From: Mohsin Bashir <mohsin.bashr@gmail.com>
To: Willem de Bruijn <willemdebruijn.kernel@gmail.com>,
netdev@vger.kernel.org
Cc: davem@davemloft.net, kuba@kernel.org, edumazet@google.com,
pabeni@redhat.com, horms@kernel.org, andrew@lunn.ch,
Willem de Bruijn <willemb@google.com>
Subject: Re: [PATCH net-next v2 1/7] net: ethtool: add hardware pacing offload support to rings
Date: Thu, 23 Jul 2026 07:16:38 -0700 [thread overview]
Message-ID: <9c0a0bb5-737b-41b7-828c-1953d885fb1c@gmail.com> (raw)
In-Reply-To: <20260722204454.3234605-2-willemdebruijn.kernel@gmail.com>
On 7/22/26 1:43 PM, Willem de Bruijn wrote:
> From: Willem de Bruijn <willemb@google.com>
>
> Add two ethtool rings operations:
>
> - get pacing offload horizon: active and max
> - set pacing offload horizon: active
>
> Replace u64 max_pacing_offload_horizon with u32 active and max fields.
> This occupies the same original 8 bytes. Reduce precision from nsec to
> usec, which is sufficient.
>
> Update the FQ scheduler to test against the active limit instead of
> max. It is the administrator responsibility to set the active limit
> before installing FQ.
>
> Assisted-by: Gemini:gemini-3
> Signed-off-by: Willem de Bruijn <willemb@google.com>
>
If I understand it correctly, the horizon lives in dev and is read at Tx
time, so it shouldn't require touching ring config. But some drivers
reallocate rings unconditionally in set_ringparam. Can we highlight
somewhere (maybe in commit message) that a horizon-only change is meant
to be non-disruptive? I may be missing the underlying reason though.
> ---
>
> Changes
> v1 -> v2
> - (minor) style: remove period from end of ethtool comments
> - (minor) style: fix missing space at end of HDS_THRESH comment
> v1: https://lore.kernel.org/netdev/20260706133433.3142805-2-willemdebruijn.kernel@gmail.com/
> ---
> Documentation/netlink/specs/ethtool.yaml | 6 ++++
> .../networking/net_cachelines/net_device.rst | 3 +-
> include/linux/ethtool.h | 6 ++++
> include/linux/netdevice.h | 6 ++--
> .../uapi/linux/ethtool_netlink_generated.h | 2 ++
> net/core/rtnetlink.c | 2 +-
> net/ethtool/common.c | 3 ++
> net/ethtool/netlink.h | 2 +-
> net/ethtool/rings.c | 30 +++++++++++++++++--
> net/sched/sch_fq.c | 3 +-
> 10 files changed, 55 insertions(+), 8 deletions(-)
>
> diff --git a/Documentation/netlink/specs/ethtool.yaml b/Documentation/netlink/specs/ethtool.yaml
> index 5dd4d1b5d94b..48f43dda2e6e 100644
> --- a/Documentation/netlink/specs/ethtool.yaml
> +++ b/Documentation/netlink/specs/ethtool.yaml
> @@ -448,6 +448,12 @@ attribute-sets:
> -
> name: hds-thresh-max
> type: u32
> + -
> + name: pacing-offload-horizon
> + type: u32
> + -
> + name: pacing-offload-horizon-max
> + type: u32
>
> -
> name: mm-stat
