From: Jie Wang <[email protected]>
Currently tx push is a standard driver feature which controls use of a fast
path descriptor push. So this patch extends the ringparam APIs and data
structures to support set/get tx push by ethtool -G/g.
Signed-off-by: Jie Wang <[email protected]>
Signed-off-by: Guangbin Huang <[email protected]>
---
Documentation/networking/ethtool-netlink.rst | 8 ++++++++
include/linux/ethtool.h | 3 +++
include/uapi/linux/ethtool_netlink.h | 1 +
net/ethtool/netlink.h | 2 +-
net/ethtool/rings.c | 18 ++++++++++++++++--
5 files changed, 29 insertions(+), 3 deletions(-)
diff --git a/Documentation/networking/ethtool-netlink.rst b/Documentation/networking/ethtool-netlink.rst
index 24d9be69065d..fcd4fdf96c8e 100644
--- a/Documentation/networking/ethtool-netlink.rst
+++ b/Documentation/networking/ethtool-netlink.rst
@@ -862,6 +862,7 @@ Kernel response contents:
``ETHTOOL_A_RINGS_RX_BUF_LEN`` u32 size of buffers on the ring
``ETHTOOL_A_RINGS_TCP_DATA_SPLIT`` u8 TCP header / data split
``ETHTOOL_A_RINGS_CQE_SIZE`` u32 Size of TX/RX CQE
+ ``ETHTOOL_A_RINGS_TX_PUSH`` u8 flag of TX Push mode
==================================== ====== ===========================
``ETHTOOL_A_RINGS_TCP_DATA_SPLIT`` indicates whether the device is usable with
@@ -871,6 +872,12 @@ separate buffers. The device configuration must make it possible to receive
full memory pages of data, for example because MTU is high enough or through
HW-GRO.
+``ETHTOOL_A_RINGS_TX_PUSH`` flag is used to choose the ordinary path or the fast
+path to send packets. In ordinary path, driver fills BDs to DDR memory and
+notifies NIC hardware. In fast path, driver pushes BDs to the device memory
+directly and thus reducing the sending latencies. Setting tx push attribute "on"
+will enable tx push mode and send packets in fast path if packet size matches.
+For those not supported hardwares, this attributes is "off" by default settings.
RINGS_SET
=========
@@ -887,6 +894,7 @@ Request contents:
``ETHTOOL_A_RINGS_TX`` u32 size of TX ring
``ETHTOOL_A_RINGS_RX_BUF_LEN`` u32 size of buffers on the ring
``ETHTOOL_A_RINGS_CQE_SIZE`` u32 Size of TX/RX CQE
+ ``ETHTOOL_A_RINGS_TX_PUSH`` u8 flag of TX Push mode
==================================== ====== ===========================
Kernel checks that requested ring sizes do not exceed limits reported by
diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h
index 4af58459a1e7..ede4f9154cd2 100644
--- a/include/linux/ethtool.h
+++ b/include/linux/ethtool.h
@@ -71,11 +71,13 @@ enum {
* struct kernel_ethtool_ringparam - RX/TX ring configuration
* @rx_buf_len: Current length of buffers on the rx ring.
