* [PATCH V2 net-next 00/11] net: hns3: updates for -next
@ 2020-11-09 3:22 Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 01/11] net: hns3: add support for configuring interrupt quantity limiting Huazhong Tan
` (10 more replies)
0 siblings, 11 replies; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
There are several updates relating to the interrupt coalesce for
the HNS3 ethernet driver.
#1 adds support for QL(quantity limiting, interrupt coalesce
based on the frame quantity).
#2 queries the maximum value of GL from the firmware instead of
a fixed value in code.
#3 adds support for 1us unit GL(gap limiting, interrupt coalesce
based on the gap time).
#4 renames gl_adapt_enable in struct hns3_enet_coalesce to fit
its new usage.
#5 & #6 adds support for the dynamic interrupt moderation,
and adds a control private flag in ethtool.
#7 adds wrapper function for state initialization.
#8 adds a check for the read-only private flag.
#9 & #10 adds support for EQ/CQ configuration, and adds a control
private flag in ethtool.
#11 adds debugfs support for interrupt coalesce.
change log:
V2 - reorder #2 & #3 to fix compiler error.
fix some checkpatch warnings in #10 & #11.
previous version:
V1: https://patchwork.ozlabs.org/project/netdev/cover/1604730681-32559-1-git-send-email-tanhuazhong@huawei.com/
Huazhong Tan (11):
net: hns3: add support for configuring interrupt quantity limiting
net: hns3: add support for querying maximum value of GL
net: hns3: add support for 1us unit GL configuration
net: hns3: rename gl_adapt_enable in struct hns3_enet_coalesce
net: hns3: add support for dynamic interrupt moderation
net: hns3: add ethtool priv-flag for DIM
net: hns3: add hns3_state_init() to do state initialization
net: hns3: add a check for ethtool priv-flag interface
net: hns3: add support for EQ/CQ mode configuration
net: hns3: add ethtool priv-flag for EQ/CQ
net: hns3: add debugfs support for interrupt coalesce
drivers/net/ethernet/hisilicon/Kconfig | 1 +
drivers/net/ethernet/hisilicon/hns3/hnae3.h | 12 +
drivers/net/ethernet/hisilicon/hns3/hns3_debugfs.c | 127 ++++++++++
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c | 258 ++++++++++++++++++---
drivers/net/ethernet/hisilicon/hns3/hns3_enet.h | 31 ++-
drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c | 185 ++++++++++++++-
.../net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h | 8 +
.../ethernet/hisilicon/hns3/hns3pf/hclge_main.c | 8 +
.../ethernet/hisilicon/hns3/hns3vf/hclgevf_cmd.h | 8 +
.../ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c | 8 +
10 files changed, 607 insertions(+), 39 deletions(-)
--
2.7.4
^ permalink raw reply [flat|nested] 22+ messages in thread
* [PATCH V2 net-next 01/11] net: hns3: add support for configuring interrupt quantity limiting
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
@ 2020-11-09 3:22 ` Huazhong Tan
2020-11-11 1:13 ` Jakub Kicinski
2020-11-09 3:22 ` [PATCH V2 net-next 02/11] net: hns3: add support for querying maximum value of GL Huazhong Tan
` (9 subsequent siblings)
10 siblings, 1 reply; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
QL(quantity limiting) means that hardware supports the interrupt
coalesce based on the frame quantity. QL can be configured when
int_ql_max in device's specification is non-zero, so add support
to configure it. Also, rename two coalesce init function to fit
their purpose.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
---
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c | 65 ++++++++++++++++------
drivers/net/ethernet/hisilicon/hns3/hns3_enet.h | 13 ++++-
drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c | 43 +++++++++++++-
.../ethernet/hisilicon/hns3/hns3pf/hclge_main.c | 1 +
.../ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c | 1 +
5 files changed, 105 insertions(+), 18 deletions(-)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index a362516..6e08719 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -237,35 +237,68 @@ void hns3_set_vector_coalesce_tx_gl(struct hns3_enet_tqp_vector *tqp_vector,
writel(tx_gl_reg, tqp_vector->mask_addr + HNS3_VECTOR_GL1_OFFSET);
}
-static void hns3_vector_gl_rl_init(struct hns3_enet_tqp_vector *tqp_vector,
- struct hns3_nic_priv *priv)
+void hns3_set_vector_coalesce_tx_ql(struct hns3_enet_tqp_vector *tqp_vector,
+ u32 ql_value)
{
+ writel(ql_value, tqp_vector->mask_addr + HNS3_VECTOR_TX_QL_OFFSET);
+}
+
+void hns3_set_vector_coalesce_rx_ql(struct hns3_enet_tqp_vector *tqp_vector,
+ u32 ql_value)
+{
+ writel(ql_value, tqp_vector->mask_addr + HNS3_VECTOR_RX_QL_OFFSET);
+}
+
+static void hns3_vector_coalesce_init(struct hns3_enet_tqp_vector *tqp_vector,
+ struct hns3_nic_priv *priv)
+{
+ struct hnae3_ae_dev *ae_dev = pci_get_drvdata(priv->ae_handle->pdev);
+ struct hns3_enet_coalesce *tx_coal = &tqp_vector->tx_group.coal;
+ struct hns3_enet_coalesce *rx_coal = &tqp_vector->rx_group.coal;
+
/* initialize the configuration for interrupt coalescing.
* 1. GL (Interrupt Gap Limiter)
* 2. RL (Interrupt Rate Limiter)
+ * 3. QL (Interrupt Quantity Limiter)
*
* Default: enable interrupt coalescing self-adaptive and GL
*/
- tqp_vector->tx_group.coal.gl_adapt_enable = 1;
- tqp_vector->rx_group.coal.gl_adapt_enable = 1;
+ tx_coal->gl_adapt_enable = 1;
+ rx_coal->gl_adapt_enable = 1;
+
+ tx_coal->int_gl = HNS3_INT_GL_50K;
+ rx_coal->int_gl = HNS3_INT_GL_50K;
- tqp_vector->tx_group.coal.int_gl = HNS3_INT_GL_50K;
- tqp_vector->rx_group.coal.int_gl = HNS3_INT_GL_50K;
+ rx_coal->flow_level = HNS3_FLOW_LOW;
+ tx_coal->flow_level = HNS3_FLOW_LOW;
- tqp_vector->rx_group.coal.flow_level = HNS3_FLOW_LOW;
- tqp_vector->tx_group.coal.flow_level = HNS3_FLOW_LOW;
+ if (ae_dev->dev_specs.int_ql_max) {
+ tx_coal->ql_enable = 1;
+ rx_coal->ql_enable = 1;
+ tx_coal->int_ql_max = ae_dev->dev_specs.int_ql_max;
+ rx_coal->int_ql_max = ae_dev->dev_specs.int_ql_max;
+ tx_coal->int_ql = HNS3_INT_QL_DEFAULT_CFG;
+ rx_coal->int_ql = HNS3_INT_QL_DEFAULT_CFG;
+ }
}
-static void hns3_vector_gl_rl_init_hw(struct hns3_enet_tqp_vector *tqp_vector,
- struct hns3_nic_priv *priv)
+static void
+hns3_vector_coalesce_init_hw(struct hns3_enet_tqp_vector *tqp_vector,
+ struct hns3_nic_priv *priv)
{
+ struct hns3_enet_coalesce *tx_coal = &tqp_vector->tx_group.coal;
+ struct hns3_enet_coalesce *rx_coal = &tqp_vector->rx_group.coal;
struct hnae3_handle *h = priv->ae_handle;
- hns3_set_vector_coalesce_tx_gl(tqp_vector,
- tqp_vector->tx_group.coal.int_gl);
- hns3_set_vector_coalesce_rx_gl(tqp_vector,
- tqp_vector->rx_group.coal.int_gl);
+ hns3_set_vector_coalesce_tx_gl(tqp_vector, tx_coal->int_gl);
+ hns3_set_vector_coalesce_rx_gl(tqp_vector, rx_coal->int_gl);
hns3_set_vector_coalesce_rl(tqp_vector, h->kinfo.int_rl_setting);
+
+ if (tx_coal->ql_enable)
+ hns3_set_vector_coalesce_tx_ql(tqp_vector, tx_coal->int_ql);
+
+ if (rx_coal->ql_enable)
+ hns3_set_vector_coalesce_rx_ql(tqp_vector, rx_coal->int_ql);
}
static int hns3_nic_set_real_num_queue(struct net_device *netdev)
@@ -3536,7 +3569,7 @@ static int hns3_nic_init_vector_data(struct hns3_nic_priv *priv)
for (i = 0; i < priv->vector_num; i++) {
tqp_vector = &priv->tqp_vector[i];
- hns3_vector_gl_rl_init_hw(tqp_vector, priv);
+ hns3_vector_coalesce_init_hw(tqp_vector, priv);
tqp_vector->num_tqps = 0;
}
@@ -3632,7 +3665,7 @@ static int hns3_nic_alloc_vector_data(struct hns3_nic_priv *priv)
tqp_vector->idx = i;
tqp_vector->mask_addr = vector[i].io_addr;
tqp_vector->vector_irq = vector[i].vector;
- hns3_vector_gl_rl_init(tqp_vector, priv);
+ hns3_vector_coalesce_init(tqp_vector, priv);
}
out:
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
index 1c81dea..10990bd 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
@@ -181,6 +181,8 @@ enum hns3_nic_state {
#define HNS3_VECTOR_GL2_OFFSET 0x300
#define HNS3_VECTOR_RL_OFFSET 0x900
#define HNS3_VECTOR_RL_EN_B 6
+#define HNS3_VECTOR_TX_QL_OFFSET 0xe00
+#define HNS3_VECTOR_RX_QL_OFFSET 0xf00
#define HNS3_RING_EN_B 0
@@ -427,9 +429,14 @@ enum hns3_flow_level_range {
#define HNS3_INT_RL_MAX 0x00EC
#define HNS3_INT_RL_ENABLE_MASK 0x40
+#define HNS3_INT_QL_DEFAULT_CFG 0x20
+
struct hns3_enet_coalesce {
u16 int_gl;
- u8 gl_adapt_enable;
+ u16 int_ql;
+ u16 int_ql_max;
+ u8 gl_adapt_enable:1;
+ u8 ql_enable:1;
enum hns3_flow_level_range flow_level;
};
@@ -595,6 +602,10 @@ void hns3_set_vector_coalesce_tx_gl(struct hns3_enet_tqp_vector *tqp_vector,
u32 gl_value);
void hns3_set_vector_coalesce_rl(struct hns3_enet_tqp_vector *tqp_vector,
u32 rl_value);
+void hns3_set_vector_coalesce_rx_ql(struct hns3_enet_tqp_vector *tqp_vector,
+ u32 ql_value);
+void hns3_set_vector_coalesce_tx_ql(struct hns3_enet_tqp_vector *tqp_vector,
+ u32 ql_value);
void hns3_enable_vlan_filter(struct net_device *netdev, bool enable);
void hns3_request_update_promisc_mode(struct hnae3_handle *handle);
