* [PATCH net-next 1/4] gve: Move ptp_schedule_worker to gve_init_clock
2025-11-14 21:11 [PATCH net-next 0/4] gve: Implement XDP HW RX Timestamping support for DQ joshwash
@ 2025-11-14 21:11 ` joshwash
2025-11-14 21:11 ` [PATCH net-next 2/4] gve: Wrap struct xdp_buff joshwash
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: joshwash @ 2025-11-14 21:11 UTC (permalink / raw)
To: netdev
Cc: Joshua Washington, Harshitha Ramamurthy, Andrew Lunn,
David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Richard Cochran, Alexei Starovoitov, Daniel Borkmann,
Jesper Dangaard Brouer, John Fastabend, Stanislav Fomichev,
Willem de Bruijn, Praveen Kaligineedi, Tim Hostetler, Kevin Yang,
linux-kernel, bpf
From: Tim Hostetler <thostet@google.com>
Previously, gve had been only initializing ptp aux work when
hardware timestamping was initialized through ndo_hwtsatmp_set. As this
patch series introduces XDP hardware timestamp metadata which will
require the ptp aux work, the work can't be gated on the
kernel_hwtstamp_config being set and must be initialized elsewhere.
For simplicity, ptp_schedule_worker is invoked right after the ptp_clock
is registered with the kernel (which happens during gve_probe or
following reset). The worker is scheduled in GVE_NIC_TS_SYNC_INTERVAL_MS
as the synchronous call to gve_clock_nic_ts_read makes the worker
redundant if scheduled immediately.
If gve cannot read the device clock immediately, it errors out of
gve_init_clock.
Signed-off-by: Tim Hostetler <thostet@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
---
drivers/net/ethernet/google/gve/gve_main.c | 4 ----
drivers/net/ethernet/google/gve/gve_ptp.c | 12 ++++++++++++
2 files changed, 12 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/google/gve/gve_main.c b/drivers/net/ethernet/google/gve/gve_main.c
index 6fb8fbb38a7d..2b41a42fb516 100644
--- a/drivers/net/ethernet/google/gve/gve_main.c
+++ b/drivers/net/ethernet/google/gve/gve_main.c
@@ -2188,10 +2188,6 @@ static int gve_set_ts_config(struct net_device *dev,
}
kernel_config->rx_filter = HWTSTAMP_FILTER_ALL;
- gve_clock_nic_ts_read(priv);
- ptp_schedule_worker(priv->ptp->clock, 0);
- } else {
- ptp_cancel_worker_sync(priv->ptp->clock);
}
priv->ts_config.rx_filter = kernel_config->rx_filter;
diff --git a/drivers/net/ethernet/google/gve/gve_ptp.c b/drivers/net/ethernet/google/gve/gve_ptp.c
index a384a9ed4914..073677d82ee8 100644
--- a/drivers/net/ethernet/google/gve/gve_ptp.c
+++ b/drivers/net/ethernet/google/gve/gve_ptp.c
@@ -133,9 +133,21 @@ int gve_init_clock(struct gve_priv *priv)
err = -ENOMEM;
goto release_ptp;
}
+ err = gve_clock_nic_ts_read(priv);
+ if (err) {
+ dev_err(&priv->pdev->dev, "failed to read NIC clock %d\n", err);
+ goto release_nic_ts_report;
+ }
+ ptp_schedule_worker(priv->ptp->clock,
+ msecs_to_jiffies(GVE_NIC_TS_SYNC_INTERVAL_MS));
return 0;
+release_nic_ts_report:
+ dma_free_coherent(&priv->pdev->dev,
+ sizeof(struct gve_nic_ts_report),
+ priv->nic_ts_report, priv->nic_ts_report_bus);
+ priv->nic_ts_report = NULL;
release_ptp:
gve_ptp_release(priv);
return err;
--
2.51.2.1041.gc1ab5b90ca-goog
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH net-next 2/4] gve: Wrap struct xdp_buff
2025-11-14 21:11 [PATCH net-next 0/4] gve: Implement XDP HW RX Timestamping support for DQ joshwash
2025-11-14 21:11 ` [PATCH net-next 1/4] gve: Move ptp_schedule_worker to gve_init_clock joshwash
@ 2025-11-14 21:11 ` joshwash
2025-11-14 21:11 ` [PATCH net-next 3/4] gve: Prepare bpf_xdp_metadata_rx_timestamp support joshwash
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: joshwash @ 2025-11-14 21:11 UTC (permalink / raw)
To: netdev
Cc: Joshua Washington, Harshitha Ramamurthy, Andrew Lunn,
David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Richard Cochran, Alexei Starovoitov, Daniel Borkmann,
Jesper Dangaard Brouer, John Fastabend, Stanislav Fomichev,
Willem de Bruijn, Praveen Kaligineedi, Tim Hostetler, Kevin Yang,
linux-kernel, bpf
From: Tim Hostetler <thostet@google.com>
RX timestamping will need to keep track of extra temporary information
per-packet. In preparation for this, introduce gve_xdp_buff to wrap the
xdp_buff. This is similar in function to stmmac_xdp_buff and
ice_xdp_buff.
