>
> rx-usecs-high: 0
> rx-frame-high: 0
> tx-usecs-high: 0
> tx-frame-high: 0
>
> In this patch series:
> Patch-1: Extract interrupt coalescing settings to a structure.
> Patch-2: Extract get/set interrupt coalesce to a function.
> Patch-3: Support per queue interrupt coalesce command.
> Patch-4: Enable per queue interrupt coalesce feature.
>
> Gavin Li (4):
> virtio_net: extract interrupt coalescing settings to a structure
> virtio_net: extract get/set interrupt coalesce to a function
> virtio_net: support per queue interrupt coalesce command
> virtio_net: enable per queue interrupt coalesce feature
>
> drivers/net/virtio_net.c | 169 ++++++++++++++++++++++++++------
> include/uapi/linux/virtio_net.h | 14 +++
> 2 files changed, 154 insertions(+), 29 deletions(-)
>
On 7/12/2023 5:39 PM, Xuan Zhuo wrote:
> On Mon, 10 Jul 2023 12:20:04 +0300, Gavin Li <[email protected]> wrote:
>> Add interrupt_coalesce config in send_queue and receive_queue to cache user
>> config.
>>
>> Send per virtqueue interrupt moderation config to underline device in order
>> to have more efficient interrupt moderation and cpu utilization of guest
>> VM.
>>
>> Signed-off-by: Gavin Li <[email protected]>
>> Reviewed-by: Dragos Tatulea <[email protected]>
>> Reviewed-by: Jiri Pirko <[email protected]>
>> ---
>> drivers/net/virtio_net.c | 117 ++++++++++++++++++++++++++++----
>> include/uapi/linux/virtio_net.h | 14 ++++
>> 2 files changed, 119 insertions(+), 12 deletions(-)
>>
>> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
>> index 802ed21453f5..333a38e1941f 100644
>> --- a/drivers/net/virtio_net.c
>> +++ b/drivers/net/virtio_net.c
>> @@ -144,6 +144,8 @@ struct send_queue {
>>
>> struct virtnet_sq_stats stats;
>>
>> + struct virtnet_interrupt_coalesce intr_coal;
>> +
>> struct napi_struct napi;
>>
>> /* Record whether sq is in reset state. */
>> @@ -161,6 +163,8 @@ struct receive_queue {
>>
>> struct virtnet_rq_stats stats;
>>
>> + struct virtnet_interrupt_coalesce intr_coal;
>> +
>> /* Chain pages by the private ptr. */
>> struct page *pages;
>>
>> @@ -3078,6 +3082,56 @@ static int virtnet_send_notf_coal_cmds(struct virtnet_info *vi,
>> return 0;
>> }
>>
>> +static int virtnet_send_ctrl_coal_vq_cmd(struct virtnet_info *vi,
>> + u16 vqn, u32 max_usecs, u32 max_packets)
>> +{
>> + struct virtio_net_ctrl_coal_vq coal_vq = {};
> We should alloc this on the heap.
>
> Thanks.