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
index 6b07b27..9af7cb9 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
@@ -1115,6 +1115,9 @@ static int hns3_get_coalesce_per_queue(struct net_device *netdev, u32 queue,
cmd->tx_coalesce_usecs_high = h->kinfo.int_rl_setting;
cmd->rx_coalesce_usecs_high = h->kinfo.int_rl_setting;
+ cmd->tx_max_coalesced_frames = tx_vector->tx_group.coal.int_ql;
+ cmd->rx_max_coalesced_frames = rx_vector->rx_group.coal.int_ql;
+
return 0;
}
@@ -1188,6 +1191,29 @@ static int hns3_check_rl_coalesce_para(struct net_device *netdev,
return 0;
}
+static int hns3_check_ql_coalesce_param(struct net_device *netdev,
+ struct ethtool_coalesce *cmd)
+{
+ struct hnae3_handle *handle = hns3_get_handle(netdev);
+ struct hnae3_ae_dev *ae_dev = pci_get_drvdata(handle->pdev);
+
+ if ((cmd->tx_max_coalesced_frames || cmd->rx_max_coalesced_frames) &&
+ !ae_dev->dev_specs.int_ql_max) {
+ netdev_err(netdev, "coalesced frames is not supported\n");
+ return -EOPNOTSUPP;
+ }
+
+ if (cmd->tx_max_coalesced_frames > ae_dev->dev_specs.int_ql_max ||
+ cmd->rx_max_coalesced_frames > ae_dev->dev_specs.int_ql_max) {
+ netdev_err(netdev,
+ "invalid coalesced_frames value, range is 0-%u\n",
+ ae_dev->dev_specs.int_ql_max);
+ return -ERANGE;
+ }
+
+ return 0;
+}
+
static int hns3_check_coalesce_para(struct net_device *netdev,
struct ethtool_coalesce *cmd)
{
@@ -1207,6 +1233,10 @@ static int hns3_check_coalesce_para(struct net_device *netdev,
return ret;
}
+ ret = hns3_check_ql_coalesce_param(netdev, cmd);
+ if (ret)
+ return ret;
+
if (cmd->use_adaptive_tx_coalesce == 1 ||
cmd->use_adaptive_rx_coalesce == 1) {
netdev_info(netdev,
@@ -1238,6 +1268,9 @@ static void hns3_set_coalesce_per_queue(struct net_device *netdev,
tx_vector->tx_group.coal.int_gl = cmd->tx_coalesce_usecs;
rx_vector->rx_group.coal.int_gl = cmd->rx_coalesce_usecs;
+ tx_vector->tx_group.coal.int_ql = cmd->tx_max_coalesced_frames;
+ rx_vector->rx_group.coal.int_ql = cmd->rx_max_coalesced_frames;
+
hns3_set_vector_coalesce_tx_gl(tx_vector,
tx_vector->tx_group.coal.int_gl);
hns3_set_vector_coalesce_rx_gl(rx_vector,
@@ -1245,6 +1278,13 @@ static void hns3_set_coalesce_per_queue(struct net_device *netdev,
hns3_set_vector_coalesce_rl(tx_vector, h->kinfo.int_rl_setting);
hns3_set_vector_coalesce_rl(rx_vector, h->kinfo.int_rl_setting);
+
+ if (tx_vector->tx_group.coal.ql_enable)
+ hns3_set_vector_coalesce_tx_ql(tx_vector,
+ tx_vector->tx_group.coal.int_ql);
+ if (rx_vector->tx_group.coal.ql_enable)
+ hns3_set_vector_coalesce_rx_ql(rx_vector,
+ rx_vector->rx_group.coal.int_ql);
}
static int hns3_set_coalesce(struct net_device *netdev,
@@ -1471,7 +1511,8 @@ static int hns3_get_module_eeprom(struct net_device *netdev,
#define HNS3_ETHTOOL_COALESCE (ETHTOOL_COALESCE_USECS | \
ETHTOOL_COALESCE_USE_ADAPTIVE | \
ETHTOOL_COALESCE_RX_USECS_HIGH | \
- ETHTOOL_COALESCE_TX_USECS_HIGH)
+ ETHTOOL_COALESCE_TX_USECS_HIGH | \
+ ETHTOOL_COALESCE_MAX_FRAMES)
static const struct ethtool_ops hns3vf_ethtool_ops = {
.supported_coalesce_params = HNS3_ETHTOOL_COALESCE,
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
index 1f02640..8bcdb28 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
@@ -1379,6 +1379,7 @@ static void hclge_parse_dev_specs(struct hclge_dev *hdev,
ae_dev->dev_specs.max_non_tso_bd_num = req0->max_non_tso_bd_num;
ae_dev->dev_specs.rss_ind_tbl_size =
le16_to_cpu(req0->rss_ind_tbl_size);
+ ae_dev->dev_specs.int_ql_max = le16_to_cpu(req0->int_ql_max);
ae_dev->dev_specs.rss_key_size = le16_to_cpu(req0->rss_key_size);
ae_dev->dev_specs.max_tm_rate = le32_to_cpu(req0->max_tm_rate);
}
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
index c8e3fdd..8209be9 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
@@ -3004,6 +3004,7 @@ static void hclgevf_parse_dev_specs(struct hclgevf_dev *hdev,
ae_dev->dev_specs.max_non_tso_bd_num = req0->max_non_tso_bd_num;
ae_dev->dev_specs.rss_ind_tbl_size =
le16_to_cpu(req0->rss_ind_tbl_size);
+ ae_dev->dev_specs.int_ql_max = le16_to_cpu(req0->int_ql_max);
ae_dev->dev_specs.rss_key_size = le16_to_cpu(req0->rss_key_size);
}
--
2.7.4
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH V2 net-next 02/11] net: hns3: add support for querying maximum value of GL
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 01/11] net: hns3: add support for configuring interrupt quantity limiting Huazhong Tan
@ 2020-11-09 3:22 ` Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 03/11] net: hns3: add support for 1us unit GL configuration Huazhong Tan
` (8 subsequent siblings)
10 siblings, 0 replies; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
For maintainability and compatibility, add support for querying
the maximum value of GL.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
---
drivers/net/ethernet/hisilicon/hns3/hnae3.h | 1 +
drivers/net/ethernet/hisilicon/hns3/hns3_debugfs.c | 1 +
drivers/net/ethernet/hisilicon/hns3/hns3_enet.h | 1 -
drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c | 14 ++++++++------
drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h | 8 ++++++++
drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c | 6 ++++++
drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_cmd.h | 8 ++++++++
drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c | 6 ++++++
8 files changed, 38 insertions(+), 7 deletions(-)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hnae3.h b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
index 912c51e..f9d4d23 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hnae3.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
@@ -278,6 +278,7 @@ struct hnae3_dev_specs {
u16 rss_ind_tbl_size;
u16 rss_key_size;
u16 int_ql_max; /* max value of interrupt coalesce based on INT_QL */
+ u16 max_int_gl; /* max value of interrupt coalesce based on INT_GL */
u8 max_non_tso_bd_num; /* max BD number of one non-TSO packet */
};
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_debugfs.c b/drivers/net/ethernet/hisilicon/hns3/hns3_debugfs.c
index dc9a857..a5ebca8 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_debugfs.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_debugfs.c
@@ -349,6 +349,7 @@ static void hns3_dbg_dev_specs(struct hnae3_handle *h)
dev_info(priv->dev, "Desc num per RX queue: %u\n", kinfo->num_rx_desc);
dev_info(priv->dev, "Total number of enabled TCs: %u\n", kinfo->num_tc);
dev_info(priv->dev, "MAX INT QL: %u\n", dev_specs->int_ql_max);
+ dev_info(priv->dev, "MAX INT GL: %u\n", dev_specs->max_int_gl);
}
static ssize_t hns3_dbg_cmd_read(struct file *filp, char __user *buffer,
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
index 10990bd..be099dd 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
@@ -420,7 +420,6 @@ enum hns3_flow_level_range {
HNS3_FLOW_ULTRA = 3,
};
-#define HNS3_INT_GL_MAX 0x1FE0
#define HNS3_INT_GL_50K 0x0014
#define HNS3_INT_GL_20K 0x0032
#define HNS3_INT_GL_18K 0x0036
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
index 9af7cb9..9e90d67 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
@@ -1130,19 +1130,21 @@ static int hns3_get_coalesce(struct net_device *netdev,
static int hns3_check_gl_coalesce_para(struct net_device *netdev,
struct ethtool_coalesce *cmd)
{
+ struct hnae3_handle *handle = hns3_get_handle(netdev);
+ struct hnae3_ae_dev *ae_dev = pci_get_drvdata(handle->pdev);
u32 rx_gl, tx_gl;
- if (cmd->rx_coalesce_usecs > HNS3_INT_GL_MAX) {
+ if (cmd->rx_coalesce_usecs > ae_dev->dev_specs.max_int_gl) {
netdev_err(netdev,
- "Invalid rx-usecs value, rx-usecs range is 0-%d\n",
- HNS3_INT_GL_MAX);
+ "invalid rx-usecs value, rx-usecs range is 0-%u\n",
+ ae_dev->dev_specs.max_int_gl);
return -EINVAL;
}
- if (cmd->tx_coalesce_usecs > HNS3_INT_GL_MAX) {
+ if (cmd->tx_coalesce_usecs > ae_dev->dev_specs.max_int_gl) {
netdev_err(netdev,
- "Invalid tx-usecs value, tx-usecs range is 0-%d\n",
- HNS3_INT_GL_MAX);
+ "invalid tx-usecs value, tx-usecs range is 0-%u\n",
+ ae_dev->dev_specs.max_int_gl);
return -EINVAL;
}
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h
index 096e26a..5b7967c 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h
@@ -1103,6 +1103,14 @@ struct hclge_dev_specs_0_cmd {
__le32 max_tm_rate;
};
+#define HCLGE_DEF_MAX_INT_GL 0x1FE0U
+
+struct hclge_dev_specs_1_cmd {
+ __le32 rsv0;
+ __le16 max_int_gl;
+ u8 rsv1[18];
+};
+
int hclge_cmd_init(struct hclge_dev *hdev);
static inline void hclge_write_reg(void __iomem *base, u32 reg, u32 value)
{
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
index 8bcdb28..7102001 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
@@ -1366,6 +1366,7 @@ static void hclge_set_default_dev_specs(struct hclge_dev *hdev)
ae_dev->dev_specs.rss_ind_tbl_size = HCLGE_RSS_IND_TBL_SIZE;
ae_dev->dev_specs.rss_key_size = HCLGE_RSS_KEY_SIZE;
ae_dev->dev_specs.max_tm_rate = HCLGE_ETHER_MAX_RATE;