Signed-off-by: Tim Hostetler <thostet@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
---
drivers/net/ethernet/google/gve/gve.h | 5 +++++
drivers/net/ethernet/google/gve/gve_rx_dqo.c | 16 ++++++++--------
2 files changed, 13 insertions(+), 8 deletions(-)
diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h
index a33b44c1eb86..a21e599cf710 100644
--- a/drivers/net/ethernet/google/gve/gve.h
+++ b/drivers/net/ethernet/google/gve/gve.h
@@ -205,6 +205,11 @@ struct gve_rx_buf_state_dqo {
s16 next;
};
+/* Wrapper for XDP Rx metadata */
+struct gve_xdp_buff {
+ struct xdp_buff xdp;
+};
+
/* `head` and `tail` are indices into an array, or -1 if empty. */
struct gve_index_list {
s16 head;
diff --git a/drivers/net/ethernet/google/gve/gve_rx_dqo.c b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
index 1aff3bbb8cfc..76b26896f572 100644
--- a/drivers/net/ethernet/google/gve/gve_rx_dqo.c
+++ b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
@@ -840,23 +840,23 @@ static int gve_rx_dqo(struct napi_struct *napi, struct gve_rx_ring *rx,
}
if (xprog) {
- struct xdp_buff xdp;
+ struct gve_xdp_buff gve_xdp;
void *old_data;
int xdp_act;
- xdp_init_buff(&xdp, buf_state->page_info.buf_size,
+ xdp_init_buff(&gve_xdp.xdp, buf_state->page_info.buf_size,
&rx->xdp_rxq);
- xdp_prepare_buff(&xdp,
+ xdp_prepare_buff(&gve_xdp.xdp,
buf_state->page_info.page_address +
buf_state->page_info.page_offset,
buf_state->page_info.pad,
buf_len, false);
- old_data = xdp.data;
- xdp_act = bpf_prog_run_xdp(xprog, &xdp);
- buf_state->page_info.pad += xdp.data - old_data;
- buf_len = xdp.data_end - xdp.data;
+ old_data = gve_xdp.xdp.data;
+ xdp_act = bpf_prog_run_xdp(xprog, &gve_xdp.xdp);
+ buf_state->page_info.pad += gve_xdp.xdp.data - old_data;
+ buf_len = gve_xdp.xdp.data_end - gve_xdp.xdp.data;
if (xdp_act != XDP_PASS) {
- gve_xdp_done_dqo(priv, rx, &xdp, xprog, xdp_act,
+ gve_xdp_done_dqo(priv, rx, &gve_xdp.xdp, xprog, xdp_act,
buf_state);
return 0;
}
--
2.51.2.1041.gc1ab5b90ca-goog
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH net-next 3/4] gve: Prepare bpf_xdp_metadata_rx_timestamp support
2025-11-14 21:11 [PATCH net-next 0/4] gve: Implement XDP HW RX Timestamping support for DQ joshwash
2025-11-14 21:11 ` [PATCH net-next 1/4] gve: Move ptp_schedule_worker to gve_init_clock joshwash
2025-11-14 21:11 ` [PATCH net-next 2/4] gve: Wrap struct xdp_buff joshwash
@ 2025-11-14 21:11 ` joshwash
2025-11-14 21:11 ` [PATCH net-next 4/4] gve: Add Rx HWTS metadata to AF_XDP ZC mode joshwash
2025-11-18 15:00 ` [PATCH net-next 0/4] gve: Implement XDP HW RX Timestamping support for DQ patchwork-bot+netdevbpf
4 siblings, 0 replies; 6+ messages in thread
From: joshwash @ 2025-11-14 21:11 UTC (permalink / raw)
To: netdev
Cc: Joshua Washington, Harshitha Ramamurthy, Andrew Lunn,
David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Richard Cochran, Alexei Starovoitov, Daniel Borkmann,
Jesper Dangaard Brouer, John Fastabend, Stanislav Fomichev,
Willem de Bruijn, Praveen Kaligineedi, Tim Hostetler, Kevin Yang,
linux-kernel, bpf
From: Tim Hostetler <thostet@google.com>
Support populating XDP RX metadata with hardware RX timestamps. This
patch utilizes the same underlying logic to calculate hardware
timestamps as the regular RX path.
xdp_metadata_ops is registered with the net_device in a future patch.
gve_rx_calculate_hwtstamp was pulled out so as to not duplicate logic
between gve_xdp_rx_timestamp and gve_rx_hwtstamp.
Signed-off-by: Tim Hostetler <thostet@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
---
drivers/net/ethernet/google/gve/gve.h | 2 +
drivers/net/ethernet/google/gve/gve_dqo.h | 1 +
drivers/net/ethernet/google/gve/gve_rx_dqo.c | 41 +++++++++++++++-----
3 files changed, 34 insertions(+), 10 deletions(-)
diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h
index a21e599cf710..970d5ca8cdde 100644
--- a/drivers/net/ethernet/google/gve/gve.h
+++ b/drivers/net/ethernet/google/gve/gve.h
@@ -208,6 +208,8 @@ struct gve_rx_buf_state_dqo {
/* Wrapper for XDP Rx metadata */
struct gve_xdp_buff {
struct xdp_buff xdp;
+ struct gve_priv *gve;
+ const struct gve_rx_compl_desc_dqo *compl_desc;
};
/* `head` and `tail` are indices into an array, or -1 if empty. */
diff --git a/drivers/net/ethernet/google/gve/gve_dqo.h b/drivers/net/ethernet/google/gve/gve_dqo.h
index 6eb442096e02..5871f773f0c7 100644
--- a/drivers/net/ethernet/google/gve/gve_dqo.h
+++ b/drivers/net/ethernet/google/gve/gve_dqo.h
@@ -36,6 +36,7 @@ netdev_tx_t gve_tx_dqo(struct sk_buff *skb, struct net_device *dev);
netdev_features_t gve_features_check_dqo(struct sk_buff *skb,
struct net_device *dev,
netdev_features_t features);
+int gve_xdp_rx_timestamp(const struct xdp_md *_ctx, u64 *timestamp);
bool gve_tx_poll_dqo(struct gve_notify_block *block, bool do_clean);
bool gve_xdp_poll_dqo(struct gve_notify_block *block);
bool gve_xsk_tx_poll_dqo(struct gve_notify_block *block, int budget);
diff --git a/drivers/net/ethernet/google/gve/gve_rx_dqo.c b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
index 76b26896f572..f20d1b1d06e6 100644
--- a/drivers/net/ethernet/google/gve/gve_rx_dqo.c
+++ b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
@@ -456,20 +456,38 @@ static void gve_rx_skb_hash(struct sk_buff *skb,
* Note that this means if the time delta between packet reception and the last
* clock read is greater than ~2 seconds, this will provide invalid results.
*/
+static ktime_t gve_rx_get_hwtstamp(struct gve_priv *gve, u32 hwts)
+{
+ u64 last_read = READ_ONCE(gve->last_sync_nic_counter);
+ u32 low = (u32)last_read;
+ s32 diff = hwts - low;
+
+ return ns_to_ktime(last_read + diff);
+}
+
static void gve_rx_skb_hwtstamp(struct gve_rx_ring *rx,
const struct gve_rx_compl_desc_dqo *desc)
{
- u64 last_read = READ_ONCE(rx->gve->last_sync_nic_counter);
struct sk_buff *skb = rx->ctx.skb_head;
- u32 ts, low;
- s32 diff;
-
- if (desc->ts_sub_nsecs_low & GVE_DQO_RX_HWTSTAMP_VALID) {
- ts = le32_to_cpu(desc->ts);
- low = (u32)last_read;
- diff = ts - low;
- skb_hwtstamps(skb)->hwtstamp = ns_to_ktime(last_read + diff);
- }
+
+ if (desc->ts_sub_nsecs_low & GVE_DQO_RX_HWTSTAMP_VALID)
+ skb_hwtstamps(skb)->hwtstamp =
+ gve_rx_get_hwtstamp(rx->gve, le32_to_cpu(desc->ts));
+}
+
+int gve_xdp_rx_timestamp(const struct xdp_md *_ctx, u64 *timestamp)
+{
+ const struct gve_xdp_buff *ctx = (void *)_ctx;
+
+ if (!ctx->gve->nic_ts_report)
+ return -ENODATA;
+
+ if (!(ctx->compl_desc->ts_sub_nsecs_low & GVE_DQO_RX_HWTSTAMP_VALID))
+ return -ENODATA;
+
+ *timestamp = gve_rx_get_hwtstamp(ctx->gve,
+ le32_to_cpu(ctx->compl_desc->ts));
+ return 0;
}
static void gve_rx_free_skb(struct napi_struct *napi, struct gve_rx_ring *rx)
@@ -851,6 +869,9 @@ static int gve_rx_dqo(struct napi_struct *napi, struct gve_rx_ring *rx,
buf_state->page_info.page_offset,
buf_state->page_info.pad,
buf_len, false);
+ gve_xdp.gve = priv;
+ gve_xdp.compl_desc = compl_desc;
+
old_data = gve_xdp.xdp.data;
xdp_act = bpf_prog_run_xdp(xprog, &gve_xdp.xdp);
buf_state->page_info.pad += gve_xdp.xdp.data - old_data;
--
2.51.2.1041.gc1ab5b90ca-goog
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH net-next 4/4] gve: Add Rx HWTS metadata to AF_XDP ZC mode
2025-11-14 21:11 [PATCH net-next 0/4] gve: Implement XDP HW RX Timestamping support for DQ joshwash
` (2 preceding siblings ...)
2025-11-14 21:11 ` [PATCH net-next 3/4] gve: Prepare bpf_xdp_metadata_rx_timestamp support joshwash
@ 2025-11-14 21:11 ` joshwash
2025-11-18 15:00 ` [PATCH net-next 0/4] gve: Implement XDP HW RX Timestamping support for DQ patchwork-bot+netdevbpf
4 siblings, 0 replies; 6+ messages in thread
From: joshwash @ 2025-11-14 21:11 UTC (permalink / raw)
To: netdev
Cc: Joshua Washington, Harshitha Ramamurthy, Andrew Lunn,
David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Richard Cochran, Alexei Starovoitov, Daniel Borkmann,
Jesper Dangaard Brouer, John Fastabend, Stanislav Fomichev,
Willem de Bruijn, Praveen Kaligineedi, Tim Hostetler, Kevin Yang,
linux-kernel, bpf
From: Tim Hostetler <thostet@google.com>
By overlaying the struct gve_xdp_buff on top of the struct xdp_buff_xsk
that AF_XDP utilizes, the driver records the 32 bit timestamp via the
completion descriptor and the cached 64 bit NIC timestamp via gve_priv.
The driver's implementation of xmo_rx_timestamp extends the timestamp to
the full and up to date 64 bit timestamp and returns it to the user.
gve_rx_xsk_dqo is modified to accept a pointer to the completion
descriptor and no longer takes a buf_len explicitly as it can be pulled
out of the descriptor.
With this patch gve now supports bpf_xdp_metadata_rx_timestamp.
Signed-off-by: Tim Hostetler <thostet@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
---
drivers/net/ethernet/google/gve/gve_main.c | 7 +++++++
drivers/net/ethernet/google/gve/gve_rx_dqo.c | 16 ++++++++++++++--
2 files changed, 21 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/google/gve/gve_main.c b/drivers/net/ethernet/google/gve/gve_main.c
index 2b41a42fb516..a5a2b18d309b 100644
--- a/drivers/net/ethernet/google/gve/gve_main.c
+++ b/drivers/net/ethernet/google/gve/gve_main.c
@@ -2348,6 +2348,10 @@ static void gve_set_netdev_xdp_features(struct gve_priv *priv)
xdp_set_features_flag_locked(priv->dev, xdp_features);
}
+static const struct xdp_metadata_ops gve_xdp_metadata_ops = {
+ .xmo_rx_timestamp = gve_xdp_rx_timestamp,
+};
+
static int gve_init_priv(struct gve_priv *priv, bool skip_describe_device)
{
int num_ntfy;
@@ -2443,6 +2447,9 @@ static int gve_init_priv(struct gve_priv *priv, bool skip_describe_device)
}
gve_set_netdev_xdp_features(priv);
+ if (!gve_is_gqi(priv))
+ priv->dev->xdp_metadata_ops = &gve_xdp_metadata_ops;
+
err = gve_setup_device_resources(priv);
if (err)
goto err_free_xsk_bitmap;
diff --git a/drivers/net/ethernet/google/gve/gve_rx_dqo.c b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
index f20d1b1d06e6..f1bd8f5d5732 100644
--- a/drivers/net/ethernet/google/gve/gve_rx_dqo.c
+++ b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
@@ -240,6 +240,11 @@ int gve_rx_alloc_ring_dqo(struct gve_priv *priv,
rx->rx_headroom = 0;
}
+ /* struct gve_xdp_buff is overlaid on struct xdp_buff_xsk and utilizes
+ * the 24 byte field cb to store gve specific data.
+ */
+ XSK_CHECK_PRIV_TYPE(struct gve_xdp_buff);
+
rx->dqo.num_buf_states = cfg->raw_addressing ? buffer_queue_slots :
gve_get_rx_pages_per_qpl_dqo(cfg->ring_size);
rx->dqo.buf_states = kvcalloc_node(rx->dqo.num_buf_states,
@@ -701,16 +706,23 @@ static void gve_xdp_done_dqo(struct gve_priv *priv, struct gve_rx_ring *rx,
}
static int gve_rx_xsk_dqo(struct napi_struct *napi, struct gve_rx_ring *rx,
- struct gve_rx_buf_state_dqo *buf_state, int buf_len,
+ const struct gve_rx_compl_desc_dqo *compl_desc,
+ struct gve_rx_buf_state_dqo *buf_state,
struct bpf_prog *xprog)
{
struct xdp_buff *xdp = buf_state->xsk_buff;
+ int buf_len = compl_desc->packet_len;
struct gve_priv *priv = rx->gve;
+ struct gve_xdp_buff *gve_xdp;
int xdp_act;
xdp->data_end = xdp->data + buf_len;
xsk_buff_dma_sync_for_cpu(xdp);
+ gve_xdp = (void *)xdp;
+ gve_xdp->gve = priv;
+ gve_xdp->compl_desc = compl_desc;
+
if (xprog) {
xdp_act = bpf_prog_run_xdp(xprog, xdp);
buf_len = xdp->data_end - xdp->data;
@@ -800,7 +812,7 @@ static int gve_rx_dqo(struct napi_struct *napi, struct gve_rx_ring *rx,
xprog = READ_ONCE(priv->xdp_prog);
if (buf_state->xsk_buff)
- return gve_rx_xsk_dqo(napi, rx, buf_state, buf_len, xprog);
+ return gve_rx_xsk_dqo(napi, rx, compl_desc, buf_state, xprog);
/* Page might have not been used for awhile and was likely last written
* by a different thread.
--
2.51.2.1041.gc1ab5b90ca-goog
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [PATCH net-next 0/4] gve: Implement XDP HW RX Timestamping support for DQ
2025-11-14 21:11 [PATCH net-next 0/4] gve: Implement XDP HW RX Timestamping support for DQ joshwash
` (3 preceding siblings ...)
2025-11-14 21:11 ` [PATCH net-next 4/4] gve: Add Rx HWTS metadata to AF_XDP ZC mode joshwash
@ 2025-11-18 15:00 ` patchwork-bot+netdevbpf
4 siblings, 0 replies; 6+ messages in thread
From: patchwork-bot+netdevbpf @ 2025-11-18 15:00 UTC (permalink / raw)
To: Joshua Washington
Cc: netdev, hramamurthy, andrew+netdev, davem, edumazet, kuba, pabeni,
richardcochran, ast, daniel, hawk, john.fastabend, sdf, willemb,
pkaligineedi, thostet, yyd, linux-kernel, bpf
Hello:
This series was applied to netdev/net-next.git (main)
by Paolo Abeni <pabeni@redhat.com>:
On Fri, 14 Nov 2025 13:11:42 -0800 you wrote:
> From: Tim Hostetler <thostet@google.com>
>
> This patch series adds support for bpf_xdp_metadata_rx_timestamp from an
> XDP program loaded into the driver on its own or bound to an XSK. This
> is only supported for DQ.
>
> Tim Hostetler (4):
> gve: Move ptp_schedule_worker to gve_init_clock
> gve: Wrap struct xdp_buff
> gve: Prepare bpf_xdp_metadata_rx_timestamp support
> gve: Add Rx HWTS metadata to AF_XDP ZC mode
Here is the summary with links:
- [net-next,1/4] gve: Move ptp_schedule_worker to gve_init_clock
https://git.kernel.org/netdev/net-next/c/46e7860ef941
- [net-next,2/4] gve: Wrap struct xdp_buff
https://git.kernel.org/netdev/net-next/c/f356a66b87bb
- [net-next,3/4] gve: Prepare bpf_xdp_metadata_rx_timestamp support
https://git.kernel.org/netdev/net-next/c/66adaf102128
- [net-next,4/4] gve: Add Rx HWTS metadata to AF_XDP ZC mode
https://git.kernel.org/netdev/net-next/c/1b42e07af1ee
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html
^ permalink raw reply [flat|nested] 6+ messages in thread