* @tcp_data_split: Scatter packet headers and data to separate buffers
+ * @tx_push: The flag of tx push mode
* @cqe_size: Size of TX/RX completion queue event
*/
struct kernel_ethtool_ringparam {
u32 rx_buf_len;
u8 tcp_data_split;
+ u8 tx_push;
u32 cqe_size;
};
@@ -87,6 +89,7 @@ struct kernel_ethtool_ringparam {
enum ethtool_supported_ring_param {
ETHTOOL_RING_USE_RX_BUF_LEN = BIT(0),
ETHTOOL_RING_USE_CQE_SIZE = BIT(1),
+ ETHTOOL_RING_USE_TX_PUSH = BIT(2),
};
#define __ETH_RSS_HASH_BIT(bit) ((u32)1 << (bit))
diff --git a/include/uapi/linux/ethtool_netlink.h b/include/uapi/linux/ethtool_netlink.h
index 979850221b8d..d2fb4f7be61b 100644
--- a/include/uapi/linux/ethtool_netlink.h
+++ b/include/uapi/linux/ethtool_netlink.h
@@ -338,6 +338,7 @@ enum {
ETHTOOL_A_RINGS_RX_BUF_LEN, /* u32 */
ETHTOOL_A_RINGS_TCP_DATA_SPLIT, /* u8 */
ETHTOOL_A_RINGS_CQE_SIZE, /* u32 */
+ ETHTOOL_A_RINGS_TX_PUSH, /* u8 */
/* add new constants above here */
__ETHTOOL_A_RINGS_CNT,
diff --git a/net/ethtool/netlink.h b/net/ethtool/netlink.h
index 29d01662a48b..7919ddb2371c 100644
--- a/net/ethtool/netlink.h
+++ b/net/ethtool/netlink.h
@@ -363,7 +363,7 @@ extern const struct nla_policy ethnl_features_set_policy[ETHTOOL_A_FEATURES_WANT
extern const struct nla_policy ethnl_privflags_get_policy[ETHTOOL_A_PRIVFLAGS_HEADER + 1];
extern const struct nla_policy ethnl_privflags_set_policy[ETHTOOL_A_PRIVFLAGS_FLAGS + 1];
extern const struct nla_policy ethnl_rings_get_policy[ETHTOOL_A_RINGS_HEADER + 1];
-extern const struct nla_policy ethnl_rings_set_policy[ETHTOOL_A_RINGS_CQE_SIZE + 1];
+extern const struct nla_policy ethnl_rings_set_policy[ETHTOOL_A_RINGS_TX_PUSH + 1];
extern const struct nla_policy ethnl_channels_get_policy[ETHTOOL_A_CHANNELS_HEADER + 1];
extern const struct nla_policy ethnl_channels_set_policy[ETHTOOL_A_CHANNELS_COMBINED_COUNT + 1];
extern const struct nla_policy ethnl_coalesce_get_policy[ETHTOOL_A_COALESCE_HEADER + 1];
diff --git a/net/ethtool/rings.c b/net/ethtool/rings.c
index 9f33c9689b56..9ed60c507d97 100644
--- a/net/ethtool/rings.c
+++ b/net/ethtool/rings.c
@@ -55,7 +55,8 @@ static int rings_reply_size(const struct ethnl_req_info *req_base,
nla_total_size(sizeof(u32)) + /* _RINGS_TX */
nla_total_size(sizeof(u32)) + /* _RINGS_RX_BUF_LEN */
nla_total_size(sizeof(u8)) + /* _RINGS_TCP_DATA_SPLIT */
- nla_total_size(sizeof(u32)); /* _RINGS_CQE_SIZE */
+ nla_total_size(sizeof(u32) + /* _RINGS_CQE_SIZE */
+ nla_total_size(sizeof(u8))); /* _RINGS_TX_PUSH */
}
static int rings_fill_reply(struct sk_buff *skb,
@@ -94,7 +95,8 @@ static int rings_fill_reply(struct sk_buff *skb,
(nla_put_u8(skb, ETHTOOL_A_RINGS_TCP_DATA_SPLIT,
kr->tcp_data_split))) ||
(kr->cqe_size &&
- (nla_put_u32(skb, ETHTOOL_A_RINGS_CQE_SIZE, kr->cqe_size))))
+ (nla_put_u32(skb, ETHTOOL_A_RINGS_CQE_SIZE, kr->cqe_size))) ||
+ nla_put_u8(skb, ETHTOOL_A_RINGS_TX_PUSH, !!kr->tx_push))
return -EMSGSIZE;
return 0;
@@ -123,6 +125,7 @@ const struct nla_policy ethnl_rings_set_policy[] = {