+ ae_dev->dev_specs.max_int_gl = HCLGE_DEF_MAX_INT_GL;
}
static void hclge_parse_dev_specs(struct hclge_dev *hdev,
@@ -1373,8 +1374,10 @@ static void hclge_parse_dev_specs(struct hclge_dev *hdev,
{
struct hnae3_ae_dev *ae_dev = pci_get_drvdata(hdev->pdev);
struct hclge_dev_specs_0_cmd *req0;
+ struct hclge_dev_specs_1_cmd *req1;
req0 = (struct hclge_dev_specs_0_cmd *)desc[0].data;
+ req1 = (struct hclge_dev_specs_1_cmd *)desc[1].data;
ae_dev->dev_specs.max_non_tso_bd_num = req0->max_non_tso_bd_num;
ae_dev->dev_specs.rss_ind_tbl_size =
@@ -1382,6 +1385,7 @@ static void hclge_parse_dev_specs(struct hclge_dev *hdev,
ae_dev->dev_specs.int_ql_max = le16_to_cpu(req0->int_ql_max);
ae_dev->dev_specs.rss_key_size = le16_to_cpu(req0->rss_key_size);
ae_dev->dev_specs.max_tm_rate = le32_to_cpu(req0->max_tm_rate);
+ ae_dev->dev_specs.max_int_gl = le16_to_cpu(req1->max_int_gl);
}
static void hclge_check_dev_specs(struct hclge_dev *hdev)
@@ -1396,6 +1400,8 @@ static void hclge_check_dev_specs(struct hclge_dev *hdev)
dev_specs->rss_key_size = HCLGE_RSS_KEY_SIZE;
if (!dev_specs->max_tm_rate)
dev_specs->max_tm_rate = HCLGE_ETHER_MAX_RATE;
+ if (!dev_specs->max_int_gl)
+ dev_specs->max_int_gl = HCLGE_DEF_MAX_INT_GL;
}
static int hclge_query_dev_specs(struct hclge_dev *hdev)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_cmd.h b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_cmd.h
index 9460c12..f94f5d4 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_cmd.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_cmd.h
@@ -285,6 +285,14 @@ struct hclgevf_dev_specs_0_cmd {
u8 rsv1[5];
};
+#define HCLGEVF_DEF_MAX_INT_GL 0x1FE0U
+
+struct hclgevf_dev_specs_1_cmd {
+ __le32 rsv0;
+ __le16 max_int_gl;
+ u8 rsv1[18];
+};
+
static inline void hclgevf_write_reg(void __iomem *base, u32 reg, u32 value)
{
writel(value, base + reg);
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
index 8209be9..71007e7 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
@@ -2991,6 +2991,7 @@ static void hclgevf_set_default_dev_specs(struct hclgevf_dev *hdev)
HCLGEVF_MAX_NON_TSO_BD_NUM;
ae_dev->dev_specs.rss_ind_tbl_size = HCLGEVF_RSS_IND_TBL_SIZE;
ae_dev->dev_specs.rss_key_size = HCLGEVF_RSS_KEY_SIZE;
+ ae_dev->dev_specs.max_int_gl = HCLGEVF_DEF_MAX_INT_GL;
}
static void hclgevf_parse_dev_specs(struct hclgevf_dev *hdev,
@@ -2998,14 +2999,17 @@ static void hclgevf_parse_dev_specs(struct hclgevf_dev *hdev,
{
struct hnae3_ae_dev *ae_dev = pci_get_drvdata(hdev->pdev);
struct hclgevf_dev_specs_0_cmd *req0;
+ struct hclgevf_dev_specs_1_cmd *req1;
req0 = (struct hclgevf_dev_specs_0_cmd *)desc[0].data;
+ req1 = (struct hclgevf_dev_specs_1_cmd *)desc[1].data;
ae_dev->dev_specs.max_non_tso_bd_num = req0->max_non_tso_bd_num;
ae_dev->dev_specs.rss_ind_tbl_size =
le16_to_cpu(req0->rss_ind_tbl_size);
ae_dev->dev_specs.int_ql_max = le16_to_cpu(req0->int_ql_max);
ae_dev->dev_specs.rss_key_size = le16_to_cpu(req0->rss_key_size);
+ ae_dev->dev_specs.max_int_gl = le16_to_cpu(req1->max_int_gl);
}
static void hclgevf_check_dev_specs(struct hclgevf_dev *hdev)
@@ -3018,6 +3022,8 @@ static void hclgevf_check_dev_specs(struct hclgevf_dev *hdev)
dev_specs->rss_ind_tbl_size = HCLGEVF_RSS_IND_TBL_SIZE;
if (!dev_specs->rss_key_size)
dev_specs->rss_key_size = HCLGEVF_RSS_KEY_SIZE;
+ if (!dev_specs->max_int_gl)
+ dev_specs->max_int_gl = HCLGEVF_DEF_MAX_INT_GL;
}
static int hclgevf_query_dev_specs(struct hclgevf_dev *hdev)
--
2.7.4
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH V2 net-next 03/11] net: hns3: add support for 1us unit GL configuration
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 01/11] net: hns3: add support for configuring interrupt quantity limiting Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 02/11] net: hns3: add support for querying maximum value of GL Huazhong Tan
@ 2020-11-09 3:22 ` Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 04/11] net: hns3: rename gl_adapt_enable in struct hns3_enet_coalesce Huazhong Tan
` (7 subsequent siblings)
10 siblings, 0 replies; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
For device whose version is above V3(include V3), the GL
configuration can set as 1us unit, so adds support for
configuring this field.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
---
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c | 26 ++++++++++++++++++----
drivers/net/ethernet/hisilicon/hns3/hns3_enet.h | 3 +++
drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c | 6 +++++
3 files changed, 31 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index 6e08719..2813fe5 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -224,17 +224,27 @@ void hns3_set_vector_coalesce_rl(struct hns3_enet_tqp_vector *tqp_vector,
void hns3_set_vector_coalesce_rx_gl(struct hns3_enet_tqp_vector *tqp_vector,
u32 gl_value)
{
- u32 rx_gl_reg = hns3_gl_usec_to_reg(gl_value);
+ u32 new_val;
- writel(rx_gl_reg, tqp_vector->mask_addr + HNS3_VECTOR_GL0_OFFSET);
+ if (tqp_vector->rx_group.coal.unit_1us)
+ new_val = gl_value | HNS3_INT_GL_1US;
+ else
+ new_val = hns3_gl_usec_to_reg(gl_value);
+
+ writel(new_val, tqp_vector->mask_addr + HNS3_VECTOR_GL0_OFFSET);
}
void hns3_set_vector_coalesce_tx_gl(struct hns3_enet_tqp_vector *tqp_vector,
u32 gl_value)
{
- u32 tx_gl_reg = hns3_gl_usec_to_reg(gl_value);
+ u32 new_val;
+
+ if (tqp_vector->tx_group.coal.unit_1us)
+ new_val = gl_value | HNS3_INT_GL_1US;
+ else
+ new_val = hns3_gl_usec_to_reg(gl_value);
- writel(tx_gl_reg, tqp_vector->mask_addr + HNS3_VECTOR_GL1_OFFSET);
+ writel(new_val, tqp_vector->mask_addr + HNS3_VECTOR_GL1_OFFSET);
}
void hns3_set_vector_coalesce_tx_ql(struct hns3_enet_tqp_vector *tqp_vector,
@@ -272,6 +282,14 @@ static void hns3_vector_coalesce_init(struct hns3_enet_tqp_vector *tqp_vector,
rx_coal->flow_level = HNS3_FLOW_LOW;
tx_coal->flow_level = HNS3_FLOW_LOW;
+ /* device version above V3(include V3), GL can configure 1us
+ * unit, so uses 1us unit.
+ */
+ if (ae_dev->dev_version >= HNAE3_DEVICE_VERSION_V3) {
+ tx_coal->unit_1us = 1;
+ rx_coal->unit_1us = 1;
+ }
+
if (ae_dev->dev_specs.int_ql_max) {
tx_coal->ql_enable = 1;
rx_coal->ql_enable = 1;
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
index be099dd..4651ad1 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
@@ -425,6 +425,8 @@ enum hns3_flow_level_range {
#define HNS3_INT_GL_18K 0x0036
#define HNS3_INT_GL_8K 0x007C
+#define HNS3_INT_GL_1US BIT(31)
+
#define HNS3_INT_RL_MAX 0x00EC
#define HNS3_INT_RL_ENABLE_MASK 0x40
@@ -436,6 +438,7 @@ struct hns3_enet_coalesce {
u16 int_ql_max;
u8 gl_adapt_enable:1;
u8 ql_enable:1;
+ u8 unit_1us:1;
enum hns3_flow_level_range flow_level;
};
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
index 9e90d67..8d5c194 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
@@ -1148,6 +1148,12 @@ static int hns3_check_gl_coalesce_para(struct net_device *netdev,
return -EINVAL;
}
+ /* device version above V3(include V3), GL uses 1us unit,
+ * so the round down is not needed.
+ */
+ if (ae_dev->dev_version >= HNAE3_DEVICE_VERSION_V3)
+ return 0;
+
rx_gl = hns3_gl_round_down(cmd->rx_coalesce_usecs);
if (rx_gl != cmd->rx_coalesce_usecs) {
netdev_info(netdev,
--
2.7.4
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH V2 net-next 04/11] net: hns3: rename gl_adapt_enable in struct hns3_enet_coalesce
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
` (2 preceding siblings ...)
2020-11-09 3:22 ` [PATCH V2 net-next 03/11] net: hns3: add support for 1us unit GL configuration Huazhong Tan
@ 2020-11-09 3:22 ` Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 05/11] net: hns3: add support for dynamic interrupt moderation Huazhong Tan
` (6 subsequent siblings)
10 siblings, 0 replies; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
Besides GL(Gap Limiting), QL(Quantity Limiting) can be modified
dynamically when DIM is supported. So rename gl_adapt_enable as
adapt_enable in struct hns3_enet_coalesce.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
---
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c | 12 ++++++------
drivers/net/ethernet/hisilicon/hns3/hns3_enet.h | 2 +-
drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c | 8 ++++----
3 files changed, 11 insertions(+), 11 deletions(-)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index 2813fe5..999a2aa 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -211,8 +211,8 @@ void hns3_set_vector_coalesce_rl(struct hns3_enet_tqp_vector *tqp_vector,
* GL and RL(Rate Limiter) are 2 ways to acheive interrupt coalescing
*/
- if (rl_reg > 0 && !tqp_vector->tx_group.coal.gl_adapt_enable &&
- !tqp_vector->rx_group.coal.gl_adapt_enable)
+ if (rl_reg > 0 && !tqp_vector->tx_group.coal.adapt_enable &&
+ !tqp_vector->rx_group.coal.adapt_enable)
/* According to the hardware, the range of rl_reg is
* 0-59 and the unit is 4.
