virtualization.lists.linux-foundation.org archive mirror
 help / color / mirror / Atom feed
* [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel
@ 2024-04-15 13:38 Heng Qi
  2024-04-15 13:38 ` [PATCH RESEND net-next v7 1/4] linux/dim: move useful macros to .h file Heng Qi
                   ` (4 more replies)
  0 siblings, 5 replies; 7+ messages in thread
From: Heng Qi @ 2024-04-15 13:38 UTC (permalink / raw)
  To: netdev, virtualization
  Cc: Jakub Kicinski, David S . Miller, Eric Dumazet, Paolo Abeni,
	Jason Wang, Michael S . Tsirkin, Brett Creeley, Ratheesh Kannoth,
	Alexander Lobakin, Xuan Zhuo

The NetDIM library provides excellent acceleration for many modern
network cards. However, the default profiles of DIM limits its maximum
capabilities for different NICs, so providing a way which the NIC can
be custom configured is necessary.

Currently, interaction with the driver is still based on the commonly
used "ethtool -C".

Since the profile now exists in netdevice, adding a function similar
to net_dim_get_rx_moderation_dev() with netdevice as argument is
nice, but this would be better along with cleaning up the rest of
the drivers, which we can get to very soon after this set.

Please review, thank you very much!

Changelog
=====
v6->v7:
  - A new wrapper struct pointer is used in struct net_device.
  - Add IS_ENABLED(CONFIG_DIMLIB) to avoid compiler warnings.
  - Profile fields changed from u16 to u32.

v5->v6:
  - Place the profile in netdevice to bypass the driver.
    The interaction code of ethtool <-> kernel has not changed at all,
    only the interaction part of kernel <-> driver has changed.

v4->v5:
  - Update some snippets from Kuba, Thanks.

v3->v4:
  - Some tiny updates and patch 1 only add a new comment.

v2->v3:
  - Break up the attributes to avoid the use of raw c structs.
  - Use per-device profile instead of global profile in the driver.

v1->v2:
  - Use ethtool tool instead of net-sysfs

Heng Qi (4):
  linux/dim: move useful macros to .h file
  ethtool: provide customized dim profile management
  virtio-net: refactor dim initialization/destruction
  virtio-net: support dim profile fine-tuning

Heng Qi (4):
  linux/dim: move useful macros to .h file
  ethtool: provide customized dim profile management
  virtio-net: refactor dim initialization/destruction
  virtio-net: support dim profile fine-tuning

 Documentation/netlink/specs/ethtool.yaml     |  33 +++
 Documentation/networking/ethtool-netlink.rst |   8 +
 drivers/net/virtio_net.c                     |  46 +++--
 include/linux/dim.h                          |  13 ++
 include/linux/ethtool.h                      |  11 +-
 include/linux/netdevice.h                    |  24 +++
 include/uapi/linux/ethtool_netlink.h         |  24 +++
 lib/dim/net_dim.c                            |  10 +-
 net/core/dev.c                               |  83 ++++++++
 net/ethtool/coalesce.c                       | 201 ++++++++++++++++++-
 10 files changed, 430 insertions(+), 23 deletions(-)

-- 
2.32.0.3.g01195cf9f


^ permalink raw reply	[flat|nested] 7+ messages in thread

* [PATCH RESEND net-next v7 1/4] linux/dim: move useful macros to .h file
  2024-04-15 13:38 [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel Heng Qi
@ 2024-04-15 13:38 ` Heng Qi
  2024-04-15 13:38 ` [PATCH RESEND net-next v7 2/4] ethtool: provide customized dim profile management Heng Qi
                   ` (3 subsequent siblings)
  4 siblings, 0 replies; 7+ messages in thread
From: Heng Qi @ 2024-04-15 13:38 UTC (permalink / raw)
  To: netdev, virtualization
  Cc: Jakub Kicinski, David S . Miller, Eric Dumazet, Paolo Abeni,
	Jason Wang, Michael S . Tsirkin, Brett Creeley, Ratheesh Kannoth,
	Alexander Lobakin, Xuan Zhuo

These will be used in subsequent patches, including
newly declared profile arrays.

Signed-off-by: Heng Qi <hengqi@linux.alibaba.com>
---
 include/linux/dim.h | 13 +++++++++++++
 lib/dim/net_dim.c   | 10 ++--------
 2 files changed, 15 insertions(+), 8 deletions(-)

diff --git a/include/linux/dim.h b/include/linux/dim.h
index f343bc9aa2ec..8149d2d6589c 100644
--- a/include/linux/dim.h
+++ b/include/linux/dim.h
@@ -10,6 +10,13 @@
 #include <linux/types.h>
 #include <linux/workqueue.h>
 
+/* Number of DIM profiles and period mode. */
+#define NET_DIM_PARAMS_NUM_PROFILES 5
+#define NET_DIM_DEFAULT_RX_CQ_PKTS_FROM_EQE 256
+#define NET_DIM_DEFAULT_TX_CQ_PKTS_FROM_EQE 128
+#define NET_DIM_DEF_PROFILE_CQE 1
+#define NET_DIM_DEF_PROFILE_EQE 1
+
 /*
  * Number of events between DIM iterations.
  * Causes a moderation of the algorithm run.
@@ -127,6 +134,12 @@ enum dim_cq_period_mode {
 	DIM_CQ_PERIOD_NUM_MODES
 };
 
+extern const struct dim_cq_moder
+rx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES];
+
+extern const struct dim_cq_moder
+tx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES];
+
 /**
  * enum dim_state - DIM algorithm states
  *
diff --git a/lib/dim/net_dim.c b/lib/dim/net_dim.c
index 4e32f7aaac86..a649d9069c06 100644
--- a/lib/dim/net_dim.c
+++ b/lib/dim/net_dim.c
@@ -11,12 +11,6 @@
  *        There are different set of profiles for RX/TX CQs.
  *        Each profile size must be of NET_DIM_PARAMS_NUM_PROFILES
  */
-#define NET_DIM_PARAMS_NUM_PROFILES 5
-#define NET_DIM_DEFAULT_RX_CQ_PKTS_FROM_EQE 256
-#define NET_DIM_DEFAULT_TX_CQ_PKTS_FROM_EQE 128
-#define NET_DIM_DEF_PROFILE_CQE 1
-#define NET_DIM_DEF_PROFILE_EQE 1
-
 #define NET_DIM_RX_EQE_PROFILES { \
 	{.usec = 1,   .pkts = NET_DIM_DEFAULT_RX_CQ_PKTS_FROM_EQE,}, \
 	{.usec = 8,   .pkts = NET_DIM_DEFAULT_RX_CQ_PKTS_FROM_EQE,}, \
@@ -49,13 +43,13 @@
 	{.usec = 64, .pkts = 32,}   \
 }
 
-static const struct dim_cq_moder
+const struct dim_cq_moder
 rx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
 	NET_DIM_RX_EQE_PROFILES,
 	NET_DIM_RX_CQE_PROFILES,
 };
 
-static const struct dim_cq_moder
+const struct dim_cq_moder
 tx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
 	NET_DIM_TX_EQE_PROFILES,
 	NET_DIM_TX_CQE_PROFILES,
-- 
2.32.0.3.g01195cf9f