[ETHTOOL_A_RINGS_TX] = { .type = NLA_U32 },
[ETHTOOL_A_RINGS_RX_BUF_LEN] = NLA_POLICY_MIN(NLA_U32, 1),
[ETHTOOL_A_RINGS_CQE_SIZE] = NLA_POLICY_MIN(NLA_U32, 1),
+ [ETHTOOL_A_RINGS_TX_PUSH] = NLA_POLICY_MAX(NLA_U8, 1),
};
int ethnl_set_rings(struct sk_buff *skb, struct genl_info *info)
@@ -149,6 +152,15 @@ int ethnl_set_rings(struct sk_buff *skb, struct genl_info *info)
if (!ops->get_ringparam || !ops->set_ringparam)
goto out_dev;
+ if (tb[ETHTOOL_A_RINGS_TX_PUSH] &&
+ !(ops->supported_ring_params & ETHTOOL_RING_USE_TX_PUSH)) {
+ ret = -EOPNOTSUPP;
+ NL_SET_ERR_MSG_ATTR(info->extack,
+ tb[ETHTOOL_A_RINGS_TX_PUSH],
+ "setting tx push not supported");
+ goto out_dev;
+ }
+
rtnl_lock();
ret = ethnl_ops_begin(dev);
if (ret < 0)
@@ -165,6 +177,8 @@ int ethnl_set_rings(struct sk_buff *skb, struct genl_info *info)
tb[ETHTOOL_A_RINGS_RX_BUF_LEN], &mod);
ethnl_update_u32(&kernel_ringparam.cqe_size,
tb[ETHTOOL_A_RINGS_CQE_SIZE], &mod);
+ ethnl_update_u8(&kernel_ringparam.tx_push,
+ tb[ETHTOOL_A_RINGS_TX_PUSH], &mod);
ret = 0;
if (!mod)
goto out_ops;
--
2.33.0
On Fri, 8 Apr 2022 15:12:43 +0800 Guangbin Huang wrote:
> From: Jie Wang <[email protected]>
>
> Currently tx push is a standard driver feature which controls use of a fast
> path descriptor push. So this patch extends the ringparam APIs and data
> structures to support set/get tx push by ethtool -G/g.
>
> Signed-off-by: Jie Wang <[email protected]>
> Signed-off-by: Guangbin Huang <[email protected]>
> +``ETHTOOL_A_RINGS_TX_PUSH`` flag is used to choose the ordinary path or the fast
> +path to send packets. In ordinary path, driver fills BDs to DDR memory and
> +notifies NIC hardware. In fast path, driver pushes BDs to the device memory
> +directly and thus reducing the sending latencies. Setting tx push attribute "on"
> +will enable tx push mode and send packets in fast path if packet size matches.
> +For those not supported hardwares, this attributes is "off" by default settings.
Since you need to respin to fix the kdoc warning - could you also add
a mention that enabling this feature may increase CPU cost? Unless it's
not the case for your implementation, I thought it usually is..
> RINGS_SET
> =========
> @@ -887,6 +894,7 @@ Request contents:
> ``ETHTOOL_A_RINGS_TX`` u32 size of TX ring
> ``ETHTOOL_A_RINGS_RX_BUF_LEN`` u32 size of buffers on the ring
> ``ETHTOOL_A_RINGS_CQE_SIZE`` u32 Size of TX/RX CQE
> + ``ETHTOOL_A_RINGS_TX_PUSH`` u8 flag of TX Push mode
> ==================================== ====== ===========================
>
> Kernel checks that requested ring sizes do not exceed limits reported by
> diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h
> index 4af58459a1e7..ede4f9154cd2 100644
> --- a/include/linux/ethtool.h
> +++ b/include/linux/ethtool.h
> @@ -71,11 +71,13 @@ enum {
> * struct kernel_ethtool_ringparam - RX/TX ring configuration
> * @rx_buf_len: Current length of buffers on the rx ring.
> * @tcp_data_split: Scatter packet headers and data to separate buffers
> + * @tx_push: The flag of tx push mode
> * @cqe_size: Size of TX/RX completion queue event
> */
> struct kernel_ethtool_ringparam {
> u32 rx_buf_len;
> u8 tcp_data_split;
> + u8 tx_push;
> u32 cqe_size;
> };
>
> @@ -87,6 +89,7 @@ struct kernel_ethtool_ringparam {
> enum ethtool_supported_ring_param {
> ETHTOOL_RING_USE_RX_BUF_LEN = BIT(0),
> ETHTOOL_RING_USE_CQE_SIZE = BIT(1),
> + ETHTOOL_RING_USE_TX_PUSH = BIT(2),
include/linux/ethtool.h:94: warning: Enum value 'ETHTOOL_RING_USE_TX_PUSH' not described in enum 'ethtool_supported_ring_param'
On 2022/4/9 5:55, Jakub Kicinski wrote:
> On Fri, 8 Apr 2022 15:12:43 +0800 Guangbin Huang wrote:
>> From: Jie Wang <[email protected]>
>>
>> Currently tx push is a standard driver feature which controls use of a fast
>> path descriptor push. So this patch extends the ringparam APIs and data
>> structures to support set/get tx push by ethtool -G/g.
>>
>> Signed-off-by: Jie Wang <[email protected]>
>> Signed-off-by: Guangbin Huang <[email protected]>
>
>> +``ETHTOOL_A_RINGS_TX_PUSH`` flag is used to choose the ordinary path or the fast
>> +path to send packets. In ordinary path, driver fills BDs to DDR memory and
>> +notifies NIC hardware. In fast path, driver pushes BDs to the device memory
>> +directly and thus reducing the sending latencies. Setting tx push attribute "on"
>> +will enable tx push mode and send packets in fast path if packet size matches.
>> +For those not supported hardwares, this attributes is "off" by default settings.
>
> Since you need to respin to fix the kdoc warning - could you also add
> a mention that enabling this feature may increase CPU cost? Unless it's
> not the case for your implementation, I thought it usually is..
>
ok, i will add it in v2
>> RINGS_SET
>> =========
>> @@ -887,6 +894,7 @@ Request contents:
>> ``ETHTOOL_A_RINGS_TX`` u32 size of TX ring
>> ``ETHTOOL_A_RINGS_RX_BUF_LEN`` u32 size of buffers on the ring
>> ``ETHTOOL_A_RINGS_CQE_SIZE`` u32 Size of TX/RX CQE
>> + ``ETHTOOL_A_RINGS_TX_PUSH`` u8 flag of TX Push mode
>> ==================================== ====== ===========================
>>
>> Kernel checks that requested ring sizes do not exceed limits reported by
>> diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h
>> index 4af58459a1e7..ede4f9154cd2 100644
>> --- a/include/linux/ethtool.h
>> +++ b/include/linux/ethtool.h
>> @@ -71,11 +71,13 @@ enum {
>> * struct kernel_ethtool_ringparam - RX/TX ring configuration
>> * @rx_buf_len: Current length of buffers on the rx ring.
>> * @tcp_data_split: Scatter packet headers and data to separate buffers
>> + * @tx_push: The flag of tx push mode
>> * @cqe_size: Size of TX/RX completion queue event
>> */
>> struct kernel_ethtool_ringparam {
>> u32 rx_buf_len;
>> u8 tcp_data_split;
>> + u8 tx_push;
>> u32 cqe_size;
>> };
>>
>> @@ -87,6 +89,7 @@ struct kernel_ethtool_ringparam {
>> enum ethtool_supported_ring_param {
>> ETHTOOL_RING_USE_RX_BUF_LEN = BIT(0),
>> ETHTOOL_RING_USE_CQE_SIZE = BIT(1),
>> + ETHTOOL_RING_USE_TX_PUSH = BIT(2),
>
> include/linux/ethtool.h:94: warning: Enum value 'ETHTOOL_RING_USE_TX_PUSH' not described in enum 'ethtool_supported_ring_param'
thx, I will fix it in v2
>
>
> .
>