*/
@@ -273,8 +273,8 @@ static void hns3_vector_coalesce_init(struct hns3_enet_tqp_vector *tqp_vector,
*
* Default: enable interrupt coalescing self-adaptive and GL
*/
- tx_coal->gl_adapt_enable = 1;
- rx_coal->gl_adapt_enable = 1;
+ tx_coal->adapt_enable = 1;
+ rx_coal->adapt_enable = 1;
tx_coal->int_gl = HNS3_INT_GL_50K;
rx_coal->int_gl = HNS3_INT_GL_50K;
@@ -3384,14 +3384,14 @@ static void hns3_update_new_int_gl(struct hns3_enet_tqp_vector *tqp_vector)
tqp_vector->last_jiffies + msecs_to_jiffies(1000)))
return;
- if (rx_group->coal.gl_adapt_enable) {
+ if (rx_group->coal.adapt_enable) {
rx_update = hns3_get_new_int_gl(rx_group);
if (rx_update)
hns3_set_vector_coalesce_rx_gl(tqp_vector,
rx_group->coal.int_gl);
}
- if (tx_group->coal.gl_adapt_enable) {
+ if (tx_group->coal.adapt_enable) {
tx_update = hns3_get_new_int_gl(tx_group);
if (tx_update)
hns3_set_vector_coalesce_tx_gl(tqp_vector,
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
index 4651ad1..8d33652 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
@@ -436,7 +436,7 @@ struct hns3_enet_coalesce {
u16 int_gl;
u16 int_ql;
u16 int_ql_max;
- u8 gl_adapt_enable:1;
+ u8 adapt_enable:1;
u8 ql_enable:1;
u8 unit_1us:1;
enum hns3_flow_level_range flow_level;
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
index 8d5c194..30ffaaf 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
@@ -1105,9 +1105,9 @@ static int hns3_get_coalesce_per_queue(struct net_device *netdev, u32 queue,
rx_vector = priv->ring[queue_num + queue].tqp_vector;
cmd->use_adaptive_tx_coalesce =
- tx_vector->tx_group.coal.gl_adapt_enable;
+ tx_vector->tx_group.coal.adapt_enable;
cmd->use_adaptive_rx_coalesce =
- rx_vector->rx_group.coal.gl_adapt_enable;
+ rx_vector->rx_group.coal.adapt_enable;
cmd->tx_coalesce_usecs = tx_vector->tx_group.coal.int_gl;
cmd->rx_coalesce_usecs = rx_vector->rx_group.coal.int_gl;
@@ -1268,9 +1268,9 @@ static void hns3_set_coalesce_per_queue(struct net_device *netdev,
tx_vector = priv->ring[queue].tqp_vector;
rx_vector = priv->ring[queue_num + queue].tqp_vector;
- tx_vector->tx_group.coal.gl_adapt_enable =
+ tx_vector->tx_group.coal.adapt_enable =
cmd->use_adaptive_tx_coalesce;
- rx_vector->rx_group.coal.gl_adapt_enable =
+ rx_vector->rx_group.coal.adapt_enable =
cmd->use_adaptive_rx_coalesce;
tx_vector->tx_group.coal.int_gl = cmd->tx_coalesce_usecs;
--
2.7.4
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH V2 net-next 05/11] net: hns3: add support for dynamic interrupt moderation
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
` (3 preceding siblings ...)
2020-11-09 3:22 ` [PATCH V2 net-next 04/11] net: hns3: rename gl_adapt_enable in struct hns3_enet_coalesce Huazhong Tan
@ 2020-11-09 3:22 ` Huazhong Tan
2020-11-11 1:20 ` Jakub Kicinski
2020-11-09 3:22 ` [PATCH V2 net-next 06/11] net: hns3: add ethtool priv-flag for DIM Huazhong Tan
` (5 subsequent siblings)
10 siblings, 1 reply; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
Add dynamic interrupt moderation support for the HNS3 driver.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
---
drivers/net/ethernet/hisilicon/Kconfig | 1 +
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c | 87 ++++++++++++++++++++++++-
drivers/net/ethernet/hisilicon/hns3/hns3_enet.h | 4 ++
3 files changed, 91 insertions(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/hisilicon/Kconfig b/drivers/net/ethernet/hisilicon/Kconfig
index 44f9279..fa6025d 100644
--- a/drivers/net/ethernet/hisilicon/Kconfig
+++ b/drivers/net/ethernet/hisilicon/Kconfig
@@ -130,6 +130,7 @@ config HNS3_ENET
default m
depends on 64BIT && PCI
depends on INET
+ select DIMLIB
help
This selects the Ethernet Driver for Hisilicon Network Subsystem 3 for hip08
family of SoCs. This module depends upon HNAE3 driver to access the HNAE3
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index 999a2aa..9e895b9 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -96,6 +96,7 @@ static irqreturn_t hns3_irq_handle(int irq, void *vector)
struct hns3_enet_tqp_vector *tqp_vector = vector;
napi_schedule_irqoff(&tqp_vector->napi);
+ tqp_vector->event_cnt++;
return IRQ_HANDLED;
}
@@ -199,6 +200,8 @@ static void hns3_vector_disable(struct hns3_enet_tqp_vector *tqp_vector)
disable_irq(tqp_vector->vector_irq);
napi_disable(&tqp_vector->napi);
+ cancel_work_sync(&tqp_vector->rx_group.dim.work);
+ cancel_work_sync(&tqp_vector->tx_group.dim.work);
}
void hns3_set_vector_coalesce_rl(struct hns3_enet_tqp_vector *tqp_vector,
@@ -3401,6 +3404,32 @@ static void hns3_update_new_int_gl(struct hns3_enet_tqp_vector *tqp_vector)
tqp_vector->last_jiffies = jiffies;
}
+static void hns3_update_rx_int_coalesce(struct hns3_enet_tqp_vector *tqp_vector)
+{
+ struct hns3_enet_ring_group *rx_group = &tqp_vector->rx_group;
+ struct dim_sample sample = {};
+
+ if (!rx_group->coal.adapt_enable)
+ return;
+
+ dim_update_sample(tqp_vector->event_cnt, rx_group->total_packets,
+ rx_group->total_bytes, &sample);
+ net_dim(&rx_group->dim, sample);
+}
+
+static void hns3_update_tx_int_coalesce(struct hns3_enet_tqp_vector *tqp_vector)
+{
+ struct hns3_enet_ring_group *tx_group = &tqp_vector->tx_group;
+ struct dim_sample sample = {};
+
+ if (!tx_group->coal.adapt_enable)
+ return;
+
+ dim_update_sample(tqp_vector->event_cnt, tx_group->total_packets,
+ tx_group->total_bytes, &sample);
+ net_dim(&tx_group->dim, sample);
+}
+
static int hns3_nic_common_poll(struct napi_struct *napi, int budget)
{
struct hns3_nic_priv *priv = netdev_priv(napi->dev);
@@ -3444,7 +3473,13 @@ static int hns3_nic_common_poll(struct napi_struct *napi, int budget)
if (napi_complete(napi) &&
likely(!test_bit(HNS3_NIC_STATE_DOWN, &priv->state))) {
- hns3_update_new_int_gl(tqp_vector);
+ if (test_bit(HNS3_NIC_STATE_DIM_ENABLE, &priv->state)) {
+ hns3_update_rx_int_coalesce(tqp_vector);
+ hns3_update_tx_int_coalesce(tqp_vector);
+ } else {
+ hns3_update_new_int_gl(tqp_vector);
+ }
+
hns3_mask_vector_irq(tqp_vector, 1);
}
@@ -3575,6 +3610,54 @@ static void hns3_nic_set_cpumask(struct hns3_nic_priv *priv)
}
}
+static void hns3_rx_dim_work(struct work_struct *work)
+{
+ struct dim *dim = container_of(work, struct dim, work);
+ struct hns3_enet_ring_group *group = container_of(dim,
+ struct hns3_enet_ring_group, dim);
+ struct hns3_enet_tqp_vector *tqp_vector = group->ring->tqp_vector;
+ struct dim_cq_moder cur_moder =
+ net_dim_get_rx_moderation(dim->mode, dim->profile_ix);
+
+ hns3_set_vector_coalesce_rx_gl(group->ring->tqp_vector, cur_moder.usec);
+ tqp_vector->rx_group.coal.int_gl = cur_moder.usec;
+
+ if (cur_moder.pkts < tqp_vector->rx_group.coal.int_ql_max) {
+ hns3_set_vector_coalesce_rx_ql(tqp_vector, cur_moder.pkts);
+ tqp_vector->rx_group.coal.int_ql = cur_moder.pkts;
+ }
+
+ dim->state = DIM_START_MEASURE;
+}
+
+static void hns3_tx_dim_work(struct work_struct *work)
+{
+ struct dim *dim = container_of(work, struct dim, work);
+ struct hns3_enet_ring_group *group = container_of(dim,
+ struct hns3_enet_ring_group, dim);
+ struct hns3_enet_tqp_vector *tqp_vector = group->ring->tqp_vector;
+ struct dim_cq_moder cur_moder =
+ net_dim_get_tx_moderation(dim->mode, dim->profile_ix);
+
+ hns3_set_vector_coalesce_tx_gl(tqp_vector, cur_moder.usec);
+ tqp_vector->tx_group.coal.int_gl = cur_moder.usec;
+
+ if (cur_moder.pkts < tqp_vector->tx_group.coal.int_ql_max) {
+ hns3_set_vector_coalesce_tx_ql(tqp_vector, cur_moder.pkts);
+ tqp_vector->tx_group.coal.int_ql = cur_moder.pkts;
+ }
+
+ dim->state = DIM_START_MEASURE;
+}
+
+static void hns3_nic_init_dim(struct hns3_enet_tqp_vector *tqp_vector)
+{
+ INIT_WORK(&tqp_vector->rx_group.dim.work, hns3_rx_dim_work);
+ tqp_vector->rx_group.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
+ INIT_WORK(&tqp_vector->tx_group.dim.work, hns3_tx_dim_work);
+ tqp_vector->tx_group.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
+}
+
static int hns3_nic_init_vector_data(struct hns3_nic_priv *priv)
{
struct hnae3_ring_chain_node vector_ring_chain;
@@ -3589,6 +3672,7 @@ static int hns3_nic_init_vector_data(struct hns3_nic_priv *priv)
tqp_vector = &priv->tqp_vector[i];
hns3_vector_coalesce_init_hw(tqp_vector, priv);
tqp_vector->num_tqps = 0;
+ hns3_nic_init_dim(tqp_vector);
}
for (i = 0; i < h->kinfo.num_tqps; i++) {
@@ -4161,6 +4245,7 @@ static int hns3_client_init(struct hnae3_handle *handle)
netdev->max_mtu = HNS3_MAX_MTU;
set_bit(HNS3_NIC_STATE_INITED, &priv->state);
+ set_bit(HNS3_NIC_STATE_DIM_ENABLE, &priv->state);
if (netif_msg_drv(handle))
hns3_info_show(priv);
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
index 8d33652..eb4e7ef 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
@@ -4,6 +4,7 @@
#ifndef __HNS3_ENET_H
#define __HNS3_ENET_H
+#include <linux/dim.h>
#include <linux/if_vlan.h>
#include "hnae3.h"
@@ -18,6 +19,7 @@ enum hns3_nic_state {
HNS3_NIC_STATE_SERVICE_INITED,
HNS3_NIC_STATE_SERVICE_SCHED,
HNS3_NIC_STATE2_RESET_REQUESTED,
+ HNS3_NIC_STATE_DIM_ENABLE,
HNS3_NIC_STATE_MAX
};
@@ -449,6 +451,7 @@ struct hns3_enet_ring_group {
u64 total_packets; /* total packets processed this group */
u16 count;
struct hns3_enet_coalesce coal;
+ struct dim dim;
};
struct hns3_enet_tqp_vector {
@@ -471,6 +474,7 @@ struct hns3_enet_tqp_vector {
char name[HNAE3_INT_NAME_LEN];
unsigned long last_jiffies;
+ u64 event_cnt;
} ____cacheline_internodealigned_in_smp;
struct hns3_nic_priv {
--
2.7.4
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH V2 net-next 06/11] net: hns3: add ethtool priv-flag for DIM
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
` (4 preceding siblings ...)