^ permalink raw reply related	[flat|nested] 7+ messages in thread

* [PATCH RESEND net-next v7 2/4] ethtool: provide customized dim profile management
  2024-04-15 13:38 [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel Heng Qi
  2024-04-15 13:38 ` [PATCH RESEND net-next v7 1/4] linux/dim: move useful macros to .h file Heng Qi
@ 2024-04-15 13:38 ` Heng Qi
  2024-04-15 13:38 ` [PATCH RESEND net-next v7 3/4] virtio-net: refactor dim initialization/destruction Heng Qi
                   ` (2 subsequent siblings)
  4 siblings, 0 replies; 7+ messages in thread
From: Heng Qi @ 2024-04-15 13:38 UTC (permalink / raw)
  To: netdev, virtualization
  Cc: Jakub Kicinski, David S . Miller, Eric Dumazet, Paolo Abeni,
	Jason Wang, Michael S . Tsirkin, Brett Creeley, Ratheesh Kannoth,
	Alexander Lobakin, Xuan Zhuo

The NetDIM library, currently leveraged by an array of NICs, delivers
excellent acceleration benefits. Nevertheless, NICs vary significantly
in their dim profile list prerequisites.

Specifically, virtio-net backends may present diverse sw or hw device
implementation, making a one-size-fits-all parameter list impractical.
On Alibaba Cloud, the virtio DPU's performance under the default DIM
profile falls short of expectations, partly due to a mismatch in
parameter configuration.

I also noticed that ice/idpf/ena and other NICs have customized
profilelist or placed some restrictions on dim capabilities.

Motivated by this, I tried adding new params for "ethtool -C" that provides
a per-device control to modify and access a device's interrupt parameters.

Usage
========
The target NIC is named ethx.

Assume that ethx only declares support for ETHTOOL_COALESCE_RX_EQE_PROFILE
in ethtool_ops->supported_coalesce_params.

1. Query the currently customized list of the device

$ ethtool -c ethx
...
rx-eqe-profile:
{.usec =   1, .pkts = 256, .comps =   0,},
{.usec =   8, .pkts = 256, .comps =   0,},
{.usec =  64, .pkts = 256, .comps =   0,},
{.usec = 128, .pkts = 256, .comps =   0,},
{.usec = 256, .pkts = 256, .comps =   0,}
rx-cqe-profile:   n/a
tx-eqe-profile:   n/a
tx-cqe-profile:   n/a

2. Tune
$ ethtool -C ethx rx-eqe-profile 1,1,0_2,2,0_3,3,0_4,4,0_5,5,0
$ ethtool -c ethx
...
rx-eqe-profile:
{.usec =   1, .pkts =   1, .comps =   0,},
{.usec =   2, .pkts =   2, .comps =   0,},
{.usec =   3, .pkts =   3, .comps =   0,},
{.usec =   4, .pkts =   4, .comps =   0,},
{.usec =   5, .pkts =   5, .comps =   0,}
rx-cqe-profile:   n/a
tx-eqe-profile:   n/a
tx-cqe-profile:   n/a

3. Hint
If the device does not support some type of customized dim
profiles, the corresponding "n/a" will display.

Signed-off-by: Heng Qi <hengqi@linux.alibaba.com>
---
 Documentation/netlink/specs/ethtool.yaml     |  33 +++
 Documentation/networking/ethtool-netlink.rst |   8 +
 include/linux/ethtool.h                      |  11 +-
 include/linux/netdevice.h                    |  24 +++
 include/uapi/linux/ethtool_netlink.h         |  24 +++
 net/core/dev.c                               |  83 ++++++++
 net/ethtool/coalesce.c                       | 201 ++++++++++++++++++-
 7 files changed, 382 insertions(+), 2 deletions(-)

diff --git a/Documentation/netlink/specs/ethtool.yaml b/Documentation/netlink/specs/ethtool.yaml
index 87ae7b397984..8165b598dab7 100644
--- a/Documentation/netlink/specs/ethtool.yaml
+++ b/Documentation/netlink/specs/ethtool.yaml
@@ -413,6 +413,18 @@ attribute-sets:
       -
         name: combined-count
         type: u32
+  -
+    name: moderation
+    attributes:
+      -
+        name: usec
+        type: u32
+      -
+        name: pkts
+        type: u32
+      -
+        name: comps
+        type: u32
 
   -
     name: coalesce
@@ -502,6 +514,23 @@ attribute-sets:
       -
         name: tx-aggr-time-usecs
         type: u32
+      -
+        name: rx-eqe-profile
+        type: nest
+        nested-attributes: moderation
+      -
+        name: rx-cqe-profile
+        type: nest
+        nested-attributes: moderation
+      -
+        name: tx-eqe-profile
+        type: nest
+        nested-attributes: moderation
+      -
+        name: tx-cqe-profile
+        type: nest
+        nested-attributes: moderation
+
   -
     name: pause-stat
     attributes:
@@ -1313,6 +1342,10 @@ operations:
             - tx-aggr-max-bytes
             - tx-aggr-max-frames
             - tx-aggr-time-usecs
+            - rx-eqe-profile
+            - rx-cqe-profile
+            - tx-eqe-profile
+            - tx-cqe-profile
       dump: *coalesce-get-op
     -
       name: coalesce-set
diff --git a/Documentation/networking/ethtool-netlink.rst b/Documentation/networking/ethtool-netlink.rst
index 4e63d3708ed9..98a619198465 100644
--- a/Documentation/networking/ethtool-netlink.rst
+++ b/Documentation/networking/ethtool-netlink.rst
@@ -1040,6 +1040,10 @@ Kernel response contents:
   ``ETHTOOL_A_COALESCE_TX_AGGR_MAX_BYTES``     u32     max aggr size, Tx
   ``ETHTOOL_A_COALESCE_TX_AGGR_MAX_FRAMES``    u32     max aggr packets, Tx
   ``ETHTOOL_A_COALESCE_TX_AGGR_TIME_USECS``    u32     time (us), aggr, Tx
+  ``ETHTOOL_A_COALESCE_RX_EQE_PROFILE``        nested  profile of DIM EQE, Rx
+  ``ETHTOOL_A_COALESCE_RX_CQE_PROFILE``        nested  profile of DIM CQE, Rx
+  ``ETHTOOL_A_COALESCE_TX_EQE_PROFILE``        nested  profile of DIM EQE, Tx
+  ``ETHTOOL_A_COALESCE_TX_CQE_PROFILE``        nested  profile of DIM CQE, Tx
   ===========================================  ======  =======================
 
 Attributes are only included in reply if their value is not zero or the
@@ -1105,6 +1109,10 @@ Request contents:
   ``ETHTOOL_A_COALESCE_TX_AGGR_MAX_BYTES``     u32     max aggr size, Tx
   ``ETHTOOL_A_COALESCE_TX_AGGR_MAX_FRAMES``    u32     max aggr packets, Tx
   ``ETHTOOL_A_COALESCE_TX_AGGR_TIME_USECS``    u32     time (us), aggr, Tx
+  ``ETHTOOL_A_COALESCE_RX_EQE_PROFILE``        nested  profile of DIM EQE, Rx
+  ``ETHTOOL_A_COALESCE_RX_CQE_PROFILE``        nested  profile of DIM CQE, Rx
+  ``ETHTOOL_A_COALESCE_TX_EQE_PROFILE``        nested  profile of DIM EQE, Tx
+  ``ETHTOOL_A_COALESCE_TX_CQE_PROFILE``        nested  profile of DIM CQE, Tx
   ===========================================  ======  =======================
 