> diff --git a/Documentation/networking/net_cachelines/net_device.rst b/Documentation/networking/net_cachelines/net_device.rst
> index 512f6d6fa3d8..3b9ce1c6e105 100644
> --- a/Documentation/networking/net_cachelines/net_device.rst
> +++ b/Documentation/networking/net_cachelines/net_device.rst
> @@ -183,7 +183,8 @@ struct devlink_port* devlink_port
> struct dpll_pin* dpll_pin
> struct hlist_head page_pools
> struct dim_irq_moder* irq_moder
> -u64 max_pacing_offload_horizon
> +u32 pacing_offload_horizon
> +u32 max_pacing_offload_horizon
> struct_napi_config* napi_config
> unsigned_long gro_flush_timeout
> u32 napi_defer_hard_irqs
> diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h
> index 5d491a98265e..adeebbf53dad 100644
> --- a/include/linux/ethtool.h
> +++ b/include/linux/ethtool.h
> @@ -85,6 +85,8 @@ enum {
> * @tx_push_buf_max_len: Maximum allowed size of TX push buffer
> * @hds_thresh: Packet size threshold for header data split (HDS)
> * @hds_thresh_max: Maximum supported setting for @hds_threshold
> + * @pacing_offload_horizon: pacing offload horizon value in usec
> + * @max_pacing_offload_horizon: max pacing offload horizon value in usec
> *
> */
> struct kernel_ethtool_ringparam {
> @@ -97,6 +99,8 @@ struct kernel_ethtool_ringparam {
> u32 tx_push_buf_max_len;
> u32 hds_thresh;
> u32 hds_thresh_max;
> + u32 pacing_offload_horizon;
> + u32 max_pacing_offload_horizon;
> };
>
> /**
> @@ -108,6 +112,7 @@ struct kernel_ethtool_ringparam {
> * @ETHTOOL_RING_USE_TX_PUSH_BUF_LEN: capture for setting tx_push_buf_len
> * @ETHTOOL_RING_USE_TCP_DATA_SPLIT: capture for setting tcp_data_split
> * @ETHTOOL_RING_USE_HDS_THRS: capture for setting header-data-split-thresh
> + * @ETHTOOL_RING_USE_PACING_OFFLOAD_HORIZON: capture for setting pacing offload horizon
> */
> enum ethtool_supported_ring_param {
> ETHTOOL_RING_USE_RX_BUF_LEN = BIT(0),
> @@ -117,6 +122,7 @@ enum ethtool_supported_ring_param {
> ETHTOOL_RING_USE_TX_PUSH_BUF_LEN = BIT(4),
> ETHTOOL_RING_USE_TCP_DATA_SPLIT = BIT(5),
> ETHTOOL_RING_USE_HDS_THRS = BIT(6),
> + ETHTOOL_RING_USE_PACING_OFFLOAD_HORIZON = BIT(7),
> };
>
> #define __ETH_RSS_HASH_BIT(bit) ((u32)1 << (bit))
> diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
> index 8db25b79573e..3c2ed45b55dc 100644
> --- a/include/linux/netdevice.h
> +++ b/include/linux/netdevice.h
> @@ -2129,7 +2129,8 @@ enum netdev_reg_state {
> * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
> * where the clock is recovered.
> *
> - * @max_pacing_offload_horizon: max EDT offload horizon in nsec.
> + * @pacing_offload_horizon: active pacing offload horizon in usec.
> + * @max_pacing_offload_horizon: max pacing offload horizon in usec.
> * @napi_config: An array of napi_config structures containing per-NAPI
> * settings.