2020-11-09 3:22 ` [PATCH V2 net-next 05/11] net: hns3: add support for dynamic interrupt moderation Huazhong Tan
@ 2020-11-09 3:22 ` Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 07/11] net: hns3: add hns3_state_init() to do state initialization Huazhong Tan
` (4 subsequent siblings)
10 siblings, 0 replies; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
Add a control private flag in ethtool for enable/disable
DIM feature.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
---
drivers/net/ethernet/hisilicon/hns3/hnae3.h | 7 +++
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c | 1 +
drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c | 71 ++++++++++++++++++++++
3 files changed, 79 insertions(+)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hnae3.h b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
index f9d4d23..18b3e43 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hnae3.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
@@ -716,6 +716,11 @@ struct hnae3_roce_private_info {
#define HNAE3_UPE (HNAE3_USER_UPE | HNAE3_OVERFLOW_UPE)
#define HNAE3_MPE (HNAE3_USER_MPE | HNAE3_OVERFLOW_MPE)
+enum hnae3_pflag {
+ HNAE3_PFLAG_DIM_ENABLE,
+ HNAE3_PFLAG_MAX
+};
+
struct hnae3_handle {
struct hnae3_client *client;
struct pci_dev *pdev;
@@ -738,6 +743,8 @@ struct hnae3_handle {
/* Network interface message level enabled bits */
u32 msg_enable;
+
+ unsigned long priv_flags;
};
#define hnae3_set_field(origin, mask, shift, val) \
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index 9e895b9..a567557 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -4246,6 +4246,7 @@ static int hns3_client_init(struct hnae3_handle *handle)
set_bit(HNS3_NIC_STATE_INITED, &priv->state);
set_bit(HNS3_NIC_STATE_DIM_ENABLE, &priv->state);
+ handle->priv_flags |= BIT(HNAE3_PFLAG_DIM_ENABLE);
if (netif_msg_drv(handle))
hns3_info_show(priv);
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
index 30ffaaf..427b72c 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
@@ -18,6 +18,11 @@ struct hns3_sfp_type {
u8 ext_type;
};
+struct hns3_pflag_desc {
+ char name[ETH_GSTRING_LEN];
+ void (*handler)(struct net_device *netdev, bool enable);
+};
+
/* tqp related stats */
#define HNS3_TQP_STAT(_string, _member) { \
.stats_string = _string, \
@@ -59,6 +64,8 @@ static const struct hns3_stats hns3_rxq_stats[] = {
HNS3_TQP_STAT("non_reuse_pg", non_reuse_pg),
};
+#define HNS3_PRIV_FLAGS_LEN ARRAY_SIZE(hns3_priv_flags)
+
#define HNS3_RXQ_STATS_COUNT ARRAY_SIZE(hns3_rxq_stats)
#define HNS3_TQP_STATS_COUNT (HNS3_TXQ_STATS_COUNT + HNS3_RXQ_STATS_COUNT)
@@ -394,6 +401,26 @@ static void hns3_self_test(struct net_device *ndev,
netif_dbg(h, drv, ndev, "self test end\n");
}
+static void hns3_update_state(struct net_device *netdev,
+ enum hns3_nic_state state, bool enable)
+{
+ struct hns3_nic_priv *priv = netdev_priv(netdev);
+
+ if (enable)
+ set_bit(state, &priv->state);
+ else
+ clear_bit(state, &priv->state);
+}
+
+static void hns3_update_dim_state(struct net_device *netdev, bool enable)
+{
+ hns3_update_state(netdev, HNS3_NIC_STATE_DIM_ENABLE, enable);
+}
+
+static const struct hns3_pflag_desc hns3_priv_flags[HNAE3_PFLAG_MAX] = {
+ { "dim_enable", hns3_update_dim_state },
+};
+
static int hns3_get_sset_count(struct net_device *netdev, int stringset)
{
struct hnae3_handle *h = hns3_get_handle(netdev);
@@ -410,6 +437,9 @@ static int hns3_get_sset_count(struct net_device *netdev, int stringset)
case ETH_SS_TEST:
return ops->get_sset_count(h, stringset);
+ case ETH_SS_PRIV_FLAGS:
+ return HNAE3_PFLAG_MAX;
+
default:
return -EOPNOTSUPP;
}
@@ -463,6 +493,7 @@ static void hns3_get_strings(struct net_device *netdev, u32 stringset, u8 *data)
struct hnae3_handle *h = hns3_get_handle(netdev);
const struct hnae3_ae_ops *ops = h->ae_algo->ops;
char *buff = (char *)data;
+ int i;
if (!ops->get_strings)
return;
@@ -475,6 +506,13 @@ static void hns3_get_strings(struct net_device *netdev, u32 stringset, u8 *data)
case ETH_SS_TEST:
ops->get_strings(h, stringset, data);
break;
+ case ETH_SS_PRIV_FLAGS:
+ for (i = 0; i < HNS3_PRIV_FLAGS_LEN; i++) {
+ snprintf(buff, ETH_GSTRING_LEN, "%s",
+ hns3_priv_flags[i].name);
+ buff += ETH_GSTRING_LEN;
+ }
+ break;
default:
break;
}
@@ -1516,6 +1554,35 @@ static int hns3_get_module_eeprom(struct net_device *netdev,
return ops->get_module_eeprom(handle, ee->offset, ee->len, data);
}
+static u32 hns3_get_priv_flags(struct net_device *netdev)
+{
+ struct hnae3_handle *handle = hns3_get_handle(netdev);
+
+ return handle->priv_flags;
+}
+
+static int hns3_set_priv_flags(struct net_device *netdev, u32 pflags)
+{
+ struct hnae3_handle *handle = hns3_get_handle(netdev);
+ u32 changed = pflags ^ handle->priv_flags;
+ u32 i;
+
+ for (i = 0; i < HNAE3_PFLAG_MAX; i++) {
+ if (changed & BIT(i)) {
+ bool enable = !(handle->priv_flags & BIT(i));
+
+ if (enable)
+ handle->priv_flags |= BIT(i);
+ else
+ handle->priv_flags &= ~BIT(i);
+
+ hns3_priv_flags[i].handler(netdev, enable);
+ }
+ }
+
+ return 0;
+}
+
#define HNS3_ETHTOOL_COALESCE (ETHTOOL_COALESCE_USECS | \
ETHTOOL_COALESCE_USE_ADAPTIVE | \
ETHTOOL_COALESCE_RX_USECS_HIGH | \
@@ -1546,6 +1613,8 @@ static const struct ethtool_ops hns3vf_ethtool_ops = {
.get_link = hns3_get_link,
.get_msglevel = hns3_get_msglevel,
.set_msglevel = hns3_set_msglevel,
+ .get_priv_flags = hns3_get_priv_flags,
+ .set_priv_flags = hns3_set_priv_flags,
};
static const struct ethtool_ops hns3_ethtool_ops = {
@@ -1582,6 +1651,8 @@ static const struct ethtool_ops hns3_ethtool_ops = {
.set_fecparam = hns3_set_fecparam,
.get_module_info = hns3_get_module_info,
.get_module_eeprom = hns3_get_module_eeprom,
+ .get_priv_flags = hns3_get_priv_flags,
+ .set_priv_flags = hns3_set_priv_flags,
};
void hns3_ethtool_set_ops(struct net_device *netdev)
--
2.7.4
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH V2 net-next 07/11] net: hns3: add hns3_state_init() to do state initialization
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
` (5 preceding siblings ...)
2020-11-09 3:22 ` [PATCH V2 net-next 06/11] net: hns3: add ethtool priv-flag for DIM Huazhong Tan
@ 2020-11-09 3:22 ` Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 08/11] net: hns3: add a check for ethtool priv-flag interface Huazhong Tan
` (3 subsequent siblings)
10 siblings, 0 replies; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
To improve the readability and maintainability, add hns3_state_init()
to initialize the state.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
---
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c | 14 +++++++++++---
1 file changed, 11 insertions(+), 3 deletions(-)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index a567557..f686723 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -4144,6 +4144,16 @@ static void hns3_info_show(struct hns3_nic_priv *priv)
dev_info(priv->dev, "Max mtu size: %u\n", priv->netdev->max_mtu);
}
+static void hns3_state_init(struct hnae3_handle *handle)
+{
+ struct net_device *netdev = handle->kinfo.netdev;
+ struct hns3_nic_priv *priv = netdev_priv(netdev);
+
+ set_bit(HNS3_NIC_STATE_INITED, &priv->state);
+ set_bit(HNS3_NIC_STATE_DIM_ENABLE, &priv->state);
+ handle->priv_flags |= BIT(HNAE3_PFLAG_DIM_ENABLE);
+}
+
static int hns3_client_init(struct hnae3_handle *handle)
{
struct pci_dev *pdev = handle->pdev;
@@ -4244,9 +4254,7 @@ static int hns3_client_init(struct hnae3_handle *handle)
/* MTU range: (ETH_MIN_MTU(kernel default) - 9702) */
netdev->max_mtu = HNS3_MAX_MTU;
- set_bit(HNS3_NIC_STATE_INITED, &priv->state);
- set_bit(HNS3_NIC_STATE_DIM_ENABLE, &priv->state);
- handle->priv_flags |= BIT(HNAE3_PFLAG_DIM_ENABLE);
+ hns3_state_init(handle);
if (netif_msg_drv(handle))
hns3_info_show(priv);
--
2.7.4
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH V2 net-next 08/11] net: hns3: add a check for ethtool priv-flag interface
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
` (6 preceding siblings ...)
2020-11-09 3:22 ` [PATCH V2 net-next 07/11] net: hns3: add hns3_state_init() to do state initialization Huazhong Tan
@ 2020-11-09 3:22 ` Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 09/11] net: hns3: add support for EQ/CQ mode configuration Huazhong Tan
` (2 subsequent siblings)
10 siblings, 0 replies; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
Add a check for hns3_set_priv_flags() since if the capability
is unsupported its private flags should not be modified as well.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
---
drivers/net/ethernet/hisilicon/hns3/hnae3.h | 1 +
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c | 1 +
drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c | 19 +++++++++++++++++++
3 files changed, 21 insertions(+)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hnae3.h b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
index 18b3e43..3642740 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hnae3.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
@@ -744,6 +744,7 @@ struct hnae3_handle {
/* Network interface message level enabled bits */
u32 msg_enable;
+ unsigned long supported_pflags;
unsigned long priv_flags;
};
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index f686723..c30cf9e 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -4152,6 +4152,7 @@ static void hns3_state_init(struct hnae3_handle *handle)
set_bit(HNS3_NIC_STATE_INITED, &priv->state);
set_bit(HNS3_NIC_STATE_DIM_ENABLE, &priv->state);
handle->priv_flags |= BIT(HNAE3_PFLAG_DIM_ENABLE);
+ set_bit(HNAE3_PFLAG_DIM_ENABLE, &handle->supported_pflags);
}
static int hns3_client_init(struct hnae3_handle *handle)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
index 427b72c..6904c0a 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
@@ -1561,12 +1561,31 @@ static u32 hns3_get_priv_flags(struct net_device *netdev)
return handle->priv_flags;
}
+static int hns3_check_priv_flags(struct hnae3_handle *h, u32 changed)
+{
+ u32 i;
+
+ for (i = 0; i < HNAE3_PFLAG_MAX; i++)
+ if ((changed & BIT(i)) && !test_bit(i, &h->supported_pflags)) {
+ netdev_err(h->netdev, "%s is unsupported\n",
+ hns3_priv_flags[i].name);
+ return -EOPNOTSUPP;
+ }
+
+ return 0;
+}
+
static int hns3_set_priv_flags(struct net_device *netdev, u32 pflags)
{
struct hnae3_handle *handle = hns3_get_handle(netdev);
u32 changed = pflags ^ handle->priv_flags;
+ int ret;
u32 i;
+ ret = hns3_check_priv_flags(handle, changed);
+ if (ret)
+ return ret;
+
for (i = 0; i < HNAE3_PFLAG_MAX; i++) {
if (changed & BIT(i)) {
bool enable = !(handle->priv_flags & BIT(i));
--
2.7.4
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH V2 net-next 09/11] net: hns3: add support for EQ/CQ mode configuration
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
` (7 preceding siblings ...)