 Request is rejected if it attributes declared as unsupported by driver (i.e.
diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h
index 6fd9107d3cc0..614a113eda29 100644
--- a/include/linux/ethtool.h
+++ b/include/linux/ethtool.h
@@ -284,7 +284,11 @@ bool ethtool_convert_link_mode_to_legacy_u32(u32 *legacy_u32,
 #define ETHTOOL_COALESCE_TX_AGGR_MAX_BYTES	BIT(24)
 #define ETHTOOL_COALESCE_TX_AGGR_MAX_FRAMES	BIT(25)
 #define ETHTOOL_COALESCE_TX_AGGR_TIME_USECS	BIT(26)
-#define ETHTOOL_COALESCE_ALL_PARAMS		GENMASK(26, 0)
+#define ETHTOOL_COALESCE_RX_EQE_PROFILE         BIT(27)
+#define ETHTOOL_COALESCE_RX_CQE_PROFILE         BIT(28)
+#define ETHTOOL_COALESCE_TX_EQE_PROFILE         BIT(29)
+#define ETHTOOL_COALESCE_TX_CQE_PROFILE         BIT(30)
+#define ETHTOOL_COALESCE_ALL_PARAMS		GENMASK(30, 0)
 
 #define ETHTOOL_COALESCE_USECS						\
 	(ETHTOOL_COALESCE_RX_USECS | ETHTOOL_COALESCE_TX_USECS)
@@ -316,6 +320,11 @@ bool ethtool_convert_link_mode_to_legacy_u32(u32 *legacy_u32,
 	(ETHTOOL_COALESCE_TX_AGGR_MAX_BYTES |	\
 	 ETHTOOL_COALESCE_TX_AGGR_MAX_FRAMES |	\
 	 ETHTOOL_COALESCE_TX_AGGR_TIME_USECS)
+#define ETHTOOL_COALESCE_PROFILE		\
+	(ETHTOOL_COALESCE_RX_EQE_PROFILE |	\
+	 ETHTOOL_COALESCE_RX_CQE_PROFILE |	\
+	 ETHTOOL_COALESCE_TX_EQE_PROFILE |	\
+	 ETHTOOL_COALESCE_TX_CQE_PROFILE)
 
 #define ETHTOOL_STAT_NOT_SET	(~0ULL)
 
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index d45f330d083d..a1c7e9c2be86 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -80,6 +80,25 @@ struct xdp_frame;
 struct xdp_metadata_ops;
 struct xdp_md;
 
+#if IS_ENABLED(CONFIG_DIMLIB)
+struct dim_cq_moder;
+
+#define NETDEV_PROFILE_USEC	BIT(0)	/* device supports usec field modification */
+#define NETDEV_PROFILE_PKTS	BIT(1)	/* device supports pkts field modification */
+#define NETDEV_PROFILE_COMPS	BIT(2)	/* device supports comps field modification */
+
+struct netdev_profile_moder {
+	/* See NETDEV_PROFILE_* */
+	unsigned int flags;
+
+	/* DIM profile lists for different dim cq modes */
+	struct dim_cq_moder *rx_eqe_profile;
+	struct dim_cq_moder *rx_cqe_profile;
+	struct dim_cq_moder *tx_eqe_profile;
+	struct dim_cq_moder *tx_cqe_profile;
+};
+#endif
+
 typedef u32 xdp_features_t;
 
 void synchronize_net(void);
@@ -2400,6 +2419,11 @@ struct net_device {
 	/** @page_pools: page pools created for this netdevice */
 	struct hlist_head	page_pools;
 #endif
+
+#if IS_ENABLED(CONFIG_DIMLIB)
+	/** @moderation: dim tunable parameters for this netdevice */
+	struct netdev_profile_moder	*moderation;
+#endif
 };
 #define to_net_dev(d) container_of(d, struct net_device, dev)
 
diff --git a/include/uapi/linux/ethtool_netlink.h b/include/uapi/linux/ethtool_netlink.h
index b4f0d233d048..d4c6e30a55cb 100644
--- a/include/uapi/linux/ethtool_netlink.h
+++ b/include/uapi/linux/ethtool_netlink.h
@@ -416,12 +416,36 @@ enum {
 	ETHTOOL_A_COALESCE_TX_AGGR_MAX_BYTES,		/* u32 */
 	ETHTOOL_A_COALESCE_TX_AGGR_MAX_FRAMES,		/* u32 */
 	ETHTOOL_A_COALESCE_TX_AGGR_TIME_USECS,		/* u32 */
+	ETHTOOL_A_COALESCE_RX_EQE_PROFILE,              /* nest - _A_MODERATIONS_MODERATION */
+	ETHTOOL_A_COALESCE_RX_CQE_PROFILE,              /* nest - _A_MODERATIONS_MODERATION */
+	ETHTOOL_A_COALESCE_TX_EQE_PROFILE,              /* nest - _A_MODERATIONS_MODERATION */
+	ETHTOOL_A_COALESCE_TX_CQE_PROFILE,              /* nest - _A_MODERATIONS_MODERATION */
 
 	/* add new constants above here */
 	__ETHTOOL_A_COALESCE_CNT,
 	ETHTOOL_A_COALESCE_MAX = (__ETHTOOL_A_COALESCE_CNT - 1)
 };
 
+enum {
+	ETHTOOL_A_MODERATIONS_UNSPEC,
+	ETHTOOL_A_MODERATIONS_MODERATION,		/* nest, _A_MODERATION_* */
+
+	/* add new constants above here */
+	__ETHTOOL_A_MODERATIONS_CNT,
+	ETHTOOL_A_MODERATIONS_MAX = (__ETHTOOL_A_MODERATIONS_CNT - 1)
+};
+
+enum {
+	ETHTOOL_A_MODERATION_UNSPEC,
+	ETHTOOL_A_MODERATION_USEC,			/* u32 */
+	ETHTOOL_A_MODERATION_PKTS,			/* u32 */
+	ETHTOOL_A_MODERATION_COMPS,			/* u32 */
+
+	/* add new constants above here */
+	__ETHTOOL_A_MODERATION_CNT,
+	ETHTOOL_A_MODERATION_MAX = (__ETHTOOL_A_MODERATION_CNT - 1)
+};
+
 /* PAUSE */
 
 enum {
diff --git a/net/core/dev.c b/net/core/dev.c
index 854a3a28a8d8..dba27150f035 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -96,6 +96,7 @@
 #include <linux/kthread.h>
 #include <linux/bpf.h>
 #include <linux/bpf_trace.h>
+#include <linux/dim.h>
 #include <net/net_namespace.h>
 #include <net/sock.h>
 #include <net/busy_poll.h>
@@ -10229,6 +10230,61 @@ static void netdev_do_free_pcpu_stats(struct net_device *dev)
 	}
 }
 