ACK
>
>
>> + struct scatterlist sgs;
>> +
>> + coal_vq.vqn = cpu_to_le16(vqn);
>> + coal_vq.coal.max_usecs = cpu_to_le32(max_usecs);
>> + coal_vq.coal.max_packets = cpu_to_le32(max_packets);
>> + sg_init_one(&sgs, &coal_vq, sizeof(coal_vq));
>> +
>> + if (!virtnet_send_command(vi, VIRTIO_NET_CTRL_NOTF_COAL,
>> + VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET,
>> + &sgs))
>> + return -EINVAL;
>> +
>> + return 0;
>> +}
>> +
>> +static int virtnet_send_notf_coal_vq_cmds(struct virtnet_info *vi,
>> + struct ethtool_coalesce *ec,
>> + u16 queue)
>> +{
>> + int err;
>> +
>> + if (ec->rx_coalesce_usecs || ec->rx_max_coalesced_frames) {
>> + err = virtnet_send_ctrl_coal_vq_cmd(vi, rxq2vq(queue),
>> + ec->rx_coalesce_usecs,
>> + ec->rx_max_coalesced_frames);
>> + if (err)
>> + return err;
>> + /* Save parameters */
>> + vi->rq[queue].intr_coal.max_usecs = ec->rx_coalesce_usecs;
>> + vi->rq[queue].intr_coal.max_packets = ec->rx_max_coalesced_frames;
>> + }
>> +
>> + if (ec->tx_coalesce_usecs || ec->tx_max_coalesced_frames) {
>> + err = virtnet_send_ctrl_coal_vq_cmd(vi, txq2vq(queue),
>> + ec->tx_coalesce_usecs,
>> + ec->tx_max_coalesced_frames);
>> + if (err)
>> + return err;
>> + /* Save parameters */
>> + vi->sq[queue].intr_coal.max_usecs = ec->tx_coalesce_usecs;
>> + vi->sq[queue].intr_coal.max_packets = ec->tx_max_coalesced_frames;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> static int virtnet_coal_params_supported(struct ethtool_coalesce *ec)
>> {
>> /* usecs coalescing is supported only if VIRTIO_NET_F_NOTF_COAL
>> @@ -3094,23 +3148,36 @@ static int virtnet_coal_params_supported(struct ethtool_coalesce *ec)
>> }
>>
>> static int virtnet_set_coalesce_one(struct net_device *dev,
>> - struct ethtool_coalesce *ec)
>> + struct ethtool_coalesce *ec,
>> + bool per_queue,
>> + u32 queue)
>> {
>> struct virtnet_info *vi = netdev_priv(dev);
>> - int ret, i, napi_weight;
>> + int queue_count = per_queue ? 1 : vi->max_queue_pairs;
>> + int queue_number = per_queue ? queue : 0;
>> bool update_napi = false;
>> + int ret, i, napi_weight;
>> +
>> + if (queue >= vi->max_queue_pairs)
>> + return -EINVAL;
>>
>> /* Can't change NAPI weight if the link is up */
>> napi_weight = ec->tx_max_coalesced_frames ? NAPI_POLL_WEIGHT : 0;
>> - if (napi_weight ^ vi->sq[0].napi.weight) {
>> - if (dev->flags & IFF_UP)
>> - return -EBUSY;
>> - else
>> + for (i = queue_number; i < queue_count; i++) {
>> + if (napi_weight ^ vi->sq[i].napi.weight) {
>> + if (dev->flags & IFF_UP)
>> + return -EBUSY;
>> +
>> update_napi = true;
>> + queue_number = i;
>> + break;
>> + }
>> }
>>
>> - if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL))
>> + if (!per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL))
>> ret = virtnet_send_notf_coal_cmds(vi, ec);
>> + else if (per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL))
>> + ret = virtnet_send_notf_coal_vq_cmds(vi, ec, queue);
>> else
>> ret = virtnet_coal_params_supported(ec);
>>
>> @@ -3118,7 +3185,7 @@ static int virtnet_set_coalesce_one(struct net_device *dev,
>> return ret;
>>
>> if (update_napi) {
>> - for (i = 0; i < vi->max_queue_pairs; i++)
>> + for (i = queue_number; i < queue_count; i++)
>> vi->sq[i].napi.weight = napi_weight;
>> }
>>
>> @@ -3130,19 +3197,29 @@ static int virtnet_set_coalesce(struct net_device *dev,
>> struct kernel_ethtool_coalesce *kernel_coal,
>> struct netlink_ext_ack *extack)
>> {
>> - return virtnet_set_coalesce_one(dev, ec);
>> + return virtnet_set_coalesce_one(dev, ec, false, 0);
>> }
>>
>> static int virtnet_get_coalesce_one(struct net_device *dev,