2020-11-09 3:22 ` [PATCH V2 net-next 08/11] net: hns3: add a check for ethtool priv-flag interface Huazhong Tan
@ 2020-11-09 3:22 ` Huazhong Tan
2020-11-11 1:25 ` Jakub Kicinski
2020-11-09 3:22 ` [PATCH V2 net-next 10/11] net: hns3: add ethtool priv-flag for EQ/CQ Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 11/11] net: hns3: add debugfs support for interrupt coalesce Huazhong Tan
10 siblings, 1 reply; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
For device whose version is above V3(include V3), the GL can
select EQ or CQ mode, so adds support for it.
In CQ mode, the coalesced timer will restart upon new completion,
while in EQ mode, the timer will not restart.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
---
drivers/net/ethernet/hisilicon/hns3/hnae3.h | 1 +
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c | 49 +++++++++++++++++++++-
drivers/net/ethernet/hisilicon/hns3/hns3_enet.h | 8 ++++
.../ethernet/hisilicon/hns3/hns3pf/hclge_main.c | 1 +
.../ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c | 1 +
5 files changed, 58 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hnae3.h b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
index 3642740..345e8a4 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hnae3.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
@@ -684,6 +684,7 @@ struct hnae3_knic_private_info {
u16 int_rl_setting;
enum pkt_hash_types rss_type;
+ void __iomem *io_base;
};
struct hnae3_roce_private_info {
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index c30cf9e..d1243ea 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -3653,9 +3653,7 @@ static void hns3_tx_dim_work(struct work_struct *work)
static void hns3_nic_init_dim(struct hns3_enet_tqp_vector *tqp_vector)
{
INIT_WORK(&tqp_vector->rx_group.dim.work, hns3_rx_dim_work);
- tqp_vector->rx_group.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
INIT_WORK(&tqp_vector->tx_group.dim.work, hns3_tx_dim_work);
- tqp_vector->tx_group.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
}
static int hns3_nic_init_vector_data(struct hns3_nic_priv *priv)
@@ -4155,6 +4153,48 @@ static void hns3_state_init(struct hnae3_handle *handle)
set_bit(HNAE3_PFLAG_DIM_ENABLE, &handle->supported_pflags);
}
+static void hns3_set_cq_period_mode(struct hns3_nic_priv *priv,
+ enum dim_cq_period_mode mode, bool is_tx)
+{
+ struct hnae3_ae_dev *ae_dev = pci_get_drvdata(priv->ae_handle->pdev);
+ struct hnae3_handle *handle = priv->ae_handle;
+ int i;
+
+ if (is_tx) {
+ priv->tx_cqe_mode = mode;
+
+ for (i = 0; i < priv->vector_num; i++)
+ priv->tqp_vector[i].tx_group.dim.mode = mode;
+ } else {
+ priv->rx_cqe_mode = mode;
+
+ for (i = 0; i < priv->vector_num; i++)
+ priv->tqp_vector[i].rx_group.dim.mode = mode;
+ }
+
+ /* only device version above V3(include V3), GL can switch CQ/EQ
+ * period mode.
+ */
+ if (ae_dev->dev_version >= HNAE3_DEVICE_VERSION_V3) {
+ u32 new_mode;
+ u64 reg;
+
+ new_mode = (mode == DIM_CQ_PERIOD_MODE_START_FROM_CQE) ?
+ HNS3_CQ_MODE_CQE : HNS3_CQ_MODE_EQE;
+ reg = is_tx ? HNS3_GL1_CQ_MODE_REG : HNS3_GL0_CQ_MODE_REG;
+
+ writel(new_mode, handle->kinfo.io_base + reg);
+ }
+}
+
+static void hns3_cq_period_mode_init(struct hns3_nic_priv *priv,
+ enum dim_cq_period_mode tx_mode,
+ enum dim_cq_period_mode rx_mode)
+{
+ hns3_set_cq_period_mode(priv, tx_mode, true);
+ hns3_set_cq_period_mode(priv, rx_mode, false);
+}
+
static int hns3_client_init(struct hnae3_handle *handle)
{
struct pci_dev *pdev = handle->pdev;
@@ -4220,6 +4260,9 @@ static int hns3_client_init(struct hnae3_handle *handle)
goto out_init_ring;
}
+ hns3_cq_period_mode_init(priv, DIM_CQ_PERIOD_MODE_START_FROM_EQE,
+ DIM_CQ_PERIOD_MODE_START_FROM_EQE);
+
ret = hns3_init_phy(netdev);
if (ret)
goto out_init_phy;
@@ -4580,6 +4623,8 @@ static int hns3_reset_notify_init_enet(struct hnae3_handle *handle)
if (ret)
goto err_uninit_vector;
+ hns3_cq_period_mode_init(priv, priv->tx_cqe_mode, priv->rx_cqe_mode);
+
/* the device can work without cpu rmap, only aRFS needs it */
ret = hns3_set_rx_cpu_rmap(netdev);
if (ret)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
index eb4e7ef..c6c082a 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
@@ -188,6 +188,12 @@ enum hns3_nic_state {
#define HNS3_RING_EN_B 0
+#define HNS3_GL0_CQ_MODE_REG 0x20d00
+#define HNS3_GL1_CQ_MODE_REG 0x20d04
+#define HNS3_GL2_CQ_MODE_REG 0x20d08
+#define HNS3_CQ_MODE_EQE 1U
+#define HNS3_CQ_MODE_CQE 0U
+
enum hns3_pkt_l2t_type {
HNS3_L2_TYPE_UNICAST,
HNS3_L2_TYPE_MULTICAST,
@@ -495,6 +501,8 @@ struct hns3_nic_priv {
unsigned long state;
+ enum dim_cq_period_mode tx_cqe_mode;
+ enum dim_cq_period_mode rx_cqe_mode;
struct hns3_enet_coalesce tx_coal;
struct hns3_enet_coalesce rx_coal;
};
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
index 7102001..0573e68 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
@@ -1738,6 +1738,7 @@ static int hclge_vport_setup(struct hclge_vport *vport, u16 num_tqps)
nic->pdev = hdev->pdev;
nic->ae_algo = &ae_algo;
nic->numa_node_mask = hdev->numa_node_mask;
+ nic->kinfo.io_base = hdev->hw.io_base;
ret = hclge_knic_setup(vport, num_tqps,
hdev->num_tx_desc, hdev->num_rx_desc);
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
index 71007e7..b628e52 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
@@ -515,6 +515,7 @@ static int hclgevf_set_handle_info(struct hclgevf_dev *hdev)
nic->pdev = hdev->pdev;
nic->numa_node_mask = hdev->numa_node_mask;
nic->flags |= HNAE3_SUPPORT_VF;
+ nic->kinfo.io_base = hdev->hw.io_base;
ret = hclgevf_knic_setup(hdev);
if (ret)
--
2.7.4
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH V2 net-next 10/11] net: hns3: add ethtool priv-flag for EQ/CQ
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
` (8 preceding siblings ...)
2020-11-09 3:22 ` [PATCH V2 net-next 09/11] net: hns3: add support for EQ/CQ mode configuration Huazhong Tan
@ 2020-11-09 3:22 ` Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 11/11] net: hns3: add debugfs support for interrupt coalesce Huazhong Tan
10 siblings, 0 replies; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
Add a control private flag in ethtool for switching EQ/CQ mode.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
---
drivers/net/ethernet/hisilicon/hns3/hnae3.h | 2 ++
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c | 19 +++++++++++++++--
drivers/net/ethernet/hisilicon/hns3/hns3_enet.h | 2 ++
drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c | 24 ++++++++++++++++++++++
4 files changed, 45 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hnae3.h b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
index 345e8a4..a452874 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hnae3.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
@@ -719,6 +719,8 @@ struct hnae3_roce_private_info {
enum hnae3_pflag {
HNAE3_PFLAG_DIM_ENABLE,
+ HNAE3_PFLAG_TX_CQE_MODE,
+ HNAE3_PFLAG_RX_CQE_MODE,
HNAE3_PFLAG_MAX
};
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index d1243ea..93f7731 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -4144,6 +4144,7 @@ static void hns3_info_show(struct hns3_nic_priv *priv)
static void hns3_state_init(struct hnae3_handle *handle)
{
+ struct hnae3_ae_dev *ae_dev = pci_get_drvdata(handle->pdev);
struct net_device *netdev = handle->kinfo.netdev;
struct hns3_nic_priv *priv = netdev_priv(netdev);
@@ -4151,10 +4152,24 @@ static void hns3_state_init(struct hnae3_handle *handle)
set_bit(HNS3_NIC_STATE_DIM_ENABLE, &priv->state);
handle->priv_flags |= BIT(HNAE3_PFLAG_DIM_ENABLE);
set_bit(HNAE3_PFLAG_DIM_ENABLE, &handle->supported_pflags);
+
+ /* device version above V3(include V3), GL can switch CQ/EQ period
+ * mode.