+static int dev_dim_profile_init(struct net_device *dev)
+{
+#if IS_ENABLED(CONFIG_DIMLIB)
+	u32 supported = dev->ethtool_ops->supported_coalesce_params;
+	struct netdev_profile_moder *moder;
+	int length;
+
+	dev->moderation = kzalloc(sizeof(*dev->moderation), GFP_KERNEL);
+	if (!dev->moderation)
+		goto err_moder;
+
+	moder = dev->moderation;
+	length = NET_DIM_PARAMS_NUM_PROFILES * sizeof(*moder->rx_eqe_profile);
+
+	if (supported & ETHTOOL_COALESCE_RX_EQE_PROFILE) {
+		moder->rx_eqe_profile = kzalloc(length, GFP_KERNEL);
+		if (!moder->rx_eqe_profile)
+			goto err_rx_eqe;
+		memcpy(moder->rx_eqe_profile, rx_profile[0], length);
+	}
+	if (supported & ETHTOOL_COALESCE_RX_CQE_PROFILE) {
+		moder->rx_cqe_profile = kzalloc(length, GFP_KERNEL);
+		if (!moder->rx_cqe_profile)
+			goto err_rx_cqe;
+		memcpy(moder->rx_cqe_profile, rx_profile[1], length);
+	}
+	if (supported & ETHTOOL_COALESCE_TX_EQE_PROFILE) {
+		moder->tx_eqe_profile = kzalloc(length, GFP_KERNEL);
+		if (!moder->tx_eqe_profile)
+			goto err_tx_eqe;
+		memcpy(moder->tx_eqe_profile, tx_profile[0], length);
+	}
+	if (supported & ETHTOOL_COALESCE_TX_CQE_PROFILE) {
+		moder->tx_cqe_profile = kzalloc(length, GFP_KERNEL);
+		if (!moder->tx_cqe_profile)
+			goto err_tx_cqe;
+		memcpy(moder->tx_cqe_profile, tx_profile[1], length);
+	}
+#endif
+	return 0;
+
+#if IS_ENABLED(CONFIG_DIMLIB)
+err_tx_cqe:
+	kfree(moder->tx_eqe_profile);
+err_tx_eqe:
+	kfree(moder->rx_cqe_profile);
+err_rx_cqe:
+	kfree(moder->rx_eqe_profile);
+err_rx_eqe:
+	kfree(moder);
+err_moder:
+	return -ENOMEM;
+#endif
+}
+
 /**
  * register_netdevice() - register a network device
  * @dev: device to register
@@ -10258,6 +10314,10 @@ int register_netdevice(struct net_device *dev)
 	if (ret)
 		return ret;
 
+	ret = dev_dim_profile_init(dev);
+	if (ret)
+		return ret;
+
 	spin_lock_init(&dev->addr_list_lock);
 	netdev_set_addr_lockdep_class(dev);
 
@@ -11011,6 +11071,27 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
 }
 EXPORT_SYMBOL(alloc_netdev_mqs);
 
+static void netif_free_profile(struct net_device *dev)
+{
+#if IS_ENABLED(CONFIG_DIMLIB)
+	u32 supported = dev->ethtool_ops->supported_coalesce_params;
+
+	if (supported & ETHTOOL_COALESCE_RX_EQE_PROFILE)
+		kfree(dev->moderation->rx_eqe_profile);
+
+	if (supported & ETHTOOL_COALESCE_RX_CQE_PROFILE)
+		kfree(dev->moderation->rx_cqe_profile);
+
+	if (supported & ETHTOOL_COALESCE_TX_EQE_PROFILE)
+		kfree(dev->moderation->tx_eqe_profile);
+
+	if (supported & ETHTOOL_COALESCE_TX_CQE_PROFILE)
+		kfree(dev->moderation->tx_cqe_profile);
+
+	kfree(dev->moderation);
+#endif
+}
+
 /**
  * free_netdev - free network device
  * @dev: device
@@ -11036,6 +11117,8 @@ void free_netdev(struct net_device *dev)
 		return;
 	}
 
+	netif_free_profile(dev);
+
 	netif_free_tx_queues(dev);
 	netif_free_rx_queues(dev);
 
diff --git a/net/ethtool/coalesce.c b/net/ethtool/coalesce.c
index 83112c1a71ae..d750045fdad8 100644
--- a/net/ethtool/coalesce.c
+++ b/net/ethtool/coalesce.c
@@ -1,5 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0-only
 
+#include <linux/dim.h>
 #include "netlink.h"
 #include "common.h"
 
@@ -51,6 +52,10 @@ __CHECK_SUPPORTED_OFFSET(COALESCE_RX_MAX_FRAMES_HIGH);
 __CHECK_SUPPORTED_OFFSET(COALESCE_TX_USECS_HIGH);
 __CHECK_SUPPORTED_OFFSET(COALESCE_TX_MAX_FRAMES_HIGH);
 __CHECK_SUPPORTED_OFFSET(COALESCE_RATE_SAMPLE_INTERVAL);
+__CHECK_SUPPORTED_OFFSET(COALESCE_RX_EQE_PROFILE);
+__CHECK_SUPPORTED_OFFSET(COALESCE_RX_CQE_PROFILE);
+__CHECK_SUPPORTED_OFFSET(COALESCE_TX_EQE_PROFILE);
+__CHECK_SUPPORTED_OFFSET(COALESCE_TX_CQE_PROFILE);
 
 const struct nla_policy ethnl_coalesce_get_policy[] = {
 	[ETHTOOL_A_COALESCE_HEADER]		=
@@ -82,6 +87,14 @@ static int coalesce_prepare_data(const struct ethnl_req_info *req_base,
 static int coalesce_reply_size(const struct ethnl_req_info *req_base,
 			       const struct ethnl_reply_data *reply_base)
 {
+	int modersz = nla_total_size(0) + /* _MODERATIONS_MODERATION, nest */
+		      nla_total_size(sizeof(u32)) + /* _MODERATION_USEC */
+		      nla_total_size(sizeof(u32)) + /* _MODERATION_PKTS */
+		      nla_total_size(sizeof(u32));  /* _MODERATION_COMPS */
+
+	int total_modersz = nla_total_size(0) +  /* _{R,T}X_{E,C}QE_PROFILE, nest */
+			modersz * NET_DIM_PARAMS_NUM_PROFILES;
+
 	return nla_total_size(sizeof(u32)) +	/* _RX_USECS */
 	       nla_total_size(sizeof(u32)) +	/* _RX_MAX_FRAMES */
 	       nla_total_size(sizeof(u32)) +	/* _RX_USECS_IRQ */
@@ -108,7 +121,8 @@ static int coalesce_reply_size(const struct ethnl_req_info *req_base,
 	       nla_total_size(sizeof(u8)) +	/* _USE_CQE_MODE_RX */
 	       nla_total_size(sizeof(u32)) +	/* _TX_AGGR_MAX_BYTES */
 	       nla_total_size(sizeof(u32)) +	/* _TX_AGGR_MAX_FRAMES */
-	       nla_total_size(sizeof(u32));	/* _TX_AGGR_TIME_USECS */
+	       nla_total_size(sizeof(u32)) +	/* _TX_AGGR_TIME_USECS */
+	       total_modersz * 4;		/* _{R,T}X_{E,C}QE_PROFILE */
 }
 
 static bool coalesce_put_u32(struct sk_buff *skb, u16 attr_type, u32 val,
@@ -127,6 +141,62 @@ static bool coalesce_put_bool(struct sk_buff *skb, u16 attr_type, u32 val,
 	return nla_put_u8(skb, attr_type, !!val);
 }
 