>> - struct ethtool_coalesce *ec)
>> + struct ethtool_coalesce *ec,
>> + bool per_queue,
>> + u32 queue)
>> {
>> struct virtnet_info *vi = netdev_priv(dev);
>>
>> - if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL)) {
>> + if (queue >= vi->max_queue_pairs)
>> + return -EINVAL;
>> +
>> + if (!per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL)) {
>> ec->rx_coalesce_usecs = vi->intr_coal_rx.max_usecs;
>> ec->tx_coalesce_usecs = vi->intr_coal_tx.max_usecs;
>> ec->tx_max_coalesced_frames = vi->intr_coal_tx.max_packets;
>> ec->rx_max_coalesced_frames = vi->intr_coal_rx.max_packets;
>> + } else if (per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL)) {
>> + ec->rx_coalesce_usecs = vi->rq[queue].intr_coal.max_usecs;
>> + ec->tx_coalesce_usecs = vi->sq[queue].intr_coal.max_usecs;
>> + ec->tx_max_coalesced_frames = vi->sq[queue].intr_coal.max_packets;
>> + ec->rx_max_coalesced_frames = vi->rq[queue].intr_coal.max_packets;
>> } else {
>> ec->rx_max_coalesced_frames = 1;
>>
>> @@ -3158,7 +3235,21 @@ static int virtnet_get_coalesce(struct net_device *dev,
>> struct kernel_ethtool_coalesce *kernel_coal,
>> struct netlink_ext_ack *extack)
>> {
>> - return virtnet_get_coalesce_one(dev, ec);
>> + return virtnet_get_coalesce_one(dev, ec, false, 0);
>> +}
>> +
>> +static int virtnet_set_per_queue_coalesce(struct net_device *dev,
>> + u32 queue,
>> + struct ethtool_coalesce *ec)
>> +{
>> + return virtnet_set_coalesce_one(dev, ec, true, queue);
>> +}
>> +
>> +static int virtnet_get_per_queue_coalesce(struct net_device *dev,
>> + u32 queue,
>> + struct ethtool_coalesce *ec)
>> +{
>> + return virtnet_get_coalesce_one(dev, ec, true, queue);
>> }
>>
>> static void virtnet_init_settings(struct net_device *dev)
>> @@ -3291,6 +3382,8 @@ static const struct ethtool_ops virtnet_ethtool_ops = {
>> .set_link_ksettings = virtnet_set_link_ksettings,
>> .set_coalesce = virtnet_set_coalesce,
>> .get_coalesce = virtnet_get_coalesce,
>> + .set_per_queue_coalesce = virtnet_set_per_queue_coalesce,
>> + .get_per_queue_coalesce = virtnet_get_per_queue_coalesce,
>> .get_rxfh_key_size = virtnet_get_rxfh_key_size,
>> .get_rxfh_indir_size = virtnet_get_rxfh_indir_size,
>> .get_rxfh = virtnet_get_rxfh,
>> diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h
>> index 12c1c9699935..cc65ef0f3c3e 100644
>> --- a/include/uapi/linux/virtio_net.h
>> +++ b/include/uapi/linux/virtio_net.h
>> @@ -56,6 +56,7 @@
>> #define VIRTIO_NET_F_MQ 22 /* Device supports Receive Flow
>> * Steering */
>> #define VIRTIO_NET_F_CTRL_MAC_ADDR 23 /* Set MAC address */
>> +#define VIRTIO_NET_F_VQ_NOTF_COAL 52 /* Device supports virtqueue notification coalescing */
>> #define VIRTIO_NET_F_NOTF_COAL 53 /* Device supports notifications coalescing */
>> #define VIRTIO_NET_F_GUEST_USO4 54 /* Guest can handle USOv4 in. */
>> #define VIRTIO_NET_F_GUEST_USO6 55 /* Guest can handle USOv6 in. */
>> @@ -391,5 +392,18 @@ struct virtio_net_ctrl_coal_rx {
>> };
>>
>> #define VIRTIO_NET_CTRL_NOTF_COAL_RX_SET 1
>> +#define VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET 2
>> +#define VIRTIO_NET_CTRL_NOTF_COAL_VQ_GET 3
>> +
>> +struct virtio_net_ctrl_coal {
>> + __le32 max_packets;
>> + __le32 max_usecs;
>> +};
>> +
>> +struct virtio_net_ctrl_coal_vq {
>> + __le16 vqn;
>> + __le16 reserved;
>> + struct virtio_net_ctrl_coal coal;
>> +};
>>
>> #endif /* _UAPI_LINUX_VIRTIO_NET_H */
>> --
>> 2.39.1
>>
On Mon, 10 Jul 2023 12:20:04 +0300, Gavin Li <[email protected]> wrote:
> Add interrupt_coalesce config in send_queue and receive_queue to cache user
> config.