+ */
+ if (ae_dev->dev_version >= HNAE3_DEVICE_VERSION_V3) {
+ set_bit(HNAE3_PFLAG_TX_CQE_MODE, &handle->supported_pflags);
+ set_bit(HNAE3_PFLAG_RX_CQE_MODE, &handle->supported_pflags);
+ }
+
+ if (priv->tx_cqe_mode == DIM_CQ_PERIOD_MODE_START_FROM_CQE)
+ handle->priv_flags |= BIT(HNAE3_PFLAG_TX_CQE_MODE);
+
+ if (priv->rx_cqe_mode == DIM_CQ_PERIOD_MODE_START_FROM_CQE)
+ handle->priv_flags |= BIT(HNAE3_PFLAG_RX_CQE_MODE);
}
-static void hns3_set_cq_period_mode(struct hns3_nic_priv *priv,
- enum dim_cq_period_mode mode, bool is_tx)
+void hns3_set_cq_period_mode(struct hns3_nic_priv *priv,
+ enum dim_cq_period_mode mode, bool is_tx)
{
struct hnae3_ae_dev *ae_dev = pci_get_drvdata(priv->ae_handle->pdev);
struct hnae3_handle *handle = priv->ae_handle;
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
index c6c082a..ecdb544 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
@@ -635,4 +635,6 @@ void hns3_dbg_uninit(struct hnae3_handle *handle);
void hns3_dbg_register_debugfs(const char *debugfs_dir_name);
void hns3_dbg_unregister_debugfs(void);
void hns3_shinfo_pack(struct skb_shared_info *shinfo, __u32 *size);
+void hns3_set_cq_period_mode(struct hns3_nic_priv *priv,
+ enum dim_cq_period_mode mode, bool is_tx);
#endif
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
index 6904c0a..cfc0766 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
@@ -417,8 +417,32 @@ static void hns3_update_dim_state(struct net_device *netdev, bool enable)
hns3_update_state(netdev, HNS3_NIC_STATE_DIM_ENABLE, enable);
}
+static void hns3_update_cqe_mode(struct net_device *netdev, bool enable,
+ bool is_tx)
+{
+ struct hns3_nic_priv *priv = netdev_priv(netdev);
+ enum dim_cq_period_mode mode;
+
+ mode = enable ? DIM_CQ_PERIOD_MODE_START_FROM_CQE :
+ DIM_CQ_PERIOD_MODE_START_FROM_EQE;
+
+ hns3_set_cq_period_mode(priv, mode, is_tx);
+}
+
+static void hns3_update_tx_cqe_mode(struct net_device *netdev, bool enable)
+{
+ hns3_update_cqe_mode(netdev, enable, true);
+}
+
+static void hns3_update_rx_cqe_mode(struct net_device *netdev, bool enable)
+{
+ hns3_update_cqe_mode(netdev, enable, false);
+}
+
static const struct hns3_pflag_desc hns3_priv_flags[HNAE3_PFLAG_MAX] = {
{ "dim_enable", hns3_update_dim_state },
+ { "tx_cqe_mode", hns3_update_tx_cqe_mode },
+ { "rx_cqe_mode", hns3_update_rx_cqe_mode },
};
static int hns3_get_sset_count(struct net_device *netdev, int stringset)
--
2.7.4
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH V2 net-next 11/11] net: hns3: add debugfs support for interrupt coalesce
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
` (9 preceding siblings ...)
2020-11-09 3:22 ` [PATCH V2 net-next 10/11] net: hns3: add ethtool priv-flag for EQ/CQ Huazhong Tan
@ 2020-11-09 3:22 ` Huazhong Tan
2020-11-11 1:28 ` Jakub Kicinski
10 siblings, 1 reply; 22+ messages in thread
From: Huazhong Tan @ 2020-11-09 3:22 UTC (permalink / raw)
To: davem
Cc: netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm, kuba,
Huazhong Tan
Since user may need to check the current configuration of the
interrupt coalesce, so add debugfs support for query this info,
which includes DIM profile, coalesce configuration of both software
and hardware.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
---
drivers/net/ethernet/hisilicon/hns3/hns3_debugfs.c | 126 +++++++++++++++++++++
1 file changed, 126 insertions(+)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_debugfs.c b/drivers/net/ethernet/hisilicon/hns3/hns3_debugfs.c
index a5ebca8..26fa69a 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_debugfs.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_debugfs.c
@@ -12,6 +12,93 @@
static struct dentry *hns3_dbgfs_root;
+static ssize_t hns3_dbg_coal_write(struct file *filp, const char __user *buffer,
+ size_t count, loff_t *ppos)
+{
+ struct hnae3_handle *h = filp->private_data;
+ struct hns3_nic_priv *priv = h->priv;
+ struct hns3_enet_tqp_vector *tqp_vector;
+ struct hns3_enet_coalesce *coal;
+ u8 __iomem *base_addr;
+ int uncopied_bytes;
+ unsigned int idx;
+ struct dim *dim;
+ char *cmd_buf;
+
+ if (*ppos != 0)
+ return 0;
+
+ if (!test_bit(HNS3_NIC_STATE_INITED, &priv->state)) {
+ dev_err(&h->pdev->dev, "device is not initialized\n");
+ return -EFAULT;
+ }
+
+ cmd_buf = kzalloc(count + 1, GFP_KERNEL);
+ if (!cmd_buf)
+ return -ENOMEM;
+
+ uncopied_bytes = copy_from_user(cmd_buf, buffer, count);
+ if (uncopied_bytes) {
+ kfree(cmd_buf);
+ return -EFAULT;
+ }
+
+ cmd_buf[count] = '\0';
+
+ if (kstrtouint(cmd_buf, 0, &idx))
+ idx = 0;
+
+ if (idx >= priv->vector_num) {
+ dev_err(&h->pdev->dev,
+ "vector index(%u) is out of range(0-%u)\n", idx,
+ priv->vector_num - 1);
+ kfree(cmd_buf);
+ return -EINVAL;
+ }
+
+ tqp_vector = &priv->tqp_vector[idx];
+ coal = &tqp_vector->tx_group.coal;
+ dim = &tqp_vector->tx_group.dim;
+ base_addr = tqp_vector->mask_addr;
+
+ dev_info(&h->pdev->dev, "vector[%u] interrupt coalesce info:\n", idx);
+ dev_info(&h->pdev->dev,
+ "TX DIM info state = %d profile_ix = %d mode = %d tune_state = %d steps_right = %d steps_left = %d tired = %d\n",
+ dim->state, dim->profile_ix, dim->mode, dim->tune_state,
+ dim->steps_right, dim->steps_left, dim->tired);
+
+ dev_info(&h->pdev->dev, "TX GL info sw_gl = %u, hw_gl = %u\n",
+ coal->int_gl,
+ readl(base_addr + HNS3_VECTOR_GL1_OFFSET));
+
+ if (coal->ql_enable)
+ dev_info(&h->pdev->dev, "TX QL info sw_ql = %u, hw_ql = %u\n",
+ coal->int_ql,
+ readl(base_addr + HNS3_VECTOR_TX_QL_OFFSET));
+
+ coal = &tqp_vector->rx_group.coal;
+ dim = &tqp_vector->rx_group.dim;
+
+ dev_info(&h->pdev->dev,
+ "RX dim_info state = %d profile_ix = %d mode = %d tune_state = %d steps_right = %d steps_left = %d tired = %d\n",
+ dim->state, dim->profile_ix, dim->mode, dim->tune_state,
+ dim->steps_right, dim->steps_left, dim->tired);
+
+ dev_info(&h->pdev->dev, "RX GL info sw_gl = %u, hw_gl = %u\n",
+ coal->int_gl,
+ readl(base_addr + HNS3_VECTOR_GL0_OFFSET));
+
+ if (coal->ql_enable)
+ dev_info(&h->pdev->dev, "RX QL info sw_ql = %u, hw_ql = %u\n",
+ coal->int_ql,
+ readl(base_addr + HNS3_VECTOR_RX_QL_OFFSET));
+
+ kfree(cmd_buf);
+ cmd_buf = NULL;
+
+ return count;
+}
+
static int hns3_dbg_queue_info(struct hnae3_handle *h,
const char *cmd_buf)
{
@@ -352,6 +439,35 @@ static void hns3_dbg_dev_specs(struct hnae3_handle *h)
dev_info(priv->dev, "MAX INT GL: %u\n", dev_specs->max_int_gl);
}
+static ssize_t hns3_dbg_coal_read(struct file *filp, char __user *buffer,
+ size_t count, loff_t *ppos)
+{
+ int uncopy_bytes;
+ char *buf;
+ int len;
+
+ if (*ppos != 0)
+ return 0;
+
+ if (count < HNS3_DBG_READ_LEN)
+ return -ENOSPC;
+
+ buf = kzalloc(HNS3_DBG_READ_LEN, GFP_KERNEL);
+ if (!buf)
+ return -ENOMEM;
+
+ len = scnprintf(buf, HNS3_DBG_READ_LEN, "%s\n",
+ "Please echo index to coal");
+ uncopy_bytes = copy_to_user(buffer, buf, len);
+
+ kfree(buf);
+
+ if (uncopy_bytes)
+ return -EFAULT;
+
+ return (*ppos = len);
+}
+
static ssize_t hns3_dbg_cmd_read(struct file *filp, char __user *buffer,
size_t count, loff_t *ppos)
{
@@ -452,6 +568,13 @@ static const struct file_operations hns3_dbg_cmd_fops = {
.write = hns3_dbg_cmd_write,
};
+static const struct file_operations hns3_dbg_coal_fops = {
+ .owner = THIS_MODULE,
+ .open = simple_open,
+ .read = hns3_dbg_coal_read,
+ .write = hns3_dbg_coal_write,
+};
+
void hns3_dbg_init(struct hnae3_handle *handle)
{
const char *name = pci_name(handle->pdev);
@@ -460,6 +583,9 @@ void hns3_dbg_init(struct hnae3_handle *handle)
debugfs_create_file("cmd", 0600, handle->hnae3_dbgfs, handle,
&hns3_dbg_cmd_fops);
+
+ debugfs_create_file("coal", 0600, handle->hnae3_dbgfs, handle,
+ &hns3_dbg_coal_fops);
}
void hns3_dbg_uninit(struct hnae3_handle *handle)
--
2.7.4
^ permalink raw reply related [flat|nested] 22+ messages in thread
* Re: [PATCH V2 net-next 01/11] net: hns3: add support for configuring interrupt quantity limiting
2020-11-09 3:22 ` [PATCH V2 net-next 01/11] net: hns3: add support for configuring interrupt quantity limiting Huazhong Tan
@ 2020-11-11 1:13 ` Jakub Kicinski
2020-11-11 1:15 ` tanhuazhong
0 siblings, 1 reply; 22+ messages in thread
From: Jakub Kicinski @ 2020-11-11 1:13 UTC (permalink / raw)
To: Huazhong Tan
Cc: davem, netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm
On Mon, 9 Nov 2020 11:22:29 +0800 Huazhong Tan wrote:
> + if (rx_vector->tx_group.coal.ql_enable)
^^^^^^^^
Is this supposed to be rx_group, not tx?
> + hns3_set_vector_coalesce_rx_ql(rx_vector,
> + rx_vector->rx_group.coal.int_ql);
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH V2 net-next 01/11] net: hns3: add support for configuring interrupt quantity limiting
2020-11-11 1:13 ` Jakub Kicinski
@ 2020-11-11 1:15 ` tanhuazhong
0 siblings, 0 replies; 22+ messages in thread
From: tanhuazhong @ 2020-11-11 1:15 UTC (permalink / raw)
To: Jakub Kicinski
Cc: davem, netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm
On 2020/11/11 9:13, Jakub Kicinski wrote:
> On Mon, 9 Nov 2020 11:22:29 +0800 Huazhong Tan wrote:
>> + if (rx_vector->tx_group.coal.ql_enable)
> ^^^^^^^^
>
> Is this supposed to be rx_group, not tx?
>
yes, will fix it.
thanks.
>> + hns3_set_vector_coalesce_rx_ql(rx_vector,
>> + rx_vector->rx_group.coal.int_ql);
>
> .
>
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH V2 net-next 05/11] net: hns3: add support for dynamic interrupt moderation
2020-11-09 3:22 ` [PATCH V2 net-next 05/11] net: hns3: add support for dynamic interrupt moderation Huazhong Tan
@ 2020-11-11 1:20 ` Jakub Kicinski
2020-11-11 2:31 ` tanhuazhong
0 siblings, 1 reply; 22+ messages in thread
From: Jakub Kicinski @ 2020-11-11 1:20 UTC (permalink / raw)
To: Huazhong Tan
Cc: davem, netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm
On Mon, 9 Nov 2020 11:22:33 +0800 Huazhong Tan wrote:
> Add dynamic interrupt moderation support for the HNS3 driver.