> * @num_napi_configs: number of allocated NAPI config structs,
> @@ -2552,7 +2553,8 @@ struct net_device {
> /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */
> struct dim_irq_moder *irq_moder;
>
> - u64 max_pacing_offload_horizon;
> + u32 pacing_offload_horizon;
> + u32 max_pacing_offload_horizon;
> struct napi_config *napi_config;
> u32 num_napi_configs;
> u32 napi_defer_hard_irqs;
> diff --git a/include/uapi/linux/ethtool_netlink_generated.h b/include/uapi/linux/ethtool_netlink_generated.h
> index 8134baf7860f..6f1a2fa36157 100644
> --- a/include/uapi/linux/ethtool_netlink_generated.h
> +++ b/include/uapi/linux/ethtool_netlink_generated.h
> @@ -193,6 +193,8 @@ enum {
> ETHTOOL_A_RINGS_TX_PUSH_BUF_LEN_MAX,
> ETHTOOL_A_RINGS_HDS_THRESH,
> ETHTOOL_A_RINGS_HDS_THRESH_MAX,
> + ETHTOOL_A_RINGS_PACING_OFFLOAD_HORIZON,
> + ETHTOOL_A_RINGS_PACING_OFFLOAD_HORIZON_MAX,
>
> __ETHTOOL_A_RINGS_CNT,
> ETHTOOL_A_RINGS_MAX = (__ETHTOOL_A_RINGS_CNT - 1)
> diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
> index 31c65a545a10..2c7a245c5ffc 100644
> --- a/net/core/rtnetlink.c
> +++ b/net/core/rtnetlink.c
> @@ -2155,7 +2155,7 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb,
> nla_put_u32(skb, IFLA_TSO_MAX_SEGS,
> READ_ONCE(dev->tso_max_segs)) ||
> nla_put_uint(skb, IFLA_MAX_PACING_OFFLOAD_HORIZON,
> - READ_ONCE(dev->max_pacing_offload_horizon)) ||
> + (u64)READ_ONCE(dev->max_pacing_offload_horizon) * NSEC_PER_USEC) ||
> #ifdef CONFIG_RPS
> nla_put_u32(skb, IFLA_NUM_RX_QUEUES,
> READ_ONCE(dev->num_rx_queues)) ||
> diff --git a/net/ethtool/common.c b/net/ethtool/common.c
> index 23db40618fed..5c2323aaa02b 100644
> --- a/net/ethtool/common.c
> +++ b/net/ethtool/common.c
> @@ -954,6 +954,9 @@ void ethtool_ringparam_get_cfg(struct net_device *dev,
> /* Driver gives us current state, we want to return current config */
> kparam->tcp_data_split = dev->cfg->hds_config;
> kparam->hds_thresh = dev->cfg->hds_thresh;
> +
> + kparam->pacing_offload_horizon = dev->pacing_offload_horizon;
> + kparam->max_pacing_offload_horizon = dev->max_pacing_offload_horizon;
> }
>
> static void ethtool_init_tsinfo(struct kernel_ethtool_ts_info *info)
> diff --git a/net/ethtool/netlink.h b/net/ethtool/netlink.h
> index 3e969a070f9f..ae7b651c658f 100644
> --- a/net/ethtool/netlink.h
> +++ b/net/ethtool/netlink.h
> @@ -494,7 +494,7 @@ 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_HDS_THRESH_MAX + 1];
> +extern const struct nla_policy ethnl_rings_set_policy[ETHTOOL_A_RINGS_PACING_OFFLOAD_HORIZON_MAX + 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 9054c89c5d7b..52158572fe10 100644
> --- a/net/ethtool/rings.c
> +++ b/net/ethtool/rings.c
> @@ -42,6 +42,8 @@ static int rings_prepare_data(const struct ethnl_req_info *req_base,
>
> data->kernel_ringparam.tcp_data_split = dev->cfg->hds_config;
> data->kernel_ringparam.hds_thresh = dev->cfg->hds_thresh;
> + data->kernel_ringparam.pacing_offload_horizon = dev->pacing_offload_horizon;
> + data->kernel_ringparam.max_pacing_offload_horizon = dev->max_pacing_offload_horizon;
>
> dev->ethtool_ops->get_ringparam(dev, &data->ringparam,
> &data->kernel_ringparam, info->extack);
> @@ -69,7 +71,9 @@ static int rings_reply_size(const struct ethnl_req_info *req_base,