>
> Send per virtqueue interrupt moderation config to underline device in order
> to have more efficient interrupt moderation and cpu utilization of guest
> VM.
>
> Signed-off-by: Gavin Li <[email protected]>
> Reviewed-by: Dragos Tatulea <[email protected]>
> Reviewed-by: Jiri Pirko <[email protected]>
> ---
> drivers/net/virtio_net.c | 117 ++++++++++++++++++++++++++++----
> include/uapi/linux/virtio_net.h | 14 ++++
> 2 files changed, 119 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index 802ed21453f5..333a38e1941f 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -144,6 +144,8 @@ struct send_queue {
>
> struct virtnet_sq_stats stats;
>
> + struct virtnet_interrupt_coalesce intr_coal;
> +
> struct napi_struct napi;
>
> /* Record whether sq is in reset state. */
> @@ -161,6 +163,8 @@ struct receive_queue {
>
> struct virtnet_rq_stats stats;
>
> + struct virtnet_interrupt_coalesce intr_coal;
> +
> /* Chain pages by the private ptr. */
> struct page *pages;
>
> @@ -3078,6 +3082,56 @@ static int virtnet_send_notf_coal_cmds(struct virtnet_info *vi,
> return 0;
> }
>
> +static int virtnet_send_ctrl_coal_vq_cmd(struct virtnet_info *vi,
> + u16 vqn, u32 max_usecs, u32 max_packets)
> +{
> + struct virtio_net_ctrl_coal_vq coal_vq = {};
We should alloc this on the heap.
Thanks.
> + struct scatterlist sgs;
> +
> + coal_vq.vqn = cpu_to_le16(vqn);
> + coal_vq.coal.max_usecs = cpu_to_le32(max_usecs);
> + coal_vq.coal.max_packets = cpu_to_le32(max_packets);
> + sg_init_one(&sgs, &coal_vq, sizeof(coal_vq));
> +
> + if (!virtnet_send_command(vi, VIRTIO_NET_CTRL_NOTF_COAL,
> + VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET,
> + &sgs))
> + return -EINVAL;
> +
> + return 0;
> +}
> +
> +static int virtnet_send_notf_coal_vq_cmds(struct virtnet_info *vi,
> + struct ethtool_coalesce *ec,
> + u16 queue)
> +{
> + int err;
> +
> + if (ec->rx_coalesce_usecs || ec->rx_max_coalesced_frames) {
> + err = virtnet_send_ctrl_coal_vq_cmd(vi, rxq2vq(queue),
> + ec->rx_coalesce_usecs,
> + ec->rx_max_coalesced_frames);
> + if (err)
> + return err;
> + /* Save parameters */
> + vi->rq[queue].intr_coal.max_usecs = ec->rx_coalesce_usecs;
> + vi->rq[queue].intr_coal.max_packets = ec->rx_max_coalesced_frames;
> + }
> +
> + if (ec->tx_coalesce_usecs || ec->tx_max_coalesced_frames) {
> + err = virtnet_send_ctrl_coal_vq_cmd(vi, txq2vq(queue),
> + ec->tx_coalesce_usecs,
> + ec->tx_max_coalesced_frames);
> + if (err)
> + return err;
> + /* Save parameters */
> + vi->sq[queue].intr_coal.max_usecs = ec->tx_coalesce_usecs;
> + vi->sq[queue].intr_coal.max_packets = ec->tx_max_coalesced_frames;
> + }
> +
> + return 0;
> +}
> +
> static int virtnet_coal_params_supported(struct ethtool_coalesce *ec)
> {
> /* usecs coalescing is supported only if VIRTIO_NET_F_NOTF_COAL