>
> Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
I'm slightly confused here. What does the adaptive moderation do in
your driver/device if you still need DIM on top of it?
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH V2 net-next 09/11] net: hns3: add support for EQ/CQ mode configuration
2020-11-09 3:22 ` [PATCH V2 net-next 09/11] net: hns3: add support for EQ/CQ mode configuration Huazhong Tan
@ 2020-11-11 1:25 ` Jakub Kicinski
2020-11-11 3:03 ` tanhuazhong
0 siblings, 1 reply; 22+ messages in thread
From: Jakub Kicinski @ 2020-11-11 1:25 UTC (permalink / raw)
To: Huazhong Tan
Cc: davem, netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm
On Mon, 9 Nov 2020 11:22:37 +0800 Huazhong Tan wrote:
> For device whose version is above V3(include V3), the GL can
> select EQ or CQ mode, so adds support for it.
>
> In CQ mode, the coalesced timer will restart upon new completion,
> while in EQ mode, the timer will not restart.
>
> Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Let's see if I understand - in CQ mode the timer is restarted very time
new frame gets received/transmitted? IOW for a continuous stream of
frames it will only generate an interrupt once it reaches max_frames?
I think that if you need such a configuration knob we should add this as
an option to the official ethtool -c/-C interface, now that we have the
ability to extend the netlink API.
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH V2 net-next 11/11] net: hns3: add debugfs support for interrupt coalesce
2020-11-09 3:22 ` [PATCH V2 net-next 11/11] net: hns3: add debugfs support for interrupt coalesce Huazhong Tan
@ 2020-11-11 1:28 ` Jakub Kicinski
2020-11-11 3:16 ` tanhuazhong
0 siblings, 1 reply; 22+ messages in thread
From: Jakub Kicinski @ 2020-11-11 1:28 UTC (permalink / raw)
To: Huazhong Tan
Cc: davem, netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm
On Mon, 9 Nov 2020 11:22:39 +0800 Huazhong Tan wrote:
> Since user may need to check the current configuration of the
> interrupt coalesce, so add debugfs support for query this info,
> which includes DIM profile, coalesce configuration of both software
> and hardware.
>
> Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Please create a file per vector so that users can just read it instead
of dumping the info into the logs.
Even better we should put as much of this information as possible into
the ethtool API. dim state is hardly hardware-specific.
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH V2 net-next 05/11] net: hns3: add support for dynamic interrupt moderation
2020-11-11 1:20 ` Jakub Kicinski
@ 2020-11-11 2:31 ` tanhuazhong
0 siblings, 0 replies; 22+ messages in thread
From: tanhuazhong @ 2020-11-11 2:31 UTC (permalink / raw)
To: Jakub Kicinski
Cc: davem, netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm
On 2020/11/11 9:20, Jakub Kicinski wrote:
> On Mon, 9 Nov 2020 11:22:33 +0800 Huazhong Tan wrote:
>> Add dynamic interrupt moderation support for the HNS3 driver.
>>
>> Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
>
> I'm slightly confused here. What does the adaptive moderation do in
> your driver/device if you still need DIM on top of it?
>
The driver's adaptive moderation and DIM are mutually-exclusive options,
and DIM is used by default. Since it is hard to verify all cases that
DIM can get a better result, so keep the adaptive moderation as an
option now, which can be swicthed by the ethtool priv-flag. If DIM is ok
for a long time, then we will send a separate patch to remove the
adaptive moderation from the driver.
Thanks.
>
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH V2 net-next 09/11] net: hns3: add support for EQ/CQ mode configuration
2020-11-11 1:25 ` Jakub Kicinski
@ 2020-11-11 3:03 ` tanhuazhong
0 siblings, 0 replies; 22+ messages in thread
From: tanhuazhong @ 2020-11-11 3:03 UTC (permalink / raw)
To: Jakub Kicinski
Cc: davem, netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm
On 2020/11/11 9:25, Jakub Kicinski wrote:
> On Mon, 9 Nov 2020 11:22:37 +0800 Huazhong Tan wrote:
>> For device whose version is above V3(include V3), the GL can
>> select EQ or CQ mode, so adds support for it.
>>
>> In CQ mode, the coalesced timer will restart upon new completion,
>> while in EQ mode, the timer will not restart.
>>
>> Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
>
> Let's see if I understand - in CQ mode the timer is restarted very time
> new frame gets received/transmitted? IOW for a continuous stream of
> frames it will only generate an interrupt once it reaches max_frames?
>
Hi, Jakub.
This EQ/CQ is related to the GL(gap limiting, interrupt coalesce based
on the gap time).
More exactly, the coalesced timer will restart when the first new
completion occurs. Will fix the commit log.
> I think that if you need such a configuration knob we should add this as
> an option to the official ethtool -c/-C interface, now that we have the
> ability to extend the netlink API.
>
It seems need more jobs to be did later.
Thanks.
> .
>
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH V2 net-next 11/11] net: hns3: add debugfs support for interrupt coalesce
2020-11-11 1:28 ` Jakub Kicinski
@ 2020-11-11 3:16 ` tanhuazhong
2020-11-11 16:15 ` Jakub Kicinski
0 siblings, 1 reply; 22+ messages in thread
From: tanhuazhong @ 2020-11-11 3:16 UTC (permalink / raw)
To: Jakub Kicinski
Cc: davem, netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm
On 2020/11/11 9:28, Jakub Kicinski wrote:
> On Mon, 9 Nov 2020 11:22:39 +0800 Huazhong Tan wrote:
>> Since user may need to check the current configuration of the
>> interrupt coalesce, so add debugfs support for query this info,
>> which includes DIM profile, coalesce configuration of both software
>> and hardware.
>>
>> Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
>
> Please create a file per vector so that users can just read it instead
> of dumping the info into the logs.
>
This patch should be removed from this series right now. Since this new
read method needs some adaptations and verifications, and there maybe
another better ways to dump this info.
> Even better we should put as much of this information as possible into
> the ethtool API. dim state is hardly hardware-specific.
>
Should the ethtool API used to dump the hardware info? Could you provide
some hints to do it?
> .
>
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH V2 net-next 11/11] net: hns3: add debugfs support for interrupt coalesce
2020-11-11 3:16 ` tanhuazhong
@ 2020-11-11 16:15 ` Jakub Kicinski
2020-11-12 0:48 ` tanhuazhong
0 siblings, 1 reply; 22+ messages in thread
From: Jakub Kicinski @ 2020-11-11 16:15 UTC (permalink / raw)
To: tanhuazhong
Cc: davem, netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm
On Wed, 11 Nov 2020 11:16:37 +0800 tanhuazhong wrote:
> On 2020/11/11 9:28, Jakub Kicinski wrote:
> > On Mon, 9 Nov 2020 11:22:39 +0800 Huazhong Tan wrote:
> >> Since user may need to check the current configuration of the
> >> interrupt coalesce, so add debugfs support for query this info,
> >> which includes DIM profile, coalesce configuration of both software
> >> and hardware.
> >>
> >> Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
> >
> > Please create a file per vector so that users can just read it instead
> > of dumping the info into the logs.
> >
>
> This patch should be removed from this series right now. Since this new
> read method needs some adaptations and verifications, and there maybe
> another better ways to dump this info.
>
> > Even better we should put as much of this information as possible into
> > the ethtool API. dim state is hardly hardware-specific.
> >
>
> Should the ethtool API used to dump the hardware info? Could you provide
> some hints to do it?
Not necessarily hardware info but if there is a use case for inspecting
DIM state we can extend
coalesce_fill_reply() in net/ethtool/coalesce.c
to report it.
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH V2 net-next 11/11] net: hns3: add debugfs support for interrupt coalesce
2020-11-11 16:15 ` Jakub Kicinski
@ 2020-11-12 0:48 ` tanhuazhong
0 siblings, 0 replies; 22+ messages in thread
From: tanhuazhong @ 2020-11-12 0:48 UTC (permalink / raw)
To: Jakub Kicinski
Cc: davem, netdev, linux-kernel, salil.mehta, yisen.zhuang, linuxarm
On 2020/11/12 0:15, Jakub Kicinski wrote:
> On Wed, 11 Nov 2020 11:16:37 +0800 tanhuazhong wrote:
>> On 2020/11/11 9:28, Jakub Kicinski wrote:
>>> On Mon, 9 Nov 2020 11:22:39 +0800 Huazhong Tan wrote:
>>>> Since user may need to check the current configuration of the
>>>> interrupt coalesce, so add debugfs support for query this info,
>>>> which includes DIM profile, coalesce configuration of both software
>>>> and hardware.
>>>>
>>>> Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
>>>
>>> Please create a file per vector so that users can just read it instead
>>> of dumping the info into the logs.
>>>
>>
>> This patch should be removed from this series right now. Since this new
>> read method needs some adaptations and verifications, and there maybe
>> another better ways to dump this info.
>>
>>> Even better we should put as much of this information as possible into
>>> the ethtool API. dim state is hardly hardware-specific.
>>>
>>
>> Should the ethtool API used to dump the hardware info? Could you provide
>> some hints to do it?
>
> Not necessarily hardware info but if there is a use case for inspecting
> DIM state we can extend
>
> coalesce_fill_reply() in net/ethtool/coalesce.c
>
> to report it.
>
ok, thanks.
> .
>
^ permalink raw reply [flat|nested] 22+ messages in thread
end of thread, other threads:[~2020-11-12 1:39 UTC | newest]
Thread overview: 22+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-11-09 3:22 [PATCH V2 net-next 00/11] net: hns3: updates for -next Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 01/11] net: hns3: add support for configuring interrupt quantity limiting Huazhong Tan
2020-11-11 1:13 ` Jakub Kicinski
2020-11-11 1:15 ` tanhuazhong
2020-11-09 3:22 ` [PATCH V2 net-next 02/11] net: hns3: add support for querying maximum value of GL Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 03/11] net: hns3: add support for 1us unit GL configuration Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 04/11] net: hns3: rename gl_adapt_enable in struct hns3_enet_coalesce Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 05/11] net: hns3: add support for dynamic interrupt moderation Huazhong Tan
2020-11-11 1:20 ` Jakub Kicinski
2020-11-11 2:31 ` tanhuazhong
2020-11-09 3:22 ` [PATCH V2 net-next 06/11] net: hns3: add ethtool priv-flag for DIM Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 07/11] net: hns3: add hns3_state_init() to do state initialization Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 08/11] net: hns3: add a check for ethtool priv-flag interface Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 09/11] net: hns3: add support for EQ/CQ mode configuration Huazhong Tan
2020-11-11 1:25 ` Jakub Kicinski
2020-11-11 3:03 ` tanhuazhong
2020-11-09 3:22 ` [PATCH V2 net-next 10/11] net: hns3: add ethtool priv-flag for EQ/CQ Huazhong Tan
2020-11-09 3:22 ` [PATCH V2 net-next 11/11] net: hns3: add debugfs support for interrupt coalesce Huazhong Tan
2020-11-11 1:28 ` Jakub Kicinski
2020-11-11 3:16 ` tanhuazhong
2020-11-11 16:15 ` Jakub Kicinski
2020-11-12 0:48 ` tanhuazhong
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).