+#if IS_ENABLED(CONFIG_DIMLIB)
+/**
+ * coalesce_put_profile - fill reply with a nla nest with four child nla nests.
+ * @skb: socket buffer the message is stored in
+ * @attr_type: nest attr type ETHTOOL_A_COALESCE_*X_*QE_PROFILE
+ * @profile: data passed to userspace
+ * @supported_params: modifiable parameters supported by the driver
+ *
+ * Put a dim profile nest attribute. Refer to ETHTOOL_A_MODERATIONS_MODERATION.
+ *
+ * Returns false to indicate successful placement or no placement, and
+ * returns true to pass the -EMSGSIZE error to the wrapper.
+ */
+static bool coalesce_put_profile(struct sk_buff *skb, u16 attr_type,
+				 const struct dim_cq_moder *profile,
+				 u32 supported_params)
+{
+	struct nlattr *profile_attr, *moder_attr;
+	bool emsg = !!-EMSGSIZE;
+	int i;
+
+	if (!profile)
+		return false;
+
+	if (!(supported_params & attr_to_mask(attr_type)))
+		return false;
+
+	profile_attr = nla_nest_start(skb, attr_type);
+	if (!profile_attr)
+		return emsg;
+
+	for (i = 0; i < NET_DIM_PARAMS_NUM_PROFILES; i++) {
+		moder_attr = nla_nest_start(skb, ETHTOOL_A_MODERATIONS_MODERATION);
+		if (!moder_attr)
+			goto nla_cancel_profile;
+
+		if (nla_put_u32(skb, ETHTOOL_A_MODERATION_USEC, profile[i].usec) ||
+		    nla_put_u32(skb, ETHTOOL_A_MODERATION_PKTS, profile[i].pkts) ||
+		    nla_put_u32(skb, ETHTOOL_A_MODERATION_COMPS, profile[i].comps))
+			goto nla_cancel_moder;
+
+		nla_nest_end(skb, moder_attr);
+	}
+
+	nla_nest_end(skb, profile_attr);
+
+	return 0;
+
+nla_cancel_moder:
+	nla_nest_cancel(skb, moder_attr);
+nla_cancel_profile:
+	nla_nest_cancel(skb, profile_attr);
+	return emsg;
+}
+#endif
+
 static int coalesce_fill_reply(struct sk_buff *skb,
 			       const struct ethnl_req_info *req_base,
 			       const struct ethnl_reply_data *reply_base)
@@ -134,6 +204,9 @@ static int coalesce_fill_reply(struct sk_buff *skb,
 	const struct coalesce_reply_data *data = COALESCE_REPDATA(reply_base);
 	const struct kernel_ethtool_coalesce *kcoal = &data->kernel_coalesce;
 	const struct ethtool_coalesce *coal = &data->coalesce;
+#if IS_ENABLED(CONFIG_DIMLIB)
+	struct net_device *dev = req_base->dev;
+#endif
 	u32 supported = data->supported_params;
 
 	if (coalesce_put_u32(skb, ETHTOOL_A_COALESCE_RX_USECS,
@@ -192,6 +265,21 @@ static int coalesce_fill_reply(struct sk_buff *skb,
 			     kcoal->tx_aggr_time_usecs, supported))
 		return -EMSGSIZE;
 
+#if IS_ENABLED(CONFIG_DIMLIB)
+	if (!(dev->moderation->flags & (NETDEV_PROFILE_USEC | NETDEV_PROFILE_PKTS |
+					NETDEV_PROFILE_COMPS)))
+		return 0;
+
+	if (coalesce_put_profile(skb, ETHTOOL_A_COALESCE_RX_EQE_PROFILE,
+				 dev->moderation->rx_eqe_profile, supported) ||
+	    coalesce_put_profile(skb, ETHTOOL_A_COALESCE_RX_CQE_PROFILE,
+				 dev->moderation->rx_cqe_profile, supported) ||
+	    coalesce_put_profile(skb, ETHTOOL_A_COALESCE_TX_EQE_PROFILE,
+				 dev->moderation->tx_eqe_profile, supported) ||
+	    coalesce_put_profile(skb, ETHTOOL_A_COALESCE_TX_CQE_PROFILE,
+				 dev->moderation->tx_cqe_profile, supported))
+		return -EMSGSIZE;
+#endif
 	return 0;
 }
 
@@ -227,7 +315,19 @@ const struct nla_policy ethnl_coalesce_set_policy[] = {
 	[ETHTOOL_A_COALESCE_TX_AGGR_MAX_BYTES] = { .type = NLA_U32 },
 	[ETHTOOL_A_COALESCE_TX_AGGR_MAX_FRAMES] = { .type = NLA_U32 },
 	[ETHTOOL_A_COALESCE_TX_AGGR_TIME_USECS] = { .type = NLA_U32 },
+	[ETHTOOL_A_COALESCE_RX_EQE_PROFILE]     = { .type = NLA_NESTED },
+	[ETHTOOL_A_COALESCE_RX_CQE_PROFILE]     = { .type = NLA_NESTED },
+	[ETHTOOL_A_COALESCE_TX_EQE_PROFILE]     = { .type = NLA_NESTED },
+	[ETHTOOL_A_COALESCE_TX_CQE_PROFILE]     = { .type = NLA_NESTED },
+};
+
+#if IS_ENABLED(CONFIG_DIMLIB)
+static const struct nla_policy coalesce_set_profile_policy[] = {
+	[ETHTOOL_A_MODERATION_USEC]	= {.type = NLA_U32},
+	[ETHTOOL_A_MODERATION_PKTS]	= {.type = NLA_U32},
+	[ETHTOOL_A_MODERATION_COMPS]	= {.type = NLA_U32},
 };
+#endif
 
 static int
 ethnl_set_coalesce_validate(struct ethnl_req_info *req_info,
@@ -253,6 +353,76 @@ ethnl_set_coalesce_validate(struct ethnl_req_info *req_info,
 	return 1;
 }
 
+#if IS_ENABLED(CONFIG_DIMLIB)
+/**
+ * ethnl_update_profile - get a nla nest with four child nla nests from userspace.
+ * @dev: netdevice to update the profile
+ * @dst: data get from the driver and modified by ethnl_update_profile.
+ * @nests: nest attr ETHTOOL_A_COALESCE_*X_*QE_PROFILE to set driver's profile.
+ * @extack: Netlink extended ack
+ *
+ * Layout of nests:
+ *   Nested ETHTOOL_A_COALESCE_*X_*QE_PROFILE attr
+ *     Nested ETHTOOL_A_MODERATIONS_MODERATION attr
+ *       ETHTOOL_A_MODERATION_USEC attr
+ *       ETHTOOL_A_MODERATION_PKTS attr
+ *       ETHTOOL_A_MODERATION_COMPS attr
+ *     ...
+ *     Nested ETHTOOL_A_MODERATIONS_MODERATION attr
+ *       ETHTOOL_A_MODERATION_USEC attr
+ *       ETHTOOL_A_MODERATION_PKTS attr
+ *       ETHTOOL_A_MODERATION_COMPS attr
+ *
+ * Returns 0 on success or a negative error code.
+ */
+static inline int ethnl_update_profile(struct net_device *dev,
+				       struct dim_cq_moder *dst,
+				       const struct nlattr *nests,
+				       struct netlink_ext_ack *extack)
+{
+	struct nlattr *tb_moder[ARRAY_SIZE(coalesce_set_profile_policy)];
+	struct dim_cq_moder profile[NET_DIM_PARAMS_NUM_PROFILES];
+	struct netdev_profile_moder *moder = dev->moderation;
+	struct nlattr *nest;
+	int ret, rem, i = 0;
+
+	if (!nests)
+		return 0;
+
+	if (!dst)
+		return -EOPNOTSUPP;
+
+	nla_for_each_nested_type(nest, ETHTOOL_A_MODERATIONS_MODERATION, nests, rem) {
+		ret = nla_parse_nested(tb_moder,
+				       ARRAY_SIZE(coalesce_set_profile_policy) - 1,
+				       nest, coalesce_set_profile_policy,
+				       extack);
+		if (ret)
+			return ret;
+
+		if (NL_REQ_ATTR_CHECK(extack, nest, tb_moder, ETHTOOL_A_MODERATION_USEC) ||
+		    NL_REQ_ATTR_CHECK(extack, nest, tb_moder, ETHTOOL_A_MODERATION_PKTS) ||
+		    NL_REQ_ATTR_CHECK(extack, nest, tb_moder, ETHTOOL_A_MODERATION_COMPS))
+			return -EINVAL;
+
+		profile[i].usec = nla_get_u32(tb_moder[ETHTOOL_A_MODERATION_USEC]);
+		profile[i].pkts = nla_get_u32(tb_moder[ETHTOOL_A_MODERATION_PKTS]);
+		profile[i].comps = nla_get_u32(tb_moder[ETHTOOL_A_MODERATION_COMPS]);
+
+		if ((dst[i].usec != profile[i].usec && !(moder->flags & NETDEV_PROFILE_USEC)) ||
+		    (dst[i].pkts != profile[i].pkts && !(moder->flags & NETDEV_PROFILE_PKTS)) ||
+		    (dst[i].comps != profile[i].comps && !(moder->flags & NETDEV_PROFILE_COMPS)))
+			return -EOPNOTSUPP;
+
+		i++;
+	}
+
+	memcpy(dst, profile, sizeof(profile));
+
+	return 0;
+}
+#endif
+
 static int
 __ethnl_set_coalesce(struct ethnl_req_info *req_info, struct genl_info *info,
 		     bool *dual_change)
@@ -317,6 +487,35 @@ __ethnl_set_coalesce(struct ethnl_req_info *req_info, struct genl_info *info,
 	ethnl_update_u32(&kernel_coalesce.tx_aggr_time_usecs,
 			 tb[ETHTOOL_A_COALESCE_TX_AGGR_TIME_USECS], &mod);
 
+#if IS_ENABLED(CONFIG_DIMLIB)
+	ret = ethnl_update_profile(dev, dev->moderation->rx_eqe_profile,
+				   tb[ETHTOOL_A_COALESCE_RX_EQE_PROFILE],
+				   info->extack);
+	if (ret < 0)
+		return ret;
+	ret = ethnl_update_profile(dev, dev->moderation->rx_cqe_profile,
+				   tb[ETHTOOL_A_COALESCE_RX_CQE_PROFILE],
+				   info->extack);
+	if (ret < 0)
+		return ret;
+	ret = ethnl_update_profile(dev, dev->moderation->tx_eqe_profile,
+				   tb[ETHTOOL_A_COALESCE_TX_EQE_PROFILE],
+				   info->extack);
+	if (ret < 0)
+		return ret;
+	ret = ethnl_update_profile(dev, dev->moderation->tx_cqe_profile,
+				   tb[ETHTOOL_A_COALESCE_TX_CQE_PROFILE],
+				   info->extack);
+	if (ret < 0)
+		return ret;
+#else
+	if (tb[ETHTOOL_A_COALESCE_RX_EQE_PROFILE] ||
+	    tb[ETHTOOL_A_COALESCE_RX_CQE_PROFILE] ||
+	    tb[ETHTOOL_A_COALESCE_TX_EQE_PROFILE] ||
+	    tb[ETHTOOL_A_COALESCE_TX_CQE_PROFILE])
+		return -EOPNOTSUPP;
+
+#endif
 	/* Update operation modes */
 	ethnl_update_bool32(&coalesce.use_adaptive_rx_coalesce,
 			    tb[ETHTOOL_A_COALESCE_USE_ADAPTIVE_RX], &mod_mode);
-- 
2.32.0.3.g01195cf9f


^ permalink raw reply related	[flat|nested] 7+ messages in thread

* [PATCH RESEND net-next v7 3/4] virtio-net: refactor dim initialization/destruction
  2024-04-15 13:38 [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel Heng Qi
  2024-04-15 13:38 ` [PATCH RESEND net-next v7 1/4] linux/dim: move useful macros to .h file Heng Qi
  2024-04-15 13:38 ` [PATCH RESEND net-next v7 2/4] ethtool: provide customized dim profile management Heng Qi
@ 2024-04-15 13:38 ` Heng Qi
  2024-04-15 13:38 ` [PATCH RESEND net-next v7 4/4] virtio-net: support dim profile fine-tuning Heng Qi
  2024-04-22 20:55 ` [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel Michael S. Tsirkin
  4 siblings, 0 replies; 7+ messages in thread
From: Heng Qi @ 2024-04-15 13:38 UTC (permalink / raw)
  To: netdev, virtualization
  Cc: Jakub Kicinski, David S . Miller, Eric Dumazet, Paolo Abeni,
	Jason Wang, Michael S . Tsirkin, Brett Creeley, Ratheesh Kannoth,
	Alexander Lobakin, Xuan Zhuo

Extract the initialization and destruction actions
of dim for use in the next patch.

Signed-off-by: Heng Qi <hengqi@linux.alibaba.com>
---
 drivers/net/virtio_net.c | 38 +++++++++++++++++++++++++++-----------
 1 file changed, 27 insertions(+), 11 deletions(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 115c3c5414f2..858d7ab89e34 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -2274,6 +2274,13 @@ static int virtnet_enable_queue_pair(struct virtnet_info *vi, int qp_index)
 	return err;
 }
 
+static void virtnet_dim_clean(struct virtnet_info *vi,
+			      int start_qnum, int end_qnum)
+{
+	for (; start_qnum <= end_qnum; start_qnum++)
+		cancel_work_sync(&vi->rq[start_qnum].dim.work);
+}
+
 static int virtnet_open(struct net_device *dev)
 {
 	struct virtnet_info *vi = netdev_priv(dev);
@@ -2297,11 +2304,9 @@ static int virtnet_open(struct net_device *dev)
 err_enable_qp:
 	disable_delayed_refill(vi);
 	cancel_delayed_work_sync(&vi->refill);
-
-	for (i--; i >= 0; i--) {
+	virtnet_dim_clean(vi, 0, i - 1);
+	for (i--; i >= 0; i--)
 		virtnet_disable_queue_pair(vi, i);
-		cancel_work_sync(&vi->rq[i].dim.work);
-	}
 
 	return err;
 }
@@ -2466,7 +2471,7 @@ static int virtnet_rx_resize(struct virtnet_info *vi,
 
 	if (running) {
 		napi_disable(&rq->napi);
-		cancel_work_sync(&rq->dim.work);
+		virtnet_dim_clean(vi, qindex, qindex);
 	}
 
 	err = virtqueue_resize(rq->vq, ring_num, virtnet_rq_unmap_free_buf);
@@ -2716,10 +2721,9 @@ static int virtnet_close(struct net_device *dev)
 	/* Make sure refill_work doesn't re-enable napi! */
 	cancel_delayed_work_sync(&vi->refill);
 