> @@ -3094,23 +3148,36 @@ static int virtnet_coal_params_supported(struct ethtool_coalesce *ec)
> }
>
> static int virtnet_set_coalesce_one(struct net_device *dev,
> - struct ethtool_coalesce *ec)
> + struct ethtool_coalesce *ec,
> + bool per_queue,
> + u32 queue)
> {
> struct virtnet_info *vi = netdev_priv(dev);
> - int ret, i, napi_weight;
> + int queue_count = per_queue ? 1 : vi->max_queue_pairs;
> + int queue_number = per_queue ? queue : 0;
> bool update_napi = false;
> + int ret, i, napi_weight;
> +
> + if (queue >= vi->max_queue_pairs)
> + return -EINVAL;
>
> /* Can't change NAPI weight if the link is up */
> napi_weight = ec->tx_max_coalesced_frames ? NAPI_POLL_WEIGHT : 0;
> - if (napi_weight ^ vi->sq[0].napi.weight) {
> - if (dev->flags & IFF_UP)
> - return -EBUSY;
> - else
> + for (i = queue_number; i < queue_count; i++) {
> + if (napi_weight ^ vi->sq[i].napi.weight) {
> + if (dev->flags & IFF_UP)
> + return -EBUSY;
> +
> update_napi = true;
> + queue_number = i;
> + break;
> + }
> }
>
> - if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL))
> + if (!per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL))
> ret = virtnet_send_notf_coal_cmds(vi, ec);
> + else if (per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL))
> + ret = virtnet_send_notf_coal_vq_cmds(vi, ec, queue);
> else
> ret = virtnet_coal_params_supported(ec);
>
> @@ -3118,7 +3185,7 @@ static int virtnet_set_coalesce_one(struct net_device *dev,
> return ret;
>
> if (update_napi) {
> - for (i = 0; i < vi->max_queue_pairs; i++)
> + for (i = queue_number; i < queue_count; i++)
> vi->sq[i].napi.weight = napi_weight;
> }
>
> @@ -3130,19 +3197,29 @@ static int virtnet_set_coalesce(struct net_device *dev,
> struct kernel_ethtool_coalesce *kernel_coal,
> struct netlink_ext_ack *extack)
> {
> - return virtnet_set_coalesce_one(dev, ec);
> + return virtnet_set_coalesce_one(dev, ec, false, 0);
> }
>
> static int virtnet_get_coalesce_one(struct net_device *dev,
> - struct ethtool_coalesce *ec)
> + struct ethtool_coalesce *ec,
> + bool per_queue,
> + u32 queue)
> {
> struct virtnet_info *vi = netdev_priv(dev);
>
> - if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL)) {
> + if (queue >= vi->max_queue_pairs)
> + return -EINVAL;
> +
> + if (!per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_NOTF_COAL)) {
> ec->rx_coalesce_usecs = vi->intr_coal_rx.max_usecs;
> ec->tx_coalesce_usecs = vi->intr_coal_tx.max_usecs;
> ec->tx_max_coalesced_frames = vi->intr_coal_tx.max_packets;
> ec->rx_max_coalesced_frames = vi->intr_coal_rx.max_packets;
> + } else if (per_queue && virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL)) {
> + ec->rx_coalesce_usecs = vi->rq[queue].intr_coal.max_usecs;
> + ec->tx_coalesce_usecs = vi->sq[queue].intr_coal.max_usecs;
> + ec->tx_max_coalesced_frames = vi->sq[queue].intr_coal.max_packets;
> + ec->rx_max_coalesced_frames = vi->rq[queue].intr_coal.max_packets;
> } else {
> ec->rx_max_coalesced_frames = 1;
>
> @@ -3158,7 +3235,21 @@ static int virtnet_get_coalesce(struct net_device *dev,
> struct kernel_ethtool_coalesce *kernel_coal,
> struct netlink_ext_ack *extack)
> {
> - return virtnet_get_coalesce_one(dev, ec);
> + return virtnet_get_coalesce_one(dev, ec, false, 0);
> +}
> +
> +static int virtnet_set_per_queue_coalesce(struct net_device *dev,
> + u32 queue,
> + struct ethtool_coalesce *ec)
> +{
> + return virtnet_set_coalesce_one(dev, ec, true, queue);
> +}
> +
> +static int virtnet_get_per_queue_coalesce(struct net_device *dev,
> + u32 queue,
> + struct ethtool_coalesce *ec)
> +{
> + return virtnet_get_coalesce_one(dev, ec, true, queue);
> }
>
> static void virtnet_init_settings(struct net_device *dev)
> @@ -3291,6 +3382,8 @@ static const struct ethtool_ops virtnet_ethtool_ops = {
> .set_link_ksettings = virtnet_set_link_ksettings,
> .set_coalesce = virtnet_set_coalesce,
> .get_coalesce = virtnet_get_coalesce,
> + .set_per_queue_coalesce = virtnet_set_per_queue_coalesce,
> + .get_per_queue_coalesce = virtnet_get_per_queue_coalesce,
> .get_rxfh_key_size = virtnet_get_rxfh_key_size,
> .get_rxfh_indir_size = virtnet_get_rxfh_indir_size,
> .get_rxfh = virtnet_get_rxfh,
> diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h
> index 12c1c9699935..cc65ef0f3c3e 100644
> --- a/include/uapi/linux/virtio_net.h
> +++ b/include/uapi/linux/virtio_net.h
> @@ -56,6 +56,7 @@
> #define VIRTIO_NET_F_MQ 22 /* Device supports Receive Flow
> * Steering */
> #define VIRTIO_NET_F_CTRL_MAC_ADDR 23 /* Set MAC address */
> +#define VIRTIO_NET_F_VQ_NOTF_COAL 52 /* Device supports virtqueue notification coalescing */
> #define VIRTIO_NET_F_NOTF_COAL 53 /* Device supports notifications coalescing */
> #define VIRTIO_NET_F_GUEST_USO4 54 /* Guest can handle USOv4 in. */
> #define VIRTIO_NET_F_GUEST_USO6 55 /* Guest can handle USOv6 in. */
> @@ -391,5 +392,18 @@ struct virtio_net_ctrl_coal_rx {
> };
>
> #define VIRTIO_NET_CTRL_NOTF_COAL_RX_SET 1
> +#define VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET 2
> +#define VIRTIO_NET_CTRL_NOTF_COAL_VQ_GET 3
> +
> +struct virtio_net_ctrl_coal {
> + __le32 max_packets;
> + __le32 max_usecs;
> +};
> +
> +struct virtio_net_ctrl_coal_vq {
> + __le16 vqn;
> + __le16 reserved;
> + struct virtio_net_ctrl_coal coal;
> +};
>
> #endif /* _UAPI_LINUX_VIRTIO_NET_H */
> --
> 2.39.1
>
On 7/12/2023 4:34 PM, Heng Qi wrote:
>
>
> 在 2023/7/10 下午5:20, Gavin Li 写道:
>> Currently, coalescing parameters are grouped for all transmit and receive
>> virtqueues. This patch series add support to set or get the parameters
>> for
>> a specified virtqueue.
>>
>> When the traffic between virtqueues is unbalanced, for example, one
>> virtqueue
>> is busy and another virtqueue is idle, then it will be very useful to
>> control coalescing parameters at the virtqueue granularity.
>
>
> We definitely did the same thing, and I'm waiting for our hardware
> implementation to be ready to
> push the ethtool + netdim implementation.
>
> Since this commit log is completely copied from the implementation of
> the virtio spec, consulting
> each other's scheduling plan in advance can be friendly to avoid us
> doing the same thing and consuming an extra effort.
>
> Thanks.
>
I didn't know about the effort you made and the progress of it.
Since the code has been done and tested, I'll continue submission.
>>
>> Example command:
>> $ ethtool -Q eth5 queue_mask 0x1 --coalesce tx-packets 10
>> Would set max_packets=10 to VQ 1.
>> $ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10
>> Would set max_packets=10 to VQ 0.
>> $ ethtool -Q eth5 queue_mask 0x1 --show-coalesce
>> Queue: 0
>> Adaptive RX: off TX: off
>> stats-block-usecs: 0
>> sample-interval: 0
>> pkt-rate-low: 0
>> pkt-rate-high: 0
>>
>> rx-usecs: 222
>> rx-frames: 0
>> rx-usecs-irq: 0
>> rx-frames-irq: 256
>>
>> tx-usecs: 222
>> tx-frames: 0
>> tx-usecs-irq: 0
>> tx-frames-irq: 256
>>
>> rx-usecs-low: 0
>> rx-frame-low: 0
>> tx-usecs-low: 0
>> tx-frame-low: 0
>>
>> rx-usecs-high: 0
>> rx-frame-high: 0
>> tx-usecs-high: 0
>> tx-frame-high: 0
>>
>> In this patch series:
>> Patch-1: Extract interrupt coalescing settings to a structure.
>> Patch-2: Extract get/set interrupt coalesce to a function.
>> Patch-3: Support per queue interrupt coalesce command.
>> Patch-4: Enable per queue interrupt coalesce feature.
>>
>> Gavin Li (4):
>> virtio_net: extract interrupt coalescing settings to a structure
>> virtio_net: extract get/set interrupt coalesce to a function
>> virtio_net: support per queue interrupt coalesce command
>> virtio_net: enable per queue interrupt coalesce feature
>>
>> drivers/net/virtio_net.c | 169 ++++++++++++++++++++++++++------
>> include/uapi/linux/virtio_net.h | 14 +++
>> 2 files changed, 154 insertions(+), 29 deletions(-)
>>
>
On Mon, Jul 10, 2023 at 12:20:01PM +0300, Gavin Li wrote:
> Currently, coalescing parameters are grouped for all transmit and receive
> virtqueues. This patch series add support to set or get the parameters for
> a specified virtqueue.
>
> When the traffic between virtqueues is unbalanced, for example, one virtqueue
> is busy and another virtqueue is idle, then it will be very useful to
> control coalescing parameters at the virtqueue granularity.
series:
Acked-by: Michael S. Tsirkin <[email protected]>
> Example command:
> $ ethtool -Q eth5 queue_mask 0x1 --coalesce tx-packets 10
> Would set max_packets=10 to VQ 1.
> $ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10
> Would set max_packets=10 to VQ 0.