> nla_total_size(sizeof(u32)) + /* _RINGS_TX_PUSH_BUF_LEN */
> nla_total_size(sizeof(u32)) + /* _RINGS_TX_PUSH_BUF_LEN_MAX */
> nla_total_size(sizeof(u32)) + /* _RINGS_HDS_THRESH */
> - nla_total_size(sizeof(u32)); /* _RINGS_HDS_THRESH_MAX*/
> + nla_total_size(sizeof(u32)) + /* _RINGS_HDS_THRESH_MAX */
> + nla_total_size(sizeof(u32)) + /* _RINGS_PACING_OFFLOAD_HORIZON */
> + nla_total_size(sizeof(u32)); /* _RINGS_PACING_OFFLOAD_HORIZON_MAX */
> }
>
> static int rings_fill_reply(struct sk_buff *skb,
> @@ -121,7 +125,12 @@ static int rings_fill_reply(struct sk_buff *skb,
> (nla_put_u32(skb, ETHTOOL_A_RINGS_HDS_THRESH,
> kr->hds_thresh) ||
> nla_put_u32(skb, ETHTOOL_A_RINGS_HDS_THRESH_MAX,
> - kr->hds_thresh_max))))
> + kr->hds_thresh_max))) ||
> + ((supported_ring_params & ETHTOOL_RING_USE_PACING_OFFLOAD_HORIZON) &&
> + (nla_put_u32(skb, ETHTOOL_A_RINGS_PACING_OFFLOAD_HORIZON,
> + kr->pacing_offload_horizon) ||
> + nla_put_u32(skb, ETHTOOL_A_RINGS_PACING_OFFLOAD_HORIZON_MAX,
> + kr->max_pacing_offload_horizon))))
> return -EMSGSIZE;
>
> return 0;
> @@ -144,6 +153,7 @@ const struct nla_policy ethnl_rings_set_policy[] = {
> [ETHTOOL_A_RINGS_RX_PUSH] = NLA_POLICY_MAX(NLA_U8, 1),
> [ETHTOOL_A_RINGS_TX_PUSH_BUF_LEN] = { .type = NLA_U32 },
> [ETHTOOL_A_RINGS_HDS_THRESH] = { .type = NLA_U32 },
> + [ETHTOOL_A_RINGS_PACING_OFFLOAD_HORIZON] = { .type = NLA_U32 },
> };
>
> static int
> @@ -177,6 +187,14 @@ ethnl_set_rings_validate(struct ethnl_req_info *req_info,
> return -EOPNOTSUPP;
> }
>
> + if (tb[ETHTOOL_A_RINGS_PACING_OFFLOAD_HORIZON] &&
> + !(ops->supported_ring_params & ETHTOOL_RING_USE_PACING_OFFLOAD_HORIZON)) {
> + NL_SET_ERR_MSG_ATTR(info->extack,
> + tb[ETHTOOL_A_RINGS_PACING_OFFLOAD_HORIZON],
> + "setting pacing offload horizon is not supported");
> + return -EOPNOTSUPP;
> + }
> +
> if (tb[ETHTOOL_A_RINGS_CQE_SIZE] &&
> !(ops->supported_ring_params & ETHTOOL_RING_USE_CQE_SIZE)) {
> NL_SET_ERR_MSG_ATTR(info->extack,
> @@ -246,6 +264,8 @@ ethnl_set_rings(struct ethnl_req_info *req_info, struct genl_info *info)
> tb[ETHTOOL_A_RINGS_TX_PUSH_BUF_LEN], &mod);
> ethnl_update_u32(&kernel_ringparam.hds_thresh,
> tb[ETHTOOL_A_RINGS_HDS_THRESH], &mod);
> + ethnl_update_u32(&kernel_ringparam.pacing_offload_horizon,
> + tb[ETHTOOL_A_RINGS_PACING_OFFLOAD_HORIZON], &mod);
> if (!mod)
> return 0;
>
> @@ -281,6 +301,9 @@ ethnl_set_rings(struct ethnl_req_info *req_info, struct genl_info *info)
> err_attr = tb[ETHTOOL_A_RINGS_TX];
> else if (kernel_ringparam.hds_thresh > kernel_ringparam.hds_thresh_max)
> err_attr = tb[ETHTOOL_A_RINGS_HDS_THRESH];
> + else if (kernel_ringparam.pacing_offload_horizon >
> + kernel_ringparam.max_pacing_offload_horizon)
> + err_attr = tb[ETHTOOL_A_RINGS_PACING_OFFLOAD_HORIZON];
> else
> err_attr = NULL;
> if (err_attr) {
> @@ -302,6 +325,9 @@ ethnl_set_rings(struct ethnl_req_info *req_info, struct genl_info *info)
>
> ret = dev->ethtool_ops->set_ringparam(dev, &ringparam,
> &kernel_ringparam, info->extack);
> + if (!ret)
> + dev->pacing_offload_horizon = kernel_ringparam.pacing_offload_horizon;
Since a driver may be reading dev->pacing_offload_horizon on the hotpath
without a lock, should we use WRITE_ONCE() here? and have the driver use
READ_ONCE()?