-	for (i = 0; i < vi->max_queue_pairs; i++) {
+	virtnet_dim_clean(vi, 0, vi->max_queue_pairs - 1);
+	for (i = 0; i < vi->max_queue_pairs; i++)
 		virtnet_disable_queue_pair(vi, i);
-		cancel_work_sync(&vi->rq[i].dim.work);
-	}
 
 	return 0;
 }
@@ -4434,6 +4438,19 @@ static int virtnet_find_vqs(struct virtnet_info *vi)
 	return ret;
 }
 
+static void virtnet_dim_init(struct virtnet_info *vi)
+{
+	int i;
+
+	if (!virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL))
+		return;
+
+	for (i = 0; i < vi->max_queue_pairs; i++) {
+		INIT_WORK(&vi->rq[i].dim.work, virtnet_rx_dim_work);
+		vi->rq[i].dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
+	}
+}
+
 static int virtnet_alloc_queues(struct virtnet_info *vi)
 {
 	int i;
@@ -4461,9 +4478,6 @@ static int virtnet_alloc_queues(struct virtnet_info *vi)
 					 virtnet_poll_tx,
 					 napi_tx ? napi_weight : 0);
 
-		INIT_WORK(&vi->rq[i].dim.work, virtnet_rx_dim_work);
-		vi->rq[i].dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
-
 		sg_init_table(vi->rq[i].sg, ARRAY_SIZE(vi->rq[i].sg));
 		ewma_pkt_len_init(&vi->rq[i].mrg_avg_pkt_len);
 		sg_init_table(vi->sq[i].sg, ARRAY_SIZE(vi->sq[i].sg));
@@ -4873,6 +4887,8 @@ static int virtnet_probe(struct virtio_device *vdev)
 
 	virtio_device_ready(vdev);
 
+	virtnet_dim_init(vi);
+
 	_virtnet_set_queues(vi, vi->curr_queue_pairs);
 
 	/* a random MAC address has been assigned, notify the device.
-- 
2.32.0.3.g01195cf9f


^ permalink raw reply related	[flat|nested] 7+ messages in thread

* [PATCH RESEND net-next v7 4/4] virtio-net: support dim profile fine-tuning
  2024-04-15 13:38 [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel Heng Qi
                   ` (2 preceding siblings ...)
  2024-04-15 13:38 ` [PATCH RESEND net-next v7 3/4] virtio-net: refactor dim initialization/destruction Heng Qi
@ 2024-04-15 13:38 ` Heng Qi
  2024-04-22 20:55 ` [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel Michael S. Tsirkin
  4 siblings, 0 replies; 7+ messages in thread
From: Heng Qi @ 2024-04-15 13:38 UTC (permalink / raw)
  To: netdev, virtualization
  Cc: Jakub Kicinski, David S . Miller, Eric Dumazet, Paolo Abeni,
	Jason Wang, Michael S . Tsirkin, Brett Creeley, Ratheesh Kannoth,
	Alexander Lobakin, Xuan Zhuo

Virtio-net has different types of back-end device implementations.
In order to effectively optimize the dim library's gains for
different device implementations, let's use the new interface
params to fine-tune the profile list.

Since the profile now exists in netdevice, adding a function similar
to net_dim_get_rx_moderation_dev() with netdevice as argument is
nice, but this would be better along with cleaning up the rest of
the drivers, which we can get to very soon after this set.

Signed-off-by: Heng Qi <hengqi@linux.alibaba.com>
---
 drivers/net/virtio_net.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 858d7ab89e34..1f7201b82be6 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -3584,7 +3584,7 @@ static void virtnet_rx_dim_work(struct work_struct *work)
 		if (!rq->dim_enabled)
 			continue;
 
-		update_moder = net_dim_get_rx_moderation(dim->mode, dim->profile_ix);
+		update_moder = dev->moderation->rx_eqe_profile[dim->profile_ix];
 		if (update_moder.usec != rq->intr_coal.max_usecs ||
 		    update_moder.pkts != rq->intr_coal.max_packets) {
 			err = virtnet_send_rx_ctrl_coal_vq_cmd(vi, qnum,
@@ -3884,7 +3884,8 @@ static int virtnet_set_rxnfc(struct net_device *dev, struct ethtool_rxnfc *info)
 
 static const struct ethtool_ops virtnet_ethtool_ops = {
 	.supported_coalesce_params = ETHTOOL_COALESCE_MAX_FRAMES |
-		ETHTOOL_COALESCE_USECS | ETHTOOL_COALESCE_USE_ADAPTIVE_RX,
+		ETHTOOL_COALESCE_USECS | ETHTOOL_COALESCE_USE_ADAPTIVE_RX |
+		ETHTOOL_COALESCE_RX_EQE_PROFILE,
 	.get_drvinfo = virtnet_get_drvinfo,
 	.get_link = ethtool_op_get_link,
 	.get_ringparam = virtnet_get_ringparam,
@@ -4440,6 +4441,7 @@ static int virtnet_find_vqs(struct virtnet_info *vi)
 
 static void virtnet_dim_init(struct virtnet_info *vi)
 {
+	struct netdev_profile_moder *moder = vi->dev->moderation;
 	int i;
 
 	if (!virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL))
@@ -4449,6 +4451,8 @@ static void virtnet_dim_init(struct virtnet_info *vi)
 		INIT_WORK(&vi->rq[i].dim.work, virtnet_rx_dim_work);
 		vi->rq[i].dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
 	}
+
+	moder->flags |= NETDEV_PROFILE_USEC | NETDEV_PROFILE_PKTS;
 }
 
 static int virtnet_alloc_queues(struct virtnet_info *vi)
-- 
2.32.0.3.g01195cf9f


^ permalink raw reply related	[flat|nested] 7+ messages in thread

* Re: [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel
  2024-04-15 13:38 [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel Heng Qi
                   ` (3 preceding siblings ...)
  2024-04-15 13:38 ` [PATCH RESEND net-next v7 4/4] virtio-net: support dim profile fine-tuning Heng Qi
@ 2024-04-22 20:55 ` Michael S. Tsirkin
  2024-04-24 14:26   ` Heng Qi
  4 siblings, 1 reply; 7+ messages in thread
From: Michael S. Tsirkin @ 2024-04-22 20:55 UTC (permalink / raw)
  To: Heng Qi
  Cc: netdev, virtualization, Jakub Kicinski, David S . Miller,
	Eric Dumazet, Paolo Abeni, Jason Wang, Brett Creeley,
	Ratheesh Kannoth, Alexander Lobakin, Xuan Zhuo

On Mon, Apr 15, 2024 at 09:38:03PM +0800, Heng Qi wrote:
> The NetDIM library provides excellent acceleration for many modern
> network cards. However, the default profiles of DIM limits its maximum
> capabilities for different NICs, so providing a way which the NIC can
> be custom configured is necessary.
> 
> Currently, interaction with the driver is still based on the commonly
> used "ethtool -C".
> 
> Since the profile now exists in netdevice, adding a function similar
> to net_dim_get_rx_moderation_dev() with netdevice as argument is
> nice, but this would be better along with cleaning up the rest of
> the drivers, which we can get to very soon after this set.
> 
> Please review, thank you very much!


Acked-by: Michael S. Tsirkin <mst@redhat.com>

> Changelog
> =====
> v6->v7:
>   - A new wrapper struct pointer is used in struct net_device.
>   - Add IS_ENABLED(CONFIG_DIMLIB) to avoid compiler warnings.
>   - Profile fields changed from u16 to u32.
> 
> v5->v6:
>   - Place the profile in netdevice to bypass the driver.
>     The interaction code of ethtool <-> kernel has not changed at all,
>     only the interaction part of kernel <-> driver has changed.
> 
> v4->v5:
>   - Update some snippets from Kuba, Thanks.
> 
> v3->v4:
>   - Some tiny updates and patch 1 only add a new comment.
> 
> v2->v3:
>   - Break up the attributes to avoid the use of raw c structs.
>   - Use per-device profile instead of global profile in the driver.
> 
> v1->v2:
>   - Use ethtool tool instead of net-sysfs
> 
> Heng Qi (4):
>   linux/dim: move useful macros to .h file
>   ethtool: provide customized dim profile management
>   virtio-net: refactor dim initialization/destruction
>   virtio-net: support dim profile fine-tuning
> 
> Heng Qi (4):
>   linux/dim: move useful macros to .h file
>   ethtool: provide customized dim profile management
>   virtio-net: refactor dim initialization/destruction
>   virtio-net: support dim profile fine-tuning
> 
>  Documentation/netlink/specs/ethtool.yaml     |  33 +++
>  Documentation/networking/ethtool-netlink.rst |   8 +
>  drivers/net/virtio_net.c                     |  46 +++--
>  include/linux/dim.h                          |  13 ++
>  include/linux/ethtool.h                      |  11 +-
>  include/linux/netdevice.h                    |  24 +++
>  include/uapi/linux/ethtool_netlink.h         |  24 +++
>  lib/dim/net_dim.c                            |  10 +-
>  net/core/dev.c                               |  83 ++++++++
>  net/ethtool/coalesce.c                       | 201 ++++++++++++++++++-
>  10 files changed, 430 insertions(+), 23 deletions(-)
> 
> -- 
> 2.32.0.3.g01195cf9f


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel
  2024-04-22 20:55 ` [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel Michael S. Tsirkin
@ 2024-04-24 14:26   ` Heng Qi
  0 siblings, 0 replies; 7+ messages in thread
From: Heng Qi @ 2024-04-24 14:26 UTC (permalink / raw)
  To: Michael S. Tsirkin
  Cc: netdev, virtualization, Jakub Kicinski, David S . Miller,
	Eric Dumazet, Paolo Abeni, Jason Wang, Brett Creeley,
	Ratheesh Kannoth, Alexander Lobakin, Xuan Zhuo


在 2024/4/23 上午4:55, Michael S. Tsirkin 写道:
> On Mon, Apr 15, 2024 at 09:38:03PM +0800, Heng Qi wrote:
>> The NetDIM library provides excellent acceleration for many modern
>> network cards. However, the default profiles of DIM limits its maximum
>> capabilities for different NICs, so providing a way which the NIC can
>> be custom configured is necessary.
>>
>> Currently, interaction with the driver is still based on the commonly
>> used "ethtool -C".
>>
>> Since the profile now exists in netdevice, adding a function similar
>> to net_dim_get_rx_moderation_dev() with netdevice as argument is
>> nice, but this would be better along with cleaning up the rest of
>> the drivers, which we can get to very soon after this set.
>>
>> Please review, thank you very much!
>
> Acked-by: Michael S. Tsirkin <mst@redhat.com>

Hi Michael,

Thanks for your ACK, the profile series has progressed to v9 and is still

being iterated. it would be better to collect your tag again in a newer 
version.

Thanks!

>
>> Changelog
>> =====
>> v6->v7:
>>    - A new wrapper struct pointer is used in struct net_device.
>>    - Add IS_ENABLED(CONFIG_DIMLIB) to avoid compiler warnings.
>>    - Profile fields changed from u16 to u32.
>>
>> v5->v6:
>>    - Place the profile in netdevice to bypass the driver.
>>      The interaction code of ethtool <-> kernel has not changed at all,
>>      only the interaction part of kernel <-> driver has changed.
>>
>> v4->v5:
>>    - Update some snippets from Kuba, Thanks.
>>
>> v3->v4:
>>    - Some tiny updates and patch 1 only add a new comment.
>>
>> v2->v3:
>>    - Break up the attributes to avoid the use of raw c structs.
>>    - Use per-device profile instead of global profile in the driver.
>>
>> v1->v2:
>>    - Use ethtool tool instead of net-sysfs
>>
>> Heng Qi (4):
>>    linux/dim: move useful macros to .h file
>>    ethtool: provide customized dim profile management
>>    virtio-net: refactor dim initialization/destruction
>>    virtio-net: support dim profile fine-tuning
>>
>> Heng Qi (4):
>>    linux/dim: move useful macros to .h file
>>    ethtool: provide customized dim profile management
>>    virtio-net: refactor dim initialization/destruction
>>    virtio-net: support dim profile fine-tuning
>>
>>   Documentation/netlink/specs/ethtool.yaml     |  33 +++
>>   Documentation/networking/ethtool-netlink.rst |   8 +
>>   drivers/net/virtio_net.c                     |  46 +++--
>>   include/linux/dim.h                          |  13 ++
>>   include/linux/ethtool.h                      |  11 +-
>>   include/linux/netdevice.h                    |  24 +++
>>   include/uapi/linux/ethtool_netlink.h         |  24 +++
>>   lib/dim/net_dim.c                            |  10 +-
>>   net/core/dev.c                               |  83 ++++++++
>>   net/ethtool/coalesce.c                       | 201 ++++++++++++++++++-
>>   10 files changed, 430 insertions(+), 23 deletions(-)
>>
>> -- 
>> 2.32.0.3.g01195cf9f

^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2024-04-24 14:26 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-04-15 13:38 [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel Heng Qi
2024-04-15 13:38 ` [PATCH RESEND net-next v7 1/4] linux/dim: move useful macros to .h file Heng Qi
2024-04-15 13:38 ` [PATCH RESEND net-next v7 2/4] ethtool: provide customized dim profile management Heng Qi
2024-04-15 13:38 ` [PATCH RESEND net-next v7 3/4] virtio-net: refactor dim initialization/destruction Heng Qi
2024-04-15 13:38 ` [PATCH RESEND net-next v7 4/4] virtio-net: support dim profile fine-tuning Heng Qi
2024-04-22 20:55 ` [PATCH RESEND net-next v7 0/4] ethtool: provide the dim profile fine-tuning channel Michael S. Tsirkin
2024-04-24 14:26   ` Heng Qi

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).