> $ ethtool -Q eth5 queue_mask 0x1 --show-coalesce
> Queue: 0
> Adaptive RX: off TX: off
> stats-block-usecs: 0
> sample-interval: 0
> pkt-rate-low: 0
> pkt-rate-high: 0
>
> rx-usecs: 222
> rx-frames: 0
> rx-usecs-irq: 0
> rx-frames-irq: 256
>
> tx-usecs: 222
> tx-frames: 0
> tx-usecs-irq: 0
> tx-frames-irq: 256
>
> rx-usecs-low: 0
> rx-frame-low: 0
> tx-usecs-low: 0
> tx-frame-low: 0
>
> rx-usecs-high: 0
> rx-frame-high: 0
> tx-usecs-high: 0
> tx-frame-high: 0
>
> In this patch series:
> Patch-1: Extract interrupt coalescing settings to a structure.
> Patch-2: Extract get/set interrupt coalesce to a function.
> Patch-3: Support per queue interrupt coalesce command.
> Patch-4: Enable per queue interrupt coalesce feature.
>
> Gavin Li (4):
> virtio_net: extract interrupt coalescing settings to a structure
> virtio_net: extract get/set interrupt coalesce to a function
> virtio_net: support per queue interrupt coalesce command
> virtio_net: enable per queue interrupt coalesce feature
>
> drivers/net/virtio_net.c | 169 ++++++++++++++++++++++++++------
> include/uapi/linux/virtio_net.h | 14 +++
> 2 files changed, 154 insertions(+), 29 deletions(-)
>
> --
> 2.39.1
On Thu, 13 Jul 2023 07:40:12 -0400, "Michael S. Tsirkin" <[email protected]> wrote:
> On Mon, Jul 10, 2023 at 12:20:01PM +0300, Gavin Li wrote:
> > Currently, coalescing parameters are grouped for all transmit and receive
> > virtqueues. This patch series add support to set or get the parameters for
> > a specified virtqueue.
> >
> > When the traffic between virtqueues is unbalanced, for example, one virtqueue
> > is busy and another virtqueue is idle, then it will be very useful to
> > control coalescing parameters at the virtqueue granularity.
>
> series:
>
> Acked-by: Michael S. Tsirkin <[email protected]>
Why?
This series has the bug I reported.
Are you thinking that is ok? Or this is not a bug?
Thanks.
>
>
>
> > Example command:
> > $ ethtool -Q eth5 queue_mask 0x1 --coalesce tx-packets 10
> > Would set max_packets=10 to VQ 1.
> > $ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10
> > Would set max_packets=10 to VQ 0.
> > $ ethtool -Q eth5 queue_mask 0x1 --show-coalesce
> > Queue: 0
> > Adaptive RX: off TX: off
> > stats-block-usecs: 0
> > sample-interval: 0
> > pkt-rate-low: 0
> > pkt-rate-high: 0
> >
> > rx-usecs: 222
> > rx-frames: 0
> > rx-usecs-irq: 0
> > rx-frames-irq: 256
> >
> > tx-usecs: 222
> > tx-frames: 0
> > tx-usecs-irq: 0
> > tx-frames-irq: 256
> >
> > rx-usecs-low: 0
> > rx-frame-low: 0
> > tx-usecs-low: 0
> > tx-frame-low: 0
> >
> > rx-usecs-high: 0
> > rx-frame-high: 0
> > tx-usecs-high: 0
> > tx-frame-high: 0
> >
> > In this patch series:
> > Patch-1: Extract interrupt coalescing settings to a structure.
> > Patch-2: Extract get/set interrupt coalesce to a function.
> > Patch-3: Support per queue interrupt coalesce command.
> > Patch-4: Enable per queue interrupt coalesce feature.
> >
> > Gavin Li (4):
> > virtio_net: extract interrupt coalescing settings to a structure
> > virtio_net: extract get/set interrupt coalesce to a function
> > virtio_net: support per queue interrupt coalesce command
> > virtio_net: enable per queue interrupt coalesce feature
> >
> > drivers/net/virtio_net.c | 169 ++++++++++++++++++++++++++------
> > include/uapi/linux/virtio_net.h | 14 +++
> > 2 files changed, 154 insertions(+), 29 deletions(-)
> >
> > --
> > 2.39.1
>