> +
> return ret < 0 ? ret : 1;
> }
>
> diff --git a/net/sched/sch_fq.c b/net/sched/sch_fq.c
> index 7cae082a9847..d4cc8bc9fa06 100644
> --- a/net/sched/sch_fq.c
> +++ b/net/sched/sch_fq.c
> @@ -1179,7 +1179,8 @@ static int fq_change(struct Qdisc *sch, struct nlattr *opt,
> u64 offload_horizon = (u64)NSEC_PER_USEC *
> nla_get_u32(tb[TCA_FQ_OFFLOAD_HORIZON]);
>
> - if (offload_horizon <= qdisc_dev(sch)->max_pacing_offload_horizon) {
> + if (offload_horizon <=
> + (u64)qdisc_dev(sch)->pacing_offload_horizon * NSEC_PER_USEC) {
> WRITE_ONCE(q->offload_horizon, offload_horizon);
> } else {
> NL_SET_ERR_MSG_MOD(extack, "invalid offload_horizon");
next prev parent reply other threads:[~2026-07-23 14:16 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-07-22 20:43 [PATCH net-next v2 0/7] hardware pacing offload Willem de Bruijn
2026-07-22 20:43 ` [PATCH net-next v2 1/7] net: ethtool: add hardware pacing offload support to rings Willem de Bruijn
2026-07-23 14:16 ` Mohsin Bashir [this message]
2026-07-23 21:29 ` Willem de Bruijn
2026-07-23 22:11 ` Mohsin Bashir
2026-07-22 20:43 ` [PATCH net-next v2 2/7] net_sched: sch_fq: clear past skb->tstamp if offloading pacing Willem de Bruijn
2026-07-22 20:43 ` [PATCH net-next v2 3/7] idpf: support pacing offload Willem de Bruijn
2026-07-23 14:32 ` Mohsin Bashir
2026-07-23 21:27 ` Willem de Bruijn
2026-07-22 20:43 ` [PATCH net-next v2 4/7] selftests: drv-net: refactor so_txtime errqueue handling Willem de Bruijn
2026-07-23 14:40 ` Mohsin Bashir
2026-07-22 20:43 ` [PATCH net-next v2 5/7] selftests: drv-net: in so_txtime tell apart sw from hw pacing Willem de Bruijn
2026-07-22 20:43 ` [PATCH net-next v2 6/7] selftests: drv-net: extend so_txtime with hw offload Willem de Bruijn
2026-07-22 20:43 ` [PATCH net-next v2 7/7] net: pktgen: add support for SO_TXTIME Willem de Bruijn
2026-07-23 14:12 ` [PATCH net-next v2 0/7] hardware pacing offload Mohsin Bashir
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=9c0a0bb5-737b-41b7-828c-1953d885fb1c@gmail.com \
--to=mohsin.bashr@gmail.com \
--cc=andrew@lunn.ch \
--cc=davem@davemloft.net \
--cc=edumazet@google.com \
--cc=horms@kernel.org \
--cc=kuba@kernel.org \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=willemb@google.com \
--cc=willemdebruijn.kernel@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox